From eb838e73dc2121d2bae47d5678952cd7d48793b5 Mon Sep 17 00:00:00 2001 From: Josef Bacik Date: Tue, 31 Jul 2012 16:28:48 -0400 Subject: Btrfs: lock extents as we map them in DIO A deadlock in xfstests 113 was uncovered by commit d187663ef24cd3d033f0cbf2867e70b36a3a90b8 This is because we would not return EIOCBQUEUED for short AIO reads, instead we'd wait for the DIO to complete and then return the amount of data we transferred, which would allow our stuff to unlock the remaning amount. But with this change this no longer happens, so if we have a short AIO read (for example if we try to read past EOF), we could leave the section from EOF to the end of where we tried to read locked. Fixing this is tricky since there is no clear way to know exactly how much data DIO truly submitted for IO, so to make this less hard on ourselves and less combersome we need to lock the extents as we try to map them, and then we unlock any areas we didn't actually map. This makes us completely safe from deadlocks and reliance on a particular behavior of the DIO code. This also lays the groundwork for allowing us to use the normal csum storage method for reads which means we can remove an allocation. Thanks, Signed-off-by: Josef Bacik --- fs/btrfs/inode.c | 256 +++++++++++++++++++++++++++---------------------------- 1 file changed, 127 insertions(+), 129 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index dac1fc21d80..09182449cbd 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -5773,18 +5773,109 @@ out: return ret; } +static int lock_extent_direct(struct inode *inode, u64 lockstart, u64 lockend, + struct extent_state **cached_state, int writing) +{ + struct btrfs_ordered_extent *ordered; + int ret = 0; + + while (1) { + lock_extent_bits(&BTRFS_I(inode)->io_tree, lockstart, lockend, + 0, cached_state); + /* + * We're concerned with the entire range that we're going to be + * doing DIO to, so we need to make sure theres no ordered + * extents in this range. + */ + ordered = btrfs_lookup_ordered_range(inode, lockstart, + lockend - lockstart + 1); + + /* + * We need to make sure there are no buffered pages in this + * range either, we could have raced between the invalidate in + * generic_file_direct_write and locking the extent. The + * invalidate needs to happen so that reads after a write do not + * get stale data. + */ + if (!ordered && (!writing || + !test_range_bit(&BTRFS_I(inode)->io_tree, + lockstart, lockend, EXTENT_UPTODATE, 0, + *cached_state))) + break; + + unlock_extent_cached(&BTRFS_I(inode)->io_tree, lockstart, lockend, + cached_state, GFP_NOFS); + + if (ordered) { + btrfs_start_ordered_extent(inode, ordered, 1); + btrfs_put_ordered_extent(ordered); + } else { + /* Screw you mmap */ + ret = filemap_write_and_wait_range(inode->i_mapping, + lockstart, + lockend); + if (ret) + break; + + /* + * If we found a page that couldn't be invalidated just + * fall back to buffered. + */ + ret = invalidate_inode_pages2_range(inode->i_mapping, + lockstart >> PAGE_CACHE_SHIFT, + lockend >> PAGE_CACHE_SHIFT); + if (ret) + break; + } + + cond_resched(); + } + + return ret; +} + static int btrfs_get_blocks_direct(struct inode *inode, sector_t iblock, struct buffer_head *bh_result, int create) { struct extent_map *em; struct btrfs_root *root = BTRFS_I(inode)->root; + struct extent_state *cached_state = NULL; u64 start = iblock << inode->i_blkbits; + u64 lockstart, lockend; u64 len = bh_result->b_size; struct btrfs_trans_handle *trans; + int unlock_bits = EXTENT_LOCKED; + int ret; + + lockstart = start; + lockend = start + len - 1; + if (create) { + ret = btrfs_delalloc_reserve_space(inode, len); + if (ret) + return ret; + unlock_bits |= EXTENT_DELALLOC | EXTENT_DIRTY; + } + + /* + * If this errors out it's because we couldn't invalidate pagecache for + * this range and we need to fallback to buffered. + */ + if (lock_extent_direct(inode, lockstart, lockend, &cached_state, create)) + return -ENOTBLK; + + if (create) { + ret = set_extent_bit(&BTRFS_I(inode)->io_tree, lockstart, + lockend, EXTENT_DELALLOC, NULL, + &cached_state, GFP_NOFS); + if (ret) + goto unlock_err; + } em = btrfs_get_extent(inode, NULL, 0, start, len, 0); - if (IS_ERR(em)) - return PTR_ERR(em); + if (IS_ERR(em)) { + ret = PTR_ERR(em); + goto unlock_err; + } /* * Ok for INLINE and COMPRESSED extents we need to fallback on buffered @@ -5803,17 +5894,16 @@ static int btrfs_get_blocks_direct(struct inode *inode, sector_t iblock, if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags) || em->block_start == EXTENT_MAP_INLINE) { free_extent_map(em); - return -ENOTBLK; + ret = -ENOTBLK; + goto unlock_err; } /* Just a good old fashioned hole, return */ if (!create && (em->block_start == EXTENT_MAP_HOLE || test_bit(EXTENT_FLAG_PREALLOC, &em->flags))) { free_extent_map(em); - /* DIO will do one hole at a time, so just unlock a sector */ - unlock_extent(&BTRFS_I(inode)->io_tree, start, - start + root->sectorsize - 1); - return 0; + ret = 0; + goto unlock_err; } /* @@ -5826,8 +5916,9 @@ static int btrfs_get_blocks_direct(struct inode *inode, sector_t iblock, * */ if (!create) { - len = em->len - (start - em->start); - goto map; + len = min(len, em->len - (start - em->start)); + lockstart = start + len; + goto unlock; } if (test_bit(EXTENT_FLAG_PREALLOC, &em->flags) || @@ -5859,7 +5950,7 @@ static int btrfs_get_blocks_direct(struct inode *inode, sector_t iblock, btrfs_end_transaction(trans, root); if (ret) { free_extent_map(em); - return ret; + goto unlock_err; } goto unlock; } @@ -5872,14 +5963,12 @@ must_cow: */ len = bh_result->b_size; em = btrfs_new_extent_direct(inode, em, start, len); - if (IS_ERR(em)) - return PTR_ERR(em); + if (IS_ERR(em)) { + ret = PTR_ERR(em); + goto unlock_err; + } len = min(len, em->len - (start - em->start)); unlock: - clear_extent_bit(&BTRFS_I(inode)->io_tree, start, start + len - 1, - EXTENT_LOCKED | EXTENT_DELALLOC | EXTENT_DIRTY, 1, - 0, NULL, GFP_NOFS); -map: bh_result->b_blocknr = (em->block_start + (start - em->start)) >> inode->i_blkbits; bh_result->b_size = len; @@ -5897,9 +5986,28 @@ map: i_size_write(inode, start + len); } + /* + * In the case of write we need to clear and unlock the entire range, + * in the case of read we need to unlock only the end area that we + * aren't using if there is any left over space. + */ + if (lockstart < lockend) + clear_extent_bit(&BTRFS_I(inode)->io_tree, lockstart, lockend, + unlock_bits, 1, 0, &cached_state, GFP_NOFS); + else + free_extent_state(cached_state); + free_extent_map(em); return 0; + +unlock_err: + if (create) + unlock_bits |= EXTENT_DO_ACCOUNTING; + + clear_extent_bit(&BTRFS_I(inode)->io_tree, lockstart, lockend, + unlock_bits, 1, 0, &cached_state, GFP_NOFS); + return ret; } struct btrfs_dio_private { @@ -6340,132 +6448,22 @@ static ssize_t check_direct_IO(struct btrfs_root *root, int rw, struct kiocb *io out: return retval; } + static ssize_t btrfs_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov, loff_t offset, unsigned long nr_segs) { struct file *file = iocb->ki_filp; struct inode *inode = file->f_mapping->host; - struct btrfs_ordered_extent *ordered; - struct extent_state *cached_state = NULL; - u64 lockstart, lockend; - ssize_t ret; - int writing = rw & WRITE; - int write_bits = 0; - size_t count = iov_length(iov, nr_segs); if (check_direct_IO(BTRFS_I(inode)->root, rw, iocb, iov, - offset, nr_segs)) { + offset, nr_segs)) return 0; - } - - lockstart = offset; - lockend = offset + count - 1; - - if (writing) { - ret = btrfs_delalloc_reserve_space(inode, count); - if (ret) - goto out; - } - while (1) { - lock_extent_bits(&BTRFS_I(inode)->io_tree, lockstart, lockend, - 0, &cached_state); - /* - * We're concerned with the entire range that we're going to be - * doing DIO to, so we need to make sure theres no ordered - * extents in this range. - */ - ordered = btrfs_lookup_ordered_range(inode, lockstart, - lockend - lockstart + 1); - - /* - * We need to make sure there are no buffered pages in this - * range either, we could have raced between the invalidate in - * generic_file_direct_write and locking the extent. The - * invalidate needs to happen so that reads after a write do not - * get stale data. - */ - if (!ordered && (!writing || - !test_range_bit(&BTRFS_I(inode)->io_tree, - lockstart, lockend, EXTENT_UPTODATE, 0, - cached_state))) - break; - - unlock_extent_cached(&BTRFS_I(inode)->io_tree, lockstart, lockend, - &cached_state, GFP_NOFS); - - if (ordered) { - btrfs_start_ordered_extent(inode, ordered, 1); - btrfs_put_ordered_extent(ordered); - } else { - /* Screw you mmap */ - ret = filemap_write_and_wait_range(file->f_mapping, - lockstart, - lockend); - if (ret) - goto out; - - /* - * If we found a page that couldn't be invalidated just - * fall back to buffered. - */ - ret = invalidate_inode_pages2_range(file->f_mapping, - lockstart >> PAGE_CACHE_SHIFT, - lockend >> PAGE_CACHE_SHIFT); - if (ret) { - if (ret == -EBUSY) - ret = 0; - goto out; - } - } - - cond_resched(); - } - - /* - * we don't use btrfs_set_extent_delalloc because we don't want - * the dirty or uptodate bits - */ - if (writing) { - write_bits = EXTENT_DELALLOC | EXTENT_DO_ACCOUNTING; - ret = set_extent_bit(&BTRFS_I(inode)->io_tree, lockstart, lockend, - EXTENT_DELALLOC, NULL, &cached_state, - GFP_NOFS); - if (ret) { - clear_extent_bit(&BTRFS_I(inode)->io_tree, lockstart, - lockend, EXTENT_LOCKED | write_bits, - 1, 0, &cached_state, GFP_NOFS); - goto out; - } - } - - free_extent_state(cached_state); - cached_state = NULL; - - ret = __blockdev_direct_IO(rw, iocb, inode, + return __blockdev_direct_IO(rw, iocb, inode, BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev, iov, offset, nr_segs, btrfs_get_blocks_direct, NULL, btrfs_submit_direct, 0); - - if (ret < 0 && ret != -EIOCBQUEUED) { - clear_extent_bit(&BTRFS_I(inode)->io_tree, offset, - offset + iov_length(iov, nr_segs) - 1, - EXTENT_LOCKED | write_bits, 1, 0, - &cached_state, GFP_NOFS); - } else if (ret >= 0 && ret < iov_length(iov, nr_segs)) { - /* - * We're falling back to buffered, unlock the section we didn't - * do IO on. - */ - clear_extent_bit(&BTRFS_I(inode)->io_tree, offset + ret, - offset + iov_length(iov, nr_segs) - 1, - EXTENT_LOCKED | write_bits, 1, 0, - &cached_state, GFP_NOFS); - } -out: - free_extent_state(cached_state); - return ret; } static int btrfs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo, -- cgit v1.2.3-70-g09d2 From c329861da40623cd838b8c9ee31a850242fd88cf Mon Sep 17 00:00:00 2001 From: Josef Bacik Date: Fri, 3 Aug 2012 16:49:19 -0400 Subject: Btrfs: don't allocate a seperate csums array for direct reads We've been allocating a big array for csums instead of storing them in the io_tree like we do for buffered reads because previously we were locking the entire range, so we didn't have an extent state for each sector of the range. But now that we do the range locking as we map the buffers we can limit the mapping lenght to sectorsize and use the private part of the io_tree for our csums. This allows us to avoid an extra memory allocation for direct reads which could incur latency. Thanks, Signed-off-by: Josef Bacik --- fs/btrfs/ctree.h | 2 +- fs/btrfs/file-item.c | 4 ++-- fs/btrfs/inode.c | 45 ++++++++++++++++----------------------------- 3 files changed, 19 insertions(+), 32 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index adb1cd7ceb9..348196350bf 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -3192,7 +3192,7 @@ int btrfs_del_csums(struct btrfs_trans_handle *trans, int btrfs_lookup_bio_sums(struct btrfs_root *root, struct inode *inode, struct bio *bio, u32 *dst); int btrfs_lookup_bio_sums_dio(struct btrfs_root *root, struct inode *inode, - struct bio *bio, u64 logical_offset, u32 *dst); + struct bio *bio, u64 logical_offset); int btrfs_insert_file_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 objectid, u64 pos, diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c index b45b9de0c21..857d93cd01d 100644 --- a/fs/btrfs/file-item.c +++ b/fs/btrfs/file-item.c @@ -272,9 +272,9 @@ int btrfs_lookup_bio_sums(struct btrfs_root *root, struct inode *inode, } int btrfs_lookup_bio_sums_dio(struct btrfs_root *root, struct inode *inode, - struct bio *bio, u64 offset, u32 *dst) + struct bio *bio, u64 offset) { - return __btrfs_lookup_bio_sums(root, inode, bio, offset, dst, 1); + return __btrfs_lookup_bio_sums(root, inode, bio, offset, NULL, 1); } int btrfs_lookup_csums_range(struct btrfs_root *root, u64 start, u64 end, diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 09182449cbd..2d65c52b094 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -5847,15 +5847,18 @@ static int btrfs_get_blocks_direct(struct inode *inode, sector_t iblock, int unlock_bits = EXTENT_LOCKED; int ret; - lockstart = start; - lockend = start + len - 1; if (create) { ret = btrfs_delalloc_reserve_space(inode, len); if (ret) return ret; unlock_bits |= EXTENT_DELALLOC | EXTENT_DIRTY; + } else { + len = min_t(u64, len, root->sectorsize); } + lockstart = start; + lockend = start + len - 1; + /* * If this errors out it's because we couldn't invalidate pagecache for * this range and we need to fallback to buffered. @@ -6015,7 +6018,6 @@ struct btrfs_dio_private { u64 logical_offset; u64 disk_bytenr; u64 bytes; - u32 *csums; void *private; /* number of bios pending for this dio */ @@ -6035,7 +6037,6 @@ static void btrfs_endio_direct_read(struct bio *bio, int err) struct inode *inode = dip->inode; struct btrfs_root *root = BTRFS_I(inode)->root; u64 start; - u32 *private = dip->csums; start = dip->logical_offset; do { @@ -6043,8 +6044,12 @@ static void btrfs_endio_direct_read(struct bio *bio, int err) struct page *page = bvec->bv_page; char *kaddr; u32 csum = ~(u32)0; + u64 private = ~(u32)0; unsigned long flags; + if (get_state_private(&BTRFS_I(inode)->io_tree, + start, &private)) + goto failed; local_irq_save(flags); kaddr = kmap_atomic(page); csum = btrfs_csum_data(root, kaddr + bvec->bv_offset, @@ -6054,18 +6059,18 @@ static void btrfs_endio_direct_read(struct bio *bio, int err) local_irq_restore(flags); flush_dcache_page(bvec->bv_page); - if (csum != *private) { + if (csum != private) { +failed: printk(KERN_ERR "btrfs csum failed ino %llu off" " %llu csum %u private %u\n", (unsigned long long)btrfs_ino(inode), (unsigned long long)start, - csum, *private); + csum, (unsigned)private); err = -EIO; } } start += bvec->bv_len; - private++; bvec++; } while (bvec <= bvec_end); @@ -6073,7 +6078,6 @@ static void btrfs_endio_direct_read(struct bio *bio, int err) dip->logical_offset + dip->bytes - 1); bio->bi_private = dip->private; - kfree(dip->csums); kfree(dip); /* If we had a csum failure make sure to clear the uptodate flag */ @@ -6179,7 +6183,7 @@ static struct bio *btrfs_dio_bio_alloc(struct block_device *bdev, static inline int __btrfs_submit_dio_bio(struct bio *bio, struct inode *inode, int rw, u64 file_offset, int skip_sum, - u32 *csums, int async_submit) + int async_submit) { int write = rw & REQ_WRITE; struct btrfs_root *root = BTRFS_I(inode)->root; @@ -6212,8 +6216,7 @@ static inline int __btrfs_submit_dio_bio(struct bio *bio, struct inode *inode, if (ret) goto err; } else if (!skip_sum) { - ret = btrfs_lookup_bio_sums_dio(root, inode, bio, - file_offset, csums); + ret = btrfs_lookup_bio_sums_dio(root, inode, bio, file_offset); if (ret) goto err; } @@ -6239,10 +6242,8 @@ static int btrfs_submit_direct_hook(int rw, struct btrfs_dio_private *dip, u64 submit_len = 0; u64 map_length; int nr_pages = 0; - u32 *csums = dip->csums; int ret = 0; int async_submit = 0; - int write = rw & REQ_WRITE; map_length = orig_bio->bi_size; ret = btrfs_map_block(map_tree, READ, start_sector << 9, @@ -6278,16 +6279,13 @@ static int btrfs_submit_direct_hook(int rw, struct btrfs_dio_private *dip, atomic_inc(&dip->pending_bios); ret = __btrfs_submit_dio_bio(bio, inode, rw, file_offset, skip_sum, - csums, async_submit); + async_submit); if (ret) { bio_put(bio); atomic_dec(&dip->pending_bios); goto out_err; } - /* Write's use the ordered csums */ - if (!write && !skip_sum) - csums = csums + nr_pages; start_sector += submit_len >> 9; file_offset += submit_len; @@ -6317,7 +6315,7 @@ static int btrfs_submit_direct_hook(int rw, struct btrfs_dio_private *dip, submit: ret = __btrfs_submit_dio_bio(bio, inode, rw, file_offset, skip_sum, - csums, async_submit); + async_submit); if (!ret) return 0; @@ -6353,17 +6351,6 @@ static void btrfs_submit_direct(int rw, struct bio *bio, struct inode *inode, ret = -ENOMEM; goto free_ordered; } - dip->csums = NULL; - - /* Write's use the ordered csum stuff, so we don't need dip->csums */ - if (!write && !skip_sum) { - dip->csums = kmalloc(sizeof(u32) * bio->bi_vcnt, GFP_NOFS); - if (!dip->csums) { - kfree(dip); - ret = -ENOMEM; - goto free_ordered; - } - } dip->private = bio->bi_private; dip->inode = inode; -- cgit v1.2.3-70-g09d2 From 66657b318e0e443ada229fccd40c8be86cfebdbf Mon Sep 17 00:00:00 2001 From: Josef Bacik Date: Wed, 1 Aug 2012 15:36:24 -0400 Subject: Btrfs: barrier before waitqueue_active We need a barrir before calling waitqueue_active otherwise we will miss wakeups. So in places that do atomic_dec(); then atomic_read() use atomic_dec_return() which imply a memory barrier (see memory-barriers.txt) and then add an explicit memory barrier everywhere else that need them. Thanks, Signed-off-by: Josef Bacik --- fs/btrfs/compression.c | 1 + fs/btrfs/delayed-inode.c | 7 +++---- fs/btrfs/disk-io.c | 7 ++++--- fs/btrfs/inode.c | 4 +--- fs/btrfs/volumes.c | 3 +-- 5 files changed, 10 insertions(+), 12 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/compression.c b/fs/btrfs/compression.c index 86eff48dab7..43d1c5a3a03 100644 --- a/fs/btrfs/compression.c +++ b/fs/btrfs/compression.c @@ -818,6 +818,7 @@ static void free_workspace(int type, struct list_head *workspace) btrfs_compress_op[idx]->free_workspace(workspace); atomic_dec(alloc_workspace); wake: + smp_mb(); if (waitqueue_active(workspace_wait)) wake_up(workspace_wait); } diff --git a/fs/btrfs/delayed-inode.c b/fs/btrfs/delayed-inode.c index 00deed4ef3e..07d5eeb1e6f 100644 --- a/fs/btrfs/delayed-inode.c +++ b/fs/btrfs/delayed-inode.c @@ -512,8 +512,8 @@ static void __btrfs_remove_delayed_item(struct btrfs_delayed_item *delayed_item) rb_erase(&delayed_item->rb_node, root); delayed_item->delayed_node->count--; - atomic_dec(&delayed_root->items); - if (atomic_read(&delayed_root->items) < BTRFS_DELAYED_BACKGROUND && + if (atomic_dec_return(&delayed_root->items) < + BTRFS_DELAYED_BACKGROUND && waitqueue_active(&delayed_root->wait)) wake_up(&delayed_root->wait); } @@ -1056,8 +1056,7 @@ static void btrfs_release_delayed_inode(struct btrfs_delayed_node *delayed_node) delayed_node->count--; delayed_root = delayed_node->root->fs_info->delayed_root; - atomic_dec(&delayed_root->items); - if (atomic_read(&delayed_root->items) < + if (atomic_dec_return(&delayed_root->items) < BTRFS_DELAYED_BACKGROUND && waitqueue_active(&delayed_root->wait)) wake_up(&delayed_root->wait); diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index a7ad8fc8dc5..dd86a5d8842 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -754,9 +754,7 @@ static void run_one_async_done(struct btrfs_work *work) limit = btrfs_async_submit_limit(fs_info); limit = limit * 2 / 3; - atomic_dec(&fs_info->nr_async_submits); - - if (atomic_read(&fs_info->nr_async_submits) < limit && + if (atomic_dec_return(&fs_info->nr_async_submits) < limit && waitqueue_active(&fs_info->async_submit_wait)) wake_up(&fs_info->async_submit_wait); @@ -3783,14 +3781,17 @@ int btrfs_cleanup_transaction(struct btrfs_root *root) /* FIXME: cleanup wait for commit */ t->in_commit = 1; t->blocked = 1; + smp_mb(); if (waitqueue_active(&root->fs_info->transaction_blocked_wait)) wake_up(&root->fs_info->transaction_blocked_wait); t->blocked = 0; + smp_mb(); if (waitqueue_active(&root->fs_info->transaction_wait)) wake_up(&root->fs_info->transaction_wait); t->commit_done = 1; + smp_mb(); if (waitqueue_active(&t->commit_wait)) wake_up(&t->commit_wait); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 2d65c52b094..97baf00b40d 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1007,9 +1007,7 @@ static noinline void async_cow_submit(struct btrfs_work *work) nr_pages = (async_cow->end - async_cow->start + PAGE_CACHE_SIZE) >> PAGE_CACHE_SHIFT; - atomic_sub(nr_pages, &root->fs_info->async_delalloc_pages); - - if (atomic_read(&root->fs_info->async_delalloc_pages) < + if (atomic_sub_return(nr_pages, &root->fs_info->async_delalloc_pages) < 5 * 1024 * 1024 && waitqueue_active(&root->fs_info->async_submit_wait)) wake_up(&root->fs_info->async_submit_wait); diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index 3b394503bd4..0b1e69d380d 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -227,9 +227,8 @@ loop_lock: cur = pending; pending = pending->bi_next; cur->bi_next = NULL; - atomic_dec(&fs_info->nr_async_bios); - if (atomic_read(&fs_info->nr_async_bios) < limit && + if (atomic_dec_return(&fs_info->nr_async_bios) < limit && waitqueue_active(&fs_info->async_submit_wait)) wake_up(&fs_info->async_submit_wait); -- cgit v1.2.3-70-g09d2 From 5a24e84c55f57cc49bd1cab531b6ef28b6b7bdaa Mon Sep 17 00:00:00 2001 From: Josef Bacik Date: Wed, 8 Aug 2012 10:12:59 -0600 Subject: Btrfs: fix enospc problems when deleting a subvol Subvol delete is a special kind of awful where we use the global reserve to cover the ENOSPC requirements. The problem is once we're done removing everything we do a btrfs_update_inode(), which by default will try to do the delayed update stuff which will use it's own reserve. There will be no space in this reserve and we'll return ENOSPC. So instead use btrfs_update_inode_fallback() which will just fallback to updating the inode item in the case of enospc. This is fine because the global reserve covers the space requirements for this. With this patch I can now delete a subvol on a problem image Dave Sterba sent me. Thanks, Reported-by: David Sterba Signed-off-by: Josef Bacik Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 97baf00b40d..0808f483daf 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -3171,7 +3171,7 @@ int btrfs_unlink_subvol(struct btrfs_trans_handle *trans, btrfs_i_size_write(dir, dir->i_size - name_len * 2); inode_inc_iversion(dir); dir->i_mtime = dir->i_ctime = CURRENT_TIME; - ret = btrfs_update_inode(trans, root, dir); + ret = btrfs_update_inode_fallback(trans, root, dir); if (ret) btrfs_abort_transaction(trans, root, ret); out: -- cgit v1.2.3-70-g09d2 From 24c03fa5cf3d02c327cf9f2fc39f72664b1bd7e1 Mon Sep 17 00:00:00 2001 From: Liu Bo Date: Wed, 22 Aug 2012 20:10:38 -0600 Subject: Btrfs: fix a dio write regression This bug is introduced by commit 3b8bde746f6f9bd36a9f05f5f3b6e334318176a9 (Btrfs: lock extents as we map them in DIO). In dio write, we should unlock the section which we didn't do IO on in case that we fall back to buffered write. But we need to not only unlock the section but also cleanup reserved space for the section. This bug was found while running xfstests 133, with this 133 no longer complains. Signed-off-by: Liu Bo Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 24 ++++++++++++++++++++---- 1 file changed, 20 insertions(+), 4 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 0808f483daf..38cda78de5e 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -5992,11 +5992,27 @@ unlock: * in the case of read we need to unlock only the end area that we * aren't using if there is any left over space. */ - if (lockstart < lockend) - clear_extent_bit(&BTRFS_I(inode)->io_tree, lockstart, lockend, - unlock_bits, 1, 0, &cached_state, GFP_NOFS); - else + if (lockstart < lockend) { + if (create && len < lockend - lockstart) { + clear_extent_bit(&BTRFS_I(inode)->io_tree, lockstart, + lockstart + len - 1, unlock_bits, 1, 0, + &cached_state, GFP_NOFS); + /* + * Beside unlock, we also need to cleanup reserved space + * for the left range by attaching EXTENT_DO_ACCOUNTING. + */ + clear_extent_bit(&BTRFS_I(inode)->io_tree, + lockstart + len, lockend, + unlock_bits | EXTENT_DO_ACCOUNTING, + 1, 0, NULL, GFP_NOFS); + } else { + clear_extent_bit(&BTRFS_I(inode)->io_tree, lockstart, + lockend, unlock_bits, 1, 0, + &cached_state, GFP_NOFS); + } + } else { free_extent_state(cached_state); + } free_extent_map(em); -- cgit v1.2.3-70-g09d2 From d280e5be940931c84bb2e9831ead9d02bc785484 Mon Sep 17 00:00:00 2001 From: Liu Bo Date: Tue, 21 Aug 2012 21:13:25 -0600 Subject: Btrfs: fix ordered extent leak when failing to start a transaction We cannot just return error before freeing ordered extent and releasing reserved space when we fail to start a transacion. Signed-off-by: Liu Bo Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 38cda78de5e..6ba80b90287 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1882,8 +1882,11 @@ static int btrfs_finish_ordered_io(struct btrfs_ordered_extent *ordered_extent) trans = btrfs_join_transaction_nolock(root); else trans = btrfs_join_transaction(root); - if (IS_ERR(trans)) - return PTR_ERR(trans); + if (IS_ERR(trans)) { + ret = PTR_ERR(trans); + trans = NULL; + goto out; + } trans->block_rsv = &root->fs_info->delalloc_block_rsv; ret = btrfs_update_inode_fallback(trans, root, inode); if (ret) /* -ENOMEM or corruption */ -- cgit v1.2.3-70-g09d2