From: Willem de Bruijn <will...@google.com> skb_segment by default transfers allocated wmem from the gso skb to the tail of the segment list. This underreports real truesize of the list, especially if the tail might be dropped.
Similar to tcp_gso_segment, update wmem_alloc with the aggregate list truesize and make each segment responsible for its own share by setting skb->destructor. Clear gso_skb->destructor prior to calling skb_segment to skip the default assignment to tail. Signed-off-by: Willem de Bruijn <will...@google.com> --- net/ipv4/udp_offload.c | 20 ++++++++++++++++++-- 1 file changed, 18 insertions(+), 2 deletions(-) diff --git a/net/ipv4/udp_offload.c b/net/ipv4/udp_offload.c index a3244768e45f..acfef71459f8 100644 --- a/net/ipv4/udp_offload.c +++ b/net/ipv4/udp_offload.c @@ -192,7 +192,9 @@ struct sk_buff *__udp_gso_segment(struct sk_buff *gso_skb, unsigned int mss, __sum16 check) { struct udphdr *uh = udp_hdr(gso_skb); - struct sk_buff *segs; + struct sock *sk = gso_skb->sk; + struct sk_buff *segs, *seg; + unsigned int sum_truesize = 0; unsigned int hdrlen; if (gso_skb->len <= sizeof(*uh) + mss) @@ -203,9 +205,23 @@ struct sk_buff *__udp_gso_segment(struct sk_buff *gso_skb, skb_pull(gso_skb, sizeof(*uh)); hdrlen = gso_skb->data - skb_mac_header(gso_skb); + /* clear destructor to avoid skb_segment assigning it to tail */ + WARN_ON_ONCE(gso_skb->destructor != sock_wfree); + gso_skb->destructor = NULL; + segs = skb_segment(gso_skb, features); - if (unlikely(IS_ERR_OR_NULL(segs))) + if (unlikely(IS_ERR_OR_NULL(segs))) { + gso_skb->destructor = sock_wfree; return segs; + } + + for (seg = segs; seg; seg = seg->next) { + seg->destructor = sock_wfree; + seg->sk = sk; + sum_truesize += seg->truesize; + } + + refcount_add(sum_truesize - gso_skb->truesize, &sk->sk_wmem_alloc); /* If last packet is not full, fix up its header */ if (segs->prev->len != hdrlen + mss) { -- 2.17.0.441.gb46fe60e1d-goog