Strange VM stats in /proc/zoneinfo

Strange VM stats in /proc/zoneinfo

[Petr Tesarik]

Hi folks,

while looking at my /proc/zoneinfo, I noticed that the counters are a bit 
strange:

Node 0, zone      DMA
  pages free     3945
        min      7
        low      8
        high     10
        scanned  0
        spanned  4080
        present  3905
    nr_free_pages 3945

OK, you'll probably argue that the rest is hidden in PCP differentials... BUT:

1. this machine has only 2 CPUs
2. stat_threshold = 4
3. vm_stat_diff[NR_FREE_PAGES] = 0 on both CPUs

Is this only me? Or do I misrepresent what these number actually tell?

TIA,
Petr Tesarik

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: Strange VM stats in /proc/zoneinfo

[Petr Tesarik]

Dne St 15. srpna 2012 11:05:27 Petr Tesarik napsal(a):
> Hi folks,
> 
> while looking at my /proc/zoneinfo, I noticed that the counters are a bit
> strange:
> 
> Node 0, zone      DMA
>   pages free     3945
>         min      7
>         low      8
>         high     10
>         scanned  0
>         spanned  4080
>         present  3905
>     nr_free_pages 3945
> 
> OK, you'll probably argue that the rest is hidden in PCP differentials...
> BUT:
> 
> 1. this machine has only 2 CPUs
> 2. stat_threshold = 4
> 3. vm_stat_diff[NR_FREE_PAGES] = 0 on both CPUs
> 
> Is this only me? Or do I misrepresent what these number actually tell?

I verified the free lists, and this part seems sane, so the "present" count 
must be wrong. FYI I can see this on various x86_64 kernels between 2.6.32 and 
3.5.0.

Petr Tesarik

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: Strange VM stats in /proc/zoneinfo

[Rik van Riel]

On 08/15/2012 05:05 AM, Petr Tesarik wrote:
> Hi folks,
>
> while looking at my /proc/zoneinfo, I noticed that the counters are a bit
> strange:
>
> Node 0, zone      DMA
>    pages free     3945
>          min      7
>          low      8
>          high     10
>          scanned  0
>          spanned  4080
>          present  3905
>      nr_free_pages 3945
>
> OK, you'll probably argue that the rest is hidden in PCP differentials... BUT:
>
> 1. this machine has only 2 CPUs
> 2. stat_threshold = 4
> 3. vm_stat_diff[NR_FREE_PAGES] = 0 on both CPUs
>
> Is this only me? Or do I misrepresent what these number actually tell?

Present should always be equal to or smaller
than spanned, as well as equal to or larger
than free.

Something looks odd...

I wonder if the statistic ends up going off
at bootmem free time?


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>