Skip to contentSkip to conversation

mkfs: fix zone capacity check for sequential zones

PATCH Superseded [email protected] 2 participants 5 messages Nov 12, 2025

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 mkfs: fix zone capacity check for sequential zones

From: Carlos Maiolino <[email protected]>

Sequential zones can have a different, smaller capacity than conventional zones.

Currently mkfs assumes both sequential and conventional zones will have the same capacity and and set the zone_info to the capacity of the first found zone and use that value to validate all the remaining zones's capacity.

Because conventional zones can't have a different capacity than its size, the first zone always have the largest possible capacity, so, mkfs will fail to validate any consecutive sequential zone if its capacity is smaller than the conventional zones.

What we should do instead, is set the zone info capacity accordingly to the settings of first zone found of the respective type and validate the capacity based on that instead of assuming all zones will have the same capacity.

---

FWIS, writing the patch description I'm assuming that the conventional zones always come first and we can't have a zoned device starting with sequential zones followed by conventional zones.

 mkfs/xfs_mkfs.c | 50 +++++++++++++++++++++++++++++++------------------
 1 file changed, 32 insertions(+), 18 deletions(-)
mkfs/xfs_mkfs.c +32 -18
index cd4f3ba4a549..1378e788eb95 100644
@@ -2545,6 +2545,28 @@ struct zone_topology {
 /* random size that allows efficient processing */
 #define ZONES_PER_IOCTL			16384
 
+static void
+zone_validate_capacity(
+	struct zone_info	*zi,
+	__u64			capacity,
+	bool			conventional)
+{
+	if (conventional && (zi->zone_capacity != zi->zone_size)) {
+		fprintf(stderr, _("Zone capacity equal to Zone size required for conventional zones.\n"));
+		exit(1);
+	}
+
+	if (zi->zone_capacity > zi->zone_size) {
+		fprintf(stderr, _("Zone capacity larger than zone size!\n"));
+		exit(1);
+	}
+
+	if (zi->zone_capacity != capacity) {
+		fprintf(stderr, _("Inconsistent zone capacity!\n"));
+		exit(1);
+	}
+}
+
 static void
 report_zones(
 	const char		*name,
@@ -2621,6 +2643,11 @@ _("Inconsistent zone size!\n"));
 
 			switch (zones[i].type) {
 			case BLK_ZONE_TYPE_CONVENTIONAL:
+				if (!zi->zone_capacity)
+					zi->zone_capacity = zones[i].capacity;
+				zone_validate_capacity(zi, zones[i].capacity,
+						       true);
+
 				/*
 				 * We can only use the conventional space at the
 				 * start of the device for metadata, so don't
@@ -2632,6 +2659,11 @@ _("Inconsistent zone size!\n"));
 					zi->nr_conv_zones++;
 				break;
 			case BLK_ZONE_TYPE_SEQWRITE_REQ:
+				if (!found_seq)
+					zi->zone_capacity = zones[i].capacity;
+				zone_validate_capacity(zi, zones[i].capacity,
+						       false);
+
 				found_seq = true;
 				break;
 			case BLK_ZONE_TYPE_SEQWRITE_PREF:
@@ -2644,19 +2676,6 @@ _("Unknown zone type (0x%x) found.\n"), zones[i].type);
 				exit(1);
 			}
 
-			if (!n) {
-				zi->zone_capacity = zones[i].capacity;
-				if (zi->zone_capacity > zi->zone_size) {
-					fprintf(stderr,
-_("Zone capacity larger than zone size!\n"));
-					exit(1);
-				}
-			} else if (zones[i].capacity != zi->zone_capacity) {
-				fprintf(stderr,
-_("Inconsistent zone capacity!\n"));
-				exit(1);
-			}
-
 			n++;
 		}
 		sector = zones[rep->nr_zones - 1].start +
@@ -2683,11 +2702,6 @@ validate_zoned(
 _("Data devices requires conventional zones.\n"));
 				usage();
 			}
-			if (zt->data.zone_capacity != zt->data.zone_size) {
-				fprintf(stderr,
-_("Zone capacity equal to Zone size required for conventional zones.\n"));
-				usage();
-			}
 
 			cli->sb_feat.zoned = true;
 			cfg->rtstart =
Signature · 1 line
2.51.0
Reviewed-by on v1 1/1
#1 lore raw
Quoted 3 lines · On Wed, Nov 12, 2025 at 01:33:16PM +0100, [email protected] wrote:

FWIS, writing the patch description I'm assuming that the conventional zones always come first and we can't have a zoned device starting with sequential zones followed by conventional zones.

XFS kinda assumes this by not counting counting conventional zones found after sequential zones, so we'll error out due to the lack of conventional space later unless a separate device is used.

mkfs/xfs_mkfs.c struct zone_topology { · line 2554

Quoted 7 lines
mkfs/xfs_mkfs.c +7 -0
+static void
+zone_validate_capacity(
+	struct zone_info	*zi,
+	__u64			capacity,
+	bool			conventional)
+{
+	if (conventional && (zi->zone_capacity != zi->zone_size)) {

No need for the inner braces.

Otherwise this looks good:

Note that we should also verify that if a device is used purely as RT device (aka no internal RT device), that there are either no conventional zones, or zone capacity == zone_size. I can look into that as an incremental patch.

#2 lore raw
Quoted 19 lines · Christoph Hellwig
Quoted 3 lines · On Wed, Nov 12, 2025 at 01:33:16PM +0100, [email protected] wrote:

FWIS, writing the patch description I'm assuming that the conventional zones always come first and we can't have a zoned device starting with sequential zones followed by conventional zones.

XFS kinda assumes this by not counting counting conventional zones found after sequential zones, so we'll error out due to the lack of conventional space later unless a separate device is used.

mkfs/xfs_mkfs.c struct zone_topology { · line 2554

Quoted 7 lines
mkfs/xfs_mkfs.c +7 -0
+static void
+zone_validate_capacity(
+	struct zone_info	*zi,
+	__u64			capacity,
+	bool			conventional)
+{
+	if (conventional && (zi->zone_capacity != zi->zone_size)) {

No need for the inner braces.

Quoted 8 lines

Otherwise this looks good:

Note that we should also verify that if a device is used purely as RT device (aka no internal RT device), that there are either no conventional zones, or zone capacity == zone_size. I can look into that as an incremental patch.

Your call. I can add it to the V2, giving I'll rewrite the above to remove the unneeded braces.

Couldn't we leverage this check in validate_zoned() for that?

if (cli->xi->rt.name && !cli->xi->rt.isfile) {
	[...]

I'm not sure though if just detecting a rtdev was passed on the CLI would suffice to cover what you proposed.

Quoted 1 line
#3 lore raw
Quoted 2 lines · Carlos Maiolino

Your call. I can add it to the V2, giving I'll rewrite the above to remove the unneeded braces.

Nah, let's do one thing at a time.

#4 lore raw
Quoted 6 lines · Christoph Hellwig
Quoted 2 lines · Carlos Maiolino

Your call. I can add it to the V2, giving I'll rewrite the above to remove the unneeded braces.

Nah, let's do one thing at a time.

okie!