2019-05-21 01:08:05 +08:00
|
|
|
// SPDX-License-Identifier: GPL-2.0-or-later
|
2008-03-01 22:42:48 +08:00
|
|
|
/*
|
|
|
|
* Freescale MPC85xx, MPC83xx DMA Engine support
|
|
|
|
*
|
2010-11-11 20:16:29 +08:00
|
|
|
* Copyright (C) 2007-2010 Freescale Semiconductor, Inc. All rights reserved.
|
2008-03-01 22:42:48 +08:00
|
|
|
*
|
|
|
|
* Author:
|
|
|
|
* Zhang Wei <wei.zhang@freescale.com>, Jul 2007
|
|
|
|
* Ebony Zhu <ebony.zhu@freescale.com>, May 2007
|
|
|
|
*
|
|
|
|
* Description:
|
|
|
|
* DMA engine driver for Freescale MPC8540 DMA controller, which is
|
|
|
|
* also fit for MPC8560, MPC8555, MPC8548, MPC8641, and etc.
|
2010-08-04 01:44:52 +08:00
|
|
|
* The support for MPC8349 DMA controller is also added.
|
2008-03-01 22:42:48 +08:00
|
|
|
*
|
2009-04-24 07:17:54 +08:00
|
|
|
* This driver instructs the DMA controller to issue the PCI Read Multiple
|
|
|
|
* command for PCI read operations, instead of using the default PCI Read Line
|
|
|
|
* command. Please be aware that this setting may result in read pre-fetching
|
|
|
|
* on some platforms.
|
2008-03-01 22:42:48 +08:00
|
|
|
*/
|
|
|
|
|
|
|
|
#include <linux/init.h>
|
|
|
|
#include <linux/module.h>
|
|
|
|
#include <linux/pci.h>
|
include cleanup: Update gfp.h and slab.h includes to prepare for breaking implicit slab.h inclusion from percpu.h
percpu.h is included by sched.h and module.h and thus ends up being
included when building most .c files. percpu.h includes slab.h which
in turn includes gfp.h making everything defined by the two files
universally available and complicating inclusion dependencies.
percpu.h -> slab.h dependency is about to be removed. Prepare for
this change by updating users of gfp and slab facilities include those
headers directly instead of assuming availability. As this conversion
needs to touch large number of source files, the following script is
used as the basis of conversion.
http://userweb.kernel.org/~tj/misc/slabh-sweep.py
The script does the followings.
* Scan files for gfp and slab usages and update includes such that
only the necessary includes are there. ie. if only gfp is used,
gfp.h, if slab is used, slab.h.
* When the script inserts a new include, it looks at the include
blocks and try to put the new include such that its order conforms
to its surrounding. It's put in the include block which contains
core kernel includes, in the same order that the rest are ordered -
alphabetical, Christmas tree, rev-Xmas-tree or at the end if there
doesn't seem to be any matching order.
* If the script can't find a place to put a new include (mostly
because the file doesn't have fitting include block), it prints out
an error message indicating which .h file needs to be added to the
file.
The conversion was done in the following steps.
1. The initial automatic conversion of all .c files updated slightly
over 4000 files, deleting around 700 includes and adding ~480 gfp.h
and ~3000 slab.h inclusions. The script emitted errors for ~400
files.
2. Each error was manually checked. Some didn't need the inclusion,
some needed manual addition while adding it to implementation .h or
embedding .c file was more appropriate for others. This step added
inclusions to around 150 files.
3. The script was run again and the output was compared to the edits
from #2 to make sure no file was left behind.
4. Several build tests were done and a couple of problems were fixed.
e.g. lib/decompress_*.c used malloc/free() wrappers around slab
APIs requiring slab.h to be added manually.
5. The script was run on all .h files but without automatically
editing them as sprinkling gfp.h and slab.h inclusions around .h
files could easily lead to inclusion dependency hell. Most gfp.h
inclusion directives were ignored as stuff from gfp.h was usually
wildly available and often used in preprocessor macros. Each
slab.h inclusion directive was examined and added manually as
necessary.
6. percpu.h was updated not to include slab.h.
7. Build test were done on the following configurations and failures
were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my
distributed build env didn't work with gcov compiles) and a few
more options had to be turned off depending on archs to make things
build (like ipr on powerpc/64 which failed due to missing writeq).
* x86 and x86_64 UP and SMP allmodconfig and a custom test config.
* powerpc and powerpc64 SMP allmodconfig
* sparc and sparc64 SMP allmodconfig
* ia64 SMP allmodconfig
* s390 SMP allmodconfig
* alpha SMP allmodconfig
* um on x86_64 SMP allmodconfig
8. percpu.h modifications were reverted so that it could be applied as
a separate patch and serve as bisection point.
Given the fact that I had only a couple of failures from tests on step
6, I'm fairly confident about the coverage of this conversion patch.
If there is a breakage, it's likely to be something in one of the arch
headers which should be easily discoverable easily on most builds of
the specific arch.
Signed-off-by: Tejun Heo <tj@kernel.org>
Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
2010-03-24 16:04:11 +08:00
|
|
|
#include <linux/slab.h>
|
2008-03-01 22:42:48 +08:00
|
|
|
#include <linux/interrupt.h>
|
|
|
|
#include <linux/dmaengine.h>
|
|
|
|
#include <linux/delay.h>
|
|
|
|
#include <linux/dma-mapping.h>
|
|
|
|
#include <linux/dmapool.h>
|
2013-09-18 03:28:33 +08:00
|
|
|
#include <linux/of_address.h>
|
|
|
|
#include <linux/of_irq.h>
|
2008-03-01 22:42:48 +08:00
|
|
|
#include <linux/of_platform.h>
|
2014-10-11 23:46:44 +08:00
|
|
|
#include <linux/fsldma.h>
|
2012-03-07 06:34:26 +08:00
|
|
|
#include "dmaengine.h"
|
2008-03-01 22:42:48 +08:00
|
|
|
#include "fsldma.h"
|
|
|
|
|
2011-03-03 15:54:55 +08:00
|
|
|
#define chan_dbg(chan, fmt, arg...) \
|
|
|
|
dev_dbg(chan->dev, "%s: " fmt, chan->name, ##arg)
|
|
|
|
#define chan_err(chan, fmt, arg...) \
|
|
|
|
dev_err(chan->dev, "%s: " fmt, chan->name, ##arg)
|
2010-09-30 19:46:45 +08:00
|
|
|
|
2011-03-03 15:54:55 +08:00
|
|
|
static const char msg_ld_oom[] = "No free memory for link descriptor";
|
2008-03-01 22:42:48 +08:00
|
|
|
|
2011-03-03 15:54:54 +08:00
|
|
|
/*
|
|
|
|
* Register Helpers
|
|
|
|
*/
|
2008-03-01 22:42:48 +08:00
|
|
|
|
2010-01-06 21:34:05 +08:00
|
|
|
static void set_sr(struct fsldma_chan *chan, u32 val)
|
2008-03-01 22:42:48 +08:00
|
|
|
{
|
2018-10-30 10:35:58 +08:00
|
|
|
FSL_DMA_OUT(chan, &chan->regs->sr, val, 32);
|
2008-03-01 22:42:48 +08:00
|
|
|
}
|
|
|
|
|
2010-01-06 21:34:05 +08:00
|
|
|
static u32 get_sr(struct fsldma_chan *chan)
|
2008-03-01 22:42:48 +08:00
|
|
|
{
|
2018-10-30 10:35:58 +08:00
|
|
|
return FSL_DMA_IN(chan, &chan->regs->sr, 32);
|
2008-03-01 22:42:48 +08:00
|
|
|
}
|
|
|
|
|
2014-04-18 16:17:45 +08:00
|
|
|
static void set_mr(struct fsldma_chan *chan, u32 val)
|
|
|
|
{
|
2018-10-30 10:35:58 +08:00
|
|
|
FSL_DMA_OUT(chan, &chan->regs->mr, val, 32);
|
2014-04-18 16:17:45 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
static u32 get_mr(struct fsldma_chan *chan)
|
|
|
|
{
|
2018-10-30 10:35:58 +08:00
|
|
|
return FSL_DMA_IN(chan, &chan->regs->mr, 32);
|
2014-04-18 16:17:45 +08:00
|
|
|
}
|
|
|
|
|
2011-03-03 15:54:54 +08:00
|
|
|
static void set_cdar(struct fsldma_chan *chan, dma_addr_t addr)
|
|
|
|
{
|
2018-10-30 10:35:58 +08:00
|
|
|
FSL_DMA_OUT(chan, &chan->regs->cdar, addr | FSL_DMA_SNEN, 64);
|
2011-03-03 15:54:54 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
static dma_addr_t get_cdar(struct fsldma_chan *chan)
|
|
|
|
{
|
2018-10-30 10:35:58 +08:00
|
|
|
return FSL_DMA_IN(chan, &chan->regs->cdar, 64) & ~FSL_DMA_SNEN;
|
2011-03-03 15:54:54 +08:00
|
|
|
}
|
|
|
|
|
2014-04-18 16:17:45 +08:00
|
|
|
static void set_bcr(struct fsldma_chan *chan, u32 val)
|
|
|
|
{
|
2018-10-30 10:35:58 +08:00
|
|
|
FSL_DMA_OUT(chan, &chan->regs->bcr, val, 32);
|
2014-04-18 16:17:45 +08:00
|
|
|
}
|
|
|
|
|
2011-03-03 15:54:54 +08:00
|
|
|
static u32 get_bcr(struct fsldma_chan *chan)
|
|
|
|
{
|
2018-10-30 10:35:58 +08:00
|
|
|
return FSL_DMA_IN(chan, &chan->regs->bcr, 32);
|
2011-03-03 15:54:54 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Descriptor Helpers
|
|
|
|
*/
|
|
|
|
|
2010-01-06 21:34:05 +08:00
|
|
|
static void set_desc_cnt(struct fsldma_chan *chan,
|
2008-03-01 22:42:48 +08:00
|
|
|
struct fsl_dma_ld_hw *hw, u32 count)
|
|
|
|
{
|
2010-01-06 21:34:05 +08:00
|
|
|
hw->count = CPU_TO_DMA(chan, count, 32);
|
2008-03-01 22:42:48 +08:00
|
|
|
}
|
|
|
|
|
2010-01-06 21:34:05 +08:00
|
|
|
static void set_desc_src(struct fsldma_chan *chan,
|
2011-03-03 15:54:57 +08:00
|
|
|
struct fsl_dma_ld_hw *hw, dma_addr_t src)
|
2008-03-01 22:42:48 +08:00
|
|
|
{
|
|
|
|
u64 snoop_bits;
|
|
|
|
|
2010-01-06 21:34:05 +08:00
|
|
|
snoop_bits = ((chan->feature & FSL_DMA_IP_MASK) == FSL_DMA_IP_85XX)
|
2008-03-01 22:42:48 +08:00
|
|
|
? ((u64)FSL_DMA_SATR_SREADTYPE_SNOOP_READ << 32) : 0;
|
2010-01-06 21:34:05 +08:00
|
|
|
hw->src_addr = CPU_TO_DMA(chan, snoop_bits | src, 64);
|
2008-03-01 22:42:48 +08:00
|
|
|
}
|
|
|
|
|
2010-01-06 21:34:05 +08:00
|
|
|
static void set_desc_dst(struct fsldma_chan *chan,
|
2011-03-03 15:54:57 +08:00
|
|
|
struct fsl_dma_ld_hw *hw, dma_addr_t dst)
|
2008-03-01 22:42:48 +08:00
|
|
|
{
|
|
|
|
u64 snoop_bits;
|
|
|
|
|
2010-01-06 21:34:05 +08:00
|
|
|
snoop_bits = ((chan->feature & FSL_DMA_IP_MASK) == FSL_DMA_IP_85XX)
|
2008-03-01 22:42:48 +08:00
|
|
|
? ((u64)FSL_DMA_DATR_DWRITETYPE_SNOOP_WRITE << 32) : 0;
|
2010-01-06 21:34:05 +08:00
|
|
|
hw->dst_addr = CPU_TO_DMA(chan, snoop_bits | dst, 64);
|
2008-03-01 22:42:48 +08:00
|
|
|
}
|
|
|
|
|
2010-01-06 21:34:05 +08:00
|
|
|
static void set_desc_next(struct fsldma_chan *chan,
|
2011-03-03 15:54:57 +08:00
|
|
|
struct fsl_dma_ld_hw *hw, dma_addr_t next)
|
2008-03-01 22:42:48 +08:00
|
|
|
{
|
|
|
|
u64 snoop_bits;
|
|
|
|
|
2010-01-06 21:34:05 +08:00
|
|
|
snoop_bits = ((chan->feature & FSL_DMA_IP_MASK) == FSL_DMA_IP_83XX)
|
2008-03-01 22:42:48 +08:00
|
|
|
? FSL_DMA_SNEN : 0;
|
2010-01-06 21:34:05 +08:00
|
|
|
hw->next_ln_addr = CPU_TO_DMA(chan, snoop_bits | next, 64);
|
2008-03-01 22:42:48 +08:00
|
|
|
}
|
|
|
|
|
2011-03-03 15:54:57 +08:00
|
|
|
static void set_ld_eol(struct fsldma_chan *chan, struct fsl_desc_sw *desc)
|
2008-03-01 22:42:48 +08:00
|
|
|
{
|
2011-03-03 15:54:54 +08:00
|
|
|
u64 snoop_bits;
|
2008-03-01 22:42:48 +08:00
|
|
|
|
2011-03-03 15:54:54 +08:00
|
|
|
snoop_bits = ((chan->feature & FSL_DMA_IP_MASK) == FSL_DMA_IP_83XX)
|
|
|
|
? FSL_DMA_SNEN : 0;
|
2008-03-01 22:42:48 +08:00
|
|
|
|
2011-03-03 15:54:54 +08:00
|
|
|
desc->hw.next_ln_addr = CPU_TO_DMA(chan,
|
|
|
|
DMA_TO_CPU(chan, desc->hw.next_ln_addr, 64) | FSL_DMA_EOL
|
|
|
|
| snoop_bits, 64);
|
2008-03-01 22:42:48 +08:00
|
|
|
}
|
|
|
|
|
2011-03-03 15:54:54 +08:00
|
|
|
/*
|
|
|
|
* DMA Engine Hardware Control Helpers
|
|
|
|
*/
|
|
|
|
|
|
|
|
static void dma_init(struct fsldma_chan *chan)
|
2008-03-19 09:45:00 +08:00
|
|
|
{
|
2011-03-03 15:54:54 +08:00
|
|
|
/* Reset the channel */
|
2014-04-18 16:17:45 +08:00
|
|
|
set_mr(chan, 0);
|
2011-03-03 15:54:54 +08:00
|
|
|
|
|
|
|
switch (chan->feature & FSL_DMA_IP_MASK) {
|
|
|
|
case FSL_DMA_IP_85XX:
|
|
|
|
/* Set the channel to below modes:
|
|
|
|
* EIE - Error interrupt enable
|
|
|
|
* EOLNIE - End of links interrupt enable
|
|
|
|
* BWC - Bandwidth sharing among channels
|
|
|
|
*/
|
2014-04-18 16:17:45 +08:00
|
|
|
set_mr(chan, FSL_DMA_MR_BWC | FSL_DMA_MR_EIE
|
|
|
|
| FSL_DMA_MR_EOLNIE);
|
2011-03-03 15:54:54 +08:00
|
|
|
break;
|
|
|
|
case FSL_DMA_IP_83XX:
|
|
|
|
/* Set the channel to below modes:
|
|
|
|
* EOTIE - End-of-transfer interrupt enable
|
|
|
|
* PRC_RM - PCI read multiple
|
|
|
|
*/
|
2014-04-18 16:17:45 +08:00
|
|
|
set_mr(chan, FSL_DMA_MR_EOTIE | FSL_DMA_MR_PRC_RM);
|
2011-03-03 15:54:54 +08:00
|
|
|
break;
|
|
|
|
}
|
2008-03-19 09:45:00 +08:00
|
|
|
}
|
|
|
|
|
2010-01-06 21:34:05 +08:00
|
|
|
static int dma_is_idle(struct fsldma_chan *chan)
|
2008-03-01 22:42:48 +08:00
|
|
|
{
|
2010-01-06 21:34:05 +08:00
|
|
|
u32 sr = get_sr(chan);
|
2008-03-01 22:42:48 +08:00
|
|
|
return (!(sr & FSL_DMA_SR_CB)) || (sr & FSL_DMA_SR_CH);
|
|
|
|
}
|
|
|
|
|
2011-03-03 15:54:58 +08:00
|
|
|
/*
|
|
|
|
* Start the DMA controller
|
|
|
|
*
|
|
|
|
* Preconditions:
|
|
|
|
* - the CDAR register must point to the start descriptor
|
|
|
|
* - the MRn[CS] bit must be cleared
|
|
|
|
*/
|
2010-01-06 21:34:05 +08:00
|
|
|
static void dma_start(struct fsldma_chan *chan)
|
2008-03-01 22:42:48 +08:00
|
|
|
{
|
2010-01-06 21:33:59 +08:00
|
|
|
u32 mode;
|
|
|
|
|
2014-04-18 16:17:45 +08:00
|
|
|
mode = get_mr(chan);
|
2010-01-06 21:33:59 +08:00
|
|
|
|
2011-03-03 15:54:58 +08:00
|
|
|
if (chan->feature & FSL_DMA_CHAN_PAUSE_EXT) {
|
2014-04-18 16:17:45 +08:00
|
|
|
set_bcr(chan, 0);
|
2011-03-03 15:54:58 +08:00
|
|
|
mode |= FSL_DMA_MR_EMP_EN;
|
|
|
|
} else {
|
|
|
|
mode &= ~FSL_DMA_MR_EMP_EN;
|
2009-05-28 17:26:40 +08:00
|
|
|
}
|
2008-03-01 22:42:48 +08:00
|
|
|
|
2011-03-03 15:54:58 +08:00
|
|
|
if (chan->feature & FSL_DMA_CHAN_START_EXT) {
|
2010-01-06 21:33:59 +08:00
|
|
|
mode |= FSL_DMA_MR_EMS_EN;
|
2011-03-03 15:54:58 +08:00
|
|
|
} else {
|
|
|
|
mode &= ~FSL_DMA_MR_EMS_EN;
|
2010-01-06 21:33:59 +08:00
|
|
|
mode |= FSL_DMA_MR_CS;
|
2011-03-03 15:54:58 +08:00
|
|
|
}
|
2008-03-01 22:42:48 +08:00
|
|
|
|
2014-04-18 16:17:45 +08:00
|
|
|
set_mr(chan, mode);
|
2008-03-01 22:42:48 +08:00
|
|
|
}
|
|
|
|
|
2010-01-06 21:34:05 +08:00
|
|
|
static void dma_halt(struct fsldma_chan *chan)
|
2008-03-01 22:42:48 +08:00
|
|
|
{
|
2010-01-06 21:33:59 +08:00
|
|
|
u32 mode;
|
2009-03-03 06:33:46 +08:00
|
|
|
int i;
|
|
|
|
|
2011-03-03 15:55:01 +08:00
|
|
|
/* read the mode register */
|
2014-04-18 16:17:45 +08:00
|
|
|
mode = get_mr(chan);
|
2010-01-06 21:33:59 +08:00
|
|
|
|
2011-03-03 15:55:01 +08:00
|
|
|
/*
|
|
|
|
* The 85xx controller supports channel abort, which will stop
|
|
|
|
* the current transfer. On 83xx, this bit is the transfer error
|
|
|
|
* mask bit, which should not be changed.
|
|
|
|
*/
|
|
|
|
if ((chan->feature & FSL_DMA_IP_MASK) == FSL_DMA_IP_85XX) {
|
|
|
|
mode |= FSL_DMA_MR_CA;
|
2014-04-18 16:17:45 +08:00
|
|
|
set_mr(chan, mode);
|
2011-03-03 15:55:01 +08:00
|
|
|
|
|
|
|
mode &= ~FSL_DMA_MR_CA;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* stop the DMA controller */
|
|
|
|
mode &= ~(FSL_DMA_MR_CS | FSL_DMA_MR_EMS_EN);
|
2014-04-18 16:17:45 +08:00
|
|
|
set_mr(chan, mode);
|
2008-03-01 22:42:48 +08:00
|
|
|
|
2011-03-03 15:55:01 +08:00
|
|
|
/* wait for the DMA controller to become idle */
|
2009-03-03 06:33:46 +08:00
|
|
|
for (i = 0; i < 100; i++) {
|
2010-01-06 21:34:05 +08:00
|
|
|
if (dma_is_idle(chan))
|
2010-01-06 21:34:06 +08:00
|
|
|
return;
|
|
|
|
|
2008-03-01 22:42:48 +08:00
|
|
|
udelay(10);
|
2009-03-03 06:33:46 +08:00
|
|
|
}
|
2010-01-06 21:33:59 +08:00
|
|
|
|
2010-01-06 21:34:06 +08:00
|
|
|
if (!dma_is_idle(chan))
|
2011-03-03 15:54:55 +08:00
|
|
|
chan_err(chan, "DMA halt timeout!\n");
|
2008-03-01 22:42:48 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* fsl_chan_set_src_loop_size - Set source address hold transfer size
|
2010-01-06 21:34:05 +08:00
|
|
|
* @chan : Freescale DMA channel
|
2008-03-01 22:42:48 +08:00
|
|
|
* @size : Address loop size, 0 for disable loop
|
|
|
|
*
|
|
|
|
* The set source address hold transfer size. The source
|
|
|
|
* address hold or loop transfer size is when the DMA transfer
|
|
|
|
* data from source address (SA), if the loop size is 4, the DMA will
|
|
|
|
* read data from SA, SA + 1, SA + 2, SA + 3, then loop back to SA,
|
|
|
|
* SA + 1 ... and so on.
|
|
|
|
*/
|
2010-01-06 21:34:05 +08:00
|
|
|
static void fsl_chan_set_src_loop_size(struct fsldma_chan *chan, int size)
|
2008-03-01 22:42:48 +08:00
|
|
|
{
|
2010-01-06 21:33:59 +08:00
|
|
|
u32 mode;
|
|
|
|
|
2014-04-18 16:17:45 +08:00
|
|
|
mode = get_mr(chan);
|
2010-01-06 21:33:59 +08:00
|
|
|
|
2008-03-01 22:42:48 +08:00
|
|
|
switch (size) {
|
|
|
|
case 0:
|
2010-01-06 21:33:59 +08:00
|
|
|
mode &= ~FSL_DMA_MR_SAHE;
|
2008-03-01 22:42:48 +08:00
|
|
|
break;
|
|
|
|
case 1:
|
|
|
|
case 2:
|
|
|
|
case 4:
|
|
|
|
case 8:
|
2017-06-19 22:40:04 +08:00
|
|
|
mode &= ~FSL_DMA_MR_SAHTS_MASK;
|
2010-01-06 21:33:59 +08:00
|
|
|
mode |= FSL_DMA_MR_SAHE | (__ilog2(size) << 14);
|
2008-03-01 22:42:48 +08:00
|
|
|
break;
|
|
|
|
}
|
2010-01-06 21:33:59 +08:00
|
|
|
|
2014-04-18 16:17:45 +08:00
|
|
|
set_mr(chan, mode);
|
2008-03-01 22:42:48 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
2010-01-06 21:34:02 +08:00
|
|
|
* fsl_chan_set_dst_loop_size - Set destination address hold transfer size
|
2010-01-06 21:34:05 +08:00
|
|
|
* @chan : Freescale DMA channel
|
2008-03-01 22:42:48 +08:00
|
|
|
* @size : Address loop size, 0 for disable loop
|
|
|
|
*
|
|
|
|
* The set destination address hold transfer size. The destination
|
|
|
|
* address hold or loop transfer size is when the DMA transfer
|
|
|
|
* data to destination address (TA), if the loop size is 4, the DMA will
|
|
|
|
* write data to TA, TA + 1, TA + 2, TA + 3, then loop back to TA,
|
|
|
|
* TA + 1 ... and so on.
|
|
|
|
*/
|
2010-01-06 21:34:05 +08:00
|
|
|
static void fsl_chan_set_dst_loop_size(struct fsldma_chan *chan, int size)
|
2008-03-01 22:42:48 +08:00
|
|
|
{
|
2010-01-06 21:33:59 +08:00
|
|
|
u32 mode;
|
|
|
|
|
2014-04-18 16:17:45 +08:00
|
|
|
mode = get_mr(chan);
|
2010-01-06 21:33:59 +08:00
|
|
|
|
2008-03-01 22:42:48 +08:00
|
|
|
switch (size) {
|
|
|
|
case 0:
|
2010-01-06 21:33:59 +08:00
|
|
|
mode &= ~FSL_DMA_MR_DAHE;
|
2008-03-01 22:42:48 +08:00
|
|
|
break;
|
|
|
|
case 1:
|
|
|
|
case 2:
|
|
|
|
case 4:
|
|
|
|
case 8:
|
2017-06-19 22:40:04 +08:00
|
|
|
mode &= ~FSL_DMA_MR_DAHTS_MASK;
|
2010-01-06 21:33:59 +08:00
|
|
|
mode |= FSL_DMA_MR_DAHE | (__ilog2(size) << 16);
|
2008-03-01 22:42:48 +08:00
|
|
|
break;
|
|
|
|
}
|
2010-01-06 21:33:59 +08:00
|
|
|
|
2014-04-18 16:17:45 +08:00
|
|
|
set_mr(chan, mode);
|
2008-03-01 22:42:48 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
2009-09-09 08:53:04 +08:00
|
|
|
* fsl_chan_set_request_count - Set DMA Request Count for external control
|
2010-01-06 21:34:05 +08:00
|
|
|
* @chan : Freescale DMA channel
|
2009-09-09 08:53:04 +08:00
|
|
|
* @size : Number of bytes to transfer in a single request
|
|
|
|
*
|
|
|
|
* The Freescale DMA channel can be controlled by the external signal DREQ#.
|
|
|
|
* The DMA request count is how many bytes are allowed to transfer before
|
|
|
|
* pausing the channel, after which a new assertion of DREQ# resumes channel
|
|
|
|
* operation.
|
2008-03-01 22:42:48 +08:00
|
|
|
*
|
2009-09-09 08:53:04 +08:00
|
|
|
* A size of 0 disables external pause control. The maximum size is 1024.
|
2008-03-01 22:42:48 +08:00
|
|
|
*/
|
2010-01-06 21:34:05 +08:00
|
|
|
static void fsl_chan_set_request_count(struct fsldma_chan *chan, int size)
|
2008-03-01 22:42:48 +08:00
|
|
|
{
|
2010-01-06 21:33:59 +08:00
|
|
|
u32 mode;
|
|
|
|
|
2009-09-09 08:53:04 +08:00
|
|
|
BUG_ON(size > 1024);
|
2010-01-06 21:33:59 +08:00
|
|
|
|
2014-04-18 16:17:45 +08:00
|
|
|
mode = get_mr(chan);
|
2017-06-19 22:40:04 +08:00
|
|
|
mode &= ~FSL_DMA_MR_BWC_MASK;
|
|
|
|
mode |= (__ilog2(size) << 24) & FSL_DMA_MR_BWC_MASK;
|
2010-01-06 21:33:59 +08:00
|
|
|
|
2014-04-18 16:17:45 +08:00
|
|
|
set_mr(chan, mode);
|
2009-09-09 08:53:04 +08:00
|
|
|
}
|
2008-03-01 22:42:48 +08:00
|
|
|
|
2009-09-09 08:53:04 +08:00
|
|
|
/**
|
|
|
|
* fsl_chan_toggle_ext_pause - Toggle channel external pause status
|
2010-01-06 21:34:05 +08:00
|
|
|
* @chan : Freescale DMA channel
|
2009-09-09 08:53:04 +08:00
|
|
|
* @enable : 0 is disabled, 1 is enabled.
|
|
|
|
*
|
|
|
|
* The Freescale DMA channel can be controlled by the external signal DREQ#.
|
|
|
|
* The DMA Request Count feature should be used in addition to this feature
|
|
|
|
* to set the number of bytes to transfer before pausing the channel.
|
|
|
|
*/
|
2010-01-06 21:34:05 +08:00
|
|
|
static void fsl_chan_toggle_ext_pause(struct fsldma_chan *chan, int enable)
|
2009-09-09 08:53:04 +08:00
|
|
|
{
|
|
|
|
if (enable)
|
2010-01-06 21:34:05 +08:00
|
|
|
chan->feature |= FSL_DMA_CHAN_PAUSE_EXT;
|
2009-09-09 08:53:04 +08:00
|
|
|
else
|
2010-01-06 21:34:05 +08:00
|
|
|
chan->feature &= ~FSL_DMA_CHAN_PAUSE_EXT;
|
2008-03-01 22:42:48 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* fsl_chan_toggle_ext_start - Toggle channel external start status
|
2010-01-06 21:34:05 +08:00
|
|
|
* @chan : Freescale DMA channel
|
2008-03-01 22:42:48 +08:00
|
|
|
* @enable : 0 is disabled, 1 is enabled.
|
|
|
|
*
|
|
|
|
* If enable the external start, the channel can be started by an
|
|
|
|
* external DMA start pin. So the dma_start() does not start the
|
|
|
|
* transfer immediately. The DMA channel will wait for the
|
|
|
|
* control pin asserted.
|
|
|
|
*/
|
2010-01-06 21:34:05 +08:00
|
|
|
static void fsl_chan_toggle_ext_start(struct fsldma_chan *chan, int enable)
|
2008-03-01 22:42:48 +08:00
|
|
|
{
|
|
|
|
if (enable)
|
2010-01-06 21:34:05 +08:00
|
|
|
chan->feature |= FSL_DMA_CHAN_START_EXT;
|
2008-03-01 22:42:48 +08:00
|
|
|
else
|
2010-01-06 21:34:05 +08:00
|
|
|
chan->feature &= ~FSL_DMA_CHAN_START_EXT;
|
2008-03-01 22:42:48 +08:00
|
|
|
}
|
|
|
|
|
2014-10-11 23:46:44 +08:00
|
|
|
int fsl_dma_external_start(struct dma_chan *dchan, int enable)
|
|
|
|
{
|
|
|
|
struct fsldma_chan *chan;
|
|
|
|
|
|
|
|
if (!dchan)
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
chan = to_fsl_chan(dchan);
|
|
|
|
|
|
|
|
fsl_chan_toggle_ext_start(chan, enable);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(fsl_dma_external_start);
|
|
|
|
|
2011-03-03 15:54:57 +08:00
|
|
|
static void append_ld_queue(struct fsldma_chan *chan, struct fsl_desc_sw *desc)
|
2010-01-06 21:34:06 +08:00
|
|
|
{
|
|
|
|
struct fsl_desc_sw *tail = to_fsl_desc(chan->ld_pending.prev);
|
|
|
|
|
|
|
|
if (list_empty(&chan->ld_pending))
|
|
|
|
goto out_splice;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Add the hardware descriptor to the chain of hardware descriptors
|
|
|
|
* that already exists in memory.
|
|
|
|
*
|
|
|
|
* This will un-set the EOL bit of the existing transaction, and the
|
|
|
|
* last link in this transaction will become the EOL descriptor.
|
|
|
|
*/
|
|
|
|
set_desc_next(chan, &tail->hw, desc->async_tx.phys);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Add the software descriptor and all children to the list
|
|
|
|
* of pending transactions
|
|
|
|
*/
|
|
|
|
out_splice:
|
|
|
|
list_splice_tail_init(&desc->tx_list, &chan->ld_pending);
|
|
|
|
}
|
|
|
|
|
2008-03-01 22:42:48 +08:00
|
|
|
static dma_cookie_t fsl_dma_tx_submit(struct dma_async_tx_descriptor *tx)
|
|
|
|
{
|
2010-01-06 21:34:05 +08:00
|
|
|
struct fsldma_chan *chan = to_fsl_chan(tx->chan);
|
2009-09-09 08:53:02 +08:00
|
|
|
struct fsl_desc_sw *desc = tx_to_fsl_desc(tx);
|
|
|
|
struct fsl_desc_sw *child;
|
2013-12-10 03:16:00 +08:00
|
|
|
dma_cookie_t cookie = -EINVAL;
|
2008-03-01 22:42:48 +08:00
|
|
|
|
2014-05-21 16:03:01 +08:00
|
|
|
spin_lock_bh(&chan->desc_lock);
|
2008-03-01 22:42:48 +08:00
|
|
|
|
2014-05-21 16:03:02 +08:00
|
|
|
#ifdef CONFIG_PM
|
|
|
|
if (unlikely(chan->pm_state != RUNNING)) {
|
|
|
|
chan_dbg(chan, "cannot submit due to suspend\n");
|
|
|
|
spin_unlock_bh(&chan->desc_lock);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
2010-01-06 21:34:06 +08:00
|
|
|
/*
|
|
|
|
* assign cookies to all of the software descriptors
|
|
|
|
* that make up this transaction
|
|
|
|
*/
|
2009-09-09 08:53:02 +08:00
|
|
|
list_for_each_entry(child, &desc->tx_list, node) {
|
2012-03-07 06:34:46 +08:00
|
|
|
cookie = dma_cookie_assign(&child->async_tx);
|
fsldma: fix infinite loop on multi-descriptor DMA chain completion
When creating a DMA transaction with multiple descriptors, the async_tx
cookie is set to 0 for each descriptor in the chain, excluding the last
descriptor, whose cookie is set to -EBUSY.
When fsl_dma_tx_submit() is run, it only assigns a cookie to the first
descriptor. All of the remaining descriptors keep their original value,
including the last descriptor, which is set to -EBUSY.
After the DMA completes, the driver will update the last completed cookie
to be -EBUSY, which is an error code instead of a valid cookie. This causes
dma_async_is_complete() to always return DMA_IN_PROGRESS.
This causes the fsldma driver to never cleanup the queue of link
descriptors, and the driver will re-run the DMA transaction on the hardware
each time it receives the End-of-Chain interrupt. This causes an infinite
loop.
With this patch, fsl_dma_tx_submit() is changed to assign a cookie to every
descriptor in the chain. The rest of the code then works without problems.
Signed-off-by: Ira W. Snyder <iws@ovro.caltech.edu>
Signed-off-by: Li Yang <leoli@freescale.com>
2009-05-16 05:27:16 +08:00
|
|
|
}
|
|
|
|
|
2010-01-06 21:34:06 +08:00
|
|
|
/* put this transaction onto the tail of the pending queue */
|
2010-01-06 21:34:05 +08:00
|
|
|
append_ld_queue(chan, desc);
|
2008-03-01 22:42:48 +08:00
|
|
|
|
2014-05-21 16:03:01 +08:00
|
|
|
spin_unlock_bh(&chan->desc_lock);
|
2008-03-01 22:42:48 +08:00
|
|
|
|
|
|
|
return cookie;
|
|
|
|
}
|
|
|
|
|
2014-04-18 16:17:47 +08:00
|
|
|
/**
|
|
|
|
* fsl_dma_free_descriptor - Free descriptor from channel's DMA pool.
|
|
|
|
* @chan : Freescale DMA channel
|
|
|
|
* @desc: descriptor to be freed
|
|
|
|
*/
|
|
|
|
static void fsl_dma_free_descriptor(struct fsldma_chan *chan,
|
|
|
|
struct fsl_desc_sw *desc)
|
|
|
|
{
|
|
|
|
list_del(&desc->node);
|
|
|
|
chan_dbg(chan, "LD %p free\n", desc);
|
|
|
|
dma_pool_free(chan->desc_pool, desc, desc->async_tx.phys);
|
|
|
|
}
|
|
|
|
|
2008-03-01 22:42:48 +08:00
|
|
|
/**
|
|
|
|
* fsl_dma_alloc_descriptor - Allocate descriptor from channel's DMA pool.
|
2010-01-06 21:34:05 +08:00
|
|
|
* @chan : Freescale DMA channel
|
2008-03-01 22:42:48 +08:00
|
|
|
*
|
|
|
|
* Return - The descriptor allocated. NULL for failed.
|
|
|
|
*/
|
2011-03-03 15:54:57 +08:00
|
|
|
static struct fsl_desc_sw *fsl_dma_alloc_descriptor(struct fsldma_chan *chan)
|
2008-03-01 22:42:48 +08:00
|
|
|
{
|
2010-01-06 21:34:06 +08:00
|
|
|
struct fsl_desc_sw *desc;
|
2008-03-01 22:42:48 +08:00
|
|
|
dma_addr_t pdesc;
|
2010-01-06 21:34:06 +08:00
|
|
|
|
2016-04-30 04:09:12 +08:00
|
|
|
desc = dma_pool_zalloc(chan->desc_pool, GFP_ATOMIC, &pdesc);
|
2010-01-06 21:34:06 +08:00
|
|
|
if (!desc) {
|
2011-03-03 15:54:55 +08:00
|
|
|
chan_dbg(chan, "out of memory for link descriptor\n");
|
2010-01-06 21:34:06 +08:00
|
|
|
return NULL;
|
2008-03-01 22:42:48 +08:00
|
|
|
}
|
|
|
|
|
2010-01-06 21:34:06 +08:00
|
|
|
INIT_LIST_HEAD(&desc->tx_list);
|
|
|
|
dma_async_tx_descriptor_init(&desc->async_tx, &chan->common);
|
|
|
|
desc->async_tx.tx_submit = fsl_dma_tx_submit;
|
|
|
|
desc->async_tx.phys = pdesc;
|
|
|
|
|
2011-03-03 15:54:56 +08:00
|
|
|
chan_dbg(chan, "LD %p allocated\n", desc);
|
|
|
|
|
2010-01-06 21:34:06 +08:00
|
|
|
return desc;
|
2008-03-01 22:42:48 +08:00
|
|
|
}
|
|
|
|
|
dmaengine: Freescale: change descriptor release process for supporting async_tx
Fix the potential risk when enable config NET_DMA and ASYNC_TX. Async_tx is
lack of support in current release process of dma descriptor, all descriptors
will be released whatever is acked or no-acked by async_tx, so there is a
potential race condition when dma engine is uesd by others clients (e.g. when
enable NET_DMA to offload TCP).
In our case, a race condition which is raised when use both of talitos and
dmaengine to offload xor is because napi scheduler will sync all pending
requests in dma channels, it affects the process of raid operations due to
ack_tx is not checked in fsl dma. The no-acked descriptor is freed which is
submitted just now, as a dependent tx, this freed descriptor trigger
BUG_ON(async_tx_test_ack(depend_tx)) in async_tx_submit().
TASK = ee1a94a0[1390] 'md0_raid5' THREAD: ecf40000 CPU: 0
GPR00: 00000001 ecf41ca0 ee44/921a94a0 0000003f 00000001 c00593e4 00000000 00000001
GPR08: 00000000 a7a7a7a7 00000001 045/920000002 42028042 100a38d4 ed576d98 00000000
GPR16: ed5a11b0 00000000 2b162000 00000200 046/920000000 2d555000 ed3015e8 c15a7aa0
GPR24: 00000000 c155fc40 00000000 ecb63220 ecf41d28 e47/92f640bb0 ef640c30 ecf41ca0
NIP [c02b048c] async_tx_submit+0x6c/0x2b4
LR [c02b068c] async_tx_submit+0x26c/0x2b4
Call Trace:
[ecf41ca0] [c02b068c] async_tx_submit+0x26c/0x2b448/92 (unreliable)
[ecf41cd0] [c02b0a4c] async_memcpy+0x240/0x25c
[ecf41d20] [c0421064] async_copy_data+0xa0/0x17c
[ecf41d70] [c0421cf4] __raid_run_ops+0x874/0xe10
[ecf41df0] [c0426ee4] handle_stripe+0x820/0x25e8
[ecf41e90] [c0429080] raid5d+0x3d4/0x5b4
[ecf41f40] [c04329b8] md_thread+0x138/0x16c
[ecf41f90] [c008277c] kthread+0x8c/0x90
[ecf41ff0] [c0011630] kernel_thread+0x4c/0x68
Another modification in this patch is the change of completed descriptors,
there is a potential risk which caused by exception interrupt, all descriptors
in ld_running list are seemed completed when an interrupt raised, it works fine
under normal condition, but if there is an exception occured, it cannot work as
our excepted. Hardware should not be depend on s/w list, the right way is to
read current descriptor address register to find the last completed descriptor.
If an interrupt is raised by an error, all descriptors in ld_running should not
be seemed finished, or these unfinished descriptors in ld_running will be
released wrongly.
A simple way to reproduce:
Enable dmatest first, then insert some bad descriptors which can trigger
Programming Error interrupts before the good descriptors. Last, the good
descriptors will be freed before they are processsed because of the exception
intrerrupt.
Note: the bad descriptors are only for simulating an exception interrupt. This
case can illustrate the potential risk in current fsl-dma very well.
Signed-off-by: Hongbo Zhang <hongbo.zhang@freescale.com>
Signed-off-by: Qiang Liu <qiang.liu@freescale.com>
Signed-off-by: Ira W. Snyder <iws@ovro.caltech.edu>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2014-05-21 16:03:03 +08:00
|
|
|
/**
|
|
|
|
* fsldma_clean_completed_descriptor - free all descriptors which
|
|
|
|
* has been completed and acked
|
|
|
|
* @chan: Freescale DMA channel
|
|
|
|
*
|
|
|
|
* This function is used on all completed and acked descriptors.
|
|
|
|
* All descriptors should only be freed in this function.
|
|
|
|
*/
|
|
|
|
static void fsldma_clean_completed_descriptor(struct fsldma_chan *chan)
|
|
|
|
{
|
|
|
|
struct fsl_desc_sw *desc, *_desc;
|
|
|
|
|
|
|
|
/* Run the callback for each descriptor, in order */
|
|
|
|
list_for_each_entry_safe(desc, _desc, &chan->ld_completed, node)
|
|
|
|
if (async_tx_test_ack(&desc->async_tx))
|
|
|
|
fsl_dma_free_descriptor(chan, desc);
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* fsldma_run_tx_complete_actions - cleanup a single link descriptor
|
|
|
|
* @chan: Freescale DMA channel
|
|
|
|
* @desc: descriptor to cleanup and free
|
|
|
|
* @cookie: Freescale DMA transaction identifier
|
|
|
|
*
|
|
|
|
* This function is used on a descriptor which has been executed by the DMA
|
|
|
|
* controller. It will run any callbacks, submit any dependencies.
|
|
|
|
*/
|
|
|
|
static dma_cookie_t fsldma_run_tx_complete_actions(struct fsldma_chan *chan,
|
|
|
|
struct fsl_desc_sw *desc, dma_cookie_t cookie)
|
|
|
|
{
|
|
|
|
struct dma_async_tx_descriptor *txd = &desc->async_tx;
|
|
|
|
dma_cookie_t ret = cookie;
|
|
|
|
|
|
|
|
BUG_ON(txd->cookie < 0);
|
|
|
|
|
|
|
|
if (txd->cookie > 0) {
|
|
|
|
ret = txd->cookie;
|
|
|
|
|
2016-07-26 01:33:57 +08:00
|
|
|
dma_descriptor_unmap(txd);
|
dmaengine: Freescale: change descriptor release process for supporting async_tx
Fix the potential risk when enable config NET_DMA and ASYNC_TX. Async_tx is
lack of support in current release process of dma descriptor, all descriptors
will be released whatever is acked or no-acked by async_tx, so there is a
potential race condition when dma engine is uesd by others clients (e.g. when
enable NET_DMA to offload TCP).
In our case, a race condition which is raised when use both of talitos and
dmaengine to offload xor is because napi scheduler will sync all pending
requests in dma channels, it affects the process of raid operations due to
ack_tx is not checked in fsl dma. The no-acked descriptor is freed which is
submitted just now, as a dependent tx, this freed descriptor trigger
BUG_ON(async_tx_test_ack(depend_tx)) in async_tx_submit().
TASK = ee1a94a0[1390] 'md0_raid5' THREAD: ecf40000 CPU: 0
GPR00: 00000001 ecf41ca0 ee44/921a94a0 0000003f 00000001 c00593e4 00000000 00000001
GPR08: 00000000 a7a7a7a7 00000001 045/920000002 42028042 100a38d4 ed576d98 00000000
GPR16: ed5a11b0 00000000 2b162000 00000200 046/920000000 2d555000 ed3015e8 c15a7aa0
GPR24: 00000000 c155fc40 00000000 ecb63220 ecf41d28 e47/92f640bb0 ef640c30 ecf41ca0
NIP [c02b048c] async_tx_submit+0x6c/0x2b4
LR [c02b068c] async_tx_submit+0x26c/0x2b4
Call Trace:
[ecf41ca0] [c02b068c] async_tx_submit+0x26c/0x2b448/92 (unreliable)
[ecf41cd0] [c02b0a4c] async_memcpy+0x240/0x25c
[ecf41d20] [c0421064] async_copy_data+0xa0/0x17c
[ecf41d70] [c0421cf4] __raid_run_ops+0x874/0xe10
[ecf41df0] [c0426ee4] handle_stripe+0x820/0x25e8
[ecf41e90] [c0429080] raid5d+0x3d4/0x5b4
[ecf41f40] [c04329b8] md_thread+0x138/0x16c
[ecf41f90] [c008277c] kthread+0x8c/0x90
[ecf41ff0] [c0011630] kernel_thread+0x4c/0x68
Another modification in this patch is the change of completed descriptors,
there is a potential risk which caused by exception interrupt, all descriptors
in ld_running list are seemed completed when an interrupt raised, it works fine
under normal condition, but if there is an exception occured, it cannot work as
our excepted. Hardware should not be depend on s/w list, the right way is to
read current descriptor address register to find the last completed descriptor.
If an interrupt is raised by an error, all descriptors in ld_running should not
be seemed finished, or these unfinished descriptors in ld_running will be
released wrongly.
A simple way to reproduce:
Enable dmatest first, then insert some bad descriptors which can trigger
Programming Error interrupts before the good descriptors. Last, the good
descriptors will be freed before they are processsed because of the exception
intrerrupt.
Note: the bad descriptors are only for simulating an exception interrupt. This
case can illustrate the potential risk in current fsl-dma very well.
Signed-off-by: Hongbo Zhang <hongbo.zhang@freescale.com>
Signed-off-by: Qiang Liu <qiang.liu@freescale.com>
Signed-off-by: Ira W. Snyder <iws@ovro.caltech.edu>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2014-05-21 16:03:03 +08:00
|
|
|
/* Run the link descriptor callback function */
|
2016-07-21 04:11:17 +08:00
|
|
|
dmaengine_desc_get_callback_invoke(txd, NULL);
|
dmaengine: Freescale: change descriptor release process for supporting async_tx
Fix the potential risk when enable config NET_DMA and ASYNC_TX. Async_tx is
lack of support in current release process of dma descriptor, all descriptors
will be released whatever is acked or no-acked by async_tx, so there is a
potential race condition when dma engine is uesd by others clients (e.g. when
enable NET_DMA to offload TCP).
In our case, a race condition which is raised when use both of talitos and
dmaengine to offload xor is because napi scheduler will sync all pending
requests in dma channels, it affects the process of raid operations due to
ack_tx is not checked in fsl dma. The no-acked descriptor is freed which is
submitted just now, as a dependent tx, this freed descriptor trigger
BUG_ON(async_tx_test_ack(depend_tx)) in async_tx_submit().
TASK = ee1a94a0[1390] 'md0_raid5' THREAD: ecf40000 CPU: 0
GPR00: 00000001 ecf41ca0 ee44/921a94a0 0000003f 00000001 c00593e4 00000000 00000001
GPR08: 00000000 a7a7a7a7 00000001 045/920000002 42028042 100a38d4 ed576d98 00000000
GPR16: ed5a11b0 00000000 2b162000 00000200 046/920000000 2d555000 ed3015e8 c15a7aa0
GPR24: 00000000 c155fc40 00000000 ecb63220 ecf41d28 e47/92f640bb0 ef640c30 ecf41ca0
NIP [c02b048c] async_tx_submit+0x6c/0x2b4
LR [c02b068c] async_tx_submit+0x26c/0x2b4
Call Trace:
[ecf41ca0] [c02b068c] async_tx_submit+0x26c/0x2b448/92 (unreliable)
[ecf41cd0] [c02b0a4c] async_memcpy+0x240/0x25c
[ecf41d20] [c0421064] async_copy_data+0xa0/0x17c
[ecf41d70] [c0421cf4] __raid_run_ops+0x874/0xe10
[ecf41df0] [c0426ee4] handle_stripe+0x820/0x25e8
[ecf41e90] [c0429080] raid5d+0x3d4/0x5b4
[ecf41f40] [c04329b8] md_thread+0x138/0x16c
[ecf41f90] [c008277c] kthread+0x8c/0x90
[ecf41ff0] [c0011630] kernel_thread+0x4c/0x68
Another modification in this patch is the change of completed descriptors,
there is a potential risk which caused by exception interrupt, all descriptors
in ld_running list are seemed completed when an interrupt raised, it works fine
under normal condition, but if there is an exception occured, it cannot work as
our excepted. Hardware should not be depend on s/w list, the right way is to
read current descriptor address register to find the last completed descriptor.
If an interrupt is raised by an error, all descriptors in ld_running should not
be seemed finished, or these unfinished descriptors in ld_running will be
released wrongly.
A simple way to reproduce:
Enable dmatest first, then insert some bad descriptors which can trigger
Programming Error interrupts before the good descriptors. Last, the good
descriptors will be freed before they are processsed because of the exception
intrerrupt.
Note: the bad descriptors are only for simulating an exception interrupt. This
case can illustrate the potential risk in current fsl-dma very well.
Signed-off-by: Hongbo Zhang <hongbo.zhang@freescale.com>
Signed-off-by: Qiang Liu <qiang.liu@freescale.com>
Signed-off-by: Ira W. Snyder <iws@ovro.caltech.edu>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2014-05-21 16:03:03 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
/* Run any dependencies */
|
|
|
|
dma_run_dependencies(txd);
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* fsldma_clean_running_descriptor - move the completed descriptor from
|
|
|
|
* ld_running to ld_completed
|
|
|
|
* @chan: Freescale DMA channel
|
|
|
|
* @desc: the descriptor which is completed
|
|
|
|
*
|
|
|
|
* Free the descriptor directly if acked by async_tx api, or move it to
|
|
|
|
* queue ld_completed.
|
|
|
|
*/
|
|
|
|
static void fsldma_clean_running_descriptor(struct fsldma_chan *chan,
|
|
|
|
struct fsl_desc_sw *desc)
|
|
|
|
{
|
|
|
|
/* Remove from the list of transactions */
|
|
|
|
list_del(&desc->node);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* the client is allowed to attach dependent operations
|
|
|
|
* until 'ack' is set
|
|
|
|
*/
|
|
|
|
if (!async_tx_test_ack(&desc->async_tx)) {
|
|
|
|
/*
|
|
|
|
* Move this descriptor to the list of descriptors which is
|
|
|
|
* completed, but still awaiting the 'ack' bit to be set.
|
|
|
|
*/
|
|
|
|
list_add_tail(&desc->node, &chan->ld_completed);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
dma_pool_free(chan->desc_pool, desc, desc->async_tx.phys);
|
|
|
|
}
|
|
|
|
|
2014-04-18 16:17:48 +08:00
|
|
|
/**
|
|
|
|
* fsl_chan_xfer_ld_queue - transfer any pending transactions
|
|
|
|
* @chan : Freescale DMA channel
|
|
|
|
*
|
|
|
|
* HARDWARE STATE: idle
|
|
|
|
* LOCKING: must hold chan->desc_lock
|
|
|
|
*/
|
|
|
|
static void fsl_chan_xfer_ld_queue(struct fsldma_chan *chan)
|
|
|
|
{
|
|
|
|
struct fsl_desc_sw *desc;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If the list of pending descriptors is empty, then we
|
|
|
|
* don't need to do any work at all
|
|
|
|
*/
|
|
|
|
if (list_empty(&chan->ld_pending)) {
|
|
|
|
chan_dbg(chan, "no pending LDs\n");
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* The DMA controller is not idle, which means that the interrupt
|
|
|
|
* handler will start any queued transactions when it runs after
|
|
|
|
* this transaction finishes
|
|
|
|
*/
|
|
|
|
if (!chan->idle) {
|
|
|
|
chan_dbg(chan, "DMA controller still busy\n");
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If there are some link descriptors which have not been
|
|
|
|
* transferred, we need to start the controller
|
|
|
|
*/
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Move all elements from the queue of pending transactions
|
|
|
|
* onto the list of running transactions
|
|
|
|
*/
|
|
|
|
chan_dbg(chan, "idle, starting controller\n");
|
|
|
|
desc = list_first_entry(&chan->ld_pending, struct fsl_desc_sw, node);
|
|
|
|
list_splice_tail_init(&chan->ld_pending, &chan->ld_running);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* The 85xx DMA controller doesn't clear the channel start bit
|
|
|
|
* automatically at the end of a transfer. Therefore we must clear
|
|
|
|
* it in software before starting the transfer.
|
|
|
|
*/
|
|
|
|
if ((chan->feature & FSL_DMA_IP_MASK) == FSL_DMA_IP_85XX) {
|
|
|
|
u32 mode;
|
|
|
|
|
|
|
|
mode = get_mr(chan);
|
|
|
|
mode &= ~FSL_DMA_MR_CS;
|
|
|
|
set_mr(chan, mode);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Program the descriptor's address into the DMA controller,
|
|
|
|
* then start the DMA transaction
|
|
|
|
*/
|
|
|
|
set_cdar(chan, desc->async_tx.phys);
|
|
|
|
get_cdar(chan);
|
|
|
|
|
|
|
|
dma_start(chan);
|
|
|
|
chan->idle = false;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
dmaengine: Freescale: change descriptor release process for supporting async_tx
Fix the potential risk when enable config NET_DMA and ASYNC_TX. Async_tx is
lack of support in current release process of dma descriptor, all descriptors
will be released whatever is acked or no-acked by async_tx, so there is a
potential race condition when dma engine is uesd by others clients (e.g. when
enable NET_DMA to offload TCP).
In our case, a race condition which is raised when use both of talitos and
dmaengine to offload xor is because napi scheduler will sync all pending
requests in dma channels, it affects the process of raid operations due to
ack_tx is not checked in fsl dma. The no-acked descriptor is freed which is
submitted just now, as a dependent tx, this freed descriptor trigger
BUG_ON(async_tx_test_ack(depend_tx)) in async_tx_submit().
TASK = ee1a94a0[1390] 'md0_raid5' THREAD: ecf40000 CPU: 0
GPR00: 00000001 ecf41ca0 ee44/921a94a0 0000003f 00000001 c00593e4 00000000 00000001
GPR08: 00000000 a7a7a7a7 00000001 045/920000002 42028042 100a38d4 ed576d98 00000000
GPR16: ed5a11b0 00000000 2b162000 00000200 046/920000000 2d555000 ed3015e8 c15a7aa0
GPR24: 00000000 c155fc40 00000000 ecb63220 ecf41d28 e47/92f640bb0 ef640c30 ecf41ca0
NIP [c02b048c] async_tx_submit+0x6c/0x2b4
LR [c02b068c] async_tx_submit+0x26c/0x2b4
Call Trace:
[ecf41ca0] [c02b068c] async_tx_submit+0x26c/0x2b448/92 (unreliable)
[ecf41cd0] [c02b0a4c] async_memcpy+0x240/0x25c
[ecf41d20] [c0421064] async_copy_data+0xa0/0x17c
[ecf41d70] [c0421cf4] __raid_run_ops+0x874/0xe10
[ecf41df0] [c0426ee4] handle_stripe+0x820/0x25e8
[ecf41e90] [c0429080] raid5d+0x3d4/0x5b4
[ecf41f40] [c04329b8] md_thread+0x138/0x16c
[ecf41f90] [c008277c] kthread+0x8c/0x90
[ecf41ff0] [c0011630] kernel_thread+0x4c/0x68
Another modification in this patch is the change of completed descriptors,
there is a potential risk which caused by exception interrupt, all descriptors
in ld_running list are seemed completed when an interrupt raised, it works fine
under normal condition, but if there is an exception occured, it cannot work as
our excepted. Hardware should not be depend on s/w list, the right way is to
read current descriptor address register to find the last completed descriptor.
If an interrupt is raised by an error, all descriptors in ld_running should not
be seemed finished, or these unfinished descriptors in ld_running will be
released wrongly.
A simple way to reproduce:
Enable dmatest first, then insert some bad descriptors which can trigger
Programming Error interrupts before the good descriptors. Last, the good
descriptors will be freed before they are processsed because of the exception
intrerrupt.
Note: the bad descriptors are only for simulating an exception interrupt. This
case can illustrate the potential risk in current fsl-dma very well.
Signed-off-by: Hongbo Zhang <hongbo.zhang@freescale.com>
Signed-off-by: Qiang Liu <qiang.liu@freescale.com>
Signed-off-by: Ira W. Snyder <iws@ovro.caltech.edu>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2014-05-21 16:03:03 +08:00
|
|
|
* fsldma_cleanup_descriptors - cleanup link descriptors which are completed
|
|
|
|
* and move them to ld_completed to free until flag 'ack' is set
|
2014-04-18 16:17:48 +08:00
|
|
|
* @chan: Freescale DMA channel
|
|
|
|
*
|
dmaengine: Freescale: change descriptor release process for supporting async_tx
Fix the potential risk when enable config NET_DMA and ASYNC_TX. Async_tx is
lack of support in current release process of dma descriptor, all descriptors
will be released whatever is acked or no-acked by async_tx, so there is a
potential race condition when dma engine is uesd by others clients (e.g. when
enable NET_DMA to offload TCP).
In our case, a race condition which is raised when use both of talitos and
dmaengine to offload xor is because napi scheduler will sync all pending
requests in dma channels, it affects the process of raid operations due to
ack_tx is not checked in fsl dma. The no-acked descriptor is freed which is
submitted just now, as a dependent tx, this freed descriptor trigger
BUG_ON(async_tx_test_ack(depend_tx)) in async_tx_submit().
TASK = ee1a94a0[1390] 'md0_raid5' THREAD: ecf40000 CPU: 0
GPR00: 00000001 ecf41ca0 ee44/921a94a0 0000003f 00000001 c00593e4 00000000 00000001
GPR08: 00000000 a7a7a7a7 00000001 045/920000002 42028042 100a38d4 ed576d98 00000000
GPR16: ed5a11b0 00000000 2b162000 00000200 046/920000000 2d555000 ed3015e8 c15a7aa0
GPR24: 00000000 c155fc40 00000000 ecb63220 ecf41d28 e47/92f640bb0 ef640c30 ecf41ca0
NIP [c02b048c] async_tx_submit+0x6c/0x2b4
LR [c02b068c] async_tx_submit+0x26c/0x2b4
Call Trace:
[ecf41ca0] [c02b068c] async_tx_submit+0x26c/0x2b448/92 (unreliable)
[ecf41cd0] [c02b0a4c] async_memcpy+0x240/0x25c
[ecf41d20] [c0421064] async_copy_data+0xa0/0x17c
[ecf41d70] [c0421cf4] __raid_run_ops+0x874/0xe10
[ecf41df0] [c0426ee4] handle_stripe+0x820/0x25e8
[ecf41e90] [c0429080] raid5d+0x3d4/0x5b4
[ecf41f40] [c04329b8] md_thread+0x138/0x16c
[ecf41f90] [c008277c] kthread+0x8c/0x90
[ecf41ff0] [c0011630] kernel_thread+0x4c/0x68
Another modification in this patch is the change of completed descriptors,
there is a potential risk which caused by exception interrupt, all descriptors
in ld_running list are seemed completed when an interrupt raised, it works fine
under normal condition, but if there is an exception occured, it cannot work as
our excepted. Hardware should not be depend on s/w list, the right way is to
read current descriptor address register to find the last completed descriptor.
If an interrupt is raised by an error, all descriptors in ld_running should not
be seemed finished, or these unfinished descriptors in ld_running will be
released wrongly.
A simple way to reproduce:
Enable dmatest first, then insert some bad descriptors which can trigger
Programming Error interrupts before the good descriptors. Last, the good
descriptors will be freed before they are processsed because of the exception
intrerrupt.
Note: the bad descriptors are only for simulating an exception interrupt. This
case can illustrate the potential risk in current fsl-dma very well.
Signed-off-by: Hongbo Zhang <hongbo.zhang@freescale.com>
Signed-off-by: Qiang Liu <qiang.liu@freescale.com>
Signed-off-by: Ira W. Snyder <iws@ovro.caltech.edu>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2014-05-21 16:03:03 +08:00
|
|
|
* This function is used on descriptors which have been executed by the DMA
|
|
|
|
* controller. It will run any callbacks, submit any dependencies, then
|
|
|
|
* free these descriptors if flag 'ack' is set.
|
2014-04-18 16:17:48 +08:00
|
|
|
*/
|
dmaengine: Freescale: change descriptor release process for supporting async_tx
Fix the potential risk when enable config NET_DMA and ASYNC_TX. Async_tx is
lack of support in current release process of dma descriptor, all descriptors
will be released whatever is acked or no-acked by async_tx, so there is a
potential race condition when dma engine is uesd by others clients (e.g. when
enable NET_DMA to offload TCP).
In our case, a race condition which is raised when use both of talitos and
dmaengine to offload xor is because napi scheduler will sync all pending
requests in dma channels, it affects the process of raid operations due to
ack_tx is not checked in fsl dma. The no-acked descriptor is freed which is
submitted just now, as a dependent tx, this freed descriptor trigger
BUG_ON(async_tx_test_ack(depend_tx)) in async_tx_submit().
TASK = ee1a94a0[1390] 'md0_raid5' THREAD: ecf40000 CPU: 0
GPR00: 00000001 ecf41ca0 ee44/921a94a0 0000003f 00000001 c00593e4 00000000 00000001
GPR08: 00000000 a7a7a7a7 00000001 045/920000002 42028042 100a38d4 ed576d98 00000000
GPR16: ed5a11b0 00000000 2b162000 00000200 046/920000000 2d555000 ed3015e8 c15a7aa0
GPR24: 00000000 c155fc40 00000000 ecb63220 ecf41d28 e47/92f640bb0 ef640c30 ecf41ca0
NIP [c02b048c] async_tx_submit+0x6c/0x2b4
LR [c02b068c] async_tx_submit+0x26c/0x2b4
Call Trace:
[ecf41ca0] [c02b068c] async_tx_submit+0x26c/0x2b448/92 (unreliable)
[ecf41cd0] [c02b0a4c] async_memcpy+0x240/0x25c
[ecf41d20] [c0421064] async_copy_data+0xa0/0x17c
[ecf41d70] [c0421cf4] __raid_run_ops+0x874/0xe10
[ecf41df0] [c0426ee4] handle_stripe+0x820/0x25e8
[ecf41e90] [c0429080] raid5d+0x3d4/0x5b4
[ecf41f40] [c04329b8] md_thread+0x138/0x16c
[ecf41f90] [c008277c] kthread+0x8c/0x90
[ecf41ff0] [c0011630] kernel_thread+0x4c/0x68
Another modification in this patch is the change of completed descriptors,
there is a potential risk which caused by exception interrupt, all descriptors
in ld_running list are seemed completed when an interrupt raised, it works fine
under normal condition, but if there is an exception occured, it cannot work as
our excepted. Hardware should not be depend on s/w list, the right way is to
read current descriptor address register to find the last completed descriptor.
If an interrupt is raised by an error, all descriptors in ld_running should not
be seemed finished, or these unfinished descriptors in ld_running will be
released wrongly.
A simple way to reproduce:
Enable dmatest first, then insert some bad descriptors which can trigger
Programming Error interrupts before the good descriptors. Last, the good
descriptors will be freed before they are processsed because of the exception
intrerrupt.
Note: the bad descriptors are only for simulating an exception interrupt. This
case can illustrate the potential risk in current fsl-dma very well.
Signed-off-by: Hongbo Zhang <hongbo.zhang@freescale.com>
Signed-off-by: Qiang Liu <qiang.liu@freescale.com>
Signed-off-by: Ira W. Snyder <iws@ovro.caltech.edu>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2014-05-21 16:03:03 +08:00
|
|
|
static void fsldma_cleanup_descriptors(struct fsldma_chan *chan)
|
2014-04-18 16:17:48 +08:00
|
|
|
{
|
dmaengine: Freescale: change descriptor release process for supporting async_tx
Fix the potential risk when enable config NET_DMA and ASYNC_TX. Async_tx is
lack of support in current release process of dma descriptor, all descriptors
will be released whatever is acked or no-acked by async_tx, so there is a
potential race condition when dma engine is uesd by others clients (e.g. when
enable NET_DMA to offload TCP).
In our case, a race condition which is raised when use both of talitos and
dmaengine to offload xor is because napi scheduler will sync all pending
requests in dma channels, it affects the process of raid operations due to
ack_tx is not checked in fsl dma. The no-acked descriptor is freed which is
submitted just now, as a dependent tx, this freed descriptor trigger
BUG_ON(async_tx_test_ack(depend_tx)) in async_tx_submit().
TASK = ee1a94a0[1390] 'md0_raid5' THREAD: ecf40000 CPU: 0
GPR00: 00000001 ecf41ca0 ee44/921a94a0 0000003f 00000001 c00593e4 00000000 00000001
GPR08: 00000000 a7a7a7a7 00000001 045/920000002 42028042 100a38d4 ed576d98 00000000
GPR16: ed5a11b0 00000000 2b162000 00000200 046/920000000 2d555000 ed3015e8 c15a7aa0
GPR24: 00000000 c155fc40 00000000 ecb63220 ecf41d28 e47/92f640bb0 ef640c30 ecf41ca0
NIP [c02b048c] async_tx_submit+0x6c/0x2b4
LR [c02b068c] async_tx_submit+0x26c/0x2b4
Call Trace:
[ecf41ca0] [c02b068c] async_tx_submit+0x26c/0x2b448/92 (unreliable)
[ecf41cd0] [c02b0a4c] async_memcpy+0x240/0x25c
[ecf41d20] [c0421064] async_copy_data+0xa0/0x17c
[ecf41d70] [c0421cf4] __raid_run_ops+0x874/0xe10
[ecf41df0] [c0426ee4] handle_stripe+0x820/0x25e8
[ecf41e90] [c0429080] raid5d+0x3d4/0x5b4
[ecf41f40] [c04329b8] md_thread+0x138/0x16c
[ecf41f90] [c008277c] kthread+0x8c/0x90
[ecf41ff0] [c0011630] kernel_thread+0x4c/0x68
Another modification in this patch is the change of completed descriptors,
there is a potential risk which caused by exception interrupt, all descriptors
in ld_running list are seemed completed when an interrupt raised, it works fine
under normal condition, but if there is an exception occured, it cannot work as
our excepted. Hardware should not be depend on s/w list, the right way is to
read current descriptor address register to find the last completed descriptor.
If an interrupt is raised by an error, all descriptors in ld_running should not
be seemed finished, or these unfinished descriptors in ld_running will be
released wrongly.
A simple way to reproduce:
Enable dmatest first, then insert some bad descriptors which can trigger
Programming Error interrupts before the good descriptors. Last, the good
descriptors will be freed before they are processsed because of the exception
intrerrupt.
Note: the bad descriptors are only for simulating an exception interrupt. This
case can illustrate the potential risk in current fsl-dma very well.
Signed-off-by: Hongbo Zhang <hongbo.zhang@freescale.com>
Signed-off-by: Qiang Liu <qiang.liu@freescale.com>
Signed-off-by: Ira W. Snyder <iws@ovro.caltech.edu>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2014-05-21 16:03:03 +08:00
|
|
|
struct fsl_desc_sw *desc, *_desc;
|
|
|
|
dma_cookie_t cookie = 0;
|
|
|
|
dma_addr_t curr_phys = get_cdar(chan);
|
|
|
|
int seen_current = 0;
|
2014-04-18 16:17:48 +08:00
|
|
|
|
dmaengine: Freescale: change descriptor release process for supporting async_tx
Fix the potential risk when enable config NET_DMA and ASYNC_TX. Async_tx is
lack of support in current release process of dma descriptor, all descriptors
will be released whatever is acked or no-acked by async_tx, so there is a
potential race condition when dma engine is uesd by others clients (e.g. when
enable NET_DMA to offload TCP).
In our case, a race condition which is raised when use both of talitos and
dmaengine to offload xor is because napi scheduler will sync all pending
requests in dma channels, it affects the process of raid operations due to
ack_tx is not checked in fsl dma. The no-acked descriptor is freed which is
submitted just now, as a dependent tx, this freed descriptor trigger
BUG_ON(async_tx_test_ack(depend_tx)) in async_tx_submit().
TASK = ee1a94a0[1390] 'md0_raid5' THREAD: ecf40000 CPU: 0
GPR00: 00000001 ecf41ca0 ee44/921a94a0 0000003f 00000001 c00593e4 00000000 00000001
GPR08: 00000000 a7a7a7a7 00000001 045/920000002 42028042 100a38d4 ed576d98 00000000
GPR16: ed5a11b0 00000000 2b162000 00000200 046/920000000 2d555000 ed3015e8 c15a7aa0
GPR24: 00000000 c155fc40 00000000 ecb63220 ecf41d28 e47/92f640bb0 ef640c30 ecf41ca0
NIP [c02b048c] async_tx_submit+0x6c/0x2b4
LR [c02b068c] async_tx_submit+0x26c/0x2b4
Call Trace:
[ecf41ca0] [c02b068c] async_tx_submit+0x26c/0x2b448/92 (unreliable)
[ecf41cd0] [c02b0a4c] async_memcpy+0x240/0x25c
[ecf41d20] [c0421064] async_copy_data+0xa0/0x17c
[ecf41d70] [c0421cf4] __raid_run_ops+0x874/0xe10
[ecf41df0] [c0426ee4] handle_stripe+0x820/0x25e8
[ecf41e90] [c0429080] raid5d+0x3d4/0x5b4
[ecf41f40] [c04329b8] md_thread+0x138/0x16c
[ecf41f90] [c008277c] kthread+0x8c/0x90
[ecf41ff0] [c0011630] kernel_thread+0x4c/0x68
Another modification in this patch is the change of completed descriptors,
there is a potential risk which caused by exception interrupt, all descriptors
in ld_running list are seemed completed when an interrupt raised, it works fine
under normal condition, but if there is an exception occured, it cannot work as
our excepted. Hardware should not be depend on s/w list, the right way is to
read current descriptor address register to find the last completed descriptor.
If an interrupt is raised by an error, all descriptors in ld_running should not
be seemed finished, or these unfinished descriptors in ld_running will be
released wrongly.
A simple way to reproduce:
Enable dmatest first, then insert some bad descriptors which can trigger
Programming Error interrupts before the good descriptors. Last, the good
descriptors will be freed before they are processsed because of the exception
intrerrupt.
Note: the bad descriptors are only for simulating an exception interrupt. This
case can illustrate the potential risk in current fsl-dma very well.
Signed-off-by: Hongbo Zhang <hongbo.zhang@freescale.com>
Signed-off-by: Qiang Liu <qiang.liu@freescale.com>
Signed-off-by: Ira W. Snyder <iws@ovro.caltech.edu>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2014-05-21 16:03:03 +08:00
|
|
|
fsldma_clean_completed_descriptor(chan);
|
|
|
|
|
|
|
|
/* Run the callback for each descriptor, in order */
|
|
|
|
list_for_each_entry_safe(desc, _desc, &chan->ld_running, node) {
|
|
|
|
/*
|
|
|
|
* do not advance past the current descriptor loaded into the
|
|
|
|
* hardware channel, subsequent descriptors are either in
|
|
|
|
* process or have not been submitted
|
|
|
|
*/
|
|
|
|
if (seen_current)
|
|
|
|
break;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* stop the search if we reach the current descriptor and the
|
|
|
|
* channel is busy
|
|
|
|
*/
|
|
|
|
if (desc->async_tx.phys == curr_phys) {
|
|
|
|
seen_current = 1;
|
|
|
|
if (!dma_is_idle(chan))
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
cookie = fsldma_run_tx_complete_actions(chan, desc, cookie);
|
|
|
|
|
|
|
|
fsldma_clean_running_descriptor(chan, desc);
|
2014-04-18 16:17:48 +08:00
|
|
|
}
|
|
|
|
|
dmaengine: Freescale: change descriptor release process for supporting async_tx
Fix the potential risk when enable config NET_DMA and ASYNC_TX. Async_tx is
lack of support in current release process of dma descriptor, all descriptors
will be released whatever is acked or no-acked by async_tx, so there is a
potential race condition when dma engine is uesd by others clients (e.g. when
enable NET_DMA to offload TCP).
In our case, a race condition which is raised when use both of talitos and
dmaengine to offload xor is because napi scheduler will sync all pending
requests in dma channels, it affects the process of raid operations due to
ack_tx is not checked in fsl dma. The no-acked descriptor is freed which is
submitted just now, as a dependent tx, this freed descriptor trigger
BUG_ON(async_tx_test_ack(depend_tx)) in async_tx_submit().
TASK = ee1a94a0[1390] 'md0_raid5' THREAD: ecf40000 CPU: 0
GPR00: 00000001 ecf41ca0 ee44/921a94a0 0000003f 00000001 c00593e4 00000000 00000001
GPR08: 00000000 a7a7a7a7 00000001 045/920000002 42028042 100a38d4 ed576d98 00000000
GPR16: ed5a11b0 00000000 2b162000 00000200 046/920000000 2d555000 ed3015e8 c15a7aa0
GPR24: 00000000 c155fc40 00000000 ecb63220 ecf41d28 e47/92f640bb0 ef640c30 ecf41ca0
NIP [c02b048c] async_tx_submit+0x6c/0x2b4
LR [c02b068c] async_tx_submit+0x26c/0x2b4
Call Trace:
[ecf41ca0] [c02b068c] async_tx_submit+0x26c/0x2b448/92 (unreliable)
[ecf41cd0] [c02b0a4c] async_memcpy+0x240/0x25c
[ecf41d20] [c0421064] async_copy_data+0xa0/0x17c
[ecf41d70] [c0421cf4] __raid_run_ops+0x874/0xe10
[ecf41df0] [c0426ee4] handle_stripe+0x820/0x25e8
[ecf41e90] [c0429080] raid5d+0x3d4/0x5b4
[ecf41f40] [c04329b8] md_thread+0x138/0x16c
[ecf41f90] [c008277c] kthread+0x8c/0x90
[ecf41ff0] [c0011630] kernel_thread+0x4c/0x68
Another modification in this patch is the change of completed descriptors,
there is a potential risk which caused by exception interrupt, all descriptors
in ld_running list are seemed completed when an interrupt raised, it works fine
under normal condition, but if there is an exception occured, it cannot work as
our excepted. Hardware should not be depend on s/w list, the right way is to
read current descriptor address register to find the last completed descriptor.
If an interrupt is raised by an error, all descriptors in ld_running should not
be seemed finished, or these unfinished descriptors in ld_running will be
released wrongly.
A simple way to reproduce:
Enable dmatest first, then insert some bad descriptors which can trigger
Programming Error interrupts before the good descriptors. Last, the good
descriptors will be freed before they are processsed because of the exception
intrerrupt.
Note: the bad descriptors are only for simulating an exception interrupt. This
case can illustrate the potential risk in current fsl-dma very well.
Signed-off-by: Hongbo Zhang <hongbo.zhang@freescale.com>
Signed-off-by: Qiang Liu <qiang.liu@freescale.com>
Signed-off-by: Ira W. Snyder <iws@ovro.caltech.edu>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2014-05-21 16:03:03 +08:00
|
|
|
/*
|
|
|
|
* Start any pending transactions automatically
|
|
|
|
*
|
|
|
|
* In the ideal case, we keep the DMA controller busy while we go
|
|
|
|
* ahead and free the descriptors below.
|
|
|
|
*/
|
|
|
|
fsl_chan_xfer_ld_queue(chan);
|
2014-04-18 16:17:48 +08:00
|
|
|
|
dmaengine: Freescale: change descriptor release process for supporting async_tx
Fix the potential risk when enable config NET_DMA and ASYNC_TX. Async_tx is
lack of support in current release process of dma descriptor, all descriptors
will be released whatever is acked or no-acked by async_tx, so there is a
potential race condition when dma engine is uesd by others clients (e.g. when
enable NET_DMA to offload TCP).
In our case, a race condition which is raised when use both of talitos and
dmaengine to offload xor is because napi scheduler will sync all pending
requests in dma channels, it affects the process of raid operations due to
ack_tx is not checked in fsl dma. The no-acked descriptor is freed which is
submitted just now, as a dependent tx, this freed descriptor trigger
BUG_ON(async_tx_test_ack(depend_tx)) in async_tx_submit().
TASK = ee1a94a0[1390] 'md0_raid5' THREAD: ecf40000 CPU: 0
GPR00: 00000001 ecf41ca0 ee44/921a94a0 0000003f 00000001 c00593e4 00000000 00000001
GPR08: 00000000 a7a7a7a7 00000001 045/920000002 42028042 100a38d4 ed576d98 00000000
GPR16: ed5a11b0 00000000 2b162000 00000200 046/920000000 2d555000 ed3015e8 c15a7aa0
GPR24: 00000000 c155fc40 00000000 ecb63220 ecf41d28 e47/92f640bb0 ef640c30 ecf41ca0
NIP [c02b048c] async_tx_submit+0x6c/0x2b4
LR [c02b068c] async_tx_submit+0x26c/0x2b4
Call Trace:
[ecf41ca0] [c02b068c] async_tx_submit+0x26c/0x2b448/92 (unreliable)
[ecf41cd0] [c02b0a4c] async_memcpy+0x240/0x25c
[ecf41d20] [c0421064] async_copy_data+0xa0/0x17c
[ecf41d70] [c0421cf4] __raid_run_ops+0x874/0xe10
[ecf41df0] [c0426ee4] handle_stripe+0x820/0x25e8
[ecf41e90] [c0429080] raid5d+0x3d4/0x5b4
[ecf41f40] [c04329b8] md_thread+0x138/0x16c
[ecf41f90] [c008277c] kthread+0x8c/0x90
[ecf41ff0] [c0011630] kernel_thread+0x4c/0x68
Another modification in this patch is the change of completed descriptors,
there is a potential risk which caused by exception interrupt, all descriptors
in ld_running list are seemed completed when an interrupt raised, it works fine
under normal condition, but if there is an exception occured, it cannot work as
our excepted. Hardware should not be depend on s/w list, the right way is to
read current descriptor address register to find the last completed descriptor.
If an interrupt is raised by an error, all descriptors in ld_running should not
be seemed finished, or these unfinished descriptors in ld_running will be
released wrongly.
A simple way to reproduce:
Enable dmatest first, then insert some bad descriptors which can trigger
Programming Error interrupts before the good descriptors. Last, the good
descriptors will be freed before they are processsed because of the exception
intrerrupt.
Note: the bad descriptors are only for simulating an exception interrupt. This
case can illustrate the potential risk in current fsl-dma very well.
Signed-off-by: Hongbo Zhang <hongbo.zhang@freescale.com>
Signed-off-by: Qiang Liu <qiang.liu@freescale.com>
Signed-off-by: Ira W. Snyder <iws@ovro.caltech.edu>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2014-05-21 16:03:03 +08:00
|
|
|
if (cookie > 0)
|
|
|
|
chan->common.completed_cookie = cookie;
|
2014-04-18 16:17:48 +08:00
|
|
|
}
|
|
|
|
|
2008-03-01 22:42:48 +08:00
|
|
|
/**
|
|
|
|
* fsl_dma_alloc_chan_resources - Allocate resources for DMA channel.
|
2010-01-06 21:34:05 +08:00
|
|
|
* @chan : Freescale DMA channel
|
2008-03-01 22:42:48 +08:00
|
|
|
*
|
|
|
|
* This function will create a dma pool for descriptor allocation.
|
|
|
|
*
|
|
|
|
* Return - The number of descriptors allocated.
|
|
|
|
*/
|
2010-01-06 21:34:05 +08:00
|
|
|
static int fsl_dma_alloc_chan_resources(struct dma_chan *dchan)
|
2008-03-01 22:42:48 +08:00
|
|
|
{
|
2010-01-06 21:34:05 +08:00
|
|
|
struct fsldma_chan *chan = to_fsl_chan(dchan);
|
2008-09-27 08:00:11 +08:00
|
|
|
|
|
|
|
/* Has this channel already been allocated? */
|
2010-01-06 21:34:05 +08:00
|
|
|
if (chan->desc_pool)
|
2008-09-27 08:00:11 +08:00
|
|
|
return 1;
|
2008-03-01 22:42:48 +08:00
|
|
|
|
2010-01-06 21:34:06 +08:00
|
|
|
/*
|
|
|
|
* We need the descriptor to be aligned to 32bytes
|
2008-03-01 22:42:48 +08:00
|
|
|
* for meeting FSL DMA specification requirement.
|
|
|
|
*/
|
2011-03-03 15:54:55 +08:00
|
|
|
chan->desc_pool = dma_pool_create(chan->name, chan->dev,
|
2010-01-06 21:34:06 +08:00
|
|
|
sizeof(struct fsl_desc_sw),
|
|
|
|
__alignof__(struct fsl_desc_sw), 0);
|
2010-01-06 21:34:05 +08:00
|
|
|
if (!chan->desc_pool) {
|
2011-03-03 15:54:55 +08:00
|
|
|
chan_err(chan, "unable to allocate descriptor pool\n");
|
2010-01-06 21:34:06 +08:00
|
|
|
return -ENOMEM;
|
2008-03-01 22:42:48 +08:00
|
|
|
}
|
|
|
|
|
2010-01-06 21:34:06 +08:00
|
|
|
/* there is at least one descriptor free to be allocated */
|
2008-03-01 22:42:48 +08:00
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
2010-01-06 21:34:06 +08:00
|
|
|
/**
|
|
|
|
* fsldma_free_desc_list - Free all descriptors in a queue
|
|
|
|
* @chan: Freescae DMA channel
|
|
|
|
* @list: the list to free
|
|
|
|
*
|
|
|
|
* LOCKING: must hold chan->desc_lock
|
|
|
|
*/
|
|
|
|
static void fsldma_free_desc_list(struct fsldma_chan *chan,
|
|
|
|
struct list_head *list)
|
|
|
|
{
|
|
|
|
struct fsl_desc_sw *desc, *_desc;
|
|
|
|
|
2014-04-18 16:17:47 +08:00
|
|
|
list_for_each_entry_safe(desc, _desc, list, node)
|
|
|
|
fsl_dma_free_descriptor(chan, desc);
|
2010-01-06 21:34:06 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
static void fsldma_free_desc_list_reverse(struct fsldma_chan *chan,
|
|
|
|
struct list_head *list)
|
|
|
|
{
|
|
|
|
struct fsl_desc_sw *desc, *_desc;
|
|
|
|
|
2014-04-18 16:17:47 +08:00
|
|
|
list_for_each_entry_safe_reverse(desc, _desc, list, node)
|
|
|
|
fsl_dma_free_descriptor(chan, desc);
|
2010-01-06 21:34:06 +08:00
|
|
|
}
|
|
|
|
|
2008-03-01 22:42:48 +08:00
|
|
|
/**
|
|
|
|
* fsl_dma_free_chan_resources - Free all resources of the channel.
|
2010-01-06 21:34:05 +08:00
|
|
|
* @chan : Freescale DMA channel
|
2008-03-01 22:42:48 +08:00
|
|
|
*/
|
2010-01-06 21:34:05 +08:00
|
|
|
static void fsl_dma_free_chan_resources(struct dma_chan *dchan)
|
2008-03-01 22:42:48 +08:00
|
|
|
{
|
2010-01-06 21:34:05 +08:00
|
|
|
struct fsldma_chan *chan = to_fsl_chan(dchan);
|
2008-03-01 22:42:48 +08:00
|
|
|
|
2011-03-03 15:54:55 +08:00
|
|
|
chan_dbg(chan, "free all channel resources\n");
|
2014-05-21 16:03:01 +08:00
|
|
|
spin_lock_bh(&chan->desc_lock);
|
dmaengine: Freescale: change descriptor release process for supporting async_tx
Fix the potential risk when enable config NET_DMA and ASYNC_TX. Async_tx is
lack of support in current release process of dma descriptor, all descriptors
will be released whatever is acked or no-acked by async_tx, so there is a
potential race condition when dma engine is uesd by others clients (e.g. when
enable NET_DMA to offload TCP).
In our case, a race condition which is raised when use both of talitos and
dmaengine to offload xor is because napi scheduler will sync all pending
requests in dma channels, it affects the process of raid operations due to
ack_tx is not checked in fsl dma. The no-acked descriptor is freed which is
submitted just now, as a dependent tx, this freed descriptor trigger
BUG_ON(async_tx_test_ack(depend_tx)) in async_tx_submit().
TASK = ee1a94a0[1390] 'md0_raid5' THREAD: ecf40000 CPU: 0
GPR00: 00000001 ecf41ca0 ee44/921a94a0 0000003f 00000001 c00593e4 00000000 00000001
GPR08: 00000000 a7a7a7a7 00000001 045/920000002 42028042 100a38d4 ed576d98 00000000
GPR16: ed5a11b0 00000000 2b162000 00000200 046/920000000 2d555000 ed3015e8 c15a7aa0
GPR24: 00000000 c155fc40 00000000 ecb63220 ecf41d28 e47/92f640bb0 ef640c30 ecf41ca0
NIP [c02b048c] async_tx_submit+0x6c/0x2b4
LR [c02b068c] async_tx_submit+0x26c/0x2b4
Call Trace:
[ecf41ca0] [c02b068c] async_tx_submit+0x26c/0x2b448/92 (unreliable)
[ecf41cd0] [c02b0a4c] async_memcpy+0x240/0x25c
[ecf41d20] [c0421064] async_copy_data+0xa0/0x17c
[ecf41d70] [c0421cf4] __raid_run_ops+0x874/0xe10
[ecf41df0] [c0426ee4] handle_stripe+0x820/0x25e8
[ecf41e90] [c0429080] raid5d+0x3d4/0x5b4
[ecf41f40] [c04329b8] md_thread+0x138/0x16c
[ecf41f90] [c008277c] kthread+0x8c/0x90
[ecf41ff0] [c0011630] kernel_thread+0x4c/0x68
Another modification in this patch is the change of completed descriptors,
there is a potential risk which caused by exception interrupt, all descriptors
in ld_running list are seemed completed when an interrupt raised, it works fine
under normal condition, but if there is an exception occured, it cannot work as
our excepted. Hardware should not be depend on s/w list, the right way is to
read current descriptor address register to find the last completed descriptor.
If an interrupt is raised by an error, all descriptors in ld_running should not
be seemed finished, or these unfinished descriptors in ld_running will be
released wrongly.
A simple way to reproduce:
Enable dmatest first, then insert some bad descriptors which can trigger
Programming Error interrupts before the good descriptors. Last, the good
descriptors will be freed before they are processsed because of the exception
intrerrupt.
Note: the bad descriptors are only for simulating an exception interrupt. This
case can illustrate the potential risk in current fsl-dma very well.
Signed-off-by: Hongbo Zhang <hongbo.zhang@freescale.com>
Signed-off-by: Qiang Liu <qiang.liu@freescale.com>
Signed-off-by: Ira W. Snyder <iws@ovro.caltech.edu>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2014-05-21 16:03:03 +08:00
|
|
|
fsldma_cleanup_descriptors(chan);
|
2010-01-06 21:34:06 +08:00
|
|
|
fsldma_free_desc_list(chan, &chan->ld_pending);
|
|
|
|
fsldma_free_desc_list(chan, &chan->ld_running);
|
dmaengine: Freescale: change descriptor release process for supporting async_tx
Fix the potential risk when enable config NET_DMA and ASYNC_TX. Async_tx is
lack of support in current release process of dma descriptor, all descriptors
will be released whatever is acked or no-acked by async_tx, so there is a
potential race condition when dma engine is uesd by others clients (e.g. when
enable NET_DMA to offload TCP).
In our case, a race condition which is raised when use both of talitos and
dmaengine to offload xor is because napi scheduler will sync all pending
requests in dma channels, it affects the process of raid operations due to
ack_tx is not checked in fsl dma. The no-acked descriptor is freed which is
submitted just now, as a dependent tx, this freed descriptor trigger
BUG_ON(async_tx_test_ack(depend_tx)) in async_tx_submit().
TASK = ee1a94a0[1390] 'md0_raid5' THREAD: ecf40000 CPU: 0
GPR00: 00000001 ecf41ca0 ee44/921a94a0 0000003f 00000001 c00593e4 00000000 00000001
GPR08: 00000000 a7a7a7a7 00000001 045/920000002 42028042 100a38d4 ed576d98 00000000
GPR16: ed5a11b0 00000000 2b162000 00000200 046/920000000 2d555000 ed3015e8 c15a7aa0
GPR24: 00000000 c155fc40 00000000 ecb63220 ecf41d28 e47/92f640bb0 ef640c30 ecf41ca0
NIP [c02b048c] async_tx_submit+0x6c/0x2b4
LR [c02b068c] async_tx_submit+0x26c/0x2b4
Call Trace:
[ecf41ca0] [c02b068c] async_tx_submit+0x26c/0x2b448/92 (unreliable)
[ecf41cd0] [c02b0a4c] async_memcpy+0x240/0x25c
[ecf41d20] [c0421064] async_copy_data+0xa0/0x17c
[ecf41d70] [c0421cf4] __raid_run_ops+0x874/0xe10
[ecf41df0] [c0426ee4] handle_stripe+0x820/0x25e8
[ecf41e90] [c0429080] raid5d+0x3d4/0x5b4
[ecf41f40] [c04329b8] md_thread+0x138/0x16c
[ecf41f90] [c008277c] kthread+0x8c/0x90
[ecf41ff0] [c0011630] kernel_thread+0x4c/0x68
Another modification in this patch is the change of completed descriptors,
there is a potential risk which caused by exception interrupt, all descriptors
in ld_running list are seemed completed when an interrupt raised, it works fine
under normal condition, but if there is an exception occured, it cannot work as
our excepted. Hardware should not be depend on s/w list, the right way is to
read current descriptor address register to find the last completed descriptor.
If an interrupt is raised by an error, all descriptors in ld_running should not
be seemed finished, or these unfinished descriptors in ld_running will be
released wrongly.
A simple way to reproduce:
Enable dmatest first, then insert some bad descriptors which can trigger
Programming Error interrupts before the good descriptors. Last, the good
descriptors will be freed before they are processsed because of the exception
intrerrupt.
Note: the bad descriptors are only for simulating an exception interrupt. This
case can illustrate the potential risk in current fsl-dma very well.
Signed-off-by: Hongbo Zhang <hongbo.zhang@freescale.com>
Signed-off-by: Qiang Liu <qiang.liu@freescale.com>
Signed-off-by: Ira W. Snyder <iws@ovro.caltech.edu>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2014-05-21 16:03:03 +08:00
|
|
|
fsldma_free_desc_list(chan, &chan->ld_completed);
|
2014-05-21 16:03:01 +08:00
|
|
|
spin_unlock_bh(&chan->desc_lock);
|
2008-09-27 08:00:11 +08:00
|
|
|
|
2010-01-06 21:34:06 +08:00
|
|
|
dma_pool_destroy(chan->desc_pool);
|
2010-01-06 21:34:05 +08:00
|
|
|
chan->desc_pool = NULL;
|
2008-03-01 22:42:48 +08:00
|
|
|
}
|
|
|
|
|
2011-03-03 15:54:57 +08:00
|
|
|
static struct dma_async_tx_descriptor *
|
|
|
|
fsl_dma_prep_memcpy(struct dma_chan *dchan,
|
|
|
|
dma_addr_t dma_dst, dma_addr_t dma_src,
|
2008-03-01 22:42:48 +08:00
|
|
|
size_t len, unsigned long flags)
|
|
|
|
{
|
2010-01-06 21:34:05 +08:00
|
|
|
struct fsldma_chan *chan;
|
2008-03-01 22:42:48 +08:00
|
|
|
struct fsl_desc_sw *first = NULL, *prev = NULL, *new;
|
|
|
|
size_t copy;
|
|
|
|
|
2010-01-06 21:34:05 +08:00
|
|
|
if (!dchan)
|
2008-03-01 22:42:48 +08:00
|
|
|
return NULL;
|
|
|
|
|
|
|
|
if (!len)
|
|
|
|
return NULL;
|
|
|
|
|
2010-01-06 21:34:05 +08:00
|
|
|
chan = to_fsl_chan(dchan);
|
2008-03-01 22:42:48 +08:00
|
|
|
|
|
|
|
do {
|
|
|
|
|
|
|
|
/* Allocate the link descriptor from DMA pool */
|
2010-01-06 21:34:05 +08:00
|
|
|
new = fsl_dma_alloc_descriptor(chan);
|
2008-03-01 22:42:48 +08:00
|
|
|
if (!new) {
|
2011-03-03 15:54:55 +08:00
|
|
|
chan_err(chan, "%s\n", msg_ld_oom);
|
2009-05-16 00:59:46 +08:00
|
|
|
goto fail;
|
2008-03-01 22:42:48 +08:00
|
|
|
}
|
|
|
|
|
2008-03-14 01:45:27 +08:00
|
|
|
copy = min(len, (size_t)FSL_DMA_BCR_MAX_CNT);
|
2008-03-01 22:42:48 +08:00
|
|
|
|
2010-01-06 21:34:05 +08:00
|
|
|
set_desc_cnt(chan, &new->hw, copy);
|
|
|
|
set_desc_src(chan, &new->hw, dma_src);
|
|
|
|
set_desc_dst(chan, &new->hw, dma_dst);
|
2008-03-01 22:42:48 +08:00
|
|
|
|
|
|
|
if (!first)
|
|
|
|
first = new;
|
|
|
|
else
|
2010-01-06 21:34:05 +08:00
|
|
|
set_desc_next(chan, &prev->hw, new->async_tx.phys);
|
2008-03-01 22:42:48 +08:00
|
|
|
|
|
|
|
new->async_tx.cookie = 0;
|
2008-04-18 11:17:26 +08:00
|
|
|
async_tx_ack(&new->async_tx);
|
2008-03-01 22:42:48 +08:00
|
|
|
|
|
|
|
prev = new;
|
|
|
|
len -= copy;
|
|
|
|
dma_src += copy;
|
2010-01-06 21:34:02 +08:00
|
|
|
dma_dst += copy;
|
2008-03-01 22:42:48 +08:00
|
|
|
|
|
|
|
/* Insert the link descriptor to the LD ring */
|
2009-09-09 08:53:02 +08:00
|
|
|
list_add_tail(&new->node, &first->tx_list);
|
2008-03-01 22:42:48 +08:00
|
|
|
} while (len);
|
|
|
|
|
2008-04-18 11:17:26 +08:00
|
|
|
new->async_tx.flags = flags; /* client is in control of this ack */
|
2008-03-01 22:42:48 +08:00
|
|
|
new->async_tx.cookie = -EBUSY;
|
|
|
|
|
2011-03-03 15:54:57 +08:00
|
|
|
/* Set End-of-link to the last link descriptor of new list */
|
2010-01-06 21:34:05 +08:00
|
|
|
set_ld_eol(chan, new);
|
2008-03-01 22:42:48 +08:00
|
|
|
|
2009-05-16 00:59:46 +08:00
|
|
|
return &first->async_tx;
|
|
|
|
|
|
|
|
fail:
|
|
|
|
if (!first)
|
|
|
|
return NULL;
|
|
|
|
|
2010-01-06 21:34:06 +08:00
|
|
|
fsldma_free_desc_list_reverse(chan, &first->tx_list);
|
2009-05-16 00:59:46 +08:00
|
|
|
return NULL;
|
2008-03-01 22:42:48 +08:00
|
|
|
}
|
|
|
|
|
2014-11-17 21:42:24 +08:00
|
|
|
static int fsl_dma_device_terminate_all(struct dma_chan *dchan)
|
2009-09-09 08:53:04 +08:00
|
|
|
{
|
2010-01-06 21:34:05 +08:00
|
|
|
struct fsldma_chan *chan;
|
2010-03-27 07:44:01 +08:00
|
|
|
|
2010-01-06 21:34:05 +08:00
|
|
|
if (!dchan)
|
2010-03-27 07:44:01 +08:00
|
|
|
return -EINVAL;
|
2009-09-09 08:53:04 +08:00
|
|
|
|
2010-01-06 21:34:05 +08:00
|
|
|
chan = to_fsl_chan(dchan);
|
2009-09-09 08:53:04 +08:00
|
|
|
|
2014-11-17 21:42:24 +08:00
|
|
|
spin_lock_bh(&chan->desc_lock);
|
2009-09-09 08:53:04 +08:00
|
|
|
|
2014-11-17 21:42:24 +08:00
|
|
|
/* Halt the DMA engine */
|
|
|
|
dma_halt(chan);
|
2009-09-09 08:53:04 +08:00
|
|
|
|
2014-11-17 21:42:24 +08:00
|
|
|
/* Remove and free all of the descriptors in the LD queue */
|
|
|
|
fsldma_free_desc_list(chan, &chan->ld_pending);
|
|
|
|
fsldma_free_desc_list(chan, &chan->ld_running);
|
|
|
|
fsldma_free_desc_list(chan, &chan->ld_completed);
|
|
|
|
chan->idle = true;
|
2010-09-30 19:46:46 +08:00
|
|
|
|
2014-11-17 21:42:24 +08:00
|
|
|
spin_unlock_bh(&chan->desc_lock);
|
|
|
|
return 0;
|
|
|
|
}
|
2010-09-30 19:46:46 +08:00
|
|
|
|
2014-11-17 21:42:24 +08:00
|
|
|
static int fsl_dma_device_config(struct dma_chan *dchan,
|
|
|
|
struct dma_slave_config *config)
|
|
|
|
{
|
|
|
|
struct fsldma_chan *chan;
|
|
|
|
int size;
|
2010-09-30 19:46:46 +08:00
|
|
|
|
2014-11-17 21:42:24 +08:00
|
|
|
if (!dchan)
|
|
|
|
return -EINVAL;
|
2010-09-30 19:46:46 +08:00
|
|
|
|
2014-11-17 21:42:24 +08:00
|
|
|
chan = to_fsl_chan(dchan);
|
2010-09-30 19:46:46 +08:00
|
|
|
|
2014-11-17 21:42:24 +08:00
|
|
|
/* make sure the channel supports setting burst size */
|
|
|
|
if (!chan->set_request_count)
|
2010-09-30 19:46:46 +08:00
|
|
|
return -ENXIO;
|
2010-03-27 07:44:01 +08:00
|
|
|
|
2014-11-17 21:42:24 +08:00
|
|
|
/* we set the controller burst size depending on direction */
|
|
|
|
if (config->direction == DMA_MEM_TO_DEV)
|
|
|
|
size = config->dst_addr_width * config->dst_maxburst;
|
|
|
|
else
|
|
|
|
size = config->src_addr_width * config->src_maxburst;
|
|
|
|
|
|
|
|
chan->set_request_count(chan, size);
|
2010-03-27 07:44:01 +08:00
|
|
|
return 0;
|
2009-09-09 08:53:04 +08:00
|
|
|
}
|
|
|
|
|
2014-11-17 21:42:24 +08:00
|
|
|
|
2008-03-01 22:42:48 +08:00
|
|
|
/**
|
|
|
|
* fsl_dma_memcpy_issue_pending - Issue the DMA start command
|
2010-01-06 21:34:05 +08:00
|
|
|
* @chan : Freescale DMA channel
|
2008-03-01 22:42:48 +08:00
|
|
|
*/
|
2010-01-06 21:34:05 +08:00
|
|
|
static void fsl_dma_memcpy_issue_pending(struct dma_chan *dchan)
|
2008-03-01 22:42:48 +08:00
|
|
|
{
|
2010-01-06 21:34:05 +08:00
|
|
|
struct fsldma_chan *chan = to_fsl_chan(dchan);
|
2011-03-03 15:55:00 +08:00
|
|
|
|
2014-05-21 16:03:01 +08:00
|
|
|
spin_lock_bh(&chan->desc_lock);
|
2010-01-06 21:34:05 +08:00
|
|
|
fsl_chan_xfer_ld_queue(chan);
|
2014-05-21 16:03:01 +08:00
|
|
|
spin_unlock_bh(&chan->desc_lock);
|
2008-03-01 22:42:48 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
2010-03-27 07:50:49 +08:00
|
|
|
* fsl_tx_status - Determine the DMA status
|
2010-01-06 21:34:05 +08:00
|
|
|
* @chan : Freescale DMA channel
|
2008-03-01 22:42:48 +08:00
|
|
|
*/
|
2010-03-27 07:50:49 +08:00
|
|
|
static enum dma_status fsl_tx_status(struct dma_chan *dchan,
|
2008-03-01 22:42:48 +08:00
|
|
|
dma_cookie_t cookie,
|
2010-03-27 07:50:49 +08:00
|
|
|
struct dma_tx_state *txstate)
|
2008-03-01 22:42:48 +08:00
|
|
|
{
|
dmaengine: Freescale: change descriptor release process for supporting async_tx
Fix the potential risk when enable config NET_DMA and ASYNC_TX. Async_tx is
lack of support in current release process of dma descriptor, all descriptors
will be released whatever is acked or no-acked by async_tx, so there is a
potential race condition when dma engine is uesd by others clients (e.g. when
enable NET_DMA to offload TCP).
In our case, a race condition which is raised when use both of talitos and
dmaengine to offload xor is because napi scheduler will sync all pending
requests in dma channels, it affects the process of raid operations due to
ack_tx is not checked in fsl dma. The no-acked descriptor is freed which is
submitted just now, as a dependent tx, this freed descriptor trigger
BUG_ON(async_tx_test_ack(depend_tx)) in async_tx_submit().
TASK = ee1a94a0[1390] 'md0_raid5' THREAD: ecf40000 CPU: 0
GPR00: 00000001 ecf41ca0 ee44/921a94a0 0000003f 00000001 c00593e4 00000000 00000001
GPR08: 00000000 a7a7a7a7 00000001 045/920000002 42028042 100a38d4 ed576d98 00000000
GPR16: ed5a11b0 00000000 2b162000 00000200 046/920000000 2d555000 ed3015e8 c15a7aa0
GPR24: 00000000 c155fc40 00000000 ecb63220 ecf41d28 e47/92f640bb0 ef640c30 ecf41ca0
NIP [c02b048c] async_tx_submit+0x6c/0x2b4
LR [c02b068c] async_tx_submit+0x26c/0x2b4
Call Trace:
[ecf41ca0] [c02b068c] async_tx_submit+0x26c/0x2b448/92 (unreliable)
[ecf41cd0] [c02b0a4c] async_memcpy+0x240/0x25c
[ecf41d20] [c0421064] async_copy_data+0xa0/0x17c
[ecf41d70] [c0421cf4] __raid_run_ops+0x874/0xe10
[ecf41df0] [c0426ee4] handle_stripe+0x820/0x25e8
[ecf41e90] [c0429080] raid5d+0x3d4/0x5b4
[ecf41f40] [c04329b8] md_thread+0x138/0x16c
[ecf41f90] [c008277c] kthread+0x8c/0x90
[ecf41ff0] [c0011630] kernel_thread+0x4c/0x68
Another modification in this patch is the change of completed descriptors,
there is a potential risk which caused by exception interrupt, all descriptors
in ld_running list are seemed completed when an interrupt raised, it works fine
under normal condition, but if there is an exception occured, it cannot work as
our excepted. Hardware should not be depend on s/w list, the right way is to
read current descriptor address register to find the last completed descriptor.
If an interrupt is raised by an error, all descriptors in ld_running should not
be seemed finished, or these unfinished descriptors in ld_running will be
released wrongly.
A simple way to reproduce:
Enable dmatest first, then insert some bad descriptors which can trigger
Programming Error interrupts before the good descriptors. Last, the good
descriptors will be freed before they are processsed because of the exception
intrerrupt.
Note: the bad descriptors are only for simulating an exception interrupt. This
case can illustrate the potential risk in current fsl-dma very well.
Signed-off-by: Hongbo Zhang <hongbo.zhang@freescale.com>
Signed-off-by: Qiang Liu <qiang.liu@freescale.com>
Signed-off-by: Ira W. Snyder <iws@ovro.caltech.edu>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2014-05-21 16:03:03 +08:00
|
|
|
struct fsldma_chan *chan = to_fsl_chan(dchan);
|
|
|
|
enum dma_status ret;
|
|
|
|
|
|
|
|
ret = dma_cookie_status(dchan, cookie, txstate);
|
|
|
|
if (ret == DMA_COMPLETE)
|
|
|
|
return ret;
|
|
|
|
|
|
|
|
spin_lock_bh(&chan->desc_lock);
|
|
|
|
fsldma_cleanup_descriptors(chan);
|
|
|
|
spin_unlock_bh(&chan->desc_lock);
|
|
|
|
|
2013-05-27 20:14:35 +08:00
|
|
|
return dma_cookie_status(dchan, cookie, txstate);
|
2008-03-01 22:42:48 +08:00
|
|
|
}
|
|
|
|
|
2010-01-06 21:34:04 +08:00
|
|
|
/*----------------------------------------------------------------------------*/
|
|
|
|
/* Interrupt Handling */
|
|
|
|
/*----------------------------------------------------------------------------*/
|
|
|
|
|
2010-01-06 21:34:03 +08:00
|
|
|
static irqreturn_t fsldma_chan_irq(int irq, void *data)
|
2008-03-01 22:42:48 +08:00
|
|
|
{
|
2010-01-06 21:34:05 +08:00
|
|
|
struct fsldma_chan *chan = data;
|
|
|
|
u32 stat;
|
2008-03-01 22:42:48 +08:00
|
|
|
|
2010-01-06 21:34:06 +08:00
|
|
|
/* save and clear the status register */
|
2010-01-06 21:34:05 +08:00
|
|
|
stat = get_sr(chan);
|
2010-01-06 21:34:06 +08:00
|
|
|
set_sr(chan, stat);
|
2011-03-03 15:54:55 +08:00
|
|
|
chan_dbg(chan, "irq: stat = 0x%x\n", stat);
|
2008-03-01 22:42:48 +08:00
|
|
|
|
2011-03-03 15:54:58 +08:00
|
|
|
/* check that this was really our device */
|
2008-03-01 22:42:48 +08:00
|
|
|
stat &= ~(FSL_DMA_SR_CB | FSL_DMA_SR_CH);
|
|
|
|
if (!stat)
|
|
|
|
return IRQ_NONE;
|
|
|
|
|
|
|
|
if (stat & FSL_DMA_SR_TE)
|
2011-03-03 15:54:55 +08:00
|
|
|
chan_err(chan, "Transfer Error!\n");
|
2008-03-01 22:42:48 +08:00
|
|
|
|
2010-01-06 21:34:06 +08:00
|
|
|
/*
|
|
|
|
* Programming Error
|
2008-03-19 09:45:00 +08:00
|
|
|
* The DMA_INTERRUPT async_tx is a NULL transfer, which will
|
2012-08-04 22:37:53 +08:00
|
|
|
* trigger a PE interrupt.
|
2008-03-19 09:45:00 +08:00
|
|
|
*/
|
|
|
|
if (stat & FSL_DMA_SR_PE) {
|
2011-03-03 15:54:55 +08:00
|
|
|
chan_dbg(chan, "irq: Programming Error INT\n");
|
2008-03-19 09:45:00 +08:00
|
|
|
stat &= ~FSL_DMA_SR_PE;
|
2011-03-03 15:54:58 +08:00
|
|
|
if (get_bcr(chan) != 0)
|
|
|
|
chan_err(chan, "Programming Error!\n");
|
2008-04-18 11:17:25 +08:00
|
|
|
}
|
|
|
|
|
2010-01-06 21:34:06 +08:00
|
|
|
/*
|
|
|
|
* For MPC8349, EOCDI event need to update cookie
|
2008-04-18 11:17:25 +08:00
|
|
|
* and start the next transfer if it exist.
|
|
|
|
*/
|
|
|
|
if (stat & FSL_DMA_SR_EOCDI) {
|
2011-03-03 15:54:55 +08:00
|
|
|
chan_dbg(chan, "irq: End-of-Chain link INT\n");
|
2008-04-18 11:17:25 +08:00
|
|
|
stat &= ~FSL_DMA_SR_EOCDI;
|
2008-03-01 22:42:48 +08:00
|
|
|
}
|
|
|
|
|
2010-01-06 21:34:06 +08:00
|
|
|
/*
|
|
|
|
* If it current transfer is the end-of-transfer,
|
2008-03-01 22:42:48 +08:00
|
|
|
* we should clear the Channel Start bit for
|
|
|
|
* prepare next transfer.
|
|
|
|
*/
|
2008-04-18 11:17:25 +08:00
|
|
|
if (stat & FSL_DMA_SR_EOLNI) {
|
2011-03-03 15:54:55 +08:00
|
|
|
chan_dbg(chan, "irq: End-of-link INT\n");
|
2008-03-01 22:42:48 +08:00
|
|
|
stat &= ~FSL_DMA_SR_EOLNI;
|
|
|
|
}
|
|
|
|
|
2011-03-03 15:54:58 +08:00
|
|
|
/* check that the DMA controller is really idle */
|
|
|
|
if (!dma_is_idle(chan))
|
|
|
|
chan_err(chan, "irq: controller not idle!\n");
|
|
|
|
|
|
|
|
/* check that we handled all of the bits */
|
2008-03-01 22:42:48 +08:00
|
|
|
if (stat)
|
2011-03-03 15:54:58 +08:00
|
|
|
chan_err(chan, "irq: unhandled sr 0x%08x\n", stat);
|
2008-03-01 22:42:48 +08:00
|
|
|
|
2011-03-03 15:54:58 +08:00
|
|
|
/*
|
|
|
|
* Schedule the tasklet to handle all cleanup of the current
|
|
|
|
* transaction. It will start a new transaction if there is
|
|
|
|
* one pending.
|
|
|
|
*/
|
2010-01-06 21:34:05 +08:00
|
|
|
tasklet_schedule(&chan->tasklet);
|
2011-03-03 15:54:58 +08:00
|
|
|
chan_dbg(chan, "irq: Exit\n");
|
2008-03-01 22:42:48 +08:00
|
|
|
return IRQ_HANDLED;
|
|
|
|
}
|
|
|
|
|
2020-08-31 21:47:45 +08:00
|
|
|
static void dma_do_tasklet(struct tasklet_struct *t)
|
2010-01-06 21:34:04 +08:00
|
|
|
{
|
2020-08-31 21:47:45 +08:00
|
|
|
struct fsldma_chan *chan = from_tasklet(chan, t, tasklet);
|
2011-03-03 15:54:58 +08:00
|
|
|
|
|
|
|
chan_dbg(chan, "tasklet entry\n");
|
|
|
|
|
2018-08-17 21:00:26 +08:00
|
|
|
spin_lock(&chan->desc_lock);
|
2011-03-03 15:55:00 +08:00
|
|
|
|
|
|
|
/* the hardware is now idle and ready for more */
|
2011-03-03 15:54:58 +08:00
|
|
|
chan->idle = true;
|
|
|
|
|
dmaengine: Freescale: change descriptor release process for supporting async_tx
Fix the potential risk when enable config NET_DMA and ASYNC_TX. Async_tx is
lack of support in current release process of dma descriptor, all descriptors
will be released whatever is acked or no-acked by async_tx, so there is a
potential race condition when dma engine is uesd by others clients (e.g. when
enable NET_DMA to offload TCP).
In our case, a race condition which is raised when use both of talitos and
dmaengine to offload xor is because napi scheduler will sync all pending
requests in dma channels, it affects the process of raid operations due to
ack_tx is not checked in fsl dma. The no-acked descriptor is freed which is
submitted just now, as a dependent tx, this freed descriptor trigger
BUG_ON(async_tx_test_ack(depend_tx)) in async_tx_submit().
TASK = ee1a94a0[1390] 'md0_raid5' THREAD: ecf40000 CPU: 0
GPR00: 00000001 ecf41ca0 ee44/921a94a0 0000003f 00000001 c00593e4 00000000 00000001
GPR08: 00000000 a7a7a7a7 00000001 045/920000002 42028042 100a38d4 ed576d98 00000000
GPR16: ed5a11b0 00000000 2b162000 00000200 046/920000000 2d555000 ed3015e8 c15a7aa0
GPR24: 00000000 c155fc40 00000000 ecb63220 ecf41d28 e47/92f640bb0 ef640c30 ecf41ca0
NIP [c02b048c] async_tx_submit+0x6c/0x2b4
LR [c02b068c] async_tx_submit+0x26c/0x2b4
Call Trace:
[ecf41ca0] [c02b068c] async_tx_submit+0x26c/0x2b448/92 (unreliable)
[ecf41cd0] [c02b0a4c] async_memcpy+0x240/0x25c
[ecf41d20] [c0421064] async_copy_data+0xa0/0x17c
[ecf41d70] [c0421cf4] __raid_run_ops+0x874/0xe10
[ecf41df0] [c0426ee4] handle_stripe+0x820/0x25e8
[ecf41e90] [c0429080] raid5d+0x3d4/0x5b4
[ecf41f40] [c04329b8] md_thread+0x138/0x16c
[ecf41f90] [c008277c] kthread+0x8c/0x90
[ecf41ff0] [c0011630] kernel_thread+0x4c/0x68
Another modification in this patch is the change of completed descriptors,
there is a potential risk which caused by exception interrupt, all descriptors
in ld_running list are seemed completed when an interrupt raised, it works fine
under normal condition, but if there is an exception occured, it cannot work as
our excepted. Hardware should not be depend on s/w list, the right way is to
read current descriptor address register to find the last completed descriptor.
If an interrupt is raised by an error, all descriptors in ld_running should not
be seemed finished, or these unfinished descriptors in ld_running will be
released wrongly.
A simple way to reproduce:
Enable dmatest first, then insert some bad descriptors which can trigger
Programming Error interrupts before the good descriptors. Last, the good
descriptors will be freed before they are processsed because of the exception
intrerrupt.
Note: the bad descriptors are only for simulating an exception interrupt. This
case can illustrate the potential risk in current fsl-dma very well.
Signed-off-by: Hongbo Zhang <hongbo.zhang@freescale.com>
Signed-off-by: Qiang Liu <qiang.liu@freescale.com>
Signed-off-by: Ira W. Snyder <iws@ovro.caltech.edu>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2014-05-21 16:03:03 +08:00
|
|
|
/* Run all cleanup for descriptors which have been completed */
|
|
|
|
fsldma_cleanup_descriptors(chan);
|
2011-03-03 15:55:00 +08:00
|
|
|
|
2018-08-17 21:00:26 +08:00
|
|
|
spin_unlock(&chan->desc_lock);
|
2011-03-03 15:55:00 +08:00
|
|
|
|
2011-03-03 15:54:58 +08:00
|
|
|
chan_dbg(chan, "tasklet exit\n");
|
2010-01-06 21:34:04 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
static irqreturn_t fsldma_ctrl_irq(int irq, void *data)
|
2008-03-01 22:42:48 +08:00
|
|
|
{
|
2010-01-06 21:34:01 +08:00
|
|
|
struct fsldma_device *fdev = data;
|
2010-01-06 21:34:04 +08:00
|
|
|
struct fsldma_chan *chan;
|
|
|
|
unsigned int handled = 0;
|
|
|
|
u32 gsr, mask;
|
|
|
|
int i;
|
2008-03-01 22:42:48 +08:00
|
|
|
|
2010-01-06 21:34:03 +08:00
|
|
|
gsr = (fdev->feature & FSL_DMA_BIG_ENDIAN) ? in_be32(fdev->regs)
|
2010-01-06 21:34:04 +08:00
|
|
|
: in_le32(fdev->regs);
|
|
|
|
mask = 0xff000000;
|
|
|
|
dev_dbg(fdev->dev, "IRQ: gsr 0x%.8x\n", gsr);
|
2008-03-01 22:42:48 +08:00
|
|
|
|
2010-01-06 21:34:04 +08:00
|
|
|
for (i = 0; i < FSL_DMA_MAX_CHANS_PER_DEVICE; i++) {
|
|
|
|
chan = fdev->chan[i];
|
|
|
|
if (!chan)
|
|
|
|
continue;
|
|
|
|
|
|
|
|
if (gsr & mask) {
|
|
|
|
dev_dbg(fdev->dev, "IRQ: chan %d\n", chan->id);
|
|
|
|
fsldma_chan_irq(irq, chan);
|
|
|
|
handled++;
|
|
|
|
}
|
|
|
|
|
|
|
|
gsr &= ~mask;
|
|
|
|
mask >>= 8;
|
|
|
|
}
|
|
|
|
|
|
|
|
return IRQ_RETVAL(handled);
|
2008-03-01 22:42:48 +08:00
|
|
|
}
|
|
|
|
|
2010-01-06 21:34:04 +08:00
|
|
|
static void fsldma_free_irqs(struct fsldma_device *fdev)
|
2008-03-01 22:42:48 +08:00
|
|
|
{
|
2010-01-06 21:34:04 +08:00
|
|
|
struct fsldma_chan *chan;
|
|
|
|
int i;
|
|
|
|
|
2016-09-10 17:56:04 +08:00
|
|
|
if (fdev->irq) {
|
2010-01-06 21:34:04 +08:00
|
|
|
dev_dbg(fdev->dev, "free per-controller IRQ\n");
|
|
|
|
free_irq(fdev->irq, fdev);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
for (i = 0; i < FSL_DMA_MAX_CHANS_PER_DEVICE; i++) {
|
|
|
|
chan = fdev->chan[i];
|
2016-09-10 17:56:04 +08:00
|
|
|
if (chan && chan->irq) {
|
2011-03-03 15:54:55 +08:00
|
|
|
chan_dbg(chan, "free per-channel IRQ\n");
|
2010-01-06 21:34:04 +08:00
|
|
|
free_irq(chan->irq, chan);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static int fsldma_request_irqs(struct fsldma_device *fdev)
|
|
|
|
{
|
|
|
|
struct fsldma_chan *chan;
|
|
|
|
int ret;
|
|
|
|
int i;
|
|
|
|
|
|
|
|
/* if we have a per-controller IRQ, use that */
|
2016-09-10 17:56:04 +08:00
|
|
|
if (fdev->irq) {
|
2010-01-06 21:34:04 +08:00
|
|
|
dev_dbg(fdev->dev, "request per-controller IRQ\n");
|
|
|
|
ret = request_irq(fdev->irq, fsldma_ctrl_irq, IRQF_SHARED,
|
|
|
|
"fsldma-controller", fdev);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* no per-controller IRQ, use the per-channel IRQs */
|
|
|
|
for (i = 0; i < FSL_DMA_MAX_CHANS_PER_DEVICE; i++) {
|
|
|
|
chan = fdev->chan[i];
|
|
|
|
if (!chan)
|
|
|
|
continue;
|
|
|
|
|
2016-09-10 17:56:04 +08:00
|
|
|
if (!chan->irq) {
|
2011-03-03 15:54:55 +08:00
|
|
|
chan_err(chan, "interrupts property missing in device tree\n");
|
2010-01-06 21:34:04 +08:00
|
|
|
ret = -ENODEV;
|
|
|
|
goto out_unwind;
|
|
|
|
}
|
|
|
|
|
2011-03-03 15:54:55 +08:00
|
|
|
chan_dbg(chan, "request per-channel IRQ\n");
|
2010-01-06 21:34:04 +08:00
|
|
|
ret = request_irq(chan->irq, fsldma_chan_irq, IRQF_SHARED,
|
|
|
|
"fsldma-chan", chan);
|
|
|
|
if (ret) {
|
2011-03-03 15:54:55 +08:00
|
|
|
chan_err(chan, "unable to request per-channel IRQ\n");
|
2010-01-06 21:34:04 +08:00
|
|
|
goto out_unwind;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
out_unwind:
|
|
|
|
for (/* none */; i >= 0; i--) {
|
|
|
|
chan = fdev->chan[i];
|
|
|
|
if (!chan)
|
|
|
|
continue;
|
|
|
|
|
2016-09-10 17:56:04 +08:00
|
|
|
if (!chan->irq)
|
2010-01-06 21:34:04 +08:00
|
|
|
continue;
|
|
|
|
|
|
|
|
free_irq(chan->irq, chan);
|
|
|
|
}
|
|
|
|
|
|
|
|
return ret;
|
2008-03-01 22:42:48 +08:00
|
|
|
}
|
|
|
|
|
2010-01-06 21:34:01 +08:00
|
|
|
/*----------------------------------------------------------------------------*/
|
|
|
|
/* OpenFirmware Subsystem */
|
|
|
|
/*----------------------------------------------------------------------------*/
|
|
|
|
|
2012-11-20 02:22:55 +08:00
|
|
|
static int fsl_dma_chan_probe(struct fsldma_device *fdev,
|
2008-09-27 08:00:11 +08:00
|
|
|
struct device_node *node, u32 feature, const char *compatible)
|
2008-03-01 22:42:48 +08:00
|
|
|
{
|
2010-01-06 21:34:05 +08:00
|
|
|
struct fsldma_chan *chan;
|
2010-01-06 21:34:00 +08:00
|
|
|
struct resource res;
|
2008-03-01 22:42:48 +08:00
|
|
|
int err;
|
|
|
|
|
|
|
|
/* alloc channel */
|
2010-01-06 21:34:05 +08:00
|
|
|
chan = kzalloc(sizeof(*chan), GFP_KERNEL);
|
|
|
|
if (!chan) {
|
2010-01-06 21:34:03 +08:00
|
|
|
err = -ENOMEM;
|
|
|
|
goto out_return;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* ioremap registers for use */
|
2010-01-06 21:34:05 +08:00
|
|
|
chan->regs = of_iomap(node, 0);
|
|
|
|
if (!chan->regs) {
|
2010-01-06 21:34:03 +08:00
|
|
|
dev_err(fdev->dev, "unable to ioremap registers\n");
|
|
|
|
err = -ENOMEM;
|
2010-01-06 21:34:05 +08:00
|
|
|
goto out_free_chan;
|
2008-03-01 22:42:48 +08:00
|
|
|
}
|
|
|
|
|
2010-01-06 21:34:00 +08:00
|
|
|
err = of_address_to_resource(node, 0, &res);
|
2008-03-01 22:42:48 +08:00
|
|
|
if (err) {
|
2010-01-06 21:34:03 +08:00
|
|
|
dev_err(fdev->dev, "unable to find 'reg' property\n");
|
|
|
|
goto out_iounmap_regs;
|
2008-03-01 22:42:48 +08:00
|
|
|
}
|
|
|
|
|
2010-01-06 21:34:05 +08:00
|
|
|
chan->feature = feature;
|
2008-03-01 22:42:48 +08:00
|
|
|
if (!fdev->feature)
|
2010-01-06 21:34:05 +08:00
|
|
|
fdev->feature = chan->feature;
|
2008-03-01 22:42:48 +08:00
|
|
|
|
2010-01-06 21:34:03 +08:00
|
|
|
/*
|
|
|
|
* If the DMA device's feature is different than the feature
|
|
|
|
* of its channels, report the bug
|
2008-03-01 22:42:48 +08:00
|
|
|
*/
|
2010-01-06 21:34:05 +08:00
|
|
|
WARN_ON(fdev->feature != chan->feature);
|
2010-01-06 21:34:03 +08:00
|
|
|
|
2010-01-06 21:34:05 +08:00
|
|
|
chan->dev = fdev->dev;
|
2013-09-26 17:33:43 +08:00
|
|
|
chan->id = (res.start & 0xfff) < 0x300 ?
|
|
|
|
((res.start - 0x100) & 0xfff) >> 7 :
|
|
|
|
((res.start - 0x200) & 0xfff) >> 7;
|
2010-01-06 21:34:05 +08:00
|
|
|
if (chan->id >= FSL_DMA_MAX_CHANS_PER_DEVICE) {
|
2010-01-06 21:34:03 +08:00
|
|
|
dev_err(fdev->dev, "too many channels for device\n");
|
2008-03-01 22:42:48 +08:00
|
|
|
err = -EINVAL;
|
2010-01-06 21:34:03 +08:00
|
|
|
goto out_iounmap_regs;
|
2008-03-01 22:42:48 +08:00
|
|
|
}
|
|
|
|
|
2010-01-06 21:34:05 +08:00
|
|
|
fdev->chan[chan->id] = chan;
|
2020-08-31 21:47:45 +08:00
|
|
|
tasklet_setup(&chan->tasklet, dma_do_tasklet);
|
2011-03-03 15:54:55 +08:00
|
|
|
snprintf(chan->name, sizeof(chan->name), "chan%d", chan->id);
|
2010-01-06 21:34:03 +08:00
|
|
|
|
|
|
|
/* Initialize the channel */
|
2010-01-06 21:34:05 +08:00
|
|
|
dma_init(chan);
|
2008-03-01 22:42:48 +08:00
|
|
|
|
|
|
|
/* Clear cdar registers */
|
2010-01-06 21:34:05 +08:00
|
|
|
set_cdar(chan, 0);
|
2008-03-01 22:42:48 +08:00
|
|
|
|
2010-01-06 21:34:05 +08:00
|
|
|
switch (chan->feature & FSL_DMA_IP_MASK) {
|
2008-03-01 22:42:48 +08:00
|
|
|
case FSL_DMA_IP_85XX:
|
2010-01-06 21:34:05 +08:00
|
|
|
chan->toggle_ext_pause = fsl_chan_toggle_ext_pause;
|
2020-08-24 06:36:59 +08:00
|
|
|
fallthrough;
|
2008-03-01 22:42:48 +08:00
|
|
|
case FSL_DMA_IP_83XX:
|
2010-01-06 21:34:05 +08:00
|
|
|
chan->toggle_ext_start = fsl_chan_toggle_ext_start;
|
|
|
|
chan->set_src_loop_size = fsl_chan_set_src_loop_size;
|
|
|
|
chan->set_dst_loop_size = fsl_chan_set_dst_loop_size;
|
|
|
|
chan->set_request_count = fsl_chan_set_request_count;
|
2008-03-01 22:42:48 +08:00
|
|
|
}
|
|
|
|
|
2010-01-06 21:34:05 +08:00
|
|
|
spin_lock_init(&chan->desc_lock);
|
2010-01-06 21:34:06 +08:00
|
|
|
INIT_LIST_HEAD(&chan->ld_pending);
|
|
|
|
INIT_LIST_HEAD(&chan->ld_running);
|
dmaengine: Freescale: change descriptor release process for supporting async_tx
Fix the potential risk when enable config NET_DMA and ASYNC_TX. Async_tx is
lack of support in current release process of dma descriptor, all descriptors
will be released whatever is acked or no-acked by async_tx, so there is a
potential race condition when dma engine is uesd by others clients (e.g. when
enable NET_DMA to offload TCP).
In our case, a race condition which is raised when use both of talitos and
dmaengine to offload xor is because napi scheduler will sync all pending
requests in dma channels, it affects the process of raid operations due to
ack_tx is not checked in fsl dma. The no-acked descriptor is freed which is
submitted just now, as a dependent tx, this freed descriptor trigger
BUG_ON(async_tx_test_ack(depend_tx)) in async_tx_submit().
TASK = ee1a94a0[1390] 'md0_raid5' THREAD: ecf40000 CPU: 0
GPR00: 00000001 ecf41ca0 ee44/921a94a0 0000003f 00000001 c00593e4 00000000 00000001
GPR08: 00000000 a7a7a7a7 00000001 045/920000002 42028042 100a38d4 ed576d98 00000000
GPR16: ed5a11b0 00000000 2b162000 00000200 046/920000000 2d555000 ed3015e8 c15a7aa0
GPR24: 00000000 c155fc40 00000000 ecb63220 ecf41d28 e47/92f640bb0 ef640c30 ecf41ca0
NIP [c02b048c] async_tx_submit+0x6c/0x2b4
LR [c02b068c] async_tx_submit+0x26c/0x2b4
Call Trace:
[ecf41ca0] [c02b068c] async_tx_submit+0x26c/0x2b448/92 (unreliable)
[ecf41cd0] [c02b0a4c] async_memcpy+0x240/0x25c
[ecf41d20] [c0421064] async_copy_data+0xa0/0x17c
[ecf41d70] [c0421cf4] __raid_run_ops+0x874/0xe10
[ecf41df0] [c0426ee4] handle_stripe+0x820/0x25e8
[ecf41e90] [c0429080] raid5d+0x3d4/0x5b4
[ecf41f40] [c04329b8] md_thread+0x138/0x16c
[ecf41f90] [c008277c] kthread+0x8c/0x90
[ecf41ff0] [c0011630] kernel_thread+0x4c/0x68
Another modification in this patch is the change of completed descriptors,
there is a potential risk which caused by exception interrupt, all descriptors
in ld_running list are seemed completed when an interrupt raised, it works fine
under normal condition, but if there is an exception occured, it cannot work as
our excepted. Hardware should not be depend on s/w list, the right way is to
read current descriptor address register to find the last completed descriptor.
If an interrupt is raised by an error, all descriptors in ld_running should not
be seemed finished, or these unfinished descriptors in ld_running will be
released wrongly.
A simple way to reproduce:
Enable dmatest first, then insert some bad descriptors which can trigger
Programming Error interrupts before the good descriptors. Last, the good
descriptors will be freed before they are processsed because of the exception
intrerrupt.
Note: the bad descriptors are only for simulating an exception interrupt. This
case can illustrate the potential risk in current fsl-dma very well.
Signed-off-by: Hongbo Zhang <hongbo.zhang@freescale.com>
Signed-off-by: Qiang Liu <qiang.liu@freescale.com>
Signed-off-by: Ira W. Snyder <iws@ovro.caltech.edu>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
2014-05-21 16:03:03 +08:00
|
|
|
INIT_LIST_HEAD(&chan->ld_completed);
|
2011-03-03 15:54:58 +08:00
|
|
|
chan->idle = true;
|
2014-05-21 16:03:02 +08:00
|
|
|
#ifdef CONFIG_PM
|
|
|
|
chan->pm_state = RUNNING;
|
|
|
|
#endif
|
2008-03-01 22:42:48 +08:00
|
|
|
|
2010-01-06 21:34:05 +08:00
|
|
|
chan->common.device = &fdev->common;
|
2012-03-07 06:36:27 +08:00
|
|
|
dma_cookie_init(&chan->common);
|
2008-03-01 22:42:48 +08:00
|
|
|
|
2010-01-06 21:34:04 +08:00
|
|
|
/* find the IRQ line, if it exists in the device tree */
|
2010-01-06 21:34:05 +08:00
|
|
|
chan->irq = irq_of_parse_and_map(node, 0);
|
2010-01-06 21:34:04 +08:00
|
|
|
|
2008-03-01 22:42:48 +08:00
|
|
|
/* Add the channel to DMA device channel list */
|
2010-01-06 21:34:05 +08:00
|
|
|
list_add_tail(&chan->common.device_node, &fdev->common.channels);
|
2008-03-01 22:42:48 +08:00
|
|
|
|
2010-01-06 21:34:05 +08:00
|
|
|
dev_info(fdev->dev, "#%d (%s), irq %d\n", chan->id, compatible,
|
2016-09-10 17:56:04 +08:00
|
|
|
chan->irq ? chan->irq : fdev->irq);
|
2008-03-01 22:42:48 +08:00
|
|
|
|
|
|
|
return 0;
|
2008-05-30 14:25:45 +08:00
|
|
|
|
2010-01-06 21:34:03 +08:00
|
|
|
out_iounmap_regs:
|
2010-01-06 21:34:05 +08:00
|
|
|
iounmap(chan->regs);
|
|
|
|
out_free_chan:
|
|
|
|
kfree(chan);
|
2010-01-06 21:34:03 +08:00
|
|
|
out_return:
|
2008-03-01 22:42:48 +08:00
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
2010-01-06 21:34:05 +08:00
|
|
|
static void fsl_dma_chan_remove(struct fsldma_chan *chan)
|
2008-03-01 22:42:48 +08:00
|
|
|
{
|
2010-01-06 21:34:05 +08:00
|
|
|
irq_dispose_mapping(chan->irq);
|
|
|
|
list_del(&chan->common.device_node);
|
|
|
|
iounmap(chan->regs);
|
|
|
|
kfree(chan);
|
2008-03-01 22:42:48 +08:00
|
|
|
}
|
|
|
|
|
2012-11-20 02:22:55 +08:00
|
|
|
static int fsldma_of_probe(struct platform_device *op)
|
2008-03-01 22:42:48 +08:00
|
|
|
{
|
2010-01-06 21:34:01 +08:00
|
|
|
struct fsldma_device *fdev;
|
2008-09-27 08:00:11 +08:00
|
|
|
struct device_node *child;
|
2010-01-06 21:34:03 +08:00
|
|
|
int err;
|
2008-03-01 22:42:48 +08:00
|
|
|
|
2010-01-06 21:34:01 +08:00
|
|
|
fdev = kzalloc(sizeof(*fdev), GFP_KERNEL);
|
2008-03-01 22:42:48 +08:00
|
|
|
if (!fdev) {
|
2010-01-06 21:34:03 +08:00
|
|
|
err = -ENOMEM;
|
|
|
|
goto out_return;
|
2008-03-01 22:42:48 +08:00
|
|
|
}
|
2010-01-06 21:34:03 +08:00
|
|
|
|
|
|
|
fdev->dev = &op->dev;
|
2008-03-01 22:42:48 +08:00
|
|
|
INIT_LIST_HEAD(&fdev->common.channels);
|
|
|
|
|
2010-01-06 21:34:03 +08:00
|
|
|
/* ioremap the registers for use */
|
2010-04-14 07:12:29 +08:00
|
|
|
fdev->regs = of_iomap(op->dev.of_node, 0);
|
2010-01-06 21:34:03 +08:00
|
|
|
if (!fdev->regs) {
|
|
|
|
dev_err(&op->dev, "unable to ioremap registers\n");
|
|
|
|
err = -ENOMEM;
|
2016-09-28 18:45:11 +08:00
|
|
|
goto out_free;
|
2008-03-01 22:42:48 +08:00
|
|
|
}
|
|
|
|
|
2010-01-06 21:34:04 +08:00
|
|
|
/* map the channel IRQ if it exists, but don't hookup the handler yet */
|
2010-04-14 07:12:29 +08:00
|
|
|
fdev->irq = irq_of_parse_and_map(op->dev.of_node, 0);
|
2010-01-06 21:34:04 +08:00
|
|
|
|
2008-03-01 22:42:48 +08:00
|
|
|
dma_cap_set(DMA_MEMCPY, fdev->common.cap_mask);
|
2009-09-09 08:53:04 +08:00
|
|
|
dma_cap_set(DMA_SLAVE, fdev->common.cap_mask);
|
2008-03-01 22:42:48 +08:00
|
|
|
fdev->common.device_alloc_chan_resources = fsl_dma_alloc_chan_resources;
|
|
|
|
fdev->common.device_free_chan_resources = fsl_dma_free_chan_resources;
|
|
|
|
fdev->common.device_prep_dma_memcpy = fsl_dma_prep_memcpy;
|
2010-03-27 07:50:49 +08:00
|
|
|
fdev->common.device_tx_status = fsl_tx_status;
|
2008-03-01 22:42:48 +08:00
|
|
|
fdev->common.device_issue_pending = fsl_dma_memcpy_issue_pending;
|
2014-11-17 21:42:24 +08:00
|
|
|
fdev->common.device_config = fsl_dma_device_config;
|
|
|
|
fdev->common.device_terminate_all = fsl_dma_device_terminate_all;
|
2010-01-06 21:34:03 +08:00
|
|
|
fdev->common.dev = &op->dev;
|
2008-03-01 22:42:48 +08:00
|
|
|
|
2015-01-08 18:38:16 +08:00
|
|
|
fdev->common.src_addr_widths = FSL_DMA_BUSWIDTHS;
|
|
|
|
fdev->common.dst_addr_widths = FSL_DMA_BUSWIDTHS;
|
|
|
|
fdev->common.directions = BIT(DMA_DEV_TO_MEM) | BIT(DMA_MEM_TO_DEV);
|
|
|
|
fdev->common.residue_granularity = DMA_RESIDUE_GRANULARITY_DESCRIPTOR;
|
|
|
|
|
2010-11-11 20:16:29 +08:00
|
|
|
dma_set_mask(&(op->dev), DMA_BIT_MASK(36));
|
|
|
|
|
2013-05-24 09:10:13 +08:00
|
|
|
platform_set_drvdata(op, fdev);
|
2008-09-27 08:00:11 +08:00
|
|
|
|
2010-01-06 21:34:03 +08:00
|
|
|
/*
|
|
|
|
* We cannot use of_platform_bus_probe() because there is no
|
|
|
|
* of_platform_bus_remove(). Instead, we manually instantiate every DMA
|
2008-09-27 08:00:11 +08:00
|
|
|
* channel object.
|
|
|
|
*/
|
2010-04-14 07:12:29 +08:00
|
|
|
for_each_child_of_node(op->dev.of_node, child) {
|
2010-01-06 21:34:03 +08:00
|
|
|
if (of_device_is_compatible(child, "fsl,eloplus-dma-channel")) {
|
2008-09-27 08:00:11 +08:00
|
|
|
fsl_dma_chan_probe(fdev, child,
|
|
|
|
FSL_DMA_IP_85XX | FSL_DMA_BIG_ENDIAN,
|
|
|
|
"fsl,eloplus-dma-channel");
|
2010-01-06 21:34:03 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
if (of_device_is_compatible(child, "fsl,elo-dma-channel")) {
|
2008-09-27 08:00:11 +08:00
|
|
|
fsl_dma_chan_probe(fdev, child,
|
|
|
|
FSL_DMA_IP_83XX | FSL_DMA_LITTLE_ENDIAN,
|
|
|
|
"fsl,elo-dma-channel");
|
2010-01-06 21:34:03 +08:00
|
|
|
}
|
2008-09-27 08:00:11 +08:00
|
|
|
}
|
2008-03-01 22:42:48 +08:00
|
|
|
|
2010-01-06 21:34:04 +08:00
|
|
|
/*
|
|
|
|
* Hookup the IRQ handler(s)
|
|
|
|
*
|
|
|
|
* If we have a per-controller interrupt, we prefer that to the
|
|
|
|
* per-channel interrupts to reduce the number of shared interrupt
|
|
|
|
* handlers on the same IRQ line
|
|
|
|
*/
|
|
|
|
err = fsldma_request_irqs(fdev);
|
|
|
|
if (err) {
|
|
|
|
dev_err(fdev->dev, "unable to request IRQs\n");
|
|
|
|
goto out_free_fdev;
|
|
|
|
}
|
|
|
|
|
2008-03-01 22:42:48 +08:00
|
|
|
dma_async_device_register(&fdev->common);
|
|
|
|
return 0;
|
|
|
|
|
2010-01-06 21:34:03 +08:00
|
|
|
out_free_fdev:
|
2010-01-06 21:34:04 +08:00
|
|
|
irq_dispose_mapping(fdev->irq);
|
2016-09-28 18:45:11 +08:00
|
|
|
iounmap(fdev->regs);
|
|
|
|
out_free:
|
2008-03-01 22:42:48 +08:00
|
|
|
kfree(fdev);
|
2010-01-06 21:34:03 +08:00
|
|
|
out_return:
|
2008-03-01 22:42:48 +08:00
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
2010-08-06 23:25:50 +08:00
|
|
|
static int fsldma_of_remove(struct platform_device *op)
|
2008-09-27 08:00:11 +08:00
|
|
|
{
|
2010-01-06 21:34:01 +08:00
|
|
|
struct fsldma_device *fdev;
|
2008-09-27 08:00:11 +08:00
|
|
|
unsigned int i;
|
|
|
|
|
2013-05-24 09:10:13 +08:00
|
|
|
fdev = platform_get_drvdata(op);
|
2008-09-27 08:00:11 +08:00
|
|
|
dma_async_device_unregister(&fdev->common);
|
|
|
|
|
2010-01-06 21:34:04 +08:00
|
|
|
fsldma_free_irqs(fdev);
|
|
|
|
|
2010-01-06 21:34:03 +08:00
|
|
|
for (i = 0; i < FSL_DMA_MAX_CHANS_PER_DEVICE; i++) {
|
2008-09-27 08:00:11 +08:00
|
|
|
if (fdev->chan[i])
|
|
|
|
fsl_dma_chan_remove(fdev->chan[i]);
|
2010-01-06 21:34:03 +08:00
|
|
|
}
|
2008-09-27 08:00:11 +08:00
|
|
|
|
2010-01-06 21:34:03 +08:00
|
|
|
iounmap(fdev->regs);
|
2008-09-27 08:00:11 +08:00
|
|
|
kfree(fdev);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2014-05-21 16:03:02 +08:00
|
|
|
#ifdef CONFIG_PM
|
|
|
|
static int fsldma_suspend_late(struct device *dev)
|
|
|
|
{
|
2018-04-22 17:14:11 +08:00
|
|
|
struct fsldma_device *fdev = dev_get_drvdata(dev);
|
2014-05-21 16:03:02 +08:00
|
|
|
struct fsldma_chan *chan;
|
|
|
|
int i;
|
|
|
|
|
|
|
|
for (i = 0; i < FSL_DMA_MAX_CHANS_PER_DEVICE; i++) {
|
|
|
|
chan = fdev->chan[i];
|
|
|
|
if (!chan)
|
|
|
|
continue;
|
|
|
|
|
|
|
|
spin_lock_bh(&chan->desc_lock);
|
|
|
|
if (unlikely(!chan->idle))
|
|
|
|
goto out;
|
|
|
|
chan->regs_save.mr = get_mr(chan);
|
|
|
|
chan->pm_state = SUSPENDED;
|
|
|
|
spin_unlock_bh(&chan->desc_lock);
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
out:
|
|
|
|
for (; i >= 0; i--) {
|
|
|
|
chan = fdev->chan[i];
|
|
|
|
if (!chan)
|
|
|
|
continue;
|
|
|
|
chan->pm_state = RUNNING;
|
|
|
|
spin_unlock_bh(&chan->desc_lock);
|
|
|
|
}
|
|
|
|
return -EBUSY;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int fsldma_resume_early(struct device *dev)
|
|
|
|
{
|
2018-04-22 17:14:11 +08:00
|
|
|
struct fsldma_device *fdev = dev_get_drvdata(dev);
|
2014-05-21 16:03:02 +08:00
|
|
|
struct fsldma_chan *chan;
|
|
|
|
u32 mode;
|
|
|
|
int i;
|
|
|
|
|
|
|
|
for (i = 0; i < FSL_DMA_MAX_CHANS_PER_DEVICE; i++) {
|
|
|
|
chan = fdev->chan[i];
|
|
|
|
if (!chan)
|
|
|
|
continue;
|
|
|
|
|
|
|
|
spin_lock_bh(&chan->desc_lock);
|
|
|
|
mode = chan->regs_save.mr
|
|
|
|
& ~FSL_DMA_MR_CS & ~FSL_DMA_MR_CC & ~FSL_DMA_MR_CA;
|
|
|
|
set_mr(chan, mode);
|
|
|
|
chan->pm_state = RUNNING;
|
|
|
|
spin_unlock_bh(&chan->desc_lock);
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static const struct dev_pm_ops fsldma_pm_ops = {
|
|
|
|
.suspend_late = fsldma_suspend_late,
|
|
|
|
.resume_early = fsldma_resume_early,
|
|
|
|
};
|
|
|
|
#endif
|
|
|
|
|
2010-02-03 14:41:06 +08:00
|
|
|
static const struct of_device_id fsldma_of_ids[] = {
|
2013-09-26 17:33:43 +08:00
|
|
|
{ .compatible = "fsl,elo3-dma", },
|
2008-04-01 00:13:21 +08:00
|
|
|
{ .compatible = "fsl,eloplus-dma", },
|
|
|
|
{ .compatible = "fsl,elo-dma", },
|
2008-03-01 22:42:48 +08:00
|
|
|
{}
|
|
|
|
};
|
2015-09-17 04:57:17 +08:00
|
|
|
MODULE_DEVICE_TABLE(of, fsldma_of_ids);
|
2008-03-01 22:42:48 +08:00
|
|
|
|
2011-04-08 01:33:03 +08:00
|
|
|
static struct platform_driver fsldma_of_driver = {
|
2010-04-14 07:13:02 +08:00
|
|
|
.driver = {
|
|
|
|
.name = "fsl-elo-dma",
|
|
|
|
.of_match_table = fsldma_of_ids,
|
2014-05-21 16:03:02 +08:00
|
|
|
#ifdef CONFIG_PM
|
|
|
|
.pm = &fsldma_pm_ops,
|
|
|
|
#endif
|
2010-04-14 07:13:02 +08:00
|
|
|
},
|
|
|
|
.probe = fsldma_of_probe,
|
|
|
|
.remove = fsldma_of_remove,
|
2008-03-01 22:42:48 +08:00
|
|
|
};
|
|
|
|
|
2010-01-06 21:34:01 +08:00
|
|
|
/*----------------------------------------------------------------------------*/
|
|
|
|
/* Module Init / Exit */
|
|
|
|
/*----------------------------------------------------------------------------*/
|
|
|
|
|
|
|
|
static __init int fsldma_init(void)
|
2008-03-01 22:42:48 +08:00
|
|
|
{
|
2013-09-26 17:33:43 +08:00
|
|
|
pr_info("Freescale Elo series DMA driver\n");
|
2011-02-23 10:59:54 +08:00
|
|
|
return platform_driver_register(&fsldma_of_driver);
|
2008-09-27 08:00:11 +08:00
|
|
|
}
|
|
|
|
|
2010-01-06 21:34:01 +08:00
|
|
|
static void __exit fsldma_exit(void)
|
2008-09-27 08:00:11 +08:00
|
|
|
{
|
2011-02-23 10:59:54 +08:00
|
|
|
platform_driver_unregister(&fsldma_of_driver);
|
2008-03-01 22:42:48 +08:00
|
|
|
}
|
|
|
|
|
2010-01-06 21:34:01 +08:00
|
|
|
subsys_initcall(fsldma_init);
|
|
|
|
module_exit(fsldma_exit);
|
2008-09-27 08:00:11 +08:00
|
|
|
|
2013-09-26 17:33:43 +08:00
|
|
|
MODULE_DESCRIPTION("Freescale Elo series DMA driver");
|
2008-09-27 08:00:11 +08:00
|
|
|
MODULE_LICENSE("GPL");
|