| Message ID | 20220419050900.3136024-1-yhs@fb.com (mailing list archive) |
|---|---|
| State | Accepted |
| Commit | 44df171a10f8969d1456e0a5af7fbac142d7fa18 |
| Delegated to: | BPF |
| Headers | show |
| Series | [bpf-next] selftests/bpf: Workaround a verifier issue for test exhandler | expand |
Hello: This patch was applied to bpf/bpf-next.git (master) by Alexei Starovoitov <ast@kernel.org>: On Mon, 18 Apr 2022 22:09:00 -0700 you wrote: > The llvm patch [1] enabled opaque pointer which caused selftest > 'exhandler' failure. > ... > ; work = task->task_works; > 7: (79) r1 = *(u64 *)(r6 +2120) ; R1_w=ptr_callback_head(off=0,imm=0) R6_w=ptr_task_struct(off=0,imm=0) > ; func = work->func; > 8: (79) r2 = *(u64 *)(r1 +8) ; R1_w=ptr_callback_head(off=0,imm=0) R2_w=scalar() > ; if (!work && !func) > 9: (4f) r1 |= r2 > math between ptr_ pointer and register with unbounded min value is not allowed > > [...] Here is the summary with links: - [bpf-next] selftests/bpf: Workaround a verifier issue for test exhandler https://git.kernel.org/bpf/bpf-next/c/44df171a10f8 You are awesome, thank you!
diff --git a/tools/testing/selftests/bpf/progs/exhandler_kern.c b/tools/testing/selftests/bpf/progs/exhandler_kern.c index f5ca142abf8f..dd9b30a0f0fc 100644 --- a/tools/testing/selftests/bpf/progs/exhandler_kern.c +++ b/tools/testing/selftests/bpf/progs/exhandler_kern.c @@ -7,6 +7,8 @@ #include <bpf/bpf_tracing.h> #include <bpf/bpf_core_read.h> +#define barrier_var(var) asm volatile("" : "=r"(var) : "0"(var)) + char _license[] SEC("license") = "GPL"; unsigned int exception_triggered; @@ -37,7 +39,16 @@ int BPF_PROG(trace_task_newtask, struct task_struct *task, u64 clone_flags) */ work = task->task_works; func = work->func; - if (!work && !func) - exception_triggered++; + /* Currently verifier will fail for `btf_ptr |= btf_ptr` * instruction. + * To workaround the issue, use barrier_var() and rewrite as below to + * prevent compiler from generating verifier-unfriendly code. + */ + barrier_var(work); + if (work) + return 0; + barrier_var(func); + if (func) + return 0; + exception_triggered++; return 0; }
The llvm patch [1] enabled opaque pointer which caused selftest 'exhandler' failure. ... ; work = task->task_works; 7: (79) r1 = *(u64 *)(r6 +2120) ; R1_w=ptr_callback_head(off=0,imm=0) R6_w=ptr_task_struct(off=0,imm=0) ; func = work->func; 8: (79) r2 = *(u64 *)(r1 +8) ; R1_w=ptr_callback_head(off=0,imm=0) R2_w=scalar() ; if (!work && !func) 9: (4f) r1 |= r2 math between ptr_ pointer and register with unbounded min value is not allowed below is insn 10 and 11 10: (55) if r1 != 0 goto +5 11: (18) r1 = 0 ll ... In llvm, the code generation of 'r1 |= r2' happened in codegen selectiondag phase due to difference of opaque pointer vs. non-opaque pointer. Without [1], the related code looks like: r2 = *(u64 *)(r6 + 2120) r1 = *(u64 *)(r2 + 8) if r2 != 0 goto +6 <LBB0_4> if r1 != 0 goto +5 <LBB0_4> r1 = 0 ll ... I haven't found a good way in llvm to fix this issue. So let us workaround the problem first so bpf CI won't be blocked. [1] https://reviews.llvm.org/D123300 Signed-off-by: Yonghong Song <yhs@fb.com> --- .../testing/selftests/bpf/progs/exhandler_kern.c | 15 +++++++++++++-- 1 file changed, 13 insertions(+), 2 deletions(-)