void jbd_preclean_buffer_check(struct buffer_head *bh)
{
if (buffer_jbd(bh)) {
struct journal_head *jh = bh2jh(bh);
transaction_t *transaction = jh->b_transaction;
journal_t *journal;
if (jh->b_jlist == 0 && transaction == NULL)
return;
J_ASSERT_JH(jh, transaction != NULL);
/* The kernel may be unmapping old data. We expect it
* to be dirty in that case, unless the buffer has
* already been forgotten by a transaction. */
if (jh->b_jlist != BJ_Forget) {
#if 1
if (!buffer_dirty(bh)) {
printk("%s: clean of clean buffer\n",
__FUNCTION__);
print_buffer_trace(bh);
return;
}
#endif
J_ASSERT_BH(bh, buffer_dirty(bh));
}
journal = transaction->t_journal;
J_ASSERT_JH(jh,
transaction == journal->j_running_transaction ||
transaction == journal->j_committing_transaction);
}
}
void print_buffer_fields(struct buffer_head *bh)
{
printk("b_next:%p, b_blocknr:%lu b_count:%d b_flushtime:%lu\n",
bh->b_next, bh->b_blocknr, atomic_read(&bh->b_count),
bh->b_flushtime);
printk("b_next_free:%p b_prev_free:%p b_this_page:%p b_reqnext:%p\n",
bh->b_next_free, bh->b_prev_free, bh->b_this_page,
bh->b_reqnext);
printk("b_pprev:%p b_data:%p b_page:%p b_inode:%p b_list:%d\n",
bh->b_pprev, bh->b_data, bh->b_page, bh->b_inode, bh->b_list);
#if defined(CONFIG_JBD) || defined(CONFIG_JBD_MODULE)
if (buffer_jbd(bh)) {
struct journal_head *jh = bh2jh(bh);
printk("b_jlist:%u b_frozen_data:%p b_committed_data:%p\n",
jh->b_jlist, jh->b_frozen_data, jh->b_committed_data);
printk(" b_transaction:%p b_next_transaction:%p "
"b_cp_transaction:%p\n",
jh->b_transaction, jh->b_next_transaction,
jh->b_cp_transaction);
printk("b_cpnext:%p b_cpprev:%p\n",
jh->b_cpnext, jh->b_cpprev);
}
#endif
}
int jbd2_journal_try_to_free_buffers(journal_t *journal,
struct page *page, gfp_t gfp_mask)
{
struct buffer_head *head;
struct buffer_head *bh;
int ret = 0;
J_ASSERT(PageLocked(page));
head = page_buffers(page);
bh = head;
do {
struct journal_head *jh;
jh = jbd2_journal_grab_journal_head(bh);
if (!jh)
continue;
jbd_lock_bh_state(bh);
__journal_try_to_free_buffer(journal, bh);
jbd2_journal_put_journal_head(jh);
jbd_unlock_bh_state(bh);
if (buffer_jbd(bh))
goto busy;
} while ((bh = bh->b_this_page) != head);
ret = try_to_free_buffers(page);
busy:
return ret;
}
/*
* Grab a ref against this buffer_head's journal_head. If it ended up not
* having a journal_head, return NULL
*/
struct journal_head *journal_grab_journal_head(struct buffer_head *bh)
{
struct journal_head *jh = NULL;
jbd_lock_bh_journal_head(bh);
if (buffer_jbd(bh)) {
jh = bh2jh(bh);
jh->b_jcount++;
}
jbd_unlock_bh_journal_head(bh);
return jh;
}
/*
* More or less lifted from ext3. I'll leave their description below:
*
* "For ext3 allocations, we must not reuse any blocks which are
* allocated in the bitmap buffer's "last committed data" copy. This
* prevents deletes from freeing up the page for reuse until we have
* committed the delete transaction.
*
* If we didn't do this, then deleting something and reallocating it as
* data would allow the old block to be overwritten before the
* transaction committed (because we force data to disk before commit).
* This would lead to corruption if we crashed between overwriting the
* data and committing the delete.
*
* @@@ We may want to make this allocation behaviour conditional on
* data-writes at some point, and disable it for metadata allocations or
* sync-data inodes."
*
* Note: OCFS2 already does this differently for metadata vs data
* allocations, as those bitmaps are seperate and undo access is never
* called on a metadata group descriptor.
*/
static int ocfs2_test_bg_bit_allocatable(struct buffer_head *bg_bh,
int nr)
{
struct ocfs2_group_desc *bg = (struct ocfs2_group_desc *) bg_bh->b_data;
if (ocfs2_test_bit(nr, (unsigned long *)bg->bg_bitmap))
return 0;
if (!buffer_jbd(bg_bh) || !bh2jh(bg_bh)->b_committed_data)
return 1;
bg = (struct ocfs2_group_desc *) bh2jh(bg_bh)->b_committed_data;
return !ocfs2_test_bit(nr, (unsigned long *)bg->bg_bitmap);
}
/*
* Give a buffer_head a journal_head.
*
* Doesn't need the journal lock.
* May sleep.
*/
struct journal_head *journal_add_journal_head(struct buffer_head *bh)
{
struct journal_head *jh;
struct journal_head *new_jh = NULL;
repeat:
if (!buffer_jbd(bh)) {
new_jh = journal_alloc_journal_head();
memset(new_jh, 0, sizeof(*new_jh));
}
jbd_lock_bh_journal_head(bh);
if (buffer_jbd(bh)) {
jh = bh2jh(bh);
} else {
J_ASSERT_BH(bh,
(atomic_read(&bh->b_count) > 0) ||
(bh->b_page && bh->b_page->mapping));
if (!new_jh) {
jbd_unlock_bh_journal_head(bh);
goto repeat;
}
jh = new_jh;
new_jh = NULL; /* We consumed it */
set_buffer_jbd(bh);
bh->b_private = jh;
jh->b_bh = bh;
get_bh(bh);
BUFFER_TRACE(bh, "added journal_head");
}
jh->b_jcount++;
jbd_unlock_bh_journal_head(bh);
if (new_jh)
journal_free_journal_head(new_jh);
return bh->b_private;
}
int ocfs2_write_block(struct ocfs2_super *osb, struct buffer_head *bh,
struct inode *inode)
{
int ret = 0;
mlog_entry("(bh->b_blocknr = %llu, inode=%p)\n",
(unsigned long long)bh->b_blocknr, inode);
BUG_ON(bh->b_blocknr < OCFS2_SUPER_BLOCK_BLKNO);
BUG_ON(buffer_jbd(bh));
/* No need to check for a soft readonly file system here. non
* journalled writes are only ever done on system files which
* can get modified during recovery even if read-only. */
if (ocfs2_is_hard_readonly(osb)) {
ret = -EROFS;
goto out;
}
mutex_lock(&OCFS2_I(inode)->ip_io_mutex);
lock_buffer(bh);
set_buffer_uptodate(bh);
/* remove from dirty list before I/O. */
clear_buffer_dirty(bh);
get_bh(bh); /* for end_buffer_write_sync() */
bh->b_end_io = end_buffer_write_sync;
submit_bh(WRITE, bh);
wait_on_buffer(bh);
if (buffer_uptodate(bh)) {
ocfs2_set_buffer_uptodate(inode, bh);
} else {
/* We don't need to remove the clustered uptodate
* information for this bh as it's not marked locally
* uptodate. */
ret = -EIO;
put_bh(bh);
}
mutex_unlock(&OCFS2_I(inode)->ip_io_mutex);
out:
mlog_exit(ret);
return ret;
}
int ocfs2_write_block(struct ocfs2_super *osb, struct buffer_head *bh,
struct ocfs2_caching_info *ci)
{
int ret = 0;
trace_ocfs2_write_block((unsigned long long)bh->b_blocknr, ci);
BUG_ON(bh->b_blocknr < OCFS2_SUPER_BLOCK_BLKNO);
BUG_ON(buffer_jbd(bh));
/* No need to check for a soft readonly file system here. non
* journalled writes are only ever done on system files which
* can get modified during recovery even if read-only. */
if (ocfs2_is_hard_readonly(osb)) {
ret = -EROFS;
mlog_errno(ret);
goto out;
}
ocfs2_metadata_cache_io_lock(ci);
lock_buffer(bh);
set_buffer_uptodate(bh);
/* remove from dirty list before I/O. */
clear_buffer_dirty(bh);
get_bh(bh); /* for end_buffer_write_sync() */
bh->b_end_io = end_buffer_write_sync;
submit_bh(REQ_OP_WRITE, 0, bh);
wait_on_buffer(bh);
if (buffer_uptodate(bh)) {
ocfs2_set_buffer_uptodate(ci, bh);
} else {
/* We don't need to remove the clustered uptodate
* information for this bh as it's not marked locally
* uptodate. */
ret = -EIO;
mlog_errno(ret);
}
ocfs2_metadata_cache_io_unlock(ci);
out:
return ret;
}
/* Warning: even if it returns true, this does *not* guarantee that
* the block is stored in our inode metadata cache.
*
* This can be called under lock_buffer()
*/
int ocfs2_buffer_uptodate(struct inode *inode,
struct buffer_head *bh)
{
/* Doesn't matter if the bh is in our cache or not -- if it's
* not marked uptodate then we know it can't have correct
* data. */
if (!buffer_uptodate(bh))
return 0;
/* OCFS2 does not allow multiple nodes to be changing the same
* block at the same time. */
if (buffer_jbd(bh))
return 1;
/* Ok, locally the buffer is marked as up to date, now search
* our cache to see if we can trust that. */
return ocfs2_buffer_cached(OCFS2_I(inode), bh);
}
void print_buffer_fields(struct buffer_head *bh)
{
printk("b_blocknr:%llu b_count:%d\n",
(unsigned long long)bh->b_blocknr, atomic_read(&bh->b_count));
printk("b_this_page:%p b_data:%p b_page:%p\n",
bh->b_this_page, bh->b_data, bh->b_page);
#if defined(CONFIG_JBD2) || defined(CONFIG_JBD2_MODULE)
if (buffer_jbd(bh)) {
struct journal_head *jh = bh2jh(bh);
printk("b_jlist:%u b_frozen_data:%p b_committed_data:%p\n",
jh->b_jlist, jh->b_frozen_data, jh->b_committed_data);
printk(" b_transaction:%p b_next_transaction:%p "
"b_cp_transaction:%p\n",
jh->b_transaction, jh->b_next_transaction,
jh->b_cp_transaction);
printk("b_cpnext:%p b_cpprev:%p\n",
jh->b_cpnext, jh->b_cpprev);
}
#endif
}
/*
* More or less lifted from ext3. I'll leave their description below:
*
* "For ext3 allocations, we must not reuse any blocks which are
* allocated in the bitmap buffer's "last committed data" copy. This
* prevents deletes from freeing up the page for reuse until we have
* committed the delete transaction.
*
* If we didn't do this, then deleting something and reallocating it as
* data would allow the old block to be overwritten before the
* transaction committed (because we force data to disk before commit).
* This would lead to corruption if we crashed between overwriting the
* data and committing the delete.
*
* @@@ We may want to make this allocation behaviour conditional on
* data-writes at some point, and disable it for metadata allocations or
* sync-data inodes."
*
* Note: OCFS2 already does this differently for metadata vs data
* allocations, as those bitmaps are separate and undo access is never
* called on a metadata group descriptor.
*/
static int ocfs2_test_bg_bit_allocatable(struct buffer_head *bg_bh,
int nr)
{
struct ocfs2_group_desc *bg = (struct ocfs2_group_desc *) bg_bh->b_data;
int ret;
if (ocfs2_test_bit(nr, (unsigned long *)bg->bg_bitmap))
return 0;
if (!buffer_jbd(bg_bh))
return 1;
jbd_lock_bh_state(bg_bh);
bg = (struct ocfs2_group_desc *) bh2jh(bg_bh)->b_committed_data;
if (bg)
ret = !ocfs2_test_bit(nr, (unsigned long *)bg->bg_bitmap);
else
ret = 1;
jbd_unlock_bh_state(bg_bh);
return ret;
}
static void __journal_remove_journal_head(struct buffer_head *bh)
{
struct journal_head *jh = bh2jh(bh);
J_ASSERT_JH(jh, jh->b_jcount >= 0);
get_bh(bh);
if (jh->b_jcount == 0) {
if (jh->b_transaction == NULL &&
jh->b_next_transaction == NULL &&
jh->b_cp_transaction == NULL) {
J_ASSERT_JH(jh, jh->b_jlist == BJ_None);
J_ASSERT_BH(bh, buffer_jbd(bh));
J_ASSERT_BH(bh, jh2bh(jh) == bh);
BUFFER_TRACE(bh, "remove journal_head");
if (jh->b_frozen_data) {
printk(KERN_WARNING "%s: freeing "
"b_frozen_data\n",
__FUNCTION__);
jbd_free(jh->b_frozen_data, bh->b_size);
}
if (jh->b_committed_data) {
printk(KERN_WARNING "%s: freeing "
"b_committed_data\n",
__FUNCTION__);
jbd_free(jh->b_committed_data, bh->b_size);
}
bh->b_private = NULL;
jh->b_bh = NULL; /* debug, really */
clear_buffer_jbd(bh);
__brelse(bh);
journal_free_journal_head(jh);
} else {
BUFFER_TRACE(bh, "journal_head was locked");
}
}
}
/*
* Write super block and backups doesn't need to collaborate with journal,
* so we don't need to lock ip_io_mutex and inode doesn't need to bea passed
* into this function.
*/
int ocfs2_write_super_or_backup(struct ocfs2_super *osb,
struct buffer_head *bh)
{
int ret = 0;
mlog_entry_void();
BUG_ON(buffer_jbd(bh));
ocfs2_check_super_or_backup(osb->sb, bh->b_blocknr);
if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb)) {
ret = -EROFS;
goto out;
}
lock_buffer(bh);
set_buffer_uptodate(bh);
/* remove from dirty list before I/O. */
clear_buffer_dirty(bh);
get_bh(bh); /* for end_buffer_write_sync() */
bh->b_end_io = end_buffer_write_sync;
submit_bh(WRITE, bh);
wait_on_buffer(bh);
if (!buffer_uptodate(bh)) {
ret = -EIO;
put_bh(bh);
}
out:
mlog_exit(ret);
return ret;
}
int journal_try_to_free_buffers(journal_t *journal,
struct page *page, gfp_t gfp_mask)
{
struct buffer_head *head;
struct buffer_head *bh;
int ret = 0;
J_ASSERT(PageLocked(page));
head = page_buffers(page);
bh = head;
do {
struct journal_head *jh;
/*
* We take our own ref against the journal_head here to avoid
* having to add tons of locking around each instance of
* journal_remove_journal_head() and journal_put_journal_head().
*/
jh = journal_grab_journal_head(bh);
if (!jh)
continue;
jbd_lock_bh_state(bh);
__journal_try_to_free_buffer(journal, bh);
journal_put_journal_head(jh);
jbd_unlock_bh_state(bh);
if (buffer_jbd(bh))
goto busy;
} while ((bh = bh->b_this_page) != head);
ret = try_to_free_buffers(page);
busy:
return ret;
}
/*
* Write super block and backups doesn't need to collaborate with journal,
* so we don't need to lock ip_io_mutex and ci doesn't need to bea passed
* into this function.
*/
int ocfs2_write_super_or_backup(struct ocfs2_super *osb,
struct buffer_head *bh)
{
int ret = 0;
struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
BUG_ON(buffer_jbd(bh));
ocfs2_check_super_or_backup(osb->sb, bh->b_blocknr);
if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb)) {
ret = -EROFS;
mlog_errno(ret);
goto out;
}
lock_buffer(bh);
set_buffer_uptodate(bh);
/* remove from dirty list before I/O. */
clear_buffer_dirty(bh);
get_bh(bh); /* for end_buffer_write_sync() */
bh->b_end_io = end_buffer_write_sync;
ocfs2_compute_meta_ecc(osb->sb, bh->b_data, &di->i_check);
submit_bh(REQ_OP_WRITE, 0, bh);
wait_on_buffer(bh);
if (!buffer_uptodate(bh)) {
ret = -EIO;
mlog_errno(ret);
}
out:
return ret;
}
int ocfs2_read_blocks(struct inode *inode, u64 block, int nr,
struct buffer_head *bhs[], int flags,
int (*validate)(struct super_block *sb,
struct buffer_head *bh))
{
int status = 0;
int i, ignore_cache = 0;
struct buffer_head *bh;
mlog_entry("(inode=%p, block=(%llu), nr=(%d), flags=%d)\n",
inode, (unsigned long long)block, nr, flags);
BUG_ON(!inode);
BUG_ON((flags & OCFS2_BH_READAHEAD) &&
(flags & OCFS2_BH_IGNORE_CACHE));
if (bhs == NULL) {
status = -EINVAL;
mlog_errno(status);
goto bail;
}
if (nr < 0) {
mlog(ML_ERROR, "asked to read %d blocks!\n", nr);
status = -EINVAL;
mlog_errno(status);
goto bail;
}
if (nr == 0) {
mlog(ML_BH_IO, "No buffers will be read!\n");
status = 0;
goto bail;
}
mutex_lock(&OCFS2_I(inode)->ip_io_mutex);
for (i = 0 ; i < nr ; i++) {
if (bhs[i] == NULL) {
bhs[i] = sb_getblk(inode->i_sb, block++);
if (bhs[i] == NULL) {
mutex_unlock(&OCFS2_I(inode)->ip_io_mutex);
status = -EIO;
mlog_errno(status);
goto bail;
}
}
bh = bhs[i];
ignore_cache = (flags & OCFS2_BH_IGNORE_CACHE);
/* There are three read-ahead cases here which we need to
* be concerned with. All three assume a buffer has
* previously been submitted with OCFS2_BH_READAHEAD
* and it hasn't yet completed I/O.
*
* 1) The current request is sync to disk. This rarely
* happens these days, and never when performance
* matters - the code can just wait on the buffer
* lock and re-submit.
*
* 2) The current request is cached, but not
* readahead. ocfs2_buffer_uptodate() will return
* false anyway, so we'll wind up waiting on the
* buffer lock to do I/O. We re-check the request
* with after getting the lock to avoid a re-submit.
*
* 3) The current request is readahead (and so must
* also be a caching one). We short circuit if the
* buffer is locked (under I/O) and if it's in the
* uptodate cache. The re-check from #2 catches the
* case that the previous read-ahead completes just
* before our is-it-in-flight check.
*/
if (!ignore_cache && !ocfs2_buffer_uptodate(inode, bh)) {
mlog(ML_UPTODATE,
"bh (%llu), inode %llu not uptodate\n",
(unsigned long long)bh->b_blocknr,
(unsigned long long)OCFS2_I(inode)->ip_blkno);
/* We're using ignore_cache here to say
* "go to disk" */
ignore_cache = 1;
}
if (buffer_jbd(bh)) {
if (ignore_cache)
mlog(ML_BH_IO, "trying to sync read a jbd "
"managed bh (blocknr = %llu)\n",
(unsigned long long)bh->b_blocknr);
continue;
}
if (ignore_cache) {
if (buffer_dirty(bh)) {
/* This should probably be a BUG, or
* at least return an error. */
mlog(ML_BH_IO, "asking me to sync read a dirty "
"buffer! (blocknr = %llu)\n",
(unsigned long long)bh->b_blocknr);
continue;
}
/* A read-ahead request was made - if the
* buffer is already under read-ahead from a
* previously submitted request than we are
* done here. */
if ((flags & OCFS2_BH_READAHEAD)
&& ocfs2_buffer_read_ahead(inode, bh))
continue;
lock_buffer(bh);
if (buffer_jbd(bh)) {
#ifdef CATCH_BH_JBD_RACES
mlog(ML_ERROR, "block %llu had the JBD bit set "
"while I was in lock_buffer!",
(unsigned long long)bh->b_blocknr);
BUG();
#else
unlock_buffer(bh);
continue;
#endif
}
/* Re-check ocfs2_buffer_uptodate() as a
* previously read-ahead buffer may have
* completed I/O while we were waiting for the
* buffer lock. */
if (!(flags & OCFS2_BH_IGNORE_CACHE)
&& !(flags & OCFS2_BH_READAHEAD)
&& ocfs2_buffer_uptodate(inode, bh)) {
unlock_buffer(bh);
continue;
}
clear_buffer_uptodate(bh);
get_bh(bh); /* for end_buffer_read_sync() */
if (validate)
set_buffer_needs_validate(bh);
bh->b_end_io = end_buffer_read_sync;
submit_bh(READ, bh);
continue;
}
}
status = 0;
for (i = (nr - 1); i >= 0; i--) {
bh = bhs[i];
if (!(flags & OCFS2_BH_READAHEAD)) {
/* We know this can't have changed as we hold the
* inode sem. Avoid doing any work on the bh if the
* journal has it. */
if (!buffer_jbd(bh))
wait_on_buffer(bh);
if (!buffer_uptodate(bh)) {
/* Status won't be cleared from here on out,
* so we can safely record this and loop back
* to cleanup the other buffers. Don't need to
* remove the clustered uptodate information
* for this bh as it's not marked locally
* uptodate. */
status = -EIO;
put_bh(bh);
bhs[i] = NULL;
continue;
}
if (buffer_needs_validate(bh)) {
/* We never set NeedsValidate if the
* buffer was held by the journal, so
* that better not have changed */
BUG_ON(buffer_jbd(bh));
clear_buffer_needs_validate(bh);
status = validate(inode->i_sb, bh);
if (status) {
put_bh(bh);
bhs[i] = NULL;
continue;
}
}
}
/* Always set the buffer in the cache, even if it was
* a forced read, or read-ahead which hasn't yet
* completed. */
ocfs2_set_buffer_uptodate(inode, bh);
}
mutex_unlock(&OCFS2_I(inode)->ip_io_mutex);
mlog(ML_BH_IO, "block=(%llu), nr=(%d), cached=%s, flags=0x%x\n",
(unsigned long long)block, nr,
((flags & OCFS2_BH_IGNORE_CACHE) || ignore_cache) ? "no" : "yes",
flags);
bail:
mlog_exit(status);
return status;
}
int ocfs2_read_blocks_sync(struct ocfs2_super *osb, u64 block,
unsigned int nr, struct buffer_head *bhs[])
{
int status = 0;
unsigned int i;
struct buffer_head *bh;
if (!nr) {
mlog(ML_BH_IO, "No buffers will be read!\n");
goto bail;
}
for (i = 0 ; i < nr ; i++) {
if (bhs[i] == NULL) {
bhs[i] = sb_getblk(osb->sb, block++);
if (bhs[i] == NULL) {
status = -EIO;
mlog_errno(status);
goto bail;
}
}
bh = bhs[i];
if (buffer_jbd(bh)) {
mlog(ML_BH_IO,
"trying to sync read a jbd "
"managed bh (blocknr = %llu), skipping\n",
(unsigned long long)bh->b_blocknr);
continue;
}
if (buffer_dirty(bh)) {
/* This should probably be a BUG, or
* at least return an error. */
mlog(ML_ERROR,
"trying to sync read a dirty "
"buffer! (blocknr = %llu), skipping\n",
(unsigned long long)bh->b_blocknr);
continue;
}
lock_buffer(bh);
if (buffer_jbd(bh)) {
mlog(ML_ERROR,
"block %llu had the JBD bit set "
"while I was in lock_buffer!",
(unsigned long long)bh->b_blocknr);
BUG();
}
clear_buffer_uptodate(bh);
get_bh(bh); /* for end_buffer_read_sync() */
bh->b_end_io = end_buffer_read_sync;
submit_bh(READ, bh);
}
for (i = nr; i > 0; i--) {
bh = bhs[i - 1];
/* No need to wait on the buffer if it's managed by JBD. */
if (!buffer_jbd(bh))
wait_on_buffer(bh);
if (!buffer_uptodate(bh)) {
/* Status won't be cleared from here on out,
* so we can safely record this and loop back
* to cleanup the other buffers. */
status = -EIO;
put_bh(bh);
bhs[i - 1] = NULL;
}
}
bail:
return status;
}
/*
* journal_commit_transaction
*
* The primary function for committing a transaction to the log. This
* function is called by the journal thread to begin a complete commit.
*/
void journal_commit_transaction(journal_t *journal)
{
transaction_t *commit_transaction;
struct journal_head *jh, *new_jh, *descriptor;
struct buffer_head **wbuf = journal->j_wbuf;
int bufs;
int flags;
int err;
unsigned int blocknr;
ktime_t start_time;
u64 commit_time;
char *tagp = NULL;
journal_header_t *header;
journal_block_tag_t *tag = NULL;
int space_left = 0;
int first_tag = 0;
int tag_flag;
int i;
struct blk_plug plug;
/*
* First job: lock down the current transaction and wait for
* all outstanding updates to complete.
*/
/* Do we need to erase the effects of a prior journal_flush? */
if (journal->j_flags & JFS_FLUSHED) {
jbd_debug(3, "super block updated\n");
journal_update_superblock(journal, 1);
} else {
jbd_debug(3, "superblock not updated\n");
}
J_ASSERT(journal->j_running_transaction != NULL);
J_ASSERT(journal->j_committing_transaction == NULL);
commit_transaction = journal->j_running_transaction;
J_ASSERT(commit_transaction->t_state == T_RUNNING);
trace_jbd_start_commit(journal, commit_transaction);
jbd_debug(1, "JBD: starting commit of transaction %d\n",
commit_transaction->t_tid);
spin_lock(&journal->j_state_lock);
commit_transaction->t_state = T_LOCKED;
trace_jbd_commit_locking(journal, commit_transaction);
spin_lock(&commit_transaction->t_handle_lock);
while (commit_transaction->t_updates) {
DEFINE_WAIT(wait);
prepare_to_wait(&journal->j_wait_updates, &wait,
TASK_UNINTERRUPTIBLE);
if (commit_transaction->t_updates) {
spin_unlock(&commit_transaction->t_handle_lock);
spin_unlock(&journal->j_state_lock);
schedule();
spin_lock(&journal->j_state_lock);
spin_lock(&commit_transaction->t_handle_lock);
}
finish_wait(&journal->j_wait_updates, &wait);
}
spin_unlock(&commit_transaction->t_handle_lock);
J_ASSERT (commit_transaction->t_outstanding_credits <=
journal->j_max_transaction_buffers);
/*
* First thing we are allowed to do is to discard any remaining
* BJ_Reserved buffers. Note, it is _not_ permissible to assume
* that there are no such buffers: if a large filesystem
* operation like a truncate needs to split itself over multiple
* transactions, then it may try to do a journal_restart() while
* there are still BJ_Reserved buffers outstanding. These must
* be released cleanly from the current transaction.
*
* In this case, the filesystem must still reserve write access
* again before modifying the buffer in the new transaction, but
* we do not require it to remember exactly which old buffers it
* has reserved. This is consistent with the existing behaviour
* that multiple journal_get_write_access() calls to the same
* buffer are perfectly permissible.
*/
while (commit_transaction->t_reserved_list) {
jh = commit_transaction->t_reserved_list;
JBUFFER_TRACE(jh, "reserved, unused: refile");
/*
* A journal_get_undo_access()+journal_release_buffer() may
* leave undo-committed data.
*/
if (jh->b_committed_data) {
struct buffer_head *bh = jh2bh(jh);
jbd_lock_bh_state(bh);
jbd_free(jh->b_committed_data, bh->b_size);
jh->b_committed_data = NULL;
jbd_unlock_bh_state(bh);
}
journal_refile_buffer(journal, jh);
}
/*
* Now try to drop any written-back buffers from the journal's
* checkpoint lists. We do this *before* commit because it potentially
* frees some memory
*/
spin_lock(&journal->j_list_lock);
__journal_clean_checkpoint_list(journal);
spin_unlock(&journal->j_list_lock);
jbd_debug (3, "JBD: commit phase 1\n");
/*
* Clear revoked flag to reflect there is no revoked buffers
* in the next transaction which is going to be started.
*/
journal_clear_buffer_revoked_flags(journal);
/*
* Switch to a new revoke table.
*/
journal_switch_revoke_table(journal);
trace_jbd_commit_flushing(journal, commit_transaction);
commit_transaction->t_state = T_FLUSH;
journal->j_committing_transaction = commit_transaction;
journal->j_running_transaction = NULL;
start_time = ktime_get();
commit_transaction->t_log_start = journal->j_head;
wake_up(&journal->j_wait_transaction_locked);
spin_unlock(&journal->j_state_lock);
jbd_debug (3, "JBD: commit phase 2\n");
/*
* Now start flushing things to disk, in the order they appear
* on the transaction lists. Data blocks go first.
*/
blk_start_plug(&plug);
err = journal_submit_data_buffers(journal, commit_transaction,
WRITE_SYNC);
blk_finish_plug(&plug);
/*
* Wait for all previously submitted IO to complete.
*/
spin_lock(&journal->j_list_lock);
while (commit_transaction->t_locked_list) {
struct buffer_head *bh;
jh = commit_transaction->t_locked_list->b_tprev;
bh = jh2bh(jh);
get_bh(bh);
if (buffer_locked(bh)) {
spin_unlock(&journal->j_list_lock);
wait_on_buffer(bh);
spin_lock(&journal->j_list_lock);
}
if (unlikely(!buffer_uptodate(bh))) {
if (!trylock_page(bh->b_page)) {
spin_unlock(&journal->j_list_lock);
lock_page(bh->b_page);
spin_lock(&journal->j_list_lock);
}
if (bh->b_page->mapping)
set_bit(AS_EIO, &bh->b_page->mapping->flags);
unlock_page(bh->b_page);
SetPageError(bh->b_page);
err = -EIO;
}
if (!inverted_lock(journal, bh)) {
put_bh(bh);
spin_lock(&journal->j_list_lock);
continue;
}
if (buffer_jbd(bh) && bh2jh(bh) == jh &&
jh->b_transaction == commit_transaction &&
jh->b_jlist == BJ_Locked)
__journal_unfile_buffer(jh);
jbd_unlock_bh_state(bh);
release_data_buffer(bh);
cond_resched_lock(&journal->j_list_lock);
}
spin_unlock(&journal->j_list_lock);
if (err) {
char b[BDEVNAME_SIZE];
printk(KERN_WARNING
"JBD: Detected IO errors while flushing file data "
"on %s\n", bdevname(journal->j_fs_dev, b));
if (journal->j_flags & JFS_ABORT_ON_SYNCDATA_ERR)
journal_abort(journal, err);
err = 0;
}
blk_start_plug(&plug);
journal_write_revoke_records(journal, commit_transaction, WRITE_SYNC);
/*
* If we found any dirty or locked buffers, then we should have
* looped back up to the write_out_data label. If there weren't
* any then journal_clean_data_list should have wiped the list
* clean by now, so check that it is in fact empty.
*/
J_ASSERT (commit_transaction->t_sync_datalist == NULL);
jbd_debug (3, "JBD: commit phase 3\n");
/*
* Way to go: we have now written out all of the data for a
* transaction! Now comes the tricky part: we need to write out
* metadata. Loop over the transaction's entire buffer list:
*/
spin_lock(&journal->j_state_lock);
commit_transaction->t_state = T_COMMIT;
spin_unlock(&journal->j_state_lock);
trace_jbd_commit_logging(journal, commit_transaction);
J_ASSERT(commit_transaction->t_nr_buffers <=
commit_transaction->t_outstanding_credits);
descriptor = NULL;
bufs = 0;
while (commit_transaction->t_buffers) {
/* Find the next buffer to be journaled... */
jh = commit_transaction->t_buffers;
/* If we're in abort mode, we just un-journal the buffer and
release it. */
if (is_journal_aborted(journal)) {
clear_buffer_jbddirty(jh2bh(jh));
JBUFFER_TRACE(jh, "journal is aborting: refile");
journal_refile_buffer(journal, jh);
/* If that was the last one, we need to clean up
* any descriptor buffers which may have been
* already allocated, even if we are now
* aborting. */
if (!commit_transaction->t_buffers)
goto start_journal_io;
continue;
}
/* Make sure we have a descriptor block in which to
record the metadata buffer. */
if (!descriptor) {
struct buffer_head *bh;
J_ASSERT (bufs == 0);
jbd_debug(4, "JBD: get descriptor\n");
descriptor = journal_get_descriptor_buffer(journal);
if (!descriptor) {
journal_abort(journal, -EIO);
continue;
}
bh = jh2bh(descriptor);
jbd_debug(4, "JBD: got buffer %llu (%p)\n",
(unsigned long long)bh->b_blocknr, bh->b_data);
header = (journal_header_t *)&bh->b_data[0];
header->h_magic = cpu_to_be32(JFS_MAGIC_NUMBER);
header->h_blocktype = cpu_to_be32(JFS_DESCRIPTOR_BLOCK);
header->h_sequence = cpu_to_be32(commit_transaction->t_tid);
tagp = &bh->b_data[sizeof(journal_header_t)];
space_left = bh->b_size - sizeof(journal_header_t);
first_tag = 1;
set_buffer_jwrite(bh);
set_buffer_dirty(bh);
wbuf[bufs++] = bh;
/* Record it so that we can wait for IO
completion later */
BUFFER_TRACE(bh, "ph3: file as descriptor");
journal_file_buffer(descriptor, commit_transaction,
BJ_LogCtl);
}
/* Where is the buffer to be written? */
err = journal_next_log_block(journal, &blocknr);
/* If the block mapping failed, just abandon the buffer
and repeat this loop: we'll fall into the
refile-on-abort condition above. */
if (err) {
journal_abort(journal, err);
continue;
}
/*
* start_this_handle() uses t_outstanding_credits to determine
* the free space in the log, but this counter is changed
* by journal_next_log_block() also.
*/
commit_transaction->t_outstanding_credits--;
/* Bump b_count to prevent truncate from stumbling over
the shadowed buffer! @@@ This can go if we ever get
rid of the BJ_IO/BJ_Shadow pairing of buffers. */
get_bh(jh2bh(jh));
/* Make a temporary IO buffer with which to write it out
(this will requeue both the metadata buffer and the
temporary IO buffer). new_bh goes on BJ_IO*/
set_buffer_jwrite(jh2bh(jh));
/*
* akpm: journal_write_metadata_buffer() sets
* new_bh->b_transaction to commit_transaction.
* We need to clean this up before we release new_bh
* (which is of type BJ_IO)
*/
JBUFFER_TRACE(jh, "ph3: write metadata");
flags = journal_write_metadata_buffer(commit_transaction,
jh, &new_jh, blocknr);
set_buffer_jwrite(jh2bh(new_jh));
wbuf[bufs++] = jh2bh(new_jh);
/* Record the new block's tag in the current descriptor
buffer */
tag_flag = 0;
if (flags & 1)
tag_flag |= JFS_FLAG_ESCAPE;
if (!first_tag)
tag_flag |= JFS_FLAG_SAME_UUID;
tag = (journal_block_tag_t *) tagp;
tag->t_blocknr = cpu_to_be32(jh2bh(jh)->b_blocknr);
tag->t_flags = cpu_to_be32(tag_flag);
tagp += sizeof(journal_block_tag_t);
space_left -= sizeof(journal_block_tag_t);
if (first_tag) {
memcpy (tagp, journal->j_uuid, 16);
tagp += 16;
space_left -= 16;
first_tag = 0;
}
/* If there's no more to do, or if the descriptor is full,
let the IO rip! */
if (bufs == journal->j_wbufsize ||
commit_transaction->t_buffers == NULL ||
space_left < sizeof(journal_block_tag_t) + 16) {
jbd_debug(4, "JBD: Submit %d IOs\n", bufs);
/* Write an end-of-descriptor marker before
submitting the IOs. "tag" still points to
the last tag we set up. */
tag->t_flags |= cpu_to_be32(JFS_FLAG_LAST_TAG);
start_journal_io:
for (i = 0; i < bufs; i++) {
struct buffer_head *bh = wbuf[i];
lock_buffer(bh);
clear_buffer_dirty(bh);
set_buffer_uptodate(bh);
bh->b_end_io = journal_end_buffer_io_sync;
submit_bh(WRITE_SYNC, bh);
}
cond_resched();
/* Force a new descriptor to be generated next
time round the loop. */
descriptor = NULL;
bufs = 0;
}
}
blk_finish_plug(&plug);
/* Lo and behold: we have just managed to send a transaction to
the log. Before we can commit it, wait for the IO so far to
complete. Control buffers being written are on the
transaction's t_log_list queue, and metadata buffers are on
the t_iobuf_list queue.
Wait for the buffers in reverse order. That way we are
less likely to be woken up until all IOs have completed, and
so we incur less scheduling load.
*/
jbd_debug(3, "JBD: commit phase 4\n");
/*
* akpm: these are BJ_IO, and j_list_lock is not needed.
* See __journal_try_to_free_buffer.
*/
wait_for_iobuf:
while (commit_transaction->t_iobuf_list != NULL) {
struct buffer_head *bh;
jh = commit_transaction->t_iobuf_list->b_tprev;
bh = jh2bh(jh);
if (buffer_locked(bh)) {
wait_on_buffer(bh);
goto wait_for_iobuf;
}
if (cond_resched())
goto wait_for_iobuf;
if (unlikely(!buffer_uptodate(bh)))
err = -EIO;
clear_buffer_jwrite(bh);
JBUFFER_TRACE(jh, "ph4: unfile after journal write");
journal_unfile_buffer(journal, jh);
/*
* ->t_iobuf_list should contain only dummy buffer_heads
* which were created by journal_write_metadata_buffer().
*/
BUFFER_TRACE(bh, "dumping temporary bh");
journal_put_journal_head(jh);
__brelse(bh);
J_ASSERT_BH(bh, atomic_read(&bh->b_count) == 0);
free_buffer_head(bh);
/* We also have to unlock and free the corresponding
shadowed buffer */
jh = commit_transaction->t_shadow_list->b_tprev;
bh = jh2bh(jh);
clear_buffer_jwrite(bh);
J_ASSERT_BH(bh, buffer_jbddirty(bh));
/* The metadata is now released for reuse, but we need
to remember it against this transaction so that when
we finally commit, we can do any checkpointing
required. */
JBUFFER_TRACE(jh, "file as BJ_Forget");
journal_file_buffer(jh, commit_transaction, BJ_Forget);
/*
* Wake up any transactions which were waiting for this
* IO to complete. The barrier must be here so that changes
* by journal_file_buffer() take effect before wake_up_bit()
* does the waitqueue check.
*/
smp_mb();
wake_up_bit(&bh->b_state, BH_Unshadow);
JBUFFER_TRACE(jh, "brelse shadowed buffer");
__brelse(bh);
}
J_ASSERT (commit_transaction->t_shadow_list == NULL);
jbd_debug(3, "JBD: commit phase 5\n");
/* Here we wait for the revoke record and descriptor record buffers */
wait_for_ctlbuf:
while (commit_transaction->t_log_list != NULL) {
struct buffer_head *bh;
jh = commit_transaction->t_log_list->b_tprev;
bh = jh2bh(jh);
if (buffer_locked(bh)) {
wait_on_buffer(bh);
goto wait_for_ctlbuf;
}
if (cond_resched())
goto wait_for_ctlbuf;
if (unlikely(!buffer_uptodate(bh)))
err = -EIO;
BUFFER_TRACE(bh, "ph5: control buffer writeout done: unfile");
clear_buffer_jwrite(bh);
journal_unfile_buffer(journal, jh);
journal_put_journal_head(jh);
__brelse(bh); /* One for getblk */
/* AKPM: bforget here */
}
if (err)
journal_abort(journal, err);
jbd_debug(3, "JBD: commit phase 6\n");
/* All metadata is written, now write commit record and do cleanup */
spin_lock(&journal->j_state_lock);
J_ASSERT(commit_transaction->t_state == T_COMMIT);
commit_transaction->t_state = T_COMMIT_RECORD;
spin_unlock(&journal->j_state_lock);
if (journal_write_commit_record(journal, commit_transaction))
err = -EIO;
if (err)
journal_abort(journal, err);
/* End of a transaction! Finally, we can do checkpoint
processing: any buffers committed as a result of this
transaction can be removed from any checkpoint list it was on
before. */
jbd_debug(3, "JBD: commit phase 7\n");
J_ASSERT(commit_transaction->t_sync_datalist == NULL);
J_ASSERT(commit_transaction->t_buffers == NULL);
J_ASSERT(commit_transaction->t_checkpoint_list == NULL);
J_ASSERT(commit_transaction->t_iobuf_list == NULL);
J_ASSERT(commit_transaction->t_shadow_list == NULL);
J_ASSERT(commit_transaction->t_log_list == NULL);
restart_loop:
/*
* As there are other places (journal_unmap_buffer()) adding buffers
* to this list we have to be careful and hold the j_list_lock.
*/
spin_lock(&journal->j_list_lock);
while (commit_transaction->t_forget) {
transaction_t *cp_transaction;
struct buffer_head *bh;
int try_to_free = 0;
jh = commit_transaction->t_forget;
spin_unlock(&journal->j_list_lock);
bh = jh2bh(jh);
/*
* Get a reference so that bh cannot be freed before we are
* done with it.
*/
get_bh(bh);
jbd_lock_bh_state(bh);
J_ASSERT_JH(jh, jh->b_transaction == commit_transaction ||
jh->b_transaction == journal->j_running_transaction);
/*
* If there is undo-protected committed data against
* this buffer, then we can remove it now. If it is a
* buffer needing such protection, the old frozen_data
* field now points to a committed version of the
* buffer, so rotate that field to the new committed
* data.
*
* Otherwise, we can just throw away the frozen data now.
*/
if (jh->b_committed_data) {
jbd_free(jh->b_committed_data, bh->b_size);
jh->b_committed_data = NULL;
if (jh->b_frozen_data) {
jh->b_committed_data = jh->b_frozen_data;
jh->b_frozen_data = NULL;
}
} else if (jh->b_frozen_data) {
jbd_free(jh->b_frozen_data, bh->b_size);
jh->b_frozen_data = NULL;
}
spin_lock(&journal->j_list_lock);
cp_transaction = jh->b_cp_transaction;
if (cp_transaction) {
JBUFFER_TRACE(jh, "remove from old cp transaction");
__journal_remove_checkpoint(jh);
}
/* Only re-checkpoint the buffer_head if it is marked
* dirty. If the buffer was added to the BJ_Forget list
* by journal_forget, it may no longer be dirty and
* there's no point in keeping a checkpoint record for
* it. */
/* A buffer which has been freed while still being
* journaled by a previous transaction may end up still
* being dirty here, but we want to avoid writing back
* that buffer in the future after the "add to orphan"
* operation been committed, That's not only a performance
* gain, it also stops aliasing problems if the buffer is
* left behind for writeback and gets reallocated for another
* use in a different page. */
if (buffer_freed(bh) && !jh->b_next_transaction) {
clear_buffer_freed(bh);
clear_buffer_jbddirty(bh);
}
if (buffer_jbddirty(bh)) {
JBUFFER_TRACE(jh, "add to new checkpointing trans");
__journal_insert_checkpoint(jh, commit_transaction);
if (is_journal_aborted(journal))
clear_buffer_jbddirty(bh);
} else {
J_ASSERT_BH(bh, !buffer_dirty(bh));
/*
* The buffer on BJ_Forget list and not jbddirty means
* it has been freed by this transaction and hence it
* could not have been reallocated until this
* transaction has committed. *BUT* it could be
* reallocated once we have written all the data to
* disk and before we process the buffer on BJ_Forget
* list.
*/
if (!jh->b_next_transaction)
try_to_free = 1;
}
JBUFFER_TRACE(jh, "refile or unfile freed buffer");
__journal_refile_buffer(jh);
jbd_unlock_bh_state(bh);
if (try_to_free)
release_buffer_page(bh);
else
__brelse(bh);
cond_resched_lock(&journal->j_list_lock);
}
spin_unlock(&journal->j_list_lock);
/*
* This is a bit sleazy. We use j_list_lock to protect transition
* of a transaction into T_FINISHED state and calling
* __journal_drop_transaction(). Otherwise we could race with
* other checkpointing code processing the transaction...
*/
spin_lock(&journal->j_state_lock);
spin_lock(&journal->j_list_lock);
/*
* Now recheck if some buffers did not get attached to the transaction
* while the lock was dropped...
*/
if (commit_transaction->t_forget) {
spin_unlock(&journal->j_list_lock);
spin_unlock(&journal->j_state_lock);
goto restart_loop;
}
/* Done with this transaction! */
jbd_debug(3, "JBD: commit phase 8\n");
J_ASSERT(commit_transaction->t_state == T_COMMIT_RECORD);
commit_transaction->t_state = T_FINISHED;
J_ASSERT(commit_transaction == journal->j_committing_transaction);
journal->j_commit_sequence = commit_transaction->t_tid;
journal->j_committing_transaction = NULL;
commit_time = ktime_to_ns(ktime_sub(ktime_get(), start_time));
/*
* weight the commit time higher than the average time so we don't
* react too strongly to vast changes in commit time
*/
if (likely(journal->j_average_commit_time))
journal->j_average_commit_time = (commit_time*3 +
journal->j_average_commit_time) / 4;
else
journal->j_average_commit_time = commit_time;
spin_unlock(&journal->j_state_lock);
if (commit_transaction->t_checkpoint_list == NULL &&
commit_transaction->t_checkpoint_io_list == NULL) {
__journal_drop_transaction(journal, commit_transaction);
} else {
if (journal->j_checkpoint_transactions == NULL) {
journal->j_checkpoint_transactions = commit_transaction;
commit_transaction->t_cpnext = commit_transaction;
commit_transaction->t_cpprev = commit_transaction;
} else {
commit_transaction->t_cpnext =
journal->j_checkpoint_transactions;
commit_transaction->t_cpprev =
commit_transaction->t_cpnext->t_cpprev;
commit_transaction->t_cpnext->t_cpprev =
commit_transaction;
commit_transaction->t_cpprev->t_cpnext =
commit_transaction;
}
}
spin_unlock(&journal->j_list_lock);
trace_jbd_end_commit(journal, commit_transaction);
jbd_debug(1, "JBD: commit %d complete, head %d\n",
journal->j_commit_sequence, journal->j_tail_sequence);
wake_up(&journal->j_wait_done_commit);
}
/*
* Submit all the data buffers to disk
*/
static int journal_submit_data_buffers(journal_t *journal,
transaction_t *commit_transaction,
int write_op)
{
struct journal_head *jh;
struct buffer_head *bh;
int locked;
int bufs = 0;
struct buffer_head **wbuf = journal->j_wbuf;
int err = 0;
/*
* Whenever we unlock the journal and sleep, things can get added
* onto ->t_sync_datalist, so we have to keep looping back to
* write_out_data until we *know* that the list is empty.
*
* Cleanup any flushed data buffers from the data list. Even in
* abort mode, we want to flush this out as soon as possible.
*/
write_out_data:
cond_resched();
spin_lock(&journal->j_list_lock);
while (commit_transaction->t_sync_datalist) {
jh = commit_transaction->t_sync_datalist;
bh = jh2bh(jh);
locked = 0;
/* Get reference just to make sure buffer does not disappear
* when we are forced to drop various locks */
get_bh(bh);
/* If the buffer is dirty, we need to submit IO and hence
* we need the buffer lock. We try to lock the buffer without
* blocking. If we fail, we need to drop j_list_lock and do
* blocking lock_buffer().
*/
if (buffer_dirty(bh)) {
if (!trylock_buffer(bh)) {
BUFFER_TRACE(bh, "needs blocking lock");
spin_unlock(&journal->j_list_lock);
trace_jbd_do_submit_data(journal,
commit_transaction);
/* Write out all data to prevent deadlocks */
journal_do_submit_data(wbuf, bufs, write_op);
bufs = 0;
lock_buffer(bh);
spin_lock(&journal->j_list_lock);
}
locked = 1;
}
/* We have to get bh_state lock. Again out of order, sigh. */
if (!inverted_lock(journal, bh)) {
jbd_lock_bh_state(bh);
spin_lock(&journal->j_list_lock);
}
/* Someone already cleaned up the buffer? */
if (!buffer_jbd(bh) || bh2jh(bh) != jh
|| jh->b_transaction != commit_transaction
|| jh->b_jlist != BJ_SyncData) {
jbd_unlock_bh_state(bh);
if (locked)
unlock_buffer(bh);
BUFFER_TRACE(bh, "already cleaned up");
release_data_buffer(bh);
continue;
}
if (locked && test_clear_buffer_dirty(bh)) {
BUFFER_TRACE(bh, "needs writeout, adding to array");
wbuf[bufs++] = bh;
__journal_file_buffer(jh, commit_transaction,
BJ_Locked);
jbd_unlock_bh_state(bh);
if (bufs == journal->j_wbufsize) {
spin_unlock(&journal->j_list_lock);
trace_jbd_do_submit_data(journal,
commit_transaction);
journal_do_submit_data(wbuf, bufs, write_op);
bufs = 0;
goto write_out_data;
}
} else if (!locked && buffer_locked(bh)) {
__journal_file_buffer(jh, commit_transaction,
BJ_Locked);
jbd_unlock_bh_state(bh);
put_bh(bh);
} else {
BUFFER_TRACE(bh, "writeout complete: unfile");
if (unlikely(!buffer_uptodate(bh)))
err = -EIO;
__journal_unfile_buffer(jh);
jbd_unlock_bh_state(bh);
if (locked)
unlock_buffer(bh);
release_data_buffer(bh);
}
if (need_resched() || spin_needbreak(&journal->j_list_lock)) {
spin_unlock(&journal->j_list_lock);
goto write_out_data;
}
}
spin_unlock(&journal->j_list_lock);
trace_jbd_do_submit_data(journal, commit_transaction);
journal_do_submit_data(wbuf, bufs, write_op);
return err;
}
static int journal_unmap_buffer(journal_t *journal, struct buffer_head *bh)
{
transaction_t *transaction;
struct journal_head *jh;
int may_free = 1;
int ret;
BUFFER_TRACE(bh, "entry");
/*
* It is safe to proceed here without the j_list_lock because the
* buffers cannot be stolen by try_to_free_buffers as long as we are
* holding the page lock. --sct
*/
if (!buffer_jbd(bh))
goto zap_buffer_unlocked;
spin_lock(&journal->j_state_lock);
jbd_lock_bh_state(bh);
spin_lock(&journal->j_list_lock);
jh = journal_grab_journal_head(bh);
if (!jh)
goto zap_buffer_no_jh;
/*
* We cannot remove the buffer from checkpoint lists until the
* transaction adding inode to orphan list (let's call it T)
* is committed. Otherwise if the transaction changing the
* buffer would be cleaned from the journal before T is
* committed, a crash will cause that the correct contents of
* the buffer will be lost. On the other hand we have to
* clear the buffer dirty bit at latest at the moment when the
* transaction marking the buffer as freed in the filesystem
* structures is committed because from that moment on the
* buffer can be reallocated and used by a different page.
* Since the block hasn't been freed yet but the inode has
* already been added to orphan list, it is safe for us to add
* the buffer to BJ_Forget list of the newest transaction.
*/
transaction = jh->b_transaction;
if (transaction == NULL) {
/* First case: not on any transaction. If it
* has no checkpoint link, then we can zap it:
* it's a writeback-mode buffer so we don't care
* if it hits disk safely. */
if (!jh->b_cp_transaction) {
JBUFFER_TRACE(jh, "not on any transaction: zap");
goto zap_buffer;
}
if (!buffer_dirty(bh)) {
/* bdflush has written it. We can drop it now */
goto zap_buffer;
}
/* OK, it must be in the journal but still not
* written fully to disk: it's metadata or
* journaled data... */
if (journal->j_running_transaction) {
/* ... and once the current transaction has
* committed, the buffer won't be needed any
* longer. */
JBUFFER_TRACE(jh, "checkpointed: add to BJ_Forget");
ret = __dispose_buffer(jh,
journal->j_running_transaction);
journal_put_journal_head(jh);
spin_unlock(&journal->j_list_lock);
jbd_unlock_bh_state(bh);
spin_unlock(&journal->j_state_lock);
return ret;
} else {
/* There is no currently-running transaction. So the
* orphan record which we wrote for this file must have
* passed into commit. We must attach this buffer to
* the committing transaction, if it exists. */
if (journal->j_committing_transaction) {
JBUFFER_TRACE(jh, "give to committing trans");
ret = __dispose_buffer(jh,
journal->j_committing_transaction);
journal_put_journal_head(jh);
spin_unlock(&journal->j_list_lock);
jbd_unlock_bh_state(bh);
spin_unlock(&journal->j_state_lock);
return ret;
} else {
/* The orphan record's transaction has
* committed. We can cleanse this buffer */
clear_buffer_jbddirty(bh);
goto zap_buffer;
}
}
} else if (transaction == journal->j_committing_transaction) {
JBUFFER_TRACE(jh, "on committing transaction");
if (jh->b_jlist == BJ_Locked) {
/*
* The buffer is on the committing transaction's locked
* list. We have the buffer locked, so I/O has
* completed. So we can nail the buffer now.
*/
may_free = __dispose_buffer(jh, transaction);
goto zap_buffer;
}
/*
* The buffer is committing, we simply cannot touch
* it. So we just set j_next_transaction to the
* running transaction (if there is one) and mark
* buffer as freed so that commit code knows it should
* clear dirty bits when it is done with the buffer.
*/
set_buffer_freed(bh);
if (journal->j_running_transaction && buffer_jbddirty(bh))
jh->b_next_transaction = journal->j_running_transaction;
journal_put_journal_head(jh);
spin_unlock(&journal->j_list_lock);
jbd_unlock_bh_state(bh);
spin_unlock(&journal->j_state_lock);
return 0;
} else {
/* Good, the buffer belongs to the running transaction.
* We are writing our own transaction's data, not any
* previous one's, so it is safe to throw it away
* (remember that we expect the filesystem to have set
* i_size already for this truncate so recovery will not
* expose the disk blocks we are discarding here.) */
J_ASSERT_JH(jh, transaction == journal->j_running_transaction);
JBUFFER_TRACE(jh, "on running transaction");
may_free = __dispose_buffer(jh, transaction);
}
zap_buffer:
journal_put_journal_head(jh);
zap_buffer_no_jh:
spin_unlock(&journal->j_list_lock);
jbd_unlock_bh_state(bh);
spin_unlock(&journal->j_state_lock);
zap_buffer_unlocked:
clear_buffer_dirty(bh);
J_ASSERT_BH(bh, !buffer_jbddirty(bh));
clear_buffer_mapped(bh);
clear_buffer_req(bh);
clear_buffer_new(bh);
bh->b_bdev = NULL;
return may_free;
}
int jbd2_journal_forget (handle_t *handle, struct buffer_head *bh)
{
transaction_t *transaction = handle->h_transaction;
journal_t *journal = transaction->t_journal;
struct journal_head *jh;
int drop_reserve = 0;
int err = 0;
int was_modified = 0;
BUFFER_TRACE(bh, "entry");
jbd_lock_bh_state(bh);
spin_lock(&journal->j_list_lock);
if (!buffer_jbd(bh))
goto not_jbd;
jh = bh2jh(bh);
if (!J_EXPECT_JH(jh, !jh->b_committed_data,
"inconsistent data on disk")) {
err = -EIO;
goto not_jbd;
}
was_modified = jh->b_modified;
jh->b_modified = 0;
if (jh->b_transaction == handle->h_transaction) {
J_ASSERT_JH(jh, !jh->b_frozen_data);
clear_buffer_dirty(bh);
clear_buffer_jbddirty(bh);
JBUFFER_TRACE(jh, "belongs to current transaction: unfile");
if (was_modified)
drop_reserve = 1;
if (jh->b_cp_transaction) {
__jbd2_journal_temp_unlink_buffer(jh);
__jbd2_journal_file_buffer(jh, transaction, BJ_Forget);
} else {
__jbd2_journal_unfile_buffer(jh);
if (!buffer_jbd(bh)) {
spin_unlock(&journal->j_list_lock);
jbd_unlock_bh_state(bh);
__bforget(bh);
goto drop;
}
}
} else if (jh->b_transaction) {
J_ASSERT_JH(jh, (jh->b_transaction ==
journal->j_committing_transaction));
JBUFFER_TRACE(jh, "belongs to older transaction");
if (jh->b_next_transaction) {
J_ASSERT(jh->b_next_transaction == transaction);
jh->b_next_transaction = NULL;
if (was_modified)
drop_reserve = 1;
}
}
not_jbd:
spin_unlock(&journal->j_list_lock);
jbd_unlock_bh_state(bh);
__brelse(bh);
drop:
if (drop_reserve) {
handle->h_buffer_credits++;
}
return err;
}
static int journal_unmap_buffer(journal_t *journal, struct buffer_head *bh)
{
transaction_t *transaction;
struct journal_head *jh;
int may_free = 1;
int ret;
BUFFER_TRACE(bh, "entry");
if (!buffer_jbd(bh))
goto zap_buffer_unlocked;
write_lock(&journal->j_state_lock);
jbd_lock_bh_state(bh);
spin_lock(&journal->j_list_lock);
jh = jbd2_journal_grab_journal_head(bh);
if (!jh)
goto zap_buffer_no_jh;
transaction = jh->b_transaction;
if (transaction == NULL) {
if (!jh->b_cp_transaction) {
JBUFFER_TRACE(jh, "not on any transaction: zap");
goto zap_buffer;
}
if (!buffer_dirty(bh)) {
goto zap_buffer;
}
if (journal->j_running_transaction) {
JBUFFER_TRACE(jh, "checkpointed: add to BJ_Forget");
ret = __dispose_buffer(jh,
journal->j_running_transaction);
jbd2_journal_put_journal_head(jh);
spin_unlock(&journal->j_list_lock);
jbd_unlock_bh_state(bh);
write_unlock(&journal->j_state_lock);
return ret;
} else {
if (journal->j_committing_transaction) {
JBUFFER_TRACE(jh, "give to committing trans");
ret = __dispose_buffer(jh,
journal->j_committing_transaction);
jbd2_journal_put_journal_head(jh);
spin_unlock(&journal->j_list_lock);
jbd_unlock_bh_state(bh);
write_unlock(&journal->j_state_lock);
return ret;
} else {
clear_buffer_jbddirty(bh);
goto zap_buffer;
}
}
} else if (transaction == journal->j_committing_transaction) {
JBUFFER_TRACE(jh, "on committing transaction");
set_buffer_freed(bh);
if (journal->j_running_transaction && buffer_jbddirty(bh))
jh->b_next_transaction = journal->j_running_transaction;
jbd2_journal_put_journal_head(jh);
spin_unlock(&journal->j_list_lock);
jbd_unlock_bh_state(bh);
write_unlock(&journal->j_state_lock);
return 0;
} else {
J_ASSERT_JH(jh, transaction == journal->j_running_transaction);
JBUFFER_TRACE(jh, "on running transaction");
may_free = __dispose_buffer(jh, transaction);
}
zap_buffer:
jbd2_journal_put_journal_head(jh);
zap_buffer_no_jh:
spin_unlock(&journal->j_list_lock);
jbd_unlock_bh_state(bh);
write_unlock(&journal->j_state_lock);
zap_buffer_unlocked:
clear_buffer_dirty(bh);
J_ASSERT_BH(bh, !buffer_jbddirty(bh));
clear_buffer_mapped(bh);
clear_buffer_req(bh);
clear_buffer_new(bh);
clear_buffer_delay(bh);
clear_buffer_unwritten(bh);
bh->b_bdev = NULL;
return may_free;
}
void buffer_trace(struct buffer_head *dest,
struct buffer_head *src, char *info)
{
struct buffer_history_item *bhist_i;
unsigned long flags;
if (dest == 0 || src == 0)
return;
spin_lock_irqsave(&trace_lock, flags);
/*
* Sometimes we don't initialise the ring pointers. (locally declared
* temp buffer_heads). Feebly attempt to detect and correct that here.
*/
if ((dest->b_history.b_history_head - dest->b_history.b_history_tail >
BUFFER_HISTORY_SIZE)) {
dest->b_history.b_history_head = 0;
dest->b_history.b_history_tail = 0;
}
bhist_i = dest->b_history.b +
(dest->b_history.b_history_head & (BUFFER_HISTORY_SIZE - 1));
bhist_i->info = info;
bhist_i->b_state = src->b_state;
bhist_i->b_list = src->b_list;
#if defined(CONFIG_JBD) || defined(CONFIG_JBD_MODULE)
bhist_i->b_trans_is_running = 0;
bhist_i->b_trans_is_committing = 0;
bhist_i->b_blocknr = src->b_blocknr;
if (buffer_jbd(src)) {
struct journal_head *jh;
journal_t *journal;
transaction_t *transaction;
/* Footwork to avoid racing with journal_remove_journal_head */
jh = src->b_private;
if (jh == 0)
goto raced;
transaction = jh->b_transaction;
if (src->b_private == 0)
goto raced;
bhist_i->b_jcount = jh->b_jcount;
bhist_i->b_jbd = 1;
bhist_i->b_jlist = jh->b_jlist;
bhist_i->b_frozen_data = jh->b_frozen_data;
bhist_i->b_committed_data = jh->b_committed_data;
bhist_i->b_transaction = !!jh->b_transaction;
bhist_i->b_next_transaction = !!jh->b_next_transaction;
bhist_i->b_cp_transaction = !!jh->b_cp_transaction;
if (transaction) {
journal = transaction->t_journal;
bhist_i->b_trans_is_running = transaction ==
journal->j_running_transaction;
bhist_i->b_trans_is_committing = transaction ==
journal->j_committing_transaction;
}
} else {
raced:
bhist_i->b_jcount = 0;
bhist_i->b_jbd = 0;
bhist_i->b_jlist = 0;
bhist_i->b_frozen_data = 0;
bhist_i->b_committed_data = 0;
bhist_i->b_transaction = 0;
bhist_i->b_next_transaction = 0;
bhist_i->b_cp_transaction = 0;
}
#endif /* defined(CONFIG_JBD) || defined(CONFIG_JBD_MODULE) */
bhist_i->on_lru = (src->b_prev_free != 0 && src->b_next_free != 0);
bhist_i->on_hash = (src->b_pprev != 0);
bhist_i->cpu = smp_processor_id();
bhist_i->b_count = atomic_read(&src->b_count);
dest->b_history.b_history_head++;
if (dest->b_history.b_history_head - dest->b_history.b_history_tail >
BUFFER_HISTORY_SIZE)
dest->b_history.b_history_tail =
dest->b_history.b_history_head - BUFFER_HISTORY_SIZE;
spin_unlock_irqrestore(&trace_lock, flags);
}
static int ocfs2_read_locked_inode(struct inode *inode,
struct ocfs2_find_inode_args *args)
{
struct super_block *sb;
struct ocfs2_super *osb;
struct ocfs2_dinode *fe;
struct buffer_head *bh = NULL;
int status, can_lock;
u32 generation = 0;
status = -EINVAL;
if (inode == NULL || inode->i_sb == NULL) {
mlog(ML_ERROR, "bad inode\n");
return status;
}
sb = inode->i_sb;
osb = OCFS2_SB(sb);
if (!args) {
mlog(ML_ERROR, "bad inode args\n");
make_bad_inode(inode);
return status;
}
/*
* To improve performance of cold-cache inode stats, we take
* the cluster lock here if possible.
*
* Generally, OCFS2 never trusts the contents of an inode
* unless it's holding a cluster lock, so taking it here isn't
* a correctness issue as much as it is a performance
* improvement.
*
* There are three times when taking the lock is not a good idea:
*
* 1) During startup, before we have initialized the DLM.
*
* 2) If we are reading certain system files which never get
* cluster locks (local alloc, truncate log).
*
* 3) If the process doing the iget() is responsible for
* orphan dir recovery. We're holding the orphan dir lock and
* can get into a deadlock with another process on another
* node in ->delete_inode().
*
* #1 and #2 can be simply solved by never taking the lock
* here for system files (which are the only type we read
* during mount). It's a heavier approach, but our main
* concern is user-accessible files anyway.
*
* #3 works itself out because we'll eventually take the
* cluster lock before trusting anything anyway.
*/
can_lock = !(args->fi_flags & OCFS2_FI_FLAG_SYSFILE)
&& !(args->fi_flags & OCFS2_FI_FLAG_ORPHAN_RECOVERY)
&& !ocfs2_mount_local(osb);
trace_ocfs2_read_locked_inode(
(unsigned long long)OCFS2_I(inode)->ip_blkno, can_lock);
/*
* To maintain backwards compatibility with older versions of
* ocfs2-tools, we still store the generation value for system
* files. The only ones that actually matter to userspace are
* the journals, but it's easier and inexpensive to just flag
* all system files similarly.
*/
if (args->fi_flags & OCFS2_FI_FLAG_SYSFILE)
generation = osb->fs_generation;
ocfs2_inode_lock_res_init(&OCFS2_I(inode)->ip_inode_lockres,
OCFS2_LOCK_TYPE_META,
generation, inode);
ocfs2_inode_lock_res_init(&OCFS2_I(inode)->ip_open_lockres,
OCFS2_LOCK_TYPE_OPEN,
0, inode);
if (can_lock) {
status = ocfs2_open_lock(inode);
if (status) {
make_bad_inode(inode);
mlog_errno(status);
return status;
}
status = ocfs2_inode_lock(inode, NULL, 0);
if (status) {
make_bad_inode(inode);
mlog_errno(status);
return status;
}
}
if (args->fi_flags & OCFS2_FI_FLAG_ORPHAN_RECOVERY) {
status = ocfs2_try_open_lock(inode, 0);
if (status) {
make_bad_inode(inode);
return status;
}
}
if (can_lock) {
status = ocfs2_read_inode_block_full(inode, &bh,
OCFS2_BH_IGNORE_CACHE);
} else {
status = ocfs2_read_blocks_sync(osb, args->fi_blkno, 1, &bh);
/*
* If buffer is in jbd, then its checksum may not have been
* computed as yet.
*/
if (!status && !buffer_jbd(bh))
status = ocfs2_validate_inode_block(osb->sb, bh);
}
if (status < 0) {
mlog_errno(status);
goto bail;
}
status = -EINVAL;
fe = (struct ocfs2_dinode *) bh->b_data;
/*
* This is a code bug. Right now the caller needs to
* understand whether it is asking for a system file inode or
* not so the proper lock names can be built.
*/
mlog_bug_on_msg(!!(fe->i_flags & cpu_to_le32(OCFS2_SYSTEM_FL)) !=
!!(args->fi_flags & OCFS2_FI_FLAG_SYSFILE),
"Inode %llu: system file state is ambigous\n",
(unsigned long long)args->fi_blkno);
if (S_ISCHR(le16_to_cpu(fe->i_mode)) ||
S_ISBLK(le16_to_cpu(fe->i_mode)))
inode->i_rdev = huge_decode_dev(le64_to_cpu(fe->id1.dev1.i_rdev));
ocfs2_populate_inode(inode, fe, 0);
BUG_ON(args->fi_blkno != le64_to_cpu(fe->i_blkno));
status = 0;
bail:
if (can_lock)
ocfs2_inode_unlock(inode, 0);
if (status < 0)
make_bad_inode(inode);
if (args && bh)
brelse(bh);
return status;
}
static int ocfs2_symlink_get_block(struct inode *inode, sector_t iblock,
struct buffer_head *bh_result, int create)
{
int err = -EIO;
int status;
struct ocfs2_dinode *fe = NULL;
struct buffer_head *bh = NULL;
struct buffer_head *buffer_cache_bh = NULL;
struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
void *kaddr;
trace_ocfs2_symlink_get_block(
(unsigned long long)OCFS2_I(inode)->ip_blkno,
(unsigned long long)iblock, bh_result, create);
BUG_ON(ocfs2_inode_is_fast_symlink(inode));
if ((iblock << inode->i_sb->s_blocksize_bits) > PATH_MAX + 1) {
mlog(ML_ERROR, "block offset > PATH_MAX: %llu",
(unsigned long long)iblock);
goto bail;
}
status = ocfs2_read_inode_block(inode, &bh);
if (status < 0) {
mlog_errno(status);
goto bail;
}
fe = (struct ocfs2_dinode *) bh->b_data;
if ((u64)iblock >= ocfs2_clusters_to_blocks(inode->i_sb,
le32_to_cpu(fe->i_clusters))) {
mlog(ML_ERROR, "block offset is outside the allocated size: "
"%llu\n", (unsigned long long)iblock);
goto bail;
}
if (!buffer_uptodate(bh_result) && ocfs2_inode_is_new(inode)) {
u64 blkno = le64_to_cpu(fe->id2.i_list.l_recs[0].e_blkno) +
iblock;
buffer_cache_bh = sb_getblk(osb->sb, blkno);
if (!buffer_cache_bh) {
mlog(ML_ERROR, "couldn't getblock for symlink!\n");
goto bail;
}
if (buffer_jbd(buffer_cache_bh)
&& ocfs2_inode_is_new(inode)) {
kaddr = kmap_atomic(bh_result->b_page);
if (!kaddr) {
mlog(ML_ERROR, "couldn't kmap!\n");
goto bail;
}
memcpy(kaddr + (bh_result->b_size * iblock),
buffer_cache_bh->b_data,
bh_result->b_size);
kunmap_atomic(kaddr);
set_buffer_uptodate(bh_result);
}
brelse(buffer_cache_bh);
}
map_bh(bh_result, inode->i_sb,
le64_to_cpu(fe->id2.i_list.l_recs[0].e_blkno) + iblock);
err = 0;
bail:
brelse(bh);
return err;
}
/*
* journal_commit_transaction
*
* The primary function for committing a transaction to the log. This
* function is called by the journal thread to begin a complete commit.
*/
void journal_commit_transaction(journal_t *journal)
{
transaction_t *commit_transaction;
struct journal_head *jh, *new_jh, *descriptor;
struct buffer_head **wbuf = journal->j_wbuf;
int bufs;
int flags;
int err;
unsigned int blocknr;
ktime_t start_time;
u64 commit_time;
char *tagp = NULL;
journal_header_t *header;
journal_block_tag_t *tag = NULL;
int space_left = 0;
int first_tag = 0;
int tag_flag;
int i;
struct blk_plug plug;
/*
* First job: lock down the current transaction and wait for
* all outstanding updates to complete.
*/
/* Do we need to erase the effects of a prior journal_flush? */
if (journal->j_flags & JFS_FLUSHED) {
jbd_debug(3, "super block updated\n");
journal_update_superblock(journal, 1);
} else {
jbd_debug(3, "superblock not updated\n");
}
J_ASSERT(journal->j_running_transaction != NULL);
J_ASSERT(journal->j_committing_transaction == NULL);
commit_transaction = journal->j_running_transaction;
J_ASSERT(commit_transaction->t_state == T_RUNNING);
trace_jbd_start_commit(journal, commit_transaction);
jbd_debug(1, "JBD: starting commit of transaction %d\n",
commit_transaction->t_tid);
spin_lock(&journal->j_state_lock);
commit_transaction->t_state = T_LOCKED;
trace_jbd_commit_locking(journal, commit_transaction);
spin_lock(&commit_transaction->t_handle_lock);
while (commit_transaction->t_updates) {
DEFINE_WAIT(wait);
prepare_to_wait(&journal->j_wait_updates, &wait,
TASK_UNINTERRUPTIBLE);
if (commit_transaction->t_updates) {
spin_unlock(&commit_transaction->t_handle_lock);
spin_unlock(&journal->j_state_lock);
schedule();
spin_lock(&journal->j_state_lock);
spin_lock(&commit_transaction->t_handle_lock);
}
finish_wait(&journal->j_wait_updates, &wait);
}
spin_unlock(&commit_transaction->t_handle_lock);
J_ASSERT (commit_transaction->t_outstanding_credits <=
journal->j_max_transaction_buffers);
/*
* First thing we are allowed to do is to discard any remaining
* BJ_Reserved buffers. Note, it is _not_ permissible to assume
* that there are no such buffers: if a large filesystem
* operation like a truncate needs to split itself over multiple
* transactions, then it may try to do a journal_restart() while
* there are still BJ_Reserved buffers outstanding. These must
* be released cleanly from the current transaction.
*
* In this case, the filesystem must still reserve write access
* again before modifying the buffer in the new transaction, but
* we do not require it to remember exactly which old buffers it
* has reserved. This is consistent with the existing behaviour
* that multiple journal_get_write_access() calls to the same
* buffer are perfectly permissible.
*/
while (commit_transaction->t_reserved_list) {
jh = commit_transaction->t_reserved_list;
JBUFFER_TRACE(jh, "reserved, unused: refile");
/*
* A journal_get_undo_access()+journal_release_buffer() may
* leave undo-committed data.
*/
if (jh->b_committed_data) {
struct buffer_head *bh = jh2bh(jh);
jbd_lock_bh_state(bh);
jbd_free(jh->b_committed_data, bh->b_size);
jh->b_committed_data = NULL;
jbd_unlock_bh_state(bh);
}
journal_refile_buffer(journal, jh);
}
/*
* Now try to drop any written-back buffers from the journal's
* checkpoint lists. We do this *before* commit because it potentially
* frees some memory
*/
spin_lock(&journal->j_list_lock);
__journal_clean_checkpoint_list(journal);
spin_unlock(&journal->j_list_lock);
jbd_debug (3, "JBD: commit phase 1\n");
/*
* Switch to a new revoke table.
*/
journal_switch_revoke_table(journal);
trace_jbd_commit_flushing(journal, commit_transaction);
commit_transaction->t_state = T_FLUSH;
journal->j_committing_transaction = commit_transaction;
journal->j_running_transaction = NULL;
start_time = ktime_get();
commit_transaction->t_log_start = journal->j_head;
wake_up(&journal->j_wait_transaction_locked);
spin_unlock(&journal->j_state_lock);
jbd_debug (3, "JBD: commit phase 2\n");
/*
* Now start flushing things to disk, in the order they appear
* on the transaction lists. Data blocks go first.
*/
blk_start_plug(&plug);
err = journal_submit_data_buffers(journal, commit_transaction,
WRITE_SYNC);
blk_finish_plug(&plug);
/*
* Wait for all previously submitted IO to complete.
*/
spin_lock(&journal->j_list_lock);
while (commit_transaction->t_locked_list) {
struct buffer_head *bh;
jh = commit_transaction->t_locked_list->b_tprev;
bh = jh2bh(jh);
get_bh(bh);
if (buffer_locked(bh)) {
spin_unlock(&journal->j_list_lock);
wait_on_buffer(bh);
spin_lock(&journal->j_list_lock);
}
if (unlikely(!buffer_uptodate(bh))) {
if (!trylock_page(bh->b_page)) {
spin_unlock(&journal->j_list_lock);
lock_page(bh->b_page);
spin_lock(&journal->j_list_lock);
}
if (bh->b_page->mapping)
set_bit(AS_EIO, &bh->b_page->mapping->flags);
unlock_page(bh->b_page);
SetPageError(bh->b_page);
err = -EIO;
}
if (!inverted_lock(journal, bh)) {
put_bh(bh);
spin_lock(&journal->j_list_lock);
continue;
}
if (buffer_jbd(bh) && bh2jh(bh) == jh &&
jh->b_transaction == commit_transaction &&
jh->b_jlist == BJ_Locked)
__journal_unfile_buffer(jh);
jbd_unlock_bh_state(bh);
release_data_buffer(bh);
cond_resched_lock(&journal->j_list_lock);
}
spin_unlock(&journal->j_list_lock);
if (err) {
char b[BDEVNAME_SIZE];
printk(KERN_WARNING
"JBD: Detected IO errors while flushing file data "
"on %s\n", bdevname(journal->j_fs_dev, b));
if (journal->j_flags & JFS_ABORT_ON_SYNCDATA_ERR)
journal_abort(journal, err);
err = 0;
}
blk_start_plug(&plug);
journal_write_revoke_records(journal, commit_transaction, WRITE_SYNC);
/*
* If we found any dirty or locked buffers, then we should have
* looped back up to the write_out_data label. If there weren't
* any then journal_clean_data_list should have wiped the list
* clean by now, so check that it is in fact empty.
*/
J_ASSERT (commit_transaction->t_sync_datalist == NULL);
jbd_debug (3, "JBD: commit phase 3\n");
/*
* Way to go: we have now written out all of the data for a
* transaction! Now comes the tricky part: we need to write out
* metadata. Loop over the transaction's entire buffer list:
*/
spin_lock(&journal->j_state_lock);
commit_transaction->t_state = T_COMMIT;
spin_unlock(&journal->j_state_lock);
trace_jbd_commit_logging(journal, commit_transaction);
J_ASSERT(commit_transaction->t_nr_buffers <=
commit_transaction->t_outstanding_credits);
descriptor = NULL;
bufs = 0;
while (commit_transaction->t_buffers) {
/* Find the next buffer to be journaled... */
jh = commit_transaction->t_buffers;
/* If we're in abort mode, we just un-journal the buffer and
release it. */
if (is_journal_aborted(journal)) {
clear_buffer_jbddirty(jh2bh(jh));
JBUFFER_TRACE(jh, "journal is aborting: refile");
journal_refile_buffer(journal, jh);
/* If that was the last one, we need to clean up
* any descriptor buffers which may have been
* already allocated, even if we are now
* aborting. */
if (!commit_transaction->t_buffers)
goto start_journal_io;
continue;
}
/* Make sure we have a descriptor block in which to
record the metadata buffer. */
if (!descriptor) {
struct buffer_head *bh;
J_ASSERT (bufs == 0);
jbd_debug(4, "JBD: get descriptor\n");
descriptor = journal_get_descriptor_buffer(journal);
if (!descriptor) {
journal_abort(journal, -EIO);
continue;
}
bh = jh2bh(descriptor);
jbd_debug(4, "JBD: got buffer %llu (%p)\n",
(unsigned long long)bh->b_blocknr, bh->b_data);
header = (journal_header_t *)&bh->b_data[0];
header->h_magic = cpu_to_be32(JFS_MAGIC_NUMBER);
header->h_blocktype = cpu_to_be32(JFS_DESCRIPTOR_BLOCK);
header->h_sequence = cpu_to_be32(commit_transaction->t_tid);
tagp = &bh->b_data[sizeof(journal_header_t)];
space_left = bh->b_size - sizeof(journal_header_t);
first_tag = 1;
set_buffer_jwrite(bh);
set_buffer_dirty(bh);
wbuf[bufs++] = bh;
/* Record it so that we can wait for IO
completion later */
BUFFER_TRACE(bh, "ph3: file as descriptor");
journal_file_buffer(descriptor, commit_transaction,
BJ_LogCtl);
}
/* Where is the buffer to be written? */
err = journal_next_log_block(journal, &blocknr);
/* If the block mapping failed, just abandon the buffer
and repeat this loop: we'll fall into the
refile-on-abort condition above. */
if (err) {
journal_abort(journal, err);
continue;
}
/*
static int ocfs2_symlink_get_block(struct inode *inode, sector_t iblock,
struct buffer_head *bh_result, int create)
{
int err = -EIO;
int status;
struct ocfs2_dinode *fe = NULL;
struct buffer_head *bh = NULL;
struct buffer_head *buffer_cache_bh = NULL;
struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
void *kaddr;
mlog_entry("(0x%p, %llu, 0x%p, %d)\n", inode,
(unsigned long long)iblock, bh_result, create);
BUG_ON(ocfs2_inode_is_fast_symlink(inode));
if ((iblock << inode->i_sb->s_blocksize_bits) > PATH_MAX + 1) {
mlog(ML_ERROR, "block offset > PATH_MAX: %llu",
(unsigned long long)iblock);
goto bail;
}
status = ocfs2_read_inode_block(inode, &bh);
if (status < 0) {
mlog_errno(status);
goto bail;
}
fe = (struct ocfs2_dinode *) bh->b_data;
if ((u64)iblock >= ocfs2_clusters_to_blocks(inode->i_sb,
le32_to_cpu(fe->i_clusters))) {
mlog(ML_ERROR, "block offset is outside the allocated size: "
"%llu\n", (unsigned long long)iblock);
goto bail;
}
/* We don't use the page cache to create symlink data, so if
* need be, copy it over from the buffer cache. */
if (!buffer_uptodate(bh_result) && ocfs2_inode_is_new(inode)) {
u64 blkno = le64_to_cpu(fe->id2.i_list.l_recs[0].e_blkno) +
iblock;
buffer_cache_bh = sb_getblk(osb->sb, blkno);
if (!buffer_cache_bh) {
mlog(ML_ERROR, "couldn't getblock for symlink!\n");
goto bail;
}
/* we haven't locked out transactions, so a commit
* could've happened. Since we've got a reference on
* the bh, even if it commits while we're doing the
* copy, the data is still good. */
if (buffer_jbd(buffer_cache_bh)
&& ocfs2_inode_is_new(inode)) {
kaddr = kmap_atomic(bh_result->b_page, KM_USER0);
if (!kaddr) {
mlog(ML_ERROR, "couldn't kmap!\n");
goto bail;
}
memcpy(kaddr + (bh_result->b_size * iblock),
buffer_cache_bh->b_data,
bh_result->b_size);
kunmap_atomic(kaddr, KM_USER0);
set_buffer_uptodate(bh_result);
}
brelse(buffer_cache_bh);
}
map_bh(bh_result, inode->i_sb,
le64_to_cpu(fe->id2.i_list.l_recs[0].e_blkno) + iblock);
err = 0;
bail:
brelse(bh);
mlog_exit(err);
return err;
}
int journal_forget (handle_t *handle, struct buffer_head *bh)
{
transaction_t *transaction = handle->h_transaction;
journal_t *journal = transaction->t_journal;
struct journal_head *jh;
int drop_reserve = 0;
int err = 0;
int was_modified = 0;
BUFFER_TRACE(bh, "entry");
jbd_lock_bh_state(bh);
spin_lock(&journal->j_list_lock);
if (!buffer_jbd(bh))
goto not_jbd;
jh = bh2jh(bh);
/* Critical error: attempting to delete a bitmap buffer, maybe?
* Don't do any jbd operations, and return an error. */
if (!J_EXPECT_JH(jh, !jh->b_committed_data,
"inconsistent data on disk")) {
err = -EIO;
goto not_jbd;
}
/* keep track of wether or not this transaction modified us */
was_modified = jh->b_modified;
/*
* The buffer's going from the transaction, we must drop
* all references -bzzz
*/
jh->b_modified = 0;
if (jh->b_transaction == handle->h_transaction) {
J_ASSERT_JH(jh, !jh->b_frozen_data);
/* If we are forgetting a buffer which is already part
* of this transaction, then we can just drop it from
* the transaction immediately. */
clear_buffer_dirty(bh);
clear_buffer_jbddirty(bh);
JBUFFER_TRACE(jh, "belongs to current transaction: unfile");
/*
* we only want to drop a reference if this transaction
* modified the buffer
*/
if (was_modified)
drop_reserve = 1;
/*
* We are no longer going to journal this buffer.
* However, the commit of this transaction is still
* important to the buffer: the delete that we are now
* processing might obsolete an old log entry, so by
* committing, we can satisfy the buffer's checkpoint.
*
* So, if we have a checkpoint on the buffer, we should
* now refile the buffer on our BJ_Forget list so that
* we know to remove the checkpoint after we commit.
*/
if (jh->b_cp_transaction) {
__journal_temp_unlink_buffer(jh);
__journal_file_buffer(jh, transaction, BJ_Forget);
} else {
__journal_unfile_buffer(jh);
journal_remove_journal_head(bh);
__brelse(bh);
if (!buffer_jbd(bh)) {
spin_unlock(&journal->j_list_lock);
jbd_unlock_bh_state(bh);
__bforget(bh);
goto drop;
}
}
} else if (jh->b_transaction) {
J_ASSERT_JH(jh, (jh->b_transaction ==
journal->j_committing_transaction));
/* However, if the buffer is still owned by a prior
* (committing) transaction, we can't drop it yet... */
JBUFFER_TRACE(jh, "belongs to older transaction");
/* ... but we CAN drop it from the new transaction if we
* have also modified it since the original commit. */
if (jh->b_next_transaction) {
J_ASSERT(jh->b_next_transaction == transaction);
jh->b_next_transaction = NULL;
/*
* only drop a reference if this transaction modified
* the buffer
*/
if (was_modified)
drop_reserve = 1;
}
}
not_jbd:
spin_unlock(&journal->j_list_lock);
jbd_unlock_bh_state(bh);
__brelse(bh);
drop:
if (drop_reserve) {
/* no need to reserve log space for this block -bzzz */
handle->h_buffer_credits++;
}
return err;
}
static int ocfs2_filecheck_repair_inode_block(struct super_block *sb,
struct buffer_head *bh)
{
int changed = 0;
struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
if (!ocfs2_filecheck_validate_inode_block(sb, bh))
return 0;
trace_ocfs2_filecheck_repair_inode_block(
(unsigned long long)bh->b_blocknr);
if (ocfs2_is_hard_readonly(OCFS2_SB(sb)) ||
ocfs2_is_soft_readonly(OCFS2_SB(sb))) {
mlog(ML_ERROR,
"Filecheck: cannot repair dinode #%llu "
"on readonly filesystem\n",
(unsigned long long)bh->b_blocknr);
return -OCFS2_FILECHECK_ERR_READONLY;
}
if (buffer_jbd(bh)) {
mlog(ML_ERROR,
"Filecheck: cannot repair dinode #%llu, "
"its buffer is in jbd\n",
(unsigned long long)bh->b_blocknr);
return -OCFS2_FILECHECK_ERR_INJBD;
}
if (!OCFS2_IS_VALID_DINODE(di)) {
/* Cannot fix invalid inode block */
return -OCFS2_FILECHECK_ERR_INVALIDINO;
}
if (!(di->i_flags & cpu_to_le32(OCFS2_VALID_FL))) {
/* Cannot just add VALID_FL flag back as a fix,
* need more things to check here.
*/
return -OCFS2_FILECHECK_ERR_VALIDFLAG;
}
if (le64_to_cpu(di->i_blkno) != bh->b_blocknr) {
di->i_blkno = cpu_to_le64(bh->b_blocknr);
changed = 1;
mlog(ML_ERROR,
"Filecheck: reset dinode #%llu: i_blkno to %llu\n",
(unsigned long long)bh->b_blocknr,
(unsigned long long)le64_to_cpu(di->i_blkno));
}
if (le32_to_cpu(di->i_fs_generation) !=
OCFS2_SB(sb)->fs_generation) {
di->i_fs_generation = cpu_to_le32(OCFS2_SB(sb)->fs_generation);
changed = 1;
mlog(ML_ERROR,
"Filecheck: reset dinode #%llu: fs_generation to %u\n",
(unsigned long long)bh->b_blocknr,
le32_to_cpu(di->i_fs_generation));
}
if (changed || ocfs2_validate_meta_ecc(sb, bh->b_data, &di->i_check)) {
ocfs2_compute_meta_ecc(sb, bh->b_data, &di->i_check);
mark_buffer_dirty(bh);
mlog(ML_ERROR,
"Filecheck: reset dinode #%llu: compute meta ecc\n",
(unsigned long long)bh->b_blocknr);
}
return 0;
}
int jbd2_journal_dirty_metadata(handle_t *handle, struct buffer_head *bh)
{
transaction_t *transaction = handle->h_transaction;
journal_t *journal = transaction->t_journal;
struct journal_head *jh = bh2jh(bh);
int ret = 0;
jbd_debug(5, "journal_head %p\n", jh);
JBUFFER_TRACE(jh, "entry");
if (is_handle_aborted(handle))
goto out;
if (!buffer_jbd(bh)) {
ret = -EUCLEAN;
goto out;
}
jbd_lock_bh_state(bh);
if (jh->b_modified == 0) {
jh->b_modified = 1;
J_ASSERT_JH(jh, handle->h_buffer_credits > 0);
handle->h_buffer_credits--;
}
if (jh->b_transaction == transaction && jh->b_jlist == BJ_Metadata) {
JBUFFER_TRACE(jh, "fastpath");
if (unlikely(jh->b_transaction !=
journal->j_running_transaction)) {
printk(KERN_EMERG "JBD: %s: "
"jh->b_transaction (%llu, %p, %u) != "
"journal->j_running_transaction (%p, %u)",
journal->j_devname,
(unsigned long long) bh->b_blocknr,
jh->b_transaction,
jh->b_transaction ? jh->b_transaction->t_tid : 0,
journal->j_running_transaction,
journal->j_running_transaction ?
journal->j_running_transaction->t_tid : 0);
ret = -EINVAL;
}
goto out_unlock_bh;
}
set_buffer_jbddirty(bh);
if (jh->b_transaction != transaction) {
JBUFFER_TRACE(jh, "already on other transaction");
if (unlikely(jh->b_transaction !=
journal->j_committing_transaction)) {
printk(KERN_EMERG "JBD: %s: "
"jh->b_transaction (%llu, %p, %u) != "
"journal->j_committing_transaction (%p, %u)",
journal->j_devname,
(unsigned long long) bh->b_blocknr,
jh->b_transaction,
jh->b_transaction ? jh->b_transaction->t_tid : 0,
journal->j_committing_transaction,
journal->j_committing_transaction ?
journal->j_committing_transaction->t_tid : 0);
ret = -EINVAL;
}
if (unlikely(jh->b_next_transaction != transaction)) {
printk(KERN_EMERG "JBD: %s: "
"jh->b_next_transaction (%llu, %p, %u) != "
"transaction (%p, %u)",
journal->j_devname,
(unsigned long long) bh->b_blocknr,
jh->b_next_transaction,
jh->b_next_transaction ?
jh->b_next_transaction->t_tid : 0,
transaction, transaction->t_tid);
ret = -EINVAL;
}
goto out_unlock_bh;
}
J_ASSERT_JH(jh, jh->b_frozen_data == NULL);
JBUFFER_TRACE(jh, "file as BJ_Metadata");
spin_lock(&journal->j_list_lock);
__jbd2_journal_file_buffer(jh, handle->h_transaction, BJ_Metadata);
spin_unlock(&journal->j_list_lock);
out_unlock_bh:
jbd_unlock_bh_state(bh);
out:
JBUFFER_TRACE(jh, "exit");
WARN_ON(ret);
return ret;
}