mtest05 fail on read only remote machine

[ngh1hc]

2016-08-26T11:37:31 Test: mtest05 Testparm: mmstress
mmstress 0 TINFO : run mmstress -h for all options
mmstress 0 TINFO : test1: Test case tests the race condition between simultaneous read faults in the same address space.
mmstress 1 TFAIL : mmstress.c:927: TEST 1 Failed
mmstress 0 TINFO : test2: Test case tests the race condition between simultaneous write faults in the same address space.
map_and_thread(): open(): Read-only file system
map_and_thread(): open(): Read-only file system
......
I run tests on read-only remote machine

Looks like the test tries to create new file in test binary folder..
map_and_thread("./tmp.file.1", thread_fault, READ_FAULT, NUMTHREAD, &retval);

[metan-ucw]

Indeed the test opens files in $CWD it should be fixed to honour $TMPDIR and create a temporary directory there before it starts testing.

[bssrikanth]

trying to recreate this issue locally.. I am having CWD which is read-only.. but able to run tests successfully
/tmp/imgmnt# touch d
touch: cannot touch 'd': Read-only file system <-- CWD is readonly

/tmp/imgmnt# /opt/ltp/runltp -s mtest05
mmstress 0 TINFO : run mmstress -h for all options
mmstress 0 TINFO : test1: Test case tests the race condition between simultaneous read faults in the same address space.
mmstress 1 TPASS : TEST 1 Passed
mmstress 0 TINFO : test2: Test case tests the race condition between simultaneous write faults in the same address space.
mmstress 2 TPASS : TEST 2 Passed
mmstress 0 TINFO : test3: Test case tests the race condition between simultaneous COW faults in the same address space.
mmstress 3 TPASS : TEST 3 Passed
mmstress 0 TINFO : test4: Test case tests the race condition between simultaneous READ faults in the same address space. The file mapped is /dev/zero
mmstress 4 TPASS : TEST 4 Passed
mmstress 0 TINFO : test5: Test case tests the race condition between simultaneous fork - exit faults in the same address space.
mmstress 5 TPASS : TEST 5 Passed
mmstress 0 TINFO : test6: Test case tests the race condition between simultaneous fork -exec - exit faults in the same address space.
mmstress 6 TPASS : TEST 6 Passed
mmstress 7 TPASS : Test Passed
<<<execution_status>>>
initiation_status="ok"
duration=1 termination_type=exited termination_id=0 corefile=no
cutime=89 cstime=882
<<<test_end>>>
INFO: ltp-pan reported all tests PASS
LTP Version: 20160920-140-g65767d1

   ###############################################################

        Done executing testcases.
        LTP Version:  20160920-140-g65767d1
   ###############################################################

[metan-ucw]

That is because the runltp script changes working directory to the /opt/ltp/. But you should be able to run the test binary directly in the read only directory, i.e. /tmp/imgmnt# /opt/ltp/testcases/bin/mtest05 which should then fail.

[athira-rajeev]

I was able to reproduce this issue . Created a file system and mounted as read only.
Executed mmstress testcase.

mmstress 0 TINFO : run /root/athira/LTP/ltp/testcases/kernel/mem/mtest05/mmstress -h for all options
mmstress 0 TINFO : test1: Test case tests the race condition between simultaneous read faults in the same address space.
map_and_thread(): open(): Read-only file system
mmstress 1 TFAIL : mmstress.c:927: TEST 1 Failed
mmstress 0 TINFO : test2: Test case tests the race condition between simultaneous write faults in the same address space.
map_and_thread(): open(): Read-only file system
mmstress 2 TFAIL : mmstress.c:927: TEST 2 Failed
mmstress 0 TINFO : test3: Test case tests the race condition between simultaneous COW faults in the same address space.
map_and_thread(): open(): Read-only file system

Created patch for this testcase where:

• Create a directory under tmp using tst_tmpdir()
• Instead of creating file with "./tmp.file.1", create under tmpdir

Will post the patch

[ghost]

Not sure when $CWD is updated (trying to export it to /tmp and run binary in another read-only directory is not help), but the testcase can run normally in /tmp
/tmp# /opt/ltp/testcases/bin/mmstress
mmstress 0 TINFO : run /opt/ltp/testcases/bin/mmstress -h for all options
mmstress 0 TINFO : test1: Test case tests the race condition between simultaneous read faults in the same address space.
mmstress 1 TPASS : TEST 1 Passed
mmstress 0 TINFO : test2: Test case tests the race condition between simultaneous write faults in the same address space.
mmstress 2 TPASS : TEST 2 Passed
mmstress 0 TINFO : test3: Test case tests the race condition between simultaneous COW faults in the same address space.
mmstress 3 TPASS : TEST 3 Passed
mmstress 0 TINFO : test4: Test case tests the race condition between simultaneous READ faults in the same address space. The file mapped is /dev/zero
mmstress 4 TPASS : TEST 4 Passed
mmstress 0 TINFO : test5: Test case tests the race condition between simultaneous fork - exit faults in the same address space.
mmstress 5 TPASS : TEST 5 Passed
mmstress 0 TINFO : test6: Test case tests the race condition between simultaneous fork -exec - exit faults in the same address space.
test6(): execvp(): No such file or directory
test6(): execvp(): No such file or directory

(test6 is failed just because "mmstress_dummy" is not in /tmp/)
So using tst_tmpdir() as @athira-rajeev recommended may help to fix all current issues
Waiting for your new patch :-)

[ghost]

Hello @athira-rajeev

Created patch for this testcase where:

Create a directory under tmp using tst_tmpdir()
Instead of creating file with "./tmp.file.1", create under tmpdir

Will post the patch

I also failed on this testcase, see that you already created the patch, did you post it ? (so that we'll not duplicate the work.)