From: Yonghong Song <yhs@fb.com>
To: Kui-Feng Lee <kuifeng@fb.com>,
"bpf@vger.kernel.org" <bpf@vger.kernel.org>
Cc: "daniel@iogearbox.net" <daniel@iogearbox.net>,
Kernel Team <Kernel-team@fb.com>,
"ast@kernel.org" <ast@kernel.org>,
"andrii@kernel.org" <andrii@kernel.org>
Subject: Re: [RFC PATCH bpf-next 0/7] bpf: Support struct value argument for trampoline base progs
Date: Thu, 28 Jul 2022 16:30:49 -0700 [thread overview]
Message-ID: <a31dfd32-3065-1881-e2e2-3c420568232a@fb.com> (raw)
In-Reply-To: <17dbbb831db12049ebfb5161e380c9078fbddad5.camel@fb.com>
On 7/28/22 12:57 PM, Kui-Feng Lee wrote:
> On Thu, 2022-07-28 at 10:46 -0700, Yonghong Song wrote:
>>
>>
>> On 7/28/22 8:46 AM, Kui-Feng Lee wrote:
>>> On Tue, 2022-07-26 at 10:11 -0700, Yonghong Song wrote:
>>>> Currently struct arguments are not supported for trampoline based
>>>> progs.
>>>> One of major reason is that struct argument may pass by value
>>>> which
>>>> may
>>>> use more than one registers. This breaks trampoline progs where
>>>> each argument is assumed to take one register. bcc community
>>>> reported
>>>> the
>>>> issue ([1]) where struct argument is not supported for fentry
>>>> program.
>>>> typedef struct {
>>>> uid_t val;
>>>> } kuid_t;
>>>> typedef struct {
>>>> gid_t val;
>>>> } kgid_t;
>>>> int security_path_chown(struct path *path, kuid_t uid, kgid_t
>>>> gid);
>>>> Inside Meta, we also have a use case to attach to
>>>> tcp_setsockopt()
>>>> typedef struct {
>>>> union {
>>>> void *kernel;
>>>> void __user *user;
>>>> };
>>>> bool is_kernel : 1;
>>>> } sockptr_t;
>>>> int tcp_setsockopt(struct sock *sk, int level, int optname,
>>>> sockptr_t optval, unsigned int optlen);
>>>>
>>>> This patch added struct value support for bpf tracing programs
>>>> which
>>>> uses trampoline. struct argument size needs to be 16 or less so
>>>> it can fit in one or two registers. Based on analysis on llvm and
>>>> experiments, atruct argument size greater than 16 will be passed
>>>> as pointer to the struct.
>>>
>>> Is it possible to force llvm to always pass a pointer to a struct
>>> over
>>> 8 bytes (the size of single register) for the BPF traget?
>>
>> This is already the case for bpf target. Any struct parameter (1
>> byte, 2
>> bytes, ..., 8 types, ..., 16 bytes, ...) will be passed as a
>> reference.
>>
>> But this is not the case for most other architectures. For example,
>> for
>> x86_64, in most cases, struct size <= 16 will be passed with two
>> registers instead of as a reference.
>
> I ask this question because you modify the signature of a bpf program
> to a pointer to a struct in patch #4. Is that necessary if the
> compiler passes a struct paramter as a reference?
Note that The true bpf program signature is only one.
long bpf_prog(<ctx_type> *ctx)
BPF_PROG is a macro for user friendly purpose.
For example,
+int BPF_PROG(test_struct_arg_1, struct bpf_testmod_struct_arg_2 *a, int
b, int c)
after macro expansion:
int test_struct_arg_1(unsigned long long *ctx);
static __attribute__((always_inline))
typeof(test_struct_arg_1(0)) ____test_struct_arg_1(
unsigned long long *ctx,
struct bpf_testmod_struct_arg_2 *a, int b, int c);
...
next prev parent reply other threads:[~2022-07-28 23:31 UTC|newest]
Thread overview: 20+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-07-26 17:11 [RFC PATCH bpf-next 0/7] bpf: Support struct value argument for trampoline base progs Yonghong Song
2022-07-26 17:11 ` [RFC PATCH bpf-next 1/7] bpf: Always return corresponding btf_type in __get_type_size() Yonghong Song
2022-07-26 17:11 ` [RFC PATCH bpf-next 2/7] bpf: Add struct argument info in btf_func_model Yonghong Song
2022-08-09 0:02 ` Andrii Nakryiko
2022-08-09 17:38 ` Yonghong Song
2022-08-10 0:25 ` Andrii Nakryiko
2022-08-11 6:24 ` Yonghong Song
2022-07-26 17:11 ` [RFC PATCH bpf-next 3/7] bpf: x86: Rename stack_size to regs_off in {save,restore}_regs() Yonghong Song
2022-07-26 17:11 ` [RFC PATCH bpf-next 4/7] bpf: x86: Support in-register struct arguments Yonghong Song
2022-07-29 11:10 ` Jiri Olsa
2022-07-31 17:00 ` Yonghong Song
2022-07-26 17:11 ` [RFC PATCH bpf-next 5/7] bpf: arm64: No support of struct value argument Yonghong Song
2022-07-26 17:12 ` [RFC PATCH bpf-next 6/7] bpf: Populate struct value info in btf_func_model Yonghong Song
2022-07-26 17:12 ` [RFC PATCH bpf-next 7/7] selftests/bpf: Add struct value tests with fentry programs Yonghong Song
2022-07-28 15:46 ` [RFC PATCH bpf-next 0/7] bpf: Support struct value argument for trampoline base progs Kui-Feng Lee
2022-07-28 17:46 ` Yonghong Song
2022-07-28 19:57 ` Kui-Feng Lee
2022-07-28 23:30 ` Yonghong Song [this message]
2022-07-29 18:04 ` Kui-Feng Lee
2022-08-02 23:46 ` Yonghong Song
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=a31dfd32-3065-1881-e2e2-3c420568232a@fb.com \
--to=yhs@fb.com \
--cc=Kernel-team@fb.com \
--cc=andrii@kernel.org \
--cc=ast@kernel.org \
--cc=bpf@vger.kernel.org \
--cc=daniel@iogearbox.net \
--cc=kuifeng@fb.com \
/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).