Ext2 vs. ext3 recovery after crash

Ext2 vs. ext3 recovery after crash

[Bill Davidsen]

I have a laptop (Dell Inspiron C600) which, like most Dell laptops,
crashes every time I log out of X. On some occasions on reboot I get a
message about replaying the journal, while occasionally I get a full ext2
style multi-pass 12 minute recovery. I don't see why the ext3 isn't always
used, I know it's going to crash, I always do a sync and wait ten seconds
for journal writes, etc, to take place.

I have tried all the usual, Redhat kernels, 2.4.17, 2.4.19, -aa, -ac,
disable io-apc, disable apic, disable all power management, boot noapic
(someone swore it wasn't enough to pull it out of the kernel ;-) all
producing about 20% chance of slow reboot.

Since I would have to spend my own money to replace this device with
something functional before 2003, is there something I'm missing about why
it does the slow cleanup? It was Redhat 7.1, updated fsutils and modutils,
pcmcia packed, etc, to latest of Mar 15 this year, in case that matters.
All kernels have ext3 compiled in, all work "most of the time."

-- 
bill davidsen <davidsen@tmr.com>
  CTO, TMR Associates, Inc
Doing interesting things with little computers since 1979.

Re: Ext2 vs. ext3 recovery after crash

[Jauder Ho]

Bill, you do know that it will do a full fsck every x mounts right?

[root@turtle /lib]# tune2fs -l /dev/hda6 | grep -i mount
Last mounted on:          <not available>
Last mount time:          Sun Mar  3 11:34:50 2002
Mount count:              1
Maximum mount count:      20

--Jauder

On Tue, 2 Apr 2002, Bill Davidsen wrote:

> I have a laptop (Dell Inspiron C600) which, like most Dell laptops,
> crashes every time I log out of X. On some occasions on reboot I get a
> message about replaying the journal, while occasionally I get a full ext2
> style multi-pass 12 minute recovery. I don't see why the ext3 isn't always
> used, I know it's going to crash, I always do a sync and wait ten seconds
> for journal writes, etc, to take place.
>
> I have tried all the usual, Redhat kernels, 2.4.17, 2.4.19, -aa, -ac,
> disable io-apc, disable apic, disable all power management, boot noapic
> (someone swore it wasn't enough to pull it out of the kernel ;-) all
> producing about 20% chance of slow reboot.
>
> Since I would have to spend my own money to replace this device with
> something functional before 2003, is there something I'm missing about why
> it does the slow cleanup? It was Redhat 7.1, updated fsutils and modutils,
> pcmcia packed, etc, to latest of Mar 15 this year, in case that matters.
> All kernels have ext3 compiled in, all work "most of the time."
>
> --
> bill davidsen <davidsen@tmr.com>
>   CTO, TMR Associates, Inc
> Doing interesting things with little computers since 1979.
>
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
>
>

Re: Ext2 vs. ext3 recovery after crash

[Bill Davidsen]

On Tue, 2 Apr 2002, Jauder Ho wrote:

> 
> Bill, you do know that it will do a full fsck every x mounts right?
> 
> [root@turtle /lib]# tune2fs -l /dev/hda6 | grep -i mount
> Last mounted on:          <not available>
> Last mount time:          Sun Mar  3 11:34:50 2002
> Mount count:              1
> Maximum mount count:      20

I assure you I have not set my max-mount down to three or four, and since
it crashes several times a day you can forget forced by date, too.

C'mon, I've been doing this stuff since the MCC four floppy distribution
was king, I've seen a max mount counts message before. This is purely a
case of the *first* mount message being EXT2 instead of EXT3, as if the
journal wasn't detected in the first place. However, the r/w mount is
always ext3 per fstab.

I never get close to max-mounts, this problem is in the 20-30% of the time
range. No one knows why a lot of the Dells hang on X-logout, guess this is
one more thing it does poorly.

-- 
bill davidsen <davidsen@tmr.com>
  CTO, TMR Associates, Inc
Doing interesting things with little computers since 1979.

Re: Ext2 vs. ext3 recovery after crash

[Andreas Dilger]

On Apr 03, 2002  05:05 -0500, Bill Davidsen wrote:
> C'mon, I've been doing this stuff since the MCC four floppy distribution
> was king, I've seen a max mount counts message before.

OK, well I've seen your name around a lot, but you never know...

> This is purely a case of the *first* mount message being EXT2 instead
> of EXT3, as if the journal wasn't detected in the first place.  However,
> the r/w mount is always ext3 per fstab.

Well, 'mount' output is useless w.r.t. the root filesystem, because it is
simply copied from /etc/fstab.  You need to check /proc/mounts to see if
it is _ever_ being mounted as ext3 (lots of people have this problem,
especially if they use initrds and ext3 as a module).

If it _is_ being mounted as ext2 sometimes and ext3 other times, it
would be informative to get the dmesg output, because it will tell
you why it couldn't load the journal.  Note that if there _is_ a
journal in use, it would normally prevent the filesystem from being
mounted as ext2 entirely after a crash.

Cheers, Andreas
--
Andreas Dilger  \ "If a man ate a pound of pasta and a pound of antipasto,
                 \  would they cancel out, leaving him still hungry?"
http://www-mddsp.enel.ucalgary.ca/People/adilger/               -- Dogbert

Re: Ext2 vs. ext3 recovery after crash

[Bill Davidsen]

On Wed, 3 Apr 2002, Andreas Dilger wrote:


> Well, 'mount' output is useless w.r.t. the root filesystem, because it is
> simply copied from /etc/fstab.  You need to check /proc/mounts to see if
> it is _ever_ being mounted as ext3 (lots of people have this problem,
> especially if they use initrds and ext3 as a module).

  The problem is that the initial mount is changing. I was at one point
making ext3 a module, and building initrd files with the ext3 modules and
fstab in the initrd. Didn't seem the way to go so I put ext3 in the
kernel, and that (usually) works. So I'm not making that particular error.

  Thanks for the ideas, I'm going to collect dmesg output and post so
someone can tell me I missed something obvious. Unless someone has a way
to so it from crashing, in which case I'll duck the problem for the
moment.

-- 
bill davidsen <davidsen@tmr.com>
  CTO, TMR Associates, Inc
Doing interesting things with little computers since 1979.

Re: Ext2 vs. ext3 recovery after crash

[Andreas Dilger]

On Apr 02, 2002  23:02 -0500, Bill Davidsen wrote:
> I have a laptop (Dell Inspiron C600) which, like most Dell laptops,
> crashes every time I log out of X. On some occasions on reboot I get a
> message about replaying the journal, while occasionally I get a full ext2
> style multi-pass 12 minute recovery. I don't see why the ext3 isn't always
> used, I know it's going to crash, I always do a sync and wait ten seconds
> for journal writes, etc, to take place.

You should be able to see easily if it is being checked by a forced
fsck, from a message like "maximal mount count reached, forcing fsck".

You can use the "tune2fs -c" option to increase the number of mounts
between full checks (or "tune2fs -c 0" to turn it off, and then use
"tune2fs -i 1m" to reduce the interval between checks to once a month or
whatever).  Note that it is _strongly_ recommended NOT to turn off both
mount- and time-based forced fscks on your filesystem, because things
like bad kernels, bad disks, RAM, cables, etc can still cause corruption.

Note also that some laptops can lose data at shutdown (even on a clean
shutdown) because their disks have problems with flushing the cache.

Cheers, Andreas
--
Andreas Dilger  \ "If a man ate a pound of pasta and a pound of antipasto,
                 \  would they cancel out, leaving him still hungry?"
http://www-mddsp.enel.ucalgary.ca/People/adilger/               -- Dogbert

AW: Ext2 vs. ext3 recovery after crash

[Dipl.-Inform. Guus Leeuw jr.]

Bill,

Just for the sake of it, the C600 should be a Latitude :)

Anyways, I had a C600 before as well, first on SuSE then on Redhat
because
For RedHat it is easier to recompile new kernels etc, being the reason I
kicked of SuSE.
I installed 7RH.2 and downloaded 2.4.17 sources to run on it, since
newer is better :)
Never had any problems. (I hope you did not specify DELL Support in the
options, did you?)

Than I bought myself an Inspiron4100 which is now running 2.4.18. The
only flaky things I've seen
Is sound support on KDE, but that might have been related to aRts, and
the ide-scsi driver does not
Seem to work anymore since 2.4.17. Rest is all fine, even APM and the
stuff.
I usually have uptimes of 8 hours, during the week, and 2.5 days over
the weekend.

I compile newer kernels for my Inspiron always with DELL Laptop Support,
which I never do for my C600.
Both are currently happily running 2.4.18. Both kernels supports ext3
and all filesystems except /dev/cdrom
Are running ext3. X Environment is XFree4 with ATI Mobility Radeon M6
Drivers. Kernel doesn't do any
Framebuffering whatsoever, but supports the AGP Drivers for Radeon. X
Desktop is KDE2, or KDE3RC3 depending on what I want to do :)

Compiler: gcc 2.95.x

Just my $0.02,
Guus

-----Ursprüngliche Nachricht-----
Von: linux-kernel-owner@vger.kernel.org
[mailto:linux-kernel-owner@vger.kernel.org] Im Auftrag von Bill Davidsen
Gesendet: Mittwoch, 3. April 2002 06:03
An: Linux-Kernel Mailing List
Betreff: Ext2 vs. ext3 recovery after crash


I have a laptop (Dell Inspiron C600) which, like most Dell laptops,
crashes every time I log out of X. On some occasions on reboot I get a
message about replaying the journal, while occasionally I get a full
ext2 style multi-pass 12 minute recovery. I don't see why the ext3 isn't
always used, I know it's going to crash, I always do a sync and wait ten
seconds for journal writes, etc, to take place.

I have tried all the usual, Redhat kernels, 2.4.17, 2.4.19, -aa, -ac,
disable io-apc, disable apic, disable all power management, boot noapic
(someone swore it wasn't enough to pull it out of the kernel ;-) all
producing about 20% chance of slow reboot.

Since I would have to spend my own money to replace this device with
something functional before 2003, is there something I'm missing about
why it does the slow cleanup? It was Redhat 7.1, updated fsutils and
modutils, pcmcia packed, etc, to latest of Mar 15 this year, in case
that matters. All kernels have ext3 compiled in, all work "most of the
time."

-- 
bill davidsen <davidsen@tmr.com>
  CTO, TMR Associates, Inc
Doing interesting things with little computers since 1979.

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel"
in the body of a message to majordomo@vger.kernel.org More majordomo
info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Dell laptop crashes -- data point (was Re: Ext2 vs. ext3 recovery after crash)

[Sean Neakums]

commence  Bill Davidsen quotation:

> I have a laptop (Dell Inspiron C600) which, like most Dell laptops,
> crashes every time I log out of X.

I have never had this happen with my Dell Inspiron 4100.  Running the
latest Debian unstable X packages (with a 4.1.99 server + modules
build overlaid for the non-VESA ATI Mobility support).

-- 
 /////////////////  |                  | The spark of a pin
<sneakums@zork.net> |  (require 'gnu)  | dropping, falling feather-like.
 \\\\\\\\\\\\\\\\\  |                  | There is too much noise.

Re: Ext2 vs. ext3 recovery after crash

[Alan Cox]

> style multi-pass 12 minute recovery. I don't see why the ext3 isn't always
> used, I know it's going to crash, I always do a sync and wait ten seconds
> for journal writes, etc, to take place.

tune2fs - the file system has a pair of values that force a check every
n mounts and every n days. You can modify these.

Alan

Re: Ext2 vs. ext3 recovery after crash

[Andreas Jellinghaus]

>I have a laptop (Dell Inspiron C600) which, like most Dell laptops,
>crashes every time I log out of X.

There is an Inspirion C600?
i have a litutude c600, and it works fine.

beside an issue with the graphic(*), it works very fine.

andreas
(*) sometimes with dvd/avi/video software and with opening/closeing the
display and with changeing to console/back to x, the screen gets wiered:
in a vertical cut, maybe 30% of the middle of the screen is missing,
filled with a duplication of the right part. everything still works, but
that 30% of the screen is not shown. after a reboot everything is ok,
sometimes a close display/open display helps.

Re: Ext2 vs. ext3 recovery after crash

[Bill Davidsen]

On Wed, 3 Apr 2002, Andreas Jellinghaus wrote:

> >I have a laptop (Dell Inspiron C600) which, like most Dell laptops,
> >crashes every time I log out of X.
> 
> There is an Inspirion C600?
> i have a litutude c600, and it works fine.

Man, good to see I'm not the only one who can't type ;-)

Yeah, the other laptop is an Inspiron I guess, I was googling for Dell and
Linux problems and had a brain fart. That said, yes, I have a "Latitude
C600" here, and when I exit X via logout it almost always hangs, the
screen turns pink, and I need to power cycle.

Given that yours doesn't do that, did you use:
- a custom XF86 config
- XFree from someone other than Redhat
- some custom config of X
- anything else you did?

Obviously I would rather not have the system hang all the time, better
even than having it boot fast ;-)

I've been running *very* recent kernels, mainly because the RH didn't work
any better. If you have a better install I'd like to hear it, although the
issues which pushed me to NSLG and 19-pre2 or news kernels are hard to
ignore.

Then again, several people have told me that they had multiple Dell
laptops and only some of them did this. Can you say "quality control" or
"running production changes?"

Thanks much for anything you can suggest, but in the meantime I will be
saving some dmesg output from ext2 and ext3 boots.

-- 
bill davidsen <davidsen@tmr.com>
  CTO, TMR Associates, Inc
Doing interesting things with little computers since 1979.

Re: Ext2 vs. ext3 recovery after crash

[Andreas Jellinghaus]

ok, i have a "LATITUDE C600",

i'm useing current debian woody, not redhat.
i didn't do anythign special. the xfree86 config is handwritten, not
generated with the debian tool (which wasn't working at that time).
will send you a copy in personal email (and anyone else interested).

my kernels are recent 2.4.* kernels, usualy the latest (currently
2.4.18) with some patch (freeswan, swsup or acpi, whatever i'm looking
at).

i have only one laptop, but at the university i dealt with several others,
but never had a pink-screen-logout--requires-reset-problem.

best regards, andreas