[Bug 610055] Re: computer stuttering, freezing, hanging, and otherwise being uncooperative

[Adam B Butler]

Same thing is happening with me -- only started with upgrade to Maverick
(on Dell E6500 laptop, x64).

After some poking around, it seems that the problem stems from this
little guy:

  /usr/lib/pm-utils/power.d/journal-commit

...which (unless I misunderstand the syntax) is remounting all
drives/partitions every time there is a power "event" -- which I suppose
includes all kinds of stuff, eg, shutting the lid on a laptop,
suspend/hibernate, resuming from suspend/hibernate, etc.  The annoying
thing (for me at least) is that this occurs every time the power cord is
disconnected.  (So when I bump the power cord loose and then immediately
plug it back in, this seemingly innocent action results in my drives
being mounted/remounted with commit=600, then mounted/remounted with
commit=0 ... and if done too quickly, this results in (at best) errors
in dmesg like everyone has already listed above, or (at worst) system
freezing/hanging, unable to switch from X to a VT, etc.)

Not to hijack this bug  -- because I think this is closely (if not
directly) related to the same problem -- but has anyone taken a look in
their /etc/fstab file (or the output of "mount") lately?   Because at
one point yesterday, when the system was slowing and about to lock up
again, I ran "mount" and saw the following craziness:

/dev/sda5 on / type ext4 (commit=0,commit=600,commit=0,commit=0,commit=0,commit=0,commit=0,commit=600,commit=0,commit=0)
/dev/sda5 on / type ext4 (commit=0,commit=600,commit=0)

Compared with today's "mount" output:
/dev/sda5 on / type ext4 (rw,errors=remount-ro,commit=0,commit=600)
/dev/sdb1 on /media/tera type ext4 (rw,errors=remount-ro,commit=0,commit=600)

My /etc/fstab file wasn't quite as crazy as all that (above) --  but it
was still gratuitous.  The original file read something like this --
sorry, having to provide this from memory because my /etc/fstab file was
either overwritten during the Maverick upgrade or this script is
rewriting it (to attach that "commit=XXX" stuff) over and over.

Original: 
/dev/sda5   /                  ext4   defaults,errors=remount-ro
/dev/sdb1   /tera           ext4   defaults,errors=remount-ro

Now:
/dev/sda5   /                  ext4    rw,commit=0,commit=600,commit=0
/dev/sdb1   /tera            ext4    rw,commit=0,commit=0

*Also note how all the original options (defaults,errors=remount-ro,etc)
have been clobbered by simply "rw,commit=x,commit=y,commit=x".

It seems to me that (at least part of) this stems from that particular
pm-utils/journal-commit script.  At first glance, it seems that when
entering a powersave state it remounts all disks (with commit=600) and
then once again when returning to a full-power state (eg, plugging the
system back into the wall), this time remounting all disks with
commit=0.  Somewhere in this process, the original option values
(defaults,errors=remount-ro) seem to get clobbered; maybe this has
something to do with the script failing to notice there's already a
"commit" value, so it tacks on additional ones, resulting in all the
gratuitous commit=X,commit=y,commit=X stuff getting "tacked on."

But I'm no expert; just trying to make an educated guess here...
Assigning this to pm-utils in the hopes of catching the eye of someone
far smarter than myself...  :)

** Changed in: linux (Ubuntu)
       Status: New => Confirmed

-- 
[pm-utils][maverick] system stutters, freezes, hangs, gratuitous 'Commit=0' errors
https://bugs.launchpad.net/bugs/610055
You received this bug notification because you are a member of Kernel
Bugs, which is subscribed to linux in ubuntu.