Re: [RFC/PATCH] - revert generic_fillattr stat->blksize to PAGE_CACHE_SIZE

[Eric Sandeen <sandeen@redhat.com>]

Andreas Dilger wrote:
> On Nov 06, 2006  17:15 -0600, Eric Sandeen wrote:
>> Christoph Hellwig wrote:
>>> I agree with the conclusion, but the patch is incomplete.  You went down
>>> all the way to find out what the fileystems do in this messages, so add
>>> the hunks to override the defaults for non-standard filesystems to the
>>> patch aswell to restore the pre-inode diet state.
>> Well, agreed.  I put 80% or more back to pre-patch state, but not all.
>> :)  So it's less broken with my patch than without, so at least it's
>> moving forward.  So... Ted's patches get in w/o fixing up all the other
>> filesystems (left as an exercise to the patch reader) but mine can't? :)
> 
> Actually, rather than blindly revert to pre-patch behaviour it would be
> worthwhile to determine if PAGE_SIZE isn't the better value.  In some
> cases people don't understand that i_blksize is the "optimal IO size"
> and instead assume it is the filesystem blocksize.  I saw a few that were
> e.g. 512 and that can't be very useful.

I'm willing to either revert everyting to pre-inode-diet behavior, or leave it 
at the (newly re-proposed) page size default and let the other fs maintainers 
sort it out for their own codebase, but I don't pretend to know what is best 
for, say, qnx4 etc...  I'd be willing to cc: all maintainers asking them to take 
another long hard look at their code. :)

As we saw with cifs, these changes can have unintended consequences (not picking 
on cifs, it's just one that ran into issues with the broad-stroke change).

-Eric