SUMMARY: Trap from _XentMM, locore.s (1296) in Xdec, twice in 48h

From: Irene A. Shilikhina <irene_at_alpha.iae.nsk.su>
Date: Wed, 13 May 1998 10:30:29 +0700 (NSD)

Hello managers,
this is Dr. Tom Blinn, 603-884-0646 <tpb_at_zk3.dec.com> who explained my
problem and offered its solution. Thanks, Dr. Blinn, I'll follow your advice.
Since his reply contains my original questions too, I reduce my summary to
forwarding it,
Irene
*****************************************************************************
> Dear managers,
> we experienced crash twice in less than 48h. The crash data look similar.
> I can suppose it's a well known error which is already fixed in the
> system. The current process is Xdec.
> It's important to notice that in the second case nobody was logged from
> the console, we have DEC 2000 model 300, version 3.2c + enhanced security.
> Following is a fragment from the crash-data file.
> I'm anxious also: couldn't it be a form of an attack? And what does "Invalid
> character in input" mean in this case?

It's nearly impossible to tell what "Invalid character in input" means,
since you did not provide ALL of the context.

However, the fact that it appears Xdec was running, together with the fact
that it was doing "gpc" input, makes me suspect that either the keyboard or
the mouse on your system is starting to malfunction. If it were my system
I would consider replacing the keyboard, it generates input, and you seem
to have had an invalid character in input in a routine that's trying to get
input from the "console" device:

> > 0 boot(0x0, 0x0, 0xfffffc00005fe4b0, 0x30, 0x0)
> 1 panic(s = 0xfffffc00005fe4b0 = "kernel memory fault")
> 2 trap() ["../../../../src/kernel/arch/alpha/trap.c":1281,
> 3 _XentMM(0x4, 0xfffffc0000563c80, 0xfffffc0000631200, 0x4,
> 4 wsselect(0x1, 0xffffffff8e6b75cc, 0xffffffff8e6b75ce, 0x1, 0x3)
> 5 gpcselect(0x1, 0xffffffff8e6b75cc, 0xffffffff8e6b75ce, 0x1, 0x0)
> 6 abc_cnselect(0x1, 0xffffffff8e6b75cc, 0xffffffff8e6b75ce, 0x1, 0x1)
> 7 cnselect(0xffffffff8e6b75ce, 0x1, 0x1, 0xffffffff8e6b75cc,
> 8 spec_select(0xffffffff8c557c00, 0xffffffff8e6b75cc,
> 9 vn_select(0xffffffff00000004, 0xffffffff8e6b75c4,
> 10 selscan(0xfffffc000024df80, 0xffffffff8e6b76e8, 0xffffffff8e6b73f0,
> 11 do_scan(0xfffffc00002539cc, 0x0, 0x6, 0xffffffff8c5f4300,
> 12 select(0xffffffff85d34d90, 0xffffffff8e6b78c8, 0xffffffff8e6b78b8,
> 13 syscall(0x3ff81f86dc4, 0x8, 0xfffffc000070cb18, 0x51, 0x5d)
> 14 _Xsyscall(0x8, 0x3ff800d24b8, 0x3ffc089b630, 0x80, 0x3ffc0894990)

Something did a syscall, presumably to get input, judging by where it got to
in the kernel, and while it was in the "wsselect" routine (which is found in
the kernel/io/dec/ws/ws_device.c source module, it's general purpose code to
deal with workstation specific devices) which it got to through some code to
deal with the DEC 2000 system console (in the abc_cnselect routine in the
kernel/arch/alpha/hal/dec2000_cons.c module). So it was clearly trying to
do some sort of keyboard or mouse input, and that's when it died. That's
why I suspect you've got bad hardware, which would fit with the message:

> Invalid character in input

that you ask about. Unless you can call failing hardware a form of attack,
you don't have an attack, but I think you have a bad keyboard or mouse.

Tom
 
Received on Wed May 13 1998 - 05:32:08 NZST

This archive was generated by hypermail 2.4.0 : Wed Nov 08 2023 - 11:53:37 NZDT