Re: [RFT] Intel 3945abg wireless driver (wpi)

Michael Butler wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Benjamin Close wrote:
>   
>> Howdy All,           I'm pleased to announce the first 'official'
>> experimental version of the wpi wireless driver and hence require your
>> help in making it become stable.
>> Expect a few things not to work (ie bg scanning, setting txpower) but in
>> general the driver should be usable in station mode (hostap is not yet
>> supported).
>>     
>
> With respect to 16k alignment .. it may be acceptable for this to fail
> at boot-time (IMHO, barely) but it usually isn't if a machine has been
> 'up' for a while.
>
> Does the UMA sub-system reliably work with 4k aligned requests? As an
> interim hack, I'd be happy to lose 12kB if it meant a 100% success rate
>  on my 2GB laptop. There are much worse things ..
>
> The only questions are, not having looked through the code recently (or
> the UMA code at all), how many buffer pools are allocated and is it
> possible to allocate a chunk of memory contiguously (for all pools) and
> assign dma tags once that chunk is 'trimmed' to an acceptable alignment?
>   
4K alignment to works reliably and I've plans to manually do the 
alignment with the 12k loss in the future. At the moment the driver 
tries the 16k alignment, if that fails it progressively increases the 
size request by 4k at a time. Generally alignment kicks in within 3 
attempts. It's a code I intend to clean up, but would prefer to fix uma 
rather than needing to work around it in the first place.

Cheers,
    Benjamin