Merge branch 'guilt/5.19-miscellaneous' into xfs-5.19-for-next

fs/xfs/libxfs/xfs_rtbitmap.c

@@ -1008,6 +1008,7 @@ xfs_rtfree_extent(
 /* Find all the free records within a given range. */
 int
 xfs_rtalloc_query_range(
+	struct xfs_mount		*mp,
 	struct xfs_trans		*tp,
 	const struct xfs_rtalloc_rec	*low_rec,
 	const struct xfs_rtalloc_rec	*high_rec,
@@ -1015,7 +1016,6 @@ xfs_rtalloc_query_range(
 	void				*priv)
 {
 	struct xfs_rtalloc_rec		rec;
-	struct xfs_mount		*mp = tp->t_mountp;
 	xfs_rtblock_t			rtstart;
 	xfs_rtblock_t			rtend;
 	xfs_rtblock_t			high_key;
@@ -1048,7 +1048,7 @@ xfs_rtalloc_query_range(
 			rec.ar_startext = rtstart;
 			rec.ar_extcount = rtend - rtstart + 1;
 
-			error = fn(tp, &rec, priv);
+			error = fn(mp, tp, &rec, priv);
 			if (error)
 				break;
 		}
@@ -1062,6 +1062,7 @@ xfs_rtalloc_query_range(
 /* Find all the free records. */
 int
 xfs_rtalloc_query_all(
+	struct xfs_mount		*mp,
 	struct xfs_trans		*tp,
 	xfs_rtalloc_query_range_fn	fn,
 	void				*priv)
@@ -1069,10 +1070,10 @@ xfs_rtalloc_query_all(
 	struct xfs_rtalloc_rec		keys[2];
 
 	keys[0].ar_startext = 0;
-	keys[1].ar_startext = tp->t_mountp->m_sb.sb_rextents - 1;
+	keys[1].ar_startext = mp->m_sb.sb_rextents - 1;
 	keys[0].ar_extcount = keys[1].ar_extcount = 0;
 
-	return xfs_rtalloc_query_range(tp, &keys[0], &keys[1], fn, priv);
+	return xfs_rtalloc_query_range(mp, tp, &keys[0], &keys[1], fn, priv);
 }
 
 /* Is the given extent all free? */

fs/xfs/libxfs/xfs_sb.c

@@ -911,6 +911,11 @@ xfs_log_sb(
 	 * reservations that have been taken out percpu counters. If we have an
 	 * unclean shutdown, this will be corrected by log recovery rebuilding
 	 * the counters from the AGF block counts.
+	 *
+	 * Do not update sb_frextents here because it is not part of the lazy
+	 * sb counters, despite having a percpu counter. It is always kept
+	 * consistent with the ondisk rtbitmap by xfs_trans_apply_sb_deltas()
+	 * and hence we don't need have to update it here.
 	 */
 	if (xfs_has_lazysbcount(mp)) {
 		mp->m_sb.sb_icount = percpu_counter_sum(&mp->m_icount);

fs/xfs/scrub/rtbitmap.c

@@ -40,6 +40,7 @@ xchk_setup_rt(
 /* Scrub a free extent record from the realtime bitmap. */
 STATIC int
 xchk_rtbitmap_rec(
+	struct xfs_mount	*mp,
 	struct xfs_trans	*tp,
 	const struct xfs_rtalloc_rec *rec,
 	void			*priv)
@@ -48,10 +49,10 @@ xchk_rtbitmap_rec(
 	xfs_rtblock_t		startblock;
 	xfs_rtblock_t		blockcount;
 
-	startblock = rec->ar_startext * tp->t_mountp->m_sb.sb_rextsize;
-	blockcount = rec->ar_extcount * tp->t_mountp->m_sb.sb_rextsize;
+	startblock = rec->ar_startext * mp->m_sb.sb_rextsize;
+	blockcount = rec->ar_extcount * mp->m_sb.sb_rextsize;
 
-	if (!xfs_verify_rtext(sc->mp, startblock, blockcount))
+	if (!xfs_verify_rtext(mp, startblock, blockcount))
 		xchk_fblock_set_corrupt(sc, XFS_DATA_FORK, 0);
 	return 0;
 }
@@ -114,7 +115,7 @@ xchk_rtbitmap(
 	if (error || (sc->sm->sm_flags & XFS_SCRUB_OFLAG_CORRUPT))
 		return error;
 
-	error = xfs_rtalloc_query_all(sc->tp, xchk_rtbitmap_rec, sc);
+	error = xfs_rtalloc_query_all(sc->mp, sc->tp, xchk_rtbitmap_rec, sc);
 	if (!xchk_fblock_process_error(sc, XFS_DATA_FORK, 0, &error))
 		goto out;
 

fs/xfs/xfs_file.c

@@ -694,9 +694,7 @@ xfs_file_buffered_write(
 	struct kiocb		*iocb,
 	struct iov_iter		*from)
 {
-	struct file		*file = iocb->ki_filp;
-	struct address_space	*mapping = file->f_mapping;
-	struct inode		*inode = mapping->host;
+	struct inode		*inode = iocb->ki_filp->f_mapping->host;
 	struct xfs_inode	*ip = XFS_I(inode);
 	ssize_t			ret;
 	bool			cleared_space = false;
@@ -767,9 +765,7 @@ xfs_file_write_iter(
 	struct kiocb		*iocb,
 	struct iov_iter		*from)
 {
-	struct file		*file = iocb->ki_filp;
-	struct address_space	*mapping = file->f_mapping;
-	struct inode		*inode = mapping->host;
+	struct inode		*inode = iocb->ki_filp->f_mapping->host;
 	struct xfs_inode	*ip = XFS_I(inode);
 	ssize_t			ret;
 	size_t			ocount = iov_iter_count(from);
@@ -1167,12 +1163,10 @@ xfs_file_open(
 	struct inode	*inode,
 	struct file	*file)
 {
-	if (!(file->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
-		return -EFBIG;
 	if (xfs_is_shutdown(XFS_M(inode->i_sb)))
 		return -EIO;
 	file->f_mode |= FMODE_NOWAIT | FMODE_BUF_RASYNC;
-	return 0;
+	return generic_file_open(inode, file);
 }
 
 STATIC int

fs/xfs/xfs_fsmap.c

@@ -450,11 +450,11 @@ xfs_getfsmap_logdev(
 /* Transform a rtbitmap "record" into a fsmap */
 STATIC int
 xfs_getfsmap_rtdev_rtbitmap_helper(
+	struct xfs_mount		*mp,
 	struct xfs_trans		*tp,
 	const struct xfs_rtalloc_rec	*rec,
 	void				*priv)
 {
-	struct xfs_mount		*mp = tp->t_mountp;
 	struct xfs_getfsmap_info	*info = priv;
 	struct xfs_rmap_irec		irec;
 	xfs_daddr_t			rec_daddr;
@@ -535,7 +535,7 @@ xfs_getfsmap_rtdev_rtbitmap_query(
 	do_div(alow.ar_startext, mp->m_sb.sb_rextsize);
 	if (do_div(ahigh.ar_startext, mp->m_sb.sb_rextsize))
 		ahigh.ar_startext++;
-	error = xfs_rtalloc_query_range(tp, &alow, &ahigh,
+	error = xfs_rtalloc_query_range(mp, tp, &alow, &ahigh,
 			xfs_getfsmap_rtdev_rtbitmap_helper, info);
 	if (error)
 		goto err;
@@ -547,7 +547,7 @@ xfs_getfsmap_rtdev_rtbitmap_query(
 	info->last = true;
 	ahigh.ar_startext = min(mp->m_sb.sb_rextents, ahigh.ar_startext);
 
-	error = xfs_getfsmap_rtdev_rtbitmap_helper(tp, &ahigh, info);
+	error = xfs_getfsmap_rtdev_rtbitmap_helper(mp, tp, &ahigh, info);
 	if (error)
 		goto err;
 err:

fs/xfs/xfs_fsops.c

@@ -349,10 +349,7 @@ xfs_fs_counts(
 	cnt->freeino = percpu_counter_read_positive(&mp->m_ifree);
 	cnt->freedata = percpu_counter_read_positive(&mp->m_fdblocks) -
 						xfs_fdblocks_unavailable(mp);
-
-	spin_lock(&mp->m_sb_lock);
-	cnt->freertx = mp->m_sb.sb_frextents;
-	spin_unlock(&mp->m_sb_lock);
+	cnt->freertx = percpu_counter_read_positive(&mp->m_frextents);
 }
 
 /*

fs/xfs/xfs_icache.c

@@ -1916,13 +1916,16 @@ xfs_inodegc_want_queue_rt_file(
 	struct xfs_inode	*ip)
 {
 	struct xfs_mount	*mp = ip->i_mount;
-	uint64_t		freertx;
 
 	if (!XFS_IS_REALTIME_INODE(ip))
 		return false;
 
-	freertx = READ_ONCE(mp->m_sb.sb_frextents);
-	return freertx < mp->m_low_rtexts[XFS_LOWSP_5_PCNT];
+	if (__percpu_counter_compare(&mp->m_frextents,
+				mp->m_low_rtexts[XFS_LOWSP_5_PCNT],
+				XFS_FDBLOCKS_BATCH) < 0)
+		return true;
+
+	return false;
 }
 #else
 # define xfs_inodegc_want_queue_rt_file(ip)	(false)

fs/xfs/xfs_message.c

@@ -27,40 +27,32 @@ __xfs_printk(
 	printk("%sXFS: %pV\n", level, vaf);
 }
 
-#define define_xfs_printk_level(func, kern_level)		\
-void func(const struct xfs_mount *mp, const char *fmt, ...)	\
-{								\
-	struct va_format	vaf;				\
-	va_list			args;				\
-	int			level;				\
-								\
-	va_start(args, fmt);					\
-								\
-	vaf.fmt = fmt;						\
-	vaf.va = &args;						\
-								\
-	__xfs_printk(kern_level, mp, &vaf);			\
-	va_end(args);						\
-								\
-	if (!kstrtoint(kern_level, 0, &level) &&		\
-	    level <= LOGLEVEL_ERR &&				\
-	    xfs_error_level >= XFS_ERRLEVEL_HIGH)		\
-		xfs_stack_trace();				\
-}								\
-
-define_xfs_printk_level(xfs_emerg, KERN_EMERG);
-define_xfs_printk_level(xfs_alert, KERN_ALERT);
-define_xfs_printk_level(xfs_crit, KERN_CRIT);
-define_xfs_printk_level(xfs_err, KERN_ERR);
-define_xfs_printk_level(xfs_warn, KERN_WARNING);
-define_xfs_printk_level(xfs_notice, KERN_NOTICE);
-define_xfs_printk_level(xfs_info, KERN_INFO);
-#ifdef DEBUG
-define_xfs_printk_level(xfs_debug, KERN_DEBUG);
-#endif
+void
+xfs_printk_level(
+	const char *kern_level,
+	const struct xfs_mount *mp,
+	const char *fmt, ...)
+{
+	struct va_format	vaf;
+	va_list			args;
+	int			level;
+
+	va_start(args, fmt);
+	vaf.fmt = fmt;
+	vaf.va = &args;
+
+	__xfs_printk(kern_level, mp, &vaf);
+
+	va_end(args);
+
+	if (!kstrtoint(kern_level, 0, &level) &&
+	    level <= LOGLEVEL_ERR &&
+	    xfs_error_level >= XFS_ERRLEVEL_HIGH)
+		xfs_stack_trace();
+}
 
 void
-xfs_alert_tag(
+_xfs_alert_tag(
 	const struct xfs_mount	*mp,
 	int			panic_tag,
 	const char		*fmt, ...)

fs/xfs/xfs_message.h

@@ -6,33 +6,45 @@
 
 struct xfs_mount;
 
-extern __printf(2, 3)
-void xfs_emerg(const struct xfs_mount *mp, const char *fmt, ...);
-extern __printf(2, 3)
-void xfs_alert(const struct xfs_mount *mp, const char *fmt, ...);
 extern __printf(3, 4)
-void xfs_alert_tag(const struct xfs_mount *mp, int tag, const char *fmt, ...);
-extern __printf(2, 3)
-void xfs_crit(const struct xfs_mount *mp, const char *fmt, ...);
-extern __printf(2, 3)
-void xfs_err(const struct xfs_mount *mp, const char *fmt, ...);
-extern __printf(2, 3)
-void xfs_warn(const struct xfs_mount *mp, const char *fmt, ...);
-extern __printf(2, 3)
-void xfs_notice(const struct xfs_mount *mp, const char *fmt, ...);
-extern __printf(2, 3)
-void xfs_info(const struct xfs_mount *mp, const char *fmt, ...);
+void xfs_printk_level(const char *kern_level, const struct xfs_mount *mp,
+			const char *fmt, ...);
 
+#define xfs_printk_index_wrap(kern_level, mp, fmt, ...)		\
+({								\
+	printk_index_subsys_emit("%sXFS%s: ", kern_level, fmt);	\
+	xfs_printk_level(kern_level, mp, fmt, ##__VA_ARGS__);	\
+})
+#define xfs_emerg(mp, fmt, ...) \
+	xfs_printk_index_wrap(KERN_EMERG, mp, fmt, ##__VA_ARGS__)
+#define xfs_alert(mp, fmt, ...) \
+	xfs_printk_index_wrap(KERN_ALERT, mp, fmt, ##__VA_ARGS__)
+#define xfs_crit(mp, fmt, ...) \
+	xfs_printk_index_wrap(KERN_CRIT, mp, fmt, ##__VA_ARGS__)
+#define xfs_err(mp, fmt, ...) \
+	xfs_printk_index_wrap(KERN_ERR, mp, fmt, ##__VA_ARGS__)
+#define xfs_warn(mp, fmt, ...) \
+	xfs_printk_index_wrap(KERN_WARNING, mp, fmt, ##__VA_ARGS__)
+#define xfs_notice(mp, fmt, ...) \
+	xfs_printk_index_wrap(KERN_NOTICE, mp, fmt, ##__VA_ARGS__)
+#define xfs_info(mp, fmt, ...) \
+	xfs_printk_index_wrap(KERN_INFO, mp, fmt, ##__VA_ARGS__)
 #ifdef DEBUG
-extern __printf(2, 3)
-void xfs_debug(const struct xfs_mount *mp, const char *fmt, ...);
+#define xfs_debug(mp, fmt, ...) \
+	xfs_printk_index_wrap(KERN_DEBUG, mp, fmt, ##__VA_ARGS__)
 #else
-static inline __printf(2, 3)
-void xfs_debug(const struct xfs_mount *mp, const char *fmt, ...)
-{
-}
+#define xfs_debug(mp, fmt, ...) do {} while (0)
 #endif
 
+#define xfs_alert_tag(mp, tag, fmt, ...)			\
+({								\
+	printk_index_subsys_emit("%sXFS%s: ", KERN_ALERT, fmt);	\
+	_xfs_alert_tag(mp, tag, fmt, ##__VA_ARGS__);		\
+})
+
+extern __printf(3, 4)
+void _xfs_alert_tag(const struct xfs_mount *mp, int tag, const char *fmt, ...);
+
 #define xfs_printk_ratelimited(func, dev, fmt, ...)			\
 do {									\
 	static DEFINE_RATELIMIT_STATE(_rs,				\

fs/xfs/xfs_mount.c

@@ -468,6 +468,8 @@ STATIC int
 xfs_check_summary_counts(
 	struct xfs_mount	*mp)
 {
+	int			error = 0;
+
 	/*
 	 * The AG0 superblock verifier rejects in-progress filesystems,
 	 * so we should never see the flag set this far into mounting.
@@ -506,11 +508,32 @@ xfs_check_summary_counts(
 	 * superblock to be correct and we don't need to do anything here.
 	 * Otherwise, recalculate the summary counters.
 	 */
-	if ((!xfs_has_lazysbcount(mp) || xfs_is_clean(mp)) &&
-	    !xfs_fs_has_sickness(mp, XFS_SICK_FS_COUNTERS))
-		return 0;
+	if ((xfs_has_lazysbcount(mp) && !xfs_is_clean(mp)) ||
+	    xfs_fs_has_sickness(mp, XFS_SICK_FS_COUNTERS)) {
+		error = xfs_initialize_perag_data(mp, mp->m_sb.sb_agcount);
+		if (error)
+			return error;
+	}
 
-	return xfs_initialize_perag_data(mp, mp->m_sb.sb_agcount);
+	/*
+	 * Older kernels misused sb_frextents to reflect both incore
+	 * reservations made by running transactions and the actual count of
+	 * free rt extents in the ondisk metadata.  Transactions committed
+	 * during runtime can therefore contain a superblock update that
+	 * undercounts the number of free rt extents tracked in the rt bitmap.
+	 * A clean unmount record will have the correct frextents value since
+	 * there can be no other transactions running at that point.
+	 *
+	 * If we're mounting the rt volume after recovering the log, recompute
+	 * frextents from the rtbitmap file to fix the inconsistency.
+	 */
+	if (xfs_has_realtime(mp) && !xfs_is_clean(mp)) {
+		error = xfs_rtalloc_reinit_frextents(mp);
+		if (error)
+			return error;
+	}
+
+	return 0;
 }
 
 /*
@@ -784,11 +807,6 @@ xfs_mountfs(
 		goto out_inodegc_shrinker;
 	}
 
-	/* Make sure the summary counts are ok. */
-	error = xfs_check_summary_counts(mp);
-	if (error)
-		goto out_log_dealloc;
-
 	/* Enable background inode inactivation workers. */
 	xfs_inodegc_start(mp);
 	xfs_blockgc_start(mp);
@@ -844,6 +862,11 @@ xfs_mountfs(
 		goto out_rele_rip;
 	}
 
+	/* Make sure the summary counts are ok. */
+	error = xfs_check_summary_counts(mp);
+	if (error)
+		goto out_rtunmount;
+
 	/*
 	 * If this is a read-only mount defer the superblock updates until
 	 * the next remount into writeable mode.  Otherwise we would never
@@ -1087,24 +1110,33 @@ xfs_fs_writable(
 	return true;
 }
 
+/* Adjust m_fdblocks or m_frextents. */
 int
-xfs_mod_fdblocks(
+xfs_mod_freecounter(
 	struct xfs_mount	*mp,
+	struct percpu_counter	*counter,
 	int64_t			delta,
 	bool			rsvd)
 {
 	int64_t			lcounter;
 	long long		res_used;
+	uint64_t		set_aside = 0;
 	s32			batch;
-	uint64_t		set_aside;
+	bool			has_resv_pool;
+
+	ASSERT(counter == &mp->m_fdblocks || counter == &mp->m_frextents);
+	has_resv_pool = (counter == &mp->m_fdblocks);
+	if (rsvd)
+		ASSERT(has_resv_pool);
 
 	if (delta > 0) {
 		/*
 		 * If the reserve pool is depleted, put blocks back into it
 		 * first. Most of the time the pool is full.
 		 */
-		if (likely(mp->m_resblks == mp->m_resblks_avail)) {
-			percpu_counter_add(&mp->m_fdblocks, delta);
+		if (likely(!has_resv_pool ||
+			   mp->m_resblks == mp->m_resblks_avail)) {
+			percpu_counter_add(counter, delta);
 			return 0;
 		}
 
@@ -1116,7 +1148,7 @@ xfs_mod_fdblocks(
 		} else {
 			delta -= res_used;
 			mp->m_resblks_avail = mp->m_resblks;
-			percpu_counter_add(&mp->m_fdblocks, delta);
+			percpu_counter_add(counter, delta);
 		}
 		spin_unlock(&mp->m_sb_lock);
 		return 0;
@@ -1130,7 +1162,7 @@ xfs_mod_fdblocks(
 	 * then make everything serialise as we are real close to
 	 * ENOSPC.
 	 */
-	if (__percpu_counter_compare(&mp->m_fdblocks, 2 * XFS_FDBLOCKS_BATCH,
+	if (__percpu_counter_compare(counter, 2 * XFS_FDBLOCKS_BATCH,
 				     XFS_FDBLOCKS_BATCH) < 0)
 		batch = 1;
 	else
@@ -1147,9 +1179,10 @@ xfs_mod_fdblocks(
 	 * problems (i.e. transaction abort, pagecache discards, etc.) than
 	 * slightly premature -ENOSPC.
 	 */
-	set_aside = xfs_fdblocks_unavailable(mp);
-	percpu_counter_add_batch(&mp->m_fdblocks, delta, batch);
-	if (__percpu_counter_compare(&mp->m_fdblocks, set_aside,
+	if (has_resv_pool)
+		set_aside = xfs_fdblocks_unavailable(mp);
+	percpu_counter_add_batch(counter, delta, batch);
+	if (__percpu_counter_compare(counter, set_aside,
 				     XFS_FDBLOCKS_BATCH) >= 0) {
 		/* we had space! */
 		return 0;
@@ -1160,8 +1193,8 @@ xfs_mod_fdblocks(
 	 * that took us to ENOSPC.
 	 */
 	spin_lock(&mp->m_sb_lock);
-	percpu_counter_add(&mp->m_fdblocks, -delta);
-	if (!rsvd)
+	percpu_counter_add(counter, -delta);
+	if (!has_resv_pool || !rsvd)
 		goto fdblocks_enospc;
 
 	lcounter = (long long)mp->m_resblks_avail + delta;
@@ -1178,24 +1211,6 @@ xfs_mod_fdblocks(
 	return -ENOSPC;
 }
 
-int
-xfs_mod_frextents(
-	struct xfs_mount	*mp,
-	int64_t			delta)
-{
-	int64_t			lcounter;
-	int			ret = 0;
-
-	spin_lock(&mp->m_sb_lock);
-	lcounter = mp->m_sb.sb_frextents + delta;
-	if (lcounter < 0)
-		ret = -ENOSPC;
-	else
-		mp->m_sb.sb_frextents = lcounter;
-	spin_unlock(&mp->m_sb_lock);
-	return ret;
-}
-
 /*
  * Used to free the superblock along various error paths.
  */

fs/xfs/xfs_mount.h

@@ -183,6 +183,8 @@ typedef struct xfs_mount {
 	struct percpu_counter	m_icount;	/* allocated inodes counter */
 	struct percpu_counter	m_ifree;	/* free inodes counter */
 	struct percpu_counter	m_fdblocks;	/* free block counter */
+	struct percpu_counter	m_frextents;	/* free rt extent counter */
+
 	/*
 	 * Count of data device blocks reserved for delayed allocations,
 	 * including indlen blocks.  Does not include allocated CoW staging
@@ -494,9 +496,20 @@ xfs_fdblocks_unavailable(
 	return mp->m_alloc_set_aside + atomic64_read(&mp->m_allocbt_blks);
 }
 
-extern int	xfs_mod_fdblocks(struct xfs_mount *mp, int64_t delta,
-				 bool reserved);
-extern int	xfs_mod_frextents(struct xfs_mount *mp, int64_t delta);
+int xfs_mod_freecounter(struct xfs_mount *mp, struct percpu_counter *counter,
+		int64_t delta, bool rsvd);
+
+static inline int
+xfs_mod_fdblocks(struct xfs_mount *mp, int64_t delta, bool reserved)
+{
+	return xfs_mod_freecounter(mp, &mp->m_fdblocks, delta, reserved);
+}
+
+static inline int
+xfs_mod_frextents(struct xfs_mount *mp, int64_t delta)
+{
+	return xfs_mod_freecounter(mp, &mp->m_frextents, delta, false);
+}
 
 extern int	xfs_readsb(xfs_mount_t *, int);
 extern void	xfs_freesb(xfs_mount_t *);

fs/xfs/xfs_rtalloc.c

@@ -1284,6 +1284,44 @@ xfs_rtmount_init(
 	return 0;
 }
 
+static int
+xfs_rtalloc_count_frextent(
+	struct xfs_mount		*mp,
+	struct xfs_trans		*tp,
+	const struct xfs_rtalloc_rec	*rec,
+	void				*priv)
+{
+	uint64_t			*valp = priv;
+
+	*valp += rec->ar_extcount;
+	return 0;
+}
+
+/*
+ * Reinitialize the number of free realtime extents from the realtime bitmap.
+ * Callers must ensure that there is no other activity in the filesystem.
+ */
+int
+xfs_rtalloc_reinit_frextents(
+	struct xfs_mount	*mp)
+{
+	uint64_t		val = 0;
+	int			error;
+
+	xfs_ilock(mp->m_rbmip, XFS_ILOCK_EXCL);
+	error = xfs_rtalloc_query_all(mp, NULL, xfs_rtalloc_count_frextent,
+			&val);
+	xfs_iunlock(mp->m_rbmip, XFS_ILOCK_EXCL);
+	if (error)
+		return error;
+
+	spin_lock(&mp->m_sb_lock);
+	mp->m_sb.sb_frextents = val;
+	spin_unlock(&mp->m_sb_lock);
+	percpu_counter_set(&mp->m_frextents, mp->m_sb.sb_frextents);
+	return 0;
+}
+
 /*
  * Get the bitmap and summary inodes and the summary cache into the mount
  * structure at mount time.

fs/xfs/xfs_rtalloc.h

@@ -22,6 +22,7 @@ struct xfs_rtalloc_rec {
 };
 
 typedef int (*xfs_rtalloc_query_range_fn)(
+	struct xfs_mount		*mp,
 	struct xfs_trans		*tp,
 	const struct xfs_rtalloc_rec	*rec,
 	void				*priv);
@@ -123,27 +124,29 @@ int xfs_rtmodify_summary(struct xfs_mount *mp, struct xfs_trans *tp, int log,
 int xfs_rtfree_range(struct xfs_mount *mp, struct xfs_trans *tp,
 		     xfs_rtblock_t start, xfs_extlen_t len,
 		     struct xfs_buf **rbpp, xfs_fsblock_t *rsb);
-int xfs_rtalloc_query_range(struct xfs_trans *tp,
+int xfs_rtalloc_query_range(struct xfs_mount *mp, struct xfs_trans *tp,
 		const struct xfs_rtalloc_rec *low_rec,
 		const struct xfs_rtalloc_rec *high_rec,
 		xfs_rtalloc_query_range_fn fn, void *priv);
-int xfs_rtalloc_query_all(struct xfs_trans *tp,
+int xfs_rtalloc_query_all(struct xfs_mount *mp, struct xfs_trans *tp,
 			  xfs_rtalloc_query_range_fn fn,
 			  void *priv);
 bool xfs_verify_rtbno(struct xfs_mount *mp, xfs_rtblock_t rtbno);
 int xfs_rtalloc_extent_is_free(struct xfs_mount *mp, struct xfs_trans *tp,
 			       xfs_rtblock_t start, xfs_extlen_t len,
 			       bool *is_free);
+int xfs_rtalloc_reinit_frextents(struct xfs_mount *mp);
 #else
 # define xfs_rtallocate_extent(t,b,min,max,l,f,p,rb)    (ENOSYS)
 # define xfs_rtfree_extent(t,b,l)                       (ENOSYS)
 # define xfs_rtpick_extent(m,t,l,rb)                    (ENOSYS)
 # define xfs_growfs_rt(mp,in)                           (ENOSYS)
 # define xfs_rtalloc_query_range(t,l,h,f,p)             (ENOSYS)
-# define xfs_rtalloc_query_all(t,f,p)                   (ENOSYS)
+# define xfs_rtalloc_query_all(m,t,f,p)                 (ENOSYS)
 # define xfs_rtbuf_get(m,t,b,i,p)                       (ENOSYS)
 # define xfs_verify_rtbno(m, r)			(false)
 # define xfs_rtalloc_extent_is_free(m,t,s,l,i)          (ENOSYS)
+# define xfs_rtalloc_reinit_frextents(m)                (0)
 static inline int		/* error */
 xfs_rtmount_init(
 	xfs_mount_t	*mp)	/* file system mount structure */

fs/xfs/xfs_super.c

@@ -843,9 +843,11 @@ xfs_fs_statfs(
 
 	if (XFS_IS_REALTIME_MOUNT(mp) &&
 	    (ip->i_diflags & (XFS_DIFLAG_RTINHERIT | XFS_DIFLAG_REALTIME))) {
+		s64	freertx;
+
 		statp->f_blocks = sbp->sb_rblocks;
-		statp->f_bavail = statp->f_bfree =
-			sbp->sb_frextents * sbp->sb_rextsize;
+		freertx = percpu_counter_sum_positive(&mp->m_frextents);
+		statp->f_bavail = statp->f_bfree = freertx * sbp->sb_rextsize;
 	}
 
 	return 0;
@@ -1015,8 +1017,14 @@ xfs_init_percpu_counters(
 	if (error)
 		goto free_fdblocks;
 
+	error = percpu_counter_init(&mp->m_frextents, 0, GFP_KERNEL);
+	if (error)
+		goto free_delalloc;
+
 	return 0;
 
+free_delalloc:
+	percpu_counter_destroy(&mp->m_delalloc_blks);
 free_fdblocks:
 	percpu_counter_destroy(&mp->m_fdblocks);
 free_ifree:
@@ -1033,6 +1041,7 @@ xfs_reinit_percpu_counters(
 	percpu_counter_set(&mp->m_icount, mp->m_sb.sb_icount);
 	percpu_counter_set(&mp->m_ifree, mp->m_sb.sb_ifree);
 	percpu_counter_set(&mp->m_fdblocks, mp->m_sb.sb_fdblocks);
+	percpu_counter_set(&mp->m_frextents, mp->m_sb.sb_frextents);
 }
 
 static void
@@ -1045,6 +1054,7 @@ xfs_destroy_percpu_counters(
 	ASSERT(xfs_is_shutdown(mp) ||
 	       percpu_counter_sum(&mp->m_delalloc_blks) == 0);
 	percpu_counter_destroy(&mp->m_delalloc_blks);
+	percpu_counter_destroy(&mp->m_frextents);
 }
 
 static int

fs/xfs/xfs_trans.c

@@ -498,10 +498,31 @@ xfs_trans_apply_sb_deltas(
 			be64_add_cpu(&sbp->sb_fdblocks, tp->t_res_fdblocks_delta);
 	}
 
-	if (tp->t_frextents_delta)
-		be64_add_cpu(&sbp->sb_frextents, tp->t_frextents_delta);
-	if (tp->t_res_frextents_delta)
-		be64_add_cpu(&sbp->sb_frextents, tp->t_res_frextents_delta);
+	/*
+	 * Updating frextents requires careful handling because it does not
+	 * behave like the lazysb counters because we cannot rely on log
+	 * recovery in older kenels to recompute the value from the rtbitmap.
+	 * This means that the ondisk frextents must be consistent with the
+	 * rtbitmap.
+	 *
+	 * Therefore, log the frextents change to the ondisk superblock and
+	 * update the incore superblock so that future calls to xfs_log_sb
+	 * write the correct value ondisk.
+	 *
+	 * Don't touch m_frextents because it includes incore reservations,
+	 * and those are handled by the unreserve function.
+	 */
+	if (tp->t_frextents_delta || tp->t_res_frextents_delta) {
+		struct xfs_mount	*mp = tp->t_mountp;
+		int64_t			rtxdelta;
+
+		rtxdelta = tp->t_frextents_delta + tp->t_res_frextents_delta;
+
+		spin_lock(&mp->m_sb_lock);
+		be64_add_cpu(&sbp->sb_frextents, rtxdelta);
+		mp->m_sb.sb_frextents += rtxdelta;
+		spin_unlock(&mp->m_sb_lock);
+	}
 
 	if (tp->t_dblocks_delta) {
 		be64_add_cpu(&sbp->sb_dblocks, tp->t_dblocks_delta);
@@ -614,7 +635,12 @@ xfs_trans_unreserve_and_mod_sb(
 	if (ifreedelta)
 		percpu_counter_add(&mp->m_ifree, ifreedelta);
 
-	if (rtxdelta == 0 && !(tp->t_flags & XFS_TRANS_SB_DIRTY))
+	if (rtxdelta) {
+		error = xfs_mod_frextents(mp, rtxdelta);
+		ASSERT(!error);
+	}
+
+	if (!(tp->t_flags & XFS_TRANS_SB_DIRTY))
 		return;
 
 	/* apply remaining deltas */
@@ -622,7 +648,12 @@ xfs_trans_unreserve_and_mod_sb(
 	mp->m_sb.sb_fdblocks += tp->t_fdblocks_delta + tp->t_res_fdblocks_delta;
 	mp->m_sb.sb_icount += idelta;
 	mp->m_sb.sb_ifree += ifreedelta;
-	mp->m_sb.sb_frextents += rtxdelta;
+	/*
+	 * Do not touch sb_frextents here because we are dealing with incore
+	 * reservation.  sb_frextents is not part of the lazy sb counters so it
+	 * must be consistent with the ondisk rtbitmap and must never include
+	 * incore reservations.
+	 */
 	mp->m_sb.sb_dblocks += tp->t_dblocks_delta;
 	mp->m_sb.sb_agcount += tp->t_agcount_delta;
 	mp->m_sb.sb_imax_pct += tp->t_imaxpct_delta;