[PATCH 00/12] qla2xxx misc features and bug fixes

[PATCH 00/12] qla2xxx misc features and bug fixes

[Nilesh Javali]

Martin,

Please apply the attached miscellaneous qla2xxx features and bug fixes
to the scsi tree at your earliest convenience.

Thanks,
Nilesh

Arun Easi (7):
  qla2xxx: Fix I/O failures during remote port toggle testing
  qla2xxx: Setup debugfs entries for remote ports
  qla2xxx: Allow dev_loss_tmo setting for FC-NVMe devices
  qla2xxx: Honor status qualifier in FCP_RSP per spec
  qla2xxx: Fix I/O errors during LIP reset tests
  qla2xxx: Make tgt_port_database available in initiator mode
  qla2xxx: Add rport fields in debugfs

Quinn Tran (4):
  qla2xxx: Reduce duplicate code in reporting speed
  qla2xxx: Fix memory size truncation
  qla2xxx: performance tweak
  qla2xxx: Add IOCB resource tracking

Saurav Kashyap (1):
  qla2xxx: Add SLER and PI control support

 drivers/scsi/qla2xxx/qla_attr.c   |  87 ++++++------
 drivers/scsi/qla2xxx/qla_dbg.c    |   2 +-
 drivers/scsi/qla2xxx/qla_def.h    |  59 ++++++--
 drivers/scsi/qla2xxx/qla_dfs.c    | 227 +++++++++++++++++++++++++-----
 drivers/scsi/qla2xxx/qla_fw.h     |   2 +-
 drivers/scsi/qla2xxx/qla_gbl.h    |   7 +
 drivers/scsi/qla2xxx/qla_gs.c     |   7 +-
 drivers/scsi/qla2xxx/qla_init.c   |  33 ++++-
 drivers/scsi/qla2xxx/qla_inline.h |  93 +++++++++++-
 drivers/scsi/qla2xxx/qla_iocb.c   |  54 +++++--
 drivers/scsi/qla2xxx/qla_isr.c    |  28 ++--
 drivers/scsi/qla2xxx/qla_mbx.c    |  28 +++-
 drivers/scsi/qla2xxx/qla_mid.c    |   4 +-
 drivers/scsi/qla2xxx/qla_nvme.c   |  36 ++++-
 drivers/scsi/qla2xxx/qla_nvme.h   |   4 +-
 drivers/scsi/qla2xxx/qla_os.c     | 118 +++-------------
 drivers/scsi/qla2xxx/qla_target.c |   2 +
 17 files changed, 549 insertions(+), 242 deletions(-)


base-commit: dca93232b361d260413933903cd4bdbd92ebcc7f
-- 
2.19.0.rc0

[PATCH 01/12] qla2xxx: Fix I/O failures during remote port toggle testing

[Nilesh Javali]

From: Arun Easi <aeasi@marvell.com>

Driver was using a lower value for dev_loss_tmo making it more prone to
I/O failures during remote port toggle testing. Set dev_loss_tmo to zero
during remote port registration to allow nvme-fc default dev_loss_tmo to
be used, which is higher than what driver was using.

Signed-off-by: Arun Easi <aeasi@marvell.com>
Signed-off-by: Himanshu Madhani <hmadhani@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
---
 drivers/scsi/qla2xxx/qla_nvme.c | 2 +-
 drivers/scsi/qla2xxx/qla_nvme.h | 3 ---
 2 files changed, 1 insertion(+), 4 deletions(-)

diff --git a/drivers/scsi/qla2xxx/qla_nvme.c b/drivers/scsi/qla2xxx/qla_nvme.c
index 90bbc61f361b..435f17f87301 100644
--- a/drivers/scsi/qla2xxx/qla_nvme.c
+++ b/drivers/scsi/qla2xxx/qla_nvme.c
@@ -42,7 +42,7 @@ int qla_nvme_register_remote(struct scsi_qla_host *vha, struct fc_port *fcport)
 	req.port_name = wwn_to_u64(fcport->port_name);
 	req.node_name = wwn_to_u64(fcport->node_name);
 	req.port_role = 0;
-	req.dev_loss_tmo = NVME_FC_DEV_LOSS_TMO;
+	req.dev_loss_tmo = 0;
 
 	if (fcport->nvme_prli_service_param & NVME_PRLI_SP_INITIATOR)
 		req.port_role = FC_PORT_ROLE_NVME_INITIATOR;
diff --git a/drivers/scsi/qla2xxx/qla_nvme.h b/drivers/scsi/qla2xxx/qla_nvme.h
index fbb844226630..cf45a5b277f1 100644
--- a/drivers/scsi/qla2xxx/qla_nvme.h
+++ b/drivers/scsi/qla2xxx/qla_nvme.h
@@ -14,9 +14,6 @@
 #include "qla_def.h"
 #include "qla_dsd.h"
 
-/* default dev loss time (seconds) before transport tears down ctrl */
-#define NVME_FC_DEV_LOSS_TMO  30
-
 #define NVME_ATIO_CMD_OFF 32
 #define NVME_FIRST_PACKET_CMDLEN (64 - NVME_ATIO_CMD_OFF)
 #define Q2T_NVME_NUM_TAGS 2048
-- 
2.19.0.rc0

[PATCH 02/12] qla2xxx: Setup debugfs entries for remote ports

[Nilesh Javali]

From: Arun Easi <aeasi@marvell.com>

Create a base for adding remote port related entries in debugfs.

Signed-off-by: Arun Easi <aeasi@marvell.com>
Signed-off-by: Himanshu Madhani <hmadhani@marvell.com>
Signed-off-by: Saurav Kashyap <skashyap@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
---
 drivers/scsi/qla2xxx/qla_def.h    |  4 +++
 drivers/scsi/qla2xxx/qla_dfs.c    | 42 ++++++++++++++++++++++++++++++-
 drivers/scsi/qla2xxx/qla_gbl.h    |  2 ++
 drivers/scsi/qla2xxx/qla_init.c   |  2 ++
 drivers/scsi/qla2xxx/qla_target.c |  2 ++
 5 files changed, 51 insertions(+), 1 deletion(-)

diff --git a/drivers/scsi/qla2xxx/qla_def.h b/drivers/scsi/qla2xxx/qla_def.h
index 1bc090d8a71b..074d8753cfc3 100644
--- a/drivers/scsi/qla2xxx/qla_def.h
+++ b/drivers/scsi/qla2xxx/qla_def.h
@@ -2544,6 +2544,8 @@ typedef struct fc_port {
 	u8 last_login_state;
 	u16 n2n_link_reset_cnt;
 	u16 n2n_chip_reset;
+
+	struct dentry *dfs_rport_dir;
 } fc_port_t;
 
 enum {
@@ -4780,6 +4782,8 @@ typedef struct scsi_qla_host {
 	uint16_t ql2xexchoffld;
 	uint16_t ql2xiniexchg;
 
+	struct dentry *dfs_rport_root;
+
 	struct purex_list {
 		struct list_head head;
 		spinlock_t lock;
diff --git a/drivers/scsi/qla2xxx/qla_dfs.c b/drivers/scsi/qla2xxx/qla_dfs.c
index e62b2115235e..3c4b9b549b17 100644
--- a/drivers/scsi/qla2xxx/qla_dfs.c
+++ b/drivers/scsi/qla2xxx/qla_dfs.c
@@ -12,6 +12,29 @@
 static struct dentry *qla2x00_dfs_root;
 static atomic_t qla2x00_dfs_root_count;
 
+void
+qla2x00_dfs_create_rport(scsi_qla_host_t *vha, struct fc_port *fp)
+{
+	char wwn[32];
+
+	if (!vha->dfs_rport_root || fp->dfs_rport_dir)
+		return;
+
+	sprintf(wwn, "pn-%016llx", wwn_to_u64(fp->port_name));
+	fp->dfs_rport_dir = debugfs_create_dir(wwn, vha->dfs_rport_root);
+	if (!fp->dfs_rport_dir)
+		return;
+}
+
+void
+qla2x00_dfs_remove_rport(scsi_qla_host_t *vha, struct fc_port *fp)
+{
+	if (!vha->dfs_rport_root || !fp->dfs_rport_dir)
+		return;
+	debugfs_remove_recursive(fp->dfs_rport_dir);
+	fp->dfs_rport_dir = NULL;
+}
+
 static int
 qla2x00_dfs_tgt_sess_show(struct seq_file *s, void *unused)
 {
@@ -473,9 +496,21 @@ qla2x00_dfs_setup(scsi_qla_host_t *vha)
 	ha->tgt.dfs_tgt_sess = debugfs_create_file("tgt_sess",
 		S_IRUSR, ha->dfs_dir, vha, &dfs_tgt_sess_ops);
 
-	if (IS_QLA27XX(ha) || IS_QLA83XX(ha) || IS_QLA28XX(ha))
+	if (IS_QLA27XX(ha) || IS_QLA83XX(ha) || IS_QLA28XX(ha)) {
 		ha->tgt.dfs_naqp = debugfs_create_file("naqp",
 		    0400, ha->dfs_dir, vha, &dfs_naqp_ops);
+		if (!ha->tgt.dfs_naqp) {
+			ql_log(ql_log_warn, vha, 0xd011,
+			       "Unable to create debugFS naqp node.\n");
+			goto out;
+		}
+	}
+	vha->dfs_rport_root = debugfs_create_dir("rports", ha->dfs_dir);
+	if (!vha->dfs_rport_root) {
+		ql_log(ql_log_warn, vha, 0xd012,
+		       "Unable to create debugFS rports node.\n");
+		goto out;
+	}
 out:
 	return 0;
 }
@@ -515,6 +550,11 @@ qla2x00_dfs_remove(scsi_qla_host_t *vha)
 		ha->dfs_fce = NULL;
 	}
 
+	if (vha->dfs_rport_root) {
+		debugfs_remove_recursive(vha->dfs_rport_root);
+		vha->dfs_rport_root = NULL;
+	}
+
 	if (ha->dfs_dir) {
 		debugfs_remove(ha->dfs_dir);
 		ha->dfs_dir = NULL;
diff --git a/drivers/scsi/qla2xxx/qla_gbl.h b/drivers/scsi/qla2xxx/qla_gbl.h
index 0ced18f3104e..36c210c24f72 100644
--- a/drivers/scsi/qla2xxx/qla_gbl.h
+++ b/drivers/scsi/qla2xxx/qla_gbl.h
@@ -935,6 +935,8 @@ void qlt_clr_qp_table(struct scsi_qla_host *vha);
 void qlt_set_mode(struct scsi_qla_host *);
 int qla2x00_set_data_rate(scsi_qla_host_t *vha, uint16_t mode);
 extern void qla24xx_process_purex_list(struct purex_list *);
+extern void qla2x00_dfs_create_rport(scsi_qla_host_t *vha, struct fc_port *fp);
+extern void qla2x00_dfs_remove_rport(scsi_qla_host_t *vha, struct fc_port *fp);
 
 /* nvme.c */
 void qla_nvme_unregister_remote_port(struct fc_port *fcport);
diff --git a/drivers/scsi/qla2xxx/qla_init.c b/drivers/scsi/qla2xxx/qla_init.c
index 57a2d76aa691..b4d53eb4e53e 100644
--- a/drivers/scsi/qla2xxx/qla_init.c
+++ b/drivers/scsi/qla2xxx/qla_init.c
@@ -5486,6 +5486,8 @@ qla2x00_update_fcport(scsi_qla_host_t *vha, fc_port_t *fcport)
 
 	qla2x00_iidma_fcport(vha, fcport);
 
+	qla2x00_dfs_create_rport(vha, fcport);
+
 	if (NVME_TARGET(vha->hw, fcport)) {
 		qla_nvme_register_remote(vha, fcport);
 		qla2x00_set_fcport_disc_state(fcport, DSC_LOGIN_COMPLETE);
diff --git a/drivers/scsi/qla2xxx/qla_target.c b/drivers/scsi/qla2xxx/qla_target.c
index 90289162dbd4..7711f95033c0 100644
--- a/drivers/scsi/qla2xxx/qla_target.c
+++ b/drivers/scsi/qla2xxx/qla_target.c
@@ -1111,6 +1111,8 @@ void qlt_free_session_done(struct work_struct *work)
 	spin_unlock_irqrestore(&ha->tgt.sess_lock, flags);
 	sess->free_pending = 0;
 
+	qla2x00_dfs_remove_rport(vha, sess);
+
 	ql_dbg(ql_dbg_disc, vha, 0xf001,
 	    "Unregistration of sess %p %8phC finished fcp_cnt %d\n",
 		sess, sess->port_name, vha->fcport_count);
-- 
2.19.0.rc0

[PATCH 03/12] qla2xxx: Allow dev_loss_tmo setting for FC-NVMe devices

[Nilesh Javali]

From: Arun Easi <aeasi@marvell.com>

Add a remote port debugfs entry to get/set dev_loss_tmo for NVMe
devices.

Signed-off-by: Arun Easi <aeasi@marvell.com>
Signed-off-by: Himanshu Madhani <hmadhani@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
---
 drivers/scsi/qla2xxx/qla_dfs.c | 54 ++++++++++++++++++++++++++++++++++
 1 file changed, 54 insertions(+)

diff --git a/drivers/scsi/qla2xxx/qla_dfs.c b/drivers/scsi/qla2xxx/qla_dfs.c
index 3c4b9b549b17..7482e6bf7f7f 100644
--- a/drivers/scsi/qla2xxx/qla_dfs.c
+++ b/drivers/scsi/qla2xxx/qla_dfs.c
@@ -12,6 +12,57 @@
 static struct dentry *qla2x00_dfs_root;
 static atomic_t qla2x00_dfs_root_count;
 
+#define QLA_DFS_RPORT_DEVLOSS_TMO	1
+
+static int
+qla_dfs_rport_get(struct fc_port *fp, int attr_id, u64 *val)
+{
+	switch (attr_id) {
+	case QLA_DFS_RPORT_DEVLOSS_TMO:
+		/* Only supported for FC-NVMe devices that are registered. */
+		if (!(fp->nvme_flag & NVME_FLAG_REGISTERED))
+			return -EIO;
+		*val = fp->nvme_remote_port->dev_loss_tmo;
+		break;
+	default:
+		return -EINVAL;
+	}
+	return 0;
+}
+
+static int
+qla_dfs_rport_set(struct fc_port *fp, int attr_id, u64 val)
+{
+	switch (attr_id) {
+	case QLA_DFS_RPORT_DEVLOSS_TMO:
+		/* Only supported for FC-NVMe devices that are registered. */
+		if (!(fp->nvme_flag & NVME_FLAG_REGISTERED))
+			return -EIO;
+		return nvme_fc_set_remoteport_devloss(fp->nvme_remote_port,
+						      val);
+	default:
+		return -EINVAL;
+	}
+	return 0;
+}
+
+#define DEFINE_QLA_DFS_RPORT_RW_ATTR(_attr_id, _attr)		\
+static int qla_dfs_rport_##_attr##_get(void *data, u64 *val)	\
+{								\
+	struct fc_port *fp = data;				\
+	return qla_dfs_rport_get(fp, _attr_id, val);		\
+}								\
+static int qla_dfs_rport_##_attr##_set(void *data, u64 val)	\
+{								\
+	struct fc_port *fp = data;				\
+	return qla_dfs_rport_set(fp, _attr_id, val);		\
+}								\
+DEFINE_DEBUGFS_ATTRIBUTE(qla_dfs_rport_##_attr##_fops,		\
+		qla_dfs_rport_##_attr##_get,			\
+		qla_dfs_rport_##_attr##_set, "%llu\n")
+
+DEFINE_QLA_DFS_RPORT_RW_ATTR(QLA_DFS_RPORT_DEVLOSS_TMO, dev_loss_tmo);
+
 void
 qla2x00_dfs_create_rport(scsi_qla_host_t *vha, struct fc_port *fp)
 {
@@ -24,6 +75,9 @@ qla2x00_dfs_create_rport(scsi_qla_host_t *vha, struct fc_port *fp)
 	fp->dfs_rport_dir = debugfs_create_dir(wwn, vha->dfs_rport_root);
 	if (!fp->dfs_rport_dir)
 		return;
+	if (NVME_TARGET(vha->hw, fp))
+		debugfs_create_file("dev_loss_tmo", 0600, fp->dfs_rport_dir,
+				    fp, &qla_dfs_rport_dev_loss_tmo_fops);
 }
 
 void
-- 
2.19.0.rc0

[PATCH 04/12] qla2xxx: Honor status qualifier in FCP_RSP per spec

[Nilesh Javali]

From: Arun Easi <aeasi@marvell.com>

FCP-4 (referred FCP-4 rev-2b) identifies the earlier known "retry delay
timer" field as "status qualifier", which is described in SAM-5 and
later specs. This fix makes appropriate driver side modifications to
honor the new definition. The SAM document referred was SAM-6 rev-5.

Signed-off-by: Arun Easi <aeasi@marvell.com>
Signed-off-by: Himanshu Madhani <hmadhani@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
---
 drivers/scsi/qla2xxx/qla_fw.h     |  2 +-
 drivers/scsi/qla2xxx/qla_inline.h | 38 +++++++++++++++++++++++++++----
 drivers/scsi/qla2xxx/qla_isr.c    | 18 ++++-----------
 3 files changed, 40 insertions(+), 18 deletions(-)

diff --git a/drivers/scsi/qla2xxx/qla_fw.h b/drivers/scsi/qla2xxx/qla_fw.h
index bba1b77fba7e..f0052d75849c 100644
--- a/drivers/scsi/qla2xxx/qla_fw.h
+++ b/drivers/scsi/qla2xxx/qla_fw.h
@@ -619,7 +619,7 @@ struct sts_entry_24xx {
 #define SF_NVME_ERSP            BIT_6
 #define SF_FCP_RSP_DMA		BIT_0
 
-	__le16	retry_delay;
+	__le16	status_qualifier;
 	__le16	scsi_status;		/* SCSI status. */
 #define SS_CONFIRMATION_REQ		BIT_12
 
diff --git a/drivers/scsi/qla2xxx/qla_inline.h b/drivers/scsi/qla2xxx/qla_inline.h
index 861dc522723c..5501b4c581ec 100644
--- a/drivers/scsi/qla2xxx/qla_inline.h
+++ b/drivers/scsi/qla2xxx/qla_inline.h
@@ -266,11 +266,41 @@ qla2x00_handle_mbx_completion(struct qla_hw_data *ha, int status)
 }
 
 static inline void
-qla2x00_set_retry_delay_timestamp(fc_port_t *fcport, uint16_t retry_delay)
+qla2x00_set_retry_delay_timestamp(fc_port_t *fcport, uint16_t sts_qual)
 {
-	if (retry_delay)
-		fcport->retry_delay_timestamp = jiffies +
-		    (retry_delay * HZ / 10);
+	u8 scope;
+	u16 qual;
+#define SQ_SCOPE_MASK		0xc000 /* SAM-6 rev5 5.3.2 */
+#define SQ_SCOPE_SHIFT		14
+#define SQ_QUAL_MASK		0x3fff
+
+#define SQ_MAX_WAIT_SEC		60 /* Max I/O hold off time in seconds. */
+#define SQ_MAX_WAIT_TIME	(SQ_MAX_WAIT_SEC * 10) /* in 100ms. */
+
+	if (!sts_qual) /* Common case. */
+		return;
+
+	scope = (sts_qual & SQ_SCOPE_MASK) >> SQ_SCOPE_SHIFT;
+	/* Handle only scope 1 or 2, which is for I-T nexus. */
+	if (scope != 1 && scope != 2)
+		return;
+
+	/* Skip processing, if retry delay timer is already in effect. */
+	if (fcport->retry_delay_timestamp &&
+	    time_before(jiffies, fcport->retry_delay_timestamp))
+		return;
+
+	qual = sts_qual & SQ_QUAL_MASK;
+	if (qual < 1 || qual > 0x3fef)
+		return;
+	qual = min(qual, (u16)SQ_MAX_WAIT_TIME);
+
+	/* qual is expressed in 100ms increments. */
+	fcport->retry_delay_timestamp = jiffies + (qual * HZ / 10);
+
+	ql_log(ql_log_warn, fcport->vha, 0x5101,
+	       "%8phC: I/O throttling requested (status qualifier = %04xh), holding off I/Os for %ums.\n",
+	       fcport->port_name, sts_qual, qual * 100);
 }
 
 static inline bool
diff --git a/drivers/scsi/qla2xxx/qla_isr.c b/drivers/scsi/qla2xxx/qla_isr.c
index ab5275dbc338..d38dd6520b53 100644
--- a/drivers/scsi/qla2xxx/qla_isr.c
+++ b/drivers/scsi/qla2xxx/qla_isr.c
@@ -2855,7 +2855,7 @@ qla2x00_status_entry(scsi_qla_host_t *vha, struct rsp_que *rsp, void *pkt)
 	int logit = 1;
 	int res = 0;
 	uint16_t state_flags = 0;
-	uint16_t retry_delay = 0;
+	uint16_t sts_qual = 0;
 
 	if (IS_FWI2_CAPABLE(ha)) {
 		comp_status = le16_to_cpu(sts24->comp_status);
@@ -2953,8 +2953,6 @@ qla2x00_status_entry(scsi_qla_host_t *vha, struct rsp_que *rsp, void *pkt)
 	sense_len = par_sense_len = rsp_info_len = resid_len =
 	    fw_resid_len = 0;
 	if (IS_FWI2_CAPABLE(ha)) {
-		u16 sts24_retry_delay = le16_to_cpu(sts24->retry_delay);
-
 		if (scsi_status & SS_SENSE_LEN_VALID)
 			sense_len = le32_to_cpu(sts24->sense_len);
 		if (scsi_status & SS_RESPONSE_INFO_LEN_VALID)
@@ -2968,13 +2966,7 @@ qla2x00_status_entry(scsi_qla_host_t *vha, struct rsp_que *rsp, void *pkt)
 		host_to_fcp_swap(sts24->data, sizeof(sts24->data));
 		ox_id = le16_to_cpu(sts24->ox_id);
 		par_sense_len = sizeof(sts24->data);
-		/* Valid values of the retry delay timer are 0x1-0xffef */
-		if (sts24_retry_delay > 0 && sts24_retry_delay < 0xfff1) {
-			retry_delay = sts24_retry_delay & 0x3fff;
-			ql_dbg(ql_dbg_io, sp->vha, 0x3033,
-			    "%s: scope=%#x retry_delay=%#x\n", __func__,
-			    sts24_retry_delay >> 14, retry_delay);
-		}
+		sts_qual = le16_to_cpu(sts24->status_qualifier);
 	} else {
 		if (scsi_status & SS_SENSE_LEN_VALID)
 			sense_len = le16_to_cpu(sts->req_sense_length);
@@ -3012,9 +3004,9 @@ qla2x00_status_entry(scsi_qla_host_t *vha, struct rsp_que *rsp, void *pkt)
 	 * Check retry_delay_timer value if we receive a busy or
 	 * queue full.
 	 */
-	if (lscsi_status == SAM_STAT_TASK_SET_FULL ||
-	    lscsi_status == SAM_STAT_BUSY)
-		qla2x00_set_retry_delay_timestamp(fcport, retry_delay);
+	if (unlikely(lscsi_status == SAM_STAT_TASK_SET_FULL ||
+		     lscsi_status == SAM_STAT_BUSY))
+		qla2x00_set_retry_delay_timestamp(fcport, sts_qual);
 
 	/*
 	 * Based on Host and scsi status generate status code for Linux
-- 
2.19.0.rc0

[PATCH 05/12] qla2xxx: Reduce duplicate code in reporting speed

[Nilesh Javali]

From: Quinn Tran <qutran@marvell.com>

Indicate correct speed for 16G Mezz card.

Signed-off-by: Quinn Tran <qutran@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
---
 drivers/scsi/qla2xxx/qla_attr.c | 41 +-------------
 drivers/scsi/qla2xxx/qla_gbl.h  |  2 +
 drivers/scsi/qla2xxx/qla_gs.c   |  7 +--
 drivers/scsi/qla2xxx/qla_os.c   | 96 +--------------------------------
 4 files changed, 9 insertions(+), 137 deletions(-)

diff --git a/drivers/scsi/qla2xxx/qla_attr.c b/drivers/scsi/qla2xxx/qla_attr.c
index 5d93ccc73153..d006ae193677 100644
--- a/drivers/scsi/qla2xxx/qla_attr.c
+++ b/drivers/scsi/qla2xxx/qla_attr.c
@@ -3214,46 +3214,7 @@ qla2x00_init_host_attr(scsi_qla_host_t *vha)
 	fc_host_max_npiv_vports(vha->host) = ha->max_npiv_vports;
 	fc_host_npiv_vports_inuse(vha->host) = ha->cur_vport_count;
 
-	if (IS_CNA_CAPABLE(ha))
-		speeds = FC_PORTSPEED_10GBIT;
-	else if (IS_QLA28XX(ha) || IS_QLA27XX(ha)) {
-		if (ha->max_supported_speed == 2) {
-			if (ha->min_supported_speed <= 6)
-				speeds |= FC_PORTSPEED_64GBIT;
-		}
-		if (ha->max_supported_speed == 2 ||
-		    ha->max_supported_speed == 1) {
-			if (ha->min_supported_speed <= 5)
-				speeds |= FC_PORTSPEED_32GBIT;
-		}
-		if (ha->max_supported_speed == 2 ||
-		    ha->max_supported_speed == 1 ||
-		    ha->max_supported_speed == 0) {
-			if (ha->min_supported_speed <= 4)
-				speeds |= FC_PORTSPEED_16GBIT;
-		}
-		if (ha->max_supported_speed == 1 ||
-		    ha->max_supported_speed == 0) {
-			if (ha->min_supported_speed <= 3)
-				speeds |= FC_PORTSPEED_8GBIT;
-		}
-		if (ha->max_supported_speed == 0) {
-			if (ha->min_supported_speed <= 2)
-				speeds |= FC_PORTSPEED_4GBIT;
-		}
-	} else if (IS_QLA2031(ha))
-		speeds = FC_PORTSPEED_16GBIT|FC_PORTSPEED_8GBIT|
-			FC_PORTSPEED_4GBIT;
-	else if (IS_QLA25XX(ha) || IS_QLAFX00(ha))
-		speeds = FC_PORTSPEED_8GBIT|FC_PORTSPEED_4GBIT|
-			FC_PORTSPEED_2GBIT|FC_PORTSPEED_1GBIT;
-	else if (IS_QLA24XX_TYPE(ha))
-		speeds = FC_PORTSPEED_4GBIT|FC_PORTSPEED_2GBIT|
-			FC_PORTSPEED_1GBIT;
-	else if (IS_QLA23XX(ha))
-		speeds = FC_PORTSPEED_2GBIT|FC_PORTSPEED_1GBIT;
-	else
-		speeds = FC_PORTSPEED_1GBIT;
+	speeds = qla25xx_fdmi_port_speed_capability(ha);
 
 	fc_host_supported_speeds(vha->host) = speeds;
 }
diff --git a/drivers/scsi/qla2xxx/qla_gbl.h b/drivers/scsi/qla2xxx/qla_gbl.h
index 36c210c24f72..3360857c4405 100644
--- a/drivers/scsi/qla2xxx/qla_gbl.h
+++ b/drivers/scsi/qla2xxx/qla_gbl.h
@@ -704,6 +704,8 @@ int qla24xx_async_gfpnid(scsi_qla_host_t *, fc_port_t *);
 void qla24xx_handle_gfpnid_event(scsi_qla_host_t *, struct event_arg *);
 void qla24xx_sp_unmap(scsi_qla_host_t *, srb_t *);
 void qla_scan_work_fn(struct work_struct *);
+uint qla25xx_fdmi_port_speed_capability(struct qla_hw_data *);
+uint qla25xx_fdmi_port_speed_currently(struct qla_hw_data *);
 
 /*
  * Global Function Prototypes in qla_attr.c source file.
diff --git a/drivers/scsi/qla2xxx/qla_gs.c b/drivers/scsi/qla2xxx/qla_gs.c
index de9fd7f688d0..700d4247a791 100644
--- a/drivers/scsi/qla2xxx/qla_gs.c
+++ b/drivers/scsi/qla2xxx/qla_gs.c
@@ -1502,7 +1502,7 @@ qla2x00_prep_ct_fdmi_req(struct ct_sns_pkt *p, uint16_t cmd,
 	return &p->p.req;
 }
 
-static uint
+uint
 qla25xx_fdmi_port_speed_capability(struct qla_hw_data *ha)
 {
 	uint speeds = 0;
@@ -1546,7 +1546,7 @@ qla25xx_fdmi_port_speed_capability(struct qla_hw_data *ha)
 		}
 		return speeds;
 	}
-	if (IS_QLA25XX(ha))
+	if (IS_QLA25XX(ha) || IS_QLAFX00(ha))
 		return FDMI_PORT_SPEED_8GB|FDMI_PORT_SPEED_4GB|
 			FDMI_PORT_SPEED_2GB|FDMI_PORT_SPEED_1GB;
 	if (IS_QLA24XX_TYPE(ha))
@@ -1556,7 +1556,8 @@ qla25xx_fdmi_port_speed_capability(struct qla_hw_data *ha)
 		return FDMI_PORT_SPEED_2GB|FDMI_PORT_SPEED_1GB;
 	return FDMI_PORT_SPEED_1GB;
 }
-static uint
+
+uint
 qla25xx_fdmi_port_speed_currently(struct qla_hw_data *ha)
 {
 	switch (ha->link_data_rate) {
diff --git a/drivers/scsi/qla2xxx/qla_os.c b/drivers/scsi/qla2xxx/qla_os.c
index 8da00ba54aec..f10a43ac1ec7 100644
--- a/drivers/scsi/qla2xxx/qla_os.c
+++ b/drivers/scsi/qla2xxx/qla_os.c
@@ -5810,98 +5810,6 @@ qla25xx_rdp_rsp_reduce_size(struct scsi_qla_host *vha,
 	return true;
 }
 
-static uint
-qla25xx_rdp_port_speed_capability(struct qla_hw_data *ha)
-{
-	if (IS_CNA_CAPABLE(ha))
-		return RDP_PORT_SPEED_10GB;
-
-	if (IS_QLA27XX(ha) || IS_QLA28XX(ha)) {
-		unsigned int speeds = 0;
-
-		if (ha->max_supported_speed == 2) {
-			if (ha->min_supported_speed <= 6)
-				speeds |= RDP_PORT_SPEED_64GB;
-		}
-
-		if (ha->max_supported_speed == 2 ||
-		    ha->max_supported_speed == 1) {
-			if (ha->min_supported_speed <= 5)
-				speeds |= RDP_PORT_SPEED_32GB;
-		}
-
-		if (ha->max_supported_speed == 2 ||
-		    ha->max_supported_speed == 1 ||
-		    ha->max_supported_speed == 0) {
-			if (ha->min_supported_speed <= 4)
-				speeds |= RDP_PORT_SPEED_16GB;
-		}
-
-		if (ha->max_supported_speed == 1 ||
-		    ha->max_supported_speed == 0) {
-			if (ha->min_supported_speed <= 3)
-				speeds |= RDP_PORT_SPEED_8GB;
-		}
-
-		if (ha->max_supported_speed == 0) {
-			if (ha->min_supported_speed <= 2)
-				speeds |= RDP_PORT_SPEED_4GB;
-		}
-
-		return speeds;
-	}
-
-	if (IS_QLA2031(ha))
-		return RDP_PORT_SPEED_16GB|RDP_PORT_SPEED_8GB|
-		       RDP_PORT_SPEED_4GB;
-
-	if (IS_QLA25XX(ha))
-		return RDP_PORT_SPEED_8GB|RDP_PORT_SPEED_4GB|
-		       RDP_PORT_SPEED_2GB|RDP_PORT_SPEED_1GB;
-
-	if (IS_QLA24XX_TYPE(ha))
-		return RDP_PORT_SPEED_4GB|RDP_PORT_SPEED_2GB|
-		       RDP_PORT_SPEED_1GB;
-
-	if (IS_QLA23XX(ha))
-		return RDP_PORT_SPEED_2GB|RDP_PORT_SPEED_1GB;
-
-	return RDP_PORT_SPEED_1GB;
-}
-
-static uint
-qla25xx_rdp_port_speed_currently(struct qla_hw_data *ha)
-{
-	switch (ha->link_data_rate) {
-	case PORT_SPEED_1GB:
-		return RDP_PORT_SPEED_1GB;
-
-	case PORT_SPEED_2GB:
-		return RDP_PORT_SPEED_2GB;
-
-	case PORT_SPEED_4GB:
-		return RDP_PORT_SPEED_4GB;
-
-	case PORT_SPEED_8GB:
-		return RDP_PORT_SPEED_8GB;
-
-	case PORT_SPEED_10GB:
-		return RDP_PORT_SPEED_10GB;
-
-	case PORT_SPEED_16GB:
-		return RDP_PORT_SPEED_16GB;
-
-	case PORT_SPEED_32GB:
-		return RDP_PORT_SPEED_32GB;
-
-	case PORT_SPEED_64GB:
-		return RDP_PORT_SPEED_64GB;
-
-	default:
-		return RDP_PORT_SPEED_UNKNOWN;
-	}
-}
-
 /*
  * Function Name: qla24xx_process_purex_iocb
  *
@@ -6068,9 +5976,9 @@ void qla24xx_process_purex_rdp(struct scsi_qla_host *vha,
 	rsp_payload->port_speed_desc.desc_len =
 	    cpu_to_be32(RDP_DESC_LEN(rsp_payload->port_speed_desc));
 	rsp_payload->port_speed_desc.speed_capab = cpu_to_be16(
-	    qla25xx_rdp_port_speed_capability(ha));
+	    qla25xx_fdmi_port_speed_capability(ha));
 	rsp_payload->port_speed_desc.operating_speed = cpu_to_be16(
-	    qla25xx_rdp_port_speed_currently(ha));
+	    qla25xx_fdmi_port_speed_currently(ha));
 
 	/* Link Error Status Descriptor */
 	rsp_payload->ls_err_desc.desc_tag = cpu_to_be32(0x10002);
-- 
2.19.0.rc0

[PATCH 06/12] qla2xxx: Fix memory size truncation

[Nilesh Javali]

From: Quinn Tran <qutran@marvell.com>

Memory size calculation for Extended Login use in hardware
offload was truncated when the setting was set with higher
value.

Signed-off-by: Quinn Tran <qutran@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
---
 drivers/scsi/qla2xxx/qla_def.h | 2 +-
 drivers/scsi/qla2xxx/qla_mbx.c | 7 ++++---
 drivers/scsi/qla2xxx/qla_os.c  | 5 +++--
 3 files changed, 8 insertions(+), 6 deletions(-)

diff --git a/drivers/scsi/qla2xxx/qla_def.h b/drivers/scsi/qla2xxx/qla_def.h
index 074d8753cfc3..ad4a0ba7203c 100644
--- a/drivers/scsi/qla2xxx/qla_def.h
+++ b/drivers/scsi/qla2xxx/qla_def.h
@@ -4216,7 +4216,7 @@ struct qla_hw_data {
 	/* Extended Logins  */
 	void		*exlogin_buf;
 	dma_addr_t	exlogin_buf_dma;
-	int		exlogin_size;
+	uint32_t	exlogin_size;
 
 #define ENABLE_EXCHANGE_OFFLD	BIT_2
 
diff --git a/drivers/scsi/qla2xxx/qla_mbx.c b/drivers/scsi/qla2xxx/qla_mbx.c
index 226f1428d3e5..6ce106096ecc 100644
--- a/drivers/scsi/qla2xxx/qla_mbx.c
+++ b/drivers/scsi/qla2xxx/qla_mbx.c
@@ -845,7 +845,7 @@ qla_get_exlogin_status(scsi_qla_host_t *vha, uint16_t *buf_sz,
  * Context:
  *	Kernel context.
  */
-#define CONFIG_XLOGINS_MEM	0x3
+#define CONFIG_XLOGINS_MEM	0x9
 int
 qla_set_exlogin_mem_cfg(scsi_qla_host_t *vha, dma_addr_t phys_addr)
 {
@@ -872,8 +872,9 @@ qla_set_exlogin_mem_cfg(scsi_qla_host_t *vha, dma_addr_t phys_addr)
 	mcp->flags = 0;
 	rval = qla2x00_mailbox_command(vha, mcp);
 	if (rval != QLA_SUCCESS) {
-		/*EMPTY*/
-		ql_dbg(ql_dbg_mbx, vha, 0x111b, "Failed=%x.\n", rval);
+		ql_dbg(ql_dbg_mbx, vha, 0x111b,
+		       "EXlogin Failed=%x. MB0=%x MB11=%x\n",
+		       rval, mcp->mb[0], mcp->mb[11]);
 	} else {
 		ql_dbg(ql_dbg_mbx + ql_dbg_verbose, vha, 0x118c,
 		    "Done %s.\n", __func__);
diff --git a/drivers/scsi/qla2xxx/qla_os.c b/drivers/scsi/qla2xxx/qla_os.c
index f10a43ac1ec7..858549e21aea 100644
--- a/drivers/scsi/qla2xxx/qla_os.c
+++ b/drivers/scsi/qla2xxx/qla_os.c
@@ -4379,11 +4379,12 @@ int
 qla2x00_set_exlogins_buffer(scsi_qla_host_t *vha)
 {
 	int rval;
-	uint16_t	size, max_cnt, temp;
+	uint16_t	size, max_cnt;
+	uint32_t temp;
 	struct qla_hw_data *ha = vha->hw;
 
 	/* Return if we don't need to alloacate any extended logins */
-	if (!ql2xexlogins)
+	if (ql2xexlogins <= MAX_FIBRE_DEVICES_2400)
 		return QLA_SUCCESS;
 
 	if (!IS_EXLOGIN_OFFLD_CAPABLE(ha))
-- 
2.19.0.rc0

[PATCH 07/12] qla2xxx: performance tweak

[Nilesh Javali]

From: Quinn Tran <qutran@marvell.com>

This patch move statistic fields from vha struct to qpair
to reduce memory thrashing.

Signed-off-by: Quinn Tran <qutran@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
---
 drivers/scsi/qla2xxx/qla_attr.c | 46 ++++++++++++++++++++++++++++-----
 drivers/scsi/qla2xxx/qla_def.h  | 32 +++++++++++++++++------
 drivers/scsi/qla2xxx/qla_init.c |  5 ++--
 drivers/scsi/qla2xxx/qla_iocb.c | 18 +++++++------
 drivers/scsi/qla2xxx/qla_isr.c  |  8 +++---
 drivers/scsi/qla2xxx/qla_mid.c  |  4 +--
 drivers/scsi/qla2xxx/qla_nvme.c |  8 +++---
 drivers/scsi/qla2xxx/qla_os.c   | 10 ++++---
 8 files changed, 92 insertions(+), 39 deletions(-)

diff --git a/drivers/scsi/qla2xxx/qla_attr.c b/drivers/scsi/qla2xxx/qla_attr.c
index d006ae193677..1ee747ba4ecc 100644
--- a/drivers/scsi/qla2xxx/qla_attr.c
+++ b/drivers/scsi/qla2xxx/qla_attr.c
@@ -2726,6 +2726,9 @@ qla2x00_get_fc_host_stats(struct Scsi_Host *shost)
 	struct link_statistics *stats;
 	dma_addr_t stats_dma;
 	struct fc_host_statistics *p = &vha->fc_host_stat;
+	struct qla_qpair *qpair;
+	int i;
+	u64 ib = 0, ob = 0, ir = 0, or = 0;
 
 	memset(p, -1, sizeof(*p));
 
@@ -2762,6 +2765,27 @@ qla2x00_get_fc_host_stats(struct Scsi_Host *shost)
 	if (rval != QLA_SUCCESS)
 		goto done_free;
 
+	/* --- */
+	for (i = 0; i < vha->hw->max_qpairs; i++) {
+		qpair = vha->hw->queue_pair_map[i];
+		if (!qpair)
+			continue;
+		ir += qpair->counters.input_requests;
+		or += qpair->counters.output_requests;
+		ib += qpair->counters.input_bytes;
+		ob += qpair->counters.output_bytes;
+	}
+	ir += ha->base_qpair->counters.input_requests;
+	or += ha->base_qpair->counters.output_requests;
+	ib += ha->base_qpair->counters.input_bytes;
+	ob += ha->base_qpair->counters.output_bytes;
+
+	ir += vha->qla_stats.input_requests;
+	or += vha->qla_stats.output_requests;
+	ib += vha->qla_stats.input_bytes;
+	ob += vha->qla_stats.output_bytes;
+	/* --- */
+
 	p->link_failure_count = le32_to_cpu(stats->link_fail_cnt);
 	p->loss_of_sync_count = le32_to_cpu(stats->loss_sync_cnt);
 	p->loss_of_signal_count = le32_to_cpu(stats->loss_sig_cnt);
@@ -2781,15 +2805,16 @@ qla2x00_get_fc_host_stats(struct Scsi_Host *shost)
 			p->rx_words = le64_to_cpu(stats->fpm_recv_word_cnt);
 			p->tx_words = le64_to_cpu(stats->fpm_xmit_word_cnt);
 		} else {
-			p->rx_words = vha->qla_stats.input_bytes;
-			p->tx_words = vha->qla_stats.output_bytes;
+			p->rx_words = ib >> 2;
+			p->tx_words = ob >> 2;
 		}
 	}
+
 	p->fcp_control_requests = vha->qla_stats.control_requests;
-	p->fcp_input_requests = vha->qla_stats.input_requests;
-	p->fcp_output_requests = vha->qla_stats.output_requests;
-	p->fcp_input_megabytes = vha->qla_stats.input_bytes >> 20;
-	p->fcp_output_megabytes = vha->qla_stats.output_bytes >> 20;
+	p->fcp_input_requests = ir;
+	p->fcp_output_requests = or;
+	p->fcp_input_megabytes  = ib >> 20;
+	p->fcp_output_megabytes = ob >> 20;
 	p->seconds_since_last_reset =
 	    get_jiffies_64() - vha->qla_stats.jiffies_at_last_reset;
 	do_div(p->seconds_since_last_reset, HZ);
@@ -2809,9 +2834,18 @@ qla2x00_reset_host_stats(struct Scsi_Host *shost)
 	struct scsi_qla_host *base_vha = pci_get_drvdata(ha->pdev);
 	struct link_statistics *stats;
 	dma_addr_t stats_dma;
+	int i;
+	struct qla_qpair *qpair;
 
 	memset(&vha->qla_stats, 0, sizeof(vha->qla_stats));
 	memset(&vha->fc_host_stat, 0, sizeof(vha->fc_host_stat));
+	for (i = 0; i < vha->hw->max_qpairs; i++) {
+		qpair = vha->hw->queue_pair_map[i];
+		if (!qpair)
+			continue;
+		memset(&qpair->counters, 0, sizeof(qpair->counters));
+	}
+	memset(&ha->base_qpair->counters, 0, sizeof(qpair->counters));
 
 	vha->qla_stats.jiffies_at_last_reset = get_jiffies_64();
 
diff --git a/drivers/scsi/qla2xxx/qla_def.h b/drivers/scsi/qla2xxx/qla_def.h
index ad4a0ba7203c..3ca8665638c4 100644
--- a/drivers/scsi/qla2xxx/qla_def.h
+++ b/drivers/scsi/qla2xxx/qla_def.h
@@ -2443,12 +2443,6 @@ typedef struct fc_port {
 	struct list_head list;
 	struct scsi_qla_host *vha;
 
-	uint8_t node_name[WWN_SIZE];
-	uint8_t port_name[WWN_SIZE];
-	port_id_t d_id;
-	uint16_t loop_id;
-	uint16_t old_loop_id;
-
 	unsigned int conf_compl_supported:1;
 	unsigned int deleted:2;
 	unsigned int free_pending:1;
@@ -2465,6 +2459,13 @@ typedef struct fc_port {
 	unsigned int n2n_flag:1;
 	unsigned int explicit_logout:1;
 	unsigned int prli_pend_timer:1;
+	uint8_t nvme_flag;
+
+	uint8_t node_name[WWN_SIZE];
+	uint8_t port_name[WWN_SIZE];
+	port_id_t d_id;
+	uint16_t loop_id;
+	uint16_t old_loop_id;
 
 	struct completion nvme_del_done;
 	uint32_t nvme_prli_service_param;
@@ -2473,7 +2474,7 @@ typedef struct fc_port {
 #define NVME_PRLI_SP_TARGET     BIT_4
 #define NVME_PRLI_SP_DISCOVERY  BIT_3
 #define NVME_PRLI_SP_FIRST_BURST	BIT_0
-	uint8_t nvme_flag;
+
 	uint32_t nvme_first_burst_size;
 #define NVME_FLAG_REGISTERED 4
 #define NVME_FLAG_DELETING 2
@@ -3510,6 +3511,14 @@ struct qla_tgt_counters {
 	uint64_t num_term_xchg_sent;
 };
 
+struct qla_counters {
+	uint64_t input_bytes;
+	uint64_t input_requests;
+	uint64_t output_bytes;
+	uint64_t output_requests;
+
+};
+
 struct qla_qpair;
 
 /* Response queue data structure */
@@ -3594,6 +3603,7 @@ struct qla_qpair {
 	uint32_t enable_class_2:1;
 	uint32_t enable_explicit_conf:1;
 	uint32_t use_shadow_reg:1;
+	uint32_t rcv_intr:1;
 
 	uint16_t id;			/* qp number used with FW */
 	uint16_t vp_idx;		/* vport ID */
@@ -3609,13 +3619,16 @@ struct qla_qpair {
 	struct qla_msix_entry *msix; /* point to &ha->msix_entries[x] */
 	struct qla_hw_data *hw;
 	struct work_struct q_work;
+	struct qla_counters counters;
+
 	struct list_head qp_list_elem; /* vha->qp_list */
 	struct list_head hints_list;
-	uint16_t cpuid;
+
 	uint16_t retry_term_cnt;
 	__le32	retry_term_exchg_addr;
 	uint64_t retry_term_jiff;
 	struct qla_tgt_counters tgt_counters;
+	uint16_t cpuid;
 };
 
 /* Place holder for FW buffer parameters */
@@ -4129,6 +4142,9 @@ struct qla_hw_data {
 #define USE_ASYNC_SCAN(ha) (IS_QLA25XX(ha) || IS_QLA81XX(ha) ||\
 	IS_QLA83XX(ha) || IS_QLA27XX(ha) || IS_QLA28XX(ha))
 
+#define IS_ZIO_THRESHOLD_CAPABLE(ha) \
+	(IS_QLA83XX(ha) || IS_QLA27XX(ha) || IS_QLA28XX(ha))
+
 	/* HBA serial number */
 	uint8_t		serial0;
 	uint8_t		serial1;
diff --git a/drivers/scsi/qla2xxx/qla_init.c b/drivers/scsi/qla2xxx/qla_init.c
index b4d53eb4e53e..99f322fb74ab 100644
--- a/drivers/scsi/qla2xxx/qla_init.c
+++ b/drivers/scsi/qla2xxx/qla_init.c
@@ -3690,9 +3690,8 @@ qla2x00_setup_chip(scsi_qla_host_t *vha)
 					goto execute_fw_with_lr;
 				}
 
-				if ((IS_QLA83XX(ha) || IS_QLA27XX(ha) ||
-				    IS_QLA28XX(ha)) &&
-				    (ha->zio_mode == QLA_ZIO_MODE_6))
+				if (IS_ZIO_THRESHOLD_CAPABLE(ha) &&
+				    ha->zio_mode == QLA_ZIO_MODE_6)
 					qla27xx_set_zio_threshold(vha,
 					    ha->last_zio_threshold);
 
diff --git a/drivers/scsi/qla2xxx/qla_iocb.c b/drivers/scsi/qla2xxx/qla_iocb.c
index e3d2dea0b057..d69e16e844aa 100644
--- a/drivers/scsi/qla2xxx/qla_iocb.c
+++ b/drivers/scsi/qla2xxx/qla_iocb.c
@@ -594,6 +594,7 @@ qla24xx_build_scsi_type_6_iocbs(srb_t *sp, struct cmd_type_6 *cmd_pkt,
 	uint32_t dsd_list_len;
 	struct dsd_dma *dsd_ptr;
 	struct ct6_dsd *ctx;
+	struct qla_qpair *qpair = sp->qpair;
 
 	cmd = GET_CMD_SP(sp);
 
@@ -612,12 +613,12 @@ qla24xx_build_scsi_type_6_iocbs(srb_t *sp, struct cmd_type_6 *cmd_pkt,
 	/* Set transfer direction */
 	if (cmd->sc_data_direction == DMA_TO_DEVICE) {
 		cmd_pkt->control_flags = cpu_to_le16(CF_WRITE_DATA);
-		vha->qla_stats.output_bytes += scsi_bufflen(cmd);
-		vha->qla_stats.output_requests++;
+		qpair->counters.output_bytes += scsi_bufflen(cmd);
+		qpair->counters.output_requests++;
 	} else if (cmd->sc_data_direction == DMA_FROM_DEVICE) {
 		cmd_pkt->control_flags = cpu_to_le16(CF_READ_DATA);
-		vha->qla_stats.input_bytes += scsi_bufflen(cmd);
-		vha->qla_stats.input_requests++;
+		qpair->counters.input_bytes += scsi_bufflen(cmd);
+		qpair->counters.input_requests++;
 	}
 
 	cur_seg = scsi_sglist(cmd);
@@ -704,6 +705,7 @@ qla24xx_build_scsi_iocbs(srb_t *sp, struct cmd_type_7 *cmd_pkt,
 	struct scsi_cmnd *cmd;
 	struct scatterlist *sg;
 	int i;
+	struct qla_qpair *qpair = sp->qpair;
 
 	cmd = GET_CMD_SP(sp);
 
@@ -721,12 +723,12 @@ qla24xx_build_scsi_iocbs(srb_t *sp, struct cmd_type_7 *cmd_pkt,
 	/* Set transfer direction */
 	if (cmd->sc_data_direction == DMA_TO_DEVICE) {
 		cmd_pkt->task_mgmt_flags = cpu_to_le16(TMF_WRITE_DATA);
-		vha->qla_stats.output_bytes += scsi_bufflen(cmd);
-		vha->qla_stats.output_requests++;
+		qpair->counters.output_bytes += scsi_bufflen(cmd);
+		qpair->counters.output_requests++;
 	} else if (cmd->sc_data_direction == DMA_FROM_DEVICE) {
 		cmd_pkt->task_mgmt_flags = cpu_to_le16(TMF_READ_DATA);
-		vha->qla_stats.input_bytes += scsi_bufflen(cmd);
-		vha->qla_stats.input_requests++;
+		qpair->counters.input_bytes += scsi_bufflen(cmd);
+		qpair->counters.input_requests++;
 	}
 
 	/* One DSD is available in the Command Type 3 IOCB */
diff --git a/drivers/scsi/qla2xxx/qla_isr.c b/drivers/scsi/qla2xxx/qla_isr.c
index d38dd6520b53..a63f2000fadf 100644
--- a/drivers/scsi/qla2xxx/qla_isr.c
+++ b/drivers/scsi/qla2xxx/qla_isr.c
@@ -3414,8 +3414,10 @@ void qla24xx_process_response_queue(struct scsi_qla_host *vha,
 	if (!ha->flags.fw_started)
 		return;
 
-	if (rsp->qpair->cpuid != smp_processor_id())
+	if (rsp->qpair->cpuid != smp_processor_id() || !rsp->qpair->rcv_intr) {
+		rsp->qpair->rcv_intr = 1;
 		qla_cpu_update(rsp->qpair, smp_processor_id());
+	}
 
 	while (rsp->ring_ptr->signature != RESPONSE_PROCESSED) {
 		pkt = (struct sts_entry_24xx *)rsp->ring_ptr;
@@ -3865,7 +3867,7 @@ qla2xxx_msix_rsp_q(int irq, void *dev_id)
 	}
 	ha = qpair->hw;
 
-	queue_work(ha->wq, &qpair->q_work);
+	queue_work_on(smp_processor_id(), ha->wq, &qpair->q_work);
 
 	return IRQ_HANDLED;
 }
@@ -3891,7 +3893,7 @@ qla2xxx_msix_rsp_q_hs(int irq, void *dev_id)
 	wrt_reg_dword(&reg->hccr, HCCRX_CLR_RISC_INT);
 	spin_unlock_irqrestore(&ha->hardware_lock, flags);
 
-	queue_work(ha->wq, &qpair->q_work);
+	queue_work_on(smp_processor_id(), ha->wq, &qpair->q_work);
 
 	return IRQ_HANDLED;
 }
diff --git a/drivers/scsi/qla2xxx/qla_mid.c b/drivers/scsi/qla2xxx/qla_mid.c
index 15efe2f04b86..08cfe043ac66 100644
--- a/drivers/scsi/qla2xxx/qla_mid.c
+++ b/drivers/scsi/qla2xxx/qla_mid.c
@@ -808,11 +808,9 @@ static void qla_do_work(struct work_struct *work)
 {
 	unsigned long flags;
 	struct qla_qpair *qpair = container_of(work, struct qla_qpair, q_work);
-	struct scsi_qla_host *vha;
-	struct qla_hw_data *ha = qpair->hw;
+	struct scsi_qla_host *vha = qpair->vha;
 
 	spin_lock_irqsave(&qpair->qp_lock, flags);
-	vha = pci_get_drvdata(ha->pdev);
 	qla24xx_process_response_queue(vha, qpair->rsp);
 	spin_unlock_irqrestore(&qpair->qp_lock, flags);
 
diff --git a/drivers/scsi/qla2xxx/qla_nvme.c b/drivers/scsi/qla2xxx/qla_nvme.c
index 435f17f87301..1c742875817f 100644
--- a/drivers/scsi/qla2xxx/qla_nvme.c
+++ b/drivers/scsi/qla2xxx/qla_nvme.c
@@ -428,8 +428,8 @@ static inline int qla2x00_start_nvme_mq(srb_t *sp)
 	/* No data transfer how do we check buffer len == 0?? */
 	if (fd->io_dir == NVMEFC_FCP_READ) {
 		cmd_pkt->control_flags = cpu_to_le16(CF_READ_DATA);
-		vha->qla_stats.input_bytes += fd->payload_length;
-		vha->qla_stats.input_requests++;
+		qpair->counters.input_bytes += fd->payload_length;
+		qpair->counters.input_requests++;
 	} else if (fd->io_dir == NVMEFC_FCP_WRITE) {
 		cmd_pkt->control_flags = cpu_to_le16(CF_WRITE_DATA);
 		if ((vha->flags.nvme_first_burst) &&
@@ -441,8 +441,8 @@ static inline int qla2x00_start_nvme_mq(srb_t *sp)
 				cmd_pkt->control_flags |=
 					cpu_to_le16(CF_NVME_FIRST_BURST_ENABLE);
 		}
-		vha->qla_stats.output_bytes += fd->payload_length;
-		vha->qla_stats.output_requests++;
+		qpair->counters.output_bytes += fd->payload_length;
+		qpair->counters.output_requests++;
 	} else if (fd->io_dir == 0) {
 		cmd_pkt->control_flags = 0;
 	}
diff --git a/drivers/scsi/qla2xxx/qla_os.c b/drivers/scsi/qla2xxx/qla_os.c
index 858549e21aea..077f50ca4ef1 100644
--- a/drivers/scsi/qla2xxx/qla_os.c
+++ b/drivers/scsi/qla2xxx/qla_os.c
@@ -7198,8 +7198,10 @@ qla2x00_timer(struct timer_list *t)
 	 * FC-NVME
 	 * see if the active AEN count has changed from what was last reported.
 	 */
+	index = atomic_read(&ha->nvme_active_aen_cnt);
 	if (!vha->vp_idx &&
-	    (atomic_read(&ha->nvme_active_aen_cnt) != ha->nvme_last_rptd_aen) &&
+	    (index != ha->nvme_last_rptd_aen) &&
+	    (index >= DEFAULT_ZIO_THRESHOLD) &&
 	    ha->zio_mode == QLA_ZIO_MODE_6 &&
 	    !ha->flags.host_shutting_down) {
 		ql_log(ql_log_info, vha, 0x3002,
@@ -7211,9 +7213,9 @@ qla2x00_timer(struct timer_list *t)
 	}
 
 	if (!vha->vp_idx &&
-	    (atomic_read(&ha->zio_threshold) != ha->last_zio_threshold) &&
-	    (ha->zio_mode == QLA_ZIO_MODE_6) &&
-	    (IS_QLA83XX(ha) || IS_QLA27XX(ha) || IS_QLA28XX(ha))) {
+	    atomic_read(&ha->zio_threshold) != ha->last_zio_threshold &&
+	    ha->zio_mode == QLA_ZIO_MODE_6 &&
+	    IS_ZIO_THRESHOLD_CAPABLE(ha)) {
 		ql_log(ql_log_info, vha, 0x3002,
 		    "Sched: Set ZIO exchange threshold to %d.\n",
 		    ha->last_zio_threshold);
-- 
2.19.0.rc0

[PATCH 08/12] qla2xxx: Fix I/O errors during LIP reset tests

[Nilesh Javali]

From: Arun Easi <aeasi@marvell.com>

In .fcp_io(), returning ENODEV as soon as remote port delete has started
can cause I/O errors. Fix this by returning EBUSY until the remote port
delete finishes.

Signed-off-by: Arun Easi <aeasi@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
---
 drivers/scsi/qla2xxx/qla_nvme.c | 10 ++++++++++
 1 file changed, 10 insertions(+)

diff --git a/drivers/scsi/qla2xxx/qla_nvme.c b/drivers/scsi/qla2xxx/qla_nvme.c
index 1c742875817f..12906ee73cbd 100644
--- a/drivers/scsi/qla2xxx/qla_nvme.c
+++ b/drivers/scsi/qla2xxx/qla_nvme.c
@@ -548,6 +548,16 @@ static int qla_nvme_post_cmd(struct nvme_fc_local_port *lport,
 		return rval;
 
 	vha = fcport->vha;
+
+	if (!(fcport->nvme_flag & NVME_FLAG_REGISTERED))
+		return rval;
+
+	if (test_bit(ABORT_ISP_ACTIVE, &vha->dpc_flags))
+		return -EBUSY;
+
+	if ((qpair && !qpair->fw_started) || fcport->deleted)
+		return -EBUSY;
+
 	/*
 	 * If we know the dev is going away while the transport is still sending
 	 * IO's return busy back to stall the IO Q.  This happens when the
-- 
2.19.0.rc0

[PATCH 09/12] qla2xxx: Make tgt_port_database available in initiator mode

[Nilesh Javali]

From: Arun Easi <aeasi@marvell.com>

tgt_port_database data is today exported only in target mode, allow it
to be shown in initiator mode, as well.

Signed-off-by: Arun Easi <aeasi@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
---
 drivers/scsi/qla2xxx/qla_dfs.c | 64 +++++++++++++++++-----------------
 1 file changed, 32 insertions(+), 32 deletions(-)

diff --git a/drivers/scsi/qla2xxx/qla_dfs.c b/drivers/scsi/qla2xxx/qla_dfs.c
index 7482e6bf7f7f..610440c647a6 100644
--- a/drivers/scsi/qla2xxx/qla_dfs.c
+++ b/drivers/scsi/qla2xxx/qla_dfs.c
@@ -134,51 +134,51 @@ qla2x00_dfs_tgt_port_database_show(struct seq_file *s, void *unused)
 {
 	scsi_qla_host_t *vha = s->private;
 	struct qla_hw_data *ha = vha->hw;
-	struct gid_list_info *gid_list, *gid;
+	struct gid_list_info *gid_list;
 	dma_addr_t gid_list_dma;
 	fc_port_t fc_port;
+	char *id_iter;
 	int rc, i;
 	uint16_t entries, loop_id;
-	struct qla_tgt *tgt = vha->vha_tgt.qla_tgt;
 
 	seq_printf(s, "%s\n", vha->host_str);
-	if (tgt) {
-		gid_list = dma_alloc_coherent(&ha->pdev->dev,
-		    qla2x00_gid_list_size(ha),
-		    &gid_list_dma, GFP_KERNEL);
-		if (!gid_list) {
-			ql_dbg(ql_dbg_user, vha, 0x7018,
-			    "DMA allocation failed for %u\n",
-			     qla2x00_gid_list_size(ha));
-			return 0;
-		}
+	gid_list = dma_alloc_coherent(&ha->pdev->dev,
+				      qla2x00_gid_list_size(ha),
+				      &gid_list_dma, GFP_KERNEL);
+	if (!gid_list) {
+		ql_dbg(ql_dbg_user, vha, 0x7018,
+		       "DMA allocation failed for %u\n",
+		       qla2x00_gid_list_size(ha));
+		return 0;
+	}
 
-		rc = qla24xx_gidlist_wait(vha, gid_list, gid_list_dma,
-		    &entries);
-		if (rc != QLA_SUCCESS)
-			goto out_free_id_list;
+	rc = qla24xx_gidlist_wait(vha, gid_list, gid_list_dma,
+				  &entries);
+	if (rc != QLA_SUCCESS)
+		goto out_free_id_list;
 
-		gid = gid_list;
+	id_iter = (char *)gid_list;
 
-		seq_puts(s, "Port Name	Port ID 	Loop ID\n");
+	seq_puts(s, "Port Name	Port ID		Loop ID\n");
 
-		for (i = 0; i < entries; i++) {
-			loop_id = le16_to_cpu(gid->loop_id);
-			memset(&fc_port, 0, sizeof(fc_port_t));
+	for (i = 0; i < entries; i++) {
+		struct gid_list_info *gid =
+			(struct gid_list_info *)id_iter;
+		loop_id = le16_to_cpu(gid->loop_id);
+		memset(&fc_port, 0, sizeof(fc_port_t));
 
-			fc_port.loop_id = loop_id;
+		fc_port.loop_id = loop_id;
 
-			rc = qla24xx_gpdb_wait(vha, &fc_port, 0);
-			seq_printf(s, "%8phC  %02x%02x%02x  %d\n",
-				fc_port.port_name, fc_port.d_id.b.domain,
-				fc_port.d_id.b.area, fc_port.d_id.b.al_pa,
-				fc_port.loop_id);
-			gid = (void *)gid + ha->gid_list_info_size;
-		}
-out_free_id_list:
-		dma_free_coherent(&ha->pdev->dev, qla2x00_gid_list_size(ha),
-		    gid_list, gid_list_dma);
+		rc = qla24xx_gpdb_wait(vha, &fc_port, 0);
+		seq_printf(s, "%8phC  %02x%02x%02x  %d\n",
+			   fc_port.port_name, fc_port.d_id.b.domain,
+			   fc_port.d_id.b.area, fc_port.d_id.b.al_pa,
+			   fc_port.loop_id);
+		id_iter += ha->gid_list_info_size;
 	}
+out_free_id_list:
+	dma_free_coherent(&ha->pdev->dev, qla2x00_gid_list_size(ha),
+			  gid_list, gid_list_dma);
 
 	return 0;
 }
-- 
2.19.0.rc0

[PATCH 10/12] qla2xxx: Add rport fields in debugfs

[Nilesh Javali]

From: Arun Easi <aeasi@marvell.com>

This patch adds rport fields in debugfs.

Signed-off-by: Arun Easi <aeasi@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
---
 drivers/scsi/qla2xxx/qla_dfs.c | 53 ++++++++++++++++++++++++++++++++++
 1 file changed, 53 insertions(+)

diff --git a/drivers/scsi/qla2xxx/qla_dfs.c b/drivers/scsi/qla2xxx/qla_dfs.c
index 610440c647a6..b9b3a3dd93db 100644
--- a/drivers/scsi/qla2xxx/qla_dfs.c
+++ b/drivers/scsi/qla2xxx/qla_dfs.c
@@ -61,13 +61,53 @@ DEFINE_DEBUGFS_ATTRIBUTE(qla_dfs_rport_##_attr##_fops,		\
 		qla_dfs_rport_##_attr##_get,			\
 		qla_dfs_rport_##_attr##_set, "%llu\n")
 
+/*
+ * Wrapper for getting fc_port fields.
+ *
+ * _attr    : Attribute name.
+ * _get_val : Accessor macro to retrieve the value.
+ */
+#define DEFINE_QLA_DFS_RPORT_FIELD_GET(_attr, _get_val)			\
+static int qla_dfs_rport_field_##_attr##_get(void *data, u64 *val)	\
+{									\
+	struct fc_port *fp = data;					\
+	*val = _get_val;						\
+	return 0;							\
+}									\
+DEFINE_DEBUGFS_ATTRIBUTE(qla_dfs_rport_field_##_attr##_fops,		\
+		qla_dfs_rport_field_##_attr##_get,			\
+		NULL, "%llu\n")
+
+#define DEFINE_QLA_DFS_RPORT_ACCESS(_attr, _get_val) \
+	DEFINE_QLA_DFS_RPORT_FIELD_GET(_attr, _get_val)
+
+#define DEFINE_QLA_DFS_RPORT_FIELD(_attr) \
+	DEFINE_QLA_DFS_RPORT_FIELD_GET(_attr, fp->_attr)
+
 DEFINE_QLA_DFS_RPORT_RW_ATTR(QLA_DFS_RPORT_DEVLOSS_TMO, dev_loss_tmo);
 
+DEFINE_QLA_DFS_RPORT_FIELD(disc_state);
+DEFINE_QLA_DFS_RPORT_FIELD(scan_state);
+DEFINE_QLA_DFS_RPORT_FIELD(fw_login_state);
+DEFINE_QLA_DFS_RPORT_FIELD(login_pause);
+DEFINE_QLA_DFS_RPORT_FIELD(flags);
+DEFINE_QLA_DFS_RPORT_FIELD(nvme_flag);
+DEFINE_QLA_DFS_RPORT_FIELD(last_rscn_gen);
+DEFINE_QLA_DFS_RPORT_FIELD(rscn_gen);
+DEFINE_QLA_DFS_RPORT_FIELD(login_gen);
+DEFINE_QLA_DFS_RPORT_FIELD(loop_id);
+DEFINE_QLA_DFS_RPORT_FIELD_GET(port_id, fp->d_id.b24);
+DEFINE_QLA_DFS_RPORT_FIELD_GET(sess_kref, kref_read(&fp->sess_kref));
+
 void
 qla2x00_dfs_create_rport(scsi_qla_host_t *vha, struct fc_port *fp)
 {
 	char wwn[32];
 
+#define QLA_CREATE_RPORT_FIELD_ATTR(_attr)			\
+	debugfs_create_file(#_attr, 0400, fp->dfs_rport_dir,	\
+		fp, &qla_dfs_rport_field_##_attr##_fops)
+
 	if (!vha->dfs_rport_root || fp->dfs_rport_dir)
 		return;
 
@@ -78,6 +118,19 @@ qla2x00_dfs_create_rport(scsi_qla_host_t *vha, struct fc_port *fp)
 	if (NVME_TARGET(vha->hw, fp))
 		debugfs_create_file("dev_loss_tmo", 0600, fp->dfs_rport_dir,
 				    fp, &qla_dfs_rport_dev_loss_tmo_fops);
+
+	QLA_CREATE_RPORT_FIELD_ATTR(disc_state);
+	QLA_CREATE_RPORT_FIELD_ATTR(scan_state);
+	QLA_CREATE_RPORT_FIELD_ATTR(fw_login_state);
+	QLA_CREATE_RPORT_FIELD_ATTR(login_pause);
+	QLA_CREATE_RPORT_FIELD_ATTR(flags);
+	QLA_CREATE_RPORT_FIELD_ATTR(nvme_flag);
+	QLA_CREATE_RPORT_FIELD_ATTR(last_rscn_gen);
+	QLA_CREATE_RPORT_FIELD_ATTR(rscn_gen);
+	QLA_CREATE_RPORT_FIELD_ATTR(login_gen);
+	QLA_CREATE_RPORT_FIELD_ATTR(loop_id);
+	QLA_CREATE_RPORT_FIELD_ATTR(port_id);
+	QLA_CREATE_RPORT_FIELD_ATTR(sess_kref);
 }
 
 void
-- 
2.19.0.rc0

[PATCH 11/12] qla2xxx: Add IOCB resource tracking

[Nilesh Javali]

From: Quinn Tran <qutran@marvell.com>

This patch tracks number of IOCB resources used in the IO
fast path. If the number of used IOCBs reach a high water
limit, driver would return the IO as busy and let upper layer
retry. This prevents over subscription of IOCB resources where
any future error recovery command is unable to cut through.
Enable IOCB throttling by default.

Signed-off-by: Quinn Tran <qutran@marvell.com>
Signed-off-by: Arun Easi <aeasi@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
---
 drivers/scsi/qla2xxx/qla_def.h    | 17 ++++++++++
 drivers/scsi/qla2xxx/qla_dfs.c    | 14 ++++++++
 drivers/scsi/qla2xxx/qla_gbl.h    |  3 ++
 drivers/scsi/qla2xxx/qla_init.c   | 26 +++++++++++++++
 drivers/scsi/qla2xxx/qla_inline.h | 55 +++++++++++++++++++++++++++++++
 drivers/scsi/qla2xxx/qla_iocb.c   | 28 ++++++++++++++++
 drivers/scsi/qla2xxx/qla_isr.c    |  2 ++
 drivers/scsi/qla2xxx/qla_os.c     |  6 ++++
 8 files changed, 151 insertions(+)

diff --git a/drivers/scsi/qla2xxx/qla_def.h b/drivers/scsi/qla2xxx/qla_def.h
index 3ca8665638c4..863b9c7766e1 100644
--- a/drivers/scsi/qla2xxx/qla_def.h
+++ b/drivers/scsi/qla2xxx/qla_def.h
@@ -624,6 +624,12 @@ enum {
 	TYPE_TGT_TMCMD,		/* task management */
 };
 
+struct iocb_resource {
+	u8 res_type;
+	u8 pad;
+	u16 iocb_cnt;
+};
+
 typedef struct srb {
 	/*
 	 * Do not move cmd_type field, it needs to
@@ -631,6 +637,7 @@ typedef struct srb {
 	 */
 	uint8_t cmd_type;
 	uint8_t pad[3];
+	struct iocb_resource iores;
 	struct kref cmd_kref;	/* need to migrate ref_count over to this */
 	void *priv;
 	wait_queue_head_t nvme_ls_waitq;
@@ -3577,6 +3584,15 @@ struct req_que {
 	uint8_t req_pkt[REQUEST_ENTRY_SIZE];
 };
 
+struct qla_fw_resources {
+	u16 iocbs_total;
+	u16 iocbs_limit;
+	u16 iocbs_qp_limit;
+	u16 iocbs_used;
+};
+
+#define QLA_IOCB_PCT_LIMIT 95
+
 /*Queue pair data structure */
 struct qla_qpair {
 	spinlock_t qp_lock;
@@ -3629,6 +3645,7 @@ struct qla_qpair {
 	uint64_t retry_term_jiff;
 	struct qla_tgt_counters tgt_counters;
 	uint16_t cpuid;
+	struct qla_fw_resources fwres ____cacheline_aligned;
 };
 
 /* Place holder for FW buffer parameters */
diff --git a/drivers/scsi/qla2xxx/qla_dfs.c b/drivers/scsi/qla2xxx/qla_dfs.c
index b9b3a3dd93db..4a86dca4a445 100644
--- a/drivers/scsi/qla2xxx/qla_dfs.c
+++ b/drivers/scsi/qla2xxx/qla_dfs.c
@@ -257,6 +257,8 @@ qla_dfs_fw_resource_cnt_show(struct seq_file *s, void *unused)
 	struct scsi_qla_host *vha = s->private;
 	uint16_t mb[MAX_IOCB_MB_REG];
 	int rc;
+	struct qla_hw_data *ha = vha->hw;
+	u16 iocbs_used, i;
 
 	rc = qla24xx_res_count_wait(vha, mb, SIZEOF_IOCB_MB_REG);
 	if (rc != QLA_SUCCESS) {
@@ -281,6 +283,18 @@ qla_dfs_fw_resource_cnt_show(struct seq_file *s, void *unused)
 		    mb[23]);
 	}
 
+	if (ql2xenforce_iocb_limit) {
+		/* lock is not require. It's an estimate. */
+		iocbs_used = ha->base_qpair->fwres.iocbs_used;
+		for (i = 0; i < ha->max_qpairs; i++) {
+			if (ha->queue_pair_map[i])
+				iocbs_used += ha->queue_pair_map[i]->fwres.iocbs_used;
+		}
+
+		seq_printf(s, "Driver: estimate iocb used [%d] high water limit [%d]\n",
+			   iocbs_used, ha->base_qpair->fwres.iocbs_limit);
+	}
+
 	return 0;
 }
 
diff --git a/drivers/scsi/qla2xxx/qla_gbl.h b/drivers/scsi/qla2xxx/qla_gbl.h
index 3360857c4405..9c4d077edf9e 100644
--- a/drivers/scsi/qla2xxx/qla_gbl.h
+++ b/drivers/scsi/qla2xxx/qla_gbl.h
@@ -129,6 +129,8 @@ int qla2x00_reserve_mgmt_server_loop_id(scsi_qla_host_t *);
 void qla_rscn_replay(fc_port_t *fcport);
 void qla24xx_free_purex_item(struct purex_item *item);
 extern bool qla24xx_risc_firmware_invalid(uint32_t *);
+void qla_init_iocb_limit(scsi_qla_host_t *);
+
 
 /*
  * Global Data in qla_os.c source file.
@@ -175,6 +177,7 @@ extern int qla2xuseresexchforels;
 extern int ql2xexlogins;
 extern int ql2xdifbundlinginternalbuffers;
 extern int ql2xfulldump_on_mpifail;
+extern int ql2xenforce_iocb_limit;
 
 extern int qla2x00_loop_reset(scsi_qla_host_t *);
 extern void qla2x00_abort_all_cmds(scsi_qla_host_t *, int);
diff --git a/drivers/scsi/qla2xxx/qla_init.c b/drivers/scsi/qla2xxx/qla_init.c
index 99f322fb74ab..a1603bad3ee6 100644
--- a/drivers/scsi/qla2xxx/qla_init.c
+++ b/drivers/scsi/qla2xxx/qla_init.c
@@ -3622,6 +3622,31 @@ qla24xx_detect_sfp(scsi_qla_host_t *vha)
 	return ha->flags.lr_detected;
 }
 
+void qla_init_iocb_limit(scsi_qla_host_t *vha)
+{
+	u16 i, num_qps;
+	u32 limit;
+	struct qla_hw_data *ha = vha->hw;
+
+	num_qps = ha->num_qpairs + 1;
+	limit = (ha->orig_fw_iocb_count * QLA_IOCB_PCT_LIMIT) / 100;
+
+	ha->base_qpair->fwres.iocbs_total = ha->orig_fw_iocb_count;
+	ha->base_qpair->fwres.iocbs_limit = limit;
+	ha->base_qpair->fwres.iocbs_qp_limit = limit / num_qps;
+	ha->base_qpair->fwres.iocbs_used = 0;
+	for (i = 0; i < ha->max_qpairs; i++) {
+		if (ha->queue_pair_map[i])  {
+			ha->queue_pair_map[i]->fwres.iocbs_total =
+				ha->orig_fw_iocb_count;
+			ha->queue_pair_map[i]->fwres.iocbs_limit = limit;
+			ha->queue_pair_map[i]->fwres.iocbs_qp_limit =
+				limit / num_qps;
+			ha->queue_pair_map[i]->fwres.iocbs_used = 0;
+		}
+	}
+}
+
 /**
  * qla2x00_setup_chip() - Load and start RISC firmware.
  * @vha: HA context
@@ -3722,6 +3747,7 @@ qla2x00_setup_chip(scsi_qla_host_t *vha)
 						    MIN_MULTI_ID_FABRIC - 1;
 				}
 				qla2x00_get_resource_cnts(vha);
+				qla_init_iocb_limit(vha);
 
 				/*
 				 * Allocate the array of outstanding commands
diff --git a/drivers/scsi/qla2xxx/qla_inline.h b/drivers/scsi/qla2xxx/qla_inline.h
index 5501b4c581ec..9e9a5d3fb802 100644
--- a/drivers/scsi/qla2xxx/qla_inline.h
+++ b/drivers/scsi/qla2xxx/qla_inline.h
@@ -373,3 +373,58 @@ qla2xxx_get_fc4_priority(struct scsi_qla_host *vha)
 
 	return (data >> 6) & BIT_0 ? FC4_PRIORITY_FCP : FC4_PRIORITY_NVME;
 }
+
+enum {
+	RESOURCE_NONE,
+	RESOURCE_INI,
+};
+
+static inline int
+qla_get_iocbs(struct qla_qpair *qp, struct iocb_resource *iores)
+{
+	u16 iocbs_used, i;
+	struct qla_hw_data *ha = qp->vha->hw;
+
+	if (!ql2xenforce_iocb_limit) {
+		iores->res_type = RESOURCE_NONE;
+		return 0;
+	}
+
+	if ((iores->iocb_cnt + qp->fwres.iocbs_used) < qp->fwres.iocbs_qp_limit) {
+		qp->fwres.iocbs_used += iores->iocb_cnt;
+		return 0;
+	} else {
+		/* no need to acquire qpair lock. It's just rough calculation */
+		iocbs_used = ha->base_qpair->fwres.iocbs_used;
+		for (i = 0; i < ha->max_qpairs; i++) {
+			if (ha->queue_pair_map[i])
+				iocbs_used += ha->queue_pair_map[i]->fwres.iocbs_used;
+		}
+
+		if ((iores->iocb_cnt + iocbs_used) < qp->fwres.iocbs_limit) {
+			qp->fwres.iocbs_used += iores->iocb_cnt;
+			return 0;
+		} else {
+			iores->res_type = RESOURCE_NONE;
+			return -ENOSPC;
+		}
+	}
+}
+
+static inline void
+qla_put_iocbs(struct qla_qpair *qp, struct iocb_resource *iores)
+{
+	switch (iores->res_type) {
+	case RESOURCE_NONE:
+		break;
+	default:
+		if (qp->fwres.iocbs_used >= iores->iocb_cnt) {
+			qp->fwres.iocbs_used -= iores->iocb_cnt;
+		} else {
+			// should not happen
+			qp->fwres.iocbs_used = 0;
+		}
+		break;
+	}
+	iores->res_type = RESOURCE_NONE;
+}
diff --git a/drivers/scsi/qla2xxx/qla_iocb.c b/drivers/scsi/qla2xxx/qla_iocb.c
index d69e16e844aa..b60a332e5846 100644
--- a/drivers/scsi/qla2xxx/qla_iocb.c
+++ b/drivers/scsi/qla2xxx/qla_iocb.c
@@ -1637,6 +1637,12 @@ qla24xx_start_scsi(srb_t *sp)
 
 	tot_dsds = nseg;
 	req_cnt = qla24xx_calc_iocbs(vha, tot_dsds);
+
+	sp->iores.res_type = RESOURCE_INI;
+	sp->iores.iocb_cnt = req_cnt;
+	if (qla_get_iocbs(sp->qpair, &sp->iores))
+		goto queuing_error;
+
 	if (req->cnt < (req_cnt + 2)) {
 		cnt = IS_SHADOW_REG_CAPABLE(ha) ? *req->out_ptr :
 		    rd_reg_dword_relaxed(req->req_q_out);
@@ -1709,6 +1715,7 @@ qla24xx_start_scsi(srb_t *sp)
 	if (tot_dsds)
 		scsi_dma_unmap(cmd);
 
+	qla_put_iocbs(sp->qpair, &sp->iores);
 	spin_unlock_irqrestore(&ha->hardware_lock, flags);
 
 	return QLA_FUNCTION_FAILED;
@@ -1822,6 +1829,12 @@ qla24xx_dif_start_scsi(srb_t *sp)
 	/* Total Data and protection sg segment(s) */
 	tot_prot_dsds = nseg;
 	tot_dsds += nseg;
+
+	sp->iores.res_type = RESOURCE_INI;
+	sp->iores.iocb_cnt = qla24xx_calc_iocbs(vha, tot_dsds);
+	if (qla_get_iocbs(sp->qpair, &sp->iores))
+		goto queuing_error;
+
 	if (req->cnt < (req_cnt + 2)) {
 		cnt = IS_SHADOW_REG_CAPABLE(ha) ? *req->out_ptr :
 		    rd_reg_dword_relaxed(req->req_q_out);
@@ -1896,6 +1909,7 @@ qla24xx_dif_start_scsi(srb_t *sp)
 	}
 	/* Cleanup will be performed by the caller (queuecommand) */
 
+	qla_put_iocbs(sp->qpair, &sp->iores);
 	spin_unlock_irqrestore(&ha->hardware_lock, flags);
 	return QLA_FUNCTION_FAILED;
 }
@@ -1957,6 +1971,12 @@ qla2xxx_start_scsi_mq(srb_t *sp)
 
 	tot_dsds = nseg;
 	req_cnt = qla24xx_calc_iocbs(vha, tot_dsds);
+
+	sp->iores.res_type = RESOURCE_INI;
+	sp->iores.iocb_cnt = req_cnt;
+	if (qla_get_iocbs(sp->qpair, &sp->iores))
+		goto queuing_error;
+
 	if (req->cnt < (req_cnt + 2)) {
 		cnt = IS_SHADOW_REG_CAPABLE(ha) ? *req->out_ptr :
 		    rd_reg_dword_relaxed(req->req_q_out);
@@ -2029,6 +2049,7 @@ qla2xxx_start_scsi_mq(srb_t *sp)
 	if (tot_dsds)
 		scsi_dma_unmap(cmd);
 
+	qla_put_iocbs(sp->qpair, &sp->iores);
 	spin_unlock_irqrestore(&qpair->qp_lock, flags);
 
 	return QLA_FUNCTION_FAILED;
@@ -2157,6 +2178,12 @@ qla2xxx_dif_start_scsi_mq(srb_t *sp)
 	/* Total Data and protection sg segment(s) */
 	tot_prot_dsds = nseg;
 	tot_dsds += nseg;
+
+	sp->iores.res_type = RESOURCE_INI;
+	sp->iores.iocb_cnt = qla24xx_calc_iocbs(vha, tot_dsds);
+	if (qla_get_iocbs(sp->qpair, &sp->iores))
+		goto queuing_error;
+
 	if (req->cnt < (req_cnt + 2)) {
 		cnt = IS_SHADOW_REG_CAPABLE(ha) ? *req->out_ptr :
 		    rd_reg_dword_relaxed(req->req_q_out);
@@ -2234,6 +2261,7 @@ qla2xxx_dif_start_scsi_mq(srb_t *sp)
 	}
 	/* Cleanup will be performed by the caller (queuecommand) */
 
+	qla_put_iocbs(sp->qpair, &sp->iores);
 	spin_unlock_irqrestore(&qpair->qp_lock, flags);
 	return QLA_FUNCTION_FAILED;
 }
diff --git a/drivers/scsi/qla2xxx/qla_isr.c b/drivers/scsi/qla2xxx/qla_isr.c
index a63f2000fadf..bb3beaa77d39 100644
--- a/drivers/scsi/qla2xxx/qla_isr.c
+++ b/drivers/scsi/qla2xxx/qla_isr.c
@@ -2901,6 +2901,7 @@ qla2x00_status_entry(scsi_qla_host_t *vha, struct rsp_que *rsp, void *pkt)
 		}
 		return;
 	}
+	qla_put_iocbs(sp->qpair, &sp->iores);
 
 	if (sp->cmd_type != TYPE_SRB) {
 		req->outstanding_cmds[handle] = NULL;
@@ -3313,6 +3314,7 @@ qla2x00_error_entry(scsi_qla_host_t *vha, struct rsp_que *rsp, sts_entry_t *pkt)
 	default:
 		sp = qla2x00_get_sp_from_handle(vha, func, req, pkt);
 		if (sp) {
+			qla_put_iocbs(sp->qpair, &sp->iores);
 			sp->done(sp, res);
 			return 0;
 		}
diff --git a/drivers/scsi/qla2xxx/qla_os.c b/drivers/scsi/qla2xxx/qla_os.c
index 077f50ca4ef1..b9373a2dcbc1 100644
--- a/drivers/scsi/qla2xxx/qla_os.c
+++ b/drivers/scsi/qla2xxx/qla_os.c
@@ -40,6 +40,11 @@ module_param(ql2xfulldump_on_mpifail, int, S_IRUGO | S_IWUSR);
 MODULE_PARM_DESC(ql2xfulldump_on_mpifail,
 		 "Set this to take full dump on MPI hang.");
 
+int ql2xenforce_iocb_limit = 1;
+module_param(ql2xenforce_iocb_limit, int, S_IRUGO | S_IWUSR);
+MODULE_PARM_DESC(ql2xenforce_iocb_limit,
+		 "Enforce IOCB throttling, to avoid FW congestion. (default: 0)");
+
 /*
  * CT6 CTX allocation cache
  */
@@ -3316,6 +3321,7 @@ qla2x00_probe_one(struct pci_dev *pdev, const struct pci_device_id *id)
 		for (i = 0; i < ha->max_qpairs; i++)
 			qla2xxx_create_qpair(base_vha, 5, 0, startit);
 	}
+	qla_init_iocb_limit(base_vha);
 
 	if (ha->flags.running_gold_fw)
 		goto skip_dpc;
-- 
2.19.0.rc0

[PATCH 12/12] qla2xxx: Add SLER and PI control support

[Nilesh Javali]

From: Saurav Kashyap <skashyap@marvell.com>

- BIT_13 of extended FW attribute informs about NVMe-2 support.
- Set BIT_15 of specical feature control block for enabling SLER in FW.
- Set bit 8 (SLER supported) to 1 for the service parameter information
  when sending NVMe PRLI request.
- Set BIT_14 of special feature control block for enabling
  PI Control in FW.
- Driver should set bit 9 (PI Control supported) to 1 for the service
  parameter information when sending NVMe PRLI request.
- Set BIT_13 for NVMe Async events.

Signed-off-by: Saurav Kashyap <skashyap@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
---
 drivers/scsi/qla2xxx/qla_dbg.c  |  2 +-
 drivers/scsi/qla2xxx/qla_def.h  |  4 ++++
 drivers/scsi/qla2xxx/qla_iocb.c |  8 ++++++++
 drivers/scsi/qla2xxx/qla_mbx.c  | 21 +++++++++++++++++++--
 drivers/scsi/qla2xxx/qla_nvme.c | 16 ++++++++++++++--
 drivers/scsi/qla2xxx/qla_nvme.h |  1 +
 drivers/scsi/qla2xxx/qla_os.c   |  1 +
 7 files changed, 48 insertions(+), 5 deletions(-)

diff --git a/drivers/scsi/qla2xxx/qla_dbg.c b/drivers/scsi/qla2xxx/qla_dbg.c
index 1be811a5d69d..369040250ab9 100644
--- a/drivers/scsi/qla2xxx/qla_dbg.c
+++ b/drivers/scsi/qla2xxx/qla_dbg.c
@@ -16,7 +16,7 @@
  * | Device Discovery             |       0x2134       | 0x210e-0x2116  |
  * |				  | 		       | 0x211a         |
  * |                              |                    | 0x211c-0x2128  |
- * |                              |                    | 0x212a-0x2134  |
+ * |                              |                    | 0x212c-0x2134  |
  * | Queue Command and IO tracing |       0x3074       | 0x300b         |
  * |                              |                    | 0x3027-0x3028  |
  * |                              |                    | 0x303d-0x3041  |
diff --git a/drivers/scsi/qla2xxx/qla_def.h b/drivers/scsi/qla2xxx/qla_def.h
index 863b9c7766e1..19b5c0e3dc99 100644
--- a/drivers/scsi/qla2xxx/qla_def.h
+++ b/drivers/scsi/qla2xxx/qla_def.h
@@ -2476,6 +2476,8 @@ typedef struct fc_port {
 
 	struct completion nvme_del_done;
 	uint32_t nvme_prli_service_param;
+#define NVME_PRLI_SP_PI_CTRL	BIT_9
+#define NVME_PRLI_SP_SLER	BIT_8
 #define NVME_PRLI_SP_CONF       BIT_7
 #define NVME_PRLI_SP_INITIATOR  BIT_5
 #define NVME_PRLI_SP_TARGET     BIT_4
@@ -4308,6 +4310,7 @@ struct qla_hw_data {
 #define FW_ATTR_EXT0_SCM_BROCADE	0x00001000
 	/* Cisco fabric attached */
 #define FW_ATTR_EXT0_SCM_CISCO		0x00002000
+#define FW_ATTR_EXT0_NVME2	BIT_13
 	uint16_t	fw_attributes_ext[2];
 	uint32_t	fw_memory_size;
 	uint32_t	fw_transfer_size;
@@ -4657,6 +4660,7 @@ typedef struct scsi_qla_host {
 		uint32_t	qpairs_rsp_created:1;
 		uint32_t	nvme_enabled:1;
 		uint32_t        nvme_first_burst:1;
+		uint32_t        nvme2_enabled:1;
 	} flags;
 
 	atomic_t	loop_state;
diff --git a/drivers/scsi/qla2xxx/qla_iocb.c b/drivers/scsi/qla2xxx/qla_iocb.c
index b60a332e5846..a46f01a25582 100644
--- a/drivers/scsi/qla2xxx/qla_iocb.c
+++ b/drivers/scsi/qla2xxx/qla_iocb.c
@@ -2378,6 +2378,14 @@ qla24xx_prli_iocb(srb_t *sp, struct logio_entry_24xx *logio)
 		if (sp->vha->flags.nvme_first_burst)
 			logio->io_parameter[0] =
 				cpu_to_le32(NVME_PRLI_SP_FIRST_BURST);
+		if (sp->vha->flags.nvme2_enabled) {
+		/* Set service parameter BIT_8 for SLER support */
+			logio->io_parameter[0] |=
+				cpu_to_le32(NVME_PRLI_SP_SLER);
+		/* Set service parameter BIT_9 for PI control support */
+			logio->io_parameter[0] |=
+				cpu_to_le32(NVME_PRLI_SP_PI_CTRL);
+		}
 	}
 
 	logio->nport_handle = cpu_to_le16(sp->fcport->loop_id);
diff --git a/drivers/scsi/qla2xxx/qla_mbx.c b/drivers/scsi/qla2xxx/qla_mbx.c
index 6ce106096ecc..d427a5bca0d9 100644
--- a/drivers/scsi/qla2xxx/qla_mbx.c
+++ b/drivers/scsi/qla2xxx/qla_mbx.c
@@ -1093,6 +1093,17 @@ qla2x00_get_fw_version(scsi_qla_host_t *vha)
 			    "%s: FC-NVMe is Enabled (0x%x)\n",
 			     __func__, ha->fw_attributes_h);
 		}
+
+		/* BIT_13 of Extended FW Attributes informs about NVMe2 support */
+		if (ha->fw_attributes_ext[0] & FW_ATTR_EXT0_NVME2) {
+			ql_dbg(ql_dbg_mbx + ql_dbg_verbose, vha, 0x1191,
+			       "%s: Firmware supports NVMe2 0x%x\n",
+			      __func__, ha->fw_attributes_ext[0]);
+			ql_log(ql_log_info, vha, 0xd302,
+			       "Firmware supports NVMe2 0x%x\n",
+			       ha->fw_attributes_ext[0]);
+			vha->flags.nvme2_enabled = 1;
+		}
 	}
 
 	if (IS_QLA27XX(ha) || IS_QLA28XX(ha)) {
@@ -1122,12 +1133,18 @@ qla2x00_get_fw_version(scsi_qla_host_t *vha)
 		if (ha->flags.scm_supported_a &&
 		    (ha->fw_attributes_ext[0] & FW_ATTR_EXT0_SCM_SUPPORTED)) {
 			ha->flags.scm_supported_f = 1;
-			memset(ha->sf_init_cb, 0, sizeof(struct init_sf_cb));
 			ha->sf_init_cb->flags |= BIT_13;
 		}
 		ql_log(ql_log_info, vha, 0x11a3, "SCM in FW: %s\n",
 		       (ha->flags.scm_supported_f) ? "Supported" :
 		       "Not Supported");
+
+		if (vha->flags.nvme2_enabled) {
+		/* set BIT_15 of special feature control block for SLER */
+			ha->sf_init_cb->flags |= BIT_15;
+		/* set BIT_14 of special feature control block for PI CTRL*/
+			ha->sf_init_cb->flags |= BIT_14;
+		}
 	}
 
 failed:
@@ -1823,7 +1840,7 @@ qla2x00_init_firmware(scsi_qla_host_t *vha, uint16_t size)
 		mcp->out_mb |= MBX_14|MBX_13|MBX_12|MBX_11|MBX_10;
 	}
 
-	if (ha->flags.scm_supported_f) {
+	if (ha->flags.scm_supported_f || vha->flags.nvme2_enabled) {
 		mcp->mb[1] |= BIT_1;
 		mcp->mb[16] = MSW(ha->sf_init_cb_dma);
 		mcp->mb[17] = LSW(ha->sf_init_cb_dma);
diff --git a/drivers/scsi/qla2xxx/qla_nvme.c b/drivers/scsi/qla2xxx/qla_nvme.c
index 12906ee73cbd..541d3f18b32c 100644
--- a/drivers/scsi/qla2xxx/qla_nvme.c
+++ b/drivers/scsi/qla2xxx/qla_nvme.c
@@ -69,6 +69,14 @@ int qla_nvme_register_remote(struct scsi_qla_host *vha, struct fc_port *fcport)
 		return ret;
 	}
 
+	if (fcport->nvme_prli_service_param & NVME_PRLI_SP_SLER)
+		ql_log(ql_log_info, vha, 0x212a,
+		       "PortID:%06x Supports SLER\n", req.port_id);
+
+	if (fcport->nvme_prli_service_param & NVME_PRLI_SP_PI_CTRL)
+		ql_log(ql_log_info, vha, 0x212b,
+		       "PortID:%06x Supports PI control\n", req.port_id);
+
 	rport = fcport->nvme_remote_port->private;
 	rport->fcport = fcport;
 
@@ -368,6 +376,7 @@ static inline int qla2x00_start_nvme_mq(srb_t *sp)
 	struct srb_iocb *nvme = &sp->u.iocb_cmd;
 	struct scatterlist *sgl, *sg;
 	struct nvmefc_fcp_req *fd = nvme->u.nvme.desc;
+	struct nvme_fc_cmd_iu *cmd = fd->cmdaddr;
 	uint32_t        rval = QLA_SUCCESS;
 
 	/* Setup qpair pointers */
@@ -399,8 +408,6 @@ static inline int qla2x00_start_nvme_mq(srb_t *sp)
 	}
 
 	if (unlikely(!fd->sqid)) {
-		struct nvme_fc_cmd_iu *cmd = fd->cmdaddr;
-
 		if (cmd->sqe.common.opcode == nvme_admin_async_event) {
 			nvme->u.nvme.aen_op = 1;
 			atomic_inc(&ha->nvme_active_aen_cnt);
@@ -446,6 +453,11 @@ static inline int qla2x00_start_nvme_mq(srb_t *sp)
 	} else if (fd->io_dir == 0) {
 		cmd_pkt->control_flags = 0;
 	}
+	/* Set BIT_13 of control flags for Async event */
+	if (vha->flags.nvme2_enabled &&
+	    cmd->sqe.common.opcode == nvme_admin_async_event) {
+		cmd_pkt->control_flags |= cpu_to_le16(CF_ADMIN_ASYNC_EVENT);
+	}
 
 	/* Set NPORT-ID */
 	cmd_pkt->nport_handle = cpu_to_le16(sp->fcport->loop_id);
diff --git a/drivers/scsi/qla2xxx/qla_nvme.h b/drivers/scsi/qla2xxx/qla_nvme.h
index cf45a5b277f1..5d5f115a77c3 100644
--- a/drivers/scsi/qla2xxx/qla_nvme.h
+++ b/drivers/scsi/qla2xxx/qla_nvme.h
@@ -54,6 +54,7 @@ struct cmd_nvme {
 	uint64_t rsvd;
 
 	__le16	control_flags;		/* Control Flags */
+#define CF_ADMIN_ASYNC_EVENT		BIT_13
 #define CF_NVME_FIRST_BURST_ENABLE	BIT_11
 #define CF_DIF_SEG_DESCR_ENABLE         BIT_3
 #define CF_DATA_SEG_DESCR_ENABLE        BIT_2
diff --git a/drivers/scsi/qla2xxx/qla_os.c b/drivers/scsi/qla2xxx/qla_os.c
index b9373a2dcbc1..bbd512917a90 100644
--- a/drivers/scsi/qla2xxx/qla_os.c
+++ b/drivers/scsi/qla2xxx/qla_os.c
@@ -4231,6 +4231,7 @@ qla2x00_mem_alloc(struct qla_hw_data *ha, uint16_t req_len, uint16_t rsp_len,
 						&ha->sf_init_cb_dma);
 		if (!ha->sf_init_cb)
 			goto fail_sf_init_cb;
+		memset(ha->sf_init_cb, 0, sizeof(struct init_sf_cb));
 		ql_dbg_pci(ql_dbg_init, ha->pdev, 0x0199,
 			   "sf_init_cb=%p.\n", ha->sf_init_cb);
 	}
-- 
2.19.0.rc0

Re: [PATCH 03/12] qla2xxx: Allow dev_loss_tmo setting for FC-NVMe devices

[Roman Bolshakov]

On Tue, Aug 18, 2020 at 05:31:54AM -0700, Nilesh Javali wrote:
> From: Arun Easi <aeasi@marvell.com>
> 
> Add a remote port debugfs entry to get/set dev_loss_tmo for NVMe
> devices.
> 
> Signed-off-by: Arun Easi <aeasi@marvell.com>
> Signed-off-by: Himanshu Madhani <hmadhani@marvell.com>
> Signed-off-by: Nilesh Javali <njavali@marvell.com>
> ---
>  drivers/scsi/qla2xxx/qla_dfs.c | 54 ++++++++++++++++++++++++++++++++++
>  1 file changed, 54 insertions(+)
> 
> diff --git a/drivers/scsi/qla2xxx/qla_dfs.c b/drivers/scsi/qla2xxx/qla_dfs.c
> index 3c4b9b549b17..7482e6bf7f7f 100644
> --- a/drivers/scsi/qla2xxx/qla_dfs.c
> +++ b/drivers/scsi/qla2xxx/qla_dfs.c
> @@ -12,6 +12,57 @@
>  static struct dentry *qla2x00_dfs_root;
>  static atomic_t qla2x00_dfs_root_count;
>  
> +#define QLA_DFS_RPORT_DEVLOSS_TMO	1
> +
> +static int
> +qla_dfs_rport_get(struct fc_port *fp, int attr_id, u64 *val)
> +{
> +	switch (attr_id) {
> +	case QLA_DFS_RPORT_DEVLOSS_TMO:
> +		/* Only supported for FC-NVMe devices that are registered. */
> +		if (!(fp->nvme_flag & NVME_FLAG_REGISTERED))
> +			return -EIO;
> +		*val = fp->nvme_remote_port->dev_loss_tmo;
> +		break;
> +	default:
> +		return -EINVAL;
> +	}
> +	return 0;
> +}
> +
> +static int
> +qla_dfs_rport_set(struct fc_port *fp, int attr_id, u64 val)
> +{
> +	switch (attr_id) {
> +	case QLA_DFS_RPORT_DEVLOSS_TMO:
> +		/* Only supported for FC-NVMe devices that are registered. */
> +		if (!(fp->nvme_flag & NVME_FLAG_REGISTERED))
> +			return -EIO;
> +		return nvme_fc_set_remoteport_devloss(fp->nvme_remote_port,
> +						      val);
> +	default:
> +		return -EINVAL;
> +	}
> +	return 0;
> +}
> +
> +#define DEFINE_QLA_DFS_RPORT_RW_ATTR(_attr_id, _attr)		\
> +static int qla_dfs_rport_##_attr##_get(void *data, u64 *val)	\
> +{								\
> +	struct fc_port *fp = data;				\
> +	return qla_dfs_rport_get(fp, _attr_id, val);		\
> +}								\
> +static int qla_dfs_rport_##_attr##_set(void *data, u64 val)	\
> +{								\
> +	struct fc_port *fp = data;				\
> +	return qla_dfs_rport_set(fp, _attr_id, val);		\
> +}								\
> +DEFINE_DEBUGFS_ATTRIBUTE(qla_dfs_rport_##_attr##_fops,		\
> +		qla_dfs_rport_##_attr##_get,			\
> +		qla_dfs_rport_##_attr##_set, "%llu\n")
> +
> +DEFINE_QLA_DFS_RPORT_RW_ATTR(QLA_DFS_RPORT_DEVLOSS_TMO, dev_loss_tmo);
> +
>  void
>  qla2x00_dfs_create_rport(scsi_qla_host_t *vha, struct fc_port *fp)
>  {
> @@ -24,6 +75,9 @@ qla2x00_dfs_create_rport(scsi_qla_host_t *vha, struct fc_port *fp)
>  	fp->dfs_rport_dir = debugfs_create_dir(wwn, vha->dfs_rport_root);
>  	if (!fp->dfs_rport_dir)
>  		return;
> +	if (NVME_TARGET(vha->hw, fp))
> +		debugfs_create_file("dev_loss_tmo", 0600, fp->dfs_rport_dir,
> +				    fp, &qla_dfs_rport_dev_loss_tmo_fops);
>  }
>  
>  void
> -- 
> 2.19.0.rc0
> 

Hi Arun,

I don't think that the setting should be in debugfs. IMO there should be
separate fc_remote_ports entry for FCP and FC-NVMe, one per PRLI for
multi-protocol targets.

That'd be consistent with what exists for FCP and would allow
similar configuration of dev_loss_tmo from multipath. It'd also provide
semantics of per-rport PRLO to allow logout from either of the protocol
but leaving second rport/session online.

Thanks,
Roman

Re: [PATCH 03/12] qla2xxx: Allow dev_loss_tmo setting for FC-NVMe devices

[kernel test robot]

[-- Attachment #1: Type: text/plain, Size: 2169 bytes --]

Hi Nilesh,

Thank you for the patch! Yet something to improve:

[auto build test ERROR on dca93232b361d260413933903cd4bdbd92ebcc7f]

url:    https://github.com/0day-ci/linux/commits/Nilesh-Javali/qla2xxx-misc-features-and-bug-fixes/20200818-203851
base:    dca93232b361d260413933903cd4bdbd92ebcc7f
config: s390-randconfig-r034-20200818 (attached as .config)
compiler: clang version 12.0.0 (https://github.com/llvm/llvm-project 790878f291fa5dc58a1c560cb6cc76fd1bfd1c5a)
reproduce (this is a W=1 build):
        wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross
        chmod +x ~/bin/make.cross
        # install s390 cross compiling tool for clang build
        # apt-get install binutils-s390x-linux-gnu
        # save the attached .config to linux build tree
        COMPILER_INSTALL_PATH=$HOME/0day COMPILER=clang make.cross ARCH=s390 

If you fix the issue, kindly add following tag as appropriate
Reported-by: kernel test robot <lkp@intel.com>

All errors (new ones prefixed by >>):

   s390x-linux-gnu-ld: drivers/scsi/qla2xxx/qla_dfs.o: in function `qla_dfs_rport_set':
>> drivers/scsi/qla2xxx/qla_dfs.c:41: undefined reference to `nvme_fc_set_remoteport_devloss'

# https://github.com/0day-ci/linux/commit/c8724abdb8c3a827186052c58054832ee10bd817
git remote add linux-review https://github.com/0day-ci/linux
git fetch --no-tags linux-review Nilesh-Javali/qla2xxx-misc-features-and-bug-fixes/20200818-203851
git checkout c8724abdb8c3a827186052c58054832ee10bd817
vim +41 drivers/scsi/qla2xxx/qla_dfs.c

    32	
    33	static int
    34	qla_dfs_rport_set(struct fc_port *fp, int attr_id, u64 val)
    35	{
    36		switch (attr_id) {
    37		case QLA_DFS_RPORT_DEVLOSS_TMO:
    38			/* Only supported for FC-NVMe devices that are registered. */
    39			if (!(fp->nvme_flag & NVME_FLAG_REGISTERED))
    40				return -EIO;
  > 41			return nvme_fc_set_remoteport_devloss(fp->nvme_remote_port,
    42							      val);
    43		default:
    44			return -EINVAL;
    45		}
    46		return 0;
    47	}
    48	

---
0-DAY CI Kernel Test Service, Intel Corporation
https://lists.01.org/hyperkitty/list/kbuild-all@lists.01.org

[-- Attachment #2: .config.gz --]
[-- Type: application/gzip, Size: 30070 bytes --]

Re: [PATCH 03/12] qla2xxx: Allow dev_loss_tmo setting for FC-NVMe devices

[kernel test robot]

[-- Attachment #1: Type: text/plain, Size: 886 bytes --]

Hi Nilesh,

Thank you for the patch! Yet something to improve:

[auto build test ERROR on dca93232b361d260413933903cd4bdbd92ebcc7f]

url:    https://github.com/0day-ci/linux/commits/Nilesh-Javali/qla2xxx-misc-features-and-bug-fixes/20200818-203851
base:    dca93232b361d260413933903cd4bdbd92ebcc7f
config: x86_64-kexec (attached as .config)
compiler: gcc-9 (Debian 9.3.0-15) 9.3.0
reproduce (this is a W=1 build):
        # save the attached .config to linux build tree
        make W=1 ARCH=x86_64 

If you fix the issue, kindly add following tag as appropriate
Reported-by: kernel test robot <lkp@intel.com>

All errors (new ones prefixed by >>, old ones prefixed by <<):

>> ERROR: modpost: "nvme_fc_set_remoteport_devloss" [drivers/scsi/qla2xxx/qla2xxx.ko] undefined!

---
0-DAY CI Kernel Test Service, Intel Corporation
https://lists.01.org/hyperkitty/list/kbuild-all@lists.01.org

[-- Attachment #2: .config.gz --]
[-- Type: application/gzip, Size: 28200 bytes --]

Re: [PATCH 03/12] qla2xxx: Allow dev_loss_tmo setting for FC-NVMe devices

[Arun Easi]

+James S.

Hi Roman,

On Tue, 18 Aug 2020, 6:24am, Roman Bolshakov wrote:

> 
> On Tue, Aug 18, 2020 at 05:31:54AM -0700, Nilesh Javali wrote:
> > From: Arun Easi <aeasi@marvell.com>
> > 
> > Add a remote port debugfs entry to get/set dev_loss_tmo for NVMe
> > devices.
> > 
> > Signed-off-by: Arun Easi <aeasi@marvell.com>
> > Signed-off-by: Himanshu Madhani <hmadhani@marvell.com>
> > Signed-off-by: Nilesh Javali <njavali@marvell.com>
> > ---
> >  drivers/scsi/qla2xxx/qla_dfs.c | 54 ++++++++++++++++++++++++++++++++++
> >  1 file changed, 54 insertions(+)
> > 
> > diff --git a/drivers/scsi/qla2xxx/qla_dfs.c b/drivers/scsi/qla2xxx/qla_dfs.c
> > index 3c4b9b549b17..7482e6bf7f7f 100644
> > --- a/drivers/scsi/qla2xxx/qla_dfs.c
> > +++ b/drivers/scsi/qla2xxx/qla_dfs.c
> > @@ -12,6 +12,57 @@
> >  static struct dentry *qla2x00_dfs_root;
> >  static atomic_t qla2x00_dfs_root_count;
> >  
> > +#define QLA_DFS_RPORT_DEVLOSS_TMO	1
> > +
> > +static int
> > +qla_dfs_rport_get(struct fc_port *fp, int attr_id, u64 *val)
> > +{
> > +	switch (attr_id) {
> > +	case QLA_DFS_RPORT_DEVLOSS_TMO:
> > +		/* Only supported for FC-NVMe devices that are registered. */
> > +		if (!(fp->nvme_flag & NVME_FLAG_REGISTERED))
> > +			return -EIO;
> > +		*val = fp->nvme_remote_port->dev_loss_tmo;
> > +		break;
> > +	default:
> > +		return -EINVAL;
> > +	}
> > +	return 0;
> > +}
> > +
> > +static int
> > +qla_dfs_rport_set(struct fc_port *fp, int attr_id, u64 val)
> > +{
> > +	switch (attr_id) {
> > +	case QLA_DFS_RPORT_DEVLOSS_TMO:
> > +		/* Only supported for FC-NVMe devices that are registered. */
> > +		if (!(fp->nvme_flag & NVME_FLAG_REGISTERED))
> > +			return -EIO;
> > +		return nvme_fc_set_remoteport_devloss(fp->nvme_remote_port,
> > +						      val);
> > +	default:
> > +		return -EINVAL;
> > +	}
> > +	return 0;
> > +}
> > +
> > +#define DEFINE_QLA_DFS_RPORT_RW_ATTR(_attr_id, _attr)		\
> > +static int qla_dfs_rport_##_attr##_get(void *data, u64 *val)	\
> > +{								\
> > +	struct fc_port *fp = data;				\
> > +	return qla_dfs_rport_get(fp, _attr_id, val);		\
> > +}								\
> > +static int qla_dfs_rport_##_attr##_set(void *data, u64 val)	\
> > +{								\
> > +	struct fc_port *fp = data;				\
> > +	return qla_dfs_rport_set(fp, _attr_id, val);		\
> > +}								\
> > +DEFINE_DEBUGFS_ATTRIBUTE(qla_dfs_rport_##_attr##_fops,		\
> > +		qla_dfs_rport_##_attr##_get,			\
> > +		qla_dfs_rport_##_attr##_set, "%llu\n")
> > +
> > +DEFINE_QLA_DFS_RPORT_RW_ATTR(QLA_DFS_RPORT_DEVLOSS_TMO, dev_loss_tmo);
> > +
> >  void
> >  qla2x00_dfs_create_rport(scsi_qla_host_t *vha, struct fc_port *fp)
> >  {
> > @@ -24,6 +75,9 @@ qla2x00_dfs_create_rport(scsi_qla_host_t *vha, struct fc_port *fp)
> >  	fp->dfs_rport_dir = debugfs_create_dir(wwn, vha->dfs_rport_root);
> >  	if (!fp->dfs_rport_dir)
> >  		return;
> > +	if (NVME_TARGET(vha->hw, fp))
> > +		debugfs_create_file("dev_loss_tmo", 0600, fp->dfs_rport_dir,
> > +				    fp, &qla_dfs_rport_dev_loss_tmo_fops);
> >  }
> >  
> >  void
> > -- 
> > 2.19.0.rc0
> > 
> 
> Hi Arun,
> 
> I don't think that the setting should be in debugfs. IMO there should be
> separate fc_remote_ports entry for FCP and FC-NVMe, one per PRLI for
> multi-protocol targets.
> 
> That'd be consistent with what exists for FCP and would allow
> similar configuration of dev_loss_tmo from multipath. It'd also provide
> semantics of per-rport PRLO to allow logout from either of the protocol
> but leaving second rport/session online.
> 

Thanks for your comments. This debugfs way is sort of a crutch, but that 
is something a user could use to set for FC-NVME dev_loss_tmo until a 
FC-NVME sysfs node hierarchy for a FC initiator (similar to FCP) comes 
into place.

BTW, the above setting is specific for FC-NVME, and is separate from FCP; 
so PRLO semantics can still be met.

Copying James for his thoughts on the topic (on dev_loss_tmo setting). 
James, would you mind sharing your ideas you'd considered?

Regards,
-Arun