On Thu, May 09, 2013 at 02:52:39PM -0500, Bruce Dubbs wrote: > Pierre M.R. wrote: > > Bruce Dubbs wrote: > >> Arthur Radley wrote: > >>> [ 3.411285] EXT4-fs (sda1): INFO: recovery required on readonly > >>> filesystem > >> I think it's normal. The messages may vary with the driver/kernel. On > >> kernel 3.8.3-LFS-7.3, I have: > >> > >> [ 3.580354] EXT3-fs (sda16): recovery required on readonly filesystem > >> [ 3.580494] EXT3-fs (sda16): write access will be enabled during > >> recovery > >> [ 5.769971] EXT3-fs (sda16): recovery complete
To me, that time delay from 3.58 to 5.77 seconds suggests that it _did_ do recovery, as does 'recovery complete'. I first *noticed* these messages on one of my desktop machines last month, on a boot where I was 99% sure I'd shut down normally. So far, I haven't noticed them on the other desktop machine [ both have been running 3.9-rc and now 3.9 kernels during this ]. That doesn't mean they aren't "the new normal" but I think something is going on. Fortunately, ext3 and ext4 recover from this situation. But for me it doesn't happen on every boot, and isn't repeatable enough to debug. > >> [ 5.770128] EXT3-fs (sda16): mounted filesystem with writeback data mode > >> > >> but not the INFO part of the statement. > > > The ext3-msg and ext4-msg from the kernel differs on the INFO. The > > kernel doesn't give the write access before the possible recovery > > required after a crash. > > Some interesting points here. > > 1. /dev/sda16 on / type ext4 (rw) > 2. Until you mentioned it, I thought I shutdown normally at the last > boot, but I did have a power failure from a thunderstorm about > 10 days ago > 3. When a did a manual mount of /mnt/lfs I got: > > [ 1255.952175] EXT4-fs (sda13): mounting ext2 file system using the ext4 > subsystem > [ 1255.956668] EXT4-fs (sda13): warning: mounting unchecked fs, running > e2fsck is recommended That message suggests you exceeded either the maximal mount count or the number of days. > [ 1255.957730] EXT4-fs (sda13): mounted filesystem without journal. > Opts: (null) > > I'm doing an lfs build right now, but I'll umount/fsck/mount when done > and see if the messages are different. > > -- Bruce > > For my own builds I use a sed on the checkfs bootscript to remove your quietening of fsck [ I like the messages telling me an fsck will be due soon, and while a slow fsck is running ]. I know you don't like it, and it won't help directly here, but in a spirit of "don't hide things, this my development machine" I'll mention it again : sed -i '/fsck /s%>/dev/null%%' lfs/init.d/checkfs Good luck to anyone who tries to debug this. A repeatable test situation (apart from unclean shutdown) eludes me. ĸen -- das eine Mal als Tragödie, das andere Mal als Farce -- http://linuxfromscratch.org/mailman/listinfo/blfs-support FAQ: http://www.linuxfromscratch.org/blfs/faq.html Unsubscribe: See the above information page
