nfit: move to nfit/ sub-directory
authorDan Williams <dan.j.williams@intel.com>
Sun, 24 Jul 2016 04:24:19 +0000 (21:24 -0700)
committerDan Williams <dan.j.williams@intel.com>
Sun, 24 Jul 2016 15:04:04 +0000 (08:04 -0700)
With the arrival of x86-machine-check support the nfit driver will add a
(conditionally-compiled) source file.  Prepare for this by moving all
nfit source to drivers/acpi/nfit/.  This is pure code movement, no
functional changes.

Signed-off-by: Dan Williams <dan.j.williams@intel.com>
drivers/acpi/Kconfig
drivers/acpi/Makefile
drivers/acpi/nfit.c [deleted file]
drivers/acpi/nfit.h [deleted file]
drivers/acpi/nfit/Kconfig [new file with mode: 0644]
drivers/acpi/nfit/Makefile [new file with mode: 0644]
drivers/acpi/nfit/core.c [new file with mode: 0644]
drivers/acpi/nfit/nfit.h [new file with mode: 0644]
tools/testing/nvdimm/Kbuild
tools/testing/nvdimm/test/Kbuild

index b7e2e77..415b148 100644 (file)
@@ -447,32 +447,7 @@ config ACPI_REDUCED_HARDWARE_ONLY
 
          If you are unsure what to do, do not enable this option.
 
-config ACPI_NFIT
-       tristate "ACPI NVDIMM Firmware Interface Table (NFIT)"
-       depends on PHYS_ADDR_T_64BIT
-       depends on BLK_DEV
-       depends on ARCH_HAS_MMIO_FLUSH
-       select LIBNVDIMM
-       help
-         Infrastructure to probe ACPI 6 compliant platforms for
-         NVDIMMs (NFIT) and register a libnvdimm device tree.  In
-         addition to storage devices this also enables libnvdimm to pass
-         ACPI._DSM messages for platform/dimm configuration.
-
-         To compile this driver as a module, choose M here:
-         the module will be called nfit.
-
-config ACPI_NFIT_DEBUG
-       bool "NFIT DSM debug"
-       depends on ACPI_NFIT
-       depends on DYNAMIC_DEBUG
-       default n
-       help
-         Enabling this option causes the nfit driver to dump the
-         input and output buffers of _DSM operations on the ACPI0012
-         device and its children.  This can be very verbose, so leave
-         it disabled unless you are debugging a hardware / firmware
-         issue.
+source "drivers/acpi/nfit/Kconfig"
 
 source "drivers/acpi/apei/Kconfig"
 
index 251ce85..64a575a 100644 (file)
@@ -70,7 +70,7 @@ obj-$(CONFIG_ACPI_PCI_SLOT)   += pci_slot.o
 obj-$(CONFIG_ACPI_PROCESSOR)   += processor.o
 obj-$(CONFIG_ACPI)             += container.o
 obj-$(CONFIG_ACPI_THERMAL)     += thermal.o
-obj-$(CONFIG_ACPI_NFIT)                += nfit.o
+obj-$(CONFIG_ACPI_NFIT)                += nfit/
 obj-$(CONFIG_ACPI)             += acpi_memhotplug.o
 obj-$(CONFIG_ACPI_HOTPLUG_IOAPIC) += ioapic.o
 obj-$(CONFIG_ACPI_BATTERY)     += battery.o
diff --git a/drivers/acpi/nfit.c b/drivers/acpi/nfit.c
deleted file mode 100644 (file)
index 19d0dfd..0000000
+++ /dev/null
@@ -1,2765 +0,0 @@
-/*
- * Copyright(c) 2013-2015 Intel Corporation. All rights reserved.
- *
- * This program is free software; you can redistribute it and/or modify
- * it under the terms of version 2 of the GNU General Public License as
- * published by the Free Software Foundation.
- *
- * This program is distributed in the hope that 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/list_sort.h>
-#include <linux/libnvdimm.h>
-#include <linux/module.h>
-#include <linux/mutex.h>
-#include <linux/ndctl.h>
-#include <linux/sysfs.h>
-#include <linux/delay.h>
-#include <linux/list.h>
-#include <linux/acpi.h>
-#include <linux/sort.h>
-#include <linux/pmem.h>
-#include <linux/io.h>
-#include <linux/nd.h>
-#include <asm/cacheflush.h>
-#include "nfit.h"
-
-/*
- * For readq() and writeq() on 32-bit builds, the hi-lo, lo-hi order is
- * irrelevant.
- */
-#include <linux/io-64-nonatomic-hi-lo.h>
-
-static bool force_enable_dimms;
-module_param(force_enable_dimms, bool, S_IRUGO|S_IWUSR);
-MODULE_PARM_DESC(force_enable_dimms, "Ignore _STA (ACPI DIMM device) status");
-
-static unsigned int scrub_timeout = NFIT_ARS_TIMEOUT;
-module_param(scrub_timeout, uint, S_IRUGO|S_IWUSR);
-MODULE_PARM_DESC(scrub_timeout, "Initial scrub timeout in seconds");
-
-/* after three payloads of overflow, it's dead jim */
-static unsigned int scrub_overflow_abort = 3;
-module_param(scrub_overflow_abort, uint, S_IRUGO|S_IWUSR);
-MODULE_PARM_DESC(scrub_overflow_abort,
-               "Number of times we overflow ARS results before abort");
-
-static bool disable_vendor_specific;
-module_param(disable_vendor_specific, bool, S_IRUGO);
-MODULE_PARM_DESC(disable_vendor_specific,
-               "Limit commands to the publicly specified set\n");
-
-static struct workqueue_struct *nfit_wq;
-
-struct nfit_table_prev {
-       struct list_head spas;
-       struct list_head memdevs;
-       struct list_head dcrs;
-       struct list_head bdws;
-       struct list_head idts;
-       struct list_head flushes;
-};
-
-static u8 nfit_uuid[NFIT_UUID_MAX][16];
-
-const u8 *to_nfit_uuid(enum nfit_uuids id)
-{
-       return nfit_uuid[id];
-}
-EXPORT_SYMBOL(to_nfit_uuid);
-
-static struct acpi_nfit_desc *to_acpi_nfit_desc(
-               struct nvdimm_bus_descriptor *nd_desc)
-{
-       return container_of(nd_desc, struct acpi_nfit_desc, nd_desc);
-}
-
-static struct acpi_device *to_acpi_dev(struct acpi_nfit_desc *acpi_desc)
-{
-       struct nvdimm_bus_descriptor *nd_desc = &acpi_desc->nd_desc;
-
-       /*
-        * If provider == 'ACPI.NFIT' we can assume 'dev' is a struct
-        * acpi_device.
-        */
-       if (!nd_desc->provider_name
-                       || strcmp(nd_desc->provider_name, "ACPI.NFIT") != 0)
-               return NULL;
-
-       return to_acpi_device(acpi_desc->dev);
-}
-
-static int xlat_status(void *buf, unsigned int cmd)
-{
-       struct nd_cmd_clear_error *clear_err;
-       struct nd_cmd_ars_status *ars_status;
-       struct nd_cmd_ars_start *ars_start;
-       struct nd_cmd_ars_cap *ars_cap;
-       u16 flags;
-
-       switch (cmd) {
-       case ND_CMD_ARS_CAP:
-               ars_cap = buf;
-               if ((ars_cap->status & 0xffff) == NFIT_ARS_CAP_NONE)
-                       return -ENOTTY;
-
-               /* Command failed */
-               if (ars_cap->status & 0xffff)
-                       return -EIO;
-
-               /* No supported scan types for this range */
-               flags = ND_ARS_PERSISTENT | ND_ARS_VOLATILE;
-               if ((ars_cap->status >> 16 & flags) == 0)
-                       return -ENOTTY;
-               break;
-       case ND_CMD_ARS_START:
-               ars_start = buf;
-               /* ARS is in progress */
-               if ((ars_start->status & 0xffff) == NFIT_ARS_START_BUSY)
-                       return -EBUSY;
-
-               /* Command failed */
-               if (ars_start->status & 0xffff)
-                       return -EIO;
-               break;
-       case ND_CMD_ARS_STATUS:
-               ars_status = buf;
-               /* Command failed */
-               if (ars_status->status & 0xffff)
-                       return -EIO;
-               /* Check extended status (Upper two bytes) */
-               if (ars_status->status == NFIT_ARS_STATUS_DONE)
-                       return 0;
-
-               /* ARS is in progress */
-               if (ars_status->status == NFIT_ARS_STATUS_BUSY)
-                       return -EBUSY;
-
-               /* No ARS performed for the current boot */
-               if (ars_status->status == NFIT_ARS_STATUS_NONE)
-                       return -EAGAIN;
-
-               /*
-                * ARS interrupted, either we overflowed or some other
-                * agent wants the scan to stop.  If we didn't overflow
-                * then just continue with the returned results.
-                */
-               if (ars_status->status == NFIT_ARS_STATUS_INTR) {
-                       if (ars_status->flags & NFIT_ARS_F_OVERFLOW)
-                               return -ENOSPC;
-                       return 0;
-               }
-
-               /* Unknown status */
-               if (ars_status->status >> 16)
-                       return -EIO;
-               break;
-       case ND_CMD_CLEAR_ERROR:
-               clear_err = buf;
-               if (clear_err->status & 0xffff)
-                       return -EIO;
-               if (!clear_err->cleared)
-                       return -EIO;
-               if (clear_err->length > clear_err->cleared)
-                       return clear_err->cleared;
-               break;
-       default:
-               break;
-       }
-
-       return 0;
-}
-
-static int acpi_nfit_ctl(struct nvdimm_bus_descriptor *nd_desc,
-               struct nvdimm *nvdimm, unsigned int cmd, void *buf,
-               unsigned int buf_len, int *cmd_rc)
-{
-       struct acpi_nfit_desc *acpi_desc = to_acpi_nfit_desc(nd_desc);
-       union acpi_object in_obj, in_buf, *out_obj;
-       const struct nd_cmd_desc *desc = NULL;
-       struct device *dev = acpi_desc->dev;
-       struct nd_cmd_pkg *call_pkg = NULL;
-       const char *cmd_name, *dimm_name;
-       unsigned long cmd_mask, dsm_mask;
-       acpi_handle handle;
-       unsigned int func;
-       const u8 *uuid;
-       u32 offset;
-       int rc, i;
-
-       func = cmd;
-       if (cmd == ND_CMD_CALL) {
-               call_pkg = buf;
-               func = call_pkg->nd_command;
-       }
-
-       if (nvdimm) {
-               struct nfit_mem *nfit_mem = nvdimm_provider_data(nvdimm);
-               struct acpi_device *adev = nfit_mem->adev;
-
-               if (!adev)
-                       return -ENOTTY;
-               if (call_pkg && nfit_mem->family != call_pkg->nd_family)
-                       return -ENOTTY;
-
-               dimm_name = nvdimm_name(nvdimm);
-               cmd_name = nvdimm_cmd_name(cmd);
-               cmd_mask = nvdimm_cmd_mask(nvdimm);
-               dsm_mask = nfit_mem->dsm_mask;
-               desc = nd_cmd_dimm_desc(cmd);
-               uuid = to_nfit_uuid(nfit_mem->family);
-               handle = adev->handle;
-       } else {
-               struct acpi_device *adev = to_acpi_dev(acpi_desc);
-
-               cmd_name = nvdimm_bus_cmd_name(cmd);
-               cmd_mask = nd_desc->cmd_mask;
-               dsm_mask = cmd_mask;
-               desc = nd_cmd_bus_desc(cmd);
-               uuid = to_nfit_uuid(NFIT_DEV_BUS);
-               handle = adev->handle;
-               dimm_name = "bus";
-       }
-
-       if (!desc || (cmd && (desc->out_num + desc->in_num == 0)))
-               return -ENOTTY;
-
-       if (!test_bit(cmd, &cmd_mask) || !test_bit(func, &dsm_mask))
-               return -ENOTTY;
-
-       in_obj.type = ACPI_TYPE_PACKAGE;
-       in_obj.package.count = 1;
-       in_obj.package.elements = &in_buf;
-       in_buf.type = ACPI_TYPE_BUFFER;
-       in_buf.buffer.pointer = buf;
-       in_buf.buffer.length = 0;
-
-       /* libnvdimm has already validated the input envelope */
-       for (i = 0; i < desc->in_num; i++)
-               in_buf.buffer.length += nd_cmd_in_size(nvdimm, cmd, desc,
-                               i, buf);
-
-       if (call_pkg) {
-               /* skip over package wrapper */
-               in_buf.buffer.pointer = (void *) &call_pkg->nd_payload;
-               in_buf.buffer.length = call_pkg->nd_size_in;
-       }
-
-       if (IS_ENABLED(CONFIG_ACPI_NFIT_DEBUG)) {
-               dev_dbg(dev, "%s:%s cmd: %d: func: %d input length: %d\n",
-                               __func__, dimm_name, cmd, func,
-                               in_buf.buffer.length);
-               print_hex_dump_debug("nvdimm in  ", DUMP_PREFIX_OFFSET, 4, 4,
-                       in_buf.buffer.pointer,
-                       min_t(u32, 256, in_buf.buffer.length), true);
-       }
-
-       out_obj = acpi_evaluate_dsm(handle, uuid, 1, func, &in_obj);
-       if (!out_obj) {
-               dev_dbg(dev, "%s:%s _DSM failed cmd: %s\n", __func__, dimm_name,
-                               cmd_name);
-               return -EINVAL;
-       }
-
-       if (call_pkg) {
-               call_pkg->nd_fw_size = out_obj->buffer.length;
-               memcpy(call_pkg->nd_payload + call_pkg->nd_size_in,
-                       out_obj->buffer.pointer,
-                       min(call_pkg->nd_fw_size, call_pkg->nd_size_out));
-
-               ACPI_FREE(out_obj);
-               /*
-                * Need to support FW function w/o known size in advance.
-                * Caller can determine required size based upon nd_fw_size.
-                * If we return an error (like elsewhere) then caller wouldn't
-                * be able to rely upon data returned to make calculation.
-                */
-               return 0;
-       }
-
-       if (out_obj->package.type != ACPI_TYPE_BUFFER) {
-               dev_dbg(dev, "%s:%s unexpected output object type cmd: %s type: %d\n",
-                               __func__, dimm_name, cmd_name, out_obj->type);
-               rc = -EINVAL;
-               goto out;
-       }
-
-       if (IS_ENABLED(CONFIG_ACPI_NFIT_DEBUG)) {
-               dev_dbg(dev, "%s:%s cmd: %s output length: %d\n", __func__,
-                               dimm_name, cmd_name, out_obj->buffer.length);
-               print_hex_dump_debug(cmd_name, DUMP_PREFIX_OFFSET, 4,
-                               4, out_obj->buffer.pointer, min_t(u32, 128,
-                                       out_obj->buffer.length), true);
-       }
-
-       for (i = 0, offset = 0; i < desc->out_num; i++) {
-               u32 out_size = nd_cmd_out_size(nvdimm, cmd, desc, i, buf,
-                               (u32 *) out_obj->buffer.pointer);
-
-               if (offset + out_size > out_obj->buffer.length) {
-                       dev_dbg(dev, "%s:%s output object underflow cmd: %s field: %d\n",
-                                       __func__, dimm_name, cmd_name, i);
-                       break;
-               }
-
-               if (in_buf.buffer.length + offset + out_size > buf_len) {
-                       dev_dbg(dev, "%s:%s output overrun cmd: %s field: %d\n",
-                                       __func__, dimm_name, cmd_name, i);
-                       rc = -ENXIO;
-                       goto out;
-               }
-               memcpy(buf + in_buf.buffer.length + offset,
-                               out_obj->buffer.pointer + offset, out_size);
-               offset += out_size;
-       }
-       if (offset + in_buf.buffer.length < buf_len) {
-               if (i >= 1) {
-                       /*
-                        * status valid, return the number of bytes left
-                        * unfilled in the output buffer
-                        */
-                       rc = buf_len - offset - in_buf.buffer.length;
-                       if (cmd_rc)
-                               *cmd_rc = xlat_status(buf, cmd);
-               } else {
-                       dev_err(dev, "%s:%s underrun cmd: %s buf_len: %d out_len: %d\n",
-                                       __func__, dimm_name, cmd_name, buf_len,
-                                       offset);
-                       rc = -ENXIO;
-               }
-       } else {
-               rc = 0;
-               if (cmd_rc)
-                       *cmd_rc = xlat_status(buf, cmd);
-       }
-
- out:
-       ACPI_FREE(out_obj);
-
-       return rc;
-}
-
-static const char *spa_type_name(u16 type)
-{
-       static const char *to_name[] = {
-               [NFIT_SPA_VOLATILE] = "volatile",
-               [NFIT_SPA_PM] = "pmem",
-               [NFIT_SPA_DCR] = "dimm-control-region",
-               [NFIT_SPA_BDW] = "block-data-window",
-               [NFIT_SPA_VDISK] = "volatile-disk",
-               [NFIT_SPA_VCD] = "volatile-cd",
-               [NFIT_SPA_PDISK] = "persistent-disk",
-               [NFIT_SPA_PCD] = "persistent-cd",
-
-       };
-
-       if (type > NFIT_SPA_PCD)
-               return "unknown";
-
-       return to_name[type];
-}
-
-static int nfit_spa_type(struct acpi_nfit_system_address *spa)
-{
-       int i;
-
-       for (i = 0; i < NFIT_UUID_MAX; i++)
-               if (memcmp(to_nfit_uuid(i), spa->range_guid, 16) == 0)
-                       return i;
-       return -1;
-}
-
-static bool add_spa(struct acpi_nfit_desc *acpi_desc,
-               struct nfit_table_prev *prev,
-               struct acpi_nfit_system_address *spa)
-{
-       struct device *dev = acpi_desc->dev;
-       struct nfit_spa *nfit_spa;
-
-       if (spa->header.length != sizeof(*spa))
-               return false;
-
-       list_for_each_entry(nfit_spa, &prev->spas, list) {
-               if (memcmp(nfit_spa->spa, spa, sizeof(*spa)) == 0) {
-                       list_move_tail(&nfit_spa->list, &acpi_desc->spas);
-                       return true;
-               }
-       }
-
-       nfit_spa = devm_kzalloc(dev, sizeof(*nfit_spa) + sizeof(*spa),
-                       GFP_KERNEL);
-       if (!nfit_spa)
-               return false;
-       INIT_LIST_HEAD(&nfit_spa->list);
-       memcpy(nfit_spa->spa, spa, sizeof(*spa));
-       list_add_tail(&nfit_spa->list, &acpi_desc->spas);
-       dev_dbg(dev, "%s: spa index: %d type: %s\n", __func__,
-                       spa->range_index,
-                       spa_type_name(nfit_spa_type(spa)));
-       return true;
-}
-
-static bool add_memdev(struct acpi_nfit_desc *acpi_desc,
-               struct nfit_table_prev *prev,
-               struct acpi_nfit_memory_map *memdev)
-{
-       struct device *dev = acpi_desc->dev;
-       struct nfit_memdev *nfit_memdev;
-
-       if (memdev->header.length != sizeof(*memdev))
-               return false;
-
-       list_for_each_entry(nfit_memdev, &prev->memdevs, list)
-               if (memcmp(nfit_memdev->memdev, memdev, sizeof(*memdev)) == 0) {
-                       list_move_tail(&nfit_memdev->list, &acpi_desc->memdevs);
-                       return true;
-               }
-
-       nfit_memdev = devm_kzalloc(dev, sizeof(*nfit_memdev) + sizeof(*memdev),
-                       GFP_KERNEL);
-       if (!nfit_memdev)
-               return false;
-       INIT_LIST_HEAD(&nfit_memdev->list);
-       memcpy(nfit_memdev->memdev, memdev, sizeof(*memdev));
-       list_add_tail(&nfit_memdev->list, &acpi_desc->memdevs);
-       dev_dbg(dev, "%s: memdev handle: %#x spa: %d dcr: %d\n",
-                       __func__, memdev->device_handle, memdev->range_index,
-                       memdev->region_index);
-       return true;
-}
-
-/*
- * An implementation may provide a truncated control region if no block windows
- * are defined.
- */
-static size_t sizeof_dcr(struct acpi_nfit_control_region *dcr)
-{
-       if (dcr->header.length < offsetof(struct acpi_nfit_control_region,
-                               window_size))
-               return 0;
-       if (dcr->windows)
-               return sizeof(*dcr);
-       return offsetof(struct acpi_nfit_control_region, window_size);
-}
-
-static bool add_dcr(struct acpi_nfit_desc *acpi_desc,
-               struct nfit_table_prev *prev,
-               struct acpi_nfit_control_region *dcr)
-{
-       struct device *dev = acpi_desc->dev;
-       struct nfit_dcr *nfit_dcr;
-
-       if (!sizeof_dcr(dcr))
-               return false;
-
-       list_for_each_entry(nfit_dcr, &prev->dcrs, list)
-               if (memcmp(nfit_dcr->dcr, dcr, sizeof_dcr(dcr)) == 0) {
-                       list_move_tail(&nfit_dcr->list, &acpi_desc->dcrs);
-                       return true;
-               }
-
-       nfit_dcr = devm_kzalloc(dev, sizeof(*nfit_dcr) + sizeof(*dcr),
-                       GFP_KERNEL);
-       if (!nfit_dcr)
-               return false;
-       INIT_LIST_HEAD(&nfit_dcr->list);
-       memcpy(nfit_dcr->dcr, dcr, sizeof_dcr(dcr));
-       list_add_tail(&nfit_dcr->list, &acpi_desc->dcrs);
-       dev_dbg(dev, "%s: dcr index: %d windows: %d\n", __func__,
-                       dcr->region_index, dcr->windows);
-       return true;
-}
-
-static bool add_bdw(struct acpi_nfit_desc *acpi_desc,
-               struct nfit_table_prev *prev,
-               struct acpi_nfit_data_region *bdw)
-{
-       struct device *dev = acpi_desc->dev;
-       struct nfit_bdw *nfit_bdw;
-
-       if (bdw->header.length != sizeof(*bdw))
-               return false;
-       list_for_each_entry(nfit_bdw, &prev->bdws, list)
-               if (memcmp(nfit_bdw->bdw, bdw, sizeof(*bdw)) == 0) {
-                       list_move_tail(&nfit_bdw->list, &acpi_desc->bdws);
-                       return true;
-               }
-
-       nfit_bdw = devm_kzalloc(dev, sizeof(*nfit_bdw) + sizeof(*bdw),
-                       GFP_KERNEL);
-       if (!nfit_bdw)
-               return false;
-       INIT_LIST_HEAD(&nfit_bdw->list);
-       memcpy(nfit_bdw->bdw, bdw, sizeof(*bdw));
-       list_add_tail(&nfit_bdw->list, &acpi_desc->bdws);
-       dev_dbg(dev, "%s: bdw dcr: %d windows: %d\n", __func__,
-                       bdw->region_index, bdw->windows);
-       return true;
-}
-
-static size_t sizeof_idt(struct acpi_nfit_interleave *idt)
-{
-       if (idt->header.length < sizeof(*idt))
-               return 0;
-       return sizeof(*idt) + sizeof(u32) * (idt->line_count - 1);
-}
-
-static bool add_idt(struct acpi_nfit_desc *acpi_desc,
-               struct nfit_table_prev *prev,
-               struct acpi_nfit_interleave *idt)
-{
-       struct device *dev = acpi_desc->dev;
-       struct nfit_idt *nfit_idt;
-
-       if (!sizeof_idt(idt))
-               return false;
-
-       list_for_each_entry(nfit_idt, &prev->idts, list) {
-               if (sizeof_idt(nfit_idt->idt) != sizeof_idt(idt))
-                       continue;
-
-               if (memcmp(nfit_idt->idt, idt, sizeof_idt(idt)) == 0) {
-                       list_move_tail(&nfit_idt->list, &acpi_desc->idts);
-                       return true;
-               }
-       }
-
-       nfit_idt = devm_kzalloc(dev, sizeof(*nfit_idt) + sizeof_idt(idt),
-                       GFP_KERNEL);
-       if (!nfit_idt)
-               return false;
-       INIT_LIST_HEAD(&nfit_idt->list);
-       memcpy(nfit_idt->idt, idt, sizeof_idt(idt));
-       list_add_tail(&nfit_idt->list, &acpi_desc->idts);
-       dev_dbg(dev, "%s: idt index: %d num_lines: %d\n", __func__,
-                       idt->interleave_index, idt->line_count);
-       return true;
-}
-
-static size_t sizeof_flush(struct acpi_nfit_flush_address *flush)
-{
-       if (flush->header.length < sizeof(*flush))
-               return 0;
-       return sizeof(*flush) + sizeof(u64) * (flush->hint_count - 1);
-}
-
-static bool add_flush(struct acpi_nfit_desc *acpi_desc,
-               struct nfit_table_prev *prev,
-               struct acpi_nfit_flush_address *flush)
-{
-       struct device *dev = acpi_desc->dev;
-       struct nfit_flush *nfit_flush;
-
-       if (!sizeof_flush(flush))
-               return false;
-
-       list_for_each_entry(nfit_flush, &prev->flushes, list) {
-               if (sizeof_flush(nfit_flush->flush) != sizeof_flush(flush))
-                       continue;
-
-               if (memcmp(nfit_flush->flush, flush,
-                                       sizeof_flush(flush)) == 0) {
-                       list_move_tail(&nfit_flush->list, &acpi_desc->flushes);
-                       return true;
-               }
-       }
-
-       nfit_flush = devm_kzalloc(dev, sizeof(*nfit_flush)
-                       + sizeof_flush(flush), GFP_KERNEL);
-       if (!nfit_flush)
-               return false;
-       INIT_LIST_HEAD(&nfit_flush->list);
-       memcpy(nfit_flush->flush, flush, sizeof_flush(flush));
-       list_add_tail(&nfit_flush->list, &acpi_desc->flushes);
-       dev_dbg(dev, "%s: nfit_flush handle: %d hint_count: %d\n", __func__,
-                       flush->device_handle, flush->hint_count);
-       return true;
-}
-
-static void *add_table(struct acpi_nfit_desc *acpi_desc,
-               struct nfit_table_prev *prev, void *table, const void *end)
-{
-       struct device *dev = acpi_desc->dev;
-       struct acpi_nfit_header *hdr;
-       void *err = ERR_PTR(-ENOMEM);
-
-       if (table >= end)
-               return NULL;
-
-       hdr = table;
-       if (!hdr->length) {
-               dev_warn(dev, "found a zero length table '%d' parsing nfit\n",
-                       hdr->type);
-               return NULL;
-       }
-
-       switch (hdr->type) {
-       case ACPI_NFIT_TYPE_SYSTEM_ADDRESS:
-               if (!add_spa(acpi_desc, prev, table))
-                       return err;
-               break;
-       case ACPI_NFIT_TYPE_MEMORY_MAP:
-               if (!add_memdev(acpi_desc, prev, table))
-                       return err;
-               break;
-       case ACPI_NFIT_TYPE_CONTROL_REGION:
-               if (!add_dcr(acpi_desc, prev, table))
-                       return err;
-               break;
-       case ACPI_NFIT_TYPE_DATA_REGION:
-               if (!add_bdw(acpi_desc, prev, table))
-                       return err;
-               break;
-       case ACPI_NFIT_TYPE_INTERLEAVE:
-               if (!add_idt(acpi_desc, prev, table))
-                       return err;
-               break;
-       case ACPI_NFIT_TYPE_FLUSH_ADDRESS:
-               if (!add_flush(acpi_desc, prev, table))
-                       return err;
-               break;
-       case ACPI_NFIT_TYPE_SMBIOS:
-               dev_dbg(dev, "%s: smbios\n", __func__);
-               break;
-       default:
-               dev_err(dev, "unknown table '%d' parsing nfit\n", hdr->type);
-               break;
-       }
-
-       return table + hdr->length;
-}
-
-static void nfit_mem_find_spa_bdw(struct acpi_nfit_desc *acpi_desc,
-               struct nfit_mem *nfit_mem)
-{
-       u32 device_handle = __to_nfit_memdev(nfit_mem)->device_handle;
-       u16 dcr = nfit_mem->dcr->region_index;
-       struct nfit_spa *nfit_spa;
-
-       list_for_each_entry(nfit_spa, &acpi_desc->spas, list) {
-               u16 range_index = nfit_spa->spa->range_index;
-               int type = nfit_spa_type(nfit_spa->spa);
-               struct nfit_memdev *nfit_memdev;
-
-               if (type != NFIT_SPA_BDW)
-                       continue;
-
-               list_for_each_entry(nfit_memdev, &acpi_desc->memdevs, list) {
-                       if (nfit_memdev->memdev->range_index != range_index)
-                               continue;
-                       if (nfit_memdev->memdev->device_handle != device_handle)
-                               continue;
-                       if (nfit_memdev->memdev->region_index != dcr)
-                               continue;
-
-                       nfit_mem->spa_bdw = nfit_spa->spa;
-                       return;
-               }
-       }
-
-       dev_dbg(acpi_desc->dev, "SPA-BDW not found for SPA-DCR %d\n",
-                       nfit_mem->spa_dcr->range_index);
-       nfit_mem->bdw = NULL;
-}
-
-static void nfit_mem_init_bdw(struct acpi_nfit_desc *acpi_desc,
-               struct nfit_mem *nfit_mem, struct acpi_nfit_system_address *spa)
-{
-       u16 dcr = __to_nfit_memdev(nfit_mem)->region_index;
-       struct nfit_memdev *nfit_memdev;
-       struct nfit_bdw *nfit_bdw;
-       struct nfit_idt *nfit_idt;
-       u16 idt_idx, range_index;
-
-       list_for_each_entry(nfit_bdw, &acpi_desc->bdws, list) {
-               if (nfit_bdw->bdw->region_index != dcr)
-                       continue;
-               nfit_mem->bdw = nfit_bdw->bdw;
-               break;
-       }
-
-       if (!nfit_mem->bdw)
-               return;
-
-       nfit_mem_find_spa_bdw(acpi_desc, nfit_mem);
-
-       if (!nfit_mem->spa_bdw)
-               return;
-
-       range_index = nfit_mem->spa_bdw->range_index;
-       list_for_each_entry(nfit_memdev, &acpi_desc->memdevs, list) {
-               if (nfit_memdev->memdev->range_index != range_index ||
-                               nfit_memdev->memdev->region_index != dcr)
-                       continue;
-               nfit_mem->memdev_bdw = nfit_memdev->memdev;
-               idt_idx = nfit_memdev->memdev->interleave_index;
-               list_for_each_entry(nfit_idt, &acpi_desc->idts, list) {
-                       if (nfit_idt->idt->interleave_index != idt_idx)
-                               continue;
-                       nfit_mem->idt_bdw = nfit_idt->idt;
-                       break;
-               }
-               break;
-       }
-}
-
-static int nfit_mem_dcr_init(struct acpi_nfit_desc *acpi_desc,
-               struct acpi_nfit_system_address *spa)
-{
-       struct nfit_mem *nfit_mem, *found;
-       struct nfit_memdev *nfit_memdev;
-       int type = nfit_spa_type(spa);
-
-       switch (type) {
-       case NFIT_SPA_DCR:
-       case NFIT_SPA_PM:
-               break;
-       default:
-               return 0;
-       }
-
-       list_for_each_entry(nfit_memdev, &acpi_desc->memdevs, list) {
-               struct nfit_flush *nfit_flush;
-               struct nfit_dcr *nfit_dcr;
-               u32 device_handle;
-               u16 dcr;
-
-               if (nfit_memdev->memdev->range_index != spa->range_index)
-                       continue;
-               found = NULL;
-               dcr = nfit_memdev->memdev->region_index;
-               device_handle = nfit_memdev->memdev->device_handle;
-               list_for_each_entry(nfit_mem, &acpi_desc->dimms, list)
-                       if (__to_nfit_memdev(nfit_mem)->device_handle
-                                       == device_handle) {
-                               found = nfit_mem;
-                               break;
-                       }
-
-               if (found)
-                       nfit_mem = found;
-               else {
-                       nfit_mem = devm_kzalloc(acpi_desc->dev,
-                                       sizeof(*nfit_mem), GFP_KERNEL);
-                       if (!nfit_mem)
-                               return -ENOMEM;
-                       INIT_LIST_HEAD(&nfit_mem->list);
-                       nfit_mem->acpi_desc = acpi_desc;
-                       list_add(&nfit_mem->list, &acpi_desc->dimms);
-               }
-
-               list_for_each_entry(nfit_dcr, &acpi_desc->dcrs, list) {
-                       if (nfit_dcr->dcr->region_index != dcr)
-                               continue;
-                       /*
-                        * Record the control region for the dimm.  For
-                        * the ACPI 6.1 case, where there are separate
-                        * control regions for the pmem vs blk
-                        * interfaces, be sure to record the extended
-                        * blk details.
-                        */
-                       if (!nfit_mem->dcr)
-                               nfit_mem->dcr = nfit_dcr->dcr;
-                       else if (nfit_mem->dcr->windows == 0
-                                       && nfit_dcr->dcr->windows)
-                               nfit_mem->dcr = nfit_dcr->dcr;
-                       break;
-               }
-
-               list_for_each_entry(nfit_flush, &acpi_desc->flushes, list) {
-                       struct acpi_nfit_flush_address *flush;
-                       u16 i;
-
-                       if (nfit_flush->flush->device_handle != device_handle)
-                               continue;
-                       nfit_mem->nfit_flush = nfit_flush;
-                       flush = nfit_flush->flush;
-                       nfit_mem->flush_wpq = devm_kzalloc(acpi_desc->dev,
-                                       flush->hint_count
-                                       * sizeof(struct resource), GFP_KERNEL);
-                       if (!nfit_mem->flush_wpq)
-                               return -ENOMEM;
-                       for (i = 0; i < flush->hint_count; i++) {
-                               struct resource *res = &nfit_mem->flush_wpq[i];
-
-                               res->start = flush->hint_address[i];
-                               res->end = res->start + 8 - 1;
-                       }
-                       break;
-               }
-
-               if (dcr && !nfit_mem->dcr) {
-                       dev_err(acpi_desc->dev, "SPA %d missing DCR %d\n",
-                                       spa->range_index, dcr);
-                       return -ENODEV;
-               }
-
-               if (type == NFIT_SPA_DCR) {
-                       struct nfit_idt *nfit_idt;
-                       u16 idt_idx;
-
-                       /* multiple dimms may share a SPA when interleaved */
-                       nfit_mem->spa_dcr = spa;
-                       nfit_mem->memdev_dcr = nfit_memdev->memdev;
-                       idt_idx = nfit_memdev->memdev->interleave_index;
-                       list_for_each_entry(nfit_idt, &acpi_desc->idts, list) {
-                               if (nfit_idt->idt->interleave_index != idt_idx)
-                                       continue;
-                               nfit_mem->idt_dcr = nfit_idt->idt;
-                               break;
-                       }
-                       nfit_mem_init_bdw(acpi_desc, nfit_mem, spa);
-               } else {
-                       /*
-                        * A single dimm may belong to multiple SPA-PM
-                        * ranges, record at least one in addition to
-                        * any SPA-DCR range.
-                        */
-                       nfit_mem->memdev_pmem = nfit_memdev->memdev;
-               }
-       }
-
-       return 0;
-}
-
-static int nfit_mem_cmp(void *priv, struct list_head *_a, struct list_head *_b)
-{
-       struct nfit_mem *a = container_of(_a, typeof(*a), list);
-       struct nfit_mem *b = container_of(_b, typeof(*b), list);
-       u32 handleA, handleB;
-
-       handleA = __to_nfit_memdev(a)->device_handle;
-       handleB = __to_nfit_memdev(b)->device_handle;
-       if (handleA < handleB)
-               return -1;
-       else if (handleA > handleB)
-               return 1;
-       return 0;
-}
-
-static int nfit_mem_init(struct acpi_nfit_desc *acpi_desc)
-{
-       struct nfit_spa *nfit_spa;
-
-       /*
-        * For each SPA-DCR or SPA-PMEM address range find its
-        * corresponding MEMDEV(s).  From each MEMDEV find the
-        * corresponding DCR.  Then, if we're operating on a SPA-DCR,
-        * try to find a SPA-BDW and a corresponding BDW that references
-        * the DCR.  Throw it all into an nfit_mem object.  Note, that
-        * BDWs are optional.
-        */
-       list_for_each_entry(nfit_spa, &acpi_desc->spas, list) {
-               int rc;
-
-               rc = nfit_mem_dcr_init(acpi_desc, nfit_spa->spa);
-               if (rc)
-                       return rc;
-       }
-
-       list_sort(NULL, &acpi_desc->dimms, nfit_mem_cmp);
-
-       return 0;
-}
-
-static ssize_t revision_show(struct device *dev,
-               struct device_attribute *attr, char *buf)
-{
-       struct nvdimm_bus *nvdimm_bus = to_nvdimm_bus(dev);
-       struct nvdimm_bus_descriptor *nd_desc = to_nd_desc(nvdimm_bus);
-       struct acpi_nfit_desc *acpi_desc = to_acpi_desc(nd_desc);
-
-       return sprintf(buf, "%d\n", acpi_desc->acpi_header.revision);
-}
-static DEVICE_ATTR_RO(revision);
-
-/*
- * This shows the number of full Address Range Scrubs that have been
- * completed since driver load time. Userspace can wait on this using
- * select/poll etc. A '+' at the end indicates an ARS is in progress
- */
-static ssize_t scrub_show(struct device *dev,
-               struct device_attribute *attr, char *buf)
-{
-       struct nvdimm_bus_descriptor *nd_desc;
-       ssize_t rc = -ENXIO;
-
-       device_lock(dev);
-       nd_desc = dev_get_drvdata(dev);
-       if (nd_desc) {
-               struct acpi_nfit_desc *acpi_desc = to_acpi_desc(nd_desc);
-
-               rc = sprintf(buf, "%d%s", acpi_desc->scrub_count,
-                               (work_busy(&acpi_desc->work)) ? "+\n" : "\n");
-       }
-       device_unlock(dev);
-       return rc;
-}
-
-static int acpi_nfit_ars_rescan(struct acpi_nfit_desc *acpi_desc);
-
-static ssize_t scrub_store(struct device *dev,
-               struct device_attribute *attr, const char *buf, size_t size)
-{
-       struct nvdimm_bus_descriptor *nd_desc;
-       ssize_t rc;
-       long val;
-
-       rc = kstrtol(buf, 0, &val);
-       if (rc)
-               return rc;
-       if (val != 1)
-               return -EINVAL;
-
-       device_lock(dev);
-       nd_desc = dev_get_drvdata(dev);
-       if (nd_desc) {
-               struct acpi_nfit_desc *acpi_desc = to_acpi_desc(nd_desc);
-
-               rc = acpi_nfit_ars_rescan(acpi_desc);
-       }
-       device_unlock(dev);
-       if (rc)
-               return rc;
-       return size;
-}
-static DEVICE_ATTR_RW(scrub);
-
-static bool ars_supported(struct nvdimm_bus *nvdimm_bus)
-{
-       struct nvdimm_bus_descriptor *nd_desc = to_nd_desc(nvdimm_bus);
-       const unsigned long mask = 1 << ND_CMD_ARS_CAP | 1 << ND_CMD_ARS_START
-               | 1 << ND_CMD_ARS_STATUS;
-
-       return (nd_desc->cmd_mask & mask) == mask;
-}
-
-static umode_t nfit_visible(struct kobject *kobj, struct attribute *a, int n)
-{
-       struct device *dev = container_of(kobj, struct device, kobj);
-       struct nvdimm_bus *nvdimm_bus = to_nvdimm_bus(dev);
-
-       if (a == &dev_attr_scrub.attr && !ars_supported(nvdimm_bus))
-               return 0;
-       return a->mode;
-}
-
-static struct attribute *acpi_nfit_attributes[] = {
-       &dev_attr_revision.attr,
-       &dev_attr_scrub.attr,
-       NULL,
-};
-
-static struct attribute_group acpi_nfit_attribute_group = {
-       .name = "nfit",
-       .attrs = acpi_nfit_attributes,
-       .is_visible = nfit_visible,
-};
-
-static const struct attribute_group *acpi_nfit_attribute_groups[] = {
-       &nvdimm_bus_attribute_group,
-       &acpi_nfit_attribute_group,
-       NULL,
-};
-
-static struct acpi_nfit_memory_map *to_nfit_memdev(struct device *dev)
-{
-       struct nvdimm *nvdimm = to_nvdimm(dev);
-       struct nfit_mem *nfit_mem = nvdimm_provider_data(nvdimm);
-
-       return __to_nfit_memdev(nfit_mem);
-}
-
-static struct acpi_nfit_control_region *to_nfit_dcr(struct device *dev)
-{
-       struct nvdimm *nvdimm = to_nvdimm(dev);
-       struct nfit_mem *nfit_mem = nvdimm_provider_data(nvdimm);
-
-       return nfit_mem->dcr;
-}
-
-static ssize_t handle_show(struct device *dev,
-               struct device_attribute *attr, char *buf)
-{
-       struct acpi_nfit_memory_map *memdev = to_nfit_memdev(dev);
-
-       return sprintf(buf, "%#x\n", memdev->device_handle);
-}
-static DEVICE_ATTR_RO(handle);
-
-static ssize_t phys_id_show(struct device *dev,
-               struct device_attribute *attr, char *buf)
-{
-       struct acpi_nfit_memory_map *memdev = to_nfit_memdev(dev);
-
-       return sprintf(buf, "%#x\n", memdev->physical_id);
-}
-static DEVICE_ATTR_RO(phys_id);
-
-static ssize_t vendor_show(struct device *dev,
-               struct device_attribute *attr, char *buf)
-{
-       struct acpi_nfit_control_region *dcr = to_nfit_dcr(dev);
-
-       return sprintf(buf, "0x%04x\n", be16_to_cpu(dcr->vendor_id));
-}
-static DEVICE_ATTR_RO(vendor);
-
-static ssize_t rev_id_show(struct device *dev,
-               struct device_attribute *attr, char *buf)
-{
-       struct acpi_nfit_control_region *dcr = to_nfit_dcr(dev);
-
-       return sprintf(buf, "0x%04x\n", be16_to_cpu(dcr->revision_id));
-}
-static DEVICE_ATTR_RO(rev_id);
-
-static ssize_t device_show(struct device *dev,
-               struct device_attribute *attr, char *buf)
-{
-       struct acpi_nfit_control_region *dcr = to_nfit_dcr(dev);
-
-       return sprintf(buf, "0x%04x\n", be16_to_cpu(dcr->device_id));
-}
-static DEVICE_ATTR_RO(device);
-
-static ssize_t subsystem_vendor_show(struct device *dev,
-               struct device_attribute *attr, char *buf)
-{
-       struct acpi_nfit_control_region *dcr = to_nfit_dcr(dev);
-
-       return sprintf(buf, "0x%04x\n", be16_to_cpu(dcr->subsystem_vendor_id));
-}
-static DEVICE_ATTR_RO(subsystem_vendor);
-
-static ssize_t subsystem_rev_id_show(struct device *dev,
-               struct device_attribute *attr, char *buf)
-{
-       struct acpi_nfit_control_region *dcr = to_nfit_dcr(dev);
-
-       return sprintf(buf, "0x%04x\n",
-                       be16_to_cpu(dcr->subsystem_revision_id));
-}
-static DEVICE_ATTR_RO(subsystem_rev_id);
-
-static ssize_t subsystem_device_show(struct device *dev,
-               struct device_attribute *attr, char *buf)
-{
-       struct acpi_nfit_control_region *dcr = to_nfit_dcr(dev);
-
-       return sprintf(buf, "0x%04x\n", be16_to_cpu(dcr->subsystem_device_id));
-}
-static DEVICE_ATTR_RO(subsystem_device);
-
-static int num_nvdimm_formats(struct nvdimm *nvdimm)
-{
-       struct nfit_mem *nfit_mem = nvdimm_provider_data(nvdimm);
-       int formats = 0;
-
-       if (nfit_mem->memdev_pmem)
-               formats++;
-       if (nfit_mem->memdev_bdw)
-               formats++;
-       return formats;
-}
-
-static ssize_t format_show(struct device *dev,
-               struct device_attribute *attr, char *buf)
-{
-       struct acpi_nfit_control_region *dcr = to_nfit_dcr(dev);
-
-       return sprintf(buf, "0x%04x\n", be16_to_cpu(dcr->code));
-}
-static DEVICE_ATTR_RO(format);
-
-static ssize_t format1_show(struct device *dev,
-               struct device_attribute *attr, char *buf)
-{
-       u32 handle;
-       ssize_t rc = -ENXIO;
-       struct nfit_mem *nfit_mem;
-       struct nfit_memdev *nfit_memdev;
-       struct acpi_nfit_desc *acpi_desc;
-       struct nvdimm *nvdimm = to_nvdimm(dev);
-       struct acpi_nfit_control_region *dcr = to_nfit_dcr(dev);
-
-       nfit_mem = nvdimm_provider_data(nvdimm);
-       acpi_desc = nfit_mem->acpi_desc;
-       handle = to_nfit_memdev(dev)->device_handle;
-
-       /* assumes DIMMs have at most 2 published interface codes */
-       mutex_lock(&acpi_desc->init_mutex);
-       list_for_each_entry(nfit_memdev, &acpi_desc->memdevs, list) {
-               struct acpi_nfit_memory_map *memdev = nfit_memdev->memdev;
-               struct nfit_dcr *nfit_dcr;
-
-               if (memdev->device_handle != handle)
-                       continue;
-
-               list_for_each_entry(nfit_dcr, &acpi_desc->dcrs, list) {
-                       if (nfit_dcr->dcr->region_index != memdev->region_index)
-                               continue;
-                       if (nfit_dcr->dcr->code == dcr->code)
-                               continue;
-                       rc = sprintf(buf, "%#x\n",
-                                       be16_to_cpu(nfit_dcr->dcr->code));
-                       break;
-               }
-               if (rc != ENXIO)
-                       break;
-       }
-       mutex_unlock(&acpi_desc->init_mutex);
-       return rc;
-}
-static DEVICE_ATTR_RO(format1);
-
-static ssize_t formats_show(struct device *dev,
-               struct device_attribute *attr, char *buf)
-{
-       struct nvdimm *nvdimm = to_nvdimm(dev);
-
-       return sprintf(buf, "%d\n", num_nvdimm_formats(nvdimm));
-}
-static DEVICE_ATTR_RO(formats);
-
-static ssize_t serial_show(struct device *dev,
-               struct device_attribute *attr, char *buf)
-{
-       struct acpi_nfit_control_region *dcr = to_nfit_dcr(dev);
-
-       return sprintf(buf, "0x%08x\n", be32_to_cpu(dcr->serial_number));
-}
-static DEVICE_ATTR_RO(serial);
-
-static ssize_t family_show(struct device *dev,
-               struct device_attribute *attr, char *buf)
-{
-       struct nvdimm *nvdimm = to_nvdimm(dev);
-       struct nfit_mem *nfit_mem = nvdimm_provider_data(nvdimm);
-
-       if (nfit_mem->family < 0)
-               return -ENXIO;
-       return sprintf(buf, "%d\n", nfit_mem->family);
-}
-static DEVICE_ATTR_RO(family);
-
-static ssize_t dsm_mask_show(struct device *dev,
-               struct device_attribute *attr, char *buf)
-{
-       struct nvdimm *nvdimm = to_nvdimm(dev);
-       struct nfit_mem *nfit_mem = nvdimm_provider_data(nvdimm);
-
-       if (nfit_mem->family < 0)
-               return -ENXIO;
-       return sprintf(buf, "%#lx\n", nfit_mem->dsm_mask);
-}
-static DEVICE_ATTR_RO(dsm_mask);
-
-static ssize_t flags_show(struct device *dev,
-               struct device_attribute *attr, char *buf)
-{
-       u16 flags = to_nfit_memdev(dev)->flags;
-
-       return sprintf(buf, "%s%s%s%s%s\n",
-               flags & ACPI_NFIT_MEM_SAVE_FAILED ? "save_fail " : "",
-               flags & ACPI_NFIT_MEM_RESTORE_FAILED ? "restore_fail " : "",
-               flags & ACPI_NFIT_MEM_FLUSH_FAILED ? "flush_fail " : "",
-               flags & ACPI_NFIT_MEM_NOT_ARMED ? "not_armed " : "",
-               flags & ACPI_NFIT_MEM_HEALTH_OBSERVED ? "smart_event " : "");
-}
-static DEVICE_ATTR_RO(flags);
-
-static ssize_t id_show(struct device *dev,
-               struct device_attribute *attr, char *buf)
-{
-       struct acpi_nfit_control_region *dcr = to_nfit_dcr(dev);
-
-       if (dcr->valid_fields & ACPI_NFIT_CONTROL_MFG_INFO_VALID)
-               return sprintf(buf, "%04x-%02x-%04x-%08x\n",
-                               be16_to_cpu(dcr->vendor_id),
-                               dcr->manufacturing_location,
-                               be16_to_cpu(dcr->manufacturing_date),
-                               be32_to_cpu(dcr->serial_number));
-       else
-               return sprintf(buf, "%04x-%08x\n",
-                               be16_to_cpu(dcr->vendor_id),
-                               be32_to_cpu(dcr->serial_number));
-}
-static DEVICE_ATTR_RO(id);
-
-static struct attribute *acpi_nfit_dimm_attributes[] = {
-       &dev_attr_handle.attr,
-       &dev_attr_phys_id.attr,
-       &dev_attr_vendor.attr,
-       &dev_attr_device.attr,
-       &dev_attr_rev_id.attr,
-       &dev_attr_subsystem_vendor.attr,
-       &dev_attr_subsystem_device.attr,
-       &dev_attr_subsystem_rev_id.attr,
-       &dev_attr_format.attr,
-       &dev_attr_formats.attr,
-       &dev_attr_format1.attr,
-       &dev_attr_serial.attr,
-       &dev_attr_flags.attr,
-       &dev_attr_id.attr,
-       &dev_attr_family.attr,
-       &dev_attr_dsm_mask.attr,
-       NULL,
-};
-
-static umode_t acpi_nfit_dimm_attr_visible(struct kobject *kobj,
-               struct attribute *a, int n)
-{
-       struct device *dev = container_of(kobj, struct device, kobj);
-       struct nvdimm *nvdimm = to_nvdimm(dev);
-
-       if (!to_nfit_dcr(dev))
-               return 0;
-       if (a == &dev_attr_format1.attr && num_nvdimm_formats(nvdimm) <= 1)
-               return 0;
-       return a->mode;
-}
-
-static struct attribute_group acpi_nfit_dimm_attribute_group = {
-       .name = "nfit",
-       .attrs = acpi_nfit_dimm_attributes,
-       .is_visible = acpi_nfit_dimm_attr_visible,
-};
-
-static const struct attribute_group *acpi_nfit_dimm_attribute_groups[] = {
-       &nvdimm_attribute_group,
-       &nd_device_attribute_group,
-       &acpi_nfit_dimm_attribute_group,
-       NULL,
-};
-
-static struct nvdimm *acpi_nfit_dimm_by_handle(struct acpi_nfit_desc *acpi_desc,
-               u32 device_handle)
-{
-       struct nfit_mem *nfit_mem;
-
-       list_for_each_entry(nfit_mem, &acpi_desc->dimms, list)
-               if (__to_nfit_memdev(nfit_mem)->device_handle == device_handle)
-                       return nfit_mem->nvdimm;
-
-       return NULL;
-}
-
-static int acpi_nfit_add_dimm(struct acpi_nfit_desc *acpi_desc,
-               struct nfit_mem *nfit_mem, u32 device_handle)
-{
-       struct acpi_device *adev, *adev_dimm;
-       struct device *dev = acpi_desc->dev;
-       unsigned long dsm_mask;
-       const u8 *uuid;
-       int i;
-
-       /* nfit test assumes 1:1 relationship between commands and dsms */
-       nfit_mem->dsm_mask = acpi_desc->dimm_cmd_force_en;
-       nfit_mem->family = NVDIMM_FAMILY_INTEL;
-       adev = to_acpi_dev(acpi_desc);
-       if (!adev)
-               return 0;
-
-       adev_dimm = acpi_find_child_device(adev, device_handle, false);
-       nfit_mem->adev = adev_dimm;
-       if (!adev_dimm) {
-               dev_err(dev, "no ACPI.NFIT device with _ADR %#x, disabling...\n",
-                               device_handle);
-               return force_enable_dimms ? 0 : -ENODEV;
-       }
-
-       /*
-        * Until standardization materializes we need to consider 4
-        * different command sets.  Note, that checking for function0 (bit0)
-        * tells us if any commands are reachable through this uuid.
-        */
-       for (i = NVDIMM_FAMILY_INTEL; i <= NVDIMM_FAMILY_MSFT; i++)
-               if (acpi_check_dsm(adev_dimm->handle, to_nfit_uuid(i), 1, 1))
-                       break;
-
-       /* limit the supported commands to those that are publicly documented */
-       nfit_mem->family = i;
-       if (nfit_mem->family == NVDIMM_FAMILY_INTEL) {
-               dsm_mask = 0x3fe;
-               if (disable_vendor_specific)
-                       dsm_mask &= ~(1 << ND_CMD_VENDOR);
-       } else if (nfit_mem->family == NVDIMM_FAMILY_HPE1) {
-               dsm_mask = 0x1c3c76;
-       } else if (nfit_mem->family == NVDIMM_FAMILY_HPE2) {
-               dsm_mask = 0x1fe;
-               if (disable_vendor_specific)
-                       dsm_mask &= ~(1 << 8);
-       } else if (nfit_mem->family == NVDIMM_FAMILY_MSFT) {
-               dsm_mask = 0xffffffff;
-       } else {
-               dev_err(dev, "unknown dimm command family\n");
-               nfit_mem->family = -1;
-               return force_enable_dimms ? 0 : -ENODEV;
-       }
-
-       uuid = to_nfit_uuid(nfit_mem->family);
-       for_each_set_bit(i, &dsm_mask, BITS_PER_LONG)
-               if (acpi_check_dsm(adev_dimm->handle, uuid, 1, 1ULL << i))
-                       set_bit(i, &nfit_mem->dsm_mask);
-
-       return 0;
-}
-
-static int acpi_nfit_register_dimms(struct acpi_nfit_desc *acpi_desc)
-{
-       struct nfit_mem *nfit_mem;
-       int dimm_count = 0;
-
-       list_for_each_entry(nfit_mem, &acpi_desc->dimms, list) {
-               struct acpi_nfit_flush_address *flush;
-               unsigned long flags = 0, cmd_mask;
-               struct nvdimm *nvdimm;
-               u32 device_handle;
-               u16 mem_flags;
-               int rc;
-
-               device_handle = __to_nfit_memdev(nfit_mem)->device_handle;
-               nvdimm = acpi_nfit_dimm_by_handle(acpi_desc, device_handle);
-               if (nvdimm) {
-                       dimm_count++;
-                       continue;
-               }
-
-               if (nfit_mem->bdw && nfit_mem->memdev_pmem)
-                       flags |= NDD_ALIASING;
-
-               mem_flags = __to_nfit_memdev(nfit_mem)->flags;
-               if (mem_flags & ACPI_NFIT_MEM_NOT_ARMED)
-                       flags |= NDD_UNARMED;
-
-               rc = acpi_nfit_add_dimm(acpi_desc, nfit_mem, device_handle);
-               if (rc)
-                       continue;
-
-               /*
-                * TODO: provide translation for non-NVDIMM_FAMILY_INTEL
-                * devices (i.e. from nd_cmd to acpi_dsm) to standardize the
-                * userspace interface.
-                */
-               cmd_mask = 1UL << ND_CMD_CALL;
-               if (nfit_mem->family == NVDIMM_FAMILY_INTEL)
-                       cmd_mask |= nfit_mem->dsm_mask;
-
-               flush = nfit_mem->nfit_flush ? nfit_mem->nfit_flush->flush
-                       : NULL;
-               nvdimm = nvdimm_create(acpi_desc->nvdimm_bus, nfit_mem,
-                               acpi_nfit_dimm_attribute_groups,
-                               flags, cmd_mask, flush ? flush->hint_count : 0,
-                               nfit_mem->flush_wpq);
-               if (!nvdimm)
-                       return -ENOMEM;
-
-               nfit_mem->nvdimm = nvdimm;
-               dimm_count++;
-
-               if ((mem_flags & ACPI_NFIT_MEM_FAILED_MASK) == 0)
-                       continue;
-
-               dev_info(acpi_desc->dev, "%s flags:%s%s%s%s\n",
-                               nvdimm_name(nvdimm),
-                 mem_flags & ACPI_NFIT_MEM_SAVE_FAILED ? " save_fail" : "",
-                 mem_flags & ACPI_NFIT_MEM_RESTORE_FAILED ? " restore_fail":"",
-                 mem_flags & ACPI_NFIT_MEM_FLUSH_FAILED ? " flush_fail" : "",
-                 mem_flags & ACPI_NFIT_MEM_NOT_ARMED ? " not_armed" : "");
-
-       }
-
-       return nvdimm_bus_check_dimm_count(acpi_desc->nvdimm_bus, dimm_count);
-}
-
-static void acpi_nfit_init_dsms(struct acpi_nfit_desc *acpi_desc)
-{
-       struct nvdimm_bus_descriptor *nd_desc = &acpi_desc->nd_desc;
-       const u8 *uuid = to_nfit_uuid(NFIT_DEV_BUS);
-       struct acpi_device *adev;
-       int i;
-
-       nd_desc->cmd_mask = acpi_desc->bus_cmd_force_en;
-       adev = to_acpi_dev(acpi_desc);
-       if (!adev)
-               return;
-
-       for (i = ND_CMD_ARS_CAP; i <= ND_CMD_CLEAR_ERROR; i++)
-               if (acpi_check_dsm(adev->handle, uuid, 1, 1ULL << i))
-                       set_bit(i, &nd_desc->cmd_mask);
-}
-
-static ssize_t range_index_show(struct device *dev,
-               struct device_attribute *attr, char *buf)
-{
-       struct nd_region *nd_region = to_nd_region(dev);
-       struct nfit_spa *nfit_spa = nd_region_provider_data(nd_region);
-
-       return sprintf(buf, "%d\n", nfit_spa->spa->range_index);
-}
-static DEVICE_ATTR_RO(range_index);
-
-static struct attribute *acpi_nfit_region_attributes[] = {
-       &dev_attr_range_index.attr,
-       NULL,
-};
-
-static struct attribute_group acpi_nfit_region_attribute_group = {
-       .name = "nfit",
-       .attrs = acpi_nfit_region_attributes,
-};
-
-static const struct attribute_group *acpi_nfit_region_attribute_groups[] = {
-       &nd_region_attribute_group,
-       &nd_mapping_attribute_group,
-       &nd_device_attribute_group,
-       &nd_numa_attribute_group,
-       &acpi_nfit_region_attribute_group,
-       NULL,
-};
-
-/* enough info to uniquely specify an interleave set */
-struct nfit_set_info {
-       struct nfit_set_info_map {
-               u64 region_offset;
-               u32 serial_number;
-               u32 pad;
-       } mapping[0];
-};
-
-static size_t sizeof_nfit_set_info(int num_mappings)
-{
-       return sizeof(struct nfit_set_info)
-               + num_mappings * sizeof(struct nfit_set_info_map);
-}
-
-static int cmp_map(const void *m0, const void *m1)
-{
-       const struct nfit_set_info_map *map0 = m0;
-       const struct nfit_set_info_map *map1 = m1;
-
-       return memcmp(&map0->region_offset, &map1->region_offset,
-                       sizeof(u64));
-}
-
-/* Retrieve the nth entry referencing this spa */
-static struct acpi_nfit_memory_map *memdev_from_spa(
-               struct acpi_nfit_desc *acpi_desc, u16 range_index, int n)
-{
-       struct nfit_memdev *nfit_memdev;
-
-       list_for_each_entry(nfit_memdev, &acpi_desc->memdevs, list)
-               if (nfit_memdev->memdev->range_index == range_index)
-                       if (n-- == 0)
-                               return nfit_memdev->memdev;
-       return NULL;
-}
-
-static int acpi_nfit_init_interleave_set(struct acpi_nfit_desc *acpi_desc,
-               struct nd_region_desc *ndr_desc,
-               struct acpi_nfit_system_address *spa)
-{
-       int i, spa_type = nfit_spa_type(spa);
-       struct device *dev = acpi_desc->dev;
-       struct nd_interleave_set *nd_set;
-       u16 nr = ndr_desc->num_mappings;
-       struct nfit_set_info *info;
-
-       if (spa_type == NFIT_SPA_PM || spa_type == NFIT_SPA_VOLATILE)
-               /* pass */;
-       else
-               return 0;
-
-       nd_set = devm_kzalloc(dev, sizeof(*nd_set), GFP_KERNEL);
-       if (!nd_set)
-               return -ENOMEM;
-
-       info = devm_kzalloc(dev, sizeof_nfit_set_info(nr), GFP_KERNEL);
-       if (!info)
-               return -ENOMEM;
-       for (i = 0; i < nr; i++) {
-               struct nd_mapping *nd_mapping = &ndr_desc->nd_mapping[i];
-               struct nfit_set_info_map *map = &info->mapping[i];
-               struct nvdimm *nvdimm = nd_mapping->nvdimm;
-               struct nfit_mem *nfit_mem = nvdimm_provider_data(nvdimm);
-               struct acpi_nfit_memory_map *memdev = memdev_from_spa(acpi_desc,
-                               spa->range_index, i);
-
-               if (!memdev || !nfit_mem->dcr) {
-                       dev_err(dev, "%s: failed to find DCR\n", __func__);
-                       return -ENODEV;
-               }
-
-               map->region_offset = memdev->region_offset;
-               map->serial_number = nfit_mem->dcr->serial_number;
-       }
-
-       sort(&info->mapping[0], nr, sizeof(struct nfit_set_info_map),
-                       cmp_map, NULL);
-       nd_set->cookie = nd_fletcher64(info, sizeof_nfit_set_info(nr), 0);
-       ndr_desc->nd_set = nd_set;
-       devm_kfree(dev, info);
-
-       return 0;
-}
-
-static u64 to_interleave_offset(u64 offset, struct nfit_blk_mmio *mmio)
-{
-       struct acpi_nfit_interleave *idt = mmio->idt;
-       u32 sub_line_offset, line_index, line_offset;
-       u64 line_no, table_skip_count, table_offset;
-
-       line_no = div_u64_rem(offset, mmio->line_size, &sub_line_offset);
-       table_skip_count = div_u64_rem(line_no, mmio->num_lines, &line_index);
-       line_offset = idt->line_offset[line_index]
-               * mmio->line_size;
-       table_offset = table_skip_count * mmio->table_size;
-
-       return mmio->base_offset + line_offset + table_offset + sub_line_offset;
-}
-
-static u32 read_blk_stat(struct nfit_blk *nfit_blk, unsigned int bw)
-{
-       struct nfit_blk_mmio *mmio = &nfit_blk->mmio[DCR];
-       u64 offset = nfit_blk->stat_offset + mmio->size * bw;
-
-       if (mmio->num_lines)
-               offset = to_interleave_offset(offset, mmio);
-
-       return readl(mmio->addr.base + offset);
-}
-
-static void write_blk_ctl(struct nfit_blk *nfit_blk, unsigned int bw,
-               resource_size_t dpa, unsigned int len, unsigned int write)
-{
-       u64 cmd, offset;
-       struct nfit_blk_mmio *mmio = &nfit_blk->mmio[DCR];
-
-       enum {
-               BCW_OFFSET_MASK = (1ULL << 48)-1,
-               BCW_LEN_SHIFT = 48,
-               BCW_LEN_MASK = (1ULL << 8) - 1,
-               BCW_CMD_SHIFT = 56,
-       };
-
-       cmd = (dpa >> L1_CACHE_SHIFT) & BCW_OFFSET_MASK;
-       len = len >> L1_CACHE_SHIFT;
-       cmd |= ((u64) len & BCW_LEN_MASK) << BCW_LEN_SHIFT;
-       cmd |= ((u64) write) << BCW_CMD_SHIFT;
-
-       offset = nfit_blk->cmd_offset + mmio->size * bw;
-       if (mmio->num_lines)
-               offset = to_interleave_offset(offset, mmio);
-
-       writeq(cmd, mmio->addr.base + offset);
-       nvdimm_flush(nfit_blk->nd_region);
-
-       if (nfit_blk->dimm_flags & NFIT_BLK_DCR_LATCH)
-               readq(mmio->addr.base + offset);
-}
-
-static int acpi_nfit_blk_single_io(struct nfit_blk *nfit_blk,
-               resource_size_t dpa, void *iobuf, size_t len, int rw,
-               unsigned int lane)
-{
-       struct nfit_blk_mmio *mmio = &nfit_blk->mmio[BDW];
-       unsigned int copied = 0;
-       u64 base_offset;
-       int rc;
-
-       base_offset = nfit_blk->bdw_offset + dpa % L1_CACHE_BYTES
-               + lane * mmio->size;
-       write_blk_ctl(nfit_blk, lane, dpa, len, rw);
-       while (len) {
-               unsigned int c;
-               u64 offset;
-
-               if (mmio->num_lines) {
-                       u32 line_offset;
-
-                       offset = to_interleave_offset(base_offset + copied,
-                                       mmio);
-                       div_u64_rem(offset, mmio->line_size, &line_offset);
-                       c = min_t(size_t, len, mmio->line_size - line_offset);
-               } else {
-                       offset = base_offset + nfit_blk->bdw_offset;
-                       c = len;
-               }
-
-               if (rw)
-                       memcpy_to_pmem(mmio->addr.aperture + offset,
-                                       iobuf + copied, c);
-               else {
-                       if (nfit_blk->dimm_flags & NFIT_BLK_READ_FLUSH)
-                               mmio_flush_range((void __force *)
-                                       mmio->addr.aperture + offset, c);
-
-                       memcpy_from_pmem(iobuf + copied,
-                                       mmio->addr.aperture + offset, c);
-               }
-
-               copied += c;
-               len -= c;
-       }
-
-       if (rw)
-               nvdimm_flush(nfit_blk->nd_region);
-
-       rc = read_blk_stat(nfit_blk, lane) ? -EIO : 0;
-       return rc;
-}
-
-static int acpi_nfit_blk_region_do_io(struct nd_blk_region *ndbr,
-               resource_size_t dpa, void *iobuf, u64 len, int rw)
-{
-       struct nfit_blk *nfit_blk = nd_blk_region_provider_data(ndbr);
-       struct nfit_blk_mmio *mmio = &nfit_blk->mmio[BDW];
-       struct nd_region *nd_region = nfit_blk->nd_region;
-       unsigned int lane, copied = 0;
-       int rc = 0;
-
-       lane = nd_region_acquire_lane(nd_region);
-       while (len) {
-               u64 c = min(len, mmio->size);
-
-               rc = acpi_nfit_blk_single_io(nfit_blk, dpa + copied,
-                               iobuf + copied, c, rw, lane);
-               if (rc)
-                       break;
-
-               copied += c;
-               len -= c;
-       }
-       nd_region_release_lane(nd_region, lane);
-
-       return rc;
-}
-
-static int nfit_blk_init_interleave(struct nfit_blk_mmio *mmio,
-               struct acpi_nfit_interleave *idt, u16 interleave_ways)
-{
-       if (idt) {
-               mmio->num_lines = idt->line_count;
-               mmio->line_size = idt->line_size;
-               if (interleave_ways == 0)
-                       return -ENXIO;
-               mmio->table_size = mmio->num_lines * interleave_ways
-                       * mmio->line_size;
-       }
-
-       return 0;
-}
-
-static int acpi_nfit_blk_get_flags(struct nvdimm_bus_descriptor *nd_desc,
-               struct nvdimm *nvdimm, struct nfit_blk *nfit_blk)
-{
-       struct nd_cmd_dimm_flags flags;
-       int rc;
-
-       memset(&flags, 0, sizeof(flags));
-       rc = nd_desc->ndctl(nd_desc, nvdimm, ND_CMD_DIMM_FLAGS, &flags,
-                       sizeof(flags), NULL);
-
-       if (rc >= 0 && flags.status == 0)
-               nfit_blk->dimm_flags = flags.flags;
-       else if (rc == -ENOTTY) {
-               /* fall back to a conservative default */
-               nfit_blk->dimm_flags = NFIT_BLK_DCR_LATCH | NFIT_BLK_READ_FLUSH;
-               rc = 0;
-       } else
-               rc = -ENXIO;
-
-       return rc;
-}
-
-static int acpi_nfit_blk_region_enable(struct nvdimm_bus *nvdimm_bus,
-               struct device *dev)
-{
-       struct nvdimm_bus_descriptor *nd_desc = to_nd_desc(nvdimm_bus);
-       struct nd_blk_region *ndbr = to_nd_blk_region(dev);
-       struct nfit_blk_mmio *mmio;
-       struct nfit_blk *nfit_blk;
-       struct nfit_mem *nfit_mem;
-       struct nvdimm *nvdimm;
-       int rc;
-
-       nvdimm = nd_blk_region_to_dimm(ndbr);
-       nfit_mem = nvdimm_provider_data(nvdimm);
-       if (!nfit_mem || !nfit_mem->dcr || !nfit_mem->bdw) {
-               dev_dbg(dev, "%s: missing%s%s%s\n", __func__,
-                               nfit_mem ? "" : " nfit_mem",
-                               (nfit_mem && nfit_mem->dcr) ? "" : " dcr",
-                               (nfit_mem && nfit_mem->bdw) ? "" : " bdw");
-               return -ENXIO;
-       }
-
-       nfit_blk = devm_kzalloc(dev, sizeof(*nfit_blk), GFP_KERNEL);
-       if (!nfit_blk)
-               return -ENOMEM;
-       nd_blk_region_set_provider_data(ndbr, nfit_blk);
-       nfit_blk->nd_region = to_nd_region(dev);
-
-       /* map block aperture memory */
-       nfit_blk->bdw_offset = nfit_mem->bdw->offset;
-       mmio = &nfit_blk->mmio[BDW];
-       mmio->addr.base = devm_nvdimm_memremap(dev, nfit_mem->spa_bdw->address,
-                        nfit_mem->spa_bdw->length, ARCH_MEMREMAP_PMEM);
-       if (!mmio->addr.base) {
-               dev_dbg(dev, "%s: %s failed to map bdw\n", __func__,
-                               nvdimm_name(nvdimm));
-               return -ENOMEM;
-       }
-       mmio->size = nfit_mem->bdw->size;
-       mmio->base_offset = nfit_mem->memdev_bdw->region_offset;
-       mmio->idt = nfit_mem->idt_bdw;
-       mmio->spa = nfit_mem->spa_bdw;
-       rc = nfit_blk_init_interleave(mmio, nfit_mem->idt_bdw,
-                       nfit_mem->memdev_bdw->interleave_ways);
-       if (rc) {
-               dev_dbg(dev, "%s: %s failed to init bdw interleave\n",
-                               __func__, nvdimm_name(nvdimm));
-               return rc;
-       }
-
-       /* map block control memory */
-       nfit_blk->cmd_offset = nfit_mem->dcr->command_offset;
-       nfit_blk->stat_offset = nfit_mem->dcr->status_offset;
-       mmio = &nfit_blk->mmio[DCR];
-       mmio->addr.base = devm_nvdimm_ioremap(dev, nfit_mem->spa_dcr->address,
-                       nfit_mem->spa_dcr->length);
-       if (!mmio->addr.base) {
-               dev_dbg(dev, "%s: %s failed to map dcr\n", __func__,
-                               nvdimm_name(nvdimm));
-               return -ENOMEM;
-       }
-       mmio->size = nfit_mem->dcr->window_size;
-       mmio->base_offset = nfit_mem->memdev_dcr->region_offset;
-       mmio->idt = nfit_mem->idt_dcr;
-       mmio->spa = nfit_mem->spa_dcr;
-       rc = nfit_blk_init_interleave(mmio, nfit_mem->idt_dcr,
-                       nfit_mem->memdev_dcr->interleave_ways);
-       if (rc) {
-               dev_dbg(dev, "%s: %s failed to init dcr interleave\n",
-                               __func__, nvdimm_name(nvdimm));
-               return rc;
-       }
-
-       rc = acpi_nfit_blk_get_flags(nd_desc, nvdimm, nfit_blk);
-       if (rc < 0) {
-               dev_dbg(dev, "%s: %s failed get DIMM flags\n",
-                               __func__, nvdimm_name(nvdimm));
-               return rc;
-       }
-
-       if (nvdimm_has_flush(nfit_blk->nd_region) < 0)
-               dev_warn(dev, "unable to guarantee persistence of writes\n");
-
-       if (mmio->line_size == 0)
-               return 0;
-
-       if ((u32) nfit_blk->cmd_offset % mmio->line_size
-                       + 8 > mmio->line_size) {
-               dev_dbg(dev, "cmd_offset crosses interleave boundary\n");
-               return -ENXIO;
-       } else if ((u32) nfit_blk->stat_offset % mmio->line_size
-                       + 8 > mmio->line_size) {
-               dev_dbg(dev, "stat_offset crosses interleave boundary\n");
-               return -ENXIO;
-       }
-
-       return 0;
-}
-
-static int ars_get_cap(struct acpi_nfit_desc *acpi_desc,
-               struct nd_cmd_ars_cap *cmd, struct nfit_spa *nfit_spa)
-{
-       struct nvdimm_bus_descriptor *nd_desc = &acpi_desc->nd_desc;
-       struct acpi_nfit_system_address *spa = nfit_spa->spa;
-       int cmd_rc, rc;
-
-       cmd->address = spa->address;
-       cmd->length = spa->length;
-       rc = nd_desc->ndctl(nd_desc, NULL, ND_CMD_ARS_CAP, cmd,
-                       sizeof(*cmd), &cmd_rc);
-       if (rc < 0)
-               return rc;
-       return cmd_rc;
-}
-
-static int ars_start(struct acpi_nfit_desc *acpi_desc, struct nfit_spa *nfit_spa)
-{
-       int rc;
-       int cmd_rc;
-       struct nd_cmd_ars_start ars_start;
-       struct acpi_nfit_system_address *spa = nfit_spa->spa;
-       struct nvdimm_bus_descriptor *nd_desc = &acpi_desc->nd_desc;
-
-       memset(&ars_start, 0, sizeof(ars_start));
-       ars_start.address = spa->address;
-       ars_start.length = spa->length;
-       if (nfit_spa_type(spa) == NFIT_SPA_PM)
-               ars_start.type = ND_ARS_PERSISTENT;
-       else if (nfit_spa_type(spa) == NFIT_SPA_VOLATILE)
-               ars_start.type = ND_ARS_VOLATILE;
-       else
-               return -ENOTTY;
-
-       rc = nd_desc->ndctl(nd_desc, NULL, ND_CMD_ARS_START, &ars_start,
-                       sizeof(ars_start), &cmd_rc);
-
-       if (rc < 0)
-               return rc;
-       return cmd_rc;
-}
-
-static int ars_continue(struct acpi_nfit_desc *acpi_desc)
-{
-       int rc, cmd_rc;
-       struct nd_cmd_ars_start ars_start;
-       struct nvdimm_bus_descriptor *nd_desc = &acpi_desc->nd_desc;
-       struct nd_cmd_ars_status *ars_status = acpi_desc->ars_status;
-
-       memset(&ars_start, 0, sizeof(ars_start));
-       ars_start.address = ars_status->restart_address;
-       ars_start.length = ars_status->restart_length;
-       ars_start.type = ars_status->type;
-       rc = nd_desc->ndctl(nd_desc, NULL, ND_CMD_ARS_START, &ars_start,
-                       sizeof(ars_start), &cmd_rc);
-       if (rc < 0)
-               return rc;
-       return cmd_rc;
-}
-
-static int ars_get_status(struct acpi_nfit_desc *acpi_desc)
-{
-       struct nvdimm_bus_descriptor *nd_desc = &acpi_desc->nd_desc;
-       struct nd_cmd_ars_status *ars_status = acpi_desc->ars_status;
-       int rc, cmd_rc;
-
-       rc = nd_desc->ndctl(nd_desc, NULL, ND_CMD_ARS_STATUS, ars_status,
-                       acpi_desc->ars_status_size, &cmd_rc);
-       if (rc < 0)
-               return rc;
-       return cmd_rc;
-}
-
-static int ars_status_process_records(struct nvdimm_bus *nvdimm_bus,
-               struct nd_cmd_ars_status *ars_status)
-{
-       int rc;
-       u32 i;
-
-       for (i = 0; i < ars_status->num_records; i++) {
-               rc = nvdimm_bus_add_poison(nvdimm_bus,
-                               ars_status->records[i].err_address,
-                               ars_status->records[i].length);
-               if (rc)
-                       return rc;
-       }
-
-       return 0;
-}
-
-static void acpi_nfit_remove_resource(void *data)
-{
-       struct resource *res = data;
-
-       remove_resource(res);
-}
-
-static int acpi_nfit_insert_resource(struct acpi_nfit_desc *acpi_desc,
-               struct nd_region_desc *ndr_desc)
-{
-       struct resource *res, *nd_res = ndr_desc->res;
-       int is_pmem, ret;
-
-       /* No operation if the region is already registered as PMEM */
-       is_pmem = region_intersects(nd_res->start, resource_size(nd_res),
-                               IORESOURCE_MEM, IORES_DESC_PERSISTENT_MEMORY);
-       if (is_pmem == REGION_INTERSECTS)
-               return 0;
-
-       res = devm_kzalloc(acpi_desc->dev, sizeof(*res), GFP_KERNEL);
-       if (!res)
-               return -ENOMEM;
-
-       res->name = "Persistent Memory";
-       res->start = nd_res->start;
-       res->end = nd_res->end;
-       res->flags = IORESOURCE_MEM;
-       res->desc = IORES_DESC_PERSISTENT_MEMORY;
-
-       ret = insert_resource(&iomem_resource, res);
-       if (ret)
-               return ret;
-
-       ret = devm_add_action_or_reset(acpi_desc->dev,
-                                       acpi_nfit_remove_resource,
-                                       res);
-       if (ret)
-               return ret;
-
-       return 0;
-}
-
-static int acpi_nfit_init_mapping(struct acpi_nfit_desc *acpi_desc,
-               struct nd_mapping *nd_mapping, struct nd_region_desc *ndr_desc,
-               struct acpi_nfit_memory_map *memdev,
-               struct nfit_spa *nfit_spa)
-{
-       struct nvdimm *nvdimm = acpi_nfit_dimm_by_handle(acpi_desc,
-                       memdev->device_handle);
-       struct acpi_nfit_system_address *spa = nfit_spa->spa;
-       struct nd_blk_region_desc *ndbr_desc;
-       struct nfit_mem *nfit_mem;
-       int blk_valid = 0;
-
-       if (!nvdimm) {
-               dev_err(acpi_desc->dev, "spa%d dimm: %#x not found\n",
-                               spa->range_index, memdev->device_handle);
-               return -ENODEV;
-       }
-
-       nd_mapping->nvdimm = nvdimm;
-       switch (nfit_spa_type(spa)) {
-       case NFIT_SPA_PM:
-       case NFIT_SPA_VOLATILE:
-               nd_mapping->start = memdev->address;
-               nd_mapping->size = memdev->region_size;
-               break;
-       case NFIT_SPA_DCR:
-               nfit_mem = nvdimm_provider_data(nvdimm);
-               if (!nfit_mem || !nfit_mem->bdw) {
-                       dev_dbg(acpi_desc->dev, "spa%d %s missing bdw\n",
-                                       spa->range_index, nvdimm_name(nvdimm));
-               } else {
-                       nd_mapping->size = nfit_mem->bdw->capacity;
-                       nd_mapping->start = nfit_mem->bdw->start_address;
-                       ndr_desc->num_lanes = nfit_mem->bdw->windows;
-                       blk_valid = 1;
-               }
-
-               ndr_desc->nd_mapping = nd_mapping;
-               ndr_desc->num_mappings = blk_valid;
-               ndbr_desc = to_blk_region_desc(ndr_desc);
-               ndbr_desc->enable = acpi_nfit_blk_region_enable;
-               ndbr_desc->do_io = acpi_desc->blk_do_io;
-               nfit_spa->nd_region = nvdimm_blk_region_create(acpi_desc->nvdimm_bus,
-                               ndr_desc);
-               if (!nfit_spa->nd_region)
-                       return -ENOMEM;
-               break;
-       }
-
-       return 0;
-}
-
-static bool nfit_spa_is_virtual(struct acpi_nfit_system_address *spa)
-{
-       return (nfit_spa_type(spa) == NFIT_SPA_VDISK ||
-               nfit_spa_type(spa) == NFIT_SPA_VCD   ||
-               nfit_spa_type(spa) == NFIT_SPA_PDISK ||
-               nfit_spa_type(spa) == NFIT_SPA_PCD);
-}
-
-static int acpi_nfit_register_region(struct acpi_nfit_desc *acpi_desc,
-               struct nfit_spa *nfit_spa)
-{
-       static struct nd_mapping nd_mappings[ND_MAX_MAPPINGS];
-       struct acpi_nfit_system_address *spa = nfit_spa->spa;
-       struct nd_blk_region_desc ndbr_desc;
-       struct nd_region_desc *ndr_desc;
-       struct nfit_memdev *nfit_memdev;
-       struct nvdimm_bus *nvdimm_bus;
-       struct resource res;
-       int count = 0, rc;
-
-       if (nfit_spa->nd_region)
-               return 0;
-
-       if (spa->range_index == 0 && !nfit_spa_is_virtual(spa)) {
-               dev_dbg(acpi_desc->dev, "%s: detected invalid spa index\n",
-                               __func__);
-               return 0;
-       }
-
-       memset(&res, 0, sizeof(res));
-       memset(&nd_mappings, 0, sizeof(nd_mappings));
-       memset(&ndbr_desc, 0, sizeof(ndbr_desc));
-       res.start = spa->address;
-       res.end = res.start + spa->length - 1;
-       ndr_desc = &ndbr_desc.ndr_desc;
-       ndr_desc->res = &res;
-       ndr_desc->provider_data = nfit_spa;
-       ndr_desc->attr_groups = acpi_nfit_region_attribute_groups;
-       if (spa->flags & ACPI_NFIT_PROXIMITY_VALID)
-               ndr_desc->numa_node = acpi_map_pxm_to_online_node(
-                                               spa->proximity_domain);
-       else
-               ndr_desc->numa_node = NUMA_NO_NODE;
-
-       list_for_each_entry(nfit_memdev, &acpi_desc->memdevs, list) {
-               struct acpi_nfit_memory_map *memdev = nfit_memdev->memdev;
-               struct nd_mapping *nd_mapping;
-
-               if (memdev->range_index != spa->range_index)
-                       continue;
-               if (count >= ND_MAX_MAPPINGS) {
-                       dev_err(acpi_desc->dev, "spa%d exceeds max mappings %d\n",
-                                       spa->range_index, ND_MAX_MAPPINGS);
-                       return -ENXIO;
-               }
-               nd_mapping = &nd_mappings[count++];
-               rc = acpi_nfit_init_mapping(acpi_desc, nd_mapping, ndr_desc,
-                               memdev, nfit_spa);
-               if (rc)
-                       goto out;
-       }
-
-       ndr_desc->nd_mapping = nd_mappings;
-       ndr_desc->num_mappings = count;
-       rc = acpi_nfit_init_interleave_set(acpi_desc, ndr_desc, spa);
-       if (rc)
-               goto out;
-
-       nvdimm_bus = acpi_desc->nvdimm_bus;
-       if (nfit_spa_type(spa) == NFIT_SPA_PM) {
-               rc = acpi_nfit_insert_resource(acpi_desc, ndr_desc);
-               if (rc) {
-                       dev_warn(acpi_desc->dev,
-                               "failed to insert pmem resource to iomem: %d\n",
-                               rc);
-                       goto out;
-               }
-
-               nfit_spa->nd_region = nvdimm_pmem_region_create(nvdimm_bus,
-                               ndr_desc);
-               if (!nfit_spa->nd_region)
-                       rc = -ENOMEM;
-       } else if (nfit_spa_type(spa) == NFIT_SPA_VOLATILE) {
-               nfit_spa->nd_region = nvdimm_volatile_region_create(nvdimm_bus,
-                               ndr_desc);
-               if (!nfit_spa->nd_region)
-                       rc = -ENOMEM;
-       } else if (nfit_spa_is_virtual(spa)) {
-               nfit_spa->nd_region = nvdimm_pmem_region_create(nvdimm_bus,
-                               ndr_desc);
-               if (!nfit_spa->nd_region)
-                       rc = -ENOMEM;
-       }
-
- out:
-       if (rc)
-               dev_err(acpi_desc->dev, "failed to register spa range %d\n",
-                               nfit_spa->spa->range_index);
-       return rc;
-}
-
-static int ars_status_alloc(struct acpi_nfit_desc *acpi_desc,
-               u32 max_ars)
-{
-       struct device *dev = acpi_desc->dev;
-       struct nd_cmd_ars_status *ars_status;
-
-       if (acpi_desc->ars_status && acpi_desc->ars_status_size >= max_ars) {
-               memset(acpi_desc->ars_status, 0, acpi_desc->ars_status_size);
-               return 0;
-       }
-
-       if (acpi_desc->ars_status)
-               devm_kfree(dev, acpi_desc->ars_status);
-       acpi_desc->ars_status = NULL;
-       ars_status = devm_kzalloc(dev, max_ars, GFP_KERNEL);
-       if (!ars_status)
-               return -ENOMEM;
-       acpi_desc->ars_status = ars_status;
-       acpi_desc->ars_status_size = max_ars;
-       return 0;
-}
-
-static int acpi_nfit_query_poison(struct acpi_nfit_desc *acpi_desc,
-               struct nfit_spa *nfit_spa)
-{
-       struct acpi_nfit_system_address *spa = nfit_spa->spa;
-       int rc;
-
-       if (!nfit_spa->max_ars) {
-               struct nd_cmd_ars_cap ars_cap;
-
-               memset(&ars_cap, 0, sizeof(ars_cap));
-               rc = ars_get_cap(acpi_desc, &ars_cap, nfit_spa);
-               if (rc < 0)
-                       return rc;
-               nfit_spa->max_ars = ars_cap.max_ars_out;
-               nfit_spa->clear_err_unit = ars_cap.clear_err_unit;
-               /* check that the supported scrub types match the spa type */
-               if (nfit_spa_type(spa) == NFIT_SPA_VOLATILE &&
-                               ((ars_cap.status >> 16) & ND_ARS_VOLATILE) == 0)
-                       return -ENOTTY;
-               else if (nfit_spa_type(spa) == NFIT_SPA_PM &&
-                               ((ars_cap.status >> 16) & ND_ARS_PERSISTENT) == 0)
-                       return -ENOTTY;
-       }
-
-       if (ars_status_alloc(acpi_desc, nfit_spa->max_ars))
-               return -ENOMEM;
-
-       rc = ars_get_status(acpi_desc);
-       if (rc < 0 && rc != -ENOSPC)
-               return rc;
-
-       if (ars_status_process_records(acpi_desc->nvdimm_bus,
-                               acpi_desc->ars_status))
-               return -ENOMEM;
-
-       return 0;
-}
-
-static void acpi_nfit_async_scrub(struct acpi_nfit_desc *acpi_desc,
-               struct nfit_spa *nfit_spa)
-{
-       struct acpi_nfit_system_address *spa = nfit_spa->spa;
-       unsigned int overflow_retry = scrub_overflow_abort;
-       u64 init_ars_start = 0, init_ars_len = 0;
-       struct device *dev = acpi_desc->dev;
-       unsigned int tmo = scrub_timeout;
-       int rc;
-
-       if (!nfit_spa->ars_required || !nfit_spa->nd_region)
-               return;
-
-       rc = ars_start(acpi_desc, nfit_spa);
-       /*
-        * If we timed out the initial scan we'll still be busy here,
-        * and will wait another timeout before giving up permanently.
-        */
-       if (rc < 0 && rc != -EBUSY)
-               return;
-
-       do {
-               u64 ars_start, ars_len;
-
-               if (acpi_desc->cancel)
-                       break;
-               rc = acpi_nfit_query_poison(acpi_desc, nfit_spa);
-               if (rc == -ENOTTY)
-                       break;
-               if (rc == -EBUSY && !tmo) {
-                       dev_warn(dev, "range %d ars timeout, aborting\n",
-                                       spa->range_index);
-                       break;
-               }
-
-               if (rc == -EBUSY) {
-                       /*
-                        * Note, entries may be appended to the list
-                        * while the lock is dropped, but the workqueue
-                        * being active prevents entries being deleted /
-                        * freed.
-                        */
-                       mutex_unlock(&acpi_desc->init_mutex);
-                       ssleep(1);
-                       tmo--;
-                       mutex_lock(&acpi_desc->init_mutex);
-                       continue;
-               }
-
-               /* we got some results, but there are more pending... */
-               if (rc == -ENOSPC && overflow_retry--) {
-                       if (!init_ars_len) {
-                               init_ars_len = acpi_desc->ars_status->length;
-                               init_ars_start = acpi_desc->ars_status->address;
-                       }
-                       rc = ars_continue(acpi_desc);
-               }
-
-               if (rc < 0) {
-                       dev_warn(dev, "range %d ars continuation failed\n",
-                                       spa->range_index);
-                       break;
-               }
-
-               if (init_ars_len) {
-                       ars_start = init_ars_start;
-                       ars_len = init_ars_len;
-               } else {
-                       ars_start = acpi_desc->ars_status->address;
-                       ars_len = acpi_desc->ars_status->length;
-               }
-               dev_dbg(dev, "spa range: %d ars from %#llx + %#llx complete\n",
-                               spa->range_index, ars_start, ars_len);
-               /* notify the region about new poison entries */
-               nvdimm_region_notify(nfit_spa->nd_region,
-                               NVDIMM_REVALIDATE_POISON);
-               break;
-       } while (1);
-}
-
-static void acpi_nfit_scrub(struct work_struct *work)
-{
-       struct device *dev;
-       u64 init_scrub_length = 0;
-       struct nfit_spa *nfit_spa;
-       u64 init_scrub_address = 0;
-       bool init_ars_done = false;
-       struct acpi_nfit_desc *acpi_desc;
-       unsigned int tmo = scrub_timeout;
-       unsigned int overflow_retry = scrub_overflow_abort;
-
-       acpi_desc = container_of(work, typeof(*acpi_desc), work);
-       dev = acpi_desc->dev;
-
-       /*
-        * We scrub in 2 phases.  The first phase waits for any platform
-        * firmware initiated scrubs to complete and then we go search for the
-        * affected spa regions to mark them scanned.  In the second phase we
-        * initiate a directed scrub for every range that was not scrubbed in
-        * phase 1. If we're called for a 'rescan', we harmlessly pass through
-        * the first phase, but really only care about running phase 2, where
-        * regions can be notified of new poison.
-        */
-
-       /* process platform firmware initiated scrubs */
- retry:
-       mutex_lock(&acpi_desc->init_mutex);
-       list_for_each_entry(nfit_spa, &acpi_desc->spas, list) {
-               struct nd_cmd_ars_status *ars_status;
-               struct acpi_nfit_system_address *spa;
-               u64 ars_start, ars_len;
-               int rc;
-
-               if (acpi_desc->cancel)
-                       break;
-
-               if (nfit_spa->nd_region)
-                       continue;
-
-               if (init_ars_done) {
-                       /*
-                        * No need to re-query, we're now just
-                        * reconciling all the ranges covered by the
-                        * initial scrub
-                        */
-                       rc = 0;
-               } else
-                       rc = acpi_nfit_query_poison(acpi_desc, nfit_spa);
-
-               if (rc == -ENOTTY) {
-                       /* no ars capability, just register spa and move on */
-                       acpi_nfit_register_region(acpi_desc, nfit_spa);
-                       continue;
-               }
-
-               if (rc == -EBUSY && !tmo) {
-                       /* fallthrough to directed scrub in phase 2 */
-                       dev_warn(dev, "timeout awaiting ars results, continuing...\n");
-                       break;
-               } else if (rc == -EBUSY) {
-                       mutex_unlock(&acpi_desc->init_mutex);
-                       ssleep(1);
-                       tmo--;
-                       goto retry;
-               }
-
-               /* we got some results, but there are more pending... */
-               if (rc == -ENOSPC && overflow_retry--) {
-                       ars_status = acpi_desc->ars_status;
-                       /*
-                        * Record the original scrub range, so that we
-                        * can recall all the ranges impacted by the
-                        * initial scrub.
-                        */
-                       if (!init_scrub_length) {
-                               init_scrub_length = ars_status->length;
-                               init_scrub_address = ars_status->address;
-                       }
-                       rc = ars_continue(acpi_desc);
-                       if (rc == 0) {
-                               mutex_unlock(&acpi_desc->init_mutex);
-                               goto retry;
-                       }
-               }
-
-               if (rc < 0) {
-                       /*
-                        * Initial scrub failed, we'll give it one more
-                        * try below...
-                        */
-                       break;
-               }
-
-               /* We got some final results, record completed ranges */
-               ars_status = acpi_desc->ars_status;
-               if (init_scrub_length) {
-                       ars_start = init_scrub_address;
-                       ars_len = ars_start + init_scrub_length;
-               } else {
-                       ars_start = ars_status->address;
-                       ars_len = ars_status->length;
-               }
-               spa = nfit_spa->spa;
-
-               if (!init_ars_done) {
-                       init_ars_done = true;
-                       dev_dbg(dev, "init scrub %#llx + %#llx complete\n",
-                                       ars_start, ars_len);
-               }
-               if (ars_start <= spa->address && ars_start + ars_len
-                               >= spa->address + spa->length)
-                       acpi_nfit_register_region(acpi_desc, nfit_spa);
-       }
-
-       /*
-        * For all the ranges not covered by an initial scrub we still
-        * want to see if there are errors, but it's ok to discover them
-        * asynchronously.
-        */
-       list_for_each_entry(nfit_spa, &acpi_desc->spas, list) {
-               /*
-                * Flag all the ranges that still need scrubbing, but
-                * register them now to make data available.
-                */
-               if (!nfit_spa->nd_region) {
-                       nfit_spa->ars_required = 1;
-                       acpi_nfit_register_region(acpi_desc, nfit_spa);
-               }
-       }
-
-       list_for_each_entry(nfit_spa, &acpi_desc->spas, list)
-               acpi_nfit_async_scrub(acpi_desc, nfit_spa);
-       acpi_desc->scrub_count++;
-       if (acpi_desc->scrub_count_state)
-               sysfs_notify_dirent(acpi_desc->scrub_count_state);
-       mutex_unlock(&acpi_desc->init_mutex);
-}
-
-static int acpi_nfit_register_regions(struct acpi_nfit_desc *acpi_desc)
-{
-       struct nfit_spa *nfit_spa;
-       int rc;
-
-       list_for_each_entry(nfit_spa, &acpi_desc->spas, list)
-               if (nfit_spa_type(nfit_spa->spa) == NFIT_SPA_DCR) {
-                       /* BLK regions don't need to wait for ars results */
-                       rc = acpi_nfit_register_region(acpi_desc, nfit_spa);
-                       if (rc)
-                               return rc;
-               }
-
-       queue_work(nfit_wq, &acpi_desc->work);
-       return 0;
-}
-
-static int acpi_nfit_check_deletions(struct acpi_nfit_desc *acpi_desc,
-               struct nfit_table_prev *prev)
-{
-       struct device *dev = acpi_desc->dev;
-
-       if (!list_empty(&prev->spas) ||
-                       !list_empty(&prev->memdevs) ||
-                       !list_empty(&prev->dcrs) ||
-                       !list_empty(&prev->bdws) ||
-                       !list_empty(&prev->idts) ||
-                       !list_empty(&prev->flushes)) {
-               dev_err(dev, "new nfit deletes entries (unsupported)\n");
-               return -ENXIO;
-       }
-       return 0;
-}
-
-static int acpi_nfit_desc_init_scrub_attr(struct acpi_nfit_desc *acpi_desc)
-{
-       struct device *dev = acpi_desc->dev;
-       struct kernfs_node *nfit;
-       struct device *bus_dev;
-
-       if (!ars_supported(acpi_desc->nvdimm_bus))
-               return 0;
-
-       bus_dev = to_nvdimm_bus_dev(acpi_desc->nvdimm_bus);
-       nfit = sysfs_get_dirent(bus_dev->kobj.sd, "nfit");
-       if (!nfit) {
-               dev_err(dev, "sysfs_get_dirent 'nfit' failed\n");
-               return -ENODEV;
-       }
-       acpi_desc->scrub_count_state = sysfs_get_dirent(nfit, "scrub");
-       sysfs_put(nfit);
-       if (!acpi_desc->scrub_count_state) {
-               dev_err(dev, "sysfs_get_dirent 'scrub' failed\n");
-               return -ENODEV;
-       }
-
-       return 0;
-}
-
-static void acpi_nfit_destruct(void *data)
-{
-       struct acpi_nfit_desc *acpi_desc = data;
-       struct device *bus_dev = to_nvdimm_bus_dev(acpi_desc->nvdimm_bus);
-
-       acpi_desc->cancel = 1;
-       /*
-        * Bounce the nvdimm bus lock to make sure any in-flight
-        * acpi_nfit_ars_rescan() submissions have had a chance to
-        * either submit or see ->cancel set.
-        */
-       device_lock(bus_dev);
-       device_unlock(bus_dev);
-
-       flush_workqueue(nfit_wq);
-       if (acpi_desc->scrub_count_state)
-               sysfs_put(acpi_desc->scrub_count_state);
-       nvdimm_bus_unregister(acpi_desc->nvdimm_bus);
-       acpi_desc->nvdimm_bus = NULL;
-}
-
-int acpi_nfit_init(struct acpi_nfit_desc *acpi_desc, void *data, acpi_size sz)
-{
-       struct device *dev = acpi_desc->dev;
-       struct nfit_table_prev prev;
-       const void *end;
-       int rc;
-
-       if (!acpi_desc->nvdimm_bus) {
-               acpi_nfit_init_dsms(acpi_desc);
-
-               acpi_desc->nvdimm_bus = nvdimm_bus_register(dev,
-                               &acpi_desc->nd_desc);
-               if (!acpi_desc->nvdimm_bus)
-                       return -ENOMEM;
-
-               rc = devm_add_action_or_reset(dev, acpi_nfit_destruct,
-                               acpi_desc);
-               if (rc)
-                       return rc;
-
-               rc = acpi_nfit_desc_init_scrub_attr(acpi_desc);
-               if (rc)
-                       return rc;
-       }
-
-       mutex_lock(&acpi_desc->init_mutex);
-
-       INIT_LIST_HEAD(&prev.spas);
-       INIT_LIST_HEAD(&prev.memdevs);
-       INIT_LIST_HEAD(&prev.dcrs);
-       INIT_LIST_HEAD(&prev.bdws);
-       INIT_LIST_HEAD(&prev.idts);
-       INIT_LIST_HEAD(&prev.flushes);
-
-       list_cut_position(&prev.spas, &acpi_desc->spas,
-                               acpi_desc->spas.prev);
-       list_cut_position(&prev.memdevs, &acpi_desc->memdevs,
-                               acpi_desc->memdevs.prev);
-       list_cut_position(&prev.dcrs, &acpi_desc->dcrs,
-                               acpi_desc->dcrs.prev);
-       list_cut_position(&prev.bdws, &acpi_desc->bdws,
-                               acpi_desc->bdws.prev);
-       list_cut_position(&prev.idts, &acpi_desc->idts,
-                               acpi_desc->idts.prev);
-       list_cut_position(&prev.flushes, &acpi_desc->flushes,
-                               acpi_desc->flushes.prev);
-
-       end = data + sz;
-       while (!IS_ERR_OR_NULL(data))
-               data = add_table(acpi_desc, &prev, data, end);
-
-       if (IS_ERR(data)) {
-               dev_dbg(dev, "%s: nfit table parsing error: %ld\n", __func__,
-                               PTR_ERR(data));
-               rc = PTR_ERR(data);
-               goto out_unlock;
-       }
-
-       rc = acpi_nfit_check_deletions(acpi_desc, &prev);
-       if (rc)
-               goto out_unlock;
-
-       rc = nfit_mem_init(acpi_desc);
-       if (rc)
-               goto out_unlock;
-
-       rc = acpi_nfit_register_dimms(acpi_desc);
-       if (rc)
-               goto out_unlock;
-
-       rc = acpi_nfit_register_regions(acpi_desc);
-
- out_unlock:
-       mutex_unlock(&acpi_desc->init_mutex);
-       return rc;
-}
-EXPORT_SYMBOL_GPL(acpi_nfit_init);
-
-struct acpi_nfit_flush_work {
-       struct work_struct work;
-       struct completion cmp;
-};
-
-static void flush_probe(struct work_struct *work)
-{
-       struct acpi_nfit_flush_work *flush;
-
-       flush = container_of(work, typeof(*flush), work);
-       complete(&flush->cmp);
-}
-
-static int acpi_nfit_flush_probe(struct nvdimm_bus_descriptor *nd_desc)
-{
-       struct acpi_nfit_desc *acpi_desc = to_acpi_nfit_desc(nd_desc);
-       struct device *dev = acpi_desc->dev;
-       struct acpi_nfit_flush_work flush;
-
-       /* bounce the device lock to flush acpi_nfit_add / acpi_nfit_notify */
-       device_lock(dev);
-       device_unlock(dev);
-
-       /*
-        * Scrub work could take 10s of seconds, userspace may give up so we
-        * need to be interruptible while waiting.
-        */
-       INIT_WORK_ONSTACK(&flush.work, flush_probe);
-       COMPLETION_INITIALIZER_ONSTACK(flush.cmp);
-       queue_work(nfit_wq, &flush.work);
-       return wait_for_completion_interruptible(&flush.cmp);
-}
-
-static int acpi_nfit_clear_to_send(struct nvdimm_bus_descriptor *nd_desc,
-               struct nvdimm *nvdimm, unsigned int cmd)
-{
-       struct acpi_nfit_desc *acpi_desc = to_acpi_nfit_desc(nd_desc);
-
-       if (nvdimm)
-               return 0;
-       if (cmd != ND_CMD_ARS_START)
-               return 0;
-
-       /*
-        * The kernel and userspace may race to initiate a scrub, but
-        * the scrub thread is prepared to lose that initial race.  It
-        * just needs guarantees that any ars it initiates are not
-        * interrupted by any intervening start reqeusts from userspace.
-        */
-       if (work_busy(&acpi_desc->work))
-               return -EBUSY;
-
-       return 0;
-}
-
-static int acpi_nfit_ars_rescan(struct acpi_nfit_desc *acpi_desc)
-{
-       struct device *dev = acpi_desc->dev;
-       struct nfit_spa *nfit_spa;
-
-       if (work_busy(&acpi_desc->work))
-               return -EBUSY;
-
-       if (acpi_desc->cancel)
-               return 0;
-
-       mutex_lock(&acpi_desc->init_mutex);
-       list_for_each_entry(nfit_spa, &acpi_desc->spas, list) {
-               struct acpi_nfit_system_address *spa = nfit_spa->spa;
-
-               if (nfit_spa_type(spa) != NFIT_SPA_PM)
-                       continue;
-
-               nfit_spa->ars_required = 1;
-       }
-       queue_work(nfit_wq, &acpi_desc->work);
-       dev_dbg(dev, "%s: ars_scan triggered\n", __func__);
-       mutex_unlock(&acpi_desc->init_mutex);
-
-       return 0;
-}
-
-void acpi_nfit_desc_init(struct acpi_nfit_desc *acpi_desc, struct device *dev)
-{
-       struct nvdimm_bus_descriptor *nd_desc;
-
-       dev_set_drvdata(dev, acpi_desc);
-       acpi_desc->dev = dev;
-       acpi_desc->blk_do_io = acpi_nfit_blk_region_do_io;
-       nd_desc = &acpi_desc->nd_desc;
-       nd_desc->provider_name = "ACPI.NFIT";
-       nd_desc->module = THIS_MODULE;
-       nd_desc->ndctl = acpi_nfit_ctl;
-       nd_desc->flush_probe = acpi_nfit_flush_probe;
-       nd_desc->clear_to_send = acpi_nfit_clear_to_send;
-       nd_desc->attr_groups = acpi_nfit_attribute_groups;
-
-       INIT_LIST_HEAD(&acpi_desc->spas);
-       INIT_LIST_HEAD(&acpi_desc->dcrs);
-       INIT_LIST_HEAD(&acpi_desc->bdws);
-       INIT_LIST_HEAD(&acpi_desc->idts);
-       INIT_LIST_HEAD(&acpi_desc->flushes);
-       INIT_LIST_HEAD(&acpi_desc->memdevs);
-       INIT_LIST_HEAD(&acpi_desc->dimms);
-       mutex_init(&acpi_desc->init_mutex);
-       INIT_WORK(&acpi_desc->work, acpi_nfit_scrub);
-}
-EXPORT_SYMBOL_GPL(acpi_nfit_desc_init);
-
-static int acpi_nfit_add(struct acpi_device *adev)
-{
-       struct acpi_buffer buf = { ACPI_ALLOCATE_BUFFER, NULL };
-       struct acpi_nfit_desc *acpi_desc;
-       struct device *dev = &adev->dev;
-       struct acpi_table_header *tbl;
-       acpi_status status = AE_OK;
-       acpi_size sz;
-       int rc = 0;
-
-       status = acpi_get_table_with_size(ACPI_SIG_NFIT, 0, &tbl, &sz);
-       if (ACPI_FAILURE(status)) {
-               /* This is ok, we could have an nvdimm hotplugged later */
-               dev_dbg(dev, "failed to find NFIT at startup\n");
-               return 0;
-       }
-
-       acpi_desc = devm_kzalloc(dev, sizeof(*acpi_desc), GFP_KERNEL);
-       if (!acpi_desc)
-               return -ENOMEM;
-       acpi_nfit_desc_init(acpi_desc, &adev->dev);
-
-       /* Save the acpi header for exporting the revision via sysfs */
-       acpi_desc->acpi_header = *tbl;
-
-       /* Evaluate _FIT and override with that if present */
-       status = acpi_evaluate_object(adev->handle, "_FIT", NULL, &buf);
-       if (ACPI_SUCCESS(status) && buf.length > 0) {
-               union acpi_object *obj = buf.pointer;
-
-               if (obj->type == ACPI_TYPE_BUFFER)
-                       rc = acpi_nfit_init(acpi_desc, obj->buffer.pointer,
-                                       obj->buffer.length);
-               else
-                       dev_dbg(dev, "%s invalid type %d, ignoring _FIT\n",
-                                __func__, (int) obj->type);
-               kfree(buf.pointer);
-       } else
-               /* skip over the lead-in header table */
-               rc = acpi_nfit_init(acpi_desc, (void *) tbl
-                               + sizeof(struct acpi_table_nfit),
-                               sz - sizeof(struct acpi_table_nfit));
-       return rc;
-}
-
-static int acpi_nfit_remove(struct acpi_device *adev)
-{
-       /* see acpi_nfit_destruct */
-       return 0;
-}
-
-static void acpi_nfit_notify(struct acpi_device *adev, u32 event)
-{
-       struct acpi_nfit_desc *acpi_desc = dev_get_drvdata(&adev->dev);
-       struct acpi_buffer buf = { ACPI_ALLOCATE_BUFFER, NULL };
-       struct device *dev = &adev->dev;
-       union acpi_object *obj;
-       acpi_status status;
-       int ret;
-
-       dev_dbg(dev, "%s: event: %d\n", __func__, event);
-
-       device_lock(dev);
-       if (!dev->driver) {
-               /* dev->driver may be null if we're being removed */
-               dev_dbg(dev, "%s: no driver found for dev\n", __func__);
-               goto out_unlock;
-       }
-
-       if (!acpi_desc) {
-               acpi_desc = devm_kzalloc(dev, sizeof(*acpi_desc), GFP_KERNEL);
-               if (!acpi_desc)
-                       goto out_unlock;
-               acpi_nfit_desc_init(acpi_desc, &adev->dev);
-       } else {
-               /*
-                * Finish previous registration before considering new
-                * regions.
-                */
-               flush_workqueue(nfit_wq);
-       }
-
-       /* Evaluate _FIT */
-       status = acpi_evaluate_object(adev->handle, "_FIT", NULL, &buf);
-       if (ACPI_FAILURE(status)) {
-               dev_err(dev, "failed to evaluate _FIT\n");
-               goto out_unlock;
-       }
-
-       obj = buf.pointer;
-       if (obj->type == ACPI_TYPE_BUFFER) {
-               ret = acpi_nfit_init(acpi_desc, obj->buffer.pointer,
-                               obj->buffer.length);
-               if (ret)
-                       dev_err(dev, "failed to merge updated NFIT\n");
-       } else
-               dev_err(dev, "Invalid _FIT\n");
-       kfree(buf.pointer);
-
- out_unlock:
-       device_unlock(dev);
-}
-
-static const struct acpi_device_id acpi_nfit_ids[] = {
-       { "ACPI0012", 0 },
-       { "", 0 },
-};
-MODULE_DEVICE_TABLE(acpi, acpi_nfit_ids);
-
-static struct acpi_driver acpi_nfit_driver = {
-       .name = KBUILD_MODNAME,
-       .ids = acpi_nfit_ids,
-       .ops = {
-               .add = acpi_nfit_add,
-               .remove = acpi_nfit_remove,
-               .notify = acpi_nfit_notify,
-       },
-};
-
-static __init int nfit_init(void)
-{
-       BUILD_BUG_ON(sizeof(struct acpi_table_nfit) != 40);
-       BUILD_BUG_ON(sizeof(struct acpi_nfit_system_address) != 56);
-       BUILD_BUG_ON(sizeof(struct acpi_nfit_memory_map) != 48);
-       BUILD_BUG_ON(sizeof(struct acpi_nfit_interleave) != 20);
-       BUILD_BUG_ON(sizeof(struct acpi_nfit_smbios) != 9);
-       BUILD_BUG_ON(sizeof(struct acpi_nfit_control_region) != 80);
-       BUILD_BUG_ON(sizeof(struct acpi_nfit_data_region) != 40);
-
-       acpi_str_to_uuid(UUID_VOLATILE_MEMORY, nfit_uuid[NFIT_SPA_VOLATILE]);
-       acpi_str_to_uuid(UUID_PERSISTENT_MEMORY, nfit_uuid[NFIT_SPA_PM]);
-       acpi_str_to_uuid(UUID_CONTROL_REGION, nfit_uuid[NFIT_SPA_DCR]);
-       acpi_str_to_uuid(UUID_DATA_REGION, nfit_uuid[NFIT_SPA_BDW]);
-       acpi_str_to_uuid(UUID_VOLATILE_VIRTUAL_DISK, nfit_uuid[NFIT_SPA_VDISK]);
-       acpi_str_to_uuid(UUID_VOLATILE_VIRTUAL_CD, nfit_uuid[NFIT_SPA_VCD]);
-       acpi_str_to_uuid(UUID_PERSISTENT_VIRTUAL_DISK, nfit_uuid[NFIT_SPA_PDISK]);
-       acpi_str_to_uuid(UUID_PERSISTENT_VIRTUAL_CD, nfit_uuid[NFIT_SPA_PCD]);
-       acpi_str_to_uuid(UUID_NFIT_BUS, nfit_uuid[NFIT_DEV_BUS]);
-       acpi_str_to_uuid(UUID_NFIT_DIMM, nfit_uuid[NFIT_DEV_DIMM]);
-       acpi_str_to_uuid(UUID_NFIT_DIMM_N_HPE1, nfit_uuid[NFIT_DEV_DIMM_N_HPE1]);
-       acpi_str_to_uuid(UUID_NFIT_DIMM_N_HPE2, nfit_uuid[NFIT_DEV_DIMM_N_HPE2]);
-       acpi_str_to_uuid(UUID_NFIT_DIMM_N_MSFT, nfit_uuid[NFIT_DEV_DIMM_N_MSFT]);
-
-       nfit_wq = create_singlethread_workqueue("nfit");
-       if (!nfit_wq)
-               return -ENOMEM;
-
-       return acpi_bus_register_driver(&acpi_nfit_driver);
-}
-
-static __exit void nfit_exit(void)
-{
-       acpi_bus_unregister_driver(&acpi_nfit_driver);
-       destroy_workqueue(nfit_wq);
-}
-
-module_init(nfit_init);
-module_exit(nfit_exit);
-MODULE_LICENSE("GPL v2");
-MODULE_AUTHOR("Intel Corporation");
diff --git a/drivers/acpi/nfit.h b/drivers/acpi/nfit.h
deleted file mode 100644 (file)
index 6ecf337..0000000
+++ /dev/null
@@ -1,207 +0,0 @@
-/*
- * NVDIMM Firmware Interface Table - NFIT
- *
- * Copyright(c) 2013-2015 Intel Corporation. All rights reserved.
- *
- * This program is free software; you can redistribute it and/or modify
- * it under the terms of version 2 of the GNU General Public License as
- * published by the Free Software Foundation.
- *
- * This program is distributed in the hope that 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.
- */
-#ifndef __NFIT_H__
-#define __NFIT_H__
-#include <linux/workqueue.h>
-#include <linux/libnvdimm.h>
-#include <linux/types.h>
-#include <linux/uuid.h>
-#include <linux/acpi.h>
-#include <acpi/acuuid.h>
-
-/* ACPI 6.1 */
-#define UUID_NFIT_BUS "2f10e7a4-9e91-11e4-89d3-123b93f75cba"
-
-/* http://pmem.io/documents/NVDIMM_DSM_Interface_Example.pdf */
-#define UUID_NFIT_DIMM "4309ac30-0d11-11e4-9191-0800200c9a66"
-
-/* https://github.com/HewlettPackard/hpe-nvm/blob/master/Documentation/ */
-#define UUID_NFIT_DIMM_N_HPE1 "9002c334-acf3-4c0e-9642-a235f0d53bc6"
-#define UUID_NFIT_DIMM_N_HPE2 "5008664b-b758-41a0-a03c-27c2f2d04f7e"
-
-/* https://msdn.microsoft.com/library/windows/hardware/mt604741 */
-#define UUID_NFIT_DIMM_N_MSFT "1ee68b36-d4bd-4a1a-9a16-4f8e53d46e05"
-
-#define ACPI_NFIT_MEM_FAILED_MASK (ACPI_NFIT_MEM_SAVE_FAILED \
-               | ACPI_NFIT_MEM_RESTORE_FAILED | ACPI_NFIT_MEM_FLUSH_FAILED \
-               | ACPI_NFIT_MEM_NOT_ARMED)
-
-enum nfit_uuids {
-       /* for simplicity alias the uuid index with the family id */
-       NFIT_DEV_DIMM = NVDIMM_FAMILY_INTEL,
-       NFIT_DEV_DIMM_N_HPE1 = NVDIMM_FAMILY_HPE1,
-       NFIT_DEV_DIMM_N_HPE2 = NVDIMM_FAMILY_HPE2,
-       NFIT_DEV_DIMM_N_MSFT = NVDIMM_FAMILY_MSFT,
-       NFIT_SPA_VOLATILE,
-       NFIT_SPA_PM,
-       NFIT_SPA_DCR,
-       NFIT_SPA_BDW,
-       NFIT_SPA_VDISK,
-       NFIT_SPA_VCD,
-       NFIT_SPA_PDISK,
-       NFIT_SPA_PCD,
-       NFIT_DEV_BUS,
-       NFIT_UUID_MAX,
-};
-
-/*
- * Region format interface codes are stored as an array of bytes in the
- * NFIT DIMM Control Region structure
- */
-#define NFIT_FIC_BYTE cpu_to_be16(0x101) /* byte-addressable energy backed */
-#define NFIT_FIC_BLK cpu_to_be16(0x201) /* block-addressable non-energy backed */
-#define NFIT_FIC_BYTEN cpu_to_be16(0x301) /* byte-addressable non-energy backed */
-
-enum {
-       NFIT_BLK_READ_FLUSH = 1,
-       NFIT_BLK_DCR_LATCH = 2,
-       NFIT_ARS_STATUS_DONE = 0,
-       NFIT_ARS_STATUS_BUSY = 1 << 16,
-       NFIT_ARS_STATUS_NONE = 2 << 16,
-       NFIT_ARS_STATUS_INTR = 3 << 16,
-       NFIT_ARS_START_BUSY = 6,
-       NFIT_ARS_CAP_NONE = 1,
-       NFIT_ARS_F_OVERFLOW = 1,
-       NFIT_ARS_TIMEOUT = 90,
-};
-
-struct nfit_spa {
-       struct list_head list;
-       struct nd_region *nd_region;
-       unsigned int ars_required:1;
-       u32 clear_err_unit;
-       u32 max_ars;
-       struct acpi_nfit_system_address spa[0];
-};
-
-struct nfit_dcr {
-       struct list_head list;
-       struct acpi_nfit_control_region dcr[0];
-};
-
-struct nfit_bdw {
-       struct list_head list;
-       struct acpi_nfit_data_region bdw[0];
-};
-
-struct nfit_idt {
-       struct list_head list;
-       struct acpi_nfit_interleave idt[0];
-};
-
-struct nfit_flush {
-       struct list_head list;
-       struct acpi_nfit_flush_address flush[0];
-};
-
-struct nfit_memdev {
-       struct list_head list;
-       struct acpi_nfit_memory_map memdev[0];
-};
-
-/* assembled tables for a given dimm/memory-device */
-struct nfit_mem {
-       struct nvdimm *nvdimm;
-       struct acpi_nfit_memory_map *memdev_dcr;
-       struct acpi_nfit_memory_map *memdev_pmem;
-       struct acpi_nfit_memory_map *memdev_bdw;
-       struct acpi_nfit_control_region *dcr;
-       struct acpi_nfit_data_region *bdw;
-       struct acpi_nfit_system_address *spa_dcr;
-       struct acpi_nfit_system_address *spa_bdw;
-       struct acpi_nfit_interleave *idt_dcr;
-       struct acpi_nfit_interleave *idt_bdw;
-       struct nfit_flush *nfit_flush;
-       struct list_head list;
-       struct acpi_device *adev;
-       struct acpi_nfit_desc *acpi_desc;
-       struct resource *flush_wpq;
-       unsigned long dsm_mask;
-       int family;
-};
-
-struct acpi_nfit_desc {
-       struct nvdimm_bus_descriptor nd_desc;
-       struct acpi_table_header acpi_header;
-       struct mutex init_mutex;
-       struct list_head memdevs;
-       struct list_head flushes;
-       struct list_head dimms;
-       struct list_head spas;
-       struct list_head dcrs;
-       struct list_head bdws;
-       struct list_head idts;
-       struct nvdimm_bus *nvdimm_bus;
-       struct device *dev;
-       struct nd_cmd_ars_status *ars_status;
-       size_t ars_status_size;
-       struct work_struct work;
-       struct kernfs_node *scrub_count_state;
-       unsigned int scrub_count;
-       unsigned int cancel:1;
-       unsigned long dimm_cmd_force_en;
-       unsigned long bus_cmd_force_en;
-       int (*blk_do_io)(struct nd_blk_region *ndbr, resource_size_t dpa,
-                       void *iobuf, u64 len, int rw);
-};
-
-enum nd_blk_mmio_selector {
-       BDW,
-       DCR,
-};
-
-struct nd_blk_addr {
-       union {
-               void __iomem *base;
-               void *aperture;
-       };
-};
-
-struct nfit_blk {
-       struct nfit_blk_mmio {
-               struct nd_blk_addr addr;
-               u64 size;
-               u64 base_offset;
-               u32 line_size;
-               u32 num_lines;
-               u32 table_size;
-               struct acpi_nfit_interleave *idt;
-               struct acpi_nfit_system_address *spa;
-       } mmio[2];
-       struct nd_region *nd_region;
-       u64 bdw_offset; /* post interleave offset */
-       u64 stat_offset;
-       u64 cmd_offset;
-       u32 dimm_flags;
-};
-
-static inline struct acpi_nfit_memory_map *__to_nfit_memdev(
-               struct nfit_mem *nfit_mem)
-{
-       if (nfit_mem->memdev_dcr)
-               return nfit_mem->memdev_dcr;
-       return nfit_mem->memdev_pmem;
-}
-
-static inline struct acpi_nfit_desc *to_acpi_desc(
-               struct nvdimm_bus_descriptor *nd_desc)
-{
-       return container_of(nd_desc, struct acpi_nfit_desc, nd_desc);
-}
-
-const u8 *to_nfit_uuid(enum nfit_uuids id);
-int acpi_nfit_init(struct acpi_nfit_desc *acpi_desc, void *nfit, acpi_size sz);
-void acpi_nfit_desc_init(struct acpi_nfit_desc *acpi_desc, struct device *dev);
-#endif /* __NFIT_H__ */
diff --git a/drivers/acpi/nfit/Kconfig b/drivers/acpi/nfit/Kconfig
new file mode 100644 (file)
index 0000000..dd0d53c
--- /dev/null
@@ -0,0 +1,26 @@
+config ACPI_NFIT
+       tristate "ACPI NVDIMM Firmware Interface Table (NFIT)"
+       depends on PHYS_ADDR_T_64BIT
+       depends on BLK_DEV
+       depends on ARCH_HAS_MMIO_FLUSH
+       select LIBNVDIMM
+       help
+         Infrastructure to probe ACPI 6 compliant platforms for
+         NVDIMMs (NFIT) and register a libnvdimm device tree.  In
+         addition to storage devices this also enables libnvdimm to pass
+         ACPI._DSM messages for platform/dimm configuration.
+
+         To compile this driver as a module, choose M here:
+         the module will be called nfit.
+
+config ACPI_NFIT_DEBUG
+       bool "NFIT DSM debug"
+       depends on ACPI_NFIT
+       depends on DYNAMIC_DEBUG
+       default n
+       help
+         Enabling this option causes the nfit driver to dump the
+         input and output buffers of _DSM operations on the ACPI0012
+         device and its children.  This can be very verbose, so leave
+         it disabled unless you are debugging a hardware / firmware
+         issue.
diff --git a/drivers/acpi/nfit/Makefile b/drivers/acpi/nfit/Makefile
new file mode 100644 (file)
index 0000000..eb95c5a
--- /dev/null
@@ -0,0 +1,2 @@
+obj-$(CONFIG_ACPI_NFIT) := nfit.o
+nfit-y := core.o
diff --git a/drivers/acpi/nfit/core.c b/drivers/acpi/nfit/core.c
new file mode 100644 (file)
index 0000000..19d0dfd
--- /dev/null
@@ -0,0 +1,2765 @@
+/*
+ * Copyright(c) 2013-2015 Intel Corporation. All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that 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/list_sort.h>
+#include <linux/libnvdimm.h>
+#include <linux/module.h>
+#include <linux/mutex.h>
+#include <linux/ndctl.h>
+#include <linux/sysfs.h>
+#include <linux/delay.h>
+#include <linux/list.h>
+#include <linux/acpi.h>
+#include <linux/sort.h>
+#include <linux/pmem.h>
+#include <linux/io.h>
+#include <linux/nd.h>
+#include <asm/cacheflush.h>
+#include "nfit.h"
+
+/*
+ * For readq() and writeq() on 32-bit builds, the hi-lo, lo-hi order is
+ * irrelevant.
+ */
+#include <linux/io-64-nonatomic-hi-lo.h>
+
+static bool force_enable_dimms;
+module_param(force_enable_dimms, bool, S_IRUGO|S_IWUSR);
+MODULE_PARM_DESC(force_enable_dimms, "Ignore _STA (ACPI DIMM device) status");
+
+static unsigned int scrub_timeout = NFIT_ARS_TIMEOUT;
+module_param(scrub_timeout, uint, S_IRUGO|S_IWUSR);
+MODULE_PARM_DESC(scrub_timeout, "Initial scrub timeout in seconds");
+
+/* after three payloads of overflow, it's dead jim */
+static unsigned int scrub_overflow_abort = 3;
+module_param(scrub_overflow_abort, uint, S_IRUGO|S_IWUSR);
+MODULE_PARM_DESC(scrub_overflow_abort,
+               "Number of times we overflow ARS results before abort");
+
+static bool disable_vendor_specific;
+module_param(disable_vendor_specific, bool, S_IRUGO);
+MODULE_PARM_DESC(disable_vendor_specific,
+               "Limit commands to the publicly specified set\n");
+
+static struct workqueue_struct *nfit_wq;
+
+struct nfit_table_prev {
+       struct list_head spas;
+       struct list_head memdevs;
+       struct list_head dcrs;
+       struct list_head bdws;
+       struct list_head idts;
+       struct list_head flushes;
+};
+
+static u8 nfit_uuid[NFIT_UUID_MAX][16];
+
+const u8 *to_nfit_uuid(enum nfit_uuids id)
+{
+       return nfit_uuid[id];
+}
+EXPORT_SYMBOL(to_nfit_uuid);
+
+static struct acpi_nfit_desc *to_acpi_nfit_desc(
+               struct nvdimm_bus_descriptor *nd_desc)
+{
+       return container_of(nd_desc, struct acpi_nfit_desc, nd_desc);
+}
+
+static struct acpi_device *to_acpi_dev(struct acpi_nfit_desc *acpi_desc)
+{
+       struct nvdimm_bus_descriptor *nd_desc = &acpi_desc->nd_desc;
+
+       /*
+        * If provider == 'ACPI.NFIT' we can assume 'dev' is a struct
+        * acpi_device.
+        */
+       if (!nd_desc->provider_name
+                       || strcmp(nd_desc->provider_name, "ACPI.NFIT") != 0)
+               return NULL;
+
+       return to_acpi_device(acpi_desc->dev);
+}
+
+static int xlat_status(void *buf, unsigned int cmd)
+{
+       struct nd_cmd_clear_error *clear_err;
+       struct nd_cmd_ars_status *ars_status;
+       struct nd_cmd_ars_start *ars_start;
+       struct nd_cmd_ars_cap *ars_cap;
+       u16 flags;
+
+       switch (cmd) {
+       case ND_CMD_ARS_CAP:
+               ars_cap = buf;
+               if ((ars_cap->status & 0xffff) == NFIT_ARS_CAP_NONE)
+                       return -ENOTTY;
+
+               /* Command failed */
+               if (ars_cap->status & 0xffff)
+                       return -EIO;
+
+               /* No supported scan types for this range */
+               flags = ND_ARS_PERSISTENT | ND_ARS_VOLATILE;
+               if ((ars_cap->status >> 16 & flags) == 0)
+                       return -ENOTTY;
+               break;
+       case ND_CMD_ARS_START:
+               ars_start = buf;
+               /* ARS is in progress */
+               if ((ars_start->status & 0xffff) == NFIT_ARS_START_BUSY)
+                       return -EBUSY;
+
+               /* Command failed */
+               if (ars_start->status & 0xffff)
+                       return -EIO;
+               break;
+       case ND_CMD_ARS_STATUS:
+               ars_status = buf;
+               /* Command failed */
+               if (ars_status->status & 0xffff)
+                       return -EIO;
+               /* Check extended status (Upper two bytes) */
+               if (ars_status->status == NFIT_ARS_STATUS_DONE)
+                       return 0;
+
+               /* ARS is in progress */
+               if (ars_status->status == NFIT_ARS_STATUS_BUSY)
+                       return -EBUSY;
+
+               /* No ARS performed for the current boot */
+               if (ars_status->status == NFIT_ARS_STATUS_NONE)
+                       return -EAGAIN;
+
+               /*
+                * ARS interrupted, either we overflowed or some other
+                * agent wants the scan to stop.  If we didn't overflow
+                * then just continue with the returned results.
+                */
+               if (ars_status->status == NFIT_ARS_STATUS_INTR) {
+                       if (ars_status->flags & NFIT_ARS_F_OVERFLOW)
+                               return -ENOSPC;
+                       return 0;
+               }
+
+               /* Unknown status */
+               if (ars_status->status >> 16)
+                       return -EIO;
+               break;
+       case ND_CMD_CLEAR_ERROR:
+               clear_err = buf;
+               if (clear_err->status & 0xffff)
+                       return -EIO;
+               if (!clear_err->cleared)
+                       return -EIO;
+               if (clear_err->length > clear_err->cleared)
+                       return clear_err->cleared;
+               break;
+       default:
+               break;
+       }
+
+       return 0;
+}
+
+static int acpi_nfit_ctl(struct nvdimm_bus_descriptor *nd_desc,
+               struct nvdimm *nvdimm, unsigned int cmd, void *buf,
+               unsigned int buf_len, int *cmd_rc)
+{
+       struct acpi_nfit_desc *acpi_desc = to_acpi_nfit_desc(nd_desc);
+       union acpi_object in_obj, in_buf, *out_obj;
+       const struct nd_cmd_desc *desc = NULL;
+       struct device *dev = acpi_desc->dev;
+       struct nd_cmd_pkg *call_pkg = NULL;
+       const char *cmd_name, *dimm_name;
+       unsigned long cmd_mask, dsm_mask;
+       acpi_handle handle;
+       unsigned int func;
+       const u8 *uuid;
+       u32 offset;
+       int rc, i;
+
+       func = cmd;
+       if (cmd == ND_CMD_CALL) {
+               call_pkg = buf;
+               func = call_pkg->nd_command;
+       }
+
+       if (nvdimm) {
+               struct nfit_mem *nfit_mem = nvdimm_provider_data(nvdimm);
+               struct acpi_device *adev = nfit_mem->adev;
+
+               if (!adev)
+                       return -ENOTTY;
+               if (call_pkg && nfit_mem->family != call_pkg->nd_family)
+                       return -ENOTTY;
+
+               dimm_name = nvdimm_name(nvdimm);
+               cmd_name = nvdimm_cmd_name(cmd);
+               cmd_mask = nvdimm_cmd_mask(nvdimm);
+               dsm_mask = nfit_mem->dsm_mask;
+               desc = nd_cmd_dimm_desc(cmd);
+               uuid = to_nfit_uuid(nfit_mem->family);
+               handle = adev->handle;
+       } else {
+               struct acpi_device *adev = to_acpi_dev(acpi_desc);
+
+               cmd_name = nvdimm_bus_cmd_name(cmd);
+               cmd_mask = nd_desc->cmd_mask;
+               dsm_mask = cmd_mask;
+               desc = nd_cmd_bus_desc(cmd);
+               uuid = to_nfit_uuid(NFIT_DEV_BUS);
+               handle = adev->handle;
+               dimm_name = "bus";
+       }
+
+       if (!desc || (cmd && (desc->out_num + desc->in_num == 0)))
+               return -ENOTTY;
+
+       if (!test_bit(cmd, &cmd_mask) || !test_bit(func, &dsm_mask))
+               return -ENOTTY;
+
+       in_obj.type = ACPI_TYPE_PACKAGE;
+       in_obj.package.count = 1;
+       in_obj.package.elements = &in_buf;
+       in_buf.type = ACPI_TYPE_BUFFER;
+       in_buf.buffer.pointer = buf;
+       in_buf.buffer.length = 0;
+
+       /* libnvdimm has already validated the input envelope */
+       for (i = 0; i < desc->in_num; i++)
+               in_buf.buffer.length += nd_cmd_in_size(nvdimm, cmd, desc,
+                               i, buf);
+
+       if (call_pkg) {
+               /* skip over package wrapper */
+               in_buf.buffer.pointer = (void *) &call_pkg->nd_payload;
+               in_buf.buffer.length = call_pkg->nd_size_in;
+       }
+
+       if (IS_ENABLED(CONFIG_ACPI_NFIT_DEBUG)) {
+               dev_dbg(dev, "%s:%s cmd: %d: func: %d input length: %d\n",
+                               __func__, dimm_name, cmd, func,
+                               in_buf.buffer.length);
+               print_hex_dump_debug("nvdimm in  ", DUMP_PREFIX_OFFSET, 4, 4,
+                       in_buf.buffer.pointer,
+                       min_t(u32, 256, in_buf.buffer.length), true);
+       }
+
+       out_obj = acpi_evaluate_dsm(handle, uuid, 1, func, &in_obj);
+       if (!out_obj) {
+               dev_dbg(dev, "%s:%s _DSM failed cmd: %s\n", __func__, dimm_name,
+                               cmd_name);
+               return -EINVAL;
+       }
+
+       if (call_pkg) {
+               call_pkg->nd_fw_size = out_obj->buffer.length;
+               memcpy(call_pkg->nd_payload + call_pkg->nd_size_in,
+                       out_obj->buffer.pointer,
+                       min(call_pkg->nd_fw_size, call_pkg->nd_size_out));
+
+               ACPI_FREE(out_obj);
+               /*
+                * Need to support FW function w/o known size in advance.
+                * Caller can determine required size based upon nd_fw_size.
+                * If we return an error (like elsewhere) then caller wouldn't
+                * be able to rely upon data returned to make calculation.
+                */
+               return 0;
+       }
+
+       if (out_obj->package.type != ACPI_TYPE_BUFFER) {
+               dev_dbg(dev, "%s:%s unexpected output object type cmd: %s type: %d\n",
+                               __func__, dimm_name, cmd_name, out_obj->type);
+               rc = -EINVAL;
+               goto out;
+       }
+
+       if (IS_ENABLED(CONFIG_ACPI_NFIT_DEBUG)) {
+               dev_dbg(dev, "%s:%s cmd: %s output length: %d\n", __func__,
+                               dimm_name, cmd_name, out_obj->buffer.length);
+               print_hex_dump_debug(cmd_name, DUMP_PREFIX_OFFSET, 4,
+                               4, out_obj->buffer.pointer, min_t(u32, 128,
+                                       out_obj->buffer.length), true);
+       }
+
+       for (i = 0, offset = 0; i < desc->out_num; i++) {
+               u32 out_size = nd_cmd_out_size(nvdimm, cmd, desc, i, buf,
+                               (u32 *) out_obj->buffer.pointer);
+
+               if (offset + out_size > out_obj->buffer.length) {
+                       dev_dbg(dev, "%s:%s output object underflow cmd: %s field: %d\n",
+                                       __func__, dimm_name, cmd_name, i);
+                       break;
+               }
+
+               if (in_buf.buffer.length + offset + out_size > buf_len) {
+                       dev_dbg(dev, "%s:%s output overrun cmd: %s field: %d\n",
+                                       __func__, dimm_name, cmd_name, i);
+                       rc = -ENXIO;
+                       goto out;
+               }
+               memcpy(buf + in_buf.buffer.length + offset,
+                               out_obj->buffer.pointer + offset, out_size);
+               offset += out_size;
+       }
+       if (offset + in_buf.buffer.length < buf_len) {
+               if (i >= 1) {
+                       /*
+                        * status valid, return the number of bytes left
+                        * unfilled in the output buffer
+                        */
+                       rc = buf_len - offset - in_buf.buffer.length;
+                       if (cmd_rc)
+                               *cmd_rc = xlat_status(buf, cmd);
+               } else {
+                       dev_err(dev, "%s:%s underrun cmd: %s buf_len: %d out_len: %d\n",
+                                       __func__, dimm_name, cmd_name, buf_len,
+                                       offset);
+                       rc = -ENXIO;
+               }
+       } else {
+               rc = 0;
+               if (cmd_rc)
+                       *cmd_rc = xlat_status(buf, cmd);
+       }
+
+ out:
+       ACPI_FREE(out_obj);
+
+       return rc;
+}
+
+static const char *spa_type_name(u16 type)
+{
+       static const char *to_name[] = {
+               [NFIT_SPA_VOLATILE] = "volatile",
+               [NFIT_SPA_PM] = "pmem",
+               [NFIT_SPA_DCR] = "dimm-control-region",
+               [NFIT_SPA_BDW] = "block-data-window",
+               [NFIT_SPA_VDISK] = "volatile-disk",
+               [NFIT_SPA_VCD] = "volatile-cd",
+               [NFIT_SPA_PDISK] = "persistent-disk",
+               [NFIT_SPA_PCD] = "persistent-cd",
+
+       };
+
+       if (type > NFIT_SPA_PCD)
+               return "unknown";
+
+       return to_name[type];
+}
+
+static int nfit_spa_type(struct acpi_nfit_system_address *spa)
+{
+       int i;
+
+       for (i = 0; i < NFIT_UUID_MAX; i++)
+               if (memcmp(to_nfit_uuid(i), spa->range_guid, 16) == 0)
+                       return i;
+       return -1;
+}
+
+static bool add_spa(struct acpi_nfit_desc *acpi_desc,
+               struct nfit_table_prev *prev,
+               struct acpi_nfit_system_address *spa)
+{
+       struct device *dev = acpi_desc->dev;
+       struct nfit_spa *nfit_spa;
+
+       if (spa->header.length != sizeof(*spa))
+               return false;
+
+       list_for_each_entry(nfit_spa, &prev->spas, list) {
+               if (memcmp(nfit_spa->spa, spa, sizeof(*spa)) == 0) {
+                       list_move_tail(&nfit_spa->list, &acpi_desc->spas);
+                       return true;
+               }
+       }
+
+       nfit_spa = devm_kzalloc(dev, sizeof(*nfit_spa) + sizeof(*spa),
+                       GFP_KERNEL);
+       if (!nfit_spa)
+               return false;
+       INIT_LIST_HEAD(&nfit_spa->list);
+       memcpy(nfit_spa->spa, spa, sizeof(*spa));
+       list_add_tail(&nfit_spa->list, &acpi_desc->spas);
+       dev_dbg(dev, "%s: spa index: %d type: %s\n", __func__,
+                       spa->range_index,
+                       spa_type_name(nfit_spa_type(spa)));
+       return true;
+}
+
+static bool add_memdev(struct acpi_nfit_desc *acpi_desc,
+               struct nfit_table_prev *prev,
+               struct acpi_nfit_memory_map *memdev)
+{
+       struct device *dev = acpi_desc->dev;
+       struct nfit_memdev *nfit_memdev;
+
+       if (memdev->header.length != sizeof(*memdev))
+               return false;
+
+       list_for_each_entry(nfit_memdev, &prev->memdevs, list)
+               if (memcmp(nfit_memdev->memdev, memdev, sizeof(*memdev)) == 0) {
+                       list_move_tail(&nfit_memdev->list, &acpi_desc->memdevs);
+                       return true;
+               }
+
+       nfit_memdev = devm_kzalloc(dev, sizeof(*nfit_memdev) + sizeof(*memdev),
+                       GFP_KERNEL);
+       if (!nfit_memdev)
+               return false;
+       INIT_LIST_HEAD(&nfit_memdev->list);
+       memcpy(nfit_memdev->memdev, memdev, sizeof(*memdev));
+       list_add_tail(&nfit_memdev->list, &acpi_desc->memdevs);
+       dev_dbg(dev, "%s: memdev handle: %#x spa: %d dcr: %d\n",
+                       __func__, memdev->device_handle, memdev->range_index,
+                       memdev->region_index);
+       return true;
+}
+
+/*
+ * An implementation may provide a truncated control region if no block windows
+ * are defined.
+ */
+static size_t sizeof_dcr(struct acpi_nfit_control_region *dcr)
+{
+       if (dcr->header.length < offsetof(struct acpi_nfit_control_region,
+                               window_size))
+               return 0;
+       if (dcr->windows)
+               return sizeof(*dcr);
+       return offsetof(struct acpi_nfit_control_region, window_size);
+}
+
+static bool add_dcr(struct acpi_nfit_desc *acpi_desc,
+               struct nfit_table_prev *prev,
+               struct acpi_nfit_control_region *dcr)
+{
+       struct device *dev = acpi_desc->dev;
+       struct nfit_dcr *nfit_dcr;
+
+       if (!sizeof_dcr(dcr))
+               return false;
+
+       list_for_each_entry(nfit_dcr, &prev->dcrs, list)
+               if (memcmp(nfit_dcr->dcr, dcr, sizeof_dcr(dcr)) == 0) {
+                       list_move_tail(&nfit_dcr->list, &acpi_desc->dcrs);
+                       return true;
+               }
+
+       nfit_dcr = devm_kzalloc(dev, sizeof(*nfit_dcr) + sizeof(*dcr),
+                       GFP_KERNEL);
+       if (!nfit_dcr)
+               return false;
+       INIT_LIST_HEAD(&nfit_dcr->list);
+       memcpy(nfit_dcr->dcr, dcr, sizeof_dcr(dcr));
+       list_add_tail(&nfit_dcr->list, &acpi_desc->dcrs);
+       dev_dbg(dev, "%s: dcr index: %d windows: %d\n", __func__,
+                       dcr->region_index, dcr->windows);
+       return true;
+}
+
+static bool add_bdw(struct acpi_nfit_desc *acpi_desc,
+               struct nfit_table_prev *prev,
+               struct acpi_nfit_data_region *bdw)
+{
+       struct device *dev = acpi_desc->dev;
+       struct nfit_bdw *nfit_bdw;
+
+       if (bdw->header.length != sizeof(*bdw))
+               return false;
+       list_for_each_entry(nfit_bdw, &prev->bdws, list)
+               if (memcmp(nfit_bdw->bdw, bdw, sizeof(*bdw)) == 0) {
+                       list_move_tail(&nfit_bdw->list, &acpi_desc->bdws);
+                       return true;
+               }
+
+       nfit_bdw = devm_kzalloc(dev, sizeof(*nfit_bdw) + sizeof(*bdw),
+                       GFP_KERNEL);
+       if (!nfit_bdw)
+               return false;
+       INIT_LIST_HEAD(&nfit_bdw->list);
+       memcpy(nfit_bdw->bdw, bdw, sizeof(*bdw));
+       list_add_tail(&nfit_bdw->list, &acpi_desc->bdws);
+       dev_dbg(dev, "%s: bdw dcr: %d windows: %d\n", __func__,
+                       bdw->region_index, bdw->windows);
+       return true;
+}
+
+static size_t sizeof_idt(struct acpi_nfit_interleave *idt)
+{
+       if (idt->header.length < sizeof(*idt))
+               return 0;
+       return sizeof(*idt) + sizeof(u32) * (idt->line_count - 1);
+}
+
+static bool add_idt(struct acpi_nfit_desc *acpi_desc,
+               struct nfit_table_prev *prev,
+               struct acpi_nfit_interleave *idt)
+{
+       struct device *dev = acpi_desc->dev;
+       struct nfit_idt *nfit_idt;
+
+       if (!sizeof_idt(idt))
+               return false;
+
+       list_for_each_entry(nfit_idt, &prev->idts, list) {
+               if (sizeof_idt(nfit_idt->idt) != sizeof_idt(idt))
+                       continue;
+
+               if (memcmp(nfit_idt->idt, idt, sizeof_idt(idt)) == 0) {
+                       list_move_tail(&nfit_idt->list, &acpi_desc->idts);
+                       return true;
+               }
+       }
+
+       nfit_idt = devm_kzalloc(dev, sizeof(*nfit_idt) + sizeof_idt(idt),
+                       GFP_KERNEL);
+       if (!nfit_idt)
+               return false;
+       INIT_LIST_HEAD(&nfit_idt->list);
+       memcpy(nfit_idt->idt, idt, sizeof_idt(idt));
+       list_add_tail(&nfit_idt->list, &acpi_desc->idts);
+       dev_dbg(dev, "%s: idt index: %d num_lines: %d\n", __func__,
+                       idt->interleave_index, idt->line_count);
+       return true;
+}
+
+static size_t sizeof_flush(struct acpi_nfit_flush_address *flush)
+{
+       if (flush->header.length < sizeof(*flush))
+               return 0;
+       return sizeof(*flush) + sizeof(u64) * (flush->hint_count - 1);
+}
+
+static bool add_flush(struct acpi_nfit_desc *acpi_desc,
+               struct nfit_table_prev *prev,
+               struct acpi_nfit_flush_address *flush)
+{
+       struct device *dev = acpi_desc->dev;
+       struct nfit_flush *nfit_flush;
+
+       if (!sizeof_flush(flush))
+               return false;
+
+       list_for_each_entry(nfit_flush, &prev->flushes, list) {
+               if (sizeof_flush(nfit_flush->flush) != sizeof_flush(flush))
+                       continue;
+
+               if (memcmp(nfit_flush->flush, flush,
+                                       sizeof_flush(flush)) == 0) {
+                       list_move_tail(&nfit_flush->list, &acpi_desc->flushes);
+                       return true;
+               }
+       }
+
+       nfit_flush = devm_kzalloc(dev, sizeof(*nfit_flush)
+                       + sizeof_flush(flush), GFP_KERNEL);
+       if (!nfit_flush)
+               return false;
+       INIT_LIST_HEAD(&nfit_flush->list);
+       memcpy(nfit_flush->flush, flush, sizeof_flush(flush));
+       list_add_tail(&nfit_flush->list, &acpi_desc->flushes);
+       dev_dbg(dev, "%s: nfit_flush handle: %d hint_count: %d\n", __func__,
+                       flush->device_handle, flush->hint_count);
+       return true;
+}
+
+static void *add_table(struct acpi_nfit_desc *acpi_desc,
+               struct nfit_table_prev *prev, void *table, const void *end)
+{
+       struct device *dev = acpi_desc->dev;
+       struct acpi_nfit_header *hdr;
+       void *err = ERR_PTR(-ENOMEM);
+
+       if (table >= end)
+               return NULL;
+
+       hdr = table;
+       if (!hdr->length) {
+               dev_warn(dev, "found a zero length table '%d' parsing nfit\n",
+                       hdr->type);
+               return NULL;
+       }
+
+       switch (hdr->type) {
+       case ACPI_NFIT_TYPE_SYSTEM_ADDRESS:
+               if (!add_spa(acpi_desc, prev, table))
+                       return err;
+               break;
+       case ACPI_NFIT_TYPE_MEMORY_MAP:
+               if (!add_memdev(acpi_desc, prev, table))
+                       return err;
+               break;
+       case ACPI_NFIT_TYPE_CONTROL_REGION:
+               if (!add_dcr(acpi_desc, prev, table))
+                       return err;
+               break;
+       case ACPI_NFIT_TYPE_DATA_REGION:
+               if (!add_bdw(acpi_desc, prev, table))
+                       return err;
+               break;
+       case ACPI_NFIT_TYPE_INTERLEAVE:
+               if (!add_idt(acpi_desc, prev, table))
+                       return err;
+               break;
+       case ACPI_NFIT_TYPE_FLUSH_ADDRESS:
+               if (!add_flush(acpi_desc, prev, table))
+                       return err;
+               break;
+       case ACPI_NFIT_TYPE_SMBIOS:
+               dev_dbg(dev, "%s: smbios\n", __func__);
+               break;
+       default:
+               dev_err(dev, "unknown table '%d' parsing nfit\n", hdr->type);
+               break;
+       }
+
+       return table + hdr->length;
+}
+
+static void nfit_mem_find_spa_bdw(struct acpi_nfit_desc *acpi_desc,
+               struct nfit_mem *nfit_mem)
+{
+       u32 device_handle = __to_nfit_memdev(nfit_mem)->device_handle;
+       u16 dcr = nfit_mem->dcr->region_index;
+       struct nfit_spa *nfit_spa;
+
+       list_for_each_entry(nfit_spa, &acpi_desc->spas, list) {
+               u16 range_index = nfit_spa->spa->range_index;
+               int type = nfit_spa_type(nfit_spa->spa);
+               struct nfit_memdev *nfit_memdev;
+
+               if (type != NFIT_SPA_BDW)
+                       continue;
+
+               list_for_each_entry(nfit_memdev, &acpi_desc->memdevs, list) {
+                       if (nfit_memdev->memdev->range_index != range_index)
+                               continue;
+                       if (nfit_memdev->memdev->device_handle != device_handle)
+                               continue;
+                       if (nfit_memdev->memdev->region_index != dcr)
+                               continue;
+
+                       nfit_mem->spa_bdw = nfit_spa->spa;
+                       return;
+               }
+       }
+
+       dev_dbg(acpi_desc->dev, "SPA-BDW not found for SPA-DCR %d\n",
+                       nfit_mem->spa_dcr->range_index);
+       nfit_mem->bdw = NULL;
+}
+
+static void nfit_mem_init_bdw(struct acpi_nfit_desc *acpi_desc,
+               struct nfit_mem *nfit_mem, struct acpi_nfit_system_address *spa)
+{
+       u16 dcr = __to_nfit_memdev(nfit_mem)->region_index;
+       struct nfit_memdev *nfit_memdev;
+       struct nfit_bdw *nfit_bdw;
+       struct nfit_idt *nfit_idt;
+       u16 idt_idx, range_index;
+
+       list_for_each_entry(nfit_bdw, &acpi_desc->bdws, list) {
+               if (nfit_bdw->bdw->region_index != dcr)
+                       continue;
+               nfit_mem->bdw = nfit_bdw->bdw;
+               break;
+       }
+
+       if (!nfit_mem->bdw)
+               return;
+
+       nfit_mem_find_spa_bdw(acpi_desc, nfit_mem);
+
+       if (!nfit_mem->spa_bdw)
+               return;
+
+       range_index = nfit_mem->spa_bdw->range_index;
+       list_for_each_entry(nfit_memdev, &acpi_desc->memdevs, list) {
+               if (nfit_memdev->memdev->range_index != range_index ||
+                               nfit_memdev->memdev->region_index != dcr)
+                       continue;
+               nfit_mem->memdev_bdw = nfit_memdev->memdev;
+               idt_idx = nfit_memdev->memdev->interleave_index;
+               list_for_each_entry(nfit_idt, &acpi_desc->idts, list) {
+                       if (nfit_idt->idt->interleave_index != idt_idx)
+                               continue;
+                       nfit_mem->idt_bdw = nfit_idt->idt;
+                       break;
+               }
+               break;
+       }
+}
+
+static int nfit_mem_dcr_init(struct acpi_nfit_desc *acpi_desc,
+               struct acpi_nfit_system_address *spa)
+{
+       struct nfit_mem *nfit_mem, *found;
+       struct nfit_memdev *nfit_memdev;
+       int type = nfit_spa_type(spa);
+
+       switch (type) {
+       case NFIT_SPA_DCR:
+       case NFIT_SPA_PM:
+               break;
+       default:
+               return 0;
+       }
+
+       list_for_each_entry(nfit_memdev, &acpi_desc->memdevs, list) {
+               struct nfit_flush *nfit_flush;
+               struct nfit_dcr *nfit_dcr;
+               u32 device_handle;
+               u16 dcr;
+
+               if (nfit_memdev->memdev->range_index != spa->range_index)
+                       continue;
+               found = NULL;
+               dcr = nfit_memdev->memdev->region_index;
+               device_handle = nfit_memdev->memdev->device_handle;
+               list_for_each_entry(nfit_mem, &acpi_desc->dimms, list)
+                       if (__to_nfit_memdev(nfit_mem)->device_handle
+                                       == device_handle) {
+                               found = nfit_mem;
+                               break;
+                       }
+
+               if (found)
+                       nfit_mem = found;
+               else {
+                       nfit_mem = devm_kzalloc(acpi_desc->dev,
+                                       sizeof(*nfit_mem), GFP_KERNEL);
+                       if (!nfit_mem)
+                               return -ENOMEM;
+                       INIT_LIST_HEAD(&nfit_mem->list);
+                       nfit_mem->acpi_desc = acpi_desc;
+                       list_add(&nfit_mem->list, &acpi_desc->dimms);
+               }
+
+               list_for_each_entry(nfit_dcr, &acpi_desc->dcrs, list) {
+                       if (nfit_dcr->dcr->region_index != dcr)
+                               continue;
+                       /*
+                        * Record the control region for the dimm.  For
+                        * the ACPI 6.1 case, where there are separate
+                        * control regions for the pmem vs blk
+                        * interfaces, be sure to record the extended
+                        * blk details.
+                        */
+                       if (!nfit_mem->dcr)
+                               nfit_mem->dcr = nfit_dcr->dcr;
+                       else if (nfit_mem->dcr->windows == 0
+                                       && nfit_dcr->dcr->windows)
+                               nfit_mem->dcr = nfit_dcr->dcr;
+                       break;
+               }
+
+               list_for_each_entry(nfit_flush, &acpi_desc->flushes, list) {
+                       struct acpi_nfit_flush_address *flush;
+                       u16 i;
+
+                       if (nfit_flush->flush->device_handle != device_handle)
+                               continue;
+                       nfit_mem->nfit_flush = nfit_flush;
+                       flush = nfit_flush->flush;
+                       nfit_mem->flush_wpq = devm_kzalloc(acpi_desc->dev,
+                                       flush->hint_count
+                                       * sizeof(struct resource), GFP_KERNEL);
+                       if (!nfit_mem->flush_wpq)
+                               return -ENOMEM;
+                       for (i = 0; i < flush->hint_count; i++) {
+                               struct resource *res = &nfit_mem->flush_wpq[i];
+
+                               res->start = flush->hint_address[i];
+                               res->end = res->start + 8 - 1;
+                       }
+                       break;
+               }
+
+               if (dcr && !nfit_mem->dcr) {
+                       dev_err(acpi_desc->dev, "SPA %d missing DCR %d\n",
+                                       spa->range_index, dcr);
+                       return -ENODEV;
+               }
+
+               if (type == NFIT_SPA_DCR) {
+                       struct nfit_idt *nfit_idt;
+                       u16 idt_idx;
+
+                       /* multiple dimms may share a SPA when interleaved */
+                       nfit_mem->spa_dcr = spa;
+                       nfit_mem->memdev_dcr = nfit_memdev->memdev;
+                       idt_idx = nfit_memdev->memdev->interleave_index;
+                       list_for_each_entry(nfit_idt, &acpi_desc->idts, list) {
+                               if (nfit_idt->idt->interleave_index != idt_idx)
+                                       continue;
+                               nfit_mem->idt_dcr = nfit_idt->idt;
+                               break;
+                       }
+                       nfit_mem_init_bdw(acpi_desc, nfit_mem, spa);
+               } else {
+                       /*
+                        * A single dimm may belong to multiple SPA-PM
+                        * ranges, record at least one in addition to
+                        * any SPA-DCR range.
+                        */
+                       nfit_mem->memdev_pmem = nfit_memdev->memdev;
+               }
+       }
+
+       return 0;
+}
+
+static int nfit_mem_cmp(void *priv, struct list_head *_a, struct list_head *_b)
+{
+       struct nfit_mem *a = container_of(_a, typeof(*a), list);
+       struct nfit_mem *b = container_of(_b, typeof(*b), list);
+       u32 handleA, handleB;
+
+       handleA = __to_nfit_memdev(a)->device_handle;
+       handleB = __to_nfit_memdev(b)->device_handle;
+       if (handleA < handleB)
+               return -1;
+       else if (handleA > handleB)
+               return 1;
+       return 0;
+}
+
+static int nfit_mem_init(struct acpi_nfit_desc *acpi_desc)
+{
+       struct nfit_spa *nfit_spa;
+
+       /*
+        * For each SPA-DCR or SPA-PMEM address range find its
+        * corresponding MEMDEV(s).  From each MEMDEV find the
+        * corresponding DCR.  Then, if we're operating on a SPA-DCR,
+        * try to find a SPA-BDW and a corresponding BDW that references
+        * the DCR.  Throw it all into an nfit_mem object.  Note, that
+        * BDWs are optional.
+        */
+       list_for_each_entry(nfit_spa, &acpi_desc->spas, list) {
+               int rc;
+
+               rc = nfit_mem_dcr_init(acpi_desc, nfit_spa->spa);
+               if (rc)
+                       return rc;
+       }
+
+       list_sort(NULL, &acpi_desc->dimms, nfit_mem_cmp);
+
+       return 0;
+}
+
+static ssize_t revision_show(struct device *dev,
+               struct device_attribute *attr, char *buf)
+{
+       struct nvdimm_bus *nvdimm_bus = to_nvdimm_bus(dev);
+       struct nvdimm_bus_descriptor *nd_desc = to_nd_desc(nvdimm_bus);
+       struct acpi_nfit_desc *acpi_desc = to_acpi_desc(nd_desc);
+
+       return sprintf(buf, "%d\n", acpi_desc->acpi_header.revision);
+}
+static DEVICE_ATTR_RO(revision);
+
+/*
+ * This shows the number of full Address Range Scrubs that have been
+ * completed since driver load time. Userspace can wait on this using
+ * select/poll etc. A '+' at the end indicates an ARS is in progress
+ */
+static ssize_t scrub_show(struct device *dev,
+               struct device_attribute *attr, char *buf)
+{
+       struct nvdimm_bus_descriptor *nd_desc;
+       ssize_t rc = -ENXIO;
+
+       device_lock(dev);
+       nd_desc = dev_get_drvdata(dev);
+       if (nd_desc) {
+               struct acpi_nfit_desc *acpi_desc = to_acpi_desc(nd_desc);
+
+               rc = sprintf(buf, "%d%s", acpi_desc->scrub_count,
+                               (work_busy(&acpi_desc->work)) ? "+\n" : "\n");
+       }
+       device_unlock(dev);
+       return rc;
+}
+
+static int acpi_nfit_ars_rescan(struct acpi_nfit_desc *acpi_desc);
+
+static ssize_t scrub_store(struct device *dev,
+               struct device_attribute *attr, const char *buf, size_t size)
+{
+       struct nvdimm_bus_descriptor *nd_desc;
+       ssize_t rc;
+       long val;
+
+       rc = kstrtol(buf, 0, &val);
+       if (rc)
+               return rc;
+       if (val != 1)
+               return -EINVAL;
+
+       device_lock(dev);
+       nd_desc = dev_get_drvdata(dev);
+       if (nd_desc) {
+               struct acpi_nfit_desc *acpi_desc = to_acpi_desc(nd_desc);
+
+               rc = acpi_nfit_ars_rescan(acpi_desc);
+       }
+       device_unlock(dev);
+       if (rc)
+               return rc;
+       return size;
+}
+static DEVICE_ATTR_RW(scrub);
+
+static bool ars_supported(struct nvdimm_bus *nvdimm_bus)
+{
+       struct nvdimm_bus_descriptor *nd_desc = to_nd_desc(nvdimm_bus);
+       const unsigned long mask = 1 << ND_CMD_ARS_CAP | 1 << ND_CMD_ARS_START
+               | 1 << ND_CMD_ARS_STATUS;
+
+       return (nd_desc->cmd_mask & mask) == mask;
+}
+
+static umode_t nfit_visible(struct kobject *kobj, struct attribute *a, int n)
+{
+       struct device *dev = container_of(kobj, struct device, kobj);
+       struct nvdimm_bus *nvdimm_bus = to_nvdimm_bus(dev);
+
+       if (a == &dev_attr_scrub.attr && !ars_supported(nvdimm_bus))
+               return 0;
+       return a->mode;
+}
+
+static struct attribute *acpi_nfit_attributes[] = {
+       &dev_attr_revision.attr,
+       &dev_attr_scrub.attr,
+       NULL,
+};
+
+static struct attribute_group acpi_nfit_attribute_group = {
+       .name = "nfit",
+       .attrs = acpi_nfit_attributes,
+       .is_visible = nfit_visible,
+};
+
+static const struct attribute_group *acpi_nfit_attribute_groups[] = {
+       &nvdimm_bus_attribute_group,
+       &acpi_nfit_attribute_group,
+       NULL,
+};
+
+static struct acpi_nfit_memory_map *to_nfit_memdev(struct device *dev)
+{
+       struct nvdimm *nvdimm = to_nvdimm(dev);
+       struct nfit_mem *nfit_mem = nvdimm_provider_data(nvdimm);
+
+       return __to_nfit_memdev(nfit_mem);
+}
+
+static struct acpi_nfit_control_region *to_nfit_dcr(struct device *dev)
+{
+       struct nvdimm *nvdimm = to_nvdimm(dev);
+       struct nfit_mem *nfit_mem = nvdimm_provider_data(nvdimm);
+
+       return nfit_mem->dcr;
+}
+
+static ssize_t handle_show(struct device *dev,
+               struct device_attribute *attr, char *buf)
+{
+       struct acpi_nfit_memory_map *memdev = to_nfit_memdev(dev);
+
+       return sprintf(buf, "%#x\n", memdev->device_handle);
+}
+static DEVICE_ATTR_RO(handle);
+
+static ssize_t phys_id_show(struct device *dev,
+               struct device_attribute *attr, char *buf)
+{
+       struct acpi_nfit_memory_map *memdev = to_nfit_memdev(dev);
+
+       return sprintf(buf, "%#x\n", memdev->physical_id);
+}
+static DEVICE_ATTR_RO(phys_id);
+
+static ssize_t vendor_show(struct device *dev,
+               struct device_attribute *attr, char *buf)
+{
+       struct acpi_nfit_control_region *dcr = to_nfit_dcr(dev);
+
+       return sprintf(buf, "0x%04x\n", be16_to_cpu(dcr->vendor_id));
+}
+static DEVICE_ATTR_RO(vendor);
+
+static ssize_t rev_id_show(struct device *dev,
+               struct device_attribute *attr, char *buf)
+{
+       struct acpi_nfit_control_region *dcr = to_nfit_dcr(dev);
+
+       return sprintf(buf, "0x%04x\n", be16_to_cpu(dcr->revision_id));
+}
+static DEVICE_ATTR_RO(rev_id);
+
+static ssize_t device_show(struct device *dev,
+               struct device_attribute *attr, char *buf)
+{
+       struct acpi_nfit_control_region *dcr = to_nfit_dcr(dev);
+
+       return sprintf(buf, "0x%04x\n", be16_to_cpu(dcr->device_id));
+}
+static DEVICE_ATTR_RO(device);
+
+static ssize_t subsystem_vendor_show(struct device *dev,
+               struct device_attribute *attr, char *buf)
+{
+       struct acpi_nfit_control_region *dcr = to_nfit_dcr(dev);
+
+       return sprintf(buf, "0x%04x\n", be16_to_cpu(dcr->subsystem_vendor_id));
+}
+static DEVICE_ATTR_RO(subsystem_vendor);
+
+static ssize_t subsystem_rev_id_show(struct device *dev,
+               struct device_attribute *attr, char *buf)
+{
+       struct acpi_nfit_control_region *dcr = to_nfit_dcr(dev);
+
+       return sprintf(buf, "0x%04x\n",
+                       be16_to_cpu(dcr->subsystem_revision_id));
+}
+static DEVICE_ATTR_RO(subsystem_rev_id);
+
+static ssize_t subsystem_device_show(struct device *dev,
+               struct device_attribute *attr, char *buf)
+{
+       struct acpi_nfit_control_region *dcr = to_nfit_dcr(dev);
+
+       return sprintf(buf, "0x%04x\n", be16_to_cpu(dcr->subsystem_device_id));
+}
+static DEVICE_ATTR_RO(subsystem_device);
+
+static int num_nvdimm_formats(struct nvdimm *nvdimm)
+{
+       struct nfit_mem *nfit_mem = nvdimm_provider_data(nvdimm);
+       int formats = 0;
+
+       if (nfit_mem->memdev_pmem)
+               formats++;
+       if (nfit_mem->memdev_bdw)
+               formats++;
+       return formats;
+}
+
+static ssize_t format_show(struct device *dev,
+               struct device_attribute *attr, char *buf)
+{
+       struct acpi_nfit_control_region *dcr = to_nfit_dcr(dev);
+
+       return sprintf(buf, "0x%04x\n", be16_to_cpu(dcr->code));
+}
+static DEVICE_ATTR_RO(format);
+
+static ssize_t format1_show(struct device *dev,
+               struct device_attribute *attr, char *buf)
+{
+       u32 handle;
+       ssize_t rc = -ENXIO;
+       struct nfit_mem *nfit_mem;
+       struct nfit_memdev *nfit_memdev;
+       struct acpi_nfit_desc *acpi_desc;
+       struct nvdimm *nvdimm = to_nvdimm(dev);
+       struct acpi_nfit_control_region *dcr = to_nfit_dcr(dev);
+
+       nfit_mem = nvdimm_provider_data(nvdimm);
+       acpi_desc = nfit_mem->acpi_desc;
+       handle = to_nfit_memdev(dev)->device_handle;
+
+       /* assumes DIMMs have at most 2 published interface codes */
+       mutex_lock(&acpi_desc->init_mutex);
+       list_for_each_entry(nfit_memdev, &acpi_desc->memdevs, list) {
+               struct acpi_nfit_memory_map *memdev = nfit_memdev->memdev;
+               struct nfit_dcr *nfit_dcr;
+
+               if (memdev->device_handle != handle)
+                       continue;
+
+               list_for_each_entry(nfit_dcr, &acpi_desc->dcrs, list) {
+                       if (nfit_dcr->dcr->region_index != memdev->region_index)
+                               continue;
+                       if (nfit_dcr->dcr->code == dcr->code)
+                               continue;
+                       rc = sprintf(buf, "%#x\n",
+                                       be16_to_cpu(nfit_dcr->dcr->code));
+                       break;
+               }
+               if (rc != ENXIO)
+                       break;
+       }
+       mutex_unlock(&acpi_desc->init_mutex);
+       return rc;
+}
+static DEVICE_ATTR_RO(format1);
+
+static ssize_t formats_show(struct device *dev,
+               struct device_attribute *attr, char *buf)
+{
+       struct nvdimm *nvdimm = to_nvdimm(dev);
+
+       return sprintf(buf, "%d\n", num_nvdimm_formats(nvdimm));
+}
+static DEVICE_ATTR_RO(formats);
+
+static ssize_t serial_show(struct device *dev,
+               struct device_attribute *attr, char *buf)
+{
+       struct acpi_nfit_control_region *dcr = to_nfit_dcr(dev);
+
+       return sprintf(buf, "0x%08x\n", be32_to_cpu(dcr->serial_number));
+}
+static DEVICE_ATTR_RO(serial);
+
+static ssize_t family_show(struct device *dev,
+               struct device_attribute *attr, char *buf)
+{
+       struct nvdimm *nvdimm = to_nvdimm(dev);
+       struct nfit_mem *nfit_mem = nvdimm_provider_data(nvdimm);
+
+       if (nfit_mem->family < 0)
+               return -ENXIO;
+       return sprintf(buf, "%d\n", nfit_mem->family);
+}
+static DEVICE_ATTR_RO(family);
+
+static ssize_t dsm_mask_show(struct device *dev,
+               struct device_attribute *attr, char *buf)
+{
+       struct nvdimm *nvdimm = to_nvdimm(dev);
+       struct nfit_mem *nfit_mem = nvdimm_provider_data(nvdimm);
+
+       if (nfit_mem->family < 0)
+               return -ENXIO;
+       return sprintf(buf, "%#lx\n", nfit_mem->dsm_mask);
+}
+static DEVICE_ATTR_RO(dsm_mask);
+
+static ssize_t flags_show(struct device *dev,
+               struct device_attribute *attr, char *buf)
+{
+       u16 flags = to_nfit_memdev(dev)->flags;
+
+       return sprintf(buf, "%s%s%s%s%s\n",
+               flags & ACPI_NFIT_MEM_SAVE_FAILED ? "save_fail " : "",
+               flags & ACPI_NFIT_MEM_RESTORE_FAILED ? "restore_fail " : "",
+               flags & ACPI_NFIT_MEM_FLUSH_FAILED ? "flush_fail " : "",
+               flags & ACPI_NFIT_MEM_NOT_ARMED ? "not_armed " : "",
+               flags & ACPI_NFIT_MEM_HEALTH_OBSERVED ? "smart_event " : "");
+}
+static DEVICE_ATTR_RO(flags);
+
+static ssize_t id_show(struct device *dev,
+               struct device_attribute *attr, char *buf)
+{
+       struct acpi_nfit_control_region *dcr = to_nfit_dcr(dev);
+
+       if (dcr->valid_fields & ACPI_NFIT_CONTROL_MFG_INFO_VALID)
+               return sprintf(buf, "%04x-%02x-%04x-%08x\n",
+                               be16_to_cpu(dcr->vendor_id),
+                               dcr->manufacturing_location,
+                               be16_to_cpu(dcr->manufacturing_date),
+                               be32_to_cpu(dcr->serial_number));
+       else
+               return sprintf(buf, "%04x-%08x\n",
+                               be16_to_cpu(dcr->vendor_id),
+                               be32_to_cpu(dcr->serial_number));
+}
+static DEVICE_ATTR_RO(id);
+
+static struct attribute *acpi_nfit_dimm_attributes[] = {
+       &dev_attr_handle.attr,
+       &dev_attr_phys_id.attr,
+       &dev_attr_vendor.attr,
+       &dev_attr_device.attr,
+       &dev_attr_rev_id.attr,
+       &dev_attr_subsystem_vendor.attr,
+       &dev_attr_subsystem_device.attr,
+       &dev_attr_subsystem_rev_id.attr,
+       &dev_attr_format.attr,
+       &dev_attr_formats.attr,
+       &dev_attr_format1.attr,
+       &dev_attr_serial.attr,
+       &dev_attr_flags.attr,
+       &dev_attr_id.attr,
+       &dev_attr_family.attr,
+       &dev_attr_dsm_mask.attr,
+       NULL,
+};
+
+static umode_t acpi_nfit_dimm_attr_visible(struct kobject *kobj,
+               struct attribute *a, int n)
+{
+       struct device *dev = container_of(kobj, struct device, kobj);
+       struct nvdimm *nvdimm = to_nvdimm(dev);
+
+       if (!to_nfit_dcr(dev))
+               return 0;
+       if (a == &dev_attr_format1.attr && num_nvdimm_formats(nvdimm) <= 1)
+               return 0;
+       return a->mode;
+}
+
+static struct attribute_group acpi_nfit_dimm_attribute_group = {
+       .name = "nfit",
+       .attrs = acpi_nfit_dimm_attributes,
+       .is_visible = acpi_nfit_dimm_attr_visible,
+};
+
+static const struct attribute_group *acpi_nfit_dimm_attribute_groups[] = {
+       &nvdimm_attribute_group,
+       &nd_device_attribute_group,
+       &acpi_nfit_dimm_attribute_group,
+       NULL,
+};
+
+static struct nvdimm *acpi_nfit_dimm_by_handle(struct acpi_nfit_desc *acpi_desc,
+               u32 device_handle)
+{
+       struct nfit_mem *nfit_mem;
+
+       list_for_each_entry(nfit_mem, &acpi_desc->dimms, list)
+               if (__to_nfit_memdev(nfit_mem)->device_handle == device_handle)
+                       return nfit_mem->nvdimm;
+
+       return NULL;
+}
+
+static int acpi_nfit_add_dimm(struct acpi_nfit_desc *acpi_desc,
+               struct nfit_mem *nfit_mem, u32 device_handle)
+{
+       struct acpi_device *adev, *adev_dimm;
+       struct device *dev = acpi_desc->dev;
+       unsigned long dsm_mask;
+       const u8 *uuid;
+       int i;
+
+       /* nfit test assumes 1:1 relationship between commands and dsms */
+       nfit_mem->dsm_mask = acpi_desc->dimm_cmd_force_en;
+       nfit_mem->family = NVDIMM_FAMILY_INTEL;
+       adev = to_acpi_dev(acpi_desc);
+       if (!adev)
+               return 0;
+
+       adev_dimm = acpi_find_child_device(adev, device_handle, false);
+       nfit_mem->adev = adev_dimm;
+       if (!adev_dimm) {
+               dev_err(dev, "no ACPI.NFIT device with _ADR %#x, disabling...\n",
+                               device_handle);
+               return force_enable_dimms ? 0 : -ENODEV;
+       }
+
+       /*
+        * Until standardization materializes we need to consider 4
+        * different command sets.  Note, that checking for function0 (bit0)
+        * tells us if any commands are reachable through this uuid.
+        */
+       for (i = NVDIMM_FAMILY_INTEL; i <= NVDIMM_FAMILY_MSFT; i++)
+               if (acpi_check_dsm(adev_dimm->handle, to_nfit_uuid(i), 1, 1))
+                       break;
+
+       /* limit the supported commands to those that are publicly documented */
+       nfit_mem->family = i;
+       if (nfit_mem->family == NVDIMM_FAMILY_INTEL) {
+               dsm_mask = 0x3fe;
+               if (disable_vendor_specific)
+                       dsm_mask &= ~(1 << ND_CMD_VENDOR);
+       } else if (nfit_mem->family == NVDIMM_FAMILY_HPE1) {
+               dsm_mask = 0x1c3c76;
+       } else if (nfit_mem->family == NVDIMM_FAMILY_HPE2) {
+               dsm_mask = 0x1fe;
+               if (disable_vendor_specific)
+                       dsm_mask &= ~(1 << 8);
+       } else if (nfit_mem->family == NVDIMM_FAMILY_MSFT) {
+               dsm_mask = 0xffffffff;
+       } else {
+               dev_err(dev, "unknown dimm command family\n");
+               nfit_mem->family = -1;
+               return force_enable_dimms ? 0 : -ENODEV;
+       }
+
+       uuid = to_nfit_uuid(nfit_mem->family);
+       for_each_set_bit(i, &dsm_mask, BITS_PER_LONG)
+               if (acpi_check_dsm(adev_dimm->handle, uuid, 1, 1ULL << i))
+                       set_bit(i, &nfit_mem->dsm_mask);
+
+       return 0;
+}
+
+static int acpi_nfit_register_dimms(struct acpi_nfit_desc *acpi_desc)
+{
+       struct nfit_mem *nfit_mem;
+       int dimm_count = 0;
+
+       list_for_each_entry(nfit_mem, &acpi_desc->dimms, list) {
+               struct acpi_nfit_flush_address *flush;
+               unsigned long flags = 0, cmd_mask;
+               struct nvdimm *nvdimm;
+               u32 device_handle;
+               u16 mem_flags;
+               int rc;
+
+               device_handle = __to_nfit_memdev(nfit_mem)->device_handle;
+               nvdimm = acpi_nfit_dimm_by_handle(acpi_desc, device_handle);
+               if (nvdimm) {
+                       dimm_count++;
+                       continue;
+               }
+
+               if (nfit_mem->bdw && nfit_mem->memdev_pmem)
+                       flags |= NDD_ALIASING;
+
+               mem_flags = __to_nfit_memdev(nfit_mem)->flags;
+               if (mem_flags & ACPI_NFIT_MEM_NOT_ARMED)
+                       flags |= NDD_UNARMED;
+
+               rc = acpi_nfit_add_dimm(acpi_desc, nfit_mem, device_handle);
+               if (rc)
+                       continue;
+
+               /*
+                * TODO: provide translation for non-NVDIMM_FAMILY_INTEL
+                * devices (i.e. from nd_cmd to acpi_dsm) to standardize the
+                * userspace interface.
+                */
+               cmd_mask = 1UL << ND_CMD_CALL;
+               if (nfit_mem->family == NVDIMM_FAMILY_INTEL)
+                       cmd_mask |= nfit_mem->dsm_mask;
+
+               flush = nfit_mem->nfit_flush ? nfit_mem->nfit_flush->flush
+                       : NULL;
+               nvdimm = nvdimm_create(acpi_desc->nvdimm_bus, nfit_mem,
+                               acpi_nfit_dimm_attribute_groups,
+                               flags, cmd_mask, flush ? flush->hint_count : 0,
+                               nfit_mem->flush_wpq);
+               if (!nvdimm)
+                       return -ENOMEM;
+
+               nfit_mem->nvdimm = nvdimm;
+               dimm_count++;
+
+               if ((mem_flags & ACPI_NFIT_MEM_FAILED_MASK) == 0)
+                       continue;
+
+               dev_info(acpi_desc->dev, "%s flags:%s%s%s%s\n",
+                               nvdimm_name(nvdimm),
+                 mem_flags & ACPI_NFIT_MEM_SAVE_FAILED ? " save_fail" : "",
+                 mem_flags & ACPI_NFIT_MEM_RESTORE_FAILED ? " restore_fail":"",
+                 mem_flags & ACPI_NFIT_MEM_FLUSH_FAILED ? " flush_fail" : "",
+                 mem_flags & ACPI_NFIT_MEM_NOT_ARMED ? " not_armed" : "");
+
+       }
+
+       return nvdimm_bus_check_dimm_count(acpi_desc->nvdimm_bus, dimm_count);
+}
+
+static void acpi_nfit_init_dsms(struct acpi_nfit_desc *acpi_desc)
+{
+       struct nvdimm_bus_descriptor *nd_desc = &acpi_desc->nd_desc;
+       const u8 *uuid = to_nfit_uuid(NFIT_DEV_BUS);
+       struct acpi_device *adev;
+       int i;
+
+       nd_desc->cmd_mask = acpi_desc->bus_cmd_force_en;
+       adev = to_acpi_dev(acpi_desc);
+       if (!adev)
+               return;
+
+       for (i = ND_CMD_ARS_CAP; i <= ND_CMD_CLEAR_ERROR; i++)
+               if (acpi_check_dsm(adev->handle, uuid, 1, 1ULL << i))
+                       set_bit(i, &nd_desc->cmd_mask);
+}
+
+static ssize_t range_index_show(struct device *dev,
+               struct device_attribute *attr, char *buf)
+{
+       struct nd_region *nd_region = to_nd_region(dev);
+       struct nfit_spa *nfit_spa = nd_region_provider_data(nd_region);
+
+       return sprintf(buf, "%d\n", nfit_spa->spa->range_index);
+}
+static DEVICE_ATTR_RO(range_index);
+
+static struct attribute *acpi_nfit_region_attributes[] = {
+       &dev_attr_range_index.attr,
+       NULL,
+};
+
+static struct attribute_group acpi_nfit_region_attribute_group = {
+       .name = "nfit",
+       .attrs = acpi_nfit_region_attributes,
+};
+
+static const struct attribute_group *acpi_nfit_region_attribute_groups[] = {
+       &nd_region_attribute_group,
+       &nd_mapping_attribute_group,
+       &nd_device_attribute_group,
+       &nd_numa_attribute_group,
+       &acpi_nfit_region_attribute_group,
+       NULL,
+};
+
+/* enough info to uniquely specify an interleave set */
+struct nfit_set_info {
+       struct nfit_set_info_map {
+               u64 region_offset;
+               u32 serial_number;
+               u32 pad;
+       } mapping[0];
+};
+
+static size_t sizeof_nfit_set_info(int num_mappings)
+{
+       return sizeof(struct nfit_set_info)
+               + num_mappings * sizeof(struct nfit_set_info_map);
+}
+
+static int cmp_map(const void *m0, const void *m1)
+{
+       const struct nfit_set_info_map *map0 = m0;
+       const struct nfit_set_info_map *map1 = m1;
+
+       return memcmp(&map0->region_offset, &map1->region_offset,
+                       sizeof(u64));
+}
+
+/* Retrieve the nth entry referencing this spa */
+static struct acpi_nfit_memory_map *memdev_from_spa(
+               struct acpi_nfit_desc *acpi_desc, u16 range_index, int n)
+{
+       struct nfit_memdev *nfit_memdev;
+
+       list_for_each_entry(nfit_memdev, &acpi_desc->memdevs, list)
+               if (nfit_memdev->memdev->range_index == range_index)
+                       if (n-- == 0)
+                               return nfit_memdev->memdev;
+       return NULL;
+}
+
+static int acpi_nfit_init_interleave_set(struct acpi_nfit_desc *acpi_desc,
+               struct nd_region_desc *ndr_desc,
+               struct acpi_nfit_system_address *spa)
+{
+       int i, spa_type = nfit_spa_type(spa);
+       struct device *dev = acpi_desc->dev;
+       struct nd_interleave_set *nd_set;
+       u16 nr = ndr_desc->num_mappings;
+       struct nfit_set_info *info;
+
+       if (spa_type == NFIT_SPA_PM || spa_type == NFIT_SPA_VOLATILE)
+               /* pass */;
+       else
+               return 0;
+
+       nd_set = devm_kzalloc(dev, sizeof(*nd_set), GFP_KERNEL);
+       if (!nd_set)
+               return -ENOMEM;
+
+       info = devm_kzalloc(dev, sizeof_nfit_set_info(nr), GFP_KERNEL);
+       if (!info)
+               return -ENOMEM;
+       for (i = 0; i < nr; i++) {
+               struct nd_mapping *nd_mapping = &ndr_desc->nd_mapping[i];
+               struct nfit_set_info_map *map = &info->mapping[i];
+               struct nvdimm *nvdimm = nd_mapping->nvdimm;
+               struct nfit_mem *nfit_mem = nvdimm_provider_data(nvdimm);
+               struct acpi_nfit_memory_map *memdev = memdev_from_spa(acpi_desc,
+                               spa->range_index, i);
+
+               if (!memdev || !nfit_mem->dcr) {
+                       dev_err(dev, "%s: failed to find DCR\n", __func__);
+                       return -ENODEV;
+               }
+
+               map->region_offset = memdev->region_offset;
+               map->serial_number = nfit_mem->dcr->serial_number;
+       }
+
+       sort(&info->mapping[0], nr, sizeof(struct nfit_set_info_map),
+                       cmp_map, NULL);
+       nd_set->cookie = nd_fletcher64(info, sizeof_nfit_set_info(nr), 0);
+       ndr_desc->nd_set = nd_set;
+       devm_kfree(dev, info);
+
+       return 0;
+}
+
+static u64 to_interleave_offset(u64 offset, struct nfit_blk_mmio *mmio)
+{
+       struct acpi_nfit_interleave *idt = mmio->idt;
+       u32 sub_line_offset, line_index, line_offset;
+       u64 line_no, table_skip_count, table_offset;
+
+       line_no = div_u64_rem(offset, mmio->line_size, &sub_line_offset);
+       table_skip_count = div_u64_rem(line_no, mmio->num_lines, &line_index);
+       line_offset = idt->line_offset[line_index]
+               * mmio->line_size;
+       table_offset = table_skip_count * mmio->table_size;
+
+       return mmio->base_offset + line_offset + table_offset + sub_line_offset;
+}
+
+static u32 read_blk_stat(struct nfit_blk *nfit_blk, unsigned int bw)
+{
+       struct nfit_blk_mmio *mmio = &nfit_blk->mmio[DCR];
+       u64 offset = nfit_blk->stat_offset + mmio->size * bw;
+
+       if (mmio->num_lines)
+               offset = to_interleave_offset(offset, mmio);
+
+       return readl(mmio->addr.base + offset);
+}
+
+static void write_blk_ctl(struct nfit_blk *nfit_blk, unsigned int bw,
+               resource_size_t dpa, unsigned int len, unsigned int write)
+{
+       u64 cmd, offset;
+       struct nfit_blk_mmio *mmio = &nfit_blk->mmio[DCR];
+
+       enum {
+               BCW_OFFSET_MASK = (1ULL << 48)-1,
+               BCW_LEN_SHIFT = 48,
+               BCW_LEN_MASK = (1ULL << 8) - 1,
+               BCW_CMD_SHIFT = 56,
+       };
+
+       cmd = (dpa >> L1_CACHE_SHIFT) & BCW_OFFSET_MASK;
+       len = len >> L1_CACHE_SHIFT;
+       cmd |= ((u64) len & BCW_LEN_MASK) << BCW_LEN_SHIFT;
+       cmd |= ((u64) write) << BCW_CMD_SHIFT;
+
+       offset = nfit_blk->cmd_offset + mmio->size * bw;
+       if (mmio->num_lines)
+               offset = to_interleave_offset(offset, mmio);
+
+       writeq(cmd, mmio->addr.base + offset);
+       nvdimm_flush(nfit_blk->nd_region);
+
+       if (nfit_blk->dimm_flags & NFIT_BLK_DCR_LATCH)
+               readq(mmio->addr.base + offset);
+}
+
+static int acpi_nfit_blk_single_io(struct nfit_blk *nfit_blk,
+               resource_size_t dpa, void *iobuf, size_t len, int rw,
+               unsigned int lane)
+{
+       struct nfit_blk_mmio *mmio = &nfit_blk->mmio[BDW];
+       unsigned int copied = 0;
+       u64 base_offset;
+       int rc;
+
+       base_offset = nfit_blk->bdw_offset + dpa % L1_CACHE_BYTES
+               + lane * mmio->size;
+       write_blk_ctl(nfit_blk, lane, dpa, len, rw);
+       while (len) {
+               unsigned int c;
+               u64 offset;
+
+               if (mmio->num_lines) {
+                       u32 line_offset;
+
+                       offset = to_interleave_offset(base_offset + copied,
+                                       mmio);
+                       div_u64_rem(offset, mmio->line_size, &line_offset);
+                       c = min_t(size_t, len, mmio->line_size - line_offset);
+               } else {
+                       offset = base_offset + nfit_blk->bdw_offset;
+                       c = len;
+               }
+
+               if (rw)
+                       memcpy_to_pmem(mmio->addr.aperture + offset,
+                                       iobuf + copied, c);
+               else {
+                       if (nfit_blk->dimm_flags & NFIT_BLK_READ_FLUSH)
+                               mmio_flush_range((void __force *)
+                                       mmio->addr.aperture + offset, c);
+
+                       memcpy_from_pmem(iobuf + copied,
+                                       mmio->addr.aperture + offset, c);
+               }
+
+               copied += c;
+               len -= c;
+       }
+
+       if (rw)
+               nvdimm_flush(nfit_blk->nd_region);
+
+       rc = read_blk_stat(nfit_blk, lane) ? -EIO : 0;
+       return rc;
+}
+
+static int acpi_nfit_blk_region_do_io(struct nd_blk_region *ndbr,
+               resource_size_t dpa, void *iobuf, u64 len, int rw)
+{
+       struct nfit_blk *nfit_blk = nd_blk_region_provider_data(ndbr);
+       struct nfit_blk_mmio *mmio = &nfit_blk->mmio[BDW];
+       struct nd_region *nd_region = nfit_blk->nd_region;
+       unsigned int lane, copied = 0;
+       int rc = 0;
+
+       lane = nd_region_acquire_lane(nd_region);
+       while (len) {
+               u64 c = min(len, mmio->size);
+
+               rc = acpi_nfit_blk_single_io(nfit_blk, dpa + copied,
+                               iobuf + copied, c, rw, lane);
+               if (rc)
+                       break;
+
+               copied += c;
+               len -= c;
+       }
+       nd_region_release_lane(nd_region, lane);
+
+       return rc;
+}
+
+static int nfit_blk_init_interleave(struct nfit_blk_mmio *mmio,
+               struct acpi_nfit_interleave *idt, u16 interleave_ways)
+{
+       if (idt) {
+               mmio->num_lines = idt->line_count;
+               mmio->line_size = idt->line_size;
+               if (interleave_ways == 0)
+                       return -ENXIO;
+               mmio->table_size = mmio->num_lines * interleave_ways
+                       * mmio->line_size;
+       }
+
+       return 0;
+}
+
+static int acpi_nfit_blk_get_flags(struct nvdimm_bus_descriptor *nd_desc,
+               struct nvdimm *nvdimm, struct nfit_blk *nfit_blk)
+{
+       struct nd_cmd_dimm_flags flags;
+       int rc;
+
+       memset(&flags, 0, sizeof(flags));
+       rc = nd_desc->ndctl(nd_desc, nvdimm, ND_CMD_DIMM_FLAGS, &flags,
+                       sizeof(flags), NULL);
+
+       if (rc >= 0 && flags.status == 0)
+               nfit_blk->dimm_flags = flags.flags;
+       else if (rc == -ENOTTY) {
+               /* fall back to a conservative default */
+               nfit_blk->dimm_flags = NFIT_BLK_DCR_LATCH | NFIT_BLK_READ_FLUSH;
+               rc = 0;
+       } else
+               rc = -ENXIO;
+
+       return rc;
+}
+
+static int acpi_nfit_blk_region_enable(struct nvdimm_bus *nvdimm_bus,
+               struct device *dev)
+{
+       struct nvdimm_bus_descriptor *nd_desc = to_nd_desc(nvdimm_bus);
+       struct nd_blk_region *ndbr = to_nd_blk_region(dev);
+       struct nfit_blk_mmio *mmio;
+       struct nfit_blk *nfit_blk;
+       struct nfit_mem *nfit_mem;
+       struct nvdimm *nvdimm;
+       int rc;
+
+       nvdimm = nd_blk_region_to_dimm(ndbr);
+       nfit_mem = nvdimm_provider_data(nvdimm);
+       if (!nfit_mem || !nfit_mem->dcr || !nfit_mem->bdw) {
+               dev_dbg(dev, "%s: missing%s%s%s\n", __func__,
+                               nfit_mem ? "" : " nfit_mem",
+                               (nfit_mem && nfit_mem->dcr) ? "" : " dcr",
+                               (nfit_mem && nfit_mem->bdw) ? "" : " bdw");
+               return -ENXIO;
+       }
+
+       nfit_blk = devm_kzalloc(dev, sizeof(*nfit_blk), GFP_KERNEL);
+       if (!nfit_blk)
+               return -ENOMEM;
+       nd_blk_region_set_provider_data(ndbr, nfit_blk);
+       nfit_blk->nd_region = to_nd_region(dev);
+
+       /* map block aperture memory */
+       nfit_blk->bdw_offset = nfit_mem->bdw->offset;
+       mmio = &nfit_blk->mmio[BDW];
+       mmio->addr.base = devm_nvdimm_memremap(dev, nfit_mem->spa_bdw->address,
+                        nfit_mem->spa_bdw->length, ARCH_MEMREMAP_PMEM);
+       if (!mmio->addr.base) {
+               dev_dbg(dev, "%s: %s failed to map bdw\n", __func__,
+                               nvdimm_name(nvdimm));
+               return -ENOMEM;
+       }
+       mmio->size = nfit_mem->bdw->size;
+       mmio->base_offset = nfit_mem->memdev_bdw->region_offset;
+       mmio->idt = nfit_mem->idt_bdw;
+       mmio->spa = nfit_mem->spa_bdw;
+       rc = nfit_blk_init_interleave(mmio, nfit_mem->idt_bdw,
+                       nfit_mem->memdev_bdw->interleave_ways);
+       if (rc) {
+               dev_dbg(dev, "%s: %s failed to init bdw interleave\n",
+                               __func__, nvdimm_name(nvdimm));
+               return rc;
+       }
+
+       /* map block control memory */
+       nfit_blk->cmd_offset = nfit_mem->dcr->command_offset;
+       nfit_blk->stat_offset = nfit_mem->dcr->status_offset;
+       mmio = &nfit_blk->mmio[DCR];
+       mmio->addr.base = devm_nvdimm_ioremap(dev, nfit_mem->spa_dcr->address,
+                       nfit_mem->spa_dcr->length);
+       if (!mmio->addr.base) {
+               dev_dbg(dev, "%s: %s failed to map dcr\n", __func__,
+                               nvdimm_name(nvdimm));
+               return -ENOMEM;
+       }
+       mmio->size = nfit_mem->dcr->window_size;
+       mmio->base_offset = nfit_mem->memdev_dcr->region_offset;
+       mmio->idt = nfit_mem->idt_dcr;
+       mmio->spa = nfit_mem->spa_dcr;
+       rc = nfit_blk_init_interleave(mmio, nfit_mem->idt_dcr,
+                       nfit_mem->memdev_dcr->interleave_ways);
+       if (rc) {
+               dev_dbg(dev, "%s: %s failed to init dcr interleave\n",
+                               __func__, nvdimm_name(nvdimm));
+               return rc;
+       }
+
+       rc = acpi_nfit_blk_get_flags(nd_desc, nvdimm, nfit_blk);
+       if (rc < 0) {
+               dev_dbg(dev, "%s: %s failed get DIMM flags\n",
+                               __func__, nvdimm_name(nvdimm));
+               return rc;
+       }
+
+       if (nvdimm_has_flush(nfit_blk->nd_region) < 0)
+               dev_warn(dev, "unable to guarantee persistence of writes\n");
+
+       if (mmio->line_size == 0)
+               return 0;
+
+       if ((u32) nfit_blk->cmd_offset % mmio->line_size
+                       + 8 > mmio->line_size) {
+               dev_dbg(dev, "cmd_offset crosses interleave boundary\n");
+               return -ENXIO;
+       } else if ((u32) nfit_blk->stat_offset % mmio->line_size
+                       + 8 > mmio->line_size) {
+               dev_dbg(dev, "stat_offset crosses interleave boundary\n");
+               return -ENXIO;
+       }
+
+       return 0;
+}
+
+static int ars_get_cap(struct acpi_nfit_desc *acpi_desc,
+               struct nd_cmd_ars_cap *cmd, struct nfit_spa *nfit_spa)
+{
+       struct nvdimm_bus_descriptor *nd_desc = &acpi_desc->nd_desc;
+       struct acpi_nfit_system_address *spa = nfit_spa->spa;
+       int cmd_rc, rc;
+
+       cmd->address = spa->address;
+       cmd->length = spa->length;
+       rc = nd_desc->ndctl(nd_desc, NULL, ND_CMD_ARS_CAP, cmd,
+                       sizeof(*cmd), &cmd_rc);
+       if (rc < 0)
+               return rc;
+       return cmd_rc;
+}
+
+static int ars_start(struct acpi_nfit_desc *acpi_desc, struct nfit_spa *nfit_spa)
+{
+       int rc;
+       int cmd_rc;
+       struct nd_cmd_ars_start ars_start;
+       struct acpi_nfit_system_address *spa = nfit_spa->spa;
+       struct nvdimm_bus_descriptor *nd_desc = &acpi_desc->nd_desc;
+
+       memset(&ars_start, 0, sizeof(ars_start));
+       ars_start.address = spa->address;
+       ars_start.length = spa->length;
+       if (nfit_spa_type(spa) == NFIT_SPA_PM)
+               ars_start.type = ND_ARS_PERSISTENT;
+       else if (nfit_spa_type(spa) == NFIT_SPA_VOLATILE)
+               ars_start.type = ND_ARS_VOLATILE;
+       else
+               return -ENOTTY;
+
+       rc = nd_desc->ndctl(nd_desc, NULL, ND_CMD_ARS_START, &ars_start,
+                       sizeof(ars_start), &cmd_rc);
+
+       if (rc < 0)
+               return rc;
+       return cmd_rc;
+}
+
+static int ars_continue(struct acpi_nfit_desc *acpi_desc)
+{
+       int rc, cmd_rc;
+       struct nd_cmd_ars_start ars_start;
+       struct nvdimm_bus_descriptor *nd_desc = &acpi_desc->nd_desc;
+       struct nd_cmd_ars_status *ars_status = acpi_desc->ars_status;
+
+       memset(&ars_start, 0, sizeof(ars_start));
+       ars_start.address = ars_status->restart_address;
+       ars_start.length = ars_status->restart_length;
+       ars_start.type = ars_status->type;
+       rc = nd_desc->ndctl(nd_desc, NULL, ND_CMD_ARS_START, &ars_start,
+                       sizeof(ars_start), &cmd_rc);
+       if (rc < 0)
+               return rc;
+       return cmd_rc;
+}
+
+static int ars_get_status(struct acpi_nfit_desc *acpi_desc)
+{
+       struct nvdimm_bus_descriptor *nd_desc = &acpi_desc->nd_desc;
+       struct nd_cmd_ars_status *ars_status = acpi_desc->ars_status;
+       int rc, cmd_rc;
+
+       rc = nd_desc->ndctl(nd_desc, NULL, ND_CMD_ARS_STATUS, ars_status,
+                       acpi_desc->ars_status_size, &cmd_rc);
+       if (rc < 0)
+               return rc;
+       return cmd_rc;
+}
+
+static int ars_status_process_records(struct nvdimm_bus *nvdimm_bus,
+               struct nd_cmd_ars_status *ars_status)
+{
+       int rc;
+       u32 i;
+
+       for (i = 0; i < ars_status->num_records; i++) {
+               rc = nvdimm_bus_add_poison(nvdimm_bus,
+                               ars_status->records[i].err_address,
+                               ars_status->records[i].length);
+               if (rc)
+                       return rc;
+       }
+
+       return 0;
+}
+
+static void acpi_nfit_remove_resource(void *data)
+{
+       struct resource *res = data;
+
+       remove_resource(res);
+}
+
+static int acpi_nfit_insert_resource(struct acpi_nfit_desc *acpi_desc,
+               struct nd_region_desc *ndr_desc)
+{
+       struct resource *res, *nd_res = ndr_desc->res;
+       int is_pmem, ret;
+
+       /* No operation if the region is already registered as PMEM */
+       is_pmem = region_intersects(nd_res->start, resource_size(nd_res),
+                               IORESOURCE_MEM, IORES_DESC_PERSISTENT_MEMORY);
+       if (is_pmem == REGION_INTERSECTS)
+               return 0;
+
+       res = devm_kzalloc(acpi_desc->dev, sizeof(*res), GFP_KERNEL);
+       if (!res)
+               return -ENOMEM;
+
+       res->name = "Persistent Memory";
+       res->start = nd_res->start;
+       res->end = nd_res->end;
+       res->flags = IORESOURCE_MEM;
+       res->desc = IORES_DESC_PERSISTENT_MEMORY;
+
+       ret = insert_resource(&iomem_resource, res);
+       if (ret)
+               return ret;
+
+       ret = devm_add_action_or_reset(acpi_desc->dev,
+                                       acpi_nfit_remove_resource,
+                                       res);
+       if (ret)
+               return ret;
+
+       return 0;
+}
+
+static int acpi_nfit_init_mapping(struct acpi_nfit_desc *acpi_desc,
+               struct nd_mapping *nd_mapping, struct nd_region_desc *ndr_desc,
+               struct acpi_nfit_memory_map *memdev,
+               struct nfit_spa *nfit_spa)
+{
+       struct nvdimm *nvdimm = acpi_nfit_dimm_by_handle(acpi_desc,
+                       memdev->device_handle);
+       struct acpi_nfit_system_address *spa = nfit_spa->spa;
+       struct nd_blk_region_desc *ndbr_desc;
+       struct nfit_mem *nfit_mem;
+       int blk_valid = 0;
+
+       if (!nvdimm) {
+               dev_err(acpi_desc->dev, "spa%d dimm: %#x not found\n",
+                               spa->range_index, memdev->device_handle);
+               return -ENODEV;
+       }
+
+       nd_mapping->nvdimm = nvdimm;
+       switch (nfit_spa_type(spa)) {
+       case NFIT_SPA_PM:
+       case NFIT_SPA_VOLATILE:
+               nd_mapping->start = memdev->address;
+               nd_mapping->size = memdev->region_size;
+               break;
+       case NFIT_SPA_DCR:
+               nfit_mem = nvdimm_provider_data(nvdimm);
+               if (!nfit_mem || !nfit_mem->bdw) {
+                       dev_dbg(acpi_desc->dev, "spa%d %s missing bdw\n",
+                                       spa->range_index, nvdimm_name(nvdimm));
+               } else {
+                       nd_mapping->size = nfit_mem->bdw->capacity;
+                       nd_mapping->start = nfit_mem->bdw->start_address;
+                       ndr_desc->num_lanes = nfit_mem->bdw->windows;
+                       blk_valid = 1;
+               }
+
+               ndr_desc->nd_mapping = nd_mapping;
+               ndr_desc->num_mappings = blk_valid;
+               ndbr_desc = to_blk_region_desc(ndr_desc);
+               ndbr_desc->enable = acpi_nfit_blk_region_enable;
+               ndbr_desc->do_io = acpi_desc->blk_do_io;
+               nfit_spa->nd_region = nvdimm_blk_region_create(acpi_desc->nvdimm_bus,
+                               ndr_desc);
+               if (!nfit_spa->nd_region)
+                       return -ENOMEM;
+               break;
+       }
+
+       return 0;
+}
+
+static bool nfit_spa_is_virtual(struct acpi_nfit_system_address *spa)
+{
+       return (nfit_spa_type(spa) == NFIT_SPA_VDISK ||
+               nfit_spa_type(spa) == NFIT_SPA_VCD   ||
+               nfit_spa_type(spa) == NFIT_SPA_PDISK ||
+               nfit_spa_type(spa) == NFIT_SPA_PCD);
+}
+
+static int acpi_nfit_register_region(struct acpi_nfit_desc *acpi_desc,
+               struct nfit_spa *nfit_spa)
+{
+       static struct nd_mapping nd_mappings[ND_MAX_MAPPINGS];
+       struct acpi_nfit_system_address *spa = nfit_spa->spa;
+       struct nd_blk_region_desc ndbr_desc;
+       struct nd_region_desc *ndr_desc;
+       struct nfit_memdev *nfit_memdev;
+       struct nvdimm_bus *nvdimm_bus;
+       struct resource res;
+       int count = 0, rc;
+
+       if (nfit_spa->nd_region)
+               return 0;
+
+       if (spa->range_index == 0 && !nfit_spa_is_virtual(spa)) {
+               dev_dbg(acpi_desc->dev, "%s: detected invalid spa index\n",
+                               __func__);
+               return 0;
+       }
+
+       memset(&res, 0, sizeof(res));
+       memset(&nd_mappings, 0, sizeof(nd_mappings));
+       memset(&ndbr_desc, 0, sizeof(ndbr_desc));
+       res.start = spa->address;
+       res.end = res.start + spa->length - 1;
+       ndr_desc = &ndbr_desc.ndr_desc;
+       ndr_desc->res = &res;
+       ndr_desc->provider_data = nfit_spa;
+       ndr_desc->attr_groups = acpi_nfit_region_attribute_groups;
+       if (spa->flags & ACPI_NFIT_PROXIMITY_VALID)
+               ndr_desc->numa_node = acpi_map_pxm_to_online_node(
+                                               spa->proximity_domain);
+       else
+               ndr_desc->numa_node = NUMA_NO_NODE;
+
+       list_for_each_entry(nfit_memdev, &acpi_desc->memdevs, list) {
+               struct acpi_nfit_memory_map *memdev = nfit_memdev->memdev;
+               struct nd_mapping *nd_mapping;
+
+               if (memdev->range_index != spa->range_index)
+                       continue;
+               if (count >= ND_MAX_MAPPINGS) {
+                       dev_err(acpi_desc->dev, "spa%d exceeds max mappings %d\n",
+                                       spa->range_index, ND_MAX_MAPPINGS);
+                       return -ENXIO;
+               }
+               nd_mapping = &nd_mappings[count++];
+               rc = acpi_nfit_init_mapping(acpi_desc, nd_mapping, ndr_desc,
+                               memdev, nfit_spa);
+               if (rc)
+                       goto out;
+       }
+
+       ndr_desc->nd_mapping = nd_mappings;
+       ndr_desc->num_mappings = count;
+       rc = acpi_nfit_init_interleave_set(acpi_desc, ndr_desc, spa);
+       if (rc)
+               goto out;
+
+       nvdimm_bus = acpi_desc->nvdimm_bus;
+       if (nfit_spa_type(spa) == NFIT_SPA_PM) {
+               rc = acpi_nfit_insert_resource(acpi_desc, ndr_desc);
+               if (rc) {
+                       dev_warn(acpi_desc->dev,
+                               "failed to insert pmem resource to iomem: %d\n",
+                               rc);
+                       goto out;
+               }
+
+               nfit_spa->nd_region = nvdimm_pmem_region_create(nvdimm_bus,
+                               ndr_desc);
+               if (!nfit_spa->nd_region)
+                       rc = -ENOMEM;
+       } else if (nfit_spa_type(spa) == NFIT_SPA_VOLATILE) {
+               nfit_spa->nd_region = nvdimm_volatile_region_create(nvdimm_bus,
+                               ndr_desc);
+               if (!nfit_spa->nd_region)
+                       rc = -ENOMEM;
+       } else if (nfit_spa_is_virtual(spa)) {
+               nfit_spa->nd_region = nvdimm_pmem_region_create(nvdimm_bus,
+                               ndr_desc);
+               if (!nfit_spa->nd_region)
+                       rc = -ENOMEM;
+       }
+
+ out:
+       if (rc)
+               dev_err(acpi_desc->dev, "failed to register spa range %d\n",
+                               nfit_spa->spa->range_index);
+       return rc;
+}
+
+static int ars_status_alloc(struct acpi_nfit_desc *acpi_desc,
+               u32 max_ars)
+{
+       struct device *dev = acpi_desc->dev;
+       struct nd_cmd_ars_status *ars_status;
+
+       if (acpi_desc->ars_status && acpi_desc->ars_status_size >= max_ars) {
+               memset(acpi_desc->ars_status, 0, acpi_desc->ars_status_size);
+               return 0;
+       }
+
+       if (acpi_desc->ars_status)
+               devm_kfree(dev, acpi_desc->ars_status);
+       acpi_desc->ars_status = NULL;
+       ars_status = devm_kzalloc(dev, max_ars, GFP_KERNEL);
+       if (!ars_status)
+               return -ENOMEM;
+       acpi_desc->ars_status = ars_status;
+       acpi_desc->ars_status_size = max_ars;
+       return 0;
+}
+
+static int acpi_nfit_query_poison(struct acpi_nfit_desc *acpi_desc,
+               struct nfit_spa *nfit_spa)
+{
+       struct acpi_nfit_system_address *spa = nfit_spa->spa;
+       int rc;
+
+       if (!nfit_spa->max_ars) {
+               struct nd_cmd_ars_cap ars_cap;
+
+               memset(&ars_cap, 0, sizeof(ars_cap));
+               rc = ars_get_cap(acpi_desc, &ars_cap, nfit_spa);
+               if (rc < 0)
+                       return rc;
+               nfit_spa->max_ars = ars_cap.max_ars_out;
+               nfit_spa->clear_err_unit = ars_cap.clear_err_unit;
+               /* check that the supported scrub types match the spa type */
+               if (nfit_spa_type(spa) == NFIT_SPA_VOLATILE &&
+                               ((ars_cap.status >> 16) & ND_ARS_VOLATILE) == 0)
+                       return -ENOTTY;
+               else if (nfit_spa_type(spa) == NFIT_SPA_PM &&
+                               ((ars_cap.status >> 16) & ND_ARS_PERSISTENT) == 0)
+                       return -ENOTTY;
+       }
+
+       if (ars_status_alloc(acpi_desc, nfit_spa->max_ars))
+               return -ENOMEM;
+
+       rc = ars_get_status(acpi_desc);
+       if (rc < 0 && rc != -ENOSPC)
+               return rc;
+
+       if (ars_status_process_records(acpi_desc->nvdimm_bus,
+                               acpi_desc->ars_status))
+               return -ENOMEM;
+
+       return 0;
+}
+
+static void acpi_nfit_async_scrub(struct acpi_nfit_desc *acpi_desc,
+               struct nfit_spa *nfit_spa)
+{
+       struct acpi_nfit_system_address *spa = nfit_spa->spa;
+       unsigned int overflow_retry = scrub_overflow_abort;
+       u64 init_ars_start = 0, init_ars_len = 0;
+       struct device *dev = acpi_desc->dev;
+       unsigned int tmo = scrub_timeout;
+       int rc;
+
+       if (!nfit_spa->ars_required || !nfit_spa->nd_region)
+               return;
+
+       rc = ars_start(acpi_desc, nfit_spa);
+       /*
+        * If we timed out the initial scan we'll still be busy here,
+        * and will wait another timeout before giving up permanently.
+        */
+       if (rc < 0 && rc != -EBUSY)
+               return;
+
+       do {
+               u64 ars_start, ars_len;
+
+               if (acpi_desc->cancel)
+                       break;
+               rc = acpi_nfit_query_poison(acpi_desc, nfit_spa);
+               if (rc == -ENOTTY)
+                       break;
+               if (rc == -EBUSY && !tmo) {
+                       dev_warn(dev, "range %d ars timeout, aborting\n",
+                                       spa->range_index);
+                       break;
+               }
+
+               if (rc == -EBUSY) {
+                       /*
+                        * Note, entries may be appended to the list
+                        * while the lock is dropped, but the workqueue
+                        * being active prevents entries being deleted /
+                        * freed.
+                        */
+                       mutex_unlock(&acpi_desc->init_mutex);
+                       ssleep(1);
+                       tmo--;
+                       mutex_lock(&acpi_desc->init_mutex);
+                       continue;
+               }
+
+               /* we got some results, but there are more pending... */
+               if (rc == -ENOSPC && overflow_retry--) {
+                       if (!init_ars_len) {
+                               init_ars_len = acpi_desc->ars_status->length;
+                               init_ars_start = acpi_desc->ars_status->address;
+                       }
+                       rc = ars_continue(acpi_desc);
+               }
+
+               if (rc < 0) {
+                       dev_warn(dev, "range %d ars continuation failed\n",
+                                       spa->range_index);
+                       break;
+               }
+
+               if (init_ars_len) {
+                       ars_start = init_ars_start;
+                       ars_len = init_ars_len;
+               } else {
+                       ars_start = acpi_desc->ars_status->address;
+                       ars_len = acpi_desc->ars_status->length;
+               }
+               dev_dbg(dev, "spa range: %d ars from %#llx + %#llx complete\n",
+                               spa->range_index, ars_start, ars_len);
+               /* notify the region about new poison entries */
+               nvdimm_region_notify(nfit_spa->nd_region,
+                               NVDIMM_REVALIDATE_POISON);
+               break;
+       } while (1);
+}
+
+static void acpi_nfit_scrub(struct work_struct *work)
+{
+       struct device *dev;
+       u64 init_scrub_length = 0;
+       struct nfit_spa *nfit_spa;
+       u64 init_scrub_address = 0;
+       bool init_ars_done = false;
+       struct acpi_nfit_desc *acpi_desc;
+       unsigned int tmo = scrub_timeout;
+       unsigned int overflow_retry = scrub_overflow_abort;
+
+       acpi_desc = container_of(work, typeof(*acpi_desc), work);
+       dev = acpi_desc->dev;
+
+       /*
+        * We scrub in 2 phases.  The first phase waits for any platform
+        * firmware initiated scrubs to complete and then we go search for the
+        * affected spa regions to mark them scanned.  In the second phase we
+        * initiate a directed scrub for every range that was not scrubbed in
+        * phase 1. If we're called for a 'rescan', we harmlessly pass through
+        * the first phase, but really only care about running phase 2, where
+        * regions can be notified of new poison.
+        */
+
+       /* process platform firmware initiated scrubs */
+ retry:
+       mutex_lock(&acpi_desc->init_mutex);
+       list_for_each_entry(nfit_spa, &acpi_desc->spas, list) {
+               struct nd_cmd_ars_status *ars_status;
+               struct acpi_nfit_system_address *spa;
+               u64 ars_start, ars_len;
+               int rc;
+
+               if (acpi_desc->cancel)
+                       break;
+
+               if (nfit_spa->nd_region)
+                       continue;
+
+               if (init_ars_done) {
+                       /*
+                        * No need to re-query, we're now just
+                        * reconciling all the ranges covered by the
+                        * initial scrub
+                        */
+                       rc = 0;
+               } else
+                       rc = acpi_nfit_query_poison(acpi_desc, nfit_spa);
+
+               if (rc == -ENOTTY) {
+                       /* no ars capability, just register spa and move on */
+                       acpi_nfit_register_region(acpi_desc, nfit_spa);
+                       continue;
+               }
+
+               if (rc == -EBUSY && !tmo) {
+                       /* fallthrough to directed scrub in phase 2 */
+                       dev_warn(dev, "timeout awaiting ars results, continuing...\n");
+                       break;
+               } else if (rc == -EBUSY) {
+                       mutex_unlock(&acpi_desc->init_mutex);
+                       ssleep(1);
+                       tmo--;
+                       goto retry;
+               }
+
+               /* we got some results, but there are more pending... */
+               if (rc == -ENOSPC && overflow_retry--) {
+                       ars_status = acpi_desc->ars_status;
+                       /*
+                        * Record the original scrub range, so that we
+                        * can recall all the ranges impacted by the
+                        * initial scrub.
+                        */
+                       if (!init_scrub_length) {
+                               init_scrub_length = ars_status->length;
+                               init_scrub_address = ars_status->address;
+                       }
+                       rc = ars_continue(acpi_desc);
+                       if (rc == 0) {
+                               mutex_unlock(&acpi_desc->init_mutex);
+                               goto retry;
+                       }
+               }
+
+               if (rc < 0) {
+                       /*
+                        * Initial scrub failed, we'll give it one more
+                        * try below...
+                        */
+                       break;
+               }
+
+               /* We got some final results, record completed ranges */
+               ars_status = acpi_desc->ars_status;
+               if (init_scrub_length) {
+                       ars_start = init_scrub_address;
+                       ars_len = ars_start + init_scrub_length;
+               } else {
+                       ars_start = ars_status->address;
+                       ars_len = ars_status->length;
+               }
+               spa = nfit_spa->spa;
+
+               if (!init_ars_done) {
+                       init_ars_done = true;
+                       dev_dbg(dev, "init scrub %#llx + %#llx complete\n",
+                                       ars_start, ars_len);
+               }
+               if (ars_start <= spa->address && ars_start + ars_len
+                               >= spa->address + spa->length)
+                       acpi_nfit_register_region(acpi_desc, nfit_spa);
+       }
+
+       /*
+        * For all the ranges not covered by an initial scrub we still
+        * want to see if there are errors, but it's ok to discover them
+        * asynchronously.
+        */
+       list_for_each_entry(nfit_spa, &acpi_desc->spas, list) {
+               /*
+                * Flag all the ranges that still need scrubbing, but
+                * register them now to make data available.
+                */
+               if (!nfit_spa->nd_region) {
+                       nfit_spa->ars_required = 1;
+                       acpi_nfit_register_region(acpi_desc, nfit_spa);
+               }
+       }
+
+       list_for_each_entry(nfit_spa, &acpi_desc->spas, list)
+               acpi_nfit_async_scrub(acpi_desc, nfit_spa);
+       acpi_desc->scrub_count++;
+       if (acpi_desc->scrub_count_state)
+               sysfs_notify_dirent(acpi_desc->scrub_count_state);
+       mutex_unlock(&acpi_desc->init_mutex);
+}
+
+static int acpi_nfit_register_regions(struct acpi_nfit_desc *acpi_desc)
+{
+       struct nfit_spa *nfit_spa;
+       int rc;
+
+       list_for_each_entry(nfit_spa, &acpi_desc->spas, list)
+               if (nfit_spa_type(nfit_spa->spa) == NFIT_SPA_DCR) {
+                       /* BLK regions don't need to wait for ars results */
+                       rc = acpi_nfit_register_region(acpi_desc, nfit_spa);
+                       if (rc)
+                               return rc;
+               }
+
+       queue_work(nfit_wq, &acpi_desc->work);
+       return 0;
+}
+
+static int acpi_nfit_check_deletions(struct acpi_nfit_desc *acpi_desc,
+               struct nfit_table_prev *prev)
+{
+       struct device *dev = acpi_desc->dev;
+
+       if (!list_empty(&prev->spas) ||
+                       !list_empty(&prev->memdevs) ||
+                       !list_empty(&prev->dcrs) ||
+                       !list_empty(&prev->bdws) ||
+                       !list_empty(&prev->idts) ||
+                       !list_empty(&prev->flushes)) {
+               dev_err(dev, "new nfit deletes entries (unsupported)\n");
+               return -ENXIO;
+       }
+       return 0;
+}
+
+static int acpi_nfit_desc_init_scrub_attr(struct acpi_nfit_desc *acpi_desc)
+{
+       struct device *dev = acpi_desc->dev;
+       struct kernfs_node *nfit;
+       struct device *bus_dev;
+
+       if (!ars_supported(acpi_desc->nvdimm_bus))
+               return 0;
+
+       bus_dev = to_nvdimm_bus_dev(acpi_desc->nvdimm_bus);
+       nfit = sysfs_get_dirent(bus_dev->kobj.sd, "nfit");
+       if (!nfit) {
+               dev_err(dev, "sysfs_get_dirent 'nfit' failed\n");
+               return -ENODEV;
+       }
+       acpi_desc->scrub_count_state = sysfs_get_dirent(nfit, "scrub");
+       sysfs_put(nfit);
+       if (!acpi_desc->scrub_count_state) {
+               dev_err(dev, "sysfs_get_dirent 'scrub' failed\n");
+               return -ENODEV;
+       }
+
+       return 0;
+}
+
+static void acpi_nfit_destruct(void *data)
+{
+       struct acpi_nfit_desc *acpi_desc = data;
+       struct device *bus_dev = to_nvdimm_bus_dev(acpi_desc->nvdimm_bus);
+
+       acpi_desc->cancel = 1;
+       /*
+        * Bounce the nvdimm bus lock to make sure any in-flight
+        * acpi_nfit_ars_rescan() submissions have had a chance to
+        * either submit or see ->cancel set.
+        */
+       device_lock(bus_dev);
+       device_unlock(bus_dev);
+
+       flush_workqueue(nfit_wq);
+       if (acpi_desc->scrub_count_state)
+               sysfs_put(acpi_desc->scrub_count_state);
+       nvdimm_bus_unregister(acpi_desc->nvdimm_bus);
+       acpi_desc->nvdimm_bus = NULL;
+}
+
+int acpi_nfit_init(struct acpi_nfit_desc *acpi_desc, void *data, acpi_size sz)
+{
+       struct device *dev = acpi_desc->dev;
+       struct nfit_table_prev prev;
+       const void *end;
+       int rc;
+
+       if (!acpi_desc->nvdimm_bus) {
+               acpi_nfit_init_dsms(acpi_desc);
+
+               acpi_desc->nvdimm_bus = nvdimm_bus_register(dev,
+                               &acpi_desc->nd_desc);
+               if (!acpi_desc->nvdimm_bus)
+                       return -ENOMEM;
+
+               rc = devm_add_action_or_reset(dev, acpi_nfit_destruct,
+                               acpi_desc);
+               if (rc)
+                       return rc;
+
+               rc = acpi_nfit_desc_init_scrub_attr(acpi_desc);
+               if (rc)
+                       return rc;
+       }
+
+       mutex_lock(&acpi_desc->init_mutex);
+
+       INIT_LIST_HEAD(&prev.spas);
+       INIT_LIST_HEAD(&prev.memdevs);
+       INIT_LIST_HEAD(&prev.dcrs);
+       INIT_LIST_HEAD(&prev.bdws);
+       INIT_LIST_HEAD(&prev.idts);
+       INIT_LIST_HEAD(&prev.flushes);
+
+       list_cut_position(&prev.spas, &acpi_desc->spas,
+                               acpi_desc->spas.prev);
+       list_cut_position(&prev.memdevs, &acpi_desc->memdevs,
+                               acpi_desc->memdevs.prev);
+       list_cut_position(&prev.dcrs, &acpi_desc->dcrs,
+                               acpi_desc->dcrs.prev);
+       list_cut_position(&prev.bdws, &acpi_desc->bdws,
+                               acpi_desc->bdws.prev);
+       list_cut_position(&prev.idts, &acpi_desc->idts,
+                               acpi_desc->idts.prev);
+       list_cut_position(&prev.flushes, &acpi_desc->flushes,
+                               acpi_desc->flushes.prev);
+
+       end = data + sz;
+       while (!IS_ERR_OR_NULL(data))
+               data = add_table(acpi_desc, &prev, data, end);
+
+       if (IS_ERR(data)) {
+               dev_dbg(dev, "%s: nfit table parsing error: %ld\n", __func__,
+                               PTR_ERR(data));
+               rc = PTR_ERR(data);
+               goto out_unlock;
+       }
+
+       rc = acpi_nfit_check_deletions(acpi_desc, &prev);
+       if (rc)
+               goto out_unlock;
+
+       rc = nfit_mem_init(acpi_desc);
+       if (rc)
+               goto out_unlock;
+
+       rc = acpi_nfit_register_dimms(acpi_desc);
+       if (rc)
+               goto out_unlock;
+
+       rc = acpi_nfit_register_regions(acpi_desc);
+
+ out_unlock:
+       mutex_unlock(&acpi_desc->init_mutex);
+       return rc;
+}
+EXPORT_SYMBOL_GPL(acpi_nfit_init);
+
+struct acpi_nfit_flush_work {
+       struct work_struct work;
+       struct completion cmp;
+};
+
+static void flush_probe(struct work_struct *work)
+{
+       struct acpi_nfit_flush_work *flush;
+
+       flush = container_of(work, typeof(*flush), work);
+       complete(&flush->cmp);
+}
+
+static int acpi_nfit_flush_probe(struct nvdimm_bus_descriptor *nd_desc)
+{
+       struct acpi_nfit_desc *acpi_desc = to_acpi_nfit_desc(nd_desc);
+       struct device *dev = acpi_desc->dev;
+       struct acpi_nfit_flush_work flush;
+
+       /* bounce the device lock to flush acpi_nfit_add / acpi_nfit_notify */
+       device_lock(dev);
+       device_unlock(dev);
+
+       /*
+        * Scrub work could take 10s of seconds, userspace may give up so we
+        * need to be interruptible while waiting.
+        */
+       INIT_WORK_ONSTACK(&flush.work, flush_probe);
+       COMPLETION_INITIALIZER_ONSTACK(flush.cmp);
+       queue_work(nfit_wq, &flush.work);
+       return wait_for_completion_interruptible(&flush.cmp);
+}
+
+static int acpi_nfit_clear_to_send(struct nvdimm_bus_descriptor *nd_desc,
+               struct nvdimm *nvdimm, unsigned int cmd)
+{
+       struct acpi_nfit_desc *acpi_desc = to_acpi_nfit_desc(nd_desc);
+
+       if (nvdimm)
+               return 0;
+       if (cmd != ND_CMD_ARS_START)
+               return 0;
+
+       /*
+        * The kernel and userspace may race to initiate a scrub, but
+        * the scrub thread is prepared to lose that initial race.  It
+        * just needs guarantees that any ars it initiates are not
+        * interrupted by any intervening start reqeusts from userspace.
+        */
+       if (work_busy(&acpi_desc->work))
+               return -EBUSY;
+
+       return 0;
+}
+
+static int acpi_nfit_ars_rescan(struct acpi_nfit_desc *acpi_desc)
+{
+       struct device *dev = acpi_desc->dev;
+       struct nfit_spa *nfit_spa;
+
+       if (work_busy(&acpi_desc->work))
+               return -EBUSY;
+
+       if (acpi_desc->cancel)
+               return 0;
+
+       mutex_lock(&acpi_desc->init_mutex);
+       list_for_each_entry(nfit_spa, &acpi_desc->spas, list) {
+               struct acpi_nfit_system_address *spa = nfit_spa->spa;
+
+               if (nfit_spa_type(spa) != NFIT_SPA_PM)
+                       continue;
+
+               nfit_spa->ars_required = 1;
+       }
+       queue_work(nfit_wq, &acpi_desc->work);
+       dev_dbg(dev, "%s: ars_scan triggered\n", __func__);
+       mutex_unlock(&acpi_desc->init_mutex);
+
+       return 0;
+}
+
+void acpi_nfit_desc_init(struct acpi_nfit_desc *acpi_desc, struct device *dev)
+{
+       struct nvdimm_bus_descriptor *nd_desc;
+
+       dev_set_drvdata(dev, acpi_desc);
+       acpi_desc->dev = dev;
+       acpi_desc->blk_do_io = acpi_nfit_blk_region_do_io;
+       nd_desc = &acpi_desc->nd_desc;
+       nd_desc->provider_name = "ACPI.NFIT";
+       nd_desc->module = THIS_MODULE;
+       nd_desc->ndctl = acpi_nfit_ctl;
+       nd_desc->flush_probe = acpi_nfit_flush_probe;
+       nd_desc->clear_to_send = acpi_nfit_clear_to_send;
+       nd_desc->attr_groups = acpi_nfit_attribute_groups;
+
+       INIT_LIST_HEAD(&acpi_desc->spas);
+       INIT_LIST_HEAD(&acpi_desc->dcrs);
+       INIT_LIST_HEAD(&acpi_desc->bdws);
+       INIT_LIST_HEAD(&acpi_desc->idts);
+       INIT_LIST_HEAD(&acpi_desc->flushes);
+       INIT_LIST_HEAD(&acpi_desc->memdevs);
+       INIT_LIST_HEAD(&acpi_desc->dimms);
+       mutex_init(&acpi_desc->init_mutex);
+       INIT_WORK(&acpi_desc->work, acpi_nfit_scrub);
+}
+EXPORT_SYMBOL_GPL(acpi_nfit_desc_init);
+
+static int acpi_nfit_add(struct acpi_device *adev)
+{
+       struct acpi_buffer buf = { ACPI_ALLOCATE_BUFFER, NULL };
+       struct acpi_nfit_desc *acpi_desc;
+       struct device *dev = &adev->dev;
+       struct acpi_table_header *tbl;
+       acpi_status status = AE_OK;
+       acpi_size sz;
+       int rc = 0;
+
+       status = acpi_get_table_with_size(ACPI_SIG_NFIT, 0, &tbl, &sz);
+       if (ACPI_FAILURE(status)) {
+               /* This is ok, we could have an nvdimm hotplugged later */
+               dev_dbg(dev, "failed to find NFIT at startup\n");
+               return 0;
+       }
+
+       acpi_desc = devm_kzalloc(dev, sizeof(*acpi_desc), GFP_KERNEL);
+       if (!acpi_desc)
+               return -ENOMEM;
+       acpi_nfit_desc_init(acpi_desc, &adev->dev);
+
+       /* Save the acpi header for exporting the revision via sysfs */
+       acpi_desc->acpi_header = *tbl;
+
+       /* Evaluate _FIT and override with that if present */
+       status = acpi_evaluate_object(adev->handle, "_FIT", NULL, &buf);
+       if (ACPI_SUCCESS(status) && buf.length > 0) {
+               union acpi_object *obj = buf.pointer;
+
+               if (obj->type == ACPI_TYPE_BUFFER)
+                       rc = acpi_nfit_init(acpi_desc, obj->buffer.pointer,
+                                       obj->buffer.length);
+               else
+                       dev_dbg(dev, "%s invalid type %d, ignoring _FIT\n",
+                                __func__, (int) obj->type);
+               kfree(buf.pointer);
+       } else
+               /* skip over the lead-in header table */
+               rc = acpi_nfit_init(acpi_desc, (void *) tbl
+                               + sizeof(struct acpi_table_nfit),
+                               sz - sizeof(struct acpi_table_nfit));
+       return rc;
+}
+
+static int acpi_nfit_remove(struct acpi_device *adev)
+{
+       /* see acpi_nfit_destruct */
+       return 0;
+}
+
+static void acpi_nfit_notify(struct acpi_device *adev, u32 event)
+{
+       struct acpi_nfit_desc *acpi_desc = dev_get_drvdata(&adev->dev);
+       struct acpi_buffer buf = { ACPI_ALLOCATE_BUFFER, NULL };
+       struct device *dev = &adev->dev;
+       union acpi_object *obj;
+       acpi_status status;
+       int ret;
+
+       dev_dbg(dev, "%s: event: %d\n", __func__, event);
+
+       device_lock(dev);
+       if (!dev->driver) {
+               /* dev->driver may be null if we're being removed */
+               dev_dbg(dev, "%s: no driver found for dev\n", __func__);
+               goto out_unlock;
+       }
+
+       if (!acpi_desc) {
+               acpi_desc = devm_kzalloc(dev, sizeof(*acpi_desc), GFP_KERNEL);
+               if (!acpi_desc)
+                       goto out_unlock;
+               acpi_nfit_desc_init(acpi_desc, &adev->dev);
+       } else {
+               /*
+                * Finish previous registration before considering new
+                * regions.
+                */
+               flush_workqueue(nfit_wq);
+       }
+
+       /* Evaluate _FIT */
+       status = acpi_evaluate_object(adev->handle, "_FIT", NULL, &buf);
+       if (ACPI_FAILURE(status)) {
+               dev_err(dev, "failed to evaluate _FIT\n");
+               goto out_unlock;
+       }
+
+       obj = buf.pointer;
+       if (obj->type == ACPI_TYPE_BUFFER) {
+               ret = acpi_nfit_init(acpi_desc, obj->buffer.pointer,
+                               obj->buffer.length);
+               if (ret)
+                       dev_err(dev, "failed to merge updated NFIT\n");
+       } else
+               dev_err(dev, "Invalid _FIT\n");
+       kfree(buf.pointer);
+
+ out_unlock:
+       device_unlock(dev);
+}
+
+static const struct acpi_device_id acpi_nfit_ids[] = {
+       { "ACPI0012", 0 },
+       { "", 0 },
+};
+MODULE_DEVICE_TABLE(acpi, acpi_nfit_ids);
+
+static struct acpi_driver acpi_nfit_driver = {
+       .name = KBUILD_MODNAME,
+       .ids = acpi_nfit_ids,
+       .ops = {
+               .add = acpi_nfit_add,
+               .remove = acpi_nfit_remove,
+               .notify = acpi_nfit_notify,
+       },
+};
+
+static __init int nfit_init(void)
+{
+       BUILD_BUG_ON(sizeof(struct acpi_table_nfit) != 40);
+       BUILD_BUG_ON(sizeof(struct acpi_nfit_system_address) != 56);
+       BUILD_BUG_ON(sizeof(struct acpi_nfit_memory_map) != 48);
+       BUILD_BUG_ON(sizeof(struct acpi_nfit_interleave) != 20);
+       BUILD_BUG_ON(sizeof(struct acpi_nfit_smbios) != 9);
+       BUILD_BUG_ON(sizeof(struct acpi_nfit_control_region) != 80);
+       BUILD_BUG_ON(sizeof(struct acpi_nfit_data_region) != 40);
+
+       acpi_str_to_uuid(UUID_VOLATILE_MEMORY, nfit_uuid[NFIT_SPA_VOLATILE]);
+       acpi_str_to_uuid(UUID_PERSISTENT_MEMORY, nfit_uuid[NFIT_SPA_PM]);
+       acpi_str_to_uuid(UUID_CONTROL_REGION, nfit_uuid[NFIT_SPA_DCR]);
+       acpi_str_to_uuid(UUID_DATA_REGION, nfit_uuid[NFIT_SPA_BDW]);
+       acpi_str_to_uuid(UUID_VOLATILE_VIRTUAL_DISK, nfit_uuid[NFIT_SPA_VDISK]);
+       acpi_str_to_uuid(UUID_VOLATILE_VIRTUAL_CD, nfit_uuid[NFIT_SPA_VCD]);
+       acpi_str_to_uuid(UUID_PERSISTENT_VIRTUAL_DISK, nfit_uuid[NFIT_SPA_PDISK]);
+       acpi_str_to_uuid(UUID_PERSISTENT_VIRTUAL_CD, nfit_uuid[NFIT_SPA_PCD]);
+       acpi_str_to_uuid(UUID_NFIT_BUS, nfit_uuid[NFIT_DEV_BUS]);
+       acpi_str_to_uuid(UUID_NFIT_DIMM, nfit_uuid[NFIT_DEV_DIMM]);
+       acpi_str_to_uuid(UUID_NFIT_DIMM_N_HPE1, nfit_uuid[NFIT_DEV_DIMM_N_HPE1]);
+       acpi_str_to_uuid(UUID_NFIT_DIMM_N_HPE2, nfit_uuid[NFIT_DEV_DIMM_N_HPE2]);
+       acpi_str_to_uuid(UUID_NFIT_DIMM_N_MSFT, nfit_uuid[NFIT_DEV_DIMM_N_MSFT]);
+
+       nfit_wq = create_singlethread_workqueue("nfit");
+       if (!nfit_wq)
+               return -ENOMEM;
+
+       return acpi_bus_register_driver(&acpi_nfit_driver);
+}
+
+static __exit void nfit_exit(void)
+{
+       acpi_bus_unregister_driver(&acpi_nfit_driver);
+       destroy_workqueue(nfit_wq);
+}
+
+module_init(nfit_init);
+module_exit(nfit_exit);
+MODULE_LICENSE("GPL v2");
+MODULE_AUTHOR("Intel Corporation");
diff --git a/drivers/acpi/nfit/nfit.h b/drivers/acpi/nfit/nfit.h
new file mode 100644 (file)
index 0000000..6ecf337
--- /dev/null
@@ -0,0 +1,207 @@
+/*
+ * NVDIMM Firmware Interface Table - NFIT
+ *
+ * Copyright(c) 2013-2015 Intel Corporation. All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that 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.
+ */
+#ifndef __NFIT_H__
+#define __NFIT_H__
+#include <linux/workqueue.h>
+#include <linux/libnvdimm.h>
+#include <linux/types.h>
+#include <linux/uuid.h>
+#include <linux/acpi.h>
+#include <acpi/acuuid.h>
+
+/* ACPI 6.1 */
+#define UUID_NFIT_BUS "2f10e7a4-9e91-11e4-89d3-123b93f75cba"
+
+/* http://pmem.io/documents/NVDIMM_DSM_Interface_Example.pdf */
+#define UUID_NFIT_DIMM "4309ac30-0d11-11e4-9191-0800200c9a66"
+
+/* https://github.com/HewlettPackard/hpe-nvm/blob/master/Documentation/ */
+#define UUID_NFIT_DIMM_N_HPE1 "9002c334-acf3-4c0e-9642-a235f0d53bc6"
+#define UUID_NFIT_DIMM_N_HPE2 "5008664b-b758-41a0-a03c-27c2f2d04f7e"
+
+/* https://msdn.microsoft.com/library/windows/hardware/mt604741 */
+#define UUID_NFIT_DIMM_N_MSFT "1ee68b36-d4bd-4a1a-9a16-4f8e53d46e05"
+
+#define ACPI_NFIT_MEM_FAILED_MASK (ACPI_NFIT_MEM_SAVE_FAILED \
+               | ACPI_NFIT_MEM_RESTORE_FAILED | ACPI_NFIT_MEM_FLUSH_FAILED \
+               | ACPI_NFIT_MEM_NOT_ARMED)
+
+enum nfit_uuids {
+       /* for simplicity alias the uuid index with the family id */
+       NFIT_DEV_DIMM = NVDIMM_FAMILY_INTEL,
+       NFIT_DEV_DIMM_N_HPE1 = NVDIMM_FAMILY_HPE1,
+       NFIT_DEV_DIMM_N_HPE2 = NVDIMM_FAMILY_HPE2,
+       NFIT_DEV_DIMM_N_MSFT = NVDIMM_FAMILY_MSFT,
+       NFIT_SPA_VOLATILE,
+       NFIT_SPA_PM,
+       NFIT_SPA_DCR,
+       NFIT_SPA_BDW,
+       NFIT_SPA_VDISK,
+       NFIT_SPA_VCD,
+       NFIT_SPA_PDISK,
+       NFIT_SPA_PCD,
+       NFIT_DEV_BUS,
+       NFIT_UUID_MAX,
+};
+
+/*
+ * Region format interface codes are stored as an array of bytes in the
+ * NFIT DIMM Control Region structure
+ */
+#define NFIT_FIC_BYTE cpu_to_be16(0x101) /* byte-addressable energy backed */
+#define NFIT_FIC_BLK cpu_to_be16(0x201) /* block-addressable non-energy backed */
+#define NFIT_FIC_BYTEN cpu_to_be16(0x301) /* byte-addressable non-energy backed */
+
+enum {
+       NFIT_BLK_READ_FLUSH = 1,
+       NFIT_BLK_DCR_LATCH = 2,
+       NFIT_ARS_STATUS_DONE = 0,
+       NFIT_ARS_STATUS_BUSY = 1 << 16,
+       NFIT_ARS_STATUS_NONE = 2 << 16,
+       NFIT_ARS_STATUS_INTR = 3 << 16,
+       NFIT_ARS_START_BUSY = 6,
+       NFIT_ARS_CAP_NONE = 1,
+       NFIT_ARS_F_OVERFLOW = 1,
+       NFIT_ARS_TIMEOUT = 90,
+};
+
+struct nfit_spa {
+       struct list_head list;
+       struct nd_region *nd_region;
+       unsigned int ars_required:1;
+       u32 clear_err_unit;
+       u32 max_ars;
+       struct acpi_nfit_system_address spa[0];
+};
+
+struct nfit_dcr {
+       struct list_head list;
+       struct acpi_nfit_control_region dcr[0];
+};
+
+struct nfit_bdw {
+       struct list_head list;
+       struct acpi_nfit_data_region bdw[0];
+};
+
+struct nfit_idt {
+       struct list_head list;
+       struct acpi_nfit_interleave idt[0];
+};
+
+struct nfit_flush {
+       struct list_head list;
+       struct acpi_nfit_flush_address flush[0];
+};
+
+struct nfit_memdev {
+       struct list_head list;
+       struct acpi_nfit_memory_map memdev[0];
+};
+
+/* assembled tables for a given dimm/memory-device */
+struct nfit_mem {
+       struct nvdimm *nvdimm;
+       struct acpi_nfit_memory_map *memdev_dcr;
+       struct acpi_nfit_memory_map *memdev_pmem;
+       struct acpi_nfit_memory_map *memdev_bdw;
+       struct acpi_nfit_control_region *dcr;
+       struct acpi_nfit_data_region *bdw;
+       struct acpi_nfit_system_address *spa_dcr;
+       struct acpi_nfit_system_address *spa_bdw;
+       struct acpi_nfit_interleave *idt_dcr;
+       struct acpi_nfit_interleave *idt_bdw;
+       struct nfit_flush *nfit_flush;
+       struct list_head list;
+       struct acpi_device *adev;
+       struct acpi_nfit_desc *acpi_desc;
+       struct resource *flush_wpq;
+       unsigned long dsm_mask;
+       int family;
+};
+
+struct acpi_nfit_desc {
+       struct nvdimm_bus_descriptor nd_desc;
+       struct acpi_table_header acpi_header;
+       struct mutex init_mutex;
+       struct list_head memdevs;
+       struct list_head flushes;
+       struct list_head dimms;
+       struct list_head spas;
+       struct list_head dcrs;
+       struct list_head bdws;
+       struct list_head idts;
+       struct nvdimm_bus *nvdimm_bus;
+       struct device *dev;
+       struct nd_cmd_ars_status *ars_status;
+       size_t ars_status_size;
+       struct work_struct work;
+       struct kernfs_node *scrub_count_state;
+       unsigned int scrub_count;
+       unsigned int cancel:1;
+       unsigned long dimm_cmd_force_en;
+       unsigned long bus_cmd_force_en;
+       int (*blk_do_io)(struct nd_blk_region *ndbr, resource_size_t dpa,
+                       void *iobuf, u64 len, int rw);
+};
+
+enum nd_blk_mmio_selector {
+       BDW,
+       DCR,
+};
+
+struct nd_blk_addr {
+       union {
+               void __iomem *base;
+               void *aperture;
+       };
+};
+
+struct nfit_blk {
+       struct nfit_blk_mmio {
+               struct nd_blk_addr addr;
+               u64 size;
+               u64 base_offset;
+               u32 line_size;
+               u32 num_lines;
+               u32 table_size;
+               struct acpi_nfit_interleave *idt;
+               struct acpi_nfit_system_address *spa;
+       } mmio[2];
+       struct nd_region *nd_region;
+       u64 bdw_offset; /* post interleave offset */
+       u64 stat_offset;
+       u64 cmd_offset;
+       u32 dimm_flags;
+};
+
+static inline struct acpi_nfit_memory_map *__to_nfit_memdev(
+               struct nfit_mem *nfit_mem)
+{
+       if (nfit_mem->memdev_dcr)
+               return nfit_mem->memdev_dcr;
+       return nfit_mem->memdev_pmem;
+}
+
+static inline struct acpi_nfit_desc *to_acpi_desc(
+               struct nvdimm_bus_descriptor *nd_desc)
+{
+       return container_of(nd_desc, struct acpi_nfit_desc, nd_desc);
+}
+
+const u8 *to_nfit_uuid(enum nfit_uuids id);
+int acpi_nfit_init(struct acpi_nfit_desc *acpi_desc, void *nfit, acpi_size sz);
+void acpi_nfit_desc_init(struct acpi_nfit_desc *acpi_desc, struct device *dev);
+#endif /* __NFIT_H__ */
index 62d5ac3..0dca8ff 100644 (file)
@@ -16,7 +16,7 @@ ldflags-y += --wrap=remove_resource
 
 DRIVERS := ../../../drivers
 NVDIMM_SRC := $(DRIVERS)/nvdimm
-ACPI_SRC := $(DRIVERS)/acpi
+ACPI_SRC := $(DRIVERS)/acpi/nfit
 DAX_SRC := $(DRIVERS)/dax
 ccflags-y := -I$(src)/$(NVDIMM_SRC)/
 
@@ -29,7 +29,7 @@ obj-$(CONFIG_ACPI_NFIT) += nfit.o
 obj-$(CONFIG_DEV_DAX) += dax.o
 obj-$(CONFIG_DEV_DAX_PMEM) += dax_pmem.o
 
-nfit-y := $(ACPI_SRC)/nfit.o
+nfit-y := $(ACPI_SRC)/core.o
 nfit-y += config_check.o
 
 nd_pmem-y := $(NVDIMM_SRC)/pmem.o
index 9241064..d32f25b 100644 (file)
@@ -1,5 +1,5 @@
 ccflags-y := -I$(src)/../../../../drivers/nvdimm/
-ccflags-y += -I$(src)/../../../../drivers/acpi/
+ccflags-y += -I$(src)/../../../../drivers/acpi/nfit/
 
 obj-m += nfit_test.o
 obj-m += nfit_test_iomap.o