Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 18 Apr 1997 04:22:51 +1000
From:      Bruce Evans <bde@zeta.org.au>
To:        cshenton@it.hq.nasa.gov, hackers@FreeBSD.ORG
Subject:   Re: 2.2.1-R crash with looped serial gettys
Message-ID:  <199704171822.EAA09021@godzilla.zeta.org.au>

next in thread | raw e-mail | index | archive | help
>I've built a little box for pppd/mpd and routing out of a 486sx33 and
>a whopping 8MB RAM; seems to mostly run fine with 2.2.1-RELEASE. It's
>a fairly minimal install (Kernel User option I think) and I rebuilt
>the kernel to fit the hardware.  I've added a dual 16550A serial port
>card to which I'll connect a pair of modems. I have getty running on
>the ports, cuaa1 and cuaa2.
>
>PROBLEM:
>
>If I connect a null-modem cable between the ports, the system crashes
>within a couple seconds.  It's repeatable.
>
>It leaves no message in /var/log/messages, or dmesg, and there's no
>core file.
>
>I presume looping the serial gettys is causing them to scream at each
>other "Login:"; "Invalid Login:"; no *you* login; dammit, `login' is
>not a valid login... and consumes too many resources.

4 problems:

1. Yes there is a getty war (actually probably a login war) that consumes
   too many resources (at 115200 bps, about 100% of a 486/33 for echoing
   and receiving echoes alone according to my estimates.  On a 486SX/33
   it is unlikely that the system runs any cycles in user mode once the
   war starts.  57600 bps might work).
2. There is a timeout leak in siopoll().  Oops.  This causes a panic
   instead of a hang.  With this fixed, the hang is fixable by unplugging
   the cable.
3. You should have seen a panic message about (2).
4. This problem should have been reported to freebsd-bugs :-).

Bruce



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199704171822.EAA09021>