Re: [PATCH bpf-next v4 1/2] bpf: support BPF_PROG_QUERY for progs attached to sockmap

[Yonghong Song <yhs@fb.com>]

On 11/2/21 7:23 PM, zhudi (E) wrote:
>> On 11/2/21 1:48 AM, Di Zhu wrote:
>>> Right now there is no way to query whether BPF programs are
>>> attached to a sockmap or not.
>>>
>>> we can use the standard interface in libbpf to query, such as:
>>> bpf_prog_query(mapFd, BPF_SK_SKB_STREAM_PARSER, 0, NULL, ...);
>>> the mapFd is the fd of sockmap.
>>>
>>> Signed-off-by: Di Zhu <zhudi2@huawei.com>
>>> ---
>>>    include/linux/bpf.h  |  9 +++++
>>>    kernel/bpf/syscall.c |  5 +++
>>>    net/core/sock_map.c  | 88
>> ++++++++++++++++++++++++++++++++++++++++----
>>>    3 files changed, 95 insertions(+), 7 deletions(-)
>>>
>>> diff --git a/include/linux/bpf.h b/include/linux/bpf.h
>>> index d604c8251d88..594ca91992db 100644
>>> --- a/include/linux/bpf.h
>>> +++ b/include/linux/bpf.h
>>> @@ -1961,6 +1961,9 @@ int bpf_prog_test_run_syscall(struct bpf_prog
>> *prog,
>>>    int sock_map_get_from_fd(const union bpf_attr *attr, struct bpf_prog
>> *prog);
>>>    int sock_map_prog_detach(const union bpf_attr *attr, enum
>> bpf_prog_type ptype);
>>>    int sock_map_update_elem_sys(struct bpf_map *map, void *key, void
>> *value, u64 flags);
>>> +int sockmap_bpf_prog_query(const union bpf_attr *attr,
>>> +			   union bpf_attr __user *uattr);
>>
>> All previous functions are with prefix "sock_map". Why you choose
>> a different prefix "sockmap"?
>>
> 
> Thanks for all your suggestions, I will make changes to the inappropriate code.
> 
>>> +
>>>    void sock_map_unhash(struct sock *sk);
>>>    void sock_map_close(struct sock *sk, long timeout);
>>>    #else
>>> @@ -2014,6 +2017,12 @@ static inline int
>> sock_map_update_elem_sys(struct bpf_map *map, void *key, void
>>>    {
>>>    	return -EOPNOTSUPP;
>>>    }
>>> +
>>> +static inline int sockmap_bpf_prog_query(const union bpf_attr *attr,
>>> +					 union bpf_attr __user *uattr)
>>> +{
>>> +	return -EINVAL;
>>> +}
>>>    #endif /* CONFIG_BPF_SYSCALL */
>>>    #endif /* CONFIG_NET && CONFIG_BPF_SYSCALL */
>>>
>>> diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
>>> index 4e50c0bfdb7d..17faeff8f85f 100644
>>> --- a/kernel/bpf/syscall.c
>>> +++ b/kernel/bpf/syscall.c
>>> @@ -3275,6 +3275,11 @@ static int bpf_prog_query(const union bpf_attr
>> *attr,
>>>    	case BPF_FLOW_DISSECTOR:
>>>    	case BPF_SK_LOOKUP:
>>>    		return netns_bpf_prog_query(attr, uattr);
>>> +	case BPF_SK_SKB_STREAM_PARSER:
>>> +	case BPF_SK_SKB_STREAM_VERDICT:
>>> +	case BPF_SK_MSG_VERDICT:
>>> +	case BPF_SK_SKB_VERDICT:
>>> +		return sockmap_bpf_prog_query(attr, uattr);
>>>    	default:
>>>    		return -EINVAL;
>>>    	}
>>> diff --git a/net/core/sock_map.c b/net/core/sock_map.c
>>> index e252b8ec2b85..ca65ed0004d3 100644
>>> --- a/net/core/sock_map.c
>>> +++ b/net/core/sock_map.c
>>> @@ -1412,38 +1412,50 @@ static struct sk_psock_progs
>> *sock_map_progs(struct bpf_map *map)
>>>    	return NULL;
>>>    }
>>>
>>> -static int sock_map_prog_update(struct bpf_map *map, struct bpf_prog
>> *prog,
>>> -				struct bpf_prog *old, u32 which)
>>> +static int sock_map_prog_lookup(struct bpf_map *map, struct bpf_prog
>> **pprog[],
>>
>> Can we just change "**pprog[]" to "***pprog"? In the code, you really
>> just pass the address of the decl "struct bpf_prog **pprog;" to the
>> function.
>>
>>> +				u32 which)
>>
>> Some format issue here?
> 
> 
> Format is right, passed the checkpatch script check.

Sorry about this. I guess my reply formating cheated me:

 >>> +static int sock_map_prog_lookup(struct bpf_map *map, struct bpf_prog
 >> **pprog[],
 >>> +				u32 which)

I see a larger misalignment between "struct bpf_map *map" and
"u32 which" in the reply email. But looking at original
patch, there are no issues.

> 
> 
>>
>>>    {
>>>    	struct sk_psock_progs *progs = sock_map_progs(map);
>>> -	struct bpf_prog **pprog;
>>>
>>>    	if (!progs)
>>>    		return -EOPNOTSUPP;
[...]