Re: [PATCH v3 0/5] optimize memblock_next_valid_pfn and early_pfn_valid

[Jia He <hejianet@gmail.com>]

On 3/28/2018 8:30 AM, Wei Yang Wrote:
> On Tue, Mar 27, 2018 at 03:15:08PM +0800, Jia He wrote:
>>
>> On 3/27/2018 9:02 AM, Wei Yang Wrote:
>>> On Sun, Mar 25, 2018 at 08:02:14PM -0700, Jia He wrote:
>>>> Commit b92df1de5d28 ("mm: page_alloc: skip over regions of invalid pfns
>>>> where possible") tried to optimize the loop in memmap_init_zone(). But
>>>> there is still some room for improvement.
>>>>
>>>> Patch 1 remain the memblock_next_valid_pfn when CONFIG_HAVE_ARCH_PFN_VALID
>>>>          is enabled
>>>> Patch 2 optimizes the memblock_next_valid_pfn()
>>>> Patch 3~5 optimizes the early_pfn_valid(), I have to split it into parts
>>>>          because the changes are located across subsystems.
>>>>
>>>> I tested the pfn loop process in memmap_init(), the same as before.
>>>> As for the performance improvement, after this set, I can see the time
>>>> overhead of memmap_init() is reduced from 41313 us to 24345 us in my
>>>> armv8a server(QDF2400 with 96G memory).
>>>>
>>>> Attached the memblock region information in my server.
>>>> [   86.956758] Zone ranges:
>>>> [   86.959452]   DMA      [mem 0x0000000000200000-0x00000000ffffffff]
>>>> [   86.966041]   Normal   [mem 0x0000000100000000-0x00000017ffffffff]
>>>> [   86.972631] Movable zone start for each node
>>>> [   86.977179] Early memory node ranges
>>>> [   86.980985]   node   0: [mem 0x0000000000200000-0x000000000021ffff]
>>>> [   86.987666]   node   0: [mem 0x0000000000820000-0x000000000307ffff]
>>>> [   86.994348]   node   0: [mem 0x0000000003080000-0x000000000308ffff]
>>>> [   87.001029]   node   0: [mem 0x0000000003090000-0x00000000031fffff]
>>>> [   87.007710]   node   0: [mem 0x0000000003200000-0x00000000033fffff]
>>>> [   87.014392]   node   0: [mem 0x0000000003410000-0x000000000563ffff]
>>>> [   87.021073]   node   0: [mem 0x0000000005640000-0x000000000567ffff]
>>>> [   87.027754]   node   0: [mem 0x0000000005680000-0x00000000056dffff]
>>>> [   87.034435]   node   0: [mem 0x00000000056e0000-0x00000000086fffff]
>>>> [   87.041117]   node   0: [mem 0x0000000008700000-0x000000000871ffff]
>>>> [   87.047798]   node   0: [mem 0x0000000008720000-0x000000000894ffff]
>>>> [   87.054479]   node   0: [mem 0x0000000008950000-0x0000000008baffff]
>>>> [   87.061161]   node   0: [mem 0x0000000008bb0000-0x0000000008bcffff]
>>>> [   87.067842]   node   0: [mem 0x0000000008bd0000-0x0000000008c4ffff]
>>>> [   87.074524]   node   0: [mem 0x0000000008c50000-0x0000000008e2ffff]
>>>> [   87.081205]   node   0: [mem 0x0000000008e30000-0x0000000008e4ffff]
>>>> [   87.087886]   node   0: [mem 0x0000000008e50000-0x0000000008fcffff]
>>>> [   87.094568]   node   0: [mem 0x0000000008fd0000-0x000000000910ffff]
>>>> [   87.101249]   node   0: [mem 0x0000000009110000-0x00000000092effff]
>>>> [   87.107930]   node   0: [mem 0x00000000092f0000-0x000000000930ffff]
>>>> [   87.114612]   node   0: [mem 0x0000000009310000-0x000000000963ffff]
>>>> [   87.121293]   node   0: [mem 0x0000000009640000-0x000000000e61ffff]
>>>> [   87.127975]   node   0: [mem 0x000000000e620000-0x000000000e64ffff]
>>>> [   87.134657]   node   0: [mem 0x000000000e650000-0x000000000fffffff]
>>>> [   87.141338]   node   0: [mem 0x0000000010800000-0x0000000017feffff]
>>>> [   87.148019]   node   0: [mem 0x000000001c000000-0x000000001c00ffff]
>>>> [   87.154701]   node   0: [mem 0x000000001c010000-0x000000001c7fffff]
>>>> [   87.161383]   node   0: [mem 0x000000001c810000-0x000000007efbffff]
>>>> [   87.168064]   node   0: [mem 0x000000007efc0000-0x000000007efdffff]
>>>> [   87.174746]   node   0: [mem 0x000000007efe0000-0x000000007efeffff]
>>>> [   87.181427]   node   0: [mem 0x000000007eff0000-0x000000007effffff]
>>>> [   87.188108]   node   0: [mem 0x000000007f000000-0x00000017ffffffff]
>>> Hi, Jia
>>>
>>> I haven't taken a deep look into your code, just one curious question on your
>>> memory layout.
>>>
>>> The log above is printed out in free_area_init_nodes(), which iterates on
>>> memblock.memory and prints them. If I am not wrong, memory regions added to
>>> memblock.memory are ordered and merged if possible.
>>>
>>> While from your log, I see many regions could be merged but are isolated. For
>>> example, the last two region:
>>>
>>>     node   0: [mem 0x000000007eff0000-0x000000007effffff]
>>>     node   0: [mem 0x000000007f000000-0x00000017ffffffff]
>>>
>>> So I am curious why they are isolated instead of combined to one.
>>>
>>> >From the code, the possible reason is the region's flag differs from each
>>> other. If you have time, would you mind taking a look into this?
>>>
>> Hi Wei
>> I thought these 2 have different flags
>> [A A A  0.000000] idx=30,region [7eff0000:10000]flag=4A A A A  <--- aka
>> MEMBLOCK_NOMAP
>> [A A A  0.000000]A A  nodeA A  0: [mem 0x000000007eff0000-0x000000007effffff]
>> [A A A  0.000000] idx=31,region [7f000000:81000000]flag=0 <--- aka MEMBLOCK_NONE
>> [A A A  0.000000]A A  nodeA A  0: [mem 0x000000007f000000-0x00000017ffffffff]
> Thanks.
>
> Hmm, I am not that familiar with those flags, while they look like to indicate
> the physical capability of this range.
>
> 	MEMBLOCK_NONE		no special
> 	MEMBLOCK_HOTPLUG	hotplug-able
> 	MEMBLOCK_MIRROR		high reliable
> 	MEMBLOCK_NOMAP		no direct map
>
> While these flags are not there when they are first added into the memory
> region. When you look at the memblock_add_range(), the last parameter passed
> is always 0. This means current several separated ranges reflect the physical
> memory capability layout.
>
> Then, why this layout is so scattered? As you can see several ranges are less
> than 1M.
>
> If, just my assumption, we could merge some of them, we could have a better
> performance. Less ranges, less searching time.
Thanks for your suggestions, Wei
Need further digging and will consider to improve it in another patchset.

-- 
Cheers,
Jia