Skip to contentSkip to conversation

nvme-fabrics: allow to queue requests for live queues

PATCH Applied Sagi Grimberg 4 participants 14 messages Jul 28, 2020 – Jul 29, 2020 NVMe

v1 of 2 Applied Latest is v2 →

lore
Share & tools
lore

Sagi Grimberg

author maintainer
#0 lore raw

PATCH 1/1 nvme-fabrics: allow to queue requests for live queues

Right now we are failing requests based on the controller state (which is checked inline in nvmf_check_ready) however we should definitely accept requests if the queue is live.

When entering controller reset, we transition the controller into NVME_CTRL_RESETTING, and then return BLK_STS_RESOURCE for non-mpath requests (have blk_noretry_request set).

This is also the case for NVME_REQ_USER for some reason, but there shouldn't be any reason for us to reject this I/O in a controller reset.

In a non-mpath setup, this means that the requests will simply be requeued over and over forever not allowing the q_usage_counter to drop its final reference, causing controller reset to hang if running concurrently with heavy I/O.

While we are at it, remove the redundant NVME_CTRL_NEW case, which should never see any I/O as it must first transition to NVME_CTRL_CONNECTING.

Fixes 35897b920c8a ("nvme-fabrics: fix and refine state checks in __nvmf_check_ready")
---
 drivers/nvme/host/fabrics.c | 10 +---------
 1 file changed, 1 insertion(+), 9 deletions(-)
drivers/nvme/host/fabrics.c +1 -9
index 4ec4829d6233..2e7838f42e36 100644
@@ -564,21 +564,13 @@ bool __nvmf_check_ready(struct nvme_ctrl *ctrl, struct request *rq,
 {
 	struct nvme_request *req = nvme_req(rq);
 
-	/*
-	 * If we are in some state of setup or teardown only allow
-	 * internally generated commands.
-	 */
-	if (!blk_rq_is_passthrough(rq) || (req->flags & NVME_REQ_USERCMD))
-		return false;
-
 	/*
 	 * Only allow commands on a live queue, except for the connect command,
 	 * which is require to set the queue live in the appropinquate states.
 	 */
 	switch (ctrl->state) {
-	case NVME_CTRL_NEW:
 	case NVME_CTRL_CONNECTING:
-		if (nvme_is_fabrics(req->cmd) &&
+		if (blk_rq_is_passthrough(rq) && nvme_is_fabrics(req->cmd) &&
 		    req->cmd->fabrics.fctype == nvme_fabrics_type_connect)
 			return true;
 		break;
Signature · 7 lines
2.25.1


_______________________________________________
Linux-nvme mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-nvme
#1 lore raw
Quoted 32 lines · Sagi Grimberg

Right now we are failing requests based on the controller state (which is checked inline in nvmf_check_ready) however we should definitely accept requests if the queue is live.

When entering controller reset, we transition the controller into NVME_CTRL_RESETTING, and then return BLK_STS_RESOURCE for non-mpath requests (have blk_noretry_request set).

This is also the case for NVME_REQ_USER for some reason, but there shouldn't be any reason for us to reject this I/O in a controller reset.

In a non-mpath setup, this means that the requests will simply be requeued over and over forever not allowing the q_usage_counter to drop its final reference, causing controller reset to hang if running concurrently with heavy I/O.

While we are at it, remove the redundant NVME_CTRL_NEW case, which should never see any I/O as it must first transition to NVME_CTRL_CONNECTING.

Fixes 35897b920c8a ("nvme-fabrics: fix and refine state checks in __nvmf_check_ready")
---
  drivers/nvme/host/fabrics.c | 10 +---------
  1 file changed, 1 insertion(+), 9 deletions(-)
drivers/nvme/host/fabrics.c +0 -0
index 4ec4829d6233..2e7838f42e36 100644

drivers/nvme/host/fabrics.c bool __nvmf_check_ready(struct nvme_ctrl *ctrl, struct request *rq, · line 571

Quoted 8 lines
drivers/nvme/host/fabrics.c +0 -5
@@ -564,21 +564,13 @@ bool __nvmf_check_ready(struct nvme_ctrl *ctrl, struct request *rq,
  {
  	struct nvme_request *req = nvme_req(rq);
 
-	/*
-	 * If we are in some state of setup or teardown only allow
-	 * internally generated commands.
-	 */
-	if (!blk_rq_is_passthrough(rq) || (req->flags & NVME_REQ_USERCMD))

"if (!blk_rq_is_passthrough(rq))" should not delete. Because if we delete, the normal io will be send to target, the target can not treat the io if the queue is not NVME_CTRL_LIVE.

drivers/nvme/host/fabrics.c bool __nvmf_check_ready(struct nvme_ctrl *ctrl, struct request *rq, · line 573

Quoted 11 lines
drivers/nvme/host/fabrics.c +1 -4
-		return false;
-
  	/*
  	 * Only allow commands on a live queue, except for the connect command,
  	 * which is require to set the queue live in the appropinquate states.
  	 */
  	switch (ctrl->state) {
-	case NVME_CTRL_NEW:
  	case NVME_CTRL_CONNECTING:
-		if (nvme_is_fabrics(req->cmd) &&
+		if (blk_rq_is_passthrough(rq) && nvme_is_fabrics(req->cmd) &&

If nvme_is_fabrics(req->cmd) is true, blk_rq_is_passthrough(rq) must be ture. why need add check blk_rq_is_passthrough(rq)?

drivers/nvme/host/fabrics.c bool __nvmf_check_ready(struct nvme_ctrl *ctrl, struct request *rq, · line 576

Quoted 4 lines
drivers/nvme/host/fabrics.c +0 -0
  		    req->cmd->fabrics.fctype == nvme_fabrics_type_connect)
  			return true;
  		break;

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

author maintainer
#2 lore raw
Quoted 17 lines
Quoted 6 lines
drivers/nvme/host/fabrics.c +0 -0
index 4ec4829d6233..2e7838f42e36 100644

drivers/nvme/host/fabrics.c bool __nvmf_check_ready(struct nvme_ctrl *ctrl, · line 571

Quoted 8 lines
drivers/nvme/host/fabrics.c +0 -5
@@ -564,21 +564,13 @@ bool __nvmf_check_ready(struct nvme_ctrl *ctrl, 
 struct request *rq,
   {
       struct nvme_request *req = nvme_req(rq);
-    /*
-     * If we are in some state of setup or teardown only allow
-     * internally generated commands.
-     */
-    if (!blk_rq_is_passthrough(rq) || (req->flags & NVME_REQ_USERCMD))

"if (!blk_rq_is_passthrough(rq))" should not delete. Because if we delete, the normal io will be send to target, the target can not treat the io if the queue is not NVME_CTRL_LIVE.

Sure it does, the only reason for us to deny this I/O, is if the queue is not live. The controller state should only _advise_ us if we need to look at the queue state.

Quoted 11 lines
Quoted 11 lines
drivers/nvme/host/fabrics.c +0 -3
-        return false;
-
       /*
        * Only allow commands on a live queue, except for the connect command,
        * which is require to set the queue live in the appropinquate states.
        */
       switch (ctrl->state) {
-    case NVME_CTRL_NEW:

      case NVME_CTRL_CONNECTING:
-        if (nvme_is_fabrics(req->cmd) &&
+        if (blk_rq_is_passthrough(rq) && nvme_is_fabrics(req->cmd) &&

Quoted 2 lines

If nvme_is_fabrics(req->cmd) is true, blk_rq_is_passthrough(rq) must be ture. why need add check blk_rq_is_passthrough(rq)?

req->cmd is only valid for passthru commands, hence we need to check for that before we reference req->cmd.

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

#3 lore raw
Quoted 18 lines · Sagi Grimberg
Quoted 17 lines
Quoted 6 lines
drivers/nvme/host/fabrics.c +0 -0
index 4ec4829d6233..2e7838f42e36 100644

drivers/nvme/host/fabrics.c bool __nvmf_check_ready(struct nvme_ctrl · line 571

Quoted 8 lines
drivers/nvme/host/fabrics.c +0 -5
@@ -564,21 +564,13 @@ bool __nvmf_check_ready(struct nvme_ctrl 
 *ctrl, struct request *rq,
   {
       struct nvme_request *req = nvme_req(rq);
-    /*
-     * If we are in some state of setup or teardown only allow
-     * internally generated commands.
-     */
-    if (!blk_rq_is_passthrough(rq) || (req->flags & NVME_REQ_USERCMD))

"if (!blk_rq_is_passthrough(rq))" should not delete. Because if we delete, the normal io will be send to target, the target can not treat the io if the queue is not NVME_CTRL_LIVE.

Quoted 3 lines

Sure it does, the only reason for us to deny this I/O, is if the queue is not live. The controller state should only _advise_ us if we need to look at the queue state.

I disagree strongly with removing the check on NVME_REQ_USERCMD. We've seen cli ioctls going to the admin queue while we're in the middle of doing controller initialization and it's has hosed the controller state in some cases. We've seen commands issued before the controller is in the proper state.  The admin queue may be live - but we don't necessarily want other io sneaking in.

As for the blk_rq_is_passthrough check - I guess I can see it being based on the queue state, and the check looks ok  (we should never see !blk_rq_is_passthrough on the admin q). But...
- I don't know why it was that important to change it. On the connecting path, all you're doing is letting io start flowing before all the queues have been created.  Did you really need to start that much sooner ?

- But on the resetting path, or deleting cases, you've added a condition now where the controller state was changed, but there was a delay before the transport marked the queue live. It's common practice in the transports to change state then schedule a work element to perform the actual state change.  Why would you want io to continue to flow during that window ?   This may bring out other problems we've avoided in the past.

-- james

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

author maintainer
#4 lore raw
Quoted 18 lines
Quoted 18 lines
Quoted 17 lines
Quoted 6 lines
drivers/nvme/host/fabrics.c +0 -0
index 4ec4829d6233..2e7838f42e36 100644

drivers/nvme/host/fabrics.c bool __nvmf_check_ready(struct nvme_ctrl · line 571

Quoted 8 lines
drivers/nvme/host/fabrics.c +0 -5
@@ -564,21 +564,13 @@ bool __nvmf_check_ready(struct nvme_ctrl 
 *ctrl, struct request *rq,
   {
       struct nvme_request *req = nvme_req(rq);
-    /*
-     * If we are in some state of setup or teardown only allow
-     * internally generated commands.
-     */
-    if (!blk_rq_is_passthrough(rq) || (req->flags & NVME_REQ_USERCMD))

"if (!blk_rq_is_passthrough(rq))" should not delete. Because if we delete, the normal io will be send to target, the target can not treat the io if the queue is not NVME_CTRL_LIVE.

Quoted 5 lines
Quoted 3 lines

Sure it does, the only reason for us to deny this I/O, is if the queue is not live. The controller state should only _advise_ us if we need to look at the queue state.

I disagree strongly with removing the check on NVME_REQ_USERCMD. We've seen cli ioctls going to the admin queue while we're in the middle of doing controller initialization and it's has hosed the controller state in some cases. We've seen commands issued before the controller is in the proper state.  The admin queue may be live - but we don't necessarily want other io sneaking in.

Can you please give an example? NVME_REQ_USERCMD should not be any different from any other type of I/O.

Also, do note that pci does allow to queue any type of command based on the queue state only. fabrics should be slightly different because we have the CONNECTING state where we want to let the connect command only to be issued.

Quoted 4 lines

As for the blk_rq_is_passthrough check - I guess I can see it being based on the queue state, and the check looks ok  (we should never see !blk_rq_is_passthrough on the admin q). But...
- I don't know why it was that important to change it. On the connecting
path, all you're doing is letting io start flowing before all the queues have been created.  Did you really need to start that much sooner ?

The issue is that controller in RESETTING state will have requests that are being issued, and if we don't let it pass through, it will hang around forever being requeued preventing queue freeze to complete.

See bug report from Krishnamraju Eraparaju.

Quoted 3 lines

- But on the resetting path, or deleting cases, you've added a condition
now where the controller state was changed, but there was a delay before the transport marked the queue live. It's common practice in the transports to change state then schedule a work element to perform the actual state change.  Why would you want io to continue to flow during that window ?   This may bring out other problems we've avoided in the past.

What are you referring to? the change here? the controller reset must allow requests that came in before we started queue freeze to pass, otherwise freeze will never complete.

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

#5 lore raw
Quoted 19 lines · Sagi Grimberg
Quoted 19 lines
Quoted 19 lines
Quoted 18 lines
Quoted 6 lines
drivers/nvme/host/fabrics.c +0 -0
index 4ec4829d6233..2e7838f42e36 100644

drivers/nvme/host/fabrics.c bool __nvmf_check_ready(struct nvme_ctrl · line 567

Quoted 9 lines
drivers/nvme/host/fabrics.c +0 -5
@@ -564,21 +564,13 @@ bool __nvmf_check_ready(struct nvme_ctrl 
 *ctrl, struct request *rq,
   {
       struct nvme_request *req = nvme_req(rq);
-    /*
-     * If we are in some state of setup or teardown only allow
-     * internally generated commands.
-     */
-    if (!blk_rq_is_passthrough(rq) || (req->flags & 
 NVME_REQ_USERCMD))

"if (!blk_rq_is_passthrough(rq))" should not delete. Because if we delete, the normal io will be send to target, the target can not treat the io if the queue is not NVME_CTRL_LIVE.

Quoted 8 lines
Quoted 5 lines
Quoted 3 lines

Sure it does, the only reason for us to deny this I/O, is if the queue is not live. The controller state should only _advise_ us if we need to look at the queue state.

I disagree strongly with removing the check on NVME_REQ_USERCMD. We've seen cli ioctls going to the admin queue while we're in the middle of doing controller initialization and it's has hosed the controller state in some cases. We've seen commands issued before the controller is in the proper state.  The admin queue may be live - but we don't necessarily want other io sneaking in.

Can you please give an example? NVME_REQ_USERCMD should not be any different from any other type of I/O.

Keep in mind  - in order to send any command, Fabric connect included, the admin queue must be unquiesced.

Here's what we have today:
At the time of the reset or immediately during the reset, an ioctl is done - it can (and has been seen) to be queued on the request queue as there are not state checks. When this io gets scheduled, as of the resetting or even the connecting state - it's currently failing the NVME_REQ_USERCMD check, which fails back into nvmf_fail_nonready_command, which fails back with BLK_STS_RESOURCE, which reschedules it.  This rescheduling reoccurs, until the its tried again with the controller state changed (usually LIVE) which then passes the admin-queue live check. It's good to go.

With your patch:
The ioctl will at least be requeued with BLK_STS_RESOURCE until after the Fabric Connect is completed (Connect is explicitly allowed by the state-specific check) as hopefully queue->live is false. It may actually be issued if we're in this resetting state only to be terminated afterward. All of the transports set the queue live flag as of Fabric Connect complete. If the ioctl is retried as of that point - it may be immediately sent on the admin queue - before or in-between any of the commands issued by nvme_enable_ctrl() - including before controller enabled, nvme_init_identify(), and before nvme_start_ctrl() has been called.    Hopefully none of these commands try to set/change configuration.   They simply shouldn't be allowed until after the controller has transitioned to LIVE. There's no need to allow them in-between.

Quoted 6 lines

Also, do note that pci does allow to queue any type of command based on the queue state only. fabrics should be slightly different because we have the CONNECTING state where we want to let the connect command only to be issued.

That doesn't make me feel any better.  PCI bit-bangs register writes, the odds they can fail/be discarded on the "fabric" is near zero. Admin Commands - both sending/processing/responding is much much faster than on a fabric, where the cmd can be dropped/delayed by fabric routing. Odds of having core-generated commands fail while reconnecting/reinitializing are much much lower.   All says - fabrics have a much higher command and inter-command latency than pcie does - so pcie is going to have a harder time seeing these cases.

Quoted 9 lines
Quoted 5 lines

As for the blk_rq_is_passthrough check - I guess I can see it being based on the queue state, and the check looks ok  (we should never see !blk_rq_is_passthrough on the admin q). But...
- I don't know why it was that important to change it. On the
connecting path, all you're doing is letting io start flowing before all the queues have been created.  Did you really need to start that much sooner ?

The issue is that controller in RESETTING state will have requests that are being issued, and if we don't let it pass through, it will hang around forever being requeued preventing queue freeze to complete.

is it really "forever" or is it until controller loss timeout ? granted 10minutes does seem like forever

Quoted 2 lines

See bug report from Krishnamraju Eraparaju.

I'll go look.

Quoted 8 lines
Quoted 4 lines

- But on the resetting path, or deleting cases, you've added a
condition now where the controller state was changed, but there was a delay before the transport marked the queue live. It's common practice in the transports to change state then schedule a work element to perform the actual state change.  Why would you want io to continue to flow during that window ?   This may bring out other problems we've avoided in the past.

What are you referring to? the change here? the controller reset must allow requests that came in before we started queue freeze to pass, otherwise freeze will never complete.

I'm referring to the io's that now pass through with the patch as the controller state is no longer blocking them (e.g. the removal of the !blk_rq_is_passthrough and NVME_REQ_USERCMD checks). They will pass the ready checks and go out on the wire, creating more active io that has to be deleted when the reset finally quiesces and terminates the active io list. For FC, we have to real link traffic for any io that gets started, so it raises the demands.   Should be fine, but it is creating more load at a point where it could have been blocked.

-- james

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

author maintainer
#6 lore raw
Quoted 11 lines
Quoted 3 lines

Can you please give an example? NVME_REQ_USERCMD should not be any different from any other type of I/O.

Keep in mind  - in order to send any command, Fabric connect included, the admin queue must be unquiesced.

Here's what we have today:
At the time of the reset or immediately during the reset, an ioctl is done - it can (and has been seen) to be queued on the request queue as there are not state checks. When this io gets scheduled, as of the resetting or even the connecting state - it's currently failing the NVME_REQ_USERCMD check, which fails back into nvmf_fail_nonready_command, which fails back with BLK_STS_RESOURCE, which reschedules it.  This rescheduling reoccurs, until the its tried again with the controller state changed (usually LIVE) which then passes the admin-queue live check. It's good to go.

With your patch:
The ioctl will at least be requeued with BLK_STS_RESOURCE until after the Fabric Connect is completed (Connect is explicitly allowed by the state-specific check) as hopefully queue->live is false. It may actually be issued if we're in this resetting state only to be terminated afterward. All of the transports set the queue live flag as of Fabric Connect complete. If the ioctl is retried as of that point - it may be immediately sent on the admin queue - before or in-between any of the commands issued by nvme_enable_ctrl() - including before controller enabled, nvme_init_identify(), and before nvme_start_ctrl() has been called.    Hopefully none of these commands try to set/change configuration.   They simply shouldn't be allowed until after the controller has transitioned to LIVE. There's no need to allow them in-between.

What you are describing is correct, that is why we have connect_q to save us from it for I/O queues.

The problem I have with this code is that checking for NVME_REQ_USERCMD is a big hammer that you land on what can be normal I/O (e.g. nvme read/write/write-zeros), and we must not rely on this indication to prevent what you are describing.

It maybe (just maybe) OK to check NVME_REQ_USERCMD only for the admin queue, but we are really circling over the fact that we cannot reliably send admin connect before to go off first, because we have to unquiesce it before we issue the admin connect, and there might be a stray command going into execution before we submit the admin connect.

Quoted 7 lines
Quoted 5 lines

Also, do note that pci does allow to queue any type of command based on the queue state only. fabrics should be slightly different because we have the CONNECTING state where we want to let the connect command only to be issued.

That doesn't make me feel any better.  PCI bit-bangs register writes, the odds they can fail/be discarded on the "fabric" is near zero. Admin Commands - both sending/processing/responding is much much faster than on a fabric, where the cmd can be dropped/delayed by fabric routing. Odds of having core-generated commands fail while reconnecting/reinitializing are much much lower.   All says - fabrics have a much higher command and inter-command latency than pcie does - so pcie is going to have a harder time seeing these cases.

Timing here is not the issue, we should take a step back and understand that we fail any user command for a very specific reason. We should solve the root cause and not paper around it.

Quoted 10 lines
Quoted 8 lines
Quoted 4 lines

As for the blk_rq_is_passthrough check - I guess I can see it being based on the queue state, and the check looks ok  (we should never see !blk_rq_is_passthrough on the admin q). But...
- I don't know why it was that important to change it. On the
connecting path, all you're doing is letting io start flowing before all the queues have been created.  Did you really need to start that much sooner ?

The issue is that controller in RESETTING state will have requests that are being issued, and if we don't let it pass through, it will hang around forever being requeued preventing queue freeze to complete.

is it really "forever" or is it until controller loss timeout ? granted 10minutes does seem like forever

ctrl_loss_tmo can be a 1000000 years, it cannot rescue us.

Quoted 8 lines
Quoted 6 lines
Quoted 2 lines

- But on the resetting path, or deleting cases, you've added a
condition now where the controller state was changed, but there was a delay before the transport marked the queue live. It's common practice in the transports to change state then schedule a work element to perform the actual state change.  Why would you want io to continue to flow during that window ?   This may bring out other problems we've avoided in the past.

What are you referring to? the change here? the controller reset must allow requests that came in before we started queue freeze to pass, otherwise freeze will never complete.

I'm referring to the io's that now pass through with the patch as the controller state is no longer blocking them (e.g. the removal of the !blk_rq_is_passthrough and NVME_REQ_USERCMD checks).

First of all, you have a bug hiding there if blk_mq_update_nr_hw_queues actually ever does anything, because it won't be able to reliably freeze the queue (because the !blk_rq_is_passthrough keeps the commands endlessly rescheduling itself, not ever completing hence the queue cannot be frozen). This is what started this, because today we support multiple queue maps, it always ends up freezing the queue, so it can happen much more frequently.

They will pass the

Quoted 1 line

ready checks and go out on the wire, creating more active io that has to be deleted when the reset finally quiesces and terminates the active io list. For FC, we have to real link traffic for any io that gets started, so it raises the demands.   Should be fine, but it is creating more load at a point where it could have been blocked.

As I said, freezing the queue is impossible when we let these commands hang in limbo state, and queue freeze may occur in a lot of places, so we must either let them pass, or fence them when starting the freeze.

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

#7 lore raw
Quoted 10 lines · Sagi Grimberg
Quoted 4 lines
Quoted 3 lines

Can you please give an example? NVME_REQ_USERCMD should not be any different from any other type of I/O.

...

The problem I have with this code is that checking for NVME_REQ_USERCMD is a big hammer that you land on what can be normal I/O (e.g. nvme read/write/write-zeros), and we must not rely on this indication to prevent what you are describing.

Quoted 5 lines

It maybe (just maybe) OK to check NVME_REQ_USERCMD only for the admin queue, but we are really circling over the fact that we cannot reliably send admin connect before to go off first, because we have to unquiesce it before we issue the admin connect, and there might be a stray command going into execution before we submit the admin connect.

agree.

Didn't you suggest something separate to isolate the connect commands/init commands ?   Can we use a different tag set for "initialization" commands?  We can use different queue_rq routines with different check ready checks.  I don't like the overhead, but sure makes the logic more straightforward.

Quoted 9 lines
Quoted 9 lines
Quoted 8 lines
Quoted 4 lines

As for the blk_rq_is_passthrough check - I guess I can see it being based on the queue state, and the check looks ok  (we should never see !blk_rq_is_passthrough on the admin q). But...
- I don't know why it was that important to change it. On the
connecting path, all you're doing is letting io start flowing before all the queues have been created.  Did you really need to start that much sooner ?

The issue is that controller in RESETTING state will have requests that are being issued, and if we don't let it pass through, it will hang around forever being requeued preventing queue freeze to complete.

Isn't this the real problem:  we require the request to complete in order to get off the request queue in order to freeze. But to complete - it has to finish with some status. This is ok if running multipath, as mpath queues and retries - but what mpath is not running ? The issuer of the io will get the error ?   am I understanding it ?   If so I don't think a simple check reorg such as proposed will work as ultimately we can't return an error to something that doesn't know how to requeue/retry.

Should we always be putting even non-mp devices under mp, thus mp catches the errors rather than the blk queue, and can have a short delay while the controller reconnects ?  At least normal io can be handled this way and we'd have to figure out the ioctl path.   But it should still allow the core io, such as scanning, to get the error and fail out, to be picked up after the reconnect.

-- james

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

author maintainer
#8 lore raw
Quoted 15 lines
Quoted 9 lines
Quoted 3 lines
Quoted 2 lines

Can you please give an example? NVME_REQ_USERCMD should not be any different from any other type of I/O.

...

The problem I have with this code is that checking for NVME_REQ_USERCMD is a big hammer that you land on what can be normal I/O (e.g. nvme read/write/write-zeros), and we must not rely on this indication to prevent what you are describing.

Quoted 5 lines

It maybe (just maybe) OK to check NVME_REQ_USERCMD only for the admin queue, but we are really circling over the fact that we cannot reliably send admin connect before to go off first, because we have to unquiesce it before we issue the admin connect, and there might be a stray command going into execution before we submit the admin connect.

Quoted 3 lines

agree.

Didn't you suggest something separate to isolate the connect commands/init commands ?   Can we use a different tag set for "initialization" commands?  We can use different queue_rq routines with different check ready checks.  I don't like the overhead, but sure makes the logic more straightforward.

I think I've added this now that I'm looking back:
e7832cb48a65 ("nvme: make fabrics command run on a separate request queue")

But it still doesn't solve the user commands you are referring to.

Quoted 11 lines
Quoted 9 lines
Quoted 9 lines
Quoted 8 lines
Quoted 4 lines

As for the blk_rq_is_passthrough check - I guess I can see it being based on the queue state, and the check looks ok  (we should never see !blk_rq_is_passthrough on the admin q). But...
- I don't know why it was that important to change it. On the
connecting path, all you're doing is letting io start flowing before all the queues have been created.  Did you really need to start that much sooner ?

The issue is that controller in RESETTING state will have requests that are being issued, and if we don't let it pass through, it will hang around forever being requeued preventing queue freeze to complete.

Isn't this the real problem:  we require the request to complete in order to get off the request queue in order to freeze. But to complete - it has to finish with some status. This is ok if running multipath, as mpath queues and retries - but what mpath is not running ? The issuer of the io will get the error ?

We already fail noretry errors when we cancel (or abort) anything that was submitted before, so there is nothing new here. non-mpath does not set FAILFAST so it gets requeued, user commands do, so they are failed, this is not different than before.

Quoted 1 line

Should we always be putting even non-mp devices under mp, thus mp catches the errors rather than the blk queue, and can have a short delay while the controller reconnects ?  At least normal io can be handled this way and we'd have to figure out the ioctl path.   But it should still allow the core io, such as scanning, to get the error and fail out, to be picked up after the reconnect.

non-mpath still gets requeued because it doesn't set any sort of FAILFAST flag.
or

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

#11 lore raw
In reply to Sagi Grimberg

Dropped from nvme-5.9 to wait to the discussion to finish as I want to send a PR to Jens today..

Quoted 56 lines · Sagi Grimberg

Right now we are failing requests based on the controller state (which is checked inline in nvmf_check_ready) however we should definitely accept requests if the queue is live.

When entering controller reset, we transition the controller into NVME_CTRL_RESETTING, and then return BLK_STS_RESOURCE for non-mpath requests (have blk_noretry_request set).

This is also the case for NVME_REQ_USER for some reason, but there shouldn't be any reason for us to reject this I/O in a controller reset.

In a non-mpath setup, this means that the requests will simply be requeued over and over forever not allowing the q_usage_counter to drop its final reference, causing controller reset to hang if running concurrently with heavy I/O.

While we are at it, remove the redundant NVME_CTRL_NEW case, which should never see any I/O as it must first transition to NVME_CTRL_CONNECTING.

Fixes 35897b920c8a ("nvme-fabrics: fix and refine state checks in __nvmf_check_ready")
---
 drivers/nvme/host/fabrics.c | 10 +---------
 1 file changed, 1 insertion(+), 9 deletions(-)
drivers/nvme/host/fabrics.c +1 -9
index 4ec4829d6233..2e7838f42e36 100644
@@ -564,21 +564,13 @@ bool __nvmf_check_ready(struct nvme_ctrl *ctrl, struct request *rq,
 {
 	struct nvme_request *req = nvme_req(rq);
 
-	/*
-	 * If we are in some state of setup or teardown only allow
-	 * internally generated commands.
-	 */
-	if (!blk_rq_is_passthrough(rq) || (req->flags & NVME_REQ_USERCMD))
-		return false;
-
 	/*
 	 * Only allow commands on a live queue, except for the connect command,
 	 * which is require to set the queue live in the appropinquate states.
 	 */
 	switch (ctrl->state) {
-	case NVME_CTRL_NEW:
 	case NVME_CTRL_CONNECTING:
-		if (nvme_is_fabrics(req->cmd) &&
+		if (blk_rq_is_passthrough(rq) && nvme_is_fabrics(req->cmd) &&
 		    req->cmd->fabrics.fctype == nvme_fabrics_type_connect)
 			return true;
 		break;
Signature · 1 line
2.25.1

---end quoted text---

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

author maintainer
#12 lore raw
Quoted 3 lines

Dropped from nvme-5.9 to wait to the discussion to finish as I want to send a PR to Jens today..

OK, this is a bug fix, so we can send it later for sure.

I can also send a patch that doesn't touch the USERCMDs that James is concerned about and we will fix that later...

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

#13 lore raw
Quoted 8 lines · Sagi Grimberg
Quoted 3 lines

Dropped from nvme-5.9 to wait to the discussion to finish as I want to send a PR to Jens today..

OK, this is a bug fix, so we can send it later for sure.

I can also send a patch that doesn't touch the USERCMDs that James is concerned about and we will fix that later...

I need to catch up on the actual discussion. I'm heading off to a vacation today and just need to get everything in order first.

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