diff options
Diffstat (limited to 'drivers/nvme')
-rw-r--r-- | drivers/nvme/host/core.c | 326 | ||||
-rw-r--r-- | drivers/nvme/host/fc.c | 577 | ||||
-rw-r--r-- | drivers/nvme/host/fc.h | 227 | ||||
-rw-r--r-- | drivers/nvme/host/lightnvm.c | 7 | ||||
-rw-r--r-- | drivers/nvme/host/multipath.c | 16 | ||||
-rw-r--r-- | drivers/nvme/host/nvme.h | 28 | ||||
-rw-r--r-- | drivers/nvme/host/pci.c | 139 | ||||
-rw-r--r-- | drivers/nvme/host/rdma.c | 321 | ||||
-rw-r--r-- | drivers/nvme/host/tcp.c | 117 | ||||
-rw-r--r-- | drivers/nvme/target/Kconfig | 1 | ||||
-rw-r--r-- | drivers/nvme/target/admin-cmd.c | 42 | ||||
-rw-r--r-- | drivers/nvme/target/configfs.c | 272 | ||||
-rw-r--r-- | drivers/nvme/target/core.c | 166 | ||||
-rw-r--r-- | drivers/nvme/target/discovery.c | 8 | ||||
-rw-r--r-- | drivers/nvme/target/fabrics-cmd.c | 15 | ||||
-rw-r--r-- | drivers/nvme/target/fc.c | 805 | ||||
-rw-r--r-- | drivers/nvme/target/fcloop.c | 155 | ||||
-rw-r--r-- | drivers/nvme/target/io-cmd-bdev.c | 120 | ||||
-rw-r--r-- | drivers/nvme/target/io-cmd-file.c | 23 | ||||
-rw-r--r-- | drivers/nvme/target/nvmet.h | 36 | ||||
-rw-r--r-- | drivers/nvme/target/rdma.c | 416 | ||||
-rw-r--r-- | drivers/nvme/target/tcp.c | 107 | ||||
-rw-r--r-- | drivers/nvme/target/trace.h | 28 |
23 files changed, 2992 insertions, 960 deletions
diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c index f2adea96b04c..0585efa47d8f 100644 --- a/drivers/nvme/host/core.c +++ b/drivers/nvme/host/core.c @@ -19,7 +19,6 @@ #include <linux/pr.h> #include <linux/ptrace.h> #include <linux/nvme_ioctl.h> -#include <linux/t10-pi.h> #include <linux/pm_qos.h> #include <asm/unaligned.h> @@ -204,11 +203,6 @@ static void nvme_delete_ctrl_sync(struct nvme_ctrl *ctrl) nvme_put_ctrl(ctrl); } -static inline bool nvme_ns_has_pi(struct nvme_ns *ns) -{ - return ns->pi_type && ns->ms == sizeof(struct t10_pi_tuple); -} - static blk_status_t nvme_error_status(u16 status) { switch (status & 0x7ff) { @@ -310,7 +304,7 @@ bool nvme_cancel_request(struct request *req, void *data, bool reserved) return true; nvme_req(req)->status = NVME_SC_HOST_ABORTED_CMD; - blk_mq_complete_request(req); + blk_mq_force_complete_rq(req); return true; } EXPORT_SYMBOL_GPL(nvme_cancel_request); @@ -433,7 +427,6 @@ static void nvme_free_ns_head(struct kref *ref) nvme_mpath_remove_disk(head); ida_simple_remove(&head->subsys->ns_ida, head->instance); - list_del_init(&head->entry); cleanup_srcu_struct(&head->srcu); nvme_put_subsystem(head->subsys); kfree(head); @@ -530,7 +523,7 @@ static int nvme_get_stream_params(struct nvme_ctrl *ctrl, c.directive.opcode = nvme_admin_directive_recv; c.directive.nsid = cpu_to_le32(nsid); - c.directive.numd = cpu_to_le32((sizeof(*s) >> 2) - 1); + c.directive.numd = cpu_to_le32(nvme_bytes_to_numd(sizeof(*s))); c.directive.doper = NVME_DIR_RCV_ST_OP_PARAM; c.directive.dtype = NVME_DIR_STREAMS; @@ -553,19 +546,22 @@ static int nvme_configure_directives(struct nvme_ctrl *ctrl) ret = nvme_get_stream_params(ctrl, &s, NVME_NSID_ALL); if (ret) - return ret; + goto out_disable_stream; ctrl->nssa = le16_to_cpu(s.nssa); if (ctrl->nssa < BLK_MAX_WRITE_HINTS - 1) { dev_info(ctrl->device, "too few streams (%u) available\n", ctrl->nssa); - nvme_disable_streams(ctrl); - return 0; + goto out_disable_stream; } ctrl->nr_streams = min_t(unsigned, ctrl->nssa, BLK_MAX_WRITE_HINTS - 1); dev_info(ctrl->device, "Using %u streams\n", ctrl->nr_streams); return 0; + +out_disable_stream: + nvme_disable_streams(ctrl); + return ret; } /* @@ -1027,6 +1023,19 @@ void nvme_stop_keep_alive(struct nvme_ctrl *ctrl) } EXPORT_SYMBOL_GPL(nvme_stop_keep_alive); +/* + * In NVMe 1.0 the CNS field was just a binary controller or namespace + * flag, thus sending any new CNS opcodes has a big chance of not working. + * Qemu unfortunately had that bug after reporting a 1.1 version compliance + * (but not for any later version). + */ +static bool nvme_ctrl_limited_cns(struct nvme_ctrl *ctrl) +{ + if (ctrl->quirks & NVME_QUIRK_IDENTIFY_CNS) + return ctrl->vs < NVME_VS(1, 2, 0); + return ctrl->vs < NVME_VS(1, 1, 0); +} + static int nvme_identify_ctrl(struct nvme_ctrl *dev, struct nvme_id_ctrl **id) { struct nvme_command c = { }; @@ -1110,7 +1119,7 @@ static int nvme_identify_ns_descs(struct nvme_ctrl *ctrl, unsigned nsid, * Don't treat an error as fatal, as we potentially already * have a NGUID or EUI-64. */ - if (status > 0) + if (status > 0 && !(status & NVME_SC_DNR)) status = 0; goto free_data; } @@ -1290,7 +1299,7 @@ static int nvme_submit_io(struct nvme_ns *ns, struct nvme_user_io __user *uio) meta_len = (io.nblocks + 1) * ns->ms; metadata = nvme_to_user_ptr(io.metadata); - if (ns->ext) { + if (ns->features & NVME_NS_EXT_LBAS) { length += meta_len; meta_len = 0; } else if (meta_len) { @@ -1392,8 +1401,10 @@ static void nvme_passthru_end(struct nvme_ctrl *ctrl, u32 effects) } if (effects & NVME_CMD_EFFECTS_CCC) nvme_init_identify(ctrl); - if (effects & (NVME_CMD_EFFECTS_NIC | NVME_CMD_EFFECTS_NCC)) + if (effects & (NVME_CMD_EFFECTS_NIC | NVME_CMD_EFFECTS_NCC)) { nvme_queue_scan(ctrl); + flush_work(&ctrl->scan_work); + } } static int nvme_user_cmd(struct nvme_ctrl *ctrl, struct nvme_ns *ns, @@ -1682,7 +1693,8 @@ static int nvme_getgeo(struct block_device *bdev, struct hd_geometry *geo) } #ifdef CONFIG_BLK_DEV_INTEGRITY -static void nvme_init_integrity(struct gendisk *disk, u16 ms, u8 pi_type) +static void nvme_init_integrity(struct gendisk *disk, u16 ms, u8 pi_type, + u32 max_integrity_segments) { struct blk_integrity integrity; @@ -1705,20 +1717,15 @@ static void nvme_init_integrity(struct gendisk *disk, u16 ms, u8 pi_type) } integrity.tuple_size = ms; blk_integrity_register(disk, &integrity); - blk_queue_max_integrity_segments(disk->queue, 1); + blk_queue_max_integrity_segments(disk->queue, max_integrity_segments); } #else -static void nvme_init_integrity(struct gendisk *disk, u16 ms, u8 pi_type) +static void nvme_init_integrity(struct gendisk *disk, u16 ms, u8 pi_type, + u32 max_integrity_segments) { } #endif /* CONFIG_BLK_DEV_INTEGRITY */ -static void nvme_set_chunk_size(struct nvme_ns *ns) -{ - u32 chunk_size = nvme_lba_to_sect(ns, ns->noiob); - blk_queue_chunk_sectors(ns->queue, rounddown_pow_of_two(chunk_size)); -} - static void nvme_config_discard(struct gendisk *disk, struct nvme_ns *ns) { struct nvme_ctrl *ctrl = ns->ctrl; @@ -1804,12 +1811,37 @@ static bool nvme_ns_ids_equal(struct nvme_ns_ids *a, struct nvme_ns_ids *b) memcmp(&a->eui64, &b->eui64, sizeof(a->eui64)) == 0; } +static int nvme_setup_streams_ns(struct nvme_ctrl *ctrl, struct nvme_ns *ns, + u32 *phys_bs, u32 *io_opt) +{ + struct streams_directive_params s; + int ret; + + if (!ctrl->nr_streams) + return 0; + + ret = nvme_get_stream_params(ctrl, &s, ns->head->ns_id); + if (ret) + return ret; + + ns->sws = le32_to_cpu(s.sws); + ns->sgs = le16_to_cpu(s.sgs); + + if (ns->sws) { + *phys_bs = ns->sws * (1 << ns->lba_shift); + if (ns->sgs) + *io_opt = *phys_bs * ns->sgs; + } + + return 0; +} + static void nvme_update_disk_info(struct gendisk *disk, struct nvme_ns *ns, struct nvme_id_ns *id) { sector_t capacity = nvme_lba_to_sect(ns, le64_to_cpu(id->nsze)); unsigned short bs = 1 << ns->lba_shift; - u32 atomic_bs, phys_bs, io_opt; + u32 atomic_bs, phys_bs, io_opt = 0; if (ns->lba_shift > PAGE_SHIFT) { /* unsupported block size, set capacity to 0 later */ @@ -1818,26 +1850,25 @@ static void nvme_update_disk_info(struct gendisk *disk, blk_mq_freeze_queue(disk->queue); blk_integrity_unregister(disk); + atomic_bs = phys_bs = bs; + nvme_setup_streams_ns(ns->ctrl, ns, &phys_bs, &io_opt); if (id->nabo == 0) { /* * Bit 1 indicates whether NAWUPF is defined for this namespace * and whether it should be used instead of AWUPF. If NAWUPF == * 0 then AWUPF must be used instead. */ - if (id->nsfeat & (1 << 1) && id->nawupf) + if (id->nsfeat & NVME_NS_FEAT_ATOMICS && id->nawupf) atomic_bs = (1 + le16_to_cpu(id->nawupf)) * bs; else atomic_bs = (1 + ns->ctrl->subsys->awupf) * bs; - } else { - atomic_bs = bs; } - phys_bs = bs; - io_opt = bs; - if (id->nsfeat & (1 << 4)) { + + if (id->nsfeat & NVME_NS_FEAT_IO_OPT) { /* NPWG = Namespace Preferred Write Granularity */ - phys_bs *= 1 + le16_to_cpu(id->npwg); + phys_bs = bs * (1 + le16_to_cpu(id->npwg)); /* NOWS = Namespace Optimal Write Size */ - io_opt *= 1 + le16_to_cpu(id->nows); + io_opt = bs * (1 + le16_to_cpu(id->nows)); } blk_queue_logical_block_size(disk->queue, bs); @@ -1850,19 +1881,34 @@ static void nvme_update_disk_info(struct gendisk *disk, blk_queue_io_min(disk->queue, phys_bs); blk_queue_io_opt(disk->queue, io_opt); - if (ns->ms && !ns->ext && - (ns->ctrl->ops->flags & NVME_F_METADATA_SUPPORTED)) - nvme_init_integrity(disk, ns->ms, ns->pi_type); - if ((ns->ms && !nvme_ns_has_pi(ns) && !blk_get_integrity(disk)) || - ns->lba_shift > PAGE_SHIFT) + /* + * The block layer can't support LBA sizes larger than the page size + * yet, so catch this early and don't allow block I/O. + */ + if (ns->lba_shift > PAGE_SHIFT) capacity = 0; + /* + * Register a metadata profile for PI, or the plain non-integrity NVMe + * metadata masquerading as Type 0 if supported, otherwise reject block + * I/O to namespaces with metadata except when the namespace supports + * PI, as it can strip/insert in that case. + */ + if (ns->ms) { + if (IS_ENABLED(CONFIG_BLK_DEV_INTEGRITY) && + (ns->features & NVME_NS_METADATA_SUPPORTED)) + nvme_init_integrity(disk, ns->ms, ns->pi_type, + ns->ctrl->max_integrity_segments); + else if (!nvme_ns_has_pi(ns)) + capacity = 0; + } + set_capacity_revalidate_and_notify(disk, capacity, false); nvme_config_discard(disk, ns); nvme_config_write_zeroes(disk, ns); - if (id->nsattr & (1 << 0)) + if (id->nsattr & NVME_NS_ATTR_RO) set_disk_ro(disk, true); else set_disk_ro(disk, false); @@ -1870,9 +1916,11 @@ static void nvme_update_disk_info(struct gendisk *disk, blk_mq_unfreeze_queue(disk->queue); } -static void __nvme_revalidate_disk(struct gendisk *disk, struct nvme_id_ns *id) +static int __nvme_revalidate_disk(struct gendisk *disk, struct nvme_id_ns *id) { struct nvme_ns *ns = disk->private_data; + struct nvme_ctrl *ctrl = ns->ctrl; + u32 iob; /* * If identify namespace failed, use default 512 byte block size so @@ -1881,32 +1929,55 @@ static void __nvme_revalidate_disk(struct gendisk *disk, struct nvme_id_ns *id) ns->lba_shift = id->lbaf[id->flbas & NVME_NS_FLBAS_LBA_MASK].ds; if (ns->lba_shift == 0) ns->lba_shift = 9; - ns->noiob = le16_to_cpu(id->noiob); + + if ((ctrl->quirks & NVME_QUIRK_STRIPE_SIZE) && + is_power_of_2(ctrl->max_hw_sectors)) + iob = ctrl->max_hw_sectors; + else + iob = nvme_lba_to_sect(ns, le16_to_cpu(id->noiob)); + + ns->features = 0; ns->ms = le16_to_cpu(id->lbaf[id->flbas & NVME_NS_FLBAS_LBA_MASK].ms); - ns->ext = ns->ms && (id->flbas & NVME_NS_FLBAS_META_EXT); /* the PI implementation requires metadata equal t10 pi tuple size */ if (ns->ms == sizeof(struct t10_pi_tuple)) ns->pi_type = id->dps & NVME_NS_DPS_PI_MASK; else ns->pi_type = 0; - if (ns->noiob) - nvme_set_chunk_size(ns); + if (ns->ms) { + /* + * For PCIe only the separate metadata pointer is supported, + * as the block layer supplies metadata in a separate bio_vec + * chain. For Fabrics, only metadata as part of extended data + * LBA is supported on the wire per the Fabrics specification, + * but the HBA/HCA will do the remapping from the separate + * metadata buffers for us. + */ + if (id->flbas & NVME_NS_FLBAS_META_EXT) { + ns->features |= NVME_NS_EXT_LBAS; + if ((ctrl->ops->flags & NVME_F_FABRICS) && + (ctrl->ops->flags & NVME_F_METADATA_SUPPORTED) && + ctrl->max_integrity_segments) + ns->features |= NVME_NS_METADATA_SUPPORTED; + } else { + if (WARN_ON_ONCE(ctrl->ops->flags & NVME_F_FABRICS)) + return -EINVAL; + if (ctrl->ops->flags & NVME_F_METADATA_SUPPORTED) + ns->features |= NVME_NS_METADATA_SUPPORTED; + } + } + + if (iob) + blk_queue_chunk_sectors(ns->queue, rounddown_pow_of_two(iob)); nvme_update_disk_info(disk, ns, id); #ifdef CONFIG_NVME_MULTIPATH if (ns->head->disk) { nvme_update_disk_info(ns->head->disk, ns, id); blk_queue_stack_limits(ns->head->disk->queue, ns->queue); - if (bdi_cap_stable_pages_required(ns->queue->backing_dev_info)) { - struct backing_dev_info *info = - ns->head->disk->queue->backing_dev_info; - - info->capabilities |= BDI_CAP_STABLE_WRITES; - } - revalidate_disk(ns->head->disk); } #endif + return 0; } static int nvme_revalidate_disk(struct gendisk *disk) @@ -1931,7 +2002,6 @@ static int nvme_revalidate_disk(struct gendisk *disk) goto free_id; } - __nvme_revalidate_disk(disk, id); ret = nvme_report_ns_ids(ctrl, ns->head->ns_id, id, &ids); if (ret) goto free_id; @@ -1940,8 +2010,10 @@ static int nvme_revalidate_disk(struct gendisk *disk) dev_err(ctrl->device, "identifiers changed for nsid %d\n", ns->head->ns_id); ret = -ENODEV; + goto free_id; } + ret = __nvme_revalidate_disk(disk, id); free_id: kfree(id); out: @@ -2249,10 +2321,8 @@ static void nvme_set_queue_limits(struct nvme_ctrl *ctrl, blk_queue_max_hw_sectors(q, ctrl->max_hw_sectors); blk_queue_max_segments(q, min_t(u32, max_segments, USHRT_MAX)); } - if ((ctrl->quirks & NVME_QUIRK_STRIPE_SIZE) && - is_power_of_2(ctrl->max_hw_sectors)) - blk_queue_chunk_sectors(q, ctrl->max_hw_sectors); blk_queue_virt_boundary(q, ctrl->page_size - 1); + blk_queue_dma_alignment(q, 7); if (ctrl->vwc & NVME_CTRL_VWC_PRESENT) vwc = true; blk_queue_write_cache(q, vwc, vwc); @@ -2655,7 +2725,7 @@ static bool nvme_validate_cntlid(struct nvme_subsystem *subsys, return false; } - if ((id->cmic & (1 << 1)) || + if ((id->cmic & NVME_CTRL_CMIC_MULTI_CTRL) || (ctrl->opts && ctrl->opts->discovery_nqn)) continue; @@ -2746,7 +2816,7 @@ int nvme_get_log(struct nvme_ctrl *ctrl, u32 nsid, u8 log_page, u8 lsp, void *log, size_t size, u64 offset) { struct nvme_command c = { }; - unsigned long dwlen = size / 4 - 1; + u32 dwlen = nvme_bytes_to_numd(size); c.get_log_page.opcode = nvme_admin_get_log_page; c.get_log_page.nsid = cpu_to_le32(nsid); @@ -3401,7 +3471,6 @@ static int __nvme_check_ids(struct nvme_subsystem *subsys, list_for_each_entry(h, &subsys->nsheads, entry) { if (nvme_ns_ids_valid(&new->ids) && - !list_empty(&h->list) && nvme_ns_ids_equal(&new->ids, &h->ids)) return -EINVAL; } @@ -3410,8 +3479,7 @@ static int __nvme_check_ids(struct nvme_subsystem *subsys, } static struct nvme_ns_head *nvme_alloc_ns_head(struct nvme_ctrl *ctrl, - unsigned nsid, struct nvme_id_ns *id, - struct nvme_ns_ids *ids) + unsigned nsid, struct nvme_ns_ids *ids) { struct nvme_ns_head *head; size_t size = sizeof(*head); @@ -3469,42 +3537,51 @@ static int nvme_init_ns_head(struct nvme_ns *ns, unsigned nsid, struct nvme_id_ns *id) { struct nvme_ctrl *ctrl = ns->ctrl; - bool is_shared = id->nmic & (1 << 0); + bool is_shared = id->nmic & NVME_NS_NMIC_SHARED; struct nvme_ns_head *head = NULL; struct nvme_ns_ids ids; int ret = 0; ret = nvme_report_ns_ids(ctrl, nsid, id, &ids); - if (ret) - goto out; + if (ret) { + if (ret < 0) + return ret; + return blk_status_to_errno(nvme_error_status(ret)); + } mutex_lock(&ctrl->subsys->lock); - if (is_shared) - head = nvme_find_ns_head(ctrl->subsys, nsid); + head = nvme_find_ns_head(ctrl->subsys, nsid); if (!head) { - head = nvme_alloc_ns_head(ctrl, nsid, id, &ids); + head = nvme_alloc_ns_head(ctrl, nsid, &ids); if (IS_ERR(head)) { ret = PTR_ERR(head); goto out_unlock; } + head->shared = is_shared; } else { + ret = -EINVAL; + if (!is_shared || !head->shared) { + dev_err(ctrl->device, + "Duplicate unshared namespace %d\n", nsid); + goto out_put_ns_head; + } if (!nvme_ns_ids_equal(&head->ids, &ids)) { dev_err(ctrl->device, "IDs don't match for shared namespace %d\n", nsid); - ret = -EINVAL; - goto out_unlock; + goto out_put_ns_head; } } list_add_tail(&ns->siblings, &head->list); ns->head = head; + mutex_unlock(&ctrl->subsys->lock); + return 0; +out_put_ns_head: + nvme_put_ns_head(head); out_unlock: mutex_unlock(&ctrl->subsys->lock); -out: - if (ret > 0) - ret = blk_status_to_errno(nvme_error_status(ret)); return ret; } @@ -3535,32 +3612,6 @@ static struct nvme_ns *nvme_find_get_ns(struct nvme_ctrl *ctrl, unsigned nsid) return ret; } -static int nvme_setup_streams_ns(struct nvme_ctrl *ctrl, struct nvme_ns *ns) -{ - struct streams_directive_params s; - int ret; - - if (!ctrl->nr_streams) - return 0; - - ret = nvme_get_stream_params(ctrl, &s, ns->head->ns_id); - if (ret) - return ret; - - ns->sws = le32_to_cpu(s.sws); - ns->sgs = le16_to_cpu(s.sgs); - - if (ns->sws) { - unsigned int bs = 1 << ns->lba_shift; - - blk_queue_io_min(ns->queue, bs * ns->sws); - if (ns->sgs) - blk_queue_io_opt(ns->queue, bs * ns->sws * ns->sgs); - } - - return 0; -} - static void nvme_alloc_ns(struct nvme_ctrl *ctrl, unsigned nsid) { struct nvme_ns *ns; @@ -3604,7 +3655,6 @@ static void nvme_alloc_ns(struct nvme_ctrl *ctrl, unsigned nsid) ret = nvme_init_ns_head(ns, nsid, id); if (ret) goto out_free_id; - nvme_setup_streams_ns(ctrl, ns); nvme_set_disk_name(disk_name, ns, ctrl, &flags); disk = alloc_disk_node(0, node); @@ -3618,7 +3668,8 @@ static void nvme_alloc_ns(struct nvme_ctrl *ctrl, unsigned nsid) memcpy(disk->disk_name, disk_name, DISK_NAME_LEN); ns->disk = disk; - __nvme_revalidate_disk(disk, id); + if (__nvme_revalidate_disk(disk, id)) + goto out_free_disk; if ((ctrl->quirks & NVME_QUIRK_LIGHTNVM) && id->vs[0] == 0x1) { ret = nvme_nvm_register(ns, disk_name, node); @@ -3645,9 +3696,13 @@ static void nvme_alloc_ns(struct nvme_ctrl *ctrl, unsigned nsid) /* prevent double queue cleanup */ ns->disk->queue = NULL; put_disk(ns->disk); + out_free_disk: + del_gendisk(ns->disk); out_unlink_ns: mutex_lock(&ctrl->subsys->lock); list_del_rcu(&ns->siblings); + if (list_empty(&ns->head->list)) + list_del_init(&ns->head->entry); mutex_unlock(&ctrl->subsys->lock); nvme_put_ns_head(ns->head); out_free_id: @@ -3667,7 +3722,10 @@ static void nvme_ns_remove(struct nvme_ns *ns) mutex_lock(&ns->ctrl->subsys->lock); list_del_rcu(&ns->siblings); + if (list_empty(&ns->head->list)) + list_del_init(&ns->head->entry); mutex_unlock(&ns->ctrl->subsys->lock); + synchronize_rcu(); /* guarantee not available in head->list */ nvme_mpath_clear_current_path(ns); synchronize_srcu(&ns->head->srcu); /* wait for concurrent submissions */ @@ -3687,6 +3745,16 @@ static void nvme_ns_remove(struct nvme_ns *ns) nvme_put_ns(ns); } +static void nvme_ns_remove_by_nsid(struct nvme_ctrl *ctrl, u32 nsid) +{ + struct nvme_ns *ns = nvme_find_get_ns(ctrl, nsid); + + if (ns) { + nvme_ns_remove(ns); + nvme_put_ns(ns); + } +} + static void nvme_validate_ns(struct nvme_ctrl *ctrl, unsigned nsid) { struct nvme_ns *ns; @@ -3718,39 +3786,34 @@ static void nvme_remove_invalid_namespaces(struct nvme_ctrl *ctrl, } -static int nvme_scan_ns_list(struct nvme_ctrl *ctrl, unsigned nn) +static int nvme_scan_ns_list(struct nvme_ctrl *ctrl) { - struct nvme_ns *ns; + const int nr_entries = NVME_IDENTIFY_DATA_SIZE / sizeof(__le32); __le32 *ns_list; - unsigned i, j, nsid, prev = 0; - unsigned num_lists = DIV_ROUND_UP_ULL((u64)nn, 1024); - int ret = 0; + u32 prev = 0; + int ret = 0, i; + + if (nvme_ctrl_limited_cns(ctrl)) + return -EOPNOTSUPP; ns_list = kzalloc(NVME_IDENTIFY_DATA_SIZE, GFP_KERNEL); if (!ns_list) return -ENOMEM; - for (i = 0; i < num_lists; i++) { + for (;;) { ret = nvme_identify_ns_list(ctrl, prev, ns_list); if (ret) goto free; - for (j = 0; j < min(nn, 1024U); j++) { - nsid = le32_to_cpu(ns_list[j]); - if (!nsid) - goto out; + for (i = 0; i < nr_entries; i++) { + u32 nsid = le32_to_cpu(ns_list[i]); + if (!nsid) /* end of the list? */ + goto out; nvme_validate_ns(ctrl, nsid); - - while (++prev < nsid) { - ns = nvme_find_get_ns(ctrl, prev); - if (ns) { - nvme_ns_remove(ns); - nvme_put_ns(ns); - } - } + while (++prev < nsid) + nvme_ns_remove_by_nsid(ctrl, prev); } - nn -= j; } out: nvme_remove_invalid_namespaces(ctrl, prev); @@ -3759,9 +3822,15 @@ static int nvme_scan_ns_list(struct nvme_ctrl *ctrl, unsigned nn) return ret; } -static void nvme_scan_ns_sequential(struct nvme_ctrl *ctrl, unsigned nn) +static void nvme_scan_ns_sequential(struct nvme_ctrl *ctrl) { - unsigned i; + struct nvme_id_ctrl *id; + u32 nn, i; + + if (nvme_identify_ctrl(ctrl, &id)) + return; + nn = le32_to_cpu(id->nn); + kfree(id); for (i = 1; i <= nn; i++) nvme_validate_ns(ctrl, i); @@ -3798,8 +3867,6 @@ static void nvme_scan_work(struct work_struct *work) { struct nvme_ctrl *ctrl = container_of(work, struct nvme_ctrl, scan_work); - struct nvme_id_ctrl *id; - unsigned nn; /* No tagset on a live ctrl means IO queues could not created */ if (ctrl->state != NVME_CTRL_LIVE || !ctrl->tagset) @@ -3810,20 +3877,11 @@ static void nvme_scan_work(struct work_struct *work) nvme_clear_changed_ns_log(ctrl); } - if (nvme_identify_ctrl(ctrl, &id)) - return; - mutex_lock(&ctrl->scan_lock); - nn = le32_to_cpu(id->nn); - if (ctrl->vs >= NVME_VS(1, 1, 0) && - !(ctrl->quirks & NVME_QUIRK_IDENTIFY_CNS)) { - if (!nvme_scan_ns_list(ctrl, nn)) - goto out_free_id; - } - nvme_scan_ns_sequential(ctrl, nn); -out_free_id: + if (nvme_scan_ns_list(ctrl) != 0) + nvme_scan_ns_sequential(ctrl); mutex_unlock(&ctrl->scan_lock); - kfree(id); + down_write(&ctrl->namespaces_rwsem); list_sort(NULL, &ctrl->namespaces, ns_cmp); up_write(&ctrl->namespaces_rwsem); diff --git a/drivers/nvme/host/fc.c b/drivers/nvme/host/fc.c index 7dfc4a2ecf1e..cb0007592c12 100644 --- a/drivers/nvme/host/fc.c +++ b/drivers/nvme/host/fc.c @@ -14,6 +14,7 @@ #include "fabrics.h" #include <linux/nvme-fc-driver.h> #include <linux/nvme-fc.h> +#include "fc.h" #include <scsi/scsi_transport_fc.h> /* *************************** Data Structures/Defines ****************** */ @@ -61,6 +62,17 @@ struct nvmefc_ls_req_op { bool req_queued; }; +struct nvmefc_ls_rcv_op { + struct nvme_fc_rport *rport; + struct nvmefc_ls_rsp *lsrsp; + union nvmefc_ls_requests *rqstbuf; + union nvmefc_ls_responses *rspbuf; + u16 rqstdatalen; + bool handled; + dma_addr_t rspdma; + struct list_head lsrcv_list; /* rport->ls_rcv_list */ +} __aligned(sizeof(u64)); /* alignment for other things alloc'd with */ + enum nvme_fcpop_state { FCPOP_STATE_UNINIT = 0, FCPOP_STATE_IDLE = 1, @@ -96,7 +108,7 @@ struct nvme_fc_fcp_op { struct nvme_fcp_op_w_sgl { struct nvme_fc_fcp_op op; struct scatterlist sgl[NVME_INLINE_SG_CNT]; - uint8_t priv[0]; + uint8_t priv[]; }; struct nvme_fc_lport { @@ -117,6 +129,7 @@ struct nvme_fc_rport { struct list_head endp_list; /* for lport->endp_list */ struct list_head ctrl_list; struct list_head ls_req_list; + struct list_head ls_rcv_list; struct list_head disc_list; struct device *dev; /* physical device for dma */ struct nvme_fc_lport *lport; @@ -124,11 +137,12 @@ struct nvme_fc_rport { struct kref ref; atomic_t act_ctrl_cnt; unsigned long dev_loss_end; + struct work_struct lsrcv_work; } __aligned(sizeof(u64)); /* alignment for other things alloc'd with */ -enum nvme_fcctrl_flags { - FCCTRL_TERMIO = (1 << 0), -}; +/* fc_ctrl flags values - specified as bit positions */ +#define ASSOC_ACTIVE 0 +#define FCCTRL_TERMIO 1 struct nvme_fc_ctrl { spinlock_t lock; @@ -139,9 +153,9 @@ struct nvme_fc_ctrl { u32 cnum; bool ioq_live; - bool assoc_active; atomic_t err_work_active; u64 association_id; + struct nvmefc_ls_rcv_op *rcv_disconn; struct list_head ctrl_list; /* rport->ctrl_list */ @@ -152,7 +166,7 @@ struct nvme_fc_ctrl { struct work_struct err_work; struct kref ref; - u32 flags; + unsigned long flags; u32 iocnt; wait_queue_head_t ioabort_wait; @@ -219,6 +233,9 @@ static struct device *fc_udev_device; static void __nvme_fc_delete_hw_queue(struct nvme_fc_ctrl *, struct nvme_fc_queue *, unsigned int); +static void nvme_fc_handle_ls_rqst_work(struct work_struct *work); + + static void nvme_fc_free_lport(struct kref *ref) { @@ -394,7 +411,10 @@ nvme_fc_register_localport(struct nvme_fc_port_info *pinfo, newrec->ops = template; newrec->dev = dev; ida_init(&newrec->endp_cnt); - newrec->localport.private = &newrec[1]; + if (template->local_priv_sz) + newrec->localport.private = &newrec[1]; + else + newrec->localport.private = NULL; newrec->localport.node_name = pinfo->node_name; newrec->localport.port_name = pinfo->port_name; newrec->localport.port_role = pinfo->port_role; @@ -701,9 +721,13 @@ nvme_fc_register_remoteport(struct nvme_fc_local_port *localport, atomic_set(&newrec->act_ctrl_cnt, 0); spin_lock_init(&newrec->lock); newrec->remoteport.localport = &lport->localport; + INIT_LIST_HEAD(&newrec->ls_rcv_list); newrec->dev = lport->dev; newrec->lport = lport; - newrec->remoteport.private = &newrec[1]; + if (lport->ops->remote_priv_sz) + newrec->remoteport.private = &newrec[1]; + else + newrec->remoteport.private = NULL; newrec->remoteport.port_role = pinfo->port_role; newrec->remoteport.node_name = pinfo->node_name; newrec->remoteport.port_name = pinfo->port_name; @@ -711,6 +735,7 @@ nvme_fc_register_remoteport(struct nvme_fc_local_port *localport, newrec->remoteport.port_state = FC_OBJSTATE_ONLINE; newrec->remoteport.port_num = idx; __nvme_fc_set_dev_loss_tmo(newrec, pinfo); + INIT_WORK(&newrec->lsrcv_work, nvme_fc_handle_ls_rqst_work); spin_lock_irqsave(&nvme_fc_lock, flags); list_add_tail(&newrec->endp_list, &lport->endp_list); @@ -1000,6 +1025,7 @@ fc_dma_unmap_sg(struct device *dev, struct scatterlist *sg, int nents, static void nvme_fc_ctrl_put(struct nvme_fc_ctrl *); static int nvme_fc_ctrl_get(struct nvme_fc_ctrl *); +static void nvme_fc_error_recovery(struct nvme_fc_ctrl *ctrl, char *errmsg); static void __nvme_fc_finish_ls_req(struct nvmefc_ls_req_op *lsop) @@ -1140,41 +1166,6 @@ nvme_fc_send_ls_req_async(struct nvme_fc_rport *rport, return __nvme_fc_send_ls_req(rport, lsop, done); } -/* Validation Error indexes into the string table below */ -enum { - VERR_NO_ERROR = 0, - VERR_LSACC = 1, - VERR_LSDESC_RQST = 2, - VERR_LSDESC_RQST_LEN = 3, - VERR_ASSOC_ID = 4, - VERR_ASSOC_ID_LEN = 5, - VERR_CONN_ID = 6, - VERR_CONN_ID_LEN = 7, - VERR_CR_ASSOC = 8, - VERR_CR_ASSOC_ACC_LEN = 9, - VERR_CR_CONN = 10, - VERR_CR_CONN_ACC_LEN = 11, - VERR_DISCONN = 12, - VERR_DISCONN_ACC_LEN = 13, -}; - -static char *validation_errors[] = { - "OK", - "Not LS_ACC", - "Not LSDESC_RQST", - "Bad LSDESC_RQST Length", - "Not Association ID", - "Bad Association ID Length", - "Not Connection ID", - "Bad Connection ID Length", - "Not CR_ASSOC Rqst", - "Bad CR_ASSOC ACC Length", - "Not CR_CONN Rqst", - "Bad CR_CONN ACC Length", - "Not Disconnect Rqst", - "Bad Disconnect ACC Length", -}; - static int nvme_fc_connect_admin_queue(struct nvme_fc_ctrl *ctrl, struct nvme_fc_queue *queue, u16 qsize, u16 ersp_ratio) @@ -1183,21 +1174,27 @@ nvme_fc_connect_admin_queue(struct nvme_fc_ctrl *ctrl, struct nvmefc_ls_req *lsreq; struct fcnvme_ls_cr_assoc_rqst *assoc_rqst; struct fcnvme_ls_cr_assoc_acc *assoc_acc; + unsigned long flags; int ret, fcret = 0; lsop = kzalloc((sizeof(*lsop) + - ctrl->lport->ops->lsrqst_priv_sz + - sizeof(*assoc_rqst) + sizeof(*assoc_acc)), GFP_KERNEL); + sizeof(*assoc_rqst) + sizeof(*assoc_acc) + + ctrl->lport->ops->lsrqst_priv_sz), GFP_KERNEL); if (!lsop) { + dev_info(ctrl->ctrl.device, + "NVME-FC{%d}: send Create Association failed: ENOMEM\n", + ctrl->cnum); ret = -ENOMEM; goto out_no_memory; } - lsreq = &lsop->ls_req; - lsreq->private = (void *)&lsop[1]; - assoc_rqst = (struct fcnvme_ls_cr_assoc_rqst *) - (lsreq->private + ctrl->lport->ops->lsrqst_priv_sz); + assoc_rqst = (struct fcnvme_ls_cr_assoc_rqst *)&lsop[1]; assoc_acc = (struct fcnvme_ls_cr_assoc_acc *)&assoc_rqst[1]; + lsreq = &lsop->ls_req; + if (ctrl->lport->ops->lsrqst_priv_sz) + lsreq->private = &assoc_acc[1]; + else + lsreq->private = NULL; assoc_rqst->w0.ls_cmd = FCNVME_LS_CREATE_ASSOCIATION; assoc_rqst->desc_list_len = @@ -1267,11 +1264,13 @@ nvme_fc_connect_admin_queue(struct nvme_fc_ctrl *ctrl, "q %d Create Association LS failed: %s\n", queue->qnum, validation_errors[fcret]); } else { + spin_lock_irqsave(&ctrl->lock, flags); ctrl->association_id = be64_to_cpu(assoc_acc->associd.association_id); queue->connection_id = be64_to_cpu(assoc_acc->connectid.connection_id); set_bit(NVME_FC_Q_CONNECTED, &queue->flags); + spin_unlock_irqrestore(&ctrl->lock, flags); } out_free_buffer: @@ -1295,18 +1294,23 @@ nvme_fc_connect_queue(struct nvme_fc_ctrl *ctrl, struct nvme_fc_queue *queue, int ret, fcret = 0; lsop = kzalloc((sizeof(*lsop) + - ctrl->lport->ops->lsrqst_priv_sz + - sizeof(*conn_rqst) + sizeof(*conn_acc)), GFP_KERNEL); + sizeof(*conn_rqst) + sizeof(*conn_acc) + + ctrl->lport->ops->lsrqst_priv_sz), GFP_KERNEL); if (!lsop) { + dev_info(ctrl->ctrl.device, + "NVME-FC{%d}: send Create Connection failed: ENOMEM\n", + ctrl->cnum); ret = -ENOMEM; goto out_no_memory; } - lsreq = &lsop->ls_req; - lsreq->private = (void *)&lsop[1]; - conn_rqst = (struct fcnvme_ls_cr_conn_rqst *) - (lsreq->private + ctrl->lport->ops->lsrqst_priv_sz); + conn_rqst = (struct fcnvme_ls_cr_conn_rqst *)&lsop[1]; conn_acc = (struct fcnvme_ls_cr_conn_acc *)&conn_rqst[1]; + lsreq = &lsop->ls_req; + if (ctrl->lport->ops->lsrqst_priv_sz) + lsreq->private = (void *)&conn_acc[1]; + else + lsreq->private = NULL; conn_rqst->w0.ls_cmd = FCNVME_LS_CREATE_CONNECTION; conn_rqst->desc_list_len = cpu_to_be32( @@ -1420,54 +1424,385 @@ nvme_fc_xmt_disconnect_assoc(struct nvme_fc_ctrl *ctrl) int ret; lsop = kzalloc((sizeof(*lsop) + - ctrl->lport->ops->lsrqst_priv_sz + - sizeof(*discon_rqst) + sizeof(*discon_acc)), - GFP_KERNEL); - if (!lsop) - /* couldn't sent it... too bad */ + sizeof(*discon_rqst) + sizeof(*discon_acc) + + ctrl->lport->ops->lsrqst_priv_sz), GFP_KERNEL); + if (!lsop) { + dev_info(ctrl->ctrl.device, + "NVME-FC{%d}: send Disconnect Association " + "failed: ENOMEM\n", + ctrl->cnum); return; + } + discon_rqst = (struct fcnvme_ls_disconnect_assoc_rqst *)&lsop[1]; + discon_acc = (struct fcnvme_ls_disconnect_assoc_acc *)&discon_rqst[1]; lsreq = &lsop->ls_req; + if (ctrl->lport->ops->lsrqst_priv_sz) + lsreq->private = (void *)&discon_acc[1]; + else + lsreq->private = NULL; - lsreq->private = (void *)&lsop[1]; - discon_rqst = (struct fcnvme_ls_disconnect_assoc_rqst *) - (lsreq->private + ctrl->lport->ops->lsrqst_priv_sz); - discon_acc = (struct fcnvme_ls_disconnect_assoc_acc *)&discon_rqst[1]; + nvmefc_fmt_lsreq_discon_assoc(lsreq, discon_rqst, discon_acc, + ctrl->association_id); - discon_rqst->w0.ls_cmd = FCNVME_LS_DISCONNECT_ASSOC; - discon_rqst->desc_list_len = cpu_to_be32( - sizeof(struct fcnvme_lsdesc_assoc_id) + - sizeof(struct fcnvme_lsdesc_disconn_cmd)); + ret = nvme_fc_send_ls_req_async(ctrl->rport, lsop, + nvme_fc_disconnect_assoc_done); + if (ret) + kfree(lsop); +} - discon_rqst->associd.desc_tag = cpu_to_be32(FCNVME_LSDESC_ASSOC_ID); - discon_rqst->associd.desc_len = - fcnvme_lsdesc_len( - sizeof(struct fcnvme_lsdesc_assoc_id)); +static void +nvme_fc_xmt_ls_rsp_done(struct nvmefc_ls_rsp *lsrsp) +{ + struct nvmefc_ls_rcv_op *lsop = lsrsp->nvme_fc_private; + struct nvme_fc_rport *rport = lsop->rport; + struct nvme_fc_lport *lport = rport->lport; + unsigned long flags; + + spin_lock_irqsave(&rport->lock, flags); + list_del(&lsop->lsrcv_list); + spin_unlock_irqrestore(&rport->lock, flags); + + fc_dma_sync_single_for_cpu(lport->dev, lsop->rspdma, + sizeof(*lsop->rspbuf), DMA_TO_DEVICE); + fc_dma_unmap_single(lport->dev, lsop->rspdma, + sizeof(*lsop->rspbuf), DMA_TO_DEVICE); + + kfree(lsop); + + nvme_fc_rport_put(rport); +} + +static void +nvme_fc_xmt_ls_rsp(struct nvmefc_ls_rcv_op *lsop) +{ + struct nvme_fc_rport *rport = lsop->rport; + struct nvme_fc_lport *lport = rport->lport; + struct fcnvme_ls_rqst_w0 *w0 = &lsop->rqstbuf->w0; + int ret; + + fc_dma_sync_single_for_device(lport->dev, lsop->rspdma, + sizeof(*lsop->rspbuf), DMA_TO_DEVICE); + + ret = lport->ops->xmt_ls_rsp(&lport->localport, &rport->remoteport, + lsop->lsrsp); + if (ret) { + dev_warn(lport->dev, + "LLDD rejected LS RSP xmt: LS %d status %d\n", + w0->ls_cmd, ret); + nvme_fc_xmt_ls_rsp_done(lsop->lsrsp); + return; + } +} + +static struct nvme_fc_ctrl * +nvme_fc_match_disconn_ls(struct nvme_fc_rport *rport, + struct nvmefc_ls_rcv_op *lsop) +{ + struct fcnvme_ls_disconnect_assoc_rqst *rqst = + &lsop->rqstbuf->rq_dis_assoc; + struct nvme_fc_ctrl *ctrl, *ret = NULL; + struct nvmefc_ls_rcv_op *oldls = NULL; + u64 association_id = be64_to_cpu(rqst->associd.association_id); + unsigned long flags; + + spin_lock_irqsave(&rport->lock, flags); + + list_for_each_entry(ctrl, &rport->ctrl_list, ctrl_list) { + if (!nvme_fc_ctrl_get(ctrl)) + continue; + spin_lock(&ctrl->lock); + if (association_id == ctrl->association_id) { + oldls = ctrl->rcv_disconn; + ctrl->rcv_disconn = lsop; + ret = ctrl; + } + spin_unlock(&ctrl->lock); + if (ret) + /* leave the ctrl get reference */ + break; + nvme_fc_ctrl_put(ctrl); + } + + spin_unlock_irqrestore(&rport->lock, flags); + + /* transmit a response for anything that was pending */ + if (oldls) { + dev_info(rport->lport->dev, + "NVME-FC{%d}: Multiple Disconnect Association " + "LS's received\n", ctrl->cnum); + /* overwrite good response with bogus failure */ + oldls->lsrsp->rsplen = nvme_fc_format_rjt(oldls->rspbuf, + sizeof(*oldls->rspbuf), + rqst->w0.ls_cmd, + FCNVME_RJT_RC_UNAB, + FCNVME_RJT_EXP_NONE, 0); + nvme_fc_xmt_ls_rsp(oldls); + } + + return ret; +} + +/* + * returns true to mean LS handled and ls_rsp can be sent + * returns false to defer ls_rsp xmt (will be done as part of + * association termination) + */ +static bool +nvme_fc_ls_disconnect_assoc(struct nvmefc_ls_rcv_op *lsop) +{ + struct nvme_fc_rport *rport = lsop->rport; + struct fcnvme_ls_disconnect_assoc_rqst *rqst = + &lsop->rqstbuf->rq_dis_assoc; + struct fcnvme_ls_disconnect_assoc_acc *acc = + &lsop->rspbuf->rsp_dis_assoc; + struct nvme_fc_ctrl *ctrl = NULL; + int ret = 0; + + memset(acc, 0, sizeof(*acc)); + + ret = nvmefc_vldt_lsreq_discon_assoc(lsop->rqstdatalen, rqst); + if (!ret) { + /* match an active association */ + ctrl = nvme_fc_match_disconn_ls(rport, lsop); + if (!ctrl) + ret = VERR_NO_ASSOC; + } + + if (ret) { + dev_info(rport->lport->dev, + "Disconnect LS failed: %s\n", + validation_errors[ret]); + lsop->lsrsp->rsplen = nvme_fc_format_rjt(acc, + sizeof(*acc), rqst->w0.ls_cmd, + (ret == VERR_NO_ASSOC) ? + FCNVME_RJT_RC_INV_ASSOC : + FCNVME_RJT_RC_LOGIC, + FCNVME_RJT_EXP_NONE, 0); + return true; + } - discon_rqst->associd.association_id = cpu_to_be64(ctrl->association_id); + /* format an ACCept response */ - discon_rqst->discon_cmd.desc_tag = cpu_to_be32( - FCNVME_LSDESC_DISCONN_CMD); - discon_rqst->discon_cmd.desc_len = + lsop->lsrsp->rsplen = sizeof(*acc); + + nvme_fc_format_rsp_hdr(acc, FCNVME_LS_ACC, fcnvme_lsdesc_len( - sizeof(struct fcnvme_lsdesc_disconn_cmd)); + sizeof(struct fcnvme_ls_disconnect_assoc_acc)), + FCNVME_LS_DISCONNECT_ASSOC); - lsreq->rqstaddr = discon_rqst; - lsreq->rqstlen = sizeof(*discon_rqst); - lsreq->rspaddr = discon_acc; - lsreq->rsplen = sizeof(*discon_acc); - lsreq->timeout = NVME_FC_LS_TIMEOUT_SEC; + /* + * the transmit of the response will occur after the exchanges + * for the association have been ABTS'd by + * nvme_fc_delete_association(). + */ - ret = nvme_fc_send_ls_req_async(ctrl->rport, lsop, - nvme_fc_disconnect_assoc_done); - if (ret) - kfree(lsop); + /* fail the association */ + nvme_fc_error_recovery(ctrl, "Disconnect Association LS received"); + + /* release the reference taken by nvme_fc_match_disconn_ls() */ + nvme_fc_ctrl_put(ctrl); + + return false; } +/* + * Actual Processing routine for received FC-NVME LS Requests from the LLD + * returns true if a response should be sent afterward, false if rsp will + * be sent asynchronously. + */ +static bool +nvme_fc_handle_ls_rqst(struct nvmefc_ls_rcv_op *lsop) +{ + struct fcnvme_ls_rqst_w0 *w0 = &lsop->rqstbuf->w0; + bool ret = true; + + lsop->lsrsp->nvme_fc_private = lsop; + lsop->lsrsp->rspbuf = lsop->rspbuf; + lsop->lsrsp->rspdma = lsop->rspdma; + lsop->lsrsp->done = nvme_fc_xmt_ls_rsp_done; + /* Be preventative. handlers will later set to valid length */ + lsop->lsrsp->rsplen = 0; -/* *********************** NVME Ctrl Routines **************************** */ + /* + * handlers: + * parse request input, execute the request, and format the + * LS response + */ + switch (w0->ls_cmd) { + case FCNVME_LS_DISCONNECT_ASSOC: + ret = nvme_fc_ls_disconnect_assoc(lsop); + break; + case FCNVME_LS_DISCONNECT_CONN: + lsop->lsrsp->rsplen = nvme_fc_format_rjt(lsop->rspbuf, + sizeof(*lsop->rspbuf), w0->ls_cmd, + FCNVME_RJT_RC_UNSUP, FCNVME_RJT_EXP_NONE, 0); + break; + case FCNVME_LS_CREATE_ASSOCIATION: + case FCNVME_LS_CREATE_CONNECTION: + lsop->lsrsp->rsplen = nvme_fc_format_rjt(lsop->rspbuf, + sizeof(*lsop->rspbuf), w0->ls_cmd, + FCNVME_RJT_RC_LOGIC, FCNVME_RJT_EXP_NONE, 0); + break; + default: + lsop->lsrsp->rsplen = nvme_fc_format_rjt(lsop->rspbuf, + sizeof(*lsop->rspbuf), w0->ls_cmd, + FCNVME_RJT_RC_INVAL, FCNVME_RJT_EXP_NONE, 0); + break; + } -static void nvme_fc_error_recovery(struct nvme_fc_ctrl *ctrl, char *errmsg); + return(ret); +} + +static void +nvme_fc_handle_ls_rqst_work(struct work_struct *work) +{ + struct nvme_fc_rport *rport = + container_of(work, struct nvme_fc_rport, lsrcv_work); + struct fcnvme_ls_rqst_w0 *w0; + struct nvmefc_ls_rcv_op *lsop; + unsigned long flags; + bool sendrsp; + +restart: + sendrsp = true; + spin_lock_irqsave(&rport->lock, flags); + list_for_each_entry(lsop, &rport->ls_rcv_list, lsrcv_list) { + if (lsop->handled) + continue; + + lsop->handled = true; + if (rport->remoteport.port_state == FC_OBJSTATE_ONLINE) { + spin_unlock_irqrestore(&rport->lock, flags); + sendrsp = nvme_fc_handle_ls_rqst(lsop); + } else { + spin_unlock_irqrestore(&rport->lock, flags); + w0 = &lsop->rqstbuf->w0; + lsop->lsrsp->rsplen = nvme_fc_format_rjt( + lsop->rspbuf, + sizeof(*lsop->rspbuf), + w0->ls_cmd, + FCNVME_RJT_RC_UNAB, + FCNVME_RJT_EXP_NONE, 0); + } + if (sendrsp) + nvme_fc_xmt_ls_rsp(lsop); + goto restart; + } + spin_unlock_irqrestore(&rport->lock, flags); +} + +/** + * nvme_fc_rcv_ls_req - transport entry point called by an LLDD + * upon the reception of a NVME LS request. + * + * The nvme-fc layer will copy payload to an internal structure for + * processing. As such, upon completion of the routine, the LLDD may + * immediately free/reuse the LS request buffer passed in the call. + * + * If this routine returns error, the LLDD should abort the exchange. + * + * @remoteport: pointer to the (registered) remote port that the LS + * was received from. The remoteport is associated with + * a specific localport. + * @lsrsp: pointer to a nvmefc_ls_rsp response structure to be + * used to reference the exchange corresponding to the LS + * when issuing an ls response. + * @lsreqbuf: pointer to the buffer containing the LS Request + * @lsreqbuf_len: length, in bytes, of the received LS request + */ +int +nvme_fc_rcv_ls_req(struct nvme_fc_remote_port *portptr, + struct nvmefc_ls_rsp *lsrsp, + void *lsreqbuf, u32 lsreqbuf_len) +{ + struct nvme_fc_rport *rport = remoteport_to_rport(portptr); + struct nvme_fc_lport *lport = rport->lport; + struct fcnvme_ls_rqst_w0 *w0 = (struct fcnvme_ls_rqst_w0 *)lsreqbuf; + struct nvmefc_ls_rcv_op *lsop; + unsigned long flags; + int ret; + + nvme_fc_rport_get(rport); + + /* validate there's a routine to transmit a response */ + if (!lport->ops->xmt_ls_rsp) { + dev_info(lport->dev, + "RCV %s LS failed: no LLDD xmt_ls_rsp\n", + (w0->ls_cmd <= NVME_FC_LAST_LS_CMD_VALUE) ? + nvmefc_ls_names[w0->ls_cmd] : ""); + ret = -EINVAL; + goto out_put; + } + + if (lsreqbuf_len > sizeof(union nvmefc_ls_requests)) { + dev_info(lport->dev, + "RCV %s LS failed: payload too large\n", + (w0->ls_cmd <= NVME_FC_LAST_LS_CMD_VALUE) ? + nvmefc_ls_names[w0->ls_cmd] : ""); + ret = -E2BIG; + goto out_put; + } + + lsop = kzalloc(sizeof(*lsop) + + sizeof(union nvmefc_ls_requests) + + sizeof(union nvmefc_ls_responses), + GFP_KERNEL); + if (!lsop) { + dev_info(lport->dev, + "RCV %s LS failed: No memory\n", + (w0->ls_cmd <= NVME_FC_LAST_LS_CMD_VALUE) ? + nvmefc_ls_names[w0->ls_cmd] : ""); + ret = -ENOMEM; + goto out_put; + } + lsop->rqstbuf = (union nvmefc_ls_requests *)&lsop[1]; + lsop->rspbuf = (union nvmefc_ls_responses *)&lsop->rqstbuf[1]; + + lsop->rspdma = fc_dma_map_single(lport->dev, lsop->rspbuf, + sizeof(*lsop->rspbuf), + DMA_TO_DEVICE); + if (fc_dma_mapping_error(lport->dev, lsop->rspdma)) { + dev_info(lport->dev, + "RCV %s LS failed: DMA mapping failure\n", + (w0->ls_cmd <= NVME_FC_LAST_LS_CMD_VALUE) ? + nvmefc_ls_names[w0->ls_cmd] : ""); + ret = -EFAULT; + goto out_free; + } + + lsop->rport = rport; + lsop->lsrsp = lsrsp; + + memcpy(lsop->rqstbuf, lsreqbuf, lsreqbuf_len); + lsop->rqstdatalen = lsreqbuf_len; + + spin_lock_irqsave(&rport->lock, flags); + if (rport->remoteport.port_state != FC_OBJSTATE_ONLINE) { + spin_unlock_irqrestore(&rport->lock, flags); + ret = -ENOTCONN; + goto out_unmap; + } + list_add_tail(&lsop->lsrcv_list, &rport->ls_rcv_list); + spin_unlock_irqrestore(&rport->lock, flags); + + schedule_work(&rport->lsrcv_work); + + return 0; + +out_unmap: + fc_dma_unmap_single(lport->dev, lsop->rspdma, + sizeof(*lsop->rspbuf), DMA_TO_DEVICE); +out_free: + kfree(lsop); +out_put: + nvme_fc_rport_put(rport); + return ret; +} +EXPORT_SYMBOL_GPL(nvme_fc_rcv_ls_req); + + +/* *********************** NVME Ctrl Routines **************************** */ static void __nvme_fc_exit_request(struct nvme_fc_ctrl *ctrl, @@ -1500,7 +1835,7 @@ __nvme_fc_abort_op(struct nvme_fc_ctrl *ctrl, struct nvme_fc_fcp_op *op) opstate = atomic_xchg(&op->state, FCPOP_STATE_ABORTED); if (opstate != FCPOP_STATE_ACTIVE) atomic_set(&op->state, opstate); - else if (ctrl->flags & FCCTRL_TERMIO) + else if (test_bit(FCCTRL_TERMIO, &ctrl->flags)) ctrl->iocnt++; spin_unlock_irqrestore(&ctrl->lock, flags); @@ -1537,7 +1872,7 @@ __nvme_fc_fcpop_chk_teardowns(struct nvme_fc_ctrl *ctrl, if (opstate == FCPOP_STATE_ABORTED) { spin_lock_irqsave(&ctrl->lock, flags); - if (ctrl->flags & FCCTRL_TERMIO) { + if (test_bit(FCCTRL_TERMIO, &ctrl->flags)) { if (!--ctrl->iocnt) wake_up(&ctrl->ioabort_wait); } @@ -1771,7 +2106,7 @@ nvme_fc_init_request(struct blk_mq_tag_set *set, struct request *rq, res = __nvme_fc_init_request(ctrl, queue, &op->op, rq, queue->rqcnt++); if (res) return res; - op->op.fcp_req.first_sgl = &op->sgl[0]; + op->op.fcp_req.first_sgl = op->sgl; op->op.fcp_req.private = &op->priv[0]; nvme_req(rq)->ctrl = &ctrl->ctrl; return res; @@ -1783,15 +2118,17 @@ nvme_fc_init_aen_ops(struct nvme_fc_ctrl *ctrl) struct nvme_fc_fcp_op *aen_op; struct nvme_fc_cmd_iu *cmdiu; struct nvme_command *sqe; - void *private; + void *private = NULL; int i, ret; aen_op = ctrl->aen_ops; for (i = 0; i < NVME_NR_AEN_COMMANDS; i++, aen_op++) { - private = kzalloc(ctrl->lport->ops->fcprqst_priv_sz, + if (ctrl->lport->ops->fcprqst_priv_sz) { + private = kzalloc(ctrl->lport->ops->fcprqst_priv_sz, GFP_KERNEL); - if (!private) - return -ENOMEM; + if (!private) + return -ENOMEM; + } cmdiu = &aen_op->cmd_iu; sqe = &cmdiu->sqe; @@ -1822,9 +2159,6 @@ nvme_fc_term_aen_ops(struct nvme_fc_ctrl *ctrl) aen_op = ctrl->aen_ops; for (i = 0; i < NVME_NR_AEN_COMMANDS; i++, aen_op++) { - if (!aen_op->fcp_req.private) - continue; - __nvme_fc_exit_request(ctrl, aen_op); kfree(aen_op->fcp_req.private); @@ -2366,16 +2700,9 @@ nvme_fc_submit_async_event(struct nvme_ctrl *arg) { struct nvme_fc_ctrl *ctrl = to_fc_ctrl(arg); struct nvme_fc_fcp_op *aen_op; - unsigned long flags; - bool terminating = false; blk_status_t ret; - spin_lock_irqsave(&ctrl->lock, flags); - if (ctrl->flags & FCCTRL_TERMIO) - terminating = true; - spin_unlock_irqrestore(&ctrl->lock, flags); - - if (terminating) + if (test_bit(FCCTRL_TERMIO, &ctrl->flags)) return; aen_op = &ctrl->aen_ops[0]; @@ -2584,10 +2911,9 @@ nvme_fc_ctlr_active_on_rport(struct nvme_fc_ctrl *ctrl) struct nvme_fc_rport *rport = ctrl->rport; u32 cnt; - if (ctrl->assoc_active) + if (test_and_set_bit(ASSOC_ACTIVE, &ctrl->flags)) return 1; - ctrl->assoc_active = true; cnt = atomic_inc_return(&rport->act_ctrl_cnt); if (cnt == 1) nvme_fc_rport_active_on_lport(rport); @@ -2602,7 +2928,7 @@ nvme_fc_ctlr_inactive_on_rport(struct nvme_fc_ctrl *ctrl) struct nvme_fc_lport *lport = rport->lport; u32 cnt; - /* ctrl->assoc_active=false will be set independently */ + /* clearing of ctrl->flags ASSOC_ACTIVE bit is in association delete */ cnt = atomic_dec_return(&rport->act_ctrl_cnt); if (cnt == 0) { @@ -2622,6 +2948,8 @@ static int nvme_fc_create_association(struct nvme_fc_ctrl *ctrl) { struct nvmf_ctrl_options *opts = ctrl->ctrl.opts; + struct nvmefc_ls_rcv_op *disls = NULL; + unsigned long flags; int ret; bool changed; @@ -2739,12 +3067,18 @@ out_term_aen_ops: out_disconnect_admin_queue: /* send a Disconnect(association) LS to fc-nvme target */ nvme_fc_xmt_disconnect_assoc(ctrl); + spin_lock_irqsave(&ctrl->lock, flags); ctrl->association_id = 0; + disls = ctrl->rcv_disconn; + ctrl->rcv_disconn = NULL; + spin_unlock_irqrestore(&ctrl->lock, flags); + if (disls) + nvme_fc_xmt_ls_rsp(disls); out_delete_hw_queue: __nvme_fc_delete_hw_queue(ctrl, &ctrl->queues[0], 0); out_free_queue: nvme_fc_free_queue(&ctrl->queues[0]); - ctrl->assoc_active = false; + clear_bit(ASSOC_ACTIVE, &ctrl->flags); nvme_fc_ctlr_inactive_on_rport(ctrl); return ret; @@ -2759,14 +3093,14 @@ out_free_queue: static void nvme_fc_delete_association(struct nvme_fc_ctrl *ctrl) { + struct nvmefc_ls_rcv_op *disls = NULL; unsigned long flags; - if (!ctrl->assoc_active) + if (!test_and_clear_bit(ASSOC_ACTIVE, &ctrl->flags)) return; - ctrl->assoc_active = false; spin_lock_irqsave(&ctrl->lock, flags); - ctrl->flags |= FCCTRL_TERMIO; + set_bit(FCCTRL_TERMIO, &ctrl->flags); ctrl->iocnt = 0; spin_unlock_irqrestore(&ctrl->lock, flags); @@ -2817,7 +3151,7 @@ nvme_fc_delete_association(struct nvme_fc_ctrl *ctrl) /* wait for all io that had to be aborted */ spin_lock_irq(&ctrl->lock); wait_event_lock_irq(ctrl->ioabort_wait, ctrl->iocnt == 0, ctrl->lock); - ctrl->flags &= ~FCCTRL_TERMIO; + clear_bit(FCCTRL_TERMIO, &ctrl->flags); spin_unlock_irq(&ctrl->lock); nvme_fc_term_aen_ops(ctrl); @@ -2831,7 +3165,17 @@ nvme_fc_delete_association(struct nvme_fc_ctrl *ctrl) if (ctrl->association_id) nvme_fc_xmt_disconnect_assoc(ctrl); + spin_lock_irqsave(&ctrl->lock, flags); ctrl->association_id = 0; + disls = ctrl->rcv_disconn; + ctrl->rcv_disconn = NULL; + spin_unlock_irqrestore(&ctrl->lock, flags); + if (disls) + /* + * if a Disconnect Request was waiting for a response, send + * now that all ABTS's have been issued (and are complete). + */ + nvme_fc_xmt_ls_rsp(disls); if (ctrl->ctrl.tagset) { nvme_fc_delete_hw_io_queues(ctrl); @@ -2902,7 +3246,9 @@ nvme_fc_reconnect_or_delete(struct nvme_fc_ctrl *ctrl, int status) dev_warn(ctrl->ctrl.device, "NVME-FC{%d}: dev_loss_tmo (%d) expired " "while waiting for remoteport connectivity.\n", - ctrl->cnum, portptr->dev_loss_tmo); + ctrl->cnum, min_t(int, portptr->dev_loss_tmo, + (ctrl->ctrl.opts->max_reconnects * + ctrl->ctrl.opts->reconnect_delay))); WARN_ON(nvme_delete_ctrl(&ctrl->ctrl)); } } @@ -3089,7 +3435,6 @@ nvme_fc_init_ctrl(struct device *dev, struct nvmf_ctrl_options *opts, ctrl->dev = lport->dev; ctrl->cnum = idx; ctrl->ioq_live = false; - ctrl->assoc_active = false; atomic_set(&ctrl->err_work_active, 0); init_waitqueue_head(&ctrl->ioabort_wait); diff --git a/drivers/nvme/host/fc.h b/drivers/nvme/host/fc.h new file mode 100644 index 000000000000..05ce566f2caf --- /dev/null +++ b/drivers/nvme/host/fc.h @@ -0,0 +1,227 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +/* + * Copyright (c) 2016, Avago Technologies + */ + +#ifndef _NVME_FC_TRANSPORT_H +#define _NVME_FC_TRANSPORT_H 1 + + +/* + * Common definitions between the nvme_fc (host) transport and + * nvmet_fc (target) transport implementation. + */ + +/* + * ****************** FC-NVME LS HANDLING ****************** + */ + +union nvmefc_ls_requests { + struct fcnvme_ls_rqst_w0 w0; + struct fcnvme_ls_cr_assoc_rqst rq_cr_assoc; + struct fcnvme_ls_cr_conn_rqst rq_cr_conn; + struct fcnvme_ls_disconnect_assoc_rqst rq_dis_assoc; + struct fcnvme_ls_disconnect_conn_rqst rq_dis_conn; +} __aligned(128); /* alignment for other things alloc'd with */ + +union nvmefc_ls_responses { + struct fcnvme_ls_rjt rsp_rjt; + struct fcnvme_ls_cr_assoc_acc rsp_cr_assoc; + struct fcnvme_ls_cr_conn_acc rsp_cr_conn; + struct fcnvme_ls_disconnect_assoc_acc rsp_dis_assoc; + struct fcnvme_ls_disconnect_conn_acc rsp_dis_conn; +} __aligned(128); /* alignment for other things alloc'd with */ + +static inline void +nvme_fc_format_rsp_hdr(void *buf, u8 ls_cmd, __be32 desc_len, u8 rqst_ls_cmd) +{ + struct fcnvme_ls_acc_hdr *acc = buf; + + acc->w0.ls_cmd = ls_cmd; + acc->desc_list_len = desc_len; + acc->rqst.desc_tag = cpu_to_be32(FCNVME_LSDESC_RQST); + acc->rqst.desc_len = + fcnvme_lsdesc_len(sizeof(struct fcnvme_lsdesc_rqst)); + acc->rqst.w0.ls_cmd = rqst_ls_cmd; +} + +static inline int +nvme_fc_format_rjt(void *buf, u16 buflen, u8 ls_cmd, + u8 reason, u8 explanation, u8 vendor) +{ + struct fcnvme_ls_rjt *rjt = buf; + + nvme_fc_format_rsp_hdr(buf, FCNVME_LSDESC_RQST, + fcnvme_lsdesc_len(sizeof(struct fcnvme_ls_rjt)), + ls_cmd); + rjt->rjt.desc_tag = cpu_to_be32(FCNVME_LSDESC_RJT); + rjt->rjt.desc_len = fcnvme_lsdesc_len(sizeof(struct fcnvme_lsdesc_rjt)); + rjt->rjt.reason_code = reason; + rjt->rjt.reason_explanation = explanation; + rjt->rjt.vendor = vendor; + + return sizeof(struct fcnvme_ls_rjt); +} + +/* Validation Error indexes into the string table below */ +enum { + VERR_NO_ERROR = 0, + VERR_CR_ASSOC_LEN = 1, + VERR_CR_ASSOC_RQST_LEN = 2, + VERR_CR_ASSOC_CMD = 3, + VERR_CR_ASSOC_CMD_LEN = 4, + VERR_ERSP_RATIO = 5, + VERR_ASSOC_ALLOC_FAIL = 6, + VERR_QUEUE_ALLOC_FAIL = 7, + VERR_CR_CONN_LEN = 8, + VERR_CR_CONN_RQST_LEN = 9, + VERR_ASSOC_ID = 10, + VERR_ASSOC_ID_LEN = 11, + VERR_NO_ASSOC = 12, + VERR_CONN_ID = 13, + VERR_CONN_ID_LEN = 14, + VERR_INVAL_CONN = 15, + VERR_CR_CONN_CMD = 16, + VERR_CR_CONN_CMD_LEN = 17, + VERR_DISCONN_LEN = 18, + VERR_DISCONN_RQST_LEN = 19, + VERR_DISCONN_CMD = 20, + VERR_DISCONN_CMD_LEN = 21, + VERR_DISCONN_SCOPE = 22, + VERR_RS_LEN = 23, + VERR_RS_RQST_LEN = 24, + VERR_RS_CMD = 25, + VERR_RS_CMD_LEN = 26, + VERR_RS_RCTL = 27, + VERR_RS_RO = 28, + VERR_LSACC = 29, + VERR_LSDESC_RQST = 30, + VERR_LSDESC_RQST_LEN = 31, + VERR_CR_ASSOC = 32, + VERR_CR_ASSOC_ACC_LEN = 33, + VERR_CR_CONN = 34, + VERR_CR_CONN_ACC_LEN = 35, + VERR_DISCONN = 36, + VERR_DISCONN_ACC_LEN = 37, +}; + +static char *validation_errors[] = { + "OK", + "Bad CR_ASSOC Length", + "Bad CR_ASSOC Rqst Length", + "Not CR_ASSOC Cmd", + "Bad CR_ASSOC Cmd Length", + "Bad Ersp Ratio", + "Association Allocation Failed", + "Queue Allocation Failed", + "Bad CR_CONN Length", + "Bad CR_CONN Rqst Length", + "Not Association ID", + "Bad Association ID Length", + "No Association", + "Not Connection ID", + "Bad Connection ID Length", + "Invalid Connection ID", + "Not CR_CONN Cmd", + "Bad CR_CONN Cmd Length", + "Bad DISCONN Length", + "Bad DISCONN Rqst Length", + "Not DISCONN Cmd", + "Bad DISCONN Cmd Length", + "Bad Disconnect Scope", + "Bad RS Length", + "Bad RS Rqst Length", + "Not RS Cmd", + "Bad RS Cmd Length", + "Bad RS R_CTL", + "Bad RS Relative Offset", + "Not LS_ACC", + "Not LSDESC_RQST", + "Bad LSDESC_RQST Length", + "Not CR_ASSOC Rqst", + "Bad CR_ASSOC ACC Length", + "Not CR_CONN Rqst", + "Bad CR_CONN ACC Length", + "Not Disconnect Rqst", + "Bad Disconnect ACC Length", +}; + +#define NVME_FC_LAST_LS_CMD_VALUE FCNVME_LS_DISCONNECT_CONN + +static char *nvmefc_ls_names[] = { + "Reserved (0)", + "RJT (1)", + "ACC (2)", + "Create Association", + "Create Connection", + "Disconnect Association", + "Disconnect Connection", +}; + +static inline void +nvmefc_fmt_lsreq_discon_assoc(struct nvmefc_ls_req *lsreq, + struct fcnvme_ls_disconnect_assoc_rqst *discon_rqst, + struct fcnvme_ls_disconnect_assoc_acc *discon_acc, + u64 association_id) +{ + lsreq->rqstaddr = discon_rqst; + lsreq->rqstlen = sizeof(*discon_rqst); + lsreq->rspaddr = discon_acc; + lsreq->rsplen = sizeof(*discon_acc); + lsreq->timeout = NVME_FC_LS_TIMEOUT_SEC; + + discon_rqst->w0.ls_cmd = FCNVME_LS_DISCONNECT_ASSOC; + discon_rqst->desc_list_len = cpu_to_be32( + sizeof(struct fcnvme_lsdesc_assoc_id) + + sizeof(struct fcnvme_lsdesc_disconn_cmd)); + + discon_rqst->associd.desc_tag = cpu_to_be32(FCNVME_LSDESC_ASSOC_ID); + discon_rqst->associd.desc_len = + fcnvme_lsdesc_len( + sizeof(struct fcnvme_lsdesc_assoc_id)); + + discon_rqst->associd.association_id = cpu_to_be64(association_id); + + discon_rqst->discon_cmd.desc_tag = cpu_to_be32( + FCNVME_LSDESC_DISCONN_CMD); + discon_rqst->discon_cmd.desc_len = + fcnvme_lsdesc_len( + sizeof(struct fcnvme_lsdesc_disconn_cmd)); +} + +static inline int +nvmefc_vldt_lsreq_discon_assoc(u32 rqstlen, + struct fcnvme_ls_disconnect_assoc_rqst *rqst) +{ + int ret = 0; + + if (rqstlen < sizeof(struct fcnvme_ls_disconnect_assoc_rqst)) + ret = VERR_DISCONN_LEN; + else if (rqst->desc_list_len != + fcnvme_lsdesc_len( + sizeof(struct fcnvme_ls_disconnect_assoc_rqst))) + ret = VERR_DISCONN_RQST_LEN; + else if (rqst->associd.desc_tag != cpu_to_be32(FCNVME_LSDESC_ASSOC_ID)) + ret = VERR_ASSOC_ID; + else if (rqst->associd.desc_len != + fcnvme_lsdesc_len( + sizeof(struct fcnvme_lsdesc_assoc_id))) + ret = VERR_ASSOC_ID_LEN; + else if (rqst->discon_cmd.desc_tag != + cpu_to_be32(FCNVME_LSDESC_DISCONN_CMD)) + ret = VERR_DISCONN_CMD; + else if (rqst->discon_cmd.desc_len != + fcnvme_lsdesc_len( + sizeof(struct fcnvme_lsdesc_disconn_cmd))) + ret = VERR_DISCONN_CMD_LEN; + /* + * As the standard changed on the LS, check if old format and scope + * something other than Association (e.g. 0). + */ + else if (rqst->discon_cmd.rsvd8[0]) + ret = VERR_DISCONN_SCOPE; + + return ret; +} + +#endif /* _NVME_FC_TRANSPORT_H */ diff --git a/drivers/nvme/host/lightnvm.c b/drivers/nvme/host/lightnvm.c index ec46693f6b64..69608755d415 100644 --- a/drivers/nvme/host/lightnvm.c +++ b/drivers/nvme/host/lightnvm.c @@ -171,7 +171,7 @@ struct nvme_nvm_bb_tbl { __le32 tdresv; __le32 thresv; __le32 rsvd2[8]; - __u8 blk[0]; + __u8 blk[]; }; struct nvme_nvm_id20_addrf { @@ -961,7 +961,10 @@ int nvme_nvm_register(struct nvme_ns *ns, char *disk_name, int node) geo = &dev->geo; geo->csecs = 1 << ns->lba_shift; geo->sos = ns->ms; - geo->ext = ns->ext; + if (ns->features & NVME_NS_EXT_LBAS) + geo->ext = true; + else + geo->ext = false; geo->mdts = ns->ctrl->max_hw_sectors; dev->q = q; diff --git a/drivers/nvme/host/multipath.c b/drivers/nvme/host/multipath.c index 54603bd3e02d..da78e499947a 100644 --- a/drivers/nvme/host/multipath.c +++ b/drivers/nvme/host/multipath.c @@ -3,6 +3,7 @@ * Copyright (c) 2017-2018 Christoph Hellwig. */ +#include <linux/backing-dev.h> #include <linux/moduleparam.h> #include <trace/events/block.h> #include "nvme.h" @@ -293,7 +294,7 @@ static bool nvme_available_path(struct nvme_ns_head *head) static blk_qc_t nvme_ns_head_make_request(struct request_queue *q, struct bio *bio) { - struct nvme_ns_head *head = q->queuedata; + struct nvme_ns_head *head = bio->bi_disk->private_data; struct device *dev = disk_to_dev(head->disk); struct nvme_ns *ns; blk_qc_t ret = BLK_QC_T_NONE; @@ -371,13 +372,12 @@ int nvme_mpath_alloc_disk(struct nvme_ctrl *ctrl, struct nvme_ns_head *head) * We also do this for private namespaces as the namespace sharing data could * change after a rescan. */ - if (!(ctrl->subsys->cmic & (1 << 1)) || !multipath) + if (!(ctrl->subsys->cmic & NVME_CTRL_CMIC_MULTI_CTRL) || !multipath) return 0; q = blk_alloc_queue(nvme_ns_head_make_request, ctrl->numa_node); if (!q) goto out; - q->queuedata = head; blk_queue_flag_set(QUEUE_FLAG_NONROT, q); /* set to a default value for 512 until disk is validated */ blk_queue_logical_block_size(q, 512); @@ -666,6 +666,13 @@ void nvme_mpath_add_disk(struct nvme_ns *ns, struct nvme_id_ns *id) nvme_mpath_set_live(ns); mutex_unlock(&ns->head->lock); } + + if (bdi_cap_stable_pages_required(ns->queue->backing_dev_info)) { + struct backing_dev_info *info = + ns->head->disk->queue->backing_dev_info; + + info->capabilities |= BDI_CAP_STABLE_WRITES; + } } void nvme_mpath_remove_disk(struct nvme_ns_head *head) @@ -687,7 +694,8 @@ int nvme_mpath_init(struct nvme_ctrl *ctrl, struct nvme_id_ctrl *id) int error; /* check if multipath is enabled and we have the capability */ - if (!multipath || !ctrl->subsys || !(ctrl->subsys->cmic & (1 << 3))) + if (!multipath || !ctrl->subsys || + !(ctrl->subsys->cmic & NVME_CTRL_CMIC_ANA)) return 0; ctrl->anacap = id->anacap; diff --git a/drivers/nvme/host/nvme.h b/drivers/nvme/host/nvme.h index 2e04a36296d9..fa5c75501049 100644 --- a/drivers/nvme/host/nvme.h +++ b/drivers/nvme/host/nvme.h @@ -16,6 +16,7 @@ #include <linux/fault-inject.h> #include <linux/rcupdate.h> #include <linux/wait.h> +#include <linux/t10-pi.h> #include <trace/events/block.h> @@ -30,8 +31,10 @@ extern unsigned int admin_timeout; #ifdef CONFIG_ARCH_NO_SG_CHAIN #define NVME_INLINE_SG_CNT 0 +#define NVME_INLINE_METADATA_SG_CNT 0 #else #define NVME_INLINE_SG_CNT 2 +#define NVME_INLINE_METADATA_SG_CNT 1 #endif extern struct workqueue_struct *nvme_wq; @@ -228,6 +231,7 @@ struct nvme_ctrl { u32 page_size; u32 max_hw_sectors; u32 max_segments; + u32 max_integrity_segments; u16 crdt[3]; u16 oncs; u16 oacs; @@ -352,6 +356,7 @@ struct nvme_ns_head { struct nvme_ns_ids ids; struct list_head entry; struct kref ref; + bool shared; int instance; #ifdef CONFIG_NVME_MULTIPATH struct gendisk *disk; @@ -363,6 +368,11 @@ struct nvme_ns_head { #endif }; +enum nvme_ns_features { + NVME_NS_EXT_LBAS = 1 << 0, /* support extended LBA format */ + NVME_NS_METADATA_SUPPORTED = 1 << 1, /* support getting generated md */ +}; + struct nvme_ns { struct list_head list; @@ -382,18 +392,23 @@ struct nvme_ns { u16 ms; u16 sgs; u32 sws; - bool ext; u8 pi_type; + unsigned long features; unsigned long flags; #define NVME_NS_REMOVING 0 #define NVME_NS_DEAD 1 #define NVME_NS_ANA_PENDING 2 - u16 noiob; struct nvme_fault_inject fault_inject; }; +/* NVMe ns supports metadata actions by the controller (generate/strip) */ +static inline bool nvme_ns_has_pi(struct nvme_ns *ns) +{ + return ns->pi_type && ns->ms == sizeof(struct t10_pi_tuple); +} + struct nvme_ctrl_ops { const char *name; struct module *module; @@ -449,6 +464,14 @@ static inline sector_t nvme_lba_to_sect(struct nvme_ns *ns, u64 lba) return lba << (ns->lba_shift - SECTOR_SHIFT); } +/* + * Convert byte length to nvme's 0-based num dwords + */ +static inline u32 nvme_bytes_to_numd(size_t len) +{ + return (len >> 2) - 1; +} + static inline void nvme_end_request(struct request *req, __le16 status, union nvme_result result) { @@ -489,7 +512,6 @@ int nvme_init_ctrl(struct nvme_ctrl *ctrl, struct device *dev, void nvme_uninit_ctrl(struct nvme_ctrl *ctrl); void nvme_start_ctrl(struct nvme_ctrl *ctrl); void nvme_stop_ctrl(struct nvme_ctrl *ctrl); -void nvme_put_ctrl(struct nvme_ctrl *ctrl); int nvme_init_identify(struct nvme_ctrl *ctrl); void nvme_remove_namespaces(struct nvme_ctrl *ctrl); diff --git a/drivers/nvme/host/pci.c b/drivers/nvme/host/pci.c index 4e79e412b276..d690d5593a80 100644 --- a/drivers/nvme/host/pci.c +++ b/drivers/nvme/host/pci.c @@ -68,14 +68,30 @@ static int io_queue_depth = 1024; module_param_cb(io_queue_depth, &io_queue_depth_ops, &io_queue_depth, 0644); MODULE_PARM_DESC(io_queue_depth, "set io queue depth, should >= 2"); +static int io_queue_count_set(const char *val, const struct kernel_param *kp) +{ + unsigned int n; + int ret; + + ret = kstrtouint(val, 10, &n); + if (ret != 0 || n > num_possible_cpus()) + return -EINVAL; + return param_set_uint(val, kp); +} + +static const struct kernel_param_ops io_queue_count_ops = { + .set = io_queue_count_set, + .get = param_get_uint, +}; + static unsigned int write_queues; -module_param(write_queues, uint, 0644); +module_param_cb(write_queues, &io_queue_count_ops, &write_queues, 0644); MODULE_PARM_DESC(write_queues, "Number of queues to use for writes. If not set, reads and writes " "will share a queue set."); static unsigned int poll_queues; -module_param(poll_queues, uint, 0644); +module_param_cb(poll_queues, &io_queue_count_ops, &poll_queues, 0644); MODULE_PARM_DESC(poll_queues, "Number of queues to use for polled IO."); struct nvme_dev; @@ -128,6 +144,9 @@ struct nvme_dev { dma_addr_t host_mem_descs_dma; struct nvme_host_mem_buf_desc *host_mem_descs; void **host_mem_desc_bufs; + unsigned int nr_allocated_queues; + unsigned int nr_write_queues; + unsigned int nr_poll_queues; }; static int io_queue_depth_set(const char *val, const struct kernel_param *kp) @@ -166,14 +185,13 @@ struct nvme_queue { void *sq_cmds; /* only used for poll queues: */ spinlock_t cq_poll_lock ____cacheline_aligned_in_smp; - volatile struct nvme_completion *cqes; + struct nvme_completion *cqes; dma_addr_t sq_dma_addr; dma_addr_t cq_dma_addr; u32 __iomem *q_db; u16 q_depth; u16 cq_vector; u16 sq_tail; - u16 last_sq_tail; u16 cq_head; u16 qid; u8 cq_phase; @@ -209,25 +227,14 @@ struct nvme_iod { struct scatterlist *sg; }; -static unsigned int max_io_queues(void) +static inline unsigned int nvme_dbbuf_size(struct nvme_dev *dev) { - return num_possible_cpus() + write_queues + poll_queues; -} - -static unsigned int max_queue_count(void) -{ - /* IO queues + admin queue */ - return 1 + max_io_queues(); -} - -static inline unsigned int nvme_dbbuf_size(u32 stride) -{ - return (max_queue_count() * 8 * stride); + return dev->nr_allocated_queues * 8 * dev->db_stride; } static int nvme_dbbuf_dma_alloc(struct nvme_dev *dev) { - unsigned int mem_size = nvme_dbbuf_size(dev->db_stride); + unsigned int mem_size = nvme_dbbuf_size(dev); if (dev->dbbuf_dbs) return 0; @@ -252,7 +259,7 @@ static int nvme_dbbuf_dma_alloc(struct nvme_dev *dev) static void nvme_dbbuf_dma_free(struct nvme_dev *dev) { - unsigned int mem_size = nvme_dbbuf_size(dev->db_stride); + unsigned int mem_size = nvme_dbbuf_size(dev); if (dev->dbbuf_dbs) { dma_free_coherent(dev->dev, mem_size, @@ -446,24 +453,11 @@ static int nvme_pci_map_queues(struct blk_mq_tag_set *set) return 0; } -/* - * Write sq tail if we are asked to, or if the next command would wrap. - */ -static inline void nvme_write_sq_db(struct nvme_queue *nvmeq, bool write_sq) +static inline void nvme_write_sq_db(struct nvme_queue *nvmeq) { - if (!write_sq) { - u16 next_tail = nvmeq->sq_tail + 1; - - if (next_tail == nvmeq->q_depth) - next_tail = 0; - if (next_tail != nvmeq->last_sq_tail) - return; - } - if (nvme_dbbuf_update_and_check_event(nvmeq->sq_tail, nvmeq->dbbuf_sq_db, nvmeq->dbbuf_sq_ei)) writel(nvmeq->sq_tail, nvmeq->q_db); - nvmeq->last_sq_tail = nvmeq->sq_tail; } /** @@ -480,7 +474,8 @@ static void nvme_submit_cmd(struct nvme_queue *nvmeq, struct nvme_command *cmd, cmd, sizeof(*cmd)); if (++nvmeq->sq_tail == nvmeq->q_depth) nvmeq->sq_tail = 0; - nvme_write_sq_db(nvmeq, write_sq); + if (write_sq) + nvme_write_sq_db(nvmeq); spin_unlock(&nvmeq->sq_lock); } @@ -489,8 +484,7 @@ static void nvme_commit_rqs(struct blk_mq_hw_ctx *hctx) struct nvme_queue *nvmeq = hctx->driver_data; spin_lock(&nvmeq->sq_lock); - if (nvmeq->sq_tail != nvmeq->last_sq_tail) - nvme_write_sq_db(nvmeq, true); + nvme_write_sq_db(nvmeq); spin_unlock(&nvmeq->sq_lock); } @@ -922,8 +916,9 @@ static void nvme_pci_complete_rq(struct request *req) /* We read the CQE phase first to check if the rest of the entry is valid */ static inline bool nvme_cqe_pending(struct nvme_queue *nvmeq) { - return (le16_to_cpu(nvmeq->cqes[nvmeq->cq_head].status) & 1) == - nvmeq->cq_phase; + struct nvme_completion *hcqe = &nvmeq->cqes[nvmeq->cq_head]; + + return (le16_to_cpu(READ_ONCE(hcqe->status)) & 1) == nvmeq->cq_phase; } static inline void nvme_ring_cq_doorbell(struct nvme_queue *nvmeq) @@ -944,7 +939,7 @@ static inline struct blk_mq_tags *nvme_queue_tagset(struct nvme_queue *nvmeq) static inline void nvme_handle_cqe(struct nvme_queue *nvmeq, u16 idx) { - volatile struct nvme_completion *cqe = &nvmeq->cqes[idx]; + struct nvme_completion *cqe = &nvmeq->cqes[idx]; struct request *req; if (unlikely(cqe->command_id >= nvmeq->q_depth)) { @@ -973,9 +968,13 @@ static inline void nvme_handle_cqe(struct nvme_queue *nvmeq, u16 idx) static inline void nvme_update_cq_head(struct nvme_queue *nvmeq) { - if (++nvmeq->cq_head == nvmeq->q_depth) { + u16 tmp = nvmeq->cq_head + 1; + + if (tmp == nvmeq->q_depth) { nvmeq->cq_head = 0; nvmeq->cq_phase ^= 1; + } else { + nvmeq->cq_head = tmp; } } @@ -985,6 +984,11 @@ static inline int nvme_process_cq(struct nvme_queue *nvmeq) while (nvme_cqe_pending(nvmeq)) { found++; + /* + * load-load control dependency between phase and the rest of + * the cqe requires a full read memory barrier + */ + dma_rmb(); nvme_handle_cqe(nvmeq, nvmeq->cq_head); nvme_update_cq_head(nvmeq); } @@ -1373,16 +1377,19 @@ static void nvme_disable_admin_queue(struct nvme_dev *dev, bool shutdown) /* * Called only on a device that has been disabled and after all other threads - * that can check this device's completion queues have synced. This is the - * last chance for the driver to see a natural completion before - * nvme_cancel_request() terminates all incomplete requests. + * that can check this device's completion queues have synced, except + * nvme_poll(). This is the last chance for the driver to see a natural + * completion before nvme_cancel_request() terminates all incomplete requests. */ static void nvme_reap_pending_cqes(struct nvme_dev *dev) { int i; - for (i = dev->ctrl.queue_count - 1; i > 0; i--) + for (i = dev->ctrl.queue_count - 1; i > 0; i--) { + spin_lock(&dev->queues[i].cq_poll_lock); nvme_process_cq(&dev->queues[i]); + spin_unlock(&dev->queues[i].cq_poll_lock); + } } static int nvme_cmb_qdepth(struct nvme_dev *dev, int nr_io_queues, @@ -1489,7 +1496,6 @@ static void nvme_init_queue(struct nvme_queue *nvmeq, u16 qid) struct nvme_dev *dev = nvmeq->dev; nvmeq->sq_tail = 0; - nvmeq->last_sq_tail = 0; nvmeq->cq_head = 0; nvmeq->cq_phase = 1; nvmeq->q_db = &dev->dbs[qid * 2 * dev->db_stride]; @@ -1991,7 +1997,7 @@ static int nvme_setup_host_mem(struct nvme_dev *dev) static void nvme_calc_irq_sets(struct irq_affinity *affd, unsigned int nrirqs) { struct nvme_dev *dev = affd->priv; - unsigned int nr_read_queues; + unsigned int nr_read_queues, nr_write_queues = dev->nr_write_queues; /* * If there is no interupt available for queues, ensure that @@ -2007,12 +2013,12 @@ static void nvme_calc_irq_sets(struct irq_affinity *affd, unsigned int nrirqs) if (!nrirqs) { nrirqs = 1; nr_read_queues = 0; - } else if (nrirqs == 1 || !write_queues) { + } else if (nrirqs == 1 || !nr_write_queues) { nr_read_queues = 0; - } else if (write_queues >= nrirqs) { + } else if (nr_write_queues >= nrirqs) { nr_read_queues = 1; } else { - nr_read_queues = nrirqs - write_queues; + nr_read_queues = nrirqs - nr_write_queues; } dev->io_queues[HCTX_TYPE_DEFAULT] = nrirqs - nr_read_queues; @@ -2036,7 +2042,7 @@ static int nvme_setup_irqs(struct nvme_dev *dev, unsigned int nr_io_queues) * Poll queues don't need interrupts, but we need at least one IO * queue left over for non-polled IO. */ - this_p_queues = poll_queues; + this_p_queues = dev->nr_poll_queues; if (this_p_queues >= nr_io_queues) { this_p_queues = nr_io_queues - 1; irq_queues = 1; @@ -2066,14 +2072,25 @@ static void nvme_disable_io_queues(struct nvme_dev *dev) __nvme_disable_io_queues(dev, nvme_admin_delete_cq); } +static unsigned int nvme_max_io_queues(struct nvme_dev *dev) +{ + return num_possible_cpus() + dev->nr_write_queues + dev->nr_poll_queues; +} + static int nvme_setup_io_queues(struct nvme_dev *dev) { struct nvme_queue *adminq = &dev->queues[0]; struct pci_dev *pdev = to_pci_dev(dev->dev); - int result, nr_io_queues; + unsigned int nr_io_queues; unsigned long size; + int result; - nr_io_queues = max_io_queues(); + /* + * Sample the module parameters once at reset time so that we have + * stable values to work with. + */ + dev->nr_write_queues = write_queues; + dev->nr_poll_queues = poll_queues; /* * If tags are shared with admin queue (Apple bug), then @@ -2081,6 +2098,9 @@ static int nvme_setup_io_queues(struct nvme_dev *dev) */ if (dev->ctrl.quirks & NVME_QUIRK_SHARED_TAGS) nr_io_queues = 1; + else + nr_io_queues = min(nvme_max_io_queues(dev), + dev->nr_allocated_queues - 1); result = nvme_set_queue_count(&dev->ctrl, &nr_io_queues); if (result < 0) @@ -2553,6 +2573,12 @@ static void nvme_reset_work(struct work_struct *work) goto out; } + /* + * We do not support an SGL for metadata (yet), so we are limited to a + * single integrity segment for the separate metadata pointer. + */ + dev->ctrl.max_integrity_segments = 1; + result = nvme_init_identify(&dev->ctrl); if (result) goto out; @@ -2755,8 +2781,11 @@ static int nvme_probe(struct pci_dev *pdev, const struct pci_device_id *id) if (!dev) return -ENOMEM; - dev->queues = kcalloc_node(max_queue_count(), sizeof(struct nvme_queue), - GFP_KERNEL, node); + dev->nr_write_queues = write_queues; + dev->nr_poll_queues = poll_queues; + dev->nr_allocated_queues = nvme_max_io_queues(dev) + 1; + dev->queues = kcalloc_node(dev->nr_allocated_queues, + sizeof(struct nvme_queue), GFP_KERNEL, node); if (!dev->queues) goto free; @@ -3119,8 +3148,6 @@ static int __init nvme_init(void) BUILD_BUG_ON(sizeof(struct nvme_delete_queue) != 64); BUILD_BUG_ON(IRQ_AFFINITY_MAX_SETS < 2); - write_queues = min(write_queues, num_possible_cpus()); - poll_queues = min(poll_queues, num_possible_cpus()); return pci_register_driver(&nvme_driver); } diff --git a/drivers/nvme/host/rdma.c b/drivers/nvme/host/rdma.c index cac8a930396a..f8f856dc0c67 100644 --- a/drivers/nvme/host/rdma.c +++ b/drivers/nvme/host/rdma.c @@ -34,6 +34,11 @@ #define NVME_RDMA_MAX_INLINE_SEGMENTS 4 +#define NVME_RDMA_DATA_SGL_SIZE \ + (sizeof(struct scatterlist) * NVME_INLINE_SG_CNT) +#define NVME_RDMA_METADATA_SGL_SIZE \ + (sizeof(struct scatterlist) * NVME_INLINE_METADATA_SG_CNT) + struct nvme_rdma_device { struct ib_device *dev; struct ib_pd *pd; @@ -48,6 +53,11 @@ struct nvme_rdma_qe { u64 dma; }; +struct nvme_rdma_sgl { + int nents; + struct sg_table sg_table; +}; + struct nvme_rdma_queue; struct nvme_rdma_request { struct nvme_request req; @@ -58,12 +68,12 @@ struct nvme_rdma_request { refcount_t ref; struct ib_sge sge[1 + NVME_RDMA_MAX_INLINE_SEGMENTS]; u32 num_sge; - int nents; struct ib_reg_wr reg_wr; struct ib_cqe reg_cqe; struct nvme_rdma_queue *queue; - struct sg_table sg_table; - struct scatterlist first_sgl[]; + struct nvme_rdma_sgl data_sgl; + struct nvme_rdma_sgl *metadata_sgl; + bool use_sig_mr; }; enum nvme_rdma_queue_flags { @@ -85,6 +95,7 @@ struct nvme_rdma_queue { struct rdma_cm_id *cm_id; int cm_error; struct completion cm_done; + bool pi_support; }; struct nvme_rdma_ctrl { @@ -261,6 +272,8 @@ static int nvme_rdma_create_qp(struct nvme_rdma_queue *queue, const int factor) init_attr.qp_type = IB_QPT_RC; init_attr.send_cq = queue->ib_cq; init_attr.recv_cq = queue->ib_cq; + if (queue->pi_support) + init_attr.create_flags |= IB_QP_CREATE_INTEGRITY_EN; ret = rdma_create_qp(queue->cm_id, dev->pd, &init_attr); @@ -290,6 +303,12 @@ static int nvme_rdma_init_request(struct blk_mq_tag_set *set, if (!req->sqe.data) return -ENOMEM; + /* metadata nvme_rdma_sgl struct is located after command's data SGL */ + if (queue->pi_support) + req->metadata_sgl = (void *)nvme_req(rq) + + sizeof(struct nvme_rdma_request) + + NVME_RDMA_DATA_SGL_SIZE; + req->queue = queue; return 0; @@ -400,6 +419,8 @@ static void nvme_rdma_destroy_queue_ib(struct nvme_rdma_queue *queue) dev = queue->device; ibdev = dev->dev; + if (queue->pi_support) + ib_mr_pool_destroy(queue->qp, &queue->qp->sig_mrs); ib_mr_pool_destroy(queue->qp, &queue->qp->rdma_mrs); /* @@ -416,10 +437,16 @@ static void nvme_rdma_destroy_queue_ib(struct nvme_rdma_queue *queue) nvme_rdma_dev_put(dev); } -static int nvme_rdma_get_max_fr_pages(struct ib_device *ibdev) +static int nvme_rdma_get_max_fr_pages(struct ib_device *ibdev, bool pi_support) { - return min_t(u32, NVME_RDMA_MAX_SEGMENTS, - ibdev->attrs.max_fast_reg_page_list_len - 1); + u32 max_page_list_len; + + if (pi_support) + max_page_list_len = ibdev->attrs.max_pi_fast_reg_page_list_len; + else + max_page_list_len = ibdev->attrs.max_fast_reg_page_list_len; + + return min_t(u32, NVME_RDMA_MAX_SEGMENTS, max_page_list_len - 1); } static int nvme_rdma_create_queue_ib(struct nvme_rdma_queue *queue) @@ -476,7 +503,7 @@ static int nvme_rdma_create_queue_ib(struct nvme_rdma_queue *queue) * misaligned we'll end up using two entries for a single data page, * so one additional entry is required. */ - pages_per_mr = nvme_rdma_get_max_fr_pages(ibdev) + 1; + pages_per_mr = nvme_rdma_get_max_fr_pages(ibdev, queue->pi_support) + 1; ret = ib_mr_pool_init(queue->qp, &queue->qp->rdma_mrs, queue->queue_size, IB_MR_TYPE_MEM_REG, @@ -488,10 +515,24 @@ static int nvme_rdma_create_queue_ib(struct nvme_rdma_queue *queue) goto out_destroy_ring; } + if (queue->pi_support) { + ret = ib_mr_pool_init(queue->qp, &queue->qp->sig_mrs, + queue->queue_size, IB_MR_TYPE_INTEGRITY, + pages_per_mr, pages_per_mr); + if (ret) { + dev_err(queue->ctrl->ctrl.device, + "failed to initialize PI MR pool sized %d for QID %d\n", + queue->queue_size, idx); + goto out_destroy_mr_pool; + } + } + set_bit(NVME_RDMA_Q_TR_READY, &queue->flags); return 0; +out_destroy_mr_pool: + ib_mr_pool_destroy(queue->qp, &queue->qp->rdma_mrs); out_destroy_ring: nvme_rdma_free_ring(ibdev, queue->rsp_ring, queue->queue_size, sizeof(struct nvme_completion), DMA_FROM_DEVICE); @@ -513,6 +554,10 @@ static int nvme_rdma_alloc_queue(struct nvme_rdma_ctrl *ctrl, queue = &ctrl->queues[idx]; queue->ctrl = ctrl; + if (idx && ctrl->ctrl.max_integrity_segments) + queue->pi_support = true; + else + queue->pi_support = false; init_completion(&queue->cm_done); if (idx > 0) @@ -723,7 +768,7 @@ static struct blk_mq_tag_set *nvme_rdma_alloc_tagset(struct nvme_ctrl *nctrl, set->reserved_tags = 2; /* connect + keep-alive */ set->numa_node = nctrl->numa_node; set->cmd_size = sizeof(struct nvme_rdma_request) + - NVME_INLINE_SG_CNT * sizeof(struct scatterlist); + NVME_RDMA_DATA_SGL_SIZE; set->driver_data = ctrl; set->nr_hw_queues = 1; set->timeout = ADMIN_TIMEOUT; @@ -737,7 +782,10 @@ static struct blk_mq_tag_set *nvme_rdma_alloc_tagset(struct nvme_ctrl *nctrl, set->numa_node = nctrl->numa_node; set->flags = BLK_MQ_F_SHOULD_MERGE; set->cmd_size = sizeof(struct nvme_rdma_request) + - NVME_INLINE_SG_CNT * sizeof(struct scatterlist); + NVME_RDMA_DATA_SGL_SIZE; + if (nctrl->max_integrity_segments) + set->cmd_size += sizeof(struct nvme_rdma_sgl) + + NVME_RDMA_METADATA_SGL_SIZE; set->driver_data = ctrl; set->nr_hw_queues = nctrl->queue_count - 1; set->timeout = NVME_IO_TIMEOUT; @@ -770,6 +818,7 @@ static void nvme_rdma_destroy_admin_queue(struct nvme_rdma_ctrl *ctrl, static int nvme_rdma_configure_admin_queue(struct nvme_rdma_ctrl *ctrl, bool new) { + bool pi_capable = false; int error; error = nvme_rdma_alloc_queue(ctrl, 0, NVME_AQ_DEPTH); @@ -779,7 +828,13 @@ static int nvme_rdma_configure_admin_queue(struct nvme_rdma_ctrl *ctrl, ctrl->device = ctrl->queues[0].device; ctrl->ctrl.numa_node = dev_to_node(ctrl->device->dev->dma_device); - ctrl->max_fr_pages = nvme_rdma_get_max_fr_pages(ctrl->device->dev); + /* T10-PI support */ + if (ctrl->device->dev->attrs.device_cap_flags & + IB_DEVICE_INTEGRITY_HANDOVER) + pi_capable = true; + + ctrl->max_fr_pages = nvme_rdma_get_max_fr_pages(ctrl->device->dev, + pi_capable); /* * Bind the async event SQE DMA mapping to the admin queue lifetime. @@ -821,6 +876,10 @@ static int nvme_rdma_configure_admin_queue(struct nvme_rdma_ctrl *ctrl, ctrl->ctrl.max_segments = ctrl->max_fr_pages; ctrl->ctrl.max_hw_sectors = ctrl->max_fr_pages << (ilog2(SZ_4K) - 9); + if (pi_capable) + ctrl->ctrl.max_integrity_segments = ctrl->max_fr_pages; + else + ctrl->ctrl.max_integrity_segments = 0; blk_mq_unquiesce_queue(ctrl->ctrl.admin_q); @@ -1149,17 +1208,29 @@ static void nvme_rdma_unmap_data(struct nvme_rdma_queue *queue, struct nvme_rdma_request *req = blk_mq_rq_to_pdu(rq); struct nvme_rdma_device *dev = queue->device; struct ib_device *ibdev = dev->dev; + struct list_head *pool = &queue->qp->rdma_mrs; if (!blk_rq_nr_phys_segments(rq)) return; + if (blk_integrity_rq(rq)) { + ib_dma_unmap_sg(ibdev, req->metadata_sgl->sg_table.sgl, + req->metadata_sgl->nents, rq_dma_dir(rq)); + sg_free_table_chained(&req->metadata_sgl->sg_table, + NVME_INLINE_METADATA_SG_CNT); + } + + if (req->use_sig_mr) + pool = &queue->qp->sig_mrs; + if (req->mr) { - ib_mr_pool_put(queue->qp, &queue->qp->rdma_mrs, req->mr); + ib_mr_pool_put(queue->qp, pool, req->mr); req->mr = NULL; } - ib_dma_unmap_sg(ibdev, req->sg_table.sgl, req->nents, rq_dma_dir(rq)); - sg_free_table_chained(&req->sg_table, NVME_INLINE_SG_CNT); + ib_dma_unmap_sg(ibdev, req->data_sgl.sg_table.sgl, req->data_sgl.nents, + rq_dma_dir(rq)); + sg_free_table_chained(&req->data_sgl.sg_table, NVME_INLINE_SG_CNT); } static int nvme_rdma_set_sg_null(struct nvme_command *c) @@ -1178,7 +1249,7 @@ static int nvme_rdma_map_sg_inline(struct nvme_rdma_queue *queue, int count) { struct nvme_sgl_desc *sg = &c->common.dptr.sgl; - struct scatterlist *sgl = req->sg_table.sgl; + struct scatterlist *sgl = req->data_sgl.sg_table.sgl; struct ib_sge *sge = &req->sge[1]; u32 len = 0; int i; @@ -1203,8 +1274,8 @@ static int nvme_rdma_map_sg_single(struct nvme_rdma_queue *queue, { struct nvme_keyed_sgl_desc *sg = &c->common.dptr.ksgl; - sg->addr = cpu_to_le64(sg_dma_address(req->sg_table.sgl)); - put_unaligned_le24(sg_dma_len(req->sg_table.sgl), sg->length); + sg->addr = cpu_to_le64(sg_dma_address(req->data_sgl.sg_table.sgl)); + put_unaligned_le24(sg_dma_len(req->data_sgl.sg_table.sgl), sg->length); put_unaligned_le32(queue->device->pd->unsafe_global_rkey, sg->key); sg->type = NVME_KEY_SGL_FMT_DATA_DESC << 4; return 0; @@ -1225,7 +1296,8 @@ static int nvme_rdma_map_sg_fr(struct nvme_rdma_queue *queue, * Align the MR to a 4K page size to match the ctrl page size and * the block virtual boundary. */ - nr = ib_map_mr_sg(req->mr, req->sg_table.sgl, count, NULL, SZ_4K); + nr = ib_map_mr_sg(req->mr, req->data_sgl.sg_table.sgl, count, NULL, + SZ_4K); if (unlikely(nr < count)) { ib_mr_pool_put(queue->qp, &queue->qp->rdma_mrs, req->mr); req->mr = NULL; @@ -1256,12 +1328,125 @@ static int nvme_rdma_map_sg_fr(struct nvme_rdma_queue *queue, return 0; } +static void nvme_rdma_set_sig_domain(struct blk_integrity *bi, + struct nvme_command *cmd, struct ib_sig_domain *domain, + u16 control, u8 pi_type) +{ + domain->sig_type = IB_SIG_TYPE_T10_DIF; + domain->sig.dif.bg_type = IB_T10DIF_CRC; + domain->sig.dif.pi_interval = 1 << bi->interval_exp; + domain->sig.dif.ref_tag = le32_to_cpu(cmd->rw.reftag); + if (control & NVME_RW_PRINFO_PRCHK_REF) + domain->sig.dif.ref_remap = true; + + domain->sig.dif.app_tag = le16_to_cpu(cmd->rw.apptag); + domain->sig.dif.apptag_check_mask = le16_to_cpu(cmd->rw.appmask); + domain->sig.dif.app_escape = true; + if (pi_type == NVME_NS_DPS_PI_TYPE3) + domain->sig.dif.ref_escape = true; +} + +static void nvme_rdma_set_sig_attrs(struct blk_integrity *bi, + struct nvme_command *cmd, struct ib_sig_attrs *sig_attrs, + u8 pi_type) +{ + u16 control = le16_to_cpu(cmd->rw.control); + + memset(sig_attrs, 0, sizeof(*sig_attrs)); + if (control & NVME_RW_PRINFO_PRACT) { + /* for WRITE_INSERT/READ_STRIP no memory domain */ + sig_attrs->mem.sig_type = IB_SIG_TYPE_NONE; + nvme_rdma_set_sig_domain(bi, cmd, &sig_attrs->wire, control, + pi_type); + /* Clear the PRACT bit since HCA will generate/verify the PI */ + control &= ~NVME_RW_PRINFO_PRACT; + cmd->rw.control = cpu_to_le16(control); + } else { + /* for WRITE_PASS/READ_PASS both wire/memory domains exist */ + nvme_rdma_set_sig_domain(bi, cmd, &sig_attrs->wire, control, + pi_type); + nvme_rdma_set_sig_domain(bi, cmd, &sig_attrs->mem, control, + pi_type); + } +} + +static void nvme_rdma_set_prot_checks(struct nvme_command *cmd, u8 *mask) +{ + *mask = 0; + if (le16_to_cpu(cmd->rw.control) & NVME_RW_PRINFO_PRCHK_REF) + *mask |= IB_SIG_CHECK_REFTAG; + if (le16_to_cpu(cmd->rw.control) & NVME_RW_PRINFO_PRCHK_GUARD) + *mask |= IB_SIG_CHECK_GUARD; +} + +static void nvme_rdma_sig_done(struct ib_cq *cq, struct ib_wc *wc) +{ + if (unlikely(wc->status != IB_WC_SUCCESS)) + nvme_rdma_wr_error(cq, wc, "SIG"); +} + +static int nvme_rdma_map_sg_pi(struct nvme_rdma_queue *queue, + struct nvme_rdma_request *req, struct nvme_command *c, + int count, int pi_count) +{ + struct nvme_rdma_sgl *sgl = &req->data_sgl; + struct ib_reg_wr *wr = &req->reg_wr; + struct request *rq = blk_mq_rq_from_pdu(req); + struct nvme_ns *ns = rq->q->queuedata; + struct bio *bio = rq->bio; + struct nvme_keyed_sgl_desc *sg = &c->common.dptr.ksgl; + int nr; + + req->mr = ib_mr_pool_get(queue->qp, &queue->qp->sig_mrs); + if (WARN_ON_ONCE(!req->mr)) + return -EAGAIN; + + nr = ib_map_mr_sg_pi(req->mr, sgl->sg_table.sgl, count, NULL, + req->metadata_sgl->sg_table.sgl, pi_count, NULL, + SZ_4K); + if (unlikely(nr)) + goto mr_put; + + nvme_rdma_set_sig_attrs(blk_get_integrity(bio->bi_disk), c, + req->mr->sig_attrs, ns->pi_type); + nvme_rdma_set_prot_checks(c, &req->mr->sig_attrs->check_mask); + + ib_update_fast_reg_key(req->mr, ib_inc_rkey(req->mr->rkey)); + + req->reg_cqe.done = nvme_rdma_sig_done; + memset(wr, 0, sizeof(*wr)); + wr->wr.opcode = IB_WR_REG_MR_INTEGRITY; + wr->wr.wr_cqe = &req->reg_cqe; + wr->wr.num_sge = 0; + wr->wr.send_flags = 0; + wr->mr = req->mr; + wr->key = req->mr->rkey; + wr->access = IB_ACCESS_LOCAL_WRITE | + IB_ACCESS_REMOTE_READ | + IB_ACCESS_REMOTE_WRITE; + + sg->addr = cpu_to_le64(req->mr->iova); + put_unaligned_le24(req->mr->length, sg->length); + put_unaligned_le32(req->mr->rkey, sg->key); + sg->type = NVME_KEY_SGL_FMT_DATA_DESC << 4; + + return 0; + +mr_put: + ib_mr_pool_put(queue->qp, &queue->qp->sig_mrs, req->mr); + req->mr = NULL; + if (nr < 0) + return nr; + return -EINVAL; +} + static int nvme_rdma_map_data(struct nvme_rdma_queue *queue, struct request *rq, struct nvme_command *c) { struct nvme_rdma_request *req = blk_mq_rq_to_pdu(rq); struct nvme_rdma_device *dev = queue->device; struct ib_device *ibdev = dev->dev; + int pi_count = 0; int count, ret; req->num_sge = 1; @@ -1272,22 +1457,52 @@ static int nvme_rdma_map_data(struct nvme_rdma_queue *queue, if (!blk_rq_nr_phys_segments(rq)) return nvme_rdma_set_sg_null(c); - req->sg_table.sgl = req->first_sgl; - ret = sg_alloc_table_chained(&req->sg_table, - blk_rq_nr_phys_segments(rq), req->sg_table.sgl, + req->data_sgl.sg_table.sgl = (struct scatterlist *)(req + 1); + ret = sg_alloc_table_chained(&req->data_sgl.sg_table, + blk_rq_nr_phys_segments(rq), req->data_sgl.sg_table.sgl, NVME_INLINE_SG_CNT); if (ret) return -ENOMEM; - req->nents = blk_rq_map_sg(rq->q, rq, req->sg_table.sgl); + req->data_sgl.nents = blk_rq_map_sg(rq->q, rq, + req->data_sgl.sg_table.sgl); - count = ib_dma_map_sg(ibdev, req->sg_table.sgl, req->nents, - rq_dma_dir(rq)); + count = ib_dma_map_sg(ibdev, req->data_sgl.sg_table.sgl, + req->data_sgl.nents, rq_dma_dir(rq)); if (unlikely(count <= 0)) { ret = -EIO; goto out_free_table; } + if (blk_integrity_rq(rq)) { + req->metadata_sgl->sg_table.sgl = + (struct scatterlist *)(req->metadata_sgl + 1); + ret = sg_alloc_table_chained(&req->metadata_sgl->sg_table, + blk_rq_count_integrity_sg(rq->q, rq->bio), + req->metadata_sgl->sg_table.sgl, + NVME_INLINE_METADATA_SG_CNT); + if (unlikely(ret)) { + ret = -ENOMEM; + goto out_unmap_sg; + } + + req->metadata_sgl->nents = blk_rq_map_integrity_sg(rq->q, + rq->bio, req->metadata_sgl->sg_table.sgl); + pi_count = ib_dma_map_sg(ibdev, + req->metadata_sgl->sg_table.sgl, + req->metadata_sgl->nents, + rq_dma_dir(rq)); + if (unlikely(pi_count <= 0)) { + ret = -EIO; + goto out_free_pi_table; + } + } + + if (req->use_sig_mr) { + ret = nvme_rdma_map_sg_pi(queue, req, c, count, pi_count); + goto out; + } + if (count <= dev->num_inline_segments) { if (rq_data_dir(rq) == WRITE && nvme_rdma_queue_idx(queue) && queue->ctrl->use_inline_data && @@ -1306,14 +1521,23 @@ static int nvme_rdma_map_data(struct nvme_rdma_queue *queue, ret = nvme_rdma_map_sg_fr(queue, req, c, count); out: if (unlikely(ret)) - goto out_unmap_sg; + goto out_unmap_pi_sg; return 0; +out_unmap_pi_sg: + if (blk_integrity_rq(rq)) + ib_dma_unmap_sg(ibdev, req->metadata_sgl->sg_table.sgl, + req->metadata_sgl->nents, rq_dma_dir(rq)); +out_free_pi_table: + if (blk_integrity_rq(rq)) + sg_free_table_chained(&req->metadata_sgl->sg_table, + NVME_INLINE_METADATA_SG_CNT); out_unmap_sg: - ib_dma_unmap_sg(ibdev, req->sg_table.sgl, req->nents, rq_dma_dir(rq)); + ib_dma_unmap_sg(ibdev, req->data_sgl.sg_table.sgl, req->data_sgl.nents, + rq_dma_dir(rq)); out_free_table: - sg_free_table_chained(&req->sg_table, NVME_INLINE_SG_CNT); + sg_free_table_chained(&req->data_sgl.sg_table, NVME_INLINE_SG_CNT); return ret; } @@ -1761,6 +1985,15 @@ static blk_status_t nvme_rdma_queue_rq(struct blk_mq_hw_ctx *hctx, blk_mq_start_request(rq); + if (IS_ENABLED(CONFIG_BLK_DEV_INTEGRITY) && + queue->pi_support && + (c->common.opcode == nvme_cmd_write || + c->common.opcode == nvme_cmd_read) && + nvme_ns_has_pi(ns)) + req->use_sig_mr = true; + else + req->use_sig_mr = false; + err = nvme_rdma_map_data(queue, rq, c); if (unlikely(err < 0)) { dev_err(queue->ctrl->ctrl.device, @@ -1801,12 +2034,46 @@ static int nvme_rdma_poll(struct blk_mq_hw_ctx *hctx) return ib_process_cq_direct(queue->ib_cq, -1); } +static void nvme_rdma_check_pi_status(struct nvme_rdma_request *req) +{ + struct request *rq = blk_mq_rq_from_pdu(req); + struct ib_mr_status mr_status; + int ret; + + ret = ib_check_mr_status(req->mr, IB_MR_CHECK_SIG_STATUS, &mr_status); + if (ret) { + pr_err("ib_check_mr_status failed, ret %d\n", ret); + nvme_req(rq)->status = NVME_SC_INVALID_PI; + return; + } + + if (mr_status.fail_status & IB_MR_CHECK_SIG_STATUS) { + switch (mr_status.sig_err.err_type) { + case IB_SIG_BAD_GUARD: + nvme_req(rq)->status = NVME_SC_GUARD_CHECK; + break; + case IB_SIG_BAD_REFTAG: + nvme_req(rq)->status = NVME_SC_REFTAG_CHECK; + break; + case IB_SIG_BAD_APPTAG: + nvme_req(rq)->status = NVME_SC_APPTAG_CHECK; + break; + } + pr_err("PI error found type %d expected 0x%x vs actual 0x%x\n", + mr_status.sig_err.err_type, mr_status.sig_err.expected, + mr_status.sig_err.actual); + } +} + static void nvme_rdma_complete_rq(struct request *rq) { struct nvme_rdma_request *req = blk_mq_rq_to_pdu(rq); struct nvme_rdma_queue *queue = req->queue; struct ib_device *ibdev = queue->device->dev; + if (req->use_sig_mr) + nvme_rdma_check_pi_status(req); + nvme_rdma_unmap_data(queue, rq); ib_dma_unmap_single(ibdev, req->sqe.dma, sizeof(struct nvme_command), DMA_TO_DEVICE); @@ -1926,7 +2193,7 @@ out_fail: static const struct nvme_ctrl_ops nvme_rdma_ctrl_ops = { .name = "rdma", .module = THIS_MODULE, - .flags = NVME_F_FABRICS, + .flags = NVME_F_FABRICS | NVME_F_METADATA_SUPPORTED, .reg_read32 = nvmf_reg_read32, .reg_read64 = nvmf_reg_read64, .reg_write32 = nvmf_reg_write32, diff --git a/drivers/nvme/host/tcp.c b/drivers/nvme/host/tcp.c index c15a92163c1f..1843110ec34f 100644 --- a/drivers/nvme/host/tcp.c +++ b/drivers/nvme/host/tcp.c @@ -60,6 +60,7 @@ struct nvme_tcp_request { enum nvme_tcp_queue_flags { NVME_TCP_Q_ALLOCATED = 0, NVME_TCP_Q_LIVE = 1, + NVME_TCP_Q_POLLING = 2, }; enum nvme_tcp_recv_state { @@ -75,6 +76,7 @@ struct nvme_tcp_queue { int io_cpu; spinlock_t lock; + struct mutex send_mutex; struct list_head send_list; /* recv state */ @@ -131,6 +133,7 @@ static DEFINE_MUTEX(nvme_tcp_ctrl_mutex); static struct workqueue_struct *nvme_tcp_wq; static struct blk_mq_ops nvme_tcp_mq_ops; static struct blk_mq_ops nvme_tcp_admin_mq_ops; +static int nvme_tcp_try_send(struct nvme_tcp_queue *queue); static inline struct nvme_tcp_ctrl *to_tcp_ctrl(struct nvme_ctrl *ctrl) { @@ -257,15 +260,29 @@ static inline void nvme_tcp_advance_req(struct nvme_tcp_request *req, } } -static inline void nvme_tcp_queue_request(struct nvme_tcp_request *req) +static inline void nvme_tcp_queue_request(struct nvme_tcp_request *req, + bool sync) { struct nvme_tcp_queue *queue = req->queue; + bool empty; spin_lock(&queue->lock); + empty = list_empty(&queue->send_list) && !queue->request; list_add_tail(&req->entry, &queue->send_list); spin_unlock(&queue->lock); - queue_work_on(queue->io_cpu, nvme_tcp_wq, &queue->io_work); + /* + * if we're the first on the send_list and we can try to send + * directly, otherwise queue io_work. Also, only do that if we + * are on the same cpu, so we don't introduce contention. + */ + if (queue->io_cpu == smp_processor_id() && + sync && empty && mutex_trylock(&queue->send_mutex)) { + nvme_tcp_try_send(queue); + mutex_unlock(&queue->send_mutex); + } else { + queue_work_on(queue->io_cpu, nvme_tcp_wq, &queue->io_work); + } } static inline struct nvme_tcp_request * @@ -578,7 +595,7 @@ static int nvme_tcp_handle_r2t(struct nvme_tcp_queue *queue, req->state = NVME_TCP_SEND_H2C_PDU; req->offset = 0; - nvme_tcp_queue_request(req); + nvme_tcp_queue_request(req, false); return 0; } @@ -794,11 +811,12 @@ static void nvme_tcp_data_ready(struct sock *sk) { struct nvme_tcp_queue *queue; - read_lock(&sk->sk_callback_lock); + read_lock_bh(&sk->sk_callback_lock); queue = sk->sk_user_data; - if (likely(queue && queue->rd_enabled)) + if (likely(queue && queue->rd_enabled) && + !test_bit(NVME_TCP_Q_POLLING, &queue->flags)) queue_work_on(queue->io_cpu, nvme_tcp_wq, &queue->io_work); - read_unlock(&sk->sk_callback_lock); + read_unlock_bh(&sk->sk_callback_lock); } static void nvme_tcp_write_space(struct sock *sk) @@ -867,7 +885,7 @@ static int nvme_tcp_try_send_data(struct nvme_tcp_request *req) if (last && !queue->data_digest) flags |= MSG_EOR; else - flags |= MSG_MORE; + flags |= MSG_MORE | MSG_SENDPAGE_NOTLAST; /* can't zcopy slab pages */ if (unlikely(PageSlab(page))) { @@ -906,11 +924,16 @@ static int nvme_tcp_try_send_cmd_pdu(struct nvme_tcp_request *req) struct nvme_tcp_queue *queue = req->queue; struct nvme_tcp_cmd_pdu *pdu = req->pdu; bool inline_data = nvme_tcp_has_inline_data(req); - int flags = MSG_DONTWAIT | (inline_data ? MSG_MORE : MSG_EOR); u8 hdgst = nvme_tcp_hdgst_len(queue); int len = sizeof(*pdu) + hdgst - req->offset; + int flags = MSG_DONTWAIT; int ret; + if (inline_data) + flags |= MSG_MORE | MSG_SENDPAGE_NOTLAST; + else + flags |= MSG_EOR; + if (queue->hdr_digest && !req->offset) nvme_tcp_hdgst(queue->snd_hash, pdu, sizeof(*pdu)); @@ -949,7 +972,7 @@ static int nvme_tcp_try_send_data_pdu(struct nvme_tcp_request *req) ret = kernel_sendpage(queue->sock, virt_to_page(pdu), offset_in_page(pdu) + req->offset, len, - MSG_DONTWAIT | MSG_MORE); + MSG_DONTWAIT | MSG_MORE | MSG_SENDPAGE_NOTLAST); if (unlikely(ret <= 0)) return ret; @@ -1063,11 +1086,14 @@ static void nvme_tcp_io_work(struct work_struct *w) bool pending = false; int result; - result = nvme_tcp_try_send(queue); - if (result > 0) - pending = true; - else if (unlikely(result < 0)) - break; + if (mutex_trylock(&queue->send_mutex)) { + result = nvme_tcp_try_send(queue); + mutex_unlock(&queue->send_mutex); + if (result > 0) + pending = true; + else if (unlikely(result < 0)) + break; + } result = nvme_tcp_try_recv(queue); if (result > 0) @@ -1313,12 +1339,12 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl, { struct nvme_tcp_ctrl *ctrl = to_tcp_ctrl(nctrl); struct nvme_tcp_queue *queue = &ctrl->queues[qid]; - struct linger sol = { .l_onoff = 1, .l_linger = 0 }; - int ret, opt, rcv_pdu_size; + int ret, rcv_pdu_size; queue->ctrl = ctrl; INIT_LIST_HEAD(&queue->send_list); spin_lock_init(&queue->lock); + mutex_init(&queue->send_mutex); INIT_WORK(&queue->io_work, nvme_tcp_io_work); queue->queue_size = queue_size; @@ -1337,60 +1363,24 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl, } /* Single syn retry */ - opt = 1; - ret = kernel_setsockopt(queue->sock, IPPROTO_TCP, TCP_SYNCNT, - (char *)&opt, sizeof(opt)); - if (ret) { - dev_err(nctrl->device, - "failed to set TCP_SYNCNT sock opt %d\n", ret); - goto err_sock; - } + tcp_sock_set_syncnt(queue->sock->sk, 1); /* Set TCP no delay */ - opt = 1; - ret = kernel_setsockopt(queue->sock, IPPROTO_TCP, - TCP_NODELAY, (char *)&opt, sizeof(opt)); - if (ret) { - dev_err(nctrl->device, - "failed to set TCP_NODELAY sock opt %d\n", ret); - goto err_sock; - } + tcp_sock_set_nodelay(queue->sock->sk); /* * Cleanup whatever is sitting in the TCP transmit queue on socket * close. This is done to prevent stale data from being sent should * the network connection be restored before TCP times out. */ - ret = kernel_setsockopt(queue->sock, SOL_SOCKET, SO_LINGER, - (char *)&sol, sizeof(sol)); - if (ret) { - dev_err(nctrl->device, - "failed to set SO_LINGER sock opt %d\n", ret); - goto err_sock; - } + sock_no_linger(queue->sock->sk); - if (so_priority > 0) { - ret = kernel_setsockopt(queue->sock, SOL_SOCKET, SO_PRIORITY, - (char *)&so_priority, sizeof(so_priority)); - if (ret) { - dev_err(ctrl->ctrl.device, - "failed to set SO_PRIORITY sock opt, ret %d\n", - ret); - goto err_sock; - } - } + if (so_priority > 0) + sock_set_priority(queue->sock->sk, so_priority); /* Set socket type of service */ - if (nctrl->opts->tos >= 0) { - opt = nctrl->opts->tos; - ret = kernel_setsockopt(queue->sock, SOL_IP, IP_TOS, - (char *)&opt, sizeof(opt)); - if (ret) { - dev_err(nctrl->device, - "failed to set IP_TOS sock opt %d\n", ret); - goto err_sock; - } - } + if (nctrl->opts->tos >= 0) + ip_sock_set_tos(queue->sock->sk, nctrl->opts->tos); queue->sock->sk->sk_allocation = GFP_ATOMIC; nvme_tcp_set_queue_io_cpu(queue); @@ -1543,6 +1533,7 @@ static struct blk_mq_tag_set *nvme_tcp_alloc_tagset(struct nvme_ctrl *nctrl, set->queue_depth = NVME_AQ_MQ_TAG_DEPTH; set->reserved_tags = 2; /* connect + keep-alive */ set->numa_node = NUMA_NO_NODE; + set->flags = BLK_MQ_F_BLOCKING; set->cmd_size = sizeof(struct nvme_tcp_request); set->driver_data = ctrl; set->nr_hw_queues = 1; @@ -1554,7 +1545,7 @@ static struct blk_mq_tag_set *nvme_tcp_alloc_tagset(struct nvme_ctrl *nctrl, set->queue_depth = nctrl->sqsize + 1; set->reserved_tags = 1; /* fabric connect */ set->numa_node = NUMA_NO_NODE; - set->flags = BLK_MQ_F_SHOULD_MERGE; + set->flags = BLK_MQ_F_SHOULD_MERGE | BLK_MQ_F_BLOCKING; set->cmd_size = sizeof(struct nvme_tcp_request); set->driver_data = ctrl; set->nr_hw_queues = nctrl->queue_count - 1; @@ -2113,7 +2104,7 @@ static void nvme_tcp_submit_async_event(struct nvme_ctrl *arg) ctrl->async_req.curr_bio = NULL; ctrl->async_req.data_len = 0; - nvme_tcp_queue_request(&ctrl->async_req); + nvme_tcp_queue_request(&ctrl->async_req, true); } static enum blk_eh_timer_return @@ -2244,7 +2235,7 @@ static blk_status_t nvme_tcp_queue_rq(struct blk_mq_hw_ctx *hctx, blk_mq_start_request(rq); - nvme_tcp_queue_request(req); + nvme_tcp_queue_request(req, true); return BLK_STS_OK; } @@ -2302,9 +2293,11 @@ static int nvme_tcp_poll(struct blk_mq_hw_ctx *hctx) if (!test_bit(NVME_TCP_Q_LIVE, &queue->flags)) return 0; + set_bit(NVME_TCP_Q_POLLING, &queue->flags); if (sk_can_busy_loop(sk) && skb_queue_empty_lockless(&sk->sk_receive_queue)) sk_busy_loop(sk, true); nvme_tcp_try_recv(queue); + clear_bit(NVME_TCP_Q_POLLING, &queue->flags); return queue->nr_cqe; } diff --git a/drivers/nvme/target/Kconfig b/drivers/nvme/target/Kconfig index d7f48c0fb311..4474952d64c6 100644 --- a/drivers/nvme/target/Kconfig +++ b/drivers/nvme/target/Kconfig @@ -4,6 +4,7 @@ config NVME_TARGET tristate "NVMe Target support" depends on BLOCK depends on CONFIGFS_FS + select BLK_DEV_INTEGRITY_T10 if BLK_DEV_INTEGRITY select SGL_ALLOC help This enabled target side support for the NVMe protocol, that is diff --git a/drivers/nvme/target/admin-cmd.c b/drivers/nvme/target/admin-cmd.c index 9d6f75cfa77c..1db8c0498668 100644 --- a/drivers/nvme/target/admin-cmd.c +++ b/drivers/nvme/target/admin-cmd.c @@ -295,7 +295,7 @@ out: static void nvmet_execute_get_log_page(struct nvmet_req *req) { - if (!nvmet_check_data_len(req, nvmet_get_log_page_len(req->cmd))) + if (!nvmet_check_transfer_len(req, nvmet_get_log_page_len(req->cmd))) return; switch (req->cmd->get_log_page.lid) { @@ -341,6 +341,7 @@ static void nvmet_execute_identify_ctrl(struct nvmet_req *req) { struct nvmet_ctrl *ctrl = req->sq->ctrl; struct nvme_id_ctrl *id; + u32 cmd_capsule_size; u16 status = 0; id = kzalloc(sizeof(*id), GFP_KERNEL); @@ -433,9 +434,15 @@ static void nvmet_execute_identify_ctrl(struct nvmet_req *req) strlcpy(id->subnqn, ctrl->subsys->subsysnqn, sizeof(id->subnqn)); - /* Max command capsule size is sqe + single page of in-capsule data */ - id->ioccsz = cpu_to_le32((sizeof(struct nvme_command) + - req->port->inline_data_size) / 16); + /* + * Max command capsule size is sqe + in-capsule data size. + * Disable in-capsule data for Metadata capable controllers. + */ + cmd_capsule_size = sizeof(struct nvme_command); + if (!ctrl->pi_support) + cmd_capsule_size += req->port->inline_data_size; + id->ioccsz = cpu_to_le32(cmd_capsule_size / 16); + /* Max response capsule size is cqe */ id->iorcsz = cpu_to_le32(sizeof(struct nvme_completion) / 16); @@ -465,6 +472,7 @@ out: static void nvmet_execute_identify_ns(struct nvmet_req *req) { + struct nvmet_ctrl *ctrl = req->sq->ctrl; struct nvmet_ns *ns; struct nvme_id_ns *id; u16 status = 0; @@ -482,10 +490,12 @@ static void nvmet_execute_identify_ns(struct nvmet_req *req) } /* return an all zeroed buffer if we can't find an active namespace */ - ns = nvmet_find_namespace(req->sq->ctrl, req->cmd->identify.nsid); + ns = nvmet_find_namespace(ctrl, req->cmd->identify.nsid); if (!ns) goto done; + nvmet_ns_revalidate(ns); + /* * nuse = ncap = nsze isn't always true, but we have no way to find * that out from the underlying device. @@ -521,6 +531,16 @@ static void nvmet_execute_identify_ns(struct nvmet_req *req) id->lbaf[0].ds = ns->blksize_shift; + if (ctrl->pi_support && nvmet_ns_has_pi(ns)) { + id->dpc = NVME_NS_DPC_PI_FIRST | NVME_NS_DPC_PI_LAST | + NVME_NS_DPC_PI_TYPE1 | NVME_NS_DPC_PI_TYPE2 | + NVME_NS_DPC_PI_TYPE3; + id->mc = NVME_MC_EXTENDED_LBA; + id->dps = ns->pi_type; + id->flbas = NVME_NS_FLBAS_META_EXT; + id->lbaf[0].ms = cpu_to_le16(ns->metadata_size); + } + if (ns->readonly) id->nsattr |= (1 << 0); nvmet_put_namespace(ns); @@ -625,7 +645,7 @@ out: static void nvmet_execute_identify(struct nvmet_req *req) { - if (!nvmet_check_data_len(req, NVME_IDENTIFY_DATA_SIZE)) + if (!nvmet_check_transfer_len(req, NVME_IDENTIFY_DATA_SIZE)) return; switch (req->cmd->identify.cns) { @@ -654,7 +674,7 @@ static void nvmet_execute_identify(struct nvmet_req *req) */ static void nvmet_execute_abort(struct nvmet_req *req) { - if (!nvmet_check_data_len(req, 0)) + if (!nvmet_check_transfer_len(req, 0)) return; nvmet_set_result(req, 1); nvmet_req_complete(req, 0); @@ -743,7 +763,7 @@ static void nvmet_execute_set_features(struct nvmet_req *req) u16 nsqr; u16 ncqr; - if (!nvmet_check_data_len(req, 0)) + if (!nvmet_check_transfer_len(req, 0)) return; switch (cdw10 & 0xff) { @@ -815,7 +835,7 @@ static void nvmet_execute_get_features(struct nvmet_req *req) u32 cdw10 = le32_to_cpu(req->cmd->common.cdw10); u16 status = 0; - if (!nvmet_check_data_len(req, nvmet_feat_data_len(req, cdw10))) + if (!nvmet_check_transfer_len(req, nvmet_feat_data_len(req, cdw10))) return; switch (cdw10 & 0xff) { @@ -882,7 +902,7 @@ void nvmet_execute_async_event(struct nvmet_req *req) { struct nvmet_ctrl *ctrl = req->sq->ctrl; - if (!nvmet_check_data_len(req, 0)) + if (!nvmet_check_transfer_len(req, 0)) return; mutex_lock(&ctrl->lock); @@ -901,7 +921,7 @@ void nvmet_execute_keep_alive(struct nvmet_req *req) { struct nvmet_ctrl *ctrl = req->sq->ctrl; - if (!nvmet_check_data_len(req, 0)) + if (!nvmet_check_transfer_len(req, 0)) return; pr_debug("ctrl %d update keep-alive timer for %d secs\n", diff --git a/drivers/nvme/target/configfs.c b/drivers/nvme/target/configfs.c index 58cabd7b6fc5..419e0d4ce79b 100644 --- a/drivers/nvme/target/configfs.c +++ b/drivers/nvme/target/configfs.c @@ -20,61 +20,71 @@ static const struct config_item_type nvmet_subsys_type; static LIST_HEAD(nvmet_ports_list); struct list_head *nvmet_ports = &nvmet_ports_list; -static const struct nvmet_transport_name { +struct nvmet_type_name_map { u8 type; const char *name; -} nvmet_transport_names[] = { +}; + +static struct nvmet_type_name_map nvmet_transport[] = { { NVMF_TRTYPE_RDMA, "rdma" }, { NVMF_TRTYPE_FC, "fc" }, { NVMF_TRTYPE_TCP, "tcp" }, { NVMF_TRTYPE_LOOP, "loop" }, }; +static const struct nvmet_type_name_map nvmet_addr_family[] = { + { NVMF_ADDR_FAMILY_PCI, "pcie" }, + { NVMF_ADDR_FAMILY_IP4, "ipv4" }, + { NVMF_ADDR_FAMILY_IP6, "ipv6" }, + { NVMF_ADDR_FAMILY_IB, "ib" }, + { NVMF_ADDR_FAMILY_FC, "fc" }, + { NVMF_ADDR_FAMILY_LOOP, "loop" }, +}; + +static bool nvmet_is_port_enabled(struct nvmet_port *p, const char *caller) +{ + if (p->enabled) + pr_err("Disable port '%u' before changing attribute in %s\n", + le16_to_cpu(p->disc_addr.portid), caller); + return p->enabled; +} + /* * nvmet_port Generic ConfigFS definitions. * Used in any place in the ConfigFS tree that refers to an address. */ -static ssize_t nvmet_addr_adrfam_show(struct config_item *item, - char *page) +static ssize_t nvmet_addr_adrfam_show(struct config_item *item, char *page) { - switch (to_nvmet_port(item)->disc_addr.adrfam) { - case NVMF_ADDR_FAMILY_IP4: - return sprintf(page, "ipv4\n"); - case NVMF_ADDR_FAMILY_IP6: - return sprintf(page, "ipv6\n"); - case NVMF_ADDR_FAMILY_IB: - return sprintf(page, "ib\n"); - case NVMF_ADDR_FAMILY_FC: - return sprintf(page, "fc\n"); - default: - return sprintf(page, "\n"); + u8 adrfam = to_nvmet_port(item)->disc_addr.adrfam; + int i; + + for (i = 1; i < ARRAY_SIZE(nvmet_addr_family); i++) { + if (nvmet_addr_family[i].type == adrfam) + return sprintf(page, "%s\n", nvmet_addr_family[i].name); } + + return sprintf(page, "\n"); } static ssize_t nvmet_addr_adrfam_store(struct config_item *item, const char *page, size_t count) { struct nvmet_port *port = to_nvmet_port(item); + int i; - if (port->enabled) { - pr_err("Cannot modify address while enabled\n"); - pr_err("Disable the address before modifying\n"); + if (nvmet_is_port_enabled(port, __func__)) return -EACCES; - } - if (sysfs_streq(page, "ipv4")) { - port->disc_addr.adrfam = NVMF_ADDR_FAMILY_IP4; - } else if (sysfs_streq(page, "ipv6")) { - port->disc_addr.adrfam = NVMF_ADDR_FAMILY_IP6; - } else if (sysfs_streq(page, "ib")) { - port->disc_addr.adrfam = NVMF_ADDR_FAMILY_IB; - } else if (sysfs_streq(page, "fc")) { - port->disc_addr.adrfam = NVMF_ADDR_FAMILY_FC; - } else { - pr_err("Invalid value '%s' for adrfam\n", page); - return -EINVAL; + for (i = 1; i < ARRAY_SIZE(nvmet_addr_family); i++) { + if (sysfs_streq(page, nvmet_addr_family[i].name)) + goto found; } + pr_err("Invalid value '%s' for adrfam\n", page); + return -EINVAL; + +found: + port->disc_addr.adrfam = nvmet_addr_family[i].type; return count; } @@ -100,11 +110,9 @@ static ssize_t nvmet_addr_portid_store(struct config_item *item, return -EINVAL; } - if (port->enabled) { - pr_err("Cannot modify address while enabled\n"); - pr_err("Disable the address before modifying\n"); + if (nvmet_is_port_enabled(port, __func__)) return -EACCES; - } + port->disc_addr.portid = cpu_to_le16(portid); return count; } @@ -130,11 +138,8 @@ static ssize_t nvmet_addr_traddr_store(struct config_item *item, return -EINVAL; } - if (port->enabled) { - pr_err("Cannot modify address while enabled\n"); - pr_err("Disable the address before modifying\n"); + if (nvmet_is_port_enabled(port, __func__)) return -EACCES; - } if (sscanf(page, "%s\n", port->disc_addr.traddr) != 1) return -EINVAL; @@ -143,20 +148,24 @@ static ssize_t nvmet_addr_traddr_store(struct config_item *item, CONFIGFS_ATTR(nvmet_, addr_traddr); -static ssize_t nvmet_addr_treq_show(struct config_item *item, - char *page) +static const struct nvmet_type_name_map nvmet_addr_treq[] = { + { NVMF_TREQ_NOT_SPECIFIED, "not specified" }, + { NVMF_TREQ_REQUIRED, "required" }, + { NVMF_TREQ_NOT_REQUIRED, "not required" }, +}; + +static ssize_t nvmet_addr_treq_show(struct config_item *item, char *page) { - switch (to_nvmet_port(item)->disc_addr.treq & - NVME_TREQ_SECURE_CHANNEL_MASK) { - case NVMF_TREQ_NOT_SPECIFIED: - return sprintf(page, "not specified\n"); - case NVMF_TREQ_REQUIRED: - return sprintf(page, "required\n"); - case NVMF_TREQ_NOT_REQUIRED: - return sprintf(page, "not required\n"); - default: - return sprintf(page, "\n"); + u8 treq = to_nvmet_port(item)->disc_addr.treq & + NVME_TREQ_SECURE_CHANNEL_MASK; + int i; + + for (i = 0; i < ARRAY_SIZE(nvmet_addr_treq); i++) { + if (treq == nvmet_addr_treq[i].type) + return sprintf(page, "%s\n", nvmet_addr_treq[i].name); } + + return sprintf(page, "\n"); } static ssize_t nvmet_addr_treq_store(struct config_item *item, @@ -164,25 +173,22 @@ static ssize_t nvmet_addr_treq_store(struct config_item *item, { struct nvmet_port *port = to_nvmet_port(item); u8 treq = port->disc_addr.treq & ~NVME_TREQ_SECURE_CHANNEL_MASK; + int i; - if (port->enabled) { - pr_err("Cannot modify address while enabled\n"); - pr_err("Disable the address before modifying\n"); + if (nvmet_is_port_enabled(port, __func__)) return -EACCES; - } - if (sysfs_streq(page, "not specified")) { - treq |= NVMF_TREQ_NOT_SPECIFIED; - } else if (sysfs_streq(page, "required")) { - treq |= NVMF_TREQ_REQUIRED; - } else if (sysfs_streq(page, "not required")) { - treq |= NVMF_TREQ_NOT_REQUIRED; - } else { - pr_err("Invalid value '%s' for treq\n", page); - return -EINVAL; + for (i = 0; i < ARRAY_SIZE(nvmet_addr_treq); i++) { + if (sysfs_streq(page, nvmet_addr_treq[i].name)) + goto found; } - port->disc_addr.treq = treq; + pr_err("Invalid value '%s' for treq\n", page); + return -EINVAL; + +found: + treq |= nvmet_addr_treq[i].type; + port->disc_addr.treq = treq; return count; } @@ -206,11 +212,8 @@ static ssize_t nvmet_addr_trsvcid_store(struct config_item *item, pr_err("Invalid value '%s' for trsvcid\n", page); return -EINVAL; } - if (port->enabled) { - pr_err("Cannot modify address while enabled\n"); - pr_err("Disable the address before modifying\n"); + if (nvmet_is_port_enabled(port, __func__)) return -EACCES; - } if (sscanf(page, "%s\n", port->disc_addr.trsvcid) != 1) return -EINVAL; @@ -233,11 +236,8 @@ static ssize_t nvmet_param_inline_data_size_store(struct config_item *item, struct nvmet_port *port = to_nvmet_port(item); int ret; - if (port->enabled) { - pr_err("Cannot modify inline_data_size while port enabled\n"); - pr_err("Disable the port before modifying\n"); + if (nvmet_is_port_enabled(port, __func__)) return -EACCES; - } ret = kstrtoint(page, 0, &port->inline_data_size); if (ret) { pr_err("Invalid value '%s' for inline_data_size\n", page); @@ -248,16 +248,45 @@ static ssize_t nvmet_param_inline_data_size_store(struct config_item *item, CONFIGFS_ATTR(nvmet_, param_inline_data_size); +#ifdef CONFIG_BLK_DEV_INTEGRITY +static ssize_t nvmet_param_pi_enable_show(struct config_item *item, + char *page) +{ + struct nvmet_port *port = to_nvmet_port(item); + + return snprintf(page, PAGE_SIZE, "%d\n", port->pi_enable); +} + +static ssize_t nvmet_param_pi_enable_store(struct config_item *item, + const char *page, size_t count) +{ + struct nvmet_port *port = to_nvmet_port(item); + bool val; + + if (strtobool(page, &val)) + return -EINVAL; + + if (port->enabled) { + pr_err("Disable port before setting pi_enable value.\n"); + return -EACCES; + } + + port->pi_enable = val; + return count; +} + +CONFIGFS_ATTR(nvmet_, param_pi_enable); +#endif + static ssize_t nvmet_addr_trtype_show(struct config_item *item, char *page) { struct nvmet_port *port = to_nvmet_port(item); int i; - for (i = 0; i < ARRAY_SIZE(nvmet_transport_names); i++) { - if (port->disc_addr.trtype != nvmet_transport_names[i].type) - continue; - return sprintf(page, "%s\n", nvmet_transport_names[i].name); + for (i = 0; i < ARRAY_SIZE(nvmet_transport); i++) { + if (port->disc_addr.trtype == nvmet_transport[i].type) + return sprintf(page, "%s\n", nvmet_transport[i].name); } return sprintf(page, "\n"); @@ -276,22 +305,20 @@ static ssize_t nvmet_addr_trtype_store(struct config_item *item, struct nvmet_port *port = to_nvmet_port(item); int i; - if (port->enabled) { - pr_err("Cannot modify address while enabled\n"); - pr_err("Disable the address before modifying\n"); + if (nvmet_is_port_enabled(port, __func__)) return -EACCES; - } - for (i = 0; i < ARRAY_SIZE(nvmet_transport_names); i++) { - if (sysfs_streq(page, nvmet_transport_names[i].name)) + for (i = 0; i < ARRAY_SIZE(nvmet_transport); i++) { + if (sysfs_streq(page, nvmet_transport[i].name)) goto found; } pr_err("Invalid value '%s' for trtype\n", page); return -EINVAL; + found: memset(&port->disc_addr.tsas, 0, NVMF_TSAS_SIZE); - port->disc_addr.trtype = nvmet_transport_names[i].type; + port->disc_addr.trtype = nvmet_transport[i].type; if (port->disc_addr.trtype == NVMF_TRTYPE_RDMA) nvmet_port_init_tsas_rdma(port); return count; @@ -327,7 +354,7 @@ static ssize_t nvmet_ns_device_path_store(struct config_item *item, kfree(ns->device_path); ret = -ENOMEM; - ns->device_path = kstrndup(page, len, GFP_KERNEL); + ns->device_path = kmemdup_nul(page, len, GFP_KERNEL); if (!ns->device_path) goto out_unlock; @@ -543,6 +570,31 @@ static ssize_t nvmet_ns_buffered_io_store(struct config_item *item, CONFIGFS_ATTR(nvmet_ns_, buffered_io); +static ssize_t nvmet_ns_revalidate_size_store(struct config_item *item, + const char *page, size_t count) +{ + struct nvmet_ns *ns = to_nvmet_ns(item); + bool val; + + if (strtobool(page, &val)) + return -EINVAL; + + if (!val) + return -EINVAL; + + mutex_lock(&ns->subsys->lock); + if (!ns->enabled) { + pr_err("enable ns before revalidate.\n"); + mutex_unlock(&ns->subsys->lock); + return -EINVAL; + } + nvmet_ns_revalidate(ns); + mutex_unlock(&ns->subsys->lock); + return count; +} + +CONFIGFS_ATTR_WO(nvmet_ns_, revalidate_size); + static struct configfs_attribute *nvmet_ns_attrs[] = { &nvmet_ns_attr_device_path, &nvmet_ns_attr_device_nguid, @@ -550,6 +602,7 @@ static struct configfs_attribute *nvmet_ns_attrs[] = { &nvmet_ns_attr_ana_grpid, &nvmet_ns_attr_enable, &nvmet_ns_attr_buffered_io, + &nvmet_ns_attr_revalidate_size, #ifdef CONFIG_PCI_P2PDMA &nvmet_ns_attr_p2pmem, #endif @@ -963,7 +1016,7 @@ static ssize_t nvmet_subsys_attr_model_store(struct config_item *item, return -EINVAL; } - new_model_number = kstrndup(page, len, GFP_KERNEL); + new_model_number = kmemdup_nul(page, len, GFP_KERNEL); if (!new_model_number) return -ENOMEM; @@ -987,6 +1040,28 @@ static ssize_t nvmet_subsys_attr_model_store(struct config_item *item, } CONFIGFS_ATTR(nvmet_subsys_, attr_model); +#ifdef CONFIG_BLK_DEV_INTEGRITY +static ssize_t nvmet_subsys_attr_pi_enable_show(struct config_item *item, + char *page) +{ + return snprintf(page, PAGE_SIZE, "%d\n", to_subsys(item)->pi_support); +} + +static ssize_t nvmet_subsys_attr_pi_enable_store(struct config_item *item, + const char *page, size_t count) +{ + struct nvmet_subsys *subsys = to_subsys(item); + bool pi_enable; + + if (strtobool(page, &pi_enable)) + return -EINVAL; + + subsys->pi_support = pi_enable; + return count; +} +CONFIGFS_ATTR(nvmet_subsys_, attr_pi_enable); +#endif + static struct configfs_attribute *nvmet_subsys_attrs[] = { &nvmet_subsys_attr_attr_allow_any_host, &nvmet_subsys_attr_attr_version, @@ -994,6 +1069,9 @@ static struct configfs_attribute *nvmet_subsys_attrs[] = { &nvmet_subsys_attr_attr_cntlid_min, &nvmet_subsys_attr_attr_cntlid_max, &nvmet_subsys_attr_attr_model, +#ifdef CONFIG_BLK_DEV_INTEGRITY + &nvmet_subsys_attr_attr_pi_enable, +#endif NULL, }; @@ -1149,10 +1227,7 @@ static const struct config_item_type nvmet_referrals_type = { .ct_group_ops = &nvmet_referral_group_ops, }; -static struct { - enum nvme_ana_state state; - const char *name; -} nvmet_ana_state_names[] = { +static struct nvmet_type_name_map nvmet_ana_state[] = { { NVME_ANA_OPTIMIZED, "optimized" }, { NVME_ANA_NONOPTIMIZED, "non-optimized" }, { NVME_ANA_INACCESSIBLE, "inaccessible" }, @@ -1167,10 +1242,9 @@ static ssize_t nvmet_ana_group_ana_state_show(struct config_item *item, enum nvme_ana_state state = grp->port->ana_state[grp->grpid]; int i; - for (i = 0; i < ARRAY_SIZE(nvmet_ana_state_names); i++) { - if (state != nvmet_ana_state_names[i].state) - continue; - return sprintf(page, "%s\n", nvmet_ana_state_names[i].name); + for (i = 0; i < ARRAY_SIZE(nvmet_ana_state); i++) { + if (state == nvmet_ana_state[i].type) + return sprintf(page, "%s\n", nvmet_ana_state[i].name); } return sprintf(page, "\n"); @@ -1180,10 +1254,11 @@ static ssize_t nvmet_ana_group_ana_state_store(struct config_item *item, const char *page, size_t count) { struct nvmet_ana_group *grp = to_ana_group(item); + enum nvme_ana_state *ana_state = grp->port->ana_state; int i; - for (i = 0; i < ARRAY_SIZE(nvmet_ana_state_names); i++) { - if (sysfs_streq(page, nvmet_ana_state_names[i].name)) + for (i = 0; i < ARRAY_SIZE(nvmet_ana_state); i++) { + if (sysfs_streq(page, nvmet_ana_state[i].name)) goto found; } @@ -1192,10 +1267,9 @@ static ssize_t nvmet_ana_group_ana_state_store(struct config_item *item, found: down_write(&nvmet_ana_sem); - grp->port->ana_state[grp->grpid] = nvmet_ana_state_names[i].state; + ana_state[grp->grpid] = (enum nvme_ana_state) nvmet_ana_state[i].type; nvmet_ana_chgcnt++; up_write(&nvmet_ana_sem); - nvmet_port_send_ana_event(grp->port); return count; } @@ -1297,6 +1371,9 @@ static struct configfs_attribute *nvmet_port_attrs[] = { &nvmet_attr_addr_trsvcid, &nvmet_attr_addr_trtype, &nvmet_attr_param_inline_data_size, +#ifdef CONFIG_BLK_DEV_INTEGRITY + &nvmet_attr_param_pi_enable, +#endif NULL, }; @@ -1346,6 +1423,7 @@ static struct config_group *nvmet_ports_make(struct config_group *group, port->inline_data_size = -1; /* < 0 == let the transport choose */ port->disc_addr.portid = cpu_to_le16(portid); + port->disc_addr.adrfam = NVMF_ADDR_FAMILY_MAX; port->disc_addr.treq = NVMF_TREQ_DISABLE_SQFLOW; config_group_init_type_name(&port->group, name, &nvmet_port_type); diff --git a/drivers/nvme/target/core.c b/drivers/nvme/target/core.c index b685f99d56a1..6392bcd30bd7 100644 --- a/drivers/nvme/target/core.c +++ b/drivers/nvme/target/core.c @@ -134,15 +134,10 @@ static void nvmet_async_events_process(struct nvmet_ctrl *ctrl, u16 status) struct nvmet_async_event *aen; struct nvmet_req *req; - while (1) { - mutex_lock(&ctrl->lock); - aen = list_first_entry_or_null(&ctrl->async_events, - struct nvmet_async_event, entry); - if (!aen || !ctrl->nr_async_event_cmds) { - mutex_unlock(&ctrl->lock); - break; - } - + mutex_lock(&ctrl->lock); + while (ctrl->nr_async_event_cmds && !list_empty(&ctrl->async_events)) { + aen = list_first_entry(&ctrl->async_events, + struct nvmet_async_event, entry); req = ctrl->async_event_cmds[--ctrl->nr_async_event_cmds]; if (status == 0) nvmet_set_result(req, nvmet_async_event_result(aen)); @@ -151,20 +146,21 @@ static void nvmet_async_events_process(struct nvmet_ctrl *ctrl, u16 status) kfree(aen); mutex_unlock(&ctrl->lock); + trace_nvmet_async_event(ctrl, req->cqe->result.u32); nvmet_req_complete(req, status); + mutex_lock(&ctrl->lock); } + mutex_unlock(&ctrl->lock); } static void nvmet_async_events_free(struct nvmet_ctrl *ctrl) { - struct nvmet_req *req; + struct nvmet_async_event *aen, *tmp; mutex_lock(&ctrl->lock); - while (ctrl->nr_async_event_cmds) { - req = ctrl->async_event_cmds[--ctrl->nr_async_event_cmds]; - mutex_unlock(&ctrl->lock); - nvmet_req_complete(req, NVME_SC_INTERNAL | NVME_SC_DNR); - mutex_lock(&ctrl->lock); + list_for_each_entry_safe(aen, tmp, &ctrl->async_events, entry) { + list_del(&aen->entry); + kfree(aen); } mutex_unlock(&ctrl->lock); } @@ -322,12 +318,21 @@ int nvmet_enable_port(struct nvmet_port *port) if (!try_module_get(ops->owner)) return -EINVAL; - ret = ops->add_port(port); - if (ret) { - module_put(ops->owner); - return ret; + /* + * If the user requested PI support and the transport isn't pi capable, + * don't enable the port. + */ + if (port->pi_enable && !ops->metadata_support) { + pr_err("T10-PI is not supported by transport type %d\n", + port->disc_addr.trtype); + ret = -EINVAL; + goto out_put; } + ret = ops->add_port(port); + if (ret) + goto out_put; + /* If the transport didn't set inline_data_size, then disable it. */ if (port->inline_data_size < 0) port->inline_data_size = 0; @@ -335,6 +340,10 @@ int nvmet_enable_port(struct nvmet_port *port) port->enabled = true; port->tr_ops = ops; return 0; + +out_put: + module_put(ops->owner); + return ret; } void nvmet_disable_port(struct nvmet_port *port) @@ -514,6 +523,19 @@ static void nvmet_p2pmem_ns_add_p2p(struct nvmet_ctrl *ctrl, ns->nsid); } +void nvmet_ns_revalidate(struct nvmet_ns *ns) +{ + loff_t oldsize = ns->size; + + if (ns->bdev) + nvmet_bdev_ns_revalidate(ns); + else + nvmet_file_ns_revalidate(ns); + + if (oldsize != ns->size) + nvmet_ns_changed(ns->subsys, ns->nsid); +} + int nvmet_ns_enable(struct nvmet_ns *ns) { struct nvmet_subsys *subsys = ns->subsys; @@ -764,10 +786,8 @@ void nvmet_sq_destroy(struct nvmet_sq *sq) * If this is the admin queue, complete all AERs so that our * queue doesn't have outstanding requests on it. */ - if (ctrl && ctrl->sqs && ctrl->sqs[0] == sq) { + if (ctrl && ctrl->sqs && ctrl->sqs[0] == sq) nvmet_async_events_process(ctrl, status); - nvmet_async_events_free(ctrl); - } percpu_ref_kill_and_confirm(&sq->ref, nvmet_confirm_sq); wait_for_completion(&sq->confirm_done); wait_for_completion(&sq->free_done); @@ -873,8 +893,11 @@ bool nvmet_req_init(struct nvmet_req *req, struct nvmet_cq *cq, req->sq = sq; req->ops = ops; req->sg = NULL; + req->metadata_sg = NULL; req->sg_cnt = 0; + req->metadata_sg_cnt = 0; req->transfer_len = 0; + req->metadata_len = 0; req->cqe->status = 0; req->cqe->sq_head = 0; req->ns = NULL; @@ -936,9 +959,9 @@ void nvmet_req_uninit(struct nvmet_req *req) } EXPORT_SYMBOL_GPL(nvmet_req_uninit); -bool nvmet_check_data_len(struct nvmet_req *req, size_t data_len) +bool nvmet_check_transfer_len(struct nvmet_req *req, size_t len) { - if (unlikely(data_len != req->transfer_len)) { + if (unlikely(len != req->transfer_len)) { req->error_loc = offsetof(struct nvme_common_command, dptr); nvmet_req_complete(req, NVME_SC_SGL_INVALID_DATA | NVME_SC_DNR); return false; @@ -946,7 +969,7 @@ bool nvmet_check_data_len(struct nvmet_req *req, size_t data_len) return true; } -EXPORT_SYMBOL_GPL(nvmet_check_data_len); +EXPORT_SYMBOL_GPL(nvmet_check_transfer_len); bool nvmet_check_data_len_lte(struct nvmet_req *req, size_t data_len) { @@ -959,50 +982,90 @@ bool nvmet_check_data_len_lte(struct nvmet_req *req, size_t data_len) return true; } -int nvmet_req_alloc_sgl(struct nvmet_req *req) +static unsigned int nvmet_data_transfer_len(struct nvmet_req *req) { - struct pci_dev *p2p_dev = NULL; + return req->transfer_len - req->metadata_len; +} - if (IS_ENABLED(CONFIG_PCI_P2PDMA)) { - if (req->sq->ctrl && req->ns) - p2p_dev = radix_tree_lookup(&req->sq->ctrl->p2p_ns_map, - req->ns->nsid); +static int nvmet_req_alloc_p2pmem_sgls(struct nvmet_req *req) +{ + req->sg = pci_p2pmem_alloc_sgl(req->p2p_dev, &req->sg_cnt, + nvmet_data_transfer_len(req)); + if (!req->sg) + goto out_err; - req->p2p_dev = NULL; - if (req->sq->qid && p2p_dev) { - req->sg = pci_p2pmem_alloc_sgl(p2p_dev, &req->sg_cnt, - req->transfer_len); - if (req->sg) { - req->p2p_dev = p2p_dev; - return 0; - } - } + if (req->metadata_len) { + req->metadata_sg = pci_p2pmem_alloc_sgl(req->p2p_dev, + &req->metadata_sg_cnt, req->metadata_len); + if (!req->metadata_sg) + goto out_free_sg; + } + return 0; +out_free_sg: + pci_p2pmem_free_sgl(req->p2p_dev, req->sg); +out_err: + return -ENOMEM; +} - /* - * If no P2P memory was available we fallback to using - * regular memory - */ +static bool nvmet_req_find_p2p_dev(struct nvmet_req *req) +{ + if (!IS_ENABLED(CONFIG_PCI_P2PDMA)) + return false; + + if (req->sq->ctrl && req->sq->qid && req->ns) { + req->p2p_dev = radix_tree_lookup(&req->sq->ctrl->p2p_ns_map, + req->ns->nsid); + if (req->p2p_dev) + return true; } - req->sg = sgl_alloc(req->transfer_len, GFP_KERNEL, &req->sg_cnt); + req->p2p_dev = NULL; + return false; +} + +int nvmet_req_alloc_sgls(struct nvmet_req *req) +{ + if (nvmet_req_find_p2p_dev(req) && !nvmet_req_alloc_p2pmem_sgls(req)) + return 0; + + req->sg = sgl_alloc(nvmet_data_transfer_len(req), GFP_KERNEL, + &req->sg_cnt); if (unlikely(!req->sg)) - return -ENOMEM; + goto out; + + if (req->metadata_len) { + req->metadata_sg = sgl_alloc(req->metadata_len, GFP_KERNEL, + &req->metadata_sg_cnt); + if (unlikely(!req->metadata_sg)) + goto out_free; + } return 0; +out_free: + sgl_free(req->sg); +out: + return -ENOMEM; } -EXPORT_SYMBOL_GPL(nvmet_req_alloc_sgl); +EXPORT_SYMBOL_GPL(nvmet_req_alloc_sgls); -void nvmet_req_free_sgl(struct nvmet_req *req) +void nvmet_req_free_sgls(struct nvmet_req *req) { - if (req->p2p_dev) + if (req->p2p_dev) { pci_p2pmem_free_sgl(req->p2p_dev, req->sg); - else + if (req->metadata_sg) + pci_p2pmem_free_sgl(req->p2p_dev, req->metadata_sg); + } else { sgl_free(req->sg); + if (req->metadata_sg) + sgl_free(req->metadata_sg); + } req->sg = NULL; + req->metadata_sg = NULL; req->sg_cnt = 0; + req->metadata_sg_cnt = 0; } -EXPORT_SYMBOL_GPL(nvmet_req_free_sgl); +EXPORT_SYMBOL_GPL(nvmet_req_free_sgls); static inline bool nvmet_cc_en(u32 cc) { @@ -1357,6 +1420,7 @@ static void nvmet_ctrl_free(struct kref *ref) ida_simple_remove(&cntlid_ida, ctrl->cntlid); + nvmet_async_events_free(ctrl); kfree(ctrl->sqs); kfree(ctrl->cqs); kfree(ctrl->changed_ns_list); diff --git a/drivers/nvme/target/discovery.c b/drivers/nvme/target/discovery.c index 0c2274b21e15..40cf0b6e6c9d 100644 --- a/drivers/nvme/target/discovery.c +++ b/drivers/nvme/target/discovery.c @@ -171,7 +171,7 @@ static void nvmet_execute_disc_get_log_page(struct nvmet_req *req) u16 status = 0; void *buffer; - if (!nvmet_check_data_len(req, data_len)) + if (!nvmet_check_transfer_len(req, data_len)) return; if (req->cmd->get_log_page.lid != NVME_LOG_DISC) { @@ -244,7 +244,7 @@ static void nvmet_execute_disc_identify(struct nvmet_req *req) const char model[] = "Linux"; u16 status = 0; - if (!nvmet_check_data_len(req, NVME_IDENTIFY_DATA_SIZE)) + if (!nvmet_check_transfer_len(req, NVME_IDENTIFY_DATA_SIZE)) return; if (req->cmd->identify.cns != NVME_ID_CNS_CTRL) { @@ -298,7 +298,7 @@ static void nvmet_execute_disc_set_features(struct nvmet_req *req) u32 cdw10 = le32_to_cpu(req->cmd->common.cdw10); u16 stat; - if (!nvmet_check_data_len(req, 0)) + if (!nvmet_check_transfer_len(req, 0)) return; switch (cdw10 & 0xff) { @@ -324,7 +324,7 @@ static void nvmet_execute_disc_get_features(struct nvmet_req *req) u32 cdw10 = le32_to_cpu(req->cmd->common.cdw10); u16 stat = 0; - if (!nvmet_check_data_len(req, 0)) + if (!nvmet_check_transfer_len(req, 0)) return; switch (cdw10 & 0xff) { diff --git a/drivers/nvme/target/fabrics-cmd.c b/drivers/nvme/target/fabrics-cmd.c index feef15c38ec9..42bd12b8bf00 100644 --- a/drivers/nvme/target/fabrics-cmd.c +++ b/drivers/nvme/target/fabrics-cmd.c @@ -12,7 +12,7 @@ static void nvmet_execute_prop_set(struct nvmet_req *req) u64 val = le64_to_cpu(req->cmd->prop_set.value); u16 status = 0; - if (!nvmet_check_data_len(req, 0)) + if (!nvmet_check_transfer_len(req, 0)) return; if (req->cmd->prop_set.attrib & 1) { @@ -41,7 +41,7 @@ static void nvmet_execute_prop_get(struct nvmet_req *req) u16 status = 0; u64 val = 0; - if (!nvmet_check_data_len(req, 0)) + if (!nvmet_check_transfer_len(req, 0)) return; if (req->cmd->prop_get.attrib & 1) { @@ -156,7 +156,7 @@ static void nvmet_execute_admin_connect(struct nvmet_req *req) struct nvmet_ctrl *ctrl = NULL; u16 status = 0; - if (!nvmet_check_data_len(req, sizeof(struct nvmf_connect_data))) + if (!nvmet_check_transfer_len(req, sizeof(struct nvmf_connect_data))) return; d = kmalloc(sizeof(*d), GFP_KERNEL); @@ -197,6 +197,8 @@ static void nvmet_execute_admin_connect(struct nvmet_req *req) goto out; } + ctrl->pi_support = ctrl->port->pi_enable && ctrl->subsys->pi_support; + uuid_copy(&ctrl->hostid, &d->hostid); status = nvmet_install_queue(ctrl, req); @@ -205,8 +207,9 @@ static void nvmet_execute_admin_connect(struct nvmet_req *req) goto out; } - pr_info("creating controller %d for subsystem %s for NQN %s.\n", - ctrl->cntlid, ctrl->subsys->subsysnqn, ctrl->hostnqn); + pr_info("creating controller %d for subsystem %s for NQN %s%s.\n", + ctrl->cntlid, ctrl->subsys->subsysnqn, ctrl->hostnqn, + ctrl->pi_support ? " T10-PI is enabled" : ""); req->cqe->result.u16 = cpu_to_le16(ctrl->cntlid); out: @@ -223,7 +226,7 @@ static void nvmet_execute_io_connect(struct nvmet_req *req) u16 qid = le16_to_cpu(c->qid); u16 status = 0; - if (!nvmet_check_data_len(req, sizeof(struct nvmf_connect_data))) + if (!nvmet_check_transfer_len(req, sizeof(struct nvmf_connect_data))) return; d = kmalloc(sizeof(*d), GFP_KERNEL); diff --git a/drivers/nvme/target/fc.c b/drivers/nvme/target/fc.c index a8ceb7721640..27fd3b5aa621 100644 --- a/drivers/nvme/target/fc.c +++ b/drivers/nvme/target/fc.c @@ -14,6 +14,7 @@ #include "nvmet.h" #include <linux/nvme-fc-driver.h> #include <linux/nvme-fc.h> +#include "../host/fc.h" /* *************************** Data Structures/Defines ****************** */ @@ -21,23 +22,21 @@ #define NVMET_LS_CTX_COUNT 256 -/* for this implementation, assume small single frame rqst/rsp */ -#define NVME_FC_MAX_LS_BUFFER_SIZE 2048 - struct nvmet_fc_tgtport; struct nvmet_fc_tgt_assoc; -struct nvmet_fc_ls_iod { - struct nvmefc_tgt_ls_req *lsreq; +struct nvmet_fc_ls_iod { /* for an LS RQST RCV */ + struct nvmefc_ls_rsp *lsrsp; struct nvmefc_tgt_fcp_req *fcpreq; /* only if RS */ - struct list_head ls_list; /* tgtport->ls_list */ + struct list_head ls_rcv_list; /* tgtport->ls_rcv_list */ struct nvmet_fc_tgtport *tgtport; struct nvmet_fc_tgt_assoc *assoc; + void *hosthandle; - u8 *rqstbuf; - u8 *rspbuf; + union nvmefc_ls_requests *rqstbuf; + union nvmefc_ls_responses *rspbuf; u16 rqstdatalen; dma_addr_t rspdma; @@ -46,6 +45,18 @@ struct nvmet_fc_ls_iod { struct work_struct work; } __aligned(sizeof(unsigned long long)); +struct nvmet_fc_ls_req_op { /* for an LS RQST XMT */ + struct nvmefc_ls_req ls_req; + + struct nvmet_fc_tgtport *tgtport; + void *hosthandle; + + int ls_error; + struct list_head lsreq_list; /* tgtport->ls_req_list */ + bool req_queued; +}; + + /* desired maximum for a single sequence - if sg list allows it */ #define NVMET_FC_MAX_SEQ_LENGTH (256 * 1024) @@ -83,7 +94,6 @@ struct nvmet_fc_fcp_iod { }; struct nvmet_fc_tgtport { - struct nvmet_fc_target_port fc_target_port; struct list_head tgt_list; /* nvmet_fc_target_list */ @@ -92,9 +102,11 @@ struct nvmet_fc_tgtport { struct nvmet_fc_ls_iod *iod; spinlock_t lock; - struct list_head ls_list; + struct list_head ls_rcv_list; + struct list_head ls_req_list; struct list_head ls_busylist; struct list_head assoc_list; + struct list_head host_list; struct ida assoc_cnt; struct nvmet_fc_port_entry *pe; struct kref ref; @@ -136,14 +148,26 @@ struct nvmet_fc_tgt_queue { struct nvmet_fc_fcp_iod fod[]; /* array of fcp_iods */ } __aligned(sizeof(unsigned long long)); +struct nvmet_fc_hostport { + struct nvmet_fc_tgtport *tgtport; + void *hosthandle; + struct list_head host_list; + struct kref ref; + u8 invalid; +}; + struct nvmet_fc_tgt_assoc { u64 association_id; u32 a_id; + atomic_t terminating; struct nvmet_fc_tgtport *tgtport; + struct nvmet_fc_hostport *hostport; + struct nvmet_fc_ls_iod *rcv_disconn; struct list_head a_list; struct nvmet_fc_tgt_queue *queues[NVMET_NR_QUEUES + 1]; struct kref ref; struct work_struct del_work; + atomic_t del_work_active; }; @@ -227,6 +251,8 @@ static int nvmet_fc_tgtport_get(struct nvmet_fc_tgtport *tgtport); static void nvmet_fc_handle_fcp_rqst(struct nvmet_fc_tgtport *tgtport, struct nvmet_fc_fcp_iod *fod); static void nvmet_fc_delete_target_assoc(struct nvmet_fc_tgt_assoc *assoc); +static void nvmet_fc_xmt_ls_rsp(struct nvmet_fc_tgtport *tgtport, + struct nvmet_fc_ls_iod *iod); /* *********************** FC-NVME DMA Handling **************************** */ @@ -318,6 +344,188 @@ fc_dma_unmap_sg(struct device *dev, struct scatterlist *sg, int nents, } +/* ********************** FC-NVME LS XMT Handling ************************* */ + + +static void +__nvmet_fc_finish_ls_req(struct nvmet_fc_ls_req_op *lsop) +{ + struct nvmet_fc_tgtport *tgtport = lsop->tgtport; + struct nvmefc_ls_req *lsreq = &lsop->ls_req; + unsigned long flags; + + spin_lock_irqsave(&tgtport->lock, flags); + + if (!lsop->req_queued) { + spin_unlock_irqrestore(&tgtport->lock, flags); + return; + } + + list_del(&lsop->lsreq_list); + + lsop->req_queued = false; + + spin_unlock_irqrestore(&tgtport->lock, flags); + + fc_dma_unmap_single(tgtport->dev, lsreq->rqstdma, + (lsreq->rqstlen + lsreq->rsplen), + DMA_BIDIRECTIONAL); + + nvmet_fc_tgtport_put(tgtport); +} + +static int +__nvmet_fc_send_ls_req(struct nvmet_fc_tgtport *tgtport, + struct nvmet_fc_ls_req_op *lsop, + void (*done)(struct nvmefc_ls_req *req, int status)) +{ + struct nvmefc_ls_req *lsreq = &lsop->ls_req; + unsigned long flags; + int ret = 0; + + if (!tgtport->ops->ls_req) + return -EOPNOTSUPP; + + if (!nvmet_fc_tgtport_get(tgtport)) + return -ESHUTDOWN; + + lsreq->done = done; + lsop->req_queued = false; + INIT_LIST_HEAD(&lsop->lsreq_list); + + lsreq->rqstdma = fc_dma_map_single(tgtport->dev, lsreq->rqstaddr, + lsreq->rqstlen + lsreq->rsplen, + DMA_BIDIRECTIONAL); + if (fc_dma_mapping_error(tgtport->dev, lsreq->rqstdma)) { + ret = -EFAULT; + goto out_puttgtport; + } + lsreq->rspdma = lsreq->rqstdma + lsreq->rqstlen; + + spin_lock_irqsave(&tgtport->lock, flags); + + list_add_tail(&lsop->lsreq_list, &tgtport->ls_req_list); + + lsop->req_queued = true; + + spin_unlock_irqrestore(&tgtport->lock, flags); + + ret = tgtport->ops->ls_req(&tgtport->fc_target_port, lsop->hosthandle, + lsreq); + if (ret) + goto out_unlink; + + return 0; + +out_unlink: + lsop->ls_error = ret; + spin_lock_irqsave(&tgtport->lock, flags); + lsop->req_queued = false; + list_del(&lsop->lsreq_list); + spin_unlock_irqrestore(&tgtport->lock, flags); + fc_dma_unmap_single(tgtport->dev, lsreq->rqstdma, + (lsreq->rqstlen + lsreq->rsplen), + DMA_BIDIRECTIONAL); +out_puttgtport: + nvmet_fc_tgtport_put(tgtport); + + return ret; +} + +static int +nvmet_fc_send_ls_req_async(struct nvmet_fc_tgtport *tgtport, + struct nvmet_fc_ls_req_op *lsop, + void (*done)(struct nvmefc_ls_req *req, int status)) +{ + /* don't wait for completion */ + + return __nvmet_fc_send_ls_req(tgtport, lsop, done); +} + +static void +nvmet_fc_disconnect_assoc_done(struct nvmefc_ls_req *lsreq, int status) +{ + struct nvmet_fc_ls_req_op *lsop = + container_of(lsreq, struct nvmet_fc_ls_req_op, ls_req); + + __nvmet_fc_finish_ls_req(lsop); + + /* fc-nvme target doesn't care about success or failure of cmd */ + + kfree(lsop); +} + +/* + * This routine sends a FC-NVME LS to disconnect (aka terminate) + * the FC-NVME Association. Terminating the association also + * terminates the FC-NVME connections (per queue, both admin and io + * queues) that are part of the association. E.g. things are torn + * down, and the related FC-NVME Association ID and Connection IDs + * become invalid. + * + * The behavior of the fc-nvme target is such that it's + * understanding of the association and connections will implicitly + * be torn down. The action is implicit as it may be due to a loss of + * connectivity with the fc-nvme host, so the target may never get a + * response even if it tried. As such, the action of this routine + * is to asynchronously send the LS, ignore any results of the LS, and + * continue on with terminating the association. If the fc-nvme host + * is present and receives the LS, it too can tear down. + */ +static void +nvmet_fc_xmt_disconnect_assoc(struct nvmet_fc_tgt_assoc *assoc) +{ + struct nvmet_fc_tgtport *tgtport = assoc->tgtport; + struct fcnvme_ls_disconnect_assoc_rqst *discon_rqst; + struct fcnvme_ls_disconnect_assoc_acc *discon_acc; + struct nvmet_fc_ls_req_op *lsop; + struct nvmefc_ls_req *lsreq; + int ret; + + /* + * If ls_req is NULL or no hosthandle, it's an older lldd and no + * message is normal. Otherwise, send unless the hostport has + * already been invalidated by the lldd. + */ + if (!tgtport->ops->ls_req || !assoc->hostport || + assoc->hostport->invalid) + return; + + lsop = kzalloc((sizeof(*lsop) + + sizeof(*discon_rqst) + sizeof(*discon_acc) + + tgtport->ops->lsrqst_priv_sz), GFP_KERNEL); + if (!lsop) { + dev_info(tgtport->dev, + "{%d:%d} send Disconnect Association failed: ENOMEM\n", + tgtport->fc_target_port.port_num, assoc->a_id); + return; + } + + discon_rqst = (struct fcnvme_ls_disconnect_assoc_rqst *)&lsop[1]; + discon_acc = (struct fcnvme_ls_disconnect_assoc_acc *)&discon_rqst[1]; + lsreq = &lsop->ls_req; + if (tgtport->ops->lsrqst_priv_sz) + lsreq->private = (void *)&discon_acc[1]; + else + lsreq->private = NULL; + + lsop->tgtport = tgtport; + lsop->hosthandle = assoc->hostport->hosthandle; + + nvmefc_fmt_lsreq_discon_assoc(lsreq, discon_rqst, discon_acc, + assoc->association_id); + + ret = nvmet_fc_send_ls_req_async(tgtport, lsop, + nvmet_fc_disconnect_assoc_done); + if (ret) { + dev_info(tgtport->dev, + "{%d:%d} XMT Disconnect Association failed: %d\n", + tgtport->fc_target_port.port_num, assoc->a_id, ret); + kfree(lsop); + } +} + + /* *********************** FC-NVME Port Management ************************ */ @@ -337,17 +545,18 @@ nvmet_fc_alloc_ls_iodlist(struct nvmet_fc_tgtport *tgtport) for (i = 0; i < NVMET_LS_CTX_COUNT; iod++, i++) { INIT_WORK(&iod->work, nvmet_fc_handle_ls_rqst_work); iod->tgtport = tgtport; - list_add_tail(&iod->ls_list, &tgtport->ls_list); + list_add_tail(&iod->ls_rcv_list, &tgtport->ls_rcv_list); - iod->rqstbuf = kcalloc(2, NVME_FC_MAX_LS_BUFFER_SIZE, - GFP_KERNEL); + iod->rqstbuf = kzalloc(sizeof(union nvmefc_ls_requests) + + sizeof(union nvmefc_ls_responses), + GFP_KERNEL); if (!iod->rqstbuf) goto out_fail; - iod->rspbuf = iod->rqstbuf + NVME_FC_MAX_LS_BUFFER_SIZE; + iod->rspbuf = (union nvmefc_ls_responses *)&iod->rqstbuf[1]; iod->rspdma = fc_dma_map_single(tgtport->dev, iod->rspbuf, - NVME_FC_MAX_LS_BUFFER_SIZE, + sizeof(*iod->rspbuf), DMA_TO_DEVICE); if (fc_dma_mapping_error(tgtport->dev, iod->rspdma)) goto out_fail; @@ -357,12 +566,12 @@ nvmet_fc_alloc_ls_iodlist(struct nvmet_fc_tgtport *tgtport) out_fail: kfree(iod->rqstbuf); - list_del(&iod->ls_list); + list_del(&iod->ls_rcv_list); for (iod--, i--; i >= 0; iod--, i--) { fc_dma_unmap_single(tgtport->dev, iod->rspdma, - NVME_FC_MAX_LS_BUFFER_SIZE, DMA_TO_DEVICE); + sizeof(*iod->rspbuf), DMA_TO_DEVICE); kfree(iod->rqstbuf); - list_del(&iod->ls_list); + list_del(&iod->ls_rcv_list); } kfree(iod); @@ -378,10 +587,10 @@ nvmet_fc_free_ls_iodlist(struct nvmet_fc_tgtport *tgtport) for (i = 0; i < NVMET_LS_CTX_COUNT; iod++, i++) { fc_dma_unmap_single(tgtport->dev, - iod->rspdma, NVME_FC_MAX_LS_BUFFER_SIZE, + iod->rspdma, sizeof(*iod->rspbuf), DMA_TO_DEVICE); kfree(iod->rqstbuf); - list_del(&iod->ls_list); + list_del(&iod->ls_rcv_list); } kfree(tgtport->iod); } @@ -393,10 +602,10 @@ nvmet_fc_alloc_ls_iod(struct nvmet_fc_tgtport *tgtport) unsigned long flags; spin_lock_irqsave(&tgtport->lock, flags); - iod = list_first_entry_or_null(&tgtport->ls_list, - struct nvmet_fc_ls_iod, ls_list); + iod = list_first_entry_or_null(&tgtport->ls_rcv_list, + struct nvmet_fc_ls_iod, ls_rcv_list); if (iod) - list_move_tail(&iod->ls_list, &tgtport->ls_busylist); + list_move_tail(&iod->ls_rcv_list, &tgtport->ls_busylist); spin_unlock_irqrestore(&tgtport->lock, flags); return iod; } @@ -409,7 +618,7 @@ nvmet_fc_free_ls_iod(struct nvmet_fc_tgtport *tgtport, unsigned long flags; spin_lock_irqsave(&tgtport->lock, flags); - list_move(&iod->ls_list, &tgtport->ls_list); + list_move(&iod->ls_rcv_list, &tgtport->ls_rcv_list); spin_unlock_irqrestore(&tgtport->lock, flags); } @@ -678,31 +887,33 @@ nvmet_fc_delete_target_queue(struct nvmet_fc_tgt_queue *queue) struct nvmet_fc_fcp_iod *fod = queue->fod; struct nvmet_fc_defer_fcp_req *deferfcp, *tempptr; unsigned long flags; - int i, writedataactive; + int i; bool disconnect; disconnect = atomic_xchg(&queue->connected, 0); + /* if not connected, nothing to do */ + if (!disconnect) + return; + spin_lock_irqsave(&queue->qlock, flags); /* abort outstanding io's */ for (i = 0; i < queue->sqsize; fod++, i++) { if (fod->active) { spin_lock(&fod->flock); fod->abort = true; - writedataactive = fod->writedataactive; - spin_unlock(&fod->flock); /* * only call lldd abort routine if waiting for * writedata. other outstanding ops should finish * on their own. */ - if (writedataactive) { - spin_lock(&fod->flock); + if (fod->writedataactive) { fod->aborted = true; spin_unlock(&fod->flock); tgtport->ops->fcp_abort( &tgtport->fc_target_port, fod->fcpreq); - } + } else + spin_unlock(&fod->flock); } } @@ -742,8 +953,7 @@ nvmet_fc_delete_target_queue(struct nvmet_fc_tgt_queue *queue) flush_workqueue(queue->work_q); - if (disconnect) - nvmet_sq_destroy(&queue->nvme_sq); + nvmet_sq_destroy(&queue->nvme_sq); nvmet_fc_tgt_q_put(queue); } @@ -778,17 +988,114 @@ nvmet_fc_find_target_queue(struct nvmet_fc_tgtport *tgtport, } static void +nvmet_fc_hostport_free(struct kref *ref) +{ + struct nvmet_fc_hostport *hostport = + container_of(ref, struct nvmet_fc_hostport, ref); + struct nvmet_fc_tgtport *tgtport = hostport->tgtport; + unsigned long flags; + + spin_lock_irqsave(&tgtport->lock, flags); + list_del(&hostport->host_list); + spin_unlock_irqrestore(&tgtport->lock, flags); + if (tgtport->ops->host_release && hostport->invalid) + tgtport->ops->host_release(hostport->hosthandle); + kfree(hostport); + nvmet_fc_tgtport_put(tgtport); +} + +static void +nvmet_fc_hostport_put(struct nvmet_fc_hostport *hostport) +{ + kref_put(&hostport->ref, nvmet_fc_hostport_free); +} + +static int +nvmet_fc_hostport_get(struct nvmet_fc_hostport *hostport) +{ + return kref_get_unless_zero(&hostport->ref); +} + +static void +nvmet_fc_free_hostport(struct nvmet_fc_hostport *hostport) +{ + /* if LLDD not implemented, leave as NULL */ + if (!hostport->hosthandle) + return; + + nvmet_fc_hostport_put(hostport); +} + +static struct nvmet_fc_hostport * +nvmet_fc_alloc_hostport(struct nvmet_fc_tgtport *tgtport, void *hosthandle) +{ + struct nvmet_fc_hostport *newhost, *host, *match = NULL; + unsigned long flags; + + /* if LLDD not implemented, leave as NULL */ + if (!hosthandle) + return NULL; + + /* take reference for what will be the newly allocated hostport */ + if (!nvmet_fc_tgtport_get(tgtport)) + return ERR_PTR(-EINVAL); + + newhost = kzalloc(sizeof(*newhost), GFP_KERNEL); + if (!newhost) { + spin_lock_irqsave(&tgtport->lock, flags); + list_for_each_entry(host, &tgtport->host_list, host_list) { + if (host->hosthandle == hosthandle && !host->invalid) { + if (nvmet_fc_hostport_get(host)) { + match = host; + break; + } + } + } + spin_unlock_irqrestore(&tgtport->lock, flags); + /* no allocation - release reference */ + nvmet_fc_tgtport_put(tgtport); + return (match) ? match : ERR_PTR(-ENOMEM); + } + + newhost->tgtport = tgtport; + newhost->hosthandle = hosthandle; + INIT_LIST_HEAD(&newhost->host_list); + kref_init(&newhost->ref); + + spin_lock_irqsave(&tgtport->lock, flags); + list_for_each_entry(host, &tgtport->host_list, host_list) { + if (host->hosthandle == hosthandle && !host->invalid) { + if (nvmet_fc_hostport_get(host)) { + match = host; + break; + } + } + } + if (match) { + kfree(newhost); + newhost = NULL; + /* releasing allocation - release reference */ + nvmet_fc_tgtport_put(tgtport); + } else + list_add_tail(&newhost->host_list, &tgtport->host_list); + spin_unlock_irqrestore(&tgtport->lock, flags); + + return (match) ? match : newhost; +} + +static void nvmet_fc_delete_assoc(struct work_struct *work) { struct nvmet_fc_tgt_assoc *assoc = container_of(work, struct nvmet_fc_tgt_assoc, del_work); nvmet_fc_delete_target_assoc(assoc); + atomic_set(&assoc->del_work_active, 0); nvmet_fc_tgt_a_put(assoc); } static struct nvmet_fc_tgt_assoc * -nvmet_fc_alloc_target_assoc(struct nvmet_fc_tgtport *tgtport) +nvmet_fc_alloc_target_assoc(struct nvmet_fc_tgtport *tgtport, void *hosthandle) { struct nvmet_fc_tgt_assoc *assoc, *tmpassoc; unsigned long flags; @@ -805,13 +1112,19 @@ nvmet_fc_alloc_target_assoc(struct nvmet_fc_tgtport *tgtport) goto out_free_assoc; if (!nvmet_fc_tgtport_get(tgtport)) - goto out_ida_put; + goto out_ida; + + assoc->hostport = nvmet_fc_alloc_hostport(tgtport, hosthandle); + if (IS_ERR(assoc->hostport)) + goto out_put; assoc->tgtport = tgtport; assoc->a_id = idx; INIT_LIST_HEAD(&assoc->a_list); kref_init(&assoc->ref); INIT_WORK(&assoc->del_work, nvmet_fc_delete_assoc); + atomic_set(&assoc->del_work_active, 0); + atomic_set(&assoc->terminating, 0); while (needrandom) { get_random_bytes(&ran, sizeof(ran) - BYTES_FOR_QID); @@ -819,11 +1132,12 @@ nvmet_fc_alloc_target_assoc(struct nvmet_fc_tgtport *tgtport) spin_lock_irqsave(&tgtport->lock, flags); needrandom = false; - list_for_each_entry(tmpassoc, &tgtport->assoc_list, a_list) + list_for_each_entry(tmpassoc, &tgtport->assoc_list, a_list) { if (ran == tmpassoc->association_id) { needrandom = true; break; } + } if (!needrandom) { assoc->association_id = ran; list_add_tail(&assoc->a_list, &tgtport->assoc_list); @@ -833,7 +1147,9 @@ nvmet_fc_alloc_target_assoc(struct nvmet_fc_tgtport *tgtport) return assoc; -out_ida_put: +out_put: + nvmet_fc_tgtport_put(tgtport); +out_ida: ida_simple_remove(&tgtport->assoc_cnt, idx); out_free_assoc: kfree(assoc); @@ -846,12 +1162,24 @@ nvmet_fc_target_assoc_free(struct kref *ref) struct nvmet_fc_tgt_assoc *assoc = container_of(ref, struct nvmet_fc_tgt_assoc, ref); struct nvmet_fc_tgtport *tgtport = assoc->tgtport; + struct nvmet_fc_ls_iod *oldls; unsigned long flags; + /* Send Disconnect now that all i/o has completed */ + nvmet_fc_xmt_disconnect_assoc(assoc); + + nvmet_fc_free_hostport(assoc->hostport); spin_lock_irqsave(&tgtport->lock, flags); list_del(&assoc->a_list); + oldls = assoc->rcv_disconn; spin_unlock_irqrestore(&tgtport->lock, flags); + /* if pending Rcv Disconnect Association LS, send rsp now */ + if (oldls) + nvmet_fc_xmt_ls_rsp(tgtport, oldls); ida_simple_remove(&tgtport->assoc_cnt, assoc->a_id); + dev_info(tgtport->dev, + "{%d:%d} Association freed\n", + tgtport->fc_target_port.port_num, assoc->a_id); kfree(assoc); nvmet_fc_tgtport_put(tgtport); } @@ -874,7 +1202,13 @@ nvmet_fc_delete_target_assoc(struct nvmet_fc_tgt_assoc *assoc) struct nvmet_fc_tgtport *tgtport = assoc->tgtport; struct nvmet_fc_tgt_queue *queue; unsigned long flags; - int i; + int i, terminating; + + terminating = atomic_xchg(&assoc->terminating, 1); + + /* if already terminating, do nothing */ + if (terminating) + return; spin_lock_irqsave(&tgtport->lock, flags); for (i = NVMET_NR_QUEUES; i >= 0; i--) { @@ -890,6 +1224,10 @@ nvmet_fc_delete_target_assoc(struct nvmet_fc_tgt_assoc *assoc) } spin_unlock_irqrestore(&tgtport->lock, flags); + dev_info(tgtport->dev, + "{%d:%d} Association deleted\n", + tgtport->fc_target_port.port_num, assoc->a_id); + nvmet_fc_tgt_a_put(assoc); } @@ -1048,16 +1386,21 @@ nvmet_fc_register_targetport(struct nvmet_fc_port_info *pinfo, newrec->fc_target_port.node_name = pinfo->node_name; newrec->fc_target_port.port_name = pinfo->port_name; - newrec->fc_target_port.private = &newrec[1]; + if (template->target_priv_sz) + newrec->fc_target_port.private = &newrec[1]; + else + newrec->fc_target_port.private = NULL; newrec->fc_target_port.port_id = pinfo->port_id; newrec->fc_target_port.port_num = idx; INIT_LIST_HEAD(&newrec->tgt_list); newrec->dev = dev; newrec->ops = template; spin_lock_init(&newrec->lock); - INIT_LIST_HEAD(&newrec->ls_list); + INIT_LIST_HEAD(&newrec->ls_rcv_list); + INIT_LIST_HEAD(&newrec->ls_req_list); INIT_LIST_HEAD(&newrec->ls_busylist); INIT_LIST_HEAD(&newrec->assoc_list); + INIT_LIST_HEAD(&newrec->host_list); kref_init(&newrec->ref); ida_init(&newrec->assoc_cnt); newrec->max_sg_cnt = template->max_sgl_segments; @@ -1134,17 +1477,90 @@ __nvmet_fc_free_assocs(struct nvmet_fc_tgtport *tgtport) { struct nvmet_fc_tgt_assoc *assoc, *next; unsigned long flags; + int ret; + + spin_lock_irqsave(&tgtport->lock, flags); + list_for_each_entry_safe(assoc, next, + &tgtport->assoc_list, a_list) { + if (!nvmet_fc_tgt_a_get(assoc)) + continue; + ret = atomic_cmpxchg(&assoc->del_work_active, 0, 1); + if (ret == 0) { + if (!schedule_work(&assoc->del_work)) + nvmet_fc_tgt_a_put(assoc); + } else { + /* already deleting - release local reference */ + nvmet_fc_tgt_a_put(assoc); + } + } + spin_unlock_irqrestore(&tgtport->lock, flags); +} + +/** + * nvmet_fc_invalidate_host - transport entry point called by an LLDD + * to remove references to a hosthandle for LS's. + * + * The nvmet-fc layer ensures that any references to the hosthandle + * on the targetport are forgotten (set to NULL). The LLDD will + * typically call this when a login with a remote host port has been + * lost, thus LS's for the remote host port are no longer possible. + * + * If an LS request is outstanding to the targetport/hosthandle (or + * issued concurrently with the call to invalidate the host), the + * LLDD is responsible for terminating/aborting the LS and completing + * the LS request. It is recommended that these terminations/aborts + * occur after calling to invalidate the host handle to avoid additional + * retries by the nvmet-fc transport. The nvmet-fc transport may + * continue to reference host handle while it cleans up outstanding + * NVME associations. The nvmet-fc transport will call the + * ops->host_release() callback to notify the LLDD that all references + * are complete and the related host handle can be recovered. + * Note: if there are no references, the callback may be called before + * the invalidate host call returns. + * + * @target_port: pointer to the (registered) target port that a prior + * LS was received on and which supplied the transport the + * hosthandle. + * @hosthandle: the handle (pointer) that represents the host port + * that no longer has connectivity and that LS's should + * no longer be directed to. + */ +void +nvmet_fc_invalidate_host(struct nvmet_fc_target_port *target_port, + void *hosthandle) +{ + struct nvmet_fc_tgtport *tgtport = targetport_to_tgtport(target_port); + struct nvmet_fc_tgt_assoc *assoc, *next; + unsigned long flags; + bool noassoc = true; + int ret; spin_lock_irqsave(&tgtport->lock, flags); list_for_each_entry_safe(assoc, next, &tgtport->assoc_list, a_list) { + if (!assoc->hostport || + assoc->hostport->hosthandle != hosthandle) + continue; if (!nvmet_fc_tgt_a_get(assoc)) continue; - if (!schedule_work(&assoc->del_work)) + assoc->hostport->invalid = 1; + noassoc = false; + ret = atomic_cmpxchg(&assoc->del_work_active, 0, 1); + if (ret == 0) { + if (!schedule_work(&assoc->del_work)) + nvmet_fc_tgt_a_put(assoc); + } else { + /* already deleting - release local reference */ nvmet_fc_tgt_a_put(assoc); + } } spin_unlock_irqrestore(&tgtport->lock, flags); + + /* if there's nothing to wait for - call the callback */ + if (noassoc && tgtport->ops->host_release) + tgtport->ops->host_release(hosthandle); } +EXPORT_SYMBOL_GPL(nvmet_fc_invalidate_host); /* * nvmet layer has called to terminate an association @@ -1157,6 +1573,7 @@ nvmet_fc_delete_ctrl(struct nvmet_ctrl *ctrl) struct nvmet_fc_tgt_queue *queue; unsigned long flags; bool found_ctrl = false; + int ret; /* this is a bit ugly, but don't want to make locks layered */ spin_lock_irqsave(&nvmet_fc_tgtlock, flags); @@ -1180,8 +1597,14 @@ nvmet_fc_delete_ctrl(struct nvmet_ctrl *ctrl) nvmet_fc_tgtport_put(tgtport); if (found_ctrl) { - if (!schedule_work(&assoc->del_work)) + ret = atomic_cmpxchg(&assoc->del_work_active, 0, 1); + if (ret == 0) { + if (!schedule_work(&assoc->del_work)) + nvmet_fc_tgt_a_put(assoc); + } else { + /* already deleting - release local reference */ nvmet_fc_tgt_a_put(assoc); + } return; } @@ -1211,6 +1634,13 @@ nvmet_fc_unregister_targetport(struct nvmet_fc_target_port *target_port) /* terminate any outstanding associations */ __nvmet_fc_free_assocs(tgtport); + /* + * should terminate LS's as well. However, LS's will be generated + * at the tail end of association termination, so they likely don't + * exist yet. And even if they did, it's worthwhile to just let + * them finish and targetport ref counting will clean things up. + */ + nvmet_fc_tgtport_put(tgtport); return 0; @@ -1218,113 +1648,15 @@ nvmet_fc_unregister_targetport(struct nvmet_fc_target_port *target_port) EXPORT_SYMBOL_GPL(nvmet_fc_unregister_targetport); -/* *********************** FC-NVME LS Handling **************************** */ - - -static void -nvmet_fc_format_rsp_hdr(void *buf, u8 ls_cmd, __be32 desc_len, u8 rqst_ls_cmd) -{ - struct fcnvme_ls_acc_hdr *acc = buf; - - acc->w0.ls_cmd = ls_cmd; - acc->desc_list_len = desc_len; - acc->rqst.desc_tag = cpu_to_be32(FCNVME_LSDESC_RQST); - acc->rqst.desc_len = - fcnvme_lsdesc_len(sizeof(struct fcnvme_lsdesc_rqst)); - acc->rqst.w0.ls_cmd = rqst_ls_cmd; -} +/* ********************** FC-NVME LS RCV Handling ************************* */ -static int -nvmet_fc_format_rjt(void *buf, u16 buflen, u8 ls_cmd, - u8 reason, u8 explanation, u8 vendor) -{ - struct fcnvme_ls_rjt *rjt = buf; - - nvmet_fc_format_rsp_hdr(buf, FCNVME_LSDESC_RQST, - fcnvme_lsdesc_len(sizeof(struct fcnvme_ls_rjt)), - ls_cmd); - rjt->rjt.desc_tag = cpu_to_be32(FCNVME_LSDESC_RJT); - rjt->rjt.desc_len = fcnvme_lsdesc_len(sizeof(struct fcnvme_lsdesc_rjt)); - rjt->rjt.reason_code = reason; - rjt->rjt.reason_explanation = explanation; - rjt->rjt.vendor = vendor; - - return sizeof(struct fcnvme_ls_rjt); -} - -/* Validation Error indexes into the string table below */ -enum { - VERR_NO_ERROR = 0, - VERR_CR_ASSOC_LEN = 1, - VERR_CR_ASSOC_RQST_LEN = 2, - VERR_CR_ASSOC_CMD = 3, - VERR_CR_ASSOC_CMD_LEN = 4, - VERR_ERSP_RATIO = 5, - VERR_ASSOC_ALLOC_FAIL = 6, - VERR_QUEUE_ALLOC_FAIL = 7, - VERR_CR_CONN_LEN = 8, - VERR_CR_CONN_RQST_LEN = 9, - VERR_ASSOC_ID = 10, - VERR_ASSOC_ID_LEN = 11, - VERR_NO_ASSOC = 12, - VERR_CONN_ID = 13, - VERR_CONN_ID_LEN = 14, - VERR_NO_CONN = 15, - VERR_CR_CONN_CMD = 16, - VERR_CR_CONN_CMD_LEN = 17, - VERR_DISCONN_LEN = 18, - VERR_DISCONN_RQST_LEN = 19, - VERR_DISCONN_CMD = 20, - VERR_DISCONN_CMD_LEN = 21, - VERR_DISCONN_SCOPE = 22, - VERR_RS_LEN = 23, - VERR_RS_RQST_LEN = 24, - VERR_RS_CMD = 25, - VERR_RS_CMD_LEN = 26, - VERR_RS_RCTL = 27, - VERR_RS_RO = 28, -}; - -static char *validation_errors[] = { - "OK", - "Bad CR_ASSOC Length", - "Bad CR_ASSOC Rqst Length", - "Not CR_ASSOC Cmd", - "Bad CR_ASSOC Cmd Length", - "Bad Ersp Ratio", - "Association Allocation Failed", - "Queue Allocation Failed", - "Bad CR_CONN Length", - "Bad CR_CONN Rqst Length", - "Not Association ID", - "Bad Association ID Length", - "No Association", - "Not Connection ID", - "Bad Connection ID Length", - "No Connection", - "Not CR_CONN Cmd", - "Bad CR_CONN Cmd Length", - "Bad DISCONN Length", - "Bad DISCONN Rqst Length", - "Not DISCONN Cmd", - "Bad DISCONN Cmd Length", - "Bad Disconnect Scope", - "Bad RS Length", - "Bad RS Rqst Length", - "Not RS Cmd", - "Bad RS Cmd Length", - "Bad RS R_CTL", - "Bad RS Relative Offset", -}; static void nvmet_fc_ls_create_association(struct nvmet_fc_tgtport *tgtport, struct nvmet_fc_ls_iod *iod) { - struct fcnvme_ls_cr_assoc_rqst *rqst = - (struct fcnvme_ls_cr_assoc_rqst *)iod->rqstbuf; - struct fcnvme_ls_cr_assoc_acc *acc = - (struct fcnvme_ls_cr_assoc_acc *)iod->rspbuf; + struct fcnvme_ls_cr_assoc_rqst *rqst = &iod->rqstbuf->rq_cr_assoc; + struct fcnvme_ls_cr_assoc_acc *acc = &iod->rspbuf->rsp_cr_assoc; struct nvmet_fc_tgt_queue *queue; int ret = 0; @@ -1356,7 +1688,8 @@ nvmet_fc_ls_create_association(struct nvmet_fc_tgtport *tgtport, else { /* new association w/ admin queue */ - iod->assoc = nvmet_fc_alloc_target_assoc(tgtport); + iod->assoc = nvmet_fc_alloc_target_assoc( + tgtport, iod->hosthandle); if (!iod->assoc) ret = VERR_ASSOC_ALLOC_FAIL; else { @@ -1371,8 +1704,8 @@ nvmet_fc_ls_create_association(struct nvmet_fc_tgtport *tgtport, dev_err(tgtport->dev, "Create Association LS failed: %s\n", validation_errors[ret]); - iod->lsreq->rsplen = nvmet_fc_format_rjt(acc, - NVME_FC_MAX_LS_BUFFER_SIZE, rqst->w0.ls_cmd, + iod->lsrsp->rsplen = nvme_fc_format_rjt(acc, + sizeof(*acc), rqst->w0.ls_cmd, FCNVME_RJT_RC_LOGIC, FCNVME_RJT_EXP_NONE, 0); return; @@ -1382,11 +1715,15 @@ nvmet_fc_ls_create_association(struct nvmet_fc_tgtport *tgtport, atomic_set(&queue->connected, 1); queue->sqhd = 0; /* best place to init value */ + dev_info(tgtport->dev, + "{%d:%d} Association created\n", + tgtport->fc_target_port.port_num, iod->assoc->a_id); + /* format a response */ - iod->lsreq->rsplen = sizeof(*acc); + iod->lsrsp->rsplen = sizeof(*acc); - nvmet_fc_format_rsp_hdr(acc, FCNVME_LS_ACC, + nvme_fc_format_rsp_hdr(acc, FCNVME_LS_ACC, fcnvme_lsdesc_len( sizeof(struct fcnvme_ls_cr_assoc_acc)), FCNVME_LS_CREATE_ASSOCIATION); @@ -1407,10 +1744,8 @@ static void nvmet_fc_ls_create_connection(struct nvmet_fc_tgtport *tgtport, struct nvmet_fc_ls_iod *iod) { - struct fcnvme_ls_cr_conn_rqst *rqst = - (struct fcnvme_ls_cr_conn_rqst *)iod->rqstbuf; - struct fcnvme_ls_cr_conn_acc *acc = - (struct fcnvme_ls_cr_conn_acc *)iod->rspbuf; + struct fcnvme_ls_cr_conn_rqst *rqst = &iod->rqstbuf->rq_cr_conn; + struct fcnvme_ls_cr_conn_acc *acc = &iod->rspbuf->rsp_cr_conn; struct nvmet_fc_tgt_queue *queue; int ret = 0; @@ -1462,8 +1797,8 @@ nvmet_fc_ls_create_connection(struct nvmet_fc_tgtport *tgtport, dev_err(tgtport->dev, "Create Connection LS failed: %s\n", validation_errors[ret]); - iod->lsreq->rsplen = nvmet_fc_format_rjt(acc, - NVME_FC_MAX_LS_BUFFER_SIZE, rqst->w0.ls_cmd, + iod->lsrsp->rsplen = nvme_fc_format_rjt(acc, + sizeof(*acc), rqst->w0.ls_cmd, (ret == VERR_NO_ASSOC) ? FCNVME_RJT_RC_INV_ASSOC : FCNVME_RJT_RC_LOGIC, @@ -1477,9 +1812,9 @@ nvmet_fc_ls_create_connection(struct nvmet_fc_tgtport *tgtport, /* format a response */ - iod->lsreq->rsplen = sizeof(*acc); + iod->lsrsp->rsplen = sizeof(*acc); - nvmet_fc_format_rsp_hdr(acc, FCNVME_LS_ACC, + nvme_fc_format_rsp_hdr(acc, FCNVME_LS_ACC, fcnvme_lsdesc_len(sizeof(struct fcnvme_ls_cr_conn_acc)), FCNVME_LS_CREATE_CONNECTION); acc->connectid.desc_tag = cpu_to_be32(FCNVME_LSDESC_CONN_ID); @@ -1491,46 +1826,28 @@ nvmet_fc_ls_create_connection(struct nvmet_fc_tgtport *tgtport, be16_to_cpu(rqst->connect_cmd.qid))); } -static void +/* + * Returns true if the LS response is to be transmit + * Returns false if the LS response is to be delayed + */ +static int nvmet_fc_ls_disconnect(struct nvmet_fc_tgtport *tgtport, struct nvmet_fc_ls_iod *iod) { struct fcnvme_ls_disconnect_assoc_rqst *rqst = - (struct fcnvme_ls_disconnect_assoc_rqst *)iod->rqstbuf; + &iod->rqstbuf->rq_dis_assoc; struct fcnvme_ls_disconnect_assoc_acc *acc = - (struct fcnvme_ls_disconnect_assoc_acc *)iod->rspbuf; - struct nvmet_fc_tgt_assoc *assoc; + &iod->rspbuf->rsp_dis_assoc; + struct nvmet_fc_tgt_assoc *assoc = NULL; + struct nvmet_fc_ls_iod *oldls = NULL; + unsigned long flags; int ret = 0; memset(acc, 0, sizeof(*acc)); - if (iod->rqstdatalen < sizeof(struct fcnvme_ls_disconnect_assoc_rqst)) - ret = VERR_DISCONN_LEN; - else if (rqst->desc_list_len != - fcnvme_lsdesc_len( - sizeof(struct fcnvme_ls_disconnect_assoc_rqst))) - ret = VERR_DISCONN_RQST_LEN; - else if (rqst->associd.desc_tag != cpu_to_be32(FCNVME_LSDESC_ASSOC_ID)) - ret = VERR_ASSOC_ID; - else if (rqst->associd.desc_len != - fcnvme_lsdesc_len( - sizeof(struct fcnvme_lsdesc_assoc_id))) - ret = VERR_ASSOC_ID_LEN; - else if (rqst->discon_cmd.desc_tag != - cpu_to_be32(FCNVME_LSDESC_DISCONN_CMD)) - ret = VERR_DISCONN_CMD; - else if (rqst->discon_cmd.desc_len != - fcnvme_lsdesc_len( - sizeof(struct fcnvme_lsdesc_disconn_cmd))) - ret = VERR_DISCONN_CMD_LEN; - /* - * As the standard changed on the LS, check if old format and scope - * something other than Association (e.g. 0). - */ - else if (rqst->discon_cmd.rsvd8[0]) - ret = VERR_DISCONN_SCOPE; - else { - /* match an active association */ + ret = nvmefc_vldt_lsreq_discon_assoc(iod->rqstdatalen, rqst); + if (!ret) { + /* match an active association - takes an assoc ref if !NULL */ assoc = nvmet_fc_find_target_assoc(tgtport, be64_to_cpu(rqst->associd.association_id)); iod->assoc = assoc; @@ -1538,34 +1855,63 @@ nvmet_fc_ls_disconnect(struct nvmet_fc_tgtport *tgtport, ret = VERR_NO_ASSOC; } - if (ret) { + if (ret || !assoc) { dev_err(tgtport->dev, "Disconnect LS failed: %s\n", validation_errors[ret]); - iod->lsreq->rsplen = nvmet_fc_format_rjt(acc, - NVME_FC_MAX_LS_BUFFER_SIZE, rqst->w0.ls_cmd, + iod->lsrsp->rsplen = nvme_fc_format_rjt(acc, + sizeof(*acc), rqst->w0.ls_cmd, (ret == VERR_NO_ASSOC) ? FCNVME_RJT_RC_INV_ASSOC : - (ret == VERR_NO_CONN) ? - FCNVME_RJT_RC_INV_CONN : - FCNVME_RJT_RC_LOGIC, + FCNVME_RJT_RC_LOGIC, FCNVME_RJT_EXP_NONE, 0); - return; + return true; } /* format a response */ - iod->lsreq->rsplen = sizeof(*acc); + iod->lsrsp->rsplen = sizeof(*acc); - nvmet_fc_format_rsp_hdr(acc, FCNVME_LS_ACC, + nvme_fc_format_rsp_hdr(acc, FCNVME_LS_ACC, fcnvme_lsdesc_len( sizeof(struct fcnvme_ls_disconnect_assoc_acc)), FCNVME_LS_DISCONNECT_ASSOC); /* release get taken in nvmet_fc_find_target_assoc */ - nvmet_fc_tgt_a_put(iod->assoc); + nvmet_fc_tgt_a_put(assoc); + + /* + * The rules for LS response says the response cannot + * go back until ABTS's have been sent for all outstanding + * I/O and a Disconnect Association LS has been sent. + * So... save off the Disconnect LS to send the response + * later. If there was a prior LS already saved, replace + * it with the newer one and send a can't perform reject + * on the older one. + */ + spin_lock_irqsave(&tgtport->lock, flags); + oldls = assoc->rcv_disconn; + assoc->rcv_disconn = iod; + spin_unlock_irqrestore(&tgtport->lock, flags); + + nvmet_fc_delete_target_assoc(assoc); - nvmet_fc_delete_target_assoc(iod->assoc); + if (oldls) { + dev_info(tgtport->dev, + "{%d:%d} Multiple Disconnect Association LS's " + "received\n", + tgtport->fc_target_port.port_num, assoc->a_id); + /* overwrite good response with bogus failure */ + oldls->lsrsp->rsplen = nvme_fc_format_rjt(oldls->rspbuf, + sizeof(*iod->rspbuf), + /* ok to use rqst, LS is same */ + rqst->w0.ls_cmd, + FCNVME_RJT_RC_UNAB, + FCNVME_RJT_EXP_NONE, 0); + nvmet_fc_xmt_ls_rsp(tgtport, oldls); + } + + return false; } @@ -1577,13 +1923,13 @@ static void nvmet_fc_fcp_nvme_cmd_done(struct nvmet_req *nvme_req); static const struct nvmet_fabrics_ops nvmet_fc_tgt_fcp_ops; static void -nvmet_fc_xmt_ls_rsp_done(struct nvmefc_tgt_ls_req *lsreq) +nvmet_fc_xmt_ls_rsp_done(struct nvmefc_ls_rsp *lsrsp) { - struct nvmet_fc_ls_iod *iod = lsreq->nvmet_fc_private; + struct nvmet_fc_ls_iod *iod = lsrsp->nvme_fc_private; struct nvmet_fc_tgtport *tgtport = iod->tgtport; fc_dma_sync_single_for_cpu(tgtport->dev, iod->rspdma, - NVME_FC_MAX_LS_BUFFER_SIZE, DMA_TO_DEVICE); + sizeof(*iod->rspbuf), DMA_TO_DEVICE); nvmet_fc_free_ls_iod(tgtport, iod); nvmet_fc_tgtport_put(tgtport); } @@ -1595,11 +1941,11 @@ nvmet_fc_xmt_ls_rsp(struct nvmet_fc_tgtport *tgtport, int ret; fc_dma_sync_single_for_device(tgtport->dev, iod->rspdma, - NVME_FC_MAX_LS_BUFFER_SIZE, DMA_TO_DEVICE); + sizeof(*iod->rspbuf), DMA_TO_DEVICE); - ret = tgtport->ops->xmt_ls_rsp(&tgtport->fc_target_port, iod->lsreq); + ret = tgtport->ops->xmt_ls_rsp(&tgtport->fc_target_port, iod->lsrsp); if (ret) - nvmet_fc_xmt_ls_rsp_done(iod->lsreq); + nvmet_fc_xmt_ls_rsp_done(iod->lsrsp); } /* @@ -1609,15 +1955,15 @@ static void nvmet_fc_handle_ls_rqst(struct nvmet_fc_tgtport *tgtport, struct nvmet_fc_ls_iod *iod) { - struct fcnvme_ls_rqst_w0 *w0 = - (struct fcnvme_ls_rqst_w0 *)iod->rqstbuf; + struct fcnvme_ls_rqst_w0 *w0 = &iod->rqstbuf->rq_cr_assoc.w0; + bool sendrsp = true; - iod->lsreq->nvmet_fc_private = iod; - iod->lsreq->rspbuf = iod->rspbuf; - iod->lsreq->rspdma = iod->rspdma; - iod->lsreq->done = nvmet_fc_xmt_ls_rsp_done; + iod->lsrsp->nvme_fc_private = iod; + iod->lsrsp->rspbuf = iod->rspbuf; + iod->lsrsp->rspdma = iod->rspdma; + iod->lsrsp->done = nvmet_fc_xmt_ls_rsp_done; /* Be preventative. handlers will later set to valid length */ - iod->lsreq->rsplen = 0; + iod->lsrsp->rsplen = 0; iod->assoc = NULL; @@ -1637,15 +1983,16 @@ nvmet_fc_handle_ls_rqst(struct nvmet_fc_tgtport *tgtport, break; case FCNVME_LS_DISCONNECT_ASSOC: /* Terminate a Queue/Connection or the Association */ - nvmet_fc_ls_disconnect(tgtport, iod); + sendrsp = nvmet_fc_ls_disconnect(tgtport, iod); break; default: - iod->lsreq->rsplen = nvmet_fc_format_rjt(iod->rspbuf, - NVME_FC_MAX_LS_BUFFER_SIZE, w0->ls_cmd, + iod->lsrsp->rsplen = nvme_fc_format_rjt(iod->rspbuf, + sizeof(*iod->rspbuf), w0->ls_cmd, FCNVME_RJT_RC_INVAL, FCNVME_RJT_EXP_NONE, 0); } - nvmet_fc_xmt_ls_rsp(tgtport, iod); + if (sendrsp) + nvmet_fc_xmt_ls_rsp(tgtport, iod); } /* @@ -1674,35 +2021,53 @@ nvmet_fc_handle_ls_rqst_work(struct work_struct *work) * * @target_port: pointer to the (registered) target port the LS was * received on. - * @lsreq: pointer to a lsreq request structure to be used to reference + * @lsrsp: pointer to a lsrsp structure to be used to reference * the exchange corresponding to the LS. * @lsreqbuf: pointer to the buffer containing the LS Request * @lsreqbuf_len: length, in bytes, of the received LS request */ int nvmet_fc_rcv_ls_req(struct nvmet_fc_target_port *target_port, - struct nvmefc_tgt_ls_req *lsreq, + void *hosthandle, + struct nvmefc_ls_rsp *lsrsp, void *lsreqbuf, u32 lsreqbuf_len) { struct nvmet_fc_tgtport *tgtport = targetport_to_tgtport(target_port); struct nvmet_fc_ls_iod *iod; - - if (lsreqbuf_len > NVME_FC_MAX_LS_BUFFER_SIZE) + struct fcnvme_ls_rqst_w0 *w0 = (struct fcnvme_ls_rqst_w0 *)lsreqbuf; + + if (lsreqbuf_len > sizeof(union nvmefc_ls_requests)) { + dev_info(tgtport->dev, + "RCV %s LS failed: payload too large (%d)\n", + (w0->ls_cmd <= NVME_FC_LAST_LS_CMD_VALUE) ? + nvmefc_ls_names[w0->ls_cmd] : "", + lsreqbuf_len); return -E2BIG; + } - if (!nvmet_fc_tgtport_get(tgtport)) + if (!nvmet_fc_tgtport_get(tgtport)) { + dev_info(tgtport->dev, + "RCV %s LS failed: target deleting\n", + (w0->ls_cmd <= NVME_FC_LAST_LS_CMD_VALUE) ? + nvmefc_ls_names[w0->ls_cmd] : ""); return -ESHUTDOWN; + } iod = nvmet_fc_alloc_ls_iod(tgtport); if (!iod) { + dev_info(tgtport->dev, + "RCV %s LS failed: context allocation failed\n", + (w0->ls_cmd <= NVME_FC_LAST_LS_CMD_VALUE) ? + nvmefc_ls_names[w0->ls_cmd] : ""); nvmet_fc_tgtport_put(tgtport); return -ENOENT; } - iod->lsreq = lsreq; + iod->lsrsp = lsrsp; iod->fcpreq = NULL; memcpy(iod->rqstbuf, lsreqbuf, lsreqbuf_len); iod->rqstdatalen = lsreqbuf_len; + iod->hosthandle = hosthandle; schedule_work(&iod->work); diff --git a/drivers/nvme/target/fcloop.c b/drivers/nvme/target/fcloop.c index f69ce66e2d44..2ff1d1334a03 100644 --- a/drivers/nvme/target/fcloop.c +++ b/drivers/nvme/target/fcloop.c @@ -208,10 +208,13 @@ struct fcloop_rport { }; struct fcloop_tport { - struct nvmet_fc_target_port *targetport; - struct nvme_fc_remote_port *remoteport; - struct fcloop_nport *nport; - struct fcloop_lport *lport; + struct nvmet_fc_target_port *targetport; + struct nvme_fc_remote_port *remoteport; + struct fcloop_nport *nport; + struct fcloop_lport *lport; + spinlock_t lock; + struct list_head ls_list; + struct work_struct ls_work; }; struct fcloop_nport { @@ -228,7 +231,8 @@ struct fcloop_nport { struct fcloop_lsreq { struct nvmefc_ls_req *lsreq; - struct nvmefc_tgt_ls_req tgt_ls_req; + struct nvmefc_ls_rsp ls_rsp; + int lsdir; /* H2T or T2H */ int status; struct list_head ls_list; /* fcloop_rport->ls_list */ }; @@ -267,9 +271,9 @@ struct fcloop_ini_fcpreq { }; static inline struct fcloop_lsreq * -tgt_ls_req_to_lsreq(struct nvmefc_tgt_ls_req *tgt_lsreq) +ls_rsp_to_lsreq(struct nvmefc_ls_rsp *lsrsp) { - return container_of(tgt_lsreq, struct fcloop_lsreq, tgt_ls_req); + return container_of(lsrsp, struct fcloop_lsreq, ls_rsp); } static inline struct fcloop_fcpreq * @@ -323,7 +327,7 @@ fcloop_rport_lsrqst_work(struct work_struct *work) } static int -fcloop_ls_req(struct nvme_fc_local_port *localport, +fcloop_h2t_ls_req(struct nvme_fc_local_port *localport, struct nvme_fc_remote_port *remoteport, struct nvmefc_ls_req *lsreq) { @@ -344,27 +348,28 @@ fcloop_ls_req(struct nvme_fc_local_port *localport, } tls_req->status = 0; - ret = nvmet_fc_rcv_ls_req(rport->targetport, &tls_req->tgt_ls_req, - lsreq->rqstaddr, lsreq->rqstlen); + ret = nvmet_fc_rcv_ls_req(rport->targetport, rport, + &tls_req->ls_rsp, + lsreq->rqstaddr, lsreq->rqstlen); return ret; } static int -fcloop_xmt_ls_rsp(struct nvmet_fc_target_port *targetport, - struct nvmefc_tgt_ls_req *tgt_lsreq) +fcloop_h2t_xmt_ls_rsp(struct nvmet_fc_target_port *targetport, + struct nvmefc_ls_rsp *lsrsp) { - struct fcloop_lsreq *tls_req = tgt_ls_req_to_lsreq(tgt_lsreq); + struct fcloop_lsreq *tls_req = ls_rsp_to_lsreq(lsrsp); struct nvmefc_ls_req *lsreq = tls_req->lsreq; struct fcloop_tport *tport = targetport->private; struct nvme_fc_remote_port *remoteport = tport->remoteport; struct fcloop_rport *rport; - memcpy(lsreq->rspaddr, tgt_lsreq->rspbuf, - ((lsreq->rsplen < tgt_lsreq->rsplen) ? - lsreq->rsplen : tgt_lsreq->rsplen)); + memcpy(lsreq->rspaddr, lsrsp->rspbuf, + ((lsreq->rsplen < lsrsp->rsplen) ? + lsreq->rsplen : lsrsp->rsplen)); - tgt_lsreq->done(tgt_lsreq); + lsrsp->done(lsrsp); if (remoteport) { rport = remoteport->private; @@ -377,6 +382,99 @@ fcloop_xmt_ls_rsp(struct nvmet_fc_target_port *targetport, return 0; } +static void +fcloop_tport_lsrqst_work(struct work_struct *work) +{ + struct fcloop_tport *tport = + container_of(work, struct fcloop_tport, ls_work); + struct fcloop_lsreq *tls_req; + + spin_lock(&tport->lock); + for (;;) { + tls_req = list_first_entry_or_null(&tport->ls_list, + struct fcloop_lsreq, ls_list); + if (!tls_req) + break; + + list_del(&tls_req->ls_list); + spin_unlock(&tport->lock); + + tls_req->lsreq->done(tls_req->lsreq, tls_req->status); + /* + * callee may free memory containing tls_req. + * do not reference lsreq after this. + */ + + spin_lock(&tport->lock); + } + spin_unlock(&tport->lock); +} + +static int +fcloop_t2h_ls_req(struct nvmet_fc_target_port *targetport, void *hosthandle, + struct nvmefc_ls_req *lsreq) +{ + struct fcloop_lsreq *tls_req = lsreq->private; + struct fcloop_tport *tport = targetport->private; + int ret = 0; + + /* + * hosthandle should be the dst.rport value. + * hosthandle ignored as fcloop currently is + * 1:1 tgtport vs remoteport + */ + tls_req->lsreq = lsreq; + INIT_LIST_HEAD(&tls_req->ls_list); + + if (!tport->remoteport) { + tls_req->status = -ECONNREFUSED; + spin_lock(&tport->lock); + list_add_tail(&tport->ls_list, &tls_req->ls_list); + spin_unlock(&tport->lock); + schedule_work(&tport->ls_work); + return ret; + } + + tls_req->status = 0; + ret = nvme_fc_rcv_ls_req(tport->remoteport, &tls_req->ls_rsp, + lsreq->rqstaddr, lsreq->rqstlen); + + return ret; +} + +static int +fcloop_t2h_xmt_ls_rsp(struct nvme_fc_local_port *localport, + struct nvme_fc_remote_port *remoteport, + struct nvmefc_ls_rsp *lsrsp) +{ + struct fcloop_lsreq *tls_req = ls_rsp_to_lsreq(lsrsp); + struct nvmefc_ls_req *lsreq = tls_req->lsreq; + struct fcloop_rport *rport = remoteport->private; + struct nvmet_fc_target_port *targetport = rport->targetport; + struct fcloop_tport *tport; + + memcpy(lsreq->rspaddr, lsrsp->rspbuf, + ((lsreq->rsplen < lsrsp->rsplen) ? + lsreq->rsplen : lsrsp->rsplen)); + lsrsp->done(lsrsp); + + if (targetport) { + tport = targetport->private; + spin_lock(&tport->lock); + list_add_tail(&tport->ls_list, &tls_req->ls_list); + spin_unlock(&tport->lock); + schedule_work(&tport->ls_work); + } + + return 0; +} + +static void +fcloop_t2h_host_release(void *hosthandle) +{ + /* host handle ignored for now */ +} + /* * Simulate reception of RSCN and converting it to a initiator transport * call to rescan a remote port. @@ -762,13 +860,19 @@ fcloop_fcp_req_release(struct nvmet_fc_target_port *tgtport, } static void -fcloop_ls_abort(struct nvme_fc_local_port *localport, +fcloop_h2t_ls_abort(struct nvme_fc_local_port *localport, struct nvme_fc_remote_port *remoteport, struct nvmefc_ls_req *lsreq) { } static void +fcloop_t2h_ls_abort(struct nvmet_fc_target_port *targetport, + void *hosthandle, struct nvmefc_ls_req *lsreq) +{ +} + +static void fcloop_fcp_abort(struct nvme_fc_local_port *localport, struct nvme_fc_remote_port *remoteport, void *hw_queue_handle, @@ -867,6 +971,7 @@ fcloop_targetport_delete(struct nvmet_fc_target_port *targetport) { struct fcloop_tport *tport = targetport->private; + flush_work(&tport->ls_work); fcloop_nport_put(tport->nport); } @@ -879,10 +984,11 @@ static struct nvme_fc_port_template fctemplate = { .remoteport_delete = fcloop_remoteport_delete, .create_queue = fcloop_create_queue, .delete_queue = fcloop_delete_queue, - .ls_req = fcloop_ls_req, + .ls_req = fcloop_h2t_ls_req, .fcp_io = fcloop_fcp_req, - .ls_abort = fcloop_ls_abort, + .ls_abort = fcloop_h2t_ls_abort, .fcp_abort = fcloop_fcp_abort, + .xmt_ls_rsp = fcloop_t2h_xmt_ls_rsp, .max_hw_queues = FCLOOP_HW_QUEUES, .max_sgl_segments = FCLOOP_SGL_SEGS, .max_dif_sgl_segments = FCLOOP_SGL_SEGS, @@ -896,11 +1002,14 @@ static struct nvme_fc_port_template fctemplate = { static struct nvmet_fc_target_template tgttemplate = { .targetport_delete = fcloop_targetport_delete, - .xmt_ls_rsp = fcloop_xmt_ls_rsp, + .xmt_ls_rsp = fcloop_h2t_xmt_ls_rsp, .fcp_op = fcloop_fcp_op, .fcp_abort = fcloop_tgt_fcp_abort, .fcp_req_release = fcloop_fcp_req_release, .discovery_event = fcloop_tgt_discovery_evt, + .ls_req = fcloop_t2h_ls_req, + .ls_abort = fcloop_t2h_ls_abort, + .host_release = fcloop_t2h_host_release, .max_hw_queues = FCLOOP_HW_QUEUES, .max_sgl_segments = FCLOOP_SGL_SEGS, .max_dif_sgl_segments = FCLOOP_SGL_SEGS, @@ -909,6 +1018,7 @@ static struct nvmet_fc_target_template tgttemplate = { .target_features = 0, /* sizes of additional private data for data structures */ .target_priv_sz = sizeof(struct fcloop_tport), + .lsrqst_priv_sz = sizeof(struct fcloop_lsreq), }; static ssize_t @@ -1258,6 +1368,9 @@ fcloop_create_target_port(struct device *dev, struct device_attribute *attr, tport->nport = nport; tport->lport = nport->lport; nport->tport = tport; + spin_lock_init(&tport->lock); + INIT_WORK(&tport->ls_work, fcloop_tport_lsrqst_work); + INIT_LIST_HEAD(&tport->ls_list); return count; } diff --git a/drivers/nvme/target/io-cmd-bdev.c b/drivers/nvme/target/io-cmd-bdev.c index ea0e596be15d..3dd6f566a240 100644 --- a/drivers/nvme/target/io-cmd-bdev.c +++ b/drivers/nvme/target/io-cmd-bdev.c @@ -47,6 +47,22 @@ void nvmet_bdev_set_limits(struct block_device *bdev, struct nvme_id_ns *id) id->nows = to0based(ql->io_opt / ql->logical_block_size); } +static void nvmet_bdev_ns_enable_integrity(struct nvmet_ns *ns) +{ + struct blk_integrity *bi = bdev_get_integrity(ns->bdev); + + if (bi) { + ns->metadata_size = bi->tuple_size; + if (bi->profile == &t10_pi_type1_crc) + ns->pi_type = NVME_NS_DPS_PI_TYPE1; + else if (bi->profile == &t10_pi_type3_crc) + ns->pi_type = NVME_NS_DPS_PI_TYPE3; + else + /* Unsupported metadata type */ + ns->metadata_size = 0; + } +} + int nvmet_bdev_ns_enable(struct nvmet_ns *ns) { int ret; @@ -64,6 +80,12 @@ int nvmet_bdev_ns_enable(struct nvmet_ns *ns) } ns->size = i_size_read(ns->bdev->bd_inode); ns->blksize_shift = blksize_bits(bdev_logical_block_size(ns->bdev)); + + ns->pi_type = 0; + ns->metadata_size = 0; + if (IS_ENABLED(CONFIG_BLK_DEV_INTEGRITY_T10)) + nvmet_bdev_ns_enable_integrity(ns); + return 0; } @@ -75,6 +97,11 @@ void nvmet_bdev_ns_disable(struct nvmet_ns *ns) } } +void nvmet_bdev_ns_revalidate(struct nvmet_ns *ns) +{ + ns->size = i_size_read(ns->bdev->bd_inode); +} + static u16 blk_to_nvme_status(struct nvmet_req *req, blk_status_t blk_sts) { u16 status = NVME_SC_SUCCESS; @@ -142,6 +169,61 @@ static void nvmet_bio_done(struct bio *bio) bio_put(bio); } +#ifdef CONFIG_BLK_DEV_INTEGRITY +static int nvmet_bdev_alloc_bip(struct nvmet_req *req, struct bio *bio, + struct sg_mapping_iter *miter) +{ + struct blk_integrity *bi; + struct bio_integrity_payload *bip; + struct block_device *bdev = req->ns->bdev; + int rc; + size_t resid, len; + + bi = bdev_get_integrity(bdev); + if (unlikely(!bi)) { + pr_err("Unable to locate bio_integrity\n"); + return -ENODEV; + } + + bip = bio_integrity_alloc(bio, GFP_NOIO, + min_t(unsigned int, req->metadata_sg_cnt, BIO_MAX_PAGES)); + if (IS_ERR(bip)) { + pr_err("Unable to allocate bio_integrity_payload\n"); + return PTR_ERR(bip); + } + + bip->bip_iter.bi_size = bio_integrity_bytes(bi, bio_sectors(bio)); + /* virtual start sector must be in integrity interval units */ + bip_set_seed(bip, bio->bi_iter.bi_sector >> + (bi->interval_exp - SECTOR_SHIFT)); + + resid = bip->bip_iter.bi_size; + while (resid > 0 && sg_miter_next(miter)) { + len = min_t(size_t, miter->length, resid); + rc = bio_integrity_add_page(bio, miter->page, len, + offset_in_page(miter->addr)); + if (unlikely(rc != len)) { + pr_err("bio_integrity_add_page() failed; %d\n", rc); + sg_miter_stop(miter); + return -ENOMEM; + } + + resid -= len; + if (len < miter->length) + miter->consumed -= miter->length - len; + } + sg_miter_stop(miter); + + return 0; +} +#else +static int nvmet_bdev_alloc_bip(struct nvmet_req *req, struct bio *bio, + struct sg_mapping_iter *miter) +{ + return -EINVAL; +} +#endif /* CONFIG_BLK_DEV_INTEGRITY */ + static void nvmet_bdev_execute_rw(struct nvmet_req *req) { int sg_cnt = req->sg_cnt; @@ -149,9 +231,12 @@ static void nvmet_bdev_execute_rw(struct nvmet_req *req) struct scatterlist *sg; struct blk_plug plug; sector_t sector; - int op, i; + int op, i, rc; + struct sg_mapping_iter prot_miter; + unsigned int iter_flags; + unsigned int total_len = nvmet_rw_data_len(req) + req->metadata_len; - if (!nvmet_check_data_len(req, nvmet_rw_len(req))) + if (!nvmet_check_transfer_len(req, total_len)) return; if (!req->sg_cnt) { @@ -163,8 +248,10 @@ static void nvmet_bdev_execute_rw(struct nvmet_req *req) op = REQ_OP_WRITE | REQ_SYNC | REQ_IDLE; if (req->cmd->rw.control & cpu_to_le16(NVME_RW_FUA)) op |= REQ_FUA; + iter_flags = SG_MITER_TO_SG; } else { op = REQ_OP_READ; + iter_flags = SG_MITER_FROM_SG; } if (is_pci_p2pdma_page(sg_page(req->sg))) @@ -186,11 +273,24 @@ static void nvmet_bdev_execute_rw(struct nvmet_req *req) bio->bi_opf = op; blk_start_plug(&plug); + if (req->metadata_len) + sg_miter_start(&prot_miter, req->metadata_sg, + req->metadata_sg_cnt, iter_flags); + for_each_sg(req->sg, sg, req->sg_cnt, i) { while (bio_add_page(bio, sg_page(sg), sg->length, sg->offset) != sg->length) { struct bio *prev = bio; + if (req->metadata_len) { + rc = nvmet_bdev_alloc_bip(req, bio, + &prot_miter); + if (unlikely(rc)) { + bio_io_error(bio); + return; + } + } + bio = bio_alloc(GFP_KERNEL, min(sg_cnt, BIO_MAX_PAGES)); bio_set_dev(bio, req->ns->bdev); bio->bi_iter.bi_sector = sector; @@ -204,6 +304,14 @@ static void nvmet_bdev_execute_rw(struct nvmet_req *req) sg_cnt--; } + if (req->metadata_len) { + rc = nvmet_bdev_alloc_bip(req, bio, &prot_miter); + if (unlikely(rc)) { + bio_io_error(bio); + return; + } + } + submit_bio(bio); blk_finish_plug(&plug); } @@ -212,7 +320,7 @@ static void nvmet_bdev_execute_flush(struct nvmet_req *req) { struct bio *bio = &req->b.inline_bio; - if (!nvmet_check_data_len(req, 0)) + if (!nvmet_check_transfer_len(req, 0)) return; bio_init(bio, req->inline_bvec, ARRAY_SIZE(req->inline_bvec)); @@ -226,7 +334,7 @@ static void nvmet_bdev_execute_flush(struct nvmet_req *req) u16 nvmet_bdev_flush(struct nvmet_req *req) { - if (blkdev_issue_flush(req->ns->bdev, GFP_KERNEL, NULL)) + if (blkdev_issue_flush(req->ns->bdev, GFP_KERNEL)) return NVME_SC_INTERNAL | NVME_SC_DNR; return 0; } @@ -304,7 +412,7 @@ static void nvmet_bdev_execute_write_zeroes(struct nvmet_req *req) sector_t nr_sector; int ret; - if (!nvmet_check_data_len(req, 0)) + if (!nvmet_check_transfer_len(req, 0)) return; sector = le64_to_cpu(write_zeroes->slba) << @@ -331,6 +439,8 @@ u16 nvmet_bdev_parse_io_cmd(struct nvmet_req *req) case nvme_cmd_read: case nvme_cmd_write: req->execute = nvmet_bdev_execute_rw; + if (req->sq->ctrl->pi_support && nvmet_ns_has_pi(req->ns)) + req->metadata_len = nvmet_rw_metadata_len(req); return 0; case nvme_cmd_flush: req->execute = nvmet_bdev_execute_flush; diff --git a/drivers/nvme/target/io-cmd-file.c b/drivers/nvme/target/io-cmd-file.c index cd5670b83118..0abbefd9925e 100644 --- a/drivers/nvme/target/io-cmd-file.c +++ b/drivers/nvme/target/io-cmd-file.c @@ -13,6 +13,18 @@ #define NVMET_MAX_MPOOL_BVEC 16 #define NVMET_MIN_MPOOL_OBJ 16 +int nvmet_file_ns_revalidate(struct nvmet_ns *ns) +{ + struct kstat stat; + int ret; + + ret = vfs_getattr(&ns->file->f_path, &stat, STATX_SIZE, + AT_STATX_FORCE_SYNC); + if (!ret) + ns->size = stat.size; + return ret; +} + void nvmet_file_ns_disable(struct nvmet_ns *ns) { if (ns->file) { @@ -30,7 +42,6 @@ void nvmet_file_ns_disable(struct nvmet_ns *ns) int nvmet_file_ns_enable(struct nvmet_ns *ns) { int flags = O_RDWR | O_LARGEFILE; - struct kstat stat; int ret; if (!ns->buffered_io) @@ -43,12 +54,10 @@ int nvmet_file_ns_enable(struct nvmet_ns *ns) return PTR_ERR(ns->file); } - ret = vfs_getattr(&ns->file->f_path, - &stat, STATX_SIZE, AT_STATX_FORCE_SYNC); + ret = nvmet_file_ns_revalidate(ns); if (ret) goto err; - ns->size = stat.size; /* * i_blkbits can be greater than the universally accepted upper bound, * so make sure we export a sane namespace lba_shift. @@ -232,7 +241,7 @@ static void nvmet_file_execute_rw(struct nvmet_req *req) { ssize_t nr_bvec = req->sg_cnt; - if (!nvmet_check_data_len(req, nvmet_rw_len(req))) + if (!nvmet_check_transfer_len(req, nvmet_rw_data_len(req))) return; if (!req->sg_cnt || !nr_bvec) { @@ -276,7 +285,7 @@ static void nvmet_file_flush_work(struct work_struct *w) static void nvmet_file_execute_flush(struct nvmet_req *req) { - if (!nvmet_check_data_len(req, 0)) + if (!nvmet_check_transfer_len(req, 0)) return; INIT_WORK(&req->f.work, nvmet_file_flush_work); schedule_work(&req->f.work); @@ -366,7 +375,7 @@ static void nvmet_file_write_zeroes_work(struct work_struct *w) static void nvmet_file_execute_write_zeroes(struct nvmet_req *req) { - if (!nvmet_check_data_len(req, 0)) + if (!nvmet_check_transfer_len(req, 0)) return; INIT_WORK(&req->f.work, nvmet_file_write_zeroes_work); schedule_work(&req->f.work); diff --git a/drivers/nvme/target/nvmet.h b/drivers/nvme/target/nvmet.h index 421dff3ea143..809691291e73 100644 --- a/drivers/nvme/target/nvmet.h +++ b/drivers/nvme/target/nvmet.h @@ -19,6 +19,7 @@ #include <linux/rcupdate.h> #include <linux/blkdev.h> #include <linux/radix-tree.h> +#include <linux/t10-pi.h> #define NVMET_ASYNC_EVENTS 4 #define NVMET_ERROR_LOG_SLOTS 128 @@ -77,6 +78,8 @@ struct nvmet_ns { int use_p2pmem; struct pci_dev *p2p_dev; + int pi_type; + int metadata_size; }; static inline struct nvmet_ns *to_nvmet_ns(struct config_item *item) @@ -142,6 +145,7 @@ struct nvmet_port { bool enabled; int inline_data_size; const struct nvmet_fabrics_ops *tr_ops; + bool pi_enable; }; static inline struct nvmet_port *to_nvmet_port(struct config_item *item) @@ -201,6 +205,7 @@ struct nvmet_ctrl { spinlock_t error_lock; u64 err_counter; struct nvme_error_slot slots[NVMET_ERROR_LOG_SLOTS]; + bool pi_support; }; struct nvmet_subsys_model { @@ -230,6 +235,7 @@ struct nvmet_subsys { u64 ver; u64 serial; char *subsysnqn; + bool pi_support; struct config_group group; @@ -281,6 +287,7 @@ struct nvmet_fabrics_ops { unsigned int type; unsigned int msdbd; bool has_keyed_sgls : 1; + bool metadata_support : 1; void (*queue_response)(struct nvmet_req *req); int (*add_port)(struct nvmet_port *port); void (*remove_port)(struct nvmet_port *port); @@ -302,6 +309,7 @@ struct nvmet_req { struct nvmet_cq *cq; struct nvmet_ns *ns; struct scatterlist *sg; + struct scatterlist *metadata_sg; struct bio_vec inline_bvec[NVMET_MAX_INLINE_BIOVEC]; union { struct { @@ -315,8 +323,10 @@ struct nvmet_req { } f; }; int sg_cnt; + int metadata_sg_cnt; /* data length as parsed from the SGL descriptor: */ size_t transfer_len; + size_t metadata_len; struct nvmet_port *port; @@ -384,11 +394,11 @@ u16 nvmet_parse_fabrics_cmd(struct nvmet_req *req); bool nvmet_req_init(struct nvmet_req *req, struct nvmet_cq *cq, struct nvmet_sq *sq, const struct nvmet_fabrics_ops *ops); void nvmet_req_uninit(struct nvmet_req *req); -bool nvmet_check_data_len(struct nvmet_req *req, size_t data_len); +bool nvmet_check_transfer_len(struct nvmet_req *req, size_t len); bool nvmet_check_data_len_lte(struct nvmet_req *req, size_t data_len); void nvmet_req_complete(struct nvmet_req *req, u16 status); -int nvmet_req_alloc_sgl(struct nvmet_req *req); -void nvmet_req_free_sgl(struct nvmet_req *req); +int nvmet_req_alloc_sgls(struct nvmet_req *req); +void nvmet_req_free_sgls(struct nvmet_req *req); void nvmet_execute_keep_alive(struct nvmet_req *req); @@ -498,13 +508,24 @@ void nvmet_file_ns_disable(struct nvmet_ns *ns); u16 nvmet_bdev_flush(struct nvmet_req *req); u16 nvmet_file_flush(struct nvmet_req *req); void nvmet_ns_changed(struct nvmet_subsys *subsys, u32 nsid); +void nvmet_bdev_ns_revalidate(struct nvmet_ns *ns); +int nvmet_file_ns_revalidate(struct nvmet_ns *ns); +void nvmet_ns_revalidate(struct nvmet_ns *ns); -static inline u32 nvmet_rw_len(struct nvmet_req *req) +static inline u32 nvmet_rw_data_len(struct nvmet_req *req) { return ((u32)le16_to_cpu(req->cmd->rw.length) + 1) << req->ns->blksize_shift; } +static inline u32 nvmet_rw_metadata_len(struct nvmet_req *req) +{ + if (!IS_ENABLED(CONFIG_BLK_DEV_INTEGRITY)) + return 0; + return ((u32)le16_to_cpu(req->cmd->rw.length) + 1) * + req->ns->metadata_size; +} + static inline u32 nvmet_dsm_len(struct nvmet_req *req) { return (le32_to_cpu(req->cmd->dsm.nr) + 1) * @@ -519,4 +540,11 @@ static inline __le16 to0based(u32 a) return cpu_to_le16(max(1U, min(1U << 16, a)) - 1); } +static inline bool nvmet_ns_has_pi(struct nvmet_ns *ns) +{ + if (!IS_ENABLED(CONFIG_BLK_DEV_INTEGRITY)) + return false; + return ns->pi_type && ns->metadata_size == sizeof(struct t10_pi_tuple); +} + #endif /* _NVMET_H */ diff --git a/drivers/nvme/target/rdma.c b/drivers/nvme/target/rdma.c index fd47de0e4e4e..d5141780592e 100644 --- a/drivers/nvme/target/rdma.c +++ b/drivers/nvme/target/rdma.c @@ -33,6 +33,9 @@ /* Assume mpsmin == device_page_size == 4KB */ #define NVMET_RDMA_MAX_MDTS 8 +#define NVMET_RDMA_MAX_METADATA_MDTS 5 + +struct nvmet_rdma_srq; struct nvmet_rdma_cmd { struct ib_sge sge[NVMET_RDMA_MAX_INLINE_SGE + 1]; @@ -41,6 +44,7 @@ struct nvmet_rdma_cmd { struct scatterlist inline_sg[NVMET_RDMA_MAX_INLINE_SGE]; struct nvme_command *nvme_cmd; struct nvmet_rdma_queue *queue; + struct nvmet_rdma_srq *nsrq; }; enum { @@ -57,6 +61,7 @@ struct nvmet_rdma_rsp { struct nvmet_rdma_queue *queue; struct ib_cqe read_cqe; + struct ib_cqe write_cqe; struct rdma_rw_ctx rw; struct nvmet_req req; @@ -83,6 +88,7 @@ struct nvmet_rdma_queue { struct ib_cq *cq; atomic_t sq_wr_avail; struct nvmet_rdma_device *dev; + struct nvmet_rdma_srq *nsrq; spinlock_t state_lock; enum nvmet_rdma_queue_state state; struct nvmet_cq nvme_cq; @@ -100,6 +106,7 @@ struct nvmet_rdma_queue { int idx; int host_qid; + int comp_vector; int recv_queue_size; int send_queue_size; @@ -113,11 +120,17 @@ struct nvmet_rdma_port { struct delayed_work repair_work; }; +struct nvmet_rdma_srq { + struct ib_srq *srq; + struct nvmet_rdma_cmd *cmds; + struct nvmet_rdma_device *ndev; +}; + struct nvmet_rdma_device { struct ib_device *device; struct ib_pd *pd; - struct ib_srq *srq; - struct nvmet_rdma_cmd *srq_cmds; + struct nvmet_rdma_srq **srqs; + int srq_count; size_t srq_size; struct kref ref; struct list_head entry; @@ -129,6 +142,16 @@ static bool nvmet_rdma_use_srq; module_param_named(use_srq, nvmet_rdma_use_srq, bool, 0444); MODULE_PARM_DESC(use_srq, "Use shared receive queue."); +static int srq_size_set(const char *val, const struct kernel_param *kp); +static const struct kernel_param_ops srq_size_ops = { + .set = srq_size_set, + .get = param_get_int, +}; + +static int nvmet_rdma_srq_size = 1024; +module_param_cb(srq_size, &srq_size_ops, &nvmet_rdma_srq_size, 0644); +MODULE_PARM_DESC(srq_size, "set Shared Receive Queue (SRQ) size, should >= 256 (default: 1024)"); + static DEFINE_IDA(nvmet_rdma_queue_ida); static LIST_HEAD(nvmet_rdma_queue_list); static DEFINE_MUTEX(nvmet_rdma_queue_mutex); @@ -140,6 +163,7 @@ static bool nvmet_rdma_execute_command(struct nvmet_rdma_rsp *rsp); static void nvmet_rdma_send_done(struct ib_cq *cq, struct ib_wc *wc); static void nvmet_rdma_recv_done(struct ib_cq *cq, struct ib_wc *wc); static void nvmet_rdma_read_data_done(struct ib_cq *cq, struct ib_wc *wc); +static void nvmet_rdma_write_data_done(struct ib_cq *cq, struct ib_wc *wc); static void nvmet_rdma_qp_event(struct ib_event *event, void *priv); static void nvmet_rdma_queue_disconnect(struct nvmet_rdma_queue *queue); static void nvmet_rdma_free_rsp(struct nvmet_rdma_device *ndev, @@ -149,6 +173,17 @@ static int nvmet_rdma_alloc_rsp(struct nvmet_rdma_device *ndev, static const struct nvmet_fabrics_ops nvmet_rdma_ops; +static int srq_size_set(const char *val, const struct kernel_param *kp) +{ + int n = 0, ret; + + ret = kstrtoint(val, 10, &n); + if (ret != 0 || n < 256) + return -EINVAL; + + return param_set_int(val, kp); +} + static int num_pages(int len) { return 1 + (((len - 1) & PAGE_MASK) >> PAGE_SHIFT); @@ -391,6 +426,9 @@ static int nvmet_rdma_alloc_rsp(struct nvmet_rdma_device *ndev, /* Data In / RDMA READ */ r->read_cqe.done = nvmet_rdma_read_data_done; + /* Data Out / RDMA WRITE */ + r->write_cqe.done = nvmet_rdma_write_data_done; + return 0; out_free_rsp: @@ -466,8 +504,8 @@ static int nvmet_rdma_post_recv(struct nvmet_rdma_device *ndev, cmd->sge[0].addr, cmd->sge[0].length, DMA_FROM_DEVICE); - if (ndev->srq) - ret = ib_post_srq_recv(ndev->srq, &cmd->wr, NULL); + if (cmd->nsrq) + ret = ib_post_srq_recv(cmd->nsrq->srq, &cmd->wr, NULL); else ret = ib_post_recv(cmd->queue->qp, &cmd->wr, NULL); @@ -500,6 +538,129 @@ static void nvmet_rdma_process_wr_wait_list(struct nvmet_rdma_queue *queue) spin_unlock(&queue->rsp_wr_wait_lock); } +static u16 nvmet_rdma_check_pi_status(struct ib_mr *sig_mr) +{ + struct ib_mr_status mr_status; + int ret; + u16 status = 0; + + ret = ib_check_mr_status(sig_mr, IB_MR_CHECK_SIG_STATUS, &mr_status); + if (ret) { + pr_err("ib_check_mr_status failed, ret %d\n", ret); + return NVME_SC_INVALID_PI; + } + + if (mr_status.fail_status & IB_MR_CHECK_SIG_STATUS) { + switch (mr_status.sig_err.err_type) { + case IB_SIG_BAD_GUARD: + status = NVME_SC_GUARD_CHECK; + break; + case IB_SIG_BAD_REFTAG: + status = NVME_SC_REFTAG_CHECK; + break; + case IB_SIG_BAD_APPTAG: + status = NVME_SC_APPTAG_CHECK; + break; + } + pr_err("PI error found type %d expected 0x%x vs actual 0x%x\n", + mr_status.sig_err.err_type, + mr_status.sig_err.expected, + mr_status.sig_err.actual); + } + + return status; +} + +static void nvmet_rdma_set_sig_domain(struct blk_integrity *bi, + struct nvme_command *cmd, struct ib_sig_domain *domain, + u16 control, u8 pi_type) +{ + domain->sig_type = IB_SIG_TYPE_T10_DIF; + domain->sig.dif.bg_type = IB_T10DIF_CRC; + domain->sig.dif.pi_interval = 1 << bi->interval_exp; + domain->sig.dif.ref_tag = le32_to_cpu(cmd->rw.reftag); + if (control & NVME_RW_PRINFO_PRCHK_REF) + domain->sig.dif.ref_remap = true; + + domain->sig.dif.app_tag = le16_to_cpu(cmd->rw.apptag); + domain->sig.dif.apptag_check_mask = le16_to_cpu(cmd->rw.appmask); + domain->sig.dif.app_escape = true; + if (pi_type == NVME_NS_DPS_PI_TYPE3) + domain->sig.dif.ref_escape = true; +} + +static void nvmet_rdma_set_sig_attrs(struct nvmet_req *req, + struct ib_sig_attrs *sig_attrs) +{ + struct nvme_command *cmd = req->cmd; + u16 control = le16_to_cpu(cmd->rw.control); + u8 pi_type = req->ns->pi_type; + struct blk_integrity *bi; + + bi = bdev_get_integrity(req->ns->bdev); + + memset(sig_attrs, 0, sizeof(*sig_attrs)); + + if (control & NVME_RW_PRINFO_PRACT) { + /* for WRITE_INSERT/READ_STRIP no wire domain */ + sig_attrs->wire.sig_type = IB_SIG_TYPE_NONE; + nvmet_rdma_set_sig_domain(bi, cmd, &sig_attrs->mem, control, + pi_type); + /* Clear the PRACT bit since HCA will generate/verify the PI */ + control &= ~NVME_RW_PRINFO_PRACT; + cmd->rw.control = cpu_to_le16(control); + /* PI is added by the HW */ + req->transfer_len += req->metadata_len; + } else { + /* for WRITE_PASS/READ_PASS both wire/memory domains exist */ + nvmet_rdma_set_sig_domain(bi, cmd, &sig_attrs->wire, control, + pi_type); + nvmet_rdma_set_sig_domain(bi, cmd, &sig_attrs->mem, control, + pi_type); + } + + if (control & NVME_RW_PRINFO_PRCHK_REF) + sig_attrs->check_mask |= IB_SIG_CHECK_REFTAG; + if (control & NVME_RW_PRINFO_PRCHK_GUARD) + sig_attrs->check_mask |= IB_SIG_CHECK_GUARD; + if (control & NVME_RW_PRINFO_PRCHK_APP) + sig_attrs->check_mask |= IB_SIG_CHECK_APPTAG; +} + +static int nvmet_rdma_rw_ctx_init(struct nvmet_rdma_rsp *rsp, u64 addr, u32 key, + struct ib_sig_attrs *sig_attrs) +{ + struct rdma_cm_id *cm_id = rsp->queue->cm_id; + struct nvmet_req *req = &rsp->req; + int ret; + + if (req->metadata_len) + ret = rdma_rw_ctx_signature_init(&rsp->rw, cm_id->qp, + cm_id->port_num, req->sg, req->sg_cnt, + req->metadata_sg, req->metadata_sg_cnt, sig_attrs, + addr, key, nvmet_data_dir(req)); + else + ret = rdma_rw_ctx_init(&rsp->rw, cm_id->qp, cm_id->port_num, + req->sg, req->sg_cnt, 0, addr, key, + nvmet_data_dir(req)); + + return ret; +} + +static void nvmet_rdma_rw_ctx_destroy(struct nvmet_rdma_rsp *rsp) +{ + struct rdma_cm_id *cm_id = rsp->queue->cm_id; + struct nvmet_req *req = &rsp->req; + + if (req->metadata_len) + rdma_rw_ctx_destroy_signature(&rsp->rw, cm_id->qp, + cm_id->port_num, req->sg, req->sg_cnt, + req->metadata_sg, req->metadata_sg_cnt, + nvmet_data_dir(req)); + else + rdma_rw_ctx_destroy(&rsp->rw, cm_id->qp, cm_id->port_num, + req->sg, req->sg_cnt, nvmet_data_dir(req)); +} static void nvmet_rdma_release_rsp(struct nvmet_rdma_rsp *rsp) { @@ -507,14 +668,11 @@ static void nvmet_rdma_release_rsp(struct nvmet_rdma_rsp *rsp) atomic_add(1 + rsp->n_rdma, &queue->sq_wr_avail); - if (rsp->n_rdma) { - rdma_rw_ctx_destroy(&rsp->rw, queue->qp, - queue->cm_id->port_num, rsp->req.sg, - rsp->req.sg_cnt, nvmet_data_dir(&rsp->req)); - } + if (rsp->n_rdma) + nvmet_rdma_rw_ctx_destroy(rsp); if (rsp->req.sg != rsp->cmd->inline_sg) - nvmet_req_free_sgl(&rsp->req); + nvmet_req_free_sgls(&rsp->req); if (unlikely(!list_empty_careful(&queue->rsp_wr_wait_list))) nvmet_rdma_process_wr_wait_list(queue); @@ -566,11 +724,16 @@ static void nvmet_rdma_queue_response(struct nvmet_req *req) rsp->send_wr.opcode = IB_WR_SEND; } - if (nvmet_rdma_need_data_out(rsp)) - first_wr = rdma_rw_ctx_wrs(&rsp->rw, cm_id->qp, - cm_id->port_num, NULL, &rsp->send_wr); - else + if (nvmet_rdma_need_data_out(rsp)) { + if (rsp->req.metadata_len) + first_wr = rdma_rw_ctx_wrs(&rsp->rw, cm_id->qp, + cm_id->port_num, &rsp->write_cqe, NULL); + else + first_wr = rdma_rw_ctx_wrs(&rsp->rw, cm_id->qp, + cm_id->port_num, NULL, &rsp->send_wr); + } else { first_wr = &rsp->send_wr; + } nvmet_rdma_post_recv(rsp->queue->dev, rsp->cmd); @@ -589,15 +752,14 @@ static void nvmet_rdma_read_data_done(struct ib_cq *cq, struct ib_wc *wc) struct nvmet_rdma_rsp *rsp = container_of(wc->wr_cqe, struct nvmet_rdma_rsp, read_cqe); struct nvmet_rdma_queue *queue = cq->cq_context; + u16 status = 0; WARN_ON(rsp->n_rdma <= 0); atomic_add(rsp->n_rdma, &queue->sq_wr_avail); - rdma_rw_ctx_destroy(&rsp->rw, queue->qp, - queue->cm_id->port_num, rsp->req.sg, - rsp->req.sg_cnt, nvmet_data_dir(&rsp->req)); rsp->n_rdma = 0; if (unlikely(wc->status != IB_WC_SUCCESS)) { + nvmet_rdma_rw_ctx_destroy(rsp); nvmet_req_uninit(&rsp->req); nvmet_rdma_release_rsp(rsp); if (wc->status != IB_WC_WR_FLUSH_ERR) { @@ -608,7 +770,58 @@ static void nvmet_rdma_read_data_done(struct ib_cq *cq, struct ib_wc *wc) return; } - rsp->req.execute(&rsp->req); + if (rsp->req.metadata_len) + status = nvmet_rdma_check_pi_status(rsp->rw.reg->mr); + nvmet_rdma_rw_ctx_destroy(rsp); + + if (unlikely(status)) + nvmet_req_complete(&rsp->req, status); + else + rsp->req.execute(&rsp->req); +} + +static void nvmet_rdma_write_data_done(struct ib_cq *cq, struct ib_wc *wc) +{ + struct nvmet_rdma_rsp *rsp = + container_of(wc->wr_cqe, struct nvmet_rdma_rsp, write_cqe); + struct nvmet_rdma_queue *queue = cq->cq_context; + struct rdma_cm_id *cm_id = rsp->queue->cm_id; + u16 status; + + if (!IS_ENABLED(CONFIG_BLK_DEV_INTEGRITY)) + return; + + WARN_ON(rsp->n_rdma <= 0); + atomic_add(rsp->n_rdma, &queue->sq_wr_avail); + rsp->n_rdma = 0; + + if (unlikely(wc->status != IB_WC_SUCCESS)) { + nvmet_rdma_rw_ctx_destroy(rsp); + nvmet_req_uninit(&rsp->req); + nvmet_rdma_release_rsp(rsp); + if (wc->status != IB_WC_WR_FLUSH_ERR) { + pr_info("RDMA WRITE for CQE 0x%p failed with status %s (%d).\n", + wc->wr_cqe, ib_wc_status_msg(wc->status), + wc->status); + nvmet_rdma_error_comp(queue); + } + return; + } + + /* + * Upon RDMA completion check the signature status + * - if succeeded send good NVMe response + * - if failed send bad NVMe response with appropriate error + */ + status = nvmet_rdma_check_pi_status(rsp->rw.reg->mr); + if (unlikely(status)) + rsp->req.cqe->status = cpu_to_le16(status << 1); + nvmet_rdma_rw_ctx_destroy(rsp); + + if (unlikely(ib_post_send(cm_id->qp, &rsp->send_wr, NULL))) { + pr_err("sending cmd response failed\n"); + nvmet_rdma_release_rsp(rsp); + } } static void nvmet_rdma_use_inline_sg(struct nvmet_rdma_rsp *rsp, u32 len, @@ -665,9 +878,9 @@ static u16 nvmet_rdma_map_sgl_inline(struct nvmet_rdma_rsp *rsp) static u16 nvmet_rdma_map_sgl_keyed(struct nvmet_rdma_rsp *rsp, struct nvme_keyed_sgl_desc *sgl, bool invalidate) { - struct rdma_cm_id *cm_id = rsp->queue->cm_id; u64 addr = le64_to_cpu(sgl->addr); u32 key = get_unaligned_le32(sgl->key); + struct ib_sig_attrs sig_attrs; int ret; rsp->req.transfer_len = get_unaligned_le24(sgl->length); @@ -676,13 +889,14 @@ static u16 nvmet_rdma_map_sgl_keyed(struct nvmet_rdma_rsp *rsp, if (!rsp->req.transfer_len) return 0; - ret = nvmet_req_alloc_sgl(&rsp->req); + if (rsp->req.metadata_len) + nvmet_rdma_set_sig_attrs(&rsp->req, &sig_attrs); + + ret = nvmet_req_alloc_sgls(&rsp->req); if (unlikely(ret < 0)) goto error_out; - ret = rdma_rw_ctx_init(&rsp->rw, cm_id->qp, cm_id->port_num, - rsp->req.sg, rsp->req.sg_cnt, 0, addr, key, - nvmet_data_dir(&rsp->req)); + ret = nvmet_rdma_rw_ctx_init(rsp, addr, key, &sig_attrs); if (unlikely(ret < 0)) goto error_out; rsp->n_rdma += ret; @@ -845,23 +1059,40 @@ static void nvmet_rdma_recv_done(struct ib_cq *cq, struct ib_wc *wc) nvmet_rdma_handle_command(queue, rsp); } -static void nvmet_rdma_destroy_srq(struct nvmet_rdma_device *ndev) +static void nvmet_rdma_destroy_srq(struct nvmet_rdma_srq *nsrq) { - if (!ndev->srq) + nvmet_rdma_free_cmds(nsrq->ndev, nsrq->cmds, nsrq->ndev->srq_size, + false); + ib_destroy_srq(nsrq->srq); + + kfree(nsrq); +} + +static void nvmet_rdma_destroy_srqs(struct nvmet_rdma_device *ndev) +{ + int i; + + if (!ndev->srqs) return; - nvmet_rdma_free_cmds(ndev, ndev->srq_cmds, ndev->srq_size, false); - ib_destroy_srq(ndev->srq); + for (i = 0; i < ndev->srq_count; i++) + nvmet_rdma_destroy_srq(ndev->srqs[i]); + + kfree(ndev->srqs); } -static int nvmet_rdma_init_srq(struct nvmet_rdma_device *ndev) +static struct nvmet_rdma_srq * +nvmet_rdma_init_srq(struct nvmet_rdma_device *ndev) { struct ib_srq_init_attr srq_attr = { NULL, }; + size_t srq_size = ndev->srq_size; + struct nvmet_rdma_srq *nsrq; struct ib_srq *srq; - size_t srq_size; int ret, i; - srq_size = 4095; /* XXX: tune */ + nsrq = kzalloc(sizeof(*nsrq), GFP_KERNEL); + if (!nsrq) + return ERR_PTR(-ENOMEM); srq_attr.attr.max_wr = srq_size; srq_attr.attr.max_sge = 1 + ndev->inline_page_count; @@ -869,35 +1100,73 @@ static int nvmet_rdma_init_srq(struct nvmet_rdma_device *ndev) srq_attr.srq_type = IB_SRQT_BASIC; srq = ib_create_srq(ndev->pd, &srq_attr); if (IS_ERR(srq)) { - /* - * If SRQs aren't supported we just go ahead and use normal - * non-shared receive queues. - */ - pr_info("SRQ requested but not supported.\n"); - return 0; + ret = PTR_ERR(srq); + goto out_free; } - ndev->srq_cmds = nvmet_rdma_alloc_cmds(ndev, srq_size, false); - if (IS_ERR(ndev->srq_cmds)) { - ret = PTR_ERR(ndev->srq_cmds); + nsrq->cmds = nvmet_rdma_alloc_cmds(ndev, srq_size, false); + if (IS_ERR(nsrq->cmds)) { + ret = PTR_ERR(nsrq->cmds); goto out_destroy_srq; } - ndev->srq = srq; - ndev->srq_size = srq_size; + nsrq->srq = srq; + nsrq->ndev = ndev; for (i = 0; i < srq_size; i++) { - ret = nvmet_rdma_post_recv(ndev, &ndev->srq_cmds[i]); + nsrq->cmds[i].nsrq = nsrq; + ret = nvmet_rdma_post_recv(ndev, &nsrq->cmds[i]); if (ret) goto out_free_cmds; } - return 0; + return nsrq; out_free_cmds: - nvmet_rdma_free_cmds(ndev, ndev->srq_cmds, ndev->srq_size, false); + nvmet_rdma_free_cmds(ndev, nsrq->cmds, srq_size, false); out_destroy_srq: ib_destroy_srq(srq); +out_free: + kfree(nsrq); + return ERR_PTR(ret); +} + +static int nvmet_rdma_init_srqs(struct nvmet_rdma_device *ndev) +{ + int i, ret; + + if (!ndev->device->attrs.max_srq_wr || !ndev->device->attrs.max_srq) { + /* + * If SRQs aren't supported we just go ahead and use normal + * non-shared receive queues. + */ + pr_info("SRQ requested but not supported.\n"); + return 0; + } + + ndev->srq_size = min(ndev->device->attrs.max_srq_wr, + nvmet_rdma_srq_size); + ndev->srq_count = min(ndev->device->num_comp_vectors, + ndev->device->attrs.max_srq); + + ndev->srqs = kcalloc(ndev->srq_count, sizeof(*ndev->srqs), GFP_KERNEL); + if (!ndev->srqs) + return -ENOMEM; + + for (i = 0; i < ndev->srq_count; i++) { + ndev->srqs[i] = nvmet_rdma_init_srq(ndev); + if (IS_ERR(ndev->srqs[i])) { + ret = PTR_ERR(ndev->srqs[i]); + goto err_srq; + } + } + + return 0; + +err_srq: + while (--i >= 0) + nvmet_rdma_destroy_srq(ndev->srqs[i]); + kfree(ndev->srqs); return ret; } @@ -910,7 +1179,7 @@ static void nvmet_rdma_free_dev(struct kref *ref) list_del(&ndev->entry); mutex_unlock(&device_list_mutex); - nvmet_rdma_destroy_srq(ndev); + nvmet_rdma_destroy_srqs(ndev); ib_dealloc_pd(ndev->pd); kfree(ndev); @@ -957,7 +1226,7 @@ nvmet_rdma_find_get_device(struct rdma_cm_id *cm_id) goto out_free_dev; if (nvmet_rdma_use_srq) { - ret = nvmet_rdma_init_srq(ndev); + ret = nvmet_rdma_init_srqs(ndev); if (ret) goto out_free_pd; } @@ -981,14 +1250,7 @@ static int nvmet_rdma_create_queue_ib(struct nvmet_rdma_queue *queue) { struct ib_qp_init_attr qp_attr; struct nvmet_rdma_device *ndev = queue->dev; - int comp_vector, nr_cqe, ret, i, factor; - - /* - * Spread the io queues across completion vectors, - * but still keep all admin queues on vector 0. - */ - comp_vector = !queue->host_qid ? 0 : - queue->idx % ndev->device->num_comp_vectors; + int nr_cqe, ret, i, factor; /* * Reserve CQ slots for RECV + RDMA_READ/RDMA_WRITE + RDMA_SEND. @@ -996,7 +1258,7 @@ static int nvmet_rdma_create_queue_ib(struct nvmet_rdma_queue *queue) nr_cqe = queue->recv_queue_size + 2 * queue->send_queue_size; queue->cq = ib_alloc_cq(ndev->device, queue, - nr_cqe + 1, comp_vector, + nr_cqe + 1, queue->comp_vector, IB_POLL_WORKQUEUE); if (IS_ERR(queue->cq)) { ret = PTR_ERR(queue->cq); @@ -1020,14 +1282,17 @@ static int nvmet_rdma_create_queue_ib(struct nvmet_rdma_queue *queue) qp_attr.cap.max_send_sge = max(ndev->device->attrs.max_sge_rd, ndev->device->attrs.max_send_sge); - if (ndev->srq) { - qp_attr.srq = ndev->srq; + if (queue->nsrq) { + qp_attr.srq = queue->nsrq->srq; } else { /* +1 for drain */ qp_attr.cap.max_recv_wr = 1 + queue->recv_queue_size; qp_attr.cap.max_recv_sge = 1 + ndev->inline_page_count; } + if (queue->port->pi_enable && queue->host_qid) + qp_attr.create_flags |= IB_QP_CREATE_INTEGRITY_EN; + ret = rdma_create_qp(queue->cm_id, ndev->pd, &qp_attr); if (ret) { pr_err("failed to create_qp ret= %d\n", ret); @@ -1041,7 +1306,7 @@ static int nvmet_rdma_create_queue_ib(struct nvmet_rdma_queue *queue) __func__, queue->cq->cqe, qp_attr.cap.max_send_sge, qp_attr.cap.max_send_wr, queue->cm_id); - if (!ndev->srq) { + if (!queue->nsrq) { for (i = 0; i < queue->recv_queue_size; i++) { queue->cmds[i].queue = queue; ret = nvmet_rdma_post_recv(ndev, &queue->cmds[i]); @@ -1076,7 +1341,7 @@ static void nvmet_rdma_free_queue(struct nvmet_rdma_queue *queue) nvmet_sq_destroy(&queue->nvme_sq); nvmet_rdma_destroy_queue_ib(queue); - if (!queue->dev->srq) { + if (!queue->nsrq) { nvmet_rdma_free_cmds(queue->dev, queue->cmds, queue->recv_queue_size, !queue->host_qid); @@ -1146,6 +1411,7 @@ nvmet_rdma_alloc_queue(struct nvmet_rdma_device *ndev, struct rdma_cm_id *cm_id, struct rdma_cm_event *event) { + struct nvmet_rdma_port *port = cm_id->context; struct nvmet_rdma_queue *queue; int ret; @@ -1172,6 +1438,7 @@ nvmet_rdma_alloc_queue(struct nvmet_rdma_device *ndev, INIT_WORK(&queue->release_work, nvmet_rdma_release_queue_work); queue->dev = ndev; queue->cm_id = cm_id; + queue->port = port->nport; spin_lock_init(&queue->state_lock); queue->state = NVMET_RDMA_Q_CONNECTING; @@ -1188,13 +1455,23 @@ nvmet_rdma_alloc_queue(struct nvmet_rdma_device *ndev, goto out_destroy_sq; } + /* + * Spread the io queues across completion vectors, + * but still keep all admin queues on vector 0. + */ + queue->comp_vector = !queue->host_qid ? 0 : + queue->idx % ndev->device->num_comp_vectors; + + ret = nvmet_rdma_alloc_rsps(queue); if (ret) { ret = NVME_RDMA_CM_NO_RSC; goto out_ida_remove; } - if (!ndev->srq) { + if (ndev->srqs) { + queue->nsrq = ndev->srqs[queue->comp_vector % ndev->srq_count]; + } else { queue->cmds = nvmet_rdma_alloc_cmds(ndev, queue->recv_queue_size, !queue->host_qid); @@ -1215,7 +1492,7 @@ nvmet_rdma_alloc_queue(struct nvmet_rdma_device *ndev, return queue; out_free_cmds: - if (!ndev->srq) { + if (!queue->nsrq) { nvmet_rdma_free_cmds(queue->dev, queue->cmds, queue->recv_queue_size, !queue->host_qid); @@ -1241,6 +1518,10 @@ static void nvmet_rdma_qp_event(struct ib_event *event, void *priv) case IB_EVENT_COMM_EST: rdma_notify(queue->cm_id, event->event); break; + case IB_EVENT_QP_LAST_WQE_REACHED: + pr_debug("received last WQE reached event for queue=0x%p\n", + queue); + break; default: pr_err("received IB QP event: %s (%d)\n", ib_event_msg(event->event), event->event); @@ -1275,7 +1556,6 @@ static int nvmet_rdma_cm_accept(struct rdma_cm_id *cm_id, static int nvmet_rdma_queue_connect(struct rdma_cm_id *cm_id, struct rdma_cm_event *event) { - struct nvmet_rdma_port *port = cm_id->context; struct nvmet_rdma_device *ndev; struct nvmet_rdma_queue *queue; int ret = -EINVAL; @@ -1291,7 +1571,6 @@ static int nvmet_rdma_queue_connect(struct rdma_cm_id *cm_id, ret = -ENOMEM; goto put_device; } - queue->port = port->nport; if (queue->host_qid == 0) { /* Let inflight controller teardown complete */ @@ -1563,6 +1842,14 @@ static int nvmet_rdma_enable_port(struct nvmet_rdma_port *port) goto out_destroy_id; } + if (port->nport->pi_enable && + !(cm_id->device->attrs.device_cap_flags & + IB_DEVICE_INTEGRITY_HANDOVER)) { + pr_err("T10-PI is not supported for %pISpcs\n", addr); + ret = -EINVAL; + goto out_destroy_id; + } + port->cm_id = cm_id; return 0; @@ -1672,6 +1959,8 @@ static void nvmet_rdma_disc_port_addr(struct nvmet_req *req, static u8 nvmet_rdma_get_mdts(const struct nvmet_ctrl *ctrl) { + if (ctrl->pi_support) + return NVMET_RDMA_MAX_METADATA_MDTS; return NVMET_RDMA_MAX_MDTS; } @@ -1680,6 +1969,7 @@ static const struct nvmet_fabrics_ops nvmet_rdma_ops = { .type = NVMF_TRTYPE_RDMA, .msdbd = 1, .has_keyed_sgls = 1, + .metadata_support = 1, .add_port = nvmet_rdma_add_port, .remove_port = nvmet_rdma_remove_port, .queue_response = nvmet_rdma_queue_response, diff --git a/drivers/nvme/target/tcp.c b/drivers/nvme/target/tcp.c index f0da04e960f4..1669177cd26c 100644 --- a/drivers/nvme/target/tcp.c +++ b/drivers/nvme/target/tcp.c @@ -325,6 +325,14 @@ static void nvmet_tcp_fatal_error(struct nvmet_tcp_queue *queue) kernel_sock_shutdown(queue->sock, SHUT_RDWR); } +static void nvmet_tcp_socket_error(struct nvmet_tcp_queue *queue, int status) +{ + if (status == -EPIPE || status == -ECONNRESET) + kernel_sock_shutdown(queue->sock, SHUT_RDWR); + else + nvmet_tcp_fatal_error(queue); +} + static int nvmet_tcp_map_data(struct nvmet_tcp_cmd *cmd) { struct nvme_sgl_desc *sgl = &cmd->req.cmd->common.dptr.sgl; @@ -510,7 +518,7 @@ static int nvmet_try_send_data_pdu(struct nvmet_tcp_cmd *cmd) ret = kernel_sendpage(cmd->queue->sock, virt_to_page(cmd->data_pdu), offset_in_page(cmd->data_pdu) + cmd->offset, - left, MSG_DONTWAIT | MSG_MORE); + left, MSG_DONTWAIT | MSG_MORE | MSG_SENDPAGE_NOTLAST); if (ret <= 0) return ret; @@ -538,7 +546,7 @@ static int nvmet_try_send_data(struct nvmet_tcp_cmd *cmd, bool last_in_batch) if ((!last_in_batch && cmd->queue->send_list_len) || cmd->wbytes_done + left < cmd->req.transfer_len || queue->data_digest || !queue->nvme_sq.sqhd_disabled) - flags |= MSG_MORE; + flags |= MSG_MORE | MSG_SENDPAGE_NOTLAST; ret = kernel_sendpage(cmd->queue->sock, page, cmd->offset, left, flags); @@ -585,7 +593,7 @@ static int nvmet_try_send_response(struct nvmet_tcp_cmd *cmd, int ret; if (!last_in_batch && cmd->queue->send_list_len) - flags |= MSG_MORE; + flags |= MSG_MORE | MSG_SENDPAGE_NOTLAST; else flags |= MSG_EOR; @@ -614,7 +622,7 @@ static int nvmet_try_send_r2t(struct nvmet_tcp_cmd *cmd, bool last_in_batch) int ret; if (!last_in_batch && cmd->queue->send_list_len) - flags |= MSG_MORE; + flags |= MSG_MORE | MSG_SENDPAGE_NOTLAST; else flags |= MSG_EOR; @@ -644,6 +652,8 @@ static int nvmet_try_send_ddgst(struct nvmet_tcp_cmd *cmd, bool last_in_batch) if (!last_in_batch && cmd->queue->send_list_len) msg.msg_flags |= MSG_MORE; + else + msg.msg_flags |= MSG_EOR; ret = kernel_sendmsg(queue->sock, &msg, &iov, 1, iov.iov_len); if (unlikely(ret <= 0)) @@ -716,11 +726,15 @@ static int nvmet_tcp_try_send(struct nvmet_tcp_queue *queue, for (i = 0; i < budget; i++) { ret = nvmet_tcp_try_send_one(queue, i == budget - 1); - if (ret <= 0) + if (unlikely(ret < 0)) { + nvmet_tcp_socket_error(queue, ret); + goto done; + } else if (ret == 0) { break; + } (*sends)++; } - +done: return ret; } @@ -1157,11 +1171,15 @@ static int nvmet_tcp_try_recv(struct nvmet_tcp_queue *queue, for (i = 0; i < budget; i++) { ret = nvmet_tcp_try_recv_one(queue); - if (ret <= 0) + if (unlikely(ret < 0)) { + nvmet_tcp_socket_error(queue, ret); + goto done; + } else if (ret == 0) { break; + } (*recvs)++; } - +done: return ret; } @@ -1186,27 +1204,16 @@ static void nvmet_tcp_io_work(struct work_struct *w) pending = false; ret = nvmet_tcp_try_recv(queue, NVMET_TCP_RECV_BUDGET, &ops); - if (ret > 0) { + if (ret > 0) pending = true; - } else if (ret < 0) { - if (ret == -EPIPE || ret == -ECONNRESET) - kernel_sock_shutdown(queue->sock, SHUT_RDWR); - else - nvmet_tcp_fatal_error(queue); + else if (ret < 0) return; - } ret = nvmet_tcp_try_send(queue, NVMET_TCP_SEND_BUDGET, &ops); - if (ret > 0) { - /* transmitted message/data */ + if (ret > 0) pending = true; - } else if (ret < 0) { - if (ret == -EPIPE || ret == -ECONNRESET) - kernel_sock_shutdown(queue->sock, SHUT_RDWR); - else - nvmet_tcp_fatal_error(queue); + else if (ret < 0) return; - } } while (pending && ops < NVMET_TCP_IO_WORK_BUDGET); @@ -1429,7 +1436,6 @@ static int nvmet_tcp_set_queue_sock(struct nvmet_tcp_queue *queue) { struct socket *sock = queue->sock; struct inet_sock *inet = inet_sk(sock->sk); - struct linger sol = { .l_onoff = 1, .l_linger = 0 }; int ret; ret = kernel_getsockname(sock, @@ -1447,27 +1453,14 @@ static int nvmet_tcp_set_queue_sock(struct nvmet_tcp_queue *queue) * close. This is done to prevent stale data from being sent should * the network connection be restored before TCP times out. */ - ret = kernel_setsockopt(sock, SOL_SOCKET, SO_LINGER, - (char *)&sol, sizeof(sol)); - if (ret) - return ret; + sock_no_linger(sock->sk); - if (so_priority > 0) { - ret = kernel_setsockopt(sock, SOL_SOCKET, SO_PRIORITY, - (char *)&so_priority, sizeof(so_priority)); - if (ret) - return ret; - } + if (so_priority > 0) + sock_set_priority(sock->sk, so_priority); /* Set socket type of service */ - if (inet->rcv_tos > 0) { - int tos = inet->rcv_tos; - - ret = kernel_setsockopt(sock, SOL_IP, IP_TOS, - (char *)&tos, sizeof(tos)); - if (ret) - return ret; - } + if (inet->rcv_tos > 0) + ip_sock_set_tos(sock->sk, inet->rcv_tos); write_lock_bh(&sock->sk->sk_callback_lock); sock->sk->sk_user_data = queue; @@ -1588,7 +1581,7 @@ static int nvmet_tcp_add_port(struct nvmet_port *nport) { struct nvmet_tcp_port *port; __kernel_sa_family_t af; - int opt, ret; + int ret; port = kzalloc(sizeof(*port), GFP_KERNEL); if (!port) @@ -1632,30 +1625,10 @@ static int nvmet_tcp_add_port(struct nvmet_port *nport) port->sock->sk->sk_user_data = port; port->data_ready = port->sock->sk->sk_data_ready; port->sock->sk->sk_data_ready = nvmet_tcp_listen_data_ready; - - opt = 1; - ret = kernel_setsockopt(port->sock, IPPROTO_TCP, - TCP_NODELAY, (char *)&opt, sizeof(opt)); - if (ret) { - pr_err("failed to set TCP_NODELAY sock opt %d\n", ret); - goto err_sock; - } - - ret = kernel_setsockopt(port->sock, SOL_SOCKET, SO_REUSEADDR, - (char *)&opt, sizeof(opt)); - if (ret) { - pr_err("failed to set SO_REUSEADDR sock opt %d\n", ret); - goto err_sock; - } - - if (so_priority > 0) { - ret = kernel_setsockopt(port->sock, SOL_SOCKET, SO_PRIORITY, - (char *)&so_priority, sizeof(so_priority)); - if (ret) { - pr_err("failed to set SO_PRIORITY sock opt %d\n", ret); - goto err_sock; - } - } + sock_set_reuseaddr(port->sock->sk); + tcp_sock_set_nodelay(port->sock->sk); + if (so_priority > 0) + sock_set_priority(port->sock->sk, so_priority); ret = kernel_bind(port->sock, (struct sockaddr *)&port->addr, sizeof(port->addr)); diff --git a/drivers/nvme/target/trace.h b/drivers/nvme/target/trace.h index e645caa882dd..0458046d6501 100644 --- a/drivers/nvme/target/trace.h +++ b/drivers/nvme/target/trace.h @@ -130,6 +130,34 @@ TRACE_EVENT(nvmet_req_complete, ); +#define aer_name(aer) { aer, #aer } + +TRACE_EVENT(nvmet_async_event, + TP_PROTO(struct nvmet_ctrl *ctrl, __le32 result), + TP_ARGS(ctrl, result), + TP_STRUCT__entry( + __field(int, ctrl_id) + __field(u32, result) + ), + TP_fast_assign( + __entry->ctrl_id = ctrl->cntlid; + __entry->result = (le32_to_cpu(result) & 0xff00) >> 8; + ), + TP_printk("nvmet%d: NVME_AEN=%#08x [%s]", + __entry->ctrl_id, __entry->result, + __print_symbolic(__entry->result, + aer_name(NVME_AER_NOTICE_NS_CHANGED), + aer_name(NVME_AER_NOTICE_ANA), + aer_name(NVME_AER_NOTICE_FW_ACT_STARTING), + aer_name(NVME_AER_NOTICE_DISC_CHANGED), + aer_name(NVME_AER_ERROR), + aer_name(NVME_AER_SMART), + aer_name(NVME_AER_CSS), + aer_name(NVME_AER_VS)) + ) +); +#undef aer_name + #endif /* _TRACE_NVMET_H */ #undef TRACE_INCLUDE_PATH |