#include "buffer_head_io.h"
-#define OCFS2_FI_FLAG_NOWAIT 0x1
-#define OCFS2_FI_FLAG_DELETE 0x2
struct ocfs2_find_inode_args
{
u64 fi_blkno;
struct inode *inode,
struct buffer_head *fe_bh);
+void ocfs2_set_inode_flags(struct inode *inode)
+{
+ unsigned int flags = OCFS2_I(inode)->ip_attr;
+
+ inode->i_flags &= ~(S_IMMUTABLE |
+ S_SYNC | S_APPEND | S_NOATIME | S_DIRSYNC);
+
+ if (flags & OCFS2_IMMUTABLE_FL)
+ inode->i_flags |= S_IMMUTABLE;
+
+ if (flags & OCFS2_SYNC_FL)
+ inode->i_flags |= S_SYNC;
+ if (flags & OCFS2_APPEND_FL)
+ inode->i_flags |= S_APPEND;
+ if (flags & OCFS2_NOATIME_FL)
+ inode->i_flags |= S_NOATIME;
+ if (flags & OCFS2_DIRSYNC_FL)
+ inode->i_flags |= S_DIRSYNC;
+}
+
struct inode *ocfs2_ilookup_for_vote(struct ocfs2_super *osb,
u64 blkno,
int delete_vote)
return ilookup5(osb->sb, args.fi_ino, ocfs2_find_actor, &args);
}
-struct inode *ocfs2_iget(struct ocfs2_super *osb, u64 blkno)
+struct inode *ocfs2_iget(struct ocfs2_super *osb, u64 blkno, int flags)
{
struct inode *inode = NULL;
struct super_block *sb = osb->sb;
struct ocfs2_find_inode_args args;
- mlog_entry("(blkno = %"MLFu64")\n", blkno);
+ mlog_entry("(blkno = %llu)\n", (unsigned long long)blkno);
/* Ok. By now we've either got the offsets passed to us by the
* caller, or we just pulled them off the bh. Lets do some
}
args.fi_blkno = blkno;
- args.fi_flags = 0;
+ args.fi_flags = flags;
args.fi_ino = ino_from_blkno(sb, blkno);
inode = iget5_locked(sb, args.fi_ino, ocfs2_find_actor,
bail:
if (!IS_ERR(inode)) {
- mlog(0, "returning inode with number %"MLFu64"\n",
- OCFS2_I(inode)->ip_blkno);
+ mlog(0, "returning inode with number %llu\n",
+ (unsigned long long)OCFS2_I(inode)->ip_blkno);
mlog_exit_ptr(inode);
} else
mlog_errno(PTR_ERR(inode));
struct ocfs2_super *osb;
int status = -EINVAL;
- mlog_entry("(0x%p, size:%"MLFu64")\n", inode, fe->i_size);
+ mlog_entry("(0x%p, size:%llu)\n", inode,
+ (unsigned long long)fe->i_size);
sb = inode->i_sb;
osb = OCFS2_SB(sb);
* today. change if needed. */
if (!OCFS2_IS_VALID_DINODE(fe) ||
!(fe->i_flags & cpu_to_le32(OCFS2_VALID_FL))) {
- mlog(ML_ERROR, "Invalid dinode: i_ino=%lu, i_blkno=%"MLFu64", "
+ mlog(ML_ERROR, "Invalid dinode: i_ino=%lu, i_blkno=%llu, "
"signature = %.*s, flags = 0x%x\n",
- inode->i_ino, le64_to_cpu(fe->i_blkno), 7,
+ inode->i_ino,
+ (unsigned long long)le64_to_cpu(fe->i_blkno), 7,
fe->i_signature, le32_to_cpu(fe->i_flags));
goto bail;
}
inode->i_mode = le16_to_cpu(fe->i_mode);
inode->i_uid = le32_to_cpu(fe->i_uid);
inode->i_gid = le32_to_cpu(fe->i_gid);
- inode->i_blksize = (u32)osb->s_clustersize;
/* Fast symlinks will have i_size but no allocated clusters. */
if (S_ISLNK(inode->i_mode) && !fe->i_clusters)
inode->i_blocks =
ocfs2_align_bytes_to_sectors(le64_to_cpu(fe->i_size));
inode->i_mapping->a_ops = &ocfs2_aops;
- inode->i_flags |= S_NOATIME;
inode->i_atime.tv_sec = le64_to_cpu(fe->i_atime);
inode->i_atime.tv_nsec = le32_to_cpu(fe->i_atime_nsec);
inode->i_mtime.tv_sec = le64_to_cpu(fe->i_mtime);
if (OCFS2_I(inode)->ip_blkno != le64_to_cpu(fe->i_blkno))
mlog(ML_ERROR,
- "ip_blkno %"MLFu64" != i_blkno %"MLFu64"!\n",
- OCFS2_I(inode)->ip_blkno, fe->i_blkno);
+ "ip_blkno %llu != i_blkno %llu!\n",
+ (unsigned long long)OCFS2_I(inode)->ip_blkno,
+ (unsigned long long)fe->i_blkno);
OCFS2_I(inode)->ip_clusters = le32_to_cpu(fe->i_clusters);
OCFS2_I(inode)->ip_orphaned_slot = OCFS2_INVALID_SLOT;
-
- if (create_ino)
- inode->i_ino = ino_from_blkno(inode->i_sb,
- le64_to_cpu(fe->i_blkno));
-
- mlog(0, "blkno = %"MLFu64", ino = %lu, create_ino = %s\n",
- fe->i_blkno, inode->i_ino, create_ino ? "true" : "false");
+ OCFS2_I(inode)->ip_attr = le32_to_cpu(fe->i_attr);
inode->i_nlink = le16_to_cpu(fe->i_links_count);
+ if (fe->i_flags & cpu_to_le32(OCFS2_SYSTEM_FL))
+ OCFS2_I(inode)->ip_flags |= OCFS2_INODE_SYSTEM_FILE;
+
if (fe->i_flags & cpu_to_le32(OCFS2_LOCAL_ALLOC_FL)) {
OCFS2_I(inode)->ip_flags |= OCFS2_INODE_BITMAP;
mlog(0, "local alloc inode: i_ino=%lu\n", inode->i_ino);
break;
}
+ if (create_ino) {
+ inode->i_ino = ino_from_blkno(inode->i_sb,
+ le64_to_cpu(fe->i_blkno));
+
+ /*
+ * If we ever want to create system files from kernel,
+ * the generation argument to
+ * ocfs2_inode_lock_res_init() will have to change.
+ */
+ BUG_ON(fe->i_flags & cpu_to_le32(OCFS2_SYSTEM_FL));
+
+ ocfs2_inode_lock_res_init(&OCFS2_I(inode)->ip_meta_lockres,
+ OCFS2_LOCK_TYPE_META, 0, inode);
+ }
+
ocfs2_inode_lock_res_init(&OCFS2_I(inode)->ip_rw_lockres,
- OCFS2_LOCK_TYPE_RW, inode);
- ocfs2_inode_lock_res_init(&OCFS2_I(inode)->ip_meta_lockres,
- OCFS2_LOCK_TYPE_META, inode);
+ OCFS2_LOCK_TYPE_RW, inode->i_generation,
+ inode);
+
ocfs2_inode_lock_res_init(&OCFS2_I(inode)->ip_data_lockres,
- OCFS2_LOCK_TYPE_DATA, inode);
+ OCFS2_LOCK_TYPE_DATA, inode->i_generation,
+ inode);
+
+ ocfs2_set_inode_flags(inode);
+ inode->i_flags |= S_NOATIME;
status = 0;
bail:
struct ocfs2_super *osb;
struct ocfs2_dinode *fe;
struct buffer_head *bh = NULL;
- int status;
- int sysfile = 0;
+ int status, can_lock;
+ u32 generation = 0;
mlog_entry("(0x%p, 0x%p)\n", inode, args);
status = -EINVAL;
if (inode == NULL || inode->i_sb == NULL) {
mlog(ML_ERROR, "bad inode\n");
- goto bail;
+ return status;
}
sb = inode->i_sb;
osb = OCFS2_SB(sb);
if (!args) {
mlog(ML_ERROR, "bad inode args\n");
make_bad_inode(inode);
- goto bail;
+ 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-accesible 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_NOLOCK);
+
+ /*
+ * 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_meta_lockres,
+ OCFS2_LOCK_TYPE_META,
+ generation, inode);
+
+ if (can_lock) {
+ status = ocfs2_meta_lock(inode, NULL, NULL, 0);
+ if (status) {
+ make_bad_inode(inode);
+ mlog_errno(status);
+ return status;
+ }
}
- /* Read the FE off disk. This is safe because the kernel only
- * does one read_inode2 for a new inode, and if it doesn't
- * exist yet then nobody can be working on it! */
- status = ocfs2_read_block(osb, args->fi_blkno, &bh, 0, NULL);
+ status = ocfs2_read_block(osb, args->fi_blkno, &bh, 0,
+ can_lock ? inode : NULL);
if (status < 0) {
mlog_errno(status);
- make_bad_inode(inode);
goto bail;
}
+ status = -EINVAL;
fe = (struct ocfs2_dinode *) bh->b_data;
if (!OCFS2_IS_VALID_DINODE(fe)) {
- mlog(ML_ERROR, "Invalid dinode #%"MLFu64": signature = %.*s\n",
- fe->i_blkno, 7, fe->i_signature);
- make_bad_inode(inode);
+ mlog(ML_ERROR, "Invalid dinode #%llu: signature = %.*s\n",
+ (unsigned long long)fe->i_blkno, 7, fe->i_signature);
goto bail;
}
- if (fe->i_flags & cpu_to_le32(OCFS2_SYSTEM_FL))
- sysfile = 1;
+ /*
+ * 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));
- status = -EINVAL;
if (ocfs2_populate_inode(inode, fe, 0) < 0) {
- mlog(ML_ERROR, "populate inode failed! i_blkno=%"MLFu64", "
- "i_ino=%lu\n", fe->i_blkno, inode->i_ino);
- make_bad_inode(inode);
+ mlog(ML_ERROR, "populate failed! i_blkno=%llu, i_ino=%lu\n",
+ (unsigned long long)fe->i_blkno, inode->i_ino);
goto bail;
}
BUG_ON(args->fi_blkno != le64_to_cpu(fe->i_blkno));
- if (sysfile)
- OCFS2_I(inode)->ip_flags |= OCFS2_INODE_SYSTEM_FILE;
-
status = 0;
bail:
+ if (can_lock)
+ ocfs2_meta_unlock(inode, 0);
+
+ if (status < 0)
+ make_bad_inode(inode);
+
if (args && bh)
brelse(bh);
* never get here as system file inodes should always have a
* positive link count. */
if (oi->ip_flags & OCFS2_INODE_SYSTEM_FILE) {
- mlog(ML_ERROR, "Skipping delete of system file %"MLFu64".\n",
- oi->ip_blkno);
+ mlog(ML_ERROR, "Skipping delete of system file %llu\n",
+ (unsigned long long)oi->ip_blkno);
goto bail_unlock;
}
* 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 %"MLFu64" because flags changed\n",
- oi->ip_blkno);
+ 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 %"MLFu64" because nlink = %u\n",
- oi->ip_blkno, inode->i_nlink);
+ mlog(0, "Skipping delete of %llu because nlink = %u\n",
+ (unsigned long long)oi->ip_blkno, inode->i_nlink);
goto bail;
}
/* for lack of a better error? */
status = -EEXIST;
mlog(ML_ERROR,
- "Inode %"MLFu64" (on-disk %"MLFu64") not orphaned! "
+ "Inode %llu (on-disk %llu) not orphaned! "
"Disk flags 0x%x, inode flags 0x%x\n",
- oi->ip_blkno, di->i_blkno, di->i_flags, oi->ip_flags);
+ (unsigned long long)oi->ip_blkno,
+ (unsigned long long)di->i_blkno, di->i_flags,
+ oi->ip_flags);
goto bail;
}
* disk and let them worry about deleting it. */
if (status == -EBUSY) {
status = 0;
- mlog(0, "Skipping delete of %"MLFu64" because it is in use on"
- "other nodes\n", oi->ip_blkno);
+ 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) {
* into. This may happen during node death and
* recovery knows how to clean it up so we can safely
* ignore this inode for now on. */
- mlog(0, "Nobody knew where inode %"MLFu64" was orphaned!\n",
- oi->ip_blkno);
+ mlog(0, "Nobody knew where inode %llu was orphaned!\n",
+ (unsigned long long)oi->ip_blkno);
} else {
*wipe = 1;
- mlog(0, "Inode %"MLFu64" is ok to wipe from orphan dir %d\n",
- oi->ip_blkno, oi->ip_orphaned_slot);
+ mlog(0, "Inode %llu is ok to wipe from orphan dir %d\n",
+ (unsigned long long)oi->ip_blkno, oi->ip_orphaned_slot);
}
spin_unlock(&oi->ip_lock);
static void ocfs2_cleanup_delete_inode(struct inode *inode,
int sync_data)
{
- mlog(0, "Cleanup inode %"MLFu64", sync = %d\n",
- OCFS2_I(inode)->ip_blkno, sync_data);
+ mlog(0, "Cleanup inode %llu, sync = %d\n",
+ (unsigned long long)OCFS2_I(inode)->ip_blkno, sync_data);
if (sync_data)
write_inode_now(inode, 1);
truncate_inode_pages(&inode->i_data, 0);
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 */
+ /*
+ * 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:
if (!inode)
goto bail;
- mlog(0, "Clearing inode: %"MLFu64", nlink = %u\n",
- OCFS2_I(inode)->ip_blkno, inode->i_nlink);
+ mlog(0, "Clearing inode: %llu, nlink = %u\n",
+ (unsigned long long)OCFS2_I(inode)->ip_blkno, inode->i_nlink);
mlog_bug_on_msg(OCFS2_SB(inode->i_sb) == NULL,
"Inode=%lu\n", inode->i_ino);
ocfs2_checkpoint_inode(inode);
mlog_bug_on_msg(!list_empty(&oi->ip_io_markers),
- "Clear inode of %"MLFu64", inode has io markers\n",
- oi->ip_blkno);
+ "Clear inode of %llu, inode has io markers\n",
+ (unsigned long long)oi->ip_blkno);
ocfs2_extent_map_drop(inode, 0);
ocfs2_extent_map_init(inode);
ocfs2_metadata_cache_purge(inode);
mlog_bug_on_msg(oi->ip_metadata_cache.ci_num_cached,
- "Clear inode of %"MLFu64", inode has %u cache items\n",
- oi->ip_blkno, oi->ip_metadata_cache.ci_num_cached);
+ "Clear inode of %llu, inode has %u cache items\n",
+ (unsigned long long)oi->ip_blkno, oi->ip_metadata_cache.ci_num_cached);
mlog_bug_on_msg(!(oi->ip_flags & OCFS2_INODE_CACHE_INLINE),
- "Clear inode of %"MLFu64", inode has a bad flag\n",
- oi->ip_blkno);
+ "Clear inode of %llu, inode has a bad flag\n",
+ (unsigned long long)oi->ip_blkno);
mlog_bug_on_msg(spin_is_locked(&oi->ip_lock),
- "Clear inode of %"MLFu64", inode is locked\n",
- oi->ip_blkno);
+ "Clear inode of %llu, inode is locked\n",
+ (unsigned long long)oi->ip_blkno);
mlog_bug_on_msg(!mutex_trylock(&oi->ip_io_mutex),
- "Clear inode of %"MLFu64", io_mutex is locked\n",
- oi->ip_blkno);
+ "Clear inode of %llu, io_mutex is locked\n",
+ (unsigned long long)oi->ip_blkno);
mutex_unlock(&oi->ip_io_mutex);
/*
* kernel 1, world 0
*/
mlog_bug_on_msg(!down_write_trylock(&oi->ip_alloc_sem),
- "Clear inode of %"MLFu64", alloc_sem is locked\n",
- oi->ip_blkno);
+ "Clear inode of %llu, alloc_sem is locked\n",
+ (unsigned long long)oi->ip_blkno);
up_write(&oi->ip_alloc_sem);
mlog_bug_on_msg(oi->ip_open_count,
- "Clear inode of %"MLFu64" has open count %d\n",
- oi->ip_blkno, oi->ip_open_count);
+ "Clear inode of %llu has open count %d\n",
+ (unsigned long long)oi->ip_blkno, oi->ip_open_count);
mlog_bug_on_msg(!list_empty(&oi->ip_handle_list),
- "Clear inode of %"MLFu64" has non empty handle list\n",
- oi->ip_blkno);
+ "Clear inode of %llu has non empty handle list\n",
+ (unsigned long long)oi->ip_blkno);
mlog_bug_on_msg(oi->ip_handle,
- "Clear inode of %"MLFu64" has non empty handle pointer\n",
- oi->ip_blkno);
+ "Clear inode of %llu has non empty handle pointer\n",
+ (unsigned long long)oi->ip_blkno);
/* Clear all other flags. */
oi->ip_flags = OCFS2_INODE_CACHE_INLINE;
mlog_entry_void();
- mlog(0, "Drop inode %"MLFu64", nlink = %u, ip_flags = 0x%x\n",
- oi->ip_blkno, inode->i_nlink, oi->ip_flags);
+ mlog(0, "Drop inode %llu, nlink = %u, ip_flags = 0x%x\n",
+ (unsigned long long)oi->ip_blkno, inode->i_nlink, oi->ip_flags);
/* Testing ip_orphaned_slot here wouldn't work because we may
* not have gotten a delete_inode vote from any other nodes
* yet. */
- if (oi->ip_flags & OCFS2_INODE_MAYBE_ORPHANED) {
- mlog(0, "Inode was orphaned on another node, clearing nlink.\n");
- inode->i_nlink = 0;
- }
-
- generic_drop_inode(inode);
+ if (oi->ip_flags & OCFS2_INODE_MAYBE_ORPHANED)
+ generic_delete_inode(inode);
+ else
+ generic_drop_inode(inode);
mlog_exit_void();
}
u64 p_blkno;
int readflags = OCFS2_BH_CACHED;
-#if 0
- /* only turn this on if we know we can deal with read_block
- * returning nothing */
if (reada)
readflags |= OCFS2_BH_READAHEAD;
-#endif
if (((u64)block << inode->i_sb->s_blocksize_bits) >=
i_size_read(inode)) {
struct inode *inode = dentry->d_inode;
int status = 0;
- mlog_entry("(inode = 0x%p, ino = %"MLFu64")\n", inode,
- inode ? OCFS2_I(inode)->ip_blkno : 0ULL);
+ mlog_entry("(inode = 0x%p, ino = %llu)\n", inode,
+ inode ? (unsigned long long)OCFS2_I(inode)->ip_blkno : 0ULL);
if (!inode) {
mlog(0, "eep, no inode!\n");
int status;
struct ocfs2_dinode *fe = (struct ocfs2_dinode *) bh->b_data;
- mlog_entry("(inode %"MLFu64")\n", OCFS2_I(inode)->ip_blkno);
+ mlog_entry("(inode %llu)\n",
+ (unsigned long long)OCFS2_I(inode)->ip_blkno);
status = ocfs2_journal_access(handle, inode, bh,
OCFS2_JOURNAL_ACCESS_WRITE);
spin_lock(&OCFS2_I(inode)->ip_lock);
fe->i_clusters = cpu_to_le32(OCFS2_I(inode)->ip_clusters);
+ fe->i_attr = cpu_to_le32(OCFS2_I(inode)->ip_attr);
spin_unlock(&OCFS2_I(inode)->ip_lock);
fe->i_size = cpu_to_le64(i_size_read(inode));
void ocfs2_refresh_inode(struct inode *inode,
struct ocfs2_dinode *fe)
{
- struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
-
spin_lock(&OCFS2_I(inode)->ip_lock);
OCFS2_I(inode)->ip_clusters = le32_to_cpu(fe->i_clusters);
+ OCFS2_I(inode)->ip_attr = le32_to_cpu(fe->i_attr);
+ ocfs2_set_inode_flags(inode);
i_size_write(inode, le64_to_cpu(fe->i_size));
inode->i_nlink = le16_to_cpu(fe->i_links_count);
inode->i_uid = le32_to_cpu(fe->i_uid);
inode->i_gid = le32_to_cpu(fe->i_gid);
inode->i_mode = le16_to_cpu(fe->i_mode);
- inode->i_blksize = (u32) osb->s_clustersize;
if (S_ISLNK(inode->i_mode) && le32_to_cpu(fe->i_clusters) == 0)
inode->i_blocks = 0;
else