All of lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH net] tuntap: properly align skb->head before building skb
@ 2017-10-26 12:15 Jason Wang
  2017-10-26 14:11 ` Eric Dumazet
  0 siblings, 1 reply; 3+ messages in thread
From: Jason Wang @ 2017-10-26 12:15 UTC (permalink / raw)
  To: netdev, linux-kernel
  Cc: dotweiba, Jason Wang, Eric Dumazet, Willem de Bruijn,
	Dmitry Vyukov, Mark Rutland

An unaligned alloc_frag->offset caused by previous allocation will
result an unaligned skb->head. This will lead unaligned
skb_shared_info and then unaligned dataref which requires to be
aligned for accessing on some architecture. Fix this by aligning
alloc_frag->offset before the frag refilling.

Fixes: 0bbd7dad34f8 ("tun: make tun_build_skb() thread safe")
Cc: Eric Dumazet <edumazet@google.com>
Cc: Willem de Bruijn <willemdebruijn.kernel@gmail.com>
Cc: Wei Wei <dotweiba@gmail.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Reported-by: Wei Wei <dotweiba@gmail.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
---
- The patch is needed for -stable.
- Wei, can you try this patch to see if it solves your issue?
---
 drivers/net/tun.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/net/tun.c b/drivers/net/tun.c
index b9973fb..60e44f2 100644
--- a/drivers/net/tun.c
+++ b/drivers/net/tun.c
@@ -1286,6 +1286,7 @@ static struct sk_buff *tun_build_skb(struct tun_struct *tun,
 	buflen += SKB_DATA_ALIGN(len + pad);
 	rcu_read_unlock();
 
+	alloc_frag->offset = ALIGN((u64)alloc_frag->offset, TUN_RX_PAD);
 	if (unlikely(!skb_page_frag_refill(buflen, alloc_frag, GFP_KERNEL)))
 		return ERR_PTR(-ENOMEM);
 
-- 
2.7.4

^ permalink raw reply related	[flat|nested] 3+ messages in thread

* Re: [PATCH net] tuntap: properly align skb->head before building skb
  2017-10-26 12:15 [PATCH net] tuntap: properly align skb->head before building skb Jason Wang
@ 2017-10-26 14:11 ` Eric Dumazet
  2017-10-27  3:03   ` Jason Wang
  0 siblings, 1 reply; 3+ messages in thread
From: Eric Dumazet @ 2017-10-26 14:11 UTC (permalink / raw)
  To: Jason Wang
  Cc: netdev, LKML, Wei Wei, Willem de Bruijn, Dmitry Vyukov, Mark Rutland

On Thu, Oct 26, 2017 at 5:15 AM, Jason Wang <jasowang@redhat.com> wrote:
> An unaligned alloc_frag->offset caused by previous allocation will
> result an unaligned skb->head. This will lead unaligned
> skb_shared_info and then unaligned dataref which requires to be
> aligned for accessing on some architecture. Fix this by aligning
> alloc_frag->offset before the frag refilling.
>
> Fixes: 0bbd7dad34f8 ("tun: make tun_build_skb() thread safe")
> Cc: Eric Dumazet <edumazet@google.com>
> Cc: Willem de Bruijn <willemdebruijn.kernel@gmail.com>
> Cc: Wei Wei <dotweiba@gmail.com>
> Cc: Dmitry Vyukov <dvyukov@google.com>
> Cc: Mark Rutland <mark.rutland@arm.com>
> Reported-by: Wei Wei <dotweiba@gmail.com>
> Signed-off-by: Jason Wang <jasowang@redhat.com>
> ---
> - The patch is needed for -stable.
> - Wei, can you try this patch to see if it solves your issue?
> ---
>  drivers/net/tun.c | 1 +
>  1 file changed, 1 insertion(+)
>
> diff --git a/drivers/net/tun.c b/drivers/net/tun.c
> index b9973fb..60e44f2 100644
> --- a/drivers/net/tun.c
> +++ b/drivers/net/tun.c
> @@ -1286,6 +1286,7 @@ static struct sk_buff *tun_build_skb(struct tun_struct *tun,
>         buflen += SKB_DATA_ALIGN(len + pad);
>         rcu_read_unlock();
>
> +       alloc_frag->offset = ALIGN((u64)alloc_frag->offset, TUN_RX_PAD);

You have to align to one cache line (SMP_CACHE_BYTES), or SKB_DATA_ALIGN(1)

Then eventually use skb_reserve() for NET_IP_ALIGN, but I guess it is
already done.

^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: [PATCH net] tuntap: properly align skb->head before building skb
  2017-10-26 14:11 ` Eric Dumazet
@ 2017-10-27  3:03   ` Jason Wang
  0 siblings, 0 replies; 3+ messages in thread
From: Jason Wang @ 2017-10-27  3:03 UTC (permalink / raw)
  To: Eric Dumazet
  Cc: netdev, LKML, Wei Wei, Willem de Bruijn, Dmitry Vyukov, Mark Rutland



On 2017年10月26日 22:11, Eric Dumazet wrote:
> On Thu, Oct 26, 2017 at 5:15 AM, Jason Wang <jasowang@redhat.com> wrote:
>> An unaligned alloc_frag->offset caused by previous allocation will
>> result an unaligned skb->head. This will lead unaligned
>> skb_shared_info and then unaligned dataref which requires to be
>> aligned for accessing on some architecture. Fix this by aligning
>> alloc_frag->offset before the frag refilling.
>>
>> Fixes: 0bbd7dad34f8 ("tun: make tun_build_skb() thread safe")
>> Cc: Eric Dumazet <edumazet@google.com>
>> Cc: Willem de Bruijn <willemdebruijn.kernel@gmail.com>
>> Cc: Wei Wei <dotweiba@gmail.com>
>> Cc: Dmitry Vyukov <dvyukov@google.com>
>> Cc: Mark Rutland <mark.rutland@arm.com>
>> Reported-by: Wei Wei <dotweiba@gmail.com>
>> Signed-off-by: Jason Wang <jasowang@redhat.com>
>> ---
>> - The patch is needed for -stable.
>> - Wei, can you try this patch to see if it solves your issue?
>> ---
>>   drivers/net/tun.c | 1 +
>>   1 file changed, 1 insertion(+)
>>
>> diff --git a/drivers/net/tun.c b/drivers/net/tun.c
>> index b9973fb..60e44f2 100644
>> --- a/drivers/net/tun.c
>> +++ b/drivers/net/tun.c
>> @@ -1286,6 +1286,7 @@ static struct sk_buff *tun_build_skb(struct tun_struct *tun,
>>          buflen += SKB_DATA_ALIGN(len + pad);
>>          rcu_read_unlock();
>>
>> +       alloc_frag->offset = ALIGN((u64)alloc_frag->offset, TUN_RX_PAD);
> You have to align to one cache line (SMP_CACHE_BYTES), or SKB_DATA_ALIGN(1)

Oh right.

> Then eventually use skb_reserve() for NET_IP_ALIGN, but I guess it is
> already done.

Yes.

Thanks

^ permalink raw reply	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2017-10-27  3:04 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2017-10-26 12:15 [PATCH net] tuntap: properly align skb->head before building skb Jason Wang
2017-10-26 14:11 ` Eric Dumazet
2017-10-27  3:03   ` Jason Wang

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.