[Xenomai] Xenomai 3.0-rc3 Interruption on Arietta G25 (AT91SAM9G25 chip)

[Xenomai] Xenomai 3.0-rc3 Interruption on Arietta G25 (AT91SAM9G25 chip)

[Tianchi Huang]

*Hello,*

*I'm now working on project using Arietta G25 board (AT91SAM9G25 chip) on
which i need to capture 1000 interruptions per second of an external ADC. *
*I have patched the kernel 3.14.17 with the patch ipipe-core-3.14.17-arm-5
which i found on GNA. The compilation passed and Xenomai seems to run all
right. The latency test gave me about 100us as worst value. *

*I make un modele of interruption test as below*
#include <linux/fs.h>
#include <linux/fcntl.h>
#include <linux/gpio.h>
#include <linux/interrupt.h>
#include <linux/module.h>
#include <rtdm/driver.h>

#define GPIO_IN  27
static rtdm_irq_t irq_rtdm;
static int handler_interruption(rtdm_irq_t * irq){
printk(KERN_ALERT "IRQ received!\n");
return RTDM_IRQ_HANDLED;
}

static int __init exemple_init (void){
int err;
int num_interruption;
if ((err = gpio_request(GPIO_IN, "test_irq")) != 0) {
return err;
}
if ((err = gpio_direction_input(GPIO_IN)) != 0) {
gpio_free(GPIO_IN);
return err;
}
numero_interruption = gpio_to_irq(GPIO_IN);
irq_set_irq_type(numero_interruption, IRQF_TRIGGER_FALLING);
if ((err = rtdm_irq_request(& irq_rtdm, num_interruption,
handler_interruption, RTDM_IRQTYPE_EDGE, "test_irq", NULL)) != 0) {
gpio_free(GPIO_IN);
return err;
}
printk(KERN_ALERT "Driver registed\n");
rtdm_irq_enable(& irq_rtdm);
return 0;
}

static void __exit exemple_exit (void){
rtdm_irq_disable(& irq_rtdm);
rtdm_irq_free(&irq_rtdm);
gpio_free(GPIO_IN);
}
module_init(exemple_init);
module_exit(exemple_exit);
MODULE_LICENSE("GPL");


*It seems that it has been registed and run without no error, as you can
see just here:*
root@arietta:/home# cat /proc/xenomai/irq
  IRQ         CPU0
   17:           1         [timer/0]
   28:           0         test_irq
 1027:           0         [virtual]
root@arietta:/home# cat /proc/interrupts
           CPU0
 16:      31031       AIC   1  at91_tick, at91_rtc, ttyS0
 18:       1579       AIC  20  at_hdmac
 19:          0       AIC  21  at_hdmac
 23:      15338       AIC  12  f0008000.mmc
 25:          0       AIC  14  f0004000.spi
 26:      20732       AIC  23  atmel_usba_udc
 27:       8119       AIC  22  ehci_hcd:usb1, ohci_hcd:usb2
 28:          1      GPIO  27
Err:          0

*But, when i generate a signal on GPIO 27, i can see the counter of
/proc/interrupts increase but not the counter of /proc/xenomai/irq nor the
message i put in the interrupt handler, as you can see below:*

root@arietta:/home# cat /proc/xenomai/irq
  IRQ         CPU0
   17:           1         [timer/0]
   28:           0         test_irq
 1027:           0         [virtual]
root@arietta:/home# cat /proc/interrupts
           CPU0
 16:      31031       AIC   1  at91_tick, at91_rtc, ttyS0
 18:       1579       AIC  20  at_hdmac
 19:          0       AIC  21  at_hdmac
 23:      15338       AIC  12  f0008000.mmc
 25:          0       AIC  14  f0004000.spi
 26:      20732       AIC  23  atmel_usba_udc
 27:       8119       AIC  22  ehci_hcd:usb1, ohci_hcd:usb2
 28:       5459      GPIO  27
Err:          0

*Do you have any idea of that? I've tried Xenomai 2.6.4 too, but i had the
same problem. Hope you can help me with this.*

*Best regards,*



*Eddie HUANG*

Re: [Xenomai] Xenomai 3.0-rc3 Interruption on Arietta G25 (AT91SAM9G25 chip)

[Gilles Chanteperdrix]

On Mon, Mar 23, 2015 at 12:07:43PM +0100, Tianchi Huang wrote:
> *Hello,*
> 
> *I'm now working on project using Arietta G25 board (AT91SAM9G25 chip) on
> which i need to capture 1000 interruptions per second of an external ADC. *
> *I have patched the kernel 3.14.17 with the patch ipipe-core-3.14.17-arm-5
> which i found on GNA. The compilation passed and Xenomai seems to run all
> right. The latency test gave me about 100us as worst value. *
> 
> *I make un modele of interruption test as below*
> #include <linux/fs.h>
> #include <linux/fcntl.h>
> #include <linux/gpio.h>
> #include <linux/interrupt.h>
> #include <linux/module.h>
> #include <rtdm/driver.h>
> 
> #define GPIO_IN  27
> static rtdm_irq_t irq_rtdm;
> static int handler_interruption(rtdm_irq_t * irq){
> printk(KERN_ALERT "IRQ received!\n");
> return RTDM_IRQ_HANDLED;
> }
> 
> static int __init exemple_init (void){
> int err;
> int num_interruption;
> if ((err = gpio_request(GPIO_IN, "test_irq")) != 0) {
> return err;
> }
> if ((err = gpio_direction_input(GPIO_IN)) != 0) {
> gpio_free(GPIO_IN);
> return err;
> }
> numero_interruption = gpio_to_irq(GPIO_IN);
> irq_set_irq_type(numero_interruption, IRQF_TRIGGER_FALLING);
> if ((err = rtdm_irq_request(& irq_rtdm, num_interruption,
> handler_interruption, RTDM_IRQTYPE_EDGE, "test_irq", NULL)) != 0) {
> gpio_free(GPIO_IN);
> return err;
> }
> printk(KERN_ALERT "Driver registed\n");
> rtdm_irq_enable(& irq_rtdm);
> return 0;
> }
> 
> static void __exit exemple_exit (void){
> rtdm_irq_disable(& irq_rtdm);
> rtdm_irq_free(&irq_rtdm);
> gpio_free(GPIO_IN);
> }
> module_init(exemple_init);
> module_exit(exemple_exit);
> MODULE_LICENSE("GPL");
> 
> 
> *It seems that it has been registed and run without no error, as you can
> see just here:*
> root@arietta:/home# cat /proc/xenomai/irq
>   IRQ         CPU0
>    17:           1         [timer/0]
>    28:           0         test_irq
>  1027:           0         [virtual]
> root@arietta:/home# cat /proc/interrupts
>            CPU0
>  16:      31031       AIC   1  at91_tick, at91_rtc, ttyS0
>  18:       1579       AIC  20  at_hdmac
>  19:          0       AIC  21  at_hdmac
>  23:      15338       AIC  12  f0008000.mmc
>  25:          0       AIC  14  f0004000.spi
>  26:      20732       AIC  23  atmel_usba_udc
>  27:       8119       AIC  22  ehci_hcd:usb1, ohci_hcd:usb2
>  28:          1      GPIO  27
> Err:          0
> 
> *But, when i generate a signal on GPIO 27, i can see the counter of
> /proc/interrupts increase but not the counter of /proc/xenomai/irq nor the
> message i put in the interrupt handler, as you can see below:*
> 
> root@arietta:/home# cat /proc/xenomai/irq
>   IRQ         CPU0
>    17:           1         [timer/0]
>    28:           0         test_irq
>  1027:           0         [virtual]
> root@arietta:/home# cat /proc/interrupts
>            CPU0
>  16:      31031       AIC   1  at91_tick, at91_rtc, ttyS0
>  18:       1579       AIC  20  at_hdmac
>  19:          0       AIC  21  at_hdmac
>  23:      15338       AIC  12  f0008000.mmc
>  25:          0       AIC  14  f0004000.spi
>  26:      20732       AIC  23  atmel_usba_udc
>  27:       8119       AIC  22  ehci_hcd:usb1, ohci_hcd:usb2
>  28:       5459      GPIO  27
> Err:          0
> 
> *Do you have any idea of that? I've tried Xenomai 2.6.4 too, but i had the
> same problem. Hope you can help me with this.*

Have you tried calling both rtdm_irq_request and request_irq in the
same driver ?

-- 
					    Gilles.

Re: [Xenomai] Xenomai 3.0-rc3 Interruption on Arietta G25 (AT91SAM9G25 chip)

[Tianchi Huang]

Thank you for your response, Gilles.

I added irq_request in the init function and an other callback handler.
Interruptions have been handled in the new handler, but not the elder. And
there was always the same problem with the counter in /proc/xenomai/irq, it
doesn't increase.

*Tianchi HUANG*

Ingénieur Informatique en Apprentissage

eddie.huang@bioserenity.com

Tel : +33(0) 6 52 86 29 86



BioSerenity

ICM-iPEPS

47-85 boulevard de l'Hôpital

75013 Paris


www.bioserenity.com

2015-03-23 23:09 GMT+01:00 Gilles Chanteperdrix <
gilles.chanteperdrix@xenomai.org>:

> On Mon, Mar 23, 2015 at 12:07:43PM +0100, Tianchi Huang wrote:
> > *Hello,*
> >
> > *I'm now working on project using Arietta G25 board (AT91SAM9G25 chip) on
> > which i need to capture 1000 interruptions per second of an external
> ADC. *
> > *I have patched the kernel 3.14.17 with the patch
> ipipe-core-3.14.17-arm-5
> > which i found on GNA. The compilation passed and Xenomai seems to run all
> > right. The latency test gave me about 100us as worst value. *
> >
> > *I make un modele of interruption test as below*
> > #include <linux/fs.h>
> > #include <linux/fcntl.h>
> > #include <linux/gpio.h>
> > #include <linux/interrupt.h>
> > #include <linux/module.h>
> > #include <rtdm/driver.h>
> >
> > #define GPIO_IN  27
> > static rtdm_irq_t irq_rtdm;
> > static int handler_interruption(rtdm_irq_t * irq){
> > printk(KERN_ALERT "IRQ received!\n");
> > return RTDM_IRQ_HANDLED;
> > }
> >
> > static int __init exemple_init (void){
> > int err;
> > int num_interruption;
> > if ((err = gpio_request(GPIO_IN, "test_irq")) != 0) {
> > return err;
> > }
> > if ((err = gpio_direction_input(GPIO_IN)) != 0) {
> > gpio_free(GPIO_IN);
> > return err;
> > }
> > numero_interruption = gpio_to_irq(GPIO_IN);
> > irq_set_irq_type(numero_interruption, IRQF_TRIGGER_FALLING);
> > if ((err = rtdm_irq_request(& irq_rtdm, num_interruption,
> > handler_interruption, RTDM_IRQTYPE_EDGE, "test_irq", NULL)) != 0) {
> > gpio_free(GPIO_IN);
> > return err;
> > }
> > printk(KERN_ALERT "Driver registed\n");
> > rtdm_irq_enable(& irq_rtdm);
> > return 0;
> > }
> >
> > static void __exit exemple_exit (void){
> > rtdm_irq_disable(& irq_rtdm);
> > rtdm_irq_free(&irq_rtdm);
> > gpio_free(GPIO_IN);
> > }
> > module_init(exemple_init);
> > module_exit(exemple_exit);
> > MODULE_LICENSE("GPL");
> >
> >
> > *It seems that it has been registed and run without no error, as you can
> > see just here:*
> > root@arietta:/home# cat /proc/xenomai/irq
> >   IRQ         CPU0
> >    17:           1         [timer/0]
> >    28:           0         test_irq
> >  1027:           0         [virtual]
> > root@arietta:/home# cat /proc/interrupts
> >            CPU0
> >  16:      31031       AIC   1  at91_tick, at91_rtc, ttyS0
> >  18:       1579       AIC  20  at_hdmac
> >  19:          0       AIC  21  at_hdmac
> >  23:      15338       AIC  12  f0008000.mmc
> >  25:          0       AIC  14  f0004000.spi
> >  26:      20732       AIC  23  atmel_usba_udc
> >  27:       8119       AIC  22  ehci_hcd:usb1, ohci_hcd:usb2
> >  28:          1      GPIO  27
> > Err:          0
> >
> > *But, when i generate a signal on GPIO 27, i can see the counter of
> > /proc/interrupts increase but not the counter of /proc/xenomai/irq nor
> the
> > message i put in the interrupt handler, as you can see below:*
> >
> > root@arietta:/home# cat /proc/xenomai/irq
> >   IRQ         CPU0
> >    17:           1         [timer/0]
> >    28:           0         test_irq
> >  1027:           0         [virtual]
> > root@arietta:/home# cat /proc/interrupts
> >            CPU0
> >  16:      31031       AIC   1  at91_tick, at91_rtc, ttyS0
> >  18:       1579       AIC  20  at_hdmac
> >  19:          0       AIC  21  at_hdmac
> >  23:      15338       AIC  12  f0008000.mmc
> >  25:          0       AIC  14  f0004000.spi
> >  26:      20732       AIC  23  atmel_usba_udc
> >  27:       8119       AIC  22  ehci_hcd:usb1, ohci_hcd:usb2
> >  28:       5459      GPIO  27
> > Err:          0
> >
> > *Do you have any idea of that? I've tried Xenomai 2.6.4 too, but i had
> the
> > same problem. Hope you can help me with this.*
>
> Have you tried calling both rtdm_irq_request and request_irq in the
> same driver ?
>
> --
>                                             Gilles.
>

Re: [Xenomai] Xenomai 3.0-rc3 Interruption on Arietta G25 (AT91SAM9G25 chip)

[Gilles Chanteperdrix]

On Tue, Mar 24, 2015 at 04:55:55PM +0100, Tianchi Huang wrote:
> Thank you for your response, Gilles.
> 
> I added irq_request in the init function and an other callback handler.
> Interruptions have been handled in the new handler, but not the elder. And
> there was always the same problem with the counter in /proc/xenomai/irq, it
> doesn't increase.

Ok, does the latency test run correctly? If yes, is the timer
counter incrementing in /proc/xenomai/irq?

-- 
					    Gilles.

Re: [Xenomai] Xenomai 3.0-rc3 Interruption on Arietta G25 (AT91SAM9G25 chip)

[Gilles Chanteperdrix]

On Tue, Mar 24, 2015 at 05:01:53PM +0100, Gilles Chanteperdrix wrote:
> On Tue, Mar 24, 2015 at 04:55:55PM +0100, Tianchi Huang wrote:
> > Thank you for your response, Gilles.
> > 
> > I added irq_request in the init function and an other callback handler.
> > Interruptions have been handled in the new handler, but not the elder. And
> > there was always the same problem with the counter in /proc/xenomai/irq, it
> > doesn't increase.
> 
> Ok, does the latency test run correctly? If yes, is the timer
> counter incrementing in /proc/xenomai/irq?

Forget about that, please send me your kernel configuration.

-- 
					    Gilles.

Re: [Xenomai] Xenomai 3.0-rc3 Interruption on Arietta G25 (AT91SAM9G25 chip)

[Tianchi Huang]

*the latency test did run correctly and after the test, the timer counter
increase, like this:*

root@arietta:/usr/xenomai/bin# ./latency
== Sampling period: 1000 us
== Test mode: periodic user-mode task
== All results in microseconds
warming up...
RTT|  00:00:01  (periodic user-mode task, 1000 us period, priority 99)
RTH|----lat min|----lat avg|----lat max|-overrun|---msw|---lat best|--lat
worst
RTD|      5.824|     14.176|     65.151|       0|     0|      5.824|
65.151
RTD|      5.806|     13.493|     65.251|       0|     0|      5.806|
65.251
RTD|      5.717|     12.141|     64.391|       0|     0|      5.717|
65.251
RTD|      5.627|     11.203|     65.611|       0|     0|      5.627|
65.611
RTD|      5.904|     15.964|     64.271|       0|     0|      5.627|
65.611
RTD|      5.783|     13.737|     63.711|       0|     0|      5.627|
65.611
^C---|-----------|-----------|-----------|--------|------|-------------------------
RTS|      5.627|     13.452|     65.611|       0|     0|
 00:00:06/00:00:06
root@arietta:/usr/xenomai/bin# cat /proc/xenomai/irq
  IRQ         CPU0
   17:        7616         [timer/0]
 1027:          16         [virtual]
root@arietta:/usr/xenomai/bin#

Here is my .config, for kernel 3.14.17.

Best regards,

*Tianchi HUANG*

Ingénieur Informatique en Apprentissage

eddie.huang@bioserenity.com

Tel : +33(0) 6 52 86 29 86



BioSerenity

ICM-iPEPS

47-85 boulevard de l'Hôpital

75013 Paris


www.bioserenity.com

2015-03-24 17:13 GMT+01:00 Gilles Chanteperdrix <
gilles.chanteperdrix@xenomai.org>:

> On Tue, Mar 24, 2015 at 05:01:53PM +0100, Gilles Chanteperdrix wrote:
> > On Tue, Mar 24, 2015 at 04:55:55PM +0100, Tianchi Huang wrote:
> > > Thank you for your response, Gilles.
> > >
> > > I added irq_request in the init function and an other callback handler.
> > > Interruptions have been handled in the new handler, but not the elder.
> And
> > > there was always the same problem with the counter in
> /proc/xenomai/irq, it
> > > doesn't increase.
> >
> > Ok, does the latency test run correctly? If yes, is the timer
> > counter incrementing in /proc/xenomai/irq?
>
> Forget about that, please send me your kernel configuration.
>
> --
>                                             Gilles.
>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: .config
Type: application/xml
Size: 68993 bytes
Desc: not available
URL: <http://www.xenomai.org/pipermail/xenomai/attachments/20150325/3bf0397a/attachment.wsdl>

Re: [Xenomai] Xenomai 3.0-rc3 Interruption on Arietta G25 (AT91SAM9G25 chip)

[Gilles Chanteperdrix]

On Thu, Mar 26, 2015 at 03:21:31PM +0100, Tianchi Huang wrote:
> Hello Gilles,
> 
> Did you find something about that bug?

No. I did not have a look yet. Very much likely there is a piece of
code in the gpio code that calls generic_handle_irq instead of
ipipe_handle_demuxed_irq, however from what I can tell the at91 gpio
driver does not have this issue. You are using a vanilla kernel,
right, not a patched one ?

-- 
					    Gilles.

Re: [Xenomai] Xenomai 3.0-rc3 Interruption on Arietta G25 (AT91SAM9G25 chip)

[Tianchi Huang]

Yes, I downloaded the vanilla kernel from kernel.org and then I patched it
directly with ipipe. What kinds of patches I need to patch?
I've looked into the /arch/arm/mach-at91/irq.c
and /arch/arm/mach-at91/gipo.c, but i can't find any generic_handle_irq but
ipipe_handle_multi_irq and ipipe_handle_demuxed_irq.

Best regards,

*Eddie HUANG*



2015-03-26 15:31 GMT+01:00 Gilles Chanteperdrix <
gilles.chanteperdrix@xenomai.org>:

> On Thu, Mar 26, 2015 at 03:21:31PM +0100, Tianchi Huang wrote:
> > Hello Gilles,
> >
> > Did you find something about that bug?
>
> No. I did not have a look yet. Very much likely there is a piece of
> code in the gpio code that calls generic_handle_irq instead of
> ipipe_handle_demuxed_irq, however from what I can tell the at91 gpio
> driver does not have this issue. You are using a vanilla kernel,
> right, not a patched one ?
>
> --
>                                             Gilles.
>

Re: [Xenomai] Xenomai 3.0-rc3 Interruption on Arietta G25 (AT91SAM9G25 chip)

[Gilles Chanteperdrix]

On Thu, Mar 26, 2015 at 03:52:06PM +0100, Tianchi Huang wrote:
> Yes, I downloaded the vanilla kernel from kernel.org and then I patched it
> directly with ipipe. What kinds of patches I need to patch?
> I've looked into the /arch/arm/mach-at91/irq.c
> and /arch/arm/mach-at91/gipo.c, but i can't find any generic_handle_irq but
> ipipe_handle_multi_irq and ipipe_handle_demuxed_irq.

Yes, this is what I tell you. Maybe some other code is compiled for
your board, but I do not find anything called at91 in drivers/gpio
either. Which is why I asked for your kernel configuration to see
what GPIO driver is compiled.

-- 
					    Gilles.

Re: [Xenomai] Xenomai 3.0-rc3 Interruption on Arietta G25 (AT91SAM9G25 chip)

[Tianchi Huang]

After kernel compilation, the following .o files have been created:
- at91_ipipe.o
- at91sam9_alt_reset.o
- at91sam9g45.o
- at91sam9g45_reset.o
- at91sam9n12.o
- at91sam9rl.o
- at91sam9x5.o
- at91sam926x_time.o
- at91sam9260.o
- at91sam9261.o
- at91sam9263.o
- board-dt-sam9.o
- sam9_smc.o
- gpio.o
- clock.o
- irq.o
- leds.o
- pm.o
- setup.o
- sysirq_mask.o



*Tianchi HUANG*

Ingénieur Informatique en Apprentissage

eddie.huang@bioserenity.com

Tel : +33(0) 6 52 86 29 86



BioSerenity

ICM-iPEPS

47-85 boulevard de l'Hôpital

75013 Paris


www.bioserenity.com

2015-03-26 15:54 GMT+01:00 Gilles Chanteperdrix <
gilles.chanteperdrix@xenomai.org>:

> On Thu, Mar 26, 2015 at 03:52:06PM +0100, Tianchi Huang wrote:
> > Yes, I downloaded the vanilla kernel from kernel.org and then I patched
> it
> > directly with ipipe. What kinds of patches I need to patch?
> > I've looked into the /arch/arm/mach-at91/irq.c
> > and /arch/arm/mach-at91/gipo.c, but i can't find any generic_handle_irq
> but
> > ipipe_handle_multi_irq and ipipe_handle_demuxed_irq.
>
> Yes, this is what I tell you. Maybe some other code is compiled for
> your board, but I do not find anything called at91 in drivers/gpio
> either. Which is why I asked for your kernel configuration to see
> what GPIO driver is compiled.
>
> --
>                                             Gilles.
>

Re: [Xenomai] Xenomai 3.0-rc3 Interruption on Arietta G25 (AT91SAM9G25 chip)

[Tianchi Huang]

*If i call request_irq and rtdm_irq_request, i've got this warning when the
irq comes in :*

------------[ cut here ]------------
WARNING: CPU: 0 PID: 629 at kernel/irq/handle.c:146
handle_irq_event_percpu+0x70/0x1c0()
irq 28 handler irq_handler+0x0/0x18 [BioS_spi] enabled interrupts
Modules linked in: BioS_spi(O) usb_f_eem g_ether usb_f_rndis u_ether
libcomposite atmel_usba_udc ohci_at91 ohci_hcd
CPU: 0 PID: 629 Comm: udevd Tainted: G           O 3.14.17 #3
[<c000dad4>] (unwind_backtrace) from [<c000b954>] (show_stack+0x10/0x14)
[<c000b954>] (show_stack) from [<c0018b54>] (warn_slowpath_common+0x60/0x80)
[<c0018b54>] (warn_slowpath_common) from [<c0018bf4>]
(warn_slowpath_fmt+0x2c/0x3c)
[<c0018bf4>] (warn_slowpath_fmt) from [<c0042490>]
(handle_irq_event_percpu+0x70/0x1c0)
[<c0042490>] (handle_irq_event_percpu) from [<c0042608>]
(handle_irq_event+0x28/0x38)
[<c0042608>] (handle_irq_event) from [<c004484c>]
(handle_simple_irq+0x88/0x90)
[<c004484c>] (handle_simple_irq) from [<c0041dbc>]
(generic_handle_irq+0x20/0x30)
[<c0041dbc>] (generic_handle_irq) from [<c019681c>]
(gpio_irq_handler+0x98/0xd0)
[<c019681c>] (gpio_irq_handler) from [<c005913c>]
(__ipipe_dispatch_irq+0x7c/0x258)
[<c005913c>] (__ipipe_dispatch_irq) from [<c0008624>]
(__ipipe_grab_irq+0x58/0x70)
[<c0008624>] (__ipipe_grab_irq) from [<c000c4a0>] (__irq_svc+0x40/0x4c)
Exception stack(0xcfac1e00 to 0xcfac1e48)
1e00: cf401038 00000001 cfa7fc00 00000011 cfac1ef0 00000000 ceda001d
cfac1ef8
1e20: cfac0000 ceda001c ceda001d cfac1e78 00000000 cfac1e48 c00b34f0
c00b320c
1e40: 60000013 ffffffff
[<c000c4a0>] (__irq_svc) from [<c00b320c>] (__inode_permission+0x8/0x58)
[<c00b320c>] (__inode_permission) from [<c00b34f0>]
(link_path_walk+0x4c/0x724)
[<c00b34f0>] (link_path_walk) from [<c00b6700>] (path_openat+0x1c8/0x524)
[<c00b6700>] (path_openat) from [<c00b6d58>] (do_filp_open+0x2c/0x78)
[<c00b6d58>] (do_filp_open) from [<c00a9f08>] (do_sys_open+0x11c/0x1bc)
[<c00a9f08>] (do_sys_open) from [<c0009620>] (ret_fast_syscall+0x0/0x2c)
---[ end trace 68f2788b079d4fa7 ]---



*Tianchi HUANG*

Ingénieur Informatique en Apprentissage

eddie.huang@bioserenity.com

Tel : +33(0) 6 52 86 29 86



BioSerenity

ICM-iPEPS

47-85 boulevard de l'Hôpital

75013 Paris


www.bioserenity.com

2015-03-26 15:54 GMT+01:00 Gilles Chanteperdrix <
gilles.chanteperdrix@xenomai.org>:

> On Thu, Mar 26, 2015 at 03:52:06PM +0100, Tianchi Huang wrote:
> > Yes, I downloaded the vanilla kernel from kernel.org and then I patched
> it
> > directly with ipipe. What kinds of patches I need to patch?
> > I've looked into the /arch/arm/mach-at91/irq.c
> > and /arch/arm/mach-at91/gipo.c, but i can't find any generic_handle_irq
> but
> > ipipe_handle_multi_irq and ipipe_handle_demuxed_irq.
>
> Yes, this is what I tell you. Maybe some other code is compiled for
> your board, but I do not find anything called at91 in drivers/gpio
> either. Which is why I asked for your kernel configuration to see
> what GPIO driver is compiled.
>
> --
>                                             Gilles.
>

Re: [Xenomai] Xenomai 3.0-rc3 Interruption on Arietta G25 (AT91SAM9G25 chip)

[Gilles Chanteperdrix]

On Thu, Mar 26, 2015 at 04:40:58PM +0100, Tianchi Huang wrote:
> [<c0041dbc>] (generic_handle_irq) from [<c019681c>]
> (gpio_irq_handler+0x98/0xd0)
> [<c019681c>] (gpio_irq_handler) from [<c005913c>]
> (__ipipe_dispatch_irq+0x7c/0x258)

As I explained, this is the problem.

Could you search with objdump what file defines gpio_irq_handler ?

-- 
					    Gilles.

Re: [Xenomai] Xenomai 3.0-rc3 Interruption on Arietta G25 (AT91SAM9G25 chip)

[Tianchi Huang]

The function gpio_irq_handler is defined in gpio.c, but i can't find
generic_handle_irq in it.

*Tianchi HUANG*

Ingénieur Informatique en Apprentissage

eddie.huang@bioserenity.com

Tel : +33(0) 6 52 86 29 86



BioSerenity

ICM-iPEPS

47-85 boulevard de l'Hôpital

75013 Paris


www.bioserenity.com

2015-03-26 16:44 GMT+01:00 Gilles Chanteperdrix <
gilles.chanteperdrix@xenomai.org>:

> On Thu, Mar 26, 2015 at 04:40:58PM +0100, Tianchi Huang wrote:
> > [<c0041dbc>] (generic_handle_irq) from [<c019681c>]
> > (gpio_irq_handler+0x98/0xd0)
> > [<c019681c>] (gpio_irq_handler) from [<c005913c>]
> > (__ipipe_dispatch_irq+0x7c/0x258)
>
> As I explained, this is the problem.
>
> Could you search with objdump what file defines gpio_irq_handler ?
>
> --
>                                             Gilles.
>

Re: [Xenomai] Xenomai 3.0-rc3 Interruption on Arietta G25 (AT91SAM9G25 chip)

[Gilles Chanteperdrix]

On Thu, Mar 26, 2015 at 04:58:36PM +0100, Tianchi Huang wrote:
> The function gpio_irq_handler is defined in gpio.c, but i can't find
> generic_handle_irq in it.

You confirmed that with the binutils ?

-- 
					    Gilles.

Re: [Xenomai] Xenomai 3.0-rc3 Interruption on Arietta G25 (AT91SAM9G25 chip)

[Tianchi Huang]

eddie@eddie-VirtualBox:~/Desktop/xenomai-3/linux-3.14.17/arch/arm/mach-at91$
objdump -t *.o | grep gpio_irq_handler -H
(standard input):0000c17c l     F .text 000000d4 gpio_irq_handler
(standard input):0018db84 l     F .text 000000d0 gpio_irq_handler
eddie@eddie-VirtualBox:~/Desktop/xenomai-3/linux-3.14.17/arch/arm/mach-at91$
objdump -t *.o > log

*The first one is defined in gpio.c, the second one is in built-in.o*

*Tianchi HUANG*

Ingénieur Informatique en Apprentissage

eddie.huang@bioserenity.com

Tel : +33(0) 6 52 86 29 86



BioSerenity

ICM-iPEPS

47-85 boulevard de l'Hôpital

75013 Paris


www.bioserenity.com

2015-03-26 17:00 GMT+01:00 Gilles Chanteperdrix <
gilles.chanteperdrix@xenomai.org>:

> On Thu, Mar 26, 2015 at 04:58:36PM +0100, Tianchi Huang wrote:
> > The function gpio_irq_handler is defined in gpio.c, but i can't find
> > generic_handle_irq in it.
>
> You confirmed that with the binutils ?
>
> --
>                                             Gilles.
>

Re: [Xenomai] Xenomai 3.0-rc3 Interruption on Arietta G25 (AT91SAM9G25 chip)

[Gilles Chanteperdrix]

On Thu, Mar 26, 2015 at 05:26:57PM +0100, Tianchi Huang wrote:
> eddie@eddie-VirtualBox:~/Desktop/xenomai-3/linux-3.14.17/arch/arm/mach-at91$
> objdump -t *.o | grep gpio_irq_handler -H
> (standard input):0000c17c l     F .text 000000d4 gpio_irq_handler
> (standard input):0018db84 l     F .text 000000d0 gpio_irq_handler
> eddie@eddie-VirtualBox:~/Desktop/xenomai-3/linux-3.14.17/arch/arm/mach-at91$
> objdump -t *.o > log

Look for the symbol gpio_irq_handler in the vmlinux file, then run
addr2line to see where the symbol comes from.

-- 
					    Gilles.

Re: [Xenomai] Xenomai 3.0-rc3 Interruption on Arietta G25 (AT91SAM9G25 chip)

[Gilles Chanteperdrix]

On Thu, Mar 26, 2015 at 05:35:48PM +0100, Gilles Chanteperdrix wrote:
> On Thu, Mar 26, 2015 at 05:26:57PM +0100, Tianchi Huang wrote:
> > eddie@eddie-VirtualBox:~/Desktop/xenomai-3/linux-3.14.17/arch/arm/mach-at91$
> > objdump -t *.o | grep gpio_irq_handler -H
> > (standard input):0000c17c l     F .text 000000d4 gpio_irq_handler
> > (standard input):0018db84 l     F .text 000000d0 gpio_irq_handler
> > eddie@eddie-VirtualBox:~/Desktop/xenomai-3/linux-3.14.17/arch/arm/mach-at91$
> > objdump -t *.o > log
> 
> Look for the symbol gpio_irq_handler in the vmlinux file, then run
> addr2line to see where the symbol comes from.

Actually you already have the address of the problematic
gpio_irq_handler in the backtrace you sent. If you did not change
any kernel option, it is still valid, so you just need to run
addr2line against the vmlinux binary to find where the function is
defined. 

-- 
					    Gilles.

Re: [Xenomai] Xenomai 3.0-rc3 Interruption on Arietta G25 (AT91SAM9G25 chip)

[Tianchi Huang]

eddie@eddie-VirtualBox:~/Desktop/xenomai-3/linux-3.14.17$
arm-linux-gnueabi-objdump -t vmlinux | grep gpio_irq_handler
c0014d7c l     F .text 000000d4 gpio_irq_handler
c0196784 l     F .text 000000d0 gpio_irq_handler
eddie@eddie-VirtualBox:~/Desktop/xenomai-3/linux-3.14.17$
arm-linux-gnueabi-addr2line 0xc0194784 -e vmlinux -f
$d
pinctrl-utils.c:?
eddie@eddie-VirtualBox:~/Desktop/xenomai-3/linux-3.14.17$
arm-linux-gnueabi-addr2line 0xc0014d7c -e vmlinux -f
gpio_irq_handler
gpio.c:?

*I found nothing in pinctrl-utils.c*


*Tianchi HUANG*

Ingénieur Informatique en Apprentissage

eddie.huang@bioserenity.com

Tel : +33(0) 6 52 86 29 86



BioSerenity

ICM-iPEPS

47-85 boulevard de l'Hôpital

75013 Paris


www.bioserenity.com

2015-03-26 17:35 GMT+01:00 Gilles Chanteperdrix <
gilles.chanteperdrix@xenomai.org>:

> On Thu, Mar 26, 2015 at 05:26:57PM +0100, Tianchi Huang wrote:
> > eddie@eddie-VirtualBox
> :~/Desktop/xenomai-3/linux-3.14.17/arch/arm/mach-at91$
> > objdump -t *.o | grep gpio_irq_handler -H
> > (standard input):0000c17c l     F .text 000000d4 gpio_irq_handler
> > (standard input):0018db84 l     F .text 000000d0 gpio_irq_handler
> > eddie@eddie-VirtualBox
> :~/Desktop/xenomai-3/linux-3.14.17/arch/arm/mach-at91$
> > objdump -t *.o > log
>
> Look for the symbol gpio_irq_handler in the vmlinux file, then run
> addr2line to see where the symbol comes from.
>
> --
>                                             Gilles.
>

Re: [Xenomai] Xenomai 3.0-rc3 Interruption on Arietta G25 (AT91SAM9G25 chip)

[Gilles Chanteperdrix]

On Thu, Mar 26, 2015 at 05:56:05PM +0100, Tianchi Huang wrote:
> eddie@eddie-VirtualBox:~/Desktop/xenomai-3/linux-3.14.17$
> arm-linux-gnueabi-objdump -t vmlinux | grep gpio_irq_handler
> c0014d7c l     F .text 000000d4 gpio_irq_handler
> c0196784 l     F .text 000000d0 gpio_irq_handler
> eddie@eddie-VirtualBox:~/Desktop/xenomai-3/linux-3.14.17$
> arm-linux-gnueabi-addr2line 0xc0194784 -e vmlinux -f
> $d
> pinctrl-utils.c:?
> eddie@eddie-VirtualBox:~/Desktop/xenomai-3/linux-3.14.17$
> arm-linux-gnueabi-addr2line 0xc0014d7c -e vmlinux -f
> gpio_irq_handler
> gpio.c:?
> 
> *I found nothing in pinctrl-utils.c*

Please enable compilation with debug symbols.

-- 
					    Gilles.

Re: [Xenomai] Xenomai 3.0-rc3 Interruption on Arietta G25 (AT91SAM9G25 chip)

[Tianchi Huang]

*Found it !*

eddie@eddie-VirtualBox:~/Desktop/xenomai-3/linux-3.14.17$
arm-linux-gnueabi-addr2line 0xc019681c -e vmlinux -f
gpio_irq_handler
/home/eddie/Desktop/xenomai-3/linux-3.14.17/drivers/pinctrl/pinctrl-at91.c:1433
(discriminator 2)

*0xc019681c shows up in the earlier bug message. *





*Tianchi HUANG*

Ingénieur Informatique en Apprentissage

eddie.huang@bioserenity.com

Tel : +33(0) 6 52 86 29 86



BioSerenity

ICM-iPEPS

47-85 boulevard de l'Hôpital

75013 Paris


www.bioserenity.com

2015-03-26 17:57 GMT+01:00 Gilles Chanteperdrix <
gilles.chanteperdrix@xenomai.org>:

> On Thu, Mar 26, 2015 at 05:56:05PM +0100, Tianchi Huang wrote:
> > eddie@eddie-VirtualBox:~/Desktop/xenomai-3/linux-3.14.17$
> > arm-linux-gnueabi-objdump -t vmlinux | grep gpio_irq_handler
> > c0014d7c l     F .text 000000d4 gpio_irq_handler
> > c0196784 l     F .text 000000d0 gpio_irq_handler
> > eddie@eddie-VirtualBox:~/Desktop/xenomai-3/linux-3.14.17$
> > arm-linux-gnueabi-addr2line 0xc0194784 -e vmlinux -f
> > $d
> > pinctrl-utils.c:?
> > eddie@eddie-VirtualBox:~/Desktop/xenomai-3/linux-3.14.17$
> > arm-linux-gnueabi-addr2line 0xc0014d7c -e vmlinux -f
> > gpio_irq_handler
> > gpio.c:?
> >
> > *I found nothing in pinctrl-utils.c*
>
> Please enable compilation with debug symbols.
>
> --
>                                             Gilles.
>

Re: [Xenomai] Xenomai 3.0-rc3 Interruption on Arietta G25 (AT91SAM9G25 chip)

[Gilles Chanteperdrix]

On Thu, Mar 26, 2015 at 03:21:31PM +0100, Tianchi Huang wrote:
> Hello Gilles,
> 
> Did you find something about that bug?

Next mail to my private address will not be answered.

-- 
					    Gilles.

Re: [Xenomai] Xenomai 3.0-rc3 Interruption on Arietta G25 (AT91SAM9G25 chip)

[Paul Janzen]

Gilles Chanteperdrix <gilles.chanteperdrix@xenomai.org> writes:
>
> Next mail to my private address will not be answered.
>

Neither your MUA nor the mailing list software insert a

Reply-To: xenomai@xenomai.org

header into messages coming from your address. If you want replies to
go to the list, why not use the provided mechanism to say that?

-- Paul

Re: [Xenomai] Xenomai 3.0-rc3 Interruption on Arietta G25 (AT91SAM9G25 chip)

[Gilles Chanteperdrix]

On Thu, Mar 26, 2015 at 10:08:30AM -0700, Paul Janzen wrote:
> Gilles Chanteperdrix <gilles.chanteperdrix@xenomai.org> writes:
> >
> > Next mail to my private address will not be answered.
> >
> 
> Neither your MUA nor the mailing list software insert a
> 
> Reply-To: xenomai@xenomai.org

You should get a resent-to.

> 
> header into messages coming from your address. If you want replies to
> go to the list, why not use the provided mechanism to say that?

Because I already replied privately, and I want to bounce the
message and avoiding to type it again. Bouncing is the fastest
solution.

A proper MUA should display the resent-to header indicating that the
mail was bounced. Unfortunately, it seems few people use proper MUAs
these days.

-- 
					    Gilles.

Re: [Xenomai] Xenomai 3.0-rc3 Interruption on Arietta G25 (AT91SAM9G25 chip)

[Paul Janzen]

Gilles Chanteperdrix <gilles.chanteperdrix@xenomai.org> writes:
> On Thu, Mar 26, 2015 at 10:08:30AM -0700, Paul Janzen wrote:
>> Neither your MUA nor the mailing list software insert a
>> 
>> Reply-To: xenomai@xenomai.org
>
> You should get a resent-to.

I see Resent-To: in only 0.3% of my xenomai messages.  I assume those
are the ones that people first replied to the original sender, then
later on noticed and forwarded again to the mailing list.

> A proper MUA should display the resent-to header indicating that the
> mail was bounced. Unfortunately, it seems few people use proper MUAs
> these days.

Proper MUA's do not create a "Reply" type message with a To: field
initialized to the value of Resent-To:.

Proper MUA's create a "Reply" message with a To: field initalized with
the value of Reply-To:.

I'm simply suggesting that xenomai mailman be configured to include
this header. Then people would have to consciously edit header fields
to send to your private account only.

-- Paul

Re: [Xenomai] Xenomai 3.0-rc3 Interruption on Arietta G25 (AT91SAM9G25 chip)

[Gilles Chanteperdrix]

On Thu, Mar 26, 2015 at 11:51:12AM -0700, Paul Janzen wrote:
> Gilles Chanteperdrix <gilles.chanteperdrix@xenomai.org> writes:
> > On Thu, Mar 26, 2015 at 10:08:30AM -0700, Paul Janzen wrote:
> >> Neither your MUA nor the mailing list software insert a
> >> 
> >> Reply-To: xenomai@xenomai.org
> >
> > You should get a resent-to.
> 
> I see Resent-To: in only 0.3% of my xenomai messages.  I assume those
> are the ones that people first replied to the original sender, then
> later on noticed and forwarded again to the mailing list.
> 
> > A proper MUA should display the resent-to header indicating that the
> > mail was bounced. Unfortunately, it seems few people use proper MUAs
> > these days.
> 
> Proper MUA's do not create a "Reply" type message with a To: field
> initialized to the value of Resent-To:.

Of course.

> 
> Proper MUA's create a "Reply" message with a To: field initalized with
> the value of Reply-To:.

The message I was replying to was not a message sent on the xenomai
mailing list. It was a private message. So, the To: field was set to
the sender only. I did not notice, sent the mail. Then did not see
the mail coming to the list, looked at the mail, so it had not been
sent to the list, so bounced the message, again, because it is
easy and has the advantage that the new mail has the same id than
the first mail, so allows eliminating the duplicate.

> 
> I'm simply suggesting that xenomai mailman be configured to include
> this header. Then people would have to consciously edit header fields
> to send to your private account only.

No, the mailman documentation recommends against that, it is a
recipe for sending private mails to the list unintentionally, which
is way worse than the opposite.

-- 
					    Gilles.

Re: [Xenomai] Xenomai 3.0-rc3 Interruption on Arietta G25 (AT91SAM9G25 chip)

[Paul Janzen]

Gilles Chanteperdrix <gilles.chanteperdrix@xenomai.org> writes:
>> I'm simply suggesting that xenomai mailman be configured to include
>> this header. Then people would have to consciously edit header fields
>> to send to your private account only.
>
> No, the mailman documentation recommends against that, it is a
> recipe for sending private mails to the list unintentionally, which
> is way worse than the opposite.

Your responses to those who make the trivial mistake of clicking on
"reply" rather than "reply all" seem to indicate that you believe
otherwise.

The rationale behind the mailman developers' stance against Reply-To:
is that users should be encouraged to choose either to reply to the
original sender's personal address, or to the entire list.  This is
the opposite of the stated policy of the xenomai list.

If you personally do not want replies sent only to your personal
address, then feel free to configure your MUA to insert your own
Reply-To: header.  If you want to maximize the amount of
xenomai-related traffic captured in the digests, configure mailman to
do it.

-- Paul

Re: [Xenomai] Xenomai 3.0-rc3 Interruption on Arietta G25 (AT91SAM9G25 chip)

[Gilles Chanteperdrix]

On Thu, Mar 26, 2015 at 12:32:24PM -0700, Paul Janzen wrote:
> Gilles Chanteperdrix <gilles.chanteperdrix@xenomai.org> writes:
> >> I'm simply suggesting that xenomai mailman be configured to include
> >> this header. Then people would have to consciously edit header fields
> >> to send to your private account only.
> >
> > No, the mailman documentation recommends against that, it is a
> > recipe for sending private mails to the list unintentionally, which
> > is way worse than the opposite.
> 
> Your responses to those who make the trivial mistake of clicking on
> "reply" rather than "reply all" seem to indicate that you believe
> otherwise.

My response are a warning, just that: next private mail will not
receive an answer. But otherwise, again, I prefer this error than
the opposite.

> 
> The rationale behind the mailman developers' stance against Reply-To:
> is that users should be encouraged to choose either to reply to the
> original sender's personal address, or to the entire list.  This is
> the opposite of the stated policy of the xenomai list.

There are cases where a private answer is appropriate, asking a
question about xenomai, is not one of them.

> 
> If you personally do not want replies sent only to your personal
> address, then feel free to configure your MUA to insert your own
> Reply-To: header.  If you want to maximize the amount of
> xenomai-related traffic captured in the digests, configure mailman to
> do it.

No, I think I will handle things differently: private answers to the
xenomai mailing list will go to my "slow" imap folder only polled once a
day.

-- 
					    Gilles.