Re: Problem with ata layer in 2.6.24

[Mark Lord <liml@rtr.ca>]

rgheck wrote:
> Mark Lord wrote:
>> rgheck wrote:
>>> Alan Cox wrote:
>>>>> not one problem but lots---is sufficiently widespread that a Mini 
>>>>> HOWTO, say, would be really welcome and, I'm guessing, widely used.
>>>>>     
>>>>
>>>> We don't see very many libata problems at the distro level and they for
>>>> the most part boil down to
>>>>
>>>> - sata_nv with >4GB of RAM, knowing being worked on, no old IDE driver
>>>> anyway
>>>>   
>>> Is this >4GB or >=4GB? I've seen contradictory reports, and I've got 
>>> 4GB.
>> ..
>>
>> For all practical purposes, most memory over 3GB (or sometimes even 2GB)
>> on a 32-bit x86 system is treated as >4GB by the motherboard.
>>
>> Because it's not the amount of *memory* that matters so much,
>> but rather the amount of *used address space*.  Video cards,
>> PCI devices, other motherboard resources etc.. can all subtract
>> from the available address space, leaving much less than 4GB
>> for your RAM.
> 
> Right. So it looks like I do have this issue, though I haven't seen any 
> actual problems on 24. Is there a known workaround?
..

For now, the workaround is to not enable the RAM above 4GB.
Your kernel .config file should therefore have these two lines:

CONFIG_HIGHMEM4G=y
# CONFIG_HIGHMEM64G is not set

Later, once the issue is fixed at the driver level (soon),
you can get your high memory back again by enabling CONFIG_HIGHMEM64G,
though this will cost a few percent of performance in the extra
page table overhead it creates.

Cheers