[LU-4746] Lustre not using kernel current_umask() function breaks GRSecurity umask handling

• Status: Resolved
• Project: Lustre
• Affects Version/s: Lustre 2.1.5, Lustre 1.8.9, Lustre 2.5.0, Lustre 2.4.2
• Fix Version/s: Lustre 2.7.0
• Type: Bug
• Priority: Minor
• Reporter: Andrew Prout
• Assignee: Bob Glossman (Inactive)
• Resolution: Fixed
• Labels: patch
• Severity: 3

Description

Lustre 2.5.0 (and before) do not use the kernel's default current_umask() function, instead accessing the current process pointer directly with current->fs->umask. On a standard kernel this is equivalent, however on a GRSecurity-enabled kernel there's additional logic in the kernel's current_umask() function that's being skipped.

Comments

[Andrew Prout]

Patches to fix attached. All affected versions compile normally when calling the Linux kernel's current_umask() function.

[Andreas Dilger]

Andrew,
thanks for the patches. Like the kernel, we need a Signed-off-by: line in the patch in order to accept it. Please see https://wiki.hpdd.intel.com/display/PUB/Using+Gerrit for details on how best to submit patches for Lustre.

[Andrew Prout]

Try #2 on the patches...

[Cliff White (Inactive)]

Can you submit the patches to Gerrit for our review process?

[Cliff White (Inactive)]

I will push these into Gerritt

[Cliff White (Inactive)]

Patches pushed, in the autotest process

[Cliff White (Inactive)]

http://review.whamcloud.com/#/c/10320/
http://review.whamcloud.com/#/c/10321/
http://review.whamcloud.com/#/c/10322/
http://review.whamcloud.com/#/c/10323/

[Oleg Drokin]

I take it this issue is also present in current master version (upcoming 2.6)?
Can we get a patch for master branch too? Thanks.
(2.5 and 2.4 versions are not supposed to be landed without a corresponding master version as tehy are technically supposed to be backports).

[James A Simmons]

Seems I never got around to cleaning that up in LU-3963. Also we need more fixes in a few other places as well in all the branches. Also is their a wrapper to change the umask value?

[James A Simmons]

Created the patch for master at

http://review.whamcloud.com/#/c/11642

Need to do it anyways for LU-3963. Once landed we can back port to the branches of interest. Even with this patch we should look to cleanup sec_ctx.c. I really don't think we need to keep push_ctx and pop_ctx around anymore.

[James A Simmons]

Patch landed for master. Patches for b2_5 and b2_4 still outstanding.

[Jodi Levi (Inactive)]

Patch landed to Master. Other version patch landings will be tracked external to this ticket.