[PATCHv2 0/3] nvme: rework __nvme_submit_sync_cmd()

[PATCHv2 0/3] nvme: rework __nvme_submit_sync_cmd()

[Hannes Reinecke]

Hi all,

as Jens and hch complained about the argument list to __nvme_submit_sync_cmd()
getting too long, here's now a patchset to clean things up.
The first patch is to split off __nvme_alloc_rq() from __nvme_submit_sync_cmd(),
which serves to reduce the number of arguments; it's also a nice cleanup as this
function can be utilized to replace some open-coded versions.
The second patch is to introduce a 'retry' argument to __nvme_alloc_rq(), to
indicate whether retries for this command should be allowd.
And the third is another cleanup; the 'at_head' argument to __nvme_submit_sync_cmd()
really should be a boolean, as it can take only two values, and it's being
converted into a boolean within __nvme_submit_sync_cmd() anyway.

As usual, comments and reviews are welcome.

Hannes Reinecke (3):
  nvme: split __nvme_submit_sync_cmd()
  nvme: retry authentication commands if DNR status bit is not set
  nvme: make 'at_head' parameter for __nvme_submit_sync_cmd() boolean

 drivers/nvme/host/auth.c       | 11 +++++--
 drivers/nvme/host/core.c       | 60 ++++++++++++++++++++++------------
 drivers/nvme/host/fabrics.c    | 38 +++++++++++++++------
 drivers/nvme/host/nvme.h       |  8 ++---
 drivers/nvme/host/pci.c        |  8 ++---
 drivers/nvme/target/passthru.c |  3 +-
 6 files changed, 83 insertions(+), 45 deletions(-)

-- 
2.35.3

[PATCH 1/3] nvme: split __nvme_submit_sync_cmd()

[Hannes Reinecke]

Split a __nvme_alloc_rq() function from __nvme_submit_sync_cmd()
to reduce the number of arguments.

Signed-off-by: Hannes Reinecke <hare@suse.de>
---
 drivers/nvme/host/auth.c       |  9 +++--
 drivers/nvme/host/core.c       | 60 ++++++++++++++++++++++------------
 drivers/nvme/host/fabrics.c    | 40 +++++++++++++++++------
 drivers/nvme/host/nvme.h       |  8 ++---
 drivers/nvme/host/pci.c        |  8 ++---
 drivers/nvme/target/passthru.c |  3 +-
 6 files changed, 83 insertions(+), 45 deletions(-)

diff --git a/drivers/nvme/host/auth.c b/drivers/nvme/host/auth.c
index 787537454f7f..dcea6873ca2c 100644
--- a/drivers/nvme/host/auth.c
+++ b/drivers/nvme/host/auth.c
@@ -62,6 +62,7 @@ static int nvme_auth_submit(struct nvme_ctrl *ctrl, int qid,
 	struct nvme_command cmd = {};
 	blk_mq_req_flags_t flags = nvme_auth_flags_from_qid(qid);
 	struct request_queue *q = nvme_auth_queue_from_qid(ctrl, qid);
+	struct request *req;
 	int ret;
 
 	cmd.auth_common.opcode = nvme_fabrics_command;
@@ -76,9 +77,11 @@ static int nvme_auth_submit(struct nvme_ctrl *ctrl, int qid,
 		cmd.auth_receive.al = cpu_to_le32(data_len);
 	}
 
-	ret = __nvme_submit_sync_cmd(q, &cmd, NULL, data, data_len,
-				     qid == 0 ? NVME_QID_ANY : qid,
-				     0, flags);
+	req = __nvme_alloc_rq(q, &cmd, qid == 0 ? NVME_QID_ANY : qid, flags);
+	if (IS_ERR(req))
+		return PTR_ERR(req);
+
+	ret = __nvme_submit_sync_cmd(req, NULL, data, data_len, 0);
 	if (ret > 0)
 		dev_warn(ctrl->device,
 			"qid %d auth_send failed with status %d\n", qid, ret);
diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c
index d1c9402389f9..278f7aa674c6 100644
--- a/drivers/nvme/host/core.c
+++ b/drivers/nvme/host/core.c
@@ -1017,29 +1017,35 @@ int nvme_execute_rq(struct request *rq, bool at_head)
 }
 EXPORT_SYMBOL_NS_GPL(nvme_execute_rq, NVME_TARGET_PASSTHRU);
 
-/*
- * Returns 0 on success.  If the result is negative, it's a Linux error code;
- * if the result is positive, it's an NVM Express status code
- */
-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)
+struct request *__nvme_alloc_rq(struct request_queue *q,
+				struct nvme_command *cmd, int qid,
+				blk_mq_req_flags_t flags)
 {
 	struct request *req;
-	int ret;
 
 	if (qid == NVME_QID_ANY)
 		req = blk_mq_alloc_request(q, nvme_req_op(cmd), flags);
 	else
 		req = blk_mq_alloc_request_hctx(q, nvme_req_op(cmd), flags,
 						qid - 1);
+	if (!IS_ERR(req))
+		nvme_init_request(req, cmd);
 
-	if (IS_ERR(req))
-		return PTR_ERR(req);
-	nvme_init_request(req, cmd);
+	return req;
+}
+EXPORT_SYMBOL_GPL(__nvme_alloc_rq);
+
+/*
+ * Returns 0 on success.  If the result is negative, it's a Linux error code;
+ * if the result is positive, it's an NVM Express status code
+ */
+int __nvme_submit_sync_cmd(struct request *req, union nvme_result *result,
+			   void *buffer, unsigned bufflen, int at_head)
+{
+	int ret;
 
 	if (buffer && bufflen) {
-		ret = blk_rq_map_kern(q, req, buffer, bufflen, GFP_KERNEL);
+		ret = blk_rq_map_kern(req->q, req, buffer, bufflen, GFP_KERNEL);
 		if (ret)
 			goto out;
 	}
@@ -1056,8 +1062,13 @@ EXPORT_SYMBOL_GPL(__nvme_submit_sync_cmd);
 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);
+	struct request *req;
+
+	req = __nvme_alloc_rq(q, cmd, NVME_QID_ANY, 0);
+	if (IS_ERR(req))
+		return PTR_ERR(req);
+
+	return __nvme_submit_sync_cmd(req, NULL, buffer, bufflen, 0);
 }
 EXPORT_SYMBOL_GPL(nvme_submit_sync_cmd);
 
@@ -1199,15 +1210,14 @@ static void nvme_keep_alive_work(struct work_struct *work)
 		return;
 	}
 
-	rq = blk_mq_alloc_request(ctrl->admin_q, nvme_req_op(&ctrl->ka_cmd),
-				  BLK_MQ_REQ_RESERVED | BLK_MQ_REQ_NOWAIT);
+	rq = __nvme_alloc_rq(ctrl->admin_q, &ctrl->ka_cmd, NVME_QID_ANY,
+			     BLK_MQ_REQ_RESERVED | BLK_MQ_REQ_NOWAIT);
 	if (IS_ERR(rq)) {
 		/* allocation failure, reset the controller */
 		dev_err(ctrl->device, "keep-alive failed: %ld\n", PTR_ERR(rq));
 		nvme_reset_ctrl(ctrl);
 		return;
 	}
-	nvme_init_request(rq, &ctrl->ka_cmd);
 
 	rq->timeout = ctrl->kato * HZ;
 	rq->end_io = nvme_keep_alive_end_io;
@@ -1475,6 +1485,7 @@ static int nvme_ns_info_from_id_cs_indep(struct nvme_ctrl *ctrl,
 static int nvme_features(struct nvme_ctrl *dev, u8 op, unsigned int fid,
 		unsigned int dword11, void *buffer, size_t buflen, u32 *result)
 {
+	struct request *req;
 	union nvme_result res = { 0 };
 	struct nvme_command c = { };
 	int ret;
@@ -1483,8 +1494,11 @@ static int nvme_features(struct nvme_ctrl *dev, u8 op, unsigned int fid,
 	c.features.fid = cpu_to_le32(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);
+	req = __nvme_alloc_rq(dev->admin_q, &c, NVME_QID_ANY, 0);
+	if (IS_ERR(req))
+		return PTR_ERR(req);
+
+	ret = __nvme_submit_sync_cmd(req, &res, buffer, buflen, 0);
 	if (ret >= 0 && result)
 		*result = le32_to_cpu(res.u32);
 	return ret;
@@ -2206,6 +2220,7 @@ static int nvme_sec_submit(void *data, u16 spsp, u8 secp, void *buffer, size_t l
 {
 	struct nvme_ctrl *ctrl = data;
 	struct nvme_command cmd = { };
+	struct request *req;
 
 	if (send)
 		cmd.common.opcode = nvme_admin_security_send;
@@ -2215,8 +2230,11 @@ static int nvme_sec_submit(void *data, u16 spsp, u8 secp, void *buffer, size_t l
 	cmd.common.cdw10 = cpu_to_le32(((u32)secp) << 24 | ((u32)spsp) << 8);
 	cmd.common.cdw11 = cpu_to_le32(len);
 
-	return __nvme_submit_sync_cmd(ctrl->admin_q, &cmd, NULL, buffer, len,
-			NVME_QID_ANY, 1, 0);
+	req = __nvme_alloc_rq(ctrl->admin_q, &cmd, NVME_QID_ANY, 0);
+	if (IS_ERR(req))
+		return PTR_ERR(req);
+
+	return __nvme_submit_sync_cmd(req, NULL, buffer, len, 1);
 }
 
 static void nvme_configure_opal(struct nvme_ctrl *ctrl, bool was_suspended)
diff --git a/drivers/nvme/host/fabrics.c b/drivers/nvme/host/fabrics.c
index bbaa04a0c502..7d2fd4b83bbc 100644
--- a/drivers/nvme/host/fabrics.c
+++ b/drivers/nvme/host/fabrics.c
@@ -144,6 +144,7 @@ EXPORT_SYMBOL_GPL(nvmf_get_address);
  */
 int nvmf_reg_read32(struct nvme_ctrl *ctrl, u32 off, u32 *val)
 {
+	struct request *req;
 	struct nvme_command cmd = { };
 	union nvme_result res;
 	int ret;
@@ -152,8 +153,11 @@ int nvmf_reg_read32(struct nvme_ctrl *ctrl, u32 off, u32 *val)
 	cmd.prop_get.fctype = nvme_fabrics_type_property_get;
 	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);
+	req = __nvme_alloc_rq(ctrl->fabrics_q, &cmd, NVME_QID_ANY, 0);
+	if (IS_ERR(req))
+		return PTR_ERR(req);
+
+	ret = __nvme_submit_sync_cmd(req, &res, NULL, 0, 0);
 
 	if (ret >= 0)
 		*val = le64_to_cpu(res.u64);
@@ -189,6 +193,7 @@ EXPORT_SYMBOL_GPL(nvmf_reg_read32);
  */
 int nvmf_reg_read64(struct nvme_ctrl *ctrl, u32 off, u64 *val)
 {
+	struct request *req;
 	struct nvme_command cmd = { };
 	union nvme_result res;
 	int ret;
@@ -198,8 +203,11 @@ int nvmf_reg_read64(struct nvme_ctrl *ctrl, u32 off, u64 *val)
 	cmd.prop_get.attrib = 1;
 	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);
+	req = __nvme_alloc_rq(ctrl->fabrics_q, &cmd, NVME_QID_ANY, 0);
+	if (IS_ERR(req))
+		return PTR_ERR(req);
+
+	ret = __nvme_submit_sync_cmd(req, &res, NULL, 0, 0);
 
 	if (ret >= 0)
 		*val = le64_to_cpu(res.u64);
@@ -234,6 +242,7 @@ EXPORT_SYMBOL_GPL(nvmf_reg_read64);
  */
 int nvmf_reg_write32(struct nvme_ctrl *ctrl, u32 off, u32 val)
 {
+	struct request *req;
 	struct nvme_command cmd = { };
 	int ret;
 
@@ -243,8 +252,11 @@ int nvmf_reg_write32(struct nvme_ctrl *ctrl, u32 off, u32 val)
 	cmd.prop_set.offset = cpu_to_le32(off);
 	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);
+	req = __nvme_alloc_rq(ctrl->fabrics_q, &cmd, NVME_QID_ANY, 0);
+	if (IS_ERR(req))
+		return PTR_ERR(req);
+
+	ret = __nvme_submit_sync_cmd(req, NULL, NULL, 0, 0);
 	if (unlikely(ret))
 		dev_err(ctrl->device,
 			"Property Set error: %d, offset %#x\n",
@@ -374,6 +386,7 @@ int nvmf_connect_admin_queue(struct nvme_ctrl *ctrl)
 	struct nvme_command cmd = { };
 	union nvme_result res;
 	struct nvmf_connect_data *data;
+	struct request *req;
 	int ret;
 	u32 result;
 
@@ -399,9 +412,12 @@ int nvmf_connect_admin_queue(struct nvme_ctrl *ctrl)
 	strncpy(data->subsysnqn, ctrl->opts->subsysnqn, NVMF_NQN_SIZE);
 	strncpy(data->hostnqn, ctrl->opts->host->nqn, NVMF_NQN_SIZE);
 
-	ret = __nvme_submit_sync_cmd(ctrl->fabrics_q, &cmd, &res,
-			data, sizeof(*data), NVME_QID_ANY, 1,
+	req = __nvme_alloc_rq(ctrl->fabrics_q, &cmd, NVME_QID_ANY,
 			BLK_MQ_REQ_RESERVED | BLK_MQ_REQ_NOWAIT);
+	if (IS_ERR(req))
+		return PTR_ERR(req);
+
+	ret = __nvme_submit_sync_cmd(req, &res, data, sizeof(*data), 1);
 	if (ret) {
 		nvmf_log_connect_error(ctrl, ret, le32_to_cpu(res.u32),
 				       &cmd, data);
@@ -465,6 +481,7 @@ int nvmf_connect_io_queue(struct nvme_ctrl *ctrl, u16 qid)
 	struct nvme_command cmd = { };
 	struct nvmf_connect_data *data;
 	union nvme_result res;
+	struct request *req;
 	int ret;
 	u32 result;
 
@@ -485,9 +502,12 @@ int nvmf_connect_io_queue(struct nvme_ctrl *ctrl, u16 qid)
 	strncpy(data->subsysnqn, ctrl->opts->subsysnqn, NVMF_NQN_SIZE);
 	strncpy(data->hostnqn, ctrl->opts->host->nqn, NVMF_NQN_SIZE);
 
-	ret = __nvme_submit_sync_cmd(ctrl->connect_q, &cmd, &res,
-			data, sizeof(*data), qid, 1,
+	req = __nvme_alloc_rq(ctrl->fabrics_q, &cmd, NVME_QID_ANY,
 			BLK_MQ_REQ_RESERVED | BLK_MQ_REQ_NOWAIT);
+	if (IS_ERR(req))
+		return PTR_ERR(req);
+
+	ret = __nvme_submit_sync_cmd(req, &res, data, sizeof(*data), 1);
 	if (ret) {
 		nvmf_log_connect_error(ctrl, ret, le32_to_cpu(res.u32),
 				       &cmd, data);
diff --git a/drivers/nvme/host/nvme.h b/drivers/nvme/host/nvme.h
index bf46f122e9e1..d10f930d1edb 100644
--- a/drivers/nvme/host/nvme.h
+++ b/drivers/nvme/host/nvme.h
@@ -813,10 +813,10 @@ 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(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);
+struct request *__nvme_alloc_rq(struct request_queue *q,
+		struct nvme_command *cmd, int qid, blk_mq_req_flags_t flags);
+int __nvme_submit_sync_cmd(struct request *req, union nvme_result *result,
+		void *buffer, unsigned bufflen, int at_head);
 int nvme_set_features(struct nvme_ctrl *dev, unsigned int fid,
 		      unsigned int dword11, void *buffer, size_t buflen,
 		      u32 *result);
diff --git a/drivers/nvme/host/pci.c b/drivers/nvme/host/pci.c
index a331fbfa9a66..8e3bae7d489b 100644
--- a/drivers/nvme/host/pci.c
+++ b/drivers/nvme/host/pci.c
@@ -1376,13 +1376,12 @@ static enum blk_eh_timer_return nvme_timeout(struct request *req)
 		 nvme_get_opcode_str(nvme_req(req)->cmd->common.opcode),
 		 nvmeq->qid);
 
-	abort_req = blk_mq_alloc_request(dev->ctrl.admin_q, nvme_req_op(&cmd),
-					 BLK_MQ_REQ_NOWAIT);
+	abort_req = __nvme_alloc_rq(dev->ctrl.admin_q, &cmd,
+			NVME_QID_ANY, BLK_MQ_REQ_NOWAIT);
 	if (IS_ERR(abort_req)) {
 		atomic_inc(&dev->ctrl.abort_limit);
 		return BLK_EH_RESET_TIMER;
 	}
-	nvme_init_request(abort_req, &cmd);
 
 	abort_req->end_io = abort_endio;
 	abort_req->end_io_data = NULL;
@@ -2391,10 +2390,9 @@ static int nvme_delete_queue(struct nvme_queue *nvmeq, u8 opcode)
 	cmd.delete_queue.opcode = opcode;
 	cmd.delete_queue.qid = cpu_to_le16(nvmeq->qid);
 
-	req = blk_mq_alloc_request(q, nvme_req_op(&cmd), BLK_MQ_REQ_NOWAIT);
+	req = __nvme_alloc_rq(q, &cmd, NVME_QID_ANY, BLK_MQ_REQ_NOWAIT);
 	if (IS_ERR(req))
 		return PTR_ERR(req);
-	nvme_init_request(req, &cmd);
 
 	if (opcode == nvme_admin_delete_cq)
 		req->end_io = nvme_del_cq_end;
diff --git a/drivers/nvme/target/passthru.c b/drivers/nvme/target/passthru.c
index 511c980d538d..0dae824a2d05 100644
--- a/drivers/nvme/target/passthru.c
+++ b/drivers/nvme/target/passthru.c
@@ -316,12 +316,11 @@ static void nvmet_passthru_execute_cmd(struct nvmet_req *req)
 		timeout = nvmet_req_subsys(req)->admin_timeout;
 	}
 
-	rq = blk_mq_alloc_request(q, nvme_req_op(req->cmd), 0);
+	rq = __nvme_alloc_rq(q, req->cmd, NVME_QID_ANY, 0);
 	if (IS_ERR(rq)) {
 		status = NVME_SC_INTERNAL;
 		goto out_put_ns;
 	}
-	nvme_init_request(rq, req->cmd);
 
 	if (timeout)
 		rq->timeout = timeout;
-- 
2.35.3

[PATCH 2/3] nvme: retry authentication commands if DNR status bit is not set

[Hannes Reinecke]

Clear the FAILFAST_DRIVER bit for authentication commands
allowing them to be retried in nvme_decide_disposition() if the DNR
bit is not set in the command result.

Signed-off-by: Hannes Reinecke <hare@suse.de>
---
 drivers/nvme/host/auth.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/drivers/nvme/host/auth.c b/drivers/nvme/host/auth.c
index dcea6873ca2c..8a1edbf69b47 100644
--- a/drivers/nvme/host/auth.c
+++ b/drivers/nvme/host/auth.c
@@ -81,6 +81,8 @@ static int nvme_auth_submit(struct nvme_ctrl *ctrl, int qid,
 	if (IS_ERR(req))
 		return PTR_ERR(req);
 
+	/* Clear failfast flag to allow for retries */
+	req->cmd_flags &= ~REQ_FAILFAST_DRIVER;
 	ret = __nvme_submit_sync_cmd(req, NULL, data, data_len, 0);
 	if (ret > 0)
 		dev_warn(ctrl->device,
-- 
2.35.3

[PATCH 3/3] nvme: make 'at_head' parameter for __nvme_submit_sync_cmd() boolean

[Hannes Reinecke]

Use a boolean for the 'at_head' parameter in __nvme_submit_sync_cmd()
to improve readability and consistency; it's being converted into a
boolean anyway.

Signed-off-by: Hannes Reinecke <hare@suse.de>
---
 drivers/nvme/host/auth.c    |  2 +-
 drivers/nvme/host/core.c    |  8 ++++----
 drivers/nvme/host/fabrics.c | 12 +++++-------
 drivers/nvme/host/nvme.h    |  2 +-
 4 files changed, 11 insertions(+), 13 deletions(-)

diff --git a/drivers/nvme/host/auth.c b/drivers/nvme/host/auth.c
index 8a1edbf69b47..d4002845064a 100644
--- a/drivers/nvme/host/auth.c
+++ b/drivers/nvme/host/auth.c
@@ -83,7 +83,7 @@ static int nvme_auth_submit(struct nvme_ctrl *ctrl, int qid,
 
 	/* Clear failfast flag to allow for retries */
 	req->cmd_flags &= ~REQ_FAILFAST_DRIVER;
-	ret = __nvme_submit_sync_cmd(req, NULL, data, data_len, 0);
+	ret = __nvme_submit_sync_cmd(req, NULL, data, data_len, false);
 	if (ret > 0)
 		dev_warn(ctrl->device,
 			"qid %d auth_send failed with status %d\n", qid, ret);
diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c
index 278f7aa674c6..3a4b5e99e648 100644
--- a/drivers/nvme/host/core.c
+++ b/drivers/nvme/host/core.c
@@ -1040,7 +1040,7 @@ EXPORT_SYMBOL_GPL(__nvme_alloc_rq);
  * if the result is positive, it's an NVM Express status code
  */
 int __nvme_submit_sync_cmd(struct request *req, union nvme_result *result,
-			   void *buffer, unsigned bufflen, int at_head)
+			   void *buffer, unsigned bufflen, bool at_head)
 {
 	int ret;
 
@@ -1068,7 +1068,7 @@ int nvme_submit_sync_cmd(struct request_queue *q, struct nvme_command *cmd,
 	if (IS_ERR(req))
 		return PTR_ERR(req);
 
-	return __nvme_submit_sync_cmd(req, NULL, buffer, bufflen, 0);
+	return __nvme_submit_sync_cmd(req, NULL, buffer, bufflen, false);
 }
 EXPORT_SYMBOL_GPL(nvme_submit_sync_cmd);
 
@@ -1498,7 +1498,7 @@ static int nvme_features(struct nvme_ctrl *dev, u8 op, unsigned int fid,
 	if (IS_ERR(req))
 		return PTR_ERR(req);
 
-	ret = __nvme_submit_sync_cmd(req, &res, buffer, buflen, 0);
+	ret = __nvme_submit_sync_cmd(req, &res, buffer, buflen, false);
 	if (ret >= 0 && result)
 		*result = le32_to_cpu(res.u32);
 	return ret;
@@ -2234,7 +2234,7 @@ static int nvme_sec_submit(void *data, u16 spsp, u8 secp, void *buffer, size_t l
 	if (IS_ERR(req))
 		return PTR_ERR(req);
 
-	return __nvme_submit_sync_cmd(req, NULL, buffer, len, 1);
+	return __nvme_submit_sync_cmd(req, NULL, buffer, len, true);
 }
 
 static void nvme_configure_opal(struct nvme_ctrl *ctrl, bool was_suspended)
diff --git a/drivers/nvme/host/fabrics.c b/drivers/nvme/host/fabrics.c
index 7d2fd4b83bbc..db11c54b8d0e 100644
--- a/drivers/nvme/host/fabrics.c
+++ b/drivers/nvme/host/fabrics.c
@@ -157,8 +157,7 @@ int nvmf_reg_read32(struct nvme_ctrl *ctrl, u32 off, u32 *val)
 	if (IS_ERR(req))
 		return PTR_ERR(req);
 
-	ret = __nvme_submit_sync_cmd(req, &res, NULL, 0, 0);
-
+	ret = __nvme_submit_sync_cmd(req, &res, NULL, 0, false);
 	if (ret >= 0)
 		*val = le64_to_cpu(res.u64);
 	if (unlikely(ret != 0))
@@ -207,8 +206,7 @@ int nvmf_reg_read64(struct nvme_ctrl *ctrl, u32 off, u64 *val)
 	if (IS_ERR(req))
 		return PTR_ERR(req);
 
-	ret = __nvme_submit_sync_cmd(req, &res, NULL, 0, 0);
-
+	ret = __nvme_submit_sync_cmd(req, &res, NULL, 0, false);
 	if (ret >= 0)
 		*val = le64_to_cpu(res.u64);
 	if (unlikely(ret != 0))
@@ -256,7 +254,7 @@ int nvmf_reg_write32(struct nvme_ctrl *ctrl, u32 off, u32 val)
 	if (IS_ERR(req))
 		return PTR_ERR(req);
 
-	ret = __nvme_submit_sync_cmd(req, NULL, NULL, 0, 0);
+	ret = __nvme_submit_sync_cmd(req, NULL, NULL, 0, false);
 	if (unlikely(ret))
 		dev_err(ctrl->device,
 			"Property Set error: %d, offset %#x\n",
@@ -417,7 +415,7 @@ int nvmf_connect_admin_queue(struct nvme_ctrl *ctrl)
 	if (IS_ERR(req))
 		return PTR_ERR(req);
 
-	ret = __nvme_submit_sync_cmd(req, &res, data, sizeof(*data), 1);
+	ret = __nvme_submit_sync_cmd(req, &res, data, sizeof(*data), true);
 	if (ret) {
 		nvmf_log_connect_error(ctrl, ret, le32_to_cpu(res.u32),
 				       &cmd, data);
@@ -507,7 +505,7 @@ int nvmf_connect_io_queue(struct nvme_ctrl *ctrl, u16 qid)
 	if (IS_ERR(req))
 		return PTR_ERR(req);
 
-	ret = __nvme_submit_sync_cmd(req, &res, data, sizeof(*data), 1);
+	ret = __nvme_submit_sync_cmd(req, &res, data, sizeof(*data), true);
 	if (ret) {
 		nvmf_log_connect_error(ctrl, ret, le32_to_cpu(res.u32),
 				       &cmd, data);
diff --git a/drivers/nvme/host/nvme.h b/drivers/nvme/host/nvme.h
index d10f930d1edb..bf984ae5cf04 100644
--- a/drivers/nvme/host/nvme.h
+++ b/drivers/nvme/host/nvme.h
@@ -816,7 +816,7 @@ int nvme_submit_sync_cmd(struct request_queue *q, struct nvme_command *cmd,
 struct request *__nvme_alloc_rq(struct request_queue *q,
 		struct nvme_command *cmd, int qid, blk_mq_req_flags_t flags);
 int __nvme_submit_sync_cmd(struct request *req, union nvme_result *result,
-		void *buffer, unsigned bufflen, int at_head);
+		void *buffer, unsigned bufflen, bool at_head);
 int nvme_set_features(struct nvme_ctrl *dev, unsigned int fid,
 		      unsigned int dword11, void *buffer, size_t buflen,
 		      u32 *result);
-- 
2.35.3

Re: [PATCHv2 0/3] nvme: rework __nvme_submit_sync_cmd()

[Jens Axboe]

On 2/8/23 8:17?AM, Hannes Reinecke wrote:
> Hi all,
> 
> as Jens and hch complained about the argument list to __nvme_submit_sync_cmd()
> getting too long, here's now a patchset to clean things up.
> The first patch is to split off __nvme_alloc_rq() from __nvme_submit_sync_cmd(),
> which serves to reduce the number of arguments; it's also a nice cleanup as this
> function can be utilized to replace some open-coded versions.
> The second patch is to introduce a 'retry' argument to __nvme_alloc_rq(), to
> indicate whether retries for this command should be allowd.
> And the third is another cleanup; the 'at_head' argument to __nvme_submit_sync_cmd()
> really should be a boolean, as it can take only two values, and it's being
> converted into a boolean within __nvme_submit_sync_cmd() anyway.

This looks WAY better than piling on to __nvme_submit_sync_cmd(). I
haven't taken a closer look yet, just a note that I'd reorder patch 2
and 3 so that you just have two prep patches and then the final patch is
what you really wanted to do. Doing cleanup, change, cleanup is a bit
odd.

-- 
Jens Axboe

Re: [PATCH 3/3] nvme: make 'at_head' parameter for __nvme_submit_sync_cmd() boolean

[Christoph Hellwig]

On Wed, Feb 08, 2023 at 04:17:20PM +0100, Hannes Reinecke wrote:
> Use a boolean for the 'at_head' parameter in __nvme_submit_sync_cmd()
> to improve readability and consistency; it's being converted into a
> boolean anyway.

I'd much rather have an INSERT_AT_HEAD bit flag, going all the
way down to the block layer.  The use of a boolean for this (and
also some other things in the block layer issue path) is really confusing.

Re: [PATCH 1/3] nvme: split __nvme_submit_sync_cmd()

[Christoph Hellwig]

On Wed, Feb 08, 2023 at 04:17:18PM +0100, Hannes Reinecke wrote:
> Split a __nvme_alloc_rq() function from __nvme_submit_sync_cmd()
> to reduce the number of arguments.

But now everyone has to call both?  

>  6 files changed, 83 insertions(+), 45 deletions(-)

.. and the code is a lot longer.  So this doesn't really seem like
much of a win?
> +struct request *__nvme_alloc_rq(struct request_queue *q,
> +				struct nvme_command *cmd, int qid,
> +				blk_mq_req_flags_t flags)

Why the double underscore profix?  Why _rq instead of _request
like blk_mq_alloc_request and nvme_init_request?

>  {
>  	struct request *req;
>  
>  	if (qid == NVME_QID_ANY)
>  		req = blk_mq_alloc_request(q, nvme_req_op(cmd), flags);
>  	else
>  		req = blk_mq_alloc_request_hctx(q, nvme_req_op(cmd), flags,
>  						qid - 1);
> +	if (!IS_ERR(req))
> +		nvme_init_request(req, cmd);
>  
> +	return req;

And I'd at very least split out the qid case as that is substantially
different and only used in very specific places.

> + */
> +int __nvme_submit_sync_cmd(struct request *req, union nvme_result *result,
> +			   void *buffer, unsigned bufflen, int at_head)
> +{
> +	int ret;
>  
>  	if (buffer && bufflen) {
> -		ret = blk_rq_map_kern(q, req, buffer, bufflen, GFP_KERNEL);
> +		ret = blk_rq_map_kern(req->q, req, buffer, bufflen, GFP_KERNEL);
>  		if (ret)
>  			goto out;

This new __nvme_submit_sync_cmd now consumes the request, which
is an odd calling conventions.

What do you think about:

 - passing the "union nvme_result *result" to nvme_execute_rq, and do the
   conditional assignment to it there, where it fits along with all the
   status management
 - make blk_rq_map_kern handle a NULL kbuf gracefully instead of doing
   that in a lot of the callers
 - just open code __nvme_submit_sync_cmd using these building blocks.

Re: [PATCH 1/3] nvme: split __nvme_submit_sync_cmd()

[Hannes Reinecke]

On 2/9/23 06:33, Christoph Hellwig wrote:
> On Wed, Feb 08, 2023 at 04:17:18PM +0100, Hannes Reinecke wrote:
>> Split a __nvme_alloc_rq() function from __nvme_submit_sync_cmd()
>> to reduce the number of arguments.
> 
> But now everyone has to call both?
> 
Yes, that's the downside of it.

>>   6 files changed, 83 insertions(+), 45 deletions(-)
> 
> .. and the code is a lot longer.  So this doesn't really seem like
> much of a win?
>> +struct request *__nvme_alloc_rq(struct request_queue *q,
>> +				struct nvme_command *cmd, int qid,
>> +				blk_mq_req_flags_t flags)
> 
> Why the double underscore profix?  Why _rq instead of _request
> like blk_mq_alloc_request and nvme_init_request?
> 
Ok, will be renaming to nvme_alloc_request().

>>   {
>>   	struct request *req;
>>   
>>   	if (qid == NVME_QID_ANY)
>>   		req = blk_mq_alloc_request(q, nvme_req_op(cmd), flags);
>>   	else
>>   		req = blk_mq_alloc_request_hctx(q, nvme_req_op(cmd), flags,
>>   						qid - 1);
>> +	if (!IS_ERR(req))
>> +		nvme_init_request(req, cmd);
>>   
>> +	return req;
> 
> And I'd at very least split out the qid case as that is substantially
> different and only used in very specific places.
> 
>> + */
>> +int __nvme_submit_sync_cmd(struct request *req, union nvme_result *result,
>> +			   void *buffer, unsigned bufflen, int at_head)
>> +{
>> +	int ret;
>>   
>>   	if (buffer && bufflen) {
>> -		ret = blk_rq_map_kern(q, req, buffer, bufflen, GFP_KERNEL);
>> +		ret = blk_rq_map_kern(req->q, req, buffer, bufflen, GFP_KERNEL);
>>   		if (ret)
>>   			goto out;
> 
> This new __nvme_submit_sync_cmd now consumes the request, which
> is an odd calling conventions.
> 
> What do you think about:
> 
>   - passing the "union nvme_result *result" to nvme_execute_rq, and do the
>     conditional assignment to it there, where it fits along with all the
>     status management
>   - make blk_rq_map_kern handle a NULL kbuf gracefully instead of doing
>     that in a lot of the callers
>   - just open code __nvme_submit_sync_cmd using these building blocks.

Hmm. Let me see how this pans out.

Cheers,

Hannes
-- 
Dr. Hannes Reinecke                Kernel Storage Architect
hare@suse.de                              +49 911 74053 688
SUSE Software Solutions GmbH, Maxfeldstr. 5, 90409 Nürnberg
HRB 36809 (AG Nürnberg), Geschäftsführer: Ivo Totev, Andrew
Myers, Andrew McDonald, Martje Boudien Moerman

Re: [PATCH 3/3] nvme: make 'at_head' parameter for __nvme_submit_sync_cmd() boolean

[Hannes Reinecke]

On 2/9/23 06:20, Christoph Hellwig wrote:
> On Wed, Feb 08, 2023 at 04:17:20PM +0100, Hannes Reinecke wrote:
>> Use a boolean for the 'at_head' parameter in __nvme_submit_sync_cmd()
>> to improve readability and consistency; it's being converted into a
>> boolean anyway.
> 
> I'd much rather have an INSERT_AT_HEAD bit flag, going all the
> way down to the block layer.  The use of a boolean for this (and
> also some other things in the block layer issue path) is really confusing.

I do agree. But that turns out to be a larger and more complex patchset, 
which I'd rather send as a separate thingie after this one is done.
Otherwise it'll hold of this patchset, and really is an unrelated issue.

Cheers,

Hannes

Re: [PATCH 3/3] nvme: make 'at_head' parameter for __nvme_submit_sync_cmd() boolean

[Kanchan Joshi]

[-- Attachment #1: Type: text/plain, Size: 828 bytes --]

On Thu, Feb 09, 2023 at 11:59:49AM +0100, Hannes Reinecke wrote:
>On 2/9/23 06:20, Christoph Hellwig wrote:
>>On Wed, Feb 08, 2023 at 04:17:20PM +0100, Hannes Reinecke wrote:
>>>Use a boolean for the 'at_head' parameter in __nvme_submit_sync_cmd()
>>>to improve readability and consistency; it's being converted into a
>>>boolean anyway.
>>
>>I'd much rather have an INSERT_AT_HEAD bit flag, going all the
>>way down to the block layer.  The use of a boolean for this (and
>>also some other things in the block layer issue path) is really confusing.
>
>I do agree. But that turns out to be a larger and more complex 
>patchset, which I'd rather send as a separate thingie after this one 
>is done.
>Otherwise it'll hold of this patchset, and really is an unrelated issue.

main patch (#2) is no longer coupled to other two now.

[-- Attachment #2: Type: text/plain, Size: 0 bytes --]