Skip to contentSkip to conversation

nvme: retry commands if DNR bit is not set

PATCH In review Hannes Reinecke 2 participants 3 messages Nov 2, 2022

Subsystems (2)
lore
Share & tools
lore

Hannes Reinecke

author maintainer
#0 lore raw

PATCH 1/1 nvme: retry commands if DNR bit is not set

Add a 'retries' argument to __nvme_submit_sync_cmd() to instruct the function to retry the command if the DNR bit is not set in the command result, and modify the authentication code to allow for retries.

---
 drivers/nvme/host/auth.c    |  2 +-
 drivers/nvme/host/core.c    | 29 ++++++++++++++++++++++++-----
 drivers/nvme/host/fabrics.c | 10 +++++-----
 drivers/nvme/host/nvme.h    |  6 +++++-
 4 files changed, 35 insertions(+), 12 deletions(-)
drivers/nvme/host/auth.c +1 -1
index c8a6db7c4498..3b63aa155beb 100644
@@ -69,7 +69,7 @@ static int nvme_auth_submit(struct nvme_ctrl *ctrl, int qid,
 
 	ret = __nvme_submit_sync_cmd(q, &cmd, NULL, data, data_len,
 				     qid == 0 ? NVME_QID_ANY : qid,
-				     0, flags);
+				     0, flags, nvme_max_retries);
 	if (ret > 0)
 		dev_warn(ctrl->device,
 			"qid %d auth_send failed with status %d\n", qid, ret);
drivers/nvme/host/core.c +24 -5
index dc4220600585..d0de1a85596a 100644
@@ -54,7 +54,7 @@ static unsigned char shutdown_timeout = 5;
 module_param(shutdown_timeout, byte, 0644);
 MODULE_PARM_DESC(shutdown_timeout, "timeout in seconds for controller shutdown");
 
-static u8 nvme_max_retries = 5;
+u8 nvme_max_retries = 5;
 module_param_named(max_retries, nvme_max_retries, byte, 0644);
 MODULE_PARM_DESC(max_retries, "max number of retries a command may have");
 
@@ -1016,7 +1016,7 @@ static int nvme_execute_rq(struct request *rq, bool at_head)
  */
 int __nvme_submit_sync_cmd(struct request_queue *q, struct nvme_command *cmd,
 		union nvme_result *result, void *buffer, unsigned bufflen,
-		int qid, int at_head, blk_mq_req_flags_t flags)
+		int qid, int at_head, blk_mq_req_flags_t flags, int retries)
 {
 	struct request *req;
 	int ret;
@@ -1030,6 +1030,7 @@ int __nvme_submit_sync_cmd(struct request_queue *q, struct nvme_command *cmd,
 	if (IS_ERR(req))
 		return PTR_ERR(req);
 	nvme_init_request(req, cmd);
+	nvme_req(req)->retries = retries;
 
 	if (buffer && bufflen) {
 		ret = blk_rq_map_kern(q, req, buffer, bufflen, GFP_KERNEL);
@@ -1037,8 +1038,18 @@ int __nvme_submit_sync_cmd(struct request_queue *q, struct nvme_command *cmd,
 			goto out;
 	}
 
+retry:
 	req->rq_flags |= RQF_QUIET;
 	ret = nvme_execute_rq(req, at_head);
+	if (ret > 0) {
+		struct nvme_ctrl *ctrl = nvme_req(req)->ctrl;
+
+		if (ctrl->kas)
+			ctrl->comp_seen = true;
+
+		if (retries-- && nvme_decide_disposition(req) != COMPLETE)
+			goto retry;
+	}
 	if (result && ret >= 0)
 		*result = nvme_req(req)->result;
  out:
@@ -1051,10 +1062,18 @@ int nvme_submit_sync_cmd(struct request_queue *q, struct nvme_command *cmd,
 		void *buffer, unsigned bufflen)
 {
 	return __nvme_submit_sync_cmd(q, cmd, NULL, buffer, bufflen,
-			NVME_QID_ANY, 0, 0);
+			NVME_QID_ANY, 0, 0, 0);
 }
 EXPORT_SYMBOL_GPL(nvme_submit_sync_cmd);
 
+int nvme_submit_sync_cmd_retry(struct request_queue *q,
+		struct nvme_command *cmd, void *buffer, unsigned bufflen)
+{
+	return __nvme_submit_sync_cmd(q, cmd, NULL, buffer, bufflen,
+			NVME_QID_ANY, 0, 0, nvme_max_retries);
+}
+EXPORT_SYMBOL_GPL(nvme_submit_sync_cmd_retry);
+
 static u32 nvme_known_admin_effects(u8 opcode)
 {
 	switch (opcode) {
@@ -1500,7 +1519,7 @@ static int nvme_features(struct nvme_ctrl *dev, u8 op, unsigned int fid,
 	c.features.dword11 = cpu_to_le32(dword11);
 
 	ret = __nvme_submit_sync_cmd(dev->admin_q, &c, &res,
-			buffer, buflen, NVME_QID_ANY, 0, 0);
+			buffer, buflen, NVME_QID_ANY, 0, 0, 0);
 	if (ret >= 0 && result)
 		*result = le32_to_cpu(res.u32);
 	return ret;
@@ -2195,7 +2214,7 @@ int nvme_sec_submit(void *data, u16 spsp, u8 secp, void *buffer, size_t len,
 	cmd.common.cdw11 = cpu_to_le32(len);
 
 	return __nvme_submit_sync_cmd(ctrl->admin_q, &cmd, NULL, buffer, len,
-			NVME_QID_ANY, 1, 0);
+			NVME_QID_ANY, 1, 0, 0);
 }
 EXPORT_SYMBOL_GPL(nvme_sec_submit);
 #endif /* CONFIG_BLK_SED_OPAL */
drivers/nvme/host/fabrics.c +5 -5
index ce27276f552d..f4bfbaf733e9 100644
@@ -153,7 +153,7 @@ int nvmf_reg_read32(struct nvme_ctrl *ctrl, u32 off, u32 *val)
 	cmd.prop_get.offset = cpu_to_le32(off);
 
 	ret = __nvme_submit_sync_cmd(ctrl->fabrics_q, &cmd, &res, NULL, 0,
-			NVME_QID_ANY, 0, 0);
+			NVME_QID_ANY, 0, 0, 0);
 
 	if (ret >= 0)
 		*val = le64_to_cpu(res.u64);
@@ -199,7 +199,7 @@ int nvmf_reg_read64(struct nvme_ctrl *ctrl, u32 off, u64 *val)
 	cmd.prop_get.offset = cpu_to_le32(off);
 
 	ret = __nvme_submit_sync_cmd(ctrl->fabrics_q, &cmd, &res, NULL, 0,
-			NVME_QID_ANY, 0, 0);
+			NVME_QID_ANY, 0, 0, 0);
 
 	if (ret >= 0)
 		*val = le64_to_cpu(res.u64);
@@ -244,7 +244,7 @@ int nvmf_reg_write32(struct nvme_ctrl *ctrl, u32 off, u32 val)
 	cmd.prop_set.value = cpu_to_le64(val);
 
 	ret = __nvme_submit_sync_cmd(ctrl->fabrics_q, &cmd, NULL, NULL, 0,
-			NVME_QID_ANY, 0, 0);
+			NVME_QID_ANY, 0, 0, 0);
 	if (unlikely(ret))
 		dev_err(ctrl->device,
 			"Property Set error: %d, offset %#x\n",
@@ -401,7 +401,7 @@ int nvmf_connect_admin_queue(struct nvme_ctrl *ctrl)
 
 	ret = __nvme_submit_sync_cmd(ctrl->fabrics_q, &cmd, &res,
 			data, sizeof(*data), NVME_QID_ANY, 1,
-			BLK_MQ_REQ_RESERVED | BLK_MQ_REQ_NOWAIT);
+			BLK_MQ_REQ_RESERVED | BLK_MQ_REQ_NOWAIT, 0);
 	if (ret) {
 		nvmf_log_connect_error(ctrl, ret, le32_to_cpu(res.u32),
 				       &cmd, data);
@@ -480,7 +480,7 @@ int nvmf_connect_io_queue(struct nvme_ctrl *ctrl, u16 qid)
 
 	ret = __nvme_submit_sync_cmd(ctrl->connect_q, &cmd, &res,
 			data, sizeof(*data), qid, 1,
-			BLK_MQ_REQ_RESERVED | BLK_MQ_REQ_NOWAIT);
+			BLK_MQ_REQ_RESERVED | BLK_MQ_REQ_NOWAIT, 0);
 	if (ret) {
 		nvmf_log_connect_error(ctrl, ret, le32_to_cpu(res.u32),
 				       &cmd, data);
drivers/nvme/host/nvme.h +5 -1
index a29877217ee6..32d9dc2d957e 100644
@@ -47,6 +47,8 @@ extern struct workqueue_struct *nvme_wq;
 extern struct workqueue_struct *nvme_reset_wq;
 extern struct workqueue_struct *nvme_delete_wq;
 
+extern u8 nvme_max_retries;
+
 /*
  * List of workarounds for devices that required behavior not specified in
  * the standard.
@@ -811,10 +813,12 @@ static inline bool nvme_is_unique_nsid(struct nvme_ctrl *ctrl,
 
 int nvme_submit_sync_cmd(struct request_queue *q, struct nvme_command *cmd,
 		void *buf, unsigned bufflen);
+int nvme_submit_sync_cmd_retry(struct request_queue *q,
+		struct nvme_command *cmd, void *buf, unsigned bufflen);
 int __nvme_submit_sync_cmd(struct request_queue *q, struct nvme_command *cmd,
 		union nvme_result *result, void *buffer, unsigned bufflen,
 		int qid, int at_head,
-		blk_mq_req_flags_t flags);
+		blk_mq_req_flags_t flags, int retries);
 int nvme_set_features(struct nvme_ctrl *dev, unsigned int fid,
 		      unsigned int dword11, void *buffer, size_t buflen,
 		      u32 *result);
Signature · 1 line
2.35.3
#1 lore raw
Quoted 64 lines · Hannes Reinecke

Add a 'retries' argument to __nvme_submit_sync_cmd() to instruct the function to retry the command if the DNR bit is not set in the command result, and modify the authentication code to allow for retries.

---
  drivers/nvme/host/auth.c    |  2 +-
  drivers/nvme/host/core.c    | 29 ++++++++++++++++++++++++-----
  drivers/nvme/host/fabrics.c | 10 +++++-----
  drivers/nvme/host/nvme.h    |  6 +++++-
  4 files changed, 35 insertions(+), 12 deletions(-)
drivers/nvme/host/auth.c +1 -1
index c8a6db7c4498..3b63aa155beb 100644
@@ -69,7 +69,7 @@ static int nvme_auth_submit(struct nvme_ctrl *ctrl, int qid,
 
  	ret = __nvme_submit_sync_cmd(q, &cmd, NULL, data, data_len,
  				     qid == 0 ? NVME_QID_ANY : qid,
-				     0, flags);
+				     0, flags, nvme_max_retries);
  	if (ret > 0)
  		dev_warn(ctrl->device,
  			"qid %d auth_send failed with status %d\n", qid, ret);
drivers/nvme/host/core.c +5 -2
index dc4220600585..d0de1a85596a 100644
@@ -54,7 +54,7 @@ static unsigned char shutdown_timeout = 5;
  module_param(shutdown_timeout, byte, 0644);
  MODULE_PARM_DESC(shutdown_timeout, "timeout in seconds for controller shutdown");
 
-static u8 nvme_max_retries = 5;
+u8 nvme_max_retries = 5;
  module_param_named(max_retries, nvme_max_retries, byte, 0644);
  MODULE_PARM_DESC(max_retries, "max number of retries a command may have");
 
@@ -1016,7 +1016,7 @@ static int nvme_execute_rq(struct request *rq, bool at_head)
   */
  int __nvme_submit_sync_cmd(struct request_queue *q, struct nvme_command *cmd,
  		union nvme_result *result, void *buffer, unsigned bufflen,
-		int qid, int at_head, blk_mq_req_flags_t flags)
+		int qid, int at_head, blk_mq_req_flags_t flags, int retries)
  {
  	struct request *req;
  	int ret;
@@ -1030,6 +1030,7 @@ int __nvme_submit_sync_cmd(struct request_queue *q, struct nvme_command *cmd,
  	if (IS_ERR(req))
  		return PTR_ERR(req);
  	nvme_init_request(req, cmd);
+	nvme_req(req)->retries = retries;
 
  	if (buffer && bufflen) {
  		ret = blk_rq_map_kern(q, req, buffer, bufflen, GFP_KERNEL);
@@ -1037,8 +1038,18 @@ int __nvme_submit_sync_cmd(struct request_queue *q, struct nvme_command *cmd,
  			goto out;
  	}
 
+retry:
  	req->rq_flags |= RQF_QUIET;
  	ret = nvme_execute_rq(req, at_head);
+	if (ret > 0) {

drivers/nvme/host/core.c int __nvme_submit_sync_cmd(struct request_queue *q, struct nvme_command *cmd, · line 1054

Quoted 10 lines
drivers/nvme/host/core.c +8 -0
+		struct nvme_ctrl *ctrl = nvme_req(req)->ctrl;
+
+		if (ctrl->kas)
+			ctrl->comp_seen = true;
+
+		if (retries-- && nvme_decide_disposition(req) != COMPLETE)
+			goto retry;
+	}
  	if (result && ret >= 0)
  		*result = nvme_req(req)->result;

Reusing the existing retry mechanism may be better.
We can do like this:
Do not set REQ_FAILFAST_DRIVER in nvme_init_request If need retry.

	if (noretry)
		req->cmd_flags |= REQ_FAILFAST_DRIVER;
author maintainer
#2 lore raw
Quoted 73 lines · Chao Leng
Quoted 70 lines · Hannes Reinecke

Add a 'retries' argument to __nvme_submit_sync_cmd() to instruct the function to retry the command if the DNR bit is not set in the command result, and modify the authentication code to allow for retries.

---
  drivers/nvme/host/auth.c    |  2 +- drivers/nvme/host/core.c    | 29 ++++++++++++++++++++++++----- drivers/nvme/host/fabrics.c | 10 +++++----- drivers/nvme/host/nvme.h    |  6 +++++-

  4 files changed, 35 insertions(+), 12 deletions(-)
drivers/nvme/host/auth.c +1 -1
index c8a6db7c4498..3b63aa155beb 100644
@@ -69,7 +69,7 @@ static int nvme_auth_submit(struct nvme_ctrl *ctrl, 
 int qid,
       ret = __nvme_submit_sync_cmd(q, &cmd, NULL, data, data_len,
                        qid == 0 ? NVME_QID_ANY : qid,
-                     0, flags);
+                     0, flags, nvme_max_retries);
       if (ret > 0)
           dev_warn(ctrl->device,
               "qid %d auth_send failed with status %d\n", qid, ret);
drivers/nvme/host/core.c +2 -2
index dc4220600585..d0de1a85596a 100644
@@ -54,7 +54,7 @@ static unsigned char shutdown_timeout = 5;
   module_param(shutdown_timeout, byte, 0644);
   MODULE_PARM_DESC(shutdown_timeout, "timeout in seconds for controller shutdown");
-static u8 nvme_max_retries = 5;
+u8 nvme_max_retries = 5;
   module_param_named(max_retries, nvme_max_retries, byte, 0644);
   MODULE_PARM_DESC(max_retries, "max number of retries a command may have");
@@ -1016,7 +1016,7 @@ static int nvme_execute_rq(struct request *rq, 
 bool at_head)
    */
   int __nvme_submit_sync_cmd(struct request_queue *q, struct nvme_command *cmd,
           union nvme_result *result, void *buffer, unsigned bufflen,
-        int qid, int at_head, blk_mq_req_flags_t flags)
+        int qid, int at_head, blk_mq_req_flags_t flags, int retries)
   {
       struct request *req;

      int ret;

drivers/nvme/host/core.c +10 -0
@@ -1030,6 +1030,7 @@ int __nvme_submit_sync_cmd(struct request_queue 
 *q, struct nvme_command *cmd,
       if (IS_ERR(req))
           return PTR_ERR(req);
       nvme_init_request(req, cmd);
+    nvme_req(req)->retries = retries;
       if (buffer && bufflen) {
           ret = blk_rq_map_kern(q, req, buffer, bufflen, GFP_KERNEL);
@@ -1037,8 +1038,18 @@ int __nvme_submit_sync_cmd(struct request_queue 
 *q, struct nvme_command *cmd,
               goto out;
       }
+retry:
       req->rq_flags |= RQF_QUIET;
       ret = nvme_execute_rq(req, at_head);
+    if (ret > 0) {
+        struct nvme_ctrl *ctrl = nvme_req(req)->ctrl;
+
+        if (ctrl->kas)
+            ctrl->comp_seen = true;
+
+        if (retries-- && nvme_decide_disposition(req) != COMPLETE)
+            goto retry;
Quoted 9 lines

drivers/nvme/host/core.c int __nvme_submit_sync_cmd(struct request_queue *q, struct nvme_command *cmd, · line 1054

Quoted 3 lines
drivers/nvme/host/core.c +1 -0
+    }
       if (result && ret >= 0)
           *result = nvme_req(req)->result;

Reusing the existing retry mechanism may be better.
We can do like this:
Do not set REQ_FAILFAST_DRIVER in nvme_init_request If need retry.
    if (noretry)
        req->cmd_flags |= REQ_FAILFAST_DRIVER;

But then we still would need to set the retries; however, it's a good suggestion. Will be reworking the patch.

Cheers,

Hannes

Signature · 4 lines
Dr. Hannes Reinecke		           Kernel Storage Architect
[email protected]			                  +49 911 74053 688
SUSE Software Solutions Germany GmbH, Maxfeldstr. 5, 90409 Nürnberg
HRB 36809 (AG Nürnberg), GF: Felix Imendörffer