Skip to content

Atomics for eBPF

Brendan Jackman bpf lkml 8 versions – Applied

Status from this message: This series was applied to bpf/bpf-next.git (refs/heads/master):

Get series with b4
b4 am [email protected]

Versions newest first · linked by author and title

  1. v7 Latest11 patches 1 replyApplied 11/11
  2. v611 patches 2 repliesSuperseded
  3. v511 patches 6 repliesSuperseded
Earlier versions · 5
VersionPostedPatchesRepliesStatus
v4 posting 2/20 of 11 patches0 repliesSuperseded
v4 posting 1/211 patches27 repliesSuperseded
v314 patches31 repliesSuperseded
v213 patches41 repliesSuperseded
v17 patches15 repliesSuperseded

Patches in v7 All 11 applied

#Subject and filesSizeRepliesApplied
1/11bpf: x86: Factor out emission of ModR/M for *(reg + off)arch/x86/net/bpf_jit_comp.c+25 -180 repliesApplied
2/11bpf: x86: Factor out emission of REX bytearch/x86/net/bpf_jit_comp.c+23 -160 repliesApplied
3/11bpf: x86: Factor out a lookup table for some ALU opcodesarch/x86/net/bpf_jit_comp.c+15 -180 repliesApplied
4/11bpf: Rename BPF_XADD and prepare to encode other atomics in .immDocumentation/networking/filter.rst arch/arm/net/bpf_jit_32.c arch/arm64/net/bpf_jit_comp.c +32 more+291 -1520 repliesApplied
5/11bpf: Move BPF_STX reserved field check into BPF_STX verifier codekernel/bpf/verifier.c+6 -70 repliesApplied
6/11bpf: Add BPF_FETCH field / create atomic_fetch_add instructionarch/x86/net/bpf_jit_comp.c include/linux/filter.h include/uapi/linux/bpf.h +5 more+56 -90 repliesApplied
7/11bpf: Add instructions for atomic_[cmp]xchgarch/x86/net/bpf_jit_comp.c include/linux/filter.h include/uapi/linux/bpf.h +5 more+70 -40 repliesApplied
8/11bpf: Pull out a macro for interpreting atomic ALU operationskernel/bpf/core.c+39 -410 repliesApplied
9/11bpf: Add bitwise atomic instructionsarch/x86/net/bpf_jit_comp.c include/linux/filter.h kernel/bpf/core.c +3 more+87 -50 repliesApplied
10/11bpf: Add tests for new BPF atomic operationstools/testing/selftests/bpf/Makefile tools/testing/selftests/bpf/prog_tests/atomics.c tools/testing/selftests/bpf/progs/atomics.c +6 more+881 -00 repliesApplied
11/11bpf: Document new atomic instructionsDocumentation/networking/filter.rst+31 -00 repliesApplied

Reviews trailers per patch

All 11 patches have a Reviewed-by or Acked-by.

PatchJohn FastabendYonghong SongBjörn Töpel
1/11A-b
2/11A-b
3/11A-b
4/11A-b
5/11A-b A-b
6/11A-b
7/11A-b
8/11A-b
9/11A-b
10/11A-b
11/11A-b

Trailers given in replies are filled; those carried in the patches from earlier versions are outlined with dashes. Scroll the table sideways for all 3 reviewers.

Read the v7 thread

Cover letter [PATCH bpf-next v7 00/11] Atomics for eBPF

There's still one unresolved review comment from John[3] which I will resolve with a followup patch.

Differences from v6->v7 [1]:

* Fixed riscv build error detected by 0-day robot.

Differences from v5->v6 [1]:

* Carried Björn Töpel's ack for RISC-V code, plus a couple more acks from Yonhgong.

* Doc fixups.

* Trivial cleanups.

Differences from v4->v5 [1]:

* Fixed bogus type casts in interpreter that led to warnings from the 0day robot.

* Dropped feature-detection for Clang per Andrii's suggestion in [4]. The selftests will now fail to build unless you have llvm-project commit 286daafd6512. The ENABLE_ATOMICS_TEST macro is still needed to support the no_alu32 tests.

* Carried some Acks from John and Yonghong.

* Dropped confusing usage of __atomic_exchange from prog_test in favour of __sync_lock_test_and_set.

* [Really] got rid of all the forest of instruction macros (BPF_ATOMIC_FETCH_ADD and friends); now there's just BPF_ATOMIC_OP to define all the instructions as we use them in the verifier tests. This makes the atomic ops less special in that API, and I don't think the resulting usage is actually any harder to read.

Differences from v3->v4 [1]:

* Added one Ack from Yonghong. He acked some other patches but those have now changed non-trivally so I didn't add those acks.

* Fixups to commit messages.

* Fixed disassembly and comments: first arg to atomic_fetch_* is a pointer.

* Improved prog_test efficiency. BPF progs are now all loaded in a single call, then the skeleton is re-used for each subtest.

* Dropped use of tools/build/feature in favour of a one-liner in the Makefile.

* Dropped the commit that created an emit_neg helper in the x86 JIT. It's not used any more (it wasn't used in v3 either).

* Combined all the different filter.h macros (used to be BPF_ATOMIC_ADD, BPF_ATOMIC_FETCH_ADD, BPF_ATOMIC_AND, etc) into just BPF_ATOMIC32 and BPF_ATOMIC64.

* Removed some references to BPF_STX_XADD from tools/, samples/ and lib/ that I missed before.

Differences from v2->v3 [1]:

* More minor fixes and naming/comment changes

* Dropped atomic subtract: compilers can implement this by preceding an atomic add with a NEG instruction (which is what the x86 JIT did under the hood anyway).

* Dropped the use of -mcpu=v4 in the Clang BPF command-line; there is no longer an architecture version bump. Instead a feature test is added to Kbuild - it builds a source file to check if Clang supports BPF atomics.

* Fixed the prog_test so it no longer breaks test_progs-no_alu32. This requires some ifdef acrobatics to avoid complicating the prog_tests model where the same userspace code exercises both the normal and no_alu32 BPF test objects, using the same skeleton header.

Differences from v1->v2 [1]:

* Fixed mistakes in the netronome driver

* Addd sub, add, or, xor operations

* The above led to some refactors to keep things readable. (Maybe I should have just waited until I'd implemented these before starting the review...)

* Replaced BPF_[CMP]SET | BPF_FETCH with just BPF_[CMP]XCHG, which include the BPF_FETCH flag

* Added a bit of documentation. Suggestions welcome for more places to dump this info...

The prog_test that's added depends on Clang/LLVM features added by Yonghong in commit 286daafd6512 (was
https://reviews.llvm.org/D72184).

This only includes a JIT implementation for x86_64 - I don't plan to implement JIT support myself for other architectures.

Operations
==========

This patchset adds atomic operations to the eBPF instruction set. The use-case that motivated this work was a trivial and efficient way to generate globally-unique cookies in BPF progs, but I think it's obvious that these features are pretty widely applicable. The instructions that are added here can be summarised with this list of kernel operations:

* atomic[64]_[fetch_]add
* atomic[64]_[fetch_]and
* atomic[64]_[fetch_]or
* atomic[64]_xchg
* atomic[64]_cmpxchg

The following are left out of scope for this effort:

* 16 and 8 bit operations
* Explicit memory barriers

Encoding
========

I originally planned to add new values for bpf_insn.opcode. This was rather unpleasant: the opcode space has holes in it but no entire instruction classes[2]. Yonghong Song had a better idea: use the immediate field of the existing STX XADD instruction to encode the operation. This works nicely, without breaking existing programs, because the immediate field is currently reserved-must-be-zero, and extra-nicely because BPF_ADD happens to be zero.

Note that this of course makes immediate-source atomic operations impossible. It's hard to imagine a measurable speedup from such instructions, and if it existed it would certainly not benefit x86, which has no support for them.

The BPF_OP opcode fields are re-used in the immediate, and an additional flag BPF_FETCH is used to mark instructions that should fetch a pre-modification value from memory.

So, BPF_XADD is now called BPF_ATOMIC (the old name is kept to avoid breaking userspace builds), and where we previously had .imm = 0, we now have .imm = BPF_ADD (which is 0).

Operands
========

Reg-source eBPF instructions only have two operands, while these atomic operations have up to four. To avoid needing to encode additional operands, then:

- One of the input registers is re-used as an output register (e.g. atomic_fetch_add both reads from and writes to the source register).

- Where necessary (i.e. for cmpxchg) , R0 is "hard-coded" as one of the operands.

This approach also allows the new eBPF instructions to map directly to single x86 instructions.

[1] Previous iterations:

    v1: https://lore.kernel.org/bpf/[email protected]/
    v2: https://lore.kernel.org/bpf/[email protected]/
    v3: https://lore.kernel.org/bpf/[email protected]/
    v4: https://lore.kernel.org/bpf/[email protected]/
    v5: https://lore.kernel.org/bpf/[email protected]/
    v6: https://lore.kernel.org/bpf/[email protected]/

[2] Visualisation of eBPF opcode space:

    https://gist.github.com/bjackman/00fdad2d5dfff601c1918bc29b16e778

[3] Comment from John about propagating bounds in verifier:

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

[4] Mail from Andrii about not supporting old Clang in selftests:
https://lore.kernel.org/bpf/CAEf4BzYBddPaEzRUs=jaWSo5kbf=LZdb7geAUVj85GxLQztuAQ@mail.gmail.com/

Brendan Jackman (11):
bpf: x86: Factor out emission of ModR/M for *(reg + off)
bpf: x86: Factor out emission of REX byte
bpf: x86: Factor out a lookup table for some ALU opcodes
bpf: Rename BPF_XADD and prepare to encode other atomics in .imm
bpf: Move BPF_STX reserved field check into BPF_STX verifier code
bpf: Add BPF_FETCH field / create atomic_fetch_add instruction
bpf: Add instructions for atomic_[cmp]xchg
bpf: Pull out a macro for interpreting atomic ALU operations
bpf: Add bitwise atomic instructions
bpf: Add tests for new BPF atomic operations
bpf: Document new atomic instructions

 Documentation/networking/filter.rst           |  61 ++++-
 arch/arm/net/bpf_jit_32.c                     |   7 +-
 arch/arm64/net/bpf_jit_comp.c                 |  16 +-
 arch/mips/net/ebpf_jit.c                      |  11 +-
 arch/powerpc/net/bpf_jit_comp64.c             |  25 +-
 arch/riscv/net/bpf_jit_comp32.c               |  20 +-
 arch/riscv/net/bpf_jit_comp64.c               |  16 +-
 arch/s390/net/bpf_jit_comp.c                  |  27 +-
 arch/sparc/net/bpf_jit_comp_64.c              |  17 +-
 arch/x86/net/bpf_jit_comp.c                   | 217 ++++++++++-----
 arch/x86/net/bpf_jit_comp32.c                 |   6 +-
 drivers/net/ethernet/netronome/nfp/bpf/jit.c  |  14 +-
 drivers/net/ethernet/netronome/nfp/bpf/main.h |   4 +-
 .../net/ethernet/netronome/nfp/bpf/verifier.c |  15 +-
 include/linux/filter.h                        |  25 +-
 include/uapi/linux/bpf.h                      |  10 +-
 kernel/bpf/core.c                             |  67 ++++-
 kernel/bpf/disasm.c                           |  43 ++-
 kernel/bpf/verifier.c                         |  75 ++++--
 lib/test_bpf.c                                |  14 +-
 samples/bpf/bpf_insn.h                        |   4 +-
 samples/bpf/cookie_uid_helper_example.c       |   8 +-
 samples/bpf/sock_example.c                    |   2 +-
 samples/bpf/test_cgrp2_attach.c               |   5 +-
 tools/include/linux/filter.h                  |  24 +-
 tools/include/uapi/linux/bpf.h                |  10 +-
 tools/testing/selftests/bpf/Makefile          |   2 +
 .../selftests/bpf/prog_tests/atomics.c        | 246 ++++++++++++++++++
 .../bpf/prog_tests/cgroup_attach_multi.c      |   4 +-
 tools/testing/selftests/bpf/progs/atomics.c   | 154 +++++++++++
 .../selftests/bpf/test_cgroup_storage.c       |   2 +-
 .../selftests/bpf/verifier/atomic_and.c       |  77 ++++++
 .../selftests/bpf/verifier/atomic_cmpxchg.c   |  96 +++++++
 .../selftests/bpf/verifier/atomic_fetch_add.c | 106 ++++++++
 .../selftests/bpf/verifier/atomic_or.c        |  77 ++++++
 .../selftests/bpf/verifier/atomic_xchg.c      |  46 ++++
 .../selftests/bpf/verifier/atomic_xor.c       |  77 ++++++
 tools/testing/selftests/bpf/verifier/ctx.c    |   7 +-
 .../bpf/verifier/direct_packet_access.c       |   4 +-
 .../testing/selftests/bpf/verifier/leak_ptr.c |  10 +-
 .../selftests/bpf/verifier/meta_access.c      |   4 +-
 tools/testing/selftests/bpf/verifier/unpriv.c |   3 +-
 .../bpf/verifier/value_illegal_alu.c          |   2 +-
 tools/testing/selftests/bpf/verifier/xadd.c   |  18 +-
 44 files changed, 1466 insertions(+), 212 deletions(-)
 create mode 100644 tools/testing/selftests/bpf/prog_tests/atomics.c
 create mode 100644 tools/testing/selftests/bpf/progs/atomics.c
 create mode 100644 tools/testing/selftests/bpf/verifier/atomic_and.c
 create mode 100644 tools/testing/selftests/bpf/verifier/atomic_cmpxchg.c
 create mode 100644 tools/testing/selftests/bpf/verifier/atomic_fetch_add.c
 create mode 100644 tools/testing/selftests/bpf/verifier/atomic_or.c
 create mode 100644 tools/testing/selftests/bpf/verifier/atomic_xchg.c
 create mode 100644 tools/testing/selftests/bpf/verifier/atomic_xor.c

base-commit: bade5c554f1ac70a50cefe96517957629dbc0d8f --
2.30.0.284.gd98b1dd5eaa7-goog