: drop test pidns03 ?

LTP Sukadev Bhattiprolu 2 participants 2 messages Aug 5, 2010 – Aug 16, 2010

lore

Share & tools

lore

Sukadev Bhattiprolu

author

2010-08-05 23:55 UTC #0 lore raw

I am not sure about the original intent of pidns03 test case but I think the test is expected to fail :-(

The header for this test says:

* Verify that:
* 1. When parent, clone a process with flag CLONE_NEWPID, see the
* process id of the parent is existing after mounting /proc

The test passes in the parent-pid from init-pid-ns and looks for that pid in the child's mount of /proc.

But for the CLONE_NEWPID child, parent process should appear to have a pid == 0. The pid of the parent in the initial pid namespace must NOT be visible to the child - in fact a different process can have that pid.

I am not surprised the test fails - I am curious to know the version of the kernel when this test reliably passed - if at all. I tried this on 2.6.29 and 2.6.34-rc5 and it fails reliably.

I dug through my history and see that this test was part of a 6-patch set, We did have some comments on the coding style on the set, but we seem to have missed the intent of this test :-(.

Do we have any other history of this test ? If not, I think we should either drop the test or modify the test to verify the opposite (i.e. verify that the opendir() in the test fails with ENOENT).

Sukadev

------------------------------------------------------------------------------
This SF.net email is sponsored by

Make an app they can't live without
Enter the BlackBerry Developer Challenge
http://p.sf.net/sfu/RIM-dev2dev
_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list

Subrata Modak

2010-08-16 14:41 UTC #1 lore raw

Dropped.

Regards--
Subrata

commit b47b875e00fbea4bb3398ac2f3ef4b4dd32826f0

Author: Subrata Modak <[email protected]>
Date:   Mon Aug 16 20:08:51 2010 +0530

Quoted 33 lines · Sukadev Bhattiprolu

I am not sure about the original intent of pidns03 test case but I think the test is expected to fail :-(
The header for this test says:
* Verify that:
* 1. When parent, clone a process with flag CLONE_NEWPID, see the
* process id of the parent is existing after mounting /proc
The test passes in the parent-pid from init-pid-ns and looks for that pid in the child's mount of /proc.
But for the CLONE_NEWPID child, parent process should appear to have a pid == 0. The pid of the parent in the initial pid namespace must NOT be visible to the child - in fact a different process can have that pid.
I am not surprised the test fails - I am curious to know the version of the kernel when this test reliably passed - if at all. I tried this on 2.6.29 and 2.6.34-rc5 and it fails reliably.
I dug through my history and see that this test was part of a 6-patch set, We did have some comments on the coding style on the set, but we seem to have missed the intent of this test :-(.
Do we have any other history of this test ? If not, I think we should either drop the test or modify the test to verify the opposite (i.e. verify that the opendir() in the test fails with ENOENT).
Sukadev
------------------------------------------------------------------------------
This SF.net email is sponsored by

Quoted 7 lines

Make an app they can't live without
Enter the BlackBerry Developer Challenge
http://p.sf.net/sfu/RIM-dev2dev
_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list

------------------------------------------------------------------------------
This SF.net email is sponsored by