Re: [PATCH 3/3 v2] netxen: qlogic ethernet : Fix endian bug.

[santosh prasad nayak]

Subject and comments are provided below for review.
This is for review not a formal one. Please let me know
if any more changes are required so that I can include it in
a formal patch.

--------------------------------------------------------------------------------------------------------

Subject: [PATCH 3/3 v2] netxen: qlogic ethernet : Fix endian bug.

Change the datatype of "ip_addr" to __be32 as 'ip' should be in
big endian format.

Adapter needs "ip address" in big endian format stored at lower 32bit
of req.word[1]. netxen_config_ipaddr() now receives 'ip' in big endian
format. To satisfy adapter's need, use memcpy() to copy byte by byte
of 'ip' into lower 32bit of req.word[1].

Mac address and serial number of adapter need to be in little endian format.
Change the data type of the related  variables to __le32 / __le64 or cast it
explicitly to __le32 / __le64 depending upon the requirement.

---------------------------------------------------------------------------------------------------------------


Regards
Santosh



On Mon, Mar 12, 2012 at 12:22 PM, santosh prasad nayak
<santoshprasadnayak@xxxxxxxxx> wrote:
> Sure David.
> I will resend it with proper comment.
>
> Regards
> Santosh
>
> On Mon, Mar 12, 2012 at 12:18 PM, David Miller <davem@xxxxxxxxxxxxx> wrote:
>> From: santosh nayak <santoshprasadnayak@xxxxxxxxx>
>> Date: Mon, 12 Mar 2012 12:08:23 +0530
>>
>>> From: Santosh Nayak <santoshprasadnayak@xxxxxxxxx>
>>>
>>> Fix endian bug.
>>>
>>> Signed-off-by: Santosh Nayak <santoshprasadnayak@xxxxxxxxx>
>>
>> This is a very non-trivial patch, therefore the terse commit
>> message and the lack of any comments in this code is unacceptable.
>>
>> This commit message is not only terse, it's wrong.  You're not
>> fixing an endian bug, you're fixing several of them.
>>
>> Explain everything in your commit message.
>>
>>>       req.words[0] = cpu_to_le64(cmd);
>>> -     req.words[1] = cpu_to_le64(ip);
>>> +     memcpy(&req.words[1], &ip, sizeof(u32));
>>
>> Nobody is going to read that and have any idea why it's correct unless
>> they happen to be able to discover the long thread you guys had this
>> past week working out how to do this change correctly.
>>
>> Putting a comment here for the rest of us mere mortals is therefore
>> absolutely required.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/