Skip to contentSkip to conversation

Fix Altix boot breakage in ACPI

PATCH In review Russ Anderson 4 participants 5 messages May 16, 2008 – May 19, 2008

lore
Share & tools
lore

Russ Anderson

author
#0 lore raw

PATCH 1/1 Fix Altix boot breakage in ACPI

A recent ACPI change (commit 5dc90c0b2d4bd0127624bab67cec159b2c6c4daf) inadvertently broke Altix boot. This patch allow Altix to boot.

Signed-off-by Russ Anderson <[email protected]>
Acked-by Ingo Molnar <[email protected]>
---
 arch/ia64/Kconfig    |    4 ++++
 include/linux/acpi.h |    2 +-
 2 files changed, 5 insertions(+), 1 deletion(-)

Index: 2.6.26-rc2/arch/ia64/Kconfig
=================================--- 2.6.26-rc2.orig/arch/ia64/Kconfig 2008-05-16 09:36:37.285558037 -0500
+++ 2.6.26-rc2/arch/ia64/Kconfig 2008-05-16 09:37:39.149204580 -0500
@@ -60,6 +60,10 @@ config RWSEM_XCHGADD_ALGORITHM
bool
default y

+config HAVE_ARCH_PARSE_SRAT
+ bool
+ default y
+
config ARCH_HAS_ILOG2_U32
bool
default n
Index: 2.6.26-rc2/include/linux/acpi.h
=================================--- 2.6.26-rc2.orig/include/linux/acpi.h 2008-05-16 09:37:12.989971159 -0500
+++ 2.6.26-rc2/include/linux/acpi.h 2008-05-16 09:50:17.082906375 -0500
@@ -93,7 +93,7 @@ int acpi_parse_mcfg (struct acpi_table_h
void acpi_table_print_madt_entry (struct acpi_subtable_header *madt);

/* the following four functions are architecture-dependent */
-#ifdef CONFIG_HAVE_ARCH_PARSE_SRAT
+#if defined(CONFIG_HAVE_ARCH_PARSE_SRAT) && !defined(NR_NODE_MEMBLKS)
#define NR_NODE_MEMBLKS MAX_NUMNODES
#define acpi_numa_slit_init(slit) do {} while (0)
#define acpi_numa_processor_affinity_init(pa) do {} while (0)

Signature · 2 lines
Russ Anderson, OS RAS/Partitioning Project Lead  
SGI - Silicon Graphics Inc          [email protected]
#1 lore raw
Quoted 1 line

A recent ACPI change (commit 5dc90c0b2d4bd0127624bab67cec159b2c6c4daf)

-ENOSUCHCOMMIT ... well I can't see it in Linus' tree.

-Tony

author Acked-by Ingo Molnar on v1 1/1
#2 lore raw
In reply to Russ Anderson
Quoted 3 lines · Tony
Quoted 1 line

A recent ACPI change (commit 5dc90c0b2d4bd0127624bab67cec159b2c6c4daf)

-ENOSUCHCOMMIT ... well I can't see it in Linus' tree.

Sorry, I should have added more information.

It is in Andrew's tree (2.6.26-rc2-mm1). It has yet to make it to Linus' tree.

http://lkml.org/lkml/2008/5/14/366

The patch name is acpi-acpi_numa_init-build-fix
http://kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.26-rc2/2.6.26-rc2-mm1/patch-list

git bisect found:
------------------------------------------------------------------------------------
5dc90c0b2d4bd0127624bab67cec159b2c6c4daf is first bad commit commit 5dc90c0b2d4bd0127624bab67cec159b2c6c4daf

Author: Ingo Molnar <[email protected]>
Date:   Thu May 1 09:51:47 2008 +0000

acpi-acpi_numa_init-build-fix

x86.git testing found the following build error on latest -git:

drivers/acpi/numa.c: In function 'acpi_numa_init':
drivers/acpi/numa.c:226: error: 'NR_NODE_MEMBLKS' undeclared (first use in this function) drivers/acpi/numa.c:226: error: (Each undeclared identifier is reported only once drivers/acpi/numa.c:226: error: for each function it appears in.)

with this config:

http://redhat.com/~mingo/misc/config-Wed_Apr_30_22_42_42_CEST_2008.bad

i suspect we dont want SRAT parsing when CONFIG_HAVE_ARCH_PARSE_SRAT is unset - but the fix looks a bit ugly. Perhaps we should define NR_NODE_MEMBLKS even in this case and just let the code fall back to some sane behavior?

Signed-off-by Ingo Molnar <[email protected]>
Signed-off-by Andrew Morton <[email protected]>

:040000 040000 65397f028a8bc082be2bde667554c9115e88064b 58d35b65618fafdcbb449dd114e0381df467431d M drivers
------------------------------------------------------------------------------------

The inadvertent breakage is the patch puts #ifdef CONFIG_HAVE_ARCH_PARSE_SRAT around acpi_parse_memory_affinity() and acpi_table_parse_srat(), but Altix needs to parse the SRAT. The result is that Altix dies early in boot in the ACPI code.

The fix defines CONFIG_HAVE_ARCH_PARSE_SRAT in ia64 config and prevents the parsing routines from being defined as stubs if NR_NODE_MEMBLKS is defined.

Since the problem effects Altix, I sent the fix to linux-ia64. Would it have been better to send the fix directly to lkml & Andrew?

This patch fixes the problem
------------------------------------------------------------------------------------
A recent ACPI change (commit 5dc90c0b2d4bd0127624bab67cec159b2c6c4daf) inadvertently broke Altix boot. This patch allow Altix to boot.

Signed-off-by Russ Anderson <[email protected]>
Acked-by Ingo Molnar <[email protected]>
---
 arch/ia64/Kconfig    |    4 ++++
 include/linux/acpi.h |    2 +-
 2 files changed, 5 insertions(+), 1 deletion(-)

Index: 2.6.26-rc2/arch/ia64/Kconfig
=================================--- 2.6.26-rc2.orig/arch/ia64/Kconfig 2008-05-16 09:36:37.285558037 -0500
+++ 2.6.26-rc2/arch/ia64/Kconfig 2008-05-16 09:37:39.149204580 -0500
@@ -60,6 +60,10 @@ config RWSEM_XCHGADD_ALGORITHM
bool
default y

+config HAVE_ARCH_PARSE_SRAT
+ bool
+ default y
+
config ARCH_HAS_ILOG2_U32
bool
default n
Index: 2.6.26-rc2/include/linux/acpi.h
=================================--- 2.6.26-rc2.orig/include/linux/acpi.h 2008-05-16 09:37:12.989971159 -0500
+++ 2.6.26-rc2/include/linux/acpi.h 2008-05-16 09:50:17.082906375 -0500
@@ -93,7 +93,7 @@ int acpi_parse_mcfg (struct acpi_table_h
void acpi_table_print_madt_entry (struct acpi_subtable_header *madt);

/* the following four functions are architecture-dependent */
-#ifdef CONFIG_HAVE_ARCH_PARSE_SRAT
+#if defined(CONFIG_HAVE_ARCH_PARSE_SRAT) && !defined(NR_NODE_MEMBLKS)
#define NR_NODE_MEMBLKS MAX_NUMNODES
#define acpi_numa_slit_init(slit) do {} while (0)
#define acpi_numa_processor_affinity_init(pa) do {} while (0)
------------------------------------------------------------------------------------

Signature · 2 lines
Russ Anderson, OS RAS/Partitioning Project Lead  
SGI - Silicon Graphics Inc          [email protected]
#3 lore raw
In reply to Russ Anderson
Quoted 5 lines · On Sat, 17 May 2008 09:10:35 -0500 Russ Anderson <[email protected]> wrote:
Quoted 3 lines · Tony
Quoted 1 line

A recent ACPI change (commit 5dc90c0b2d4bd0127624bab67cec159b2c6c4daf)

-ENOSUCHCOMMIT ... well I can't see it in Linus' tree.

Quoted 8 lines

Sorry, I should have added more information.

It is in Andrew's tree (2.6.26-rc2-mm1). It has yet to make it to Linus' tree.

http://lkml.org/lkml/2008/5/14/366

The patch name is acpi-acpi_numa_init-build-fix
http://kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.26-rc2/2.6.26-rc2-mm1/patch-list

That was dropped because it broke things. umm...

	From: "Torsten Kaiser" <[email protected]>
	To: "Andrew Morton" <[email protected]>
	Cc: [email protected]
	Subject: Re: 2.6.26-rc2-mm1
	Date: 	Wed, 14 May 2008 21:12:13 +0200
	Sender: [email protected]
	...
	But somehow the NUMAness of my system is gone.
Quoted 30 lines

A recent ACPI change (commit 5dc90c0b2d4bd0127624bab67cec159b2c6c4daf) inadvertently broke Altix boot. This patch allow Altix to boot.

Signed-off-by Russ Anderson <[email protected]>
Acked-by Ingo Molnar <[email protected]>
---
 arch/ia64/Kconfig    |    4 ++++
 include/linux/acpi.h |    2 +-
 2 files changed, 5 insertions(+), 1 deletion(-)

Index: 2.6.26-rc2/arch/ia64/Kconfig
=================================> --- 2.6.26-rc2.orig/arch/ia64/Kconfig 2008-05-16 09:36:37.285558037 -0500
+++ 2.6.26-rc2/arch/ia64/Kconfig 2008-05-16 09:37:39.149204580 -0500

(quoted diff) +4 -0
@@ -60,6 +60,10 @@ config RWSEM_XCHGADD_ALGORITHM
 	bool
 	default y
 
+config HAVE_ARCH_PARSE_SRAT
+	bool
+	default y
+
 config ARCH_HAS_ILOG2_U32
 	bool
 	default n

Index: 2.6.26-rc2/include/linux/acpi.h
=================================> --- 2.6.26-rc2.orig/include/linux/acpi.h 2008-05-16 09:37:12.989971159 -0500
+++ 2.6.26-rc2/include/linux/acpi.h 2008-05-16 09:50:17.082906375 -0500

Review int acpi_parse_mcfg (struct acpi_table_h · line 99

Quoted 8 lines
(quoted diff) +1 -1
@@ -93,7 +93,7 @@ int acpi_parse_mcfg (struct acpi_table_h
 void acpi_table_print_madt_entry (struct acpi_subtable_header *madt);
 
 /* the following four functions are architecture-dependent */
-#ifdef CONFIG_HAVE_ARCH_PARSE_SRAT
+#if defined(CONFIG_HAVE_ARCH_PARSE_SRAT) && !defined(NR_NODE_MEMBLKS)
 #define NR_NODE_MEMBLKS MAX_NUMNODES
 #define acpi_numa_slit_init(slit) do {} while (0)
 #define acpi_numa_processor_affinity_init(pa) do {} while (0)

So this affects machines other than altix?

Who gets to merge this? I'll plan on sending it in on Mondayish unless I hear otherwise.

(I wish the changelog was more specific than "inadvertently broke". It should explain what the failure was).

Tested-by on v1 1/1
#4 lore raw
In reply to Russ Anderson

Hi Russ,

Quoted 6 lines

------------------------------------------------------------------------------------

The inadvertent breakage is the patch puts #ifdef CONFIG_HAVE_ARCH_PARSE_SRAT around acpi_parse_memory_affinity() and acpi_table_parse_srat(), but Altix needs to parse the SRAT. The result is that Altix dies early in boot in the ACPI code.

your patch fix fujitsu box problem too.
I think almost IA64 box use ACPI SRAT, not only altrix.

Tested-by KOSAKI Motohiro <[email protected]>

and IMHO, your patch is not altrix independent. it is generic more.

if possible, I hope change patch description. What do you think it?