Re: [PATCH] rdma_rxe: call comp_handler without holding cq->cq_lock

[Zhu Yanjun]

在 2025/8/12 8:54, Daisuke Matsuda 写道:
> On 2025/08/11 22:48, Zhu Yanjun wrote:
> > 在 2025/8/10 22:26, Philipp Reisner 写道:
> > > On Thu, Aug 7, 2025 at 3:09 AM Zhu Yanjun <yanjun.zhu@xxxxxxxxx> wrote:
> > > > 在 2025/8/6 5:39, Philipp Reisner 写道:
> > > > > Allow the comp_handler callback implementation to call ib_poll_cq().
> > > > > A call to ib_poll_cq() calls rxe_poll_cq() with the rdma_rxe driver.
> > > > > And rxe_poll_cq() locks cq->cq_lock. That leads to a spinlock 
> > > > > deadlock.
> > > > >
> > > > > The Mellanox and Intel drivers allow a comp_handler callback
> > > > > implementation to call ib_poll_cq().
> > > > >
> > > > > Avoid the deadlock by calling the comp_handler callback without
> > > > > holding cq->cw_lock.
> > > > >
> > > > > Signed-off-by: Philipp Reisner <philipp.reisner@xxxxxxxxxx>
> > > >
> > > > ERROR: test_resize_cq (tests.test_cq.CQTest.test_resize_cq)
> > > > Test resize CQ, start with specific value and then increase and 
> > > > decrease
> > > > ----------------------------------------------------------------------
> > > > Traceback (most recent call last):
> > > >     File "/root/deb/rdma-core/tests/test_cq.py", line 135, in 
> > > > test_resize_cq
> > > >       u.poll_cq(self.client.cq)
> > > >     File "/root/deb/rdma-core/tests/utils.py", line 687, in poll_cq
> > > >       wcs = _poll_cq(cq, count, data)
> > > >             ^^^^^^^^^^^^^^^^^^^^^^^^^
> > > >     File "/root/deb/rdma-core/tests/utils.py", line 669, in _poll_cq
> > > >       raise PyverbsError(f'Got timeout on polling ({count} CQEs 
> > > > remaining)')
> > > > pyverbs.pyverbs_error.PyverbsError: Got timeout on polling (1 CQEs
> > > > remaining)
> > > >
> > > > After I applied your patch in kervel v6.16, I got the above errors.
> > > >
> > > > Zhu Yanjun
> > >
> > > Hello Zhu,
> > >
> > > When I run the test_resize_cq test in a loop (100 runs each) on the
> > > original code and with my patch, I get about the same failure rate.
> >
> > Add Daisuke Matsuda
> >
> > If I remember it correctly, when Daisuke and I discussed ODP patches, 
> > we both made tests with rxe, from our tests results, it seems that 
> > this test_resize_cq error does not occur.
>
> Hi Zhu and Philipp,
>
> As far as I know, this error has been present for some time.
> It might be possible to investigate further by capturing a memory dump 
> while the polling is stuck, but I have not had time to do that yet.
> At least, I can confirm that this is not a regression caused by 
> Philipp's patch.

Hi, Daisuke

Thanks a lot. I’m now able to consistently reproduce this problem. I 
have created a commit here: 
https://github.com/zhuyj/linux/commit/8db3abc00bf49cac6ea1d5718d28c6516c94fb4e.

After applying this commit, I ran test_resize_cq 10,000 times, and the 
problem did not occur.

I’m not sure if there’s a better way to fix this issue. If anyone has a 
better solution, please share it.

Thanks a lot.
Zhu Yanjun

> Thanks,
> Daisuke