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

["zhudi (E)" <zhudi2@huawei.com>]

> 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.


> 
> >   {
> >   	struct sk_psock_progs *progs = sock_map_progs(map);
> > -	struct bpf_prog **pprog;
> >
> >   	if (!progs)
> >   		return -EOPNOTSUPP;
> >
> >   	switch (which) {
> >   	case BPF_SK_MSG_VERDICT:
> > -		pprog = &progs->msg_parser;
> > +		*pprog = &progs->msg_parser;
> >   		break;
> >   #if IS_ENABLED(CONFIG_BPF_STREAM_PARSER)
> >   	case BPF_SK_SKB_STREAM_PARSER:
> > -		pprog = &progs->stream_parser;
> > +		*pprog = &progs->stream_parser;
> >   		break;
> >   #endif
> >   	case BPF_SK_SKB_STREAM_VERDICT:
> >   		if (progs->skb_verdict)
> >   			return -EBUSY;
> > -		pprog = &progs->stream_verdict;
> > +		*pprog = &progs->stream_verdict;
> >   		break;
> >   	case BPF_SK_SKB_VERDICT:
> >   		if (progs->stream_verdict)
> >   			return -EBUSY;
> > -		pprog = &progs->skb_verdict;
> > +		*pprog = &progs->skb_verdict;
> >   		break;
> >   	default:
> >   		return -EOPNOTSUPP;
> >   	}
> >
> > +	return 0;
> > +}
> > +
> > +static int sock_map_prog_update(struct bpf_map *map, struct bpf_prog
> *prog,
> > +				struct bpf_prog *old, u32 which)
> 
> Some format issue here?
> 
> > +{
> > +	struct bpf_prog **pprog;
> > +	int ret;
> > +
> > +	ret = sock_map_prog_lookup(map, &pprog, which);
> > +	if (ret)
> > +		return ret;
> > +
> >   	if (old)
> >   		return psock_replace_prog(pprog, prog, old);
> >
> > @@ -1451,6 +1463,68 @@ static int sock_map_prog_update(struct
> bpf_map *map, struct bpf_prog *prog,
> >   	return 0;
> >   }
> >
> > +int sockmap_bpf_prog_query(const union bpf_attr *attr,
> > +			   union bpf_attr __user *uattr)
> 
> Format issue here?
> 
> > +{
> > +	__u32 __user *prog_ids = u64_to_user_ptr(attr->query.prog_ids);
> 
> Typically we use u32 in the kernel code. But I know there are __u32
> usage as well, esp. with __user attributes. I put a comment here just
> in case that somebody else has a different opinion.
> 
> > +	u32 prog_cnt = 0, flags = 0;
> > +	u32 ufd = attr->target_fd;
> 
> You can merge the above u32 together.
> 
> > +	struct bpf_prog **pprog;
> > +	struct bpf_prog *prog;
> > +	struct bpf_map *map;
> > +	struct fd f;
> > +	int ret;
> > +	u32 id = 0;
> 
> to maintain reverse christmas tree?
> 
> > +
> > +	if (attr->query.query_flags)
> > +		return -EINVAL;
> > +
> > +	f = fdget(ufd);
> > +	map = __bpf_map_get(f);
> > +	if (IS_ERR(map))
> > +		return PTR_ERR(map);
> > +
> > +	rcu_read_lock();
> > +
> > +	ret = sock_map_prog_lookup(map, &pprog, attr->query.attach_type);
> > +	if (ret)
> > +		goto end;
> > +
> > +	prog = *pprog;
> > +	prog_cnt = (!prog) ? 0 : 1;
> > +
> > +	if (!attr->query.prog_cnt || !prog_ids || !prog_cnt)
> > +		goto end;
> > +
> > +	prog = bpf_prog_inc_not_zero(prog);
> 
> Could you explain why we need bpf_prog_inc_not_zero here?
> We are inside rcu_read_lock/unlock region. We got a program
> from *pprog. If this program is not NULL, this program should
> not disappear since we are in rcu read lock region, right?
> Maybe I missed something, it would be good you can explain
> the scenario you try to pretect here.


bpf_prog_inc_not_zero() return a failure indicating that the program is
being released and prog->aux->id will be set to 0.
Yes, it is just ok for accessing prog->aux->id directly.


> > +	if (IS_ERR(prog)) {
> > +		ret = PTR_ERR(prog);
> > +		goto end;
> > +	}
> > +	id = prog->aux->id;
> > +	bpf_prog_put(prog);
> > +
> > +end:
> > +	rcu_read_unlock();
> > +
> > +	if (copy_to_user(&uattr->query.attach_flags, &flags, sizeof(flags))) {
> > +		ret = -EFAULT;
> > +		goto err;
> > +	}
> > +	if (id != 0 && copy_to_user(prog_ids, &id, sizeof(u32))) {
> > +		ret = -EFAULT;
> > +		goto err;
> > +	}
> > +	if (copy_to_user(&uattr->query.prog_cnt, &prog_cnt, sizeof(prog_cnt))) {
> > +		ret = -EFAULT;
> > +		goto err;
> > +	}
> 
> You can do
> 
> 	if (copy_to_user(&uattr->query.attach_flags, &flags, sizeof(flags)) ||
> 	    (id != 0 && copy_to_user(prog_ids, &id, sizeof(u32))) ||
> 	    copy_to_user(&uattr->query.prog_cnt, &prog_cnt, sizeof(prog_cnt)))
> 		ret = -EFAULT;
> 
> to make code a little bit concise.
> 
> > +
> > +err:
> > +	fdput(f);
> > +	return ret;
> > +}
> > +
> >   static void sock_map_unlink(struct sock *sk, struct sk_psock_link *link)
> >   {
> >   	switch (link->map->map_type) {
> >