[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

POSIX signal handling



How about this version?

RTEMS is multi-threaded within a single process and no thread
had ever enabled SIGALRM. 

This MIGHT be a buglet in relation to the default signal mask
but we would have to consult the Open Group specification to
be sure.

S?rgio Santos wrote:
> Thanks again for your kind answer. 
>
> Actually we've had that very output... since the beginning :)
> Ever since the problem for us is we're expecting not the signal to be pending, but otherwise that the handler *sig_handler* runs when the signal is received!
>  
> We're running the same code, and - glad to know by comparing - no problems coming from our makefile. So we'd expect something like the output in Linux (every second):
>
> [sbrotas at imaserver2 sigact]$ ./sigact
> HANDLER: caught SIGALRM
> SIGALRM not pending :)
>
> HANDLER: caught SIGALRM
> SIGALRM not pending :)
>
> ^C
> [sbrotas at imaserver2 sigact]$
>
> ---
> Best Regards
> S?rgio
>
>
>
>   
>> -----Original Message-----
>> From: Joel Sherrill [mailto:joel.sherrill at OARcorp.com]
>> Sent: ter?a-feira, 31 de Mar?o de 2009 14:11
>> To: S?rgio Santos
>> Cc: rtems-users at rtems.com
>> Subject: Re: POSIX signal handling
>>
>> S?rgio Santos wrote:
>>     
>>> Hi, Joel
>>> I'm using qemu (mostly) but have tried running directly on the target
>>>       
>> too.
>>     
>>> Our current guess is some build configuration conflict or mistake, but
>>>       
>> we're still to find what exactly. Still, we haven't debugged much for some
>> clues or looked at the rtems code.
>>     
>> Runs for me.  I attached my version of your code and the
>> Makefile.
>>
>> Built on an installed BSP like this:
>>
>>
>>  RTEMS_MAKEFILE_PATH=/home/joel/rtems-4.10-work/bsp-install/i386-
>> rtems4.10/pc386
>> make clean all
>>
>> [joel at iceland sigaction]$ runqemu -i */*exe
>> Running o-optimize/sigaction.exe
>> Could not open '/dev/kqemu' - QEMU acceleration layer not activated: No
>> such file or directory
>> SIGALRM is pending
>>
>> SIGALRM is pending
>>
>> SIGALRM is pending
>>
>> SIGALRM is pending
>>
>> SIGALRM is pending
>>
>> SIGALRM is pending
>>
>> /home/joel/bin/runqemu: line 65: 15963 Killed                  ${QEMU}
>> ${ARGS}
>> Could not open '/dev/kqemu' - QEMU acceleration layer not activated: No
>> such file or directory
>> SIGALRM is pending
>>
>>     
>>> Our option is now getting closer to drop POSIX and go for the RTEMS
>>>       
>> classic API on the async signals matter in particular, and for the port of
>> our Linux app overall. Still we'll report any findings.
>>     
>>>       
>>> Thank you much
>>>
>>> S?rgio
>>>
>>>
>>>
>>>       
>>>> -----Original Message-----
>>>> From: Joel Sherrill [mailto:joel.sherrill at OARcorp.com]
>>>> Sent: segunda-feira, 30 de Mar?o de 2009 20:30
>>>> To: S?rgio Santos
>>>> Cc: rtems-users at rtems.com
>>>> Subject: Re: POSIX signal handling
>>>>
>>>> S?rgio Santos wrote:
>>>>
>>>>         
>>>>>> -----Original Message-----
>>>>>> From: Joel Sherrill [mailto:joel.sherrill at OARcorp.com]
>>>>>> Sent: segunda-feira, 30 de Mar?o de 2009 19:11
>>>>>> To: S?rgio Santos
>>>>>> Cc: rtems-users at rtems.com
>>>>>> Subject: Re: POSIX signal handling
>>>>>>
>>>>>> Sergios,
>>>>>>
>>>>>> I meant to ask.. Is it ok to include this in the RTEMS examples
>>>>>> as a simple example of alarm and signals?
>>>>>>
>>>>>>
>>>>>>             
>>>>> Joel,
>>>>>
>>>>> Sure! Can't help feeling it's a privilege, despite the simplicity of
>>>>>           
>> the
>>     
>>>> code :) Nevertheless I must say I'd be happier if it would work at all
>>>>         
>> in
>>     
>>>> my system! (please see below)
>>>>
>>>>         
>>>>>           
>>>>>> Joel Sherrill wrote:
>>>>>>
>>>>>>
>>>>>>             
>>>>>>> OK.. runs for me with the CVS head.
>>>>>>>
>>>>>>> You forget to include <rtems/confdefs.h> at the bottom after
>>>>>>>
>>>>>>>
>>>>>>>               
>>>>>> CONFIGURE_INIT.
>>>>>>
>>>>>>
>>>>>>             
>>>>> You're right in pointing this, it was actually an unfortunate typo...
>>>>>
>>>>>
>>>>>
>>>>>           
>>>> :)  I expected as much.  :)
>>>>
>>>>         
>>>>>>> And delete the timer configuration.
>>>>>>>
>>>>>>>
>>>>>>>               
>>>>> Done already! So a conclusion is still open for me, the code is ok,
>>>>>
>>>>>           
>>>> configuration is (now) ok, but instead of the current 4.8.1, I should
>>>>         
>> try
>>     
>>>> it with the CVS head (4.9.99?)
>>>>
>>>>         
>>>> How are you running it?
>>>>
>>>>         
>>>>> Thanks again for your help,
>>>>>
>>>>> S?rgio
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>           
>>>>>>> Folks .. please consider support agreements. :-D
>>>>>>> --joel
>>>>>>>
>>>>>>> S?rgio Santos wrote:
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>               
>>>>>>>> Hello all,
>>>>>>>>
>>>>>>>> I'm using rtems 4.8.1, target is i386 (single core, with FC9),
>>>>>>>>                 
>> using
>>     
>>>>>>>>                 
>>>>>> bsp's between pc386 and pc686 and the RTEMS makefile templates.
>>>>>>
>>>>>>
>>>>>>             
>>>>>>>> I have been trying to port a Linux based application to RTEMS and
>>>>>>>>
>>>>>>>>                 
>>>> found
>>>>
>>>>         
>>>>>> at some point I couldn't get POSIX signal handlers (set with
>>>>>>             
>> sigaction)
>>     
>>>> to
>>>>
>>>>         
>>>>>> ever execute. From what I've read in rtems-users and user docs, no
>>>>>> problems should arise from using POSIX asynchronous signal handling
>>>>>>             
>> in
>>     
>>>>>> RTEMS, so my guess is I'm making some na?ve mistake on code or
>>>>>> configuration...
>>>>>>
>>>>>>
>>>>>>             
>>>>>>>> Also, searching on the list, came across a similar problem, but
>>>>>>>>                 
>> this
>>     
>>>>>>>>                 
>>>>>> was a unique case and a long while ago:
>>>>>>
>>>>>>
>>>>>>             
>>>>>>>> http://www.rtems.com/ml/rtems-users/2002/september/msg00015.html
>>>>>>>> Unfortunately that thread was not conclusive.
>>>>>>>>
>>>>>>>> After a few other tests I've got down to a simple test application
>>>>>>>>
>>>>>>>>                 
>>>> (see
>>>>
>>>>         
>>>>>> below), similar to the code in the post I mentioned above: a thread
>>>>>>
>>>>>>             
>>>> sets
>>>>
>>>>         
>>>>>> sigaction and just does alarm(1), sleep(3) in a cycle. Alike the
>>>>>>             
>> other
>>     
>>>>>> tests, in RTEMS the handler never gets to run (although the signal
>>>>>> (SIGALRM) is seen pending); in Linux all tests will work.
>>>>>>
>>>>>>
>>>>>>             
>>>>>>>> POSIX was enabled in the build configuration:
>>>>>>>> ../../../rtems-4.8.1/c/configure --prefix=/opt/rtems-4.8 --
>>>>>>>>                 
>> host=i386-
>>     
>>>>>>>>                 
>>>>>> rtems4.8 --build=i686-pc-linux-gnu --target=i386-rtems4.8 --enable-
>>>>>>
>>>>>>             
>>>> posix -
>>>>
>>>>         
>>>>>> -disable-itron --enable-networking --enable-cxx --enable-rtems-debug
>>>>>>             
>> --
>>     
>>>>>> enable-rdbg --enable-rtemsbsp=pc386 --with-target-subdir=i386-
>>>>>>             
>> rtems4.8
>>     
>>>> --
>>>>
>>>>         
>>>>>> exec-prefix=/opt/rtems-4.8/i386-rtems4.8 --includedir=/opt/rtems-
>>>>>>
>>>>>>             
>>>> 4.8/i386-
>>>>
>>>>         
>>>>>> rtems4.8/include --cache-file=/dev/null --srcdir=../../../rtems-
>>>>>>             
>> 4.8.1/c
>>     
>>>>>>             
>>>>>>>> I post below the src for the simple test, having the RTEMS
>>>>>>>>
>>>>>>>>                 
>>>> definitions
>>>>
>>>>         
>>>>>> in the end.
>>>>>>
>>>>>>
>>>>>>             
>>>>>>>> Thanks in advance for your help!
>>>>>>>>
>>>>>>>> S?rgio Santos
>>>>>>>>
>>>>>>>>
>>>>>>>> --------------
>>>>>>>>
>>>>>>>>
>>>>>>>> #include <signal.h>
>>>>>>>> #include <stdio.h>
>>>>>>>> #include <unistd.h>
>>>>>>>> #include <stdlib.h>
>>>>>>>>
>>>>>>>> #include <bsp.h>
>>>>>>>>
>>>>>>>> void sig_handler(int sig){
>>>>>>>> 	switch(sig){
>>>>>>>> 	case SIGALRM:
>>>>>>>> 		printf("HANDLER: caught SIGALRM\n");
>>>>>>>> 		break;
>>>>>>>> 	default:
>>>>>>>> 		printf("Received unexpected signal (%d)\nBye!", sig);
>>>>>>>> 		exit(1);
>>>>>>>> 	}
>>>>>>>> }
>>>>>>>>
>>>>>>>> void sig_init(){
>>>>>>>>
>>>>>>>> 	int rc;
>>>>>>>> 	sigset_t 			sigset;
>>>>>>>> 	struct sigaction 	sigact;
>>>>>>>>
>>>>>>>> 	rc = sigemptyset(&sigset);
>>>>>>>> 	if (rc != 0) {
>>>>>>>> 		printf("Can't empty set\n");exit(1);
>>>>>>>> 	}
>>>>>>>>
>>>>>>>> 	sigact.sa_mask 		= sigset;
>>>>>>>> 	sigact.sa_flags 	= 0;
>>>>>>>> 	sigact.sa_handler 	= &sig_handler;
>>>>>>>>
>>>>>>>> 	rc = sigaction(SIGALRM, &sigact, NULL);
>>>>>>>> 	if (rc != 0) {
>>>>>>>> 		printf("Can't install handler\n"); exit(1);
>>>>>>>> 	}
>>>>>>>> }
>>>>>>>>
>>>>>>>> void *POSIX_Init(void *arg){
>>>>>>>>
>>>>>>>> 	int rc = 0;
>>>>>>>> 	sigset_t pendingset;
>>>>>>>>
>>>>>>>> 	sig_init();
>>>>>>>>
>>>>>>>> 	rc = sigemptyset(&pendingset);
>>>>>>>> 	if (rc != 0) {
>>>>>>>> 		printf("Can't empty set\n");exit(1);
>>>>>>>> 	}
>>>>>>>>
>>>>>>>> 	while(1){
>>>>>>>> 		alarm(1);
>>>>>>>> 		sleep(3);
>>>>>>>> 		sigpending(&pendingset);
>>>>>>>> 		printf("%s\n\n", ( sigismember(&pendingset, SIGALRM) ?
>>>>>>>> 				"SIGALRM is pending" : "SIGALRM not pending"
>>>>>>>>                 
>> ));
>>     
>>>>>>>> 	}
>>>>>>>> 	return 0;
>>>>>>>> }
>>>>>>>>
>>>>>>>> #define CONFIGURE_MAXIMUM_POSIX_THREADS 10
>>>>>>>> #define CONFIGURE_MAXIMUM_POSIX_MUTEXES 10
>>>>>>>> #define CONFIGURE_MAXIMUM_POSIX_TIMERS  10
>>>>>>>> #define CONFIGURE_MAXIMUM_POSIX_QUEUED_SIGNALS 40
>>>>>>>>
>>>>>>>> #define CONFIGURE_APPLICATION_NEEDS_CLOCK_DRIVER
>>>>>>>> #define CONFIGURE_APPLICATION_NEEDS_TIMER_DRIVER
>>>>>>>> #define CONFIGURE_APPLICATION_NEEDS_CONSOLE_DRIVER
>>>>>>>> #define CONFIGURE_POSIX_INIT_THREAD_TABLE
>>>>>>>> #define CONFIGURE_INIT
>>>>>>>>
>>>>>>>> // EOF
>>>>>>>> ----
>>>>>>>>
>>>>>>>> _______________________________________________
>>>>>>>> rtems-users mailing list
>>>>>>>> rtems-users at rtems.com
>>>>>>>> http://rtems.rtems.org/mailman/listinfo/rtems-users
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>                 
>>>>>> --
>>>>>> Joel Sherrill, Ph.D.             Director of Research & Development
>>>>>> joel.sherrill at OARcorp.com        On-Line Applications Research
>>>>>> Ask me about RTEMS: a free RTOS  Huntsville AL 35805
>>>>>>    Support Available             (256) 722-9985
>>>>>>
>>>>>>
>>>>>>
>>>>>>             
>>>> --
>>>> Joel Sherrill, Ph.D.             Director of Research & Development
>>>> joel.sherrill at OARcorp.com        On-Line Applications Research
>>>> Ask me about RTEMS: a free RTOS  Huntsville AL 35805
>>>>    Support Available             (256) 722-9985
>>>>
>>>>
>>>>         
>>>       
>> --
>> Joel Sherrill, Ph.D.             Director of Research & Development
>> joel.sherrill at OARcorp.com        On-Line Applications Research
>> Ask me about RTEMS: a free RTOS  Huntsville AL 35805
>>    Support Available             (256) 722-9985
>>
>>     
>
>   


-- 
Joel Sherrill, Ph.D.             Director of Research & Development
joel.sherrill at OARcorp.com        On-Line Applications Research
Ask me about RTEMS: a free RTOS  Huntsville AL 35805
   Support Available             (256) 722-9985


-------------- next part --------------
A non-text attachment was scrubbed...
Name: init.c
Type: text/x-csrc
Size: 1685 bytes
Desc: not available
Url : http://rtems.rtems.org/pipermail/rtems-users/attachments/20090331/6650c7d5/attachment-0001.bin