[Xenomai-help] Exception handlers in primary domain / user-space signals

[Xenomai-help] Exception handlers in primary domain / user-space signals

[Henri Roosen]

We are using signal handlers for catching exceptions which our
application is allowed to make and which we know how to handle.

The current Xenomai implementation is to switch to the secondary
domain and call the handlers from there.
Unfortunately this takes too much time for our application and we
would like to handle the exception without the switch to the secondary
domain, in primary domain.

Can anyone give some advice how to implement that?
Will "user-space signals" which was planned for Xenomai 2.6 fulfill this need?
Is there already code available for user-space signals?

Thanks,
Henri.

Xenomai: 2.5.5.2
Linux-kernel: 2.6.32.15
Platform: x86

Re: [Xenomai-help] Exception handlers in primary domain / user-space signals

[Gilles Chanteperdrix]

Henri Roosen wrote:
> We are using signal handlers for catching exceptions which our
> application is allowed to make and which we know how to handle.
> 
> The current Xenomai implementation is to switch to the secondary
> domain and call the handlers from there.
> Unfortunately this takes too much time for our application and we
> would like to handle the exception without the switch to the secondary
> domain, in primary domain.
> 
> Can anyone give some advice how to implement that?
> Will "user-space signals" which was planned for Xenomai 2.6 fulfill this need?
> Is there already code available for user-space signals?

In the 2.5 series, we added some code to support signals. The signals
are multiplexed per-skin in kernel-space, and demultiplexed in
user-space, upon exit of system calls. We implemented a unit test of
this functionality with the "sigtest" skin and user-space test, but they
only work upon return from system calls.

Then we added support for the "mayday" page, which made us realize, that
maybe implementing signals handling at any time, not only when returning
from system calls, was possible. But then came the realization that in
order to implement that, we would have to fiddle with the FPU, which is
an area where we have a certain tradition for not getting the things
right at the first attempt. So, we kind of stopped here.

So, if you want some ad-hoc signals upon return from system call, the
task is pretty easy. If you want the full posix signals interface, then
things are going to be a bit harder.

-- 
                                                                Gilles.

Re: [Xenomai-help] Exception handlers in primary domain / user-space signals

[Henri Roosen]

On Mon, Feb 7, 2011 at 7:27 PM, Gilles Chanteperdrix
<gilles.chanteperdrix@xenomai.org> wrote:
> Henri Roosen wrote:
>> We are using signal handlers for catching exceptions which our
>> application is allowed to make and which we know how to handle.
>>
>> The current Xenomai implementation is to switch to the secondary
>> domain and call the handlers from there.
>> Unfortunately this takes too much time for our application and we
>> would like to handle the exception without the switch to the secondary
>> domain, in primary domain.
>>
>> Can anyone give some advice how to implement that?
>> Will "user-space signals" which was planned for Xenomai 2.6 fulfill this need?
>> Is there already code available for user-space signals?
>
> In the 2.5 series, we added some code to support signals. The signals
> are multiplexed per-skin in kernel-space, and demultiplexed in
> user-space, upon exit of system calls. We implemented a unit test of
> this functionality with the "sigtest" skin and user-space test, but they
> only work upon return from system calls.
>
> Then we added support for the "mayday" page, which made us realize, that
> maybe implementing signals handling at any time, not only when returning
> from system calls, was possible. But then came the realization that in
> order to implement that, we would have to fiddle with the FPU, which is
> an area where we have a certain tradition for not getting the things
> right at the first attempt. So, we kind of stopped here.
>
> So, if you want some ad-hoc signals upon return from system call, the
> task is pretty easy. If you want the full posix signals interface, then
> things are going to be a bit harder.
>
I am afraid it's going to be a bit harder; we would need it when the
exception occurs and that is in most cases not at a place in the code
where there is a system call :-(.

I was thinking of adding a hook in Xenomai's exception handler before
it makes the switch to the secondary domain... but that would of
course be a very ugly hack and I don't know if it can be done. Do you
have a suggestion?

What are the plans with the full posix signals interface?

Thanks,
> --
>                                                                Gilles.
>

Re: [Xenomai-help] Exception handlers in primary domain / user-space signals

[Gilles Chanteperdrix]

Henri Roosen wrote:
> On Mon, Feb 7, 2011 at 7:27 PM, Gilles Chanteperdrix
> <gilles.chanteperdrix@xenomai.org> wrote:
>> Henri Roosen wrote:
>>> We are using signal handlers for catching exceptions which our
>>> application is allowed to make and which we know how to handle.
>>>
>>> The current Xenomai implementation is to switch to the secondary
>>> domain and call the handlers from there.
>>> Unfortunately this takes too much time for our application and we
>>> would like to handle the exception without the switch to the secondary
>>> domain, in primary domain.
>>>
>>> Can anyone give some advice how to implement that?
>>> Will "user-space signals" which was planned for Xenomai 2.6 fulfill this need?
>>> Is there already code available for user-space signals?
>> In the 2.5 series, we added some code to support signals. The signals
>> are multiplexed per-skin in kernel-space, and demultiplexed in
>> user-space, upon exit of system calls. We implemented a unit test of
>> this functionality with the "sigtest" skin and user-space test, but they
>> only work upon return from system calls.
>>
>> Then we added support for the "mayday" page, which made us realize, that
>> maybe implementing signals handling at any time, not only when returning
>> from system calls, was possible. But then came the realization that in
>> order to implement that, we would have to fiddle with the FPU, which is
>> an area where we have a certain tradition for not getting the things
>> right at the first attempt. So, we kind of stopped here.
>>
>> So, if you want some ad-hoc signals upon return from system call, the
>> task is pretty easy. If you want the full posix signals interface, then
>> things are going to be a bit harder.
>>
> I am afraid it's going to be a bit harder; we would need it when the
> exception occurs and that is in most cases not at a place in the code
> where there is a system call :-(.

What kind of exception is it? Could not the exception be signalled at
the next system call?

> 
> I was thinking of adding a hook in Xenomai's exception handler before
> it makes the switch to the secondary domain... but that would of
> course be a very ugly hack and I don't know if it can be done. Do you
> have a suggestion?
> 
> What are the plans with the full posix signals interface?

Plans were to get it during the 2.6 branch, but of course if someone is
able to contribute it before, there is no problem.

> 
> Thanks,
>> --
>>                                                                Gilles.
>>
> 


-- 
                                                                Gilles.

Re: [Xenomai-help] Exception handlers in primary domain / user-space signals

[Henri Roosen]

On Mon, Feb 7, 2011 at 8:08 PM, Gilles Chanteperdrix
<gilles.chanteperdrix@xenomai.org> wrote:
> Henri Roosen wrote:
>> On Mon, Feb 7, 2011 at 7:27 PM, Gilles Chanteperdrix
>> <gilles.chanteperdrix@xenomai.org> wrote:
>>> Henri Roosen wrote:
>>>> We are using signal handlers for catching exceptions which our
>>>> application is allowed to make and which we know how to handle.
>>>>
>>>> The current Xenomai implementation is to switch to the secondary
>>>> domain and call the handlers from there.
>>>> Unfortunately this takes too much time for our application and we
>>>> would like to handle the exception without the switch to the secondary
>>>> domain, in primary domain.
>>>>
>>>> Can anyone give some advice how to implement that?
>>>> Will "user-space signals" which was planned for Xenomai 2.6 fulfill this need?
>>>> Is there already code available for user-space signals?
>>> In the 2.5 series, we added some code to support signals. The signals
>>> are multiplexed per-skin in kernel-space, and demultiplexed in
>>> user-space, upon exit of system calls. We implemented a unit test of
>>> this functionality with the "sigtest" skin and user-space test, but they
>>> only work upon return from system calls.
>>>
>>> Then we added support for the "mayday" page, which made us realize, that
>>> maybe implementing signals handling at any time, not only when returning
>>> from system calls, was possible. But then came the realization that in
>>> order to implement that, we would have to fiddle with the FPU, which is
>>> an area where we have a certain tradition for not getting the things
>>> right at the first attempt. So, we kind of stopped here.
>>>
>>> So, if you want some ad-hoc signals upon return from system call, the
>>> task is pretty easy. If you want the full posix signals interface, then
>>> things are going to be a bit harder.
>>>
>> I am afraid it's going to be a bit harder; we would need it when the
>> exception occurs and that is in most cases not at a place in the code
>> where there is a system call :-(.
>
> What kind of exception is it? Could not the exception be signalled at
> the next system call?

Our customers provide the application code, we provide more or less
the framework. Customers can install exception handlers for for
instance floating point exceptions (SIGFPE).
Besides that we provide a means of tracing the application code, which
is handled by breakpoints in the code which then does some bookkeeping
and lets the task run again. Of course that has some overhead also
when using our old OS, but Linux-Xenomai has so much overhead because
of the secondary domain switch. Therefore we would like to handle it
in primary domain.

>
>>
>> I was thinking of adding a hook in Xenomai's exception handler before
>> it makes the switch to the secondary domain... but that would of
>> course be a very ugly hack and I don't know if it can be done. Do you
>> have a suggestion?
>>
>> What are the plans with the full posix signals interface?
>
> Plans were to get it during the 2.6 branch, but of course if someone is
> able to contribute it before, there is no problem.

I would like to help out of course, but first would like some quick
tests if it would be feasible in our application. Any hints on that?

Thanks,
Henri.

>
>>
>> Thanks,
>>> --
>>>                                                                Gilles.
>>>
>>
>
>
> --
>                                                                Gilles.
>

Re: [Xenomai-help] Exception handlers in primary domain / user-space signals

[Philippe Gerum]

On Tue, 2011-02-08 at 09:21 +0100, Henri Roosen wrote:
> On Mon, Feb 7, 2011 at 8:08 PM, Gilles Chanteperdrix
> <gilles.chanteperdrix@xenomai.org> wrote:
> > Henri Roosen wrote:
> >> On Mon, Feb 7, 2011 at 7:27 PM, Gilles Chanteperdrix
> >> <gilles.chanteperdrix@xenomai.org> wrote:
> >>> Henri Roosen wrote:
> >>>> We are using signal handlers for catching exceptions which our
> >>>> application is allowed to make and which we know how to handle.
> >>>>
> >>>> The current Xenomai implementation is to switch to the secondary
> >>>> domain and call the handlers from there.
> >>>> Unfortunately this takes too much time for our application and we
> >>>> would like to handle the exception without the switch to the secondary
> >>>> domain, in primary domain.
> >>>>
> >>>> Can anyone give some advice how to implement that?
> >>>> Will "user-space signals" which was planned for Xenomai 2.6 fulfill this need?
> >>>> Is there already code available for user-space signals?
> >>> In the 2.5 series, we added some code to support signals. The signals
> >>> are multiplexed per-skin in kernel-space, and demultiplexed in
> >>> user-space, upon exit of system calls. We implemented a unit test of
> >>> this functionality with the "sigtest" skin and user-space test, but they
> >>> only work upon return from system calls.
> >>>
> >>> Then we added support for the "mayday" page, which made us realize, that
> >>> maybe implementing signals handling at any time, not only when returning
> >>> from system calls, was possible. But then came the realization that in
> >>> order to implement that, we would have to fiddle with the FPU, which is
> >>> an area where we have a certain tradition for not getting the things
> >>> right at the first attempt. So, we kind of stopped here.
> >>>
> >>> So, if you want some ad-hoc signals upon return from system call, the
> >>> task is pretty easy. If you want the full posix signals interface, then
> >>> things are going to be a bit harder.
> >>>
> >> I am afraid it's going to be a bit harder; we would need it when the
> >> exception occurs and that is in most cases not at a place in the code
> >> where there is a system call :-(.
> >
> > What kind of exception is it? Could not the exception be signalled at
> > the next system call?
> 
> Our customers provide the application code, we provide more or less
> the framework. Customers can install exception handlers for for
> instance floating point exceptions (SIGFPE).
> Besides that we provide a means of tracing the application code, which
> is handled by breakpoints in the code which then does some bookkeeping
> and lets the task run again. Of course that has some overhead also
> when using our old OS, but Linux-Xenomai has so much overhead because
> of the secondary domain switch. Therefore we would like to handle it
> in primary domain.

Connect a high priority shadow task in userland to an exception handler
installed in kernel space via some synchronization (semaphore, event,
whatever). The handler would be called upon exception, then would wake
up your task in userland, which would preempt immediately any other task
activity due to its higher priority. This would not entail any mode
switch, only a fast context switch between Xenomai contexts.

Over this "exception server" task context, you should be able to execute
any kind of user-space handler to mimic the POSIX signal interface as
much as required. Of course this would not run over the faulting context
like POSIX signals do (unless SIGEV_THREAD is used), but this might be
ok for your purpose.

> 
> >
> >>
> >> I was thinking of adding a hook in Xenomai's exception handler before
> >> it makes the switch to the secondary domain... but that would of
> >> course be a very ugly hack and I don't know if it can be done. Do you
> >> have a suggestion?
> >>
> >> What are the plans with the full posix signals interface?
> >
> > Plans were to get it during the 2.6 branch, but of course if someone is
> > able to contribute it before, there is no problem.
> 
> I would like to help out of course, but first would like some quick
> tests if it would be feasible in our application. Any hints on that?
> 
> Thanks,
> Henri.
> 
> >
> >>
> >> Thanks,
> >>> --
> >>>                                                                Gilles.
> >>>
> >>
> >
> >
> > --
> >                                                                Gilles.
> >
> 
> _______________________________________________
> Xenomai-help mailing list
> Xenomai-help@domain.hid
> https://mail.gna.org/listinfo/xenomai-help

-- 
Philippe.

Re: [Xenomai-help] Exception handlers in primary domain / user-space signals

[Henri Roosen]

On Tue, Feb 8, 2011 at 9:38 AM, Philippe Gerum <rpm@xenomai.org> wrote:
> On Tue, 2011-02-08 at 09:21 +0100, Henri Roosen wrote:
>> On Mon, Feb 7, 2011 at 8:08 PM, Gilles Chanteperdrix
>> <gilles.chanteperdrix@xenomai.org> wrote:
>> > Henri Roosen wrote:
>> >> On Mon, Feb 7, 2011 at 7:27 PM, Gilles Chanteperdrix
>> >> <gilles.chanteperdrix@xenomai.org> wrote:
>> >>> Henri Roosen wrote:
>> >>>> We are using signal handlers for catching exceptions which our
>> >>>> application is allowed to make and which we know how to handle.
>> >>>>
>> >>>> The current Xenomai implementation is to switch to the secondary
>> >>>> domain and call the handlers from there.
>> >>>> Unfortunately this takes too much time for our application and we
>> >>>> would like to handle the exception without the switch to the secondary
>> >>>> domain, in primary domain.
>> >>>>
>> >>>> Can anyone give some advice how to implement that?
>> >>>> Will "user-space signals" which was planned for Xenomai 2.6 fulfill this need?
>> >>>> Is there already code available for user-space signals?
>> >>> In the 2.5 series, we added some code to support signals. The signals
>> >>> are multiplexed per-skin in kernel-space, and demultiplexed in
>> >>> user-space, upon exit of system calls. We implemented a unit test of
>> >>> this functionality with the "sigtest" skin and user-space test, but they
>> >>> only work upon return from system calls.
>> >>>
>> >>> Then we added support for the "mayday" page, which made us realize, that
>> >>> maybe implementing signals handling at any time, not only when returning
>> >>> from system calls, was possible. But then came the realization that in
>> >>> order to implement that, we would have to fiddle with the FPU, which is
>> >>> an area where we have a certain tradition for not getting the things
>> >>> right at the first attempt. So, we kind of stopped here.
>> >>>
>> >>> So, if you want some ad-hoc signals upon return from system call, the
>> >>> task is pretty easy. If you want the full posix signals interface, then
>> >>> things are going to be a bit harder.
>> >>>
>> >> I am afraid it's going to be a bit harder; we would need it when the
>> >> exception occurs and that is in most cases not at a place in the code
>> >> where there is a system call :-(.
>> >
>> > What kind of exception is it? Could not the exception be signalled at
>> > the next system call?
>>
>> Our customers provide the application code, we provide more or less
>> the framework. Customers can install exception handlers for for
>> instance floating point exceptions (SIGFPE).
>> Besides that we provide a means of tracing the application code, which
>> is handled by breakpoints in the code which then does some bookkeeping
>> and lets the task run again. Of course that has some overhead also
>> when using our old OS, but Linux-Xenomai has so much overhead because
>> of the secondary domain switch. Therefore we would like to handle it
>> in primary domain.
>
> Connect a high priority shadow task in userland to an exception handler
> installed in kernel space via some synchronization (semaphore, event,
> whatever). The handler would be called upon exception, then would wake
> up your task in userland, which would preempt immediately any other task
> activity due to its higher priority. This would not entail any mode
> switch, only a fast context switch between Xenomai contexts.
>
> Over this "exception server" task context, you should be able to execute
> any kind of user-space handler to mimic the POSIX signal interface as
> much as required. Of course this would not run over the faulting context
> like POSIX signals do (unless SIGEV_THREAD is used), but this might be
> ok for your purpose.
>

Unfortunately we do need the faulting context for the SIGFPE signal
and SIGTRAP (x86) / SIGILL (arm) signals...

For some quick tests, where in Xenomai code would be best to place a
hook for catching exceptions in primary domain which would also
provide the faulting context? Would that be xnpod_trap_fault?

>>
>> >
>> >>
>> >> I was thinking of adding a hook in Xenomai's exception handler before
>> >> it makes the switch to the secondary domain... but that would of
>> >> course be a very ugly hack and I don't know if it can be done. Do you
>> >> have a suggestion?
>> >>
>> >> What are the plans with the full posix signals interface?
>> >
>> > Plans were to get it during the 2.6 branch, but of course if someone is
>> > able to contribute it before, there is no problem.
>>
>> I would like to help out of course, but first would like some quick
>> tests if it would be feasible in our application. Any hints on that?
>>
>> Thanks,
>> Henri.
>>
>> >
>> >>
>> >> Thanks,
>> >>> --
>> >>>                                                                Gilles.
>> >>>
>> >>
>> >
>> >
>> > --
>> >                                                                Gilles.
>> >
>>
>> _______________________________________________
>> Xenomai-help mailing list
>> Xenomai-help@domain.hid
>> https://mail.gna.org/listinfo/xenomai-help
>
> --
> Philippe.
>
>
>

Re: [Xenomai-help] Exception handlers in primary domain / user-space signals

[Philippe Gerum]

On Tue, 2011-02-08 at 10:10 +0100, Henri Roosen wrote:
> On Tue, Feb 8, 2011 at 9:38 AM, Philippe Gerum <rpm@xenomai.org> wrote:
> > On Tue, 2011-02-08 at 09:21 +0100, Henri Roosen wrote:
> >> On Mon, Feb 7, 2011 at 8:08 PM, Gilles Chanteperdrix
> >> <gilles.chanteperdrix@xenomai.org> wrote:
> >> > Henri Roosen wrote:
> >> >> On Mon, Feb 7, 2011 at 7:27 PM, Gilles Chanteperdrix
> >> >> <gilles.chanteperdrix@xenomai.org> wrote:
> >> >>> Henri Roosen wrote:
> >> >>>> We are using signal handlers for catching exceptions which our
> >> >>>> application is allowed to make and which we know how to handle.
> >> >>>>
> >> >>>> The current Xenomai implementation is to switch to the secondary
> >> >>>> domain and call the handlers from there.
> >> >>>> Unfortunately this takes too much time for our application and we
> >> >>>> would like to handle the exception without the switch to the secondary
> >> >>>> domain, in primary domain.
> >> >>>>
> >> >>>> Can anyone give some advice how to implement that?
> >> >>>> Will "user-space signals" which was planned for Xenomai 2.6 fulfill this need?
> >> >>>> Is there already code available for user-space signals?
> >> >>> In the 2.5 series, we added some code to support signals. The signals
> >> >>> are multiplexed per-skin in kernel-space, and demultiplexed in
> >> >>> user-space, upon exit of system calls. We implemented a unit test of
> >> >>> this functionality with the "sigtest" skin and user-space test, but they
> >> >>> only work upon return from system calls.
> >> >>>
> >> >>> Then we added support for the "mayday" page, which made us realize, that
> >> >>> maybe implementing signals handling at any time, not only when returning
> >> >>> from system calls, was possible. But then came the realization that in
> >> >>> order to implement that, we would have to fiddle with the FPU, which is
> >> >>> an area where we have a certain tradition for not getting the things
> >> >>> right at the first attempt. So, we kind of stopped here.
> >> >>>
> >> >>> So, if you want some ad-hoc signals upon return from system call, the
> >> >>> task is pretty easy. If you want the full posix signals interface, then
> >> >>> things are going to be a bit harder.
> >> >>>
> >> >> I am afraid it's going to be a bit harder; we would need it when the
> >> >> exception occurs and that is in most cases not at a place in the code
> >> >> where there is a system call :-(.
> >> >
> >> > What kind of exception is it? Could not the exception be signalled at
> >> > the next system call?
> >>
> >> Our customers provide the application code, we provide more or less
> >> the framework. Customers can install exception handlers for for
> >> instance floating point exceptions (SIGFPE).
> >> Besides that we provide a means of tracing the application code, which
> >> is handled by breakpoints in the code which then does some bookkeeping
> >> and lets the task run again. Of course that has some overhead also
> >> when using our old OS, but Linux-Xenomai has so much overhead because
> >> of the secondary domain switch. Therefore we would like to handle it
> >> in primary domain.
> >
> > Connect a high priority shadow task in userland to an exception handler
> > installed in kernel space via some synchronization (semaphore, event,
> > whatever). The handler would be called upon exception, then would wake
> > up your task in userland, which would preempt immediately any other task
> > activity due to its higher priority. This would not entail any mode
> > switch, only a fast context switch between Xenomai contexts.
> >
> > Over this "exception server" task context, you should be able to execute
> > any kind of user-space handler to mimic the POSIX signal interface as
> > much as required. Of course this would not run over the faulting context
> > like POSIX signals do (unless SIGEV_THREAD is used), but this might be
> > ok for your purpose.
> >
> 
> Unfortunately we do need the faulting context for the SIGFPE signal
> and SIGTRAP (x86) / SIGILL (arm) signals...


It's too much to ask in the current implementation. There is no quick
fix to this, I mean if you want to have a standard signal frame to
exploit in userland. Otherwise, you could pull some relevant bits from
the exception frame in kernel space (you have the struct pt_regs of the
faulting context avail there), and pass them through your
synchronization mechanism to userland, so as to fake some kind of
pseudo-signal frame.

> 
> For some quick tests, where in Xenomai code would be best to place a
> hook for catching exceptions in primary domain which would also
> provide the faulting context? Would that be xnpod_trap_fault?
> 

Yes.

> >>
> >> >
> >> >>
> >> >> I was thinking of adding a hook in Xenomai's exception handler before
> >> >> it makes the switch to the secondary domain... but that would of
> >> >> course be a very ugly hack and I don't know if it can be done. Do you
> >> >> have a suggestion?
> >> >>
> >> >> What are the plans with the full posix signals interface?
> >> >
> >> > Plans were to get it during the 2.6 branch, but of course if someone is
> >> > able to contribute it before, there is no problem.
> >>
> >> I would like to help out of course, but first would like some quick
> >> tests if it would be feasible in our application. Any hints on that?
> >>
> >> Thanks,
> >> Henri.
> >>
> >> >
> >> >>
> >> >> Thanks,
> >> >>> --
> >> >>>                                                                Gilles.
> >> >>>
> >> >>
> >> >
> >> >
> >> > --
> >> >                                                                Gilles.
> >> >
> >>
> >> _______________________________________________
> >> Xenomai-help mailing list
> >> Xenomai-help@domain.hid
> >> https://mail.gna.org/listinfo/xenomai-help
> >
> > --
> > Philippe.
> >
> >
> >

-- 
Philippe.

Re: [Xenomai-help] Exception handlers in primary domain / user-space signals

[Gilles Chanteperdrix]

Philippe Gerum wrote:
> On Tue, 2011-02-08 at 10:10 +0100, Henri Roosen wrote:
>> On Tue, Feb 8, 2011 at 9:38 AM, Philippe Gerum <rpm@xenomai.org> wrote:
>>> On Tue, 2011-02-08 at 09:21 +0100, Henri Roosen wrote:
>>>> On Mon, Feb 7, 2011 at 8:08 PM, Gilles Chanteperdrix
>>>> <gilles.chanteperdrix@xenomai.org> wrote:
>>>>> Henri Roosen wrote:
>>>>>> On Mon, Feb 7, 2011 at 7:27 PM, Gilles Chanteperdrix
>>>>>> <gilles.chanteperdrix@xenomai.org> wrote:
>>>>>>> Henri Roosen wrote:
>>>>>>>> We are using signal handlers for catching exceptions which our
>>>>>>>> application is allowed to make and which we know how to handle.
>>>>>>>>
>>>>>>>> The current Xenomai implementation is to switch to the secondary
>>>>>>>> domain and call the handlers from there.
>>>>>>>> Unfortunately this takes too much time for our application and we
>>>>>>>> would like to handle the exception without the switch to the secondary
>>>>>>>> domain, in primary domain.
>>>>>>>>
>>>>>>>> Can anyone give some advice how to implement that?
>>>>>>>> Will "user-space signals" which was planned for Xenomai 2.6 fulfill this need?
>>>>>>>> Is there already code available for user-space signals?
>>>>>>> In the 2.5 series, we added some code to support signals. The signals
>>>>>>> are multiplexed per-skin in kernel-space, and demultiplexed in
>>>>>>> user-space, upon exit of system calls. We implemented a unit test of
>>>>>>> this functionality with the "sigtest" skin and user-space test, but they
>>>>>>> only work upon return from system calls.
>>>>>>>
>>>>>>> Then we added support for the "mayday" page, which made us realize, that
>>>>>>> maybe implementing signals handling at any time, not only when returning
>>>>>>> from system calls, was possible. But then came the realization that in
>>>>>>> order to implement that, we would have to fiddle with the FPU, which is
>>>>>>> an area where we have a certain tradition for not getting the things
>>>>>>> right at the first attempt. So, we kind of stopped here.
>>>>>>>
>>>>>>> So, if you want some ad-hoc signals upon return from system call, the
>>>>>>> task is pretty easy. If you want the full posix signals interface, then
>>>>>>> things are going to be a bit harder.
>>>>>>>
>>>>>> I am afraid it's going to be a bit harder; we would need it when the
>>>>>> exception occurs and that is in most cases not at a place in the code
>>>>>> where there is a system call :-(.
>>>>> What kind of exception is it? Could not the exception be signalled at
>>>>> the next system call?
>>>> Our customers provide the application code, we provide more or less
>>>> the framework. Customers can install exception handlers for for
>>>> instance floating point exceptions (SIGFPE).
>>>> Besides that we provide a means of tracing the application code, which
>>>> is handled by breakpoints in the code which then does some bookkeeping
>>>> and lets the task run again. Of course that has some overhead also
>>>> when using our old OS, but Linux-Xenomai has so much overhead because
>>>> of the secondary domain switch. Therefore we would like to handle it
>>>> in primary domain.
>>> Connect a high priority shadow task in userland to an exception handler
>>> installed in kernel space via some synchronization (semaphore, event,
>>> whatever). The handler would be called upon exception, then would wake
>>> up your task in userland, which would preempt immediately any other task
>>> activity due to its higher priority. This would not entail any mode
>>> switch, only a fast context switch between Xenomai contexts.
>>>
>>> Over this "exception server" task context, you should be able to execute
>>> any kind of user-space handler to mimic the POSIX signal interface as
>>> much as required. Of course this would not run over the faulting context
>>> like POSIX signals do (unless SIGEV_THREAD is used), but this might be
>>> ok for your purpose.
>>>
>> Unfortunately we do need the faulting context for the SIGFPE signal
>> and SIGTRAP (x86) / SIGILL (arm) signals...
> 
> 
> It's too much to ask in the current implementation. There is no quick
> fix to this, I mean if you want to have a standard signal frame to
> exploit in userland. Otherwise, you could pull some relevant bits from
> the exception frame in kernel space (you have the struct pt_regs of the
> faulting context avail there), and pass them through your
> synchronization mechanism to userland, so as to fake some kind of
> pseudo-signal frame.

All this is certainly doable, but even without Xenomai, going to
kernel-space in case of exception then building a signal frame, going
back to user-space, executing the signal handler, then returning from
the signal (possibly going to kernel-space again) is not exactly a light
operation. So, surely, exceptions should remain exceptional and using
them routinely does not look like the right thing to do.

For the signals themselve:
- implementing SIGFPE will require clearing up exceptions at the FPU
level, which may not be easy, depending on the architecture;
- what will you do with SIGTRAP, stop the task? Why would you need to
remain in primary mode?
- why do you get a SIGILL on ARM? This is an abnormal condition...

-- 
                                                                Gilles.

Re: [Xenomai-help] Exception handlers in primary domain / user-space signals

[Philippe Gerum]

On Tue, 2011-02-08 at 13:09 +0100, Gilles Chanteperdrix wrote:
> Philippe Gerum wrote:
> > On Tue, 2011-02-08 at 10:10 +0100, Henri Roosen wrote:
> >> On Tue, Feb 8, 2011 at 9:38 AM, Philippe Gerum <rpm@xenomai.org> wrote:
> >>> On Tue, 2011-02-08 at 09:21 +0100, Henri Roosen wrote:
> >>>> On Mon, Feb 7, 2011 at 8:08 PM, Gilles Chanteperdrix
> >>>> <gilles.chanteperdrix@xenomai.org> wrote:
> >>>>> Henri Roosen wrote:
> >>>>>> On Mon, Feb 7, 2011 at 7:27 PM, Gilles Chanteperdrix
> >>>>>> <gilles.chanteperdrix@xenomai.org> wrote:
> >>>>>>> Henri Roosen wrote:
> >>>>>>>> We are using signal handlers for catching exceptions which our
> >>>>>>>> application is allowed to make and which we know how to handle.
> >>>>>>>>
> >>>>>>>> The current Xenomai implementation is to switch to the secondary
> >>>>>>>> domain and call the handlers from there.
> >>>>>>>> Unfortunately this takes too much time for our application and we
> >>>>>>>> would like to handle the exception without the switch to the secondary
> >>>>>>>> domain, in primary domain.
> >>>>>>>>
> >>>>>>>> Can anyone give some advice how to implement that?
> >>>>>>>> Will "user-space signals" which was planned for Xenomai 2.6 fulfill this need?
> >>>>>>>> Is there already code available for user-space signals?
> >>>>>>> In the 2.5 series, we added some code to support signals. The signals
> >>>>>>> are multiplexed per-skin in kernel-space, and demultiplexed in
> >>>>>>> user-space, upon exit of system calls. We implemented a unit test of
> >>>>>>> this functionality with the "sigtest" skin and user-space test, but they
> >>>>>>> only work upon return from system calls.
> >>>>>>>
> >>>>>>> Then we added support for the "mayday" page, which made us realize, that
> >>>>>>> maybe implementing signals handling at any time, not only when returning
> >>>>>>> from system calls, was possible. But then came the realization that in
> >>>>>>> order to implement that, we would have to fiddle with the FPU, which is
> >>>>>>> an area where we have a certain tradition for not getting the things
> >>>>>>> right at the first attempt. So, we kind of stopped here.
> >>>>>>>
> >>>>>>> So, if you want some ad-hoc signals upon return from system call, the
> >>>>>>> task is pretty easy. If you want the full posix signals interface, then
> >>>>>>> things are going to be a bit harder.
> >>>>>>>
> >>>>>> I am afraid it's going to be a bit harder; we would need it when the
> >>>>>> exception occurs and that is in most cases not at a place in the code
> >>>>>> where there is a system call :-(.
> >>>>> What kind of exception is it? Could not the exception be signalled at
> >>>>> the next system call?
> >>>> Our customers provide the application code, we provide more or less
> >>>> the framework. Customers can install exception handlers for for
> >>>> instance floating point exceptions (SIGFPE).
> >>>> Besides that we provide a means of tracing the application code, which
> >>>> is handled by breakpoints in the code which then does some bookkeeping
> >>>> and lets the task run again. Of course that has some overhead also
> >>>> when using our old OS, but Linux-Xenomai has so much overhead because
> >>>> of the secondary domain switch. Therefore we would like to handle it
> >>>> in primary domain.
> >>> Connect a high priority shadow task in userland to an exception handler
> >>> installed in kernel space via some synchronization (semaphore, event,
> >>> whatever). The handler would be called upon exception, then would wake
> >>> up your task in userland, which would preempt immediately any other task
> >>> activity due to its higher priority. This would not entail any mode
> >>> switch, only a fast context switch between Xenomai contexts.
> >>>
> >>> Over this "exception server" task context, you should be able to execute
> >>> any kind of user-space handler to mimic the POSIX signal interface as
> >>> much as required. Of course this would not run over the faulting context
> >>> like POSIX signals do (unless SIGEV_THREAD is used), but this might be
> >>> ok for your purpose.
> >>>
> >> Unfortunately we do need the faulting context for the SIGFPE signal
> >> and SIGTRAP (x86) / SIGILL (arm) signals...
> > 
> > 
> > It's too much to ask in the current implementation. There is no quick
> > fix to this, I mean if you want to have a standard signal frame to
> > exploit in userland. Otherwise, you could pull some relevant bits from
> > the exception frame in kernel space (you have the struct pt_regs of the
> > faulting context avail there), and pass them through your
> > synchronization mechanism to userland, so as to fake some kind of
> > pseudo-signal frame.
> 
> All this is certainly doable, but even without Xenomai, going to
> kernel-space in case of exception then building a signal frame, going
> back to user-space, executing the signal handler, then returning from
> the signal (possibly going to kernel-space again) is not exactly a light
> operation. So, surely, exceptions should remain exceptional and using
> them routinely does not look like the right thing to do.

No, you missed the point. The idea is not to forge a stack frame in
kernel space. The idea is to propagate enough information to userland in
order to provide whatever bits are needed there.

> 
> For the signals themselve:
> - implementing SIGFPE will require clearing up exceptions at the FPU
> level, which may not be easy, depending on the architecture;
> - what will you do with SIGTRAP, stop the task? Why would you need to
> remain in primary mode?
> - why do you get a SIGILL on ARM? This is an abnormal condition...
> 

-- 
Philippe.

Re: [Xenomai-help] Exception handlers in primary domain / user-space signals

[Gilles Chanteperdrix]

Philippe Gerum wrote:
> On Tue, 2011-02-08 at 13:09 +0100, Gilles Chanteperdrix wrote:
>> Philippe Gerum wrote:
>>> On Tue, 2011-02-08 at 10:10 +0100, Henri Roosen wrote:
>>>> On Tue, Feb 8, 2011 at 9:38 AM, Philippe Gerum <rpm@xenomai.org> wrote:
>>>>> On Tue, 2011-02-08 at 09:21 +0100, Henri Roosen wrote:
>>>>>> On Mon, Feb 7, 2011 at 8:08 PM, Gilles Chanteperdrix
>>>>>> <gilles.chanteperdrix@xenomai.org> wrote:
>>>>>>> Henri Roosen wrote:
>>>>>>>> On Mon, Feb 7, 2011 at 7:27 PM, Gilles Chanteperdrix
>>>>>>>> <gilles.chanteperdrix@xenomai.org> wrote:
>>>>>>>>> Henri Roosen wrote:
>>>>>>>>>> We are using signal handlers for catching exceptions which our
>>>>>>>>>> application is allowed to make and which we know how to handle.
>>>>>>>>>>
>>>>>>>>>> The current Xenomai implementation is to switch to the secondary
>>>>>>>>>> domain and call the handlers from there.
>>>>>>>>>> Unfortunately this takes too much time for our application and we
>>>>>>>>>> would like to handle the exception without the switch to the secondary
>>>>>>>>>> domain, in primary domain.
>>>>>>>>>>
>>>>>>>>>> Can anyone give some advice how to implement that?
>>>>>>>>>> Will "user-space signals" which was planned for Xenomai 2.6 fulfill this need?
>>>>>>>>>> Is there already code available for user-space signals?
>>>>>>>>> In the 2.5 series, we added some code to support signals. The signals
>>>>>>>>> are multiplexed per-skin in kernel-space, and demultiplexed in
>>>>>>>>> user-space, upon exit of system calls. We implemented a unit test of
>>>>>>>>> this functionality with the "sigtest" skin and user-space test, but they
>>>>>>>>> only work upon return from system calls.
>>>>>>>>>
>>>>>>>>> Then we added support for the "mayday" page, which made us realize, that
>>>>>>>>> maybe implementing signals handling at any time, not only when returning
>>>>>>>>> from system calls, was possible. But then came the realization that in
>>>>>>>>> order to implement that, we would have to fiddle with the FPU, which is
>>>>>>>>> an area where we have a certain tradition for not getting the things
>>>>>>>>> right at the first attempt. So, we kind of stopped here.
>>>>>>>>>
>>>>>>>>> So, if you want some ad-hoc signals upon return from system call, the
>>>>>>>>> task is pretty easy. If you want the full posix signals interface, then
>>>>>>>>> things are going to be a bit harder.
>>>>>>>>>
>>>>>>>> I am afraid it's going to be a bit harder; we would need it when the
>>>>>>>> exception occurs and that is in most cases not at a place in the code
>>>>>>>> where there is a system call :-(.
>>>>>>> What kind of exception is it? Could not the exception be signalled at
>>>>>>> the next system call?
>>>>>> Our customers provide the application code, we provide more or less
>>>>>> the framework. Customers can install exception handlers for for
>>>>>> instance floating point exceptions (SIGFPE).
>>>>>> Besides that we provide a means of tracing the application code, which
>>>>>> is handled by breakpoints in the code which then does some bookkeeping
>>>>>> and lets the task run again. Of course that has some overhead also
>>>>>> when using our old OS, but Linux-Xenomai has so much overhead because
>>>>>> of the secondary domain switch. Therefore we would like to handle it
>>>>>> in primary domain.
>>>>> Connect a high priority shadow task in userland to an exception handler
>>>>> installed in kernel space via some synchronization (semaphore, event,
>>>>> whatever). The handler would be called upon exception, then would wake
>>>>> up your task in userland, which would preempt immediately any other task
>>>>> activity due to its higher priority. This would not entail any mode
>>>>> switch, only a fast context switch between Xenomai contexts.
>>>>>
>>>>> Over this "exception server" task context, you should be able to execute
>>>>> any kind of user-space handler to mimic the POSIX signal interface as
>>>>> much as required. Of course this would not run over the faulting context
>>>>> like POSIX signals do (unless SIGEV_THREAD is used), but this might be
>>>>> ok for your purpose.
>>>>>
>>>> Unfortunately we do need the faulting context for the SIGFPE signal
>>>> and SIGTRAP (x86) / SIGILL (arm) signals...
>>>
>>> It's too much to ask in the current implementation. There is no quick
>>> fix to this, I mean if you want to have a standard signal frame to
>>> exploit in userland. Otherwise, you could pull some relevant bits from
>>> the exception frame in kernel space (you have the struct pt_regs of the
>>> faulting context avail there), and pass them through your
>>> synchronization mechanism to userland, so as to fake some kind of
>>> pseudo-signal frame.
>> All this is certainly doable, but even without Xenomai, going to
>> kernel-space in case of exception then building a signal frame, going
>> back to user-space, executing the signal handler, then returning from
>> the signal (possibly going to kernel-space again) is not exactly a light
>> operation. So, surely, exceptions should remain exceptional and using
>> them routinely does not look like the right thing to do.
> 
> No, you missed the point. The idea is not to forge a stack frame in
> kernel space. The idea is to propagate enough information to userland in
> order to provide whatever bits are needed there.

I was not talking about the Xenomai case specifically, but since Henri
would like to have the full signals implementation with Xenomai, this
does a apply to Xenomai too.


-- 
                                                                Gilles.

Re: [Xenomai-help] Exception handlers in primary domain / user-space signals

[Philippe Gerum]

On Tue, 2011-02-08 at 13:16 +0100, Gilles Chanteperdrix wrote:
> Philippe Gerum wrote:
> > On Tue, 2011-02-08 at 13:09 +0100, Gilles Chanteperdrix wrote:
> >> Philippe Gerum wrote:
> >>> On Tue, 2011-02-08 at 10:10 +0100, Henri Roosen wrote:
> >>>> On Tue, Feb 8, 2011 at 9:38 AM, Philippe Gerum <rpm@xenomai.org> wrote:
> >>>>> On Tue, 2011-02-08 at 09:21 +0100, Henri Roosen wrote:
> >>>>>> On Mon, Feb 7, 2011 at 8:08 PM, Gilles Chanteperdrix
> >>>>>> <gilles.chanteperdrix@xenomai.org> wrote:
> >>>>>>> Henri Roosen wrote:
> >>>>>>>> On Mon, Feb 7, 2011 at 7:27 PM, Gilles Chanteperdrix
> >>>>>>>> <gilles.chanteperdrix@xenomai.org> wrote:
> >>>>>>>>> Henri Roosen wrote:
> >>>>>>>>>> We are using signal handlers for catching exceptions which our
> >>>>>>>>>> application is allowed to make and which we know how to handle.
> >>>>>>>>>>
> >>>>>>>>>> The current Xenomai implementation is to switch to the secondary
> >>>>>>>>>> domain and call the handlers from there.
> >>>>>>>>>> Unfortunately this takes too much time for our application and we
> >>>>>>>>>> would like to handle the exception without the switch to the secondary
> >>>>>>>>>> domain, in primary domain.
> >>>>>>>>>>
> >>>>>>>>>> Can anyone give some advice how to implement that?
> >>>>>>>>>> Will "user-space signals" which was planned for Xenomai 2.6 fulfill this need?
> >>>>>>>>>> Is there already code available for user-space signals?
> >>>>>>>>> In the 2.5 series, we added some code to support signals. The signals
> >>>>>>>>> are multiplexed per-skin in kernel-space, and demultiplexed in
> >>>>>>>>> user-space, upon exit of system calls. We implemented a unit test of
> >>>>>>>>> this functionality with the "sigtest" skin and user-space test, but they
> >>>>>>>>> only work upon return from system calls.
> >>>>>>>>>
> >>>>>>>>> Then we added support for the "mayday" page, which made us realize, that
> >>>>>>>>> maybe implementing signals handling at any time, not only when returning
> >>>>>>>>> from system calls, was possible. But then came the realization that in
> >>>>>>>>> order to implement that, we would have to fiddle with the FPU, which is
> >>>>>>>>> an area where we have a certain tradition for not getting the things
> >>>>>>>>> right at the first attempt. So, we kind of stopped here.
> >>>>>>>>>
> >>>>>>>>> So, if you want some ad-hoc signals upon return from system call, the
> >>>>>>>>> task is pretty easy. If you want the full posix signals interface, then
> >>>>>>>>> things are going to be a bit harder.
> >>>>>>>>>
> >>>>>>>> I am afraid it's going to be a bit harder; we would need it when the
> >>>>>>>> exception occurs and that is in most cases not at a place in the code
> >>>>>>>> where there is a system call :-(.
> >>>>>>> What kind of exception is it? Could not the exception be signalled at
> >>>>>>> the next system call?
> >>>>>> Our customers provide the application code, we provide more or less
> >>>>>> the framework. Customers can install exception handlers for for
> >>>>>> instance floating point exceptions (SIGFPE).
> >>>>>> Besides that we provide a means of tracing the application code, which
> >>>>>> is handled by breakpoints in the code which then does some bookkeeping
> >>>>>> and lets the task run again. Of course that has some overhead also
> >>>>>> when using our old OS, but Linux-Xenomai has so much overhead because
> >>>>>> of the secondary domain switch. Therefore we would like to handle it
> >>>>>> in primary domain.
> >>>>> Connect a high priority shadow task in userland to an exception handler
> >>>>> installed in kernel space via some synchronization (semaphore, event,
> >>>>> whatever). The handler would be called upon exception, then would wake
> >>>>> up your task in userland, which would preempt immediately any other task
> >>>>> activity due to its higher priority. This would not entail any mode
> >>>>> switch, only a fast context switch between Xenomai contexts.
> >>>>>
> >>>>> Over this "exception server" task context, you should be able to execute
> >>>>> any kind of user-space handler to mimic the POSIX signal interface as
> >>>>> much as required. Of course this would not run over the faulting context
> >>>>> like POSIX signals do (unless SIGEV_THREAD is used), but this might be
> >>>>> ok for your purpose.
> >>>>>
> >>>> Unfortunately we do need the faulting context for the SIGFPE signal
> >>>> and SIGTRAP (x86) / SIGILL (arm) signals...
> >>>
> >>> It's too much to ask in the current implementation. There is no quick
> >>> fix to this, I mean if you want to have a standard signal frame to
> >>> exploit in userland. Otherwise, you could pull some relevant bits from
> >>> the exception frame in kernel space (you have the struct pt_regs of the
> >>> faulting context avail there), and pass them through your
> >>> synchronization mechanism to userland, so as to fake some kind of
> >>> pseudo-signal frame.
> >> All this is certainly doable, but even without Xenomai, going to
> >> kernel-space in case of exception then building a signal frame, going
> >> back to user-space, executing the signal handler, then returning from
> >> the signal (possibly going to kernel-space again) is not exactly a light
> >> operation. So, surely, exceptions should remain exceptional and using
> >> them routinely does not look like the right thing to do.
> > 
> > No, you missed the point. The idea is not to forge a stack frame in
> > kernel space. The idea is to propagate enough information to userland in
> > order to provide whatever bits are needed there.
> 
> I was not talking about the Xenomai case specifically, but since Henri
> would like to have the full signals implementation with Xenomai, this
> does a apply to Xenomai too.
> 
> 

I think we all agree that having a complete signal implementation for
Xenomai in pure rt mode won't happen overnight. So the point is now: how
could it be mimicked, at least for the most useful part.

-- 
Philippe.

Re: [Xenomai-help] Exception handlers in primary domain / user-space signals

[Gilles Chanteperdrix]

Philippe Gerum wrote:
> On Tue, 2011-02-08 at 13:16 +0100, Gilles Chanteperdrix wrote:
>> I was not talking about the Xenomai case specifically, but since Henri
>> would like to have the full signals implementation with Xenomai, this
>> does a apply to Xenomai too.
>>
>>
> 
> I think we all agree that having a complete signal implementation for
> Xenomai in pure rt mode won't happen overnight. So the point is now: how
> could it be mimicked, at least for the most useful part.
> 

My point is that whatever you do, a switch user-kernel, then kernel-user
is not going to be lightweight, so avoiding it in the application in the
first place may be a better idea.

My aim with implementing complete signals was rather for things like
timer_* and mq_notify, where the interface requires them, I did not even
imagine implementing SIGFPE, SIGILL, SIGTRAP, which I thought could not
be time critical anyway, for the reasons explained earlier. So, my
question (rather to Henri) is: what would we need SIGFPE, SIGILL,
SIGTRAP in an real-time application for?

-- 
                                                                Gilles.

Re: [Xenomai-help] Exception handlers in primary domain / user-space signals

[Henri Roosen]

On Tue, Feb 8, 2011 at 1:31 PM, Gilles Chanteperdrix
<gilles.chanteperdrix@xenomai.org> wrote:
> Philippe Gerum wrote:
>> On Tue, 2011-02-08 at 13:16 +0100, Gilles Chanteperdrix wrote:
>>> I was not talking about the Xenomai case specifically, but since Henri
>>> would like to have the full signals implementation with Xenomai, this
>>> does a apply to Xenomai too.
>>>
>>>
>>
>> I think we all agree that having a complete signal implementation for
>> Xenomai in pure rt mode won't happen overnight. So the point is now: how
>> could it be mimicked, at least for the most useful part.
>>
>
> My point is that whatever you do, a switch user-kernel, then kernel-user
> is not going to be lightweight, so avoiding it in the application in the
> first place may be a better idea.
>
> My aim with implementing complete signals was rather for things like
> timer_* and mq_notify, where the interface requires them, I did not even
> imagine implementing SIGFPE, SIGILL, SIGTRAP, which I thought could not
> be time critical anyway, for the reasons explained earlier. So, my
> question (rather to Henri) is: what would we need SIGFPE, SIGILL,
> SIGTRAP in an real-time application for?

I agree it might be unusual. For the tracing use case: the SIGTRAP we
use as a means for tracing whether code is actually executed, just
like breakpoints, we exchange the code to 0xcc and handle the
exceptions do book-keeping but don't stop the task. We know this has
overhead, it also had when using our old OS. The old OS handled it in
an accepted amount of time. Using the Xenomai kernel it also works,
however the overhead is not acceptable anymore.
Installing a floating point exception handler was also provided to our
customers with the old OS and we have to make that available now too.
So actually it is all because of legacy reasons, we have to provide
similar functionality as with the old OS.

I'm afraid we cannot mimic enough so it suits our use cases. We need
the fault context to handle the exception and to set the IP one
instruction back.

Thanks,
Henri.

>
> --
>                                                                Gilles.
>

Re: [Xenomai-help] Exception handlers in primary domain / user-space signals

[Philippe Gerum]

On Tue, 2011-02-08 at 13:31 +0100, Gilles Chanteperdrix wrote:
> Philippe Gerum wrote:
> > On Tue, 2011-02-08 at 13:16 +0100, Gilles Chanteperdrix wrote:
> >> I was not talking about the Xenomai case specifically, but since Henri
> >> would like to have the full signals implementation with Xenomai, this
> >> does a apply to Xenomai too.
> >>
> >>
> > 
> > I think we all agree that having a complete signal implementation for
> > Xenomai in pure rt mode won't happen overnight. So the point is now: how
> > could it be mimicked, at least for the most useful part.
> > 
> 
> My point is that whatever you do, a switch user-kernel, then kernel-user
> is not going to be lightweight, so avoiding it in the application in the
> first place may be a better idea.

Sometimes you can't decide what the application code wants, that is the
issue.

> 
> My aim with implementing complete signals was rather for things like
> timer_* and mq_notify, where the interface requires them, I did not even
> imagine implementing SIGFPE, SIGILL, SIGTRAP, which I thought could not
> be time critical anyway, for the reasons explained earlier. So, my
> question (rather to Henri) is: what would we need SIGFPE, SIGILL,
> SIGTRAP in an real-time application for?
> 

-- 
Philippe.

Re: [Xenomai-help] Exception handlers in primary domain / user-space signals

[Philippe Gerum]

On Tue, 2011-02-08 at 13:51 +0100, Henri Roosen wrote:
> On Tue, Feb 8, 2011 at 1:31 PM, Gilles Chanteperdrix
> <gilles.chanteperdrix@xenomai.org> wrote:
> > Philippe Gerum wrote:
> >> On Tue, 2011-02-08 at 13:16 +0100, Gilles Chanteperdrix wrote:
> >>> I was not talking about the Xenomai case specifically, but since Henri
> >>> would like to have the full signals implementation with Xenomai, this
> >>> does a apply to Xenomai too.
> >>>
> >>>
> >>
> >> I think we all agree that having a complete signal implementation for
> >> Xenomai in pure rt mode won't happen overnight. So the point is now: how
> >> could it be mimicked, at least for the most useful part.
> >>
> >
> > My point is that whatever you do, a switch user-kernel, then kernel-user
> > is not going to be lightweight, so avoiding it in the application in the
> > first place may be a better idea.
> >
> > My aim with implementing complete signals was rather for things like
> > timer_* and mq_notify, where the interface requires them, I did not even
> > imagine implementing SIGFPE, SIGILL, SIGTRAP, which I thought could not
> > be time critical anyway, for the reasons explained earlier. So, my
> > question (rather to Henri) is: what would we need SIGFPE, SIGILL,
> > SIGTRAP in an real-time application for?
> 
> I agree it might be unusual. For the tracing use case: the SIGTRAP we
> use as a means for tracing whether code is actually executed, just
> like breakpoints, we exchange the code to 0xcc and handle the
> exceptions do book-keeping but don't stop the task. We know this has
> overhead, it also had when using our old OS. The old OS handled it in
> an accepted amount of time. Using the Xenomai kernel it also works,
> however the overhead is not acceptable anymore.
> Installing a floating point exception handler was also provided to our
> customers with the old OS and we have to make that available now too.
> So actually it is all because of legacy reasons, we have to provide
> similar functionality as with the old OS.
> 
> I'm afraid we cannot mimic enough so it suits our use cases. We need
> the fault context to handle the exception and to set the IP one
> instruction back.

So you need the signal rebase over the mayday support I merged a few
months ago. Back to square one I'm afraid, this won't be available soon,
albeit this might happen in the 2.6 timeframe. We'll see.

> 
> Thanks,
> Henri.
> 
> >
> > --
> >                                                                Gilles.
> >

-- 
Philippe.

Re: [Xenomai-help] Exception handlers in primary domain / user-space signals

[Gilles Chanteperdrix]

Philippe Gerum wrote:
> On Tue, 2011-02-08 at 13:51 +0100, Henri Roosen wrote:
>> On Tue, Feb 8, 2011 at 1:31 PM, Gilles Chanteperdrix
>> <gilles.chanteperdrix@xenomai.org> wrote:
>>> Philippe Gerum wrote:
>>>> On Tue, 2011-02-08 at 13:16 +0100, Gilles Chanteperdrix wrote:
>>>>> I was not talking about the Xenomai case specifically, but since Henri
>>>>> would like to have the full signals implementation with Xenomai, this
>>>>> does a apply to Xenomai too.
>>>>>
>>>>>
>>>> I think we all agree that having a complete signal implementation for
>>>> Xenomai in pure rt mode won't happen overnight. So the point is now: how
>>>> could it be mimicked, at least for the most useful part.
>>>>
>>> My point is that whatever you do, a switch user-kernel, then kernel-user
>>> is not going to be lightweight, so avoiding it in the application in the
>>> first place may be a better idea.
>>>
>>> My aim with implementing complete signals was rather for things like
>>> timer_* and mq_notify, where the interface requires them, I did not even
>>> imagine implementing SIGFPE, SIGILL, SIGTRAP, which I thought could not
>>> be time critical anyway, for the reasons explained earlier. So, my
>>> question (rather to Henri) is: what would we need SIGFPE, SIGILL,
>>> SIGTRAP in an real-time application for?
>> I agree it might be unusual. For the tracing use case: the SIGTRAP we
>> use as a means for tracing whether code is actually executed, just
>> like breakpoints, we exchange the code to 0xcc and handle the
>> exceptions do book-keeping but don't stop the task. We know this has
>> overhead, it also had when using our old OS. The old OS handled it in
>> an accepted amount of time. Using the Xenomai kernel it also works,
>> however the overhead is not acceptable anymore.
>> Installing a floating point exception handler was also provided to our
>> customers with the old OS and we have to make that available now too.
>> So actually it is all because of legacy reasons, we have to provide
>> similar functionality as with the old OS.
>>
>> I'm afraid we cannot mimic enough so it suits our use cases. We need
>> the fault context to handle the exception and to set the IP one
>> instruction back.
> 
> So you need the signal rebase over the mayday support I merged a few
> months ago. Back to square one I'm afraid, this won't be available soon,
> albeit this might happen in the 2.6 timeframe. We'll see.

Well, not necessarily, the fault handler may set the IP, suspend the
task, wake-up the dedicated fault-handler thread, then, the dedicated
fault-handler may wake-up the suspended task when the work has been done.

-- 
					    Gilles.

Re: [Xenomai-help] Exception handlers in primary domain / user-space signals

[Philippe Gerum]

On Tue, 2011-02-08 at 14:11 +0100, Gilles Chanteperdrix wrote:
> Philippe Gerum wrote:
> > On Tue, 2011-02-08 at 13:51 +0100, Henri Roosen wrote:
> >> On Tue, Feb 8, 2011 at 1:31 PM, Gilles Chanteperdrix
> >> <gilles.chanteperdrix@xenomai.org> wrote:
> >>> Philippe Gerum wrote:
> >>>> On Tue, 2011-02-08 at 13:16 +0100, Gilles Chanteperdrix wrote:
> >>>>> I was not talking about the Xenomai case specifically, but since Henri
> >>>>> would like to have the full signals implementation with Xenomai, this
> >>>>> does a apply to Xenomai too.
> >>>>>
> >>>>>
> >>>> I think we all agree that having a complete signal implementation for
> >>>> Xenomai in pure rt mode won't happen overnight. So the point is now: how
> >>>> could it be mimicked, at least for the most useful part.
> >>>>
> >>> My point is that whatever you do, a switch user-kernel, then kernel-user
> >>> is not going to be lightweight, so avoiding it in the application in the
> >>> first place may be a better idea.
> >>>
> >>> My aim with implementing complete signals was rather for things like
> >>> timer_* and mq_notify, where the interface requires them, I did not even
> >>> imagine implementing SIGFPE, SIGILL, SIGTRAP, which I thought could not
> >>> be time critical anyway, for the reasons explained earlier. So, my
> >>> question (rather to Henri) is: what would we need SIGFPE, SIGILL,
> >>> SIGTRAP in an real-time application for?
> >> I agree it might be unusual. For the tracing use case: the SIGTRAP we
> >> use as a means for tracing whether code is actually executed, just
> >> like breakpoints, we exchange the code to 0xcc and handle the
> >> exceptions do book-keeping but don't stop the task. We know this has
> >> overhead, it also had when using our old OS. The old OS handled it in
> >> an accepted amount of time. Using the Xenomai kernel it also works,
> >> however the overhead is not acceptable anymore.
> >> Installing a floating point exception handler was also provided to our
> >> customers with the old OS and we have to make that available now too.
> >> So actually it is all because of legacy reasons, we have to provide
> >> similar functionality as with the old OS.
> >>
> >> I'm afraid we cannot mimic enough so it suits our use cases. We need
> >> the fault context to handle the exception and to set the IP one
> >> instruction back.
> > 
> > So you need the signal rebase over the mayday support I merged a few
> > months ago. Back to square one I'm afraid, this won't be available soon,
> > albeit this might happen in the 2.6 timeframe. We'll see.
> 
> Well, not necessarily, the fault handler may set the IP, suspend the
> task, wake-up the dedicated fault-handler thread, then, the dedicated
> fault-handler may wake-up the suspended task when the work has been done.
> 

This is not exactly what I'd call a straightforward solution (which was
the point of offloading the work to userland) . If one knows how to do
that within the Xenomai core, he could just re-route the mayday
mechanism, no need for sideways.

-- 
Philippe.

Re: [Xenomai-help] Exception handlers in primary domain / user-space signals

[Henri Roosen]

On Tue, Feb 8, 2011 at 2:25 PM, Philippe Gerum <rpm@xenomai.org> wrote:
> On Tue, 2011-02-08 at 14:11 +0100, Gilles Chanteperdrix wrote:
>> Philippe Gerum wrote:
>> > On Tue, 2011-02-08 at 13:51 +0100, Henri Roosen wrote:
>> >> On Tue, Feb 8, 2011 at 1:31 PM, Gilles Chanteperdrix
>> >> <gilles.chanteperdrix@xenomai.org> wrote:
>> >>> Philippe Gerum wrote:
>> >>>> On Tue, 2011-02-08 at 13:16 +0100, Gilles Chanteperdrix wrote:
>> >>>>> I was not talking about the Xenomai case specifically, but since Henri
>> >>>>> would like to have the full signals implementation with Xenomai, this
>> >>>>> does a apply to Xenomai too.
>> >>>>>
>> >>>>>
>> >>>> I think we all agree that having a complete signal implementation for
>> >>>> Xenomai in pure rt mode won't happen overnight. So the point is now: how
>> >>>> could it be mimicked, at least for the most useful part.
>> >>>>
>> >>> My point is that whatever you do, a switch user-kernel, then kernel-user
>> >>> is not going to be lightweight, so avoiding it in the application in the
>> >>> first place may be a better idea.
>> >>>
>> >>> My aim with implementing complete signals was rather for things like
>> >>> timer_* and mq_notify, where the interface requires them, I did not even
>> >>> imagine implementing SIGFPE, SIGILL, SIGTRAP, which I thought could not
>> >>> be time critical anyway, for the reasons explained earlier. So, my
>> >>> question (rather to Henri) is: what would we need SIGFPE, SIGILL,
>> >>> SIGTRAP in an real-time application for?
>> >> I agree it might be unusual. For the tracing use case: the SIGTRAP we
>> >> use as a means for tracing whether code is actually executed, just
>> >> like breakpoints, we exchange the code to 0xcc and handle the
>> >> exceptions do book-keeping but don't stop the task. We know this has
>> >> overhead, it also had when using our old OS. The old OS handled it in
>> >> an accepted amount of time. Using the Xenomai kernel it also works,
>> >> however the overhead is not acceptable anymore.
>> >> Installing a floating point exception handler was also provided to our
>> >> customers with the old OS and we have to make that available now too.
>> >> So actually it is all because of legacy reasons, we have to provide
>> >> similar functionality as with the old OS.
>> >>
>> >> I'm afraid we cannot mimic enough so it suits our use cases. We need
>> >> the fault context to handle the exception and to set the IP one
>> >> instruction back.
>> >
>> > So you need the signal rebase over the mayday support I merged a few
>> > months ago. Back to square one I'm afraid, this won't be available soon,
>> > albeit this might happen in the 2.6 timeframe. We'll see.
>>
>> Well, not necessarily, the fault handler may set the IP, suspend the
>> task, wake-up the dedicated fault-handler thread, then, the dedicated
>> fault-handler may wake-up the suspended task when the work has been done.
>>
>
> This is not exactly what I'd call a straightforward solution (which was
> the point of offloading the work to userland) . If one knows how to do
> that within the Xenomai core, he could just re-route the mayday
> mechanism, no need for sideways.
>

Unfortunately are facing some other problems we need to work on first.
But I would like to investigate the fault handling over the mayday
mechanism when I find some spare time. Could you point me to the
'signal rebase over the mayday' patches? I can find them in the
xenomai-head branch, right?

> --
> Philippe.
>
>
>

Re: [Xenomai-help] Exception handlers in primary domain / user-space signals

[Henri Roosen]

On Fri, Feb 11, 2011 at 10:44 AM, Henri Roosen <henriroosen@domain.hid> wrote:
> On Tue, Feb 8, 2011 at 2:25 PM, Philippe Gerum <rpm@xenomai.org> wrote:
>> On Tue, 2011-02-08 at 14:11 +0100, Gilles Chanteperdrix wrote:
>>> Philippe Gerum wrote:
>>> > On Tue, 2011-02-08 at 13:51 +0100, Henri Roosen wrote:
>>> >> On Tue, Feb 8, 2011 at 1:31 PM, Gilles Chanteperdrix
>>> >> <gilles.chanteperdrix@xenomai.org> wrote:
>>> >>> Philippe Gerum wrote:
>>> >>>> On Tue, 2011-02-08 at 13:16 +0100, Gilles Chanteperdrix wrote:
>>> >>>>> I was not talking about the Xenomai case specifically, but since Henri
>>> >>>>> would like to have the full signals implementation with Xenomai, this
>>> >>>>> does a apply to Xenomai too.
>>> >>>>>
>>> >>>>>
>>> >>>> I think we all agree that having a complete signal implementation for
>>> >>>> Xenomai in pure rt mode won't happen overnight. So the point is now: how
>>> >>>> could it be mimicked, at least for the most useful part.
>>> >>>>
>>> >>> My point is that whatever you do, a switch user-kernel, then kernel-user
>>> >>> is not going to be lightweight, so avoiding it in the application in the
>>> >>> first place may be a better idea.
>>> >>>
>>> >>> My aim with implementing complete signals was rather for things like
>>> >>> timer_* and mq_notify, where the interface requires them, I did not even
>>> >>> imagine implementing SIGFPE, SIGILL, SIGTRAP, which I thought could not
>>> >>> be time critical anyway, for the reasons explained earlier. So, my
>>> >>> question (rather to Henri) is: what would we need SIGFPE, SIGILL,
>>> >>> SIGTRAP in an real-time application for?
>>> >> I agree it might be unusual. For the tracing use case: the SIGTRAP we
>>> >> use as a means for tracing whether code is actually executed, just
>>> >> like breakpoints, we exchange the code to 0xcc and handle the
>>> >> exceptions do book-keeping but don't stop the task. We know this has
>>> >> overhead, it also had when using our old OS. The old OS handled it in
>>> >> an accepted amount of time. Using the Xenomai kernel it also works,
>>> >> however the overhead is not acceptable anymore.
>>> >> Installing a floating point exception handler was also provided to our
>>> >> customers with the old OS and we have to make that available now too.
>>> >> So actually it is all because of legacy reasons, we have to provide
>>> >> similar functionality as with the old OS.
>>> >>
>>> >> I'm afraid we cannot mimic enough so it suits our use cases. We need
>>> >> the fault context to handle the exception and to set the IP one
>>> >> instruction back.
>>> >
>>> > So you need the signal rebase over the mayday support I merged a few
>>> > months ago. Back to square one I'm afraid, this won't be available soon,
>>> > albeit this might happen in the 2.6 timeframe. We'll see.
>>>
>>> Well, not necessarily, the fault handler may set the IP, suspend the
>>> task, wake-up the dedicated fault-handler thread, then, the dedicated
>>> fault-handler may wake-up the suspended task when the work has been done.
>>>
>>
>> This is not exactly what I'd call a straightforward solution (which was
>> the point of offloading the work to userland) . If one knows how to do
>> that within the Xenomai core, he could just re-route the mayday
>> mechanism, no need for sideways.
>>
>
> Unfortunately are facing some other problems we need to work on first.
> But I would like to investigate the fault handling over the mayday
> mechanism when I find some spare time. Could you point me to the
> 'signal rebase over the mayday' patches? I can find them in the
> xenomai-head branch, right?
>
>> --
>> Philippe.
>>
>>
>>
>

Ok, I'm back on this one now.

I couldn't find the 'the signal rebase over the mayday support'
patches which Philippe mentioned. I searched the xenomai-2.5 and
xenomai-head trees. Is it somewhere available? It might be a good
starting point.

The way I see it, currently available is an implementation like used
in sigtest. This means using a special skin that gets a handler
installed for signals at xeno_bind_skin(..., handler). Current status
seems to me a framework for supporting kill system calls. As far as I
can see the signal reporting part has no support for sigcontext (which
is a requirement for us). Also there is no connection of events like
SIGFPE to this mechanism.

Where I see the mayday implementation comes in use is when sending
signals to other threads: it forces the target thread to make a kernel
call from userspace. Then the handle_rt_signals() function can handle
the pending signals.
Is the mayday implementation also needed for exception signals
(SIGFPE), which are generated by the task itself? I see the current
implementation needs a system call from userspace to provide the
struct xnsigs to kernel space. So I guess answer to the last question
is yes... but that might also have some overhead as we were in kernel
space already..

Is it a good idea to implement SIGFPE support using the signal-skin?

So for me there are some unclarities of how to implement primary
domain exception handling from userspace. Would be good when this
email triggers a discussion of how to correctly implement it so I can
then make a useful contribution to the project.

Thanks,
Henri.

Re: [Xenomai-help] Exception handlers in primary domain / user-space signals

[Gilles Chanteperdrix]

Henri Roosen wrote:
> Ok, I'm back on this one now.
> 
> I couldn't find the 'the signal rebase over the mayday support'
> patches which Philippe mentioned. I searched the xenomai-2.5 and
> xenomai-head trees. Is it somewhere available? It might be a good
> starting point.

When Philippe said: "So you need the signal rebase over the mayday
support I merged a few months ago.". The "I merged a few months ago" was
referring to the mayday support, not the signal rebase. This signal
rebase does not exist. Because, especially due to FPU handling, it is
not a trivial thing to do. sigcontext makes things even more complicated.

> 
> The way I see it, currently available is an implementation like used
> in sigtest. This means using a special skin that gets a handler
> installed for signals at xeno_bind_skin(..., handler). Current status
> seems to me a framework for supporting kill system calls. As far as I
> can see the signal reporting part has no support for sigcontext (which
> is a requirement for us). Also there is no connection of events like
> SIGFPE to this mechanism.

The current framework allows:
- sending a signal to a thread
- getting the handler executed by the thread.

But, nothing sends signals to a thread, and nothings allows registering
handlers for handling different signals.

What the mayday support would changes is the way a signal handler could
get executed by a thread.

> 
> Where I see the mayday implementation comes in use is when sending
> signals to other threads: it forces the target thread to make a kernel
> call from userspace. Then the handle_rt_signals() function can handle
> the pending signals.
> Is the mayday implementation also needed for exception signals
> (SIGFPE), which are generated by the task itself? I see the current
> implementation needs a system call from userspace to provide the
> struct xnsigs to kernel space. So I guess answer to the last question
> is yes... but that might also have some overhead as we were in kernel
> space already..

No real overhead. Since we are in kernel-space, the signal handler will
get executed when going back to user-space. The exception causes a
switch to kernel-space anyway.

> 
> Is it a good idea to implement SIGFPE support using the signal-skin?

I think it is a better idea to implement this with whatever skin you are
using native, posix, etc...

> 
> So for me there are some unclarities of how to implement primary
> domain exception handling from userspace. Would be good when this
> email triggers a discussion of how to correctly implement it so I can
> then make a useful contribution to the project.
> 
> Thanks,
> Henri.
> 


-- 
                                                                Gilles.

[Xenomai-help] Xenomai rt_printf() don't print

[davide doninelli]

Hi,
With xenomai on the APF27, the commands rt_printf() don't print on console? 
Have someone  the same issue?
Or i need to change some output ? (on which serial  print out) 


rt_printk() works fine.



thanks in advance 
Davide Doninelli

Re: [Xenomai-help] Xenomai rt_printf() don't print

[Gilles Chanteperdrix]

davide doninelli wrote:
> Hi,
> With xenomai on the APF27, the commands rt_printf() don't print on console? 
> Have someone  the same issue?
> Or i need to change some output ? (on which serial  print out) 
> 
> 
> rt_printk() works fine.
> 

http://www.xenomai.org/index.php/Request_for_information

-- 
                                                                Gilles.

Re: [Xenomai-help] Xenomai rt_printf() don't print

[davide doninelli]

[-- Attachment #1: Type: text/plain, Size: 629 bytes --]

Hi,
sorry i precise
Linux kernel v.2.6.29.6
Adeos patch 2.6.29 2.6.26-mxc
Xenomai v. 2.4.8

	compiler 4.2.1
thanks in advance
davide
Il giorno 19-apr-2011, alle ore 19.13, Gilles Chanteperdrix ha scritto:

> davide doninelli wrote:
>> Hi,
>> With xenomai on the APF27, the commands rt_printf() don't print on console? 
>> Have someone  the same issue?
>> Or i need to change some output ? (on which serial  print out) 
>> 
>> 
>> rt_printk() works fine.
>> 
> 
> http://www.xenomai.org/index.php/Request_for_information
> 
> -- 
>                                                                Gilles.


[-- Attachment #2: Type: text/html, Size: 2248 bytes --]

Re: [Xenomai-help] Xenomai rt_printf() don't print

[Gilles Chanteperdrix]

davide doninelli wrote:
> Hi,
> sorry i precise
> Linux kernel v.2.6.29.6
> Adeos patch 2.6.29 2.6.26-mxc
> Xenomai v. 2.4.8

As mentioned in the first paragraph, please try and verify that the
problem exists with the latest version of that branch (2.4.10).

Please also send us a small self-contained test case.

-- 
                                                                Gilles.

Re: [Xenomai-help] Xenomai rt_printf() don't print

[davide doninelli]

[-- Attachment #1: Type: text/plain, Size: 1835 bytes --]


Hi,
with Xenomai 2.4.10 same error.

i solved with this patch that you have made in 2009. 

best regards 
davide

diff --git a/src/rtdk/rt_print.c b/src/rtdk/rt_print.c
index 0615247..a0aeec3 100644
--- a/src/rtdk/rt_print.c
+++ b/src/rtdk/rt_print.c
@@ -422,6 +422,7 @@ void __rt_print_init(void)
        pthread_attr_t thattr;
        const char *value_str;
        unsigned long long period;
+       unsigned stksize;

        first_buffer = NULL;
        seq_no = 0;
@@ -457,6 +458,9 @@ void __rt_print_init(void)
        pthread_cond_init(&printer_wakeup, NULL);

        pthread_attr_init(&thattr);
-       pthread_attr_setstacksize(&thattr, PTHREAD_STACK_MIN);
+       stksize = 32768;
+       if (stksize < PTHREAD_STACK_MIN)
+               stksize = PTHREAD_STACK_MIN;
+       pthread_attr_setstacksize(&thattr, stksize);
        pthread_create(&printer_thread, &thattr, printer_loop, NULL);
 }


Il giorno 19-apr-2011, alle ore 19.39, davide doninelli ha scritto:

> Hi,
> sorry i precise
> Linux kernel v.2.6.29.6
> Adeos patch 2.6.29 2.6.26-mxc
> Xenomai v. 2.4.8
> 
> 	compiler 4.2.1
> thanks in advance
> davide
> Il giorno 19-apr-2011, alle ore 19.13, Gilles Chanteperdrix ha scritto:
> 
>> davide doninelli wrote:
>>> Hi,
>>> With xenomai on the APF27, the commands rt_printf() don't print on console? 
>>> Have someone  the same issue?
>>> Or i need to change some output ? (on which serial  print out) 
>>> 
>>> 
>>> rt_printk() works fine.
>>> 
>> 
>> http://www.xenomai.org/index.php/Request_for_information
>> 
>> -- 
>>                                                                Gilles.
> 
> _______________________________________________
> Xenomai-help mailing list
> Xenomai-help@domain.hid
> https://mail.gna.org/listinfo/xenomai-help


[-- Attachment #2: Type: text/html, Size: 3904 bytes --]

Re: [Xenomai-help] Exception handlers in primary domain / user-space signals

[Henri Roosen]

[-- Attachment #1: Type: text/plain, Size: 3639 bytes --]

On Fri, Apr 15, 2011 at 3:33 PM, Gilles Chanteperdrix
<gilles.chanteperdrix@xenomai.org> wrote:
> Henri Roosen wrote:
>> Ok, I'm back on this one now.
>>
>> I couldn't find the 'the signal rebase over the mayday support'
>> patches which Philippe mentioned. I searched the xenomai-2.5 and
>> xenomai-head trees. Is it somewhere available? It might be a good
>> starting point.
>
> When Philippe said: "So you need the signal rebase over the mayday
> support I merged a few months ago.". The "I merged a few months ago" was
> referring to the mayday support, not the signal rebase. This signal
> rebase does not exist. Because, especially due to FPU handling, it is
> not a trivial thing to do. sigcontext makes things even more complicated.
>
>>
>> The way I see it, currently available is an implementation like used
>> in sigtest. This means using a special skin that gets a handler
>> installed for signals at xeno_bind_skin(..., handler). Current status
>> seems to me a framework for supporting kill system calls. As far as I
>> can see the signal reporting part has no support for sigcontext (which
>> is a requirement for us). Also there is no connection of events like
>> SIGFPE to this mechanism.
>
> The current framework allows:
> - sending a signal to a thread
> - getting the handler executed by the thread.
>
> But, nothing sends signals to a thread, and nothings allows registering
> handlers for handling different signals.
>
> What the mayday support would changes is the way a signal handler could
> get executed by a thread.
>
>>
>> Where I see the mayday implementation comes in use is when sending
>> signals to other threads: it forces the target thread to make a kernel
>> call from userspace. Then the handle_rt_signals() function can handle
>> the pending signals.
>> Is the mayday implementation also needed for exception signals
>> (SIGFPE), which are generated by the task itself? I see the current
>> implementation needs a system call from userspace to provide the
>> struct xnsigs to kernel space. So I guess answer to the last question
>> is yes... but that might also have some overhead as we were in kernel
>> space already..
>
> No real overhead. Since we are in kernel-space, the signal handler will
> get executed when going back to user-space. The exception causes a
> switch to kernel-space anyway.
>
>>
>> Is it a good idea to implement SIGFPE support using the signal-skin?
>
> I think it is a better idea to implement this with whatever skin you are
> using native, posix, etc...
>

I did some tests with the RT-signal handling currently available in
Xenomai w.r.t. exception handling, but unfortunately didn't lead to
any solution. Then I decided base it more on the native Linux
exception handling.

I've made an experimental implementation for the native skin x86-32
exception handling in primary domain. I used the 2.6.32.20 kernel and
did some basic tests with divide by zero. Because we need the
exception context the patch unfortunately contains a lot of arch
specific code.

Please find the patch attached to this email for anyone who wants to
experiment with this. Let me know of any comments and findings.

Thanks,
Henri.

>>
>> So for me there are some unclarities of how to implement primary
>> domain exception handling from userspace. Would be good when this
>> email triggers a discussion of how to correctly implement it so I can
>> then make a useful contribution to the project.
>>
>> Thanks,
>> Henri.
>>
>
>
> --
>                                                                Gilles.
>

[-- Attachment #2: rt_exception_handling_x86-32.patch --]
[-- Type: text/x-diff, Size: 22266 bytes --]

diff --git a/include/asm-arm/bits/shadow.h b/include/asm-arm/bits/shadow.h
index 270156e..b2056dd 100644
--- a/include/asm-arm/bits/shadow.h
+++ b/include/asm-arm/bits/shadow.h
@@ -291,6 +291,23 @@ static inline void xnarch_fixup_mayday(struct xnarchtcb *tcb,
 #endif
 }
 
+static inline int xnshadow_setup_trap_info(unsigned int vector,
+		struct pt_regs *regs, long errcode, int *sig, siginfo_t *info)
+{
+	return -ENOSYS;
+}
+
+static inline int xnshadow_setup_rt_frame(int sig, void *handler, siginfo_t *info,
+		sigset_t *set, struct pt_regs *regs, void __user *restorer)
+{
+	return -ENOSYS;
+}
+
+static inline long xnshadow_rt_sigreturn(struct pt_regs *regs)
+{
+	return -ENOSYS;
+}
+
 #endif /* XNARCH_HAVE_MAYDAY */
 
 #endif /* !_XENO_ASM_ARM_BITS_SHADOW_H */
diff --git a/include/asm-blackfin/bits/shadow.h b/include/asm-blackfin/bits/shadow.h
index 778cf30..aead1b2 100644
--- a/include/asm-blackfin/bits/shadow.h
+++ b/include/asm-blackfin/bits/shadow.h
@@ -142,6 +142,23 @@ static inline void xnarch_fixup_mayday(struct xnarchtcb *tcb,
 	regs->r5 = tcb->mayday.r5;
 }
 
+static inline int xnshadow_setup_trap_info(unsigned int vector,
+                struct pt_regs *regs, long errcode, int *sig, siginfo_t *info)
+{
+        return -ENOSYS;
+}
+
+static inline int xnshadow_setup_rt_frame(int sig, void *handler, siginfo_t *info,
+                sigset_t *set, struct pt_regs *regs, void __user *restorer)
+{
+        return -ENOSYS;
+}
+
+static inline long xnshadow_rt_sigreturn(struct pt_regs *regs)
+{
+        return -ENOSYS;
+}
+
 #endif /* XNARCH_HAVE_MAYDAY */
 
 #endif /* !_XENO_ASM_BLACKFIN_BITS_SHADOW_H */
diff --git a/include/asm-generic/syscall.h b/include/asm-generic/syscall.h
index 7e3d01a..d4ae122 100644
--- a/include/asm-generic/syscall.h
+++ b/include/asm-generic/syscall.h
@@ -48,6 +48,7 @@
 #define __xn_sys_get_next_sigs 10	/* only unqueue pending signals. */
 #define __xn_sys_drop_u_mode   11	/* stop updating thread->u_mode */
 #define __xn_sys_mayday        12	/* request mayday fixup */
+#define __xn_sys_rt_sig_return 13	/* request signal return fixup */
 
 #define XENOMAI_LINUX_DOMAIN  0
 #define XENOMAI_XENO_DOMAIN   1
diff --git a/include/asm-powerpc/bits/shadow.h b/include/asm-powerpc/bits/shadow.h
index 319f48d..57aa30f 100644
--- a/include/asm-powerpc/bits/shadow.h
+++ b/include/asm-powerpc/bits/shadow.h
@@ -118,6 +118,23 @@ static inline void xnarch_fixup_mayday(struct xnarchtcb *tcb,
 	regs->gpr[8] = tcb->mayday.r8;
 }
 
+static inline int xnshadow_setup_trap_info(unsigned int vector,
+                struct pt_regs *regs, long errcode, int *sig, siginfo_t *info)
+{
+        return -ENOSYS;
+}
+
+static inline int xnshadow_setup_rt_frame(int sig, void *handler, siginfo_t *info,
+                sigset_t *set, struct pt_regs *regs, void __user *restorer)
+{
+        return -ENOSYS;
+}
+
+static inline long xnshadow_rt_sigreturn(struct pt_regs *regs)
+{
+        return -ENOSYS;
+}
+
 #endif /* XNARCH_HAVE_MAYDAY */
 
 #endif /* !_XENO_ASM_POWERPC_BITS_SHADOW_H */
diff --git a/include/asm-x86/bits/shadow_32.h b/include/asm-x86/bits/shadow_32.h
index 41080b1..0717b26 100644
--- a/include/asm-x86/bits/shadow_32.h
+++ b/include/asm-x86/bits/shadow_32.h
@@ -51,6 +51,8 @@ static void xnarch_schedule_tail(struct task_struct *prev)
 
 #ifdef XNARCH_HAVE_MAYDAY
 
+#include <asm/sigframe.h>
+
 static inline void xnarch_setup_mayday_page(void *page)
 {
 	/*
@@ -71,7 +73,7 @@ static inline void xnarch_setup_mayday_page(void *page)
 	 * Also note that if SEP is present, we always assume NPTL on
 	 * the user side.
 	 */
-	static const struct {
+	static struct {
 		struct __attribute__ ((__packed__)) {
 			u8 op;
 			u32 imm;
@@ -103,7 +105,7 @@ static inline void xnarch_setup_mayday_page(void *page)
 		.bug = 0x0b0f,
 	};
 
-	static const struct {
+	static struct {
 		struct __attribute__ ((__packed__)) {
 			u8 op;
 			u32 imm;
@@ -127,10 +129,19 @@ static inline void xnarch_setup_mayday_page(void *page)
 		.bug = 0x0b0f,
 	};
 
-	if (cpu_has_sep)
+	if (cpu_has_sep) {
 		memcpy(page, &code_sep, sizeof(code_sep));
-	else
+		code_sep.mov_eax.imm =
+				__xn_mux_code(0, __xn_sys_rt_sig_return);
+		memcpy((char *)page + XNSHADOW_MD_SIGRETURN_OFFS, &code_sep,
+				sizeof(code_sep));
+	} else {
 		memcpy(page, &code_nosep, sizeof(code_nosep));
+		code_nosep.mov_eax.imm =
+				__xn_mux_code(0, __xn_sys_rt_sig_return);
+		memcpy((char *)page + XNSHADOW_MD_SIGRETURN_OFFS, &code_nosep,
+				sizeof(code_nosep));
+	}
 
 	/* no cache flush required. */
 }
@@ -158,6 +169,320 @@ static inline void xnarch_fixup_mayday(struct xnarchtcb *tcb,
 	regs->x86reg_bp = tcb->mayday.ebp;
 }
 
+/*
+ * Non-blocking signal stack functions based on linux-2.6.32.20
+ */
+#define __FIX_EFLAGS	(X86_EFLAGS_AC | X86_EFLAGS_OF | \
+			 X86_EFLAGS_DF | X86_EFLAGS_TF | X86_EFLAGS_SF | \
+			 X86_EFLAGS_ZF | X86_EFLAGS_AF | X86_EFLAGS_PF | \
+			 X86_EFLAGS_CF)
+
+#ifdef CONFIG_X86_32
+# define FIX_EFLAGS	(__FIX_EFLAGS | X86_EFLAGS_RF)
+#else
+# define FIX_EFLAGS	__FIX_EFLAGS
+#endif
+
+#define COPY(x)			do {			\
+	get_user_ex(regs->x, &sc->x);			\
+} while (0)
+
+#define GET_SEG(seg)		({			\
+	unsigned short tmp;				\
+	get_user_ex(tmp, &sc->seg);			\
+	tmp;						\
+})
+
+#define COPY_SEG(seg)		do {			\
+	regs->seg = GET_SEG(seg);			\
+} while (0)
+
+#define COPY_SEG_CPL3(seg)	do {			\
+	regs->seg = GET_SEG(seg) | 3;			\
+} while (0)
+
+
+static inline int
+xnshadow_restore_sigcontext(struct pt_regs *regs, struct sigcontext __user *sc,
+		   unsigned long *pax)
+{
+	void __user *buf;
+	unsigned int tmpflags;
+	unsigned int err = 0;
+
+	get_user_try {
+
+		set_user_gs(regs, GET_SEG(gs));
+		COPY_SEG(fs);
+		COPY_SEG(es);
+		COPY_SEG(ds);
+
+		COPY(di); COPY(si); COPY(bp); COPY(sp); COPY(bx);
+		COPY(dx); COPY(cx); COPY(ip);
+
+		COPY_SEG_CPL3(cs);
+		COPY_SEG_CPL3(ss);
+
+		get_user_ex(tmpflags, &sc->flags);
+		regs->flags = (regs->flags & ~FIX_EFLAGS) | (tmpflags & FIX_EFLAGS);
+		regs->orig_ax = -1;		/* disable syscall checks */
+
+		get_user_ex(buf, &sc->fpstate);
+
+		/* Do we have to disable / enable preemption here? */
+		if (buf)
+			xnpod_restore_fpu(buf);
+
+		get_user_ex(*pax, &sc->ax);
+	} get_user_catch(err);
+
+	return err;
+}
+
+static inline int xnshadow_setup_sigcontext(struct sigcontext __user *sc, void __user *fpstate,
+		 struct pt_regs *regs, unsigned long mask)
+{
+	int err = 0;
+
+	put_user_try {
+
+		put_user_ex(get_user_gs(regs), (unsigned int __user *)&sc->gs);
+		put_user_ex(regs->fs, (unsigned int __user *)&sc->fs);
+		put_user_ex(regs->es, (unsigned int __user *)&sc->es);
+		put_user_ex(regs->ds, (unsigned int __user *)&sc->ds);
+
+		put_user_ex(regs->di, &sc->di);
+		put_user_ex(regs->si, &sc->si);
+		put_user_ex(regs->bp, &sc->bp);
+		put_user_ex(regs->sp, &sc->sp);
+		put_user_ex(regs->bx, &sc->bx);
+		put_user_ex(regs->dx, &sc->dx);
+		put_user_ex(regs->cx, &sc->cx);
+		put_user_ex(regs->ax, &sc->ax);
+
+		put_user_ex(current->thread.trap_no, &sc->trapno);
+		put_user_ex(current->thread.error_code, &sc->err);
+		put_user_ex(regs->ip, &sc->ip);
+		put_user_ex(regs->cs, (unsigned int __user *)&sc->cs);
+		put_user_ex(regs->flags, &sc->flags);
+		put_user_ex(regs->sp, &sc->sp_at_signal);
+		put_user_ex(regs->ss, (unsigned int __user *)&sc->ss);
+
+		put_user_ex(fpstate, &sc->fpstate);
+
+		/* non-iBCS2 extensions.. */
+		put_user_ex(mask, &sc->oldmask);
+		put_user_ex(current->thread.cr2, &sc->cr2);
+	} put_user_catch(err);
+
+	return err;
+}
+
+static inline long xnshadow_rt_sigreturn(struct pt_regs *regs)
+{
+	struct xnthread *cur;
+	struct rt_sigframe __user *frame;
+	unsigned long ax;
+
+	frame = (struct rt_sigframe __user *)(regs->sp - sizeof(long));
+	if (!access_rok(frame, sizeof(*frame)))
+		goto badframe;
+
+	if (xnshadow_restore_sigcontext(regs, &frame->uc.uc_mcontext, &ax))
+		goto badframe;
+
+	return ax;
+
+badframe:
+	/* TODO: what can we do when we have a bad frame? */
+	cur = xnshadow_thread(current);
+	if (likely(cur != NULL))
+		xnshadow_call_mayday(cur, SIGDEBUG_WATCHDOG);
+
+	return 0;
+}
+
+static inline void __user *
+xnshadow_get_sigframe(struct pt_regs *regs, size_t frame_size,
+	     void __user **fpstate)
+{
+	/* Default to using normal stack */
+	unsigned long sp = regs->sp;
+
+	if (wrap_test_fpu_used(current)) {
+		sp -= sizeof(x86_fpustate);
+		*fpstate = (void __user *)sp;
+	}
+
+	/* Offset the stackpointer so there is room for the frame and align */
+	sp = ((sp - frame_size + 4) & -16ul) - 4;
+
+	/* save i387 state */
+	if (wrap_test_fpu_used(current))
+		xnpod_save_fpu(*fpstate);
+
+	return (void __user *)sp;
+}
+
+static const struct {
+	u8  movl;
+	u32 val;
+	u16 int80;
+	u8  pad;
+} __attribute__((packed)) rt_retcode = {
+	0xb8,		/* movl $..., %eax */
+	__NR_rt_sigreturn,
+	0x80cd,		/* int $0x80 */
+	0
+};
+
+static inline int xnshadow_setup_rt_frame(int sig, void *handler, siginfo_t *info,
+			    sigset_t *set, struct pt_regs *regs, void __user *restorer)
+{
+	struct rt_sigframe __user *frame;
+	int err = 0;
+	void __user *fpstate = NULL;
+
+	frame = xnshadow_get_sigframe(regs, sizeof(*frame), &fpstate);
+	if (!access_wok(frame, sizeof(*frame)))
+		return -EFAULT;
+
+	put_user_try {
+		put_user_ex(sig, &frame->sig);
+		put_user_ex(&frame->info, &frame->pinfo);
+		put_user_ex(&frame->uc, &frame->puc);
+		err |= copy_siginfo_to_user(&frame->info, info);
+
+		/* Create the ucontext. */
+		if (cpu_has_xsave)
+			put_user_ex(UC_FP_XSTATE, &frame->uc.uc_flags);
+		else
+			put_user_ex(0, &frame->uc.uc_flags);
+		put_user_ex(0, &frame->uc.uc_link);
+		put_user_ex(current->sas_ss_sp, &frame->uc.uc_stack.ss_sp);
+		put_user_ex(sas_ss_flags(regs->sp),
+			    &frame->uc.uc_stack.ss_flags);
+		put_user_ex(current->sas_ss_size, &frame->uc.uc_stack.ss_size);
+		err |= xnshadow_setup_sigcontext(&frame->uc.uc_mcontext, fpstate,
+					regs, set->sig[0]);
+		err |= __copy_to_user(&frame->uc.uc_sigmask, set, sizeof(*set));
+
+		/* Set up to return from userspace.  */
+		put_user_ex(restorer, &frame->pretcode);
+
+		/*
+		 * This is movl $__NR_rt_sigreturn, %ax ; int $0x80
+		 *
+		 * WE DO NOT USE IT ANY MORE! It's only left here for historical
+		 * reasons and because gdb uses it as a signature to notice
+		 * signal handler stack frames.
+		 */
+		put_user_ex(*((u64 *)&rt_retcode), (u64 *)frame->retcode);
+	} put_user_catch(err);
+
+	if (err)
+		return -EFAULT;
+
+	/* Set up registers for signal handler */
+	regs->sp = (unsigned long)frame;
+	regs->ip = (unsigned long)handler;
+	regs->ax = (unsigned long)sig;
+	regs->dx = (unsigned long)&frame->info;
+	regs->cx = (unsigned long)&frame->uc;
+
+	regs->ds = __USER_DS;
+	regs->es = __USER_DS;
+	regs->ss = __USER_DS;
+	regs->cs = __USER_CS;
+
+	return 0;
+}
+
+static inline int xnshadow_setup_trap_info(unsigned int vector,
+		struct pt_regs *regs, long errcode, int *sig, siginfo_t *info)
+{
+	switch (vector) {
+
+	case 0: /* divide_error */
+		*sig = SIGFPE;
+		info->si_signo = *sig;
+		info->si_errno = 0;
+		info->si_code = FPE_INTDIV;
+		info->si_addr = (void __user *)regs->ip;
+		return 0;
+	case 6: /* invalid_op */
+		*sig = SIGILL;
+		info->si_signo = *sig;
+		info->si_errno = 0;
+		info->si_code = ILL_ILLOPN;
+		info->si_addr = (void __user *)regs->ip;
+		return 0;
+	case 9: /* coprocessor_segment_overrun */
+		*sig = SIGFPE;
+		return 0;
+	case 16: { /* coprocessor_error */
+		unsigned short cwd, swd, err;
+		xnarchtcb_t *tcb;
+
+		*sig = SIGFPE;
+		/*
+		 * Save the info for the exception handler and clear the error.
+		 */
+		tcb = xnthread_archtcb(xnpod_current_thread());
+		xnpod_save_fpu(NULL);
+
+		info->si_signo = *sig;
+		info->si_errno = 0;
+		info->si_addr = (void __user *)regs->ip;
+		/*
+		 * (~cwd & swd) will mask out exceptions that are not set to unmasked
+		 * status.  0x3f is the exception bits in these regs, 0x200 is the
+		 * C1 reg you need in case of a stack fault, 0x040 is the stack
+		 * fault bit.  We should only be taking one exception at a time,
+		 * so if this combination doesn't produce any single exception,
+		 * then we have a bad program that isn't synchronizing its FPU usage
+		 * and it will suffer the consequences since we won't be able to
+		 * fully reproduce the context of the exception
+		 */
+		if (cpu_has_fxsr) {
+			cwd = tcb->fpup->fxsave.cwd;
+			swd = tcb->fpup->fxsave.swd;
+		} else {
+			cwd = tcb->fpup->fsave.cwd;
+			swd = tcb->fpup->fsave.swd;
+		}
+		err = swd & ~cwd;
+
+		if (err & 0x001) {	/* Invalid op */
+			/*
+			 * swd & 0x240 == 0x040: Stack Underflow
+			 * swd & 0x240 == 0x240: Stack Overflow
+			 * User must clear the SF bit (0x40) if set
+			 */
+			info->si_code = FPE_FLTINV;
+		} else if (err & 0x004) { /* Divide by Zero */
+			info->si_code = FPE_FLTDIV;
+		} else if (err & 0x008) { /* Overflow */
+			info->si_code = FPE_FLTOVF;
+		} else if (err & 0x012) { /* Denormal, Underflow */
+			info->si_code = FPE_FLTUND;
+		} else if (err & 0x020) { /* Precision */
+			info->si_code = FPE_FLTRES;
+		} else {
+			/*
+			 * If we're using IRQ 13, or supposedly even some trap 16
+			 * implementations, it's possible we get a spurious trap...
+			 */
+			return -ENOSYS; /* Spurious trap, no error */
+		}
+		return 0;
+	}
+	default:
+		break;
+	}
+	return -ENOSYS;
+}
+
 #endif /* XNARCH_HAVE_MAYDAY */
 
 #endif /* !_XENO_ASM_X86_BITS_SHADOW_32_H */
diff --git a/include/asm-x86/bits/shadow_64.h b/include/asm-x86/bits/shadow_64.h
index 69881db..a53994b 100644
--- a/include/asm-x86/bits/shadow_64.h
+++ b/include/asm-x86/bits/shadow_64.h
@@ -121,6 +121,23 @@ static inline void xnarch_fixup_mayday(struct xnarchtcb *tcb,
 	regs->r9 = tcb->mayday.r9;
 }
 
+static inline int xnshadow_setup_trap_info(unsigned int vector,
+                struct pt_regs *regs, long errcode, int *sig, siginfo_t *info)
+{
+        return -ENOSYS;
+}
+
+static inline int xnshadow_setup_rt_frame(int sig, void *handler, siginfo_t *info,
+                sigset_t *set, struct pt_regs *regs, void __user *restorer)
+{
+        return -ENOSYS;
+}
+
+static inline long xnshadow_rt_sigreturn(struct pt_regs *regs)
+{
+        return -ENOSYS;
+}
+
 #endif /* XNARCH_HAVE_MAYDAY */
 
 #endif /* !_XENO_ASM_X86_BITS_SHADOW_64_H */
diff --git a/include/native/misc.h b/include/native/misc.h
index 00c4a72..115da49 100644
--- a/include/native/misc.h
+++ b/include/native/misc.h
@@ -92,6 +92,8 @@ extern "C" {
 
 /* Public interface. */
 
+int rt_sigaction(int signum, void *action);
+
 int rt_io_get_region(RT_IOREGION *iorn,
 		     const char *name,
 		     uint64_t start,
diff --git a/include/native/syscall.h b/include/native/syscall.h
index 2e3d7ec..df9b259 100644
--- a/include/native/syscall.h
+++ b/include/native/syscall.h
@@ -128,6 +128,7 @@
 #define __native_buffer_inquire     102
 #define __native_queue_flush        103
 #define __native_cond_wait_epilogue 104
+#define __native_sigaction          105
 
 struct rt_arg_bulk {
 
diff --git a/include/nucleus/pod.h b/include/nucleus/pod.h
index 056e376..2701946 100644
--- a/include/nucleus/pod.h
+++ b/include/nucleus/pod.h
@@ -135,6 +135,8 @@ void __xnpod_reset_thread(struct xnthread *thread);
 
 #ifdef CONFIG_XENO_HW_FPU
 void xnpod_switch_fpu(xnsched_t *sched);
+void xnpod_save_fpu(void __user *buf);
+void xnpod_restore_fpu(void __user *buf);
 #endif /* CONFIG_XENO_HW_FPU */
 
 void __xnpod_schedule(struct xnsched *sched);
diff --git a/include/nucleus/ppd.h b/include/nucleus/ppd.h
index 972b0f2..7e9f19a 100644
--- a/include/nucleus/ppd.h
+++ b/include/nucleus/ppd.h
@@ -16,6 +16,7 @@ typedef struct xnshadow_ppd_t {
     xnholder_t link;
 #ifdef XNARCH_HAVE_MAYDAY
     unsigned long mayday_addr;
+	void *sighand[_NSIG];
 #endif
 #define link2ppd(ln)	container_of(ln, xnshadow_ppd_t, link)
 } xnshadow_ppd_t;
diff --git a/include/nucleus/shadow.h b/include/nucleus/shadow.h
index cd88522..6697f45 100644
--- a/include/nucleus/shadow.h
+++ b/include/nucleus/shadow.h
@@ -112,7 +112,10 @@ void xnshadow_clear_sig(struct xnthread *thread, unsigned muxid);
 
 #ifdef RTHAL_HAVE_RETURN_EVENT
 #define XNARCH_HAVE_MAYDAY  1
+#define XNSHADOW_MD_MAYDAYRETURN_OFFS 0
+#define XNSHADOW_MD_SIGRETURN_OFFS    50
 void xnshadow_call_mayday(struct xnthread *thread, int sigtype);
+int xnshadow_handle_rt_signals(unsigned int vector, struct pt_regs *regs, long errcode);
 #else
 static inline void xnshadow_call_mayday(struct xnthread *thread,
 					int sigtype)
diff --git a/ksrc/nucleus/pod.c b/ksrc/nucleus/pod.c
index b5ddbaa..1bae36c 100644
--- a/ksrc/nucleus/pod.c
+++ b/ksrc/nucleus/pod.c
@@ -166,6 +166,25 @@ static inline int __xnpod_fault_init_fpu(struct xnthread *thread)
 	return 0;
 }
 
+void xnpod_save_fpu(void __user *buf)
+{
+	xnarch_save_fpu(xnthread_archtcb(xnpod_current_thread()));
+	if (buf)
+		__xn_safe_copy_to_user(buf,
+				xnthread_archtcb(xnpod_current_thread())->fpup,
+				sizeof(*xnthread_archtcb(xnpod_current_thread())->fpup));
+}
+
+void xnpod_restore_fpu(void __user *buf)
+{
+	/* TODO: should we do locking/preempt disable? */
+	if (!buf || __xn_safe_copy_from_user(
+			xnthread_archtcb(xnpod_current_thread())->fpup, buf,
+			sizeof(*xnthread_archtcb(xnpod_current_thread())->fpup)))
+		return;
+	xnarch_restore_fpu(xnthread_archtcb(xnpod_current_thread()));
+}
+
 #else /* !CONFIG_XENO_HW_FPU */
 
 static inline void __xnpod_init_fpu(struct xnsched *sched,
@@ -2625,6 +2644,12 @@ int xnpod_trap_fault(xnarch_fltinfo_t *fltinfo)
 			     xnarch_fault_pc(fltinfo),
 			     xnthread_user_pid(thread));
 #endif /* XENO_DEBUG(NUCLEUS) */
+
+		/* It might be we have to handle this fault in realtime */
+		if (xnshadow_handle_rt_signals(xnarch_fault_trap(fltinfo),
+				fltinfo->regs, xnarch_fault_code(fltinfo)))
+			return 1;
+
 		if (xnarch_fault_pf_p(fltinfo))
 			/* The page fault counter is not SMP-safe, but it's a
 			   simple indicator that something went wrong wrt memory
diff --git a/ksrc/nucleus/shadow.c b/ksrc/nucleus/shadow.c
index d8d7116..5264a50 100644
--- a/ksrc/nucleus/shadow.c
+++ b/ksrc/nucleus/shadow.c
@@ -1675,6 +1675,46 @@ static int xnshadow_sys_mayday(struct pt_regs *regs)
 	return -EPERM;
 }
 
+static int xnshadow_sys_rt_sig_return(struct pt_regs *regs)
+{
+	xnshadow_rt_sigreturn(regs);
+
+	return 0;
+}
+
+int xnshadow_handle_rt_signals(unsigned int vector, struct pt_regs *regs, long errcode)
+{
+	int sig = 0;
+	struct xnshadow_ppd_t *sys_ppd;
+	siginfo_t info;
+	sigset_t set;
+	int rval;
+
+	/* Check if we can get a signo for the vector */
+	rval = xnshadow_setup_trap_info(vector, regs, errcode, &sig, &info);
+	if (rval || !sig)
+		return 0;
+
+	/* We enter the event handler with hw IRQs off. */
+	xnlock_get(&nklock);
+	sys_ppd = xnshadow_ppd_get(0);
+	xnlock_put(&nklock);
+	XENO_BUGON(NUCLEUS, sys_ppd == NULL);
+
+	/* Check if we have a handler installed for this sig */
+	if (!sys_ppd->sighand[sig] || (sys_ppd->sighand[sig] == SIG_DFL))
+		return 0;
+
+	/* Signal delivery */
+	rval = xnshadow_setup_rt_frame(sig, sys_ppd->sighand[sig], &info, &set, regs,
+		(void *)(sys_ppd->mayday_addr + XNSHADOW_MD_SIGRETURN_OFFS));
+
+	if (rval)
+		return 0;
+	else
+		return 1;
+}
+
 static inline int mayday_init_page(void)
 {
 	mayday_page = vmalloc(PAGE_SIZE);
@@ -1706,7 +1746,8 @@ static inline void do_mayday_event(struct pt_regs *regs)
 	xnlock_put(&nklock);
 	XENO_BUGON(NUCLEUS, sys_ppd == NULL);
 
-	xnarch_handle_mayday(tcb, regs, sys_ppd->mayday_addr);
+	xnarch_handle_mayday(tcb, regs,
+			sys_ppd->mayday_addr + XNSHADOW_MD_MAYDAYRETURN_OFFS);
 }
 
 RTHAL_DECLARE_MAYDAY_EVENT(mayday_event);
@@ -1727,6 +1768,10 @@ static inline void mayday_cleanup_page(void)
 {
 }
 
+static int xnshadow_sys_rt_sig_return(struct pt_regs *regs)
+{
+	return -ENOSYS;
+}
 #endif /* XNARCH_HAVE_MAYDAY */
 
 static int xnshadow_sys_bind(struct pt_regs *regs)
@@ -1829,6 +1874,11 @@ static int xnshadow_sys_bind(struct pt_regs *regs)
 		err = -ENOMEM;
 		goto fail_destroy_sys_ppd;
 	}
+	{
+		int i;
+		for (i = 0; i < ARRAY_SIZE(sys_ppd->sighand); i++)
+			sys_ppd->sighand[i] = 0;
+	}
 #endif /* XNARCH_HAVE_MAYDAY */
 
 	if (ppd_insert(sys_ppd) == -EBUSY) {
@@ -2160,6 +2210,7 @@ static xnsysent_t __systab[] = {
 	[__xn_sys_drop_u_mode] =
 		{&xnshadow_sys_drop_u_mode, __xn_exec_shadow},
 	[__xn_sys_mayday] = {&xnshadow_sys_mayday, __xn_exec_any|__xn_exec_norestart},
+	[__xn_sys_rt_sig_return] = {&xnshadow_sys_rt_sig_return, __xn_exec_primary|__xn_exec_norestart},
 };
 
 static void post_ppd_release(struct xnheap *h)
diff --git a/ksrc/skins/native/syscall.c b/ksrc/skins/native/syscall.c
index e6c4257..324f8dc 100644
--- a/ksrc/skins/native/syscall.c
+++ b/ksrc/skins/native/syscall.c
@@ -3972,6 +3972,29 @@ static int __rt_io_put_region(struct pt_regs *regs)
 	return 0;
 }
 
+static int __rt_sigaction(struct pt_regs *regs)
+{
+	int signum = 0;
+	void *handler = NULL;
+	struct xnshadow_ppd_t *sys_ppd;
+
+	/* We enter the event handler with hw IRQs off. */
+	xnlock_get(&nklock);
+	sys_ppd = xnshadow_ppd_get(0);
+	xnlock_put(&nklock);
+	XENO_BUGON(NUCLEUS, sys_ppd == NULL);
+
+	if (__xn_safe_copy_from_user(&signum, (void __user *)__xn_reg_arg1(regs),
+				     sizeof(signum)))
+		return -EFAULT;
+	if (__xn_safe_copy_from_user(&handler, (void __user *)__xn_reg_arg2(regs),
+				     sizeof(handler)))
+		return -EFAULT;
+
+	sys_ppd->sighand[signum] = handler;
+	return 0;
+}
+
 static __attribute__ ((unused))
 int __rt_call_not_available(struct pt_regs *regs)
 {
@@ -4147,6 +4170,7 @@ static xnsysent_t __systab[] = {
 	[__native_buffer_write] = {&__rt_buffer_write, __xn_exec_conforming},
 	[__native_buffer_clear] = {&__rt_buffer_clear, __xn_exec_any},
 	[__native_buffer_inquire] = {&__rt_buffer_inquire, __xn_exec_any},
+	[__native_sigaction] = {&__rt_sigaction, __xn_exec_any},
 };
 
 static struct xnskin_props __props = {
diff --git a/src/skins/native/misc.c b/src/skins/native/misc.c
index 1570d14..71bfee0 100644
--- a/src/skins/native/misc.c
+++ b/src/skins/native/misc.c
@@ -18,9 +18,16 @@
 
 #include <native/syscall.h>
 #include <native/misc.h>
+/* #include <signal.h> */
 
 extern int __native_muxid;
 
+int rt_sigaction(int signum, void *action)
+{
+	return XENOMAI_SKINCALL2(__native_muxid,
+			__native_sigaction, &signum, &action);
+}
+
 int rt_io_get_region(RT_IOREGION *iorn, const char *name,
 		     uint64_t start, uint64_t len, int flags)
 {