Skip to contentSkip to conversation

potential sockmap memleak and proc stats fix

PATCH Superseded John Fastabend 2 participants 7 messages Jul 2, 2021 – Jul 5, 2021

Subsystems (5)

v2 of 4 Superseded Latest is v4 →

lore
Share & tools
lore

John Fastabend

author maintainer
#0 lore raw

v2 cover potential sockmap memleak and proc stats fix

While investigating a memleak in sockmap I found these two issues. Patch 1 found doing code review, I wasn't able to get KASAN to trigger a memleak here, but should be necessary. Patch 2 fixes proc stats so when we use sockstats for debugging we get correct values.

The fix for observered memleak will come after these, but requires some more discussion and potentially patch revert so I'll try to get the set here going now.

John Fastabend (2):
bpf, sockmap: fix potential memory leak on unlikely error case bpf, sockmap: sk_prot needs inuse_idx set for proc stats

 net/core/skmsg.c    | 4 +++-
 net/core/sock_map.c | 9 +++++++++
 2 files changed, 12 insertions(+), 1 deletion(-)
Signature · 1 line
2.25.1
author maintainer
#1 lore raw

PATCH v2 1/2 bpf, sockmap: fix potential msg memory leak

If skb_linearize is needed and fails we could leak a msg on the error handling. To fix ensure we kfree the msg block before returning error. Found during code review.

Fixes 4363023d2668e ("bpf, sockmap: Avoid failures from skb_to_sgvec when skb has frag_list")
---
 net/core/skmsg.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
net/core/skmsg.c +3 -1
index 9b6160a191f8..22603289c2b2 100644
@@ -505,8 +505,10 @@ static int sk_psock_skb_ingress_enqueue(struct sk_buff *skb,
 	 * drop the skb. We need to linearize the skb so that the mapping
 	 * in skb_to_sgvec can not error.
 	 */
-	if (skb_linearize(skb))
+	if (skb_linearize(skb)) {
+		kfree(msg);
 		return -EAGAIN;
+	}
 	num_sge = skb_to_sgvec(skb, msg->sg.data, 0, skb->len);
 	if (unlikely(num_sge < 0)) {
 		kfree(msg);
Signature · 1 line
2.25.1
#2 lore raw

Re: [PATCH v2 bpf 1/2] bpf, sockmap: fix potential msg memory leak

Quoted 17 lines · John Fastabend

If skb_linearize is needed and fails we could leak a msg on the error handling. To fix ensure we kfree the msg block before returning error. Found during code review.

Fixes 4363023d2668e ("bpf, sockmap: Avoid failures from skb_to_sgvec when skb has frag_list")
---
 net/core/skmsg.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
net/core/skmsg.c +0 -0
index 9b6160a191f8..22603289c2b2 100644
@@ -505,8 +505,10 @@ static int sk_psock_skb_ingress_enqueue(struct sk_buff *skb,
         * drop the skb. We need to linearize the skb so that the mapping

net/core/skmsg.c static int sk_psock_skb_ingress_enqueue(struct sk_buff *skb, · line 514

Quoted 10 lines
net/core/skmsg.c +3 -1
         * in skb_to_sgvec can not error.
         */
-       if (skb_linearize(skb))
+       if (skb_linearize(skb)) {
+               kfree(msg);
                return -EAGAIN;
+       }
        num_sge = skb_to_sgvec(skb, msg->sg.data, 0, skb->len);
        if (unlikely(num_sge < 0)) {
                kfree(msg);

I think it is better to let whoever allocates msg free it, IOW, let sk_psock_skb_ingress_enqueue()'s callers handle its failure.

Thanks.

author maintainer
#3 lore raw

Re: [PATCH v2 bpf 1/2] bpf, sockmap: fix potential msg memory leak

Quoted 23 lines · Cong Wang
Quoted 22 lines · John Fastabend

If skb_linearize is needed and fails we could leak a msg on the error handling. To fix ensure we kfree the msg block before returning error. Found during code review.

Fixes 4363023d2668e ("bpf, sockmap: Avoid failures from skb_to_sgvec when skb has frag_list")
---
 net/core/skmsg.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
net/core/skmsg.c +2 -1
index 9b6160a191f8..22603289c2b2 100644
@@ -505,8 +505,10 @@ static int sk_psock_skb_ingress_enqueue(struct sk_buff *skb,
         * drop the skb. We need to linearize the skb so that the mapping
         * in skb_to_sgvec can not error.
         */
-       if (skb_linearize(skb))
+       if (skb_linearize(skb)) {
+               kfree(msg);
Quoted 8 lines

net/core/skmsg.c static int sk_psock_skb_ingress_enqueue(struct sk_buff *skb, · line 514

Quoted 5 lines
net/core/skmsg.c +1 -0
                return -EAGAIN;
+       }
        num_sge = skb_to_sgvec(skb, msg->sg.data, 0, skb->len);
        if (unlikely(num_sge < 0)) {
                kfree(msg);

I think it is better to let whoever allocates msg free it, IOW, let sk_psock_skb_ingress_enqueue()'s callers handle its failure.

Sure, although we already have the one kfree(msg) below. Anyways I'll just move these back a bit. Agree it is slightly nicer.

Thanks.

author maintainer
#4 lore raw

PATCH v2 2/2 bpf, sockmap: sk_prot needs inuse_idx for proc stats

In reply to John Fastabend

Proc socket stats use sk_prot->inuse_idx value to record inuse sock stats. We currently do not set this correctly from sockmap side. The result is reading sock stats '/proc/net/sockstat' gives incorrect values. The socket counter is incremented correctly, but because we don't set the counter correctly when we replace sk_prot we may omit the decrement.

Fixes 604326b41a6fb ("bpf, sockmap: convert to generic sk_msg interface")
---
 net/core/sock_map.c | 9 +++++++++
 1 file changed, 9 insertions(+)
net/core/sock_map.c +9 -0
index 60decd6420ca..016ea5460f8f 100644
@@ -222,6 +222,9 @@ static int sock_map_link(struct bpf_map *map, struct sock *sk)
 	struct bpf_prog *msg_parser = NULL;
 	struct sk_psock *psock;
 	int ret;
+#ifdef CONFIG_PROC_FS
+	int idx;
+#endif
 
 	/* Only sockets we can redirect into/from in BPF need to hold
 	 * refs to parser/verdict progs and have their sk_data_ready
@@ -293,9 +296,15 @@ static int sock_map_link(struct bpf_map *map, struct sock *sk)
 	if (msg_parser)
 		psock_set_prog(&psock->progs.msg_parser, msg_parser);
 
+#ifdef CONFIG_PROC_FS
+	idx = sk->sk_prot->inuse_idx;
+#endif
 	ret = sock_map_init_proto(sk, psock);
 	if (ret < 0)
 		goto out_drop;
+#ifdef CONFIG_PROC_FS
+	sk->sk_prot->inuse_idx = idx;
+#endif
 
 	write_lock_bh(&sk->sk_callback_lock);
 	if (stream_parser && stream_verdict && !psock->saved_data_ready) {
Signature · 1 line
2.25.1
#5 lore raw

Re: [PATCH v2 bpf 2/2] bpf, sockmap: sk_prot needs inuse_idx for proc stats

Quoted 30 lines · John Fastabend

Proc socket stats use sk_prot->inuse_idx value to record inuse sock stats. We currently do not set this correctly from sockmap side. The result is reading sock stats '/proc/net/sockstat' gives incorrect values. The socket counter is incremented correctly, but because we don't set the counter correctly when we replace sk_prot we may omit the decrement.

Fixes 604326b41a6fb ("bpf, sockmap: convert to generic sk_msg interface")
---
 net/core/sock_map.c | 9 +++++++++
 1 file changed, 9 insertions(+)
net/core/sock_map.c +3 -0
index 60decd6420ca..016ea5460f8f 100644
@@ -222,6 +222,9 @@ static int sock_map_link(struct bpf_map *map, struct sock *sk)
        struct bpf_prog *msg_parser = NULL;
        struct sk_psock *psock;
        int ret;
+#ifdef CONFIG_PROC_FS
+       int idx;
+#endif
 
        /* Only sockets we can redirect into/from in BPF need to hold
         * refs to parser/verdict progs and have their sk_data_ready
@@ -293,9 +296,15 @@ static int sock_map_link(struct bpf_map *map, struct sock *sk)
        if (msg_parser)
                psock_set_prog(&psock->progs.msg_parser, msg_parser);

net/core/sock_map.c static int sock_map_link(struct bpf_map *map, struct sock *sk) · line 307

Quoted 10 lines
net/core/sock_map.c +6 -0
 
+#ifdef CONFIG_PROC_FS
+       idx = sk->sk_prot->inuse_idx;
+#endif
        ret = sock_map_init_proto(sk, psock);
        if (ret < 0)
                goto out_drop;
+#ifdef CONFIG_PROC_FS
+       sk->sk_prot->inuse_idx = idx;
+#endif

I think it is better to put these into sock_map_init_proto() so that sock_map_link() does not need to worry about the sk_prot details.

Thanks.

author maintainer
#6 lore raw

Re: [PATCH v2 bpf 2/2] bpf, sockmap: sk_prot needs inuse_idx for proc stats

Quoted 46 lines · Cong Wang
Quoted 30 lines · John Fastabend

Proc socket stats use sk_prot->inuse_idx value to record inuse sock stats. We currently do not set this correctly from sockmap side. The result is reading sock stats '/proc/net/sockstat' gives incorrect values. The socket counter is incremented correctly, but because we don't set the counter correctly when we replace sk_prot we may omit the decrement.

Fixes 604326b41a6fb ("bpf, sockmap: convert to generic sk_msg interface")
---
 net/core/sock_map.c | 9 +++++++++
 1 file changed, 9 insertions(+)
net/core/sock_map.c +3 -0
index 60decd6420ca..016ea5460f8f 100644
@@ -222,6 +222,9 @@ static int sock_map_link(struct bpf_map *map, struct sock *sk)
        struct bpf_prog *msg_parser = NULL;
        struct sk_psock *psock;
        int ret;
+#ifdef CONFIG_PROC_FS
+       int idx;
+#endif
 
        /* Only sockets we can redirect into/from in BPF need to hold
         * refs to parser/verdict progs and have their sk_data_ready
@@ -293,9 +296,15 @@ static int sock_map_link(struct bpf_map *map, struct sock *sk)
        if (msg_parser)
                psock_set_prog(&psock->progs.msg_parser, msg_parser);

net/core/sock_map.c static int sock_map_link(struct bpf_map *map, struct sock *sk) · line 307

Quoted 10 lines
net/core/sock_map.c +6 -0
 
+#ifdef CONFIG_PROC_FS
+       idx = sk->sk_prot->inuse_idx;
+#endif
        ret = sock_map_init_proto(sk, psock);
        if (ret < 0)
                goto out_drop;
+#ifdef CONFIG_PROC_FS
+       sk->sk_prot->inuse_idx = idx;
+#endif

I think it is better to put these into sock_map_init_proto() so that sock_map_link() does not need to worry about the sk_prot details.

Quoted 1 line

Thanks.

Sure, that is fine.