Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1471702 > unrolled thread

Re: [Xen-devel] [PATCH net-next] xen-netfront: avoid packet loss when ethernet header crosses page boundary

Back to article view | Back to linux.kernel

This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by below is the oldest one visible, not the original post.

Contents

  Re: [Xen-devel] [PATCH net-next] xen-netfront: avoid packet loss when ethernet header crosses page boundary Vitaly Kuznetsov <vkuznets@redhat.com> - 2016-08-29 12:30 +0200
    Re: [Xen-devel] [PATCH net-next] xen-netfront: avoid packet loss when ethernet header crosses page boundary Vitaly Kuznetsov <vkuznets@redhat.com> - 2016-09-09 15:40 +0200

#1471702 — Re: [Xen-devel] [PATCH net-next] xen-netfront: avoid packet loss when ethernet header crosses page boundary

Vitaly Kuznetsov <vkuznets@redhat.com> writes:

> David Vrabel <david.vrabel@citrix.com> writes:
>
>> On 22/08/16 16:42, Vitaly Kuznetsov wrote:
>>> 
>>> I see two ways to fix the issue:
>>> - Change the 'wire' protocol between netfront and netback to start keeping
>>>   the original SKB structure. We'll have to add a flag indicating the fact
>>>   that the particular request is a part of the original linear part and not
>>>   a frag. We'll need to know the length of the linear part to pre-allocate
>>>   memory.
>>
>> I don't think there needs to be a protocol change.  I think the check in
>> netback is bogus -- it's the total packet length that must be >
>> HLEN_ETH.  The upper layers will pull any headers from the frags as
>> needed
>
> I'm afraid this is not always true, just removing the check leads us to
> the following:
>
> [  495.442186] kernel BUG at ./include/linux/skbuff.h:1927! 
> [  495.468789] invalid opcode: 0000 [#1] SMP 

What I wanted to say here is that this test makes me think the
description of the patch I suggested is correct: an SKB can't have its
linear part shorter than ETH_HLEN as the header is being pointed directly,
upper network layers don't assemble it from frags, the check in netback
is valid.

So, how can we proceed here?

-- 
  Vitaly

[toc] | [next] | [standalone]

#1480000

Vitaly Kuznetsov <vkuznets@redhat.com> writes:

> Vitaly Kuznetsov <vkuznets@redhat.com> writes:
>
>> David Vrabel <david.vrabel@citrix.com> writes:
>>
>>> On 22/08/16 16:42, Vitaly Kuznetsov wrote:
>>>> 
>>>> I see two ways to fix the issue:
>>>> - Change the 'wire' protocol between netfront and netback to start keeping
>>>>   the original SKB structure. We'll have to add a flag indicating the fact
>>>>   that the particular request is a part of the original linear part and not
>>>>   a frag. We'll need to know the length of the linear part to pre-allocate
>>>>   memory.
>>>
>>> I don't think there needs to be a protocol change.  I think the check in
>>> netback is bogus -- it's the total packet length that must be >
>>> HLEN_ETH.  The upper layers will pull any headers from the frags as
>>> needed
>>
>> I'm afraid this is not always true, just removing the check leads us to
>> the following:
>>
>> [  495.442186] kernel BUG at ./include/linux/skbuff.h:1927! 
>> [  495.468789] invalid opcode: 0000 [#1] SMP 
>
> What I wanted to say here is that this test makes me think the
> description of the patch I suggested is correct: an SKB can't have its
> linear part shorter than ETH_HLEN as the header is being pointed directly,
> upper network layers don't assemble it from frags, the check in netback
> is valid.
>
> So, how can we proceed here?

Sorry for the second ping but I'd really like to see this moving
forward...

-- 
  Vitaly

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web