Re: ECN target bug report

[Patrick McHardy <kaber@trash.net>]

Hi,

Andrea Rossato wrote:

> Patrick McHardy wrote:
>
>> the first attached patch fixes the issue you reported (verified), 
>> altough i'm not entirely sure why ;).
>> the second one is untested but probably couldn't hurt neither.
>
>
> great patrick! faster then light!
>
> BTW, your second patch seems to imply that byte order is just a matter 
> of taste ;)
>
> iph and tcph should give little endian int on i386 arch, so tons 
> should be needed when writing to the sk_buff. but in computer 
> programming there are misteries a lawyar will never be able to grasp...

I thought calculating checksum for x and htons(x) should give me same 
result, obiously that is wrong (which i learned after trying it out).
For the first patch: *tcpflags are bitfields (position declared 
dependant of endianess), so no htons here.
For the second patch, it's working with or without, although i think the 
correct way is with the patch:

u_int16_t diffs[2];
diffs[0] = htons(iph->tos) ^ 0xFFFF;
...
diffs[1] = htons(iph->tos);
...

iph->tos is of type u_int8_t. so no htons is necessary.

Bye,
Patrick