[uml-devel] unkillable UML threads

[uml-devel] unkillable UML threads

[Michael Richardson]

FYI: This is without the skas patch, with a stock RH9 kernel on the host.

Any thoughts or coroboration would be appreciated. Clearly this is a
bug in the RH9 kernel. 

I'm not sure the revision.


Date: Tue, 23 Sep 2003 03:11:57 -0400 (EDT)
From: "D. Hugh Redelmeier" <hugh@mimosa.com>
Reply-To: "D. Hugh Redelmeier" <hugh@mimosa.com>
To: "Michael C. Richardson" <mcr@sandelman.ottawa.on.ca>
cc: FreeS/WAN Design <design@lists.freeswan.org>
Subject: Re: [design] test failures in HEAD
In-Reply-To: <Pine.LNX.4.44.0309220422270.26952-100000@redshift.mimosa.com>
Message-ID: <Pine.LNX.4.44.0309230304340.26952-100000@redshift.mimosa.com>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII

-----BEGIN PGP SIGNED MESSAGE-----


I've made some progress with make check on Red Hat Linux 9, but I'm at
a problem that looks bad.

I have some UML kernel processes appear to be unkillable.  They don't
respond to kill -9 by me, nor to kill -9 by the superuser.  They are
not zombies.

Here's one that I"ve attacked:

F   UID   PID  PPID PRI  NI   VSZ  RSS WCHAN  STAT TTY        TIME COMMAND
1   104  7622     1  15   0 33404 7200 get_si T    ?          0:00 /home/hugh/swan/umls/uml2.4.22-sept/umlbuild/east/linux (east) [lwdnsq]                                                                                        !

As far as I know, that means we have a RHL9 kernel bug.  No process
should be unkillable.

Second opinion, anyone?

Hugh Redelmeier
hugh@mimosa.com  voice: +1 416 482-8253

-----BEGIN PGP SIGNATURE-----
Version: 2.6.3ia
Charset: noconv

iQCVAwUBP2/yQsFAuQPManGZAQG3AQP+M2Qja5DZ+o5+JBAFJnQ+22jGsdxP/GnM
0nJIeKo0lbvRimCnCtM7VaXz4GGZmqyAzwXS4g4iOEXrNuv67hLUpdlPqj9/Qm5w
Wk3aXo1F/lDNiwQ+xG3+fjt3dWMTXvGE0wqXp2iRQNIciBtipymUqsEM/0Bpxwc1
cbKmbnG3ELk=
=jpnE
-----END PGP SIGNATURE-----



-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] unkillable UML threads

[Matt Zimmerman]

On Wed, Sep 24, 2003 at 08:40:42AM -0400, Michael Richardson wrote:

> 
> FYI: This is without the skas patch, with a stock RH9 kernel on the host.
> 
> Any thoughts or coroboration would be appreciated. Clearly this is a
> bug in the RH9 kernel. 

No, it's what we call "suspended".  It can't die until it's continued.

It is presumably a bug in UML that it leaves these suspended processes
behind (happens to me sometimes also).

-- 
 - mdz


-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] unkillable UML threads

[Michael Richardson]

-----BEGIN PGP SIGNED MESSAGE-----


>>>>> "Matt" == Matt Zimmerman <mdz@debian.org> writes:
    >> FYI: This is without the skas patch, with a stock RH9 kernel on the host.
    >> 
    >> Any thoughts or coroboration would be appreciated. Clearly this is a
    >> bug in the RH9 kernel. 

    Matt> No, it's what we call "suspended".  It can't die until it's
    Matt> continued. 

  kill -9 should always be able to get rid of any process.
  If not, it is due to a kernel bug. It sounds like the -9 handler needs
to "continue" it first, and I see no reason why not.

  Otherwise, you are running windows!

  It may *also* be a bug in UML that it doesn't clean up nicely. It may not
always get a chance.

] Train travel features AC outlets with no take-off restrictions|  firewalls  [
]   Michael Richardson, Sandelman Software Works, Ottawa, ON    |net architect[
] mcr@sandelman.ottawa.on.ca http://www.sandelman.ottawa.on.ca/ |device driver[
] panic("Just another Debian/notebook using, kernel hacking, security guy");  [


-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.2 (GNU/Linux)
Comment: Finger me for keys - custom hacks make this fully PGP2 compat

iQCVAwUBP3L+bYqHRg3pndX9AQFhyAP/QgBis8ToxrwMPlyvFWpxRd8nU/BeM7fI
jDqP37d8s9KCpQ83GR8J07TrAh8ckKY87QLyNyJXEWKTKiEwAOAfY3PEGkf0ZUgc
YUnbHK03WdDR5+AjMhlHTBhVuZOk4A9vIdeTsb0qv8FLkAvcoVAwqj5DtfYMsQjR
wND3psLQAqs=
=eJDY
-----END PGP SIGNATURE-----


-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] unkillable UML threads

[Jeff Dike]

mcr@sandelman.ottawa.on.ca said:
> FYI: This is without the skas patch, with a stock RH9 kernel on the
> host.
>
> Any thoughts or coroboration would be appreciated. Clearly this is a
> bug in the RH9 kernel.  

It could easily be a UML bug.  I do have code which continues threads
so that they die properly, but it seems not to be happening here.  Can
you tell me how to reproduce it?

				Jeff



-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] unkillable UML threads

[Michael Richardson]

-----BEGIN PGP SIGNED MESSAGE-----


>>>>> "Jeff" == Jeff Dike <jdike@addtoit.com> writes:
    Jeff> mcr@sandelman.ottawa.on.ca said:
    >> FYI: This is without the skas patch, with a stock RH9 kernel on the
    >> host.
    >> 
    >> Any thoughts or coroboration would be appreciated. Clearly this is a
    >> bug in the RH9 kernel.

    Jeff> It could easily be a UML bug.  I do have code which continues

  Again, if any process becomes unkillable on *nix system, there is at
*least* a bug in the kernel. This should never be the case. 

  yes, there is likely also a bug in UML.

    Jeff> threads so that they die properly, but it seems not to be happening
    Jeff> here.  Can you tell me how to reproduce it?

  I don't have a formula. Hugh sees is pretty much at least once if not more
often in regression testing of FreeS/WAN. 

  Btw, I experience segment faults with 2.4.22-7 on a 2.4.20 host with skas
patch. Again, not trivially reproducible, but they do occur
frequently. Enough that since this is isn't RH9, I went back to 2.4.19-47,
which was pretty solid. I haven't tried things in between, but I could.

] Train travel features AC outlets with no take-off restrictions|  firewalls  [
]   Michael Richardson, Sandelman Software Works, Ottawa, ON    |net architect[
] mcr@sandelman.ottawa.on.ca http://www.sandelman.ottawa.on.ca/ |device driver[
] panic("Just another Debian/notebook using, kernel hacking, security guy");  [

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.2 (GNU/Linux)
Comment: Finger me for keys - custom hacks make this fully PGP2 compat

iQCVAwUBP4ISMIqHRg3pndX9AQGFzAP8C6tutx3wvBYK5qKIlk9AgZ6J+HdcDtiG
C9/vLNtoBnMhU8t4i0W+gFEGvTHbL6uMi+dyuLMx4eKA4OvmJHRQVYid/aoe5P33
4tD8MhfWDqGgFGBj8m1gYX2v9psUo0PUwK6gBtmI6uu9DKyRi6bCYMqn0JXqFk+p
Tknwh/R0uMo=
=h6vh
-----END PGP SIGNATURE-----


-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] unkillable UML threads

[Jeff Dike]

mcr@sandelman.ottawa.on.ca said:
> Again, if any process becomes unkillable on *nix system, there is at
> *least* a bug in the kernel.

No.  At least not in this case, since the process is quite killable.  It
just needs a SIGCONT before it actually dies.

				Jeff



-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] unkillable UML threads

[Michael Richardson]

-----BEGIN PGP SIGNED MESSAGE-----


>>>>> "Jeff" == Jeff Dike <jdike@addtoit.com> writes:
    >> Again, if any process becomes unkillable on *nix system, there is at
    >> *least* a bug in the kernel.

    Jeff> No.  At least not in this case, since the process is quite
    Jeff> killable.  It just needs a SIGCONT before it actually dies.

  I'll try that, thank you for the suggestion.
  However -9 isn't supposed to care. 

] Train travel features AC outlets with no take-off restrictions|  firewalls  [
]   Michael Richardson, Sandelman Software Works, Ottawa, ON    |net architect[
] mcr@sandelman.ottawa.on.ca http://www.sandelman.ottawa.on.ca/ |device driver[
] panic("Just another Debian/notebook using, kernel hacking, security guy");  [


-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.2 (GNU/Linux)
Comment: Finger me for keys - custom hacks make this fully PGP2 compat

iQCVAwUBP4NOwYqHRg3pndX9AQHCyAP/Q8BTqBg1xhRZXOdT68DC/PGJjhD7MBXJ
j8w4HlroIVsZeolUKyxV5IKQZECyug7Y44F9vtrgiE/IqGace/7850qaWix1ZSRl
6ZglJ7JR1QGNe3V+ZJAkE9m9ePwsRu4qOGW9IFDWzb8rf18lIgM2HghgO1id/dad
uQZNeAfRo5c=
=pMFi
-----END PGP SIGNATURE-----


-------------------------------------------------------
This SF.net email is sponsored by: SF.net Giveback Program.
SourceForge.net hosts over 70,000 Open Source Projects.
See the people who have HELPED US provide better services:
Click here: http://sourceforge.net/supporters.php
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] unkillable UML threads

[Matt Zimmerman]

On Mon, Oct 06, 2003 at 05:18:54PM -0400, Jeff Dike wrote:

> mcr@sandelman.ottawa.on.ca said:
> > FYI: This is without the skas patch, with a stock RH9 kernel on the
> > host.
> >
> > Any thoughts or coroboration would be appreciated. Clearly this is a
> > bug in the RH9 kernel.  
> 
> It could easily be a UML bug.  I do have code which continues threads
> so that they die properly, but it seems not to be happening here.  Can
> you tell me how to reproduce it?

I know of no reliable way to reproduce it, but I commonly do things like
"killall linux" and end up with a suspended process left behind.  Maybe it
has something to do with multiple threads receiving the signal?  I think it
has also happened when killing only the first process, but I can't test
right now.

-- 
 - mdz


-------------------------------------------------------
This SF.net email is sponsored by: SF.net Giveback Program.
SourceForge.net hosts over 70,000 Open Source Projects.
See the people who have HELPED US provide better services:
Click here: http://sourceforge.net/supporters.php
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] unkillable UML threads

[Jan Hudec]

On Sat, Oct 11, 2003 at 00:15:47 -0400, Matt Zimmerman wrote:
> On Mon, Oct 06, 2003 at 05:18:54PM -0400, Jeff Dike wrote:
> 
> > mcr@sandelman.ottawa.on.ca said:
> > > FYI: This is without the skas patch, with a stock RH9 kernel on the
> > > host.
> > >
> > > Any thoughts or coroboration would be appreciated. Clearly this is a
> > > bug in the RH9 kernel.  
> > 
> > It could easily be a UML bug.  I do have code which continues threads
> > so that they die properly, but it seems not to be happening here.  Can
> > you tell me how to reproduce it?
> 
> I know of no reliable way to reproduce it, but I commonly do things like
> "killall linux" and end up with a suspended process left behind.  Maybe it
> has something to do with multiple threads receiving the signal?  I think it
> has also happened when killing only the first process, but I can't test
> right now.

The way out is to also kill with SIGCONT signal. Since umlinux uses
tracing a lot, some processes are often stopped. When you kill the
tracer while it's victim is stopped, it's not awoken. So you have to do
it yourself.

-------------------------------------------------------------------------------
						 Jan 'Bulb' Hudec <bulb@ucw.cz>


-------------------------------------------------------
This SF.net email is sponsored by: SF.net Giveback Program.
SourceForge.net hosts over 70,000 Open Source Projects.
See the people who have HELPED US provide better services:
Click here: http://sourceforge.net/supporters.php
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] unkillable UML threads

[Matt Zimmerman]

On Thu, Oct 16, 2003 at 10:15:10PM +0200, Jan Hudec wrote:

> On Sat, Oct 11, 2003 at 00:15:47 -0400, Matt Zimmerman wrote:
> > On Mon, Oct 06, 2003 at 05:18:54PM -0400, Jeff Dike wrote:
> > > It could easily be a UML bug.  I do have code which continues threads
> > > so that they die properly, but it seems not to be happening here.  Can
> > > you tell me how to reproduce it?
> > 
> > I know of no reliable way to reproduce it, but I commonly do things like
> > "killall linux" and end up with a suspended process left behind.  Maybe it
> > has something to do with multiple threads receiving the signal?  I think it
> > has also happened when killing only the first process, but I can't test
> > right now.
> 
> The way out is to also kill with SIGCONT signal. Since umlinux uses
> tracing a lot, some processes are often stopped. When you kill the
> tracer while it's victim is stopped, it's not awoken. So you have to do
> it yourself.

Yes, I understand how signals work, thank you.  Jeff asked how to reproduce
the situation where a suspended process is left behind, because that should
not happen.

-- 
 - mdz


-------------------------------------------------------
This SF.net email is sponsored by: SF.net Giveback Program.
SourceForge.net hosts over 70,000 Open Source Projects.
See the people who have HELPED US provide better services:
Click here: http://sourceforge.net/supporters.php
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel