mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Yonghong Song <yonghong.song@linux.dev>
To: chenyuan_fl@163.com
Cc: andrii@kernel.org, ast@kernel.org, bpf@vger.kernel.org,
	chenyuan@kylinos.cn, daniel@iogearbox.net, eddyz87@gmail.com,
	jolsa@kernel.org, linux-kernel@vger.kernel.org,
	martin.lau@linux.dev, memxor@gmail.com, song@kernel.org
Subject: Re: [PATCH bpf v2 2/2] selftests/bpf: Add regression test for kfunc implicit arg injection with stale register
Date: Mon, 1 Jun 2026 10:17:07 -0700	[thread overview]
Message-ID: <9cc9ce47-c45e-4878-ad3a-6e6967399ab9@linux.dev> (raw)
In-Reply-To: <20260601064611.807637-3-chenyuan_fl@163.com>



On 5/31/26 11:46 PM, chenyuan_fl@163.com wrote:
> From: Yuan Chen <chenyuan@kylinos.cn>
>
> The preceding patch fixes a bug where the BPF verifier failed to detect
> an implicit struct bpf_prog_aux * argument in a module kfunc, causing
> the kernel to dereference a stale register value as a pointer and crash.
>
> Add a selftest that deliberately contaminates BPF R2 (the register slot
> for the 2nd kfunc argument, which maps to the implicit bpf_prog_aux *)
> with a known magic value 0xDEAD via inline assembly, then calls a
> KF_IMPLICIT_ARGS kfunc that validates whether injection occurred:
>
>    - If the kernel correctly injects env->prog->aux into R2, the kfunc
>      receives a real bpf_prog_aux pointer (≠ 0xDEAD) and returns the
>      caller-supplied marker value.
>
>    - If injection is skipped (the original bug), the kfunc receives the
>      stale 0xDEAD value and returns -EINVAL, which the BPF program
>      detects as a test failure.
>
> The magic value 0xDEAD is chosen with bit 31 clear to avoid BPF ALU64
> sign-extension when used as a 32-bit immediate, ensuring the comparison
> in the kfunc matches the actual register value.  The kfunc is associated
> with a struct_ops map to exercise the exact call path (struct_ops →
> kfunc with KF_IMPLICIT_ARGS → bpf_prog_aux dereference) that triggered
> the original crash.

Not sure whether this patch makes sense or not. The CI has pahole 1.31
so the test will always succeed.

>
> Signed-off-by: Yuan Chen <chenyuan@kylinos.cn>
> ---
>   .../bpf/prog_tests/test_struct_ops_assoc.c    |  5 +++
>   .../selftests/bpf/progs/struct_ops_assoc.c    | 40 +++++++++++++++++++
>   .../selftests/bpf/test_kmods/bpf_testmod.c    |  9 +++++
>   .../bpf/test_kmods/bpf_testmod_kfunc.h        |  1 +
>   4 files changed, 55 insertions(+)
>
> diff --git a/tools/testing/selftests/bpf/prog_tests/test_struct_ops_assoc.c b/tools/testing/selftests/bpf/prog_tests/test_struct_ops_assoc.c
> index 461ded722351..123bd2c7a292 100644
> --- a/tools/testing/selftests/bpf/prog_tests/test_struct_ops_assoc.c
> +++ b/tools/testing/selftests/bpf/prog_tests/test_struct_ops_assoc.c
> @@ -35,6 +35,10 @@ static void test_st_ops_assoc(void)
>   					    skel->maps.st_ops_map_b, NULL);
>   	ASSERT_OK(err, "bpf_program__assoc_struct_ops(sys_enter_prog_b, st_ops_map_b)");
>   
> +	err = bpf_program__assoc_struct_ops(skel->progs.sys_enter_prog_test_aux_inject,
> +					    skel->maps.st_ops_map_a, NULL);
> +	ASSERT_OK(err, "bpf_program__assoc_struct_ops(sys_enter_prog_test_aux_inject, st_ops_map_a)");
> +
>   	/* sys_enter_prog_a already associated with map_a */
>   	err = bpf_program__assoc_struct_ops(skel->progs.sys_enter_prog_a,
>   					    skel->maps.st_ops_map_b, NULL);
> @@ -52,6 +56,7 @@ static void test_st_ops_assoc(void)
>   
>   	ASSERT_EQ(skel->bss->test_err_a, 0, "skel->bss->test_err_a");
>   	ASSERT_EQ(skel->bss->test_err_b, 0, "skel->bss->test_err_b");
> +	ASSERT_EQ(skel->bss->test_err_inject, 0, "skel->bss->test_err_inject");
>   
>   	/* run syscall_prog that calls .test_1 and checks return */
>   	err = bpf_prog_test_run_opts(bpf_program__fd(skel->progs.syscall_prog_a), NULL);
> diff --git a/tools/testing/selftests/bpf/progs/struct_ops_assoc.c b/tools/testing/selftests/bpf/progs/struct_ops_assoc.c
> index 68842e3f936b..dd322d43ff5e 100644
> --- a/tools/testing/selftests/bpf/progs/struct_ops_assoc.c
> +++ b/tools/testing/selftests/bpf/progs/struct_ops_assoc.c
> @@ -103,3 +103,43 @@ SEC(".struct_ops.link")
>   struct bpf_testmod_multi_st_ops st_ops_map_b = {
>   	.test_1 = (void *)test_1_b,
>   };
> +
> +/* Test for aux injection with stale register contamination.
> + *
> + * This test reproduces the scenario where the BPF verifier fails to
> + * inject the implicit bpf_prog_aux pointer for kfuncs with
> + * KF_IMPLICIT_ARGS.  The program uses inline assembly to explicitly
> + * set R2 (the register for the 2nd kfunc argument, which maps to
> + * the implicit bpf_prog_aux *) to a known magic value (0xDEAD,
> + * chosen with bit 31 clear to avoid BPF ALU64 sign-extension):
> + *
> + *   asm volatile("r2 = %[magic]" :: [magic] "ri"(0xDEAD) : "r2");
> + *
> + * Then bpf_kfunc_aux_inject_stale() is called.  The kernel verifier
> + * should inject the real bpf_prog_aux into R2, overriding the magic
> + * value.  The kfunc compares aux against 0xDEAD:
> + *
> + *   - aux == 0xDEAD  → kernel failed to inject aux → test fails
> + *   - aux != 0xDEAD  → kernel correctly injected aux → test passes
> + */
> +int test_err_inject;
> +
> +SEC("tp_btf/sys_enter")
> +int BPF_PROG(sys_enter_prog_test_aux_inject, struct pt_regs *regs, long id)
> +{
> +	struct task_struct *task;
> +	int marker = 0x5A5A;
> +	int ret;
> +
> +	task = bpf_get_current_task_btf();
> +	if (!test_pid || task->pid != test_pid)
> +		return 0;
> +
> +	asm volatile("r2 = %[magic]" :: [magic] "ri"(0xDEAD) : "r2");
> +
> +	ret = bpf_kfunc_aux_inject_stale(marker);
> +	if (ret != marker)
> +		test_err_inject++;
> +
> +	return 0;
> +}
> diff --git a/tools/testing/selftests/bpf/test_kmods/bpf_testmod.c b/tools/testing/selftests/bpf/test_kmods/bpf_testmod.c
> index 0be918fe3021..da95a6de3bbf 100644
> --- a/tools/testing/selftests/bpf/test_kmods/bpf_testmod.c
> +++ b/tools/testing/selftests/bpf/test_kmods/bpf_testmod.c
> @@ -1316,6 +1316,7 @@ __bpf_kfunc int bpf_kfunc_multi_st_ops_test_1_assoc(struct st_ops_args *args, st
>   __bpf_kfunc int bpf_kfunc_implicit_arg(int a, struct bpf_prog_aux *aux);
>   __bpf_kfunc int bpf_kfunc_implicit_arg_legacy(int a, int b, struct bpf_prog_aux *aux);
>   __bpf_kfunc int bpf_kfunc_implicit_arg_legacy_impl(int a, int b, struct bpf_prog_aux *aux);
> +__bpf_kfunc int bpf_kfunc_aux_inject_stale(int marker, struct bpf_prog_aux *aux);
>   
>   /* hook targets */
>   noinline void bpf_testmod_test_hardirq_fn(void) { barrier(); }
> @@ -1399,6 +1400,7 @@ BTF_ID_FLAGS(func, bpf_kfunc_multi_st_ops_test_1_assoc, KF_IMPLICIT_ARGS)
>   BTF_ID_FLAGS(func, bpf_kfunc_implicit_arg, KF_IMPLICIT_ARGS)
>   BTF_ID_FLAGS(func, bpf_kfunc_implicit_arg_legacy, KF_IMPLICIT_ARGS)
>   BTF_ID_FLAGS(func, bpf_kfunc_implicit_arg_legacy_impl)
> +BTF_ID_FLAGS(func, bpf_kfunc_aux_inject_stale, KF_IMPLICIT_ARGS)
>   BTF_ID_FLAGS(func, bpf_kfunc_trigger_ctx_check)
>   BTF_KFUNCS_END(bpf_testmod_check_kfunc_ids)
>   
> @@ -1916,6 +1918,13 @@ int bpf_kfunc_implicit_arg_legacy_impl(int a, int b, struct bpf_prog_aux *aux)
>   	return bpf_kfunc_implicit_arg_legacy(a, b, aux);
>   }
>   
> +int bpf_kfunc_aux_inject_stale(int marker, struct bpf_prog_aux *aux)
> +{
> +	if ((unsigned long)aux == 0xDEAD)
> +		return -EINVAL;
> +	return marker;
> +}
> +
>   static int multi_st_ops_reg(void *kdata, struct bpf_link *link)
>   {
>   	struct bpf_testmod_multi_st_ops *st_ops =
> diff --git a/tools/testing/selftests/bpf/test_kmods/bpf_testmod_kfunc.h b/tools/testing/selftests/bpf/test_kmods/bpf_testmod_kfunc.h
> index 2edc36b66de9..c18791e96b21 100644
> --- a/tools/testing/selftests/bpf/test_kmods/bpf_testmod_kfunc.h
> +++ b/tools/testing/selftests/bpf/test_kmods/bpf_testmod_kfunc.h
> @@ -192,6 +192,7 @@ int *bpf_kfunc_ret_rcu_test_nostruct(int rdonly_buf_size) __ksym;
>   #ifndef __KERNEL__
>   extern int bpf_kfunc_multi_st_ops_test_1(struct st_ops_args *args, u32 id) __weak __ksym;
>   extern int bpf_kfunc_multi_st_ops_test_1_assoc(struct st_ops_args *args) __weak __ksym;
> +extern int bpf_kfunc_aux_inject_stale(int marker) __weak __ksym;
>   #endif
>   
>   struct prog_test_member *bpf_kfunc_get_default_trusted_ptr_test(void) __ksym;


      reply	other threads:[~2026-06-01 17:17 UTC|newest]

Thread overview: 48+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-04-07  8:09 [PATCH] bpf: fix btf_types_are_same for cross-BTF type comparison chenyuan_fl
2026-04-07  8:58 ` Leon Hwang
2026-04-07  9:01 ` bot+bpf-ci
2026-04-07 11:19 ` Alan Maguire
2026-05-15 18:27 ` Ihor Solodrai
2026-06-01  6:46 ` [PATCH bpf v2 0/2] bpf: Fix kfunc implicit arg injection and add selftest chenyuan_fl
2026-06-01  6:46   ` [PATCH bpf v2 1/2] bpf: Fix kfunc implicit arg inject type detection to prevent invalid pointer deref chenyuan_fl
2026-06-01  7:42     ` bot+bpf-ci
2026-06-01 19:32       ` Eduard Zingerman
2026-06-02  8:58         ` [PATCH bpf v3 0/2] bpf: Fix kfunc implicit arg injection and add selftest chenyuan_fl
2026-06-02  8:58           ` [PATCH bpf v3 1/2] bpf: Fix kfunc implicit arg inject type detection to prevent invalid pointer deref chenyuan_fl
2026-06-02  9:44             ` bot+bpf-ci
2026-06-02 18:52             ` Ihor Solodrai
2026-06-04  9:14               ` chenyuan
2026-06-04 10:21                 ` Alan Maguire
2026-06-09  0:46                   ` Ihor Solodrai
2026-06-02  8:58           ` [PATCH bpf v3 2/2] selftests/bpf: Add regression test for kfunc implicit arg injection chenyuan_fl
2026-06-02  9:44             ` bot+bpf-ci
2026-06-02  9:38         ` [PATCH bpf v4 0/2] bpf: Fix kfunc implicit arg injection and add selftest chenyuan_fl
2026-06-02  9:38           ` [PATCH bpf v4 1/2] bpf: Fix kfunc implicit arg inject type detection to prevent invalid pointer deref chenyuan_fl
2026-06-02 10:42             ` bot+bpf-ci
2026-06-05  0:42             ` Eduard Zingerman
2026-06-02  9:38           ` [PATCH bpf v4 2/2] selftests/bpf: Add regression test for kfunc implicit arg injection chenyuan_fl
2026-06-02 10:27             ` bot+bpf-ci
2026-06-02 17:36             ` kernel test robot
2026-06-02 18:37             ` kernel test robot
2026-06-05  1:29             ` Eduard Zingerman
2026-06-08 14:26           ` [PATCH bpf v5 0/2] bpf: Fix kfunc implicit arg injection and add selftest chenyuan_fl
2026-06-08 14:26             ` [PATCH bpf v5 1/2] bpf: Fix kfunc implicit arg inject type detection to prevent invalid pointer deref chenyuan_fl
2026-06-08 14:52               ` bot+bpf-ci
2026-06-08 17:28               ` Eduard Zingerman
2026-06-09  0:54               ` Ihor Solodrai
2026-06-09  1:00                 ` Eduard Zingerman
2026-06-09  1:18                   ` Ihor Solodrai
2026-06-09  6:20                     ` Eduard Zingerman
2026-06-08 14:26             ` [PATCH bpf v5 2/2] selftests/bpf: strengthen bpf_kfunc_implicit_arg to verify aux injection chenyuan_fl
2026-06-08 17:34               ` Eduard Zingerman
2026-06-09  0:59               ` Ihor Solodrai
2026-06-08 19:58             ` [PATCH bpf v5 0/2] bpf: Fix kfunc implicit arg injection and add selftest Alexei Starovoitov
2026-06-09 12:52             ` [PATCH bpf v6 " chenyuan_fl
2026-06-09 12:52               ` [PATCH bpf v6 1/2] bpf: Fix kfunc implicit arg inject type detection to prevent invalid pointer deref chenyuan_fl
2026-06-09 13:32                 ` bot+bpf-ci
2026-06-09 16:39                 ` Eduard Zingerman
2026-06-09 12:52               ` [PATCH bpf v6 2/2] selftests/bpf: strengthen bpf_kfunc_implicit_arg to verify aux injection chenyuan_fl
2026-06-01 17:12     ` [PATCH bpf v2 1/2] bpf: Fix kfunc implicit arg inject type detection to prevent invalid pointer deref Yonghong Song
2026-06-01 21:36     ` Eduard Zingerman
2026-06-01  6:46   ` [PATCH bpf v2 2/2] selftests/bpf: Add regression test for kfunc implicit arg injection with stale register chenyuan_fl
2026-06-01 17:17     ` Yonghong Song [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=9cc9ce47-c45e-4878-ad3a-6e6967399ab9@linux.dev \
    --to=yonghong.song@linux.dev \
    --cc=andrii@kernel.org \
    --cc=ast@kernel.org \
    --cc=bpf@vger.kernel.org \
    --cc=chenyuan@kylinos.cn \
    --cc=chenyuan_fl@163.com \
    --cc=daniel@iogearbox.net \
    --cc=eddyz87@gmail.com \
    --cc=jolsa@kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=martin.lau@linux.dev \
    --cc=memxor@gmail.com \
    --cc=song@kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

Powered by JetHome