On Fri, Oct 18, 2019 at 03:55:13PM -0700, Omar Sandoval wrote:
> On Wed, Oct 16, 2019 at 01:44:56PM +0300, Nikolay Borisov wrote:
> >
> >
> > On 15.10.19 г. 21:42 ч., Omar Sandoval wrote:
> > > From: Omar Sandoval <osandov@xxxxxx>
> > >
> > > The implementation resembles direct I/O: we have to flush any ordered
> > > extents, invalidate the page cache, and do the io tree/delalloc/extent
> > > map/ordered extent dance. From there, we can reuse the compression code
> > > with a minor modification to distinguish the write from writeback.
> > >
> > > Now that read and write are implemented, this also sets the
> > > FMODE_ENCODED_IO flag in btrfs_file_open().
> > >
> > > Signed-off-by: Omar Sandoval <osandov@xxxxxx>
> > > ---
> > > fs/btrfs/compression.c | 6 +-
> > > fs/btrfs/compression.h | 5 +-
> > > fs/btrfs/ctree.h | 2 +
> > > fs/btrfs/file.c | 40 +++++++--
> > > fs/btrfs/inode.c | 197 ++++++++++++++++++++++++++++++++++++++++-
> > > 5 files changed, 237 insertions(+), 13 deletions(-)
> > >
[snip]
> > > + for (;;) {
> > > + struct btrfs_ordered_extent *ordered;
> > > +
> > > + ret = btrfs_wait_ordered_range(inode, start, end - start + 1);
> > > + if (ret)
> > > + goto out_pages;
> > > + ret = invalidate_inode_pages2_range(inode->i_mapping,
> > > + start >> PAGE_SHIFT,
> > > + end >> PAGE_SHIFT);
> > > + if (ret)
> > > + goto out_pages;
> > > + lock_extent_bits(io_tree, start, end, &cached_state);
> > > + ordered = btrfs_lookup_ordered_range(BTRFS_I(inode), start,
> > > + end - start + 1);
> > > + if (!ordered &&
> > > + !filemap_range_has_page(inode->i_mapping, start, end))
> > > + break;
> > > + if (ordered)
> > > + btrfs_put_ordered_extent(ordered);
> > > + unlock_extent_cached(io_tree, start, end, &cached_state);
> > > + cond_resched();
> > > + }
> > > +
> > > + ret = btrfs_delalloc_reserve_space(inode, &data_reserved, start,
> > > + num_bytes);
> > > + if (ret)
> > > + goto out_unlock;
> > > +
> > > + ret = btrfs_reserve_extent(root, num_bytes, disk_num_bytes,
> > > + disk_num_bytes, 0, 0, &ins, 1, 1);
> > > + if (ret)
> > > + goto out_delalloc_release;
> > > +
> > > + em = create_io_em(inode, start, num_bytes, start, ins.objectid,
> > > + ins.offset, ins.offset, num_bytes, compression,
> > > + BTRFS_ORDERED_COMPRESSED);
> > > + if (IS_ERR(em)) {
> > > + ret = PTR_ERR(em);
> > > + goto out_free_reserve;
> > > + }
> > > + free_extent_map(em);
> > > +
> > > + ret = btrfs_add_ordered_extent_compress(inode, start, ins.objectid,
> > > + num_bytes, ins.offset,
> > > + BTRFS_ORDERED_COMPRESSED,
> > > + compression);
> > > + if (ret) {
> > > + btrfs_drop_extent_cache(BTRFS_I(inode), start, end, 0);
> > > + goto out_free_reserve;
> > > + }
> > > + btrfs_dec_block_group_reservations(fs_info, ins.objectid);
> > > +
> > > + if (start + encoded->len > inode->i_size)
> > > + i_size_write(inode, start + encoded->len);
> >
> > Don't we want the inode size to be updated once data hits disk and
> > btrfs_finish_ordered_io is called?
>
> Yup, you're right, this is too early.
Actually, no, this part is fine. Compare to the call to i_size_write()
in btrfs_get_blocks_direct_write(): we lock the extent in the io_tree,
create the ordered extent, update i_size, then unlock the extent. Anyone
else who comes in is going to find the ordered extent and wait on that.
> > > +
> > > + unlock_extent_cached(io_tree, start, end, &cached_state);
> > > +
> > > + btrfs_delalloc_release_extents(BTRFS_I(inode), num_bytes, false);
> > > +
> > > + if (btrfs_submit_compressed_write(inode, start, num_bytes, ins.objectid,
> > > + ins.offset, pages, nr_pages, 0,
> > > + false)) {
> > > + struct page *page = pages[0];
> > > +
> > > + page->mapping = inode->i_mapping;
> > > + btrfs_writepage_endio_finish_ordered(page, start, end, 0);
> > > + page->mapping = NULL;
> > > + ret = -EIO;
> > > + goto out_pages;
> > > + }
>
> I also need to wait for the I/O to finish here.
>
> > > + iocb->ki_pos += encoded->len;
> > > + return orig_count;
> > > +
> > > +out_free_reserve:
> > > + btrfs_dec_block_group_reservations(fs_info, ins.objectid);
> > > + btrfs_free_reserved_extent(fs_info, ins.objectid, ins.offset, 1);
> > > +out_delalloc_release:
> > > + btrfs_delalloc_release_space(inode, data_reserved, start, num_bytes,
> > > + true);
> > > +out_unlock:
> > > + unlock_extent_cached(io_tree, start, end, &cached_state);
> > > +out_pages:
> > > + for (i = 0; i < nr_pages; i++) {
> > > + if (pages[i])
> > > + put_page(pages[i]);
> > > + }
> > > + kvfree(pages);
> > > + return ret;
> > > +}
> > > +
> > > #ifdef CONFIG_SWAP
> > > /*
> > > * Add an entry indicating a block group or device which is pinned by a
> > >
