Merge tag 'dmaengine-4.6-rc1' of git://git.infradead.org/users/vkoul/slave-dma
authorLinus Torvalds <torvalds@linux-foundation.org>
Thu, 17 Mar 2016 19:34:54 +0000 (12:34 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Thu, 17 Mar 2016 19:34:54 +0000 (12:34 -0700)
Pull dmaengine updates from Vinod Koul:
 "This is smallish update with minor changes to core and new driver and
  usual updates.  Nothing super exciting here..

   - We have made slave address as physical to enable driver to do the
     mapping.

   - We now expose the maxburst for slave dma as new capability so
     clients can know this and program accordingly

   - addition of device synchronize callbacks on omap and edma.

   - pl330 updates to support DMAFLUSHP for Rockchip platforms.

   - Updates and improved sg handling in Xilinx VDMA driver.

   - New hidma qualcomm dma driver, though some bits are still in
     progress"

* tag 'dmaengine-4.6-rc1' of git://git.infradead.org/users/vkoul/slave-dma: (40 commits)
  dmaengine: IOATDMA: revise channel reset workaround on CB3.3 platforms
  dmaengine: add Qualcomm Technologies HIDMA channel driver
  dmaengine: add Qualcomm Technologies HIDMA management driver
  dmaengine: hidma: Add Device Tree binding
  dmaengine: qcom_bam_dma: move to qcom directory
  dmaengine: tegra: Move of_device_id table near to its user
  dmaengine: xilinx_vdma: Remove unnecessary variable initializations
  dmaengine: sirf: use __maybe_unused to hide pm functions
  dmaengine: rcar-dmac: clear pertinence number of channels
  dmaengine: sh: shdmac: don't open code of_device_get_match_data()
  dmaengine: tegra: don't open code of_device_get_match_data()
  dmaengine: qcom_bam_dma: Make driver work for BE
  dmaengine: sun4i: support module autoloading
  dma/mic_x100_dma: IS_ERR() vs PTR_ERR() typo
  dmaengine: xilinx_vdma: Use readl_poll_timeout instead of do while loop's
  dmaengine: xilinx_vdma: Simplify spin lock handling
  dmaengine: xilinx_vdma: Fix issues with non-parking mode
  dmaengine: xilinx_vdma: Improve SG engine handling
  dmaengine: pl330: fix to support the burst mode
  dmaengine: make slave address physical
  ...

1  2 
drivers/dma/edma.c
drivers/dma/ioat/dma.c
drivers/dma/qcom/bam_dma.c
drivers/spi/spi-rockchip.c

Simple merge
Simple merge
index 0000000,2d691a3..d5e0a9c
mode 000000,100644..100644
--- /dev/null
@@@ -1,0 -1,1262 +1,1262 @@@
 -      bchan->fifo_virt = dma_alloc_writecombine(bdev->dev, BAM_DESC_FIFO_SIZE,
 -                              &bchan->fifo_phys, GFP_KERNEL);
+ /*
+  * Copyright (c) 2013-2014, The Linux Foundation. All rights reserved.
+  *
+  * This program is free software; you can redistribute it and/or modify
+  * it under the terms of the GNU General Public License version 2 and
+  * only version 2 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.
+  *
+  */
+ /*
+  * QCOM BAM DMA engine driver
+  *
+  * QCOM BAM DMA blocks are distributed amongst a number of the on-chip
+  * peripherals on the MSM 8x74.  The configuration of the channels are dependent
+  * on the way they are hard wired to that specific peripheral.  The peripheral
+  * device tree entries specify the configuration of each channel.
+  *
+  * The DMA controller requires the use of external memory for storage of the
+  * hardware descriptors for each channel.  The descriptor FIFO is accessed as a
+  * circular buffer and operations are managed according to the offset within the
+  * FIFO.  After pipe/channel reset, all of the pipe registers and internal state
+  * are back to defaults.
+  *
+  * During DMA operations, we write descriptors to the FIFO, being careful to
+  * handle wrapping and then write the last FIFO offset to that channel's
+  * P_EVNT_REG register to kick off the transaction.  The P_SW_OFSTS register
+  * indicates the current FIFO offset that is being processed, so there is some
+  * indication of where the hardware is currently working.
+  */
+ #include <linux/kernel.h>
+ #include <linux/io.h>
+ #include <linux/init.h>
+ #include <linux/slab.h>
+ #include <linux/module.h>
+ #include <linux/interrupt.h>
+ #include <linux/dma-mapping.h>
+ #include <linux/scatterlist.h>
+ #include <linux/device.h>
+ #include <linux/platform_device.h>
+ #include <linux/of.h>
+ #include <linux/of_address.h>
+ #include <linux/of_irq.h>
+ #include <linux/of_dma.h>
+ #include <linux/clk.h>
+ #include <linux/dmaengine.h>
+ #include "../dmaengine.h"
+ #include "../virt-dma.h"
+ struct bam_desc_hw {
+       __le32 addr;            /* Buffer physical address */
+       __le16 size;            /* Buffer size in bytes */
+       __le16 flags;
+ };
+ #define DESC_FLAG_INT BIT(15)
+ #define DESC_FLAG_EOT BIT(14)
+ #define DESC_FLAG_EOB BIT(13)
+ #define DESC_FLAG_NWD BIT(12)
+ struct bam_async_desc {
+       struct virt_dma_desc vd;
+       u32 num_desc;
+       u32 xfer_len;
+       /* transaction flags, EOT|EOB|NWD */
+       u16 flags;
+       struct bam_desc_hw *curr_desc;
+       enum dma_transfer_direction dir;
+       size_t length;
+       struct bam_desc_hw desc[0];
+ };
+ enum bam_reg {
+       BAM_CTRL,
+       BAM_REVISION,
+       BAM_NUM_PIPES,
+       BAM_DESC_CNT_TRSHLD,
+       BAM_IRQ_SRCS,
+       BAM_IRQ_SRCS_MSK,
+       BAM_IRQ_SRCS_UNMASKED,
+       BAM_IRQ_STTS,
+       BAM_IRQ_CLR,
+       BAM_IRQ_EN,
+       BAM_CNFG_BITS,
+       BAM_IRQ_SRCS_EE,
+       BAM_IRQ_SRCS_MSK_EE,
+       BAM_P_CTRL,
+       BAM_P_RST,
+       BAM_P_HALT,
+       BAM_P_IRQ_STTS,
+       BAM_P_IRQ_CLR,
+       BAM_P_IRQ_EN,
+       BAM_P_EVNT_DEST_ADDR,
+       BAM_P_EVNT_REG,
+       BAM_P_SW_OFSTS,
+       BAM_P_DATA_FIFO_ADDR,
+       BAM_P_DESC_FIFO_ADDR,
+       BAM_P_EVNT_GEN_TRSHLD,
+       BAM_P_FIFO_SIZES,
+ };
+ struct reg_offset_data {
+       u32 base_offset;
+       unsigned int pipe_mult, evnt_mult, ee_mult;
+ };
+ static const struct reg_offset_data bam_v1_3_reg_info[] = {
+       [BAM_CTRL]              = { 0x0F80, 0x00, 0x00, 0x00 },
+       [BAM_REVISION]          = { 0x0F84, 0x00, 0x00, 0x00 },
+       [BAM_NUM_PIPES]         = { 0x0FBC, 0x00, 0x00, 0x00 },
+       [BAM_DESC_CNT_TRSHLD]   = { 0x0F88, 0x00, 0x00, 0x00 },
+       [BAM_IRQ_SRCS]          = { 0x0F8C, 0x00, 0x00, 0x00 },
+       [BAM_IRQ_SRCS_MSK]      = { 0x0F90, 0x00, 0x00, 0x00 },
+       [BAM_IRQ_SRCS_UNMASKED] = { 0x0FB0, 0x00, 0x00, 0x00 },
+       [BAM_IRQ_STTS]          = { 0x0F94, 0x00, 0x00, 0x00 },
+       [BAM_IRQ_CLR]           = { 0x0F98, 0x00, 0x00, 0x00 },
+       [BAM_IRQ_EN]            = { 0x0F9C, 0x00, 0x00, 0x00 },
+       [BAM_CNFG_BITS]         = { 0x0FFC, 0x00, 0x00, 0x00 },
+       [BAM_IRQ_SRCS_EE]       = { 0x1800, 0x00, 0x00, 0x80 },
+       [BAM_IRQ_SRCS_MSK_EE]   = { 0x1804, 0x00, 0x00, 0x80 },
+       [BAM_P_CTRL]            = { 0x0000, 0x80, 0x00, 0x00 },
+       [BAM_P_RST]             = { 0x0004, 0x80, 0x00, 0x00 },
+       [BAM_P_HALT]            = { 0x0008, 0x80, 0x00, 0x00 },
+       [BAM_P_IRQ_STTS]        = { 0x0010, 0x80, 0x00, 0x00 },
+       [BAM_P_IRQ_CLR]         = { 0x0014, 0x80, 0x00, 0x00 },
+       [BAM_P_IRQ_EN]          = { 0x0018, 0x80, 0x00, 0x00 },
+       [BAM_P_EVNT_DEST_ADDR]  = { 0x102C, 0x00, 0x40, 0x00 },
+       [BAM_P_EVNT_REG]        = { 0x1018, 0x00, 0x40, 0x00 },
+       [BAM_P_SW_OFSTS]        = { 0x1000, 0x00, 0x40, 0x00 },
+       [BAM_P_DATA_FIFO_ADDR]  = { 0x1024, 0x00, 0x40, 0x00 },
+       [BAM_P_DESC_FIFO_ADDR]  = { 0x101C, 0x00, 0x40, 0x00 },
+       [BAM_P_EVNT_GEN_TRSHLD] = { 0x1028, 0x00, 0x40, 0x00 },
+       [BAM_P_FIFO_SIZES]      = { 0x1020, 0x00, 0x40, 0x00 },
+ };
+ static const struct reg_offset_data bam_v1_4_reg_info[] = {
+       [BAM_CTRL]              = { 0x0000, 0x00, 0x00, 0x00 },
+       [BAM_REVISION]          = { 0x0004, 0x00, 0x00, 0x00 },
+       [BAM_NUM_PIPES]         = { 0x003C, 0x00, 0x00, 0x00 },
+       [BAM_DESC_CNT_TRSHLD]   = { 0x0008, 0x00, 0x00, 0x00 },
+       [BAM_IRQ_SRCS]          = { 0x000C, 0x00, 0x00, 0x00 },
+       [BAM_IRQ_SRCS_MSK]      = { 0x0010, 0x00, 0x00, 0x00 },
+       [BAM_IRQ_SRCS_UNMASKED] = { 0x0030, 0x00, 0x00, 0x00 },
+       [BAM_IRQ_STTS]          = { 0x0014, 0x00, 0x00, 0x00 },
+       [BAM_IRQ_CLR]           = { 0x0018, 0x00, 0x00, 0x00 },
+       [BAM_IRQ_EN]            = { 0x001C, 0x00, 0x00, 0x00 },
+       [BAM_CNFG_BITS]         = { 0x007C, 0x00, 0x00, 0x00 },
+       [BAM_IRQ_SRCS_EE]       = { 0x0800, 0x00, 0x00, 0x80 },
+       [BAM_IRQ_SRCS_MSK_EE]   = { 0x0804, 0x00, 0x00, 0x80 },
+       [BAM_P_CTRL]            = { 0x1000, 0x1000, 0x00, 0x00 },
+       [BAM_P_RST]             = { 0x1004, 0x1000, 0x00, 0x00 },
+       [BAM_P_HALT]            = { 0x1008, 0x1000, 0x00, 0x00 },
+       [BAM_P_IRQ_STTS]        = { 0x1010, 0x1000, 0x00, 0x00 },
+       [BAM_P_IRQ_CLR]         = { 0x1014, 0x1000, 0x00, 0x00 },
+       [BAM_P_IRQ_EN]          = { 0x1018, 0x1000, 0x00, 0x00 },
+       [BAM_P_EVNT_DEST_ADDR]  = { 0x182C, 0x00, 0x1000, 0x00 },
+       [BAM_P_EVNT_REG]        = { 0x1818, 0x00, 0x1000, 0x00 },
+       [BAM_P_SW_OFSTS]        = { 0x1800, 0x00, 0x1000, 0x00 },
+       [BAM_P_DATA_FIFO_ADDR]  = { 0x1824, 0x00, 0x1000, 0x00 },
+       [BAM_P_DESC_FIFO_ADDR]  = { 0x181C, 0x00, 0x1000, 0x00 },
+       [BAM_P_EVNT_GEN_TRSHLD] = { 0x1828, 0x00, 0x1000, 0x00 },
+       [BAM_P_FIFO_SIZES]      = { 0x1820, 0x00, 0x1000, 0x00 },
+ };
+ static const struct reg_offset_data bam_v1_7_reg_info[] = {
+       [BAM_CTRL]              = { 0x00000, 0x00, 0x00, 0x00 },
+       [BAM_REVISION]          = { 0x01000, 0x00, 0x00, 0x00 },
+       [BAM_NUM_PIPES]         = { 0x01008, 0x00, 0x00, 0x00 },
+       [BAM_DESC_CNT_TRSHLD]   = { 0x00008, 0x00, 0x00, 0x00 },
+       [BAM_IRQ_SRCS]          = { 0x03010, 0x00, 0x00, 0x00 },
+       [BAM_IRQ_SRCS_MSK]      = { 0x03014, 0x00, 0x00, 0x00 },
+       [BAM_IRQ_SRCS_UNMASKED] = { 0x03018, 0x00, 0x00, 0x00 },
+       [BAM_IRQ_STTS]          = { 0x00014, 0x00, 0x00, 0x00 },
+       [BAM_IRQ_CLR]           = { 0x00018, 0x00, 0x00, 0x00 },
+       [BAM_IRQ_EN]            = { 0x0001C, 0x00, 0x00, 0x00 },
+       [BAM_CNFG_BITS]         = { 0x0007C, 0x00, 0x00, 0x00 },
+       [BAM_IRQ_SRCS_EE]       = { 0x03000, 0x00, 0x00, 0x1000 },
+       [BAM_IRQ_SRCS_MSK_EE]   = { 0x03004, 0x00, 0x00, 0x1000 },
+       [BAM_P_CTRL]            = { 0x13000, 0x1000, 0x00, 0x00 },
+       [BAM_P_RST]             = { 0x13004, 0x1000, 0x00, 0x00 },
+       [BAM_P_HALT]            = { 0x13008, 0x1000, 0x00, 0x00 },
+       [BAM_P_IRQ_STTS]        = { 0x13010, 0x1000, 0x00, 0x00 },
+       [BAM_P_IRQ_CLR]         = { 0x13014, 0x1000, 0x00, 0x00 },
+       [BAM_P_IRQ_EN]          = { 0x13018, 0x1000, 0x00, 0x00 },
+       [BAM_P_EVNT_DEST_ADDR]  = { 0x1382C, 0x00, 0x1000, 0x00 },
+       [BAM_P_EVNT_REG]        = { 0x13818, 0x00, 0x1000, 0x00 },
+       [BAM_P_SW_OFSTS]        = { 0x13800, 0x00, 0x1000, 0x00 },
+       [BAM_P_DATA_FIFO_ADDR]  = { 0x13824, 0x00, 0x1000, 0x00 },
+       [BAM_P_DESC_FIFO_ADDR]  = { 0x1381C, 0x00, 0x1000, 0x00 },
+       [BAM_P_EVNT_GEN_TRSHLD] = { 0x13828, 0x00, 0x1000, 0x00 },
+       [BAM_P_FIFO_SIZES]      = { 0x13820, 0x00, 0x1000, 0x00 },
+ };
+ /* BAM CTRL */
+ #define BAM_SW_RST                    BIT(0)
+ #define BAM_EN                                BIT(1)
+ #define BAM_EN_ACCUM                  BIT(4)
+ #define BAM_TESTBUS_SEL_SHIFT         5
+ #define BAM_TESTBUS_SEL_MASK          0x3F
+ #define BAM_DESC_CACHE_SEL_SHIFT      13
+ #define BAM_DESC_CACHE_SEL_MASK               0x3
+ #define BAM_CACHED_DESC_STORE         BIT(15)
+ #define IBC_DISABLE                   BIT(16)
+ /* BAM REVISION */
+ #define REVISION_SHIFT                0
+ #define REVISION_MASK         0xFF
+ #define NUM_EES_SHIFT         8
+ #define NUM_EES_MASK          0xF
+ #define CE_BUFFER_SIZE                BIT(13)
+ #define AXI_ACTIVE            BIT(14)
+ #define USE_VMIDMT            BIT(15)
+ #define SECURED                       BIT(16)
+ #define BAM_HAS_NO_BYPASS     BIT(17)
+ #define HIGH_FREQUENCY_BAM    BIT(18)
+ #define INACTIV_TMRS_EXST     BIT(19)
+ #define NUM_INACTIV_TMRS      BIT(20)
+ #define DESC_CACHE_DEPTH_SHIFT        21
+ #define DESC_CACHE_DEPTH_1    (0 << DESC_CACHE_DEPTH_SHIFT)
+ #define DESC_CACHE_DEPTH_2    (1 << DESC_CACHE_DEPTH_SHIFT)
+ #define DESC_CACHE_DEPTH_3    (2 << DESC_CACHE_DEPTH_SHIFT)
+ #define DESC_CACHE_DEPTH_4    (3 << DESC_CACHE_DEPTH_SHIFT)
+ #define CMD_DESC_EN           BIT(23)
+ #define INACTIV_TMR_BASE_SHIFT        24
+ #define INACTIV_TMR_BASE_MASK 0xFF
+ /* BAM NUM PIPES */
+ #define BAM_NUM_PIPES_SHIFT           0
+ #define BAM_NUM_PIPES_MASK            0xFF
+ #define PERIPH_NON_PIPE_GRP_SHIFT     16
+ #define PERIPH_NON_PIP_GRP_MASK               0xFF
+ #define BAM_NON_PIPE_GRP_SHIFT                24
+ #define BAM_NON_PIPE_GRP_MASK         0xFF
+ /* BAM CNFG BITS */
+ #define BAM_PIPE_CNFG         BIT(2)
+ #define BAM_FULL_PIPE         BIT(11)
+ #define BAM_NO_EXT_P_RST      BIT(12)
+ #define BAM_IBC_DISABLE               BIT(13)
+ #define BAM_SB_CLK_REQ                BIT(14)
+ #define BAM_PSM_CSW_REQ               BIT(15)
+ #define BAM_PSM_P_RES         BIT(16)
+ #define BAM_AU_P_RES          BIT(17)
+ #define BAM_SI_P_RES          BIT(18)
+ #define BAM_WB_P_RES          BIT(19)
+ #define BAM_WB_BLK_CSW                BIT(20)
+ #define BAM_WB_CSW_ACK_IDL    BIT(21)
+ #define BAM_WB_RETR_SVPNT     BIT(22)
+ #define BAM_WB_DSC_AVL_P_RST  BIT(23)
+ #define BAM_REG_P_EN          BIT(24)
+ #define BAM_PSM_P_HD_DATA     BIT(25)
+ #define BAM_AU_ACCUMED                BIT(26)
+ #define BAM_CMD_ENABLE                BIT(27)
+ #define BAM_CNFG_BITS_DEFAULT (BAM_PIPE_CNFG |        \
+                                BAM_NO_EXT_P_RST |     \
+                                BAM_IBC_DISABLE |      \
+                                BAM_SB_CLK_REQ |       \
+                                BAM_PSM_CSW_REQ |      \
+                                BAM_PSM_P_RES |        \
+                                BAM_AU_P_RES |         \
+                                BAM_SI_P_RES |         \
+                                BAM_WB_P_RES |         \
+                                BAM_WB_BLK_CSW |       \
+                                BAM_WB_CSW_ACK_IDL |   \
+                                BAM_WB_RETR_SVPNT |    \
+                                BAM_WB_DSC_AVL_P_RST | \
+                                BAM_REG_P_EN |         \
+                                BAM_PSM_P_HD_DATA |    \
+                                BAM_AU_ACCUMED |       \
+                                BAM_CMD_ENABLE)
+ /* PIPE CTRL */
+ #define P_EN                  BIT(1)
+ #define P_DIRECTION           BIT(3)
+ #define P_SYS_STRM            BIT(4)
+ #define P_SYS_MODE            BIT(5)
+ #define P_AUTO_EOB            BIT(6)
+ #define P_AUTO_EOB_SEL_SHIFT  7
+ #define P_AUTO_EOB_SEL_512    (0 << P_AUTO_EOB_SEL_SHIFT)
+ #define P_AUTO_EOB_SEL_256    (1 << P_AUTO_EOB_SEL_SHIFT)
+ #define P_AUTO_EOB_SEL_128    (2 << P_AUTO_EOB_SEL_SHIFT)
+ #define P_AUTO_EOB_SEL_64     (3 << P_AUTO_EOB_SEL_SHIFT)
+ #define P_PREFETCH_LIMIT_SHIFT        9
+ #define P_PREFETCH_LIMIT_32   (0 << P_PREFETCH_LIMIT_SHIFT)
+ #define P_PREFETCH_LIMIT_16   (1 << P_PREFETCH_LIMIT_SHIFT)
+ #define P_PREFETCH_LIMIT_4    (2 << P_PREFETCH_LIMIT_SHIFT)
+ #define P_WRITE_NWD           BIT(11)
+ #define P_LOCK_GROUP_SHIFT    16
+ #define P_LOCK_GROUP_MASK     0x1F
+ /* BAM_DESC_CNT_TRSHLD */
+ #define CNT_TRSHLD            0xffff
+ #define DEFAULT_CNT_THRSHLD   0x4
+ /* BAM_IRQ_SRCS */
+ #define BAM_IRQ                       BIT(31)
+ #define P_IRQ                 0x7fffffff
+ /* BAM_IRQ_SRCS_MSK */
+ #define BAM_IRQ_MSK           BAM_IRQ
+ #define P_IRQ_MSK             P_IRQ
+ /* BAM_IRQ_STTS */
+ #define BAM_TIMER_IRQ         BIT(4)
+ #define BAM_EMPTY_IRQ         BIT(3)
+ #define BAM_ERROR_IRQ         BIT(2)
+ #define BAM_HRESP_ERR_IRQ     BIT(1)
+ /* BAM_IRQ_CLR */
+ #define BAM_TIMER_CLR         BIT(4)
+ #define BAM_EMPTY_CLR         BIT(3)
+ #define BAM_ERROR_CLR         BIT(2)
+ #define BAM_HRESP_ERR_CLR     BIT(1)
+ /* BAM_IRQ_EN */
+ #define BAM_TIMER_EN          BIT(4)
+ #define BAM_EMPTY_EN          BIT(3)
+ #define BAM_ERROR_EN          BIT(2)
+ #define BAM_HRESP_ERR_EN      BIT(1)
+ /* BAM_P_IRQ_EN */
+ #define P_PRCSD_DESC_EN               BIT(0)
+ #define P_TIMER_EN            BIT(1)
+ #define P_WAKE_EN             BIT(2)
+ #define P_OUT_OF_DESC_EN      BIT(3)
+ #define P_ERR_EN              BIT(4)
+ #define P_TRNSFR_END_EN               BIT(5)
+ #define P_DEFAULT_IRQS_EN     (P_PRCSD_DESC_EN | P_ERR_EN | P_TRNSFR_END_EN)
+ /* BAM_P_SW_OFSTS */
+ #define P_SW_OFSTS_MASK               0xffff
+ #define BAM_DESC_FIFO_SIZE    SZ_32K
+ #define MAX_DESCRIPTORS (BAM_DESC_FIFO_SIZE / sizeof(struct bam_desc_hw) - 1)
+ #define BAM_MAX_DATA_SIZE     (SZ_32K - 8)
+ struct bam_chan {
+       struct virt_dma_chan vc;
+       struct bam_device *bdev;
+       /* configuration from device tree */
+       u32 id;
+       struct bam_async_desc *curr_txd;        /* current running dma */
+       /* runtime configuration */
+       struct dma_slave_config slave;
+       /* fifo storage */
+       struct bam_desc_hw *fifo_virt;
+       dma_addr_t fifo_phys;
+       /* fifo markers */
+       unsigned short head;            /* start of active descriptor entries */
+       unsigned short tail;            /* end of active descriptor entries */
+       unsigned int initialized;       /* is the channel hw initialized? */
+       unsigned int paused;            /* is the channel paused? */
+       unsigned int reconfigure;       /* new slave config? */
+       struct list_head node;
+ };
+ static inline struct bam_chan *to_bam_chan(struct dma_chan *common)
+ {
+       return container_of(common, struct bam_chan, vc.chan);
+ }
+ struct bam_device {
+       void __iomem *regs;
+       struct device *dev;
+       struct dma_device common;
+       struct device_dma_parameters dma_parms;
+       struct bam_chan *channels;
+       u32 num_channels;
+       /* execution environment ID, from DT */
+       u32 ee;
+       const struct reg_offset_data *layout;
+       struct clk *bamclk;
+       int irq;
+       /* dma start transaction tasklet */
+       struct tasklet_struct task;
+ };
+ /**
+  * bam_addr - returns BAM register address
+  * @bdev: bam device
+  * @pipe: pipe instance (ignored when register doesn't have multiple instances)
+  * @reg:  register enum
+  */
+ static inline void __iomem *bam_addr(struct bam_device *bdev, u32 pipe,
+               enum bam_reg reg)
+ {
+       const struct reg_offset_data r = bdev->layout[reg];
+       return bdev->regs + r.base_offset +
+               r.pipe_mult * pipe +
+               r.evnt_mult * pipe +
+               r.ee_mult * bdev->ee;
+ }
+ /**
+  * bam_reset_channel - Reset individual BAM DMA channel
+  * @bchan: bam channel
+  *
+  * This function resets a specific BAM channel
+  */
+ static void bam_reset_channel(struct bam_chan *bchan)
+ {
+       struct bam_device *bdev = bchan->bdev;
+       lockdep_assert_held(&bchan->vc.lock);
+       /* reset channel */
+       writel_relaxed(1, bam_addr(bdev, bchan->id, BAM_P_RST));
+       writel_relaxed(0, bam_addr(bdev, bchan->id, BAM_P_RST));
+       /* don't allow cpu to reorder BAM register accesses done after this */
+       wmb();
+       /* make sure hw is initialized when channel is used the first time  */
+       bchan->initialized = 0;
+ }
+ /**
+  * bam_chan_init_hw - Initialize channel hardware
+  * @bchan: bam channel
+  *
+  * This function resets and initializes the BAM channel
+  */
+ static void bam_chan_init_hw(struct bam_chan *bchan,
+       enum dma_transfer_direction dir)
+ {
+       struct bam_device *bdev = bchan->bdev;
+       u32 val;
+       /* Reset the channel to clear internal state of the FIFO */
+       bam_reset_channel(bchan);
+       /*
+        * write out 8 byte aligned address.  We have enough space for this
+        * because we allocated 1 more descriptor (8 bytes) than we can use
+        */
+       writel_relaxed(ALIGN(bchan->fifo_phys, sizeof(struct bam_desc_hw)),
+                       bam_addr(bdev, bchan->id, BAM_P_DESC_FIFO_ADDR));
+       writel_relaxed(BAM_DESC_FIFO_SIZE,
+                       bam_addr(bdev, bchan->id, BAM_P_FIFO_SIZES));
+       /* enable the per pipe interrupts, enable EOT, ERR, and INT irqs */
+       writel_relaxed(P_DEFAULT_IRQS_EN,
+                       bam_addr(bdev, bchan->id, BAM_P_IRQ_EN));
+       /* unmask the specific pipe and EE combo */
+       val = readl_relaxed(bam_addr(bdev, 0, BAM_IRQ_SRCS_MSK_EE));
+       val |= BIT(bchan->id);
+       writel_relaxed(val, bam_addr(bdev, 0, BAM_IRQ_SRCS_MSK_EE));
+       /* don't allow cpu to reorder the channel enable done below */
+       wmb();
+       /* set fixed direction and mode, then enable channel */
+       val = P_EN | P_SYS_MODE;
+       if (dir == DMA_DEV_TO_MEM)
+               val |= P_DIRECTION;
+       writel_relaxed(val, bam_addr(bdev, bchan->id, BAM_P_CTRL));
+       bchan->initialized = 1;
+       /* init FIFO pointers */
+       bchan->head = 0;
+       bchan->tail = 0;
+ }
+ /**
+  * bam_alloc_chan - Allocate channel resources for DMA channel.
+  * @chan: specified channel
+  *
+  * This function allocates the FIFO descriptor memory
+  */
+ static int bam_alloc_chan(struct dma_chan *chan)
+ {
+       struct bam_chan *bchan = to_bam_chan(chan);
+       struct bam_device *bdev = bchan->bdev;
+       if (bchan->fifo_virt)
+               return 0;
+       /* allocate FIFO descriptor space, but only if necessary */
 -      dma_free_writecombine(bdev->dev, BAM_DESC_FIFO_SIZE, bchan->fifo_virt,
 -                              bchan->fifo_phys);
++      bchan->fifo_virt = dma_alloc_wc(bdev->dev, BAM_DESC_FIFO_SIZE,
++                                      &bchan->fifo_phys, GFP_KERNEL);
+       if (!bchan->fifo_virt) {
+               dev_err(bdev->dev, "Failed to allocate desc fifo\n");
+               return -ENOMEM;
+       }
+       return 0;
+ }
+ /**
+  * bam_free_chan - Frees dma resources associated with specific channel
+  * @chan: specified channel
+  *
+  * Free the allocated fifo descriptor memory and channel resources
+  *
+  */
+ static void bam_free_chan(struct dma_chan *chan)
+ {
+       struct bam_chan *bchan = to_bam_chan(chan);
+       struct bam_device *bdev = bchan->bdev;
+       u32 val;
+       unsigned long flags;
+       vchan_free_chan_resources(to_virt_chan(chan));
+       if (bchan->curr_txd) {
+               dev_err(bchan->bdev->dev, "Cannot free busy channel\n");
+               return;
+       }
+       spin_lock_irqsave(&bchan->vc.lock, flags);
+       bam_reset_channel(bchan);
+       spin_unlock_irqrestore(&bchan->vc.lock, flags);
 -              dma_free_writecombine(bdev->dev, BAM_DESC_FIFO_SIZE,
 -                      bdev->channels[i].fifo_virt,
 -                      bdev->channels[i].fifo_phys);
++      dma_free_wc(bdev->dev, BAM_DESC_FIFO_SIZE, bchan->fifo_virt,
++                  bchan->fifo_phys);
+       bchan->fifo_virt = NULL;
+       /* mask irq for pipe/channel */
+       val = readl_relaxed(bam_addr(bdev, 0, BAM_IRQ_SRCS_MSK_EE));
+       val &= ~BIT(bchan->id);
+       writel_relaxed(val, bam_addr(bdev, 0, BAM_IRQ_SRCS_MSK_EE));
+       /* disable irq */
+       writel_relaxed(0, bam_addr(bdev, bchan->id, BAM_P_IRQ_EN));
+ }
+ /**
+  * bam_slave_config - set slave configuration for channel
+  * @chan: dma channel
+  * @cfg: slave configuration
+  *
+  * Sets slave configuration for channel
+  *
+  */
+ static int bam_slave_config(struct dma_chan *chan,
+                           struct dma_slave_config *cfg)
+ {
+       struct bam_chan *bchan = to_bam_chan(chan);
+       unsigned long flag;
+       spin_lock_irqsave(&bchan->vc.lock, flag);
+       memcpy(&bchan->slave, cfg, sizeof(*cfg));
+       bchan->reconfigure = 1;
+       spin_unlock_irqrestore(&bchan->vc.lock, flag);
+       return 0;
+ }
+ /**
+  * bam_prep_slave_sg - Prep slave sg transaction
+  *
+  * @chan: dma channel
+  * @sgl: scatter gather list
+  * @sg_len: length of sg
+  * @direction: DMA transfer direction
+  * @flags: DMA flags
+  * @context: transfer context (unused)
+  */
+ static struct dma_async_tx_descriptor *bam_prep_slave_sg(struct dma_chan *chan,
+       struct scatterlist *sgl, unsigned int sg_len,
+       enum dma_transfer_direction direction, unsigned long flags,
+       void *context)
+ {
+       struct bam_chan *bchan = to_bam_chan(chan);
+       struct bam_device *bdev = bchan->bdev;
+       struct bam_async_desc *async_desc;
+       struct scatterlist *sg;
+       u32 i;
+       struct bam_desc_hw *desc;
+       unsigned int num_alloc = 0;
+       if (!is_slave_direction(direction)) {
+               dev_err(bdev->dev, "invalid dma direction\n");
+               return NULL;
+       }
+       /* calculate number of required entries */
+       for_each_sg(sgl, sg, sg_len, i)
+               num_alloc += DIV_ROUND_UP(sg_dma_len(sg), BAM_MAX_DATA_SIZE);
+       /* allocate enough room to accomodate the number of entries */
+       async_desc = kzalloc(sizeof(*async_desc) +
+                       (num_alloc * sizeof(struct bam_desc_hw)), GFP_NOWAIT);
+       if (!async_desc)
+               goto err_out;
+       if (flags & DMA_PREP_FENCE)
+               async_desc->flags |= DESC_FLAG_NWD;
+       if (flags & DMA_PREP_INTERRUPT)
+               async_desc->flags |= DESC_FLAG_EOT;
+       else
+               async_desc->flags |= DESC_FLAG_INT;
+       async_desc->num_desc = num_alloc;
+       async_desc->curr_desc = async_desc->desc;
+       async_desc->dir = direction;
+       /* fill in temporary descriptors */
+       desc = async_desc->desc;
+       for_each_sg(sgl, sg, sg_len, i) {
+               unsigned int remainder = sg_dma_len(sg);
+               unsigned int curr_offset = 0;
+               do {
+                       desc->addr = cpu_to_le32(sg_dma_address(sg) +
+                                                curr_offset);
+                       if (remainder > BAM_MAX_DATA_SIZE) {
+                               desc->size = cpu_to_le16(BAM_MAX_DATA_SIZE);
+                               remainder -= BAM_MAX_DATA_SIZE;
+                               curr_offset += BAM_MAX_DATA_SIZE;
+                       } else {
+                               desc->size = cpu_to_le16(remainder);
+                               remainder = 0;
+                       }
+                       async_desc->length += desc->size;
+                       desc++;
+               } while (remainder > 0);
+       }
+       return vchan_tx_prep(&bchan->vc, &async_desc->vd, flags);
+ err_out:
+       kfree(async_desc);
+       return NULL;
+ }
+ /**
+  * bam_dma_terminate_all - terminate all transactions on a channel
+  * @bchan: bam dma channel
+  *
+  * Dequeues and frees all transactions
+  * No callbacks are done
+  *
+  */
+ static int bam_dma_terminate_all(struct dma_chan *chan)
+ {
+       struct bam_chan *bchan = to_bam_chan(chan);
+       unsigned long flag;
+       LIST_HEAD(head);
+       /* remove all transactions, including active transaction */
+       spin_lock_irqsave(&bchan->vc.lock, flag);
+       if (bchan->curr_txd) {
+               list_add(&bchan->curr_txd->vd.node, &bchan->vc.desc_issued);
+               bchan->curr_txd = NULL;
+       }
+       vchan_get_all_descriptors(&bchan->vc, &head);
+       spin_unlock_irqrestore(&bchan->vc.lock, flag);
+       vchan_dma_desc_free_list(&bchan->vc, &head);
+       return 0;
+ }
+ /**
+  * bam_pause - Pause DMA channel
+  * @chan: dma channel
+  *
+  */
+ static int bam_pause(struct dma_chan *chan)
+ {
+       struct bam_chan *bchan = to_bam_chan(chan);
+       struct bam_device *bdev = bchan->bdev;
+       unsigned long flag;
+       spin_lock_irqsave(&bchan->vc.lock, flag);
+       writel_relaxed(1, bam_addr(bdev, bchan->id, BAM_P_HALT));
+       bchan->paused = 1;
+       spin_unlock_irqrestore(&bchan->vc.lock, flag);
+       return 0;
+ }
+ /**
+  * bam_resume - Resume DMA channel operations
+  * @chan: dma channel
+  *
+  */
+ static int bam_resume(struct dma_chan *chan)
+ {
+       struct bam_chan *bchan = to_bam_chan(chan);
+       struct bam_device *bdev = bchan->bdev;
+       unsigned long flag;
+       spin_lock_irqsave(&bchan->vc.lock, flag);
+       writel_relaxed(0, bam_addr(bdev, bchan->id, BAM_P_HALT));
+       bchan->paused = 0;
+       spin_unlock_irqrestore(&bchan->vc.lock, flag);
+       return 0;
+ }
+ /**
+  * process_channel_irqs - processes the channel interrupts
+  * @bdev: bam controller
+  *
+  * This function processes the channel interrupts
+  *
+  */
+ static u32 process_channel_irqs(struct bam_device *bdev)
+ {
+       u32 i, srcs, pipe_stts;
+       unsigned long flags;
+       struct bam_async_desc *async_desc;
+       srcs = readl_relaxed(bam_addr(bdev, 0, BAM_IRQ_SRCS_EE));
+       /* return early if no pipe/channel interrupts are present */
+       if (!(srcs & P_IRQ))
+               return srcs;
+       for (i = 0; i < bdev->num_channels; i++) {
+               struct bam_chan *bchan = &bdev->channels[i];
+               if (!(srcs & BIT(i)))
+                       continue;
+               /* clear pipe irq */
+               pipe_stts = readl_relaxed(bam_addr(bdev, i, BAM_P_IRQ_STTS));
+               writel_relaxed(pipe_stts, bam_addr(bdev, i, BAM_P_IRQ_CLR));
+               spin_lock_irqsave(&bchan->vc.lock, flags);
+               async_desc = bchan->curr_txd;
+               if (async_desc) {
+                       async_desc->num_desc -= async_desc->xfer_len;
+                       async_desc->curr_desc += async_desc->xfer_len;
+                       bchan->curr_txd = NULL;
+                       /* manage FIFO */
+                       bchan->head += async_desc->xfer_len;
+                       bchan->head %= MAX_DESCRIPTORS;
+                       /*
+                        * if complete, process cookie.  Otherwise
+                        * push back to front of desc_issued so that
+                        * it gets restarted by the tasklet
+                        */
+                       if (!async_desc->num_desc)
+                               vchan_cookie_complete(&async_desc->vd);
+                       else
+                               list_add(&async_desc->vd.node,
+                                       &bchan->vc.desc_issued);
+               }
+               spin_unlock_irqrestore(&bchan->vc.lock, flags);
+       }
+       return srcs;
+ }
+ /**
+  * bam_dma_irq - irq handler for bam controller
+  * @irq: IRQ of interrupt
+  * @data: callback data
+  *
+  * IRQ handler for the bam controller
+  */
+ static irqreturn_t bam_dma_irq(int irq, void *data)
+ {
+       struct bam_device *bdev = data;
+       u32 clr_mask = 0, srcs = 0;
+       srcs |= process_channel_irqs(bdev);
+       /* kick off tasklet to start next dma transfer */
+       if (srcs & P_IRQ)
+               tasklet_schedule(&bdev->task);
+       if (srcs & BAM_IRQ)
+               clr_mask = readl_relaxed(bam_addr(bdev, 0, BAM_IRQ_STTS));
+       /* don't allow reorder of the various accesses to the BAM registers */
+       mb();
+       writel_relaxed(clr_mask, bam_addr(bdev, 0, BAM_IRQ_CLR));
+       return IRQ_HANDLED;
+ }
+ /**
+  * bam_tx_status - returns status of transaction
+  * @chan: dma channel
+  * @cookie: transaction cookie
+  * @txstate: DMA transaction state
+  *
+  * Return status of dma transaction
+  */
+ static enum dma_status bam_tx_status(struct dma_chan *chan, dma_cookie_t cookie,
+               struct dma_tx_state *txstate)
+ {
+       struct bam_chan *bchan = to_bam_chan(chan);
+       struct virt_dma_desc *vd;
+       int ret;
+       size_t residue = 0;
+       unsigned int i;
+       unsigned long flags;
+       ret = dma_cookie_status(chan, cookie, txstate);
+       if (ret == DMA_COMPLETE)
+               return ret;
+       if (!txstate)
+               return bchan->paused ? DMA_PAUSED : ret;
+       spin_lock_irqsave(&bchan->vc.lock, flags);
+       vd = vchan_find_desc(&bchan->vc, cookie);
+       if (vd)
+               residue = container_of(vd, struct bam_async_desc, vd)->length;
+       else if (bchan->curr_txd && bchan->curr_txd->vd.tx.cookie == cookie)
+               for (i = 0; i < bchan->curr_txd->num_desc; i++)
+                       residue += bchan->curr_txd->curr_desc[i].size;
+       spin_unlock_irqrestore(&bchan->vc.lock, flags);
+       dma_set_residue(txstate, residue);
+       if (ret == DMA_IN_PROGRESS && bchan->paused)
+               ret = DMA_PAUSED;
+       return ret;
+ }
+ /**
+  * bam_apply_new_config
+  * @bchan: bam dma channel
+  * @dir: DMA direction
+  */
+ static void bam_apply_new_config(struct bam_chan *bchan,
+       enum dma_transfer_direction dir)
+ {
+       struct bam_device *bdev = bchan->bdev;
+       u32 maxburst;
+       if (dir == DMA_DEV_TO_MEM)
+               maxburst = bchan->slave.src_maxburst;
+       else
+               maxburst = bchan->slave.dst_maxburst;
+       writel_relaxed(maxburst, bam_addr(bdev, 0, BAM_DESC_CNT_TRSHLD));
+       bchan->reconfigure = 0;
+ }
+ /**
+  * bam_start_dma - start next transaction
+  * @bchan - bam dma channel
+  */
+ static void bam_start_dma(struct bam_chan *bchan)
+ {
+       struct virt_dma_desc *vd = vchan_next_desc(&bchan->vc);
+       struct bam_device *bdev = bchan->bdev;
+       struct bam_async_desc *async_desc;
+       struct bam_desc_hw *desc;
+       struct bam_desc_hw *fifo = PTR_ALIGN(bchan->fifo_virt,
+                                       sizeof(struct bam_desc_hw));
+       lockdep_assert_held(&bchan->vc.lock);
+       if (!vd)
+               return;
+       list_del(&vd->node);
+       async_desc = container_of(vd, struct bam_async_desc, vd);
+       bchan->curr_txd = async_desc;
+       /* on first use, initialize the channel hardware */
+       if (!bchan->initialized)
+               bam_chan_init_hw(bchan, async_desc->dir);
+       /* apply new slave config changes, if necessary */
+       if (bchan->reconfigure)
+               bam_apply_new_config(bchan, async_desc->dir);
+       desc = bchan->curr_txd->curr_desc;
+       if (async_desc->num_desc > MAX_DESCRIPTORS)
+               async_desc->xfer_len = MAX_DESCRIPTORS;
+       else
+               async_desc->xfer_len = async_desc->num_desc;
+       /* set any special flags on the last descriptor */
+       if (async_desc->num_desc == async_desc->xfer_len)
+               desc[async_desc->xfer_len - 1].flags =
+                                       cpu_to_le16(async_desc->flags);
+       else
+               desc[async_desc->xfer_len - 1].flags |=
+                                       cpu_to_le16(DESC_FLAG_INT);
+       if (bchan->tail + async_desc->xfer_len > MAX_DESCRIPTORS) {
+               u32 partial = MAX_DESCRIPTORS - bchan->tail;
+               memcpy(&fifo[bchan->tail], desc,
+                               partial * sizeof(struct bam_desc_hw));
+               memcpy(fifo, &desc[partial], (async_desc->xfer_len - partial) *
+                               sizeof(struct bam_desc_hw));
+       } else {
+               memcpy(&fifo[bchan->tail], desc,
+                       async_desc->xfer_len * sizeof(struct bam_desc_hw));
+       }
+       bchan->tail += async_desc->xfer_len;
+       bchan->tail %= MAX_DESCRIPTORS;
+       /* ensure descriptor writes and dma start not reordered */
+       wmb();
+       writel_relaxed(bchan->tail * sizeof(struct bam_desc_hw),
+                       bam_addr(bdev, bchan->id, BAM_P_EVNT_REG));
+ }
+ /**
+  * dma_tasklet - DMA IRQ tasklet
+  * @data: tasklet argument (bam controller structure)
+  *
+  * Sets up next DMA operation and then processes all completed transactions
+  */
+ static void dma_tasklet(unsigned long data)
+ {
+       struct bam_device *bdev = (struct bam_device *)data;
+       struct bam_chan *bchan;
+       unsigned long flags;
+       unsigned int i;
+       /* go through the channels and kick off transactions */
+       for (i = 0; i < bdev->num_channels; i++) {
+               bchan = &bdev->channels[i];
+               spin_lock_irqsave(&bchan->vc.lock, flags);
+               if (!list_empty(&bchan->vc.desc_issued) && !bchan->curr_txd)
+                       bam_start_dma(bchan);
+               spin_unlock_irqrestore(&bchan->vc.lock, flags);
+       }
+ }
+ /**
+  * bam_issue_pending - starts pending transactions
+  * @chan: dma channel
+  *
+  * Calls tasklet directly which in turn starts any pending transactions
+  */
+ static void bam_issue_pending(struct dma_chan *chan)
+ {
+       struct bam_chan *bchan = to_bam_chan(chan);
+       unsigned long flags;
+       spin_lock_irqsave(&bchan->vc.lock, flags);
+       /* if work pending and idle, start a transaction */
+       if (vchan_issue_pending(&bchan->vc) && !bchan->curr_txd)
+               bam_start_dma(bchan);
+       spin_unlock_irqrestore(&bchan->vc.lock, flags);
+ }
+ /**
+  * bam_dma_free_desc - free descriptor memory
+  * @vd: virtual descriptor
+  *
+  */
+ static void bam_dma_free_desc(struct virt_dma_desc *vd)
+ {
+       struct bam_async_desc *async_desc = container_of(vd,
+                       struct bam_async_desc, vd);
+       kfree(async_desc);
+ }
+ static struct dma_chan *bam_dma_xlate(struct of_phandle_args *dma_spec,
+               struct of_dma *of)
+ {
+       struct bam_device *bdev = container_of(of->of_dma_data,
+                                       struct bam_device, common);
+       unsigned int request;
+       if (dma_spec->args_count != 1)
+               return NULL;
+       request = dma_spec->args[0];
+       if (request >= bdev->num_channels)
+               return NULL;
+       return dma_get_slave_channel(&(bdev->channels[request].vc.chan));
+ }
+ /**
+  * bam_init
+  * @bdev: bam device
+  *
+  * Initialization helper for global bam registers
+  */
+ static int bam_init(struct bam_device *bdev)
+ {
+       u32 val;
+       /* read revision and configuration information */
+       val = readl_relaxed(bam_addr(bdev, 0, BAM_REVISION)) >> NUM_EES_SHIFT;
+       val &= NUM_EES_MASK;
+       /* check that configured EE is within range */
+       if (bdev->ee >= val)
+               return -EINVAL;
+       val = readl_relaxed(bam_addr(bdev, 0, BAM_NUM_PIPES));
+       bdev->num_channels = val & BAM_NUM_PIPES_MASK;
+       /* s/w reset bam */
+       /* after reset all pipes are disabled and idle */
+       val = readl_relaxed(bam_addr(bdev, 0, BAM_CTRL));
+       val |= BAM_SW_RST;
+       writel_relaxed(val, bam_addr(bdev, 0, BAM_CTRL));
+       val &= ~BAM_SW_RST;
+       writel_relaxed(val, bam_addr(bdev, 0, BAM_CTRL));
+       /* make sure previous stores are visible before enabling BAM */
+       wmb();
+       /* enable bam */
+       val |= BAM_EN;
+       writel_relaxed(val, bam_addr(bdev, 0, BAM_CTRL));
+       /* set descriptor threshhold, start with 4 bytes */
+       writel_relaxed(DEFAULT_CNT_THRSHLD,
+                       bam_addr(bdev, 0, BAM_DESC_CNT_TRSHLD));
+       /* Enable default set of h/w workarounds, ie all except BAM_FULL_PIPE */
+       writel_relaxed(BAM_CNFG_BITS_DEFAULT, bam_addr(bdev, 0, BAM_CNFG_BITS));
+       /* enable irqs for errors */
+       writel_relaxed(BAM_ERROR_EN | BAM_HRESP_ERR_EN,
+                       bam_addr(bdev, 0, BAM_IRQ_EN));
+       /* unmask global bam interrupt */
+       writel_relaxed(BAM_IRQ_MSK, bam_addr(bdev, 0, BAM_IRQ_SRCS_MSK_EE));
+       return 0;
+ }
+ static void bam_channel_init(struct bam_device *bdev, struct bam_chan *bchan,
+       u32 index)
+ {
+       bchan->id = index;
+       bchan->bdev = bdev;
+       vchan_init(&bchan->vc, &bdev->common);
+       bchan->vc.desc_free = bam_dma_free_desc;
+ }
+ static const struct of_device_id bam_of_match[] = {
+       { .compatible = "qcom,bam-v1.3.0", .data = &bam_v1_3_reg_info },
+       { .compatible = "qcom,bam-v1.4.0", .data = &bam_v1_4_reg_info },
+       { .compatible = "qcom,bam-v1.7.0", .data = &bam_v1_7_reg_info },
+       {}
+ };
+ MODULE_DEVICE_TABLE(of, bam_of_match);
+ static int bam_dma_probe(struct platform_device *pdev)
+ {
+       struct bam_device *bdev;
+       const struct of_device_id *match;
+       struct resource *iores;
+       int ret, i;
+       bdev = devm_kzalloc(&pdev->dev, sizeof(*bdev), GFP_KERNEL);
+       if (!bdev)
+               return -ENOMEM;
+       bdev->dev = &pdev->dev;
+       match = of_match_node(bam_of_match, pdev->dev.of_node);
+       if (!match) {
+               dev_err(&pdev->dev, "Unsupported BAM module\n");
+               return -ENODEV;
+       }
+       bdev->layout = match->data;
+       iores = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+       bdev->regs = devm_ioremap_resource(&pdev->dev, iores);
+       if (IS_ERR(bdev->regs))
+               return PTR_ERR(bdev->regs);
+       bdev->irq = platform_get_irq(pdev, 0);
+       if (bdev->irq < 0)
+               return bdev->irq;
+       ret = of_property_read_u32(pdev->dev.of_node, "qcom,ee", &bdev->ee);
+       if (ret) {
+               dev_err(bdev->dev, "Execution environment unspecified\n");
+               return ret;
+       }
+       bdev->bamclk = devm_clk_get(bdev->dev, "bam_clk");
+       if (IS_ERR(bdev->bamclk))
+               return PTR_ERR(bdev->bamclk);
+       ret = clk_prepare_enable(bdev->bamclk);
+       if (ret) {
+               dev_err(bdev->dev, "failed to prepare/enable clock\n");
+               return ret;
+       }
+       ret = bam_init(bdev);
+       if (ret)
+               goto err_disable_clk;
+       tasklet_init(&bdev->task, dma_tasklet, (unsigned long)bdev);
+       bdev->channels = devm_kcalloc(bdev->dev, bdev->num_channels,
+                               sizeof(*bdev->channels), GFP_KERNEL);
+       if (!bdev->channels) {
+               ret = -ENOMEM;
+               goto err_tasklet_kill;
+       }
+       /* allocate and initialize channels */
+       INIT_LIST_HEAD(&bdev->common.channels);
+       for (i = 0; i < bdev->num_channels; i++)
+               bam_channel_init(bdev, &bdev->channels[i], i);
+       ret = devm_request_irq(bdev->dev, bdev->irq, bam_dma_irq,
+                       IRQF_TRIGGER_HIGH, "bam_dma", bdev);
+       if (ret)
+               goto err_bam_channel_exit;
+       /* set max dma segment size */
+       bdev->common.dev = bdev->dev;
+       bdev->common.dev->dma_parms = &bdev->dma_parms;
+       ret = dma_set_max_seg_size(bdev->common.dev, BAM_MAX_DATA_SIZE);
+       if (ret) {
+               dev_err(bdev->dev, "cannot set maximum segment size\n");
+               goto err_bam_channel_exit;
+       }
+       platform_set_drvdata(pdev, bdev);
+       /* set capabilities */
+       dma_cap_zero(bdev->common.cap_mask);
+       dma_cap_set(DMA_SLAVE, bdev->common.cap_mask);
+       /* initialize dmaengine apis */
+       bdev->common.directions = BIT(DMA_DEV_TO_MEM) | BIT(DMA_MEM_TO_DEV);
+       bdev->common.residue_granularity = DMA_RESIDUE_GRANULARITY_SEGMENT;
+       bdev->common.src_addr_widths = DMA_SLAVE_BUSWIDTH_4_BYTES;
+       bdev->common.dst_addr_widths = DMA_SLAVE_BUSWIDTH_4_BYTES;
+       bdev->common.device_alloc_chan_resources = bam_alloc_chan;
+       bdev->common.device_free_chan_resources = bam_free_chan;
+       bdev->common.device_prep_slave_sg = bam_prep_slave_sg;
+       bdev->common.device_config = bam_slave_config;
+       bdev->common.device_pause = bam_pause;
+       bdev->common.device_resume = bam_resume;
+       bdev->common.device_terminate_all = bam_dma_terminate_all;
+       bdev->common.device_issue_pending = bam_issue_pending;
+       bdev->common.device_tx_status = bam_tx_status;
+       bdev->common.dev = bdev->dev;
+       ret = dma_async_device_register(&bdev->common);
+       if (ret) {
+               dev_err(bdev->dev, "failed to register dma async device\n");
+               goto err_bam_channel_exit;
+       }
+       ret = of_dma_controller_register(pdev->dev.of_node, bam_dma_xlate,
+                                       &bdev->common);
+       if (ret)
+               goto err_unregister_dma;
+       return 0;
+ err_unregister_dma:
+       dma_async_device_unregister(&bdev->common);
+ err_bam_channel_exit:
+       for (i = 0; i < bdev->num_channels; i++)
+               tasklet_kill(&bdev->channels[i].vc.task);
+ err_tasklet_kill:
+       tasklet_kill(&bdev->task);
+ err_disable_clk:
+       clk_disable_unprepare(bdev->bamclk);
+       return ret;
+ }
+ static int bam_dma_remove(struct platform_device *pdev)
+ {
+       struct bam_device *bdev = platform_get_drvdata(pdev);
+       u32 i;
+       of_dma_controller_free(pdev->dev.of_node);
+       dma_async_device_unregister(&bdev->common);
+       /* mask all interrupts for this execution environment */
+       writel_relaxed(0, bam_addr(bdev, 0,  BAM_IRQ_SRCS_MSK_EE));
+       devm_free_irq(bdev->dev, bdev->irq, bdev);
+       for (i = 0; i < bdev->num_channels; i++) {
+               bam_dma_terminate_all(&bdev->channels[i].vc.chan);
+               tasklet_kill(&bdev->channels[i].vc.task);
++              dma_free_wc(bdev->dev, BAM_DESC_FIFO_SIZE,
++                          bdev->channels[i].fifo_virt,
++                          bdev->channels[i].fifo_phys);
+       }
+       tasklet_kill(&bdev->task);
+       clk_disable_unprepare(bdev->bamclk);
+       return 0;
+ }
+ static struct platform_driver bam_dma_driver = {
+       .probe = bam_dma_probe,
+       .remove = bam_dma_remove,
+       .driver = {
+               .name = "bam-dma-engine",
+               .of_match_table = bam_of_match,
+       },
+ };
+ module_platform_driver(bam_dma_driver);
+ MODULE_AUTHOR("Andy Gross <agross@codeaurora.org>");
+ MODULE_DESCRIPTION("QCOM BAM DMA engine driver");
+ MODULE_LICENSE("GPL v2");
Simple merge