Merge branch 'nvme-5.4' of git://git.infradead.org/nvme into for-linus

Pull NVMe updates from Keith:

"This is a collection of bug fixes committed since the previous pull
 request that address deadlocks, double resets, memory leaks, and other
 regression."

* 'nvme-5.4' of git://git.infradead.org/nvme:
  nvme-pci: Set the prp2 correctly when using more than 4k page
  nvme-tcp: fix possible leakage during error flow
  nvmet-loop: fix possible leakage during error flow
  nvme-tcp: Initialize sk->sk_ll_usec only with NET_RX_BUSY_POLL
  nvme: Wait for reset state when required
  nvme: Prevent resets during paused controller state
  nvme: Restart request timers in resetting state
  nvme: Remove ADMIN_ONLY state
  nvme-pci: Free tagset if no IO queues
  nvme: retain split access workaround for capability reads
  nvme: fix possible deadlock when nvme_update_formats fails

Author: axboe

drivers/nvme/host/core.c

@@ -116,10 +116,26 @@ static void nvme_queue_scan(struct nvme_ctrl *ctrl)
 	/*
 	 * Only new queue scan work when admin and IO queues are both alive
 	 */
-	if (ctrl->state == NVME_CTRL_LIVE)
+	if (ctrl->state == NVME_CTRL_LIVE && ctrl->tagset)
 		queue_work(nvme_wq, &ctrl->scan_work);
 }
 
+/*
+ * Use this function to proceed with scheduling reset_work for a controller
+ * that had previously been set to the resetting state. This is intended for
+ * code paths that can't be interrupted by other reset attempts. A hot removal
+ * may prevent this from succeeding.
+ */
+int nvme_try_sched_reset(struct nvme_ctrl *ctrl)
+{
+	if (ctrl->state != NVME_CTRL_RESETTING)
+		return -EBUSY;
+	if (!queue_work(nvme_reset_wq, &ctrl->reset_work))
+		return -EBUSY;
+	return 0;
+}
+EXPORT_SYMBOL_GPL(nvme_try_sched_reset);
+
 int nvme_reset_ctrl(struct nvme_ctrl *ctrl)
 {
 	if (!nvme_change_ctrl_state(ctrl, NVME_CTRL_RESETTING))
@@ -137,8 +153,7 @@ int nvme_reset_ctrl_sync(struct nvme_ctrl *ctrl)
 	ret = nvme_reset_ctrl(ctrl);
 	if (!ret) {
 		flush_work(&ctrl->reset_work);
-		if (ctrl->state != NVME_CTRL_LIVE &&
-		    ctrl->state != NVME_CTRL_ADMIN_ONLY)
+		if (ctrl->state != NVME_CTRL_LIVE)
 			ret = -ENETRESET;
 	}
 
@@ -315,15 +330,6 @@ bool nvme_change_ctrl_state(struct nvme_ctrl *ctrl,
 
 	old_state = ctrl->state;
 	switch (new_state) {
-	case NVME_CTRL_ADMIN_ONLY:
-		switch (old_state) {
-		case NVME_CTRL_CONNECTING:
-			changed = true;
-			/* FALLTHRU */
-		default:
-			break;
-		}
-		break;
 	case NVME_CTRL_LIVE:
 		switch (old_state) {
 		case NVME_CTRL_NEW:
@@ -339,7 +345,6 @@ bool nvme_change_ctrl_state(struct nvme_ctrl *ctrl,
 		switch (old_state) {
 		case NVME_CTRL_NEW:
 		case NVME_CTRL_LIVE:
-		case NVME_CTRL_ADMIN_ONLY:
 			changed = true;
 			/* FALLTHRU */
 		default:
@@ -359,7 +364,6 @@ bool nvme_change_ctrl_state(struct nvme_ctrl *ctrl,
 	case NVME_CTRL_DELETING:
 		switch (old_state) {
 		case NVME_CTRL_LIVE:
-		case NVME_CTRL_ADMIN_ONLY:
 		case NVME_CTRL_RESETTING:
 		case NVME_CTRL_CONNECTING:
 			changed = true;
@@ -381,8 +385,10 @@ bool nvme_change_ctrl_state(struct nvme_ctrl *ctrl,
 		break;
 	}
 
-	if (changed)
+	if (changed) {
 		ctrl->state = new_state;
+		wake_up_all(&ctrl->state_wq);
+	}
 
 	spin_unlock_irqrestore(&ctrl->lock, flags);
 	if (changed && ctrl->state == NVME_CTRL_LIVE)
@@ -391,6 +397,39 @@ bool nvme_change_ctrl_state(struct nvme_ctrl *ctrl,
 }
 EXPORT_SYMBOL_GPL(nvme_change_ctrl_state);
 
+/*
+ * Returns true for sink states that can't ever transition back to live.
+ */
+static bool nvme_state_terminal(struct nvme_ctrl *ctrl)
+{
+	switch (ctrl->state) {
+	case NVME_CTRL_NEW:
+	case NVME_CTRL_LIVE:
+	case NVME_CTRL_RESETTING:
+	case NVME_CTRL_CONNECTING:
+		return false;
+	case NVME_CTRL_DELETING:
+	case NVME_CTRL_DEAD:
+		return true;
+	default:
+		WARN_ONCE(1, "Unhandled ctrl state:%d", ctrl->state);
+		return true;
+	}
+}
+
+/*
+ * Waits for the controller state to be resetting, or returns false if it is
+ * not possible to ever transition to that state.
+ */
+bool nvme_wait_reset(struct nvme_ctrl *ctrl)
+{
+	wait_event(ctrl->state_wq,
+		   nvme_change_ctrl_state(ctrl, NVME_CTRL_RESETTING) ||
+		   nvme_state_terminal(ctrl));
+	return ctrl->state == NVME_CTRL_RESETTING;
+}
+EXPORT_SYMBOL_GPL(nvme_wait_reset);
+
 static void nvme_free_ns_head(struct kref *ref)
 {
 	struct nvme_ns_head *head =
@@ -1306,8 +1345,6 @@ static void nvme_update_formats(struct nvme_ctrl *ctrl)
 		if (ns->disk && nvme_revalidate_disk(ns->disk))
 			nvme_set_queue_dying(ns);
 	up_read(&ctrl->namespaces_rwsem);
-
-	nvme_remove_invalid_namespaces(ctrl, NVME_NSID_ALL);
 }
 
 static void nvme_passthru_end(struct nvme_ctrl *ctrl, u32 effects)
@@ -1323,6 +1360,7 @@ static void nvme_passthru_end(struct nvme_ctrl *ctrl, u32 effects)
 		nvme_unfreeze(ctrl);
 		nvme_mpath_unfreeze(ctrl->subsys);
 		mutex_unlock(&ctrl->subsys->lock);
+		nvme_remove_invalid_namespaces(ctrl, NVME_NSID_ALL);
 		mutex_unlock(&ctrl->scan_lock);
 	}
 	if (effects & NVME_CMD_EFFECTS_CCC)
@@ -2874,7 +2912,6 @@ static int nvme_dev_open(struct inode *inode, struct file *file)
 
 	switch (ctrl->state) {
 	case NVME_CTRL_LIVE:
-	case NVME_CTRL_ADMIN_ONLY:
 		break;
 	default:
 		return -EWOULDBLOCK;
@@ -3168,7 +3205,6 @@ static ssize_t nvme_sysfs_show_state(struct device *dev,
 	static const char *const state_name[] = {
 		[NVME_CTRL_NEW]		= "new",
 		[NVME_CTRL_LIVE]	= "live",
-		[NVME_CTRL_ADMIN_ONLY]	= "only-admin",
 		[NVME_CTRL_RESETTING]	= "resetting",
 		[NVME_CTRL_CONNECTING]	= "connecting",
 		[NVME_CTRL_DELETING]	= "deleting",
@@ -3679,11 +3715,10 @@ static void nvme_scan_work(struct work_struct *work)
 	struct nvme_id_ctrl *id;
 	unsigned nn;
 
-	if (ctrl->state != NVME_CTRL_LIVE)
+	/* No tagset on a live ctrl means IO queues could not created */
+	if (ctrl->state != NVME_CTRL_LIVE || !ctrl->tagset)
 		return;
 
-	WARN_ON_ONCE(!ctrl->tagset);
-
 	if (test_and_clear_bit(NVME_AER_NOTICE_NS_CHANGED, &ctrl->events)) {
 		dev_info(ctrl->device, "rescanning namespaces.\n");
 		nvme_clear_changed_ns_log(ctrl);
@@ -3844,13 +3879,13 @@ static void nvme_fw_act_work(struct work_struct *work)
 		if (time_after(jiffies, fw_act_timeout)) {
 			dev_warn(ctrl->device,
 				"Fw activation timeout, reset controller\n");
-			nvme_reset_ctrl(ctrl);
-			break;
+			nvme_try_sched_reset(ctrl);
+			return;
 		}
 		msleep(100);
 	}
 
-	if (ctrl->state != NVME_CTRL_LIVE)
+	if (!nvme_change_ctrl_state(ctrl, NVME_CTRL_LIVE))
 		return;
 
 	nvme_start_queues(ctrl);
@@ -3870,7 +3905,13 @@ static void nvme_handle_aen_notice(struct nvme_ctrl *ctrl, u32 result)
 		nvme_queue_scan(ctrl);
 		break;
 	case NVME_AER_NOTICE_FW_ACT_STARTING:
-		queue_work(nvme_wq, &ctrl->fw_act_work);
+		/*
+		 * We are (ab)using the RESETTING state to prevent subsequent
+		 * recovery actions from interfering with the controller's
+		 * firmware activation.
+		 */
+		if (nvme_change_ctrl_state(ctrl, NVME_CTRL_RESETTING))
+			queue_work(nvme_wq, &ctrl->fw_act_work);
 		break;
 #ifdef CONFIG_NVME_MULTIPATH
 	case NVME_AER_NOTICE_ANA:
@@ -3993,6 +4034,7 @@ int nvme_init_ctrl(struct nvme_ctrl *ctrl, struct device *dev,
 	INIT_WORK(&ctrl->async_event_work, nvme_async_event_work);
 	INIT_WORK(&ctrl->fw_act_work, nvme_fw_act_work);
 	INIT_WORK(&ctrl->delete_work, nvme_delete_ctrl_work);
+	init_waitqueue_head(&ctrl->state_wq);
 
 	INIT_DELAYED_WORK(&ctrl->ka_work, nvme_keep_alive_work);
 	memset(&ctrl->ka_cmd, 0, sizeof(ctrl->ka_cmd));

drivers/nvme/host/fabrics.h

@@ -182,8 +182,7 @@ bool nvmf_ip_options_match(struct nvme_ctrl *ctrl,
 static inline bool nvmf_check_ready(struct nvme_ctrl *ctrl, struct request *rq,
 		bool queue_live)
 {
-	if (likely(ctrl->state == NVME_CTRL_LIVE ||
-		   ctrl->state == NVME_CTRL_ADMIN_ONLY))
+	if (likely(ctrl->state == NVME_CTRL_LIVE))
 		return true;
 	return __nvmf_check_ready(ctrl, rq, queue_live);
 }

drivers/nvme/host/nvme.h

@@ -15,6 +15,7 @@
 #include <linux/sed-opal.h>
 #include <linux/fault-inject.h>
 #include <linux/rcupdate.h>
+#include <linux/wait.h>
 
 #include <trace/events/block.h>
 
@@ -161,7 +162,6 @@ static inline u16 nvme_req_qid(struct request *req)
 enum nvme_ctrl_state {
 	NVME_CTRL_NEW,
 	NVME_CTRL_LIVE,
-	NVME_CTRL_ADMIN_ONLY,    /* Only admin queue live */
 	NVME_CTRL_RESETTING,
 	NVME_CTRL_CONNECTING,
 	NVME_CTRL_DELETING,
@@ -199,6 +199,7 @@ struct nvme_ctrl {
 	struct cdev cdev;
 	struct work_struct reset_work;
 	struct work_struct delete_work;
+	wait_queue_head_t state_wq;
 
 	struct nvme_subsystem *subsys;
 	struct list_head subsys_entry;
@@ -449,6 +450,7 @@ void nvme_complete_rq(struct request *req);
 bool nvme_cancel_request(struct request *req, void *data, bool reserved);
 bool nvme_change_ctrl_state(struct nvme_ctrl *ctrl,
 		enum nvme_ctrl_state new_state);
+bool nvme_wait_reset(struct nvme_ctrl *ctrl);
 int nvme_disable_ctrl(struct nvme_ctrl *ctrl);
 int nvme_enable_ctrl(struct nvme_ctrl *ctrl);
 int nvme_shutdown_ctrl(struct nvme_ctrl *ctrl);
@@ -499,6 +501,7 @@ int nvme_set_queue_count(struct nvme_ctrl *ctrl, int *count);
 void nvme_stop_keep_alive(struct nvme_ctrl *ctrl);
 int nvme_reset_ctrl(struct nvme_ctrl *ctrl);
 int nvme_reset_ctrl_sync(struct nvme_ctrl *ctrl);
+int nvme_try_sched_reset(struct nvme_ctrl *ctrl);
 int nvme_delete_ctrl(struct nvme_ctrl *ctrl);
 
 int nvme_get_log(struct nvme_ctrl *ctrl, u32 nsid, u8 log_page, u8 lsp,