Re: [Xenomai-help] gpio-keys with xenomai

Re: [Xenomai-help] gpio-keys with xenomai

[Peter Breuer]

Hello, 
I finally had the time to test some different settings...

I wrote the following program:
#include <sys/mman.h>
#include <native/task.h>
#include <native/intr.h>

#define IRQ_NUMBER 100  /* Intercept interrupt */
#define IRQ_MODE   I_NOAUTOENA //I_PROPAGATE //0
#define TASK_PRIO  99 /* Highest RT priority */
#define TASK_MODE  T_FPU  /*  */
#define TASK_STKSZ 0  /* Stack size (use default one) */
#define PIPE_MINOR 0

RT_INTR intr_desc;

RT_TASK server_desc;


void irq_server (void *cookie)

{
        int err =
rt_intr_create(&intr_desc,"MyIrq",IRQ_NUMBER,IRQ_MODE);
        printf("int create: %d\n", err);
        err = rt_intr_enable(&intr_desc);
        printf("enable int: %d\n", err);
        printf("begin..\n");
        err = rt_intr_wait(&intr_desc,TM_INFINITE);
        printf("int wait:%d\n", err);

}

int main (int argc, char *argv[])

{
        int err;
        int a;
        printf("Programmstart\n");
        mlockall(MCL_CURRENT|MCL_FUTURE);

        err = rt_task_create(&server_desc,
                                "MyIrqServer",
                                TASK_STKSZ,
                                TASK_PRIO,
                                TASK_MODE);
        printf("task create: %d\n", err);
        rt_task_start(&server_desc,&irq_server,NULL);
}

void cleanup (void)

{
    rt_intr_delete(&intr_desc);
    rt_task_delete(&server_desc);
}

When I start it with the compiled xenomai, it runs until the
rt_intr_wait() and then quits, without any message. It doesnt even
execute printf("int wait:%d\n", err)
create_task, create_intr and enable_intr all return 0. 
When I start the programm with a kernel, where xenomai isnt compiled in,
then it quits and complains about it. 
When I start the programm with xenomai debug settings compiled in the
kernel, the system freezes at task creating. Heres the output:
Programmstart
taskXenomai: fatal: bug at kernel/xenomai/nucleus/shadow.c
:1128 (!irqs_disabled_hw())
CPU PID PRI TIMEOUT STAT NAME
0 0 -1 0 00400088 ROOT
> 0 923 99 0 00300180 MyIrqServer
Master time base: clock=114327768
[<c002bde4>] (show_stack+0x0/0x48) from [<c006e69c>] (xnshadow_harden
+0x1d0/0x21c)
[<c006e4cc>] (xnshadow_harden+0x0/0x21c) from [<c006ec00>]
(xnshadow_wait_barrier+0x124/0x138)
[<c006eadc>] (xnshadow_wait_barrier+0x0/0x138) from [<c006fbe4>]
(xnshadow_sys_barrier+0x14/0x18)
r6:c309a000 r5:c309bfb0 r4:00000001
[<c006fbd0>] (xnshadow_sys_barrier+0x0/0x18) from [<c006ff70>]
(losyscall_event+0xb4/0x1ac)
[<c006febc>] (losyscall_event+0x0/0x1ac) from [<c0061214>]
(__ipipe_dispatch_event+0xd0/0x1c8)
[<c0061144>] (__ipipe_dispatch_event+0x0/0x1c8) from [<c002cc5c>]
(__ipipe_syscall_root+0x80/0x100)
[<c002cbdc>] (__ipipe_syscall_root+0x0/0x100) from [<c0028108>]
(vector_swi+0x68/0xa8)

Any ideas? 
is the irq-number, that I need for the rt_intr_create() the same number
that is shown in /proc/interrupts?
thx, greetings, 
Peter



Am Donnerstag, den 18.03.2010, 21:14 +0100 schrieb Gilles Chanteperdrix:
> Peter Breuer wrote:
> > Hello everyone,
> > I am new to xenomai, so please be patient.
> > I am working with an AT91SAM9621S (Board SBC6000X) and I
successfully
> > patched my linux 2.6.24 with xenomai 2.4.9.1. I also have a signal
on on
> > of the GPIOs and I want a program to react, every time the signal
> > arrives (~10Hz).
> > Before compiling I added some lines to the board description file
> > (arch/arm/mach-at91/board-sbc9621.c):
> > ...
> > static struct gpio_keys_button sbc9261_buttons[] = {
> > ...
> > //begin changes
> >     {
> >         .gpio        = AT91_PIN_PC4,
> >         .code        = BTN_2,
> >         .desc        = "GPIO 1",
> >         .active_low    = 1,
> >     },
> > //end changes
> > };
> > ...
> > static void __init sbc9261_add_device_buttons(void)
> > {
> > ...
> > //begin changes
> >     at91_set_gpio_input(AT91_PIN_PC4, 1);    /* btn1 */
> >     at91_set_deglitch(AT91_PIN_PC4, 1);
> > //end changes
> > 
> >     platform_device_register(&sbc9261_button_device);
> > }
> > 
> > Now a new event showed up in /dev/input/
> > Without the xenomai patch there was an interrupt, that I could
listen to
> > with a program (open(/dev/input/event2) and select for listening)
and
> > react to the signal.
> 
> That is one thing. It could be due to a edge vs level issue. First
when
> you tell this, you must be 100% sure that you are exactly using the
same
> configuration except for the xenomai settings. Second, please try and
> recompile a kernel with only CONFIG_IPIPE enabled, but not with
> CONFIG_XENOMAI.

> > With the xenomai patch nothing is coming trough, and even though
GPIO 1
> > still shows up in /proc/interrupts, it doesnt show up
> > in /proc/xenomai/irq.
> 
> That is not your problem. Irqs appear in /proc/xenomai/irq only when
the
> IRQ is used in the real-time domain.
> 
> > I tried to use the native API (I used the usr_irq.c) to create and
wait
> > for the signal, but nothin happend (rt_intr_wait() returned -1).
> 
> That is yet another story. Error codes are documented in the API
> documentation. -1 is -EPERM. It usually means that the task calling
the
> function is not a real-time task and should be one.
> 
> > I would be very happy about every hint in the right direction.
> > greetings,
> > Peter 
> > 
> > 

Re: [Xenomai-help] gpio-keys with xenomai

[Gilles Chanteperdrix]

Peter Breuer wrote:
> Hello, 
> I finally had the time to test some different settings...
> 
> I wrote the following program:
> #include <sys/mman.h>
> #include <native/task.h>
> #include <native/intr.h>
> 
> #define IRQ_NUMBER 100  /* Intercept interrupt */
> #define IRQ_MODE   I_NOAUTOENA //I_PROPAGATE //0
> #define TASK_PRIO  99 /* Highest RT priority */
> #define TASK_MODE  T_FPU  /*  */
> #define TASK_STKSZ 0  /* Stack size (use default one) */
> #define PIPE_MINOR 0
> 
> RT_INTR intr_desc;
> 
> RT_TASK server_desc;
> 
> 
> void irq_server (void *cookie)
> 
> {
>         int err =
> rt_intr_create(&intr_desc,"MyIrq",IRQ_NUMBER,IRQ_MODE);
>         printf("int create: %d\n", err);
>         err = rt_intr_enable(&intr_desc);
>         printf("enable int: %d\n", err);
>         printf("begin..\n");
>         err = rt_intr_wait(&intr_desc,TM_INFINITE);
>         printf("int wait:%d\n", err);
> 
> }
> 
> int main (int argc, char *argv[])
> 
> {
>         int err;
>         int a;
>         printf("Programmstart\n");
>         mlockall(MCL_CURRENT|MCL_FUTURE);
> 
>         err = rt_task_create(&server_desc,
>                                 "MyIrqServer",
>                                 TASK_STKSZ,
>                                 TASK_PRIO,
>                                 TASK_MODE);
>         printf("task create: %d\n", err);
>         rt_task_start(&server_desc,&irq_server,NULL);
> }
> 
> void cleanup (void)
> 
> {
>     rt_intr_delete(&intr_desc);
>     rt_task_delete(&server_desc);
> }
> 
> When I start it with the compiled xenomai, it runs until the
> rt_intr_wait() and then quits, without any message. It doesnt even
> execute printf("int wait:%d\n", err)
> create_task, create_intr and enable_intr all return 0. 
> When I start the programm with a kernel, where xenomai isnt compiled in,
> then it quits and complains about it. 
> When I start the programm with xenomai debug settings compiled in the
> kernel, the system freezes at task creating. Heres the output:
> Programmstart
> taskXenomai: fatal: bug at kernel/xenomai/nucleus/shadow.c
> :1128 (!irqs_disabled_hw())
> CPU PID PRI TIMEOUT STAT NAME
> 0 0 -1 0 00400088 ROOT
>> 0 923 99 0 00300180 MyIrqServer
> Master time base: clock=114327768
> [<c002bde4>] (show_stack+0x0/0x48) from [<c006e69c>] (xnshadow_harden
> +0x1d0/0x21c)
> [<c006e4cc>] (xnshadow_harden+0x0/0x21c) from [<c006ec00>]
> (xnshadow_wait_barrier+0x124/0x138)
> [<c006eadc>] (xnshadow_wait_barrier+0x0/0x138) from [<c006fbe4>]
> (xnshadow_sys_barrier+0x14/0x18)
> r6:c309a000 r5:c309bfb0 r4:00000001
> [<c006fbd0>] (xnshadow_sys_barrier+0x0/0x18) from [<c006ff70>]
> (losyscall_event+0xb4/0x1ac)
> [<c006febc>] (losyscall_event+0x0/0x1ac) from [<c0061214>]
> (__ipipe_dispatch_event+0xd0/0x1c8)
> [<c0061144>] (__ipipe_dispatch_event+0x0/0x1c8) from [<c002cc5c>]
> (__ipipe_syscall_root+0x80/0x100)
> [<c002cbdc>] (__ipipe_syscall_root+0x0/0x100) from [<c0028108>]
> (vector_swi+0x68/0xa8)
> 
> Any ideas? 
> is the irq-number, that I need for the rt_intr_create() the same number
> that is shown in /proc/interrupts?
> thx, greetings, 
> Peter

You are using a too old patch. Please use an I-pipe patch for linux
2.6.28 or above.



-- 
					    Gilles.

[Xenomai-help] gpio-keys with xenomai

[Peter Breuer]

Hello everyone,
I am new to xenomai, so please be patient.
I am working with an AT91SAM9621S (Board SBC6000X) and I successfully
patched my linux 2.6.24 with xenomai 2.4.9.1. I also have a signal on on
of the GPIOs and I want a program to react, every time the signal
arrives (~10Hz).
Before compiling I added some lines to the board description file
(arch/arm/mach-at91/board-sbc9621.c):
...
static struct gpio_keys_button sbc9261_buttons[] = {
...
//begin changes
    {
        .gpio        = AT91_PIN_PC4,
        .code        = BTN_2,
        .desc        = "GPIO 1",
        .active_low    = 1,
    },
//end changes
};
...
static void __init sbc9261_add_device_buttons(void)
{
...
//begin changes
    at91_set_gpio_input(AT91_PIN_PC4, 1);    /* btn1 */
    at91_set_deglitch(AT91_PIN_PC4, 1);
//end changes

    platform_device_register(&sbc9261_button_device);
}

Now a new event showed up in /dev/input/
Without the xenomai patch there was an interrupt, that I could listen to
with a program (open(/dev/input/event2) and select for listening) and
react to the signal.
With the xenomai patch nothing is coming trough, and even though GPIO 1
still shows up in /proc/interrupts, it doesnt show up
in /proc/xenomai/irq.
I tried to use the native API (I used the usr_irq.c) to create and wait
for the signal, but nothin happend (rt_intr_wait() returned -1).
I would be very happy about every hint in the right direction.
greetings,
Peter 

Re: [Xenomai-help] gpio-keys with xenomai

[Gilles Chanteperdrix]

Peter Breuer wrote:
> Hello everyone,
> I am new to xenomai, so please be patient.
> I am working with an AT91SAM9621S (Board SBC6000X) and I successfully
> patched my linux 2.6.24 with xenomai 2.4.9.1. I also have a signal on on
> of the GPIOs and I want a program to react, every time the signal
> arrives (~10Hz).
> Before compiling I added some lines to the board description file
> (arch/arm/mach-at91/board-sbc9621.c):
> ...
> static struct gpio_keys_button sbc9261_buttons[] = {
> ...
> //begin changes
>     {
>         .gpio        = AT91_PIN_PC4,
>         .code        = BTN_2,
>         .desc        = "GPIO 1",
>         .active_low    = 1,
>     },
> //end changes
> };
> ...
> static void __init sbc9261_add_device_buttons(void)
> {
> ...
> //begin changes
>     at91_set_gpio_input(AT91_PIN_PC4, 1);    /* btn1 */
>     at91_set_deglitch(AT91_PIN_PC4, 1);
> //end changes
> 
>     platform_device_register(&sbc9261_button_device);
> }
> 
> Now a new event showed up in /dev/input/
> Without the xenomai patch there was an interrupt, that I could listen to
> with a program (open(/dev/input/event2) and select for listening) and
> react to the signal.

That is one thing. It could be due to a edge vs level issue. First when
you tell this, you must be 100% sure that you are exactly using the same
configuration except for the xenomai settings. Second, please try and
recompile a kernel with only CONFIG_IPIPE enabled, but not with
CONFIG_XENOMAI.

> With the xenomai patch nothing is coming trough, and even though GPIO 1
> still shows up in /proc/interrupts, it doesnt show up
> in /proc/xenomai/irq.

That is not your problem. Irqs appear in /proc/xenomai/irq only when the
IRQ is used in the real-time domain.

> I tried to use the native API (I used the usr_irq.c) to create and wait
> for the signal, but nothin happend (rt_intr_wait() returned -1).

That is yet another story. Error codes are documented in the API
documentation. -1 is -EPERM. It usually means that the task calling the
function is not a real-time task and should be one.

> I would be very happy about every hint in the right direction.
> greetings,
> Peter 
> 
> 
> 
> _______________________________________________
> Xenomai-help mailing list
> Xenomai-help@domain.hid
> https://mail.gna.org/listinfo/xenomai-help
> 


-- 
Gilles Chanteperdrix, Free Electrons
Kernel, drivers, real-time and embedded Linux
development, consulting, training and support.
http://free-electrons.com

Re: [Xenomai-help] gpio-keys with xenomai

[Philippe Gerum]

On Wed, 2010-03-17 at 23:35 +0100, Peter Breuer wrote:
> Hello everyone,
> I am new to xenomai, so please be patient.
> I am working with an AT91SAM9621S (Board SBC6000X) and I successfully
> patched my linux 2.6.24 with xenomai 2.4.9.1. I also have a signal on on
> of the GPIOs and I want a program to react, every time the signal
> arrives (~10Hz).
> Before compiling I added some lines to the board description file
> (arch/arm/mach-at91/board-sbc9621.c):
> ...
> static struct gpio_keys_button sbc9261_buttons[] = {
> ...
> //begin changes
>     {
>         .gpio        = AT91_PIN_PC4,
>         .code        = BTN_2,
>         .desc        = "GPIO 1",
>         .active_low    = 1,
>     },
> //end changes
> };
> ...
> static void __init sbc9261_add_device_buttons(void)
> {
> ...
> //begin changes
>     at91_set_gpio_input(AT91_PIN_PC4, 1);    /* btn1 */
>     at91_set_deglitch(AT91_PIN_PC4, 1);
> //end changes
> 
>     platform_device_register(&sbc9261_button_device);
> }
> 
> Now a new event showed up in /dev/input/
> Without the xenomai patch there was an interrupt, that I could listen to
> with a program (open(/dev/input/event2) and select for listening) and
> react to the signal.
> With the xenomai patch nothing is coming trough, and even though GPIO 1
> still shows up in /proc/interrupts, it doesnt show up
> in /proc/xenomai/irq.
> I tried to use the native API (I used the usr_irq.c) to create and wait
> for the signal, but nothin happend (rt_intr_wait() returned -1).
> I would be very happy about every hint in the right direction.

rt_intr_wait() should be called from a real-time task, obtained from
rt_task_create/spawn; failure to do so causes the Xenomai core to
dismiss your request with -EPERM.

There is a simple rule of thumb to know whether a Xenomai syscall shall
be called from a real-time task: if a Xenomai service may block the
caller, or has an effect on some of its dynamic properties (like its
real-time scheduling priority, some of its runtime options or current
execution mode within the Xenomai domain etc.), then it may only be
called on behalf of a real-time task.

Additionally, you should call rt_intr_enable() to explicitly unmask the
interrupt channel, since it is left masked upon creation. Maybe that is
what is missing (user_irq.c is wrong in this respect) to get the first
IRQ.

> greetings,
> Peter 
> 
> 
> 
> _______________________________________________
> Xenomai-help mailing list
> Xenomai-help@domain.hid
> https://mail.gna.org/listinfo/xenomai-help


-- 
Philippe.