From 207910ddeeda38fd54544d94f8c8ca5a9632cc25 Mon Sep 17 00:00:00 2001 From: Mark Fasheh Date: Tue, 30 Jun 2015 14:42:04 -0700 Subject: btrfs: pass unaligned length to btrfs_cmp_data() In the case that we dedupe the tail of a file, we might expand the dedupe len out to the end of our last block. We don't want to compare data past i_size however, so pass the original length to btrfs_cmp_data(). Signed-off-by: Mark Fasheh Reviewed-by: David Sterba Signed-off-by: Chris Mason --- fs/btrfs/ioctl.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'fs/btrfs/ioctl.c') diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index c86b835da7a8..55504338491d 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -2943,7 +2943,8 @@ static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen, goto out_unlock; } - ret = btrfs_cmp_data(src, loff, dst, dst_loff, len); + /* pass original length for comparison so we stay within i_size */ + ret = btrfs_cmp_data(src, loff, dst, dst_loff, olen); if (ret == 0) ret = btrfs_clone(src, dst, loff, olen, len, dst_loff); -- cgit v1.2.3 From f441460202cb787c49963bcc1f54cb48c52f7512 Mon Sep 17 00:00:00 2001 From: Mark Fasheh Date: Tue, 30 Jun 2015 14:42:05 -0700 Subject: btrfs: fix deadlock with extent-same and readpage ->readpage() does page_lock() before extent_lock(), we do the opposite in extent-same. We want to reverse the order in btrfs_extent_same() but it's not quite straightforward since the page locks are taken inside btrfs_cmp_data(). So I split btrfs_cmp_data() into 3 parts with a small context structure that is passed between them. The first, btrfs_cmp_data_prepare() gathers up the pages needed (taking page lock as required) and puts them on our context structure. At this point, we are safe to lock the extent range. Afterwards, we use btrfs_cmp_data() to do the data compare as usual and btrfs_cmp_data_free() to clean up our context. Signed-off-by: Mark Fasheh Reviewed-by: David Sterba Signed-off-by: Chris Mason --- fs/btrfs/ioctl.c | 148 +++++++++++++++++++++++++++++++++++++++++++------------ 1 file changed, 117 insertions(+), 31 deletions(-) (limited to 'fs/btrfs/ioctl.c') diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index 55504338491d..9ebe2dd31f2a 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -2765,14 +2765,11 @@ out: return ret; } -static struct page *extent_same_get_page(struct inode *inode, u64 off) +static struct page *extent_same_get_page(struct inode *inode, pgoff_t index) { struct page *page; - pgoff_t index; struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree; - index = off >> PAGE_CACHE_SHIFT; - page = grab_cache_page(inode->i_mapping, index); if (!page) return NULL; @@ -2793,6 +2790,20 @@ static struct page *extent_same_get_page(struct inode *inode, u64 off) return page; } +static int gather_extent_pages(struct inode *inode, struct page **pages, + int num_pages, u64 off) +{ + int i; + pgoff_t index = off >> PAGE_CACHE_SHIFT; + + for (i = 0; i < num_pages; i++) { + pages[i] = extent_same_get_page(inode, index + i); + if (!pages[i]) + return -ENOMEM; + } + return 0; +} + static inline void lock_extent_range(struct inode *inode, u64 off, u64 len) { /* do any pending delalloc/csum calc on src, one way or @@ -2818,52 +2829,120 @@ static inline void lock_extent_range(struct inode *inode, u64 off, u64 len) } } -static void btrfs_double_unlock(struct inode *inode1, u64 loff1, - struct inode *inode2, u64 loff2, u64 len) +static void btrfs_double_inode_unlock(struct inode *inode1, struct inode *inode2) { - unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1); - unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1); - mutex_unlock(&inode1->i_mutex); mutex_unlock(&inode2->i_mutex); } -static void btrfs_double_lock(struct inode *inode1, u64 loff1, - struct inode *inode2, u64 loff2, u64 len) +static void btrfs_double_inode_lock(struct inode *inode1, struct inode *inode2) +{ + if (inode1 < inode2) + swap(inode1, inode2); + + mutex_lock_nested(&inode1->i_mutex, I_MUTEX_PARENT); + if (inode1 != inode2) + mutex_lock_nested(&inode2->i_mutex, I_MUTEX_CHILD); +} + +static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1, + struct inode *inode2, u64 loff2, u64 len) +{ + unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1); + unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1); +} + +static void btrfs_double_extent_lock(struct inode *inode1, u64 loff1, + struct inode *inode2, u64 loff2, u64 len) { if (inode1 < inode2) { swap(inode1, inode2); swap(loff1, loff2); } - - mutex_lock_nested(&inode1->i_mutex, I_MUTEX_PARENT); lock_extent_range(inode1, loff1, len); - if (inode1 != inode2) { - mutex_lock_nested(&inode2->i_mutex, I_MUTEX_CHILD); + if (inode1 != inode2) lock_extent_range(inode2, loff2, len); +} + +struct cmp_pages { + int num_pages; + struct page **src_pages; + struct page **dst_pages; +}; + +static void btrfs_cmp_data_free(struct cmp_pages *cmp) +{ + int i; + struct page *pg; + + for (i = 0; i < cmp->num_pages; i++) { + pg = cmp->src_pages[i]; + if (pg) + page_cache_release(pg); + pg = cmp->dst_pages[i]; + if (pg) + page_cache_release(pg); + } + kfree(cmp->src_pages); + kfree(cmp->dst_pages); +} + +static int btrfs_cmp_data_prepare(struct inode *src, u64 loff, + struct inode *dst, u64 dst_loff, + u64 len, struct cmp_pages *cmp) +{ + int ret; + int num_pages = PAGE_CACHE_ALIGN(len) >> PAGE_CACHE_SHIFT; + struct page **src_pgarr, **dst_pgarr; + + /* + * We must gather up all the pages before we initiate our + * extent locking. We use an array for the page pointers. Size + * of the array is bounded by len, which is in turn bounded by + * BTRFS_MAX_DEDUPE_LEN. + */ + src_pgarr = kzalloc(num_pages * sizeof(struct page *), GFP_NOFS); + dst_pgarr = kzalloc(num_pages * sizeof(struct page *), GFP_NOFS); + if (!src_pgarr || !dst_pgarr) { + kfree(src_pgarr); + kfree(dst_pgarr); + return -ENOMEM; } + cmp->num_pages = num_pages; + cmp->src_pages = src_pgarr; + cmp->dst_pages = dst_pgarr; + + ret = gather_extent_pages(src, cmp->src_pages, cmp->num_pages, loff); + if (ret) + goto out; + + ret = gather_extent_pages(dst, cmp->dst_pages, cmp->num_pages, dst_loff); + +out: + if (ret) + btrfs_cmp_data_free(cmp); + return 0; } static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst, - u64 dst_loff, u64 len) + u64 dst_loff, u64 len, struct cmp_pages *cmp) { int ret = 0; + int i; struct page *src_page, *dst_page; unsigned int cmp_len = PAGE_CACHE_SIZE; void *addr, *dst_addr; + i = 0; while (len) { if (len < PAGE_CACHE_SIZE) cmp_len = len; - src_page = extent_same_get_page(src, loff); - if (!src_page) - return -EINVAL; - dst_page = extent_same_get_page(dst, dst_loff); - if (!dst_page) { - page_cache_release(src_page); - return -EINVAL; - } + BUG_ON(i >= cmp->num_pages); + + src_page = cmp->src_pages[i]; + dst_page = cmp->dst_pages[i]; + addr = kmap_atomic(src_page); dst_addr = kmap_atomic(dst_page); @@ -2875,15 +2954,12 @@ static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst, kunmap_atomic(addr); kunmap_atomic(dst_addr); - page_cache_release(src_page); - page_cache_release(dst_page); if (ret) break; - loff += cmp_len; - dst_loff += cmp_len; len -= cmp_len; + i++; } return ret; @@ -2914,6 +2990,7 @@ static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen, { int ret; u64 len = olen; + struct cmp_pages cmp; /* * btrfs_clone() can't handle extents in the same file @@ -2926,7 +3003,7 @@ static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen, if (len == 0) return 0; - btrfs_double_lock(src, loff, dst, dst_loff, len); + btrfs_double_inode_lock(src, dst); ret = extent_same_check_offsets(src, loff, &len, olen); if (ret) @@ -2943,13 +3020,22 @@ static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen, goto out_unlock; } + ret = btrfs_cmp_data_prepare(src, loff, dst, dst_loff, olen, &cmp); + if (ret) + goto out_unlock; + + btrfs_double_extent_lock(src, loff, dst, dst_loff, len); + /* pass original length for comparison so we stay within i_size */ - ret = btrfs_cmp_data(src, loff, dst, dst_loff, olen); + ret = btrfs_cmp_data(src, loff, dst, dst_loff, olen, &cmp); if (ret == 0) ret = btrfs_clone(src, dst, loff, olen, len, dst_loff); + btrfs_double_extent_unlock(src, loff, dst, dst_loff, len); + + btrfs_cmp_data_free(&cmp); out_unlock: - btrfs_double_unlock(src, loff, dst, dst_loff, len); + btrfs_double_inode_unlock(src, dst); return ret; } -- cgit v1.2.3 From 0efa9f48c7e6c15e75946dd2b1c82d3d19e13545 Mon Sep 17 00:00:00 2001 From: Mark Fasheh Date: Tue, 30 Jun 2015 14:42:07 -0700 Subject: btrfs: allow dedupe of same inode clone() supports cloning within an inode so extent-same can do the same now. This patch fixes up the locking in extent-same to know about the single-inode case. In addition to that, we add a check for overlapping ranges, which clone does not allow. Signed-off-by: Mark Fasheh Reviewed-by: David Sterba Signed-off-by: Chris Mason --- fs/btrfs/ioctl.c | 76 ++++++++++++++++++++++++++++++++++++++++++++------------ 1 file changed, 60 insertions(+), 16 deletions(-) (limited to 'fs/btrfs/ioctl.c') diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index 9ebe2dd31f2a..af064946c9b2 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -2991,27 +2991,61 @@ static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen, int ret; u64 len = olen; struct cmp_pages cmp; + int same_inode = 0; + u64 same_lock_start = 0; + u64 same_lock_len = 0; - /* - * btrfs_clone() can't handle extents in the same file - * yet. Once that works, we can drop this check and replace it - * with a check for the same inode, but overlapping extents. - */ if (src == dst) - return -EINVAL; + same_inode = 1; if (len == 0) return 0; - btrfs_double_inode_lock(src, dst); + if (same_inode) { + mutex_lock(&src->i_mutex); - ret = extent_same_check_offsets(src, loff, &len, olen); - if (ret) - goto out_unlock; + ret = extent_same_check_offsets(src, loff, &len, olen); + if (ret) + goto out_unlock; - ret = extent_same_check_offsets(dst, dst_loff, &len, olen); - if (ret) - goto out_unlock; + /* + * Single inode case wants the same checks, except we + * don't want our length pushed out past i_size as + * comparing that data range makes no sense. + * + * extent_same_check_offsets() will do this for an + * unaligned length at i_size, so catch it here and + * reject the request. + * + * This effectively means we require aligned extents + * for the single-inode case, whereas the other cases + * allow an unaligned length so long as it ends at + * i_size. + */ + if (len != olen) { + ret = -EINVAL; + goto out_unlock; + } + + /* Check for overlapping ranges */ + if (dst_loff + len > loff && dst_loff < loff + len) { + ret = -EINVAL; + goto out_unlock; + } + + same_lock_start = min_t(u64, loff, dst_loff); + same_lock_len = max_t(u64, loff, dst_loff) + len - same_lock_start; + } else { + btrfs_double_inode_lock(src, dst); + + ret = extent_same_check_offsets(src, loff, &len, olen); + if (ret) + goto out_unlock; + + ret = extent_same_check_offsets(dst, dst_loff, &len, olen); + if (ret) + goto out_unlock; + } /* don't make the dst file partly checksummed */ if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) != @@ -3024,18 +3058,28 @@ static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen, if (ret) goto out_unlock; - btrfs_double_extent_lock(src, loff, dst, dst_loff, len); + if (same_inode) + lock_extent_range(src, same_lock_start, same_lock_len); + else + btrfs_double_extent_lock(src, loff, dst, dst_loff, len); /* pass original length for comparison so we stay within i_size */ ret = btrfs_cmp_data(src, loff, dst, dst_loff, olen, &cmp); if (ret == 0) ret = btrfs_clone(src, dst, loff, olen, len, dst_loff); - btrfs_double_extent_unlock(src, loff, dst, dst_loff, len); + if (same_inode) + unlock_extent(&BTRFS_I(src)->io_tree, same_lock_start, + same_lock_start + same_lock_len - 1); + else + btrfs_double_extent_unlock(src, loff, dst, dst_loff, len); btrfs_cmp_data_free(&cmp); out_unlock: - btrfs_double_inode_unlock(src, dst); + if (same_inode) + mutex_unlock(&src->i_mutex); + else + btrfs_double_inode_unlock(src, dst); return ret; } -- cgit v1.2.3 From 1c919a5e13702caffbe2d2c7c305f9d0d2925160 Mon Sep 17 00:00:00 2001 From: Mark Fasheh Date: Tue, 30 Jun 2015 14:42:08 -0700 Subject: btrfs: don't update mtime/ctime on deduped inodes One issue users have reported is that dedupe changes mtime on files, resulting in tools like rsync thinking that their contents have changed when in fact the data is exactly the same. We also skip the ctime update as no user-visible metadata changes here and we want dedupe to be transparent to the user. Clone still wants time changes, so we special case this in the code. This was tested with the btrfs-extent-same tool. Signed-off-by: Mark Fasheh Signed-off-by: Chris Mason --- fs/btrfs/ioctl.c | 24 ++++++++++++++---------- 1 file changed, 14 insertions(+), 10 deletions(-) (limited to 'fs/btrfs/ioctl.c') diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index af064946c9b2..5d91776e12a2 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -87,7 +87,8 @@ struct btrfs_ioctl_received_subvol_args_32 { static int btrfs_clone(struct inode *src, struct inode *inode, - u64 off, u64 olen, u64 olen_aligned, u64 destoff); + u64 off, u64 olen, u64 olen_aligned, u64 destoff, + int no_time_update); /* Mask out flags that are inappropriate for the given type of inode. */ static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags) @@ -3066,7 +3067,7 @@ static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen, /* pass original length for comparison so we stay within i_size */ ret = btrfs_cmp_data(src, loff, dst, dst_loff, olen, &cmp); if (ret == 0) - ret = btrfs_clone(src, dst, loff, olen, len, dst_loff); + ret = btrfs_clone(src, dst, loff, olen, len, dst_loff, 1); if (same_inode) unlock_extent(&BTRFS_I(src)->io_tree, same_lock_start, @@ -3231,13 +3232,15 @@ static int clone_finish_inode_update(struct btrfs_trans_handle *trans, struct inode *inode, u64 endoff, const u64 destoff, - const u64 olen) + const u64 olen, + int no_time_update) { struct btrfs_root *root = BTRFS_I(inode)->root; int ret; inode_inc_iversion(inode); - inode->i_mtime = inode->i_ctime = CURRENT_TIME; + if (!no_time_update) + inode->i_mtime = inode->i_ctime = CURRENT_TIME; /* * We round up to the block size at eof when determining which * extents to clone above, but shouldn't round up the file size. @@ -3322,13 +3325,13 @@ static void clone_update_extent_map(struct inode *inode, * @inode: Inode to clone to * @off: Offset within source to start clone from * @olen: Original length, passed by user, of range to clone - * @olen_aligned: Block-aligned value of olen, extent_same uses - * identical values here + * @olen_aligned: Block-aligned value of olen * @destoff: Offset within @inode to start clone + * @no_time_update: Whether to update mtime/ctime on the target inode */ static int btrfs_clone(struct inode *src, struct inode *inode, const u64 off, const u64 olen, const u64 olen_aligned, - const u64 destoff) + const u64 destoff, int no_time_update) { struct btrfs_root *root = BTRFS_I(inode)->root; struct btrfs_path *path = NULL; @@ -3652,7 +3655,8 @@ process_slot: root->sectorsize); ret = clone_finish_inode_update(trans, inode, last_dest_end, - destoff, olen); + destoff, olen, + no_time_update); if (ret) goto out; if (new_key.offset + datal >= destoff + len) @@ -3690,7 +3694,7 @@ process_slot: clone_update_extent_map(inode, trans, NULL, last_dest_end, destoff + len - last_dest_end); ret = clone_finish_inode_update(trans, inode, destoff + len, - destoff, olen); + destoff, olen, no_time_update); } out: @@ -3827,7 +3831,7 @@ static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd, lock_extent_range(inode, destoff, len); } - ret = btrfs_clone(src, inode, off, olen, len, destoff); + ret = btrfs_clone(src, inode, off, olen, len, destoff, 0); if (same_inode) { u64 lock_start = min_t(u64, off, destoff); -- cgit v1.2.3