Insert a padding between data and the stored_xfer_buffer pointer to
ensure they are not on the same cache line.
Otherwise, the stored_xfer_buffer gets corrupted for IN URBs on
non-cache-coherent systems. (In my case: Lantiq xRX200 MIPS)
Fixes: 3bc04e28a030 ("usb: dwc2: host: Get aligned DMA in a more supported way")
Fixes: 56406e017a88 ("usb: dwc2: Fix DMA alignment to start at allocated boundary")
Signed-off-by: Martin Schiller <[email protected]>
---
drivers/usb/dwc2/hcd.c | 10 +++++++---
1 file changed, 7 insertions(+), 3 deletions(-)
diff --git a/drivers/usb/dwc2/hcd.c b/drivers/usb/dwc2/hcd.c
index dd82fa516f3f..f3035dd4db25 100644
--- a/drivers/usb/dwc2/hcd.c
+++ b/drivers/usb/dwc2/hcd.c
@@ -2664,8 +2664,10 @@ static void dwc2_free_dma_aligned_buffer(struct urb *urb)
return;
/* Restore urb->transfer_buffer from the end of the allocated area */
- memcpy(&stored_xfer_buffer, urb->transfer_buffer +
- urb->transfer_buffer_length, sizeof(urb->transfer_buffer));
+ memcpy(&stored_xfer_buffer,
+ PTR_ALIGN(urb->transfer_buffer + urb->transfer_buffer_length,
+ dma_get_cache_alignment()),
+ sizeof(urb->transfer_buffer));
if (usb_urb_dir_in(urb)) {
if (usb_pipeisoc(urb->pipe))
@@ -2697,6 +2699,7 @@ static int dwc2_alloc_dma_aligned_buffer(struct urb *urb, gfp_t mem_flags)
* DMA
*/
kmalloc_size = urb->transfer_buffer_length +
+ (dma_get_cache_alignment() - 1) +
sizeof(urb->transfer_buffer);
kmalloc_ptr = kmalloc(kmalloc_size, mem_flags);
@@ -2707,7 +2710,8 @@ static int dwc2_alloc_dma_aligned_buffer(struct urb *urb, gfp_t mem_flags)
* Position value of original urb->transfer_buffer pointer to the end
* of allocation for later referencing
*/
- memcpy(kmalloc_ptr + urb->transfer_buffer_length,
+ memcpy(PTR_ALIGN(kmalloc_ptr + urb->transfer_buffer_length,
+ dma_get_cache_alignment()),
&urb->transfer_buffer, sizeof(urb->transfer_buffer));
if (usb_urb_dir_out(urb))
--
2.11.0
Hi,
On Sun, Feb 17, 2019 at 10:37 PM Martin Schiller <[email protected]> wrote:
>
> Insert a padding between data and the stored_xfer_buffer pointer to
> ensure they are not on the same cache line.
>
> Otherwise, the stored_xfer_buffer gets corrupted for IN URBs on
> non-cache-coherent systems. (In my case: Lantiq xRX200 MIPS)
>
> Fixes: 3bc04e28a030 ("usb: dwc2: host: Get aligned DMA in a more supported way")
> Fixes: 56406e017a88 ("usb: dwc2: Fix DMA alignment to start at allocated boundary")
> Signed-off-by: Martin Schiller <[email protected]>
> ---
> drivers/usb/dwc2/hcd.c | 10 +++++++---
> 1 file changed, 7 insertions(+), 3 deletions(-)
This patch has been in the back of my mind for a while bug I never got
around to it. Today I was debugging memory corruption problems when
using a webcam on dwc2 on rk3288-veyron-jerry. This patch appears to
solve my problems nicely. Thanks!
Tested-by: Douglas Anderson <[email protected]>
Reviewed-by: Douglas Anderson <[email protected]>
Cc: <[email protected]>
On 5/29/2019 10:31 PM, Doug Anderson wrote:
> Hi,
>
> On Sun, Feb 17, 2019 at 10:37 PM Martin Schiller <[email protected]> wrote:
>>
>> Insert a padding between data and the stored_xfer_buffer pointer to
>> ensure they are not on the same cache line.
>>
>> Otherwise, the stored_xfer_buffer gets corrupted for IN URBs on
>> non-cache-coherent systems. (In my case: Lantiq xRX200 MIPS)
>>
>> Fixes: 3bc04e28a030 ("usb: dwc2: host: Get aligned DMA in a more supported way")
>> Fixes: 56406e017a88 ("usb: dwc2: Fix DMA alignment to start at allocated boundary")
>> Signed-off-by: Martin Schiller <[email protected]>
>> ---
>> drivers/usb/dwc2/hcd.c | 10 +++++++---
>> 1 file changed, 7 insertions(+), 3 deletions(-)
>
> This patch has been in the back of my mind for a while bug I never got
> around to it. Today I was debugging memory corruption problems when
> using a webcam on dwc2 on rk3288-veyron-jerry. This patch appears to
> solve my problems nicely. Thanks!
>
> Tested-by: Douglas Anderson <[email protected]>
> Reviewed-by: Douglas Anderson <[email protected]>
> Cc: <[email protected]>
>
Acked-by: Minas Harutyunyan <[email protected]>