scsi: qla2xxx: Move target stat counters from vha to qpair.

Move counters to qpair to reduce cache miss.

Signed-off-by: Quinn Tran <quinn.tran@cavium.com>
Signed-off-by: Himanshu Madhani <himanshu.madhani@cavium.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

drivers/scsi/qla2xxx/qla_def.h

@@ -3183,6 +3183,18 @@ struct qla_tc_param {
 #define QLA_MAX_VPORTS_QLA24XX	128
 #define QLA_MAX_VPORTS_QLA25XX	256
 
+struct qla_tgt_counters {
+	uint64_t qla_core_sbt_cmd;
+	uint64_t core_qla_que_buf;
+	uint64_t qla_core_ret_ctio;
+	uint64_t core_qla_snd_status;
+	uint64_t qla_core_ret_sta_ctio;
+	uint64_t core_qla_free_cmd;
+	uint64_t num_q_full_sent;
+	uint64_t num_alloc_iocb_failed;
+	uint64_t num_term_xchg_sent;
+};
+
 struct qla_qpair;
 
 /* Response queue data structure */
@@ -3285,6 +3297,7 @@ struct qla_qpair {
 	struct list_head qp_list_elem; /* vha->qp_list */
 	struct list_head hints_list;
 	uint16_t cpuid;
+	struct qla_tgt_counters tgt_counters;
 };
 
 /* Place holder for FW buffer parameters */
@@ -3964,18 +3977,6 @@ struct qla_hw_data {
 	int	allow_cna_fw_dump;
 };
 
-struct qla_tgt_counters {
-	uint64_t qla_core_sbt_cmd;
-	uint64_t core_qla_que_buf;
-	uint64_t qla_core_ret_ctio;
-	uint64_t core_qla_snd_status;
-	uint64_t qla_core_ret_sta_ctio;
-	uint64_t core_qla_free_cmd;
-	uint64_t num_q_full_sent;
-	uint64_t num_alloc_iocb_failed;
-	uint64_t num_term_xchg_sent;
-};
-
 /*
  * Qlogic scsi host structure
  */
@@ -4140,10 +4141,8 @@ typedef struct scsi_qla_host {
 	struct fc_host_statistics fc_host_stat;
 	struct qla_statistics qla_stats;
 	struct bidi_statistics bidi_stats;
-
 	atomic_t	vref_count;
 	struct qla8044_reset_template reset_tmplt;
-	struct qla_tgt_counters tgt_counters;
 	uint16_t	bbcr;
 	struct name_list_extended gnl;
 	/* Count of active session/fcport */

drivers/scsi/qla2xxx/qla_dfs.c

@@ -164,26 +164,56 @@ static int
 qla_dfs_tgt_counters_show(struct seq_file *s, void *unused)
 {
 	struct scsi_qla_host *vha = s->private;
+	struct qla_qpair *qpair = vha->hw->base_qpair;
+	uint64_t qla_core_sbt_cmd, core_qla_que_buf, qla_core_ret_ctio,
+		core_qla_snd_status, qla_core_ret_sta_ctio, core_qla_free_cmd,
+		num_q_full_sent, num_alloc_iocb_failed, num_term_xchg_sent;
+	u16 i;
+
+	qla_core_sbt_cmd = qpair->tgt_counters.qla_core_sbt_cmd;
+	core_qla_que_buf = qpair->tgt_counters.core_qla_que_buf;
+	qla_core_ret_ctio = qpair->tgt_counters.qla_core_ret_ctio;
+	core_qla_snd_status = qpair->tgt_counters.core_qla_snd_status;
+	qla_core_ret_sta_ctio = qpair->tgt_counters.qla_core_ret_sta_ctio;
+	core_qla_free_cmd = qpair->tgt_counters.core_qla_free_cmd;
+	num_q_full_sent = qpair->tgt_counters.num_q_full_sent;
+	num_alloc_iocb_failed = qpair->tgt_counters.num_alloc_iocb_failed;
+	num_term_xchg_sent = qpair->tgt_counters.num_term_xchg_sent;
+
+	for (i = 0; i < vha->hw->max_qpairs; i++) {
+		qpair = vha->hw->queue_pair_map[i];
+		qla_core_sbt_cmd += qpair->tgt_counters.qla_core_sbt_cmd;
+		core_qla_que_buf += qpair->tgt_counters.core_qla_que_buf;
+		qla_core_ret_ctio += qpair->tgt_counters.qla_core_ret_ctio;
+		core_qla_snd_status += qpair->tgt_counters.core_qla_snd_status;
+		qla_core_ret_sta_ctio +=
+		    qpair->tgt_counters.qla_core_ret_sta_ctio;
+		core_qla_free_cmd += qpair->tgt_counters.core_qla_free_cmd;
+		num_q_full_sent += qpair->tgt_counters.num_q_full_sent;
+		num_alloc_iocb_failed +=
+		    qpair->tgt_counters.num_alloc_iocb_failed;
+		num_term_xchg_sent += qpair->tgt_counters.num_term_xchg_sent;
+	}
 
 	seq_puts(s, "Target Counters\n");
 	seq_printf(s, "qla_core_sbt_cmd = %lld\n",
-		vha->tgt_counters.qla_core_sbt_cmd);
+		qla_core_sbt_cmd);
 	seq_printf(s, "qla_core_ret_sta_ctio = %lld\n",
-		vha->tgt_counters.qla_core_ret_sta_ctio);
+		qla_core_ret_sta_ctio);
 	seq_printf(s, "qla_core_ret_ctio = %lld\n",
-		vha->tgt_counters.qla_core_ret_ctio);
+		qla_core_ret_ctio);
 	seq_printf(s, "core_qla_que_buf = %lld\n",
-		vha->tgt_counters.core_qla_que_buf);
+		core_qla_que_buf);
 	seq_printf(s, "core_qla_snd_status = %lld\n",
-		vha->tgt_counters.core_qla_snd_status);
+		core_qla_snd_status);
 	seq_printf(s, "core_qla_free_cmd = %lld\n",
-		vha->tgt_counters.core_qla_free_cmd);
+		core_qla_free_cmd);
 	seq_printf(s, "num alloc iocb failed = %lld\n",
-		vha->tgt_counters.num_alloc_iocb_failed);
+		num_alloc_iocb_failed);
 	seq_printf(s, "num term exchange sent = %lld\n",
-		vha->tgt_counters.num_term_xchg_sent);
+		num_term_xchg_sent);
 	seq_printf(s, "num Q full sent = %lld\n",
-		vha->tgt_counters.num_q_full_sent);
+		num_q_full_sent);
 
 	/* DIF stats */
 	seq_printf(s, "DIF Inp Bytes = %lld\n",

drivers/scsi/qla2xxx/qla_iocb.c

@@ -2189,7 +2189,7 @@ skip_cmd_array:
 	}
 
 queuing_error:
-	vha->tgt_counters.num_alloc_iocb_failed++;
+	qpair->tgt_counters.num_alloc_iocb_failed++;
 	return pkt;
 }
 

drivers/scsi/qla2xxx/qla_target.c

@@ -3078,9 +3078,9 @@ int qlt_xmit_response(struct qla_tgt_cmd *cmd, int xmit_type,
 	spin_lock_irqsave(qpair->qp_lock_ptr, flags);
 
 	if (xmit_type == QLA_TGT_XMIT_STATUS)
-		vha->tgt_counters.core_qla_snd_status++;
+		qpair->tgt_counters.core_qla_snd_status++;
 	else
-		vha->tgt_counters.core_qla_que_buf++;
+		qpair->tgt_counters.core_qla_que_buf++;
 
 	if (!qpair->fw_started || cmd->reset_count != qpair->chip_reset) {
 		/*
@@ -3500,7 +3500,7 @@ static int __qlt_send_term_exchange(struct qla_qpair *qpair,
 			ret = 1;
 	}
 
-	vha->tgt_counters.num_term_xchg_sent++;
+	qpair->tgt_counters.num_term_xchg_sent++;
 	pkt->entry_count = 1;
 	pkt->handle = QLA_TGT_SKIP_HANDLE | CTIO_COMPLETION_HANDLE_MARK;
 
@@ -5103,7 +5103,7 @@ static int __qlt_send_busy(struct qla_qpair *qpair,
 		return -ENOMEM;
 	}
 
-	vha->tgt_counters.num_q_full_sent++;
+	qpair->tgt_counters.num_q_full_sent++;
 	pkt->entry_count = 1;
 	pkt->handle = QLA_TGT_SKIP_HANDLE | CTIO_COMPLETION_HANDLE_MARK;
 
@@ -5466,13 +5466,12 @@ static void qlt_24xx_atio_pkt(struct scsi_qla_host *vha,
 static void qlt_response_pkt(struct scsi_qla_host *vha,
 	struct rsp_que *rsp, response_t *pkt)
 {
-	struct qla_hw_data *ha = vha->hw;
 	struct qla_tgt *tgt = vha->vha_tgt.qla_tgt;
 
 	if (unlikely(tgt == NULL)) {
 		ql_dbg(ql_dbg_tgt, vha, 0xe05d,
-		    "qla_target(%d): Response pkt %x received, but no "
-		    "tgt (ha %p)\n", vha->vp_idx, pkt->entry_type, ha);
+		    "qla_target(%d): Response pkt %x received, but no tgt (ha %p)\n",
+		    vha->vp_idx, pkt->entry_type, vha->hw);
 		return;
 	}
 

drivers/scsi/qla2xxx/tcm_qla2xxx.c

@@ -283,7 +283,7 @@ static void tcm_qla2xxx_complete_free(struct work_struct *work)
 
 	WARN_ON(cmd->trc_flags & TRC_CMD_FREE);
 
-	cmd->vha->tgt_counters.qla_core_ret_sta_ctio++;
+	cmd->qpair->tgt_counters.qla_core_ret_sta_ctio++;
 	cmd->trc_flags |= TRC_CMD_FREE;
 	transport_generic_free_cmd(&cmd->se_cmd, 0);
 }
@@ -295,7 +295,7 @@ static void tcm_qla2xxx_complete_free(struct work_struct *work)
  */
 static void tcm_qla2xxx_free_cmd(struct qla_tgt_cmd *cmd)
 {
-	cmd->vha->tgt_counters.core_qla_free_cmd++;
+	cmd->qpair->tgt_counters.core_qla_free_cmd++;
 	cmd->cmd_in_wq = 1;
 
 	WARN_ON(cmd->trc_flags & TRC_CMD_DONE);
@@ -491,7 +491,7 @@ static int tcm_qla2xxx_handle_cmd(scsi_qla_host_t *vha, struct qla_tgt_cmd *cmd,
 	}
 #endif
 
-	cmd->vha->tgt_counters.qla_core_sbt_cmd++;
+	cmd->qpair->tgt_counters.qla_core_sbt_cmd++;
 	return target_submit_cmd(se_cmd, se_sess, cdb, &cmd->sense_buffer[0],
 				cmd->unpacked_lun, data_length, fcp_task_attr,
 				data_dir, flags);
@@ -519,7 +519,7 @@ static void tcm_qla2xxx_handle_data_work(struct work_struct *work)
 	}
 	spin_unlock_irqrestore(&cmd->cmd_lock, flags);
 
-	cmd->vha->tgt_counters.qla_core_ret_ctio++;
+	cmd->qpair->tgt_counters.qla_core_ret_ctio++;
 	if (!cmd->write_data_transferred) {
 		/*
 		 * Check if se_cmd has already been aborted via LUN_RESET, and