FAILED: patch "[PATCH] Revert "net: skb: introduce and use a single page frag cache"" failed to apply to 6.6-stable tree

The patch below does not apply to the 6.6-stable tree.
If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <[email protected]>.

To reproduce the conflict and resubmit, you may use the following commands:

git fetch https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/ linux-6.6.y
git checkout FETCH_HEAD
git cherry-pick -x 011b0335903832facca86cd8ed05d7d8d94c9c76
# <resolve conflicts, build, test, etc.>
git commit -s
git send-email --to '<[email protected]>' --in-reply-to '2025021859-upturned-trailside-5b21@gregkh' --subject-prefix 'PATCH 6.6.y' HEAD^..

Possible dependencies:

thanks,

greg k-h

------------------ original commit in Linus's tree ------------------

From 011b0335903832facca86cd8ed05d7d8d94c9c76 Mon Sep 17 00:00:00 2001
From: Paolo Abeni <[email protected]>
Date: Thu, 6 Feb 2025 22:28:48 +0100
Subject: [PATCH] Revert "net: skb: introduce and use a single page frag cache"

This reverts commit dbae2b062824 ("net: skb: introduce and use a single page frag cache"). The intended goal of such change was to counter a performance regression introduced by commit 3226b158e67c ("net: avoid 32 x truesize under-estimation for tiny skbs").

Unfortunately, the blamed commit introduces another regression for the virtio_net driver. Such a driver calls napi_alloc_skb() with a tiny size, so that the whole head frag could fit a 512-byte block.

The single page frag cache uses a 1K fragment for such allocation, and the additional overhead, under small UDP packets flood, makes the page allocator a bottleneck.

Thanks to commit bf9f1baa279f ("net: add dedicated kmem_cache for typical/small skb->head"), this revert does not re-introduce the original regression. Actually, in the relevant test on top of this revert, I measure a small but noticeable positive delta, just above noise level.

The revert itself required some additional mangling due to the introduction of the SKB_HEAD_ALIGN() helper and local lock infra in the affected code.

index c0a86afb85da..365f0e2098d1 100644
@@ -4115,7 +4115,6 @@ void netif_receive_skb_list(struct list_head *head);
 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
 struct sk_buff *napi_get_frags(struct napi_struct *napi);
-void napi_get_frags_check(struct napi_struct *napi);
 gro_result_t napi_gro_frags(struct napi_struct *napi);
 
 static inline void napi_free_frags(struct napi_struct *napi)

index b91658e8aedb..55e356a68db6 100644
@@ -6920,6 +6920,23 @@ netif_napi_dev_list_add(struct net_device *dev, struct napi_struct *napi)
 	list_add_rcu(&napi->dev_list, higher); /* adds after higher */
 }
 
+/* Double check that napi_get_frags() allocates skbs with
+ * skb->head being backed by slab, not a page fragment.
+ * This is to make sure bug fixed in 3226b158e67c
+ * ("net: avoid 32 x truesize under-estimation for tiny skbs")
+ * does not accidentally come back.
+ */
+static void napi_get_frags_check(struct napi_struct *napi)
+{
+	struct sk_buff *skb;
+
+	local_bh_disable();
+	skb = napi_get_frags(napi);
+	WARN_ON_ONCE(skb && skb->head_frag);
+	napi_free_frags(napi);
+	local_bh_enable();
+}
+
 void netif_napi_add_weight_locked(struct net_device *dev,
 				  struct napi_struct *napi,
 				  int (*poll)(struct napi_struct *, int),