From 3c911608085bf2d5a0822c418129f96a2a89d1b5 Mon Sep 17 00:00:00 2001 From: Eric Sandeen Date: Thu, 31 Jan 2013 00:55:02 +0000 Subject: btrfs: don't try to notify udev about missing devices If we remove a missing device, bdev is null, and if we send that off to btrfs_kobject_uevent we'll panic. Signed-off-by: Eric Sandeen Signed-off-by: Josef Bacik Signed-off-by: Chris Mason --- fs/btrfs/volumes.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'fs/btrfs/volumes.c') diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index 5cce6aa74012..485a5423e3c6 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -1556,7 +1556,8 @@ int btrfs_rm_device(struct btrfs_root *root, char *device_path) ret = 0; /* Notify udev that device has changed */ - btrfs_kobject_uevent(bdev, KOBJ_CHANGE); + if (bdev) + btrfs_kobject_uevent(bdev, KOBJ_CHANGE); error_brelse: brelse(bh); -- 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/volumes.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 1f0905ec156eec8f12cd593bc564551770319720 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 5 Feb 2013 10:01:42 -0500 Subject: Btrfs: remove conflicting check for minimum number of devices in raid56 The device removal code was incorrectly checking against two different limits for raid5 and raid6. Signed-off-by: Chris Mason --- fs/btrfs/volumes.c | 8 -------- 1 file changed, 8 deletions(-) (limited to 'fs/btrfs/volumes.c') diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index c372264b85bf..77620f2d8af9 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -1392,14 +1392,6 @@ 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"); -- cgit v1.2.3 From 6f60cbd3ae442cb35861bb522f388db123d42ec1 Mon Sep 17 00:00:00 2001 From: David Sterba Date: Fri, 15 Feb 2013 11:31:02 -0700 Subject: btrfs: access superblock via pagecache in scan_one_device btrfs_scan_one_device is calling set_blocksize() which can race with a concurrent process making dirty page cache pages. It can end up dropping dirty page cache pages on the floor, which isn't very nice when someone is just running btrfs dev scan to find filesystems on the box. Now that udev is registering btrfs devices as it discovers them, we can actually end up racing with our own mkfs program too. When this happens, we drop some of the important blocks written by mkfs. This commit changes scan_one_device to read the super out of the page cache instead of trying to use bread. This way we don't have to care about the blocksize of the device. This also drops the invalidate_bdev() call. It wasn't very polite to invalidate during the scan either. mkfs is putting the super into the page cache, there's no reason to invalidate at this point. Signed-off-by: David Sterba Signed-off-by: Chris Mason --- fs/btrfs/volumes.c | 70 +++++++++++++++++++++++++++++++++++++++++++++++++----- 1 file changed, 64 insertions(+), 6 deletions(-) (limited to 'fs/btrfs/volumes.c') diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index 5cbb7f4b1672..5349e17d8863 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -792,26 +792,77 @@ int btrfs_open_devices(struct btrfs_fs_devices *fs_devices, return ret; } +/* + * Look for a btrfs signature on a device. This may be called out of the mount path + * and we are not allowed to call set_blocksize during the scan. The superblock + * is read via pagecache + */ int btrfs_scan_one_device(const char *path, fmode_t flags, void *holder, struct btrfs_fs_devices **fs_devices_ret) { struct btrfs_super_block *disk_super; struct block_device *bdev; - struct buffer_head *bh; - int ret; + struct page *page; + void *p; + int ret = -EINVAL; u64 devid; u64 transid; u64 total_devices; + u64 bytenr; + pgoff_t index; + /* + * we would like to check all the supers, but that would make + * a btrfs mount succeed after a mkfs from a different FS. + * So, we need to add a special mount option to scan for + * later supers, using BTRFS_SUPER_MIRROR_MAX instead + */ + bytenr = btrfs_sb_offset(0); flags |= FMODE_EXCL; mutex_lock(&uuid_mutex); - ret = btrfs_get_bdev_and_sb(path, flags, holder, 0, &bdev, &bh); - if (ret) + + bdev = blkdev_get_by_path(path, flags, holder); + + if (IS_ERR(bdev)) { + ret = PTR_ERR(bdev); + printk(KERN_INFO "btrfs: open %s failed\n", path); goto error; - disk_super = (struct btrfs_super_block *)bh->b_data; + } + + /* make sure our super fits in the device */ + if (bytenr + PAGE_CACHE_SIZE >= i_size_read(bdev->bd_inode)) + goto error_bdev_put; + + /* make sure our super fits in the page */ + if (sizeof(*disk_super) > PAGE_CACHE_SIZE) + goto error_bdev_put; + + /* make sure our super doesn't straddle pages on disk */ + index = bytenr >> PAGE_CACHE_SHIFT; + if ((bytenr + sizeof(*disk_super) - 1) >> PAGE_CACHE_SHIFT != index) + goto error_bdev_put; + + /* pull in the page with our super */ + page = read_cache_page_gfp(bdev->bd_inode->i_mapping, + index, GFP_NOFS); + + if (IS_ERR_OR_NULL(page)) + goto error_bdev_put; + + p = kmap(page); + + /* align our pointer to the offset of the super block */ + disk_super = p + (bytenr & ~PAGE_CACHE_MASK); + + if (btrfs_super_bytenr(disk_super) != bytenr || + strncmp((char *)(&disk_super->magic), BTRFS_MAGIC, + sizeof(disk_super->magic))) + goto error_unmap; + devid = btrfs_stack_device_id(&disk_super->dev_item); transid = btrfs_super_generation(disk_super); total_devices = btrfs_super_num_devices(disk_super); + if (disk_super->label[0]) { if (disk_super->label[BTRFS_LABEL_SIZE - 1]) disk_super->label[BTRFS_LABEL_SIZE - 1] = '\0'; @@ -819,12 +870,19 @@ int btrfs_scan_one_device(const char *path, fmode_t flags, void *holder, } else { printk(KERN_INFO "device fsid %pU ", disk_super->fsid); } + printk(KERN_CONT "devid %llu transid %llu %s\n", (unsigned long long)devid, (unsigned long long)transid, path); + ret = device_list_add(path, disk_super, devid, fs_devices_ret); if (!ret && fs_devices_ret) (*fs_devices_ret)->total_devices = total_devices; - brelse(bh); + +error_unmap: + kunmap(page); + page_cache_release(page); + +error_bdev_put: blkdev_put(bdev, flags); error: mutex_unlock(&uuid_mutex); -- cgit v1.2.3 From e6ec716f0ddbe51741ef261d0804f0c28038dda4 Mon Sep 17 00:00:00 2001 From: Miao Xie Date: Thu, 17 Jan 2013 05:38:51 +0000 Subject: Btrfs: make raid attr array more readable The current code of raid attr arry is hard to understand and it is easy to introduce some problem if we modify the array. So I changed it and made it more readable. Cc: Liu Bo Signed-off-by: Miao Xie Signed-off-by: Josef Bacik --- fs/btrfs/ctree.h | 10 +++++++++- fs/btrfs/extent-tree.c | 22 +++++++++------------- fs/btrfs/volumes.c | 47 +++++++++++++++++++++++++++++++++++++++++------ 3 files changed, 59 insertions(+), 20 deletions(-) (limited to 'fs/btrfs/volumes.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 9ee099f3f834..541ce9a9949e 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -953,7 +953,15 @@ struct btrfs_dev_replace_item { #define BTRFS_BLOCK_GROUP_DUP (1ULL << 5) #define BTRFS_BLOCK_GROUP_RAID10 (1ULL << 6) #define BTRFS_BLOCK_GROUP_RESERVED BTRFS_AVAIL_ALLOC_BIT_SINGLE -#define BTRFS_NR_RAID_TYPES 5 + +enum btrfs_raid_types { + BTRFS_RAID_RAID10, + BTRFS_RAID_RAID1, + BTRFS_RAID_DUP, + BTRFS_RAID_RAID0, + BTRFS_RAID_SINGLE, + BTRFS_NR_RAID_TYPES +}; #define BTRFS_BLOCK_GROUP_TYPE_MASK (BTRFS_BLOCK_GROUP_DATA | \ BTRFS_BLOCK_GROUP_SYSTEM | \ diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 82400b2b2517..174c4d5c692c 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -5545,20 +5545,16 @@ wait_block_group_cache_done(struct btrfs_block_group_cache *cache) int __get_raid_index(u64 flags) { - int index; - if (flags & BTRFS_BLOCK_GROUP_RAID10) - index = 0; + return BTRFS_RAID_RAID10; else if (flags & BTRFS_BLOCK_GROUP_RAID1) - index = 1; + return BTRFS_RAID_RAID1; else if (flags & BTRFS_BLOCK_GROUP_DUP) - index = 2; + return BTRFS_RAID_DUP; else if (flags & BTRFS_BLOCK_GROUP_RAID0) - index = 3; + return BTRFS_RAID_RAID0; else - index = 4; - - return index; + return BTRFS_RAID_SINGLE; } static int get_block_group_index(struct btrfs_block_group_cache *cache) @@ -7518,16 +7514,16 @@ int btrfs_can_relocate(struct btrfs_root *root, u64 bytenr) index = get_block_group_index(block_group); } - if (index == 0) { + if (index == BTRFS_RAID_RAID10) { dev_min = 4; /* Divide by 2 */ min_free >>= 1; - } else if (index == 1) { + } else if (index == BTRFS_RAID_RAID1) { dev_min = 2; - } else if (index == 2) { + } else if (index == BTRFS_RAID_DUP) { /* Multiply by 2 */ min_free <<= 1; - } else if (index == 3) { + } else if (index == BTRFS_RAID_RAID0) { dev_min = fs_devices->rw_devices; do_div(min_free, dev_min); } diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index 5349e17d8863..1ce581d58d87 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -3562,13 +3562,48 @@ static int btrfs_cmp_device_info(const void *a, const void *b) } struct btrfs_raid_attr btrfs_raid_array[BTRFS_NR_RAID_TYPES] = { - { 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, 1, 1, 1, 1 /* single */ }, + [BTRFS_RAID_RAID10] = { + .sub_stripes = 2, + .dev_stripes = 1, + .devs_max = 0, /* 0 == as many as possible */ + .devs_min = 4, + .devs_increment = 2, + .ncopies = 2, + }, + [BTRFS_RAID_RAID1] = { + .sub_stripes = 1, + .dev_stripes = 1, + .devs_max = 2, + .devs_min = 2, + .devs_increment = 2, + .ncopies = 2, + }, + [BTRFS_RAID_DUP] = { + .sub_stripes = 1, + .dev_stripes = 2, + .devs_max = 1, + .devs_min = 1, + .devs_increment = 1, + .ncopies = 2, + }, + [BTRFS_RAID_RAID0] = { + .sub_stripes = 1, + .dev_stripes = 1, + .devs_max = 0, + .devs_min = 2, + .devs_increment = 1, + .ncopies = 1, + }, + [BTRFS_RAID_SINGLE] = { + .sub_stripes = 1, + .dev_stripes = 1, + .devs_max = 1, + .devs_min = 1, + .devs_increment = 1, + .ncopies = 1, + }, }; - + static int __btrfs_alloc_chunk(struct btrfs_trans_handle *trans, struct btrfs_root *extent_root, struct map_lookup **map_ret, -- cgit v1.2.3 From de98ced9e743656d108de41841797def0f5cb951 Mon Sep 17 00:00:00 2001 From: Miao Xie Date: Tue, 29 Jan 2013 10:13:12 +0000 Subject: Btrfs: use seqlock to protect fs_info->avail_{data, metadata, system}_alloc_bits There is no lock to protect fs_info->avail_{data, metadata, system}_alloc_bits, it may introduce some problem, such as the wrong profile information, so we add a seqlock to protect them. Signed-off-by: Zhao Lei Signed-off-by: Miao Xie Signed-off-by: Josef Bacik --- fs/btrfs/ctree.h | 2 ++ fs/btrfs/disk-io.c | 1 + fs/btrfs/extent-tree.c | 22 ++++++++++++++------ fs/btrfs/volumes.c | 56 +++++++++++++++++++++++++++----------------------- 4 files changed, 49 insertions(+), 32 deletions(-) (limited to 'fs/btrfs/volumes.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 50def99f5379..5f5c30aaef36 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1483,6 +1483,8 @@ struct btrfs_fs_info { struct rb_root defrag_inodes; atomic_t defrag_running; + /* Used to protect avail_{data, metadata, system}_alloc_bits */ + seqlock_t profiles_lock; /* * these three are in extended format (availability of single * chunks is denoted by BTRFS_AVAIL_ALLOC_BIT_SINGLE bit, other diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 11f6dbcb1191..00b6742fdde7 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -2040,6 +2040,7 @@ int open_ctree(struct super_block *sb, spin_lock_init(&fs_info->tree_mod_seq_lock); rwlock_init(&fs_info->tree_mod_log_lock); mutex_init(&fs_info->reloc_mutex); + seqlock_init(&fs_info->profiles_lock); init_completion(&fs_info->kobj_unregister); INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 115d1646bf50..faff98f720de 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -3227,12 +3227,14 @@ static void set_avail_alloc_bits(struct btrfs_fs_info *fs_info, u64 flags) u64 extra_flags = chunk_to_extended(flags) & BTRFS_EXTENDED_PROFILE_MASK; + write_seqlock(&fs_info->profiles_lock); if (flags & BTRFS_BLOCK_GROUP_DATA) fs_info->avail_data_alloc_bits |= extra_flags; if (flags & BTRFS_BLOCK_GROUP_METADATA) fs_info->avail_metadata_alloc_bits |= extra_flags; if (flags & BTRFS_BLOCK_GROUP_SYSTEM) fs_info->avail_system_alloc_bits |= extra_flags; + write_sequnlock(&fs_info->profiles_lock); } /* @@ -3324,12 +3326,18 @@ u64 btrfs_reduce_alloc_profile(struct btrfs_root *root, u64 flags) static u64 get_alloc_profile(struct btrfs_root *root, u64 flags) { - if (flags & BTRFS_BLOCK_GROUP_DATA) - flags |= root->fs_info->avail_data_alloc_bits; - else if (flags & BTRFS_BLOCK_GROUP_SYSTEM) - flags |= root->fs_info->avail_system_alloc_bits; - else if (flags & BTRFS_BLOCK_GROUP_METADATA) - flags |= root->fs_info->avail_metadata_alloc_bits; + unsigned seq; + + do { + seq = read_seqbegin(&root->fs_info->profiles_lock); + + if (flags & BTRFS_BLOCK_GROUP_DATA) + flags |= root->fs_info->avail_data_alloc_bits; + else if (flags & BTRFS_BLOCK_GROUP_SYSTEM) + flags |= root->fs_info->avail_system_alloc_bits; + else if (flags & BTRFS_BLOCK_GROUP_METADATA) + flags |= root->fs_info->avail_metadata_alloc_bits; + } while (read_seqretry(&root->fs_info->profiles_lock, seq)); return btrfs_reduce_alloc_profile(root, flags); } @@ -7967,12 +7975,14 @@ static void clear_avail_alloc_bits(struct btrfs_fs_info *fs_info, u64 flags) u64 extra_flags = chunk_to_extended(flags) & BTRFS_EXTENDED_PROFILE_MASK; + write_seqlock(&fs_info->profiles_lock); if (flags & BTRFS_BLOCK_GROUP_DATA) fs_info->avail_data_alloc_bits &= ~extra_flags; if (flags & BTRFS_BLOCK_GROUP_METADATA) fs_info->avail_metadata_alloc_bits &= ~extra_flags; if (flags & BTRFS_BLOCK_GROUP_SYSTEM) fs_info->avail_system_alloc_bits &= ~extra_flags; + write_sequnlock(&fs_info->profiles_lock); } int btrfs_remove_block_group(struct btrfs_trans_handle *trans, diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index 1ce581d58d87..8c9ea4cd66bb 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -1430,14 +1430,19 @@ int btrfs_rm_device(struct btrfs_root *root, char *device_path) u64 devid; u64 num_devices; u8 *dev_uuid; + unsigned seq; int ret = 0; bool clear_super = false; mutex_lock(&uuid_mutex); - all_avail = root->fs_info->avail_data_alloc_bits | - root->fs_info->avail_system_alloc_bits | - root->fs_info->avail_metadata_alloc_bits; + do { + seq = read_seqbegin(&root->fs_info->profiles_lock); + + all_avail = root->fs_info->avail_data_alloc_bits | + root->fs_info->avail_system_alloc_bits | + root->fs_info->avail_metadata_alloc_bits; + } while (read_seqretry(&root->fs_info->profiles_lock, seq)); num_devices = root->fs_info->fs_devices->num_devices; btrfs_dev_replace_lock(&root->fs_info->dev_replace); @@ -3043,6 +3048,7 @@ int btrfs_balance(struct btrfs_balance_control *bctl, int mixed = 0; int ret; u64 num_devices; + unsigned seq; if (btrfs_fs_closing(fs_info) || atomic_read(&fs_info->balance_pause_req) || @@ -3126,22 +3132,26 @@ 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; - if (((bctl->sys.flags & BTRFS_BALANCE_ARGS_CONVERT) && - (fs_info->avail_system_alloc_bits & allowed) && - !(bctl->sys.target & allowed)) || - ((bctl->meta.flags & BTRFS_BALANCE_ARGS_CONVERT) && - (fs_info->avail_metadata_alloc_bits & allowed) && - !(bctl->meta.target & allowed))) { - if (bctl->flags & BTRFS_BALANCE_FORCE) { - printk(KERN_INFO "btrfs: force reducing metadata " - "integrity\n"); - } else { - printk(KERN_ERR "btrfs: balance will reduce metadata " - "integrity, use force if you want this\n"); - ret = -EINVAL; - goto out; + do { + seq = read_seqbegin(&fs_info->profiles_lock); + + if (((bctl->sys.flags & BTRFS_BALANCE_ARGS_CONVERT) && + (fs_info->avail_system_alloc_bits & allowed) && + !(bctl->sys.target & allowed)) || + ((bctl->meta.flags & BTRFS_BALANCE_ARGS_CONVERT) && + (fs_info->avail_metadata_alloc_bits & allowed) && + !(bctl->meta.target & allowed))) { + if (bctl->flags & BTRFS_BALANCE_FORCE) { + printk(KERN_INFO "btrfs: force reducing metadata " + "integrity\n"); + } else { + printk(KERN_ERR "btrfs: balance will reduce metadata " + "integrity, use force if you want this\n"); + ret = -EINVAL; + goto out; + } } - } + } while (read_seqretry(&fs_info->profiles_lock, seq)); if (bctl->sys.flags & BTRFS_BALANCE_ARGS_CONVERT) { int num_tolerated_disk_barrier_failures; @@ -3980,10 +3990,7 @@ static noinline int init_first_rw_device(struct btrfs_trans_handle *trans, if (ret) return ret; - alloc_profile = BTRFS_BLOCK_GROUP_METADATA | - fs_info->avail_metadata_alloc_bits; - alloc_profile = btrfs_reduce_alloc_profile(root, alloc_profile); - + alloc_profile = btrfs_get_alloc_profile(extent_root, 0); ret = __btrfs_alloc_chunk(trans, extent_root, &map, &chunk_size, &stripe_size, chunk_offset, alloc_profile); if (ret) @@ -3991,10 +3998,7 @@ static noinline int init_first_rw_device(struct btrfs_trans_handle *trans, sys_chunk_offset = chunk_offset + chunk_size; - alloc_profile = BTRFS_BLOCK_GROUP_SYSTEM | - fs_info->avail_system_alloc_bits; - alloc_profile = btrfs_reduce_alloc_profile(root, alloc_profile); - + alloc_profile = btrfs_get_alloc_profile(fs_info->chunk_root, 0); ret = __btrfs_alloc_chunk(trans, extent_root, &sys_map, &sys_chunk_size, &sys_stripe_size, sys_chunk_offset, alloc_profile); -- cgit v1.2.3 From 0448748849ef7c593be40e2c1404f7974bd3aac6 Mon Sep 17 00:00:00 2001 From: Josef Bacik Date: Tue, 29 Jan 2013 15:03:37 -0500 Subject: Btrfs: fix chunk allocation error handling If we error out allocating a dev extent we will have already created the block group and such which will cause problems since the allocator may have tried to allocate out of the block group that no longer exists. This will cause BUG_ON()'s in the bio submission path. This also makes a failure to allocate a dev extent a non-abort error, we will just clean up the dev extents we did allocate and exit. Now if we fail to delete the dev extents we will abort since we can't have half of the dev extents hanging around, but this will make us much less likely to abort. Thanks, Signed-off-by: Josef Bacik --- fs/btrfs/volumes.c | 32 ++++++++++++++++++++++---------- 1 file changed, 22 insertions(+), 10 deletions(-) (limited to 'fs/btrfs/volumes.c') diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index 8c9ea4cd66bb..13efbcf03122 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -3825,12 +3825,6 @@ static int __btrfs_alloc_chunk(struct btrfs_trans_handle *trans, if (ret) goto error; - ret = btrfs_make_block_group(trans, extent_root, 0, type, - BTRFS_FIRST_CHUNK_TREE_OBJECTID, - start, num_bytes); - if (ret) - goto error; - for (i = 0; i < map->num_stripes; ++i) { struct btrfs_device *device; u64 dev_offset; @@ -3842,15 +3836,33 @@ static int __btrfs_alloc_chunk(struct btrfs_trans_handle *trans, info->chunk_root->root_key.objectid, BTRFS_FIRST_CHUNK_TREE_OBJECTID, start, dev_offset, stripe_size); - if (ret) { - btrfs_abort_transaction(trans, extent_root, ret); - goto error; - } + if (ret) + goto error_dev_extent; + } + + ret = btrfs_make_block_group(trans, extent_root, 0, type, + BTRFS_FIRST_CHUNK_TREE_OBJECTID, + start, num_bytes); + if (ret) { + i = map->num_stripes - 1; + goto error_dev_extent; } kfree(devices_info); return 0; +error_dev_extent: + for (; i >= 0; i--) { + struct btrfs_device *device; + int err; + + device = map->stripes[i].dev; + err = btrfs_free_dev_extent(trans, device, start); + if (err) { + btrfs_abort_transaction(trans, extent_root, err); + break; + } + } error: kfree(map); kfree(devices_info); -- cgit v1.2.3 From 0f5d42b287f32417e54485d79f2318cf2970b37d Mon Sep 17 00:00:00 2001 From: Josef Bacik Date: Thu, 31 Jan 2013 10:23:04 -0500 Subject: Btrfs: remove extent mapping if we fail to add chunk I got a double free error when unmounting a file system that failed to add a chunk during its operation. This is because we will kfree the mapping that we created but leave the extent_map in the em_tree for chunks. So to fix this just remove the extent_map when we error out so we don't run into this problem. Thanks, Signed-off-by: Josef Bacik --- fs/btrfs/volumes.c | 14 ++++++++++++-- 1 file changed, 12 insertions(+), 2 deletions(-) (limited to 'fs/btrfs/volumes.c') diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index 13efbcf03122..c7843349c795 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -3821,9 +3821,10 @@ static int __btrfs_alloc_chunk(struct btrfs_trans_handle *trans, write_lock(&em_tree->lock); ret = add_extent_mapping(em_tree, em); write_unlock(&em_tree->lock); - free_extent_map(em); - if (ret) + if (ret) { + free_extent_map(em); goto error; + } for (i = 0; i < map->num_stripes; ++i) { struct btrfs_device *device; @@ -3848,6 +3849,7 @@ static int __btrfs_alloc_chunk(struct btrfs_trans_handle *trans, goto error_dev_extent; } + free_extent_map(em); kfree(devices_info); return 0; @@ -3863,6 +3865,14 @@ error_dev_extent: break; } } + write_lock(&em_tree->lock); + remove_extent_mapping(em_tree, em); + write_unlock(&em_tree->lock); + + /* One for our allocation */ + free_extent_map(em); + /* One for the tree reference */ + free_extent_map(em); error: kfree(map); kfree(devices_info); -- cgit v1.2.3 From 063d006fa06fbf73fab370921120380333a33e85 Mon Sep 17 00:00:00 2001 From: Eric Sandeen Date: Thu, 31 Jan 2013 00:55:01 +0000 Subject: btrfs: ensure we don't overrun devices_info[] in __btrfs_alloc_chunk WARN_ON isn't enough, we need to stop the loop if for any reason we would overrun the devices_info array. I tried to track down the connection between the length of the alloc_devices list and the rw_devices counter but it wasn't immediately obvious, so be defensive about it. Signed-off-by: Eric Sandeen Signed-off-by: Josef Bacik --- fs/btrfs/volumes.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) (limited to 'fs/btrfs/volumes.c') diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index c7843349c795..305b6a63ab1f 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -3734,12 +3734,16 @@ static int __btrfs_alloc_chunk(struct btrfs_trans_handle *trans, if (max_avail < BTRFS_STRIPE_LEN * dev_stripes) continue; + if (ndevs == fs_devices->rw_devices) { + WARN(1, "%s: found more than %llu devices\n", + __func__, fs_devices->rw_devices); + break; + } devices_info[ndevs].dev_offset = dev_offset; devices_info[ndevs].max_avail = max_avail; devices_info[ndevs].total_avail = total_avail; devices_info[ndevs].dev = device; ++ndevs; - WARN_ON(ndevs > fs_devices->rw_devices); } /* -- cgit v1.2.3 From bf023ecfcaf114c410759f375073e78b0871c120 Mon Sep 17 00:00:00 2001 From: Ilya Dryomov Date: Tue, 12 Feb 2013 16:27:46 +0000 Subject: Btrfs: eliminate a use-after-free in btrfs_balance() Commit 5af3e8cc introduced a use-after-free at volumes.c:3139: bctl is freed above in __cancel_balance() in all cases except for balance pause. Fix this by moving the offending check a couple statements above, the meaning of the check is preserved. Reported-by: Chris Mason Signed-off-by: Ilya Dryomov Signed-off-by: Josef Bacik --- fs/btrfs/volumes.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) (limited to 'fs/btrfs/volumes.c') diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index 305b6a63ab1f..d778e9666597 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -3195,6 +3195,11 @@ int btrfs_balance(struct btrfs_balance_control *bctl, mutex_lock(&fs_info->balance_mutex); atomic_dec(&fs_info->balance_running); + 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 (bargs) { memset(bargs, 0, sizeof(*bargs)); update_ioctl_balance_args(fs_info, 0, bargs); @@ -3205,11 +3210,6 @@ int btrfs_balance(struct btrfs_balance_control *bctl, __cancel_balance(fs_info); } - if (bctl->sys.flags & BTRFS_BALANCE_ARGS_CONVERT) { - fs_info->num_tolerated_disk_barrier_failures = - btrfs_calc_num_tolerated_disk_barrier_failures(fs_info); - } - wake_up(&fs_info->balance_wait_q); return ret; -- cgit v1.2.3 From 3e39cea61c4ecb5774c08f3e30507fdd58c1cca9 Mon Sep 17 00:00:00 2001 From: Ilya Dryomov Date: Tue, 12 Feb 2013 16:28:59 +0000 Subject: Btrfs: allow for selecting only completely empty chunks Enhance balance usage filter by making it possible to balance out only completely empty chunks. Today, usage filter properly acts on values from 1 to 99 inclusive, usage=100 selects all chunks, and usage=0 selects no chunks. This commit changes the usage=0 case: the new meaning is to restripe only completely empty chunks and nothing else. Suggested-by: David Sterba Signed-off-by: Ilya Dryomov Signed-off-by: Josef Bacik --- fs/btrfs/volumes.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs/btrfs/volumes.c') diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index d778e9666597..567076542d1d 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -2679,7 +2679,7 @@ static int chunk_usage_filter(struct btrfs_fs_info *fs_info, u64 chunk_offset, chunk_used = btrfs_block_group_used(&cache->item); if (bargs->usage == 0) - user_thresh = 0; + user_thresh = 1; else if (bargs->usage > 100) user_thresh = cache->key.offset; else -- cgit v1.2.3 From cdb4c5748cb3ac533889a6b0b95aa10651e68785 Mon Sep 17 00:00:00 2001 From: Zach Brown Date: Wed, 20 Feb 2013 00:55:13 +0000 Subject: btrfs: define BTRFS_MAGIC as a u64 value super.magic is an le64 but it's treated as an unterminated string when compared against BTRFS_MAGIC which is defined as a string. Instead define BTRFS_MAGIC as a normal hex value and use endian helpers to compare it to the super's magic. I tested this by mounting an fs made before the change and made sure that it didn't introduce sparse errors. This matches a similar cleanup that is pending in btrfs-progs. David Sterba pointed out that we should fix the kernel side as well :). Signed-off-by: Zach Brown Signed-off-by: Josef Bacik --- fs/btrfs/check-integrity.c | 3 +-- fs/btrfs/ctree.h | 2 +- fs/btrfs/disk-io.c | 6 ++---- fs/btrfs/volumes.c | 3 +-- 4 files changed, 5 insertions(+), 9 deletions(-) (limited to 'fs/btrfs/volumes.c') diff --git a/fs/btrfs/check-integrity.c b/fs/btrfs/check-integrity.c index 11d47bfb62b4..18af6f48781a 100644 --- a/fs/btrfs/check-integrity.c +++ b/fs/btrfs/check-integrity.c @@ -813,8 +813,7 @@ static int btrfsic_process_superblock_dev_mirror( (bh->b_data + (dev_bytenr & 4095)); if (btrfs_super_bytenr(super_tmp) != dev_bytenr || - strncmp((char *)(&(super_tmp->magic)), BTRFS_MAGIC, - sizeof(super_tmp->magic)) || + super_tmp->magic != cpu_to_le64(BTRFS_MAGIC) || memcmp(device->uuid, super_tmp->dev_item.uuid, BTRFS_UUID_SIZE) || btrfs_super_nodesize(super_tmp) != state->metablock_size || btrfs_super_leafsize(super_tmp) != state->metablock_size || diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 961ff2986341..1679051f4d39 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -46,7 +46,7 @@ extern struct kmem_cache *btrfs_path_cachep; extern struct kmem_cache *btrfs_free_space_cachep; struct btrfs_ordered_sum; -#define BTRFS_MAGIC "_BHRfS_M" +#define BTRFS_MAGIC 0x4D5F53665248425FULL /* ascii _BHRfS_M, no null */ #define BTRFS_MAX_MIRRORS 3 diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index e511d9f78c19..39ff34a62a24 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -2407,8 +2407,7 @@ int open_ctree(struct super_block *sb, sb->s_blocksize = sectorsize; sb->s_blocksize_bits = blksize_bits(sectorsize); - if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC, - sizeof(disk_super->magic))) { + if (disk_super->magic != cpu_to_le64(BTRFS_MAGIC)) { printk(KERN_INFO "btrfs: valid FS not found on %s\n", sb->s_id); goto fail_sb_buffer; } @@ -2815,8 +2814,7 @@ struct buffer_head *btrfs_read_dev_super(struct block_device *bdev) super = (struct btrfs_super_block *)bh->b_data; if (btrfs_super_bytenr(super) != bytenr || - strncmp((char *)(&super->magic), BTRFS_MAGIC, - sizeof(super->magic))) { + super->magic != cpu_to_le64(BTRFS_MAGIC)) { brelse(bh); continue; } diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index 567076542d1d..72b1cf1b2b5e 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -855,8 +855,7 @@ int btrfs_scan_one_device(const char *path, fmode_t flags, void *holder, disk_super = p + (bytenr & ~PAGE_CACHE_MASK); if (btrfs_super_bytenr(disk_super) != bytenr || - strncmp((char *)(&disk_super->magic), BTRFS_MAGIC, - sizeof(disk_super->magic))) + disk_super->magic != cpu_to_le64(BTRFS_MAGIC)) goto error_unmap; devid = btrfs_stack_device_id(&disk_super->dev_item); -- cgit v1.2.3 From 1cba0cdf5e4dbcd9e5fa5b54d7a028e55e2ca057 Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Wed, 20 Feb 2013 14:06:20 -0500 Subject: btrfs: Init io_lock after cloning btrfs device struct __btrfs_close_devices() clones btrfs device structs with memcpy(). Some of the fields in the clone are reinitialized, but it's missing to init io_lock. In mainline this goes unnoticed, but on RT it leaves the plist pointing to the original about to be freed lock struct. Initialize io_lock after cloning, so no references to the original struct are left. Reported-and-tested-by: Mike Galbraith Cc: stable@vger.kernel.org Signed-off-by: Thomas Gleixner Signed-off-by: Chris Mason --- fs/btrfs/volumes.c | 1 + 1 file changed, 1 insertion(+) (limited to 'fs/btrfs/volumes.c') diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index 7992dc4ea4cc..5d6010ba8b7e 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -650,6 +650,7 @@ static int __btrfs_close_devices(struct btrfs_fs_devices *fs_devices) new_device->writeable = 0; new_device->in_fs_metadata = 0; new_device->can_discard = 0; + spin_lock_init(&new_device->io_lock); list_replace_rcu(&device->dev_list, &new_device->dev_list); call_rcu(&device->rcu, free_device); -- cgit v1.2.3 From 86db25785a6e7cacc1ee868fe437e8a2957eae94 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 20 Feb 2013 16:23:40 -0500 Subject: Btrfs: fix max chunk size on raid5/6 We try to limit the size of a chunk to 10GB, which keeps the unit of work reasonable during balance and resize operations. The limit checks were taking into account the number of copies of the data we had but what they really should be doing is comparing against the logical size of the chunk we're creating. This moves the code around a little to use the count of data stripes from raid5/6. Signed-off-by: Chris Mason --- fs/btrfs/volumes.c | 25 +++++++++++++++++++++---- 1 file changed, 21 insertions(+), 4 deletions(-) (limited to 'fs/btrfs/volumes.c') diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index 5d6010ba8b7e..538c5cfa005f 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -3837,10 +3837,6 @@ static int __btrfs_alloc_chunk(struct btrfs_trans_handle *trans, */ 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)); @@ -3851,6 +3847,27 @@ static int __btrfs_alloc_chunk(struct btrfs_trans_handle *trans, btrfs_super_stripesize(info->super_copy)); data_stripes = num_stripes - 2; } + + /* + * Use the number of data stripes to figure out how big this chunk + * is really going to be in terms of logical address space, + * and compare that answer with the max chunk size + */ + if (stripe_size * data_stripes > max_chunk_size) { + u64 mask = (1ULL << 24) - 1; + stripe_size = max_chunk_size; + do_div(stripe_size, data_stripes); + + /* bump the answer up to a 16MB boundary */ + stripe_size = (stripe_size + mask) & ~mask; + + /* but don't go higher than the limits we found + * while searching for free extents + */ + if (stripe_size > devices_info[ndevs-1].max_avail) + stripe_size = devices_info[ndevs-1].max_avail; + } + do_div(stripe_size, dev_stripes); /* align to BTRFS_STRIPE_LEN */ -- cgit v1.2.3 From fda2832febb1928da0625b2c5d15559b29d7e740 Mon Sep 17 00:00:00 2001 From: Qu Wenruo Date: Tue, 26 Feb 2013 08:10:22 +0000 Subject: btrfs: cleanup for open-coded alignment Though most of the btrfs codes are using ALIGN macro for page alignment, there are still some codes using open-coded alignment like the following: ------ u64 mask = ((u64)root->stripesize - 1); u64 ret = (val + mask) & ~mask; ------ Or even hidden one: ------ num_bytes = (end - start + blocksize) & ~(blocksize - 1); ------ Sometimes these open-coded alignment is not so easy to understand for newbie like me. This commit changes the open-coded alignment to the ALIGN macro for a better readability. Also there is a previous patch from David Sterba with similar changes, but the patch is for 3.2 kernel and seems not merged. http://www.spinics.net/lists/linux-btrfs/msg12747.html Cc: David Sterba Signed-off-by: Qu Wenruo Signed-off-by: Josef Bacik --- fs/btrfs/extent-tree.c | 9 +++------ fs/btrfs/extent_io.c | 8 ++++---- fs/btrfs/file.c | 3 +-- fs/btrfs/inode.c | 37 +++++++++++++++---------------------- fs/btrfs/tree-log.c | 3 +-- fs/btrfs/volumes.c | 3 +-- 6 files changed, 25 insertions(+), 38 deletions(-) (limited to 'fs/btrfs/volumes.c') diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 8520354f086e..5681a91ed400 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -3431,7 +3431,7 @@ int btrfs_check_data_free_space(struct inode *inode, u64 bytes) int ret = 0, committed = 0, alloc_chunk = 1; /* make sure bytes are sectorsize aligned */ - bytes = (bytes + root->sectorsize - 1) & ~((u64)root->sectorsize - 1); + bytes = ALIGN(bytes, root->sectorsize); if (root == root->fs_info->tree_root || BTRFS_I(inode)->location.objectid == BTRFS_FREE_INO_OBJECTID) { @@ -3526,7 +3526,7 @@ void btrfs_free_reserved_data_space(struct inode *inode, u64 bytes) struct btrfs_space_info *data_sinfo; /* make sure bytes are sectorsize aligned */ - bytes = (bytes + root->sectorsize - 1) & ~((u64)root->sectorsize - 1); + bytes = ALIGN(bytes, root->sectorsize); data_sinfo = root->fs_info->data_sinfo; spin_lock(&data_sinfo->lock); @@ -5607,10 +5607,7 @@ static u64 stripe_align(struct btrfs_root *root, struct btrfs_block_group_cache *cache, u64 val, u64 num_bytes) { - u64 mask; - u64 ret; - mask = ((u64)root->stripesize - 1); - ret = (val + mask) & ~mask; + u64 ret = ALIGN(val, root->stripesize); return ret; } diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 66f999b97cbb..597ab8966c80 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -2686,7 +2686,7 @@ static int __extent_read_full_page(struct extent_io_tree *tree, iosize = min(extent_map_end(em) - cur, end - cur + 1); cur_end = min(extent_map_end(em) - 1, end); - iosize = (iosize + blocksize - 1) & ~((u64)blocksize - 1); + iosize = ALIGN(iosize, blocksize); if (this_bio_flag & EXTENT_BIO_COMPRESSED) { disk_io_size = em->block_len; sector = em->block_start >> 9; @@ -2977,7 +2977,7 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, BUG_ON(extent_map_end(em) <= cur); BUG_ON(end < cur); iosize = min(extent_map_end(em) - cur, end - cur + 1); - iosize = (iosize + blocksize - 1) & ~((u64)blocksize - 1); + iosize = ALIGN(iosize, blocksize); sector = (em->block_start + extent_offset) >> 9; bdev = em->bdev; block_start = em->block_start; @@ -3664,7 +3664,7 @@ int extent_invalidatepage(struct extent_io_tree *tree, u64 end = start + PAGE_CACHE_SIZE - 1; size_t blocksize = page->mapping->host->i_sb->s_blocksize; - start += (offset + blocksize - 1) & ~(blocksize - 1); + start += ALIGN(offset, blocksize); if (start > end) return 0; @@ -3783,7 +3783,7 @@ static struct extent_map *get_extent_skip_holes(struct inode *inode, len = last - offset; if (len == 0) break; - len = (len + sectorsize - 1) & ~(sectorsize - 1); + len = ALIGN(len, sectorsize); em = get_extent(inode, NULL, 0, offset, len, 0); if (IS_ERR_OR_NULL(em)) return em; diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 6e6dd8cdad92..83c790d84038 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -510,8 +510,7 @@ int btrfs_dirty_pages(struct btrfs_root *root, struct inode *inode, loff_t isize = i_size_read(inode); start_pos = pos & ~((u64)root->sectorsize - 1); - num_bytes = (write_bytes + pos - start_pos + - root->sectorsize - 1) & ~((u64)root->sectorsize - 1); + num_bytes = ALIGN(write_bytes + pos - start_pos, root->sectorsize); end_of_last_block = start_pos + num_bytes - 1; err = btrfs_set_extent_delalloc(inode, start_pos, end_of_last_block, diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index be09654e11b9..9ef7a5b1b77e 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -233,8 +233,7 @@ static noinline int cow_file_range_inline(struct btrfs_trans_handle *trans, u64 isize = i_size_read(inode); u64 actual_end = min(end + 1, isize); u64 inline_len = actual_end - start; - u64 aligned_end = (end + root->sectorsize - 1) & - ~((u64)root->sectorsize - 1); + u64 aligned_end = ALIGN(end, root->sectorsize); u64 data_len = inline_len; int ret; @@ -391,7 +390,7 @@ again: * a compressed extent to 128k. */ total_compressed = min(total_compressed, max_uncompressed); - num_bytes = (end - start + blocksize) & ~(blocksize - 1); + num_bytes = ALIGN(end - start + 1, blocksize); num_bytes = max(blocksize, num_bytes); total_in = 0; ret = 0; @@ -490,15 +489,13 @@ cont: * up to a block size boundary so the allocator does sane * things */ - total_compressed = (total_compressed + blocksize - 1) & - ~(blocksize - 1); + total_compressed = ALIGN(total_compressed, blocksize); /* * one last check to make sure the compression is really a * win, compare the page count read with the blocks on disk */ - total_in = (total_in + PAGE_CACHE_SIZE - 1) & - ~(PAGE_CACHE_SIZE - 1); + total_in = ALIGN(total_in, PAGE_CACHE_SIZE); if (total_compressed >= total_in) { will_compress = 0; } else { @@ -856,7 +853,7 @@ static noinline int __cow_file_range(struct btrfs_trans_handle *trans, BUG_ON(btrfs_is_free_space_inode(inode)); - num_bytes = (end - start + blocksize) & ~(blocksize - 1); + num_bytes = ALIGN(end - start + 1, blocksize); num_bytes = max(blocksize, num_bytes); disk_num_bytes = num_bytes; @@ -4015,7 +4012,6 @@ int btrfs_truncate_inode_items(struct btrfs_trans_handle *trans, u64 extent_num_bytes = 0; u64 extent_offset = 0; u64 item_end = 0; - u64 mask = root->sectorsize - 1; u32 found_type = (u8)-1; int found_extent; int del_item; @@ -4039,7 +4035,8 @@ int btrfs_truncate_inode_items(struct btrfs_trans_handle *trans, * extent just the way it is. */ if (root->ref_cows || root == root->fs_info->tree_root) - btrfs_drop_extent_cache(inode, (new_size + mask) & (~mask), (u64)-1, 0); + btrfs_drop_extent_cache(inode, ALIGN(new_size, + root->sectorsize), (u64)-1, 0); /* * This function is also used to drop the items in the log tree before @@ -4118,10 +4115,9 @@ search_again: if (!del_item) { u64 orig_num_bytes = btrfs_file_extent_num_bytes(leaf, fi); - extent_num_bytes = new_size - - found_key.offset + root->sectorsize - 1; - extent_num_bytes = extent_num_bytes & - ~((u64)root->sectorsize - 1); + extent_num_bytes = ALIGN(new_size - + found_key.offset, + root->sectorsize); btrfs_set_file_extent_num_bytes(leaf, fi, extent_num_bytes); num_dec = (orig_num_bytes - @@ -4357,9 +4353,8 @@ int btrfs_cont_expand(struct inode *inode, loff_t oldsize, loff_t size) struct extent_map *em = NULL; struct extent_state *cached_state = NULL; struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; - u64 mask = root->sectorsize - 1; - u64 hole_start = (oldsize + mask) & ~mask; - u64 block_end = (size + mask) & ~mask; + u64 hole_start = ALIGN(oldsize, root->sectorsize); + u64 block_end = ALIGN(size, root->sectorsize); u64 last_byte; u64 cur_offset; u64 hole_size; @@ -4392,7 +4387,7 @@ int btrfs_cont_expand(struct inode *inode, loff_t oldsize, loff_t size) break; } last_byte = min(extent_map_end(em), block_end); - last_byte = (last_byte + mask) & ~mask; + last_byte = ALIGN(last_byte , root->sectorsize); if (!test_bit(EXTENT_FLAG_PREALLOC, &em->flags)) { struct extent_map *hole_em; hole_size = last_byte - cur_offset; @@ -6111,8 +6106,7 @@ again: } else if (found_type == BTRFS_FILE_EXTENT_INLINE) { size_t size; size = btrfs_file_extent_inline_len(leaf, item); - extent_end = (extent_start + size + root->sectorsize - 1) & - ~((u64)root->sectorsize - 1); + extent_end = ALIGN(extent_start + size, root->sectorsize); } if (start >= extent_end) { @@ -6184,8 +6178,7 @@ again: copy_size = min_t(u64, PAGE_CACHE_SIZE - pg_offset, size - extent_offset); em->start = extent_start + extent_offset; - em->len = (copy_size + root->sectorsize - 1) & - ~((u64)root->sectorsize - 1); + em->len = ALIGN(copy_size, root->sectorsize); em->orig_block_len = em->len; em->orig_start = em->start; if (compress_type) { diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c index 1a79087c4575..e8b7a68e1b37 100644 --- a/fs/btrfs/tree-log.c +++ b/fs/btrfs/tree-log.c @@ -484,7 +484,6 @@ static noinline int replay_one_extent(struct btrfs_trans_handle *trans, struct btrfs_key *key) { int found_type; - u64 mask = root->sectorsize - 1; u64 extent_end; u64 start = key->offset; u64 saved_nbytes; @@ -501,7 +500,7 @@ static noinline int replay_one_extent(struct btrfs_trans_handle *trans, extent_end = start + btrfs_file_extent_num_bytes(eb, item); else if (found_type == BTRFS_FILE_EXTENT_INLINE) { size = btrfs_file_extent_inline_len(eb, item); - extent_end = (start + size + mask) & ~mask; + extent_end = ALIGN(start + size, root->sectorsize); } else { ret = 0; goto out; diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index 538c5cfa005f..db72e0cc6f87 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -4556,8 +4556,7 @@ static int __btrfs_map_block(struct btrfs_fs_info *fs_info, int rw, num_stripes = 1; stripe_index = 0; stripe_nr_orig = stripe_nr; - stripe_nr_end = (offset + *length + map->stripe_len - 1) & - (~(map->stripe_len - 1)); + stripe_nr_end = ALIGN(offset + *length, map->stripe_len); do_div(stripe_nr_end, map->stripe_len); stripe_end_offset = stripe_nr_end * map->stripe_len - (offset + *length); -- cgit v1.2.3 From 2d8946c59743bc635a5e5701e6ef8e71e0a16ab7 Mon Sep 17 00:00:00 2001 From: David Sterba Date: Wed, 27 Feb 2013 23:13:55 +0000 Subject: btrfs: remove a printk from scan_one_device Dave pointed out that he saw messages from btrfs although there was no such filesystem on his computers. The automatic device scan is called on every new blockdevice if the usual distro udev rule set is used. The printk introduced in 6f60cbd3ae442c was a remainder from copying portions of code from btrfs_get_bdev_and_sb which is used under different conditions and the warning makes sense there. Reported-by: Dave Chinner Signed-off-by: David Sterba Signed-off-by: Josef Bacik --- fs/btrfs/volumes.c | 1 - 1 file changed, 1 deletion(-) (limited to 'fs/btrfs/volumes.c') diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index db72e0cc6f87..35bb2d4ed29f 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -829,7 +829,6 @@ int btrfs_scan_one_device(const char *path, fmode_t flags, void *holder, if (IS_ERR(bdev)) { ret = PTR_ERR(bdev); - printk(KERN_INFO "btrfs: open %s failed\n", path); goto error; } -- cgit v1.2.3 From 0f788c58194e4ccc5b3ab23f872c5e18542335e4 Mon Sep 17 00:00:00 2001 From: Liu Bo Date: Mon, 4 Mar 2013 16:25:40 +0000 Subject: Btrfs: do not BUG_ON on aborted situation Btrfs balance can easily hit BUG_ON in these places, but we want to it bail out gracefully after we force the whole filesystem to readonly. So we use btrfs_std_error hook in place of BUG_ON. Signed-off-by: Liu Bo Signed-off-by: Josef Bacik --- fs/btrfs/relocation.c | 6 +++++- fs/btrfs/volumes.c | 9 +++++++-- 2 files changed, 12 insertions(+), 3 deletions(-) (limited to 'fs/btrfs/volumes.c') diff --git a/fs/btrfs/relocation.c b/fs/btrfs/relocation.c index 9d13786eec73..3ebe87977aae 100644 --- a/fs/btrfs/relocation.c +++ b/fs/btrfs/relocation.c @@ -3771,7 +3771,11 @@ static noinline_for_stack int relocate_block_group(struct reloc_control *rc) while (1) { progress++; trans = btrfs_start_transaction(rc->extent_root, 0); - BUG_ON(IS_ERR(trans)); + if (IS_ERR(trans)) { + err = PTR_ERR(trans); + trans = NULL; + break; + } restart: if (update_backref_cache(trans, &rc->backref_cache)) { btrfs_end_transaction(trans, rc->extent_root); diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index 35bb2d4ed29f..9ff454df6756 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -2379,7 +2379,11 @@ static int btrfs_relocate_chunk(struct btrfs_root *root, return ret; trans = btrfs_start_transaction(root, 0); - BUG_ON(IS_ERR(trans)); + if (IS_ERR(trans)) { + ret = PTR_ERR(trans); + btrfs_std_error(root->fs_info, ret); + return ret; + } lock_chunks(root); @@ -3050,7 +3054,8 @@ static void __cancel_balance(struct btrfs_fs_info *fs_info) unset_balance_control(fs_info); ret = del_balance_item(fs_info->tree_root); - BUG_ON(ret); + if (ret) + btrfs_std_error(fs_info, ret); atomic_set(&fs_info->mutually_exclusive_operation_running, 0); } -- cgit v1.2.3 From 3a01aa7a25274308fe813a6237f678aed901cea3 Mon Sep 17 00:00:00 2001 From: Ilya Dryomov Date: Wed, 6 Mar 2013 01:57:55 -0700 Subject: Btrfs: fix a mismerge in btrfs_balance() Raid56 merge (merge commit e942f88) had mistakenly removed a call to __cancel_balance(), which resulted in balance not cleaning up after itself after a successful finish. (Cleanup includes switching the state, removing the balance item and releasing mut_ex_op testnset lock.) Bring it back. Reported-by: David Sterba Signed-off-by: Ilya Dryomov Signed-off-by: Chris Mason --- fs/btrfs/volumes.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'fs/btrfs/volumes.c') diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index 9ff454df6756..6b9cff42265d 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -3235,6 +3235,11 @@ int btrfs_balance(struct btrfs_balance_control *bctl, update_ioctl_balance_args(fs_info, 0, bargs); } + if ((ret && ret != -ECANCELED && ret != -ENOSPC) || + balance_need_close(fs_info)) { + __cancel_balance(fs_info); + } + wake_up(&fs_info->balance_wait_q); return ret; -- cgit v1.2.3 From bc178622d40d87e75abc131007342429c9b03351 Mon Sep 17 00:00:00 2001 From: Eric Sandeen Date: Sat, 9 Mar 2013 15:18:39 +0000 Subject: btrfs: use rcu_barrier() to wait for bdev puts at unmount Doing this would reliably fail with -EBUSY for me: # mount /dev/sdb2 /mnt/scratch; umount /mnt/scratch; mkfs.btrfs -f /dev/sdb2 ... unable to open /dev/sdb2: Device or resource busy because mkfs.btrfs tries to open the device O_EXCL, and somebody still has it. Using systemtap to track bdev gets & puts shows a kworker thread doing a blkdev put after mkfs attempts a get; this is left over from the unmount path: btrfs_close_devices __btrfs_close_devices call_rcu(&device->rcu, free_device); free_device INIT_WORK(&device->rcu_work, __free_device); schedule_work(&device->rcu_work); so unmount might complete before __free_device fires & does its blkdev_put. Adding an rcu_barrier() to btrfs_close_devices() causes unmount to wait until all blkdev_put()s are done, and the device is truly free once unmount completes. Cc: stable@vger.kernel.org Signed-off-by: Eric Sandeen Signed-off-by: Josef Bacik Signed-off-by: Chris Mason --- fs/btrfs/volumes.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'fs/btrfs/volumes.c') diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index 6b9cff42265d..5989a92236f7 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -684,6 +684,12 @@ int btrfs_close_devices(struct btrfs_fs_devices *fs_devices) __btrfs_close_devices(fs_devices); free_fs_devices(fs_devices); } + /* + * Wait for rcu kworkers under __btrfs_close_devices + * to finish all blkdev_puts so device is really + * free when umount is done. + */ + rcu_barrier(); return ret; } -- cgit v1.2.3