Hello I found this failure on stable 5.13.16-rc1[1] and cannot reproduce it on 5.14, seems we are missing commit[2] on 5.13.y, could anyone help check it?
[1] # nvme_trtype=tcp ./check nvme/014 nvme/014 (flush a NVMeOF block device-backed ns) [passed] runtime 30.745s ... 22.110s # lsblk NAME MAJ:MIN RM SIZE RO TYPE MOUNTPOINT sda 8:0 0 465.8G 0 disk ├─sda1 8:1 0 1G 0 part /boot ├─sda2 8:2 0 16G 0 part /mnt/dmtest/data ├─sda3 8:3 0 15G 0 part /mnt/xfstests/mnt1 ├─sda4 8:4 0 1K 0 part ├─sda5 8:5 0 15G 0 part /mnt/xfstests/mnt2 ├─sda6 8:6 0 7.9G 0 part [SWAP] ├─sda7 8:7 0 5G 0 part /mnt/xfstests/logwrites ├─sda8 8:8 0 1G 0 part /mnt/dmtest/metadata └─sda9 8:9 0 404.9G 0 part / zram0 252:0 0 8G 0 disk [SWAP] nvme0n1 259:3 0 1G 0 disk
[ 133.791507] run blktests nvme/014 at 2021-09-12 08:08:17 [ 133.908750] loop0: detected capacity change from 0 to 2097152 [ 133.927551] nvmet: adding nsid 1 to subsystem blktests-subsystem-1 [ 133.942518] nvmet_tcp: enabling port 0 (127.0.0.1:4420) [ 133.977746] nvmet: creating controller 1 for subsystem blktests-subsystem-1 for NQN nqn.2014-08.org.nvmexpress:uuid:c4c5023e9bcd4cb5889e9510622eb88c. [ 133.994170] nvme nvme0: creating 4 I/O queues. [ 133.999881] nvme nvme0: mapped 4/0/0 default/read/poll queues. [ 134.008683] nvme nvme0: new ctrl: NQN "blktests-subsystem-1", addr 127.0.0.1:4420 [ 151.223049] nvme nvme0: using deprecated NVME_IOCTL_IO_CMD ioctl on the char device! [ 155.675634] nvme nvme0: Removing ctrl: NQN "blktests-subsystem-1" [ 155.682902] block nvme0n1: no available path - failing I/O [ 155.689137] block nvme0n1: no available path - failing I/O [ 155.695295] Buffer I/O error on dev nvme0n1, logical block 32, async page read
[2] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?i... 5396fdac56d8 nvme: fix refcounting imbalance when all paths are down