Re: [PATCH 3/3] dmaengine: Add Freescale i.MX SDMA support

From: Linus Walleij
Date: Mon Aug 16 2010 - 08:21:15 EST


2010/8/16 Sascha Hauer <s.hauer@xxxxxxxxxxxxxx>:

> This patch adds support for the Freescale i.MX SDMA engine.

I like it!

> The SDMA engine is a scatter/gather DMA engine which is implemented
> as a seperate coprocessor. SDMA needs its own firmware which is
> requested using the standard request_firmware mechanism. The firmware
> has different entry points for each peripheral type, so drivers
> have to pass the peripheral type to the DMA engine which in turn
> picks the correct firmware entry point from a table contained in
> the firmware image itself.

Quite fun, if the spec for the microcode is open this opens up
for dynamic firmware generation for specific DMA jobs does it
not?

> I took a very simple approach to implement dmaengine support. Only
> a single descriptor is statically assigned to a each channel. This
> means that transfers can't be queued up but only a single transfer
> is in progress. This simplifies implementation a lot and is sufficient
> for the usual device/memory transfers.

If you want to add memcpy() capability later you're gonna need
this I think, but you can take that when that need arise.

>(...)
> +++ b/arch/arm/plat-mxc/include/mach/dma.h
> @@ -0,0 +1,64 @@
> +/*
> + * Copyright 2004-2009 Freescale Semiconductor, Inc. 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 as
> + * published by the Free Software Foundation.
> + */
> +
> +#ifndef __ASM_ARCH_MXC_DMA_H__
> +#define __ASM_ARCH_MXC_DMA_H__
> +
> +#include <linux/scatterlist.h>
> +
> +/*
> + * This enumerates peripheral types. Used for SDMA.
> + */
> +typedef enum {

The kernel is not really happy about typedefs, can't this be a
regular enum?

> +       IMX_DMATYPE_SSI,        /* MCU domain SSI */
> +       IMX_DMATYPE_SSI_SP,     /* Shared SSI */
> +       IMX_DMATYPE_MMC,        /* MMC */
> +       IMX_DMATYPE_SDHC,       /* SDHC */
> +       IMX_DMATYPE_UART,       /* MCU domain UART */
> +       IMX_DMATYPE_UART_SP,    /* Shared UART */
> +       IMX_DMATYPE_FIRI,       /* FIRI */
> +       IMX_DMATYPE_CSPI,       /* MCU domain CSPI */
> +       IMX_DMATYPE_CSPI_SP,    /* Shared CSPI */
> +       IMX_DMATYPE_SIM,        /* SIM */
> +       IMX_DMATYPE_ATA,        /* ATA */
> +       IMX_DMATYPE_CCM,        /* CCM */
> +       IMX_DMATYPE_EXT,        /* External peripheral */
> +       IMX_DMATYPE_MSHC,       /* Memory Stick Host Controller */
> +       IMX_DMATYPE_MSHC_SP,    /* Shared Memory Stick Host Controller */
> +       IMX_DMATYPE_DSP,        /* DSP */
> +       IMX_DMATYPE_MEMORY,     /* Memory */
> +       IMX_DMATYPE_FIFO_MEMORY,/* FIFO type Memory */
> +       IMX_DMATYPE_SPDIF,      /* SPDIF */
> +       IMX_DMATYPE_IPU_MEMORY, /* IPU Memory */
> +       IMX_DMATYPE_ASRC,       /* ASRC */
> +       IMX_DMATYPE_ESAI,       /* ESAI */
> +} sdma_peripheral_type;
> +
> +enum imx_dma_prio {
> +       DMA_PRIO_HIGH = 0,
> +       DMA_PRIO_MEDIUM = 1,
> +       DMA_PRIO_LOW = 2
> +};
> +
> +struct imx_dma_data {
> +       int dma_request; /* DMA request line */

Can this be negative and what is the range? I would
suspect something like u8 or u16 would surely be more
apropriate...

> +       sdma_peripheral_type peripheral_type;
> +       int priority;

Isn't this an enum imx_dma_prio?

> +};
> +
> +static inline int imx_dma_is_ipu(struct dma_chan *chan)
> +{
> +       return !strcmp(dev_name(chan->device->dev), "ipu-core");
> +}
> +
> +static inline int imx_dma_is_general_purpose(struct dma_chan *chan)
> +{
> +       return !strcmp(dev_name(chan->device->dev), "imx-sdma");
> +}
> +
> +#endif
> diff --git a/arch/arm/plat-mxc/include/mach/sdma.h b/arch/arm/plat-mxc/include/mach/sdma.h
> new file mode 100644
> index 0000000..5d542b8
> --- /dev/null
> +++ b/arch/arm/plat-mxc/include/mach/sdma.h
> @@ -0,0 +1,8 @@
> +#ifndef __MACH_MXC_SDMA_H__
> +#define __MACH_MXC_SDMA_H__
> +
> +struct sdma_platform_data {
> +       int sdma_version;

Do you have negative versions or can it be unsigned?

> +};
> +
> +#endif /* __MACH_MXC_SDMA_H__ */
> diff --git a/drivers/dma/Kconfig b/drivers/dma/Kconfig
> index 9520cf0..f76bda9 100644
> --- a/drivers/dma/Kconfig
> +++ b/drivers/dma/Kconfig
> @@ -195,6 +195,14 @@ config PCH_DMA
>        help
>          Enable support for the Topcliff PCH DMA engine.
>
> +config IMX_SDMA
> +       tristate "Atmel AHB DMA support"
> +       depends on ARCH_MXC
> +       select DMA_ENGINE
> +       help
> +         Support the i.MX SDMA engine. This engine is integrated into
> +         Freescale i.MX25/31/35/51 chips.
> +
>  config DMA_ENGINE
>        bool
>
> diff --git a/drivers/dma/Makefile b/drivers/dma/Makefile
> index 72bd703..14d7a1b 100644
> --- a/drivers/dma/Makefile
> +++ b/drivers/dma/Makefile
> @@ -25,3 +25,4 @@ obj-$(CONFIG_TIMB_DMA) += timb_dma.o
>  obj-$(CONFIG_STE_DMA40) += ste_dma40.o ste_dma40_ll.o
>  obj-$(CONFIG_PL330_DMA) += pl330.o
>  obj-$(CONFIG_PCH_DMA) += pch_dma.o
> +obj-$(CONFIG_IMX_SDMA) += imx-sdma.o
> diff --git a/drivers/dma/imx-sdma.c b/drivers/dma/imx-sdma.c
> new file mode 100644
> index 0000000..3ba7905
> --- /dev/null
> +++ b/drivers/dma/imx-sdma.c
> @@ -0,0 +1,1383 @@
> +/*
> + * drivers/dma/imx-sdma.c
> + *
> + * This file contains a driver for the Freescale Smart DMA engine
> + *
> + * Copyright 2010 Sascha Hauer, Pengutronix <s.hauer@xxxxxxxxxxxxxx>
> + *
> + * Based on code from Freescale:
> + *
> + * Copyright 2004-2009 Freescale Semiconductor, Inc. All Rights Reserved.
> + *
> + * The code contained herein is licensed under the GNU General Public
> + * License. You may obtain a copy of the GNU General Public License
> + * Version 2 or later at the following locations:
> + *
> + * http://www.opensource.org/licenses/gpl-license.html
> + * http://www.gnu.org/copyleft/gpl.html
> + */
> +#include <linux/init.h>
> +#include <linux/types.h>
> +#include <linux/mm.h>
> +#include <linux/interrupt.h>
> +#include <linux/clk.h>
> +#include <linux/semaphore.h>
> +#include <linux/spinlock.h>
> +#include <linux/device.h>
> +#include <linux/dma-mapping.h>
> +#include <linux/firmware.h>
> +#include <linux/slab.h>
> +#include <linux/platform_device.h>
> +#include <linux/dmaengine.h>
> +
> +#include <asm/irq.h>
> +#include <mach/sdma.h>
> +#include <mach/dma.h>
> +#include <mach/hardware.h>
> +
> +/* SDMA registers */
> +#define SDMA_H_C0PTR           (sdma_base + 0x000)
> +#define SDMA_H_INTR            (sdma_base + 0x004)
> +#define SDMA_H_STATSTOP                (sdma_base + 0x008)
> +#define SDMA_H_START           (sdma_base + 0x00c)
> +#define SDMA_H_EVTOVR          (sdma_base + 0x010)
> +#define SDMA_H_DSPOVR          (sdma_base + 0x014)
> +#define SDMA_H_HOSTOVR         (sdma_base + 0x018)
> +#define SDMA_H_EVTPEND         (sdma_base + 0x01c)
> +#define SDMA_H_DSPENBL         (sdma_base + 0x020)
> +#define SDMA_H_RESET           (sdma_base + 0x024)
> +#define SDMA_H_EVTERR          (sdma_base + 0x028)
> +#define SDMA_H_INTRMSK         (sdma_base + 0x02c)
> +#define SDMA_H_PSW             (sdma_base + 0x030)
> +#define SDMA_H_EVTERRDBG       (sdma_base + 0x034)
> +#define SDMA_H_CONFIG          (sdma_base + 0x038)
> +#define SDMA_ONCE_ENB          (sdma_base + 0x040)
> +#define SDMA_ONCE_DATA         (sdma_base + 0x044)
> +#define SDMA_ONCE_INSTR                (sdma_base + 0x048)
> +#define SDMA_ONCE_STAT         (sdma_base + 0x04c)
> +#define SDMA_ONCE_CMD          (sdma_base + 0x050)
> +#define SDMA_EVT_MIRROR                (sdma_base + 0x054)
> +#define SDMA_ILLINSTADDR       (sdma_base + 0x058)
> +#define SDMA_CHN0ADDR          (sdma_base + 0x05c)
> +#define SDMA_ONCE_RTB          (sdma_base + 0x060)
> +#define SDMA_XTRIG_CONF1       (sdma_base + 0x070)
> +#define SDMA_XTRIG_CONF2       (sdma_base + 0x074)
> +#define SDMA_CHNENBL_0         (sdma_base + (sdma_version == 2 ? 0x200 : 0x80))
> +#define SDMA_CHNPRI_0          (sdma_base + 0x100)

All these rely on a fixed sdma_base which makes the driver
a singleton. This is not so good if you imagine the situation with a
platform with two SDMA engines on different addresses.

Can't you create a runtime allocated stateholder to hold
the base and access relative to the offset?

> +
> +/*
> + * Buffer descriptor status values.
> + */
> +#define BD_DONE  0x01
> +#define BD_WRAP  0x02
> +#define BD_CONT  0x04
> +#define BD_INTR  0x08
> +#define BD_RROR  0x10
> +#define BD_LAST  0x20
> +#define BD_EXTD  0x80
> +
> +/*
> + * Data Node descriptor status values.
> + */
> +#define DND_END_OF_FRAME  0x80
> +#define DND_END_OF_XFER   0x40
> +#define DND_DONE          0x20
> +#define DND_UNUSED        0x01
> +
> +/*
> + * IPCV2 descriptor status values.
> + */
> +#define BD_IPCV2_END_OF_FRAME  0x40
> +
> +#define IPCV2_MAX_NODES        50
> +/*
> + * Error bit set in the CCB status field by the SDMA,
> + * in setbd routine, in case of a transfer error
> + */
> +#define DATA_ERROR  0x10000000
> +
> +/*
> + * Buffer descriptor commands.
> + */
> +#define C0_ADDR             0x01
> +#define C0_LOAD             0x02
> +#define C0_DUMP             0x03
> +#define C0_SETCTX           0x07
> +#define C0_GETCTX           0x03
> +#define C0_SETDM            0x01
> +#define C0_SETPM            0x04
> +#define C0_GETDM            0x02
> +#define C0_GETPM            0x08
> +/*
> + * Change endianness indicator in the BD command field
> + */
> +#define CHANGE_ENDIANNESS   0x80
> +
> +/*
> + * Mode/Count of data node descriptors - IPCv2
> + */
> +#ifdef __BIG_ENDIAN
> +struct sdma_mode_count {
> +       u32 command :  8; /* command mostlky used for channel 0 */

There are a lot of inline commented struct members, please
use kerneldoc, that's simple. (Applies all over the patch...)
Documentation/kernel-doc-nano-HOWTO

> +       u32 status  :  8; /* E,R,I,C,W,D status bits stored here */
> +       u32 count   : 16; /* size of the buffer pointed by this BD */
> +};
> +#else
> +struct sdma_mode_count {
> +       u32 count   : 16; /* size of the buffer pointed by this BD */
> +       u32 status  :  8; /* E,R,I,C,W,D status bits stored here */
> +       u32 command :  8; /* command mostlky used for channel 0 */
> +};
> +#endif

This use of #ifdef is odd to me but others are probably more
experienced. Anyway, the way it is used with different
:n suffixes makes me believe that you need a packed
compiler directive for this layout to be explicitly coherent.

Atleast add some comment on what this #ifdef construction
does so guys like me can understand what's going on.

> +
> +/*
> + * Buffer descriptor
> + */
> +struct sdma_buffer_descriptor {
> +       struct sdma_mode_count  mode;
> +       u32 buffer_addr;    /* address of the buffer described */
> +       u32 ext_buffer_addr; /* extended buffer address */

Shouldn't these be dma_addr_t? OK that's probably u32
anyway but just to make a marker...

> +};
> +
> +/*
> + * Channel control Block
> + */
> +struct sdma_channel_control {
> +       u32 current_bd_ptr; /* current buffer descriptor processed */
> +       u32 base_bd_ptr;    /* first element of buffer descriptor array */
> +       void *unused;
> +       void *unused1;

Hm, can you comment on what these unused things are for...?

> +};
> +
> +/**
> + * Context structure.
> + */
> +#ifdef __BIG_ENDIAN
> +struct sdma_state_registers {
> +       u32 sf     : 1; /* source fault while loading data */
> +       u32 unused0: 1;
> +       u32 rpc    :14; /* return program counter */
> +       u32 t      : 1; /* test bit:status of arithmetic & test instruction*/
> +       u32 unused1: 1;
> +       u32 pc     :14; /* program counter */
> +       u32 lm     : 2; /* loop mode */
> +       u32 epc    :14; /* loop end program counter */
> +       u32 df     : 1; /* destination fault while storing data */
> +       u32 unused2: 1;
> +       u32 spc    :14; /* loop start program counter */
> +};
> +#else
> +struct sdma_state_registers {
> +       u32 pc     :14; /* program counter */
> +       u32 unused1: 1;
> +       u32 t      : 1; /* test bit: status of arithmetic & test instruction*/
> +       u32 rpc    :14; /* return program counter */
> +       u32 unused0: 1;
> +       u32 sf     : 1; /* source fault while loading data */
> +       u32 spc    :14; /* loop start program counter */
> +       u32 unused2: 1;
> +       u32 df     : 1; /* destination fault while storing data */
> +       u32 epc    :14; /* loop end program counter */
> +       u32 lm     : 2; /* loop mode */
> +};
> +#endif

Again this is odd to me...

> +
> +struct sdma_context_data {
> +       struct sdma_state_registers  channel_state; /* channel state bits */
> +       u32  gReg[8]; /* general registers */
> +       u32  mda; /* burst dma destination address register */
> +       u32  msa; /* burst dma source address register */
> +       u32  ms;  /* burst dma status  register */
> +       u32  md;  /* burst dma data    register */
> +       u32  pda; /* peripheral dma destination address register */
> +       u32  psa; /* peripheral dma source address register */
> +       u32  ps;  /* peripheral dma  status  register */
> +       u32  pd;  /* peripheral dma  data    register */
> +       u32  ca;  /* CRC polynomial  register */
> +       u32  cs;  /* CRC accumulator register */
> +       u32  dda; /* dedicated core destination address register */
> +       u32  dsa; /* dedicated core source address register */
> +       u32  ds;  /* dedicated core status  register */
> +       u32  dd;  /* dedicated core data    register */
> +       u32  scratch0;
> +       u32  scratch1;
> +       u32  scratch2;
> +       u32  scratch3;
> +       u32  scratch4;
> +       u32  scratch5;
> +       u32  scratch6;
> +       u32  scratch7;
> +};
> +
> +struct sdma_channel {
> +       /* Channel number */
> +       int channel;

Unsigned?

> +       /* Transfer type. Needed for setting SDMA script */
> +       enum dma_data_direction direction;
> +       /* Peripheral type. Needed for setting SDMA script */
> +       sdma_peripheral_type peripheral_type;
> +       /* Peripheral event id */
> +       int event_id;

Unsigned?

> +       /* Peripheral event id2 (for channels that use 2 events) */
> +       int event_id2;

Unsigned?

> +       /* SDMA data access word size */
> +       unsigned long word_size;

Is this in bits, bytes etc? Isn't e.g. an u8 enough to hold this,
and further, isn't it possible to recycle enum dma_slave_buswidth
from dmaengine.h instead?

> +
> +       /* ID of the buffer that was processed */
> +       unsigned int buf_tail;
> +
> +       wait_queue_head_t waitq;        /* channel completion waitqeue */
> +
> +       int num_bd;

Unsigned? Range?

> +
> +       struct sdma_buffer_descriptor *bd;
> +       dma_addr_t      bd_phys;
> +
> +       int pc_from_device, pc_to_device;

Unsigned?

> +
> +       unsigned long flags;

Is this an u32?

> +       dma_addr_t per_address;
> +
> +       uint32_t event_mask1, event_mask2;
> +       uint32_t watermark_level;
> +       uint32_t shp_addr, per_addr;
> +
> +       /* DMA-Engine Channel */
> +       struct dma_chan chan;
> +
> +       spinlock_t              lock;
> +       struct dma_async_tx_descriptor desc;
> +       dma_cookie_t            last_completed;
> +       int busy;

Shouldn't this be a bool?

> +};
> +
> +#define IMX_DMA_SG_LOOP                (1 << 0)
> +
> +#define MAX_DMA_CHANNELS 32
> +#define MXC_SDMA_DEFAULT_PRIORITY 1
> +#define MXC_SDMA_MIN_PRIORITY 1
> +#define MXC_SDMA_MAX_PRIORITY 7
> +
> +/*
> + * This enumerates transfer types
> + */
> +typedef enum {

Again a typedef, please plain enum is fine.

> +       emi_2_per = 0,          /* EMI memory to peripheral */
> +       emi_2_int,              /* EMI memory to internal RAM */
> +       emi_2_emi,              /* EMI memory to EMI memory */
> +       emi_2_dsp,              /* EMI memory to DSP memory */
> +       per_2_int,              /* Peripheral to internal RAM */
> +       per_2_emi,              /* Peripheral to internal EMI memory */
> +       per_2_dsp,              /* Peripheral to DSP memory */
> +       per_2_per,              /* Peripheral to Peripheral */
> +       int_2_per,              /* Internal RAM to peripheral */
> +       int_2_int,              /* Internal RAM to Internal RAM */
> +       int_2_emi,              /* Internal RAM to EMI memory */
> +       int_2_dsp,              /* Internal RAM to DSP memory */
> +       dsp_2_per,              /* DSP memory to peripheral */
> +       dsp_2_int,              /* DSP memory to internal RAM */
> +       dsp_2_emi,              /* DSP memory to EMI memory */
> +       dsp_2_dsp,              /* DSP memory to DSP memory */
> +       emi_2_dsp_loop,         /* EMI memory to DSP memory loopback */
> +       dsp_2_emi_loop,         /* DSP memory to EMI memory loopback */
> +       dvfs_pll,               /* DVFS script with PLL change       */
> +       dvfs_pdr                /* DVFS script without PLL change    */
> +} sdma_transfer_type;
> +
> +/*
> + * Structure containing sdma request  parameters.
> + */
> +struct sdma_script_start_addrs {
> +       int ap_2_ap_addr;
> +       int ap_2_bp_addr;
> +       int ap_2_ap_fixed_addr;
> +       int bp_2_ap_addr;
> +       int loopback_on_dsp_side_addr;
> +       int mcu_interrupt_only_addr;
> +
> +       int firi_2_per_addr;
> +       int firi_2_mcu_addr;
> +       int per_2_firi_addr;
> +       int mcu_2_firi_addr;
> +
> +       int uart_2_per_addr;
> +       int uart_2_mcu_addr;
> +       int per_2_app_addr;
> +       int mcu_2_app_addr;
> +       int per_2_per_addr;
> +
> +       int uartsh_2_per_addr;
> +       int uartsh_2_mcu_addr;
> +       int per_2_shp_addr;
> +       int mcu_2_shp_addr;
> +
> +       int ata_2_mcu_addr;
> +       int mcu_2_ata_addr;
> +
> +       int app_2_per_addr;
> +       int app_2_mcu_addr;
> +       int shp_2_per_addr;
> +       int shp_2_mcu_addr;
> +
> +       int mshc_2_mcu_addr;
> +       int mcu_2_mshc_addr;
> +
> +       int spdif_2_mcu_addr;
> +       int mcu_2_spdif_addr;
> +
> +       int asrc_2_mcu_addr;
> +
> +       int ext_mem_2_ipu_addr;
> +
> +       int descrambler_addr;
> +
> +       int dptc_dvfs_addr;
> +
> +       int utra_addr;
> +
> +       int ram_code_start_addr;

All these addresses, are they really integers with
valid negative values... Aren't they dma_addr_t or
atleast u32?

> +};
> +
> +#define SDMA_FIRMWARE_MAGIC 0x414d4453
> +
> +struct sdma_firmware_header {
> +       uint32_t        magic; /* "SDMA" */
> +       uint32_t        version_major;  /* increased whenever layout of struct sdma_script_start_addrs changes */
> +       uint32_t        version_minor;  /* firmware version */
> +       uint32_t        script_addrs_start; /* offset of struct sdma_script_start_addrs in this image */
> +       uint32_t        num_script_addrs; /* Number of script addresses in this image */
> +       uint32_t        ram_code_start; /* offset of SDMA ram image in this firmware image */
> +       uint32_t        ram_code_size; /* size of SDMA ram image */

Please use u32. uint32_t is not the preferred kernel type.
(Still I've seen people use it in some cases so I might be wrong,
feel welcome to bit back on this.)

> +};
> +
> +static struct sdma_channel sdma_data[MAX_DMA_CHANNELS];
> +static struct sdma_channel_control *channel_control;
> +static void __iomem *sdma_base;
> +static int sdma_version;

Unsigned?

> +static int sdma_num_events;

Unsigned?

> +static struct sdma_context_data *sdma_context;
> +dma_addr_t sdma_context_phys;
> +static struct dma_device __sdma_dma_device;
> +static struct dma_device *sdma_dma_device = &__sdma_dma_device;

This is what I suspected: local variables making the entire driver
a singleton, which means you can never have more than one
SDMA. Atleast collect all of these in a struct, call it
"struct sdma" simply (if you ask me) and use as a stateholder.
This makes it easier to kzalloc() that struct later if you
want to support non-singletons.

I know this require some work but I've done it to several drivers
(always asked on mailinglists to do this) and I don't regret a single
rewrite. Last time was for the PL18x DMAengine driver actually.

> +
> +#define SDMA_H_CONFIG_DSPDMA   (1 << 12) /* indicates if the DSPDMA is used */
> +#define SDMA_H_CONFIG_RTD_PINS (1 << 11) /* indicates if Real-Time Debug pins are enabled */
> +#define SDMA_H_CONFIG_ACR      (1 << 4)  /* indicates if AHB freq /core freq = 2 or 1 */
> +#define SDMA_H_CONFIG_CSM      (3)       /* indicates which context switch mode is selected*/
> +
> +static int sdma_config_ownership(int channel, int event_override,
> +                  int mcu_verride, int dsp_override)
> +{
> +       u32 evt, mcu, dsp;
> +
> +       if (event_override && mcu_verride && dsp_override)
> +               return -EINVAL;
> +
> +       evt = readl(SDMA_H_EVTOVR);
> +       mcu = readl(SDMA_H_HOSTOVR);
> +       dsp = readl(SDMA_H_DSPOVR);
> +
> +       if (dsp_override)
> +               dsp &= ~(1 << channel);
> +       else
> +               dsp |= (1 << channel);
> +
> +       if (event_override)
> +               evt &= ~(1 << channel);
> +       else
> +               evt |= (1 << channel);
> +
> +       if (mcu_verride)
> +               mcu &= ~(1 << channel);
> +       else
> +               mcu |= (1 << channel);
> +
> +       writel(evt, SDMA_H_EVTOVR);
> +       writel(mcu, SDMA_H_HOSTOVR);
> +       writel(dsp, SDMA_H_DSPOVR);
> +
> +       return 0;
> +}
> +
> +/*
> + * sdma_run_channel - run a channel and wait till it's done
> + */
> +static int sdma_run_channel(int channel)
> +{
> +       struct sdma_channel *sdma = &sdma_data[channel];
> +       int ret;
> +
> +       writel(1 << channel, SDMA_H_START);
> +
> +       ret = wait_event_interruptible(sdma->waitq,
> +                       !(readl(SDMA_H_STATSTOP) & (1 << channel)));

OK not the biggest thing in the world, but can't you use a
completion for this? (I'm not so clever with waitqueues so
forgive me if this is malinformed.)

> +       return ret;
> +}
> +
> +static int sdma_load_script(void *buf, int size, u32 address)
> +{
> +       struct sdma_buffer_descriptor *bd0 = sdma_data[0].bd;
> +       void *buf_virt;
> +       dma_addr_t buf_phys;
> +       int ret;
> +
> +       buf_virt = dma_alloc_coherent(NULL,
> +                       size,
> +                       &buf_phys, GFP_KERNEL);
> +       if (!buf_virt)
> +               return -ENOMEM;
> +
> +       bd0->mode.command = C0_SETPM;
> +       bd0->mode.status = BD_DONE | BD_INTR | BD_WRAP | BD_EXTD;
> +       bd0->mode.count = size / 2;
> +       bd0->buffer_addr = buf_phys;
> +       bd0->ext_buffer_addr = address;
> +
> +       memcpy(buf_virt, buf, size);
> +
> +       ret = sdma_run_channel(0);
> +
> +       dma_free_coherent(NULL, size, buf_virt, buf_phys);
> +
> +       return ret;
> +}
> +
> +static void sdma_event_enable(int channel, int event)
> +{
> +       u32 val;
> +
> +       val = readl(SDMA_CHNENBL_0 + event * 4);

This use indicates that event should probably be
unsigned, and probably not greater than u16 atleast.
I suspect it is never more than an u8 really.

> +       val |= (1 << channel);
> +       writel(val, SDMA_CHNENBL_0 + event * 4);
> +}
> +
> +static void sdma_event_disable(int channel, int event)
> +{
> +       u32 val;
> +
> +       val = readl(SDMA_CHNENBL_0 + event * 4);
> +       val &= ~(1 << channel);
> +       writel(val, SDMA_CHNENBL_0 + event * 4);

Same comment here.

> +}
> +
> +static void mxc_sdma_handle_channel_loop(int channel)
> +{
> +       struct sdma_channel *sdma = &sdma_data[channel];

This indicates that channel should be unsigned.

> +       struct sdma_buffer_descriptor *bd;
> +       int error = 0;

Unused variable?

> +
> +       /*
> +        * loop mode. Iterate over descriptors, re-setup them and
> +        * call callback function.
> +        */
> +       while (1) {
> +               bd = &sdma->bd[sdma->buf_tail];
> +
> +               if (bd->mode.status & BD_DONE)
> +                       break;
> +
> +               if (bd->mode.status & BD_RROR)
> +                       error = -EIO;
> +
> +               bd->mode.status |= BD_DONE;
> +               sdma->buf_tail++;
> +               sdma->buf_tail %= sdma->num_bd;
> +
> +               if (sdma->desc.callback)
> +                       sdma->desc.callback(sdma->desc.callback_param);
> +       }
> +}
> +
> +static void mxc_sdma_handle_channel_normal(int channel)
> +{
> +       struct sdma_channel *sdma = &sdma_data[channel];
> +       struct sdma_buffer_descriptor *bd;
> +       int i, error = 0;
> +
> +       /*
> +        * non loop mode. Iterate over all descriptors, collect
> +        * errors and call callback function
> +        */
> +       for (i = 0; i < sdma->num_bd; i++) {
> +               bd = &sdma->bd[i];
> +
> +                if (bd->mode.status & (BD_DONE | BD_RROR))
> +                       error = -EIO;
> +       }
> +
> +       if (sdma->desc.callback)
> +               sdma->desc.callback(sdma->desc.callback_param);
> +       sdma->last_completed = sdma->desc.cookie;
> +
> +       sdma->busy = 0;

= true if you switch this to bool..

> +}
> +
> +static void mxc_sdma_handle_channel(int channel)
> +{
> +       struct sdma_channel *sdma = &sdma_data[channel];
> +
> +       wake_up_interruptible(&sdma->waitq);
> +
> +       /* not interested in channel 0 interrupts */
> +       if (!channel)
> +               return;
> +
> +       if (sdma->flags & IMX_DMA_SG_LOOP)
> +               mxc_sdma_handle_channel_loop(channel);
> +       else
> +               mxc_sdma_handle_channel_normal(channel);
> +}
> +
> +static irqreturn_t sdma_int_handler(int irq, void *dev_id)
> +{
> +       u32 stat;
> +
> +       stat = readl(SDMA_H_INTR);
> +       writel(stat, SDMA_H_INTR);
> +
> +       while (stat) {
> +               int channel = fls(stat) - 1;
> +
> +               mxc_sdma_handle_channel(channel);
> +
> +               stat &= ~(1 << channel);
> +       }
> +
> +       return IRQ_HANDLED;
> +}
> +
> +static struct clk *sdma_clk;
> +
> +/*
> + * Stores the start address of the SDMA scripts
> + */
> +static struct sdma_script_start_addrs __sdma_script_addrs;
> +static struct sdma_script_start_addrs *sdma_script_addrs = &__sdma_script_addrs;
> +
> +/*
> + * sets the pc of SDMA script according to the peripheral type
> + */
> +static void sdma_get_pc(struct sdma_channel *sdma,
> +               sdma_peripheral_type peripheral_type)
> +{
> +       int res = 0;
> +       int per_2_emi = 0, emi_2_per = 0;
> +       int per_2_int = 0, int_2_per = 0;
> +       int per_2_per = 0, emi_2_emi = 0;
> +
> +       sdma->pc_from_device = 0;
> +       sdma->pc_to_device = 0;

There are a *lot* of local variables here, and only two of them
are used eventually, at the end of the function. I cannot quite
follow this, what is going on?

Some like emi_2_emi seem to be totally unused.

The types here look like some kind of enum or other
similar construction is really what's being asked for
here.

> +
> +       switch (peripheral_type) {
> +       case IMX_DMATYPE_MEMORY:
> +               emi_2_emi = sdma_script_addrs->ap_2_ap_addr;
> +               break;
> +       case IMX_DMATYPE_DSP:
> +               emi_2_per = sdma_script_addrs->bp_2_ap_addr;
> +               per_2_emi = sdma_script_addrs->ap_2_bp_addr;
> +               break;
> +       case IMX_DMATYPE_FIRI:
> +               per_2_int = sdma_script_addrs->firi_2_per_addr;
> +               per_2_emi = sdma_script_addrs->firi_2_mcu_addr;
> +               int_2_per = sdma_script_addrs->per_2_firi_addr;
> +               emi_2_per = sdma_script_addrs->mcu_2_firi_addr;
> +               break;
> +       case IMX_DMATYPE_UART:
> +               per_2_int = sdma_script_addrs->uart_2_per_addr;
> +               per_2_emi = sdma_script_addrs->uart_2_mcu_addr;
> +               int_2_per = sdma_script_addrs->per_2_app_addr;
> +               emi_2_per = sdma_script_addrs->mcu_2_app_addr;
> +               break;
> +       case IMX_DMATYPE_UART_SP:
> +               per_2_int = sdma_script_addrs->uartsh_2_per_addr;
> +               per_2_emi = sdma_script_addrs->uartsh_2_mcu_addr;
> +               int_2_per = sdma_script_addrs->per_2_shp_addr;
> +               emi_2_per = sdma_script_addrs->mcu_2_shp_addr;
> +               break;
> +       case IMX_DMATYPE_ATA:
> +               per_2_emi = sdma_script_addrs->ata_2_mcu_addr;
> +               emi_2_per = sdma_script_addrs->mcu_2_ata_addr;
> +               break;
> +       case IMX_DMATYPE_CSPI:
> +       case IMX_DMATYPE_EXT:
> +       case IMX_DMATYPE_SSI:
> +               per_2_int = sdma_script_addrs->app_2_per_addr;
> +               per_2_emi = sdma_script_addrs->app_2_mcu_addr;
> +               int_2_per = sdma_script_addrs->per_2_app_addr;
> +               emi_2_per = sdma_script_addrs->mcu_2_app_addr;
> +               break;
> +       case IMX_DMATYPE_SSI_SP:
> +       case IMX_DMATYPE_MMC:
> +       case IMX_DMATYPE_SDHC:
> +       case IMX_DMATYPE_CSPI_SP:
> +       case IMX_DMATYPE_ESAI:
> +       case IMX_DMATYPE_MSHC_SP:
> +               per_2_int = sdma_script_addrs->shp_2_per_addr;
> +               per_2_emi = sdma_script_addrs->shp_2_mcu_addr;
> +               int_2_per = sdma_script_addrs->per_2_shp_addr;
> +               emi_2_per = sdma_script_addrs->mcu_2_shp_addr;
> +               break;
> +       case IMX_DMATYPE_ASRC:
> +               per_2_emi = sdma_script_addrs->asrc_2_mcu_addr;
> +               emi_2_per = sdma_script_addrs->asrc_2_mcu_addr;
> +               per_2_per = sdma_script_addrs->per_2_per_addr;
> +               break;
> +       case IMX_DMATYPE_MSHC:
> +               per_2_emi = sdma_script_addrs->mshc_2_mcu_addr;
> +               emi_2_per = sdma_script_addrs->mcu_2_mshc_addr;
> +               break;
> +       case IMX_DMATYPE_CCM:
> +               per_2_emi = sdma_script_addrs->dptc_dvfs_addr;
> +               break;
> +       case IMX_DMATYPE_FIFO_MEMORY:
> +               res = sdma_script_addrs->ap_2_ap_fixed_addr;

res? This thing is never used.

> +               break;
> +       case IMX_DMATYPE_SPDIF:
> +               per_2_emi = sdma_script_addrs->spdif_2_mcu_addr;
> +               emi_2_per = sdma_script_addrs->mcu_2_spdif_addr;
> +               break;
> +       case IMX_DMATYPE_IPU_MEMORY:
> +               emi_2_per = sdma_script_addrs->ext_mem_2_ipu_addr;
> +               break;
> +       default:
> +               break;
> +       }
> +
> +       sdma->pc_from_device = per_2_emi;
> +       sdma->pc_to_device = emi_2_per;

Return res? You're assigning it a value in some cases.

> +}
> +
> +static int sdma_load_context(int channel)
> +{
> +       struct sdma_channel *sdma = &sdma_data[channel];
> +       int load_address;
> +       struct sdma_buffer_descriptor *bd0 = sdma_data[0].bd;
> +       int ret;
> +
> +       if (sdma->direction == DMA_FROM_DEVICE) {
> +               load_address = sdma->pc_from_device;
> +       } else {
> +               load_address = sdma->pc_to_device;
> +       }
> +
> +       if (load_address < 0)
> +               return load_address;
> +
> +       pr_debug("%s: load_address = %d\n", __func__, load_address);
> +       pr_debug("%s: wml = 0x%08x\n", __func__, sdma->watermark_level);
> +       pr_debug("%s: shp_addr = 0x%08x\n", __func__, sdma->shp_addr);
> +       pr_debug("%s: per_addr = 0x%08x\n", __func__, sdma->per_addr);
> +       pr_debug("%s: event_mask1 = 0x%08x\n", __func__, sdma->event_mask1);
> +       pr_debug("%s: event_mask2 = 0x%08x\n", __func__, sdma->event_mask2);

Surely it must be possible to get the struct device * pointer for the
channels host and use dev_dbg() instead?

> +
> +       memset(sdma_context, 0, sizeof(*sdma_context));
> +       sdma_context->channel_state.pc = load_address;
> +
> +       /* Send by context the event mask,base address for peripheral
> +        * and watermark level
> +        */
> +       sdma_context->gReg[0] = sdma->event_mask2;
> +       sdma_context->gReg[1] = sdma->event_mask1;
> +       sdma_context->gReg[2] = sdma->per_addr;
> +       sdma_context->gReg[6] = sdma->shp_addr;
> +       sdma_context->gReg[7] = sdma->watermark_level;
> +
> +       bd0->mode.command = C0_SETDM;
> +       bd0->mode.status = BD_DONE | BD_INTR | BD_WRAP | BD_EXTD;
> +       bd0->mode.count = sizeof(*sdma_context) / 4;
> +       bd0->buffer_addr = sdma_context_phys;
> +       bd0->ext_buffer_addr = 2048 + (sizeof(*sdma_context) / 4) * channel;
> +
> +       ret = sdma_run_channel(0);
> +
> +       return ret;
> +}
> +
> +static void sdma_disable_channel(int channel)
> +{
> +       struct sdma_channel *sdma = &sdma_data[channel];
> +
> +       writel(1 << channel, SDMA_H_STATSTOP);
> +       sdma->busy = 0;
> +}
> +
> +static int sdma_config_channel(int channel)
> +{
> +       struct sdma_channel *sdma = &sdma_data[channel];
> +       int ret;
> +
> +       sdma_disable_channel(channel);
> +
> +       sdma->event_mask1 = 0;
> +       sdma->event_mask2 = 0;
> +       sdma->shp_addr = 0;
> +       sdma->per_addr = 0;
> +
> +       if (sdma->event_id)
> +               sdma_event_enable(channel, sdma->event_id);
> +
> +       switch (sdma->peripheral_type) {
> +       case IMX_DMATYPE_DSP:
> +               sdma_config_ownership(channel, 0, 1, 1);

The parameters here makes yoy believe that the types should
be bool rather than int...

> +               break;
> +       case IMX_DMATYPE_MEMORY:
> +               sdma_config_ownership(channel, 0, 1, 0);
> +               break;
> +       default:
> +               sdma_config_ownership(channel, 1, 1, 0);
> +               break;
> +       }
> +
> +       sdma_get_pc(sdma, sdma->peripheral_type);
> +
> +       if ((sdma->peripheral_type != IMX_DMATYPE_MEMORY) &&
> +                       (sdma->peripheral_type != IMX_DMATYPE_DSP)) {
> +               /* Handle multiple event channels differently */
> +               if (sdma->event_id2) {
> +                       sdma->event_mask2 = 1 << (sdma->event_id2 % 32);
> +                       if (sdma->event_id2 > 31)
> +                               sdma->watermark_level |= 1 << 31;
> +                       sdma->event_mask1 = 1 << (sdma->event_id % 32);
> +                       if (sdma->event_id > 31)
> +                               sdma->watermark_level |= 1 << 30;
> +               } else {
> +                       sdma->event_mask1 = 1 << sdma->event_id;
> +                       sdma->event_mask2 = 1 << (sdma->event_id - 32);
> +               }
> +               /* Watermark Level */
> +               sdma->watermark_level |= sdma->watermark_level;
> +               /* Address */
> +               sdma->shp_addr = sdma->per_address;
> +       } else {
> +               sdma->watermark_level = 0; /* FIXME: M3_BASE_ADDRESS */
> +       }
> +
> +       ret = sdma_load_context(channel);
> +
> +       return ret;
> +}
> +
> +static int sdma_set_channel_priority(unsigned int channel, unsigned int priority)
> +{
> +       if (priority < MXC_SDMA_MIN_PRIORITY
> +           || priority > MXC_SDMA_MAX_PRIORITY) {
> +               return -EINVAL;
> +       }
> +
> +       writel(priority, SDMA_CHNPRI_0 + 4 * channel);
> +
> +       return 0;
> +}
> +
> +static int sdma_request_channel(int channel)
> +{
> +       struct sdma_channel *sdma = &sdma_data[channel];
> +       int ret = -EBUSY;
> +
> +       sdma->bd = dma_alloc_coherent(NULL, PAGE_SIZE, &sdma->bd_phys, GFP_KERNEL);
> +       if (!sdma->bd) {
> +               ret = -ENOMEM;
> +               goto out;
> +       }
> +
> +       memset(sdma->bd, 0, PAGE_SIZE);
> +
> +       channel_control[channel].base_bd_ptr = sdma->bd_phys;
> +       channel_control[channel].current_bd_ptr = sdma->bd_phys;
> +
> +       clk_enable(sdma_clk);

Aha you're enabling it once for every channel and rely on
clk reference counting that's clever!

> +
> +       sdma_set_channel_priority(channel, MXC_SDMA_DEFAULT_PRIORITY);
> +
> +       init_waitqueue_head(&sdma->waitq);
> +
> +       sdma->buf_tail = 0;
> +
> +       return 0;
> +out:
> +
> +       return ret;
> +}
> +
> +static void sdma_enable_channel(int channel)
> +{
> +       writel(1 << channel, SDMA_H_START);
> +}
> +
> +static int __init sdma_init(unsigned long phys_base, int irq, int version,
> +               void *ram_code,
> +               int ram_code_size)
> +{
> +       int i, ret;
> +       int channel;
> +       dma_addr_t ccb_phys;
> +
> +       sdma_version = version;
> +       switch (sdma_version) {
> +       case 1:
> +               sdma_num_events = 32;
> +               break;
> +       case 2:
> +               sdma_num_events = 48;
> +               break;
> +       default:
> +               pr_err("SDMA: Unknown version %d. aborting\n", sdma_version);
> +               return -ENODEV;
> +       }
> +
> +       clk_enable(sdma_clk);
> +
> +       sdma_base = ioremap(phys_base, 4096);

Use SZ_4K instead of 4096.

> +       if (!sdma_base) {
> +               ret = -ENOMEM;
> +               goto err_ioremap;
> +       }
> +
> +       /* Initialize SDMA private data */
> +       memset(sdma_data, 0, sizeof(struct sdma_channel) * MAX_DMA_CHANNELS);
> +
> +       for (channel = 0; channel < MAX_DMA_CHANNELS; channel++)
> +               sdma_data[channel].channel = channel;
> +
> +       ret = request_irq(irq, sdma_int_handler, 0, "sdma", NULL);
> +       if (ret)
> +               goto err_request_irq;
> +
> +       /* Be sure SDMA has not started yet */
> +       writel(0, SDMA_H_C0PTR);
> +
> +       channel_control = dma_alloc_coherent(NULL,
> +                       MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control) +
> +                       sizeof(struct sdma_context_data),
> +                       &ccb_phys, GFP_KERNEL);
> +
> +       if (!channel_control) {
> +               ret = -ENOMEM;
> +               goto err_dma_alloc;
> +       }
> +
> +       sdma_context = (void *)channel_control +
> +               MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control);
> +       sdma_context_phys = ccb_phys +
> +               MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control);
> +
> +       /* Zero-out the CCB structures array just allocated */
> +       memset(channel_control, 0,
> +                       MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control));
> +
> +       /* disable all channels */
> +       for (i = 0; i < sdma_num_events; i++)
> +               writel(0, SDMA_CHNENBL_0 + i * 4);
> +
> +       /* All channels have priority 0 */
> +       for (i = 0; i < MAX_DMA_CHANNELS; i++)
> +               writel(0, SDMA_CHNPRI_0 + i * 4);
> +
> +       ret = sdma_request_channel(0);
> +       if (ret)
> +               goto err_dma_alloc;
> +
> +       sdma_config_ownership(0, 0, 1, 0);
> +
> +       /* Set Command Channel (Channel Zero) */
> +       writel(0x4050, SDMA_CHN0ADDR);
> +
> +       /* Set bits of CONFIG register but with static context switching */
> +       /* FIXME: Check whether to set ACR bit depending on clock ratios */
> +       writel(0, SDMA_H_CONFIG);
> +
> +       writel(ccb_phys, SDMA_H_C0PTR);
> +
> +       /* download the RAM image for SDMA */
> +       sdma_load_script(ram_code,
> +                       ram_code_size,
> +                       sdma_script_addrs->ram_code_start_addr);
> +
> +       /* Set bits of CONFIG register with given context switching mode */
> +       writel(SDMA_H_CONFIG_CSM, SDMA_H_CONFIG);
> +
> +       /* Initializes channel's priorities */
> +       sdma_set_channel_priority(0, 7);
> +
> +       clk_disable(sdma_clk);
> +
> +       return 0;
> +
> +err_dma_alloc:
> +       free_irq(irq, NULL);
> +err_request_irq:
> +       iounmap(sdma_base);
> +err_ioremap:
> +       clk_disable(sdma_clk);
> +       pr_err("%s failed with %d\n", __func__, ret);
> +       return ret;
> +}
> +
> +static dma_cookie_t sdma_assign_cookie(struct sdma_channel *sdma)
> +{
> +       dma_cookie_t cookie = sdma->chan.cookie;
> +
> +       if (++cookie < 0)
> +               cookie = 1;
> +
> +       sdma->chan.cookie = cookie;
> +       sdma->desc.cookie = cookie;
> +
> +       return cookie;
> +}
> +
> +static struct sdma_channel *to_sdma_chan(struct dma_chan *chan)
> +{
> +       return container_of(chan, struct sdma_channel, chan);
> +}
> +
> +static dma_cookie_t sdma_tx_submit(struct dma_async_tx_descriptor *tx)
> +{
> +       struct sdma_channel *sdma = to_sdma_chan(tx->chan);
> +       dma_cookie_t cookie;
> +
> +       spin_lock_irq(&sdma->lock);
> +
> +       cookie = sdma_assign_cookie(sdma);
> +
> +       sdma_enable_channel(tx->chan->chan_id);
> +
> +       spin_unlock_irq(&sdma->lock);
> +
> +       return cookie;
> +}
> +
> +static int sdma_alloc_chan_resources(struct dma_chan *chan)
> +{
> +       struct sdma_channel *sdma = to_sdma_chan(chan);
> +       struct imx_dma_data *data = chan->private;
> +       int prio, ret;
> +
> +       /* No need to execute this for internal channel 0 */
> +       if (!chan->chan_id)
> +               return 0;
> +
> +       if (!data)
> +               return -EINVAL;
> +
> +       switch (data->priority) {
> +       case DMA_PRIO_HIGH:
> +               prio = 3;

Wait, aren't these enumerated?
Add some enum sdma_channel_prio {}..


> +               break;
> +       case DMA_PRIO_MEDIUM:
> +               prio = 2;
> +               break;
> +       case DMA_PRIO_LOW:
> +       default:
> +               prio = 1;
> +               break;
> +       }
> +
> +       sdma->peripheral_type = data->peripheral_type;
> +       sdma->event_id = data->dma_request;
> +       ret = sdma_set_channel_priority(chan->chan_id, prio);
> +       if (ret)
> +               return ret;
> +
> +       if (chan->chan_id) {
> +               ret = sdma_request_channel(chan->chan_id);
> +               if (ret)
> +                       return ret;
> +       }
> +
> +       dma_async_tx_descriptor_init(&sdma->desc, chan);
> +       sdma->desc.tx_submit = sdma_tx_submit;
> +       /* txd.flags will be overwritten in prep funcs */
> +       sdma->desc.flags = DMA_CTRL_ACK;
> +
> +       return 0;
> +}
> +
> +static void sdma_free_chan_resources(struct dma_chan *chan)
> +{
> +       struct sdma_channel *sdma = to_sdma_chan(chan);
> +       int channel = chan->chan_id;
> +
> +       sdma_disable_channel(channel);
> +
> +       if (sdma->event_id)
> +               sdma_event_disable(channel, sdma->event_id);
> +       if (sdma->event_id2)
> +               sdma_event_disable(channel, sdma->event_id2);
> +
> +       sdma->event_id = 0;
> +       sdma->event_id2 = 0;
> +
> +       sdma_set_channel_priority(channel, 0);
> +
> +       dma_free_coherent(NULL, PAGE_SIZE, sdma->bd, sdma->bd_phys);
> +
> +       clk_disable(sdma_clk);
> +}
> +
> +#define NUM_BD (int)(PAGE_SIZE / sizeof(struct sdma_buffer_descriptor))
> +
> +static struct dma_async_tx_descriptor *sdma_prep_slave_sg(
> +               struct dma_chan *chan, struct scatterlist *sgl,
> +               unsigned int sg_len, enum dma_data_direction direction,
> +               unsigned long flags)
> +{
> +       struct sdma_channel *sdma = to_sdma_chan(chan);
> +       int ret, i, count;
> +       int channel = chan->chan_id;
> +       struct scatterlist *sg;
> +
> +       if (sdma->busy)
> +               return NULL;
> +       sdma->busy = 1;
> +
> +       sdma->flags = 0;

What are those flags anyway? I think you will need some
#define:s for them.

> +
> +       pr_debug("SDMA: setting up %d entries for channel %d.\n",
> +                       sg_len, channel);
> +
> +       sdma->direction = direction;
> +       ret = sdma_load_context(channel);
> +       if (ret)
> +               goto err_out;
> +
> +       if (sg_len > NUM_BD) {
> +               pr_err("SDMA channel %d: maximum number of sg exceeded: %d > %d\n",
> +                               channel, sg_len, NUM_BD);
> +               ret = -EINVAL;
> +               goto err_out;
> +       }
> +
> +       for_each_sg(sgl, sg, sg_len, i) {
> +               struct sdma_buffer_descriptor *bd = &sdma->bd[i];
> +               int param;
> +
> +               bd->buffer_addr = sgl->dma_address;
> +
> +               count = sg->length;
> +
> +               if (count > 0xffff) {
> +                       pr_err("SDMA channel %d: maximum bytes for sg entry exceeded: %d > %d\n",
> +                                       channel, count, 0xffff);
> +                       ret = -EINVAL;
> +                       goto err_out;
> +               }
> +
> +               bd->mode.count = count;
> +
> +               if (sdma->word_size > 4) {
> +                       ret =  -EINVAL;
> +                       goto err_out;
> +               }
> +               if (sdma->word_size == 4)
> +                       bd->mode.command = 0;
> +               else
> +                       bd->mode.command = sdma->word_size;
> +
> +               param = BD_DONE | BD_EXTD | BD_CONT;
> +
> +               if (sdma->flags & IMX_DMA_SG_LOOP) {
> +                       param |= BD_INTR;
> +                       if (i + 1 == sg_len)
> +                               param |= BD_WRAP;
> +               }
> +
> +               if (i + 1 == sg_len)
> +                       param |= BD_INTR;
> +
> +               pr_debug("entry %d: count: %d dma: 0x%08x %s%s\n",
> +                               i, count, sg->dma_address,
> +                               param & BD_WRAP ? "wrap" : "",
> +                               param & BD_INTR ? " intr" : "");
> +
> +               bd->mode.status = param;
> +       }
> +
> +       sdma->num_bd = sg_len;
> +       channel_control[channel].current_bd_ptr = sdma->bd_phys;
> +
> +       return &sdma->desc;
> +err_out:
> +       return NULL;
> +}
> +
> +static struct dma_async_tx_descriptor *sdma_prep_dma_cyclic(
> +               struct dma_chan *chan, dma_addr_t dma_addr, size_t buf_len,
> +               size_t period_len, enum dma_data_direction direction)
> +{
> +       int num_periods = buf_len / period_len;
> +       struct sdma_channel *sdma = to_sdma_chan(chan);
> +       int channel = chan->chan_id;
> +       int ret, i = 0, buf = 0;
> +
> +       pr_debug("%s channel: %d\n", __func__, channel);

Must be possible to find struct device * and use dev_dbg()

> +
> +       if (sdma->busy)
> +               return NULL;
> +
> +       sdma->busy = 1;
> +
> +       sdma->flags |= IMX_DMA_SG_LOOP;
> +       sdma->direction = direction;
> +       ret = sdma_load_context(channel);
> +       if (ret)
> +               goto err_out;
> +
> +       if (num_periods > NUM_BD) {
> +               pr_err("SDMA channel %d: maximum number of sg exceeded: %d > %d\n",
> +                               channel, num_periods, NUM_BD);
> +               goto err_out;
> +       }
> +
> +       if (period_len > 0xffff) {
> +               pr_err("SDMA channel %d: maximum period size exceeded: %d > %d\n",
> +                               channel, period_len, 0xffff);
> +               goto err_out;
> +       }
> +
> +       while (buf < buf_len) {
> +               struct sdma_buffer_descriptor *bd = &sdma->bd[i];
> +               int param;
> +
> +               bd->buffer_addr = dma_addr;
> +
> +               bd->mode.count = period_len;
> +
> +               if (sdma->word_size > 4)
> +                       goto err_out;
> +               if (sdma->word_size == 4)
> +                       bd->mode.command = 0;
> +               else
> +                       bd->mode.command = sdma->word_size;
> +
> +               param = BD_DONE | BD_EXTD | BD_CONT | BD_INTR;
> +               if (i + 1 == num_periods)
> +                       param |= BD_WRAP;
> +
> +               pr_debug("entry %d: count: %d dma: 0x%08x %s%s\n",
> +                               i, period_len, dma_addr,
> +                               param & BD_WRAP ? "wrap" : "",
> +                               param & BD_INTR ? " intr" : "");
> +
> +               bd->mode.status = param;
> +
> +               dma_addr += period_len;
> +               buf += period_len;
> +
> +               i++;
> +       }
> +
> +       sdma->num_bd = num_periods;
> +       channel_control[channel].current_bd_ptr = sdma->bd_phys;
> +
> +       return &sdma->desc;
> +err_out:
> +       sdma->busy = 0;
> +       return NULL;
> +}
> +
> +static int sdma_control(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
> +               unsigned long arg)
> +{
> +       struct sdma_channel *sdma = to_sdma_chan(chan);
> +       struct dma_slave_config *dmaengine_cfg = (void *)arg;
> +
> +       switch (cmd) {
> +       case DMA_TERMINATE_ALL:
> +               sdma_disable_channel(chan->chan_id);
> +               return 0;
> +       case DMA_SLAVE_CONFIG:
> +               if (dmaengine_cfg->direction == DMA_FROM_DEVICE) {
> +                       sdma->per_address = dmaengine_cfg->src_addr;
> +                       sdma->watermark_level = dmaengine_cfg->src_maxburst;
> +                       sdma->word_size = dmaengine_cfg->src_addr_width;
> +               } else {
> +                       sdma->per_address = dmaengine_cfg->dst_addr;
> +                       sdma->watermark_level = dmaengine_cfg->dst_maxburst;
> +                       sdma->word_size = dmaengine_cfg->dst_addr_width;
> +               }
> +               return sdma_config_channel(chan->chan_id);
> +       default:
> +               return -ENOSYS;
> +       }
> +
> +       return -EINVAL;
> +}
> +
> +static enum dma_status sdma_tx_status(struct dma_chan *chan,
> +                                           dma_cookie_t cookie,
> +                                           struct dma_tx_state *txstate)
> +{
> +       struct sdma_channel *sdma = to_sdma_chan(chan);
> +       dma_cookie_t last_used;
> +       enum dma_status ret;
> +
> +       last_used = chan->cookie;
> +
> +       ret = dma_async_is_complete(cookie, sdma->last_completed, last_used);
> +       dma_set_tx_state(txstate, sdma->last_completed, last_used, 0);
> +
> +       return ret;
> +}
> +
> +static void sdma_issue_pending(struct dma_chan *chan)
> +{
> +       /*
> +        * Nothing to do. We only have a single descriptor
> +        */
> +}
> +
> +static int __devinit sdma_probe(struct platform_device *pdev)
> +{
> +       int ret;
> +       const struct firmware *fw;
> +       const struct sdma_firmware_header *header;
> +       const struct sdma_script_start_addrs *addr;
> +       int irq;
> +       unsigned short *ram_code;
> +       struct resource *iores;
> +       struct sdma_platform_data *pdata = pdev->dev.platform_data;
> +       int version;
> +       char *cpustr, *fwname;
> +       int i;
> +       dma_cap_mask_t mask;
> +
> +       /* there can be only one */
> +       BUG_ON(sdma_base);
> +
> +       iores = platform_get_resource(pdev, IORESOURCE_MEM, 0);
> +       irq = platform_get_irq(pdev, 0);
> +       if (!iores || irq < 0 || !pdata)
> +               return -EINVAL;
> +
> +       sdma_clk = clk_get(&pdev->dev, NULL);
> +       if (IS_ERR(sdma_clk)) {
> +               ret = PTR_ERR(sdma_clk);
> +               goto err_clk;
> +       }
> +
> +       if (cpu_is_mx31()) {
> +               cpustr = "imx31";
> +               version = mx31_revision() >> 4;
> +       } else if (cpu_is_mx35()) {
> +               cpustr = "imx35";
> +/* FIXME:      version = mx35_revision(); */
> +               version = 2;
> +       } else {
> +               ret = -EINVAL;
> +               goto err_cputype;
> +       }
> +
> +       fwname = kasprintf(GFP_KERNEL, "sdma-%s-to%d.bin", cpustr, version);
> +       if (!fwname) {
> +               ret = -ENOMEM;
> +               goto err_cputype;
> +       }
> +
> +       ret = request_firmware(&fw, fwname, &pdev->dev);
> +       if (ret) {
> +               dev_err(&pdev->dev, "request firmware \"%s\" failed with %d\n",
> +                               fwname, ret);
> +               kfree(fwname);
> +               goto err_cputype;
> +       }
> +       kfree(fwname);
> +
> +       if (fw->size < sizeof(*header))
> +               goto err_firmware;
> +
> +       header = (struct sdma_firmware_header *)fw->data;
> +
> +       if (header->magic != SDMA_FIRMWARE_MAGIC)
> +               goto err_firmware;
> +       if (header->ram_code_start + header->ram_code_size > fw->size)
> +               goto err_firmware;
> +
> +       addr = (void *)header + header->script_addrs_start;
> +       ram_code = (void *)header + header->ram_code_start;
> +       memcpy(&__sdma_script_addrs, addr, sizeof(*addr));
> +
> +       ret = sdma_init(iores->start, irq, pdata->sdma_version,
> +                       ram_code, header->ram_code_size);
> +       if (ret)
> +               goto err_firmware;
> +
> +       INIT_LIST_HEAD(&sdma_dma_device->channels);
> +
> +       /* Initialize channel parameters */
> +       for (i = 0; i < MAX_DMA_CHANNELS; i++) {
> +               struct sdma_channel *sdma = &sdma_data[i];
> +
> +               spin_lock_init(&sdma->lock);
> +
> +               dma_cap_set(DMA_SLAVE, sdma_dma_device->cap_mask);
> +               dma_cap_set(DMA_CYCLIC, sdma_dma_device->cap_mask);
> +
> +               sdma->chan.device = sdma_dma_device;
> +               sdma->chan.chan_id = i;
> +
> +               /* Add the channel to the DMAC list */
> +               list_add_tail(&sdma->chan.device_node, &sdma_dma_device->channels);
> +       }
> +
> +       sdma_dma_device->dev = &pdev->dev;
> +
> +       sdma_dma_device->device_alloc_chan_resources = sdma_alloc_chan_resources;
> +       sdma_dma_device->device_free_chan_resources = sdma_free_chan_resources;
> +       sdma_dma_device->device_tx_status = sdma_tx_status;
> +       sdma_dma_device->device_prep_slave_sg = sdma_prep_slave_sg;
> +       sdma_dma_device->device_prep_dma_cyclic = sdma_prep_dma_cyclic;
> +       sdma_dma_device->device_control = sdma_control;
> +       sdma_dma_device->device_issue_pending = sdma_issue_pending;
> +
> +       ret = dma_async_device_register(sdma_dma_device);
> +       if (ret) {
> +               dev_err(&pdev->dev, "unable to register DMAC\n");

SDMAC even?

> +               goto err_firmware;
> +       }
> +
> +       dev_info(&pdev->dev, "initialized (firmware %d.%d)\n",
> +                       header->version_major,
> +                       header->version_minor);
> +
> +       /* request channel 0. This is an internal control channel
> +        * to the SDMA engine and not available to clients.
> +        */
> +       dma_cap_zero(mask);
> +       dma_cap_set(DMA_SLAVE, mask);
> +       dma_request_channel(mask, NULL, NULL);
> +
> +       release_firmware(fw);
> +
> +       return 0;
> +
> +err_firmware:
> +       release_firmware(fw);
> +err_cputype:
> +       clk_put(sdma_clk);
> +err_clk:
> +       return 0;
> +}
> +
> +static int __devexit sdma_remove(struct platform_device *pdev)
> +{
> +       return -EBUSY;
> +}
> +
> +static struct platform_driver sdma_driver = {
> +       .driver         = {
> +               .name   = "imx-sdma",
> +       },
> +       .probe          = sdma_probe,
> +       .remove         = __devexit_p(sdma_remove),
> +};
> +
> +static int __init sdma_module_init(void)
> +{
> +       return platform_driver_register(&sdma_driver);
> +}
> +subsys_initcall(sdma_module_init);
> +
> +MODULE_AUTHOR("Sascha Hauer, Pengutronix <s.hauer@xxxxxxxxxxxxxx>");
> +MODULE_DESCRIPTION("i.MX SDMA driver");
> +MODULE_LICENSE("GPL");
> --
> 1.7.1

Thanks for using this API
Sascha!

Yours,
Linus Walleij
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/