joy/android_packages_modules_Connectivity
Watch
1
Fork
You've already forked android_packages_modules_Connectivity
 0
Code Issues Pull requests Projects Releases Wiki Activity

bpf/progs/offload.c: add comment about pmtu

Browse source 
Bug: 263884894
Flag: EXEMPT mainline
Test: TreeHugger
Signed-off-by: Maciej Żenczykowski <maze@google.com>
Change-Id: I34c8d033ee0da34fc6db4d6093ed6a10c57c0dc1
This commit is contained in:
Maciej Żenczykowski 2025-06-02 15:45:36 -07:00
commit f1f59d721d
1 changed files with 4 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files

4
bpf/progs/offload.c
View file

@ -176,6 +176,8 @@ static inline __always_inline int do_forward6(struct __sk_buff* skb,
// Handling of IPv6 overhead for incoming LRO/GRO packets
uint64_t packets = 1;
uint64_t L3_bytes = skb->len - l2_header_size;
// I don't trust v->pmtu to not be higher then it should be: it might just
// be device (instead of path/route/ipv6) mtu and that itself might be wrong...
if (L3_bytes > v->pmtu) {
if (gso_hdr_size == sizeof(struct ipv6hdr)) TC_PUNT(UNKNOWN_IPV6_GSO);
const int mss = v->pmtu - gso_hdr_size;
@ -423,6 +425,8 @@ static inline __always_inline int do_forward4_bottom(struct __sk_buff* skb,
// Handling of IPv4 overhead for incoming LRO/GRO packets
uint64_t packets = 1;
uint64_t L3_bytes = skb->len - l2_header_size;
// I don't trust v->pmtu to not be higher then it should be: it might just
// be device (instead of path/route) mtu and that itself might be wrong...
if (L3_bytes > v->pmtu) {
// we know it is IPv4 without IP options, and either TCP or UDP
const int hdr_sz = sizeof(struct iphdr) + (is_tcp ? tcph->doff * 4 : sizeof(struct udphdr));