| 1 |
Dale wrote: |
| 2 |
> Kerin Millar wrote: |
| 3 |
>> Helmut Jarausch wrote: |
| 4 |
>>> On 10/24/2012 03:54:39 PM, Nikos Chantziaras wrote: |
| 5 |
>>>> Kernels 3.4, 3.5, and 3.6 can result in severe data corruption if |
| 6 |
>>>> you're using the EXT4 filesystem: |
| 7 |
>>> It looks as if Eric Sandeen has found the culprit and Theodore Ts'o has |
| 8 |
>>> suggested this patch yesterday |
| 9 |
>>> https://lkml.org/lkml/2012/10/28/309 |
| 10 |
>>> |
| 11 |
>>> Helmut. |
| 12 |
>>> |
| 13 |
>> Here's the final patch: |
| 14 |
>> |
| 15 |
>> https://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=ffb5387 |
| 16 |
>> |
| 17 |
>> |
| 18 |
>> It turns out to have had nothing to do with nobarrier. |
| 19 |
>> |
| 20 |
>> --Kerin |
| 21 |
>> |
| 22 |
>> |
| 23 |
> |
| 24 |
> |
| 25 |
> Could you explain a little on who it could have affected? Is it more |
| 26 |
> serious or less serious than originally thought? This is for those of |
| 27 |
> us who don't subscribe to the kernel mailing list, which I have read is |
| 28 |
> hugely active. |
| 29 |
|
| 30 |
I would describe the bug itself as serious but it will affect few users |
| 31 |
because journal_checksum isn't enabled by default. Ted submitted a patch |
| 32 |
to enable the option by default back in 2009 but it was reverted a few |
| 33 |
months later by Linus due to this bug: |
| 34 |
|
| 35 |
https://bugzilla.kernel.org/show_bug.cgi?id=14354 |
| 36 |
|
| 37 |
I'm pretty sure that it hasn't since been re-enabled as a default. As |
| 38 |
long as you haven't enabled this option, you shouldn't have anything to |
| 39 |
worry about. |
| 40 |
|
| 41 |
--Kerin |
Archives