Re: [gentoo-user] Heads-up: Several kernel versions have severe EXT4 data corruption bug

[Kerin Millar]

Canek Peláez Valdés wrote:
> On Wed, Oct 24, 2012 at 8:09 PM, Kerin Millar<kerframil@×××××××××××.uk>  wrote:
>> Canek Peláez Valdés wrote:
>>> On Wed, Oct 24, 2012 at 8:54 AM, Nikos Chantziaras<realnc@×××××.com>
>>> wrote:
>>>> Kernels 3.4, 3.5, and 3.6 can result in severe data corruption if you're
>>>> using the EXT4 filesystem:
>>>>
>>>> http://www.phoronix.com/scan.php?page=news_item&px=MTIxNDQ
>>>>
>>>> This includes gentoo-sources.  I hope the Gentoo developers are on top of
>>>> this.  In the meantime, avoid doing reboots after too short an uptime.
>>>
>>> Doesn't seem to be that serious:
>>>
>>> https://plus.google.com/u/0/117091380454742934025/posts/Wcc5tMiCgq7
>>
>> Might I enquire as to the manner in which this comment impartially
>> establishes that the consequences of the bug upon those affected is not
>> serious?
>
> Oh, and about "impartiality"; this is a technical issue, not a
> philosophical one. I will always trust the expert's opinion over
> almost everyone's else.
>

The comment you linked to was fairly bereft of technical content, other 
than to assert that the circumstances under which the bug triggers are 
so limited that there is no general cause for concern. Given that (a) 
the investigation chronicled by the lkml thread remains ongoing (b) a 
remedy has yet to be conclusively determined, it is illogical that any 
statement as to the scope of the bug can anything more than a hypothesis 
at best, irrespective of how well-informed said hypothesis might be.

As for impartiality, it is entirely conceivable that someone in Ted's 
position would be riled by what they perceive (not necessarily 
correctly) as negative publicity and to respond in kind. Particularly 
when one carries a burden of responsibility of the subsystem in question.

Until such time as the matter is concluded, ext4 users that value their 
data will exercise due concern, naturally. The petty sniping about 
drumming up ad-revenue and silly 4chan style image memes do not strike 
me as a constructive way in which to assuage those concerns.

Further, the notion that nobarrier is an "esoteric" option is 
questionable. In my experience, it is common practice to employ it as a 
performance-enhancing measure on systems equipped with a battery-backed 
write cache; especially MySQL servers that must contend with a heavy 
workload. One wonders what he would have made of the notion of running 
ext4 without a journal, had it not been at the behest of Google.

In summary, I maintain that his fatuous Google+ post does nothing to 
establish just why it is that those of us in the peanut gallery should 
be unconcerned as to the impact of the bug. On my part, I will continue 
to be concerned until the investigation has fully run its course.

--Kerin