Re: [PATCH v3 0/4] LoongArch bpf kptr xchg inline support

On Mon, Jun 22, 2026 at 3:18=E2=80=AFPM Chenguang Zhao <zhaochenguang@kylin=
os.cn> wrote:
>
> This series enables the BPF verifier to inline bpf_kptr_xchg() into BPF_X=
CHG
> on LoongArch64, and fixes the underlying JIT atomic ordering that makes s=
uch
> inlining safe.
> The BPF verifier can lower bpf_kptr_xchg() to a single BPF_XCHG atomic wh=
en
> the JIT advertises ptr xchg support via bpf_jit_supports_ptr_xchg(). This
> removes helper-call overhead from the kptr exchange fast path. Inlining i=
s
> only correct when the JITed exchange provides the same sequentially consi=
stent
> ordering as the bpf_kptr_xchg() helper.
> Patch 1 fixes memory ordering for all value-returning BPF atomic RMW oper=
ations
> emitted by the LoongArch bpf JIT. Per LKMM, value-returning atomic RMW mu=
st
> provide sequentially consistent ordering. Plain AMO instructions and bare
> ll/sc loops on LoongArch do not satisfy this by themselves:
>  - BPF_FETCH (ADD/AND/OR/XOR): switch to am*_db.{b,h,w,d}
>  - BPF_XCHG: switch to amswap_db.{b,h,w,d}
>  - BPF_CMPXCHG: emit dbar 0x700 after the ll/sc loop, matching
>    __WEAK_LLSC_MB in cmpxchg.h
> Non-value-returning RMW ops (plain BPF_ADD, BPF_AND, etc.) remain weakly
> ordered, consistent with LKMM. This fix is independent of kptr inlining a=
nd
> benefits all BPF programs using value-returning atomics on LoongArch.
> Patch 2 implements bpf_jit_supports_ptr_xchg() so the verifier may inline
> bpf_kptr_xchg() on LoongArch64.
> Patches 3 and 4 extend bpf selftests: functional coverage via kptr_xchg_i=
nline,
> and an optional kptr-xchg benchmark to compare helper vs inlined paths.
>
> Chenguang Zhao (4):
>   LoongArch: bpf: Fix memory ordering for value-returning atomics
>   LoongArch: bpf: Advertise JIT support for kptr xchg inline
>   selftests/bpf: Enable kptr_xchg_inline test on LoongArch
>   selftests/bpf: Add kptr-xchg benchmark
>

Hello, sorry for the long delay. Please rebase and resend this series. Than=
ks.

>  arch/loongarch/include/asm/inst.h             | 18 ++++
>  arch/loongarch/net/bpf_jit.c                  | 37 ++++---
>  tools/testing/selftests/bpf/Makefile          |  2 +
>  tools/testing/selftests/bpf/bench.c           |  2 +
>  .../selftests/bpf/benchs/bench_kptr_xchg.c    | 96 +++++++++++++++++++
>  .../bpf/prog_tests/kptr_xchg_inline.c         |  3 +-
>  .../selftests/bpf/progs/kptr_xchg_bench.c     | 49 ++++++++++
>  7 files changed, 192 insertions(+), 15 deletions(-)
>  create mode 100644 tools/testing/selftests/bpf/benchs/bench_kptr_xchg.c
>  create mode 100644 tools/testing/selftests/bpf/progs/kptr_xchg_bench.c
>
> ---
> v3:
>  - Expand the JIT ordering fix beyond BPF_XCHG w/d to all value-returning
>    atomic RMW ops (BPF_FETCH, BPF_XCHG, BPF_CMPXCHG), using barrier-carry=
ing
>    AMO variants and dbar after ll/sc as required by LKMM.
>  - Add __btf_root() in the benchmark BPF program so bpf_obj_drop() remain=
s
>    visible to libbpf's kfunc linker when bpf_kptr_xchg() is inlined.
>
> v2:
>  - https://lore.kernel.org/all/20260603100438.2177817-1-zhaochenguang@kyl=
inos.cn/
>
> v1:
>  - https://lore.kernel.org/all/20260602021515.214560-1-zhaochenguang@kyli=
nos.cn/
>
> --
> 2.25.1
>
>