[PATCH v3 0/5] mm: Enable CONFIG_NODES_SPAN_OTHER_NODES by default for NUMA

Mike Rapoport rppt at linux.ibm.com
Fri Apr 10 02:27:41 AEST 2020


On Tue, Mar 31, 2020 at 04:21:38PM +0200, Michal Hocko wrote:
> On Tue 31-03-20 22:03:32, Baoquan He wrote:
> > Hi Michal,
> > 
> > On 03/31/20 at 10:55am, Michal Hocko wrote:
> > > On Tue 31-03-20 11:14:23, Mike Rapoport wrote:
> > > > Maybe I mis-read the code, but I don't see how this could happen. In the
> > > > HAVE_MEMBLOCK_NODE_MAP=y case, free_area_init_node() calls
> > > > calculate_node_totalpages() that ensures that node->node_zones are entirely
> > > > within the node because this is checked in zone_spanned_pages_in_node().
> > > 
> > > zone_spanned_pages_in_node does chech the zone boundaries are within the
> > > node boundaries. But that doesn't really tell anything about other
> > > potential zones interleaving with the physical memory range.
> > > zone->spanned_pages simply gives the physical range for the zone
> > > including holes. Interleaving nodes are essentially a hole
> > > (__absent_pages_in_range is going to skip those).
> > > 
> > > That means that when free_area_init_core simply goes over the whole
> > > physical zone range including holes and that is why we need to check
> > > both for physical and logical holes (aka other nodes).
> > > 
> > > The life would be so much easier if the whole thing would simply iterate
> > > over memblocks...
> > 
> > The memblock iterating sounds a great idea. I tried with putting the
> > memblock iterating in the upper layer, memmap_init(), which is used for
> > boot mem only anyway. Do you think it's doable and OK? It yes, I can
> > work out a formal patch to make this simpler as you said. The draft code
> > is as below. Like this it uses the existing code and involves little change.
> 
> Doing this would be a step in the right direction! I haven't checked the
> code very closely though. The below sounds way too simple to be truth I
> am afraid. First for_each_mem_pfn_range is available only for
> CONFIG_HAVE_MEMBLOCK_NODE_MAP (which is one of the reasons why I keep
> saying that I really hate that being conditional). Also I haven't really
> checked the deferred initialization path - I have a very vague
> recollection that it has been converted to the memblock api but I have
> happilly dropped all that memory.

The Baoquan's patch almost did it, at least for simple case of qemu with 2
nodes. It's only missing the adjustment to the size passed to
memmap_init_zone() as it may change because of clamping.

I've drafted something that removes HAVE_MEMBLOCK_NODE_MAP and added this
patch there [1]. For several memory configurations I could emulate with
qemu it worked.
I'm going to wait a bit to see of kbuild is happy and then I'll send the
patches.

Baoquan, I took liberty to add your SoB, hope you don't mind.

[1] https://git.kernel.org/pub/scm/linux/kernel/git/rppt/linux.git/log/?h=memblock/all-have-node-map 
  
> > diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> > index 138a56c0f48f..558d421f294b 100644
> > --- a/mm/page_alloc.c
> > +++ b/mm/page_alloc.c
> > @@ -6007,14 +6007,6 @@ void __meminit memmap_init_zone(unsigned long size, int nid, unsigned long zone,
> >  		 * function.  They do not exist on hotplugged memory.
> >  		 */
> >  		if (context == MEMMAP_EARLY) {
> > -			if (!early_pfn_valid(pfn)) {
> > -				pfn = next_pfn(pfn);
> > -				continue;
> > -			}
> > -			if (!early_pfn_in_nid(pfn, nid)) {
> > -				pfn++;
> > -				continue;
> > -			}
> >  			if (overlap_memmap_init(zone, &pfn))
> >  				continue;
> >  			if (defer_init(nid, pfn, end_pfn))
> > @@ -6130,9 +6122,17 @@ static void __meminit zone_init_free_lists(struct zone *zone)
> >  }
> >  
> >  void __meminit __weak memmap_init(unsigned long size, int nid,
> > -				  unsigned long zone, unsigned long start_pfn)
> > +				  unsigned long zone, unsigned long range_start_pfn)
> >  {
> > -	memmap_init_zone(size, nid, zone, start_pfn, MEMMAP_EARLY, NULL);
> > +	unsigned long start_pfn, end_pfn;
> > +	unsigned long range_end_pfn = range_start_pfn + size;
> > +	int i;
> > +	for_each_mem_pfn_range(i, nid, &start_pfn, &end_pfn, NULL) {
> > +		start_pfn = clamp(start_pfn, range_start_pfn, range_end_pfn);
> > +		end_pfn = clamp(end_pfn, range_start_pfn, range_end_pfn);
> > +		if (end_pfn > start_pfn)
> > +			memmap_init_zone(size, nid, zone, start_pfn, MEMMAP_EARLY, NULL);
> > +	}
> >  }
> >  
> >  static int zone_batchsize(struct zone *zone)
> 
> -- 
> Michal Hocko
> SUSE Labs

-- 
Sincerely yours,
Mike.



More information about the Linuxppc-dev mailing list