On Mon, Apr 09, 2018 at 03:52:15PM +0200, Michal Hocko wrote:
> On Mon 09-04-18 06:41:14, Matthew Wilcox wrote:
> > On Mon, Apr 09, 2018 at 02:48:52PM +0200, Michal Hocko wrote:
> > > On Mon 09-04-18 20:25:06, Chao Yu wrote:
> > > [...]
> > > > diff --git a/fs/f2fs/inode.c b/fs/f2fs/inode.c
> > > > index c85cccc2e800..cc63f8c448f0 100644
> > > > --- a/fs/f2fs/inode.c
> > > > +++ b/fs/f2fs/inode.c
> > > > @@ -339,10 +339,10 @@ struct inode *f2fs_iget(struct super_block *sb, unsigned long ino)
> > > > make_now:
> > > > if (ino == F2FS_NODE_INO(sbi)) {
> > > > inode->i_mapping->a_ops = &f2fs_node_aops;
> > > > - mapping_set_gfp_mask(inode->i_mapping, GFP_F2FS_ZERO);
> > > > + mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
> > >
> > > An unrelated question. Why do you make all allocations for the mapping
> > > NOFS automatically? What kind of reclaim recursion problems are you
> > > trying to prevent?
> >
> > It's worth noting that this is endemic in filesystems.
>
> Yes, and I have strong suspicion that this is a mindless copy&pasting...
> Well, xfs had a good reason for it in the past - mostly to handle deep
> call stacks on complicated storage setups in the past when we used to
> trigger IO from the direct reclaim. I am not sure whether there are
> other reasons to keep the status quo except for finding somebody brave
> enough to post the patch, do all the due testing.
>
> > $ git grep mapping_set_gfp_mask.*FS
> > drivers/block/loop.c: mapping_set_gfp_mask(mapping, lo->old_gfp_mask & ~(__GFP_IO|__GFP_FS));
> > fs/btrfs/disk-io.c: mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
Thi was added in 1561deda687eef0
(https://git.kernel.org/linus/1561deda687eef0e9506) and probably after a
deadlock report.
The changelog mentions the potential recursion from fs -> allocation -> fs,
but I'm not sure if this still happens on the MM side today.
For the filesystem part, I think the key functions of the callchain are
still there.
The code was been added in 2011 and the 2nd hunk of the patch added a
code that's not present today AFAICS, so this is worth revisiting.
I still don't understand how it's related to the GFP_HIGHUSER_MOVABLE,
this patch is from time where the metadata pages were possibly allocated
from HIGHMEM but this was removed later in a65917156e345946db
(https://git.kernel.org/linus/a65917156e345946db).
