X-Git-Url: http://ftp.safe.ca/?a=blobdiff_plain;f=fs%2Fbtrfs%2Fextent_io.c;h=96577e8bf9fdb62819ab2dbd5f9da91200624596;hb=ba230c3f6dc88ec008806adb27b12088486d508e;hp=3153b4fdd0b375227b9ae97d55474edab5cc4cee;hpb=211f90e68b679d27fe23c5505f86d6ce62c98bae;p=safe%2Fjmp%2Flinux-2.6 diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 3153b4f..96577e8 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -9,24 +9,24 @@ #include #include #include -#include #include #include #include "extent_io.h" #include "extent_map.h" - -/* temporary define until extent_map moves out of btrfs */ -struct kmem_cache *btrfs_cache_create(const char *name, size_t size, - unsigned long extra_flags, - void (*ctor)(void *, struct kmem_cache *, - unsigned long)); +#include "compat.h" +#include "ctree.h" +#include "btrfs_inode.h" static struct kmem_cache *extent_state_cache; static struct kmem_cache *extent_buffer_cache; static LIST_HEAD(buffers); static LIST_HEAD(states); -static spinlock_t leak_lock = SPIN_LOCK_UNLOCKED; + +#define LEAK_DEBUG 0 +#if LEAK_DEBUG +static DEFINE_SPINLOCK(leak_lock); +#endif #define BUFFER_LRU_MAX 64 @@ -40,19 +40,27 @@ struct extent_page_data { struct bio *bio; struct extent_io_tree *tree; get_extent_t *get_extent; + + /* tells writepage not to lock the state bits for this range + * it still does the unlocking + */ + unsigned int extent_locked:1; + + /* tells the submit_bio code to use a WRITE_SYNC */ + unsigned int sync_io:1; }; int __init extent_io_init(void) { - extent_state_cache = btrfs_cache_create("extent_state", - sizeof(struct extent_state), 0, - NULL); + extent_state_cache = kmem_cache_create("extent_state", + sizeof(struct extent_state), 0, + SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL); if (!extent_state_cache) return -ENOMEM; - extent_buffer_cache = btrfs_cache_create("extent_buffers", - sizeof(struct extent_buffer), 0, - NULL); + extent_buffer_cache = kmem_cache_create("extent_buffers", + sizeof(struct extent_buffer), 0, + SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL); if (!extent_buffer_cache) goto free_state_cache; return 0; @@ -69,7 +77,11 @@ void extent_io_exit(void) while (!list_empty(&states)) { state = list_entry(states.next, struct extent_state, leak_list); - printk("state leak: start %Lu end %Lu state %lu in tree %p refs %d\n", state->start, state->end, state->state, state->tree, atomic_read(&state->refs)); + printk(KERN_ERR "btrfs state leak: start %llu end %llu " + "state %lu in tree %p refs %d\n", + (unsigned long long)state->start, + (unsigned long long)state->end, + state->state, state->tree, atomic_read(&state->refs)); list_del(&state->leak_list); kmem_cache_free(extent_state_cache, state); @@ -77,7 +89,9 @@ void extent_io_exit(void) while (!list_empty(&buffers)) { eb = list_entry(buffers.next, struct extent_buffer, leak_list); - printk("buffer leak start %Lu len %lu refs %d\n", eb->start, eb->len, atomic_read(&eb->refs)); + printk(KERN_ERR "btrfs buffer leak start %llu len %lu " + "refs %d\n", (unsigned long long)eb->start, + eb->len, atomic_read(&eb->refs)); list_del(&eb->leak_list); kmem_cache_free(extent_buffer_cache, eb); } @@ -91,33 +105,20 @@ void extent_io_tree_init(struct extent_io_tree *tree, struct address_space *mapping, gfp_t mask) { tree->state.rb_node = NULL; + tree->buffer.rb_node = NULL; tree->ops = NULL; tree->dirty_bytes = 0; spin_lock_init(&tree->lock); - spin_lock_init(&tree->lru_lock); + spin_lock_init(&tree->buffer_lock); tree->mapping = mapping; - INIT_LIST_HEAD(&tree->buffer_lru); - tree->lru_size = 0; - tree->last = NULL; -} -EXPORT_SYMBOL(extent_io_tree_init); - -void extent_io_tree_empty_lru(struct extent_io_tree *tree) -{ - struct extent_buffer *eb; - while(!list_empty(&tree->buffer_lru)) { - eb = list_entry(tree->buffer_lru.next, struct extent_buffer, - lru); - list_del_init(&eb->lru); - free_extent_buffer(eb); - } } -EXPORT_SYMBOL(extent_io_tree_empty_lru); -struct extent_state *alloc_extent_state(gfp_t mask) +static struct extent_state *alloc_extent_state(gfp_t mask) { struct extent_state *state; +#if LEAK_DEBUG unsigned long flags; +#endif state = kmem_cache_alloc(extent_state_cache, mask); if (!state) @@ -125,39 +126,42 @@ struct extent_state *alloc_extent_state(gfp_t mask) state->state = 0; state->private = 0; state->tree = NULL; +#if LEAK_DEBUG spin_lock_irqsave(&leak_lock, flags); list_add(&state->leak_list, &states); spin_unlock_irqrestore(&leak_lock, flags); - +#endif atomic_set(&state->refs, 1); init_waitqueue_head(&state->wq); return state; } -EXPORT_SYMBOL(alloc_extent_state); -void free_extent_state(struct extent_state *state) +static void free_extent_state(struct extent_state *state) { if (!state) return; if (atomic_dec_and_test(&state->refs)) { +#if LEAK_DEBUG unsigned long flags; +#endif WARN_ON(state->tree); +#if LEAK_DEBUG spin_lock_irqsave(&leak_lock, flags); list_del(&state->leak_list); spin_unlock_irqrestore(&leak_lock, flags); +#endif kmem_cache_free(extent_state_cache, state); } } -EXPORT_SYMBOL(free_extent_state); static struct rb_node *tree_insert(struct rb_root *root, u64 offset, struct rb_node *node) { - struct rb_node ** p = &root->rb_node; - struct rb_node * parent = NULL; + struct rb_node **p = &root->rb_node; + struct rb_node *parent = NULL; struct tree_entry *entry; - while(*p) { + while (*p) { parent = *p; entry = rb_entry(parent, struct tree_entry, rb_node); @@ -180,19 +184,13 @@ static struct rb_node *__etree_search(struct extent_io_tree *tree, u64 offset, struct rb_node **next_ret) { struct rb_root *root = &tree->state; - struct rb_node * n = root->rb_node; + struct rb_node *n = root->rb_node; struct rb_node *prev = NULL; struct rb_node *orig_prev = NULL; struct tree_entry *entry; struct tree_entry *prev_entry = NULL; - if (tree->last) { - struct extent_state *state; - state = tree->last; - if (state->start <= offset && offset <= state->end) - return &tree->last->rb_node; - } - while(n) { + while (n) { entry = rb_entry(n, struct tree_entry, rb_node); prev = n; prev_entry = entry; @@ -201,15 +199,13 @@ static struct rb_node *__etree_search(struct extent_io_tree *tree, u64 offset, n = n->rb_left; else if (offset > entry->end) n = n->rb_right; - else { - tree->last = rb_entry(n, struct extent_state, rb_node); + else return n; - } } if (prev_ret) { orig_prev = prev; - while(prev && offset > prev_entry->end) { + while (prev && offset > prev_entry->end) { prev = rb_next(prev); prev_entry = rb_entry(prev, struct tree_entry, rb_node); } @@ -219,7 +215,7 @@ static struct rb_node *__etree_search(struct extent_io_tree *tree, u64 offset, if (next_ret) { prev_entry = rb_entry(prev, struct tree_entry, rb_node); - while(prev && offset < prev_entry->start) { + while (prev && offset < prev_entry->start) { prev = rb_prev(prev); prev_entry = rb_entry(prev, struct tree_entry, rb_node); } @@ -235,16 +231,63 @@ static inline struct rb_node *tree_search(struct extent_io_tree *tree, struct rb_node *ret; ret = __etree_search(tree, offset, &prev, NULL); - if (!ret) { - if (prev) { - tree->last = rb_entry(prev, struct extent_state, - rb_node); - } + if (!ret) return prev; - } return ret; } +static struct extent_buffer *buffer_tree_insert(struct extent_io_tree *tree, + u64 offset, struct rb_node *node) +{ + struct rb_root *root = &tree->buffer; + struct rb_node **p = &root->rb_node; + struct rb_node *parent = NULL; + struct extent_buffer *eb; + + while (*p) { + parent = *p; + eb = rb_entry(parent, struct extent_buffer, rb_node); + + if (offset < eb->start) + p = &(*p)->rb_left; + else if (offset > eb->start) + p = &(*p)->rb_right; + else + return eb; + } + + rb_link_node(node, parent, p); + rb_insert_color(node, root); + return NULL; +} + +static struct extent_buffer *buffer_search(struct extent_io_tree *tree, + u64 offset) +{ + struct rb_root *root = &tree->buffer; + struct rb_node *n = root->rb_node; + struct extent_buffer *eb; + + while (n) { + eb = rb_entry(n, struct extent_buffer, rb_node); + if (offset < eb->start) + n = n->rb_left; + else if (offset > eb->start) + n = n->rb_right; + else + return eb; + } + return NULL; +} + +static void merge_cb(struct extent_io_tree *tree, struct extent_state *new, + struct extent_state *other) +{ + if (tree->ops && tree->ops->merge_extent_hook) + tree->ops->merge_extent_hook(tree->mapping->host, new, + other); +} + /* * utility function to look for merge candidates inside a given range. * Any extents with matching state are merged together into a single @@ -260,7 +303,7 @@ static int merge_state(struct extent_io_tree *tree, struct extent_state *other; struct rb_node *other_node; - if (state->state & EXTENT_IOBITS) + if (state->state & (EXTENT_IOBITS | EXTENT_BOUNDARY)) return 0; other_node = rb_prev(&state->rb_node); @@ -268,10 +311,9 @@ static int merge_state(struct extent_io_tree *tree, other = rb_entry(other_node, struct extent_state, rb_node); if (other->end == state->start - 1 && other->state == state->state) { + merge_cb(tree, state, other); state->start = other->start; other->tree = NULL; - if (tree->last == other) - tree->last = state; rb_erase(&other->rb_node, &tree->state); free_extent_state(other); } @@ -281,35 +323,37 @@ static int merge_state(struct extent_io_tree *tree, other = rb_entry(other_node, struct extent_state, rb_node); if (other->start == state->end + 1 && other->state == state->state) { + merge_cb(tree, state, other); other->start = state->start; state->tree = NULL; - if (tree->last == state) - tree->last = other; rb_erase(&state->rb_node, &tree->state); free_extent_state(state); + state = NULL; } } + return 0; } -static void set_state_cb(struct extent_io_tree *tree, +static int set_state_cb(struct extent_io_tree *tree, struct extent_state *state, unsigned long bits) { if (tree->ops && tree->ops->set_bit_hook) { - tree->ops->set_bit_hook(tree->mapping->host, state->start, - state->end, state->state, bits); + return tree->ops->set_bit_hook(tree->mapping->host, + state->start, state->end, + state->state, bits); } + + return 0; } static void clear_state_cb(struct extent_io_tree *tree, struct extent_state *state, unsigned long bits) { - if (tree->ops && tree->ops->set_bit_hook) { - tree->ops->clear_bit_hook(tree->mapping->host, state->start, - state->end, state->state, bits); - } + if (tree->ops && tree->ops->clear_bit_hook) + tree->ops->clear_bit_hook(tree->mapping->host, state, bits); } /* @@ -327,31 +371,48 @@ static int insert_state(struct extent_io_tree *tree, int bits) { struct rb_node *node; + int ret; if (end < start) { - printk("end < start %Lu %Lu\n", end, start); + printk(KERN_ERR "btrfs end < start %llu %llu\n", + (unsigned long long)end, + (unsigned long long)start); WARN_ON(1); } + state->start = start; + state->end = end; + ret = set_state_cb(tree, state, bits); + if (ret) + return ret; + if (bits & EXTENT_DIRTY) tree->dirty_bytes += end - start + 1; - set_state_cb(tree, state, bits); state->state |= bits; - state->start = start; - state->end = end; node = tree_insert(&tree->state, end, &state->rb_node); if (node) { struct extent_state *found; found = rb_entry(node, struct extent_state, rb_node); - printk("found node %Lu %Lu on insert of %Lu %Lu\n", found->start, found->end, start, end); + printk(KERN_ERR "btrfs found node %llu %llu on insert of " + "%llu %llu\n", (unsigned long long)found->start, + (unsigned long long)found->end, + (unsigned long long)start, (unsigned long long)end); free_extent_state(state); return -EEXIST; } state->tree = tree; - tree->last = state; merge_state(tree, state); return 0; } +static int split_cb(struct extent_io_tree *tree, struct extent_state *orig, + u64 split) +{ + if (tree->ops && tree->ops->split_extent_hook) + return tree->ops->split_extent_hook(tree->mapping->host, + orig, split); + return 0; +} + /* * split a given extent state struct in two, inserting the preallocated * struct 'prealloc' as the newly created second half. 'split' indicates an @@ -370,6 +431,9 @@ static int split_state(struct extent_io_tree *tree, struct extent_state *orig, struct extent_state *prealloc, u64 split) { struct rb_node *node; + + split_cb(tree, orig, split); + prealloc->start = orig->start; prealloc->end = split - 1; prealloc->state = orig->state; @@ -377,9 +441,6 @@ static int split_state(struct extent_io_tree *tree, struct extent_state *orig, node = tree_insert(&tree->state, prealloc->end, &prealloc->rb_node); if (node) { - struct extent_state *found; - found = rb_entry(node, struct extent_state, rb_node); - printk("found node %Lu %Lu on insert of %Lu %Lu\n", found->start, found->end, prealloc->start, prealloc->end); free_extent_state(prealloc); return -EEXIST; } @@ -399,7 +460,8 @@ static int clear_state_bit(struct extent_io_tree *tree, struct extent_state *state, int bits, int wake, int delete) { - int ret = state->state & bits; + int bits_to_clear = bits & ~EXTENT_DO_ACCOUNTING; + int ret = state->state & bits_to_clear; if ((bits & EXTENT_DIRTY) && (state->state & EXTENT_DIRTY)) { u64 range = state->end - state->start + 1; @@ -407,15 +469,12 @@ static int clear_state_bit(struct extent_io_tree *tree, tree->dirty_bytes -= range; } clear_state_cb(tree, state, bits); - state->state &= ~bits; + state->state &= ~bits_to_clear; if (wake) wake_up(&state->wq); if (delete || state->state == 0) { if (state->tree) { clear_state_cb(tree, state, state->state); - if (tree->last == state) { - tree->last = extent_state_next(state); - } rb_erase(&state->rb_node, &tree->state); state->tree = NULL; free_extent_state(state); @@ -442,12 +501,16 @@ static int clear_state_bit(struct extent_io_tree *tree, * bits were already set, or zero if none of the bits were already set. */ int clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, - int bits, int wake, int delete, gfp_t mask) + int bits, int wake, int delete, + struct extent_state **cached_state, + gfp_t mask) { struct extent_state *state; + struct extent_state *cached; struct extent_state *prealloc = NULL; + struct rb_node *next_node; struct rb_node *node; - unsigned long flags; + u64 last_end; int err; int set = 0; @@ -458,7 +521,18 @@ again: return -ENOMEM; } - spin_lock_irqsave(&tree->lock, flags); + spin_lock(&tree->lock); + if (cached_state) { + cached = *cached_state; + *cached_state = NULL; + cached_state = NULL; + if (cached && cached->tree && cached->start == start) { + atomic_dec(&cached->refs); + state = cached; + goto hit_next; + } + free_extent_state(cached); + } /* * this search will find the extents that end after * our range starts @@ -467,9 +541,11 @@ again: if (!node) goto out; state = rb_entry(node, struct extent_state, rb_node); +hit_next: if (state->start > end) goto out; WARN_ON(state->end < start); + last_end = state->end; /* * | ---- desired range ---- | @@ -496,11 +572,11 @@ again: if (err) goto out; if (state->end <= end) { - start = state->end + 1; - set |= clear_state_bit(tree, state, bits, - wake, delete); - } else { - start = state->start; + set |= clear_state_bit(tree, state, bits, wake, + delete); + if (last_end == (u64)-1) + goto out; + start = last_end + 1; } goto search_again; } @@ -515,21 +591,34 @@ again: prealloc = alloc_extent_state(GFP_ATOMIC); err = split_state(tree, state, prealloc, end + 1); BUG_ON(err == -EEXIST); - if (wake) wake_up(&state->wq); - set |= clear_state_bit(tree, prealloc, bits, - wake, delete); + + set |= clear_state_bit(tree, prealloc, bits, wake, delete); + prealloc = NULL; goto out; } - start = state->end + 1; + if (state->end < end && prealloc && !need_resched()) + next_node = rb_next(&state->rb_node); + else + next_node = NULL; + set |= clear_state_bit(tree, state, bits, wake, delete); + if (last_end == (u64)-1) + goto out; + start = last_end + 1; + if (start <= end && next_node) { + state = rb_entry(next_node, struct extent_state, + rb_node); + if (state->start == start) + goto hit_next; + } goto search_again; out: - spin_unlock_irqrestore(&tree->lock, flags); + spin_unlock(&tree->lock); if (prealloc) free_extent_state(prealloc); @@ -538,21 +627,22 @@ out: search_again: if (start > end) goto out; - spin_unlock_irqrestore(&tree->lock, flags); + spin_unlock(&tree->lock); if (mask & __GFP_WAIT) cond_resched(); goto again; } -EXPORT_SYMBOL(clear_extent_bit); static int wait_on_state(struct extent_io_tree *tree, struct extent_state *state) + __releases(tree->lock) + __acquires(tree->lock) { DEFINE_WAIT(wait); prepare_to_wait(&state->wq, &wait, TASK_UNINTERRUPTIBLE); - spin_unlock_irq(&tree->lock); + spin_unlock(&tree->lock); schedule(); - spin_lock_irq(&tree->lock); + spin_lock(&tree->lock); finish_wait(&state->wq, &wait); return 0; } @@ -567,7 +657,7 @@ int wait_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, int bits) struct extent_state *state; struct rb_node *node; - spin_lock_irq(&tree->lock); + spin_lock(&tree->lock); again: while (1) { /* @@ -596,51 +686,69 @@ again: break; if (need_resched()) { - spin_unlock_irq(&tree->lock); + spin_unlock(&tree->lock); cond_resched(); - spin_lock_irq(&tree->lock); + spin_lock(&tree->lock); } } out: - spin_unlock_irq(&tree->lock); + spin_unlock(&tree->lock); return 0; } -EXPORT_SYMBOL(wait_extent_bit); -static void set_state_bits(struct extent_io_tree *tree, +static int set_state_bits(struct extent_io_tree *tree, struct extent_state *state, int bits) { + int ret; + + ret = set_state_cb(tree, state, bits); + if (ret) + return ret; + if ((bits & EXTENT_DIRTY) && !(state->state & EXTENT_DIRTY)) { u64 range = state->end - state->start + 1; tree->dirty_bytes += range; } - set_state_cb(tree, state, bits); state->state |= bits; + + return 0; +} + +static void cache_state(struct extent_state *state, + struct extent_state **cached_ptr) +{ + if (cached_ptr && !(*cached_ptr)) { + if (state->state & (EXTENT_IOBITS | EXTENT_BOUNDARY)) { + *cached_ptr = state; + atomic_inc(&state->refs); + } + } } /* - * set some bits on a range in the tree. This may require allocations - * or sleeping, so the gfp mask is used to indicate what is allowed. + * set some bits on a range in the tree. This may require allocations or + * sleeping, so the gfp mask is used to indicate what is allowed. * - * If 'exclusive' == 1, this will fail with -EEXIST if some part of the - * range already has the desired bits set. The start of the existing - * range is returned in failed_start in this case. + * If any of the exclusive bits are set, this will fail with -EEXIST if some + * part of the range already has the desired bits set. The start of the + * existing range is returned in failed_start in this case. * - * [start, end] is inclusive - * This takes the tree lock. + * [start, end] is inclusive This takes the tree lock. */ -int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, int bits, - int exclusive, u64 *failed_start, gfp_t mask) + +static int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, + int bits, int exclusive_bits, u64 *failed_start, + struct extent_state **cached_state, + gfp_t mask) { struct extent_state *state; struct extent_state *prealloc = NULL; struct rb_node *node; - unsigned long flags; int err = 0; - int set; u64 last_start; u64 last_end; + again: if (!prealloc && (mask & __GFP_WAIT)) { prealloc = alloc_extent_state(mask); @@ -648,7 +756,14 @@ again: return -ENOMEM; } - spin_lock_irqsave(&tree->lock, flags); + spin_lock(&tree->lock); + if (cached_state && *cached_state) { + state = *cached_state; + if (state->start == start && state->tree) { + node = &state->rb_node; + goto hit_next; + } + } /* * this search will find all the extents that end after * our range starts. @@ -660,8 +775,8 @@ again: BUG_ON(err == -EEXIST); goto out; } - state = rb_entry(node, struct extent_state, rb_node); +hit_next: last_start = state->start; last_end = state->end; @@ -672,15 +787,32 @@ again: * Just lock what we found and keep going */ if (state->start == start && state->end <= end) { - set = state->state & bits; - if (set && exclusive) { + struct rb_node *next_node; + if (state->state & exclusive_bits) { *failed_start = state->start; err = -EEXIST; goto out; } - set_state_bits(tree, state, bits); - start = state->end + 1; + + err = set_state_bits(tree, state, bits); + if (err) + goto out; + + cache_state(state, cached_state); merge_state(tree, state); + if (last_end == (u64)-1) + goto out; + + start = last_end + 1; + if (start < end && prealloc && !need_resched()) { + next_node = rb_next(node); + if (next_node) { + state = rb_entry(next_node, struct extent_state, + rb_node); + if (state->start == start) + goto hit_next; + } + } goto search_again; } @@ -701,8 +833,7 @@ again: * desired bit on it. */ if (state->start < start) { - set = state->state & bits; - if (exclusive && set) { + if (state->state & exclusive_bits) { *failed_start = start; err = -EEXIST; goto out; @@ -713,11 +844,14 @@ again: if (err) goto out; if (state->end <= end) { - set_state_bits(tree, state, bits); - start = state->end + 1; + err = set_state_bits(tree, state, bits); + if (err) + goto out; + cache_state(state, cached_state); merge_state(tree, state); - } else { - start = state->start; + if (last_end == (u64)-1) + goto out; + start = last_end + 1; } goto search_again; } @@ -733,13 +867,16 @@ again: if (end < last_start) this_end = end; else - this_end = last_start -1; + this_end = last_start - 1; err = insert_state(tree, prealloc, start, this_end, bits); - prealloc = NULL; BUG_ON(err == -EEXIST); - if (err) + if (err) { + prealloc = NULL; goto out; + } + cache_state(prealloc, cached_state); + prealloc = NULL; start = this_end + 1; goto search_again; } @@ -750,8 +887,7 @@ again: * on the first half */ if (state->start <= end && state->end > end) { - set = state->state & bits; - if (exclusive && set) { + if (state->state & exclusive_bits) { *failed_start = start; err = -EEXIST; goto out; @@ -759,7 +895,12 @@ again: err = split_state(tree, state, prealloc, end + 1); BUG_ON(err == -EEXIST); - set_state_bits(tree, prealloc, bits); + err = set_state_bits(tree, prealloc, bits); + if (err) { + prealloc = NULL; + goto out; + } + cache_state(prealloc, cached_state); merge_state(tree, prealloc); prealloc = NULL; goto out; @@ -768,7 +909,7 @@ again: goto search_again; out: - spin_unlock_irqrestore(&tree->lock, flags); + spin_unlock(&tree->lock); if (prealloc) free_extent_state(prealloc); @@ -777,126 +918,96 @@ out: search_again: if (start > end) goto out; - spin_unlock_irqrestore(&tree->lock, flags); + spin_unlock(&tree->lock); if (mask & __GFP_WAIT) cond_resched(); goto again; } -EXPORT_SYMBOL(set_extent_bit); /* wrappers around set/clear extent bit */ int set_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask) { return set_extent_bit(tree, start, end, EXTENT_DIRTY, 0, NULL, - mask); -} -EXPORT_SYMBOL(set_extent_dirty); - -int set_extent_ordered(struct extent_io_tree *tree, u64 start, u64 end, - gfp_t mask) -{ - return set_extent_bit(tree, start, end, EXTENT_ORDERED, 0, NULL, mask); + NULL, mask); } -EXPORT_SYMBOL(set_extent_ordered); int set_extent_bits(struct extent_io_tree *tree, u64 start, u64 end, int bits, gfp_t mask) { return set_extent_bit(tree, start, end, bits, 0, NULL, - mask); + NULL, mask); } -EXPORT_SYMBOL(set_extent_bits); int clear_extent_bits(struct extent_io_tree *tree, u64 start, u64 end, int bits, gfp_t mask) { - return clear_extent_bit(tree, start, end, bits, 0, 0, mask); + return clear_extent_bit(tree, start, end, bits, 0, 0, NULL, mask); } -EXPORT_SYMBOL(clear_extent_bits); int set_extent_delalloc(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask) { return set_extent_bit(tree, start, end, - EXTENT_DELALLOC | EXTENT_DIRTY, - 0, NULL, mask); + EXTENT_DELALLOC | EXTENT_DIRTY | EXTENT_UPTODATE, + 0, NULL, NULL, mask); } -EXPORT_SYMBOL(set_extent_delalloc); int clear_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask) { return clear_extent_bit(tree, start, end, - EXTENT_DIRTY | EXTENT_DELALLOC, 0, 0, mask); + EXTENT_DIRTY | EXTENT_DELALLOC | + EXTENT_DO_ACCOUNTING, 0, 0, + NULL, mask); } -EXPORT_SYMBOL(clear_extent_dirty); - -int clear_extent_ordered(struct extent_io_tree *tree, u64 start, u64 end, - gfp_t mask) -{ - return clear_extent_bit(tree, start, end, EXTENT_ORDERED, 1, 0, mask); -} -EXPORT_SYMBOL(clear_extent_ordered); int set_extent_new(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask) { return set_extent_bit(tree, start, end, EXTENT_NEW, 0, NULL, - mask); + NULL, mask); } -EXPORT_SYMBOL(set_extent_new); -int clear_extent_new(struct extent_io_tree *tree, u64 start, u64 end, +static int clear_extent_new(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask) { - return clear_extent_bit(tree, start, end, EXTENT_NEW, 0, 0, mask); + return clear_extent_bit(tree, start, end, EXTENT_NEW, 0, 0, + NULL, mask); } -EXPORT_SYMBOL(clear_extent_new); int set_extent_uptodate(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask) { return set_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, NULL, - mask); + NULL, mask); } -EXPORT_SYMBOL(set_extent_uptodate); -int clear_extent_uptodate(struct extent_io_tree *tree, u64 start, u64 end, - gfp_t mask) +static int clear_extent_uptodate(struct extent_io_tree *tree, u64 start, + u64 end, gfp_t mask) { - return clear_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, 0, mask); + return clear_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, 0, + NULL, mask); } -EXPORT_SYMBOL(clear_extent_uptodate); - -int set_extent_writeback(struct extent_io_tree *tree, u64 start, u64 end, - gfp_t mask) -{ - return set_extent_bit(tree, start, end, EXTENT_WRITEBACK, - 0, NULL, mask); -} -EXPORT_SYMBOL(set_extent_writeback); - -int clear_extent_writeback(struct extent_io_tree *tree, u64 start, u64 end, - gfp_t mask) -{ - return clear_extent_bit(tree, start, end, EXTENT_WRITEBACK, 1, 0, mask); -} -EXPORT_SYMBOL(clear_extent_writeback); int wait_on_extent_writeback(struct extent_io_tree *tree, u64 start, u64 end) { return wait_extent_bit(tree, start, end, EXTENT_WRITEBACK); } -EXPORT_SYMBOL(wait_on_extent_writeback); -int lock_extent(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask) +/* + * either insert or lock state struct between start and end use mask to tell + * us if waiting is desired. + */ +int lock_extent_bits(struct extent_io_tree *tree, u64 start, u64 end, + int bits, struct extent_state **cached_state, gfp_t mask) { int err; u64 failed_start; while (1) { - err = set_extent_bit(tree, start, end, EXTENT_LOCKED, 1, - &failed_start, mask); + err = set_extent_bit(tree, start, end, EXTENT_LOCKED | bits, + EXTENT_LOCKED, &failed_start, + cached_state, mask); if (err == -EEXIST && (mask & __GFP_WAIT)) { wait_extent_bit(tree, failed_start, end, EXTENT_LOCKED); start = failed_start; @@ -907,14 +1018,42 @@ int lock_extent(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask) } return err; } -EXPORT_SYMBOL(lock_extent); + +int lock_extent(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask) +{ + return lock_extent_bits(tree, start, end, 0, NULL, mask); +} + +int try_lock_extent(struct extent_io_tree *tree, u64 start, u64 end, + gfp_t mask) +{ + int err; + u64 failed_start; + + err = set_extent_bit(tree, start, end, EXTENT_LOCKED, EXTENT_LOCKED, + &failed_start, NULL, mask); + if (err == -EEXIST) { + if (failed_start > start) + clear_extent_bit(tree, start, failed_start - 1, + EXTENT_LOCKED, 1, 0, NULL, mask); + return 0; + } + return 1; +} + +int unlock_extent_cached(struct extent_io_tree *tree, u64 start, u64 end, + struct extent_state **cached, gfp_t mask) +{ + return clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, cached, + mask); +} int unlock_extent(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask) { - return clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, mask); + return clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, NULL, + mask); } -EXPORT_SYMBOL(unlock_extent); /* * helper function to set pages and extents in the tree dirty @@ -932,15 +1071,13 @@ int set_range_dirty(struct extent_io_tree *tree, u64 start, u64 end) page_cache_release(page); index++; } - set_extent_dirty(tree, start, end, GFP_NOFS); return 0; } -EXPORT_SYMBOL(set_range_dirty); /* * helper function to set both pages and extents in the tree writeback */ -int set_range_writeback(struct extent_io_tree *tree, u64 start, u64 end) +static int set_range_writeback(struct extent_io_tree *tree, u64 start, u64 end) { unsigned long index = start >> PAGE_CACHE_SHIFT; unsigned long end_index = end >> PAGE_CACHE_SHIFT; @@ -953,11 +1090,16 @@ int set_range_writeback(struct extent_io_tree *tree, u64 start, u64 end) page_cache_release(page); index++; } - set_extent_writeback(tree, start, end, GFP_NOFS); return 0; } -EXPORT_SYMBOL(set_range_writeback); +/* + * find the first offset in the io tree with 'bits' set. zero is + * returned if we find something, and *start_ret and *end_ret are + * set to reflect the state struct that was found. + * + * If nothing was found, 1 is returned, < 0 on error + */ int find_first_extent_bit(struct extent_io_tree *tree, u64 start, u64 *start_ret, u64 *end_ret, int bits) { @@ -965,17 +1107,16 @@ int find_first_extent_bit(struct extent_io_tree *tree, u64 start, struct extent_state *state; int ret = 1; - spin_lock_irq(&tree->lock); + spin_lock(&tree->lock); /* * this search will find all the extents that end after * our range starts. */ node = tree_search(tree, start); - if (!node) { + if (!node) goto out; - } - while(1) { + while (1) { state = rb_entry(node, struct extent_state, rb_node); if (state->end >= start && (state->state & bits)) { *start_ret = state->start; @@ -988,11 +1129,14 @@ int find_first_extent_bit(struct extent_io_tree *tree, u64 start, break; } out: - spin_unlock_irq(&tree->lock); + spin_unlock(&tree->lock); return ret; } -EXPORT_SYMBOL(find_first_extent_bit); +/* find the first state struct with 'bits' set after 'start', and + * return it. tree->lock must be held. NULL will returned if + * nothing was found after 'start' + */ struct extent_state *find_first_extent_bit_state(struct extent_io_tree *tree, u64 start, int bits) { @@ -1004,15 +1148,14 @@ struct extent_state *find_first_extent_bit_state(struct extent_io_tree *tree, * our range starts. */ node = tree_search(tree, start); - if (!node) { + if (!node) goto out; - } - while(1) { + while (1) { state = rb_entry(node, struct extent_state, rb_node); - if (state->end >= start && (state->state & bits)) { + if (state->end >= start && (state->state & bits)) return state; - } + node = rb_next(node); if (!node) break; @@ -1020,10 +1163,15 @@ struct extent_state *find_first_extent_bit_state(struct extent_io_tree *tree, out: return NULL; } -EXPORT_SYMBOL(find_first_extent_bit_state); -u64 find_lock_delalloc_range(struct extent_io_tree *tree, - u64 *start, u64 *end, u64 max_bytes) +/* + * find a contiguous range of bytes in the file marked as delalloc, not + * more than 'max_bytes'. start and end are used to return the range, + * + * 1 is returned if we find something, 0 if nothing was in the tree + */ +static noinline u64 find_delalloc_range(struct extent_io_tree *tree, + u64 *start, u64 *end, u64 max_bytes) { struct rb_node *node; struct extent_state *state; @@ -1031,12 +1179,12 @@ u64 find_lock_delalloc_range(struct extent_io_tree *tree, u64 found = 0; u64 total_bytes = 0; - spin_lock_irq(&tree->lock); + spin_lock(&tree->lock); + /* * this search will find all the extents that end after * our range starts. */ -search_again: node = tree_search(tree, cur_start); if (!node) { if (!found) @@ -1044,9 +1192,10 @@ search_again: goto out; } - while(1) { + while (1) { state = rb_entry(node, struct extent_state, rb_node); - if (found && state->start != cur_start) { + if (found && (state->start != cur_start || + (state->state & EXTENT_BOUNDARY))) { goto out; } if (!(state->state & EXTENT_DELALLOC)) { @@ -1054,36 +1203,6 @@ search_again: *end = state->end; goto out; } - if (!found) { - struct extent_state *prev_state; - struct rb_node *prev_node = node; - while(1) { - prev_node = rb_prev(prev_node); - if (!prev_node) - break; - prev_state = rb_entry(prev_node, - struct extent_state, - rb_node); - if (!(prev_state->state & EXTENT_DELALLOC)) - break; - state = prev_state; - node = prev_node; - } - } - if (state->state & EXTENT_LOCKED) { - DEFINE_WAIT(wait); - atomic_inc(&state->refs); - prepare_to_wait(&state->wq, &wait, - TASK_UNINTERRUPTIBLE); - spin_unlock_irq(&tree->lock); - schedule(); - spin_lock_irq(&tree->lock); - finish_wait(&state->wq, &wait); - free_extent_state(state); - goto search_again; - } - set_state_cb(tree, state, EXTENT_LOCKED); - state->state |= EXTENT_LOCKED; if (!found) *start = state->start; found++; @@ -1097,10 +1216,255 @@ search_again: break; } out: - spin_unlock_irq(&tree->lock); + spin_unlock(&tree->lock); return found; } +static noinline int __unlock_for_delalloc(struct inode *inode, + struct page *locked_page, + u64 start, u64 end) +{ + int ret; + struct page *pages[16]; + unsigned long index = start >> PAGE_CACHE_SHIFT; + unsigned long end_index = end >> PAGE_CACHE_SHIFT; + unsigned long nr_pages = end_index - index + 1; + int i; + + if (index == locked_page->index && end_index == index) + return 0; + + while (nr_pages > 0) { + ret = find_get_pages_contig(inode->i_mapping, index, + min_t(unsigned long, nr_pages, + ARRAY_SIZE(pages)), pages); + for (i = 0; i < ret; i++) { + if (pages[i] != locked_page) + unlock_page(pages[i]); + page_cache_release(pages[i]); + } + nr_pages -= ret; + index += ret; + cond_resched(); + } + return 0; +} + +static noinline int lock_delalloc_pages(struct inode *inode, + struct page *locked_page, + u64 delalloc_start, + u64 delalloc_end) +{ + unsigned long index = delalloc_start >> PAGE_CACHE_SHIFT; + unsigned long start_index = index; + unsigned long end_index = delalloc_end >> PAGE_CACHE_SHIFT; + unsigned long pages_locked = 0; + struct page *pages[16]; + unsigned long nrpages; + int ret; + int i; + + /* the caller is responsible for locking the start index */ + if (index == locked_page->index && index == end_index) + return 0; + + /* skip the page at the start index */ + nrpages = end_index - index + 1; + while (nrpages > 0) { + ret = find_get_pages_contig(inode->i_mapping, index, + min_t(unsigned long, + nrpages, ARRAY_SIZE(pages)), pages); + if (ret == 0) { + ret = -EAGAIN; + goto done; + } + /* now we have an array of pages, lock them all */ + for (i = 0; i < ret; i++) { + /* + * the caller is taking responsibility for + * locked_page + */ + if (pages[i] != locked_page) { + lock_page(pages[i]); + if (!PageDirty(pages[i]) || + pages[i]->mapping != inode->i_mapping) { + ret = -EAGAIN; + unlock_page(pages[i]); + page_cache_release(pages[i]); + goto done; + } + } + page_cache_release(pages[i]); + pages_locked++; + } + nrpages -= ret; + index += ret; + cond_resched(); + } + ret = 0; +done: + if (ret && pages_locked) { + __unlock_for_delalloc(inode, locked_page, + delalloc_start, + ((u64)(start_index + pages_locked - 1)) << + PAGE_CACHE_SHIFT); + } + return ret; +} + +/* + * find a contiguous range of bytes in the file marked as delalloc, not + * more than 'max_bytes'. start and end are used to return the range, + * + * 1 is returned if we find something, 0 if nothing was in the tree + */ +static noinline u64 find_lock_delalloc_range(struct inode *inode, + struct extent_io_tree *tree, + struct page *locked_page, + u64 *start, u64 *end, + u64 max_bytes) +{ + u64 delalloc_start; + u64 delalloc_end; + u64 found; + struct extent_state *cached_state = NULL; + int ret; + int loops = 0; + +again: + /* step one, find a bunch of delalloc bytes starting at start */ + delalloc_start = *start; + delalloc_end = 0; + found = find_delalloc_range(tree, &delalloc_start, &delalloc_end, + max_bytes); + if (!found || delalloc_end <= *start) { + *start = delalloc_start; + *end = delalloc_end; + return found; + } + + /* + * start comes from the offset of locked_page. We have to lock + * pages in order, so we can't process delalloc bytes before + * locked_page + */ + if (delalloc_start < *start) + delalloc_start = *start; + + /* + * make sure to limit the number of pages we try to lock down + * if we're looping. + */ + if (delalloc_end + 1 - delalloc_start > max_bytes && loops) + delalloc_end = delalloc_start + PAGE_CACHE_SIZE - 1; + + /* step two, lock all the pages after the page that has start */ + ret = lock_delalloc_pages(inode, locked_page, + delalloc_start, delalloc_end); + if (ret == -EAGAIN) { + /* some of the pages are gone, lets avoid looping by + * shortening the size of the delalloc range we're searching + */ + free_extent_state(cached_state); + if (!loops) { + unsigned long offset = (*start) & (PAGE_CACHE_SIZE - 1); + max_bytes = PAGE_CACHE_SIZE - offset; + loops = 1; + goto again; + } else { + found = 0; + goto out_failed; + } + } + BUG_ON(ret); + + /* step three, lock the state bits for the whole range */ + lock_extent_bits(tree, delalloc_start, delalloc_end, + 0, &cached_state, GFP_NOFS); + + /* then test to make sure it is all still delalloc */ + ret = test_range_bit(tree, delalloc_start, delalloc_end, + EXTENT_DELALLOC, 1, cached_state); + if (!ret) { + unlock_extent_cached(tree, delalloc_start, delalloc_end, + &cached_state, GFP_NOFS); + __unlock_for_delalloc(inode, locked_page, + delalloc_start, delalloc_end); + cond_resched(); + goto again; + } + free_extent_state(cached_state); + *start = delalloc_start; + *end = delalloc_end; +out_failed: + return found; +} + +int extent_clear_unlock_delalloc(struct inode *inode, + struct extent_io_tree *tree, + u64 start, u64 end, struct page *locked_page, + unsigned long op) +{ + int ret; + struct page *pages[16]; + unsigned long index = start >> PAGE_CACHE_SHIFT; + unsigned long end_index = end >> PAGE_CACHE_SHIFT; + unsigned long nr_pages = end_index - index + 1; + int i; + int clear_bits = 0; + + if (op & EXTENT_CLEAR_UNLOCK) + clear_bits |= EXTENT_LOCKED; + if (op & EXTENT_CLEAR_DIRTY) + clear_bits |= EXTENT_DIRTY; + + if (op & EXTENT_CLEAR_DELALLOC) + clear_bits |= EXTENT_DELALLOC; + + if (op & EXTENT_CLEAR_ACCOUNTING) + clear_bits |= EXTENT_DO_ACCOUNTING; + + clear_extent_bit(tree, start, end, clear_bits, 1, 0, NULL, GFP_NOFS); + if (!(op & (EXTENT_CLEAR_UNLOCK_PAGE | EXTENT_CLEAR_DIRTY | + EXTENT_SET_WRITEBACK | EXTENT_END_WRITEBACK | + EXTENT_SET_PRIVATE2))) + return 0; + + while (nr_pages > 0) { + ret = find_get_pages_contig(inode->i_mapping, index, + min_t(unsigned long, + nr_pages, ARRAY_SIZE(pages)), pages); + for (i = 0; i < ret; i++) { + + if (op & EXTENT_SET_PRIVATE2) + SetPagePrivate2(pages[i]); + + if (pages[i] == locked_page) { + page_cache_release(pages[i]); + continue; + } + if (op & EXTENT_CLEAR_DIRTY) + clear_page_dirty_for_io(pages[i]); + if (op & EXTENT_SET_WRITEBACK) + set_page_writeback(pages[i]); + if (op & EXTENT_END_WRITEBACK) + end_page_writeback(pages[i]); + if (op & EXTENT_CLEAR_UNLOCK_PAGE) + unlock_page(pages[i]); + page_cache_release(pages[i]); + } + nr_pages -= ret; + index += ret; + cond_resched(); + } + return 0; +} + +/* + * count the number of bytes in the tree that have a given bit(s) + * set. This can be fairly slow, except for EXTENT_DIRTY which is + * cached. The total number found is returned. + */ u64 count_range_bits(struct extent_io_tree *tree, u64 *start, u64 search_end, u64 max_bytes, unsigned long bits) @@ -1112,12 +1476,11 @@ u64 count_range_bits(struct extent_io_tree *tree, int found = 0; if (search_end <= cur_start) { - printk("search_end %Lu start %Lu\n", search_end, cur_start); WARN_ON(1); return 0; } - spin_lock_irq(&tree->lock); + spin_lock(&tree->lock); if (cur_start == 0 && bits == EXTENT_DIRTY) { total_bytes = tree->dirty_bytes; goto out; @@ -1127,11 +1490,10 @@ u64 count_range_bits(struct extent_io_tree *tree, * our range starts. */ node = tree_search(tree, cur_start); - if (!node) { + if (!node) goto out; - } - while(1) { + while (1) { state = rb_entry(node, struct extent_state, rb_node); if (state->start > search_end) break; @@ -1150,79 +1512,21 @@ u64 count_range_bits(struct extent_io_tree *tree, break; } out: - spin_unlock_irq(&tree->lock); + spin_unlock(&tree->lock); return total_bytes; } -/* - * helper function to lock both pages and extents in the tree. - * pages must be locked first. - */ -int lock_range(struct extent_io_tree *tree, u64 start, u64 end) -{ - unsigned long index = start >> PAGE_CACHE_SHIFT; - unsigned long end_index = end >> PAGE_CACHE_SHIFT; - struct page *page; - int err; - - while (index <= end_index) { - page = grab_cache_page(tree->mapping, index); - if (!page) { - err = -ENOMEM; - goto failed; - } - if (IS_ERR(page)) { - err = PTR_ERR(page); - goto failed; - } - index++; - } - lock_extent(tree, start, end, GFP_NOFS); - return 0; - -failed: - /* - * we failed above in getting the page at 'index', so we undo here - * up to but not including the page at 'index' - */ - end_index = index; - index = start >> PAGE_CACHE_SHIFT; - while (index < end_index) { - page = find_get_page(tree->mapping, index); - unlock_page(page); - page_cache_release(page); - index++; - } - return err; -} -EXPORT_SYMBOL(lock_range); /* - * helper function to unlock both pages and extents in the tree. + * set the private field for a given byte offset in the tree. If there isn't + * an extent_state there already, this does nothing. */ -int unlock_range(struct extent_io_tree *tree, u64 start, u64 end) -{ - unsigned long index = start >> PAGE_CACHE_SHIFT; - unsigned long end_index = end >> PAGE_CACHE_SHIFT; - struct page *page; - - while (index <= end_index) { - page = find_get_page(tree->mapping, index); - unlock_page(page); - page_cache_release(page); - index++; - } - unlock_extent(tree, start, end, GFP_NOFS); - return 0; -} -EXPORT_SYMBOL(unlock_range); - int set_state_private(struct extent_io_tree *tree, u64 start, u64 private) { struct rb_node *node; struct extent_state *state; int ret = 0; - spin_lock_irq(&tree->lock); + spin_lock(&tree->lock); /* * this search will find all the extents that end after * our range starts. @@ -1239,7 +1543,7 @@ int set_state_private(struct extent_io_tree *tree, u64 start, u64 private) } state->private = private; out: - spin_unlock_irq(&tree->lock); + spin_unlock(&tree->lock); return ret; } @@ -1249,7 +1553,7 @@ int get_state_private(struct extent_io_tree *tree, u64 start, u64 *private) struct extent_state *state; int ret = 0; - spin_lock_irq(&tree->lock); + spin_lock(&tree->lock); /* * this search will find all the extents that end after * our range starts. @@ -1266,7 +1570,7 @@ int get_state_private(struct extent_io_tree *tree, u64 start, u64 *private) } *private = state->private; out: - spin_unlock_irq(&tree->lock); + spin_unlock(&tree->lock); return ret; } @@ -1277,15 +1581,17 @@ out: * range is found set. */ int test_range_bit(struct extent_io_tree *tree, u64 start, u64 end, - int bits, int filled) + int bits, int filled, struct extent_state *cached) { struct extent_state *state = NULL; struct rb_node *node; int bitset = 0; - unsigned long flags; - spin_lock_irqsave(&tree->lock, flags); - node = tree_search(tree, start); + spin_lock(&tree->lock); + if (cached && cached->tree && cached->start == start) + node = &cached->rb_node; + else + node = tree_search(tree, start); while (node && start <= end) { state = rb_entry(node, struct extent_state, rb_node); @@ -1305,6 +1611,10 @@ int test_range_bit(struct extent_io_tree *tree, u64 start, u64 end, bitset = 0; break; } + + if (state->end == (u64)-1) + break; + start = state->end + 1; if (start > end) break; @@ -1315,10 +1625,9 @@ int test_range_bit(struct extent_io_tree *tree, u64 start, u64 end, break; } } - spin_unlock_irqrestore(&tree->lock, flags); + spin_unlock(&tree->lock); return bitset; } -EXPORT_SYMBOL(test_range_bit); /* * helper function to set a given page up to date if all the @@ -1329,7 +1638,7 @@ static int check_page_uptodate(struct extent_io_tree *tree, { u64 start = (u64)page->index << PAGE_CACHE_SHIFT; u64 end = start + PAGE_CACHE_SIZE - 1; - if (test_range_bit(tree, start, end, EXTENT_UPTODATE, 1)) + if (test_range_bit(tree, start, end, EXTENT_UPTODATE, 1, NULL)) SetPageUptodate(page); return 0; } @@ -1343,7 +1652,7 @@ static int check_page_locked(struct extent_io_tree *tree, { u64 start = (u64)page->index << PAGE_CACHE_SHIFT; u64 end = start + PAGE_CACHE_SIZE - 1; - if (!test_range_bit(tree, start, end, EXTENT_LOCKED, 0)) + if (!test_range_bit(tree, start, end, EXTENT_LOCKED, 0, NULL)) unlock_page(page); return 0; } @@ -1355,10 +1664,7 @@ static int check_page_locked(struct extent_io_tree *tree, static int check_page_writeback(struct extent_io_tree *tree, struct page *page) { - u64 start = (u64)page->index << PAGE_CACHE_SHIFT; - u64 end = start + PAGE_CACHE_SIZE - 1; - if (!test_range_bit(tree, start, end, EXTENT_WRITEBACK, 0)) - end_page_writeback(page); + end_page_writeback(page); return 0; } @@ -1373,31 +1679,20 @@ static int check_page_writeback(struct extent_io_tree *tree, * Scheduling is not allowed, so the extent state tree is expected * to have one and only one object corresponding to this IO. */ -#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) static void end_bio_extent_writepage(struct bio *bio, int err) -#else -static int end_bio_extent_writepage(struct bio *bio, - unsigned int bytes_done, int err) -#endif { int uptodate = err == 0; struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1; - struct extent_state *state = bio->bi_private; - struct extent_io_tree *tree = state->tree; - struct rb_node *node; + struct extent_io_tree *tree; u64 start; u64 end; - u64 cur; int whole_page; int ret; - unsigned long flags; -#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) - if (bio->bi_size) - return 1; -#endif do { struct page *page = bvec->bv_page; + tree = &BTRFS_I(page->mapping->host)->io_tree; + start = ((u64)page->index << PAGE_CACHE_SHIFT) + bvec->bv_offset; end = start + bvec->bv_len - 1; @@ -1411,7 +1706,7 @@ static int end_bio_extent_writepage(struct bio *bio, prefetchw(&bvec->bv_page->flags); if (tree->ops && tree->ops->writepage_end_io_hook) { ret = tree->ops->writepage_end_io_hook(page, start, - end, state, uptodate); + end, NULL, uptodate); if (ret) uptodate = 0; } @@ -1419,92 +1714,26 @@ static int end_bio_extent_writepage(struct bio *bio, if (!uptodate && tree->ops && tree->ops->writepage_io_failed_hook) { ret = tree->ops->writepage_io_failed_hook(bio, page, - start, end, state); + start, end, NULL); if (ret == 0) { - state = NULL; uptodate = (err == 0); continue; } } if (!uptodate) { - clear_extent_uptodate(tree, start, end, GFP_ATOMIC); + clear_extent_uptodate(tree, start, end, GFP_NOFS); ClearPageUptodate(page); SetPageError(page); } - /* - * bios can get merged in funny ways, and so we need to - * be careful with the state variable. We know the - * state won't be merged with others because it has - * WRITEBACK set, but we can't be sure each biovec is - * sequential in the file. So, if our cached state - * doesn't match the expected end, search the tree - * for the correct one. - */ - - spin_lock_irqsave(&tree->lock, flags); - if (!state || state->end != end) { - state = NULL; - node = __etree_search(tree, start, NULL, NULL); - if (node) { - state = rb_entry(node, struct extent_state, - rb_node); - if (state->end != end || - !(state->state & EXTENT_WRITEBACK)) - state = NULL; - } - if (!state) { - spin_unlock_irqrestore(&tree->lock, flags); - clear_extent_writeback(tree, start, - end, GFP_ATOMIC); - goto next_io; - } - } - cur = end; - while(1) { - struct extent_state *clear = state; - cur = state->start; - node = rb_prev(&state->rb_node); - if (node) { - state = rb_entry(node, - struct extent_state, - rb_node); - } else { - state = NULL; - } - - clear_state_bit(tree, clear, EXTENT_WRITEBACK, - 1, 0); - if (cur == start) - break; - if (cur < start) { - WARN_ON(1); - break; - } - if (!node) - break; - } - /* before releasing the lock, make sure the next state - * variable has the expected bits set and corresponds - * to the correct offsets in the file - */ - if (state && (state->end + 1 != start || - !(state->state & EXTENT_WRITEBACK))) { - state = NULL; - } - spin_unlock_irqrestore(&tree->lock, flags); -next_io: - if (whole_page) end_page_writeback(page); else check_page_writeback(tree, page); } while (bvec >= bio->bi_io_vec); + bio_put(bio); -#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) - return 0; -#endif } /* @@ -1518,32 +1747,23 @@ next_io: * Scheduling is not allowed, so the extent state tree is expected * to have one and only one object corresponding to this IO. */ -#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) static void end_bio_extent_readpage(struct bio *bio, int err) -#else -static int end_bio_extent_readpage(struct bio *bio, - unsigned int bytes_done, int err) -#endif { int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags); struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1; - struct extent_state *state = bio->bi_private; - struct extent_io_tree *tree = state->tree; - struct rb_node *node; + struct extent_io_tree *tree; u64 start; u64 end; - u64 cur; - unsigned long flags; int whole_page; int ret; -#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) - if (bio->bi_size) - return 1; -#endif + if (err) + uptodate = 0; do { struct page *page = bvec->bv_page; + tree = &BTRFS_I(page->mapping->host)->io_tree; + start = ((u64)page->index << PAGE_CACHE_SHIFT) + bvec->bv_offset; end = start + bvec->bv_len - 1; @@ -1558,80 +1778,29 @@ static int end_bio_extent_readpage(struct bio *bio, if (uptodate && tree->ops && tree->ops->readpage_end_io_hook) { ret = tree->ops->readpage_end_io_hook(page, start, end, - state); + NULL); if (ret) uptodate = 0; } if (!uptodate && tree->ops && tree->ops->readpage_io_failed_hook) { ret = tree->ops->readpage_io_failed_hook(bio, page, - start, end, state); + start, end, NULL); if (ret == 0) { - state = NULL; uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags); + if (err) + uptodate = 0; continue; } } - spin_lock_irqsave(&tree->lock, flags); - if (!state || state->end != end) { - state = NULL; - node = __etree_search(tree, start, NULL, NULL); - if (node) { - state = rb_entry(node, struct extent_state, - rb_node); - if (state->end != end || - !(state->state & EXTENT_LOCKED)) - state = NULL; - } - if (!state) { - spin_unlock_irqrestore(&tree->lock, flags); - if (uptodate) - set_extent_uptodate(tree, start, end, - GFP_ATOMIC); - unlock_extent(tree, start, end, GFP_ATOMIC); - goto next_io; - } + if (uptodate) { + set_extent_uptodate(tree, start, end, + GFP_ATOMIC); } + unlock_extent(tree, start, end, GFP_ATOMIC); - cur = end; - while(1) { - struct extent_state *clear = state; - cur = state->start; - node = rb_prev(&state->rb_node); - if (node) { - state = rb_entry(node, - struct extent_state, - rb_node); - } else { - state = NULL; - } - if (uptodate) { - set_state_cb(tree, clear, EXTENT_UPTODATE); - clear->state |= EXTENT_UPTODATE; - } - clear_state_bit(tree, clear, EXTENT_LOCKED, - 1, 0); - if (cur == start) - break; - if (cur < start) { - WARN_ON(1); - break; - } - if (!node) - break; - } - /* before releasing the lock, make sure the next state - * variable has the expected bits set and corresponds - * to the correct offsets in the file - */ - if (state && (state->end + 1 != start || - !(state->state & EXTENT_LOCKED))) { - state = NULL; - } - spin_unlock_irqrestore(&tree->lock, flags); -next_io: if (whole_page) { if (uptodate) { SetPageUptodate(page); @@ -1652,9 +1821,6 @@ next_io: } while (bvec >= bio->bi_io_vec); bio_put(bio); -#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) - return 0; -#endif } /* @@ -1662,27 +1828,18 @@ next_io: * the structs in the extent tree when done, and set the uptodate bits * as appropriate. */ -#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) static void end_bio_extent_preparewrite(struct bio *bio, int err) -#else -static int end_bio_extent_preparewrite(struct bio *bio, - unsigned int bytes_done, int err) -#endif { const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags); struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1; - struct extent_state *state = bio->bi_private; - struct extent_io_tree *tree = state->tree; + struct extent_io_tree *tree; u64 start; u64 end; -#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) - if (bio->bi_size) - return 1; -#endif - do { struct page *page = bvec->bv_page; + tree = &BTRFS_I(page->mapping->host)->io_tree; + start = ((u64)page->index << PAGE_CACHE_SHIFT) + bvec->bv_offset; end = start + bvec->bv_len - 1; @@ -1702,9 +1859,6 @@ static int end_bio_extent_preparewrite(struct bio *bio, } while (bvec >= bio->bi_io_vec); bio_put(bio); -#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) - return 0; -#endif } static struct bio * @@ -1728,38 +1882,26 @@ extent_bio_alloc(struct block_device *bdev, u64 first_sector, int nr_vecs, return bio; } -static int submit_one_bio(int rw, struct bio *bio, int mirror_num) +static int submit_one_bio(int rw, struct bio *bio, int mirror_num, + unsigned long bio_flags) { int ret = 0; struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1; struct page *page = bvec->bv_page; struct extent_io_tree *tree = bio->bi_private; - struct rb_node *node; - struct extent_state *state; u64 start; u64 end; start = ((u64)page->index << PAGE_CACHE_SHIFT) + bvec->bv_offset; end = start + bvec->bv_len - 1; - spin_lock_irq(&tree->lock); - node = __etree_search(tree, start, NULL, NULL); - BUG_ON(!node); - state = rb_entry(node, struct extent_state, rb_node); - while(state->end < end) { - node = rb_next(node); - state = rb_entry(node, struct extent_state, rb_node); - } - BUG_ON(state->end != end); - spin_unlock_irq(&tree->lock); - - bio->bi_private = state; + bio->bi_private = NULL; bio_get(bio); if (tree->ops && tree->ops->submit_bio_hook) tree->ops->submit_bio_hook(page->mapping->host, rw, bio, - mirror_num); + mirror_num, bio_flags); else submit_bio(rw, bio); if (bio_flagged(bio, BIO_EOPNOTSUPP)) @@ -1775,40 +1917,53 @@ static int submit_extent_page(int rw, struct extent_io_tree *tree, struct bio **bio_ret, unsigned long max_pages, bio_end_io_t end_io_func, - int mirror_num) + int mirror_num, + unsigned long prev_bio_flags, + unsigned long bio_flags) { int ret = 0; struct bio *bio; int nr; + int contig = 0; + int this_compressed = bio_flags & EXTENT_BIO_COMPRESSED; + int old_compressed = prev_bio_flags & EXTENT_BIO_COMPRESSED; + size_t page_size = min_t(size_t, size, PAGE_CACHE_SIZE); if (bio_ret && *bio_ret) { bio = *bio_ret; - if (bio->bi_sector + (bio->bi_size >> 9) != sector || + if (old_compressed) + contig = bio->bi_sector == sector; + else + contig = bio->bi_sector + (bio->bi_size >> 9) == + sector; + + if (prev_bio_flags != bio_flags || !contig || (tree->ops && tree->ops->merge_bio_hook && - tree->ops->merge_bio_hook(page, offset, size, bio)) || - bio_add_page(bio, page, size, offset) < size) { - ret = submit_one_bio(rw, bio, mirror_num); + tree->ops->merge_bio_hook(page, offset, page_size, bio, + bio_flags)) || + bio_add_page(bio, page, page_size, offset) < page_size) { + ret = submit_one_bio(rw, bio, mirror_num, + prev_bio_flags); bio = NULL; } else { return 0; } } - nr = bio_get_nr_vecs(bdev); - bio = extent_bio_alloc(bdev, sector, nr, GFP_NOFS | __GFP_HIGH); - if (!bio) { - printk("failed to allocate bio nr %d\n", nr); - } + if (this_compressed) + nr = BIO_MAX_PAGES; + else + nr = bio_get_nr_vecs(bdev); + bio = extent_bio_alloc(bdev, sector, nr, GFP_NOFS | __GFP_HIGH); - bio_add_page(bio, page, size, offset); + bio_add_page(bio, page, page_size, offset); bio->bi_end_io = end_io_func; bio->bi_private = tree; - if (bio_ret) { + if (bio_ret) *bio_ret = bio; - } else { - ret = submit_one_bio(rw, bio, mirror_num); - } + else + ret = submit_one_bio(rw, bio, mirror_num, bio_flags); return ret; } @@ -1817,13 +1972,12 @@ void set_page_extent_mapped(struct page *page) { if (!PagePrivate(page)) { SetPagePrivate(page); - WARN_ON(!page->mapping->a_ops->invalidatepage); - set_page_private(page, EXTENT_PAGE_PRIVATE); page_cache_get(page); + set_page_private(page, EXTENT_PAGE_PRIVATE); } } -void set_page_extent_head(struct page *page, unsigned long len) +static void set_page_extent_head(struct page *page, unsigned long len) { set_page_private(page, EXTENT_PAGE_PRIVATE_FIRST_PAGE | len << 2); } @@ -1836,7 +1990,8 @@ void set_page_extent_head(struct page *page, unsigned long len) static int __extent_read_full_page(struct extent_io_tree *tree, struct page *page, get_extent_t *get_extent, - struct bio **bio, int mirror_num) + struct bio **bio, int mirror_num, + unsigned long *bio_flags) { struct inode *inode = page->mapping->host; u64 start = (u64)page->index << PAGE_CACHE_SHIFT; @@ -1854,13 +2009,27 @@ static int __extent_read_full_page(struct extent_io_tree *tree, int nr = 0; size_t page_offset = 0; size_t iosize; + size_t disk_io_size; size_t blocksize = inode->i_sb->s_blocksize; + unsigned long this_bio_flag = 0; set_page_extent_mapped(page); end = page_end; lock_extent(tree, start, end, GFP_NOFS); + if (page->index == last_byte >> PAGE_CACHE_SHIFT) { + char *userpage; + size_t zero_offset = last_byte & (PAGE_CACHE_SIZE - 1); + + if (zero_offset) { + iosize = PAGE_CACHE_SIZE - zero_offset; + userpage = kmap_atomic(page, KM_USER0); + memset(userpage + zero_offset, 0, iosize); + flush_dcache_page(page); + kunmap_atomic(userpage, KM_USER0); + } + } while (cur <= end) { if (cur >= last_byte) { char *userpage; @@ -1882,21 +2051,26 @@ static int __extent_read_full_page(struct extent_io_tree *tree, break; } extent_offset = cur - em->start; - if (extent_map_end(em) <= cur) { -printk("bad mapping em [%Lu %Lu] cur %Lu\n", em->start, extent_map_end(em), cur); - } BUG_ON(extent_map_end(em) <= cur); - if (end < cur) { -printk("2bad mapping end %Lu cur %Lu\n", end, cur); - } BUG_ON(end < cur); + if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) + this_bio_flag = EXTENT_BIO_COMPRESSED; + 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); - sector = (em->block_start + extent_offset) >> 9; + if (this_bio_flag & EXTENT_BIO_COMPRESSED) { + disk_io_size = em->block_len; + sector = em->block_start >> 9; + } else { + sector = (em->block_start + extent_offset) >> 9; + disk_io_size = iosize; + } bdev = em->bdev; block_start = em->block_start; + if (test_bit(EXTENT_FLAG_PREALLOC, &em->flags)) + block_start = EXTENT_MAP_HOLE; free_extent_map(em); em = NULL; @@ -1916,7 +2090,9 @@ printk("2bad mapping end %Lu cur %Lu\n", end, cur); continue; } /* the get_extent function already copied into the page */ - if (test_range_bit(tree, cur, cur_end, EXTENT_UPTODATE, 1)) { + if (test_range_bit(tree, cur, cur_end, + EXTENT_UPTODATE, 1, NULL)) { + check_page_uptodate(tree, page); unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS); cur = cur + iosize; page_offset += iosize; @@ -1939,18 +2115,21 @@ printk("2bad mapping end %Lu cur %Lu\n", end, cur); cur + iosize - 1); } if (!ret) { - unsigned long nr = (last_byte >> PAGE_CACHE_SHIFT) + 1; - nr -= page->index; + unsigned long pnr = (last_byte >> PAGE_CACHE_SHIFT) + 1; + pnr -= page->index; ret = submit_extent_page(READ, tree, page, - sector, iosize, page_offset, - bdev, bio, nr, - end_bio_extent_readpage, mirror_num); + sector, disk_io_size, page_offset, + bdev, bio, pnr, + end_bio_extent_readpage, mirror_num, + *bio_flags, + this_bio_flag); + nr++; + *bio_flags = this_bio_flag; } if (ret) SetPageError(page); cur = cur + iosize; page_offset += iosize; - nr++; } if (!nr) { if (!PageError(page)) @@ -1964,14 +2143,25 @@ int extent_read_full_page(struct extent_io_tree *tree, struct page *page, get_extent_t *get_extent) { struct bio *bio = NULL; + unsigned long bio_flags = 0; int ret; - ret = __extent_read_full_page(tree, page, get_extent, &bio, 0); + ret = __extent_read_full_page(tree, page, get_extent, &bio, 0, + &bio_flags); if (bio) - submit_one_bio(READ, bio, 0); + submit_one_bio(READ, bio, 0, bio_flags); return ret; } -EXPORT_SYMBOL(extent_read_full_page); + +static noinline void update_nr_written(struct page *page, + struct writeback_control *wbc, + unsigned long nr_written) +{ + wbc->nr_to_write -= nr_written; + if (wbc->range_cyclic || (wbc->nr_to_write > 0 && + wbc->range_start == 0 && wbc->range_end == LLONG_MAX)) + page->mapping->writeback_index = page->index + nr_written; +} /* * the writepage semantics are similar to regular writepage. extent @@ -1996,21 +2186,31 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, u64 iosize; u64 unlock_start; sector_t sector; + struct extent_state *cached_state = NULL; struct extent_map *em; struct block_device *bdev; int ret; int nr = 0; - size_t page_offset = 0; + size_t pg_offset = 0; size_t blocksize; loff_t i_size = i_size_read(inode); unsigned long end_index = i_size >> PAGE_CACHE_SHIFT; u64 nr_delalloc; u64 delalloc_end; + int page_started; + int compressed; + int write_flags; + unsigned long nr_written = 0; + + if (wbc->sync_mode == WB_SYNC_ALL) + write_flags = WRITE_SYNC_PLUG; + else + write_flags = WRITE; WARN_ON(!PageLocked(page)); - page_offset = i_size & (PAGE_CACHE_SIZE - 1); + pg_offset = i_size & (PAGE_CACHE_SIZE - 1); if (page->index > end_index || - (page->index == end_index && !page_offset)) { + (page->index == end_index && !pg_offset)) { page->mapping->a_ops->invalidatepage(page, 0); unlock_page(page); return 0; @@ -2020,54 +2220,92 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, char *userpage; userpage = kmap_atomic(page, KM_USER0); - memset(userpage + page_offset, 0, - PAGE_CACHE_SIZE - page_offset); + memset(userpage + pg_offset, 0, + PAGE_CACHE_SIZE - pg_offset); kunmap_atomic(userpage, KM_USER0); flush_dcache_page(page); } - page_offset = 0; + pg_offset = 0; set_page_extent_mapped(page); delalloc_start = start; delalloc_end = 0; - while(delalloc_end < page_end) { - nr_delalloc = find_lock_delalloc_range(tree, &delalloc_start, + page_started = 0; + if (!epd->extent_locked) { + u64 delalloc_to_write = 0; + /* + * make sure the wbc mapping index is at least updated + * to this page. + */ + update_nr_written(page, wbc, 0); + + while (delalloc_end < page_end) { + nr_delalloc = find_lock_delalloc_range(inode, tree, + page, + &delalloc_start, &delalloc_end, 128 * 1024 * 1024); - if (nr_delalloc == 0) { + if (nr_delalloc == 0) { + delalloc_start = delalloc_end + 1; + continue; + } + tree->ops->fill_delalloc(inode, page, delalloc_start, + delalloc_end, &page_started, + &nr_written); + /* + * delalloc_end is already one less than the total + * length, so we don't subtract one from + * PAGE_CACHE_SIZE + */ + delalloc_to_write += (delalloc_end - delalloc_start + + PAGE_CACHE_SIZE) >> + PAGE_CACHE_SHIFT; delalloc_start = delalloc_end + 1; - continue; } - tree->ops->fill_delalloc(inode, delalloc_start, - delalloc_end); - clear_extent_bit(tree, delalloc_start, - delalloc_end, - EXTENT_LOCKED | EXTENT_DELALLOC, - 1, 0, GFP_NOFS); - delalloc_start = delalloc_end + 1; + if (wbc->nr_to_write < delalloc_to_write) { + int thresh = 8192; + + if (delalloc_to_write < thresh * 2) + thresh = delalloc_to_write; + wbc->nr_to_write = min_t(u64, delalloc_to_write, + thresh); + } + + /* did the fill delalloc function already unlock and start + * the IO? + */ + if (page_started) { + ret = 0; + /* + * we've unlocked the page, so we can't update + * the mapping's writeback index, just update + * nr_to_write. + */ + wbc->nr_to_write -= nr_written; + goto done_unlocked; + } } - lock_extent(tree, start, page_end, GFP_NOFS); - unlock_start = start; - if (tree->ops && tree->ops->writepage_start_hook) { - ret = tree->ops->writepage_start_hook(page, start, page_end); + ret = tree->ops->writepage_start_hook(page, start, + page_end); if (ret == -EAGAIN) { - unlock_extent(tree, start, page_end, GFP_NOFS); redirty_page_for_writepage(wbc, page); + update_nr_written(page, wbc, nr_written); unlock_page(page); - return 0; + ret = 0; + goto done_unlocked; } } - end = page_end; - if (test_range_bit(tree, start, page_end, EXTENT_DELALLOC, 0)) { - printk("found delalloc bits after lock_extent\n"); - } + /* + * we don't want to touch the inode after unlocking the page, + * so we update the mapping writeback index now + */ + update_nr_written(page, wbc, nr_written + 1); + end = page_end; if (last_byte <= start) { - clear_extent_dirty(tree, start, page_end, GFP_NOFS); - unlock_extent(tree, start, page_end, GFP_NOFS); if (tree->ops && tree->ops->writepage_end_io_hook) tree->ops->writepage_end_io_hook(page, start, page_end, NULL, 1); @@ -2075,20 +2313,17 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, goto done; } - set_extent_uptodate(tree, start, page_end, GFP_NOFS); blocksize = inode->i_sb->s_blocksize; while (cur <= end) { if (cur >= last_byte) { - clear_extent_dirty(tree, cur, page_end, GFP_NOFS); - unlock_extent(tree, unlock_start, page_end, GFP_NOFS); if (tree->ops && tree->ops->writepage_end_io_hook) tree->ops->writepage_end_io_hook(page, cur, page_end, NULL, 1); unlock_start = page_end + 1; break; } - em = epd->get_extent(inode, page, page_offset, cur, + em = epd->get_extent(inode, page, pg_offset, cur, end - cur + 1, 1); if (IS_ERR(em) || !em) { SetPageError(page); @@ -2103,34 +2338,46 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, sector = (em->block_start + extent_offset) >> 9; bdev = em->bdev; block_start = em->block_start; + compressed = test_bit(EXTENT_FLAG_COMPRESSED, &em->flags); free_extent_map(em); em = NULL; - if (block_start == EXTENT_MAP_HOLE || + /* + * compressed and inline extents are written through other + * paths in the FS + */ + if (compressed || block_start == EXTENT_MAP_HOLE || block_start == EXTENT_MAP_INLINE) { - clear_extent_dirty(tree, cur, - cur + iosize - 1, GFP_NOFS); - - unlock_extent(tree, unlock_start, cur + iosize -1, - GFP_NOFS); - if (tree->ops && tree->ops->writepage_end_io_hook) + /* + * end_io notification does not happen here for + * compressed extents + */ + if (!compressed && tree->ops && + tree->ops->writepage_end_io_hook) tree->ops->writepage_end_io_hook(page, cur, cur + iosize - 1, NULL, 1); - cur = cur + iosize; - page_offset += iosize; + else if (compressed) { + /* we don't want to end_page_writeback on + * a compressed extent. this happens + * elsewhere + */ + nr++; + } + + cur += iosize; + pg_offset += iosize; unlock_start = cur; continue; } - /* leave this out until we have a page_mkwrite call */ if (0 && !test_range_bit(tree, cur, cur + iosize - 1, - EXTENT_DIRTY, 0)) { + EXTENT_DIRTY, 0, NULL)) { cur = cur + iosize; - page_offset += iosize; + pg_offset += iosize; continue; } - clear_extent_dirty(tree, cur, cur + iosize - 1, GFP_NOFS); + if (tree->ops && tree->ops->writepage_io_hook) { ret = tree->ops->writepage_io_hook(page, cur, cur + iosize - 1); @@ -2141,23 +2388,25 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, SetPageError(page); } else { unsigned long max_nr = end_index + 1; + set_range_writeback(tree, cur, cur + iosize - 1); if (!PageWriteback(page)) { - printk("warning page %lu not writeback, " - "cur %llu end %llu\n", page->index, - (unsigned long long)cur, + printk(KERN_ERR "btrfs warning page %lu not " + "writeback, cur %llu end %llu\n", + page->index, (unsigned long long)cur, (unsigned long long)end); } - ret = submit_extent_page(WRITE, tree, page, sector, - iosize, page_offset, bdev, - &epd->bio, max_nr, - end_bio_extent_writepage, 0); + ret = submit_extent_page(write_flags, tree, page, + sector, iosize, pg_offset, + bdev, &epd->bio, max_nr, + end_bio_extent_writepage, + 0, 0, 0); if (ret) SetPageError(page); } cur = cur + iosize; - page_offset += iosize; + pg_offset += iosize; nr++; } done: @@ -2166,20 +2415,17 @@ done: set_page_writeback(page); end_page_writeback(page); } - if (unlock_start <= page_end) - unlock_extent(tree, unlock_start, page_end, GFP_NOFS); unlock_page(page); + +done_unlocked: + + /* drop our reference on any cached states */ + free_extent_state(cached_state); return 0; } -#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,22) -/* Taken directly from 2.6.23 for 2.6.18 back port */ -typedef int (*writepage_t)(struct page *page, struct writeback_control *wbc, - void *data); - /** - * write_cache_pages - walk the list of dirty pages of the given address space - * and write all of them. + * write_cache_pages - walk the list of dirty pages of the given address space and write all of them. * @mapping: address space structure to write * @wbc: subtract the number of written pages from *@wbc->nr_to_write * @writepage: function called for each page @@ -2193,13 +2439,15 @@ typedef int (*writepage_t)(struct page *page, struct writeback_control *wbc, * WB_SYNC_ALL then we were called for data integrity and we must wait for * existing IO to complete. */ -static int write_cache_pages(struct address_space *mapping, - struct writeback_control *wbc, writepage_t writepage, - void *data) +static int extent_write_cache_pages(struct extent_io_tree *tree, + struct address_space *mapping, + struct writeback_control *wbc, + writepage_t writepage, void *data, + void (*flush_fn)(void *)) { - struct backing_dev_info *bdi = mapping->backing_dev_info; int ret = 0; int done = 0; + int nr_to_write_done = 0; struct pagevec pvec; int nr_pages; pgoff_t index; @@ -2207,11 +2455,6 @@ static int write_cache_pages(struct address_space *mapping, int scanned = 0; int range_whole = 0; - if (wbc->nonblocking && bdi_write_congested(bdi)) { - wbc->encountered_congestion = 1; - return 0; - } - pagevec_init(&pvec, 0); if (wbc->range_cyclic) { index = mapping->writeback_index; /* Start from prev offset */ @@ -2224,10 +2467,10 @@ static int write_cache_pages(struct address_space *mapping, scanned = 1; } retry: - while (!done && (index <= end) && + while (!done && !nr_to_write_done && (index <= end) && (nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, - PAGECACHE_TAG_DIRTY, - min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1))) { + PAGECACHE_TAG_DIRTY, min(end - index, + (pgoff_t)PAGEVEC_SIZE-1) + 1))) { unsigned i; scanned = 1; @@ -2241,7 +2484,10 @@ retry: * swizzled back from swapper_space to tmpfs file * mapping */ - lock_page(page); + if (tree->ops && tree->ops->write_cache_pages_lock_hook) + tree->ops->write_cache_pages_lock_hook(page); + else + lock_page(page); if (unlikely(page->mapping != mapping)) { unlock_page(page); @@ -2254,8 +2500,11 @@ retry: continue; } - if (wbc->sync_mode != WB_SYNC_NONE) + if (wbc->sync_mode != WB_SYNC_NONE) { + if (PageWriteback(page)) + flush_fn(data); wait_on_page_writeback(page); + } if (PageWriteback(page) || !clear_page_dirty_for_io(page)) { @@ -2269,12 +2518,15 @@ retry: unlock_page(page); ret = 0; } - if (ret || (--(wbc->nr_to_write) <= 0)) - done = 1; - if (wbc->nonblocking && bdi_write_congested(bdi)) { - wbc->encountered_congestion = 1; + if (ret) done = 1; - } + + /* + * the filesystem may choose to bump up nr_to_write. + * We have to make sure to honor the new nr_to_write + * at any time + */ + nr_to_write_done = wbc->nr_to_write <= 0; } pagevec_release(&pvec); cond_resched(); @@ -2288,11 +2540,25 @@ retry: index = 0; goto retry; } - if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0)) - mapping->writeback_index = index; return ret; } -#endif + +static void flush_epd_write_bio(struct extent_page_data *epd) +{ + if (epd->bio) { + if (epd->sync_io) + submit_one_bio(WRITE_SYNC, epd->bio, 0, 0); + else + submit_one_bio(WRITE, epd->bio, 0, 0); + epd->bio = NULL; + } +} + +static noinline void flush_write_bio(void *data) +{ + struct extent_page_data *epd = data; + flush_epd_write_bio(epd); +} int extent_write_full_page(struct extent_io_tree *tree, struct page *page, get_extent_t *get_extent, @@ -2304,27 +2570,70 @@ int extent_write_full_page(struct extent_io_tree *tree, struct page *page, .bio = NULL, .tree = tree, .get_extent = get_extent, + .extent_locked = 0, + .sync_io = wbc->sync_mode == WB_SYNC_ALL, }; struct writeback_control wbc_writepages = { .bdi = wbc->bdi, - .sync_mode = WB_SYNC_NONE, + .sync_mode = wbc->sync_mode, .older_than_this = NULL, .nr_to_write = 64, .range_start = page_offset(page) + PAGE_CACHE_SIZE, .range_end = (loff_t)-1, }; - ret = __extent_writepage(page, wbc, &epd); - write_cache_pages(mapping, &wbc_writepages, __extent_writepage, &epd); - if (epd.bio) { - submit_one_bio(WRITE, epd.bio, 0); - } + extent_write_cache_pages(tree, mapping, &wbc_writepages, + __extent_writepage, &epd, flush_write_bio); + flush_epd_write_bio(&epd); return ret; } -EXPORT_SYMBOL(extent_write_full_page); +int extent_write_locked_range(struct extent_io_tree *tree, struct inode *inode, + u64 start, u64 end, get_extent_t *get_extent, + int mode) +{ + int ret = 0; + struct address_space *mapping = inode->i_mapping; + struct page *page; + unsigned long nr_pages = (end - start + PAGE_CACHE_SIZE) >> + PAGE_CACHE_SHIFT; + + struct extent_page_data epd = { + .bio = NULL, + .tree = tree, + .get_extent = get_extent, + .extent_locked = 1, + .sync_io = mode == WB_SYNC_ALL, + }; + struct writeback_control wbc_writepages = { + .bdi = inode->i_mapping->backing_dev_info, + .sync_mode = mode, + .older_than_this = NULL, + .nr_to_write = nr_pages * 2, + .range_start = start, + .range_end = end + 1, + }; + + while (start <= end) { + page = find_get_page(mapping, start >> PAGE_CACHE_SHIFT); + if (clear_page_dirty_for_io(page)) + ret = __extent_writepage(page, &wbc_writepages, &epd); + else { + if (tree->ops && tree->ops->writepage_end_io_hook) + tree->ops->writepage_end_io_hook(page, start, + start + PAGE_CACHE_SIZE - 1, + NULL, 1); + unlock_page(page); + } + page_cache_release(page); + start += PAGE_CACHE_SIZE; + } + + flush_epd_write_bio(&epd); + return ret; +} int extent_writepages(struct extent_io_tree *tree, struct address_space *mapping, @@ -2336,15 +2645,16 @@ int extent_writepages(struct extent_io_tree *tree, .bio = NULL, .tree = tree, .get_extent = get_extent, + .extent_locked = 0, + .sync_io = wbc->sync_mode == WB_SYNC_ALL, }; - ret = write_cache_pages(mapping, wbc, __extent_writepage, &epd); - if (epd.bio) { - submit_one_bio(WRITE, epd.bio, 0); - } + ret = extent_write_cache_pages(tree, mapping, wbc, + __extent_writepage, &epd, + flush_write_bio); + flush_epd_write_bio(&epd); return ret; } -EXPORT_SYMBOL(extent_writepages); int extent_readpages(struct extent_io_tree *tree, struct address_space *mapping, @@ -2354,6 +2664,7 @@ int extent_readpages(struct extent_io_tree *tree, struct bio *bio = NULL; unsigned page_idx; struct pagevec pvec; + unsigned long bio_flags = 0; pagevec_init(&pvec, 0); for (page_idx = 0; page_idx < nr_pages; page_idx++) { @@ -2371,20 +2682,19 @@ int extent_readpages(struct extent_io_tree *tree, /* open coding of lru_cache_add, also not exported */ page_cache_get(page); if (!pagevec_add(&pvec, page)) - __pagevec_lru_add(&pvec); + __pagevec_lru_add_file(&pvec); __extent_read_full_page(tree, page, get_extent, - &bio, 0); + &bio, 0, &bio_flags); } page_cache_release(page); } if (pagevec_count(&pvec)) - __pagevec_lru_add(&pvec); + __pagevec_lru_add_file(&pvec); BUG_ON(!list_empty(pages)); if (bio) - submit_one_bio(READ, bio, 0); + submit_one_bio(READ, bio, 0, bio_flags); return 0; } -EXPORT_SYMBOL(extent_readpages); /* * basic invalidatepage code, this waits on any locked or writeback @@ -2398,18 +2708,18 @@ 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 += (offset + blocksize - 1) & ~(blocksize - 1); if (start > end) return 0; lock_extent(tree, start, end, GFP_NOFS); - wait_on_extent_writeback(tree, start, end); + wait_on_page_writeback(page); clear_extent_bit(tree, start, end, - EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC, - 1, 1, GFP_NOFS); + EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC | + EXTENT_DO_ACCOUNTING, + 1, 1, NULL, GFP_NOFS); return 0; } -EXPORT_SYMBOL(extent_invalidatepage); /* * simple commit_write call, set_range_dirty is used to mark both @@ -2430,7 +2740,6 @@ int extent_commit_write(struct extent_io_tree *tree, } return 0; } -EXPORT_SYMBOL(extent_commit_write); int extent_prepare_write(struct extent_io_tree *tree, struct inode *inode, struct page *page, @@ -2459,12 +2768,12 @@ int extent_prepare_write(struct extent_io_tree *tree, orig_block_start = block_start; lock_extent(tree, page_start, page_end, GFP_NOFS); - while(block_start <= block_end) { + while (block_start <= block_end) { em = get_extent(inode, page, page_offset, block_start, block_end - block_start + 1, 1); - if (IS_ERR(em) || !em) { + if (IS_ERR(em) || !em) goto err; - } + cur_end = min(block_end, extent_map_end(em) - 1); block_off_start = block_start & (PAGE_CACHE_SIZE - 1); block_off_end = block_off_start + blocksize; @@ -2488,7 +2797,7 @@ int extent_prepare_write(struct extent_io_tree *tree, !isnew && !PageUptodate(page) && (block_off_end > to || block_off_start < from) && !test_range_bit(tree, block_start, cur_end, - EXTENT_UPTODATE, 1)) { + EXTENT_UPTODATE, 1, NULL)) { u64 sector; u64 extent_offset = block_start - em->start; size_t iosize; @@ -2502,11 +2811,12 @@ int extent_prepare_write(struct extent_io_tree *tree, */ set_extent_bit(tree, block_start, block_start + iosize - 1, - EXTENT_LOCKED, 0, NULL, GFP_NOFS); + EXTENT_LOCKED, 0, NULL, NULL, GFP_NOFS); ret = submit_extent_page(READ, tree, page, sector, iosize, page_offset, em->bdev, NULL, 1, - end_bio_extent_preparewrite, 0); + end_bio_extent_preparewrite, 0, + 0, 0); iocount++; block_start = block_start + iosize; } else { @@ -2527,7 +2837,6 @@ err: /* FIXME, zero out newly allocated blocks on error */ return err; } -EXPORT_SYMBOL(extent_prepare_write); /* * a helper for releasepage, this tests for areas of the page that @@ -2543,17 +2852,21 @@ int try_release_extent_state(struct extent_map_tree *map, int ret = 1; if (test_range_bit(tree, start, end, - EXTENT_IOBITS | EXTENT_ORDERED, 0)) + EXTENT_IOBITS, 0, NULL)) ret = 0; else { if ((mask & GFP_NOFS) == GFP_NOFS) mask = GFP_NOFS; - clear_extent_bit(tree, start, end, EXTENT_UPTODATE, - 1, 1, mask); + /* + * at this point we can safely clear everything except the + * locked bit and the nodatasum bit + */ + clear_extent_bit(tree, start, end, + ~(EXTENT_LOCKED | EXTENT_NODATASUM), + 0, 0, NULL, mask); } return ret; } -EXPORT_SYMBOL(try_release_extent_state); /* * a helper for releasepage. As long as there are no locked extents @@ -2573,26 +2886,28 @@ int try_release_extent_mapping(struct extent_map_tree *map, u64 len; while (start <= end) { len = end - start + 1; - spin_lock(&map->lock); + write_lock(&map->lock); em = lookup_extent_mapping(map, start, len); if (!em || IS_ERR(em)) { - spin_unlock(&map->lock); + write_unlock(&map->lock); break; } - if (em->start != start) { - spin_unlock(&map->lock); + if (test_bit(EXTENT_FLAG_PINNED, &em->flags) || + em->start != start) { + write_unlock(&map->lock); free_extent_map(em); break; } if (!test_range_bit(tree, em->start, extent_map_end(em) - 1, - EXTENT_LOCKED, 0)) { + EXTENT_LOCKED | EXTENT_WRITEBACK, + 0, NULL)) { remove_extent_mapping(map, em); /* once for the rb tree */ free_extent_map(em); } start = extent_map_end(em); - spin_unlock(&map->lock); + write_unlock(&map->lock); /* once for us */ free_extent_map(em); @@ -2600,7 +2915,6 @@ int try_release_extent_mapping(struct extent_map_tree *map, } return try_release_extent_state(map, tree, page, mask); } -EXPORT_SYMBOL(try_release_extent_mapping); sector_t extent_bmap(struct address_space *mapping, sector_t iblock, get_extent_t *get_extent) @@ -2608,14 +2922,18 @@ sector_t extent_bmap(struct address_space *mapping, sector_t iblock, struct inode *inode = mapping->host; u64 start = iblock << inode->i_blkbits; sector_t sector = 0; + size_t blksize = (1 << inode->i_blkbits); struct extent_map *em; - em = get_extent(inode, NULL, 0, start, (1 << inode->i_blkbits), 0); + lock_extent(&BTRFS_I(inode)->io_tree, start, start + blksize - 1, + GFP_NOFS); + em = get_extent(inode, NULL, 0, start, blksize, 0); + unlock_extent(&BTRFS_I(inode)->io_tree, start, start + blksize - 1, + GFP_NOFS); if (!em || IS_ERR(em)) return 0; - if (em->block_start == EXTENT_MAP_INLINE || - em->block_start == EXTENT_MAP_HOLE) + if (em->block_start > EXTENT_MAP_LAST_BYTE) goto out; sector = (em->block_start + start - em->start) >> inode->i_blkbits; @@ -2624,49 +2942,90 @@ out: return sector; } -static int add_lru(struct extent_io_tree *tree, struct extent_buffer *eb) +int extent_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo, + __u64 start, __u64 len, get_extent_t *get_extent) { - if (list_empty(&eb->lru)) { - extent_buffer_get(eb); - list_add(&eb->lru, &tree->buffer_lru); - tree->lru_size++; - if (tree->lru_size >= BUFFER_LRU_MAX) { - struct extent_buffer *rm; - rm = list_entry(tree->buffer_lru.prev, - struct extent_buffer, lru); - tree->lru_size--; - list_del_init(&rm->lru); - free_extent_buffer(rm); + int ret; + u64 off = start; + u64 max = start + len; + u32 flags = 0; + u64 disko = 0; + struct extent_map *em = NULL; + int end = 0; + u64 em_start = 0, em_len = 0; + unsigned long emflags; + ret = 0; + + if (len == 0) + return -EINVAL; + + lock_extent(&BTRFS_I(inode)->io_tree, start, start + len, + GFP_NOFS); + em = get_extent(inode, NULL, 0, off, max - off, 0); + if (!em) + goto out; + if (IS_ERR(em)) { + ret = PTR_ERR(em); + goto out; + } + while (!end) { + off = em->start + em->len; + if (off >= max) + end = 1; + + em_start = em->start; + em_len = em->len; + + disko = 0; + flags = 0; + + if (em->block_start == EXTENT_MAP_LAST_BYTE) { + end = 1; + flags |= FIEMAP_EXTENT_LAST; + } else if (em->block_start == EXTENT_MAP_HOLE) { + flags |= FIEMAP_EXTENT_UNWRITTEN; + } else if (em->block_start == EXTENT_MAP_INLINE) { + flags |= (FIEMAP_EXTENT_DATA_INLINE | + FIEMAP_EXTENT_NOT_ALIGNED); + } else if (em->block_start == EXTENT_MAP_DELALLOC) { + flags |= (FIEMAP_EXTENT_DELALLOC | + FIEMAP_EXTENT_UNKNOWN); + } else { + disko = em->block_start; } - } else - list_move(&eb->lru, &tree->buffer_lru); - return 0; -} -static struct extent_buffer *find_lru(struct extent_io_tree *tree, - u64 start, unsigned long len) -{ - struct list_head *lru = &tree->buffer_lru; - struct list_head *cur = lru->next; - struct extent_buffer *eb; + if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) + flags |= FIEMAP_EXTENT_ENCODED; - if (list_empty(lru)) - return NULL; + emflags = em->flags; + free_extent_map(em); + em = NULL; - do { - eb = list_entry(cur, struct extent_buffer, lru); - if (eb->start == start && eb->len == len) { - extent_buffer_get(eb); - return eb; + if (!end) { + em = get_extent(inode, NULL, 0, off, max - off, 0); + if (!em) + goto out; + if (IS_ERR(em)) { + ret = PTR_ERR(em); + goto out; + } + emflags = em->flags; + } + if (test_bit(EXTENT_FLAG_VACANCY, &emflags)) { + flags |= FIEMAP_EXTENT_LAST; + end = 1; } - cur = cur->next; - } while (cur != lru); - return NULL; -} -static inline unsigned long num_extent_pages(u64 start, u64 len) -{ - return ((start + len + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT) - - (start >> PAGE_CACHE_SHIFT); + ret = fiemap_fill_next_extent(fieinfo, em_start, disko, + em_len, flags); + if (ret) + goto out_free; + } +out_free: + free_extent_map(em); +out: + unlock_extent(&BTRFS_I(inode)->io_tree, start, start + len, + GFP_NOFS); + return ret; } static inline struct page *extent_buffer_page(struct extent_buffer *eb, @@ -2679,50 +3038,25 @@ static inline struct page *extent_buffer_page(struct extent_buffer *eb, return eb->first_page; i += eb->start >> PAGE_CACHE_SHIFT; mapping = eb->first_page->mapping; - read_lock_irq(&mapping->tree_lock); - p = radix_tree_lookup(&mapping->page_tree, i); - read_unlock_irq(&mapping->tree_lock); - return p; -} + if (!mapping) + return NULL; -int release_extent_buffer_tail_pages(struct extent_buffer *eb) -{ - unsigned long num_pages = num_extent_pages(eb->start, eb->len); - struct page *page; - unsigned long i; + /* + * extent_buffer_page is only called after pinning the page + * by increasing the reference count. So we know the page must + * be in the radix tree. + */ + rcu_read_lock(); + p = radix_tree_lookup(&mapping->page_tree, i); + rcu_read_unlock(); - if (num_pages == 1) - return 0; - for (i = 1; i < num_pages; i++) { - page = extent_buffer_page(eb, i); - page_cache_release(page); - } - return 0; + return p; } - -int invalidate_extent_lru(struct extent_io_tree *tree, u64 start, - unsigned long len) +static inline unsigned long num_extent_pages(u64 start, u64 len) { - struct list_head *lru = &tree->buffer_lru; - struct list_head *cur = lru->next; - struct extent_buffer *eb; - int found = 0; - - spin_lock(&tree->lru_lock); - if (list_empty(lru)) - goto out; - - do { - eb = list_entry(cur, struct extent_buffer, lru); - if (eb->start <= start && eb->start + eb->len > start) { - eb->flags &= ~EXTENT_UPTODATE; - } - cur = cur->next; - } while (cur != lru); -out: - spin_unlock(&tree->lru_lock); - return found; + return ((start + len + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT) - + (start >> PAGE_CACHE_SHIFT); } static struct extent_buffer *__alloc_extent_buffer(struct extent_io_tree *tree, @@ -2731,22 +3065,21 @@ static struct extent_buffer *__alloc_extent_buffer(struct extent_io_tree *tree, gfp_t mask) { struct extent_buffer *eb = NULL; +#if LEAK_DEBUG unsigned long flags; - - spin_lock(&tree->lru_lock); - eb = find_lru(tree, start, len); - spin_unlock(&tree->lru_lock); - if (eb) { - return eb; - } +#endif eb = kmem_cache_zalloc(extent_buffer_cache, mask); - INIT_LIST_HEAD(&eb->lru); eb->start = start; eb->len = len; + spin_lock_init(&eb->lock); + init_waitqueue_head(&eb->lock_wq); + +#if LEAK_DEBUG spin_lock_irqsave(&leak_lock, flags); list_add(&eb->leak_list, &buffers); spin_unlock_irqrestore(&leak_lock, flags); +#endif atomic_set(&eb->refs, 1); return eb; @@ -2754,10 +3087,12 @@ static struct extent_buffer *__alloc_extent_buffer(struct extent_io_tree *tree, static void __free_extent_buffer(struct extent_buffer *eb) { +#if LEAK_DEBUG unsigned long flags; spin_lock_irqsave(&leak_lock, flags); list_del(&eb->leak_list); spin_unlock_irqrestore(&leak_lock, flags); +#endif kmem_cache_free(extent_buffer_cache, eb); } @@ -2770,17 +3105,25 @@ struct extent_buffer *alloc_extent_buffer(struct extent_io_tree *tree, unsigned long i; unsigned long index = start >> PAGE_CACHE_SHIFT; struct extent_buffer *eb; + struct extent_buffer *exists = NULL; struct page *p; struct address_space *mapping = tree->mapping; int uptodate = 1; + spin_lock(&tree->buffer_lock); + eb = buffer_search(tree, start); + if (eb) { + atomic_inc(&eb->refs); + spin_unlock(&tree->buffer_lock); + mark_page_accessed(eb->first_page); + return eb; + } + spin_unlock(&tree->buffer_lock); + eb = __alloc_extent_buffer(tree, start, len, mask); if (!eb) return NULL; - if (eb->flags & EXTENT_BUFFER_FILLED) - goto lru_add; - if (page0) { eb->first_page = page0; i = 1; @@ -2797,7 +3140,7 @@ struct extent_buffer *alloc_extent_buffer(struct extent_io_tree *tree, p = find_or_create_page(mapping, index, mask | __GFP_HIGHMEM); if (!p) { WARN_ON(1); - goto fail; + goto free_eb; } set_page_extent_mapped(p); mark_page_accessed(p); @@ -2812,169 +3155,93 @@ struct extent_buffer *alloc_extent_buffer(struct extent_io_tree *tree, unlock_page(p); } if (uptodate) - eb->flags |= EXTENT_UPTODATE; - eb->flags |= EXTENT_BUFFER_FILLED; + set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags); -lru_add: - spin_lock(&tree->lru_lock); - add_lru(tree, eb); - spin_unlock(&tree->lru_lock); + spin_lock(&tree->buffer_lock); + exists = buffer_tree_insert(tree, start, &eb->rb_node); + if (exists) { + /* add one reference for the caller */ + atomic_inc(&exists->refs); + spin_unlock(&tree->buffer_lock); + goto free_eb; + } + spin_unlock(&tree->buffer_lock); + + /* add one reference for the tree */ + atomic_inc(&eb->refs); return eb; -fail: - spin_lock(&tree->lru_lock); - list_del_init(&eb->lru); - spin_unlock(&tree->lru_lock); +free_eb: if (!atomic_dec_and_test(&eb->refs)) - return NULL; - for (index = 1; index < i; index++) { + return exists; + for (index = 1; index < i; index++) page_cache_release(extent_buffer_page(eb, index)); - } - if (i > 0) - page_cache_release(extent_buffer_page(eb, 0)); + page_cache_release(extent_buffer_page(eb, 0)); __free_extent_buffer(eb); - return NULL; + return exists; } -EXPORT_SYMBOL(alloc_extent_buffer); struct extent_buffer *find_extent_buffer(struct extent_io_tree *tree, u64 start, unsigned long len, gfp_t mask) { - unsigned long num_pages = num_extent_pages(start, len); - unsigned long i; - unsigned long index = start >> PAGE_CACHE_SHIFT; struct extent_buffer *eb; - struct page *p; - struct address_space *mapping = tree->mapping; - int uptodate = 1; - - eb = __alloc_extent_buffer(tree, start, len, mask); - if (!eb) - return NULL; - - if (eb->flags & EXTENT_BUFFER_FILLED) - goto lru_add; - - for (i = 0; i < num_pages; i++, index++) { - p = find_get_page(mapping, index); - if (!p) { - goto fail; - } - if (TestSetPageLocked(p)) { - page_cache_release(p); - goto fail; - } - - set_page_extent_mapped(p); - mark_page_accessed(p); - if (i == 0) { - eb->first_page = p; - set_page_extent_head(p, len); - } else { - set_page_private(p, EXTENT_PAGE_PRIVATE); - } + spin_lock(&tree->buffer_lock); + eb = buffer_search(tree, start); + if (eb) + atomic_inc(&eb->refs); + spin_unlock(&tree->buffer_lock); - if (!PageUptodate(p)) - uptodate = 0; - unlock_page(p); - } - if (uptodate) - eb->flags |= EXTENT_UPTODATE; - eb->flags |= EXTENT_BUFFER_FILLED; + if (eb) + mark_page_accessed(eb->first_page); -lru_add: - spin_lock(&tree->lru_lock); - add_lru(tree, eb); - spin_unlock(&tree->lru_lock); return eb; -fail: - spin_lock(&tree->lru_lock); - list_del_init(&eb->lru); - spin_unlock(&tree->lru_lock); - if (!atomic_dec_and_test(&eb->refs)) - return NULL; - for (index = 1; index < i; index++) { - page_cache_release(extent_buffer_page(eb, index)); - } - if (i > 0) - page_cache_release(extent_buffer_page(eb, 0)); - __free_extent_buffer(eb); - return NULL; } -EXPORT_SYMBOL(find_extent_buffer); void free_extent_buffer(struct extent_buffer *eb) { - unsigned long i; - unsigned long num_pages; - if (!eb) return; if (!atomic_dec_and_test(&eb->refs)) return; - WARN_ON(!list_empty(&eb->lru)); - num_pages = num_extent_pages(eb->start, eb->len); - - for (i = 1; i < num_pages; i++) { - page_cache_release(extent_buffer_page(eb, i)); - } - page_cache_release(extent_buffer_page(eb, 0)); - __free_extent_buffer(eb); + WARN_ON(1); } -EXPORT_SYMBOL(free_extent_buffer); int clear_extent_buffer_dirty(struct extent_io_tree *tree, struct extent_buffer *eb) { - int set; unsigned long i; unsigned long num_pages; struct page *page; - u64 start = eb->start; - u64 end = start + eb->len - 1; - - set = clear_extent_dirty(tree, start, end, GFP_NOFS); num_pages = num_extent_pages(eb->start, eb->len); for (i = 0; i < num_pages; i++) { page = extent_buffer_page(eb, i); + if (!PageDirty(page)) + continue; + + lock_page(page); if (i == 0) set_page_extent_head(page, eb->len); else set_page_private(page, EXTENT_PAGE_PRIVATE); - /* - * if we're on the last page or the first page and the - * block isn't aligned on a page boundary, do extra checks - * to make sure we don't clean page that is partially dirty - */ - if ((i == 0 && (eb->start & (PAGE_CACHE_SIZE - 1))) || - ((i == num_pages - 1) && - ((eb->start + eb->len) & (PAGE_CACHE_SIZE - 1)))) { - start = (u64)page->index << PAGE_CACHE_SHIFT; - end = start + PAGE_CACHE_SIZE - 1; - if (test_range_bit(tree, start, end, - EXTENT_DIRTY, 0)) { - continue; - } - } clear_page_dirty_for_io(page); - read_lock_irq(&page->mapping->tree_lock); + spin_lock_irq(&page->mapping->tree_lock); if (!PageDirty(page)) { radix_tree_tag_clear(&page->mapping->page_tree, page_index(page), PAGECACHE_TAG_DIRTY); } - read_unlock_irq(&page->mapping->tree_lock); + spin_unlock_irq(&page->mapping->tree_lock); + unlock_page(page); } return 0; } -EXPORT_SYMBOL(clear_extent_buffer_dirty); int wait_on_extent_buffer_writeback(struct extent_io_tree *tree, struct extent_buffer *eb) @@ -2982,34 +3249,20 @@ int wait_on_extent_buffer_writeback(struct extent_io_tree *tree, return wait_on_extent_writeback(tree, eb->start, eb->start + eb->len - 1); } -EXPORT_SYMBOL(wait_on_extent_buffer_writeback); int set_extent_buffer_dirty(struct extent_io_tree *tree, struct extent_buffer *eb) { unsigned long i; unsigned long num_pages; + int was_dirty = 0; + was_dirty = test_and_set_bit(EXTENT_BUFFER_DIRTY, &eb->bflags); num_pages = num_extent_pages(eb->start, eb->len); - for (i = 0; i < num_pages; i++) { - struct page *page = extent_buffer_page(eb, i); - /* writepage may need to do something special for the - * first page, we have to make sure page->private is - * properly set. releasepage may drop page->private - * on us if the page isn't already dirty. - */ - if (i == 0) { - set_page_extent_head(page, eb->len); - } else if (PagePrivate(page) && - page->private != EXTENT_PAGE_PRIVATE) { - set_page_extent_mapped(page); - } + for (i = 0; i < num_pages; i++) __set_page_dirty_nobuffers(extent_buffer_page(eb, i)); - } - return set_extent_dirty(tree, eb->start, - eb->start + eb->len - 1, GFP_NOFS); + return was_dirty; } -EXPORT_SYMBOL(set_extent_buffer_dirty); int clear_extent_buffer_uptodate(struct extent_io_tree *tree, struct extent_buffer *eb) @@ -3019,13 +3272,14 @@ int clear_extent_buffer_uptodate(struct extent_io_tree *tree, unsigned long num_pages; num_pages = num_extent_pages(eb->start, eb->len); - eb->flags &= ~EXTENT_UPTODATE; + clear_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags); clear_extent_uptodate(tree, eb->start, eb->start + eb->len - 1, GFP_NOFS); for (i = 0; i < num_pages; i++) { page = extent_buffer_page(eb, i); - ClearPageUptodate(page); + if (page) + ClearPageUptodate(page); } return 0; } @@ -3053,7 +3307,6 @@ int set_extent_buffer_uptodate(struct extent_io_tree *tree, } return 0; } -EXPORT_SYMBOL(set_extent_buffer_uptodate); int extent_range_uptodate(struct extent_io_tree *tree, u64 start, u64 end) @@ -3064,10 +3317,10 @@ int extent_range_uptodate(struct extent_io_tree *tree, int uptodate; unsigned long index; - ret = test_range_bit(tree, start, end, EXTENT_UPTODATE, 1); + ret = test_range_bit(tree, start, end, EXTENT_UPTODATE, 1, NULL); if (ret) return 1; - while(start <= end) { + while (start <= end) { index = start >> PAGE_CACHE_SHIFT; page = find_get_page(tree->mapping, index); uptodate = PageUptodate(page); @@ -3090,11 +3343,11 @@ int extent_buffer_uptodate(struct extent_io_tree *tree, struct page *page; int pg_uptodate = 1; - if (eb->flags & EXTENT_UPTODATE) + if (test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags)) return 1; ret = test_range_bit(tree, eb->start, eb->start + eb->len - 1, - EXTENT_UPTODATE, 1); + EXTENT_UPTODATE, 1, NULL); if (ret) return ret; @@ -3108,7 +3361,6 @@ int extent_buffer_uptodate(struct extent_io_tree *tree, } return pg_uptodate; } -EXPORT_SYMBOL(extent_buffer_uptodate); int read_extent_buffer_pages(struct extent_io_tree *tree, struct extent_buffer *eb, @@ -3125,12 +3377,13 @@ int read_extent_buffer_pages(struct extent_io_tree *tree, int inc_all_pages = 0; unsigned long num_pages; struct bio *bio = NULL; + unsigned long bio_flags = 0; - if (eb->flags & EXTENT_UPTODATE) + if (test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags)) return 0; if (test_range_bit(tree, eb->start, eb->start + eb->len - 1, - EXTENT_UPTODATE, 1)) { + EXTENT_UPTODATE, 1, NULL)) { return 0; } @@ -3146,19 +3399,18 @@ int read_extent_buffer_pages(struct extent_io_tree *tree, for (i = start_i; i < num_pages; i++) { page = extent_buffer_page(eb, i); if (!wait) { - if (TestSetPageLocked(page)) + if (!trylock_page(page)) goto unlock_exit; } else { lock_page(page); } locked_pages++; - if (!PageUptodate(page)) { + if (!PageUptodate(page)) all_uptodate = 0; - } } if (all_uptodate) { if (start_i == 0) - eb->flags |= EXTENT_UPTODATE; + set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags); goto unlock_exit; } @@ -3172,35 +3424,34 @@ int read_extent_buffer_pages(struct extent_io_tree *tree, ClearPageError(page); err = __extent_read_full_page(tree, page, get_extent, &bio, - mirror_num); - if (err) { + mirror_num, &bio_flags); + if (err) ret = err; - } } else { unlock_page(page); } } if (bio) - submit_one_bio(READ, bio, mirror_num); + submit_one_bio(READ, bio, mirror_num, bio_flags); - if (ret || !wait) { + if (ret || !wait) return ret; - } + for (i = start_i; i < num_pages; i++) { page = extent_buffer_page(eb, i); wait_on_page_locked(page); - if (!PageUptodate(page)) { + if (!PageUptodate(page)) ret = -EIO; - } } + if (!ret) - eb->flags |= EXTENT_UPTODATE; + set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags); return ret; unlock_exit: i = start_i; - while(locked_pages > 0) { + while (locked_pages > 0) { page = extent_buffer_page(eb, i); i++; unlock_page(page); @@ -3208,7 +3459,6 @@ unlock_exit: } return ret; } -EXPORT_SYMBOL(read_extent_buffer_pages); void read_extent_buffer(struct extent_buffer *eb, void *dstv, unsigned long start, @@ -3227,7 +3477,7 @@ void read_extent_buffer(struct extent_buffer *eb, void *dstv, offset = (start_offset + start) & ((unsigned long)PAGE_CACHE_SIZE - 1); - while(len > 0) { + while (len > 0) { page = extent_buffer_page(eb, i); cur = min(len, (PAGE_CACHE_SIZE - offset)); @@ -3241,7 +3491,6 @@ void read_extent_buffer(struct extent_buffer *eb, void *dstv, i++; } } -EXPORT_SYMBOL(read_extent_buffer); int map_private_extent_buffer(struct extent_buffer *eb, unsigned long start, unsigned long min_len, char **token, char **map, @@ -3266,8 +3515,11 @@ int map_private_extent_buffer(struct extent_buffer *eb, unsigned long start, offset = 0; *map_start = ((u64)i << PAGE_CACHE_SHIFT) - start_offset; } + if (start + min_len > eb->len) { -printk("bad mapping eb start %Lu len %lu, wanted %lu %lu\n", eb->start, eb->len, start, min_len); + printk(KERN_ERR "btrfs bad mapping eb start %llu len %lu, " + "wanted %lu %lu\n", (unsigned long long)eb->start, + eb->len, start, min_len); WARN_ON(1); } @@ -3278,7 +3530,6 @@ printk("bad mapping eb start %Lu len %lu, wanted %lu %lu\n", eb->start, eb->len, *map_len = PAGE_CACHE_SIZE - offset; return 0; } -EXPORT_SYMBOL(map_private_extent_buffer); int map_extent_buffer(struct extent_buffer *eb, unsigned long start, unsigned long min_len, @@ -3303,13 +3554,11 @@ int map_extent_buffer(struct extent_buffer *eb, unsigned long start, } return err; } -EXPORT_SYMBOL(map_extent_buffer); void unmap_extent_buffer(struct extent_buffer *eb, char *token, int km) { kunmap_atomic(token, km); } -EXPORT_SYMBOL(unmap_extent_buffer); int memcmp_extent_buffer(struct extent_buffer *eb, const void *ptrv, unsigned long start, @@ -3329,7 +3578,7 @@ int memcmp_extent_buffer(struct extent_buffer *eb, const void *ptrv, offset = (start_offset + start) & ((unsigned long)PAGE_CACHE_SIZE - 1); - while(len > 0) { + while (len > 0) { page = extent_buffer_page(eb, i); cur = min(len, (PAGE_CACHE_SIZE - offset)); @@ -3347,7 +3596,6 @@ int memcmp_extent_buffer(struct extent_buffer *eb, const void *ptrv, } return ret; } -EXPORT_SYMBOL(memcmp_extent_buffer); void write_extent_buffer(struct extent_buffer *eb, const void *srcv, unsigned long start, unsigned long len) @@ -3365,7 +3613,7 @@ void write_extent_buffer(struct extent_buffer *eb, const void *srcv, offset = (start_offset + start) & ((unsigned long)PAGE_CACHE_SIZE - 1); - while(len > 0) { + while (len > 0) { page = extent_buffer_page(eb, i); WARN_ON(!PageUptodate(page)); @@ -3380,7 +3628,6 @@ void write_extent_buffer(struct extent_buffer *eb, const void *srcv, i++; } } -EXPORT_SYMBOL(write_extent_buffer); void memset_extent_buffer(struct extent_buffer *eb, char c, unsigned long start, unsigned long len) @@ -3397,7 +3644,7 @@ void memset_extent_buffer(struct extent_buffer *eb, char c, offset = (start_offset + start) & ((unsigned long)PAGE_CACHE_SIZE - 1); - while(len > 0) { + while (len > 0) { page = extent_buffer_page(eb, i); WARN_ON(!PageUptodate(page)); @@ -3411,7 +3658,6 @@ void memset_extent_buffer(struct extent_buffer *eb, char c, i++; } } -EXPORT_SYMBOL(memset_extent_buffer); void copy_extent_buffer(struct extent_buffer *dst, struct extent_buffer *src, unsigned long dst_offset, unsigned long src_offset, @@ -3430,7 +3676,7 @@ void copy_extent_buffer(struct extent_buffer *dst, struct extent_buffer *src, offset = (start_offset + dst_offset) & ((unsigned long)PAGE_CACHE_SIZE - 1); - while(len > 0) { + while (len > 0) { page = extent_buffer_page(dst, i); WARN_ON(!PageUptodate(page)); @@ -3446,7 +3692,6 @@ void copy_extent_buffer(struct extent_buffer *dst, struct extent_buffer *src, i++; } } -EXPORT_SYMBOL(copy_extent_buffer); static void move_pages(struct page *dst_page, struct page *src_page, unsigned long dst_off, unsigned long src_off, @@ -3497,17 +3742,17 @@ void memcpy_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset, unsigned long src_i; if (src_offset + len > dst->len) { - printk("memmove bogus src_offset %lu move len %lu len %lu\n", - src_offset, len, dst->len); + printk(KERN_ERR "btrfs memmove bogus src_offset %lu move " + "len %lu dst len %lu\n", src_offset, len, dst->len); BUG_ON(1); } if (dst_offset + len > dst->len) { - printk("memmove bogus dst_offset %lu move len %lu len %lu\n", - dst_offset, len, dst->len); + printk(KERN_ERR "btrfs memmove bogus dst_offset %lu move " + "len %lu dst len %lu\n", dst_offset, len, dst->len); BUG_ON(1); } - while(len > 0) { + while (len > 0) { dst_off_in_page = (start_offset + dst_offset) & ((unsigned long)PAGE_CACHE_SIZE - 1); src_off_in_page = (start_offset + src_offset) & @@ -3530,7 +3775,6 @@ void memcpy_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset, len -= cur; } } -EXPORT_SYMBOL(memcpy_extent_buffer); void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset, unsigned long src_offset, unsigned long len) @@ -3545,20 +3789,20 @@ void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset, unsigned long src_i; if (src_offset + len > dst->len) { - printk("memmove bogus src_offset %lu move len %lu len %lu\n", - src_offset, len, dst->len); + printk(KERN_ERR "btrfs memmove bogus src_offset %lu move " + "len %lu len %lu\n", src_offset, len, dst->len); BUG_ON(1); } if (dst_offset + len > dst->len) { - printk("memmove bogus dst_offset %lu move len %lu len %lu\n", - dst_offset, len, dst->len); + printk(KERN_ERR "btrfs memmove bogus dst_offset %lu move " + "len %lu len %lu\n", dst_offset, len, dst->len); BUG_ON(1); } if (dst_offset < src_offset) { memcpy_extent_buffer(dst, dst_offset, src_offset, len); return; } - while(len > 0) { + while (len > 0) { dst_i = (start_offset + dst_end) >> PAGE_CACHE_SHIFT; src_i = (start_offset + src_end) >> PAGE_CACHE_SHIFT; @@ -3579,4 +3823,35 @@ void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset, len -= cur; } } -EXPORT_SYMBOL(memmove_extent_buffer); + +int try_release_extent_buffer(struct extent_io_tree *tree, struct page *page) +{ + u64 start = page_offset(page); + struct extent_buffer *eb; + int ret = 1; + unsigned long i; + unsigned long num_pages; + + spin_lock(&tree->buffer_lock); + eb = buffer_search(tree, start); + if (!eb) + goto out; + + if (atomic_read(&eb->refs) > 1) { + ret = 0; + goto out; + } + if (test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) { + ret = 0; + goto out; + } + /* at this point we can safely release the extent buffer */ + num_pages = num_extent_pages(eb->start, eb->len); + for (i = 0; i < num_pages; i++) + page_cache_release(extent_buffer_page(eb, i)); + rb_erase(&eb->rb_node, &tree->buffer); + __free_extent_buffer(eb); +out: + spin_unlock(&tree->buffer_lock); + return ret; +}