Re: xend gets SIGBUS and cores under -current

[Sarton O'Brien <bsd-xen%roguewrt.org@localhost>]

On 22/05/2009 3:56 PM, Sarton O'Brien wrote:
> Built and running as of Fri May 22 11:29:03 EST 2009, all seems to be
> fine. I haven't updated userland ... yet. uname and package list below.
>
> Just before updating I did see some vnds stay tied up when domu failed
> to honor 'shutdown -r now', shutting down completely and then being
> unable to start them back up. Bit of a pain ... I'll check if I
> replicate the problem with current.

I updated userland and various problems appeared. No domu started but:

# ps ax|grep xen
 241 ?       I    0:00.36 xenstored --pid-file /var/run/xenstore.pid
 243 ?       Il   0:00.00 xenconsoled
 255 ?       I    0:00.12 /usr/pkg/bin/python2.5 /usr/pkg/sbin/xend start
 274 ?       Isl  0:00.01 /usr/pkg/sbin/xenbackendd
 718 ?       Il   0:00.42 /usr/pkg/bin/python2.5 /usr/pkg/sbin/xend start
 304 ttyp1   S+   0:00.00 grep xen
 732 xencons Is+  0:00.00 /usr/libexec/getty Pc console

# xm list
Xend has probably crashed!  Invalid or missing HTTP status code.

Then 'vnconfig -l' responded with rvnd0 device not configured.

I ran the new MAKEDEV (yes the new one that build.sh neglects ;) over 
everything (-f, all, xen, vnd0) and then 'vnconfig -l' hung:

# ktruss vnconfig -l
  1610      1 ktruss   emul(netbsd)
  1610      1 ktruss   fktrace                     = 0
  1610      1 ktruss   fcntl(0x4, 0x3, 0)          = 1
  1610      1 ktruss   fcntl(0x4, 0x4, 0x1)        = 0
  1610      1 ktruss   execve("/sbin/vnconfig", 0x7f7fffffd910, 
0x7f7fffffd928) Err#2 ENOENT
  1610      1 vnconfig emul(netbsd)
  1610      1 vnconfig execve("/usr/sbin/vnconfig", 0x7f7fffffd910, 
0x7f7fffffd928) JUSTRETURN
  1610      1 vnconfig mmap(0, 0x8000, 0x3, 0x1002, 0xffffffff, 0, 0) = 
0x7f7ffdff8000
  1610      1 vnconfig open("/etc/ld.so.conf", 0, 0) = 3
  1610      1 vnconfig read(0x3, 0x7f7fffffd7a0, 0x80) = 0
       ""
  1610      1 vnconfig close(0x3)                  = 0
  1610      1 vnconfig open("/usr/lib/libutil.so.7", 0, 0x6f6900) = 3
  1610      1 vnconfig __fstat50(0x3, 0x7f7fffffd050) = 0
  1610      1 vnconfig mmap(0, 0x1000, 0x1, 0x1, 0x3, 0, 0) = 
0x7f7ffdff7000
  1610      1 vnconfig munmap(0x7f7ffdff7000, 0x1000) = 0
  1610      1 vnconfig mmap(0, 0x117000, 0x5, 0x14000002, 0x3, 0, 0) = 
0x7f7ffdc00000
  1610      1 vnconfig mmap(0x7f7ffdd13000, 0x3000, 0x3, 0x12, 0x3, 0, 
0x13000) = 0x7f7ffdd13000
  1610      1 vnconfig mmap(0x7f7ffdd16000, 0x1000, 0x3, 0x1012, 
0xffffffff, 0, 0) = 0x7f7ffdd16000
  1610      1 vnconfig mprotect(0x7f7ffdc13000, 0x100000, 0) = 0
  1610      1 vnconfig close(0x3)                  = 0
  1610      1 vnconfig open("/usr/lib/libc.so.12", 0, 0) = 3
  1610      1 vnconfig __fstat50(0x3, 0x7f7fffffd050) = 0
  1610      1 vnconfig mmap(0, 0x1000, 0x1, 0x1, 0x3, 0, 0) = 
0x7f7ffdff7000
  1610      1 vnconfig munmap(0x7f7ffdff7000, 0x1000) = 0
  1610      1 vnconfig mmap(0, 0x11e000, 0x5, 0x14000002, 0x3, 0, 0) = 
0x7f7ffda00000
  1610      1 vnconfig mmap(0x7f7ffdb00000, 0xc000, 0x3, 0x12, 0x3, 0, 
0x100000) = 0x7f7ffdb00000
  1610      1 vnconfig mmap(0x7f7ffdb0c000, 0x12000, 0x3, 0x1012, 
0xffffffff, 0, 0) = 0x7f7ffdb0c000
  1610      1 vnconfig close(0x3)                  = 0
  1610      1 vnconfig open("/lib/libc.so.12", 0, 0x61746164655f00) = 3
  1610      1 vnconfig __fstat50(0x3, 0x7f7fffffd050) = 0
  1610      1 vnconfig close(0x3)                  = 0
  1610      1 vnconfig __sysctl(0x7f7fffffd7f0, 0x2, 0x7f7ffdb18a20, 
0x7f7fffffd7e8, 0, 0) = 0
  1610      1 vnconfig __sysctl(0x7f7fffffd7b0, 0x2, 0x7f7fffffd7bc, 
0x7f7fffffd7a8, 0, 0) = 0
  1610      1 vnconfig open("vnd0", 0, 0x3)        Err#2 ENOENT
  1610      1 vnconfig open("vnd0d", 0, 0)         Err#2 ENOENT
  1610      1 vnconfig open("/dev/rvnd0", 0, 0)    Err#2 ENOENT

So I'm guessing vnd/xen related? It ran fine up until the new userland 
and prior to that I was seeing vnd wierdness.

Rebooting into generic all seems fine from the get go:

# vnconfig -l
vnd0: not in use
vnd1: not in use
vnd2: not in use
vnd3: not in use
# vnconfig vnd0 /usr/pkg/etc/xen/spike-vnd.img
# vnconfig -l
vnd0: /usr (/dev/raid0e) inode 11043390
# mount -o log /dev/vnd0a /mnt

> NetBSD gogeta.internal 5.99.11 NetBSD 5.99.11 (XEN3_DOM0) #0: Fri May 22
> 15:09:04 EST 2009
> root%spike.internal@localhost:/usr/obj/sys/arch/amd64/compile/XEN3_DOM0 amd64
>
> xenkernel33-3.3.1nb1 Xen 3.3.1 Kernel
> xentools33-3.3.1nb6 Userland Tools for Xen 3.3.x

Sarton