* [PATCH bpf-next] selftests/bpf: Fix the invalid operand for instruction issue
@ 2025-08-27 3:15 Feng Yang
2025-08-27 4:03 ` Alexei Starovoitov
2025-08-27 4:48 ` Eduard Zingerman
0 siblings, 2 replies; 5+ messages in thread
From: Feng Yang @ 2025-08-27 3:15 UTC (permalink / raw)
To: ast, daniel, john.fastabend, andrii, martin.lau, eddyz87, song,
yonghong.song, kpsingh, sdf, haoluo, jolsa
Cc: bpf, linux-kernel, linux-kselftest, llvm
From: Feng Yang <yangfeng@kylinos.cn>
The following issue occurs when compiling with clang version 17.0.6,
but not with version 18.1.8. Add a version restriction to fix this problem.
progs/compute_live_registers.c:251:3: error: invalid operand for instruction
251 | "r0 = 1;"
| ^
<inline asm>:1:22: note: instantiated into assembly here
1 | r0 = 1;r2 = 2;if r1 & 0x7 goto +1;exit;r0 = r2;exit;
| ^
1 error generated.
Fixes: 4a4b84ba9e453 ("selftests/bpf: verify jset handling in CFG computation")
Signed-off-by: Feng Yang <yangfeng@kylinos.cn>
---
tools/testing/selftests/bpf/progs/compute_live_registers.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/tools/testing/selftests/bpf/progs/compute_live_registers.c b/tools/testing/selftests/bpf/progs/compute_live_registers.c
index 6884ab99a421..56aec43f206f 100644
--- a/tools/testing/selftests/bpf/progs/compute_live_registers.c
+++ b/tools/testing/selftests/bpf/progs/compute_live_registers.c
@@ -240,6 +240,7 @@ __naked void if2(void)
::: __clobber_all);
}
+#if __clang_major__ >= 18
/* Verifier misses that r2 is alive if jset is not handled properly */
SEC("socket")
__log_level(2)
@@ -255,6 +256,7 @@ __naked void if3_jset_bug(void)
"exit;"
::: __clobber_all);
}
+#endif
SEC("socket")
__log_level(2)
--
2.43.0
^ permalink raw reply related [flat|nested] 5+ messages in thread
* Re: [PATCH bpf-next] selftests/bpf: Fix the invalid operand for instruction issue
2025-08-27 3:15 [PATCH bpf-next] selftests/bpf: Fix the invalid operand for instruction issue Feng Yang
@ 2025-08-27 4:03 ` Alexei Starovoitov
2025-08-27 4:48 ` Eduard Zingerman
1 sibling, 0 replies; 5+ messages in thread
From: Alexei Starovoitov @ 2025-08-27 4:03 UTC (permalink / raw)
To: Feng Yang
Cc: Alexei Starovoitov, Daniel Borkmann, John Fastabend,
Andrii Nakryiko, Martin KaFai Lau, Eduard, Song Liu,
Yonghong Song, KP Singh, Stanislav Fomichev, Hao Luo, Jiri Olsa,
bpf, LKML, open list:KERNEL SELFTEST FRAMEWORK, clang-built-linux
On Tue, Aug 26, 2025 at 8:16 PM Feng Yang <yangfeng59949@163.com> wrote:
>
> From: Feng Yang <yangfeng@kylinos.cn>
>
> The following issue occurs when compiling with clang version 17.0.6,
> but not with version 18.1.8. Add a version restriction to fix this problem.
>
> progs/compute_live_registers.c:251:3: error: invalid operand for instruction
> 251 | "r0 = 1;"
> | ^
> <inline asm>:1:22: note: instantiated into assembly here
> 1 | r0 = 1;r2 = 2;if r1 & 0x7 goto +1;exit;r0 = r2;exit;
> | ^
> 1 error generated.
and once it compiles what happens next ?
pw-bot: cr
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH bpf-next] selftests/bpf: Fix the invalid operand for instruction issue
2025-08-27 3:15 [PATCH bpf-next] selftests/bpf: Fix the invalid operand for instruction issue Feng Yang
2025-08-27 4:03 ` Alexei Starovoitov
@ 2025-08-27 4:48 ` Eduard Zingerman
2025-08-27 8:24 ` Feng Yang
1 sibling, 1 reply; 5+ messages in thread
From: Eduard Zingerman @ 2025-08-27 4:48 UTC (permalink / raw)
To: Feng Yang, ast, daniel, john.fastabend, andrii, martin.lau, song,
yonghong.song, kpsingh, sdf, haoluo, jolsa
Cc: bpf, linux-kernel, linux-kselftest, llvm
On Wed, 2025-08-27 at 11:15 +0800, Feng Yang wrote:
> From: Feng Yang <yangfeng@kylinos.cn>
>
> The following issue occurs when compiling with clang version 17.0.6,
> but not with version 18.1.8. Add a version restriction to fix this problem.
>
> progs/compute_live_registers.c:251:3: error: invalid operand for instruction
> 251 | "r0 = 1;"
> | ^
> <inline asm>:1:22: note: instantiated into assembly here
> 1 | r0 = 1;r2 = 2;if r1 & 0x7 goto +1;exit;r0 = r2;exit;
> | ^
> 1 error generated.
>
> Fixes: 4a4b84ba9e453 ("selftests/bpf: verify jset handling in CFG computation")
> Signed-off-by: Feng Yang <yangfeng@kylinos.cn>
> ---
> tools/testing/selftests/bpf/progs/compute_live_registers.c | 2 ++
> 1 file changed, 2 insertions(+)
>
> diff --git a/tools/testing/selftests/bpf/progs/compute_live_registers.c b/tools/testing/selftests/bpf/progs/compute_live_registers.c
> index 6884ab99a421..56aec43f206f 100644
> --- a/tools/testing/selftests/bpf/progs/compute_live_registers.c
> +++ b/tools/testing/selftests/bpf/progs/compute_live_registers.c
> @@ -240,6 +240,7 @@ __naked void if2(void)
> ::: __clobber_all);
> }
>
> +#if __clang_major__ >= 18
Instead of guarding this with compiler version, could you please use
progs/bpf_misc.h:__imm_insn() macro for the jset instruction?
> /* Verifier misses that r2 is alive if jset is not handled properly */
> SEC("socket")
> __log_level(2)
> @@ -255,6 +256,7 @@ __naked void if3_jset_bug(void)
> "exit;"
> ::: __clobber_all);
> }
> +#endif
>
> SEC("socket")
> __log_level(2)
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH bpf-next] selftests/bpf: Fix the invalid operand for instruction issue
2025-08-27 4:48 ` Eduard Zingerman
@ 2025-08-27 8:24 ` Feng Yang
2025-08-27 10:21 ` Eduard Zingerman
0 siblings, 1 reply; 5+ messages in thread
From: Feng Yang @ 2025-08-27 8:24 UTC (permalink / raw)
To: eddyz87
Cc: andrii, ast, bpf, daniel, haoluo, john.fastabend, jolsa, kpsingh,
linux-kernel, linux-kselftest, llvm, martin.lau, sdf, song,
yangfeng59949, yonghong.song
On Tue, 26 Aug 2025 21:48:43 -0700, Eduard Zingerman <eddyz87@gmail.com> wrote:
> On Wed, 2025-08-27 at 11:15 +0800, Feng Yang wrote:
> > From: Feng Yang <yangfeng@kylinos.cn>
> >
> > The following issue occurs when compiling with clang version 17.0.6,
> > but not with version 18.1.8. Add a version restriction to fix this problem.
> >
> > progs/compute_live_registers.c:251:3: error: invalid operand for instruction
> > 251 | "r0 = 1;"
> > | ^
> > <inline asm>:1:22: note: instantiated into assembly here
> > 1 | r0 = 1;r2 = 2;if r1 & 0x7 goto +1;exit;r0 = r2;exit;
> > | ^
> > 1 error generated.
> >
> > Fixes: 4a4b84ba9e453 ("selftests/bpf: verify jset handling in CFG computation")
> > Signed-off-by: Feng Yang <yangfeng@kylinos.cn>
> > ---
> > tools/testing/selftests/bpf/progs/compute_live_registers.c | 2 ++
> > 1 file changed, 2 insertions(+)
> >
> > diff --git a/tools/testing/selftests/bpf/progs/compute_live_registers.c b/tools/testing/selftests/bpf/progs/compute_live_registers.c
> > index 6884ab99a421..56aec43f206f 100644
> > --- a/tools/testing/selftests/bpf/progs/compute_live_registers.c
> > +++ b/tools/testing/selftests/bpf/progs/compute_live_registers.c
> > @@ -240,6 +240,7 @@ __naked void if2(void)
> > ::: __clobber_all);
> > }
> >
> > +#if __clang_major__ >= 18
> Instead of guarding this with compiler version, could you please use
> progs/bpf_misc.h:__imm_insn() macro for the jset instruction?
I don't know much about assembly language. Could you tell me if the following changes are correct?
diff --git a/tools/testing/selftests/bpf/progs/compute_live_registers.c b/tools/testing/selftests/bpf/progs/compute_live_registers.c
index 6884ab99a421..01d73ad76faf 100644
--- a/tools/testing/selftests/bpf/progs/compute_live_registers.c
+++ b/tools/testing/selftests/bpf/progs/compute_live_registers.c
@@ -249,11 +249,13 @@ __naked void if3_jset_bug(void)
asm volatile (
"r0 = 1;"
"r2 = 2;"
- "if r1 & 0x7 goto +1;"
+ ".8byte %[jset];" /* same as 'if r1 & 0x7 goto +1;' */
"exit;"
"r0 = r2;"
"exit;"
- ::: __clobber_all);
+ :
+ : __imm_insn(jset, BPF_JMP_IMM(BPF_JSET, BPF_REG_1, 0x7, 1))
+ : __clobber_all);
}
^ permalink raw reply related [flat|nested] 5+ messages in thread
* Re: [PATCH bpf-next] selftests/bpf: Fix the invalid operand for instruction issue
2025-08-27 8:24 ` Feng Yang
@ 2025-08-27 10:21 ` Eduard Zingerman
0 siblings, 0 replies; 5+ messages in thread
From: Eduard Zingerman @ 2025-08-27 10:21 UTC (permalink / raw)
To: Feng Yang
Cc: andrii, ast, bpf, daniel, haoluo, john.fastabend, jolsa, kpsingh,
linux-kernel, linux-kselftest, llvm, martin.lau, sdf, song,
yonghong.song
On Wed, 2025-08-27 at 16:24 +0800, Feng Yang wrote:
[...]
> I don't know much about assembly language. Could you tell me if the following changes are correct?
Looks correct, should be similar to verifier_search_pruning.c:short_loop1().
Unfortunately, I'm afraid that the best source for assembly syntax doc
are llvm backend tests and sources, e.g.:
- https://github.com/llvm/llvm-project/blob/main/llvm/test/CodeGen/BPF/assembler-disassembler.s
- https://github.com/llvm/llvm-project/blob/main/llvm/test/CodeGen/BPF/assembler-disassembler-v4.s
- https://github.com/llvm/llvm-project/blob/main/llvm/lib/Target/BPF/BPFInstrInfo.td
The directives should gas compatible (subset supported by llvm):
- https://sourceware.org/binutils/docs/as/8byte.html
> diff --git a/tools/testing/selftests/bpf/progs/compute_live_registers.c b/tools/testing/selftests/bpf/progs/compute_live_registers.c
> index 6884ab99a421..01d73ad76faf 100644
> --- a/tools/testing/selftests/bpf/progs/compute_live_registers.c
> +++ b/tools/testing/selftests/bpf/progs/compute_live_registers.c
> @@ -249,11 +249,13 @@ __naked void if3_jset_bug(void)
> asm volatile (
> "r0 = 1;"
> "r2 = 2;"
> - "if r1 & 0x7 goto +1;"
> + ".8byte %[jset];" /* same as 'if r1 & 0x7 goto +1;' */
> "exit;"
> "r0 = r2;"
> "exit;"
> - ::: __clobber_all);
> + :
> + : __imm_insn(jset, BPF_JMP_IMM(BPF_JSET, BPF_REG_1, 0x7, 1))
> + : __clobber_all);
> }
>
^ permalink raw reply [flat|nested] 5+ messages in thread
end of thread, other threads:[~2025-08-27 10:21 UTC | newest]
Thread overview: 5+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2025-08-27 3:15 [PATCH bpf-next] selftests/bpf: Fix the invalid operand for instruction issue Feng Yang
2025-08-27 4:03 ` Alexei Starovoitov
2025-08-27 4:48 ` Eduard Zingerman
2025-08-27 8:24 ` Feng Yang
2025-08-27 10:21 ` Eduard Zingerman
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).