This is a note to let you know that I've just added the patch titled net: add and use skb_unclone_keeptruesize() helper to the 5.15-stable tree which can be found at: http://www.kernel.org/git/?p=linux/kernel/git/stable/stable-queue.git;a=summary The filename of the patch is: net-add-and-use-skb_unclone_keeptruesize-helper.patch and it can be found in the queue-5.15 subdirectory. If you, or anyone else, feels it should not be added to the stable tree, please let <stable@xxxxxxxxxxxxxxx> know about it. >From c4777efa751d293e369aec464ce6875e957be255 Mon Sep 17 00:00:00 2001 From: Eric Dumazet <edumazet@xxxxxxxxxx> Date: Mon, 1 Nov 2021 17:45:55 -0700 Subject: net: add and use skb_unclone_keeptruesize() helper From: Eric Dumazet <edumazet@xxxxxxxxxx> commit c4777efa751d293e369aec464ce6875e957be255 upstream. While commit 097b9146c0e2 ("net: fix up truesize of cloned skb in skb_prepare_for_shift()") fixed immediate issues found when KFENCE was enabled/tested, there are still similar issues, when tcp_trim_head() hits KFENCE while the master skb is cloned. This happens under heavy networking TX workloads, when the TX completion might be delayed after incoming ACK. This patch fixes the WARNING in sk_stream_kill_queues when sk->sk_mem_queued/sk->sk_forward_alloc are not zero. Fixes: d3fb45f370d9 ("mm, kfence: insert KFENCE hooks for SLAB") Signed-off-by: Eric Dumazet <edumazet@xxxxxxxxxx> Acked-by: Marco Elver <elver@xxxxxxxxxx> Link: https://lore.kernel.org/r/20211102004555.1359210-1-eric.dumazet@xxxxxxxxx Signed-off-by: Jakub Kicinski <kuba@xxxxxxxxxx> Signed-off-by: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx> --- include/linux/skbuff.h | 16 ++++++++++++++++ net/core/skbuff.c | 14 +------------- net/ipv4/tcp_output.c | 6 +++--- 3 files changed, 20 insertions(+), 16 deletions(-) --- a/include/linux/skbuff.h +++ b/include/linux/skbuff.h @@ -1671,6 +1671,22 @@ static inline int skb_unclone(struct sk_ return 0; } +/* This variant of skb_unclone() makes sure skb->truesize is not changed */ +static inline int skb_unclone_keeptruesize(struct sk_buff *skb, gfp_t pri) +{ + might_sleep_if(gfpflags_allow_blocking(pri)); + + if (skb_cloned(skb)) { + unsigned int save = skb->truesize; + int res; + + res = pskb_expand_head(skb, 0, 0, pri); + skb->truesize = save; + return res; + } + return 0; +} + /** * skb_header_cloned - is the header a clone * @skb: buffer to check --- a/net/core/skbuff.c +++ b/net/core/skbuff.c @@ -3449,19 +3449,7 @@ EXPORT_SYMBOL(skb_split); */ static int skb_prepare_for_shift(struct sk_buff *skb) { - int ret = 0; - - if (skb_cloned(skb)) { - /* Save and restore truesize: pskb_expand_head() may reallocate - * memory where ksize(kmalloc(S)) != ksize(kmalloc(S)), but we - * cannot change truesize at this point. - */ - unsigned int save_truesize = skb->truesize; - - ret = pskb_expand_head(skb, 0, 0, GFP_ATOMIC); - skb->truesize = save_truesize; - } - return ret; + return skb_unclone_keeptruesize(skb, GFP_ATOMIC); } /** --- a/net/ipv4/tcp_output.c +++ b/net/ipv4/tcp_output.c @@ -1562,7 +1562,7 @@ int tcp_fragment(struct sock *sk, enum t return -ENOMEM; } - if (skb_unclone(skb, gfp)) + if (skb_unclone_keeptruesize(skb, gfp)) return -ENOMEM; /* Get a new skb... force flag on. */ @@ -1672,7 +1672,7 @@ int tcp_trim_head(struct sock *sk, struc { u32 delta_truesize; - if (skb_unclone(skb, GFP_ATOMIC)) + if (skb_unclone_keeptruesize(skb, GFP_ATOMIC)) return -ENOMEM; delta_truesize = __pskb_trim_head(skb, len); @@ -3184,7 +3184,7 @@ int __tcp_retransmit_skb(struct sock *sk cur_mss, GFP_ATOMIC)) return -ENOMEM; /* We'll try again later. */ } else { - if (skb_unclone(skb, GFP_ATOMIC)) + if (skb_unclone_keeptruesize(skb, GFP_ATOMIC)) return -ENOMEM; diff = tcp_skb_pcount(skb); Patches currently in stable-queue which might be from edumazet@xxxxxxxxxx are queue-5.15/tcp-fix-uninitialized-access-in-skb-frags-array-for-.patch queue-5.15/net-add-and-use-skb_unclone_keeptruesize-helper.patch