[SCSI] qla2xxx: Add CPU affinity support.

Set the module parameter ql2xmultique_tag to 1 to enable this
feature. In this mode, the total number of response queues
created is equal to the number of online cpus. Turning the block
layer's rq_affinity mode on enables requests to be routed to the
proper cpu and at the same time it enables completion of the IO
in a response queue that is affined to the cpu in the request
path.

Signed-off-by: Anirban Chakraborty <anirban.chakraborty@qlogic.com>
Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>

1 files changed, 76 insertions, 2 deletions

diff --git a/drivers/scsi/qla2xxx/qla_os.c b/drivers/scsi/qla2xxx/qla_os.c
index e2647e02dac9..d6817df95e30 100644
--- a/drivers/scsi/qla2xxx/qla_os.c
+++ b/drivers/scsi/qla2xxx/qla_os.c
@@ -96,6 +96,13 @@ MODULE_PARM_DESC(ql2xmaxqueues,
                 "Enables MQ settings "
                 "Default is 1 for single queue. Set it to number \
                         of queues in MQ mode.");
+
+int ql2xmultique_tag;
+module_param(ql2xmultique_tag, int, S_IRUGO|S_IRUSR);
+MODULE_PARM_DESC(ql2xmultique_tag,
+                "Enables CPU affinity settings for the driver "
+                "Default is 0 for no affinity of request and response IO. "
+                "Set it to 1 to turn on the cpu affinity.");
 /*
  * SCSI host template entry points
  */
@@ -256,6 +263,47 @@ static void qla2x00_free_queues(struct qla_hw_data *ha)
         ha->rsp_q_map = NULL;
 }
 
+static int qla25xx_setup_mode(struct scsi_qla_host *vha)
+{
+        uint16_t options = 0;
+        int ques, req, ret;
+        struct qla_hw_data *ha = vha->hw;
+
+        if (ql2xmultique_tag) {
+                /* CPU affinity mode */
+                ha->wq = create_workqueue("qla2xxx_wq");
+                /* create a request queue for IO */
+                options |= BIT_7;
+                req = qla25xx_create_req_que(ha, options, 0, 0, -1,
+                        QLA_DEFAULT_QUE_QOS);
+                if (!req) {
+                        qla_printk(KERN_WARNING, ha,
+                                "Can't create request queue\n");
+                        goto fail;
+                }
+                vha->req = ha->req_q_map[req];
+                options |= BIT_1;
+                for (ques = 1; ques < ha->max_rsp_queues; ques++) {
+                        ret = qla25xx_create_rsp_que(ha, options, 0, 0, req);
+                        if (!ret) {
+                                qla_printk(KERN_WARNING, ha,
+                                        "Response Queue create failed\n");
+                                goto fail2;
+                        }
+                }
+                DEBUG2(qla_printk(KERN_INFO, ha,
+                        "CPU affinity mode enabled, no. of response"
+                        " queues:%d, no. of request queues:%d\n",
+                        ha->max_rsp_queues, ha->max_req_queues));
+        }
+        return 0;
+fail2:
+        qla25xx_delete_queues(vha);
+fail:
+        ha->mqenable = 0;
+        return 1;
+}
+
 static char *
 qla2x00_pci_info_str(struct scsi_qla_host *vha, char *str)
 {
@@ -998,6 +1046,9 @@ qla2xxx_eh_host_reset(struct scsi_cmnd *cmd)
                 if (qla2x00_vp_abort_isp(vha))
                         goto eh_host_reset_lock;
         } else {
+                if (ha->wq)
+                        flush_workqueue(ha->wq);
+
                 set_bit(ABORT_ISP_ACTIVE, &base_vha->dpc_flags);
                 if (qla2x00_abort_isp(base_vha)) {
                         clear_bit(ABORT_ISP_ACTIVE, &base_vha->dpc_flags);
@@ -1521,6 +1572,7 @@ qla2x00_iospace_config(struct qla_hw_data *ha)
 {
         resource_size_t pio;
         uint16_t msix;
+        int cpus;
 
         if (pci_request_selected_regions(ha->pdev, ha->bars,
             QLA2XXX_DRIVER_NAME)) {
@@ -1575,7 +1627,7 @@ skip_pio:
 
         /* Determine queue resources */
         ha->max_req_queues = ha->max_rsp_queues = 1;
-        if (ql2xmaxqueues <= 1  &&
+        if ((ql2xmaxqueues <= 1 || ql2xmultique_tag < 1) &&
                 (!IS_QLA25XX(ha) && !IS_QLA81XX(ha)))
                 goto mqiobase_exit;
         ha->mqiobase = ioremap(pci_resource_start(ha->pdev, 3),
@@ -1584,12 +1636,21 @@ skip_pio:
                 /* Read MSIX vector size of the board */
                 pci_read_config_word(ha->pdev, QLA_PCI_MSIX_CONTROL, &msix);
                 ha->msix_count = msix;
-                if (ql2xmaxqueues > 1) {
+                /* Max queues are bounded by available msix vectors */
+                /* queue 0 uses two msix vectors */
+                if (ql2xmultique_tag) {
+                        cpus = num_online_cpus();
+                        ha->max_rsp_queues = (ha->msix_count - 1 - cpus) ?
+                                (cpus + 1) : (ha->msix_count - 1);
+                        ha->max_req_queues = 2;
+                } else if (ql2xmaxqueues > 1) {
                         ha->max_req_queues = ql2xmaxqueues > QLA_MQ_SIZE ?
                                                 QLA_MQ_SIZE : ql2xmaxqueues;
                         DEBUG2(qla_printk(KERN_INFO, ha, "QoS mode set, max no"
                         " of request queues:%d\n", ha->max_req_queues));
                 }
+                qla_printk(KERN_INFO, ha,
+                        "MSI-X vector count: %d\n", msix);
         } else
                 qla_printk(KERN_INFO, ha, "BAR 3 not enabled\n");
 
@@ -1871,6 +1932,12 @@ qla2x00_probe_one(struct pci_dev *pdev, const struct pci_device_id *id)
                 goto probe_failed;
         }
 
+        if (ha->mqenable)
+                if (qla25xx_setup_mode(base_vha))
+                        qla_printk(KERN_WARNING, ha,
+                                "Can't create queues, falling back to single"
+                                " queue mode\n");
+
         /*
          * Startup the kernel thread for this host adapter
          */
@@ -1982,6 +2049,13 @@ qla2x00_remove_one(struct pci_dev *pdev)
 
         base_vha->flags.online = 0;
 
+        /* Flush the work queue and remove it */
+        if (ha->wq) {
+                flush_workqueue(ha->wq);
+                destroy_workqueue(ha->wq);
+                ha->wq = NULL;
+        }
+
         /* Kill the kernel thread for this host */
         if (ha->dpc_thread) {
                 struct task_struct *t = ha->dpc_thread;