[LTP] [PATCH] epoll_pwait06: Skip BPF map file descriptor

[Martin Doucha]

On 9/24/25 14:08, Cyril Hrubis wrote:
> Hi!
>> I'll accept EPERM only for the file descriptor types which are now
>> unconditionally skipped. The only file descriptor type which could then
>> get skipped incorrectly will be BPF. But that's not a problem because
>> verifying epoll support is out of scope of this test. The primary
>> purpose is to verify that small epoll_pwait() timeouts won't get
>> misinterpreted as infinity. In theory, verifying that on a single file
>> descriptor type should be sufficient.
> 
> Ah, I had no idea that this is a regression test for a generic epoll
> code, it never occured to me that we run something like that for all
> types of file descriptors. For that test it makes sense to just skip
> unsupported fds.
> 
> I guess that we are actually mixing two tests and it would make sense to
> separate these two. I.e. one that is a regression test for the epoll bug
> and second, probably called epoll_ctl06.c that would hammer the
> EPOLL_CTL_ADD with all kinds of fds and expect either success or error.

Sure, we can add another test for epoll_ctl() later. I wrote 
epoll_pwait06 because I actually ran into the timeout regression while 
playing a game at home. I decided to loop over multiple file descriptor 
types because it's simple enough and provides better coverage for edge 
cases.

Now, should I resubmit with that EPERM check or is the one-line fix good 
enough as is?

-- 
Martin Doucha   mdoucha@suse.cz
SW Quality Engineer
SUSE LINUX, s.r.o.
CORSO IIa
Krizikova 148/34
186 00 Prague 8
Czech Republic