| Message ID | 20161101101842.12630-2-wangxg.fnst@cn.fujitsu.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
hi, I have rebased these 2 patches against Linux 4.9-rc2, sorry for being late. After applying these patches, Stefan does not see any ENOSPC error :) If you have free time, please have a check, thanks. Regards, Xiaoguang Wang On 11/01/2016 06:18 PM, Wang Xiaoguang wrote: > When testing btrfs compression, sometimes we got ENOSPC error, though fs > still has much free space, xfstests generic/171, generic/172, generic/173, > generic/174, generic/175 can reveal this bug in my test environment when > compression is enabled. > > After some debuging work, we found that it's btrfs_delalloc_reserve_metadata() > which sometimes tries to reserve plenty of metadata space, even for very small > data range. In btrfs_delalloc_reserve_metadata(), the number of metadata bytes > we try to reserve is calculated by the difference between outstanding_extents > and reserved_extents. Please see below case for how ENOSPC occurs: > > 1, Buffered write 128MB data in unit of 128KB, so finially we'll have inode > outstanding extents be 1, and reserved_extents be 1024. Note it's > btrfs_merge_extent_hook() that merges these 128KB units into one big > outstanding extent, but do not change reserved_extents. > > 2, When writing dirty pages, for compression, cow_file_range_async() will > split above big extent in unit of 128KB(compression extent size is 128KB). > When first split opeartion finishes, we'll have 2 outstanding extents and 1024 > reserved extents, and just right now the currently generated ordered extent is > dispatched to run and complete, then btrfs_delalloc_release_metadata()(see > btrfs_finish_ordered_io()) will be called to release metadata, after that we > will have 1 outstanding extents and 1 reserved extents(also see logic in > drop_outstanding_extent()). Later cow_file_range_async() continues to handles > left data range[128KB, 128MB), and if no other ordered extent was dispatched > to run, there will be 1023 outstanding extents and 1 reserved extent. > > 3, Now if another bufferd write for this file enters, then > btrfs_delalloc_reserve_metadata() will at least try to reserve metadata > for 1023 outstanding extents' metadata, for 16KB node size, it'll be 1023*16384*2*8, > about 255MB, for 64K node size, it'll be 1023*65536*8*2, about 1GB metadata, so > obviously it's not sane and can easily result in enospc error. > > The root cause is that for compression, its max extent size will no longer be > BTRFS_MAX_EXTENT_SIZE(128MB), it'll be 128KB, so current metadata reservation > method in btrfs is not appropriate or correct, here we introduce: > enum btrfs_metadata_reserve_type { > BTRFS_RESERVE_NORMAL, > BTRFS_RESERVE_COMPRESS, > }; > and expand btrfs_delalloc_reserve_metadata() and btrfs_delalloc_reserve_space() > by adding a new enum btrfs_metadata_reserve_type argument. When a data range will > go through compression, we use BTRFS_RESERVE_COMPRESS to reserve metatata. > Meanwhile we introduce EXTENT_COMPRESS flag to mark a data range that will go > through compression path. > > With this patch, we can fix these false enospc error for compression. > > Signed-off-by: Wang Xiaoguang <wangxg.fnst@cn.fujitsu.com> > Tested-by: Holger Hoffstätte <holger@applied-asynchrony.com> > Tested-by: Stefan Priebe <s.priebe@profihost.ag> > --- > fs/btrfs/ctree.h | 31 ++++++-- > fs/btrfs/extent-tree.c | 55 +++++++++---- > fs/btrfs/extent_io.c | 59 +++++++++++++- > fs/btrfs/extent_io.h | 2 + > fs/btrfs/file.c | 28 +++++-- > fs/btrfs/free-space-cache.c | 6 +- > fs/btrfs/inode-map.c | 5 +- > fs/btrfs/inode.c | 185 ++++++++++++++++++++++++++++++++----------- > fs/btrfs/ioctl.c | 12 ++- > fs/btrfs/relocation.c | 14 +++- > fs/btrfs/tests/inode-tests.c | 15 ++-- > 11 files changed, 315 insertions(+), 97 deletions(-) > > diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h > index 52a8535..457b106 100644 > --- a/fs/btrfs/ctree.h > +++ b/fs/btrfs/ctree.h > @@ -98,6 +98,19 @@ static const int btrfs_csum_sizes[] = { 4 }; > > #define BTRFS_DIRTY_METADATA_THRESH SZ_32M > > +/* > + * for compression, max file extent size would be limited to 128K, so when > + * reserving metadata for such delalloc writes, pass BTRFS_RESERVE_COMPRESS to > + * btrfs_delalloc_reserve_metadata() or btrfs_delalloc_reserve_space() to > + * calculate metadata, for none-compression, use BTRFS_RESERVE_NORMAL. > + */ > +enum btrfs_metadata_reserve_type { > + BTRFS_RESERVE_NORMAL, > + BTRFS_RESERVE_COMPRESS, > +}; > +int inode_need_compress(struct inode *inode); > +u64 btrfs_max_extent_size(enum btrfs_metadata_reserve_type reserve_type); > + > #define BTRFS_MAX_EXTENT_SIZE SZ_128M > > struct btrfs_mapping_tree { > @@ -2693,10 +2706,14 @@ int btrfs_subvolume_reserve_metadata(struct btrfs_root *root, > void btrfs_subvolume_release_metadata(struct btrfs_root *root, > struct btrfs_block_rsv *rsv, > u64 qgroup_reserved); > -int btrfs_delalloc_reserve_metadata(struct inode *inode, u64 num_bytes); > -void btrfs_delalloc_release_metadata(struct inode *inode, u64 num_bytes); > -int btrfs_delalloc_reserve_space(struct inode *inode, u64 start, u64 len); > -void btrfs_delalloc_release_space(struct inode *inode, u64 start, u64 len); > +int btrfs_delalloc_reserve_metadata(struct inode *inode, u64 num_bytes, > + enum btrfs_metadata_reserve_type reserve_type); > +void btrfs_delalloc_release_metadata(struct inode *inode, u64 num_bytes, > + enum btrfs_metadata_reserve_type reserve_type); > +int btrfs_delalloc_reserve_space(struct inode *inode, u64 start, u64 len, > + enum btrfs_metadata_reserve_type reserve_type); > +void btrfs_delalloc_release_space(struct inode *inode, u64 start, u64 len, > + enum btrfs_metadata_reserve_type reserve_type); > void btrfs_init_block_rsv(struct btrfs_block_rsv *rsv, unsigned short type); > struct btrfs_block_rsv *btrfs_alloc_block_rsv(struct btrfs_root *root, > unsigned short type); > @@ -3138,9 +3155,9 @@ int btrfs_start_delalloc_inodes(struct btrfs_root *root, int delay_iput); > int btrfs_start_delalloc_roots(struct btrfs_fs_info *fs_info, int delay_iput, > int nr); > int btrfs_set_extent_delalloc(struct inode *inode, u64 start, u64 end, > - struct extent_state **cached_state, int dedupe); > + struct extent_state **cached_state, int flag); > int btrfs_set_extent_defrag(struct inode *inode, u64 start, u64 end, > - struct extent_state **cached_state); > + struct extent_state **cached_state, int flag); > int btrfs_create_subvol_root(struct btrfs_trans_handle *trans, > struct btrfs_root *new_root, > struct btrfs_root *parent_root, > @@ -3232,7 +3249,7 @@ int btrfs_release_file(struct inode *inode, struct file *file); > int btrfs_dirty_pages(struct btrfs_root *root, struct inode *inode, > struct page **pages, size_t num_pages, > loff_t pos, size_t write_bytes, > - struct extent_state **cached); > + struct extent_state **cached, int flag); > int btrfs_fdatawrite_range(struct inode *inode, loff_t start, loff_t end); > ssize_t btrfs_copy_file_range(struct file *file_in, loff_t pos_in, > struct file *file_out, loff_t pos_out, > diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c > index 210c94a..9aa6d2c 100644 > --- a/fs/btrfs/extent-tree.c > +++ b/fs/btrfs/extent-tree.c > @@ -5838,15 +5838,16 @@ void btrfs_subvolume_release_metadata(struct btrfs_root *root, > * reserved extents that need to be freed. This must be called with > * BTRFS_I(inode)->lock held. > */ > -static unsigned drop_outstanding_extent(struct inode *inode, u64 num_bytes) > +static unsigned drop_outstanding_extent(struct inode *inode, u64 num_bytes, > + enum btrfs_metadata_reserve_type reserve_type) > { > unsigned drop_inode_space = 0; > unsigned dropped_extents = 0; > unsigned num_extents = 0; > + u64 max_extent_size = btrfs_max_extent_size(reserve_type); > > - num_extents = (unsigned)div64_u64(num_bytes + > - BTRFS_MAX_EXTENT_SIZE - 1, > - BTRFS_MAX_EXTENT_SIZE); > + num_extents = (unsigned)div64_u64(num_bytes + max_extent_size - 1, > + max_extent_size); > ASSERT(num_extents); > ASSERT(BTRFS_I(inode)->outstanding_extents >= num_extents); > BTRFS_I(inode)->outstanding_extents -= num_extents; > @@ -5916,7 +5917,21 @@ static u64 calc_csum_metadata_size(struct inode *inode, u64 num_bytes, > return btrfs_calc_trans_metadata_size(root, old_csums - num_csums); > } > > -int btrfs_delalloc_reserve_metadata(struct inode *inode, u64 num_bytes) > +u64 btrfs_max_extent_size(enum btrfs_metadata_reserve_type reserve_type) > +{ > + if (reserve_type == BTRFS_RESERVE_COMPRESS) > + return SZ_128K; > + > + return BTRFS_MAX_EXTENT_SIZE; > +} > + > +/* > + * @reserve_type: normally reserve_type should be BTRFS_RESERVE_NORMAL, but for > + * compression path, its max extent size is limited to 128KB, not 128MB, when > + * reserving metadata, we should set reserve_type to BTRFS_RESERVE_COMPRESS. > + */ > +int btrfs_delalloc_reserve_metadata(struct inode *inode, u64 num_bytes, > + enum btrfs_metadata_reserve_type reserve_type) > { > struct btrfs_root *root = BTRFS_I(inode)->root; > struct btrfs_block_rsv *block_rsv = &root->fs_info->delalloc_block_rsv; > @@ -5929,6 +5944,7 @@ int btrfs_delalloc_reserve_metadata(struct inode *inode, u64 num_bytes) > u64 to_free = 0; > unsigned dropped; > bool release_extra = false; > + u64 max_extent_size = btrfs_max_extent_size(reserve_type); > > /* If we are a free space inode we need to not flush since we will be in > * the middle of a transaction commit. We also don't need the delalloc > @@ -5955,9 +5971,8 @@ int btrfs_delalloc_reserve_metadata(struct inode *inode, u64 num_bytes) > num_bytes = ALIGN(num_bytes, root->sectorsize); > > spin_lock(&BTRFS_I(inode)->lock); > - nr_extents = (unsigned)div64_u64(num_bytes + > - BTRFS_MAX_EXTENT_SIZE - 1, > - BTRFS_MAX_EXTENT_SIZE); > + nr_extents = (unsigned)div64_u64(num_bytes + max_extent_size - 1, > + max_extent_size); > BTRFS_I(inode)->outstanding_extents += nr_extents; > > nr_extents = 0; > @@ -6008,7 +6023,7 @@ int btrfs_delalloc_reserve_metadata(struct inode *inode, u64 num_bytes) > > out_fail: > spin_lock(&BTRFS_I(inode)->lock); > - dropped = drop_outstanding_extent(inode, num_bytes); > + dropped = drop_outstanding_extent(inode, num_bytes, reserve_type); > /* > * If the inodes csum_bytes is the same as the original > * csum_bytes then we know we haven't raced with any free()ers > @@ -6074,12 +6089,15 @@ int btrfs_delalloc_reserve_metadata(struct inode *inode, u64 num_bytes) > * btrfs_delalloc_release_metadata - release a metadata reservation for an inode > * @inode: the inode to release the reservation for > * @num_bytes: the number of bytes we're releasing > + * @reserve_type: this value must be same to the value passing to > + * btrfs_delalloc_reserve_metadata(). > * > * This will release the metadata reservation for an inode. This can be called > * once we complete IO for a given set of bytes to release their metadata > * reservations. > */ > -void btrfs_delalloc_release_metadata(struct inode *inode, u64 num_bytes) > +void btrfs_delalloc_release_metadata(struct inode *inode, u64 num_bytes, > + enum btrfs_metadata_reserve_type reserve_type) > { > struct btrfs_root *root = BTRFS_I(inode)->root; > u64 to_free = 0; > @@ -6087,7 +6105,7 @@ void btrfs_delalloc_release_metadata(struct inode *inode, u64 num_bytes) > > num_bytes = ALIGN(num_bytes, root->sectorsize); > spin_lock(&BTRFS_I(inode)->lock); > - dropped = drop_outstanding_extent(inode, num_bytes); > + dropped = drop_outstanding_extent(inode, num_bytes, reserve_type); > > if (num_bytes) > to_free = calc_csum_metadata_size(inode, num_bytes, 0); > @@ -6111,6 +6129,9 @@ void btrfs_delalloc_release_metadata(struct inode *inode, u64 num_bytes) > * @inode: inode we're writing to > * @start: start range we are writing to > * @len: how long the range we are writing to > + * @reserve_type: normally reserve_type should be BTRFS_RESERVE_NORMAL, but for > + * compression path, its max extent size is limited to 128KB, not 128MB, when > + * reserving metadata, we should set reserve_type to BTRFS_RESERVE_COMPRESS. > * > * This will do the following things > * > @@ -6128,14 +6149,15 @@ void btrfs_delalloc_release_metadata(struct inode *inode, u64 num_bytes) > * Return 0 for success > * Return <0 for error(-ENOSPC or -EQUOT) > */ > -int btrfs_delalloc_reserve_space(struct inode *inode, u64 start, u64 len) > +int btrfs_delalloc_reserve_space(struct inode *inode, u64 start, u64 len, > + enum btrfs_metadata_reserve_type reserve_type) > { > int ret; > > ret = btrfs_check_data_free_space(inode, start, len); > if (ret < 0) > return ret; > - ret = btrfs_delalloc_reserve_metadata(inode, len); > + ret = btrfs_delalloc_reserve_metadata(inode, len, reserve_type); > if (ret < 0) > btrfs_free_reserved_data_space(inode, start, len); > return ret; > @@ -6146,6 +6168,8 @@ int btrfs_delalloc_reserve_space(struct inode *inode, u64 start, u64 len) > * @inode: inode we're releasing space for > * @start: start position of the space already reserved > * @len: the len of the space already reserved > + * @reserve_type: this value must be same to the value passing to > + * btrfs_delalloc_reserve_space(). > * > * This must be matched with a call to btrfs_delalloc_reserve_space. This is > * called in the case that we don't need the metadata AND data reservations > @@ -6156,9 +6180,10 @@ int btrfs_delalloc_reserve_space(struct inode *inode, u64 start, u64 len) > * list if there are no delalloc bytes left. > * Also it will handle the qgroup reserved space. > */ > -void btrfs_delalloc_release_space(struct inode *inode, u64 start, u64 len) > +void btrfs_delalloc_release_space(struct inode *inode, u64 start, u64 len, > + enum btrfs_metadata_reserve_type reserve_type) > { > - btrfs_delalloc_release_metadata(inode, len); > + btrfs_delalloc_release_metadata(inode, len, reserve_type); > btrfs_free_reserved_data_space(inode, start, len); > } > > diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c > index 66a7551..8c07306 100644 > --- a/fs/btrfs/extent_io.c > +++ b/fs/btrfs/extent_io.c > @@ -603,7 +603,7 @@ static int __clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, > btrfs_debug_check_extent_io_range(tree, start, end); > > if (bits & EXTENT_DELALLOC) > - bits |= EXTENT_NORESERVE; > + bits |= EXTENT_NORESERVE | EXTENT_COMPRESS; > > if (delete) > bits |= ~EXTENT_CTLBITS; > @@ -742,6 +742,58 @@ static int __clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, > > } > > +static void adjust_one_outstanding_extent(struct inode *inode, u64 len) > +{ > + unsigned old_extents, new_extents; > + > + old_extents = div64_u64(len + SZ_128K - 1, SZ_128K); > + new_extents = div64_u64(len + BTRFS_MAX_EXTENT_SIZE - 1, > + BTRFS_MAX_EXTENT_SIZE); > + if (old_extents <= new_extents) > + return; > + > + spin_lock(&BTRFS_I(inode)->lock); > + BTRFS_I(inode)->outstanding_extents -= old_extents - new_extents; > + spin_unlock(&BTRFS_I(inode)->lock); > +} > + > +/* > + * For a extent with EXTENT_COMPRESS flag, if later it does not go through > + * compress path, we need to adjust the number of outstanding_extents. > + * It's because for extent with EXTENT_COMPRESS flag, its number of outstanding > + * extents is calculated by 128KB, so here we need to adjust it. > + */ > +void adjust_outstanding_extents(struct inode *inode, > + u64 start, u64 end) > +{ > + struct rb_node *node; > + struct extent_state *state; > + struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree; > + > + spin_lock(&tree->lock); > + node = tree_search(tree, start); > + if (!node) > + goto out; > + > + while (1) { > + state = rb_entry(node, struct extent_state, rb_node); > + if (state->start > end) > + goto out; > + /* > + * The whole range is locked, so we can safely clear > + * EXTENT_COMPRESS flag. > + */ > + state->state &= ~EXTENT_COMPRESS; > + adjust_one_outstanding_extent(inode, > + state->end - state->start + 1); > + node = rb_next(node); > + if (!node) > + break; > + } > +out: > + spin_unlock(&tree->lock); > +} > + > static void wait_on_state(struct extent_io_tree *tree, > struct extent_state *state) > __releases(tree->lock) > @@ -1504,6 +1556,7 @@ static noinline u64 find_delalloc_range(struct extent_io_tree *tree, > u64 cur_start = *start; > u64 found = 0; > u64 total_bytes = 0; > + unsigned pre_state; > > spin_lock(&tree->lock); > > @@ -1521,7 +1574,8 @@ static noinline u64 find_delalloc_range(struct extent_io_tree *tree, > while (1) { > state = rb_entry(node, struct extent_state, rb_node); > if (found && (state->start != cur_start || > - (state->state & EXTENT_BOUNDARY))) { > + (state->state & EXTENT_BOUNDARY) || > + (state->state ^ pre_state) & EXTENT_COMPRESS)) { > goto out; > } > if (!(state->state & EXTENT_DELALLOC)) { > @@ -1537,6 +1591,7 @@ static noinline u64 find_delalloc_range(struct extent_io_tree *tree, > found++; > *end = state->end; > cur_start = state->end + 1; > + pre_state = state->state; > node = rb_next(node); > total_bytes += state->end - state->start + 1; > if (total_bytes >= max_bytes) > diff --git a/fs/btrfs/extent_io.h b/fs/btrfs/extent_io.h > index ab31d14..d9f846c 100644 > --- a/fs/btrfs/extent_io.h > +++ b/fs/btrfs/extent_io.h > @@ -21,6 +21,7 @@ > #define EXTENT_NORESERVE (1U << 15) > #define EXTENT_QGROUP_RESERVED (1U << 16) > #define EXTENT_CLEAR_DATA_RESV (1U << 17) > +#define EXTENT_COMPRESS (1U << 18) > #define EXTENT_IOBITS (EXTENT_LOCKED | EXTENT_WRITEBACK) > #define EXTENT_CTLBITS (EXTENT_DO_ACCOUNTING | EXTENT_FIRST_DELALLOC) > > @@ -247,6 +248,7 @@ int clear_record_extent_bits(struct extent_io_tree *tree, u64 start, u64 end, > int clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, > unsigned bits, int wake, int delete, > struct extent_state **cached, gfp_t mask); > +void adjust_outstanding_extents(struct inode *inode, u64 start, u64 end); > > static inline int unlock_extent(struct extent_io_tree *tree, u64 start, u64 end) > { > diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c > index 3a14c87..e4c998d 100644 > --- a/fs/btrfs/file.c > +++ b/fs/btrfs/file.c > @@ -484,11 +484,15 @@ static void btrfs_drop_pages(struct page **pages, size_t num_pages) > * > * this also makes the decision about creating an inline extent vs > * doing real data extents, marking pages dirty and delalloc as required. > + * > + * normally flag should be 0, but if a data range will go through compress path, > + * set flag to 1. Note: here we should ensure enum btrfs_metadata_reserve_type > + * and flag's values are consistent. > */ > int btrfs_dirty_pages(struct btrfs_root *root, struct inode *inode, > struct page **pages, size_t num_pages, > loff_t pos, size_t write_bytes, > - struct extent_state **cached) > + struct extent_state **cached, int flag) > { > int err = 0; > int i; > @@ -503,7 +507,7 @@ int btrfs_dirty_pages(struct btrfs_root *root, struct inode *inode, > > end_of_last_block = start_pos + num_bytes - 1; > err = btrfs_set_extent_delalloc(inode, start_pos, end_of_last_block, > - cached, 0); > + cached, flag); > if (err) > return err; > > @@ -1521,6 +1525,7 @@ static noinline ssize_t __btrfs_buffered_write(struct file *file, > bool only_release_metadata = false; > bool force_page_uptodate = false; > bool need_unlock; > + enum btrfs_metadata_reserve_type reserve_type = BTRFS_RESERVE_NORMAL; > > nrptrs = min(DIV_ROUND_UP(iov_iter_count(i), PAGE_SIZE), > PAGE_SIZE / (sizeof(struct page *))); > @@ -1530,6 +1535,9 @@ static noinline ssize_t __btrfs_buffered_write(struct file *file, > if (!pages) > return -ENOMEM; > > + if (inode_need_compress(inode)) > + reserve_type = BTRFS_RESERVE_COMPRESS; > + > while (iov_iter_count(i) > 0) { > size_t offset = pos & (PAGE_SIZE - 1); > size_t sector_offset; > @@ -1583,7 +1591,8 @@ static noinline ssize_t __btrfs_buffered_write(struct file *file, > } > } > > - ret = btrfs_delalloc_reserve_metadata(inode, reserve_bytes); > + ret = btrfs_delalloc_reserve_metadata(inode, reserve_bytes, > + reserve_type); > if (ret) { > if (!only_release_metadata) > btrfs_free_reserved_data_space(inode, pos, > @@ -1666,14 +1675,16 @@ static noinline ssize_t __btrfs_buffered_write(struct file *file, > } > if (only_release_metadata) { > btrfs_delalloc_release_metadata(inode, > - release_bytes); > + release_bytes, > + reserve_type); > } else { > u64 __pos; > > __pos = round_down(pos, root->sectorsize) + > (dirty_pages << PAGE_SHIFT); > btrfs_delalloc_release_space(inode, __pos, > - release_bytes); > + release_bytes, > + reserve_type); > } > } > > @@ -1683,7 +1694,7 @@ static noinline ssize_t __btrfs_buffered_write(struct file *file, > if (copied > 0) > ret = btrfs_dirty_pages(root, inode, pages, > dirty_pages, pos, copied, > - NULL); > + NULL, reserve_type); > if (need_unlock) > unlock_extent_cached(&BTRFS_I(inode)->io_tree, > lockstart, lockend, &cached_state, > @@ -1724,11 +1735,12 @@ static noinline ssize_t __btrfs_buffered_write(struct file *file, > if (release_bytes) { > if (only_release_metadata) { > btrfs_end_write_no_snapshoting(root); > - btrfs_delalloc_release_metadata(inode, release_bytes); > + btrfs_delalloc_release_metadata(inode, release_bytes, > + reserve_type); > } else { > btrfs_delalloc_release_space(inode, > round_down(pos, root->sectorsize), > - release_bytes); > + release_bytes, reserve_type); > } > } > > diff --git a/fs/btrfs/free-space-cache.c b/fs/btrfs/free-space-cache.c > index e4b48f3..e8947c2 100644 > --- a/fs/btrfs/free-space-cache.c > +++ b/fs/btrfs/free-space-cache.c > @@ -1297,7 +1297,7 @@ static int __btrfs_write_out_cache(struct btrfs_root *root, struct inode *inode, > > /* Everything is written out, now we dirty the pages in the file. */ > ret = btrfs_dirty_pages(root, inode, io_ctl->pages, io_ctl->num_pages, > - 0, i_size_read(inode), &cached_state); > + 0, i_size_read(inode), &cached_state, 0); > if (ret) > goto out_nospc; > > @@ -3516,6 +3516,7 @@ int btrfs_write_out_ino_cache(struct btrfs_root *root, > int ret; > struct btrfs_io_ctl io_ctl; > bool release_metadata = true; > + enum btrfs_metadata_reserve_type reserve_type = BTRFS_RESERVE_NORMAL; > > if (!btrfs_test_opt(root->fs_info, INODE_MAP_CACHE)) > return 0; > @@ -3536,7 +3537,8 @@ int btrfs_write_out_ino_cache(struct btrfs_root *root, > > if (ret) { > if (release_metadata) > - btrfs_delalloc_release_metadata(inode, inode->i_size); > + btrfs_delalloc_release_metadata(inode, inode->i_size, > + reserve_type); > #ifdef DEBUG > btrfs_err(root->fs_info, > "failed to write free ino cache for root %llu", > diff --git a/fs/btrfs/inode-map.c b/fs/btrfs/inode-map.c > index d27014b..39c4b1c 100644 > --- a/fs/btrfs/inode-map.c > +++ b/fs/btrfs/inode-map.c > @@ -404,6 +404,7 @@ int btrfs_save_ino_cache(struct btrfs_root *root, > int ret; > int prealloc; > bool retry = false; > + enum btrfs_metadata_reserve_type reserve_type = BTRFS_RESERVE_NORMAL; > > /* only fs tree and subvol/snap needs ino cache */ > if (root->root_key.objectid != BTRFS_FS_TREE_OBJECTID && > @@ -491,14 +492,14 @@ int btrfs_save_ino_cache(struct btrfs_root *root, > /* Just to make sure we have enough space */ > prealloc += 8 * PAGE_SIZE; > > - ret = btrfs_delalloc_reserve_space(inode, 0, prealloc); > + ret = btrfs_delalloc_reserve_space(inode, 0, prealloc, reserve_type); > if (ret) > goto out_put; > > ret = btrfs_prealloc_file_range_trans(inode, trans, 0, 0, prealloc, > prealloc, prealloc, &alloc_hint); > if (ret) { > - btrfs_delalloc_release_metadata(inode, prealloc); > + btrfs_delalloc_release_metadata(inode, prealloc, reserve_type); > goto out_put; > } > > diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c > index e427b81..608fd84 100644 > --- a/fs/btrfs/inode.c > +++ b/fs/btrfs/inode.c > @@ -315,7 +315,7 @@ static noinline int cow_file_range_inline(struct btrfs_root *root, > } > > set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &BTRFS_I(inode)->runtime_flags); > - btrfs_delalloc_release_metadata(inode, end + 1 - start); > + btrfs_delalloc_release_metadata(inode, end + 1 - start, 0); > btrfs_drop_extent_cache(inode, start, aligned_end - 1, 0); > out: > /* > @@ -371,7 +371,7 @@ static noinline int add_async_extent(struct async_cow *cow, > return 0; > } > > -static inline int inode_need_compress(struct inode *inode) > +int inode_need_compress(struct inode *inode) > { > struct btrfs_root *root = BTRFS_I(inode)->root; > > @@ -710,6 +710,16 @@ static noinline void submit_compressed_extents(struct inode *inode, > async_extent->start + > async_extent->ram_size - 1); > > + /* > + * We use 128KB as max extent size to calculate number > + * of outstanding extents for this extent before, now > + * it'll go throuth uncompressed IO, we need to use > + * 128MB as max extent size to re-calculate number of > + * outstanding extents for this extent. > + */ > + adjust_outstanding_extents(inode, async_extent->start, > + async_extent->start + > + async_extent->ram_size - 1); > /* allocate blocks */ > ret = cow_file_range(inode, async_cow->locked_page, > async_extent->start, > @@ -1573,14 +1583,24 @@ static int run_delalloc_range(struct inode *inode, struct page *locked_page, > { > int ret; > int force_cow = need_force_cow(inode, start, end); > + struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; > + int need_compress; > > + need_compress = test_range_bit(io_tree, start, end, > + EXTENT_COMPRESS, 1, NULL); > if (BTRFS_I(inode)->flags & BTRFS_INODE_NODATACOW && !force_cow) { > + if (need_compress) > + adjust_outstanding_extents(inode, start, end); > + > ret = run_delalloc_nocow(inode, locked_page, start, end, > page_started, 1, nr_written); > } else if (BTRFS_I(inode)->flags & BTRFS_INODE_PREALLOC && !force_cow) { > + if (need_compress) > + adjust_outstanding_extents(inode, start, end); > + > ret = run_delalloc_nocow(inode, locked_page, start, end, > page_started, 0, nr_written); > - } else if (!inode_need_compress(inode)) { > + } else if (!need_compress) { > ret = cow_file_range(inode, locked_page, start, end, end, > page_started, nr_written, 1, NULL); > } else { > @@ -1596,6 +1616,7 @@ static void btrfs_split_extent_hook(struct inode *inode, > struct extent_state *orig, u64 split) > { > u64 size; > + u64 max_extent_size = BTRFS_MAX_EXTENT_SIZE; > > /* not delalloc, ignore it */ > if (!(orig->state & EXTENT_DELALLOC)) > @@ -1604,8 +1625,11 @@ static void btrfs_split_extent_hook(struct inode *inode, > if (btrfs_is_free_space_inode(inode)) > return; > > + if (orig->state & EXTENT_COMPRESS) > + max_extent_size = SZ_128K; > + > size = orig->end - orig->start + 1; > - if (size > BTRFS_MAX_EXTENT_SIZE) { > + if (size > max_extent_size) { > u64 num_extents; > u64 new_size; > > @@ -1614,13 +1638,13 @@ static void btrfs_split_extent_hook(struct inode *inode, > * applies here, just in reverse. > */ > new_size = orig->end - split + 1; > - num_extents = div64_u64(new_size + BTRFS_MAX_EXTENT_SIZE - 1, > - BTRFS_MAX_EXTENT_SIZE); > + num_extents = div64_u64(new_size + max_extent_size - 1, > + max_extent_size); > new_size = split - orig->start; > - num_extents += div64_u64(new_size + BTRFS_MAX_EXTENT_SIZE - 1, > - BTRFS_MAX_EXTENT_SIZE); > - if (div64_u64(size + BTRFS_MAX_EXTENT_SIZE - 1, > - BTRFS_MAX_EXTENT_SIZE) >= num_extents) > + num_extents += div64_u64(new_size + max_extent_size - 1, > + max_extent_size); > + if (div64_u64(size + max_extent_size - 1, > + max_extent_size) >= num_extents) > return; > } > > @@ -1641,6 +1665,7 @@ static void btrfs_merge_extent_hook(struct inode *inode, > { > u64 new_size, old_size; > u64 num_extents; > + u64 max_extent_size = BTRFS_MAX_EXTENT_SIZE; > > /* not delalloc, ignore it */ > if (!(other->state & EXTENT_DELALLOC)) > @@ -1649,13 +1674,16 @@ static void btrfs_merge_extent_hook(struct inode *inode, > if (btrfs_is_free_space_inode(inode)) > return; > > + if (other->state & EXTENT_COMPRESS) > + max_extent_size = SZ_128K; > + > if (new->start > other->start) > new_size = new->end - other->start + 1; > else > new_size = other->end - new->start + 1; > > /* we're not bigger than the max, unreserve the space and go */ > - if (new_size <= BTRFS_MAX_EXTENT_SIZE) { > + if (new_size <= max_extent_size) { > spin_lock(&BTRFS_I(inode)->lock); > BTRFS_I(inode)->outstanding_extents--; > spin_unlock(&BTRFS_I(inode)->lock); > @@ -1681,14 +1709,14 @@ static void btrfs_merge_extent_hook(struct inode *inode, > * this case. > */ > old_size = other->end - other->start + 1; > - num_extents = div64_u64(old_size + BTRFS_MAX_EXTENT_SIZE - 1, > - BTRFS_MAX_EXTENT_SIZE); > + num_extents = div64_u64(old_size + max_extent_size - 1, > + max_extent_size); > old_size = new->end - new->start + 1; > - num_extents += div64_u64(old_size + BTRFS_MAX_EXTENT_SIZE - 1, > - BTRFS_MAX_EXTENT_SIZE); > + num_extents += div64_u64(old_size + max_extent_size - 1, > + max_extent_size); > > - if (div64_u64(new_size + BTRFS_MAX_EXTENT_SIZE - 1, > - BTRFS_MAX_EXTENT_SIZE) >= num_extents) > + if (div64_u64(new_size + max_extent_size - 1, > + max_extent_size) >= num_extents) > return; > > spin_lock(&BTRFS_I(inode)->lock); > @@ -1754,10 +1782,15 @@ static void btrfs_set_bit_hook(struct inode *inode, > if (!(state->state & EXTENT_DELALLOC) && (*bits & EXTENT_DELALLOC)) { > struct btrfs_root *root = BTRFS_I(inode)->root; > u64 len = state->end + 1 - state->start; > - u64 num_extents = div64_u64(len + BTRFS_MAX_EXTENT_SIZE - 1, > - BTRFS_MAX_EXTENT_SIZE); > + u64 max_extent_size = BTRFS_MAX_EXTENT_SIZE; > + u64 num_extents; > bool do_list = !btrfs_is_free_space_inode(inode); > > + if (*bits & EXTENT_COMPRESS) > + max_extent_size = SZ_128K; > + num_extents = div64_u64(len + max_extent_size - 1, > + max_extent_size); > + > if (*bits & EXTENT_FIRST_DELALLOC) > *bits &= ~EXTENT_FIRST_DELALLOC; > > @@ -1792,8 +1825,9 @@ static void btrfs_clear_bit_hook(struct inode *inode, > unsigned *bits) > { > u64 len = state->end + 1 - state->start; > - u64 num_extents = div64_u64(len + BTRFS_MAX_EXTENT_SIZE -1, > - BTRFS_MAX_EXTENT_SIZE); > + u64 max_extent_size = BTRFS_MAX_EXTENT_SIZE; > + u64 num_extents; > + enum btrfs_metadata_reserve_type reserve_type = BTRFS_RESERVE_NORMAL; > > spin_lock(&BTRFS_I(inode)->lock); > if ((state->state & EXTENT_DEFRAG) && (*bits & EXTENT_DEFRAG)) > @@ -1809,6 +1843,14 @@ static void btrfs_clear_bit_hook(struct inode *inode, > struct btrfs_root *root = BTRFS_I(inode)->root; > bool do_list = !btrfs_is_free_space_inode(inode); > > + if (state->state & EXTENT_COMPRESS) { > + max_extent_size = SZ_128K; > + reserve_type = BTRFS_RESERVE_COMPRESS; > + } > + > + num_extents = div64_u64(len + max_extent_size - 1, > + max_extent_size); > + > if (*bits & EXTENT_FIRST_DELALLOC) { > *bits &= ~EXTENT_FIRST_DELALLOC; > } else if (!(*bits & EXTENT_DO_ACCOUNTING) && do_list) { > @@ -1824,7 +1866,8 @@ static void btrfs_clear_bit_hook(struct inode *inode, > */ > if (*bits & EXTENT_DO_ACCOUNTING && > root != root->fs_info->tree_root) > - btrfs_delalloc_release_metadata(inode, len); > + btrfs_delalloc_release_metadata(inode, len, > + reserve_type); > > /* For sanity tests. */ > if (btrfs_is_testing(root->fs_info)) > @@ -2006,16 +2049,34 @@ static noinline int add_pending_csums(struct btrfs_trans_handle *trans, > return 0; > } > > +/* > + * Normally flag should be 0, but if a data range will go through compress path, > + * set flag to 1. Note: here we should ensure enum btrfs_metadata_reserve_type > + * and flag's values are consistent. > + */ > int btrfs_set_extent_delalloc(struct inode *inode, u64 start, u64 end, > - struct extent_state **cached_state, int dedupe) > + struct extent_state **cached_state, int flag) > { > int ret; > - u64 num_extents = div64_u64(end - start + BTRFS_MAX_EXTENT_SIZE, > - BTRFS_MAX_EXTENT_SIZE); > + unsigned bits; > + u64 max_extent_size = BTRFS_MAX_EXTENT_SIZE; > + u64 num_extents; > + > + if (flag == 1) > + max_extent_size = SZ_128K; > + > + num_extents = div64_u64(end - start + max_extent_size, > + max_extent_size); > + > + /* compression path */ > + if (flag == 1) > + bits = EXTENT_DELALLOC | EXTENT_COMPRESS | EXTENT_UPTODATE; > + else > + bits = EXTENT_DELALLOC | EXTENT_UPTODATE; > > WARN_ON((end & (PAGE_SIZE - 1)) == 0); > - ret = set_extent_delalloc(&BTRFS_I(inode)->io_tree, start, end, > - cached_state); > + ret = set_extent_bit(&BTRFS_I(inode)->io_tree, start, end, > + bits, NULL, cached_state, GFP_NOFS); > > /* > * btrfs_delalloc_reserve_metadata() will first add number of > @@ -2038,16 +2099,28 @@ int btrfs_set_extent_delalloc(struct inode *inode, u64 start, u64 end, > } > > int btrfs_set_extent_defrag(struct inode *inode, u64 start, u64 end, > - struct extent_state **cached_state) > + struct extent_state **cached_state, int flag) > { > int ret; > - u64 num_extents = div64_u64(end - start + BTRFS_MAX_EXTENT_SIZE, > - BTRFS_MAX_EXTENT_SIZE); > + u64 max_extent_size = BTRFS_MAX_EXTENT_SIZE; > + u64 num_extents; > + unsigned bits; > + > + if (flag == 1) > + max_extent_size = SZ_128K; > + > + num_extents = div64_u64(end - start + max_extent_size, > + max_extent_size); > > WARN_ON((end & (PAGE_SIZE - 1)) == 0); > - ret = set_extent_defrag(&BTRFS_I(inode)->io_tree, start, end, > - cached_state); > + if (flag == 1) > + bits = EXTENT_DELALLOC | EXTENT_UPTODATE | EXTENT_DEFRAG | > + EXTENT_COMPRESS; > + else > + bits = EXTENT_DELALLOC | EXTENT_UPTODATE | EXTENT_DEFRAG; > > + ret = set_extent_bit(&BTRFS_I(inode)->io_tree, start, end, > + bits, NULL, cached_state, GFP_NOFS); > if (ret == 0 && !btrfs_is_free_space_inode(inode)) { > spin_lock(&BTRFS_I(inode)->lock); > BTRFS_I(inode)->outstanding_extents -= num_extents; > @@ -2073,6 +2146,7 @@ static void btrfs_writepage_fixup_worker(struct btrfs_work *work) > u64 page_start; > u64 page_end; > int ret; > + enum btrfs_metadata_reserve_type reserve_type = BTRFS_RESERVE_NORMAL; > > fixup = container_of(work, struct btrfs_writepage_fixup, work); > page = fixup->page; > @@ -2105,8 +2179,10 @@ static void btrfs_writepage_fixup_worker(struct btrfs_work *work) > goto again; > } > > + if (inode_need_compress(inode)) > + reserve_type = BTRFS_RESERVE_COMPRESS; > ret = btrfs_delalloc_reserve_space(inode, page_start, > - PAGE_SIZE); > + PAGE_SIZE, reserve_type); > if (ret) { > mapping_set_error(page->mapping, ret); > end_extent_writepage(page, ret, page_start, page_end); > @@ -2115,7 +2191,7 @@ static void btrfs_writepage_fixup_worker(struct btrfs_work *work) > } > > btrfs_set_extent_delalloc(inode, page_start, page_end, &cached_state, > - 0); > + reserve_type); > ClearPageChecked(page); > set_page_dirty(page); > out: > @@ -2925,6 +3001,7 @@ static int btrfs_finish_ordered_io(struct btrfs_ordered_extent *ordered_extent) > u64 logical_len = ordered_extent->len; > bool nolock; > bool truncated = false; > + enum btrfs_metadata_reserve_type reserve_type = BTRFS_RESERVE_NORMAL; > > nolock = btrfs_is_free_space_inode(inode); > > @@ -3002,8 +3079,11 @@ static int btrfs_finish_ordered_io(struct btrfs_ordered_extent *ordered_extent) > > trans->block_rsv = &root->fs_info->delalloc_block_rsv; > > - if (test_bit(BTRFS_ORDERED_COMPRESSED, &ordered_extent->flags)) > + if (test_bit(BTRFS_ORDERED_COMPRESSED, &ordered_extent->flags)) { > compress_type = ordered_extent->compress_type; > + reserve_type = BTRFS_RESERVE_COMPRESS; > + } > + > if (test_bit(BTRFS_ORDERED_PREALLOC, &ordered_extent->flags)) { > BUG_ON(compress_type); > ret = btrfs_mark_extent_written(trans, inode, > @@ -3048,7 +3128,8 @@ static int btrfs_finish_ordered_io(struct btrfs_ordered_extent *ordered_extent) > ordered_extent->len - 1, &cached_state, GFP_NOFS); > out: > if (root != root->fs_info->tree_root) > - btrfs_delalloc_release_metadata(inode, ordered_extent->len); > + btrfs_delalloc_release_metadata(inode, ordered_extent->len, > + reserve_type); > if (trans) > btrfs_end_transaction(trans, root); > > @@ -4762,13 +4843,17 @@ int btrfs_truncate_block(struct inode *inode, loff_t from, loff_t len, > int ret = 0; > u64 block_start; > u64 block_end; > + enum btrfs_metadata_reserve_type reserve_type = BTRFS_RESERVE_NORMAL; > + > + if (inode_need_compress(inode)) > + reserve_type = BTRFS_RESERVE_COMPRESS; > > if ((offset & (blocksize - 1)) == 0 && > (!len || ((len & (blocksize - 1)) == 0))) > goto out; > > ret = btrfs_delalloc_reserve_space(inode, > - round_down(from, blocksize), blocksize); > + round_down(from, blocksize), blocksize, reserve_type); > if (ret) > goto out; > > @@ -4777,7 +4862,7 @@ int btrfs_truncate_block(struct inode *inode, loff_t from, loff_t len, > if (!page) { > btrfs_delalloc_release_space(inode, > round_down(from, blocksize), > - blocksize); > + blocksize, reserve_type); > ret = -ENOMEM; > goto out; > } > @@ -4820,7 +4905,7 @@ int btrfs_truncate_block(struct inode *inode, loff_t from, loff_t len, > 0, 0, &cached_state, GFP_NOFS); > > ret = btrfs_set_extent_delalloc(inode, block_start, block_end, > - &cached_state, 0); > + &cached_state, reserve_type); > if (ret) { > unlock_extent_cached(io_tree, block_start, block_end, > &cached_state, GFP_NOFS); > @@ -4848,7 +4933,7 @@ int btrfs_truncate_block(struct inode *inode, loff_t from, loff_t len, > out_unlock: > if (ret) > btrfs_delalloc_release_space(inode, block_start, > - blocksize); > + blocksize, reserve_type); > unlock_page(page); > put_page(page); > out: > @@ -8744,7 +8829,8 @@ static ssize_t btrfs_direct_IO(struct kiocb *iocb, struct iov_iter *iter) > inode_unlock(inode); > relock = true; > } > - ret = btrfs_delalloc_reserve_space(inode, offset, count); > + ret = btrfs_delalloc_reserve_space(inode, offset, count, > + BTRFS_RESERVE_NORMAL); > if (ret) > goto out; > dio_data.outstanding_extents = div64_u64(count + > @@ -8776,7 +8862,7 @@ static ssize_t btrfs_direct_IO(struct kiocb *iocb, struct iov_iter *iter) > if (ret < 0 && ret != -EIOCBQUEUED) { > if (dio_data.reserve) > btrfs_delalloc_release_space(inode, offset, > - dio_data.reserve); > + dio_data.reserve, BTRFS_RESERVE_NORMAL); > /* > * On error we might have left some ordered extents > * without submitting corresponding bios for them, so > @@ -8792,7 +8878,7 @@ static ssize_t btrfs_direct_IO(struct kiocb *iocb, struct iov_iter *iter) > 0); > } else if (ret >= 0 && (size_t)ret < count) > btrfs_delalloc_release_space(inode, offset, > - count - (size_t)ret); > + count - (size_t)ret, BTRFS_RESERVE_NORMAL); > } > out: > if (wakeup) > @@ -9035,6 +9121,7 @@ int btrfs_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf) > u64 page_start; > u64 page_end; > u64 end; > + enum btrfs_metadata_reserve_type reserve_type = BTRFS_RESERVE_NORMAL; > > reserved_space = PAGE_SIZE; > > @@ -9043,6 +9130,8 @@ int btrfs_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf) > page_end = page_start + PAGE_SIZE - 1; > end = page_end; > > + if (inode_need_compress(inode)) > + reserve_type = BTRFS_RESERVE_COMPRESS; > /* > * Reserving delalloc space after obtaining the page lock can lead to > * deadlock. For example, if a dirty page is locked by this function > @@ -9052,7 +9141,7 @@ int btrfs_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf) > * being processed by btrfs_page_mkwrite() function. > */ > ret = btrfs_delalloc_reserve_space(inode, page_start, > - reserved_space); > + reserved_space, reserve_type); > if (!ret) { > ret = file_update_time(vma->vm_file); > reserved = 1; > @@ -9104,7 +9193,8 @@ int btrfs_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf) > BTRFS_I(inode)->outstanding_extents++; > spin_unlock(&BTRFS_I(inode)->lock); > btrfs_delalloc_release_space(inode, page_start, > - PAGE_SIZE - reserved_space); > + PAGE_SIZE - reserved_space, > + reserve_type); > } > } > > @@ -9121,7 +9211,7 @@ int btrfs_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf) > 0, 0, &cached_state, GFP_NOFS); > > ret = btrfs_set_extent_delalloc(inode, page_start, end, > - &cached_state, 0); > + &cached_state, reserve_type); > if (ret) { > unlock_extent_cached(io_tree, page_start, page_end, > &cached_state, GFP_NOFS); > @@ -9159,7 +9249,8 @@ int btrfs_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf) > } > unlock_page(page); > out: > - btrfs_delalloc_release_space(inode, page_start, reserved_space); > + btrfs_delalloc_release_space(inode, page_start, reserved_space, > + reserve_type); > out_noreserve: > sb_end_pagefault(inode->i_sb); > return ret; > diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c > index fdf1901..91d121c 100644 > --- a/fs/btrfs/ioctl.c > +++ b/fs/btrfs/ioctl.c > @@ -1132,6 +1132,7 @@ static int cluster_pages_for_defrag(struct inode *inode, > struct extent_state *cached_state = NULL; > struct extent_io_tree *tree; > gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping); > + enum btrfs_metadata_reserve_type reserve_type = BTRFS_RESERVE_NORMAL; > > file_end = (isize - 1) >> PAGE_SHIFT; > if (!isize || start_index > file_end) > @@ -1139,9 +1140,11 @@ static int cluster_pages_for_defrag(struct inode *inode, > > page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1); > > + if (inode_need_compress(inode)) > + reserve_type = BTRFS_RESERVE_COMPRESS; > ret = btrfs_delalloc_reserve_space(inode, > start_index << PAGE_SHIFT, > - page_cnt << PAGE_SHIFT); > + page_cnt << PAGE_SHIFT, reserve_type); > if (ret) > return ret; > i_done = 0; > @@ -1232,11 +1235,12 @@ static int cluster_pages_for_defrag(struct inode *inode, > spin_unlock(&BTRFS_I(inode)->lock); > btrfs_delalloc_release_space(inode, > start_index << PAGE_SHIFT, > - (page_cnt - i_done) << PAGE_SHIFT); > + (page_cnt - i_done) << PAGE_SHIFT, > + reserve_type); > } > > btrfs_set_extent_defrag(inode, page_start, > - page_end - 1, &cached_state); > + page_end - 1, &cached_state, reserve_type); > unlock_extent_cached(&BTRFS_I(inode)->io_tree, > page_start, page_end - 1, &cached_state, > GFP_NOFS); > @@ -1257,7 +1261,7 @@ static int cluster_pages_for_defrag(struct inode *inode, > } > btrfs_delalloc_release_space(inode, > start_index << PAGE_SHIFT, > - page_cnt << PAGE_SHIFT); > + page_cnt << PAGE_SHIFT, reserve_type); > return ret; > > } > diff --git a/fs/btrfs/relocation.c b/fs/btrfs/relocation.c > index 0ec8ffa..db90b47 100644 > --- a/fs/btrfs/relocation.c > +++ b/fs/btrfs/relocation.c > @@ -3144,10 +3144,14 @@ static int relocate_file_extent_cluster(struct inode *inode, > gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping); > int nr = 0; > int ret = 0; > + enum btrfs_metadata_reserve_type reserve_type = BTRFS_RESERVE_NORMAL; > > if (!cluster->nr) > return 0; > > + if (inode_need_compress(inode)) > + reserve_type = BTRFS_RESERVE_COMPRESS; > + > ra = kzalloc(sizeof(*ra), GFP_NOFS); > if (!ra) > return -ENOMEM; > @@ -3166,7 +3170,8 @@ static int relocate_file_extent_cluster(struct inode *inode, > index = (cluster->start - offset) >> PAGE_SHIFT; > last_index = (cluster->end - offset) >> PAGE_SHIFT; > while (index <= last_index) { > - ret = btrfs_delalloc_reserve_metadata(inode, PAGE_SIZE); > + ret = btrfs_delalloc_reserve_metadata(inode, PAGE_SIZE, > + reserve_type); > if (ret) > goto out; > > @@ -3179,7 +3184,7 @@ static int relocate_file_extent_cluster(struct inode *inode, > mask); > if (!page) { > btrfs_delalloc_release_metadata(inode, > - PAGE_SIZE); > + PAGE_SIZE, reserve_type); > ret = -ENOMEM; > goto out; > } > @@ -3198,7 +3203,7 @@ static int relocate_file_extent_cluster(struct inode *inode, > unlock_page(page); > put_page(page); > btrfs_delalloc_release_metadata(inode, > - PAGE_SIZE); > + PAGE_SIZE, reserve_type); > ret = -EIO; > goto out; > } > @@ -3219,7 +3224,8 @@ static int relocate_file_extent_cluster(struct inode *inode, > nr++; > } > > - btrfs_set_extent_delalloc(inode, page_start, page_end, NULL, 0); > + btrfs_set_extent_delalloc(inode, page_start, page_end, NULL, > + reserve_type); > set_page_dirty(page); > > unlock_extent(&BTRFS_I(inode)->io_tree, > diff --git a/fs/btrfs/tests/inode-tests.c b/fs/btrfs/tests/inode-tests.c > index 0bf4680..9a1a01d 100644 > --- a/fs/btrfs/tests/inode-tests.c > +++ b/fs/btrfs/tests/inode-tests.c > @@ -943,6 +943,7 @@ static int test_extent_accounting(u32 sectorsize, u32 nodesize) > struct inode *inode = NULL; > struct btrfs_root *root = NULL; > int ret = -ENOMEM; > + enum btrfs_metadata_reserve_type reserve_type = BTRFS_RESERVE_NORMAL; > > inode = btrfs_new_test_inode(); > if (!inode) { > @@ -968,7 +969,7 @@ static int test_extent_accounting(u32 sectorsize, u32 nodesize) > /* [BTRFS_MAX_EXTENT_SIZE] */ > BTRFS_I(inode)->outstanding_extents++; > ret = btrfs_set_extent_delalloc(inode, 0, BTRFS_MAX_EXTENT_SIZE - 1, > - NULL, 0); > + NULL, reserve_type); > if (ret) { > test_msg("btrfs_set_extent_delalloc returned %d\n", ret); > goto out; > @@ -984,7 +985,7 @@ static int test_extent_accounting(u32 sectorsize, u32 nodesize) > BTRFS_I(inode)->outstanding_extents++; > ret = btrfs_set_extent_delalloc(inode, BTRFS_MAX_EXTENT_SIZE, > BTRFS_MAX_EXTENT_SIZE + sectorsize - 1, > - NULL, 0); > + NULL, reserve_type); > if (ret) { > test_msg("btrfs_set_extent_delalloc returned %d\n", ret); > goto out; > @@ -1019,7 +1020,7 @@ static int test_extent_accounting(u32 sectorsize, u32 nodesize) > ret = btrfs_set_extent_delalloc(inode, BTRFS_MAX_EXTENT_SIZE >> 1, > (BTRFS_MAX_EXTENT_SIZE >> 1) > + sectorsize - 1, > - NULL, 0); > + NULL, reserve_type); > if (ret) { > test_msg("btrfs_set_extent_delalloc returned %d\n", ret); > goto out; > @@ -1042,7 +1043,7 @@ static int test_extent_accounting(u32 sectorsize, u32 nodesize) > ret = btrfs_set_extent_delalloc(inode, > BTRFS_MAX_EXTENT_SIZE + 2 * sectorsize, > (BTRFS_MAX_EXTENT_SIZE << 1) + 3 * sectorsize - 1, > - NULL, 0); > + NULL, reserve_type); > if (ret) { > test_msg("btrfs_set_extent_delalloc returned %d\n", ret); > goto out; > @@ -1060,7 +1061,8 @@ static int test_extent_accounting(u32 sectorsize, u32 nodesize) > BTRFS_I(inode)->outstanding_extents++; > ret = btrfs_set_extent_delalloc(inode, > BTRFS_MAX_EXTENT_SIZE + sectorsize, > - BTRFS_MAX_EXTENT_SIZE + 2 * sectorsize - 1, NULL, 0); > + BTRFS_MAX_EXTENT_SIZE + 2 * sectorsize - 1, > + NULL, reserve_type); > if (ret) { > test_msg("btrfs_set_extent_delalloc returned %d\n", ret); > goto out; > @@ -1097,7 +1099,8 @@ static int test_extent_accounting(u32 sectorsize, u32 nodesize) > BTRFS_I(inode)->outstanding_extents++; > ret = btrfs_set_extent_delalloc(inode, > BTRFS_MAX_EXTENT_SIZE + sectorsize, > - BTRFS_MAX_EXTENT_SIZE + 2 * sectorsize - 1, NULL, 0); > + BTRFS_MAX_EXTENT_SIZE + 2 * sectorsize - 1, > + NULL, reserve_type); > if (ret) { > test_msg("btrfs_set_extent_delalloc returned %d\n", ret); > goto out; -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 52a8535..457b106 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -98,6 +98,19 @@ static const int btrfs_csum_sizes[] = { 4 }; #define BTRFS_DIRTY_METADATA_THRESH SZ_32M +/* + * for compression, max file extent size would be limited to 128K, so when + * reserving metadata for such delalloc writes, pass BTRFS_RESERVE_COMPRESS to + * btrfs_delalloc_reserve_metadata() or btrfs_delalloc_reserve_space() to + * calculate metadata, for none-compression, use BTRFS_RESERVE_NORMAL. + */ +enum btrfs_metadata_reserve_type { + BTRFS_RESERVE_NORMAL, + BTRFS_RESERVE_COMPRESS, +}; +int inode_need_compress(struct inode *inode); +u64 btrfs_max_extent_size(enum btrfs_metadata_reserve_type reserve_type); + #define BTRFS_MAX_EXTENT_SIZE SZ_128M struct btrfs_mapping_tree { @@ -2693,10 +2706,14 @@ int btrfs_subvolume_reserve_metadata(struct btrfs_root *root, void btrfs_subvolume_release_metadata(struct btrfs_root *root, struct btrfs_block_rsv *rsv, u64 qgroup_reserved); -int btrfs_delalloc_reserve_metadata(struct inode *inode, u64 num_bytes); -void btrfs_delalloc_release_metadata(struct inode *inode, u64 num_bytes); -int btrfs_delalloc_reserve_space(struct inode *inode, u64 start, u64 len); -void btrfs_delalloc_release_space(struct inode *inode, u64 start, u64 len); +int btrfs_delalloc_reserve_metadata(struct inode *inode, u64 num_bytes, + enum btrfs_metadata_reserve_type reserve_type); +void btrfs_delalloc_release_metadata(struct inode *inode, u64 num_bytes, + enum btrfs_metadata_reserve_type reserve_type); +int btrfs_delalloc_reserve_space(struct inode *inode, u64 start, u64 len, + enum btrfs_metadata_reserve_type reserve_type); +void btrfs_delalloc_release_space(struct inode *inode, u64 start, u64 len, + enum btrfs_metadata_reserve_type reserve_type); void btrfs_init_block_rsv(struct btrfs_block_rsv *rsv, unsigned short type); struct btrfs_block_rsv *btrfs_alloc_block_rsv(struct btrfs_root *root, unsigned short type); @@ -3138,9 +3155,9 @@ int btrfs_start_delalloc_inodes(struct btrfs_root *root, int delay_iput); int btrfs_start_delalloc_roots(struct btrfs_fs_info *fs_info, int delay_iput, int nr); int btrfs_set_extent_delalloc(struct inode *inode, u64 start, u64 end, - struct extent_state **cached_state, int dedupe); + struct extent_state **cached_state, int flag); int btrfs_set_extent_defrag(struct inode *inode, u64 start, u64 end, - struct extent_state **cached_state); + struct extent_state **cached_state, int flag); int btrfs_create_subvol_root(struct btrfs_trans_handle *trans, struct btrfs_root *new_root, struct btrfs_root *parent_root, @@ -3232,7 +3249,7 @@ int btrfs_release_file(struct inode *inode, struct file *file); int btrfs_dirty_pages(struct btrfs_root *root, struct inode *inode, struct page **pages, size_t num_pages, loff_t pos, size_t write_bytes, - struct extent_state **cached); + struct extent_state **cached, int flag); int btrfs_fdatawrite_range(struct inode *inode, loff_t start, loff_t end); ssize_t btrfs_copy_file_range(struct file *file_in, loff_t pos_in, struct file *file_out, loff_t pos_out, diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 210c94a..9aa6d2c 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -5838,15 +5838,16 @@ void btrfs_subvolume_release_metadata(struct btrfs_root *root, * reserved extents that need to be freed. This must be called with * BTRFS_I(inode)->lock held. */ -static unsigned drop_outstanding_extent(struct inode *inode, u64 num_bytes) +static unsigned drop_outstanding_extent(struct inode *inode, u64 num_bytes, + enum btrfs_metadata_reserve_type reserve_type) { unsigned drop_inode_space = 0; unsigned dropped_extents = 0; unsigned num_extents = 0; + u64 max_extent_size = btrfs_max_extent_size(reserve_type); - num_extents = (unsigned)div64_u64(num_bytes + - BTRFS_MAX_EXTENT_SIZE - 1, - BTRFS_MAX_EXTENT_SIZE); + num_extents = (unsigned)div64_u64(num_bytes + max_extent_size - 1, + max_extent_size); ASSERT(num_extents); ASSERT(BTRFS_I(inode)->outstanding_extents >= num_extents); BTRFS_I(inode)->outstanding_extents -= num_extents; @@ -5916,7 +5917,21 @@ static u64 calc_csum_metadata_size(struct inode *inode, u64 num_bytes, return btrfs_calc_trans_metadata_size(root, old_csums - num_csums); } -int btrfs_delalloc_reserve_metadata(struct inode *inode, u64 num_bytes) +u64 btrfs_max_extent_size(enum btrfs_metadata_reserve_type reserve_type) +{ + if (reserve_type == BTRFS_RESERVE_COMPRESS) + return SZ_128K; + + return BTRFS_MAX_EXTENT_SIZE; +} + +/* + * @reserve_type: normally reserve_type should be BTRFS_RESERVE_NORMAL, but for + * compression path, its max extent size is limited to 128KB, not 128MB, when + * reserving metadata, we should set reserve_type to BTRFS_RESERVE_COMPRESS. + */ +int btrfs_delalloc_reserve_metadata(struct inode *inode, u64 num_bytes, + enum btrfs_metadata_reserve_type reserve_type) { struct btrfs_root *root = BTRFS_I(inode)->root; struct btrfs_block_rsv *block_rsv = &root->fs_info->delalloc_block_rsv; @@ -5929,6 +5944,7 @@ int btrfs_delalloc_reserve_metadata(struct inode *inode, u64 num_bytes) u64 to_free = 0; unsigned dropped; bool release_extra = false; + u64 max_extent_size = btrfs_max_extent_size(reserve_type); /* If we are a free space inode we need to not flush since we will be in * the middle of a transaction commit. We also don't need the delalloc @@ -5955,9 +5971,8 @@ int btrfs_delalloc_reserve_metadata(struct inode *inode, u64 num_bytes) num_bytes = ALIGN(num_bytes, root->sectorsize); spin_lock(&BTRFS_I(inode)->lock); - nr_extents = (unsigned)div64_u64(num_bytes + - BTRFS_MAX_EXTENT_SIZE - 1, - BTRFS_MAX_EXTENT_SIZE); + nr_extents = (unsigned)div64_u64(num_bytes + max_extent_size - 1, + max_extent_size); BTRFS_I(inode)->outstanding_extents += nr_extents; nr_extents = 0; @@ -6008,7 +6023,7 @@ int btrfs_delalloc_reserve_metadata(struct inode *inode, u64 num_bytes) out_fail: spin_lock(&BTRFS_I(inode)->lock); - dropped = drop_outstanding_extent(inode, num_bytes); + dropped = drop_outstanding_extent(inode, num_bytes, reserve_type); /* * If the inodes csum_bytes is the same as the original * csum_bytes then we know we haven't raced with any free()ers @@ -6074,12 +6089,15 @@ int btrfs_delalloc_reserve_metadata(struct inode *inode, u64 num_bytes) * btrfs_delalloc_release_metadata - release a metadata reservation for an inode * @inode: the inode to release the reservation for * @num_bytes: the number of bytes we're releasing + * @reserve_type: this value must be same to the value passing to + * btrfs_delalloc_reserve_metadata(). * * This will release the metadata reservation for an inode. This can be called * once we complete IO for a given set of bytes to release their metadata * reservations. */ -void btrfs_delalloc_release_metadata(struct inode *inode, u64 num_bytes) +void btrfs_delalloc_release_metadata(struct inode *inode, u64 num_bytes, + enum btrfs_metadata_reserve_type reserve_type) { struct btrfs_root *root = BTRFS_I(inode)->root; u64 to_free = 0; @@ -6087,7 +6105,7 @@ void btrfs_delalloc_release_metadata(struct inode *inode, u64 num_bytes) num_bytes = ALIGN(num_bytes, root->sectorsize); spin_lock(&BTRFS_I(inode)->lock); - dropped = drop_outstanding_extent(inode, num_bytes); + dropped = drop_outstanding_extent(inode, num_bytes, reserve_type); if (num_bytes) to_free = calc_csum_metadata_size(inode, num_bytes, 0); @@ -6111,6 +6129,9 @@ void btrfs_delalloc_release_metadata(struct inode *inode, u64 num_bytes) * @inode: inode we're writing to * @start: start range we are writing to * @len: how long the range we are writing to + * @reserve_type: normally reserve_type should be BTRFS_RESERVE_NORMAL, but for + * compression path, its max extent size is limited to 128KB, not 128MB, when + * reserving metadata, we should set reserve_type to BTRFS_RESERVE_COMPRESS. * * This will do the following things * @@ -6128,14 +6149,15 @@ void btrfs_delalloc_release_metadata(struct inode *inode, u64 num_bytes) * Return 0 for success * Return <0 for error(-ENOSPC or -EQUOT) */ -int btrfs_delalloc_reserve_space(struct inode *inode, u64 start, u64 len) +int btrfs_delalloc_reserve_space(struct inode *inode, u64 start, u64 len, + enum btrfs_metadata_reserve_type reserve_type) { int ret; ret = btrfs_check_data_free_space(inode, start, len); if (ret < 0) return ret; - ret = btrfs_delalloc_reserve_metadata(inode, len); + ret = btrfs_delalloc_reserve_metadata(inode, len, reserve_type); if (ret < 0) btrfs_free_reserved_data_space(inode, start, len); return ret; @@ -6146,6 +6168,8 @@ int btrfs_delalloc_reserve_space(struct inode *inode, u64 start, u64 len) * @inode: inode we're releasing space for * @start: start position of the space already reserved * @len: the len of the space already reserved + * @reserve_type: this value must be same to the value passing to + * btrfs_delalloc_reserve_space(). * * This must be matched with a call to btrfs_delalloc_reserve_space. This is * called in the case that we don't need the metadata AND data reservations @@ -6156,9 +6180,10 @@ int btrfs_delalloc_reserve_space(struct inode *inode, u64 start, u64 len) * list if there are no delalloc bytes left. * Also it will handle the qgroup reserved space. */ -void btrfs_delalloc_release_space(struct inode *inode, u64 start, u64 len) +void btrfs_delalloc_release_space(struct inode *inode, u64 start, u64 len, + enum btrfs_metadata_reserve_type reserve_type) { - btrfs_delalloc_release_metadata(inode, len); + btrfs_delalloc_release_metadata(inode, len, reserve_type); btrfs_free_reserved_data_space(inode, start, len); } diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 66a7551..8c07306 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -603,7 +603,7 @@ static int __clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, btrfs_debug_check_extent_io_range(tree, start, end); if (bits & EXTENT_DELALLOC) - bits |= EXTENT_NORESERVE; + bits |= EXTENT_NORESERVE | EXTENT_COMPRESS; if (delete) bits |= ~EXTENT_CTLBITS; @@ -742,6 +742,58 @@ static int __clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, } +static void adjust_one_outstanding_extent(struct inode *inode, u64 len) +{ + unsigned old_extents, new_extents; + + old_extents = div64_u64(len + SZ_128K - 1, SZ_128K); + new_extents = div64_u64(len + BTRFS_MAX_EXTENT_SIZE - 1, + BTRFS_MAX_EXTENT_SIZE); + if (old_extents <= new_extents) + return; + + spin_lock(&BTRFS_I(inode)->lock); + BTRFS_I(inode)->outstanding_extents -= old_extents - new_extents; + spin_unlock(&BTRFS_I(inode)->lock); +} + +/* + * For a extent with EXTENT_COMPRESS flag, if later it does not go through + * compress path, we need to adjust the number of outstanding_extents. + * It's because for extent with EXTENT_COMPRESS flag, its number of outstanding + * extents is calculated by 128KB, so here we need to adjust it. + */ +void adjust_outstanding_extents(struct inode *inode, + u64 start, u64 end) +{ + struct rb_node *node; + struct extent_state *state; + struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree; + + spin_lock(&tree->lock); + node = tree_search(tree, start); + if (!node) + goto out; + + while (1) { + state = rb_entry(node, struct extent_state, rb_node); + if (state->start > end) + goto out; + /* + * The whole range is locked, so we can safely clear + * EXTENT_COMPRESS flag. + */ + state->state &= ~EXTENT_COMPRESS; + adjust_one_outstanding_extent(inode, + state->end - state->start + 1); + node = rb_next(node); + if (!node) + break; + } +out: + spin_unlock(&tree->lock); +} + static void wait_on_state(struct extent_io_tree *tree, struct extent_state *state) __releases(tree->lock) @@ -1504,6 +1556,7 @@ static noinline u64 find_delalloc_range(struct extent_io_tree *tree, u64 cur_start = *start; u64 found = 0; u64 total_bytes = 0; + unsigned pre_state; spin_lock(&tree->lock); @@ -1521,7 +1574,8 @@ static noinline u64 find_delalloc_range(struct extent_io_tree *tree, while (1) { state = rb_entry(node, struct extent_state, rb_node); if (found && (state->start != cur_start || - (state->state & EXTENT_BOUNDARY))) { + (state->state & EXTENT_BOUNDARY) || + (state->state ^ pre_state) & EXTENT_COMPRESS)) { goto out; } if (!(state->state & EXTENT_DELALLOC)) { @@ -1537,6 +1591,7 @@ static noinline u64 find_delalloc_range(struct extent_io_tree *tree, found++; *end = state->end; cur_start = state->end + 1; + pre_state = state->state; node = rb_next(node); total_bytes += state->end - state->start + 1; if (total_bytes >= max_bytes) diff --git a/fs/btrfs/extent_io.h b/fs/btrfs/extent_io.h index ab31d14..d9f846c 100644 --- a/fs/btrfs/extent_io.h +++ b/fs/btrfs/extent_io.h @@ -21,6 +21,7 @@ #define EXTENT_NORESERVE (1U << 15) #define EXTENT_QGROUP_RESERVED (1U << 16) #define EXTENT_CLEAR_DATA_RESV (1U << 17) +#define EXTENT_COMPRESS (1U << 18) #define EXTENT_IOBITS (EXTENT_LOCKED | EXTENT_WRITEBACK) #define EXTENT_CTLBITS (EXTENT_DO_ACCOUNTING | EXTENT_FIRST_DELALLOC) @@ -247,6 +248,7 @@ int clear_record_extent_bits(struct extent_io_tree *tree, u64 start, u64 end, int clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, unsigned bits, int wake, int delete, struct extent_state **cached, gfp_t mask); +void adjust_outstanding_extents(struct inode *inode, u64 start, u64 end); static inline int unlock_extent(struct extent_io_tree *tree, u64 start, u64 end) { diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 3a14c87..e4c998d 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -484,11 +484,15 @@ static void btrfs_drop_pages(struct page **pages, size_t num_pages) * * this also makes the decision about creating an inline extent vs * doing real data extents, marking pages dirty and delalloc as required. + * + * normally flag should be 0, but if a data range will go through compress path, + * set flag to 1. Note: here we should ensure enum btrfs_metadata_reserve_type + * and flag's values are consistent. */ int btrfs_dirty_pages(struct btrfs_root *root, struct inode *inode, struct page **pages, size_t num_pages, loff_t pos, size_t write_bytes, - struct extent_state **cached) + struct extent_state **cached, int flag) { int err = 0; int i; @@ -503,7 +507,7 @@ int btrfs_dirty_pages(struct btrfs_root *root, struct inode *inode, end_of_last_block = start_pos + num_bytes - 1; err = btrfs_set_extent_delalloc(inode, start_pos, end_of_last_block, - cached, 0); + cached, flag); if (err) return err; @@ -1521,6 +1525,7 @@ static noinline ssize_t __btrfs_buffered_write(struct file *file, bool only_release_metadata = false; bool force_page_uptodate = false; bool need_unlock; + enum btrfs_metadata_reserve_type reserve_type = BTRFS_RESERVE_NORMAL; nrptrs = min(DIV_ROUND_UP(iov_iter_count(i), PAGE_SIZE), PAGE_SIZE / (sizeof(struct page *))); @@ -1530,6 +1535,9 @@ static noinline ssize_t __btrfs_buffered_write(struct file *file, if (!pages) return -ENOMEM; + if (inode_need_compress(inode)) + reserve_type = BTRFS_RESERVE_COMPRESS; + while (iov_iter_count(i) > 0) { size_t offset = pos & (PAGE_SIZE - 1); size_t sector_offset; @@ -1583,7 +1591,8 @@ static noinline ssize_t __btrfs_buffered_write(struct file *file, } } - ret = btrfs_delalloc_reserve_metadata(inode, reserve_bytes); + ret = btrfs_delalloc_reserve_metadata(inode, reserve_bytes, + reserve_type); if (ret) { if (!only_release_metadata) btrfs_free_reserved_data_space(inode, pos, @@ -1666,14 +1675,16 @@ static noinline ssize_t __btrfs_buffered_write(struct file *file, } if (only_release_metadata) { btrfs_delalloc_release_metadata(inode, - release_bytes); + release_bytes, + reserve_type); } else { u64 __pos; __pos = round_down(pos, root->sectorsize) + (dirty_pages << PAGE_SHIFT); btrfs_delalloc_release_space(inode, __pos, - release_bytes); + release_bytes, + reserve_type); } } @@ -1683,7 +1694,7 @@ static noinline ssize_t __btrfs_buffered_write(struct file *file, if (copied > 0) ret = btrfs_dirty_pages(root, inode, pages, dirty_pages, pos, copied, - NULL); + NULL, reserve_type); if (need_unlock) unlock_extent_cached(&BTRFS_I(inode)->io_tree, lockstart, lockend, &cached_state, @@ -1724,11 +1735,12 @@ static noinline ssize_t __btrfs_buffered_write(struct file *file, if (release_bytes) { if (only_release_metadata) { btrfs_end_write_no_snapshoting(root); - btrfs_delalloc_release_metadata(inode, release_bytes); + btrfs_delalloc_release_metadata(inode, release_bytes, + reserve_type); } else { btrfs_delalloc_release_space(inode, round_down(pos, root->sectorsize), - release_bytes); + release_bytes, reserve_type); } } diff --git a/fs/btrfs/free-space-cache.c b/fs/btrfs/free-space-cache.c index e4b48f3..e8947c2 100644 --- a/fs/btrfs/free-space-cache.c +++ b/fs/btrfs/free-space-cache.c @@ -1297,7 +1297,7 @@ static int __btrfs_write_out_cache(struct btrfs_root *root, struct inode *inode, /* Everything is written out, now we dirty the pages in the file. */ ret = btrfs_dirty_pages(root, inode, io_ctl->pages, io_ctl->num_pages, - 0, i_size_read(inode), &cached_state); + 0, i_size_read(inode), &cached_state, 0); if (ret) goto out_nospc; @@ -3516,6 +3516,7 @@ int btrfs_write_out_ino_cache(struct btrfs_root *root, int ret; struct btrfs_io_ctl io_ctl; bool release_metadata = true; + enum btrfs_metadata_reserve_type reserve_type = BTRFS_RESERVE_NORMAL; if (!btrfs_test_opt(root->fs_info, INODE_MAP_CACHE)) return 0; @@ -3536,7 +3537,8 @@ int btrfs_write_out_ino_cache(struct btrfs_root *root, if (ret) { if (release_metadata) - btrfs_delalloc_release_metadata(inode, inode->i_size); + btrfs_delalloc_release_metadata(inode, inode->i_size, + reserve_type); #ifdef DEBUG btrfs_err(root->fs_info, "failed to write free ino cache for root %llu", diff --git a/fs/btrfs/inode-map.c b/fs/btrfs/inode-map.c index d27014b..39c4b1c 100644 --- a/fs/btrfs/inode-map.c +++ b/fs/btrfs/inode-map.c @@ -404,6 +404,7 @@ int btrfs_save_ino_cache(struct btrfs_root *root, int ret; int prealloc; bool retry = false; + enum btrfs_metadata_reserve_type reserve_type = BTRFS_RESERVE_NORMAL; /* only fs tree and subvol/snap needs ino cache */ if (root->root_key.objectid != BTRFS_FS_TREE_OBJECTID && @@ -491,14 +492,14 @@ int btrfs_save_ino_cache(struct btrfs_root *root, /* Just to make sure we have enough space */ prealloc += 8 * PAGE_SIZE; - ret = btrfs_delalloc_reserve_space(inode, 0, prealloc); + ret = btrfs_delalloc_reserve_space(inode, 0, prealloc, reserve_type); if (ret) goto out_put; ret = btrfs_prealloc_file_range_trans(inode, trans, 0, 0, prealloc, prealloc, prealloc, &alloc_hint); if (ret) { - btrfs_delalloc_release_metadata(inode, prealloc); + btrfs_delalloc_release_metadata(inode, prealloc, reserve_type); goto out_put; } diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index e427b81..608fd84 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -315,7 +315,7 @@ static noinline int cow_file_range_inline(struct btrfs_root *root, } set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &BTRFS_I(inode)->runtime_flags); - btrfs_delalloc_release_metadata(inode, end + 1 - start); + btrfs_delalloc_release_metadata(inode, end + 1 - start, 0); btrfs_drop_extent_cache(inode, start, aligned_end - 1, 0); out: /* @@ -371,7 +371,7 @@ static noinline int add_async_extent(struct async_cow *cow, return 0; } -static inline int inode_need_compress(struct inode *inode) +int inode_need_compress(struct inode *inode) { struct btrfs_root *root = BTRFS_I(inode)->root; @@ -710,6 +710,16 @@ static noinline void submit_compressed_extents(struct inode *inode, async_extent->start + async_extent->ram_size - 1); + /* + * We use 128KB as max extent size to calculate number + * of outstanding extents for this extent before, now + * it'll go throuth uncompressed IO, we need to use + * 128MB as max extent size to re-calculate number of + * outstanding extents for this extent. + */ + adjust_outstanding_extents(inode, async_extent->start, + async_extent->start + + async_extent->ram_size - 1); /* allocate blocks */ ret = cow_file_range(inode, async_cow->locked_page, async_extent->start, @@ -1573,14 +1583,24 @@ static int run_delalloc_range(struct inode *inode, struct page *locked_page, { int ret; int force_cow = need_force_cow(inode, start, end); + struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; + int need_compress; + need_compress = test_range_bit(io_tree, start, end, + EXTENT_COMPRESS, 1, NULL); if (BTRFS_I(inode)->flags & BTRFS_INODE_NODATACOW && !force_cow) { + if (need_compress) + adjust_outstanding_extents(inode, start, end); + ret = run_delalloc_nocow(inode, locked_page, start, end, page_started, 1, nr_written); } else if (BTRFS_I(inode)->flags & BTRFS_INODE_PREALLOC && !force_cow) { + if (need_compress) + adjust_outstanding_extents(inode, start, end); + ret = run_delalloc_nocow(inode, locked_page, start, end, page_started, 0, nr_written); - } else if (!inode_need_compress(inode)) { + } else if (!need_compress) { ret = cow_file_range(inode, locked_page, start, end, end, page_started, nr_written, 1, NULL); } else { @@ -1596,6 +1616,7 @@ static void btrfs_split_extent_hook(struct inode *inode, struct extent_state *orig, u64 split) { u64 size; + u64 max_extent_size = BTRFS_MAX_EXTENT_SIZE; /* not delalloc, ignore it */ if (!(orig->state & EXTENT_DELALLOC)) @@ -1604,8 +1625,11 @@ static void btrfs_split_extent_hook(struct inode *inode, if (btrfs_is_free_space_inode(inode)) return; + if (orig->state & EXTENT_COMPRESS) + max_extent_size = SZ_128K; + size = orig->end - orig->start + 1; - if (size > BTRFS_MAX_EXTENT_SIZE) { + if (size > max_extent_size) { u64 num_extents; u64 new_size; @@ -1614,13 +1638,13 @@ static void btrfs_split_extent_hook(struct inode *inode, * applies here, just in reverse. */ new_size = orig->end - split + 1; - num_extents = div64_u64(new_size + BTRFS_MAX_EXTENT_SIZE - 1, - BTRFS_MAX_EXTENT_SIZE); + num_extents = div64_u64(new_size + max_extent_size - 1, + max_extent_size); new_size = split - orig->start; - num_extents += div64_u64(new_size + BTRFS_MAX_EXTENT_SIZE - 1, - BTRFS_MAX_EXTENT_SIZE); - if (div64_u64(size + BTRFS_MAX_EXTENT_SIZE - 1, - BTRFS_MAX_EXTENT_SIZE) >= num_extents) + num_extents += div64_u64(new_size + max_extent_size - 1, + max_extent_size); + if (div64_u64(size + max_extent_size - 1, + max_extent_size) >= num_extents) return; } @@ -1641,6 +1665,7 @@ static void btrfs_merge_extent_hook(struct inode *inode, { u64 new_size, old_size; u64 num_extents; + u64 max_extent_size = BTRFS_MAX_EXTENT_SIZE; /* not delalloc, ignore it */ if (!(other->state & EXTENT_DELALLOC)) @@ -1649,13 +1674,16 @@ static void btrfs_merge_extent_hook(struct inode *inode, if (btrfs_is_free_space_inode(inode)) return; + if (other->state & EXTENT_COMPRESS) + max_extent_size = SZ_128K; + if (new->start > other->start) new_size = new->end - other->start + 1; else new_size = other->end - new->start + 1; /* we're not bigger than the max, unreserve the space and go */ - if (new_size <= BTRFS_MAX_EXTENT_SIZE) { + if (new_size <= max_extent_size) { spin_lock(&BTRFS_I(inode)->lock); BTRFS_I(inode)->outstanding_extents--; spin_unlock(&BTRFS_I(inode)->lock); @@ -1681,14 +1709,14 @@ static void btrfs_merge_extent_hook(struct inode *inode, * this case. */ old_size = other->end - other->start + 1; - num_extents = div64_u64(old_size + BTRFS_MAX_EXTENT_SIZE - 1, - BTRFS_MAX_EXTENT_SIZE); + num_extents = div64_u64(old_size + max_extent_size - 1, + max_extent_size); old_size = new->end - new->start + 1; - num_extents += div64_u64(old_size + BTRFS_MAX_EXTENT_SIZE - 1, - BTRFS_MAX_EXTENT_SIZE); + num_extents += div64_u64(old_size + max_extent_size - 1, + max_extent_size); - if (div64_u64(new_size + BTRFS_MAX_EXTENT_SIZE - 1, - BTRFS_MAX_EXTENT_SIZE) >= num_extents) + if (div64_u64(new_size + max_extent_size - 1, + max_extent_size) >= num_extents) return; spin_lock(&BTRFS_I(inode)->lock); @@ -1754,10 +1782,15 @@ static void btrfs_set_bit_hook(struct inode *inode, if (!(state->state & EXTENT_DELALLOC) && (*bits & EXTENT_DELALLOC)) { struct btrfs_root *root = BTRFS_I(inode)->root; u64 len = state->end + 1 - state->start; - u64 num_extents = div64_u64(len + BTRFS_MAX_EXTENT_SIZE - 1, - BTRFS_MAX_EXTENT_SIZE); + u64 max_extent_size = BTRFS_MAX_EXTENT_SIZE; + u64 num_extents; bool do_list = !btrfs_is_free_space_inode(inode); + if (*bits & EXTENT_COMPRESS) + max_extent_size = SZ_128K; + num_extents = div64_u64(len + max_extent_size - 1, + max_extent_size); + if (*bits & EXTENT_FIRST_DELALLOC) *bits &= ~EXTENT_FIRST_DELALLOC; @@ -1792,8 +1825,9 @@ static void btrfs_clear_bit_hook(struct inode *inode, unsigned *bits) { u64 len = state->end + 1 - state->start; - u64 num_extents = div64_u64(len + BTRFS_MAX_EXTENT_SIZE -1, - BTRFS_MAX_EXTENT_SIZE); + u64 max_extent_size = BTRFS_MAX_EXTENT_SIZE; + u64 num_extents; + enum btrfs_metadata_reserve_type reserve_type = BTRFS_RESERVE_NORMAL; spin_lock(&BTRFS_I(inode)->lock); if ((state->state & EXTENT_DEFRAG) && (*bits & EXTENT_DEFRAG)) @@ -1809,6 +1843,14 @@ static void btrfs_clear_bit_hook(struct inode *inode, struct btrfs_root *root = BTRFS_I(inode)->root; bool do_list = !btrfs_is_free_space_inode(inode); + if (state->state & EXTENT_COMPRESS) { + max_extent_size = SZ_128K; + reserve_type = BTRFS_RESERVE_COMPRESS; + } + + num_extents = div64_u64(len + max_extent_size - 1, + max_extent_size); + if (*bits & EXTENT_FIRST_DELALLOC) { *bits &= ~EXTENT_FIRST_DELALLOC; } else if (!(*bits & EXTENT_DO_ACCOUNTING) && do_list) { @@ -1824,7 +1866,8 @@ static void btrfs_clear_bit_hook(struct inode *inode, */ if (*bits & EXTENT_DO_ACCOUNTING && root != root->fs_info->tree_root) - btrfs_delalloc_release_metadata(inode, len); + btrfs_delalloc_release_metadata(inode, len, + reserve_type); /* For sanity tests. */ if (btrfs_is_testing(root->fs_info)) @@ -2006,16 +2049,34 @@ static noinline int add_pending_csums(struct btrfs_trans_handle *trans, return 0; } +/* + * Normally flag should be 0, but if a data range will go through compress path, + * set flag to 1. Note: here we should ensure enum btrfs_metadata_reserve_type + * and flag's values are consistent. + */ int btrfs_set_extent_delalloc(struct inode *inode, u64 start, u64 end, - struct extent_state **cached_state, int dedupe) + struct extent_state **cached_state, int flag) { int ret; - u64 num_extents = div64_u64(end - start + BTRFS_MAX_EXTENT_SIZE, - BTRFS_MAX_EXTENT_SIZE); + unsigned bits; + u64 max_extent_size = BTRFS_MAX_EXTENT_SIZE; + u64 num_extents; + + if (flag == 1) + max_extent_size = SZ_128K; + + num_extents = div64_u64(end - start + max_extent_size, + max_extent_size); + + /* compression path */ + if (flag == 1) + bits = EXTENT_DELALLOC | EXTENT_COMPRESS | EXTENT_UPTODATE; + else + bits = EXTENT_DELALLOC | EXTENT_UPTODATE; WARN_ON((end & (PAGE_SIZE - 1)) == 0); - ret = set_extent_delalloc(&BTRFS_I(inode)->io_tree, start, end, - cached_state); + ret = set_extent_bit(&BTRFS_I(inode)->io_tree, start, end, + bits, NULL, cached_state, GFP_NOFS); /* * btrfs_delalloc_reserve_metadata() will first add number of @@ -2038,16 +2099,28 @@ int btrfs_set_extent_delalloc(struct inode *inode, u64 start, u64 end, } int btrfs_set_extent_defrag(struct inode *inode, u64 start, u64 end, - struct extent_state **cached_state) + struct extent_state **cached_state, int flag) { int ret; - u64 num_extents = div64_u64(end - start + BTRFS_MAX_EXTENT_SIZE, - BTRFS_MAX_EXTENT_SIZE); + u64 max_extent_size = BTRFS_MAX_EXTENT_SIZE; + u64 num_extents; + unsigned bits; + + if (flag == 1) + max_extent_size = SZ_128K; + + num_extents = div64_u64(end - start + max_extent_size, + max_extent_size); WARN_ON((end & (PAGE_SIZE - 1)) == 0); - ret = set_extent_defrag(&BTRFS_I(inode)->io_tree, start, end, - cached_state); + if (flag == 1) + bits = EXTENT_DELALLOC | EXTENT_UPTODATE | EXTENT_DEFRAG | + EXTENT_COMPRESS; + else + bits = EXTENT_DELALLOC | EXTENT_UPTODATE | EXTENT_DEFRAG; + ret = set_extent_bit(&BTRFS_I(inode)->io_tree, start, end, + bits, NULL, cached_state, GFP_NOFS); if (ret == 0 && !btrfs_is_free_space_inode(inode)) { spin_lock(&BTRFS_I(inode)->lock); BTRFS_I(inode)->outstanding_extents -= num_extents; @@ -2073,6 +2146,7 @@ static void btrfs_writepage_fixup_worker(struct btrfs_work *work) u64 page_start; u64 page_end; int ret; + enum btrfs_metadata_reserve_type reserve_type = BTRFS_RESERVE_NORMAL; fixup = container_of(work, struct btrfs_writepage_fixup, work); page = fixup->page; @@ -2105,8 +2179,10 @@ static void btrfs_writepage_fixup_worker(struct btrfs_work *work) goto again; } + if (inode_need_compress(inode)) + reserve_type = BTRFS_RESERVE_COMPRESS; ret = btrfs_delalloc_reserve_space(inode, page_start, - PAGE_SIZE); + PAGE_SIZE, reserve_type); if (ret) { mapping_set_error(page->mapping, ret); end_extent_writepage(page, ret, page_start, page_end); @@ -2115,7 +2191,7 @@ static void btrfs_writepage_fixup_worker(struct btrfs_work *work) } btrfs_set_extent_delalloc(inode, page_start, page_end, &cached_state, - 0); + reserve_type); ClearPageChecked(page); set_page_dirty(page); out: @@ -2925,6 +3001,7 @@ static int btrfs_finish_ordered_io(struct btrfs_ordered_extent *ordered_extent) u64 logical_len = ordered_extent->len; bool nolock; bool truncated = false; + enum btrfs_metadata_reserve_type reserve_type = BTRFS_RESERVE_NORMAL; nolock = btrfs_is_free_space_inode(inode); @@ -3002,8 +3079,11 @@ static int btrfs_finish_ordered_io(struct btrfs_ordered_extent *ordered_extent) trans->block_rsv = &root->fs_info->delalloc_block_rsv; - if (test_bit(BTRFS_ORDERED_COMPRESSED, &ordered_extent->flags)) + if (test_bit(BTRFS_ORDERED_COMPRESSED, &ordered_extent->flags)) { compress_type = ordered_extent->compress_type; + reserve_type = BTRFS_RESERVE_COMPRESS; + } + if (test_bit(BTRFS_ORDERED_PREALLOC, &ordered_extent->flags)) { BUG_ON(compress_type); ret = btrfs_mark_extent_written(trans, inode, @@ -3048,7 +3128,8 @@ static int btrfs_finish_ordered_io(struct btrfs_ordered_extent *ordered_extent) ordered_extent->len - 1, &cached_state, GFP_NOFS); out: if (root != root->fs_info->tree_root) - btrfs_delalloc_release_metadata(inode, ordered_extent->len); + btrfs_delalloc_release_metadata(inode, ordered_extent->len, + reserve_type); if (trans) btrfs_end_transaction(trans, root); @@ -4762,13 +4843,17 @@ int btrfs_truncate_block(struct inode *inode, loff_t from, loff_t len, int ret = 0; u64 block_start; u64 block_end; + enum btrfs_metadata_reserve_type reserve_type = BTRFS_RESERVE_NORMAL; + + if (inode_need_compress(inode)) + reserve_type = BTRFS_RESERVE_COMPRESS; if ((offset & (blocksize - 1)) == 0 && (!len || ((len & (blocksize - 1)) == 0))) goto out; ret = btrfs_delalloc_reserve_space(inode, - round_down(from, blocksize), blocksize); + round_down(from, blocksize), blocksize, reserve_type); if (ret) goto out; @@ -4777,7 +4862,7 @@ int btrfs_truncate_block(struct inode *inode, loff_t from, loff_t len, if (!page) { btrfs_delalloc_release_space(inode, round_down(from, blocksize), - blocksize); + blocksize, reserve_type); ret = -ENOMEM; goto out; } @@ -4820,7 +4905,7 @@ int btrfs_truncate_block(struct inode *inode, loff_t from, loff_t len, 0, 0, &cached_state, GFP_NOFS); ret = btrfs_set_extent_delalloc(inode, block_start, block_end, - &cached_state, 0); + &cached_state, reserve_type); if (ret) { unlock_extent_cached(io_tree, block_start, block_end, &cached_state, GFP_NOFS); @@ -4848,7 +4933,7 @@ int btrfs_truncate_block(struct inode *inode, loff_t from, loff_t len, out_unlock: if (ret) btrfs_delalloc_release_space(inode, block_start, - blocksize); + blocksize, reserve_type); unlock_page(page); put_page(page); out: @@ -8744,7 +8829,8 @@ static ssize_t btrfs_direct_IO(struct kiocb *iocb, struct iov_iter *iter) inode_unlock(inode); relock = true; } - ret = btrfs_delalloc_reserve_space(inode, offset, count); + ret = btrfs_delalloc_reserve_space(inode, offset, count, + BTRFS_RESERVE_NORMAL); if (ret) goto out; dio_data.outstanding_extents = div64_u64(count + @@ -8776,7 +8862,7 @@ static ssize_t btrfs_direct_IO(struct kiocb *iocb, struct iov_iter *iter) if (ret < 0 && ret != -EIOCBQUEUED) { if (dio_data.reserve) btrfs_delalloc_release_space(inode, offset, - dio_data.reserve); + dio_data.reserve, BTRFS_RESERVE_NORMAL); /* * On error we might have left some ordered extents * without submitting corresponding bios for them, so @@ -8792,7 +8878,7 @@ static ssize_t btrfs_direct_IO(struct kiocb *iocb, struct iov_iter *iter) 0); } else if (ret >= 0 && (size_t)ret < count) btrfs_delalloc_release_space(inode, offset, - count - (size_t)ret); + count - (size_t)ret, BTRFS_RESERVE_NORMAL); } out: if (wakeup) @@ -9035,6 +9121,7 @@ int btrfs_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf) u64 page_start; u64 page_end; u64 end; + enum btrfs_metadata_reserve_type reserve_type = BTRFS_RESERVE_NORMAL; reserved_space = PAGE_SIZE; @@ -9043,6 +9130,8 @@ int btrfs_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf) page_end = page_start + PAGE_SIZE - 1; end = page_end; + if (inode_need_compress(inode)) + reserve_type = BTRFS_RESERVE_COMPRESS; /* * Reserving delalloc space after obtaining the page lock can lead to * deadlock. For example, if a dirty page is locked by this function @@ -9052,7 +9141,7 @@ int btrfs_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf) * being processed by btrfs_page_mkwrite() function. */ ret = btrfs_delalloc_reserve_space(inode, page_start, - reserved_space); + reserved_space, reserve_type); if (!ret) { ret = file_update_time(vma->vm_file); reserved = 1; @@ -9104,7 +9193,8 @@ int btrfs_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf) BTRFS_I(inode)->outstanding_extents++; spin_unlock(&BTRFS_I(inode)->lock); btrfs_delalloc_release_space(inode, page_start, - PAGE_SIZE - reserved_space); + PAGE_SIZE - reserved_space, + reserve_type); } } @@ -9121,7 +9211,7 @@ int btrfs_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf) 0, 0, &cached_state, GFP_NOFS); ret = btrfs_set_extent_delalloc(inode, page_start, end, - &cached_state, 0); + &cached_state, reserve_type); if (ret) { unlock_extent_cached(io_tree, page_start, page_end, &cached_state, GFP_NOFS); @@ -9159,7 +9249,8 @@ int btrfs_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf) } unlock_page(page); out: - btrfs_delalloc_release_space(inode, page_start, reserved_space); + btrfs_delalloc_release_space(inode, page_start, reserved_space, + reserve_type); out_noreserve: sb_end_pagefault(inode->i_sb); return ret; diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index fdf1901..91d121c 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -1132,6 +1132,7 @@ static int cluster_pages_for_defrag(struct inode *inode, struct extent_state *cached_state = NULL; struct extent_io_tree *tree; gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping); + enum btrfs_metadata_reserve_type reserve_type = BTRFS_RESERVE_NORMAL; file_end = (isize - 1) >> PAGE_SHIFT; if (!isize || start_index > file_end) @@ -1139,9 +1140,11 @@ static int cluster_pages_for_defrag(struct inode *inode, page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1); + if (inode_need_compress(inode)) + reserve_type = BTRFS_RESERVE_COMPRESS; ret = btrfs_delalloc_reserve_space(inode, start_index << PAGE_SHIFT, - page_cnt << PAGE_SHIFT); + page_cnt << PAGE_SHIFT, reserve_type); if (ret) return ret; i_done = 0; @@ -1232,11 +1235,12 @@ static int cluster_pages_for_defrag(struct inode *inode, spin_unlock(&BTRFS_I(inode)->lock); btrfs_delalloc_release_space(inode, start_index << PAGE_SHIFT, - (page_cnt - i_done) << PAGE_SHIFT); + (page_cnt - i_done) << PAGE_SHIFT, + reserve_type); } btrfs_set_extent_defrag(inode, page_start, - page_end - 1, &cached_state); + page_end - 1, &cached_state, reserve_type); unlock_extent_cached(&BTRFS_I(inode)->io_tree, page_start, page_end - 1, &cached_state, GFP_NOFS); @@ -1257,7 +1261,7 @@ static int cluster_pages_for_defrag(struct inode *inode, } btrfs_delalloc_release_space(inode, start_index << PAGE_SHIFT, - page_cnt << PAGE_SHIFT); + page_cnt << PAGE_SHIFT, reserve_type); return ret; } diff --git a/fs/btrfs/relocation.c b/fs/btrfs/relocation.c index 0ec8ffa..db90b47 100644 --- a/fs/btrfs/relocation.c +++ b/fs/btrfs/relocation.c @@ -3144,10 +3144,14 @@ static int relocate_file_extent_cluster(struct inode *inode, gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping); int nr = 0; int ret = 0; + enum btrfs_metadata_reserve_type reserve_type = BTRFS_RESERVE_NORMAL; if (!cluster->nr) return 0; + if (inode_need_compress(inode)) + reserve_type = BTRFS_RESERVE_COMPRESS; + ra = kzalloc(sizeof(*ra), GFP_NOFS); if (!ra) return -ENOMEM; @@ -3166,7 +3170,8 @@ static int relocate_file_extent_cluster(struct inode *inode, index = (cluster->start - offset) >> PAGE_SHIFT; last_index = (cluster->end - offset) >> PAGE_SHIFT; while (index <= last_index) { - ret = btrfs_delalloc_reserve_metadata(inode, PAGE_SIZE); + ret = btrfs_delalloc_reserve_metadata(inode, PAGE_SIZE, + reserve_type); if (ret) goto out; @@ -3179,7 +3184,7 @@ static int relocate_file_extent_cluster(struct inode *inode, mask); if (!page) { btrfs_delalloc_release_metadata(inode, - PAGE_SIZE); + PAGE_SIZE, reserve_type); ret = -ENOMEM; goto out; } @@ -3198,7 +3203,7 @@ static int relocate_file_extent_cluster(struct inode *inode, unlock_page(page); put_page(page); btrfs_delalloc_release_metadata(inode, - PAGE_SIZE); + PAGE_SIZE, reserve_type); ret = -EIO; goto out; } @@ -3219,7 +3224,8 @@ static int relocate_file_extent_cluster(struct inode *inode, nr++; } - btrfs_set_extent_delalloc(inode, page_start, page_end, NULL, 0); + btrfs_set_extent_delalloc(inode, page_start, page_end, NULL, + reserve_type); set_page_dirty(page); unlock_extent(&BTRFS_I(inode)->io_tree, diff --git a/fs/btrfs/tests/inode-tests.c b/fs/btrfs/tests/inode-tests.c index 0bf4680..9a1a01d 100644 --- a/fs/btrfs/tests/inode-tests.c +++ b/fs/btrfs/tests/inode-tests.c @@ -943,6 +943,7 @@ static int test_extent_accounting(u32 sectorsize, u32 nodesize) struct inode *inode = NULL; struct btrfs_root *root = NULL; int ret = -ENOMEM; + enum btrfs_metadata_reserve_type reserve_type = BTRFS_RESERVE_NORMAL; inode = btrfs_new_test_inode(); if (!inode) { @@ -968,7 +969,7 @@ static int test_extent_accounting(u32 sectorsize, u32 nodesize) /* [BTRFS_MAX_EXTENT_SIZE] */ BTRFS_I(inode)->outstanding_extents++; ret = btrfs_set_extent_delalloc(inode, 0, BTRFS_MAX_EXTENT_SIZE - 1, - NULL, 0); + NULL, reserve_type); if (ret) { test_msg("btrfs_set_extent_delalloc returned %d\n", ret); goto out; @@ -984,7 +985,7 @@ static int test_extent_accounting(u32 sectorsize, u32 nodesize) BTRFS_I(inode)->outstanding_extents++; ret = btrfs_set_extent_delalloc(inode, BTRFS_MAX_EXTENT_SIZE, BTRFS_MAX_EXTENT_SIZE + sectorsize - 1, - NULL, 0); + NULL, reserve_type); if (ret) { test_msg("btrfs_set_extent_delalloc returned %d\n", ret); goto out; @@ -1019,7 +1020,7 @@ static int test_extent_accounting(u32 sectorsize, u32 nodesize) ret = btrfs_set_extent_delalloc(inode, BTRFS_MAX_EXTENT_SIZE >> 1, (BTRFS_MAX_EXTENT_SIZE >> 1) + sectorsize - 1, - NULL, 0); + NULL, reserve_type); if (ret) { test_msg("btrfs_set_extent_delalloc returned %d\n", ret); goto out; @@ -1042,7 +1043,7 @@ static int test_extent_accounting(u32 sectorsize, u32 nodesize) ret = btrfs_set_extent_delalloc(inode, BTRFS_MAX_EXTENT_SIZE + 2 * sectorsize, (BTRFS_MAX_EXTENT_SIZE << 1) + 3 * sectorsize - 1, - NULL, 0); + NULL, reserve_type); if (ret) { test_msg("btrfs_set_extent_delalloc returned %d\n", ret); goto out; @@ -1060,7 +1061,8 @@ static int test_extent_accounting(u32 sectorsize, u32 nodesize) BTRFS_I(inode)->outstanding_extents++; ret = btrfs_set_extent_delalloc(inode, BTRFS_MAX_EXTENT_SIZE + sectorsize, - BTRFS_MAX_EXTENT_SIZE + 2 * sectorsize - 1, NULL, 0); + BTRFS_MAX_EXTENT_SIZE + 2 * sectorsize - 1, + NULL, reserve_type); if (ret) { test_msg("btrfs_set_extent_delalloc returned %d\n", ret); goto out; @@ -1097,7 +1099,8 @@ static int test_extent_accounting(u32 sectorsize, u32 nodesize) BTRFS_I(inode)->outstanding_extents++; ret = btrfs_set_extent_delalloc(inode, BTRFS_MAX_EXTENT_SIZE + sectorsize, - BTRFS_MAX_EXTENT_SIZE + 2 * sectorsize - 1, NULL, 0); + BTRFS_MAX_EXTENT_SIZE + 2 * sectorsize - 1, + NULL, reserve_type); if (ret) { test_msg("btrfs_set_extent_delalloc returned %d\n", ret); goto out;
When testing btrfs compression, sometimes we got ENOSPC error, though fs still has much free space, xfstests generic/171, generic/172, generic/173, generic/174, generic/175 can reveal this bug in my test environment when compression is enabled. After some debuging work, we found that it's btrfs_delalloc_reserve_metadata() which sometimes tries to reserve plenty of metadata space, even for very small data range. In btrfs_delalloc_reserve_metadata(), the number of metadata bytes we try to reserve is calculated by the difference between outstanding_extents and reserved_extents. Please see below case for how ENOSPC occurs: 1, Buffered write 128MB data in unit of 128KB, so finially we'll have inode outstanding extents be 1, and reserved_extents be 1024. Note it's btrfs_merge_extent_hook() that merges these 128KB units into one big outstanding extent, but do not change reserved_extents. 2, When writing dirty pages, for compression, cow_file_range_async() will split above big extent in unit of 128KB(compression extent size is 128KB). When first split opeartion finishes, we'll have 2 outstanding extents and 1024 reserved extents, and just right now the currently generated ordered extent is dispatched to run and complete, then btrfs_delalloc_release_metadata()(see btrfs_finish_ordered_io()) will be called to release metadata, after that we will have 1 outstanding extents and 1 reserved extents(also see logic in drop_outstanding_extent()). Later cow_file_range_async() continues to handles left data range[128KB, 128MB), and if no other ordered extent was dispatched to run, there will be 1023 outstanding extents and 1 reserved extent. 3, Now if another bufferd write for this file enters, then btrfs_delalloc_reserve_metadata() will at least try to reserve metadata for 1023 outstanding extents' metadata, for 16KB node size, it'll be 1023*16384*2*8, about 255MB, for 64K node size, it'll be 1023*65536*8*2, about 1GB metadata, so obviously it's not sane and can easily result in enospc error. The root cause is that for compression, its max extent size will no longer be BTRFS_MAX_EXTENT_SIZE(128MB), it'll be 128KB, so current metadata reservation method in btrfs is not appropriate or correct, here we introduce: enum btrfs_metadata_reserve_type { BTRFS_RESERVE_NORMAL, BTRFS_RESERVE_COMPRESS, }; and expand btrfs_delalloc_reserve_metadata() and btrfs_delalloc_reserve_space() by adding a new enum btrfs_metadata_reserve_type argument. When a data range will go through compression, we use BTRFS_RESERVE_COMPRESS to reserve metatata. Meanwhile we introduce EXTENT_COMPRESS flag to mark a data range that will go through compression path. With this patch, we can fix these false enospc error for compression. Signed-off-by: Wang Xiaoguang <wangxg.fnst@cn.fujitsu.com> Tested-by: Holger Hoffstätte <holger@applied-asynchrony.com> Tested-by: Stefan Priebe <s.priebe@profihost.ag> --- fs/btrfs/ctree.h | 31 ++++++-- fs/btrfs/extent-tree.c | 55 +++++++++---- fs/btrfs/extent_io.c | 59 +++++++++++++- fs/btrfs/extent_io.h | 2 + fs/btrfs/file.c | 28 +++++-- fs/btrfs/free-space-cache.c | 6 +- fs/btrfs/inode-map.c | 5 +- fs/btrfs/inode.c | 185 ++++++++++++++++++++++++++++++++----------- fs/btrfs/ioctl.c | 12 ++- fs/btrfs/relocation.c | 14 +++- fs/btrfs/tests/inode-tests.c | 15 ++-- 11 files changed, 315 insertions(+), 97 deletions(-)