Re: [PATCH v12 00/10] enable bs > ps in XFS

["Pankaj Raghav (Samsung)" <kernel@pankajraghav.com>]

On Mon, Aug 19, 2024 at 09:24:11AM +0200, Hannes Reinecke wrote:
> On 8/18/24 22:16, David Howells wrote:
> > Pankaj Raghav (Samsung) <kernel@pankajraghav.com> wrote:
> > 
> > > I am no expert in network filesystems but are you sure there are no
> > > PAGE_SIZE assumption when manipulating folios from the page cache in
> > > AFS?
> > 
> > Note that I've removed the knowledge of the pagecache from 9p, afs and cifs to
> > netfslib and intend to do the same to ceph.  The client filesystems just
> > provide read and write ops to netfslib and netfslib uses those to do ordinary
> > buffered I/O, unbuffered I/O (selectable by mount option on some filesystems)
> > and DIO.
> > 
> > That said, I'm not sure that I haven't made some PAGE_SIZE assumptions.  I
> > don't *think* I have since netfslib is fully multipage folio capable, but I
> > can't guarantee it.
> > 
> I guess you did:
> 
> static int afs_fill_super(struct super_block *sb, struct afs_fs_context
> *ctx)
> {
>         struct afs_super_info *as = AFS_FS_S(sb);
>         struct inode *inode = NULL;
>         int ret;
> 
>         _enter("");
> 
>         /* fill in the superblock */
>         sb->s_blocksize         = PAGE_SIZE;
>         sb->s_blocksize_bits    = PAGE_SHIFT;
>         sb->s_maxbytes          = MAX_LFS_FILESIZE;
>         sb->s_magic             = AFS_FS_MAGIC;
>         sb->s_op                = &afs_super_ops;
> 
> IE you essentially nail AFS to use PAGE_SIZE.
> Not sure how you would tell AFS to use a different block size;
> maybe a mount option?

I saw this as well, but I didn't see this variable being used anywhere.
Probably this has no meaning in a network-based FSs?

> And there are several other places which will need to be modified;
> eg afs_mntpt_set_params() is trying to read from a page which
> won't fly with large blocks (converted to read_full_folio()?),
> and, of course, the infamous AFS_DIR_BLOCKS_PER_PAGE which will
> overflow for large blocks.

But the min folio order is set only for AFS_FTYPE_FILE and not
for AFS_FTYPE_DIR.

> 
> So some work is required, but everything looks doable.
> Maybe I can find some time until LPC.
> 
> > Mostly this was just a note to you that there might be an issue with your code
> > - but I haven't investigated it yet and it could well be in my code.
> > 
> Hmm. I'd rather fix the obvious places in afs first; just do a quick
> grep for 'PAGE_', that'll give you a good impression of places to look at.
> 
Agree.

> Cheers,
> 
> Hannes
> -- 
> Dr. Hannes Reinecke                  Kernel Storage Architect
> hare@suse.de                                +49 911 74053 688
> SUSE Software Solutions GmbH, Frankenstr. 146, 90461 Nürnberg
> HRB 36809 (AG Nürnberg), GF: I. Totev, A. McDonald, W. Knoblich
>