From 4cff280a5fccf6513ed9e895bb3a4e7ad8b0cedc Mon Sep 17 00:00:00 2001 From: James Smart Date: Wed, 14 Nov 2018 16:35:10 -0800 Subject: nvme-fc: resolve io failures during connect If an io error occurs on an io issued while connecting, recovery of the io falls flat as the state checking ends up nooping the error handler. Create an err_work work item that is scheduled upon an io error while connecting. The work thread terminates all io on all queues and marks the queues as not connected. The termination of the io will return back to the callee, which will then back out of the connection attempt and will reschedule, if possible, the connection attempt. The changes: - in case there are several commands hitting the error handler, a state flag is kept so that the error work is only scheduled once, on the first error. The subsequent errors can be ignored. - The calling sequence to stop keep alive and terminate the queues and their io is lifted from the reset routine. Made a small service routine used by both reset and err_work. - During debugging, found that the teardown path can reference an uninitialized pointer, resulting in a NULL pointer oops. The aen_ops weren't initialized yet. Add validation on their initialization before calling the teardown routine. Signed-off-by: James Smart Signed-off-by: Christoph Hellwig --- drivers/nvme/host/fc.c | 73 +++++++++++++++++++++++++++++++++++++++++++------- 1 file changed, 63 insertions(+), 10 deletions(-) (limited to 'drivers/nvme/host/fc.c') diff --git a/drivers/nvme/host/fc.c b/drivers/nvme/host/fc.c index 0b70c8bab045..54032c466636 100644 --- a/drivers/nvme/host/fc.c +++ b/drivers/nvme/host/fc.c @@ -152,6 +152,7 @@ struct nvme_fc_ctrl { bool ioq_live; bool assoc_active; + atomic_t err_work_active; u64 association_id; struct list_head ctrl_list; /* rport->ctrl_list */ @@ -160,6 +161,7 @@ struct nvme_fc_ctrl { struct blk_mq_tag_set tag_set; struct delayed_work connect_work; + struct work_struct err_work; struct kref ref; u32 flags; @@ -1531,6 +1533,10 @@ nvme_fc_abort_aen_ops(struct nvme_fc_ctrl *ctrl) struct nvme_fc_fcp_op *aen_op = ctrl->aen_ops; int i; + /* ensure we've initialized the ops once */ + if (!(aen_op->flags & FCOP_FLAGS_AEN)) + return; + for (i = 0; i < NVME_NR_AEN_COMMANDS; i++, aen_op++) __nvme_fc_abort_op(ctrl, aen_op); } @@ -2049,7 +2055,25 @@ nvme_fc_nvme_ctrl_freed(struct nvme_ctrl *nctrl) static void nvme_fc_error_recovery(struct nvme_fc_ctrl *ctrl, char *errmsg) { - /* only proceed if in LIVE state - e.g. on first error */ + int active; + + /* + * if an error (io timeout, etc) while (re)connecting, + * it's an error on creating the new association. + * Start the error recovery thread if it hasn't already + * been started. It is expected there could be multiple + * ios hitting this path before things are cleaned up. + */ + if (ctrl->ctrl.state == NVME_CTRL_CONNECTING) { + active = atomic_xchg(&ctrl->err_work_active, 1); + if (!active && !schedule_work(&ctrl->err_work)) { + atomic_set(&ctrl->err_work_active, 0); + WARN_ON(1); + } + return; + } + + /* Otherwise, only proceed if in LIVE state - e.g. on first error */ if (ctrl->ctrl.state != NVME_CTRL_LIVE) return; @@ -2814,6 +2838,7 @@ nvme_fc_delete_ctrl(struct nvme_ctrl *nctrl) { struct nvme_fc_ctrl *ctrl = to_fc_ctrl(nctrl); + cancel_work_sync(&ctrl->err_work); cancel_delayed_work_sync(&ctrl->connect_work); /* * kill the association on the link side. this will block @@ -2866,23 +2891,30 @@ nvme_fc_reconnect_or_delete(struct nvme_fc_ctrl *ctrl, int status) } static void -nvme_fc_reset_ctrl_work(struct work_struct *work) +__nvme_fc_terminate_io(struct nvme_fc_ctrl *ctrl) { - struct nvme_fc_ctrl *ctrl = - container_of(work, struct nvme_fc_ctrl, ctrl.reset_work); - int ret; - - nvme_stop_ctrl(&ctrl->ctrl); + nvme_stop_keep_alive(&ctrl->ctrl); /* will block will waiting for io to terminate */ nvme_fc_delete_association(ctrl); - if (!nvme_change_ctrl_state(&ctrl->ctrl, NVME_CTRL_CONNECTING)) { + if (ctrl->ctrl.state != NVME_CTRL_CONNECTING && + !nvme_change_ctrl_state(&ctrl->ctrl, NVME_CTRL_CONNECTING)) dev_err(ctrl->ctrl.device, "NVME-FC{%d}: error_recovery: Couldn't change state " "to CONNECTING\n", ctrl->cnum); - return; - } +} + +static void +nvme_fc_reset_ctrl_work(struct work_struct *work) +{ + struct nvme_fc_ctrl *ctrl = + container_of(work, struct nvme_fc_ctrl, ctrl.reset_work); + int ret; + + __nvme_fc_terminate_io(ctrl); + + nvme_stop_ctrl(&ctrl->ctrl); if (ctrl->rport->remoteport.port_state == FC_OBJSTATE_ONLINE) ret = nvme_fc_create_association(ctrl); @@ -2897,6 +2929,24 @@ nvme_fc_reset_ctrl_work(struct work_struct *work) ctrl->cnum); } +static void +nvme_fc_connect_err_work(struct work_struct *work) +{ + struct nvme_fc_ctrl *ctrl = + container_of(work, struct nvme_fc_ctrl, err_work); + + __nvme_fc_terminate_io(ctrl); + + atomic_set(&ctrl->err_work_active, 0); + + /* + * Rescheduling the connection after recovering + * from the io error is left to the reconnect work + * item, which is what should have stalled waiting on + * the io that had the error that scheduled this work. + */ +} + static const struct nvme_ctrl_ops nvme_fc_ctrl_ops = { .name = "fc", .module = THIS_MODULE, @@ -3007,6 +3057,7 @@ nvme_fc_init_ctrl(struct device *dev, struct nvmf_ctrl_options *opts, ctrl->cnum = idx; ctrl->ioq_live = false; ctrl->assoc_active = false; + atomic_set(&ctrl->err_work_active, 0); init_waitqueue_head(&ctrl->ioabort_wait); get_device(ctrl->dev); @@ -3014,6 +3065,7 @@ nvme_fc_init_ctrl(struct device *dev, struct nvmf_ctrl_options *opts, INIT_WORK(&ctrl->ctrl.reset_work, nvme_fc_reset_ctrl_work); INIT_DELAYED_WORK(&ctrl->connect_work, nvme_fc_connect_ctrl_work); + INIT_WORK(&ctrl->err_work, nvme_fc_connect_err_work); spin_lock_init(&ctrl->lock); /* io queue count */ @@ -3103,6 +3155,7 @@ nvme_fc_init_ctrl(struct device *dev, struct nvmf_ctrl_options *opts, fail_ctrl: nvme_change_ctrl_state(&ctrl->ctrl, NVME_CTRL_DELETING); cancel_work_sync(&ctrl->ctrl.reset_work); + cancel_work_sync(&ctrl->err_work); cancel_delayed_work_sync(&ctrl->connect_work); ctrl->ctrl.opts = NULL; -- cgit v1.2.3