Re: [PATCH net-next] xen-netback: improve guest-receive-side flow control

[David Vrabel <david.vrabel@citrix.com>]

On 28/11/13 16:33, Paul Durrant wrote:
>> -----Original Message-----
>> From: Wei Liu [mailto:wei.liu2@citrix.com]
>> Sent: 28 November 2013 16:27
>> To: Paul Durrant
>> Cc: Wei Liu; xen-devel@lists.xen.org; Ian Campbell; David Vrabel
>> Subject: Re: [PATCH net-next] xen-netback: improve guest-receive-side flow
>> control
> [snip]
> 
>>>> I'm afraid this starting point is not correct. Consider you have a SKB
>>>> with very large linear buffer, you might need more than 2 slots to fit
>>>> that in, right?
>>>>
>>>
>>> I was wondering about that. Would a roundup of mtu to the nearest 4k
>> boundary be sufficient?
>>>
>>
>> I think we can reuse the logic in xenvif_gop_skb, there's a snippet used
>> to calculate how many slots are necessary for linear area. We just need
>> to avoid pulling requrests from the ring.
>>
> 
> Actually, DIV_ROUND(skb_headlen(skb), PAGE_SIZE) should give a cheap
> upper bound, shouldn't it? I don't want the code here to get too
> complex.

Make sure it does the right thing if the head is not page aligned etc.

David