[PATCH v4] tun: avoid high-order page allocation for packet header

[PATCH v4] tun: avoid high-order page allocation for packet header

[Tahsin Erdogan]

When gso.hdr_len is zero and a packet is transmitted via write() or
writev(), all payload is treated as header which requires a contiguous
memory allocation. This allocation request is harder to satisfy, and may
even fail if there is enough fragmentation.

Note that sendmsg() code path limits the linear copy length, so this change
makes write()/writev() and sendmsg() paths more consistent.

Signed-off-by: Tahsin Erdogan <trdgn@amazon.com>
---
v4: updated commit message address comments from Willem
v3: rebase to latest net-next
v2: replace linear == 0 with !linear
v1: https://lore.kernel.org/all/20230726030936.1587269-1-trdgn@amazon.com/
 drivers/net/tun.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/drivers/net/tun.c b/drivers/net/tun.c
index 973b2fc74de3..62106464f1b9 100644
--- a/drivers/net/tun.c
+++ b/drivers/net/tun.c
@@ -1523,7 +1523,7 @@ static struct sk_buff *tun_alloc_skb(struct tun_file *tfile,
 	int err;
 
 	/* Under a page?  Don't bother with paged skb. */
-	if (prepad + len < PAGE_SIZE || !linear)
+	if (prepad + len < PAGE_SIZE)
 		linear = len;
 
 	if (len - linear > MAX_SKB_FRAGS * (PAGE_SIZE << PAGE_ALLOC_COSTLY_ORDER))
@@ -1840,6 +1840,9 @@ static ssize_t tun_get_user(struct tun_struct *tun, struct tun_file *tfile,
 			 */
 			zerocopy = false;
 		} else {
+			if (!linear)
+				linear = min_t(size_t, good_linear, copylen);
+
 			skb = tun_alloc_skb(tfile, align, copylen, linear,
 					    noblock);
 		}
-- 
2.41.0

Re: [PATCH v4] tun: avoid high-order page allocation for packet header

[Jason Wang]

On Thu, Aug 10, 2023 at 12:48 AM Tahsin Erdogan <trdgn@amazon.com> wrote:
>
> When gso.hdr_len is zero and a packet is transmitted via write() or
> writev(), all payload is treated as header which requires a contiguous
> memory allocation. This allocation request is harder to satisfy, and may
> even fail if there is enough fragmentation.
>
> Note that sendmsg() code path limits the linear copy length, so this change
> makes write()/writev() and sendmsg() paths more consistent.
>
> Signed-off-by: Tahsin Erdogan <trdgn@amazon.com>
> ---
> v4: updated commit message address comments from Willem
> v3: rebase to latest net-next
> v2: replace linear == 0 with !linear
> v1: https://lore.kernel.org/all/20230726030936.1587269-1-trdgn@amazon.com/

Acked-by: Jason Wang <jasowang@redhat.com>

Thanks

Re: [PATCH v4] tun: avoid high-order page allocation for packet header

[Eric Dumazet]

On Wed, Aug 9, 2023 at 6:48 PM Tahsin Erdogan <trdgn@amazon.com> wrote:
>
> When gso.hdr_len is zero and a packet is transmitted via write() or
> writev(), all payload is treated as header which requires a contiguous
> memory allocation. This allocation request is harder to satisfy, and may
> even fail if there is enough fragmentation.
>
> Note that sendmsg() code path limits the linear copy length, so this change
> makes write()/writev() and sendmsg() paths more consistent.
>
> Signed-off-by: Tahsin Erdogan <trdgn@amazon.com>

Reviewed-by: Eric Dumazet <edumazet@google.com>

Thanks.

Re: [PATCH v4] tun: avoid high-order page allocation for packet header

[Willem de Bruijn]

Tahsin Erdogan wrote:
> When gso.hdr_len is zero and a packet is transmitted via write() or
> writev(), all payload is treated as header which requires a contiguous
> memory allocation. This allocation request is harder to satisfy, and may
> even fail if there is enough fragmentation.
> 
> Note that sendmsg() code path limits the linear copy length, so this change
> makes write()/writev() and sendmsg() paths more consistent.
> 
> Signed-off-by: Tahsin Erdogan <trdgn@amazon.com>

Reviewed-by: Willem de Bruijn <willemb@google.com>

Re: [PATCH v4] tun: avoid high-order page allocation for packet header

[patchwork-bot+netdevbpf]

Hello:

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

On Wed, 9 Aug 2023 09:47:52 -0700 you wrote:
> When gso.hdr_len is zero and a packet is transmitted via write() or
> writev(), all payload is treated as header which requires a contiguous
> memory allocation. This allocation request is harder to satisfy, and may
> even fail if there is enough fragmentation.
> 
> Note that sendmsg() code path limits the linear copy length, so this change
> makes write()/writev() and sendmsg() paths more consistent.
> 
> [...]

Here is the summary with links:
  - [v4] tun: avoid high-order page allocation for packet header
    https://git.kernel.org/netdev/net-next/c/6231e47b6fad

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