Skip to contentSkip to conversation

nvme: fixup zns namespace initialisation

PATCH In review Hannes Reinecke 3 participants 3 messages May 24, 2024 NVMe

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 nvme: fixup zns namespace initialisation

nvme_set_chunk_sectors() needs to know whether it's a zoned namespace or not, but the information is only set in a later call. So move the calls around and ensure that the namespace is marked as non-zoned initially to avoid calculation errors when trying to derive a non-existing zone geometry.

Fixes c85c9ab926a5 ("nvme: split nvme_update_zone_info")
---
 drivers/nvme/host/core.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
drivers/nvme/host/core.c +3 -3
index 111bf4197052..a968d33914a7 100644
@@ -2115,13 +2115,13 @@ static int nvme_update_ns_info_block(struct nvme_ns *ns,
 	lim = queue_limits_start_update(ns->disk->queue);
 	nvme_set_ctrl_limits(ns->ctrl, &lim);
 	nvme_configure_metadata(ns->ctrl, ns->head, id, nvm);
+	if (IS_ENABLED(CONFIG_BLK_DEV_ZONED) &&
+	    ns->head->ids.csi == NVME_CSI_ZNS)
+		nvme_update_zone_info(ns, &lim, &zi);
 	nvme_set_chunk_sectors(ns, id, &lim);
 	if (!nvme_update_disk_info(ns, id, &lim))
 		capacity = 0;
 	nvme_config_discard(ns, &lim);
-	if (IS_ENABLED(CONFIG_BLK_DEV_ZONED) &&
-	    ns->head->ids.csi == NVME_CSI_ZNS)
-		nvme_update_zone_info(ns, &lim, &zi);
 	ret = queue_limits_commit_update(ns->disk->queue, &lim);
 	if (ret) {
 		blk_mq_unfreeze_queue(ns->disk->queue);
Signature · 1 line
2.35.3
maintainer Reviewed-by on v2 1/1
#1 lore raw
Quoted 2 lines · Hannes Reinecke

nvme_set_chunk_sectors() needs to know whether it's a zoned namespace or not, but the information is only set in a later call.

It doesn't really "need to know it". It's kinda handy to warn if a zone namespace sets an IOB.

Quoted 5 lines

So move the calls around and ensure that the namespace is marked as non-zoned initially to avoid calculation errors when trying to derive a non-existing zone geometry.

Fixes c85c9ab926a5 ("nvme: split nvme_update_zone_info")

Even before that commit, nvme_update_zone_info is clearly called only after nvme_set_chunk_sectors. The order goes back to 73d90386b559d6f4c3c5db5e6bb1b68aae8fd3e7 as far as I can tell.

Note that until the atomic limits series we kinda had to do it that way as nvme_update_zone_info needs the sector size and other earlier information.

The changes themselves looks fine, though:

Reviewed-by on v2 1/1
#2 lore raw
In reply to Hannes Reinecke
Quoted 23 lines · Hannes Reinecke

nvme_set_chunk_sectors() needs to know whether it's a zoned namespace or not, but the information is only set in a later call. So move the calls around and ensure that the namespace is marked as non-zoned initially to avoid calculation errors when trying to derive a non-existing zone geometry.

Fixes c85c9ab926a5 ("nvme: split nvme_update_zone_info")

---
drivers/nvme/host/core.c | 6 +++---

1 file changed, 3 insertions(+), 3 deletions(-)
drivers/nvme/host/core.c +3 -0
index 111bf4197052..a968d33914a7 100644
@@ -2115,13 +2115,13 @@ static int nvme_update_ns_info_block(struct nvme_ns *ns,
 	lim = queue_limits_start_update(ns->disk->queue);
 	nvme_set_ctrl_limits(ns->ctrl, &lim);
 	nvme_configure_metadata(ns->ctrl, ns->head, id, nvm);
+	if (IS_ENABLED(CONFIG_BLK_DEV_ZONED) &&
+	    ns->head->ids.csi == NVME_CSI_ZNS)
+		nvme_update_zone_info(ns, &lim, &zi);

drivers/nvme/host/core.c static int nvme_update_ns_info_block(struct nvme_ns *ns, · line 2127

Quoted 10 lines
drivers/nvme/host/core.c +0 -3
 	nvme_set_chunk_sectors(ns, id, &lim);
 	if (!nvme_update_disk_info(ns, id, &lim))
 		capacity = 0;
 	nvme_config_discard(ns, &lim);
-	if (IS_ENABLED(CONFIG_BLK_DEV_ZONED) &&
-	    ns->head->ids.csi == NVME_CSI_ZNS)
-		nvme_update_zone_info(ns, &lim, &zi);
 	ret = queue_limits_commit_update(ns->disk->queue, &lim);
 	if (ret) {
 		blk_mq_unfreeze_queue(ns->disk->queue);