Re: Lemote Yeeloong observations

On Thu, 10 May 2012 18:42:28 +0000 (UTC), christos%astron.com@localhost
(Christos Zoulas) wrote:

>In article <Pine.NEB.4.64.1205092317360.6809%david.technoskunk.fur@localhost>,
>John D. Baker <jdbaker%mylinuxisp.com@localhost> wrote:
>
>>kernel) hangs at the client end and produces the following on console:
>>
>>   sshd[pid]: error: do_exec_pty: Dup #1: Bad file descriptor
>
> You don't need PTYFS if your kernel has COMPAT_BSDPTY.
> ktrace it and see what it is trying to dup

Trying to attach 'sudo ktrace -i -p <pid>' exited immediately, leaving
an empty trace file.  Using 'ktruss' instead and 'tee'ing its output
to a file (-o outfile caused immediate exit as well) produced the
following output:

   http://bobdbob.com/~jdbaker/nbsd-debug/sshd.ktruss.gz

which indicates that one of the child processes forked to handle the
connection obtains a file descriptor via netbsd32___socket(),
netbsd32_connect()s with it, does a netbsd32_sendto() on it,
netbsd32_close()s it _and then attempts to perform netbsd32_ioctl() on it
it_, resulting in EBADF.  It further attempts to netbsd32_dup() it,
again resulting in EBADF.  Possible compiler pessimization reorder some
calls?

>>   ntpd[pid]: step-systime: Invalid argument
>>
>>Not sure what this indicates.  Is 'clockctl' not included?  I'll have
>>to check the kernel config.
>
> Something is wrong with the ioctl? ktrace it.

Likewise, produced:

   http://bobdbob.com/~jdbaker/nbsd-debug/ntpd.ktruss.gz

Searching for EINVAL revealed calls to netbsd32_ioctl() with commands
"CLOCKCTL_CLOCK_SETTIME",  "CLOCKCTL_SETTIMEOFDAY", and "_IOWR" as
failing in this way.  A few lines earlier, ioctl() calls with command
"CLOCKCTL_NTP_ADJTIME" fails with EFAULT.  Alignment issues?

-- 
|/"\ John D. Baker, KN5UKS               NetBSD     Darwin/MacOS X
|\ / jdbaker[snail]mylinuxisp[flyspeck]com    OpenBSD            FreeBSD
| X  No HTML/proprietary data in email.   BSD just sits there and works!
|/ \ GPGkeyID:  D703 4A7E 479F 63F8 D3F4  BD99 9572 8F23 E4AD 1645