Home » Mailing list archives » linux-kernel » 2007 » January » 7

Re: [PATCH] Fix sparsemem on Cell (take 3)

Previous thread: [2.6 patch] SCSI seagate.c: remove SEAGATE_USE_ASM by Adrian Bunk on Friday, January 5, 2007 - 9:06 pm. (1 message)

Next thread: [2.6 patch] cleanup include/linux/reiserfs_xattr.h by Adrian Bunk on Friday, January 5, 2007 - 9:09 pm. (1 message)
[view original message]
From: Dave Hansen <haveblue@...>
To: Andrew Morton <akpm@...>
Cc: <linuxppc-dev@...>, KAMEZAWA Hiroyuki <kamezawa.hiroyu@...>, Andrew Morton <akpm@...>, <kmannth@...>, <linux-kernel@...>, <hch@...>, <linux-mm@...>, <paulus@...>, <mkravetz@...>, <gone@...>, <cbe-oss-dev@...>, John Rose <johnrose@...>, Arnd Bergmann <arnd@...>
Subject: [PATCH] Fix sparsemem on Cell (take 3)
Date: Friday, January 5, 2007 - 9:10 pm

I dropped this on the floor over Christmas.  This has had a few smoke

tests on ppc64 and i386 and is ready for -mm.  Against 2.6.20-rc2-mm1.


The following patch fixes an oops experienced on the Cell architecture

when init-time functions, early_*(), are called at runtime.  It alters

the call paths to make sure that the callers explicitly say whether the

call is being made on behalf of a hotplug even, or happening at

boot-time. 


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


---


 lxc-dave/arch/s390/mm/vmem.c    |    3 ++-

 lxc-dave/include/linux/mm.h     |    7 ++++++-

 lxc-dave/include/linux/mmzone.h |    3 ++-

 lxc-dave/mm/memory_hotplug.c    |    6 ++++--

 lxc-dave/mm/page_alloc.c        |   25 +++++++++++++++++--------

 5 files changed, 31 insertions(+), 13 deletions(-)


diff -puN mm/page_alloc.c~sparsemem-enum1 mm/page_alloc.c

--- lxc/mm/page_alloc.c~sparsemem-enum1	2006-12-19 09:38:34.000000000 -0800

+++ lxc-dave/mm/page_alloc.c	2006-12-19 11:18:24.000000000 -0800

@@ -2062,17 +2062,24 @@ static inline unsigned long wait_table_b

  * done. Non-atomic initialization, single-pass.

  */

 void __meminit memmap_init_zone(unsigned long size, int nid, unsigned long zone,

-		unsigned long start_pfn)

+		unsigned long start_pfn, enum memmap_context context)

 {

 	struct page *page;

 	unsigned long end_pfn = start_pfn + size;

 	unsigned long pfn;


 	for (pfn = start_pfn; pfn < end_pfn; pfn++) {

-		if (!early_pfn_valid(pfn))

-			continue;

-		if (!early_pfn_in_nid(pfn, nid))

-			continue;

+		/*

+		 * There can be holes in boot-time mem_map[]s

+		 * handed to this function.  They do not

+		 * exist on hotplugged memory.

+		 */

+		if (context == MEMMAP_EARLY) {

+			if (!early_pfn_valid(pfn))

+				continue;

+			if (!early_pfn_in_nid(pfn, nid))

+				continue;

+		}

 		page = pfn_to_page(pfn);

 		set_page_links(page, zone, nid, pfn);

 		init_page_count(page);

@@ -2102,7 +2109,7 @@ void zone_init_free_lists(struct pglist_


 #ifndef __HAVE_ARCH_MEMMAP_INIT

 #de...
[ message continues ]
[view original message]
From: John Rose <johnrose@...>
To: Dave Hansen <haveblue@...>
Cc: Andrew Morton <akpm@...>, External List <linuxppc-dev@...>, KAMEZAWA Hiroyuki <kamezawa.hiroyu@...>, <kmannth@...>, lkml <linux-kernel@...>, <hch@...>, <linux-mm@...>, Paul Mackerras <paulus@...>, <mkravetz@...>, <gone@...>, <cbe-oss-dev@...>, Arnd Bergmann <arnd@...>
Subject: Re: [PATCH] Fix sparsemem on Cell (take 3)
Date: Saturday, January 6, 2007 - 12:52 am

Could this break ia64, given that it uses memmap_init_zone()?


-
[view original message]
From: Dave Hansen <haveblue@...>
To: John Rose <johnrose@...>
Cc: Andrew Morton <akpm@...>, External List <linuxppc-dev@...>, KAMEZAWA Hiroyuki <kamezawa.hiroyu@...>, <kmannth@...>, lkml <linux-kernel@...>, <hch@...>, <linux-mm@...>, Paul Mackerras <paulus@...>, <mkravetz@...>, <gone@...>, Arnd Bergmann <arnd@...>
Subject: Re: [PATCH] Fix sparsemem on Cell (take 3)
Date: Sunday, January 7, 2007 - 4:58 am

You are right, I think it does.


Here's an updated patch to replace the earlier one.  I had to move the

enum definition over to a different header because ia64 evidently has a

different include order.


---


The following patch fixes an oops experienced on the Cell architecture

when init-time functions, early_*(), are called at runtime.  It alters

the call paths to make sure that the callers explicitly say whether the

call is being made on behalf of a hotplug even, or happening at

boot-time. 


It has been compile tested on ia64, s390, i386 and x86_64.


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


---


 lxc-dave/arch/ia64/mm/init.c    |    5 +++--

 lxc-dave/arch/s390/mm/vmem.c    |    3 ++-

 lxc-dave/include/linux/mm.h     |    3 ++-

 lxc-dave/include/linux/mmzone.h |    8 ++++++--

 lxc-dave/mm/memory_hotplug.c    |    6 ++++--

 lxc-dave/mm/page_alloc.c        |   25 +++++++++++++++++--------

 6 files changed, 34 insertions(+), 16 deletions(-)


diff -puN arch/s390/mm/vmem.c~Re-_PATCH_Fix_sparsemem_on_Cell arch/s390/mm/vmem.c

--- lxc/arch/s390/mm/vmem.c~Re-_PATCH_Fix_sparsemem_on_Cell	2007-01-05 15:38:23.000000000 -0800

+++ lxc-dave/arch/s390/mm/vmem.c	2007-01-07 00:47:02.000000000 -0800

@@ -61,7 +61,8 @@ void memmap_init(unsigned long size, int


 		if (map_start < map_end)

 			memmap_init_zone((unsigned long)(map_end - map_start),

-					 nid, zone, page_to_pfn(map_start));

+					 nid, zone, page_to_pfn(map_start),

+					 MEMMAP_EARLY);

 	}

 }


diff -puN include/linux/mm.h~Re-_PATCH_Fix_sparsemem_on_Cell include/linux/mm.h

--- lxc/include/linux/mm.h~Re-_PATCH_Fix_sparsemem_on_Cell	2007-01-05 15:38:23.000000000 -0800

+++ lxc-dave/include/linux/mm.h	2007-01-06 23:57:59.000000000 -0800

@@ -979,7 +979,8 @@ extern int early_pfn_to_nid(unsigned lon

 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */

 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */

 extern void set_dma_reserve(unsigned long new_dma_reserve);

-extern void memmap_init_zone(unsigned long, int, unsigned...
[ message continues ]
[view original message]
From: Tim Pepper <tpepper@...>
To: Dave Hansen <haveblue@...>
Cc: John Rose <johnrose@...>, Andrew Morton <akpm@...>, External List <linuxppc-dev@...>, KAMEZAWA Hiroyuki <kamezawa.hiroyu@...>, <kmannth@...>, lkml <linux-kernel@...>, <hch@...>, <linux-mm@...>, Paul Mackerras <paulus@...>, <mkravetz@...>, <gone@...>, Arnd Bergmann <arnd@...>
Subject: Re: [PATCH] Fix sparsemem on Cell (take 3)
Date: Monday, January 8, 2007 - 2:47 am

Boot tested OK on ia64 with this latest version of the patch.


(forgot to click plain text on gmail the first time..sorry if you got

html mail or repeat)


Tim

-
[view original message]
From: Arnd Bergmann <arnd@...>
To: Dave Hansen <haveblue@...>
Cc: John Rose <johnrose@...>, Andrew Morton <akpm@...>, External List <linuxppc-dev@...>, KAMEZAWA Hiroyuki <kamezawa.hiroyu@...>, <kmannth@...>, lkml <linux-kernel@...>, <hch@...>, <linux-mm@...>, Paul Mackerras <paulus@...>, <mkravetz@...>, <gone@...>
Subject: Re: [PATCH] Fix sparsemem on Cell (take 3)
Date: Sunday, January 7, 2007 - 8:07 am

I can't test it here, since I'm travelling at the moment, but


Acked-by: Arnd Bergmann <arndb@de.ibm.com>

-
Previous thread: [2.6 patch] SCSI seagate.c: remove SEAGATE_USE_ASM by Adrian Bunk on Friday, January 5, 2007 - 9:06 pm. (1 message)

Next thread: [2.6 patch] cleanup include/linux/reiserfs_xattr.h by Adrian Bunk on Friday, January 5, 2007 - 9:09 pm. (1 message)