Re: fs/stat: Reduce memory requirements for stat_open

[Stefan Bader <stefan.bader@canonical.com>]

[-- Attachment #1: Type: text/plain, Size: 1640 bytes --]

On 25.06.2014 01:44, David Rientjes wrote:
> On Thu, 12 Jun 2014, Stefan Bader wrote:
> 
>> doh, so you guys have been hit by that before. And I have missed the fact that
>> single_open is special. Which makes the change for the upper limit do the wrong
>> thing. While long-term it sounds like changing it to vmalloc or iterative reads
>> sounds better, maybe the change from possible to online cpus might be something
>> that is better acceptable as a quick fix... Not sure. At least this giving back
>> a bit of attention to the matter and it is not only affecting zSeries. x86
>> starts to see hw that requires a similar high possible cpus... :)
>>
> 
> Heiko's patches that should fix this problem are now in -mm, so it would 
> be nice to know if there are any existing issues that haven't been 
> addressed yet with your bug report.  See 
> http://www.ozlabs.org/~akpm/mmotm/
> 

Heiko and I both had the same issue. Since some x86 hardware also reaches a lot
of CPUs (hyperthreads included), we bumped the possible number of CPUs to 256 at
least for the 64bit kernel. And that resulted in failed accesses to /proc/stat
when memory became fragmented.
So the first patch will avoid this on most systems. I have not seen this myself,
but I would expect him to be happy with 1/2 already. For really excessive
hardware 2/2 will close the gap.
Since this is no critical bug, I am fine with 3.17, too. I have not done so,
yet, but I could let our reporter try the patches (again, probably not verifying
the second part). Just waited to do so to see whether the code settles down to
these changes.

-Stefan


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 901 bytes --]