From 955b1b5a00ba694159a7d3763412597f707c294d Mon Sep 17 00:00:00 2001 From: Minwoo Im Date: Wed, 20 Dec 2017 16:30:50 +0900 Subject: nvme-pci: move use_sgl initialization to nvme_init_iod() A flag "use_sgl" of "struct nvme_iod" has been used in nvme_init_iod() without being set to any value. It seems like "use_sgl" has been set in either nvme_pci_setup_prps() or nvme_pci_setup_sgls() which occur later than nvme_init_iod(). Make "iod->use_sgl" being set in a proper place, nvme_init_iod(). Also move nvme_pci_use_sgls() up above nvme_init_iod() to make it possible to be called by nvme_init_iod(). Signed-off-by: Minwoo Im Reviewed-by: Sagi Grimberg Signed-off-by: Christoph Hellwig --- drivers/nvme/host/pci.c | 42 ++++++++++++++++++++---------------------- 1 file changed, 20 insertions(+), 22 deletions(-) diff --git a/drivers/nvme/host/pci.c b/drivers/nvme/host/pci.c index f5800c3c9082..d53550e612bc 100644 --- a/drivers/nvme/host/pci.c +++ b/drivers/nvme/host/pci.c @@ -448,12 +448,31 @@ static void **nvme_pci_iod_list(struct request *req) return (void **)(iod->sg + blk_rq_nr_phys_segments(req)); } +static inline bool nvme_pci_use_sgls(struct nvme_dev *dev, struct request *req) +{ + struct nvme_iod *iod = blk_mq_rq_to_pdu(req); + unsigned int avg_seg_size; + + avg_seg_size = DIV_ROUND_UP(blk_rq_payload_bytes(req), + blk_rq_nr_phys_segments(req)); + + if (!(dev->ctrl.sgls & ((1 << 0) | (1 << 1)))) + return false; + if (!iod->nvmeq->qid) + return false; + if (!sgl_threshold || avg_seg_size < sgl_threshold) + return false; + return true; +} + static blk_status_t nvme_init_iod(struct request *rq, struct nvme_dev *dev) { struct nvme_iod *iod = blk_mq_rq_to_pdu(rq); int nseg = blk_rq_nr_phys_segments(rq); unsigned int size = blk_rq_payload_bytes(rq); + iod->use_sgl = nvme_pci_use_sgls(dev, rq); + if (nseg > NVME_INT_PAGES || size > NVME_INT_BYTES(dev)) { size_t alloc_size = nvme_pci_iod_alloc_size(dev, size, nseg, iod->use_sgl); @@ -604,8 +623,6 @@ static blk_status_t nvme_pci_setup_prps(struct nvme_dev *dev, dma_addr_t prp_dma; int nprps, i; - iod->use_sgl = false; - length -= (page_size - offset); if (length <= 0) { iod->first_dma = 0; @@ -715,8 +732,6 @@ static blk_status_t nvme_pci_setup_sgls(struct nvme_dev *dev, int entries = iod->nents, i = 0; dma_addr_t sgl_dma; - iod->use_sgl = true; - /* setting the transfer type as SGL */ cmd->flags = NVME_CMD_SGL_METABUF; @@ -770,23 +785,6 @@ static blk_status_t nvme_pci_setup_sgls(struct nvme_dev *dev, return BLK_STS_OK; } -static inline bool nvme_pci_use_sgls(struct nvme_dev *dev, struct request *req) -{ - struct nvme_iod *iod = blk_mq_rq_to_pdu(req); - unsigned int avg_seg_size; - - avg_seg_size = DIV_ROUND_UP(blk_rq_payload_bytes(req), - blk_rq_nr_phys_segments(req)); - - if (!(dev->ctrl.sgls & ((1 << 0) | (1 << 1)))) - return false; - if (!iod->nvmeq->qid) - return false; - if (!sgl_threshold || avg_seg_size < sgl_threshold) - return false; - return true; -} - static blk_status_t nvme_map_data(struct nvme_dev *dev, struct request *req, struct nvme_command *cmnd) { @@ -806,7 +804,7 @@ static blk_status_t nvme_map_data(struct nvme_dev *dev, struct request *req, DMA_ATTR_NO_WARN)) goto out; - if (nvme_pci_use_sgls(dev, req)) + if (iod->use_sgl) ret = nvme_pci_setup_sgls(dev, req, &cmnd->rw); else ret = nvme_pci_setup_prps(dev, req, &cmnd->rw); -- cgit v1.2.3 From cee160fd34b459ace029653436319557a643795a Mon Sep 17 00:00:00 2001 From: Jeff Lien Date: Tue, 19 Dec 2017 13:24:15 -0600 Subject: nvme: fix sector units when going between formats If you format a device with a 4k sector size back to 512 bytes, the queue limit values for physical block size and minimum IO size were not getting updated; only the logical block size was being updated. This patch adds code to update the physical block and IO minimum sizes. Signed-off-by: Jeff Lien Reviewed-by: Martin K. Petersen Signed-off-by: Christoph Hellwig --- drivers/nvme/host/core.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c index 1e46e60b8f10..961d6a4af19c 100644 --- a/drivers/nvme/host/core.c +++ b/drivers/nvme/host/core.c @@ -1335,6 +1335,7 @@ static void nvme_update_disk_info(struct gendisk *disk, struct nvme_ns *ns, struct nvme_id_ns *id) { sector_t capacity = le64_to_cpup(&id->nsze) << (ns->lba_shift - 9); + unsigned short bs = 1 << ns->lba_shift; unsigned stream_alignment = 0; if (ns->ctrl->nr_streams && ns->sws && ns->sgs) @@ -1343,7 +1344,10 @@ static void nvme_update_disk_info(struct gendisk *disk, blk_mq_freeze_queue(disk->queue); blk_integrity_unregister(disk); - blk_queue_logical_block_size(disk->queue, 1 << ns->lba_shift); + blk_queue_logical_block_size(disk->queue, bs); + blk_queue_physical_block_size(disk->queue, bs); + blk_queue_io_min(disk->queue, bs); + if (ns->ms && !ns->ext && (ns->ctrl->ops->flags & NVME_F_METADATA_SUPPORTED)) nvme_init_integrity(disk, ns->ms, ns->pi_type); -- cgit v1.2.3 From d5bf4b7f437c250821d40c3e32158729e6b484ce Mon Sep 17 00:00:00 2001 From: Sagi Grimberg Date: Thu, 21 Dec 2017 14:54:15 +0200 Subject: nvme-rdma: fix concurrent reset and reconnect Now ctrl state machine allows to transition from RESETTING to RECONNECTING. In nvme-rdma when we receive a rdma cm DISONNECTED event, we trigger nvme_rdma_error_recovery. This happens also when we execute a controller reset, issue a cm diconnect request and receive a cm disconnect reply, as a result, the reset work and the error recovery work can run concurrently. Until now the state machine prevented from the error recovery work from running as a result of a controller reset (RESETTING -> RECONNECTING was not allowed). To fix this, we adopt the FC state machine approach, we always transition from LIVE to RESETTING and only then to RECONNECTING. We do this both for the error recovery work and the controller reset work: 1. transition to RESETTING 2. teardown the controller association 3. transition to RECONNECTING This will restore the protection against reset work and error recovery work from concurrently running together. Fixes: 3cec7f9de448 ("nvme: allow controller RESETTING to RECONNECTING transition") Signed-off-by: Sagi Grimberg Signed-off-by: Christoph Hellwig --- drivers/nvme/host/rdma.c | 14 +++++++++++++- 1 file changed, 13 insertions(+), 1 deletion(-) diff --git a/drivers/nvme/host/rdma.c b/drivers/nvme/host/rdma.c index 37af56596be6..2a0bba7f50cf 100644 --- a/drivers/nvme/host/rdma.c +++ b/drivers/nvme/host/rdma.c @@ -974,12 +974,18 @@ static void nvme_rdma_error_recovery_work(struct work_struct *work) blk_mq_unquiesce_queue(ctrl->ctrl.admin_q); nvme_start_queues(&ctrl->ctrl); + if (!nvme_change_ctrl_state(&ctrl->ctrl, NVME_CTRL_RECONNECTING)) { + /* state change failure should never happen */ + WARN_ON_ONCE(1); + return; + } + nvme_rdma_reconnect_or_remove(ctrl); } static void nvme_rdma_error_recovery(struct nvme_rdma_ctrl *ctrl) { - if (!nvme_change_ctrl_state(&ctrl->ctrl, NVME_CTRL_RECONNECTING)) + if (!nvme_change_ctrl_state(&ctrl->ctrl, NVME_CTRL_RESETTING)) return; queue_work(nvme_wq, &ctrl->err_work); @@ -1753,6 +1759,12 @@ static void nvme_rdma_reset_ctrl_work(struct work_struct *work) nvme_stop_ctrl(&ctrl->ctrl); nvme_rdma_shutdown_ctrl(ctrl, false); + if (!nvme_change_ctrl_state(&ctrl->ctrl, NVME_CTRL_RECONNECTING)) { + /* state change failure should never happen */ + WARN_ON_ONCE(1); + return; + } + ret = nvme_rdma_configure_admin_queue(ctrl, false); if (ret) goto out_fail; -- cgit v1.2.3 From 479a322fb729d657d34706ccf8dd12916f36628f Mon Sep 17 00:00:00 2001 From: Sagi Grimberg Date: Thu, 21 Dec 2017 15:07:27 +0200 Subject: nvme-mpath: fix last path removal during traffic In case our last path is removed during traffic, we can end up requeueing the bio(s) but never schedule the actual requeue work as upper layers still have open handles on the mpath device node. Fix this by scheduling requeue work if the namespace being removed is the last path in the ns_head path list. Fixes: 32acab3181c7 ("nvme: implement multipath access to nvme subsystems") Signed-off-by: Sagi Grimberg Signed-off-by: Christoph Hellwig --- drivers/nvme/host/core.c | 1 + drivers/nvme/host/nvme.h | 12 ++++++++++++ 2 files changed, 13 insertions(+) diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c index 961d6a4af19c..839650e0926a 100644 --- a/drivers/nvme/host/core.c +++ b/drivers/nvme/host/core.c @@ -2991,6 +2991,7 @@ static void nvme_ns_remove(struct nvme_ns *ns) mutex_unlock(&ns->ctrl->namespaces_mutex); synchronize_srcu(&ns->head->srcu); + nvme_mpath_check_last_path(ns); nvme_put_ns(ns); } diff --git a/drivers/nvme/host/nvme.h b/drivers/nvme/host/nvme.h index ea1aa5283e8e..a00eabd06427 100644 --- a/drivers/nvme/host/nvme.h +++ b/drivers/nvme/host/nvme.h @@ -417,6 +417,15 @@ static inline void nvme_mpath_clear_current_path(struct nvme_ns *ns) rcu_assign_pointer(head->current_path, NULL); } struct nvme_ns *nvme_find_path(struct nvme_ns_head *head); + +static inline void nvme_mpath_check_last_path(struct nvme_ns *ns) +{ + struct nvme_ns_head *head = ns->head; + + if (head->disk && list_empty(&head->list)) + kblockd_schedule_work(&head->requeue_work); +} + #else static inline void nvme_failover_req(struct request *req) { @@ -448,6 +457,9 @@ static inline void nvme_mpath_remove_disk_links(struct nvme_ns *ns) static inline void nvme_mpath_clear_current_path(struct nvme_ns *ns) { } +static inline void nvme_mpath_check_last_path(struct nvme_ns *ns) +{ +} #endif /* CONFIG_NVME_MULTIPATH */ #ifdef CONFIG_NVM -- cgit v1.2.3 From 254beb84faccbe2f4eda0b51924857bdfb679969 Mon Sep 17 00:00:00 2001 From: James Smart Date: Thu, 21 Dec 2017 14:15:47 -0800 Subject: nvme-fcloop: avoid possible uninitialized variable warning The kbuild test robot send mail of a potential use of an uninitialized variable - "tport" in fcloop_delete_targetport() which then calls __targetport_unreg() which uses the variable. It will never be the case it is uninitialized as the call to __targetport_unreg() only occurs if there is a valid nport pointer. And at the time the nport pointer is assigned, the tport variable is set. Remove the warning by assigning a NULL value initially. Signed-off-by: James Smart Signed-off-by: Christoph Hellwig --- drivers/nvme/target/fcloop.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/nvme/target/fcloop.c b/drivers/nvme/target/fcloop.c index 7b75d9de55ab..6a018a0bd6ce 100644 --- a/drivers/nvme/target/fcloop.c +++ b/drivers/nvme/target/fcloop.c @@ -1085,7 +1085,7 @@ fcloop_delete_target_port(struct device *dev, struct device_attribute *attr, const char *buf, size_t count) { struct fcloop_nport *nport = NULL, *tmpport; - struct fcloop_tport *tport; + struct fcloop_tport *tport = NULL; u64 nodename, portname; unsigned long flags; int ret; -- cgit v1.2.3