Currently, hash_item() processes the 6-byte Ethernet address and the
2-byte session ID byte-wise to compute a hash.

Optimize this by using 16-bit word operations: XOR three 16-bit words
from the Ethernet address and the 16-bit session ID, then fold the
result. This reduces the total number of loads and XORs. The Ethernet
addresses in a skb and struct pppoe_addr are both 2-byte aligned, so the
u16 pointer cast is safe.

Signed-off-by: Qingfang Deng <qingfang.deng@linux.dev>
---
 drivers/net/ppp/pppoe.c | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

diff --git a/drivers/net/ppp/pppoe.c b/drivers/net/ppp/pppoe.c
index d546a7af0d54..e2e70628958b 100644
--- a/drivers/net/ppp/pppoe.c
+++ b/drivers/net/ppp/pppoe.c
@@ -136,15 +136,15 @@ static inline int cmp_addr(struct pppoe_addr *a, __be16 sid, char *addr)
 #error 8 must be a multiple of PPPOE_HASH_BITS
 #endif
 
-static int hash_item(__be16 sid, unsigned char *addr)
+static u8 hash_item(__be16 sid, const u8 addr[ETH_ALEN])
 {
-	unsigned char hash = 0;
+	const u16 *addr16 = (const u16 *)addr;
 	unsigned int i;
+	u16 hash16;
+	u8 hash;
 
-	for (i = 0; i < ETH_ALEN; i++)
-		hash ^= addr[i];
-	for (i = 0; i < sizeof(sid_t) * 8; i += 8)
-		hash ^= (__force __u32)sid >> i;
+	hash16 = addr16[0] ^ addr16[1] ^ addr16[2] ^ (__force u16)sid;
+	hash = (hash16 >> 8) ^ hash16;
 	for (i = 8; (i >>= 1) >= PPPOE_HASH_BITS;)
 		hash ^= hash >> i;
 
-- 
2.43.0

On Sun, Apr 12, 2026 at 8:52 PM Qingfang Deng <qingfang.deng@linux.dev> wrote:
>
> Currently, hash_item() processes the 6-byte Ethernet address and the
> 2-byte session ID byte-wise to compute a hash.
>
> Optimize this by using 16-bit word operations: XOR three 16-bit words
> from the Ethernet address and the 16-bit session ID, then fold the
> result. This reduces the total number of loads and XORs. The Ethernet
> addresses in a skb and struct pppoe_addr are both 2-byte aligned, so the
> u16 pointer cast is safe.
>
> Signed-off-by: Qingfang Deng <qingfang.deng@linux.dev>

net-next is closed.

https://lore.kernel.org/netdev/20260412142250.131bf997@kernel.org/

Also I would suggest using hash32(hash, PPPOE_HASH_BITS)

April 13, 2026 at 4:42 PM, Eric Dumazet wrote:

> 
> On Sun, Apr 12, 2026 at 8:52 PM Qingfang Deng <qingfang.deng@linux.dev> wrote:
> 
> net-next is closed.
> 
> https://lore.kernel.org/netdev/20260412142250.131bf997@kernel.org/
> 
> Also I would suggest using hash32(hash, PPPOE_HASH_BITS)

Thanks for the info, but I would like to keep the same algorithm.