fcoll/two_phase: fix coverity errors #963

hjelmn · 2015-10-01T16:55:20Z

Fixes CIDs 72300, 72344, 1196764-1196768, 72300: Resource leaks

Mulitple allocated arrays are going out of scope at the end of
mca_fcoll_two_phase_file_write_all. Free these arrays. Also removed
the extraneous NULL checks since free (NULL) is safe in C.

Change returns to goto exit where the allocated resources are freed.

Fixes CIDs 72285-72292, 72297, 72298: Resource leaks

Change all appropriate return statements to goto exit to ensure that
all resources are freed. Also removed the NULL checks since free
(NULL) is safe in C.

Fixes CIDs 72295, 72296: Resource leaks

Moved free of requests and recv_types to after exit label. This will
ensure these are freed on error.

Also added a loop and statement to free send_buf which is going out of
scope at the end of the function.

Fixes CIDs 72336-72240, 735197, 735198: Resource leaks

Moved the exit label before to before the resources are released and
changed all appropriate return statements to goto exit. Also removed
extraneous NULL checks because free (NULL) is safe in C.

Fixes CIDs 72341, 72343, 1196805-1196809: Resource leaks

Free all resources after exit label and change return statements to
goto exit to ensure all resources are freed on error.

Fixes CID 1269973: Unused value

Check return code of ompi_request_wait_all. If it fails jump to the
exit.

Fixes CID 714119: Dereference before NULL check

Wrong value checked in conditional.

Signed-off-by: Nathan Hjelm [email protected]

hjelmn · 2015-10-01T16:55:58Z

@edgargabriel Please review. This fixes a large number of issues identified by coverity. There are others that remain but the fixes are not as straight-forward.

hjelmn · 2015-10-01T17:57:41Z

Fixed typos.

lanl-ompi · 2015-10-01T18:16:17Z

Test FAILed.

edgargabriel · 2015-10-01T18:53:12Z

What is the reason that you removed all the checks whether the pointer is a NULL pointer before freeing the memory for all variables? I do not see any benefit on that, and in fact, I am pretty sure e.g. in the dynamic segmentation algorithm they required for correctness inside the while loop.

hjelmn · 2015-10-01T19:24:28Z

free (NULL) is completely safe in C. free checks for NULL already and returns if the pointer is NULL. With the check before free you end up checking the conditional twice for the common case (non-NULL pointer).

edgargabriel · 2015-10-01T19:26:58Z

ok, grabbing a copy of your repository and testing whether our tests still work.

edgargabriel · 2015-10-01T20:01:14Z

the changes to dynamic module seem to be ok, the two_phase module generates now however a segfault in a free statement.

gabriel@salmon:~/ompi-tests/mpi2basic_tests/file> mpirun --mca fcoll two_phase -np 6 ./filetest
Checking for MPI_File_open:
using MPI_MODE_CREATE..................working
using MPI_MODE_WRONLY..................working
using MPI_MODE_RDONLY..................working
using MPI_MODE_RDWR....................working
using MPI_MODE_APPEND..................working
using MPI_MODE_EXCL....................working
using MPI_MODE_DELETE_ON_CLOSE.........working
Checking for MPI_File_close................working
Checking for MPI_File_delete...............working
Checking for MPI_File_write:
using single proc no fview.............working
verifying status fields................working
opening file in append mode............working
writing a large file...................working
[salmon:07724] *** Process received signal ***
[salmon:07724] Signal: Segmentation fault (11)
[salmon:07724] Signal code: Address not mapped (1)
[salmon:07724] Failing at address: 0xfffffffffffffff9
[salmon:07724] [ 0] /lib64/libpthread.so.0(+0xf890)[0x7f61c8306890]
[salmon:07724] [ 1] /lib64/libc.so.6(cfree+0x14)[0x7f61c7fcbd34]
[salmon:07724] [ 2] /home/gabriel/OpenMPI/lib64/libopen-pal.so.0(opal_free+0x1f)[0x7f61c79aee9d]
[salmon:07724] [ 3] /home/gabriel/OpenMPI/lib64/openmpi/mca_fcoll_two_phase.so(mca_fcoll_two_phase_file_write_all+0xc04)[0x7f61afdf9264]
[salmon:07724] [ 4] /home/gabriel/OpenMPI/lib64/openmpi/mca_io_ompio.so(ompio_io_ompio_file_write_at_all+0x70)[0x7f61bc5e5b4d]
[salmon:07724] [ 5] /home/gabriel/OpenMPI/lib64/openmpi/mca_io_ompio.so(mca_io_ompio_file_write_at_all+0x5a)[0x7f61bc5e5a29]
[salmon:07724] [ 6] /home/gabriel/OpenMPI/lib64/libmpi.so.0(PMPI_File_write_at_all+0x188)[0x7f61c85dbbdd]
[salmon:07724] [ 7] ./filetest[0x4093e9]
[salmon:07724] [ 8] ./filetest[0x40230a]
[salmon:07724] [ 9] /lib64/libc.so.6(__libc_start_main+0xf5)[0x7f61c7f71b05]
[salmon:07724] [10] ./filetest[0x402199]

[salmon:07724] * End of error message *

mpirun noticed that process rank 0 with PID 0 on node salmon exited on signal 11 (Segmentation fault).

hjelmn · 2015-10-01T20:15:17Z

@edgargabriel Ok, I will double-check. Thanks for testing. I will let you know when I have determined what I did wrong.

edgargabriel · 2015-10-01T20:32:42Z

I can also check, but it might be next week before I get to it. Thanks!

Fixes CIDs 72300, 72344, 1196764-1196768, 72300: Resource leaks Mulitple allocated arrays are going out of scope at the end of mca_fcoll_two_phase_file_write_all. Free these arrays. Also removed the extraneous NULL checks since free (NULL) is safe in C. Change returns to goto exit where the allocated resources are freed. Fixes CIDs 72285-72292, 72297, 72298: Resource leaks Change all appropriate return statements to goto exit to ensure that all resources are freed. Also removed the NULL checks since free (NULL) is safe in C. Fixes CIDs 72295, 72296: Resource leaks Moved free of requests and recv_types to after exit label. This will ensure these are freed on error. Also added a loop and statement to free send_buf which is going out of scope at the end of the function. Fixes CIDs 72336-72240, 735197, 735198: Resource leaks Moved the exit label before to before the resources are released and changed all appropriate return statements to goto exit. Also removed extraneous NULL checks because free (NULL) is safe in C. Fixes CIDs 72341, 72343, 1196805-1196809: Resource leaks Free all resources after exit label and change return statements to goto exit to ensure all resources are freed on error. Fixes CID 1269973: Unused value Check return code of ompi_request_wait_all. If it fails jump to the exit. Fixes CID 714119: Dereference before NULL check Wrong value checked in conditional. Signed-off-by: Nathan Hjelm <[email protected]>

Fixes CID 72320: Explicit NULL dereferenced On error it is possible that the blocklen_per_process array is NULL. Change the NULL check before the free to check for non-NULL on the array not the array element. Also clean up allocation of this array to use calloc instead of malloc + setting each element to NULL. Signed-off-by: Nathan Hjelm <[email protected]>

hjelmn · 2015-10-01T20:41:42Z

Found a typo in one of the changes. Fixed that error. Now I see another error but that also happens without any of my changes.

edgargabriel · 2015-10-01T20:48:51Z

you mean the error in the append test? I see that too, not sure what is triggering that, will check. It used to pass.

hjelmn · 2015-10-01T20:56:39Z

Seeing the error in the write shared test on my mac:

Checking for MPI_File_write_ordered........false
Process 42213 stopped
* thread #1: tid = 0xdaf8d, 0x0000000100051bc9 libmpi.0.dylib`ompi_request_default_wait_all(count=3, requests=0x00007fff5fbff170, statuses=0x00007fff5fbff190) + 105 at req_wait.c:233, queue = 'com.apple.main-thread', stop reason = EXC_BAD_ACCESS (code=1, address=0x1100000088)
    frame #0: 0x0000000100051bc9 libmpi.0.dylib`ompi_request_default_wait_all(count=3, requests=0x00007fff5fbff170, statuses=0x00007fff5fbff190) + 105 at req_wait.c:233
   230      for (i = 0; i < count; i++) {
   231          request = *rptr++;
   232 
-> 233          if (request->req_complete == true) {
   234              if( OPAL_UNLIKELY( MPI_SUCCESS != request->req_status.MPI_ERROR ) ) {
   235                  failed++;
   236              }
(lldb) bt
* thread #1: tid = 0xdaf8d, 0x0000000100051bc9 libmpi.0.dylib`ompi_request_default_wait_all(count=3, requests=0x00007fff5fbff170, statuses=0x00007fff5fbff190) + 105 at req_wait.c:233, queue = 'com.apple.main-thread', stop reason = EXC_BAD_ACCESS (code=1, address=0x1100000088)
  * frame #0: 0x0000000100051bc9 libmpi.0.dylib`ompi_request_default_wait_all(count=3, requests=0x00007fff5fbff170, statuses=0x00007fff5fbff190) + 105 at req_wait.c:233
    frame #1: 0x00000001000dd18c libmpi.0.dylib`MPI_Waitall(count=3, requests=0x00007fff5fbff170, statuses=0x00007fff5fbff190) + 444 at waitall.c:76
    frame #2: 0x000000010000c6a5 filetest`write_shared_test(comm=0x0000000100212a98, root=0) + 2917 at write_shared.c:228
    frame #3: 0x0000000100000a29 filetest`main(argc=1, argv=0x00007fff5fbff278) + 345 at filetest.c:84
    frame #4: 0x00007fff8265b5c9 libdyld.dylib`start + 1
    frame #5: 0x00007fff8265b5c9 libdyld.dylib`start + 1

edgargabriel · 2015-10-01T20:58:45Z

so actually it works for me mostly, just changed the free to my_req_per_proc, correct? Which sharedfp component is selected in your case? sm or lockedfile (I don't have a mac to test unfortunately).

edgargabriel · 2015-10-01T21:02:34Z

the write_shared test passes for both components on linux, so not sure where to start to look into that. can you post a complete traceback?

hjelmn · 2015-10-01T21:14:18Z

Output of rank 0 with -mca sharedfp_base_verbose 100: https://gist.github.com/hjelmn/9ffb5b42299058b89701

edgargabriel · 2015-10-02T16:03:04Z

I looked at the file that you attached, and I think I have an idea on what might be going wrong on your system. But I will need somehow access to a MAC to test it :-( Anyway, it is not related to this PR, it has to do with creating the shared memory region through mmap inside of the sm component. So I think we should be good to go with this PR. I can open a new issue for the sm sharedfp component.

+1

fcoll/two_phase: fix coverity errors

Update the PMIx support to the PMIx 1.1.2 release.

hjelmn force-pushed the ompi_coverity branch from 4a38543 to 63e2bcc Compare October 1, 2015 17:57

hjelmn added 2 commits October 1, 2015 14:38

hjelmn force-pushed the ompi_coverity branch from 63e2bcc to 95b95e1 Compare October 1, 2015 20:38

hjelmn added the bug label Oct 2, 2015

hjelmn added this to the Open MPI v2.0.0 milestone Oct 2, 2015

hjelmn assigned edgargabriel Oct 2, 2015

hjelmn added a commit that referenced this pull request Oct 2, 2015

Merge pull request #963 from hjelmn/ompi_coverity

eb79edf

fcoll/two_phase: fix coverity errors

hjelmn merged commit eb79edf into open-mpi:master Oct 2, 2015

jsquyres added a commit to jsquyres/ompi that referenced this pull request Aug 23, 2016

Merge pull request open-mpi#963 from rhc54/cmr2.x/pmix

980766e

Update the PMIx support to the PMIx 1.1.2 release.

fcoll/two_phase: fix coverity errors #963

fcoll/two_phase: fix coverity errors #963

Uh oh!

Conversation

hjelmn commented Oct 1, 2015

Uh oh!

hjelmn commented Oct 1, 2015

Uh oh!

hjelmn commented Oct 1, 2015

Uh oh!

lanl-ompi commented Oct 1, 2015

Uh oh!

edgargabriel commented Oct 1, 2015

Uh oh!

hjelmn commented Oct 1, 2015

Uh oh!

edgargabriel commented Oct 1, 2015

Uh oh!

edgargabriel commented Oct 1, 2015

[salmon:07724] *** End of error message ***

mpirun noticed that process rank 0 with PID 0 on node salmon exited on signal 11 (Segmentation fault).

Uh oh!

hjelmn commented Oct 1, 2015

Uh oh!

edgargabriel commented Oct 1, 2015

Uh oh!

hjelmn commented Oct 1, 2015

Uh oh!

edgargabriel commented Oct 1, 2015

Uh oh!

hjelmn commented Oct 1, 2015

Uh oh!

edgargabriel commented Oct 1, 2015

Uh oh!

edgargabriel commented Oct 1, 2015

Uh oh!

hjelmn commented Oct 1, 2015

Uh oh!

edgargabriel commented Oct 2, 2015

Uh oh!

Uh oh!

[salmon:07724] * End of error message *