From d858e5f04e58a42a6e0c8ec74ea15e3ea4bb45d0 Mon Sep 17 00:00:00 2001 From: Sagi Grimberg Date: Mon, 24 Apr 2017 10:58:29 +0300 Subject: nvme: move queue_count to the nvme_ctrl All all transports use the queue_count in exactly the same, so move it to the generic struct nvme_ctrl. In the future it will also be maintained by the core. Reviewed-by: Christoph Hellwig Reviewed-By: James Smart Reviewed-by: Johannes Thumshirn Reviewed-by: Max Gurtovoy Signed-off-by: Sagi Grimberg --- drivers/nvme/host/rdma.c | 39 +++++++++++++++++++-------------------- 1 file changed, 19 insertions(+), 20 deletions(-) (limited to 'drivers/nvme/host/rdma.c') diff --git a/drivers/nvme/host/rdma.c b/drivers/nvme/host/rdma.c index 6d4119dfbdaa..c4bacad3fe23 100644 --- a/drivers/nvme/host/rdma.c +++ b/drivers/nvme/host/rdma.c @@ -103,7 +103,6 @@ struct nvme_rdma_queue { struct nvme_rdma_ctrl { /* read only in the hot path */ struct nvme_rdma_queue *queues; - u32 queue_count; /* other member variables */ struct blk_mq_tag_set tag_set; @@ -349,7 +348,7 @@ static int nvme_rdma_init_hctx(struct blk_mq_hw_ctx *hctx, void *data, struct nvme_rdma_ctrl *ctrl = data; struct nvme_rdma_queue *queue = &ctrl->queues[hctx_idx + 1]; - BUG_ON(hctx_idx >= ctrl->queue_count); + BUG_ON(hctx_idx >= ctrl->ctrl.queue_count); hctx->driver_data = queue; return 0; @@ -587,7 +586,7 @@ static void nvme_rdma_free_io_queues(struct nvme_rdma_ctrl *ctrl) { int i; - for (i = 1; i < ctrl->queue_count; i++) + for (i = 1; i < ctrl->ctrl.queue_count; i++) nvme_rdma_stop_and_free_queue(&ctrl->queues[i]); } @@ -595,7 +594,7 @@ static int nvme_rdma_connect_io_queues(struct nvme_rdma_ctrl *ctrl) { int i, ret = 0; - for (i = 1; i < ctrl->queue_count; i++) { + for (i = 1; i < ctrl->ctrl.queue_count; i++) { ret = nvmf_connect_io_queue(&ctrl->ctrl, i); if (ret) { dev_info(ctrl->ctrl.device, @@ -623,14 +622,14 @@ static int nvme_rdma_init_io_queues(struct nvme_rdma_ctrl *ctrl) if (ret) return ret; - ctrl->queue_count = nr_io_queues + 1; - if (ctrl->queue_count < 2) + ctrl->ctrl.queue_count = nr_io_queues + 1; + if (ctrl->ctrl.queue_count < 2) return 0; dev_info(ctrl->ctrl.device, "creating %d I/O queues.\n", nr_io_queues); - for (i = 1; i < ctrl->queue_count; i++) { + for (i = 1; i < ctrl->ctrl.queue_count; i++) { ret = nvme_rdma_init_queue(ctrl, i, ctrl->ctrl.opts->queue_size); if (ret) { @@ -705,7 +704,7 @@ static void nvme_rdma_reconnect_ctrl_work(struct work_struct *work) ++ctrl->ctrl.nr_reconnects; - if (ctrl->queue_count > 1) { + if (ctrl->ctrl.queue_count > 1) { nvme_rdma_free_io_queues(ctrl); ret = blk_mq_reinit_tagset(&ctrl->tag_set); @@ -735,7 +734,7 @@ static void nvme_rdma_reconnect_ctrl_work(struct work_struct *work) nvme_start_keep_alive(&ctrl->ctrl); - if (ctrl->queue_count > 1) { + if (ctrl->ctrl.queue_count > 1) { ret = nvme_rdma_init_io_queues(ctrl); if (ret) goto requeue; @@ -749,7 +748,7 @@ static void nvme_rdma_reconnect_ctrl_work(struct work_struct *work) WARN_ON_ONCE(!changed); ctrl->ctrl.nr_reconnects = 0; - if (ctrl->queue_count > 1) { + if (ctrl->ctrl.queue_count > 1) { nvme_queue_scan(&ctrl->ctrl); nvme_queue_async_events(&ctrl->ctrl); } @@ -772,15 +771,15 @@ static void nvme_rdma_error_recovery_work(struct work_struct *work) nvme_stop_keep_alive(&ctrl->ctrl); - for (i = 0; i < ctrl->queue_count; i++) + for (i = 0; i < ctrl->ctrl.queue_count; i++) clear_bit(NVME_RDMA_Q_LIVE, &ctrl->queues[i].flags); - if (ctrl->queue_count > 1) + if (ctrl->ctrl.queue_count > 1) nvme_stop_queues(&ctrl->ctrl); blk_mq_stop_hw_queues(ctrl->ctrl.admin_q); /* We must take care of fastfail/requeue all our inflight requests */ - if (ctrl->queue_count > 1) + if (ctrl->ctrl.queue_count > 1) blk_mq_tagset_busy_iter(&ctrl->tag_set, nvme_cancel_request, &ctrl->ctrl); blk_mq_tagset_busy_iter(&ctrl->admin_tag_set, @@ -1624,7 +1623,7 @@ static void nvme_rdma_shutdown_ctrl(struct nvme_rdma_ctrl *ctrl) cancel_work_sync(&ctrl->err_work); cancel_delayed_work_sync(&ctrl->reconnect_work); - if (ctrl->queue_count > 1) { + if (ctrl->ctrl.queue_count > 1) { nvme_stop_queues(&ctrl->ctrl); blk_mq_tagset_busy_iter(&ctrl->tag_set, nvme_cancel_request, &ctrl->ctrl); @@ -1716,7 +1715,7 @@ static void nvme_rdma_reset_ctrl_work(struct work_struct *work) goto del_dead_ctrl; } - if (ctrl->queue_count > 1) { + if (ctrl->ctrl.queue_count > 1) { ret = blk_mq_reinit_tagset(&ctrl->tag_set); if (ret) goto del_dead_ctrl; @@ -1733,7 +1732,7 @@ static void nvme_rdma_reset_ctrl_work(struct work_struct *work) changed = nvme_change_ctrl_state(&ctrl->ctrl, NVME_CTRL_LIVE); WARN_ON_ONCE(!changed); - if (ctrl->queue_count > 1) { + if (ctrl->ctrl.queue_count > 1) { nvme_start_queues(&ctrl->ctrl); nvme_queue_scan(&ctrl->ctrl); nvme_queue_async_events(&ctrl->ctrl); @@ -1785,7 +1784,7 @@ static int nvme_rdma_create_io_queues(struct nvme_rdma_ctrl *ctrl) ctrl->tag_set.cmd_size = sizeof(struct nvme_rdma_request) + SG_CHUNK_SIZE * sizeof(struct scatterlist); ctrl->tag_set.driver_data = ctrl; - ctrl->tag_set.nr_hw_queues = ctrl->queue_count - 1; + ctrl->tag_set.nr_hw_queues = ctrl->ctrl.queue_count - 1; ctrl->tag_set.timeout = NVME_IO_TIMEOUT; ret = blk_mq_alloc_tag_set(&ctrl->tag_set); @@ -1863,12 +1862,12 @@ static struct nvme_ctrl *nvme_rdma_create_ctrl(struct device *dev, INIT_WORK(&ctrl->delete_work, nvme_rdma_del_ctrl_work); INIT_WORK(&ctrl->ctrl.reset_work, nvme_rdma_reset_ctrl_work); - ctrl->queue_count = opts->nr_io_queues + 1; /* +1 for admin queue */ + ctrl->ctrl.queue_count = opts->nr_io_queues + 1; /* +1 for admin queue */ ctrl->ctrl.sqsize = opts->queue_size - 1; ctrl->ctrl.kato = opts->kato; ret = -ENOMEM; - ctrl->queues = kcalloc(ctrl->queue_count, sizeof(*ctrl->queues), + ctrl->queues = kcalloc(ctrl->ctrl.queue_count, sizeof(*ctrl->queues), GFP_KERNEL); if (!ctrl->queues) goto out_uninit_ctrl; @@ -1925,7 +1924,7 @@ static struct nvme_ctrl *nvme_rdma_create_ctrl(struct device *dev, list_add_tail(&ctrl->list, &nvme_rdma_ctrl_list); mutex_unlock(&nvme_rdma_ctrl_mutex); - if (opts->nr_io_queues) { + if (ctrl->ctrl.queue_count > 1) { nvme_queue_scan(&ctrl->ctrl); nvme_queue_async_events(&ctrl->ctrl); } -- cgit v1.2.3 From 20d0dfe65afd3fb59d14720570a6921eb6bf5c1f Mon Sep 17 00:00:00 2001 From: Sagi Grimberg Date: Tue, 27 Jun 2017 22:16:38 +0300 Subject: nvme: move ctrl cap to struct nvme_ctrl All transports use either a private cache of controller cap or an on-stack copy, move it to the generic struct nvme_ctrl. In the future it will also be maintained by the core. Reviewed-by: Christoph Hellwig Reviewed-by: Johannes Thumshirn Reviewed-by: Max Gurtovoy Signed-off-by: Sagi Grimberg --- drivers/nvme/host/fc.c | 8 +++----- drivers/nvme/host/nvme.h | 1 + drivers/nvme/host/pci.c | 20 +++++++++----------- drivers/nvme/host/rdma.c | 10 +++++----- drivers/nvme/target/loop.c | 7 +++---- 5 files changed, 21 insertions(+), 25 deletions(-) (limited to 'drivers/nvme/host/rdma.c') diff --git a/drivers/nvme/host/fc.c b/drivers/nvme/host/fc.c index 7eb006427caf..2f990d979037 100644 --- a/drivers/nvme/host/fc.c +++ b/drivers/nvme/host/fc.c @@ -152,8 +152,6 @@ struct nvme_fc_ctrl { u64 association_id; - u64 cap; - struct list_head ctrl_list; /* rport->ctrl_list */ struct blk_mq_tag_set admin_tag_set; @@ -2328,7 +2326,7 @@ nvme_fc_create_association(struct nvme_fc_ctrl *ctrl) * prior connection values */ - ret = nvmf_reg_read64(&ctrl->ctrl, NVME_REG_CAP, &ctrl->cap); + ret = nvmf_reg_read64(&ctrl->ctrl, NVME_REG_CAP, &ctrl->ctrl.cap); if (ret) { dev_err(ctrl->ctrl.device, "prop_get NVME_REG_CAP failed\n"); @@ -2336,9 +2334,9 @@ nvme_fc_create_association(struct nvme_fc_ctrl *ctrl) } ctrl->ctrl.sqsize = - min_t(int, NVME_CAP_MQES(ctrl->cap) + 1, ctrl->ctrl.sqsize); + min_t(int, NVME_CAP_MQES(ctrl->ctrl.cap) + 1, ctrl->ctrl.sqsize); - ret = nvme_enable_ctrl(&ctrl->ctrl, ctrl->cap); + ret = nvme_enable_ctrl(&ctrl->ctrl, ctrl->ctrl.cap); if (ret) goto out_disconnect_admin_queue; diff --git a/drivers/nvme/host/nvme.h b/drivers/nvme/host/nvme.h index 6c51d92b7fab..e0b83311d5de 100644 --- a/drivers/nvme/host/nvme.h +++ b/drivers/nvme/host/nvme.h @@ -144,6 +144,7 @@ struct nvme_ctrl { u32 ctrl_config; u32 queue_count; + u64 cap; u32 page_size; u32 max_hw_sectors; u16 oncs; diff --git a/drivers/nvme/host/pci.c b/drivers/nvme/host/pci.c index 6b50c9096fe4..fe3907a082ba 100644 --- a/drivers/nvme/host/pci.c +++ b/drivers/nvme/host/pci.c @@ -1144,8 +1144,7 @@ static void nvme_disable_admin_queue(struct nvme_dev *dev, bool shutdown) if (shutdown) nvme_shutdown_ctrl(&dev->ctrl); else - nvme_disable_ctrl(&dev->ctrl, lo_hi_readq( - dev->bar + NVME_REG_CAP)); + nvme_disable_ctrl(&dev->ctrl, dev->ctrl.cap); spin_lock_irq(&nvmeq->q_lock); nvme_process_cq(nvmeq); @@ -1388,7 +1387,6 @@ static int nvme_configure_admin_queue(struct nvme_dev *dev) { int result; u32 aqa; - u64 cap = lo_hi_readq(dev->bar + NVME_REG_CAP); struct nvme_queue *nvmeq; result = nvme_remap_bar(dev, db_bar_size(dev, 0)); @@ -1396,13 +1394,13 @@ static int nvme_configure_admin_queue(struct nvme_dev *dev) return result; dev->subsystem = readl(dev->bar + NVME_REG_VS) >= NVME_VS(1, 1, 0) ? - NVME_CAP_NSSRC(cap) : 0; + NVME_CAP_NSSRC(dev->ctrl.cap) : 0; if (dev->subsystem && (readl(dev->bar + NVME_REG_CSTS) & NVME_CSTS_NSSRO)) writel(NVME_CSTS_NSSRO, dev->bar + NVME_REG_CSTS); - result = nvme_disable_ctrl(&dev->ctrl, cap); + result = nvme_disable_ctrl(&dev->ctrl, dev->ctrl.cap); if (result < 0) return result; @@ -1421,7 +1419,7 @@ static int nvme_configure_admin_queue(struct nvme_dev *dev) lo_hi_writeq(nvmeq->sq_dma_addr, dev->bar + NVME_REG_ASQ); lo_hi_writeq(nvmeq->cq_dma_addr, dev->bar + NVME_REG_ACQ); - result = nvme_enable_ctrl(&dev->ctrl, cap); + result = nvme_enable_ctrl(&dev->ctrl, dev->ctrl.cap); if (result) return result; @@ -1865,7 +1863,6 @@ static int nvme_dev_add(struct nvme_dev *dev) static int nvme_pci_enable(struct nvme_dev *dev) { - u64 cap; int result = -ENOMEM; struct pci_dev *pdev = to_pci_dev(dev->dev); @@ -1892,10 +1889,11 @@ static int nvme_pci_enable(struct nvme_dev *dev) if (result < 0) return result; - cap = lo_hi_readq(dev->bar + NVME_REG_CAP); + dev->ctrl.cap = lo_hi_readq(dev->bar + NVME_REG_CAP); - dev->q_depth = min_t(int, NVME_CAP_MQES(cap) + 1, NVME_Q_DEPTH); - dev->db_stride = 1 << NVME_CAP_STRIDE(cap); + dev->q_depth = min_t(int, NVME_CAP_MQES(dev->ctrl.cap) + 1, + NVME_Q_DEPTH); + dev->db_stride = 1 << NVME_CAP_STRIDE(dev->ctrl.cap); dev->dbs = dev->bar + 4096; /* @@ -1909,7 +1907,7 @@ static int nvme_pci_enable(struct nvme_dev *dev) dev->q_depth); } else if (pdev->vendor == PCI_VENDOR_ID_SAMSUNG && (pdev->device == 0xa821 || pdev->device == 0xa822) && - NVME_CAP_MQES(cap) == 0) { + NVME_CAP_MQES(dev->ctrl.cap) == 0) { dev->q_depth = 64; dev_err(dev->ctrl.device, "detected PM1725 NVMe controller, " "set queue depth=%u\n", dev->q_depth); diff --git a/drivers/nvme/host/rdma.c b/drivers/nvme/host/rdma.c index c4bacad3fe23..2ab0cdb4d881 100644 --- a/drivers/nvme/host/rdma.c +++ b/drivers/nvme/host/rdma.c @@ -118,7 +118,6 @@ struct nvme_rdma_ctrl { struct blk_mq_tag_set admin_tag_set; struct nvme_rdma_device *device; - u64 cap; u32 max_fr_pages; struct sockaddr_storage addr; @@ -728,7 +727,7 @@ static void nvme_rdma_reconnect_ctrl_work(struct work_struct *work) set_bit(NVME_RDMA_Q_LIVE, &ctrl->queues[0].flags); - ret = nvme_enable_ctrl(&ctrl->ctrl, ctrl->cap); + ret = nvme_enable_ctrl(&ctrl->ctrl, ctrl->ctrl.cap); if (ret) goto requeue; @@ -1573,7 +1572,8 @@ static int nvme_rdma_configure_admin_queue(struct nvme_rdma_ctrl *ctrl) set_bit(NVME_RDMA_Q_LIVE, &ctrl->queues[0].flags); - error = nvmf_reg_read64(&ctrl->ctrl, NVME_REG_CAP, &ctrl->cap); + error = nvmf_reg_read64(&ctrl->ctrl, NVME_REG_CAP, + &ctrl->ctrl.cap); if (error) { dev_err(ctrl->ctrl.device, "prop_get NVME_REG_CAP failed\n"); @@ -1581,9 +1581,9 @@ static int nvme_rdma_configure_admin_queue(struct nvme_rdma_ctrl *ctrl) } ctrl->ctrl.sqsize = - min_t(int, NVME_CAP_MQES(ctrl->cap), ctrl->ctrl.sqsize); + min_t(int, NVME_CAP_MQES(ctrl->ctrl.cap), ctrl->ctrl.sqsize); - error = nvme_enable_ctrl(&ctrl->ctrl, ctrl->cap); + error = nvme_enable_ctrl(&ctrl->ctrl, ctrl->ctrl.cap); if (error) goto out_cleanup_queue; diff --git a/drivers/nvme/target/loop.c b/drivers/nvme/target/loop.c index edf0e2ab19e3..568ed8625696 100644 --- a/drivers/nvme/target/loop.c +++ b/drivers/nvme/target/loop.c @@ -48,7 +48,6 @@ struct nvme_loop_ctrl { struct blk_mq_tag_set admin_tag_set; struct list_head list; - u64 cap; struct blk_mq_tag_set tag_set; struct nvme_loop_iod async_event_iod; struct nvme_ctrl ctrl; @@ -387,7 +386,7 @@ static int nvme_loop_configure_admin_queue(struct nvme_loop_ctrl *ctrl) if (error) goto out_cleanup_queue; - error = nvmf_reg_read64(&ctrl->ctrl, NVME_REG_CAP, &ctrl->cap); + error = nvmf_reg_read64(&ctrl->ctrl, NVME_REG_CAP, &ctrl->ctrl.cap); if (error) { dev_err(ctrl->ctrl.device, "prop_get NVME_REG_CAP failed\n"); @@ -395,9 +394,9 @@ static int nvme_loop_configure_admin_queue(struct nvme_loop_ctrl *ctrl) } ctrl->ctrl.sqsize = - min_t(int, NVME_CAP_MQES(ctrl->cap), ctrl->ctrl.sqsize); + min_t(int, NVME_CAP_MQES(ctrl->ctrl.cap), ctrl->ctrl.sqsize); - error = nvme_enable_ctrl(&ctrl->ctrl, ctrl->cap); + error = nvme_enable_ctrl(&ctrl->ctrl, ctrl->ctrl.cap); if (error) goto out_cleanup_queue; -- cgit v1.2.3 From 4c8b99f6b1ff086c1745fcb0511800fba5c4fb34 Mon Sep 17 00:00:00 2001 From: Sagi Grimberg Date: Thu, 29 Jun 2017 11:10:44 +0300 Subject: nvme-rdma: update tagset nr_hw_queues after reconnecting/resetting We might have more/less queues once we reconnect/reset. For example due to cpu going online/offline or controller constraints. Reviewed-by: Johannes Thumshirn Reviewed-by: Christoph Hellwig Signed-off-by: Sagi Grimberg --- drivers/nvme/host/rdma.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'drivers/nvme/host/rdma.c') diff --git a/drivers/nvme/host/rdma.c b/drivers/nvme/host/rdma.c index 2ab0cdb4d881..cfb22531fc16 100644 --- a/drivers/nvme/host/rdma.c +++ b/drivers/nvme/host/rdma.c @@ -741,6 +741,9 @@ static void nvme_rdma_reconnect_ctrl_work(struct work_struct *work) ret = nvme_rdma_connect_io_queues(ctrl); if (ret) goto requeue; + + blk_mq_update_nr_hw_queues(&ctrl->tag_set, + ctrl->ctrl.queue_count - 1); } changed = nvme_change_ctrl_state(&ctrl->ctrl, NVME_CTRL_LIVE); @@ -1727,6 +1730,9 @@ static void nvme_rdma_reset_ctrl_work(struct work_struct *work) ret = nvme_rdma_connect_io_queues(ctrl); if (ret) goto del_dead_ctrl; + + blk_mq_update_nr_hw_queues(&ctrl->tag_set, + ctrl->ctrl.queue_count - 1); } changed = nvme_change_ctrl_state(&ctrl->ctrl, NVME_CTRL_LIVE); -- cgit v1.2.3 From 5e599d73c1c1816af07f94ddba879499aa39b43c Mon Sep 17 00:00:00 2001 From: Marta Rybczynska Date: Tue, 6 Jun 2017 13:27:21 +0200 Subject: nvme-rdma: remove race conditions from IB signalling This patch improves the way the RDMA IB signalling is done by using atomic operations for the signalling variable. This avoids race conditions on sig_count. The signalling interval changes slightly and is now the largest power of two not larger than queue depth / 2. ilog() usage idea by Bart Van Assche. Signed-off-by: Marta Rybczynska Reviewed-by: Sagi Grimberg Signed-off-by: Christoph Hellwig Cc: stable@vger.kernel.org --- drivers/nvme/host/rdma.c | 20 ++++++++++---------- 1 file changed, 10 insertions(+), 10 deletions(-) (limited to 'drivers/nvme/host/rdma.c') diff --git a/drivers/nvme/host/rdma.c b/drivers/nvme/host/rdma.c index cfb22531fc16..0ab163e6e640 100644 --- a/drivers/nvme/host/rdma.c +++ b/drivers/nvme/host/rdma.c @@ -86,7 +86,7 @@ enum nvme_rdma_queue_flags { struct nvme_rdma_queue { struct nvme_rdma_qe *rsp_ring; - u8 sig_count; + atomic_t sig_count; int queue_size; size_t cmnd_capsule_len; struct nvme_rdma_ctrl *ctrl; @@ -523,6 +523,7 @@ static int nvme_rdma_init_queue(struct nvme_rdma_ctrl *ctrl, queue->cmnd_capsule_len = sizeof(struct nvme_command); queue->queue_size = queue_size; + atomic_set(&queue->sig_count, 0); queue->cm_id = rdma_create_id(&init_net, nvme_rdma_cm_handler, queue, RDMA_PS_TCP, IB_QPT_RC); @@ -1009,17 +1010,16 @@ static void nvme_rdma_send_done(struct ib_cq *cq, struct ib_wc *wc) nvme_rdma_wr_error(cq, wc, "SEND"); } -static inline int nvme_rdma_queue_sig_limit(struct nvme_rdma_queue *queue) +/* + * We want to signal completion at least every queue depth/2. This returns the + * largest power of two that is not above half of (queue size + 1) to optimize + * (avoid divisions). + */ +static inline bool nvme_rdma_queue_sig_limit(struct nvme_rdma_queue *queue) { - int sig_limit; + int limit = 1 << ilog2((queue->queue_size + 1) / 2); - /* - * We signal completion every queue depth/2 and also handle the - * degenerated case of a device with queue_depth=1, where we - * would need to signal every message. - */ - sig_limit = max(queue->queue_size / 2, 1); - return (++queue->sig_count % sig_limit) == 0; + return (atomic_inc_return(&queue->sig_count) & (limit - 1)) == 0; } static int nvme_rdma_post_send(struct nvme_rdma_queue *queue, -- cgit v1.2.3 From fb051339727cd8134dd6ef0305a120e6e265dded Mon Sep 17 00:00:00 2001 From: Sagi Grimberg Date: Sun, 2 Jul 2017 15:33:32 +0300 Subject: nvme-rdma: quiesce/unquiesce admin_q instead of start/stop its hw queues unlike blk_mq_stop_hw_queues and blk_mq_start_stopped_hw_queues quiescing/unquiescing respects the submission path rcu grace. Also make sure to kick the requeue list when appropriate. Reviewed-by: Ming Lei Signed-off-by: Sagi Grimberg --- drivers/nvme/host/rdma.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) (limited to 'drivers/nvme/host/rdma.c') diff --git a/drivers/nvme/host/rdma.c b/drivers/nvme/host/rdma.c index 0ab163e6e640..6f8f9ffcbc00 100644 --- a/drivers/nvme/host/rdma.c +++ b/drivers/nvme/host/rdma.c @@ -779,7 +779,7 @@ static void nvme_rdma_error_recovery_work(struct work_struct *work) if (ctrl->ctrl.queue_count > 1) nvme_stop_queues(&ctrl->ctrl); - blk_mq_stop_hw_queues(ctrl->ctrl.admin_q); + blk_mq_quiesce_queue(ctrl->ctrl.admin_q); /* We must take care of fastfail/requeue all our inflight requests */ if (ctrl->ctrl.queue_count > 1) @@ -792,7 +792,7 @@ static void nvme_rdma_error_recovery_work(struct work_struct *work) * queues are not a live anymore, so restart the queues to fail fast * new IO */ - blk_mq_start_stopped_hw_queues(ctrl->ctrl.admin_q, true); + blk_mq_unquiesce_queue(ctrl->ctrl.admin_q); nvme_start_queues(&ctrl->ctrl); nvme_rdma_reconnect_or_remove(ctrl); @@ -1636,9 +1636,10 @@ static void nvme_rdma_shutdown_ctrl(struct nvme_rdma_ctrl *ctrl) if (test_bit(NVME_RDMA_Q_LIVE, &ctrl->queues[0].flags)) nvme_shutdown_ctrl(&ctrl->ctrl); - blk_mq_stop_hw_queues(ctrl->ctrl.admin_q); + blk_mq_quiesce_queue(ctrl->ctrl.admin_q); blk_mq_tagset_busy_iter(&ctrl->admin_tag_set, nvme_cancel_request, &ctrl->ctrl); + blk_mq_unquiesce_queue(ctrl->ctrl.admin_q); nvme_rdma_destroy_admin_queue(ctrl); } -- cgit v1.2.3 From d09f2b45f346f0a9e5e1b5fcea531b1b393671dc Mon Sep 17 00:00:00 2001 From: Sagi Grimberg Date: Sun, 2 Jul 2017 10:56:43 +0300 Subject: nvme: split nvme_uninit_ctrl into stop and uninit Usually before we teardown the controller we want to: 1. complete/cancel any ctrl inflight works 2. remove ctrl namespaces (only for removal though, resets shouldn't remove any namespaces). but we do not want to destroy the controller device as we might use it for logging during the teardown stage. This patch adds nvme_start_ctrl() which queues inflight controller works (aen, ns scan, queue start and keep-alive if kato is set) and nvme_stop_ctrl() which cancels the works namespace removal is left to the callers to handle. Move nvme_uninit_ctrl after we are done with the controller device. Reviewed-by: Keith Busch Reviewed-by: Christoph Hellwig Signed-off-by: Sagi Grimberg --- drivers/nvme/host/core.c | 21 +++++++++++++++++++-- drivers/nvme/host/fc.c | 16 ++++------------ drivers/nvme/host/nvme.h | 2 ++ drivers/nvme/host/pci.c | 16 ++++------------ drivers/nvme/host/rdma.c | 29 ++++++++--------------------- drivers/nvme/target/loop.c | 19 ++++++------------- 6 files changed, 43 insertions(+), 60 deletions(-) (limited to 'drivers/nvme/host/rdma.c') diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c index 48cafaa6fbc5..cb96f4a7ae3a 100644 --- a/drivers/nvme/host/core.c +++ b/drivers/nvme/host/core.c @@ -2591,12 +2591,29 @@ static void nvme_release_instance(struct nvme_ctrl *ctrl) spin_unlock(&dev_list_lock); } -void nvme_uninit_ctrl(struct nvme_ctrl *ctrl) +void nvme_stop_ctrl(struct nvme_ctrl *ctrl) { + nvme_stop_keep_alive(ctrl); flush_work(&ctrl->async_event_work); flush_work(&ctrl->scan_work); - nvme_remove_namespaces(ctrl); +} +EXPORT_SYMBOL_GPL(nvme_stop_ctrl); + +void nvme_start_ctrl(struct nvme_ctrl *ctrl) +{ + if (ctrl->kato) + nvme_start_keep_alive(ctrl); + + if (ctrl->queue_count > 1) { + nvme_queue_scan(ctrl); + nvme_queue_async_events(ctrl); + nvme_start_queues(ctrl); + } +} +EXPORT_SYMBOL_GPL(nvme_start_ctrl); +void nvme_uninit_ctrl(struct nvme_ctrl *ctrl) +{ device_destroy(nvme_class, MKDEV(nvme_char_major, ctrl->instance)); spin_lock(&dev_list_lock); diff --git a/drivers/nvme/host/fc.c b/drivers/nvme/host/fc.c index 8d55e7827932..d666ada39a9b 100644 --- a/drivers/nvme/host/fc.c +++ b/drivers/nvme/host/fc.c @@ -2232,7 +2232,6 @@ nvme_fc_create_io_queues(struct nvme_fc_ctrl *ctrl) out_delete_hw_queues: nvme_fc_delete_hw_io_queues(ctrl); out_cleanup_blk_queue: - nvme_stop_keep_alive(&ctrl->ctrl); blk_cleanup_queue(ctrl->ctrl.connect_q); out_free_tag_set: blk_mq_free_tag_set(&ctrl->tag_set); @@ -2366,8 +2365,6 @@ nvme_fc_create_association(struct nvme_fc_ctrl *ctrl) goto out_disconnect_admin_queue; } - nvme_start_keep_alive(&ctrl->ctrl); - /* FC-NVME supports normal SGL Data Block Descriptors */ if (opts->queue_size > ctrl->ctrl.maxcmd) { @@ -2401,17 +2398,12 @@ nvme_fc_create_association(struct nvme_fc_ctrl *ctrl) ctrl->ctrl.nr_reconnects = 0; - if (ctrl->ctrl.queue_count > 1) { - nvme_start_queues(&ctrl->ctrl); - nvme_queue_scan(&ctrl->ctrl); - nvme_queue_async_events(&ctrl->ctrl); - } + nvme_start_ctrl(&ctrl->ctrl); return 0; /* Success */ out_term_aen_ops: nvme_fc_term_aen_ops(ctrl); - nvme_stop_keep_alive(&ctrl->ctrl); out_disconnect_admin_queue: /* send a Disconnect(association) LS to fc-nvme target */ nvme_fc_xmt_disconnect_assoc(ctrl); @@ -2434,8 +2426,6 @@ nvme_fc_delete_association(struct nvme_fc_ctrl *ctrl) { unsigned long flags; - nvme_stop_keep_alive(&ctrl->ctrl); - spin_lock_irqsave(&ctrl->lock, flags); ctrl->flags |= FCCTRL_TERMIO; ctrl->iocnt = 0; @@ -2517,7 +2507,8 @@ nvme_fc_delete_ctrl_work(struct work_struct *work) cancel_work_sync(&ctrl->ctrl.reset_work); cancel_delayed_work_sync(&ctrl->connect_work); - + nvme_stop_ctrl(&ctrl->ctrl); + nvme_remove_namespaces(&ctrl->ctrl); /* * kill the association on the link side. this will block * waiting for io to terminate @@ -2612,6 +2603,7 @@ nvme_fc_reset_ctrl_work(struct work_struct *work) container_of(work, struct nvme_fc_ctrl, ctrl.reset_work); int ret; + nvme_stop_ctrl(&ctrl->ctrl); /* will block will waiting for io to terminate */ nvme_fc_delete_association(ctrl); diff --git a/drivers/nvme/host/nvme.h b/drivers/nvme/host/nvme.h index e0b83311d5de..8f2a168ddc01 100644 --- a/drivers/nvme/host/nvme.h +++ b/drivers/nvme/host/nvme.h @@ -280,6 +280,8 @@ int nvme_shutdown_ctrl(struct nvme_ctrl *ctrl); int nvme_init_ctrl(struct nvme_ctrl *ctrl, struct device *dev, const struct nvme_ctrl_ops *ops, unsigned long quirks); 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); diff --git a/drivers/nvme/host/pci.c b/drivers/nvme/host/pci.c index d9c0010a9bbc..882ed3677117 100644 --- a/drivers/nvme/host/pci.c +++ b/drivers/nvme/host/pci.c @@ -2134,15 +2134,6 @@ static void nvme_reset_work(struct work_struct *work) if (result) goto out; - /* - * A controller that can not execute IO typically requires user - * intervention to correct. For such degraded controllers, the driver - * should not submit commands the user did not request, so skip - * registering for asynchronous event notification on this condition. - */ - if (dev->online_queues > 1) - nvme_queue_async_events(&dev->ctrl); - /* * Keep the controller around but remove all namespaces if we don't have * any working I/O queue. @@ -2163,8 +2154,7 @@ static void nvme_reset_work(struct work_struct *work) goto out; } - if (dev->online_queues > 1) - nvme_queue_scan(&dev->ctrl); + nvme_start_ctrl(&dev->ctrl); return; out: @@ -2341,11 +2331,13 @@ static void nvme_remove(struct pci_dev *pdev) } flush_work(&dev->ctrl.reset_work); - nvme_uninit_ctrl(&dev->ctrl); + nvme_stop_ctrl(&dev->ctrl); + nvme_remove_namespaces(&dev->ctrl); nvme_dev_disable(dev, true); nvme_free_host_mem(dev); nvme_dev_remove_admin(dev); nvme_free_queues(dev, 0); + nvme_uninit_ctrl(&dev->ctrl); nvme_release_prp_pools(dev); nvme_dev_unmap(dev); nvme_put_ctrl(&dev->ctrl); diff --git a/drivers/nvme/host/rdma.c b/drivers/nvme/host/rdma.c index 6f8f9ffcbc00..ccdbd99d5a1c 100644 --- a/drivers/nvme/host/rdma.c +++ b/drivers/nvme/host/rdma.c @@ -732,8 +732,6 @@ static void nvme_rdma_reconnect_ctrl_work(struct work_struct *work) if (ret) goto requeue; - nvme_start_keep_alive(&ctrl->ctrl); - if (ctrl->ctrl.queue_count > 1) { ret = nvme_rdma_init_io_queues(ctrl); if (ret) @@ -751,10 +749,7 @@ static void nvme_rdma_reconnect_ctrl_work(struct work_struct *work) WARN_ON_ONCE(!changed); ctrl->ctrl.nr_reconnects = 0; - if (ctrl->ctrl.queue_count > 1) { - nvme_queue_scan(&ctrl->ctrl); - nvme_queue_async_events(&ctrl->ctrl); - } + nvme_start_ctrl(&ctrl->ctrl); dev_info(ctrl->ctrl.device, "Successfully reconnected\n"); @@ -772,7 +767,7 @@ static void nvme_rdma_error_recovery_work(struct work_struct *work) struct nvme_rdma_ctrl, err_work); int i; - nvme_stop_keep_alive(&ctrl->ctrl); + nvme_stop_ctrl(&ctrl->ctrl); for (i = 0; i < ctrl->ctrl.queue_count; i++) clear_bit(NVME_RDMA_Q_LIVE, &ctrl->queues[i].flags); @@ -1603,8 +1598,6 @@ static int nvme_rdma_configure_admin_queue(struct nvme_rdma_ctrl *ctrl) if (error) goto out_cleanup_queue; - nvme_start_keep_alive(&ctrl->ctrl); - return 0; out_cleanup_queue: @@ -1622,7 +1615,6 @@ out_free_queue: static void nvme_rdma_shutdown_ctrl(struct nvme_rdma_ctrl *ctrl) { - nvme_stop_keep_alive(&ctrl->ctrl); cancel_work_sync(&ctrl->err_work); cancel_delayed_work_sync(&ctrl->reconnect_work); @@ -1645,10 +1637,12 @@ static void nvme_rdma_shutdown_ctrl(struct nvme_rdma_ctrl *ctrl) static void __nvme_rdma_remove_ctrl(struct nvme_rdma_ctrl *ctrl, bool shutdown) { - nvme_uninit_ctrl(&ctrl->ctrl); + nvme_stop_ctrl(&ctrl->ctrl); + nvme_remove_namespaces(&ctrl->ctrl); if (shutdown) nvme_rdma_shutdown_ctrl(ctrl); + nvme_uninit_ctrl(&ctrl->ctrl); if (ctrl->ctrl.tagset) { blk_cleanup_queue(ctrl->ctrl.connect_q); blk_mq_free_tag_set(&ctrl->tag_set); @@ -1710,6 +1704,7 @@ static void nvme_rdma_reset_ctrl_work(struct work_struct *work) int ret; bool changed; + nvme_stop_ctrl(&ctrl->ctrl); nvme_rdma_shutdown_ctrl(ctrl); ret = nvme_rdma_configure_admin_queue(ctrl); @@ -1739,11 +1734,7 @@ static void nvme_rdma_reset_ctrl_work(struct work_struct *work) changed = nvme_change_ctrl_state(&ctrl->ctrl, NVME_CTRL_LIVE); WARN_ON_ONCE(!changed); - if (ctrl->ctrl.queue_count > 1) { - nvme_start_queues(&ctrl->ctrl); - nvme_queue_scan(&ctrl->ctrl); - nvme_queue_async_events(&ctrl->ctrl); - } + nvme_start_ctrl(&ctrl->ctrl); return; @@ -1931,15 +1922,11 @@ static struct nvme_ctrl *nvme_rdma_create_ctrl(struct device *dev, list_add_tail(&ctrl->list, &nvme_rdma_ctrl_list); mutex_unlock(&nvme_rdma_ctrl_mutex); - if (ctrl->ctrl.queue_count > 1) { - nvme_queue_scan(&ctrl->ctrl); - nvme_queue_async_events(&ctrl->ctrl); - } + nvme_start_ctrl(&ctrl->ctrl); return &ctrl->ctrl; out_remove_admin_queue: - nvme_stop_keep_alive(&ctrl->ctrl); nvme_rdma_destroy_admin_queue(ctrl); out_kfree_queues: kfree(ctrl->queues); diff --git a/drivers/nvme/target/loop.c b/drivers/nvme/target/loop.c index 6a0b70685e77..717ed7ddb2f6 100644 --- a/drivers/nvme/target/loop.c +++ b/drivers/nvme/target/loop.c @@ -407,8 +407,6 @@ static int nvme_loop_configure_admin_queue(struct nvme_loop_ctrl *ctrl) if (error) goto out_cleanup_queue; - nvme_start_keep_alive(&ctrl->ctrl); - return 0; out_cleanup_queue: @@ -422,8 +420,6 @@ out_free_sq: static void nvme_loop_shutdown_ctrl(struct nvme_loop_ctrl *ctrl) { - nvme_stop_keep_alive(&ctrl->ctrl); - if (ctrl->ctrl.queue_count > 1) { nvme_stop_queues(&ctrl->ctrl); blk_mq_tagset_busy_iter(&ctrl->tag_set, @@ -446,8 +442,10 @@ static void nvme_loop_del_ctrl_work(struct work_struct *work) struct nvme_loop_ctrl *ctrl = container_of(work, struct nvme_loop_ctrl, delete_work); - nvme_uninit_ctrl(&ctrl->ctrl); + nvme_stop_ctrl(&ctrl->ctrl); + nvme_remove_namespaces(&ctrl->ctrl); nvme_loop_shutdown_ctrl(ctrl); + nvme_uninit_ctrl(&ctrl->ctrl); nvme_put_ctrl(&ctrl->ctrl); } @@ -495,6 +493,7 @@ static void nvme_loop_reset_ctrl_work(struct work_struct *work) bool changed; int ret; + nvme_stop_ctrl(&ctrl->ctrl); nvme_loop_shutdown_ctrl(ctrl); ret = nvme_loop_configure_admin_queue(ctrl); @@ -515,10 +514,7 @@ static void nvme_loop_reset_ctrl_work(struct work_struct *work) changed = nvme_change_ctrl_state(&ctrl->ctrl, NVME_CTRL_LIVE); WARN_ON_ONCE(!changed); - nvme_queue_scan(&ctrl->ctrl); - nvme_queue_async_events(&ctrl->ctrl); - - nvme_start_queues(&ctrl->ctrl); + nvme_start_ctrl(&ctrl->ctrl); return; @@ -653,10 +649,7 @@ static struct nvme_ctrl *nvme_loop_create_ctrl(struct device *dev, list_add_tail(&ctrl->list, &nvme_loop_ctrl_list); mutex_unlock(&nvme_loop_ctrl_mutex); - if (opts->nr_io_queues) { - nvme_queue_scan(&ctrl->ctrl); - nvme_queue_async_events(&ctrl->ctrl); - } + nvme_start_ctrl(&ctrl->ctrl); return &ctrl->ctrl; -- cgit v1.2.3 From 842594c8775b585c58459e044708c0335b6aa6b7 Mon Sep 17 00:00:00 2001 From: Sagi Grimberg Date: Wed, 5 Jul 2017 09:56:13 +0300 Subject: nvme-rdma: unconditionally recycle the request mr When our RDMA queue-pair is torn down with high load of I/O traffic, we have no way of knowing if the memory region was actually registered by the reg_mr work request as it completion flushes with error (hw might have done it or not). So in order to not deal with all this uncertanty, we simply recycle the MR in reinit_request. Reviewed-by: Christoph Hellwig Signed-off-by: Sagi Grimberg --- drivers/nvme/host/rdma.c | 3 --- 1 file changed, 3 deletions(-) (limited to 'drivers/nvme/host/rdma.c') diff --git a/drivers/nvme/host/rdma.c b/drivers/nvme/host/rdma.c index ccdbd99d5a1c..da04df1af231 100644 --- a/drivers/nvme/host/rdma.c +++ b/drivers/nvme/host/rdma.c @@ -272,9 +272,6 @@ static int nvme_rdma_reinit_request(void *data, struct request *rq) struct nvme_rdma_request *req = blk_mq_rq_to_pdu(rq); int ret = 0; - if (!req->mr->need_inval) - goto out; - ib_dereg_mr(req->mr); req->mr = ib_alloc_mr(dev->pd, IB_MR_TYPE_MEM_REG, -- cgit v1.2.3