Re: [Xenomai-help] mem usage and other application

Re: [Xenomai-help] mem usage and other application

[Gilles Chanteperdrix]

Maksym Veremeyenko wrote:
> Gilles Chanteperdrix ~~~~~:
> 
> 
>>Well, it really looks like an FPU bug. Could you try to:
>>- upgrade your trunk to latest revision, then test
>>- then apply the attached patch to your up-to-date trunk, then test
>>  again ?
>>
> 
> 
> I used configuration 2.6.17.14, xenomai-r2299
> 
> Before patching xenomai result was the same and more faulty after 
> decreasing RAM size down to 256M.
> 
> After applying patch situation become worth, i got "Bad pte" messages, 
> sometimes "Soft lockup detected", and kernel dump:

Why changing the RAM size ? Could you double-check with your RAM size
normally set ? Do you observe the "Bad pte" message with latency or
switchbench running, or with your RT application ?

-- 
                                                 Gilles Chanteperdrix

Re: [Xenomai-help] mem usage and other application

[Maksym Veremeyenko]

Gilles Chanteperdrix пишет:

> Why changing the RAM size ? 
i added RAM before to check if it could help running other non-RT 
application.

> Could you double-check with your RAM size
> normally set ? Do you observe the "Bad pte" message with latency or
> switchbench running, or with your RT application ?

I started 'latency' and then start any perl script. It behavior depends 
on configurations:
     Problematic configuration are:
       1. Kernel 2.6.17.14, Xenomai 2.3.0... r2299, Adeos 1.5-02
       2. Kernel 2.6.19.2, Xenomai r2299, Adeos 1.7-01

     No problems with:
       3. Kernel 2.6.19.2, Xenomai 2.3.0, Adeos 1.7-01
       4. kernel 2.6.17.14, xenomai 2.2.4, Adeos 1.5-00

After applying your patch to configuration #1 or #2 situation become worth:
1) i still get errors,
2) i also get a "Bad pte" with 'process' points to non-RT application 
caused fault
3) sometime perl script "core dumped":
Core was generated by `/usr/bin/perl ./m1.meta.pl U0002323'.
Program terminated with signal 11, Segmentation fault.
#0  0x42a56004 in ?? ()
(gdb) bt
#0  0x42a56004 in ?? ()
#1  0x42a547a7 in ?? ()
#2  0x0804bc0a in ?? ()
#3  0x00000000 in ?? ()

________________________________________
Maksym Veremeyenko

Re: [Xenomai-help] mem usage and other application

[Gilles Chanteperdrix]

Maksym Veremeyenko wrote:
 > Gilles Chanteperdrix ~~~~~:
 > 
 > > Why changing the RAM size ? 
 > i added RAM before to check if it could help running other non-RT 
 > application.
 > 
 > > Could you double-check with your RAM size
 > > normally set ? Do you observe the "Bad pte" message with latency or
 > > switchbench running, or with your RT application ?
 > 
 > I started 'latency' and then start any perl script. It behavior depends 
 > on configurations:
 >      Problematic configuration are:
 >        1. Kernel 2.6.17.14, Xenomai 2.3.0... r2299, Adeos 1.5-02
Here, I assume you mean Xenomai branch v2.3.x revision 2299

 >        2. Kernel 2.6.19.2, Xenomai r2299, Adeos 1.7-01
Here, I assume you mean Xenomai trunk revision 2299

The problem is that if there is no difference with regard to FPU context
management between release 2.3.0 and latest revision of branch
v2.3.x. So, the difference must be somewhere else.

Could you send me your .config ?

Also, is it possible for you to try bisecting in v2.3.x branch ? v2.3.0
release is revision 2004, and you know that revision 2299 does not work
for you, so you would be able to tell us exactly which revision is the
cause of your problem.

-- 


					    Gilles Chanteperdrix.

Re: [Xenomai-help] mem usage and other application

[Maksym Veremeyenko]

[-- Attachment #1: Type: text/plain, Size: 1204 bytes --]

Gilles Chanteperdrix пишет:

>  >        1. Kernel 2.6.17.14, Xenomai 2.3.0... r2299, Adeos 1.5-02
> Here, I assume you mean Xenomai branch v2.3.x revision 2299
release 2.3.0 and branch v2.3.x revision 2299 - both

>  >        2. Kernel 2.6.19.2, Xenomai r2299, Adeos 1.7-01
> Here, I assume you mean Xenomai trunk revision 2299
yes

> The problem is that if there is no difference with regard to FPU context
> management between release 2.3.0 and latest revision of branch
> v2.3.x. So, the difference must be somewhere else.
What is recommended configuration and version of kernel and Xenomai?
In my case
2.6.20.3 upgrade failed due to Xenomai compilation,
2.6.19 fault gdb to work properly,
2.6.18 fail due to absent adeos-ipipe version.
I have to use 2.6.17.14 .....

> Could you send me your .config ?
attached


> Also, is it possible for you to try bisecting in v2.3.x branch ? v2.3.0
> release is revision 2004, and you know that revision 2299 does not work
> for you, so you would be able to tell us exactly which revision is the
> cause of your problem.
What trunk revisions should be tried?


-- 
________________________________________
Maksym Veremeyenko


[-- Attachment #2: config-2.6.17.14.xenomai-r2299.bz2 --]
[-- Type: application/x-bzip, Size: 6399 bytes --]

Re: [Xenomai-help] mem usage and other application

[Gilles Chanteperdrix]

Maksym Veremeyenko wrote:
 > Gilles Chanteperdrix ~~~~~:
 > 
 > >  >        1. Kernel 2.6.17.14, Xenomai 2.3.0... r2299, Adeos 1.5-02
 > > Here, I assume you mean Xenomai branch v2.3.x revision 2299
 > release 2.3.0 and branch v2.3.x revision 2299 - both
 > 
 > >  >        2. Kernel 2.6.19.2, Xenomai r2299, Adeos 1.7-01
 > > Here, I assume you mean Xenomai trunk revision 2299
 > yes
 > 
 > > The problem is that if there is no difference with regard to FPU context
 > > management between release 2.3.0 and latest revision of branch
 > > v2.3.x. So, the difference must be somewhere else.
 > What is recommended configuration and version of kernel and Xenomai?
 > In my case
 > 2.6.20.3 upgrade failed due to Xenomai compilation,
 > 2.6.19 fault gdb to work properly,
 > 2.6.18 fail due to absent adeos-ipipe version.
 > I have to use 2.6.17.14 .....

As far as I understood, Adeos version does not matter for your bug,
what matters is xenomai revision. So, during your tests, you should keep
the same Adeos version and only change Xenomai version. 

However, if you have difficulties with some versions of Adeos, you
should report the errors here. We are interested in getting them
solved.

 > 
 > > Could you send me your .config ?
 > attached

I see that you you have selected the K7 processor. Be sure to use an
Adeos version greater than 1.5-02, or select a processor that does not
support 3DNow (I think K6 should do).

 > 
 > 
 > > Also, is it possible for you to try bisecting in v2.3.x branch ? v2.3.0
 > > release is revision 2004, and you know that revision 2299 does not work
 > > for you, so you would be able to tell us exactly which revision is the
 > > cause of your problem.
 > What trunk revisions should be tried?

If revision 2299 of branch v2.3.x fails, you should stay in branch
v2.3.x. Try revision 2150, if it fails, try revision 2075, if revision
2150 works, try revision 2225, etc... Do not forget to rebuild kernel
and reboot between tests. And tell us when you have found that revision
n works and revision n+1 does not work.

-- 


					    Gilles Chanteperdrix.

Re: [Xenomai-help] mem usage and other application

[Maksym Veremeyenko]

Gilles Chanteperdrix пишет:

>  > > Could you send me your .config ?
>  > attached
> 
> I see that you you have selected the K7 processor. Be sure to use an
> Adeos version greater than 1.5-02, or select a processor that does not
> support 3DNow (I think K6 should do).

Yes! Master!
Shame on me - that is a problem.
I changed to P3 and problem gone.

-- 
________________________________________
Maksym Veremeyenko

[Xenomai-help] mem usage and other application

[Maksym Veremeyenko]

Hello!

I have a question about coexisting another applications on host that is 
under xenomai.

First my problem: running some applications cause not swap usage, memory 
statistics is:

Mem:    385000k total,    91908k used,   293092k free,    21180k buffers
Swap:  2000368k total,        0k used,  2000368k free,    42684k cached

and swap is constantly 0k used. Is it normal?


Second problem:  other applications sometimes (50% attemps) fail to 
start (especially perl scripts and Eclipse) then at least one Xenomai 
application running.

I use linux 2.6.17.14, Adeos 1.5-02, Xenomai trunk 2170


________________________________________
Maksym Veremeyenko

Re: [Xenomai-help] mem usage and other application

[Gilles Chanteperdrix]

Maksym Veremeyenko wrote:
 > Hello!
 > 
 > I have a question about coexisting another applications on host that is 
 > under xenomai.
 > 
 > First my problem: running some applications cause not swap usage, memory 
 > statistics is:
 > 
 > Mem:    385000k total,    91908k used,   293092k free,    21180k buffers
 > Swap:  2000368k total,        0k used,  2000368k free,    42684k cached
 > 
 > and swap is constantly 0k used. Is it normal?

If the system does not need swap, it does not use swap, I think it is
normal.


 > 
 > 
 > Second problem:  other applications sometimes (50% attemps) fail to 
 > start (especially perl scripts and Eclipse) then at least one Xenomai 
 > application running.
 > 
 > I use linux 2.6.17.14, Adeos 1.5-02, Xenomai trunk 2170

Could you try running "switchtest" while your Xenomai application is
running ?

-- 


					    Gilles Chanteperdrix.

Re: [Xenomai-help] mem usage and other application

[Maksym Veremeyenko]

Gilles Chanteperdrix пишет:

>  > 
>  > Second problem:  other applications sometimes (50% attemps) fail to 
>  > start (especially perl scripts and Eclipse) then at least one Xenomai 
>  > application running.
>  > 
>  > I use linux 2.6.17.14, Adeos 1.5-02, Xenomai trunk 2170
> 
> Could you try running "switchtest" while your Xenomai application is
> running ?
> 

during running program i got:
[root@domain.hid switchtest]# ./switchtest
== Testing FPU check routines...
r0: 1 != 2
r1: 1 != 2
r2: 1 != 2
r3: 1 != 2
r4: 1 != 2
r5: 1 != 2
r6: 1 != 2
r7: 1 != 2
== FPU check routines: OK.
ioctl(RTTST_RTIOC_SWTEST_CREATE_KTASK): Cannot allocate memory
== Threads: sleeper_ufps-0RTT|  326087:53:20
RTH|ctx switches|-------total
RTD|           0|           0

after stopping program:
[root@domain.hid switchtest]# ./switchtest
== Testing FPU check routines...
r0: 1 != 2
r1: 1 != 2
r2: 1 != 2
r3: 1 != 2
r4: 1 != 2
r5: 1 != 2
r6: 1 != 2
r7: 1 != 2
== FPU check routines: OK.
== Threads: sleeper_ufps-0 rtk-1 rtk-2 rtk_fp-3 rtk_fp-4 rtk_fp_ufpp-5 
rtk_fp_ufpp-6 rtup-7 rtup-8 rtup_ufpp-9 rtup_ufpp-10 rtus-11 rtus-12 
rtus_ufps-13 rtus_ufps-14 rtuo-15 rtuo-16 rtuo_ufpp-17 rtuo_ufpp-18 
rtuo_ufps-19 rtuo_ufps-20 rtuo_ufpp_ufps-21 rtuo_ufpp_ufps-22
RTT|  00:00:01
RTH|ctx switches|-------total
RTD|       11500|       11500
RTD|       11523|       23023
RTD|       11523|       34546
RTD|       11523|       46069
RTD|       11500|       57569
RTD|        9108|       66677

switchtest's results are the same for two configurations:

1) kernel 2.6.17.14, xenomai 2.2.4, Adeos 1.5-00
2) kernel 2.6.17.14, xenomai 2.3.0-r2170, Adeos 1.5-02, debug enabled

except that in first configuration there are no fails (i did not found) 
of perl scrips or Eclipse.


-- 
________________________________________
Maksym Veremeyenko

Re: [Xenomai-help] mem usage and other application

[Gilles Chanteperdrix]

Maksym Veremeyenko wrote:
 > Gilles Chanteperdrix ~~~~~:
 > 
 > >  > 
 > >  > Second problem:  other applications sometimes (50% attemps) fail to 
 > >  > start (especially perl scripts and Eclipse) then at least one Xenomai 
 > >  > application running.
 > >  > 
 > >  > I use linux 2.6.17.14, Adeos 1.5-02, Xenomai trunk 2170
 > > 
 > > Could you try running "switchtest" while your Xenomai application is
 > > running ?
 > > 
 > 
 > during running program i got:
 > [root@domain.hid switchtest]# ./switchtest
 > == Testing FPU check routines...
 > r0: 1 != 2
 > r1: 1 != 2
 > r2: 1 != 2
 > r3: 1 != 2
 > r4: 1 != 2
 > r5: 1 != 2
 > r6: 1 != 2
 > r7: 1 != 2
 > == FPU check routines: OK.
 > ioctl(RTTST_RTIOC_SWTEST_CREATE_KTASK): Cannot allocate memory
 > == Threads: sleeper_ufps-0RTT|  326087:53:20
 > RTH|ctx switches|-------total
 > RTD|           0|           0

You should probably increase the system heap size, with the option
CONFIG_XENO_OPT_SYS_HEAPSZ

-- 


					    Gilles Chanteperdrix.

Re: [Xenomai-help] mem usage and other application

[Maksym Veremeyenko]

Gilles Chanteperdrix пишет:

> You should probably increase the system heap size, with the option
> CONFIG_XENO_OPT_SYS_HEAPSZ
There are problems with RT application, problem with non-RT application 
running on the same host, in the same time RT-application running...


-- 
________________________________________
Maksym Veremeyenko

Re: [Xenomai-help] mem usage and other application

[Gilles Chanteperdrix]

Maksym Veremeyenko wrote:
 > Gilles Chanteperdrix ~~~~~:
 > 
 > > You should probably increase the system heap size, with the option
 > > CONFIG_XENO_OPT_SYS_HEAPSZ
 > There are problems with RT application, problem with non-RT application 
 > running on the same host, in the same time RT-application running...

What I would like to know is if switchtest fails in the same condition
as your non real-time applications, just to see if the situation you
observe is an FPU bug. But for switchtest to start correctly, you need
to increase CONFIG_XENO_OPT_SYS_HEAPSZ.

-- 


					    Gilles Chanteperdrix.

Re: [Xenomai-help] mem usage and other application

[Maksym Veremeyenko]

Gilles Chanteperdrix пишет:

> What I would like to know is if switchtest fails in the same condition
> as your non real-time applications, just to see if the situation you
> observe is an FPU bug. But for switchtest to start correctly, you need
> to increase CONFIG_XENO_OPT_SYS_HEAPSZ.
> 

I significantly increased CONFIG_XENO_OPT_SYS_HEAPSZ and switchtest 
seems to me runs fine:

[root@domain.hid bin]# ./switchtest
== Testing FPU check routines...
r0: 1 != 2
r1: 1 != 2
r2: 1 != 2
r3: 1 != 2
r4: 1 != 2
r5: 1 != 2
r6: 1 != 2
r7: 1 != 2
== FPU check routines: OK.
== Threads: sleeper_ufps-0 rtk-1 rtk-2 rtk_fp-3 rtk_fp-4 rtk_fp_ufpp-5 
rtk_fp_ufpp-6 rtup-7 rtup-8 rtup_ufpp-9 rtup_ufpp-10 rtus-11 rtus-12 
rtus_ufps-13 rtus_ufps-14 rtuo-15 rtuo-16 rtuo_ufpp-17 rtuo_ufpp-18 
rtuo_ufps-19 rtuo_ufps-20 rtuo_ufpp_ufps-21 rtuo_ufpp_ufps-22
RTT|  00:00:01
RTH|ctx switches|-------total
RTD|        7544|        7544
RTD|        7590|       15134
RTD|        7567|       22701
RTD|        7590|       30291
RTD|        7544|       37835
RTD|        7544|       45379
RTD|        7544|       52923
RTD|        7567|       60490
RTD|        7567|       68057
RTD|        7498|       75555
RTD|        7544|       83099
RTD|        7521|       90620
RTD|        7567|       98187
RTD|        7544|      105731
RTD|        7613|      113344
RTD|        7567|      120911
RTD|        7544|      128455
RTD|        7521|      135976
RTD|        7567|      143543
RTD|        7590|      151133
RTD|        7475|      158608
RTT|  00:00:22
RTH|ctx switches|-------total
RTD|        7613|      166221
RTD|        7567|      173788
RTD|        7544|      181332
RTD|        5244|      186576

My problem with non-RT application happens only if RT application 
(running in the same time) uses native skin. Perl and Eclipse 
application fails when switchbench is running, but not problems during 
switchtest.

-- 
________________________________________
Maksym Veremeyenko

Re: [Xenomai-help] mem usage and other application

[Gilles Chanteperdrix]

Maksym Veremeyenko wrote:
 > Gilles Chanteperdrix ~~~~~:
 > 
 > > What I would like to know is if switchtest fails in the same condition
 > > as your non real-time applications, just to see if the situation you
 > > observe is an FPU bug. But for switchtest to start correctly, you need
 > > to increase CONFIG_XENO_OPT_SYS_HEAPSZ.
 > > 
 > 
 > I significantly increased CONFIG_XENO_OPT_SYS_HEAPSZ and switchtest 
 > seems to me runs fine:
 > 
 > [root@domain.hid bin]# ./switchtest
 > == Testing FPU check routines...
 > r0: 1 != 2
 > r1: 1 != 2
 > r2: 1 != 2
 > r3: 1 != 2
 > r4: 1 != 2
 > r5: 1 != 2
 > r6: 1 != 2
 > r7: 1 != 2
 > == FPU check routines: OK.
 > == Threads: sleeper_ufps-0 rtk-1 rtk-2 rtk_fp-3 rtk_fp-4 rtk_fp_ufpp-5 
 > rtk_fp_ufpp-6 rtup-7 rtup-8 rtup_ufpp-9 rtup_ufpp-10 rtus-11 rtus-12 
 > rtus_ufps-13 rtus_ufps-14 rtuo-15 rtuo-16 rtuo_ufpp-17 rtuo_ufpp-18 
 > rtuo_ufps-19 rtuo_ufps-20 rtuo_ufpp_ufps-21 rtuo_ufpp_ufps-22
 > RTT|  00:00:01
 > RTH|ctx switches|-------total
 > RTD|        7544|        7544
 > RTD|        7590|       15134
 > RTD|        7567|       22701
 > RTD|        7590|       30291
 > RTD|        7544|       37835
 > RTD|        7544|       45379
 > RTD|        7544|       52923
 > RTD|        7567|       60490
 > RTD|        7567|       68057
 > RTD|        7498|       75555
 > RTD|        7544|       83099
 > RTD|        7521|       90620
 > RTD|        7567|       98187
 > RTD|        7544|      105731
 > RTD|        7613|      113344
 > RTD|        7567|      120911
 > RTD|        7544|      128455
 > RTD|        7521|      135976
 > RTD|        7567|      143543
 > RTD|        7590|      151133
 > RTD|        7475|      158608
 > RTT|  00:00:22
 > RTH|ctx switches|-------total
 > RTD|        7613|      166221
 > RTD|        7567|      173788
 > RTD|        7544|      181332
 > RTD|        5244|      186576
 > 
 > My problem with non-RT application happens only if RT application 
 > (running in the same time) uses native skin. Perl and Eclipse 
 > application fails when switchbench is running, but not problems during 
 > switchtest.

Ok. If you can run switchtest at the same time as your application and
switchtest does not fail, it does not look like an FPU bug. What is
the error message when perl or eclipse fail to start ?

-- 


					    Gilles Chanteperdrix.

Re: [Xenomai-help] mem usage and other application

[Maksym Veremeyenko]

Gilles Chanteperdrix пишет:

>  > My problem with non-RT application happens only if RT application 
>  > (running in the same time) uses native skin. Perl and Eclipse 
>  > application fails when switchbench is running, but not problems during 
>  > switchtest.
> 
> Ok. If you can run switchtest at the same time as your application and
> switchtest does not fail, it does not look like an FPU bug. What is
> the error message when perl or eclipse fail to start ?
> 

Perl scripts fails with various compilation errors - approximately 6 
types (seems it could be much more if more modules were used).

Eclipse fails with Java VM init error:
[verem@domain.hid ~]$ /usr/local//eclipse/eclipse
Error occurred during initialization of VM
java.lang.IllegalArgumentException: Illegal Load: NaN

During runned 'latency' i even have ssh fail:
Mar 15 01:47:18 dev-3 sshd[2908]: fatal: Couldn't obtain random bytes 
(error 604389476)
Mar 15 01:52:11 dev-3 sshd[3006]: Received signal 15; terminating.



-- 
________________________________________
Maksym Veremeyenko

Re: [Xenomai-help] mem usage and other application

[Gilles Chanteperdrix]

[-- Attachment #1: message body and .signature --]
[-- Type: text/plain, Size: 1316 bytes --]

Maksym Veremeyenko wrote:
 > Gilles Chanteperdrix ~~~~~:
 > 
 > >  > My problem with non-RT application happens only if RT application 
 > >  > (running in the same time) uses native skin. Perl and Eclipse 
 > >  > application fails when switchbench is running, but not problems during 
 > >  > switchtest.
 > > 
 > > Ok. If you can run switchtest at the same time as your application and
 > > switchtest does not fail, it does not look like an FPU bug. What is
 > > the error message when perl or eclipse fail to start ?
 > > 
 > 
 > Perl scripts fails with various compilation errors - approximately 6 
 > types (seems it could be much more if more modules were used).
 > 
 > Eclipse fails with Java VM init error:
 > [verem@domain.hid ~]$ /usr/local//eclipse/eclipse
 > Error occurred during initialization of VM
 > java.lang.IllegalArgumentException: Illegal Load: NaN
 > 
 > During runned 'latency' i even have ssh fail:
 > Mar 15 01:47:18 dev-3 sshd[2908]: fatal: Couldn't obtain random bytes 
 > (error 604389476)
 > Mar 15 01:52:11 dev-3 sshd[3006]: Received signal 15; terminating.

Well, it really looks like an FPU bug. Could you try to:
- upgrade your trunk to latest revision, then test
- then apply the attached patch to your up-to-date trunk, then test
  again ?

-- 


					    Gilles Chanteperdrix.

[-- Attachment #2: xeno-revert-all-fpu-changes.diff --]
[-- Type: text/plain, Size: 1648 bytes --]

Index: include/asm-i386/bits/pod.h
===================================================================
--- include/asm-i386/bits/pod.h	(revision 2299)
+++ include/asm-i386/bits/pod.h	(working copy)
@@ -209,7 +209,6 @@
 {
 	struct task_struct *task = tcb->user_task;
 
-	if (!tcb->is_root) {
 		if (task) {
 			/* fpu not used or already saved by __switch_to. */
 			if (!wrap_test_fpu_used(task))
@@ -219,14 +218,7 @@
 		   	hardware of this task. */
 			wrap_clear_fpu_used(task);
 		}
-	} else {
-		if (tcb->cr0_ts || 
-		    (tcb->ts_usedfpu && !wrap_test_fpu_used(task)))
-			return;
 
-		wrap_clear_fpu_used(task);
-	}
-
 	clts();
 
 	if (cpu_has_fxsr)
@@ -239,7 +231,6 @@
 {
 	struct task_struct *task = tcb->user_task;
 
-	if (!tcb->is_root) {
 		if (task) {
 			if (!xnarch_fpu_init_p(task)) {
 				stts();
@@ -250,17 +241,7 @@
 			 * the FPU hardware. */
 			wrap_set_fpu_used(task);
 		}
-	} else {
-		/* Restore state of FPU only if TS bit in cr0 was clear. */
-		if (tcb->cr0_ts) {
-			stts();
-			return;
-		}
 
-		if (tcb->ts_usedfpu)
-			wrap_set_fpu_used(task);
-	}
-
 	/* Restore the FPU hardware with valid fp registers from a
 	   user-space or kernel thread. */
 	clts();
@@ -277,7 +258,6 @@
 {
 	struct task_struct *task = tcb->user_task;
 
-	if (!tcb->is_root) {
 		if (task) {
 			if (!xnarch_fpu_init_p(task))
 				return;
@@ -289,20 +269,7 @@
 				return;
 			}
 		}
-	} else {
-		if (tcb->cr0_ts)
-			return;
 
-		if (wrap_test_fpu_used(task)) {
-			/* Fpu context was not even saved, do not restore */
-			clts();
-			return;
-		}
-		
-		xnarch_restore_fpu(tcb);
-		return;
-	}
-
 	clts();
 
 	if (!cpu_has_fxsr && task)

Re: [Xenomai-help] mem usage and other application

[Maksym Veremeyenko]

Gilles Chanteperdrix пишет:

> Well, it really looks like an FPU bug. Could you try to:
> - upgrade your trunk to latest revision, then test
> - then apply the attached patch to your up-to-date trunk, then test
>   again ?
> 

I used configuration 2.6.17.14, xenomai-r2299

Before patching xenomai result was the same and more faulty after 
decreasing RAM size down to 256M.

After applying patch situation become worth, i got "Bad pte" messages, 
sometimes "Soft lockup detected", and kernel dump:

Mar 15 04:08:21 dev-3 kernel: Bad pte = 42a5fef8, process = ???, 
vm_flags = 100073, vaddr = 42f60000
Mar 15 04:08:21 dev-3 kernel:  <c0102f70> show_trace+0x12/0x14 
<c0103474> dump_stack+0x1c/0x1e
Mar 15 04:08:21 dev-3 kernel:  <c015c460> print_bad_pte+0x4f/0x5a 
<c015c4b1> vm_normal_page+0x46/0x5c
Mar 15 04:08:21 dev-3 kernel:  <c015c9d4> unmap_vmas+0x162/0x3e4 
<c015f059> exit_mmap+0x55/0xc1
Mar 15 04:08:21 dev-3 kernel:  <c010feb8> mmput+0x5a/0xb0  <c0113161> 
exit_mm+0xe8/0xee
Mar 15 04:08:21 dev-3 kernel:  <c011459f> do_exit+0x1a2/0x719 
<c010d21f> do_page_fault+0x4bf/0x542
Mar 15 04:08:21 dev-3 kernel:  <c010b1da> 
__ipipe_handle_exception+0x89/0xbf  <c0102988> error_code+0x54/0x64

Mar 15 04:08:21 dev-3 kernel: Bad pte = 42a5fef8, process = ???, 
vm_flags = 75, vaddr = 42f62000
Mar 15 04:08:21 dev-3 kernel:  <c0102f70> show_trace+0x12/0x14 
<c0103474> dump_stack+0x1c/0x1e
Mar 15 04:08:21 dev-3 kernel:  <c015c460> print_bad_pte+0x4f/0x5a 
<c015c4b1> vm_normal_page+0x46/0x5c
Mar 15 04:08:21 dev-3 kernel:  <c015c9d4> unmap_vmas+0x162/0x3e4 
<c015f059> exit_mmap+0x55/0xc1
Mar 15 04:08:21 dev-3 kernel:  <c010feb8> mmput+0x5a/0xb0  <c0113161> 
exit_mm+0xe8/0xee
Mar 15 04:08:21 dev-3 kernel:  <c011459f> do_exit+0x1a2/0x719 
<c010d21f> do_page_fault+0x4bf/0x542
Mar 15 04:08:21 dev-3 kernel:  <c010b1da> 
__ipipe_handle_exception+0x89/0xbf  <c0102988> error_code+0x54/0x64

Mar 15 04:08:23 dev-3 kernel: ------------[ cut here ]------------
Mar 15 04:08:23 dev-3 kernel: kernel BUG at mm/rmap.c:560!
Mar 15 04:08:23 dev-3 kernel: invalid opcode: 0000 [#1]
Mar 15 04:08:23 dev-3 kernel: Modules linked in: xeno_16550A
Mar 15 04:08:23 dev-3 kernel: CPU:    0
Mar 15 04:08:23 dev-3 kernel: EIP:    0060:[<c01619f0>]    Not tainted VLI
Mar 15 04:08:23 dev-3 kernel: EFLAGS: 00010286 
(2.6.17.14-xenomai-r2299-min-debug #3)
Mar 15 04:08:23 dev-3 kernel: EIP is at page_remove_rmap+0x1b/0x32
Mar 15 04:08:23 dev-3 kernel: eax: ffffffff   ebx: c1000000   ecx: 
c04d2aec   edx: c1000000
Mar 15 04:08:23 dev-3 kernel: esi: b7f79000   edi: d36f1de4   ebp: 
d351dce4   esp: d351dce4
Mar 15 04:08:23 dev-3 kernel: ds: 007b   es: 007b   ss: 0068
Mar 15 04:08:23 dev-3 kernel: Process bash (pid: 2698, 
threadinfo=d351c000 task=d500c030)
Mar 15 04:08:23 dev-3 kernel: Stack: d351dd3c c015caa6 00000000 d341f6fc 
d351dd54 00000000 00000001 b7fb3000
Mar 15 04:08:23 dev-3 kernel:        d3a5eb7c d55faa1c c04d2aec 00000000 
ffffffff d55faa6c d3a5eb7c 00000019
Mar 15 04:08:23 dev-3 kernel:        003a4c2b b7fb3000 00000000 d351dd54 
d3a08e04 d55faa1c d351dd64 c015f059
Mar 15 04:08:23 dev-3 kernel: Call Trace:
Mar 15 04:08:23 dev-3 kernel:  <c0102f26> show_stack_log_lvl+0x8a/0x94 
<c01030ba> show_registers+0x148/0x1b4
Mar 15 04:08:23 dev-3 kernel:  <c010328a> die+0x164/0x262  <c0103409> 
do_trap+0x81/0x9b
Mar 15 04:08:23 dev-3 kernel:  <c0103b5e> do_invalid_op+0x8e/0x98 
<c010b1da> __ipipe_handle_exception+0x89/0xbf
Mar 15 04:08:23 dev-3 kernel:  <c0102988> error_code+0x54/0x64 
<c015caa6> unmap_vmas+0x234/0x3e4
Mar 15 04:08:23 dev-3 kernel:  <c015f059> exit_mmap+0x55/0xc1 
<c010feb8> mmput+0x5a/0xb0
Mar 15 04:08:23 dev-3 kernel:  <c0175321> flush_old_exec+0x599/0x7c6 
<c01904ad> load_elf_binary+0x4bf/0x1478
Mar 15 04:08:23 dev-3 kernel:  <c0174718> 
search_binary_handler+0x76/0x181  <c018f8e7> load_script+0x1b7/0x1cc
Mar 15 04:08:23 dev-3 kernel:  <c0174718> 
search_binary_handler+0x76/0x181  <c01760f5> do_execve+0x13f/0x1ac
Mar 15 04:08:23 dev-3 kernel:  <c010142b> sys_execve+0x2f/0x76 
<c010269e> sysenter_past_esp+0x63/0x67
Mar 15 04:08:23 dev-3 kernel: Code: e8 7a 4b 13 00 8b 45 d0 83 c4 24 5b 
5e 5f c9 c3 55 89 e5 e8 33 98 fa ff 89 c2 83 40 08 f
Mar 15 04:08:23 dev-3 kernel: EIP: [<c01619f0>] 
page_remove_rmap+0x1b/0x32 SS:ESP 0068:d351dce4

With latest trunk and 256MB i get stable errors even with "perl 
Makefile.PL" for  XML-DOM-1.44 then any RT application used Native skin 
running.


-- 
________________________________________
Maksym Veremeyenko