Well, I set the system to the default ordered mode and the problem
went away.  EXT3 recovers nicely now.  I run across this all the time
since I develop high speed kernel stuff and have a lot of cases where
a bug crashes the system.  This time it showed up while developing the
MDB debugger with the hw_breakpoint interface which caused the system
to crash until I figured out this newer interface had hooked the
notify_die handlers and was trapping breakpoints which caused a lot of
hangs until I fixed it, so it is something I ran across coincidently.
The default ordered mode makes ext3 robust again.

Jeff


On Thu, Jun 10, 2010 at 3:04 PM, Bill Davidsen <davidsen@tmr.com> wrote:
quoted text
> Jeffrey Merkey wrote:
>>
>> ---------- Forwarded message ----------
>> From: Jeffrey Merkey <jeffmerkey@gmail.com>
>> Date: Mon, Jun 7, 2010 at 7:55 PM
>> Subject: Re: EXT3 File System Corruption 2.6.34
>> To: Eric Sandeen <sandeen@sandeen.net>
>>
>>
>>> On Jun 7, 2010, at 6:55 PM, Jeffrey Merkey <jeffmerkey@gmail.com> wrote:
>>>
>>>> ---------- Forwarded message ----------
>>>> From: Jeffrey Merkey <jeffmerkey@gmail.com>
>>>> Date: Mon, Jun 7, 2010 at 5:54 PM
>>>> Subject: Re: EXT3 File System Corruption 2.6.34
>>>> To: Eric Sandeen <sandeen@sandeen.net>
>>>>
>>>>
>>>> REPLY TO ALL
>>>>
>>>> CONFIG_EXT3_DEFAULTS_TO_ORDERED is not set
>>>>
>>>> Whether set this way or not, should not see corruption.
>>>
>>> Here you are mistaken.  Mount with data=ordered and see.  Writeback can
>>> expose stale data.
>>>
>>> -Eric
>>>
>>
>> OK.  I will set this up.  You may want to make this option the default
>> in the build scripts.  here is a corrupted file.  This was a .gif
>> image file I saved THEN AFTER SAVING THE FILE I pulled the power to
>> the machine and during recovery the file was FUCKED.  At any rate,
>> this does not happen with 2.6.28.
>>
> Having bad things happen when power is removed is not much of a surprise,
> and various options can fix that at the cost of speed. The fact that this
> didn't happen with 2.6.28 is bothersome.
>
> I actually take some care to avoid testing behavior in this area, not my
> normal intended mode of operation.
>
> --
> Bill Davidsen <davidsen@tmr.com>
>  "We have more to fear from the bungling of the incompetent than from
> the machinations of the wicked."  - from Slashdot
>
--
unsubscribe notice
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/