On Apr 21, 2012, at 5:26 PM, Brice Goglin wrote:

On 21/04/2012 23:08, Vlad wrote:

I use hwloc-1.4.1 stable on Red Hat 5 and am seeing a possible concurrency issue not covered by the "Thread Safety" guidelines:

- I start a small number (4) of threads,  each of which does some work and periodically executes hwloc_get_last_cpu_location() with HWLOC_CPUBIND_PROCESS
- occasionally, one or two of those threads will see the call fail with ENOSYS (even though the same call has already executed successfully a number of times)

These errors are transient and seem to occur only when some of the threads in the group are terminating. I've skimmed through the implementation in topology-linux.c and it seems plausible to me that the errors could be caused by failure to read /proc state "atomically" in the presence of concurrent thread starts/exits.

Of course, the latter is hard (impossible ?) to do because the state always changes and a snapshot can only be obtained with a single read() (which in turn would require knowing how many thread entries to expect in advance). However, returning ENOSYS in such cases does not seems intended but rather a flaw in retry logic. Similar issues may be present with other API methods that rely on hwloc_linux_foreach_proc_tid() or hwloc_linux_get_proc_tids().

Can you try the attached patch? It doesn't abort the loop immediately on per-tid errors anymore. This may work better when threads disappear. I don't remember if the retry logic was written while thinking about adding threads only or about adding and removing threads.

If the patch doesn't help, can you send your code to help debug things?

Will try this within a day or two. At the moment I am simply using a retry loop on ENOSYS and usually no more than one retry is needed.

An alternative explanation could be that the retry logic is correct but the implementation relies on readdir(), which is documented to not be thread-safe: http://www.gnu.org/software/libc/manual/html_node/Reading_002fClosing-Directory.html

I don't this can happen. Your threads should not be accessing the same DIR stream here.

You are probably correct. I was thinking of this code from https://svn.open-mpi.org/trac/hwloc/browser/trunk/src/topology-linux.c:

425 hwloc_linux_get_proc_tids(DIR *taskdir, unsigned *nr_tidsp, pid_t ** tidsp)
426 {
427  struct dirent *dirent;
428  unsigned nr_tids = 0;
429  unsigned max_tids = 32;
430  pid_t *tids;
431  struct stat sb;
433  /* take the number of links as a good estimate for the number of tids */
434  if (fstat(dirfd(taskdir), &sb) == 0)
435    max_tids = sb.st_nlink;
437  tids = malloc(max_tids*sizeof(pid_t));
438  if (!tids) {
439    errno = ENOMEM;
440    return -1;
441  }
443  rewinddir(taskdir);
445  while ((dirent = readdir(taskdir)) != NULL) {

"taskdir" here is /proc/<pid>/task, correct? In which case the threads will be doing readdir() on the same DIR stream...


hwloc-users mailing list