I'm using Linux kernel 2.4.10, and since the fatidic 2.4.11 release ( i
tried 2.4.11 (one day :)))) 2.4.12, 2.4.13 and 2.4.14-pre6) I get the
same bug on and on (that means I can reproduce the experience and obtain
the same results).
Procedure
In X windows (version 4.1.0 compiled from the sources) when writing
"exit" in xterm to close the terminal emulator, the window freezes, and
from that moment on, every process becomes "unkillable", including xterm
and X (ps also freezes), and there's no way to shutdown GNU/Linux in a
sane way (must hit reset or poweroff).
Environment
I used Glibc 2.2.4 and GCC 3.0.1 (tried with 2.95.3, obtained the same
results).
The odd thing is, that with the same configuration, kernel 2.4.10 works
just fine, but every other release since then ends up doing the same
thing (the system can't maintain integrity after writing "exit" and
hiting enter in xterm).
Please help me, I getting slightly mad with the situation.
Ricardo Martins
Ricardo Martins wrote:
> Procedure
> In X windows (version 4.1.0 compiled from the sources) when writing
> "exit" in xterm to close the terminal emulator, the window freezes, and
> from that moment on, every process becomes "unkillable", including xterm
> and X (ps also freezes), and there's no way to shutdown GNU/Linux in a
> sane way (must hit reset or poweroff).
I can see the problem here with 2.4.13. I don't know if it's kernel
related, I'm used using rxvt, never xterm.
It looks like xterm takes the terminal where you started X from.
Are you using devfs ?
Pierre
--
------------------------------------------------
Pierre Rousselet <[email protected]>
------------------------------------------------
>> Procedure
>> In X windows (version 4.1.0 compiled from the sources) when writing
>> "exit" in xterm to close the terminal emulator, the window freezes, and
>> from that moment on, every process becomes "unkillable", including
xterm
>> and X (ps also freezes), and there's no way to shutdown GNU/Linux in a
>> sane way (must hit reset or poweroff).
>I can see the problem here with 2.4.13. I don't know if it's kernel
>related, I'm used using rxvt, never xterm.
>It looks like xterm takes the terminal where you started X from.
>Are you using devfs ?
>Pierre
Pierre, yes, i'm using devfs that seems to be the problem, do you know
how to fix it ?
Ricardo Martins wrote:
> >> Procedure
> >> In X windows (version 4.1.0 compiled from the sources) when writing
> >> "exit" in xterm to close the terminal emulator, the window freezes, and
> >> from that moment on, every process becomes "unkillable", including
> xterm
> >> and X (ps also freezes), and there's no way to shutdown GNU/Linux in a
> >> sane way (must hit reset or poweroff).
>
>
> >I can see the problem here with 2.4.13. I don't know if it's kernel
> >related, I'm used using rxvt, never xterm.
>
> >It looks like xterm takes the terminal where you started X from.
>
> >Are you using devfs ?
>
>
> >Pierre
>
>
> Pierre, yes, i'm using devfs that seems to be the problem, do you know
> how to fix it ?
Is it devfs or xterm which needs to be fixed ? I would
suggest to switch to rxvt which works fine with/without devfs.
Pierre
--
------------------------------------------------
Pierre Rousselet <[email protected]>
------------------------------------------------
On Thursday 01 November 2001 13:31, Pierre Rousselet wrote:
> Ricardo Martins wrote:
> > >> Procedure
> > >> In X windows (version 4.1.0 compiled from the sources) when writing
> > >> "exit" in xterm to close the terminal emulator, the window freezes,
> > >> and from that moment on, every process becomes "unkillable",
> > >> including
> >
> > xterm
> >
> > >> and X (ps also freezes), and there's no way to shutdown GNU/Linux in
> > >> a sane way (must hit reset or poweroff).
> > >
> > >I can see the problem here with 2.4.13. I don't know if it's kernel
> > >related, I'm used using rxvt, never xterm.
> > >
> > >It looks like xterm takes the terminal where you started X from.
> > >
> > >Are you using devfs ?
> > >
> > >
> > >Pierre
> >
> > Pierre, yes, i'm using devfs that seems to be the problem, do you know
> > how to fix it ?
>
> Is it devfs or xterm which needs to be fixed ? I would
>
> suggest to switch to rxvt which works fine with/without devfs.
With all due respect, I'd have to differ..... Do you have any idea how many
people are running how many copies of xterm as we speak? Even if we pair
that down to all those running devfs, it's certainly a substantial number.
The kernel should, above all else, run old applications without breaking them.
-Nick
Pierre, thanks for the advice.
If the problem is with xterm, it sure kicks "Linux Stability" in the
face. Maybe (and I hope) the problem is in devfs.
Ricardo Martins
On Thu, 1 Nov 2001, Ricardo Martins wrote:
[...]
> If the problem is with xterm, it sure kicks "Linux Stability" in the
> face. Maybe (and I hope) the problem is in devfs.
xterm is not the problem, it's devfs' locking ema which results in a
deadlock.
/Per
Nick LeRoy wrote:
> On Thursday 01 November 2001 13:31, Pierre Rousselet wrote:
>
>>Ricardo Martins wrote:
>>
>>> >> Procedure
>>> >> In X windows (version 4.1.0 compiled from the sources) when writing
>>> >> "exit" in xterm to close the terminal emulator, the window freezes,
>>> >> and from that moment on, every process becomes "unkillable",
>>> >> including
>>>
>>>xterm
>>>
>>> >> and X (ps also freezes), and there's no way to shutdown GNU/Linux in
>>> >> a sane way (must hit reset or poweroff).
>>> >
>>> >I can see the problem here with 2.4.13. I don't know if it's kernel
>>> >related, I'm used using rxvt, never xterm.
>>> >
>>> >It looks like xterm takes the terminal where you started X from.
>>> >
>>> >Are you using devfs ?
>>> >
>>> >
>>> >Pierre
>>>
>>>Pierre, yes, i'm using devfs that seems to be the problem, do you know
>>>how to fix it ?
>>>
>>Is it devfs or xterm which needs to be fixed ? I would
>>
>>suggest to switch to rxvt which works fine with/without devfs.
>>
>
> With all due respect, I'd have to differ..... Do you have any idea how many
> people are running how many copies of xterm as we speak? Even if we pair
> that down to all those running devfs, it's certainly a substantial number.
> The kernel should, above all else, run old applications without breaking them.
>
> -Nick
devfs is still marked EXPERIMENTAL in the kernel building. If you select it you
must be prepared to tolerate some misbehaviour. rxvt is not newer than
xterm, it is lighter.
Pierre
--
------------------------------------------------
Pierre Rousselet <[email protected]>
------------------------------------------------
On Thursday 01 November 2001 14:00, Pierre Rousselet wrote:
> Nick LeRoy wrote:
> > On Thursday 01 November 2001 13:31, Pierre Rousselet wrote:
> >>Ricardo Martins wrote:
> >>> >> Procedure
> >>> >> In X windows (version 4.1.0 compiled from the sources) when writing
> >>> >> "exit" in xterm to close the terminal emulator, the window freezes,
> >>> >> and from that moment on, every process becomes "unkillable",
> >>> >> including
> >>>
> >>>xterm
> >>>
> >>> >> and X (ps also freezes), and there's no way to shutdown GNU/Linux in
> >>> >> a sane way (must hit reset or poweroff).
> >>> >
> >>> >I can see the problem here with 2.4.13. I don't know if it's kernel
> >>> >related, I'm used using rxvt, never xterm.
> >>> >
> >>> >It looks like xterm takes the terminal where you started X from.
> >>> >
> >>> >Are you using devfs ?
> >>> >
> >>> >
> >>> >Pierre
> >>>
> >>>Pierre, yes, i'm using devfs that seems to be the problem, do you know
> >>>how to fix it ?
> >>
> >>Is it devfs or xterm which needs to be fixed ? I would
> >>
> >>suggest to switch to rxvt which works fine with/without devfs.
> >
> > With all due respect, I'd have to differ..... Do you have any idea how
> > many people are running how many copies of xterm as we speak? Even if we
> > pair that down to all those running devfs, it's certainly a substantial
> > number. The kernel should, above all else, run old applications without
> > breaking them.
> >
> > -Nick
>
> devfs is still marked EXPERIMENTAL in the kernel building. If you select it
> you
>
> must be prepared to tolerate some misbehaviour. rxvt is not newer than
> xterm, it is lighter.
Marked experiment, for now. What about when it's no longer "experimental"?
Configuring a kernel to enable such a feature should *not* break
applications, especially something as prolific as xterm.
-Nick
On Thu, 1 Nov 2001, Nick LeRoy wrote:
> Marked experiment, for now. What about when it's no longer "experimental"?
> Configuring a kernel to enable such a feature should *not* break
> applications, especially something as prolific as xterm.
Are you sure you know what you are talking about? Devfs causes this
problem because of a defect, not by design. It is marked experimental
because it's loaded with such defects. Don't use it until the
experimental tag is removed, if you are not prepared for some malfunction.
-jwb
On Thursday 01 November 2001 14:13, you wrote:
> On Thu, 1 Nov 2001, Nick LeRoy wrote:
> > Marked experiment, for now. What about when it's no longer
> > "experimental"? Configuring a kernel to enable such a feature should
> > *not* break applications, especially something as prolific as xterm.
>
> Are you sure you know what you are talking about? Devfs causes this
> problem because of a defect, not by design. It is marked experimental
> because it's loaded with such defects. Don't use it until the
> experimental tag is removed, if you are not prepared for some malfunction.
Yeah, I think that I know what I'm talking about. The question was: Should
devfs be fixed, or should xterm be fixed. I don't know how serious it is, or
exactly what the nature of the problem is (haven't followed the thread that
closely), but, from the "mile high" point of view, this defect, be it design
or just a one-line bug, needs to be fixed before it can be tagged
"non-experimental". I don't understand why people would think otherwise, to
be honest.
-Nick
On Thu, 1 Nov 2001, Nick LeRoy wrote:
> Yeah, I think that I know what I'm talking about. The question
> was: Should devfs be fixed, or should xterm be fixed.
If any random malicious user can crash the machine through
devfs, I think the answer to this question is quite obvious.
The security hole should be fixed.
Rik
--
DMCA, SSSCA, W3C? Who cares? http://thefreeworld.net/ (volunteers needed)
http://www.surriel.com/ http://distro.conectiva.com/
Nick LeRoy wrote:
>
> On Thursday 01 November 2001 14:13, you wrote:
> > On Thu, 1 Nov 2001, Nick LeRoy wrote:
> > > Marked experiment, for now. What about when it's no longer
> > > "experimental"? Configuring a kernel to enable such a feature should
> > > *not* break applications, especially something as prolific as xterm.
> >
> > Are you sure you know what you are talking about? Devfs causes this
> > problem because of a defect, not by design. It is marked experimental
> > because it's loaded with such defects. Don't use it until the
> > experimental tag is removed, if you are not prepared for some malfunction.
>
> Yeah, I think that I know what I'm talking about. The question was: Should
> devfs be fixed, or should xterm be fixed. I don't know how serious it is, or
> exactly what the nature of the problem is (haven't followed the thread that
> closely), but, from the "mile high" point of view, this defect, be it design
> or just a one-line bug, needs to be fixed before it can be tagged
> "non-experimental". I don't understand why people would think otherwise, to
> be honest.
Fix devfs. If the kernel lets a user program crash it, it's a kernel
bug.
--
Brian Gerst
Rik van Riel writes:
> On Thu, 1 Nov 2001, Nick LeRoy wrote:
>
> > Yeah, I think that I know what I'm talking about. The question
> > was: Should devfs be fixed, or should xterm be fixed.
>
> If any random malicious user can crash the machine through
> devfs, I think the answer to this question is quite obvious.
>
> The security hole should be fixed.
Agreed. I will.
Regards,
Richard....
Permanent: [email protected]
Current: [email protected]
You should censor those words ;) You wouldn't want us Americans knowing
about it.
"The <censored> should be fixed."
David
Rik van Riel wrote:
>On Thu, 1 Nov 2001, Nick LeRoy wrote:
>
>>Yeah, I think that I know what I'm talking about. The question
>>was: Should devfs be fixed, or should xterm be fixed.
>>
>
>If any random malicious user can crash the machine through
>devfs, I think the answer to this question is quite obvious.
>
>The security hole should be fixed.
>
>Rik
>
Hi,
I see something similar with vanilla 2.4.13 (+devfs), except my
unkillable xterms don't appear until a modprobe has already failed.
Specifically, what seems to happen is:
- an open() call causes the kernel to grab the devfs rwsem for reading,
and then load a module (e.g. ide-cd).
- the modprobe process then waits forever for write-access to the devfs
rwsem.
Each xterm then waits forever in "wait_for_devfsd_finished()",
presumably when it tries to close its terminal.
Are you also using dynamic module loading? Are ALL of your unkillable
processes xterms, or do you also have a failed modprobe lurking somewhere?
Chris
On Thu, 1 Nov 2001, Nick LeRoy wrote:
> Marked experiment, for now. What about when it's no longer "experimental"?
> Configuring a kernel to enable such a feature should *not* break
> applications, especially something as prolific as xterm.
Well, we would hope it's fixed by then, wouldn't we? How about you fix it
and send us the patch? :)
Kelsey Hudson [email protected]
Software Engineer
Compendium Technologies, Inc (619) 725-0771
---------------------------------------------------------------------------