Skip to contentSkip to conversation

refactor zone reporting v2

PATCH In review Christoph Hellwig 3 participants 16 messages Jan 14 – Jan 21 XFS filesystem

lore
Share & tools
lore
#0 lore raw

Hi all,

this series refactor the zone reporting code so that it is more clearly split between sanity checking the report hardware zone information, and the XFS zoned RT information. This reduced the code size and removes an iteration over all RTGs at boot time.

It will also allow to do smarter checking of hardware zones and RTG allocation information in repair once ported to userspace.

I've also included Damien's xfsprogs patch to make xfs_zones.h compile better standalone as it touches the same area.

Changes since v1:
- merge two functions in the final version to be a little less confusing
- rename one of those functions which later goes away
- expand a comment a bit
- rename the libxfs zone validation to make it more clear it is all about the blk_zone

Diffstat:

 libxfs/xfs_rtgroup.h |   15 ++++
 libxfs/xfs_zones.c   |  149 +++++++++++---------------------------------
 libxfs/xfs_zones.h   |    6 +
 xfs_zone_alloc.c     |  171 +++++++++++++++++++++++++++++----------------------
 4 files changed, 156 insertions(+), 185 deletions(-)
#1 lore raw

PATCH 1/6 xfs: add missing forward declaration in xfs_zones.h

From: Damien Le Moal <[email protected]>

Add the missing forward declaration for struct blk_zone in xfs_zones.h. This avoids headaches with the order of header file inclusion to avoid compilation errors.

---
 fs/xfs/libxfs/xfs_zones.h | 1 +
 1 file changed, 1 insertion(+)
fs/xfs/libxfs/xfs_zones.h +1 -0
index 5fefd132e002..df10a34da71d 100644
@@ -3,6 +3,7 @@
 #define _LIBXFS_ZONES_H
 
 struct xfs_rtgroup;
+struct blk_zone;
 
 /*
  * In order to guarantee forward progress for GC we need to reserve at least
Signature · 1 line
2.47.3
maintainer Reviewed-by on v1 1/6
#2 lore raw

Re: [PATCH 1/6] xfs: add missing forward declaration in xfs_zones.h

Quoted 18 lines · Christoph Hellwig

From: Damien Le Moal <[email protected]>

Add the missing forward declaration for struct blk_zone in xfs_zones.h. This avoids headaches with the order of header file inclusion to avoid compilation errors.

---
 fs/xfs/libxfs/xfs_zones.h | 1 +
 1 file changed, 1 insertion(+)
fs/xfs/libxfs/xfs_zones.h +0 -0
index 5fefd132e002..df10a34da71d 100644

fs/xfs/libxfs/xfs_zones.h line 6

Quoted 4 lines
fs/xfs/libxfs/xfs_zones.h +1 -0
@@ -3,6 +3,7 @@
 #define _LIBXFS_ZONES_H
 
 struct xfs_rtgroup;
+struct blk_zone;
Quoted 7 lines
fs/xfs/libxfs/xfs_zones.h +0 -0
 /*
  * In order to guarantee forward progress for GC we need to reserve at least
Signature · 1 line
2.47.3
#3 lore raw

PATCH 2/6 xfs: add a xfs_rtgroup_raw_size helper

In reply to Christoph Hellwig

Add a helper to figure the on-disk size of a group, accounting for the XFS_SB_FEAT_INCOMPAT_ZONE_GAPS feature if needed.

---
 fs/xfs/libxfs/xfs_rtgroup.h | 15 +++++++++++++++
 1 file changed, 15 insertions(+)
fs/xfs/libxfs/xfs_rtgroup.h +15 -0
index 73cace4d25c7..c0b9f9f2c413 100644
@@ -371,4 +371,19 @@ xfs_rtgs_to_rfsbs(
 	return xfs_groups_to_rfsbs(mp, nr_groups, XG_TYPE_RTG);
 }
 
+/*
+ * Return the "raw" size of a group on the hardware device.  This includes the
+ * daddr gaps present for XFS_SB_FEAT_INCOMPAT_ZONE_GAPS file systems.
+ */
+static inline xfs_rgblock_t
+xfs_rtgroup_raw_size(
+	struct xfs_mount	*mp)
+{
+	struct xfs_groups	*g = &mp->m_groups[XG_TYPE_RTG];
+
+	if (g->has_daddr_gaps)
+		return 1U << g->blklog;
+	return g->blocks;
+}
+
 #endif /* __LIBXFS_RTGROUP_H */
Signature · 1 line
2.47.3
Reviewed-by on v1 2/6
#4 lore raw

Re: [PATCH 2/6] xfs: add a xfs_rtgroup_raw_size helper

Quoted 5 lines · Christoph Hellwig

Add a helper to figure the on-disk size of a group, accounting for the XFS_SB_FEAT_INCOMPAT_ZONE_GAPS feature if needed.

Looks OK to me.

Signature · 2 lines
Damien Le Moal
Western Digital Research
maintainer Reviewed-by on v1 2/6
#5 lore raw

Re: [PATCH 2/6] xfs: add a xfs_rtgroup_raw_size helper

In reply to Christoph Hellwig
Quoted 22 lines · Christoph Hellwig

Add a helper to figure the on-disk size of a group, accounting for the XFS_SB_FEAT_INCOMPAT_ZONE_GAPS feature if needed.

---
 fs/xfs/libxfs/xfs_rtgroup.h | 15 +++++++++++++++
 1 file changed, 15 insertions(+)
fs/xfs/libxfs/xfs_rtgroup.h +5 -0
index 73cace4d25c7..c0b9f9f2c413 100644
@@ -371,4 +371,19 @@ xfs_rtgs_to_rfsbs(
 	return xfs_groups_to_rfsbs(mp, nr_groups, XG_TYPE_RTG);
 }
 
+/*
+ * Return the "raw" size of a group on the hardware device.  This includes the
+ * daddr gaps present for XFS_SB_FEAT_INCOMPAT_ZONE_GAPS file systems.
+ */
+static inline xfs_rgblock_t

fs/xfs/libxfs/xfs_rtgroup.h xfs_rtgs_to_rfsbs( · line 388

Quoted 10 lines
fs/xfs/libxfs/xfs_rtgroup.h +10 -0
+xfs_rtgroup_raw_size(
+	struct xfs_mount	*mp)
+{
+	struct xfs_groups	*g = &mp->m_groups[XG_TYPE_RTG];
+
+	if (g->has_daddr_gaps)
+		return 1U << g->blklog;
+	return g->blocks;
+}
+
Quoted 5 lines
fs/xfs/libxfs/xfs_rtgroup.h +0 -0
 #endif /* __LIBXFS_RTGROUP_H */
Signature · 1 line
2.47.3
#6 lore raw

PATCH 3/6 xfs: pass the write pointer to xfs_init_zone

In reply to Christoph Hellwig

Move the two methods to query the write pointer out of xfs_init_zone into the callers, so that xfs_init_zone doesn't have to bother with the blk_zone structure and instead operates purely at the XFS realtime group level.

---
 fs/xfs/xfs_zone_alloc.c | 66 +++++++++++++++++++++++------------------
 1 file changed, 37 insertions(+), 29 deletions(-)
fs/xfs/xfs_zone_alloc.c +37 -29
index 4ca7769b5adb..87243644d88e 100644
@@ -981,43 +981,43 @@ struct xfs_init_zones {
 	uint64_t		reclaimable;
 };
 
+/*
+ * For sequential write required zones, we restart writing at the hardware write
+ * pointer returned by xfs_zone_validate().
+ *
+ * For conventional zones or conventional devices we have query the rmap to
+ * find the highest recorded block and set the write pointer to the block after
+ * that.  In case of a power loss this misses blocks where the data I/O has
+ * completed but not recorded in the rmap yet, and it also rewrites blocks if
+ * the most recently written ones got deleted again before unmount, but this is
+ * the best we can do without hardware support.
+ */
+static xfs_rgblock_t
+xfs_rmap_estimate_write_pointer(
+	struct xfs_rtgroup	*rtg)
+{
+	xfs_rgblock_t		highest_rgbno;
+
+	xfs_rtgroup_lock(rtg, XFS_RTGLOCK_RMAP);
+	highest_rgbno = xfs_rtrmap_highest_rgbno(rtg);
+	xfs_rtgroup_unlock(rtg, XFS_RTGLOCK_RMAP);
+
+	if (highest_rgbno == NULLRGBLOCK)
+		return 0;
+	return highest_rgbno + 1;
+}
+
 static int
 xfs_init_zone(
 	struct xfs_init_zones	*iz,
 	struct xfs_rtgroup	*rtg,
-	struct blk_zone		*zone)
+	xfs_rgblock_t		write_pointer)
 {
 	struct xfs_mount	*mp = rtg_mount(rtg);
 	struct xfs_zone_info	*zi = mp->m_zone_info;
 	uint32_t		used = rtg_rmap(rtg)->i_used_blocks;
-	xfs_rgblock_t		write_pointer, highest_rgbno;
 	int			error;
 
-	if (zone && !xfs_zone_validate(zone, rtg, &write_pointer))
-		return -EFSCORRUPTED;
-
-	/*
-	 * For sequential write required zones we retrieved the hardware write
-	 * pointer above.
-	 *
-	 * For conventional zones or conventional devices we don't have that
-	 * luxury.  Instead query the rmap to find the highest recorded block
-	 * and set the write pointer to the block after that.  In case of a
-	 * power loss this misses blocks where the data I/O has completed but
-	 * not recorded in the rmap yet, and it also rewrites blocks if the most
-	 * recently written ones got deleted again before unmount, but this is
-	 * the best we can do without hardware support.
-	 */
-	if (!zone || zone->cond == BLK_ZONE_COND_NOT_WP) {
-		xfs_rtgroup_lock(rtg, XFS_RTGLOCK_RMAP);
-		highest_rgbno = xfs_rtrmap_highest_rgbno(rtg);
-		if (highest_rgbno == NULLRGBLOCK)
-			write_pointer = 0;
-		else
-			write_pointer = highest_rgbno + 1;
-		xfs_rtgroup_unlock(rtg, XFS_RTGLOCK_RMAP);
-	}
-
 	/*
 	 * If there are no used blocks, but the zone is not in empty state yet
 	 * we lost power before the zoned reset.  In that case finish the work
@@ -1066,6 +1066,7 @@ xfs_get_zone_info_cb(
 	struct xfs_mount	*mp = iz->mp;
 	xfs_fsblock_t		zsbno = xfs_daddr_to_rtb(mp, zone->start);
 	xfs_rgnumber_t		rgno;
+	xfs_rgblock_t		write_pointer;
 	struct xfs_rtgroup	*rtg;
 	int			error;
 
@@ -1080,7 +1081,13 @@ xfs_get_zone_info_cb(
 		xfs_warn(mp, "realtime group not found for zone %u.", rgno);
 		return -EFSCORRUPTED;
 	}
-	error = xfs_init_zone(iz, rtg, zone);
+	if (!xfs_zone_validate(zone, rtg, &write_pointer)) {
+		xfs_rtgroup_rele(rtg);
+		return -EFSCORRUPTED;
+	}
+	if (zone->cond == BLK_ZONE_COND_NOT_WP)
+		write_pointer = xfs_rmap_estimate_write_pointer(rtg);
+	error = xfs_init_zone(iz, rtg, write_pointer);
 	xfs_rtgroup_rele(rtg);
 	return error;
 }
@@ -1290,7 +1297,8 @@ xfs_mount_zones(
 		struct xfs_rtgroup	*rtg = NULL;
 
 		while ((rtg = xfs_rtgroup_next(mp, rtg))) {
-			error = xfs_init_zone(&iz, rtg, NULL);
+			error = xfs_init_zone(&iz, rtg,
+					xfs_rmap_estimate_write_pointer(rtg));
 			if (error) {
 				xfs_rtgroup_rele(rtg);
 				goto out_free_zone_info;
Signature · 1 line
2.47.3
Reviewed-by on v1 3/6
#7 lore raw

Re: [PATCH 3/6] xfs: pass the write pointer to xfs_init_zone

Quoted 7 lines · Christoph Hellwig

Move the two methods to query the write pointer out of xfs_init_zone into the callers, so that xfs_init_zone doesn't have to bother with the blk_zone structure and instead operates purely at the XFS realtime group level.

One nit below. Otherwise, looks OK to me.

Quoted 9 lines
---
 fs/xfs/xfs_zone_alloc.c | 66 +++++++++++++++++++++++------------------
 1 file changed, 37 insertions(+), 29 deletions(-)
fs/xfs/xfs_zone_alloc.c +0 -0
index 4ca7769b5adb..87243644d88e 100644

fs/xfs/xfs_zone_alloc.c struct xfs_init_zones { · line 988

Quoted 8 lines
fs/xfs/xfs_zone_alloc.c +5 -0
@@ -981,43 +981,43 @@ struct xfs_init_zones {
 	uint64_t		reclaimable;
 };
 
+/*
+ * For sequential write required zones, we restart writing at the hardware write
+ * pointer returned by xfs_zone_validate().
+ *
+ * For conventional zones or conventional devices we have query the rmap to

Nit:
s/we have query/we have to query (or "we must")

Quoted 6 lines
fs/xfs/xfs_zone_alloc.c +6 -0
+ * find the highest recorded block and set the write pointer to the block after
+ * that.  In case of a power loss this misses blocks where the data I/O has
+ * completed but not recorded in the rmap yet, and it also rewrites blocks if
+ * the most recently written ones got deleted again before unmount, but this is
+ * the best we can do without hardware support.
+ */
Signature · 2 lines
Damien Le Moal
Western Digital Research
maintainer Reviewed-by on v1 3/6
#8 lore raw

Re: [PATCH 3/6] xfs: pass the write pointer to xfs_init_zone

In reply to Christoph Hellwig
Quoted 122 lines · Christoph Hellwig

Move the two methods to query the write pointer out of xfs_init_zone into the callers, so that xfs_init_zone doesn't have to bother with the blk_zone structure and instead operates purely at the XFS realtime group level.

---
 fs/xfs/xfs_zone_alloc.c | 66 +++++++++++++++++++++++------------------
 1 file changed, 37 insertions(+), 29 deletions(-)
fs/xfs/xfs_zone_alloc.c +37 -29
index 4ca7769b5adb..87243644d88e 100644
@@ -981,43 +981,43 @@ struct xfs_init_zones {
 	uint64_t		reclaimable;
 };
 
+/*
+ * For sequential write required zones, we restart writing at the hardware write
+ * pointer returned by xfs_zone_validate().
+ *
+ * For conventional zones or conventional devices we have query the rmap to
+ * find the highest recorded block and set the write pointer to the block after
+ * that.  In case of a power loss this misses blocks where the data I/O has
+ * completed but not recorded in the rmap yet, and it also rewrites blocks if
+ * the most recently written ones got deleted again before unmount, but this is
+ * the best we can do without hardware support.
+ */
+static xfs_rgblock_t
+xfs_rmap_estimate_write_pointer(
+	struct xfs_rtgroup	*rtg)
+{
+	xfs_rgblock_t		highest_rgbno;
+
+	xfs_rtgroup_lock(rtg, XFS_RTGLOCK_RMAP);
+	highest_rgbno = xfs_rtrmap_highest_rgbno(rtg);
+	xfs_rtgroup_unlock(rtg, XFS_RTGLOCK_RMAP);
+
+	if (highest_rgbno == NULLRGBLOCK)
+		return 0;
+	return highest_rgbno + 1;
+}
+
 static int
 xfs_init_zone(
 	struct xfs_init_zones	*iz,
 	struct xfs_rtgroup	*rtg,
-	struct blk_zone		*zone)
+	xfs_rgblock_t		write_pointer)
 {
 	struct xfs_mount	*mp = rtg_mount(rtg);
 	struct xfs_zone_info	*zi = mp->m_zone_info;
 	uint32_t		used = rtg_rmap(rtg)->i_used_blocks;
-	xfs_rgblock_t		write_pointer, highest_rgbno;
 	int			error;
 
-	if (zone && !xfs_zone_validate(zone, rtg, &write_pointer))
-		return -EFSCORRUPTED;
-
-	/*
-	 * For sequential write required zones we retrieved the hardware write
-	 * pointer above.
-	 *
-	 * For conventional zones or conventional devices we don't have that
-	 * luxury.  Instead query the rmap to find the highest recorded block
-	 * and set the write pointer to the block after that.  In case of a
-	 * power loss this misses blocks where the data I/O has completed but
-	 * not recorded in the rmap yet, and it also rewrites blocks if the most
-	 * recently written ones got deleted again before unmount, but this is
-	 * the best we can do without hardware support.
-	 */
-	if (!zone || zone->cond == BLK_ZONE_COND_NOT_WP) {
-		xfs_rtgroup_lock(rtg, XFS_RTGLOCK_RMAP);
-		highest_rgbno = xfs_rtrmap_highest_rgbno(rtg);
-		if (highest_rgbno == NULLRGBLOCK)
-			write_pointer = 0;
-		else
-			write_pointer = highest_rgbno + 1;
-		xfs_rtgroup_unlock(rtg, XFS_RTGLOCK_RMAP);
-	}
-
 	/*
 	 * If there are no used blocks, but the zone is not in empty state yet
 	 * we lost power before the zoned reset.  In that case finish the work
@@ -1066,6 +1066,7 @@ xfs_get_zone_info_cb(
 	struct xfs_mount	*mp = iz->mp;
 	xfs_fsblock_t		zsbno = xfs_daddr_to_rtb(mp, zone->start);
 	xfs_rgnumber_t		rgno;
+	xfs_rgblock_t		write_pointer;
 	struct xfs_rtgroup	*rtg;
 	int			error;
 
@@ -1080,7 +1081,13 @@ xfs_get_zone_info_cb(
 		xfs_warn(mp, "realtime group not found for zone %u.", rgno);
 		return -EFSCORRUPTED;
 	}
-	error = xfs_init_zone(iz, rtg, zone);
+	if (!xfs_zone_validate(zone, rtg, &write_pointer)) {
+		xfs_rtgroup_rele(rtg);
+		return -EFSCORRUPTED;
+	}
+	if (zone->cond == BLK_ZONE_COND_NOT_WP)
+		write_pointer = xfs_rmap_estimate_write_pointer(rtg);
+	error = xfs_init_zone(iz, rtg, write_pointer);
 	xfs_rtgroup_rele(rtg);
 	return error;
 }
@@ -1290,7 +1297,8 @@ xfs_mount_zones(
 		struct xfs_rtgroup	*rtg = NULL;
 
 		while ((rtg = xfs_rtgroup_next(mp, rtg))) {
-			error = xfs_init_zone(&iz, rtg, NULL);
+			error = xfs_init_zone(&iz, rtg,
+					xfs_rmap_estimate_write_pointer(rtg));
 			if (error) {
 				xfs_rtgroup_rele(rtg);
 				goto out_free_zone_info;
Signature · 1 line
2.47.3

With Damien's comment in place:

Quoted 1 line
#9 lore raw

PATCH 4/6 xfs: split and refactor zone validation

In reply to Christoph Hellwig

Currently xfs_zone_validate mixes validating the software zone state in the XFS realtime group with validating the hardware state reported in struct blk_zone and deriving the write pointer from that.

Move all code that works on the realtime group to xfs_init_zone, and only keep the hardware state validation in xfs_zone_validate. This makes the code more clear, and allows for better reuse in userspace.

---
 fs/xfs/libxfs/xfs_zones.c | 149 ++++++++++----------------------------
 fs/xfs/libxfs/xfs_zones.h |   5 +-
 fs/xfs/xfs_zone_alloc.c   |  28 ++++++-
 3 files changed, 68 insertions(+), 114 deletions(-)
fs/xfs/libxfs/xfs_zones.c +39 -110
index 8c3c67caf64e..24e350c31933 100644
@@ -15,173 +15,102 @@
 #include "xfs_zones.h"
 
 static bool
-xfs_zone_validate_empty(
+xfs_validate_blk_zone_seq(
+	struct xfs_mount	*mp,
 	struct blk_zone		*zone,
-	struct xfs_rtgroup	*rtg,
+	unsigned int		zone_no,
 	xfs_rgblock_t		*write_pointer)
 {
-	struct xfs_mount	*mp = rtg_mount(rtg);
-
-	if (rtg_rmap(rtg)->i_used_blocks > 0) {
-		xfs_warn(mp, "empty zone %u has non-zero used counter (0x%x).",
-			 rtg_rgno(rtg), rtg_rmap(rtg)->i_used_blocks);
-		return false;
-	}
-
-	*write_pointer = 0;
-	return true;
-}
-
-static bool
-xfs_zone_validate_wp(
-	struct blk_zone		*zone,
-	struct xfs_rtgroup	*rtg,
-	xfs_rgblock_t		*write_pointer)
-{
-	struct xfs_mount	*mp = rtg_mount(rtg);
-	xfs_rtblock_t		wp_fsb = xfs_daddr_to_rtb(mp, zone->wp);
-
-	if (rtg_rmap(rtg)->i_used_blocks > rtg->rtg_extents) {
-		xfs_warn(mp, "zone %u has too large used counter (0x%x).",
-			 rtg_rgno(rtg), rtg_rmap(rtg)->i_used_blocks);
-		return false;
-	}
-
-	if (xfs_rtb_to_rgno(mp, wp_fsb) != rtg_rgno(rtg)) {
-		xfs_warn(mp, "zone %u write pointer (0x%llx) outside of zone.",
-			 rtg_rgno(rtg), wp_fsb);
-		return false;
-	}
-
-	*write_pointer = xfs_rtb_to_rgbno(mp, wp_fsb);
-	if (*write_pointer >= rtg->rtg_extents) {
-		xfs_warn(mp, "zone %u has invalid write pointer (0x%x).",
-			 rtg_rgno(rtg), *write_pointer);
-		return false;
-	}
-
-	return true;
-}
-
-static bool
-xfs_zone_validate_full(
-	struct blk_zone		*zone,
-	struct xfs_rtgroup	*rtg,
-	xfs_rgblock_t		*write_pointer)
-{
-	struct xfs_mount	*mp = rtg_mount(rtg);
-
-	if (rtg_rmap(rtg)->i_used_blocks > rtg->rtg_extents) {
-		xfs_warn(mp, "zone %u has too large used counter (0x%x).",
-			 rtg_rgno(rtg), rtg_rmap(rtg)->i_used_blocks);
-		return false;
-	}
-
-	*write_pointer = rtg->rtg_extents;
-	return true;
-}
-
-static bool
-xfs_zone_validate_seq(
-	struct blk_zone		*zone,
-	struct xfs_rtgroup	*rtg,
-	xfs_rgblock_t		*write_pointer)
-{
-	struct xfs_mount	*mp = rtg_mount(rtg);
-
 	switch (zone->cond) {
 	case BLK_ZONE_COND_EMPTY:
-		return xfs_zone_validate_empty(zone, rtg, write_pointer);
+		*write_pointer = 0;
+		return true;
 	case BLK_ZONE_COND_IMP_OPEN:
 	case BLK_ZONE_COND_EXP_OPEN:
 	case BLK_ZONE_COND_CLOSED:
 	case BLK_ZONE_COND_ACTIVE:
-		return xfs_zone_validate_wp(zone, rtg, write_pointer);
+		if (zone->wp < zone->start ||
+		    zone->wp >= zone->start + zone->capacity) {
+			xfs_warn(mp,
+	"zone %u write pointer (%llu) outside of zone.",
+				zone_no, zone->wp);
+			return false;
+		}
+
+		*write_pointer = XFS_BB_TO_FSB(mp, zone->wp - zone->start);
+		return true;
 	case BLK_ZONE_COND_FULL:
-		return xfs_zone_validate_full(zone, rtg, write_pointer);
+		*write_pointer = XFS_BB_TO_FSB(mp, zone->capacity);
+		return true;
 	case BLK_ZONE_COND_NOT_WP:
 	case BLK_ZONE_COND_OFFLINE:
 	case BLK_ZONE_COND_READONLY:
 		xfs_warn(mp, "zone %u has unsupported zone condition 0x%x.",
-			rtg_rgno(rtg), zone->cond);
+			zone_no, zone->cond);
 		return false;
 	default:
 		xfs_warn(mp, "zone %u has unknown zone condition 0x%x.",
-			rtg_rgno(rtg), zone->cond);
+			zone_no, zone->cond);
 		return false;
 	}
 }
 
 static bool
-xfs_zone_validate_conv(
+xfs_validate_blk_zone_conv(
+	struct xfs_mount	*mp,
 	struct blk_zone		*zone,
-	struct xfs_rtgroup	*rtg)
+	unsigned int		zone_no)
 {
-	struct xfs_mount	*mp = rtg_mount(rtg);
-
 	switch (zone->cond) {
 	case BLK_ZONE_COND_NOT_WP:
 		return true;
 	default:
 		xfs_warn(mp,
 "conventional zone %u has unsupported zone condition 0x%x.",
-			 rtg_rgno(rtg), zone->cond);
+			 zone_no, zone->cond);
 		return false;
 	}
 }
 
 bool
-xfs_zone_validate(
+xfs_validate_blk_zone(
+	struct xfs_mount	*mp,
 	struct blk_zone		*zone,
-	struct xfs_rtgroup	*rtg,
+	unsigned int		zone_no,
+	uint32_t		expected_size,
+	uint32_t		expected_capacity,
 	xfs_rgblock_t		*write_pointer)
 {
-	struct xfs_mount	*mp = rtg_mount(rtg);
-	struct xfs_groups	*g = &mp->m_groups[XG_TYPE_RTG];
-	uint32_t		expected_size;
-
 	/*
 	 * Check that the zone capacity matches the rtgroup size stored in the
 	 * superblock.  Note that all zones including the last one must have a
 	 * uniform capacity.
 	 */
-	if (XFS_BB_TO_FSB(mp, zone->capacity) != g->blocks) {
+	if (XFS_BB_TO_FSB(mp, zone->capacity) != expected_capacity) {
 		xfs_warn(mp,
-"zone %u capacity (0x%llx) does not match RT group size (0x%x).",
-			rtg_rgno(rtg), XFS_BB_TO_FSB(mp, zone->capacity),
-			g->blocks);
+"zone %u capacity (%llu) does not match RT group size (%u).",
+			zone_no, XFS_BB_TO_FSB(mp, zone->capacity),
+			expected_capacity);
 		return false;
 	}
 
-	if (g->has_daddr_gaps) {
-		expected_size = 1 << g->blklog;
-	} else {
-		if (zone->len != zone->capacity) {
-			xfs_warn(mp,
-"zone %u has capacity != size ((0x%llx vs 0x%llx)",
-				rtg_rgno(rtg),
-				XFS_BB_TO_FSB(mp, zone->len),
-				XFS_BB_TO_FSB(mp, zone->capacity));
-			return false;
-		}
-		expected_size = g->blocks;
-	}
-
 	if (XFS_BB_TO_FSB(mp, zone->len) != expected_size) {
 		xfs_warn(mp,
-"zone %u length (0x%llx) does match geometry (0x%x).",
-			rtg_rgno(rtg), XFS_BB_TO_FSB(mp, zone->len),
+"zone %u length (%llu) does not match geometry (%u).",
+			zone_no, XFS_BB_TO_FSB(mp, zone->len),
 			expected_size);
+		return false;
 	}
 
 	switch (zone->type) {
 	case BLK_ZONE_TYPE_CONVENTIONAL:
-		return xfs_zone_validate_conv(zone, rtg);
+		return xfs_validate_blk_zone_conv(mp, zone, zone_no);
 	case BLK_ZONE_TYPE_SEQWRITE_REQ:
-		return xfs_zone_validate_seq(zone, rtg, write_pointer);
+		return xfs_validate_blk_zone_seq(mp, zone, zone_no,
+				write_pointer);
 	default:
 		xfs_warn(mp, "zoned %u has unsupported type 0x%x.",
-			rtg_rgno(rtg), zone->type);
+			zone_no, zone->type);
 		return false;
 	}
 }
fs/xfs/libxfs/xfs_zones.h +3 -2
index df10a34da71d..c16089c9a652 100644
@@ -37,7 +37,8 @@ struct blk_zone;
  */
 #define XFS_DEFAULT_MAX_OPEN_ZONES	128
 
-bool xfs_zone_validate(struct blk_zone *zone, struct xfs_rtgroup *rtg,
-	xfs_rgblock_t *write_pointer);
+bool xfs_validate_blk_zone(struct xfs_mount *mp, struct blk_zone *zone,
+	unsigned int zone_no, uint32_t expected_size,
+	uint32_t expected_capacity, xfs_rgblock_t *write_pointer);
 
 #endif /* _LIBXFS_ZONES_H */
fs/xfs/xfs_zone_alloc.c +26 -2
index 87243644d88e..889a681d50b6 100644
@@ -977,13 +977,15 @@ xfs_free_open_zones(
 
 struct xfs_init_zones {
 	struct xfs_mount	*mp;
+	uint32_t		zone_size;
+	uint32_t		zone_capacity;
 	uint64_t		available;
 	uint64_t		reclaimable;
 };
 
 /*
  * For sequential write required zones, we restart writing at the hardware write
- * pointer returned by xfs_zone_validate().
+ * pointer returned by xfs_validate_blk_zone().
  *
  * For conventional zones or conventional devices we have query the rmap to
  * find the highest recorded block and set the write pointer to the block after
@@ -1018,6 +1020,25 @@ xfs_init_zone(
 	uint32_t		used = rtg_rmap(rtg)->i_used_blocks;
 	int			error;
 
+	if (write_pointer > rtg->rtg_extents) {
+		xfs_warn(mp, "zone %u has invalid write pointer (0x%x).",
+			 rtg_rgno(rtg), write_pointer);
+		return -EFSCORRUPTED;
+	}
+
+	if (used > rtg->rtg_extents) {
+		xfs_warn(mp,
+"zone %u has used counter (0x%x) larger than zone capacity (0x%llx).",
+			 rtg_rgno(rtg), used, rtg->rtg_extents);
+		return -EFSCORRUPTED;
+	}
+
+	if (write_pointer == 0 && used != 0) {
+		xfs_warn(mp, "empty zone %u has non-zero used counter (0x%x).",
+			rtg_rgno(rtg), used);
+		return -EFSCORRUPTED;
+	}
+
 	/*
 	 * If there are no used blocks, but the zone is not in empty state yet
 	 * we lost power before the zoned reset.  In that case finish the work
@@ -1081,7 +1102,8 @@ xfs_get_zone_info_cb(
 		xfs_warn(mp, "realtime group not found for zone %u.", rgno);
 		return -EFSCORRUPTED;
 	}
-	if (!xfs_zone_validate(zone, rtg, &write_pointer)) {
+	if (!xfs_validate_blk_zone(mp, zone, idx, iz->zone_size,
+			iz->zone_capacity, &write_pointer)) {
 		xfs_rtgroup_rele(rtg);
 		return -EFSCORRUPTED;
 	}
@@ -1227,6 +1249,8 @@ xfs_mount_zones(
 {
 	struct xfs_init_zones	iz = {
 		.mp		= mp,
+		.zone_capacity	= mp->m_groups[XG_TYPE_RTG].blocks,
+		.zone_size	= xfs_rtgroup_raw_size(mp),
 	};
 	struct xfs_buftarg	*bt = mp->m_rtdev_targp;
 	xfs_extlen_t		zone_blocks = mp->m_groups[XG_TYPE_RTG].blocks;
Signature · 1 line
2.47.3
Reviewed-by on v1 4/6
#10 lore raw

Re: [PATCH 4/6] xfs: split and refactor zone validation

Quoted 10 lines · Christoph Hellwig

Currently xfs_zone_validate mixes validating the software zone state in the XFS realtime group with validating the hardware state reported in struct blk_zone and deriving the write pointer from that.

Move all code that works on the realtime group to xfs_init_zone, and only keep the hardware state validation in xfs_zone_validate. This makes the code more clear, and allows for better reuse in userspace.

Signature · 2 lines
Damien Le Moal
Western Digital Research
#11 lore raw

PATCH 5/6 xfs: check that used blocks are smaller than the write pointer

In reply to Christoph Hellwig

Any used block must have been written, this reject used blocks > write pointer.

---
 fs/xfs/xfs_zone_alloc.c | 7 +++++++
 1 file changed, 7 insertions(+)
fs/xfs/xfs_zone_alloc.c +7 -0
index 889a681d50b6..5ad9ae368996 100644
@@ -1033,6 +1033,13 @@ xfs_init_zone(
 		return -EFSCORRUPTED;
 	}
 
+	if (used > write_pointer) {
+		xfs_warn(mp,
+"zone %u has used counter (0x%x) larger than write pointer (0x%x).",
+			 rtg_rgno(rtg), used, write_pointer);
+		return -EFSCORRUPTED;
+	}
+
 	if (write_pointer == 0 && used != 0) {
 		xfs_warn(mp, "empty zone %u has non-zero used counter (0x%x).",
 			rtg_rgno(rtg), used);
Signature · 1 line
2.47.3
#13 lore raw

PATCH 6/6 xfs: use blkdev_get_zone_info to simplify zone reporting

In reply to Christoph Hellwig

Unwind the callback based programming model by querying the cached zone information using blkdev_get_zone_info.

---
 fs/xfs/xfs_zone_alloc.c | 114 ++++++++++++++++++----------------------
 1 file changed, 50 insertions(+), 64 deletions(-)
fs/xfs/xfs_zone_alloc.c +50 -64
index 5ad9ae368996..211ec08cb508 100644
@@ -976,7 +976,6 @@ xfs_free_open_zones(
 }
 
 struct xfs_init_zones {
-	struct xfs_mount	*mp;
 	uint32_t		zone_size;
 	uint32_t		zone_capacity;
 	uint64_t		available;
@@ -994,19 +993,52 @@ struct xfs_init_zones {
  * the most recently written ones got deleted again before unmount, but this is
  * the best we can do without hardware support.
  */
-static xfs_rgblock_t
-xfs_rmap_estimate_write_pointer(
-	struct xfs_rtgroup	*rtg)
+static int
+xfs_query_write_pointer(
+	struct xfs_init_zones	*iz,
+	struct xfs_rtgroup	*rtg,
+	xfs_rgblock_t		*write_pointer)
 {
+	struct xfs_mount	*mp = rtg_mount(rtg);
+	struct block_device	*bdev = mp->m_rtdev_targp->bt_bdev;
+	sector_t		start = xfs_gbno_to_daddr(&rtg->rtg_group, 0);
 	xfs_rgblock_t		highest_rgbno;
+	struct blk_zone		zone = {};
+	int			error;
+
+	if (bdev_is_zoned(bdev)) {
+		error = blkdev_get_zone_info(bdev, start, &zone);
+		if (error)
+			return error;
+		if (zone.start != start) {
+			xfs_warn(mp, "mismatched zone start: 0x%llx/0x%llx.",
+				zone.start, start);
+			return -EFSCORRUPTED;
+		}
+
+		if (!xfs_validate_blk_zone(mp, &zone, rtg_rgno(rtg),
+				iz->zone_size, iz->zone_capacity,
+				write_pointer))
+			return -EFSCORRUPTED;
+
+		/*
+		 * Use the hardware write pointer returned by
+		 * xfs_validate_blk_zone for sequential write required zones,
+		 * else fall through to the rmap-based estimation below.
+		 */
+		if (zone.cond != BLK_ZONE_COND_NOT_WP)
+			return 0;
+	}
 
 	xfs_rtgroup_lock(rtg, XFS_RTGLOCK_RMAP);
 	highest_rgbno = xfs_rtrmap_highest_rgbno(rtg);
 	xfs_rtgroup_unlock(rtg, XFS_RTGLOCK_RMAP);
 
 	if (highest_rgbno == NULLRGBLOCK)
-		return 0;
-	return highest_rgbno + 1;
+		*write_pointer = 0;
+	else
+		*write_pointer = highest_rgbno + 1;
+	return 0;
 }
 
 static int
@@ -1084,43 +1116,6 @@ xfs_init_zone(
 	return 0;
 }
 
-static int
-xfs_get_zone_info_cb(
-	struct blk_zone		*zone,
-	unsigned int		idx,
-	void			*data)
-{
-	struct xfs_init_zones	*iz = data;
-	struct xfs_mount	*mp = iz->mp;
-	xfs_fsblock_t		zsbno = xfs_daddr_to_rtb(mp, zone->start);
-	xfs_rgnumber_t		rgno;
-	xfs_rgblock_t		write_pointer;
-	struct xfs_rtgroup	*rtg;
-	int			error;
-
-	if (xfs_rtb_to_rgbno(mp, zsbno) != 0) {
-		xfs_warn(mp, "mismatched zone start 0x%llx.", zsbno);
-		return -EFSCORRUPTED;
-	}
-
-	rgno = xfs_rtb_to_rgno(mp, zsbno);
-	rtg = xfs_rtgroup_grab(mp, rgno);
-	if (!rtg) {
-		xfs_warn(mp, "realtime group not found for zone %u.", rgno);
-		return -EFSCORRUPTED;
-	}
-	if (!xfs_validate_blk_zone(mp, zone, idx, iz->zone_size,
-			iz->zone_capacity, &write_pointer)) {
-		xfs_rtgroup_rele(rtg);
-		return -EFSCORRUPTED;
-	}
-	if (zone->cond == BLK_ZONE_COND_NOT_WP)
-		write_pointer = xfs_rmap_estimate_write_pointer(rtg);
-	error = xfs_init_zone(iz, rtg, write_pointer);
-	xfs_rtgroup_rele(rtg);
-	return error;
-}
-
 /*
  * Calculate the max open zone limit based on the of number of backing zones
  * available.
@@ -1255,15 +1250,13 @@ xfs_mount_zones(
 	struct xfs_mount	*mp)
 {
 	struct xfs_init_zones	iz = {
-		.mp		= mp,
 		.zone_capacity	= mp->m_groups[XG_TYPE_RTG].blocks,
 		.zone_size	= xfs_rtgroup_raw_size(mp),
 	};
-	struct xfs_buftarg	*bt = mp->m_rtdev_targp;
-	xfs_extlen_t		zone_blocks = mp->m_groups[XG_TYPE_RTG].blocks;
+	struct xfs_rtgroup	*rtg = NULL;
 	int			error;
 
-	if (!bt) {
+	if (!mp->m_rtdev_targp) {
 		xfs_notice(mp, "RT device missing.");
 		return -EINVAL;
 	}
@@ -1291,7 +1284,7 @@ xfs_mount_zones(
 		return -ENOMEM;
 
 	xfs_info(mp, "%u zones of %u blocks (%u max open zones)",
-		 mp->m_sb.sb_rgcount, zone_blocks, mp->m_max_open_zones);
+		 mp->m_sb.sb_rgcount, iz.zone_capacity, mp->m_max_open_zones);
 	trace_xfs_zones_mount(mp);
 
 	/*
@@ -1315,25 +1308,18 @@ xfs_mount_zones(
 	 * or beneficial.
 	 */
 	mp->m_super->s_min_writeback_pages =
-		XFS_FSB_TO_B(mp, min(zone_blocks, XFS_MAX_BMBT_EXTLEN)) >>
+		XFS_FSB_TO_B(mp, min(iz.zone_capacity, XFS_MAX_BMBT_EXTLEN)) >>
 			PAGE_SHIFT;
 
-	if (bdev_is_zoned(bt->bt_bdev)) {
-		error = blkdev_report_zones_cached(bt->bt_bdev,
-				XFS_FSB_TO_BB(mp, mp->m_sb.sb_rtstart),
-				mp->m_sb.sb_rgcount, xfs_get_zone_info_cb, &iz);
-		if (error < 0)
+	while ((rtg = xfs_rtgroup_next(mp, rtg))) {
+		xfs_rgblock_t		write_pointer;
+
+		error = xfs_query_write_pointer(&iz, rtg, &write_pointer);
+		if (!error)
+			error = xfs_init_zone(&iz, rtg, write_pointer);
+		if (error) {
+			xfs_rtgroup_rele(rtg);
 			goto out_free_zone_info;
-	} else {
-		struct xfs_rtgroup	*rtg = NULL;
-
-		while ((rtg = xfs_rtgroup_next(mp, rtg))) {
-			error = xfs_init_zone(&iz, rtg,
-					xfs_rmap_estimate_write_pointer(rtg));
-			if (error) {
-				xfs_rtgroup_rele(rtg);
-				goto out_free_zone_info;
-			}
 		}
 	}
 
Signature · 1 line
2.47.3
Reviewed-by on v1 6/6
#14 lore raw

Re: [PATCH 6/6] xfs: use blkdev_get_zone_info to simplify zone reporting

Quoted 5 lines · Christoph Hellwig

Unwind the callback based programming model by querying the cached zone information using blkdev_get_zone_info.

Nice cleanup !

Signature · 2 lines
Damien Le Moal
Western Digital Research
maintainer Applied
#15 lore raw
In reply to Christoph Hellwig
Quoted 9 lines · Christoph Hellwig

this series refactor the zone reporting code so that it is more clearly split between sanity checking the report hardware zone information, and the XFS zoned RT information. This reduced the code size and removes an iteration over all RTGs at boot time.

It will also allow to do smarter checking of hardware zones and RTG allocation information in repair once ported to userspace.

[...]

Applied to for-next, thanks!

[1/6] xfs: add missing forward declaration in xfs_zones.h

      commit: 41263267ef26d315b1425eb9c8a8d7092f9db7c8
[2/6] xfs: add a xfs_rtgroup_raw_size helper
      commit: fc633b5c5b80c1d840b7a8bc2828be96582c6b55
[3/6] xfs: pass the write pointer to xfs_init_zone
      commit: 776b76f7547fb839954aae06f58ac7b6b35c0b25
[4/6] xfs: split and refactor zone validation
      commit: 19c5b6051ed62d8c4b1cf92e463c1bcf629107f4
[5/6] xfs: check that used blocks are smaller than the write pointer
      commit: b37c1e4e9af795ac31ddc992b0461182c45705dc
[6/6] xfs: use blkdev_get_zone_info to simplify zone reporting
      commit: 12d12dcc1508874886ebcbd2aefba74f1ed71f98

Best regards,

Signature · 1 line
Carlos Maiolino <[email protected]>