[PATCH net-next] net: minor __alloc_skb() optimization

[PATCH net-next] net: minor __alloc_skb() optimization

[Eric Dumazet]

We can directly call __finalize_skb_around()
instead of __build_skb_around() because @size is not zero.

Signed-off-by: Eric Dumazet <edumazet@google.com>
---
 net/core/skbuff.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/net/core/skbuff.c b/net/core/skbuff.c
index 4887099e8678352a62d805e1b0be2736dc985376..77508cf7c41e829a11a988d8de3d2673ff1ff121 100644
--- a/net/core/skbuff.c
+++ b/net/core/skbuff.c
@@ -714,7 +714,7 @@ struct sk_buff *__alloc_skb(unsigned int size, gfp_t gfp_mask,
 	prefetchw(data + SKB_WITH_OVERHEAD(size));
 
 	skbuff_clear(skb);
-	__build_skb_around(skb, data, size);
+	__finalize_skb_around(skb, data, size);
 	skb->pfmemalloc = pfmemalloc;
 
 	if (flags & SKB_ALLOC_FCLONE) {
-- 
2.52.0.457.g6b5491de43-goog

Re: [PATCH net-next] net: minor __alloc_skb() optimization

[Jakub Kicinski]

On Tue, 13 Jan 2026 13:10:17 +0000 Eric Dumazet wrote:
> We can directly call __finalize_skb_around()
> instead of __build_skb_around() because @size is not zero.

FWIW I've been tempted to delete the zero check from
__build_skb_around() completely recently..
It's been a few years since we added slab_build_skb()
surely any buggy driver that's actually used would have
already hit that WARN_ONCE() and gotten reported?

Re: [PATCH net-next] net: minor __alloc_skb() optimization

[Eric Dumazet]

On Wed, Jan 14, 2026 at 2:44 AM Jakub Kicinski <kuba@kernel.org> wrote:
>
> On Tue, 13 Jan 2026 13:10:17 +0000 Eric Dumazet wrote:
> > We can directly call __finalize_skb_around()
> > instead of __build_skb_around() because @size is not zero.
>
> FWIW I've been tempted to delete the zero check from
> __build_skb_around() completely recently..
> It's been a few years since we added slab_build_skb()
> surely any buggy driver that's actually used would have
> already hit that WARN_ONCE() and gotten reported?

We could keep it for a while, WDYT of

diff --git a/net/core/skbuff.c b/net/core/skbuff.c
index 77508cf7c41e829a11a988d8de3d2673ff1ff121..ccd287ff46e91c2548483c51fa32fc6167867940
100644
--- a/net/core/skbuff.c
+++ b/net/core/skbuff.c
@@ -458,7 +458,8 @@ static void __build_skb_around(struct sk_buff
*skb, void *data,
        /* frag_size == 0 is considered deprecated now. Callers
         * using slab buffer should use slab_build_skb() instead.
         */
-       if (WARN_ONCE(size == 0, "Use slab_build_skb() instead"))
+       if (IS_ENABLED(CONFIG_DEBUG_NET) &&
+           WARN_ONCE(size == 0, "Use slab_build_skb() instead"))
                data = __slab_build_skb(data, &size);

        __finalize_skb_around(skb, data, size);

Re: [PATCH net-next] net: minor __alloc_skb() optimization

[Eric Dumazet]

On Wed, Jan 14, 2026 at 2:54 AM Eric Dumazet <edumazet@google.com> wrote:
>
> On Wed, Jan 14, 2026 at 2:44 AM Jakub Kicinski <kuba@kernel.org> wrote:
> >
> > On Tue, 13 Jan 2026 13:10:17 +0000 Eric Dumazet wrote:
> > > We can directly call __finalize_skb_around()
> > > instead of __build_skb_around() because @size is not zero.
> >
> > FWIW I've been tempted to delete the zero check from
> > __build_skb_around() completely recently..
> > It's been a few years since we added slab_build_skb()
> > surely any buggy driver that's actually used would have
> > already hit that WARN_ONCE() and gotten reported?
>
> We could keep it for a while, WDYT of
>
> diff --git a/net/core/skbuff.c b/net/core/skbuff.c
> index 77508cf7c41e829a11a988d8de3d2673ff1ff121..ccd287ff46e91c2548483c51fa32fc6167867940
> 100644
> --- a/net/core/skbuff.c
> +++ b/net/core/skbuff.c
> @@ -458,7 +458,8 @@ static void __build_skb_around(struct sk_buff
> *skb, void *data,
>         /* frag_size == 0 is considered deprecated now. Callers
>          * using slab buffer should use slab_build_skb() instead.
>          */
> -       if (WARN_ONCE(size == 0, "Use slab_build_skb() instead"))
> +       if (IS_ENABLED(CONFIG_DEBUG_NET) &&
> +           WARN_ONCE(size == 0, "Use slab_build_skb() instead"))
>                 data = __slab_build_skb(data, &size);
>
>         __finalize_skb_around(skb, data, size);


BTW, for folks using clang and not up to date kernels, backporting this commit
makes sure __build_skb_around() is at least inlined.

commit 242b872239f6a7deacbc20ab9406ea40cb738ec6
Author: Xie Yuanbin <qq570070308@gmail.com>
Date:   Sun Nov 9 16:37:15 2025 +0800

    include/linux/once_lite.h: fix judgment in WARN_ONCE with clang

Re: [PATCH net-next] net: minor __alloc_skb() optimization

[Jakub Kicinski]

On Wed, 14 Jan 2026 02:54:10 +0100 Eric Dumazet wrote:
> We could keep it for a while, WDYT of
> 
> diff --git a/net/core/skbuff.c b/net/core/skbuff.c
> index 77508cf7c41e829a11a988d8de3d2673ff1ff121..ccd287ff46e91c2548483c51fa32fc6167867940
> 100644
> --- a/net/core/skbuff.c
> +++ b/net/core/skbuff.c
> @@ -458,7 +458,8 @@ static void __build_skb_around(struct sk_buff
> *skb, void *data,
>         /* frag_size == 0 is considered deprecated now. Callers
>          * using slab buffer should use slab_build_skb() instead.
>          */
> -       if (WARN_ONCE(size == 0, "Use slab_build_skb() instead"))
> +       if (IS_ENABLED(CONFIG_DEBUG_NET) &&
> +           WARN_ONCE(size == 0, "Use slab_build_skb() instead"))
>                 data = __slab_build_skb(data, &size);

Probably not worth it. People who use relevant HW + 3 year old kernels
(I just checked, we added the warning in Dec 2022) likely don't set
DEBUG_NET either. That said looks like the warning landed in 6.2,
narrowly missing the v6.1 LTS. I suppose v6.1 may still be used by some
"enterprise-ish" distroes. I guess we should wait another year :(
Sorry for the noise :(

Re: [PATCH net-next] net: minor __alloc_skb() optimization

[patchwork-bot+netdevbpf]

Hello:

This patch was applied to netdev/net-next.git (main)
by Jakub Kicinski <kuba@kernel.org>:

On Tue, 13 Jan 2026 13:10:17 +0000 you wrote:
> We can directly call __finalize_skb_around()
> instead of __build_skb_around() because @size is not zero.
> 
> Signed-off-by: Eric Dumazet <edumazet@google.com>
> ---
>  net/core/skbuff.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)

Here is the summary with links:
  - [net-next] net: minor __alloc_skb() optimization
    https://git.kernel.org/netdev/net-next/c/2db009e4c8d6

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html