Hi Eric,
On 09/23/2016 10:19 AM, Eric Nelson wrote:
> Thanks Eric,
>
> On 09/23/2016 09:54 AM, Eric Dumazet wrote:
>> On Fri, Sep 23, 2016 at 9:43 AM, Eric Nelson <e...@nelint.com> wrote:
>>>
>>> Hello all,
>>>
>>> We're seeing alignment issues from the ethernet stack on an i.MX6UL board:
>>>
>>>
>
> <snip>
>
>>>
>>> - id = ntohl(*(__be32 *)&iph->id);
>>> - flush = (u16)((ntohl(*(__be32 *)iph) ^ skb_gro_len(skb)) | (id & ~IP_DF));
>>> - id >>= 16;
>>> + id = ntohs(*(__be16 *)&iph->id);
>>> + frag = ntohs(*(__be16 *)&iph->frag_off);
>>> + flush = (u16)((ntohl(*(__be32 *)iph) ^ skb_gro_len(skb)) | (frag &
>>> ~IP_DF));
>>>
>>> for (p = *head; p; p = p->next) {
>>> struct iphdr *iph2;
>>>
>>
>> This solves nothing, because a few lines after you'll have yet another
>> unaligned access :
>>
>
> Oddly, it does prevent the vast majority (90%+) of the alignment errors.
>
> I believe this is because the compiler is generating an ldm instruction
> when the ntohl() call is used, but I'm stumped about why these aren't
> generating faults:
>
>> ((__force u32)iph->saddr ^ (__force u32)iph2->saddr) |
>> ((__force u32)iph->daddr ^ (__force u32)iph2->daddr)) {
>>
>> So you might have one less problematic access, out of hundreds of them
>> all over the places.
>>
>> Really the problem is that whole stack depends on the assumption that
>> IP headers are aligned on arches that care
>> (ie where NET_IP_ALIGN == 2)
>>
>> If your build does have NET_IP_ALIGN = 2 and you get a fault here, it
>> might be because of a buggy driver.
>>
>
> NET_IP_ALIGN is set to 2.
>
>> The other known case is some GRE encapsulations that break the
>> assumption, and this is discussed somewhere else.
>>
> I don't think that's the case.
>
> # CONFIG_IPV6_GRE is not set
>
> Hmm... Instrumenting the kernel, it seems that iphdr **is** aligned on
> a 4-byte boundary.
>
No. That was wrong.
The iphdr is aligned at offsets of 14 from the ethernet frame, which itself
is longword aligned.
I mistakenly tested before the call to skb_gro_header_slow(), when
iph was NULL.
After putting a test in the right place, I'm seeing an address of
888a364e for the first un-aligned packet.
Since the hardware requires longword alignment for its' DMA transfers,
aligning the IP header will require a memcpy, right?
You hinted at a solution in this post:
http://www.spinics.net/lists/netdev/msg213166.html
Are you aware of another driver doing this that could be used as
a reference?
Please advise,
Eric
