Description
On Ubuntu 10.04 amd64, with a git checkout (ID: 7dfe24e), I loaded the zfs module, made 7 pools named "disk1" through "disk7", and made a zvol on each, named "d1" through "d7", respectively.
I assembled an MD RAID-6 out of the zvols, made an ext4 FS on it, mounted it and started copying data to test the behavior. While it was copying, I pulled one of the seven disks. Linux appeared to not notice in a meaningful way - the pool, disk2, reported >1600 write errors, but thought the disk was still online, as did Linux, though any IO to it failed.
I rebooted, the machine hung during shutdown, I SysRq-S,U,B'd the machine.
On next bootup, I tried "modprobe zfs", and it's been sitting here for about 10 minutes, with no output. dmesg|tail reports "INFO: task modprobe:2606 blocked for more than 120 seconds.", with a trace. (As soon as I figure out how to attach things in github's issue tracker, I'll attach the trace.)