[U-Boot] [PATCH v4 3/6] mcx: Disable DCACHE since USB EHCI is enabled

[Ilya Yanok]

Hi Tom,

04.07.2012 00:43, Tom Rini wrote:
> On 07/03/2012 01:13 PM, Ilya Yanok wrote:
>> Dear Marek,
>>
>> 30.06.2012 23:28, Marek Vasut wrote:
>>>> Not exactly. It never worked (at least on my systems) with D-Cache
>>>> enabled. But at least we had a choice of run-time disabled dcache. With
>>>> the recent changes we have to disable cache support at compile time.
>>> I see what you're after. But do you consider runtime disabling the
>>> cache is the
>>> way to go or it's a way of hiding bugs?
>> Both ;) And now we are going to hide even more bugs with compile-time
>> disabling :(
> Does someone wish to argue we should disable USB support instead on
> these platforms?  I don't see anyone arguing "I have time to fix this
> for v2012.07".

I just looked at the code more carefully and it seems that most of the 
upper layers are in much better shape than I thought. So I think we 
should just extend your 2/6 patch to fix both address and size for 
structs QH and qtd and don't mess with buffer at all: if we got 
unaligned buffer -- it's definetely upper layer bug so we should produce 
some noise in this case. As I said upper layers seems to be in good 
shape so hopefully there won't be too much noise.

Hm, probably we should put buffer invalidation under 
if(dcache_enabled()) to leave run-time cache disabling as rescue option 
for broken upper-layer code..

I'm working on the patch now and hopefully will post it this night.

Regards, Ilya.