NetBSD Problem Report #9306

Received: (qmail 14180 invoked from network); 27 Jan 2000 20:28:56 -0000
Message-Id: <200001272028.PAA05076@kolvir.arlington.ma.us>
Date: Thu, 27 Jan 2000 15:28:55 -0500 (EST)
From: John Kohl <jtk@kolvir.arlington.ma.us>
Reply-To: jtk@kolvir.arlington.ma.us
To: gnats-bugs@gnats.netbsd.org
Subject: options XSERVER_DDB has no effect on wscons
X-Send-Pr-Version: 3.95

>Number:         9306
>Category:       kern
>Synopsis:       options XSERVER_DDB has no effect on wscons
>Confidential:   no
>Severity:       non-critical
>Priority:       medium
>Responsible:    mycroft
>State:          closed
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Thu Jan 27 12:30:00 +0000 2000
>Closed-Date:    Tue Jul 06 22:55:35 +0000 2004
>Last-Modified:  Tue Jul 06 22:55:35 +0000 2004
>Originator:     John Kohl
>Release:        NetBSD-current, 2000-01-26
>Organization:
NetBSD Kernel Hackers `R` Us
>Environment:

System: NetBSD kolvir.arlington.ma.us 1.4M NetBSD 1.4M (KOLVIR) #14: Sun Oct 31 09:36:01 EST 1999 jtk@kolvir.arlington.ma.us:/usr/u4/sandbox/src/sys/arch/i386/compile/KOLVIR i386


>Description:
	the wscons driver doesn't care about XSERVER_DDB.  It appears
that (for i386 port) only the pccons driver cares.

>How-To-Repeat:
	Try using F12 on an X server
>Fix:
	Implement XSERVER_DDB handling for wscons
>Release-Note:
>Audit-Trail:
State-Changed-From-To: open->feedback 
State-Changed-By: jhawk 
State-Changed-When: Mon May 22 10:28:05 PDT 2000 
State-Changed-Why:  
It's not clear to me that XSERVER_DDB is supposed to work on wscons; 
XSERVER is a pccons/pcvt -specific option, and XSERVER_DDB presumably 
is as well; it's certainly not documented. 

Do you actually find XSERVER_DDB to be a useful thing? I have not found 
it to be so; if you do, please let me know why you think so ;-) 

Per the discussion on port-i386 or current-users a few months ago, 
I think the correct thing, absent a mechanism for switching the video 
display back to text or a vga implementation of ddb, is to just panic 
and dump as normal. Do you disagree? 

--jhawk 


Responsible-Changed-From-To: kern-bug-people->jhawk 
Responsible-Changed-By: jhawk 
Responsible-Changed-When: Mon May 22 10:28:05 PDT 2000 
Responsible-Changed-Why:  
Mmm, ddb bugs. Tasty. 

From: John Kohl <jtk@kolvir.arlington.ma.us>
To: jhawk@netbsd.org, gnats-bugs@netbsd.org
Cc:  
Subject: Re: kern/9306: options XSERVER_DDB has no effect on wscons
Date: Mon, 22 May 2000 20:48:14 -0400 (EDT)

 I would say that if we want to support XSERVER_DDB, it should be
 supported on all console video drivers.  XSERVER_DDB could be useful to
 force a crash dump (via blind typing) when the X server has lost it and
 can't be terminated/switched away from.

 Maybe another option (forcibly crash and dump?) would be more useful in its
 place, however.

 -- 
 ==John Kohl <jtk@kolvir.arlington.ma.us>, <john_kohl@alum.mit.edu>
 Home page: <http://people.ne.mediaone.net/jtk/>
 Bicycling and Skiing to keep fit.

From: John Hawkinson <jhawk@MIT.EDU>
To: John Kohl <jtk@kolvir.arlington.ma.us>
Cc: gnats-bugs@netbsd.org
Subject: Re: kern/9306: options XSERVER_DDB has no effect on wscons 
Date: Wed, 24 May 2000 01:27:59 -0400

 I note you didn't rigorously answer my questions.

 In message <200005230048.UAA17254@kolvir.arlington.ma.us>, John Kohl writes:
 >I would say that if we want to support XSERVER_DDB,

 I don't think we do.

 >it should be supported on all console video drivers.

 And it is not.

 >XSERVER_DDB could be useful to force a crash dump (via blind typing)
 >when the X server has lost it and can't be terminated/switched away
 >from.
 >
 >Maybe another option (forcibly crash and dump?) would be more useful in its
 >place, however.

 I don't see any need for an option -- as far as I can tell the
 correct implementation is for the kernel to always skip ddb
 if it cannot switch to it in a useful video mode. I think the
 pain this saves vastly outweights the rare case where you
 want to actually type blind at ddb to do something other than
 dump.


 Does this seem reasonable?

 --jhawk

From: John Kohl <jtk@kolvir.arlington.ma.us>
To: jhawk@MIT.EDU
Cc: gnats-bugs@netbsd.org
Subject: Re: kern/9306: options XSERVER_DDB has no effect on wscons
Date: Wed, 24 May 2000 07:38:49 -0400 (EDT)

 You've still missed the point.  If the system is running but the X
 server is hosed, I cannot force a crash dump right now.
 That used to be possible with XSERVER_DDB.

 Do you have a suggested replacement for this function?  (If we don't
 want a generic "go to debugger" sequence, as you suggest by nuking
 XSERVER_DDB, then I'd suggest some hot key sequence, valid in all
 keyboard input modes, which forces a crash dump.)

 -- 
 ==John Kohl <jtk@kolvir.arlington.ma.us>, <john_kohl@alum.mit.edu>
 Home page: <http://people.ne.mediaone.net/jtk/>
 Bicycling and Skiing to keep fit.

From: John Hawkinson <jhawk@MIT.EDU>
To: John Kohl <jtk@kolvir.arlington.ma.us>
Cc: gnats-bugs@netbsd.org
Subject: Re: kern/9306: options XSERVER_DDB has no effect on wscons 
Date: Wed, 24 May 2000 11:03:58 -0400

 In message <200005241138.HAA22940@kolvir.arlington.ma.us>, John Kohl writes:
 >You've still missed the point.  If the system is running but the X
 >server is hosed, I cannot force a crash dump right now.
 >That used to be possible with XSERVER_DDB.

 I understand that this is currently broken.

 >
 >Do you have a suggested replacement for this function?

 Yes -- I intend to add support to wscons such that
 if you are running X on the current display, DDB
 will be skipped and the kernel will panic and crash
 and dump. The precise implementation is not yet 
 clear to me, though.

 Does this meet your requirements?

 >(If we don't want a generic "go to debugger" sequence, as you suggest
 >by nuking XSERVER_DDB, then I'd suggest some hot key sequence, valid
 >in all keyboard input modes, which forces a crash dump.)

 There's no framework for writing generic key sequences that work
 across all console drivers.

 I don't see any cases where the "force the machine to dump" sequence
 actually adds any functionality over the "dump automatically if we
 panic under X" mechanism. I'm trying to get you to tell me what those
 cases are so I can evaluate their reasonability.

 --jhawk

From: John Kohl <jtk@kolvir.arlington.ma.us>
To: jhawk@MIT.EDU
Cc: gnats-bugs@netbsd.org
Subject: Re: kern/9306: options XSERVER_DDB has no effect on wscons
Date: Wed, 24 May 2000 15:48:59 -0400 (EDT)

 >>>>> "JH" == John Hawkinson <jhawk@MIT.EDU> writes:

 JH> In message <200005241138.HAA22940@kolvir.arlington.ma.us>, John Kohl writes:
 >> You've still missed the point.  If the system is running but the X
 >> server is hosed, I cannot force a crash dump right now.
 >> That used to be possible with XSERVER_DDB.

 JH> I understand that this is currently broken.

 JH> ...

 JH> I don't see any cases where the "force the machine to dump" sequence
 JH> actually adds any functionality over the "dump automatically if we
 JH> panic under X" mechanism. I'm trying to get you to tell me what those
 JH> cases are so I can evaluate their reasonability.

 Read paragraph 1 and use just a tiny bit of imagination.  System gets
 hosed but DOES NOT PANIC.  X controls the keyboard and won't relinquish
 the console.  User wants to force crash dump.  How does he/she do it?

 -- 
 ==John Kohl <jtk@kolvir.arlington.ma.us>, <john_kohl@alum.mit.edu>
 Home page: <http://people.ne.mediaone.net/jtk/>
 Bicycling and Skiing to keep fit.

From: John Hawkinson <jhawk@MIT.EDU>
To: John Kohl <jtk@kolvir.arlington.ma.us>
Cc: gnats-bugs@netbsd.org
Subject: Re: kern/9306: options XSERVER_DDB has no effect on wscons
Date: Wed, 24 May 2000 15:58:38 -0400

 >JH> I don't see any cases where the "force the machine to dump" sequence
 >JH> actually adds any functionality over the "dump automatically if we
 >JH> panic under X" mechanism. I'm trying to get you to tell me what those
 >JH> cases are so I can evaluate their reasonability.
 >
 >Read paragraph 1 and use just a tiny bit of imagination.

 I'm trying to avoid using my imagination right now, to avoid consing
 up corner-cases that are not critical path.

 >System gets hosed but DOES NOT PANIC.  X controls the keyboard and
 >won't relinquish the console.  User wants to force crash dump.  How
 >does he/she do it?

 My judgement is that this case is sufficiently weird/rare that
 it does not merit requiring users to type blind at DDB
 for hte more common case of a panic under X.

 But I'm not sure I understand your question. X controls the keyboard
 and you cannot switch displays for some reason -- why would that happen?

 Can you type in an xterm? Presumably you just do the moral equivalent of:

     echo set hz=0 ; gdb -wk /netbsd /dev/mem

 --jhawk

From: John Kohl <jtk@kolvir.arlington.ma.us>
To: jhawk@MIT.EDU
Cc: gnats-bugs@netbsd.org
Subject: Re: kern/9306: options XSERVER_DDB has no effect on wscons
Date: Wed, 24 May 2000 16:15:22 -0400 (EDT)

 >>>>> "JH" == John Hawkinson <jhawk@MIT.EDU> writes:

 >> System gets hosed but DOES NOT PANIC.  X controls the keyboard and
 >> won't relinquish the console.  User wants to force crash dump.  How
 >> does he/she do it?

 JH> My judgement is that this case is sufficiently weird/rare that
 JH> it does not merit requiring users to type blind at DDB
 JH> for hte more common case of a panic under X.

 I actually would tend to agree here.  I think XSERVER_DDB is probably
 the wrong thing in such a case; I'd rather have a key sequence/combo
 that forces a crash dump (optionally compiled in, of course).

 JH> But I'm not sure I understand your question. X controls the keyboard
 JH> and you cannot switch displays for some reason -- why would that happen?

 Oh, say, bugs in the X server that cause it to loop in some other code,
 not reading the keyboard.  IIRC, it was something like this that got me
 irked about no XSERVER_DDB when I submitted the PR.

 -- 
 ==John Kohl <jtk@kolvir.arlington.ma.us>, <john_kohl@alum.mit.edu>
 Home page: <http://people.ne.mediaone.net/jtk/>
 Bicycling and Skiing to keep fit.

From: "Erik E. Fair" <fair@clock.org>
To: NetBSD GNATS Problem Report Tracking System <gnats-bugs@gnats.netbsd.org>
Cc:  
Subject: Re: kern/9306
Date: Sun, 6 Jan 2002 12:58:53 -0800

 Can this PR be closed?

 	Erik <fair@clock.org>
State-Changed-From-To: feedback->open 
State-Changed-By: perry 
State-Changed-When: Fri Apr 4 13:14:53 PST 2003 
State-Changed-Why:  
The problem is obviously not fixed in -current, so we are not 
waiting for feedback. 

From: "Perry E. Metzger" <perry@piermont.com>
To: John Hawkinson <jhawk@MIT.EDU>
Cc: jtk@kolvir.arlington.ma.us, gnats-bugs@gnats.netbsd.org
Subject: Re: kern/9306
Date: 04 Apr 2003 16:44:58 -0500

 jhawk dixit:
 > That's not correct. I assert "this is not a bug" it's in feedback
 > waiting for jtk to concur/disagree.

 I agree with jtk. This is pretty clearly a deficit in our system. It
 is unreasonable to have no way to force a core dump from a screwed up
 machine that has not paniced and is hung.

 I've personally wanted such a facility twice in just in the last
 month. Machine hangs, there is no way to find out what happened.

 I see your statements that the problem is not real, but I disagree
 with them.

 Perry
State-Changed-From-To: open->closed 
State-Changed-By: mycroft 
State-Changed-When: Tue Jul 6 22:49:46 UTC 2004 
State-Changed-Why:  
I've removed XSERVER and XSERVER_DDB from all config files.  It will be 
removed as an option completely when pccons dies. 


Responsible-Changed-From-To: jhawk->mycroft 
Responsible-Changed-By: mycroft 
Responsible-Changed-When: Tue Jul 6 22:49:46 UTC 2004 
Responsible-Changed-Why:  
. 
>Unformatted:

NetBSD Home
NetBSD PR Database Search

(Contact us) $NetBSD: query-full-pr,v 1.51 2026/08/10 02:28:17 riastradh Exp $
$NetBSD: gnats_config.sh,v 1.10 2026/05/13 22:00:09 riastradh Exp $
Copyright © 1994-2026 The NetBSD Foundation, Inc. ALL RIGHTS RESERVED.