[Bug 231117] I/O lockups inside bhyve vms

[bugzilla-noreply at freebsd.org]

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=231117

--- Comment #12 from Kristian K. Christensen <kkc at duckpond.dk> ---
(In reply to Kristian K. Christensen from comment #11)

So, I can reproduce the issue on a test vm, regretably I can't get ps output or
run procstat, they both hang.

On the hypervisor, I get the following from procstat -kk on the bhyve pid:
machiavelli# procstat -kk 35743
  PID    TID COMM                TDNAME              KSTACK                     
35743 100908 bhyve               mevent              mi_switch+0xe6
sleepq_catch_signals+0x40c sleepq_wait_sig+0xf _sleep+0x231 kqueue_kevent+0x290
kern_kevent+0xb5 sys_kevent+0xa4 amd64_syscall+0xa38 fast_syscall_common+0x101 
35743 101987 bhyve               vtnet-2:0 tx        mi_switch+0xe6
sleepq_catch_signals+0x40c sleepq_wait_sig+0xf _sleep+0x231 umtxq_sleep+0x143
do_wait+0x427 __umtx_op_wait_uint_private+0x53 amd64_syscall+0xa38
fast_syscall_common+0x101 
35743 101988 bhyve               blk-3:0-0           mi_switch+0xe6
sleepq_catch_signals+0x40c sleepq_wait_sig+0xf _sleep+0x231 umtxq_sleep+0x143
do_wait+0x427 __umtx_op_wait_uint_private+0x53 amd64_syscall+0xa38
fast_syscall_common+0x101 
35743 101989 bhyve               blk-3:0-1           mi_switch+0xe6
sleepq_catch_signals+0x40c sleepq_wait_sig+0xf _sleep+0x231 umtxq_sleep+0x143
do_wait+0x427 __umtx_op_wait_uint_private+0x53 amd64_syscall+0xa38
fast_syscall_common+0x101 
35743 101990 bhyve               blk-3:0-2           mi_switch+0xe6
sleepq_catch_signals+0x40c sleepq_wait_sig+0xf _sleep+0x231 umtxq_sleep+0x143
do_wait+0x427 __umtx_op_wait_uint_private+0x53 amd64_syscall+0xa38
fast_syscall_common+0x101 
35743 101991 bhyve               blk-3:0-3           mi_switch+0xe6
sleepq_catch_signals+0x40c sleepq_wait_sig+0xf _sleep+0x231 umtxq_sleep+0x143
do_wait+0x427 __umtx_op_wait_uint_private+0x53 amd64_syscall+0xa38
fast_syscall_common+0x101 
35743 101992 bhyve               blk-3:0-4           mi_switch+0xe6
sleepq_catch_signals+0x40c sleepq_wait_sig+0xf _sleep+0x231 umtxq_sleep+0x143
do_wait+0x427 __umtx_op_wait_uint_private+0x53 amd64_syscall+0xa38
fast_syscall_common+0x101 
35743 101993 bhyve               blk-3:0-5           mi_switch+0xe6
sleepq_catch_signals+0x40c sleepq_wait_sig+0xf _sleep+0x231 umtxq_sleep+0x143
do_wait+0x427 __umtx_op_wait_uint_private+0x53 amd64_syscall+0xa38
fast_syscall_common+0x101 
35743 101994 bhyve               blk-3:0-6           mi_switch+0xe6
sleepq_catch_signals+0x40c sleepq_wait_sig+0xf _sleep+0x231 umtxq_sleep+0x143
do_wait+0x427 __umtx_op_wait_uint_private+0x53 amd64_syscall+0xa38
fast_syscall_common+0x101 
35743 101995 bhyve               blk-3:0-7           mi_switch+0xe6
sleepq_catch_signals+0x40c sleepq_wait_sig+0xf _sleep+0x231 umtxq_sleep+0x143
do_wait+0x427 __umtx_op_wait_uint_private+0x53 amd64_syscall+0xa38
fast_syscall_common+0x101 
35743 101996 bhyve               vcpu 0              mi_switch+0xe6
sleepq_timedwait+0x2f msleep_spin_sbt+0x161 vm_run+0x7d7 vmmdev_ioctl+0x85f
devfs_ioctl_f+0x128 kern_ioctl+0x26d sys_ioctl+0x15c amd64_syscall+0xa38
fast_syscall_common+0x101 

The system isn't completely dead, but anything at all that requires io hangs
indefinitely. I can type on the console, provided it was logged in beforehand,
but trying to tab complete or run anything that's not in memory, hangs.

I'm going to try to run procstat and ps more frequently in cron, hoping to keep
them in memory. Any other suggestions as to how to provide you with any useful
input would be welcomed.

-- 
You are receiving this mail because:
You are the assignee for the bug.