On Wed, Mar 26, 2025 at 02:57:37PM -0700, Stanislav Fomichev wrote:

On 03/26, Samiullah Khawaja wrote:

quoted

On Tue, Mar 25, 2025 at 10:47 AM Martin Karsten [off-list ref] wrote:

quoted

On 2025-03-25 12:40, Samiullah Khawaja wrote:

quoted

On Sun, Mar 23, 2025 at 7:38 PM Martin Karsten [off-list ref] wrote:

quoted

On 2025-03-20 22:15, Samiullah Khawaja wrote:

quoted

quoted

quoted

Nice catch. It seems a recent change broke the busy polling for AF_XDP
and there was a fix for the XDP_ZEROCOPY but the XDP_COPY remained
broken and seems in my experiments I didn't pick that up. During my
experimentation I confirmed that all experiment modes are invoking the
busypoll and not going through softirqs. I confirmed this through perf
traces. I sent out a fix for XDP_COPY busy polling here in the link
below. I will resent this for the net since the original commit has
already landed in 6.13.
https://lore.kernel.org/netdev/CAAywjhSEjaSgt7fCoiqJiMufGOi=oxa164_vTfk+3P43H60qwQ@mail.gmail.com/T/#t (local)

In general, when sending the patches and numbers, try running everything
against the latest net-next. Otherwise, it is very confusing to reason
about..

I had mentioned in my previous review, but worth mentioning again...
using --base=auto when using format-patch to generate the base
commit SHA would be really useful to potentially help avoid this
problem.