bpf.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Jiri Olsa <olsajiri@gmail.com>
To: Alan Maguire <alan.maguire@oracle.com>
Cc: ast@kernel.org, andrii@kernel.org, daniel@iogearbox.net,
	martin.lau@linux.dev, eddyz87@gmail.com, song@kernel.org,
	yonghong.song@linux.dev, john.fastabend@gmail.com,
	kpsingh@kernel.org, sdf@fomichev.me, haoluo@google.com,
	mykolal@fb.com, bpf@vger.kernel.org
Subject: Re: [PATCH bpf-next] selftests/bpf: add btf dedup test covering module BTF dedup
Date: Fri, 20 Jun 2025 15:34:14 +0200	[thread overview]
Message-ID: <aFVjVoafmmPeUqiz@krava> (raw)
In-Reply-To: <20250430134249.2451066-1-alan.maguire@oracle.com>

On Wed, Apr 30, 2025 at 02:42:49PM +0100, Alan Maguire wrote:
> Recently issues were observed with module BTF deduplication failures
> [1].  Add a dedup selftest that ensures that core kernel types are
> referenced from split BTF as base BTF types.  To do this use bpf_testmod
> functions which utilize core kernel types, specifically
> 
> ssize_t
> bpf_testmod_test_write(struct file *file, struct kobject *kobj,
>                        struct bin_attribute *bin_attr,
>                        char *buf, loff_t off, size_t len);
> 
> __bpf_kfunc struct sock *bpf_kfunc_call_test3(struct sock *sk);
> 
> __bpf_kfunc void bpf_kfunc_call_test_pass_ctx(struct __sk_buff *skb);
> 
> For each of these ensure that the types they reference -
> struct file, struct kobject, struct bin_attr etc - are in base BTF.
> Note that because bpf_testmod.ko is built with distilled base BTF
> the associated reference types - i.e. the PTR that points at a
> "struct file" - will be in split BTF.  As a result the test resolves
> typedef and pointer references and verifies the pointed-at or
> typedef'ed type is in base BTF.  Because we use BTF from
> /sys/kernel/btf/bpf_testmod relocation has occurred for the
> referenced types and they will be base - not distilled base - types.
> 
> For large-scale dedup issues, we see such types appear in split BTF and
> as a result this test fails.  Hence it is proposed as a test which will
> fail when large-scale dedup issues have occurred.
> 
> [1] https://lore.kernel.org/dwarves/CAADnVQL+-LiJGXwxD3jEUrOonO-fX0SZC8496dVzUXvfkB7gYQ@mail.gmail.com/
> 
> Signed-off-by: Alan Maguire <alan.maguire@oracle.com>

hi Alan,
this one started to fail in my tests.. it's likely some screw up in
my environment, but I haven't found the cause yet, I'm using the
pahole 1.30 .. just cheking if it's known issue already ;-)

thanks,
jirka


test_split_module:PASS:vmlinux_btf 0 nsec
test_split_module:PASS:nr_base_types 0 nsec
test_split_module:PASS:split_btf 0 nsec
test_split_module:PASS:func_id 0 nsec
test_split_module:PASS:func_id_type 0 nsec
test_split_module:PASS:func_proto_id_type 0 nsec
test_split_module:PASS:is_func_proto 0 nsec
test_split_module:PASS:param_ref_type 0 nsec
test_split_module:PASS:param_ref_type 0 nsec
test_split_module:FAIL:verify_base_type unexpected verify_base_type: actual 183322 >= expected 183225
#33/5    btf_dedup_split/split_module:FAIL
#33      btf_dedup_split:FAIL


> ---
>  .../bpf/prog_tests/btf_dedup_split.c          | 101 ++++++++++++++++++
>  1 file changed, 101 insertions(+)
> 
> diff --git a/tools/testing/selftests/bpf/prog_tests/btf_dedup_split.c b/tools/testing/selftests/bpf/prog_tests/btf_dedup_split.c
> index d9024c7a892a..5bc15bb6b7ce 100644
> --- a/tools/testing/selftests/bpf/prog_tests/btf_dedup_split.c
> +++ b/tools/testing/selftests/bpf/prog_tests/btf_dedup_split.c
> @@ -440,6 +440,105 @@ static void test_split_dup_struct_in_cu()
>  	btf__free(btf1);
>  }
>  
> +/* Ensure module split BTF dedup worked correctly; when dedup fails badly
> + * core kernel types are in split BTF also, so ensure that references to
> + * such types point at base - not split - BTF.
> + *
> + * bpf_testmod_test_write() has multiple core kernel type parameters;
> + *
> + * ssize_t
> + * bpf_testmod_test_write(struct file *file, struct kobject *kobj,
> + *                        struct bin_attribute *bin_attr,
> + *                        char *buf, loff_t off, size_t len);
> + *
> + * Ensure each of the FUNC_PROTO params is a core kernel type.
> + *
> + * Do the same for
> + *
> + * __bpf_kfunc struct sock *bpf_kfunc_call_test3(struct sock *sk);
> + *
> + * ...and
> + *
> + * __bpf_kfunc void bpf_kfunc_call_test_pass_ctx(struct __sk_buff *skb);
> + *
> + */
> +const char *mod_funcs[] = {
> +	"bpf_testmod_test_write",
> +	"bpf_kfunc_call_test3",
> +	"bpf_kfunc_call_test_pass_ctx"
> +};
> +
> +static void test_split_module(void)
> +{
> +	struct btf *vmlinux_btf, *btf1 = NULL;
> +	int i, nr_base_types;
> +
> +	vmlinux_btf = btf__load_vmlinux_btf();
> +	if (!ASSERT_OK_PTR(vmlinux_btf, "vmlinux_btf"))
> +		return;
> +	nr_base_types = btf__type_cnt(vmlinux_btf);
> +	if (!ASSERT_GT(nr_base_types, 0, "nr_base_types"))
> +		goto cleanup;
> +
> +	btf1 = btf__parse_split("/sys/kernel/btf/bpf_testmod", vmlinux_btf);
> +	if (!ASSERT_OK_PTR(btf1, "split_btf"))
> +		return;
> +
> +	for (i = 0; i < ARRAY_SIZE(mod_funcs); i++) {
> +		const struct btf_param *p;
> +		const struct btf_type *t;
> +		__u16 vlen;
> +		__u32 id;
> +		int j;
> +
> +		id = btf__find_by_name_kind(btf1, mod_funcs[i], BTF_KIND_FUNC);
> +		if (!ASSERT_GE(id, nr_base_types, "func_id"))
> +			goto cleanup;
> +		t = btf__type_by_id(btf1, id);
> +		if (!ASSERT_OK_PTR(t, "func_id_type"))
> +			goto cleanup;
> +		t = btf__type_by_id(btf1, t->type);
> +		if (!ASSERT_OK_PTR(t, "func_proto_id_type"))
> +			goto cleanup;
> +		if (!ASSERT_EQ(btf_is_func_proto(t), true, "is_func_proto"))
> +			goto cleanup;
> +		vlen = btf_vlen(t);
> +
> +		for (j = 0, p = btf_params(t); j < vlen; j++, p++) {
> +			/* bpf_testmod uses resilient split BTF, so any
> +			 * reference types will be added to split BTF and their
> +			 * associated targets will be base BTF types; for example
> +			 * for a "struct sock *" the PTR will be in split BTF
> +			 * while the "struct sock" will be in base.
> +			 *
> +			 * In some cases like loff_t we have to resolve
> +			 * multiple typedefs hence the while() loop below.
> +			 *
> +			 * Note that resilient split BTF generation depends
> +			 * on pahole version, so we do not assert that
> +			 * reference types are in split BTF, as if pahole
> +			 * does not support resilient split BTF they will
> +			 * also be base BTF types.
> +			 */
> +			id = p->type;
> +			do {
> +				t = btf__type_by_id(btf1, id);
> +				if (!ASSERT_OK_PTR(t, "param_ref_type"))
> +					goto cleanup;
> +				if (!btf_is_mod(t) && !btf_is_ptr(t) && !btf_is_typedef(t))
> +					break;
> +				id = t->type;
> +			} while (true);
> +
> +			if (!ASSERT_LT(id, nr_base_types, "verify_base_type"))
> +				goto cleanup;
> +		}
> +	}
> +cleanup:
> +	btf__free(btf1);
> +	btf__free(vmlinux_btf);
> +}
> +
>  void test_btf_dedup_split()
>  {
>  	if (test__start_subtest("split_simple"))
> @@ -450,4 +549,6 @@ void test_btf_dedup_split()
>  		test_split_fwd_resolve();
>  	if (test__start_subtest("split_dup_struct_in_cu"))
>  		test_split_dup_struct_in_cu();
> +	if (test__start_subtest("split_module"))
> +		test_split_module();
>  }
> -- 
> 2.39.3
> 

  parent reply	other threads:[~2025-06-20 13:34 UTC|newest]

Thread overview: 7+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-04-30 13:42 [PATCH bpf-next] selftests/bpf: add btf dedup test covering module BTF dedup Alan Maguire
2025-04-30 21:30 ` Eduard Zingerman
2025-05-01 21:10 ` patchwork-bot+netdevbpf
2025-06-20 13:34 ` Jiri Olsa [this message]
2025-06-20 13:51   ` Jiri Olsa
2025-06-20 15:41     ` Alan Maguire
2025-06-24 11:48       ` Jiri Olsa

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=aFVjVoafmmPeUqiz@krava \
    --to=olsajiri@gmail.com \
    --cc=alan.maguire@oracle.com \
    --cc=andrii@kernel.org \
    --cc=ast@kernel.org \
    --cc=bpf@vger.kernel.org \
    --cc=daniel@iogearbox.net \
    --cc=eddyz87@gmail.com \
    --cc=haoluo@google.com \
    --cc=john.fastabend@gmail.com \
    --cc=kpsingh@kernel.org \
    --cc=martin.lau@linux.dev \
    --cc=mykolal@fb.com \
    --cc=sdf@fomichev.me \
    --cc=song@kernel.org \
    --cc=yonghong.song@linux.dev \
    /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;
as well as URLs for NNTP newsgroup(s).