Skip to contentSkip to conversation

bpf: support BPF_PROG_QUERY for BPF_FLOW_DISSECTOR attach_type

PATCH Superseded Stanislav Fomichev 3 participants 4 messages Apr 24, 2019

Subsystems (5)

v2 of 5 Superseded Latest is v5 →

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/2 bpf: support BPF_PROG_QUERY for BPF_FLOW_DISSECTOR attach_type

target_fd is target namespace. If there is a flow dissector BPF program attached to that namespace, its (single) id is returned.

v2:
* don't sleep in rcu critical section (Jakub Kicinski)
* check input prog_cnt (exit early)

---
 include/linux/skbuff.h    |  8 ++++++++
 kernel/bpf/syscall.c      |  2 ++
 net/core/flow_dissector.c | 37 +++++++++++++++++++++++++++++++++++++
 3 files changed, 47 insertions(+)
include/linux/skbuff.h +8 -0
index 998256c2820b..9f499c0ac53b 100644
@@ -1258,11 +1258,19 @@ void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
 			     unsigned int key_count);
 
 #ifdef CONFIG_NET
+int skb_flow_dissector_prog_query(const union bpf_attr *attr,
+				  union bpf_attr __user *uattr);
 int skb_flow_dissector_bpf_prog_attach(const union bpf_attr *attr,
 				       struct bpf_prog *prog);
 
 int skb_flow_dissector_bpf_prog_detach(const union bpf_attr *attr);
 #else
+static int skb_flow_dissector_prog_query(const union bpf_attr *attr,
+					 union bpf_attr __user *uattr)
+{
+	return -EOPNOTSUPP;
+}
+
 static inline int skb_flow_dissector_bpf_prog_attach(const union bpf_attr *attr,
 						     struct bpf_prog *prog)
 {
kernel/bpf/syscall.c +2 -0
index 92c9b8a32b50..b0de49598341 100644
@@ -2009,6 +2009,8 @@ static int bpf_prog_query(const union bpf_attr *attr,
 		break;
 	case BPF_LIRC_MODE2:
 		return lirc_prog_query(attr, uattr);
+	case BPF_FLOW_DISSECTOR:
+		return skb_flow_dissector_prog_query(attr, uattr);
 	default:
 		return -EINVAL;
 	}
net/core/flow_dissector.c +37 -0
index fac712cee9d5..12b92c7d1eae 100644
@@ -65,6 +65,43 @@ void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
 }
 EXPORT_SYMBOL(skb_flow_dissector_init);
 
+int skb_flow_dissector_prog_query(const union bpf_attr *attr,
+				  union bpf_attr __user *uattr)
+{
+	__u32 __user *prog_ids = u64_to_user_ptr(attr->query.prog_ids);
+	u32 prog_id, prog_cnt = 0, flags = 0;
+	struct bpf_prog *attached;
+	struct net *net;
+
+	if (attr->query.query_flags)
+		return -EINVAL;
+
+	net = get_net_ns_by_fd(attr->query.target_fd);
+	if (IS_ERR(net))
+		return PTR_ERR(net);
+
+	rcu_read_lock();
+	attached = rcu_dereference(net->flow_dissector_prog);
+	if (attached) {
+		prog_cnt = 1;
+		prog_id = attached->aux->id;
+	}
+	rcu_read_unlock();
+
+	if (copy_to_user(&uattr->query.attach_flags, &flags, sizeof(flags)))
+		return -EFAULT;
+	if (copy_to_user(&uattr->query.prog_cnt, &prog_cnt, sizeof(prog_cnt)))
+		return -EFAULT;
+
+	if (!attr->query.prog_cnt || !prog_ids || !prog_cnt)
+		return 0;
+
+	if (copy_to_user(prog_ids, &prog_id, sizeof(u32)))
+		return -EFAULT;
+
+	return 0;
+}
+
 int skb_flow_dissector_bpf_prog_attach(const union bpf_attr *attr,
 				       struct bpf_prog *prog)
 {
Signature · 1 line
2.21.0.593.g511ec345e18-goog
#1 lore raw

PATCH v2 2/2 bpftool: show flow_dissector attachment status

Right now there is no way to query whether BPF flow_dissector program is attached to a network namespace or not. In previous commit, I added support for querying that info, show it when doing bpftool net:

$ bpftool prog loadall ./bpf_flow.o \
	/sys/fs/bpf/flow type flow_dissector \
	pinmaps /sys/fs/bpf/flow
$ bpftool prog
3: flow_dissector  name _dissect  tag 8c9e917b513dd5cc  gpl
        loaded_at 2019-04-23T16:14:48-0700  uid 0
        xlated 656B  jited 461B  memlock 4096B  map_ids 1,2
        btf_id 1
...
$ bpftool net -j
[{"xdp":[],"tc":[],"flow_dissector":[]}]
$ bpftool prog attach pinned \
	/sys/fs/bpf/flow/flow_dissector flow_dissector
$ bpftool net -j
[{"xdp":[],"tc":[],"flow_dissector":["id":3]}]

Doesn't show up in a different net namespace:

$ ip netns add test
$ ip netns exec test bpftool net -j
[{"xdp":[],"tc":[],"flow_dissector":[]}]

Non-json output:

$ bpftool net
xdp:

tc:

flow_dissector:
id 3

v2:
* initialization order (Jakub Kicinski)
* clear errno for batch mode (Quentin Monnet)

---
 tools/bpf/bpftool/net.c | 54 +++++++++++++++++++++++++++++++++++++++++
 1 file changed, 54 insertions(+)
tools/bpf/bpftool/net.c +54 -0
index db0e7de49d49..67e99c56bc88 100644
@@ -3,6 +3,7 @@
 
 #define _GNU_SOURCE
 #include <errno.h>
+#include <fcntl.h>
 #include <stdlib.h>
 #include <string.h>
 #include <unistd.h>
@@ -12,6 +13,8 @@
 #include <linux/rtnetlink.h>
 #include <linux/tc_act/tc_bpf.h>
 #include <sys/socket.h>
+#include <sys/stat.h>
+#include <sys/types.h>
 
 #include <bpf.h>
 #include <nlattr.h>
@@ -48,6 +51,10 @@ struct bpf_filter_t {
 	int		ifindex;
 };
 
+struct bpf_attach_info {
+	__u32 flow_dissector_id;
+};
+
 static int dump_link_nlmsg(void *cookie, void *msg, struct nlattr **tb)
 {
 	struct bpf_netdev_t *netinfo = cookie;
@@ -180,8 +187,45 @@ static int show_dev_tc_bpf(int sock, unsigned int nl_pid,
 	return 0;
 }
 
+static int query_flow_dissector(struct bpf_attach_info *attach_info)
+{
+	__u32 attach_flags;
+	__u32 prog_ids[1];
+	__u32 prog_cnt;
+	int err;
+	int fd;
+
+	fd = open("/proc/self/ns/net", O_RDONLY);
+	if (fd < 0) {
+		p_err("can't open /proc/self/ns/net: %d",
+		      strerror(errno));
+		return -1;
+	}
+	prog_cnt = ARRAY_SIZE(prog_ids);
+	err = bpf_prog_query(fd, BPF_FLOW_DISSECTOR, 0,
+			     &attach_flags, prog_ids, &prog_cnt);
+	close(fd);
+	if (err) {
+		if (errno == EINVAL) {
+			/* Older kernel's don't support querying
+			 * flow dissector programs.
+			 */
+			errno = 0;
+			return 0;
+		}
+		p_err("can't query prog: %s", strerror(errno));
+		return -1;
+	}
+
+	if (prog_cnt == 1)
+		attach_info->flow_dissector_id = prog_ids[0];
+
+	return 0;
+}
+
 static int do_show(int argc, char **argv)
 {
+	struct bpf_attach_info attach_info = {};
 	int i, sock, ret, filter_idx = -1;
 	struct bpf_netdev_t dev_array;
 	unsigned int nl_pid;
@@ -199,6 +243,10 @@ static int do_show(int argc, char **argv)
 		usage();
 	}
 
+	ret = query_flow_dissector(&attach_info);
+	if (ret)
+		return -1;
+
 	sock = libbpf_netlink_open(&nl_pid);
 	if (sock < 0) {
 		fprintf(stderr, "failed to open netlink sock\n");
@@ -227,6 +275,12 @@ static int do_show(int argc, char **argv)
 		}
 		NET_END_ARRAY("\n");
 	}
+
+	NET_START_ARRAY("flow_dissector", "%s:\n");
+	if (attach_info.flow_dissector_id > 0)
+		NET_DUMP_UINT("id", "id %u", attach_info.flow_dissector_id);
+	NET_END_ARRAY("\n");
+
 	NET_END_OBJECT;
 	if (json_output)
 		jsonw_end_array(json_wtr);
Signature · 1 line
2.21.0.593.g511ec345e18-goog
#2 lore raw
In reply to Stanislav Fomichev
Quoted 8 lines · Stanislav Fomichev

target_fd is target namespace. If there is a flow dissector BPF program attached to that namespace, its (single) id is returned.

v2:
* don't sleep in rcu critical section (Jakub Kicinski)
* check input prog_cnt (exit early)

[...]

Quoted 3 lines
include/linux/skbuff.h +3 -0
+int skb_flow_dissector_prog_query(const union bpf_attr *attr,
+                                 union bpf_attr __user *uattr)
+{

[...]

net/core/flow_dissector.c void skb_flow_dissector_init(struct flow_dissector *flow_dissector, · line 81

Quoted 3 lines
net/core/flow_dissector.c +3 -0
+       net = get_net_ns_by_fd(attr->query.target_fd);
+       if (IS_ERR(net))
+               return PTR_ERR(net);

At this point, you're holding a refcounted reference to net. It looks like that reference is never dropped?

Quoted 11 lines
net/core/flow_dissector.c +11 -0
+
+       rcu_read_lock();
+       attached = rcu_dereference(net->flow_dissector_prog);
+       if (attached) {
+               prog_cnt = 1;
+               prog_id = attached->aux->id;
+       }
+       rcu_read_unlock();
+
+       if (copy_to_user(&uattr->query.attach_flags, &flags, sizeof(flags)))
+               return -EFAULT;

[...]

Quoted 1 line

+}

#3 lore raw
Quoted 14 lines · Jann Horn
Quoted 8 lines · Stanislav Fomichev

target_fd is target namespace. If there is a flow dissector BPF program attached to that namespace, its (single) id is returned.

v2:
* don't sleep in rcu critical section (Jakub Kicinski)
* check input prog_cnt (exit early)

[...]

Quoted 3 lines
include/linux/skbuff.h +3 -0
+int skb_flow_dissector_prog_query(const union bpf_attr *attr,
+                                 union bpf_attr __user *uattr)
+{

[...]

Quoted 6 lines

net/core/flow_dissector.c void skb_flow_dissector_init(struct flow_dissector *flow_dissector, · line 81

Quoted 3 lines
net/core/flow_dissector.c +3 -0
+       net = get_net_ns_by_fd(attr->query.target_fd);
+       if (IS_ERR(net))
+               return PTR_ERR(net);

At this point, you're holding a refcounted reference to net. It looks like that reference is never dropped?

Ah, indeed, put_net is missing, thanks!

Quoted 13 lines
Quoted 11 lines
net/core/flow_dissector.c +11 -0
+
+       rcu_read_lock();
+       attached = rcu_dereference(net->flow_dissector_prog);
+       if (attached) {
+               prog_cnt = 1;
+               prog_id = attached->aux->id;
+       }
+       rcu_read_unlock();
+
+       if (copy_to_user(&uattr->query.attach_flags, &flags, sizeof(flags)))
+               return -EFAULT;

[...]

Quoted 1 line

+}