static void ocfs2_delete_inode(struct inode *inode)
{
int wipe, status;
sigset_t oldset;
struct buffer_head *di_bh = NULL;
trace_ocfs2_delete_inode(inode->i_ino,
(unsigned long long)OCFS2_I(inode)->ip_blkno,
is_bad_inode(inode));
/* When we fail in read_inode() we mark inode as bad. The second test
* catches the case when inode allocation fails before allocating
* a block for inode. */
if (is_bad_inode(inode) || !OCFS2_I(inode)->ip_blkno)
goto bail;
dquot_initialize(inode);
if (!ocfs2_inode_is_valid_to_delete(inode)) {
/* It's probably not necessary to truncate_inode_pages
* here but we do it for safety anyway (it will most
* likely be a no-op anyway) */
ocfs2_cleanup_delete_inode(inode, 0);
goto bail;
}
/* We want to block signals in delete_inode as the lock and
* messaging paths may return us -ERESTARTSYS. Which would
* cause us to exit early, resulting in inodes being orphaned
* forever. */
ocfs2_block_signals(&oldset);
/*
* Synchronize us against ocfs2_get_dentry. We take this in
* shared mode so that all nodes can still concurrently
* process deletes.
*/
status = ocfs2_nfs_sync_lock(OCFS2_SB(inode->i_sb), 0);
if (status < 0) {
mlog(ML_ERROR, "getting nfs sync lock(PR) failed %d\n", status);
ocfs2_cleanup_delete_inode(inode, 0);
goto bail_unblock;
}
/* Lock down the inode. This gives us an up to date view of
* it's metadata (for verification), and allows us to
* serialize delete_inode on multiple nodes.
*
* Even though we might be doing a truncate, we don't take the
* allocation lock here as it won't be needed - nobody will
* have the file open.
*/
status = ocfs2_inode_lock(inode, &di_bh, 1);
if (status < 0) {
if (status != -ENOENT)
mlog_errno(status);
ocfs2_cleanup_delete_inode(inode, 0);
goto bail_unlock_nfs_sync;
}
/* Query the cluster. This will be the final decision made
* before we go ahead and wipe the inode. */
status = ocfs2_query_inode_wipe(inode, di_bh, &wipe);
if (!wipe || status < 0) {
/* Error and remote inode busy both mean we won't be
* removing the inode, so they take almost the same
* path. */
if (status < 0)
mlog_errno(status);
/* Someone in the cluster has disallowed a wipe of
* this inode, or it was never completely
* orphaned. Write out the pages and exit now. */
ocfs2_cleanup_delete_inode(inode, 1);
goto bail_unlock_inode;
}
ocfs2_cleanup_delete_inode(inode, 0);
status = ocfs2_wipe_inode(inode, di_bh);
if (status < 0) {
if (status != -EDEADLK)
mlog_errno(status);
goto bail_unlock_inode;
}
/*
* Mark the inode as successfully deleted.
*
* This is important for ocfs2_clear_inode() as it will check
* this flag and skip any checkpointing work
*
* ocfs2_stuff_meta_lvb() also uses this flag to invalidate
* the LVB for other nodes.
*/
OCFS2_I(inode)->ip_flags |= OCFS2_INODE_DELETED;
bail_unlock_inode:
ocfs2_inode_unlock(inode, 1);
brelse(di_bh);
bail_unlock_nfs_sync:
ocfs2_nfs_sync_unlock(OCFS2_SB(inode->i_sb), 0);
bail_unblock:
ocfs2_unblock_signals(&oldset);
bail:
return;
}
void ocfs2_delete_inode(struct inode *inode)
{
int wipe, status;
sigset_t blocked, oldset;
struct buffer_head *di_bh = NULL;
mlog_entry("(inode->i_ino = %lu)\n", inode->i_ino);
if (is_bad_inode(inode)) {
mlog(0, "Skipping delete of bad inode\n");
goto bail;
}
if (!ocfs2_inode_is_valid_to_delete(inode)) {
/* It's probably not necessary to truncate_inode_pages
* here but we do it for safety anyway (it will most
* likely be a no-op anyway) */
ocfs2_cleanup_delete_inode(inode, 0);
goto bail;
}
/* We want to block signals in delete_inode as the lock and
* messaging paths may return us -ERESTARTSYS. Which would
* cause us to exit early, resulting in inodes being orphaned
* forever. */
sigfillset(&blocked);
status = sigprocmask(SIG_BLOCK, &blocked, &oldset);
if (status < 0) {
mlog_errno(status);
ocfs2_cleanup_delete_inode(inode, 1);
goto bail;
}
/* Lock down the inode. This gives us an up to date view of
* it's metadata (for verification), and allows us to
* serialize delete_inode votes.
*
* Even though we might be doing a truncate, we don't take the
* allocation lock here as it won't be needed - nobody will
* have the file open.
*/
status = ocfs2_meta_lock(inode, &di_bh, 1);
if (status < 0) {
if (status != -ENOENT)
mlog_errno(status);
ocfs2_cleanup_delete_inode(inode, 0);
goto bail_unblock;
}
/* Query the cluster. This will be the final decision made
* before we go ahead and wipe the inode. */
status = ocfs2_query_inode_wipe(inode, di_bh, &wipe);
if (!wipe || status < 0) {
/* Error and inode busy vote both mean we won't be
* removing the inode, so they take almost the same
* path. */
if (status < 0)
mlog_errno(status);
/* Someone in the cluster has voted to not wipe this
* inode, or it was never completely orphaned. Write
* out the pages and exit now. */
ocfs2_cleanup_delete_inode(inode, 1);
goto bail_unlock_inode;
}
ocfs2_cleanup_delete_inode(inode, 0);
status = ocfs2_wipe_inode(inode, di_bh);
if (status < 0) {
if (status != -EDEADLK)
mlog_errno(status);
goto bail_unlock_inode;
}
/*
* Mark the inode as successfully deleted.
*
* This is important for ocfs2_clear_inode() as it will check
* this flag and skip any checkpointing work
*
* ocfs2_stuff_meta_lvb() also uses this flag to invalidate
* the LVB for other nodes.
*/
OCFS2_I(inode)->ip_flags |= OCFS2_INODE_DELETED;
bail_unlock_inode:
ocfs2_meta_unlock(inode, 1);
brelse(di_bh);
bail_unblock:
status = sigprocmask(SIG_SETMASK, &oldset, NULL);
if (status < 0)
mlog_errno(status);
bail:
clear_inode(inode);
mlog_exit_void();
}
/* Query the cluster to determine whether we should wipe an inode from
* disk or not.
*
* Requires the inode to have the cluster lock. */
static int ocfs2_query_inode_wipe(struct inode *inode,
struct buffer_head *di_bh,
int *wipe)
{
int status = 0, reason = 0;
struct ocfs2_inode_info *oi = OCFS2_I(inode);
struct ocfs2_dinode *di;
*wipe = 0;
trace_ocfs2_query_inode_wipe_begin((unsigned long long)oi->ip_blkno,
inode->i_nlink);
/* While we were waiting for the cluster lock in
* ocfs2_delete_inode, another node might have asked to delete
* the inode. Recheck our flags to catch this. */
if (!ocfs2_inode_is_valid_to_delete(inode)) {
reason = 1;
goto bail;
}
/* Now that we have an up to date inode, we can double check
* the link count. */
if (inode->i_nlink)
goto bail;
/* Do some basic inode verification... */
di = (struct ocfs2_dinode *) di_bh->b_data;
if (!(di->i_flags & cpu_to_le32(OCFS2_ORPHANED_FL)) &&
!(oi->ip_flags & OCFS2_INODE_SKIP_ORPHAN_DIR)) {
/*
* Inodes in the orphan dir must have ORPHANED_FL. The only
* inodes that come back out of the orphan dir are reflink
* targets. A reflink target may be moved out of the orphan
* dir between the time we scan the directory and the time we
* process it. This would lead to HAS_REFCOUNT_FL being set but
* ORPHANED_FL not.
*/
if (di->i_dyn_features & cpu_to_le16(OCFS2_HAS_REFCOUNT_FL)) {
reason = 2;
goto bail;
}
/* for lack of a better error? */
status = -EEXIST;
mlog(ML_ERROR,
"Inode %llu (on-disk %llu) not orphaned! "
"Disk flags 0x%x, inode flags 0x%x\n",
(unsigned long long)oi->ip_blkno,
(unsigned long long)le64_to_cpu(di->i_blkno),
le32_to_cpu(di->i_flags), oi->ip_flags);
goto bail;
}
/* has someone already deleted us?! baaad... */
if (di->i_dtime) {
status = -EEXIST;
mlog_errno(status);
goto bail;
}
/*
* This is how ocfs2 determines whether an inode is still live
* within the cluster. Every node takes a shared read lock on
* the inode open lock in ocfs2_read_locked_inode(). When we
* get to ->delete_inode(), each node tries to convert it's
* lock to an exclusive. Trylocks are serialized by the inode
* meta data lock. If the upconvert succeeds, we know the inode
* is no longer live and can be deleted.
*
* Though we call this with the meta data lock held, the
* trylock keeps us from ABBA deadlock.
*/
status = ocfs2_try_open_lock(inode, 1);
if (status == -EAGAIN) {
status = 0;
reason = 3;
goto bail;
}
if (status < 0) {
mlog_errno(status);
goto bail;
}
*wipe = 1;
trace_ocfs2_query_inode_wipe_succ(le16_to_cpu(di->i_orphaned_slot));
bail:
trace_ocfs2_query_inode_wipe_end(status, reason);
return status;
}
/* Query the cluster to determine whether we should wipe an inode from
* disk or not.
*
* Requires the inode to have the cluster lock. */
static int ocfs2_query_inode_wipe(struct inode *inode,
struct buffer_head *di_bh,
int *wipe)
{
int status = 0;
struct ocfs2_inode_info *oi = OCFS2_I(inode);
struct ocfs2_dinode *di;
*wipe = 0;
/* While we were waiting for the cluster lock in
* ocfs2_delete_inode, another node might have asked to delete
* the inode. Recheck our flags to catch this. */
if (!ocfs2_inode_is_valid_to_delete(inode)) {
mlog(0, "Skipping delete of %llu because flags changed\n",
(unsigned long long)oi->ip_blkno);
goto bail;
}
/* Now that we have an up to date inode, we can double check
* the link count. */
if (inode->i_nlink) {
mlog(0, "Skipping delete of %llu because nlink = %u\n",
(unsigned long long)oi->ip_blkno, inode->i_nlink);
goto bail;
}
/* Do some basic inode verification... */
di = (struct ocfs2_dinode *) di_bh->b_data;
if (!(di->i_flags & cpu_to_le32(OCFS2_ORPHANED_FL))) {
/* for lack of a better error? */
status = -EEXIST;
mlog(ML_ERROR,
"Inode %llu (on-disk %llu) not orphaned! "
"Disk flags 0x%x, inode flags 0x%x\n",
(unsigned long long)oi->ip_blkno,
(unsigned long long)le64_to_cpu(di->i_blkno),
le32_to_cpu(di->i_flags), oi->ip_flags);
goto bail;
}
/* has someone already deleted us?! baaad... */
if (di->i_dtime) {
status = -EEXIST;
mlog_errno(status);
goto bail;
}
/*
* This is how ocfs2 determines whether an inode is still live
* within the cluster. Every node takes a shared read lock on
* the inode open lock in ocfs2_read_locked_inode(). When we
* get to ->delete_inode(), each node tries to convert it's
* lock to an exclusive. Trylocks are serialized by the inode
* meta data lock. If the upconvert suceeds, we know the inode
* is no longer live and can be deleted.
*
* Though we call this with the meta data lock held, the
* trylock keeps us from ABBA deadlock.
*/
status = ocfs2_try_open_lock(inode, 1);
if (status == -EAGAIN) {
status = 0;
mlog(0, "Skipping delete of %llu because it is in use on "
"other nodes\n", (unsigned long long)oi->ip_blkno);
goto bail;
}
if (status < 0) {
mlog_errno(status);
goto bail;
}
*wipe = 1;
mlog(0, "Inode %llu is ok to wipe from orphan dir %u\n",
(unsigned long long)oi->ip_blkno,
le16_to_cpu(di->i_orphaned_slot));
bail:
return status;
}
