mq_open() and close_on_exec?

mq_open() and close_on_exec?

[Chris Wright]

SUSv3 doesn't seem to specify one way or the other.  I don't have the
POSIX specs, and the old docs I have suggest that mq_open() creates an
object which is to be closed upon exec.  Anyone have a clue if this is
actually required?  Patch below sets this as default (if indeed it's
valid/required).

thanks,
-chris
-- 
Linux Security Modules     http://lsm.immunix.org     http://lsm.bkbits.net

===== ipc/mqueue.c 1.6 vs edited =====
--- 1.6/ipc/mqueue.c	Mon Apr 12 10:54:17 2004
+++ edited/ipc/mqueue.c	Tue Apr 13 16:05:36 2004
@@ -641,6 +641,7 @@
 		goto out_putfd;
 	}
 
+	set_close_on_exec(fd, 1);
 	fd_install(fd, filp);
 	goto out_upsem;
 

Re: mq_open() and close_on_exec?

[Manfred Spraul]

Chris Wright wrote:

>SUSv3 doesn't seem to specify one way or the other.  I don't have the
>POSIX specs, and the old docs I have suggest that mq_open() creates an
>object which is to be closed upon exec.  Anyone have a clue if this is
>actually required?  Patch below sets this as default (if indeed it's
>valid/required).
>  
>
Did you test what other unices do? I think the patch is correct - at 
least solaris implements message queues in user space, and then an exec 
should close everything.

--
    Manfred

Re: mq_open() and close_on_exec?

[Jakub Jelinek]

On Tue, Apr 13, 2004 at 05:40:05PM -0700, Chris Wright wrote:
> SUSv3 doesn't seem to specify one way or the other.  I don't have the
> POSIX specs, and the old docs I have suggest that mq_open() creates an
> object which is to be closed upon exec.  Anyone have a clue if this is
> actually required?  Patch below sets this as default (if indeed it's
> valid/required).

I think it is valid and required:
http://www.opengroup.org/onlinepubs/007904975/functions/exec.html

^[MSG] [Option Start] All open message queue descriptors in the calling process shall be closed, as described in
mq_close() . [Option End]

I'll add a new test for this into glibc testsuite.

	Jakub

Re: mq_open() and close_on_exec?

[Chris Wright]

* Manfred Spraul (manfred@colorfullife.com) wrote:
> Chris Wright wrote:
> >SUSv3 doesn't seem to specify one way or the other.  I don't have the
> >POSIX specs, and the old docs I have suggest that mq_open() creates an
> >object which is to be closed upon exec.  Anyone have a clue if this is
> >actually required?  Patch below sets this as default (if indeed it's
> >valid/required).
> >
> Did you test what other unices do? I think the patch is correct - at 
> least solaris implements message queues in user space, and then an exec 
> should close everything.

No, I haven't.  Looks like I missed the SUSv3 specification that indeed
requires closing on exec.  So, I believe the patch is needed.

thanks,
-chris
-- 
Linux Security Modules     http://lsm.immunix.org     http://lsm.bkbits.net

Re: mq_open() and close_on_exec?

[Chris Wright]

* Jakub Jelinek (jakub@redhat.com) wrote:
> I think it is valid and required:
> http://www.opengroup.org/onlinepubs/007904975/functions/exec.html
> 
> ^[MSG] [Option Start] All open message queue descriptors in the calling process shall be closed, as described in
> mq_close() . [Option End]
> 
> I'll add a new test for this into glibc testsuite.

Thanks, I managed to miss that in the spec.

thanks,
-chris
-- 
Linux Security Modules     http://lsm.immunix.org     http://lsm.bkbits.net