On 2026/1/16 08:42, Andrii Nakryiko wrote:
> On Mon, Jan 12, 2026 at 6:58 AM Leon Hwang <[email protected]> wrote:
>>
>> To support the extended BPF syscall introduced in the previous commit,
>> introduce the following internal APIs:
>>
>> * 'sys_bpf_ext()'
>> * 'sys_bpf_ext_fd()'
>> They wrap the raw 'syscall()' interface to support passing extended
>> attributes.
>> * 'probe_sys_bpf_ext()'
>> Check whether current kernel supports the extended attributes.
>>
>> Signed-off-by: Leon Hwang <[email protected]>
>> ---
>> tools/lib/bpf/bpf.c | 34 +++++++++++++++++++++++++++++++++
>> tools/lib/bpf/features.c | 8 ++++++++
>> tools/lib/bpf/libbpf_internal.h | 3 +++
>> 3 files changed, 45 insertions(+)
>>
>> diff --git a/tools/lib/bpf/bpf.c b/tools/lib/bpf/bpf.c
>> index 21b57a629916..d44e667aaf02 100644
>> --- a/tools/lib/bpf/bpf.c
>> +++ b/tools/lib/bpf/bpf.c
>> @@ -69,6 +69,40 @@ static inline __u64 ptr_to_u64(const void *ptr)
>> return (__u64) (unsigned long) ptr;
>> }
>>
>> +static inline int sys_bpf_ext(enum bpf_cmd cmd, union bpf_attr *attr,
>> + unsigned int size,
>> + struct bpf_common_attr *common_attr,
>
> nit: kernel uses consistent attr_common/size_common pattern, but here
> you are inverting attr_common -> common_attr, let's not?
>
Ack.
I'll keep the same pattern.
>> + unsigned int size_common)
>> +{
>> + cmd = common_attr ? (cmd | BPF_COMMON_ATTRS) : (cmd &
>> ~BPF_COMMON_ATTRS);
>> + return syscall(__NR_bpf, cmd, attr, size, common_attr, size_common);
>> +}
>> +
>> +static inline int sys_bpf_ext_fd(enum bpf_cmd cmd, union bpf_attr *attr,
>> + unsigned int size,
>> + struct bpf_common_attr *common_attr,
>> + unsigned int size_common)
>> +{
>> + int fd;
>> +
>> + fd = sys_bpf_ext(cmd, attr, size, common_attr, size_common);
>> + return ensure_good_fd(fd);
>> +}
>> +
>> +int probe_sys_bpf_ext(void)
>> +{
>> + const size_t attr_sz = offsetofend(union bpf_attr, prog_token_fd);
>> + union bpf_attr attr;
>> + int fd;
>> +
>> + memset(&attr, 0, attr_sz);
>> + fd = syscall(__NR_bpf, BPF_PROG_LOAD | BPF_COMMON_ATTRS, &attr,
>> attr_sz, NULL,
>> + sizeof(struct bpf_common_attr));
>> + if (fd >= 0)
>> + close(fd);
>
> hm... close can change errno, this is fragile. If fd >= 0, something
> is wrong with our detection, just return error right away?
>
How about capture errno before closing?
err = errno;
if (fd >= 0)
close(fd);
return err = EFAULT;
Then, we can wrap all details in probe_sys_bpf_ext().
>> + return errno == EFAULT;
>> +}
>> +
>> static inline int sys_bpf(enum bpf_cmd cmd, union bpf_attr *attr,
>> unsigned int size)
>> {
>> diff --git a/tools/lib/bpf/features.c b/tools/lib/bpf/features.c
>> index b842b83e2480..d786a815f1ae 100644
>> --- a/tools/lib/bpf/features.c
>> +++ b/tools/lib/bpf/features.c
>> @@ -506,6 +506,11 @@ static int probe_kern_arg_ctx_tag(int token_fd)
>> return probe_fd(prog_fd);
>> }
>>
>> +static int probe_kern_extended_syscall(int token_fd)
>> +{
>> + return probe_sys_bpf_ext();
>> +}
>> +
>> typedef int (*feature_probe_fn)(int /* token_fd */);
>>
>> static struct kern_feature_cache feature_cache;
>> @@ -581,6 +586,9 @@ static struct kern_feature_desc {
>> [FEAT_BTF_QMARK_DATASEC] = {
>> "BTF DATASEC names starting from '?'",
>> probe_kern_btf_qmark_datasec,
>> },
>> + [FEAT_EXTENDED_SYSCALL] = {
>> + "Kernel supports extended syscall",
>> probe_kern_extended_syscall,
>
> "extended syscall" is a bit vague... We specifically detect common
> attrs support, maybe say that?
>
Ack.
I'll update it to "BPF syscall common attributes support."
>> + },
>> };
>>
>> bool feat_supported(struct kern_feature_cache *cache, enum kern_feature_id
>> feat_id)
>> diff --git a/tools/lib/bpf/libbpf_internal.h
>> b/tools/lib/bpf/libbpf_internal.h
>> index fc59b21b51b5..e2a6ef4b45ae 100644
>> --- a/tools/lib/bpf/libbpf_internal.h
>> +++ b/tools/lib/bpf/libbpf_internal.h
>> @@ -392,6 +392,8 @@ enum kern_feature_id {
>> FEAT_ARG_CTX_TAG,
>> /* Kernel supports '?' at the front of datasec names */
>> FEAT_BTF_QMARK_DATASEC,
>> + /* Kernel supports extended syscall */
>> + FEAT_EXTENDED_SYSCALL,
>
> FEAT_BPF_COMMON_ATTRS ?
>
FEAT_BPF_SYSCALL_COMMON_ATTRS seems more accurate.
Thanks,
Leon
>> __FEAT_CNT,
>> };
>>
>> @@ -757,4 +759,5 @@ int probe_fd(int fd);
>> #define SHA256_DWORD_SIZE SHA256_DIGEST_LENGTH / sizeof(__u64)
>>
>> void libbpf_sha256(const void *data, size_t len, __u8
>> out[SHA256_DIGEST_LENGTH]);
>> +int probe_sys_bpf_ext(void);
>> #endif /* __LIBBPF_LIBBPF_INTERNAL_H */
>> --
>> 2.52.0
>>
