Re: problem with printk on SMP-- somebody please help

Re: problem with printk on SMP-- somebody please help

[Arun Srinivas]

hi

I really appreciate your suggestions and as a newcomer eager to learn more 
from you people.

As I said I was able to do printk anywhere in the sched.c (including 
_activate_task ) on a non-smp kernel and on a smp-kernel I can do it only on 
the main schedule() function.
Also, I would like to add that I am not able to do the macro rdtsc() for 
reading the timestamp counter in the same function.When I compile the kernel 
it dosent show any error, but just the printk's and rdtsc()'s get subdued!

Well,  with reference to your reply, I have some basic questions:
1) on a non-smp kernel will the _activate_task not lock the given runqueue?
2) where is the best place I can do the rdtsc() and printk to read the value 
as to when a task is being scheduled for execution, on a SMP kernel?

Please help.

Thanks
Arun

>From: Mike Waychison <Michael.Waychison@Sun.COM>
>To: Arun Srinivas <getarunsri@hotmail.com>
>CC: linux-kernel@vger.kernel.org
>Subject: Re: problem with printk--  somebody please help
>Date: Mon, 08 Nov 2004 14:47:59 -0500
>
>-----BEGIN PGP SIGNED MESSAGE-----
>Hash: SHA1
>
>Arun Srinivas wrote:
> > hi
> >
> > I am new to the kernel world and would be very glad if somebody could
> > help me with this problem.
> >
> > I am unable to do printk or even a macro call like rdtsc()...(for
> > reading the time stamp counter) from within the "activate_task"
> > function on a kernel with smp support.But these work under the main
> > schedule() function.
> >
> > I was able to do all these i.e., inside "activate_task" on a kernel
> > without smp support.Can anybody suggest a solution as to what could be
> > the problem??
> >
> > somebody please help.
>
>I seem to recall that doing so would deadlock your machine.
>activate_task is called with the given cpu's runqueue locked.  printk
>eventually calls release_console_sem, which will wake_up_interruptible,
>which will eventually call try_to_wake_up which grabs the same lock.
>
>I don't think I've ever seen printk work while task_rq_lock'ed.
>
>Someone correct me if I'm wrong, this is the best I was able to figure
>out before giving up.
>
>- --
>Mike Waychison
>Sun Microsystems, Inc.
>1 (650) 352-5299 voice
>1 (416) 202-8336 voice
>
>~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>NOTICE:  The opinions expressed in this email are held by me,
>and may not represent the views of Sun Microsystems, Inc.
>~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>-----BEGIN PGP SIGNATURE-----
>Version: GnuPG v1.2.5 (GNU/Linux)
>Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org
>
>iD8DBQFBj81vdQs4kOxk3/MRAlJGAKCIYIgXCkaSXpwGLdsj/WK1BhPOlwCeK6s0
>1pc0XbERKlQKpLIBpObhwZA=
>=2DSX
>-----END PGP SIGNATURE-----

_________________________________________________________________
Mergers, takeovers, buyouts. Get all the latest biz bytes. 
http://www.msn.co.in/business/ Tune in to MSN Business!

Re: problem with printk on SMP-- somebody please help

[Mike Waychison]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Arun Srinivas wrote:
> hi
> 
> I really appreciate your suggestions and as a newcomer eager to learn
> more from you people.
> 
> As I said I was able to do printk anywhere in the sched.c (including
> _activate_task ) on a non-smp kernel and on a smp-kernel I can do it
> only on the main schedule() function.
> Also, I would like to add that I am not able to do the macro rdtsc() for
> reading the timestamp counter in the same function.When I compile the
> kernel it dosent show any error, but just the printk's and rdtsc()'s get
> subdued!
> 
> Well,  with reference to your reply, I have some basic questions:
> 1) on a non-smp kernel will the _activate_task not lock the given runqueue?

The locking is done in schedule().  Read include/linux/spinlock.h to see
how spinlocks differ in the SMP vs non-SMP case.  In the latter case
with spinlock debugging enabled, there is no lock bit and recursively
grabbing a lock is possible (though still not allowed!).

> 2) where is the best place I can do the rdtsc() and printk to read the
> value as to when a task is being scheduled for execution, on a SMP kernel?
> 

No idea wrt to rdtsc.

HTH,

- --
Mike Waychison
Sun Microsystems, Inc.
1 (650) 352-5299 voice
1 (416) 202-8336 voice

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
NOTICE:  The opinions expressed in this email are held by me,
and may not represent the views of Sun Microsystems, Inc.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.5 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org

iD8DBQFBkNpAdQs4kOxk3/MRAkCCAJ9IoN7VfyymTBuYn7R8//dbxLGwkACfUTQu
l1BiHVTfZf6IvIT2+nqsiPE=
=loXh
-----END PGP SIGNATURE-----

Re: problem with printk on SMP-- somebody please help

[arun srinivasan]

Can I do a sched_clock() inside activate_task to record the time at
which a task is ready for execution and then can read it by a prink in
schedule()??
I think sched_clock() calls rdtsc() and converts it to nanosecond scale.
or is there any other better way of reading the time at which a task
is being executed?
 


On Tue, 09 Nov 2004 09:54:56 -0500, Mike Waychison
<michael.waychison@sun.com> wrote:
> 
> 
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> Arun Srinivas wrote:
> > hi
> >
> > I really appreciate your suggestions and as a newcomer eager to learn
> > more from you people.
> >
> > As I said I was able to do printk anywhere in the sched.c (including
> > _activate_task ) on a non-smp kernel and on a smp-kernel I can do it
> > only on the main schedule() function.
> > Also, I would like to add that I am not able to do the macro rdtsc() for
> > reading the timestamp counter in the same function.When I compile the
> > kernel it dosent show any error, but just the printk's and rdtsc()'s get
> > subdued!
> >
> > Well,  with reference to your reply, I have some basic questions:
> > 1) on a non-smp kernel will the _activate_task not lock the given runqueue?
> 
> The locking is done in schedule().  Read include/linux/spinlock.h to see
> how spinlocks differ in the SMP vs non-SMP case.  In the latter case
> with spinlock debugging enabled, there is no lock bit and recursively
> grabbing a lock is possible (though still not allowed!).
> 
> > 2) where is the best place I can do the rdtsc() and printk to read the
> > value as to when a task is being scheduled for execution, on a SMP kernel?
> >
> 
> No idea wrt to rdtsc.
> 
> HTH,
> 
> 
> 
> - --
> Mike Waychison
> Sun Microsystems, Inc.
> 1 (650) 352-5299 voice
> 1 (416) 202-8336 voice
> 
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> NOTICE:  The opinions expressed in this email are held by me,
> and may not represent the views of Sun Microsystems, Inc.
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.2.5 (GNU/Linux)
> Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org
> 
> iD8DBQFBkNpAdQs4kOxk3/MRAkCCAJ9IoN7VfyymTBuYn7R8//dbxLGwkACfUTQu
> l1BiHVTfZf6IvIT2+nqsiPE=
> =loXh
> -----END PGP SIGNATURE-----
> 
> 
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
>