Re: [PATCH bpf-next] selftests/bpf: Fix build of callx tests with GCC

[Vineet Gupta <gvineet@meta.com>]

On 9/24/26 11:21 AM, Alexei Starovoitov wrote:
> From: Alexei Starovoitov <ast@kernel.org>
> 
> GCC doesn't support indirect calls in BPF:
> 
>    progs/verifier_callx.c: In function 'apply_op':
>    progs/verifier_callx.c:977:1: error: indirect call in function, which are not supported by eBPF
>    progs/verifier_callx_rodata.c: In function 'use_shape':
>    progs/verifier_callx_rodata.c:746:1: error: indirect call in function, which are not supported by eBPF

GCC supports them under -mxbpf but unfortunately even that is broken.


> Build verifier_callx, verifier_callx_rodata and callx_rodata progs
> with clang only. Skip callx_rodata_lskel test when progs are built
> with GCC.

Can we do this slightly differently.
Following on same theme have been queued up in bpf-next

2026-08-24 48b69cc8825d selftests/bpf: Enable some of the blocked cpuv4 
instruction tests for bpf-gcc
2026-08-24 8928f5ffdb65 selftests/bpf: Report placeholder tests as SKIP, 
not OK
2026-08-24 b0eeb9e0f548 selftests/bpf: Name the remaining placeholder 
programs dummy_test

> 
> Fixes: e9b22d50796d ("selftests/bpf: Add tests for callx")
> Fixes: b018e998e645 ("selftests/bpf: Add tests for callx through pointers in read-only data")
> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
> ---
>   .../bpf/prog_tests/callx_rodata_lskel.c       |  6 ++++
>   .../selftests/bpf/progs/callx_rodata.c        | 28 +++++++++++++++++--
>   .../selftests/bpf/progs/verifier_callx.c      |  3 +-
>   .../bpf/progs/verifier_callx_rodata.c         |  3 +-
>   4 files changed, 36 insertions(+), 4 deletions(-)
> 
> diff --git a/tools/testing/selftests/bpf/prog_tests/callx_rodata_lskel.c b/tools/testing/selftests/bpf/prog_tests/callx_rodata_lskel.c
> index 96f6179f0492..6a43b880d730 100644
> --- a/tools/testing/selftests/bpf/prog_tests/callx_rodata_lskel.c
> +++ b/tools/testing/selftests/bpf/prog_tests/callx_rodata_lskel.c
> @@ -35,6 +35,12 @@ void test_callx_rodata_lskel(void)
>   	if (!ASSERT_OK_PTR(skel, "open"))
>   		return;
>   
> +	/* gcc doesn't support indirect calls */
> +	if (skel->rodata->skip) {
> +		test__skip();
> +		goto out;
> +	}
> +

We now have common skipping machinery using __skip, see further below.

>   	skel->rodata->bias = 7;
>   
>   	if (!ASSERT_OK(callx_rodata_lskel__load(skel), "load"))
> diff --git a/tools/testing/selftests/bpf/progs/callx_rodata.c b/tools/testing/selftests/bpf/progs/callx_rodata.c
> index 07ae2d1b6b8e..e3f8335f80f8 100644
> --- a/tools/testing/selftests/bpf/progs/callx_rodata.c
> +++ b/tools/testing/selftests/bpf/progs/callx_rodata.c
> @@ -1,16 +1,21 @@
>   // SPDX-License-Identifier: GPL-2.0
>   /* callx through pointers to functions in .rodata, loaded by light skeleton */
>   
> +#include <stdbool.h>
>   #include <linux/bpf.h>
>   #include <bpf/bpf_helpers.h>
>   
> -typedef int (*op_fn)(int);
> -
>   /* set by user space before the programs are loaded, it's in .rodata too */
>   const volatile int bias = 1;
>   
>   int op_idx;
>   
> +#ifdef __clang__

It unfortunatete that this doesn't have a compiler feature test macro 
__BPF_FEATURE_* in clang or gcc. It seems we have existing 
CAN_USE_BPF_ST and CAN_USE_GOTOL, maybe we can define CAN_USE_CALLX in 
bpf_misc.h and use that as a gate, keeps one place for all such 
__clang/gcc workaround definitions instead of them scattered all over. I 
have a few pending patches which clean things up.

> +
> +const volatile bool skip = false;
> +
> +typedef int (*op_fn)(int);
> +
>   static __noinline int add_bias(int x)
>   {
>   	return x + bias;
> @@ -41,6 +46,25 @@ int both_ops(void *ctx)
>   	return ops[1](ops[0](4));
>   }
>   
> +#else
> +
> +/* gcc doesn't support indirect calls */
> +const volatile bool skip = true;
> +
> +SEC("socket")
> +int select_op(void *ctx)
> +{
> +	return 0;
> +}
> +
> +SEC("socket")
> +int both_ops(void *ctx)
> +{
> +	return 0;
> +}
> +
> +#endif
> +

This can be rewritten using exising skipping machinery as shown in 
bpf-next patch: 8928f5ffdb65 ("selftests/bpf: Report placeholder tests 
as SKIP, not OK")

#ifdef CAN_USE_CALLX

llvm_code

#else
__skip("gcc doesnt support callx")
int dummy_test(void *ctx)
{
     return 0;
}
#endif

This avoids the explicit skipping logic.

>   /* no callx here */
>   SEC("socket")
>   int read_bias(void *ctx)
> diff --git a/tools/testing/selftests/bpf/progs/verifier_callx.c b/tools/testing/selftests/bpf/progs/verifier_callx.c
> index ea3d3f97da31..bc31d0221f6a 100644
> --- a/tools/testing/selftests/bpf/progs/verifier_callx.c
> +++ b/tools/testing/selftests/bpf/progs/verifier_callx.c
> @@ -6,7 +6,8 @@
>   #include "bpf_misc.h"
>   #include "../../../include/linux/filter.h"
>   
> -#if defined(__TARGET_ARCH_x86) || defined(__TARGET_ARCH_arm64)
> +/* gcc doesn't support indirect calls */
> +#if (defined(__TARGET_ARCH_x86) || defined(__TARGET_ARCH_arm64)) && defined(__clang__)
>   
>   #define CALLX_INSN(DST, SRC, OFF, IMM) \
>   	BPF_RAW_INSN(BPF_JMP | BPF_CALL | BPF_X, DST, SRC, OFF, IMM)
> diff --git a/tools/testing/selftests/bpf/progs/verifier_callx_rodata.c b/tools/testing/selftests/bpf/progs/verifier_callx_rodata.c
> index 65d34769c8bf..db3169a1ddf6 100644
> --- a/tools/testing/selftests/bpf/progs/verifier_callx_rodata.c
> +++ b/tools/testing/selftests/bpf/progs/verifier_callx_rodata.c
> @@ -6,7 +6,8 @@
>   #include "bpf_misc.h"
>   #include "../../../include/linux/filter.h"
>   
> -#if defined(__TARGET_ARCH_x86) || defined(__TARGET_ARCH_arm64)
> +/* gcc doesn't support indirect calls */
> +#if (defined(__TARGET_ARCH_x86) || defined(__TARGET_ARCH_arm64)) && defined(__clang__)

Instead of __clang scattered all over, CAN_USE_CALLX in bpf_misc.h so 
its one stop shop to clean things up in future.

BTW this is

Thx,
-Vineet