sshd session hang, bad file descriptor data

["John D. Baker" <jdbaker%mylinuxisp.com@localhost>]

Attempts to connect to my Yeeloong via ssh from an i386 client still
hang.  I'm not getting any console messages, but that may be a side
effect from my 'amd' investigation.

The symptoms from the client end remain the same.  The server prompts
for password and no further operations are possible on the client end.
the client terminal is locked into an unresponsive network connection.
The usual "~." doesn't work.  In a graphical environment, one must
destroy the window.  (I haven't tried from a non-graphical client, but
I expect one must log in to the client from elsewhere and kill the
frozen session.)

Below are what may be relevant snippets of ktruss output generated with
'sudo ktruss -i -p <pid> | tee sshd.ktruss'.  In this instance, the PID
of the master sshd process was 630.  There were no messages relevant
to this logged in /var/log/messages.

The output is some different than before.  Before, I saw a distinct
pattern of user-after-close.  This time, an attempt is made to close
a file descriptor of -1 (0xffffffff)


   630      1 sshd     __select50                  = 1
   630      1 sshd     emul(netbsd32)
   630      1 sshd     netbsd32_accept(0x4, 0x7fff67f8, 0x7fff6730) = 5
   630      1 sshd     netbsd32_fcntl(0x5, 0x3, 0) = 6
   630      1 sshd     netbsd32_fcntl(0x5, 0x4, 0x2) = 0, 2015257344
   630      1 sshd     pipe()                      = 7, 8
   630      1 sshd     netbsd32_socketpair(0x1, 0x1, 0, 0x7fff6738) = 0
   630      1 sshd     fork()                      = 1061

  1061      1 sshd     fork                        = 0
  1061      1 sshd     emul(netbsd32)

  1061      1 sshd     netbsd32___sigaction_sigtramp(0x14, 0x7fff6530, 
0x7fff6548, 0x7811ea10, 0x2) = 0, 2015257344
  1061      1 sshd     netbsd32_kill(0x129, 0xf)   Err#3 ESRCH
  1061      1 sshd     netbsd32_close(0x4)         = 0
  1061      1 sshd     netbsd32_close(0xffffffff)  Err#9 EBADF
  1061      1 sshd     netbsd32_write(0x7, 0x7fff6580, 0x5) = 5, 2013921600
       "\0\0\0\^A;"

  1061      1 sshd     netbsd32_read(0x7, 0x77e6d000, 0x59b) = 1435, 268823216
       "\^X\0\0\0 
C\M-5\M-L\M-7\M-X\M-e\M-bS\M-Pl\^YZ\M^@\M-%\^S\M^\\M-H\M-/<3A\^_"
  1061      1 sshd     netbsd32_close(0x7)         = 0, 7
  1061      1 sshd     netbsd32_close(0xffffffff)  Err#9 EBADF
  1061      1 sshd     netbsd32_mmap(0, 0x140000, 0x3, 0x1001, 0xffffffff, 0, 
0) = 0x77b20000
  1061      1 sshd     netbsd32_munmap(0x77fb0000, 0x10000) = 0


I killed the client window and tried again.  This time there are
messages about "tcgetattr: Bad file descriptor", but their PIDs don't
match the recorded session.  I'll have to try again later and see if
anything is different.

--
|/"\ 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