Bug 24924 - nptl/tst-rwlock9 and nptl/tst-rwlock18 fail on aarch64
Summary: nptl/tst-rwlock9 and nptl/tst-rwlock18 fail on aarch64
Status: NEW
Alias: None
Product: glibc
Classification: Unclassified
Component: nptl (show other bugs)
Version: 2.30
: P2 normal
Target Milestone: ---
Assignee: Not yet assigned to anyone
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2019-08-21 11:01 UTC by Andreas Schwab
Modified: 2019-08-22 10:32 UTC (History)
3 users (show)

See Also:
Host: aarch64-*-*
Target:
Build:
Last reconfirmed:
fweimer: security-


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description Andreas Schwab 2019-08-21 11:01:15 UTC
nptl/tst-rwlock9 and nptl/tst-rwlock18 never finish, with an infinite stream of "reader thread %d tries again".
Comment 1 Florian Weimer 2019-08-21 11:07:24 UTC
I don't think we have seen that in our testing.  Would you please share more details about the environment (kernel and silicon)?  Thanks.
Comment 2 Andreas Schwab 2019-08-21 11:56:37 UTC
It's using the standard Factory kernel, which is 5.2.8 right now.  The hardware is probably a Cavium ThunderX1.
Comment 3 Andreas Schwab 2019-08-21 12:00:33 UTC
It looks like the tests only fail when run on a ThunderX1.  Other build workers, even those with a ThunderX2, do not cause them to fail.
Comment 4 Andreas Schwab 2019-08-21 12:19:05 UTC
That appears to be a regression introduced between 2019-07-07 and 2019-07-14.
Comment 5 Adhemerval Zanella 2019-08-21 17:33:34 UTC
The test was recently changed to cover pthread_rwlock_clock*. Is this what is failing? Which kind of failure
Comment 6 Adhemerval Zanella 2019-08-21 17:35:35 UTC
(In reply to Adhemerval Zanella from comment #5)
> The test was recently changed to cover pthread_rwlock_clock*. Is this what
> is failing? Which kind of failure

I hit send prior to finishing the last sentence.  Which return value pthread_rwlock_* shows in the loop?
Comment 7 Andreas Schwab 2019-08-22 10:16:19 UTC
It's ETIMEDOUT of course, otherwise the loop would stop.  It's the CLOCK_REALTIME test that fails.
Comment 8 Andreas Schwab 2019-08-22 10:17:56 UTC
And the CLOCK_MONOTONIC test fails in the same way.
Comment 9 Andreas Schwab 2019-08-22 10:32:54 UTC
When I comment out the "reader thread tries again" messages (and skip the timedlock test), all I see is this:

writer thread 0 tries again
writer thread 1 tries again
writer thread 0 succeeded
writer thread 2 tries again
writer thread 3 tries again
writer thread 4 tries again
writer thread 5 tries again
writer thread 6 tries again
writer thread 7 tries again
writer thread 8 tries again
writer thread 9 tries again
writer thread 10 tries again
writer thread 11 tries again
writer thread 12 tries again
writer thread 13 tries again
writer thread 14 tries again
writer thread 0 released
writer thread 0 tries again
writer thread 2 tries again
writer thread 3 tries again
writer thread 4 tries again
writer thread 5 tries again
writer thread 6 tries again
writer thread 7 tries again
writer thread 8 tries again
writer thread 9 tries again
writer thread 10 tries again
writer thread 11 tries again
writer thread 12 tries again
writer thread 13 tries again
writer thread 14 tries again