scsi: mpt3sas: Use firmware recommended queue depth

Currently, the mpt3sas driver sets the default queue depth based on the
physical interface of the attached device:

 - SAS : 254
 - SATA:  32
 - NVMe: 128

The IOC firmware provides a recommended queue depth for each device through
SAS IO Unit Page1 for SAS/SATA and PCIe IO Unit Page 1 for NVMe devices.

If the host sets the queue depth greater than the firmware recommended
value, then the IOC places the I/Os above the recommended queue depth in an
internal pending queue. This consumes outstanding host-credit/resources,
thereby leading to potential starvation of other devices.

To avoid this, use the device depth recommended by the IOC firmware.

Link: https://lore.kernel.org/r/20210809072639.21228-2-suganath-prabu.subramani@broadcom.comSigned-off-by: Suganath Prabu S <suganath-prabu.subramani@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

drivers/scsi/mpt3sas/mpt3sas_base.c

@@ -5365,6 +5365,73 @@ _base_update_diag_trigger_pages(struct MPT3SAS_ADAPTER *ioc)
 		    &ioc->diag_trigger_mpi, 1);
 }
 
+/**
+ * _base_assign_fw_reported_qd	- Get FW reported QD for SAS/SATA devices.
+ *				- On failure set default QD values.
+ * @ioc : per adapter object
+ *
+ * Returns 0 for success, non-zero for failure.
+ *
+ */
+static int _base_assign_fw_reported_qd(struct MPT3SAS_ADAPTER *ioc)
+{
+	Mpi2ConfigReply_t mpi_reply;
+	Mpi2SasIOUnitPage1_t *sas_iounit_pg1 = NULL;
+	Mpi26PCIeIOUnitPage1_t pcie_iounit_pg1;
+	int sz;
+	int rc = 0;
+
+	ioc->max_wideport_qd = MPT3SAS_SAS_QUEUE_DEPTH;
+	ioc->max_narrowport_qd = MPT3SAS_SAS_QUEUE_DEPTH;
+	ioc->max_sata_qd = MPT3SAS_SATA_QUEUE_DEPTH;
+	ioc->max_nvme_qd = MPT3SAS_NVME_QUEUE_DEPTH;
+	if (!ioc->is_gen35_ioc)
+		goto out;
+	/* sas iounit page 1 */
+	sz = offsetof(Mpi2SasIOUnitPage1_t, PhyData);
+	sas_iounit_pg1 = kzalloc(sz, GFP_KERNEL);
+	if (!sas_iounit_pg1) {
+		pr_err("%s: failure at %s:%d/%s()!\n",
+		    ioc->name, __FILE__, __LINE__, __func__);
+		return rc;
+	}
+	rc = mpt3sas_config_get_sas_iounit_pg1(ioc, &mpi_reply,
+	    sas_iounit_pg1, sz);
+	if (rc) {
+		pr_err("%s: failure at %s:%d/%s()!\n",
+		    ioc->name, __FILE__, __LINE__, __func__);
+		goto out;
+	}
+	ioc->max_wideport_qd =
+	    (le16_to_cpu(sas_iounit_pg1->SASWideMaxQueueDepth)) ?
+	    le16_to_cpu(sas_iounit_pg1->SASWideMaxQueueDepth) :
+	    MPT3SAS_SAS_QUEUE_DEPTH;
+	ioc->max_narrowport_qd =
+	    (le16_to_cpu(sas_iounit_pg1->SASNarrowMaxQueueDepth)) ?
+	    le16_to_cpu(sas_iounit_pg1->SASNarrowMaxQueueDepth) :
+	    MPT3SAS_SAS_QUEUE_DEPTH;
+	ioc->max_sata_qd = (sas_iounit_pg1->SATAMaxQDepth) ?
+	    sas_iounit_pg1->SATAMaxQDepth : MPT3SAS_SATA_QUEUE_DEPTH;
+	/* pcie iounit page 1 */
+	rc = mpt3sas_config_get_pcie_iounit_pg1(ioc, &mpi_reply,
+	    &pcie_iounit_pg1, sizeof(Mpi26PCIeIOUnitPage1_t));
+	if (rc) {
+		pr_err("%s: failure at %s:%d/%s()!\n",
+		    ioc->name, __FILE__, __LINE__, __func__);
+		goto out;
+	}
+	ioc->max_nvme_qd = (le16_to_cpu(pcie_iounit_pg1.NVMeMaxQueueDepth)) ?
+	    (le16_to_cpu(pcie_iounit_pg1.NVMeMaxQueueDepth)) :
+	    MPT3SAS_NVME_QUEUE_DEPTH;
+out:
+	dinitprintk(ioc, pr_err(
+	    "MaxWidePortQD: 0x%x MaxNarrowPortQD: 0x%x MaxSataQD: 0x%x MaxNvmeQD: 0x%x\n",
+	    ioc->max_wideport_qd, ioc->max_narrowport_qd,
+	    ioc->max_sata_qd, ioc->max_nvme_qd));
+	kfree(sas_iounit_pg1);
+	return rc;
+}
+
 /**
  * _base_static_config_pages - static start of day config pages
  * @ioc: per adapter object
@@ -5434,6 +5501,9 @@ _base_static_config_pages(struct MPT3SAS_ADAPTER *ioc)
 			ioc_warn(ioc,
 			    "TimeSync Interval in Manuf page-11 is not enabled. Periodic Time-Sync will be disabled\n");
 	}
+	rc = _base_assign_fw_reported_qd(ioc);
+	if (rc)
+		return rc;
 	rc = mpt3sas_config_get_bios_pg2(ioc, &mpi_reply, &ioc->bios_pg2);
 	if (rc)
 		return rc;

drivers/scsi/mpt3sas/mpt3sas_base.h

@@ -576,6 +576,7 @@ struct _sas_device {
 	u8	is_chassis_slot_valid;
 	u8	connector_name[5];
 	struct kref refcount;
+	u8	port_type;
 	struct hba_port *port;
 	struct sas_rphy *rphy;
 };
@@ -1443,6 +1444,10 @@ struct MPT3SAS_ADAPTER {
 	u8		tm_custom_handling;
 	u8		nvme_abort_timeout;
 	u16		max_shutdown_latency;
+	u16		max_wideport_qd;
+	u16		max_narrowport_qd;
+	u16		max_nvme_qd;
+	u8		max_sata_qd;
 
 	/* static config pages */
 	struct mpt3sas_facts facts;
@@ -1848,6 +1853,9 @@ int mpt3sas_config_get_pcie_device_pg0(struct MPT3SAS_ADAPTER *ioc,
 int mpt3sas_config_get_pcie_device_pg2(struct MPT3SAS_ADAPTER *ioc,
 	Mpi2ConfigReply_t *mpi_reply, Mpi26PCIeDevicePage2_t *config_page,
 	u32 form, u32 handle);
+int mpt3sas_config_get_pcie_iounit_pg1(struct MPT3SAS_ADAPTER *ioc,
+	Mpi2ConfigReply_t *mpi_reply, Mpi26PCIeIOUnitPage1_t *config_page,
+	u16 sz);
 int mpt3sas_config_get_sas_iounit_pg0(struct MPT3SAS_ADAPTER *ioc,
 	Mpi2ConfigReply_t *mpi_reply, Mpi2SasIOUnitPage0_t *config_page,
 	u16 sz);

drivers/scsi/mpt3sas/mpt3sas_config.c

@@ -1168,6 +1168,43 @@ mpt3sas_config_get_pcie_device_pg0(struct MPT3SAS_ADAPTER *ioc,
 	return r;
 }
 
+/**
+ * mpt3sas_config_get_pcie_iounit_pg1 - obtain pcie iounit page 1
+ * @ioc: per adapter object
+ * @mpi_reply: reply mf payload returned from firmware
+ * @config_page: contents of the config page
+ * @sz: size of buffer passed in config_page
+ * Context: sleep.
+ *
+ * Returns 0 for success, non-zero for failure.
+ */
+int
+mpt3sas_config_get_pcie_iounit_pg1(struct MPT3SAS_ADAPTER *ioc,
+	Mpi2ConfigReply_t *mpi_reply, Mpi26PCIeIOUnitPage1_t *config_page,
+	u16 sz)
+{
+	Mpi2ConfigRequest_t mpi_request;
+	int r;
+
+	memset(&mpi_request, 0, sizeof(Mpi2ConfigRequest_t));
+	mpi_request.Function = MPI2_FUNCTION_CONFIG;
+	mpi_request.Action = MPI2_CONFIG_ACTION_PAGE_HEADER;
+	mpi_request.Header.PageType = MPI2_CONFIG_PAGETYPE_EXTENDED;
+	mpi_request.ExtPageType = MPI2_CONFIG_EXTPAGETYPE_PCIE_IO_UNIT;
+	mpi_request.Header.PageVersion = MPI26_PCIEIOUNITPAGE1_PAGEVERSION;
+	mpi_request.Header.PageNumber = 1;
+	ioc->build_zero_len_sge_mpi(ioc, &mpi_request.PageBufferSGE);
+	r = _config_request(ioc, &mpi_request, mpi_reply,
+	    MPT3_CONFIG_PAGE_DEFAULT_TIMEOUT, NULL, 0);
+	if (r)
+		goto out;
+	mpi_request.Action = MPI2_CONFIG_ACTION_PAGE_READ_CURRENT;
+	r = _config_request(ioc, &mpi_request, mpi_reply,
+	    MPT3_CONFIG_PAGE_DEFAULT_TIMEOUT, config_page, sz);
+out:
+	return r;
+}
+
 /**
  * mpt3sas_config_get_pcie_device_pg2 - obtain pcie device page 2
  * @ioc: per adapter object

drivers/scsi/mpt3sas/mpt3sas_ctl.c

@@ -3820,9 +3820,10 @@ enable_sdev_max_qd_store(struct device *cdev,
 				}
 			} else if (sas_target_priv_data->flags &
 			    MPT_TARGET_FLAGS_PCIE_DEVICE)
-				qdepth = MPT3SAS_NVME_QUEUE_DEPTH;
+				qdepth = ioc->max_nvme_qd;
 			else
-				qdepth = MPT3SAS_SAS_QUEUE_DEPTH;
+				qdepth = (sas_target_priv_data->sas_dev->port_type > 1) ?
+				    ioc->max_wideport_qd : ioc->max_narrowport_qd;
 
 			mpt3sas_scsih_change_queue_depth(sdev, qdepth);
 		}

drivers/scsi/mpt3sas/mpt3sas_scsih.c

@@ -1803,7 +1803,7 @@ scsih_change_queue_depth(struct scsi_device *sdev, int qdepth)
 	 * limit max device queue for SATA to 32 if enable_sdev_max_qd
 	 * is disabled.
 	 */
-	if (ioc->enable_sdev_max_qd)
+	if (ioc->enable_sdev_max_qd || ioc->is_gen35_ioc)
 		goto not_sata;
 
 	sas_device_priv_data = sdev->hostdata;
@@ -2657,7 +2657,7 @@ scsih_slave_configure(struct scsi_device *sdev)
 			return 1;
 		}
 
-		qdepth = MPT3SAS_NVME_QUEUE_DEPTH;
+		qdepth = ioc->max_nvme_qd;
 		ds = "NVMe";
 		sdev_printk(KERN_INFO, sdev,
 			"%s: handle(0x%04x), wwid(0x%016llx), port(%d)\n",
@@ -2709,7 +2709,8 @@ scsih_slave_configure(struct scsi_device *sdev)
 	sas_device->volume_handle = volume_handle;
 	sas_device->volume_wwid = volume_wwid;
 	if (sas_device->device_info & MPI2_SAS_DEVICE_INFO_SSP_TARGET) {
-		qdepth = MPT3SAS_SAS_QUEUE_DEPTH;
+		qdepth = (sas_device->port_type > 1) ?
+			ioc->max_wideport_qd : ioc->max_narrowport_qd;
 		ssp_target = 1;
 		if (sas_device->device_info &
 				MPI2_SAS_DEVICE_INFO_SEP) {
@@ -2721,7 +2722,7 @@ scsih_slave_configure(struct scsi_device *sdev)
 		} else
 			ds = "SSP";
 	} else {
-		qdepth = MPT3SAS_SATA_QUEUE_DEPTH;
+		qdepth = ioc->max_sata_qd;
 		if (sas_device->device_info & MPI2_SAS_DEVICE_INFO_STP_TARGET)
 			ds = "STP";
 		else if (sas_device->device_info &
@@ -7371,6 +7372,10 @@ _scsih_add_device(struct MPT3SAS_ADAPTER *ioc, u16 handle, u8 phy_num,
 
 	/* get device name */
 	sas_device->device_name = le64_to_cpu(sas_device_pg0.DeviceName);
+	sas_device->port_type = sas_device_pg0.MaxPortConnections;
+	ioc_info(ioc,
+	    "handle(0x%0x) sas_address(0x%016llx) port_type(0x%0x)\n",
+	    handle, sas_device->sas_address, sas_device->port_type);
 
 	if (ioc->wait_for_discovery_to_complete)
 		_scsih_sas_device_init_add(ioc, sas_device);
@@ -9603,6 +9608,42 @@ _scsih_prep_device_scan(struct MPT3SAS_ADAPTER *ioc)
 	}
 }
 
+/**
+ * _scsih_update_device_qdepth - Update QD during Reset.
+ * @ioc: per adapter object
+ *
+ */
+static void
+_scsih_update_device_qdepth(struct MPT3SAS_ADAPTER *ioc)
+{
+	struct MPT3SAS_DEVICE *sas_device_priv_data;
+	struct MPT3SAS_TARGET *sas_target_priv_data;
+	struct _sas_device *sas_device;
+	struct scsi_device *sdev;
+	u16 qdepth;
+
+	ioc_info(ioc, "Update devices with firmware reported queue depth\n");
+	shost_for_each_device(sdev, ioc->shost) {
+		sas_device_priv_data = sdev->hostdata;
+		if (sas_device_priv_data && sas_device_priv_data->sas_target) {
+			sas_target_priv_data = sas_device_priv_data->sas_target;
+			sas_device = sas_device_priv_data->sas_target->sas_dev;
+			if (sas_target_priv_data->flags & MPT_TARGET_FLAGS_PCIE_DEVICE)
+				qdepth = ioc->max_nvme_qd;
+			else if (sas_device &&
+			    sas_device->device_info & MPI2_SAS_DEVICE_INFO_SSP_TARGET)
+				qdepth = (sas_device->port_type > 1) ?
+				    ioc->max_wideport_qd : ioc->max_narrowport_qd;
+			else if (sas_device &&
+			    sas_device->device_info & MPI2_SAS_DEVICE_INFO_SATA_DEVICE)
+				qdepth = ioc->max_sata_qd;
+			else
+				continue;
+			mpt3sas_scsih_change_queue_depth(sdev, qdepth);
+		}
+	}
+}
+
 /**
  * _scsih_mark_responding_sas_device - mark a sas_devices as responding
  * @ioc: per adapter object
@@ -10654,6 +10695,8 @@ _mpt3sas_fw_work(struct MPT3SAS_ADAPTER *ioc, struct fw_event_work *fw_event)
 		_scsih_remove_unresponding_devices(ioc);
 		_scsih_del_dirty_vphy(ioc);
 		_scsih_del_dirty_port_entries(ioc);
+		if (ioc->is_gen35_ioc)
+			_scsih_update_device_qdepth(ioc);
 		_scsih_scan_for_devices_after_reset(ioc);
 		/*
 		 * If diag reset has occurred during the driver load