Re: [PATCH bpf-next] selftests/bpf: Add F_SETFL for fcntl

[John Fastabend <john.fastabend@gmail.com>]

Jakub Sitnicki wrote:
> Hi Geliang,
> 
> On Wed, Apr 03, 2024 at 04:32 PM +08, Geliang Tang wrote:
> > From: Geliang Tang <tanggeliang@kylinos.cn>
> >
> > Incorrect arguments are passed to fcntl() in test_sockmap.c when invoking
> > it to set file status flags. If O_NONBLOCK is used as 2nd argument and
> > passed into fcntl, -EINVAL will be returned (See do_fcntl() in fs/fcntl.c).
> > The correct approach is to use F_SETFL as 2nd argument, and O_NONBLOCK as
> > 3rd one.
> >
> > Fixes: 16962b2404ac ("bpf: sockmap, add selftests")
> > Signed-off-by: Geliang Tang <tanggeliang@kylinos.cn>
> > ---
> >  tools/testing/selftests/bpf/test_sockmap.c | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/tools/testing/selftests/bpf/test_sockmap.c b/tools/testing/selftests/bpf/test_sockmap.c
> > index 024a0faafb3b..34d6a1e6f664 100644
> > --- a/tools/testing/selftests/bpf/test_sockmap.c
> > +++ b/tools/testing/selftests/bpf/test_sockmap.c
> > @@ -603,7 +603,7 @@ static int msg_loop(int fd, int iov_count, int iov_length, int cnt,
> >  		struct timeval timeout;
> >  		fd_set w;
> >  
> > -		fcntl(fd, fd_flags);
> > +		fcntl(fd, F_SETFL, fd_flags);
> >  		/* Account for pop bytes noting each iteration of apply will
> >  		 * call msg_pop_data helper so we need to account for this
> >  		 * by calculating the number of apply iterations. Note user
> 
> Good catch. But we also need to figure out why some tests failing with
> this patch applied and fix them in one go:
> 
> # 6/ 7  sockmap::txmsg test skb:FAIL
> #21/ 7 sockhash::txmsg test skb:FAIL
> #36/ 7 sockhash:ktls:txmsg test skb:FAIL
> Pass: 42 Fail: 3
> 
> I'm seeing this error message when running `test_sockmap`:
> 
> detected skb data error with skb ingress update @iov[0]:0 "00 00 00 00" != "PASS"
> data verify msg failed: Unknown error -5
> rx thread exited with err 1.

I have a theory this is a real bug in the SK_SKB_STREAM_PARSER which has an
issue with wakup logic. Maybe we wakeup the poll/select logic before the
data is copied and because the recv() is actually nonblocking now we get
the error.

> 
> I'd also:
> - add an error check for fnctl, so we don't regress,
> - get rid of fd_flags, pass O_NONBLOCK flag directly to fnctl.
> 
> Thanks,
> -jkbs