Re: Recent -current cpucontrol panic: [cpuctl, 129]: cannot bind to target cpu 1

On Sun, May 22, 2016 at 12:19:32PM +0300, Andrey Chernov wrote:
> On 22.05.2016 10:14, Konstantin Belousov wrote:
> > On Sun, May 22, 2016 at 09:47:08AM +0300, Andrey Chernov wrote:
> >> With microcode_update_enable="YES" in rc.conf:
> >>
> >> db:0:kdb.enter.panic>  show pcpu
> >> cpuid        = 1
> >> dynamic pcpu = 0xfffffe02ac1fd300
> >> curthread    = 0xfffff8000ae75a00: pid 736 "cpucontrol"
> >> curpcb       = 0xfffffe023754eb80
> >> fpcurthread  = none
> >> idlethread   = 0xfffff800022b5000: tid 100003 "idle: cpu1"
> >> curpmap      = 0xfffff8000af02138
> >> tssp         = 0xffffffff80c1cf78
> >> commontssp   = 0xffffffff80c1cf78
> >> rsp0         = 0xfffffe023754eb80
> >> gs32p        = 0xffffffff80c237d0
> >> ldt          = 0xffffffff80c23810
> >> tss          = 0xffffffff80c23800
> >> db:0:kdb.enter.panic>  bt
> >> Tracing pid 736 tid 100121 td 0xfffff8000ae75a00
> >> kdb_enter() at kdb_enter+0x3b/frame 0xfffffe023754e620
> >> vpanic() at vpanic+0x19f/frame 0xfffffe023754e6a0
> >> kassert_panic() at kassert_panic+0x126/frame 0xfffffe023754e710
> >> cpuctl_do_cpuid_count() at cpuctl_do_cpuid_count+0x85/frame
> >> 0xfffffe023754e750
> >> cpuctl_ioctl() at cpuctl_ioctl+0x389/frame 0xfffffe023754e800
> >> devfs_ioctl_f() at devfs_ioctl_f+0x156/frame 0xfffffe023754e860
> >> kern_ioctl() at kern_ioctl+0x204/frame 0xfffffe023754e8c0
> >> sys_ioctl() at sys_ioctl+0x171/frame 0xfffffe023754e9a0
> >> amd64_syscall() at amd64_syscall+0x28e/frame 0xfffffe023754eab0
> >> Xfast_syscall() at Xfast_syscall+0xfb/frame 0xfffffe023754eab0
> >> --- syscall (54, FreeBSD ELF64, sys_ioctl), rip = 0x8009778ea, rsp =
> >> 0x7fffffffe788, rbp = 0x7fffffffe7d0 ---
> > 
> > Show verbose dmesg of your boot.
> 
> Attached.
Thanks.

> 
> > What scheduler do you use ?
> 
> ULE
> 
> > In the fired KASSERT, add printout of td->td_oncpu and show the updated
> > panic message.
> 
> I add, but can't reproduce this bug again so far.
> 
Do you have core file for the panic ?  If yes, it would be at least
interesting to see the output of 'p *td' in kgdb.