Description of problem: Kernels starting at 2.6.23-0.202.rc8.fc8 (tried 2.6.23-0.202.rc8.fc8, 2.6.23-0.204.rc8.fc8 and 2.6.23-0.211.rc8.git2.fc8) hangs while running udev. Kernels before and up to 2.6.23-0.195.rc7.git3.fc8 works fine. Version-Release number of selected component (if applicable): 2.6.23-0.202.rc8.fc8 and newer. How reproducible: Always Steps to Reproduce: 1. boot kernel 202 or greater Actual results: Hangs after 'Starting udevd...' Expected results: Booting machine. Additional info: hardware: http://smolt.fedoraproject.org/show?UUID=d6158ae5-29cf-4280-9e3d-6256b4ae9d64 lspci and other info found in bug #240343 (if more info is needed just ask). I have loaded netconsole from initrd (patching it), and I have attached the full boot until the hang, and the output got pressing SysRQ-<almost-everything>
Created attachment 211271 [details] Boot log from kernel 211 until hang.
Vanilla kernel from git works, I tried: $ git-clone git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git linux-2.6 $ cd linux-2.6 $ git log | head -1 commit 1bef7dc00caa7bcbff4fdb55e599e2591461fafa $ cp /boot/config-2.6.23-0.211.rc8.git2.fc8 .config $ make oldconfig $ make $ su Password: # make modules_install install # reboot And this kernel starts udev just fine and works.
kernel 2.6.23-0.214.rc8.git2.fc8 still broken.
Process events/0 is spinning. Can you do sysrq-P multiple times to see where it is stuck?
Created attachment 212871 [details] Boot log from kernel 211, lots of sysrq-P pressed after hang There is the boot log after pressing lots of sysrq-P. Kernel is always 211. There is also a sysrq-T just in case the processes are different. overall it looks very boring to my kernel-illiterate eyes: only ESI, EDI, EBX and ECX seems to change. ESI and EDI seems to grow together, so I waited for them to warp around, but nothing happened :-/ Tell me if I can try something else.
kernel 2.6.23-6.fc8.i686 is still broken, but I did a "bisection" downloading from koji all the kernels between kernel-2.6.23-0.195.rc7.git3.fc8.i686 (the last good) and kernel-2.6.23-0.202.rc8.fc8.i686 (the first bad), and I found that kernel-2.6.23-0.197.rc7.git5.fc8.i686 works, while kernel-2.6.23-0.198.rc7.git5.fc8.i686 is broken. The only difference between those kernels is the updated linux-2.6-highres-timers.patch. Sadly the changes between the old version (in 197) and new version (in 198) are big. I will try compiling 198 with the old patch, or without if it doesn't apply.
kernel-2.6.23.1-42.fc8 still hangs, But I have some more info: - vanilla 2.6.24-rc1 hangs, while vanilla 2.6.23 works. - a git bisect on vanilla kernel points to commit 4f86d3a8e297205780cca027e974fd5f81064780 as the first bad commit. This is the first commit after v2.6.23! - passing the 'nolapic' parameter fixes the problem for all the hanging kernels. I will try reporting the regression to LKML.
That commit is the cpuidle code; Fedora 8 had that before it went into 2.6.24-rc so it makes sense that it is the cause of this problem.
I moved this bug to uptream bugzilla: http://bugzilla.kernel.org/show_bug.cgi?id=9355