scsi: lpfc: Fix erroneous cpu limit of 128 on I/O statistics

The cpu io statistics were capped by a hard define limit of 128. This
effectively was a max number of CPUs, not an actual CPU count, nor actual
CPU numbers which can be even larger than both of those values. This made
stats off/misleading and on large CPU count systems, wrong.

Fix the stats so that all CPUs can have a stats struct.  Fix the looping
such that it loops by hdwq, finds CPUs that used the hdwq, and sum the
stats, then display.

Link: https://lore.kernel.org/r/20200322181304.37655-9-jsmart2021@gmail.comSigned-off-by: James Smart <jsmart2021@gmail.com>
Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

drivers/scsi/lpfc/lpfc.h

@@ -481,7 +481,7 @@ struct lpfc_vport {
 	struct dentry *debug_nvmestat;
 	struct dentry *debug_scsistat;
 	struct dentry *debug_nvmektime;
-	struct dentry *debug_cpucheck;
+	struct dentry *debug_hdwqstat;
 	struct dentry *vport_debugfs_root;
 	struct lpfc_debugfs_trc *disc_trc;
 	atomic_t disc_trc_cnt;
@@ -1175,12 +1175,11 @@ struct lpfc_hba {
 	uint16_t sfp_warning;
 
 #ifdef CONFIG_SCSI_LPFC_DEBUG_FS
-	uint16_t cpucheck_on;
+	uint16_t hdwqstat_on;
 #define LPFC_CHECK_OFF		0
 #define LPFC_CHECK_NVME_IO	1
-#define LPFC_CHECK_NVMET_RCV	2
-#define LPFC_CHECK_NVMET_IO	4
-#define LPFC_CHECK_SCSI_IO	8
+#define LPFC_CHECK_NVMET_IO	2
+#define LPFC_CHECK_SCSI_IO	4
 	uint16_t ktime_on;
 	uint64_t ktime_data_samples;
 	uint64_t ktime_status_samples;

drivers/scsi/lpfc/lpfc_debugfs.h

@@ -47,7 +47,6 @@
 /* nvmestat output buffer size */
 #define LPFC_NVMESTAT_SIZE 8192
 #define LPFC_NVMEKTIME_SIZE 8192
-#define LPFC_CPUCHECK_SIZE 8192
 #define LPFC_NVMEIO_TRC_SIZE 8192
 
 /* scsistat output buffer size */

drivers/scsi/lpfc/lpfc_init.c

@@ -6951,6 +6951,17 @@ lpfc_sli4_driver_resource_setup(struct lpfc_hba *phba)
 		rc = -ENOMEM;
 		goto out_free_hba_cpu_map;
 	}
+
+#ifdef CONFIG_SCSI_LPFC_DEBUG_FS
+	phba->sli4_hba.c_stat = alloc_percpu(struct lpfc_hdwq_stat);
+	if (!phba->sli4_hba.c_stat) {
+		lpfc_printf_log(phba, KERN_ERR, LOG_INIT,
+				"3332 Failed allocating per cpu hdwq stats\n");
+		rc = -ENOMEM;
+		goto out_free_hba_eq_info;
+	}
+#endif
+
 	/*
 	 * Enable sr-iov virtual functions if supported and configured
 	 * through the module parameter.
@@ -6970,6 +6981,10 @@ lpfc_sli4_driver_resource_setup(struct lpfc_hba *phba)
 
 	return 0;
 
+#ifdef CONFIG_SCSI_LPFC_DEBUG_FS
+out_free_hba_eq_info:
+	free_percpu(phba->sli4_hba.eq_info);
+#endif
 out_free_hba_cpu_map:
 	kfree(phba->sli4_hba.cpu_map);
 out_free_hba_eq_hdl:
@@ -7008,6 +7023,9 @@ lpfc_sli4_driver_resource_unset(struct lpfc_hba *phba)
 	struct lpfc_fcf_conn_entry *conn_entry, *next_conn_entry;
 
 	free_percpu(phba->sli4_hba.eq_info);
+#ifdef CONFIG_SCSI_LPFC_DEBUG_FS
+	free_percpu(phba->sli4_hba.c_stat);
+#endif
 
 	/* Free memory allocated for msi-x interrupt vector to CPU mapping */
 	kfree(phba->sli4_hba.cpu_map);
@@ -10848,6 +10866,9 @@ lpfc_cpu_affinity_check(struct lpfc_hba *phba, int vectors)
 #ifdef CONFIG_X86
 	struct cpuinfo_x86 *cpuinfo;
 #endif
+#ifdef CONFIG_SCSI_LPFC_DEBUG_FS
+	struct lpfc_hdwq_stat *c_stat;
+#endif
 
 	max_phys_id = 0;
 	min_phys_id = LPFC_VECTOR_MAP_EMPTY;
@@ -11099,10 +11120,17 @@ lpfc_cpu_affinity_check(struct lpfc_hba *phba, int vectors)
 	idx = 0;
 	for_each_possible_cpu(cpu) {
 		cpup = &phba->sli4_hba.cpu_map[cpu];
+#ifdef CONFIG_SCSI_LPFC_DEBUG_FS
+		c_stat = per_cpu_ptr(phba->sli4_hba.c_stat, cpu);
+		c_stat->hdwq_no = cpup->hdwq;
+#endif
 		if (cpup->hdwq != LPFC_VECTOR_MAP_EMPTY)
 			continue;
 
 		cpup->hdwq = idx++ % phba->cfg_hdw_queue;
+#ifdef CONFIG_SCSI_LPFC_DEBUG_FS
+		c_stat->hdwq_no = cpup->hdwq;
+#endif
 		lpfc_printf_log(phba, KERN_INFO, LOG_INIT,
 				"3340 Set Affinity: not present "
 				"CPU %d hdwq %d\n",

drivers/scsi/lpfc/lpfc_nvme.c

@@ -1012,6 +1012,9 @@ lpfc_nvme_io_cmd_wqe_cmpl(struct lpfc_hba *phba, struct lpfc_iocbq *pwqeIn,
 	uint32_t code, status, idx;
 	uint16_t cid, sqhd, data;
 	uint32_t *ptr;
+#ifdef CONFIG_SCSI_LPFC_DEBUG_FS
+	int cpu;
+#endif
 
 	/* Sanity check on return of outstanding command */
 	if (!lpfc_ncmd) {
@@ -1184,19 +1187,15 @@ lpfc_nvme_io_cmd_wqe_cmpl(struct lpfc_hba *phba, struct lpfc_iocbq *pwqeIn,
 		phba->ktime_last_cmd = lpfc_ncmd->ts_data_nvme;
 		lpfc_nvme_ktime(phba, lpfc_ncmd);
 	}
-	if (unlikely(phba->cpucheck_on & LPFC_CHECK_NVME_IO)) {
-		uint32_t cpu;
-		idx = lpfc_ncmd->cur_iocbq.hba_wqidx;
+	if (unlikely(phba->hdwqstat_on & LPFC_CHECK_NVME_IO)) {
 		cpu = raw_smp_processor_id();
-		if (cpu < LPFC_CHECK_CPU_CNT) {
-			if (lpfc_ncmd->cpu != cpu)
-				lpfc_printf_vlog(vport,
-						 KERN_INFO, LOG_NVME_IOERR,
-						 "6701 CPU Check cmpl: "
-						 "cpu %d expect %d\n",
-						 cpu, lpfc_ncmd->cpu);
-			phba->sli4_hba.hdwq[idx].cpucheck_cmpl_io[cpu]++;
-		}
+		this_cpu_inc(phba->sli4_hba.c_stat->cmpl_io);
+		if (lpfc_ncmd->cpu != cpu)
+			lpfc_printf_vlog(vport,
+					 KERN_INFO, LOG_NVME_IOERR,
+					 "6701 CPU Check cmpl: "
+					 "cpu %d expect %d\n",
+					 cpu, lpfc_ncmd->cpu);
 	}
 #endif
 
@@ -1745,19 +1744,17 @@ lpfc_nvme_fcp_io_submit(struct nvme_fc_local_port *pnvme_lport,
 	if (lpfc_ncmd->ts_cmd_start)
 		lpfc_ncmd->ts_cmd_wqput = ktime_get_ns();
 
-	if (phba->cpucheck_on & LPFC_CHECK_NVME_IO) {
+	if (phba->hdwqstat_on & LPFC_CHECK_NVME_IO) {
 		cpu = raw_smp_processor_id();
-		if (cpu < LPFC_CHECK_CPU_CNT) {
-			lpfc_ncmd->cpu = cpu;
-			if (idx != cpu)
-				lpfc_printf_vlog(vport,
-						 KERN_INFO, LOG_NVME_IOERR,
-						"6702 CPU Check cmd: "
-						"cpu %d wq %d\n",
-						lpfc_ncmd->cpu,
-						lpfc_queue_info->index);
-			phba->sli4_hba.hdwq[idx].cpucheck_xmt_io[cpu]++;
-		}
+		this_cpu_inc(phba->sli4_hba.c_stat->xmt_io);
+		lpfc_ncmd->cpu = cpu;
+		if (idx != cpu)
+			lpfc_printf_vlog(vport,
+					 KERN_INFO, LOG_NVME_IOERR,
+					"6702 CPU Check cmd: "
+					"cpu %d wq %d\n",
+					lpfc_ncmd->cpu,
+					lpfc_queue_info->index);
 	}
 #endif
 	return 0;

drivers/scsi/lpfc/lpfc_nvmet.c

@@ -707,7 +707,7 @@ lpfc_nvmet_xmt_fcp_op_cmp(struct lpfc_hba *phba, struct lpfc_iocbq *cmdwqe,
 	struct lpfc_nvmet_rcv_ctx *ctxp;
 	uint32_t status, result, op, start_clean, logerr;
 #ifdef CONFIG_SCSI_LPFC_DEBUG_FS
-	uint32_t id;
+	int id;
 #endif
 
 	ctxp = cmdwqe->context2;
@@ -814,16 +814,14 @@ lpfc_nvmet_xmt_fcp_op_cmp(struct lpfc_hba *phba, struct lpfc_iocbq *cmdwqe,
 		rsp->done(rsp);
 	}
 #ifdef CONFIG_SCSI_LPFC_DEBUG_FS
-	if (phba->cpucheck_on & LPFC_CHECK_NVMET_IO) {
+	if (phba->hdwqstat_on & LPFC_CHECK_NVMET_IO) {
 		id = raw_smp_processor_id();
-		if (id < LPFC_CHECK_CPU_CNT) {
-			if (ctxp->cpu != id)
-				lpfc_printf_log(phba, KERN_INFO, LOG_NVME_IOERR,
-						"6704 CPU Check cmdcmpl: "
-						"cpu %d expect %d\n",
-						id, ctxp->cpu);
-			phba->sli4_hba.hdwq[rsp->hwqid].cpucheck_cmpl_io[id]++;
-		}
+		this_cpu_inc(phba->sli4_hba.c_stat->cmpl_io);
+		if (ctxp->cpu != id)
+			lpfc_printf_log(phba, KERN_INFO, LOG_NVME_IOERR,
+					"6704 CPU Check cmdcmpl: "
+					"cpu %d expect %d\n",
+					id, ctxp->cpu);
 	}
 #endif
 }
@@ -931,6 +929,9 @@ lpfc_nvmet_xmt_fcp_op(struct nvmet_fc_target_port *tgtport,
 	struct lpfc_sli_ring *pring;
 	unsigned long iflags;
 	int rc;
+#ifdef CONFIG_SCSI_LPFC_DEBUG_FS
+	int id;
+#endif
 
 	if (phba->pport->load_flag & FC_UNLOADING) {
 		rc = -ENODEV;
@@ -954,16 +955,14 @@ lpfc_nvmet_xmt_fcp_op(struct nvmet_fc_target_port *tgtport,
 	if (!ctxp->hdwq)
 		ctxp->hdwq = &phba->sli4_hba.hdwq[rsp->hwqid];
 
-	if (phba->cpucheck_on & LPFC_CHECK_NVMET_IO) {
-		int id = raw_smp_processor_id();
-		if (id < LPFC_CHECK_CPU_CNT) {
-			if (rsp->hwqid != id)
-				lpfc_printf_log(phba, KERN_INFO, LOG_NVME_IOERR,
-						"6705 CPU Check OP: "
-						"cpu %d expect %d\n",
-						id, rsp->hwqid);
-			phba->sli4_hba.hdwq[rsp->hwqid].cpucheck_xmt_io[id]++;
-		}
+	if (phba->hdwqstat_on & LPFC_CHECK_NVMET_IO) {
+		id = raw_smp_processor_id();
+		this_cpu_inc(phba->sli4_hba.c_stat->xmt_io);
+		if (rsp->hwqid != id)
+			lpfc_printf_log(phba, KERN_INFO, LOG_NVME_IOERR,
+					"6705 CPU Check OP: "
+					"cpu %d expect %d\n",
+					id, rsp->hwqid);
 		ctxp->cpu = id; /* Setup cpu for cmpl check */
 	}
 #endif
@@ -2270,15 +2269,13 @@ lpfc_nvmet_unsol_fcp_buffer(struct lpfc_hba *phba,
 	size = nvmebuf->bytes_recv;
 
 #ifdef CONFIG_SCSI_LPFC_DEBUG_FS
-	if (phba->cpucheck_on & LPFC_CHECK_NVMET_RCV) {
-		if (current_cpu < LPFC_CHECK_CPU_CNT) {
-			if (idx != current_cpu)
-				lpfc_printf_log(phba, KERN_INFO, LOG_NVME_IOERR,
-						"6703 CPU Check rcv: "
-						"cpu %d expect %d\n",
-						current_cpu, idx);
-			phba->sli4_hba.hdwq[idx].cpucheck_rcv_io[current_cpu]++;
-		}
+	if (phba->hdwqstat_on & LPFC_CHECK_NVMET_IO) {
+		this_cpu_inc(phba->sli4_hba.c_stat->rcv_io);
+		if (idx != current_cpu)
+			lpfc_printf_log(phba, KERN_INFO, LOG_NVME_IOERR,
+					"6703 CPU Check rcv: "
+					"cpu %d expect %d\n",
+					current_cpu, idx);
 	}
 #endif
 

drivers/scsi/lpfc/lpfc_scsi.c

@@ -3805,9 +3805,6 @@ lpfc_scsi_cmd_iocb_cmpl(struct lpfc_hba *phba, struct lpfc_iocbq *pIocbIn,
 	struct Scsi_Host *shost;
 	int idx;
 	uint32_t logit = LOG_FCP;
-#ifdef CONFIG_SCSI_LPFC_DEBUG_FS
-	int cpu;
-#endif
 
 	/* Guard against abort handler being called at same time */
 	spin_lock(&lpfc_cmd->buf_lock);
@@ -3826,11 +3823,8 @@ lpfc_scsi_cmd_iocb_cmpl(struct lpfc_hba *phba, struct lpfc_iocbq *pIocbIn,
 		phba->sli4_hba.hdwq[idx].scsi_cstat.io_cmpls++;
 
 #ifdef CONFIG_SCSI_LPFC_DEBUG_FS
-	if (unlikely(phba->cpucheck_on & LPFC_CHECK_SCSI_IO)) {
-		cpu = raw_smp_processor_id();
-		if (cpu < LPFC_CHECK_CPU_CNT && phba->sli4_hba.hdwq)
-			phba->sli4_hba.hdwq[idx].cpucheck_cmpl_io[cpu]++;
-	}
+	if (unlikely(phba->hdwqstat_on & LPFC_CHECK_SCSI_IO))
+		this_cpu_inc(phba->sli4_hba.c_stat->cmpl_io);
 #endif
 	shost = cmd->device->host;
 
@@ -4503,9 +4497,6 @@ lpfc_queuecommand(struct Scsi_Host *shost, struct scsi_cmnd *cmnd)
 	struct lpfc_io_buf *lpfc_cmd;
 	struct fc_rport *rport = starget_to_rport(scsi_target(cmnd->device));
 	int err, idx;
-#ifdef CONFIG_SCSI_LPFC_DEBUG_FS
-	int cpu;
-#endif
 
 	rdata = lpfc_rport_data_from_scsi_device(cmnd->device);
 
@@ -4626,14 +4617,8 @@ lpfc_queuecommand(struct Scsi_Host *shost, struct scsi_cmnd *cmnd)
 	lpfc_scsi_prep_cmnd(vport, lpfc_cmd, ndlp);
 
 #ifdef CONFIG_SCSI_LPFC_DEBUG_FS
-	if (unlikely(phba->cpucheck_on & LPFC_CHECK_SCSI_IO)) {
-		cpu = raw_smp_processor_id();
-		if (cpu < LPFC_CHECK_CPU_CNT) {
-			struct lpfc_sli4_hdw_queue *hdwq =
-					&phba->sli4_hba.hdwq[lpfc_cmd->hdwq_no];
-			hdwq->cpucheck_xmt_io[cpu]++;
-		}
-	}
+	if (unlikely(phba->hdwqstat_on & LPFC_CHECK_SCSI_IO))
+		this_cpu_inc(phba->sli4_hba.c_stat->xmt_io);
 #endif
 	err = lpfc_sli_issue_iocb(phba, LPFC_FCP_RING,
 				  &lpfc_cmd->cur_iocbq, SLI_IOCB_RET_IOCB);

drivers/scsi/lpfc/lpfc_sli4.h

@@ -697,13 +697,6 @@ struct lpfc_sli4_hdw_queue {
 	struct lpfc_lock_stat lock_conflict;
 #endif
 
-#ifdef CONFIG_SCSI_LPFC_DEBUG_FS
-#define LPFC_CHECK_CPU_CNT    128
-	uint32_t cpucheck_rcv_io[LPFC_CHECK_CPU_CNT];
-	uint32_t cpucheck_xmt_io[LPFC_CHECK_CPU_CNT];
-	uint32_t cpucheck_cmpl_io[LPFC_CHECK_CPU_CNT];
-#endif
-
 	/* Per HDWQ pool resources */
 	struct list_head sgl_list;
 	struct list_head cmd_rsp_buf_list;
@@ -740,6 +733,15 @@ struct lpfc_sli4_hdw_queue {
 #define lpfc_qp_spin_lock(lock, qp, lstat) spin_lock(lock)
 #endif
 
+#ifdef CONFIG_SCSI_LPFC_DEBUG_FS
+struct lpfc_hdwq_stat {
+	u32 hdwq_no;
+	u32 rcv_io;
+	u32 xmt_io;
+	u32 cmpl_io;
+};
+#endif
+
 struct lpfc_sli4_hba {
 	void __iomem *conf_regs_memmap_p; /* Kernel memory mapped address for
 					   * config space registers
@@ -921,6 +923,9 @@ struct lpfc_sli4_hba {
 	struct cpumask numa_mask;
 	uint16_t curr_disp_cpu;
 	struct lpfc_eq_intr_info __percpu *eq_info;
+#ifdef CONFIG_SCSI_LPFC_DEBUG_FS
+	struct lpfc_hdwq_stat __percpu *c_stat;
+#endif
 	uint32_t conf_trunk;
 #define lpfc_conf_trunk_port0_WORD	conf_trunk
 #define lpfc_conf_trunk_port0_SHIFT	0