linux-toradex.git/drivers/dma/k3dma.c, branch v4.16-rc4

dmaengine: k3dma: Use vchan_terminate_vdesc() instead of desc_free

2017-12-04T17:03:51+00:00

To avoid race with vchan_complete, use the race free way to terminate
running transfer.

Implement the device_synchronize callback to make sure that the terminated
descriptor is freed.

Signed-off-by: Peter Ujfalusi 
Acked-by: Zhangfei Gao 
Signed-off-by: Vinod Koul

dmaengine: k3dma: remove useless ON_WARN_ONCE()

2017-08-25T06:45:40+00:00

Commit 36387a2b1f62b5c087c5fe6f0f7b23b94f722ad7 ("k3dma: Fix
memory handling in preparation for cyclic mode") adds few
calls to ON_WARN_ONCE() to track the use of ds_run/ds_done.

After the two fixes:
- dmaengine: k3dma: fix non-cyclic mode
- dmaengine: k3dma: fix re-free of the same descriptor
the behaviour of ds_run/ds_done is properly fixed.
The remaining ON_WARN_ONCE() are never triggered and can be
removed.

Signed-off-by: Antonio Borneo 
Signed-off-by: Vinod Koul

dmaengine: k3dma: fix double free of descriptor

2017-08-25T06:45:40+00:00

Commit 36387a2b1f62b5c087c5fe6f0f7b23b94f722ad7 ("k3dma: Fix
memory handling in preparation for cyclic mode") adds code
to free the descriptor in ds_done.

In cyclic mode, ds_done is never used and it's always NULL,
so the added code is not executed.

In non-cyclic mode, ds_done is used as a flag: when not NULL
it signals that the descriptor has been consumed. No need to
free it because it would be free by vchan_complete().

The fix takes back the code changed by the commit above:
- remove the free on the descriptor;
- initialize ds_done to NULL for the next run.

Signed-off-by: Antonio Borneo 
Signed-off-by: Vinod Koul

dmaengine: k3dma: fix non-cyclic mode

2017-08-25T06:45:40+00:00

Commit 36387a2b1f62b5c087c5fe6f0f7b23b94f722ad7 ("k3dma: Fix
memory handling in preparation for cyclic mode") broke the
logic around ds_run/ds_done in case of non-cyclic DMA.

This went unnoticed as the only user of k3dma was the i2s
audio driver but, with a patch set to enable dma on SPI, the
issue popped out.

The fix re-applies the initialization to ds_run/ds_done in
k3_dma_start_txd() that were removed by the commit above.

Also, one of the calls to k3_dma_start_txd() is triggered by
(ds_done != NULL), so remove the noisy and useless call to
WARN_ON_ONCE().

Signed-off-by: Antonio Borneo 
Signed-off-by: Vinod Koul

dmaengine: k3dma: move to dma_pool_zalloc

2016-12-08T15:54:21+00:00

Replace dma_pool_alloc & memset with dma_pool_zalloc.

Acked-by: Zhangfei Gao 
Signed-off-by: Vinod Koul

dmaengine: k3dma: use correct format string for debug output

2016-09-07T08:08:43+00:00

The newly added k3_dma_prep_dma_cyclic function has some debug output
that uses incorrect typecasts, some of which cause a warning like:

drivers/dma/k3dma.c: In function 'k3_dma_prep_dma_cyclic':
drivers/dma/k3dma.c:589:671: error: cast to pointer from integer of different size [-Werror=int-to-pointer-cast]

In general, we have to print 'dma_addr_t' values using special
'%pad' format to get the correct behavior on kernels that have
a 64-bit dma_addr_t type but 32-bit pointers.

Similarly, printing size_t values should be done using the %z
modifier to get the correct behavior on 64-bit kernels.

Signed-off-by: Arnd Bergmann 
Fixes: a7e08fa6cc78 ("k3dma: Add cyclic mode for audio")
Signed-off-by: Vinod Koul

k3dma: Add cyclic mode for audio

2016-08-31T04:39:32+00:00

Currently the k3dma driver doesn't offer the cyclic mode
necessary for handling audio.

This patch adds it.

Cc: Zhangfei Gao 
Cc: Krzysztof Kozlowski 
Cc: Maxime Ripard 
Cc: Vinod Koul 
Cc: Dan Williams 
Cc: Mark Brown 
Cc: Andy Green 
Acked-by: Zhangfei Gao 
Signed-off-by: Andy Green 
[jstultz: Forward ported to mainline, removed a few
 bits of logic that didn't seem to have much effect]
Signed-off-by: John Stultz 
Signed-off-by: Vinod Koul

k3dma: Fix memory handling in preparation for cyclic mode

2016-08-31T04:39:32+00:00

With cyclic mode, the shared virt-dma logic doesn't actually
manage the descriptor state, nor the calling of the descriptor
free callback. This results in leaking a desc structure every
time we start an audio transfer.

Thus we must manage it ourselves. The k3dma driver already keeps
track of the active and finished descriptors via ds_run and ds_done
pointers, so cleanup how we handle those two values, so when we
tear down everything in terminate_all, call free_desc on the ds_run
and ds_done pointers if they are not null.

NOTE: HiKey doesn't use the non-cyclic dma modes, so I'm not been
able to test those modes. But with this patch we no longer leak
the desc structures.

Cc: Zhangfei Gao 
Cc: Krzysztof Kozlowski 
Cc: Maxime Ripard 
Cc: Vinod Koul 
Cc: Dan Williams 
Cc: Mark Brown 
Cc: Andy Green 
Acked-by: Zhangfei Gao 
Signed-off-by: John Stultz 
Signed-off-by: Vinod Koul

k3dma: Fix occasional DMA ERR issue by using proper dma api

2016-08-31T04:39:32+00:00

After lots of debugging on an occasional DMA ERR issue, I realized
that the desc structures which we point the dma hardware are being
allocated out of regular memory. This means when we fill the desc
structures, that data doesn't always get flushed out to memory by
the time we start the dma transfer, resulting in the dma engine getting
some null values, resulting in a DMA ERR on the first irq.

Thus, this patch adopts mechanism similar to the zx296702_dma of
allocating the desc structures from a dma pool, so the memory caching
rules are properly set to avoid this issue.

Cc: Zhangfei Gao 
Cc: Krzysztof Kozlowski 
Cc: Maxime Ripard 
Cc: Vinod Koul 
Cc: Dan Williams 
Cc: Mark Brown 
Cc: Andy Green 
Acked-by: Zhangfei Gao 
Signed-off-by: John Stutlz 
Signed-off-by: Vinod Koul

k3dma: Fix "nobody cared" message seen on any error

2016-08-31T04:39:32+00:00

As it was before, as soon as the DMAC IP felt there was an error
he would return IRQ_NONE since no actual transfer had completed.

After spinning on that for 100K interrupts, Linux yanks the IRQ with
a "nobody cared" error.

This patch lets it handle the interrupt and keep the IRQ alive.

Cc: Zhangfei Gao 
Cc: Krzysztof Kozlowski 
Cc: Maxime Ripard 
Cc: Vinod Koul 
Cc: Dan Williams 
Cc: Mark Brown 
Cc: Andy Green 
Acked-by: Zhangfei Gao 
Signed-off-by: Andy Green 
[jstultz: Forward ported to mainline]
Signed-off-by: John Stultz 
Signed-off-by: Vinod Koul