Re: [f2fs-dev] [PATCH] f2fs: initialize page->private when using for our internal use

[Jaegeuk Kim]

On 07/05, Matthew Wilcox wrote:
> On Mon, Jul 05, 2021 at 07:33:35PM +0800, Chao Yu wrote:
> > On 2021/7/5 16:56, Jaegeuk Kim wrote:
> > > On 07/05, Chao Yu wrote:
> > > > On 2021/7/5 13:22, Jaegeuk Kim wrote:
> > > > > We need to guarantee it's initially zero. Otherwise, it'll hurt entire flag
> > > > > operations.
> > > > 
> > > > Oops, I didn't get the point, shouldn't .private be zero after page was
> > > > just allocated by filesystem? What's the case we will encounter stall
> > > > private data left in page?
> > > 
> > > I'm seeing f2fs_migrate_page() has the newpage with some value without Private
> > > flag. That causes a kernel panic later due to wrong private flag used in f2fs.
> > 
> > I'm not familiar with that part of codes, so Cc mm mailing list for help.
> > 
> > My question is newpage in .migrate_page() may contain non-zero value in .private
> > field but w/o setting PagePrivate flag, is it a normal case?
> 
> I think freshly allocated pages have a page->private of 0.  ie this
> code in mm/page_alloc.c:
> 
>                 page = rmqueue(ac->preferred_zoneref->zone, zone, order,
>                                 gfp_mask, alloc_flags, ac->migratetype);
>                 if (page) {
>                         prep_new_page(page, order, gfp_mask, alloc_flags);
> 
> where prep_new_page() calls post_alloc_hook() which contains:
>         set_page_private(page, 0);
> 
> Now, I do see in __buffer_migrate_page() (mm/migrate.c):
> 
>         attach_page_private(newpage, detach_page_private(page));
> 
> but as far as I can tell, f2fs doesn't call any of the
> buffer_migrate_page() paths.  So I'm not sure why you're seeing
> a non-zero page->private.

Hmm, I can see it in 4.14 and 5.10 kernel.

The trace is on:

 30875 [ 1065.118750] c3     87  f2fs_migrate_page+0x354/0x45c
 30876 [ 1065.123872] c3     87  move_to_new_page+0x70/0x30c
 30877 [ 1065.128813] c3     87  migrate_pages+0x3a0/0x964
 30878 [ 1065.133583] c3     87  compact_zone+0x608/0xb04
 30879 [ 1065.138257] c3     87  kcompactd+0x378/0x4ec
 30880 [ 1065.142664] c3     87  kthread+0x11c/0x12c
 30881 [ 1065.146897] c3     87  ret_from_fork+0x10/0x18

 It seems compaction_alloc() gets a free page which doesn't reset the fields?