Skip to content

RFC refactor mseal_test

[email protected] linux-hardening linux-kselftest linux-mm lkml 1 version In review

Get series with b4
b4 am [email protected]

Versions newest first

  1. v1 RFC Latest1 patch 4 repliesIn review

Patches in v1 No applies recorded yet

#Subject and filesSizeRepliesApplied
1/1selftest/mm: refactor mseal_testtools/testing/selftests/mm/.gitignore tools/testing/selftests/mm/Makefile tools/testing/selftests/mm/memorysealing.c +2 more+301 -664 replies

Reviews trailers per patch

No Reviewed-by, Acked-by or Tested-by trailers for v1 yet.

Read the v1 thread

Cover letter [RFC PATCH v1 0/1] refactor mseal_test

From: Jeff Xu <[email protected]>

This change creates the initial version of memorysealing.c.

The introduction of memorysealing.c, which replaces mseal_test.c and uses the kselftest_harness, aims to initiate a discussion on using the selftest harness for memory sealing tests. Upon approval of this approach, the migration of tests from mseal_test.c to memorysealing.c can be implemented in a step-by-step manner.

This tests addresses following feedbacks from previous reviews:

1> Use kselftest_harness instead of custom macro, such as EXPECT_XX, ASSERT_XX, etc. (Lorenzo Stoakes, Mark Brown, etc) [1]

2> Use MAP_FAILED to check the return of mmap (Lorenzo Stoakes).

3> Adding a check for vma size and prot bits. The discussion for
this can be found in [2] [3], here is a brief summary:
This is to follow up on Pedro’s in-loop change (from can_modify_mm to can_modify_vma). When mseal_test is initially created, they have a common pattern: setup memory layout, seal the memory, perform a few mm-api steps, verify return code (not zero). Because of the nature of out-of-loop, it is sufficient to just verify the error code in a few cases.

With Pedro's in-loop change, the sealing check happens later in the stack, thus there are more things and scenarios to verify. And there were feedbacks to me that mseal_test should be extensive enough to discover all regressions. Hence I'm adding check for vma size and prot bits.

In this change: we created two fixtures:

Fixture basic: This creates a single VMA, the VMA has a
PROT_NONE page at each end to prevent auto-merging.

Fixture wo_vma: Two VMAs back to end, a PROT_NONE page at each
end to prevent auto-merging.

In addition, I add one test (mprotec) in each fixture for discussion.

[1] https://lore.kernel.org/all/[email protected]/
[2] https://lore.kernel.org/all/CABi2SkUgDZtJtRJe+J9UNdtZn=EQzZcbMB685P=1rR7DUhg=6Q@mail.gmail.com/
[3] https://lore.kernel.org/all/2qywbjb5ebtgwkh354w3lj3vhaothvubjokxq5fhyri5jeeton@duqngzo3swjz/

Jeff Xu (1):
selftest/mm: refactor mseal_test

 tools/testing/selftests/mm/.gitignore      |   1 +
 tools/testing/selftests/mm/Makefile        |   1 +
 tools/testing/selftests/mm/memorysealing.c | 182 +++++++++++++++++++++
 tools/testing/selftests/mm/memorysealing.h | 116 +++++++++++++
 tools/testing/selftests/mm/mseal_test.c    |  67 +-------
 5 files changed, 301 insertions(+), 66 deletions(-)
 create mode 100644 tools/testing/selftests/mm/memorysealing.c
 create mode 100644 tools/testing/selftests/mm/memorysealing.h
Signature · 1 line
2.47.1.613.gc27f4b7a9f-goog