#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,
- 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;
/* tells writepage not to lock the state bits for this range
* it still does the unlocking
*/
- int extent_locked;
+ unsigned int extent_locked:1;
+
+ /* tells the submit_bio code to use a WRITE_SYNC */
+ unsigned int sync_io:1;
};
int __init extent_io_init(void)
{
- extent_state_cache = btrfs_cache_create("extent_state",
- sizeof(struct extent_state), 0,
- NULL);
+ extent_state_cache = kmem_cache_create("extent_state",
+ sizeof(struct extent_state), 0,
+ SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
if (!extent_state_cache)
return -ENOMEM;
- extent_buffer_cache = btrfs_cache_create("extent_buffers",
- sizeof(struct extent_buffer), 0,
- NULL);
+ extent_buffer_cache = kmem_cache_create("extent_buffers",
+ sizeof(struct extent_buffer), 0,
+ SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
if (!extent_buffer_cache)
goto free_state_cache;
return 0;
}
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);
+ state->state |= bits;
node = tree_insert(&tree->state, end, &state->rb_node);
if (node) {
struct extent_state *found;
struct extent_state *state;
struct extent_state *prealloc = NULL;
struct rb_node *node;
+ u64 last_end;
int err;
int set = 0;
if (state->start > end)
goto out;
WARN_ON(state->end < start);
+ last_end = state->end;
/*
* | ---- desired range ---- |
if (err)
goto out;
if (state->end <= end) {
- start = state->end + 1;
set |= clear_state_bit(tree, state, bits,
wake, delete);
+ if (last_end == (u64)-1)
+ goto out;
+ start = last_end + 1;
} else {
start = state->start;
}
goto out;
}
- start = state->end + 1;
set |= clear_state_bit(tree, state, bits, wake, delete);
+ if (last_end == (u64)-1)
+ goto out;
+ start = last_end + 1;
goto search_again;
out:
}
/*
- * set some bits on a range in the tree. This may require allocations
- * or sleeping, so the gfp mask is used to indicate what is allowed.
+ * set some bits on a range in the tree. This may require allocations or
+ * sleeping, so the gfp mask is used to indicate what is allowed.
*
- * If 'exclusive' == 1, this will fail with -EEXIST if some part of the
- * range already has the desired bits set. The start of the existing
- * range is returned in failed_start in this case.
+ * If any of the exclusive bits are set, this will fail with -EEXIST if some
+ * part of the range already has the desired bits set. The start of the
+ * existing range is returned in failed_start in this case.
*
- * [start, end] is inclusive
- * This takes the tree lock.
+ * [start, end] is inclusive This takes the tree lock.
*/
+
static int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
- int bits, int exclusive, u64 *failed_start,
+ int bits, int exclusive_bits, u64 *failed_start,
gfp_t mask)
{
struct extent_state *state;
struct extent_state *prealloc = NULL;
struct rb_node *node;
int err = 0;
- int set;
u64 last_start;
u64 last_end;
again:
BUG_ON(err == -EEXIST);
goto out;
}
-
state = rb_entry(node, struct extent_state, rb_node);
+hit_next:
last_start = state->start;
last_end = state->end;
* Just lock what we found and keep going
*/
if (state->start == start && state->end <= end) {
- set = state->state & bits;
- if (set && exclusive) {
+ struct rb_node *next_node;
+ if (state->state & exclusive_bits) {
*failed_start = state->start;
err = -EEXIST;
goto out;
}
set_state_bits(tree, state, bits);
- start = state->end + 1;
merge_state(tree, state);
+ if (last_end == (u64)-1)
+ goto out;
+
+ start = last_end + 1;
+ if (start < end && prealloc && !need_resched()) {
+ next_node = rb_next(node);
+ if (next_node) {
+ state = rb_entry(next_node, struct extent_state,
+ rb_node);
+ if (state->start == start)
+ goto hit_next;
+ }
+ }
goto search_again;
}
* desired bit on it.
*/
if (state->start < start) {
- set = state->state & bits;
- if (exclusive && set) {
+ if (state->state & exclusive_bits) {
*failed_start = start;
err = -EEXIST;
goto out;
goto out;
if (state->end <= end) {
set_state_bits(tree, state, bits);
- start = state->end + 1;
merge_state(tree, state);
+ if (last_end == (u64)-1)
+ goto out;
+ start = last_end + 1;
} else {
start = state->start;
}
* on the first half
*/
if (state->start <= end && state->end > end) {
- set = state->state & bits;
- if (exclusive && set) {
+ if (state->state & exclusive_bits) {
*failed_start = start;
err = -EEXIST;
goto out;
gfp_t mask)
{
return set_extent_bit(tree, start, end,
- EXTENT_DELALLOC | EXTENT_DIRTY,
+ EXTENT_DELALLOC | EXTENT_DIRTY | EXTENT_UPTODATE,
0, NULL, mask);
}
return clear_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, 0, mask);
}
-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);
-}
-
-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);
-}
-
int wait_on_extent_writeback(struct extent_io_tree *tree, u64 start, u64 end)
{
return wait_extent_bit(tree, start, end, 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 lock_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
+ int bits, gfp_t mask)
{
int err;
u64 failed_start;
while (1) {
- err = set_extent_bit(tree, start, end, EXTENT_LOCKED, 1,
- &failed_start, mask);
+ err = set_extent_bit(tree, start, end, EXTENT_LOCKED | bits,
+ EXTENT_LOCKED, &failed_start, mask);
if (err == -EEXIST && (mask & __GFP_WAIT)) {
wait_extent_bit(tree, failed_start, end, EXTENT_LOCKED);
start = failed_start;
return err;
}
+int lock_extent(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask)
+{
+ return lock_extent_bits(tree, start, end, 0, mask);
+}
+
int try_lock_extent(struct extent_io_tree *tree, u64 start, u64 end,
gfp_t mask)
{
page_cache_release(page);
index++;
}
- set_extent_dirty(tree, start, end, GFP_NOFS);
return 0;
}
page_cache_release(page);
index++;
}
- set_extent_writeback(tree, start, end, GFP_NOFS);
return 0;
}
return total_bytes;
}
-#if 0
-/*
- * helper function to lock both pages and extents in the tree.
- * pages must be locked first.
- */
-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;
- 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;
-}
-
-/*
- * helper function to unlock both pages and extents in the tree.
- */
-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;
- 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;
-}
-#endif
-
/*
* set the private field for a given byte offset in the tree. If there isn't
* an extent_state there already, this does nothing.
static int check_page_writeback(struct extent_io_tree *tree,
struct page *page)
{
- u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
- u64 end = start + PAGE_CACHE_SIZE - 1;
- if (!test_range_bit(tree, start, end, EXTENT_WRITEBACK, 0))
- end_page_writeback(page);
+ end_page_writeback(page);
return 0;
}
}
if (!uptodate) {
- clear_extent_uptodate(tree, start, end, GFP_ATOMIC);
+ clear_extent_uptodate(tree, start, end, GFP_NOFS);
ClearPageUptodate(page);
SetPageError(page);
}
- clear_extent_writeback(tree, start, end, GFP_ATOMIC);
-
if (whole_page)
end_page_writeback(page);
else
return ret;
}
+static noinline void update_nr_written(struct page *page,
+ struct writeback_control *wbc,
+ unsigned long nr_written)
+{
+ wbc->nr_to_write -= nr_written;
+ if (wbc->range_cyclic || (wbc->nr_to_write > 0 &&
+ wbc->range_start == 0 && wbc->range_end == LLONG_MAX))
+ page->mapping->writeback_index = page->index + nr_written;
+}
+
/*
* the writepage semantics are similar to regular writepage. extent
* records are inserted to lock ranges in the tree, and as dirty areas
u64 delalloc_end;
int page_started;
int compressed;
+ int write_flags;
unsigned long nr_written = 0;
+ if (wbc->sync_mode == WB_SYNC_ALL)
+ write_flags = WRITE_SYNC_PLUG;
+ else
+ write_flags = WRITE;
+
WARN_ON(!PageLocked(page));
pg_offset = i_size & (PAGE_CACHE_SIZE - 1);
if (page->index > end_index ||
delalloc_end = 0;
page_started = 0;
if (!epd->extent_locked) {
+ u64 delalloc_to_write;
+ /*
+ * make sure the wbc mapping index is at least updated
+ * to this page.
+ */
+ update_nr_written(page, wbc, 0);
+
while (delalloc_end < page_end) {
nr_delalloc = find_lock_delalloc_range(inode, tree,
page,
tree->ops->fill_delalloc(inode, page, delalloc_start,
delalloc_end, &page_started,
&nr_written);
+ delalloc_to_write = (delalloc_end -
+ max_t(u64, page_offset(page),
+ delalloc_start) + 1) >>
+ PAGE_CACHE_SHIFT;
+ if (wbc->nr_to_write < delalloc_to_write) {
+ wbc->nr_to_write = min_t(long, 8192,
+ delalloc_to_write);
+ }
delalloc_start = delalloc_end + 1;
}
*/
if (page_started) {
ret = 0;
- goto update_nr_written;
+ /*
+ * we've unlocked the page, so we can't update
+ * the mapping's writeback index, just update
+ * nr_to_write.
+ */
+ wbc->nr_to_write -= nr_written;
+ goto done_unlocked;
}
}
lock_extent(tree, start, page_end, GFP_NOFS);
if (ret == -EAGAIN) {
unlock_extent(tree, start, page_end, GFP_NOFS);
redirty_page_for_writepage(wbc, page);
+ update_nr_written(page, wbc, nr_written);
unlock_page(page);
ret = 0;
- goto update_nr_written;
+ goto done_unlocked;
}
}
- nr_written++;
+ /*
+ * we don't want to touch the inode after unlocking the page,
+ * so we update the mapping writeback index now
+ */
+ update_nr_written(page, wbc, nr_written + 1);
end = page_end;
if (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);
- unlock_extent(tree, start, page_end, GFP_NOFS);
+ clear_extent_bit(tree, start, page_end,
+ EXTENT_LOCKED | EXTENT_DIRTY,
+ 1, 0, GFP_NOFS);
if (tree->ops && tree->ops->writepage_end_io_hook)
tree->ops->writepage_end_io_hook(page, start,
page_end, NULL, 1);
goto done;
}
- set_extent_uptodate(tree, start, page_end, GFP_NOFS);
blocksize = inode->i_sb->s_blocksize;
while (cur <= end) {
if (cur >= last_byte) {
- clear_extent_dirty(tree, cur, page_end, GFP_NOFS);
unlock_extent(tree, unlock_start, page_end, GFP_NOFS);
if (tree->ops && tree->ops->writepage_end_io_hook)
tree->ops->writepage_end_io_hook(page, cur,
*/
if (compressed || block_start == EXTENT_MAP_HOLE ||
block_start == EXTENT_MAP_INLINE) {
- clear_extent_dirty(tree, cur,
- cur + iosize - 1, GFP_NOFS);
-
unlock_extent(tree, unlock_start, cur + iosize - 1,
GFP_NOFS);
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);
(unsigned long long)end);
}
- ret = submit_extent_page(WRITE, tree, page, sector,
- iosize, pg_offset, bdev,
- &epd->bio, max_nr,
+ ret = submit_extent_page(write_flags, tree, page,
+ sector, iosize, pg_offset,
+ bdev, &epd->bio, max_nr,
end_bio_extent_writepage,
0, 0, 0);
if (ret)
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;
+done_unlocked:
+
return 0;
}
writepage_t writepage, void *data,
void (*flush_fn)(void *))
{
- struct backing_dev_info *bdi = mapping->backing_dev_info;
int ret = 0;
int done = 0;
struct pagevec pvec;
}
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();
return ret;
}
-static noinline void flush_write_bio(void *data)
+static void flush_epd_write_bio(struct extent_page_data *epd)
{
- struct extent_page_data *epd = data;
if (epd->bio) {
- submit_one_bio(WRITE, epd->bio, 0, 0);
+ if (epd->sync_io)
+ submit_one_bio(WRITE_SYNC, epd->bio, 0, 0);
+ else
+ submit_one_bio(WRITE, epd->bio, 0, 0);
epd->bio = NULL;
}
}
+static noinline void flush_write_bio(void *data)
+{
+ struct extent_page_data *epd = data;
+ flush_epd_write_bio(epd);
+}
+
int extent_write_full_page(struct extent_io_tree *tree, struct page *page,
get_extent_t *get_extent,
struct writeback_control *wbc)
.tree = tree,
.get_extent = get_extent,
.extent_locked = 0,
+ .sync_io = wbc->sync_mode == WB_SYNC_ALL,
};
struct writeback_control wbc_writepages = {
.bdi = wbc->bdi,
- .sync_mode = WB_SYNC_NONE,
+ .sync_mode = wbc->sync_mode,
.older_than_this = NULL,
.nr_to_write = 64,
.range_start = page_offset(page) + PAGE_CACHE_SIZE,
.range_end = (loff_t)-1,
};
-
ret = __extent_writepage(page, wbc, &epd);
extent_write_cache_pages(tree, mapping, &wbc_writepages,
__extent_writepage, &epd, flush_write_bio);
- if (epd.bio)
- submit_one_bio(WRITE, epd.bio, 0, 0);
+ flush_epd_write_bio(&epd);
return ret;
}
.tree = tree,
.get_extent = get_extent,
.extent_locked = 1,
+ .sync_io = mode == WB_SYNC_ALL,
};
struct writeback_control wbc_writepages = {
.bdi = inode->i_mapping->backing_dev_info,
start += PAGE_CACHE_SIZE;
}
- if (epd.bio)
- submit_one_bio(WRITE, epd.bio, 0, 0);
+ flush_epd_write_bio(&epd);
return ret;
}
.tree = tree,
.get_extent = get_extent,
.extent_locked = 0,
+ .sync_io = wbc->sync_mode == WB_SYNC_ALL,
};
ret = extent_write_cache_pages(tree, mapping, wbc,
__extent_writepage, &epd,
flush_write_bio);
- if (epd.bio)
- submit_one_bio(WRITE, epd.bio, 0, 0);
+ flush_epd_write_bio(&epd);
return ret;
}
return 0;
lock_extent(tree, start, end, GFP_NOFS);
- wait_on_extent_writeback(tree, start, end);
+ wait_on_page_writeback(page);
clear_extent_bit(tree, start, end,
EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC,
1, 1, GFP_NOFS);
u64 len;
while (start <= end) {
len = end - start + 1;
- spin_lock(&map->lock);
+ write_lock(&map->lock);
em = lookup_extent_mapping(map, start, len);
if (!em || IS_ERR(em)) {
- spin_unlock(&map->lock);
+ write_unlock(&map->lock);
break;
}
if (test_bit(EXTENT_FLAG_PINNED, &em->flags) ||
em->start != start) {
- spin_unlock(&map->lock);
+ write_unlock(&map->lock);
free_extent_map(em);
break;
}
free_extent_map(em);
}
start = extent_map_end(em);
- spin_unlock(&map->lock);
+ write_unlock(&map->lock);
/* once for us */
free_extent_map(em);
git://bbs.cooldavid.org / net-next-2.6.git / blobdiff