[Pvfs2-developers] help with kernel dentry revalidate problem

[Sam Lang]

Hi Scott,

You're right, in fact the simul test does do a barrier after the  
stat.  In the code below, the end() function does an MPI_Barrier.

-sam

void simul_file_stat(int shared) {
     char *filename;
     struct stat buf;

     begin("setup");
     filename = create_files("simul_file_stat", 0, shared);
     end("setup");

     begin("test");
     /* All stat the file */
     if (stat(filename, &buf) == -1) {
         FAIL("stat failed");
     }
     end("test");

     begin("cleanup");
     remove_files("simul_file_stat", shared);
     end("cleanup");

On Aug 15, 2007, at 7:37 PM, Scott Atchley wrote:

> On Aug 15, 2007, at 6:53 PM, Pete Wyckoff wrote:
>
>>     if (task == 0)
>> 	mkdir("foo");
>>
>>     MPI_Barrier();
>>     sleep(3);
>>
>>     stat("foo");
>>
>>     if (task == 0)
>>     	rmdir("foo");
>>
>
> Hi Pete,
>
> Isn't there a race condition between task 0 and the others after  
> the barrier? Can 0 stat the file and rmdir it before another task  
> can stat() it? If so, should there be a barrier after the stat()  
> and before the rmdir?
>
> Scott
> _______________________________________________
> Pvfs2-developers mailing list
> Pvfs2-developers at beowulf-underground.org
> http://www.beowulf-underground.org/mailman/listinfo/pvfs2-developers
>