[PATCH] NVMe: Skip orderly shutdown on failed devices

[PATCH] NVMe: Skip orderly shutdown on failed devices

[Keith Busch]

Rather than skipping shutdown only for device's that have been removed,
skip the orderly shutdown on failed devices to avoid the long timeout
handling that inevitably happens when deleting queues on such a device.

Signed-off-by: Keith Busch <keith.busch at intel.com>
---
 drivers/block/nvme-core.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/block/nvme-core.c b/drivers/block/nvme-core.c
index b634233..30b7f64 100644
--- a/drivers/block/nvme-core.c
+++ b/drivers/block/nvme-core.c
@@ -2525,7 +2525,7 @@ static void nvme_dev_shutdown(struct nvme_dev *dev)
 	dev->initialized = 0;
 	nvme_dev_list_remove(dev);
 
-	if (!dev->bar || (dev->bar && readl(&dev->bar->csts) == -1)) {
+	if (!dev->bar || (dev->bar && readl(&dev->bar->csts) & NVME_CSTS_CFS)) {
 		for (i = dev->queue_count - 1; i >= 0; i--) {
 			struct nvme_queue *nvmeq = raw_nvmeq(dev, i);
 			nvme_suspend_queue(nvmeq);
-- 
1.7.10.4

[PATCH] NVMe: Skip orderly shutdown on failed devices

[Matthew Wilcox]

On Mon, Jun 23, 2014@11:46:35AM -0600, Keith Busch wrote:
> Rather than skipping shutdown only for device's that have been removed,
> skip the orderly shutdown on failed devices to avoid the long timeout
> handling that inevitably happens when deleting queues on such a device.

How about this instead?

@@ -2500,11 +2500,14 @@ static void nvme_dev_list_remove(struct nvme_dev *dev)
 static void nvme_dev_shutdown(struct nvme_dev *dev)
 {
        int i;
+       u32 csts = -1;
 
        dev->initialized = 0;
        nvme_dev_list_remove(dev);
+       if (dev->bar)
+               csts = readl(&dev->bar->csts);
 
-       if (!dev->bar || (dev->bar && readl(&dev->bar->csts) == -1)) {
+       if (csts & NVME_CSTS_CFS) {
                for (i = dev->queue_count - 1; i >= 0; i--) {
                        struct nvme_queue *nvmeq = raw_nvmeq(dev, i);
                        nvme_suspend_queue(nvmeq);

[PATCH] NVMe: Skip orderly shutdown on failed devices

[Keith Busch]

On Wed, 25 Jun 2014, Matthew Wilcox wrote:
> On Mon, Jun 23, 2014@11:46:35AM -0600, Keith Busch wrote:
>> Rather than skipping shutdown only for device's that have been removed,
>> skip the orderly shutdown on failed devices to avoid the long timeout
>> handling that inevitably happens when deleting queues on such a device.
>
> How about this instead?

Yes, that's much better.

I also just encountered a device that dropped the ready bit. I'm not
sure why, but the device definitely wasn't ready to do anything, so I
can add the check and skip shutdown there as well.

> @@ -2500,11 +2500,14 @@ static void nvme_dev_list_remove(struct nvme_dev *dev)
> static void nvme_dev_shutdown(struct nvme_dev *dev)
> {
>        int i;
> +       u32 csts = -1;
>
>        dev->initialized = 0;
>        nvme_dev_list_remove(dev);
> +       if (dev->bar)
> +               csts = readl(&dev->bar->csts);
>
> -       if (!dev->bar || (dev->bar && readl(&dev->bar->csts) == -1)) {
> +       if (csts & NVME_CSTS_CFS) {
>                for (i = dev->queue_count - 1; i >= 0; i--) {
>                        struct nvme_queue *nvmeq = raw_nvmeq(dev, i);
>                        nvme_suspend_queue(nvmeq);