Skip to content

Fix for raw_tp PTR_MAYBE_NULL handling

Kumar Kartikeya Dwivedi bpf 3 versions – In review

Get series with b4
b4 am [email protected]

Versions newest first · linked by author and title

  1. v3 Latest3 patches 11 repliesIn review
  2. v22 patches 2 repliesSuperseded
  3. v12 patches 11 repliesSuperseded

Changes in v3 from the cover letter

v2: https://lore.kernel.org/bpf/[email protected]

* Add Acked-by for Patch 1
* Add fix for scalar dereference issue
* Roll both fixes into one, as second fix undoes first
* Fix nits

Patches in v3 No applies recorded yet

#Subject and filesSizeRepliesApplied
1/3bpf: Suppress warning for non-zero off raw_tp arg NULL checkkernel/bpf/verifier.c+31 -70 replies
2/3bpf: Do not mark NULL-checked raw_tp arg as scalarkernel/bpf/verifier.c+6 -011 replies
3/3selftests/bpf: Add raw_tp tests for PTR_MAYBE_NULL markingtools/testing/selftests/bpf/prog_tests/raw_tp_null.c tools/testing/selftests/bpf/progs/raw_tp_null_fail.c+96 -00 replies

Reviews trailers per patch

1 of 3 patches has a Reviewed-by or Acked-by.

PatchEduard Zingerman
1/3A-b
2/3
3/3

Trailers given in replies are filled; those carried in the patches from earlier versions are outlined with dashes.

Read the v3 thread

Cover letter [PATCH bpf v3 0/3] Fix for raw_tp PTR_MAYBE_NULL handling

A production BPF program had the following code produced by LLVM.

r0 = 1024;
r1 = ...; // r1 = trusted_or_null_(id=1)
r3 = r1; // r3 = trusted_or_null_(id=1) r1 = trusted_or_null_(id=1) r3 += r0; // r3 = trusted_or_null_(id=1, off=1024)
if r1 == 0 goto pc+X;

After cb4158ce8ec8 ("bpf: Mark raw_tp arguments with PTR_MAYBE_NULL"), the production BPF program began throwing a warning in the verifier because for the code above, when unmarking null mark from r1, the verifier will notice another register r3 with same id but off != 0, which is unexpected, since offset modification on PTR_MAYBE_NULL is not permitted, but the aforementioned commit relaxed that restriction to preserve compatibility with non-NULL raw_tp args.

Another production program hit a case where generic code it was calling into would perform a NULL check, while the program knows and is written with the knowledge that the raw_tp arg can never be NULL.

In earlier versions before the raw_tp change, verifier would never walk the path where raw_tp arg was seen as scalar zero, but now it will, hence code in the program that operates on the raw_tp arg later on will fail on dereferencing a scalar.

Provide a fix to suppress the warning for raw_tp args, and not mark NULL checked raw_tp args as scalars. We will follow up with a more generic fix to handle such patterns for all pointer types in the verifier, which currently involves playing whack-a-mole with suppressing such LLVM optimizations and reworking BPF programs to avoid verifier errors.

Changelog:
----------
v2 -> v3
v2: https://lore.kernel.org/bpf/[email protected]

* Add Acked-by for Patch 1
* Add fix for scalar dereference issue
* Roll both fixes into one, as second fix undoes first
* Fix nits

v1 -> v2
v1: https://lore.kernel.org/bpf/[email protected]

* Fix eager unmarking bug (Eduard)
* Generalize approach, always unmark NULL when off == 0 is checked
* Make NULL check noop if operand has off != 0
* Do not reset id when treating as noop
* Trim comment (Alexei)
* Adjust selftests

Kumar Kartikeya Dwivedi (3):
bpf: Suppress warning for non-zero off raw_tp arg NULL check
bpf: Do not mark NULL-checked raw_tp arg as scalar selftests/bpf: Add raw_tp tests for PTR_MAYBE_NULL marking

 kernel/bpf/verifier.c                         | 44 +++++++--
 .../selftests/bpf/prog_tests/raw_tp_null.c    |  6 ++
 .../selftests/bpf/progs/raw_tp_null_fail.c    | 90 +++++++++++++++++++
 3 files changed, 133 insertions(+), 7 deletions(-)
 create mode 100644 tools/testing/selftests/bpf/progs/raw_tp_null_fail.c

base-commit: 5a6ea7022ff4d2a65ae328619c586d6a8909b48b

Signature · 1 line
2.43.5