12.0-R-p2: login(1) hang on [dp->dp_spaceavail_cv]

[Ravi Pokala]

Hi all,

Yesterday, I found I was unable to SSH into one of my servers. When I had physical access, I tried to log in using the video console, but that didn't work either. I was able to use ^T to try and see what was going on, and login(1) reported that it was waiting on [dp->dp_spaceavail_cv]. I poked it periodically for several minutes, but it was unchanged. I ultimately gave up and used <ctr>-<alt>-<del> to trigger the ACPI shutdown hook, but that hung also - specifically, trying to write entropy.

Unfortunately, I was not able to get a vmcore. But, it looks like there is only one place which waits on that:

    sys/cddl/contrib/opensolaris/uts/common/fs/zfs/dmu_tx.c: dmu_tx_wait()

On the way back up, it took an inordinately long time to mount the local filesystems (~10 minutes, when it typically takes <1 second), with significant drive activity. The good news is that it *did* eventually mount, and `zfs scrub' didn't show any errors on either pool.

Anyway, just wanted to throw that out there in case anyone has any thoughts.

Thanks,

Ravi (rpokala@)