This bug has been reported at Ubuntu and Launchpad already: http://ubuntuforums.org/showthread.php?t=1475124 https://bugs.launchpad.net/ubuntu/+source/linux/+bug/515937 Systems (aggregated from other affected people): - Thinkpad T61p, Ubuntu 10.04, 64bit, ext4/ext3 (me) - ThinkPad T400, Ubuntu 10.04, 64bit, ext4 (launchpad thread) - ASUS n80vn, Ubuntu 10.04, 64bit, ext4 (ubuntu forums thread) Rest of description is solely what I experienced on my T61p. Steps to reproduce: 1) Install Ubuntu 10.04, install restricted NVidia drivers, run Update Manager 2) Start watching movie, after about 20 minutes system freezes. Actual results: System freezes, then after about 20sec it comes back, root partition is remounted read-only.
Created attachment 26437 [details] lspci -v output
Created attachment 26438 [details] dmesg
Accidentally submitted to early. Here is the rest of bug report: Expected Results: Errors should not occur. Additional Information: This happens most often when I try to watch a movie. What is interesting, I'm watching movie from my second partition (ext3), then system freezes and root partition (ext4) gets remounted read-only. This also happens while torrent client is enabled. When I keep my profile low and don't use hdd much system is stable. My HDD controller is "82801HBM/HEM (ICH8M/ICH8M-E) SATA IDE Controller", but this bug is not exclusive for it, as guy in launchpad thread reports very similar behavior on "Intel Corporation ICH9M/M-E SATA AHCI Controller [8086:2929] (rev 03)" I checked ram with memtest some time ago, and it's OK. HDD Smart test OK. In general, system is rock solid on Ubuntu 8.10. Ubuntu 9.04 doesn't work on my laptop, never tested 9.10. Also guy at launchpad thread submitted his kern.log and lspci outputs. I'm ready to give you further help, perform test, try newer kernel etc. as long as I'm pointed in right direction.
FLUSH_EXT timed out which shouldn't happen but can. libata as of 2.6.32 doesn't retry after any FLUSH failure and just returns the error to upper layer leading to fs ro remounting the device. The reason for the behavior is that FLUSH failure often indicates (abort by device always does) data loss so continuing RW operation is likely to cause massive filesystem corruption. As the behavior caused some spurious failures like this, EH was updated to distinguish between various FLUSH failure modes and retry unless it's certain that the device aborted it. So, in short, please upgrade to newer kernel or tell your distro to backport the update. Thanks.
Resolving as PATCH_ALREADY_AVAILABLE. Thanks.
Wow, that was quick. Thanks a lot!
BTW, the commit ID is 6013efd8860bf15c1f86f365332642cfe557152f.