diff options
Diffstat (limited to 'drivers/scsi/megaraid/megaraid_sas_fusion.c')
-rw-r--r-- | drivers/scsi/megaraid/megaraid_sas_fusion.c | 1334 |
1 files changed, 1080 insertions, 254 deletions
diff --git a/drivers/scsi/megaraid/megaraid_sas_fusion.c b/drivers/scsi/megaraid/megaraid_sas_fusion.c index 24778ba4b6e8..29650ba669da 100644 --- a/drivers/scsi/megaraid/megaraid_sas_fusion.c +++ b/drivers/scsi/megaraid/megaraid_sas_fusion.c @@ -47,6 +47,7 @@ #include <linux/blkdev.h> #include <linux/mutex.h> #include <linux/poll.h> +#include <linux/vmalloc.h> #include <scsi/scsi.h> #include <scsi/scsi_cmnd.h> @@ -181,32 +182,44 @@ inline void megasas_return_cmd_fusion(struct megasas_instance *instance, struct megasas_cmd_fusion *cmd) { cmd->scmd = NULL; - memset(cmd->io_request, 0, sizeof(struct MPI2_RAID_SCSI_IO_REQUEST)); + memset(cmd->io_request, 0, MEGA_MPI2_RAID_DEFAULT_IO_FRAME_SIZE); + cmd->r1_alt_dev_handle = MR_DEVHANDLE_INVALID; + cmd->cmd_completed = false; } /** * megasas_fire_cmd_fusion - Sends command to the FW + * @instance: Adapter soft state + * @req_desc: 32bit or 64bit Request descriptor + * + * Perform PCI Write. Ventura supports 32 bit Descriptor. + * Prior to Ventura (12G) MR controller supports 64 bit Descriptor. */ + static void megasas_fire_cmd_fusion(struct megasas_instance *instance, union MEGASAS_REQUEST_DESCRIPTOR_UNION *req_desc) { + if (instance->is_ventura) + writel(le32_to_cpu(req_desc->u.low), + &instance->reg_set->inbound_single_queue_port); + else { #if defined(writeq) && defined(CONFIG_64BIT) - u64 req_data = (((u64)le32_to_cpu(req_desc->u.high) << 32) | - le32_to_cpu(req_desc->u.low)); + u64 req_data = (((u64)le32_to_cpu(req_desc->u.high) << 32) | + le32_to_cpu(req_desc->u.low)); - writeq(req_data, &instance->reg_set->inbound_low_queue_port); + writeq(req_data, &instance->reg_set->inbound_low_queue_port); #else - unsigned long flags; - - spin_lock_irqsave(&instance->hba_lock, flags); - writel(le32_to_cpu(req_desc->u.low), - &instance->reg_set->inbound_low_queue_port); - writel(le32_to_cpu(req_desc->u.high), - &instance->reg_set->inbound_high_queue_port); - mmiowb(); - spin_unlock_irqrestore(&instance->hba_lock, flags); + unsigned long flags; + spin_lock_irqsave(&instance->hba_lock, flags); + writel(le32_to_cpu(req_desc->u.low), + &instance->reg_set->inbound_low_queue_port); + writel(le32_to_cpu(req_desc->u.high), + &instance->reg_set->inbound_high_queue_port); + mmiowb(); + spin_unlock_irqrestore(&instance->hba_lock, flags); #endif + } } /** @@ -229,7 +242,10 @@ megasas_fusion_update_can_queue(struct megasas_instance *instance, int fw_boot_c reg_set = instance->reg_set; - cur_max_fw_cmds = readl(&instance->reg_set->outbound_scratch_pad_3) & 0x00FFFF; + /* ventura FW does not fill outbound_scratch_pad_3 with queue depth */ + if (!instance->is_ventura) + cur_max_fw_cmds = + readl(&instance->reg_set->outbound_scratch_pad_3) & 0x00FFFF; if (dual_qdepth_disable || !cur_max_fw_cmds) cur_max_fw_cmds = instance->instancet->read_fw_status_reg(reg_set) & 0x00FFFF; @@ -243,7 +259,7 @@ megasas_fusion_update_can_queue(struct megasas_instance *instance, int fw_boot_c if (fw_boot_context == OCR_CONTEXT) { cur_max_fw_cmds = cur_max_fw_cmds - 1; - if (cur_max_fw_cmds <= instance->max_fw_cmds) { + if (cur_max_fw_cmds < instance->max_fw_cmds) { instance->cur_can_queue = cur_max_fw_cmds - (MEGASAS_FUSION_INTERNAL_CMDS + MEGASAS_FUSION_IOCTL_CMDS); @@ -255,7 +271,8 @@ megasas_fusion_update_can_queue(struct megasas_instance *instance, int fw_boot_c instance->ldio_threshold = ldio_threshold; if (!instance->is_rdpq) - instance->max_fw_cmds = min_t(u16, instance->max_fw_cmds, 1024); + instance->max_fw_cmds = + min_t(u16, instance->max_fw_cmds, 1024); if (reset_devices) instance->max_fw_cmds = min(instance->max_fw_cmds, @@ -271,7 +288,14 @@ megasas_fusion_update_can_queue(struct megasas_instance *instance, int fw_boot_c (MEGASAS_FUSION_INTERNAL_CMDS + MEGASAS_FUSION_IOCTL_CMDS); instance->cur_can_queue = instance->max_scsi_cmds; + instance->host->can_queue = instance->cur_can_queue; } + + if (instance->is_ventura) + instance->max_mpt_cmds = + instance->max_fw_cmds * RAID_1_PEER_CMDS; + else + instance->max_mpt_cmds = instance->max_fw_cmds; } /** * megasas_free_cmds_fusion - Free all the cmds in the free cmd pool @@ -285,7 +309,7 @@ megasas_free_cmds_fusion(struct megasas_instance *instance) struct megasas_cmd_fusion *cmd; /* SG, Sense */ - for (i = 0; i < instance->max_fw_cmds; i++) { + for (i = 0; i < instance->max_mpt_cmds; i++) { cmd = fusion->cmd_list[i]; if (cmd) { if (cmd->sg_frame) @@ -329,7 +353,7 @@ megasas_free_cmds_fusion(struct megasas_instance *instance) /* cmd_list */ - for (i = 0; i < instance->max_fw_cmds; i++) + for (i = 0; i < instance->max_mpt_cmds; i++) kfree(fusion->cmd_list[i]); kfree(fusion->cmd_list); @@ -343,7 +367,7 @@ megasas_free_cmds_fusion(struct megasas_instance *instance) static int megasas_create_sg_sense_fusion(struct megasas_instance *instance) { int i; - u32 max_cmd; + u16 max_cmd; struct fusion_context *fusion; struct megasas_cmd_fusion *cmd; @@ -353,7 +377,8 @@ static int megasas_create_sg_sense_fusion(struct megasas_instance *instance) fusion->sg_dma_pool = pci_pool_create("mr_sg", instance->pdev, - instance->max_chain_frame_sz, 4, 0); + instance->max_chain_frame_sz, + MR_DEFAULT_NVME_PAGE_SIZE, 0); /* SCSI_SENSE_BUFFERSIZE = 96 bytes */ fusion->sense_dma_pool = pci_pool_create("mr_sense", instance->pdev, @@ -381,33 +406,47 @@ static int megasas_create_sg_sense_fusion(struct megasas_instance *instance) return -ENOMEM; } } + + /* create sense buffer for the raid 1/10 fp */ + for (i = max_cmd; i < instance->max_mpt_cmds; i++) { + cmd = fusion->cmd_list[i]; + cmd->sense = pci_pool_alloc(fusion->sense_dma_pool, + GFP_KERNEL, &cmd->sense_phys_addr); + if (!cmd->sense) { + dev_err(&instance->pdev->dev, + "Failed from %s %d\n", __func__, __LINE__); + return -ENOMEM; + } + } + return 0; } int megasas_alloc_cmdlist_fusion(struct megasas_instance *instance) { - u32 max_cmd, i; + u32 max_mpt_cmd, i; struct fusion_context *fusion; fusion = instance->ctrl_context; - max_cmd = instance->max_fw_cmds; + max_mpt_cmd = instance->max_mpt_cmds; /* * fusion->cmd_list is an array of struct megasas_cmd_fusion pointers. * Allocate the dynamic array first and then allocate individual * commands. */ - fusion->cmd_list = kzalloc(sizeof(struct megasas_cmd_fusion *) * max_cmd, - GFP_KERNEL); + fusion->cmd_list = + kzalloc(sizeof(struct megasas_cmd_fusion *) * max_mpt_cmd, + GFP_KERNEL); if (!fusion->cmd_list) { dev_err(&instance->pdev->dev, "Failed from %s %d\n", __func__, __LINE__); return -ENOMEM; } - for (i = 0; i < max_cmd; i++) { + for (i = 0; i < max_mpt_cmd; i++) { fusion->cmd_list[i] = kzalloc(sizeof(struct megasas_cmd_fusion), GFP_KERNEL); if (!fusion->cmd_list[i]) { @@ -539,7 +578,7 @@ megasas_alloc_rdpq_fusion(struct megasas_instance *instance) } fusion->rdpq_virt[i].RDPQBaseAddress = - fusion->reply_frames_desc_phys[i]; + cpu_to_le64(fusion->reply_frames_desc_phys[i]); reply_desc = fusion->reply_frames_desc[i]; for (j = 0; j < fusion->reply_q_depth; j++, reply_desc++) @@ -642,13 +681,14 @@ megasas_alloc_cmds_fusion(struct megasas_instance *instance) */ /* SMID 0 is reserved. Set SMID/index from 1 */ - for (i = 0; i < instance->max_fw_cmds; i++) { + for (i = 0; i < instance->max_mpt_cmds; i++) { cmd = fusion->cmd_list[i]; offset = MEGA_MPI2_RAID_DEFAULT_IO_FRAME_SIZE * i; memset(cmd, 0, sizeof(struct megasas_cmd_fusion)); cmd->index = i + 1; cmd->scmd = NULL; - cmd->sync_cmd_idx = (i >= instance->max_scsi_cmds) ? + cmd->sync_cmd_idx = + (i >= instance->max_scsi_cmds && i < instance->max_fw_cmds) ? (i - instance->max_scsi_cmds) : (u32)ULONG_MAX; /* Set to Invalid */ cmd->instance = instance; @@ -658,6 +698,7 @@ megasas_alloc_cmds_fusion(struct megasas_instance *instance) memset(cmd->io_request, 0, sizeof(struct MPI2_RAID_SCSI_IO_REQUEST)); cmd->io_request_phys_addr = io_req_base_phys + offset; + cmd->r1_alt_dev_handle = MR_DEVHANDLE_INVALID; } if (megasas_create_sg_sense_fusion(instance)) @@ -725,6 +766,7 @@ megasas_ioc_init_fusion(struct megasas_instance *instance) const char *sys_info; MFI_CAPABILITIES *drv_ops; u32 scratch_pad_2; + unsigned long flags; fusion = instance->ctrl_context; @@ -781,6 +823,7 @@ megasas_ioc_init_fusion(struct megasas_instance *instance) MPI2_IOCINIT_MSGFLAG_RDPQ_ARRAY_MODE : 0; IOCInitMessage->SystemRequestFrameBaseAddress = cpu_to_le64(fusion->io_request_frames_phys); IOCInitMessage->HostMSIxVectors = instance->msix_vectors; + IOCInitMessage->HostPageSize = MR_DEFAULT_NVME_PAGE_SHIFT; init_frame = (struct megasas_init_frame *)cmd->frame; memset(init_frame, 0, MEGAMFI_FRAME_SIZE); @@ -796,7 +839,7 @@ megasas_ioc_init_fusion(struct megasas_instance *instance) drv_ops = (MFI_CAPABILITIES *) &(init_frame->driver_operations); /* driver support Extended MSIX */ - if (fusion->adapter_type == INVADER_SERIES) + if (fusion->adapter_type >= INVADER_SERIES) drv_ops->mfi_capabilities.support_additional_msix = 1; /* driver supports HA / Remote LUN over Fast Path interface */ drv_ops->mfi_capabilities.support_fp_remote_lun = 1; @@ -813,6 +856,7 @@ megasas_ioc_init_fusion(struct megasas_instance *instance) drv_ops->mfi_capabilities.support_ext_queue_depth = 1; drv_ops->mfi_capabilities.support_qd_throttling = 1; + drv_ops->mfi_capabilities.support_pd_map_target_id = 1; /* Convert capability to LE32 */ cpu_to_le32s((u32 *)&init_frame->driver_operations.mfi_capabilities); @@ -850,7 +894,14 @@ megasas_ioc_init_fusion(struct megasas_instance *instance) break; } - megasas_fire_cmd_fusion(instance, &req_desc); + /* For Ventura also IOC INIT required 64 bit Descriptor write. */ + spin_lock_irqsave(&instance->hba_lock, flags); + writel(le32_to_cpu(req_desc.u.low), + &instance->reg_set->inbound_low_queue_port); + writel(le32_to_cpu(req_desc.u.high), + &instance->reg_set->inbound_high_queue_port); + mmiowb(); + spin_unlock_irqrestore(&instance->hba_lock, flags); wait_and_poll(instance, cmd, MFI_POLL_TIMEOUT_SECS); @@ -1009,11 +1060,6 @@ megasas_get_ld_map_info(struct megasas_instance *instance) memset(ci, 0, fusion->max_map_sz); memset(dcmd->mbox.b, 0, MFI_MBOX_SIZE); -#if VD_EXT_DEBUG - dev_dbg(&instance->pdev->dev, - "%s sending MR_DCMD_LD_MAP_GET_INFO with size %d\n", - __func__, cpu_to_le32(size_map_info)); -#endif dcmd->cmd = MFI_CMD_DCMD; dcmd->cmd_status = 0xFF; dcmd->sge_count = 1; @@ -1065,10 +1111,11 @@ megasas_get_map_info(struct megasas_instance *instance) int megasas_sync_map_info(struct megasas_instance *instance) { - int ret = 0, i; + int i; struct megasas_cmd *cmd; struct megasas_dcmd_frame *dcmd; - u32 size_sync_info, num_lds; + u16 num_lds; + u32 size_sync_info; struct fusion_context *fusion; struct MR_LD_TARGET_SYNC *ci = NULL; struct MR_DRV_RAID_MAP_ALL *map; @@ -1134,7 +1181,7 @@ megasas_sync_map_info(struct megasas_instance *instance) instance->instancet->issue_dcmd(instance, cmd); - return ret; + return 0; } /* @@ -1220,7 +1267,8 @@ megasas_init_adapter_fusion(struct megasas_instance *instance) { struct megasas_register_set __iomem *reg_set; struct fusion_context *fusion; - u32 max_cmd, scratch_pad_2; + u16 max_cmd; + u32 scratch_pad_2; int i = 0, count; fusion = instance->ctrl_context; @@ -1230,13 +1278,6 @@ megasas_init_adapter_fusion(struct megasas_instance *instance) megasas_fusion_update_can_queue(instance, PROBE_CONTEXT); /* - * Reduce the max supported cmds by 1. This is to ensure that the - * reply_q_sz (1 more than the max cmd that driver may send) - * does not exceed max cmds that the FW can support - */ - instance->max_fw_cmds = instance->max_fw_cmds-1; - - /* * Only Driver's internal DCMDs and IOCTL DCMDs needs to have MFI frames */ instance->max_mfi_cmds = @@ -1247,12 +1288,12 @@ megasas_init_adapter_fusion(struct megasas_instance *instance) fusion->reply_q_depth = 2 * (((max_cmd + 1 + 15)/16)*16); fusion->request_alloc_sz = - sizeof(union MEGASAS_REQUEST_DESCRIPTOR_UNION) *max_cmd; + sizeof(union MEGASAS_REQUEST_DESCRIPTOR_UNION) * instance->max_mpt_cmds; fusion->reply_alloc_sz = sizeof(union MPI2_REPLY_DESCRIPTORS_UNION) *(fusion->reply_q_depth); fusion->io_frames_alloc_sz = MEGA_MPI2_RAID_DEFAULT_IO_FRAME_SIZE + - (MEGA_MPI2_RAID_DEFAULT_IO_FRAME_SIZE * - (max_cmd + 1)); /* Extra 1 for SMID 0 */ + (MEGA_MPI2_RAID_DEFAULT_IO_FRAME_SIZE + * (instance->max_mpt_cmds + 1)); /* Extra 1 for SMID 0 */ scratch_pad_2 = readl(&instance->reg_set->outbound_scratch_pad_2); /* If scratch_pad_2 & MEGASAS_MAX_CHAIN_SIZE_UNITS_MASK is set, @@ -1302,7 +1343,7 @@ megasas_init_adapter_fusion(struct megasas_instance *instance) fusion->last_reply_idx[i] = 0; /* - * For fusion adapters, 3 commands for IOCTL and 5 commands + * For fusion adapters, 3 commands for IOCTL and 8 commands * for driver's internal DCMDs. */ instance->max_scsi_cmds = instance->max_fw_cmds - @@ -1331,6 +1372,7 @@ megasas_init_adapter_fusion(struct megasas_instance *instance) } instance->flag_ieee = 1; + instance->r1_ldio_hint_default = MR_R1_LDIO_PIGGYBACK_DEFAULT; fusion->fast_path_io = 0; fusion->drv_map_pages = get_order(fusion->drv_map_sz); @@ -1388,96 +1430,348 @@ fail_alloc_mfi_cmds: */ void -map_cmd_status(struct megasas_cmd_fusion *cmd, u8 status, u8 ext_status) +map_cmd_status(struct fusion_context *fusion, + struct scsi_cmnd *scmd, u8 status, u8 ext_status, + u32 data_length, u8 *sense) { + u8 cmd_type; + int resid; + cmd_type = megasas_cmd_type(scmd); switch (status) { case MFI_STAT_OK: - cmd->scmd->result = DID_OK << 16; + scmd->result = DID_OK << 16; break; case MFI_STAT_SCSI_IO_FAILED: case MFI_STAT_LD_INIT_IN_PROGRESS: - cmd->scmd->result = (DID_ERROR << 16) | ext_status; + scmd->result = (DID_ERROR << 16) | ext_status; break; case MFI_STAT_SCSI_DONE_WITH_ERROR: - cmd->scmd->result = (DID_OK << 16) | ext_status; + scmd->result = (DID_OK << 16) | ext_status; if (ext_status == SAM_STAT_CHECK_CONDITION) { - memset(cmd->scmd->sense_buffer, 0, + memset(scmd->sense_buffer, 0, SCSI_SENSE_BUFFERSIZE); - memcpy(cmd->scmd->sense_buffer, cmd->sense, + memcpy(scmd->sense_buffer, sense, SCSI_SENSE_BUFFERSIZE); - cmd->scmd->result |= DRIVER_SENSE << 24; + scmd->result |= DRIVER_SENSE << 24; } + + /* + * If the IO request is partially completed, then MR FW will + * update "io_request->DataLength" field with actual number of + * bytes transferred.Driver will set residual bytes count in + * SCSI command structure. + */ + resid = (scsi_bufflen(scmd) - data_length); + scsi_set_resid(scmd, resid); + + if (resid && + ((cmd_type == READ_WRITE_LDIO) || + (cmd_type == READ_WRITE_SYSPDIO))) + scmd_printk(KERN_INFO, scmd, "BRCM Debug mfi stat 0x%x, data len" + " requested/completed 0x%x/0x%x\n", + status, scsi_bufflen(scmd), data_length); break; case MFI_STAT_LD_OFFLINE: case MFI_STAT_DEVICE_NOT_FOUND: - cmd->scmd->result = DID_BAD_TARGET << 16; + scmd->result = DID_BAD_TARGET << 16; break; case MFI_STAT_CONFIG_SEQ_MISMATCH: - cmd->scmd->result = DID_IMM_RETRY << 16; + scmd->result = DID_IMM_RETRY << 16; break; default: - dev_printk(KERN_DEBUG, &cmd->instance->pdev->dev, "FW status %#x\n", status); - cmd->scmd->result = DID_ERROR << 16; + scmd->result = DID_ERROR << 16; break; } } /** + * megasas_is_prp_possible - + * Checks if native NVMe PRPs can be built for the IO + * + * @instance: Adapter soft state + * @scmd: SCSI command from the mid-layer + * @sge_count: scatter gather element count. + * + * Returns: true: PRPs can be built + * false: IEEE SGLs needs to be built + */ +static bool +megasas_is_prp_possible(struct megasas_instance *instance, + struct scsi_cmnd *scmd, int sge_count) +{ + struct fusion_context *fusion; + int i; + u32 data_length = 0; + struct scatterlist *sg_scmd; + bool build_prp = false; + u32 mr_nvme_pg_size; + + mr_nvme_pg_size = max_t(u32, instance->nvme_page_size, + MR_DEFAULT_NVME_PAGE_SIZE); + fusion = instance->ctrl_context; + data_length = scsi_bufflen(scmd); + sg_scmd = scsi_sglist(scmd); + + /* + * NVMe uses one PRP for each page (or part of a page) + * look at the data length - if 4 pages or less then IEEE is OK + * if > 5 pages then we need to build a native SGL + * if > 4 and <= 5 pages, then check physical address of 1st SG entry + * if this first size in the page is >= the residual beyond 4 pages + * then use IEEE, otherwise use native SGL + */ + + if (data_length > (mr_nvme_pg_size * 5)) { + build_prp = true; + } else if ((data_length > (mr_nvme_pg_size * 4)) && + (data_length <= (mr_nvme_pg_size * 5))) { + /* check if 1st SG entry size is < residual beyond 4 pages */ + if (sg_dma_len(sg_scmd) < (data_length - (mr_nvme_pg_size * 4))) + build_prp = true; + } + +/* + * Below code detects gaps/holes in IO data buffers. + * What does holes/gaps mean? + * Any SGE except first one in a SGL starts at non NVME page size + * aligned address OR Any SGE except last one in a SGL ends at + * non NVME page size boundary. + * + * Driver has already informed block layer by setting boundary rules for + * bio merging done at NVME page size boundary calling kernel API + * blk_queue_virt_boundary inside slave_config. + * Still there is possibility of IO coming with holes to driver because of + * IO merging done by IO scheduler. + * + * With SCSI BLK MQ enabled, there will be no IO with holes as there is no + * IO scheduling so no IO merging. + * + * With SCSI BLK MQ disabled, IO scheduler may attempt to merge IOs and + * then sending IOs with holes. + * + * Though driver can request block layer to disable IO merging by calling- + * queue_flag_set_unlocked(QUEUE_FLAG_NOMERGES, sdev->request_queue) but + * user may tune sysfs parameter- nomerges again to 0 or 1. + * + * If in future IO scheduling is enabled with SCSI BLK MQ, + * this algorithm to detect holes will be required in driver + * for SCSI BLK MQ enabled case as well. + * + * + */ + scsi_for_each_sg(scmd, sg_scmd, sge_count, i) { + if ((i != 0) && (i != (sge_count - 1))) { + if (mega_mod64(sg_dma_len(sg_scmd), mr_nvme_pg_size) || + mega_mod64(sg_dma_address(sg_scmd), + mr_nvme_pg_size)) { + build_prp = false; + atomic_inc(&instance->sge_holes_type1); + break; + } + } + + if ((sge_count > 1) && (i == 0)) { + if ((mega_mod64((sg_dma_address(sg_scmd) + + sg_dma_len(sg_scmd)), + mr_nvme_pg_size))) { + build_prp = false; + atomic_inc(&instance->sge_holes_type2); + break; + } + } + + if ((sge_count > 1) && (i == (sge_count - 1))) { + if (mega_mod64(sg_dma_address(sg_scmd), + mr_nvme_pg_size)) { + build_prp = false; + atomic_inc(&instance->sge_holes_type3); + break; + } + } + } + + return build_prp; +} + +/** + * megasas_make_prp_nvme - + * Prepare PRPs(Physical Region Page)- SGLs specific to NVMe drives only + * + * @instance: Adapter soft state + * @scmd: SCSI command from the mid-layer + * @sgl_ptr: SGL to be filled in + * @cmd: Fusion command frame + * @sge_count: scatter gather element count. + * + * Returns: true: PRPs are built + * false: IEEE SGLs needs to be built + */ +static bool +megasas_make_prp_nvme(struct megasas_instance *instance, struct scsi_cmnd *scmd, + struct MPI25_IEEE_SGE_CHAIN64 *sgl_ptr, + struct megasas_cmd_fusion *cmd, int sge_count) +{ + int sge_len, offset, num_prp_in_chain = 0; + struct MPI25_IEEE_SGE_CHAIN64 *main_chain_element, *ptr_first_sgl; + u64 *ptr_sgl; + dma_addr_t ptr_sgl_phys; + u64 sge_addr; + u32 page_mask, page_mask_result; + struct scatterlist *sg_scmd; + u32 first_prp_len; + bool build_prp = false; + int data_len = scsi_bufflen(scmd); + struct fusion_context *fusion; + u32 mr_nvme_pg_size = max_t(u32, instance->nvme_page_size, + MR_DEFAULT_NVME_PAGE_SIZE); + + fusion = instance->ctrl_context; + + build_prp = megasas_is_prp_possible(instance, scmd, sge_count); + + if (!build_prp) + return false; + + /* + * Nvme has a very convoluted prp format. One prp is required + * for each page or partial page. Driver need to split up OS sg_list + * entries if it is longer than one page or cross a page + * boundary. Driver also have to insert a PRP list pointer entry as + * the last entry in each physical page of the PRP list. + * + * NOTE: The first PRP "entry" is actually placed in the first + * SGL entry in the main message as IEEE 64 format. The 2nd + * entry in the main message is the chain element, and the rest + * of the PRP entries are built in the contiguous pcie buffer. + */ + page_mask = mr_nvme_pg_size - 1; + ptr_sgl = (u64 *)cmd->sg_frame; + ptr_sgl_phys = cmd->sg_frame_phys_addr; + memset(ptr_sgl, 0, instance->max_chain_frame_sz); + + /* Build chain frame element which holds all prps except first*/ + main_chain_element = (struct MPI25_IEEE_SGE_CHAIN64 *) + ((u8 *)sgl_ptr + sizeof(struct MPI25_IEEE_SGE_CHAIN64)); + + main_chain_element->Address = cpu_to_le64(ptr_sgl_phys); + main_chain_element->NextChainOffset = 0; + main_chain_element->Flags = IEEE_SGE_FLAGS_CHAIN_ELEMENT | + IEEE_SGE_FLAGS_SYSTEM_ADDR | + MPI26_IEEE_SGE_FLAGS_NSF_NVME_PRP; + + /* Build first prp, sge need not to be page aligned*/ + ptr_first_sgl = sgl_ptr; + sg_scmd = scsi_sglist(scmd); + sge_addr = sg_dma_address(sg_scmd); + sge_len = sg_dma_len(sg_scmd); + + offset = (u32)(sge_addr & page_mask); + first_prp_len = mr_nvme_pg_size - offset; + + ptr_first_sgl->Address = cpu_to_le64(sge_addr); + ptr_first_sgl->Length = cpu_to_le32(first_prp_len); + + data_len -= first_prp_len; + + if (sge_len > first_prp_len) { + sge_addr += first_prp_len; + sge_len -= first_prp_len; + } else if (sge_len == first_prp_len) { + sg_scmd = sg_next(sg_scmd); + sge_addr = sg_dma_address(sg_scmd); + sge_len = sg_dma_len(sg_scmd); + } + + for (;;) { + offset = (u32)(sge_addr & page_mask); + + /* Put PRP pointer due to page boundary*/ + page_mask_result = (uintptr_t)(ptr_sgl + 1) & page_mask; + if (unlikely(!page_mask_result)) { + scmd_printk(KERN_NOTICE, + scmd, "page boundary ptr_sgl: 0x%p\n", + ptr_sgl); + ptr_sgl_phys += 8; + *ptr_sgl = cpu_to_le64(ptr_sgl_phys); + ptr_sgl++; + num_prp_in_chain++; + } + + *ptr_sgl = cpu_to_le64(sge_addr); + ptr_sgl++; + ptr_sgl_phys += 8; + num_prp_in_chain++; + + sge_addr += mr_nvme_pg_size; + sge_len -= mr_nvme_pg_size; + data_len -= mr_nvme_pg_size; + + if (data_len <= 0) + break; + + if (sge_len > 0) + continue; + + sg_scmd = sg_next(sg_scmd); + sge_addr = sg_dma_address(sg_scmd); + sge_len = sg_dma_len(sg_scmd); + } + + main_chain_element->Length = + cpu_to_le32(num_prp_in_chain * sizeof(u64)); + + atomic_inc(&instance->prp_sgl); + return build_prp; +} + +/** * megasas_make_sgl_fusion - Prepares 32-bit SGL * @instance: Adapter soft state * @scp: SCSI command from the mid-layer * @sgl_ptr: SGL to be filled in * @cmd: cmd we are working on + * @sge_count sge count * - * If successful, this function returns the number of SG elements. */ -static int +static void megasas_make_sgl_fusion(struct megasas_instance *instance, struct scsi_cmnd *scp, struct MPI25_IEEE_SGE_CHAIN64 *sgl_ptr, - struct megasas_cmd_fusion *cmd) + struct megasas_cmd_fusion *cmd, int sge_count) { - int i, sg_processed, sge_count; + int i, sg_processed; struct scatterlist *os_sgl; struct fusion_context *fusion; fusion = instance->ctrl_context; - if (fusion->adapter_type == INVADER_SERIES) { + if (fusion->adapter_type >= INVADER_SERIES) { struct MPI25_IEEE_SGE_CHAIN64 *sgl_ptr_end = sgl_ptr; sgl_ptr_end += fusion->max_sge_in_main_msg - 1; sgl_ptr_end->Flags = 0; } - sge_count = scsi_dma_map(scp); - - BUG_ON(sge_count < 0); - - if (sge_count > instance->max_num_sge || !sge_count) - return sge_count; - scsi_for_each_sg(scp, os_sgl, sge_count, i) { sgl_ptr->Length = cpu_to_le32(sg_dma_len(os_sgl)); sgl_ptr->Address = cpu_to_le64(sg_dma_address(os_sgl)); sgl_ptr->Flags = 0; - if (fusion->adapter_type == INVADER_SERIES) + if (fusion->adapter_type >= INVADER_SERIES) if (i == sge_count - 1) sgl_ptr->Flags = IEEE_SGE_FLAGS_END_OF_LIST; sgl_ptr++; - sg_processed = i + 1; if ((sg_processed == (fusion->max_sge_in_main_msg - 1)) && (sge_count > fusion->max_sge_in_main_msg)) { struct MPI25_IEEE_SGE_CHAIN64 *sg_chain; - if (fusion->adapter_type == INVADER_SERIES) { + if (fusion->adapter_type >= INVADER_SERIES) { if ((le16_to_cpu(cmd->io_request->IoFlags) & MPI25_SAS_DEVICE0_FLAGS_ENABLED_FAST_PATH) != MPI25_SAS_DEVICE0_FLAGS_ENABLED_FAST_PATH) @@ -1493,7 +1787,7 @@ megasas_make_sgl_fusion(struct megasas_instance *instance, sg_chain = sgl_ptr; /* Prepare chain element */ sg_chain->NextChainOffset = 0; - if (fusion->adapter_type == INVADER_SERIES) + if (fusion->adapter_type >= INVADER_SERIES) sg_chain->Flags = IEEE_SGE_FLAGS_CHAIN_ELEMENT; else sg_chain->Flags = @@ -1507,6 +1801,45 @@ megasas_make_sgl_fusion(struct megasas_instance *instance, memset(sgl_ptr, 0, instance->max_chain_frame_sz); } } + atomic_inc(&instance->ieee_sgl); +} + +/** + * megasas_make_sgl - Build Scatter Gather List(SGLs) + * @scp: SCSI command pointer + * @instance: Soft instance of controller + * @cmd: Fusion command pointer + * + * This function will build sgls based on device type. + * For nvme drives, there is different way of building sgls in nvme native + * format- PRPs(Physical Region Page). + * + * Returns the number of sg lists actually used, zero if the sg lists + * is NULL, or -ENOMEM if the mapping failed + */ +static +int megasas_make_sgl(struct megasas_instance *instance, struct scsi_cmnd *scp, + struct megasas_cmd_fusion *cmd) +{ + int sge_count; + bool build_prp = false; + struct MPI25_IEEE_SGE_CHAIN64 *sgl_chain64; + + sge_count = scsi_dma_map(scp); + + if ((sge_count > instance->max_num_sge) || (sge_count <= 0)) + return sge_count; + + sgl_chain64 = (struct MPI25_IEEE_SGE_CHAIN64 *)&cmd->io_request->SGL; + if ((le16_to_cpu(cmd->io_request->IoFlags) & + MPI25_SAS_DEVICE0_FLAGS_ENABLED_FAST_PATH) && + (cmd->pd_interface == NVME_PD)) + build_prp = megasas_make_prp_nvme(instance, scp, sgl_chain64, + cmd, sge_count); + + if (!build_prp) + megasas_make_sgl_fusion(instance, scp, sgl_chain64, + cmd, sge_count); return sge_count; } @@ -1525,7 +1858,7 @@ megasas_set_pd_lba(struct MPI2_RAID_SCSI_IO_REQUEST *io_request, u8 cdb_len, struct MR_DRV_RAID_MAP_ALL *local_map_ptr, u32 ref_tag) { struct MR_LD_RAID *raid; - u32 ld; + u16 ld; u64 start_blk = io_info->pdBlock; u8 *cdb = io_request->CDB.CDB32; u32 num_blocks = io_info->numBlocks; @@ -1574,6 +1907,7 @@ megasas_set_pd_lba(struct MPI2_RAID_SCSI_IO_REQUEST *io_request, u8 cdb_len, MPI2_SCSIIO_EEDPFLAGS_CHECK_REFTAG | MPI2_SCSIIO_EEDPFLAGS_CHECK_REMOVE_OP | MPI2_SCSIIO_EEDPFLAGS_CHECK_APPTAG | + MPI25_SCSIIO_EEDPFLAGS_DO_NOT_DISABLE_MODE | MPI2_SCSIIO_EEDPFLAGS_CHECK_GUARD); } else { io_request->EEDPFlags = cpu_to_le16( @@ -1688,6 +2022,166 @@ megasas_set_pd_lba(struct MPI2_RAID_SCSI_IO_REQUEST *io_request, u8 cdb_len, } /** + * megasas_stream_detect - stream detection on read and and write IOs + * @instance: Adapter soft state + * @cmd: Command to be prepared + * @io_info: IO Request info + * + */ + +/** stream detection on read and and write IOs */ +static void megasas_stream_detect(struct megasas_instance *instance, + struct megasas_cmd_fusion *cmd, + struct IO_REQUEST_INFO *io_info) +{ + struct fusion_context *fusion = instance->ctrl_context; + u32 device_id = io_info->ldTgtId; + struct LD_STREAM_DETECT *current_ld_sd + = fusion->stream_detect_by_ld[device_id]; + u32 *track_stream = ¤t_ld_sd->mru_bit_map, stream_num; + u32 shifted_values, unshifted_values; + u32 index_value_mask, shifted_values_mask; + int i; + bool is_read_ahead = false; + struct STREAM_DETECT *current_sd; + /* find possible stream */ + for (i = 0; i < MAX_STREAMS_TRACKED; ++i) { + stream_num = (*track_stream >> + (i * BITS_PER_INDEX_STREAM)) & + STREAM_MASK; + current_sd = ¤t_ld_sd->stream_track[stream_num]; + /* if we found a stream, update the raid + * context and also update the mruBitMap + */ + /* boundary condition */ + if ((current_sd->next_seq_lba) && + (io_info->ldStartBlock >= current_sd->next_seq_lba) && + (io_info->ldStartBlock <= (current_sd->next_seq_lba + 32)) && + (current_sd->is_read == io_info->isRead)) { + + if ((io_info->ldStartBlock != current_sd->next_seq_lba) && + ((!io_info->isRead) || (!is_read_ahead))) + /* + * Once the API availible we need to change this. + * At this point we are not allowing any gap + */ + continue; + + SET_STREAM_DETECTED(cmd->io_request->RaidContext.raid_context_g35); + current_sd->next_seq_lba = + io_info->ldStartBlock + io_info->numBlocks; + /* + * update the mruBitMap LRU + */ + shifted_values_mask = + (1 << i * BITS_PER_INDEX_STREAM) - 1; + shifted_values = ((*track_stream & shifted_values_mask) + << BITS_PER_INDEX_STREAM); + index_value_mask = + STREAM_MASK << i * BITS_PER_INDEX_STREAM; + unshifted_values = + *track_stream & ~(shifted_values_mask | + index_value_mask); + *track_stream = + unshifted_values | shifted_values | stream_num; + return; + } + } + /* + * if we did not find any stream, create a new one + * from the least recently used + */ + stream_num = (*track_stream >> + ((MAX_STREAMS_TRACKED - 1) * BITS_PER_INDEX_STREAM)) & + STREAM_MASK; + current_sd = ¤t_ld_sd->stream_track[stream_num]; + current_sd->is_read = io_info->isRead; + current_sd->next_seq_lba = io_info->ldStartBlock + io_info->numBlocks; + *track_stream = (((*track_stream & ZERO_LAST_STREAM) << 4) | stream_num); + return; +} + +/** + * megasas_set_raidflag_cpu_affinity - This function sets the cpu + * affinity (cpu of the controller) and raid_flags in the raid context + * based on IO type. + * + * @praid_context: IO RAID context + * @raid: LD raid map + * @fp_possible: Is fast path possible? + * @is_read: Is read IO? + * + */ +static void +megasas_set_raidflag_cpu_affinity(union RAID_CONTEXT_UNION *praid_context, + struct MR_LD_RAID *raid, bool fp_possible, + u8 is_read, u32 scsi_buff_len) +{ + u8 cpu_sel = MR_RAID_CTX_CPUSEL_0; + struct RAID_CONTEXT_G35 *rctx_g35; + + rctx_g35 = &praid_context->raid_context_g35; + if (fp_possible) { + if (is_read) { + if ((raid->cpuAffinity.pdRead.cpu0) && + (raid->cpuAffinity.pdRead.cpu1)) + cpu_sel = MR_RAID_CTX_CPUSEL_FCFS; + else if (raid->cpuAffinity.pdRead.cpu1) + cpu_sel = MR_RAID_CTX_CPUSEL_1; + } else { + if ((raid->cpuAffinity.pdWrite.cpu0) && + (raid->cpuAffinity.pdWrite.cpu1)) + cpu_sel = MR_RAID_CTX_CPUSEL_FCFS; + else if (raid->cpuAffinity.pdWrite.cpu1) + cpu_sel = MR_RAID_CTX_CPUSEL_1; + /* Fast path cache by pass capable R0/R1 VD */ + if ((raid->level <= 1) && + (raid->capability.fp_cache_bypass_capable)) { + rctx_g35->routing_flags |= + (1 << MR_RAID_CTX_ROUTINGFLAGS_SLD_SHIFT); + rctx_g35->raid_flags = + (MR_RAID_FLAGS_IO_SUB_TYPE_CACHE_BYPASS + << MR_RAID_CTX_RAID_FLAGS_IO_SUB_TYPE_SHIFT); + } + } + } else { + if (is_read) { + if ((raid->cpuAffinity.ldRead.cpu0) && + (raid->cpuAffinity.ldRead.cpu1)) + cpu_sel = MR_RAID_CTX_CPUSEL_FCFS; + else if (raid->cpuAffinity.ldRead.cpu1) + cpu_sel = MR_RAID_CTX_CPUSEL_1; + } else { + if ((raid->cpuAffinity.ldWrite.cpu0) && + (raid->cpuAffinity.ldWrite.cpu1)) + cpu_sel = MR_RAID_CTX_CPUSEL_FCFS; + else if (raid->cpuAffinity.ldWrite.cpu1) + cpu_sel = MR_RAID_CTX_CPUSEL_1; + + if (is_stream_detected(rctx_g35) && + (raid->level == 5) && + (raid->writeMode == MR_RL_WRITE_THROUGH_MODE) && + (cpu_sel == MR_RAID_CTX_CPUSEL_FCFS)) + cpu_sel = MR_RAID_CTX_CPUSEL_0; + } + } + + rctx_g35->routing_flags |= + (cpu_sel << MR_RAID_CTX_ROUTINGFLAGS_CPUSEL_SHIFT); + + /* Always give priority to MR_RAID_FLAGS_IO_SUB_TYPE_LDIO_BW_LIMIT + * vs MR_RAID_FLAGS_IO_SUB_TYPE_CACHE_BYPASS. + * IO Subtype is not bitmap. + */ + if ((raid->level == 1) && (!is_read)) { + if (scsi_buff_len > MR_LARGE_IO_MIN_SIZE) + praid_context->raid_context_g35.raid_flags = + (MR_RAID_FLAGS_IO_SUB_TYPE_LDIO_BW_LIMIT + << MR_RAID_CTX_RAID_FLAGS_IO_SUB_TYPE_SHIFT); + } +} + +/** * megasas_build_ldio_fusion - Prepares IOs to devices * @instance: Adapter soft state * @scp: SCSI command @@ -1701,29 +2195,36 @@ megasas_build_ldio_fusion(struct megasas_instance *instance, struct scsi_cmnd *scp, struct megasas_cmd_fusion *cmd) { - u8 fp_possible; + bool fp_possible; + u16 ld; u32 start_lba_lo, start_lba_hi, device_id, datalength = 0; + u32 scsi_buff_len; struct MPI2_RAID_SCSI_IO_REQUEST *io_request; union MEGASAS_REQUEST_DESCRIPTOR_UNION *req_desc; struct IO_REQUEST_INFO io_info; struct fusion_context *fusion; struct MR_DRV_RAID_MAP_ALL *local_map_ptr; u8 *raidLUN; + unsigned long spinlock_flags; + union RAID_CONTEXT_UNION *praid_context; + struct MR_LD_RAID *raid = NULL; + struct MR_PRIV_DEVICE *mrdev_priv; device_id = MEGASAS_DEV_INDEX(scp); fusion = instance->ctrl_context; io_request = cmd->io_request; - io_request->RaidContext.VirtualDiskTgtId = cpu_to_le16(device_id); - io_request->RaidContext.status = 0; - io_request->RaidContext.exStatus = 0; + io_request->RaidContext.raid_context.virtual_disk_tgt_id = + cpu_to_le16(device_id); + io_request->RaidContext.raid_context.status = 0; + io_request->RaidContext.raid_context.ex_status = 0; req_desc = (union MEGASAS_REQUEST_DESCRIPTOR_UNION *)cmd->request_desc; start_lba_lo = 0; start_lba_hi = 0; - fp_possible = 0; + fp_possible = false; /* * 6-byte READ(0x08) or WRITE(0x0A) cdb @@ -1779,22 +2280,27 @@ megasas_build_ldio_fusion(struct megasas_instance *instance, io_info.ldStartBlock = ((u64)start_lba_hi << 32) | start_lba_lo; io_info.numBlocks = datalength; io_info.ldTgtId = device_id; - io_request->DataLength = cpu_to_le32(scsi_bufflen(scp)); + io_info.r1_alt_dev_handle = MR_DEVHANDLE_INVALID; + scsi_buff_len = scsi_bufflen(scp); + io_request->DataLength = cpu_to_le32(scsi_buff_len); if (scp->sc_data_direction == PCI_DMA_FROMDEVICE) io_info.isRead = 1; local_map_ptr = fusion->ld_drv_map[(instance->map_id & 1)]; + ld = MR_TargetIdToLdGet(device_id, local_map_ptr); |