Details

Reviewers

asomers
gnn
rwatson

Commits

rS336418: audit(4): add tests for _exit(2), cap_enter(2), and cap_getmode(2)

Summary

This revision introduces atf-c(3) tests for 2 Capability mode syscalls and _exit(2).

cap_enter(2)
cap_getmode(2)
_exit(2)

Test Plan

Execute make && make install from test/sys/audit.
Execute kyua test from /usr/tests/sys/audit. All testcases should succeed.

Diff Detail

Lint

Lint Passed

Unit

No Test Coverage

Build Status

Buildable 17792
Build 17570: arc lint + arc unit

Event Timeline

aniketp created this revision.Jul 2 2018, 5:41 PM

Herald added a subscriber: jonathan. · View Herald TranscriptJul 2 2018, 5:41 PM

Harbormaster completed remote builds in B17792: Diff 44764.Jul 2 2018, 5:41 PM

For me cap_enter_success passes intermittently. On a case when it failed, the global audit trail showed the cap_enter call we were looking for (as well as the child's exit(2), which is also in class "pc"). But the auditpipe showed nothing after fork. I think we're looking at a buffering issue. When I run "./process_control cap_enter_success" the last thing I see is the fork record. Then the process pauses for 10 seconds. But while it's paused, if I run any command at all in another terminal, then the test immediately passes. So I think that the auditpipe(4) device is buffering up some amount of data before its read(2) returns.

Add the AUDITPIPE_FLUSH ioctl after calling cap_enter() to eliminate the buffering issue

Harbormaster completed remote builds in B17948: Diff 45103.Jul 10 2018, 8:47 AM

Add tests for _exit(2) using the AUDITPIPE_FLUSH approach (for now)

Harbormaster completed remote builds in B17970: Diff 45145.Jul 11 2018, 7:32 AM

aniketp retitled this revision from Introduce tests for Capability mode System calls to Introduce tests for Capability mode System calls and _exit(2).Jul 11 2018, 7:33 AM

aniketp edited the summary of this revision. (Show Details)

Updates in _exit(2) test case comment

Harbormaster completed remote builds in B17971: Diff 45146.Jul 11 2018, 7:57 AM

aniketp mentioned this in D16210: Don't acquire evclass_lock with a spinlock held.Jul 11 2018, 8:07 AM

using AUDITPIPE_FLUSH doesn't work. You should remove that part and fix the bug in utils.c instead.

This revision now requires changes to proceed.Jul 11 2018, 2:58 PM

Disable I/O buffering from /dev/auditpipe for read operations for cap_enter(2) and _exit(2)

Harbormaster completed remote builds in B18082: Diff 45389.Jul 16 2018, 8:31 PM

Forgot a comma in _exit(2) test case comments

Harbormaster completed remote builds in B18083: Diff 45390.Jul 16 2018, 8:33 PM

Some more updates to the recent comments

Harbormaster completed remote builds in B18084: Diff 45391.Jul 16 2018, 8:35 PM

There's nothing special about exit or cap_enter. You need to disable buffering everywhere, anytime you combine fread(3) with something like ppoll(2).

This revision now requires changes to proceed.Jul 16 2018, 9:55 PM

Disable I/O buffering for the whole test-suite

Harbormaster completed remote builds in B18089: Diff 45397.Jul 16 2018, 10:44 PM

Remove redundant header includes that were added for ioctl operations which are not required anymore

Harbormaster completed remote builds in B18090: Diff 45398.Jul 16 2018, 10:47 PM

asomers added inline comments.Jul 16 2018, 10:58 PM

tests/sys/audit/utils.c
216 ↗	(On Diff #45398)	Technically it's not possible, not since you use `setvbuf`. I would say something more like "disable stream buffering so fread(3) (called via au_read_rec(3)) doesn't buffer data unbeknownst to ppoll(2)"