[PATCH V4 4/7] nvme: pci: freeze queue in nvme_dev_disable() in case of error recovery

[Ming Lei]

When nvme_dev_disable() is used for error recovery, we should always
freeze queues before shutdown controller:

- reset handler supposes queues are frozen, and will wait_freeze &
unfreeze them explicitly, if queues aren't frozen during nvme_dev_disable(),
reset handler may wait forever even though there isn't any requests
allocated.

- this way may avoid to cancel lots of requests during error recovery

This patch introduces the parameter of 'freeze_queue' for fixing this
issue.

Cc: Jianchao Wang <jianchao.w.wang at oracle.com>
Cc: Christoph Hellwig <hch at lst.de>
Cc: Sagi Grimberg <sagi at grimberg.me>
Cc: linux-nvme at lists.infradead.org
Cc: Laurence Oberman <loberman at redhat.com>
Signed-off-by: Ming Lei <ming.lei at redhat.com>
---
 drivers/nvme/host/pci.c | 47 ++++++++++++++++++++++++++++++++---------------
 1 file changed, 32 insertions(+), 15 deletions(-)

diff --git a/drivers/nvme/host/pci.c b/drivers/nvme/host/pci.c
index 57bd7bebd1e5..1fafe5d01355 100644
--- a/drivers/nvme/host/pci.c
+++ b/drivers/nvme/host/pci.c
@@ -69,7 +69,8 @@ struct nvme_dev;
 struct nvme_queue;
 
 static void nvme_process_cq(struct nvme_queue *nvmeq);
-static void nvme_dev_disable(struct nvme_dev *dev, bool shutdown);
+static void nvme_dev_disable(struct nvme_dev *dev, bool shutdown, bool
+		freeze_queue);
 
 /*
  * Represents an NVM Express device.  Each nvme_dev is a PCI function.
@@ -1197,7 +1198,7 @@ static enum blk_eh_timer_return nvme_timeout(struct request *req, bool reserved)
 	 */
 	if (nvme_should_reset(dev, csts)) {
 		nvme_warn_reset(dev, csts);
-		nvme_dev_disable(dev, false);
+		nvme_dev_disable(dev, false, true);
 		nvme_reset_ctrl(&dev->ctrl);
 		return BLK_EH_HANDLED;
 	}
@@ -1224,7 +1225,7 @@ static enum blk_eh_timer_return nvme_timeout(struct request *req, bool reserved)
 		dev_warn(dev->ctrl.device,
 			 "I/O %d QID %d timeout, disable controller\n",
 			 req->tag, nvmeq->qid);
-		nvme_dev_disable(dev, false);
+		nvme_dev_disable(dev, false, false);
 		nvme_req(req)->flags |= NVME_REQ_CANCELLED;
 		return BLK_EH_HANDLED;
 	default:
@@ -1240,7 +1241,7 @@ static enum blk_eh_timer_return nvme_timeout(struct request *req, bool reserved)
 		dev_warn(dev->ctrl.device,
 			 "I/O %d QID %d timeout, reset controller\n",
 			 req->tag, nvmeq->qid);
-		nvme_dev_disable(dev, false);
+		nvme_dev_disable(dev, false, true);
 		nvme_reset_ctrl(&dev->ctrl);
 
 		/*
@@ -2239,19 +2240,35 @@ static void nvme_pci_disable(struct nvme_dev *dev)
 	}
 }
 
-static void nvme_dev_disable(struct nvme_dev *dev, bool shutdown)
+/*
+ * Resetting often follows nvme_dev_disable(), so queues need to be frozen
+ * before resetting.
+ */
+static void nvme_dev_disable(struct nvme_dev *dev, bool shutdown, bool
+		freeze_queue)
 {
 	int i;
 	bool dead = true;
 	struct pci_dev *pdev = to_pci_dev(dev->dev);
 	bool frozen = false;
 
+	/*
+	 * 'freeze_queue' is only valid for non-shutdown, and we do
+	 * inline freeze & wait_freeze_timeout for shutdown just for
+	 * completing as many as possible requests before shutdown
+	 */
+	if (shutdown)
+		freeze_queue = false;
+
+	if (freeze_queue)
+		nvme_start_freeze(&dev->ctrl);
+
 	mutex_lock(&dev->shutdown_lock);
 	if (pci_is_enabled(pdev)) {
 		u32 csts = readl(dev->bar + NVME_REG_CSTS);
 
-		if (dev->ctrl.state == NVME_CTRL_LIVE ||
-		    dev->ctrl.state == NVME_CTRL_RESETTING) {
+		if (shutdown && (dev->ctrl.state == NVME_CTRL_LIVE ||
+		    dev->ctrl.state == NVME_CTRL_RESETTING)) {
 			nvme_start_freeze(&dev->ctrl);
 			frozen = true;
 		}
@@ -2343,7 +2360,7 @@ static void nvme_remove_dead_ctrl(struct nvme_dev *dev, int status)
 	dev_warn(dev->ctrl.device, "Removing after probe failure status: %d\n", status);
 
 	nvme_get_ctrl(&dev->ctrl);
-	nvme_dev_disable(dev, false);
+	nvme_dev_disable(dev, false, false);
 	if (!queue_work(nvme_wq, &dev->remove_work))
 		nvme_put_ctrl(&dev->ctrl);
 }
@@ -2364,7 +2381,7 @@ static void nvme_reset_work(struct work_struct *work)
 	 * moving on.
 	 */
 	if (dev->ctrl.ctrl_config & NVME_CC_ENABLE)
-		nvme_dev_disable(dev, false);
+		nvme_dev_disable(dev, false, false);
 
 	/*
 	 * Introduce CONNECTING state from nvme-fc/rdma transports to mark the
@@ -2613,7 +2630,7 @@ static int nvme_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 static void nvme_reset_prepare(struct pci_dev *pdev)
 {
 	struct nvme_dev *dev = pci_get_drvdata(pdev);
-	nvme_dev_disable(dev, false);
+	nvme_dev_disable(dev, false, true);
 }
 
 static void nvme_reset_done(struct pci_dev *pdev)
@@ -2625,7 +2642,7 @@ static void nvme_reset_done(struct pci_dev *pdev)
 static void nvme_shutdown(struct pci_dev *pdev)
 {
 	struct nvme_dev *dev = pci_get_drvdata(pdev);
-	nvme_dev_disable(dev, true);
+	nvme_dev_disable(dev, true, false);
 }
 
 /*
@@ -2644,13 +2661,13 @@ static void nvme_remove(struct pci_dev *pdev)
 
 	if (!pci_device_is_present(pdev)) {
 		nvme_change_ctrl_state(&dev->ctrl, NVME_CTRL_DEAD);
-		nvme_dev_disable(dev, false);
+		nvme_dev_disable(dev, false, false);
 	}
 
 	flush_work(&dev->ctrl.reset_work);
 	nvme_stop_ctrl(&dev->ctrl);
 	nvme_remove_namespaces(&dev->ctrl);
-	nvme_dev_disable(dev, true);
+	nvme_dev_disable(dev, true, false);
 	nvme_free_host_mem(dev);
 	nvme_dev_remove_admin(dev);
 	nvme_free_queues(dev, 0);
@@ -2684,7 +2701,7 @@ static int nvme_suspend(struct device *dev)
 	struct pci_dev *pdev = to_pci_dev(dev);
 	struct nvme_dev *ndev = pci_get_drvdata(pdev);
 
-	nvme_dev_disable(ndev, true);
+	nvme_dev_disable(ndev, true, false);
 	return 0;
 }
 
@@ -2716,7 +2733,7 @@ static pci_ers_result_t nvme_error_detected(struct pci_dev *pdev,
 	case pci_channel_io_frozen:
 		dev_warn(dev->ctrl.device,
 			"frozen state error detected, reset controller\n");
-		nvme_dev_disable(dev, false);
+		nvme_dev_disable(dev, false, true);
 		return PCI_ERS_RESULT_NEED_RESET;
 	case pci_channel_io_perm_failure:
 		dev_warn(dev->ctrl.device,
-- 
2.9.5