[PATCH][v2] powerpc/mm: using two zones for freescale 64 bit kernel

[Kumar Gala]

On Sep 20, 2012, at 5:14 AM, Xie Shaohui-B21989 wrote:

>>> On Thu, 2012-08-30 at 15:49 -0500, Kumar Gala wrote:
>>>> On Aug 24, 2012, at 5:50 AM, Shaohui Xie wrote:
>>>> 
>>>>> PowerPC platform only supports ZONE_DMA zone for 64bit kernel, so
>>>>> all the memory will be put into this zone. If the memory size is
>>>>> greater than the device's DMA capability and device uses
>>>>> dma_alloc_coherent to allocate memory, it will get an address
>>>>> which is over the device's DMA addressing, the device will fail.
>>>>> 
>>>>> So we split the memory to two zones: zone ZONE_DMA32 &
>>>>> ZONE_NORMAL, since we already allocate PCICSRBAR/PEXCSRBAR right
>>>>> below the 4G boundary (if the lowest PCI address is above 4G), so
>>>>> we constrain the DMA zone ZONE_DMA32 to 2GB, also, we clear flag
>>>>> __GFP_DMA &
>>>>> __GFP_DMA32 and set __GFP_DMA32 only if the device's dma_mask <
>>>>> total memory size. By doing this, devices which cannot DMA all the
>>>>> memory will be limited to ZONE_DMA32, but devices which can DMA
>>>>> all
>>> the memory will not be affected by this limitation.
>>>>> 
>>>>> Signed-off-by: Shaohui Xie <Shaohui.Xie at freescale.com>
>>>>> Signed-off-by: Mingkai Hu <Mingkai.hu at freescale.com>
>>>>> Signed-off-by: Chen Yuanquan <B41889 at freescale.com>
>>>>> ---
>>>>> changes for v2:
>>>>> 1. use a config option for using two zones (ZONE_DMA32 &
>>>>> ZONE_NORMAL) in freescale 64 bit kernel.
>>>>> 
>>> 
>>> There must have been a misunderstanding. I think this should be a
>>> runtime choice, possibly by the platform code. Any reason that can't be
>> done ?
>>> 
>> [S.H] Do you mean this:
>> 
>> phys_addr_t platform_dma_size (maybe a default value should be used, then
>> platform code will change it)
>> 
>> if (top_of_ram > platform_dma_size)
>> 	max_zone_pfns[ZONE_DMA] = platform_dma_size >> PAGE_SHIFT; else
>> 	max_zone_pfns[ZONE_DMA] = top_of_ram >> PAGE_SHIFT;
>> 
>> max_zone_pfns[ZONE_NORMAL] = top_of_ram >> PAGE_SHIFT;
>> 
>>> Also how does Intel do it ?
>> [S.H] below are codes in Intel:
>> 
>> 403 void __init zone_sizes_init(void)
>> 404 {
>> 405         unsigned long max_zone_pfns[MAX_NR_ZONES];
>> 406
>> 407         memset(max_zone_pfns, 0, sizeof(max_zone_pfns));
>> 408
>> 409 #ifdef CONFIG_ZONE_DMA
>> 410         max_zone_pfns[ZONE_DMA]         = MAX_DMA_PFN;
>> 411 #endif
>> 412 #ifdef CONFIG_ZONE_DMA32
>> 413         max_zone_pfns[ZONE_DMA32]       = MAX_DMA32_PFN;
>> 414 #endif
>> 415         max_zone_pfns[ZONE_NORMAL]      = max_low_pfn;
>> 416 #ifdef CONFIG_HIGHMEM
>> 417         max_zone_pfns[ZONE_HIGHMEM]     = max_pfn;
>> 418 #endif
>> 419
>> 
>> For x86_64, there is no CONFIG_HIGHMEM, so there will be three zones:
>> ZONE_DMA/ZONE_DMA32/ZONE_NORMAL.
>> 
> [S.H] Hello, Ben,
> 
> I have some questions, though I'm still expecting your comments.
> PPC does not have ZONE_DMA32 by default, if we want to use it, we need to add "config ZONE_DMA32" in Kconfig first.
> If setting multiple zones without ZONE_DMA, kmalloc in "include/linux/slab_def.h" will fail if it uses flag GFP_DMA.
> For the runtime choice in 64-bit kernel, what exactly multiple zones should be used?
> "ZONE_DMA & ZONE_NORMAL" or "ZONE_DMA & ZONE_DMA32 & ZONE_NORMAL"?
> Then what the size should be set for them respectively?
> 
> Please comment, Thanks!

I think Ben is saying that Kconfig would enable ZONE_DMA32 for all PPC64, but make it runtime/per platform how we setup the zone's such that either ZONE_DMA32 is set to MAX_DMA32_PFN or it set to same value as ZONE_DMA.

However that's just a guess.

- k