Filipe Manana wrote on 2015/09/14 10:22 +0100:
On Mon, Sep 14, 2015 at 10:08 AM, Qu Wenruo <quwenruo@xxxxxxxxxxxxxx> wrote:Hi Filepe, wrote on 2015/09/14 09:29 +0100:From: Filipe Manana <fdmanana@xxxxxxxx> If a file has a range pointing to a compressed extent, followed by another range that points to the same compressed extent and a read operation attempts to read both ranges (either completely or part of them), the pages that correspond to the second range are incorrectly filled with zeroes. Consider the following example: File layout [0 - 8K] [8K - 24K] | | | | points to extent X, points to extent X, offset 4K, length of 8K offset 0, length 16K [extent X, compressed length = 4K uncompressed length = 16K] If a readpages() call spans the 2 ranges, a single bio to read the extent is submitted - extent_io.c:submit_extent_page() would only create a new bio to cover the second range pointing to the extent if the extent it points to had a different logical address than the extent associated with the first range. This has a consequence of the compressed read end io handler (compression.c:end_compressed_bio_read()) finish once the extent is decompressed into the pages covering the first range, leaving the remaining pages (belonging to the second range) filled with zeroes (done by compression.c:btrfs_clear_biovec_end()). So fix this by submitting the current bio whenever we find a range pointing to a compressed extent that was preceded by a range with a different extent map. This is the simplest solution for this corner case. Making the end io callback populate both ranges (or more, if we have multiple pointing to the same extent) is a much more complex solution since each bio is tightly coupled with a single extent map and the extent maps associated to the ranges pointing to the shared extent can have different offsets and lengths. The following test case for fstests triggers the issue: seq=`basename $0` seqres=$RESULT_DIR/$seq echo "QA output created by $seq" tmp=/tmp/$$ status=1 # failure is the default! trap "_cleanup; exit \$status" 0 1 2 3 15 _cleanup() { rm -f $tmp.* } # get standard environment, filters and checks . ./common/rc . ./common/filter # real QA test starts here _need_to_be_root _supported_fs btrfs _supported_os Linux _require_scratch _require_cloner rm -f $seqres.full test_clone_and_read_compressed_extent() { local mount_opts=$1 _scratch_mkfs >>$seqres.full 2>&1 _scratch_mount $mount_opts # Create a test file with a single extent that is compressed (the # data we write into it is highly compressible no matter which # compression algorithm is used, zlib or lzo). $XFS_IO_PROG -f -c "pwrite -S 0xaa 0K 4K" \ -c "pwrite -S 0xbb 4K 8K" \ -c "pwrite -S 0xcc 12K 4K" \ $SCRATCH_MNT/foo | _filter_xfs_io # Now clone our extent into an adjacent offset. $CLONER_PROG -s $((4 * 1024)) -d $((16 * 1024)) -l $((8 * 1024)) \ $SCRATCH_MNT/foo $SCRATCH_MNT/foo # Same as before but for this file we clone the extent into a lower # file offset. $XFS_IO_PROG -f -c "pwrite -S 0xaa 8K 4K" \ -c "pwrite -S 0xbb 12K 8K" \ -c "pwrite -S 0xcc 20K 4K" \ $SCRATCH_MNT/bar | _filter_xfs_io $CLONER_PROG -s $((12 * 1024)) -d 0 -l $((8 * 1024)) \ $SCRATCH_MNT/bar $SCRATCH_MNT/bar echo "File digests before unmounting filesystem:" md5sum $SCRATCH_MNT/foo | _filter_scratch md5sum $SCRATCH_MNT/bar | _filter_scratch # Evicting the inode or clearing the page cache before reading # again the file would also trigger the bug - reads were returning # all bytes in the range corresponding to the second reference to # the extent with a value of 0, but the correct data was persisted # (it was a bug exclusively in the read path). The issue happened # only if the same readpages() call targeted pages belonging to the # first and second ranges that point to the same compressed extent. _scratch_remount echo "File digests after mounting filesystem again:" # Must match the same digests we got before. md5sum $SCRATCH_MNT/foo | _filter_scratch md5sum $SCRATCH_MNT/bar | _filter_scratch } echo -e "\nTesting with zlib compression..." test_clone_and_read_compressed_extent "-o compress=zlib" _scratch_unmount echo -e "\nTesting with lzo compression..." test_clone_and_read_compressed_extent "-o compress=lzo" status=0 exit Cc: stable@xxxxxxxxxxxxxxx Signed-off-by: Filipe Manana <fdmanana@xxxxxxxx>Thanks for the fix. Most of it seems good for me. Feel free to add "Reviewed-by: Qu Wenruo<quwenruo@xxxxxxxxxxxxxx>" But as mentioned by yourself, I personally consider the patch more as a hot fix, other than a root fix. Although it's good enough for now. One of the problem is the impact on read performance.Most likely the block layer will do one read only of the extent often enough (i.e. multiple bios against same extent result in a single disk read). My basic testing didn't show actually any difference, but I didn't worry about that so much to be honest. I rather have it simple, correct and tested as of now. And this is a corner case that happens not only if the shared extents are consecutive in the file layout but they need to be large as well (a single readpages() call won't attempt to read 1Gb for e.g.). Lets worry first on correctness and decent testing. A lot of problems would likely be prevented if we pay more attention to these two aspects first.It may not affect many people, but for people using off-band deduplication and compression, it may bring obvious read performance impact. Although, without the patch, user won't read correct contents anyway. I think the problem will become more obvious if one day btrfs supports in-band dedup. I know it's not a easy job, but would you please consider some further fix? For example, use list/rb_tree to record the (offset,len) reference to a compressed extent, and still submit only one bio for the max needed range.As mentioned above, getting some numbers/tests that really show it's worth the added complexity (with the possibility of introducing more bugs) would have to happen first.And in your case: (With a little modification, decompressed length is 32K now) File layout [0 - 8K] [8K - 24K] | | | | points to extent X, points to extent X, offset 4K, length of 8K offset 0, length 16K [extent X, compressed length = 4K uncompressed length = 32K] We want to read [4K,12K) and [0,16K) of uncompressed extent X, then the real needed max range will be [0,16K). So submit one bio to read the extent, then extract the first 16K. Then copy [4K,12K) into pages for first extent, and [0,16K) for 2nd. Any idea for such enhancement?Yes, but I really didn't like the complexity of such solution (I tried something similar) and didn't saw any significant overhead in my testing that justifies it. thanks
Right, I didn't take block layer into consideration. That's what block layer should do and it has already done it well.So I'm completely OK with your fix, especially for your nice and clear comment in the code.
Thanks, Qu
Thanks, Qu--- fs/btrfs/extent_io.c | 65 +++++++++++++++++++++++++++++++++++++++++++++------- 1 file changed, 57 insertions(+), 8 deletions(-) diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index fa19f2f..11aa8f7 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -2805,7 +2805,8 @@ static int submit_extent_page(int rw, struct extent_io_tree *tree, bio_end_io_t end_io_func, int mirror_num, unsigned long prev_bio_flags, - unsigned long bio_flags) + unsigned long bio_flags, + bool force_bio_submit) { int ret = 0; struct bio *bio; @@ -2823,6 +2824,7 @@ static int submit_extent_page(int rw, struct extent_io_tree *tree, contig = bio_end_sector(bio) == sector; if (prev_bio_flags != bio_flags || !contig || + force_bio_submit || merge_bio(rw, tree, page, offset, page_size, bio, bio_flags) || bio_add_page(bio, page, page_size, offset) < page_size) { ret = submit_one_bio(rw, bio, mirror_num, @@ -2922,7 +2924,8 @@ static int __do_readpage(struct extent_io_tree *tree, get_extent_t *get_extent, struct extent_map **em_cached, struct bio **bio, int mirror_num, - unsigned long *bio_flags, int rw) + unsigned long *bio_flags, int rw, + u64 *prev_em_start) { struct inode *inode = page->mapping->host; u64 start = page_offset(page); @@ -2970,6 +2973,7 @@ static int __do_readpage(struct extent_io_tree *tree, } while (cur <= end) { unsigned long pnr = (last_byte >> PAGE_CACHE_SHIFT) + 1; + bool force_bio_submit = false; if (cur >= last_byte) { char *userpage; @@ -3020,6 +3024,49 @@ static int __do_readpage(struct extent_io_tree *tree, block_start = em->block_start; if (test_bit(EXTENT_FLAG_PREALLOC, &em->flags)) block_start = EXTENT_MAP_HOLE; + + /* + * If we have a file range that points to a compressed extent + * and it's followed by a consecutive file range that points to + * to the same compressed extent (possibly with a different + * offset and/or length, so it either points to the whole extent + * or only part of it), we must make sure we do not submit a + * single bio to populate the pages for the 2 ranges because + * this makes the compressed extent read zero out the pages + * belonging to the 2nd range. Imagine the following scenario: + * + * File layout + * [0 - 8K] [8K - 24K] + * | | + * | | + * points to extent X, points to extent X, + * offset 4K, length of 8K offset 0, length 16K + * + * [extent X, compressed length = 4K uncompressed length = 16K] + * + * If the bio to read the compressed extent covers both ranges, + * it will decompress extent X into the pages belonging to the + * first range and then it will stop, zeroing out the remaining + * pages that belong to the other range that points to extent X. + * So here we make sure we submit 2 bios, one for the first + * range and another one for the third range. Both will target + * the same physical extent from disk, but we can't currently + * make the compressed bio endio callback populate the pages + * for both ranges because each compressed bio is tightly + * coupled with a single extent map, and each range can have + * an extent map with a different offset value relative to the + * uncompressed data of our extent and different lengths. This + * is a corner case so we prioritize correctness over + * non-optimal behavior (submitting 2 bios for the same extent). + */ + if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags) && + prev_em_start && *prev_em_start != (u64)-1 && + *prev_em_start != em->orig_start) + force_bio_submit = true; + + if (prev_em_start) + *prev_em_start = em->orig_start; + free_extent_map(em); em = NULL; @@ -3069,7 +3116,8 @@ static int __do_readpage(struct extent_io_tree *tree, bdev, bio, pnr, end_bio_extent_readpage, mirror_num, *bio_flags, - this_bio_flag); + this_bio_flag, + force_bio_submit); if (!ret) { nr++; *bio_flags = this_bio_flag; @@ -3101,6 +3149,7 @@ static inline void __do_contiguous_readpages(struct extent_io_tree *tree, struct inode *inode; struct btrfs_ordered_extent *ordered; int index; + u64 prev_em_start = (u64)-1; inode = pages[0]->mapping->host; while (1) { @@ -3116,7 +3165,7 @@ static inline void __do_contiguous_readpages(struct extent_io_tree *tree, for (index = 0; index < nr_pages; index++) { __do_readpage(tree, pages[index], get_extent, em_cached, bio, - mirror_num, bio_flags, rw); + mirror_num, bio_flags, rw, &prev_em_start); page_cache_release(pages[index]); } } @@ -3184,7 +3233,7 @@ static int __extent_read_full_page(struct extent_io_tree *tree, } ret = __do_readpage(tree, page, get_extent, NULL, bio, mirror_num, - bio_flags, rw); + bio_flags, rw, NULL); return ret; } @@ -3210,7 +3259,7 @@ int extent_read_full_page_nolock(struct extent_io_tree *tree, struct page *page, int ret; ret = __do_readpage(tree, page, get_extent, NULL, &bio, mirror_num, - &bio_flags, READ); + &bio_flags, READ, NULL); if (bio) ret = submit_one_bio(READ, bio, mirror_num, bio_flags); return ret; @@ -3463,7 +3512,7 @@ static noinline_for_stack int __extent_writepage_io(struct inode *inode, sector, iosize, pg_offset, bdev, &epd->bio, max_nr, end_bio_extent_writepage, - 0, 0, 0); + 0, 0, 0, false); if (ret) SetPageError(page); } @@ -3765,7 +3814,7 @@ static noinline_for_stack int write_one_eb(struct extent_buffer *eb, ret = submit_extent_page(rw, tree, wbc, p, offset >> 9, PAGE_CACHE_SIZE, 0, bdev, &epd->bio, -1, end_bio_extent_buffer_writepage, - 0, epd->bio_flags, bio_flags); + 0, epd->bio_flags, bio_flags, false); epd->bio_flags = bio_flags; if (ret) { set_btree_ioerr(p);
-- To unsubscribe from this list: send the line "unsubscribe stable" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html