"Berg, Benjamin" <benjamin.berg@xxxxxxxxx> writes:
> Hi,
>
> On Wed, 2024-08-07 at 16:49 +0300, Kalle Valo wrote:
>> Chris Bainbridge <chris.bainbridge@xxxxxxxxx> writes:
>>
>> > On Wed, 7 Aug 2024 at 13:10, Kalle Valo <kvalo@xxxxxxxxxx> wrote:
>> > > [SNIP]
>> > Not really. It is functional but tx throughput is severely
>> > degraded,
>> > iperf shows ~1mbit/sec instead of the usual ~600mbit. Rx throughput
>> > (measured with `iperf -c ... -R`) doesn't seem to be affected.
>>
>> Ok, so a quite severe issue. Hopefully Intel can find a fix,
>> otherwise
>> we need to consider should we revert commit 90db50755228 ("wifi:
>> iwlwifi: use already mapped data when TXing an AMSDU") for v6.11.
>
> Yes, if we do not have a fix soon we'll need to revert. If so, then we
> should also revert:
>
> wifi: iwlwifi: keep the TSO and workaround pages mapped
> wifi: iwlwifi: map entire SKB when sending AMSDUs
>
> as well as the related bugfix:
> wifi: iwlwifi: correctly reference TSO page information
>
> Otherwise there is a performance impact as DMA mapping is effectively
> done twice.
That's good to know, thanks. But let's hope you find a fix.
--
https://patchwork.kernel.org/project/linux-wireless/list/
https://wireless.wiki.kernel.org/en/developers/documentation/submittingpatches
