/**
* write_mft_record_nolock - write out a mapped (extent) mft record
* @ni: ntfs inode describing the mapped (extent) mft record
* @m: mapped (extent) mft record to write
* @sync: if true, wait for i/o completion
*
* Write the mapped (extent) mft record @m described by the (regular or extent)
* ntfs inode @ni to backing store. If the mft record @m has a counterpart in
* the mft mirror, that is also updated.
*
* On success, clean the mft record and return 0. On error, leave the mft
* record dirty and return -errno. The caller should call make_bad_inode() on
* the base inode to ensure no more access happens to this inode. We do not do
* it here as the caller may want to finish writing other extent mft records
* first to minimize on-disk metadata inconsistencies.
*
* NOTE: We always perform synchronous i/o and ignore the @sync parameter.
* However, if the mft record has a counterpart in the mft mirror and @sync is
* true, we write the mft record, wait for i/o completion, and only then write
* the mft mirror copy. This ensures that if the system crashes either the mft
* or the mft mirror will contain a self-consistent mft record @m. If @sync is
* false on the other hand, we start i/o on both and then wait for completion
* on them. This provides a speedup but no longer guarantees that you will end
* up with a self-consistent mft record in the case of a crash but if you asked
* for asynchronous writing you probably do not care about that anyway.
*
* TODO: If @sync is false, want to do truly asynchronous i/o, i.e. just
* schedule i/o via ->writepage or do it via kntfsd or whatever.
*/
int write_mft_record_nolock(ntfs_inode *ni, MFT_RECORD *m, int sync)
{
ntfs_volume *vol = ni->vol;
struct page *page = ni->page;
unsigned int blocksize = vol->sb->s_blocksize;
int max_bhs = vol->mft_record_size / blocksize;
struct buffer_head *bhs[max_bhs];
struct buffer_head *bh, *head;
unsigned int block_start, block_end, m_start, m_end;
int i_bhs, nr_bhs, err = 0;
ntfs_debug("Entering for inode 0x%lx.", ni->mft_no);
BUG_ON(NInoAttr(ni));
BUG_ON(!max_bhs);
BUG_ON(!PageLocked(page));
/*
* If the ntfs_inode is clean no need to do anything. If it is dirty,
* mark it as clean now so that it can be redirtied later on if needed.
* There is no danger of races since the caller is holding the locks
* for the mft record @m and the page it is in.
*/
if (!NInoTestClearDirty(ni))
goto done;
/* Make sure we have mapped buffers. */
if (!page_has_buffers(page)) {
no_buffers_err_out:
ntfs_error(vol->sb, "Writing mft records without existing "
"buffers is not implemented yet. %s",
ntfs_please_email);
err = -EOPNOTSUPP;
goto err_out;
}
bh = head = page_buffers(page);
if (!bh)
goto no_buffers_err_out;
nr_bhs = 0;
block_start = 0;
m_start = ni->page_ofs;
m_end = m_start + vol->mft_record_size;
do {
block_end = block_start + blocksize;
/*
* If the buffer is outside the mft record, just skip it,
* clearing it if it is dirty to make sure it is not written
* out. It should never be marked dirty but better be safe.
*/
if ((block_end <= m_start) || (block_start >= m_end)) {
if (buffer_dirty(bh)) {
ntfs_warning(vol->sb, "Clearing dirty mft "
"record page buffer. %s",
ntfs_please_email);
clear_buffer_dirty(bh);
}
continue;
}
if (!buffer_mapped(bh)) {
ntfs_error(vol->sb, "Writing mft records without "
"existing mapped buffers is not "
"implemented yet. %s",
ntfs_please_email);
err = -EOPNOTSUPP;
continue;
}
if (!buffer_uptodate(bh)) {
ntfs_error(vol->sb, "Writing mft records without "
"existing uptodate buffers is not "
"implemented yet. %s",
ntfs_please_email);
err = -EOPNOTSUPP;
continue;
}
BUG_ON(!nr_bhs && (m_start != block_start));
BUG_ON(nr_bhs >= max_bhs);
bhs[nr_bhs++] = bh;
BUG_ON((nr_bhs >= max_bhs) && (m_end != block_end));
} while (block_start = block_end, (bh = bh->b_this_page) != head);
if (unlikely(err))
goto cleanup_out;
/* Apply the mst protection fixups. */
err = pre_write_mst_fixup((NTFS_RECORD*)m, vol->mft_record_size);
if (err) {
ntfs_error(vol->sb, "Failed to apply mst fixups!");
goto cleanup_out;
}
flush_dcache_mft_record_page(ni);
/* Lock buffers and start synchronous write i/o on them. */
for (i_bhs = 0; i_bhs < nr_bhs; i_bhs++) {
struct buffer_head *tbh = bhs[i_bhs];
if (unlikely(test_set_buffer_locked(tbh)))
BUG();
BUG_ON(!buffer_uptodate(tbh));
if (buffer_dirty(tbh))
clear_buffer_dirty(tbh);
get_bh(tbh);
tbh->b_end_io = end_buffer_write_sync;
submit_bh(WRITE, tbh);
}
/* Synchronize the mft mirror now if not @sync. */
if (!sync && ni->mft_no < vol->mftmirr_size)
sync_mft_mirror(ni, m, sync);
/* Wait on i/o completion of buffers. */
for (i_bhs = 0; i_bhs < nr_bhs; i_bhs++) {
struct buffer_head *tbh = bhs[i_bhs];
wait_on_buffer(tbh);
if (unlikely(!buffer_uptodate(tbh))) {
err = -EIO;
/*
* Set the buffer uptodate so the page & buffer states
* don't become out of sync.
*/
if (PageUptodate(page))
set_buffer_uptodate(tbh);
}
}
/* If @sync, now synchronize the mft mirror. */
if (sync && ni->mft_no < vol->mftmirr_size)
sync_mft_mirror(ni, m, sync);
/* Remove the mst protection fixups again. */
post_write_mst_fixup((NTFS_RECORD*)m);
flush_dcache_mft_record_page(ni);
if (unlikely(err)) {
/* I/O error during writing. This is really bad! */
ntfs_error(vol->sb, "I/O error while writing mft record "
"0x%lx! Marking base inode as bad. You "
"should unmount the volume and run chkdsk.",
ni->mft_no);
goto err_out;
}
done:
ntfs_debug("Done.");
return 0;
cleanup_out:
/* Clean the buffers. */
for (i_bhs = 0; i_bhs < nr_bhs; i_bhs++)
clear_buffer_dirty(bhs[i_bhs]);
err_out:
/*
* Current state: all buffers are clean, unlocked, and uptodate.
* The caller should mark the base inode as bad so that no more i/o
* happens. ->clear_inode() will still be invoked so all extent inodes
* and other allocated memory will be freed.
*/
if (err == -ENOMEM) {
ntfs_error(vol->sb, "Not enough memory to write mft record. "
"Redirtying so the write is retried later.");
mark_mft_record_dirty(ni);
err = 0;
}
return err;
}
/*
* Submit all the data buffers to disk
*/
static void journal_submit_data_buffers(journal_t *journal,
transaction_t *commit_transaction)
{
struct journal_head *jh;
struct buffer_head *bh;
int locked;
int bufs = 0;
struct buffer_head **wbuf = journal->j_wbuf;
/*
* 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 (test_set_buffer_locked(bh)) {
BUFFER_TRACE(bh, "needs blocking lock");
spin_unlock(&journal->j_list_lock);
/* Write out all data to prevent deadlocks */
journal_do_submit_data(wbuf, bufs);
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)
|| 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");
put_bh(bh);
continue;
}
if (locked && test_clear_buffer_dirty(bh)) {
BUFFER_TRACE(bh, "needs writeout, adding to array");
wbuf[bufs++] = bh;
__jbd2_journal_file_buffer(jh, commit_transaction,
BJ_Locked);
jbd_unlock_bh_state(bh);
if (bufs == journal->j_wbufsize) {
spin_unlock(&journal->j_list_lock);
journal_do_submit_data(wbuf, bufs);
bufs = 0;
goto write_out_data;
}
} else if (!locked && buffer_locked(bh)) {
__jbd2_journal_file_buffer(jh, commit_transaction,
BJ_Locked);
jbd_unlock_bh_state(bh);
put_bh(bh);
} else {
BUFFER_TRACE(bh, "writeout complete: unfile");
__jbd2_journal_unfile_buffer(jh);
jbd_unlock_bh_state(bh);
if (locked)
unlock_buffer(bh);
jbd2_journal_remove_journal_head(bh);
/* Once for our safety reference, once for
* jbd2_journal_remove_journal_head() */
put_bh(bh);
put_bh(bh);
}
if (lock_need_resched(&journal->j_list_lock)) {
spin_unlock(&journal->j_list_lock);
goto write_out_data;
}
}
spin_unlock(&journal->j_list_lock);
journal_do_submit_data(wbuf, bufs);
}
/**
* sync_mft_mirror - synchronize an mft record to the mft mirror
* @ni: ntfs inode whose mft record to synchronize
* @m: mapped, mst protected (extent) mft record to synchronize
* @sync: if true, wait for i/o completion
*
* Write the mapped, mst protected (extent) mft record @m described by the
* (regular or extent) ntfs inode @ni to the mft mirror ($MFTMirr).
*
* On success return 0. On error return -errno and set the volume errors flag
* in the ntfs_volume to which @ni belongs.
*
* NOTE: We always perform synchronous i/o and ignore the @sync parameter.
*
* TODO: If @sync is false, want to do truly asynchronous i/o, i.e. just
* schedule i/o via ->writepage or do it via kntfsd or whatever.
*/
static int sync_mft_mirror(ntfs_inode *ni, MFT_RECORD *m, int sync)
{
ntfs_volume *vol = ni->vol;
struct page *page;
unsigned int blocksize = vol->sb->s_blocksize;
int max_bhs = vol->mft_record_size / blocksize;
struct buffer_head *bhs[max_bhs];
struct buffer_head *bh, *head;
u8 *kmirr;
unsigned int block_start, block_end, m_start, m_end;
int i_bhs, nr_bhs, err = 0;
ntfs_debug("Entering for inode 0x%lx.", ni->mft_no);
BUG_ON(!max_bhs);
if (unlikely(!vol->mftmirr_ino)) {
/* This could happen during umount... */
err = sync_mft_mirror_umount(ni, m);
if (likely(!err))
return err;
goto err_out;
}
/* Get the page containing the mirror copy of the mft record @m. */
page = ntfs_map_page(vol->mftmirr_ino->i_mapping, ni->mft_no >>
(PAGE_CACHE_SHIFT - vol->mft_record_size_bits));
if (unlikely(IS_ERR(page))) {
ntfs_error(vol->sb, "Failed to map mft mirror page.");
err = PTR_ERR(page);
goto err_out;
}
/*
* Exclusion against other writers. This should never be a problem
* since the page in which the mft record @m resides is also locked and
* hence any other writers would be held up there but it is better to
* make sure no one is writing from elsewhere.
*/
lock_page(page);
/* The address in the page of the mirror copy of the mft record @m. */
kmirr = page_address(page) + ((ni->mft_no << vol->mft_record_size_bits)
& ~PAGE_CACHE_MASK);
/* Copy the mst protected mft record to the mirror. */
memcpy(kmirr, m, vol->mft_record_size);
/* Make sure we have mapped buffers. */
if (!page_has_buffers(page)) {
no_buffers_err_out:
ntfs_error(vol->sb, "Writing mft mirror records without "
"existing buffers is not implemented yet. %s",
ntfs_please_email);
err = -EOPNOTSUPP;
goto unlock_err_out;
}
bh = head = page_buffers(page);
if (!bh)
goto no_buffers_err_out;
nr_bhs = 0;
block_start = 0;
m_start = kmirr - (u8*)page_address(page);
m_end = m_start + vol->mft_record_size;
do {
block_end = block_start + blocksize;
/*
* If the buffer is outside the mft record, just skip it,
* clearing it if it is dirty to make sure it is not written
* out. It should never be marked dirty but better be safe.
*/
if ((block_end <= m_start) || (block_start >= m_end)) {
if (buffer_dirty(bh)) {
ntfs_warning(vol->sb, "Clearing dirty mft "
"record page buffer. %s",
ntfs_please_email);
clear_buffer_dirty(bh);
}
continue;
}
if (!buffer_mapped(bh)) {
ntfs_error(vol->sb, "Writing mft mirror records "
"without existing mapped buffers is "
"not implemented yet. %s",
ntfs_please_email);
err = -EOPNOTSUPP;
continue;
}
if (!buffer_uptodate(bh)) {
ntfs_error(vol->sb, "Writing mft mirror records "
"without existing uptodate buffers is "
"not implemented yet. %s",
ntfs_please_email);
err = -EOPNOTSUPP;
continue;
}
BUG_ON(!nr_bhs && (m_start != block_start));
BUG_ON(nr_bhs >= max_bhs);
bhs[nr_bhs++] = bh;
BUG_ON((nr_bhs >= max_bhs) && (m_end != block_end));
} while (block_start = block_end, (bh = bh->b_this_page) != head);
if (likely(!err)) {
/* Lock buffers and start synchronous write i/o on them. */
for (i_bhs = 0; i_bhs < nr_bhs; i_bhs++) {
struct buffer_head *tbh = bhs[i_bhs];
if (unlikely(test_set_buffer_locked(tbh)))
BUG();
BUG_ON(!buffer_uptodate(tbh));
if (buffer_dirty(tbh))
clear_buffer_dirty(tbh);
get_bh(tbh);
tbh->b_end_io = end_buffer_write_sync;
submit_bh(WRITE, tbh);
}
/* Wait on i/o completion of buffers. */
for (i_bhs = 0; i_bhs < nr_bhs; i_bhs++) {
struct buffer_head *tbh = bhs[i_bhs];
wait_on_buffer(tbh);
if (unlikely(!buffer_uptodate(tbh))) {
err = -EIO;
/*
* Set the buffer uptodate so the page & buffer
* states don't become out of sync.
*/
if (PageUptodate(page))
set_buffer_uptodate(tbh);
}
}
} else /* if (unlikely(err)) */ {
/* Clean the buffers. */
for (i_bhs = 0; i_bhs < nr_bhs; i_bhs++)
clear_buffer_dirty(bhs[i_bhs]);
}
unlock_err_out:
/* Current state: all buffers are clean, unlocked, and uptodate. */
/* Remove the mst protection fixups again. */
post_write_mst_fixup((NTFS_RECORD*)kmirr);
flush_dcache_page(page);
unlock_page(page);
ntfs_unmap_page(page);
if (unlikely(err)) {
/* I/O error during writing. This is really bad! */
ntfs_error(vol->sb, "I/O error while writing mft mirror "
"record 0x%lx! You should unmount the volume "
"and run chkdsk or ntfsfix.", ni->mft_no);
goto err_out;
}
ntfs_debug("Done.");
return 0;
err_out:
ntfs_error(vol->sb, "Failed to synchronize $MFTMirr (error code %i). "
"Volume will be left marked dirty on umount. Run "
"ntfsfix on the partition after umounting to correct "
"this.", -err);
/* We don't want to clear the dirty bit on umount. */
NVolSetErrors(vol);
return err;
}
