Re: [ckrm-tech] [PATCH 2/6] CKRM: Core framework support

[Chandra Seetharaman <sekharan@us.ibm.com>]

On Thu, Apr 07, 2005 at 05:39:22PM -0700, Dave Hansen wrote:
> xOn Fri, 2005-04-01 at 19:12 -0800, Chandra Seetharaman wrote:

Hmm... big hole.... 
will provide a temporary patch to disable mem controller in NUMA
till I make it work on NUMA.

chandra
> > +struct ckrm_mem_res {
> ...
> > +       struct ckrm_zone ckrm_zone[MAX_NR_ZONES];
> 
>  static void
>  mem_res_initcls_one(struct ckrm_mem_res *res)
>  {
> ...
> +       for_each_zone(zone) {
> ...
> +               res->ckrm_zone[zindex].memcls = res;
> +               zindex++;
> +       }
> 
> MAX_NR_ZONES is actually the max number of *kinds* of zones.  It's the
> maximum number of 'struct zones' that a single pg_data_t can have in its
> node_zones[] array.  However, each DISCONTIG or NUMA node has one of
> these arrays, and that's what for_each_zone() loops over: _all_ of the
> system's zones, not just a single node's. See:
> 
> #define for_each_zone(zone) \
>         for (zone = pgdat_list->node_zones; zone; zone = next_zone(zone))
> 
> Thus, the first call to mem_res_initcls_one() on a DISCONTIG or NUMA
> system which has a non-node-zero node will overflow that array.
> 
> I saw some of this code before, and that's when I asked about the memory
> controller's NUMA interaction.  I thought something was wrong, but I
> couldn't put my finger on it.
> 
> I addition to these overflows, the same issue exists with results from
> the page_zonenum() macro.  This badly named macro returns a "unique
> identifier" for a node, not its index in its parent pg_data_t's
> node_zones[] array (like the code expects).  So, on i386, a page on
> node0[ZONE_NORMAL] will have a page_zonenum() of 1, node0[ZONE_HIGHMEM]
> will be 2, node1[ZONE_DMA] will be 3, node100[ZONE_NORMAL] will be 301,
> etc...
> 
> Indexing any array declared array[MAX_NR_ZONES=1] as array[301] is
> likely to cause problems pretty fast.
> 
> -- Dave
> 

-- 

----------------------------------------------------------------------
    Chandra Seetharaman               | Be careful what you choose....
              - sekharan@us.ibm.com   |      .......you may get it.
----------------------------------------------------------------------
--
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:"aart@kvack.org"> aart@kvack.org </a>