Re: [PATCH v2] f2fs: use kmem_cache pool during inline xattr lookups

[Jaegeuk Kim]

On 03/19, Chao Yu wrote:
> Hi Ju Hyung,
> 
> On 2020/3/18 20:14, Ju Hyung Park wrote:
> > Hi Chao.
> > 
> > I got the time around to test this patch.
> > The v2 patch seems to work just fine, and the code looks good.
> 
> Thanks a lot for the review and test.
> 
> > 
> > On Tue, Feb 25, 2020 at 7:17 PM Chao Yu <yuchao0@xxxxxxxxxx> wrote:
> >> diff --git a/fs/f2fs/xattr.c b/fs/f2fs/xattr.c
> >> index a3360a97e624..e46a10eb0e42 100644
> >> --- a/fs/f2fs/xattr.c
> >> +++ b/fs/f2fs/xattr.c
> >> @@ -23,6 +23,25 @@
> >>  #include "xattr.h"
> >>  #include "segment.h"
> >>
> >> +static void *xattr_alloc(struct f2fs_sb_info *sbi, int size, bool *is_inline)
> >> +{
> >> +       *is_inline = (size == sbi->inline_xattr_slab_size);
> > 
> > Would it be meaningless to change this to the following code?
> > if (likely(size == sbi->inline_xattr_slab_size))
> >     *is_inline = true;
> > else
> >     *is_inline = false;
> 
> Yup, I guess it's very rare that user will change inline xattr size via remount,
> so I'm okay with this change.

Applied like this. Thanks,

 26 static void *xattr_alloc(struct f2fs_sb_info *sbi, int size, bool *is_inline)
 27 {
 28         if (likely(size == sbi->inline_xattr_slab_size)) {
 29                 *is_inline = true;
 30                 return kmem_cache_zalloc(sbi->inline_xattr_slab, GFP_NOFS);
 31         }
 32         *is_inline = false;
 33         return f2fs_kzalloc(sbi, size, GFP_NOFS);
 34 }

> 
> Jaegeuk,
> 
> Could you please help to update the patch in your git tree directly?
> 
> Thanks,
> 
> > 
> > The above statement seems to be only false during the initial mount
> > and the rest(millions) seems to be always true.
> > 
> > Thanks.
> > .
> >