[PATCH v4] iomap: support tail packing inline read

[Andreas Grünbacher]

Am Di., 20. Juli 2021 um 23:19 Uhr schrieb Matthew Wilcox <willy at infradead.org>:
> On Tue, Jul 20, 2021 at 01:42:24PM -0700, Darrick J. Wong wrote:
> > > -   BUG_ON(page_has_private(page));
> > > -   BUG_ON(page->index);
> > > -   BUG_ON(size > PAGE_SIZE - offset_in_page(iomap->inline_data));
> > > +   /* inline source data must be inside a single page */
> > > +   BUG_ON(iomap->length > PAGE_SIZE - offset_in_page(iomap->inline_data));
> >
> > Can we reduce the strength of these checks to a warning and an -EIO
> > return?

Yes, we could do that.

> I'm not entirely sure that we need this check, tbh.

I wanted to make sure the memcpy / memset won't corrupt random kernel
memory when the filesystem gets the iomap_begin wrong.

> > > +   /* handle tail-packing blocks cross the current page into the next */
> > > +   size = min_t(unsigned int, iomap->length + pos - iomap->offset,
> > > +                PAGE_SIZE - poff);
> > >
> > >     addr = kmap_atomic(page);
> > > -   memcpy(addr, iomap->inline_data, size);
> > > -   memset(addr + size, 0, PAGE_SIZE - size);
> > > +   memcpy(addr + poff, iomap->inline_data - iomap->offset + pos, size);
> > > +   memset(addr + poff + size, 0, PAGE_SIZE - poff - size);
> >
> > Hmm, so I guess the point of this is to support reading data from a
> > tail-packing block, where each file gets some arbitrary byte range
> > within the tp-block, and the range isn't aligned to an fs block?  Hence
> > you have to use the inline data code to read the relevant bytes and copy
> > them into the pagecache?
>
> I think there are two distinct cases for IOMAP_INLINE.  One is
> where the tail of the file is literally embedded into the inode.
> Like ext4 fast symbolic links.  Taking the ext4 i_blocks layout
> as an example, you could have a 4kB block stored in i_block[0]
> and then store bytes 4096-4151 in i_block[1-14] (although reading
> https://www.kernel.org/doc/html/latest/filesystems/ext4/dynamic.html
> makes me think that ext4 only supports storing 0-59 in the i_blocks;
> it doesn't support 0-4095 in i_block[0] and then 4096-4151 in i_blocks)
>
> The other is what I think erofs is doing where, for example, you'd
> specify in i_block[1] the block which contains the tail and then in
> i_block[2] what offset of the block the tail starts at.

Andreas