[RFC][PATCH] no per-arch mem_map init

[RFC][PATCH] no per-arch mem_map init

[Dave Hansen]

[-- Attachment #1: Type: text/plain, Size: 332 bytes --]

This patch has been one of the base patches in the -mhp tree for a bit
now, and seems to be working pretty well, at least on x86.  I would like
to submit it upstream, but I want to get a bit more testing first.  Is
there a chance you ia64 guys could give it a quick test boot to make
sure that it doesn't screw you over?  

-- Dave

[-- Attachment #2: A6-no_arch_mem_map_init.patch --]
[-- Type: text/x-patch, Size: 10474 bytes --]


So, this patch started out with me trying to keep from passing
contiguous, node-specific mem_map into free_area_init_node() and
cousins.  Instead, I relied on some calls to pfn_to_page().

This works fine and dandy when all you need is the pgdat->node_mem_map
to do pfn_to_page().  However, the non-NUMA/DISCONTIG architectures use
the real, global mem_map[] instead of a node_mem_map in the
pfn_to_page() calculation.  So, I ended up effectively trying to
initialize mem_map from itself, when it was NULL.  That was bad, and
caused some very pretty colors on someone's screen when he tested it.

So, I had to make sure to initialize the global mem_map[] before calling
into free_area_init_node().  Then, I realized how many architectures do
this on their own, and have comments like this:

        /* XXX: MRB-remove - this doesn't seem sane, should this be done somewhere else ?*/
        mem_map = NODE_DATA(0)->node_mem_map;

The following patch does what my first one did (don't pass mem_map into
the init functions), incorporates Jesse Barnes' ia64 fixes on top of
that, and gets rid of all but one of the global mem_map initializations
(parisc is weird).  It also magically removes more code than it adds. 
It could be smaller, but I shamelessly added some comments.  

Boot-tested on ppc64, i386 (NUMAQ, plain SMP, laptop), UML (i386).

Signed-off-by: Dave Hansen <haveblue@us.ibm.com>
---

 memhotplug-dave/arch/arm/mm/init.c           |    4 ----
 memhotplug-dave/arch/arm26/mm/init.c         |    2 --
 memhotplug-dave/arch/cris/arch-v10/mm/init.c |    1 -
 memhotplug-dave/arch/ia64/mm/contig.c        |    2 +-
 memhotplug-dave/arch/m32r/mm/init.c          |    2 --
 memhotplug-dave/arch/ppc64/mm/init.c         |    1 -
 memhotplug-dave/arch/sh/mm/init.c            |    2 --
 memhotplug-dave/arch/sh64/mm/init.c          |    3 ---
 memhotplug-dave/arch/sparc/mm/srmmu.c        |    1 -
 memhotplug-dave/arch/sparc/mm/sun4c.c        |    1 -
 memhotplug-dave/arch/sparc64/mm/init.c       |    1 -
 memhotplug-dave/arch/um/kernel/physmem.c     |    1 -
 memhotplug-dave/arch/v850/kernel/setup.c     |    1 -
 memhotplug-dave/mm/page_alloc.c              |   18 ++++++++++++++----
 14 files changed, 15 insertions(+), 25 deletions(-)

diff -puN arch/arm/mm/init.c~A6-no_arch_mem_map_init arch/arm/mm/init.c
--- memhotplug/arch/arm/mm/init.c~A6-no_arch_mem_map_init	2005-02-04 15:21:57.000000000 -0800
+++ memhotplug-dave/arch/arm/mm/init.c	2005-02-04 15:21:57.000000000 -0800
@@ -501,10 +501,6 @@ void __init paging_init(struct meminfo *
 				bdata->node_boot_start >> PAGE_SHIFT, zhole_size);
 	}
 
-#ifndef CONFIG_DISCONTIGMEM
-	mem_map = contig_page_data.node_mem_map;
-#endif
-
 	/*
 	 * finish off the bad pages once
 	 * the mem_map is initialised
diff -puN arch/arm26/mm/init.c~A6-no_arch_mem_map_init arch/arm26/mm/init.c
--- memhotplug/arch/arm26/mm/init.c~A6-no_arch_mem_map_init	2005-02-04 15:21:57.000000000 -0800
+++ memhotplug-dave/arch/arm26/mm/init.c	2005-02-04 15:21:57.000000000 -0800
@@ -309,8 +309,6 @@ void __init paging_init(struct meminfo *
 	free_area_init_node(0, pgdat, zone_size,
 			bdata->node_boot_start >> PAGE_SHIFT, zhole_size);
 
-	mem_map = NODE_DATA(0)->node_mem_map;
-
 	/*
 	 * finish off the bad pages once
 	 * the mem_map is initialised
diff -puN arch/cris/arch-v10/mm/init.c~A6-no_arch_mem_map_init arch/cris/arch-v10/mm/init.c
--- memhotplug/arch/cris/arch-v10/mm/init.c~A6-no_arch_mem_map_init	2005-02-04 15:21:57.000000000 -0800
+++ memhotplug-dave/arch/cris/arch-v10/mm/init.c	2005-02-04 15:21:57.000000000 -0800
@@ -184,7 +184,6 @@ paging_init(void)
 	 */
 
 	free_area_init_node(0, &contig_page_data, zones_size, PAGE_OFFSET >> PAGE_SHIFT, 0);
-	mem_map = contig_page_data.node_mem_map;
 }
 
 /* Initialize remaps of some I/O-ports. It is important that this
diff -puN arch/i386/mm/discontig.c~A6-no_arch_mem_map_init arch/i386/mm/discontig.c
diff -puN arch/ia64/mm/contig.c~A6-no_arch_mem_map_init arch/ia64/mm/contig.c
--- memhotplug/arch/ia64/mm/contig.c~A6-no_arch_mem_map_init	2005-02-04 15:21:57.000000000 -0800
+++ memhotplug-dave/arch/ia64/mm/contig.c	2005-02-04 15:21:57.000000000 -0800
@@ -280,7 +280,7 @@ paging_init (void)
 		vmem_map = (struct page *) vmalloc_end;
 		efi_memmap_walk(create_mem_map_page_table, NULL);
 
-		mem_map = contig_page_data.node_mem_map = vmem_map;
+		NODE_DATA(0)->node_mem_map = vmem_map;
 		free_area_init_node(0, &contig_page_data, zones_size,
 				    0, zholes_size);
 
diff -puN arch/ppc64/mm/init.c~A6-no_arch_mem_map_init arch/ppc64/mm/init.c
--- memhotplug/arch/ppc64/mm/init.c~A6-no_arch_mem_map_init	2005-02-04 15:21:57.000000000 -0800
+++ memhotplug-dave/arch/ppc64/mm/init.c	2005-02-08 09:59:02.000000000 -0800
@@ -659,7 +659,6 @@ void __init paging_init(void)
 
 	free_area_init_node(0, &contig_page_data, zones_size,
 			    __pa(PAGE_OFFSET) >> PAGE_SHIFT, zholes_size);
-	mem_map = contig_page_data.node_mem_map;
 }
 #endif /* CONFIG_DISCONTIGMEM */
 
diff -puN arch/sh/mm/init.c~A6-no_arch_mem_map_init arch/sh/mm/init.c
--- memhotplug/arch/sh/mm/init.c~A6-no_arch_mem_map_init	2005-02-04 15:21:57.000000000 -0800
+++ memhotplug-dave/arch/sh/mm/init.c	2005-02-04 15:21:57.000000000 -0800
@@ -216,8 +216,6 @@ void __init paging_init(void)
 #endif
 	NODE_DATA(0)->node_mem_map = NULL;
 	free_area_init_node(0, NODE_DATA(0), zones_size, __MEMORY_START >> PAGE_SHIFT, 0);
-	/* XXX: MRB-remove - this doesn't seem sane, should this be done somewhere else ?*/
-	mem_map = NODE_DATA(0)->node_mem_map;
 
 #ifdef CONFIG_DISCONTIGMEM
 	/*
diff -puN arch/sh64/mm/init.c~A6-no_arch_mem_map_init arch/sh64/mm/init.c
--- memhotplug/arch/sh64/mm/init.c~A6-no_arch_mem_map_init	2005-02-04 15:21:57.000000000 -0800
+++ memhotplug-dave/arch/sh64/mm/init.c	2005-02-04 15:21:57.000000000 -0800
@@ -124,9 +124,6 @@ void __init paging_init(void)
 	zones_size[ZONE_DMA] = MAX_LOW_PFN - START_PFN;
 	NODE_DATA(0)->node_mem_map = NULL;
 	free_area_init_node(0, NODE_DATA(0), zones_size, __MEMORY_START >> PAGE_SHIFT, 0);
-
-	/* XXX: MRB-remove - this doesn't seem sane, should this be done somewhere else ?*/
-	mem_map = NODE_DATA(0)->node_mem_map;
 }
 
 void __init mem_init(void)
diff -puN arch/sparc/mm/srmmu.c~A6-no_arch_mem_map_init arch/sparc/mm/srmmu.c
--- memhotplug/arch/sparc/mm/srmmu.c~A6-no_arch_mem_map_init	2005-02-04 15:21:57.000000000 -0800
+++ memhotplug-dave/arch/sparc/mm/srmmu.c	2005-02-04 15:21:57.000000000 -0800
@@ -1343,7 +1343,6 @@ void __init srmmu_paging_init(void)
 
 		free_area_init_node(0, &contig_page_data, zones_size,
 				    pfn_base, zholes_size);
-		mem_map = contig_page_data.node_mem_map;
 	}
 }
 
diff -puN arch/sparc/mm/sun4c.c~A6-no_arch_mem_map_init arch/sparc/mm/sun4c.c
--- memhotplug/arch/sparc/mm/sun4c.c~A6-no_arch_mem_map_init	2005-02-04 15:21:57.000000000 -0800
+++ memhotplug-dave/arch/sparc/mm/sun4c.c	2005-02-04 15:21:57.000000000 -0800
@@ -2116,7 +2116,6 @@ void __init sun4c_paging_init(void)
 
 		free_area_init_node(0, &contig_page_data, zones_size,
 				    pfn_base, zholes_size);
-		mem_map = contig_page_data.node_mem_map;
 	}
 
 	cnt = 0;
diff -puN arch/sparc64/mm/init.c~A6-no_arch_mem_map_init arch/sparc64/mm/init.c
--- memhotplug/arch/sparc64/mm/init.c~A6-no_arch_mem_map_init	2005-02-04 15:21:57.000000000 -0800
+++ memhotplug-dave/arch/sparc64/mm/init.c	2005-02-04 15:21:57.000000000 -0800
@@ -1512,7 +1512,6 @@ void __init paging_init(void)
 
 		free_area_init_node(0, &contig_page_data, zones_size,
 				    phys_base >> PAGE_SHIFT, zholes_size);
-		mem_map = contig_page_data.node_mem_map;
 	}
 
 	device_scan();
diff -puN arch/v850/kernel/setup.c~A6-no_arch_mem_map_init arch/v850/kernel/setup.c
--- memhotplug/arch/v850/kernel/setup.c~A6-no_arch_mem_map_init	2005-02-04 15:21:57.000000000 -0800
+++ memhotplug-dave/arch/v850/kernel/setup.c	2005-02-04 15:21:57.000000000 -0800
@@ -283,5 +283,4 @@ init_mem_alloc (unsigned long ram_start,
 	NODE_DATA(0)->node_mem_map = NULL;
 	free_area_init_node (0, NODE_DATA(0), zones_size,
 			     ADDR_TO_PAGE (PAGE_OFFSET), 0);
-	mem_map = NODE_DATA(0)->node_mem_map;
 }
diff -puN mm/page_alloc.c~A6-no_arch_mem_map_init mm/page_alloc.c
--- memhotplug/mm/page_alloc.c~A6-no_arch_mem_map_init	2005-02-04 15:21:57.000000000 -0800
+++ memhotplug-dave/mm/page_alloc.c	2005-02-08 09:59:02.000000000 -0800
@@ -1848,14 +1848,25 @@ static void __init free_area_init_core(s
 	}
 }
 
-void __init node_alloc_mem_map(struct pglist_data *pgdat)
+static void __init alloc_node_mem_map(struct pglist_data *pgdat)
 {
 	unsigned long size;
 
+	/*
+	 * Make sure that the architecture hasn't already allocated
+	 * a node_mem_map, and that the node contains memory.
+	 */
+	if (pgdat->node_mem_map || !pgdat->node_spanned_pages)
+		return;
+
 	size = (pgdat->node_spanned_pages + 1) * sizeof(struct page);
 	pgdat->node_mem_map = alloc_bootmem_node(pgdat, size);
 #ifndef CONFIG_DISCONTIGMEM
-	mem_map = contig_page_data.node_mem_map;
+	/*
+	 * With no DISCONTIG, the global mem_map is just set as node 0's
+	 */
+	if (pgdat == NODE_DATA(0))
+		mem_map = NODE_DATA(0)->node_mem_map;
 #endif
 }
 
@@ -1867,8 +1878,7 @@ void __init free_area_init_node(int nid,
 	pgdat->node_start_pfn = node_start_pfn;
 	calculate_zone_totalpages(pgdat, zones_size, zholes_size);
 
-	if (!pfn_to_page(node_start_pfn))
-		node_alloc_mem_map(pgdat);
+	alloc_node_mem_map(pgdat);
 
 	free_area_init_core(pgdat, zones_size, zholes_size);
 }
diff -puN arch/i386/kernel/mpparse.c~A6-no_arch_mem_map_init arch/i386/kernel/mpparse.c
diff -puN arch/i386/kernel/setup.c~A6-no_arch_mem_map_init arch/i386/kernel/setup.c
diff -puN arch/m32r/mm/init.c~A6-no_arch_mem_map_init arch/m32r/mm/init.c
--- memhotplug/arch/m32r/mm/init.c~A6-no_arch_mem_map_init	2005-02-08 10:06:45.000000000 -0800
+++ memhotplug-dave/arch/m32r/mm/init.c	2005-02-08 10:06:55.000000000 -0800
@@ -121,8 +121,6 @@ unsigned long __init zone_sizes_init(voi
 
 	free_area_init_node(0, NODE_DATA(0), zones_size, start_pfn, 0);
 
-	mem_map = contig_page_data.node_mem_map;
-
 	return 0;
 }
 #else	/* CONFIG_DISCONTIGMEM */
diff -puN arch/um/kernel/physmem.c~A6-no_arch_mem_map_init arch/um/kernel/physmem.c
--- memhotplug/arch/um/kernel/physmem.c~A6-no_arch_mem_map_init	2005-02-08 11:14:20.000000000 -0800
+++ memhotplug-dave/arch/um/kernel/physmem.c	2005-02-08 11:14:31.000000000 -0800
@@ -294,7 +294,6 @@ int init_maps(unsigned long physmem, uns
 		INIT_LIST_HEAD(&p->lru);
 	}
 
-	mem_map = map;
 	max_mapnr = total_pages;
 	return(0);
 }
_

Re: [Lhms-devel] [RFC][PATCH] no per-arch mem_map init

[Yasunori Goto]

Hi Dave-san.

> This patch has been one of the base patches in the -mhp tree for a bit
> now, and seems to be working pretty well, at least on x86.  I would like
> to submit it upstream, but I want to get a bit more testing first.  Is
> there a chance you ia64 guys could give it a quick test boot to make
> sure that it doesn't screw you over?  

I tried this single patch with 2.6.11-rc2-mm2 on my Tiger4, and
there is no problem in booting. In addition, I compliled other
kernel as simple workload test on this test kernel, I didn't find
any problem.

Bye.

-- 
Yasunori Goto <ygoto at us.fujitsu.com>


--
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>

Re: [RFC][PATCH] no per-arch mem_map init

[Bob Picco]

Dave,

Dave Hansen wrote:	[Tue Feb 08 2005, 02:37:14PM EST]
> This patch has been one of the base patches in the -mhp tree for a bit
> now, and seems to be working pretty well, at least on x86.  I would like
> to submit it upstream, but I want to get a bit more testing first.  Is
> there a chance you ia64 guys could give it a quick test boot to make
> sure that it doesn't screw you over?  
> 
> -- Dave
[snip]
> diff -puN arch/i386/mm/discontig.c~A6-no_arch_mem_map_init arch/i386/mm/discontig.c
> diff -puN arch/ia64/mm/contig.c~A6-no_arch_mem_map_init arch/ia64/mm/contig.c
> --- memhotplug/arch/ia64/mm/contig.c~A6-no_arch_mem_map_init	2005-02-04 15:21:57.000000000 -0800
> +++ memhotplug-dave/arch/ia64/mm/contig.c	2005-02-04 15:21:57.000000000 -0800
> @@ -280,7 +280,7 @@ paging_init (void)
>  		vmem_map = (struct page *) vmalloc_end;
>  		efi_memmap_walk(create_mem_map_page_table, NULL);
>  
> -		mem_map = contig_page_data.node_mem_map = vmem_map;
> +		NODE_DATA(0)->node_mem_map = vmem_map;
This has to be changed to.
		mem_map = NODE_DATA(0)->node_mem_map = vmem_map;
>  		free_area_init_node(0, &contig_page_data, zones_size,
>  				    0, zholes_size);
>  
[snip]
I actually submitted an identical change within my last patchset to lhms.
Not making this change requires changing use of mem_map throughout contig.c
and one BUG assertion in init.c.  I haven't tested this patch but it was
indirectly tested by me in FLATMEM configuration for lhms.

bob
--
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>

Re: [Lhms-devel] [RFC][PATCH] no per-arch mem_map init

[Dave Hansen]

On Tue, 2005-02-08 at 16:39 -0800, Yasunori Goto wrote:
> Hi Dave-san.
> 
> > This patch has been one of the base patches in the -mhp tree for a bit
> > now, and seems to be working pretty well, at least on x86.  I would like
> > to submit it upstream, but I want to get a bit more testing first.  Is
> > there a chance you ia64 guys could give it a quick test boot to make
> > sure that it doesn't screw you over?  
> 
> I tried this single patch with 2.6.11-rc2-mm2 on my Tiger4, and
> there is no problem in booting. In addition, I compliled other
> kernel as simple workload test on this test kernel, I didn't find
> any problem.

Thanks for the testing!

-- Dave

--
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>

Re: [RFC][PATCH] no per-arch mem_map init

[Dave Hansen]

On Tue, 2005-02-08 at 20:04 -0500, Bob Picco wrote:
> > -		mem_map = contig_page_data.node_mem_map = vmem_map;
> > +		NODE_DATA(0)->node_mem_map = vmem_map;
> This has to be changed to.
> 		mem_map = NODE_DATA(0)->node_mem_map = vmem_map;
> >  		free_area_init_node(0, &contig_page_data, zones_size,
> >  				    0, zholes_size);
> >  
> [snip]
> I actually submitted an identical change within my last patchset to lhms.

Good to know.  I hadn't actually noticed that bit in your patch.  It's
another good example why to split things up into as many small, logical
pieces as possible.  

> Not making this change requires changing use of mem_map throughout contig.c
> and one BUG assertion in init.c.  I haven't tested this patch but it was
> indirectly tested by me in FLATMEM configuration for lhms.

Hmm.  Do you really need the 'mem_map = ' part?  I *think*
free_area_init_node() calls alloc_node_mem_map(), which should do that
exact assignment for you.  

-- Dave

--
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>

Re: [RFC][PATCH] no per-arch mem_map init

[Bob Picco]

Dave Hansen wrote:	[Tue Feb 08 2005, 08:17:30PM EST]
> On Tue, 2005-02-08 at 20:04 -0500, Bob Picco wrote:
> > > -		mem_map = contig_page_data.node_mem_map = vmem_map;
> > > +		NODE_DATA(0)->node_mem_map = vmem_map;
> > This has to be changed to.
> > 		mem_map = NODE_DATA(0)->node_mem_map = vmem_map;
> > >  		free_area_init_node(0, &contig_page_data, zones_size,
> > >  				    0, zholes_size);
> > >  
> > [snip]
> > I actually submitted an identical change within my last patchset to lhms.
> 
> Good to know.  I hadn't actually noticed that bit in your patch.  It's
> another good example why to split things up into as many small, logical
> pieces as possible.  
> 
> > Not making this change requires changing use of mem_map throughout contig.c
> > and one BUG assertion in init.c.  I haven't tested this patch but it was
> > indirectly tested by me in FLATMEM configuration for lhms.
> 
> Hmm.  Do you really need the 'mem_map = ' part?  I *think*
> free_area_init_node() calls alloc_node_mem_map(), which should do that
> exact assignment for you.  
> 
> -- Dave
okay. I see what happened here.  alloc_node_mem_map is correct and makes my 
suggested change not required.  I think this was fallout from our bad composite
lhms patchset.

bob
> 
--
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>