From 210b907acfb414650dcdae9b470eb2fccac7996b Mon Sep 17 00:00:00 2001 From: Wang Sheng-Hui Date: Mon, 24 Sep 2012 08:47:23 +0800 Subject: btrfs: remove unnecessary cur_trans set before goto loop in join_transaction In the big loop, cur_trans will be set fs_info->running_transaction before it's used. And after kmem_cache_free it and goto loop, it will be setup again. No need to setup it immediately after freed. Signed-off-by: Wang Sheng-Hui Signed-off-by: Jiri Kosina --- fs/btrfs/transaction.c | 1 - 1 file changed, 1 deletion(-) (limited to 'fs/btrfs/transaction.c') diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 04bbfb1052eb..68704e74f0d3 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -111,7 +111,6 @@ loop: * to redo the trans_no_join checks above */ kmem_cache_free(btrfs_transaction_cachep, cur_trans); - cur_trans = fs_info->running_transaction; goto loop; } else if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) { spin_unlock(&fs_info->trans_lock); -- cgit v1.2.3 From 53b381b3abeb86f12787a6c40fee9b2f71edc23b Mon Sep 17 00:00:00 2001 From: David Woodhouse Date: Tue, 29 Jan 2013 18:40:14 -0500 Subject: Btrfs: RAID5 and RAID6 This builds on David Woodhouse's original Btrfs raid5/6 implementation. The code has changed quite a bit, blame Chris Mason for any bugs. Read/modify/write is done after the higher levels of the filesystem have prepared a given bio. This means the higher layers are not responsible for building full stripes, and they don't need to query for the topology of the extents that may get allocated during delayed allocation runs. It also means different files can easily share the same stripe. But, it does expose us to incorrect parity if we crash or lose power while doing a read/modify/write cycle. This will be addressed in a later commit. Scrub is unable to repair crc errors on raid5/6 chunks. Discard does not work on raid5/6 (yet) The stripe size is fixed at 64KiB per disk. This will be tunable in a later commit. Signed-off-by: Chris Mason --- fs/btrfs/Kconfig | 2 + fs/btrfs/Makefile | 2 +- fs/btrfs/ctree.h | 35 +- fs/btrfs/disk-io.c | 62 +- fs/btrfs/disk-io.h | 7 + fs/btrfs/extent-tree.c | 88 ++- fs/btrfs/extent_io.c | 18 +- fs/btrfs/free-space-cache.c | 50 +- fs/btrfs/inode.c | 18 +- fs/btrfs/raid56.c | 1647 +++++++++++++++++++++++++++++++++++++++++++ fs/btrfs/raid56.h | 51 ++ fs/btrfs/scrub.c | 8 + fs/btrfs/transaction.c | 3 + fs/btrfs/volumes.c | 385 ++++++++-- fs/btrfs/volumes.h | 9 +- 15 files changed, 2283 insertions(+), 102 deletions(-) create mode 100644 fs/btrfs/raid56.c create mode 100644 fs/btrfs/raid56.h (limited to 'fs/btrfs/transaction.c') diff --git a/fs/btrfs/Kconfig b/fs/btrfs/Kconfig index d33f01c08b60..4f5dc93fa2f8 100644 --- a/fs/btrfs/Kconfig +++ b/fs/btrfs/Kconfig @@ -6,6 +6,8 @@ config BTRFS_FS select ZLIB_DEFLATE select LZO_COMPRESS select LZO_DECOMPRESS + select RAID6_PQ + help Btrfs is a new filesystem with extents, writable snapshotting, support for multiple devices and many more features. diff --git a/fs/btrfs/Makefile b/fs/btrfs/Makefile index 7df3e0f0ee51..3932224f99e9 100644 --- a/fs/btrfs/Makefile +++ b/fs/btrfs/Makefile @@ -8,7 +8,7 @@ btrfs-y += super.o ctree.o extent-tree.o print-tree.o root-tree.o dir-item.o \ extent_io.o volumes.o async-thread.o ioctl.o locking.o orphan.o \ export.o tree-log.o free-space-cache.o zlib.o lzo.o \ compression.o delayed-ref.o relocation.o delayed-inode.o scrub.o \ - reada.o backref.o ulist.o qgroup.o send.o dev-replace.o + reada.o backref.o ulist.o qgroup.o send.o dev-replace.o raid56.o btrfs-$(CONFIG_BTRFS_FS_POSIX_ACL) += acl.o btrfs-$(CONFIG_BTRFS_FS_CHECK_INTEGRITY) += check-integrity.o diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 0ab51be6879f..0cce3aafbd62 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -502,6 +502,7 @@ struct btrfs_super_block { #define BTRFS_FEATURE_INCOMPAT_BIG_METADATA (1ULL << 5) #define BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF (1ULL << 6) +#define BTRFS_FEATURE_INCOMPAT_RAID56 (1ULL << 7) #define BTRFS_FEATURE_COMPAT_SUPP 0ULL #define BTRFS_FEATURE_COMPAT_RO_SUPP 0ULL @@ -511,6 +512,7 @@ struct btrfs_super_block { BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS | \ BTRFS_FEATURE_INCOMPAT_BIG_METADATA | \ BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO | \ + BTRFS_FEATURE_INCOMPAT_RAID56 | \ BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF) /* @@ -952,8 +954,10 @@ struct btrfs_dev_replace_item { #define BTRFS_BLOCK_GROUP_RAID1 (1ULL << 4) #define BTRFS_BLOCK_GROUP_DUP (1ULL << 5) #define BTRFS_BLOCK_GROUP_RAID10 (1ULL << 6) +#define BTRFS_BLOCK_GROUP_RAID5 (1 << 7) +#define BTRFS_BLOCK_GROUP_RAID6 (1 << 8) #define BTRFS_BLOCK_GROUP_RESERVED BTRFS_AVAIL_ALLOC_BIT_SINGLE -#define BTRFS_NR_RAID_TYPES 5 +#define BTRFS_NR_RAID_TYPES 7 #define BTRFS_BLOCK_GROUP_TYPE_MASK (BTRFS_BLOCK_GROUP_DATA | \ BTRFS_BLOCK_GROUP_SYSTEM | \ @@ -961,6 +965,8 @@ struct btrfs_dev_replace_item { #define BTRFS_BLOCK_GROUP_PROFILE_MASK (BTRFS_BLOCK_GROUP_RAID0 | \ BTRFS_BLOCK_GROUP_RAID1 | \ + BTRFS_BLOCK_GROUP_RAID5 | \ + BTRFS_BLOCK_GROUP_RAID6 | \ BTRFS_BLOCK_GROUP_DUP | \ BTRFS_BLOCK_GROUP_RAID10) /* @@ -1185,6 +1191,10 @@ struct btrfs_block_group_cache { u64 flags; u64 sectorsize; u64 cache_generation; + + /* for raid56, this is a full stripe, without parity */ + unsigned long full_stripe_len; + unsigned int ro:1; unsigned int dirty:1; unsigned int iref:1; @@ -1225,6 +1235,20 @@ struct seq_list { u64 seq; }; +/* used by the raid56 code to lock stripes for read/modify/write */ +struct btrfs_stripe_hash { + struct list_head hash_list; + wait_queue_head_t wait; + spinlock_t lock; +}; + +/* used by the raid56 code to lock stripes for read/modify/write */ +struct btrfs_stripe_hash_table { + struct btrfs_stripe_hash *table; +}; + +#define BTRFS_STRIPE_HASH_TABLE_BITS 11 + /* fs_info */ struct reloc_control; struct btrfs_device; @@ -1307,6 +1331,13 @@ struct btrfs_fs_info { struct mutex cleaner_mutex; struct mutex chunk_mutex; struct mutex volume_mutex; + + /* this is used during read/modify/write to make sure + * no two ios are trying to mod the same stripe at the same + * time + */ + struct btrfs_stripe_hash_table *stripe_hash_table; + /* * this protects the ordered operations list only while we are * processing all of the entries on it. This way we make @@ -1395,6 +1426,8 @@ struct btrfs_fs_info { struct btrfs_workers flush_workers; struct btrfs_workers endio_workers; struct btrfs_workers endio_meta_workers; + struct btrfs_workers endio_raid56_workers; + struct btrfs_workers rmw_workers; struct btrfs_workers endio_meta_write_workers; struct btrfs_workers endio_write_workers; struct btrfs_workers endio_freespace_worker; diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 65f03670a952..e9fa7b4d18e3 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -46,6 +46,7 @@ #include "check-integrity.h" #include "rcu-string.h" #include "dev-replace.h" +#include "raid56.h" #ifdef CONFIG_X86 #include @@ -639,8 +640,15 @@ err: btree_readahead_hook(root, eb, eb->start, ret); } - if (ret) + if (ret) { + /* + * our io error hook is going to dec the io pages + * again, we have to make sure it has something + * to decrement + */ + atomic_inc(&eb->io_pages); clear_extent_buffer_uptodate(eb); + } free_extent_buffer(eb); out: return ret; @@ -654,6 +662,7 @@ static int btree_io_failed_hook(struct page *page, int failed_mirror) eb = (struct extent_buffer *)page->private; set_bit(EXTENT_BUFFER_IOERR, &eb->bflags); eb->read_mirror = failed_mirror; + atomic_dec(&eb->io_pages); if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) btree_readahead_hook(root, eb, eb->start, -EIO); return -EIO; /* we fixed nothing */ @@ -670,17 +679,23 @@ static void end_workqueue_bio(struct bio *bio, int err) end_io_wq->work.flags = 0; if (bio->bi_rw & REQ_WRITE) { - if (end_io_wq->metadata == 1) + if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) btrfs_queue_worker(&fs_info->endio_meta_write_workers, &end_io_wq->work); - else if (end_io_wq->metadata == 2) + else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) btrfs_queue_worker(&fs_info->endio_freespace_worker, &end_io_wq->work); + else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) + btrfs_queue_worker(&fs_info->endio_raid56_workers, + &end_io_wq->work); else btrfs_queue_worker(&fs_info->endio_write_workers, &end_io_wq->work); } else { - if (end_io_wq->metadata) + if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) + btrfs_queue_worker(&fs_info->endio_raid56_workers, + &end_io_wq->work); + else if (end_io_wq->metadata) btrfs_queue_worker(&fs_info->endio_meta_workers, &end_io_wq->work); else @@ -695,6 +710,7 @@ static void end_workqueue_bio(struct bio *bio, int err) * 0 - if data * 1 - if normal metadta * 2 - if writing to the free space cache area + * 3 - raid parity work */ int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, int metadata) @@ -2165,6 +2181,12 @@ int open_ctree(struct super_block *sb, init_waitqueue_head(&fs_info->transaction_blocked_wait); init_waitqueue_head(&fs_info->async_submit_wait); + ret = btrfs_alloc_stripe_hash_table(fs_info); + if (ret) { + err = -ENOMEM; + goto fail_alloc; + } + __setup_root(4096, 4096, 4096, 4096, tree_root, fs_info, BTRFS_ROOT_TREE_OBJECTID); @@ -2332,6 +2354,12 @@ int open_ctree(struct super_block *sb, btrfs_init_workers(&fs_info->endio_meta_write_workers, "endio-meta-write", fs_info->thread_pool_size, &fs_info->generic_worker); + btrfs_init_workers(&fs_info->endio_raid56_workers, + "endio-raid56", fs_info->thread_pool_size, + &fs_info->generic_worker); + btrfs_init_workers(&fs_info->rmw_workers, + "rmw", fs_info->thread_pool_size, + &fs_info->generic_worker); btrfs_init_workers(&fs_info->endio_write_workers, "endio-write", fs_info->thread_pool_size, &fs_info->generic_worker); @@ -2350,6 +2378,8 @@ int open_ctree(struct super_block *sb, */ fs_info->endio_workers.idle_thresh = 4; fs_info->endio_meta_workers.idle_thresh = 4; + fs_info->endio_raid56_workers.idle_thresh = 4; + fs_info->rmw_workers.idle_thresh = 2; fs_info->endio_write_workers.idle_thresh = 2; fs_info->endio_meta_write_workers.idle_thresh = 2; @@ -2366,6 +2396,8 @@ int open_ctree(struct super_block *sb, ret |= btrfs_start_workers(&fs_info->fixup_workers); ret |= btrfs_start_workers(&fs_info->endio_workers); ret |= btrfs_start_workers(&fs_info->endio_meta_workers); + ret |= btrfs_start_workers(&fs_info->rmw_workers); + ret |= btrfs_start_workers(&fs_info->endio_raid56_workers); ret |= btrfs_start_workers(&fs_info->endio_meta_write_workers); ret |= btrfs_start_workers(&fs_info->endio_write_workers); ret |= btrfs_start_workers(&fs_info->endio_freespace_worker); @@ -2710,6 +2742,8 @@ fail_sb_buffer: btrfs_stop_workers(&fs_info->workers); btrfs_stop_workers(&fs_info->endio_workers); btrfs_stop_workers(&fs_info->endio_meta_workers); + btrfs_stop_workers(&fs_info->endio_raid56_workers); + btrfs_stop_workers(&fs_info->rmw_workers); btrfs_stop_workers(&fs_info->endio_meta_write_workers); btrfs_stop_workers(&fs_info->endio_write_workers); btrfs_stop_workers(&fs_info->endio_freespace_worker); @@ -2728,6 +2762,7 @@ fail_bdi: fail_srcu: cleanup_srcu_struct(&fs_info->subvol_srcu); fail: + btrfs_free_stripe_hash_table(fs_info); btrfs_close_devices(fs_info->fs_devices); return err; @@ -3076,11 +3111,16 @@ int btrfs_calc_num_tolerated_disk_barrier_failures( ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0))) num_tolerated_disk_barrier_failures = 0; - else if (num_tolerated_disk_barrier_failures > 1 - && - (flags & (BTRFS_BLOCK_GROUP_RAID1 | - BTRFS_BLOCK_GROUP_RAID10))) - num_tolerated_disk_barrier_failures = 1; + else if (num_tolerated_disk_barrier_failures > 1) { + if (flags & (BTRFS_BLOCK_GROUP_RAID1 | + BTRFS_BLOCK_GROUP_RAID5 | + BTRFS_BLOCK_GROUP_RAID10)) { + num_tolerated_disk_barrier_failures = 1; + } else if (flags & + BTRFS_BLOCK_GROUP_RAID5) { + num_tolerated_disk_barrier_failures = 2; + } + } } } up_read(&sinfo->groups_sem); @@ -3384,6 +3424,8 @@ int close_ctree(struct btrfs_root *root) btrfs_stop_workers(&fs_info->workers); btrfs_stop_workers(&fs_info->endio_workers); btrfs_stop_workers(&fs_info->endio_meta_workers); + btrfs_stop_workers(&fs_info->endio_raid56_workers); + btrfs_stop_workers(&fs_info->rmw_workers); btrfs_stop_workers(&fs_info->endio_meta_write_workers); btrfs_stop_workers(&fs_info->endio_write_workers); btrfs_stop_workers(&fs_info->endio_freespace_worker); @@ -3404,6 +3446,8 @@ int close_ctree(struct btrfs_root *root) bdi_destroy(&fs_info->bdi); cleanup_srcu_struct(&fs_info->subvol_srcu); + btrfs_free_stripe_hash_table(fs_info); + return 0; } diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h index 305c33efb0e3..034d7dc552b2 100644 --- a/fs/btrfs/disk-io.h +++ b/fs/btrfs/disk-io.h @@ -25,6 +25,13 @@ #define BTRFS_SUPER_MIRROR_MAX 3 #define BTRFS_SUPER_MIRROR_SHIFT 12 +enum { + BTRFS_WQ_ENDIO_DATA = 0, + BTRFS_WQ_ENDIO_METADATA = 1, + BTRFS_WQ_ENDIO_FREE_SPACE = 2, + BTRFS_WQ_ENDIO_RAID56 = 3, +}; + static inline u64 btrfs_sb_offset(int mirror) { u64 start = 16 * 1024; diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index d133edfcd449..3345f68fc64b 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -31,6 +31,7 @@ #include "print-tree.h" #include "transaction.h" #include "volumes.h" +#include "raid56.h" #include "locking.h" #include "free-space-cache.h" #include "math.h" @@ -1852,6 +1853,8 @@ static int btrfs_discard_extent(struct btrfs_root *root, u64 bytenr, *actual_bytes = discarded_bytes; + if (ret == -EOPNOTSUPP) + ret = 0; return ret; } @@ -3276,6 +3279,7 @@ u64 btrfs_reduce_alloc_profile(struct btrfs_root *root, u64 flags) u64 num_devices = root->fs_info->fs_devices->rw_devices + root->fs_info->fs_devices->missing_devices; u64 target; + u64 tmp; /* * see if restripe for this chunk_type is in progress, if so @@ -3292,30 +3296,32 @@ u64 btrfs_reduce_alloc_profile(struct btrfs_root *root, u64 flags) } spin_unlock(&root->fs_info->balance_lock); + /* First, mask out the RAID levels which aren't possible */ if (num_devices == 1) - flags &= ~(BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID0); + flags &= ~(BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID0 | + BTRFS_BLOCK_GROUP_RAID5); + if (num_devices < 3) + flags &= ~BTRFS_BLOCK_GROUP_RAID6; if (num_devices < 4) flags &= ~BTRFS_BLOCK_GROUP_RAID10; - if ((flags & BTRFS_BLOCK_GROUP_DUP) && - (flags & (BTRFS_BLOCK_GROUP_RAID1 | - BTRFS_BLOCK_GROUP_RAID10))) { - flags &= ~BTRFS_BLOCK_GROUP_DUP; - } - - if ((flags & BTRFS_BLOCK_GROUP_RAID1) && - (flags & BTRFS_BLOCK_GROUP_RAID10)) { - flags &= ~BTRFS_BLOCK_GROUP_RAID1; - } + tmp = flags & (BTRFS_BLOCK_GROUP_DUP | BTRFS_BLOCK_GROUP_RAID0 | + BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID5 | + BTRFS_BLOCK_GROUP_RAID6 | BTRFS_BLOCK_GROUP_RAID10); + flags &= ~tmp; - if ((flags & BTRFS_BLOCK_GROUP_RAID0) && - ((flags & BTRFS_BLOCK_GROUP_RAID1) | - (flags & BTRFS_BLOCK_GROUP_RAID10) | - (flags & BTRFS_BLOCK_GROUP_DUP))) { - flags &= ~BTRFS_BLOCK_GROUP_RAID0; - } + if (tmp & BTRFS_BLOCK_GROUP_RAID6) + tmp = BTRFS_BLOCK_GROUP_RAID6; + else if (tmp & BTRFS_BLOCK_GROUP_RAID5) + tmp = BTRFS_BLOCK_GROUP_RAID5; + else if (tmp & BTRFS_BLOCK_GROUP_RAID10) + tmp = BTRFS_BLOCK_GROUP_RAID10; + else if (tmp & BTRFS_BLOCK_GROUP_RAID1) + tmp = BTRFS_BLOCK_GROUP_RAID1; + else if (tmp & BTRFS_BLOCK_GROUP_RAID0) + tmp = BTRFS_BLOCK_GROUP_RAID0; - return extended_to_chunk(flags); + return extended_to_chunk(flags | tmp); } static u64 get_alloc_profile(struct btrfs_root *root, u64 flags) @@ -3333,6 +3339,7 @@ static u64 get_alloc_profile(struct btrfs_root *root, u64 flags) u64 btrfs_get_alloc_profile(struct btrfs_root *root, int data) { u64 flags; + u64 ret; if (data) flags = BTRFS_BLOCK_GROUP_DATA; @@ -3341,7 +3348,8 @@ u64 btrfs_get_alloc_profile(struct btrfs_root *root, int data) else flags = BTRFS_BLOCK_GROUP_METADATA; - return get_alloc_profile(root, flags); + ret = get_alloc_profile(root, flags); + return ret; } /* @@ -3516,8 +3524,10 @@ static u64 get_system_chunk_thresh(struct btrfs_root *root, u64 type) { u64 num_dev; - if (type & BTRFS_BLOCK_GROUP_RAID10 || - type & BTRFS_BLOCK_GROUP_RAID0) + if (type & (BTRFS_BLOCK_GROUP_RAID10 | + BTRFS_BLOCK_GROUP_RAID0 | + BTRFS_BLOCK_GROUP_RAID5 | + BTRFS_BLOCK_GROUP_RAID6)) num_dev = root->fs_info->fs_devices->rw_devices; else if (type & BTRFS_BLOCK_GROUP_RAID1) num_dev = 2; @@ -3667,7 +3677,9 @@ static int can_overcommit(struct btrfs_root *root, /* * If we have dup, raid1 or raid10 then only half of the free - * space is actually useable. + * space is actually useable. For raid56, the space info used + * doesn't include the parity drive, so we don't have to + * change the math */ if (profile & (BTRFS_BLOCK_GROUP_DUP | BTRFS_BLOCK_GROUP_RAID1 | @@ -5455,10 +5467,14 @@ int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, return ret; } -static u64 stripe_align(struct btrfs_root *root, u64 val) +static u64 stripe_align(struct btrfs_root *root, + struct btrfs_block_group_cache *cache, + u64 val, u64 num_bytes) { - u64 mask = ((u64)root->stripesize - 1); - u64 ret = (val + mask) & ~mask; + u64 mask; + u64 ret; + mask = ((u64)root->stripesize - 1); + ret = (val + mask) & ~mask; return ret; } @@ -5519,9 +5535,12 @@ int __get_raid_index(u64 flags) index = 2; else if (flags & BTRFS_BLOCK_GROUP_RAID0) index = 3; + else if (flags & BTRFS_BLOCK_GROUP_RAID5) + index = 5; + else if (flags & BTRFS_BLOCK_GROUP_RAID6) + index = 6; else - index = 4; - + index = 4; /* BTRFS_BLOCK_GROUP_SINGLE */ return index; } @@ -5665,6 +5684,8 @@ search: if (!block_group_bits(block_group, data)) { u64 extra = BTRFS_BLOCK_GROUP_DUP | BTRFS_BLOCK_GROUP_RAID1 | + BTRFS_BLOCK_GROUP_RAID5 | + BTRFS_BLOCK_GROUP_RAID6 | BTRFS_BLOCK_GROUP_RAID10; /* @@ -5835,7 +5856,8 @@ unclustered_alloc: goto loop; } checks: - search_start = stripe_align(root, offset); + search_start = stripe_align(root, used_block_group, + offset, num_bytes); /* move on to the next group */ if (search_start + num_bytes > @@ -7203,6 +7225,7 @@ static u64 update_block_group_flags(struct btrfs_root *root, u64 flags) root->fs_info->fs_devices->missing_devices; stripped = BTRFS_BLOCK_GROUP_RAID0 | + BTRFS_BLOCK_GROUP_RAID5 | BTRFS_BLOCK_GROUP_RAID6 | BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID10; if (num_devices == 1) { @@ -7754,7 +7777,9 @@ int btrfs_read_block_groups(struct btrfs_root *root) btrfs_release_path(path); cache->flags = btrfs_block_group_flags(&cache->item); cache->sectorsize = root->sectorsize; - + cache->full_stripe_len = btrfs_full_stripe_len(root, + &root->fs_info->mapping_tree, + found_key.objectid); btrfs_init_free_space_ctl(cache); /* @@ -7808,6 +7833,8 @@ int btrfs_read_block_groups(struct btrfs_root *root) if (!(get_alloc_profile(root, space_info->flags) & (BTRFS_BLOCK_GROUP_RAID10 | BTRFS_BLOCK_GROUP_RAID1 | + BTRFS_BLOCK_GROUP_RAID5 | + BTRFS_BLOCK_GROUP_RAID6 | BTRFS_BLOCK_GROUP_DUP))) continue; /* @@ -7883,6 +7910,9 @@ int btrfs_make_block_group(struct btrfs_trans_handle *trans, cache->key.type = BTRFS_BLOCK_GROUP_ITEM_KEY; cache->sectorsize = root->sectorsize; cache->fs_info = root->fs_info; + cache->full_stripe_len = btrfs_full_stripe_len(root, + &root->fs_info->mapping_tree, + chunk_offset); atomic_set(&cache->count, 1); spin_lock_init(&cache->lock); diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 86ecca48c604..3b9fb478b0d1 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -1895,13 +1895,11 @@ static int free_io_failure(struct inode *inode, struct io_failure_record *rec, if (ret) err = ret; - if (did_repair) { - ret = clear_extent_bits(&BTRFS_I(inode)->io_tree, rec->start, - rec->start + rec->len - 1, - EXTENT_DAMAGED, GFP_NOFS); - if (ret && !err) - err = ret; - } + ret = clear_extent_bits(&BTRFS_I(inode)->io_tree, rec->start, + rec->start + rec->len - 1, + EXTENT_DAMAGED, GFP_NOFS); + if (ret && !err) + err = ret; kfree(rec); return err; @@ -1932,10 +1930,15 @@ int repair_io_failure(struct btrfs_fs_info *fs_info, u64 start, u64 map_length = 0; u64 sector; struct btrfs_bio *bbio = NULL; + struct btrfs_mapping_tree *map_tree = &fs_info->mapping_tree; int ret; BUG_ON(!mirror_num); + /* we can't repair anything in raid56 yet */ + if (btrfs_is_parity_mirror(map_tree, logical, length, mirror_num)) + return 0; + bio = bio_alloc(GFP_NOFS, 1); if (!bio) return -EIO; @@ -2052,6 +2055,7 @@ static int clean_io_failure(u64 start, struct page *page) failrec->failed_mirror); did_repair = !ret; } + ret = 0; } out: diff --git a/fs/btrfs/free-space-cache.c b/fs/btrfs/free-space-cache.c index 59ea2e4349c9..62020b7f7036 100644 --- a/fs/btrfs/free-space-cache.c +++ b/fs/btrfs/free-space-cache.c @@ -1463,10 +1463,14 @@ static int search_bitmap(struct btrfs_free_space_ctl *ctl, } static struct btrfs_free_space * -find_free_space(struct btrfs_free_space_ctl *ctl, u64 *offset, u64 *bytes) +find_free_space(struct btrfs_free_space_ctl *ctl, u64 *offset, u64 *bytes, + unsigned long align) { struct btrfs_free_space *entry; struct rb_node *node; + u64 ctl_off; + u64 tmp; + u64 align_off; int ret; if (!ctl->free_space_offset.rb_node) @@ -1481,15 +1485,34 @@ find_free_space(struct btrfs_free_space_ctl *ctl, u64 *offset, u64 *bytes) if (entry->bytes < *bytes) continue; + /* make sure the space returned is big enough + * to match our requested alignment + */ + if (*bytes >= align) { + ctl_off = entry->offset - ctl->start; + tmp = ctl_off + align - 1;; + do_div(tmp, align); + tmp = tmp * align + ctl->start; + align_off = tmp - entry->offset; + } else { + align_off = 0; + tmp = entry->offset; + } + + if (entry->bytes < *bytes + align_off) + continue; + if (entry->bitmap) { - ret = search_bitmap(ctl, entry, offset, bytes); - if (!ret) + ret = search_bitmap(ctl, entry, &tmp, bytes); + if (!ret) { + *offset = tmp; return entry; + } continue; } - *offset = entry->offset; - *bytes = entry->bytes; + *offset = tmp; + *bytes = entry->bytes - align_off; return entry; } @@ -2091,9 +2114,12 @@ u64 btrfs_find_space_for_alloc(struct btrfs_block_group_cache *block_group, struct btrfs_free_space *entry = NULL; u64 bytes_search = bytes + empty_size; u64 ret = 0; + u64 align_gap = 0; + u64 align_gap_len = 0; spin_lock(&ctl->tree_lock); - entry = find_free_space(ctl, &offset, &bytes_search); + entry = find_free_space(ctl, &offset, &bytes_search, + block_group->full_stripe_len); if (!entry) goto out; @@ -2103,9 +2129,15 @@ u64 btrfs_find_space_for_alloc(struct btrfs_block_group_cache *block_group, if (!entry->bytes) free_bitmap(ctl, entry); } else { + unlink_free_space(ctl, entry); - entry->offset += bytes; - entry->bytes -= bytes; + align_gap_len = offset - entry->offset; + align_gap = entry->offset; + + entry->offset = offset + bytes; + WARN_ON(entry->bytes < bytes + align_gap_len); + + entry->bytes -= bytes + align_gap_len; if (!entry->bytes) kmem_cache_free(btrfs_free_space_cachep, entry); else @@ -2115,6 +2147,8 @@ u64 btrfs_find_space_for_alloc(struct btrfs_block_group_cache *block_group, out: spin_unlock(&ctl->tree_lock); + if (align_gap_len) + __btrfs_add_free_space(ctl, align_gap, align_gap_len); return ret; } diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 1b98c4ce3c6f..6f4e41dca970 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -39,6 +39,7 @@ #include #include #include +#include #include "compat.h" #include "ctree.h" #include "disk-io.h" @@ -6386,19 +6387,24 @@ static int btrfs_submit_direct_hook(int rw, struct btrfs_dio_private *dip, int async_submit = 0; map_length = orig_bio->bi_size; - ret = btrfs_map_block(root->fs_info, READ, start_sector << 9, + ret = btrfs_map_block(root->fs_info, rw, start_sector << 9, &map_length, NULL, 0); if (ret) { bio_put(orig_bio); return -EIO; } - if (map_length >= orig_bio->bi_size) { bio = orig_bio; goto submit; } - async_submit = 1; + /* async crcs make it difficult to collect full stripe writes. */ + if (btrfs_get_alloc_profile(root, 1) & + (BTRFS_BLOCK_GROUP_RAID5 | BTRFS_BLOCK_GROUP_RAID6)) + async_submit = 0; + else + async_submit = 1; + bio = btrfs_dio_bio_alloc(orig_bio->bi_bdev, start_sector, GFP_NOFS); if (!bio) return -ENOMEM; @@ -6440,7 +6446,7 @@ static int btrfs_submit_direct_hook(int rw, struct btrfs_dio_private *dip, bio->bi_end_io = btrfs_end_dio_bio; map_length = orig_bio->bi_size; - ret = btrfs_map_block(root->fs_info, READ, + ret = btrfs_map_block(root->fs_info, rw, start_sector << 9, &map_length, NULL, 0); if (ret) { @@ -6583,15 +6589,17 @@ static ssize_t btrfs_direct_IO(int rw, struct kiocb *iocb, { struct file *file = iocb->ki_filp; struct inode *inode = file->f_mapping->host; + ssize_t ret; if (check_direct_IO(BTRFS_I(inode)->root, rw, iocb, iov, offset, nr_segs)) return 0; - return __blockdev_direct_IO(rw, iocb, inode, + ret = __blockdev_direct_IO(rw, iocb, inode, BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev, iov, offset, nr_segs, btrfs_get_blocks_direct, NULL, btrfs_submit_direct, 0); + return ret; } #define BTRFS_FIEMAP_FLAGS (FIEMAP_FLAG_SYNC) diff --git a/fs/btrfs/raid56.c b/fs/btrfs/raid56.c new file mode 100644 index 000000000000..d02510f34936 --- /dev/null +++ b/fs/btrfs/raid56.c @@ -0,0 +1,1647 @@ +/* + * Copyright (C) 2012 Fusion-io All rights reserved. + * Copyright (C) 2012 Intel Corp. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include "compat.h" +#include "ctree.h" +#include "extent_map.h" +#include "disk-io.h" +#include "transaction.h" +#include "print-tree.h" +#include "volumes.h" +#include "raid56.h" +#include "async-thread.h" +#include "check-integrity.h" +#include "rcu-string.h" + +/* set when additional merges to this rbio are not allowed */ +#define RBIO_RMW_LOCKED_BIT 1 + +struct btrfs_raid_bio { + struct btrfs_fs_info *fs_info; + struct btrfs_bio *bbio; + + /* + * logical block numbers for the start of each stripe + * The last one or two are p/q. These are sorted, + * so raid_map[0] is the start of our full stripe + */ + u64 *raid_map; + + /* while we're doing rmw on a stripe + * we put it into a hash table so we can + * lock the stripe and merge more rbios + * into it. + */ + struct list_head hash_list; + + /* + * for scheduling work in the helper threads + */ + struct btrfs_work work; + + /* + * bio list and bio_list_lock are used + * to add more bios into the stripe + * in hopes of avoiding the full rmw + */ + struct bio_list bio_list; + spinlock_t bio_list_lock; + + /* + * also protected by the bio_list_lock, the + * stripe locking code uses plug_list to hand off + * the stripe lock to the next pending IO + */ + struct list_head plug_list; + + /* + * flags that tell us if it is safe to + * merge with this bio + */ + unsigned long flags; + + /* size of each individual stripe on disk */ + int stripe_len; + + /* number of data stripes (no p/q) */ + int nr_data; + + /* + * set if we're doing a parity rebuild + * for a read from higher up, which is handled + * differently from a parity rebuild as part of + * rmw + */ + int read_rebuild; + + /* first bad stripe */ + int faila; + + /* second bad stripe (for raid6 use) */ + int failb; + + /* + * number of pages needed to represent the full + * stripe + */ + int nr_pages; + + /* + * size of all the bios in the bio_list. This + * helps us decide if the rbio maps to a full + * stripe or not + */ + int bio_list_bytes; + + atomic_t refs; + + /* + * these are two arrays of pointers. We allocate the + * rbio big enough to hold them both and setup their + * locations when the rbio is allocated + */ + + /* pointers to pages that we allocated for + * reading/writing stripes directly from the disk (including P/Q) + */ + struct page **stripe_pages; + + /* + * pointers to the pages in the bio_list. Stored + * here for faster lookup + */ + struct page **bio_pages; +}; + +static int __raid56_parity_recover(struct btrfs_raid_bio *rbio); +static noinline void finish_rmw(struct btrfs_raid_bio *rbio); +static void rmw_work(struct btrfs_work *work); +static void read_rebuild_work(struct btrfs_work *work); +static void async_rmw_stripe(struct btrfs_raid_bio *rbio); +static void async_read_rebuild(struct btrfs_raid_bio *rbio); +static int fail_bio_stripe(struct btrfs_raid_bio *rbio, struct bio *bio); +static int fail_rbio_index(struct btrfs_raid_bio *rbio, int failed); +static void __free_raid_bio(struct btrfs_raid_bio *rbio); +static void index_rbio_pages(struct btrfs_raid_bio *rbio); +static int alloc_rbio_pages(struct btrfs_raid_bio *rbio); + +/* + * the stripe hash table is used for locking, and to collect + * bios in hopes of making a full stripe + */ +int btrfs_alloc_stripe_hash_table(struct btrfs_fs_info *info) +{ + struct btrfs_stripe_hash_table *table; + struct btrfs_stripe_hash_table *x; + struct btrfs_stripe_hash *cur; + struct btrfs_stripe_hash *h; + int num_entries = 1 << BTRFS_STRIPE_HASH_TABLE_BITS; + int i; + + if (info->stripe_hash_table) + return 0; + + table = kzalloc(sizeof(*table) + sizeof(*h) * num_entries, GFP_NOFS); + if (!table) + return -ENOMEM; + + table->table = (void *)(table + 1); + h = table->table; + + for (i = 0; i < num_entries; i++) { + cur = h + i; + INIT_LIST_HEAD(&cur->hash_list); + spin_lock_init(&cur->lock); + init_waitqueue_head(&cur->wait); + } + + x = cmpxchg(&info->stripe_hash_table, NULL, table); + if (x) + kfree(x); + return 0; +} + +/* + * we hash on the first logical address of the stripe + */ +static int rbio_bucket(struct btrfs_raid_bio *rbio) +{ + u64 num = rbio->raid_map[0]; + + /* + * we shift down quite a bit. We're using byte + * addressing, and most of the lower bits are zeros. + * This tends to upset hash_64, and it consistently + * returns just one or two different values. + * + * shifting off the lower bits fixes things. + */ + return hash_64(num >> 16, BTRFS_STRIPE_HASH_TABLE_BITS); +} + +/* + * merging means we take the bio_list from the victim and + * splice it into the destination. The victim should + * be discarded afterwards. + * + * must be called with dest->rbio_list_lock held + */ +static void merge_rbio(struct btrfs_raid_bio *dest, + struct btrfs_raid_bio *victim) +{ + bio_list_merge(&dest->bio_list, &victim->bio_list); + dest->bio_list_bytes += victim->bio_list_bytes; + bio_list_init(&victim->bio_list); +} + +/* + * free the hash table used by unmount + */ +void btrfs_free_stripe_hash_table(struct btrfs_fs_info *info) +{ + if (!info->stripe_hash_table) + return; + kfree(info->stripe_hash_table); + info->stripe_hash_table = NULL; +} + +/* + * helper function to run the xor_blocks api. It is only + * able to do MAX_XOR_BLOCKS at a time, so we need to + * loop through. + */ +static void run_xor(void **pages, int src_cnt, ssize_t len) +{ + int src_off = 0; + int xor_src_cnt = 0; + void *dest = pages[src_cnt]; + + while(src_cnt > 0) { + xor_src_cnt = min(src_cnt, MAX_XOR_BLOCKS); + xor_blocks(xor_src_cnt, len, dest, pages + src_off); + + src_cnt -= xor_src_cnt; + src_off += xor_src_cnt; + } +} + +/* + * returns true if the bio list inside this rbio + * covers an entire stripe (no rmw required). + * Must be called with the bio list lock held, or + * at a time when you know it is impossible to add + * new bios into the list + */ +static int __rbio_is_full(struct btrfs_raid_bio *rbio) +{ + unsigned long size = rbio->bio_list_bytes; + int ret = 1; + + if (size != rbio->nr_data * rbio->stripe_len) + ret = 0; + + BUG_ON(size > rbio->nr_data * rbio->stripe_len); + return ret; +} + +static int rbio_is_full(struct btrfs_raid_bio *rbio) +{ + unsigned long flags; + int ret; + + spin_lock_irqsave(&rbio->bio_list_lock, flags); + ret = __rbio_is_full(rbio); + spin_unlock_irqrestore(&rbio->bio_list_lock, flags); + return ret; +} + +/* + * returns 1 if it is safe to merge two rbios together. + * The merging is safe if the two rbios correspond to + * the same stripe and if they are both going in the same + * direction (read vs write), and if neither one is + * locked for final IO + * + * The caller is responsible for locking such that + * rmw_locked is safe to test + */ +static int rbio_can_merge(struct btrfs_raid_bio *last, + struct btrfs_raid_bio *cur) +{ + if (test_bit(RBIO_RMW_LOCKED_BIT, &last->flags) || + test_bit(RBIO_RMW_LOCKED_BIT, &cur->flags)) + return 0; + + if (last->raid_map[0] != + cur->raid_map[0]) + return 0; + + /* reads can't merge with writes */ + if (last->read_rebuild != + cur->read_rebuild) { + return 0; + } + + return 1; +} + +/* + * helper to index into the pstripe + */ +static struct page *rbio_pstripe_page(struct btrfs_raid_bio *rbio, int index) +{ + index += (rbio->nr_data * rbio->stripe_len) >> PAGE_CACHE_SHIFT; + return rbio->stripe_pages[index]; +} + +/* + * helper to index into the qstripe, returns null + * if there is no qstripe + */ +static struct page *rbio_qstripe_page(struct btrfs_raid_bio *rbio, int index) +{ + if (rbio->nr_data + 1 == rbio->bbio->num_stripes) + return NULL; + + index += ((rbio->nr_data + 1) * rbio->stripe_len) >> + PAGE_CACHE_SHIFT; + return rbio->stripe_pages[index]; +} + +/* + * The first stripe in the table for a logical address + * has the lock. rbios are added in one of three ways: + * + * 1) Nobody has the stripe locked yet. The rbio is given + * the lock and 0 is returned. The caller must start the IO + * themselves. + * + * 2) Someone has the stripe locked, but we're able to merge + * with the lock owner. The rbio is freed and the IO will + * start automatically along with the existing rbio. 1 is returned. + * + * 3) Someone has the stripe locked, but we're not able to merge. + * The rbio is added to the lock owner's plug list, or merged into + * an rbio already on the plug list. When the lock owner unlocks, + * the next rbio on the list is run and the IO is started automatically. + * 1 is returned + * + * If we return 0, the caller still owns the rbio and must continue with + * IO submission. If we return 1, the caller must assume the rbio has + * already been freed. + */ +static noinline int lock_stripe_add(struct btrfs_raid_bio *rbio) +{ + int bucket = rbio_bucket(rbio); + struct btrfs_stripe_hash *h = rbio->fs_info->stripe_hash_table->table + bucket; + struct btrfs_raid_bio *cur; + struct btrfs_raid_bio *pending; + unsigned long flags; + DEFINE_WAIT(wait); + struct btrfs_raid_bio *freeit = NULL; + int ret = 0; + int walk = 0; + + spin_lock_irqsave(&h->lock, flags); + list_for_each_entry(cur, &h->hash_list, hash_list) { + walk++; + if (cur->raid_map[0] == rbio->raid_map[0]) { + spin_lock(&cur->bio_list_lock); + + /* can we merge into the lock owner? */ + if (rbio_can_merge(cur, rbio)) { + merge_rbio(cur, rbio); + spin_unlock(&cur->bio_list_lock); + freeit = rbio; + ret = 1; + goto out; + } + + /* + * we couldn't merge with the running + * rbio, see if we can merge with the + * pending ones. We don't have to + * check for rmw_locked because there + * is no way they are inside finish_rmw + * right now + */ + list_for_each_entry(pending, &cur->plug_list, + plug_list) { + if (rbio_can_merge(pending, rbio)) { + merge_rbio(pending, rbio); + spin_unlock(&cur->bio_list_lock); + freeit = rbio; + ret = 1; + goto out; + } + } + + /* no merging, put us on the tail of the plug list, + * our rbio will be started with the currently + * running rbio unlocks + */ + list_add_tail(&rbio->plug_list, &cur->plug_list); + spin_unlock(&cur->bio_list_lock); + ret = 1; + goto out; + } + } + + atomic_inc(&rbio->refs); + list_add(&rbio->hash_list, &h->hash_list); +out: + spin_unlock_irqrestore(&h->lock, flags); + if (freeit) + __free_raid_bio(freeit); + return ret; +} + +/* + * called as rmw or parity rebuild is completed. If the plug list has more + * rbios waiting for this stripe, the next one on the list will be started + */ +static noinline void unlock_stripe(struct btrfs_raid_bio *rbio) +{ + int bucket; + struct btrfs_stripe_hash *h; + unsigned long flags; + + bucket = rbio_bucket(rbio); + h = rbio->fs_info->stripe_hash_table->table + bucket; + + spin_lock_irqsave(&h->lock, flags); + spin_lock(&rbio->bio_list_lock); + + if (!list_empty(&rbio->hash_list)) { + + list_del_init(&rbio->hash_list); + atomic_dec(&rbio->refs); + + /* + * we use the plug list to hold all the rbios + * waiting for the chance to lock this stripe. + * hand the lock over to one of them. + */ + if (!list_empty(&rbio->plug_list)) { + struct btrfs_raid_bio *next; + struct list_head *head = rbio->plug_list.next; + + next = list_entry(head, struct btrfs_raid_bio, + plug_list); + + list_del_init(&rbio->plug_list); + + list_add(&next->hash_list, &h->hash_list); + atomic_inc(&next->refs); + spin_unlock(&rbio->bio_list_lock); + spin_unlock_irqrestore(&h->lock, flags); + + if (next->read_rebuild) + async_read_rebuild(next); + else + async_rmw_stripe(next); + + goto done_nolock; + + } else if (waitqueue_active(&h->wait)) { + spin_unlock(&rbio->bio_list_lock); + spin_unlock_irqrestore(&h->lock, flags); + wake_up(&h->wait); + goto done_nolock; + } + } + spin_unlock(&rbio->bio_list_lock); + spin_unlock_irqrestore(&h->lock, flags); + +done_nolock: + return; +} + +static void __free_raid_bio(struct btrfs_raid_bio *rbio) +{ + int i; + + WARN_ON(atomic_read(&rbio->refs) < 0); + if (!atomic_dec_and_test(&rbio->refs)) + return; + + WARN_ON(!list_empty(&rbio->hash_list)); + WARN_ON(!bio_list_empty(&rbio->bio_list)); + + for (i = 0; i < rbio->nr_pages; i++) { + if (rbio->stripe_pages[i]) { + __free_page(rbio->stripe_pages[i]); + rbio->stripe_pages[i] = NULL; + } + } + kfree(rbio->raid_map); + kfree(rbio->bbio); + kfree(rbio); +} + +static void free_raid_bio(struct btrfs_raid_bio *rbio) +{ + unlock_stripe(rbio); + __free_raid_bio(rbio); +} + +/* + * this frees the rbio and runs through all the bios in the + * bio_list and calls end_io on them + */ +static void rbio_orig_end_io(struct btrfs_raid_bio *rbio, int err, int uptodate) +{ + struct bio *cur = bio_list_get(&rbio->bio_list); + struct bio *next; + free_raid_bio(rbio); + + while (cur) { + next = cur->bi_next; + cur->bi_next = NULL; + if (uptodate) + set_bit(BIO_UPTODATE, &cur->bi_flags); + bio_endio(cur, err); + cur = next; + } +} + +/* + * end io function used by finish_rmw. When we finally + * get here, we've written a full stripe + */ +static void raid_write_end_io(struct bio *bio, int err) +{ + struct btrfs_raid_bio *rbio = bio->bi_private; + + if (err) + fail_bio_stripe(rbio, bio); + + bio_put(bio); + + if (!atomic_dec_and_test(&rbio->bbio->stripes_pending)) + return; + + err = 0; + + /* OK, we have read all the stripes we need to. */ + if (atomic_read(&rbio->bbio->error) > rbio->bbio->max_errors) + err = -EIO; + + rbio_orig_end_io(rbio, err, 0); + return; +} + +/* + * the read/modify/write code wants to use the original bio for + * any pages it included, and then use the rbio for everything + * else. This function decides if a given index (stripe number) + * and page number in that stripe fall inside the original bio + * or the rbio. + * + * if you set bio_list_only, you'll get a NULL back for any ranges + * that are outside the bio_list + * + * This doesn't take any refs on anything, you get a bare page pointer + * and the caller must bump refs as required. + * + * You must call index_rbio_pages once before you can trust + * the answers from this function. + */ +static struct page *page_in_rbio(struct btrfs_raid_bio *rbio, + int index, int pagenr, int bio_list_only) +{ + int chunk_page; + struct page *p = NULL; + + chunk_page = index * (rbio->stripe_len >> PAGE_SHIFT) + pagenr; + + spin_lock_irq(&rbio->bio_list_lock); + p = rbio->bio_pages[chunk_page]; + spin_unlock_irq(&rbio->bio_list_lock); + + if (p || bio_list_only) + return p; + + return rbio->stripe_pages[chunk_page]; +} + +/* + * number of pages we need for the entire stripe across all the + * drives + */ +static unsigned long rbio_nr_pages(unsigned long stripe_len, int nr_stripes) +{ + unsigned long nr = stripe_len * nr_stripes; + return (nr + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT; +} + +/* + * allocation and initial setup for the btrfs_raid_bio. Not + * this does not allocate any pages for rbio->pages. + */ +static struct btrfs_raid_bio *alloc_rbio(struct btrfs_root *root, + struct btrfs_bio *bbio, u64 *raid_map, + u64 stripe_len) +{ + struct btrfs_raid_bio *rbio; + int nr_data = 0; + int num_pages = rbio_nr_pages(stripe_len, bbio->num_stripes); + void *p; + + rbio = kzalloc(sizeof(*rbio) + num_pages * sizeof(struct page *) * 2, + GFP_NOFS); + if (!rbio) { + kfree(raid_map); + kfree(bbio); + return ERR_PTR(-ENOMEM); + } + + bio_list_init(&rbio->bio_list); + INIT_LIST_HEAD(&rbio->plug_list); + spin_lock_init(&rbio->bio_list_lock); + INIT_LIST_HEAD(&rbio->hash_list); + rbio->bbio = bbio; + rbio->raid_map = raid_map; + rbio->fs_info = root->fs_info; + rbio->stripe_len = stripe_len; + rbio->nr_pages = num_pages; + rbio->faila = -1; + rbio->failb = -1; + atomic_set(&rbio->refs, 1); + + /* + * the stripe_pages and bio_pages array point to the extra + * memory we allocated past the end of the rbio + */ + p = rbio + 1; + rbio->stripe_pages = p; + rbio->bio_pages = p + sizeof(struct page *) * num_pages; + + if (raid_map[bbio->num_stripes - 1] == RAID6_Q_STRIPE) + nr_data = bbio->num_stripes - 2; + else + nr_data = bbio->num_stripes - 1; + + rbio->nr_data = nr_data; + return rbio; +} + +/* allocate pages for all the stripes in the bio, including parity */ +static int alloc_rbio_pages(struct btrfs_raid_bio *rbio) +{ + int i; + struct page *page; + + for (i = 0; i < rbio->nr_pages; i++) { + if (rbio->stripe_pages[i]) + continue; + page = alloc_page(GFP_NOFS | __GFP_HIGHMEM); + if (!page) + return -ENOMEM; + rbio->stripe_pages[i] = page; + ClearPageUptodate(page); + } + return 0; +} + +/* allocate pages for just the p/q stripes */ +static int alloc_rbio_parity_pages(struct btrfs_raid_bio *rbio) +{ + int i; + struct page *page; + + i = (rbio->nr_data * rbio->stripe_len) >> PAGE_CACHE_SHIFT; + + for (; i < rbio->nr_pages; i++) { + if (rbio->stripe_pages[i]) + continue; + page = alloc_page(GFP_NOFS | __GFP_HIGHMEM); + if (!page) + return -ENOMEM; + rbio->stripe_pages[i] = page; + } + return 0; +} + +/* + * add a single page from a specific stripe into our list of bios for IO + * this will try to merge into existing bios if possible, and returns + * zero if all went well. + */ +int rbio_add_io_page(struct btrfs_raid_bio *rbio, + struct bio_list *bio_list, + struct page *page, + int stripe_nr, + unsigned long page_index, + unsigned long bio_max_len) +{ + struct bio *last = bio_list->tail; + u64 last_end = 0; + int ret; + struct bio *bio; + struct btrfs_bio_stripe *stripe; + u64 disk_start; + + stripe = &rbio->bbio->stripes[stripe_nr]; + disk_start = stripe->physical + (page_index << PAGE_CACHE_SHIFT); + + /* if the device is missing, just fail this stripe */ + if (!stripe->dev->bdev) + return fail_rbio_index(rbio, stripe_nr); + + /* see if we can add this page onto our existing bio */ + if (last) { + last_end = (u64)last->bi_sector << 9; + last_end += last->bi_size; + + /* + * we can't merge these if they are from different + * devices or if they are not contiguous + */ + if (last_end == disk_start && stripe->dev->bdev && + test_bit(BIO_UPTODATE, &last->bi_flags) && + last->bi_bdev == stripe->dev->bdev) { + ret = bio_add_page(last, page, PAGE_CACHE_SIZE, 0); + if (ret == PAGE_CACHE_SIZE) + return 0; + } + } + + /* put a new bio on the list */ + bio = bio_alloc(GFP_NOFS, bio_max_len >> PAGE_SHIFT?:1); + if (!bio) + return -ENOMEM; + + bio->bi_size = 0; + bio->bi_bdev = stripe->dev->bdev; + bio->bi_sector = disk_start >> 9; + set_bit(BIO_UPTODATE, &bio->bi_flags); + + bio_add_page(bio, page, PAGE_CACHE_SIZE, 0); + bio_list_add(bio_list, bio); + return 0; +} + +/* + * while we're doing the read/modify/write cycle, we could + * have errors in reading pages off the disk. This checks + * for errors and if we're not able to read the page it'll + * trigger parity reconstruction. The rmw will be finished + * after we've reconstructed the failed stripes + */ +static void validate_rbio_for_rmw(struct btrfs_raid_bio *rbio) +{ + if (rbio->faila >= 0 || rbio->failb >= 0) { + BUG_ON(rbio->faila == rbio->bbio->num_stripes - 1); + __raid56_parity_recover(rbio); + } else { + finish_rmw(rbio); + } +} + +/* + * these are just the pages from the rbio array, not from anything + * the FS sent down to us + */ +static struct page *rbio_stripe_page(struct btrfs_raid_bio *rbio, int stripe, int page) +{ + int index; + index = stripe * (rbio->stripe_len >> PAGE_CACHE_SHIFT); + index += page; + return rbio->stripe_pages[index]; +} + +/* + * helper function to walk our bio list and populate the bio_pages array with + * the result. This seems expensive, but it is faster than constantly + * searching through the bio list as we setup the IO in finish_rmw or stripe + * reconstruction. + * + * This must be called before you trust the answers from page_in_rbio + */ +static void index_rbio_pages(struct btrfs_raid_bio *rbio) +{ + struct bio *bio; + u64 start; + unsigned long stripe_offset; + unsigned long page_index; + struct page *p; + int i; + + spin_lock_irq(&rbio->bio_list_lock); + bio_list_for_each(bio, &rbio->bio_list) { + start = (u64)bio->bi_sector << 9; + stripe_offset = start - rbio->raid_map[0]; + page_index = stripe_offset >> PAGE_CACHE_SHIFT; + + for (i = 0; i < bio->bi_vcnt; i++) { + p = bio->bi_io_vec[i].bv_page; + rbio->bio_pages[page_index + i] = p; + } + } + spin_unlock_irq(&rbio->bio_list_lock); +} + +/* + * this is called from one of two situations. We either + * have a full stripe from the higher layers, or we've read all + * the missing bits off disk. + * + * This will calculate the parity and then send down any + * changed blocks. + */ +static noinline void finish_rmw(struct btrfs_raid_bio *rbio) +{ + struct btrfs_bio *bbio = rbio->bbio; + void *pointers[bbio->num_stripes]; + int stripe_len = rbio->stripe_len; + int nr_data = rbio->nr_data; + int stripe; + int pagenr; + int p_stripe = -1; + int q_stripe = -1; + struct bio_list bio_list; + struct bio *bio; + int pages_per_stripe = stripe_len >> PAGE_CACHE_SHIFT; + int ret; + + bio_list_init(&bio_list); + + if (bbio->num_stripes - rbio->nr_data == 1) { + p_stripe = bbio->num_stripes - 1; + } else if (bbio->num_stripes - rbio->nr_data == 2) { + p_stripe = bbio->num_stripes - 2; + q_stripe = bbio->num_stripes - 1; + } else { + BUG(); + } + + /* at this point we either have a full stripe, + * or we've read the full stripe from the drive. + * recalculate the parity and write the new results. + * + * We're not allowed to add any new bios to the + * bio list here, anyone else that wants to + * change this stripe needs to do their own rmw. + */ + spin_lock_irq(&rbio->bio_list_lock); + set_bit(RBIO_RMW_LOCKED_BIT, &rbio->flags); + spin_unlock_irq(&rbio->bio_list_lock); + + atomic_set(&rbio->bbio->error, 0); + + /* + * now that we've set rmw_locked, run through the + * bio list one last time and map the page pointers + */ + index_rbio_pages(rbio); + + for (pagenr = 0; pagenr < pages_per_stripe; pagenr++) { + struct page *p; + /* first collect one page from each data stripe */ + for (stripe = 0; stripe < nr_data; stripe++) { + p = page_in_rbio(rbio, stripe, pagenr, 0); + pointers[stripe] = kmap(p); + } + + /* then add the parity stripe */ + p = rbio_pstripe_page(rbio, pagenr); + SetPageUptodate(p); + pointers[stripe++] = kmap(p); + + if (q_stripe != -1) { + + /* + * raid6, add the qstripe and call the + * library function to fill in our p/q + */ + p = rbio_qstripe_page(rbio, pagenr); + SetPageUptodate(p); + pointers[stripe++] = kmap(p); + + raid6_call.gen_syndrome(bbio->num_stripes, PAGE_SIZE, + pointers); + } else { + /* raid5 */ + memcpy(pointers[nr_data], pointers[0], PAGE_SIZE); + run_xor(pointers + 1, nr_data - 1, PAGE_CACHE_SIZE); + } + + + for (stripe = 0; stripe < bbio->num_stripes; stripe++) + kunmap(page_in_rbio(rbio, stripe, pagenr, 0)); + } + + /* + * time to start writing. Make bios for everything from the + * higher layers (the bio_list in our rbio) and our p/q. Ignore + * everything else. + */ + for (stripe = 0; stripe < bbio->num_stripes; stripe++) { + for (pagenr = 0; pagenr < pages_per_stripe; pagenr++) { + struct page *page; + if (stripe < rbio->nr_data) { + page = page_in_rbio(rbio, stripe, pagenr, 1); + if (!page) + continue; + } else { + page = rbio_stripe_page(rbio, stripe, pagenr); + } + + ret = rbio_add_io_page(rbio, &bio_list, + page, stripe, pagenr, rbio->stripe_len); + if (ret) + goto cleanup; + } + } + + atomic_set(&bbio->stripes_pending, bio_list_size(&bio_list)); + BUG_ON(atomic_read(&bbio->stripes_pending) == 0); + + while (1) { + bio = bio_list_pop(&bio_list); + if (!bio) + break; + + bio->bi_private = rbio; + bio->bi_end_io = raid_write_end_io; + BUG_ON(!test_bit(BIO_UPTODATE, &bio->bi_flags)); + submit_bio(WRITE, bio); + } + return; + +cleanup: + rbio_orig_end_io(rbio, -EIO, 0); +} + +/* + * helper to find the stripe number for a given bio. Used to figure out which + * stripe has failed. This expects the bio to correspond to a physical disk, + * so it looks up based on physical sector numbers. + */ +static int find_bio_stripe(struct btrfs_raid_bio *rbio, + struct bio *bio) +{ + u64 physical = bio->bi_sector; + u64 stripe_start; + int i; + struct btrfs_bio_stripe *stripe; + + physical <<= 9; + + for (i = 0; i < rbio->bbio->num_stripes; i++) { + stripe = &rbio->bbio->stripes[i]; + stripe_start = stripe->physical; + if (physical >= stripe_start && + physical < stripe_start + rbio->stripe_len) { + return i; + } + } + return -1; +} + +/* + * helper to find the stripe number for a given + * bio (before mapping). Used to figure out which stripe has + * failed. This looks up based on logical block numbers. + */ +static int find_logical_bio_stripe(struct btrfs_raid_bio *rbio, + struct bio *bio) +{ + u64 logical = bio->bi_sector; + u64 stripe_start; + int i; + + logical <<= 9; + + for (i = 0; i < rbio->nr_data; i++) { + stripe_start = rbio->raid_map[i]; + if (logical >= stripe_start && + logical < stripe_start + rbio->stripe_len) { + return i; + } + } + return -1; +} + +/* + * returns -EIO if we had too many failures + */ +static int fail_rbio_index(struct btrfs_raid_bio *rbio, int failed) +{ + unsigned long flags; + int ret = 0; + + spin_lock_irqsave(&rbio->bio_list_lock, flags); + + /* we already know this stripe is bad, move on */ + if (rbio->faila == failed || rbio->failb == failed) + goto out; + + if (rbio->faila == -1) { + /* first failure on this rbio */ + rbio->faila = failed; + atomic_inc(&rbio->bbio->error); + } else if (rbio->failb == -1) { + /* second failure on this rbio */ + rbio->failb = failed; + atomic_inc(&rbio->bbio->error); + } else { + ret = -EIO; + } +out: + spin_unlock_irqrestore(&rbio->bio_list_lock, flags); + + return ret; +} + +/* + * helper to fail a stripe based on a physical disk + * bio. + */ +static int fail_bio_stripe(struct btrfs_raid_bio *rbio, + struct bio *bio) +{ + int failed = find_bio_stripe(rbio, bio); + + if (failed < 0) + return -EIO; + + return fail_rbio_index(rbio, failed); +} + +/* + * this sets each page in the bio uptodate. It should only be used on private + * rbio pages, nothing that comes in from the higher layers + */ +static void set_bio_pages_uptodate(struct bio *bio) +{ + int i; + struct page *p; + + for (i = 0; i < bio->bi_vcnt; i++) { + p = bio->bi_io_vec[i].bv_page; + SetPageUptodate(p); + } +} + +/* + * end io for the read phase of the rmw cycle. All the bios here are physical + * stripe bios we've read from the disk so we can recalculate the parity of the + * stripe. + * + * This will usually kick off finish_rmw once all the bios are read in, but it + * may trigger parity reconstruction if we had any errors along the way + */ +static void raid_rmw_end_io(struct bio *bio, int err) +{ + struct btrfs_raid_bio *rbio = bio->bi_private; + + if (err) + fail_bio_stripe(rbio, bio); + else + set_bio_pages_uptodate(bio); + + bio_put(bio); + + if (!atomic_dec_and_test(&rbio->bbio->stripes_pending)) + return; + + err = 0; + if (atomic_read(&rbio->bbio->error) > rbio->bbio->max_errors) + goto cleanup; + + /* + * this will normally call finish_rmw to start our write + * but if there are any failed stripes we'll reconstruct + * from parity first + */ + validate_rbio_for_rmw(rbio); + return; + +cleanup: + + rbio_orig_end_io(rbio, -EIO, 0); +} + +static void async_rmw_stripe(struct btrfs_raid_bio *rbio) +{ + rbio->work.flags = 0; + rbio->work.func = rmw_work; + + btrfs_queue_worker(&rbio->fs_info->rmw_workers, + &rbio->work); +} + +static void async_read_rebuild(struct btrfs_raid_bio *rbio) +{ + rbio->work.flags = 0; + rbio->work.func = read_rebuild_work; + + btrfs_queue_worker(&rbio->fs_info->rmw_workers, + &rbio->work); +} + +/* + * the stripe must be locked by the caller. It will + * unlock after all the writes are done + */ +static int raid56_rmw_stripe(struct btrfs_raid_bio *rbio) +{ + int bios_to_read = 0; + struct btrfs_bio *bbio = rbio->bbio; + struct bio_list bio_list; + int ret; + int nr_pages = (rbio->stripe_len + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT; + int pagenr; + int stripe; + struct bio *bio; + + bio_list_init(&bio_list); + + ret = alloc_rbio_pages(rbio); + if (ret) + goto cleanup; + + index_rbio_pages(rbio); + + atomic_set(&rbio->bbio->error, 0); + /* + * build a list of bios to read all the missing parts of this + * stripe + */ + for (stripe = 0; stripe < rbio->nr_data; stripe++) { + for (pagenr = 0; pagenr < nr_pages; pagenr++) { + struct page *page; + /* + * we want to find all the pages missing from + * the rbio and read them from the disk. If + * page_in_rbio finds a page in the bio list + * we don't need to read it off the stripe. + */ + page = page_in_rbio(rbio, stripe, pagenr, 1); + if (page) + continue; + + page = rbio_stripe_page(rbio, stripe, pagenr); + ret = rbio_add_io_page(rbio, &bio_list, page, + stripe, pagenr, rbio->stripe_len); + if (ret) + goto cleanup; + } + } + + bios_to_read = bio_list_size(&bio_list); + if (!bios_to_read) { + /* + * this can happen if others have merged with + * us, it means there is nothing left to read. + * But if there are missing devices it may not be + * safe to do the full stripe write yet. + */ + goto finish; + } + + /* + * the bbio may be freed once we submit the last bio. Make sure + * not to touch it after that + */ + atomic_set(&bbio->stripes_pending, bios_to_read); + while (1) { + bio = bio_list_pop(&bio_list); + if (!bio) + break; + + bio->bi_private = rbio; + bio->bi_end_io = raid_rmw_end_io; + + btrfs_bio_wq_end_io(rbio->fs_info, bio, + BTRFS_WQ_ENDIO_RAID56); + + BUG_ON(!test_bit(BIO_UPTODATE, &bio->bi_flags)); + submit_bio(READ, bio); + } + /* the actual write will happen once the reads are done */ + return 0; + +cleanup: + rbio_orig_end_io(rbio, -EIO, 0); + return -EIO; + +finish: + validate_rbio_for_rmw(rbio); + return 0; +} + +/* + * if the upper layers pass in a full stripe, we thank them by only allocating + * enough pages to hold the parity, and sending it all down quickly. + */ +static int full_stripe_write(struct btrfs_raid_bio *rbio) +{ + int ret; + + ret = alloc_rbio_parity_pages(rbio); + if (ret) + return ret; + + ret = lock_stripe_add(rbio); + if (ret == 0) + finish_rmw(rbio); + return 0; +} + +/* + * partial stripe writes get handed over to async helpers. + * We're really hoping to merge a few more writes into this + * rbio before calculating new parity + */ +static int partial_stripe_write(struct btrfs_raid_bio *rbio) +{ + int ret; + + ret = lock_stripe_add(rbio); + if (ret == 0) + async_rmw_stripe(rbio); + return 0; +} + +/* + * sometimes while we were reading from the drive to + * recalculate parity, enough new bios come into create + * a full stripe. So we do a check here to see if we can + * go directly to finish_rmw + */ +static int __raid56_parity_write(struct btrfs_raid_bio *rbio) +{ + /* head off into rmw land if we don't have a full stripe */ + if (!rbio_is_full(rbio)) + return partial_stripe_write(rbio); + return full_stripe_write(rbio); +} + +/* + * our main entry point for writes from the rest of the FS. + */ +int raid56_parity_write(struct btrfs_root *root, struct bio *bio, + struct btrfs_bio *bbio, u64 *raid_map, + u64 stripe_len) +{ + struct btrfs_raid_bio *rbio; + + rbio = alloc_rbio(root, bbio, raid_map, stripe_len); + if (IS_ERR(rbio)) { + kfree(raid_map); + kfree(bbio); + return PTR_ERR(rbio); + } + bio_list_add(&rbio->bio_list, bio); + rbio->bio_list_bytes = bio->bi_size; + return __raid56_parity_write(rbio); +} + +/* + * all parity reconstruction happens here. We've read in everything + * we can find from the drives and this does the heavy lifting of + * sorting the good from the bad. + */ +static void __raid_recover_end_io(struct btrfs_raid_bio *rbio) +{ + int pagenr, stripe; + void **pointers; + int faila = -1, failb = -1; + int nr_pages = (rbio->stripe_len + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT; + struct page *page; + int err; + int i; + + pointers = kzalloc(rbio->bbio->num_stripes * sizeof(void *), + GFP_NOFS); + if (!pointers) { + err = -ENOMEM; + goto cleanup_io; + } + + faila = rbio->faila; + failb = rbio->failb; + + if (rbio->read_rebuild) { + spin_lock_irq(&rbio->bio_list_lock); + set_bit(RBIO_RMW_LOCKED_BIT, &rbio->flags); + spin_unlock_irq(&rbio->bio_list_lock); + } + + index_rbio_pages(rbio); + + for (pagenr = 0; pagenr < nr_pages; pagenr++) { + /* setup our array of pointers with pages + * from each stripe + */ + for (stripe = 0; stripe < rbio->bbio->num_stripes; stripe++) { + /* + * if we're rebuilding a read, we have to use + * pages from the bio list + */ + if (rbio->read_rebuild && + (stripe == faila || stripe == failb)) { + page = page_in_rbio(rbio, stripe, pagenr, 0); + } else { + page = rbio_stripe_page(rbio, stripe, pagenr); + } + pointers[stripe] = kmap(page); + } + + /* all raid6 handling here */ + if (rbio->raid_map[rbio->bbio->num_stripes - 1] == + RAID6_Q_STRIPE) { + + /* + * single failure, rebuild from parity raid5 + * style + */ + if (failb < 0) { + if (faila == rbio->nr_data) { + /* + * Just the P stripe has failed, without + * a bad data or Q stripe. + * TODO, we should redo the xor here. + */ + err = -EIO; + goto cleanup; + } + /* + * a single failure in raid6 is rebuilt + * in the pstripe code below + */ + goto pstripe; + } + + /* make sure our ps and qs are in order */ + if (faila > failb) { + int tmp = failb; + failb = faila; + faila = tmp; + } + + /* if the q stripe is failed, do a pstripe reconstruction + * from the xors. + * If both the q stripe and the P stripe are failed, we're + * here due to a crc mismatch and we can't give them the + * data they want + */ + if (rbio->raid_map[failb] == RAID6_Q_STRIPE) { + if (rbio->raid_map[faila] == RAID5_P_STRIPE) { + err = -EIO; + goto cleanup; + } + /* + * otherwise we have one bad data stripe and + * a good P stripe. raid5! + */ + goto pstripe; + } + + if (rbio->raid_map[failb] == RAID5_P_STRIPE) { + raid6_datap_recov(rbio->bbio->num_stripes, + PAGE_SIZE, faila, pointers); + } else { + raid6_2data_recov(rbio->bbio->num_stripes, + PAGE_SIZE, faila, failb, + pointers); + } + } else { + void *p; + + /* rebuild from P stripe here (raid5 or raid6) */ + BUG_ON(failb != -1); +pstripe: + /* Copy parity block into failed block to start with */ + memcpy(pointers[faila], + pointers[rbio->nr_data], + PAGE_CACHE_SIZE); + + /* rearrange the pointer array */ + p = pointers[faila]; + for (stripe = faila; stripe < rbio->nr_data - 1; stripe++) + pointers[stripe] = pointers[stripe + 1]; + pointers[rbio->nr_data - 1] = p; + + /* xor in the rest */ + run_xor(pointers, rbio->nr_data - 1, PAGE_CACHE_SIZE); + } + /* if we're doing this rebuild as part of an rmw, go through + * and set all of our private rbio pages in the + * failed stripes as uptodate. This way finish_rmw will + * know they can be trusted. If this was a read reconstruction, + * other endio functions will fiddle the uptodate bits + */ + if (!rbio->read_rebuild) { + for (i = 0; i < nr_pages; i++) { + if (faila != -1) { + page = rbio_stripe_page(rbio, faila, i); + SetPageUptodate(page); + } + if (failb != -1) { + page = rbio_stripe_page(rbio, failb, i); + SetPageUptodate(page); + } + } + } + for (stripe = 0; stripe < rbio->bbio->num_stripes; stripe++) { + /* + * if we're rebuilding a read, we have to use + * pages from the bio list + */ + if (rbio->read_rebuild && + (stripe == faila || stripe == failb)) { + page = page_in_rbio(rbio, stripe, pagenr, 0); + } else { + page = rbio_stripe_page(rbio, stripe, pagenr); + } + kunmap(page); + } + } + + err = 0; +cleanup: + kfree(pointers); + +cleanup_io: + + if (rbio->read_rebuild) { + rbio_orig_end_io(rbio, err, err == 0); + } else if (err == 0) { + rbio->faila = -1; + rbio->failb = -1; + finish_rmw(rbio); + } else { + rbio_orig_end_io(rbio, err, 0); + } +} + +/* + * This is called only for stripes we've read from disk to + * reconstruct the parity. + */ +static void raid_recover_end_io(struct bio *bio, int err) +{ + struct btrfs_raid_bio *rbio = bio->bi_private; + + /* + * we only read stripe pages off the disk, set them + * up to date if there were no errors + */ + if (err) + fail_bio_stripe(rbio, bio); + else + set_bio_pages_uptodate(bio); + bio_put(bio); + + if (!atomic_dec_and_test(&rbio->bbio->stripes_pending)) + return; + + if (atomic_read(&rbio->bbio->error) > rbio->bbio->max_errors) + rbio_orig_end_io(rbio, -EIO, 0); + else + __raid_recover_end_io(rbio); +} + +/* + * reads everything we need off the disk to reconstruct + * the parity. endio handlers trigger final reconstruction + * when the IO is done. + * + * This is used both for reads from the higher layers and for + * parity construction required to finish a rmw cycle. + */ +static int __raid56_parity_recover(struct btrfs_raid_bio *rbio) +{ + int bios_to_read = 0; + struct btrfs_bio *bbio = rbio->bbio; + struct bio_list bio_list; + int ret; + int nr_pages = (rbio->stripe_len + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT; + int pagenr; + int stripe; + struct bio *bio; + + bio_list_init(&bio_list); + + ret = alloc_rbio_pages(rbio); + if (ret) + goto cleanup; + + atomic_set(&rbio->bbio->error, 0); + + /* + * read everything that hasn't failed. + */ + for (stripe = 0; stripe < bbio->num_stripes; stripe++) { + if (rbio->faila == stripe || + rbio->failb == stripe) + continue; + + for (pagenr = 0; pagenr < nr_pages; pagenr++) { + struct page *p; + + /* + * the rmw code may have already read this + * page in + */ + p = rbio_stripe_page(rbio, stripe, pagenr); + if (PageUptodate(p)) + continue; + + ret = rbio_add_io_page(rbio, &bio_list, + rbio_stripe_page(rbio, stripe, pagenr), + stripe, pagenr, rbio->stripe_len); + if (ret < 0) + goto cleanup; + } + } + + bios_to_read = bio_list_size(&bio_list); + if (!bios_to_read) { + /* + * we might have no bios to read just because the pages + * were up to date, or we might have no bios to read because + * the devices were gone. + */ + if (atomic_read(&rbio->bbio->error) <= rbio->bbio->max_errors) { + __raid_recover_end_io(rbio); + goto out; + } else { + goto cleanup; + } + } + + /* + * the bbio may be freed once we submit the last bio. Make sure + * not to touch it after that + */ + atomic_set(&bbio->stripes_pending, bios_to_read); + while (1) { + bio = bio_list_pop(&bio_list); + if (!bio) + break; + + bio->bi_private = rbio; + bio->bi_end_io = raid_recover_end_io; + + btrfs_bio_wq_end_io(rbio->fs_info, bio, + BTRFS_WQ_ENDIO_RAID56); + + BUG_ON(!test_bit(BIO_UPTODATE, &bio->bi_flags)); + submit_bio(READ, bio); + } +out: + return 0; + +cleanup: + if (rbio->read_rebuild) + rbio_orig_end_io(rbio, -EIO, 0); + return -EIO; +} + +/* + * the main entry point for reads from the higher layers. This + * is really only called when the normal read path had a failure, + * so we assume the bio they send down corresponds to a failed part + * of the drive. + */ +int raid56_parity_recover(struct btrfs_root *root, struct bio *bio, + struct btrfs_bio *bbio, u64 *raid_map, + u64 stripe_len, int mirror_num) +{ + struct btrfs_raid_bio *rbio; + int ret; + + rbio = alloc_rbio(root, bbio, raid_map, stripe_len); + if (IS_ERR(rbio)) { + return PTR_ERR(rbio); + } + + rbio->read_rebuild = 1; + bio_list_add(&rbio->bio_list, bio); + rbio->bio_list_bytes = bio->bi_size; + + rbio->faila = find_logical_bio_stripe(rbio, bio); + if (rbio->faila == -1) { + BUG(); + kfree(rbio); + return -EIO; + } + + /* + * reconstruct from the q stripe if they are + * asking for mirror 3 + */ + if (mirror_num == 3) + rbio->failb = bbio->num_stripes - 2; + + ret = lock_stripe_add(rbio); + + /* + * __raid56_parity_recover will end the bio with + * any errors it hits. We don't want to return + * its error value up the stack because our caller + * will end up calling bio_endio with any nonzero + * return + */ + if (ret == 0) + __raid56_parity_recover(rbio); + /* + * our rbio has been added to the list of + * rbios that will be handled after the + * currently lock owner is done + */ + return 0; + +} + +static void rmw_work(struct btrfs_work *work) +{ + struct btrfs_raid_bio *rbio; + + rbio = container_of(work, struct btrfs_raid_bio, work); + raid56_rmw_stripe(rbio); +} + +static void read_rebuild_work(struct btrfs_work *work) +{ + struct btrfs_raid_bio *rbio; + + rbio = container_of(work, struct btrfs_raid_bio, work); + __raid56_parity_recover(rbio); +} diff --git a/fs/btrfs/raid56.h b/fs/btrfs/raid56.h new file mode 100644 index 000000000000..ea5d73bfdfbe --- /dev/null +++ b/fs/btrfs/raid56.h @@ -0,0 +1,51 @@ +/* + * Copyright (C) 2012 Fusion-io All rights reserved. + * Copyright (C) 2012 Intel Corp. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ + +#ifndef __BTRFS_RAID56__ +#define __BTRFS_RAID56__ +static inline int nr_parity_stripes(struct map_lookup *map) +{ + if (map->type & BTRFS_BLOCK_GROUP_RAID5) + return 1; + else if (map->type & BTRFS_BLOCK_GROUP_RAID6) + return 2; + else + return 0; +} + +static inline int nr_data_stripes(struct map_lookup *map) +{ + return map->num_stripes - nr_parity_stripes(map); +} +#define RAID5_P_STRIPE ((u64)-2) +#define RAID6_Q_STRIPE ((u64)-1) + +#define is_parity_stripe(x) (((x) == RAID5_P_STRIPE) || \ + ((x) == RAID6_Q_STRIPE)) + +int raid56_parity_recover(struct btrfs_root *root, struct bio *bio, + struct btrfs_bio *bbio, u64 *raid_map, + u64 stripe_len, int mirror_num); +int raid56_parity_write(struct btrfs_root *root, struct bio *bio, + struct btrfs_bio *bbio, u64 *raid_map, + u64 stripe_len); + +int btrfs_alloc_stripe_hash_table(struct btrfs_fs_info *info); +void btrfs_free_stripe_hash_table(struct btrfs_fs_info *info); +#endif diff --git a/fs/btrfs/scrub.c b/fs/btrfs/scrub.c index bdbb94f245c9..bc35ed4238b8 100644 --- a/fs/btrfs/scrub.c +++ b/fs/btrfs/scrub.c @@ -28,6 +28,7 @@ #include "dev-replace.h" #include "check-integrity.h" #include "rcu-string.h" +#include "raid56.h" /* * This is only the first step towards a full-features scrub. It reads all @@ -2246,6 +2247,13 @@ static noinline_for_stack int scrub_stripe(struct scrub_ctx *sctx, struct btrfs_device *extent_dev; int extent_mirror_num; + if (map->type & (BTRFS_BLOCK_GROUP_RAID5 | + BTRFS_BLOCK_GROUP_RAID6)) { + if (num >= nr_data_stripes(map)) { + return 0; + } + } + nstripes = length; offset = 0; do_div(nstripes, map->stripe_len); diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 87fac9a21ea5..a065dec0e330 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -686,7 +686,9 @@ int btrfs_write_marked_extents(struct btrfs_root *root, struct extent_state *cached_state = NULL; u64 start = 0; u64 end; + struct blk_plug plug; + blk_start_plug(&plug); while (!find_first_extent_bit(dirty_pages, start, &start, &end, mark, &cached_state)) { convert_extent_bit(dirty_pages, start, end, EXTENT_NEED_WAIT, @@ -700,6 +702,7 @@ int btrfs_write_marked_extents(struct btrfs_root *root, } if (err) werr = err; + blk_finish_plug(&plug); return werr; } diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index 485a5423e3c6..c372264b85bf 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -25,6 +25,8 @@ #include #include #include +#include +#include #include "compat.h" #include "ctree.h" #include "extent_map.h" @@ -32,6 +34,7 @@ #include "transaction.h" #include "print-tree.h" #include "volumes.h" +#include "raid56.h" #include "async-thread.h" #include "check-integrity.h" #include "rcu-string.h" @@ -1389,6 +1392,14 @@ int btrfs_rm_device(struct btrfs_root *root, char *device_path) } btrfs_dev_replace_unlock(&root->fs_info->dev_replace); + if ((all_avail & (BTRFS_BLOCK_GROUP_RAID5 | + BTRFS_BLOCK_GROUP_RAID6) && num_devices <= 3)) { + printk(KERN_ERR "btrfs: unable to go below three devices " + "on raid5 or raid6\n"); + ret = -EINVAL; + goto out; + } + if ((all_avail & BTRFS_BLOCK_GROUP_RAID10) && num_devices <= 4) { printk(KERN_ERR "btrfs: unable to go below four devices " "on raid10\n"); @@ -1403,6 +1414,21 @@ int btrfs_rm_device(struct btrfs_root *root, char *device_path) goto out; } + if ((all_avail & BTRFS_BLOCK_GROUP_RAID5) && + root->fs_info->fs_devices->rw_devices <= 2) { + printk(KERN_ERR "btrfs: unable to go below two " + "devices on raid5\n"); + ret = -EINVAL; + goto out; + } + if ((all_avail & BTRFS_BLOCK_GROUP_RAID6) && + root->fs_info->fs_devices->rw_devices <= 3) { + printk(KERN_ERR "btrfs: unable to go below three " + "devices on raid6\n"); + ret = -EINVAL; + goto out; + } + if (strcmp(device_path, "missing") == 0) { struct list_head *devices; struct btrfs_device *tmp; @@ -2657,11 +2683,15 @@ static int chunk_drange_filter(struct extent_buffer *leaf, return 0; if (btrfs_chunk_type(leaf, chunk) & (BTRFS_BLOCK_GROUP_DUP | - BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID10)) - factor = 2; - else - factor = 1; - factor = num_stripes / factor; + BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID10)) { + factor = num_stripes / 2; + } else if (btrfs_chunk_type(leaf, chunk) & BTRFS_BLOCK_GROUP_RAID5) { + factor = num_stripes - 1; + } else if (btrfs_chunk_type(leaf, chunk) & BTRFS_BLOCK_GROUP_RAID6) { + factor = num_stripes - 2; + } else { + factor = num_stripes; + } for (i = 0; i < num_stripes; i++) { stripe = btrfs_stripe_nr(chunk, i); @@ -2976,6 +3006,7 @@ int btrfs_balance(struct btrfs_balance_control *bctl, int mixed = 0; int ret; u64 num_devices; + int cancel = 0; if (btrfs_fs_closing(fs_info) || atomic_read(&fs_info->balance_pause_req) || @@ -3018,7 +3049,9 @@ int btrfs_balance(struct btrfs_balance_control *bctl, allowed |= (BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1); else allowed |= (BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 | - BTRFS_BLOCK_GROUP_RAID10); + BTRFS_BLOCK_GROUP_RAID10 | + BTRFS_BLOCK_GROUP_RAID5 | + BTRFS_BLOCK_GROUP_RAID6); if ((bctl->data.flags & BTRFS_BALANCE_ARGS_CONVERT) && (!alloc_profile_is_valid(bctl->data.target, 1) || @@ -3058,7 +3091,10 @@ int btrfs_balance(struct btrfs_balance_control *bctl, /* allow to reduce meta or sys integrity only if force set */ allowed = BTRFS_BLOCK_GROUP_DUP | BTRFS_BLOCK_GROUP_RAID1 | - BTRFS_BLOCK_GROUP_RAID10; + BTRFS_BLOCK_GROUP_RAID10 | + BTRFS_BLOCK_GROUP_RAID5 | + BTRFS_BLOCK_GROUP_RAID6; + if (((bctl->sys.flags & BTRFS_BALANCE_ARGS_CONVERT) && (fs_info->avail_system_alloc_bits & allowed) && !(bctl->sys.target & allowed)) || @@ -3124,15 +3160,17 @@ int btrfs_balance(struct btrfs_balance_control *bctl, } if ((ret && ret != -ECANCELED && ret != -ENOSPC) || - balance_need_close(fs_info)) { - __cancel_balance(fs_info); - } + balance_need_close(fs_info)) + cancel = 1; if (bctl->sys.flags & BTRFS_BALANCE_ARGS_CONVERT) { fs_info->num_tolerated_disk_barrier_failures = btrfs_calc_num_tolerated_disk_barrier_failures(fs_info); } + if (cancel) + __cancel_balance(fs_info); + wake_up(&fs_info->balance_wait_q); return ret; @@ -3493,13 +3531,45 @@ static int btrfs_cmp_device_info(const void *a, const void *b) } struct btrfs_raid_attr btrfs_raid_array[BTRFS_NR_RAID_TYPES] = { + /* + * sub_stripes info for map, + * dev_stripes -- stripes per dev, 2 for DUP, 1 other wise + * devs_max -- max devices per stripe, 0 for unlimited + * devs_min -- min devices per stripe + * devs_increment -- ndevs must be a multiple of this + * ncopies -- how many copies of the data we have + */ { 2, 1, 0, 4, 2, 2 /* raid10 */ }, { 1, 1, 2, 2, 2, 2 /* raid1 */ }, { 1, 2, 1, 1, 1, 2 /* dup */ }, { 1, 1, 0, 2, 1, 1 /* raid0 */ }, { 1, 1, 0, 1, 1, 1 /* single */ }, + { 1, 1, 0, 2, 1, 2 /* raid5 */ }, + { 1, 1, 0, 3, 1, 3 /* raid6 */ }, }; +static u32 find_raid56_stripe_len(u32 data_devices, u32 dev_stripe_target) +{ + /* TODO allow them to set a preferred stripe size */ + return 64 * 1024; +} + +static void check_raid56_incompat_flag(struct btrfs_fs_info *info, u64 type) +{ + u64 features; + + if (!(type & (BTRFS_BLOCK_GROUP_RAID5 | BTRFS_BLOCK_GROUP_RAID6))) + return; + + features = btrfs_super_incompat_flags(info->super_copy); + if (features & BTRFS_FEATURE_INCOMPAT_RAID56) + return; + + features |= BTRFS_FEATURE_INCOMPAT_RAID56; + btrfs_set_super_incompat_flags(info->super_copy, features); + printk(KERN_INFO "btrfs: setting RAID5/6 feature flag\n"); +} + static int __btrfs_alloc_chunk(struct btrfs_trans_handle *trans, struct btrfs_root *extent_root, struct map_lookup **map_ret, @@ -3515,6 +3585,8 @@ static int __btrfs_alloc_chunk(struct btrfs_trans_handle *trans, struct btrfs_device_info *devices_info = NULL; u64 total_avail; int num_stripes; /* total number of stripes to allocate */ + int data_stripes; /* number of stripes that count for + block group size */ int sub_stripes; /* sub_stripes info for map */ int dev_stripes; /* stripes per dev */ int devs_max; /* max devs to use */ @@ -3526,6 +3598,7 @@ static int __btrfs_alloc_chunk(struct btrfs_trans_handle *trans, u64 max_chunk_size; u64 stripe_size; u64 num_bytes; + u64 raid_stripe_len = BTRFS_STRIPE_LEN; int ndevs; int i; int j; @@ -3651,16 +3724,31 @@ static int __btrfs_alloc_chunk(struct btrfs_trans_handle *trans, stripe_size = devices_info[ndevs-1].max_avail; num_stripes = ndevs * dev_stripes; + /* + * this will have to be fixed for RAID1 and RAID10 over + * more drives + */ + data_stripes = num_stripes / ncopies; + if (stripe_size * ndevs > max_chunk_size * ncopies) { stripe_size = max_chunk_size * ncopies; do_div(stripe_size, ndevs); } - + if (type & BTRFS_BLOCK_GROUP_RAID5) { + raid_stripe_len = find_raid56_stripe_len(ndevs - 1, + btrfs_super_stripesize(info->super_copy)); + data_stripes = num_stripes - 1; + } + if (type & BTRFS_BLOCK_GROUP_RAID6) { + raid_stripe_len = find_raid56_stripe_len(ndevs - 2, + btrfs_super_stripesize(info->super_copy)); + data_stripes = num_stripes - 2; + } do_div(stripe_size, dev_stripes); /* align to BTRFS_STRIPE_LEN */ - do_div(stripe_size, BTRFS_STRIPE_LEN); - stripe_size *= BTRFS_STRIPE_LEN; + do_div(stripe_size, raid_stripe_len); + stripe_size *= raid_stripe_len; map = kmalloc(map_lookup_size(num_stripes), GFP_NOFS); if (!map) { @@ -3678,14 +3766,14 @@ static int __btrfs_alloc_chunk(struct btrfs_trans_handle *trans, } } map->sector_size = extent_root->sectorsize; - map->stripe_len = BTRFS_STRIPE_LEN; - map->io_align = BTRFS_STRIPE_LEN; - map->io_width = BTRFS_STRIPE_LEN; + map->stripe_len = raid_stripe_len; + map->io_align = raid_stripe_len; + map->io_width = raid_stripe_len; map->type = type; map->sub_stripes = sub_stripes; *map_ret = map; - num_bytes = stripe_size * (num_stripes / ncopies); + num_bytes = stripe_size * data_stripes; *stripe_size_out = stripe_size; *num_bytes_out = num_bytes; @@ -3734,6 +3822,8 @@ static int __btrfs_alloc_chunk(struct btrfs_trans_handle *trans, } } + check_raid56_incompat_flag(extent_root->fs_info, type); + kfree(devices_info); return 0; @@ -4003,6 +4093,10 @@ int btrfs_num_copies(struct btrfs_fs_info *fs_info, u64 logical, u64 len) ret = map->num_stripes; else if (map->type & BTRFS_BLOCK_GROUP_RAID10) ret = map->sub_stripes; + else if (map->type & BTRFS_BLOCK_GROUP_RAID5) + ret = 2; + else if (map->type & BTRFS_BLOCK_GROUP_RAID6) + ret = 3; else ret = 1; free_extent_map(em); @@ -4015,6 +4109,52 @@ int btrfs_num_copies(struct btrfs_fs_info *fs_info, u64 logical, u64 len) return ret; } +unsigned long btrfs_full_stripe_len(struct btrfs_root *root, + struct btrfs_mapping_tree *map_tree, + u64 logical) +{ + struct extent_map *em; + struct map_lookup *map; + struct extent_map_tree *em_tree = &map_tree->map_tree; + unsigned long len = root->sectorsize; + + read_lock(&em_tree->lock); + em = lookup_extent_mapping(em_tree, logical, len); + read_unlock(&em_tree->lock); + BUG_ON(!em); + + BUG_ON(em->start > logical || em->start + em->len < logical); + map = (struct map_lookup *)em->bdev; + if (map->type & (BTRFS_BLOCK_GROUP_RAID5 | + BTRFS_BLOCK_GROUP_RAID6)) { + len = map->stripe_len * nr_data_stripes(map); + } + free_extent_map(em); + return len; +} + +int btrfs_is_parity_mirror(struct btrfs_mapping_tree *map_tree, + u64 logical, u64 len, int mirror_num) +{ + struct extent_map *em; + struct map_lookup *map; + struct extent_map_tree *em_tree = &map_tree->map_tree; + int ret = 0; + + read_lock(&em_tree->lock); + em = lookup_extent_mapping(em_tree, logical, len); + read_unlock(&em_tree->lock); + BUG_ON(!em); + + BUG_ON(em->start > logical || em->start + em->len < logical); + map = (struct map_lookup *)em->bdev; + if (map->type & (BTRFS_BLOCK_GROUP_RAID5 | + BTRFS_BLOCK_GROUP_RAID6)) + ret = 1; + free_extent_map(em); + return ret; +} + static int find_live_mirror(struct btrfs_fs_info *fs_info, struct map_lookup *map, int first, int num, int optimal, int dev_replace_is_ongoing) @@ -4052,10 +4192,39 @@ static int find_live_mirror(struct btrfs_fs_info *fs_info, return optimal; } +static inline int parity_smaller(u64 a, u64 b) +{ + return a > b; +} + +/* Bubble-sort the stripe set to put the parity/syndrome stripes last */ +static void sort_parity_stripes(struct btrfs_bio *bbio, u64 *raid_map) +{ + struct btrfs_bio_stripe s; + int i; + u64 l; + int again = 1; + + while (again) { + again = 0; + for (i = 0; i < bbio->num_stripes - 1; i++) { + if (parity_smaller(raid_map[i], raid_map[i+1])) { + s = bbio->stripes[i]; + l = raid_map[i]; + bbio->stripes[i] = bbio->stripes[i+1]; + raid_map[i] = raid_map[i+1]; + bbio->stripes[i+1] = s; + raid_map[i+1] = l; + again = 1; + } + } + } +} + static int __btrfs_map_block(struct btrfs_fs_info *fs_info, int rw, u64 logical, u64 *length, struct btrfs_bio **bbio_ret, - int mirror_num) + int mirror_num, u64 **raid_map_ret) { struct extent_map *em; struct map_lookup *map; @@ -4067,6 +4236,8 @@ static int __btrfs_map_block(struct btrfs_fs_info *fs_info, int rw, u64 stripe_nr; u64 stripe_nr_orig; u64 stripe_nr_end; + u64 stripe_len; + u64 *raid_map = NULL; int stripe_index; int i; int ret = 0; @@ -4078,6 +4249,7 @@ static int __btrfs_map_block(struct btrfs_fs_info *fs_info, int rw, int num_alloc_stripes; int patch_the_first_stripe_for_dev_replace = 0; u64 physical_to_patch_in_first_stripe = 0; + u64 raid56_full_stripe_start = (u64)-1; read_lock(&em_tree->lock); em = lookup_extent_mapping(em_tree, logical, *length); @@ -4094,29 +4266,63 @@ static int __btrfs_map_block(struct btrfs_fs_info *fs_info, int rw, map = (struct map_lookup *)em->bdev; offset = logical - em->start; + if (mirror_num > map->num_stripes) + mirror_num = 0; + + stripe_len = map->stripe_len; stripe_nr = offset; /* * stripe_nr counts the total number of stripes we have to stride * to get to this block */ - do_div(stripe_nr, map->stripe_len); + do_div(stripe_nr, stripe_len); - stripe_offset = stripe_nr * map->stripe_len; + stripe_offset = stripe_nr * stripe_len; BUG_ON(offset < stripe_offset); /* stripe_offset is the offset of this block in its stripe*/ stripe_offset = offset - stripe_offset; - if (rw & REQ_DISCARD) + /* if we're here for raid56, we need to know the stripe aligned start */ + if (map->type & (BTRFS_BLOCK_GROUP_RAID5 | BTRFS_BLOCK_GROUP_RAID6)) { + unsigned long full_stripe_len = stripe_len * nr_data_stripes(map); + raid56_full_stripe_start = offset; + + /* allow a write of a full stripe, but make sure we don't + * allow straddling of stripes + */ + do_div(raid56_full_stripe_start, full_stripe_len); + raid56_full_stripe_start *= full_stripe_len; + } + + if (rw & REQ_DISCARD) { + /* we don't discard raid56 yet */ + if (map->type & + (BTRFS_BLOCK_GROUP_RAID5 | BTRFS_BLOCK_GROUP_RAID6)) { + ret = -EOPNOTSUPP; + goto out; + } *length = min_t(u64, em->len - offset, *length); - else if (map->type & BTRFS_BLOCK_GROUP_PROFILE_MASK) { - /* we limit the length of each bio to what fits in a stripe */ - *length = min_t(u64, em->len - offset, - map->stripe_len - stripe_offset); + } else if (map->type & BTRFS_BLOCK_GROUP_PROFILE_MASK) { + u64 max_len; + /* For writes to RAID[56], allow a full stripeset across all disks. + For other RAID types and for RAID[56] reads, just allow a single + stripe (on a single disk). */ + if (map->type & (BTRFS_BLOCK_GROUP_RAID5 | BTRFS_BLOCK_GROUP_RAID6) && + (rw & REQ_WRITE)) { + max_len = stripe_len * nr_data_stripes(map) - + (offset - raid56_full_stripe_start); + } else { + /* we limit the length of each bio to what fits in a stripe */ + max_len = stripe_len - stripe_offset; + } + *length = min_t(u64, em->len - offset, max_len); } else { *length = em->len - offset; } + /* This is for when we're called from btrfs_merge_bio_hook() and all + it cares about is the length */ if (!bbio_ret) goto out; @@ -4149,7 +4355,7 @@ static int __btrfs_map_block(struct btrfs_fs_info *fs_info, int rw, u64 physical_of_found = 0; ret = __btrfs_map_block(fs_info, REQ_GET_READ_MIRRORS, - logical, &tmp_length, &tmp_bbio, 0); + logical, &tmp_length, &tmp_bbio, 0, NULL); if (ret) { WARN_ON(tmp_bbio != NULL); goto out; @@ -4215,6 +4421,7 @@ static int __btrfs_map_block(struct btrfs_fs_info *fs_info, int rw, do_div(stripe_nr_end, map->stripe_len); stripe_end_offset = stripe_nr_end * map->stripe_len - (offset + *length); + if (map->type & BTRFS_BLOCK_GROUP_RAID0) { if (rw & REQ_DISCARD) num_stripes = min_t(u64, map->num_stripes, @@ -4265,6 +4472,65 @@ static int __btrfs_map_block(struct btrfs_fs_info *fs_info, int rw, dev_replace_is_ongoing); mirror_num = stripe_index - old_stripe_index + 1; } + + } else if (map->type & (BTRFS_BLOCK_GROUP_RAID5 | + BTRFS_BLOCK_GROUP_RAID6)) { + u64 tmp; + + if (bbio_ret && ((rw & REQ_WRITE) || mirror_num > 1) + && raid_map_ret) { + int i, rot; + + /* push stripe_nr back to the start of the full stripe */ + stripe_nr = raid56_full_stripe_start; + do_div(stripe_nr, stripe_len); + + stripe_index = do_div(stripe_nr, nr_data_stripes(map)); + + /* RAID[56] write or recovery. Return all stripes */ + num_stripes = map->num_stripes; + max_errors = nr_parity_stripes(map); + + raid_map = kmalloc(sizeof(u64) * num_stripes, + GFP_NOFS); + if (!raid_map) { + ret = -ENOMEM; + goto out; + } + + /* Work out the disk rotation on this stripe-set */ + tmp = stripe_nr; + rot = do_div(tmp, num_stripes); + + /* Fill in the logical address of each stripe */ + tmp = stripe_nr * nr_data_stripes(map); + for (i = 0; i < nr_data_stripes(map); i++) + raid_map[(i+rot) % num_stripes] = + em->start + (tmp + i) * map->stripe_len; + + raid_map[(i+rot) % map->num_stripes] = RAID5_P_STRIPE; + if (map->type & BTRFS_BLOCK_GROUP_RAID6) + raid_map[(i+rot+1) % num_stripes] = + RAID6_Q_STRIPE; + + *length = map->stripe_len; + stripe_index = 0; + stripe_offset = 0; + } else { + /* + * Mirror #0 or #1 means the original data block. + * Mirror #2 is RAID5 parity block. + * Mirror #3 is RAID6 Q block. + */ + stripe_index = do_div(stripe_nr, nr_data_stripes(map)); + if (mirror_num > 1) + stripe_index = nr_data_stripes(map) + + mirror_num - 2; + + /* We distribute the parity blocks across stripes */ + tmp = stripe_nr + stripe_index; + stripe_index = do_div(tmp, map->num_stripes); + } } else { /* * after this do_div call, stripe_nr is the number of stripes @@ -4373,8 +4639,11 @@ static int __btrfs_map_block(struct btrfs_fs_info *fs_info, int rw, if (rw & (REQ_WRITE | REQ_GET_READ_MIRRORS)) { if (map->type & (BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID10 | + BTRFS_BLOCK_GROUP_RAID5 | BTRFS_BLOCK_GROUP_DUP)) { max_errors = 1; + } else if (map->type & BTRFS_BLOCK_GROUP_RAID6) { + max_errors = 2; } } @@ -4475,6 +4744,10 @@ static int __btrfs_map_block(struct btrfs_fs_info *fs_info, int rw, bbio->stripes[0].physical = physical_to_patch_in_first_stripe; bbio->mirror_num = map->num_stripes + 1; } + if (raid_map) { + sort_parity_stripes(bbio, raid_map); + *raid_map_ret = raid_map; + } out: if (dev_replace_is_ongoing) btrfs_dev_replace_unlock(dev_replace); @@ -4487,7 +4760,7 @@ int btrfs_map_block(struct btrfs_fs_info *fs_info, int rw, struct btrfs_bio **bbio_ret, int mirror_num) { return __btrfs_map_block(fs_info, rw, logical, length, bbio_ret, - mirror_num); + mirror_num, NULL); } int btrfs_rmap_block(struct btrfs_mapping_tree *map_tree, @@ -4501,6 +4774,7 @@ int btrfs_rmap_block(struct btrfs_mapping_tree *map_tree, u64 bytenr; u64 length; u64 stripe_nr; + u64 rmap_len; int i, j, nr = 0; read_lock(&em_tree->lock); @@ -4511,10 +4785,17 @@ int btrfs_rmap_block(struct btrfs_mapping_tree *map_tree, map = (struct map_lookup *)em->bdev; length = em->len; + rmap_len = map->stripe_len; + if (map->type & BTRFS_BLOCK_GROUP_RAID10) do_div(length, map->num_stripes / map->sub_stripes); else if (map->type & BTRFS_BLOCK_GROUP_RAID0) do_div(length, map->num_stripes); + else if (map->type & (BTRFS_BLOCK_GROUP_RAID5 | + BTRFS_BLOCK_GROUP_RAID6)) { + do_div(length, nr_data_stripes(map)); + rmap_len = map->stripe_len * nr_data_stripes(map); + } buf = kzalloc(sizeof(u64) * map->num_stripes, GFP_NOFS); BUG_ON(!buf); /* -ENOMEM */ @@ -4534,8 +4815,11 @@ int btrfs_rmap_block(struct btrfs_mapping_tree *map_tree, do_div(stripe_nr, map->sub_stripes); } else if (map->type & BTRFS_BLOCK_GROUP_RAID0) { stripe_nr = stripe_nr * map->num_stripes + i; - } - bytenr = chunk_start + stripe_nr * map->stripe_len; + } /* else if RAID[56], multiply by nr_data_stripes(). + * Alternatively, just use rmap_len below instead of + * map->stripe_len */ + + bytenr = chunk_start + stripe_nr * rmap_len; WARN_ON(nr >= map->num_stripes); for (j = 0; j < nr; j++) { if (buf[j] == bytenr) @@ -4549,7 +4833,7 @@ int btrfs_rmap_block(struct btrfs_mapping_tree *map_tree, *logical = buf; *naddrs = nr; - *stripe_len = map->stripe_len; + *stripe_len = rmap_len; free_extent_map(em); return 0; @@ -4623,7 +4907,7 @@ static void btrfs_end_bio(struct bio *bio, int err) bio->bi_bdev = (struct block_device *) (unsigned long)bbio->mirror_num; /* only send an error to the higher layers if it is - * beyond the tolerance of the multi-bio + * beyond the tolerance of the btrfs bio */ if (atomic_read(&bbio->error) > bbio->max_errors) { err = -EIO; @@ -4657,13 +4941,18 @@ struct async_sched { * This will add one bio to the pending list for a device and make sure * the work struct is scheduled. */ -static noinline void schedule_bio(struct btrfs_root *root, +noinline void btrfs_schedule_bio(struct btrfs_root *root, struct btrfs_device *device, int rw, struct bio *bio) { int should_queue = 1; struct btrfs_pending_bios *pending_bios; + if (device->missing || !device->bdev) { + bio_endio(bio, -EIO); + return; + } + /* don't bother with additional async steps for reads, right now */ if (!(rw & REQ_WRITE)) { bio_get(bio); @@ -4761,7 +5050,7 @@ static void submit_stripe_bio(struct btrfs_root *root, struct btrfs_bio *bbio, #endif bio->bi_bdev = dev->bdev; if (async) - schedule_bio(root, dev, rw, bio); + btrfs_schedule_bio(root, dev, rw, bio); else btrfsic_submit_bio(rw, bio); } @@ -4820,6 +5109,7 @@ int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio, u64 logical = (u64)bio->bi_sector << 9; u64 length = 0; u64 map_length; + u64 *raid_map = NULL; int ret; int dev_nr = 0; int total_devs = 1; @@ -4828,12 +5118,30 @@ int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio, length = bio->bi_size; map_length = length; - ret = btrfs_map_block(root->fs_info, rw, logical, &map_length, &bbio, - mirror_num); - if (ret) + ret = __btrfs_map_block(root->fs_info, rw, logical, &map_length, &bbio, + mirror_num, &raid_map); + if (ret) /* -ENOMEM */ return ret; total_devs = bbio->num_stripes; + bbio->orig_bio = first_bio; + bbio->private = first_bio->bi_private; + bbio->end_io = first_bio->bi_end_io; + atomic_set(&bbio->stripes_pending, bbio->num_stripes); + + if (raid_map) { + /* In this case, map_length has been set to the length of + a single stripe; not the whole write */ + if (rw & WRITE) { + return raid56_parity_write(root, bio, bbio, + raid_map, map_length); + } else { + return raid56_parity_recover(root, bio, bbio, + raid_map, map_length, + mirror_num); + } + } + if (map_length < length) { printk(KERN_CRIT "btrfs: mapping failed logical %llu bio len %llu " "len %llu\n", (unsigned long long)logical, @@ -4842,11 +5150,6 @@ int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio, BUG(); } - bbio->orig_bio = first_bio; - bbio->private = first_bio->bi_private; - bbio->end_io = first_bio->bi_end_io; - atomic_set(&bbio->stripes_pending, bbio->num_stripes); - while (dev_nr < total_devs) { dev = bbio->stripes[dev_nr].dev; if (!dev || !dev->bdev || (rw & WRITE && !dev->writeable)) { diff --git a/fs/btrfs/volumes.h b/fs/btrfs/volumes.h index d3c3939ac751..0c2b856ecd98 100644 --- a/fs/btrfs/volumes.h +++ b/fs/btrfs/volumes.h @@ -321,7 +321,14 @@ void btrfs_destroy_dev_replace_tgtdev(struct btrfs_fs_info *fs_info, void btrfs_init_dev_replace_tgtdev_for_resume(struct btrfs_fs_info *fs_info, struct btrfs_device *tgtdev); int btrfs_scratch_superblock(struct btrfs_device *device); - +void btrfs_schedule_bio(struct btrfs_root *root, + struct btrfs_device *device, + int rw, struct bio *bio); +int btrfs_is_parity_mirror(struct btrfs_mapping_tree *map_tree, + u64 logical, u64 len, int mirror_num); +unsigned long btrfs_full_stripe_len(struct btrfs_root *root, + struct btrfs_mapping_tree *map_tree, + u64 logical); static inline void btrfs_dev_stat_inc(struct btrfs_device *dev, int index) { -- cgit v1.2.3 From bb721703aa551e98dc5c7fb259cf90343408baf2 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 29 Jan 2013 18:44:12 -0500 Subject: Btrfs: reduce CPU contention while waiting for delayed extent operations We batch up operations to the extent allocation tree, which allows us to deal with the recursive nature of using the extent allocation tree to allocate extents to the extent allocation tree. It also provides a mechanism to sort and collect extent operations, which makes it much more efficient to record extents that are close together. The delayed extent operations must all be finished before the running transaction commits, so we have code to make sure and run a few of the batched operations when closing our transaction handles. This creates a great deal of contention for the locks in the delayed extent operation tree, and also contention for the lock on the extent allocation tree itself. All the extra contention just slows down the operations and doesn't get things done any faster. This commit changes things to use a wait queue instead. As procs want to run the delayed operations, one of them races in and gets permission to hit the tree, and the others step back and wait for progress to be made. Signed-off-by: Chris Mason --- fs/btrfs/delayed-ref.h | 9 ++++++++ fs/btrfs/extent-tree.c | 60 ++++++++++++++++++++++++++++++++++++++++++++++---- fs/btrfs/transaction.c | 6 ++++- 3 files changed, 70 insertions(+), 5 deletions(-) (limited to 'fs/btrfs/transaction.c') diff --git a/fs/btrfs/delayed-ref.h b/fs/btrfs/delayed-ref.h index c9d703693df0..23bdeb8502a6 100644 --- a/fs/btrfs/delayed-ref.h +++ b/fs/btrfs/delayed-ref.h @@ -131,6 +131,15 @@ struct btrfs_delayed_ref_root { /* total number of head nodes ready for processing */ unsigned long num_heads_ready; + /* + * bumped when someone is making progress on the delayed + * refs, so that other procs know they are just adding to + * contention intead of helping + */ + atomic_t procs_running_refs; + atomic_t ref_seq; + wait_queue_head_t wait; + /* * set when the tree is flushing before a transaction commit, * used by the throttling code to decide if new updates need diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index f13402104c96..87b0e856b6d0 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -2438,6 +2438,16 @@ int btrfs_delayed_refs_qgroup_accounting(struct btrfs_trans_handle *trans, return ret; } +static int refs_newer(struct btrfs_delayed_ref_root *delayed_refs, int seq, + int count) +{ + int val = atomic_read(&delayed_refs->ref_seq); + + if (val < seq || val >= seq + count) + return 1; + return 0; +} + /* * this starts processing the delayed reference count updates and * extent insertions we have queued up so far. count can be @@ -2472,6 +2482,44 @@ int btrfs_run_delayed_refs(struct btrfs_trans_handle *trans, delayed_refs = &trans->transaction->delayed_refs; INIT_LIST_HEAD(&cluster); + if (count == 0) { + count = delayed_refs->num_entries * 2; + run_most = 1; + } + + if (!run_all && !run_most) { + int old; + int seq = atomic_read(&delayed_refs->ref_seq); + +progress: + old = atomic_cmpxchg(&delayed_refs->procs_running_refs, 0, 1); + if (old) { + DEFINE_WAIT(__wait); + if (delayed_refs->num_entries < 16348) + return 0; + + prepare_to_wait(&delayed_refs->wait, &__wait, + TASK_UNINTERRUPTIBLE); + + old = atomic_cmpxchg(&delayed_refs->procs_running_refs, 0, 1); + if (old) { + schedule(); + finish_wait(&delayed_refs->wait, &__wait); + + if (!refs_newer(delayed_refs, seq, 256)) + goto progress; + else + return 0; + } else { + finish_wait(&delayed_refs->wait, &__wait); + goto again; + } + } + + } else { + atomic_inc(&delayed_refs->procs_running_refs); + } + again: loops = 0; spin_lock(&delayed_refs->lock); @@ -2480,10 +2528,6 @@ again: delayed_refs->run_delayed_start = find_middle(&delayed_refs->root); #endif - if (count == 0) { - count = delayed_refs->num_entries * 2; - run_most = 1; - } while (1) { if (!(run_all || run_most) && delayed_refs->num_heads_ready < 64) @@ -2505,9 +2549,12 @@ again: if (ret < 0) { spin_unlock(&delayed_refs->lock); btrfs_abort_transaction(trans, root, ret); + atomic_dec(&delayed_refs->procs_running_refs); return ret; } + atomic_add(ret, &delayed_refs->ref_seq); + count -= min_t(unsigned long, ret, count); if (count == 0) @@ -2576,6 +2623,11 @@ again: goto again; } out: + atomic_dec(&delayed_refs->procs_running_refs); + smp_mb(); + if (waitqueue_active(&delayed_refs->wait)) + wake_up(&delayed_refs->wait); + spin_unlock(&delayed_refs->lock); assert_qgroups_uptodate(trans); return 0; diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index a065dec0e330..1e7f176bd211 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -156,6 +156,9 @@ loop: spin_lock_init(&cur_trans->commit_lock); spin_lock_init(&cur_trans->delayed_refs.lock); + atomic_set(&cur_trans->delayed_refs.procs_running_refs, 0); + atomic_set(&cur_trans->delayed_refs.ref_seq, 0); + init_waitqueue_head(&cur_trans->delayed_refs.wait); INIT_LIST_HEAD(&cur_trans->pending_snapshots); list_add_tail(&cur_trans->list, &fs_info->trans_list); @@ -577,7 +580,7 @@ static int __btrfs_end_transaction(struct btrfs_trans_handle *trans, if (!list_empty(&trans->new_bgs)) btrfs_create_pending_block_groups(trans, root); - while (count < 2) { + while (count < 1) { unsigned long cur = trans->delayed_ref_updates; trans->delayed_ref_updates = 0; if (cur && @@ -589,6 +592,7 @@ static int __btrfs_end_transaction(struct btrfs_trans_handle *trans, } count++; } + btrfs_trans_release_metadata(trans, root); trans->block_rsv = NULL; -- cgit v1.2.3 From 843fcf35733164076a77ad833c72c32da8228ad0 Mon Sep 17 00:00:00 2001 From: Miao Xie Date: Mon, 28 Jan 2013 12:36:22 +0000 Subject: Btrfs: fix missing release of the space/qgroup reservation in start_transaction() When we fail to start a transaction, we need to release the reserved free space and qgroup space, fix it. Signed-off-by: Miao Xie Reviewed-by: Jan Schmidt Signed-off-by: Josef Bacik --- fs/btrfs/transaction.c | 27 +++++++++++++++++++-------- 1 file changed, 19 insertions(+), 8 deletions(-) (limited to 'fs/btrfs/transaction.c') diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index f15494699f3b..fc03aa60b684 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -333,12 +333,14 @@ start_transaction(struct btrfs_root *root, u64 num_items, int type, &root->fs_info->trans_block_rsv, num_bytes, flush); if (ret) - return ERR_PTR(ret); + goto reserve_fail; } again: h = kmem_cache_alloc(btrfs_trans_handle_cachep, GFP_NOFS); - if (!h) - return ERR_PTR(-ENOMEM); + if (!h) { + ret = -ENOMEM; + goto alloc_fail; + } /* * If we are JOIN_NOLOCK we're already committing a transaction and @@ -365,11 +367,7 @@ again: if (ret < 0) { /* We must get the transaction if we are JOIN_NOLOCK. */ BUG_ON(type == TRANS_JOIN_NOLOCK); - - if (type < TRANS_JOIN_NOLOCK) - sb_end_intwrite(root->fs_info->sb); - kmem_cache_free(btrfs_trans_handle_cachep, h); - return ERR_PTR(ret); + goto join_fail; } cur_trans = root->fs_info->running_transaction; @@ -410,6 +408,19 @@ got_it: if (!current->journal_info && type != TRANS_USERSPACE) current->journal_info = h; return h; + +join_fail: + if (type < TRANS_JOIN_NOLOCK) + sb_end_intwrite(root->fs_info->sb); + kmem_cache_free(btrfs_trans_handle_cachep, h); +alloc_fail: + if (num_bytes) + btrfs_block_rsv_release(root, &root->fs_info->trans_block_rsv, + num_bytes); +reserve_fail: + if (qgroup_reserved) + btrfs_qgroup_free(root, qgroup_reserved); + return ERR_PTR(ret); } struct btrfs_trans_handle *btrfs_start_transaction(struct btrfs_root *root, -- cgit v1.2.3 From 7b5a1c5310a50abc96c9ca07039688027d0a4282 Mon Sep 17 00:00:00 2001 From: Miao Xie Date: Thu, 15 Nov 2012 08:14:11 +0000 Subject: Btrfs: cleanup unnecessary clear when freeing a transaction or a trans handle We clear the transaction object and the trans handle when they are about to be freed, it is unnecessary, cleanup it. Signed-off-by: Miao Xie --- fs/btrfs/transaction.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'fs/btrfs/transaction.c') diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index fc03aa60b684..a1455f1e4676 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -40,7 +40,6 @@ void put_transaction(struct btrfs_transaction *transaction) if (atomic_dec_and_test(&transaction->use_count)) { BUG_ON(!list_empty(&transaction->list)); WARN_ON(transaction->delayed_refs.root.rb_node); - memset(transaction, 0, sizeof(*transaction)); kmem_cache_free(btrfs_transaction_cachep, transaction); } } @@ -650,7 +649,6 @@ static int __btrfs_end_transaction(struct btrfs_trans_handle *trans, } assert_qgroups_uptodate(trans); - memset(trans, 0, sizeof(*trans)); kmem_cache_free(btrfs_trans_handle_cachep, trans); return err; } -- cgit v1.2.3 From 7892b5afe4a1a00af25107e27357db30434ab876 Mon Sep 17 00:00:00 2001 From: Miao Xie Date: Thu, 15 Nov 2012 08:14:47 +0000 Subject: Btrfs: use common work instead of delayed work Since we do not want to delay the async transaction commit, we should use common work, not delayed work. Signed-off-by: Miao Xie --- fs/btrfs/transaction.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'fs/btrfs/transaction.c') diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index a1455f1e4676..86bb105b3982 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -1305,13 +1305,13 @@ static void wait_current_trans_commit_start_and_unblock(struct btrfs_root *root, struct btrfs_async_commit { struct btrfs_trans_handle *newtrans; struct btrfs_root *root; - struct delayed_work work; + struct work_struct work; }; static void do_async_commit(struct work_struct *work) { struct btrfs_async_commit *ac = - container_of(work, struct btrfs_async_commit, work.work); + container_of(work, struct btrfs_async_commit, work); /* * We've got freeze protection passed with the transaction. @@ -1339,7 +1339,7 @@ int btrfs_commit_transaction_async(struct btrfs_trans_handle *trans, if (!ac) return -ENOMEM; - INIT_DELAYED_WORK(&ac->work, do_async_commit); + INIT_WORK(&ac->work, do_async_commit); ac->root = root; ac->newtrans = btrfs_join_transaction(root); if (IS_ERR(ac->newtrans)) { @@ -1363,7 +1363,7 @@ int btrfs_commit_transaction_async(struct btrfs_trans_handle *trans, &root->fs_info->sb->s_writers.lock_map[SB_FREEZE_FS-1], 1, _THIS_IP_); - schedule_delayed_work(&ac->work, 0); + schedule_work(&ac->work); /* wait for transaction to start and unblock */ if (wait_for_unblock) -- cgit v1.2.3 From c6b305a89b1903d63652691ad5eb9f05aa0326b8 Mon Sep 17 00:00:00 2001 From: Josef Bacik Date: Tue, 18 Dec 2012 09:16:16 -0500 Subject: Btrfs: don't re-enter when allocating a chunk If we start running low on metadata space we will try to allocate a chunk, which could then try to allocate a chunk to add the device entry. The thing is we allocate a chunk before we try really hard to make the allocation, so we should be able to find space for the device entry. Add a flag to the trans handle so we know we're currently allocating a chunk so we can just bail out if we try to allocate another chunk. Thanks, Signed-off-by: Josef Bacik --- fs/btrfs/extent-tree.c | 7 +++++++ fs/btrfs/transaction.c | 1 + fs/btrfs/transaction.h | 1 + 3 files changed, 9 insertions(+) (limited to 'fs/btrfs/transaction.c') diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index d5e60d25ca51..c642fc2cfc2b 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -3570,6 +3570,10 @@ static int do_chunk_alloc(struct btrfs_trans_handle *trans, int wait_for_alloc = 0; int ret = 0; + /* Don't re-enter if we're already allocating a chunk */ + if (trans->allocating_chunk) + return -ENOSPC; + space_info = __find_space_info(extent_root->fs_info, flags); if (!space_info) { ret = update_space_info(extent_root->fs_info, flags, @@ -3612,6 +3616,8 @@ again: goto again; } + trans->allocating_chunk = true; + /* * If we have mixed data/metadata chunks we want to make sure we keep * allocating mixed chunks instead of individual chunks. @@ -3638,6 +3644,7 @@ again: check_system_chunk(trans, extent_root, flags); ret = btrfs_alloc_chunk(trans, extent_root, flags); + trans->allocating_chunk = false; if (ret < 0 && ret != -ENOSPC) goto out; diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 86bb105b3982..24fde97cba81 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -385,6 +385,7 @@ again: h->qgroup_reserved = qgroup_reserved; h->delayed_ref_elem.seq = 0; h->type = type; + h->allocating_chunk = false; INIT_LIST_HEAD(&h->qgroup_ref_list); INIT_LIST_HEAD(&h->new_bgs); diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h index 0e8aa1e6c287..69700f7b20ac 100644 --- a/fs/btrfs/transaction.h +++ b/fs/btrfs/transaction.h @@ -68,6 +68,7 @@ struct btrfs_trans_handle { struct btrfs_block_rsv *orig_rsv; short aborted; short adding_csums; + bool allocating_chunk; enum btrfs_trans_type type; /* * this root is only needed to validate that the root passed to -- cgit v1.2.3 From 3edb2a68cb23cd6ca84022421eeae2604722cdc4 Mon Sep 17 00:00:00 2001 From: Miao Xie Date: Tue, 22 Jan 2013 10:49:33 +0000 Subject: Btrfs: check the return value of btrfs_start_delalloc_inodes() We forget to check the return value of btrfs_start_delalloc_inodes(), fix it. Signed-off-by: Miao Xie Signed-off-by: Josef Bacik --- fs/btrfs/dev-replace.c | 6 +++++- fs/btrfs/transaction.c | 4 +++- 2 files changed, 8 insertions(+), 2 deletions(-) (limited to 'fs/btrfs/transaction.c') diff --git a/fs/btrfs/dev-replace.c b/fs/btrfs/dev-replace.c index 66dbc8dbddf7..7ba7b3900cb8 100644 --- a/fs/btrfs/dev-replace.c +++ b/fs/btrfs/dev-replace.c @@ -465,7 +465,11 @@ static int btrfs_dev_replace_finishing(struct btrfs_fs_info *fs_info, * flush all outstanding I/O and inode extent mappings before the * copy operation is declared as being finished */ - btrfs_start_delalloc_inodes(root, 0); + ret = btrfs_start_delalloc_inodes(root, 0); + if (ret) { + mutex_unlock(&dev_replace->lock_finishing_cancel_unmount); + return ret; + } btrfs_wait_ordered_extents(root, 0); trans = btrfs_start_transaction(root, 0); diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 24fde97cba81..42dac27207ab 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -1427,7 +1427,9 @@ static int btrfs_flush_all_pending_stuffs(struct btrfs_trans_handle *trans, } if (flush_on_commit || snap_pending) { - btrfs_start_delalloc_inodes(root, 1); + ret = btrfs_start_delalloc_inodes(root, 1); + if (ret) + return ret; btrfs_wait_ordered_extents(root, 1); } -- cgit v1.2.3 From eebc60840636e7351371fc17bcd057384bf0c16a Mon Sep 17 00:00:00 2001 From: Miao Xie Date: Tue, 22 Jan 2013 10:50:01 +0000 Subject: Btrfs: check the return value of btrfs_run_ordered_operations() We forget to check the return value of btrfs_run_ordered_operations() when flushing all the pending stuffs, fix it. Signed-off-by: Miao Xie Signed-off-by: Josef Bacik --- fs/btrfs/transaction.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'fs/btrfs/transaction.c') diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 42dac27207ab..34610dc6d140 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -1451,9 +1451,9 @@ static int btrfs_flush_all_pending_stuffs(struct btrfs_trans_handle *trans, * it here and no for sure that nothing new will be added * to the list */ - btrfs_run_ordered_operations(root, 1); + ret = btrfs_run_ordered_operations(root, 1); - return 0; + return ret; } /* -- cgit v1.2.3 From 87533c475187c1420794a2e164bc67a7974f1327 Mon Sep 17 00:00:00 2001 From: Miao Xie Date: Tue, 29 Jan 2013 10:14:48 +0000 Subject: Btrfs: use bit operation for ->fs_state There is no lock to protect fs_info->fs_state, it will introduce some problems, such as the value may be covered by the other task when several tasks modify it. For example: Task0 - CPU0 Task1 - CPU1 mov %fs_state rax or $0x1 rax mov %fs_state rax or $0x2 rax mov rax %fs_state mov rax %fs_state The expected value is 3, but in fact, it is 2. Though this problem doesn't happen now (because there is only one flag currently), the code is error prone, if we add other flags, the above problem will happen to a certainty. Now we use bit operation for it to fix the above problem. In this way, we can make the code more robust and be easy to add new flags. Signed-off-by: Miao Xie Signed-off-by: Josef Bacik --- fs/btrfs/ctree.h | 4 +++- fs/btrfs/disk-io.c | 5 +++-- fs/btrfs/file.c | 2 +- fs/btrfs/scrub.c | 2 +- fs/btrfs/super.c | 4 ++-- fs/btrfs/transaction.c | 9 ++++----- 6 files changed, 14 insertions(+), 12 deletions(-) (limited to 'fs/btrfs/transaction.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 5f5c30aaef36..78fba44ad3d9 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -338,7 +338,9 @@ static inline unsigned long btrfs_chunk_item_size(int num_stripes) /* * File system states */ +#define BTRFS_FS_STATE_ERROR 0 +/* Super block flags */ /* Errors detected */ #define BTRFS_SUPER_FLAG_ERROR (1ULL << 2) @@ -1549,7 +1551,7 @@ struct btrfs_fs_info { u64 qgroup_seq; /* filesystem state */ - u64 fs_state; + unsigned long fs_state; struct btrfs_delayed_root *delayed_root; diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 00b6742fdde7..8e58a1f90546 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -2201,7 +2201,8 @@ int open_ctree(struct super_block *sb, goto fail_alloc; /* check FS state, whether FS is broken. */ - fs_info->fs_state |= btrfs_super_flags(disk_super); + if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) + set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY); if (ret) { @@ -3359,7 +3360,7 @@ int close_ctree(struct btrfs_root *root) printk(KERN_ERR "btrfs: commit super ret %d\n", ret); } - if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) + if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) btrfs_error_commit_super(root); btrfs_put_block_group_cache(fs_info); diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 13c78ea3ebce..75d0fe134be3 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -1545,7 +1545,7 @@ static ssize_t btrfs_file_aio_write(struct kiocb *iocb, * although we have opened a file as writable, we have * to stop this write operation to ensure FS consistency. */ - if (root->fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) { + if (test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state)) { mutex_unlock(&inode->i_mutex); err = -EROFS; goto out; diff --git a/fs/btrfs/scrub.c b/fs/btrfs/scrub.c index 67783e03d121..c78b2a3fc335 100644 --- a/fs/btrfs/scrub.c +++ b/fs/btrfs/scrub.c @@ -2708,7 +2708,7 @@ static noinline_for_stack int scrub_supers(struct scrub_ctx *sctx, int ret; struct btrfs_root *root = sctx->dev_root; - if (root->fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) + if (test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state)) return -EIO; gen = root->fs_info->last_trans_committed; diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index eda330df45a4..4152f9ea34f5 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -98,7 +98,7 @@ static void __save_error_info(struct btrfs_fs_info *fs_info) * today we only save the error info into ram. Long term we'll * also send it down to the disk */ - fs_info->fs_state = BTRFS_SUPER_FLAG_ERROR; + set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); } static void save_error_info(struct btrfs_fs_info *fs_info) @@ -114,7 +114,7 @@ static void btrfs_handle_error(struct btrfs_fs_info *fs_info) if (sb->s_flags & MS_RDONLY) return; - if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) { + if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { sb->s_flags |= MS_RDONLY; printk(KERN_INFO "btrfs is forced readonly\n"); /* diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 34610dc6d140..baf6d74fd0f2 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -61,7 +61,7 @@ static noinline int join_transaction(struct btrfs_root *root, int type) spin_lock(&fs_info->trans_lock); loop: /* The file system has been taken offline. No new transactions. */ - if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) { + if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { spin_unlock(&fs_info->trans_lock); return -EROFS; } @@ -113,7 +113,7 @@ loop: kmem_cache_free(btrfs_transaction_cachep, cur_trans); cur_trans = fs_info->running_transaction; goto loop; - } else if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) { + } else if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { spin_unlock(&fs_info->trans_lock); kmem_cache_free(btrfs_transaction_cachep, cur_trans); return -EROFS; @@ -301,7 +301,7 @@ start_transaction(struct btrfs_root *root, u64 num_items, int type, int ret; u64 qgroup_reserved = 0; - if (root->fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) + if (test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state)) return ERR_PTR(-EROFS); if (current->journal_info) { @@ -645,9 +645,8 @@ static int __btrfs_end_transaction(struct btrfs_trans_handle *trans, btrfs_run_delayed_iputs(root); if (trans->aborted || - root->fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) { + test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state)) err = -EIO; - } assert_qgroups_uptodate(trans); kmem_cache_free(btrfs_trans_handle_cachep, trans); -- cgit v1.2.3 From e4a2bcaca9643e7430207810653222fc5187f2be Mon Sep 17 00:00:00 2001 From: Josef Bacik Date: Wed, 6 Feb 2013 16:55:41 -0500 Subject: Btrfs: if we aren't committing just end the transaction if we error out I hit a deadlock where transaction commit was waiting on num_writers to be 0. This happened because somebody came into btrfs_commit_transaction and noticed we had aborted and it went to cleanup_transaction. This shouldn't happen because cleanup_transaction is really to fixup a bad commit, it doesn't do the normal trans handle cleanup things. So if we have an error just do the normal btrfs_end_transaction dance and return. Once we are in the actual commit path we can use cleanup_transaction and be good to go. Thanks, Signed-off-by: Josef Bacik --- fs/btrfs/transaction.c | 18 ++++++++++++------ 1 file changed, 12 insertions(+), 6 deletions(-) (limited to 'fs/btrfs/transaction.c') diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index baf6d74fd0f2..5144ad19ef47 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -1476,21 +1476,25 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, ret = btrfs_run_ordered_operations(root, 0); if (ret) { btrfs_abort_transaction(trans, root, ret); - goto cleanup_transaction; + btrfs_end_transaction(trans, root); + return ret; } /* Stop the commit early if ->aborted is set */ if (unlikely(ACCESS_ONCE(cur_trans->aborted))) { ret = cur_trans->aborted; - goto cleanup_transaction; + btrfs_end_transaction(trans, root); + return ret; } /* make a pass through all the delayed refs we have so far * any runnings procs may add more while we are here */ ret = btrfs_run_delayed_refs(trans, root, 0); - if (ret) - goto cleanup_transaction; + if (ret) { + btrfs_end_transaction(trans, root); + return ret; + } btrfs_trans_release_metadata(trans, root); trans->block_rsv = NULL; @@ -1507,8 +1511,10 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, btrfs_create_pending_block_groups(trans, root); ret = btrfs_run_delayed_refs(trans, root, 0); - if (ret) - goto cleanup_transaction; + if (ret) { + btrfs_end_transaction(trans, root); + return ret; + } spin_lock(&cur_trans->commit_lock); if (cur_trans->in_commit) { -- cgit v1.2.3 From de78b51a2852bddccd6535e9e12de65f92787a1e Mon Sep 17 00:00:00 2001 From: Eric Sandeen Date: Thu, 31 Jan 2013 18:21:12 +0000 Subject: btrfs: remove cache only arguments from defrag path The entry point at the defrag ioctl always sets "cache only" to 0; the codepaths haven't run for a long time as far as I can tell. Chris says they're dead code, so remove them. Signed-off-by: Eric Sandeen Signed-off-by: Josef Bacik --- fs/btrfs/ctree.c | 65 ++++++++------------------------------------------ fs/btrfs/ctree.h | 8 +++---- fs/btrfs/ioctl.c | 8 +++---- fs/btrfs/transaction.c | 8 +++---- fs/btrfs/transaction.h | 2 +- fs/btrfs/tree-defrag.c | 19 ++++++--------- fs/btrfs/tree-log.c | 4 ++-- 7 files changed, 32 insertions(+), 82 deletions(-) (limited to 'fs/btrfs/transaction.c') diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index 35444013f0cf..ecd25a1b4e51 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -1442,7 +1442,7 @@ int btrfs_comp_cpu_keys(struct btrfs_key *k1, struct btrfs_key *k2) */ int btrfs_realloc_node(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct extent_buffer *parent, - int start_slot, int cache_only, u64 *last_ret, + int start_slot, u64 *last_ret, struct btrfs_key *progress) { struct extent_buffer *cur; @@ -1462,8 +1462,6 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans, struct btrfs_disk_key disk_key; parent_level = btrfs_header_level(parent); - if (cache_only && parent_level != 1) - return 0; WARN_ON(trans->transaction != root->fs_info->running_transaction); WARN_ON(trans->transid != root->fs_info->generation); @@ -1509,10 +1507,6 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans, else uptodate = 0; if (!cur || !uptodate) { - if (cache_only) { - free_extent_buffer(cur); - continue; - } if (!cur) { cur = read_tree_block(root, blocknr, blocksize, gen); @@ -4826,8 +4820,8 @@ int btrfs_prev_leaf(struct btrfs_root *root, struct btrfs_path *path) /* * A helper function to walk down the tree starting at min_key, and looking - * for nodes or leaves that are either in cache or have a minimum - * transaction id. This is used by the btree defrag code, and tree logging + * for nodes or leaves that are have a minimum transaction id. + * This is used by the btree defrag code, and tree logging * * This does not cow, but it does stuff the starting key it finds back * into min_key, so you can call btrfs_search_slot with cow=1 on the @@ -4848,7 +4842,7 @@ int btrfs_prev_leaf(struct btrfs_root *root, struct btrfs_path *path) */ int btrfs_search_forward(struct btrfs_root *root, struct btrfs_key *min_key, struct btrfs_key *max_key, - struct btrfs_path *path, int cache_only, + struct btrfs_path *path, u64 min_trans) { struct extent_buffer *cur; @@ -4888,15 +4882,12 @@ again: if (sret && slot > 0) slot--; /* - * check this node pointer against the cache_only and - * min_trans parameters. If it isn't in cache or is too - * old, skip to the next one. + * check this node pointer against the min_trans parameters. + * If it is too old, old, skip to the next one. */ while (slot < nritems) { u64 blockptr; u64 gen; - struct extent_buffer *tmp; - struct btrfs_disk_key disk_key; blockptr = btrfs_node_blockptr(cur, slot); gen = btrfs_node_ptr_generation(cur, slot); @@ -4904,27 +4895,7 @@ again: slot++; continue; } - if (!cache_only) - break; - - if (max_key) { - btrfs_node_key(cur, &disk_key, slot); - if (comp_keys(&disk_key, max_key) >= 0) { - ret = 1; - goto out; - } - } - - tmp = btrfs_find_tree_block(root, blockptr, - btrfs_level_size(root, level - 1)); - - if (tmp && btrfs_buffer_uptodate(tmp, gen, 1) > 0) { - free_extent_buffer(tmp); - break; - } - if (tmp) - free_extent_buffer(tmp); - slot++; + break; } find_next_key: /* @@ -4935,7 +4906,7 @@ find_next_key: path->slots[level] = slot; btrfs_set_path_blocking(path); sret = btrfs_find_next_key(root, path, min_key, level, - cache_only, min_trans); + min_trans); if (sret == 0) { btrfs_release_path(path); goto again; @@ -5400,8 +5371,7 @@ out: /* * this is similar to btrfs_next_leaf, but does not try to preserve * and fixup the path. It looks for and returns the next key in the - * tree based on the current path and the cache_only and min_trans - * parameters. + * tree based on the current path and the min_trans parameters. * * 0 is returned if another key is found, < 0 if there are any errors * and 1 is returned if there are no higher keys in the tree @@ -5410,8 +5380,7 @@ out: * calling this function. */ int btrfs_find_next_key(struct btrfs_root *root, struct btrfs_path *path, - struct btrfs_key *key, int level, - int cache_only, u64 min_trans) + struct btrfs_key *key, int level, u64 min_trans) { int slot; struct extent_buffer *c; @@ -5462,22 +5431,8 @@ next: if (level == 0) btrfs_item_key_to_cpu(c, key, slot); else { - u64 blockptr = btrfs_node_blockptr(c, slot); u64 gen = btrfs_node_ptr_generation(c, slot); - if (cache_only) { - struct extent_buffer *cur; - cur = btrfs_find_tree_block(root, blockptr, - btrfs_level_size(root, level - 1)); - if (!cur || - btrfs_buffer_uptodate(cur, gen, 1) <= 0) { - slot++; - if (cur) - free_extent_buffer(cur); - goto next; - } - free_extent_buffer(cur); - } if (gen < min_trans) { slot++; goto next; diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 1e3a994e1899..7e2cffd2a5d8 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -3125,10 +3125,10 @@ struct extent_buffer *btrfs_root_node(struct btrfs_root *root); struct extent_buffer *btrfs_lock_root_node(struct btrfs_root *root); int btrfs_find_next_key(struct btrfs_root *root, struct btrfs_path *path, struct btrfs_key *key, int lowest_level, - int cache_only, u64 min_trans); + u64 min_trans); int btrfs_search_forward(struct btrfs_root *root, struct btrfs_key *min_key, struct btrfs_key *max_key, - struct btrfs_path *path, int cache_only, + struct btrfs_path *path, u64 min_trans); enum btrfs_compare_tree_result { BTRFS_COMPARE_TREE_NEW, @@ -3181,7 +3181,7 @@ int btrfs_search_slot_for_read(struct btrfs_root *root, int find_higher, int return_any); int btrfs_realloc_node(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct extent_buffer *parent, - int start_slot, int cache_only, u64 *last_ret, + int start_slot, u64 *last_ret, struct btrfs_key *progress); void btrfs_release_path(struct btrfs_path *p); struct btrfs_path *btrfs_alloc_path(void); @@ -3576,7 +3576,7 @@ int btrfs_dirty_pages(struct btrfs_root *root, struct inode *inode, /* tree-defrag.c */ int btrfs_defrag_leaves(struct btrfs_trans_handle *trans, - struct btrfs_root *root, int cache_only); + struct btrfs_root *root); /* sysfs.c */ int btrfs_init_sysfs(void); diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index 96ecefc1724f..daea831f3d36 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -818,7 +818,7 @@ static int find_new_extents(struct btrfs_root *root, while(1) { ret = btrfs_search_forward(root, &min_key, &max_key, - path, 0, newer_than); + path, newer_than); if (ret != 0) goto none; if (min_key.objectid != ino) @@ -1864,7 +1864,7 @@ static noinline int search_ioctl(struct inode *inode, path->keep_locks = 1; while(1) { - ret = btrfs_search_forward(root, &key, &max_key, path, 0, + ret = btrfs_search_forward(root, &key, &max_key, path, sk->min_transid); if (ret != 0) { if (ret > 0) @@ -2212,10 +2212,10 @@ static int btrfs_ioctl_defrag(struct file *file, void __user *argp) ret = -EPERM; goto out; } - ret = btrfs_defrag_root(root, 0); + ret = btrfs_defrag_root(root); if (ret) goto out; - ret = btrfs_defrag_root(root->fs_info->extent_root, 0); + ret = btrfs_defrag_root(root->fs_info->extent_root); break; case S_IFREG: if (!(file->f_mode & FMODE_WRITE)) { diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 5144ad19ef47..60481a53e004 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -959,10 +959,10 @@ static noinline int commit_fs_roots(struct btrfs_trans_handle *trans, } /* - * defrag a given btree. If cacheonly == 1, this won't read from the disk, - * otherwise every leaf in the btree is read and defragged. + * defrag a given btree. + * Every leaf in the btree is read and defragged. */ -int btrfs_defrag_root(struct btrfs_root *root, int cacheonly) +int btrfs_defrag_root(struct btrfs_root *root) { struct btrfs_fs_info *info = root->fs_info; struct btrfs_trans_handle *trans; @@ -976,7 +976,7 @@ int btrfs_defrag_root(struct btrfs_root *root, int cacheonly) if (IS_ERR(trans)) return PTR_ERR(trans); - ret = btrfs_defrag_leaves(trans, root, cacheonly); + ret = btrfs_defrag_leaves(trans, root); btrfs_end_transaction(trans, root); btrfs_btree_balance_dirty(info->tree_root); diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h index 69700f7b20ac..46628210e5d8 100644 --- a/fs/btrfs/transaction.h +++ b/fs/btrfs/transaction.h @@ -117,7 +117,7 @@ int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans, struct btrfs_root *root); int btrfs_add_dead_root(struct btrfs_root *root); -int btrfs_defrag_root(struct btrfs_root *root, int cacheonly); +int btrfs_defrag_root(struct btrfs_root *root); int btrfs_clean_old_snapshots(struct btrfs_root *root); int btrfs_commit_transaction(struct btrfs_trans_handle *trans, struct btrfs_root *root); diff --git a/fs/btrfs/tree-defrag.c b/fs/btrfs/tree-defrag.c index 3b580ee8ab1d..94e05c1f118a 100644 --- a/fs/btrfs/tree-defrag.c +++ b/fs/btrfs/tree-defrag.c @@ -23,13 +23,14 @@ #include "transaction.h" #include "locking.h" -/* defrag all the leaves in a given btree. If cache_only == 1, don't read - * things from disk, otherwise read all the leaves and try to get key order to +/* + * Defrag all the leaves in a given btree. + * Read all the leaves and try to get key order to * better reflect disk order */ int btrfs_defrag_leaves(struct btrfs_trans_handle *trans, - struct btrfs_root *root, int cache_only) + struct btrfs_root *root) { struct btrfs_path *path = NULL; struct btrfs_key key; @@ -41,9 +42,6 @@ int btrfs_defrag_leaves(struct btrfs_trans_handle *trans, u64 last_ret = 0; u64 min_trans = 0; - if (cache_only) - goto out; - if (root->fs_info->extent_root == root) { /* * there's recursion here right now in the tree locking, @@ -86,11 +84,8 @@ int btrfs_defrag_leaves(struct btrfs_trans_handle *trans, } path->keep_locks = 1; - if (cache_only) - min_trans = root->defrag_trans_start; - ret = btrfs_search_forward(root, &key, NULL, path, - cache_only, min_trans); + ret = btrfs_search_forward(root, &key, NULL, path, min_trans); if (ret < 0) goto out; if (ret > 0) { @@ -109,11 +104,11 @@ int btrfs_defrag_leaves(struct btrfs_trans_handle *trans, goto out; } path->slots[1] = btrfs_header_nritems(path->nodes[1]); - next_key_ret = btrfs_find_next_key(root, path, &key, 1, cache_only, + next_key_ret = btrfs_find_next_key(root, path, &key, 1, min_trans); ret = btrfs_realloc_node(trans, root, path->nodes[1], 0, - cache_only, &last_ret, + &last_ret, &root->defrag_progress); if (ret) { WARN_ON(ret == -EAGAIN); diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c index deb7f66356bd..1a79087c4575 100644 --- a/fs/btrfs/tree-log.c +++ b/fs/btrfs/tree-log.c @@ -2738,7 +2738,7 @@ static noinline int log_dir_items(struct btrfs_trans_handle *trans, path->keep_locks = 1; ret = btrfs_search_forward(root, &min_key, &max_key, - path, 0, trans->transid); + path, trans->transid); /* * we didn't find anything from this transaction, see if there @@ -3680,7 +3680,7 @@ static int btrfs_log_inode(struct btrfs_trans_handle *trans, while (1) { ins_nr = 0; ret = btrfs_search_forward(root, &min_key, &max_key, - path, 0, trans->transid); + path, trans->transid); if (ret != 0) break; again: -- cgit v1.2.3 From 210549ebe9047ae5a8cc47487203d3ee16a7749b Mon Sep 17 00:00:00 2001 From: David Sterba Date: Sat, 9 Feb 2013 23:38:06 +0000 Subject: btrfs: add cancellation points to defrag The defrag operation can take very long, we want to have a way how to cancel it. The code checks for a pending signal at safe points in the defrag loops and returns EAGAIN. This means a user can press ^C after running 'btrfs fi defrag', woks for both defrag modes, files and root. Returning from the command was instant in my light tests, but may take longer depending on the aging factor of the filesystem. Signed-off-by: David Sterba Signed-off-by: Josef Bacik --- fs/btrfs/ctree.h | 7 +++++++ fs/btrfs/ioctl.c | 6 ++++++ fs/btrfs/transaction.c | 6 ++++++ 3 files changed, 19 insertions(+) (limited to 'fs/btrfs/transaction.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index f1cc247f3178..14f01dc70ea6 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -3786,4 +3786,11 @@ static inline int is_fstree(u64 rootid) return 1; return 0; } + +static inline int btrfs_defrag_cancelled(struct btrfs_fs_info *fs_info) +{ + return signal_pending(current); +} + + #endif diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index 6d6314406e27..d02ec577f70f 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -1202,6 +1202,12 @@ int btrfs_defrag_file(struct inode *inode, struct file *file, if (!(inode->i_sb->s_flags & MS_ACTIVE)) break; + if (btrfs_defrag_cancelled(root->fs_info)) { + printk(KERN_DEBUG "btrfs: defrag_file cancelled\n"); + ret = -EAGAIN; + break; + } + if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT, extent_thresh, &last_len, &skip, &defrag_end, range->flags & diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 60481a53e004..d574d830a1c4 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -984,6 +984,12 @@ int btrfs_defrag_root(struct btrfs_root *root) if (btrfs_fs_closing(root->fs_info) || ret != -EAGAIN) break; + + if (btrfs_defrag_cancelled(root->fs_info)) { + printk(KERN_DEBUG "btrfs: defrag_root cancelled\n"); + ret = -EAGAIN; + break; + } } root->defrag_running = 0; return ret; -- cgit v1.2.3 From 569e0f358c0c37f6733702d4a5d2c412860f7169 Mon Sep 17 00:00:00 2001 From: Josef Bacik Date: Wed, 13 Feb 2013 11:09:14 -0500 Subject: Btrfs: place ordered operations on a per transaction list Miao made the ordered operations stuff run async, which introduced a deadlock where we could get somebody (sync) racing in and committing the transaction while a commit was already happening. The new committer would try and flush ordered operations which would hang waiting for the commit to finish because it is done asynchronously and no longer inherits the callers trans handle. To fix this we need to make the ordered operations list a per transaction list. We can get new inodes added to the ordered operation list by truncating them and then having another process writing to them, so this makes it so that anybody trying to add an ordered operation _must_ start a transaction in order to add itself to the list, which will keep new inodes from getting added to the ordered operations list after we start committing. This should fix the deadlock and also keeps us from doing a lot more work than we need to during commit. Thanks, Signed-off-by: Josef Bacik --- fs/btrfs/ctree.h | 7 ------- fs/btrfs/disk-io.c | 11 ++++++----- fs/btrfs/file.c | 15 ++++++++++++++- fs/btrfs/ordered-data.c | 13 ++++++++----- fs/btrfs/ordered-data.h | 3 ++- fs/btrfs/transaction.c | 5 +++-- fs/btrfs/transaction.h | 1 + 7 files changed, 34 insertions(+), 21 deletions(-) (limited to 'fs/btrfs/transaction.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 14f01dc70ea6..961ff2986341 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1407,13 +1407,6 @@ struct btrfs_fs_info { */ struct list_head delalloc_inodes; - /* - * special rename and truncate targets that must be on disk before - * we're allowed to commit. This is basically the ext3 style - * data=ordered list. - */ - struct list_head ordered_operations; - /* * there is a pool of worker threads for checksumming during writes * and a pool for checksumming after reads. This is because readers diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 9fcce54ecde4..e511d9f78c19 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -56,7 +56,8 @@ static void end_workqueue_fn(struct btrfs_work *work); static void free_fs_root(struct btrfs_root *root); static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info, int read_only); -static void btrfs_destroy_ordered_operations(struct btrfs_root *root); +static void btrfs_destroy_ordered_operations(struct btrfs_transaction *t, + struct btrfs_root *root); static void btrfs_destroy_ordered_extents(struct btrfs_root *root); static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, struct btrfs_root *root); @@ -2029,7 +2030,6 @@ int open_ctree(struct super_block *sb, INIT_LIST_HEAD(&fs_info->dead_roots); INIT_LIST_HEAD(&fs_info->delayed_iputs); INIT_LIST_HEAD(&fs_info->delalloc_inodes); - INIT_LIST_HEAD(&fs_info->ordered_operations); INIT_LIST_HEAD(&fs_info->caching_block_groups); spin_lock_init(&fs_info->delalloc_lock); spin_lock_init(&fs_info->trans_lock); @@ -3538,7 +3538,8 @@ void btrfs_error_commit_super(struct btrfs_root *root) btrfs_cleanup_transaction(root); } -static void btrfs_destroy_ordered_operations(struct btrfs_root *root) +static void btrfs_destroy_ordered_operations(struct btrfs_transaction *t, + struct btrfs_root *root) { struct btrfs_inode *btrfs_inode; struct list_head splice; @@ -3548,7 +3549,7 @@ static void btrfs_destroy_ordered_operations(struct btrfs_root *root) mutex_lock(&root->fs_info->ordered_operations_mutex); spin_lock(&root->fs_info->ordered_extent_lock); - list_splice_init(&root->fs_info->ordered_operations, &splice); + list_splice_init(&t->ordered_operations, &splice); while (!list_empty(&splice)) { btrfs_inode = list_entry(splice.next, struct btrfs_inode, ordered_operations); @@ -3829,7 +3830,7 @@ int btrfs_cleanup_transaction(struct btrfs_root *root) while (!list_empty(&list)) { t = list_entry(list.next, struct btrfs_transaction, list); - btrfs_destroy_ordered_operations(root); + btrfs_destroy_ordered_operations(t, root); btrfs_destroy_ordered_extents(root); diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 75d0fe134be3..b12ba52c4505 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -1628,7 +1628,20 @@ int btrfs_release_file(struct inode *inode, struct file *filp) */ if (test_and_clear_bit(BTRFS_INODE_ORDERED_DATA_CLOSE, &BTRFS_I(inode)->runtime_flags)) { - btrfs_add_ordered_operation(NULL, BTRFS_I(inode)->root, inode); + struct btrfs_trans_handle *trans; + struct btrfs_root *root = BTRFS_I(inode)->root; + + /* + * We need to block on a committing transaction to keep us from + * throwing a ordered operation on to the list and causing + * something like sync to deadlock trying to flush out this + * inode. + */ + trans = btrfs_start_transaction(root, 0); + if (IS_ERR(trans)) + return PTR_ERR(trans); + btrfs_add_ordered_operation(trans, BTRFS_I(inode)->root, inode); + btrfs_end_transaction(trans, root); if (inode->i_size > BTRFS_ORDERED_OPERATIONS_FLUSH_LIMIT) filemap_flush(inode->i_mapping); } diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c index 9489fa96e3ed..dc08d77b717e 100644 --- a/fs/btrfs/ordered-data.c +++ b/fs/btrfs/ordered-data.c @@ -612,10 +612,12 @@ void btrfs_wait_ordered_extents(struct btrfs_root *root, int delay_iput) * extra check to make sure the ordered operation list really is empty * before we return */ -int btrfs_run_ordered_operations(struct btrfs_root *root, int wait) +int btrfs_run_ordered_operations(struct btrfs_trans_handle *trans, + struct btrfs_root *root, int wait) { struct btrfs_inode *btrfs_inode; struct inode *inode; + struct btrfs_transaction *cur_trans = trans->transaction; struct list_head splice; struct list_head works; struct btrfs_delalloc_work *work, *next; @@ -626,7 +628,7 @@ int btrfs_run_ordered_operations(struct btrfs_root *root, int wait) mutex_lock(&root->fs_info->ordered_operations_mutex); spin_lock(&root->fs_info->ordered_extent_lock); - list_splice_init(&root->fs_info->ordered_operations, &splice); + list_splice_init(&cur_trans->ordered_operations, &splice); while (!list_empty(&splice)) { btrfs_inode = list_entry(splice.next, struct btrfs_inode, ordered_operations); @@ -643,7 +645,7 @@ int btrfs_run_ordered_operations(struct btrfs_root *root, int wait) if (!wait) list_add_tail(&BTRFS_I(inode)->ordered_operations, - &root->fs_info->ordered_operations); + &cur_trans->ordered_operations); spin_unlock(&root->fs_info->ordered_extent_lock); work = btrfs_alloc_delalloc_work(inode, wait, 1); @@ -653,7 +655,7 @@ int btrfs_run_ordered_operations(struct btrfs_root *root, int wait) list_add_tail(&btrfs_inode->ordered_operations, &splice); list_splice_tail(&splice, - &root->fs_info->ordered_operations); + &cur_trans->ordered_operations); spin_unlock(&root->fs_info->ordered_extent_lock); ret = -ENOMEM; goto out; @@ -1033,6 +1035,7 @@ out: void btrfs_add_ordered_operation(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct inode *inode) { + struct btrfs_transaction *cur_trans = trans->transaction; u64 last_mod; last_mod = max(BTRFS_I(inode)->generation, BTRFS_I(inode)->last_trans); @@ -1047,7 +1050,7 @@ void btrfs_add_ordered_operation(struct btrfs_trans_handle *trans, spin_lock(&root->fs_info->ordered_extent_lock); if (list_empty(&BTRFS_I(inode)->ordered_operations)) { list_add_tail(&BTRFS_I(inode)->ordered_operations, - &root->fs_info->ordered_operations); + &cur_trans->ordered_operations); } spin_unlock(&root->fs_info->ordered_extent_lock); } diff --git a/fs/btrfs/ordered-data.h b/fs/btrfs/ordered-data.h index d523dbd2314d..267ac99095f6 100644 --- a/fs/btrfs/ordered-data.h +++ b/fs/btrfs/ordered-data.h @@ -197,7 +197,8 @@ struct btrfs_ordered_extent *btrfs_lookup_ordered_range(struct inode *inode, int btrfs_ordered_update_i_size(struct inode *inode, u64 offset, struct btrfs_ordered_extent *ordered); int btrfs_find_ordered_sum(struct inode *inode, u64 offset, u64 disk_bytenr, u32 *sum); -int btrfs_run_ordered_operations(struct btrfs_root *root, int wait); +int btrfs_run_ordered_operations(struct btrfs_trans_handle *trans, + struct btrfs_root *root, int wait); void btrfs_add_ordered_operation(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct inode *inode); diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index d574d830a1c4..0c87d18d1881 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -157,6 +157,7 @@ loop: spin_lock_init(&cur_trans->delayed_refs.lock); INIT_LIST_HEAD(&cur_trans->pending_snapshots); + INIT_LIST_HEAD(&cur_trans->ordered_operations); list_add_tail(&cur_trans->list, &fs_info->trans_list); extent_io_tree_init(&cur_trans->dirty_pages, fs_info->btree_inode->i_mapping); @@ -1456,7 +1457,7 @@ static int btrfs_flush_all_pending_stuffs(struct btrfs_trans_handle *trans, * it here and no for sure that nothing new will be added * to the list */ - ret = btrfs_run_ordered_operations(root, 1); + ret = btrfs_run_ordered_operations(trans, root, 1); return ret; } @@ -1479,7 +1480,7 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, int should_grow = 0; unsigned long now = get_seconds(); - ret = btrfs_run_ordered_operations(root, 0); + ret = btrfs_run_ordered_operations(trans, root, 0); if (ret) { btrfs_abort_transaction(trans, root, ret); btrfs_end_transaction(trans, root); diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h index 46628210e5d8..3f772fd0191a 100644 --- a/fs/btrfs/transaction.h +++ b/fs/btrfs/transaction.h @@ -43,6 +43,7 @@ struct btrfs_transaction { wait_queue_head_t writer_wait; wait_queue_head_t commit_wait; struct list_head pending_snapshots; + struct list_head ordered_operations; struct btrfs_delayed_ref_root delayed_refs; int aborted; }; -- cgit v1.2.3 From 4b82490649f5c8ecbf888752c325ea68831c497e Mon Sep 17 00:00:00 2001 From: Miao Xie Date: Wed, 20 Feb 2013 09:13:32 +0000 Subject: Btrfs: fix the qgroup reserved space is released prematurely In start_transactio(), we will try to join the transaction again after the current transaction is committed, so we should not release the reserved space of the qgroup. Fix it. Cc: Arne Jansen Signed-off-by: Miao Xie Signed-off-by: Josef Bacik --- fs/btrfs/transaction.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'fs/btrfs/transaction.c') diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 0c87d18d1881..425d5b57d377 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -383,7 +383,7 @@ again: h->block_rsv = NULL; h->orig_rsv = NULL; h->aborted = 0; - h->qgroup_reserved = qgroup_reserved; + h->qgroup_reserved = 0; h->delayed_ref_elem.seq = 0; h->type = type; h->allocating_chunk = false; @@ -402,6 +402,7 @@ again: h->block_rsv = &root->fs_info->trans_block_rsv; h->bytes_reserved = num_bytes; } + h->qgroup_reserved = qgroup_reserved; got_it: btrfs_record_root_in_trans(h, root); -- cgit v1.2.3 From 178260b2c14969f29ba39a78df74ed485abc6203 Mon Sep 17 00:00:00 2001 From: Miao Xie Date: Wed, 20 Feb 2013 09:16:24 +0000 Subject: Btrfs: fix the deadlock between the transaction start/attach and commit Now btrfs_commit_transaction() does this ret = btrfs_run_ordered_operations(root, 0) which async flushes all inodes on the ordered operations list, it introduced a deadlock that transaction-start task, transaction-commit task and the flush workers waited for each other. (See the following URL to get the detail http://marc.info/?l=linux-btrfs&m=136070705732646&w=2) As we know, if ->in_commit is set, it means someone is committing the current transaction, we should not try to join it if we are not JOIN or JOIN_NOLOCK, wait is the best choice for it. In this way, we can avoid the above problem. In this way, there is another benefit: there is no new transaction handle to block the transaction which is on the way of commit, once we set ->in_commit. Signed-off-by: Miao Xie Signed-off-by: Josef Bacik --- fs/btrfs/transaction.c | 17 ++++++++++++++++- 1 file changed, 16 insertions(+), 1 deletion(-) (limited to 'fs/btrfs/transaction.c') diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 425d5b57d377..5767ea1c0150 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -50,6 +50,14 @@ static noinline void switch_commit_root(struct btrfs_root *root) root->commit_root = btrfs_root_node(root); } +static inline int can_join_transaction(struct btrfs_transaction *trans, + int type) +{ + return !(trans->in_commit && + type != TRANS_JOIN && + type != TRANS_JOIN_NOLOCK); +} + /* * either allocate a new transaction or hop into the existing one */ @@ -85,6 +93,10 @@ loop: spin_unlock(&fs_info->trans_lock); return cur_trans->aborted; } + if (!can_join_transaction(cur_trans, type)) { + spin_unlock(&fs_info->trans_lock); + return -EBUSY; + } atomic_inc(&cur_trans->use_count); atomic_inc(&cur_trans->num_writers); cur_trans->num_joined++; @@ -360,8 +372,11 @@ again: do { ret = join_transaction(root, type); - if (ret == -EBUSY) + if (ret == -EBUSY) { wait_current_trans(root); + if (unlikely(type == TRANS_ATTACH)) + ret = -ENOENT; + } } while (ret == -EBUSY); if (ret < 0) { -- cgit v1.2.3 From d4edf39bd5db443151efc993dac67ec9d6b5b8c1 Mon Sep 17 00:00:00 2001 From: Miao Xie Date: Wed, 20 Feb 2013 09:17:06 +0000 Subject: Btrfs: fix uncompleted transaction In some cases, we need commit the current transaction, but don't want to start a new one if there is no running transaction, so we introduce the function - btrfs_attach_transaction(), which can catch the current transaction, and return -ENOENT if there is no running transaction. But no running transaction doesn't mean the current transction completely, because we removed the running transaction before it completes. In some cases, it doesn't matter. But in some special cases, such as freeze fs, we hope the transaction is fully on disk, it will introduce some bugs, for example, we may feeze the fs and dump the data in the disk, if the transction doesn't complete, we would dump inconsistent data. So we need fix the above problem for those cases. We fixes this problem by introducing a function: btrfs_attach_transaction_barrier() if we hope all the transaction is fully on the disk, even they are not running, we can use this function. Signed-off-by: Miao Xie Signed-off-by: Josef Bacik --- fs/btrfs/ioctl.c | 2 +- fs/btrfs/super.c | 4 ++-- fs/btrfs/transaction.c | 32 ++++++++++++++++++++++++++++++++ fs/btrfs/transaction.h | 2 ++ 4 files changed, 37 insertions(+), 3 deletions(-) (limited to 'fs/btrfs/transaction.c') diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index 0f68729f261e..e0893036da08 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -3114,7 +3114,7 @@ static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root, u64 transid; int ret; - trans = btrfs_attach_transaction(root); + trans = btrfs_attach_transaction_barrier(root); if (IS_ERR(trans)) { if (PTR_ERR(trans) != -ENOENT) return PTR_ERR(trans); diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index e0ea1278377a..db1ba9a2ed64 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -878,7 +878,7 @@ int btrfs_sync_fs(struct super_block *sb, int wait) btrfs_wait_ordered_extents(root, 0); - trans = btrfs_attach_transaction(root); + trans = btrfs_attach_transaction_barrier(root); if (IS_ERR(trans)) { /* no transaction, don't bother */ if (PTR_ERR(trans) == -ENOENT) @@ -1563,7 +1563,7 @@ static int btrfs_freeze(struct super_block *sb) struct btrfs_trans_handle *trans; struct btrfs_root *root = btrfs_sb(sb)->tree_root; - trans = btrfs_attach_transaction(root); + trans = btrfs_attach_transaction_barrier(root); if (IS_ERR(trans)) { /* no transaction, don't bother */ if (PTR_ERR(trans) == -ENOENT) diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 5767ea1c0150..c1ce664c0c39 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -469,11 +469,43 @@ struct btrfs_trans_handle *btrfs_start_ioctl_transaction(struct btrfs_root *root return start_transaction(root, 0, TRANS_USERSPACE, 0); } +/* + * btrfs_attach_transaction() - catch the running transaction + * + * It is used when we want to commit the current the transaction, but + * don't want to start a new one. + * + * Note: If this function return -ENOENT, it just means there is no + * running transaction. But it is possible that the inactive transaction + * is still in the memory, not fully on disk. If you hope there is no + * inactive transaction in the fs when -ENOENT is returned, you should + * invoke + * btrfs_attach_transaction_barrier() + */ struct btrfs_trans_handle *btrfs_attach_transaction(struct btrfs_root *root) { return start_transaction(root, 0, TRANS_ATTACH, 0); } +/* + * btrfs_attach_transaction() - catch the running transaction + * + * It is similar to the above function, the differentia is this one + * will wait for all the inactive transactions until they fully + * complete. + */ +struct btrfs_trans_handle * +btrfs_attach_transaction_barrier(struct btrfs_root *root) +{ + struct btrfs_trans_handle *trans; + + trans = start_transaction(root, 0, TRANS_ATTACH, 0); + if (IS_ERR(trans) && PTR_ERR(trans) == -ENOENT) + btrfs_wait_for_commit(root, 0); + + return trans; +} + /* wait for a transaction commit to be fully complete */ static noinline void wait_for_commit(struct btrfs_root *root, struct btrfs_transaction *commit) diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h index 3f772fd0191a..5afd7b1dceac 100644 --- a/fs/btrfs/transaction.h +++ b/fs/btrfs/transaction.h @@ -112,6 +112,8 @@ struct btrfs_trans_handle *btrfs_start_transaction_lflush( struct btrfs_trans_handle *btrfs_join_transaction(struct btrfs_root *root); struct btrfs_trans_handle *btrfs_join_transaction_nolock(struct btrfs_root *root); struct btrfs_trans_handle *btrfs_attach_transaction(struct btrfs_root *root); +struct btrfs_trans_handle *btrfs_attach_transaction_barrier( + struct btrfs_root *root); struct btrfs_trans_handle *btrfs_start_ioctl_transaction(struct btrfs_root *root); int btrfs_wait_for_commit(struct btrfs_root *root, u64 transid); int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans, -- cgit v1.2.3 From 272d26d0ad8c0e326689f2fa3cdc6a5fcc8e74e0 Mon Sep 17 00:00:00 2001 From: Miao Xie Date: Wed, 20 Feb 2013 14:16:39 +0000 Subject: Btrfs: fix missing release of qgroup reservation in commit_transaction() We forget to free qgroup reservation in commit_transaction(),fix it. Signed-off-by: Miao Xie Signed-off-by: Wang Shilong Cc: Arne Jansen Signed-off-by: Josef Bacik --- fs/btrfs/transaction.c | 8 ++++++++ 1 file changed, 8 insertions(+) (limited to 'fs/btrfs/transaction.c') diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index c1ce664c0c39..955204ca0447 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -1553,6 +1553,10 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, btrfs_trans_release_metadata(trans, root); trans->block_rsv = NULL; + if (trans->qgroup_reserved) { + btrfs_qgroup_free(root, trans->qgroup_reserved); + trans->qgroup_reserved = 0; + } cur_trans = trans->transaction; @@ -1833,6 +1837,10 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, cleanup_transaction: btrfs_trans_release_metadata(trans, root); trans->block_rsv = NULL; + if (trans->qgroup_reserved) { + btrfs_qgroup_free(root, trans->qgroup_reserved); + trans->qgroup_reserved = 0; + } btrfs_printk(root->fs_info, "Skipping commit of aborted transaction.\n"); // WARN_ON(1); if (current->journal_info == trans) -- cgit v1.2.3 From 2382c5cc7ed0396b61a359765bf5ee125b0a2f46 Mon Sep 17 00:00:00 2001 From: Liu Bo Date: Fri, 22 Feb 2013 04:33:36 +0000 Subject: Btrfs: use reserved space for creating a snapshot While inserting dir index and updating inode for a snapshot, we'd add delayed items which consume trans->block_rsv, if we don't have any space reserved in this trans handle, we either just return or reserve space again. But before creating pending snapshots during committing transaction, we've done a release on this trans handle, so we don't have space reserved in it at this stage. What we're using is block_rsv of pending snapshots which has already reserved well enough space for both inserting dir index and updating inode, so we need to set trans handle to indicate that we have space now. Signed-off-by: Liu Bo Reviewed-by: Miao Xie Signed-off-by: Josef Bacik --- fs/btrfs/transaction.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'fs/btrfs/transaction.c') diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index a83d486cc70c..4330433b7b4f 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -1123,6 +1123,7 @@ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans, rsv = trans->block_rsv; trans->block_rsv = &pending->block_rsv; + trans->bytes_reserved = trans->block_rsv->reserved; dentry = pending->dentry; parent = dget_parent(dentry); @@ -1276,6 +1277,7 @@ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans, fail: dput(parent); trans->block_rsv = rsv; + trans->bytes_reserved = 0; no_free_objectid: kfree(new_root_item); root_item_alloc_fail: -- cgit v1.2.3 From f094ac32aba3a51c00e970a2ea029339af2ca048 Mon Sep 17 00:00:00 2001 From: Liu Bo Date: Wed, 27 Feb 2013 13:28:25 +0000 Subject: Btrfs: fix NULL pointer after aborting a transaction While doing cleanup work on an aborted transaction, we've set the global running transaction pointer to NULL _before_ waiting all other transaction handles to finish, so others'd hit NULL pointer crash when referencing the global running transaction pointer. This first sets a hint to avoid new transaction handle joining, then waits other existing handles to abort or finish so that we can safely set the above global pointer to NULL. Signed-off-by: Liu Bo Signed-off-by: Josef Bacik --- fs/btrfs/transaction.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-) (limited to 'fs/btrfs/transaction.c') diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 4330433b7b4f..3733c4939a27 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -1447,6 +1447,7 @@ static void cleanup_transaction(struct btrfs_trans_handle *trans, struct btrfs_root *root, int err) { struct btrfs_transaction *cur_trans = trans->transaction; + DEFINE_WAIT(wait); WARN_ON(trans->use_count > 1); @@ -1455,8 +1456,13 @@ static void cleanup_transaction(struct btrfs_trans_handle *trans, spin_lock(&root->fs_info->trans_lock); list_del_init(&cur_trans->list); if (cur_trans == root->fs_info->running_transaction) { + root->fs_info->trans_no_join = 1; + spin_unlock(&root->fs_info->trans_lock); + wait_event(cur_trans->writer_wait, + atomic_read(&cur_trans->num_writers) == 1); + + spin_lock(&root->fs_info->trans_lock); root->fs_info->running_transaction = NULL; - root->fs_info->trans_no_join = 0; } spin_unlock(&root->fs_info->trans_lock); -- cgit v1.2.3 From e9662f701c85ebc99f532bf8bb53208c0648846a Mon Sep 17 00:00:00 2001 From: Miao Xie Date: Thu, 28 Feb 2013 10:01:15 +0000 Subject: Btrfs: remove unnecessary dget_parent/dput when creating the pending snapshot Since we have grabbed the parent inode at the beginning of the snapshot creation, and both sync and async snapshot creation release it after the pending snapshots are actually created, it is safe to access the parent inode directly during the snapshot creation, we needn't use dget_parent/dput to fix the parent dentry and get the dir inode. Signed-off-by: Miao Xie Signed-off-by: Josef Bacik --- fs/btrfs/ioctl.c | 10 ++++++---- fs/btrfs/transaction.c | 5 +---- fs/btrfs/transaction.h | 1 + 3 files changed, 8 insertions(+), 8 deletions(-) (limited to 'fs/btrfs/transaction.c') diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index 059546aa8fdf..8dcd4ff0c3a5 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -530,9 +530,10 @@ fail: return ret; } -static int create_snapshot(struct btrfs_root *root, struct dentry *dentry, - char *name, int namelen, u64 *async_transid, - bool readonly, struct btrfs_qgroup_inherit *inherit) +static int create_snapshot(struct btrfs_root *root, struct inode *dir, + struct dentry *dentry, char *name, int namelen, + u64 *async_transid, bool readonly, + struct btrfs_qgroup_inherit *inherit) { struct inode *inode; struct btrfs_pending_snapshot *pending_snapshot; @@ -551,6 +552,7 @@ static int create_snapshot(struct btrfs_root *root, struct dentry *dentry, pending_snapshot->dentry = dentry; pending_snapshot->root = root; pending_snapshot->readonly = readonly; + pending_snapshot->dir = dir; pending_snapshot->inherit = inherit; trans = btrfs_start_transaction(root->fs_info->extent_root, 6); @@ -728,7 +730,7 @@ static noinline int btrfs_mksubvol(struct path *parent, goto out_up_read; if (snap_src) { - error = create_snapshot(snap_src, dentry, name, namelen, + error = create_snapshot(snap_src, dir, dentry, name, namelen, async_transid, readonly, inherit); } else { error = create_subvol(BTRFS_I(dir)->root, dentry, diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 3733c4939a27..71de435a291e 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -1068,7 +1068,6 @@ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans, struct inode *parent_inode; struct btrfs_path *path; struct btrfs_dir_item *dir_item; - struct dentry *parent; struct dentry *dentry; struct extent_buffer *tmp; struct extent_buffer *old; @@ -1126,8 +1125,7 @@ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans, trans->bytes_reserved = trans->block_rsv->reserved; dentry = pending->dentry; - parent = dget_parent(dentry); - parent_inode = parent->d_inode; + parent_inode = pending->dir; parent_root = BTRFS_I(parent_inode)->root; record_root_in_trans(trans, parent_root); @@ -1275,7 +1273,6 @@ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans, if (ret) btrfs_abort_transaction(trans, root, ret); fail: - dput(parent); trans->block_rsv = rsv; trans->bytes_reserved = 0; no_free_objectid: diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h index 5afd7b1dceac..5f67fba07ab4 100644 --- a/fs/btrfs/transaction.h +++ b/fs/btrfs/transaction.h @@ -84,6 +84,7 @@ struct btrfs_trans_handle { struct btrfs_pending_snapshot { struct dentry *dentry; + struct inode *dir; struct btrfs_root *root; struct btrfs_root *snap; struct btrfs_qgroup_inherit *inherit; -- cgit v1.2.3 From d5c1207017cd8387b4d3224dd7ab6cf5cd7f1c9a Mon Sep 17 00:00:00 2001 From: Miao Xie Date: Thu, 28 Feb 2013 10:04:33 +0000 Subject: Btrfs: fix wrong reserved space in qgroup during snap/subv creation There are two problems in the space reservation of the snapshot/ subvolume creation. - don't reserve the space for the root item insertion - the space which is reserved in the qgroup is different with the free space reservation. we need reserve free space for 7 items, but in qgroup reservation, we need reserve space only for 3 items. So we implement new metadata reservation functions for the snapshot/subvolume creation. Signed-off-by: Miao Xie Signed-off-by: Josef Bacik --- fs/btrfs/ctree.h | 9 +++++-- fs/btrfs/extent-tree.c | 65 ++++++++++++++++++++++++++++++++++++++++---------- fs/btrfs/ioctl.c | 62 ++++++++++++++++++++++++++++++++--------------- fs/btrfs/transaction.c | 4 +--- fs/btrfs/transaction.h | 1 + 5 files changed, 105 insertions(+), 36 deletions(-) (limited to 'fs/btrfs/transaction.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index ae8dcc406805..0d82922179db 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -3106,8 +3106,13 @@ void btrfs_trans_release_metadata(struct btrfs_trans_handle *trans, int btrfs_orphan_reserve_metadata(struct btrfs_trans_handle *trans, struct inode *inode); void btrfs_orphan_release_metadata(struct inode *inode); -int btrfs_snap_reserve_metadata(struct btrfs_trans_handle *trans, - struct btrfs_pending_snapshot *pending); +int btrfs_subvolume_reserve_metadata(struct btrfs_root *root, + struct btrfs_block_rsv *rsv, + int nitems, + u64 *qgroup_reserved); +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 num_bytes); diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 5681a91ed400..7cb9d734f6e5 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -4560,19 +4560,60 @@ void btrfs_orphan_release_metadata(struct inode *inode) btrfs_block_rsv_release(root, root->orphan_block_rsv, num_bytes); } -int btrfs_snap_reserve_metadata(struct btrfs_trans_handle *trans, - struct btrfs_pending_snapshot *pending) +/* + * btrfs_subvolume_reserve_metadata() - reserve space for subvolume operation + * root: the root of the parent directory + * rsv: block reservation + * items: the number of items that we need do reservation + * qgroup_reserved: used to return the reserved size in qgroup + * + * This function is used to reserve the space for snapshot/subvolume + * creation and deletion. Those operations are different with the + * common file/directory operations, they change two fs/file trees + * and root tree, the number of items that the qgroup reserves is + * different with the free space reservation. So we can not use + * the space reseravtion mechanism in start_transaction(). + */ +int btrfs_subvolume_reserve_metadata(struct btrfs_root *root, + struct btrfs_block_rsv *rsv, + int items, + u64 *qgroup_reserved) { - struct btrfs_root *root = pending->root; - struct btrfs_block_rsv *src_rsv = get_block_rsv(trans, root); - struct btrfs_block_rsv *dst_rsv = &pending->block_rsv; - /* - * two for root back/forward refs, two for directory entries, - * one for root of the snapshot and one for parent inode. - */ - u64 num_bytes = btrfs_calc_trans_metadata_size(root, 6); - dst_rsv->space_info = src_rsv->space_info; - return block_rsv_migrate_bytes(src_rsv, dst_rsv, num_bytes); + u64 num_bytes; + int ret; + + if (root->fs_info->quota_enabled) { + /* One for parent inode, two for dir entries */ + num_bytes = 3 * root->leafsize; + ret = btrfs_qgroup_reserve(root, num_bytes); + if (ret) + return ret; + } else { + num_bytes = 0; + } + + *qgroup_reserved = num_bytes; + + num_bytes = btrfs_calc_trans_metadata_size(root, items); + rsv->space_info = __find_space_info(root->fs_info, + BTRFS_BLOCK_GROUP_METADATA); + ret = btrfs_block_rsv_add(root, rsv, num_bytes, + BTRFS_RESERVE_FLUSH_ALL); + if (ret) { + if (*qgroup_reserved) + btrfs_qgroup_free(root, *qgroup_reserved); + } + + return ret; +} + +void btrfs_subvolume_release_metadata(struct btrfs_root *root, + struct btrfs_block_rsv *rsv, + u64 qgroup_reserved) +{ + btrfs_block_rsv_release(root, rsv, (u64)-1); + if (qgroup_reserved) + btrfs_qgroup_free(root, qgroup_reserved); } /** diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index 8dcd4ff0c3a5..56d92549389c 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -363,7 +363,7 @@ static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg) return 0; } -static noinline int create_subvol(struct btrfs_root *root, +static noinline int create_subvol(struct inode *dir, struct dentry *dentry, char *name, int namelen, u64 *async_transid, @@ -374,32 +374,39 @@ static noinline int create_subvol(struct btrfs_root *root, struct btrfs_root_item root_item; struct btrfs_inode_item *inode_item; struct extent_buffer *leaf; + struct btrfs_root *root = BTRFS_I(dir)->root; struct btrfs_root *new_root; - struct dentry *parent = dentry->d_parent; - struct inode *dir; + struct btrfs_block_rsv block_rsv; struct timespec cur_time = CURRENT_TIME; int ret; int err; u64 objectid; u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID; u64 index = 0; + u64 qgroup_reserved; uuid_le new_uuid; ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid); if (ret) return ret; - dir = parent->d_inode; - + btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); /* - * 1 - inode item - * 2 - refs - * 1 - root item - * 2 - dir items + * The same as the snapshot creation, please see the comment + * of create_snapshot(). */ - trans = btrfs_start_transaction(root, 6); - if (IS_ERR(trans)) - return PTR_ERR(trans); + ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, + 7, &qgroup_reserved); + if (ret) + return ret; + + trans = btrfs_start_transaction(root, 0); + if (IS_ERR(trans)) { + ret = PTR_ERR(trans); + goto out; + } + trans->block_rsv = &block_rsv; + trans->bytes_reserved = block_rsv.size; ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit); if (ret) @@ -515,6 +522,8 @@ static noinline int create_subvol(struct btrfs_root *root, BUG_ON(ret); fail: + trans->block_rsv = NULL; + trans->bytes_reserved = 0; if (async_transid) { *async_transid = trans->transid; err = btrfs_commit_transaction_async(trans, root, 1); @@ -526,7 +535,8 @@ fail: if (!ret) d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry)); - +out: + btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved); return ret; } @@ -549,21 +559,31 @@ static int create_snapshot(struct btrfs_root *root, struct inode *dir, btrfs_init_block_rsv(&pending_snapshot->block_rsv, BTRFS_BLOCK_RSV_TEMP); + /* + * 1 - parent dir inode + * 2 - dir entries + * 1 - root item + * 2 - root ref/backref + * 1 - root of snapshot + */ + ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root, + &pending_snapshot->block_rsv, 7, + &pending_snapshot->qgroup_reserved); + if (ret) + goto out; + pending_snapshot->dentry = dentry; pending_snapshot->root = root; pending_snapshot->readonly = readonly; pending_snapshot->dir = dir; pending_snapshot->inherit = inherit; - trans = btrfs_start_transaction(root->fs_info->extent_root, 6); + trans = btrfs_start_transaction(root, 0); if (IS_ERR(trans)) { ret = PTR_ERR(trans); goto fail; } - ret = btrfs_snap_reserve_metadata(trans, pending_snapshot); - BUG_ON(ret); - spin_lock(&root->fs_info->trans_lock); list_add(&pending_snapshot->list, &trans->transaction->pending_snapshots); @@ -600,6 +620,10 @@ static int create_snapshot(struct btrfs_root *root, struct inode *dir, d_instantiate(dentry, inode); ret = 0; fail: + btrfs_subvolume_release_metadata(BTRFS_I(dir)->root, + &pending_snapshot->block_rsv, + pending_snapshot->qgroup_reserved); +out: kfree(pending_snapshot); return ret; } @@ -733,8 +757,8 @@ static noinline int btrfs_mksubvol(struct path *parent, error = create_snapshot(snap_src, dir, dentry, name, namelen, async_transid, readonly, inherit); } else { - error = create_subvol(BTRFS_I(dir)->root, dentry, - name, namelen, async_transid, inherit); + error = create_subvol(dir, dentry, name, namelen, + async_transid, inherit); } if (!error) fsnotify_mkdir(dir, dentry); diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 71de435a291e..f11c2e0a3746 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -1082,7 +1082,7 @@ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans, path = btrfs_alloc_path(); if (!path) { ret = pending->error = -ENOMEM; - goto path_alloc_fail; + return ret; } new_root_item = kmalloc(sizeof(*new_root_item), GFP_NOFS); @@ -1279,8 +1279,6 @@ no_free_objectid: kfree(new_root_item); root_item_alloc_fail: btrfs_free_path(path); -path_alloc_fail: - btrfs_block_rsv_release(root, &pending->block_rsv, (u64)-1); return ret; } diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h index 5f67fba07ab4..3c8e0d25c8e4 100644 --- a/fs/btrfs/transaction.h +++ b/fs/btrfs/transaction.h @@ -90,6 +90,7 @@ struct btrfs_pending_snapshot { struct btrfs_qgroup_inherit *inherit; /* block reservation for the operation */ struct btrfs_block_rsv block_rsv; + u64 qgroup_reserved; /* extra metadata reseration for relocation */ int error; bool readonly; -- cgit v1.2.3 From aec8030a8745221c8658f2033b22c98528897b13 Mon Sep 17 00:00:00 2001 From: Miao Xie Date: Mon, 4 Mar 2013 09:44:29 +0000 Subject: Btrfs: fix wrong handle at error path of create_snapshot() when the commit fails There are several bugs at error path of create_snapshot() when the transaction commitment failed. - access the freed transaction handler. At the end of the transaction commitment, the transaction handler was freed, so we should not access it after the transaction commitment. - we were not aware of the error which happened during the snapshot creation if we submitted a async transaction commitment. - pending snapshot access vs pending snapshot free. when something wrong happened after we submitted a async transaction commitment, the transaction committer would cleanup the pending snapshots and free them. But the snapshot creators were not aware of it, they would access the freed pending snapshots. This patch fixes the above problems by: - remove the dangerous code that accessed the freed handler - assign ->error if the error happens during the snapshot creation - the transaction committer doesn't free the pending snapshots, just assigns the error number and evicts them before we unblock the transaction. Reported-by: Dan Carpenter Signed-off-by: Miao Xie Signed-off-by: Josef Bacik --- fs/btrfs/disk-io.c | 16 ++++++-------- fs/btrfs/ioctl.c | 6 +----- fs/btrfs/transaction.c | 58 ++++++++++++++++++++++++++++---------------------- 3 files changed, 41 insertions(+), 39 deletions(-) (limited to 'fs/btrfs/transaction.c') diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 02369a3c162e..7d84651e850b 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -62,7 +62,7 @@ static void btrfs_destroy_ordered_operations(struct btrfs_transaction *t, static void btrfs_destroy_ordered_extents(struct btrfs_root *root); static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, struct btrfs_root *root); -static void btrfs_destroy_pending_snapshots(struct btrfs_transaction *t); +static void btrfs_evict_pending_snapshots(struct btrfs_transaction *t); static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root); static int btrfs_destroy_marked_extents(struct btrfs_root *root, struct extent_io_tree *dirty_pages, @@ -3687,7 +3687,7 @@ int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, return ret; } -static void btrfs_destroy_pending_snapshots(struct btrfs_transaction *t) +static void btrfs_evict_pending_snapshots(struct btrfs_transaction *t) { struct btrfs_pending_snapshot *snapshot; struct list_head splice; @@ -3700,10 +3700,8 @@ static void btrfs_destroy_pending_snapshots(struct btrfs_transaction *t) snapshot = list_entry(splice.next, struct btrfs_pending_snapshot, list); - + snapshot->error = -ECANCELED; list_del_init(&snapshot->list); - - kfree(snapshot); } } @@ -3840,6 +3838,8 @@ void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, cur_trans->blocked = 1; wake_up(&root->fs_info->transaction_blocked_wait); + btrfs_evict_pending_snapshots(cur_trans); + cur_trans->blocked = 0; wake_up(&root->fs_info->transaction_wait); @@ -3849,8 +3849,6 @@ void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, btrfs_destroy_delayed_inodes(root); btrfs_assert_delayed_root_empty(root); - btrfs_destroy_pending_snapshots(cur_trans); - btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages, EXTENT_DIRTY); btrfs_destroy_pinned_extent(root, @@ -3894,6 +3892,8 @@ int btrfs_cleanup_transaction(struct btrfs_root *root) if (waitqueue_active(&root->fs_info->transaction_blocked_wait)) wake_up(&root->fs_info->transaction_blocked_wait); + btrfs_evict_pending_snapshots(t); + t->blocked = 0; smp_mb(); if (waitqueue_active(&root->fs_info->transaction_wait)) @@ -3907,8 +3907,6 @@ int btrfs_cleanup_transaction(struct btrfs_root *root) btrfs_destroy_delayed_inodes(root); btrfs_assert_delayed_root_empty(root); - btrfs_destroy_pending_snapshots(t); - btrfs_destroy_delalloc_inodes(root); spin_lock(&root->fs_info->trans_lock); diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index b908960c9746..94c0e42dfa1e 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -596,12 +596,8 @@ static int create_snapshot(struct btrfs_root *root, struct inode *dir, ret = btrfs_commit_transaction(trans, root->fs_info->extent_root); } - if (ret) { - /* cleanup_transaction has freed this for us */ - if (trans->aborted) - pending_snapshot = NULL; + if (ret) goto fail; - } ret = pending_snapshot->error; if (ret) diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index f11c2e0a3746..d8fce6fe9cf8 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -1053,7 +1053,12 @@ int btrfs_defrag_root(struct btrfs_root *root) /* * new snapshots need to be created at a very specific time in the - * transaction commit. This does the actual creation + * transaction commit. This does the actual creation. + * + * Note: + * If the error which may affect the commitment of the current transaction + * happens, we should return the error number. If the error which just affect + * the creation of the pending snapshots, just return 0. */ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans, struct btrfs_fs_info *fs_info, @@ -1072,7 +1077,7 @@ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans, struct extent_buffer *tmp; struct extent_buffer *old; struct timespec cur_time = CURRENT_TIME; - int ret; + int ret = 0; u64 to_reserve = 0; u64 index = 0; u64 objectid; @@ -1081,40 +1086,36 @@ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans, path = btrfs_alloc_path(); if (!path) { - ret = pending->error = -ENOMEM; - return ret; + pending->error = -ENOMEM; + return 0; } new_root_item = kmalloc(sizeof(*new_root_item), GFP_NOFS); if (!new_root_item) { - ret = pending->error = -ENOMEM; + pending->error = -ENOMEM; goto root_item_alloc_fail; } - ret = btrfs_find_free_objectid(tree_root, &objectid); - if (ret) { - pending->error = ret; + pending->error = btrfs_find_free_objectid(tree_root, &objectid); + if (pending->error) goto no_free_objectid; - } btrfs_reloc_pre_snapshot(trans, pending, &to_reserve); if (to_reserve > 0) { - ret = btrfs_block_rsv_add(root, &pending->block_rsv, - to_reserve, - BTRFS_RESERVE_NO_FLUSH); - if (ret) { - pending->error = ret; + pending->error = btrfs_block_rsv_add(root, + &pending->block_rsv, + to_reserve, + BTRFS_RESERVE_NO_FLUSH); + if (pending->error) goto no_free_objectid; - } } - ret = btrfs_qgroup_inherit(trans, fs_info, root->root_key.objectid, - objectid, pending->inherit); - if (ret) { - pending->error = ret; + pending->error = btrfs_qgroup_inherit(trans, fs_info, + root->root_key.objectid, + objectid, pending->inherit); + if (pending->error) goto no_free_objectid; - } key.objectid = objectid; key.offset = (u64)-1; @@ -1142,7 +1143,7 @@ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans, dentry->d_name.len, 0); if (dir_item != NULL && !IS_ERR(dir_item)) { pending->error = -EEXIST; - goto fail; + goto dir_item_existed; } else if (IS_ERR(dir_item)) { ret = PTR_ERR(dir_item); btrfs_abort_transaction(trans, root, ret); @@ -1273,6 +1274,8 @@ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans, if (ret) btrfs_abort_transaction(trans, root, ret); fail: + pending->error = ret; +dir_item_existed: trans->block_rsv = rsv; trans->bytes_reserved = 0; no_free_objectid: @@ -1288,12 +1291,17 @@ root_item_alloc_fail: static noinline int create_pending_snapshots(struct btrfs_trans_handle *trans, struct btrfs_fs_info *fs_info) { - struct btrfs_pending_snapshot *pending; + struct btrfs_pending_snapshot *pending, *next; struct list_head *head = &trans->transaction->pending_snapshots; + int ret = 0; - list_for_each_entry(pending, head, list) - create_pending_snapshot(trans, fs_info, pending); - return 0; + list_for_each_entry_safe(pending, next, head, list) { + list_del(&pending->list); + ret = create_pending_snapshot(trans, fs_info, pending); + if (ret) + break; + } + return ret; } static void update_super_roots(struct btrfs_root *root) -- cgit v1.2.3 From 66b6135b7cf741f6f44ba938b27583ea3b83bd12 Mon Sep 17 00:00:00 2001 From: Liu Bo Date: Mon, 4 Mar 2013 16:25:41 +0000 Subject: Btrfs: avoid deadlock on transaction waiting list Only let one trans handle to wait for other handles, otherwise we will get ABBA issues. Signed-off-by: Liu Bo Signed-off-by: Josef Bacik --- fs/btrfs/transaction.c | 7 +++++++ 1 file changed, 7 insertions(+) (limited to 'fs/btrfs/transaction.c') diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index d8fce6fe9cf8..fedede1fe178 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -1457,6 +1457,13 @@ static void cleanup_transaction(struct btrfs_trans_handle *trans, btrfs_abort_transaction(trans, root, err); spin_lock(&root->fs_info->trans_lock); + + if (list_empty(&cur_trans->list)) { + spin_unlock(&root->fs_info->trans_lock); + btrfs_end_transaction(trans, root); + return; + } + list_del_init(&cur_trans->list); if (cur_trans == root->fs_info->running_transaction) { root->fs_info->trans_no_join = 1; -- cgit v1.2.3 From 7c2ec3f0730729f4829d01f7c19538d135f86712 Mon Sep 17 00:00:00 2001 From: Liu Bo Date: Wed, 13 Mar 2013 07:43:03 -0600 Subject: Btrfs: fix warning when creating snapshots Creating snapshot passes extent_root to commit its transaction, but it can lead to the warning of checking root for quota in the __btrfs_end_transaction() when someone else is committing the current transaction. Since we've recorded the needed root in trans_handle, just use it to get rid of the warning. Signed-off-by: Liu Bo Signed-off-by: Chris Mason --- fs/btrfs/transaction.c | 11 +++++------ 1 file changed, 5 insertions(+), 6 deletions(-) (limited to 'fs/btrfs/transaction.c') diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index fedede1fe178..c4a1531c52d8 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -626,14 +626,13 @@ static int __btrfs_end_transaction(struct btrfs_trans_handle *trans, btrfs_trans_release_metadata(trans, root); trans->block_rsv = NULL; - /* - * the same root has to be passed to start_transaction and - * end_transaction. Subvolume quota depends on this. - */ - WARN_ON(trans->root != root); if (trans->qgroup_reserved) { - btrfs_qgroup_free(root, trans->qgroup_reserved); + /* + * the same root has to be passed here between start_transaction + * and end_transaction. Subvolume quota depends on this. + */ + btrfs_qgroup_free(trans->root, trans->qgroup_reserved); trans->qgroup_reserved = 0; } -- cgit v1.2.3