Skip to contentSkip to conversation

nvme: always treat DNR status as no-retryable

PATCH In review Hannes Reinecke 3 participants 6 messages Nov 26, 2019 NVMe

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 nvme: always treat DNR status as no-retryable

If the DNR bit is set in the command status we should not retry the command, irrespective of what the actual status is. So map it directly to BLK_STS_TARGET to inform upper layers to not retry the command, not even on another path.

---
 drivers/nvme/host/core.c | 3 +++
 1 file changed, 3 insertions(+)
drivers/nvme/host/core.c +3 -0
index 752a40daf2b3..f015f6e91c85 100644
@@ -198,6 +198,9 @@ static inline bool nvme_ns_has_pi(struct nvme_ns *ns)
 
 static blk_status_t nvme_error_status(u16 status)
 {
+	if (status & NVME_SC_DNR)
+		return BLK_STS_TARGET;
+
 	switch (status & 0x7ff) {
 	case NVME_SC_SUCCESS:
 		return BLK_STS_OK;
Signature · 7 lines
2.16.4


_______________________________________________
linux-nvme mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-nvme
maintainer
#1 lore raw
Quoted 4 lines · Hannes Reinecke

If the DNR bit is set in the command status we should not retry the command, irrespective of what the actual status is. So map it directly to BLK_STS_TARGET to inform upper layers to not retry the command, not even on another path.

Why can't a DNR error be path specific, like if I detach a namespace from one of the controllers?

Quoted 11 lines
---
 drivers/nvme/host/core.c | 3 +++
 1 file changed, 3 insertions(+)
drivers/nvme/host/core.c +0 -0
index 752a40daf2b3..f015f6e91c85 100644

drivers/nvme/host/core.c static inline bool nvme_ns_has_pi(struct nvme_ns *ns) · line 206

Quoted 9 lines
drivers/nvme/host/core.c +3 -0
@@ -198,6 +198,9 @@ static inline bool nvme_ns_has_pi(struct nvme_ns *ns)
 
 static blk_status_t nvme_error_status(u16 status)
 {
+	if (status & NVME_SC_DNR)
+		return BLK_STS_TARGET;
+
 	switch (status & 0x7ff) {
 	case NVME_SC_SUCCESS:
 		return BLK_STS_OK;

_______________________________________________
linux-nvme mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-nvme

#2 lore raw
Quoted 9 lines · Keith Busch
Quoted 4 lines · Hannes Reinecke

If the DNR bit is set in the command status we should not retry the command, irrespective of what the actual status is. So map it directly to BLK_STS_TARGET to inform upper layers to not retry the command, not even on another path.

Why can't a DNR error be path specific, like if I detach a namespace from one of the controllers?

I've discussed the very same question with Fred Knight, and according to him the 'DNR' bit is _command_ specific, ie _this particular command_ should not be retried. Any eventual alternative path should not be considered.

Cheers,

Hannes

Signature · 10 lines
Dr. Hannes Reinecke            Teamlead Storage & Networking
[email protected]                              +49 911 74053 688
SUSE LINUX GmbH, Maxfeldstr. 5, 90409 Nürnberg
GF: Felix Imendörffer, Mary Higgins, Sri Rasiah
HRB 21284 (AG Nürnberg)

_______________________________________________
linux-nvme mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-nvme
#3 lore raw
Quoted 7 lines · Hannes Reinecke
Quoted 6 lines · Keith Busch
Quoted 4 lines · Hannes Reinecke

If the DNR bit is set in the command status we should not retry the command, irrespective of what the actual status is. So map it directly to BLK_STS_TARGET to inform upper layers to not retry the command, not even on another path.

Quoted 7 lines
Quoted 3 lines

Why can't a DNR error be path specific, like if I detach a namespace from one of the controllers?

I've discussed the very same question with Fred Knight, and according to him the 'DNR' bit is _command_ specific, ie _this particular command_ should not be retried.
Any eventual alternative path should not be considered.

To quote from the spec:

"Do Not Retry (DNR): If set to ‘1’, indicates that if the same command is re-submitted to any controller in the NVM subsystem, then that re-submitted command is expected to fail. If cleared to ‘0’, indicates that the same command may succeed if retried."

So yes, we should generally not submit it on another path, which is in fact what we already do if you look at nvme_complete_rq and nvme_req_needs_retry.

Besides not not needing the BLK_STS_TARGET remap that remap also is a horrible bad idea given that is loses way too much information.

_______________________________________________
linux-nvme mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-nvme

#4 lore raw
In reply to Hannes Reinecke
Quoted 4 lines · Hannes Reinecke

If the DNR bit is set in the command status we should not retry the command, irrespective of what the actual status is. So map it directly to BLK_STS_TARGET to inform upper layers to not retry the command, not even on another path.

We've been there before, have we? What is your use case. I.e. what part of the kernel is retrying what command that you don't want to be retried?

_______________________________________________
linux-nvme mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-nvme

#5 lore raw
Quoted 10 lines · Christoph Hellwig
Quoted 4 lines · Hannes Reinecke

If the DNR bit is set in the command status we should not retry the command, irrespective of what the actual status is. So map it directly to BLK_STS_TARGET to inform upper layers to not retry the command, not even on another path.

We've been there before, have we? What is your use case. I.e. what part of the kernel is retrying what command that you don't want to be retried?

EG Namespace not ready.
Some NetApp arrays return "namespace not ready" for either a temporary issue (eg during array failover or whatnot), and then will have the DNR bit not set (as you can retry, and eventually the namespace will become ready). Or they return 'namepace not ready' for cases where the namespace has been unmapped, for which is retry is pointless, and hence have the DNR bit set.

Cheers,

Hannes

Signature · 10 lines
Dr. Hannes Reinecke            Teamlead Storage & Networking
[email protected]                              +49 911 74053 688
SUSE LINUX GmbH, Maxfeldstr. 5, 90409 Nürnberg
GF: Felix Imendörffer, Mary Higgins, Sri Rasiah
HRB 21284 (AG Nürnberg)

_______________________________________________
linux-nvme mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-nvme