[PATCH] nvme: set non-mdts limits at I/O queue creation

[PATCH] nvme: set non-mdts limits at I/O queue creation

[Chaitanya Kulkarni]

In current implementation we set the non-mdts limits by calling
nvme_init_non_mdts_limits() from nvme_init_ctrl_finish().
This also tries to set the limits for the discovery controller which
has no I/O queues resulting in the warning message reported by the
nvme_log_error() when running blktest nvme/002: -

[ 2005.155946] run blktests nvme/002 at 2022-04-09 16:57:47
[ 2005.192223] loop: module loaded
[ 2005.196429] nvmet: adding nsid 1 to subsystem blktests-subsystem-0
[ 2005.200334] nvmet: adding nsid 1 to subsystem blktests-subsystem-1

<------------------------------SNIP---------------------------------->

[ 2008.958108] nvmet: adding nsid 1 to subsystem blktests-subsystem-997
[ 2008.962082] nvmet: adding nsid 1 to subsystem blktests-subsystem-998
[ 2008.966102] nvmet: adding nsid 1 to subsystem blktests-subsystem-999
[ 2008.973132] nvmet: creating discovery controller 1 for subsystem nqn.2014-08.org.nvmexpress.discovery for NQN testhostnqn.
*[ 2008.973196] nvme1: Identify(0x6), Invalid Field in Command (sct 0x0 / sc 0x2) MORE DNR*
[ 2008.974595] nvme nvme1: new ctrl: "nqn.2014-08.org.nvmexpress.discovery"
[ 2009.103248] nvme nvme1: Removing ctrl: NQN "nqn.2014-08.org.nvmexpress.discovery"

Export the nvme_init_non_mdts_limits() and move the same call from
nvme_ctrl_finish() to each transport right before each transport calls
the nvme_scan_work() from nvme_start_ctrl() path, then proceeds to the
ns allocation where these limits max_discard_{segments|sectors} and
max_write_zeroes_sectors are actually used in folllwing path :-

A. nvme_scan_work()
...
nvme_validate_or_alloc_ns()
  nvme_alloc_ns()
   nvme_update_ns_info()
    nvme_update_disk_info()
     nvme_config_discard()
     blk_queue_max_write_zeroes_sectors()

1. FC:-
nvme_fc_create_association()
   call nvme_init_mdts_limits() <---
  nvme_start_ctrl()
   nvme_scan_queue()
    nvme_scan_work()
     path to ns alloc & accessing ctrl non mdts limits see above A.

2. PCIe:-
nvme_reset_work()
  nvme_dev_add()
   call nvme_init_mdts_limits() <---
  nvme_start_ctrl()
   nvme_scan_queue()
    nvme_scan_work()
     path to ns alloc & accessing ctrl non mdts limits see above A.

3. RDMA :-
nvme_rdma_setup_ctrl
  nvme_rdma_configure_io_queues
   call nvme_init_mdts_limits() <---
  nvme_start_ctrl()
   nvme_scan_queue()
    nvme_scan_work()
     path to ns alloc & accessing ctrl non mdts limits see above A.

4. TCP :-
nvme_tcp_setup_ctrl
  nvme_tcp_configure_io_queues
   call nvme_init_mdts_limits() <---
  nvme_start_ctrl()
   nvme_scan_queue()
    nvme_scan_work()
     path to ns alloc & accessing ctrl non mdts limits see above A.

Signed-off-by: Chaitanya Kulkarni <kch@nvidia.com>
---
 drivers/nvme/host/core.c   | 7 ++-----
 drivers/nvme/host/fc.c     | 4 ++++
 drivers/nvme/host/nvme.h   | 1 +
 drivers/nvme/host/pci.c    | 7 +++++++
 drivers/nvme/host/rdma.c   | 4 ++++
 drivers/nvme/host/tcp.c    | 4 ++++
 drivers/nvme/target/loop.c | 4 ++++
 7 files changed, 26 insertions(+), 5 deletions(-)

diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c
index 42f9772abc4d..8bc0d10931e7 100644
--- a/drivers/nvme/host/core.c
+++ b/drivers/nvme/host/core.c
@@ -2854,7 +2854,7 @@ static inline u32 nvme_mps_to_sectors(struct nvme_ctrl *ctrl, u32 units)
 	return val;
 }
 
-static int nvme_init_non_mdts_limits(struct nvme_ctrl *ctrl)
+int nvme_init_non_mdts_limits(struct nvme_ctrl *ctrl)
 {
 	struct nvme_command c = { };
 	struct nvme_id_ctrl_nvm *id;
@@ -2905,6 +2905,7 @@ static int nvme_init_non_mdts_limits(struct nvme_ctrl *ctrl)
 	kfree(id);
 	return ret;
 }
+EXPORT_SYMBOL_GPL(nvme_init_non_mdts_limits);
 
 static int nvme_init_identify(struct nvme_ctrl *ctrl)
 {
@@ -3082,10 +3083,6 @@ int nvme_init_ctrl_finish(struct nvme_ctrl *ctrl)
 	if (ret)
 		return ret;
 
-	ret = nvme_init_non_mdts_limits(ctrl);
-	if (ret < 0)
-		return ret;
-
 	ret = nvme_configure_apst(ctrl);
 	if (ret < 0)
 		return ret;
diff --git a/drivers/nvme/host/fc.c b/drivers/nvme/host/fc.c
index 080f85f4105f..a04a0627b0a0 100644
--- a/drivers/nvme/host/fc.c
+++ b/drivers/nvme/host/fc.c
@@ -3168,6 +3168,10 @@ nvme_fc_create_association(struct nvme_fc_ctrl *ctrl)
 	 */
 
 	if (ctrl->ctrl.queue_count > 1) {
+		ret = nvme_init_non_mdts_limits(&ctrl->ctrl);
+		if (ret < 0)
+			goto out_term_aen_ops;
+
 		if (!ctrl->ioq_live)
 			ret = nvme_fc_create_io_queues(ctrl);
 		else
diff --git a/drivers/nvme/host/nvme.h b/drivers/nvme/host/nvme.h
index 81c4f5379c0c..2f3c6fcaa440 100644
--- a/drivers/nvme/host/nvme.h
+++ b/drivers/nvme/host/nvme.h
@@ -784,6 +784,7 @@ long nvme_ns_head_chr_ioctl(struct file *file, unsigned int cmd,
 long nvme_dev_ioctl(struct file *file, unsigned int cmd,
 		unsigned long arg);
 int nvme_getgeo(struct block_device *bdev, struct hd_geometry *geo);
+int nvme_init_non_mdts_limits(struct nvme_ctrl *ctrl);
 
 extern const struct attribute_group *nvme_ns_id_attr_groups[];
 extern const struct pr_ops nvme_pr_ops;
diff --git a/drivers/nvme/host/pci.c b/drivers/nvme/host/pci.c
index 5a98a7de0964..2ceca29ad997 100644
--- a/drivers/nvme/host/pci.c
+++ b/drivers/nvme/host/pci.c
@@ -2526,6 +2526,13 @@ static void nvme_dev_add(struct nvme_dev *dev)
 {
 	int ret;
 
+	ret = nvme_init_non_mdts_limits(&dev->ctrl);
+	if (ret < 0) {
+		dev_warn(dev->ctrl.device,
+				"reading non mdts limit error: %d\n", ret);
+		return;
+	}
+
 	if (!dev->ctrl.tagset) {
 		dev->tagset.ops = &nvme_mq_ops;
 		dev->tagset.nr_hw_queues = dev->online_queues - 1;
diff --git a/drivers/nvme/host/rdma.c b/drivers/nvme/host/rdma.c
index b87c8ae41d9b..7acce264375d 100644
--- a/drivers/nvme/host/rdma.c
+++ b/drivers/nvme/host/rdma.c
@@ -971,6 +971,10 @@ static int nvme_rdma_configure_io_queues(struct nvme_rdma_ctrl *ctrl, bool new)
 	if (ret)
 		return ret;
 
+	ret = nvme_init_non_mdts_limits(&ctrl->ctrl);
+	if (ret < 0)
+		goto out_free_io_queues;
+
 	if (new) {
 		ctrl->ctrl.tagset = nvme_rdma_alloc_tagset(&ctrl->ctrl, false);
 		if (IS_ERR(ctrl->ctrl.tagset)) {
diff --git a/drivers/nvme/host/tcp.c b/drivers/nvme/host/tcp.c
index bb67538d241b..2a9c32afb6e7 100644
--- a/drivers/nvme/host/tcp.c
+++ b/drivers/nvme/host/tcp.c
@@ -1899,6 +1899,10 @@ static int nvme_tcp_configure_io_queues(struct nvme_ctrl *ctrl, bool new)
 	if (ret)
 		return ret;
 
+	ret = nvme_init_non_mdts_limits(ctrl);
+	if (ret < 0)
+		goto out_free_io_queues;
+
 	if (new) {
 		ctrl->tagset = nvme_tcp_alloc_tagset(ctrl, false);
 		if (IS_ERR(ctrl->tagset)) {
diff --git a/drivers/nvme/target/loop.c b/drivers/nvme/target/loop.c
index 59024af2da2e..2ac9017e966a 100644
--- a/drivers/nvme/target/loop.c
+++ b/drivers/nvme/target/loop.c
@@ -522,6 +522,10 @@ static int nvme_loop_create_io_queues(struct nvme_loop_ctrl *ctrl)
 {
 	int ret;
 
+	ret = nvme_init_non_mdts_limits(&ctrl->ctrl);
+	if (ret < 0)
+		return ret;
+
 	ret = nvme_loop_init_io_queues(ctrl);
 	if (ret)
 		return ret;
-- 
2.29.0

Re: [PATCH] nvme: set non-mdts limits at I/O queue creation

[Sagi Grimberg]

On 5/17/22 08:53, Chaitanya Kulkarni wrote:
> In current implementation we set the non-mdts limits by calling
> nvme_init_non_mdts_limits() from nvme_init_ctrl_finish().
> This also tries to set the limits for the discovery controller which
> has no I/O queues resulting in the warning message reported by the
> nvme_log_error() when running blktest nvme/002: -
> 
> [ 2005.155946] run blktests nvme/002 at 2022-04-09 16:57:47
> [ 2005.192223] loop: module loaded
> [ 2005.196429] nvmet: adding nsid 1 to subsystem blktests-subsystem-0
> [ 2005.200334] nvmet: adding nsid 1 to subsystem blktests-subsystem-1
> 
> <------------------------------SNIP---------------------------------->
> 
> [ 2008.958108] nvmet: adding nsid 1 to subsystem blktests-subsystem-997
> [ 2008.962082] nvmet: adding nsid 1 to subsystem blktests-subsystem-998
> [ 2008.966102] nvmet: adding nsid 1 to subsystem blktests-subsystem-999
> [ 2008.973132] nvmet: creating discovery controller 1 for subsystem nqn.2014-08.org.nvmexpress.discovery for NQN testhostnqn.
> *[ 2008.973196] nvme1: Identify(0x6), Invalid Field in Command (sct 0x0 / sc 0x2) MORE DNR*
> [ 2008.974595] nvme nvme1: new ctrl: "nqn.2014-08.org.nvmexpress.discovery"
> [ 2009.103248] nvme nvme1: Removing ctrl: NQN "nqn.2014-08.org.nvmexpress.discovery"
> 
> Export the nvme_init_non_mdts_limits() and move the same call from
> nvme_ctrl_finish() to each transport right before each transport calls
> the nvme_scan_work() from nvme_start_ctrl() path, then proceeds to the
> ns allocation where these limits max_discard_{segments|sectors} and
> max_write_zeroes_sectors are actually used in folllwing path :-

Why not just check the ctrl->queue_count and call it accordingly?

Re: [PATCH] nvme: set non-mdts limits at I/O queue creation

[Chaitanya Kulkarni]

On 5/17/22 02:56, Sagi Grimberg wrote:
> 
> 
> On 5/17/22 08:53, Chaitanya Kulkarni wrote:
>> In current implementation we set the non-mdts limits by calling
>> nvme_init_non_mdts_limits() from nvme_init_ctrl_finish().
>> This also tries to set the limits for the discovery controller which
>> has no I/O queues resulting in the warning message reported by the
>> nvme_log_error() when running blktest nvme/002: -
>>
>> [ 2005.155946] run blktests nvme/002 at 2022-04-09 16:57:47
>> [ 2005.192223] loop: module loaded
>> [ 2005.196429] nvmet: adding nsid 1 to subsystem blktests-subsystem-0
>> [ 2005.200334] nvmet: adding nsid 1 to subsystem blktests-subsystem-1
>>
>> <------------------------------SNIP---------------------------------->
>>
>> [ 2008.958108] nvmet: adding nsid 1 to subsystem blktests-subsystem-997
>> [ 2008.962082] nvmet: adding nsid 1 to subsystem blktests-subsystem-998
>> [ 2008.966102] nvmet: adding nsid 1 to subsystem blktests-subsystem-999
>> [ 2008.973132] nvmet: creating discovery controller 1 for subsystem 
>> nqn.2014-08.org.nvmexpress.discovery for NQN testhostnqn.
>> *[ 2008.973196] nvme1: Identify(0x6), Invalid Field in Command (sct 
>> 0x0 / sc 0x2) MORE DNR*
>> [ 2008.974595] nvme nvme1: new ctrl: 
>> "nqn.2014-08.org.nvmexpress.discovery"
>> [ 2009.103248] nvme nvme1: Removing ctrl: NQN 
>> "nqn.2014-08.org.nvmexpress.discovery"
>>
>> Export the nvme_init_non_mdts_limits() and move the same call from
>> nvme_ctrl_finish() to each transport right before each transport calls
>> the nvme_scan_work() from nvme_start_ctrl() path, then proceeds to the
>> ns allocation where these limits max_discard_{segments|sectors} and
>> max_write_zeroes_sectors are actually used in folllwing path :-
> 
> Why not just check the ctrl->queue_count and call it accordingly?

nvme_init_ctrl_finish() is called before the I/O queues are allocated
so queue_count value will be always equal to 1 when checked in the
nvme_init_ctrl_finish() before calling the
nvme_init_non_mdts_limits() for PCIe :-

nvme_reset_work()
...
  nvme_init_ctrl_finish()
   nvme_init_non_mdts_limits() <--- ctrl->queue_count = 1
  nvme_dev_add()
  nvme_setup_io_queues()
   nvme_create_io_queues()
    nvme_alloc_queue()
     dev->ctrl.queue_count++ <---
...

Also, this patch groups reading ctrl non-mdts limits right before I/O
queue allocation for each transport, since these limits are only
used for I/O queue allocation.

-ck

Re: [PATCH] nvme: set non-mdts limits at I/O queue creation

[Keith Busch]

On Tue, May 17, 2022 at 10:41:54PM +0000, Chaitanya Kulkarni wrote:
> Also, this patch groups reading ctrl non-mdts limits right before I/O
> queue allocation for each transport, since these limits are only
> used for I/O queue allocation.

Maybe just check that 'ctrl->cntrltype == NVME_CTRL_IO' instead?

Re: [PATCH] nvme: set non-mdts limits at I/O queue creation

[Chaitanya Kulkarni]

On 5/17/22 15:44, Keith Busch wrote:
> On Tue, May 17, 2022 at 10:41:54PM +0000, Chaitanya Kulkarni wrote:
>> Also, this patch groups reading ctrl non-mdts limits right before I/O
>> queue allocation for each transport, since these limits are only
>> used for I/O queue allocation.
> 
> Maybe just check that 'ctrl->cntrltype == NVME_CTRL_IO' instead?

I did that initially, but if I remember correctly versions prior to
1.4 don't necessarily report a valid ctrl type.

-ck

Re: [PATCH] nvme: set non-mdts limits at I/O queue creation

[Keith Busch]

On Tue, May 17, 2022 at 10:49:21PM +0000, Chaitanya Kulkarni wrote:
> On 5/17/22 15:44, Keith Busch wrote:
> > On Tue, May 17, 2022 at 10:41:54PM +0000, Chaitanya Kulkarni wrote:
> >> Also, this patch groups reading ctrl non-mdts limits right before I/O
> >> queue allocation for each transport, since these limits are only
> >> used for I/O queue allocation.
> > 
> > Maybe just check that 'ctrl->cntrltype == NVME_CTRL_IO' instead?
> 
> I did that initially, but if I remember correctly versions prior to
> 1.4 don't necessarily report a valid ctrl type.

I guess we could assume IO controller if < 1.4 since those are the only type
that could have existed.

Otherwise, I really don't like having this all be transport specific. It sounds
like updating the non_mdts limits could safely be done in scan_work since it
requires an IO tagset.

Re: [PATCH] nvme: set non-mdts limits at I/O queue creation

[Chaitanya Kulkarni]

On 5/17/22 15:58, Keith Busch wrote:
> On Tue, May 17, 2022 at 10:49:21PM +0000, Chaitanya Kulkarni wrote:
>> On 5/17/22 15:44, Keith Busch wrote:
>>> On Tue, May 17, 2022 at 10:41:54PM +0000, Chaitanya Kulkarni wrote:
>>>> Also, this patch groups reading ctrl non-mdts limits right before I/O
>>>> queue allocation for each transport, since these limits are only
>>>> used for I/O queue allocation.
>>>
>>> Maybe just check that 'ctrl->cntrltype == NVME_CTRL_IO' instead?
>>
>> I did that initially, but if I remember correctly versions prior to
>> 1.4 don't necessarily report a valid ctrl type.
> 
> I guess we could assume IO controller if < 1.4 since those are the only type
> that could have existed.
> 
> Otherwise, I really don't like having this all be transport specific. It sounds
> like updating the non_mdts limits could safely be done in scan_work since it
> requires an IO tagset.

Even better to do it in nvme_scan_work, let me read the code again and
send out V2 as nvme_scan_work is a common point for all the transports.

-ck