On 2020-06-29 15:03, Christoph Hellwig wrote:
> Hi all,
>
> this series lifts the somewhat hacky checks in the XSK code if a DMA
> streaming mapping needs dma_sync_single_for_{device,cpu} calls to the
> DMA API.
>
Thanks a lot for working on, and fixing this, Christoph!
I took the series for a spin, and there are (obviously) no performance
regressions.
Would the patches go through the net/bpf trees or somewhere else?
For the series:
Tested-by: Björn Töpel <[email protected]>
Acked-by: Björn Töpel <[email protected]>
Björn
>
> Diffstat:
> Documentation/core-api/dma-api.rst | 8 +++++
> include/linux/dma-direct.h | 1
> include/linux/dma-mapping.h | 5 +++
> include/net/xsk_buff_pool.h | 6 ++--
> kernel/dma/direct.c | 6 ++++
> kernel/dma/mapping.c | 10 ++++++
> net/xdp/xsk_buff_pool.c | 54 ++-----------------------------------
> 7 files changed, 37 insertions(+), 53 deletions(-)
>
On Mon, Jun 29, 2020 at 03:39:01PM +0200, Bj?rn T?pel wrote:
> On 2020-06-29 15:03, Christoph Hellwig wrote:
>> Hi all,
>>
>> this series lifts the somewhat hacky checks in the XSK code if a DMA
>> streaming mapping needs dma_sync_single_for_{device,cpu} calls to the
>> DMA API.
>>
>
> Thanks a lot for working on, and fixing this, Christoph!
>
> I took the series for a spin, and there are (obviously) no performance
> regressions.
>
> Would the patches go through the net/bpf trees or somewhere else?
either the net tree or (my) dma-mapping tree would be fine with me.
On Mon, Jun 29, 2020 at 03:39:01PM +0200, Bj?rn T?pel wrote:
> On 2020-06-29 15:03, Christoph Hellwig wrote:
>> Hi all,
>>
>> this series lifts the somewhat hacky checks in the XSK code if a DMA
>> streaming mapping needs dma_sync_single_for_{device,cpu} calls to the
>> DMA API.
>>
>
> Thanks a lot for working on, and fixing this, Christoph!
>
> I took the series for a spin, and there are (obviously) no performance
> regressions.
>
> Would the patches go through the net/bpf trees or somewhere else?
Where did this end up? I still don't see it in Linus' tree and this
is getting urgent now.
On 7/8/20 9:44 AM, Christoph Hellwig wrote:
> On Mon, Jun 29, 2020 at 03:39:01PM +0200, Björn Töpel wrote:
>> On 2020-06-29 15:03, Christoph Hellwig wrote:
>>> Hi all,
>>>
>>> this series lifts the somewhat hacky checks in the XSK code if a DMA
>>> streaming mapping needs dma_sync_single_for_{device,cpu} calls to the
>>> DMA API.
>>>
>>
>> Thanks a lot for working on, and fixing this, Christoph!
>>
>> I took the series for a spin, and there are (obviously) no performance
>> regressions.
>>
>> Would the patches go through the net/bpf trees or somewhere else?
>
> Where did this end up? I still don't see it in Linus' tree and this
> is getting urgent now.
It was merged into bpf tree and we sent the PR to DaveM which was merged into
net tree around a week ago [0]; I assume the PR for net might go to Linus soon
this week.
[0] https://git.kernel.org/pub/scm/linux/kernel/git/netdev/net.git/commit/?id=e708e2bd55c921f5bb554fa5837d132a878951cf