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
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)
}
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
}
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;
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);
* [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;
}
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)
}
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)
{
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);
/*
* 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;
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
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]);
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;
* the caller is taking responsibility for
* locked_page
*/
- if (pages[i] != locked_page)
+ if (pages[i] != locked_page) {
lock_page(pages[i]);
+ if (!PageDirty(pages[i]) ||
+ pages[i]->mapping != inode->i_mapping) {
+ ret = -EAGAIN;
+ unlock_page(pages[i]);
+ page_cache_release(pages[i]);
+ goto done;
+ }
+ }
page_cache_release(pages[i]);
+ pages_locked++;
}
- pages_locked += ret;
nrpages -= ret;
index += ret;
cond_resched();
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.
*/
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,
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;
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,
- 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]);
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;
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;
}
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;
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
int whole_page;
int ret;
+ if (err)
+ uptodate = 0;
+
do {
struct page *page = bvec->bv_page;
tree = &BTRFS_I(page->mapping->host)->io_tree;
if (ret == 0) {
uptodate =
test_bit(BIO_UPTODATE, &bio->bi_flags);
+ if (err)
+ uptodate = 0;
continue;
}
}
- if (uptodate)
+ if (uptodate) {
set_extent_uptodate(tree, start, end,
GFP_ATOMIC);
+ }
unlock_extent(tree, start, end, GFP_ATOMIC);
if (whole_page) {
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;
set_page_private(page, EXTENT_PAGE_PRIVATE);
}
}
+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);
}
}
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;
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);
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) {
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");
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;
}
* 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)
+ writepage_t writepage, void *data,
+ void (*flush_fn)(void *))
{
struct backing_dev_info *bdi = mapping->backing_dev_info;
int ret = 0;
continue;
}
- if (wbc->sync_mode != WB_SYNC_NONE)
+ if (wbc->sync_mode != WB_SYNC_NONE) {
+ if (PageWriteback(page))
+ flush_fn(data);
wait_on_page_writeback(page);
+ }
if (PageWriteback(page) ||
!clear_page_dirty_for_io(page)) {
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;
index = 0;
goto 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;
}
-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,
.bio = NULL,
.tree = tree,
.get_extent = get_extent,
+ .extent_locked = 0,
};
struct writeback_control wbc_writepages = {
.bdi = wbc->bdi,
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);
}
}
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,
.bio = NULL,
.tree = tree,
.get_extent = get_extent,
+ .extent_locked = 0,
};
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);
}
/* 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);
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;
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);
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);
git://ftp.safe.ca / safe / jmp / linux-2.6 / blobdiff