summaryrefslogtreecommitdiff
path: root/fs/ext4/file.c
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2016-03-18 02:31:18 +0300
committerLinus Torvalds <torvalds@linux-foundation.org>2016-03-18 02:31:18 +0300
commitfaeb20ecfa398b043c3224607f512c009c51653d (patch)
treeffd185ffb5e499a76f261c700de72241e6781ecf /fs/ext4/file.c
parent364e8dd9d636fea7def862919aac092b19b7c581 (diff)
parent0304688676bdfc8159e165313d71da19c118ba27 (diff)
downloadlinux-faeb20ecfa398b043c3224607f512c009c51653d.tar.xz
Merge tag 'ext4_for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4
Pull ext4 updates from Ted Ts'o: "Performance improvements in SEEK_DATA and xattr scalability improvements, plus a lot of clean ups and bug fixes" * tag 'ext4_for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4: (38 commits) ext4: clean up error handling in the MMP support jbd2: do not fail journal because of frozen_buffer allocation failure ext4: use __GFP_NOFAIL in ext4_free_blocks() ext4: fix compile error while opening the macro DOUBLE_CHECK ext4: print ext4 mount option data_err=abort correctly ext4: fix NULL pointer dereference in ext4_mark_inode_dirty() ext4: drop unneeded BUFFER_TRACE in ext4_delete_inline_entry() ext4: fix misspellings in comments. jbd2: fix FS corruption possibility in jbd2_journal_destroy() on umount path ext4: more efficient SEEK_DATA implementation ext4: cleanup handling of bh->b_state in DAX mmap ext4: return hole from ext4_map_blocks() ext4: factor out determining of hole size ext4: fix setting of referenced bit in ext4_es_lookup_extent() ext4: remove i_ioend_count ext4: simplify io_end handling for AIO DIO ext4: move trans handling and completion deferal out of _ext4_get_block ext4: rename and split get blocks functions ext4: use i_mutex to serialize unaligned AIO DIO ext4: pack ioend structure better ...
Diffstat (limited to 'fs/ext4/file.c')
-rw-r--r--fs/ext4/file.c129
1 files changed, 49 insertions, 80 deletions
diff --git a/fs/ext4/file.c b/fs/ext4/file.c
index 4cd318f31cbe..6659e216385e 100644
--- a/fs/ext4/file.c
+++ b/fs/ext4/file.c
@@ -93,31 +93,29 @@ ext4_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
{
struct file *file = iocb->ki_filp;
struct inode *inode = file_inode(iocb->ki_filp);
- struct mutex *aio_mutex = NULL;
struct blk_plug plug;
int o_direct = iocb->ki_flags & IOCB_DIRECT;
+ int unaligned_aio = 0;
int overwrite = 0;
ssize_t ret;
+ inode_lock(inode);
+ ret = generic_write_checks(iocb, from);
+ if (ret <= 0)
+ goto out;
+
/*
- * Unaligned direct AIO must be serialized; see comment above
- * In the case of O_APPEND, assume that we must always serialize
+ * Unaligned direct AIO must be serialized among each other as zeroing
+ * of partial blocks of two competing unaligned AIOs can result in data
+ * corruption.
*/
- if (o_direct &&
- ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS) &&
+ if (o_direct && ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS) &&
!is_sync_kiocb(iocb) &&
- (iocb->ki_flags & IOCB_APPEND ||
- ext4_unaligned_aio(inode, from, iocb->ki_pos))) {
- aio_mutex = ext4_aio_mutex(inode);
- mutex_lock(aio_mutex);
+ ext4_unaligned_aio(inode, from, iocb->ki_pos)) {
+ unaligned_aio = 1;
ext4_unwritten_wait(inode);
}
- inode_lock(inode);
- ret = generic_write_checks(iocb, from);
- if (ret <= 0)
- goto out;
-
/*
* If we have encountered a bitmap-format file, the size limit
* is smaller than s_maxbytes, which is for extent-mapped files.
@@ -139,7 +137,7 @@ ext4_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
blk_start_plug(&plug);
/* check whether we do a DIO overwrite or not */
- if (ext4_should_dioread_nolock(inode) && !aio_mutex &&
+ if (ext4_should_dioread_nolock(inode) && !unaligned_aio &&
!file->f_mapping->nrpages && pos + length <= i_size_read(inode)) {
struct ext4_map_blocks map;
unsigned int blkbits = inode->i_blkbits;
@@ -181,14 +179,10 @@ ext4_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
if (o_direct)
blk_finish_plug(&plug);
- if (aio_mutex)
- mutex_unlock(aio_mutex);
return ret;
out:
inode_unlock(inode);
- if (aio_mutex)
- mutex_unlock(aio_mutex);
return ret;
}
@@ -417,7 +411,7 @@ static int ext4_file_open(struct inode * inode, struct file * filp)
*/
static int ext4_find_unwritten_pgoff(struct inode *inode,
int whence,
- struct ext4_map_blocks *map,
+ ext4_lblk_t end_blk,
loff_t *offset)
{
struct pagevec pvec;
@@ -432,7 +426,7 @@ static int ext4_find_unwritten_pgoff(struct inode *inode,
blkbits = inode->i_sb->s_blocksize_bits;
startoff = *offset;
lastoff = startoff;
- endoff = (loff_t)(map->m_lblk + map->m_len) << blkbits;
+ endoff = (loff_t)end_blk << blkbits;
index = startoff >> PAGE_CACHE_SHIFT;
end = endoff >> PAGE_CACHE_SHIFT;
@@ -550,12 +544,11 @@ out:
static loff_t ext4_seek_data(struct file *file, loff_t offset, loff_t maxsize)
{
struct inode *inode = file->f_mapping->host;
- struct ext4_map_blocks map;
struct extent_status es;
ext4_lblk_t start, last, end;
loff_t dataoff, isize;
int blkbits;
- int ret = 0;
+ int ret;
inode_lock(inode);
@@ -572,41 +565,32 @@ static loff_t ext4_seek_data(struct file *file, loff_t offset, loff_t maxsize)
dataoff = offset;
do {
- map.m_lblk = last;
- map.m_len = end - last + 1;
- ret = ext4_map_blocks(NULL, inode, &map, 0);
- if (ret > 0 && !(map.m_flags & EXT4_MAP_UNWRITTEN)) {
- if (last != start)
- dataoff = (loff_t)last << blkbits;
- break;
+ ret = ext4_get_next_extent(inode, last, end - last + 1, &es);
+ if (ret <= 0) {
+ /* No extent found -> no data */
+ if (ret == 0)
+ ret = -ENXIO;
+ inode_unlock(inode);
+ return ret;
}
- /*
- * If there is a delay extent at this offset,
- * it will be as a data.
- */
- ext4_es_find_delayed_extent_range(inode, last, last, &es);
- if (es.es_len != 0 && in_range(last, es.es_lblk, es.es_len)) {
- if (last != start)
- dataoff = (loff_t)last << blkbits;
+ last = es.es_lblk;
+ if (last != start)
+ dataoff = (loff_t)last << blkbits;
+ if (!ext4_es_is_unwritten(&es))
break;
- }
/*
* If there is a unwritten extent at this offset,
* it will be as a data or a hole according to page
* cache that has data or not.
*/
- if (map.m_flags & EXT4_MAP_UNWRITTEN) {
- int unwritten;
- unwritten = ext4_find_unwritten_pgoff(inode, SEEK_DATA,
- &map, &dataoff);
- if (unwritten)
- break;
- }
-
- last++;
+ if (ext4_find_unwritten_pgoff(inode, SEEK_DATA,
+ es.es_lblk + es.es_len, &dataoff))
+ break;
+ last += es.es_len;
dataoff = (loff_t)last << blkbits;
+ cond_resched();
} while (last <= end);
inode_unlock(inode);
@@ -623,12 +607,11 @@ static loff_t ext4_seek_data(struct file *file, loff_t offset, loff_t maxsize)
static loff_t ext4_seek_hole(struct file *file, loff_t offset, loff_t maxsize)
{
struct inode *inode = file->f_mapping->host;
- struct ext4_map_blocks map;
struct extent_status es;
ext4_lblk_t start, last, end;
loff_t holeoff, isize;
int blkbits;
- int ret = 0;
+ int ret;
inode_lock(inode);
@@ -645,44 +628,30 @@ static loff_t ext4_seek_hole(struct file *file, loff_t offset, loff_t maxsize)
holeoff = offset;
do {
- map.m_lblk = last;
- map.m_len = end - last + 1;
- ret = ext4_map_blocks(NULL, inode, &map, 0);
- if (ret > 0 && !(map.m_flags & EXT4_MAP_UNWRITTEN)) {
- last += ret;
- holeoff = (loff_t)last << blkbits;
- continue;
+ ret = ext4_get_next_extent(inode, last, end - last + 1, &es);
+ if (ret < 0) {
+ inode_unlock(inode);
+ return ret;
}
-
- /*
- * If there is a delay extent at this offset,
- * we will skip this extent.
- */
- ext4_es_find_delayed_extent_range(inode, last, last, &es);
- if (es.es_len != 0 && in_range(last, es.es_lblk, es.es_len)) {
- last = es.es_lblk + es.es_len;
- holeoff = (loff_t)last << blkbits;
- continue;
+ /* Found a hole? */
+ if (ret == 0 || es.es_lblk > last) {
+ if (last != start)
+ holeoff = (loff_t)last << blkbits;
+ break;
}
-
/*
* If there is a unwritten extent at this offset,
* it will be as a data or a hole according to page
* cache that has data or not.
*/
- if (map.m_flags & EXT4_MAP_UNWRITTEN) {
- int unwritten;
- unwritten = ext4_find_unwritten_pgoff(inode, SEEK_HOLE,
- &map, &holeoff);
- if (!unwritten) {
- last += ret;
- holeoff = (loff_t)last << blkbits;
- continue;
- }
- }
+ if (ext4_es_is_unwritten(&es) &&
+ ext4_find_unwritten_pgoff(inode, SEEK_HOLE,
+ last + es.es_len, &holeoff))
+ break;
- /* find a hole */
- break;
+ last += es.es_len;
+ holeoff = (loff_t)last << blkbits;
+ cond_resched();
} while (last <= end);
inode_unlock(inode);