Skip to contentSkip to conversation

bpf: only test gso type on gso packets

PATCH In review Willem de Bruijn 3 participants 3 messages Mar 6, 2019 – Mar 7, 2019

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 bpf: only test gso type on gso packets

From: Willem de Bruijn <[email protected]>

BPF can adjust gso only for tcp bytestreams. Fail on other gso types.

But only on gso packets. It does not touch this field if !gso_size.

Fixes b90efd225874 ("bpf: only adjust gso_size on bytestream protocols")

---

Stupid bug on my part. Found only when adding tests for the feature. Will try to upstream those once bpf-next opens.

On a related note, also working on a flag BPF_F_ADJ_ROOM_FIXED_GSO that will allow reenabling this field for UDP (and possibly avoiding the expensive skb_cow for the TCP common case).

---
 include/linux/skbuff.h | 4 ++--
 net/core/filter.c      | 8 ++++----
 2 files changed, 6 insertions(+), 6 deletions(-)
include/linux/skbuff.h +2 -2
index 27beb549ffbe1..f32f32407dc43 100644
@@ -4232,10 +4232,10 @@ static inline bool skb_is_gso_sctp(const struct sk_buff *skb)
 	return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP;
 }
 
+/* Note: Should be called only if skb_is_gso(skb) is true */
 static inline bool skb_is_gso_tcp(const struct sk_buff *skb)
 {
-	return skb_is_gso(skb) &&
-	       skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6);
+	return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6);
 }
 
 static inline void skb_gso_reset(struct sk_buff *skb)
net/core/filter.c +4 -4
index 5ceba98069d46..f274620945ff0 100644
@@ -2804,7 +2804,7 @@ static int bpf_skb_proto_4_to_6(struct sk_buff *skb)
 	u32 off = skb_mac_header_len(skb);
 	int ret;
 
-	if (!skb_is_gso_tcp(skb))
+	if (skb_is_gso(skb) && !skb_is_gso_tcp(skb))
 		return -ENOTSUPP;
 
 	ret = skb_cow(skb, len_diff);
@@ -2845,7 +2845,7 @@ static int bpf_skb_proto_6_to_4(struct sk_buff *skb)
 	u32 off = skb_mac_header_len(skb);
 	int ret;
 
-	if (!skb_is_gso_tcp(skb))
+	if (skb_is_gso(skb) && !skb_is_gso_tcp(skb))
 		return -ENOTSUPP;
 
 	ret = skb_unclone(skb, GFP_ATOMIC);
@@ -2970,7 +2970,7 @@ static int bpf_skb_net_grow(struct sk_buff *skb, u32 len_diff)
 	u32 off = skb_mac_header_len(skb) + bpf_skb_net_base_len(skb);
 	int ret;
 
-	if (!skb_is_gso_tcp(skb))
+	if (skb_is_gso(skb) && !skb_is_gso_tcp(skb))
 		return -ENOTSUPP;
 
 	ret = skb_cow(skb, len_diff);
@@ -2999,7 +2999,7 @@ static int bpf_skb_net_shrink(struct sk_buff *skb, u32 len_diff)
 	u32 off = skb_mac_header_len(skb) + bpf_skb_net_base_len(skb);
 	int ret;
 
-	if (!skb_is_gso_tcp(skb))
+	if (skb_is_gso(skb) && !skb_is_gso_tcp(skb))
 		return -ENOTSUPP;
 
 	ret = skb_unclone(skb, GFP_ATOMIC);
Signature · 1 line
2.21.0.352.gf09ad66450-goog
Acked-by Yonghong Song on v1 1/1
#1 lore raw
Quoted 6 lines · Willem de Bruijn

From: Willem de Bruijn <[email protected]>

BPF can adjust gso only for tcp bytestreams. Fail on other gso types.

But only on gso packets. It does not touch this field if !gso_size.

Could you specify which field you are referring here? gso_type, gso_segs, gso_size?

Quoted 3 lines
Fixes b90efd225874 ("bpf: only adjust gso_size on bytestream protocols")

Other than the above commit message suggestion,

maintainer
#2 lore raw
In reply to Willem de Bruijn
Quoted 8 lines · Willem de Bruijn

From: Willem de Bruijn <[email protected]>

BPF can adjust gso only for tcp bytestreams. Fail on other gso types.

But only on gso packets. It does not touch this field if !gso_size.

Fixes b90efd225874 ("bpf: only adjust gso_size on bytestream protocols")

Good catch, applied, thanks!

Quoted 6 lines

Stupid bug on my part. Found only when adding tests for the feature. Will try to upstream those once bpf-next opens.

On a related note, also working on a flag BPF_F_ADJ_ROOM_FIXED_GSO that will allow reenabling this field for UDP (and possibly avoiding the expensive skb_cow for the TCP common case).

Nice, looking forward!