time warp on 2.6.18-rt6 (2nd try)
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Ingo,
I sent this yesterday morning and haven't seen it show up. I wonder if my attached configs/logs were too large? Now referenced as URLs...
I'm still seeing the time warp ("It's just a jump to the left!" :))* being triggered on both my Athlon64x2 (32-bit kernel) and my Athlon64 up box (64-bit kernel). A boot log that shows the failure and the config files for both systems can be found at:
http://people.redhat.com/williams/rt
To trigger this, I was running pi_stress test from:
http://people.redhat.com/williams/tests/pi_tests-1.3.tar.gz
I was running it like this:
$ sudo pi_stress --verbose
Note: Be aware that I still haven't got pi_stress to stop reliably. It will catch SIGINT sometimes and sometimes it just blithely ignores it and sails on. Since in the default run there are ten groups of three threads all running SCHED_FIFO and performing a priority inversion scenario, this will mean a uniprocessor box will be almost unusable for other tasks. On my SMP box I can ssh into the box and kill the test with "kill <pid>".
Let me know if you need me to reconfigure and try something else.
Clark
* lame "Rocky Horror Picture Show" reference
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (GNU/Linux)
Comment: Using GnuPG with Fedora - http://enigmail.mozdev.org
iD8DBQFFORWMHyuj/+TTEp0RAq8QAJ9t30GOTjRmKD0Tkif94wQVyOyRTQCfbmX7 Qm3mAzmnnH8KDqpP2hoDwSE=
=kkAj
-----END PGP SIGNATURE-----
Quoted 3 lines · Clark Williams
I'm still seeing the time warp ("It's just a jump to the left!" :))* being triggered on both my Athlon64x2 (32-bit kernel) and my Athlon64 up box (64-bit kernel). [...]
that was most likely a false positive - every time settimeofday is done. I've just uploaded -rt7, could you check it whether the time warp messages are gone?
Ingo
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Quoted 11 lines · Ingo Molnar
Quoted 3 lines · Clark Williams
I'm still seeing the time warp ("It's just a jump to the left!" :))* being triggered on both my Athlon64x2 (32-bit kernel) and my Athlon64 up box (64-bit kernel). [...]
that was most likely a false positive - every time settimeofday is done. I've just uploaded -rt7, could you check it whether the time warp messages are gone?
Ingo
I ran pi_stress on both the 32-bit kernel (Athlon64x2) and a 64-bit kernel (Athlon64 up) and am still getting a time warp bug.
I'm rebuilding now with a bit more debug info turned on and I'm trying to track down an Intel box that I can try this on to see if somehow I'm exercising a code path that hasn't been touched much. Hopefully I'll be able to test on a P3/P4 box this afternoon and can confirm or deny whether this is AMD specific.
Clark
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (GNU/Linux)
Comment: Using GnuPG with Fedora - http://enigmail.mozdev.org
iD8DBQFFPOv6Hyuj/+TTEp0RAo2SAJ4iKMHJeYY8KHkK+cR9DdjW5caZbwCdHE+U J2N4AuipJwzZJi0n31BXFYE=
=VIvH
-----END PGP SIGNATURE-----