[dpdk-dev] Get the socket id

[dpdk-dev] Get the socket id

[dvr]

[-- Attachment #1: Type: text/plain, Size: 737 bytes --]

hi,
   all
   I'm confused with the following code(Eal_memory.c in lib\librte_eal\linuxapp\eal directory, LINE 373):
 
   nodestr = strstr(buf," N");
   if(nodestr == NULL) {...}
   nodestr += 2;
   end = strstr(nodestr,"=");
   if(end == NULL) {...}
   // here nodestr = "0=1"
   end[0] = '\0';
   end = NULL;
  // here nodestr = "0"
   socket_id = strtoul(nodestr,&end,0);
 
   Through "printf", I know that:
   buf = "00200000 default file=/mnt/dpdk/huge/rtemap_1 huge dirty=1 N0=1"
   And finally socket_id=0
   My question is :why we get the socket_id before "=" instead of after "=" ?
   and if we want to get the character after "N", isn't this code too complex?
 
   I will be very appreciate for any help.
   Regards!
 
Maria
  

[-- Attachment #2: Type: text/html, Size: 1560 bytes --]

Re: [dpdk-dev] Get the socket id

[Stephen Hemminger]

On Wed, 7 Aug 2013 14:14:19 +0800 (CST)
dvr <jinxiu.1006@163.com> wrote:

> hi,
>    all
>    I'm confused with the following code(Eal_memory.c in lib\librte_eal\linuxapp\eal directory, LINE 373):
>  
>    nodestr = strstr(buf," N");
>    if(nodestr == NULL) {...}
>    nodestr += 2;
>    end = strstr(nodestr,"=");
>    if(end == NULL) {...}
>    // here nodestr = "0=1"
>    end[0] = '\0';
>    end = NULL;
>   // here nodestr = "0"
>    socket_id = strtoul(nodestr,&end,0);
>  
>    Through "printf", I know that:
>    buf = "00200000 default file=/mnt/dpdk/huge/rtemap_1 huge dirty=1 N0=1"
>    And finally socket_id=0
>    My question is :why we get the socket_id before "=" instead of after "=" ?
>    and if we want to get the character after "N", isn't this code too complex?
>  
>    I will be very appreciate for any help.
>    Regards!
>  
> Maria
>   

Good question.

The best reference for the kernel proc files is usually in the Linux
kernel source Documentation directory. But can't find any documentation
on numa_maps in detail.

Inside kernel VM, there is an array of numa_maps for each process.
The two parts of NX=Y are:
   X => the numa map index
   Y => the numa node associated with that map

Therefore it is possible to have multiple entries, like:
 "00200000 default file=/myfile dirty=1 N0=1 N1=2"
if the the memory region crosses multiple NUMA nodes.
Since huge pages are pinned to numa node, you won't see that for them.
Basically if "huge" is in the line only one NX=Y entry will be present;
and the X part is unimportant.

Re: [dpdk-dev] Get the socket id

[Damien Millescamps]

On 08/07/2013 05:52 PM, Stephen Hemminger wrote:
> But can't find any documentation
> on numa_maps in detail.
You should try with man 5 numa_maps.

-- 
Damien Millescamps

Re: [dpdk-dev] Get the socket id

[Stephen Hemminger]

On Wed, 07 Aug 2013 17:56:55 +0200
Damien Millescamps <damien.millescamps@6wind.com> wrote:

> On 08/07/2013 05:52 PM, Stephen Hemminger wrote:
> > But can't find any documentation
> > on numa_maps in detail.
> You should try with man 5 numa_maps.
> 

Thanks, knew it was there some where:
       N<node>=<nr_pages>
              The number of pages allocated on  <node>.   <nr_pages>  includes
              only  pages currently mapped by the process.  Page migration and
              memory reclaim may have temporarily  unmapped  pages  associated
              with  this  memory  range.   These  pages may only show up again
              after the process has attempted to reference them.  If the  mem‐
              ory range represents a shared memory area or file mapping, other
              processes may currently have additional pages mapped in a corre‐
              sponding memory range.