From d1310b2e0cd98eb1348553e69b73827b436dca7b Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 24 Jan 2008 16:13:08 -0500 Subject: Btrfs: Split the extent_map code into two parts There is now extent_map for mapping offsets in the file to disk and extent_io for state tracking, IO submission and extent_bufers. The new extent_map code shifts from [start,end] pairs to [start,len], and pushes the locking out into the caller. This allows a few performance optimizations and is easier to use. A number of extent_map usage bugs were fixed, mostly with failing to remove extent_map entries when changing the file. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 3089 ++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 3089 insertions(+) create mode 100644 fs/btrfs/extent_io.c (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c new file mode 100644 index 000000000000..15cc158a0498 --- /dev/null +++ b/fs/btrfs/extent_io.c @@ -0,0 +1,3089 @@ +#include +#include +#include +#include +#include +#include +#include +#include +#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)); + +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 state_lock = SPIN_LOCK_UNLOCKED; +#define BUFFER_LRU_MAX 64 + +struct tree_entry { + u64 start; + u64 end; + int in_tree; + struct rb_node rb_node; +}; + +struct extent_page_data { + struct bio *bio; + struct extent_io_tree *tree; + get_extent_t *get_extent; +}; + +int __init extent_io_init(void) +{ + extent_state_cache = btrfs_cache_create("extent_state", + sizeof(struct extent_state), 0, + NULL); + if (!extent_state_cache) + return -ENOMEM; + + extent_buffer_cache = btrfs_cache_create("extent_buffers", + sizeof(struct extent_buffer), 0, + NULL); + if (!extent_buffer_cache) + goto free_state_cache; + return 0; + +free_state_cache: + kmem_cache_destroy(extent_state_cache); + return -ENOMEM; +} + +void extent_io_exit(void) +{ + struct extent_state *state; + + while (!list_empty(&states)) { + state = list_entry(states.next, struct extent_state, list); + printk("state leak: start %Lu end %Lu state %lu in tree %d refs %d\n", state->start, state->end, state->state, state->in_tree, atomic_read(&state->refs)); + list_del(&state->list); + kmem_cache_free(extent_state_cache, state); + + } + + if (extent_state_cache) + kmem_cache_destroy(extent_state_cache); + if (extent_buffer_cache) + kmem_cache_destroy(extent_buffer_cache); +} + +void extent_io_tree_init(struct extent_io_tree *tree, + struct address_space *mapping, gfp_t mask) +{ + tree->state.rb_node = NULL; + tree->ops = NULL; + tree->dirty_bytes = 0; + rwlock_init(&tree->lock); + spin_lock_init(&tree->lru_lock); + tree->mapping = mapping; + INIT_LIST_HEAD(&tree->buffer_lru); + tree->lru_size = 0; +} +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) +{ + struct extent_state *state; + unsigned long flags; + + state = kmem_cache_alloc(extent_state_cache, mask); + if (!state || IS_ERR(state)) + return state; + state->state = 0; + state->in_tree = 0; + state->private = 0; + + spin_lock_irqsave(&state_lock, flags); + list_add(&state->list, &states); + spin_unlock_irqrestore(&state_lock, flags); + + 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) +{ + unsigned long flags; + if (!state) + return; + if (atomic_dec_and_test(&state->refs)) { + WARN_ON(state->in_tree); + spin_lock_irqsave(&state_lock, flags); + list_del(&state->list); + spin_unlock_irqrestore(&state_lock, flags); + 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 tree_entry *entry; + + while(*p) { + parent = *p; + entry = rb_entry(parent, struct tree_entry, rb_node); + + if (offset < entry->start) + p = &(*p)->rb_left; + else if (offset > entry->end) + p = &(*p)->rb_right; + else + return parent; + } + + entry = rb_entry(node, struct tree_entry, rb_node); + entry->in_tree = 1; + rb_link_node(node, parent, p); + rb_insert_color(node, root); + return NULL; +} + +static struct rb_node *__tree_search(struct rb_root *root, u64 offset, + struct rb_node **prev_ret, + struct rb_node **next_ret) +{ + 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; + + while(n) { + entry = rb_entry(n, struct tree_entry, rb_node); + prev = n; + prev_entry = entry; + + if (offset < entry->start) + n = n->rb_left; + else if (offset > entry->end) + n = n->rb_right; + else + return n; + } + + if (prev_ret) { + orig_prev = prev; + while(prev && offset > prev_entry->end) { + prev = rb_next(prev); + prev_entry = rb_entry(prev, struct tree_entry, rb_node); + } + *prev_ret = prev; + prev = orig_prev; + } + + if (next_ret) { + prev_entry = rb_entry(prev, struct tree_entry, rb_node); + while(prev && offset < prev_entry->start) { + prev = rb_prev(prev); + prev_entry = rb_entry(prev, struct tree_entry, rb_node); + } + *next_ret = prev; + } + return NULL; +} + +static inline struct rb_node *tree_search(struct rb_root *root, u64 offset) +{ + struct rb_node *prev; + struct rb_node *ret; + ret = __tree_search(root, offset, &prev, NULL); + if (!ret) + return prev; + return ret; +} + +/* + * utility function to look for merge candidates inside a given range. + * Any extents with matching state are merged together into a single + * extent in the tree. Extents with EXTENT_IO in their state field + * are not merged because the end_io handlers need to be able to do + * operations on them without sleeping (or doing allocations/splits). + * + * This should be called with the tree lock held. + */ +static int merge_state(struct extent_io_tree *tree, + struct extent_state *state) +{ + struct extent_state *other; + struct rb_node *other_node; + + if (state->state & EXTENT_IOBITS) + return 0; + + other_node = rb_prev(&state->rb_node); + if (other_node) { + other = rb_entry(other_node, struct extent_state, rb_node); + if (other->end == state->start - 1 && + other->state == state->state) { + state->start = other->start; + other->in_tree = 0; + rb_erase(&other->rb_node, &tree->state); + free_extent_state(other); + } + } + other_node = rb_next(&state->rb_node); + if (other_node) { + other = rb_entry(other_node, struct extent_state, rb_node); + if (other->start == state->end + 1 && + other->state == state->state) { + other->start = state->start; + state->in_tree = 0; + rb_erase(&state->rb_node, &tree->state); + free_extent_state(state); + } + } + return 0; +} + +/* + * insert an extent_state struct into the tree. 'bits' are set on the + * struct before it is inserted. + * + * This may return -EEXIST if the extent is already there, in which case the + * state struct is freed. + * + * The tree lock is not taken internally. This is a utility function and + * probably isn't what you want to call (see set/clear_extent_bit). + */ +static int insert_state(struct extent_io_tree *tree, + struct extent_state *state, u64 start, u64 end, + int bits) +{ + struct rb_node *node; + + if (end < start) { + printk("end < start %Lu %Lu\n", end, start); + WARN_ON(1); + } + if (bits & EXTENT_DIRTY) + tree->dirty_bytes += end - start + 1; + 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); + free_extent_state(state); + return -EEXIST; + } + merge_state(tree, state); + return 0; +} + +/* + * split a given extent state struct in two, inserting the preallocated + * struct 'prealloc' as the newly created second half. 'split' indicates an + * offset inside 'orig' where it should be split. + * + * Before calling, + * the tree has 'orig' at [orig->start, orig->end]. After calling, there + * are two extent state structs in the tree: + * prealloc: [orig->start, split - 1] + * orig: [ split, orig->end ] + * + * The tree locks are not taken by this function. They need to be held + * by the caller. + */ +static int split_state(struct extent_io_tree *tree, struct extent_state *orig, + struct extent_state *prealloc, u64 split) +{ + struct rb_node *node; + prealloc->start = orig->start; + prealloc->end = split - 1; + prealloc->state = orig->state; + orig->start = split; + + 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; + } + return 0; +} + +/* + * utility function to clear some bits in an extent state struct. + * it will optionally wake up any one waiting on this state (wake == 1), or + * forcibly remove the state from the tree (delete == 1). + * + * If no bits are set on the state struct after clearing things, the + * struct is freed and removed from the tree + */ +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; + + if ((bits & EXTENT_DIRTY) && (state->state & EXTENT_DIRTY)) { + u64 range = state->end - state->start + 1; + WARN_ON(range > tree->dirty_bytes); + tree->dirty_bytes -= range; + } + state->state &= ~bits; + if (wake) + wake_up(&state->wq); + if (delete || state->state == 0) { + if (state->in_tree) { + rb_erase(&state->rb_node, &tree->state); + state->in_tree = 0; + free_extent_state(state); + } else { + WARN_ON(1); + } + } else { + merge_state(tree, state); + } + return ret; +} + +/* + * clear some bits on a range in the tree. This may require splitting + * or inserting elements in the tree, so the gfp mask is used to + * indicate which allocations or sleeping are allowed. + * + * pass 'wake' == 1 to kick any sleepers, and 'delete' == 1 to remove + * the given range from the tree regardless of state (ie for truncate). + * + * the range [start, end] is inclusive. + * + * This takes the tree lock, and returns < 0 on error, > 0 if any of the + * 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) +{ + struct extent_state *state; + struct extent_state *prealloc = NULL; + struct rb_node *node; + unsigned long flags; + int err; + int set = 0; + +again: + if (!prealloc && (mask & __GFP_WAIT)) { + prealloc = alloc_extent_state(mask); + if (!prealloc) + return -ENOMEM; + } + + write_lock_irqsave(&tree->lock, flags); + /* + * this search will find the extents that end after + * our range starts + */ + node = tree_search(&tree->state, start); + if (!node) + goto out; + state = rb_entry(node, struct extent_state, rb_node); + if (state->start > end) + goto out; + WARN_ON(state->end < start); + + /* + * | ---- desired range ---- | + * | state | or + * | ------------- state -------------- | + * + * We need to split the extent we found, and may flip + * bits on second half. + * + * If the extent we found extends past our range, we + * just split and search again. It'll get split again + * the next time though. + * + * If the extent we found is inside our range, we clear + * the desired bit on it. + */ + + if (state->start < start) { + err = split_state(tree, state, prealloc, start); + BUG_ON(err == -EEXIST); + prealloc = NULL; + 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; + } + goto search_again; + } + /* + * | ---- desired range ---- | + * | state | + * We need to split the extent, and clear the bit + * on the first half + */ + if (state->start <= end && state->end > end) { + 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); + prealloc = NULL; + goto out; + } + + start = state->end + 1; + set |= clear_state_bit(tree, state, bits, wake, delete); + goto search_again; + +out: + write_unlock_irqrestore(&tree->lock, flags); + if (prealloc) + free_extent_state(prealloc); + + return set; + +search_again: + if (start > end) + goto out; + write_unlock_irqrestore(&tree->lock, flags); + 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) +{ + DEFINE_WAIT(wait); + prepare_to_wait(&state->wq, &wait, TASK_UNINTERRUPTIBLE); + read_unlock_irq(&tree->lock); + schedule(); + read_lock_irq(&tree->lock); + finish_wait(&state->wq, &wait); + return 0; +} + +/* + * waits for one or more bits to clear on a range in the state tree. + * The range [start, end] is inclusive. + * The tree lock is taken by this function + */ +int wait_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, int bits) +{ + struct extent_state *state; + struct rb_node *node; + + read_lock_irq(&tree->lock); +again: + while (1) { + /* + * this search will find all the extents that end after + * our range starts + */ + node = tree_search(&tree->state, start); + if (!node) + break; + + state = rb_entry(node, struct extent_state, rb_node); + + if (state->start > end) + goto out; + + if (state->state & bits) { + start = state->start; + atomic_inc(&state->refs); + wait_on_state(tree, state); + free_extent_state(state); + goto again; + } + start = state->end + 1; + + if (start > end) + break; + + if (need_resched()) { + read_unlock_irq(&tree->lock); + cond_resched(); + read_lock_irq(&tree->lock); + } + } +out: + read_unlock_irq(&tree->lock); + return 0; +} +EXPORT_SYMBOL(wait_extent_bit); + +static void set_state_bits(struct extent_io_tree *tree, + struct extent_state *state, + int bits) +{ + if ((bits & EXTENT_DIRTY) && !(state->state & EXTENT_DIRTY)) { + u64 range = state->end - state->start + 1; + tree->dirty_bytes += range; + } + state->state |= bits; +} + +/* + * 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. + * + * [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) +{ + 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); + if (!prealloc) + return -ENOMEM; + } + + write_lock_irqsave(&tree->lock, flags); + /* + * this search will find all the extents that end after + * our range starts. + */ + node = tree_search(&tree->state, start); + if (!node) { + err = insert_state(tree, prealloc, start, end, bits); + prealloc = NULL; + BUG_ON(err == -EEXIST); + goto out; + } + + state = rb_entry(node, struct extent_state, rb_node); + last_start = state->start; + last_end = state->end; + + /* + * | ---- desired range ---- | + * | state | + * + * Just lock what we found and keep going + */ + if (state->start == start && state->end <= end) { + set = state->state & bits; + if (set && exclusive) { + *failed_start = state->start; + err = -EEXIST; + goto out; + } + set_state_bits(tree, state, bits); + start = state->end + 1; + merge_state(tree, state); + goto search_again; + } + + /* + * | ---- desired range ---- | + * | state | + * or + * | ------------- state -------------- | + * + * We need to split the extent we found, and may flip bits on + * second half. + * + * If the extent we found extends past our + * range, we just split and search again. It'll get split + * again the next time though. + * + * If the extent we found is inside our range, we set the + * desired bit on it. + */ + if (state->start < start) { + set = state->state & bits; + if (exclusive && set) { + *failed_start = start; + err = -EEXIST; + goto out; + } + err = split_state(tree, state, prealloc, start); + BUG_ON(err == -EEXIST); + prealloc = NULL; + if (err) + goto out; + if (state->end <= end) { + set_state_bits(tree, state, bits); + start = state->end + 1; + merge_state(tree, state); + } else { + start = state->start; + } + goto search_again; + } + /* + * | ---- desired range ---- | + * | state | or | state | + * + * There's a hole, we need to insert something in it and + * ignore the extent we found. + */ + if (state->start > start) { + u64 this_end; + if (end < last_start) + this_end = end; + else + this_end = last_start -1; + err = insert_state(tree, prealloc, start, this_end, + bits); + prealloc = NULL; + BUG_ON(err == -EEXIST); + if (err) + goto out; + start = this_end + 1; + goto search_again; + } + /* + * | ---- desired range ---- | + * | state | + * We need to split the extent, and set the bit + * on the first half + */ + if (state->start <= end && state->end > end) { + set = state->state & bits; + if (exclusive && set) { + *failed_start = start; + err = -EEXIST; + goto out; + } + err = split_state(tree, state, prealloc, end + 1); + BUG_ON(err == -EEXIST); + + set_state_bits(tree, prealloc, bits); + merge_state(tree, prealloc); + prealloc = NULL; + goto out; + } + + goto search_again; + +out: + write_unlock_irqrestore(&tree->lock, flags); + if (prealloc) + free_extent_state(prealloc); + + return err; + +search_again: + if (start > end) + goto out; + write_unlock_irqrestore(&tree->lock, flags); + 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_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); +} +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); +} +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); +} +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); +} +EXPORT_SYMBOL(clear_extent_dirty); + +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); +} +EXPORT_SYMBOL(set_extent_new); + +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); +} +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); +} +EXPORT_SYMBOL(set_extent_uptodate); + +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); +} +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); + +/* + * locks a range in ascending order, waiting for any locked regions + * it hits on the way. [start,end] are inclusive, and this will sleep. + */ +int lock_extent(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask) +{ + int err; + u64 failed_start; + while (1) { + err = set_extent_bit(tree, start, end, EXTENT_LOCKED, 1, + &failed_start, mask); + if (err == -EEXIST && (mask & __GFP_WAIT)) { + wait_extent_bit(tree, failed_start, end, EXTENT_LOCKED); + start = failed_start; + } else { + break; + } + WARN_ON(start > end); + } + return err; +} +EXPORT_SYMBOL(lock_extent); + +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); +} +EXPORT_SYMBOL(unlock_extent); + +/* + * helper function to set pages and extents in the tree dirty + */ +int set_range_dirty(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); + BUG_ON(!page); + __set_page_dirty_nobuffers(page); + 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) +{ + 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); + BUG_ON(!page); + set_page_writeback(page); + page_cache_release(page); + index++; + } + set_extent_writeback(tree, start, end, GFP_NOFS); + return 0; +} +EXPORT_SYMBOL(set_range_writeback); + +int find_first_extent_bit(struct extent_io_tree *tree, u64 start, + u64 *start_ret, u64 *end_ret, int bits) +{ + struct rb_node *node; + struct extent_state *state; + int ret = 1; + + read_lock_irq(&tree->lock); + /* + * this search will find all the extents that end after + * our range starts. + */ + node = tree_search(&tree->state, start); + if (!node || IS_ERR(node)) { + goto out; + } + + while(1) { + state = rb_entry(node, struct extent_state, rb_node); + if (state->end >= start && (state->state & bits)) { + *start_ret = state->start; + *end_ret = state->end; + ret = 0; + break; + } + node = rb_next(node); + if (!node) + break; + } +out: + read_unlock_irq(&tree->lock); + return ret; +} +EXPORT_SYMBOL(find_first_extent_bit); + +u64 find_lock_delalloc_range(struct extent_io_tree *tree, + u64 *start, u64 *end, u64 max_bytes) +{ + struct rb_node *node; + struct extent_state *state; + u64 cur_start = *start; + u64 found = 0; + u64 total_bytes = 0; + + write_lock_irq(&tree->lock); + /* + * this search will find all the extents that end after + * our range starts. + */ +search_again: + node = tree_search(&tree->state, cur_start); + if (!node || IS_ERR(node)) { + *end = (u64)-1; + goto out; + } + + while(1) { + state = rb_entry(node, struct extent_state, rb_node); + if (found && state->start != cur_start) { + goto out; + } + if (!(state->state & EXTENT_DELALLOC)) { + if (!found) + *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); + write_unlock_irq(&tree->lock); + schedule(); + write_lock_irq(&tree->lock); + finish_wait(&state->wq, &wait); + free_extent_state(state); + goto search_again; + } + state->state |= EXTENT_LOCKED; + if (!found) + *start = state->start; + found++; + *end = state->end; + cur_start = state->end + 1; + node = rb_next(node); + if (!node) + break; + total_bytes += state->end - state->start + 1; + if (total_bytes >= max_bytes) + break; + } +out: + write_unlock_irq(&tree->lock); + return found; +} + +u64 count_range_bits(struct extent_io_tree *tree, + u64 *start, u64 search_end, u64 max_bytes, + unsigned long bits) +{ + struct rb_node *node; + struct extent_state *state; + u64 cur_start = *start; + u64 total_bytes = 0; + int found = 0; + + if (search_end <= cur_start) { + printk("search_end %Lu start %Lu\n", search_end, cur_start); + WARN_ON(1); + return 0; + } + + write_lock_irq(&tree->lock); + if (cur_start == 0 && bits == EXTENT_DIRTY) { + total_bytes = tree->dirty_bytes; + goto out; + } + /* + * this search will find all the extents that end after + * our range starts. + */ + node = tree_search(&tree->state, cur_start); + if (!node || IS_ERR(node)) { + goto out; + } + + while(1) { + state = rb_entry(node, struct extent_state, rb_node); + if (state->start > search_end) + break; + if (state->end >= cur_start && (state->state & bits)) { + total_bytes += min(search_end, state->end) + 1 - + max(cur_start, state->start); + if (total_bytes >= max_bytes) + break; + if (!found) { + *start = state->start; + found = 1; + } + } + node = rb_next(node); + if (!node) + break; + } +out: + write_unlock_irq(&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. + */ +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; + + write_lock_irq(&tree->lock); + /* + * this search will find all the extents that end after + * our range starts. + */ + node = tree_search(&tree->state, start); + if (!node || IS_ERR(node)) { + ret = -ENOENT; + goto out; + } + state = rb_entry(node, struct extent_state, rb_node); + if (state->start != start) { + ret = -ENOENT; + goto out; + } + state->private = private; +out: + write_unlock_irq(&tree->lock); + return ret; +} + +int get_state_private(struct extent_io_tree *tree, u64 start, u64 *private) +{ + struct rb_node *node; + struct extent_state *state; + int ret = 0; + + read_lock_irq(&tree->lock); + /* + * this search will find all the extents that end after + * our range starts. + */ + node = tree_search(&tree->state, start); + if (!node || IS_ERR(node)) { + ret = -ENOENT; + goto out; + } + state = rb_entry(node, struct extent_state, rb_node); + if (state->start != start) { + ret = -ENOENT; + goto out; + } + *private = state->private; +out: + read_unlock_irq(&tree->lock); + return ret; +} + +/* + * searches a range in the state tree for a given mask. + * If 'filled' == 1, this returns 1 only if ever extent in the tree + * has the bits set. Otherwise, 1 is returned if any bit in the + * range is found set. + */ +int test_range_bit(struct extent_io_tree *tree, u64 start, u64 end, + int bits, int filled) +{ + struct extent_state *state = NULL; + struct rb_node *node; + int bitset = 0; + unsigned long flags; + + read_lock_irqsave(&tree->lock, flags); + node = tree_search(&tree->state, start); + while (node && start <= end) { + state = rb_entry(node, struct extent_state, rb_node); + + if (filled && state->start > start) { + bitset = 0; + break; + } + + if (state->start > end) + break; + + if (state->state & bits) { + bitset = 1; + if (!filled) + break; + } else if (filled) { + bitset = 0; + break; + } + start = state->end + 1; + if (start > end) + break; + node = rb_next(node); + if (!node) { + if (filled) + bitset = 0; + break; + } + } + read_unlock_irqrestore(&tree->lock, flags); + return bitset; +} +EXPORT_SYMBOL(test_range_bit); + +/* + * helper function to set a given page up to date if all the + * extents in the tree for that page are up to date + */ +static int check_page_uptodate(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_UPTODATE, 1)) + SetPageUptodate(page); + return 0; +} + +/* + * helper function to unlock a page if all the extents in the tree + * for that page are unlocked + */ +static int check_page_locked(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_LOCKED, 0)) + unlock_page(page); + return 0; +} + +/* + * helper function to end page writeback if all the extents + * in the tree for that page are done with writeback + */ +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); + return 0; +} + +/* lots and lots of room for performance fixes in the end_bio funcs */ + +/* + * after a writepage IO is done, we need to: + * clear the uptodate bits on error + * clear the writeback bits in the extent tree for this IO + * end_page_writeback if the page has no more pending 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_writepage(struct bio *bio, int err) +#else +static int end_bio_extent_writepage(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_io_tree *tree = bio->bi_private; + u64 start; + u64 end; + int whole_page; + +#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) + if (bio->bi_size) + return 1; +#endif + + do { + struct page *page = bvec->bv_page; + start = ((u64)page->index << PAGE_CACHE_SHIFT) + + bvec->bv_offset; + end = start + bvec->bv_len - 1; + + if (bvec->bv_offset == 0 && bvec->bv_len == PAGE_CACHE_SIZE) + whole_page = 1; + else + whole_page = 0; + + if (--bvec >= bio->bi_io_vec) + prefetchw(&bvec->bv_page->flags); + + if (!uptodate) { + clear_extent_uptodate(tree, start, end, GFP_ATOMIC); + ClearPageUptodate(page); + SetPageError(page); + } + clear_extent_writeback(tree, start, end, GFP_ATOMIC); + + if (whole_page) + end_page_writeback(page); + else + check_page_writeback(tree, page); + if (tree->ops && tree->ops->writepage_end_io_hook) + tree->ops->writepage_end_io_hook(page, start, end); + } while (bvec >= bio->bi_io_vec); + + bio_put(bio); +#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) + return 0; +#endif +} + +/* + * after a readpage IO is done, we need to: + * clear the uptodate bits on error + * set the uptodate bits if things worked + * set the page up to date if all extents in the tree are uptodate + * clear the lock bit in the extent tree + * unlock the page if there are no other extents locked for it + * + * 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_io_tree *tree = bio->bi_private; + u64 start; + u64 end; + int whole_page; + int ret; + +#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) + if (bio->bi_size) + return 1; +#endif + + do { + struct page *page = bvec->bv_page; + start = ((u64)page->index << PAGE_CACHE_SHIFT) + + bvec->bv_offset; + end = start + bvec->bv_len - 1; + + if (bvec->bv_offset == 0 && bvec->bv_len == PAGE_CACHE_SIZE) + whole_page = 1; + else + whole_page = 0; + + if (--bvec >= bio->bi_io_vec) + prefetchw(&bvec->bv_page->flags); + + if (uptodate && tree->ops && tree->ops->readpage_end_io_hook) { + ret = tree->ops->readpage_end_io_hook(page, start, end); + if (ret) + uptodate = 0; + } + if (uptodate) { + set_extent_uptodate(tree, start, end, GFP_ATOMIC); + if (whole_page) + SetPageUptodate(page); + else + check_page_uptodate(tree, page); + } else { + ClearPageUptodate(page); + SetPageError(page); + } + + unlock_extent(tree, start, end, GFP_ATOMIC); + + if (whole_page) + unlock_page(page); + else + check_page_locked(tree, page); + } while (bvec >= bio->bi_io_vec); + + bio_put(bio); +#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) + return 0; +#endif +} + +/* + * IO done from prepare_write is pretty simple, we just unlock + * 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_io_tree *tree = bio->bi_private; + 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; + start = ((u64)page->index << PAGE_CACHE_SHIFT) + + bvec->bv_offset; + end = start + bvec->bv_len - 1; + + if (--bvec >= bio->bi_io_vec) + prefetchw(&bvec->bv_page->flags); + + if (uptodate) { + set_extent_uptodate(tree, start, end, GFP_ATOMIC); + } else { + ClearPageUptodate(page); + SetPageError(page); + } + + unlock_extent(tree, start, end, GFP_ATOMIC); + + } while (bvec >= bio->bi_io_vec); + + bio_put(bio); +#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) + return 0; +#endif +} + +static struct bio * +extent_bio_alloc(struct block_device *bdev, u64 first_sector, int nr_vecs, + gfp_t gfp_flags) +{ + struct bio *bio; + + bio = bio_alloc(gfp_flags, nr_vecs); + + if (bio == NULL && (current->flags & PF_MEMALLOC)) { + while (!bio && (nr_vecs /= 2)) + bio = bio_alloc(gfp_flags, nr_vecs); + } + + if (bio) { + bio->bi_bdev = bdev; + bio->bi_sector = first_sector; + } + return bio; +} + +static int submit_one_bio(int rw, struct bio *bio) +{ + u64 maxsector; + int ret = 0; + + bio_get(bio); + + maxsector = bio->bi_bdev->bd_inode->i_size >> 9; + if (maxsector < bio->bi_sector) { + printk("sector too large max %Lu got %llu\n", maxsector, + (unsigned long long)bio->bi_sector); + WARN_ON(1); + } + + submit_bio(rw, bio); + if (bio_flagged(bio, BIO_EOPNOTSUPP)) + ret = -EOPNOTSUPP; + bio_put(bio); + return ret; +} + +static int submit_extent_page(int rw, struct extent_io_tree *tree, + struct page *page, sector_t sector, + size_t size, unsigned long offset, + struct block_device *bdev, + struct bio **bio_ret, + unsigned long max_pages, + bio_end_io_t end_io_func) +{ + int ret = 0; + struct bio *bio; + int nr; + + if (bio_ret && *bio_ret) { + bio = *bio_ret; + if (bio->bi_sector + (bio->bi_size >> 9) != sector || + bio_add_page(bio, page, size, offset) < size) { + ret = submit_one_bio(rw, bio); + bio = NULL; + } else { + return 0; + } + } + nr = min_t(int, max_pages, 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); + } + bio_add_page(bio, page, size, offset); + bio->bi_end_io = end_io_func; + bio->bi_private = tree; + if (bio_ret) { + *bio_ret = bio; + } else { + ret = submit_one_bio(rw, bio); + } + + return ret; +} + +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); + } +} + +void set_page_extent_head(struct page *page, unsigned long len) +{ + set_page_private(page, EXTENT_PAGE_PRIVATE_FIRST_PAGE | len << 2); +} + +/* + * basic readpage implementation. Locked extent state structs are inserted + * into the tree that are removed when the IO is done (by the end_io + * handlers) + */ +static int __extent_read_full_page(struct extent_io_tree *tree, + struct page *page, + get_extent_t *get_extent, + struct bio **bio) +{ + struct inode *inode = page->mapping->host; + u64 start = (u64)page->index << PAGE_CACHE_SHIFT; + u64 page_end = start + PAGE_CACHE_SIZE - 1; + u64 end; + u64 cur = start; + u64 extent_offset; + u64 last_byte = i_size_read(inode); + u64 block_start; + u64 cur_end; + sector_t sector; + struct extent_map *em; + struct block_device *bdev; + int ret; + int nr = 0; + size_t page_offset = 0; + size_t iosize; + size_t blocksize = inode->i_sb->s_blocksize; + + set_page_extent_mapped(page); + + end = page_end; + lock_extent(tree, start, end, GFP_NOFS); + + while (cur <= end) { + if (cur >= last_byte) { + char *userpage; + iosize = PAGE_CACHE_SIZE - page_offset; + userpage = kmap_atomic(page, KM_USER0); + memset(userpage + page_offset, 0, iosize); + flush_dcache_page(page); + kunmap_atomic(userpage, KM_USER0); + set_extent_uptodate(tree, cur, cur + iosize - 1, + GFP_NOFS); + unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS); + break; + } + em = get_extent(inode, page, page_offset, cur, + end - cur + 1, 0); + if (IS_ERR(em) || !em) { + SetPageError(page); + unlock_extent(tree, cur, end, GFP_NOFS); + break; + } + + extent_offset = cur - em->start; + BUG_ON(extent_map_end(em) <= cur); + BUG_ON(end < cur); + + 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; + bdev = em->bdev; + block_start = em->block_start; + free_extent_map(em); + em = NULL; + + /* we've found a hole, just zero and go on */ + if (block_start == EXTENT_MAP_HOLE) { + char *userpage; + userpage = kmap_atomic(page, KM_USER0); + memset(userpage + page_offset, 0, iosize); + flush_dcache_page(page); + kunmap_atomic(userpage, KM_USER0); + + set_extent_uptodate(tree, cur, cur + iosize - 1, + GFP_NOFS); + unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS); + cur = cur + iosize; + page_offset += iosize; + continue; + } + /* the get_extent function already copied into the page */ + if (test_range_bit(tree, cur, cur_end, EXTENT_UPTODATE, 1)) { + unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS); + cur = cur + iosize; + page_offset += iosize; + continue; + } + + ret = 0; + if (tree->ops && tree->ops->readpage_io_hook) { + ret = tree->ops->readpage_io_hook(page, cur, + cur + iosize - 1); + } + if (!ret) { + unsigned long nr = (last_byte >> PAGE_CACHE_SHIFT) + 1; + nr -= page->index; + ret = submit_extent_page(READ, tree, page, + sector, iosize, page_offset, + bdev, bio, nr, + end_bio_extent_readpage); + } + if (ret) + SetPageError(page); + cur = cur + iosize; + page_offset += iosize; + nr++; + } + if (!nr) { + if (!PageError(page)) + SetPageUptodate(page); + unlock_page(page); + } + return 0; +} + +int extent_read_full_page(struct extent_io_tree *tree, struct page *page, + get_extent_t *get_extent) +{ + struct bio *bio = NULL; + int ret; + + ret = __extent_read_full_page(tree, page, get_extent, &bio); + if (bio) + submit_one_bio(READ, bio); + return ret; +} +EXPORT_SYMBOL(extent_read_full_page); + +/* + * the writepage semantics are similar to regular writepage. extent + * records are inserted to lock ranges in the tree, and as dirty areas + * are found, they are marked writeback. Then the lock bits are removed + * and the end_io handler clears the writeback ranges + */ +static int __extent_writepage(struct page *page, struct writeback_control *wbc, + void *data) +{ + struct inode *inode = page->mapping->host; + struct extent_page_data *epd = data; + struct extent_io_tree *tree = epd->tree; + u64 start = (u64)page->index << PAGE_CACHE_SHIFT; + u64 delalloc_start; + u64 page_end = start + PAGE_CACHE_SIZE - 1; + u64 end; + u64 cur = start; + u64 extent_offset; + u64 last_byte = i_size_read(inode); + u64 block_start; + u64 iosize; + sector_t sector; + struct extent_map *em; + struct block_device *bdev; + int ret; + int nr = 0; + size_t page_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; + + WARN_ON(!PageLocked(page)); + if (page->index > end_index) { + clear_extent_dirty(tree, start, page_end, GFP_NOFS); + unlock_page(page); + return 0; + } + + if (page->index == end_index) { + char *userpage; + + size_t offset = i_size & (PAGE_CACHE_SIZE - 1); + + userpage = kmap_atomic(page, KM_USER0); + memset(userpage + offset, 0, PAGE_CACHE_SIZE - offset); + flush_dcache_page(page); + kunmap_atomic(userpage, KM_USER0); + } + + 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, + &delalloc_end, + 128 * 1024 * 1024); + if (nr_delalloc == 0) { + 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; + } + lock_extent(tree, start, page_end, GFP_NOFS); + + end = page_end; + if (test_range_bit(tree, start, page_end, EXTENT_DELALLOC, 0)) { + printk("found delalloc bits after lock_extent\n"); + } + + if (last_byte <= start) { + clear_extent_dirty(tree, start, page_end, GFP_NOFS); + 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); + break; + } + em = epd->get_extent(inode, page, page_offset, cur, + end - cur + 1, 1); + if (IS_ERR(em) || !em) { + SetPageError(page); + break; + } + + extent_offset = cur - em->start; + BUG_ON(extent_map_end(em) <= cur); + BUG_ON(end < cur); + iosize = min(extent_map_end(em) - cur, end - cur + 1); + iosize = (iosize + blocksize - 1) & ~((u64)blocksize - 1); + sector = (em->block_start + extent_offset) >> 9; + bdev = em->bdev; + block_start = em->block_start; + free_extent_map(em); + em = NULL; + + if (block_start == EXTENT_MAP_HOLE || + block_start == EXTENT_MAP_INLINE) { + clear_extent_dirty(tree, cur, + cur + iosize - 1, GFP_NOFS); + cur = cur + iosize; + page_offset += iosize; + continue; + } + + /* leave this out until we have a page_mkwrite call */ + if (0 && !test_range_bit(tree, cur, cur + iosize - 1, + EXTENT_DIRTY, 0)) { + cur = cur + iosize; + page_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); + } else { + ret = 0; + } + if (ret) + 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, + (unsigned long long)end); + } + + ret = submit_extent_page(WRITE, tree, page, sector, + iosize, page_offset, bdev, + &epd->bio, max_nr, + end_bio_extent_writepage); + if (ret) + SetPageError(page); + } + cur = cur + iosize; + page_offset += iosize; + nr++; + } +done: + if (nr == 0) { + /* make sure the mapping tag for page dirty gets cleared */ + set_page_writeback(page); + end_page_writeback(page); + } + unlock_extent(tree, start, page_end, GFP_NOFS); + unlock_page(page); + return 0; +} + +#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) + +/* 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. + * @mapping: address space structure to write + * @wbc: subtract the number of written pages from *@wbc->nr_to_write + * @writepage: function called for each page + * @data: data passed to writepage function + * + * If a page is already under I/O, write_cache_pages() skips it, even + * if it's dirty. This is desirable behaviour for memory-cleaning writeback, + * but it is INCORRECT for data-integrity system calls such as fsync(). fsync() + * and msync() need to guarantee that all the data which was dirty at the time + * the call was made get new I/O started against them. If wbc->sync_mode is + * 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) +{ + struct backing_dev_info *bdi = mapping->backing_dev_info; + int ret = 0; + int done = 0; + struct pagevec pvec; + int nr_pages; + pgoff_t index; + pgoff_t end; /* Inclusive */ + 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 */ + end = -1; + } else { + index = wbc->range_start >> PAGE_CACHE_SHIFT; + end = wbc->range_end >> PAGE_CACHE_SHIFT; + if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX) + range_whole = 1; + scanned = 1; + } +retry: + while (!done && (index <= end) && + (nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, + PAGECACHE_TAG_DIRTY, + min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1))) { + unsigned i; + + scanned = 1; + for (i = 0; i < nr_pages; i++) { + struct page *page = pvec.pages[i]; + + /* + * At this point we hold neither mapping->tree_lock nor + * lock on the page itself: the page may be truncated or + * invalidated (changing page->mapping to NULL), or even + * swizzled back from swapper_space to tmpfs file + * mapping + */ + lock_page(page); + + if (unlikely(page->mapping != mapping)) { + unlock_page(page); + continue; + } + + if (!wbc->range_cyclic && page->index > end) { + done = 1; + unlock_page(page); + continue; + } + + if (wbc->sync_mode != WB_SYNC_NONE) + wait_on_page_writeback(page); + + if (PageWriteback(page) || + !clear_page_dirty_for_io(page)) { + unlock_page(page); + continue; + } + + ret = (*writepage)(page, wbc, data); + + if (unlikely(ret == AOP_WRITEPAGE_ACTIVATE)) { + 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; + done = 1; + } + } + pagevec_release(&pvec); + cond_resched(); + } + if (!scanned && !done) { + /* + * We hit the last page and there is more work to be done: wrap + * back to the start of the file + */ + scanned = 1; + index = 0; + goto retry; + } + if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0)) + mapping->writeback_index = index; + return ret; +} +#endif + +int extent_write_full_page(struct extent_io_tree *tree, struct page *page, + get_extent_t *get_extent, + struct writeback_control *wbc) +{ + int ret; + struct address_space *mapping = page->mapping; + struct extent_page_data epd = { + .bio = NULL, + .tree = tree, + .get_extent = get_extent, + }; + struct writeback_control wbc_writepages = { + .bdi = wbc->bdi, + .sync_mode = WB_SYNC_NONE, + .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); + } + return ret; +} +EXPORT_SYMBOL(extent_write_full_page); + + +int extent_writepages(struct extent_io_tree *tree, + struct address_space *mapping, + get_extent_t *get_extent, + struct writeback_control *wbc) +{ + int ret = 0; + struct extent_page_data epd = { + .bio = NULL, + .tree = tree, + .get_extent = get_extent, + }; + + ret = write_cache_pages(mapping, wbc, __extent_writepage, &epd); + if (epd.bio) { + submit_one_bio(WRITE, epd.bio); + } + return ret; +} +EXPORT_SYMBOL(extent_writepages); + +int extent_readpages(struct extent_io_tree *tree, + struct address_space *mapping, + struct list_head *pages, unsigned nr_pages, + get_extent_t get_extent) +{ + struct bio *bio = NULL; + unsigned page_idx; + struct pagevec pvec; + + pagevec_init(&pvec, 0); + for (page_idx = 0; page_idx < nr_pages; page_idx++) { + struct page *page = list_entry(pages->prev, struct page, lru); + + prefetchw(&page->flags); + list_del(&page->lru); + /* + * what we want to do here is call add_to_page_cache_lru, + * but that isn't exported, so we reproduce it here + */ + if (!add_to_page_cache(page, mapping, + page->index, GFP_KERNEL)) { + + /* open coding of lru_cache_add, also not exported */ + page_cache_get(page); + if (!pagevec_add(&pvec, page)) + __pagevec_lru_add(&pvec); + __extent_read_full_page(tree, page, get_extent, &bio); + } + page_cache_release(page); + } + if (pagevec_count(&pvec)) + __pagevec_lru_add(&pvec); + BUG_ON(!list_empty(pages)); + if (bio) + submit_one_bio(READ, bio); + return 0; +} +EXPORT_SYMBOL(extent_readpages); + +/* + * basic invalidatepage code, this waits on any locked or writeback + * ranges corresponding to the page, and then deletes any extent state + * records from the tree + */ +int extent_invalidatepage(struct extent_io_tree *tree, + struct page *page, unsigned long offset) +{ + u64 start = ((u64)page->index << PAGE_CACHE_SHIFT); + u64 end = start + PAGE_CACHE_SIZE - 1; + size_t blocksize = page->mapping->host->i_sb->s_blocksize; + + 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); + clear_extent_bit(tree, start, end, + EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC, + 1, 1, GFP_NOFS); + return 0; +} +EXPORT_SYMBOL(extent_invalidatepage); + +/* + * simple commit_write call, set_range_dirty is used to mark both + * the pages and the extent records as dirty + */ +int extent_commit_write(struct extent_io_tree *tree, + struct inode *inode, struct page *page, + unsigned from, unsigned to) +{ + loff_t pos = ((loff_t)page->index << PAGE_CACHE_SHIFT) + to; + + set_page_extent_mapped(page); + set_page_dirty(page); + + if (pos > inode->i_size) { + i_size_write(inode, pos); + mark_inode_dirty(inode); + } + return 0; +} +EXPORT_SYMBOL(extent_commit_write); + +int extent_prepare_write(struct extent_io_tree *tree, + struct inode *inode, struct page *page, + unsigned from, unsigned to, get_extent_t *get_extent) +{ + u64 page_start = (u64)page->index << PAGE_CACHE_SHIFT; + u64 page_end = page_start + PAGE_CACHE_SIZE - 1; + u64 block_start; + u64 orig_block_start; + u64 block_end; + u64 cur_end; + struct extent_map *em; + unsigned blocksize = 1 << inode->i_blkbits; + size_t page_offset = 0; + size_t block_off_start; + size_t block_off_end; + int err = 0; + int iocount = 0; + int ret = 0; + int isnew; + + set_page_extent_mapped(page); + + block_start = (page_start + from) & ~((u64)blocksize - 1); + block_end = (page_start + to - 1) | (blocksize - 1); + orig_block_start = block_start; + + lock_extent(tree, page_start, page_end, GFP_NOFS); + 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) { + 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; + isnew = clear_extent_new(tree, block_start, cur_end, GFP_NOFS); + + if (!PageUptodate(page) && isnew && + (block_off_end > to || block_off_start < from)) { + void *kaddr; + + kaddr = kmap_atomic(page, KM_USER0); + if (block_off_end > to) + memset(kaddr + to, 0, block_off_end - to); + if (block_off_start < from) + memset(kaddr + block_off_start, 0, + from - block_off_start); + flush_dcache_page(page); + kunmap_atomic(kaddr, KM_USER0); + } + if ((em->block_start != EXTENT_MAP_HOLE && + em->block_start != EXTENT_MAP_INLINE) && + !isnew && !PageUptodate(page) && + (block_off_end > to || block_off_start < from) && + !test_range_bit(tree, block_start, cur_end, + EXTENT_UPTODATE, 1)) { + u64 sector; + u64 extent_offset = block_start - em->start; + size_t iosize; + sector = (em->block_start + extent_offset) >> 9; + iosize = (cur_end - block_start + blocksize) & + ~((u64)blocksize - 1); + /* + * we've already got the extent locked, but we + * need to split the state such that our end_bio + * handler can clear the lock. + */ + set_extent_bit(tree, block_start, + block_start + iosize - 1, + EXTENT_LOCKED, 0, NULL, GFP_NOFS); + ret = submit_extent_page(READ, tree, page, + sector, iosize, page_offset, em->bdev, + NULL, 1, + end_bio_extent_preparewrite); + iocount++; + block_start = block_start + iosize; + } else { + set_extent_uptodate(tree, block_start, cur_end, + GFP_NOFS); + unlock_extent(tree, block_start, cur_end, GFP_NOFS); + block_start = cur_end + 1; + } + page_offset = block_start & (PAGE_CACHE_SIZE - 1); + free_extent_map(em); + } + if (iocount) { + wait_extent_bit(tree, orig_block_start, + block_end, EXTENT_LOCKED); + } + check_page_uptodate(tree, page); +err: + /* FIXME, zero out newly allocated blocks on error */ + return err; +} +EXPORT_SYMBOL(extent_prepare_write); + +/* + * a helper for releasepage. As long as there are no locked extents + * in the range corresponding to the page, both state records and extent + * map records are removed + */ +int try_release_extent_mapping(struct extent_map_tree *map, + struct extent_io_tree *tree, struct page *page) +{ + struct extent_map *em; + u64 start = (u64)page->index << PAGE_CACHE_SHIFT; + u64 end = start + PAGE_CACHE_SIZE - 1; + u64 orig_start = start; + int ret = 1; + + while (start <= end) { + spin_lock(&map->lock); + em = lookup_extent_mapping(map, start, end); + if (!em || IS_ERR(em)) { + spin_unlock(&map->lock); + break; + } + if (!test_range_bit(tree, em->start, extent_map_end(em) - 1, + EXTENT_LOCKED, 0)) { + remove_extent_mapping(map, em); + /* once for the rb tree */ + free_extent_map(em); + } + start = extent_map_end(em); + spin_unlock(&map->lock); + + /* once for us */ + free_extent_map(em); + } + if (test_range_bit(tree, orig_start, end, EXTENT_LOCKED, 0)) + ret = 0; + else + clear_extent_bit(tree, orig_start, end, EXTENT_UPTODATE, + 1, 1, GFP_NOFS); + return ret; +} +EXPORT_SYMBOL(try_release_extent_mapping); + +sector_t extent_bmap(struct address_space *mapping, sector_t iblock, + get_extent_t *get_extent) +{ + struct inode *inode = mapping->host; + u64 start = iblock << inode->i_blkbits; + sector_t sector = 0; + struct extent_map *em; + + em = get_extent(inode, NULL, 0, start, (1 << inode->i_blkbits), 0); + if (!em || IS_ERR(em)) + return 0; + + if (em->block_start == EXTENT_MAP_INLINE || + em->block_start == EXTENT_MAP_HOLE) + goto out; + + sector = (em->block_start + start - em->start) >> inode->i_blkbits; +printk("bmap finds %Lu %Lu block %Lu\n", em->start, em->len, em->block_start); +out: + free_extent_map(em); + return sector; +} + +static int add_lru(struct extent_io_tree *tree, struct extent_buffer *eb) +{ + 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); + } + } 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 (list_empty(lru)) + return NULL; + + do { + eb = list_entry(cur, struct extent_buffer, lru); + if (eb->start == start && eb->len == len) { + extent_buffer_get(eb); + return eb; + } + 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); +} + +static inline struct page *extent_buffer_page(struct extent_buffer *eb, + unsigned long i) +{ + struct page *p; + struct address_space *mapping; + + if (i == 0) + 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; +} + +static struct extent_buffer *__alloc_extent_buffer(struct extent_io_tree *tree, + u64 start, + unsigned long len, + gfp_t mask) +{ + struct extent_buffer *eb = NULL; + + spin_lock(&tree->lru_lock); + eb = find_lru(tree, start, len); + spin_unlock(&tree->lru_lock); + if (eb) { + return eb; + } + + eb = kmem_cache_zalloc(extent_buffer_cache, mask); + INIT_LIST_HEAD(&eb->lru); + eb->start = start; + eb->len = len; + atomic_set(&eb->refs, 1); + + return eb; +} + +static void __free_extent_buffer(struct extent_buffer *eb) +{ + kmem_cache_free(extent_buffer_cache, eb); +} + +struct extent_buffer *alloc_extent_buffer(struct extent_io_tree *tree, + u64 start, unsigned long len, + struct page *page0, + 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 || IS_ERR(eb)) + return NULL; + + if (eb->flags & EXTENT_BUFFER_FILLED) + goto lru_add; + + if (page0) { + eb->first_page = page0; + i = 1; + index++; + page_cache_get(page0); + mark_page_accessed(page0); + set_page_extent_mapped(page0); + WARN_ON(!PageUptodate(page0)); + set_page_extent_head(page0, len); + } else { + i = 0; + } + for (; i < num_pages; i++, index++) { + p = find_or_create_page(mapping, index, mask | __GFP_HIGHMEM); + if (!p) { + WARN_ON(1); + 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); + } + if (!PageUptodate(p)) + uptodate = 0; + unlock_page(p); + } + if (uptodate) + eb->flags |= EXTENT_UPTODATE; + eb->flags |= EXTENT_BUFFER_FILLED; + +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(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 || IS_ERR(eb)) + return NULL; + + if (eb->flags & EXTENT_BUFFER_FILLED) + goto lru_add; + + for (i = 0; i < num_pages; i++, index++) { + p = find_lock_page(mapping, index); + if (!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); + } + + if (!PageUptodate(p)) + uptodate = 0; + unlock_page(p); + } + if (uptodate) + eb->flags |= EXTENT_UPTODATE; + eb->flags |= EXTENT_BUFFER_FILLED; + +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); +} +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); + 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)) { + unlock_page(page); + continue; + } + } + clear_page_dirty_for_io(page); + write_lock_irq(&page->mapping->tree_lock); + if (!PageDirty(page)) { + radix_tree_tag_clear(&page->mapping->page_tree, + page_index(page), + PAGECACHE_TAG_DIRTY); + } + write_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) +{ + 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; + + 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) { + lock_page(page); + set_page_extent_head(page, eb->len); + } else if (PagePrivate(page) && + page->private != EXTENT_PAGE_PRIVATE) { + lock_page(page); + set_page_extent_mapped(page); + unlock_page(page); + } + __set_page_dirty_nobuffers(extent_buffer_page(eb, i)); + if (i == 0) + unlock_page(page); + } + return set_extent_dirty(tree, eb->start, + eb->start + eb->len - 1, GFP_NOFS); +} +EXPORT_SYMBOL(set_extent_buffer_dirty); + +int set_extent_buffer_uptodate(struct extent_io_tree *tree, + struct extent_buffer *eb) +{ + unsigned long i; + struct page *page; + unsigned long num_pages; + + num_pages = num_extent_pages(eb->start, eb->len); + + set_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); + if ((i == 0 && (eb->start & (PAGE_CACHE_SIZE - 1))) || + ((i == num_pages - 1) && + ((eb->start + eb->len) & (PAGE_CACHE_SIZE - 1)))) { + check_page_uptodate(tree, page); + continue; + } + SetPageUptodate(page); + } + return 0; +} +EXPORT_SYMBOL(set_extent_buffer_uptodate); + +int extent_buffer_uptodate(struct extent_io_tree *tree, + struct extent_buffer *eb) +{ + if (eb->flags & EXTENT_UPTODATE) + return 1; + return test_range_bit(tree, eb->start, eb->start + eb->len - 1, + EXTENT_UPTODATE, 1); +} +EXPORT_SYMBOL(extent_buffer_uptodate); + +int read_extent_buffer_pages(struct extent_io_tree *tree, + struct extent_buffer *eb, + u64 start, + int wait) +{ + unsigned long i; + unsigned long start_i; + struct page *page; + int err; + int ret = 0; + unsigned long num_pages; + + if (eb->flags & EXTENT_UPTODATE) + return 0; + + if (0 && test_range_bit(tree, eb->start, eb->start + eb->len - 1, + EXTENT_UPTODATE, 1)) { + return 0; + } + + if (start) { + WARN_ON(start < eb->start); + start_i = (start >> PAGE_CACHE_SHIFT) - + (eb->start >> PAGE_CACHE_SHIFT); + } else { + start_i = 0; + } + + num_pages = num_extent_pages(eb->start, eb->len); + for (i = start_i; i < num_pages; i++) { + page = extent_buffer_page(eb, i); + if (PageUptodate(page)) { + continue; + } + if (!wait) { + if (TestSetPageLocked(page)) { + continue; + } + } else { + lock_page(page); + } + if (!PageUptodate(page)) { + err = page->mapping->a_ops->readpage(NULL, page); + if (err) { + ret = err; + } + } else { + unlock_page(page); + } + } + + 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)) { + ret = -EIO; + } + } + if (!ret) + eb->flags |= EXTENT_UPTODATE; + return ret; +} +EXPORT_SYMBOL(read_extent_buffer_pages); + +void read_extent_buffer(struct extent_buffer *eb, void *dstv, + unsigned long start, + unsigned long len) +{ + size_t cur; + size_t offset; + struct page *page; + char *kaddr; + char *dst = (char *)dstv; + size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1); + unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT; + unsigned long num_pages = num_extent_pages(eb->start, eb->len); + + WARN_ON(start > eb->len); + WARN_ON(start + len > eb->start + eb->len); + + offset = (start_offset + start) & ((unsigned long)PAGE_CACHE_SIZE - 1); + + while(len > 0) { + page = extent_buffer_page(eb, i); + if (!PageUptodate(page)) { + printk("page %lu not up to date i %lu, total %lu, len %lu\n", page->index, i, num_pages, eb->len); + WARN_ON(1); + } + WARN_ON(!PageUptodate(page)); + + cur = min(len, (PAGE_CACHE_SIZE - offset)); + kaddr = kmap_atomic(page, KM_USER1); + memcpy(dst, kaddr + offset, cur); + kunmap_atomic(kaddr, KM_USER1); + + dst += cur; + len -= cur; + offset = 0; + 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, + unsigned long *map_start, + unsigned long *map_len, int km) +{ + size_t offset = start & (PAGE_CACHE_SIZE - 1); + char *kaddr; + struct page *p; + size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1); + unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT; + unsigned long end_i = (start_offset + start + min_len - 1) >> + PAGE_CACHE_SHIFT; + + if (i != end_i) + return -EINVAL; + + if (i == 0) { + offset = start_offset; + *map_start = 0; + } else { + 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); + WARN_ON(1); + } + + p = extent_buffer_page(eb, i); + WARN_ON(!PageUptodate(p)); + kaddr = kmap_atomic(p, km); + *token = kaddr; + *map = kaddr + offset; + *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, + char **token, char **map, + unsigned long *map_start, + unsigned long *map_len, int km) +{ + int err; + int save = 0; + if (eb->map_token) { + unmap_extent_buffer(eb, eb->map_token, km); + eb->map_token = NULL; + save = 1; + } + err = map_private_extent_buffer(eb, start, min_len, token, map, + map_start, map_len, km); + if (!err && save) { + eb->map_token = *token; + eb->kaddr = *map; + eb->map_start = *map_start; + eb->map_len = *map_len; + } + 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, + unsigned long len) +{ + size_t cur; + size_t offset; + struct page *page; + char *kaddr; + char *ptr = (char *)ptrv; + size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1); + unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT; + int ret = 0; + + WARN_ON(start > eb->len); + WARN_ON(start + len > eb->start + eb->len); + + offset = (start_offset + start) & ((unsigned long)PAGE_CACHE_SIZE - 1); + + while(len > 0) { + page = extent_buffer_page(eb, i); + WARN_ON(!PageUptodate(page)); + + cur = min(len, (PAGE_CACHE_SIZE - offset)); + + kaddr = kmap_atomic(page, KM_USER0); + ret = memcmp(ptr, kaddr + offset, cur); + kunmap_atomic(kaddr, KM_USER0); + if (ret) + break; + + ptr += cur; + len -= cur; + offset = 0; + i++; + } + return ret; +} +EXPORT_SYMBOL(memcmp_extent_buffer); + +void write_extent_buffer(struct extent_buffer *eb, const void *srcv, + unsigned long start, unsigned long len) +{ + size_t cur; + size_t offset; + struct page *page; + char *kaddr; + char *src = (char *)srcv; + size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1); + unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT; + + WARN_ON(start > eb->len); + WARN_ON(start + len > eb->start + eb->len); + + offset = (start_offset + start) & ((unsigned long)PAGE_CACHE_SIZE - 1); + + while(len > 0) { + page = extent_buffer_page(eb, i); + WARN_ON(!PageUptodate(page)); + + cur = min(len, PAGE_CACHE_SIZE - offset); + kaddr = kmap_atomic(page, KM_USER1); + memcpy(kaddr + offset, src, cur); + kunmap_atomic(kaddr, KM_USER1); + + src += cur; + len -= cur; + offset = 0; + i++; + } +} +EXPORT_SYMBOL(write_extent_buffer); + +void memset_extent_buffer(struct extent_buffer *eb, char c, + unsigned long start, unsigned long len) +{ + size_t cur; + size_t offset; + struct page *page; + char *kaddr; + size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1); + unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT; + + WARN_ON(start > eb->len); + WARN_ON(start + len > eb->start + eb->len); + + offset = (start_offset + start) & ((unsigned long)PAGE_CACHE_SIZE - 1); + + while(len > 0) { + page = extent_buffer_page(eb, i); + WARN_ON(!PageUptodate(page)); + + cur = min(len, PAGE_CACHE_SIZE - offset); + kaddr = kmap_atomic(page, KM_USER0); + memset(kaddr + offset, c, cur); + kunmap_atomic(kaddr, KM_USER0); + + len -= cur; + offset = 0; + 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, + unsigned long len) +{ + u64 dst_len = dst->len; + size_t cur; + size_t offset; + struct page *page; + char *kaddr; + size_t start_offset = dst->start & ((u64)PAGE_CACHE_SIZE - 1); + unsigned long i = (start_offset + dst_offset) >> PAGE_CACHE_SHIFT; + + WARN_ON(src->len != dst_len); + + offset = (start_offset + dst_offset) & + ((unsigned long)PAGE_CACHE_SIZE - 1); + + while(len > 0) { + page = extent_buffer_page(dst, i); + WARN_ON(!PageUptodate(page)); + + cur = min(len, (unsigned long)(PAGE_CACHE_SIZE - offset)); + + kaddr = kmap_atomic(page, KM_USER0); + read_extent_buffer(src, kaddr + offset, src_offset, cur); + kunmap_atomic(kaddr, KM_USER0); + + src_offset += cur; + len -= cur; + offset = 0; + 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, + unsigned long len) +{ + char *dst_kaddr = kmap_atomic(dst_page, KM_USER0); + if (dst_page == src_page) { + memmove(dst_kaddr + dst_off, dst_kaddr + src_off, len); + } else { + char *src_kaddr = kmap_atomic(src_page, KM_USER1); + char *p = dst_kaddr + dst_off + len; + char *s = src_kaddr + src_off + len; + + while (len--) + *--p = *--s; + + kunmap_atomic(src_kaddr, KM_USER1); + } + kunmap_atomic(dst_kaddr, KM_USER0); +} + +static void copy_pages(struct page *dst_page, struct page *src_page, + unsigned long dst_off, unsigned long src_off, + unsigned long len) +{ + char *dst_kaddr = kmap_atomic(dst_page, KM_USER0); + char *src_kaddr; + + if (dst_page != src_page) + src_kaddr = kmap_atomic(src_page, KM_USER1); + else + src_kaddr = dst_kaddr; + + memcpy(dst_kaddr + dst_off, src_kaddr + src_off, len); + kunmap_atomic(dst_kaddr, KM_USER0); + if (dst_page != src_page) + kunmap_atomic(src_kaddr, KM_USER1); +} + +void memcpy_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset, + unsigned long src_offset, unsigned long len) +{ + size_t cur; + size_t dst_off_in_page; + size_t src_off_in_page; + size_t start_offset = dst->start & ((u64)PAGE_CACHE_SIZE - 1); + unsigned long dst_i; + 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); + 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); + BUG_ON(1); + } + + 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) & + ((unsigned long)PAGE_CACHE_SIZE - 1); + + dst_i = (start_offset + dst_offset) >> PAGE_CACHE_SHIFT; + src_i = (start_offset + src_offset) >> PAGE_CACHE_SHIFT; + + cur = min(len, (unsigned long)(PAGE_CACHE_SIZE - + src_off_in_page)); + cur = min_t(unsigned long, cur, + (unsigned long)(PAGE_CACHE_SIZE - dst_off_in_page)); + + copy_pages(extent_buffer_page(dst, dst_i), + extent_buffer_page(dst, src_i), + dst_off_in_page, src_off_in_page, cur); + + src_offset += cur; + dst_offset += cur; + 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) +{ + size_t cur; + size_t dst_off_in_page; + size_t src_off_in_page; + unsigned long dst_end = dst_offset + len - 1; + unsigned long src_end = src_offset + len - 1; + size_t start_offset = dst->start & ((u64)PAGE_CACHE_SIZE - 1); + unsigned long dst_i; + 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); + 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); + BUG_ON(1); + } + if (dst_offset < src_offset) { + memcpy_extent_buffer(dst, dst_offset, src_offset, len); + return; + } + while(len > 0) { + dst_i = (start_offset + dst_end) >> PAGE_CACHE_SHIFT; + src_i = (start_offset + src_end) >> PAGE_CACHE_SHIFT; + + dst_off_in_page = (start_offset + dst_end) & + ((unsigned long)PAGE_CACHE_SIZE - 1); + src_off_in_page = (start_offset + src_end) & + ((unsigned long)PAGE_CACHE_SIZE - 1); + + cur = min_t(unsigned long, len, src_off_in_page + 1); + cur = min(cur, dst_off_in_page + 1); + move_pages(extent_buffer_page(dst, dst_i), + extent_buffer_page(dst, src_i), + dst_off_in_page - cur + 1, + src_off_in_page - cur + 1, cur); + + dst_end -= cur; + src_end -= cur; + len -= cur; + } +} +EXPORT_SYMBOL(memmove_extent_buffer); -- cgit v1.2.2 From 70dec8079d78691e476cc6c7cede40656078ad30 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 29 Jan 2008 09:59:12 -0500 Subject: Btrfs: extent_io and extent_state optimizations The end_bio routines are changed to take a pointer to the extent state struct, and the state tree is walked in order to set/clear appropriate bits as IO completes. This greatly reduces the number of rbtree searches done by the end_bio handlers, and reduces lock contention. The extent_io releasepage function is changed to avoid expensive searches for locked state. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 358 +++++++++++++++++++++++++++++++++++++-------------- 1 file changed, 263 insertions(+), 95 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 15cc158a0498..1f734c34dc24 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -27,13 +27,11 @@ static struct kmem_cache *extent_buffer_cache; static LIST_HEAD(buffers); static LIST_HEAD(states); -static spinlock_t state_lock = SPIN_LOCK_UNLOCKED; #define BUFFER_LRU_MAX 64 struct tree_entry { u64 start; u64 end; - int in_tree; struct rb_node rb_node; }; @@ -69,7 +67,7 @@ void extent_io_exit(void) while (!list_empty(&states)) { state = list_entry(states.next, struct extent_state, list); - printk("state leak: start %Lu end %Lu state %lu in tree %d refs %d\n", state->start, state->end, state->state, state->in_tree, atomic_read(&state->refs)); + 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)); list_del(&state->list); kmem_cache_free(extent_state_cache, state); @@ -87,7 +85,7 @@ void extent_io_tree_init(struct extent_io_tree *tree, tree->state.rb_node = NULL; tree->ops = NULL; tree->dirty_bytes = 0; - rwlock_init(&tree->lock); + spin_lock_init(&tree->lock); spin_lock_init(&tree->lru_lock); tree->mapping = mapping; INIT_LIST_HEAD(&tree->buffer_lru); @@ -110,18 +108,13 @@ EXPORT_SYMBOL(extent_io_tree_empty_lru); struct extent_state *alloc_extent_state(gfp_t mask) { struct extent_state *state; - unsigned long flags; state = kmem_cache_alloc(extent_state_cache, mask); if (!state || IS_ERR(state)) return state; state->state = 0; - state->in_tree = 0; state->private = 0; - - spin_lock_irqsave(&state_lock, flags); - list_add(&state->list, &states); - spin_unlock_irqrestore(&state_lock, flags); + state->tree = NULL; atomic_set(&state->refs, 1); init_waitqueue_head(&state->wq); @@ -131,14 +124,10 @@ EXPORT_SYMBOL(alloc_extent_state); void free_extent_state(struct extent_state *state) { - unsigned long flags; if (!state) return; if (atomic_dec_and_test(&state->refs)) { - WARN_ON(state->in_tree); - spin_lock_irqsave(&state_lock, flags); - list_del(&state->list); - spin_unlock_irqrestore(&state_lock, flags); + WARN_ON(state->tree); kmem_cache_free(extent_state_cache, state); } } @@ -164,7 +153,6 @@ static struct rb_node *tree_insert(struct rb_root *root, u64 offset, } entry = rb_entry(node, struct tree_entry, rb_node); - entry->in_tree = 1; rb_link_node(node, parent, p); rb_insert_color(node, root); return NULL; @@ -216,8 +204,9 @@ static struct rb_node *__tree_search(struct rb_root *root, u64 offset, static inline struct rb_node *tree_search(struct rb_root *root, u64 offset) { - struct rb_node *prev; + struct rb_node *prev = NULL; struct rb_node *ret; + ret = __tree_search(root, offset, &prev, NULL); if (!ret) return prev; @@ -248,7 +237,7 @@ static int merge_state(struct extent_io_tree *tree, if (other->end == state->start - 1 && other->state == state->state) { state->start = other->start; - other->in_tree = 0; + other->tree = NULL; rb_erase(&other->rb_node, &tree->state); free_extent_state(other); } @@ -259,7 +248,7 @@ static int merge_state(struct extent_io_tree *tree, if (other->start == state->end + 1 && other->state == state->state) { other->start = state->start; - state->in_tree = 0; + state->tree = NULL; rb_erase(&state->rb_node, &tree->state); free_extent_state(state); } @@ -300,6 +289,7 @@ static int insert_state(struct extent_io_tree *tree, free_extent_state(state); return -EEXIST; } + state->tree = tree; merge_state(tree, state); return 0; } @@ -335,6 +325,7 @@ static int split_state(struct extent_io_tree *tree, struct extent_state *orig, free_extent_state(prealloc); return -EEXIST; } + prealloc->tree = tree; return 0; } @@ -361,9 +352,9 @@ static int clear_state_bit(struct extent_io_tree *tree, if (wake) wake_up(&state->wq); if (delete || state->state == 0) { - if (state->in_tree) { + if (state->tree) { rb_erase(&state->rb_node, &tree->state); - state->in_tree = 0; + state->tree = NULL; free_extent_state(state); } else { WARN_ON(1); @@ -404,7 +395,7 @@ again: return -ENOMEM; } - write_lock_irqsave(&tree->lock, flags); + spin_lock_irqsave(&tree->lock, flags); /* * this search will find the extents that end after * our range starts @@ -434,6 +425,8 @@ again: */ if (state->start < start) { + if (!prealloc) + prealloc = alloc_extent_state(GFP_ATOMIC); err = split_state(tree, state, prealloc, start); BUG_ON(err == -EEXIST); prealloc = NULL; @@ -455,6 +448,8 @@ again: * on the first half */ if (state->start <= end && state->end > end) { + if (!prealloc) + prealloc = alloc_extent_state(GFP_ATOMIC); err = split_state(tree, state, prealloc, end + 1); BUG_ON(err == -EEXIST); @@ -471,7 +466,7 @@ again: goto search_again; out: - write_unlock_irqrestore(&tree->lock, flags); + spin_unlock_irqrestore(&tree->lock, flags); if (prealloc) free_extent_state(prealloc); @@ -480,7 +475,7 @@ out: search_again: if (start > end) goto out; - write_unlock_irqrestore(&tree->lock, flags); + spin_unlock_irqrestore(&tree->lock, flags); if (mask & __GFP_WAIT) cond_resched(); goto again; @@ -492,9 +487,9 @@ static int wait_on_state(struct extent_io_tree *tree, { DEFINE_WAIT(wait); prepare_to_wait(&state->wq, &wait, TASK_UNINTERRUPTIBLE); - read_unlock_irq(&tree->lock); + spin_unlock_irq(&tree->lock); schedule(); - read_lock_irq(&tree->lock); + spin_lock_irq(&tree->lock); finish_wait(&state->wq, &wait); return 0; } @@ -509,7 +504,7 @@ int wait_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, int bits) struct extent_state *state; struct rb_node *node; - read_lock_irq(&tree->lock); + spin_lock_irq(&tree->lock); again: while (1) { /* @@ -538,13 +533,13 @@ again: break; if (need_resched()) { - read_unlock_irq(&tree->lock); + spin_unlock_irq(&tree->lock); cond_resched(); - read_lock_irq(&tree->lock); + spin_lock_irq(&tree->lock); } } out: - read_unlock_irq(&tree->lock); + spin_unlock_irq(&tree->lock); return 0; } EXPORT_SYMBOL(wait_extent_bit); @@ -589,7 +584,7 @@ again: return -ENOMEM; } - write_lock_irqsave(&tree->lock, flags); + spin_lock_irqsave(&tree->lock, flags); /* * this search will find all the extents that end after * our range starts. @@ -709,7 +704,7 @@ again: goto search_again; out: - write_unlock_irqrestore(&tree->lock, flags); + spin_unlock_irqrestore(&tree->lock, flags); if (prealloc) free_extent_state(prealloc); @@ -718,7 +713,7 @@ out: search_again: if (start > end) goto out; - write_unlock_irqrestore(&tree->lock, flags); + spin_unlock_irqrestore(&tree->lock, flags); if (mask & __GFP_WAIT) cond_resched(); goto again; @@ -817,10 +812,6 @@ int wait_on_extent_writeback(struct extent_io_tree *tree, u64 start, u64 end) } EXPORT_SYMBOL(wait_on_extent_writeback); -/* - * locks a range in ascending order, waiting for any locked regions - * it hits on the way. [start,end] are inclusive, and this will sleep. - */ int lock_extent(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask) { int err; @@ -896,7 +887,7 @@ int find_first_extent_bit(struct extent_io_tree *tree, u64 start, struct extent_state *state; int ret = 1; - read_lock_irq(&tree->lock); + spin_lock_irq(&tree->lock); /* * this search will find all the extents that end after * our range starts. @@ -919,7 +910,7 @@ int find_first_extent_bit(struct extent_io_tree *tree, u64 start, break; } out: - read_unlock_irq(&tree->lock); + spin_unlock_irq(&tree->lock); return ret; } EXPORT_SYMBOL(find_first_extent_bit); @@ -933,7 +924,7 @@ u64 find_lock_delalloc_range(struct extent_io_tree *tree, u64 found = 0; u64 total_bytes = 0; - write_lock_irq(&tree->lock); + spin_lock_irq(&tree->lock); /* * this search will find all the extents that end after * our range starts. @@ -976,9 +967,9 @@ search_again: atomic_inc(&state->refs); prepare_to_wait(&state->wq, &wait, TASK_UNINTERRUPTIBLE); - write_unlock_irq(&tree->lock); + spin_unlock_irq(&tree->lock); schedule(); - write_lock_irq(&tree->lock); + spin_lock_irq(&tree->lock); finish_wait(&state->wq, &wait); free_extent_state(state); goto search_again; @@ -997,7 +988,7 @@ search_again: break; } out: - write_unlock_irq(&tree->lock); + spin_unlock_irq(&tree->lock); return found; } @@ -1017,7 +1008,7 @@ u64 count_range_bits(struct extent_io_tree *tree, return 0; } - write_lock_irq(&tree->lock); + spin_lock_irq(&tree->lock); if (cur_start == 0 && bits == EXTENT_DIRTY) { total_bytes = tree->dirty_bytes; goto out; @@ -1050,7 +1041,7 @@ u64 count_range_bits(struct extent_io_tree *tree, break; } out: - write_unlock_irq(&tree->lock); + spin_unlock_irq(&tree->lock); return total_bytes; } /* @@ -1122,7 +1113,7 @@ int set_state_private(struct extent_io_tree *tree, u64 start, u64 private) struct extent_state *state; int ret = 0; - write_lock_irq(&tree->lock); + spin_lock_irq(&tree->lock); /* * this search will find all the extents that end after * our range starts. @@ -1139,7 +1130,7 @@ int set_state_private(struct extent_io_tree *tree, u64 start, u64 private) } state->private = private; out: - write_unlock_irq(&tree->lock); + spin_unlock_irq(&tree->lock); return ret; } @@ -1149,7 +1140,7 @@ int get_state_private(struct extent_io_tree *tree, u64 start, u64 *private) struct extent_state *state; int ret = 0; - read_lock_irq(&tree->lock); + spin_lock_irq(&tree->lock); /* * this search will find all the extents that end after * our range starts. @@ -1166,13 +1157,13 @@ int get_state_private(struct extent_io_tree *tree, u64 start, u64 *private) } *private = state->private; out: - read_unlock_irq(&tree->lock); + spin_unlock_irq(&tree->lock); return ret; } /* * searches a range in the state tree for a given mask. - * If 'filled' == 1, this returns 1 only if ever extent in the tree + * If 'filled' == 1, this returns 1 only if every extent in the tree * has the bits set. Otherwise, 1 is returned if any bit in the * range is found set. */ @@ -1184,7 +1175,7 @@ int test_range_bit(struct extent_io_tree *tree, u64 start, u64 end, int bitset = 0; unsigned long flags; - read_lock_irqsave(&tree->lock, flags); + spin_lock_irqsave(&tree->lock, flags); node = tree_search(&tree->state, start); while (node && start <= end) { state = rb_entry(node, struct extent_state, rb_node); @@ -1215,7 +1206,7 @@ int test_range_bit(struct extent_io_tree *tree, u64 start, u64 end, break; } } - read_unlock_irqrestore(&tree->lock, flags); + spin_unlock_irqrestore(&tree->lock, flags); return bitset; } EXPORT_SYMBOL(test_range_bit); @@ -1282,16 +1273,19 @@ static int end_bio_extent_writepage(struct bio *bio, { const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags); struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1; - struct extent_io_tree *tree = bio->bi_private; + struct extent_state *state = bio->bi_private; + struct extent_io_tree *tree = state->tree; + struct rb_node *node; u64 start; u64 end; + u64 cur; int whole_page; + 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; start = ((u64)page->index << PAGE_CACHE_SHIFT) + @@ -1311,16 +1305,80 @@ static int end_bio_extent_writepage(struct bio *bio, ClearPageUptodate(page); SetPageError(page); } - clear_extent_writeback(tree, start, end, GFP_ATOMIC); + + if (tree->ops && tree->ops->writepage_end_io_hook) { + tree->ops->writepage_end_io_hook(page, start, end, + state); + } + + /* + * 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 = __tree_search(&tree->state, 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); - if (tree->ops && tree->ops->writepage_end_io_hook) - tree->ops->writepage_end_io_hook(page, start, end); } while (bvec >= bio->bi_io_vec); - bio_put(bio); #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) return 0; @@ -1347,9 +1405,13 @@ static int end_bio_extent_readpage(struct bio *bio, { int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags); struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1; - struct extent_io_tree *tree = bio->bi_private; + struct extent_state *state = bio->bi_private; + struct extent_io_tree *tree = state->tree; + struct rb_node *node; u64 start; u64 end; + u64 cur; + unsigned long flags; int whole_page; int ret; @@ -1373,27 +1435,83 @@ static int end_bio_extent_readpage(struct bio *bio, prefetchw(&bvec->bv_page->flags); if (uptodate && tree->ops && tree->ops->readpage_end_io_hook) { - ret = tree->ops->readpage_end_io_hook(page, start, end); + ret = tree->ops->readpage_end_io_hook(page, start, end, + state); if (ret) uptodate = 0; } - if (uptodate) { - set_extent_uptodate(tree, start, end, GFP_ATOMIC); - if (whole_page) - SetPageUptodate(page); - else - check_page_uptodate(tree, page); - } else { - ClearPageUptodate(page); - SetPageError(page); - } - unlock_extent(tree, start, end, GFP_ATOMIC); + spin_lock_irqsave(&tree->lock, flags); + if (!state || state->end != end) { + state = NULL; + node = __tree_search(&tree->state, 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); + set_extent_uptodate(tree, start, end, + GFP_ATOMIC); + unlock_extent(tree, start, end, GFP_ATOMIC); + goto next_io; + } + } - if (whole_page) + 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 |= 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_WRITEBACK)) { + state = NULL; + } + spin_unlock_irqrestore(&tree->lock, flags); +next_io: + if (whole_page) { + if (uptodate) { + SetPageUptodate(page); + } else { + ClearPageUptodate(page); + SetPageError(page); + } unlock_page(page); - else + } else { + if (uptodate) { + check_page_uptodate(tree, page); + } else { + ClearPageUptodate(page); + SetPageError(page); + } check_page_locked(tree, page); + } } while (bvec >= bio->bi_io_vec); bio_put(bio); @@ -1416,7 +1534,8 @@ static int end_bio_extent_preparewrite(struct bio *bio, { const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags); struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1; - struct extent_io_tree *tree = bio->bi_private; + struct extent_state *state = bio->bi_private; + struct extent_io_tree *tree = state->tree; u64 start; u64 end; @@ -1475,6 +1594,29 @@ static int submit_one_bio(int rw, struct bio *bio) { u64 maxsector; 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 = __tree_search(&tree->state, 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_get(bio); @@ -1519,9 +1661,12 @@ static int submit_extent_page(int rw, struct extent_io_tree *tree, if (!bio) { printk("failed to allocate bio nr %d\n", nr); } + + bio_add_page(bio, page, size, offset); bio->bi_end_io = end_io_func; bio->bi_private = tree; + if (bio_ret) { *bio_ret = bio; } else { @@ -1635,6 +1780,16 @@ static int __extent_read_full_page(struct extent_io_tree *tree, page_offset += iosize; continue; } + /* we have an inline extent but it didn't get marked up + * to date. Error out + */ + if (block_start == EXTENT_MAP_INLINE) { + SetPageError(page); + unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS); + cur = cur + iosize; + page_offset += iosize; + continue; + } ret = 0; if (tree->ops && tree->ops->readpage_io_hook) { @@ -2205,7 +2360,8 @@ EXPORT_SYMBOL(extent_prepare_write); * map records are removed */ int try_release_extent_mapping(struct extent_map_tree *map, - struct extent_io_tree *tree, struct page *page) + struct extent_io_tree *tree, struct page *page, + gfp_t mask) { struct extent_map *em; u64 start = (u64)page->index << PAGE_CACHE_SHIFT; @@ -2213,30 +2369,42 @@ int try_release_extent_mapping(struct extent_map_tree *map, u64 orig_start = start; int ret = 1; - while (start <= end) { - spin_lock(&map->lock); - em = lookup_extent_mapping(map, start, end); - if (!em || IS_ERR(em)) { + if ((mask & __GFP_WAIT) && + page->mapping->host->i_size > 16 * 1024 * 1024) { + while (start <= end) { + spin_lock(&map->lock); + em = lookup_extent_mapping(map, start, end); + if (!em || IS_ERR(em)) { + spin_unlock(&map->lock); + break; + } + if (em->start != start) { + spin_unlock(&map->lock); + free_extent_map(em); + break; + } + if (!test_range_bit(tree, em->start, + extent_map_end(em) - 1, + EXTENT_LOCKED, 0)) { + remove_extent_mapping(map, em); + /* once for the rb tree */ + free_extent_map(em); + } + start = extent_map_end(em); spin_unlock(&map->lock); - break; - } - if (!test_range_bit(tree, em->start, extent_map_end(em) - 1, - EXTENT_LOCKED, 0)) { - remove_extent_mapping(map, em); - /* once for the rb tree */ + + /* once for us */ free_extent_map(em); } - start = extent_map_end(em); - spin_unlock(&map->lock); - - /* once for us */ - free_extent_map(em); } - if (test_range_bit(tree, orig_start, end, EXTENT_LOCKED, 0)) + if (test_range_bit(tree, orig_start, end, EXTENT_IOBITS, 0)) ret = 0; - else + else { + if ((mask & GFP_NOFS) == GFP_NOFS) + mask = GFP_NOFS; clear_extent_bit(tree, orig_start, end, EXTENT_UPTODATE, - 1, 1, GFP_NOFS); + 1, 1, mask); + } return ret; } EXPORT_SYMBOL(try_release_extent_mapping); @@ -2553,13 +2721,13 @@ int clear_extent_buffer_dirty(struct extent_io_tree *tree, } } clear_page_dirty_for_io(page); - write_lock_irq(&page->mapping->tree_lock); + read_lock_irq(&page->mapping->tree_lock); if (!PageDirty(page)) { radix_tree_tag_clear(&page->mapping->page_tree, page_index(page), PAGECACHE_TAG_DIRTY); } - write_unlock_irq(&page->mapping->tree_lock); + read_unlock_irq(&page->mapping->tree_lock); unlock_page(page); } return 0; -- cgit v1.2.2 From 85e21bac165b4ba1f6f90431ad6fc658ffcbaf3a Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 29 Jan 2008 15:11:36 -0500 Subject: Btrfs: During deletes and truncate, remove many items at once from the tree Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 1 - 1 file changed, 1 deletion(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 1f734c34dc24..8aec72253a17 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -2863,7 +2863,6 @@ int read_extent_buffer_pages(struct extent_io_tree *tree, if (ret || !wait) { return ret; } - for (i = start_i; i < num_pages; i++) { page = extent_buffer_page(eb, i); wait_on_page_locked(page); -- cgit v1.2.2 From 291d673e6a22d9c6834e939f66c7cfef90669021 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 29 Jan 2008 15:55:23 -0500 Subject: Btrfs: Do delalloc accounting via hooks in the extent_state code Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 25 +++++++++++++++++++++++++ 1 file changed, 25 insertions(+) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 8aec72253a17..e81b35b06187 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -256,6 +256,26 @@ static int merge_state(struct extent_io_tree *tree, return 0; } +static void 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, bits); + } +} + +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, bits); + } +} + /* * insert an extent_state struct into the tree. 'bits' are set on the * struct before it is inserted. @@ -281,6 +301,7 @@ static int insert_state(struct extent_io_tree *tree, state->state |= bits; state->start = start; state->end = end; + set_state_cb(tree, state, bits); node = tree_insert(&tree->state, end, &state->rb_node); if (node) { struct extent_state *found; @@ -349,6 +370,7 @@ static int clear_state_bit(struct extent_io_tree *tree, tree->dirty_bytes -= range; } state->state &= ~bits; + clear_state_cb(tree, state, bits); if (wake) wake_up(&state->wq); if (delete || state->state == 0) { @@ -553,6 +575,7 @@ static void set_state_bits(struct extent_io_tree *tree, tree->dirty_bytes += range; } state->state |= bits; + set_state_cb(tree, state, bits); } /* @@ -975,6 +998,7 @@ search_again: goto search_again; } state->state |= EXTENT_LOCKED; + set_state_cb(tree, state, EXTENT_LOCKED); if (!found) *start = state->start; found++; @@ -1474,6 +1498,7 @@ static int end_bio_extent_readpage(struct bio *bio, state = NULL; } clear->state |= EXTENT_UPTODATE; + set_state_cb(tree, clear, EXTENT_UPTODATE); clear_state_bit(tree, clear, EXTENT_LOCKED, 1, 0); if (cur == start) -- cgit v1.2.2 From 6f568d35a045dbb8a13fe71bfc32e85e39a986cb Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 29 Jan 2008 16:03:38 -0500 Subject: Btrfs: mount -o max_inline=size to control the maximum inline extent size Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 1 - 1 file changed, 1 deletion(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index e81b35b06187..6999b819db62 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -2451,7 +2451,6 @@ sector_t extent_bmap(struct address_space *mapping, sector_t iblock, goto out; sector = (em->block_start + start - em->start) >> inode->i_blkbits; -printk("bmap finds %Lu %Lu block %Lu\n", em->start, em->len, em->block_start); out: free_extent_map(em); return sector; -- cgit v1.2.2 From b0c68f8bed058d9f2023b067b16ed06a8c439544 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 31 Jan 2008 11:05:37 -0500 Subject: Btrfs: Enable delalloc accounting Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 6999b819db62..624aabc3e6f2 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -262,7 +262,7 @@ static void set_state_cb(struct extent_io_tree *tree, { if (tree->ops && tree->ops->set_bit_hook) { tree->ops->set_bit_hook(tree->mapping->host, state->start, - state->end, bits); + state->end, state->state, bits); } } @@ -272,7 +272,7 @@ static void clear_state_cb(struct extent_io_tree *tree, { if (tree->ops && tree->ops->set_bit_hook) { tree->ops->clear_bit_hook(tree->mapping->host, state->start, - state->end, bits); + state->end, state->state, bits); } } @@ -298,10 +298,10 @@ static int insert_state(struct extent_io_tree *tree, } 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; - set_state_cb(tree, state, bits); node = tree_insert(&tree->state, end, &state->rb_node); if (node) { struct extent_state *found; @@ -369,8 +369,8 @@ static int clear_state_bit(struct extent_io_tree *tree, WARN_ON(range > tree->dirty_bytes); tree->dirty_bytes -= range; } - state->state &= ~bits; clear_state_cb(tree, state, bits); + state->state &= ~bits; if (wake) wake_up(&state->wq); if (delete || state->state == 0) { @@ -574,8 +574,8 @@ static void set_state_bits(struct extent_io_tree *tree, u64 range = state->end - state->start + 1; tree->dirty_bytes += range; } - state->state |= bits; set_state_cb(tree, state, bits); + state->state |= bits; } /* @@ -997,8 +997,8 @@ search_again: free_extent_state(state); goto search_again; } - state->state |= EXTENT_LOCKED; set_state_cb(tree, state, EXTENT_LOCKED); + state->state |= EXTENT_LOCKED; if (!found) *start = state->start; found++; @@ -1497,8 +1497,8 @@ static int end_bio_extent_readpage(struct bio *bio, } else { state = NULL; } - clear->state |= EXTENT_UPTODATE; set_state_cb(tree, clear, EXTENT_UPTODATE); + clear->state |= EXTENT_UPTODATE; clear_state_bit(tree, clear, EXTENT_LOCKED, 1, 0); if (cur == start) -- cgit v1.2.2 From 80ea96b1f3bd2431e0d71c9df6ab45c3de0c5840 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Fri, 1 Feb 2008 14:51:59 -0500 Subject: Btrfs: Add a lookup cache to the extent state tree Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 57 ++++++++++++++++++++++++++++++++++++---------------- 1 file changed, 40 insertions(+), 17 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 624aabc3e6f2..5f2fbf2054f2 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -90,6 +90,7 @@ void extent_io_tree_init(struct extent_io_tree *tree, tree->mapping = mapping; INIT_LIST_HEAD(&tree->buffer_lru); tree->lru_size = 0; + tree->last = NULL; } EXPORT_SYMBOL(extent_io_tree_init); @@ -158,16 +159,23 @@ static struct rb_node *tree_insert(struct rb_root *root, u64 offset, return NULL; } -static struct rb_node *__tree_search(struct rb_root *root, u64 offset, +static struct rb_node *__etree_search(struct extent_io_tree *tree, u64 offset, struct rb_node **prev_ret, struct rb_node **next_ret) { + struct rb_root *root = &tree->state; 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) { entry = rb_entry(n, struct tree_entry, rb_node); prev = n; @@ -177,8 +185,10 @@ static struct rb_node *__tree_search(struct rb_root *root, u64 offset, n = n->rb_left; else if (offset > entry->end) n = n->rb_right; - else + else { + tree->last = rb_entry(n, struct extent_state, rb_node); return n; + } } if (prev_ret) { @@ -202,14 +212,20 @@ static struct rb_node *__tree_search(struct rb_root *root, u64 offset, return NULL; } -static inline struct rb_node *tree_search(struct rb_root *root, u64 offset) +static inline struct rb_node *tree_search(struct extent_io_tree *tree, + u64 offset) { struct rb_node *prev = NULL; struct rb_node *ret; - ret = __tree_search(root, offset, &prev, NULL); - if (!ret) + ret = __etree_search(tree, offset, &prev, NULL); + if (!ret) { + if (prev) { + tree->last = rb_entry(prev, struct extent_state, + rb_node); + } return prev; + } return ret; } @@ -238,6 +254,8 @@ static int merge_state(struct extent_io_tree *tree, other->state == state->state) { state->start = other->start; other->tree = NULL; + if (tree->last == other) + tree->last = NULL; rb_erase(&other->rb_node, &tree->state); free_extent_state(other); } @@ -249,6 +267,8 @@ static int merge_state(struct extent_io_tree *tree, other->state == state->state) { other->start = state->start; state->tree = NULL; + if (tree->last == state) + tree->last = NULL; rb_erase(&state->rb_node, &tree->state); free_extent_state(state); } @@ -311,6 +331,7 @@ static int insert_state(struct extent_io_tree *tree, return -EEXIST; } state->tree = tree; + tree->last = state; merge_state(tree, state); return 0; } @@ -375,6 +396,8 @@ static int clear_state_bit(struct extent_io_tree *tree, wake_up(&state->wq); if (delete || state->state == 0) { if (state->tree) { + if (tree->last == state) + tree->last = NULL; rb_erase(&state->rb_node, &tree->state); state->tree = NULL; free_extent_state(state); @@ -422,7 +445,7 @@ again: * this search will find the extents that end after * our range starts */ - node = tree_search(&tree->state, start); + node = tree_search(tree, start); if (!node) goto out; state = rb_entry(node, struct extent_state, rb_node); @@ -533,7 +556,7 @@ again: * this search will find all the extents that end after * our range starts */ - node = tree_search(&tree->state, start); + node = tree_search(tree, start); if (!node) break; @@ -612,7 +635,7 @@ again: * this search will find all the extents that end after * our range starts. */ - node = tree_search(&tree->state, start); + node = tree_search(tree, start); if (!node) { err = insert_state(tree, prealloc, start, end, bits); prealloc = NULL; @@ -915,7 +938,7 @@ int find_first_extent_bit(struct extent_io_tree *tree, u64 start, * this search will find all the extents that end after * our range starts. */ - node = tree_search(&tree->state, start); + node = tree_search(tree, start); if (!node || IS_ERR(node)) { goto out; } @@ -953,7 +976,7 @@ u64 find_lock_delalloc_range(struct extent_io_tree *tree, * our range starts. */ search_again: - node = tree_search(&tree->state, cur_start); + node = tree_search(tree, cur_start); if (!node || IS_ERR(node)) { *end = (u64)-1; goto out; @@ -1041,7 +1064,7 @@ u64 count_range_bits(struct extent_io_tree *tree, * this search will find all the extents that end after * our range starts. */ - node = tree_search(&tree->state, cur_start); + node = tree_search(tree, cur_start); if (!node || IS_ERR(node)) { goto out; } @@ -1142,7 +1165,7 @@ int set_state_private(struct extent_io_tree *tree, u64 start, u64 private) * this search will find all the extents that end after * our range starts. */ - node = tree_search(&tree->state, start); + node = tree_search(tree, start); if (!node || IS_ERR(node)) { ret = -ENOENT; goto out; @@ -1169,7 +1192,7 @@ int get_state_private(struct extent_io_tree *tree, u64 start, u64 *private) * this search will find all the extents that end after * our range starts. */ - node = tree_search(&tree->state, start); + node = tree_search(tree, start); if (!node || IS_ERR(node)) { ret = -ENOENT; goto out; @@ -1200,7 +1223,7 @@ int test_range_bit(struct extent_io_tree *tree, u64 start, u64 end, unsigned long flags; spin_lock_irqsave(&tree->lock, flags); - node = tree_search(&tree->state, start); + node = tree_search(tree, start); while (node && start <= end) { state = rb_entry(node, struct extent_state, rb_node); @@ -1348,7 +1371,7 @@ static int end_bio_extent_writepage(struct bio *bio, spin_lock_irqsave(&tree->lock, flags); if (!state || state->end != end) { state = NULL; - node = __tree_search(&tree->state, start, NULL, NULL); + node = __etree_search(tree, start, NULL, NULL); if (node) { state = rb_entry(node, struct extent_state, rb_node); @@ -1468,7 +1491,7 @@ static int end_bio_extent_readpage(struct bio *bio, spin_lock_irqsave(&tree->lock, flags); if (!state || state->end != end) { state = NULL; - node = __tree_search(&tree->state, start, NULL, NULL); + node = __etree_search(tree, start, NULL, NULL); if (node) { state = rb_entry(node, struct extent_state, rb_node); @@ -1631,7 +1654,7 @@ static int submit_one_bio(int rw, struct bio *bio) end = start + bvec->bv_len - 1; spin_lock_irq(&tree->lock); - node = __tree_search(&tree->state, start, NULL, NULL); + node = __etree_search(tree, start, NULL, NULL); BUG_ON(!node); state = rb_entry(node, struct extent_state, rb_node); while(state->end < end) { -- cgit v1.2.2 From ae9d12853b44f4e0c06732166e8e3399ec01a680 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Fri, 1 Feb 2008 15:42:15 -0500 Subject: Btrfs: Fix delalloc account on state deletion Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 1 + 1 file changed, 1 insertion(+) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 5f2fbf2054f2..5d2fb527fd96 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -396,6 +396,7 @@ static int clear_state_bit(struct extent_io_tree *tree, 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 = NULL; rb_erase(&state->rb_node, &tree->state); -- cgit v1.2.2 From c2e639f02cff721172569add21ba5d058fd1a81c Mon Sep 17 00:00:00 2001 From: Yan Date: Mon, 4 Feb 2008 08:57:25 -0500 Subject: Btrfs: Fix typo in extent_io.c --- Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 5d2fb527fd96..36a943e51808 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -1416,7 +1416,7 @@ static int end_bio_extent_writepage(struct bio *bio, * to the correct offsets in the file */ if (state && (state->end + 1 != start || - !state->state & EXTENT_WRITEBACK)) { + !(state->state & EXTENT_WRITEBACK))) { state = NULL; } spin_unlock_irqrestore(&tree->lock, flags); @@ -1539,7 +1539,7 @@ static int end_bio_extent_readpage(struct bio *bio, * to the correct offsets in the file */ if (state && (state->end + 1 != start || - !state->state & EXTENT_WRITEBACK)) { + !(state->state & EXTENT_LOCKED))) { state = NULL; } spin_unlock_irqrestore(&tree->lock, flags); -- cgit v1.2.2 From 961d023213b94861480f5d6533f7adc8332eb527 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 6 Feb 2008 11:01:42 -0500 Subject: Btrfs: Don't case unsigned long to int in bio submission Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 36a943e51808..b68e57f66a96 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -1705,7 +1705,7 @@ static int submit_extent_page(int rw, struct extent_io_tree *tree, return 0; } } - nr = min_t(int, max_pages, bio_get_nr_vecs(bdev)); + 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); -- cgit v1.2.2 From a86c12c73d982c545a37a8ecdd66528ab260b770 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 7 Feb 2008 10:50:54 -0500 Subject: Btrfs: Create larger bios for btree blocks Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 12 +++++++++--- 1 file changed, 9 insertions(+), 3 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index b68e57f66a96..fcae999ab712 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -2859,8 +2859,8 @@ EXPORT_SYMBOL(extent_buffer_uptodate); int read_extent_buffer_pages(struct extent_io_tree *tree, struct extent_buffer *eb, - u64 start, - int wait) + u64 start, int wait, + get_extent_t *get_extent) { unsigned long i; unsigned long start_i; @@ -2868,6 +2868,8 @@ int read_extent_buffer_pages(struct extent_io_tree *tree, int err; int ret = 0; unsigned long num_pages; + struct bio *bio = NULL; + if (eb->flags & EXTENT_UPTODATE) return 0; @@ -2899,7 +2901,8 @@ int read_extent_buffer_pages(struct extent_io_tree *tree, lock_page(page); } if (!PageUptodate(page)) { - err = page->mapping->a_ops->readpage(NULL, page); + err = __extent_read_full_page(tree, page, + get_extent, &bio); if (err) { ret = err; } @@ -2908,6 +2911,9 @@ int read_extent_buffer_pages(struct extent_io_tree *tree, } } + if (bio) + submit_one_bio(READ, bio); + if (ret || !wait) { return ret; } -- cgit v1.2.2 From 39b5637f6f195852259004bb27b58e2dcf9fb378 Mon Sep 17 00:00:00 2001 From: Yan Date: Fri, 15 Feb 2008 10:40:50 -0500 Subject: Btrfs: Fix "no csum found for inode" issue. A few codes were not properly updated for changes of extent map. This may be the causes of "no csum found for inode" issue. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index fcae999ab712..9262ab37a7cd 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -2417,12 +2417,13 @@ int try_release_extent_mapping(struct extent_map_tree *map, u64 end = start + PAGE_CACHE_SIZE - 1; u64 orig_start = start; int ret = 1; - if ((mask & __GFP_WAIT) && page->mapping->host->i_size > 16 * 1024 * 1024) { + u64 len; while (start <= end) { + len = end - start + 1; spin_lock(&map->lock); - em = lookup_extent_mapping(map, start, end); + em = lookup_extent_mapping(map, start, len); if (!em || IS_ERR(em)) { spin_unlock(&map->lock); break; -- cgit v1.2.2 From d7fc640e6fed46932f7c74e14f9b58b8637c66cf Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 18 Feb 2008 12:12:38 -0500 Subject: Btrfs: Allocator improvements Reduce CPU time searching for free blocks by optimizing find_first_extent_bit Fix find_free_extent to make better use of the last_alloc hint. Before it was often finding blocks just before the hint. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 38 ++++++++++++++++++++++++++++++++++---- 1 file changed, 34 insertions(+), 4 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 9262ab37a7cd..fb6400895ed6 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -255,7 +255,7 @@ static int merge_state(struct extent_io_tree *tree, state->start = other->start; other->tree = NULL; if (tree->last == other) - tree->last = NULL; + tree->last = state; rb_erase(&other->rb_node, &tree->state); free_extent_state(other); } @@ -268,7 +268,7 @@ static int merge_state(struct extent_io_tree *tree, other->start = state->start; state->tree = NULL; if (tree->last == state) - tree->last = NULL; + tree->last = other; rb_erase(&state->rb_node, &tree->state); free_extent_state(state); } @@ -397,8 +397,9 @@ static int clear_state_bit(struct extent_io_tree *tree, if (delete || state->state == 0) { if (state->tree) { clear_state_cb(tree, state, state->state); - if (tree->last == state) - tree->last = NULL; + if (tree->last == state) { + tree->last = extent_state_next(state); + } rb_erase(&state->rb_node, &tree->state); state->tree = NULL; free_extent_state(state); @@ -962,6 +963,35 @@ out: } EXPORT_SYMBOL(find_first_extent_bit); +struct extent_state *find_first_extent_bit_state(struct extent_io_tree *tree, + u64 start, int bits) +{ + struct rb_node *node; + struct extent_state *state; + + /* + * this search will find all the extents that end after + * our range starts. + */ + node = tree_search(tree, start); + if (!node || IS_ERR(node)) { + goto out; + } + + while(1) { + state = rb_entry(node, struct extent_state, rb_node); + if (state->end >= start && (state->state & bits)) { + return state; + } + node = rb_next(node); + if (!node) + break; + } +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) { -- cgit v1.2.2 From 065631f6dccea07bfad48d8981369f6d9cfd6e2b Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 20 Feb 2008 12:07:25 -0500 Subject: Btrfs: checksum file data at bio submission time instead of during writepage When we checkum file data during writepage, the checksumming is done one page at a time, making it difficult to do bulk metadata modifications to insert checksums for large ranges of the file at once. This patch changes btrfs to checksum on a per-bio basis instead. The bios are checksummed before they are handed off to the block layer, so each bio is contiguous and only has pages from the same inode. Checksumming on a bio basis allows us to insert and modify the file checksum items in large groups. It also allows the checksumming to be done more easily by async worker threads. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index fb6400895ed6..e8130c876330 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -1705,6 +1705,8 @@ static int submit_one_bio(int rw, struct bio *bio) (unsigned long long)bio->bi_sector); WARN_ON(1); } + if (tree->ops && tree->ops->submit_bio_hook) + tree->ops->submit_bio_hook(rw, bio); submit_bio(rw, bio); if (bio_flagged(bio, BIO_EOPNOTSUPP)) -- cgit v1.2.2 From 0b86a832a1f38abec695864ec2eaedc9d2383f1b Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 24 Mar 2008 15:01:56 -0400 Subject: Btrfs: Add support for multiple devices per filesystem Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index e8130c876330..7e3a1ebde9fc 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -1706,9 +1706,9 @@ static int submit_one_bio(int rw, struct bio *bio) WARN_ON(1); } if (tree->ops && tree->ops->submit_bio_hook) - tree->ops->submit_bio_hook(rw, bio); - - submit_bio(rw, bio); + tree->ops->submit_bio_hook(page->mapping->host, rw, bio); + else + submit_bio(rw, bio); if (bio_flagged(bio, BIO_EOPNOTSUPP)) ret = -EOPNOTSUPP; bio_put(bio); -- cgit v1.2.2 From 239b14b32dc39232ebf9cce29ff77c4c564355fd Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 24 Mar 2008 15:02:07 -0400 Subject: Btrfs: Bring back mount -o ssd optimizations Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 7e3a1ebde9fc..6dab664529c1 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -1730,6 +1730,8 @@ static int submit_extent_page(int rw, struct extent_io_tree *tree, if (bio_ret && *bio_ret) { bio = *bio_ret; if (bio->bi_sector + (bio->bi_size >> 9) != sector || + (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); bio = NULL; -- cgit v1.2.2 From 2d2ae547979854d10b75d557b3abdb3eb7511bbc Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 26 Mar 2008 16:24:23 -0400 Subject: Btrfs: Add leak debugging for extent_buffer and extent_state This also fixes one leak around the super block when failing to mount the FS. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 28 ++++++++++++++++++++++++++-- 1 file changed, 26 insertions(+), 2 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 6dab664529c1..b99cfabd008e 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -26,6 +26,7 @@ static struct kmem_cache *extent_buffer_cache; static LIST_HEAD(buffers); static LIST_HEAD(states); +static spinlock_t leak_lock = SPIN_LOCK_UNLOCKED; #define BUFFER_LRU_MAX 64 @@ -64,15 +65,22 @@ free_state_cache: void extent_io_exit(void) { struct extent_state *state; + struct extent_buffer *eb; while (!list_empty(&states)) { - state = list_entry(states.next, struct extent_state, list); + 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)); - list_del(&state->list); + list_del(&state->leak_list); kmem_cache_free(extent_state_cache, state); } + 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)); + list_del(&eb->leak_list); + kmem_cache_free(extent_buffer_cache, eb); + } if (extent_state_cache) kmem_cache_destroy(extent_state_cache); if (extent_buffer_cache) @@ -109,6 +117,7 @@ EXPORT_SYMBOL(extent_io_tree_empty_lru); struct extent_state *alloc_extent_state(gfp_t mask) { struct extent_state *state; + unsigned long flags; state = kmem_cache_alloc(extent_state_cache, mask); if (!state || IS_ERR(state)) @@ -116,6 +125,9 @@ struct extent_state *alloc_extent_state(gfp_t mask) state->state = 0; state->private = 0; state->tree = NULL; + spin_lock_irqsave(&leak_lock, flags); + list_add(&state->leak_list, &states); + spin_unlock_irqrestore(&leak_lock, flags); atomic_set(&state->refs, 1); init_waitqueue_head(&state->wq); @@ -128,7 +140,11 @@ void free_extent_state(struct extent_state *state) if (!state) return; if (atomic_dec_and_test(&state->refs)) { + unsigned long flags; WARN_ON(state->tree); + spin_lock_irqsave(&leak_lock, flags); + list_del(&state->leak_list); + spin_unlock_irqrestore(&leak_lock, flags); kmem_cache_free(extent_state_cache, state); } } @@ -2582,6 +2598,7 @@ static struct extent_buffer *__alloc_extent_buffer(struct extent_io_tree *tree, gfp_t mask) { struct extent_buffer *eb = NULL; + unsigned long flags; spin_lock(&tree->lru_lock); eb = find_lru(tree, start, len); @@ -2594,6 +2611,9 @@ static struct extent_buffer *__alloc_extent_buffer(struct extent_io_tree *tree, INIT_LIST_HEAD(&eb->lru); eb->start = start; eb->len = len; + spin_lock_irqsave(&leak_lock, flags); + list_add(&eb->leak_list, &buffers); + spin_unlock_irqrestore(&leak_lock, flags); atomic_set(&eb->refs, 1); return eb; @@ -2601,6 +2621,10 @@ static struct extent_buffer *__alloc_extent_buffer(struct extent_io_tree *tree, static void __free_extent_buffer(struct extent_buffer *eb) { + unsigned long flags; + spin_lock_irqsave(&leak_lock, flags); + list_del(&eb->leak_list); + spin_unlock_irqrestore(&leak_lock, flags); kmem_cache_free(extent_buffer_cache, eb); } -- cgit v1.2.2 From 2b114d1d33551ab1dcaa1d0b3a956df3776fb767 Mon Sep 17 00:00:00 2001 From: Peter Date: Tue, 1 Apr 2008 11:21:40 -0400 Subject: Btrfs: Correct usage of IS_ERR() in extent_io.c Signed-off-by: Peter Teoh Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 18 +++++++++--------- 1 file changed, 9 insertions(+), 9 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index b99cfabd008e..776c33f1e6fe 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -120,7 +120,7 @@ struct extent_state *alloc_extent_state(gfp_t mask) unsigned long flags; state = kmem_cache_alloc(extent_state_cache, mask); - if (!state || IS_ERR(state)) + if (!state) return state; state->state = 0; state->private = 0; @@ -957,7 +957,7 @@ int find_first_extent_bit(struct extent_io_tree *tree, u64 start, * our range starts. */ node = tree_search(tree, start); - if (!node || IS_ERR(node)) { + if (!node) { goto out; } @@ -990,7 +990,7 @@ struct extent_state *find_first_extent_bit_state(struct extent_io_tree *tree, * our range starts. */ node = tree_search(tree, start); - if (!node || IS_ERR(node)) { + if (!node) { goto out; } @@ -1024,7 +1024,7 @@ u64 find_lock_delalloc_range(struct extent_io_tree *tree, */ search_again: node = tree_search(tree, cur_start); - if (!node || IS_ERR(node)) { + if (!node) { *end = (u64)-1; goto out; } @@ -1112,7 +1112,7 @@ u64 count_range_bits(struct extent_io_tree *tree, * our range starts. */ node = tree_search(tree, cur_start); - if (!node || IS_ERR(node)) { + if (!node) { goto out; } @@ -1213,7 +1213,7 @@ int set_state_private(struct extent_io_tree *tree, u64 start, u64 private) * our range starts. */ node = tree_search(tree, start); - if (!node || IS_ERR(node)) { + if (!node) { ret = -ENOENT; goto out; } @@ -1240,7 +1240,7 @@ int get_state_private(struct extent_io_tree *tree, u64 start, u64 *private) * our range starts. */ node = tree_search(tree, start); - if (!node || IS_ERR(node)) { + if (!node) { ret = -ENOENT; goto out; } @@ -2642,7 +2642,7 @@ struct extent_buffer *alloc_extent_buffer(struct extent_io_tree *tree, int uptodate = 1; eb = __alloc_extent_buffer(tree, start, len, mask); - if (!eb || IS_ERR(eb)) + if (!eb) return NULL; if (eb->flags & EXTENT_BUFFER_FILLED) @@ -2717,7 +2717,7 @@ struct extent_buffer *find_extent_buffer(struct extent_io_tree *tree, int uptodate = 1; eb = __alloc_extent_buffer(tree, start, len, mask); - if (!eb || IS_ERR(eb)) + if (!eb) return NULL; if (eb->flags & EXTENT_BUFFER_FILLED) -- cgit v1.2.2 From 728131d8e40c2a47c59ca91a806299c4708029f9 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 9 Apr 2008 16:28:12 -0400 Subject: Btrfs: Add additional debugging for metadata checksum failures Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 54 +++++++++++++++++++++++++++++++++++++++++++++++++--- 1 file changed, 51 insertions(+), 3 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 776c33f1e6fe..13cc2360e375 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -2592,6 +2592,33 @@ static inline struct page *extent_buffer_page(struct extent_buffer *eb, return p; } +int invalidate_extent_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; + 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; + } + if (eb->start == start) { + eb->flags &= ~EXTENT_CSUM; + } + cur = cur->next; + } while (cur != lru); +out: + spin_unlock(&tree->lru_lock); + return found; +} + static struct extent_buffer *__alloc_extent_buffer(struct extent_io_tree *tree, u64 start, unsigned long len, @@ -2909,10 +2936,32 @@ EXPORT_SYMBOL(set_extent_buffer_uptodate); int extent_buffer_uptodate(struct extent_io_tree *tree, struct extent_buffer *eb) { + int ret = 0; + int ret2; + int num_pages; + int i; + struct page *page; + int pg_uptodate = 1; + if (eb->flags & EXTENT_UPTODATE) - return 1; - return test_range_bit(tree, eb->start, eb->start + eb->len - 1, + ret = 1; + + ret2 = test_range_bit(tree, eb->start, eb->start + eb->len - 1, EXTENT_UPTODATE, 1); + + num_pages = num_extent_pages(eb->start, eb->len); + for (i = 0; i < num_pages; i++) { + page = extent_buffer_page(eb, i); + if (!PageUptodate(page)) { + pg_uptodate = 0; + break; + } + } + if ((ret || ret2) && !pg_uptodate) { +printk("uptodate error2 eb %Lu ret %d ret2 %d pg_uptodate %d\n", eb->start, ret, ret2, pg_uptodate); + WARN_ON(1); + } + return (ret || ret2); } EXPORT_SYMBOL(extent_buffer_uptodate); @@ -2929,7 +2978,6 @@ int read_extent_buffer_pages(struct extent_io_tree *tree, unsigned long num_pages; struct bio *bio = NULL; - if (eb->flags & EXTENT_UPTODATE) return 0; -- cgit v1.2.2 From ce9adaa5a792c2099a83246265eb4055bc38b6b8 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 9 Apr 2008 16:28:12 -0400 Subject: Btrfs: Do metadata checksums for reads via a workqueue Before, metadata checksumming was done by the callers of read_tree_block, which would set EXTENT_CSUM bits in the extent tree to show that a given range of pages was already checksummed and didn't need to be verified again. But, those bits could go away via try_to_releasepage, and the end result was bogus checksum failures on pages that never left the cache. The new code validates checksums when the page is read. It is a little tricky because metadata blocks can span pages and a single read may end up going via multiple bios. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 100 +++++++++++++++++++++++++++++++++++++++------------ 1 file changed, 78 insertions(+), 22 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 13cc2360e375..cfc383c17a3a 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -2592,6 +2592,22 @@ static inline struct page *extent_buffer_page(struct extent_buffer *eb, return p; } +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; + + 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; +} + + int invalidate_extent_lru(struct extent_io_tree *tree, u64 start, unsigned long len) { @@ -2609,9 +2625,6 @@ int invalidate_extent_lru(struct extent_io_tree *tree, u64 start, if (eb->start <= start && eb->start + eb->len > start) { eb->flags &= ~EXTENT_UPTODATE; } - if (eb->start == start) { - eb->flags &= ~EXTENT_CSUM; - } cur = cur->next; } while (cur != lru); out: @@ -2682,7 +2695,6 @@ struct extent_buffer *alloc_extent_buffer(struct extent_io_tree *tree, page_cache_get(page0); mark_page_accessed(page0); set_page_extent_mapped(page0); - WARN_ON(!PageUptodate(page0)); set_page_extent_head(page0, len); } else { i = 0; @@ -2933,13 +2945,39 @@ int set_extent_buffer_uptodate(struct extent_io_tree *tree, } EXPORT_SYMBOL(set_extent_buffer_uptodate); +int extent_range_uptodate(struct extent_io_tree *tree, + u64 start, u64 end) +{ + struct page *page; + int ret; + int pg_uptodate = 1; + int uptodate; + unsigned long index; + + ret = test_range_bit(tree, start, end, EXTENT_UPTODATE, 1); + if (ret) + return 1; + while(start <= end) { + index = start >> PAGE_CACHE_SHIFT; + page = find_get_page(tree->mapping, index); + uptodate = PageUptodate(page); + page_cache_release(page); + if (!uptodate) { + pg_uptodate = 0; + break; + } + start += PAGE_CACHE_SIZE; + } + return pg_uptodate; +} + int extent_buffer_uptodate(struct extent_io_tree *tree, - struct extent_buffer *eb) + struct extent_buffer *eb) { int ret = 0; int ret2; - int num_pages; - int i; + unsigned long num_pages; + unsigned long i; struct page *page; int pg_uptodate = 1; @@ -2975,13 +3013,16 @@ int read_extent_buffer_pages(struct extent_io_tree *tree, struct page *page; int err; int ret = 0; + int locked_pages = 0; + int all_uptodate = 1; + int inc_all_pages = 0; unsigned long num_pages; struct bio *bio = NULL; if (eb->flags & EXTENT_UPTODATE) return 0; - if (0 && test_range_bit(tree, eb->start, eb->start + eb->len - 1, + if (test_range_bit(tree, eb->start, eb->start + eb->len - 1, EXTENT_UPTODATE, 1)) { return 0; } @@ -2997,17 +3038,30 @@ int read_extent_buffer_pages(struct extent_io_tree *tree, num_pages = num_extent_pages(eb->start, eb->len); for (i = start_i; i < num_pages; i++) { page = extent_buffer_page(eb, i); - if (PageUptodate(page)) { - continue; - } if (!wait) { - if (TestSetPageLocked(page)) { - continue; - } + if (TestSetPageLocked(page)) + goto unlock_exit; } else { lock_page(page); } + locked_pages++; + if (!PageUptodate(page)) { + all_uptodate = 0; + } + } + if (all_uptodate) { + if (start_i == 0) + eb->flags |= EXTENT_UPTODATE; + goto unlock_exit; + } + + for (i = start_i; i < num_pages; i++) { + page = extent_buffer_page(eb, i); + if (inc_all_pages) + page_cache_get(page); if (!PageUptodate(page)) { + if (start_i == 0) + inc_all_pages = 1; err = __extent_read_full_page(tree, page, get_extent, &bio); if (err) { @@ -3034,6 +3088,16 @@ int read_extent_buffer_pages(struct extent_io_tree *tree, if (!ret) eb->flags |= EXTENT_UPTODATE; return ret; + +unlock_exit: + i = start_i; + while(locked_pages > 0) { + page = extent_buffer_page(eb, i); + i++; + unlock_page(page); + locked_pages--; + } + return ret; } EXPORT_SYMBOL(read_extent_buffer_pages); @@ -3048,7 +3112,6 @@ void read_extent_buffer(struct extent_buffer *eb, void *dstv, char *dst = (char *)dstv; size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1); unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT; - unsigned long num_pages = num_extent_pages(eb->start, eb->len); WARN_ON(start > eb->len); WARN_ON(start + len > eb->start + eb->len); @@ -3057,11 +3120,6 @@ void read_extent_buffer(struct extent_buffer *eb, void *dstv, while(len > 0) { page = extent_buffer_page(eb, i); - if (!PageUptodate(page)) { - printk("page %lu not up to date i %lu, total %lu, len %lu\n", page->index, i, num_pages, eb->len); - WARN_ON(1); - } - WARN_ON(!PageUptodate(page)); cur = min(len, (PAGE_CACHE_SIZE - offset)); kaddr = kmap_atomic(page, KM_USER1); @@ -3105,7 +3163,6 @@ printk("bad mapping eb start %Lu len %lu, wanted %lu %lu\n", eb->start, eb->len, } p = extent_buffer_page(eb, i); - WARN_ON(!PageUptodate(p)); kaddr = kmap_atomic(p, km); *token = kaddr; *map = kaddr + offset; @@ -3165,7 +3222,6 @@ int memcmp_extent_buffer(struct extent_buffer *eb, const void *ptrv, while(len > 0) { page = extent_buffer_page(eb, i); - WARN_ON(!PageUptodate(page)); cur = min(len, (PAGE_CACHE_SIZE - offset)); -- cgit v1.2.2 From f188591e987e21b6f7f8864c66a02858b95b530e Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 9 Apr 2008 16:28:12 -0400 Subject: Btrfs: Retry metadata reads in the face of checksum failures Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 50 +++++++++++++++++++++++++++++--------------------- 1 file changed, 29 insertions(+), 21 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index cfc383c17a3a..2f159375c878 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -1546,7 +1546,7 @@ static int end_bio_extent_readpage(struct bio *bio, !(state->state & EXTENT_LOCKED)) state = NULL; } - if (!state) { + if (!state && uptodate) { spin_unlock_irqrestore(&tree->lock, flags); set_extent_uptodate(tree, start, end, GFP_ATOMIC); @@ -1567,8 +1567,10 @@ static int end_bio_extent_readpage(struct bio *bio, } else { state = NULL; } - set_state_cb(tree, clear, EXTENT_UPTODATE); - clear->state |= EXTENT_UPTODATE; + 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) @@ -1685,7 +1687,7 @@ 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) +static int submit_one_bio(int rw, struct bio *bio, int mirror_num) { u64 maxsector; int ret = 0; @@ -1722,7 +1724,8 @@ static int submit_one_bio(int rw, struct bio *bio) WARN_ON(1); } if (tree->ops && tree->ops->submit_bio_hook) - tree->ops->submit_bio_hook(page->mapping->host, rw, bio); + tree->ops->submit_bio_hook(page->mapping->host, rw, bio, + mirror_num); else submit_bio(rw, bio); if (bio_flagged(bio, BIO_EOPNOTSUPP)) @@ -1737,7 +1740,8 @@ static int submit_extent_page(int rw, struct extent_io_tree *tree, struct block_device *bdev, struct bio **bio_ret, unsigned long max_pages, - bio_end_io_t end_io_func) + bio_end_io_t end_io_func, + int mirror_num) { int ret = 0; struct bio *bio; @@ -1749,7 +1753,7 @@ static int submit_extent_page(int rw, struct extent_io_tree *tree, (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); + ret = submit_one_bio(rw, bio, mirror_num); bio = NULL; } else { return 0; @@ -1769,7 +1773,7 @@ static int submit_extent_page(int rw, struct extent_io_tree *tree, if (bio_ret) { *bio_ret = bio; } else { - ret = submit_one_bio(rw, bio); + ret = submit_one_bio(rw, bio, mirror_num); } return ret; @@ -1798,7 +1802,7 @@ 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) + struct bio **bio, int mirror_num) { struct inode *inode = page->mapping->host; u64 start = (u64)page->index << PAGE_CACHE_SHIFT; @@ -1901,7 +1905,7 @@ static int __extent_read_full_page(struct extent_io_tree *tree, ret = submit_extent_page(READ, tree, page, sector, iosize, page_offset, bdev, bio, nr, - end_bio_extent_readpage); + end_bio_extent_readpage, mirror_num); } if (ret) SetPageError(page); @@ -1923,9 +1927,9 @@ int extent_read_full_page(struct extent_io_tree *tree, struct page *page, struct bio *bio = NULL; int ret; - ret = __extent_read_full_page(tree, page, get_extent, &bio); + ret = __extent_read_full_page(tree, page, get_extent, &bio, 0); if (bio) - submit_one_bio(READ, bio); + submit_one_bio(READ, bio, 0); return ret; } EXPORT_SYMBOL(extent_read_full_page); @@ -2077,7 +2081,7 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, ret = submit_extent_page(WRITE, tree, page, sector, iosize, page_offset, bdev, &epd->bio, max_nr, - end_bio_extent_writepage); + end_bio_extent_writepage, 0); if (ret) SetPageError(page); } @@ -2244,7 +2248,7 @@ int extent_write_full_page(struct extent_io_tree *tree, struct page *page, write_cache_pages(mapping, &wbc_writepages, __extent_writepage, &epd); if (epd.bio) { - submit_one_bio(WRITE, epd.bio); + submit_one_bio(WRITE, epd.bio, 0); } return ret; } @@ -2265,7 +2269,7 @@ int extent_writepages(struct extent_io_tree *tree, ret = write_cache_pages(mapping, wbc, __extent_writepage, &epd); if (epd.bio) { - submit_one_bio(WRITE, epd.bio); + submit_one_bio(WRITE, epd.bio, 0); } return ret; } @@ -2297,7 +2301,8 @@ int extent_readpages(struct extent_io_tree *tree, page_cache_get(page); if (!pagevec_add(&pvec, page)) __pagevec_lru_add(&pvec); - __extent_read_full_page(tree, page, get_extent, &bio); + __extent_read_full_page(tree, page, get_extent, + &bio, 0); } page_cache_release(page); } @@ -2305,7 +2310,7 @@ int extent_readpages(struct extent_io_tree *tree, __pagevec_lru_add(&pvec); BUG_ON(!list_empty(pages)); if (bio) - submit_one_bio(READ, bio); + submit_one_bio(READ, bio, 0); return 0; } EXPORT_SYMBOL(extent_readpages); @@ -2430,7 +2435,7 @@ int extent_prepare_write(struct extent_io_tree *tree, ret = submit_extent_page(READ, tree, page, sector, iosize, page_offset, em->bdev, NULL, 1, - end_bio_extent_preparewrite); + end_bio_extent_preparewrite, 0); iocount++; block_start = block_start + iosize; } else { @@ -2696,6 +2701,7 @@ struct extent_buffer *alloc_extent_buffer(struct extent_io_tree *tree, mark_page_accessed(page0); set_page_extent_mapped(page0); set_page_extent_head(page0, len); + uptodate = PageUptodate(page0); } else { i = 0; } @@ -3006,7 +3012,7 @@ EXPORT_SYMBOL(extent_buffer_uptodate); int read_extent_buffer_pages(struct extent_io_tree *tree, struct extent_buffer *eb, u64 start, int wait, - get_extent_t *get_extent) + get_extent_t *get_extent, int mirror_num) { unsigned long i; unsigned long start_i; @@ -3062,8 +3068,10 @@ int read_extent_buffer_pages(struct extent_io_tree *tree, if (!PageUptodate(page)) { if (start_i == 0) inc_all_pages = 1; + ClearPageError(page); err = __extent_read_full_page(tree, page, - get_extent, &bio); + get_extent, &bio, + mirror_num); if (err) { ret = err; } @@ -3073,7 +3081,7 @@ int read_extent_buffer_pages(struct extent_io_tree *tree, } if (bio) - submit_one_bio(READ, bio); + submit_one_bio(READ, bio, mirror_num); if (ret || !wait) { return ret; -- cgit v1.2.2 From 7e38326f5b125288a2baea1d815c500502ab9fc0 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 9 Apr 2008 16:28:12 -0400 Subject: Btrfs: Handle checksumming errors while reading data blocks Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 9 +++++++++ 1 file changed, 9 insertions(+) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 2f159375c878..866460c3d72c 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -1534,6 +1534,15 @@ static int end_bio_extent_readpage(struct bio *bio, 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); + if (ret == 0) { + state = NULL; + continue; + } + } spin_lock_irqsave(&tree->lock, flags); if (!state || state->end != end) { -- cgit v1.2.2 From 594994aa3ea4cb52cba9c83a4b6f03e6ff3a646e Mon Sep 17 00:00:00 2001 From: Miguel Date: Fri, 11 Apr 2008 15:46:48 -0400 Subject: Btrfs: define write_cache_pages for linux kernel <= 2.6.20 instead write_cache_pages doesn't exist in linux 2.6.20, change the #if condition to match that. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 866460c3d72c..88322684be6a 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -2109,8 +2109,7 @@ done: return 0; } -#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) - +#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,20) /* 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); -- cgit v1.2.2 From 3b951516ed703af0f6d82053937655ad69b60864 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 17 Apr 2008 11:29:12 -0400 Subject: Btrfs: Use the extent map cache to find the logical disk block during data retries The data read retry code needs to find the logical disk block before it can resubmit new bios. But, finding this block isn't allowed to take the fs_mutex because that will deadlock with a number of different callers. This changes the retry code to use the extent map cache instead, but that requires the extent map cache to have the extent we're looking for. This is a problem because btrfs_drop_extent_cache just drops the entire extent instead of the little tiny part it is invalidating. The bulk of the code in this patch changes btrfs_drop_extent_cache to invalidate only a portion of the extent cache, and changes btrfs_get_extent to deal with the results. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 12 ++++++++---- 1 file changed, 8 insertions(+), 4 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 88322684be6a..21597bea21fd 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -1025,7 +1025,8 @@ u64 find_lock_delalloc_range(struct extent_io_tree *tree, search_again: node = tree_search(tree, cur_start); if (!node) { - *end = (u64)-1; + if (!found) + *end = (u64)-1; goto out; } @@ -1540,6 +1541,8 @@ static int end_bio_extent_readpage(struct bio *bio, start, end, state); if (ret == 0) { state = NULL; + uptodate = + test_bit(BIO_UPTODATE, &bio->bi_flags); continue; } } @@ -1555,10 +1558,11 @@ static int end_bio_extent_readpage(struct bio *bio, !(state->state & EXTENT_LOCKED)) state = NULL; } - if (!state && uptodate) { + if (!state) { spin_unlock_irqrestore(&tree->lock, flags); - set_extent_uptodate(tree, start, end, - GFP_ATOMIC); + if (uptodate) + set_extent_uptodate(tree, start, end, + GFP_ATOMIC); unlock_extent(tree, start, end, GFP_ATOMIC); goto next_io; } -- cgit v1.2.2 From 41471e8341a6b476bcd8ecc765a8b297c22a74f9 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 17 Apr 2008 14:08:30 -0400 Subject: Btrfs: Remove bogus max_sector warnings from the extent_io code It was testing the bio before doing logical->physical mapping, so the test was always wrong. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 7 ------- 1 file changed, 7 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 21597bea21fd..a984a7bc1197 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -1702,7 +1702,6 @@ extent_bio_alloc(struct block_device *bdev, u64 first_sector, int nr_vecs, static int submit_one_bio(int rw, struct bio *bio, int mirror_num) { - u64 maxsector; int ret = 0; struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1; struct page *page = bvec->bv_page; @@ -1730,12 +1729,6 @@ static int submit_one_bio(int rw, struct bio *bio, int mirror_num) bio_get(bio); - maxsector = bio->bi_bdev->bd_inode->i_size >> 9; - if (maxsector < bio->bi_sector) { - printk("sector too large max %Lu got %llu\n", maxsector, - (unsigned long long)bio->bi_sector); - WARN_ON(1); - } if (tree->ops && tree->ops->submit_bio_hook) tree->ops->submit_bio_hook(page->mapping->host, rw, bio, mirror_num); -- cgit v1.2.2 From 7b13b7b119c932a5eca486db4113f4c1fe3b97a8 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Fri, 18 Apr 2008 10:29:50 -0400 Subject: Btrfs: Don't drop extent_map cache during releasepage on the btree inode The btree inode should only have a single extent_map in the cache, it doesn't make sense to ever drop it. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 38 +++++++++++++++++++++++++++----------- 1 file changed, 27 insertions(+), 11 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index a984a7bc1197..81f8b4fd069b 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -2463,6 +2463,31 @@ err: } EXPORT_SYMBOL(extent_prepare_write); +/* + * a helper for releasepage, this tests for areas of the page that + * are locked or under IO and drops the related state bits if it is safe + * to drop the page. + */ +int try_release_extent_state(struct extent_map_tree *map, + struct extent_io_tree *tree, struct page *page, + gfp_t mask) +{ + u64 start = (u64)page->index << PAGE_CACHE_SHIFT; + u64 end = start + PAGE_CACHE_SIZE - 1; + int ret = 1; + + if (test_range_bit(tree, start, end, EXTENT_IOBITS, 0)) + ret = 0; + else { + if ((mask & GFP_NOFS) == GFP_NOFS) + mask = GFP_NOFS; + clear_extent_bit(tree, start, end, EXTENT_UPTODATE, + 1, 1, mask); + } + return ret; +} +EXPORT_SYMBOL(try_release_extent_state); + /* * a helper for releasepage. As long as there are no locked extents * in the range corresponding to the page, both state records and extent @@ -2475,8 +2500,7 @@ int try_release_extent_mapping(struct extent_map_tree *map, struct extent_map *em; u64 start = (u64)page->index << PAGE_CACHE_SHIFT; u64 end = start + PAGE_CACHE_SIZE - 1; - u64 orig_start = start; - int ret = 1; + if ((mask & __GFP_WAIT) && page->mapping->host->i_size > 16 * 1024 * 1024) { u64 len; @@ -2507,15 +2531,7 @@ int try_release_extent_mapping(struct extent_map_tree *map, free_extent_map(em); } } - if (test_range_bit(tree, orig_start, end, EXTENT_IOBITS, 0)) - ret = 0; - else { - if ((mask & GFP_NOFS) == GFP_NOFS) - mask = GFP_NOFS; - clear_extent_bit(tree, orig_start, end, EXTENT_UPTODATE, - 1, 1, mask); - } - return ret; + return try_release_extent_state(map, tree, page, mask); } EXPORT_SYMBOL(try_release_extent_mapping); -- cgit v1.2.2 From e1c4b7451e22f5b0a9fbccfa560ee7b80c35b8cd Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 22 Apr 2008 13:26:46 -0400 Subject: Fix btrfs_get_extent and get_block corner cases, and disable O_DIRECT reads The generic O_DIRECT code assumes all the bios have the same bdev, which isn't true for multi-device btrfs. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 1 + 1 file changed, 1 insertion(+) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 81f8b4fd069b..6e4bf029c6d4 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -1694,6 +1694,7 @@ extent_bio_alloc(struct block_device *bdev, u64 first_sector, int nr_vecs, } if (bio) { + bio->bi_size = 0; bio->bi_bdev = bdev; bio->bi_sector = first_sector; } -- cgit v1.2.2 From 004fb5750615bb3cf53e2aa50f6ef1ea57d97df9 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Fri, 25 Apr 2008 09:00:55 -0400 Subject: Btrfs: write_extent_pages came in 2.6.23 Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 6e4bf029c6d4..c1c980d1fbbf 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -2107,7 +2107,7 @@ done: return 0; } -#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,20) +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23) /* 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); -- cgit v1.2.2 From 5e478dc9828ad33d7b08dcdf277e13f14a7c1be7 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Fri, 25 Apr 2008 09:10:45 -0400 Subject: Btrfs: write_cache_pages came in 2.6.22 Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index c1c980d1fbbf..2368536629e7 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -2107,7 +2107,7 @@ done: return 0; } -#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23) +#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); -- cgit v1.2.2 From 4235298e4fc3c1a09f659cfe2fd285024eeb2241 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 28 Apr 2008 16:40:52 -0400 Subject: Btrfs: Drop some verbose printks Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 13 +++++-------- 1 file changed, 5 insertions(+), 8 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 2368536629e7..dd403b426ff5 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -3003,17 +3003,18 @@ int extent_buffer_uptodate(struct extent_io_tree *tree, struct extent_buffer *eb) { int ret = 0; - int ret2; unsigned long num_pages; unsigned long i; struct page *page; int pg_uptodate = 1; if (eb->flags & EXTENT_UPTODATE) - ret = 1; + return 1; - ret2 = test_range_bit(tree, eb->start, eb->start + eb->len - 1, + ret = test_range_bit(tree, eb->start, eb->start + eb->len - 1, EXTENT_UPTODATE, 1); + if (ret) + return ret; num_pages = num_extent_pages(eb->start, eb->len); for (i = 0; i < num_pages; i++) { @@ -3023,11 +3024,7 @@ int extent_buffer_uptodate(struct extent_io_tree *tree, break; } } - if ((ret || ret2) && !pg_uptodate) { -printk("uptodate error2 eb %Lu ret %d ret2 %d pg_uptodate %d\n", eb->start, ret, ret2, pg_uptodate); - WARN_ON(1); - } - return (ret || ret2); + return pg_uptodate; } EXPORT_SYMBOL(extent_buffer_uptodate); -- cgit v1.2.2 From 1259ab75c62462b8ffad90067b5e1f6312786a18 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 12 May 2008 13:39:03 -0400 Subject: Btrfs: Handle write errors on raid1 and raid10 When duplicate copies exist, writes are allowed to fail to one of those copies. This changeset includes a few changes that allow the FS to continue even when some IOs fail. It also adds verification of the parent generation number for btree blocks. This generation is stored in the pointer to a block, and it ensures that missed writes to are detected. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 49 +++++++++++++++++++++++++++++++++++++++++-------- 1 file changed, 41 insertions(+), 8 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index dd403b426ff5..2a3624adc0cf 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -1366,7 +1366,7 @@ static int end_bio_extent_writepage(struct bio *bio, unsigned int bytes_done, int err) #endif { - const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags); + 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; @@ -1375,6 +1375,7 @@ static int end_bio_extent_writepage(struct bio *bio, u64 end; u64 cur; int whole_page; + int ret; unsigned long flags; #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) @@ -1395,17 +1396,30 @@ static int end_bio_extent_writepage(struct bio *bio, if (--bvec >= bio->bi_io_vec) 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); + if (ret) + uptodate = 0; + } + + if (!uptodate && tree->ops && + tree->ops->writepage_io_failed_hook) { + ret = tree->ops->writepage_io_failed_hook(bio, page, + start, end, state); + if (ret == 0) { + state = NULL; + uptodate = (err == 0); + continue; + } + } + if (!uptodate) { clear_extent_uptodate(tree, start, end, GFP_ATOMIC); ClearPageUptodate(page); SetPageError(page); } - if (tree->ops && tree->ops->writepage_end_io_hook) { - tree->ops->writepage_end_io_hook(page, start, end, - state); - } - /* * bios can get merged in funny ways, and so we need to * be careful with the state variable. We know the @@ -2073,9 +2087,9 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, } else { ret = 0; } - if (ret) + if (ret) { SetPageError(page); - else { + } else { unsigned long max_nr = end_index + 1; set_range_writeback(tree, cur, cur + iosize - 1); if (!PageWriteback(page)) { @@ -2948,6 +2962,25 @@ int set_extent_buffer_dirty(struct extent_io_tree *tree, } EXPORT_SYMBOL(set_extent_buffer_dirty); +int clear_extent_buffer_uptodate(struct extent_io_tree *tree, + struct extent_buffer *eb) +{ + unsigned long i; + struct page *page; + unsigned long num_pages; + + num_pages = num_extent_pages(eb->start, eb->len); + eb->flags &= ~EXTENT_UPTODATE; + + 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); + } + return 0; +} + int set_extent_buffer_uptodate(struct extent_io_tree *tree, struct extent_buffer *eb) { -- cgit v1.2.2 From 211c17f51f46dc6c308c742098273dd46b5ca59c Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 15 May 2008 09:13:45 -0400 Subject: Fix corners in writepage and btrfs_truncate_page The extent_io writepage calls needed an extra check for discarding pages that started on th last byte in the file. btrfs_truncate_page needed checks to make sure the page was still part of the file after reading it, and most importantly, needed to wait for all IO to the page to finish before freeing the corresponding extents on disk. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 15 +++++++++------ 1 file changed, 9 insertions(+), 6 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 2a3624adc0cf..17c508a941eb 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -1988,9 +1988,12 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, u64 nr_delalloc; u64 delalloc_end; + WARN_ON(!PageLocked(page)); - if (page->index > end_index) { - clear_extent_dirty(tree, start, page_end, GFP_NOFS); + page_offset = i_size & (PAGE_CACHE_SIZE - 1); + if (page->index > end_index || + (page->index == end_index && !page_offset)) { + page->mapping->a_ops->invalidatepage(page, 0); unlock_page(page); return 0; } @@ -1998,13 +2001,13 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, if (page->index == end_index) { char *userpage; - size_t offset = i_size & (PAGE_CACHE_SIZE - 1); - userpage = kmap_atomic(page, KM_USER0); - memset(userpage + offset, 0, PAGE_CACHE_SIZE - offset); - flush_dcache_page(page); + memset(userpage + page_offset, 0, + PAGE_CACHE_SIZE - page_offset); kunmap_atomic(userpage, KM_USER0); + flush_dcache_page(page); } + page_offset = 0; set_page_extent_mapped(page); -- cgit v1.2.2 From 925baeddc5b0764a53f2214a1253251bab0e0324 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 25 Jun 2008 16:01:30 -0400 Subject: Btrfs: Start btree concurrency work. The allocation trees and the chunk trees are serialized via their own dedicated mutexes. This means allocation location is still not very fine grained. The main FS btree is protected by locks on each block in the btree. Locks are taken top / down, and as processing finishes on a given level of the tree, the lock is released after locking the lower level. The end result of a search is now a path where only the lowest level is locked. Releasing or freeing the path drops any locks held. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 8 -------- 1 file changed, 8 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 17c508a941eb..bd15cdcaba95 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -2889,7 +2889,6 @@ int clear_extent_buffer_dirty(struct extent_io_tree *tree, for (i = 0; i < num_pages; i++) { page = extent_buffer_page(eb, i); - lock_page(page); if (i == 0) set_page_extent_head(page, eb->len); else @@ -2907,7 +2906,6 @@ int clear_extent_buffer_dirty(struct extent_io_tree *tree, end = start + PAGE_CACHE_SIZE - 1; if (test_range_bit(tree, start, end, EXTENT_DIRTY, 0)) { - unlock_page(page); continue; } } @@ -2919,7 +2917,6 @@ int clear_extent_buffer_dirty(struct extent_io_tree *tree, PAGECACHE_TAG_DIRTY); } read_unlock_irq(&page->mapping->tree_lock); - unlock_page(page); } return 0; } @@ -2948,17 +2945,12 @@ int set_extent_buffer_dirty(struct extent_io_tree *tree, * on us if the page isn't already dirty. */ if (i == 0) { - lock_page(page); set_page_extent_head(page, eb->len); } else if (PagePrivate(page) && page->private != EXTENT_PAGE_PRIVATE) { - lock_page(page); set_page_extent_mapped(page); - unlock_page(page); } __set_page_dirty_nobuffers(extent_buffer_page(eb, i)); - if (i == 0) - unlock_page(page); } return set_extent_dirty(tree, eb->start, eb->start + eb->len - 1, GFP_NOFS); -- cgit v1.2.2 From 079899c2384023cd8efcd3806680b4f1d2abbd54 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 25 Jun 2008 16:01:31 -0400 Subject: Btrfs: Change find_extent_buffer to use TestSetPageLocked This makes it possible for callers to check for extent_buffers in cache without deadlocking against any btree locks held. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index bd15cdcaba95..40a5f53cb040 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -2808,10 +2808,15 @@ struct extent_buffer *find_extent_buffer(struct extent_io_tree *tree, goto lru_add; for (i = 0; i < num_pages; i++, index++) { - p = find_lock_page(mapping, 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); -- cgit v1.2.2 From e6dcd2dc9c489108648e2ed543315dd134d50a9a Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 17 Jul 2008 12:53:50 -0400 Subject: Btrfs: New data=ordered implementation The old data=ordered code would force commit to wait until all the data extents from the transaction were fully on disk. This introduced large latencies into the commit and stalled new writers in the transaction for a long time. The new code changes the way data allocations and extents work: * When delayed allocation is filled, data extents are reserved, and the extent bit EXTENT_ORDERED is set on the entire range of the extent. A struct btrfs_ordered_extent is allocated an inserted into a per-inode rbtree to track the pending extents. * As each page is written EXTENT_ORDERED is cleared on the bytes corresponding to that page. * When all of the bytes corresponding to a single struct btrfs_ordered_extent are written, The previously reserved extent is inserted into the FS btree and into the extent allocation trees. The checksums for the file data are also updated. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 52 +++++++++++++++++++++++++++++++++++++++++++++------- 1 file changed, 45 insertions(+), 7 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 40a5f53cb040..3f82a6e9ca4f 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -793,6 +793,13 @@ int set_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end, } 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); +} +EXPORT_SYMBOL(set_extent_ordered); + int set_extent_bits(struct extent_io_tree *tree, u64 start, u64 end, int bits, gfp_t mask) { @@ -812,8 +819,8 @@ 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, + 0, NULL, mask); } EXPORT_SYMBOL(set_extent_delalloc); @@ -825,6 +832,13 @@ int clear_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end, } 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) { @@ -1395,10 +1409,9 @@ static int end_bio_extent_writepage(struct bio *bio, if (--bvec >= bio->bi_io_vec) 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); + end, state, uptodate); if (ret) uptodate = 0; } @@ -1868,9 +1881,14 @@ static int __extent_read_full_page(struct extent_io_tree *tree, unlock_extent(tree, cur, end, GFP_NOFS); 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); iosize = min(extent_map_end(em) - cur, end - cur + 1); @@ -1976,6 +1994,7 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, u64 last_byte = i_size_read(inode); u64 block_start; u64 iosize; + u64 unlock_start; sector_t sector; struct extent_map *em; struct block_device *bdev; @@ -1988,7 +2007,6 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, u64 nr_delalloc; u64 delalloc_end; - WARN_ON(!PageLocked(page)); page_offset = i_size & (PAGE_CACHE_SIZE - 1); if (page->index > end_index || @@ -2030,6 +2048,7 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, delalloc_start = delalloc_end + 1; } lock_extent(tree, start, page_end, GFP_NOFS); + unlock_start = start; end = page_end; if (test_range_bit(tree, start, page_end, EXTENT_DELALLOC, 0)) { @@ -2038,6 +2057,11 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, 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); + unlock_start = page_end + 1; goto done; } @@ -2047,6 +2071,11 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, 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, @@ -2071,8 +2100,16 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, 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) + tree->ops->writepage_end_io_hook(page, cur, + cur + iosize - 1, + NULL, 1); cur = cur + iosize; page_offset += iosize; + unlock_start = cur; continue; } @@ -2119,7 +2156,8 @@ done: set_page_writeback(page); end_page_writeback(page); } - unlock_extent(tree, start, page_end, GFP_NOFS); + if (unlock_start <= page_end) + unlock_extent(tree, unlock_start, page_end, GFP_NOFS); unlock_page(page); return 0; } -- cgit v1.2.2 From 247e743cbe6e655768c3679f84821e03c1577902 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 17 Jul 2008 12:53:51 -0400 Subject: Btrfs: Use async helpers to deal with pages that have been improperly dirtied Higher layers sometimes call set_page_dirty without asking the filesystem to help. This causes many problems for the data=ordered and cow code. This commit detects pages that haven't been properly setup for IO and kicks off an async helper to deal with them. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 10 ++++++++++ 1 file changed, 10 insertions(+) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 3f82a6e9ca4f..feff16cb9b40 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -2050,6 +2050,16 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, 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); + if (ret == -EAGAIN) { + unlock_extent(tree, start, page_end, GFP_NOFS); + redirty_page_for_writepage(wbc, page); + unlock_page(page); + return 0; + } + } + end = page_end; if (test_range_bit(tree, start, page_end, EXTENT_DELALLOC, 0)) { printk("found delalloc bits after lock_extent\n"); -- cgit v1.2.2 From 211f90e68b679d27fe23c5505f86d6ce62c98bae Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Fri, 18 Jul 2008 11:56:15 -0400 Subject: Btrfs: Don't allow releasepage to succeed if EXTENT_ORDERED is set Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index feff16cb9b40..3153b4fdd0b3 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -2542,7 +2542,8 @@ int try_release_extent_state(struct extent_map_tree *map, u64 end = start + PAGE_CACHE_SIZE - 1; int ret = 1; - if (test_range_bit(tree, start, end, EXTENT_IOBITS, 0)) + if (test_range_bit(tree, start, end, + EXTENT_IOBITS | EXTENT_ORDERED, 0)) ret = 0; else { if ((mask & GFP_NOFS) == GFP_NOFS) -- cgit v1.2.2 From 7f3c74fb831fa19bafe087e817c0a5ff3883f1ea Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Fri, 18 Jul 2008 12:01:11 -0400 Subject: Btrfs: Keep extent mappings in ram until pending ordered extents are done It was possible for stale mappings from disk to be used instead of the new pending ordered extent. This adds a flag to the extent map struct to keep it pinned until the pending ordered extent is actually on disk. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 27 +++++++++++++++------------ 1 file changed, 15 insertions(+), 12 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 3153b4fdd0b3..d4a63ae7ed1b 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -2000,7 +2000,7 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, 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; @@ -2008,9 +2008,9 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, u64 delalloc_end; 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; @@ -2020,12 +2020,12 @@ 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); @@ -2088,7 +2088,7 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, 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); @@ -2113,12 +2113,13 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, unlock_extent(tree, unlock_start, cur + iosize -1, GFP_NOFS); + if (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; + pg_offset += iosize; unlock_start = cur; continue; } @@ -2127,7 +2128,7 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, if (0 && !test_range_bit(tree, cur, cur + iosize - 1, EXTENT_DIRTY, 0)) { cur = cur + iosize; - page_offset += iosize; + pg_offset += iosize; continue; } clear_extent_dirty(tree, cur, cur + iosize - 1, GFP_NOFS); @@ -2141,6 +2142,7 @@ 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, " @@ -2150,14 +2152,14 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, } ret = submit_extent_page(WRITE, tree, page, sector, - iosize, page_offset, bdev, + iosize, pg_offset, bdev, &epd->bio, max_nr, end_bio_extent_writepage, 0); if (ret) SetPageError(page); } cur = cur + iosize; - page_offset += iosize; + pg_offset += iosize; nr++; } done: @@ -2579,7 +2581,8 @@ int try_release_extent_mapping(struct extent_map_tree *map, spin_unlock(&map->lock); break; } - if (em->start != start) { + if (test_bit(EXTENT_FLAG_PINNED, &em->flags) || + em->start != start) { spin_unlock(&map->lock); free_extent_map(em); break; -- cgit v1.2.2 From 6af118ce51b52ceda357c671550c79628b9c4a65 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 22 Jul 2008 11:18:07 -0400 Subject: Btrfs: Index extent buffers in an rbtree Before, extent buffers were a temporary object, meant to map a number of pages at once and collect operations on them. But, a few extra fields have crept in, and they are also the best place to store a per-tree block lock field as well. This commit puts the extent buffers into an rbtree, and ensures a single extent buffer for each tree block. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 309 +++++++++++++++++++-------------------------------- 1 file changed, 116 insertions(+), 193 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index d4a63ae7ed1b..32bb4ed3723d 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -91,29 +91,16 @@ 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) { struct extent_state *state; @@ -245,6 +232,50 @@ static inline struct rb_node *tree_search(struct extent_io_tree *tree, 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; +} + /* * utility function to look for merge candidates inside a given range. * Any extents with matching state are merged together into a single @@ -1817,9 +1848,8 @@ 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); } } @@ -2627,51 +2657,6 @@ out: return sector; } -static int add_lru(struct extent_io_tree *tree, struct extent_buffer *eb) -{ - 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); - } - } 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 (list_empty(lru)) - return NULL; - - do { - eb = list_entry(cur, struct extent_buffer, lru); - if (eb->start == start && eb->len == len) { - extent_buffer_get(eb); - return eb; - } - 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); -} - static inline struct page *extent_buffer_page(struct extent_buffer *eb, unsigned long i) { @@ -2688,44 +2673,10 @@ static inline struct page *extent_buffer_page(struct extent_buffer *eb, return p; } -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; - - 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; -} - - -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, @@ -2736,15 +2687,7 @@ static struct extent_buffer *__alloc_extent_buffer(struct extent_io_tree *tree, struct extent_buffer *eb = NULL; unsigned long flags; - spin_lock(&tree->lru_lock); - eb = find_lru(tree, start, len); - spin_unlock(&tree->lru_lock); - if (eb) { - return eb; - } - eb = kmem_cache_zalloc(extent_buffer_cache, mask); - INIT_LIST_HEAD(&eb->lru); eb->start = start; eb->len = len; spin_lock_irqsave(&leak_lock, flags); @@ -2773,17 +2716,24 @@ 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); + 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; @@ -2800,7 +2750,7 @@ 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); @@ -2818,25 +2768,28 @@ struct extent_buffer *alloc_extent_buffer(struct extent_io_tree *tree, eb->flags |= EXTENT_UPTODATE; eb->flags |= EXTENT_BUFFER_FILLED; -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); @@ -2844,89 +2797,27 @@ 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); - } - - if (!PageUptodate(p)) - uptodate = 0; - unlock_page(p); - } - if (uptodate) - eb->flags |= EXTENT_UPTODATE; - eb->flags |= EXTENT_BUFFER_FILLED; + spin_lock(&tree->buffer_lock); + eb = buffer_search(tree, start); + if (eb) + atomic_inc(&eb->refs); + spin_unlock(&tree->buffer_lock); -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); @@ -3583,3 +3474,35 @@ void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset, } } 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; + } + /* 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++) { + struct page *page = extent_buffer_page(eb, i); + page_cache_release(page); + } + rb_erase(&eb->rb_node, &tree->buffer); + __free_extent_buffer(eb); +out: + spin_unlock(&tree->buffer_lock); + return ret; +} +EXPORT_SYMBOL(try_release_extent_buffer); + -- cgit v1.2.2 From a61e6f29dc7c9d56a776a518eed92bbc61848263 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 22 Jul 2008 11:18:08 -0400 Subject: Btrfs: Use a mutex in the extent buffer for tree block locking This replaces the use of the page cache lock bit for locking, which wasn't suitable for block size < page size and couldn't be used recursively. The mutexes alone don't fix either problem, but they are the first step. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 9 +++++++++ 1 file changed, 9 insertions(+) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 32bb4ed3723d..7380449cb5b3 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -2690,6 +2690,7 @@ static struct extent_buffer *__alloc_extent_buffer(struct extent_io_tree *tree, eb = kmem_cache_zalloc(extent_buffer_cache, mask); eb->start = start; eb->len = len; + mutex_init(&eb->mutex); spin_lock_irqsave(&leak_lock, flags); list_add(&eb->leak_list, &buffers); spin_unlock_irqrestore(&leak_lock, flags); @@ -2837,6 +2838,7 @@ int clear_extent_buffer_dirty(struct extent_io_tree *tree, for (i = 0; i < num_pages; i++) { page = extent_buffer_page(eb, i); + lock_page(page); if (i == 0) set_page_extent_head(page, eb->len); else @@ -2854,6 +2856,7 @@ int clear_extent_buffer_dirty(struct extent_io_tree *tree, end = start + PAGE_CACHE_SIZE - 1; if (test_range_bit(tree, start, end, EXTENT_DIRTY, 0)) { + unlock_page(page); continue; } } @@ -2865,6 +2868,7 @@ int clear_extent_buffer_dirty(struct extent_io_tree *tree, PAGECACHE_TAG_DIRTY); } read_unlock_irq(&page->mapping->tree_lock); + unlock_page(page); } return 0; } @@ -2893,12 +2897,17 @@ int set_extent_buffer_dirty(struct extent_io_tree *tree, * on us if the page isn't already dirty. */ if (i == 0) { + lock_page(page); set_page_extent_head(page, eb->len); } else if (PagePrivate(page) && page->private != EXTENT_PAGE_PRIVATE) { + lock_page(page); set_page_extent_mapped(page); + unlock_page(page); } __set_page_dirty_nobuffers(extent_buffer_page(eb, i)); + if (i == 0) + unlock_page(page); } return set_extent_dirty(tree, eb->start, eb->start + eb->len - 1, GFP_NOFS); -- cgit v1.2.2 From f421950f86bf96a11fef932e167ab2e70d4c43a0 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 22 Jul 2008 11:18:09 -0400 Subject: Btrfs: Fix some data=ordered related data corruptions Stress testing was showing data checksum errors, most of which were caused by a lookup bug in the extent_map tree. The tree was caching the last pointer returned, and searches would check the last pointer first. But, search callers also expect the search to return the very first matching extent in the range, which wasn't always true with the last pointer usage. For now, the code to cache the last return value is just removed. It is easy to fix, but I think lookups are rare enough that it isn't required anymore. This commit also replaces do_sync_mapping_range with a local copy of the related functions. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 20 -------------------- 1 file changed, 20 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 7380449cb5b3..9965993748d0 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -97,7 +97,6 @@ void extent_io_tree_init(struct extent_io_tree *tree, spin_lock_init(&tree->lock); spin_lock_init(&tree->buffer_lock); tree->mapping = mapping; - tree->last = NULL; } EXPORT_SYMBOL(extent_io_tree_init); @@ -173,12 +172,6 @@ static struct rb_node *__etree_search(struct extent_io_tree *tree, u64 offset, 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) { entry = rb_entry(n, struct tree_entry, rb_node); prev = n; @@ -189,7 +182,6 @@ static struct rb_node *__etree_search(struct extent_io_tree *tree, u64 offset, else if (offset > entry->end) n = n->rb_right; else { - tree->last = rb_entry(n, struct extent_state, rb_node); return n; } } @@ -223,10 +215,6 @@ static inline struct rb_node *tree_search(struct extent_io_tree *tree, ret = __etree_search(tree, offset, &prev, NULL); if (!ret) { - if (prev) { - tree->last = rb_entry(prev, struct extent_state, - rb_node); - } return prev; } return ret; @@ -301,8 +289,6 @@ static int merge_state(struct extent_io_tree *tree, other->state == state->state) { 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); } @@ -314,8 +300,6 @@ static int merge_state(struct extent_io_tree *tree, other->state == state->state) { 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); } @@ -378,7 +362,6 @@ static int insert_state(struct extent_io_tree *tree, return -EEXIST; } state->tree = tree; - tree->last = state; merge_state(tree, state); return 0; } @@ -444,9 +427,6 @@ static int clear_state_bit(struct extent_io_tree *tree, 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); -- cgit v1.2.2 From 89642229a582a5c2b6d2ed8ec16986387d9a9047 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 24 Jul 2008 09:41:53 -0400 Subject: Btrfs: Search data ordered extents first for checksums on read Checksum items are not inserted into the tree until all of the io from a given extent is complete. This means one dirty page from an extent may be written, freed, and then read again before the entire extent is on disk and the checksum item is inserted. The checksums themselves are stored in the ordered extent so they can be inserted in bulk when IO is complete. On read, if a checksum item isn't found, the ordered extents were being searched for a checksum record. This all worked most of the time, but the checksum insertion code tries to reduce the number of tree operations by pre-inserting checksum items based on i_size and a few other factors. This means the read code might find a checksum item that hasn't yet really been filled in. This commit changes things to check the ordered extents first and only dive into the btree if nothing was found. This removes the need for extra locking and is more reliable. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 9965993748d0..e3547a992d5c 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -1949,18 +1949,18 @@ 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, + bdev, bio, pnr, end_bio_extent_readpage, mirror_num); + nr++; } if (ret) SetPageError(page); cur = cur + iosize; page_offset += iosize; - nr++; } if (!nr) { if (!PageError(page)) -- cgit v1.2.2 From 33958dc6d38fb4ca7e62273855fcb2db7e616263 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 30 Jul 2008 10:29:12 -0400 Subject: Btrfs: Fix verify_parent_transid It was incorrectly clearing the up to date flag on the buffer even when the buffer properly verified. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index e3547a992d5c..964ec1622d66 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -2647,6 +2647,8 @@ 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; + if (!mapping) + return NULL; read_lock_irq(&mapping->tree_lock); p = radix_tree_lookup(&mapping->page_tree, i); read_unlock_irq(&mapping->tree_lock); @@ -2908,7 +2910,8 @@ int clear_extent_buffer_uptodate(struct extent_io_tree *tree, GFP_NOFS); for (i = 0; i < num_pages; i++) { page = extent_buffer_page(eb, i); - ClearPageUptodate(page); + if (page) + ClearPageUptodate(page); } return 0; } -- cgit v1.2.2 From bcc63abbf3e9bf948a1b0129b3e6120ec7d7f698 Mon Sep 17 00:00:00 2001 From: Yan Date: Wed, 30 Jul 2008 16:29:20 -0400 Subject: Btrfs: implement memory reclaim for leaf reference cache The memory reclaiming issue happens when snapshot exists. In that case, some cache entries may not be used during old snapshot dropping, so they will remain in the cache until umount. The patch adds a field to struct btrfs_leaf_ref to record create time. Besides, the patch makes all dead roots of a given snapshot linked together in order of create time. After a old snapshot was completely dropped, we check the dead root list and remove all cache entries created before the oldest dead root in the list. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 1 - 1 file changed, 1 deletion(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 964ec1622d66..5368e3b6eb96 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -3497,4 +3497,3 @@ out: return ret; } EXPORT_SYMBOL(try_release_extent_buffer); - -- cgit v1.2.2 From 0ee0fda06b943d7ef65f3cec50ab9d427a2baf97 Mon Sep 17 00:00:00 2001 From: Sven Wegener Date: Wed, 30 Jul 2008 16:54:26 -0400 Subject: Btrfs: Add compatibility for kernels >= 2.6.27-rc1 Add a couple of #if's to follow API changes. Signed-off-by: Sven Wegener Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 23 +++++++++++++++++++++++ 1 file changed, 23 insertions(+) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 5368e3b6eb96..1cf4baba4342 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -2649,9 +2649,24 @@ static inline struct page *extent_buffer_page(struct extent_buffer *eb, mapping = eb->first_page->mapping; if (!mapping) return NULL; + + /* + * 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. + */ +#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26) + rcu_read_lock(); +#else read_lock_irq(&mapping->tree_lock); +#endif p = radix_tree_lookup(&mapping->page_tree, i); + +#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26) + rcu_read_unlock(); +#else read_unlock_irq(&mapping->tree_lock); +#endif return p; } @@ -2843,13 +2858,21 @@ int clear_extent_buffer_dirty(struct extent_io_tree *tree, } } clear_page_dirty_for_io(page); +#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26) + spin_lock_irq(&page->mapping->tree_lock); +#else read_lock_irq(&page->mapping->tree_lock); +#endif if (!PageDirty(page)) { radix_tree_tag_clear(&page->mapping->page_tree, page_index(page), PAGECACHE_TAG_DIRTY); } +#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26) + spin_unlock_irq(&page->mapping->tree_lock); +#else read_unlock_irq(&page->mapping->tree_lock); +#endif unlock_page(page); } return 0; -- cgit v1.2.2 From 2db04966ae9a3eeb57f28df9aac4e77d5b287cb1 Mon Sep 17 00:00:00 2001 From: David Woodhouse Date: Thu, 7 Aug 2008 11:19:43 -0400 Subject: Btrfs: Change TestSetPageLocked() to trylock_page() Add backwards compatibility in compat.h Signed-off-by: David Woodhouse --- compat.h | 3 +++ extent_io.c | 3 ++- 2 files changed, 5 insertions(+), 1 deletions(-) Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 1cf4baba4342..f46f88620c70 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -14,6 +14,7 @@ #include #include "extent_io.h" #include "extent_map.h" +#include "compat.h" /* temporary define until extent_map moves out of btrfs */ struct kmem_cache *btrfs_cache_create(const char *name, size_t size, @@ -3055,7 +3056,7 @@ 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); -- cgit v1.2.2 From 902b22f341efa00be802418a0a8c57bddcd269a6 Mon Sep 17 00:00:00 2001 From: David Woodhouse Date: Wed, 20 Aug 2008 08:51:49 -0400 Subject: Btrfs: Remove broken optimisations in end_bio functions. These ended up freeing objects while they were still using them. Under guidance from Chris, just rip out the 'clever' bits and do things the simple way. Signed-off-by: David Woodhouse Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 159 +++++++-------------------------------------------- 1 file changed, 21 insertions(+), 138 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index f46f88620c70..83ba0c328722 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -15,6 +15,8 @@ #include "extent_io.h" #include "extent_map.h" #include "compat.h" +#include "ctree.h" +#include "btrfs_inode.h" /* temporary define until extent_map moves out of btrfs */ struct kmem_cache *btrfs_cache_create(const char *name, size_t size, @@ -1394,15 +1396,11 @@ static int end_bio_extent_writepage(struct bio *bio, { 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) @@ -1410,6 +1408,8 @@ static int end_bio_extent_writepage(struct bio *bio, #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; @@ -1423,7 +1423,7 @@ 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; } @@ -1431,9 +1431,8 @@ 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; } @@ -1445,68 +1444,7 @@ static int end_bio_extent_writepage(struct bio *bio, 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: + clear_extent_writeback(tree, start, end, GFP_ATOMIC); if (whole_page) end_page_writeback(page); @@ -1539,13 +1477,9 @@ static int end_bio_extent_readpage(struct bio *bio, { 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; @@ -1556,6 +1490,8 @@ static int end_bio_extent_readpage(struct bio *bio, 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; @@ -1570,80 +1506,26 @@ 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); 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); @@ -1683,8 +1565,7 @@ static int end_bio_extent_preparewrite(struct bio *bio, { 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; @@ -1695,6 +1576,8 @@ static int end_bio_extent_preparewrite(struct bio *bio, 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; @@ -1765,7 +1648,7 @@ static int submit_one_bio(int rw, struct bio *bio, int mirror_num) BUG_ON(state->end != end); spin_unlock_irq(&tree->lock); - bio->bi_private = state; + bio->bi_private = NULL; bio_get(bio); -- cgit v1.2.2 From a1b32a5932cfac7c38b442582285f3da2a09dfd8 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Fri, 5 Sep 2008 16:09:51 -0400 Subject: Btrfs: Add debugging checks to track down corrupted metadata Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 21 ++++++++++++++------- 1 file changed, 14 insertions(+), 7 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 83ba0c328722..7ca89c45d401 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -1811,6 +1811,7 @@ printk("2bad mapping end %Lu cur %Lu\n", end, cur); } /* the get_extent function already copied into the page */ if (test_range_bit(tree, cur, cur_end, EXTENT_UPTODATE, 1)) { + check_page_uptodate(tree, page); unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS); cur = cur + iosize; page_offset += iosize; @@ -2785,21 +2786,20 @@ int set_extent_buffer_dirty(struct extent_io_tree *tree, * properly set. releasepage may drop page->private * on us if the page isn't already dirty. */ + lock_page(page); if (i == 0) { - lock_page(page); set_page_extent_head(page, eb->len); } else if (PagePrivate(page) && page->private != EXTENT_PAGE_PRIVATE) { - lock_page(page); set_page_extent_mapped(page); - unlock_page(page); } __set_page_dirty_nobuffers(extent_buffer_page(eb, i)); - if (i == 0) - unlock_page(page); + set_extent_dirty(tree, page_offset(page), + page_offset(page) + PAGE_CACHE_SIZE -1, + GFP_NOFS); + unlock_page(page); } - return set_extent_dirty(tree, eb->start, - eb->start + eb->len - 1, GFP_NOFS); + return 0; } EXPORT_SYMBOL(set_extent_buffer_dirty); @@ -2952,6 +2952,9 @@ int read_extent_buffer_pages(struct extent_io_tree *tree, if (all_uptodate) { if (start_i == 0) eb->flags |= EXTENT_UPTODATE; + if (ret) { + printk("all up to date but ret is %d\n", ret); + } goto unlock_exit; } @@ -2968,6 +2971,7 @@ int read_extent_buffer_pages(struct extent_io_tree *tree, mirror_num); if (err) { ret = err; + printk("err %d from __extent_read_full_page\n", ret); } } else { unlock_page(page); @@ -2978,12 +2982,15 @@ int read_extent_buffer_pages(struct extent_io_tree *tree, submit_one_bio(READ, bio, mirror_num); if (ret || !wait) { + if (ret) + printk("ret %d wait %d returning\n", 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)) { + printk("page not uptodate after wait_on_page_locked\n"); ret = -EIO; } } -- cgit v1.2.2 From b214107eda845f9a5851ae198f5b972e0dc30c45 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Fri, 5 Sep 2008 16:43:31 -0400 Subject: Btrfs: trivial sparse fixes Fix a bunch of trivial sparse complaints. Signed-off-by: Christoph Hellwig Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 7ca89c45d401..239e7c908abf 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -3400,10 +3400,8 @@ int try_release_extent_buffer(struct extent_io_tree *tree, struct page *page) } /* 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++) { - struct page *page = extent_buffer_page(eb, i); - page_cache_release(page); - } + 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: -- cgit v1.2.2 From 4bef084857ab8fe71cf49eae349c25e440a49150 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 8 Sep 2008 11:18:08 -0400 Subject: Btrfs: Tree logging fixes * Pin down data blocks to prevent them from being reallocated like so: trans 1: allocate file extent trans 2: free file extent trans 3: free file extent during old snapshot deletion trans 3: allocate file extent to new file trans 3: fsync new file Before the tree logging code, this was legal because the fsync would commit the transation that did the final data extent free and the transaction that allocated the extent to the new file at the same time. With the tree logging code, the tree log subtransaction can commit before the transaction that freed the extent. If we crash, we're left with two different files using the extent. * Don't wait in start_transaction if log replay is going on. This avoids deadlocks from iput while we're cleaning up link counts in the replay code. * Don't deadlock in replay_one_name by trying to read an inode off the disk while holding paths for the directory * Hold the buffer lock while we mark a buffer as written. This closes a race where someone is changing a buffer while we write it. They are supposed to mark it dirty again after they change it, but this violates the cow rules. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 47 ++++++++++++++++++++++++++++++++++++----------- 1 file changed, 36 insertions(+), 11 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 239e7c908abf..319a0c7a4a58 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -29,7 +29,10 @@ static struct kmem_cache *extent_buffer_cache; static LIST_HEAD(buffers); static LIST_HEAD(states); + +#ifdef LEAK_DEBUG static spinlock_t leak_lock = SPIN_LOCK_UNLOCKED; +#endif #define BUFFER_LRU_MAX 64 @@ -106,7 +109,9 @@ EXPORT_SYMBOL(extent_io_tree_init); struct extent_state *alloc_extent_state(gfp_t mask) { struct extent_state *state; +#ifdef LEAK_DEBUG unsigned long flags; +#endif state = kmem_cache_alloc(extent_state_cache, mask); if (!state) @@ -114,10 +119,11 @@ struct extent_state *alloc_extent_state(gfp_t mask) state->state = 0; state->private = 0; state->tree = NULL; +#ifdef 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; @@ -129,11 +135,15 @@ void free_extent_state(struct extent_state *state) if (!state) return; if (atomic_dec_and_test(&state->refs)) { +#ifdef LEAK_DEBUG unsigned long flags; +#endif WARN_ON(state->tree); +#ifdef 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); } } @@ -2070,13 +2080,13 @@ done: } #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,22) -/* Taken directly from 2.6.23 for 2.6.18 back port */ +/* Taken directly from 2.6.23 with a mod for a lockpage hook */ typedef int (*writepage_t)(struct page *page, struct writeback_control *wbc, void *data); +#endif /** - * 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 @@ -2090,9 +2100,10 @@ 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) +int extent_write_cache_pages(struct extent_io_tree *tree, + struct address_space *mapping, + struct writeback_control *wbc, + writepage_t writepage, void *data) { struct backing_dev_info *bdi = mapping->backing_dev_info; int ret = 0; @@ -2138,7 +2149,10 @@ 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); @@ -2187,9 +2201,12 @@ retry: } if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0)) mapping->writeback_index = index; + + if (wbc->range_cont) + wbc->range_start = index << PAGE_CACHE_SHIFT; return ret; } -#endif +EXPORT_SYMBOL(extent_write_cache_pages); int extent_write_full_page(struct extent_io_tree *tree, struct page *page, get_extent_t *get_extent, @@ -2214,7 +2231,8 @@ int extent_write_full_page(struct extent_io_tree *tree, struct page *page, ret = __extent_writepage(page, wbc, &epd); - write_cache_pages(mapping, &wbc_writepages, __extent_writepage, &epd); + extent_write_cache_pages(tree, mapping, &wbc_writepages, + __extent_writepage, &epd); if (epd.bio) { submit_one_bio(WRITE, epd.bio, 0); } @@ -2235,7 +2253,8 @@ int extent_writepages(struct extent_io_tree *tree, .get_extent = get_extent, }; - ret = write_cache_pages(mapping, wbc, __extent_writepage, &epd); + ret = extent_write_cache_pages(tree, mapping, wbc, + __extent_writepage, &epd); if (epd.bio) { submit_one_bio(WRITE, epd.bio, 0); } @@ -2567,15 +2586,19 @@ static struct extent_buffer *__alloc_extent_buffer(struct extent_io_tree *tree, gfp_t mask) { struct extent_buffer *eb = NULL; +#ifdef LEAK_DEBUG unsigned long flags; +#endif eb = kmem_cache_zalloc(extent_buffer_cache, mask); eb->start = start; eb->len = len; mutex_init(&eb->mutex); +#ifdef 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; @@ -2583,10 +2606,12 @@ static struct extent_buffer *__alloc_extent_buffer(struct extent_io_tree *tree, static void __free_extent_buffer(struct extent_buffer *eb) { +#ifdef 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); } -- cgit v1.2.2 From 0f9dd46cda36b8de3b9f48bc42bd09d20b9c3b52 Mon Sep 17 00:00:00 2001 From: Josef Bacik Date: Tue, 23 Sep 2008 13:14:11 -0400 Subject: Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 319a0c7a4a58..8624f3e88036 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -2634,6 +2634,7 @@ struct extent_buffer *alloc_extent_buffer(struct extent_io_tree *tree, if (eb) { atomic_inc(&eb->refs); spin_unlock(&tree->buffer_lock); + mark_page_accessed(eb->first_page); return eb; } spin_unlock(&tree->buffer_lock); @@ -2713,6 +2714,9 @@ struct extent_buffer *find_extent_buffer(struct extent_io_tree *tree, atomic_inc(&eb->refs); spin_unlock(&tree->buffer_lock); + if (eb) + mark_page_accessed(eb->first_page); + return eb; } EXPORT_SYMBOL(find_extent_buffer); -- cgit v1.2.2 From 31840ae1a6b433ca0e6a8d341756ff478bbf959e Mon Sep 17 00:00:00 2001 From: Zheng Yan Date: Tue, 23 Sep 2008 13:14:14 -0400 Subject: Btrfs: Full back reference support This patch makes the back reference system to explicit record the location of parent node for all types of extents. The location of parent node is placed into the offset field of backref key. Every time a tree block is balanced, the back references for the affected lower level extents are updated. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 8624f3e88036..58ad25838a41 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -2201,9 +2201,10 @@ retry: } if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0)) mapping->writeback_index = index; - +#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26) if (wbc->range_cont) wbc->range_start = index << PAGE_CACHE_SHIFT; +#endif return ret; } EXPORT_SYMBOL(extent_write_cache_pages); -- cgit v1.2.2 From 2b1f55b0f0d0d1a66470ef4ea2696cd5dd741a12 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 24 Sep 2008 11:48:04 -0400 Subject: Remove Btrfs compat code for older kernels Btrfs had compatibility code for kernels back to 2.6.18. These have been removed, and will be maintained in a separate backport git tree from now on. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 66 +++------------------------------------------------- 1 file changed, 3 insertions(+), 63 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 58ad25838a41..e3a25be5c663 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -1397,12 +1397,7 @@ 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; @@ -1412,10 +1407,6 @@ static int end_bio_extent_writepage(struct bio *bio, int whole_page; int ret; -#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; @@ -1461,10 +1452,8 @@ static int end_bio_extent_writepage(struct bio *bio, 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 } /* @@ -1478,12 +1467,7 @@ static int end_bio_extent_writepage(struct bio *bio, * 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; @@ -1493,11 +1477,6 @@ static int end_bio_extent_readpage(struct bio *bio, int whole_page; int ret; -#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; @@ -1556,9 +1535,6 @@ static int end_bio_extent_readpage(struct bio *bio, } while (bvec >= bio->bi_io_vec); bio_put(bio); -#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) - return 0; -#endif } /* @@ -1566,12 +1542,7 @@ static int end_bio_extent_readpage(struct bio *bio, * 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; @@ -1579,11 +1550,6 @@ static int end_bio_extent_preparewrite(struct bio *bio, 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; @@ -1607,9 +1573,6 @@ 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 * @@ -2079,12 +2042,6 @@ done: return 0; } -#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,22) -/* Taken directly from 2.6.23 with a mod for a lockpage hook */ -typedef int (*writepage_t)(struct page *page, struct writeback_control *wbc, - void *data); -#endif - /** * 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 @@ -2201,10 +2158,9 @@ retry: } if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0)) mapping->writeback_index = index; -#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26) + if (wbc->range_cont) wbc->range_start = index << PAGE_CACHE_SHIFT; -#endif return ret; } EXPORT_SYMBOL(extent_write_cache_pages); @@ -2560,18 +2516,10 @@ static inline struct page *extent_buffer_page(struct extent_buffer *eb, * by increasing the reference count. So we know the page must * be in the radix tree. */ -#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26) rcu_read_lock(); -#else - read_lock_irq(&mapping->tree_lock); -#endif p = radix_tree_lookup(&mapping->page_tree, i); - -#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26) rcu_read_unlock(); -#else - read_unlock_irq(&mapping->tree_lock); -#endif + return p; } @@ -2773,21 +2721,13 @@ int clear_extent_buffer_dirty(struct extent_io_tree *tree, } } clear_page_dirty_for_io(page); -#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26) spin_lock_irq(&page->mapping->tree_lock); -#else - read_lock_irq(&page->mapping->tree_lock); -#endif if (!PageDirty(page)) { radix_tree_tag_clear(&page->mapping->page_tree, page_index(page), PAGECACHE_TAG_DIRTY); } -#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26) spin_unlock_irq(&page->mapping->tree_lock); -#else - read_unlock_irq(&page->mapping->tree_lock); -#endif unlock_page(page); } return 0; -- cgit v1.2.2 From 5b21f2ed3f2947b5195b65c9fdbdd9e52904cc03 Mon Sep 17 00:00:00 2001 From: Zheng Yan Date: Fri, 26 Sep 2008 10:05:38 -0400 Subject: Btrfs: extent_map and data=ordered fixes for space balancing * Add an EXTENT_BOUNDARY state bit to keep the writepage code from merging data extents that are in the process of being relocated. This allows us to do accounting for them properly. * The balancing code relocates data extents indepdent of the underlying inode. The extent_map code was modified to properly account for things moving around (invalidating extent_map caches in the inode). * Don't take the drop_mutex in the create_subvol ioctl. It isn't required. * Fix walking of the ordered extent list to avoid races with sys_unlink * Change the lock ordering rules. Transaction start goes outside the drop_mutex. This allows btrfs_commit_transaction to directly drop the relocation trees. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 13 +++++++++---- 1 file changed, 9 insertions(+), 4 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index e3a25be5c663..8bd1b402f3fd 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -292,7 +292,7 @@ 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); @@ -1070,7 +1070,8 @@ search_again: 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)) { @@ -1078,7 +1079,7 @@ search_again: *end = state->end; goto out; } - if (!found) { + if (!found && !(state->state & EXTENT_BOUNDARY)) { struct extent_state *prev_state; struct rb_node *prev_node = node; while(1) { @@ -1088,7 +1089,11 @@ search_again: prev_state = rb_entry(prev_node, struct extent_state, rb_node); - if (!(prev_state->state & EXTENT_DELALLOC)) + if ((prev_state->end + 1 != state->start) || + !(prev_state->state & EXTENT_DELALLOC)) + break; + if ((cur_start - prev_state->start) * 2 > + max_bytes) break; state = prev_state; node = prev_node; -- cgit v1.2.2 From d352ac68148b69937d39ca5d48bcc4478e118dbf Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 29 Sep 2008 15:18:18 -0400 Subject: Btrfs: add and improve comments This improves the comments at the top of many functions. It didn't dive into the guts of functions because I was trying to avoid merging problems with the new allocator and back reference work. extent-tree.c and volumes.c were both skipped, and there is definitely more work todo in cleaning and commenting the code. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 34 ++++++++++++++++++++++++++++++++-- 1 file changed, 32 insertions(+), 2 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 8bd1b402f3fd..563b2d12f4f2 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -914,6 +914,10 @@ int wait_on_extent_writeback(struct extent_io_tree *tree, u64 start, u64 end) } EXPORT_SYMBOL(wait_on_extent_writeback); +/* + * either insert or lock state struct between start and end use mask to tell + * us if waiting is desired. + */ int lock_extent(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask) { int err; @@ -982,6 +986,13 @@ int set_range_writeback(struct extent_io_tree *tree, u64 start, u64 end) } 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) { @@ -1017,6 +1028,10 @@ out: } 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) { @@ -1046,8 +1061,14 @@ out: } 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_lock_delalloc_range(struct extent_io_tree *tree, + u64 *start, u64 *end, u64 max_bytes) { struct rb_node *node; struct extent_state *state; @@ -1130,6 +1151,11 @@ out: return found; } +/* + * 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) @@ -1245,6 +1271,10 @@ int unlock_range(struct extent_io_tree *tree, u64 start, u64 end) } EXPORT_SYMBOL(unlock_range); +/* + * 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 set_state_private(struct extent_io_tree *tree, u64 start, u64 private) { struct rb_node *node; -- cgit v1.2.2 From c8b978188c9a0fd3d535c13debd19d522b726f1f Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 29 Oct 2008 14:49:59 -0400 Subject: Btrfs: Add zlib compression support This is a large change for adding compression on reading and writing, both for inline and regular extents. It does some fairly large surgery to the writeback paths. Compression is off by default and enabled by mount -o compress. Even when the -o compress mount option is not used, it is possible to read compressed extents off the disk. If compression for a given set of pages fails to make them smaller, the file is flagged to avoid future compression attempts later. * While finding delalloc extents, the pages are locked before being sent down to the delalloc handler. This allows the delalloc handler to do complex things such as cleaning the pages, marking them writeback and starting IO on their behalf. * Inline extents are inserted at delalloc time now. This allows us to compress the data before inserting the inline extent, and it allows us to insert an inline extent that spans multiple pages. * All of the in-memory extent representations (extent_map.c, ordered-data.c etc) are changed to record both an in-memory size and an on disk size, as well as a flag for compression. From a disk format point of view, the extent pointers in the file are changed to record the on disk size of a given extent and some encoding flags. Space in the disk format is allocated for compression encoding, as well as encryption and a generic 'other' field. Neither the encryption or the 'other' field are currently used. In order to limit the amount of data read for a single random read in the file, the size of a compressed extent is limited to 128k. This is a software only limit, the disk format supports u64 sized compressed extents. In order to limit the ram consumed while processing extents, the uncompressed size of a compressed extent is limited to 256k. This is a software only limit and will be subject to tuning later. Checksumming is still done on compressed extents, and it is done on the uncompressed version of the data. This way additional encodings can be layered on without having to figure out which encoding to checksum. Compression happens at delalloc time, which is basically singled threaded because it is usually done by a single pdflush thread. This makes it tricky to spread the compression load across all the cpus on the box. We'll have to look at parallel pdflush walks of dirty inodes at a later time. Decompression is hooked into readpages and it does spread across CPUs nicely. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 411 ++++++++++++++++++++++++++++++++++++++++----------- 1 file changed, 323 insertions(+), 88 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 563b2d12f4f2..314041fdfa43 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -30,6 +30,7 @@ static struct kmem_cache *extent_buffer_cache; static LIST_HEAD(buffers); static LIST_HEAD(states); +#define LEAK_DEBUG 1 #ifdef LEAK_DEBUG static spinlock_t leak_lock = SPIN_LOCK_UNLOCKED; #endif @@ -1067,8 +1068,8 @@ EXPORT_SYMBOL(find_first_extent_bit_state); * * 1 is returned if we find something, 0 if nothing was in the tree */ -static noinline u64 find_lock_delalloc_range(struct extent_io_tree *tree, - u64 *start, u64 *end, u64 max_bytes) +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; @@ -1077,11 +1078,11 @@ static noinline u64 find_lock_delalloc_range(struct extent_io_tree *tree, u64 total_bytes = 0; spin_lock_irq(&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) @@ -1100,40 +1101,6 @@ search_again: *end = state->end; goto out; } - if (!found && !(state->state & EXTENT_BOUNDARY)) { - 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->end + 1 != state->start) || - !(prev_state->state & EXTENT_DELALLOC)) - break; - if ((cur_start - prev_state->start) * 2 > - max_bytes) - 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++; @@ -1151,6 +1118,208 @@ out: 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(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(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]); + page_cache_release(pages[i]); + } + pages_locked += ret; + 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; + 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) { + *start = delalloc_start; + *end = delalloc_end; + return found; + } + + /* + * 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) & + ~((u64)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 + */ + 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(tree, delalloc_start, delalloc_end, GFP_NOFS); + + /* then test to make sure it is all still delalloc */ + ret = test_range_bit(tree, delalloc_start, delalloc_end, + EXTENT_DELALLOC, 1); + if (!ret) { + unlock_extent(tree, delalloc_start, delalloc_end, GFP_NOFS); + __unlock_for_delalloc(inode, locked_page, + delalloc_start, delalloc_end); + cond_resched(); + goto again; + } + *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, + int clear_dirty, int set_writeback, + int end_writeback) +{ + 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 = EXTENT_LOCKED | EXTENT_DELALLOC; + + if (clear_dirty) + clear_bits |= EXTENT_DIRTY; + + clear_extent_bit(tree, start, end, clear_bits, 1, 0, GFP_NOFS); + + while(nr_pages > 0) { + ret = find_get_pages_contig(inode->i_mapping, index, + min(nr_pages, ARRAY_SIZE(pages)), pages); + for (i = 0; i < ret; i++) { + if (pages[i] == locked_page) { + page_cache_release(pages[i]); + continue; + } + if (clear_dirty) + clear_page_dirty_for_io(pages[i]); + if (set_writeback) + set_page_writeback(pages[i]); + if (end_writeback) + end_page_writeback(pages[i]); + unlock_page(pages[i]); + page_cache_release(pages[i]); + } + nr_pages -= ret; + index += ret; + cond_resched(); + } + return 0; +} +EXPORT_SYMBOL(extent_clear_unlock_delalloc); + /* * 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 @@ -1631,38 +1800,26 @@ 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 = 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)) @@ -1678,39 +1835,56 @@ 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(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); + 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); if (!bio) { printk("failed to allocate bio nr %d\n", nr); } - - 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) { *bio_ret = bio; } else { - ret = submit_one_bio(rw, bio, mirror_num); + ret = submit_one_bio(rw, bio, mirror_num, bio_flags); } return ret; @@ -1738,7 +1912,8 @@ 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; @@ -1756,13 +1931,27 @@ 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; @@ -1793,10 +1982,19 @@ 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; free_extent_map(em); @@ -1845,10 +2043,13 @@ printk("2bad mapping end %Lu cur %Lu\n", end, cur); unsigned long pnr = (last_byte >> PAGE_CACHE_SHIFT) + 1; pnr -= page->index; ret = submit_extent_page(READ, tree, page, - sector, iosize, page_offset, + sector, disk_io_size, page_offset, bdev, bio, pnr, - end_bio_extent_readpage, mirror_num); + end_bio_extent_readpage, mirror_num, + *bio_flags, + this_bio_flag); nr++; + *bio_flags = this_bio_flag; } if (ret) SetPageError(page); @@ -1867,11 +2068,13 @@ 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); @@ -1909,6 +2112,8 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, unsigned long end_index = i_size >> PAGE_CACHE_SHIFT; u64 nr_delalloc; u64 delalloc_end; + int page_started; + int compressed; WARN_ON(!PageLocked(page)); pg_offset = i_size & (PAGE_CACHE_SIZE - 1); @@ -1934,27 +2139,33 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, delalloc_start = start; delalloc_end = 0; + page_started = 0; while(delalloc_end < page_end) { - nr_delalloc = find_lock_delalloc_range(tree, &delalloc_start, + nr_delalloc = find_lock_delalloc_range(inode, tree, + page, + &delalloc_start, &delalloc_end, 128 * 1024 * 1024); if (nr_delalloc == 0) { 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); + tree->ops->fill_delalloc(inode, page, delalloc_start, + delalloc_end, &page_started); delalloc_start = delalloc_end + 1; } + + /* did the fill delalloc function already unlock and start the IO? */ + if (page_started) { + return 0; + } + 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); @@ -2006,10 +2217,15 @@ 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); @@ -2017,16 +2233,28 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, 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; + 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)) { @@ -2034,6 +2262,7 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, 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, @@ -2057,7 +2286,8 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, ret = submit_extent_page(WRITE, tree, page, sector, iosize, pg_offset, bdev, &epd->bio, max_nr, - end_bio_extent_writepage, 0); + end_bio_extent_writepage, + 0, 0, 0); if (ret) SetPageError(page); } @@ -2226,7 +2456,7 @@ int extent_write_full_page(struct extent_io_tree *tree, struct page *page, extent_write_cache_pages(tree, mapping, &wbc_writepages, __extent_writepage, &epd); if (epd.bio) { - submit_one_bio(WRITE, epd.bio, 0); + submit_one_bio(WRITE, epd.bio, 0, 0); } return ret; } @@ -2248,7 +2478,7 @@ int extent_writepages(struct extent_io_tree *tree, ret = extent_write_cache_pages(tree, mapping, wbc, __extent_writepage, &epd); if (epd.bio) { - submit_one_bio(WRITE, epd.bio, 0); + submit_one_bio(WRITE, epd.bio, 0, 0); } return ret; } @@ -2262,6 +2492,7 @@ 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++) { @@ -2281,7 +2512,7 @@ int extent_readpages(struct extent_io_tree *tree, if (!pagevec_add(&pvec, page)) __pagevec_lru_add(&pvec); __extent_read_full_page(tree, page, get_extent, - &bio, 0); + &bio, 0, &bio_flags); } page_cache_release(page); } @@ -2289,7 +2520,7 @@ int extent_readpages(struct extent_io_tree *tree, __pagevec_lru_add(&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); @@ -2414,7 +2645,8 @@ int extent_prepare_write(struct extent_io_tree *tree, 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 { @@ -2495,7 +2727,9 @@ int try_release_extent_mapping(struct extent_map_tree *map, } if (!test_range_bit(tree, em->start, extent_map_end(em) - 1, - EXTENT_LOCKED, 0)) { + EXTENT_LOCKED | EXTENT_WRITEBACK | + EXTENT_ORDERED, + 0)) { remove_extent_mapping(map, em); /* once for the rb tree */ free_extent_map(em); @@ -2923,6 +3157,7 @@ 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) return 0; @@ -2973,7 +3208,7 @@ int read_extent_buffer_pages(struct extent_io_tree *tree, ClearPageError(page); err = __extent_read_full_page(tree, page, get_extent, &bio, - mirror_num); + mirror_num, &bio_flags); if (err) { ret = err; printk("err %d from __extent_read_full_page\n", ret); @@ -2984,7 +3219,7 @@ int read_extent_buffer_pages(struct extent_io_tree *tree, } if (bio) - submit_one_bio(READ, bio, mirror_num); + submit_one_bio(READ, bio, mirror_num, bio_flags); if (ret || !wait) { if (ret) -- cgit v1.2.2 From 2517920135b0d29e70453e5b03d70d7b94207df3 Mon Sep 17 00:00:00 2001 From: Josef Bacik Date: Wed, 29 Oct 2008 14:49:05 -0400 Subject: Btrfs: nuke fs wide allocation mutex V2 This patch removes the giant fs_info->alloc_mutex and replaces it with a bunch of little locks. There is now a pinned_mutex, which is used when messing with the pinned_extents extent io tree, and the extent_ins_mutex which is used with the pending_del and extent_ins extent io trees. The locking for the extent tree stuff was inspired by a patch that Yan Zheng wrote to fix a race condition, I cleaned it up some and changed the locking around a little bit, but the idea remains the same. Basically instead of holding the extent_ins_mutex throughout the processing of an extent on the extent_ins or pending_del trees, we just hold it while we're searching and when we clear the bits on those trees, and lock the extent for the duration of the operations on the extent. Also to keep from getting hung up waiting to lock an extent, I've added a try_lock_extent so if we cannot lock the extent, move on to the next one in the tree and we'll come back to that one. I have tested this heavily and it does not appear to break anything. This has to be applied on top of my find_free_extent redo patch. I tested this patch on top of Yan's space reblancing code and it worked fine. The only thing that has changed since the last version is I pulled out all my debugging stuff, apparently I forgot to run guilt refresh before I sent the last patch out. Thank you, Signed-off-by: Josef Bacik --- fs/btrfs/extent_io.c | 14 ++++++++++++++ 1 file changed, 14 insertions(+) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 314041fdfa43..7503bd46819b 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -938,6 +938,20 @@ int lock_extent(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask) } EXPORT_SYMBOL(lock_extent); +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, 1, + &failed_start, mask); + if (err == -EEXIST) + return 0; + return 1; +} +EXPORT_SYMBOL(try_lock_extent); + int unlock_extent(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask) { -- cgit v1.2.2 From 6643558db29006825dbb10012b3f8890aca4bcd5 Mon Sep 17 00:00:00 2001 From: Yan Zheng Date: Thu, 30 Oct 2008 14:19:50 -0400 Subject: Btrfs: Fix bookend extent race v2 When dropping middle part of an extent, btrfs_drop_extents truncates the extent at first, then inserts a bookend extent. Since truncation and insertion can't be done atomically, there is a small period that the bookend extent isn't in the tree. This causes problem for functions that search the tree for file extent item. The way to fix this is lock the range of the bookend extent before truncation. Signed-off-by: Yan Zheng --- fs/btrfs/extent_io.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 7503bd46819b..65a0583027e9 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -946,8 +946,12 @@ int try_lock_extent(struct extent_io_tree *tree, u64 start, u64 end, err = set_extent_bit(tree, start, end, EXTENT_LOCKED, 1, &failed_start, mask); - if (err == -EEXIST) + if (err == -EEXIST) { + if (failed_start > start) + clear_extent_bit(tree, start, failed_start - 1, + EXTENT_LOCKED, 1, 0, mask); return 0; + } return 1; } EXPORT_SYMBOL(try_lock_extent); -- cgit v1.2.2 From d899e05215178fed903ad0e7fc1cb4d8e0cc0a88 Mon Sep 17 00:00:00 2001 From: Yan Zheng Date: Thu, 30 Oct 2008 14:25:28 -0400 Subject: Btrfs: Add fallocate support v2 This patch updates btrfs-progs for fallocate support. fallocate is a little different in Btrfs because we need to tell the COW system that a given preallocated extent doesn't need to be cow'd as long as there are no snapshots of it. This leverages the -o nodatacow checks. Signed-off-by: Yan Zheng --- fs/btrfs/extent_io.c | 12 +++++++++--- 1 file changed, 9 insertions(+), 3 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 65a0583027e9..eb3c12e7beaf 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -2015,6 +2015,8 @@ printk("2bad mapping end %Lu cur %Lu\n", end, cur); } 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; @@ -2769,14 +2771,18 @@ 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; -- cgit v1.2.2 From 70b99e6959a4c28ae1b314985eca731f3db72f1d Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Fri, 31 Oct 2008 12:46:39 -0400 Subject: Btrfs: Compression corner fixes Make sure we keep page->mapping NULL on the pages we're getting via alloc_page. It gets set so a few of the callbacks can do the right thing, but in general these pages don't have a mapping. Don't try to truncate compressed inline items in btrfs_drop_extents. The whole compressed item must be preserved. Don't try to create multipage inline compressed items. When we try to overwrite just the first page of the file, we would have to read in and recow all the pages after it in the same compressed inline items. For now, only create single page inline items. Make sure we lock pages in the correct order during delalloc. The search into the state tree for delalloc bytes can return bytes before the page we already have locked. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index eb3c12e7beaf..9b37ce6e5168 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -1242,12 +1242,21 @@ again: delalloc_end = 0; found = find_delalloc_range(tree, &delalloc_start, &delalloc_end, max_bytes); - if (!found) { + 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. -- cgit v1.2.2 From 771ed689d2cd53439e28e095bc38fbe40a71429e Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 6 Nov 2008 22:02:51 -0500 Subject: Btrfs: Optimize compressed writeback and reads When reading compressed extents, try to put pages into the page cache for any pages covered by the compressed extent that readpages didn't already preload. Add an async work queue to handle transformations at delayed allocation processing time. Right now this is just compression. The workflow is: 1) Find offsets in the file marked for delayed allocation 2) Lock the pages 3) Lock the state bits 4) Call the async delalloc code The async delalloc code clears the state lock bits and delalloc bits. It is important this happens before the range goes into the work queue because otherwise it might deadlock with other work queue items that try to lock those extent bits. The file pages are compressed, and if the compression doesn't work the pages are written back directly. An ordered work queue is used to make sure the inodes are written in the same order that pdflush or writepages sent them down. This changes extent_write_cache_pages to let the writepage function update the wbc nr_written count. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 140 ++++++++++++++++++++++++++++++++++++++++++--------- 1 file changed, 117 insertions(+), 23 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 9b37ce6e5168..bbe3bcfcf4ae 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -47,6 +47,11 @@ 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 + */ + int extent_locked; }; int __init extent_io_init(void) @@ -1198,11 +1203,18 @@ static noinline int lock_delalloc_pages(struct inode *inode, * the caller is taking responsibility for * locked_page */ - if (pages[i] != locked_page) + if (pages[i] != locked_page) { lock_page(pages[i]); + if (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++; } - pages_locked += ret; nrpages -= ret; index += ret; cond_resched(); @@ -1262,8 +1274,7 @@ again: * if we're looping. */ if (delalloc_end + 1 - delalloc_start > max_bytes && loops) { - delalloc_end = (delalloc_start + PAGE_CACHE_SIZE - 1) & - ~((u64)PAGE_CACHE_SIZE - 1); + 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, @@ -1306,7 +1317,10 @@ out_failed: int extent_clear_unlock_delalloc(struct inode *inode, struct extent_io_tree *tree, u64 start, u64 end, struct page *locked_page, - int clear_dirty, int set_writeback, + int unlock_pages, + int clear_unlock, + int clear_delalloc, int clear_dirty, + int set_writeback, int end_writeback) { int ret; @@ -1315,12 +1329,19 @@ int extent_clear_unlock_delalloc(struct inode *inode, unsigned long end_index = end >> PAGE_CACHE_SHIFT; unsigned long nr_pages = end_index - index + 1; int i; - int clear_bits = EXTENT_LOCKED | EXTENT_DELALLOC; + int clear_bits = 0; + if (clear_unlock) + clear_bits |= EXTENT_LOCKED; if (clear_dirty) clear_bits |= EXTENT_DIRTY; + if (clear_delalloc) + clear_bits |= EXTENT_DELALLOC; + clear_extent_bit(tree, start, end, clear_bits, 1, 0, GFP_NOFS); + if (!(unlock_pages || clear_dirty || set_writeback || end_writeback)) + return 0; while(nr_pages > 0) { ret = find_get_pages_contig(inode->i_mapping, index, @@ -1336,7 +1357,8 @@ int extent_clear_unlock_delalloc(struct inode *inode, set_page_writeback(pages[i]); if (end_writeback) end_page_writeback(pages[i]); - unlock_page(pages[i]); + if (unlock_pages) + unlock_page(pages[i]); page_cache_release(pages[i]); } nr_pages -= ret; @@ -1741,9 +1763,10 @@ static void end_bio_extent_readpage(struct bio *bio, int err) } } - if (uptodate) + if (uptodate) { set_extent_uptodate(tree, start, end, GFP_ATOMIC); + } unlock_extent(tree, start, end, GFP_ATOMIC); if (whole_page) { @@ -1925,6 +1948,7 @@ void set_page_extent_mapped(struct page *page) set_page_private(page, EXTENT_PAGE_PRIVATE); } } +EXPORT_SYMBOL(set_page_extent_mapped); void set_page_extent_head(struct page *page, unsigned long len) { @@ -2143,12 +2167,17 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, u64 delalloc_end; int page_started; int compressed; + unsigned long nr_written = 0; WARN_ON(!PageLocked(page)); pg_offset = i_size & (PAGE_CACHE_SIZE - 1); if (page->index > end_index || (page->index == end_index && !pg_offset)) { - page->mapping->a_ops->invalidatepage(page, 0); + if (epd->extent_locked) { + if (tree->ops && tree->ops->writepage_end_io_hook) + tree->ops->writepage_end_io_hook(page, start, + page_end, NULL, 1); + } unlock_page(page); return 0; } @@ -2169,27 +2198,33 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, delalloc_start = start; delalloc_end = 0; page_started = 0; - while(delalloc_end < page_end) { - nr_delalloc = find_lock_delalloc_range(inode, tree, + if (!epd->extent_locked) { + 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_start = delalloc_end + 1; - continue; } - tree->ops->fill_delalloc(inode, page, delalloc_start, - delalloc_end, &page_started); - delalloc_start = delalloc_end + 1; - } - /* did the fill delalloc function already unlock and start the IO? */ - if (page_started) { - return 0; + /* did the fill delalloc function already unlock and start + * the IO? + */ + if (page_started) { + ret = 0; + goto update_nr_written; + } } - lock_extent(tree, start, page_end, GFP_NOFS); + unlock_start = start; if (tree->ops && tree->ops->writepage_start_hook) { @@ -2199,10 +2234,13 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, unlock_extent(tree, start, page_end, GFP_NOFS); redirty_page_for_writepage(wbc, page); unlock_page(page); - return 0; + ret = 0; + goto update_nr_written; } } + nr_written++; + end = page_end; if (test_range_bit(tree, start, page_end, EXTENT_DELALLOC, 0)) { printk("found delalloc bits after lock_extent\n"); @@ -2333,6 +2371,12 @@ done: if (unlock_start <= page_end) unlock_extent(tree, unlock_start, page_end, GFP_NOFS); unlock_page(page); + +update_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; return 0; } @@ -2431,7 +2475,7 @@ retry: unlock_page(page); ret = 0; } - if (ret || (--(wbc->nr_to_write) <= 0)) + if (ret || wbc->nr_to_write <= 0) done = 1; if (wbc->nonblocking && bdi_write_congested(bdi)) { wbc->encountered_congestion = 1; @@ -2452,6 +2496,8 @@ retry: } if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0)) mapping->writeback_index = index; + if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX) + range_whole = 1; if (wbc->range_cont) wbc->range_start = index << PAGE_CACHE_SHIFT; @@ -2469,6 +2515,7 @@ int extent_write_full_page(struct extent_io_tree *tree, struct page *page, .bio = NULL, .tree = tree, .get_extent = get_extent, + .extent_locked = 0, }; struct writeback_control wbc_writepages = { .bdi = wbc->bdi, @@ -2491,6 +2538,52 @@ int extent_write_full_page(struct extent_io_tree *tree, struct page *page, } 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, + }; + 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; + } + + if (epd.bio) + submit_one_bio(WRITE, epd.bio, 0, 0); + return ret; +} +EXPORT_SYMBOL(extent_write_locked_range); + int extent_writepages(struct extent_io_tree *tree, struct address_space *mapping, @@ -2502,6 +2595,7 @@ int extent_writepages(struct extent_io_tree *tree, .bio = NULL, .tree = tree, .get_extent = get_extent, + .extent_locked = 0, }; ret = extent_write_cache_pages(tree, mapping, wbc, -- cgit v1.2.2 From 3b7885bf96e03271a9fff41124c38ed7176616e2 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 6 Nov 2008 21:48:27 -0500 Subject: Btrfs: enforce metadata allocation clustering The allocator uses the last allocation as a starting point for metadata allocations, and tries to allocate in clusters of at least 256k. If the search for a free block fails to find the expected block, this patch forces a new cluster to be found in the free list. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 7 ------- 1 file changed, 7 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index bbe3bcfcf4ae..ad75a9cc3604 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -2494,13 +2494,6 @@ retry: index = 0; goto retry; } - if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0)) - mapping->writeback_index = index; - if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX) - range_whole = 1; - - if (wbc->range_cont) - wbc->range_start = index << PAGE_CACHE_SHIFT; return ret; } EXPORT_SYMBOL(extent_write_cache_pages); -- cgit v1.2.2 From f2b1c41cf94d7f839fe9ede5f3ead92698a93fb3 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 10 Nov 2008 07:31:30 -0500 Subject: Btrfs: Make sure pages are dirty before doing delalloc for them This adds a PageDirty check to the writeback path that locks pages for delalloc. If a page wasn't dirty at this point, it is in the process of being truncated away. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index ad75a9cc3604..69ea09659d47 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -1205,7 +1205,8 @@ static noinline int lock_delalloc_pages(struct inode *inode, */ if (pages[i] != locked_page) { lock_page(pages[i]); - if (pages[i]->mapping != inode->i_mapping) { + if (!PageDirty(pages[i]) || + pages[i]->mapping != inode->i_mapping) { ret = -EAGAIN; unlock_page(pages[i]); page_cache_release(pages[i]); -- cgit v1.2.2 From 39be25cd89450940b0e5f8a6aad71d1ec99b17bf Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 10 Nov 2008 11:50:50 -0500 Subject: Btrfs: Use invalidatepage when writepage finds a page outside of i_size With all the recent fixes to the delalloc locking, it is now safe again to use invalidatepage inside the writepage code for pages outside of i_size. This used to deadlock against some of the code to write locked ranges of pages, but all of that has been fixed. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 6 +----- 1 file changed, 1 insertion(+), 5 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 69ea09659d47..af2d9a9300a7 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -2174,11 +2174,7 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, pg_offset = i_size & (PAGE_CACHE_SIZE - 1); if (page->index > end_index || (page->index == end_index && !pg_offset)) { - if (epd->extent_locked) { - if (tree->ops && tree->ops->writepage_end_io_hook) - tree->ops->writepage_end_io_hook(page, start, - page_end, NULL, 1); - } + page->mapping->a_ops->invalidatepage(page, 0); unlock_page(page); return 0; } -- cgit v1.2.2 From b47eda8690a10f4fc01eb7b795078fa3fa57149f Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 10 Nov 2008 12:34:40 -0500 Subject: Btrfs: Turn off extent state leak debugging The extent_io.c code has a #define to find and cleanup extent state leaks on module unmount. This adds a very highly contended spinlock to a hot path for most FS operations. Turn it off by default. A later changeset will add a .config option for it. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index af2d9a9300a7..5cc0082379ce 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -30,7 +30,7 @@ static struct kmem_cache *extent_buffer_cache; static LIST_HEAD(buffers); static LIST_HEAD(states); -#define LEAK_DEBUG 1 +#define LEAK_DEBUG 0 #ifdef LEAK_DEBUG static spinlock_t leak_lock = SPIN_LOCK_UNLOCKED; #endif -- cgit v1.2.2 From 5b050f04c8ce911c5b6831305a24d70eab95e732 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 11 Nov 2008 09:34:41 -0500 Subject: Btrfs: Fix compile warnings on 32 bit machines Simple casting here and there to fix things up. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 11 +++++++---- 1 file changed, 7 insertions(+), 4 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 5cc0082379ce..54d013c3bb88 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -1157,7 +1157,8 @@ static noinline int __unlock_for_delalloc(struct inode *inode, while(nr_pages > 0) { ret = find_get_pages_contig(inode->i_mapping, index, - min(nr_pages, ARRAY_SIZE(pages)), pages); + 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]); @@ -1192,7 +1193,8 @@ static noinline int lock_delalloc_pages(struct inode *inode, nrpages = end_index - index + 1; while(nrpages > 0) { ret = find_get_pages_contig(inode->i_mapping, index, - min(nrpages, ARRAY_SIZE(pages)), pages); + min_t(unsigned long, + nrpages, ARRAY_SIZE(pages)), pages); if (ret == 0) { ret = -EAGAIN; goto done; @@ -1346,7 +1348,8 @@ int extent_clear_unlock_delalloc(struct inode *inode, while(nr_pages > 0) { ret = find_get_pages_contig(inode->i_mapping, index, - min(nr_pages, ARRAY_SIZE(pages)), pages); + min_t(unsigned long, + nr_pages, ARRAY_SIZE(pages)), pages); for (i = 0; i < ret; i++) { if (pages[i] == locked_page) { page_cache_release(pages[i]); @@ -1896,7 +1899,7 @@ static int submit_extent_page(int rw, struct extent_io_tree *tree, 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(size, PAGE_CACHE_SIZE); + size_t page_size = min_t(size_t, size, PAGE_CACHE_SIZE); if (bio_ret && *bio_ret) { bio = *bio_ret; -- cgit v1.2.2 From d2c3f4f695edac4d75c1b3eb01a1d16072de63bb Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 19 Nov 2008 12:44:22 -0500 Subject: Btrfs: Avoid writeback stalls While building large bios in writepages, btrfs may end up waiting for other page writeback to finish if WB_SYNC_ALL is used. While it is waiting, the bio it is building has a number of pages with the writeback bit set and they aren't getting to the disk any time soon. This lowers the latencies of writeback in general by sending down the bio being built before waiting for other pages. The bio submission code tries to limit the total number of async bios in flight by waiting when we're over a certain number of async bios. But, the waits are happening while writepages is building bios, and this can easily lead to stalls and other problems for people calling wait_on_page_writeback. The current fix is to let the congestion tests take care of waiting. sync() and others make sure to drain the current async requests to make sure that everything that was pending when the sync was started really get to disk. The code would drain pending requests both before and after submitting a new request. But, if one of the requests is waiting for page writeback to finish, the draining waits might block that page writeback. This changes the draining code to only wait after submitting the bio being processed. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 24 ++++++++++++++++++++---- 1 file changed, 20 insertions(+), 4 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 54d013c3bb88..a0f3804efe4f 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -2398,7 +2398,8 @@ update_nr_written: int extent_write_cache_pages(struct extent_io_tree *tree, struct address_space *mapping, struct writeback_control *wbc, - writepage_t writepage, void *data) + writepage_t writepage, void *data, + void (*flush_fn)(void *)) { struct backing_dev_info *bdi = mapping->backing_dev_info; int ret = 0; @@ -2460,8 +2461,10 @@ retry: continue; } - if (wbc->sync_mode != WB_SYNC_NONE) + if (wbc->sync_mode != WB_SYNC_NONE) { + flush_fn(data); wait_on_page_writeback(page); + } if (PageWriteback(page) || !clear_page_dirty_for_io(page)) { @@ -2498,6 +2501,15 @@ retry: } EXPORT_SYMBOL(extent_write_cache_pages); +static noinline void flush_write_bio(void *data) +{ + struct extent_page_data *epd = data; + if (epd->bio) { + submit_one_bio(WRITE, epd->bio, 0, 0); + epd->bio = NULL; + } +} + int extent_write_full_page(struct extent_io_tree *tree, struct page *page, get_extent_t *get_extent, struct writeback_control *wbc) @@ -2523,7 +2535,7 @@ int extent_write_full_page(struct extent_io_tree *tree, struct page *page, ret = __extent_writepage(page, wbc, &epd); extent_write_cache_pages(tree, mapping, &wbc_writepages, - __extent_writepage, &epd); + __extent_writepage, &epd, flush_write_bio); if (epd.bio) { submit_one_bio(WRITE, epd.bio, 0, 0); } @@ -2592,7 +2604,8 @@ int extent_writepages(struct extent_io_tree *tree, }; ret = extent_write_cache_pages(tree, mapping, wbc, - __extent_writepage, &epd); + __extent_writepage, &epd, + flush_write_bio); if (epd.bio) { submit_one_bio(WRITE, epd.bio, 0, 0); } @@ -3087,6 +3100,9 @@ int clear_extent_buffer_dirty(struct extent_io_tree *tree, for (i = 0; i < num_pages; i++) { page = extent_buffer_page(eb, i); + if (!set && !PageDirty(page)) + continue; + lock_page(page); if (i == 0) set_page_extent_head(page, eb->len); -- cgit v1.2.2 From 15916de835a683bd8133a0d1ac0c982b795ab4ff Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 19 Nov 2008 21:17:22 -0500 Subject: Btrfs: Fixes for 2.6.28-rc API changes * open/close_bdev_excl -> open/close_bdev_exclusive * blkdev_issue_discard takes a GFP mask now * Fix blkdev_issue_discard usage now that it is enabled Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index a0f3804efe4f..3a65c10dce33 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -2639,14 +2639,14 @@ 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_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, bio_flags); -- cgit v1.2.2 From 0e6bd956ed238eb2f69386f251847fe3163532e1 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 20 Nov 2008 10:46:35 -0500 Subject: Btrfs: only flush down bios for writeback pages The btrfs write_cache_pages call has a flush function so that it submits the bio it has been building before it waits on any writeback pages. This adds a check so that flush only happens on writeback pages. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 3a65c10dce33..87dba8517935 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -2462,7 +2462,8 @@ retry: } if (wbc->sync_mode != WB_SYNC_NONE) { - flush_fn(data); + if (PageWriteback(page)) + flush_fn(data); wait_on_page_writeback(page); } -- cgit v1.2.2 From 641f5219f2ea0dfd1c024c91021ddc83f50bf1ce Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Tue, 2 Dec 2008 06:36:10 -0500 Subject: Btrfs: sparse lock verification annotations for wait_on_state Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 87dba8517935..d79ccdbfdd95 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -577,6 +577,8 @@ 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); -- cgit v1.2.2 From b2950863c61bc24cf0f63bc05947d9d50663c4c0 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Tue, 2 Dec 2008 09:54:17 -0500 Subject: Btrfs: make things static and include the right headers Shut up various sparse warnings about symbols that should be either static or have their declarations in scope. Signed-off-by: Christoph Hellwig --- fs/btrfs/extent_io.c | 35 +++++++++++++++-------------------- 1 file changed, 15 insertions(+), 20 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index d79ccdbfdd95..c3dfe2a0ec85 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -112,7 +112,7 @@ void extent_io_tree_init(struct extent_io_tree *tree, } EXPORT_SYMBOL(extent_io_tree_init); -struct extent_state *alloc_extent_state(gfp_t mask) +static struct extent_state *alloc_extent_state(gfp_t mask) { struct extent_state *state; #ifdef LEAK_DEBUG @@ -136,7 +136,7 @@ struct extent_state *alloc_extent_state(gfp_t mask) } 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; @@ -662,7 +662,7 @@ static void set_state_bits(struct extent_io_tree *tree, * [start, end] is inclusive * This takes the tree lock. */ -int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, int bits, +static int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, int bits, int exclusive, u64 *failed_start, gfp_t mask) { struct extent_state *state; @@ -879,12 +879,11 @@ int set_extent_new(struct extent_io_tree *tree, u64 start, u64 end, } 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); } -EXPORT_SYMBOL(clear_extent_new); int set_extent_uptodate(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask) @@ -894,27 +893,24 @@ int set_extent_uptodate(struct extent_io_tree *tree, u64 start, u64 end, } EXPORT_SYMBOL(set_extent_uptodate); -int clear_extent_uptodate(struct extent_io_tree *tree, u64 start, u64 end, +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); } -EXPORT_SYMBOL(clear_extent_uptodate); -int set_extent_writeback(struct extent_io_tree *tree, u64 start, u64 end, +static 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, +static 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) { @@ -994,7 +990,7 @@ 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; @@ -1010,7 +1006,6 @@ int set_range_writeback(struct extent_io_tree *tree, u64 start, u64 end) 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 @@ -1432,11 +1427,13 @@ out: spin_unlock_irq(&tree->lock); return total_bytes; } + +#if 0 /* * 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) +static 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; @@ -1473,12 +1470,11 @@ failed: } return err; } -EXPORT_SYMBOL(lock_range); /* * helper function to unlock both pages and extents in the tree. */ -int unlock_range(struct extent_io_tree *tree, u64 start, u64 end) +static 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; @@ -1493,7 +1489,7 @@ int unlock_range(struct extent_io_tree *tree, u64 start, u64 end) unlock_extent(tree, start, end, GFP_NOFS); return 0; } -EXPORT_SYMBOL(unlock_range); +#endif /* * set the private field for a given byte offset in the tree. If there isn't @@ -1956,7 +1952,7 @@ void set_page_extent_mapped(struct page *page) } EXPORT_SYMBOL(set_page_extent_mapped); -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); } @@ -2397,7 +2393,7 @@ update_nr_written: * WB_SYNC_ALL then we were called for data integrity and we must wait for * existing IO to complete. */ -int extent_write_cache_pages(struct extent_io_tree *tree, +static int extent_write_cache_pages(struct extent_io_tree *tree, struct address_space *mapping, struct writeback_control *wbc, writepage_t writepage, void *data, @@ -2502,7 +2498,6 @@ retry: } return ret; } -EXPORT_SYMBOL(extent_write_cache_pages); static noinline void flush_write_bio(void *data) { -- cgit v1.2.2 From d20f7043fa65659136c1a7c3c456eeeb5c6f431f Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 8 Dec 2008 16:58:54 -0500 Subject: Btrfs: move data checksumming into a dedicated tree Btrfs stores checksums for each data block. Until now, they have been stored in the subvolume trees, indexed by the inode that is referencing the data block. This means that when we read the inode, we've probably read in at least some checksums as well. But, this has a few problems: * The checksums are indexed by logical offset in the file. When compression is on, this means we have to do the expensive checksumming on the uncompressed data. It would be faster if we could checksum the compressed data instead. * If we implement encryption, we'll be checksumming the plain text and storing that on disk. This is significantly less secure. * For either compression or encryption, we have to get the plain text back before we can verify the checksum as correct. This makes the raid layer balancing and extent moving much more expensive. * It makes the front end caching code more complex, as we have touch the subvolume and inodes as we cache extents. * There is potentitally one copy of the checksum in each subvolume referencing an extent. The solution used here is to store the extent checksums in a dedicated tree. This allows us to index the checksums by phyiscal extent start and length. It means: * The checksum is against the data stored on disk, after any compression or encryption is done. * The checksum is stored in a central location, and can be verified without following back references, or reading inodes. This makes compression significantly faster by reducing the amount of data that needs to be checksummed. It will also allow much faster raid management code in general. The checksums are indexed by a key with a fixed objectid (a magic value in ctree.h) and offset set to the starting byte of the extent. This allows us to copy the checksum items into the fsync log tree directly (or any other tree), without having to invent a second format for them. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index c3dfe2a0ec85..7449ecf32c50 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -1732,6 +1732,9 @@ static void end_bio_extent_readpage(struct bio *bio, int err) int whole_page; int ret; + if (err) + uptodate = 0; + do { struct page *page = bvec->bv_page; tree = &BTRFS_I(page->mapping->host)->io_tree; @@ -1761,6 +1764,8 @@ static void end_bio_extent_readpage(struct bio *bio, int err) if (ret == 0) { uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags); + if (err) + uptodate = 0; continue; } } -- cgit v1.2.2 From 934d375bacf9ea8a37fbfff5f3cf1c093f324095 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 8 Dec 2008 16:43:10 -0500 Subject: Btrfs: Use map_private_extent_buffer during generic_bin_search It is possible that generic_bin_search will be called on a tree block that has not been locked. This happens because cache_block_block skips locking on the tree blocks. Since the tree block isn't locked, we aren't allowed to change the extent_buffer->map_token field. Using map_private_extent_buffer avoids any changes to the internal extent buffer fields. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 1 + 1 file changed, 1 insertion(+) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 7449ecf32c50..607f5ff2791c 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -3471,6 +3471,7 @@ int map_extent_buffer(struct extent_buffer *eb, unsigned long start, unmap_extent_buffer(eb, eb->map_token, km); eb->map_token = NULL; save = 1; + WARN_ON(!mutex_is_locked(&eb->mutex)); } err = map_private_extent_buffer(eb, start, min_len, token, map, map_start, map_len, km); -- cgit v1.2.2 From cad321ad529400c6ab24c501a67c3be720a0744c Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 17 Dec 2008 14:51:42 -0500 Subject: Btrfs: shift all end_io work to thread pools bio_end_io for reads without checksumming on and btree writes were happening without using async thread pools. This means the extent_io.c code had to use spin_lock_irq and friends on the rb tree locks for extent state. There were some irq safe vs unsafe lock inversions between the delallock lock and the extent state locks. This patch gets rid of them by moving all end_io code into the thread pools. To avoid contention and deadlocks between the data end_io processing and the metadata end_io processing yet another thread pool is added to finish off metadata writes. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 51 ++++++++++++++++++++++++--------------------------- 1 file changed, 24 insertions(+), 27 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 607f5ff2791c..25ce2d18e5b4 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -477,7 +477,6 @@ int clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, struct extent_state *state; struct extent_state *prealloc = NULL; struct rb_node *node; - unsigned long flags; int err; int set = 0; @@ -488,7 +487,7 @@ again: return -ENOMEM; } - spin_lock_irqsave(&tree->lock, flags); + spin_lock(&tree->lock); /* * this search will find the extents that end after * our range starts @@ -559,7 +558,7 @@ again: goto search_again; out: - spin_unlock_irqrestore(&tree->lock, flags); + spin_unlock(&tree->lock); if (prealloc) free_extent_state(prealloc); @@ -568,7 +567,7 @@ 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; @@ -582,9 +581,9 @@ static int wait_on_state(struct extent_io_tree *tree, { 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; } @@ -599,7 +598,7 @@ 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) { /* @@ -628,13 +627,13 @@ 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); @@ -668,7 +667,6 @@ static int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, int b struct extent_state *state; struct extent_state *prealloc = NULL; struct rb_node *node; - unsigned long flags; int err = 0; int set; u64 last_start; @@ -680,7 +678,7 @@ again: return -ENOMEM; } - spin_lock_irqsave(&tree->lock, flags); + spin_lock(&tree->lock); /* * this search will find all the extents that end after * our range starts. @@ -800,7 +798,7 @@ again: goto search_again; out: - spin_unlock_irqrestore(&tree->lock, flags); + spin_unlock(&tree->lock); if (prealloc) free_extent_state(prealloc); @@ -809,7 +807,7 @@ 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; @@ -1021,7 +1019,7 @@ 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. @@ -1044,7 +1042,7 @@ 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); @@ -1097,7 +1095,7 @@ static noinline u64 find_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 @@ -1134,7 +1132,7 @@ static noinline u64 find_delalloc_range(struct extent_io_tree *tree, break; } out: - spin_unlock_irq(&tree->lock); + spin_unlock(&tree->lock); return found; } @@ -1391,7 +1389,7 @@ u64 count_range_bits(struct extent_io_tree *tree, 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; @@ -1424,7 +1422,7 @@ u64 count_range_bits(struct extent_io_tree *tree, break; } out: - spin_unlock_irq(&tree->lock); + spin_unlock(&tree->lock); return total_bytes; } @@ -1501,7 +1499,7 @@ int set_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. @@ -1518,7 +1516,7 @@ 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; } @@ -1528,7 +1526,7 @@ 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. @@ -1545,7 +1543,7 @@ 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; } @@ -1561,9 +1559,8 @@ int test_range_bit(struct extent_io_tree *tree, u64 start, u64 end, struct extent_state *state = NULL; struct rb_node *node; int bitset = 0; - unsigned long flags; - spin_lock_irqsave(&tree->lock, flags); + spin_lock(&tree->lock); node = tree_search(tree, start); while (node && start <= end) { state = rb_entry(node, struct extent_state, rb_node); @@ -1594,7 +1591,7 @@ 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); -- cgit v1.2.2 From c584482b47f47b051cdc1d5236b99ad18f1b1cfb Mon Sep 17 00:00:00 2001 From: Liu Hui Date: Mon, 5 Jan 2009 15:49:55 -0500 Subject: Btrfs: Fix typo in clear_state_cb In clear_state_cb, we should check 'tree->ops->clear_bit_hook' instead of 'tree->ops->set_bit_hook'. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 25ce2d18e5b4..0bf7684207aa 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -340,7 +340,7 @@ 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) { + if (tree->ops && tree->ops->clear_bit_hook) { tree->ops->clear_bit_hook(tree->mapping->host, state->start, state->end, state->state, bits); } -- cgit v1.2.2 From d397712bcc6a759a560fd247e6053ecae091f958 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 5 Jan 2009 21:25:51 -0500 Subject: Btrfs: Fix checkpatch.pl warnings There were many, most are fixed now. struct-funcs.c generates some warnings but these are bogus. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 213 ++++++++++++++++++++++++--------------------------- 1 file changed, 99 insertions(+), 114 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 0bf7684207aa..39edb551dca6 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -32,7 +32,7 @@ static LIST_HEAD(states); #define LEAK_DEBUG 0 #ifdef LEAK_DEBUG -static spinlock_t leak_lock = SPIN_LOCK_UNLOCKED; +static DEFINE_SPINLOCK(leak_lock); #endif #define BUFFER_LRU_MAX 64 @@ -81,7 +81,11 @@ 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); @@ -89,7 +93,9 @@ 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); } @@ -158,11 +164,11 @@ 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); @@ -185,13 +191,13 @@ 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; - while(n) { + while (n) { entry = rb_entry(n, struct tree_entry, rb_node); prev = n; prev_entry = entry; @@ -200,14 +206,13 @@ 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 { + 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); } @@ -217,7 +222,7 @@ 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); } @@ -233,9 +238,8 @@ 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 (!ret) return prev; - } return ret; } @@ -243,11 +247,11 @@ 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 rb_node **p = &root->rb_node; + struct rb_node *parent = NULL; struct extent_buffer *eb; - while(*p) { + while (*p) { parent = *p; eb = rb_entry(parent, struct extent_buffer, rb_node); @@ -268,10 +272,10 @@ 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 rb_node *n = root->rb_node; struct extent_buffer *eb; - while(n) { + while (n) { eb = rb_entry(n, struct extent_buffer, rb_node); if (offset < eb->start) n = n->rb_left; @@ -363,7 +367,9 @@ static int insert_state(struct extent_io_tree *tree, struct rb_node *node; 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); } if (bits & EXTENT_DIRTY) @@ -376,7 +382,10 @@ static int insert_state(struct extent_io_tree *tree, 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; } @@ -412,7 +421,6 @@ static int split_state(struct extent_io_tree *tree, struct extent_state *orig, 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; } @@ -661,8 +669,9 @@ static void set_state_bits(struct extent_io_tree *tree, * [start, end] is inclusive * This takes the tree lock. */ -static 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, u64 *failed_start, + gfp_t mask) { struct extent_state *state; struct extent_state *prealloc = NULL; @@ -763,7 +772,7 @@ 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; @@ -891,8 +900,8 @@ int set_extent_uptodate(struct extent_io_tree *tree, u64 start, u64 end, } EXPORT_SYMBOL(set_extent_uptodate); -static 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); } @@ -904,8 +913,8 @@ static int set_extent_writeback(struct extent_io_tree *tree, u64 start, u64 end, 0, NULL, mask); } -static int clear_extent_writeback(struct extent_io_tree *tree, u64 start, u64 end, - gfp_t mask) +static 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); } @@ -1025,11 +1034,10 @@ int find_first_extent_bit(struct extent_io_tree *tree, u64 start, * 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; @@ -1062,15 +1070,14 @@ 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; @@ -1108,7 +1115,7 @@ static noinline u64 find_delalloc_range(struct extent_io_tree *tree, goto out; } - while(1) { + while (1) { state = rb_entry(node, struct extent_state, rb_node); if (found && (state->start != cur_start || (state->state & EXTENT_BOUNDARY))) { @@ -1150,7 +1157,7 @@ static noinline int __unlock_for_delalloc(struct inode *inode, if (index == locked_page->index && end_index == index) return 0; - while(nr_pages > 0) { + while (nr_pages > 0) { ret = find_get_pages_contig(inode->i_mapping, index, min_t(unsigned long, nr_pages, ARRAY_SIZE(pages)), pages); @@ -1186,7 +1193,7 @@ static noinline int lock_delalloc_pages(struct inode *inode, /* skip the page at the start index */ nrpages = end_index - index + 1; - while(nrpages > 0) { + while (nrpages > 0) { ret = find_get_pages_contig(inode->i_mapping, index, min_t(unsigned long, nrpages, ARRAY_SIZE(pages)), pages); @@ -1263,17 +1270,16 @@ again: * pages in order, so we can't process delalloc bytes before * locked_page */ - if (delalloc_start < *start) { + 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) { + 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); @@ -1341,7 +1347,7 @@ int extent_clear_unlock_delalloc(struct inode *inode, if (!(unlock_pages || clear_dirty || set_writeback || end_writeback)) return 0; - while(nr_pages > 0) { + while (nr_pages > 0) { ret = find_get_pages_contig(inode->i_mapping, index, min_t(unsigned long, nr_pages, ARRAY_SIZE(pages)), pages); @@ -1384,7 +1390,6 @@ 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; } @@ -1399,11 +1404,10 @@ 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; @@ -1927,19 +1931,15 @@ static int submit_extent_page(int rw, struct extent_io_tree *tree, 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); - } 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 { + else ret = submit_one_bio(rw, bio, mirror_num, bio_flags); - } return ret; } @@ -2028,13 +2028,7 @@ 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)) @@ -2199,7 +2193,7 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, delalloc_end = 0; page_started = 0; if (!epd->extent_locked) { - while(delalloc_end < page_end) { + while (delalloc_end < page_end) { nr_delalloc = find_lock_delalloc_range(inode, tree, page, &delalloc_start, @@ -2242,9 +2236,8 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, nr_written++; end = page_end; - if (test_range_bit(tree, start, page_end, EXTENT_DELALLOC, 0)) { - printk("found delalloc bits after lock_extent\n"); - } + if (test_range_bit(tree, start, page_end, EXTENT_DELALLOC, 0)) + printk(KERN_ERR "btrfs delalloc bits after lock_extent\n"); if (last_byte <= start) { clear_extent_dirty(tree, start, page_end, GFP_NOFS); @@ -2297,7 +2290,7 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, clear_extent_dirty(tree, cur, cur + iosize - 1, GFP_NOFS); - unlock_extent(tree, unlock_start, cur + iosize -1, + unlock_extent(tree, unlock_start, cur + iosize - 1, GFP_NOFS); /* @@ -2344,9 +2337,9 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, 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); } @@ -2430,8 +2423,8 @@ static int extent_write_cache_pages(struct extent_io_tree *tree, retry: while (!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; @@ -2536,9 +2529,8 @@ int extent_write_full_page(struct extent_io_tree *tree, struct page *page, extent_write_cache_pages(tree, mapping, &wbc_writepages, __extent_writepage, &epd, flush_write_bio); - if (epd.bio) { + if (epd.bio) submit_one_bio(WRITE, epd.bio, 0, 0); - } return ret; } EXPORT_SYMBOL(extent_write_full_page); @@ -2568,7 +2560,7 @@ int extent_write_locked_range(struct extent_io_tree *tree, struct inode *inode, .range_end = end + 1, }; - while(start <= end) { + 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); @@ -2606,9 +2598,8 @@ int extent_writepages(struct extent_io_tree *tree, ret = extent_write_cache_pages(tree, mapping, wbc, __extent_writepage, &epd, flush_write_bio); - if (epd.bio) { + if (epd.bio) submit_one_bio(WRITE, epd.bio, 0, 0); - } return ret; } EXPORT_SYMBOL(extent_writepages); @@ -2666,7 +2657,7 @@ int extent_invalidatepage(struct extent_io_tree *tree, u64 end = start + PAGE_CACHE_SIZE - 1; size_t blocksize = page->mapping->host->i_sb->s_blocksize; - start += (offset + blocksize -1) & ~(blocksize - 1); + start += (offset + blocksize - 1) & ~(blocksize - 1); if (start > end) return 0; @@ -2727,12 +2718,12 @@ 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; @@ -3170,7 +3161,7 @@ int set_extent_buffer_dirty(struct extent_io_tree *tree, } __set_page_dirty_nobuffers(extent_buffer_page(eb, i)); set_extent_dirty(tree, page_offset(page), - page_offset(page) + PAGE_CACHE_SIZE -1, + page_offset(page) + PAGE_CACHE_SIZE - 1, GFP_NOFS); unlock_page(page); } @@ -3235,7 +3226,7 @@ int extent_range_uptodate(struct extent_io_tree *tree, ret = test_range_bit(tree, start, end, EXTENT_UPTODATE, 1); 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); @@ -3321,16 +3312,12 @@ int read_extent_buffer_pages(struct extent_io_tree *tree, 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; - if (ret) { - printk("all up to date but ret is %d\n", ret); - } goto unlock_exit; } @@ -3345,10 +3332,8 @@ int read_extent_buffer_pages(struct extent_io_tree *tree, err = __extent_read_full_page(tree, page, get_extent, &bio, mirror_num, &bio_flags); - if (err) { + if (err) ret = err; - printk("err %d from __extent_read_full_page\n", ret); - } } else { unlock_page(page); } @@ -3357,26 +3342,23 @@ int read_extent_buffer_pages(struct extent_io_tree *tree, if (bio) submit_one_bio(READ, bio, mirror_num, bio_flags); - if (ret || !wait) { - if (ret) - printk("ret %d wait %d returning\n", 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)) { - printk("page not uptodate after wait_on_page_locked\n"); + if (!PageUptodate(page)) ret = -EIO; - } } + if (!ret) eb->flags |= EXTENT_UPTODATE; 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); @@ -3403,7 +3385,7 @@ 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)); @@ -3442,8 +3424,11 @@ 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); } @@ -3506,7 +3491,7 @@ 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)); @@ -3542,7 +3527,7 @@ 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)); @@ -3574,7 +3559,7 @@ 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)); @@ -3607,7 +3592,7 @@ 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)); @@ -3674,17 +3659,17 @@ 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) & @@ -3722,20 +3707,20 @@ 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; -- cgit v1.2.2 From 43b774ba1378dc5e952aadb829048bbe90ba33e4 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 5 Jan 2009 22:05:48 -0500 Subject: Btrfs: drop EXPORT symbols from extent_io.c They should stay out until this is turned into generic code. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 56 ---------------------------------------------------- 1 file changed, 56 deletions(-) (limited to 'fs/btrfs/extent_io.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 39edb551dca6..e086d407f1fa 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -116,7 +116,6 @@ void extent_io_tree_init(struct extent_io_tree *tree, spin_lock_init(&tree->buffer_lock); tree->mapping = mapping; } -EXPORT_SYMBOL(extent_io_tree_init); static struct extent_state *alloc_extent_state(gfp_t mask) { @@ -140,7 +139,6 @@ static struct extent_state *alloc_extent_state(gfp_t mask) init_waitqueue_head(&state->wq); return state; } -EXPORT_SYMBOL(alloc_extent_state); static void free_extent_state(struct extent_state *state) { @@ -159,7 +157,6 @@ static void free_extent_state(struct extent_state *state) 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) @@ -580,7 +577,6 @@ search_again: cond_resched(); goto again; } -EXPORT_SYMBOL(clear_extent_bit); static int wait_on_state(struct extent_io_tree *tree, struct extent_state *state) @@ -644,7 +640,6 @@ out: spin_unlock(&tree->lock); return 0; } -EXPORT_SYMBOL(wait_extent_bit); static void set_state_bits(struct extent_io_tree *tree, struct extent_state *state, @@ -821,7 +816,6 @@ search_again: 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, @@ -830,14 +824,12 @@ int set_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end, 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); } -EXPORT_SYMBOL(set_extent_ordered); int set_extent_bits(struct extent_io_tree *tree, u64 start, u64 end, int bits, gfp_t mask) @@ -845,14 +837,12 @@ int set_extent_bits(struct extent_io_tree *tree, u64 start, u64 end, return set_extent_bit(tree, start, end, bits, 0, 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); } -EXPORT_SYMBOL(clear_extent_bits); int set_extent_delalloc(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask) @@ -861,7 +851,6 @@ int set_extent_delalloc(struct extent_io_tree *tree, u64 start, u64 end, EXTENT_DELALLOC | EXTENT_DIRTY, 0, NULL, mask); } -EXPORT_SYMBOL(set_extent_delalloc); int clear_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask) @@ -869,14 +858,12 @@ int clear_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end, return clear_extent_bit(tree, start, end, EXTENT_DIRTY | EXTENT_DELALLOC, 0, 0, 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) @@ -884,7 +871,6 @@ int set_extent_new(struct extent_io_tree *tree, u64 start, u64 end, return set_extent_bit(tree, start, end, EXTENT_NEW, 0, NULL, mask); } -EXPORT_SYMBOL(set_extent_new); static int clear_extent_new(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask) @@ -898,7 +884,6 @@ int set_extent_uptodate(struct extent_io_tree *tree, u64 start, u64 end, return set_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, NULL, mask); } -EXPORT_SYMBOL(set_extent_uptodate); static int clear_extent_uptodate(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask) @@ -923,7 +908,6 @@ 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); /* * either insert or lock state struct between start and end use mask to tell @@ -946,7 +930,6 @@ int lock_extent(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask) } return err; } -EXPORT_SYMBOL(lock_extent); int try_lock_extent(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask) @@ -964,14 +947,12 @@ int try_lock_extent(struct extent_io_tree *tree, u64 start, u64 end, } return 1; } -EXPORT_SYMBOL(try_lock_extent); 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); } -EXPORT_SYMBOL(unlock_extent); /* * helper function to set pages and extents in the tree dirty @@ -992,7 +973,6 @@ int set_range_dirty(struct extent_io_tree *tree, u64 start, u64 end) 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 @@ -1053,7 +1033,6 @@ out: 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 @@ -1085,7 +1064,6 @@ struct extent_state *find_first_extent_bit_state(struct extent_io_tree *tree, out: return NULL; } -EXPORT_SYMBOL(find_first_extent_bit_state); /* * find a contiguous range of bytes in the file marked as delalloc, not @@ -1372,7 +1350,6 @@ int extent_clear_unlock_delalloc(struct inode *inode, } return 0; } -EXPORT_SYMBOL(extent_clear_unlock_delalloc); /* * count the number of bytes in the tree that have a given bit(s) @@ -1598,7 +1575,6 @@ int test_range_bit(struct extent_io_tree *tree, u64 start, u64 end, spin_unlock(&tree->lock); return bitset; } -EXPORT_SYMBOL(test_range_bit); /* * helper function to set a given page up to date if all the @@ -1952,7 +1928,6 @@ void set_page_extent_mapped(struct page *page) set_page_private(page, EXTENT_PAGE_PRIVATE); } } -EXPORT_SYMBOL(set_page_extent_mapped); static void set_page_extent_head(struct page *page, unsigned long len) { @@ -2128,7 +2103,6 @@ int extent_read_full_page(struct extent_io_tree *tree, struct page *page, submit_one_bio(READ, bio, 0, bio_flags); return ret; } -EXPORT_SYMBOL(extent_read_full_page); /* * the writepage semantics are similar to regular writepage. extent @@ -2533,7 +2507,6 @@ int extent_write_full_page(struct extent_io_tree *tree, struct page *page, submit_one_bio(WRITE, epd.bio, 0, 0); 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, @@ -2579,8 +2552,6 @@ int extent_write_locked_range(struct extent_io_tree *tree, struct inode *inode, submit_one_bio(WRITE, epd.bio, 0, 0); return ret; } -EXPORT_SYMBOL(extent_write_locked_range); - int extent_writepages(struct extent_io_tree *tree, struct address_space *mapping, @@ -2602,7 +2573,6 @@ int extent_writepages(struct extent_io_tree *tree, submit_one_bio(WRITE, epd.bio, 0, 0); return ret; } -EXPORT_SYMBOL(extent_writepages); int extent_readpages(struct extent_io_tree *tree, struct address_space *mapping, @@ -2643,7 +2613,6 @@ int extent_readpages(struct extent_io_tree *tree, submit_one_bio(READ, bio, 0, bio_flags); return 0; } -EXPORT_SYMBOL(extent_readpages); /* * basic invalidatepage code, this waits on any locked or writeback @@ -2668,7 +2637,6 @@ int extent_invalidatepage(struct extent_io_tree *tree, 1, 1, GFP_NOFS); return 0; } -EXPORT_SYMBOL(extent_invalidatepage); /* * simple commit_write call, set_range_dirty is used to mark both @@ -2689,7 +2657,6 @@ 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, @@ -2787,7 +2754,6 @@ 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 @@ -2813,7 +2779,6 @@ int try_release_extent_state(struct extent_map_tree *map, } return ret; } -EXPORT_SYMBOL(try_release_extent_state); /* * a helper for releasepage. As long as there are no locked extents @@ -2863,7 +2828,6 @@ 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) @@ -3042,7 +3006,6 @@ free_eb: __free_extent_buffer(eb); return exists; } -EXPORT_SYMBOL(alloc_extent_buffer); struct extent_buffer *find_extent_buffer(struct extent_io_tree *tree, u64 start, unsigned long len, @@ -3061,7 +3024,6 @@ struct extent_buffer *find_extent_buffer(struct extent_io_tree *tree, return eb; } -EXPORT_SYMBOL(find_extent_buffer); void free_extent_buffer(struct extent_buffer *eb) { @@ -3073,7 +3035,6 @@ void free_extent_buffer(struct 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) @@ -3128,7 +3089,6 @@ int clear_extent_buffer_dirty(struct extent_io_tree *tree, } return 0; } -EXPORT_SYMBOL(clear_extent_buffer_dirty); int wait_on_extent_buffer_writeback(struct extent_io_tree *tree, struct extent_buffer *eb) @@ -3136,7 +3096,6 @@ 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) @@ -3167,7 +3126,6 @@ int set_extent_buffer_dirty(struct extent_io_tree *tree, } return 0; } -EXPORT_SYMBOL(set_extent_buffer_dirty); int clear_extent_buffer_uptodate(struct extent_io_tree *tree, struct extent_buffer *eb) @@ -3212,7 +3170,6 @@ 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) @@ -3267,7 +3224,6 @@ 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, @@ -3366,7 +3322,6 @@ unlock_exit: } return ret; } -EXPORT_SYMBOL(read_extent_buffer_pages); void read_extent_buffer(struct extent_buffer *eb, void *dstv, unsigned long start, @@ -3399,7 +3354,6 @@ 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, @@ -3439,7 +3393,6 @@ int map_private_extent_buffer(struct extent_buffer *eb, unsigned long start, *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, @@ -3465,13 +3418,11 @@ 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, @@ -3509,7 +3460,6 @@ 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) @@ -3542,7 +3492,6 @@ 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) @@ -3573,7 +3522,6 @@ 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, @@ -3608,7 +3556,6 @@ 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, @@ -3692,7 +3639,6 @@ 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) @@ -3741,7 +3687,6 @@ 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) { @@ -3770,4 +3715,3 @@ out: spin_unlock(&tree->buffer_lock); return ret; } -EXPORT_SYMBOL(try_release_extent_buffer); -- cgit v1.2.2