2015-11-20 08:17:45

by Xishi Qiu

Subject: [RFC] mm: direct mapping count in /proc/meminfo is error

I find the direct mapping count in /proc/meminfo is error.
The value should be equal to the size of init_memory_mapping which
showed in boot log.

I add some print to show direct_pages_count[] immediately after
init_memory_mapping(). The reason is that we double counting.

Here is the log(kernel v4.4):
...
[ 0.000000] init_memory_mapping: [mem 0x00000000-0x000fffff] // called from "init_memory_mapping(0, ISA_END_ADDRESS);"
[ 0.000000] [mem 0x00000000-0x000fffff] page 4k
[ 0.000000] BRK [0x01ebf000, 0x01ebffff] PGTABLE
[ 0.000000] BRK [0x01ec0000, 0x01ec0fff] PGTABLE
[ 0.000000] BRK [0x01ec1000, 0x01ec1fff] PGTABLE
[ 0.000000] init_memory_mapping: [mem 0xc3fe00000-0xc3fffffff] // called from "memory_map_top_down(ISA_END_ADDRESS, end);"
[ 0.000000] [mem 0xc3fe00000-0xc3fffffff] page 1G // increase count of PG_LEVEL_1G in c00000000(48G)-c3fffffff(49G) one time
[ 0.000000] init_memory_mapping: [mem 0xc20000000-0xc3fdfffff]
[ 0.000000] [mem 0xc20000000-0xc3fdfffff] page 1G // increase count of PG_LEVEL_1G in c00000000(48G)-c3fffffff(49G) two time
[ 0.000000] init_memory_mapping: [mem 0x00100000-0xbf78ffff]
[ 0.000000] [mem 0x00100000-0x001fffff] page 4k
[ 0.000000] [mem 0x00200000-0x3fffffff] page 2M
[ 0.000000] [mem 0x40000000-0x7fffffff] page 1G
[ 0.000000] [mem 0x80000000-0xbf5fffff] page 2M
[ 0.000000] [mem 0xbf600000-0xbf78ffff] page 4k
[ 0.000000] init_memory_mapping: [mem 0x100000000-0xc1fffffff]
[ 0.000000] [mem 0x100000000-0xc1fffffff] page 1G // increase count of PG_LEVEL_1G in c00000000(48G)-c3fffffff(49G) three time
...
[ 0.000000] DirectMap4k: 3648 kB
[ 0.000000] DirectMap2M: 2084864 kB
[ 0.000000] DirectMap1G: 50331648 kB

euler-linux:~ # cat /proc/meminfo | grep DirectMap
DirectMap4k: 91712 kB
DirectMap2M: 4093952 kB
DirectMap1G: 48234496 kB


total DirectMap is 48234496 + 4093952 + 91712 = 52420160kb
50331648 + 2084864 + 3648 = 52420160kb
total init_memory_mapping is 50323008kb

52420160kb - 50323008kb = 2097152kb = 2G

However I haven't find a better way to fix it, any ideas?

Thanks,
Xishi Qiu

2015-11-20 08:46:32

by Tang Chen

Subject: Re: [RFC] mm: direct mapping count in /proc/meminfo is error

Hi Shi,

Would you please share where did you add the printk debug info ?

Thanks. :)

On 11/20/2015 04:17 PM, Xishi Qiu wrote:
> I find the direct mapping count in /proc/meminfo is error.
> The value should be equal to the size of init_memory_mapping which
> showed in boot log.
>
> I add some print to show direct_pages_count[] immediately after
> init_memory_mapping(). The reason is that we double counting.
>
> Here is the log(kernel v4.4):
> ...
> [ 0.000000] init_memory_mapping: [mem 0x00000000-0x000fffff] // called from "init_memory_mapping(0, ISA_END_ADDRESS);"
> [ 0.000000] [mem 0x00000000-0x000fffff] page 4k
> [ 0.000000] BRK [0x01ebf000, 0x01ebffff] PGTABLE
> [ 0.000000] BRK [0x01ec0000, 0x01ec0fff] PGTABLE
> [ 0.000000] BRK [0x01ec1000, 0x01ec1fff] PGTABLE
> [ 0.000000] init_memory_mapping: [mem 0xc3fe00000-0xc3fffffff] // called from "memory_map_top_down(ISA_END_ADDRESS, end);"
> [ 0.000000] [mem 0xc3fe00000-0xc3fffffff] page 1G // increase count of PG_LEVEL_1G in c00000000(48G)-c3fffffff(49G) one time
> [ 0.000000] init_memory_mapping: [mem 0xc20000000-0xc3fdfffff]
> [ 0.000000] [mem 0xc20000000-0xc3fdfffff] page 1G // increase count of PG_LEVEL_1G in c00000000(48G)-c3fffffff(49G) two time
> [ 0.000000] init_memory_mapping: [mem 0x00100000-0xbf78ffff]
> [ 0.000000] [mem 0x00100000-0x001fffff] page 4k
> [ 0.000000] [mem 0x00200000-0x3fffffff] page 2M
> [ 0.000000] [mem 0x40000000-0x7fffffff] page 1G
> [ 0.000000] [mem 0x80000000-0xbf5fffff] page 2M
> [ 0.000000] [mem 0xbf600000-0xbf78ffff] page 4k
> [ 0.000000] init_memory_mapping: [mem 0x100000000-0xc1fffffff]
> [ 0.000000] [mem 0x100000000-0xc1fffffff] page 1G // increase count of PG_LEVEL_1G in c00000000(48G)-c3fffffff(49G) three time
> ...
> [ 0.000000] DirectMap4k: 3648 kB
> [ 0.000000] DirectMap2M: 2084864 kB
> [ 0.000000] DirectMap1G: 50331648 kB
>
> euler-linux:~ # cat /proc/meminfo | grep DirectMap
> DirectMap4k: 91712 kB
> DirectMap2M: 4093952 kB
> DirectMap1G: 48234496 kB
>
>
> total DirectMap is 48234496 + 4093952 + 91712 = 52420160kb
> 50331648 + 2084864 + 3648 = 52420160kb
> total init_memory_mapping is 50323008kb
>
> 52420160kb - 50323008kb = 2097152kb = 2G
>
> However I haven't find a better way to fix it, any ideas?
>
> Thanks,
> Xishi Qiu
>
>
> .
>

2015-11-20 09:47:07

by Xishi Qiu

Subject: Re: [RFC] mm: direct mapping count in /proc/meminfo is error

On 2015/11/20 16:43, Tang Chen wrote:

> Hi Shi,
>
> Would you please share where did you add the printk debug info ?
>

Sure, at the end of init_mem_mapping(), I add a new function, like arch_report_meminfo(),
just change "seq_printf" to "printk".

Thanks,
Xishi Qiu

> Thanks. :)
>
> On 11/20/2015 04:17 PM, Xishi Qiu wrote:
>> I find the direct mapping count in /proc/meminfo is error.
>> The value should be equal to the size of init_memory_mapping which
>> showed in boot log.
>>
>> I add some print to show direct_pages_count[] immediately after
>> init_memory_mapping(). The reason is that we double counting.
>>
>> Here is the log(kernel v4.4):
>> ...
>> [ 0.000000] init_memory_mapping: [mem 0x00000000-0x000fffff] // called from "init_memory_mapping(0, ISA_END_ADDRESS);"
>> [ 0.000000] [mem 0x00000000-0x000fffff] page 4k
>> [ 0.000000] BRK [0x01ebf000, 0x01ebffff] PGTABLE
>> [ 0.000000] BRK [0x01ec0000, 0x01ec0fff] PGTABLE
>> [ 0.000000] BRK [0x01ec1000, 0x01ec1fff] PGTABLE
>> [ 0.000000] init_memory_mapping: [mem 0xc3fe00000-0xc3fffffff] // called from "memory_map_top_down(ISA_END_ADDRESS, end);"
>> [ 0.000000] [mem 0xc3fe00000-0xc3fffffff] page 1G // increase count of PG_LEVEL_1G in c00000000(48G)-c3fffffff(49G) one time
>> [ 0.000000] init_memory_mapping: [mem 0xc20000000-0xc3fdfffff]
>> [ 0.000000] [mem 0xc20000000-0xc3fdfffff] page 1G // increase count of PG_LEVEL_1G in c00000000(48G)-c3fffffff(49G) two time
>> [ 0.000000] init_memory_mapping: [mem 0x00100000-0xbf78ffff]
>> [ 0.000000] [mem 0x00100000-0x001fffff] page 4k
>> [ 0.000000] [mem 0x00200000-0x3fffffff] page 2M
>> [ 0.000000] [mem 0x40000000-0x7fffffff] page 1G
>> [ 0.000000] [mem 0x80000000-0xbf5fffff] page 2M
>> [ 0.000000] [mem 0xbf600000-0xbf78ffff] page 4k
>> [ 0.000000] init_memory_mapping: [mem 0x100000000-0xc1fffffff]
>> [ 0.000000] [mem 0x100000000-0xc1fffffff] page 1G // increase count of PG_LEVEL_1G in c00000000(48G)-c3fffffff(49G) three time
>> ...
>> [ 0.000000] DirectMap4k: 3648 kB
>> [ 0.000000] DirectMap2M: 2084864 kB
>> [ 0.000000] DirectMap1G: 50331648 kB
>>
>> euler-linux:~ # cat /proc/meminfo | grep DirectMap
>> DirectMap4k: 91712 kB
>> DirectMap2M: 4093952 kB
>> DirectMap1G: 48234496 kB
>>
>>
>> total DirectMap is 48234496 + 4093952 + 91712 = 52420160kb
>> 50331648 + 2084864 + 3648 = 52420160kb
>> total init_memory_mapping is 50323008kb
>>
>> 52420160kb - 50323008kb = 2097152kb = 2G
>>
>> However I haven't find a better way to fix it, any ideas?
>>
>> Thanks,
>> Xishi Qiu
>>
>>
>> .
>>
>
>
> .
>