[Xenomai-help] Xenomai test progs segfault

[Xenomai-help] Xenomai test progs segfault

[Terry Fryar]

[-- Attachment #1: Type: text/plain, Size: 989 bytes --]

Using 2.6.38 omap kernel on beagleboard xm-c, with 2.5.6 Xenomai and 2.6.38
adeos patch.  Adeos patches went through with no errors.  I do have a
"tweaked" omap3 board file (gpio's and such), but this is a pretty
straightforward kernel (PREEMPT, no power mgmt, no SMP).  Using Codesourcery
2010.09-50 as tool chain on host.
 
Looks like the kernel is ok on boot:
 
[    0.969854] I-pipe: Domain Xenomai registered.
[    0.969900] Xenomai: hal/arm started.
[    0.971906] Xenomai: scheduling class idle registered.
[    0.971937] Xenomai: scheduling class rt registered.
[    1.026001] Xenomai: real-time nucleus v2.5.6 (Wormhole Wizards) loaded.
[    1.026763] Xenomai: starting native API services.
[    1.027046] Xenomai: starting RTDM services.

 
If I try to run "latency", or some of the other test programs I get a
segfault.  GDB says it's getting a signal SIGXCPU??
 
Been through the kernel .config many, many times!  Not sure where to go from
here?   Any suggestions?
 
Thanks!
 

[-- Attachment #2: Type: text/html, Size: 2419 bytes --]

Re: [Xenomai-help] Xenomai test progs segfault

[Gilles Chanteperdrix]

On 11/10/2011 12:37 AM, Terry Fryar wrote:
> Using 2.6.38 omap kernel on beagleboard xm-c, with 2.5.6 Xenomai and 2.6.38
> adeos patch.  Adeos patches went through with no errors.  I do have a
> "tweaked" omap3 board file (gpio's and such), but this is a pretty
> straightforward kernel (PREEMPT, no power mgmt, no SMP).  Using Codesourcery
> 2010.09-50 as tool chain on host.
>  
> Looks like the kernel is ok on boot:
>  
> [    0.969854] I-pipe: Domain Xenomai registered.
> [    0.969900] Xenomai: hal/arm started.
> [    0.971906] Xenomai: scheduling class idle registered.
> [    0.971937] Xenomai: scheduling class rt registered.
> [    1.026001] Xenomai: real-time nucleus v2.5.6 (Wormhole Wizards) loaded.
> [    1.026763] Xenomai: starting native API services.
> [    1.027046] Xenomai: starting RTDM services.
> 
>  
> If I try to run "latency", or some of the other test programs I get a
> segfault.  GDB says it's getting a signal SIGXCPU??
>  
> Been through the kernel .config many, many times!  Not sure where to go from
> here?   Any suggestions?

Please post your .config.

-- 
                                                                Gilles.

Re: [Xenomai-help] Xenomai test progs segfault

[Gilles Chanteperdrix]

On 11/10/2011 12:37 AM, Terry Fryar wrote:
> Using 2.6.38 omap kernel on beagleboard xm-c, with 2.5.6 Xenomai and 2.6.38
> adeos patch.  Adeos patches went through with no errors.  I do have a
> "tweaked" omap3 board file (gpio's and such), but this is a pretty
> straightforward kernel (PREEMPT, no power mgmt, no SMP).  Using Codesourcery
> 2010.09-50 as tool chain on host.
>  
> Looks like the kernel is ok on boot:
>  
> [    0.969854] I-pipe: Domain Xenomai registered.
> [    0.969900] Xenomai: hal/arm started.
> [    0.971906] Xenomai: scheduling class idle registered.
> [    0.971937] Xenomai: scheduling class rt registered.
> [    1.026001] Xenomai: real-time nucleus v2.5.6 (Wormhole Wizards) loaded.
> [    1.026763] Xenomai: starting native API services.
> [    1.027046] Xenomai: starting RTDM services.
> 
>  
> If I try to run "latency", or some of the other test programs I get a
> segfault.  GDB says it's getting a signal SIGXCPU??
>  
> Been through the kernel .config many, many times!  Not sure where to go from
> here?   Any suggestions?

Ah, sorry, xenomai 2.5.6 does not support linux 2.6.38 on arm. You need
xenomai 2.6.0. In case 2.5.6 was a typo and you were already using
2.6.0, please post your .config.

>  
> Thanks!
>  
> 
> 
> 
> 
> _______________________________________________
> Xenomai-help mailing list
> Xenomai-help@domain.hid
> https://mail.gna.org/listinfo/xenomai-help


-- 
                                                                Gilles.

Re: [Xenomai-help] Xenomai test progs segfault

[Gilles Chanteperdrix]

On 11/10/2011 03:04 PM, Terry Fryar wrote:
> Whoops...no, I was using 2.5.6...I'll switch to 2.6.0 and see what happens.
> Also, I've been wondering if my FPU flags were consistent between my
> kernel/xenomai builds...I need to double check that also I think?

In user-space the choice is either soft-float, or hard-float with VFP.
If you want hard-float with VFP, you need CONFIG_VFP in kernel
configuration, but it is probably enabled by default. For use of VFP by
real-time threads, you need CONFIG_XENO_HW_FPU, which should be enabled
by default as well.

Please no private mails.

-- 
					    Gilles.

Re: [Xenomai-help] Xenomai test progs segfault

[Terry Fryar]

2.6.0 worked fine with 2.6.38, although I did get this:

[    4.900113] WARNING: at kernel/lockdep.c:2321 omap3_pm_idle+0x78/0x84()
[    4.907076] Modules linked in:
[    4.910532] [<c00572ac>] (unwind_backtrace+0x0/0xe0) from [<c0081164>]
(warn_slowpath_common+0x4c/0x64)
[    4.920564] [<c0081164>] (warn_slowpath_common+0x4c/0x64) from
[<c0081194>] (warn_slowpath_null+0x18/0x1c)
[    4.930865] [<c0081194>] (warn_slowpath_null+0x18/0x1c) from [<c0064268>]
(omap3_pm_idle+0x78/0x84)
[    4.940508] [<c0064268>] (omap3_pm_idle+0x78/0x84) from [<c005241c>]
(cpu_idle+0x6c/0xe4)
[    4.949116] [<c005241c>] (cpu_idle+0x6c/0xe4) from [<c00088bc>]
(start_kernel+0x2b0/0x30c)
[    4.957941] [<c00088bc>] (start_kernel+0x2b0/0x30c) from [<80008034>]
(0x80008034) 

Not sure what's up here, but the xenomai kernel stuff loaded and the xenomai
test programs work just fine now.  I'm partially wondering if there were
mis-matched floating point compiler flags involved as well (2.6.0 doesn't
force any flags now).

Thanks Gilles...your help is MUCH appreciated!


-----Original Message-----
From: Gilles Chanteperdrix [mailto:gilles.chanteperdrix@xenomai.org] 
Sent: Thursday, November 10, 2011 2:53 AM
To: Terry Fryar
Cc: xenomai@xenomai.org
Subject: Re: [Xenomai-help] Xenomai test progs segfault

On 11/10/2011 12:37 AM, Terry Fryar wrote:
> Using 2.6.38 omap kernel on beagleboard xm-c, with 2.5.6 Xenomai and 
> 2.6.38 adeos patch.  Adeos patches went through with no errors.  I do 
> have a "tweaked" omap3 board file (gpio's and such), but this is a 
> pretty straightforward kernel (PREEMPT, no power mgmt, no SMP).  Using 
> Codesourcery 2010.09-50 as tool chain on host.
>  
> Looks like the kernel is ok on boot:
>  
> [    0.969854] I-pipe: Domain Xenomai registered.
> [    0.969900] Xenomai: hal/arm started.
> [    0.971906] Xenomai: scheduling class idle registered.
> [    0.971937] Xenomai: scheduling class rt registered.
> [    1.026001] Xenomai: real-time nucleus v2.5.6 (Wormhole Wizards)
loaded.
> [    1.026763] Xenomai: starting native API services.
> [    1.027046] Xenomai: starting RTDM services.
> 
>  
> If I try to run "latency", or some of the other test programs I get a 
> segfault.  GDB says it's getting a signal SIGXCPU??
>  
> Been through the kernel .config many, many times!  Not sure where to go
from
> here?   Any suggestions?

Ah, sorry, xenomai 2.5.6 does not support linux 2.6.38 on arm. You need
xenomai 2.6.0. In case 2.5.6 was a typo and you were already using 2.6.0,
please post your .config.

>  
> Thanks!
>  
> 
> 
> 
> 
> _______________________________________________
> Xenomai-help mailing list
> Xenomai-help@domain.hid
> https://mail.gna.org/listinfo/xenomai-help


-- 
                                                                Gilles.

Re: [Xenomai-help] Xenomai test progs segfault

[Gilles Chanteperdrix]

On 11/10/2011 09:09 PM, Terry Fryar wrote:
> 2.6.0 worked fine with 2.6.38, although I did get this:
> 
> [    4.900113] WARNING: at kernel/lockdep.c:2321 omap3_pm_idle+0x78/0x84()
> [    4.907076] Modules linked in:
> [    4.910532] [<c00572ac>] (unwind_backtrace+0x0/0xe0) from [<c0081164>]
> (warn_slowpath_common+0x4c/0x64)
> [    4.920564] [<c0081164>] (warn_slowpath_common+0x4c/0x64) from
> [<c0081194>] (warn_slowpath_null+0x18/0x1c)
> [    4.930865] [<c0081194>] (warn_slowpath_null+0x18/0x1c) from [<c0064268>]
> (omap3_pm_idle+0x78/0x84)
> [    4.940508] [<c0064268>] (omap3_pm_idle+0x78/0x84) from [<c005241c>]
> (cpu_idle+0x6c/0xe4)
> [    4.949116] [<c005241c>] (cpu_idle+0x6c/0xe4) from [<c00088bc>]
> (start_kernel+0x2b0/0x30c)
> [    4.957941] [<c00088bc>] (start_kernel+0x2b0/0x30c) from [<80008034>]
> (0x80008034) 
> 
> Not sure what's up here, but the xenomai kernel stuff loaded and the xenomai
> test programs work just fine now.  I'm partially wondering if there were
> mis-matched floating point compiler flags involved as well (2.6.0 doesn't
> force any flags now).
> 
> Thanks Gilles...your help is MUCH appreciated!

You should disable whatever option causes omap3_pm_idle to be called, as
this will give bad latencies.

-- 
                                                                Gilles.

Re: [Xenomai-help] Xenomai test progs segfault

[Gilles Chanteperdrix]

On 11/10/2011 09:12 PM, Gilles Chanteperdrix wrote:
> On 11/10/2011 09:09 PM, Terry Fryar wrote:
>> 2.6.0 worked fine with 2.6.38, although I did get this:
>>
>> [    4.900113] WARNING: at kernel/lockdep.c:2321 omap3_pm_idle+0x78/0x84()
>> [    4.907076] Modules linked in:
>> [    4.910532] [<c00572ac>] (unwind_backtrace+0x0/0xe0) from [<c0081164>]
>> (warn_slowpath_common+0x4c/0x64)
>> [    4.920564] [<c0081164>] (warn_slowpath_common+0x4c/0x64) from
>> [<c0081194>] (warn_slowpath_null+0x18/0x1c)
>> [    4.930865] [<c0081194>] (warn_slowpath_null+0x18/0x1c) from [<c0064268>]
>> (omap3_pm_idle+0x78/0x84)
>> [    4.940508] [<c0064268>] (omap3_pm_idle+0x78/0x84) from [<c005241c>]
>> (cpu_idle+0x6c/0xe4)
>> [    4.949116] [<c005241c>] (cpu_idle+0x6c/0xe4) from [<c00088bc>]
>> (start_kernel+0x2b0/0x30c)
>> [    4.957941] [<c00088bc>] (start_kernel+0x2b0/0x30c) from [<80008034>]
>> (0x80008034) 
>>
>> Not sure what's up here, but the xenomai kernel stuff loaded and the xenomai
>> test programs work just fine now.  I'm partially wondering if there were
>> mis-matched floating point compiler flags involved as well (2.6.0 doesn't
>> force any flags now).
>>
>> Thanks Gilles...your help is MUCH appreciated!
> 
> You should disable whatever option causes omap3_pm_idle to be called, as
> this will give bad latencies.
> 

You can not avoid omap3_pm_idle with a configuration option.
Fortunately, it does not enter deep sleep states, if you do not mount
debugfs and do not set sleep_while_idle to 1.

-- 
                                                                Gilles.