Merge tag 'xfs-5.14-fixes-2' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux

Pull xfs fixes from Darrick Wong:
 "This contains a bunch of bug fixes in XFS.

  Dave and I have been busy the last couple of weeks to find and fix as
  many log recovery bugs as we can find; here are the results so far. Go
  fstests -g recoveryloop! ;)

   - Fix a number of coordination bugs relating to cache flushes for
     metadata writeback, cache flushes for multi-buffer log writes, and
     FUA writes for single-buffer log writes

   - Fix a bug with incorrect replay of attr3 blocks

   - Fix unnecessary stalls when flushing logs to disk

   - Fix spoofing problems when recovering realtime bitmap blocks"

* tag 'xfs-5.14-fixes-2' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux:
  xfs: prevent spoofing of rtbitmap blocks when recovering buffers
  xfs: limit iclog tail updates
  xfs: need to see iclog flags in tracing
  xfs: Enforce attr3 buffer recovery order
  xfs: logging the on disk inode LSN can make it go backwards
  xfs: avoid unnecessary waits in xfs_log_force_lsn()
  xfs: log forces imply data device cache flushes
  xfs: factor out forced iclog flushes
  xfs: fix ordering violation between cache flushes and tail updates
  xfs: fold __xlog_state_release_iclog into xlog_state_release_iclog
  xfs: external logs need to flush data device
  xfs: flush data dev on external log write

fs/xfs/libxfs/xfs_log_format.h

@@ -411,7 +411,16 @@ struct xfs_log_dinode {
 	/* start of the extended dinode, writable fields */
 	uint32_t	di_crc;		/* CRC of the inode */
 	uint64_t	di_changecount;	/* number of attribute changes */
-	xfs_lsn_t	di_lsn;		/* flush sequence */
+
+	/*
+	 * The LSN we write to this field during formatting is not a reflection
+	 * of the current on-disk LSN. It should never be used for recovery
+	 * sequencing, nor should it be recovered into the on-disk inode at all.
+	 * See xlog_recover_inode_commit_pass2() and xfs_log_dinode_to_disk()
+	 * for details.
+	 */
+	xfs_lsn_t	di_lsn;
+
 	uint64_t	di_flags2;	/* more random flags */
 	uint32_t	di_cowextsize;	/* basic cow extent size for file */
 	uint8_t		di_pad2[12];	/* more padding for future expansion */

fs/xfs/xfs_buf_item_recover.c

@@ -698,7 +698,8 @@ xlog_recover_do_inode_buffer(
 static xfs_lsn_t
 xlog_recover_get_buf_lsn(
 	struct xfs_mount	*mp,
-	struct xfs_buf		*bp)
+	struct xfs_buf		*bp,
+	struct xfs_buf_log_format *buf_f)
 {
 	uint32_t		magic32;
 	uint16_t		magic16;
@@ -706,11 +707,20 @@ xlog_recover_get_buf_lsn(
 	void			*blk = bp->b_addr;
 	uuid_t			*uuid;
 	xfs_lsn_t		lsn = -1;
+	uint16_t		blft;
 
 	/* v4 filesystems always recover immediately */
 	if (!xfs_sb_version_hascrc(&mp->m_sb))
 		goto recover_immediately;
 
+	/*
+	 * realtime bitmap and summary file blocks do not have magic numbers or
+	 * UUIDs, so we must recover them immediately.
+	 */
+	blft = xfs_blft_from_flags(buf_f);
+	if (blft == XFS_BLFT_RTBITMAP_BUF || blft == XFS_BLFT_RTSUMMARY_BUF)
+		goto recover_immediately;
+
 	magic32 = be32_to_cpu(*(__be32 *)blk);
 	switch (magic32) {
 	case XFS_ABTB_CRC_MAGIC:
@@ -796,6 +806,7 @@ xlog_recover_get_buf_lsn(
 	switch (magicda) {
 	case XFS_DIR3_LEAF1_MAGIC:
 	case XFS_DIR3_LEAFN_MAGIC:
+	case XFS_ATTR3_LEAF_MAGIC:
 	case XFS_DA3_NODE_MAGIC:
 		lsn = be64_to_cpu(((struct xfs_da3_blkinfo *)blk)->lsn);
 		uuid = &((struct xfs_da3_blkinfo *)blk)->uuid;
@@ -919,7 +930,7 @@ xlog_recover_buf_commit_pass2(
 	 * the verifier will be reset to match whatever recover turns that
 	 * buffer into.
 	 */
-	lsn = xlog_recover_get_buf_lsn(mp, bp);
+	lsn = xlog_recover_get_buf_lsn(mp, bp, buf_f);
 	if (lsn && lsn != -1 && XFS_LSN_CMP(lsn, current_lsn) >= 0) {
 		trace_xfs_log_recover_buf_skip(log, buf_f);
 		xlog_recover_validate_buf_type(mp, bp, buf_f, NULLCOMMITLSN);

fs/xfs/xfs_inode_item_recover.c

@@ -145,7 +145,8 @@ xfs_log_dinode_to_disk_ts(
 STATIC void
 xfs_log_dinode_to_disk(
 	struct xfs_log_dinode	*from,
-	struct xfs_dinode	*to)
+	struct xfs_dinode	*to,
+	xfs_lsn_t		lsn)
 {
 	to->di_magic = cpu_to_be16(from->di_magic);
 	to->di_mode = cpu_to_be16(from->di_mode);
@@ -182,7 +183,7 @@ xfs_log_dinode_to_disk(
 		to->di_flags2 = cpu_to_be64(from->di_flags2);
 		to->di_cowextsize = cpu_to_be32(from->di_cowextsize);
 		to->di_ino = cpu_to_be64(from->di_ino);
-		to->di_lsn = cpu_to_be64(from->di_lsn);
+		to->di_lsn = cpu_to_be64(lsn);
 		memcpy(to->di_pad2, from->di_pad2, sizeof(to->di_pad2));
 		uuid_copy(&to->di_uuid, &from->di_uuid);
 		to->di_flushiter = 0;
@@ -261,16 +262,25 @@ xlog_recover_inode_commit_pass2(
 	}
 
 	/*
-	 * If the inode has an LSN in it, recover the inode only if it's less
-	 * than the lsn of the transaction we are replaying. Note: we still
-	 * need to replay an owner change even though the inode is more recent
-	 * than the transaction as there is no guarantee that all the btree
-	 * blocks are more recent than this transaction, too.
+	 * If the inode has an LSN in it, recover the inode only if the on-disk
+	 * inode's LSN is older than the lsn of the transaction we are
+	 * replaying. We can have multiple checkpoints with the same start LSN,
+	 * so the current LSN being equal to the on-disk LSN doesn't necessarily
+	 * mean that the on-disk inode is more recent than the change being
+	 * replayed.
+	 *
+	 * We must check the current_lsn against the on-disk inode
+	 * here because the we can't trust the log dinode to contain a valid LSN
+	 * (see comment below before replaying the log dinode for details).
+	 *
+	 * Note: we still need to replay an owner change even though the inode
+	 * is more recent than the transaction as there is no guarantee that all
+	 * the btree blocks are more recent than this transaction, too.
 	 */
 	if (dip->di_version >= 3) {
 		xfs_lsn_t	lsn = be64_to_cpu(dip->di_lsn);
 
-		if (lsn && lsn != -1 && XFS_LSN_CMP(lsn, current_lsn) >= 0) {
+		if (lsn && lsn != -1 && XFS_LSN_CMP(lsn, current_lsn) > 0) {
 			trace_xfs_log_recover_inode_skip(log, in_f);
 			error = 0;
 			goto out_owner_change;
@@ -368,8 +378,17 @@ xlog_recover_inode_commit_pass2(
 		goto out_release;
 	}
 
-	/* recover the log dinode inode into the on disk inode */
-	xfs_log_dinode_to_disk(ldip, dip);
+	/*
+	 * Recover the log dinode inode into the on disk inode.
+	 *
+	 * The LSN in the log dinode is garbage - it can be zero or reflect
+	 * stale in-memory runtime state that isn't coherent with the changes
+	 * logged in this transaction or the changes written to the on-disk
+	 * inode.  Hence we write the current lSN into the inode because that
+	 * matches what xfs_iflush() would write inode the inode when flushing
+	 * the changes in this transaction.
+	 */
+	xfs_log_dinode_to_disk(ldip, dip, current_lsn);
 
 	fields = in_f->ilf_fields;
 	if (fields & XFS_ILOG_DEV)

fs/xfs/xfs_log_cil.c

@@ -654,8 +654,9 @@ xlog_cil_push_work(
 	struct xfs_trans_header thdr;
 	struct xfs_log_iovec	lhdr;
 	struct xfs_log_vec	lvhdr = { NULL };
+	xfs_lsn_t		preflush_tail_lsn;
 	xfs_lsn_t		commit_lsn;
-	xfs_lsn_t		push_seq;
+	xfs_csn_t		push_seq;
 	struct bio		bio;
 	DECLARE_COMPLETION_ONSTACK(bdev_flush);
 
@@ -730,7 +731,15 @@ xlog_cil_push_work(
 	 * because we hold the flush lock exclusively. Hence we can now issue
 	 * a cache flush to ensure all the completed metadata in the journal we
 	 * are about to overwrite is on stable storage.
+	 *
+	 * Because we are issuing this cache flush before we've written the
+	 * tail lsn to the iclog, we can have metadata IO completions move the
+	 * tail forwards between the completion of this flush and the iclog
+	 * being written. In this case, we need to re-issue the cache flush
+	 * before the iclog write. To detect whether the log tail moves, sample
+	 * the tail LSN *before* we issue the flush.
 	 */
+	preflush_tail_lsn = atomic64_read(&log->l_tail_lsn);
 	xfs_flush_bdev_async(&bio, log->l_mp->m_ddev_targp->bt_bdev,
 				&bdev_flush);
 
@@ -941,7 +950,7 @@ xlog_cil_push_work(
 	 * storage.
 	 */
 	commit_iclog->ic_flags |= XLOG_ICL_NEED_FUA;
-	xlog_state_release_iclog(log, commit_iclog);
+	xlog_state_release_iclog(log, commit_iclog, preflush_tail_lsn);
 	spin_unlock(&log->l_icloglock);
 	return;
 

fs/xfs/xfs_log_priv.h

@@ -59,6 +59,16 @@ enum xlog_iclog_state {
 	{ XLOG_STATE_DIRTY,	"XLOG_STATE_DIRTY" }, \
 	{ XLOG_STATE_IOERROR,	"XLOG_STATE_IOERROR" }
 
+/*
+ * In core log flags
+ */
+#define XLOG_ICL_NEED_FLUSH	(1 << 0)	/* iclog needs REQ_PREFLUSH */
+#define XLOG_ICL_NEED_FUA	(1 << 1)	/* iclog needs REQ_FUA */
+
+#define XLOG_ICL_STRINGS \
+	{ XLOG_ICL_NEED_FLUSH,	"XLOG_ICL_NEED_FLUSH" }, \
+	{ XLOG_ICL_NEED_FUA,	"XLOG_ICL_NEED_FUA" }
+
 
 /*
  * Log ticket flags
@@ -143,9 +153,6 @@ enum xlog_iclog_state {
 
 #define XLOG_COVER_OPS		5
 
-#define XLOG_ICL_NEED_FLUSH	(1 << 0)	/* iclog needs REQ_PREFLUSH */
-#define XLOG_ICL_NEED_FUA	(1 << 1)	/* iclog needs REQ_FUA */
-
 /* Ticket reservation region accounting */ 
 #define XLOG_TIC_LEN_MAX	15
 
@@ -497,7 +504,8 @@ int	xlog_commit_record(struct xlog *log, struct xlog_ticket *ticket,
 void	xfs_log_ticket_ungrant(struct xlog *log, struct xlog_ticket *ticket);
 void	xfs_log_ticket_regrant(struct xlog *log, struct xlog_ticket *ticket);
 
-int xlog_state_release_iclog(struct xlog *log, struct xlog_in_core *iclog);
+int xlog_state_release_iclog(struct xlog *log, struct xlog_in_core *iclog,
+		xfs_lsn_t log_tail_lsn);
 
 /*
  * When we crack an atomic LSN, we sample it first so that the value will not

fs/xfs/xfs_trace.h

@@ -3944,6 +3944,7 @@ DECLARE_EVENT_CLASS(xlog_iclog_class,
 		__field(uint32_t, state)
 		__field(int32_t, refcount)
 		__field(uint32_t, offset)
+		__field(uint32_t, flags)
 		__field(unsigned long long, lsn)
 		__field(unsigned long, caller_ip)
 	),
@@ -3952,15 +3953,17 @@ DECLARE_EVENT_CLASS(xlog_iclog_class,
 		__entry->state = iclog->ic_state;
 		__entry->refcount = atomic_read(&iclog->ic_refcnt);
 		__entry->offset = iclog->ic_offset;
+		__entry->flags = iclog->ic_flags;
 		__entry->lsn = be64_to_cpu(iclog->ic_header.h_lsn);
 		__entry->caller_ip = caller_ip;
 	),
-	TP_printk("dev %d:%d state %s refcnt %d offset %u lsn 0x%llx caller %pS",
+	TP_printk("dev %d:%d state %s refcnt %d offset %u lsn 0x%llx flags %s caller %pS",
 		  MAJOR(__entry->dev), MINOR(__entry->dev),
 		  __print_symbolic(__entry->state, XLOG_STATE_STRINGS),
 		  __entry->refcount,
 		  __entry->offset,
 		  __entry->lsn,
+		  __print_flags(__entry->flags, "|", XLOG_ICL_STRINGS),
 		  (char *)__entry->caller_ip)
 
 );