[rt2x00-users] padding and alignment

[Ivo Van Doorn]

On Fri, Aug 14, 2009 at 3:34 PM, Benoit
PAPILLAULT<benoit.papillault at free.fr> wrote:
> Ivo Van Doorn a écrit :
>> For some chipsets it means the frame will not be send out, I am not
>> aware of a performance hit.
>
> OK, i will do test this way. I will try every kind of addresses for
> SDPTR0 and SDPTR1 (be it a multiple of 4 + 0, 1, 2, 3).

Ok.

>> Well HW crypto is broken in rt2800pci, so you should always test it
>> with software crypto. And even when it is working, all drivers should work
>> with software encryption since it is important for testing purposes or when
>> frames should be encrypted which the hardware can't encrypt itself (due to
>> new ieee80211 specifications).
>
> Hummm... Using SW crypto use way too much CPU, but i know that some
> cases are not handled by the HW, so I think we use either use SW crypto
> for testing purpose and a mix of HW/SW crypto for real world and 802.11
> compliance.

Regardless, all drivers should work with SW crypto, and HW crypto only
if they support it. Currently HW crypto is disabled for rt2800usb because
it isn't working.

> The good news is that rt2800pci is properly working in WPA2 STA mode
> (I'm using it on a 2.6.30-wl based kernel since 2.6.31-rc5-wl was not
> working at all).

That's quite odd, HW crypto is broken in rt2800usb, and the same code
was used in rt2800pci. Also what additional patches do you have in rt2800pci?
So far everybody who used rt2800pci as-is in rt2x00.git reported failure
(except for users who use a special chipset on embedded machines that
don't load firmware).

Ivo