Merge tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm
authorLinus Torvalds <torvalds@linux-foundation.org>
Sat, 27 Aug 2016 22:51:50 +0000 (15:51 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Sat, 27 Aug 2016 22:51:50 +0000 (15:51 -0700)
Pull KVM fixes from Paolo Bonzini:
 "ARM:
   - fixes for ITS init issues, error handling, IRQ leakage, race
     conditions
   - an erratum workaround for timers
   - some removal of misleading use of errors and comments
   - a fix for GICv3 on 32-bit guests

  MIPS:
   - fix for where the guest could wrongly map the first page of
     physical memory

  x86:
   - nested virtualization fixes"

* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
  MIPS: KVM: Check for pfn noslot case
  kvm: nVMX: fix nested tsc scaling
  KVM: nVMX: postpone VMCS changes on MSR_IA32_APICBASE write
  KVM: nVMX: fix msr bitmaps to prevent L2 from accessing L0 x2APIC
  arm64: KVM: report configured SRE value to 32-bit world
  arm64: KVM: remove misleading comment on pmu status
  KVM: arm/arm64: timer: Workaround misconfigured timer interrupt
  arm64: Document workaround for Cortex-A72 erratum #853709
  KVM: arm/arm64: Change misleading use of is_error_pfn
  KVM: arm64: ITS: avoid re-mapping LPIs
  KVM: arm64: check for ITS device on MSI injection
  KVM: arm64: ITS: move ITS registration into first VCPU run
  KVM: arm64: vgic-its: Make updates to propbaser/pendbaser atomic
  KVM: arm64: vgic-its: Plug race in vgic_put_irq
  KVM: arm64: vgic-its: Handle errors from vgic_add_lpi
  KVM: arm64: ITS: return 1 on successful MSI injection

280 files changed:
Documentation/PCI/MSI-HOWTO.txt
Documentation/conf.py
Documentation/devicetree/bindings/sound/omap-mcpdm.txt
Documentation/devicetree/bindings/thermal/thermal.txt
Documentation/hwmon/ftsteutates
Documentation/kernel-documentation.rst
Documentation/kernel-parameters.txt
Documentation/power/basic-pm-debugging.txt
Documentation/power/interface.txt
Documentation/sphinx-static/theme_overrides.css
MAINTAINERS
Makefile
arch/arc/include/asm/entry.h
arch/arc/include/asm/irqflags-compact.h
arch/arc/include/asm/pgtable.h
arch/arc/include/uapi/asm/elf.h
arch/arc/kernel/arcksyms.c
arch/arc/kernel/process.c
arch/arc/kernel/setup.c
arch/arc/mm/cache.c
arch/arc/mm/highmem.c
arch/arm/kernel/entry-armv.S
arch/arm/mach-imx/gpc.c
arch/arm/mm/mmu.c
arch/arm/xen/enlighten.c
arch/arm64/kernel/head.S
arch/arm64/kernel/sleep.S
arch/arm64/mm/dump.c
arch/arm64/mm/numa.c
arch/mips/include/asm/page.h
arch/parisc/include/uapi/asm/errno.h
arch/parisc/kernel/processor.c
arch/parisc/kernel/time.c
arch/s390/kernel/setup.c
arch/um/include/asm/common.lds.S
arch/x86/crypto/sha256-mb/sha256_mb.c
arch/x86/crypto/sha256-mb/sha256_mb_mgr_flush_avx2.S
arch/x86/crypto/sha512-mb/sha512_mb.c
arch/x86/kernel/cpu/microcode/amd.c
arch/x86/kernel/smpboot.c
arch/x86/mm/kaslr.c
arch/x86/pci/vmd.c
arch/x86/power/hibernate_64.c
arch/x86/xen/enlighten.c
block/bio.c
block/blk-core.c
block/blk-merge.c
block/blk-mq.c
block/elevator.c
drivers/block/floppy.c
drivers/block/xen-blkfront.c
drivers/clocksource/bcm_kona_timer.c
drivers/clocksource/mips-gic-timer.c
drivers/clocksource/time-armada-370-xp.c
drivers/dax/pmem.c
drivers/edac/Kconfig
drivers/edac/Makefile
drivers/edac/skx_edac.c [new file with mode: 0644]
drivers/gpio/Kconfig
drivers/gpio/gpio-max730x.c
drivers/gpu/drm/amd/amdgpu/amdgpu.h
drivers/gpu/drm/amd/amdgpu/amdgpu_atpx_handler.c
drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c
drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c
drivers/gpu/drm/drm_fb_helper.c
drivers/gpu/drm/etnaviv/etnaviv_gpu.c
drivers/gpu/drm/i915/i915_drv.h
drivers/gpu/drm/i915/i915_gem.c
drivers/gpu/drm/i915/i915_gem_gtt.c
drivers/gpu/drm/i915/i915_reg.h
drivers/gpu/drm/i915/intel_audio.c
drivers/gpu/drm/i915/intel_ddi.c
drivers/gpu/drm/i915/intel_display.c
drivers/gpu/drm/i915/intel_fbc.c
drivers/gpu/drm/i915/intel_pm.c
drivers/gpu/drm/i915/intel_ringbuffer.c
drivers/gpu/drm/mediatek/Kconfig
drivers/gpu/drm/radeon/radeon_atpx_handler.c
drivers/hwmon/it87.c
drivers/i2c/busses/i2c-at91.c
drivers/i2c/busses/i2c-bcm-iproc.c
drivers/i2c/busses/i2c-bcm-kona.c
drivers/i2c/busses/i2c-brcmstb.c
drivers/i2c/busses/i2c-cros-ec-tunnel.c
drivers/i2c/busses/i2c-meson.c
drivers/i2c/busses/i2c-ocores.c
drivers/i2c/muxes/i2c-demux-pinctrl.c
drivers/infiniband/core/cma.c
drivers/infiniband/hw/cxgb4/cm.c
drivers/infiniband/hw/cxgb4/cq.c
drivers/infiniband/hw/cxgb4/t4.h
drivers/infiniband/hw/hfi1/affinity.c
drivers/infiniband/hw/hfi1/debugfs.c
drivers/infiniband/hw/hfi1/driver.c
drivers/infiniband/hw/hfi1/file_ops.c
drivers/infiniband/hw/hfi1/hfi.h
drivers/infiniband/hw/hfi1/init.c
drivers/infiniband/hw/hfi1/mad.c
drivers/infiniband/hw/hfi1/qp.c
drivers/infiniband/hw/hfi1/qsfp.c
drivers/infiniband/hw/hfi1/qsfp.h
drivers/infiniband/hw/i40iw/i40iw.h
drivers/infiniband/hw/i40iw/i40iw_cm.c
drivers/infiniband/hw/i40iw/i40iw_main.c
drivers/infiniband/hw/i40iw/i40iw_utils.c
drivers/infiniband/hw/i40iw/i40iw_verbs.c
drivers/infiniband/hw/mlx4/cq.c
drivers/infiniband/hw/mlx5/main.c
drivers/infiniband/hw/ocrdma/ocrdma_hw.c
drivers/infiniband/hw/ocrdma/ocrdma_sli.h
drivers/infiniband/hw/ocrdma/ocrdma_verbs.c
drivers/infiniband/hw/qib/qib_debugfs.c
drivers/infiniband/hw/qib/qib_fs.c
drivers/infiniband/hw/qib/qib_qp.c
drivers/infiniband/hw/usnic/usnic_ib_main.c
drivers/infiniband/sw/rdmavt/qp.c
drivers/infiniband/ulp/isert/ib_isert.c
drivers/infiniband/ulp/srpt/ib_srpt.c
drivers/input/keyboard/tegra-kbc.c
drivers/input/rmi4/rmi_driver.c
drivers/input/serio/i8042.c
drivers/input/touchscreen/ads7846.c
drivers/input/touchscreen/silead.c
drivers/iommu/arm-smmu-v3.c
drivers/iommu/arm-smmu.c
drivers/iommu/io-pgtable-arm-v7s.c
drivers/md/bcache/super.c
drivers/md/dm-crypt.c
drivers/md/dm-flakey.c
drivers/md/dm-log.c
drivers/md/dm-raid.c
drivers/md/dm-round-robin.c
drivers/mmc/card/block.c
drivers/mmc/card/queue.c
drivers/mmc/card/queue.h
drivers/nvme/host/core.c
drivers/of/base.c
drivers/of/fdt.c
drivers/of/irq.c
drivers/of/platform.c
drivers/pci/msi.c
drivers/scsi/aacraid/commctrl.c
drivers/scsi/fcoe/fcoe_ctlr.c
drivers/scsi/megaraid/megaraid_sas_base.c
drivers/scsi/megaraid/megaraid_sas_fusion.c
drivers/scsi/mpt3sas/mpt3sas_base.c
drivers/scsi/ses.c
drivers/thermal/cpu_cooling.c
drivers/thermal/imx_thermal.c
drivers/thermal/int340x_thermal/int3406_thermal.c
drivers/usb/class/cdc-acm.c
drivers/usb/class/cdc-acm.h
drivers/usb/core/config.c
drivers/usb/core/devio.c
drivers/usb/core/hub.c
drivers/usb/dwc3/dwc3-of-simple.c
drivers/usb/dwc3/dwc3-pci.c
drivers/usb/dwc3/gadget.c
drivers/usb/gadget/composite.c
drivers/usb/gadget/configfs.c
drivers/usb/gadget/function/rndis.c
drivers/usb/gadget/function/u_ether.c
drivers/usb/gadget/function/uvc_configfs.c
drivers/usb/gadget/legacy/inode.c
drivers/usb/gadget/udc/core.c
drivers/usb/gadget/udc/fsl_qe_udc.c
drivers/usb/host/ehci-hcd.c
drivers/usb/host/max3421-hcd.c
drivers/usb/host/xhci-hub.c
drivers/usb/host/xhci-pci.c
drivers/usb/host/xhci-ring.c
drivers/usb/misc/ftdi-elan.c
drivers/usb/misc/usbtest.c
drivers/usb/phy/phy-omap-otg.c
drivers/usb/renesas_usbhs/common.c
drivers/usb/renesas_usbhs/fifo.c
drivers/usb/renesas_usbhs/mod_gadget.c
drivers/usb/serial/ftdi_sio.c
drivers/usb/serial/ftdi_sio_ids.h
drivers/usb/serial/option.c
drivers/usb/serial/usb-serial.c
drivers/vhost/scsi.c
drivers/xen/xenbus/xenbus_dev_frontend.c
fs/block_dev.c
fs/btrfs/backref.c
fs/btrfs/ctree.h
fs/btrfs/delayed-ref.c
fs/btrfs/disk-io.c
fs/btrfs/disk-io.h
fs/btrfs/extent-tree.c
fs/btrfs/extent_io.h
fs/btrfs/file.c
fs/btrfs/inode-map.c
fs/btrfs/inode.c
fs/btrfs/ioctl.c
fs/btrfs/qgroup.c
fs/btrfs/qgroup.h
fs/btrfs/relocation.c
fs/btrfs/root-tree.c
fs/btrfs/super.c
fs/btrfs/transaction.c
fs/btrfs/tree-log.c
fs/btrfs/tree-log.h
fs/btrfs/volumes.c
fs/dlm/debug_fs.c
fs/f2fs/data.c
fs/f2fs/f2fs.h
fs/f2fs/file.c
fs/f2fs/node.c
fs/f2fs/super.c
fs/iomap.c
fs/seq_file.c
fs/ubifs/tnc_commit.c
fs/ubifs/xattr.c
fs/xfs/libxfs/xfs_alloc.c
fs/xfs/libxfs/xfs_format.h
fs/xfs/libxfs/xfs_rmap_btree.c
fs/xfs/xfs_buf.c
fs/xfs/xfs_file.c
fs/xfs/xfs_fsops.c
fs/xfs/xfs_iomap.c
fs/xfs/xfs_iomap.h
fs/xfs/xfs_iops.c
fs/xfs/xfs_trace.h
include/linux/bio.h
include/linux/blkdev.h
include/linux/compiler-gcc.h
include/linux/compiler.h
include/linux/pci.h
include/linux/sysctl.h
include/rdma/ib_verbs.h
include/xen/xen-ops.h
kernel/events/core.c
kernel/events/uprobes.c
kernel/power/snapshot.c
kernel/printk/braille.c
kernel/sched/cputime.c
kernel/sysctl.c
kernel/trace/blktrace.c
mm/Kconfig
mm/huge_memory.c
mm/memcontrol.c
mm/readahead.c
mm/usercopy.c
scripts/get_maintainer.pl
security/Kconfig
sound/pci/hda/patch_realtek.c
sound/soc/atmel/atmel_ssc_dai.c
sound/soc/codecs/da7213.c
sound/soc/codecs/max98371.c
sound/soc/codecs/nau8825.c
sound/soc/codecs/wm2000.c
sound/soc/generic/Makefile
sound/soc/generic/simple-card-utils.c
sound/soc/intel/skylake/skl-sst-utils.c
sound/soc/intel/skylake/skl.c
sound/soc/omap/omap-abe-twl6040.c
sound/soc/omap/omap-mcpdm.c
sound/soc/samsung/s3c24xx_uda134x.c
sound/soc/sh/rcar/src.c
sound/soc/soc-compress.c
sound/soc/soc-core.c
sound/soc/soc-dapm.c
sound/usb/line6/pcm.c
sound/usb/line6/pod.c
tools/arch/arm64/include/uapi/asm/kvm.h
tools/arch/s390/include/uapi/asm/kvm.h
tools/arch/s390/include/uapi/asm/sie.h
tools/gpio/gpio-event-mon.c
tools/perf/arch/powerpc/util/sym-handling.c
tools/perf/arch/x86/util/intel-pt.c
tools/perf/builtin-mem.c
tools/perf/builtin-script.c
tools/perf/util/intel-pt-decoder/intel-pt-decoder.c
tools/perf/util/intel-pt-decoder/intel-pt-pkt-decoder.c
tools/perf/util/jitdump.c
tools/perf/util/probe-file.c
tools/perf/util/symbol-elf.c

index c55df29..cd9c9f6 100644 (file)
@@ -94,14 +94,11 @@ has a requirements for a minimum number of vectors the driver can pass a
 min_vecs argument set to this limit, and the PCI core will return -ENOSPC
 if it can't meet the minimum number of vectors.
 
-The flags argument should normally be set to 0, but can be used to pass the
-PCI_IRQ_NOMSI and PCI_IRQ_NOMSIX flag in case a device claims to support
-MSI or MSI-X, but the support is broken, or to pass PCI_IRQ_NOLEGACY in
-case the device does not support legacy interrupt lines.
-
-By default this function will spread the interrupts around the available
-CPUs, but this feature can be disabled by passing the PCI_IRQ_NOAFFINITY
-flag.
+The flags argument is used to specify which type of interrupt can be used
+by the device and the driver (PCI_IRQ_LEGACY, PCI_IRQ_MSI, PCI_IRQ_MSIX).
+A convenient short-hand (PCI_IRQ_ALL_TYPES) is also available to ask for
+any possible kind of interrupt.  If the PCI_IRQ_AFFINITY flag is set,
+pci_alloc_irq_vectors() will spread the interrupts around the available CPUs.
 
 To get the Linux IRQ numbers passed to request_irq() and free_irq() and the
 vectors, use the following function:
@@ -131,7 +128,7 @@ larger than the number supported by the device it will automatically be
 capped to the supported limit, so there is no need to query the number of
 vectors supported beforehand:
 
-       nvec = pci_alloc_irq_vectors(pdev, 1, nvec, 0);
+       nvec = pci_alloc_irq_vectors(pdev, 1, nvec, PCI_IRQ_ALL_TYPES)
        if (nvec < 0)
                goto out_err;
 
@@ -140,7 +137,7 @@ interrupts it can request a particular number of interrupts by passing that
 number to pci_alloc_irq_vectors() function as both 'min_vecs' and
 'max_vecs' parameters:
 
-       ret = pci_alloc_irq_vectors(pdev, nvec, nvec, 0);
+       ret = pci_alloc_irq_vectors(pdev, nvec, nvec, PCI_IRQ_ALL_TYPES);
        if (ret < 0)
                goto out_err;
 
@@ -148,15 +145,14 @@ The most notorious example of the request type described above is enabling
 the single MSI mode for a device.  It could be done by passing two 1s as
 'min_vecs' and 'max_vecs':
 
-       ret = pci_alloc_irq_vectors(pdev, 1, 1, 0);
+       ret = pci_alloc_irq_vectors(pdev, 1, 1, PCI_IRQ_ALL_TYPES);
        if (ret < 0)
                goto out_err;
 
 Some devices might not support using legacy line interrupts, in which case
-the PCI_IRQ_NOLEGACY flag can be used to fail the request if the platform
-can't provide MSI or MSI-X interrupts:
+the driver can specify that only MSI or MSI-X is acceptable:
 
-       nvec = pci_alloc_irq_vectors(pdev, 1, nvec, PCI_IRQ_NOLEGACY);
+       nvec = pci_alloc_irq_vectors(pdev, 1, nvec, PCI_IRQ_MSI | PCI_IRQ_MSIX);
        if (nvec < 0)
                goto out_err;
 
index 96b7aa6..106ae9c 100644 (file)
@@ -131,7 +131,7 @@ pygments_style = 'sphinx'
 todo_include_todos = False
 
 primary_domain = 'C'
-highlight_language = 'C'
+highlight_language = 'guess'
 
 # -- Options for HTML output ----------------------------------------------
 
index 6f6c2f8..0741dff 100644 (file)
@@ -8,8 +8,6 @@ Required properties:
 - interrupts: Interrupt number for McPDM
 - interrupt-parent: The parent interrupt controller
 - ti,hwmods: Name of the hwmod associated to the McPDM
-- clocks:  phandle for the pdmclk provider, likely <&twl6040>
-- clock-names: Must be "pdmclk"
 
 Example:
 
@@ -21,11 +19,3 @@ mcpdm: mcpdm@40132000 {
        interrupt-parent = <&gic>;
        ti,hwmods = "mcpdm";
 };
-
-In board DTS file the pdmclk needs to be added:
-
-&mcpdm {
-       clocks = <&twl6040>;
-       clock-names = "pdmclk";
-       status = "okay";
-};
index 41b817f..88b6ea1 100644 (file)
@@ -62,7 +62,7 @@ For more examples of cooling devices, refer to the example sections below.
 Required properties:
 - #cooling-cells:      Used to provide cooling device specific information
   Type: unsigned       while referring to it. Must be at least 2, in order
-  Size: one cell       to specify minimum and maximum cooling state used
+  Size: one cell       to specify minimum and maximum cooling state used
                        in the reference. The first cell is the minimum
                        cooling state requested and the second cell is
                        the maximum cooling state requested in the reference.
@@ -119,7 +119,7 @@ Required properties:
 Optional property:
 - contribution:                The cooling contribution to the thermal zone of the
   Type: unsigned       referred cooling device at the referred trip point.
-  Size: one cell       The contribution is a ratio of the sum
+  Size: one cell       The contribution is a ratio of the sum
                        of all cooling contributions within a thermal zone.
 
 Note: Using the THERMAL_NO_LIMIT (-1UL) constant in the cooling-device phandle
@@ -145,7 +145,7 @@ Required properties:
   Size: one cell
 
 - thermal-sensors:     A list of thermal sensor phandles and sensor specifier
-  Type: list of        used while monitoring the thermal zone.
+  Type: list of                used while monitoring the thermal zone.
   phandles + sensor
   specifier
 
@@ -473,7 +473,7 @@ thermal-zones {
                                  <&adc>;       /* pcb north */
 
                /* hotspot = 100 * bandgap - 120 * adc + 484 */
-               coefficients =          <100    -120    484>;
+               coefficients =          <100    -120    484>;
 
                trips {
                        ...
@@ -502,7 +502,7 @@ from the ADC sensor. The binding would be then:
         thermal-sensors =  <&adc>;
 
                /* hotspot = 1 * adc + 6000 */
-       coefficients =          <1      6000>;
+       coefficients =          <1      6000>;
 
 (d) - Board thermal
 
index 2a1bf69..8c10a91 100644 (file)
@@ -19,5 +19,5 @@ enhancements. It can monitor up to 4 voltages, 16 temperatures and
 implemented in this driver.
 
 Specification of the chip can be found here:
-ftp:///pub/Mainboard-OEM-Sales/Services/Software&Tools/Linux_SystemMonitoring&Watchdog&GPIO/BMC-Teutates_Specification_V1.21.pdf
-ftp:///pub/Mainboard-OEM-Sales/Services/Software&Tools/Linux_SystemMonitoring&Watchdog&GPIO/Fujitsu_mainboards-1-Sensors_HowTo-en-US.pdf
+ftp://ftp.ts.fujitsu.com/pub/Mainboard-OEM-Sales/Services/Software&Tools/Linux_SystemMonitoring&Watchdog&GPIO/BMC-Teutates_Specification_V1.21.pdf
+ftp://ftp.ts.fujitsu.com/pub/Mainboard-OEM-Sales/Services/Software&Tools/Linux_SystemMonitoring&Watchdog&GPIO/Fujitsu_mainboards-1-Sensors_HowTo-en-US.pdf
index c4eb504..391decc 100644 (file)
@@ -366,8 +366,6 @@ Domain`_ references.
 Cross-referencing from reStructuredText
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
-.. highlight:: none
-
 To cross-reference the functions and types defined in the kernel-doc comments
 from reStructuredText documents, please use the `Sphinx C Domain`_
 references. For example::
@@ -390,8 +388,6 @@ For further details, please refer to the `Sphinx C Domain`_ documentation.
 Function documentation
 ----------------------
 
-.. highlight:: c
-
 The general format of a function and function-like macro kernel-doc comment is::
 
   /**
@@ -572,8 +568,6 @@ DocBook XML [DEPRECATED]
 Converting DocBook to Sphinx
 ----------------------------
 
-.. highlight:: none
-
 Over time, we expect all of the documents under ``Documentation/DocBook`` to be
 converted to Sphinx and reStructuredText. For most DocBook XML documents, a good
 enough solution is to use the simple ``Documentation/sphinx/tmplcvt`` script,
index 46c030a..a4f4d69 100644 (file)
@@ -3032,6 +3032,10 @@ bytes respectively. Such letter suffixes can also be entirely omitted.
                                PAGE_SIZE is used as alignment.
                                PCI-PCI bridge can be specified, if resource
                                windows need to be expanded.
+                               To specify the alignment for several
+                               instances of a device, the PCI vendor,
+                               device, subvendor, and subdevice may be
+                               specified, e.g., 4096@pci:8086:9c22:103c:198f
                ecrc=           Enable/disable PCIe ECRC (transaction layer
                                end-to-end CRC checking).
                                bios: Use BIOS/firmware settings. This is the
index b96098c..708f87f 100644 (file)
@@ -164,7 +164,32 @@ load n/2 modules more and try again.
 Again, if you find the offending module(s), it(they) must be unloaded every time
 before hibernation, and please report the problem with it(them).
 
-c) Advanced debugging
+c) Using the "test_resume" hibernation option
+
+/sys/power/disk generally tells the kernel what to do after creating a
+hibernation image.  One of the available options is "test_resume" which
+causes the just created image to be used for immediate restoration.  Namely,
+after doing:
+
+# echo test_resume > /sys/power/disk
+# echo disk > /sys/power/state
+
+a hibernation image will be created and a resume from it will be triggered
+immediately without involving the platform firmware in any way.
+
+That test can be used to check if failures to resume from hibernation are
+related to bad interactions with the platform firmware.  That is, if the above
+works every time, but resume from actual hibernation does not work or is
+unreliable, the platform firmware may be responsible for the failures.
+
+On architectures and platforms that support using different kernels to restore
+hibernation images (that is, the kernel used to read the image from storage and
+load it into memory is different from the one included in the image) or support
+kernel address space randomization, it also can be used to check if failures
+to resume may be related to the differences between the restore and image
+kernels.
+
+d) Advanced debugging
 
 In case that hibernation does not work on your system even in the minimal
 configuration and compiling more drivers as modules is not practical or some
index f1f0f59..974916f 100644 (file)
@@ -1,75 +1,76 @@
-Power Management Interface
-
-
-The power management subsystem provides a unified sysfs interface to 
-userspace, regardless of what architecture or platform one is
-running. The interface exists in /sys/power/ directory (assuming sysfs
-is mounted at /sys). 
-
-/sys/power/state controls system power state. Reading from this file
-returns what states are supported, which is hard-coded to 'freeze',
-'standby' (Power-On Suspend), 'mem' (Suspend-to-RAM), and 'disk'
-(Suspend-to-Disk). 
-
-Writing to this file one of those strings causes the system to
-transition into that state. Please see the file
-Documentation/power/states.txt for a description of each of those
-states.
-
-
-/sys/power/disk controls the operating mode of the suspend-to-disk
-mechanism. Suspend-to-disk can be handled in several ways. We have a
-few options for putting the system to sleep - using the platform driver
-(e.g. ACPI or other suspend_ops), powering off the system or rebooting the
-system (for testing).
-
-Additionally, /sys/power/disk can be used to turn on one of the two testing
-modes of the suspend-to-disk mechanism: 'testproc' or 'test'.  If the
-suspend-to-disk mechanism is in the 'testproc' mode, writing 'disk' to
-/sys/power/state will cause the kernel to disable nonboot CPUs and freeze
-tasks, wait for 5 seconds, unfreeze tasks and enable nonboot CPUs.  If it is
-in the 'test' mode, writing 'disk' to /sys/power/state will cause the kernel
-to disable nonboot CPUs and freeze tasks, shrink memory, suspend devices, wait
-for 5 seconds, resume devices, unfreeze tasks and enable nonboot CPUs.  Then,
-we are able to look in the log messages and work out, for example, which code
-is being slow and which device drivers are misbehaving.
-
-Reading from this file will display all supported modes and the currently
-selected one in brackets, for example
-
-       [shutdown] reboot test testproc
-
-Writing to this file will accept one of
-
-       'platform' (only if the platform supports it)
-       'shutdown'
-       'reboot'
-       'testproc'
-       'test'
-
-/sys/power/image_size controls the size of the image created by
-the suspend-to-disk mechanism.  It can be written a string
-representing a non-negative integer that will be used as an upper
-limit of the image size, in bytes.  The suspend-to-disk mechanism will
-do its best to ensure the image size will not exceed that number.  However,
-if this turns out to be impossible, it will try to suspend anyway using the
-smallest image possible.  In particular, if "0" is written to this file, the
-suspend image will be as small as possible.
-
-Reading from this file will display the current image size limit, which
-is set to 2/5 of available RAM by default.
-
-/sys/power/pm_trace controls the code which saves the last PM event point in
-the RTC across reboots, so that you can debug a machine that just hangs
-during suspend (or more commonly, during resume).  Namely, the RTC is only
-used to save the last PM event point if this file contains '1'.  Initially it
-contains '0' which may be changed to '1' by writing a string representing a
-nonzero integer into it.
-
-To use this debugging feature you should attempt to suspend the machine, then
-reboot it and run
-
-       dmesg -s 1000000 | grep 'hash matches'
-
-CAUTION: Using it will cause your machine's real-time (CMOS) clock to be
-set to a random invalid time after a resume.
+Power Management Interface for System Sleep
+
+Copyright (c) 2016 Intel Corp., Rafael J. Wysocki <rafael.j.wysocki@intel.com>
+
+The power management subsystem provides userspace with a unified sysfs interface
+for system sleep regardless of the underlying system architecture or platform.
+The interface is located in the /sys/power/ directory (assuming that sysfs is
+mounted at /sys).
+
+/sys/power/state is the system sleep state control file.
+
+Reading from it returns a list of supported sleep states, encoded as:
+
+'freeze' (Suspend-to-Idle)
+'standby' (Power-On Suspend)
+'mem' (Suspend-to-RAM)
+'disk' (Suspend-to-Disk)
+
+Suspend-to-Idle is always supported.  Suspend-to-Disk is always supported
+too as long the kernel has been configured to support hibernation at all
+(ie. CONFIG_HIBERNATION is set in the kernel configuration file).  Support
+for Suspend-to-RAM and Power-On Suspend depends on the capabilities of the
+platform.
+
+If one of the strings listed in /sys/power/state is written to it, the system
+will attempt to transition into the corresponding sleep state.  Refer to
+Documentation/power/states.txt for a description of each of those states.
+
+/sys/power/disk controls the operating mode of hibernation (Suspend-to-Disk).
+Specifically, it tells the kernel what to do after creating a hibernation image.
+
+Reading from it returns a list of supported options encoded as:
+
+'platform' (put the system into sleep using a platform-provided method)
+'shutdown' (shut the system down)
+'reboot' (reboot the system)
+'suspend' (trigger a Suspend-to-RAM transition)
+'test_resume' (resume-after-hibernation test mode)
+
+The currently selected option is printed in square brackets.
+
+The 'platform' option is only available if the platform provides a special
+mechanism to put the system to sleep after creating a hibernation image (ACPI
+does that, for example).  The 'suspend' option is available if Suspend-to-RAM
+is supported.  Refer to Documentation/power/basic_pm_debugging.txt for the
+description of the 'test_resume' option.
+
+To select an option, write the string representing it to /sys/power/disk.
+
+/sys/power/image_size controls the size of hibernation images.
+
+It can be written a string representing a non-negative integer that will be
+used as a best-effort upper limit of the image size, in bytes.  The hibernation
+core will do its best to ensure that the image size will not exceed that number.
+However, if that turns out to be impossible to achieve, a hibernation image will
+still be created and its size will be as small as possible.  In particular,
+writing '0' to this file will enforce hibernation images to be as small as
+possible.
+
+Reading from this file returns the current image size limit, which is set to
+around 2/5 of available RAM by default.
+
+/sys/power/pm_trace controls the PM trace mechanism saving the last suspend
+or resume event point in the RTC across reboots.
+
+It helps to debug hard lockups or reboots due to device driver failures that
+occur during system suspend or resume (which is more common) more effectively.
+
+If /sys/power/pm_trace contains '1', the fingerprint of each suspend/resume
+event point in turn will be stored in the RTC memory (overwriting the actual
+RTC information), so it will survive a system crash if one occurs right after
+storing it and it can be used later to identify the driver that caused the crash
+to happen (see Documentation/power/s2ram.txt for more information).
+
+Initially it contains '0' which may be changed to '1' by writing a string
+representing a nonzero integer into it.
index 3a2ac4b..e88461c 100644 (file)
     caption a.headerlink { opacity: 0; }
     caption a.headerlink:hover { opacity: 1; }
 
-    /* inline literal: drop the borderbox and red color */
+    /* inline literal: drop the borderbox, padding and red color */
 
     code, .rst-content tt, .rst-content code {
         color: inherit;
         border: none;
+        padding: unset;
         background: inherit;
         font-size: 85%;
     }
index a306795..fc2613d 100644 (file)
@@ -4525,6 +4525,12 @@ L:       linux-edac@vger.kernel.org
 S:     Maintained
 F:     drivers/edac/sb_edac.c
 
+EDAC-SKYLAKE
+M:     Tony Luck <tony.luck@intel.com>
+L:     linux-edac@vger.kernel.org
+S:     Maintained
+F:     drivers/edac/skx_edac.c
+
 EDAC-XGENE
 APPLIED MICRO (APM) X-GENE SOC EDAC
 M:     Loc Ho <lho@apm.com>
@@ -7655,7 +7661,7 @@ L:        linux-rdma@vger.kernel.org
 S:     Supported
 W:     https://github.com/SoftRoCE/rxe-dev/wiki/rxe-dev:-Home
 Q:     http://patchwork.kernel.org/project/linux-rdma/list/
-F:     drivers/infiniband/hw/rxe/
+F:     drivers/infiniband/sw/rxe/
 F:     include/uapi/rdma/rdma_user_rxe.h
 
 MEMBARRIER SUPPORT
index 5c18baa..3537aa2 100644 (file)
--- a/Makefile
+++ b/Makefile
@@ -1,7 +1,7 @@
 VERSION = 4
 PATCHLEVEL = 8
 SUBLEVEL = 0
-EXTRAVERSION = -rc2
+EXTRAVERSION = -rc3
 NAME = Psychotic Stoned Sheep
 
 # *DOCUMENTATION*
index ad7860c..51597f3 100644 (file)
 
 #ifdef CONFIG_ARC_CURR_IN_REG
        ; Retrieve orig r25 and save it with rest of callee_regs
-       ld.as   r12, [r12, PT_user_r25]
+       ld      r12, [r12, PT_user_r25]
        PUSH    r12
 #else
        PUSH    r25
 
        ; SP is back to start of pt_regs
 #ifdef CONFIG_ARC_CURR_IN_REG
-       st.as   r12, [sp, PT_user_r25]
+       st      r12, [sp, PT_user_r25]
 #endif
 .endm
 
index c1d3645..4c6eed8 100644 (file)
@@ -188,10 +188,10 @@ static inline int arch_irqs_disabled(void)
 .endm
 
 .macro IRQ_ENABLE  scratch
+       TRACE_ASM_IRQ_ENABLE
        lr      \scratch, [status32]
        or      \scratch, \scratch, (STATUS_E1_MASK | STATUS_E2_MASK)
        flag    \scratch
-       TRACE_ASM_IRQ_ENABLE
 .endm
 
 #endif /* __ASSEMBLY__ */
index 0f92d97..89eeb37 100644 (file)
@@ -280,7 +280,7 @@ static inline void pmd_set(pmd_t *pmdp, pte_t *ptep)
 
 #define pte_page(pte)          pfn_to_page(pte_pfn(pte))
 #define mk_pte(page, prot)     pfn_pte(page_to_pfn(page), prot)
-#define pfn_pte(pfn, prot)     (__pte(((pte_t)(pfn) << PAGE_SHIFT) | pgprot_val(prot)))
+#define pfn_pte(pfn, prot)     __pte(((pfn) << PAGE_SHIFT) | pgprot_val(prot))
 
 /* Don't use virt_to_pfn for macros below: could cause truncations for PAE40*/
 #define pte_pfn(pte)           (pte_val(pte) >> PAGE_SHIFT)
index 0f99ac8..0037a58 100644 (file)
 
 /* Machine specific ELF Hdr flags */
 #define EF_ARC_OSABI_MSK       0x00000f00
-#define EF_ARC_OSABI_ORIG      0x00000000   /* MUST be zero for back-compat */
-#define EF_ARC_OSABI_CURRENT   0x00000300   /* v3 (no legacy syscalls) */
+
+#define EF_ARC_OSABI_V3                0x00000300   /* v3 (no legacy syscalls) */
+#define EF_ARC_OSABI_V4                0x00000400   /* v4 (64bit data any reg align) */
+
+#if __GNUC__ < 6
+#define EF_ARC_OSABI_CURRENT   EF_ARC_OSABI_V3
+#else
+#define EF_ARC_OSABI_CURRENT   EF_ARC_OSABI_V4
+#endif
 
 typedef unsigned long elf_greg_t;
 typedef unsigned long elf_fpregset_t;
index 4d9e777..000dd04 100644 (file)
@@ -28,6 +28,7 @@ extern void __muldf3(void);
 extern void __divdf3(void);
 extern void __floatunsidf(void);
 extern void __floatunsisf(void);
+extern void __udivdi3(void);
 
 EXPORT_SYMBOL(__ashldi3);
 EXPORT_SYMBOL(__ashrdi3);
@@ -45,6 +46,7 @@ EXPORT_SYMBOL(__muldf3);
 EXPORT_SYMBOL(__divdf3);
 EXPORT_SYMBOL(__floatunsidf);
 EXPORT_SYMBOL(__floatunsisf);
+EXPORT_SYMBOL(__udivdi3);
 
 /* ARC optimised assembler routines */
 EXPORT_SYMBOL(memset);
index b5db9e7..be1972b 100644 (file)
@@ -199,7 +199,7 @@ int elf_check_arch(const struct elf32_hdr *x)
        }
 
        eflags = x->e_flags;
-       if ((eflags & EF_ARC_OSABI_MSK) < EF_ARC_OSABI_CURRENT) {
+       if ((eflags & EF_ARC_OSABI_MSK) != EF_ARC_OSABI_CURRENT) {
                pr_err("ABI mismatch - you need newer toolchain\n");
                force_sigsegv(SIGSEGV, current);
                return 0;
index a946400..f52a0d0 100644 (file)
@@ -291,8 +291,10 @@ static char *arc_extn_mumbojumbo(int cpu_id, char *buf, int len)
                               cpu->dccm.base_addr, TO_KB(cpu->dccm.sz),
                               cpu->iccm.base_addr, TO_KB(cpu->iccm.sz));
 
-       n += scnprintf(buf + n, len - n,
-                      "OS ABI [v3]\t: no-legacy-syscalls\n");
+       n += scnprintf(buf + n, len - n, "OS ABI [v%d]\t: %s\n",
+                       EF_ARC_OSABI_CURRENT >> 8,
+                       EF_ARC_OSABI_CURRENT == EF_ARC_OSABI_V3 ?
+                       "no-legacy-syscalls" : "64-bit data any register aligned");
 
        return buf;
 }
index 5a294b2..0b10efe 100644 (file)
@@ -921,6 +921,15 @@ void arc_cache_init(void)
 
        printk(arc_cache_mumbojumbo(0, str, sizeof(str)));
 
+       /*
+        * Only master CPU needs to execute rest of function:
+        *  - Assume SMP so all cores will have same cache config so
+        *    any geomtry checks will be same for all
+        *  - IOC setup / dma callbacks only need to be setup once
+        */
+       if (cpu)
+               return;
+
        if (IS_ENABLED(CONFIG_ARC_HAS_ICACHE)) {
                struct cpuinfo_arc_cache *ic = &cpuinfo_arc700[cpu].icache;
 
index 04f8332..77ff64a 100644 (file)
@@ -61,6 +61,7 @@ void *kmap(struct page *page)
 
        return kmap_high(page);
 }
+EXPORT_SYMBOL(kmap);
 
 void *kmap_atomic(struct page *page)
 {
index bc5f507..9f157e7 100644 (file)
@@ -295,6 +295,7 @@ __und_svc_fault:
        bl      __und_fault
 
 __und_svc_finish:
+       get_thread_info tsk
        ldr     r5, [sp, #S_PSR]                @ Get SVC cpsr
        svc_exit r5                             @ return from exception
  UNWIND(.fnend         )
index fd87205..0df062d 100644 (file)
@@ -271,6 +271,12 @@ static int __init imx_gpc_init(struct device_node *node,
        for (i = 0; i < IMR_NUM; i++)
                writel_relaxed(~0, gpc_base + GPC_IMR1 + i * 4);
 
+       /*
+        * Clear the OF_POPULATED flag set in of_irq_init so that
+        * later the GPC power domain driver will not be skipped.
+        */
+       of_node_clear_flag(node, OF_POPULATED);
+
        return 0;
 }
 IRQCHIP_DECLARE(imx_gpc, "fsl,imx6q-gpc", imx_gpc_init);
index 62f4d01..6344913 100644 (file)
@@ -728,7 +728,8 @@ static void *__init late_alloc(unsigned long sz)
 {
        void *ptr = (void *)__get_free_pages(PGALLOC_GFP, get_order(sz));
 
-       BUG_ON(!ptr);
+       if (!ptr || !pgtable_page_ctor(virt_to_page(ptr)))
+               BUG();
        return ptr;
 }
 
@@ -1155,10 +1156,19 @@ void __init sanity_check_meminfo(void)
 {
        phys_addr_t memblock_limit = 0;
        int highmem = 0;
-       phys_addr_t vmalloc_limit = __pa(vmalloc_min - 1) + 1;
+       u64 vmalloc_limit;
        struct memblock_region *reg;
        bool should_use_highmem = false;
 
+       /*
+        * Let's use our own (unoptimized) equivalent of __pa() that is
+        * not affected by wrap-arounds when sizeof(phys_addr_t) == 4.
+        * The result is used as the upper bound on physical memory address
+        * and may itself be outside the valid range for which phys_addr_t
+        * and therefore __pa() is defined.
+        */
+       vmalloc_limit = (u64)(uintptr_t)vmalloc_min - PAGE_OFFSET + PHYS_OFFSET;
+
        for_each_memblock(memory, reg) {
                phys_addr_t block_start = reg->base;
                phys_addr_t block_end = reg->base + reg->size;
@@ -1183,10 +1193,11 @@ void __init sanity_check_meminfo(void)
                        if (reg->size > size_limit) {
                                phys_addr_t overlap_size = reg->size - size_limit;
 
-                               pr_notice("Truncating RAM at %pa-%pa to -%pa",
-                                         &block_start, &block_end, &vmalloc_limit);
-                               memblock_remove(vmalloc_limit, overlap_size);
+                               pr_notice("Truncating RAM at %pa-%pa",
+                                         &block_start, &block_end);
                                block_end = vmalloc_limit;
+                               pr_cont(" to -%pa", &block_end);
+                               memblock_remove(vmalloc_limit, overlap_size);
                                should_use_highmem = true;
                        }
                }
index b0b82f5..3d2cef6 100644 (file)
@@ -50,7 +50,7 @@ DEFINE_PER_CPU(struct vcpu_info *, xen_vcpu);
 static struct vcpu_info __percpu *xen_vcpu_info;
 
 /* Linux <-> Xen vCPU id mapping */
-DEFINE_PER_CPU(int, xen_vcpu_id) = -1;
+DEFINE_PER_CPU(uint32_t, xen_vcpu_id);
 EXPORT_PER_CPU_SYMBOL(xen_vcpu_id);
 
 /* These are unused until we support booting "pre-ballooned" */
index b77f583..3e7b050 100644 (file)
@@ -757,6 +757,9 @@ ENTRY(__enable_mmu)
        isb
        bl      __create_page_tables            // recreate kernel mapping
 
+       tlbi    vmalle1                         // Remove any stale TLB entries
+       dsb     nsh
+
        msr     sctlr_el1, x19                  // re-enable the MMU
        isb
        ic      iallu                           // flush instructions fetched
index 9a3aec9..ccf79d8 100644 (file)
@@ -101,12 +101,20 @@ ENTRY(cpu_resume)
        bl      el2_setup               // if in EL2 drop to EL1 cleanly
        /* enable the MMU early - so we can access sleep_save_stash by va */
        adr_l   lr, __enable_mmu        /* __cpu_setup will return here */
-       ldr     x27, =_cpu_resume       /* __enable_mmu will branch here */
+       adr_l   x27, _resume_switched   /* __enable_mmu will branch here */
        adrp    x25, idmap_pg_dir
        adrp    x26, swapper_pg_dir
        b       __cpu_setup
 ENDPROC(cpu_resume)
 
+       .pushsection    ".idmap.text", "ax"
+_resume_switched:
+       ldr     x8, =_cpu_resume
+       br      x8
+ENDPROC(_resume_switched)
+       .ltorg
+       .popsection
+
 ENTRY(_cpu_resume)
        mrs     x1, mpidr_el1
        adrp    x8, mpidr_hash
index f94b80e..9c3e75d 100644 (file)
@@ -242,7 +242,7 @@ static void note_page(struct pg_state *st, unsigned long addr, unsigned level,
 
 static void walk_pte(struct pg_state *st, pmd_t *pmd, unsigned long start)
 {
-       pte_t *pte = pte_offset_kernel(pmd, 0);
+       pte_t *pte = pte_offset_kernel(pmd, 0UL);
        unsigned long addr;
        unsigned i;
 
@@ -254,7 +254,7 @@ static void walk_pte(struct pg_state *st, pmd_t *pmd, unsigned long start)
 
 static void walk_pmd(struct pg_state *st, pud_t *pud, unsigned long start)
 {
-       pmd_t *pmd = pmd_offset(pud, 0);
+       pmd_t *pmd = pmd_offset(pud, 0UL);
        unsigned long addr;
        unsigned i;
 
@@ -271,7 +271,7 @@ static void walk_pmd(struct pg_state *st, pud_t *pud, unsigned long start)
 
 static void walk_pud(struct pg_state *st, pgd_t *pgd, unsigned long start)
 {
-       pud_t *pud = pud_offset(pgd, 0);
+       pud_t *pud = pud_offset(pgd, 0UL);
        unsigned long addr;
        unsigned i;
 
index c7fe3ec..5bb15ea 100644 (file)
@@ -23,6 +23,8 @@
 #include <linux/module.h>
 #include <linux/of.h>
 
+#include <asm/acpi.h>
+
 struct pglist_data *node_data[MAX_NUMNODES] __read_mostly;
 EXPORT_SYMBOL(node_data);
 nodemask_t numa_nodes_parsed __initdata;
index ea0cd97..5f98759 100644 (file)
@@ -164,7 +164,7 @@ typedef struct { unsigned long pgprot; } pgprot_t;
  */
 static inline unsigned long ___pa(unsigned long x)
 {
-       if (config_enabled(CONFIG_64BIT)) {
+       if (IS_ENABLED(CONFIG_64BIT)) {
                /*
                 * For MIPS64 the virtual address may either be in one of
                 * the compatibility segements ckseg0 or ckseg1, or it may
@@ -173,7 +173,7 @@ static inline unsigned long ___pa(unsigned long x)
                return x < CKSEG0 ? XPHYSADDR(x) : CPHYSADDR(x);
        }
 
-       if (!config_enabled(CONFIG_EVA)) {
+       if (!IS_ENABLED(CONFIG_EVA)) {
                /*
                 * We're using the standard MIPS32 legacy memory map, ie.
                 * the address x is going to be in kseg0 or kseg1. We can
index c0ae625..274d5bc 100644 (file)
 #define        ENOTCONN        235     /* Transport endpoint is not connected */
 #define        ESHUTDOWN       236     /* Cannot send after transport endpoint shutdown */
 #define        ETOOMANYREFS    237     /* Too many references: cannot splice */
-#define EREFUSED       ECONNREFUSED    /* for HP's NFS apparently */
 #define        ETIMEDOUT       238     /* Connection timed out */
 #define        ECONNREFUSED    239     /* Connection refused */
-#define EREMOTERELEASE 240     /* Remote peer released connection */
+#define        EREFUSED        ECONNREFUSED    /* for HP's NFS apparently */
+#define        EREMOTERELEASE  240     /* Remote peer released connection */
 #define        EHOSTDOWN       241     /* Host is down */
 #define        EHOSTUNREACH    242     /* No route to host */
 
index 5adc339..0c2a94a 100644 (file)
@@ -51,8 +51,6 @@ EXPORT_SYMBOL(_parisc_requires_coherency);
 
 DEFINE_PER_CPU(struct cpuinfo_parisc, cpu_data);
 
-extern int update_cr16_clocksource(void);      /* from time.c */
-
 /*
 **     PARISC CPU driver - claim "device" and initialize CPU data structures.
 **
@@ -228,12 +226,6 @@ static int processor_probe(struct parisc_device *dev)
        }
 #endif
 
-       /* If we've registered more than one cpu,
-        * we'll use the jiffies clocksource since cr16
-        * is not synchronized between CPUs.
-        */
-       update_cr16_clocksource();
-
        return 0;
 }
 
index 505cf1a..4b0b963 100644 (file)
@@ -221,18 +221,6 @@ static struct clocksource clocksource_cr16 = {
        .flags                  = CLOCK_SOURCE_IS_CONTINUOUS,
 };
 
-int update_cr16_clocksource(void)
-{
-       /* since the cr16 cycle counters are not synchronized across CPUs,
-          we'll check if we should switch to a safe clocksource: */
-       if (clocksource_cr16.rating != 0 && num_online_cpus() > 1) {
-               clocksource_change_rating(&clocksource_cr16, 0);
-               return 1;
-       }
-
-       return 0;
-}
-
 void __init start_cpu_itimer(void)
 {
        unsigned int cpu = smp_processor_id();
index ba5f456..7f7ba5f 100644 (file)
@@ -204,11 +204,9 @@ static void __init conmode_default(void)
 #endif
                }
        } else if (MACHINE_IS_KVM) {
-               if (sclp.has_vt220 &&
-                   config_enabled(CONFIG_SCLP_VT220_CONSOLE))
+               if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE))
                        SET_CONSOLE_VT220;
-               else if (sclp.has_linemode &&
-                        config_enabled(CONFIG_SCLP_CONSOLE))
+               else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE))
                        SET_CONSOLE_SCLP;
                else
                        SET_CONSOLE_HVC;
index 1dd5bd8..1330553 100644 (file)
@@ -81,7 +81,7 @@
   .altinstr_replacement : { *(.altinstr_replacement) }
   /* .exit.text is discard at runtime, not link time, to deal with references
      from .altinstructions and .eh_frame */
-  .exit.text : { *(.exit.text) }
+  .exit.text : { EXIT_TEXT }
   .exit.data : { *(.exit.data) }
 
   .preinit_array : {
index 89fa85e..6f97fb3 100644 (file)
@@ -485,10 +485,10 @@ static int sha_complete_job(struct mcryptd_hash_request_ctx *rctx,
 
                        req = cast_mcryptd_ctx_to_req(req_ctx);
                        if (irqs_disabled())
-                               rctx->complete(&req->base, ret);
+                               req_ctx->complete(&req->base, ret);
                        else {
                                local_bh_disable();
-                               rctx->complete(&req->base, ret);
+                               req_ctx->complete(&req->base, ret);
                                local_bh_enable();
                        }
                }
index b691da9..a78a069 100644 (file)
@@ -265,13 +265,14 @@ ENTRY(sha256_mb_mgr_get_comp_job_avx2)
        vpinsrd $1, _args_digest+1*32(state, idx, 4), %xmm0, %xmm0
        vpinsrd $2, _args_digest+2*32(state, idx, 4), %xmm0, %xmm0
        vpinsrd $3, _args_digest+3*32(state, idx, 4), %xmm0, %xmm0
-       movl    _args_digest+4*32(state, idx, 4), tmp2_w
+       vmovd   _args_digest(state , idx, 4) , %xmm0
        vpinsrd $1, _args_digest+5*32(state, idx, 4), %xmm1, %xmm1
        vpinsrd $2, _args_digest+6*32(state, idx, 4), %xmm1, %xmm1
        vpinsrd $3, _args_digest+7*32(state, idx, 4), %xmm1, %xmm1
 
-       vmovdqu %xmm0, _result_digest(job_rax)
-       movl    tmp2_w, _result_digest+1*16(job_rax)
+        vmovdqu %xmm0, _result_digest(job_rax)
+        offset =  (_result_digest + 1*16)
+        vmovdqu %xmm1, offset(job_rax)
 
        pop     %rbx
 
index f4cf5b7..d210174 100644 (file)
@@ -497,10 +497,10 @@ static int sha_complete_job(struct mcryptd_hash_request_ctx *rctx,
 
                        req = cast_mcryptd_ctx_to_req(req_ctx);
                        if (irqs_disabled())
-                               rctx->complete(&req->base, ret);
+                               req_ctx->complete(&req->base, ret);
                        else {
                                local_bh_disable();
-                               rctx->complete(&req->base, ret);
+                               req_ctx->complete(&req->base, ret);
                                local_bh_enable();
                        }
                }
index 27a0228..b816971 100644 (file)
@@ -355,6 +355,7 @@ void load_ucode_amd_ap(void)
        unsigned int cpu = smp_processor_id();
        struct equiv_cpu_entry *eq;
        struct microcode_amd *mc;
+       u8 *cont = container;
        u32 rev, eax;
        u16 eq_id;
 
@@ -371,8 +372,11 @@ void load_ucode_amd_ap(void)
        if (check_current_patch_level(&rev, false))
                return;
 
+       /* Add CONFIG_RANDOMIZE_MEMORY offset. */
+       cont += PAGE_OFFSET - __PAGE_OFFSET_BASE;
+
        eax = cpuid_eax(0x00000001);
-       eq  = (struct equiv_cpu_entry *)(container + CONTAINER_HDR_SZ);
+       eq  = (struct equiv_cpu_entry *)(cont + CONTAINER_HDR_SZ);
 
        eq_id = find_equiv_id(eq, eax);
        if (!eq_id)
@@ -434,6 +438,9 @@ int __init save_microcode_in_initrd_amd(void)
        else
                container = cont_va;
 
+       /* Add CONFIG_RANDOMIZE_MEMORY offset. */
+       container += PAGE_OFFSET - __PAGE_OFFSET_BASE;
+
        eax   = cpuid_eax(0x00000001);
        eax   = ((eax >> 8) & 0xf) + ((eax >> 20) & 0xff);
 
index 2a6e84a..4296beb 100644 (file)
@@ -100,10 +100,11 @@ EXPORT_PER_CPU_SYMBOL(cpu_info);
 /* Logical package management. We might want to allocate that dynamically */
 static int *physical_to_logical_pkg __read_mostly;
 static unsigned long *physical_package_map __read_mostly;;
-static unsigned long *logical_package_map  __read_mostly;
 static unsigned int max_physical_pkg_id __read_mostly;
 unsigned int __max_logical_packages __read_mostly;
 EXPORT_SYMBOL(__max_logical_packages);
+static unsigned int logical_packages __read_mostly;
+static bool logical_packages_frozen __read_mostly;
 
 /* Maximum number of SMT threads on any online core */
 int __max_smt_threads __read_mostly;
@@ -277,14 +278,14 @@ int topology_update_package_map(unsigned int apicid, unsigned int cpu)
        if (test_and_set_bit(pkg, physical_package_map))
                goto found;
 
-       new = find_first_zero_bit(logical_package_map, __max_logical_packages);
-       if (new >= __max_logical_packages) {
+       if (logical_packages_frozen) {
                physical_to_logical_pkg[pkg] = -1;
-               pr_warn("APIC(%x) Package %u exceeds logical package map\n",
+               pr_warn("APIC(%x) Package %u exceeds logical package max\n",
                        apicid, pkg);
                return -ENOSPC;
        }
-       set_bit(new, logical_package_map);
+
+       new = logical_packages++;
        pr_info("APIC(%x) Converting physical %u to logical package %u\n",
                apicid, pkg, new);
        physical_to_logical_pkg[pkg] = new;
@@ -341,6 +342,7 @@ static void __init smp_init_package_map(void)
        }
 
        __max_logical_packages = DIV_ROUND_UP(total_cpus, ncpus);
+       logical_packages = 0;
 
        /*
         * Possibly larger than what we need as the number of apic ids per
@@ -352,10 +354,6 @@ static void __init smp_init_package_map(void)
        memset(physical_to_logical_pkg, 0xff, size);
        size = BITS_TO_LONGS(max_physical_pkg_id) * sizeof(unsigned long);
        physical_package_map = kzalloc(size, GFP_KERNEL);
-       size = BITS_TO_LONGS(__max_logical_packages) * sizeof(unsigned long);
-       logical_package_map = kzalloc(size, GFP_KERNEL);
-
-       pr_info("Max logical packages: %u\n", __max_logical_packages);
 
        for_each_present_cpu(cpu) {
                unsigned int apicid = apic->cpu_present_to_apicid(cpu);
@@ -369,6 +367,15 @@ static void __init smp_init_package_map(void)
                set_cpu_possible(cpu, false);
                set_cpu_present(cpu, false);
        }
+
+       if (logical_packages > __max_logical_packages) {
+               pr_warn("Detected more packages (%u), then computed by BIOS data (%u).\n",
+                       logical_packages, __max_logical_packages);
+               logical_packages_frozen = true;
+               __max_logical_packages  = logical_packages;
+       }
+
+       pr_info("Max logical packages: %u\n", __max_logical_packages);
 }
 
 void __init smp_store_boot_cpu_info(void)
index ec8654f..bda8d5e 100644 (file)
@@ -77,7 +77,7 @@ static inline unsigned long get_padding(struct kaslr_memory_region *region)
  */
 static inline bool kaslr_memory_enabled(void)
 {
-       return kaslr_enabled() && !config_enabled(CONFIG_KASAN);
+       return kaslr_enabled() && !IS_ENABLED(CONFIG_KASAN);
 }
 
 /* Initialize base and padding for each memory region randomized with KASLR */
index b814ca6..7948be3 100644 (file)
@@ -41,6 +41,7 @@ static DEFINE_RAW_SPINLOCK(list_lock);
  * @node:      list item for parent traversal.
  * @rcu:       RCU callback item for freeing.
  * @irq:       back pointer to parent.
+ * @enabled:   true if driver enabled IRQ
  * @virq:      the virtual IRQ value provided to the requesting driver.
  *
  * Every MSI/MSI-X IRQ requested for a device in a VMD domain will be mapped to
@@ -50,6 +51,7 @@ struct vmd_irq {
        struct list_head        node;
        struct rcu_head         rcu;
        struct vmd_irq_list     *irq;
+       bool                    enabled;
        unsigned int            virq;
 };
 
@@ -122,7 +124,9 @@ static void vmd_irq_enable(struct irq_data *data)
        unsigned long flags;
 
        raw_spin_lock_irqsave(&list_lock, flags);
+       WARN_ON(vmdirq->enabled);
        list_add_tail_rcu(&vmdirq->node, &vmdirq->irq->irq_list);
+       vmdirq->enabled = true;
        raw_spin_unlock_irqrestore(&list_lock, flags);
 
        data->chip->irq_unmask(data);
@@ -136,8 +140,10 @@ static void vmd_irq_disable(struct irq_data *data)
        data->chip->irq_mask(data);
 
        raw_spin_lock_irqsave(&list_lock, flags);
-       list_del_rcu(&vmdirq->node);
-       INIT_LIST_HEAD_RCU(&vmdirq->node);
+       if (vmdirq->enabled) {
+               list_del_rcu(&vmdirq->node);
+               vmdirq->enabled = false;
+       }
        raw_spin_unlock_irqrestore(&list_lock, flags);
 }
 
index a3e3ccc..9634557 100644 (file)
@@ -113,7 +113,7 @@ static int set_up_temporary_mappings(void)
                        return result;
        }
 
-       temp_level4_pgt = (unsigned long)pgd - __PAGE_OFFSET;
+       temp_level4_pgt = __pa(pgd);
        return 0;
 }
 
index 8ffb089..b86ebb1 100644 (file)
@@ -118,7 +118,7 @@ DEFINE_PER_CPU(struct vcpu_info *, xen_vcpu);
 DEFINE_PER_CPU(struct vcpu_info, xen_vcpu_info);
 
 /* Linux <-> Xen vCPU id mapping */
-DEFINE_PER_CPU(int, xen_vcpu_id) = -1;
+DEFINE_PER_CPU(uint32_t, xen_vcpu_id);
 EXPORT_PER_CPU_SYMBOL(xen_vcpu_id);
 
 enum xen_domain_type xen_domain_type = XEN_NATIVE;
index f394775..aa73540 100644 (file)
@@ -667,18 +667,19 @@ struct bio *bio_clone_bioset(struct bio *bio_src, gfp_t gfp_mask,
        bio->bi_iter.bi_sector  = bio_src->bi_iter.bi_sector;
        bio->bi_iter.bi_size    = bio_src->bi_iter.bi_size;
 
-       if (bio_op(bio) == REQ_OP_DISCARD)
-               goto integrity_clone;
-
-       if (bio_op(bio) == REQ_OP_WRITE_SAME) {
+       switch (bio_op(bio)) {
+       case REQ_OP_DISCARD:
+       case REQ_OP_SECURE_ERASE:
+               break;
+       case REQ_OP_WRITE_SAME:
                bio->bi_io_vec[bio->bi_vcnt++] = bio_src->bi_io_vec[0];
-               goto integrity_clone;
+               break;
+       default:
+               bio_for_each_segment(bv, bio_src, iter)
+                       bio->bi_io_vec[bio->bi_vcnt++] = bv;
+               break;
        }
 
-       bio_for_each_segment(bv, bio_src, iter)
-               bio->bi_io_vec[bio->bi_vcnt++] = bv;
-
-integrity_clone:
        if (bio_integrity(bio_src)) {
                int ret;
 
@@ -1788,7 +1789,7 @@ struct bio *bio_split(struct bio *bio, int sectors,
         * Discards need a mutable bio_vec to accommodate the payload
         * required by the DSM TRIM and UNMAP commands.
         */
-       if (bio_op(bio) == REQ_OP_DISCARD)
+       if (bio_op(bio) == REQ_OP_DISCARD || bio_op(bio) == REQ_OP_SECURE_ERASE)
                split = bio_clone_bioset(bio, gfp, bs);
        else
                split = bio_clone_fast(bio, gfp, bs);
index 999442e..36c7ac3 100644 (file)
@@ -515,7 +515,9 @@ EXPORT_SYMBOL_GPL(blk_queue_bypass_end);
 
 void blk_set_queue_dying(struct request_queue *q)
 {
-       queue_flag_set_unlocked(QUEUE_FLAG_DYING, q);
+       spin_lock_irq(q->queue_lock);
+       queue_flag_set(QUEUE_FLAG_DYING, q);
+       spin_unlock_irq(q->queue_lock);
 
        if (q->mq_ops)
                blk_mq_wake_waiters(q);
index 3eec75a..2642e5f 100644 (file)
@@ -94,8 +94,30 @@ static struct bio *blk_bio_segment_split(struct request_queue *q,
        bool do_split = true;
        struct bio *new = NULL;
        const unsigned max_sectors = get_max_io_size(q, bio);
+       unsigned bvecs = 0;
 
        bio_for_each_segment(bv, bio, iter) {
+               /*
+                * With arbitrary bio size, the incoming bio may be very
+                * big. We have to split the bio into small bios so that
+                * each holds at most BIO_MAX_PAGES bvecs because
+                * bio_clone() can fail to allocate big bvecs.
+                *
+                * It should have been better to apply the limit per
+                * request queue in which bio_clone() is involved,
+                * instead of globally. The biggest blocker is the
+                * bio_clone() in bio bounce.
+                *
+                * If bio is splitted by this reason, we should have
+                * allowed to continue bios merging, but don't do
+                * that now for making the change simple.
+                *
+                * TODO: deal with bio bounce's bio_clone() gracefully
+                * and convert the global limit into per-queue limit.
+                */
+               if (bvecs++ >= BIO_MAX_PAGES)
+                       goto split;
+
                /*
                 * If the queue doesn't support SG gaps and adding this
                 * offset would create a gap, disallow it.
@@ -172,12 +194,18 @@ void blk_queue_split(struct request_queue *q, struct bio **bio,
        struct bio *split, *res;
        unsigned nsegs;
 
-       if (bio_op(*bio) == REQ_OP_DISCARD)
+       switch (bio_op(*bio)) {
+       case REQ_OP_DISCARD:
+       case REQ_OP_SECURE_ERASE:
                split = blk_bio_discard_split(q, *bio, bs, &nsegs);
-       else if (bio_op(*bio) == REQ_OP_WRITE_SAME)
+               break;
+       case REQ_OP_WRITE_SAME:
                split = blk_bio_write_same_split(q, *bio, bs, &nsegs);
-       else
+               break;
+       default:
                split = blk_bio_segment_split(q, *bio, q->bio_split, &nsegs);
+               break;
+       }
 
        /* physical segments can be figured out during splitting */
        res = split ? split : *bio;
@@ -213,7 +241,7 @@ static unsigned int __blk_recalc_rq_segments(struct request_queue *q,
         * This should probably be returning 0, but blk_add_request_payload()
         * (Christoph!!!!)
         */
-       if (bio_op(bio) == REQ_OP_DISCARD)
+       if (bio_op(bio) == REQ_OP_DISCARD || bio_op(bio) == REQ_OP_SECURE_ERASE)
                return 1;
 
        if (bio_op(bio) == REQ_OP_WRITE_SAME)
@@ -385,7 +413,9 @@ static int __blk_bios_map_sg(struct request_queue *q, struct bio *bio,
        nsegs = 0;
        cluster = blk_queue_cluster(q);
 
-       if (bio_op(bio) == REQ_OP_DISCARD) {
+       switch (bio_op(bio)) {
+       case REQ_OP_DISCARD:
+       case REQ_OP_SECURE_ERASE:
                /*
                 * This is a hack - drivers should be neither modifying the
                 * biovec, nor relying on bi_vcnt - but because of
@@ -393,19 +423,16 @@ static int __blk_bios_map_sg(struct request_queue *q, struct bio *bio,
                 * a payload we need to set up here (thank you Christoph) and
                 * bi_vcnt is really the only way of telling if we need to.
                 */
-
-               if (bio->bi_vcnt)
-                       goto single_segment;
-
-               return 0;
-       }
-
-       if (bio_op(bio) == REQ_OP_WRITE_SAME) {
-single_segment:
+               if (!bio->bi_vcnt)
+                       return 0;
+               /* Fall through */
+       case REQ_OP_WRITE_SAME:
                *sg = sglist;
                bvec = bio_iovec(bio);
                sg_set_page(*sg, bvec.bv_page, bvec.bv_len, bvec.bv_offset);
                return 1;
+       default:
+               break;
        }
 
        for_each_bio(bio)
index e931a0e..13f5a6c 100644 (file)
@@ -793,11 +793,12 @@ static void __blk_mq_run_hw_queue(struct blk_mq_hw_ctx *hctx)
        struct list_head *dptr;
        int queued;
 
-       WARN_ON(!cpumask_test_cpu(raw_smp_processor_id(), hctx->cpumask));
-
        if (unlikely(test_bit(BLK_MQ_S_STOPPED, &hctx->state)))
                return;
 
+       WARN_ON(!cpumask_test_cpu(raw_smp_processor_id(), hctx->cpumask) &&
+               cpu_online(hctx->next_cpu));
+
        hctx->run++;
 
        /*
@@ -1036,10 +1037,11 @@ void blk_mq_delay_queue(struct blk_mq_hw_ctx *hctx, unsigned long msecs)
 EXPORT_SYMBOL(blk_mq_delay_queue);
 
 static inline void __blk_mq_insert_req_list(struct blk_mq_hw_ctx *hctx,
-                                           struct blk_mq_ctx *ctx,
                                            struct request *rq,
                                            bool at_head)
 {
+       struct blk_mq_ctx *ctx = rq->mq_ctx;
+
        trace_block_rq_insert(hctx->queue, rq);
 
        if (at_head)
@@ -1053,20 +1055,16 @@ static void __blk_mq_insert_request(struct blk_mq_hw_ctx *hctx,
 {
        struct blk_mq_ctx *ctx = rq->mq_ctx;
 
-       __blk_mq_insert_req_list(hctx, ctx, rq, at_head);
+       __blk_mq_insert_req_list(hctx, rq, at_head);
        blk_mq_hctx_mark_pending(hctx, ctx);
 }
 
 void blk_mq_insert_request(struct request *rq, bool at_head, bool run_queue,
-               bool async)
+                          bool async)
 {
+       struct blk_mq_ctx *ctx = rq->mq_ctx;
        struct request_queue *q = rq->q;
        struct blk_mq_hw_ctx *hctx;
-       struct blk_mq_ctx *ctx = rq->mq_ctx, *current_ctx;
-
-       current_ctx = blk_mq_get_ctx(q);
-       if (!cpu_online(ctx->cpu))
-               rq->mq_ctx = ctx = current_ctx;
 
        hctx = q->mq_ops->map_queue(q, ctx->cpu);
 
@@ -1076,8 +1074,6 @@ void blk_mq_insert_request(struct request *rq, bool at_head, bool run_queue,
 
        if (run_queue)
                blk_mq_run_hw_queue(hctx, async);
-
-       blk_mq_put_ctx(current_ctx);
 }
 
 static void blk_mq_insert_requests(struct request_queue *q,
@@ -1088,14 +1084,9 @@ static void blk_mq_insert_requests(struct request_queue *q,
 
 {
        struct blk_mq_hw_ctx *hctx;
-       struct blk_mq_ctx *current_ctx;
 
        trace_block_unplug(q, depth, !from_schedule);
 
-       current_ctx = blk_mq_get_ctx(q);
-
-       if (!cpu_online(ctx->cpu))
-               ctx = current_ctx;
        hctx = q->mq_ops->map_queue(q, ctx->cpu);
 
        /*
@@ -1107,15 +1098,14 @@ static void blk_mq_insert_requests(struct request_queue *q,
                struct request *rq;
 
                rq = list_first_entry(list, struct request, queuelist);
+               BUG_ON(rq->mq_ctx != ctx);
                list_del_init(&rq->queuelist);
-               rq->mq_ctx = ctx;
-               __blk_mq_insert_req_list(hctx, ctx, rq, false);
+               __blk_mq_insert_req_list(hctx, rq, false);
        }
        blk_mq_hctx_mark_pending(hctx, ctx);
        spin_unlock(&ctx->lock);
 
        blk_mq_run_hw_queue(hctx, from_schedule);
-       blk_mq_put_ctx(current_ctx);
 }
 
 static int plug_ctx_cmp(void *priv, struct list_head *a, struct list_head *b)
@@ -1630,16 +1620,17 @@ static int blk_mq_alloc_bitmap(struct blk_mq_ctxmap *bitmap, int node)
        return 0;
 }
 
+/*
+ * 'cpu' is going away. splice any existing rq_list entries from this
+ * software queue to the hw queue dispatch list, and ensure that it
+ * gets run.
+ */
 static int blk_mq_hctx_cpu_offline(struct blk_mq_hw_ctx *hctx, int cpu)
 {
-       struct request_queue *q = hctx->queue;
        struct blk_mq_ctx *ctx;
        LIST_HEAD(tmp);
 
-       /*
-        * Move ctx entries to new CPU, if this one is going away.
-        */
-       ctx = __blk_mq_get_ctx(q, cpu);
+       ctx = __blk_mq_get_ctx(hctx->queue, cpu);
 
        spin_lock(&ctx->lock);
        if (!list_empty(&ctx->rq_list)) {
@@ -1651,24 +1642,11 @@ static int blk_mq_hctx_cpu_offline(struct blk_mq_hw_ctx *hctx, int cpu)
        if (list_empty(&tmp))
                return NOTIFY_OK;
 
-       ctx = blk_mq_get_ctx(q);
-       spin_lock(&ctx->lock);
-
-       while (!list_empty(&tmp)) {
-               struct request *rq;
-
-               rq = list_first_entry(&tmp, struct request, queuelist);
-               rq->mq_ctx = ctx;
-               list_move_tail(&rq->queuelist, &ctx->rq_list);
-       }
-
-       hctx = q->mq_ops->map_queue(q, ctx->cpu);
-       blk_mq_hctx_mark_pending(hctx, ctx);
-
-       spin_unlock(&ctx->lock);
+       spin_lock(&hctx->lock);
+       list_splice_tail_init(&tmp, &hctx->dispatch);
+       spin_unlock(&hctx->lock);
 
        blk_mq_run_hw_queue(hctx, true);
-       blk_mq_put_ctx(ctx);
        return NOTIFY_OK;
 }
 
index 7096c22..f7d973a 100644 (file)
@@ -366,7 +366,7 @@ void elv_dispatch_sort(struct request_queue *q, struct request *rq)
        list_for_each_prev(entry, &q->queue_head) {
                struct request *pos = list_entry_rq(entry);
 
-               if ((req_op(rq) == REQ_OP_DISCARD) != (req_op(pos) == REQ_OP_DISCARD))
+               if (req_op(rq) != req_op(pos))
                        break;
                if (rq_data_dir(rq) != rq_data_dir(pos))
                        break;
index b71a9c7..e3d8e4c 100644 (file)
@@ -3706,22 +3706,21 @@ static int floppy_open(struct block_device *bdev, fmode_t mode)
        if (UFDCS->rawcmd == 1)
                UFDCS->rawcmd = 2;
 
-       if (mode & (FMODE_READ|FMODE_WRITE)) {
-               UDRS->last_checked = 0;
-               clear_bit(FD_OPEN_SHOULD_FAIL_BIT, &UDRS->flags);
-               check_disk_change(bdev);
-               if (test_bit(FD_DISK_CHANGED_BIT, &UDRS->flags))
-                       goto out;
-               if (test_bit(FD_OPEN_SHOULD_FAIL_BIT, &UDRS->flags))
+       if (!(mode & FMODE_NDELAY)) {
+               if (mode & (FMODE_READ|FMODE_WRITE)) {
+                       UDRS->last_checked = 0;
+                       clear_bit(FD_OPEN_SHOULD_FAIL_BIT, &UDRS->flags);
+                       check_disk_change(bdev);
+                       if (test_bit(FD_DISK_CHANGED_BIT, &UDRS->flags))
+                               goto out;
+                       if (test_bit(FD_OPEN_SHOULD_FAIL_BIT, &UDRS->flags))
+                               goto out;
+               }
+               res = -EROFS;
+               if ((mode & FMODE_WRITE) &&
+                   !test_bit(FD_DISK_WRITABLE_BIT, &UDRS->flags))
                        goto out;
        }
-
-       res = -EROFS;
-
-       if ((mode & FMODE_WRITE) &&
-                       !test_bit(FD_DISK_WRITABLE_BIT, &UDRS->flags))
-               goto out;
-
        mutex_unlock(&open_lock);
        mutex_unlock(&floppy_mutex);
        return 0;
index be4fea6..88ef6d4 100644 (file)
@@ -189,6 +189,8 @@ struct blkfront_info
        struct mutex mutex;
        struct xenbus_device *xbdev;
        struct gendisk *gd;
+       u16 sector_size;
+       unsigned int physical_sector_size;
        int vdevice;
        blkif_vdev_t handle;
        enum blkif_state connected;
@@ -910,9 +912,45 @@ static struct blk_mq_ops blkfront_mq_ops = {
        .map_queue = blk_mq_map_queue,
 };
 
+static void blkif_set_queue_limits(struct blkfront_info *info)
+{
+       struct request_queue *rq = info->rq;
+       struct gendisk *gd = info->gd;
+       unsigned int segments = info->max_indirect_segments ? :
+                               BLKIF_MAX_SEGMENTS_PER_REQUEST;
+
+       queue_flag_set_unlocked(QUEUE_FLAG_VIRT, rq);
+
+       if (info->feature_discard) {
+               queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, rq);
+               blk_queue_max_discard_sectors(rq, get_capacity(gd));
+               rq->limits.discard_granularity = info->discard_granularity;
+               rq->limits.discard_alignment = info->discard_alignment;
+               if (info->feature_secdiscard)
+                       queue_flag_set_unlocked(QUEUE_FLAG_SECERASE, rq);
+       }
+
+       /* Hard sector size and max sectors impersonate the equiv. hardware. */
+       blk_queue_logical_block_size(rq, info->sector_size);
+       blk_queue_physical_block_size(rq, info->physical_sector_size);
+       blk_queue_max_hw_sectors(rq, (segments * XEN_PAGE_SIZE) / 512);
+
+       /* Each segment in a request is up to an aligned page in size. */
+       blk_queue_segment_boundary(rq, PAGE_SIZE - 1);
+       blk_queue_max_segment_size(rq, PAGE_SIZE);
+
+       /* Ensure a merged request will fit in a single I/O ring slot. */
+       blk_queue_max_segments(rq, segments / GRANTS_PER_PSEG);
+
+       /* Make sure buffer addresses are sector-aligned. */
+       blk_queue_dma_alignment(rq, 511);
+
+       /* Make sure we don't use bounce buffers. */
+       blk_queue_bounce_limit(rq, BLK_BOUNCE_ANY);
+}
+
 static int xlvbd_init_blk_queue(struct gendisk *gd, u16 sector_size,
-                               unsigned int physical_sector_size,
-                               unsigned int segments)
+                               unsigned int physical_sector_size)
 {
        struct request_queue *rq;
        struct blkfront_info *info = gd->private_data;
@@ -944,36 +982,11 @@ static int xlvbd_init_blk_queue(struct gendisk *gd, u16 sector_size,
        }
 
        rq->queuedata = info;
-       queue_flag_set_unlocked(QUEUE_FLAG_VIRT, rq);
-
-       if (info->feature_discard) {
-               queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, rq);
-               blk_queue_max_discard_sectors(rq, get_capacity(gd));
-               rq->limits.discard_granularity = info->discard_granularity;
-               rq->limits.discard_alignment = info->discard_alignment;
-               if (info->feature_secdiscard)
-                       queue_flag_set_unlocked(QUEUE_FLAG_SECERASE, rq);
-       }
-
-       /* Hard sector size and max sectors impersonate the equiv. hardware. */
-       blk_queue_logical_block_size(rq, sector_size);
-       blk_queue_physical_block_size(rq, physical_sector_size);
-       blk_queue_max_hw_sectors(rq, (segments * XEN_PAGE_SIZE) / 512);
-
-       /* Each segment in a request is up to an aligned page in size. */
-       blk_queue_segment_boundary(rq, PAGE_SIZE - 1);
-       blk_queue_max_segment_size(rq, PAGE_SIZE);
-
-       /* Ensure a merged request will fit in a single I/O ring slot. */
-       blk_queue_max_segments(rq, segments / GRANTS_PER_PSEG);
-
-       /* Make sure buffer addresses are sector-aligned. */
-       blk_queue_dma_alignment(rq, 511);
-
-       /* Make sure we don't use bounce buffers. */
-       blk_queue_bounce_limit(rq, BLK_BOUNCE_ANY);
-
-       gd->queue = rq;
+       info->rq = gd->queue = rq;
+       info->gd = gd;
+       info->sector_size = sector_size;
+       info->physical_sector_size = physical_sector_size;
+       blkif_set_queue_limits(info);
 
        return 0;
 }
@@ -1136,16 +1149,11 @@ static int xlvbd_alloc_gendisk(blkif_sector_t capacity,
        gd->private_data = info;
        set_capacity(gd, capacity);
 
-       if (xlvbd_init_blk_queue(gd, sector_size, physical_sector_size,
-                                info->max_indirect_segments ? :
-                                BLKIF_MAX_SEGMENTS_PER_REQUEST)) {
+       if (xlvbd_init_blk_queue(gd, sector_size, physical_sector_size)) {
                del_gendisk(gd);
                goto release;
        }
 
-       info->rq = gd->queue;
-       info->gd = gd;
-
        xlvbd_flush(info);
 
        if (vdisk_info & VDISK_READONLY)
@@ -1315,7 +1323,7 @@ free_shadow:
                        rinfo->ring_ref[i] = GRANT_INVALID_REF;
                }
        }
-       free_pages((unsigned long)rinfo->ring.sring, get_order(info->nr_ring_pages * PAGE_SIZE));
+       free_pages((unsigned long)rinfo->ring.sring, get_order(info->nr_ring_pages * XEN_PAGE_SIZE));
        rinfo->ring.sring = NULL;
 
        if (rinfo->irq)
@@ -2007,8 +2015,10 @@ static int blkif_recover(struct blkfront_info *info)
        struct split_bio *split_bio;
 
        blkfront_gather_backend_features(info);
+       /* Reset limits changed by blk_mq_update_nr_hw_queues(). */
+       blkif_set_queue_limits(info);
        segs = info->max_indirect_segments ? : BLKIF_MAX_SEGMENTS_PER_REQUEST;
-       blk_queue_max_segments(info->rq, segs);
+       blk_queue_max_segments(info->rq, segs / GRANTS_PER_PSEG);
 
        for (r_index = 0; r_index < info->nr_rings; r_index++) {
                struct blkfront_ring_info *rinfo = &info->rinfo[r_index];
@@ -2432,7 +2442,7 @@ static void blkfront_connect(struct blkfront_info *info)
        if (err) {
                xenbus_dev_fatal(info->xbdev, err, "xlvbd_add at %s",
                                 info->xbdev->otherend);
-               return;
+               goto fail;
        }
 
        xenbus_switch_state(info->xbdev, XenbusStateConnected);
@@ -2445,6 +2455,11 @@ static void blkfront_connect(struct blkfront_info *info)
        device_add_disk(&info->xbdev->dev, info->gd);
 
        info->is_ready = 1;
+       return;
+
+fail:
+       blkif_free(info, 0);
+       return;
 }
 
 /**
index 7e3fd37..92f6e4d 100644 (file)
@@ -66,10 +66,10 @@ static void kona_timer_disable_and_clear(void __iomem *base)
 
 }
 
-static void
+static int
 kona_timer_get_counter(void __iomem *timer_base, uint32_t *msw, uint32_t *lsw)
 {
-       int loop_limit = 4;
+       int loop_limit = 3;
 
        /*
         * Read 64-bit free running counter
@@ -83,18 +83,19 @@ kona_timer_get_counter(void __iomem *timer_base, uint32_t *msw, uint32_t *lsw)
         *      if new hi-word is equal to previously read hi-word then stop.
         */
 
-       while (--loop_limit) {
+       do {
                *msw = readl(timer_base + KONA_GPTIMER_STCHI_OFFSET);
                *lsw = readl(timer_base + KONA_GPTIMER_STCLO_OFFSET);
                if (*msw == readl(timer_base + KONA_GPTIMER_STCHI_OFFSET))
                        break;
-       }
+       } while (--loop_limit);
        if (!loop_limit) {
                pr_err("bcm_kona_timer: getting counter failed.\n");
                pr_err(" Timer will be impacted\n");
+               return -ETIMEDOUT;
        }
 
-       return;
+       return 0;
 }
 
 static int kona_timer_set_next_event(unsigned long clc,
@@ -112,8 +113,11 @@ static int kona_timer_set_next_event(unsigned long clc,
 
        uint32_t lsw, msw;
        uint32_t reg;
+       int ret;
 
-       kona_timer_get_counter(timers.tmr_regs, &msw, &lsw);
+       ret = kona_timer_get_counter(timers.tmr_regs, &msw, &lsw);
+       if (ret)
+               return ret;
 
        /* Load the "next" event tick value */
        writel(lsw + clc, timers.tmr_regs + KONA_GPTIMER_STCM0_OFFSET);
index d91e872..b4b3ab5 100644 (file)
@@ -164,7 +164,7 @@ void __init gic_clocksource_init(unsigned int frequency)
        gic_start_count();
 }
 
-static void __init gic_clocksource_of_init(struct device_node *node)
+static int __init gic_clocksource_of_init(struct device_node *node)
 {
        struct clk *clk;
        int ret;
index 719b478..3c39e6f 100644 (file)
@@ -338,7 +338,6 @@ static int __init armada_xp_timer_init(struct device_node *np)
        struct clk *clk = of_clk_get_by_name(np, "fixed");
        int ret;
 
-       clk = of_clk_get(np, 0);
        if (IS_ERR(clk)) {
                pr_err("Failed to get clock");
                return PTR_ERR(clk);
index dfb1685..1f01e98 100644 (file)
@@ -116,6 +116,9 @@ static int dax_pmem_probe(struct device *dev)
        if (rc)
                return rc;
 
+       /* adjust the dax_region resource to the start of data */
+       res.start += le64_to_cpu(pfn_sb->dataoff);
+
        nd_region = to_nd_region(dev->parent);
        dax_region = alloc_dax_region(dev, nd_region->id, &res,
                        le32_to_cpu(pfn_sb->align), addr, PFN_DEV|PFN_MAP);
index d0c1dab..dff1a4a 100644 (file)
@@ -251,6 +251,14 @@ config EDAC_SBRIDGE
          Support for error detection and correction the Intel
          Sandy Bridge, Ivy Bridge and Haswell Integrated Memory Controllers.
 
+config EDAC_SKX
+       tristate "Intel Skylake server Integrated MC"
+       depends on EDAC_MM_EDAC && PCI && X86_64 && X86_MCE_INTEL
+       depends on PCI_MMCONFIG
+       help
+         Support for error detection and correction the Intel
+         Skylake server Integrated Memory Controllers.
+
 config EDAC_MPC85XX
        tristate "Freescale MPC83xx / MPC85xx"
        depends on EDAC_MM_EDAC && FSL_SOC
index f9e4a3e..9860499 100644 (file)
@@ -31,6 +31,7 @@ obj-$(CONFIG_EDAC_I5400)              += i5400_edac.o
 obj-$(CONFIG_EDAC_I7300)               += i7300_edac.o
 obj-$(CONFIG_EDAC_I7CORE)              += i7core_edac.o
 obj-$(CONFIG_EDAC_SBRIDGE)             += sb_edac.o
+obj-$(CONFIG_EDAC_SKX)                 += skx_edac.o
 obj-$(CONFIG_EDAC_E7XXX)               += e7xxx_edac.o
 obj-$(CONFIG_EDAC_E752X)               += e752x_edac.o
 obj-$(CONFIG_EDAC_I82443BXGX)          += i82443bxgx_edac.o
diff --git a/drivers/edac/skx_edac.c b/drivers/edac/skx_edac.c
new file mode 100644 (file)
index 0000000..0ff4878
--- /dev/null
@@ -0,0 +1,1121 @@
+/*
+ * EDAC driver for Intel(R) Xeon(R) Skylake processors
+ * Copyright (c) 2016, Intel Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ */
+
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/pci.h>
+#include <linux/pci_ids.h>
+#include <linux/slab.h>
+#include <linux/delay.h>
+#include <linux/edac.h>
+#include <linux/mmzone.h>
+#include <linux/smp.h>
+#include <linux/bitmap.h>
+#include <linux/math64.h>
+#include <linux/mod_devicetable.h>
+#include <asm/cpu_device_id.h>
+#include <asm/processor.h>
+#include <asm/mce.h>
+
+#include "edac_core.h"
+
+#define SKX_REVISION    " Ver: 1.0 "
+
+/*
+ * Debug macros
+ */
+#define skx_printk(level, fmt, arg...)                 \
+       edac_printk(level, "skx", fmt, ##arg)
+
+#define skx_mc_printk(mci, level, fmt, arg...)         \
+       edac_mc_chipset_printk(mci, level, "skx", fmt, ##arg)
+
+/*
+ * Get a bit field at register value <v>, from bit <lo> to bit <hi>
+ */
+#define GET_BITFIELD(v, lo, hi) \
+       (((v) & GENMASK_ULL((hi), (lo))) >> (lo))
+
+static LIST_HEAD(skx_edac_list);
+
+static u64 skx_tolm, skx_tohm;
+
+#define NUM_IMC                        2       /* memory controllers per socket */
+#define NUM_CHANNELS           3       /* channels per memory controller */
+#define NUM_DIMMS              2       /* Max DIMMS per channel */
+
+#define        MASK26  0x3FFFFFF               /* Mask for 2^26 */
+#define MASK29 0x1FFFFFFF              /* Mask for 2^29 */
+
+/*
+ * Each cpu socket contains some pci devices that provide global
+ * information, and also some that are local to each of the two
+ * memory controllers on the die.
+ */
+struct skx_dev {
+       struct list_head        list;
+       u8                      bus[4];
+       struct pci_dev  *sad_all;
+       struct pci_dev  *util_all;
+       u32     mcroute;
+       struct skx_imc {
+               struct mem_ctl_info *mci;
+               u8      mc;     /* system wide mc# */
+               u8      lmc;    /* socket relative mc# */
+               u8      src_id, node_id;
+               struct skx_channel {
+                       struct pci_dev *cdev;
+                       struct skx_dimm {
+                               u8      close_pg;
+                               u8      bank_xor_enable;
+                               u8      fine_grain_bank;
+                               u8      rowbits;
+                               u8      colbits;
+                       } dimms[NUM_DIMMS];
+               } chan[NUM_CHANNELS];
+       } imc[NUM_IMC];
+};
+static int skx_num_sockets;
+
+struct skx_pvt {
+       struct skx_imc  *imc;
+};
+
+struct decoded_addr {
+       struct skx_dev *dev;
+       u64     addr;
+       int     socket;
+       int     imc;
+       int     channel;
+       u64     chan_addr;
+       int     sktways;
+       int     chanways;
+       int     dimm;
+       int     rank;
+       int     channel_rank;
+       u64     rank_address;
+       int     row;
+       int     column;
+       int     bank_address;
+       int     bank_group;
+};
+
+static struct skx_dev *get_skx_dev(u8 bus, u8 idx)
+{
+       struct skx_dev *d;
+
+       list_for_each_entry(d, &skx_edac_list, list) {
+               if (d->bus[idx] == bus)
+                       return d;
+       }
+
+       return NULL;
+}
+
+enum munittype {
+       CHAN0, CHAN1, CHAN2, SAD_ALL, UTIL_ALL, SAD
+};
+
+struct munit {
+       u16     did;
+       u16     devfn[NUM_IMC];
+       u8      busidx;
+       u8      per_socket;
+       enum munittype mtype;
+};
+
+/*
+ * List of PCI device ids that we need together with some device
+ * number and function numbers to tell which memory controller the
+ * device belongs to.
+ */
+static const struct munit skx_all_munits[] = {
+       { 0x2054, { }, 1, 1, SAD_ALL },
+       { 0x2055, { }, 1, 1, UTIL_ALL },
+       { 0x2040, { PCI_DEVFN(10, 0), PCI_DEVFN(12, 0) }, 2, 2, CHAN0 },
+       { 0x2044, { PCI_DEVFN(10, 4), PCI_DEVFN(12, 4) }, 2, 2, CHAN1 },
+       { 0x2048, { PCI_DEVFN(11, 0), PCI_DEVFN(13, 0) }, 2, 2, CHAN2 },
+       { 0x208e, { }, 1, 0, SAD },
+       { }
+};
+
+/*
+ * We use the per-socket device 0x2016 to count how many sockets are present,
+ * and to detemine which PCI buses are associated with each socket. Allocate
+ * and build the full list of all the skx_dev structures that we need here.
+ */
+static int get_all_bus_mappings(void)
+{
+       struct pci_dev *pdev, *prev;
+       struct skx_dev *d;
+       u32 reg;
+       int ndev = 0;
+
+       prev = NULL;
+       for (;;) {
+               pdev = pci_get_device(PCI_VENDOR_ID_INTEL, 0x2016, prev);
+               if (!pdev)
+                       break;
+               ndev++;
+               d = kzalloc(sizeof(*d), GFP_KERNEL);
+               if (!d) {
+                       pci_dev_put(pdev);
+                       return -ENOMEM;
+               }
+               pci_read_config_dword(pdev, 0xCC, &reg);
+               d->bus[0] =  GET_BITFIELD(reg, 0, 7);
+               d->bus[1] =  GET_BITFIELD(reg, 8, 15);
+               d->bus[2] =  GET_BITFIELD(reg, 16, 23);
+               d->bus[3] =  GET_BITFIELD(reg, 24, 31);
+               edac_dbg(2, "busses: %x, %x, %x, %x\n",
+                        d->bus[0], d->bus[1], d->bus[2], d->bus[3]);
+               list_add_tail(&d->list, &skx_edac_list);
+               skx_num_sockets++;
+               prev = pdev;
+       }
+
+       return ndev;
+}
+
+static int get_all_munits(const struct munit *m)
+{
+       struct pci_dev *pdev, *prev;
+       struct skx_dev *d;
+       u32 reg;
+       int i = 0, ndev = 0;
+
+       prev = NULL;
+       for (;;) {
+               pdev = pci_get_device(PCI_VENDOR_ID_INTEL, m->did, prev);
+               if (!pdev)
+                       break;
+               ndev++;
+               if (m->per_socket == NUM_IMC) {
+                       for (i = 0; i < NUM_IMC; i++)
+                               if (m->devfn[i] == pdev->devfn)
+                                       break;
+                       if (i == NUM_IMC)
+                               goto fail;
+               }
+               d = get_skx_dev(pdev->bus->number, m->busidx);
+               if (!d)
+                       goto fail;
+
+               /* Be sure that the device is enabled */
+               if (unlikely(pci_enable_device(pdev) < 0)) {
+                       skx_printk(KERN_ERR,
+                               "Couldn't enable %04x:%04x\n", PCI_VENDOR_ID_INTEL, m->did);
+                       goto fail;
+               }
+
+               switch (m->mtype) {
+               case CHAN0: case CHAN1: case CHAN2:
+                       pci_dev_get(pdev);
+                       d->imc[i].chan[m->mtype].cdev = pdev;
+                       break;
+               case SAD_ALL:
+                       pci_dev_get(pdev);
+                       d->sad_all = pdev;
+                       break;
+               case UTIL_ALL:
+                       pci_dev_get(pdev);
+                       d->util_all = pdev;
+                       break;
+               case SAD:
+                       /*
+                        * one of these devices per core, including cores
+                        * that don't exist on this SKU. Ignore any that
+                        * read a route table of zero, make sure all the
+                        * non-zero values match.
+                        */
+                       pci_read_config_dword(pdev, 0xB4, &reg);
+                       if (reg != 0) {
+                               if (d->mcroute == 0)
+                                       d->mcroute = reg;
+                               else if (d->mcroute != reg) {
+                                       skx_printk(KERN_ERR,
+                                               "mcroute mismatch\n");
+                                       goto fail;
+                               }
+                       }
+                       ndev--;
+                       break;
+               }
+
+               prev = pdev;
+       }
+
+       return ndev;
+fail:
+       pci_dev_put(pdev);
+       return -ENODEV;
+}
+
+const struct x86_cpu_id skx_cpuids[] = {
+       { X86_VENDOR_INTEL, 6, 0x55, 0, 0 },    /* Skylake */
+       { }
+};
+MODULE_DEVICE_TABLE(x86cpu, skx_cpuids);
+
+static u8 get_src_id(struct skx_dev *d)
+{
+       u32 reg;
+
+       pci_read_config_dword(d->util_all, 0xF0, &reg);
+
+       return GET_BITFIELD(reg, 12, 14);
+}
+
+static u8 skx_get_node_id(struct skx_dev *d)
+{
+       u32 reg;
+
+       pci_read_config_dword(d->util_all, 0xF4, &reg);
+
+       return GET_BITFIELD(reg, 0, 2);
+}
+
+static int get_dimm_attr(u32 reg, int lobit, int hibit, int add, int minval,
+                        int maxval, char *name)
+{
+       u32 val = GET_BITFIELD(reg, lobit, hibit);
+
+       if (val < minval || val > maxval) {
+               edac_dbg(2, "bad %s = %d (raw=%x)\n", name, val, reg);
+               return -EINVAL;
+       }
+       return val + add;
+}
+
+#define IS_DIMM_PRESENT(mtr)           GET_BITFIELD((mtr), 15, 15)
+
+#define numrank(reg) get_dimm_attr((reg), 12, 13, 0, 1, 2, "ranks")
+#define numrow(reg) get_dimm_attr((reg), 2, 4, 12, 1, 6, "rows")
+#define numcol(reg) get_dimm_attr((reg), 0, 1, 10, 0, 2, "cols")
+
+static int get_width(u32 mtr)
+{
+       switch (GET_BITFIELD(mtr, 8, 9)) {
+       case 0:
+               return DEV_X4;
+       case 1:
+               return DEV_X8;
+       case 2:
+               return DEV_X16;
+       }
+       return DEV_UNKNOWN;
+}
+
+static int skx_get_hi_lo(void)
+{
+       struct pci_dev *pdev;
+       u32 reg;
+
+       pdev = pci_get_device(PCI_VENDOR_ID_INTEL, 0x2034, NULL);
+       if (!pdev) {
+               edac_dbg(0, "Can't get tolm/tohm\n");
+               return -ENODEV;
+       }
+
+       pci_read_config_dword(pdev, 0xD0, &reg);
+       skx_tolm = reg;
+       pci_read_config_dword(pdev, 0xD4, &reg);
+       skx_tohm = reg;
+       pci_read_config_dword(pdev, 0xD8, &reg);
+       skx_tohm |= (u64)reg << 32;
+
+       pci_dev_put(pdev);
+       edac_dbg(2, "tolm=%llx tohm=%llx\n", skx_tolm, skx_tohm);
+
+       return 0;
+}
+
+static int get_dimm_info(u32 mtr, u32 amap, struct dimm_info *dimm,
+                        struct skx_imc *imc, int chan, int dimmno)
+{
+       int  banks = 16, ranks, rows, cols, npages;
+       u64 size;
+
+       if (!IS_DIMM_PRESENT(mtr))
+               return 0;
+       ranks = numrank(mtr);
+       rows = numrow(mtr);
+       cols = numcol(mtr);
+
+       /*
+        * Compute size in 8-byte (2^3) words, then shift to MiB (2^20)
+        */
+       size = ((1ull << (rows + cols + ranks)) * banks) >> (20 - 3);
+       npages = MiB_TO_PAGES(size);
+
+       edac_dbg(0, "mc#%d: channel %d, dimm %d, %lld Mb (%d pages) bank: %d, rank: %d, row: %#x, col: %#x\n",
+                imc->mc, chan, dimmno, size, npages,
+                banks, ranks, rows, cols);
+
+       imc->chan[chan].dimms[dimmno].close_pg = GET_BITFIELD(mtr, 0, 0);
+       imc->chan[chan].dimms[dimmno].bank_xor_enable = GET_BITFIELD(mtr, 9, 9);
+       imc->chan[chan].dimms[dimmno].fine_grain_bank = GET_BITFIELD(amap, 0, 0);
+       imc->chan[chan].dimms[dimmno].rowbits = rows;
+       imc->chan[chan].dimms[dimmno].colbits = cols;
+
+       dimm->nr_pages = npages;
+       dimm->grain = 32;
+       dimm->dtype = get_width(mtr);
+       dimm->mtype = MEM_DDR4;
+       dimm->edac_mode = EDAC_SECDED; /* likely better than this */
+       snprintf(dimm->label, sizeof(dimm->label), "CPU_SrcID#%u_MC#%u_Chan#%u_DIMM#%u",
+                imc->src_id, imc->lmc, chan, dimmno);
+
+       return 1;
+}
+
+#define SKX_GET_MTMTR(dev, reg) \
+       pci_read_config_dword((dev), 0x87c, &reg)
+
+static bool skx_check_ecc(struct pci_dev *pdev)
+{
+       u32 mtmtr;
+
+       SKX_GET_MTMTR(pdev, mtmtr);
+
+       return !!GET_BITFIELD(mtmtr, 2, 2);
+}
+
+static int skx_get_dimm_config(struct mem_ctl_info *mci)
+{
+       struct skx_pvt *pvt = mci->pvt_info;
+       struct skx_imc *imc = pvt->imc;
+       struct dimm_info *dimm;
+       int i, j;
+       u32 mtr, amap;
+       int ndimms;
+
+       for (i = 0; i < NUM_CHANNELS; i++) {
+               ndimms = 0;
+               pci_read_config_dword(imc->chan[i].cdev, 0x8C, &amap);
+               for (j = 0; j < NUM_DIMMS; j++) {
+                       dimm = EDAC_DIMM_PTR(mci->layers, mci->dimms,
+                                            mci->n_layers, i, j, 0);
+                       pci_read_config_dword(imc->chan[i].cdev,
+                                       0x80 + 4*j, &mtr);
+                       ndimms += get_dimm_info(mtr, amap, dimm, imc, i, j);
+               }
+               if (ndimms && !skx_check_ecc(imc->chan[0].cdev)) {
+                       skx_printk(KERN_ERR, "ECC is disabled on imc %d\n", imc->mc);
+                       return -ENODEV;
+               }
+       }
+
+       return 0;
+}
+
+static void skx_unregister_mci(struct skx_imc *imc)
+{
+       struct mem_ctl_info *mci = imc->mci;
+
+       if (!mci)
+               return;
+
+       edac_dbg(0, "MC%d: mci = %p\n", imc->mc, mci);
+
+       /* Remove MC sysfs nodes */
+       edac_mc_del_mc(mci->pdev);
+
+       edac_dbg(1, "%s: free mci struct\n", mci->ctl_name);
+       kfree(mci->ctl_name);
+       edac_mc_free(mci);
+}
+
+static int skx_register_mci(struct skx_imc *imc)
+{
+       struct mem_ctl_info *mci;
+       struct edac_mc_layer layers[2];
+       struct pci_dev *pdev = imc->chan[0].cdev;
+       struct skx_pvt *pvt;
+       int rc;
+
+       /* allocate a new MC control structure */
+       layers[0].type = EDAC_MC_LAYER_CHANNEL;
+       layers[0].size = NUM_CHANNELS;
+       layers[0].is_virt_csrow = false;
+       layers[1].type = EDAC_MC_LAYER_SLOT;
+       layers[1].size = NUM_DIMMS;
+       layers[1].is_virt_csrow = true;
+       mci = edac_mc_alloc(imc->mc, ARRAY_SIZE(layers), layers,
+                           sizeof(struct skx_pvt));
+
+       if (unlikely(!mci))
+               return -ENOMEM;
+
+       edac_dbg(0, "MC#%d: mci = %p\n", imc->mc, mci);
+
+       /* Associate skx_dev and mci for future usage */
+       imc->mci = mci;
+       pvt = mci->pvt_info;
+       pvt->imc = imc;
+
+       mci->ctl_name = kasprintf(GFP_KERNEL, "Skylake Socket#%d IMC#%d",
+                                 imc->node_id, imc->lmc);
+       mci->mtype_cap = MEM_FLAG_DDR4;
+       mci->edac_ctl_cap = EDAC_FLAG_NONE;
+       mci->edac_cap = EDAC_FLAG_NONE;
+       mci->mod_name = "skx_edac.c";
+       mci->dev_name = pci_name(imc->chan[0].cdev);
+       mci->mod_ver = SKX_REVISION;
+       mci->ctl_page_to_phys = NULL;
+
+       rc = skx_get_dimm_config(mci);
+       if (rc < 0)
+               goto fail;
+
+       /* record ptr to the generic device */
+       mci->pdev = &pdev->dev;
+
+       /* add this new MC control structure to EDAC's list of MCs */
+       if (unlikely(edac_mc_add_mc(mci))) {
+               edac_dbg(0, "MC: failed edac_mc_add_mc()\n");
+               rc = -EINVAL;
+               goto fail;
+       }
+
+       return 0;
+
+fail:
+       kfree(mci->ctl_name);
+       edac_mc_free(mci);
+       imc->mci = NULL;
+       return rc;
+}
+
+#define        SKX_MAX_SAD 24
+
+#define SKX_GET_SAD(d, i, reg) \
+       pci_read_config_dword((d)->sad_all, 0x60 + 8 * (i), &reg)
+#define SKX_GET_ILV(d, i, reg) \
+       pci_read_config_dword((d)->sad_all, 0x64 + 8 * (i), &reg)
+
+#define        SKX_SAD_MOD3MODE(sad)   GET_BITFIELD((sad), 30, 31)
+#define        SKX_SAD_MOD3(sad)       GET_BITFIELD((sad), 27, 27)
+#define SKX_SAD_LIMIT(sad)     (((u64)GET_BITFIELD((sad), 7, 26) << 26) | MASK26)
+#define        SKX_SAD_MOD3ASMOD2(sad) GET_BITFIELD((sad), 5, 6)
+#define        SKX_SAD_ATTR(sad)       GET_BITFIELD((sad), 3, 4)
+#define        SKX_SAD_INTERLEAVE(sad) GET_BITFIELD((sad), 1, 2)
+#define SKX_SAD_ENABLE(sad)    GET_BITFIELD((sad), 0, 0)
+
+#define SKX_ILV_REMOTE(tgt)    (((tgt) & 8) == 0)
+#define SKX_ILV_TARGET(tgt)    ((tgt) & 7)
+
+static bool skx_sad_decode(struct decoded_addr *res)
+{
+       struct skx_dev *d = list_first_entry(&skx_edac_list, typeof(*d), list);
+       u64 addr = res->addr;
+       int i, idx, tgt, lchan, shift;
+       u32 sad, ilv;
+       u64 limit, prev_limit;
+       int remote = 0;
+
+       /* Simple sanity check for I/O space or out of range */
+       if (addr >= skx_tohm || (addr >= skx_tolm && addr < BIT_ULL(32))) {
+               edac_dbg(0, "Address %llx out of range\n", addr);
+               return false;
+       }
+
+restart:
+       prev_limit = 0;
+       for (i = 0; i < SKX_MAX_SAD; i++) {
+               SKX_GET_SAD(d, i, sad);
+               limit = SKX_SAD_LIMIT(sad);
+               if (SKX_SAD_ENABLE(sad)) {
+                       if (addr >= prev_limit && addr <= limit)
+                               goto sad_found;
+               }
+               prev_limit = limit + 1;
+       }
+       edac_dbg(0, "No SAD entry for %llx\n", addr);
+       return false;
+
+sad_found:
+       SKX_GET_ILV(d, i, ilv);
+
+       switch (SKX_SAD_INTERLEAVE(sad)) {
+       case 0:
+               idx = GET_BITFIELD(addr, 6, 8);
+               break;
+       case 1:
+               idx = GET_BITFIELD(addr, 8, 10);
+               break;
+       case 2:
+               idx = GET_BITFIELD(addr, 12, 14);
+               break;
+       case 3:
+               idx = GET_BITFIELD(addr, 30, 32);
+               break;
+       }
+
+       tgt = GET_BITFIELD(ilv, 4 * idx, 4 * idx + 3);
+
+       /* If point to another node, find it and start over */
+       if (SKX_ILV_REMOTE(tgt)) {
+               if (remote) {
+                       edac_dbg(0, "Double remote!\n");
+                       return false;
+               }
+               remote = 1;
+               list_for_each_entry(d, &skx_edac_list, list) {
+                       if (d->imc[0].src_id == SKX_ILV_TARGET(tgt))
+                               goto restart;
+               }
+               edac_dbg(0, "Can't find node %d\n", SKX_ILV_TARGET(tgt));
+               return false;
+       }
+
+       if (SKX_SAD_MOD3(sad) == 0)
+               lchan = SKX_ILV_TARGET(tgt);
+       else {
+               switch (SKX_SAD_MOD3MODE(sad)) {
+               case 0:
+                       shift = 6;
+                       break;
+               case 1:
+                       shift = 8;
+                       break;
+               case 2:
+                       shift = 12;
+                       break;
+               default:
+                       edac_dbg(0, "illegal mod3mode\n");
+                       return false;
+               }
+               switch (SKX_SAD_MOD3ASMOD2(sad)) {
+               case 0:
+                       lchan = (addr >> shift) % 3;
+                       break;
+               case 1:
+                       lchan = (addr >> shift) % 2;
+                       break;
+               case 2:
+                       lchan = (addr >> shift) % 2;
+                       lchan = (lchan << 1) | ~lchan;
+                       break;
+               case 3:
+                       lchan = ((addr >> shift) % 2) << 1;
+                       break;
+               }
+               lchan = (lchan << 1) | (SKX_ILV_TARGET(tgt) & 1);
+       }
+
+       res->dev = d;
+       res->socket = d->imc[0].src_id;
+       res->imc = GET_BITFIELD(d->mcroute, lchan * 3, lchan * 3 + 2);
+       res->channel = GET_BITFIELD(d->mcroute, lchan * 2 + 18, lchan * 2 + 19);
+
+       edac_dbg(2, "%llx: socket=%d imc=%d channel=%d\n",
+                res->addr, res->socket, res->imc, res->channel);
+       return true;
+}
+
+#define        SKX_MAX_TAD 8
+
+#define SKX_GET_TADBASE(d, mc, i, reg)                 \
+       pci_read_config_dword((d)->imc[mc].chan[0].cdev, 0x850 + 4 * (i), &reg)
+#define SKX_GET_TADWAYNESS(d, mc, i, reg)              \
+       pci_read_config_dword((d)->imc[mc].chan[0].cdev, 0x880 + 4 * (i), &reg)
+#define SKX_GET_TADCHNILVOFFSET(d, mc, ch, i, reg)     \
+       pci_read_config_dword((d)->imc[mc].chan[ch].cdev, 0x90 + 4 * (i), &reg)
+
+#define        SKX_TAD_BASE(b)         ((u64)GET_BITFIELD((b), 12, 31) << 26)
+#define SKX_TAD_SKT_GRAN(b)    GET_BITFIELD((b), 4, 5)
+#define SKX_TAD_CHN_GRAN(b)    GET_BITFIELD((b), 6, 7)
+#define        SKX_TAD_LIMIT(b)        (((u64)GET_BITFIELD((b), 12, 31) << 26) | MASK26)
+#define        SKX_TAD_OFFSET(b)       ((u64)GET_BITFIELD((b), 4, 23) << 26)
+#define        SKX_TAD_SKTWAYS(b)      (1 << GET_BITFIELD((b), 10, 11))
+#define        SKX_TAD_CHNWAYS(b)      (GET_BITFIELD((b), 8, 9) + 1)
+
+/* which bit used for both socket and channel interleave */
+static int skx_granularity[] = { 6, 8, 12, 30 };
+
+static u64 skx_do_interleave(u64 addr, int shift, int ways, u64 lowbits)
+{
+       addr >>= shift;
+       addr /= ways;
+       addr <<= shift;
+
+       return addr | (lowbits & ((1ull << shift) - 1));
+}
+
+static bool skx_tad_decode(struct decoded_addr *res)
+{
+       int i;
+       u32 base, wayness, chnilvoffset;
+       int skt_interleave_bit, chn_interleave_bit;
+       u64 channel_addr;
+
+       for (i = 0; i < SKX_MAX_TAD; i++) {
+               SKX_GET_TADBASE(res->dev, res->imc, i, base);
+               SKX_GET_TADWAYNESS(res->dev, res->imc, i, wayness);
+               if (SKX_TAD_BASE(base) <= res->addr && res->addr <= SKX_TAD_LIMIT(wayness))
+                       goto tad_found;
+       }
+       edac_dbg(0, "No TAD entry for %llx\n", res->addr);
+       return false;
+
+tad_found:
+       res->sktways = SKX_TAD_SKTWAYS(wayness);
+       res->chanways = SKX_TAD_CHNWAYS(wayness);
+       skt_interleave_bit = skx_granularity[SKX_TAD_SKT_GRAN(base)];
+       chn_interleave_bit = skx_granularity[SKX_TAD_CHN_GRAN(base)];
+
+       SKX_GET_TADCHNILVOFFSET(res->dev, res->imc, res->channel, i, chnilvoffset);
+       channel_addr = res->addr - SKX_TAD_OFFSET(chnilvoffset);
+
+       if (res->chanways == 3 && skt_interleave_bit > chn_interleave_bit) {
+               /* Must handle channel first, then socket */
+               channel_addr = skx_do_interleave(channel_addr, chn_interleave_bit,
+                                                res->chanways, channel_addr);
+               channel_addr = skx_do_interleave(channel_addr, skt_interleave_bit,
+                                                res->sktways, channel_addr);
+       } else {
+               /* Handle socket then channel. Preserve low bits from original address */
+               channel_addr = skx_do_interleave(channel_addr, skt_interleave_bit,
+                                                res->sktways, res->addr);
+               channel_addr = skx_do_interleave(channel_addr, chn_interleave_bit,
+                                                res->chanways, res->addr);
+       }
+
+       res->chan_addr = channel_addr;
+
+       edac_dbg(2, "%llx: chan_addr=%llx sktways=%d chanways=%d\n",
+                res->addr, res->chan_addr, res->sktways, res->chanways);
+       return true;
+}
+
+#define SKX_MAX_RIR 4
+
+#define SKX_GET_RIRWAYNESS(d, mc, ch, i, reg)          \
+       pci_read_config_dword((d)->imc[mc].chan[ch].cdev,       \
+                             0x108 + 4 * (i), &reg)
+#define SKX_GET_RIRILV(d, mc, ch, idx, i, reg)         \
+       pci_read_config_dword((d)->imc[mc].chan[ch].cdev,       \
+                             0x120 + 16 * idx + 4 * (i), &reg)
+
+#define        SKX_RIR_VALID(b) GET_BITFIELD((b), 31, 31)
+#define        SKX_RIR_LIMIT(b) (((u64)GET_BITFIELD((b), 1, 11) << 29) | MASK29)
+#define        SKX_RIR_WAYS(b) (1 << GET_BITFIELD((b), 28, 29))
+#define        SKX_RIR_CHAN_RANK(b) GET_BITFIELD((b), 16, 19)
+#define        SKX_RIR_OFFSET(b) ((u64)(GET_BITFIELD((b), 2, 15) << 26))
+
+static bool skx_rir_decode(struct decoded_addr *res)
+{
+       int i, idx, chan_rank;
+       int shift;
+       u32 rirway, rirlv;
+       u64 rank_addr, prev_limit = 0, limit;
+
+       if (res->dev->imc[res->imc].chan[res->channel].dimms[0].close_pg)
+               shift = 6;
+       else
+               shift = 13;
+
+       for (i = 0; i < SKX_MAX_RIR; i++) {
+               SKX_GET_RIRWAYNESS(res->dev, res->imc, res->channel, i, rirway);
+               limit = SKX_RIR_LIMIT(rirway);
+               if (SKX_RIR_VALID(rirway)) {
+                       if (prev_limit <= res->chan_addr &&
+                           res->chan_addr <= limit)
+                               goto rir_found;
+               }
+               prev_limit = limit;
+       }
+       edac_dbg(0, "No RIR entry for %llx\n", res->addr);
+       return false;
+
+rir_found:
+       rank_addr = res->chan_addr >> shift;
+       rank_addr /= SKX_RIR_WAYS(rirway);
+       rank_addr <<= shift;
+       rank_addr |= res->chan_addr & GENMASK_ULL(shift - 1, 0);
+
+       res->rank_address = rank_addr;
+       idx = (res->chan_addr >> shift) % SKX_RIR_WAYS(rirway);
+
+       SKX_GET_RIRILV(res->dev, res->imc, res->channel, idx, i, rirlv);
+       res->rank_address = rank_addr - SKX_RIR_OFFSET(rirlv);
+       chan_rank = SKX_RIR_CHAN_RANK(rirlv);
+       res->channel_rank = chan_rank;
+       res->dimm = chan_rank / 4;
+       res->rank = chan_rank % 4;
+
+       edac_dbg(2, "%llx: dimm=%d rank=%d chan_rank=%d rank_addr=%llx\n",
+                res->addr, res->dimm, res->rank,
+                res->channel_rank, res->rank_address);
+       return true;
+}
+
+static u8 skx_close_row[] = {
+       15, 16, 17, 18, 20, 21, 22, 28, 10, 11, 12, 13, 29, 30, 31, 32, 33
+};
+static u8 skx_close_column[] = {
+       3, 4, 5, 14, 19, 23, 24, 25, 26, 27
+};
+static u8 skx_open_row[] = {
+       14, 15, 16, 20, 28, 21, 22, 23, 24, 25, 26, 27, 29, 30, 31, 32, 33
+};
+static u8 skx_open_column[] = {
+       3, 4, 5, 6, 7, 8, 9, 10, 11, 12
+};
+static u8 skx_open_fine_column[] = {
+       3, 4, 5, 7, 8, 9, 10, 11, 12, 13
+};
+
+static int skx_bits(u64 addr, int nbits, u8 *bits)
+{
+       int i, res = 0;
+
+       for (i = 0; i < nbits; i++)
+               res |= ((addr >> bits[i]) & 1) << i;
+       return res;
+}
+
+static int skx_bank_bits(u64 addr, int b0, int b1, int do_xor, int x0, int x1)
+{
+       int ret = GET_BITFIELD(addr, b0, b0) | (GET_BITFIELD(addr, b1, b1) << 1);
+
+       if (do_xor)
+               ret ^= GET_BITFIELD(addr, x0, x0) | (GET_BITFIELD(addr, x1, x1) << 1);
+
+       return ret;
+}
+
+static bool skx_mad_decode(struct decoded_addr *r)
+{
+       struct skx_dimm *dimm = &r->dev->imc[r->imc].chan[r->channel].dimms[r->dimm];
+       int bg0 = dimm->fine_grain_bank ? 6 : 13;
+
+       if (dimm->close_pg) {
+               r->row = skx_bits(r->rank_address, dimm->rowbits, skx_close_row);
+               r->column = skx_bits(r->rank_address, dimm->colbits, skx_close_column);
+               r->column |= 0x400; /* C10 is autoprecharge, always set */
+               r->bank_address = skx_bank_bits(r->rank_address, 8, 9, dimm->bank_xor_enable, 22, 28);
+               r->bank_group = skx_bank_bits(r->rank_address, 6, 7, dimm->bank_xor_enable, 20, 21);
+       } else {
+               r->row = skx_bits(r->rank_address, dimm->rowbits, skx_open_row);
+               if (dimm->fine_grain_bank)
+                       r->column = skx_bits(r->rank_address, dimm->colbits, skx_open_fine_column);
+               else
+                       r->column = skx_bits(r->rank_address, dimm->colbits, skx_open_column);
+               r->bank_address = skx_bank_bits(r->rank_address, 18, 19, dimm->bank_xor_enable, 22, 23);
+               r->bank_group = skx_bank_bits(r->rank_address, bg0, 17, dimm->bank_xor_enable, 20, 21);
+       }
+       r->row &= (1u << dimm->rowbits) - 1;
+
+       edac_dbg(2, "%llx: row=%x col=%x bank_addr=%d bank_group=%d\n",
+                r->addr, r->row, r->column, r->bank_address,
+                r->bank_group);
+       return true;
+}
+
+static bool skx_decode(struct decoded_addr *res)
+{
+
+       return skx_sad_decode(res) && skx_tad_decode(res) &&
+               skx_rir_decode(res) && skx_mad_decode(res);
+}
+
+#ifdef CONFIG_EDAC_DEBUG
+/*
+ * Debug feature. Make /sys/kernel/debug/skx_edac_test/addr.
+ * Write an address to this file to exercise the address decode
+ * logic in this driver.
+ */
+static struct dentry *skx_test;
+static u64 skx_fake_addr;
+
+static int debugfs_u64_set(void *data, u64 val)
+{
+       struct decoded_addr res;
+
+       res.addr = val;
+       skx_decode(&res);
+
+       return 0;
+}
+
+DEFINE_SIMPLE_ATTRIBUTE(fops_u64_wo, NULL, debugfs_u64_set, "%llu\n");
+
+static struct dentry *mydebugfs_create(const char *name, umode_t mode,
+                                      struct dentry *parent, u64 *value)
+{
+       return debugfs_create_file(name, mode, parent, value, &fops_u64_wo);
+}
+
+static void setup_skx_debug(void)
+{
+       skx_test = debugfs_create_dir("skx_edac_test", NULL);
+       mydebugfs_create("addr", S_IWUSR, skx_test, &skx_fake_addr);
+}
+
+static void teardown_skx_debug(void)
+{
+       debugfs_remove_recursive(skx_test);
+}
+#else
+static void setup_skx_debug(void)
+{
+}
+
+static void teardown_skx_debug(void)
+{
+}
+#endif /*CONFIG_EDAC_DEBUG*/
+
+static void skx_mce_output_error(struct mem_ctl_info *mci,
+                                const struct mce *m,
+                                struct decoded_addr *res)
+{
+       enum hw_event_mc_err_type tp_event;
+       char *type, *optype, msg[256];
+       bool ripv = GET_BITFIELD(m->mcgstatus, 0, 0);
+       bool overflow = GET_BITFIELD(m->status, 62, 62);
+       bool uncorrected_error = GET_BITFIELD(m->status, 61, 61);
+       bool recoverable;
+       u32 core_err_cnt = GET_BITFIELD(m->status, 38, 52);
+       u32 mscod = GET_BITFIELD(m->status, 16, 31);
+       u32 errcode = GET_BITFIELD(m->status, 0, 15);
+       u32 optypenum = GET_BITFIELD(m->status, 4, 6);
+
+       recoverable = GET_BITFIELD(m->status, 56, 56);
+
+       if (uncorrected_error) {
+               if (ripv) {
+                       type = "FATAL";
+                       tp_event = HW_EVENT_ERR_FATAL;
+               } else {
+                       type = "NON_FATAL";
+                       tp_event = HW_EVENT_ERR_UNCORRECTED;
+               }
+       } else {
+               type = "CORRECTED";
+               tp_event = HW_EVENT_ERR_CORRECTED;
+       }
+
+       /*
+        * According with Table 15-9 of the Intel Architecture spec vol 3A,
+        * memory errors should fit in this mask:
+        *      000f 0000 1mmm cccc (binary)
+        * where:
+        *      f = Correction Report Filtering Bit. If 1, subsequent errors
+        *          won't be shown
+        *      mmm = error type
+        *      cccc = channel
+        * If the mask doesn't match, report an error to the parsing logic
+        */
+       if (!((errcode & 0xef80) == 0x80)) {
+               optype = "Can't parse: it is not a mem";
+       } else {
+               switch (optypenum) {
+               case 0:
+                       optype = "generic undef request error";
+                       break;
+               case 1:
+                       optype = "memory read error";
+                       break;
+               case 2:
+                       optype = "memory write error";
+                       break;
+               case 3:
+                       optype = "addr/cmd error";
+                       break;
+               case 4:
+                       optype = "memory scrubbing error";
+                       break;
+               default:
+                       optype = "reserved";
+                       break;
+               }
+       }
+
+       snprintf(msg, sizeof(msg),
+                "%s%s err_code:%04x:%04x socket:%d imc:%d rank:%d bg:%d ba:%d row:%x col:%x",
+                overflow ? " OVERFLOW" : "",
+                (uncorrected_error && recoverable) ? " recoverable" : "",
+                mscod, errcode,
+                res->socket, res->imc, res->rank,
+                res->bank_group, res->bank_address, res->row, res->column);
+
+       edac_dbg(0, "%s\n", msg);
+
+       /* Call the helper to output message */
+       edac_mc_handle_error(tp_event, mci, core_err_cnt,
+                            m->addr >> PAGE_SHIFT, m->addr & ~PAGE_MASK, 0,
+                            res->channel, res->dimm, -1,
+                            optype, msg);
+}
+
+static int skx_mce_check_error(struct notifier_block *nb, unsigned long val,
+                              void *data)
+{
+       struct mce *mce = (struct mce *)data;
+       struct decoded_addr res;
+       struct mem_ctl_info *mci;
+       char *type;
+
+       if (get_edac_report_status() == EDAC_REPORTING_DISABLED)
+               return NOTIFY_DONE;
+
+       /* ignore unless this is memory related with an address */
+       if ((mce->status & 0xefff) >> 7 != 1 || !(mce->status & MCI_STATUS_ADDRV))
+               return NOTIFY_DONE;
+
+       res.addr = mce->addr;
+       if (!skx_decode(&res))
+               return NOTIFY_DONE;
+       mci = res.dev->imc[res.imc].mci;
+
+       if (mce->mcgstatus & MCG_STATUS_MCIP)
+               type = "Exception";
+       else
+               type = "Event";
+
+       skx_mc_printk(mci, KERN_DEBUG, "HANDLING MCE MEMORY ERROR\n");
+
+       skx_mc_printk(mci, KERN_DEBUG, "CPU %d: Machine Check %s: %Lx "
+                         "Bank %d: %016Lx\n", mce->extcpu, type,
+                         mce->mcgstatus, mce->bank, mce->status);
+       skx_mc_printk(mci, KERN_DEBUG, "TSC %llx ", mce->tsc);
+       skx_mc_printk(mci, KERN_DEBUG, "ADDR %llx ", mce->addr);
+       skx_mc_printk(mci, KERN_DEBUG, "MISC %llx ", mce->misc);
+
+       skx_mc_printk(mci, KERN_DEBUG, "PROCESSOR %u:%x TIME %llu SOCKET "
+                         "%u APIC %x\n", mce->cpuvendor, mce->cpuid,
+                         mce->time, mce->socketid, mce->apicid);
+
+       skx_mce_output_error(mci, mce, &res);
+
+       return NOTIFY_DONE;
+}
+
+static struct notifier_block skx_mce_dec = {
+       .notifier_call = skx_mce_check_error,
+};
+
+static void skx_remove(void)
+{
+       int i, j;
+       struct skx_dev *d, *tmp;
+
+       edac_dbg(0, "\n");
+
+       list_for_each_entry_safe(d, tmp, &skx_edac_list, list) {
+               list_del(&d->list);
+               for (i = 0; i < NUM_IMC; i++) {
+                       skx_unregister_mci(&d->imc[i]);
+                       for (j = 0; j < NUM_CHANNELS; j++)
+                               pci_dev_put(d->imc[i].chan[j].cdev);
+               }
+               pci_dev_put(d->util_all);
+               pci_dev_put(d->sad_all);
+
+               kfree(d);
+       }
+}
+
+/*
+ * skx_init:
+ *     make sure we are running on the correct cpu model
+ *     search for all the devices we need
+ *     check which DIMMs are present.
+ */
+int __init skx_init(void)
+{
+       const struct x86_cpu_id *id;
+       const struct munit *m;
+       int rc = 0, i;
+       u8 mc = 0, src_id, node_id;
+       struct skx_dev *d;
+
+       edac_dbg(2, "\n");
+
+       id = x86_match_cpu(skx_cpuids);
+       if (!id)
+               return -ENODEV;
+
+       rc = skx_get_hi_lo();
+       if (rc)
+               return rc;
+
+       rc = get_all_bus_mappings();
+       if (rc < 0)
+               goto fail;
+       if (rc == 0) {
+               edac_dbg(2, "No memory controllers found\n");
+               return -ENODEV;
+       }
+
+       for (m = skx_all_munits; m->did; m++) {
+               rc = get_all_munits(m);
+               if (rc < 0)
+                       goto fail;
+               if (rc != m->per_socket * skx_num_sockets) {
+                       edac_dbg(2, "Expected %d, got %d of %x\n",
+                                m->per_socket * skx_num_sockets, rc, m->did);
+                       rc = -ENODEV;
+                       goto fail;
+               }
+       }
+
+       list_for_each_entry(d, &skx_edac_list, list) {
+               src_id = get_src_id(d);
+               node_id = skx_get_node_id(d);
+               edac_dbg(2, "src_id=%d node_id=%d\n", src_id, node_id);
+               for (i = 0; i < NUM_IMC; i++) {
+                       d->imc[i].mc = mc++;
+                       d->imc[i].lmc = i;
+                       d->imc[i].src_id = src_id;
+                       d->imc[i].node_id = node_id;
+                       rc = skx_register_mci(&d->imc[i]);
+                       if (rc < 0)
+                               goto fail;
+               }
+       }
+
+       /* Ensure that the OPSTATE is set correctly for POLL or NMI */
+       opstate_init();
+
+       setup_skx_debug();
+
+       mce_register_decode_chain(&skx_mce_dec);
+
+       return 0;
+fail:
+       skx_remove();
+       return rc;
+}
+
+static void __exit skx_exit(void)
+{
+       edac_dbg(2, "\n");
+       mce_unregister_decode_chain(&skx_mce_dec);
+       skx_remove();
+       teardown_skx_debug();
+}
+
+module_init(skx_init);
+module_exit(skx_exit);
+
+module_param(edac_op_state, int, 0444);
+MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll,1=NMI");
+
+MODULE_LICENSE("GPL v2");
+MODULE_AUTHOR("Tony Luck");
+MODULE_DESCRIPTION("MC Driver for Intel Skylake server processors");
index 98dd47a..66a9410 100644 (file)
@@ -50,6 +50,7 @@ config GPIO_DEVRES
 config OF_GPIO
        def_bool y
        depends on OF
+       depends on HAS_IOMEM
 
 config GPIO_ACPI
        def_bool y
@@ -188,7 +189,7 @@ config GPIO_EP93XX
 config GPIO_ETRAXFS
        bool "Axis ETRAX FS General I/O"
        depends on CRIS || COMPILE_TEST
-       depends on OF
+       depends on OF_GPIO
        select GPIO_GENERIC
        select GPIOLIB_IRQCHIP
        help
@@ -214,7 +215,7 @@ config GPIO_GENERIC_PLATFORM
 
 config GPIO_GRGPIO
        tristate "Aeroflex Gaisler GRGPIO support"
-       depends on OF
+       depends on OF_GPIO
        select GPIO_GENERIC
        select IRQ_DOMAIN
        help
@@ -312,7 +313,7 @@ config GPIO_MPC8XXX
 config GPIO_MVEBU
        def_bool y
        depends on PLAT_ORION
-       depends on OF
+       depends on OF_GPIO
        select GENERIC_IRQ_CHIP
 
 config GPIO_MXC
@@ -405,7 +406,7 @@ config GPIO_TEGRA
        bool "NVIDIA Tegra GPIO support"
        default ARCH_TEGRA
        depends on ARCH_TEGRA || COMPILE_TEST
-       depends on OF
+       depends on OF_GPIO
        help
          Say yes here to support GPIO pins on NVIDIA Tegra SoCs.
 
@@ -1099,7 +1100,7 @@ menu "SPI GPIO expanders"
 
 config GPIO_74X164
        tristate "74x164 serial-in/parallel-out 8-bits shift register"
-       depends on OF
+       depends on OF_GPIO
        help
          Driver for 74x164 compatible serial-in/parallel-out 8-outputs
          shift registers. This driver can be used to provide access
index 0880736..946d091 100644 (file)
@@ -192,6 +192,10 @@ int __max730x_probe(struct max7301 *ts)
        ts->chip.parent = dev;
        ts->chip.owner = THIS_MODULE;
 
+       ret = gpiochip_add_data(&ts->chip, ts);
+       if (ret)
+               goto exit_destroy;
+
        /*
         * initialize pullups according to platform data and cache the
         * register values for later use.
@@ -213,10 +217,6 @@ int __max730x_probe(struct max7301 *ts)
                }
        }
 
-       ret = gpiochip_add_data(&ts->chip, ts);
-       if (ret)
-               goto exit_destroy;
-
        return ret;
 
 exit_destroy:
index 8ebc5f1..8c704c8 100644 (file)
@@ -646,9 +646,9 @@ int amdgpu_gart_table_vram_pin(struct amdgpu_device *adev);
 void amdgpu_gart_table_vram_unpin(struct amdgpu_device *adev);
 int amdgpu_gart_init(struct amdgpu_device *adev);
 void amdgpu_gart_fini(struct amdgpu_device *adev);
-void amdgpu_gart_unbind(struct amdgpu_device *adev, unsigned offset,
+void amdgpu_gart_unbind(struct amdgpu_device *adev, uint64_t offset,
                        int pages);
-int amdgpu_gart_bind(struct amdgpu_device *adev, unsigned offset,
+int amdgpu_gart_bind(struct amdgpu_device *adev, uint64_t offset,
                     int pages, struct page **pagelist,
                     dma_addr_t *dma_addr, uint32_t flags);
 
index 49de926..10b5ddf 100644 (file)
@@ -200,16 +200,7 @@ static int amdgpu_atpx_validate(struct amdgpu_atpx *atpx)
        atpx->is_hybrid = false;
        if (valid_bits & ATPX_MS_HYBRID_GFX_SUPPORTED) {
                printk("ATPX Hybrid Graphics\n");
-#if 1
-               /* This is a temporary hack until the D3 cold support
-                * makes it upstream.  The ATPX power_control method seems
-                * to still work on even if the system should be using
-                * the new standardized hybrid D3 cold ACPI interface.
-                */
-               atpx->functions.power_cntl = true;
-#else
                atpx->functions.power_cntl = false;
-#endif
                atpx->is_hybrid = true;
        }
 
index 921bce2..0feea34 100644 (file)
@@ -221,7 +221,7 @@ void amdgpu_gart_table_vram_free(struct amdgpu_device *adev)
  * Unbinds the requested pages from the gart page table and
  * replaces them with the dummy page (all asics).
  */
-void amdgpu_gart_unbind(struct amdgpu_device *adev, unsigned offset,
+void amdgpu_gart_unbind(struct amdgpu_device *adev, uint64_t offset,
                        int pages)
 {
        unsigned t;
@@ -268,7 +268,7 @@ void amdgpu_gart_unbind(struct amdgpu_device *adev, unsigned offset,
  * (all asics).
  * Returns 0 for success, -EINVAL for failure.
  */
-int amdgpu_gart_bind(struct amdgpu_device *adev, unsigned offset,
+int amdgpu_gart_bind(struct amdgpu_device *adev, uint64_t offset,
                     int pages, struct page **pagelist, dma_addr_t *dma_addr,
                     uint32_t flags)
 {
index b11f4e8..4aa993d 100644 (file)
@@ -1187,7 +1187,8 @@ int amdgpu_uvd_ring_test_ib(struct amdgpu_ring *ring, long timeout)
                r = 0;
        }
 
-error:
        fence_put(fence);
+
+error:
        return r;
 }
index 8e642fc..80120fa 100644 (file)
@@ -1535,7 +1535,7 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm)
        r = amd_sched_entity_init(&ring->sched, &vm->entity,
                                  rq, amdgpu_sched_jobs);
        if (r)
-               return r;
+               goto err;
 
        vm->page_directory_fence = NULL;
 
@@ -1565,6 +1565,9 @@ error_free_page_directory:
 error_free_sched_entity:
        amd_sched_entity_fini(&ring->sched, &vm->entity);
 
+err:
+       drm_free_large(vm->page_tables);
+
        return r;
 }
 
index e621eba..a7d3cb3 100644 (file)
@@ -184,7 +184,7 @@ u32 __iomem *kfd_get_kernel_doorbell(struct kfd_dev *kfd,
                                                        sizeof(u32)) + inx;
 
        pr_debug("kfd: get kernel queue doorbell\n"
-                        "     doorbell offset   == 0x%08d\n"
+                        "     doorbell offset   == 0x%08X\n"
                         "     kernel address    == 0x%08lX\n",
                *doorbell_off, (uintptr_t)(kfd->doorbell_kernel_ptr + inx));
 
index ce54e98..0a06f91 100644 (file)
@@ -464,7 +464,7 @@ static bool drm_fb_helper_is_bound(struct drm_fb_helper *fb_helper)
 
        /* Sometimes user space wants everything disabled, so don't steal the
         * display if there's a master. */
-       if (lockless_dereference(dev->master))
+       if (READ_ONCE(dev->master))
                return false;
 
        drm_for_each_crtc(crtc, dev) {
index 87ef341..b382cf5 100644 (file)
@@ -1333,8 +1333,6 @@ int etnaviv_gpu_submit(struct etnaviv_gpu *gpu,
        if (ret < 0)
                return ret;
 
-       mutex_lock(&gpu->lock);
-
        /*
         * TODO
         *
@@ -1348,16 +1346,18 @@ int etnaviv_gpu_submit(struct etnaviv_gpu *gpu,
        if (unlikely(event == ~0U)) {
                DRM_ERROR("no free event\n");
                ret = -EBUSY;
-               goto out_unlock;
+               goto out_pm_put;
        }
 
        fence = etnaviv_gpu_fence_alloc(gpu);
        if (!fence) {
                event_free(gpu, event);
                ret = -ENOMEM;
-               goto out_unlock;
+               goto out_pm_put;
        }
 
+       mutex_lock(&gpu->lock);
+
        gpu->event[event].fence = fence;
        submit->fence = fence->seqno;
        gpu->active_fence = submit->fence;
@@ -1395,9 +1395,9 @@ int etnaviv_gpu_submit(struct etnaviv_gpu *gpu,
        hangcheck_timer_reset(gpu);
        ret = 0;
 
-out_unlock:
        mutex_unlock(&gpu->lock);
 
+out_pm_put:
        etnaviv_gpu_pm_put(gpu);
 
        return ret;
index 21f9390..20fe9d5 100644 (file)
@@ -1854,6 +1854,7 @@ struct drm_i915_private {
        enum modeset_restore modeset_restore;
        struct mutex modeset_restore_lock;
        struct drm_atomic_state *modeset_restore_state;
+       struct drm_modeset_acquire_ctx reset_ctx;
 
        struct list_head vm_list; /* Global list of all address spaces */
        struct i915_ggtt ggtt; /* VM representing the global address space */
index 1168150..a77ce99 100644 (file)
@@ -879,9 +879,12 @@ i915_gem_pread_ioctl(struct drm_device *dev, void *data,
        ret = i915_gem_shmem_pread(dev, obj, args, file);
 
        /* pread for non shmem backed objects */
-       if (ret == -EFAULT || ret == -ENODEV)
+       if (ret == -EFAULT || ret == -ENODEV) {
+               intel_runtime_pm_get(to_i915(dev));
                ret = i915_gem_gtt_pread(dev, obj, args->size,
                                        args->offset, args->data_ptr);
+               intel_runtime_pm_put(to_i915(dev));
+       }
 
 out:
        drm_gem_object_unreference(&obj->base);
@@ -1306,7 +1309,7 @@ i915_gem_pwrite_ioctl(struct drm_device *dev, void *data,
                 * textures). Fallback to the shmem path in that case. */
        }
 
-       if (ret == -EFAULT) {
+       if (ret == -EFAULT || ret == -ENOSPC) {
                if (obj->phys_handle)
                        ret = i915_gem_phys_pwrite(obj, args, file);
                else if (i915_gem_object_has_struct_page(obj))
@@ -3169,6 +3172,8 @@ static void i915_gem_reset_engine_cleanup(struct intel_engine_cs *engine)
        }
 
        intel_ring_init_seqno(engine, engine->last_submitted_seqno);
+
+       engine->i915->gt.active_engines &= ~intel_engine_flag(engine);
 }
 
 void i915_gem_reset(struct drm_device *dev)
@@ -3186,6 +3191,7 @@ void i915_gem_reset(struct drm_device *dev)
 
        for_each_engine(engine, dev_priv)
                i915_gem_reset_engine_cleanup(engine);
+       mod_delayed_work(dev_priv->wq, &dev_priv->gt.idle_work, 0);
 
        i915_gem_context_reset(dev);
 
index 10f1e32..7a30af7 100644 (file)
@@ -2873,6 +2873,7 @@ void i915_ggtt_cleanup_hw(struct drm_device *dev)
                struct i915_hw_ppgtt *ppgtt = dev_priv->mm.aliasing_ppgtt;
 
                ppgtt->base.cleanup(&ppgtt->base);
+               kfree(ppgtt);
        }
 
        i915_gem_cleanup_stolen(dev);
index ce14fe0..5c06413 100644 (file)
@@ -1536,6 +1536,7 @@ enum skl_disp_power_wells {
 #define BALANCE_LEG_MASK(port)         (7<<(8+3*(port)))
 /* Balance leg disable bits */
 #define BALANCE_LEG_DISABLE_SHIFT      23
+#define BALANCE_LEG_DISABLE(port)      (1 << (23 + (port)))
 
 /*
  * Fence registers
index 6700a7b..d32f586 100644 (file)
@@ -600,6 +600,8 @@ static void i915_audio_component_codec_wake_override(struct device *dev,
        if (!IS_SKYLAKE(dev_priv) && !IS_KABYLAKE(dev_priv))
                return;
 
+       i915_audio_component_get_power(dev);
+
        /*
         * Enable/disable generating the codec wake signal, overriding the
         * internal logic to generate the codec wake to controller.
@@ -615,6 +617,8 @@ static void i915_audio_component_codec_wake_override(struct device *dev,
                I915_WRITE(HSW_AUD_CHICKENBIT, tmp);
                usleep_range(1000, 1500);
        }
+
+       i915_audio_component_put_power(dev);
 }
 
 /* Get CDCLK in kHz  */
@@ -648,6 +652,7 @@ static int i915_audio_component_sync_audio_rate(struct device *dev,
            !IS_HASWELL(dev_priv))
                return 0;
 
+       i915_audio_component_get_power(dev);
        mutex_lock(&dev_priv->av_mutex);
        /* 1. get the pipe */
        intel_encoder = dev_priv->dig_port_map[port];
@@ -698,6 +703,7 @@ static int i915_audio_component_sync_audio_rate(struct device *dev,
 
  unlock:
        mutex_unlock(&dev_priv->av_mutex);
+       i915_audio_component_put_power(dev);
        return err;
 }
 
index dd1d6fe..1a7efac 100644 (file)
@@ -145,7 +145,7 @@ static const struct ddi_buf_trans skl_ddi_translations_dp[] = {
 static const struct ddi_buf_trans skl_u_ddi_translations_dp[] = {
        { 0x0000201B, 0x000000A2, 0x0 },
        { 0x00005012, 0x00000088, 0x0 },
-       { 0x80007011, 0x000000CD, 0x0 },
+       { 0x80007011, 0x000000CD, 0x1 },
        { 0x80009010, 0x000000C0, 0x1 },
        { 0x0000201B, 0x0000009D, 0x0 },
        { 0x80005012, 0x000000C0, 0x1 },
@@ -158,7 +158,7 @@ static const struct ddi_buf_trans skl_u_ddi_translations_dp[] = {
 static const struct ddi_buf_trans skl_y_ddi_translations_dp[] = {
        { 0x00000018, 0x000000A2, 0x0 },
        { 0x00005012, 0x00000088, 0x0 },
-       { 0x80007011, 0x000000CD, 0x0 },
+       { 0x80007011, 0x000000CD, 0x3 },
        { 0x80009010, 0x000000C0, 0x3 },
        { 0x00000018, 0x0000009D, 0x0 },
        { 0x80005012, 0x000000C0, 0x3 },
@@ -388,6 +388,40 @@ skl_get_buf_trans_hdmi(struct drm_i915_private *dev_priv, int *n_entries)
        }
 }
 
+static int intel_ddi_hdmi_level(struct drm_i915_private *dev_priv, enum port port)
+{
+       int n_hdmi_entries;
+       int hdmi_level;
+       int hdmi_default_entry;
+
+       hdmi_level = dev_priv->vbt.ddi_port_info[port].hdmi_level_shift;
+
+       if (IS_BROXTON(dev_priv))
+               return hdmi_level;
+
+       if (IS_SKYLAKE(dev_priv) || IS_KABYLAKE(dev_priv)) {
+               skl_get_buf_trans_hdmi(dev_priv, &n_hdmi_entries);
+               hdmi_default_entry = 8;
+       } else if (IS_BROADWELL(dev_priv)) {
+               n_hdmi_entries = ARRAY_SIZE(bdw_ddi_translations_hdmi);
+               hdmi_default_entry = 7;
+       } else if (IS_HASWELL(dev_priv)) {
+               n_hdmi_entries = ARRAY_SIZE(hsw_ddi_translations_hdmi);
+               hdmi_default_entry = 6;
+       } else {
+               WARN(1, "ddi translation table missing\n");
+               n_hdmi_entries = ARRAY_SIZE(bdw_ddi_translations_hdmi);
+               hdmi_default_entry = 7;
+       }
+
+       /* Choose a good default if VBT is badly populated */
+       if (hdmi_level == HDMI_LEVEL_SHIFT_UNKNOWN ||
+           hdmi_level >= n_hdmi_entries)
+               hdmi_level = hdmi_default_entry;
+
+       return hdmi_level;
+}
+
 /*
  * Starting with Haswell, DDI port buffers must be programmed with correct
  * values in advance. The buffer values are different for FDI and DP modes,
@@ -399,7 +433,7 @@ void intel_prepare_ddi_buffer(struct intel_encoder *encoder)
 {
        struct drm_i915_private *dev_priv = to_i915(encoder->base.dev);
        u32 iboost_bit = 0;
-       int i, n_hdmi_entries, n_dp_entries, n_edp_entries, hdmi_default_entry,
+       int i, n_hdmi_entries, n_dp_entries, n_edp_entries,
            size;
        int hdmi_level;
        enum port port;
@@ -410,7 +444,7 @@ void intel_prepare_ddi_buffer(struct intel_encoder *encoder)
        const struct ddi_buf_trans *ddi_translations;
 
        port = intel_ddi_get_encoder_port(encoder);
-       hdmi_level = dev_priv->vbt.ddi_port_info[port].hdmi_level_shift;
+       hdmi_level = intel_ddi_hdmi_level(dev_priv, port);
 
        if (IS_BROXTON(dev_priv)) {
                if (encoder->type != INTEL_OUTPUT_HDMI)
@@ -430,7 +464,6 @@ void intel_prepare_ddi_buffer(struct intel_encoder *encoder)
                                skl_get_buf_trans_edp(dev_priv, &n_edp_entries);
                ddi_translations_hdmi =
                                skl_get_buf_trans_hdmi(dev_priv, &n_hdmi_entries);
-               hdmi_default_entry = 8;
                /* If we're boosting the current, set bit 31 of trans1 */
                if (dev_priv->vbt.ddi_port_info[port].hdmi_boost_level ||
                    dev_priv->vbt.ddi_port_info[port].dp_boost_level)
@@ -456,7 +489,6 @@ void intel_prepare_ddi_buffer(struct intel_encoder *encoder)
 
                n_dp_entries = ARRAY_SIZE(bdw_ddi_translations_dp);
                n_hdmi_entries = ARRAY_SIZE(bdw_ddi_translations_hdmi);
-               hdmi_default_entry = 7;
        } else if (IS_HASWELL(dev_priv)) {
                ddi_translations_fdi = hsw_ddi_translations_fdi;
                ddi_translations_dp = hsw_ddi_translations_dp;
@@ -464,7 +496,6 @@ void intel_prepare_ddi_buffer(struct intel_encoder *encoder)
                ddi_translations_hdmi = hsw_ddi_translations_hdmi;
                n_dp_entries = n_edp_entries = ARRAY_SIZE(hsw_ddi_translations_dp);
                n_hdmi_entries = ARRAY_SIZE(hsw_ddi_translations_hdmi);
-               hdmi_default_entry = 6;
        } else {
                WARN(1, "ddi translation table missing\n");
                ddi_translations_edp = bdw_ddi_translations_dp;
@@ -474,7 +505,6 @@ void intel_prepare_ddi_buffer(struct intel_encoder *encoder)
                n_edp_entries = ARRAY_SIZE(bdw_ddi_translations_edp);
                n_dp_entries = ARRAY_SIZE(bdw_ddi_translations_dp);
                n_hdmi_entries = ARRAY_SIZE(bdw_ddi_translations_hdmi);
-               hdmi_default_entry = 7;
        }
 
        switch (encoder->type) {
@@ -505,11 +535,6 @@ void intel_prepare_ddi_buffer(struct intel_encoder *encoder)
        if (encoder->type != INTEL_OUTPUT_HDMI)
                return;
 
-       /* Choose a good default if VBT is badly populated */
-       if (hdmi_level == HDMI_LEVEL_SHIFT_UNKNOWN ||
-           hdmi_level >= n_hdmi_entries)
-               hdmi_level = hdmi_default_entry;
-
        /* Entry 9 is for HDMI: */
        I915_WRITE(DDI_BUF_TRANS_LO(port, i),
                   ddi_translations_hdmi[hdmi_level].trans1 | iboost_bit);
@@ -1379,14 +1404,30 @@ void intel_ddi_disable_pipe_clock(struct intel_crtc *intel_crtc)
                           TRANS_CLK_SEL_DISABLED);
 }
 
-static void skl_ddi_set_iboost(struct drm_i915_private *dev_priv,
-                              u32 level, enum port port, int type)
+static void _skl_ddi_set_iboost(struct drm_i915_private *dev_priv,
+                               enum port port, uint8_t iboost)
 {
+       u32 tmp;
+
+       tmp = I915_READ(DISPIO_CR_TX_BMU_CR0);
+       tmp &= ~(BALANCE_LEG_MASK(port) | BALANCE_LEG_DISABLE(port));
+       if (iboost)
+               tmp |= iboost << BALANCE_LEG_SHIFT(port);
+       else
+               tmp |= BALANCE_LEG_DISABLE(port);
+       I915_WRITE(DISPIO_CR_TX_BMU_CR0, tmp);
+}
+
+static void skl_ddi_set_iboost(struct intel_encoder *encoder, u32 level)
+{
+       struct intel_digital_port *intel_dig_port = enc_to_dig_port(&encoder->base);
+       struct drm_i915_private *dev_priv = to_i915(intel_dig_port->base.base.dev);
+       enum port port = intel_dig_port->port;
+       int type = encoder->type;
        const struct ddi_buf_trans *ddi_translations;
        uint8_t iboost;
        uint8_t dp_iboost, hdmi_iboost;
        int n_entries;
-       u32 reg;
 
        /* VBT may override standard boost values */
        dp_iboost = dev_priv->vbt.ddi_port_info[port].dp_boost_level;
@@ -1428,16 +1469,10 @@ static void skl_ddi_set_iboost(struct drm_i915_private *dev_priv,
                return;
        }
 
-       reg = I915_READ(DISPIO_CR_TX_BMU_CR0);
-       reg &= ~BALANCE_LEG_MASK(port);
-       reg &= ~(1 << (BALANCE_LEG_DISABLE_SHIFT + port));
-
-       if (iboost)
-               reg |= iboost << BALANCE_LEG_SHIFT(port);
-       else
-               reg |= 1 << (BALANCE_LEG_DISABLE_SHIFT + port);
+       _skl_ddi_set_iboost(dev_priv, port, iboost);
 
-       I915_WRITE(DISPIO_CR_TX_BMU_CR0, reg);
+       if (port == PORT_A && intel_dig_port->max_lanes == 4)
+               _skl_ddi_set_iboost(dev_priv, PORT_E, iboost);
 }
 
 static void bxt_ddi_vswing_sequence(struct drm_i915_private *dev_priv,
@@ -1568,7 +1603,7 @@ uint32_t ddi_signal_levels(struct intel_dp *intel_dp)
        level = translate_signal_level(signal_levels);
 
        if (IS_SKYLAKE(dev_priv) || IS_KABYLAKE(dev_priv))
-               skl_ddi_set_iboost(dev_priv, level, port, encoder->type);
+               skl_ddi_set_iboost(encoder, level);
        else if (IS_BROXTON(dev_priv))
                bxt_ddi_vswing_sequence(dev_priv, level, port, encoder->type);
 
@@ -1637,6 +1672,10 @@ static void intel_ddi_pre_enable(struct intel_encoder *intel_encoder)
                        intel_dp_stop_link_train(intel_dp);
        } else if (type == INTEL_OUTPUT_HDMI) {
                struct intel_hdmi *intel_hdmi = enc_to_intel_hdmi(encoder);
+               int level = intel_ddi_hdmi_level(dev_priv, port);
+
+               if (IS_SKYLAKE(dev_priv) || IS_KABYLAKE(dev_priv))
+                       skl_ddi_set_iboost(intel_encoder, level);
 
                intel_hdmi->set_infoframes(encoder,
                                           crtc->config->has_hdmi_sink,
index dcf93b3..2a751b6 100644 (file)
@@ -3093,40 +3093,110 @@ static void intel_update_primary_planes(struct drm_device *dev)
 
        for_each_crtc(dev, crtc) {
                struct intel_plane *plane = to_intel_plane(crtc->primary);
-               struct intel_plane_state *plane_state;
-
-               drm_modeset_lock_crtc(crtc, &plane->base);
-               plane_state = to_intel_plane_state(plane->base.state);
+               struct intel_plane_state *plane_state =
+                       to_intel_plane_state(plane->base.state);
 
                if (plane_state->visible)
                        plane->update_plane(&plane->base,
                                            to_intel_crtc_state(crtc->state),
                                            plane_state);
+       }
+}
+
+static int
+__intel_display_resume(struct drm_device *dev,
+                      struct drm_atomic_state *state)
+{
+       struct drm_crtc_state *crtc_state;
+       struct drm_crtc *crtc;
+       int i, ret;
+
+       intel_modeset_setup_hw_state(dev);
+       i915_redisable_vga(dev);
 
-               drm_modeset_unlock_crtc(crtc);
+       if (!state)
+               return 0;
+
+       for_each_crtc_in_state(state, crtc, crtc_state, i) {
+               /*
+                * Force recalculation even if we restore
+                * current state. With fast modeset this may not result
+                * in a modeset when the state is compatible.
+                */
+               crtc_state->mode_changed = true;
        }
+
+       /* ignore any reset values/BIOS leftovers in the WM registers */
+       to_intel_atomic_state(state)->skip_intermediate_wm = true;
+
+       ret = drm_atomic_commit(state);
+
+       WARN_ON(ret == -EDEADLK);
+       return ret;
 }
 
 void intel_prepare_reset(struct drm_i915_private *dev_priv)
 {
+       struct drm_device *dev = &dev_priv->drm;
+       struct drm_modeset_acquire_ctx *ctx = &dev_priv->reset_ctx;
+       struct drm_atomic_state *state;
+       int ret;
+
        /* no reset support for gen2 */
        if (IS_GEN2(dev_priv))
                return;
 
-       /* reset doesn't touch the display */
+       /*
+        * Need mode_config.mutex so that we don't
+        * trample ongoing ->detect() and whatnot.
+        */
+       mutex_lock(&dev->mode_config.mutex);
+       drm_modeset_acquire_init(ctx, 0);
+       while (1) {
+               ret = drm_modeset_lock_all_ctx(dev, ctx);
+               if (ret != -EDEADLK)
+                       break;
+
+               drm_modeset_backoff(ctx);
+       }
+
+       /* reset doesn't touch the display, but flips might get nuked anyway, */
        if (INTEL_GEN(dev_priv) >= 5 || IS_G4X(dev_priv))
                return;
 
-       drm_modeset_lock_all(&dev_priv->drm);
        /*
         * Disabling the crtcs gracefully seems nicer. Also the
         * g33 docs say we should at least disable all the planes.
         */
-       intel_display_suspend(&dev_priv->drm);
+       state = drm_atomic_helper_duplicate_state(dev, ctx);
+       if (IS_ERR(state)) {
+               ret = PTR_ERR(state);
+               state = NULL;
+               DRM_ERROR("Duplicating state failed with %i\n", ret);
+               goto err;
+       }
+
+       ret = drm_atomic_helper_disable_all(dev, ctx);
+       if (ret) {
+               DRM_ERROR("Suspending crtc's failed with %i\n", ret);
+               goto err;
+       }
+
+       dev_priv->modeset_restore_state = state;
+       state->acquire_ctx = ctx;
+       return;
+
+err:
+       drm_atomic_state_free(state);
 }
 
 void intel_finish_reset(struct drm_i915_private *dev_priv)
 {
+       struct drm_device *dev = &dev_priv->drm;
+       struct drm_modeset_acquire_ctx *ctx = &dev_priv->reset_ctx;
+       struct drm_atomic_state *state = dev_priv->modeset_restore_state;
+       int ret;
+
        /*
         * Flips in the rings will be nuked by the reset,
         * so complete all pending flips so that user space
@@ -3138,6 +3208,8 @@ void intel_finish_reset(struct drm_i915_private *dev_priv)
        if (IS_GEN2(dev_priv))
                return;
 
+       dev_priv->modeset_restore_state = NULL;
+
        /* reset doesn't touch the display */
        if (INTEL_GEN(dev_priv) >= 5 || IS_G4X(dev_priv)) {
                /*
@@ -3149,29 +3221,32 @@ void intel_finish_reset(struct drm_i915_private *dev_priv)
                 * FIXME: Atomic will make this obsolete since we won't schedule
                 * CS-based flips (which might get lost in gpu resets) any more.
                 */
-               intel_update_primary_planes(&dev_priv->drm);
-               return;
-       }
-
-       /*
-        * The display has been reset as well,
-        * so need a full re-initialization.
-        */
-       intel_runtime_pm_disable_interrupts(dev_priv);
-       intel_runtime_pm_enable_interrupts(dev_priv);
+               intel_update_primary_planes(dev);
+       } else {
+               /*
+                * The display has been reset as well,
+                * so need a full re-initialization.
+                */
+               intel_runtime_pm_disable_interrupts(dev_priv);
+               intel_runtime_pm_enable_interrupts(dev_priv);
 
-       intel_modeset_init_hw(&dev_priv->drm);
+               intel_modeset_init_hw(dev);
 
-       spin_lock_irq(&dev_priv->irq_lock);
-       if (dev_priv->display.hpd_irq_setup)
-               dev_priv->display.hpd_irq_setup(dev_priv);
-       spin_unlock_irq(&dev_priv->irq_lock);
+               spin_lock_irq(&dev_priv->irq_lock);
+               if (dev_priv->display.hpd_irq_setup)
+                       dev_priv->display.hpd_irq_setup(dev_priv);
+               spin_unlock_irq(&dev_priv->irq_lock);
 
-       intel_display_resume(&dev_priv->drm);
+               ret = __intel_display_resume(dev, state);
+               if (ret)
+                       DRM_ERROR("Restoring old state failed with %i\n", ret);
 
-       intel_hpd_init(dev_priv);
+               intel_hpd_init(dev_priv);
+       }
 
-       drm_modeset_unlock_all(&dev_priv->drm);
+       drm_modeset_drop_locks(ctx);
+       drm_modeset_acquire_fini(ctx);
+       mutex_unlock(&dev->mode_config.mutex);
 }
 
 static bool intel_crtc_has_pending_flip(struct drm_crtc *crtc)
@@ -16156,9 +16231,10 @@ void intel_display_resume(struct drm_device *dev)
        struct drm_atomic_state *state = dev_priv->modeset_restore_state;
        struct drm_modeset_acquire_ctx ctx;
        int ret;
-       bool setup = false;
 
        dev_priv->modeset_restore_state = NULL;
+       if (state)
+               state->acquire_ctx = &ctx;
 
        /*
         * This is a cludge because with real atomic modeset mode_config.mutex
@@ -16169,43 +16245,17 @@ void intel_display_resume(struct drm_device *dev)
        mutex_lock(&dev->mode_config.mutex);
        drm_modeset_acquire_init(&ctx, 0);
 
-retry:
-       ret = drm_modeset_lock_all_ctx(dev, &ctx);
-
-       if (ret == 0 && !setup) {
-               setup = true;
-
-               intel_modeset_setup_hw_state(dev);
-               i915_redisable_vga(dev);
-       }
-
-       if (ret == 0 && state) {
-               struct drm_crtc_state *crtc_state;
-               struct drm_crtc *crtc;
-               int i;
-
-               state->acquire_ctx = &ctx;
-
-               /* ignore any reset values/BIOS leftovers in the WM registers */
-               to_intel_atomic_state(state)->skip_intermediate_wm = true;
-
-               for_each_crtc_in_state(state, crtc, crtc_state, i) {
-                       /*
-                        * Force recalculation even if we restore
-                        * current state. With fast modeset this may not result
-                        * in a modeset when the state is compatible.
-                        */
-                       crtc_state->mode_changed = true;
-               }
-
-               ret = drm_atomic_commit(state);
-       }
+       while (1) {
+               ret = drm_modeset_lock_all_ctx(dev, &ctx);
+               if (ret != -EDEADLK)
+                       break;
 
-       if (ret == -EDEADLK) {
                drm_modeset_backoff(&ctx);
-               goto retry;
        }
 
+       if (!ret)
+               ret = __intel_display_resume(dev, state);
+
        drm_modeset_drop_locks(&ctx);
        drm_modeset_acquire_fini(&ctx);
        mutex_unlock(&dev->mode_config.mutex);
index 6a7ad3e..3836a1c 100644 (file)
@@ -1230,12 +1230,29 @@ static int intel_sanitize_fbc_option(struct drm_i915_private *dev_priv)
        if (i915.enable_fbc >= 0)
                return !!i915.enable_fbc;
 
+       if (!HAS_FBC(dev_priv))
+               return 0;
+
        if (IS_BROADWELL(dev_priv))
                return 1;
 
        return 0;
 }
 
+static bool need_fbc_vtd_wa(struct drm_i915_private *dev_priv)
+{
+#ifdef CONFIG_INTEL_IOMMU
+       /* WaFbcTurnOffFbcWhenHyperVisorIsUsed:skl,bxt */
+       if (intel_iommu_gfx_mapped &&
+           (IS_SKYLAKE(dev_priv) || IS_BROXTON(dev_priv))) {
+               DRM_INFO("Disabling framebuffer compression (FBC) to prevent screen flicker with VT-d enabled\n");
+               return true;
+       }
+#endif
+
+       return false;
+}
+
 /**
  * intel_fbc_init - Initialize FBC
  * @dev_priv: the i915 device
@@ -1253,6 +1270,9 @@ void intel_fbc_init(struct drm_i915_private *dev_priv)
        fbc->active = false;
        fbc->work.scheduled = false;
 
+       if (need_fbc_vtd_wa(dev_priv))
+               mkwrite_device_info(dev_priv)->has_fbc = false;
+
        i915.enable_fbc = intel_sanitize_fbc_option(dev_priv);
        DRM_DEBUG_KMS("Sanitized enable_fbc value: %d\n", i915.enable_fbc);
 
index 97ba6c8..d5deb58 100644 (file)
@@ -3344,6 +3344,8 @@ static uint32_t skl_wm_method2(uint32_t pixel_rate, uint32_t pipe_htotal,
                plane_bytes_per_line *= 4;
                plane_blocks_per_line = DIV_ROUND_UP(plane_bytes_per_line, 512);
                plane_blocks_per_line /= 4;
+       } else if (tiling == DRM_FORMAT_MOD_NONE) {
+               plane_blocks_per_line = DIV_ROUND_UP(plane_bytes_per_line, 512) + 1;
        } else {
                plane_blocks_per_line = DIV_ROUND_UP(plane_bytes_per_line, 512);
        }
@@ -6574,9 +6576,7 @@ void intel_init_gt_powersave(struct drm_i915_private *dev_priv)
 
 void intel_cleanup_gt_powersave(struct drm_i915_private *dev_priv)
 {
-       if (IS_CHERRYVIEW(dev_priv))
-               return;
-       else if (IS_VALLEYVIEW(dev_priv))
+       if (IS_VALLEYVIEW(dev_priv))
                valleyview_cleanup_gt_powersave(dev_priv);
 
        if (!i915.enable_rc6)
index cca7792..1d3161b 100644 (file)
@@ -1178,8 +1178,8 @@ static int bxt_init_workarounds(struct intel_engine_cs *engine)
                I915_WRITE(GEN8_L3SQCREG1, L3_GENERAL_PRIO_CREDITS(62) |
                                           L3_HIGH_PRIO_CREDITS(2));
 
-       /* WaInsertDummyPushConstPs:bxt */
-       if (IS_BXT_REVID(dev_priv, 0, BXT_REVID_B0))
+       /* WaToEnableHwFixForPushConstHWBug:bxt */
+       if (IS_BXT_REVID(dev_priv, BXT_REVID_C0, REVID_FOREVER))
                WA_SET_BIT_MASKED(COMMON_SLICE_CHICKEN2,
                                  GEN8_SBE_DISABLE_REPLAY_BUF_OPTIMIZATION);
 
@@ -1222,8 +1222,8 @@ static int kbl_init_workarounds(struct intel_engine_cs *engine)
                I915_WRITE(GEN8_L3SQCREG4, I915_READ(GEN8_L3SQCREG4) |
                           GEN8_LQSC_RO_PERF_DIS);
 
-       /* WaInsertDummyPushConstPs:kbl */
-       if (IS_KBL_REVID(dev_priv, 0, KBL_REVID_B0))
+       /* WaToEnableHwFixForPushConstHWBug:kbl */
+       if (IS_KBL_REVID(dev_priv, KBL_REVID_C0, REVID_FOREVER))
                WA_SET_BIT_MASKED(COMMON_SLICE_CHICKEN2,
                                  GEN8_SBE_DISABLE_REPLAY_BUF_OPTIMIZATION);
 
index 23ac804..294de45 100644 (file)
@@ -2,6 +2,9 @@ config DRM_MEDIATEK
        tristate "DRM Support for Mediatek SoCs"
        depends on DRM
        depends on ARCH_MEDIATEK || (ARM && COMPILE_TEST)
+       depends on COMMON_CLK
+       depends on HAVE_ARM_SMCCC
+       depends on OF
        select DRM_GEM_CMA_HELPER
        select DRM_KMS_HELPER
        select DRM_MIPI_DSI
index 6de3428..ddef0d4 100644 (file)
@@ -198,16 +198,7 @@ static int radeon_atpx_validate(struct radeon_atpx *atpx)
        atpx->is_hybrid = false;
        if (valid_bits & ATPX_MS_HYBRID_GFX_SUPPORTED) {
                printk("ATPX Hybrid Graphics\n");
-#if 1
-               /* This is a temporary hack until the D3 cold support
-                * makes it upstream.  The ATPX power_control method seems
-                * to still work on even if the system should be using
-                * the new standardized hybrid D3 cold ACPI interface.
-                */
-               atpx->functions.power_cntl = true;
-#else
                atpx->functions.power_cntl = false;
-#endif
                atpx->is_hybrid = true;
        }
 
index 730d840..d0203a1 100644 (file)
@@ -491,7 +491,7 @@ struct it87_sio_data {
 struct it87_data {
        const struct attribute_group *groups[7];
        enum chips type;
-       u16 features;
+       u32 features;
        u8 peci_mask;
        u8 old_peci_mask;
 
index f233726..1bb97f6 100644 (file)
@@ -38,6 +38,7 @@
 #define AT91_I2C_TIMEOUT       msecs_to_jiffies(100)   /* transfer timeout */
 #define AT91_I2C_DMA_THRESHOLD 8                       /* enable DMA if transfer size is bigger than this threshold */
 #define AUTOSUSPEND_TIMEOUT            2000
+#define AT91_I2C_MAX_ALT_CMD_DATA_SIZE 256
 
 /* AT91 TWI register definitions */
 #define        AT91_TWI_CR             0x0000  /* Control Register */
@@ -141,6 +142,7 @@ struct at91_twi_dev {
        unsigned twi_cwgr_reg;
        struct at91_twi_pdata *pdata;
        bool use_dma;
+       bool use_alt_cmd;
        bool recv_len_abort;
        u32 fifo_size;
        struct at91_twi_dma dma;
@@ -269,7 +271,7 @@ static void at91_twi_write_next_byte(struct at91_twi_dev *dev)
 
        /* send stop when last byte has been written */
        if (--dev->buf_len == 0)
-               if (!dev->pdata->has_alt_cmd)
+               if (!dev->use_alt_cmd)
                        at91_twi_write(dev, AT91_TWI_CR, AT91_TWI_STOP);
 
        dev_dbg(dev->dev, "wrote 0x%x, to go %d\n", *dev->buf, dev->buf_len);
@@ -292,7 +294,7 @@ static void at91_twi_write_data_dma_callback(void *data)
         * we just have to enable TXCOMP one.
         */
        at91_twi_write(dev, AT91_TWI_IER, AT91_TWI_TXCOMP);
-       if (!dev->pdata->has_alt_cmd)
+       if (!dev->use_alt_cmd)
                at91_twi_write(dev, AT91_TWI_CR, AT91_TWI_STOP);
 }
 
@@ -410,7 +412,7 @@ static void at91_twi_read_next_byte(struct at91_twi_dev *dev)
        }
 
        /* send stop if second but last byte has been read */
-       if (!dev->pdata->has_alt_cmd && dev->buf_len == 1)
+       if (!dev->use_alt_cmd && dev->buf_len == 1)
                at91_twi_write(dev, AT91_TWI_CR, AT91_TWI_STOP);
 
        dev_dbg(dev->dev, "read 0x%x, to go %d\n", *dev->buf, dev->buf_len);
@@ -426,7 +428,7 @@ static void at91_twi_read_data_dma_callback(void *data)
        dma_unmap_single(dev->dev, sg_dma_address(&dev->dma.sg[0]),
                         dev->buf_len, DMA_FROM_DEVICE);
 
-       if (!dev->pdata->has_alt_cmd) {
+       if (!dev->use_alt_cmd) {
                /* The last two bytes have to be read without using dma */
                dev->buf += dev->buf_len - 2;
                dev->buf_len = 2;
@@ -443,7 +445,7 @@ static void at91_twi_read_data_dma(struct at91_twi_dev *dev)
        struct dma_chan *chan_rx = dma->chan_rx;
        size_t buf_len;
 
-       buf_len = (dev->pdata->has_alt_cmd) ? dev->buf_len : dev->buf_len - 2;
+       buf_len = (dev->use_alt_cmd) ? dev->buf_len : dev->buf_len - 2;
        dma->direction = DMA_FROM_DEVICE;
 
        /* Keep in mind that we won't use dma to read the last two bytes */
@@ -651,7 +653,7 @@ static int at91_do_twi_transfer(struct at91_twi_dev *dev)
                unsigned start_flags = AT91_TWI_START;
 
                /* if only one byte is to be read, immediately stop transfer */
-               if (!has_alt_cmd && dev->buf_len <= 1 &&
+               if (!dev->use_alt_cmd && dev->buf_len <= 1 &&
                    !(dev->msg->flags & I2C_M_RECV_LEN))
                        start_flags |= AT91_TWI_STOP;
                at91_twi_write(dev, AT91_TWI_CR, start_flags);
@@ -745,7 +747,7 @@ static int at91_twi_xfer(struct i2c_adapter *adap, struct i2c_msg *msg, int num)
        int ret;
        unsigned int_addr_flag = 0;
        struct i2c_msg *m_start = msg;
-       bool is_read, use_alt_cmd = false;
+       bool is_read;
 
        dev_dbg(&adap->dev, "at91_xfer: processing %d messages:\n", num);
 
@@ -768,14 +770,16 @@ static int at91_twi_xfer(struct i2c_adapter *adap, struct i2c_msg *msg, int num)
                at91_twi_write(dev, AT91_TWI_IADR, internal_address);
        }
 
+       dev->use_alt_cmd = false;
        is_read = (m_start->flags & I2C_M_RD);
        if (dev->pdata->has_alt_cmd) {
-               if (m_start->len > 0) {
+               if (m_start->len > 0 &&
+                   m_start->len < AT91_I2C_MAX_ALT_CMD_DATA_SIZE) {
                        at91_twi_write(dev, AT91_TWI_CR, AT91_TWI_ACMEN);
                        at91_twi_write(dev, AT91_TWI_ACR,
                                       AT91_TWI_ACR_DATAL(m_start->len) |
                                       ((is_read) ? AT91_TWI_ACR_DIR : 0));
-                       use_alt_cmd = true;
+                       dev->use_alt_cmd = true;
                } else {
                        at91_twi_write(dev, AT91_TWI_CR, AT91_TWI_ACMDIS);
                }
@@ -784,7 +788,7 @@ static int at91_twi_xfer(struct i2c_adapter *adap, struct i2c_msg *msg, int num)
        at91_twi_write(dev, AT91_TWI_MMR,
                       (m_start->addr << 16) |
                       int_addr_flag |
-                      ((!use_alt_cmd && is_read) ? AT91_TWI_MREAD : 0));
+                      ((!dev->use_alt_cmd && is_read) ? AT91_TWI_MREAD : 0));
 
        dev->buf_len = m_start->len;
        dev->buf = m_start->buf;
index 19c8438..95f7cac 100644 (file)
@@ -158,7 +158,7 @@ static irqreturn_t bcm_iproc_i2c_isr(int irq, void *data)
 
        if (status & BIT(IS_M_START_BUSY_SHIFT)) {
                iproc_i2c->xfer_is_done = 1;
-               complete_all(&iproc_i2c->done);
+               complete(&iproc_i2c->done);
        }
 
        writel(status, iproc_i2c->base + IS_OFFSET);
index ac9f476..f987432 100644 (file)
@@ -229,7 +229,7 @@ static irqreturn_t bcm_kona_i2c_isr(int irq, void *devid)
                       dev->base + TXFCR_OFFSET);
 
        writel(status & ~ISR_RESERVED_MASK, dev->base + ISR_OFFSET);
-       complete_all(&dev->done);
+       complete(&dev->done);
 
        return IRQ_HANDLED;
 }
index 3f5a4d7..385b57b 100644 (file)
@@ -228,7 +228,7 @@ static irqreturn_t brcmstb_i2c_isr(int irq, void *devid)
                return IRQ_NONE;
 
        brcmstb_i2c_enable_disable_irq(dev, INT_DISABLE);
-       complete_all(&dev->done);
+       complete(&dev->done);
 
        dev_dbg(dev->device, "isr handled");
        return IRQ_HANDLED;
index a0d95ff..2d5ff86 100644 (file)
@@ -215,7 +215,7 @@ static int ec_i2c_xfer(struct i2c_adapter *adap, struct i2c_msg i2c_msgs[],
        msg->outsize = request_len;
        msg->insize = response_len;
 
-       result = cros_ec_cmd_xfer(bus->ec, msg);
+       result = cros_ec_cmd_xfer_status(bus->ec, msg);
        if (result < 0) {
                dev_err(dev, "Error transferring EC i2c message %d\n", result);
                goto exit;
index 71d3929..76e2898 100644 (file)
@@ -211,7 +211,7 @@ static void meson_i2c_stop(struct meson_i2c *i2c)
                meson_i2c_add_token(i2c, TOKEN_STOP);
        } else {
                i2c->state = STATE_IDLE;
-               complete_all(&i2c->done);
+               complete(&i2c->done);
        }
 }
 
@@ -238,7 +238,7 @@ static irqreturn_t meson_i2c_irq(int irqno, void *dev_id)
                dev_dbg(i2c->dev, "error bit set\n");
                i2c->error = -ENXIO;
                i2c->state = STATE_IDLE;
-               complete_all(&i2c->done);
+               complete(&i2c->done);
                goto out;
        }
 
@@ -269,7 +269,7 @@ static irqreturn_t meson_i2c_irq(int irqno, void *dev_id)
                break;
        case STATE_STOP:
                i2c->state = STATE_IDLE;
-               complete_all(&i2c->done);
+               complete(&i2c->done);
                break;
        case STATE_IDLE:
                break;
index dfa7a4b..ac88a52 100644 (file)
@@ -379,6 +379,7 @@ static int ocores_i2c_of_probe(struct platform_device *pdev,
                        if (!clock_frequency_present) {
                                dev_err(&pdev->dev,
                                        "Missing required parameter 'opencores,ip-clock-frequency'\n");
+                               clk_disable_unprepare(i2c->clk);
                                return -ENODEV;
                        }
                        i2c->ip_clock_khz = clock_frequency / 1000;
@@ -467,20 +468,21 @@ static int ocores_i2c_probe(struct platform_device *pdev)
                default:
                        dev_err(&pdev->dev, "Unsupported I/O width (%d)\n",
                                i2c->reg_io_width);
-                       return -EINVAL;
+                       ret = -EINVAL;
+                       goto err_clk;
                }
        }
 
        ret = ocores_init(&pdev->dev, i2c);
        if (ret)
-               return ret;
+               goto err_clk;
 
        init_waitqueue_head(&i2c->wait);
        ret = devm_request_irq(&pdev->dev, irq, ocores_isr, 0,
                               pdev->name, i2c);
        if (ret) {
                dev_err(&pdev->dev, "Cannot claim IRQ\n");
-               return ret;
+               goto err_clk;
        }
 
        /* hook up driver to tree */
@@ -494,7 +496,7 @@ static int ocores_i2c_probe(struct platform_device *pdev)
        ret = i2c_add_adapter(&i2c->adap);
        if (ret) {
                dev_err(&pdev->dev, "Failed to add adapter\n");
-               return ret;
+               goto err_clk;
        }
 
        /* add in known devices to the bus */
@@ -504,6 +506,10 @@ static int ocores_i2c_probe(struct platform_device *pdev)
        }
 
        return 0;
+
+err_clk:
+       clk_disable_unprepare(i2c->clk);
+       return ret;
 }
 
 static int ocores_i2c_remove(struct platform_device *pdev)
index 8de073a..215ac87 100644 (file)
@@ -68,7 +68,7 @@ static int i2c_demux_activate_master(struct i2c_demux_pinctrl_priv *priv, u32 ne
        adap = of_find_i2c_adapter_by_node(priv->chan[new_chan].parent_np);
        if (!adap) {
                ret = -ENODEV;
-               goto err;
+               goto err_with_revert;
        }
 
        p = devm_pinctrl_get_select(adap->dev.parent, priv->bus_name);
@@ -103,6 +103,8 @@ static int i2c_demux_activate_master(struct i2c_demux_pinctrl_priv *priv, u32 ne
 
  err_with_put:
        i2c_put_adapter(adap);
+ err_with_revert:
+       of_changeset_revert(&priv->chan[new_chan].chgset);
  err:
        dev_err(priv->dev, "failed to setup demux-adapter %d (%d)\n", new_chan, ret);
        return ret;
index e6dfa1b..5f65a78 100644 (file)
@@ -2462,18 +2462,24 @@ static int cma_resolve_iboe_route(struct rdma_id_private *id_priv)
 
        if (addr->dev_addr.bound_dev_if) {
                ndev = dev_get_by_index(&init_net, addr->dev_addr.bound_dev_if);
-               if (!ndev)
-                       return -ENODEV;
+               if (!ndev) {
+                       ret = -ENODEV;
+                       goto err2;
+               }
 
                if (ndev->flags & IFF_LOOPBACK) {
                        dev_put(ndev);
-                       if (!id_priv->id.device->get_netdev)
-                               return -EOPNOTSUPP;
+                       if (!id_priv->id.device->get_netdev) {
+                               ret = -EOPNOTSUPP;
+                               goto err2;
+                       }
 
                        ndev = id_priv->id.device->get_netdev(id_priv->id.device,
                                                              id_priv->id.port_num);
-                       if (!ndev)
-                               return -ENODEV;
+                       if (!ndev) {
+                               ret = -ENODEV;
+                               goto err2;
+                       }
                }
 
                route->path_rec->net = &init_net;
index 3aca7f6..b6a953a 100644 (file)
@@ -1827,8 +1827,12 @@ static int process_mpa_request(struct c4iw_ep *ep, struct sk_buff *skb)
                                (ep->mpa_pkt + sizeof(*mpa));
                        ep->ird = ntohs(mpa_v2_params->ird) &
                                MPA_V2_IRD_ORD_MASK;
+                       ep->ird = min_t(u32, ep->ird,
+                                       cur_max_read_depth(ep->com.dev));
                        ep->ord = ntohs(mpa_v2_params->ord) &
                                MPA_V2_IRD_ORD_MASK;
+                       ep->ord = min_t(u32, ep->ord,
+                                       cur_max_read_depth(ep->com.dev));
                        PDBG("%s initiator ird %u ord %u\n", __func__, ep->ird,
                             ep->ord);
                        if (ntohs(mpa_v2_params->ird) & MPA_V2_PEER2PEER_MODEL)
@@ -3136,7 +3140,7 @@ int c4iw_accept_cr(struct iw_cm_id *cm_id, struct iw_cm_conn_param *conn_param)
        if (ep->mpa_attr.version == 2 && ep->mpa_attr.enhanced_rdma_conn) {
                if (conn_param->ord > ep->ird) {
                        if (RELAXED_IRD_NEGOTIATION) {
-                               ep->ord = ep->ird;
+                               conn_param->ord = ep->ird;
                        } else {
                                ep->ird = conn_param->ird;
                                ep->ord = conn_param->ord;
index 812ab72..ac926c9 100644 (file)
@@ -1016,15 +1016,15 @@ int c4iw_resize_cq(struct ib_cq *cq, int cqe, struct ib_udata *udata)
 int c4iw_arm_cq(struct ib_cq *ibcq, enum ib_cq_notify_flags flags)
 {
        struct c4iw_cq *chp;
-       int ret;
+       int ret = 0;
        unsigned long flag;
 
        chp = to_c4iw_cq(ibcq);
        spin_lock_irqsave(&chp->lock, flag);
-       ret = t4_arm_cq(&chp->cq,
-                       (flags & IB_CQ_SOLICITED_MASK) == IB_CQ_SOLICITED);
+       t4_arm_cq(&chp->cq,
+                 (flags & IB_CQ_SOLICITED_MASK) == IB_CQ_SOLICITED);
+       if (flags & IB_CQ_REPORT_MISSED_EVENTS)
+               ret = t4_cq_notempty(&chp->cq);
        spin_unlock_irqrestore(&chp->lock, flag);
-       if (ret && !(flags & IB_CQ_REPORT_MISSED_EVENTS))
-               ret = 0;
        return ret;
 }
index 6126bbe..02173f4 100644 (file)
@@ -634,6 +634,11 @@ static inline int t4_valid_cqe(struct t4_cq *cq, struct t4_cqe *cqe)
        return (CQE_GENBIT(cqe) == cq->gen);
 }
 
+static inline int t4_cq_notempty(struct t4_cq *cq)
+{
+       return cq->sw_in_use || t4_valid_cqe(cq, &cq->queue[cq->cidx]);
+}
+
 static inline int t4_next_hw_cqe(struct t4_cq *cq, struct t4_cqe **cqe)
 {
        int ret;
index 79575ee..0566393 100644 (file)
@@ -47,7 +47,6 @@
 #include <linux/topology.h>
 #include <linux/cpumask.h>
 #include <linux/module.h>
-#include <linux/cpumask.h>
 
 #include "hfi.h"
 #include "affinity.h"
@@ -682,7 +681,7 @@ int hfi1_set_sdma_affinity(struct hfi1_devdata *dd, const char *buf,
                           size_t count)
 {
        struct hfi1_affinity_node *entry;
-       struct cpumask mask;
+       cpumask_var_t mask;
        int ret, i;
 
        spin_lock(&node_affinity.lock);
@@ -692,19 +691,24 @@ int hfi1_set_sdma_affinity(struct hfi1_devdata *dd, const char *buf,
        if (!entry)
                return -EINVAL;
 
-       ret = cpulist_parse(buf, &mask);
+       ret = zalloc_cpumask_var(&mask, GFP_KERNEL);
+       if (!ret)
+               return -ENOMEM;
+
+       ret = cpulist_parse(buf, mask);
        if (ret)
-               return ret;
+               goto out;
 
-       if (!cpumask_subset(&mask, cpu_online_mask) || cpumask_empty(&mask)) {
+       if (!cpumask_subset(mask, cpu_online_mask) || cpumask_empty(mask)) {
                dd_dev_warn(dd, "Invalid CPU mask\n");
-               return -EINVAL;
+               ret = -EINVAL;
+               goto out;
        }
 
        mutex_lock(&sdma_affinity_mutex);
        /* reset the SDMA interrupt affinity details */
        init_cpu_mask_set(&entry->def_intr);
-       cpumask_copy(&entry->def_intr.mask, &mask);
+       cpumask_copy(&entry->def_intr.mask, mask);
        /*
         * Reassign the affinity for each SDMA interrupt.
         */
@@ -720,8 +724,9 @@ int hfi1_set_sdma_affinity(struct hfi1_devdata *dd, const char *buf,
                if (ret)
                        break;
        }
-
        mutex_unlock(&sdma_affinity_mutex);
+out:
+       free_cpumask_var(mask);
        return ret ? ret : strnlen(buf, PAGE_SIZE);
 }
 
index dbab9d9..a49cc88 100644 (file)
@@ -223,28 +223,32 @@ DEBUGFS_SEQ_FILE_OPEN(ctx_stats)
 DEBUGFS_FILE_OPS(ctx_stats);
 
 static void *_qp_stats_seq_start(struct seq_file *s, loff_t *pos)
-__acquires(RCU)
+       __acquires(RCU)
 {
        struct qp_iter *iter;
        loff_t n = *pos;
 
-       rcu_read_lock();
        iter = qp_iter_init(s->private);
+
+       /* stop calls rcu_read_unlock */
+       rcu_read_lock();
+
        if (!iter)
                return NULL;
 
-       while (n--) {
+       do {
                if (qp_iter_next(iter)) {
                        kfree(iter);
                        return NULL;
                }
-       }
+       } while (n--);
 
        return iter;
 }
 
 static void *_qp_stats_seq_next(struct seq_file *s, void *iter_ptr,
                                loff_t *pos)
+       __must_hold(RCU)
 {
        struct qp_iter *iter = iter_ptr;
 
@@ -259,7 +263,7 @@ static void *_qp_stats_seq_next(struct seq_file *s, void *iter_ptr,
 }
 
 static void _qp_stats_seq_stop(struct seq_file *s, void *iter_ptr)
-__releases(RCU)
+       __releases(RCU)
 {
        rcu_read_unlock();
 }
index 8246dc7..303f105 100644 (file)
@@ -888,14 +888,15 @@ void set_all_slowpath(struct hfi1_devdata *dd)
 }
 
 static inline int set_armed_to_active(struct hfi1_ctxtdata *rcd,
-                                     struct hfi1_packet packet,
+                                     struct hfi1_packet *packet,
                                      struct hfi1_devdata *dd)
 {
        struct work_struct *lsaw = &rcd->ppd->linkstate_active_work;
-       struct hfi1_message_header *hdr = hfi1_get_msgheader(packet.rcd->dd,
-                                                            packet.rhf_addr);
+       struct hfi1_message_header *hdr = hfi1_get_msgheader(packet->rcd->dd,
+                                                            packet->rhf_addr);
+       u8 etype = rhf_rcv_type(packet->rhf);
 
-       if (hdr2sc(hdr, packet.rhf) != 0xf) {
+       if (etype == RHF_RCV_TYPE_IB && hdr2sc(hdr, packet->rhf) != 0xf) {
                int hwstate = read_logical_state(dd);
 
                if (hwstate != LSTATE_ACTIVE) {
@@ -979,7 +980,7 @@ int handle_receive_interrupt(struct hfi1_ctxtdata *rcd, int thread)
                        /* Auto activate link on non-SC15 packet receive */
                        if (unlikely(rcd->ppd->host_link_state ==
                                     HLS_UP_ARMED) &&
-                           set_armed_to_active(rcd, packet, dd))
+                           set_armed_to_active(rcd, &packet, dd))
                                goto bail;
                        last = process_rcv_packet(&packet, thread);
                }
index 1ecbec1..7e03ccd 100644 (file)
@@ -183,6 +183,7 @@ static int hfi1_file_open(struct inode *inode, struct file *fp)
        if (fd) {
                fd->rec_cpu_num = -1; /* no cpu affinity by default */
                fd->mm = current->mm;
+               atomic_inc(&fd->mm->mm_count);
        }
 
        fp->private_data = fd;
@@ -222,7 +223,7 @@ static long hfi1_file_ioctl(struct file *fp, unsigned int cmd,
                ret = assign_ctxt(fp, &uinfo);
                if (ret < 0)
                        return ret;
-               setup_ctxt(fp);
+               ret = setup_ctxt(fp);
                if (ret)
                        return ret;
                ret = user_init(fp);
@@ -779,6 +780,7 @@ static int hfi1_file_close(struct inode *inode, struct file *fp)
        mutex_unlock(&hfi1_mutex);
        hfi1_free_ctxtdata(dd, uctxt);
 done:
+       mmdrop(fdata->mm);
        kobject_put(&dd->kobj);
        kfree(fdata);
        return 0;
index 1000e0f..a021e66 100644 (file)
@@ -1272,9 +1272,26 @@ static inline int hdr2sc(struct hfi1_message_header *hdr, u64 rhf)
               ((!!(rhf_dc_info(rhf))) << 4);
 }
 
+#define HFI1_JKEY_WIDTH       16
+#define HFI1_JKEY_MASK        (BIT(16) - 1)
+#define HFI1_ADMIN_JKEY_RANGE 32
+
+/*
+ * J_KEYs are split and allocated in the following groups:
+ *   0 - 31    - users with administrator privileges
+ *  32 - 63    - kernel protocols using KDETH packets
+ *  64 - 65535 - all other users using KDETH packets
+ */
 static inline u16 generate_jkey(kuid_t uid)
 {
-       return from_kuid(current_user_ns(), uid) & 0xffff;
+       u16 jkey = from_kuid(current_user_ns(), uid) & HFI1_JKEY_MASK;
+
+       if (capable(CAP_SYS_ADMIN))
+               jkey &= HFI1_ADMIN_JKEY_RANGE - 1;
+       else if (jkey < 64)
+               jkey |= BIT(HFI1_JKEY_WIDTH - 1);
+
+       return jkey;
 }
 
 /*
@@ -1656,7 +1673,6 @@ struct cc_state *get_cc_state_protected(struct hfi1_pportdata *ppd)
 struct hfi1_devdata *hfi1_init_dd(struct pci_dev *,
                                  const struct pci_device_id *);
 void hfi1_free_devdata(struct hfi1_devdata *);
-void cc_state_reclaim(struct rcu_head *rcu);
 struct hfi1_devdata *hfi1_alloc_devdata(struct pci_dev *pdev, size_t extra);
 
 /* LED beaconing functions */
index a358d23..b793545 100644 (file)
@@ -1333,7 +1333,7 @@ static void cleanup_device_data(struct hfi1_devdata *dd)
                spin_unlock(&ppd->cc_state_lock);
 
                if (cc_state)
-                       call_rcu(&cc_state->rcu, cc_state_reclaim);
+                       kfree_rcu(cc_state, rcu);
        }
 
        free_credit_return(dd);
index 1263abe..39e42c3 100644 (file)
@@ -1819,6 +1819,11 @@ static int __subn_get_opa_cable_info(struct opa_smp *smp, u32 am, u8 *data,
        u32 len = OPA_AM_CI_LEN(am) + 1;
        int ret;
 
+       if (dd->pport->port_type != PORT_TYPE_QSFP) {
+               smp->status |= IB_SMP_INVALID_FIELD;
+               return reply((struct ib_mad_hdr *)smp);
+       }
+
 #define __CI_PAGE_SIZE BIT(7) /* 128 bytes */
 #define __CI_PAGE_MASK ~(__CI_PAGE_SIZE - 1)
 #define __CI_PAGE_NUM(a) ((a) & __CI_PAGE_MASK)
@@ -3398,7 +3403,7 @@ static void apply_cc_state(struct hfi1_pportdata *ppd)
 
        spin_unlock(&ppd->cc_state_lock);
 
-       call_rcu(&old_cc_state->rcu, cc_state_reclaim);
+       kfree_rcu(old_cc_state, rcu);
 }
 
 static int __subn_set_opa_cong_setting(struct opa_smp *smp, u32 am, u8 *data,
@@ -3553,13 +3558,6 @@ static int __subn_get_opa_cc_table(struct opa_smp *smp, u32 am, u8 *data,
        return reply((struct ib_mad_hdr *)smp);
 }
 
-void cc_state_reclaim(struct rcu_head *rcu)
-{
-       struct cc_state *cc_state = container_of(rcu, struct cc_state, rcu);
-
-       kfree(cc_state);
-}
-
 static int __subn_set_opa_cc_table(struct opa_smp *smp, u32 am, u8 *data,
                                   struct ib_device *ibdev, u8 port,
                                   u32 *resp_len)
index a5aa351..4e4d831 100644 (file)
@@ -656,10 +656,6 @@ struct qp_iter *qp_iter_init(struct hfi1_ibdev *dev)
 
        iter->dev = dev;
        iter->specials = dev->rdi.ibdev.phys_port_cnt * 2;
-       if (qp_iter_next(iter)) {
-               kfree(iter);
-               return NULL;
-       }
 
        return iter;
 }
index a207717..4e95ad8 100644 (file)
@@ -706,8 +706,8 @@ int get_cable_info(struct hfi1_devdata *dd, u32 port_num, u32 addr, u32 len,
                   u8 *data)
 {
        struct hfi1_pportdata *ppd;
-       u32 excess_len = 0;
-       int ret = 0;
+       u32 excess_len = len;
+       int ret = 0, offset = 0;
 
        if (port_num > dd->num_pports || port_num < 1) {
                dd_dev_info(dd, "%s: Invalid port number %d\n",
@@ -740,6 +740,34 @@ int get_cable_info(struct hfi1_devdata *dd, u32 port_num, u32 addr, u32 len,
        }
 
        memcpy(data, &ppd->qsfp_info.cache[addr], len);
+
+       if (addr <= QSFP_MONITOR_VAL_END &&
+           (addr + len) >= QSFP_MONITOR_VAL_START) {
+               /* Overlap with the dynamic channel monitor range */
+               if (addr < QSFP_MONITOR_VAL_START) {
+                       if (addr + len <= QSFP_MONITOR_VAL_END)
+                               len = addr + len - QSFP_MONITOR_VAL_START;
+                       else
+                               len = QSFP_MONITOR_RANGE;
+                       offset = QSFP_MONITOR_VAL_START - addr;
+                       addr = QSFP_MONITOR_VAL_START;
+               } else if (addr == QSFP_MONITOR_VAL_START) {
+                       offset = 0;
+                       if (addr + len > QSFP_MONITOR_VAL_END)
+                               len = QSFP_MONITOR_RANGE;
+               } else {
+                       offset = 0;
+                       if (addr + len > QSFP_MONITOR_VAL_END)
+                               len = QSFP_MONITOR_VAL_END - addr + 1;
+               }
+               /* Refresh the values of the dynamic monitors from the cable */
+               ret = one_qsfp_read(ppd, dd->hfi1_id, addr, data + offset, len);
+               if (ret != len) {
+                       ret = -EAGAIN;
+                       goto set_zeroes;
+               }
+       }
+
        return 0;
 
 set_zeroes:
index 69275eb..36cf523 100644 (file)
@@ -74,6 +74,9 @@
 /* Defined fields that Intel requires of qualified cables */
 /* Byte 0 is Identifier, not checked */
 /* Byte 1 is reserved "status MSB" */
+#define QSFP_MONITOR_VAL_START 22
+#define QSFP_MONITOR_VAL_END 81
+#define QSFP_MONITOR_RANGE (QSFP_MONITOR_VAL_END - QSFP_MONITOR_VAL_START + 1)
 #define QSFP_TX_CTRL_BYTE_OFFS 86
 #define QSFP_PWR_CTRL_BYTE_OFFS 93
 #define QSFP_CDR_CTRL_BYTE_OFFS 98
index b738acd..8ec09e4 100644 (file)
@@ -232,7 +232,7 @@ struct i40iw_device {
        struct i40e_client *client;
        struct i40iw_hw hw;
        struct i40iw_cm_core cm_core;
-       unsigned long *mem_resources;
+       u8 *mem_resources;
        unsigned long *allocated_qps;
        unsigned long *allocated_cqs;
        unsigned long *allocated_mrs;
@@ -435,8 +435,8 @@ static inline int i40iw_alloc_resource(struct i40iw_device *iwdev,
        *next = resource_num + 1;
        if (*next == max_resources)
                *next = 0;
-       spin_unlock_irqrestore(&iwdev->resource_lock, flags);
        *req_resource_num = resource_num;
+       spin_unlock_irqrestore(&iwdev->resource_lock, flags);
 
        return 0;
 }
index 5026dc7..7ca0638 100644 (file)
@@ -535,8 +535,8 @@ static struct i40iw_puda_buf *i40iw_form_cm_frame(struct i40iw_cm_node *cm_node,
                buf += hdr_len;
        }
 
-       if (pd_len)
-               memcpy(buf, pdata->addr, pd_len);
+       if (pdata && pdata->addr)
+               memcpy(buf, pdata->addr, pdata->size);
 
        atomic_set(&sqbuf->refcount, 1);
 
@@ -3346,26 +3346,6 @@ int i40iw_cm_disconn(struct i40iw_qp *iwqp)
        return 0;
 }
 
-/**
- * i40iw_loopback_nop - Send a nop
- * @qp: associated hw qp
- */
-static void i40iw_loopback_nop(struct i40iw_sc_qp *qp)
-{
-       u64 *wqe;
-       u64 header;
-
-       wqe = qp->qp_uk.sq_base->elem;
-       set_64bit_val(wqe, 0, 0);
-       set_64bit_val(wqe, 8, 0);
-       set_64bit_val(wqe, 16, 0);
-
-       header = LS_64(I40IWQP_OP_NOP, I40IWQPSQ_OPCODE) |
-           LS_64(0, I40IWQPSQ_SIGCOMPL) |
-           LS_64(qp->qp_uk.swqe_polarity, I40IWQPSQ_VALID);
-       set_64bit_val(wqe, 24, header);
-}
-
 /**
  * i40iw_qp_disconnect - free qp and close cm
  * @iwqp: associate qp for the connection
@@ -3638,7 +3618,7 @@ int i40iw_accept(struct iw_cm_id *cm_id, struct iw_cm_conn_param *conn_param)
        } else {
                if (iwqp->page)
                        iwqp->sc_qp.qp_uk.sq_base = kmap(iwqp->page);
-               i40iw_loopback_nop(&iwqp->sc_qp);
+               dev->iw_priv_qp_ops->qp_send_lsmm(&iwqp->sc_qp, NULL, 0, 0);
        }
 
        if (iwqp->page)
index 6e90813..0cbbe40 100644 (file)
@@ -1558,6 +1558,10 @@ static int i40iw_open(struct i40e_info *ldev, struct i40e_client *client)
        enum i40iw_status_code status;
        struct i40iw_handler *hdl;
 
+       hdl = i40iw_find_netdev(ldev->netdev);
+       if (hdl)
+               return 0;
+
        hdl = kzalloc(sizeof(*hdl), GFP_KERNEL);
        if (!hdl)
                return -ENOMEM;
index 0e8db0a..6fd043b 100644 (file)
@@ -673,8 +673,11 @@ enum i40iw_status_code i40iw_free_virt_mem(struct i40iw_hw *hw,
 {
        if (!mem)
                return I40IW_ERR_PARAM;
+       /*
+        * mem->va points to the parent of mem, so both mem and mem->va
+        * can not be touched once mem->va is freed
+        */
        kfree(mem->va);
-       mem->va = NULL;
        return 0;
 }
 
index 2360338..6329c97 100644 (file)
@@ -794,7 +794,6 @@ static struct ib_qp *i40iw_create_qp(struct ib_pd *ibpd,
        return &iwqp->ibqp;
 error:
        i40iw_free_qp_resources(iwdev, iwqp, qp_num);
-       kfree(mem);
        return ERR_PTR(err_code);
 }
 
@@ -1926,8 +1925,7 @@ static int i40iw_dereg_mr(struct ib_mr *ib_mr)
                }
                if (iwpbl->pbl_allocated)
                        i40iw_free_pble(iwdev->pble_rsrc, palloc);
-               kfree(iwpbl->iwmr);
-               iwpbl->iwmr = NULL;
+               kfree(iwmr);
                return 0;
        }
 
index d6fc8a6..006db64 100644 (file)
@@ -576,8 +576,8 @@ static int mlx4_ib_ipoib_csum_ok(__be16 status, __be16 checksum)
                checksum == cpu_to_be16(0xffff);
 }
 
-static int use_tunnel_data(struct mlx4_ib_qp *qp, struct mlx4_ib_cq *cq, struct ib_wc *wc,
-                          unsigned tail, struct mlx4_cqe *cqe, int is_eth)
+static void use_tunnel_data(struct mlx4_ib_qp *qp, struct mlx4_ib_cq *cq, struct ib_wc *wc,
+                           unsigned tail, struct mlx4_cqe *cqe, int is_eth)
 {
        struct mlx4_ib_proxy_sqp_hdr *hdr;
 
@@ -600,8 +600,6 @@ static int use_tunnel_data(struct mlx4_ib_qp *qp, struct mlx4_ib_cq *cq, struct
                wc->slid        = be16_to_cpu(hdr->tun.slid_mac_47_32);
                wc->sl          = (u8) (be16_to_cpu(hdr->tun.sl_vid) >> 12);
        }
-
-       return 0;
 }
 
 static void mlx4_ib_qp_sw_comp(struct mlx4_ib_qp *qp, int num_entries,
@@ -692,7 +690,7 @@ repoll:
        if (unlikely((cqe->owner_sr_opcode & MLX4_CQE_OPCODE_MASK) == MLX4_OPCODE_NOP &&
                     is_send)) {
                pr_warn("Completion for NOP opcode detected!\n");
-               return -EINVAL;
+               return -EAGAIN;
        }
 
        /* Resize CQ in progress */
@@ -723,7 +721,7 @@ repoll:
                if (unlikely(!mqp)) {
                        pr_warn("CQ %06x with entry for unknown QPN %06x\n",
                               cq->mcq.cqn, be32_to_cpu(cqe->vlan_my_qpn) & MLX4_CQE_QPN_MASK);
-                       return -EINVAL;
+                       return -EAGAIN;
                }
 
                *cur_qp = to_mibqp(mqp);
@@ -741,7 +739,7 @@ repoll:
                if (unlikely(!msrq)) {
                        pr_warn("CQ %06x with entry for unknown SRQN %06x\n",
                                cq->mcq.cqn, srq_num);
-                       return -EINVAL;
+                       return -EAGAIN;
                }
        }
 
@@ -852,9 +850,11 @@ repoll:
                if (mlx4_is_mfunc(to_mdev(cq->ibcq.device)->dev)) {
                        if ((*cur_qp)->mlx4_ib_qp_type &
                            (MLX4_IB_QPT_PROXY_SMI_OWNER |
-                            MLX4_IB_QPT_PROXY_SMI | MLX4_IB_QPT_PROXY_GSI))
-                               return use_tunnel_data(*cur_qp, cq, wc, tail,
-                                                      cqe, is_eth);
+                            MLX4_IB_QPT_PROXY_SMI | MLX4_IB_QPT_PROXY_GSI)) {
+                               use_tunnel_data(*cur_qp, cq, wc, tail, cqe,
+                                               is_eth);
+                               return 0;
+                       }
                }
 
                wc->slid           = be16_to_cpu(cqe->rlid);
index a84bb76..1b4094b 100644 (file)
@@ -37,7 +37,6 @@
 #include <linux/pci.h>
 #include <linux/dma-mapping.h>
 #include <linux/slab.h>
-#include <linux/io-mapping.h>
 #if defined(CONFIG_X86)
 #include <asm/pat.h>
 #endif
index 16740dc..67fc0b6 100644 (file)
@@ -1156,18 +1156,18 @@ static void ocrdma_get_attr(struct ocrdma_dev *dev,
        attr->max_srq =
                (rsp->max_srq_rpir_qps & OCRDMA_MBX_QUERY_CFG_MAX_SRQ_MASK) >>
                OCRDMA_MBX_QUERY_CFG_MAX_SRQ_OFFSET;
-       attr->max_send_sge = ((rsp->max_write_send_sge &
+       attr->max_send_sge = ((rsp->max_recv_send_sge &
                               OCRDMA_MBX_QUERY_CFG_MAX_SEND_SGE_MASK) >>
                              OCRDMA_MBX_QUERY_CFG_MAX_SEND_SGE_SHIFT);
-       attr->max_recv_sge = (rsp->max_write_send_sge &
-                             OCRDMA_MBX_QUERY_CFG_MAX_SEND_SGE_MASK) >>
-           OCRDMA_MBX_QUERY_CFG_MAX_SEND_SGE_SHIFT;
+       attr->max_recv_sge = (rsp->max_recv_send_sge &
+                             OCRDMA_MBX_QUERY_CFG_MAX_RECV_SGE_MASK) >>
+           OCRDMA_MBX_QUERY_CFG_MAX_RECV_SGE_SHIFT;
        attr->max_srq_sge = (rsp->max_srq_rqe_sge &
                              OCRDMA_MBX_QUERY_CFG_MAX_SRQ_SGE_MASK) >>
            OCRDMA_MBX_QUERY_CFG_MAX_SRQ_SGE_OFFSET;
-       attr->max_rdma_sge = (rsp->max_write_send_sge &
-                             OCRDMA_MBX_QUERY_CFG_MAX_WRITE_SGE_MASK) >>
-           OCRDMA_MBX_QUERY_CFG_MAX_WRITE_SGE_SHIFT;
+       attr->max_rdma_sge = (rsp->max_wr_rd_sge &
+                             OCRDMA_MBX_QUERY_CFG_MAX_RD_SGE_MASK) >>
+           OCRDMA_MBX_QUERY_CFG_MAX_RD_SGE_SHIFT;
        attr->max_ord_per_qp = (rsp->max_ird_ord_per_qp &
                                OCRDMA_MBX_QUERY_CFG_MAX_ORD_PER_QP_MASK) >>
            OCRDMA_MBX_QUERY_CFG_MAX_ORD_PER_QP_SHIFT;
index 0efc966..37df448 100644 (file)
@@ -554,9 +554,9 @@ enum {
        OCRDMA_MBX_QUERY_CFG_L3_TYPE_MASK               = 0x18,
        OCRDMA_MBX_QUERY_CFG_MAX_SEND_SGE_SHIFT         = 0,
        OCRDMA_MBX_QUERY_CFG_MAX_SEND_SGE_MASK          = 0xFFFF,
-       OCRDMA_MBX_QUERY_CFG_MAX_WRITE_SGE_SHIFT        = 16,
-       OCRDMA_MBX_QUERY_CFG_MAX_WRITE_SGE_MASK         = 0xFFFF <<
-                               OCRDMA_MBX_QUERY_CFG_MAX_WRITE_SGE_SHIFT,
+       OCRDMA_MBX_QUERY_CFG_MAX_RECV_SGE_SHIFT = 16,
+       OCRDMA_MBX_QUERY_CFG_MAX_RECV_SGE_MASK          = 0xFFFF <<
+                               OCRDMA_MBX_QUERY_CFG_MAX_RECV_SGE_SHIFT,
 
        OCRDMA_MBX_QUERY_CFG_MAX_ORD_PER_QP_SHIFT       = 0,
        OCRDMA_MBX_QUERY_CFG_MAX_ORD_PER_QP_MASK        = 0xFFFF,
@@ -612,6 +612,8 @@ enum {
        OCRDMA_MBX_QUERY_CFG_MAX_SRQ_SGE_OFFSET         = 0,
        OCRDMA_MBX_QUERY_CFG_MAX_SRQ_SGE_MASK           = 0xFFFF <<
                                OCRDMA_MBX_QUERY_CFG_MAX_SRQ_SGE_OFFSET,
+       OCRDMA_MBX_QUERY_CFG_MAX_RD_SGE_SHIFT           = 0,
+       OCRDMA_MBX_QUERY_CFG_MAX_RD_SGE_MASK            = 0xFFFF,
 };
 
 struct ocrdma_mbx_query_config {
@@ -619,7 +621,7 @@ struct ocrdma_mbx_query_config {
        struct ocrdma_mbx_rsp rsp;
        u32 qp_srq_cq_ird_ord;
        u32 max_pd_ca_ack_delay;
-       u32 max_write_send_sge;
+       u32 max_recv_send_sge;
        u32 max_ird_ord_per_qp;
        u32 max_shared_ird_ord;
        u32 max_mr;
@@ -639,6 +641,8 @@ struct ocrdma_mbx_query_config {
        u32 max_wqes_rqes_per_q;
        u32 max_cq_cqes_per_cq;
        u32 max_srq_rqe_sge;
+       u32 max_wr_rd_sge;
+       u32 ird_pgsz_num_pages;
 };
 
 struct ocrdma_fw_ver_rsp {
index b1a3d91..0aa8547 100644 (file)
@@ -125,8 +125,8 @@ int ocrdma_query_device(struct ib_device *ibdev, struct ib_device_attr *attr,
                                        IB_DEVICE_SYS_IMAGE_GUID |
                                        IB_DEVICE_LOCAL_DMA_LKEY |
                                        IB_DEVICE_MEM_MGT_EXTENSIONS;
-       attr->max_sge = dev->attr.max_send_sge;
-       attr->max_sge_rd = attr->max_sge;
+       attr->max_sge = min(dev->attr.max_send_sge, dev->attr.max_recv_sge);
+       attr->max_sge_rd = dev->attr.max_rdma_sge;
        attr->max_cq = dev->attr.max_cq;
        attr->max_cqe = dev->attr.max_cqe;
        attr->max_mr = dev->attr.max_mr;
index 5e75b43..5bad8e3 100644 (file)
@@ -189,27 +189,32 @@ static int _ctx_stats_seq_show(struct seq_file *s, void *v)
 DEBUGFS_FILE(ctx_stats)
 
 static void *_qp_stats_seq_start(struct seq_file *s, loff_t *pos)
+       __acquires(RCU)
 {
        struct qib_qp_iter *iter;
        loff_t n = *pos;
 
-       rcu_read_lock();
        iter = qib_qp_iter_init(s->private);
+
+       /* stop calls rcu_read_unlock */
+       rcu_read_lock();
+
        if (!iter)
                return NULL;
 
-       while (n--) {
+       do {
                if (qib_qp_iter_next(iter)) {
                        kfree(iter);
                        return NULL;
                }
-       }
+       } while (n--);
 
        return iter;
 }
 
 static void *_qp_stats_seq_next(struct seq_file *s, void *iter_ptr,
                                   loff_t *pos)
+       __must_hold(RCU)
 {
        struct qib_qp_iter *iter = iter_ptr;
 
@@ -224,6 +229,7 @@ static void *_qp_stats_seq_next(struct seq_file *s, void *iter_ptr,
 }
 
 static void _qp_stats_seq_stop(struct seq_file *s, void *iter_ptr)
+       __releases(RCU)
 {
        rcu_read_unlock();
 }
index fcdf379..c3edc03 100644 (file)
@@ -328,26 +328,12 @@ static ssize_t flash_write(struct file *file, const char __user *buf,
 
        pos = *ppos;
 
-       if (pos != 0) {
-               ret = -EINVAL;
-               goto bail;
-       }
-
-       if (count != sizeof(struct qib_flash)) {
-               ret = -EINVAL;
-               goto bail;
-       }
-
-       tmp = kmalloc(count, GFP_KERNEL);
-       if (!tmp) {
-               ret = -ENOMEM;
-               goto bail;
-       }
+       if (pos != 0 || count != sizeof(struct qib_flash))
+               return -EINVAL;
 
-       if (copy_from_user(tmp, buf, count)) {
-               ret = -EFAULT;
-               goto bail_tmp;
-       }
+       tmp = memdup_user(buf, count);
+       if (IS_ERR(tmp))
+               return PTR_ERR(tmp);
 
        dd = private2dd(file);
        if (qib_eeprom_write(dd, pos, tmp, count)) {
@@ -361,8 +347,6 @@ static ssize_t flash_write(struct file *file, const char __user *buf,
 
 bail_tmp:
        kfree(tmp);
-
-bail:
        return ret;
 }
 
index 9cc0aae..f9b8cd2 100644 (file)
@@ -573,10 +573,6 @@ struct qib_qp_iter *qib_qp_iter_init(struct qib_ibdev *dev)
                return NULL;
 
        iter->dev = dev;
-       if (qib_qp_iter_next(iter)) {
-               kfree(iter);
-               return NULL;
-       }
 
        return iter;
 }
index c229b9f..0a89a95 100644 (file)
@@ -664,7 +664,8 @@ static int __init usnic_ib_init(void)
                return err;
        }
 
-       if (pci_register_driver(&usnic_ib_pci_driver)) {
+       err = pci_register_driver(&usnic_ib_pci_driver);
+       if (err) {
                usnic_err("Unable to register with PCI\n");
                goto out_umem_fini;
        }
index bdb540f..870b4f2 100644 (file)
@@ -873,7 +873,8 @@ bail_qpn:
        free_qpn(&rdi->qp_dev->qpn_table, qp->ibqp.qp_num);
 
 bail_rq_wq:
-       vfree(qp->r_rq.wq);
+       if (!qp->ip)
+               vfree(qp->r_rq.wq);
 
 bail_driver_priv:
        rdi->driver_f.qp_priv_free(rdi, qp);
index ba6be06..7914c14 100644 (file)
@@ -448,7 +448,7 @@ isert_alloc_login_buf(struct isert_conn *isert_conn,
 
        isert_conn->login_rsp_buf = kzalloc(ISER_RX_PAYLOAD_SIZE, GFP_KERNEL);
        if (!isert_conn->login_rsp_buf) {
-               isert_err("Unable to allocate isert_conn->login_rspbuf\n");
+               ret = -ENOMEM;
                goto out_unmap_login_req_buf;
        }
 
index dfa23b0..883bbfe 100644 (file)
@@ -522,6 +522,11 @@ static int srpt_refresh_port(struct srpt_port *sport)
        if (ret)
                goto err_query_port;
 
+       snprintf(sport->port_guid, sizeof(sport->port_guid),
+               "0x%016llx%016llx",
+               be64_to_cpu(sport->gid.global.subnet_prefix),
+               be64_to_cpu(sport->gid.global.interface_id));
+
        if (!sport->mad_agent) {
                memset(&reg_req, 0, sizeof(reg_req));
                reg_req.mgmt_class = IB_MGMT_CLASS_DEVICE_MGMT;
@@ -2548,10 +2553,6 @@ static void srpt_add_one(struct ib_device *device)
                               sdev->device->name, i);
                        goto err_ring;
                }
-               snprintf(sport->port_guid, sizeof(sport->port_guid),
-                       "0x%016llx%016llx",
-                       be64_to_cpu(sport->gid.global.subnet_prefix),
-                       be64_to_cpu(sport->gid.global.interface_id));
        }
 
        spin_lock(&srpt_dev_lock);
index 7d61439..0c07e10 100644 (file)
@@ -376,7 +376,7 @@ static int tegra_kbc_start(struct tegra_kbc *kbc)
        /* Reset the KBC controller to clear all previous status.*/
        reset_control_assert(kbc->rst);
        udelay(100);
-       reset_control_assert(kbc->rst);
+       reset_control_deassert(kbc->rst);
        udelay(100);
 
        tegra_kbc_config_pins(kbc);
index faa295e..c83bce8 100644 (file)
@@ -553,7 +553,6 @@ int rmi_read_register_desc(struct rmi_device *d, u16 addr,
                goto free_struct_buff;
 
        reg = find_first_bit(rdesc->presense_map, RMI_REG_DESC_PRESENSE_BITS);
-       map_offset = 0;
        for (i = 0; i < rdesc->num_registers; i++) {
                struct rmi_register_desc_item *item = &rdesc->registers[i];
                int reg_size = struct_buf[offset];
@@ -576,6 +575,8 @@ int rmi_read_register_desc(struct rmi_device *d, u16 addr,
                item->reg = reg;
                item->reg_size = reg_size;
 
+               map_offset = 0;
+
                do {
                        for (b = 0; b < 7; b++) {
                                if (struct_buf[offset] & (0x1 << b))
index b4d3408..405252a 100644 (file)
@@ -1305,6 +1305,7 @@ static int __init i8042_create_aux_port(int idx)
        serio->write            = i8042_aux_write;
        serio->start            = i8042_start;
        serio->stop             = i8042_stop;
+       serio->ps2_cmd_mutex    = &i8042_mutex;
        serio->port_data        = port;
        serio->dev.parent       = &i8042_platform_device->dev;
        if (idx < 0) {
index a61b215..1ce3ecb 100644 (file)
@@ -1473,7 +1473,6 @@ static int ads7846_remove(struct spi_device *spi)
 
        ads784x_hwmon_unregister(spi, ts);
 
-       regulator_disable(ts->reg);
        regulator_put(ts->reg);
 
        if (!ts->get_pendown_state) {
index 7379fe1..b2744a6 100644 (file)
@@ -464,7 +464,7 @@ static int silead_ts_probe(struct i2c_client *client,
                return -ENODEV;
 
        /* Power GPIO pin */
-       data->gpio_power = gpiod_get_optional(dev, "power", GPIOD_OUT_LOW);
+       data->gpio_power = devm_gpiod_get_optional(dev, "power", GPIOD_OUT_LOW);
        if (IS_ERR(data->gpio_power)) {
                if (PTR_ERR(data->gpio_power) != -EPROBE_DEFER)
                        dev_err(dev, "Shutdown GPIO request failed\n");
index ce80117..641e887 100644 (file)
@@ -879,7 +879,7 @@ static void arm_smmu_cmdq_skip_err(struct arm_smmu_device *smmu)
         * We may have concurrent producers, so we need to be careful
         * not to touch any of the shadow cmdq state.
         */
-       queue_read(cmd, Q_ENT(q, idx), q->ent_dwords);
+       queue_read(cmd, Q_ENT(q, cons), q->ent_dwords);
        dev_err(smmu->dev, "skipping command in error state:\n");
        for (i = 0; i < ARRAY_SIZE(cmd); ++i)
                dev_err(smmu->dev, "\t0x%016llx\n", (unsigned long long)cmd[i]);
@@ -890,7 +890,7 @@ static void arm_smmu_cmdq_skip_err(struct arm_smmu_device *smmu)
                return;
        }
 
-       queue_write(cmd, Q_ENT(q, idx), q->ent_dwords);
+       queue_write(Q_ENT(q, cons), cmd, q->ent_dwords);
 }
 
 static void arm_smmu_cmdq_issue_cmd(struct arm_smmu_device *smmu,
@@ -1034,6 +1034,9 @@ static void arm_smmu_write_strtab_ent(struct arm_smmu_device *smmu, u32 sid,
                case STRTAB_STE_0_CFG_S2_TRANS:
                        ste_live = true;
                        break;
+               case STRTAB_STE_0_CFG_ABORT:
+                       if (disable_bypass)
+                               break;
                default:
                        BUG(); /* STE corruption */
                }
index 4f49fe2..2db74eb 100644 (file)
@@ -686,8 +686,7 @@ static struct iommu_gather_ops arm_smmu_gather_ops = {
 
 static irqreturn_t arm_smmu_context_fault(int irq, void *dev)
 {
-       int flags, ret;
-       u32 fsr, fsynr, resume;
+       u32 fsr, fsynr;
        unsigned long iova;
        struct iommu_domain *domain = dev;
        struct arm_smmu_domain *smmu_domain = to_smmu_domain(domain);
@@ -701,34 +700,15 @@ static irqreturn_t arm_smmu_context_fault(int irq, void *dev)
        if (!(fsr & FSR_FAULT))
                return IRQ_NONE;
 
-       if (fsr & FSR_IGN)
-               dev_err_ratelimited(smmu->dev,
-                                   "Unexpected context fault (fsr 0x%x)\n",
-                                   fsr);
-
        fsynr = readl_relaxed(cb_base + ARM_SMMU_CB_FSYNR0);
-       flags = fsynr & FSYNR0_WNR ? IOMMU_FAULT_WRITE : IOMMU_FAULT_READ;
-
        iova = readq_relaxed(cb_base + ARM_SMMU_CB_FAR);
-       if (!report_iommu_fault(domain, smmu->dev, iova, flags)) {
-               ret = IRQ_HANDLED;
-               resume = RESUME_RETRY;
-       } else {
-               dev_err_ratelimited(smmu->dev,
-                   "Unhandled context fault: iova=0x%08lx, fsynr=0x%x, cb=%d\n",
-                   iova, fsynr, cfg->cbndx);
-               ret = IRQ_NONE;
-               resume = RESUME_TERMINATE;
-       }
-
-       /* Clear the faulting FSR */
-       writel(fsr, cb_base + ARM_SMMU_CB_FSR);
 
-       /* Retry or terminate any stalled transactions */
-       if (fsr & FSR_SS)
-               writel_relaxed(resume, cb_base + ARM_SMMU_CB_RESUME);
+       dev_err_ratelimited(smmu->dev,
+       "Unhandled context fault: fsr=0x%x, iova=0x%08lx, fsynr=0x%x, cb=%d\n",
+                           fsr, iova, fsynr, cfg->cbndx);
 
-       return ret;
+       writel(fsr, cb_base + ARM_SMMU_CB_FSR);
+       return IRQ_HANDLED;
 }
 
 static irqreturn_t arm_smmu_global_fault(int irq, void *dev)
@@ -837,7 +817,7 @@ static void arm_smmu_init_context_bank(struct arm_smmu_domain *smmu_domain,
        }
 
        /* SCTLR */
-       reg = SCTLR_CFCFG | SCTLR_CFIE | SCTLR_CFRE | SCTLR_M | SCTLR_EAE_SBOP;
+       reg = SCTLR_CFIE | SCTLR_CFRE | SCTLR_M | SCTLR_EAE_SBOP;
        if (stage1)
                reg |= SCTLR_S1_ASIDPNE;
 #ifdef __BIG_ENDIAN
index 8c61399..def8ca1 100644 (file)
@@ -286,12 +286,14 @@ static int arm_v7s_pte_to_prot(arm_v7s_iopte pte, int lvl)
        int prot = IOMMU_READ;
        arm_v7s_iopte attr = pte >> ARM_V7S_ATTR_SHIFT(lvl);
 
-       if (attr & ARM_V7S_PTE_AP_RDONLY)
+       if (!(attr & ARM_V7S_PTE_AP_RDONLY))
                prot |= IOMMU_WRITE;
        if ((attr & (ARM_V7S_TEX_MASK << ARM_V7S_TEX_SHIFT)) == 0)
                prot |= IOMMU_MMIO;
        else if (pte & ARM_V7S_ATTR_C)
                prot |= IOMMU_CACHE;
+       if (pte & ARM_V7S_ATTR_XN(lvl))
+               prot |= IOMMU_NOEXEC;
 
        return prot;
 }
index 95a4ca6..849ad44 100644 (file)
@@ -760,7 +760,8 @@ static int bcache_device_init(struct bcache_device *d, unsigned block_size,
        if (!d->nr_stripes ||
            d->nr_stripes > INT_MAX ||
            d->nr_stripes > SIZE_MAX / sizeof(atomic_t)) {
-               pr_err("nr_stripes too large");
+               pr_err("nr_stripes too large or invalid: %u (start sector beyond end of disk?)",
+                       (unsigned)d->nr_stripes);
                return -ENOMEM;
        }
 
@@ -1820,7 +1821,7 @@ static int cache_alloc(struct cache *ca)
        free = roundup_pow_of_two(ca->sb.nbuckets) >> 10;
 
        if (!init_fifo(&ca->free[RESERVE_BTREE], 8, GFP_KERNEL) ||
-           !init_fifo(&ca->free[RESERVE_PRIO], prio_buckets(ca), GFP_KERNEL) ||
+           !init_fifo_exact(&ca->free[RESERVE_PRIO], prio_buckets(ca), GFP_KERNEL) ||
            !init_fifo(&ca->free[RESERVE_MOVINGGC], free, GFP_KERNEL) ||
            !init_fifo(&ca->free[RESERVE_NONE], free, GFP_KERNEL) ||
            !init_fifo(&ca->free_inc,   free << 2, GFP_KERNEL) ||
@@ -1844,7 +1845,7 @@ static int register_cache(struct cache_sb *sb, struct page *sb_page,
                                struct block_device *bdev, struct cache *ca)
 {
        char name[BDEVNAME_SIZE];
-       const char *err = NULL;
+       const char *err = NULL; /* must be set for any error case */
        int ret = 0;
 
        memcpy(&ca->sb, sb, sizeof(struct cache_sb));
@@ -1861,8 +1862,13 @@ static int register_cache(struct cache_sb *sb, struct page *sb_page,
                ca->discard = CACHE_DISCARD(&ca->sb);
 
        ret = cache_alloc(ca);
-       if (ret != 0)
+       if (ret != 0) {
+               if (ret == -ENOMEM)
+                       err = "cache_alloc(): -ENOMEM";
+               else
+                       err = "cache_alloc(): unknown error";
                goto err;
+       }
 
        if (kobject_add(&ca->kobj, &part_to_dev(bdev->bd_part)->kobj, "bcache")) {
                err = "error calling kobject_add";
index 4e9784b..eedba67 100644 (file)
@@ -181,7 +181,7 @@ struct crypt_config {
        u8 key[0];
 };
 
-#define MIN_IOS        16
+#define MIN_IOS        64
 
 static void clone_init(struct dm_crypt_io *, struct bio *);
 static void kcryptd_queue_crypt(struct dm_crypt_io *io);
index 97e446d..6a2e8dd 100644 (file)
@@ -289,15 +289,13 @@ static int flakey_map(struct dm_target *ti, struct bio *bio)
                pb->bio_submitted = true;
 
                /*
-                * Map reads as normal only if corrupt_bio_byte set.
+                * Error reads if neither corrupt_bio_byte or drop_writes are set.
+                * Otherwise, flakey_end_io() will decide if the reads should be modified.
                 */
                if (bio_data_dir(bio) == READ) {
-                       /* If flags were specified, only corrupt those that match. */
-                       if (fc->corrupt_bio_byte && (fc->corrupt_bio_rw == READ) &&
-                           all_corrupt_bio_flags_match(bio, fc))
-                               goto map_bio;
-                       else
+                       if (!fc->corrupt_bio_byte && !test_bit(DROP_WRITES, &fc->flags))
                                return -EIO;
+                       goto map_bio;
                }
 
                /*
@@ -334,14 +332,21 @@ static int flakey_end_io(struct dm_target *ti, struct bio *bio, int error)
        struct flakey_c *fc = ti->private;
        struct per_bio_data *pb = dm_per_bio_data(bio, sizeof(struct per_bio_data));
 
-       /*
-        * Corrupt successful READs while in down state.
-        */
        if (!error && pb->bio_submitted && (bio_data_dir(bio) == READ)) {
-               if (fc->corrupt_bio_byte)
+               if (fc->corrupt_bio_byte && (fc->corrupt_bio_rw == READ) &&
+                   all_corrupt_bio_flags_match(bio, fc)) {
+                       /*
+                        * Corrupt successful matching READs while in down state.
+                        */
                        corrupt_bio_data(bio, fc);
-               else
+
+               } else if (!test_bit(DROP_WRITES, &fc->flags)) {
+                       /*
+                        * Error read during the down_interval if drop_writes
+                        * wasn't configured.
+                        */
                        return -EIO;
+               }
        }
 
        return error;
index 4ca2d1d..07fc1ad 100644 (file)
@@ -291,9 +291,10 @@ static void header_from_disk(struct log_header_core *core, struct log_header_dis
        core->nr_regions = le64_to_cpu(disk->nr_regions);
 }
 
-static int rw_header(struct log_c *lc, int rw)
+static int rw_header(struct log_c *lc, int op)
 {
-       lc->io_req.bi_op = rw;
+       lc->io_req.bi_op = op;
+       lc->io_req.bi_op_flags = 0;
 
        return dm_io(&lc->io_req, 1, &lc->header_location, NULL);
 }
@@ -316,7 +317,7 @@ static int read_header(struct log_c *log)
 {
        int r;
 
-       r = rw_header(log, READ);
+       r = rw_header(log, REQ_OP_READ);
        if (r)
                return r;
 
@@ -630,7 +631,7 @@ static int disk_resume(struct dm_dirty_log *log)
        header_to_disk(&lc->header, lc->disk_header);
 
        /* write the new header */
-       r = rw_header(lc, WRITE);
+       r = rw_header(lc, REQ_OP_WRITE);
        if (!r) {
                r = flush_header(lc);
                if (r)
@@ -698,7 +699,7 @@ static int disk_flush(struct dm_dirty_log *log)
                        log_clear_bit(lc, lc->clean_bits, i);
        }
 
-       r = rw_header(lc, WRITE);
+       r = rw_header(lc, REQ_OP_WRITE);
        if (r)
                fail_log_device(lc);
        else {
index 1b9795d..8abde6b 100644 (file)
@@ -191,7 +191,6 @@ struct raid_dev {
 #define RT_FLAG_RS_BITMAP_LOADED       2
 #define RT_FLAG_UPDATE_SBS             3
 #define RT_FLAG_RESHAPE_RS             4
-#define RT_FLAG_KEEP_RS_FROZEN         5
 
 /* Array elements of 64 bit needed for rebuild/failed disk bits */
 #define DISKS_ARRAY_ELEMS ((MAX_RAID_DEVICES + (sizeof(uint64_t) * 8 - 1)) / sizeof(uint64_t) / 8)
@@ -861,6 +860,9 @@ static int validate_region_size(struct raid_set *rs, unsigned long region_size)
 {
        unsigned long min_region_size = rs->ti->len / (1 << 21);
 
+       if (rs_is_raid0(rs))
+               return 0;
+
        if (!region_size) {
                /*
                 * Choose a reasonable default.  All figures in sectors.
@@ -930,6 +932,8 @@ static int validate_raid_redundancy(struct raid_set *rs)
                        rebuild_cnt++;
 
        switch (rs->raid_type->level) {
+       case 0:
+               break;
        case 1:
                if (rebuild_cnt >= rs->md.raid_disks)
                        goto too_many;
@@ -2335,6 +2339,13 @@ static int analyse_superblocks(struct dm_target *ti, struct raid_set *rs)
                case 0:
                        break;
                default:
+                       /*
+                        * We have to keep any raid0 data/metadata device pairs or
+                        * the MD raid0 personality will fail to start the array.
+                        */
+                       if (rs_is_raid0(rs))
+                               continue;
+
                        dev = container_of(rdev, struct raid_dev, rdev);
                        if (dev->meta_dev)
                                dm_put_device(ti, dev->meta_dev);
@@ -2579,7 +2590,6 @@ static int rs_prepare_reshape(struct raid_set *rs)
                } else {
                        /* Process raid1 without delta_disks */
                        mddev->raid_disks = rs->raid_disks;
-                       set_bit(RT_FLAG_KEEP_RS_FROZEN, &rs->runtime_flags);
                        reshape = false;
                }
        } else {
@@ -2590,7 +2600,6 @@ static int rs_prepare_reshape(struct raid_set *rs)
        if (reshape) {
                set_bit(RT_FLAG_RESHAPE_RS, &rs->runtime_flags);
                set_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags);
-               set_bit(RT_FLAG_KEEP_RS_FROZEN, &rs->runtime_flags);
        } else if (mddev->raid_disks < rs->raid_disks)
                /* Create new superblocks and bitmaps, if any new disks */
                set_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags);
@@ -2902,7 +2911,6 @@ static int raid_ctr(struct dm_target *ti, unsigned int argc, char **argv)
                        goto bad;
 
                set_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags);
-               set_bit(RT_FLAG_KEEP_RS_FROZEN, &rs->runtime_flags);
                /* Takeover ain't recovery, so disable recovery */
                rs_setup_recovery(rs, MaxSector);
                rs_set_new(rs);
@@ -3386,21 +3394,28 @@ static void raid_postsuspend(struct dm_target *ti)
 {
        struct raid_set *rs = ti->private;
 
-       if (test_and_clear_bit(RT_FLAG_RS_RESUMED, &rs->runtime_flags)) {
-               if (!rs->md.suspended)
-                       mddev_suspend(&rs->md);
-               rs->md.ro = 1;
-       }
+       if (!rs->md.suspended)
+               mddev_suspend(&rs->md);
+
+       rs->md.ro = 1;
 }
 
 static void attempt_restore_of_faulty_devices(struct raid_set *rs)
 {
        int i;
-       uint64_t failed_devices, cleared_failed_devices = 0;
+       uint64_t cleared_failed_devices[DISKS_ARRAY_ELEMS];
        unsigned long flags;
+       bool cleared = false;
        struct dm_raid_superblock *sb;
+       struct mddev *mddev = &rs->md;
        struct md_rdev *r;
 
+       /* RAID personalities have to provide hot add/remove methods or we need to bail out. */
+       if (!mddev->pers || !mddev->pers->hot_add_disk || !mddev->pers->hot_remove_disk)
+               return;
+
+       memset(cleared_failed_devices, 0, sizeof(cleared_failed_devices));
+
        for (i = 0; i < rs->md.raid_disks; i++) {
                r = &rs->dev[i].rdev;
                if (test_bit(Faulty, &r->flags) && r->sb_page &&
@@ -3420,7 +3435,7 @@ static void attempt_restore_of_faulty_devices(struct raid_set *rs)
                         * ourselves.
                         */
                        if ((r->raid_disk >= 0) &&
-                           (r->mddev->pers->hot_remove_disk(r->mddev, r) != 0))
+                           (mddev->pers->hot_remove_disk(mddev, r) != 0))
                                /* Failed to revive this device, try next */
                                continue;
 
@@ -3430,22 +3445,30 @@ static void attempt_restore_of_faulty_devices(struct raid_set *rs)
                        clear_bit(Faulty, &r->flags);
                        clear_bit(WriteErrorSeen, &r->flags);
                        clear_bit(In_sync, &r->flags);
-                       if (r->mddev->pers->hot_add_disk(r->mddev, r)) {
+                       if (mddev->pers->hot_add_disk(mddev, r)) {
                                r->raid_disk = -1;
                                r->saved_raid_disk = -1;
                                r->flags = flags;
                        } else {
                                r->recovery_offset = 0;
-                               cleared_failed_devices |= 1 << i;
+                               set_bit(i, (void *) cleared_failed_devices);
+                               cleared = true;
                        }
                }
        }
-       if (cleared_failed_devices) {
+
+       /* If any failed devices could be cleared, update all sbs failed_devices bits */
+       if (cleared) {
+               uint64_t failed_devices[DISKS_ARRAY_ELEMS];
+
                rdev_for_each(r, &rs->md) {
                        sb = page_address(r->sb_page);
-                       failed_devices = le64_to_cpu(sb->failed_devices);
-                       failed_devices &= ~cleared_failed_devices;
-                       sb->failed_devices = cpu_to_le64(failed_devices);
+                       sb_retrieve_failed_devices(sb, failed_devices);
+
+                       for (i = 0; i < DISKS_ARRAY_ELEMS; i++)
+                               failed_devices[i] &= ~cleared_failed_devices[i];
+
+                       sb_update_failed_devices(sb, failed_devices);
                }
        }
 }
@@ -3610,26 +3633,15 @@ static void raid_resume(struct dm_target *ti)
                 * devices are reachable again.
                 */
                attempt_restore_of_faulty_devices(rs);
-       } else {
-               mddev->ro = 0;
-               mddev->in_sync = 0;
+       }
 
-               /*
-                * When passing in flags to the ctr, we expect userspace
-                * to reset them because they made it to the superblocks
-                * and reload the mapping anyway.
-                *
-                * -> only unfreeze recovery in case of a table reload or
-                *    we'll have a bogus recovery/reshape position
-                *    retrieved from the superblock by the ctr because
-                *    the ongoing recovery/reshape will change it after read.
-                */
-               if (!test_bit(RT_FLAG_KEEP_RS_FROZEN, &rs->runtime_flags))
-                       clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
+       mddev->ro = 0;
+       mddev->in_sync = 0;
 
-               if (mddev->suspended)
-                       mddev_resume(mddev);
-       }
+       clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
+
+       if (mddev->suspended)
+               mddev_resume(mddev);
 }
 
 static struct target_type raid_target = {
index 4ace1da..6c25213 100644 (file)
@@ -210,14 +210,17 @@ static struct dm_path *rr_select_path(struct path_selector *ps, size_t nr_bytes)
        struct path_info *pi = NULL;
        struct dm_path *current_path = NULL;
 
+       local_irq_save(flags);
        current_path = *this_cpu_ptr(s->current_path);
        if (current_path) {
                percpu_counter_dec(&s->repeat_count);
-               if (percpu_counter_read_positive(&s->repeat_count) > 0)
+               if (percpu_counter_read_positive(&s->repeat_count) > 0) {
+                       local_irq_restore(flags);
                        return current_path;
+               }
        }
 
-       spin_lock_irqsave(&s->lock, flags);
+       spin_lock(&s->lock);
        if (!list_empty(&s->valid_paths)) {
                pi = list_entry(s->valid_paths.next, struct path_info, list);
                list_move_tail(&pi->list, &s->valid_paths);
index 48a5dd7..2206d44 100644 (file)
@@ -1726,6 +1726,7 @@ static u8 mmc_blk_prep_packed_list(struct mmc_queue *mq, struct request *req)
                        break;
 
                if (req_op(next) == REQ_OP_DISCARD ||
+                   req_op(next) == REQ_OP_SECURE_ERASE ||
                    req_op(next) == REQ_OP_FLUSH)
                        break;
 
@@ -2150,6 +2151,7 @@ static int mmc_blk_issue_rq(struct mmc_queue *mq, struct request *req)
        struct mmc_card *card = md->queue.card;
        struct mmc_host *host = card->host;
        unsigned long flags;
+       bool req_is_special = mmc_req_is_special(req);
 
        if (req && !mq->mqrq_prev->req)
                /* claim host only for the first request */
@@ -2190,8 +2192,7 @@ static int mmc_blk_issue_rq(struct mmc_queue *mq, struct request *req)
        }
 
 out:
-       if ((!req && !(mq->flags & MMC_QUEUE_NEW_REQUEST)) ||
-           mmc_req_is_special(req))
+       if ((!req && !(mq->flags & MMC_QUEUE_NEW_REQUEST)) || req_is_special)
                /*
                 * Release host when there are no more requests
                 * and after special request(discard, flush) is done.
index bf14642..7080572 100644 (file)
@@ -33,7 +33,8 @@ static int mmc_prep_request(struct request_queue *q, struct request *req)
        /*
         * We only like normal block requests and discards.
         */
-       if (req->cmd_type != REQ_TYPE_FS && req_op(req) != REQ_OP_DISCARD) {
+       if (req->cmd_type != REQ_TYPE_FS && req_op(req) != REQ_OP_DISCARD &&
+           req_op(req) != REQ_OP_SECURE_ERASE) {
                blk_dump_rq_flags(req, "MMC bad request");
                return BLKPREP_KILL;
        }
@@ -64,6 +65,8 @@ static int mmc_queue_thread(void *d)
                spin_unlock_irq(q->queue_lock);
 
                if (req || mq->mqrq_prev->req) {
+                       bool req_is_special = mmc_req_is_special(req);
+
                        set_current_state(TASK_RUNNING);
                        mq->issue_fn(mq, req);
                        cond_resched();
@@ -79,7 +82,7 @@ static int mmc_queue_thread(void *d)
                         * has been finished. Do not assign it to previous
                         * request.
                         */
-                       if (mmc_req_is_special(req))
+                       if (req_is_special)
                                mq->mqrq_cur->req = NULL;
 
                        mq->mqrq_prev->brq.mrq.data = NULL;
index d625311..fee5e12 100644 (file)
@@ -4,7 +4,9 @@
 static inline bool mmc_req_is_special(struct request *req)
 {
        return req &&
-               (req_op(req) == REQ_OP_FLUSH || req_op(req) == REQ_OP_DISCARD);
+               (req_op(req) == REQ_OP_FLUSH ||
+                req_op(req) == REQ_OP_DISCARD ||
+                req_op(req) == REQ_OP_SECURE_ERASE);
 }
 
 struct request;
index 7ff2e82..2feacc7 100644 (file)
@@ -81,10 +81,12 @@ EXPORT_SYMBOL_GPL(nvme_cancel_request);
 bool nvme_change_ctrl_state(struct nvme_ctrl *ctrl,
                enum nvme_ctrl_state new_state)
 {
-       enum nvme_ctrl_state old_state = ctrl->state;
+       enum nvme_ctrl_state old_state;
        bool changed = false;
 
        spin_lock_irq(&ctrl->lock);
+
+       old_state = ctrl->state;
        switch (new_state) {
        case NVME_CTRL_LIVE:
                switch (old_state) {
@@ -140,11 +142,12 @@ bool nvme_change_ctrl_state(struct nvme_ctrl *ctrl,
        default:
                break;
        }
-       spin_unlock_irq(&ctrl->lock);
 
        if (changed)
                ctrl->state = new_state;
 
+       spin_unlock_irq(&ctrl->lock);
+
        return changed;
 }
 EXPORT_SYMBOL_GPL(nvme_change_ctrl_state);
@@ -608,7 +611,7 @@ int nvme_get_features(struct nvme_ctrl *dev, unsigned fid, unsigned nsid,
 
        ret = __nvme_submit_sync_cmd(dev->admin_q, &c, &cqe, NULL, 0, 0,
                        NVME_QID_ANY, 0, 0);
-       if (ret >= 0)
+       if (ret >= 0 && result)
                *result = le32_to_cpu(cqe.result);
        return ret;
 }
@@ -628,7 +631,7 @@ int nvme_set_features(struct nvme_ctrl *dev, unsigned fid, unsigned dword11,
 
        ret = __nvme_submit_sync_cmd(dev->admin_q, &c, &cqe, NULL, 0, 0,
                        NVME_QID_ANY, 0, 0);
-       if (ret >= 0)
+       if (ret >= 0 && result)
                *result = le32_to_cpu(cqe.result);
        return ret;
 }
index 7792266..3ce6953 100644 (file)
@@ -1631,8 +1631,7 @@ static int __of_parse_phandle_with_args(const struct device_node *np,
         */
 
  err:
-       if (it.node)
-               of_node_put(it.node);
+       of_node_put(it.node);
        return rc;
 }
 
@@ -2343,20 +2342,13 @@ struct device_node *of_graph_get_endpoint_by_regs(
        const struct device_node *parent, int port_reg, int reg)
 {
        struct of_endpoint endpoint;
-       struct device_node *node, *prev_node = NULL;
-
-       while (1) {
-               node = of_graph_get_next_endpoint(parent, prev_node);
-               of_node_put(prev_node);
-               if (!node)
-                       break;
+       struct device_node *node = NULL;
 
+       for_each_endpoint_of_node(parent, node) {
                of_graph_parse_endpoint(node, &endpoint);
                if (((port_reg == -1) || (endpoint.port == port_reg)) &&
                        ((reg == -1) || (endpoint.id == reg)))
                        return node;
-
-               prev_node = node;
        }
 
        return NULL;
index 55f1b83..085c638 100644 (file)
@@ -517,7 +517,7 @@ static void *__unflatten_device_tree(const void *blob,
                pr_warning("End of tree marker overwritten: %08x\n",
                           be32_to_cpup(mem + size));
 
-       if (detached) {
+       if (detached && mynodes) {
                of_node_set_flag(*mynodes, OF_DETACHED);
                pr_debug("unflattened tree is detached\n");
        }
index 89a71c6..a2e68f7 100644 (file)
@@ -544,12 +544,15 @@ void __init of_irq_init(const struct of_device_id *matches)
 
                        list_del(&desc->list);
 
+                       of_node_set_flag(desc->dev, OF_POPULATED);
+
                        pr_debug("of_irq_init: init %s (%p), parent %p\n",
                                 desc->dev->full_name,
                                 desc->dev, desc->interrupt_parent);
                        ret = desc->irq_init_cb(desc->dev,
                                                desc->interrupt_parent);
                        if (ret) {
+                               of_node_clear_flag(desc->dev, OF_POPULATED);
                                kfree(desc);
                                continue;
                        }
@@ -559,8 +562,6 @@ void __init of_irq_init(const struct of_device_id *matches)
                         * its children can get processed in a subsequent pass.
                         */
                        list_add_tail(&desc->list, &intc_parent_list);
-
-                       of_node_set_flag(desc->dev, OF_POPULATED);
                }
 
                /* Get the next pending parent that might have children */
index 8aa1976..f39ccd5 100644 (file)
@@ -497,6 +497,7 @@ int of_platform_default_populate(struct device_node *root,
 }
 EXPORT_SYMBOL_GPL(of_platform_default_populate);
 
+#ifndef CONFIG_PPC
 static int __init of_platform_default_populate_init(void)
 {
        struct device_node *node;
@@ -521,6 +522,7 @@ static int __init of_platform_default_populate_init(void)
        return 0;
 }
 arch_initcall_sync(of_platform_default_populate_init);
+#endif
 
 static int of_platform_device_destroy(struct device *dev, void *data)
 {
index eafa613..98f1222 100644 (file)
@@ -1069,7 +1069,7 @@ static int __pci_enable_msi_range(struct pci_dev *dev, int minvec, int maxvec,
                nvec = maxvec;
 
        for (;;) {
-               if (!(flags & PCI_IRQ_NOAFFINITY)) {
+               if (flags & PCI_IRQ_AFFINITY) {
                        dev->irq_affinity = irq_create_affinity_mask(&nvec);
                        if (nvec < minvec)
                                return -ENOSPC;
@@ -1105,7 +1105,7 @@ static int __pci_enable_msi_range(struct pci_dev *dev, int minvec, int maxvec,
  **/
 int pci_enable_msi_range(struct pci_dev *dev, int minvec, int maxvec)
 {
-       return __pci_enable_msi_range(dev, minvec, maxvec, PCI_IRQ_NOAFFINITY);
+       return __pci_enable_msi_range(dev, minvec, maxvec, 0);
 }
 EXPORT_SYMBOL(pci_enable_msi_range);
 
@@ -1120,7 +1120,7 @@ static int __pci_enable_msix_range(struct pci_dev *dev,
                return -ERANGE;
 
        for (;;) {
-               if (!(flags & PCI_IRQ_NOAFFINITY)) {
+               if (flags & PCI_IRQ_AFFINITY) {
                        dev->irq_affinity = irq_create_affinity_mask(&nvec);
                        if (nvec < minvec)
                                return -ENOSPC;
@@ -1160,8 +1160,7 @@ static int __pci_enable_msix_range(struct pci_dev *dev,
 int pci_enable_msix_range(struct pci_dev *dev, struct msix_entry *entries,
                int minvec, int maxvec)
 {
-       return __pci_enable_msix_range(dev, entries, minvec, maxvec,
-                       PCI_IRQ_NOAFFINITY);
+       return __pci_enable_msix_range(dev, entries, minvec, maxvec, 0);
 }
 EXPORT_SYMBOL(pci_enable_msix_range);
 
@@ -1187,22 +1186,25 @@ int pci_alloc_irq_vectors(struct pci_dev *dev, unsigned int min_vecs,
 {
        int vecs = -ENOSPC;
 
-       if (!(flags & PCI_IRQ_NOMSIX)) {
+       if (flags & PCI_IRQ_MSIX) {
                vecs = __pci_enable_msix_range(dev, NULL, min_vecs, max_vecs,
                                flags);
                if (vecs > 0)
                        return vecs;
        }
 
-       if (!(flags & PCI_IRQ_NOMSI)) {
+       if (flags & PCI_IRQ_MSI) {
                vecs = __pci_enable_msi_range(dev, min_vecs, max_vecs, flags);
                if (vecs > 0)
                        return vecs;
        }
 
        /* use legacy irq if allowed */
-       if (!(flags & PCI_IRQ_NOLEGACY) && min_vecs == 1)
+       if ((flags & PCI_IRQ_LEGACY) && min_vecs == 1) {
+               pci_intx(dev, 1);
                return 1;
+       }
+
        return vecs;
 }
 EXPORT_SYMBOL(pci_alloc_irq_vectors);
index b381b37..5648b71 100644 (file)
@@ -63,7 +63,7 @@ static int ioctl_send_fib(struct aac_dev * dev, void __user *arg)
        struct fib *fibptr;
        struct hw_fib * hw_fib = (struct hw_fib *)0;
        dma_addr_t hw_fib_pa = (dma_addr_t)0LL;
-       unsigned size;
+       unsigned int size, osize;
        int retval;
 
        if (dev->in_reset) {
@@ -87,7 +87,8 @@ static int ioctl_send_fib(struct aac_dev * dev, void __user *arg)
         *      will not overrun the buffer when we copy the memory. Return
         *      an error if we would.
         */
-       size = le16_to_cpu(kfib->header.Size) + sizeof(struct aac_fibhdr);
+       osize = size = le16_to_cpu(kfib->header.Size) +
+               sizeof(struct aac_fibhdr);
        if (size < le16_to_cpu(kfib->header.SenderSize))
                size = le16_to_cpu(kfib->header.SenderSize);
        if (size > dev->max_fib_size) {
@@ -118,6 +119,14 @@ static int ioctl_send_fib(struct aac_dev * dev, void __user *arg)
                goto cleanup;
        }
 
+       /* Sanity check the second copy */
+       if ((osize != le16_to_cpu(kfib->header.Size) +
+               sizeof(struct aac_fibhdr))
+               || (size < le16_to_cpu(kfib->header.SenderSize))) {
+               retval = -EINVAL;
+               goto cleanup;
+       }
+
        if (kfib->header.Command == cpu_to_le16(TakeABreakPt)) {
                aac_adapter_interrupt(dev);
                /*
index a569c65..dcf3653 100644 (file)
@@ -2923,7 +2923,7 @@ static int fcoe_ctlr_vlan_recv(struct fcoe_ctlr *fip, struct sk_buff *skb)
        mutex_unlock(&fip->ctlr_mutex);
 
 drop:
-       kfree(skb);
+       kfree_skb(skb);
        return rc;
 }
 
index 2dab3dc..c1ed25a 100644 (file)
@@ -5037,7 +5037,7 @@ static int megasas_init_fw(struct megasas_instance *instance)
        /* Find first memory bar */
        bar_list = pci_select_bars(instance->pdev, IORESOURCE_MEM);
        instance->bar = find_first_bit(&bar_list, sizeof(unsigned long));
-       if (pci_request_selected_regions(instance->pdev, instance->bar,
+       if (pci_request_selected_regions(instance->pdev, 1<<instance->bar,
                                         "megasas: LSI")) {
                dev_printk(KERN_DEBUG, &instance->pdev->dev, "IO memory region busy!\n");
                return -EBUSY;
@@ -5339,7 +5339,7 @@ fail_ready_state:
        iounmap(instance->reg_set);
 
       fail_ioremap:
-       pci_release_selected_regions(instance->pdev, instance->bar);
+       pci_release_selected_regions(instance->pdev, 1<<instance->bar);
 
        return -EINVAL;
 }
@@ -5360,7 +5360,7 @@ static void megasas_release_mfi(struct megasas_instance *instance)
 
        iounmap(instance->reg_set);
 
-       pci_release_selected_regions(instance->pdev, instance->bar);
+       pci_release_selected_regions(instance->pdev, 1<<instance->bar);
 }
 
 /**
index ec83754..52d8bbf 100644 (file)
@@ -2603,7 +2603,7 @@ megasas_release_fusion(struct megasas_instance *instance)
 
        iounmap(instance->reg_set);
 
-       pci_release_selected_regions(instance->pdev, instance->bar);
+       pci_release_selected_regions(instance->pdev, 1<<instance->bar);
 }
 
 /**
index 751f13e..750f82c 100644 (file)
@@ -2188,6 +2188,17 @@ mpt3sas_base_map_resources(struct MPT3SAS_ADAPTER *ioc)
        } else
                ioc->msix96_vector = 0;
 
+       if (ioc->is_warpdrive) {
+               ioc->reply_post_host_index[0] = (resource_size_t __iomem *)
+                   &ioc->chip->ReplyPostHostIndex;
+
+               for (i = 1; i < ioc->cpu_msix_table_sz; i++)
+                       ioc->reply_post_host_index[i] =
+                       (resource_size_t __iomem *)
+                       ((u8 __iomem *)&ioc->chip->Doorbell + (0x4000 + ((i - 1)
+                       * 4)));
+       }
+
        list_for_each_entry(reply_q, &ioc->reply_queue_list, list)
                pr_info(MPT3SAS_FMT "%s: IRQ %d\n",
                    reply_q->name,  ((ioc->msix_enable) ? "PCI-MSI-X enabled" :
@@ -5280,17 +5291,6 @@ mpt3sas_base_attach(struct MPT3SAS_ADAPTER *ioc)
        if (r)
                goto out_free_resources;
 
-       if (ioc->is_warpdrive) {
-               ioc->reply_post_host_index[0] = (resource_size_t __iomem *)
-                   &ioc->chip->ReplyPostHostIndex;
-
-               for (i = 1; i < ioc->cpu_msix_table_sz; i++)
-                       ioc->reply_post_host_index[i] =
-                       (resource_size_t __iomem *)
-                       ((u8 __iomem *)&ioc->chip->Doorbell + (0x4000 + ((i - 1)
-                       * 4)));
-       }
-
        pci_set_drvdata(ioc->pdev, ioc->shost);
        r = _base_get_ioc_facts(ioc, CAN_SLEEP);
        if (r)
index 53ef1cb..0e8601a 100644 (file)
@@ -778,6 +778,8 @@ static void ses_intf_remove_enclosure(struct scsi_device *sdev)
        if (!edev)
                return;
 
+       enclosure_unregister(edev);
+
        ses_dev = edev->scratch;
        edev->scratch = NULL;
 
@@ -789,7 +791,6 @@ static void ses_intf_remove_enclosure(struct scsi_device *sdev)
        kfree(edev->component[0].scratch);
 
        put_device(&edev->edev);
-       enclosure_unregister(edev);
 }
 
 static void ses_intf_remove(struct device *cdev,
index 3788ed7..a32b417 100644 (file)
@@ -740,12 +740,22 @@ static int cpufreq_power2state(struct thermal_cooling_device *cdev,
 }
 
 /* Bind cpufreq callbacks to thermal cooling device ops */
+
 static struct thermal_cooling_device_ops cpufreq_cooling_ops = {
        .get_max_state = cpufreq_get_max_state,
        .get_cur_state = cpufreq_get_cur_state,
        .set_cur_state = cpufreq_set_cur_state,
 };
 
+static struct thermal_cooling_device_ops cpufreq_power_cooling_ops = {
+       .get_max_state          = cpufreq_get_max_state,
+       .get_cur_state          = cpufreq_get_cur_state,
+       .set_cur_state          = cpufreq_set_cur_state,
+       .get_requested_power    = cpufreq_get_requested_power,
+       .state2power            = cpufreq_state2power,
+       .power2state            = cpufreq_power2state,
+};
+
 /* Notifier for cpufreq policy change */
 static struct notifier_block thermal_cpufreq_notifier_block = {
        .notifier_call = cpufreq_thermal_notifier,
@@ -795,6 +805,7 @@ __cpufreq_cooling_register(struct device_node *np,
        struct cpumask temp_mask;
        unsigned int freq, i, num_cpus;
        int ret;
+       struct thermal_cooling_device_ops *cooling_ops;
 
        cpumask_and(&temp_mask, clip_cpus, cpu_online_mask);
        policy = cpufreq_cpu_get(cpumask_first(&temp_mask));
@@ -850,10 +861,6 @@ __cpufreq_cooling_register(struct device_node *np,
        cpumask_copy(&cpufreq_dev->allowed_cpus, clip_cpus);
 
        if (capacitance) {
-               cpufreq_cooling_ops.get_requested_power =
-                       cpufreq_get_requested_power;
-               cpufreq_cooling_ops.state2power = cpufreq_state2power;
-               cpufreq_cooling_ops.power2state = cpufreq_power2state;
                cpufreq_dev->plat_get_static_power = plat_static_func;
 
                ret = build_dyn_power_table(cpufreq_dev, capacitance);
@@ -861,6 +868,10 @@ __cpufreq_cooling_register(struct device_node *np,
                        cool_dev = ERR_PTR(ret);
                        goto free_table;
                }
+
+               cooling_ops = &cpufreq_power_cooling_ops;
+       } else {
+               cooling_ops = &cpufreq_cooling_ops;
        }
 
        ret = get_idr(&cpufreq_idr, &cpufreq_dev->id);
@@ -885,7 +896,7 @@ __cpufreq_cooling_register(struct device_node *np,
                 cpufreq_dev->id);
 
        cool_dev = thermal_of_cooling_device_register(np, dev_name, cpufreq_dev,
-                                                     &cpufreq_cooling_ops);
+                                                     cooling_ops);
        if (IS_ERR(cool_dev))
                goto remove_idr;
 
index c5547bd..e473548 100644 (file)
@@ -471,8 +471,6 @@ MODULE_DEVICE_TABLE(of, of_imx_thermal_match);
 
 static int imx_thermal_probe(struct platform_device *pdev)
 {
-       const struct of_device_id *of_id =
-               of_match_device(of_imx_thermal_match, &pdev->dev);
        struct imx_thermal_data *data;
        struct regmap *map;
        int measure_freq;
@@ -490,7 +488,7 @@ static int imx_thermal_probe(struct platform_device *pdev)
        }
        data->tempmon = map;
 
-       data->socdata = of_id->data;
+       data->socdata = of_device_get_match_data(&pdev->dev);
 
        /* make sure the IRQ flag is clear before enabling irq on i.MX6SX */
        if (data->socdata->version == TEMPMON_IMX6SX) {
index a578cd2..1891f34 100644 (file)
@@ -225,7 +225,6 @@ static struct platform_driver int3406_thermal_driver = {
        .remove = int3406_thermal_remove,
        .driver = {
                   .name = "int3406 thermal",
-                  .owner = THIS_MODULE,
                   .acpi_match_table = int3406_thermal_match,
                   },
 };
index 7191230..0f3f62e 100644 (file)
@@ -1354,7 +1354,6 @@ made_compressed_probe:
        spin_lock_init(&acm->write_lock);
        spin_lock_init(&acm->read_lock);
        mutex_init(&acm->mutex);
-       acm->rx_endpoint = usb_rcvbulkpipe(usb_dev, epread->bEndpointAddress);
        acm->is_int_ep = usb_endpoint_xfer_int(epread);
        if (acm->is_int_ep)
                acm->bInterval = epread->bInterval;
@@ -1394,14 +1393,14 @@ made_compressed_probe:
                urb->transfer_dma = rb->dma;
                if (acm->is_int_ep) {
                        usb_fill_int_urb(urb, acm->dev,
-                                        acm->rx_endpoint,
+                                        usb_rcvintpipe(usb_dev, epread->bEndpointAddress),
                                         rb->base,
                                         acm->readsize,
                                         acm_read_bulk_callback, rb,
                                         acm->bInterval);
                } else {
                        usb_fill_bulk_urb(urb, acm->dev,
-                                         acm->rx_endpoint,
+                                         usb_rcvbulkpipe(usb_dev, epread->bEndpointAddress),
                                          rb->base,
                                          acm->readsize,
                                          acm_read_bulk_callback, rb);
index 05ce308..1f1eabf 100644 (file)
@@ -96,7 +96,6 @@ struct acm {
        struct acm_rb read_buffers[ACM_NR];
        struct acm_wb *putbuffer;                       /* for acm_tty_put_char() */
        int rx_buflimit;
-       int rx_endpoint;
        spinlock_t read_lock;
        int write_used;                                 /* number of non-empty write buffers */
        int transmitting;
index 31ccdcc..0511631 100644 (file)
@@ -171,6 +171,31 @@ static void usb_parse_ss_endpoint_companion(struct device *ddev, int cfgno,
                                                        ep, buffer, size);
 }
 
+static const unsigned short low_speed_maxpacket_maxes[4] = {
+       [USB_ENDPOINT_XFER_CONTROL] = 8,
+       [USB_ENDPOINT_XFER_ISOC] = 0,
+       [USB_ENDPOINT_XFER_BULK] = 0,
+       [USB_ENDPOINT_XFER_INT] = 8,
+};
+static const unsigned short full_speed_maxpacket_maxes[4] = {
+       [USB_ENDPOINT_XFER_CONTROL] = 64,
+       [USB_ENDPOINT_XFER_ISOC] = 1023,
+       [USB_ENDPOINT_XFER_BULK] = 64,
+       [USB_ENDPOINT_XFER_INT] = 64,
+};
+static const unsigned short high_speed_maxpacket_maxes[4] = {
+       [USB_ENDPOINT_XFER_CONTROL] = 64,
+       [USB_ENDPOINT_XFER_ISOC] = 1024,
+       [USB_ENDPOINT_XFER_BULK] = 512,
+       [USB_ENDPOINT_XFER_INT] = 1023,
+};
+static const unsigned short super_speed_maxpacket_maxes[4] = {
+       [USB_ENDPOINT_XFER_CONTROL] = 512,
+       [USB_ENDPOINT_XFER_ISOC] = 1024,
+       [USB_ENDPOINT_XFER_BULK] = 1024,
+       [USB_ENDPOINT_XFER_INT] = 1024,
+};
+
 static int usb_parse_endpoint(struct device *ddev, int cfgno, int inum,
     int asnum, struct usb_host_interface *ifp, int num_ep,
     unsigned char *buffer, int size)
@@ -179,6 +204,8 @@ static int usb_parse_endpoint(struct device *ddev, int cfgno, int inum,
        struct usb_endpoint_descriptor *d;
        struct usb_host_endpoint *endpoint;
        int n, i, j, retval;
+       unsigned int maxp;
+       const unsigned short *maxpacket_maxes;
 
        d = (struct usb_endpoint_descriptor *) buffer;
        buffer += d->bLength;
@@ -286,6 +313,42 @@ static int usb_parse_endpoint(struct device *ddev, int cfgno, int inum,
                        endpoint->desc.wMaxPacketSize = cpu_to_le16(8);
        }
 
+       /* Validate the wMaxPacketSize field */
+       maxp = usb_endpoint_maxp(&endpoint->desc);
+
+       /* Find the highest legal maxpacket size for this endpoint */
+       i = 0;          /* additional transactions per microframe */
+       switch (to_usb_device(ddev)->speed) {
+       case USB_SPEED_LOW:
+               maxpacket_maxes = low_speed_maxpacket_maxes;
+               break;
+       case USB_SPEED_FULL:
+               maxpacket_maxes = full_speed_maxpacket_maxes;
+               break;
+       case USB_SPEED_HIGH:
+               /* Bits 12..11 are allowed only for HS periodic endpoints */
+               if (usb_endpoint_xfer_int(d) || usb_endpoint_xfer_isoc(d)) {
+                       i = maxp & (BIT(12) | BIT(11));
+                       maxp &= ~i;
+               }
+               /* fallthrough */
+       default:
+               maxpacket_maxes = high_speed_maxpacket_maxes;
+               break;
+       case USB_SPEED_SUPER:
+       case USB_SPEED_SUPER_PLUS:
+               maxpacket_maxes = super_speed_maxpacket_maxes;
+               break;
+       }
+       j = maxpacket_maxes[usb_endpoint_type(&endpoint->desc)];
+
+       if (maxp > j) {
+               dev_warn(ddev, "config %d interface %d altsetting %d endpoint 0x%X has invalid maxpacket %d, setting to %d\n",
+                   cfgno, inum, asnum, d->bEndpointAddress, maxp, j);
+               maxp = j;
+               endpoint->desc.wMaxPacketSize = cpu_to_le16(i | maxp);
+       }
+
        /*
         * Some buggy high speed devices have bulk endpoints using
         * maxpacket sizes other than 512.  High speed HCDs may not
@@ -293,9 +356,6 @@ static int usb_parse_endpoint(struct device *ddev, int cfgno, int inum,
         */
        if (to_usb_device(ddev)->speed == USB_SPEED_HIGH
                        && usb_endpoint_xfer_bulk(d)) {
-               unsigned maxp;
-
-               maxp = usb_endpoint_maxp(&endpoint->desc) & 0x07ff;
                if (maxp != 512)
                        dev_warn(ddev, "config %d interface %d altsetting %d "
                                "bulk endpoint 0x%X has invalid maxpacket %d\n",
index e9f5043..e6a6d67 100644 (file)
@@ -241,7 +241,8 @@ static int usbdev_mmap(struct file *file, struct vm_area_struct *vma)
                goto error_decrease_mem;
        }
 
-       mem = usb_alloc_coherent(ps->dev, size, GFP_USER, &dma_handle);
+       mem = usb_alloc_coherent(ps->dev, size, GFP_USER | __GFP_NOWARN,
+                       &dma_handle);
        if (!mem) {
                ret = -ENOMEM;
                goto error_free_usbm;
@@ -2582,7 +2583,9 @@ static unsigned int usbdev_poll(struct file *file,
        if (file->f_mode & FMODE_WRITE && !list_empty(&ps->async_completed))
                mask |= POLLOUT | POLLWRNORM;
        if (!connected(ps))
-               mask |= POLLERR | POLLHUP;
+               mask |= POLLHUP;
+       if (list_empty(&ps->list))
+               mask |= POLLERR;
        return mask;
 }
 
index bee1351..1d5fc32 100644 (file)
@@ -1052,14 +1052,11 @@ static void hub_activate(struct usb_hub *hub, enum hub_activation_type type)
 
        /* Continue a partial initialization */
        if (type == HUB_INIT2 || type == HUB_INIT3) {
-               device_lock(hub->intfdev);
+               device_lock(&hdev->dev);
 
                /* Was the hub disconnected while we were waiting? */
-               if (hub->disconnected) {
-                       device_unlock(hub->intfdev);
-                       kref_put(&hub->kref, hub_release);
-                       return;
-               }
+               if (hub->disconnected)
+                       goto disconnected;
                if (type == HUB_INIT2)
                        goto init2;
                goto init3;
@@ -1262,7 +1259,7 @@ static void hub_activate(struct usb_hub *hub, enum hub_activation_type type)
                        queue_delayed_work(system_power_efficient_wq,
                                        &hub->init_work,
                                        msecs_to_jiffies(delay));
-                       device_unlock(hub->intfdev);
+                       device_unlock(&hdev->dev);
                        return;         /* Continues at init3: below */
                } else {
                        msleep(delay);
@@ -1281,12 +1278,12 @@ static void hub_activate(struct usb_hub *hub, enum hub_activation_type type)
        /* Scan all ports that need attention */
        kick_hub_wq(hub);
 
-       /* Allow autosuspend if it was suppressed */
-       if (type <= HUB_INIT3)
+       if (type == HUB_INIT2 || type == HUB_INIT3) {
+               /* Allow autosuspend if it was suppressed */
+ disconnected:
                usb_autopm_put_interface_async(to_usb_interface(hub->intfdev));
-
-       if (type == HUB_INIT2 || type == HUB_INIT3)
-               device_unlock(hub->intfdev);
+               device_unlock(&hdev->dev);
+       }
 
        kref_put(&hub->kref, hub_release);
 }
@@ -1315,8 +1312,6 @@ static void hub_quiesce(struct usb_hub *hub, enum hub_quiescing_type type)
        struct usb_device *hdev = hub->hdev;
        int i;
 
-       cancel_delayed_work_sync(&hub->init_work);
-
        /* hub_wq and related activity won't re-trigger */
        hub->quiescing = 1;
 
index 9743353..e56d59b 100644 (file)
@@ -61,6 +61,7 @@ static int dwc3_of_simple_probe(struct platform_device *pdev)
        if (!simple->clks)
                return -ENOMEM;
 
+       platform_set_drvdata(pdev, simple);
        simple->dev = dev;
 
        for (i = 0; i < simple->num_clocks; i++) {
index 45f5a23..2eb84d6 100644 (file)
@@ -37,6 +37,7 @@
 #define PCI_DEVICE_ID_INTEL_BXT                        0x0aaa
 #define PCI_DEVICE_ID_INTEL_BXT_M              0x1aaa
 #define PCI_DEVICE_ID_INTEL_APL                        0x5aaa
+#define PCI_DEVICE_ID_INTEL_KBP                        0xa2b0
 
 static const struct acpi_gpio_params reset_gpios = { 0, 0, false };
 static const struct acpi_gpio_params cs_gpios = { 1, 0, false };
@@ -227,6 +228,7 @@ static const struct pci_device_id dwc3_pci_id_table[] = {
        { PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_BXT), },
        { PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_BXT_M), },
        { PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_APL), },
+       { PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_KBP), },
        { PCI_DEVICE(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_NL_USB), },
        {  }    /* Terminating Entry */
 };
index 8f8c215..1f5597e 100644 (file)
@@ -829,7 +829,7 @@ static void dwc3_prepare_one_trb(struct dwc3_ep *dep,
        if (!req->request.no_interrupt && !chain)
                trb->ctrl |= DWC3_TRB_CTRL_IOC | DWC3_TRB_CTRL_ISP_IMI;
 
-       if (last)
+       if (last && !usb_endpoint_xfer_isoc(dep->endpoint.desc))
                trb->ctrl |= DWC3_TRB_CTRL_LST;
 
        if (chain)
@@ -1955,7 +1955,8 @@ static void dwc3_gadget_free_endpoints(struct dwc3 *dwc)
 
 static int __dwc3_cleanup_done_trbs(struct dwc3 *dwc, struct dwc3_ep *dep,
                struct dwc3_request *req, struct dwc3_trb *trb,
-               const struct dwc3_event_depevt *event, int status)
+               const struct dwc3_event_depevt *event, int status,
+               int chain)
 {
        unsigned int            count;
        unsigned int            s_pkt = 0;
@@ -1964,17 +1965,22 @@ static int __dwc3_cleanup_done_trbs(struct dwc3 *dwc, struct dwc3_ep *dep,
        dep->queued_requests--;
        trace_dwc3_complete_trb(dep, trb);
 
+       /*
+        * If we're in the middle of series of chained TRBs and we
+        * receive a short transfer along the way, DWC3 will skip
+        * through all TRBs including the last TRB in the chain (the
+        * where CHN bit is zero. DWC3 will also avoid clearing HWO
+        * bit and SW has to do it manually.
+        *
+        * We're going to do that here to avoid problems of HW trying
+        * to use bogus TRBs for transfers.
+        */
+       if (chain && (trb->ctrl & DWC3_TRB_CTRL_HWO))
+               trb->ctrl &= ~DWC3_TRB_CTRL_HWO;
+
        if ((trb->ctrl & DWC3_TRB_CTRL_HWO) && status != -ESHUTDOWN)
-               /*
-                * We continue despite the error. There is not much we
-                * can do. If we don't clean it up we loop forever. If
-                * we skip the TRB then it gets overwritten after a
-                * while since we use them in a ring buffer. A BUG()
-                * would help. Lets hope that if this occurs, someone
-                * fixes the root cause instead of looking away :)
-                */
-               dev_err(dwc->dev, "%s's TRB (%p) still owned by HW\n",
-                               dep->name, trb);
+               return 1;
+
        count = trb->size & DWC3_TRB_SIZE_MASK;
 
        if (dep->direction) {
@@ -2013,15 +2019,7 @@ static int __dwc3_cleanup_done_trbs(struct dwc3 *dwc, struct dwc3_ep *dep,
                        s_pkt = 1;
        }
 
-       /*
-        * We assume here we will always receive the entire data block
-        * which we should receive. Meaning, if we program RX to
-        * receive 4K but we receive only 2K, we assume that's all we
-        * should receive and we simply bounce the request back to the
-        * gadget driver for further processing.
-        */
-       req->request.actual += req->request.length - count;
-       if (s_pkt)
+       if (s_pkt && !chain)
                return 1;
        if ((event->status & DEPEVT_STATUS_LST) &&
                        (trb->ctrl & (DWC3_TRB_CTRL_LST |
@@ -2040,13 +2038,17 @@ static int dwc3_cleanup_done_reqs(struct dwc3 *dwc, struct dwc3_ep *dep,
        struct dwc3_trb         *trb;
        unsigned int            slot;
        unsigned int            i;
+       int                     count = 0;
        int                     ret;
 
        do {
+               int chain;
+
                req = next_request(&dep->started_list);
                if (WARN_ON_ONCE(!req))
                        return 1;
 
+               chain = req->request.num_mapped_sgs > 0;
                i = 0;
                do {
                        slot = req->first_trb_index + i;
@@ -2054,13 +2056,22 @@ static int dwc3_cleanup_done_reqs(struct dwc3 *dwc, struct dwc3_ep *dep,
                                slot++;
                        slot %= DWC3_TRB_NUM;
                        trb = &dep->trb_pool[slot];
+                       count += trb->size & DWC3_TRB_SIZE_MASK;
 
                        ret = __dwc3_cleanup_done_trbs(dwc, dep, req, trb,
-                                       event, status);
+                                       event, status, chain);
                        if (ret)
                                break;
                } while (++i < req->request.num_mapped_sgs);
 
+               /*
+                * We assume here we will always receive the entire data block
+                * which we should receive. Meaning, if we program RX to
+                * receive 4K but we receive only 2K, we assume that's all we
+                * should receive and we simply bounce the request back to the
+                * gadget driver for further processing.
+                */
+               req->request.actual += req->request.length - count;
                dwc3_gadget_giveback(dep, req, status);
 
                if (ret)
index eb64848..5ebe6af 100644 (file)
@@ -1913,6 +1913,8 @@ unknown:
                        break;
 
                case USB_RECIP_ENDPOINT:
+                       if (!cdev->config)
+                               break;
                        endp = ((w_index & 0x80) >> 3) | (w_index & 0x0f);
                        list_for_each_entry(f, &cdev->config->functions, list) {
                                if (test_bit(endp, f->endpoints))
@@ -2124,14 +2126,14 @@ int composite_os_desc_req_prepare(struct usb_composite_dev *cdev,
 
        cdev->os_desc_req = usb_ep_alloc_request(ep0, GFP_KERNEL);
        if (!cdev->os_desc_req) {
-               ret = PTR_ERR(cdev->os_desc_req);
+               ret = -ENOMEM;
                goto end;
        }
 
        /* OS feature descriptor length <= 4kB */
        cdev->os_desc_req->buf = kmalloc(4096, GFP_KERNEL);
        if (!cdev->os_desc_req->buf) {
-               ret = PTR_ERR(cdev->os_desc_req->buf);
+               ret = -ENOMEM;
                kfree(cdev->os_desc_req);
                goto end;
        }
index 70cf347..f9237fe 100644 (file)
@@ -1490,7 +1490,9 @@ void unregister_gadget_item(struct config_item *item)
 {
        struct gadget_info *gi = to_gadget_info(item);
 
+       mutex_lock(&gi->lock);
        unregister_gadget(gi);
+       mutex_unlock(&gi->lock);
 }
 EXPORT_SYMBOL_GPL(unregister_gadget_item);
 
index 943c21a..ab6ac1b 100644 (file)
@@ -680,6 +680,12 @@ static int rndis_reset_response(struct rndis_params *params,
 {
        rndis_reset_cmplt_type *resp;
        rndis_resp_t *r;
+       u8 *xbuf;
+       u32 length;
+
+       /* drain the response queue */
+       while ((xbuf = rndis_get_next_response(params, &length)))
+               rndis_free_response(params, xbuf);
 
        r = rndis_add_response(params, sizeof(rndis_reset_cmplt_type));
        if (!r)
index a3f7e7c..5f562c1 100644 (file)
@@ -556,7 +556,8 @@ static netdev_tx_t eth_start_xmit(struct sk_buff *skb,
                        /* Multi frame CDC protocols may store the frame for
                         * later which is not a dropped frame.
                         */
-                       if (dev->port_usb->supports_multi_frame)
+                       if (dev->port_usb &&
+                                       dev->port_usb->supports_multi_frame)
                                goto multiframe;
                        goto drop;
                }
index 66753ba..31125a4 100644 (file)
@@ -2023,7 +2023,7 @@ static int uvcg_streaming_class_allow_link(struct config_item *src,
        if (!data) {
                kfree(*class_array);
                *class_array = NULL;
-               ret = PTR_ERR(data);
+               ret = -ENOMEM;
                goto unlock;
        }
        cl_arr = *class_array;
index aa3707b..16104b5 100644 (file)
@@ -542,7 +542,7 @@ static ssize_t ep_aio(struct kiocb *iocb,
         */
        spin_lock_irq(&epdata->dev->lock);
        value = -ENODEV;
-       if (unlikely(epdata->ep))
+       if (unlikely(epdata->ep == NULL))
                goto fail;
 
        req = usb_ep_alloc_request(epdata->ep, GFP_ATOMIC);
@@ -606,7 +606,7 @@ ep_read_iter(struct kiocb *iocb, struct iov_iter *to)
        }
        if (is_sync_kiocb(iocb)) {
                value = ep_io(epdata, buf, len);
-               if (value >= 0 && copy_to_iter(buf, value, to))
+               if (value >= 0 && (copy_to_iter(buf, value, to) != value))
                        value = -EFAULT;
        } else {
                struct kiocb_priv *priv = kzalloc(sizeof *priv, GFP_KERNEL);
index ff8685e..934f838 100644 (file)
@@ -1145,7 +1145,7 @@ int usb_add_gadget_udc_release(struct device *parent, struct usb_gadget *gadget,
                        if (ret != -EPROBE_DEFER)
                                list_del(&driver->pending);
                        if (ret)
-                               goto err4;
+                               goto err5;
                        break;
                }
        }
@@ -1154,6 +1154,9 @@ int usb_add_gadget_udc_release(struct device *parent, struct usb_gadget *gadget,
 
        return 0;
 
+err5:
+       device_del(&udc->dev);
+
 err4:
        list_del(&udc->list);
        mutex_unlock(&udc_lock);
index 93d28cb..cf8819a 100644 (file)
@@ -2053,7 +2053,7 @@ static void setup_received_handle(struct qe_udc *udc,
                        struct qe_ep *ep;
 
                        if (wValue != 0 || wLength != 0
-                               || pipe > USB_MAX_ENDPOINTS)
+                               || pipe >= USB_MAX_ENDPOINTS)
                                break;
                        ep = &udc->eps[pipe];
 
index a962b89..1e5f529 100644 (file)
@@ -332,11 +332,11 @@ static void ehci_turn_off_all_ports(struct ehci_hcd *ehci)
        int     port = HCS_N_PORTS(ehci->hcs_params);
 
        while (port--) {
-               ehci_writel(ehci, PORT_RWC_BITS,
-                               &ehci->regs->port_status[port]);
                spin_unlock_irq(&ehci->lock);
                ehci_port_power(ehci, port, false);
                spin_lock_irq(&ehci->lock);
+               ehci_writel(ehci, PORT_RWC_BITS,
+                               &ehci->regs->port_status[port]);
        }
 }
 
index c369c29..2f76900 100644 (file)
@@ -1675,7 +1675,7 @@ max3421_gpout_set_value(struct usb_hcd *hcd, u8 pin_number, u8 value)
        if (pin_number > 7)
                return;
 
-       mask = 1u << pin_number;
+       mask = 1u << (pin_number % 4);
        idx = pin_number / 4;
 
        if (value)
index d61fcc4..730b9fd 100644 (file)
@@ -386,6 +386,9 @@ static int xhci_stop_device(struct xhci_hcd *xhci, int slot_id, int suspend)
 
        ret = 0;
        virt_dev = xhci->devs[slot_id];
+       if (!virt_dev)
+               return -ENODEV;
+
        cmd = xhci_alloc_command(xhci, false, true, GFP_NOIO);
        if (!cmd) {
                xhci_dbg(xhci, "Couldn't allocate command structure.\n");
index 4fd041b..d7b0f97 100644 (file)
@@ -314,11 +314,12 @@ static void xhci_pci_remove(struct pci_dev *dev)
                usb_remove_hcd(xhci->shared_hcd);
                usb_put_hcd(xhci->shared_hcd);
        }
-       usb_hcd_pci_remove(dev);
 
        /* Workaround for spurious wakeups at shutdown with HSW */
        if (xhci->quirks & XHCI_SPURIOUS_WAKEUP)
                pci_set_power_state(dev, PCI_D3hot);
+
+       usb_hcd_pci_remove(dev);
 }
 
 #ifdef CONFIG_PM
index 918e0c7..fd9fd12 100644 (file)
@@ -1334,12 +1334,6 @@ static void handle_cmd_completion(struct xhci_hcd *xhci,
 
        cmd = list_entry(xhci->cmd_list.next, struct xhci_command, cmd_list);
 
-       if (cmd->command_trb != xhci->cmd_ring->dequeue) {
-               xhci_err(xhci,
-                        "Command completion event does not match command\n");
-               return;
-       }
-
        del_timer(&xhci->cmd_timer);
 
        trace_xhci_cmd_completion(cmd_trb, (struct xhci_generic_trb *) event);
@@ -1351,6 +1345,13 @@ static void handle_cmd_completion(struct xhci_hcd *xhci,
                xhci_handle_stopped_cmd_ring(xhci, cmd);
                return;
        }
+
+       if (cmd->command_trb != xhci->cmd_ring->dequeue) {
+               xhci_err(xhci,
+                        "Command completion event does not match command\n");
+               return;
+       }
+
        /*
         * Host aborted the command ring, check if the current command was
         * supposed to be aborted, otherwise continue normally.
@@ -3243,7 +3244,8 @@ int xhci_queue_bulk_tx(struct xhci_hcd *xhci, gfp_t mem_flags,
        send_addr = addr;
 
        /* Queue the TRBs, even if they are zero-length */
-       for (enqd_len = 0; enqd_len < full_len; enqd_len += trb_buff_len) {
+       for (enqd_len = 0; first_trb || enqd_len < full_len;
+                       enqd_len += trb_buff_len) {
                field = TRB_TYPE(TRB_NORMAL);
 
                /* TRB buffer should not cross 64KB boundaries */
index 52c27ca..9b5b3b2 100644 (file)
@@ -665,7 +665,7 @@ static ssize_t ftdi_elan_read(struct file *file, char __user *buffer,
 {
        char data[30 *3 + 4];
        char *d = data;
-       int m = (sizeof(data) - 1) / 3;
+       int m = (sizeof(data) - 1) / 3 - 1;
        int bytes_read = 0;
        int retry_on_empty = 10;
        int retry_on_timeout = 5;
@@ -1684,7 +1684,7 @@ wait:if (ftdi->disconnected > 0) {
                        int i = 0;
                        char data[30 *3 + 4];
                        char *d = data;
-                       int m = (sizeof(data) - 1) / 3;
+                       int m = (sizeof(data) - 1) / 3 - 1;
                        int l = 0;
                        struct u132_target *target = &ftdi->target[ed];
                        struct u132_command *command = &ftdi->command[
@@ -1876,7 +1876,7 @@ more:{
                if (packet_bytes > 2) {
                        char diag[30 *3 + 4];
                        char *d = diag;
-                       int m = (sizeof(diag) - 1) / 3;
+                       int m = (sizeof(diag) - 1) / 3 - 1;
                        char *b = ftdi->bulk_in_buffer;
                        int bytes_read = 0;
                        diag[0] = 0;
@@ -2053,7 +2053,7 @@ static int ftdi_elan_synchronize(struct usb_ftdi *ftdi)
                        if (packet_bytes > 2) {
                                char diag[30 *3 + 4];
                                char *d = diag;
-                               int m = (sizeof(diag) - 1) / 3;
+                               int m = (sizeof(diag) - 1) / 3 - 1;
                                char *b = ftdi->bulk_in_buffer;
                                int bytes_read = 0;
                                unsigned char c = 0;
@@ -2155,7 +2155,7 @@ more:{
                if (packet_bytes > 2) {
                        char diag[30 *3 + 4];
                        char *d = diag;
-                       int m = (sizeof(diag) - 1) / 3;
+                       int m = (sizeof(diag) - 1) / 3 - 1;
                        char *b = ftdi->bulk_in_buffer;
                        int bytes_read = 0;
                        diag[0] = 0;
index 6b978f0..5c8210d 100644 (file)
@@ -585,7 +585,6 @@ static void sg_timeout(unsigned long _req)
 {
        struct usb_sg_request   *req = (struct usb_sg_request *) _req;
 
-       req->status = -ETIMEDOUT;
        usb_sg_cancel(req);
 }
 
@@ -616,8 +615,10 @@ static int perform_sglist(
                mod_timer(&sg_timer, jiffies +
                                msecs_to_jiffies(SIMPLE_IO_TIMEOUT));
                usb_sg_wait(req);
-               del_timer_sync(&sg_timer);
-               retval = req->status;
+               if (!del_timer_sync(&sg_timer))
+                       retval = -ETIMEDOUT;
+               else
+                       retval = req->status;
 
                /* FIXME check resulting data pattern */
 
@@ -2602,7 +2603,7 @@ usbtest_ioctl(struct usb_interface *intf, unsigned int code, void *buf)
        ktime_get_ts64(&start);
 
        retval = usbtest_do_ioctl(intf, param_32);
-       if (retval)
+       if (retval < 0)
                goto free_mutex;
 
        ktime_get_ts64(&end);
index 6f6d2a7..6523af4 100644 (file)
@@ -140,6 +140,8 @@ static int omap_otg_probe(struct platform_device *pdev)
                 (rev >> 4) & 0xf, rev & 0xf, config->extcon, otg_dev->id,
                 otg_dev->vbus);
 
+       platform_set_drvdata(pdev, otg_dev);
+
        return 0;
 }
 
index 8fbbc2d..ac67bab 100644 (file)
@@ -514,7 +514,8 @@ static struct renesas_usbhs_platform_info *usbhs_parse_dt(struct device *dev)
        if (gpio > 0)
                dparam->enable_gpio = gpio;
 
-       if (dparam->type == USBHS_TYPE_RCAR_GEN2)
+       if (dparam->type == USBHS_TYPE_RCAR_GEN2 ||
+           dparam->type == USBHS_TYPE_RCAR_GEN3)
                dparam->has_usb_dmac = 1;
 
        return info;
index 280ed5f..857e783 100644 (file)
@@ -871,7 +871,7 @@ static int usbhsf_dma_prepare_push(struct usbhs_pkt *pkt, int *is_done)
 
        /* use PIO if packet is less than pio_dma_border or pipe is DCP */
        if ((len < usbhs_get_dparam(priv, pio_dma_border)) ||
-           usbhs_pipe_is_dcp(pipe))
+           usbhs_pipe_type_is(pipe, USB_ENDPOINT_XFER_ISOC))
                goto usbhsf_pio_prepare_push;
 
        /* check data length if this driver don't use USB-DMAC */
@@ -976,7 +976,7 @@ static int usbhsf_dma_prepare_pop_with_usb_dmac(struct usbhs_pkt *pkt,
 
        /* use PIO if packet is less than pio_dma_border or pipe is DCP */
        if ((pkt->length < usbhs_get_dparam(priv, pio_dma_border)) ||
-           usbhs_pipe_is_dcp(pipe))
+           usbhs_pipe_type_is(pipe, USB_ENDPOINT_XFER_ISOC))
                goto usbhsf_pio_prepare_pop;
 
        fifo = usbhsf_get_dma_fifo(priv, pkt);
index 50f3363..92bc83b 100644 (file)
@@ -617,10 +617,13 @@ static int usbhsg_ep_enable(struct usb_ep *ep,
                 * use dmaengine if possible.
                 * It will use pio handler if impossible.
                 */
-               if (usb_endpoint_dir_in(desc))
+               if (usb_endpoint_dir_in(desc)) {
                        pipe->handler = &usbhs_fifo_dma_push_handler;
-               else
+               } else {
                        pipe->handler = &usbhs_fifo_dma_pop_handler;
+                       usbhs_xxxsts_clear(priv, BRDYSTS,
+                                          usbhs_pipe_number(pipe));
+               }
 
                ret = 0;
        }
index 0082080..b2d767e 100644 (file)
@@ -648,6 +648,8 @@ static const struct usb_device_id id_table_combined[] = {
        { USB_DEVICE(FTDI_VID, FTDI_ELV_TFD128_PID) },
        { USB_DEVICE(FTDI_VID, FTDI_ELV_FM3RX_PID) },
        { USB_DEVICE(FTDI_VID, FTDI_ELV_WS777_PID) },
+       { USB_DEVICE(FTDI_VID, FTDI_PALMSENS_PID) },
+       { USB_DEVICE(FTDI_VID, FTDI_IVIUM_XSTAT_PID) },
        { USB_DEVICE(FTDI_VID, LINX_SDMUSBQSS_PID) },
        { USB_DEVICE(FTDI_VID, LINX_MASTERDEVEL2_PID) },
        { USB_DEVICE(FTDI_VID, LINX_FUTURE_0_PID) },
@@ -1008,6 +1010,7 @@ static const struct usb_device_id id_table_combined[] = {
        { USB_DEVICE(ICPDAS_VID, ICPDAS_I7560U_PID) },
        { USB_DEVICE(ICPDAS_VID, ICPDAS_I7561U_PID) },
        { USB_DEVICE(ICPDAS_VID, ICPDAS_I7563U_PID) },
+       { USB_DEVICE(WICED_VID, WICED_USB20706V2_PID) },
        { }                                     /* Terminating entry */
 };
 
index c5d6c1e..f87a938 100644 (file)
 #define FTDI_4N_GALAXY_DE_2_PID        0xF3C1
 #define FTDI_4N_GALAXY_DE_3_PID        0xF3C2
 
+/*
+ * Ivium Technologies product IDs
+ */
+#define FTDI_PALMSENS_PID      0xf440
+#define FTDI_IVIUM_XSTAT_PID   0xf441
+
 /*
  * Linx Technologies product ids
  */
 #define INTREPID_VALUECAN_PID  0x0601
 #define INTREPID_NEOVI_PID     0x0701
 
+/*
+ * WICED USB UART
+ */
+#define WICED_VID              0x0A5C
+#define WICED_USB20706V2_PID   0x6422
+
 /*
  * Definitions for ID TECH (www.idt-net.com) devices
  */
index 8e07536..bc47258 100644 (file)
@@ -274,6 +274,12 @@ static void option_instat_callback(struct urb *urb);
 #define TELIT_PRODUCT_LE920                    0x1200
 #define TELIT_PRODUCT_LE910                    0x1201
 #define TELIT_PRODUCT_LE910_USBCFG4            0x1206
+#define TELIT_PRODUCT_LE920A4_1207             0x1207
+#define TELIT_PRODUCT_LE920A4_1208             0x1208
+#define TELIT_PRODUCT_LE920A4_1211             0x1211
+#define TELIT_PRODUCT_LE920A4_1212             0x1212
+#define TELIT_PRODUCT_LE920A4_1213             0x1213
+#define TELIT_PRODUCT_LE920A4_1214             0x1214
 
 /* ZTE PRODUCTS */
 #define ZTE_VENDOR_ID                          0x19d2
@@ -628,6 +634,11 @@ static const struct option_blacklist_info telit_le920_blacklist = {
        .reserved = BIT(1) | BIT(5),
 };
 
+static const struct option_blacklist_info telit_le920a4_blacklist_1 = {
+       .sendsetup = BIT(0),
+       .reserved = BIT(1),
+};
+
 static const struct option_blacklist_info telit_le922_blacklist_usbcfg0 = {
        .sendsetup = BIT(2),
        .reserved = BIT(0) | BIT(1) | BIT(3),
@@ -1203,6 +1214,16 @@ static const struct usb_device_id option_ids[] = {
                .driver_info = (kernel_ulong_t)&telit_le922_blacklist_usbcfg3 },
        { USB_DEVICE(TELIT_VENDOR_ID, TELIT_PRODUCT_LE920),
                .driver_info = (kernel_ulong_t)&telit_le920_blacklist },
+       { USB_DEVICE(TELIT_VENDOR_ID, TELIT_PRODUCT_LE920A4_1207) },
+       { USB_DEVICE(TELIT_VENDOR_ID, TELIT_PRODUCT_LE920A4_1208),
+               .driver_info = (kernel_ulong_t)&telit_le920a4_blacklist_1 },
+       { USB_DEVICE(TELIT_VENDOR_ID, TELIT_PRODUCT_LE920A4_1211),
+               .driver_info = (kernel_ulong_t)&telit_le922_blacklist_usbcfg3 },
+       { USB_DEVICE(TELIT_VENDOR_ID, TELIT_PRODUCT_LE920A4_1212),
+               .driver_info = (kernel_ulong_t)&telit_le920a4_blacklist_1 },
+       { USB_DEVICE_INTERFACE_CLASS(TELIT_VENDOR_ID, TELIT_PRODUCT_LE920A4_1213, 0xff) },
+       { USB_DEVICE(TELIT_VENDOR_ID, TELIT_PRODUCT_LE920A4_1214),
+               .driver_info = (kernel_ulong_t)&telit_le922_blacklist_usbcfg3 },
        { USB_DEVICE_AND_INTERFACE_INFO(ZTE_VENDOR_ID, ZTE_PRODUCT_MF622, 0xff, 0xff, 0xff) }, /* ZTE WCDMA products */
        { USB_DEVICE_AND_INTERFACE_INFO(ZTE_VENDOR_ID, 0x0002, 0xff, 0xff, 0xff),
                .driver_info = (kernel_ulong_t)&net_intf1_blacklist },
@@ -1966,6 +1987,7 @@ static const struct usb_device_id option_ids[] = {
          .driver_info = (kernel_ulong_t)&net_intf4_blacklist },
        { USB_DEVICE_AND_INTERFACE_INFO(0x07d1, 0x3e01, 0xff, 0xff, 0xff) }, /* D-Link DWM-152/C1 */
        { USB_DEVICE_AND_INTERFACE_INFO(0x07d1, 0x3e02, 0xff, 0xff, 0xff) }, /* D-Link DWM-156/C1 */
+       { USB_DEVICE_AND_INTERFACE_INFO(0x07d1, 0x7e11, 0xff, 0xff, 0xff) }, /* D-Link DWM-156/A3 */
        { USB_DEVICE_INTERFACE_CLASS(0x2020, 0x4000, 0xff) },                /* OLICARD300 - MT6225 */
        { USB_DEVICE(INOVIA_VENDOR_ID, INOVIA_SEW858) },
        { USB_DEVICE(VIATELECOM_VENDOR_ID, VIATELECOM_PRODUCT_CDS7) },
index b1b9bac..d213cf4 100644 (file)
@@ -1433,7 +1433,7 @@ int usb_serial_register_drivers(struct usb_serial_driver *const serial_drivers[]
 
        rc = usb_register(udriver);
        if (rc)
-               return rc;
+               goto failed_usb_register;
 
        for (sd = serial_drivers; *sd; ++sd) {
                (*sd)->usb_driver = udriver;
@@ -1451,6 +1451,8 @@ int usb_serial_register_drivers(struct usb_serial_driver *const serial_drivers[]
        while (sd-- > serial_drivers)
                usb_serial_deregister(*sd);
        usb_deregister(udriver);
+failed_usb_register:
+       kfree(udriver);
        return rc;
 }
 EXPORT_SYMBOL_GPL(usb_serial_register_drivers);
index 9d6320e..6e29d05 100644 (file)
@@ -88,7 +88,7 @@ struct vhost_scsi_cmd {
        struct scatterlist *tvc_prot_sgl;
        struct page **tvc_upages;
        /* Pointer to response header iovec */
-       struct iovec *tvc_resp_iov;
+       struct iovec tvc_resp_iov;
        /* Pointer to vhost_scsi for our device */
        struct vhost_scsi *tvc_vhost;
        /* Pointer to vhost_virtqueue for the cmd */
@@ -547,7 +547,7 @@ static void vhost_scsi_complete_cmd_work(struct vhost_work *work)
                memcpy(v_rsp.sense, cmd->tvc_sense_buf,
                       se_cmd->scsi_sense_length);
 
-               iov_iter_init(&iov_iter, READ, cmd->tvc_resp_iov,
+               iov_iter_init(&iov_iter, READ, &cmd->tvc_resp_iov,
                              cmd->tvc_in_iovs, sizeof(v_rsp));
                ret = copy_to_iter(&v_rsp, sizeof(v_rsp), &iov_iter);
                if (likely(ret == sizeof(v_rsp))) {
@@ -1044,7 +1044,7 @@ vhost_scsi_handle_vq(struct vhost_scsi *vs, struct vhost_virtqueue *vq)
                }
                cmd->tvc_vhost = vs;
                cmd->tvc_vq = vq;
-               cmd->tvc_resp_iov = &vq->iov[out];
+               cmd->tvc_resp_iov = vq->iov[out];
                cmd->tvc_in_iovs = in;
 
                pr_debug("vhost_scsi got command opcode: %#02x, lun: %d\n",
index 7487971..c1010f0 100644 (file)
@@ -316,7 +316,7 @@ static int xenbus_write_transaction(unsigned msg_type,
                        rc = -ENOMEM;
                        goto out;
                }
-       } else {
+       } else if (msg_type == XS_TRANSACTION_END) {
                list_for_each_entry(trans, &u->transactions, list)
                        if (trans->handle.id == u->u.msg.tx_id)
                                break;
index c3cdde8..08ae993 100644 (file)
@@ -249,7 +249,8 @@ struct super_block *freeze_bdev(struct block_device *bdev)
                 * thaw_bdev drops it.
                 */
                sb = get_super(bdev);
-               drop_super(sb);
+               if (sb)
+                       drop_super(sb);
                mutex_unlock(&bdev->bd_fsfreeze_mutex);
                return sb;
        }
@@ -646,7 +647,7 @@ static struct dentry *bd_mount(struct file_system_type *fs_type,
 {
        struct dentry *dent;
        dent = mount_pseudo(fs_type, "bdev:", &bdev_sops, NULL, BDEVFS_MAGIC);
-       if (dent)
+       if (!IS_ERR(dent))
                dent->d_sb->s_iflags |= SB_I_CGROUPWB;
        return dent;
 }
index 2b88439..455a6b2 100644 (file)
@@ -589,6 +589,7 @@ static void __merge_refs(struct list_head *head, int mode)
 
                        list_del(&ref2->list);
                        kmem_cache_free(btrfs_prelim_ref_cache, ref2);
+                       cond_resched();
                }
 
        }
index 2fe8f89..eff3993 100644 (file)
@@ -1028,6 +1028,7 @@ struct btrfs_fs_info {
        struct btrfs_workqueue *qgroup_rescan_workers;
        struct completion qgroup_rescan_completion;
        struct btrfs_work qgroup_rescan_work;
+       bool qgroup_rescan_running;     /* protected by qgroup_rescan_lock */
 
        /* filesystem state */
        unsigned long fs_state;
@@ -1079,6 +1080,8 @@ struct btrfs_fs_info {
        struct list_head pinned_chunks;
 
        int creating_free_space_tree;
+       /* Used to record internally whether fs has been frozen */
+       int fs_frozen;
 };
 
 struct btrfs_subvolume_writers {
@@ -2578,7 +2581,7 @@ int btrfs_alloc_logged_file_extent(struct btrfs_trans_handle *trans,
                                   struct btrfs_root *root,
                                   u64 root_objectid, u64 owner, u64 offset,
                                   struct btrfs_key *ins);
-int btrfs_reserve_extent(struct btrfs_root *root, u64 num_bytes,
+int btrfs_reserve_extent(struct btrfs_root *root, u64 ram_bytes, u64 num_bytes,
                         u64 min_alloc_size, u64 empty_size, u64 hint_byte,
                         struct btrfs_key *ins, int is_data, int delalloc);
 int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
index d9ddcfc..ac02e04 100644 (file)
@@ -541,7 +541,6 @@ add_delayed_ref_head(struct btrfs_fs_info *fs_info,
        struct btrfs_delayed_ref_head *existing;
        struct btrfs_delayed_ref_head *head_ref = NULL;
        struct btrfs_delayed_ref_root *delayed_refs;
-       struct btrfs_qgroup_extent_record *qexisting;
        int count_mod = 1;
        int must_insert_reserved = 0;
 
@@ -606,10 +605,8 @@ add_delayed_ref_head(struct btrfs_fs_info *fs_info,
                qrecord->num_bytes = num_bytes;
                qrecord->old_roots = NULL;
 
-               qexisting = btrfs_qgroup_insert_dirty_extent(fs_info,
-                                                            delayed_refs,
-                                                            qrecord);
-               if (qexisting)
+               if(btrfs_qgroup_insert_dirty_extent_nolock(fs_info,
+                                       delayed_refs, qrecord))
                        kfree(qrecord);
        }
 
index 59febfb..54bc8c7 100644 (file)
@@ -559,8 +559,29 @@ static noinline int check_leaf(struct btrfs_root *root,
        u32 nritems = btrfs_header_nritems(leaf);
        int slot;
 
-       if (nritems == 0)
+       if (nritems == 0) {
+               struct btrfs_root *check_root;
+
+               key.objectid = btrfs_header_owner(leaf);
+               key.type = BTRFS_ROOT_ITEM_KEY;
+               key.offset = (u64)-1;
+
+               check_root = btrfs_get_fs_root(root->fs_info, &key, false);
+               /*
+                * The only reason we also check NULL here is that during
+                * open_ctree() some roots has not yet been set up.
+                */
+               if (!IS_ERR_OR_NULL(check_root)) {
+                       /* if leaf is the root, then it's fine */
+                       if (leaf->start !=
+                           btrfs_root_bytenr(&check_root->root_item)) {
+                               CORRUPT("non-root leaf's nritems is 0",
+                                       leaf, root, 0);
+                               return -EIO;
+                       }
+               }
                return 0;
+       }
 
        /* Check the 0 item */
        if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) !=
@@ -612,6 +633,19 @@ static noinline int check_leaf(struct btrfs_root *root,
        return 0;
 }
 
+static int check_node(struct btrfs_root *root, struct extent_buffer *node)
+{
+       unsigned long nr = btrfs_header_nritems(node);
+
+       if (nr == 0 || nr > BTRFS_NODEPTRS_PER_BLOCK(root)) {
+               btrfs_crit(root->fs_info,
+                          "corrupt node: block %llu root %llu nritems %lu",
+                          node->start, root->objectid, nr);
+               return -EIO;
+       }
+       return 0;
+}
+
 static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio,
                                      u64 phy_offset, struct page *page,
                                      u64 start, u64 end, int mirror)
@@ -682,6 +716,9 @@ static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio,
                ret = -EIO;
        }
 
+       if (found_level > 0 && check_node(root, eb))
+               ret = -EIO;
+
        if (!ret)
                set_extent_buffer_uptodate(eb);
 err:
@@ -1618,8 +1655,8 @@ fail:
        return ret;
 }
 
-static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
-                                              u64 root_id)
+struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
+                                       u64 root_id)
 {
        struct btrfs_root *root;
 
@@ -2298,6 +2335,7 @@ static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info)
        fs_info->quota_enabled = 0;
        fs_info->pending_quota_state = 0;
        fs_info->qgroup_ulist = NULL;
+       fs_info->qgroup_rescan_running = false;
        mutex_init(&fs_info->qgroup_rescan_lock);
 }
 
@@ -2624,6 +2662,7 @@ int open_ctree(struct super_block *sb,
        atomic_set(&fs_info->qgroup_op_seq, 0);
        atomic_set(&fs_info->reada_works_cnt, 0);
        atomic64_set(&fs_info->tree_mod_seq, 0);
+       fs_info->fs_frozen = 0;
        fs_info->sb = sb;
        fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE;
        fs_info->metadata_ratio = 0;
@@ -3739,8 +3778,15 @@ void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info,
        if (btrfs_root_refs(&root->root_item) == 0)
                synchronize_srcu(&fs_info->subvol_srcu);
 
-       if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
+       if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
                btrfs_free_log(NULL, root);
+               if (root->reloc_root) {
+                       free_extent_buffer(root->reloc_root->node);
+                       free_extent_buffer(root->reloc_root->commit_root);
+                       btrfs_put_fs_root(root->reloc_root);
+                       root->reloc_root = NULL;
+               }
+       }
 
        if (root->free_ino_pinned)
                __btrfs_remove_free_space_cache(root->free_ino_pinned);
@@ -3851,7 +3897,7 @@ void close_ctree(struct btrfs_root *root)
        smp_mb();
 
        /* wait for the qgroup rescan worker to stop */
-       btrfs_qgroup_wait_for_completion(fs_info);
+       btrfs_qgroup_wait_for_completion(fs_info, false);
 
        /* wait for the uuid_scan task to finish */
        down(&fs_info->uuid_tree_rescan_sem);
index b3207a0..f19a982 100644 (file)
@@ -68,6 +68,8 @@ struct extent_buffer *btrfs_find_tree_block(struct btrfs_fs_info *fs_info,
 struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root,
                                      struct btrfs_key *location);
 int btrfs_init_fs_root(struct btrfs_root *root);
+struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
+                                       u64 root_id);
 int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info,
                         struct btrfs_root *root);
 void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info);
index 61b494e..0450dc4 100644 (file)
@@ -60,21 +60,6 @@ enum {
        CHUNK_ALLOC_FORCE = 2,
 };
 
-/*
- * Control how reservations are dealt with.
- *
- * RESERVE_FREE - freeing a reservation.
- * RESERVE_ALLOC - allocating space and we need to update bytes_may_use for
- *   ENOSPC accounting
- * RESERVE_ALLOC_NO_ACCOUNT - allocating space and we should not update
- *   bytes_may_use as the ENOSPC accounting is done elsewhere
- */
-enum {
-       RESERVE_FREE = 0,
-       RESERVE_ALLOC = 1,
-       RESERVE_ALLOC_NO_ACCOUNT = 2,
-};
-
 static int update_block_group(struct btrfs_trans_handle *trans,
                              struct btrfs_root *root, u64 bytenr,
                              u64 num_bytes, int alloc);
@@ -104,9 +89,10 @@ static int find_next_key(struct btrfs_path *path, int level,
                         struct btrfs_key *key);
 static void dump_space_info(struct btrfs_space_info *info, u64 bytes,
                            int dump_block_groups);
-static int btrfs_update_reserved_bytes(struct btrfs_block_group_cache *cache,
-                                      u64 num_bytes, int reserve,
-                                      int delalloc);
+static int btrfs_add_reserved_bytes(struct btrfs_block_group_cache *cache,
+                                   u64 ram_bytes, u64 num_bytes, int delalloc);
+static int btrfs_free_reserved_bytes(struct btrfs_block_group_cache *cache,
+                                    u64 num_bytes, int delalloc);
 static int block_rsv_use_bytes(struct btrfs_block_rsv *block_rsv,
                               u64 num_bytes);
 int btrfs_pin_extent(struct btrfs_root *root,
@@ -3501,7 +3487,6 @@ again:
                dcs = BTRFS_DC_SETUP;
        else if (ret == -ENOSPC)
                set_bit(BTRFS_TRANS_CACHE_ENOSPC, &trans->transaction->flags);
-       btrfs_free_reserved_data_space(inode, 0, num_pages);
 
 out_put:
        iput(inode);
@@ -4472,6 +4457,15 @@ void check_system_chunk(struct btrfs_trans_handle *trans,
        }
 }
 
+/*
+ * If force is CHUNK_ALLOC_FORCE:
+ *    - return 1 if it successfully allocates a chunk,
+ *    - return errors including -ENOSPC otherwise.
+ * If force is NOT CHUNK_ALLOC_FORCE:
+ *    - return 0 if it doesn't need to allocate a new chunk,
+ *    - return 1 if it successfully allocates a chunk,
+ *    - return errors including -ENOSPC otherwise.
+ */
 static int do_chunk_alloc(struct btrfs_trans_handle *trans,
                          struct btrfs_root *extent_root, u64 flags, int force)
 {
@@ -4882,7 +4876,7 @@ static int flush_space(struct btrfs_root *root,
                                     btrfs_get_alloc_profile(root, 0),
                                     CHUNK_ALLOC_NO_FORCE);
                btrfs_end_transaction(trans, root);
-               if (ret == -ENOSPC)
+               if (ret > 0 || ret == -ENOSPC)
                        ret = 0;
                break;
        case COMMIT_TRANS:
@@ -6497,19 +6491,15 @@ void btrfs_wait_block_group_reservations(struct btrfs_block_group_cache *bg)
 }
 
 /**
- * btrfs_update_reserved_bytes - update the block_group and space info counters
+ * btrfs_add_reserved_bytes - update the block_group and space info counters
  * @cache:     The cache we are manipulating
+ * @ram_bytes:  The number of bytes of file content, and will be same to
+ *              @num_bytes except for the compress path.
  * @num_bytes: The number of bytes in question
- * @reserve:   One of the reservation enums
  * @delalloc:   The blocks are allocated for the delalloc write
  *
- * This is called by the allocator when it reserves space, or by somebody who is
- * freeing space that was never actually used on disk.  For example if you
- * reserve some space for a new leaf in transaction A and before transaction A
- * commits you free that leaf, you call this with reserve set to 0 in order to
- * clear the reservation.
- *
- * Metadata reservations should be called with RESERVE_ALLOC so we do the proper
+ * This is called by the allocator when it reserves space. Metadata
+ * reservations should be called with RESERVE_ALLOC so we do the proper
  * ENOSPC accounting.  For data we handle the reservation through clearing the
  * delalloc bits in the io_tree.  We have to do this since we could end up
  * allocating less disk space for the amount of data we have reserved in the
@@ -6519,44 +6509,63 @@ void btrfs_wait_block_group_reservations(struct btrfs_block_group_cache *bg)
  * make the reservation and return -EAGAIN, otherwise this function always
  * succeeds.
  */
-static int btrfs_update_reserved_bytes(struct btrfs_block_group_cache *cache,
-                                      u64 num_bytes, int reserve, int delalloc)
+static int btrfs_add_reserved_bytes(struct btrfs_block_group_cache *cache,
+                                   u64 ram_bytes, u64 num_bytes, int delalloc)
 {
        struct btrfs_space_info *space_info = cache->space_info;
        int ret = 0;
 
        spin_lock(&space_info->lock);
        spin_lock(&cache->lock);
-       if (reserve != RESERVE_FREE) {
-               if (cache->ro) {
-                       ret = -EAGAIN;
-               } else {
-                       cache->reserved += num_bytes;
-                       space_info->bytes_reserved += num_bytes;
-                       if (reserve == RESERVE_ALLOC) {
-                               trace_btrfs_space_reservation(cache->fs_info,
-                                               "space_info", space_info->flags,
-                                               num_bytes, 0);
-                               space_info->bytes_may_use -= num_bytes;
-                       }
-
-                       if (delalloc)
-                               cache->delalloc_bytes += num_bytes;
-               }
+       if (cache->ro) {
+               ret = -EAGAIN;
        } else {
-               if (cache->ro)
-                       space_info->bytes_readonly += num_bytes;
-               cache->reserved -= num_bytes;
-               space_info->bytes_reserved -= num_bytes;
+               cache->reserved += num_bytes;
+               space_info->bytes_reserved += num_bytes;
 
+               trace_btrfs_space_reservation(cache->fs_info,
+                               "space_info", space_info->flags,
+                               ram_bytes, 0);
+               space_info->bytes_may_use -= ram_bytes;
                if (delalloc)
-                       cache->delalloc_bytes -= num_bytes;
+                       cache->delalloc_bytes += num_bytes;
        }
        spin_unlock(&cache->lock);
        spin_unlock(&space_info->lock);
        return ret;
 }
 
+/**
+ * btrfs_free_reserved_bytes - update the block_group and space info counters
+ * @cache:      The cache we are manipulating
+ * @num_bytes:  The number of bytes in question
+ * @delalloc:   The blocks are allocated for the delalloc write
+ *
+ * This is called by somebody who is freeing space that was never actually used
+ * on disk.  For example if you reserve some space for a new leaf in transaction
+ * A and before transaction A commits you free that leaf, you call this with
+ * reserve set to 0 in order to clear the reservation.
+ */
+
+static int btrfs_free_reserved_bytes(struct btrfs_block_group_cache *cache,
+                                    u64 num_bytes, int delalloc)
+{
+       struct btrfs_space_info *space_info = cache->space_info;
+       int ret = 0;
+
+       spin_lock(&space_info->lock);
+       spin_lock(&cache->lock);
+       if (cache->ro)
+               space_info->bytes_readonly += num_bytes;
+       cache->reserved -= num_bytes;
+       space_info->bytes_reserved -= num_bytes;
+
+       if (delalloc)
+               cache->delalloc_bytes -= num_bytes;
+       spin_unlock(&cache->lock);
+       spin_unlock(&space_info->lock);
+       return ret;
+}
 void btrfs_prepare_extent_commit(struct btrfs_trans_handle *trans,
                                struct btrfs_root *root)
 {
@@ -7191,7 +7200,7 @@ void btrfs_free_tree_block(struct btrfs_trans_handle *trans,
                WARN_ON(test_bit(EXTENT_BUFFER_DIRTY, &buf->bflags));
 
                btrfs_add_free_space(cache, buf->start, buf->len);
-               btrfs_update_reserved_bytes(cache, buf->len, RESERVE_FREE, 0);
+               btrfs_free_reserved_bytes(cache, buf->len, 0);
                btrfs_put_block_group(cache);
                trace_btrfs_reserved_extent_free(root, buf->start, buf->len);
                pin = 0;
@@ -7416,9 +7425,9 @@ btrfs_release_block_group(struct btrfs_block_group_cache *cache,
  * the free space extent currently.
  */
 static noinline int find_free_extent(struct btrfs_root *orig_root,
-                                    u64 num_bytes, u64 empty_size,
-                                    u64 hint_byte, struct btrfs_key *ins,
-                                    u64 flags, int delalloc)
+                               u64 ram_bytes, u64 num_bytes, u64 empty_size,
+                               u64 hint_byte, struct btrfs_key *ins,
+                               u64 flags, int delalloc)
 {
        int ret = 0;
        struct btrfs_root *root = orig_root->fs_info->extent_root;
@@ -7430,8 +7439,6 @@ static noinline int find_free_extent(struct btrfs_root *orig_root,
        struct btrfs_space_info *space_info;
        int loop = 0;
        int index = __get_raid_index(flags);
-       int alloc_type = (flags & BTRFS_BLOCK_GROUP_DATA) ?
-               RESERVE_ALLOC_NO_ACCOUNT : RESERVE_ALLOC;
        bool failed_cluster_refill = false;
        bool failed_alloc = false;
        bool use_cluster = true;
@@ -7763,8 +7770,8 @@ checks:
                                             search_start - offset);
                BUG_ON(offset > search_start);
 
-               ret = btrfs_update_reserved_bytes(block_group, num_bytes,
-                                                 alloc_type, delalloc);
+               ret = btrfs_add_reserved_bytes(block_group, ram_bytes,
+                               num_bytes, delalloc);
                if (ret == -EAGAIN) {
                        btrfs_add_free_space(block_group, offset, num_bytes);
                        goto loop;
@@ -7936,7 +7943,7 @@ again:
        up_read(&info->groups_sem);
 }
 
-int btrfs_reserve_extent(struct btrfs_root *root,
+int btrfs_reserve_extent(struct btrfs_root *root, u64 ram_bytes,
                         u64 num_bytes, u64 min_alloc_size,
                         u64 empty_size, u64 hint_byte,
                         struct btrfs_key *ins, int is_data, int delalloc)
@@ -7948,8 +7955,8 @@ int btrfs_reserve_extent(struct btrfs_root *root,
        flags = btrfs_get_alloc_profile(root, is_data);
 again:
        WARN_ON(num_bytes < root->sectorsize);
-       ret = find_free_extent(root, num_bytes, empty_size, hint_byte, ins,
-                              flags, delalloc);
+       ret = find_free_extent(root, ram_bytes, num_bytes, empty_size,
+                              hint_byte, ins, flags, delalloc);
        if (!ret && !is_data) {
                btrfs_dec_block_group_reservations(root->fs_info,
                                                   ins->objectid);
@@ -7958,6 +7965,7 @@ again:
                        num_bytes = min(num_bytes >> 1, ins->offset);
                        num_bytes = round_down(num_bytes, root->sectorsize);
                        num_bytes = max(num_bytes, min_alloc_size);
+                       ram_bytes = num_bytes;
                        if (num_bytes == min_alloc_size)
                                final_tried = true;
                        goto again;
@@ -7995,7 +8003,7 @@ static int __btrfs_free_reserved_extent(struct btrfs_root *root,
                if (btrfs_test_opt(root->fs_info, DISCARD))
                        ret = btrfs_discard_extent(root, start, len, NULL);
                btrfs_add_free_space(cache, start, len);
-               btrfs_update_reserved_bytes(cache, len, RESERVE_FREE, delalloc);
+               btrfs_free_reserved_bytes(cache, len, delalloc);
                trace_btrfs_reserved_extent_free(root, start, len);
        }
 
@@ -8223,8 +8231,8 @@ int btrfs_alloc_logged_file_extent(struct btrfs_trans_handle *trans,
        if (!block_group)
                return -EINVAL;
 
-       ret = btrfs_update_reserved_bytes(block_group, ins->offset,
-                                         RESERVE_ALLOC_NO_ACCOUNT, 0);
+       ret = btrfs_add_reserved_bytes(block_group, ins->offset,
+                                      ins->offset, 0);
        BUG_ON(ret); /* logic error */
        ret = alloc_reserved_file_extent(trans, root, 0, root_objectid,
                                         0, owner, offset, ins, 1);
@@ -8368,7 +8376,7 @@ struct extent_buffer *btrfs_alloc_tree_block(struct btrfs_trans_handle *trans,
        if (IS_ERR(block_rsv))
                return ERR_CAST(block_rsv);
 
-       ret = btrfs_reserve_extent(root, blocksize, blocksize,
+       ret = btrfs_reserve_extent(root, blocksize, blocksize, blocksize,
                                   empty_size, hint, &ins, 0, 0);
        if (ret)
                goto out_unuse;
@@ -8521,35 +8529,6 @@ reada:
        wc->reada_slot = slot;
 }
 
-/*
- * These may not be seen by the usual inc/dec ref code so we have to
- * add them here.
- */
-static int record_one_subtree_extent(struct btrfs_trans_handle *trans,
-                                    struct btrfs_root *root, u64 bytenr,
-                                    u64 num_bytes)
-{
-       struct btrfs_qgroup_extent_record *qrecord;
-       struct btrfs_delayed_ref_root *delayed_refs;
-
-       qrecord = kmalloc(sizeof(*qrecord), GFP_NOFS);
-       if (!qrecord)
-               return -ENOMEM;
-
-       qrecord->bytenr = bytenr;
-       qrecord->num_bytes = num_bytes;
-       qrecord->old_roots = NULL;
-
-       delayed_refs = &trans->transaction->delayed_refs;
-       spin_lock(&delayed_refs->lock);
-       if (btrfs_qgroup_insert_dirty_extent(trans->fs_info,
-                                            delayed_refs, qrecord))
-               kfree(qrecord);
-       spin_unlock(&delayed_refs->lock);
-
-       return 0;
-}
-
 static int account_leaf_items(struct btrfs_trans_handle *trans,
                              struct btrfs_root *root,
                              struct extent_buffer *eb)
@@ -8583,7 +8562,8 @@ static int account_leaf_items(struct btrfs_trans_handle *trans,
 
                num_bytes = btrfs_file_extent_disk_num_bytes(eb, fi);
 
-               ret = record_one_subtree_extent(trans, root, bytenr, num_bytes);
+               ret = btrfs_qgroup_insert_dirty_extent(trans, root->fs_info,
+                               bytenr, num_bytes, GFP_NOFS);
                if (ret)
                        return ret;
        }
@@ -8732,8 +8712,9 @@ walk_down:
                        btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
                        path->locks[level] = BTRFS_READ_LOCK_BLOCKING;
 
-                       ret = record_one_subtree_extent(trans, root, child_bytenr,
-                                                       root->nodesize);
+                       ret = btrfs_qgroup_insert_dirty_extent(trans,
+                                       root->fs_info, child_bytenr,
+                                       root->nodesize, GFP_NOFS);
                        if (ret)
                                goto out;
                }
@@ -9906,6 +9887,7 @@ static int find_first_block_group(struct btrfs_root *root,
                        } else {
                                ret = 0;
                        }
+                       free_extent_map(em);
                        goto out;
                }
                path->slots[0]++;
@@ -9942,6 +9924,7 @@ void btrfs_put_block_group_cache(struct btrfs_fs_info *info)
                block_group->iref = 0;
                block_group->inode = NULL;
                spin_unlock(&block_group->lock);
+               ASSERT(block_group->io_ctl.inode == NULL);
                iput(inode);
                last = block_group->key.objectid + block_group->key.offset;
                btrfs_put_block_group(block_group);
@@ -9999,6 +9982,10 @@ int btrfs_free_block_groups(struct btrfs_fs_info *info)
                        free_excluded_extents(info->extent_root, block_group);
 
                btrfs_remove_free_space_cache(block_group);
+               ASSERT(list_empty(&block_group->dirty_list));
+               ASSERT(list_empty(&block_group->io_list));
+               ASSERT(list_empty(&block_group->bg_list));
+               ASSERT(atomic_read(&block_group->count) == 1);
                btrfs_put_block_group(block_group);
 
                spin_lock(&info->block_group_cache_lock);
index bc2729a..28cd88f 100644 (file)
@@ -20,6 +20,7 @@
 #define EXTENT_DAMAGED         (1U << 14)
 #define EXTENT_NORESERVE       (1U << 15)
 #define EXTENT_QGROUP_RESERVED (1U << 16)
+#define EXTENT_CLEAR_DATA_RESV (1U << 17)
 #define EXTENT_IOBITS          (EXTENT_LOCKED | EXTENT_WRITEBACK)
 #define EXTENT_CTLBITS         (EXTENT_DO_ACCOUNTING | EXTENT_FIRST_DELALLOC)
 
index 5842423..fea31a4 100644 (file)
@@ -2070,7 +2070,7 @@ int btrfs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
        }
        trans->sync = true;
 
-       btrfs_init_log_ctx(&ctx);
+       btrfs_init_log_ctx(&ctx, inode);
 
        ret = btrfs_log_dentry_safe(trans, root, dentry, start, end, &ctx);
        if (ret < 0) {
@@ -2675,6 +2675,7 @@ static long btrfs_fallocate(struct file *file, int mode,
 
        alloc_start = round_down(offset, blocksize);
        alloc_end = round_up(offset + len, blocksize);
+       cur_offset = alloc_start;
 
        /* Make sure we aren't being give some crap mode */
        if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
@@ -2767,7 +2768,6 @@ static long btrfs_fallocate(struct file *file, int mode,
 
        /* First, check if we exceed the qgroup limit */
        INIT_LIST_HEAD(&reserve_list);
-       cur_offset = alloc_start;
        while (1) {
                em = btrfs_get_extent(inode, NULL, 0, cur_offset,
                                      alloc_end - cur_offset, 0);
@@ -2794,6 +2794,14 @@ static long btrfs_fallocate(struct file *file, int mode,
                                        last_byte - cur_offset);
                        if (ret < 0)
                                break;
+               } else {
+                       /*
+                        * Do not need to reserve unwritten extent for this
+                        * range, free reserved data space first, otherwise
+                        * it'll result in false ENOSPC error.
+                        */
+                       btrfs_free_reserved_data_space(inode, cur_offset,
+                               last_byte - cur_offset);
                }
                free_extent_map(em);
                cur_offset = last_byte;
@@ -2811,6 +2819,9 @@ static long btrfs_fallocate(struct file *file, int mode,
                                        range->start,
                                        range->len, 1 << inode->i_blkbits,
                                        offset + len, &alloc_hint);
+               else
+                       btrfs_free_reserved_data_space(inode, range->start,
+                                                      range->len);
                list_del(&range->list);
                kfree(range);
        }
@@ -2845,18 +2856,11 @@ out_unlock:
        unlock_extent_cached(&BTRFS_I(inode)->io_tree, alloc_start, locked_end,
                             &cached_state, GFP_KERNEL);
 out:
-       /*
-        * As we waited the extent range, the data_rsv_map must be empty
-        * in the range, as written data range will be released from it.
-        * And for prealloacted extent, it will also be released when
-        * its metadata is written.
-        * So this is completely used as cleanup.
-        */
-       btrfs_qgroup_free_data(inode, alloc_start, alloc_end - alloc_start);
        inode_unlock(inode);
        /* Let go of our reservation. */
-       btrfs_free_reserved_data_space(inode, alloc_start,
-                                      alloc_end - alloc_start);
+       if (ret != 0)
+               btrfs_free_reserved_data_space(inode, alloc_start,
+                                      alloc_end - cur_offset);
        return ret;
 }
 
index aa6faba..359ee86 100644 (file)
@@ -495,10 +495,9 @@ again:
        ret = btrfs_prealloc_file_range_trans(inode, trans, 0, 0, prealloc,
                                              prealloc, prealloc, &alloc_hint);
        if (ret) {
-               btrfs_delalloc_release_space(inode, 0, prealloc);
+               btrfs_delalloc_release_metadata(inode, prealloc);
                goto out_put;
        }
-       btrfs_free_reserved_data_space(inode, 0, prealloc);
 
        ret = btrfs_write_out_ino_cache(root, trans, path, inode);
 out_put:
index 08dfc57..e6811c4 100644 (file)
@@ -566,6 +566,8 @@ cont:
                                                     PAGE_SET_WRITEBACK |
                                                     page_error_op |
                                                     PAGE_END_WRITEBACK);
+                       btrfs_free_reserved_data_space_noquota(inode, start,
+                                               end - start + 1);
                        goto free_pages_out;
                }
        }
@@ -742,7 +744,7 @@ retry:
                lock_extent(io_tree, async_extent->start,
                            async_extent->start + async_extent->ram_size - 1);
 
-               ret = btrfs_reserve_extent(root,
+               ret = btrfs_reserve_extent(root, async_extent->ram_size,
                                           async_extent->compressed_size,
                                           async_extent->compressed_size,
                                           0, alloc_hint, &ins, 1, 1);
@@ -969,7 +971,8 @@ static noinline int cow_file_range(struct inode *inode,
                                     EXTENT_DEFRAG, PAGE_UNLOCK |
                                     PAGE_CLEAR_DIRTY | PAGE_SET_WRITEBACK |
                                     PAGE_END_WRITEBACK);
-
+                       btrfs_free_reserved_data_space_noquota(inode, start,
+                                               end - start + 1);
                        *nr_written = *nr_written +
                             (end - start + PAGE_SIZE) / PAGE_SIZE;
                        *page_started = 1;
@@ -989,7 +992,7 @@ static noinline int cow_file_range(struct inode *inode,
                unsigned long op;
 
                cur_alloc_size = disk_num_bytes;
-               ret = btrfs_reserve_extent(root, cur_alloc_size,
+               ret = btrfs_reserve_extent(root, cur_alloc_size, cur_alloc_size,
                                           root->sectorsize, 0, alloc_hint,
                                           &ins, 1, 1);
                if (ret < 0)
@@ -1489,8 +1492,10 @@ out_check:
                extent_clear_unlock_delalloc(inode, cur_offset,
                                             cur_offset + num_bytes - 1,
                                             locked_page, EXTENT_LOCKED |
-                                            EXTENT_DELALLOC, PAGE_UNLOCK |
-                                            PAGE_SET_PRIVATE2);
+                                            EXTENT_DELALLOC |
+                                            EXTENT_CLEAR_DATA_RESV,
+                                            PAGE_UNLOCK | PAGE_SET_PRIVATE2);
+
                if (!nolock && nocow)
                        btrfs_end_write_no_snapshoting(root);
                cur_offset = extent_end;
@@ -1807,7 +1812,9 @@ static void btrfs_clear_bit_hook(struct inode *inode,
                        return;
 
                if (root->root_key.objectid != BTRFS_DATA_RELOC_TREE_OBJECTID
-                   && do_list && !(state->state & EXTENT_NORESERVE))
+                   && do_list && !(state->state & EXTENT_NORESERVE)
+                   && (*bits & (EXTENT_DO_ACCOUNTING |
+                   EXTENT_CLEAR_DATA_RESV)))
                        btrfs_free_reserved_data_space_noquota(inode,
                                        state->start, len);
 
@@ -7251,7 +7258,7 @@ static struct extent_map *btrfs_new_extent_direct(struct inode *inode,
        int ret;
 
        alloc_hint = get_extent_allocation_hint(inode, start, len);
-       ret = btrfs_reserve_extent(root, len, root->sectorsize, 0,
+       ret = btrfs_reserve_extent(root, len, len, root->sectorsize, 0,
                                   alloc_hint, &ins, 1, 1);
        if (ret)
                return ERR_PTR(ret);
@@ -7751,6 +7758,13 @@ static int btrfs_get_blocks_direct(struct inode *inode, sector_t iblock,
                                ret = PTR_ERR(em2);
                                goto unlock_err;
                        }
+                       /*
+                        * For inode marked NODATACOW or extent marked PREALLOC,
+                        * use the existing or preallocated extent, so does not
+                        * need to adjust btrfs_space_info's bytes_may_use.
+                        */
+                       btrfs_free_reserved_data_space_noquota(inode,
+                                       start, len);
                        goto unlock;
                }
        }
@@ -7785,7 +7799,6 @@ unlock:
                        i_size_write(inode, start + len);
 
                adjust_dio_outstanding_extents(inode, dio_data, len);
-               btrfs_free_reserved_data_space(inode, start, len);
                WARN_ON(dio_data->reserve < len);
                dio_data->reserve -= len;
                dio_data->unsubmitted_oe_range_end = start + len;
@@ -10306,6 +10319,7 @@ static int __btrfs_prealloc_file_range(struct inode *inode, int mode,
        u64 last_alloc = (u64)-1;
        int ret = 0;
        bool own_trans = true;
+       u64 end = start + num_bytes - 1;
 
        if (trans)
                own_trans = false;
@@ -10327,8 +10341,8 @@ static int __btrfs_prealloc_file_range(struct inode *inode, int mode,
                 * sized chunks.
                 */
                cur_bytes = min(cur_bytes, last_alloc);
-               ret = btrfs_reserve_extent(root, cur_bytes, min_size, 0,
-                                          *alloc_hint, &ins, 1, 0);
+               ret = btrfs_reserve_extent(root, cur_bytes, cur_bytes,
+                               min_size, 0, *alloc_hint, &ins, 1, 0);
                if (ret) {
                        if (own_trans)
                                btrfs_end_transaction(trans, root);
@@ -10414,6 +10428,9 @@ next:
                if (own_trans)
                        btrfs_end_transaction(trans, root);
        }
+       if (cur_offset < end)
+               btrfs_free_reserved_data_space(inode, cur_offset,
+                       end - cur_offset + 1);
        return ret;
 }
 
index 14ed1e9..b2a2da5 100644 (file)
@@ -5084,7 +5084,7 @@ static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
        if (!capable(CAP_SYS_ADMIN))
                return -EPERM;
 
-       return btrfs_qgroup_wait_for_completion(root->fs_info);
+       return btrfs_qgroup_wait_for_completion(root->fs_info, true);
 }
 
 static long _btrfs_ioctl_set_received_subvol(struct file *file,
index 93ee1c1..8db2e29 100644 (file)
@@ -995,7 +995,7 @@ int btrfs_quota_disable(struct btrfs_trans_handle *trans,
                goto out;
        fs_info->quota_enabled = 0;
        fs_info->pending_quota_state = 0;
-       btrfs_qgroup_wait_for_completion(fs_info);
+       btrfs_qgroup_wait_for_completion(fs_info, false);
        spin_lock(&fs_info->qgroup_lock);
        quota_root = fs_info->quota_root;
        fs_info->quota_root = NULL;
@@ -1453,10 +1453,9 @@ int btrfs_qgroup_prepare_account_extents(struct btrfs_trans_handle *trans,
        return ret;
 }
 
-struct btrfs_qgroup_extent_record *
-btrfs_qgroup_insert_dirty_extent(struct btrfs_fs_info *fs_info,
-                                struct btrfs_delayed_ref_root *delayed_refs,
-                                struct btrfs_qgroup_extent_record *record)
+int btrfs_qgroup_insert_dirty_extent_nolock(struct btrfs_fs_info *fs_info,
+                               struct btrfs_delayed_ref_root *delayed_refs,
+                               struct btrfs_qgroup_extent_record *record)
 {
        struct rb_node **p = &delayed_refs->dirty_extent_root.rb_node;
        struct rb_node *parent_node = NULL;
@@ -1475,12 +1474,42 @@ btrfs_qgroup_insert_dirty_extent(struct btrfs_fs_info *fs_info,
                else if (bytenr > entry->bytenr)
                        p = &(*p)->rb_right;
                else
-                       return entry;
+                       return 1;
        }
 
        rb_link_node(&record->node, parent_node, p);
        rb_insert_color(&record->node, &delayed_refs->dirty_extent_root);
-       return NULL;
+       return 0;
+}
+
+int btrfs_qgroup_insert_dirty_extent(struct btrfs_trans_handle *trans,
+               struct btrfs_fs_info *fs_info, u64 bytenr, u64 num_bytes,
+               gfp_t gfp_flag)
+{
+       struct btrfs_qgroup_extent_record *record;
+       struct btrfs_delayed_ref_root *delayed_refs;
+       int ret;
+
+       if (!fs_info->quota_enabled || bytenr == 0 || num_bytes == 0)
+               return 0;
+       if (WARN_ON(trans == NULL))
+               return -EINVAL;
+       record = kmalloc(sizeof(*record), gfp_flag);
+       if (!record)
+               return -ENOMEM;
+
+       delayed_refs = &trans->transaction->delayed_refs;
+       record->bytenr = bytenr;
+       record->num_bytes = num_bytes;
+       record->old_roots = NULL;
+
+       spin_lock(&delayed_refs->lock);
+       ret = btrfs_qgroup_insert_dirty_extent_nolock(fs_info, delayed_refs,
+                                                     record);
+       spin_unlock(&delayed_refs->lock);
+       if (ret > 0)
+               kfree(record);
+       return 0;
 }
 
 #define UPDATE_NEW     0
@@ -2303,6 +2332,10 @@ static void btrfs_qgroup_rescan_worker(struct btrfs_work *work)
        int err = -ENOMEM;
        int ret = 0;
 
+       mutex_lock(&fs_info->qgroup_rescan_lock);
+       fs_info->qgroup_rescan_running = true;
+       mutex_unlock(&fs_info->qgroup_rescan_lock);
+
        path = btrfs_alloc_path();
        if (!path)
                goto out;
@@ -2369,6 +2402,9 @@ out:
        }
 
 done:
+       mutex_lock(&fs_info->qgroup_rescan_lock);
+       fs_info->qgroup_rescan_running = false;
+       mutex_unlock(&fs_info->qgroup_rescan_lock);
        complete_all(&fs_info->qgroup_rescan_completion);
 }
 
@@ -2487,20 +2523,26 @@ btrfs_qgroup_rescan(struct btrfs_fs_info *fs_info)
        return 0;
 }
 
-int btrfs_qgroup_wait_for_completion(struct btrfs_fs_info *fs_info)
+int btrfs_qgroup_wait_for_completion(struct btrfs_fs_info *fs_info,
+                                    bool interruptible)
 {
        int running;
        int ret = 0;
 
        mutex_lock(&fs_info->qgroup_rescan_lock);
        spin_lock(&fs_info->qgroup_lock);
-       running = fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN;
+       running = fs_info->qgroup_rescan_running;
        spin_unlock(&fs_info->qgroup_lock);
        mutex_unlock(&fs_info->qgroup_rescan_lock);
 
-       if (running)
+       if (!running)
+               return 0;
+
+       if (interruptible)
                ret = wait_for_completion_interruptible(
                                        &fs_info->qgroup_rescan_completion);
+       else
+               wait_for_completion(&fs_info->qgroup_rescan_completion);
 
        return ret;
 }
index 710887c..1bc64c8 100644 (file)
@@ -46,7 +46,8 @@ int btrfs_quota_disable(struct btrfs_trans_handle *trans,
                        struct btrfs_fs_info *fs_info);
 int btrfs_qgroup_rescan(struct btrfs_fs_info *fs_info);
 void btrfs_qgroup_rescan_resume(struct btrfs_fs_info *fs_info);
-int btrfs_qgroup_wait_for_completion(struct btrfs_fs_info *fs_info);
+int btrfs_qgroup_wait_for_completion(struct btrfs_fs_info *fs_info,
+                                    bool interruptible);
 int btrfs_add_qgroup_relation(struct btrfs_trans_handle *trans,
                              struct btrfs_fs_info *fs_info, u64 src, u64 dst);
 int btrfs_del_qgroup_relation(struct btrfs_trans_handle *trans,
@@ -63,10 +64,35 @@ void btrfs_free_qgroup_config(struct btrfs_fs_info *fs_info);
 struct btrfs_delayed_extent_op;
 int btrfs_qgroup_prepare_account_extents(struct btrfs_trans_handle *trans,
                                         struct btrfs_fs_info *fs_info);
-struct btrfs_qgroup_extent_record *
-btrfs_qgroup_insert_dirty_extent(struct btrfs_fs_info *fs_info,
-                                struct btrfs_delayed_ref_root *delayed_refs,
-                                struct btrfs_qgroup_extent_record *record);
+/*
+ * Insert one dirty extent record into @delayed_refs, informing qgroup to
+ * account that extent at commit trans time.
+ *
+ * No lock version, caller must acquire delayed ref lock and allocate memory.
+ *
+ * Return 0 for success insert
+ * Return >0 for existing record, caller can free @record safely.
+ * Error is not possible
+ */
+int btrfs_qgroup_insert_dirty_extent_nolock(
+               struct btrfs_fs_info *fs_info,
+               struct btrfs_delayed_ref_root *delayed_refs,
+               struct btrfs_qgroup_extent_record *record);
+
+/*
+ * Insert one dirty extent record into @delayed_refs, informing qgroup to
+ * account that extent at commit trans time.
+ *
+ * Better encapsulated version.
+ *
+ * Return 0 if the operation is done.
+ * Return <0 for error, like memory allocation failure or invalid parameter
+ * (NULL trans)
+ */
+int btrfs_qgroup_insert_dirty_extent(struct btrfs_trans_handle *trans,
+               struct btrfs_fs_info *fs_info, u64 bytenr, u64 num_bytes,
+               gfp_t gfp_flag);
+
 int
 btrfs_qgroup_account_extent(struct btrfs_trans_handle *trans,
                            struct btrfs_fs_info *fs_info,
index b26a5ae..8a2c2a0 100644 (file)
@@ -31,6 +31,7 @@
 #include "async-thread.h"
 #include "free-space-cache.h"
 #include "inode-map.h"
+#include "qgroup.h"
 
 /*
  * backref_node, mapping_node and tree_block start with this
@@ -3037,15 +3038,19 @@ int prealloc_file_extent_cluster(struct inode *inode,
        u64 num_bytes;
        int nr = 0;
        int ret = 0;
+       u64 prealloc_start = cluster->start - offset;
+       u64 prealloc_end = cluster->end - offset;
+       u64 cur_offset;
 
        BUG_ON(cluster->start != cluster->boundary[0]);
        inode_lock(inode);
 
-       ret = btrfs_check_data_free_space(inode, cluster->start,
-                                         cluster->end + 1 - cluster->start);
+       ret = btrfs_check_data_free_space(inode, prealloc_start,
+                                         prealloc_end + 1 - prealloc_start);
        if (ret)
                goto out;
 
+       cur_offset = prealloc_start;
        while (nr < cluster->nr) {
                start = cluster->boundary[nr] - offset;
                if (nr + 1 < cluster->nr)
@@ -3055,16 +3060,21 @@ int prealloc_file_extent_cluster(struct inode *inode,
 
                lock_extent(&BTRFS_I(inode)->io_tree, start, end);
                num_bytes = end + 1 - start;
+               if (cur_offset < start)
+                       btrfs_free_reserved_data_space(inode, cur_offset,
+                                       start - cur_offset);
                ret = btrfs_prealloc_file_range(inode, 0, start,
                                                num_bytes, num_bytes,
                                                end + 1, &alloc_hint);
+               cur_offset = end + 1;
                unlock_extent(&BTRFS_I(inode)->io_tree, start, end);
                if (ret)
                        break;
                nr++;
        }
-       btrfs_free_reserved_data_space(inode, cluster->start,
-                                      cluster->end + 1 - cluster->start);
+       if (cur_offset < prealloc_end)
+               btrfs_free_reserved_data_space(inode, cur_offset,
+                                      prealloc_end + 1 - cur_offset);
 out:
        inode_unlock(inode);
        return ret;
@@ -3916,6 +3926,90 @@ int prepare_to_relocate(struct reloc_control *rc)
        return 0;
 }
 
+/*
+ * Qgroup fixer for data chunk relocation.
+ * The data relocation is done in the following steps
+ * 1) Copy data extents into data reloc tree
+ * 2) Create tree reloc tree(special snapshot) for related subvolumes
+ * 3) Modify file extents in tree reloc tree
+ * 4) Merge tree reloc tree with original fs tree, by swapping tree blocks
+ *
+ * The problem is, data and tree reloc tree are not accounted to qgroup,
+ * and 4) will only info qgroup to track tree blocks change, not file extents
+ * in the tree blocks.
+ *
+ * The good news is, related data extents are all in data reloc tree, so we
+ * only need to info qgroup to track all file extents in data reloc tree
+ * before commit trans.
+ */
+static int qgroup_fix_relocated_data_extents(struct btrfs_trans_handle *trans,
+                                            struct reloc_control *rc)
+{
+       struct btrfs_fs_info *fs_info = rc->extent_root->fs_info;
+       struct inode *inode = rc->data_inode;
+       struct btrfs_root *data_reloc_root = BTRFS_I(inode)->root;
+       struct btrfs_path *path;
+       struct btrfs_key key;
+       int ret = 0;
+
+       if (!fs_info->quota_enabled)
+               return 0;
+
+       /*
+        * Only for stage where we update data pointers the qgroup fix is
+        * valid.
+        * For MOVING_DATA stage, we will miss the timing of swapping tree
+        * blocks, and won't fix it.
+        */
+       if (!(rc->stage == UPDATE_DATA_PTRS && rc->extents_found))
+               return 0;
+
+       path = btrfs_alloc_path();
+       if (!path)
+               return -ENOMEM;
+       key.objectid = btrfs_ino(inode);
+       key.type = BTRFS_EXTENT_DATA_KEY;
+       key.offset = 0;
+
+       ret = btrfs_search_slot(NULL, data_reloc_root, &key, path, 0, 0);
+       if (ret < 0)
+               goto out;
+
+       lock_extent(&BTRFS_I(inode)->io_tree, 0, (u64)-1);
+       while (1) {
+               struct btrfs_file_extent_item *fi;
+
+               btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
+               if (key.objectid > btrfs_ino(inode))
+                       break;
+               if (key.type != BTRFS_EXTENT_DATA_KEY)
+                       goto next;
+               fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
+                                   struct btrfs_file_extent_item);
+               if (btrfs_file_extent_type(path->nodes[0], fi) !=
+                               BTRFS_FILE_EXTENT_REG)
+                       goto next;
+               ret = btrfs_qgroup_insert_dirty_extent(trans, fs_info,
+                       btrfs_file_extent_disk_bytenr(path->nodes[0], fi),
+                       btrfs_file_extent_disk_num_bytes(path->nodes[0], fi),
+                       GFP_NOFS);
+               if (ret < 0)
+                       break;
+next:
+               ret = btrfs_next_item(data_reloc_root, path);
+               if (ret < 0)
+                       break;
+               if (ret > 0) {
+                       ret = 0;
+                       break;
+               }
+       }
+       unlock_extent(&BTRFS_I(inode)->io_tree, 0 , (u64)-1);
+out:
+       btrfs_free_path(path);
+       return ret;
+}
+
 static noinline_for_stack int relocate_block_group(struct reloc_control *rc)
 {
        struct rb_root blocks = RB_ROOT;
@@ -4102,10 +4196,16 @@ restart:
 
        /* get rid of pinned extents */
        trans = btrfs_join_transaction(rc->extent_root);
-       if (IS_ERR(trans))
+       if (IS_ERR(trans)) {
                err = PTR_ERR(trans);
-       else
-               btrfs_commit_transaction(trans, rc->extent_root);
+               goto out_free;
+       }
+       err = qgroup_fix_relocated_data_extents(trans, rc);
+       if (err < 0) {
+               btrfs_abort_transaction(trans, err);
+               goto out_free;
+       }
+       btrfs_commit_transaction(trans, rc->extent_root);
 out_free:
        btrfs_free_block_rsv(rc->extent_root, rc->block_rsv);
        btrfs_free_path(path);
@@ -4468,10 +4568,16 @@ int btrfs_recover_relocation(struct btrfs_root *root)
        unset_reloc_control(rc);
 
        trans = btrfs_join_transaction(rc->extent_root);
-       if (IS_ERR(trans))
+       if (IS_ERR(trans)) {
                err = PTR_ERR(trans);
-       else
-               err = btrfs_commit_transaction(trans, rc->extent_root);
+               goto out_free;
+       }
+       err = qgroup_fix_relocated_data_extents(trans, rc);
+       if (err < 0) {
+               btrfs_abort_transaction(trans, err);
+               goto out_free;
+       }
+       err = btrfs_commit_transaction(trans, rc->extent_root);
 out_free:
        kfree(rc);
 out:
index 7fd7e18..0912960 100644 (file)
@@ -272,6 +272,23 @@ int btrfs_find_orphan_roots(struct btrfs_root *tree_root)
                root_key.objectid = key.offset;
                key.offset++;
 
+               /*
+                * The root might have been inserted already, as before we look
+                * for orphan roots, log replay might have happened, which
+                * triggers a transaction commit and qgroup accounting, which
+                * in turn reads and inserts fs roots while doing backref
+                * walking.
+                */
+               root = btrfs_lookup_fs_root(tree_root->fs_info,
+                                           root_key.objectid);
+               if (root) {
+                       WARN_ON(!test_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED,
+                                         &root->state));
+                       if (btrfs_root_refs(&root->root_item) == 0)
+                               btrfs_add_dead_root(root);
+                       continue;
+               }
+
                root = btrfs_read_fs_root(tree_root, &root_key);
                err = PTR_ERR_OR_ZERO(root);
                if (err && err != -ENOENT) {
@@ -310,16 +327,8 @@ int btrfs_find_orphan_roots(struct btrfs_root *tree_root)
                set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state);
 
                err = btrfs_insert_fs_root(root->fs_info, root);
-               /*
-                * The root might have been inserted already, as before we look
-                * for orphan roots, log replay might have happened, which
-                * triggers a transaction commit and qgroup accounting, which
-                * in turn reads and inserts fs roots while doing backref
-                * walking.
-                */
-               if (err == -EEXIST)
-                       err = 0;
                if (err) {
+                       BUG_ON(err == -EEXIST);
                        btrfs_free_fs_root(root);
                        break;
                }
index 864ce33..4071fe2 100644 (file)
@@ -2241,6 +2241,13 @@ static int btrfs_freeze(struct super_block *sb)
        struct btrfs_trans_handle *trans;
        struct btrfs_root *root = btrfs_sb(sb)->tree_root;
 
+       root->fs_info->fs_frozen = 1;
+       /*
+        * We don't need a barrier here, we'll wait for any transaction that
+        * could be in progress on other threads (and do delayed iputs that
+        * we want to avoid on a frozen filesystem), or do the commit
+        * ourselves.
+        */
        trans = btrfs_attach_transaction_barrier(root);
        if (IS_ERR(trans)) {
                /* no transaction, don't bother */
@@ -2251,6 +2258,14 @@ static int btrfs_freeze(struct super_block *sb)
        return btrfs_commit_transaction(trans, root);
 }
 
+static int btrfs_unfreeze(struct super_block *sb)
+{
+       struct btrfs_root *root = btrfs_sb(sb)->tree_root;
+
+       root->fs_info->fs_frozen = 0;
+       return 0;
+}
+
 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
 {
        struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
@@ -2299,6 +2314,7 @@ static const struct super_operations btrfs_super_ops = {
        .statfs         = btrfs_statfs,
        .remount_fs     = btrfs_remount,
        .freeze_fs      = btrfs_freeze,
+       .unfreeze_fs    = btrfs_unfreeze,
 };
 
 static const struct file_operations btrfs_ctl_fops = {
index 9cca0a7..95d4191 100644 (file)
@@ -2278,8 +2278,13 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 
        kmem_cache_free(btrfs_trans_handle_cachep, trans);
 
+       /*
+        * If fs has been frozen, we can not handle delayed iputs, otherwise
+        * it'll result in deadlock about SB_FREEZE_FS.
+        */
        if (current != root->fs_info->transaction_kthread &&
-           current != root->fs_info->cleaner_kthread)
+           current != root->fs_info->cleaner_kthread &&
+           !root->fs_info->fs_frozen)
                btrfs_run_delayed_iputs(root);
 
        return ret;
index fff3f3e..e935035 100644 (file)
@@ -27,6 +27,7 @@
 #include "backref.h"
 #include "hash.h"
 #include "compression.h"
+#include "qgroup.h"
 
 /* magic values for the inode_only field in btrfs_log_inode:
  *
@@ -680,6 +681,21 @@ static noinline int replay_one_extent(struct btrfs_trans_handle *trans,
                ins.type = BTRFS_EXTENT_ITEM_KEY;
                offset = key->offset - btrfs_file_extent_offset(eb, item);
 
+               /*
+                * Manually record dirty extent, as here we did a shallow
+                * file extent item copy and skip normal backref update,
+                * but modifying extent tree all by ourselves.
+                * So need to manually record dirty extent for qgroup,
+                * as the owner of the file extent changed from log tree
+                * (doesn't affect qgroup) to fs/file tree(affects qgroup)
+                */
+               ret = btrfs_qgroup_insert_dirty_extent(trans, root->fs_info,
+                               btrfs_file_extent_disk_bytenr(eb, item),
+                               btrfs_file_extent_disk_num_bytes(eb, item),
+                               GFP_NOFS);
+               if (ret < 0)
+                       goto out;
+
                if (ins.objectid > 0) {
                        u64 csum_start;
                        u64 csum_end;
@@ -2807,7 +2823,7 @@ int btrfs_sync_log(struct btrfs_trans_handle *trans,
         */
        mutex_unlock(&root->log_mutex);
 
-       btrfs_init_log_ctx(&root_log_ctx);
+       btrfs_init_log_ctx(&root_log_ctx, NULL);
 
        mutex_lock(&log_root_tree->log_mutex);
        atomic_inc(&log_root_tree->log_batch);
@@ -4741,7 +4757,8 @@ again:
                        if (ret < 0) {
                                err = ret;
                                goto out_unlock;
-                       } else if (ret > 0) {
+                       } else if (ret > 0 && ctx &&
+                                  other_ino != btrfs_ino(ctx->inode)) {
                                struct btrfs_key inode_key;
                                struct inode *other_inode;
 
index a9f1b75..ab858e3 100644 (file)
@@ -30,15 +30,18 @@ struct btrfs_log_ctx {
        int log_transid;
        int io_err;
        bool log_new_dentries;
+       struct inode *inode;
        struct list_head list;
 };
 
-static inline void btrfs_init_log_ctx(struct btrfs_log_ctx *ctx)
+static inline void btrfs_init_log_ctx(struct btrfs_log_ctx *ctx,
+                                     struct inode *inode)
 {
        ctx->log_ret = 0;
        ctx->log_transid = 0;
        ctx->io_err = 0;
        ctx->log_new_dentries = false;
+       ctx->inode = inode;
        INIT_LIST_HEAD(&ctx->list);
 }
 
index 51f1255..035efce 100644 (file)
@@ -834,10 +834,6 @@ static void __free_device(struct work_struct *work)
        struct btrfs_device *device;
 
        device = container_of(work, struct btrfs_device, rcu_work);
-
-       if (device->bdev)
-               blkdev_put(device->bdev, device->mode);
-
        rcu_string_free(device->name);
        kfree(device);
 }
@@ -852,6 +848,17 @@ static void free_device(struct rcu_head *head)
        schedule_work(&device->rcu_work);
 }
 
+static void btrfs_close_bdev(struct btrfs_device *device)
+{
+       if (device->bdev && device->writeable) {
+               sync_blockdev(device->bdev);
+               invalidate_bdev(device->bdev);
+       }
+
+       if (device->bdev)
+               blkdev_put(device->bdev, device->mode);
+}
+
 static void btrfs_close_one_device(struct btrfs_device *device)
 {
        struct btrfs_fs_devices *fs_devices = device->fs_devices;
@@ -870,10 +877,7 @@ static void btrfs_close_one_device(struct btrfs_device *device)
        if (device->missing)
                fs_devices->missing_devices--;
 
-       if (device->bdev && device->writeable) {
-               sync_blockdev(device->bdev);
-               invalidate_bdev(device->bdev);
-       }
+       btrfs_close_bdev(device);
 
        new_device = btrfs_alloc_device(NULL, &device->devid,
                                        device->uuid);
@@ -1932,6 +1936,8 @@ int btrfs_rm_device(struct btrfs_root *root, char *device_path, u64 devid)
                btrfs_sysfs_rm_device_link(root->fs_info->fs_devices, device);
        }
 
+       btrfs_close_bdev(device);
+
        call_rcu(&device->rcu, free_device);
 
        num_devices = btrfs_super_num_devices(root->fs_info->super_copy) - 1;
@@ -2025,6 +2031,9 @@ void btrfs_rm_dev_replace_free_srcdev(struct btrfs_fs_info *fs_info,
                /* zero out the old super if it is writable */
                btrfs_scratch_superblocks(srcdev->bdev, srcdev->name->str);
        }
+
+       btrfs_close_bdev(srcdev);
+
        call_rcu(&srcdev->rcu, free_device);
 
        /*
@@ -2080,6 +2089,8 @@ void btrfs_destroy_dev_replace_tgtdev(struct btrfs_fs_info *fs_info,
         * the device_list_mutex lock.
         */
        btrfs_scratch_superblocks(tgtdev->bdev, tgtdev->name->str);
+
+       btrfs_close_bdev(tgtdev);
        call_rcu(&tgtdev->rcu, free_device);
 }
 
index eea6491..466f7d6 100644 (file)
@@ -607,20 +607,54 @@ static const struct file_operations format2_fops;
 static const struct file_operations format3_fops;
 static const struct file_operations format4_fops;
 
-static int table_open(struct inode *inode, struct file *file)
+static int table_open1(struct inode *inode, struct file *file)
 {
        struct seq_file *seq;
-       int ret = -1;
+       int ret;
 
-       if (file->f_op == &format1_fops)
-               ret = seq_open(file, &format1_seq_ops);
-       else if (file->f_op == &format2_fops)
-               ret = seq_open(file, &format2_seq_ops);
-       else if (file->f_op == &format3_fops)
-               ret = seq_open(file, &format3_seq_ops);
-       else if (file->f_op == &format4_fops)
-               ret = seq_open(file, &format4_seq_ops);
+       ret = seq_open(file, &format1_seq_ops);
+       if (ret)
+               return ret;
+
+       seq = file->private_data;
+       seq->private = inode->i_private; /* the dlm_ls */
+       return 0;
+}
+
+static int table_open2(struct inode *inode, struct file *file)
+{
+       struct seq_file *seq;
+       int ret;
+
+       ret = seq_open(file, &format2_seq_ops);
+       if (ret)
+               return ret;
+
+       seq = file->private_data;
+       seq->private = inode->i_private; /* the dlm_ls */
+       return 0;
+}
+
+static int table_open3(struct inode *inode, struct file *file)
+{
+       struct seq_file *seq;
+       int ret;
+
+       ret = seq_open(file, &format3_seq_ops);
+       if (ret)
+               return ret;
+
+       seq = file->private_data;
+       seq->private = inode->i_private; /* the dlm_ls */
+       return 0;
+}
+
+static int table_open4(struct inode *inode, struct file *file)
+{
+       struct seq_file *seq;
+       int ret;
 
+       ret = seq_open(file, &format4_seq_ops);
        if (ret)
                return ret;
 
@@ -631,7 +665,7 @@ static int table_open(struct inode *inode, struct file *file)
 
 static const struct file_operations format1_fops = {
        .owner   = THIS_MODULE,
-       .open    = table_open,
+       .open    = table_open1,
        .read    = seq_read,
        .llseek  = seq_lseek,
        .release = seq_release
@@ -639,7 +673,7 @@ static const struct file_operations format1_fops = {
 
 static const struct file_operations format2_fops = {
        .owner   = THIS_MODULE,
-       .open    = table_open,
+       .open    = table_open2,
        .read    = seq_read,
        .llseek  = seq_lseek,
        .release = seq_release
@@ -647,7 +681,7 @@ static const struct file_operations format2_fops = {
 
 static const struct file_operations format3_fops = {
        .owner   = THIS_MODULE,
-       .open    = table_open,
+       .open    = table_open3,
        .read    = seq_read,
        .llseek  = seq_lseek,
        .release = seq_release
@@ -655,7 +689,7 @@ static const struct file_operations format3_fops = {
 
 static const struct file_operations format4_fops = {
        .owner   = THIS_MODULE,
-       .open    = table_open,
+       .open    = table_open4,
        .read    = seq_read,
        .llseek  = seq_lseek,
        .release = seq_release
index d64d2a5..ccb401e 100644 (file)
@@ -1699,11 +1699,11 @@ static int f2fs_write_end(struct file *file,
        trace_f2fs_write_end(inode, pos, len, copied);
 
        set_page_dirty(page);
-       f2fs_put_page(page, 1);
 
        if (pos + copied > i_size_read(inode))
                f2fs_i_size_write(inode, pos + copied);
 
+       f2fs_put_page(page, 1);
        f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
        return copied;
 }
index 675fa79..14f5fe2 100644 (file)
@@ -538,7 +538,7 @@ struct f2fs_nm_info {
        /* NAT cache management */
        struct radix_tree_root nat_root;/* root of the nat entry cache */
        struct radix_tree_root nat_set_root;/* root of the nat set cache */
-       struct percpu_rw_semaphore nat_tree_lock;       /* protect nat_tree_lock */
+       struct rw_semaphore nat_tree_lock;      /* protect nat_tree_lock */
        struct list_head nat_entries;   /* cached nat entry list (clean) */
        unsigned int nat_cnt;           /* the # of cached nat entries */
        unsigned int dirty_nat_cnt;     /* total num of nat entries in set */
@@ -787,7 +787,7 @@ struct f2fs_sb_info {
        struct f2fs_checkpoint *ckpt;           /* raw checkpoint pointer */
        struct inode *meta_inode;               /* cache meta blocks */
        struct mutex cp_mutex;                  /* checkpoint procedure lock */
-       struct percpu_rw_semaphore cp_rwsem;            /* blocking FS operations */
+       struct rw_semaphore cp_rwsem;           /* blocking FS operations */
        struct rw_semaphore node_write;         /* locking node writes */
        wait_queue_head_t cp_wait;
        unsigned long last_time[MAX_TIME];      /* to store time in jiffies */
@@ -1074,22 +1074,22 @@ static inline void clear_ckpt_flags(struct f2fs_checkpoint *cp, unsigned int f)
 
 static inline void f2fs_lock_op(struct f2fs_sb_info *sbi)
 {
-       percpu_down_read(&sbi->cp_rwsem);
+       down_read(&sbi->cp_rwsem);
 }
 
 static inline void f2fs_unlock_op(struct f2fs_sb_info *sbi)
 {
-       percpu_up_read(&sbi->cp_rwsem);
+       up_read(&sbi->cp_rwsem);
 }
 
 static inline void f2fs_lock_all(struct f2fs_sb_info *sbi)
 {
-       percpu_down_write(&sbi->cp_rwsem);
+       down_write(&sbi->cp_rwsem);
 }
 
 static inline void f2fs_unlock_all(struct f2fs_sb_info *sbi)
 {
-       percpu_up_write(&sbi->cp_rwsem);
+       up_write(&sbi->cp_rwsem);
 }
 
 static inline int __get_cp_reason(struct f2fs_sb_info *sbi)
index 0e493f6..47abb96 100644 (file)
@@ -2086,15 +2086,19 @@ static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
        if (unlikely(f2fs_readonly(src->i_sb)))
                return -EROFS;
 
-       if (S_ISDIR(src->i_mode) || S_ISDIR(dst->i_mode))
-               return -EISDIR;
+       if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
+               return -EINVAL;
 
        if (f2fs_encrypted_inode(src) || f2fs_encrypted_inode(dst))
                return -EOPNOTSUPP;
 
        inode_lock(src);
-       if (src != dst)
-               inode_lock(dst);
+       if (src != dst) {
+               if (!inode_trylock(dst)) {
+                       ret = -EBUSY;
+                       goto out;
+               }
+       }
 
        ret = -EINVAL;
        if (pos_in + len > src->i_size || pos_in + len < pos_in)
@@ -2152,6 +2156,7 @@ static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
 out_unlock:
        if (src != dst)
                inode_unlock(dst);
+out:
        inode_unlock(src);
        return ret;
 }
index b2fa4b6..f75d197 100644 (file)
@@ -206,14 +206,14 @@ int need_dentry_mark(struct f2fs_sb_info *sbi, nid_t nid)
        struct nat_entry *e;
        bool need = false;
 
-       percpu_down_read(&nm_i->nat_tree_lock);
+       down_read(&nm_i->nat_tree_lock);
        e = __lookup_nat_cache(nm_i, nid);
        if (e) {
                if (!get_nat_flag(e, IS_CHECKPOINTED) &&
                                !get_nat_flag(e, HAS_FSYNCED_INODE))
                        need = true;
        }
-       percpu_up_read(&nm_i->nat_tree_lock);
+       up_read(&nm_i->nat_tree_lock);
        return need;
 }
 
@@ -223,11 +223,11 @@ bool is_checkpointed_node(struct f2fs_sb_info *sbi, nid_t nid)
        struct nat_entry *e;
        bool is_cp = true;
 
-       percpu_down_read(&nm_i->nat_tree_lock);
+       down_read(&nm_i->nat_tree_lock);
        e = __lookup_nat_cache(nm_i, nid);
        if (e && !get_nat_flag(e, IS_CHECKPOINTED))
                is_cp = false;
-       percpu_up_read(&nm_i->nat_tree_lock);
+       up_read(&nm_i->nat_tree_lock);
        return is_cp;
 }
 
@@ -237,13 +237,13 @@ bool need_inode_block_update(struct f2fs_sb_info *sbi, nid_t ino)
        struct nat_entry *e;
        bool need_update = true;
 
-       percpu_down_read(&nm_i->nat_tree_lock);
+       down_read(&nm_i->nat_tree_lock);
        e = __lookup_nat_cache(nm_i, ino);
        if (e && get_nat_flag(e, HAS_LAST_FSYNC) &&
                        (get_nat_flag(e, IS_CHECKPOINTED) ||
                         get_nat_flag(e, HAS_FSYNCED_INODE)))
                need_update = false;
-       percpu_up_read(&nm_i->nat_tree_lock);
+       up_read(&nm_i->nat_tree_lock);
        return need_update;
 }
 
@@ -284,7 +284,7 @@ static void set_node_addr(struct f2fs_sb_info *sbi, struct node_info *ni,
        struct f2fs_nm_info *nm_i = NM_I(sbi);
        struct nat_entry *e;
 
-       percpu_down_write(&nm_i->nat_tree_lock);
+       down_write(&nm_i->nat_tree_lock);
        e = __lookup_nat_cache(nm_i, ni->nid);
        if (!e) {
                e = grab_nat_entry(nm_i, ni->nid);
@@ -334,7 +334,7 @@ static void set_node_addr(struct f2fs_sb_info *sbi, struct node_info *ni,
                        set_nat_flag(e, HAS_FSYNCED_INODE, true);
                set_nat_flag(e, HAS_LAST_FSYNC, fsync_done);
        }
-       percpu_up_write(&nm_i->nat_tree_lock);
+       up_write(&nm_i->nat_tree_lock);
 }
 
 int try_to_free_nats(struct f2fs_sb_info *sbi, int nr_shrink)
@@ -342,7 +342,8 @@ int try_to_free_nats(struct f2fs_sb_info *sbi, int nr_shrink)
        struct f2fs_nm_info *nm_i = NM_I(sbi);
        int nr = nr_shrink;
 
-       percpu_down_write(&nm_i->nat_tree_lock);
+       if (!down_write_trylock(&nm_i->nat_tree_lock))
+               return 0;
 
        while (nr_shrink && !list_empty(&nm_i->nat_entries)) {
                struct nat_entry *ne;
@@ -351,7 +352,7 @@ int try_to_free_nats(struct f2fs_sb_info *sbi, int nr_shrink)
                __del_from_nat_cache(nm_i, ne);
                nr_shrink--;
        }
-       percpu_up_write(&nm_i->nat_tree_lock);
+       up_write(&nm_i->nat_tree_lock);
        return nr - nr_shrink;
 }
 
@@ -373,13 +374,13 @@ void get_node_info(struct f2fs_sb_info *sbi, nid_t nid, struct node_info *ni)
        ni->nid = nid;
 
        /* Check nat cache */
-       percpu_down_read(&nm_i->nat_tree_lock);
+       down_read(&nm_i->nat_tree_lock);
        e = __lookup_nat_cache(nm_i, nid);
        if (e) {
                ni->ino = nat_get_ino(e);
                ni->blk_addr = nat_get_blkaddr(e);
                ni->version = nat_get_version(e);
-               percpu_up_read(&nm_i->nat_tree_lock);
+               up_read(&nm_i->nat_tree_lock);
                return;
        }
 
@@ -403,11 +404,11 @@ void get_node_info(struct f2fs_sb_info *sbi, nid_t nid, struct node_info *ni)
        node_info_from_raw_nat(ni, &ne);
        f2fs_put_page(page, 1);
 cache:
-       percpu_up_read(&nm_i->nat_tree_lock);
+       up_read(&nm_i->nat_tree_lock);
        /* cache nat entry */
-       percpu_down_write(&nm_i->nat_tree_lock);
+       down_write(&nm_i->nat_tree_lock);
        cache_nat_entry(sbi, nid, &ne);
-       percpu_up_write(&nm_i->nat_tree_lock);
+       up_write(&nm_i->nat_tree_lock);
 }
 
 /*
@@ -1788,7 +1789,7 @@ void build_free_nids(struct f2fs_sb_info *sbi)
        ra_meta_pages(sbi, NAT_BLOCK_OFFSET(nid), FREE_NID_PAGES,
                                                        META_NAT, true);
 
-       percpu_down_read(&nm_i->nat_tree_lock);
+       down_read(&nm_i->nat_tree_lock);
 
        while (1) {
                struct page *page = get_current_nat_page(sbi, nid);
@@ -1820,7 +1821,7 @@ void build_free_nids(struct f2fs_sb_info *sbi)
                        remove_free_nid(nm_i, nid);
        }
        up_read(&curseg->journal_rwsem);
-       percpu_up_read(&nm_i->nat_tree_lock);
+       up_read(&nm_i->nat_tree_lock);
 
        ra_meta_pages(sbi, NAT_BLOCK_OFFSET(nm_i->next_scan_nid),
                                        nm_i->ra_nid_pages, META_NAT, false);
@@ -2209,7 +2210,7 @@ void flush_nat_entries(struct f2fs_sb_info *sbi)
        if (!nm_i->dirty_nat_cnt)
                return;
 
-       percpu_down_write(&nm_i->nat_tree_lock);
+       down_write(&nm_i->nat_tree_lock);
 
        /*
         * if there are no enough space in journal to store dirty nat
@@ -2232,7 +2233,7 @@ void flush_nat_entries(struct f2fs_sb_info *sbi)
        list_for_each_entry_safe(set, tmp, &sets, set_list)
                __flush_nat_entry_set(sbi, set);
 
-       percpu_up_write(&nm_i->nat_tree_lock);
+       up_write(&nm_i->nat_tree_lock);
 
        f2fs_bug_on(sbi, nm_i->dirty_nat_cnt);
 }
@@ -2268,8 +2269,7 @@ static int init_node_manager(struct f2fs_sb_info *sbi)
 
        mutex_init(&nm_i->build_lock);
        spin_lock_init(&nm_i->free_nid_list_lock);
-       if (percpu_init_rwsem(&nm_i->nat_tree_lock))
-               return -ENOMEM;
+       init_rwsem(&nm_i->nat_tree_lock);
 
        nm_i->next_scan_nid = le32_to_cpu(sbi->ckpt->next_free_nid);
        nm_i->bitmap_size = __bitmap_size(sbi, NAT_BITMAP);
@@ -2326,7 +2326,7 @@ void destroy_node_manager(struct f2fs_sb_info *sbi)
        spin_unlock(&nm_i->free_nid_list_lock);
 
        /* destroy nat cache */
-       percpu_down_write(&nm_i->nat_tree_lock);
+       down_write(&nm_i->nat_tree_lock);
        while ((found = __gang_lookup_nat_cache(nm_i,
                                        nid, NATVEC_SIZE, natvec))) {
                unsigned idx;
@@ -2351,9 +2351,8 @@ void destroy_node_manager(struct f2fs_sb_info *sbi)
                        kmem_cache_free(nat_entry_set_slab, setvec[idx]);
                }
        }
-       percpu_up_write(&nm_i->nat_tree_lock);
+       up_write(&nm_i->nat_tree_lock);
 
-       percpu_free_rwsem(&nm_i->nat_tree_lock);
        kfree(nm_i->nat_bitmap);
        sbi->nm_info = NULL;
        kfree(nm_i);
index 1b86d3f..7f863a6 100644 (file)
@@ -706,8 +706,6 @@ static void destroy_percpu_info(struct f2fs_sb_info *sbi)
                percpu_counter_destroy(&sbi->nr_pages[i]);
        percpu_counter_destroy(&sbi->alloc_valid_block_count);
        percpu_counter_destroy(&sbi->total_valid_inode_count);
-
-       percpu_free_rwsem(&sbi->cp_rwsem);
 }
 
 static void f2fs_put_super(struct super_block *sb)
@@ -1483,9 +1481,6 @@ static int init_percpu_info(struct f2fs_sb_info *sbi)
 {
        int i, err;
 
-       if (percpu_init_rwsem(&sbi->cp_rwsem))
-               return -ENOMEM;
-
        for (i = 0; i < NR_COUNT_TYPE; i++) {
                err = percpu_counter_init(&sbi->nr_pages[i], 0, GFP_KERNEL);
                if (err)
@@ -1686,6 +1681,7 @@ try_onemore:
                sbi->write_io[i].bio = NULL;
        }
 
+       init_rwsem(&sbi->cp_rwsem);
        init_waitqueue_head(&sbi->cp_wait);
        init_sb_info(sbi);
 
index 48141b8..0342254 100644 (file)
@@ -84,8 +84,11 @@ iomap_apply(struct inode *inode, loff_t pos, loff_t length, unsigned flags,
         * Now the data has been copied, commit the range we've copied.  This
         * should not fail unless the filesystem has had a fatal error.
         */
-       ret = ops->iomap_end(inode, pos, length, written > 0 ? written : 0,
-                       flags, &iomap);
+       if (ops->iomap_end) {
+               ret = ops->iomap_end(inode, pos, length,
+                                    written > 0 ? written : 0,
+                                    flags, &iomap);
+       }
 
        return written ? written : ret;
 }
@@ -194,12 +197,9 @@ again:
                if (mapping_writably_mapped(inode->i_mapping))
                        flush_dcache_page(page);
 
-               pagefault_disable();
                copied = iov_iter_copy_from_user_atomic(page, i, offset, bytes);
-               pagefault_enable();
 
                flush_dcache_page(page);
-               mark_page_accessed(page);
 
                status = iomap_write_end(inode, pos, bytes, copied, page);
                if (unlikely(status < 0))
@@ -470,13 +470,18 @@ int iomap_fiemap(struct inode *inode, struct fiemap_extent_info *fi,
        if (ret)
                return ret;
 
-       ret = filemap_write_and_wait(inode->i_mapping);
-       if (ret)
-               return ret;
+       if (fi->fi_flags & FIEMAP_FLAG_SYNC) {
+               ret = filemap_write_and_wait(inode->i_mapping);
+               if (ret)
+                       return ret;
+       }
 
        while (len > 0) {
                ret = iomap_apply(inode, start, len, 0, ops, &ctx,
                                iomap_fiemap_actor);
+               /* inode with no (attribute) mapping will give ENOENT */
+               if (ret == -ENOENT)
+                       break;
                if (ret < 0)
                        return ret;
                if (ret == 0)
index 19f532e..6dc4296 100644 (file)
@@ -223,8 +223,10 @@ ssize_t seq_read(struct file *file, char __user *buf, size_t size, loff_t *ppos)
                size -= n;
                buf += n;
                copied += n;
-               if (!m->count)
+               if (!m->count) {
+                       m->from = 0;
                        m->index++;
+               }
                if (!size)
                        goto Done;
        }
index b45345d..51157da 100644 (file)
@@ -370,7 +370,7 @@ static int layout_in_gaps(struct ubifs_info *c, int cnt)
 
        p = c->gap_lebs;
        do {
-               ubifs_assert(p < c->gap_lebs + sizeof(int) * c->lst.idx_lebs);
+               ubifs_assert(p < c->gap_lebs + c->lst.idx_lebs);
                written = layout_leb_in_gaps(c, p);
                if (written < 0) {
                        err = written;
index e237811..11a0041 100644 (file)
@@ -575,7 +575,8 @@ static int ubifs_xattr_get(const struct xattr_handler *handler,
        dbg_gen("xattr '%s', ino %lu ('%pd'), buf size %zd", name,
                inode->i_ino, dentry, size);
 
-       return  __ubifs_getxattr(inode, name, buffer, size);
+       name = xattr_full_name(handler, name);
+       return __ubifs_getxattr(inode, name, buffer, size);
 }
 
 static int ubifs_xattr_set(const struct xattr_handler *handler,
@@ -586,6 +587,8 @@ static int ubifs_xattr_set(const struct xattr_handler *handler,
        dbg_gen("xattr '%s', host ino %lu ('%pd'), size %zd",
                name, inode->i_ino, dentry, size);
 
+       name = xattr_full_name(handler, name);
+
        if (value)
                return __ubifs_setxattr(inode, name, value, size, flags);
        else
index 776ae2f..3dd8f1d 100644 (file)
@@ -1582,6 +1582,7 @@ xfs_alloc_ag_vextent_small(
        xfs_extlen_t    *flenp, /* result length */
        int             *stat)  /* status: 0-freelist, 1-normal/none */
 {
+       struct xfs_owner_info   oinfo;
        int             error;
        xfs_agblock_t   fbno;
        xfs_extlen_t    flen;
@@ -1624,6 +1625,18 @@ xfs_alloc_ag_vextent_small(
                                error0);
                        args->wasfromfl = 1;
                        trace_xfs_alloc_small_freelist(args);
+
+                       /*
+                        * If we're feeding an AGFL block to something that
+                        * doesn't live in the free space, we need to clear
+                        * out the OWN_AG rmap.
+                        */
+                       xfs_rmap_ag_owner(&oinfo, XFS_RMAP_OWN_AG);
+                       error = xfs_rmap_free(args->tp, args->agbp, args->agno,
+                                       fbno, 1, &oinfo);
+                       if (error)
+                               goto error0;
+
                        *stat = 0;
                        return 0;
                }
@@ -2264,6 +2277,7 @@ xfs_alloc_log_agf(
                offsetof(xfs_agf_t, agf_longest),
                offsetof(xfs_agf_t, agf_btreeblks),
                offsetof(xfs_agf_t, agf_uuid),
+               offsetof(xfs_agf_t, agf_rmap_blocks),
                sizeof(xfs_agf_t)
        };
 
index f814d42..e6a8bea 100644 (file)
@@ -640,12 +640,15 @@ typedef struct xfs_agf {
        __be32          agf_btreeblks;  /* # of blocks held in AGF btrees */
        uuid_t          agf_uuid;       /* uuid of filesystem */
 
+       __be32          agf_rmap_blocks;        /* rmapbt blocks used */
+       __be32          agf_padding;            /* padding */
+
        /*
         * reserve some contiguous space for future logged fields before we add
         * the unlogged fields. This makes the range logging via flags and
         * structure offsets much simpler.
         */
-       __be64          agf_spare64[16];
+       __be64          agf_spare64[15];
 
        /* unlogged fields, written during buffer writeback. */
        __be64          agf_lsn;        /* last write sequence */
@@ -670,7 +673,8 @@ typedef struct xfs_agf {
 #define        XFS_AGF_LONGEST         0x00000400
 #define        XFS_AGF_BTREEBLKS       0x00000800
 #define        XFS_AGF_UUID            0x00001000
-#define        XFS_AGF_NUM_BITS        13
+#define        XFS_AGF_RMAP_BLOCKS     0x00002000
+#define        XFS_AGF_NUM_BITS        14
 #define        XFS_AGF_ALL_BITS        ((1 << XFS_AGF_NUM_BITS) - 1)
 
 #define XFS_AGF_FLAGS \
@@ -686,7 +690,8 @@ typedef struct xfs_agf {
        { XFS_AGF_FREEBLKS,     "FREEBLKS" }, \
        { XFS_AGF_LONGEST,      "LONGEST" }, \
        { XFS_AGF_BTREEBLKS,    "BTREEBLKS" }, \
-       { XFS_AGF_UUID,         "UUID" }
+       { XFS_AGF_UUID,         "UUID" }, \
+       { XFS_AGF_RMAP_BLOCKS,  "RMAP_BLOCKS" }
 
 /* disk block (xfs_daddr_t) in the AG */
 #define XFS_AGF_DADDR(mp)      ((xfs_daddr_t)(1 << (mp)->m_sectbb_log))
index bc1faeb..17b8eeb 100644 (file)
@@ -98,6 +98,8 @@ xfs_rmapbt_alloc_block(
        union xfs_btree_ptr     *new,
        int                     *stat)
 {
+       struct xfs_buf          *agbp = cur->bc_private.a.agbp;
+       struct xfs_agf          *agf = XFS_BUF_TO_AGF(agbp);
        int                     error;
        xfs_agblock_t           bno;
 
@@ -124,6 +126,8 @@ xfs_rmapbt_alloc_block(
 
        xfs_trans_agbtree_delta(cur->bc_tp, 1);
        new->s = cpu_to_be32(bno);
+       be32_add_cpu(&agf->agf_rmap_blocks, 1);
+       xfs_alloc_log_agf(cur->bc_tp, agbp, XFS_AGF_RMAP_BLOCKS);
 
        XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
        *stat = 1;
@@ -143,6 +147,8 @@ xfs_rmapbt_free_block(
        bno = xfs_daddr_to_agbno(cur->bc_mp, XFS_BUF_ADDR(bp));
        trace_xfs_rmapbt_free_block(cur->bc_mp, cur->bc_private.a.agno,
                        bno, 1);
+       be32_add_cpu(&agf->agf_rmap_blocks, -1);
+       xfs_alloc_log_agf(cur->bc_tp, agbp, XFS_AGF_RMAP_BLOCKS);
        error = xfs_alloc_put_freelist(cur->bc_tp, agbp, NULL, bno, 1);
        if (error)
                return error;
index 47a318c..607cc29 100644 (file)
@@ -115,7 +115,6 @@ xfs_buf_ioacct_dec(
        if (!(bp->b_flags & _XBF_IN_FLIGHT))
                return;
 
-       ASSERT(bp->b_flags & XBF_ASYNC);
        bp->b_flags &= ~_XBF_IN_FLIGHT;
        percpu_counter_dec(&bp->b_target->bt_io_count);
 }
index ed95e5b..e612a02 100644 (file)
@@ -741,9 +741,20 @@ xfs_file_dax_write(
         * page is inserted into the pagecache when we have to serve a write
         * fault on a hole.  It should never be dirtied and can simply be
         * dropped from the pagecache once we get real data for the page.
+        *
+        * XXX: This is racy against mmap, and there's nothing we can do about
+        * it. dax_do_io() should really do this invalidation internally as
+        * it will know if we've allocated over a holei for this specific IO and
+        * if so it needs to update the mapping tree and invalidate existing
+        * PTEs over the newly allocated range. Remove this invalidation when
+        * dax_do_io() is fixed up.
         */
        if (mapping->nrpages) {
-               ret = invalidate_inode_pages2(mapping);
+               loff_t end = iocb->ki_pos + iov_iter_count(from) - 1;
+
+               ret = invalidate_inode_pages2_range(mapping,
+                                                   iocb->ki_pos >> PAGE_SHIFT,
+                                                   end >> PAGE_SHIFT);
                WARN_ON_ONCE(ret);
        }
 
index 0f96847..0b7f986 100644 (file)
@@ -248,6 +248,7 @@ xfs_growfs_data_private(
                        agf->agf_roots[XFS_BTNUM_RMAPi] =
                                                cpu_to_be32(XFS_RMAP_BLOCK(mp));
                        agf->agf_levels[XFS_BTNUM_RMAPi] = cpu_to_be32(1);
+                       agf->agf_rmap_blocks = cpu_to_be32(1);
                }
 
                agf->agf_flfirst = cpu_to_be32(1);
index 2114d53..2af0dda 100644 (file)
@@ -715,12 +715,16 @@ xfs_iomap_write_allocate(
                 * is in the delayed allocation extent on which we sit
                 * but before our buffer starts.
                 */
-
                nimaps = 0;
                while (nimaps == 0) {
                        nres = XFS_EXTENTADD_SPACE_RES(mp, XFS_DATA_FORK);
-
-                       error = xfs_trans_alloc(mp, &M_RES(mp)->tr_write, nres,
+                       /*
+                        * We have already reserved space for the extent and any
+                        * indirect blocks when creating the delalloc extent,
+                        * there is no need to reserve space in this transaction
+                        * again.
+                        */
+                       error = xfs_trans_alloc(mp, &M_RES(mp)->tr_write, 0,
                                        0, XFS_TRANS_RESERVE, &tp);
                        if (error)
                                return error;
@@ -1037,20 +1041,14 @@ xfs_file_iomap_begin(
                        return error;
 
                trace_xfs_iomap_alloc(ip, offset, length, 0, &imap);
-               xfs_bmbt_to_iomap(ip, iomap, &imap);
-       } else if (nimaps) {
-               xfs_iunlock(ip, XFS_ILOCK_EXCL);
-               trace_xfs_iomap_found(ip, offset, length, 0, &imap);
-               xfs_bmbt_to_iomap(ip, iomap, &imap);
        } else {
+               ASSERT(nimaps);
+
                xfs_iunlock(ip, XFS_ILOCK_EXCL);
-               trace_xfs_iomap_not_found(ip, offset, length, 0, &imap);
-               iomap->blkno = IOMAP_NULL_BLOCK;
-               iomap->type = IOMAP_HOLE;
-               iomap->offset = offset;
-               iomap->length = length;
+               trace_xfs_iomap_found(ip, offset, length, 0, &imap);
        }
 
+       xfs_bmbt_to_iomap(ip, iomap, &imap);
        return 0;
 }
 
@@ -1112,3 +1110,48 @@ struct iomap_ops xfs_iomap_ops = {
        .iomap_begin            = xfs_file_iomap_begin,
        .iomap_end              = xfs_file_iomap_end,
 };
+
+static int
+xfs_xattr_iomap_begin(
+       struct inode            *inode,
+       loff_t                  offset,
+       loff_t                  length,
+       unsigned                flags,
+       struct iomap            *iomap)
+{
+       struct xfs_inode        *ip = XFS_I(inode);
+       struct xfs_mount        *mp = ip->i_mount;
+       xfs_fileoff_t           offset_fsb = XFS_B_TO_FSBT(mp, offset);
+       xfs_fileoff_t           end_fsb = XFS_B_TO_FSB(mp, offset + length);
+       struct xfs_bmbt_irec    imap;
+       int                     nimaps = 1, error = 0;
+       unsigned                lockmode;
+
+       if (XFS_FORCED_SHUTDOWN(mp))
+               return -EIO;
+
+       lockmode = xfs_ilock_data_map_shared(ip);
+
+       /* if there are no attribute fork or extents, return ENOENT */
+       if (XFS_IFORK_Q(ip) || !ip->i_d.di_anextents) {
+               error = -ENOENT;
+               goto out_unlock;
+       }
+
+       ASSERT(ip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL);
+       error = xfs_bmapi_read(ip, offset_fsb, end_fsb - offset_fsb, &imap,
+                              &nimaps, XFS_BMAPI_ENTIRE | XFS_BMAPI_ATTRFORK);
+out_unlock:
+       xfs_iunlock(ip, lockmode);
+
+       if (!error) {
+               ASSERT(nimaps);
+               xfs_bmbt_to_iomap(ip, iomap, &imap);
+       }
+
+       return error;
+}
+
+struct iomap_ops xfs_xattr_iomap_ops = {
+       .iomap_begin            = xfs_xattr_iomap_begin,
+};
index e066d04..fb8aca3 100644 (file)
@@ -35,5 +35,6 @@ void xfs_bmbt_to_iomap(struct xfs_inode *, struct iomap *,
                struct xfs_bmbt_irec *);
 
 extern struct iomap_ops xfs_iomap_ops;
+extern struct iomap_ops xfs_xattr_iomap_ops;
 
 #endif /* __XFS_IOMAP_H__*/
index ab820f8..b24c310 100644 (file)
@@ -1009,7 +1009,14 @@ xfs_vn_fiemap(
        int                     error;
 
        xfs_ilock(XFS_I(inode), XFS_IOLOCK_SHARED);
-       error = iomap_fiemap(inode, fieinfo, start, length, &xfs_iomap_ops);
+       if (fieinfo->fi_flags & FIEMAP_FLAG_XATTR) {
+               fieinfo->fi_flags &= ~FIEMAP_FLAG_XATTR;
+               error = iomap_fiemap(inode, fieinfo, start, length,
+                               &xfs_xattr_iomap_ops);
+       } else {
+               error = iomap_fiemap(inode, fieinfo, start, length,
+                               &xfs_iomap_ops);
+       }
        xfs_iunlock(XFS_I(inode), XFS_IOLOCK_SHARED);
 
        return error;
index 551b7e2..7e88bec 100644 (file)
@@ -1298,7 +1298,6 @@ DEFINE_IOMAP_EVENT(xfs_get_blocks_alloc);
 DEFINE_IOMAP_EVENT(xfs_get_blocks_map_direct);
 DEFINE_IOMAP_EVENT(xfs_iomap_alloc);
 DEFINE_IOMAP_EVENT(xfs_iomap_found);
-DEFINE_IOMAP_EVENT(xfs_iomap_not_found);
 
 DECLARE_EVENT_CLASS(xfs_simple_io_class,
        TP_PROTO(struct xfs_inode *ip, xfs_off_t offset, ssize_t count),
index 59ffaa6..23ddf4b 100644 (file)
@@ -71,7 +71,8 @@ static inline bool bio_has_data(struct bio *bio)
 {
        if (bio &&
            bio->bi_iter.bi_size &&
-           bio_op(bio) != REQ_OP_DISCARD)
+           bio_op(bio) != REQ_OP_DISCARD &&
+           bio_op(bio) != REQ_OP_SECURE_ERASE)
                return true;
 
        return false;
@@ -79,7 +80,9 @@ static inline bool bio_has_data(struct bio *bio)
 
 static inline bool bio_no_advance_iter(struct bio *bio)
 {
-       return bio_op(bio) == REQ_OP_DISCARD || bio_op(bio) == REQ_OP_WRITE_SAME;
+       return bio_op(bio) == REQ_OP_DISCARD ||
+              bio_op(bio) == REQ_OP_SECURE_ERASE ||
+              bio_op(bio) == REQ_OP_WRITE_SAME;
 }
 
 static inline bool bio_is_rw(struct bio *bio)
@@ -199,6 +202,9 @@ static inline unsigned bio_segments(struct bio *bio)
        if (bio_op(bio) == REQ_OP_DISCARD)
                return 1;
 
+       if (bio_op(bio) == REQ_OP_SECURE_ERASE)
+               return 1;
+
        if (bio_op(bio) == REQ_OP_WRITE_SAME)
                return 1;
 
index 2c210b6..e79055c 100644 (file)
@@ -882,7 +882,7 @@ static inline unsigned int blk_rq_cur_sectors(const struct request *rq)
 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q,
                                                     int op)
 {
-       if (unlikely(op == REQ_OP_DISCARD))
+       if (unlikely(op == REQ_OP_DISCARD || op == REQ_OP_SECURE_ERASE))
                return min(q->limits.max_discard_sectors, UINT_MAX >> 9);
 
        if (unlikely(op == REQ_OP_WRITE_SAME))
@@ -913,7 +913,9 @@ static inline unsigned int blk_rq_get_max_sectors(struct request *rq,
        if (unlikely(rq->cmd_type != REQ_TYPE_FS))
                return q->limits.max_hw_sectors;
 
-       if (!q->limits.chunk_sectors || (req_op(rq) == REQ_OP_DISCARD))
+       if (!q->limits.chunk_sectors ||
+           req_op(rq) == REQ_OP_DISCARD ||
+           req_op(rq) == REQ_OP_SECURE_ERASE)
                return blk_queue_get_max_sectors(q, req_op(rq));
 
        return min(blk_max_size_offset(q, offset),
index e294939..8dbc892 100644 (file)
  */
 #define asm_volatile_goto(x...)        do { asm goto(x); asm (""); } while (0)
 
-#ifdef CONFIG_ARCH_USE_BUILTIN_BSWAP
+/*
+ * sparse (__CHECKER__) pretends to be gcc, but can't do constant
+ * folding in __builtin_bswap*() (yet), so don't set these for it.
+ */
+#if defined(CONFIG_ARCH_USE_BUILTIN_BSWAP) && !defined(__CHECKER__)
 #if GCC_VERSION >= 40400
 #define __HAVE_BUILTIN_BSWAP32__
 #define __HAVE_BUILTIN_BSWAP64__
 #if GCC_VERSION >= 40800
 #define __HAVE_BUILTIN_BSWAP16__
 #endif
-#endif /* CONFIG_ARCH_USE_BUILTIN_BSWAP */
+#endif /* CONFIG_ARCH_USE_BUILTIN_BSWAP && !__CHECKER__ */
 
 #if GCC_VERSION >= 50000
 #define KASAN_ABI_VERSION 4
index 1bb9548..436aa4e 100644 (file)
@@ -527,13 +527,13 @@ static __always_inline void __write_once_size(volatile void *p, void *res, int s
  * object's lifetime is managed by something other than RCU.  That
  * "something other" might be reference counting or simple immortality.
  *
- * The seemingly unused void * variable is to validate @p is indeed a pointer
- * type. All pointer types silently cast to void *.
+ * The seemingly unused size_t variable is to validate @p is indeed a pointer
+ * type by making sure it can be dereferenced.
  */
 #define lockless_dereference(p) \
 ({ \
        typeof(p) _________p1 = READ_ONCE(p); \
-       __maybe_unused const void * const _________p2 = _________p1; \
+       size_t __maybe_unused __size_of_ptr = sizeof(*(p)); \
        smp_read_barrier_depends(); /* Dependency order vs. p above. */ \
        (_________p1); \
 })
index 2599a98..fbc1fa6 100644 (file)
@@ -1251,10 +1251,12 @@ resource_size_t pcibios_iov_resource_alignment(struct pci_dev *dev, int resno);
 int pci_set_vga_state(struct pci_dev *pdev, bool decode,
                      unsigned int command_bits, u32 flags);
 
-#define PCI_IRQ_NOLEGACY       (1 << 0) /* don't use legacy interrupts */
-#define PCI_IRQ_NOMSI          (1 << 1) /* don't use MSI interrupts */
-#define PCI_IRQ_NOMSIX         (1 << 2) /* don't use MSI-X interrupts */
-#define PCI_IRQ_NOAFFINITY     (1 << 3) /* don't auto-assign affinity */
+#define PCI_IRQ_LEGACY         (1 << 0) /* allow legacy interrupts */
+#define PCI_IRQ_MSI            (1 << 1) /* allow MSI interrupts */
+#define PCI_IRQ_MSIX           (1 << 2) /* allow MSI-X interrupts */
+#define PCI_IRQ_AFFINITY       (1 << 3) /* auto-assign affinity */
+#define PCI_IRQ_ALL_TYPES \
+       (PCI_IRQ_LEGACY | PCI_IRQ_MSI | PCI_IRQ_MSIX)
 
 /* kmem_cache style wrapper around pci_alloc_consistent() */
 
index 697e160..a4f7203 100644 (file)
@@ -42,6 +42,8 @@ extern int proc_dostring(struct ctl_table *, int,
                         void __user *, size_t *, loff_t *);
 extern int proc_dointvec(struct ctl_table *, int,
                         void __user *, size_t *, loff_t *);
+extern int proc_douintvec(struct ctl_table *, int,
+                        void __user *, size_t *, loff_t *);
 extern int proc_dointvec_minmax(struct ctl_table *, int,
                                void __user *, size_t *, loff_t *);
 extern int proc_dointvec_jiffies(struct ctl_table *, int,
index 8e90dd2..e1f9673 100644 (file)
@@ -2115,22 +2115,17 @@ static inline bool ib_is_udata_cleared(struct ib_udata *udata,
                                       size_t len)
 {
        const void __user *p = udata->inbuf + offset;
-       bool ret = false;
+       bool ret;
        u8 *buf;
 
        if (len > USHRT_MAX)
                return false;
 
-       buf = kmalloc(len, GFP_KERNEL);
-       if (!buf)
+       buf = memdup_user(p, len);
+       if (IS_ERR(buf))
                return false;
 
-       if (copy_from_user(buf, p, len))
-               goto free;
-
        ret = !memchr_inv(buf, 0, len);
-
-free:
        kfree(buf);
        return ret;
 }
index 9a37c54..b5486e6 100644 (file)
@@ -9,8 +9,8 @@
 
 DECLARE_PER_CPU(struct vcpu_info *, xen_vcpu);
 
-DECLARE_PER_CPU(int, xen_vcpu_id);
-static inline int xen_vcpu_nr(int cpu)
+DECLARE_PER_CPU(uint32_t, xen_vcpu_id);
+static inline uint32_t xen_vcpu_nr(int cpu)
 {
        return per_cpu(xen_vcpu_id, cpu);
 }
index 1903b8f..5650f53 100644 (file)
@@ -242,18 +242,6 @@ unlock:
        return ret;
 }
 
-static void event_function_local(struct perf_event *event, event_f func, void *data)
-{
-       struct event_function_struct efs = {
-               .event = event,
-               .func = func,
-               .data = data,
-       };
-
-       int ret = event_function(&efs);
-       WARN_ON_ONCE(ret);
-}
-
 static void event_function_call(struct perf_event *event, event_f func, void *data)
 {
        struct perf_event_context *ctx = event->ctx;
@@ -303,6 +291,54 @@ again:
        raw_spin_unlock_irq(&ctx->lock);
 }
 
+/*
+ * Similar to event_function_call() + event_function(), but hard assumes IRQs
+ * are already disabled and we're on the right CPU.
+ */
+static void event_function_local(struct perf_event *event, event_f func, void *data)
+{
+       struct perf_event_context *ctx = event->ctx;
+       struct perf_cpu_context *cpuctx = __get_cpu_context(ctx);
+       struct task_struct *task = READ_ONCE(ctx->task);
+       struct perf_event_context *task_ctx = NULL;
+
+       WARN_ON_ONCE(!irqs_disabled());
+
+       if (task) {
+               if (task == TASK_TOMBSTONE)
+                       return;
+
+               task_ctx = ctx;
+       }
+
+       perf_ctx_lock(cpuctx, task_ctx);
+
+       task = ctx->task;
+       if (task == TASK_TOMBSTONE)
+               goto unlock;
+
+       if (task) {
+               /*
+                * We must be either inactive or active and the right task,
+                * otherwise we're screwed, since we cannot IPI to somewhere
+                * else.
+                */
+               if (ctx->is_active) {
+                       if (WARN_ON_ONCE(task != current))
+                               goto unlock;
+
+                       if (WARN_ON_ONCE(cpuctx->task_ctx != ctx))
+                               goto unlock;
+               }
+       } else {
+               WARN_ON_ONCE(&cpuctx->ctx != ctx);
+       }
+
+       func(event, cpuctx, ctx, data);
+unlock:
+       perf_ctx_unlock(cpuctx, task_ctx);
+}
+
 #define PERF_FLAG_ALL (PERF_FLAG_FD_NO_GROUP |\
                       PERF_FLAG_FD_OUTPUT  |\
                       PERF_FLAG_PID_CGROUP |\
@@ -3513,9 +3549,10 @@ static int perf_event_read(struct perf_event *event, bool group)
                        .group = group,
                        .ret = 0,
                };
-               smp_call_function_single(event->oncpu,
-                                        __perf_event_read, &data, 1);
-               ret = data.ret;
+               ret = smp_call_function_single(event->oncpu, __perf_event_read, &data, 1);
+               /* The event must have been read from an online CPU: */
+               WARN_ON_ONCE(ret);
+               ret = ret ? : data.ret;
        } else if (event->state == PERF_EVENT_STATE_INACTIVE) {
                struct perf_event_context *ctx = event->ctx;
                unsigned long flags;
@@ -6583,15 +6620,6 @@ got_name:
        kfree(buf);
 }
 
-/*
- * Whether this @filter depends on a dynamic object which is not loaded
- * yet or its load addresses are not known.
- */
-static bool perf_addr_filter_needs_mmap(struct perf_addr_filter *filter)
-{
-       return filter->filter && filter->inode;
-}
-
 /*
  * Check whether inode and address range match filter criteria.
  */
@@ -6653,6 +6681,13 @@ static void perf_addr_filters_adjust(struct vm_area_struct *vma)
        struct perf_event_context *ctx;
        int ctxn;
 
+       /*
+        * Data tracing isn't supported yet and as such there is no need
+        * to keep track of anything that isn't related to executable code:
+        */
+       if (!(vma->vm_flags & VM_EXEC))
+               return;
+
        rcu_read_lock();
        for_each_task_context_nr(ctxn) {
                ctx = rcu_dereference(current->perf_event_ctxp[ctxn]);
@@ -7805,7 +7840,11 @@ static void perf_event_addr_filters_apply(struct perf_event *event)
        list_for_each_entry(filter, &ifh->list, entry) {
                event->addr_filters_offs[count] = 0;
 
-               if (perf_addr_filter_needs_mmap(filter))
+               /*
+                * Adjust base offset if the filter is associated to a binary
+                * that needs to be mapped:
+                */
+               if (filter->inode)
                        event->addr_filters_offs[count] =
                                perf_addr_filter_apply(filter, mm);
 
@@ -7936,8 +7975,10 @@ perf_event_parse_addr_filter(struct perf_event *event, char *fstr,
                                        goto fail;
                        }
 
-                       if (token == IF_SRC_FILE) {
-                               filename = match_strdup(&args[2]);
+                       if (token == IF_SRC_FILE || token == IF_SRC_FILEADDR) {
+                               int fpos = filter->range ? 2 : 1;
+
+                               filename = match_strdup(&args[fpos]);
                                if (!filename) {
                                        ret = -ENOMEM;
                                        goto fail;
index b7a525a..8c50276 100644 (file)
@@ -172,8 +172,10 @@ static int __replace_page(struct vm_area_struct *vma, unsigned long addr,
        mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end);
        err = -EAGAIN;
        ptep = page_check_address(page, mm, addr, &ptl, 0);
-       if (!ptep)
+       if (!ptep) {
+               mem_cgroup_cancel_charge(kpage, memcg, false);
                goto unlock;
+       }
 
        get_page(kpage);
        page_add_new_anon_rmap(kpage, vma, addr, false);
@@ -200,7 +202,6 @@ static int __replace_page(struct vm_area_struct *vma, unsigned long addr,
 
        err = 0;
  unlock:
-       mem_cgroup_cancel_charge(kpage, memcg, false);
        mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end);
        unlock_page(page);
        return err;
index 9a0178c..b022284 100644 (file)
@@ -835,9 +835,9 @@ static bool memory_bm_pfn_present(struct memory_bitmap *bm, unsigned long pfn)
  */
 static bool rtree_next_node(struct memory_bitmap *bm)
 {
-       bm->cur.node = list_entry(bm->cur.node->list.next,
-                                 struct rtree_node, list);
-       if (&bm->cur.node->list != &bm->cur.zone->leaves) {
+       if (!list_is_last(&bm->cur.node->list, &bm->cur.zone->leaves)) {
+               bm->cur.node = list_entry(bm->cur.node->list.next,
+                                         struct rtree_node, list);
                bm->cur.node_pfn += BM_BITS_PER_BLOCK;
                bm->cur.node_bit  = 0;
                touch_softlockup_watchdog();
@@ -845,9 +845,9 @@ static bool rtree_next_node(struct memory_bitmap *bm)
        }
 
        /* No more nodes, goto next zone */
-       bm->cur.zone = list_entry(bm->cur.zone->list.next,
+       if (!list_is_last(&bm->cur.zone->list, &bm->zones)) {
+               bm->cur.zone = list_entry(bm->cur.zone->list.next,
                                  struct mem_zone_bm_rtree, list);
-       if (&bm->cur.zone->list != &bm->zones) {
                bm->cur.node = list_entry(bm->cur.zone->leaves.next,
                                          struct rtree_node, list);
                bm->cur.node_pfn = 0;
index 276762f..d5760c4 100644 (file)
@@ -9,10 +9,10 @@
 
 char *_braille_console_setup(char **str, char **brl_options)
 {
-       if (!memcmp(*str, "brl,", 4)) {
+       if (!strncmp(*str, "brl,", 4)) {
                *brl_options = "";
                *str += 4;
-       } else if (!memcmp(str, "brl=", 4)) {
+       } else if (!strncmp(*str, "brl=", 4)) {
                *brl_options = *str + 4;
                *str = strchr(*brl_options, ',');
                if (!*str)
index 9858266..a846cf8 100644 (file)
@@ -263,6 +263,11 @@ void account_idle_time(cputime_t cputime)
                cpustat[CPUTIME_IDLE] += (__force u64) cputime;
 }
 
+/*
+ * When a guest is interrupted for a longer amount of time, missed clock
+ * ticks are not redelivered later. Due to that, this function may on
+ * occasion account more time than the calling functions think elapsed.
+ */
 static __always_inline cputime_t steal_account_process_time(cputime_t maxtime)
 {
 #ifdef CONFIG_PARAVIRT
@@ -371,7 +376,7 @@ static void irqtime_account_process_tick(struct task_struct *p, int user_tick,
         * idle, or potentially user or system time. Due to rounding,
         * other time can exceed ticks occasionally.
         */
-       other = account_other_time(cputime);
+       other = account_other_time(ULONG_MAX);
        if (other >= cputime)
                return;
        cputime -= other;
@@ -486,7 +491,7 @@ void account_process_tick(struct task_struct *p, int user_tick)
        }
 
        cputime = cputime_one_jiffy;
-       steal = steal_account_process_time(cputime);
+       steal = steal_account_process_time(ULONG_MAX);
 
        if (steal >= cputime)
                return;
@@ -516,7 +521,7 @@ void account_idle_ticks(unsigned long ticks)
        }
 
        cputime = jiffies_to_cputime(ticks);
-       steal = steal_account_process_time(cputime);
+       steal = steal_account_process_time(ULONG_MAX);
 
        if (steal >= cputime)
                return;
@@ -614,19 +619,25 @@ static void cputime_adjust(struct task_cputime *curr,
        stime = curr->stime;
        utime = curr->utime;
 
-       if (utime == 0) {
-               stime = rtime;
+       /*
+        * If either stime or both stime and utime are 0, assume all runtime is
+        * userspace. Once a task gets some ticks, the monotonicy code at
+        * 'update' will ensure things converge to the observed ratio.
+        */
+       if (stime == 0) {
+               utime = rtime;
                goto update;
        }
 
-       if (stime == 0) {
-               utime = rtime;
+       if (utime == 0) {
+               stime = rtime;
                goto update;
        }
 
        stime = scale_stime((__force u64)stime, (__force u64)rtime,
                            (__force u64)(stime + utime));
 
+update:
        /*
         * Make sure stime doesn't go backwards; this preserves monotonicity
         * for utime because rtime is monotonic.
@@ -649,7 +660,6 @@ static void cputime_adjust(struct task_cputime *curr,
                stime = rtime - utime;
        }
 
-update:
        prev->stime = stime;
        prev->utime = utime;
 out:
@@ -694,6 +704,13 @@ static cputime_t get_vtime_delta(struct task_struct *tsk)
        unsigned long now = READ_ONCE(jiffies);
        cputime_t delta, other;
 
+       /*
+        * Unlike tick based timing, vtime based timing never has lost
+        * ticks, and no need for steal time accounting to make up for
+        * lost ticks. Vtime accounts a rounded version of actual
+        * elapsed time. Limit account_other_time to prevent rounding
+        * errors from causing elapsed vtime to go negative.
+        */
        delta = jiffies_to_cputime(now - tsk->vtime_snap);
        other = account_other_time(delta);
        WARN_ON_ONCE(tsk->vtime_snap_whence == VTIME_INACTIVE);
index b43d0b2..a13bbda 100644 (file)
@@ -2140,6 +2140,21 @@ static int do_proc_dointvec_conv(bool *negp, unsigned long *lvalp,
        return 0;
 }
 
+static int do_proc_douintvec_conv(bool *negp, unsigned long *lvalp,
+                                int *valp,
+                                int write, void *data)
+{
+       if (write) {
+               if (*negp)
+                       return -EINVAL;
+               *valp = *lvalp;
+       } else {
+               unsigned int val = *valp;
+               *lvalp = (unsigned long)val;
+       }
+       return 0;
+}
+
 static const char proc_wspace_sep[] = { ' ', '\t', '\n' };
 
 static int __do_proc_dointvec(void *tbl_data, struct ctl_table *table,
@@ -2259,8 +2274,27 @@ static int do_proc_dointvec(struct ctl_table *table, int write,
 int proc_dointvec(struct ctl_table *table, int write,
                     void __user *buffer, size_t *lenp, loff_t *ppos)
 {
-    return do_proc_dointvec(table,write,buffer,lenp,ppos,
-                           NULL,NULL);
+       return do_proc_dointvec(table, write, buffer, lenp, ppos, NULL, NULL);
+}
+
+/**
+ * proc_douintvec - read a vector of unsigned integers
+ * @table: the sysctl table
+ * @write: %TRUE if this is a write to the sysctl file
+ * @buffer: the user buffer
+ * @lenp: the size of the user buffer
+ * @ppos: file position
+ *
+ * Reads/writes up to table->maxlen/sizeof(unsigned int) unsigned integer
+ * values from/to the user buffer, treated as an ASCII string.
+ *
+ * Returns 0 on success.
+ */
+int proc_douintvec(struct ctl_table *table, int write,
+                    void __user *buffer, size_t *lenp, loff_t *ppos)
+{
+       return do_proc_dointvec(table, write, buffer, lenp, ppos,
+                               do_proc_douintvec_conv, NULL);
 }
 
 /*
@@ -2858,6 +2892,12 @@ int proc_dointvec(struct ctl_table *table, int write,
        return -ENOSYS;
 }
 
+int proc_douintvec(struct ctl_table *table, int write,
+                 void __user *buffer, size_t *lenp, loff_t *ppos)
+{
+       return -ENOSYS;
+}
+
 int proc_dointvec_minmax(struct ctl_table *table, int write,
                    void __user *buffer, size_t *lenp, loff_t *ppos)
 {
@@ -2903,6 +2943,7 @@ int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write,
  * exception granted :-)
  */
 EXPORT_SYMBOL(proc_dointvec);
+EXPORT_SYMBOL(proc_douintvec);
 EXPORT_SYMBOL(proc_dointvec_jiffies);
 EXPORT_SYMBOL(proc_dointvec_minmax);
 EXPORT_SYMBOL(proc_dointvec_userhz_jiffies);
index 7598e6c..dbafc5d 100644 (file)
@@ -223,7 +223,7 @@ static void __blk_add_trace(struct blk_trace *bt, sector_t sector, int bytes,
        what |= MASK_TC_BIT(op_flags, META);
        what |= MASK_TC_BIT(op_flags, PREFLUSH);
        what |= MASK_TC_BIT(op_flags, FUA);
-       if (op == REQ_OP_DISCARD)
+       if (op == REQ_OP_DISCARD || op == REQ_OP_SECURE_ERASE)
                what |= BLK_TC_ACT(BLK_TC_DISCARD);
        if (op == REQ_OP_FLUSH)
                what |= BLK_TC_ACT(BLK_TC_FLUSH);
index 78a23c5..be0ee11 100644 (file)
@@ -262,7 +262,14 @@ config COMPACTION
        select MIGRATION
        depends on MMU
        help
-         Allows the compaction of memory for the allocation of huge pages.
+          Compaction is the only memory management component to form
+          high order (larger physically contiguous) memory blocks
+          reliably. The page allocator relies on compaction heavily and
+          the lack of the feature can lead to unexpected OOM killer
+          invocations for high order memory requests. You shouldn't
+          disable this option unless there really is a strong reason for
+          it and then we would be really interested to hear about that at
+          linux-mm@kvack.org.
 
 #
 # support for page migration
index 2373f0a..2db2112 100644 (file)
@@ -1512,7 +1512,7 @@ static void __split_huge_pmd_locked(struct vm_area_struct *vma, pmd_t *pmd,
        struct page *page;
        pgtable_t pgtable;
        pmd_t _pmd;
-       bool young, write, dirty;
+       bool young, write, dirty, soft_dirty;
        unsigned long addr;
        int i;
 
@@ -1546,6 +1546,7 @@ static void __split_huge_pmd_locked(struct vm_area_struct *vma, pmd_t *pmd,
        write = pmd_write(*pmd);
        young = pmd_young(*pmd);
        dirty = pmd_dirty(*pmd);
+       soft_dirty = pmd_soft_dirty(*pmd);
 
        pmdp_huge_split_prepare(vma, haddr, pmd);
        pgtable = pgtable_trans_huge_withdraw(mm, pmd);
@@ -1562,6 +1563,8 @@ static void __split_huge_pmd_locked(struct vm_area_struct *vma, pmd_t *pmd,
                        swp_entry_t swp_entry;
                        swp_entry = make_migration_entry(page + i, write);
                        entry = swp_entry_to_pte(swp_entry);
+                       if (soft_dirty)
+                               entry = pte_swp_mksoft_dirty(entry);
                } else {
                        entry = mk_pte(page + i, vma->vm_page_prot);
                        entry = maybe_mkwrite(entry, vma);
@@ -1569,6 +1572,8 @@ static void __split_huge_pmd_locked(struct vm_area_struct *vma, pmd_t *pmd,
                                entry = pte_wrprotect(entry);
                        if (!young)
                                entry = pte_mkold(entry);
+                       if (soft_dirty)
+                               entry = pte_mksoft_dirty(entry);
                }
                if (dirty)
                        SetPageDirty(page + i);
index 2ff0289..9a6a51a 100644 (file)
@@ -4082,24 +4082,6 @@ static void mem_cgroup_id_get_many(struct mem_cgroup *memcg, unsigned int n)
        atomic_add(n, &memcg->id.ref);
 }
 
-static struct mem_cgroup *mem_cgroup_id_get_online(struct mem_cgroup *memcg)
-{
-       while (!atomic_inc_not_zero(&memcg->id.ref)) {
-               /*
-                * The root cgroup cannot be destroyed, so it's refcount must
-                * always be >= 1.
-                */
-               if (WARN_ON_ONCE(memcg == root_mem_cgroup)) {
-                       VM_BUG_ON(1);
-                       break;
-               }
-               memcg = parent_mem_cgroup(memcg);
-               if (!memcg)
-                       memcg = root_mem_cgroup;
-       }
-       return memcg;
-}
-
 static void mem_cgroup_id_put_many(struct mem_cgroup *memcg, unsigned int n)
 {
        if (atomic_sub_and_test(n, &memcg->id.ref)) {
@@ -5821,6 +5803,24 @@ static int __init mem_cgroup_init(void)
 subsys_initcall(mem_cgroup_init);
 
 #ifdef CONFIG_MEMCG_SWAP
+static struct mem_cgroup *mem_cgroup_id_get_online(struct mem_cgroup *memcg)
+{
+       while (!atomic_inc_not_zero(&memcg->id.ref)) {
+               /*
+                * The root cgroup cannot be destroyed, so it's refcount must
+                * always be >= 1.
+                */
+               if (WARN_ON_ONCE(memcg == root_mem_cgroup)) {
+                       VM_BUG_ON(1);
+                       break;
+               }
+               memcg = parent_mem_cgroup(memcg);
+               if (!memcg)
+                       memcg = root_mem_cgroup;
+       }
+       return memcg;
+}
+
 /**
  * mem_cgroup_swapout - transfer a memsw charge to swap
  * @page: page whose memsw charge to transfer
index 65ec288..c8a955b 100644 (file)
@@ -8,6 +8,7 @@
  */
 
 #include <linux/kernel.h>
+#include <linux/dax.h>
 #include <linux/gfp.h>
 #include <linux/export.h>
 #include <linux/blkdev.h>
@@ -544,6 +545,14 @@ do_readahead(struct address_space *mapping, struct file *filp,
        if (!mapping || !mapping->a_ops)
                return -EINVAL;
 
+       /*
+        * Readahead doesn't make sense for DAX inodes, but we don't want it
+        * to report a failure either.  Instead, we just return success and
+        * don't do any work.
+        */
+       if (dax_mapping(mapping))
+               return 0;
+
        return force_page_cache_readahead(mapping, filp, index, nr);
 }
 
index 8ebae91..a3cc305 100644 (file)
@@ -83,7 +83,7 @@ static bool overlaps(const void *ptr, unsigned long n, unsigned long low,
        unsigned long check_high = check_low + n;
 
        /* Does not overlap if entirely above or entirely below. */
-       if (check_low >= high || check_high < low)
+       if (check_low >= high || check_high <= low)
                return false;
 
        return true;
@@ -124,7 +124,7 @@ static inline const char *check_kernel_text_object(const void *ptr,
 static inline const char *check_bogus_address(const void *ptr, unsigned long n)
 {
        /* Reject if object wraps past end of memory. */
-       if (ptr + n < ptr)
+       if ((unsigned long)ptr + n < (unsigned long)ptr)
                return "<wrapped address>";
 
        /* Reject if NULL or ZERO-allocation. */
index 49a00d5..aed4511 100755 (executable)
@@ -2136,9 +2136,11 @@ sub vcs_file_exists {
 
     my $cmd = $VCS_cmds{"file_exists_cmd"};
     $cmd =~ s/(\$\w+)/$1/eeg;          # interpolate $cmd
-
+    $cmd .= " 2>&1";
     $exists = &{$VCS_cmds{"execute_cmd"}}($cmd);
 
+    return 0 if ($? != 0);
+
     return $exists;
 }
 
index df28f2b..da10d9b 100644 (file)
@@ -136,6 +136,7 @@ config HAVE_ARCH_HARDENED_USERCOPY
 config HARDENED_USERCOPY
        bool "Harden memory copies between kernel and userspace"
        depends on HAVE_ARCH_HARDENED_USERCOPY
+       depends on HAVE_HARDENED_USERCOPY_ALLOCATOR
        select BUG
        help
          This option checks for obviously wrong memory regions when
index 574b1b4..7100f05 100644 (file)
@@ -4828,7 +4828,7 @@ enum {
        ALC293_FIXUP_DELL1_MIC_NO_PRESENCE,
        ALC292_FIXUP_TPT440_DOCK,
        ALC292_FIXUP_TPT440,
-       ALC283_FIXUP_BXBT2807_MIC,
+       ALC283_FIXUP_HEADSET_MIC,
        ALC255_FIXUP_DELL_WMI_MIC_MUTE_LED,
        ALC282_FIXUP_ASPIRE_V5_PINS,
        ALC280_FIXUP_HP_GPIO4,
@@ -5321,7 +5321,7 @@ static const struct hda_fixup alc269_fixups[] = {
                .chained = true,
                .chain_id = ALC292_FIXUP_TPT440_DOCK,
        },
-       [ALC283_FIXUP_BXBT2807_MIC] = {
+       [ALC283_FIXUP_HEADSET_MIC] = {
                .type = HDA_FIXUP_PINS,
                .v.pins = (const struct hda_pintbl[]) {
                        { 0x19, 0x04a110f0 },
@@ -5651,7 +5651,8 @@ static const struct snd_pci_quirk alc269_fixup_tbl[] = {
        SND_PCI_QUIRK(0x10cf, 0x1757, "Lifebook E752", ALC269_FIXUP_LIFEBOOK_HP_PIN),
        SND_PCI_QUIRK(0x10cf, 0x1845, "Lifebook U904", ALC269_FIXUP_LIFEBOOK_EXTMIC),
        SND_PCI_QUIRK(0x144d, 0xc109, "Samsung Ativ book 9 (NP900X3G)", ALC269_FIXUP_INV_DMIC),
-       SND_PCI_QUIRK(0x1458, 0xfa53, "Gigabyte BXBT-2807", ALC283_FIXUP_BXBT2807_MIC),
+       SND_PCI_QUIRK(0x1458, 0xfa53, "Gigabyte BXBT-2807", ALC283_FIXUP_HEADSET_MIC),
+       SND_PCI_QUIRK(0x1462, 0xb120, "MSI Cubi MS-B120", ALC283_FIXUP_HEADSET_MIC),
        SND_PCI_QUIRK(0x17aa, 0x20f2, "Thinkpad SL410/510", ALC269_FIXUP_SKU_IGNORE),
        SND_PCI_QUIRK(0x17aa, 0x215e, "Thinkpad L512", ALC269_FIXUP_SKU_IGNORE),
        SND_PCI_QUIRK(0x17aa, 0x21b8, "Thinkpad Edge 14", ALC269_FIXUP_SKU_IGNORE),
index 54c09ac..16e459a 100644 (file)
@@ -299,8 +299,9 @@ static int atmel_ssc_startup(struct snd_pcm_substream *substream,
        clk_enable(ssc_p->ssc->clk);
        ssc_p->mck_rate = clk_get_rate(ssc_p->ssc->clk);
 
-       /* Reset the SSC to keep it at a clean status */
-       ssc_writel(ssc_p->ssc->regs, CR, SSC_BIT(CR_SWRST));
+       /* Reset the SSC unless initialized to keep it in a clean state */
+       if (!ssc_p->initialized)
+               ssc_writel(ssc_p->ssc->regs, CR, SSC_BIT(CR_SWRST));
 
        if (substream->stream == SNDRV_PCM_STREAM_PLAYBACK) {
                dir = 0;
index e5527bc..bcf1834 100644 (file)
@@ -1247,8 +1247,8 @@ static int da7213_set_dai_fmt(struct snd_soc_dai *codec_dai, unsigned int fmt)
                return -EINVAL;
        }
 
-       /* By default only 32 BCLK per WCLK is supported */
-       dai_clk_mode |= DA7213_DAI_BCLKS_PER_WCLK_32;
+       /* By default only 64 BCLK per WCLK is supported */
+       dai_clk_mode |= DA7213_DAI_BCLKS_PER_WCLK_64;
 
        snd_soc_write(codec, DA7213_DAI_CLK_MODE, dai_clk_mode);
        snd_soc_update_bits(codec, DA7213_DAI_CTRL, DA7213_DAI_FORMAT_MASK,
index cf0a39b..02352ed 100644 (file)
@@ -412,6 +412,7 @@ static int max98371_i2c_remove(struct i2c_client *client)
 
 static const struct i2c_device_id max98371_i2c_id[] = {
        { "max98371", 0 },
+       { }
 };
 
 MODULE_DEVICE_TABLE(i2c, max98371_i2c_id);
index 5c9707a..2e59a85 100644 (file)
@@ -212,31 +212,6 @@ static const unsigned short logtable[256] = {
        0xfa2f, 0xfaea, 0xfba5, 0xfc60, 0xfd1a, 0xfdd4, 0xfe8e, 0xff47
 };
 
-static struct snd_soc_dai *nau8825_get_codec_dai(struct nau8825 *nau8825)
-{
-       struct snd_soc_codec *codec = snd_soc_dapm_to_codec(nau8825->dapm);
-       struct snd_soc_component *component = &codec->component;
-       struct snd_soc_dai *codec_dai, *_dai;
-
-       list_for_each_entry_safe(codec_dai, _dai, &component->dai_list, list) {
-               if (!strncmp(codec_dai->name, NUVOTON_CODEC_DAI,
-                       strlen(NUVOTON_CODEC_DAI)))
-                       return codec_dai;
-       }
-       return NULL;
-}
-
-static bool nau8825_dai_is_active(struct nau8825 *nau8825)
-{
-       struct snd_soc_dai *codec_dai = nau8825_get_codec_dai(nau8825);
-
-       if (codec_dai) {
-               if (codec_dai->playback_active || codec_dai->capture_active)
-                       return true;
-       }
-       return false;
-}
-
 /**
  * nau8825_sema_acquire - acquire the semaphore of nau88l25
  * @nau8825:  component to register the codec private data with
@@ -250,19 +225,26 @@ static bool nau8825_dai_is_active(struct nau8825 *nau8825)
  * Acquires the semaphore without jiffies. If no more tasks are allowed
  * to acquire the semaphore, calling this function will put the task to
  * sleep until the semaphore is released.
- * It returns if the semaphore was acquired.
+ * If the semaphore is not released within the specified number of jiffies,
+ * this function returns -ETIME.
+ * If the sleep is interrupted by a signal, this function will return -EINTR.
+ * It returns 0 if the semaphore was acquired successfully.
  */
-static void nau8825_sema_acquire(struct nau8825 *nau8825, long timeout)
+static int nau8825_sema_acquire(struct nau8825 *nau8825, long timeout)
 {
        int ret;
 
-       if (timeout)
+       if (timeout) {
                ret = down_timeout(&nau8825->xtalk_sem, timeout);
-       else
+               if (ret < 0)
+                       dev_warn(nau8825->dev, "Acquire semaphone timeout\n");
+       } else {
                ret = down_interruptible(&nau8825->xtalk_sem);
+               if (ret < 0)
+                       dev_warn(nau8825->dev, "Acquire semaphone fail\n");
+       }
 
-       if (ret < 0)
-               dev_warn(nau8825->dev, "Acquire semaphone fail\n");
+       return ret;
 }
 
 /**
@@ -1205,6 +1187,8 @@ static int nau8825_hw_params(struct snd_pcm_substream *substream,
        struct nau8825 *nau8825 = snd_soc_codec_get_drvdata(codec);
        unsigned int val_len = 0;
 
+       nau8825_sema_acquire(nau8825, 2 * HZ);
+
        switch (params_width(params)) {
        case 16:
                val_len |= NAU8825_I2S_DL_16;
@@ -1225,6 +1209,9 @@ static int nau8825_hw_params(struct snd_pcm_substream *substream,
        regmap_update_bits(nau8825->regmap, NAU8825_REG_I2S_PCM_CTRL1,
                NAU8825_I2S_DL_MASK, val_len);
 
+       /* Release the semaphone. */
+       nau8825_sema_release(nau8825);
+
        return 0;
 }
 
@@ -1234,6 +1221,8 @@ static int nau8825_set_dai_fmt(struct snd_soc_dai *codec_dai, unsigned int fmt)
        struct nau8825 *nau8825 = snd_soc_codec_get_drvdata(codec);
        unsigned int ctrl1_val = 0, ctrl2_val = 0;
 
+       nau8825_sema_acquire(nau8825, 2 * HZ);
+
        switch (fmt & SND_SOC_DAIFMT_MASTER_MASK) {
        case SND_SOC_DAIFMT_CBM_CFM:
                ctrl2_val |= NAU8825_I2S_MS_MASTER;
@@ -1282,6 +1271,9 @@ static int nau8825_set_dai_fmt(struct snd_soc_dai *codec_dai, unsigned int fmt)
        regmap_update_bits(nau8825->regmap, NAU8825_REG_I2S_PCM_CTRL2,
                NAU8825_I2S_MS_MASK, ctrl2_val);
 
+       /* Release the semaphone. */
+       nau8825_sema_release(nau8825);
+
        return 0;
 }
 
@@ -1611,8 +1603,11 @@ static irqreturn_t nau8825_interrupt(int irq, void *data)
                                         * cess and restore changes if process
                                         * is ongoing when ejection.
                                         */
+                                       int ret;
                                        nau8825->xtalk_protect = true;
-                                       nau8825_sema_acquire(nau8825, 0);
+                                       ret = nau8825_sema_acquire(nau8825, 0);
+                                       if (ret < 0)
+                                               nau8825->xtalk_protect = false;
                                }
                                /* Startup cross talk detection process */
                                nau8825->xtalk_state = NAU8825_XTALK_PREPARE;
@@ -2238,23 +2233,14 @@ static int __maybe_unused nau8825_suspend(struct snd_soc_codec *codec)
 static int __maybe_unused nau8825_resume(struct snd_soc_codec *codec)
 {
        struct nau8825 *nau8825 = snd_soc_codec_get_drvdata(codec);
+       int ret;
 
        regcache_cache_only(nau8825->regmap, false);
        regcache_sync(nau8825->regmap);
-       if (nau8825_is_jack_inserted(nau8825->regmap)) {
-               /* If the jack is inserted, we need to check whether the play-
-                * back is active before suspend. If active, the driver has to
-                * raise the protection for cross talk function to avoid the
-                * playback recovers before cross talk process finish. Other-
-                * wise, the playback will be interfered by cross talk func-
-                * tion. It is better to apply hardware related parameters
-                * before starting playback or record.
-                */
-               if (nau8825_dai_is_active(nau8825)) {
-                       nau8825->xtalk_protect = true;
-                       nau8825_sema_acquire(nau8825, 0);
-               }
-       }
+       nau8825->xtalk_protect = true;
+       ret = nau8825_sema_acquire(nau8825, 0);
+       if (ret < 0)
+               nau8825->xtalk_protect = false;
        enable_irq(nau8825->irq);
 
        return 0;
index a67ea10..f266439 100644 (file)
@@ -581,7 +581,7 @@ static int wm2000_anc_transition(struct wm2000_priv *wm2000,
        if (anc_transitions[i].dest == ANC_OFF)
                clk_disable_unprepare(wm2000->mclk);
 
-       return ret;
+       return 0;
 }
 
 static int wm2000_anc_set_mode(struct wm2000_priv *wm2000)
index 45602ca..2d53c8d 100644 (file)
@@ -1,5 +1,5 @@
-obj-$(CONFIG_SND_SIMPLE_CARD_UTILS) := simple-card-utils.o
-
+snd-soc-simple-card-utils-objs := simple-card-utils.o
 snd-soc-simple-card-objs       := simple-card.o
 
-obj-$(CONFIG_SND_SIMPLE_CARD)  += snd-soc-simple-card.o
+obj-$(CONFIG_SND_SIMPLE_CARD_UTILS)    += snd-soc-simple-card-utils.o
+obj-$(CONFIG_SND_SIMPLE_CARD)          += snd-soc-simple-card.o
index d89a9a1..9599de6 100644 (file)
@@ -7,6 +7,7 @@
  * it under the terms of the GNU General Public License version 2 as
  * published by the Free Software Foundation.
  */
+#include <linux/module.h>
 #include <linux/of.h>
 #include <sound/simple_card_utils.h>
 
@@ -95,3 +96,8 @@ int asoc_simple_card_parse_card_name(struct snd_soc_card *card,
        return 0;
 }
 EXPORT_SYMBOL_GPL(asoc_simple_card_parse_card_name);
+
+/* Module information */
+MODULE_AUTHOR("Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>");
+MODULE_DESCRIPTION("ALSA SoC Simple Card Utils");
+MODULE_LICENSE("GPL v2");
index 25fcb79..ddcb52a 100644 (file)
@@ -123,6 +123,11 @@ int snd_skl_get_module_info(struct skl_sst *ctx, u8 *uuid,
 
        uuid_mod = (uuid_le *)uuid;
 
+       if (list_empty(&ctx->uuid_list)) {
+               dev_err(ctx->dev, "Module list is empty\n");
+               return -EINVAL;
+       }
+
        list_for_each_entry(module, &ctx->uuid_list, list) {
                if (uuid_le_cmp(*uuid_mod, module->uuid) == 0) {
                        dfw_config->module_id = module->id;
index cd59536..e3e7641 100644 (file)
@@ -672,8 +672,10 @@ static int skl_probe(struct pci_dev *pci,
 
        skl->nhlt = skl_nhlt_init(bus->dev);
 
-       if (skl->nhlt == NULL)
+       if (skl->nhlt == NULL) {
+               err = -ENODEV;
                goto out_free;
+       }
 
        skl_nhlt_update_topology_bin(skl);
 
index 0843a68..f61b3b5 100644 (file)
 struct abe_twl6040 {
        int     jack_detection; /* board can detect jack events */
        int     mclk_freq;      /* MCLK frequency speed for twl6040 */
-
-       struct platform_device *dmic_codec_dev;
 };
 
+struct platform_device *dmic_codec_dev;
+
 static int omap_abe_hw_params(struct snd_pcm_substream *substream,
        struct snd_pcm_hw_params *params)
 {
@@ -258,8 +258,6 @@ static int omap_abe_probe(struct platform_device *pdev)
        if (priv == NULL)
                return -ENOMEM;
 
-       priv->dmic_codec_dev = ERR_PTR(-EINVAL);
-
        if (snd_soc_of_parse_card_name(card, "ti,model")) {
                dev_err(&pdev->dev, "Card name is not provided\n");
                return -ENODEV;
@@ -284,13 +282,6 @@ static int omap_abe_probe(struct platform_device *pdev)
                num_links = 2;
                abe_twl6040_dai_links[1].cpu_of_node = dai_node;
                abe_twl6040_dai_links[1].platform_of_node = dai_node;
-
-               priv->dmic_codec_dev = platform_device_register_simple(
-                                               "dmic-codec", -1, NULL, 0);
-               if (IS_ERR(priv->dmic_codec_dev)) {
-                       dev_err(&pdev->dev, "Can't instantiate dmic-codec\n");
-                       return PTR_ERR(priv->dmic_codec_dev);
-               }
        } else {
                num_links = 1;
        }
@@ -299,16 +290,14 @@ static int omap_abe_probe(struct platform_device *pdev)
        of_property_read_u32(node, "ti,mclk-freq", &priv->mclk_freq);
        if (!priv->mclk_freq) {
                dev_err(&pdev->dev, "MCLK frequency not provided\n");
-               ret = -EINVAL;
-               goto err_unregister;
+               return -EINVAL;
        }
 
        card->fully_routed = 1;
 
        if (!priv->mclk_freq) {
                dev_err(&pdev->dev, "MCLK frequency missing\n");
-               ret = -ENODEV;
-               goto err_unregister;
+               return -ENODEV;
        }
 
        card->dai_link = abe_twl6040_dai_links;
@@ -317,17 +306,9 @@ static int omap_abe_probe(struct platform_device *pdev)
        snd_soc_card_set_drvdata(card, priv);
 
        ret = snd_soc_register_card(card);
-       if (ret) {
+       if (ret)
                dev_err(&pdev->dev, "snd_soc_register_card() failed: %d\n",
                        ret);
-               goto err_unregister;
-       }
-
-       return 0;
-
-err_unregister:
-       if (!IS_ERR(priv->dmic_codec_dev))
-               platform_device_unregister(priv->dmic_codec_dev);
 
        return ret;
 }
@@ -335,13 +316,9 @@ err_unregister:
 static int omap_abe_remove(struct platform_device *pdev)
 {
        struct snd_soc_card *card = platform_get_drvdata(pdev);
-       struct abe_twl6040 *priv = snd_soc_card_get_drvdata(card);
 
        snd_soc_unregister_card(card);
 
-       if (!IS_ERR(priv->dmic_codec_dev))
-               platform_device_unregister(priv->dmic_codec_dev);
-
        return 0;
 }
 
@@ -361,7 +338,33 @@ static struct platform_driver omap_abe_driver = {
        .remove = omap_abe_remove,
 };
 
-module_platform_driver(omap_abe_driver);
+static int __init omap_abe_init(void)
+{
+       int ret;
+
+       dmic_codec_dev = platform_device_register_simple("dmic-codec", -1, NULL,
+                                                        0);
+       if (IS_ERR(dmic_codec_dev)) {
+               pr_err("%s: dmic-codec device registration failed\n", __func__);
+               return PTR_ERR(dmic_codec_dev);
+       }
+
+       ret = platform_driver_register(&omap_abe_driver);
+       if (ret) {
+               pr_err("%s: platform driver registration failed\n", __func__);
+               platform_device_unregister(dmic_codec_dev);
+       }
+
+       return ret;
+}
+module_init(omap_abe_init);
+
+static void __exit omap_abe_exit(void)
+{
+       platform_driver_unregister(&omap_abe_driver);
+       platform_device_unregister(dmic_codec_dev);
+}
+module_exit(omap_abe_exit);
 
 MODULE_AUTHOR("Misael Lopez Cruz <misael.lopez@ti.com>");
 MODULE_DESCRIPTION("ALSA SoC for OMAP boards with ABE and twl6040 codec");
index e7cdc51..64609c7 100644 (file)
@@ -31,7 +31,6 @@
 #include <linux/err.h>
 #include <linux/io.h>
 #include <linux/irq.h>
-#include <linux/clk.h>
 #include <linux/slab.h>
 #include <linux/pm_runtime.h>
 #include <linux/of_device.h>
@@ -55,7 +54,6 @@ struct omap_mcpdm {
        unsigned long phys_base;
        void __iomem *io_base;
        int irq;
-       struct clk *pdmclk;
 
        struct mutex mutex;
 
@@ -390,15 +388,14 @@ static int omap_mcpdm_probe(struct snd_soc_dai *dai)
        struct omap_mcpdm *mcpdm = snd_soc_dai_get_drvdata(dai);
        int ret;
 
-       clk_prepare_enable(mcpdm->pdmclk);
        pm_runtime_enable(mcpdm->dev);
 
        /* Disable lines while request is ongoing */
        pm_runtime_get_sync(mcpdm->dev);
        omap_mcpdm_write(mcpdm, MCPDM_REG_CTRL, 0x00);
 
-       ret = devm_request_irq(mcpdm->dev, mcpdm->irq, omap_mcpdm_irq_handler,
-                               0, "McPDM", (void *)mcpdm);
+       ret = request_irq(mcpdm->irq, omap_mcpdm_irq_handler, 0, "McPDM",
+                         (void *)mcpdm);
 
        pm_runtime_put_sync(mcpdm->dev);
 
@@ -423,9 +420,9 @@ static int omap_mcpdm_remove(struct snd_soc_dai *dai)
 {
        struct omap_mcpdm *mcpdm = snd_soc_dai_get_drvdata(dai);
 
+       free_irq(mcpdm->irq, (void *)mcpdm);
        pm_runtime_disable(mcpdm->dev);
 
-       clk_disable_unprepare(mcpdm->pdmclk);
        return 0;
 }
 
@@ -445,8 +442,6 @@ static int omap_mcpdm_suspend(struct snd_soc_dai *dai)
                mcpdm->pm_active_count++;
        }
 
-       clk_disable_unprepare(mcpdm->pdmclk);
-
        return 0;
 }
 
@@ -454,8 +449,6 @@ static int omap_mcpdm_resume(struct snd_soc_dai *dai)
 {
        struct omap_mcpdm *mcpdm = snd_soc_dai_get_drvdata(dai);
 
-       clk_prepare_enable(mcpdm->pdmclk);
-
        if (mcpdm->pm_active_count) {
                while (mcpdm->pm_active_count--)
                        pm_runtime_get_sync(mcpdm->dev);
@@ -549,15 +542,6 @@ static int asoc_mcpdm_probe(struct platform_device *pdev)
 
        mcpdm->dev = &pdev->dev;
 
-       mcpdm->pdmclk = devm_clk_get(&pdev->dev, "pdmclk");
-       if (IS_ERR(mcpdm->pdmclk)) {
-               if (PTR_ERR(mcpdm->pdmclk) == -EPROBE_DEFER)
-                       return -EPROBE_DEFER;
-               dev_warn(&pdev->dev, "Error getting pdmclk (%ld)!\n",
-                        PTR_ERR(mcpdm->pdmclk));
-               mcpdm->pdmclk = NULL;
-       }
-
        ret =  devm_snd_soc_register_component(&pdev->dev,
                                               &omap_mcpdm_component,
                                               &omap_mcpdm_dai, 1);
index 50849e1..92e88bc 100644 (file)
@@ -58,10 +58,12 @@ static struct platform_device *s3c24xx_uda134x_snd_device;
 
 static int s3c24xx_uda134x_startup(struct snd_pcm_substream *substream)
 {
-       int ret = 0;
+       struct snd_soc_pcm_runtime *rtd = substream->private_data;
+       struct snd_soc_dai *cpu_dai = rtd->cpu_dai;
 #ifdef ENFORCE_RATES
        struct snd_pcm_runtime *runtime = substream->runtime;
 #endif
+       int ret = 0;
 
        mutex_lock(&clk_lock);
        pr_debug("%s %d\n", __func__, clk_users);
@@ -71,8 +73,7 @@ static int s3c24xx_uda134x_startup(struct snd_pcm_substream *substream)
                        printk(KERN_ERR "%s cannot get xtal\n", __func__);
                        ret = PTR_ERR(xtal);
                } else {
-                       pclk = clk_get(&s3c24xx_uda134x_snd_device->dev,
-                                      "pclk");
+                       pclk = clk_get(cpu_dai->dev, "iis");
                        if (IS_ERR(pclk)) {
                                printk(KERN_ERR "%s cannot get pclk\n",
                                       __func__);
index e39f916..969a516 100644 (file)
@@ -226,8 +226,12 @@ static void rsnd_src_set_convert_rate(struct rsnd_dai_stream *io,
        ifscr = 0;
        fsrate = 0;
        if (fin != fout) {
+               u64 n;
+
                ifscr = 1;
-               fsrate = 0x0400000 / fout * fin;
+               n = (u64)0x0400000 * fin;
+               do_div(n, fout);
+               fsrate = n;
        }
 
        /*
index d2df46c..bf7b52f 100644 (file)
@@ -121,7 +121,7 @@ static int soc_compr_open_fe(struct snd_compr_stream *cstream)
 
                dpcm_be_disconnect(fe, stream);
                fe->dpcm[stream].runtime = NULL;
-               goto fe_err;
+               goto path_err;
        }
 
        dpcm_clear_pending_state(fe, stream);
@@ -136,6 +136,8 @@ static int soc_compr_open_fe(struct snd_compr_stream *cstream)
 
        return 0;
 
+path_err:
+       dpcm_path_put(&list);
 fe_err:
        if (fe->dai_link->compr_ops && fe->dai_link->compr_ops->shutdown)
                fe->dai_link->compr_ops->shutdown(cstream);
index 16369ca..4afa8db 100644 (file)
@@ -1056,7 +1056,7 @@ static int soc_bind_dai_link(struct snd_soc_card *card,
        if (!rtd->platform) {
                dev_err(card->dev, "ASoC: platform %s not registered\n",
                        dai_link->platform_name);
-               return -EPROBE_DEFER;
+               goto _err_defer;
        }
 
        soc_add_pcm_runtime(card, rtd);
@@ -2083,14 +2083,13 @@ static int soc_cleanup_card_resources(struct snd_soc_card *card)
        /* remove auxiliary devices */
        soc_remove_aux_devices(card);
 
+       snd_soc_dapm_free(&card->dapm);
        soc_cleanup_card_debugfs(card);
 
        /* remove the card */
        if (card->remove)
                card->remove(card);
 
-       snd_soc_dapm_free(&card->dapm);
-
        snd_card_free(card->snd_card);
        return 0;
 
index 8698c26..d908ff8 100644 (file)
@@ -3493,6 +3493,7 @@ static int snd_soc_dai_link_event(struct snd_soc_dapm_widget *w,
        const struct snd_soc_pcm_stream *config = w->params + w->params_select;
        struct snd_pcm_substream substream;
        struct snd_pcm_hw_params *params = NULL;
+       struct snd_pcm_runtime *runtime = NULL;
        u64 fmt;
        int ret;
 
@@ -3541,6 +3542,14 @@ static int snd_soc_dai_link_event(struct snd_soc_dapm_widget *w,
 
        memset(&substream, 0, sizeof(substream));
 
+       /* Allocate a dummy snd_pcm_runtime for startup() and other ops() */
+       runtime = kzalloc(sizeof(*runtime), GFP_KERNEL);
+       if (!runtime) {
+               ret = -ENOMEM;
+               goto out;
+       }
+       substream.runtime = runtime;
+
        switch (event) {
        case SND_SOC_DAPM_PRE_PMU:
                substream.stream = SNDRV_PCM_STREAM_CAPTURE;
@@ -3606,6 +3615,7 @@ static int snd_soc_dai_link_event(struct snd_soc_dapm_widget *w,
        }
 
 out:
+       kfree(runtime);
        kfree(params);
        return ret;
 }
index 204cc07..41aa335 100644 (file)
@@ -55,7 +55,6 @@ static int snd_line6_impulse_volume_put(struct snd_kcontrol *kcontrol,
                err = line6_pcm_acquire(line6pcm, LINE6_STREAM_IMPULSE);
                if (err < 0) {
                        line6pcm->impulse_volume = 0;
-                       line6_pcm_release(line6pcm, LINE6_STREAM_IMPULSE);
                        return err;
                }
        } else {
@@ -211,7 +210,9 @@ static void line6_stream_stop(struct snd_line6_pcm *line6pcm, int direction,
        spin_lock_irqsave(&pstr->lock, flags);
        clear_bit(type, &pstr->running);
        if (!pstr->running) {
+               spin_unlock_irqrestore(&pstr->lock, flags);
                line6_unlink_audio_urbs(line6pcm, pstr);
+               spin_lock_irqsave(&pstr->lock, flags);
                if (direction == SNDRV_PCM_STREAM_CAPTURE) {
                        line6pcm->prev_fbuf = NULL;
                        line6pcm->prev_fsize = 0;
index daf81d1..45dd348 100644 (file)
@@ -244,8 +244,8 @@ static int pod_set_system_param_int(struct usb_line6_pod *pod, int value,
 static ssize_t serial_number_show(struct device *dev,
                                  struct device_attribute *attr, char *buf)
 {
-       struct usb_interface *interface = to_usb_interface(dev);
-       struct usb_line6_pod *pod = usb_get_intfdata(interface);
+       struct snd_card *card = dev_to_snd_card(dev);
+       struct usb_line6_pod *pod = card->private_data;
 
        return sprintf(buf, "%u\n", pod->serial_number);
 }
@@ -256,8 +256,8 @@ static ssize_t serial_number_show(struct device *dev,
 static ssize_t firmware_version_show(struct device *dev,
                                     struct device_attribute *attr, char *buf)
 {
-       struct usb_interface *interface = to_usb_interface(dev);
-       struct usb_line6_pod *pod = usb_get_intfdata(interface);
+       struct snd_card *card = dev_to_snd_card(dev);
+       struct usb_line6_pod *pod = card->private_data;
 
        return sprintf(buf, "%d.%02d\n", pod->firmware_version / 100,
                       pod->firmware_version % 100);
@@ -269,8 +269,8 @@ static ssize_t firmware_version_show(struct device *dev,
 static ssize_t device_id_show(struct device *dev,
                              struct device_attribute *attr, char *buf)
 {
-       struct usb_interface *interface = to_usb_interface(dev);
-       struct usb_line6_pod *pod = usb_get_intfdata(interface);
+       struct snd_card *card = dev_to_snd_card(dev);
+       struct usb_line6_pod *pod = card->private_data;
 
        return sprintf(buf, "%d\n", pod->device_id);
 }
index f209ea1..3051f86 100644 (file)
@@ -87,9 +87,11 @@ struct kvm_regs {
 /* Supported VGICv3 address types  */
 #define KVM_VGIC_V3_ADDR_TYPE_DIST     2
 #define KVM_VGIC_V3_ADDR_TYPE_REDIST   3
+#define KVM_VGIC_ITS_ADDR_TYPE         4
 
 #define KVM_VGIC_V3_DIST_SIZE          SZ_64K
 #define KVM_VGIC_V3_REDIST_SIZE                (2 * SZ_64K)
+#define KVM_VGIC_V3_ITS_SIZE           (2 * SZ_64K)
 
 #define KVM_ARM_VCPU_POWER_OFF         0 /* CPU is started in OFF state */
 #define KVM_ARM_VCPU_EL1_32BIT         1 /* CPU running a 32bit VM */
index 3b8e99e..a2ffec4 100644 (file)
@@ -93,6 +93,47 @@ struct kvm_s390_vm_cpu_machine {
        __u64 fac_list[256];
 };
 
+#define KVM_S390_VM_CPU_PROCESSOR_FEAT 2
+#define KVM_S390_VM_CPU_MACHINE_FEAT   3
+
+#define KVM_S390_VM_CPU_FEAT_NR_BITS   1024
+#define KVM_S390_VM_CPU_FEAT_ESOP      0
+#define KVM_S390_VM_CPU_FEAT_SIEF2     1
+#define KVM_S390_VM_CPU_FEAT_64BSCAO   2
+#define KVM_S390_VM_CPU_FEAT_SIIF      3
+#define KVM_S390_VM_CPU_FEAT_GPERE     4
+#define KVM_S390_VM_CPU_FEAT_GSLS      5
+#define KVM_S390_VM_CPU_FEAT_IB                6
+#define KVM_S390_VM_CPU_FEAT_CEI       7
+#define KVM_S390_VM_CPU_FEAT_IBS       8
+#define KVM_S390_VM_CPU_FEAT_SKEY      9
+#define KVM_S390_VM_CPU_FEAT_CMMA      10
+#define KVM_S390_VM_CPU_FEAT_PFMFI     11
+#define KVM_S390_VM_CPU_FEAT_SIGPIF    12
+struct kvm_s390_vm_cpu_feat {
+       __u64 feat[16];
+};
+
+#define KVM_S390_VM_CPU_PROCESSOR_SUBFUNC      4
+#define KVM_S390_VM_CPU_MACHINE_SUBFUNC                5
+/* for "test bit" instructions MSB 0 bit ordering, for "query" raw blocks */
+struct kvm_s390_vm_cpu_subfunc {
+       __u8 plo[32];           /* always */
+       __u8 ptff[16];          /* with TOD-clock steering */
+       __u8 kmac[16];          /* with MSA */
+       __u8 kmc[16];           /* with MSA */
+       __u8 km[16];            /* with MSA */
+       __u8 kimd[16];          /* with MSA */
+       __u8 klmd[16];          /* with MSA */
+       __u8 pckmo[16];         /* with MSA3 */
+       __u8 kmctr[16];         /* with MSA4 */
+       __u8 kmf[16];           /* with MSA4 */
+       __u8 kmo[16];           /* with MSA4 */
+       __u8 pcc[16];           /* with MSA4 */
+       __u8 ppno[16];          /* with MSA5 */
+       __u8 reserved[1824];
+};
+
 /* kvm attributes for crypto */
 #define KVM_S390_VM_CRYPTO_ENABLE_AES_KW       0
 #define KVM_S390_VM_CRYPTO_ENABLE_DEA_KW       1
index 8fb5d4a..3ac6343 100644 (file)
        exit_code_ipa0(0xB2, 0x4c, "TAR"),      \
        exit_code_ipa0(0xB2, 0x50, "CSP"),      \
        exit_code_ipa0(0xB2, 0x54, "MVPG"),     \
+       exit_code_ipa0(0xB2, 0x56, "STHYI"),    \
        exit_code_ipa0(0xB2, 0x58, "BSG"),      \
        exit_code_ipa0(0xB2, 0x5a, "BSA"),      \
        exit_code_ipa0(0xB2, 0x5f, "CHSC"),     \
index 448ed96..1c14c25 100644 (file)
@@ -1,5 +1,5 @@
 /*
- * gpio-hammer - example swiss army knife to shake GPIO lines on a system
+ * gpio-event-mon - monitor GPIO line events from userspace
  *
  * Copyright (C) 2016 Linus Walleij
  *
index 8d4dc97..35745a7 100644 (file)
@@ -97,6 +97,7 @@ void arch__fix_tev_from_maps(struct perf_probe_event *pev,
        }
 }
 
+#ifdef HAVE_LIBELF_SUPPORT
 void arch__post_process_probe_trace_events(struct perf_probe_event *pev,
                                           int ntevs)
 {
@@ -118,5 +119,6 @@ void arch__post_process_probe_trace_events(struct perf_probe_event *pev,
                }
        }
 }
+#endif /* HAVE_LIBELF_SUPPORT */
 
 #endif
index fb51457..a2412e9 100644 (file)
@@ -501,7 +501,7 @@ static int intel_pt_recording_options(struct auxtrace_record *itr,
        struct intel_pt_recording *ptr =
                        container_of(itr, struct intel_pt_recording, itr);
        struct perf_pmu *intel_pt_pmu = ptr->intel_pt_pmu;
-       bool have_timing_info;
+       bool have_timing_info, need_immediate = false;
        struct perf_evsel *evsel, *intel_pt_evsel = NULL;
        const struct cpu_map *cpus = evlist->cpus;
        bool privileged = geteuid() == 0 || perf_event_paranoid() < 0;
@@ -655,6 +655,7 @@ static int intel_pt_recording_options(struct auxtrace_record *itr,
                                ptr->have_sched_switch = 3;
                        } else {
                                opts->record_switch_events = true;
+                               need_immediate = true;
                                if (cpu_wide)
                                        ptr->have_sched_switch = 3;
                                else
@@ -700,6 +701,9 @@ static int intel_pt_recording_options(struct auxtrace_record *itr,
                tracking_evsel->attr.freq = 0;
                tracking_evsel->attr.sample_period = 1;
 
+               if (need_immediate)
+                       tracking_evsel->immediate = true;
+
                /* In per-cpu case, always need the time of mmap events etc */
                if (!cpu_map__empty(cpus)) {
                        perf_evsel__set_sample_bit(tracking_evsel, TIME);
index d608a2c..d1ce29b 100644 (file)
@@ -88,6 +88,9 @@ static int __cmd_record(int argc, const char **argv, struct perf_mem *mem)
        if (mem->operation & MEM_OPERATION_LOAD)
                perf_mem_events[PERF_MEM_EVENTS__LOAD].record = true;
 
+       if (mem->operation & MEM_OPERATION_STORE)
+               perf_mem_events[PERF_MEM_EVENTS__STORE].record = true;
+
        if (perf_mem_events[PERF_MEM_EVENTS__LOAD].record)
                rec_argv[i++] = "-W";
 
index 9c640a8..c859e59 100644 (file)
@@ -371,14 +371,16 @@ static int perf_session__check_output_opt(struct perf_session *session)
 
        if (!no_callchain) {
                bool use_callchain = false;
+               bool not_pipe = false;
 
                evlist__for_each_entry(session->evlist, evsel) {
+                       not_pipe = true;
                        if (evsel->attr.sample_type & PERF_SAMPLE_CALLCHAIN) {
                                use_callchain = true;
                                break;
                        }
                }
-               if (!use_callchain)
+               if (not_pipe && !use_callchain)
                        symbol_conf.use_callchain = false;
        }
 
@@ -1690,8 +1692,13 @@ static int list_available_scripts(const struct option *opt __maybe_unused,
        snprintf(scripts_path, MAXPATHLEN, "%s/scripts", get_argv_exec_path());
 
        scripts_dir = opendir(scripts_path);
-       if (!scripts_dir)
-               return -1;
+       if (!scripts_dir) {
+               fprintf(stdout,
+                       "open(%s) failed.\n"
+                       "Check \"PERF_EXEC_PATH\" env to set scripts dir.\n",
+                       scripts_path);
+               exit(-1);
+       }
 
        for_each_lang(scripts_path, scripts_dir, lang_dirent) {
                snprintf(lang_path, MAXPATHLEN, "%s/%s/bin", scripts_path,
index 9c8f15d..8ff6c6a 100644 (file)
@@ -123,8 +123,6 @@ struct intel_pt_decoder {
        bool have_calc_cyc_to_tsc;
        int exec_mode;
        unsigned int insn_bytes;
-       uint64_t sign_bit;
-       uint64_t sign_bits;
        uint64_t period;
        enum intel_pt_period_type period_type;
        uint64_t tot_insn_cnt;
@@ -191,9 +189,6 @@ struct intel_pt_decoder *intel_pt_decoder_new(struct intel_pt_params *params)
        decoder->data               = params->data;
        decoder->return_compression = params->return_compression;
 
-       decoder->sign_bit           = (uint64_t)1 << 47;
-       decoder->sign_bits          = ~(((uint64_t)1 << 48) - 1);
-
        decoder->period             = params->period;
        decoder->period_type        = params->period_type;
 
@@ -362,21 +357,30 @@ int intel_pt__strerror(int code, char *buf, size_t buflen)
        return 0;
 }
 
-static uint64_t intel_pt_calc_ip(struct intel_pt_decoder *decoder,
-                                const struct intel_pt_pkt *packet,
+static uint64_t intel_pt_calc_ip(const struct intel_pt_pkt *packet,
                                 uint64_t last_ip)
 {
        uint64_t ip;
 
        switch (packet->count) {
-       case 2:
+       case 1:
                ip = (last_ip & (uint64_t)0xffffffffffff0000ULL) |
                     packet->payload;
                break;
-       case 4:
+       case 2:
                ip = (last_ip & (uint64_t)0xffffffff00000000ULL) |
                     packet->payload;
                break;
+       case 3:
+               ip = packet->payload;
+               /* Sign-extend 6-byte ip */
+               if (ip & (uint64_t)0x800000000000ULL)
+                       ip |= (uint64_t)0xffff000000000000ULL;
+               break;
+       case 4:
+               ip = (last_ip & (uint64_t)0xffff000000000000ULL) |
+                    packet->payload;
+               break;
        case 6:
                ip = packet->payload;
                break;
@@ -384,16 +388,12 @@ static uint64_t intel_pt_calc_ip(struct intel_pt_decoder *decoder,
                return 0;
        }
 
-       if (ip & decoder->sign_bit)
-               return ip | decoder->sign_bits;
-
        return ip;
 }
 
 static inline void intel_pt_set_last_ip(struct intel_pt_decoder *decoder)
 {
-       decoder->last_ip = intel_pt_calc_ip(decoder, &decoder->packet,
-                                           decoder->last_ip);
+       decoder->last_ip = intel_pt_calc_ip(&decoder->packet, decoder->last_ip);
 }
 
 static inline void intel_pt_set_ip(struct intel_pt_decoder *decoder)
@@ -1657,6 +1657,12 @@ next:
        }
 }
 
+static inline bool intel_pt_have_ip(struct intel_pt_decoder *decoder)
+{
+       return decoder->last_ip || decoder->packet.count == 0 ||
+              decoder->packet.count == 3 || decoder->packet.count == 6;
+}
+
 /* Walk PSB+ packets to get in sync. */
 static int intel_pt_walk_psb(struct intel_pt_decoder *decoder)
 {
@@ -1677,8 +1683,7 @@ static int intel_pt_walk_psb(struct intel_pt_decoder *decoder)
 
                case INTEL_PT_FUP:
                        decoder->pge = true;
-                       if (decoder->last_ip || decoder->packet.count == 6 ||
-                           decoder->packet.count == 0) {
+                       if (intel_pt_have_ip(decoder)) {
                                uint64_t current_ip = decoder->ip;
 
                                intel_pt_set_ip(decoder);
@@ -1767,8 +1772,7 @@ static int intel_pt_walk_to_ip(struct intel_pt_decoder *decoder)
                case INTEL_PT_TIP_PGE:
                case INTEL_PT_TIP:
                        decoder->pge = decoder->packet.type != INTEL_PT_TIP_PGD;
-                       if (decoder->last_ip || decoder->packet.count == 6 ||
-                           decoder->packet.count == 0)
+                       if (intel_pt_have_ip(decoder))
                                intel_pt_set_ip(decoder);
                        if (decoder->ip)
                                return 0;
@@ -1776,9 +1780,7 @@ static int intel_pt_walk_to_ip(struct intel_pt_decoder *decoder)
 
                case INTEL_PT_FUP:
                        if (decoder->overflow) {
-                               if (decoder->last_ip ||
-                                   decoder->packet.count == 6 ||
-                                   decoder->packet.count == 0)
+                               if (intel_pt_have_ip(decoder))
                                        intel_pt_set_ip(decoder);
                                if (decoder->ip)
                                        return 0;
index b1257c8..4f7b320 100644 (file)
@@ -292,36 +292,46 @@ static int intel_pt_get_ip(enum intel_pt_pkt_type type, unsigned int byte,
                           const unsigned char *buf, size_t len,
                           struct intel_pt_pkt *packet)
 {
-       switch (byte >> 5) {
+       int ip_len;
+
+       packet->count = byte >> 5;
+
+       switch (packet->count) {
        case 0:
-               packet->count = 0;
+               ip_len = 0;
                break;
        case 1:
                if (len < 3)
                        return INTEL_PT_NEED_MORE_BYTES;
-               packet->count = 2;
+               ip_len = 2;
                packet->payload = le16_to_cpu(*(uint16_t *)(buf + 1));
                break;
        case 2:
                if (len < 5)
                        return INTEL_PT_NEED_MORE_BYTES;
-               packet->count = 4;
+               ip_len = 4;
                packet->payload = le32_to_cpu(*(uint32_t *)(buf + 1));
                break;
        case 3:
-       case 6:
+       case 4:
                if (len < 7)
                        return INTEL_PT_NEED_MORE_BYTES;
-               packet->count = 6;
+               ip_len = 6;
                memcpy_le64(&packet->payload, buf + 1, 6);
                break;
+       case 6:
+               if (len < 9)
+                       return INTEL_PT_NEED_MORE_BYTES;
+               ip_len = 8;
+               packet->payload = le64_to_cpu(*(uint64_t *)(buf + 1));
+               break;
        default:
                return INTEL_PT_BAD_PACKET;
        }
 
        packet->type = type;
 
-       return packet->count + 1;
+       return ip_len + 1;
 }
 
 static int intel_pt_get_mode(const unsigned char *buf, size_t len,
index 9f3305f..95f0884 100644 (file)
@@ -1,3 +1,4 @@
+#include <sys/sysmacros.h>
 #include <sys/types.h>
 #include <stdio.h>
 #include <stdlib.h>
index 9aed9c3..9c3b9ed 100644 (file)
@@ -133,7 +133,7 @@ int probe_file__open_both(int *kfd, int *ufd, int flag)
 /* Get raw string list of current kprobe_events  or uprobe_events */
 struct strlist *probe_file__get_rawlist(int fd)
 {
-       int ret, idx;
+       int ret, idx, fddup;
        FILE *fp;
        char buf[MAX_CMDLEN];
        char *p;
@@ -143,8 +143,17 @@ struct strlist *probe_file__get_rawlist(int fd)
                return NULL;
 
        sl = strlist__new(NULL, NULL);
+       if (sl == NULL)
+               return NULL;
+
+       fddup = dup(fd);
+       if (fddup < 0)
+               goto out_free_sl;
+
+       fp = fdopen(fddup, "r");
+       if (!fp)
+               goto out_close_fddup;
 
-       fp = fdopen(dup(fd), "r");
        while (!feof(fp)) {
                p = fgets(buf, MAX_CMDLEN, fp);
                if (!p)
@@ -156,13 +165,21 @@ struct strlist *probe_file__get_rawlist(int fd)
                ret = strlist__add(sl, buf);
                if (ret < 0) {
                        pr_debug("strlist__add failed (%d)\n", ret);
-                       strlist__delete(sl);
-                       return NULL;
+                       goto out_close_fp;
                }
        }
        fclose(fp);
 
        return sl;
+
+out_close_fp:
+       fclose(fp);
+       goto out_free_sl;
+out_close_fddup:
+       close(fddup);
+out_free_sl:
+       strlist__delete(sl);
+       return NULL;
 }
 
 static struct strlist *__probe_file__get_namelist(int fd, bool include_group)
@@ -447,12 +464,17 @@ static int probe_cache__load(struct probe_cache *pcache)
 {
        struct probe_cache_entry *entry = NULL;
        char buf[MAX_CMDLEN], *p;
-       int ret = 0;
+       int ret = 0, fddup;
        FILE *fp;
 
-       fp = fdopen(dup(pcache->fd), "r");
-       if (!fp)
+       fddup = dup(pcache->fd);
+       if (fddup < 0)
+               return -errno;
+       fp = fdopen(fddup, "r");
+       if (!fp) {
+               close(fddup);
                return -EINVAL;
+       }
 
        while (!feof(fp)) {
                if (!fgets(buf, MAX_CMDLEN, fp))
index a34321e..a811c13 100644 (file)
@@ -837,7 +837,8 @@ int dso__load_sym(struct dso *dso, struct map *map,
        sec = syms_ss->symtab;
        shdr = syms_ss->symshdr;
 
-       if (elf_section_by_name(elf, &ehdr, &tshdr, ".text", NULL))
+       if (elf_section_by_name(runtime_ss->elf, &runtime_ss->ehdr, &tshdr,
+                               ".text", NULL))
                dso->text_offset = tshdr.sh_addr - tshdr.sh_offset;
 
        if (runtime_ss->opdsec)