Hey, I just confirmed my theory that zfs/zpool is broken in stable/11. After experiencing it on my own machine, I fired up a google compute instance with 8 CPUs and checked out stable/11, and recompiled it with make -j10 buildworld. After make installworld, the zfs and zpool binaries are getting a SIGABRT: ---snip--- [root@freebsd-instance /usr/src]# zfs list internal error: Invalid argument Abort trap (core dumped) ---snip--- the problem seems to be /lib/libzfs_core.so.2 and/or /lib/libzfs.so.2, since when I put those to libraries back from base.txz, the commands start working. moreover, if I compile a kernel from the stable/11 now, at the next boot, after loading the kernel, it won't find my zpool and panics at startup. So to reproduce: - fire up a freebsd 11 instance somewhere - pkg install subversion - cd /usr - svn co https://svn0.eu.freebsd.org/base/stable/11 src - cd src - make -j10 buildworld - make installworld - zfs list <--- should end up in a SIGABRT and a coredump. Cheers, László
I can confirm the broken zfs/zpool on my system: gucky% uname -a FreeBSD gucky 11.0-STABLE FreeBSD 11.0-STABLE #1 r308107: Sun Oct 30 19:48:10 CET 2016 mathiasp@gucky:/usr/obj/usr/src/sys/GUCKY amd64 gucky% zpool internal error: failed to initialize ZFS library gucky% zfs internal error: failed to initialize ZFS library But I could boot just fine. Lucky me ;)
Okay, the fault with the rebooting is on my behalf, I forgot the 'da' driver from the kernel config, thus the kernel couldn't find my ada drives. After adding that, the reboot worked. And, after reboot, the zfs binaries work again. I don't know if this is still considered a bug, but before the reboot, the problem persists. Someone close this bug if it deems to be closed, I'll just leave it open.
And I was obviously too tired to properly check what I was doing. Everything is working fine on my end.