Re: [Valgrind-users] Semaphore error on Helgrind and Drd?

[Paul F. <pj...@wa...>]

On 25/05/2024 12:29, Paulo Ferreira wrote:
> Error message from Helgrind: 
> 
> $ valgrind --tool=helgrind  ./prog 
> ...
> ==6587== Thread #1: Bug in libpthread: sem_wait succeeded on semaphore without prior sem_post
> ==6587==    at 0x4850069: sem_wait_WRK (hg_intercepts.c:3155)
> ==6587==    by 0x4851180: sem_wait@* (hg_intercepts.c:3174)
> ==6587==    by 0x109271: main (in /media/sf_antiX/erros/duvida/prog)
> …
> 
> Error message from Drd: 
> 
> $ valgrind --tool=drd --trace-semaphore=yes  ./prog
> ...
> ==6828== [1] sem_wait      0x487f000 value 0 -> 4294967295
> ==6828== Invalid semaphore: semaphore 0x487f000
> ==6828==    at 0x4866C1A: sem_wait_intercept (drd_pthread_intercepts.c:1570)
> ==6828==    by 0x4866C1A: sem_wait@* (drd_pthread_intercepts.c:1578)
> ==6828==    by 0x109271: main (in /media/sf_antiX/erros/duvida/prog)
> ==6828== semaphore 0x487f000 was first observed at:
> ==6828==    at 0x4865A16: sem_open_intercept (drd_pthread_intercepts.c:1527)
> ==6828==    by 0x4865A16: sem_open@* (drd_pthread_intercepts.c:1538)
> ==6828==    by 0x1091D3: main (in /media/sf_antiX/erros/duvida/prog)
> …  
> 
> So the issue seems to be that according to Drd a semaphore that has the value of zero, with a sem_wait() gets decremented to 4294967295. (????????) 
> I have the sample code on the end of this message, and I have tried to run the code on several recent versions of Valgrind (32.3 and others), and on x86-64 using antiX Linux, MX Linux, Ubuntu Linux and FreeBSD. I have also tried ARM 64 bits under Ubuntu, and the behaviour is the same. The sample program seems to work ok. 
> 
> 	pid_t pid = fork(); 
>     if (pid>0) // Parent process
Hi

Semaphores can be used with both threads and processes. I don't think that Valgrind works with semaphores across processes, only threads. 

Regards

Paul