* [Xenomai] System hang when doing several rtdm_task_init() calling
@ 2016-12-19 14:04 Sebastien Schneider
2016-12-19 17:20 ` Philippe Gerum
0 siblings, 1 reply; 6+ messages in thread
From: Sebastien Schneider @ 2016-12-19 14:04 UTC (permalink / raw)
To: xenomai
Hello,
I am facing issues with the rtdm_task_init() function when trying to
developp a xenomai real-time base driver.
I firstly call the rtdm_task_init() method for a non periodic task with the
lowest priority. This lowest priority task infinitly executes code which
first consists on a rtdm_task_sleep() for a predefined amount of time,
before other various operations.
Then, I secondly call the rtdm_task_init() method for an another non
periodic task with the highest priority. And what happens after compiling
the code and trying to insert the obtained module, is that the system seems
to hang.
Has anyone some idea that can help me ?
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [Xenomai] System hang when doing several rtdm_task_init() calling
2016-12-19 14:04 [Xenomai] System hang when doing several rtdm_task_init() calling Sebastien Schneider
@ 2016-12-19 17:20 ` Philippe Gerum
2016-12-20 9:54 ` Sebastien Schneider
0 siblings, 1 reply; 6+ messages in thread
From: Philippe Gerum @ 2016-12-19 17:20 UTC (permalink / raw)
To: Sebastien Schneider, xenomai
On 12/19/2016 03:04 PM, Sebastien Schneider wrote:
> Hello,
>
> I am facing issues with the rtdm_task_init() function when trying to
> developp a xenomai real-time base driver.
> I firstly call the rtdm_task_init() method for a non periodic task with the
> lowest priority. This lowest priority task infinitly executes code which
> first consists on a rtdm_task_sleep() for a predefined amount of time,
> before other various operations.
> Then, I secondly call the rtdm_task_init() method for an another non
> periodic task with the highest priority. And what happens after compiling
> the code and trying to insert the obtained module, is that the system seems
> to hang.
>
> Has anyone some idea that can help me ?
Runaway thread #2, maybe never blocking unlike you would expect it to?
Enabling the watchdog is recommended when developing Xenomai apps
(CONFIG_XENO_OPT_WATCHDOG); it would trap when a rt thread runs away,
suspending it with a warning message sent to the kernel log.
Sharing the driver code in question would help. Such issue basically
depends on what your code is exactly doing.
--
Philippe.
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [Xenomai] System hang when doing several rtdm_task_init() calling
2016-12-19 17:20 ` Philippe Gerum
@ 2016-12-20 9:54 ` Sebastien Schneider
2016-12-20 11:18 ` Philippe Gerum
0 siblings, 1 reply; 6+ messages in thread
From: Sebastien Schneider @ 2016-12-20 9:54 UTC (permalink / raw)
To: Philippe Gerum; +Cc: xenomai
I am using xenomai-3.0.2 on a Raspberry PI 2 with patched kernel 4.1.21,
and there is no CONFIG_XENO_OPT_WATCHDOG option available.
After system has hang, I can even "dmesg" the kernel using the serial line,
that leads to such messages:
[ 258.813973] INFO: rcu_sched detected stalls on CPUs/tasks:
[ 258.814019] 2: (0 ticks this GP) idle=aa1/140000000000000/0
softirq=2028/2028 fqs=1
[ 258.814035] (detected by 3, t=2102 jiffies, g=2860, c=2859, q=460)
[ 258.814059] Task dump for CPU 2:
[ 258.814075] swapper/2 R running 0 0 1 0x00000002
[ 258.814140] [<80025028>] (xnarch_switch_to) from [<800f4688>]
(___xnsched_run+0x260/0x558)
[ 258.814169] [<800f4688>] (___xnsched_run) from [<800f4f0c>]
(__xnsched_run_handler+0x78/0xf4)
[ 258.814201] [<800f4f0c>] (__xnsched_run_handler) from [<800b2690>]
(dispatch_irq_head+0xb4/0x170)
[ 258.814233] [<800b2690>] (dispatch_irq_head) from [<800b2c5c>]
(__ipipe_dispatch_irq+0x108/0x20c)
[ 258.814263] [<800b2c5c>] (__ipipe_dispatch_irq) from [<80009204>]
(__ipipe_grab_ipi+0x40/0x90)
[ 258.814292] [<80009204>] (__ipipe_grab_ipi) from [<8059cff4>]
(__irq_svc+0x34/0x118)
[ 258.814308] Exception stack(0xb98f9f48 to 0xb98f9f90)
[ 258.814331] 9f40: 00000000 00000002 39751000 00000000
b98f8000 80858e1c
[ 258.814356] 9f60: 80858dbc 805a22e8 0000406a 808c1bbc 00000000 b98f9f9c
8005d294 b98f9f90
[ 258.814374] 9f80: 80092f0c 8005d2c0 60000013 ffffffff
[ 258.814405] [<8059cff4>] (__irq_svc) from [<8005d2c0>]
(set_cpu_sd_state_idle+0x40/0x80)
[ 258.814437] [<8005d2c0>] (set_cpu_sd_state_idle) from [<80092f0c>]
(tick_nohz_idle_enter+0x24/0x90)
[ 258.814470] [<80092f0c>] (tick_nohz_idle_enter) from [<800638a8>]
(cpu_startup_entry+0x68/0x2d0)
[ 258.814501] [<800638a8>] (cpu_startup_entry) from [<80016184>]
(secondary_start_kernel+0x12c/0x134)
[ 258.814531] [<80016184>] (secondary_start_kernel) from [<0000974c>]
(0x974c)
[ 258.814549] rcu_sched kthread starved for 2101 jiffies!
Is this log helpful ?
2016-12-19 18:20 GMT+01:00 Philippe Gerum <rpm@xenomai.org>:
> On 12/19/2016 03:04 PM, Sebastien Schneider wrote:
> > Hello,
> >
> > I am facing issues with the rtdm_task_init() function when trying to
> > developp a xenomai real-time base driver.
> > I firstly call the rtdm_task_init() method for a non periodic task with
> the
> > lowest priority. This lowest priority task infinitly executes code which
> > first consists on a rtdm_task_sleep() for a predefined amount of time,
> > before other various operations.
> > Then, I secondly call the rtdm_task_init() method for an another non
> > periodic task with the highest priority. And what happens after compiling
> > the code and trying to insert the obtained module, is that the system
> seems
> > to hang.
> >
> > Has anyone some idea that can help me ?
>
> Runaway thread #2, maybe never blocking unlike you would expect it to?
> Enabling the watchdog is recommended when developing Xenomai apps
> (CONFIG_XENO_OPT_WATCHDOG); it would trap when a rt thread runs away,
> suspending it with a warning message sent to the kernel log.
>
> Sharing the driver code in question would help. Such issue basically
> depends on what your code is exactly doing.
>
> --
> Philippe.
>
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [Xenomai] System hang when doing several rtdm_task_init() calling
2016-12-20 9:54 ` Sebastien Schneider
@ 2016-12-20 11:18 ` Philippe Gerum
2016-12-21 14:24 ` Sebastien Schneider
0 siblings, 1 reply; 6+ messages in thread
From: Philippe Gerum @ 2016-12-20 11:18 UTC (permalink / raw)
To: Sebastien Schneider; +Cc: xenomai
On 12/20/2016 10:54 AM, Sebastien Schneider wrote:
> I am using xenomai-3.0.2 on a Raspberry PI 2 with patched kernel 4.1.21,
> and there is no CONFIG_XENO_OPT_WATCHDOG option available.
There is. It may require CONFIG_XENO_OPT_DEBUG to be enabled in your
kernel config to see this knob, depending on the config interface you
are using.
> After system has hang, I can even "dmesg" the kernel using the serial
> line, that leads to such messages:
>
> [ 258.813973] INFO: rcu_sched detected stalls on CPUs/tasks:
> [ 258.814019] 2: (0 ticks this GP) idle=aa1/140000000000000/0
> softirq=2028/2028 fqs=1
> [ 258.814035] (detected by 3, t=2102 jiffies, g=2860, c=2859, q=460)
> [ 258.814059] Task dump for CPU 2:
> [ 258.814075] swapper/2 R running 0 0 1 0x00000002
> [ 258.814140] [<80025028>] (xnarch_switch_to) from [<800f4688>]
> (___xnsched_run+0x260/0x558)
> [ 258.814169] [<800f4688>] (___xnsched_run) from [<800f4f0c>]
> (__xnsched_run_handler+0x78/0xf4)
> [ 258.814201] [<800f4f0c>] (__xnsched_run_handler) from [<800b2690>]
> (dispatch_irq_head+0xb4/0x170)
> [ 258.814233] [<800b2690>] (dispatch_irq_head) from [<800b2c5c>]
> (__ipipe_dispatch_irq+0x108/0x20c)
> [ 258.814263] [<800b2c5c>] (__ipipe_dispatch_irq) from [<80009204>]
> (__ipipe_grab_ipi+0x40/0x90)
> [ 258.814292] [<80009204>] (__ipipe_grab_ipi) from [<8059cff4>]
> (__irq_svc+0x34/0x118)
The last event on the locked up CPU is a switch to a real-time task, so
the presence of a runaway code in this context seems likely. You may
want to check your RTDM driver task code, and enable the watchdog as
suggested.
--
Philippe.
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [Xenomai] System hang when doing several rtdm_task_init() calling
2016-12-20 11:18 ` Philippe Gerum
@ 2016-12-21 14:24 ` Sebastien Schneider
2016-12-21 16:11 ` Philippe Gerum
0 siblings, 1 reply; 6+ messages in thread
From: Sebastien Schneider @ 2016-12-21 14:24 UTC (permalink / raw)
To: Philippe Gerum; +Cc: xenomai
I've correctly set CONFIG_XENO_OPT_WATCHDOG thanks to CONFIG_XENO_OPT_DEBUG.
Here is the basic version of driver I try to load, and that leads to the
same effect :
*#define LOWEST_PRIORITY 0*
*#define HIGHEST_PRIORITY 99*
*rtdm_task_t task1;*
*rtdm_task_t task2;*
*void handler_task1(void *arg)*
*{*
* while(1)*
* {*
* }*
* return ;*
*}*
*void handler_task2(void *arg)*
*{*
* while(1)*
* {*
* }*
* return ;*
*}*
*static int __init vlc_driver_init(void)*
*{*
* printk(KERN_INFO "VLC: Inserting module...\n");*
* rtdm_task_init(&task1, "task1", handler_task1, NULL,
LOWEST_PRIORITY, 0);*
* rtdm_task_init(&task2, "task2", handler_task2, NULL,
HIGHEST_PRIORITY, 0);*
* return 0;*
*}*
*static void __exit vlc_driver_exit(void)*
*{*
* printk(KERN_INFO "VLC: Removing module...\n");*
* return;*
*}*
*module_init(vlc_driver_init);*
*module_exit(vlc_driver_exit);*
*MODULE_LICENSE("GPL");*
So, it must be something very basic I did not have understood !!!
2016-12-20 12:18 GMT+01:00 Philippe Gerum <rpm@xenomai.org>:
> On 12/20/2016 10:54 AM, Sebastien Schneider wrote:
> > I am using xenomai-3.0.2 on a Raspberry PI 2 with patched kernel 4.1.21,
> > and there is no CONFIG_XENO_OPT_WATCHDOG option available.
>
> There is. It may require CONFIG_XENO_OPT_DEBUG to be enabled in your
> kernel config to see this knob, depending on the config interface you
> are using.
>
> > After system has hang, I can even "dmesg" the kernel using the serial
> > line, that leads to such messages:
> >
> > [ 258.813973] INFO: rcu_sched detected stalls on CPUs/tasks:
> > [ 258.814019] 2: (0 ticks this GP) idle=aa1/140000000000000/0
> > softirq=2028/2028 fqs=1
> > [ 258.814035] (detected by 3, t=2102 jiffies, g=2860, c=2859, q=460)
> > [ 258.814059] Task dump for CPU 2:
> > [ 258.814075] swapper/2 R running 0 0 1 0x00000002
> > [ 258.814140] [<80025028>] (xnarch_switch_to) from [<800f4688>]
> > (___xnsched_run+0x260/0x558)
> > [ 258.814169] [<800f4688>] (___xnsched_run) from [<800f4f0c>]
> > (__xnsched_run_handler+0x78/0xf4)
> > [ 258.814201] [<800f4f0c>] (__xnsched_run_handler) from [<800b2690>]
> > (dispatch_irq_head+0xb4/0x170)
> > [ 258.814233] [<800b2690>] (dispatch_irq_head) from [<800b2c5c>]
> > (__ipipe_dispatch_irq+0x108/0x20c)
> > [ 258.814263] [<800b2c5c>] (__ipipe_dispatch_irq) from [<80009204>]
> > (__ipipe_grab_ipi+0x40/0x90)
> > [ 258.814292] [<80009204>] (__ipipe_grab_ipi) from [<8059cff4>]
> > (__irq_svc+0x34/0x118)
>
> The last event on the locked up CPU is a switch to a real-time task, so
> the presence of a runaway code in this context seems likely. You may
> want to check your RTDM driver task code, and enable the watchdog as
> suggested.
>
> --
> Philippe.
>
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [Xenomai] System hang when doing several rtdm_task_init() calling
2016-12-21 14:24 ` Sebastien Schneider
@ 2016-12-21 16:11 ` Philippe Gerum
0 siblings, 0 replies; 6+ messages in thread
From: Philippe Gerum @ 2016-12-21 16:11 UTC (permalink / raw)
To: Sebastien Schneider; +Cc: xenomai
On 12/21/2016 03:24 PM, Sebastien Schneider wrote:
> I've correctly set CONFIG_XENO_OPT_WATCHDOG thanks to CONFIG_XENO_OPT_DEBUG.
> Here is the basic version of driver I try to load, and that leads to the
> same effect :
>
> /#define LOWEST_PRIORITY 0/
> /#define HIGHEST_PRIORITY 99/
> /rtdm_task_t task1;/
> /rtdm_task_t task2;/
> /
> /
> /void handler_task1(void *arg)/
> /{/
> / while(1)/
> / {/
> / }/
> / return ;/
> /}/
> /
> /
> /void handler_task2(void *arg)/
> /{/
> / while(1)/
> / {/
> / }/
> / return ;/
> /}/
> /
> /
> /static int __init vlc_driver_init(void)/
> /{/
> / printk(KERN_INFO "VLC: Inserting module...\n");/
> / rtdm_task_init(&task1, "task1", handler_task1, NULL,
> LOWEST_PRIORITY, 0);/
> / rtdm_task_init(&task2, "task2", handler_task2, NULL,
> HIGHEST_PRIORITY, 0);/
> / return 0;/
> /}/
> /
> /
> /static void __exit vlc_driver_exit(void)/
> /{/
> / printk(KERN_INFO "VLC: Removing module...\n");/
> / return;/
> /}/
> /
> /
> /module_init(vlc_driver_init);/
> /module_exit(vlc_driver_exit);/
> /MODULE_LICENSE("GPL");/
> /
> /
> So, it must be something very basic I did not have understood !!!
>
If the code above is actually the one you are trying to run over a dual
kernel system, infinitely running real-time tasks which never block can
only lead to a complete lock up.
--
Philippe.
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2016-12-21 16:11 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2016-12-19 14:04 [Xenomai] System hang when doing several rtdm_task_init() calling Sebastien Schneider
2016-12-19 17:20 ` Philippe Gerum
2016-12-20 9:54 ` Sebastien Schneider
2016-12-20 11:18 ` Philippe Gerum
2016-12-21 14:24 ` Sebastien Schneider
2016-12-21 16:11 ` Philippe Gerum
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.