9:24 a.m.
Patchset to fix RAID5 on ia64. Removes dead CONFIG options and
introduces a new CONFIG option.
(Dave, we need to do this to at least make sure RAID5 is working on ia64
for test2 ...)
Patch is against 2.6.17 + 2.6.18-rc2-git5, and applies to latest nightly
trees.
Patchset submitted upstream
http://marc.theaimsgroup.com/?l=linux-raid&m=115409487823519&w=2
Signed-off-by: Prarit Bhargava <prarit(a)redhat.com>
9:24 a.m.
New subject: [RHEL5 PATCH 1/2] Fix RAID5 components
Patch to fix export issue with RAID5 + IA64.
-adds new MD_RAID5_IA64 option (default to "y")
that builds IA64 RAID5 components.
This patch should be committed by the MD maintainers.
Signed-off-by: Prarit Bhargava <prarit(a)redhat.com>
diff -urNp linux-2.6.17.ia64.orig/drivers/md/Kconfig linux-2.6.17.ia64/drivers/md/Kconfig
--- linux-2.6.17.ia64.orig/drivers/md/Kconfig 2006-07-27 09:20:03.000000000 -0400
+++ linux-2.6.17.ia64/drivers/md/Kconfig 2006-07-27 14:55:15.000000000 -0400
@@ -135,6 +135,13 @@ config MD_RAID456
If unsure, say Y.
+config MD_RAID5_IA64
+ bool "RAID 5 support for IA64"
+ depends on MD_RAID456 && IA64
+ default y
+ ---help---
+ Say "Y" if you plan on using RAID5 in IA64
+
config MD_RAID5_RESHAPE
bool "Support adding drives to a raid-5 array (experimental)"
depends on MD_RAID456 && EXPERIMENTAL
9:24 a.m.
New subject: [RHEL5 PATCH 2/2] Fix IA64 components
Patch to fix export issue with RAID5 + IA64.
-removes dead CONFIG options
This patch should be committed by the IA64 maintainer.
Signed-off-by: Prarit Bhargava <prarit(a)redhat.com>
diff -urNp linux-2.6.17.ia64.orig/arch/ia64/kernel/ia64_ksyms.c
linux-2.6.17.ia64/arch/ia64/kernel/ia64_ksyms.c
--- linux-2.6.17.ia64.orig/arch/ia64/kernel/ia64_ksyms.c 2006-07-27 09:20:01.000000000
-0400
+++ linux-2.6.17.ia64/arch/ia64/kernel/ia64_ksyms.c 2006-07-27 14:51:42.000000000 -0400
@@ -67,7 +68,7 @@ EXPORT_SYMBOL(__udivdi3);
EXPORT_SYMBOL(__moddi3);
EXPORT_SYMBOL(__umoddi3);
-#if defined(CONFIG_MD_RAID5) || defined(CONFIG_MD_RAID5_MODULE)
+#ifdef CONFIG_MD_RAID5_IA64
extern void xor_ia64_2(void);
extern void xor_ia64_3(void);
extern void xor_ia64_4(void);
diff -urNp linux-2.6.17.ia64.orig/arch/ia64/lib/Makefile
linux-2.6.17.ia64/arch/ia64/lib/Makefile
--- linux-2.6.17.ia64.orig/arch/ia64/lib/Makefile 2006-07-27 09:20:01.000000000 -0400
+++ linux-2.6.17.ia64/arch/ia64/lib/Makefile 2006-07-27 14:51:19.000000000 -0400
@@ -14,7 +14,7 @@ lib-y := __divsi3.o __udivsi3.o __modsi3
lib-$(CONFIG_ITANIUM) += copy_page.o copy_user.o memcpy.o
lib-$(CONFIG_MCKINLEY) += copy_page_mck.o memcpy_mck.o
lib-$(CONFIG_PERFMON) += carta_random.o
-lib-$(CONFIG_MD_RAID5) += xor.o
+lib-$(CONFIG_MD_RAID5_IA64) += xor.o
AFLAGS___divdi3.o =
AFLAGS___udivdi3.o = -DUNSIGNED
2:31 p.m.
New subject: [RHEL5 PATCH 1/2] Fix RAID5 components
Prarit Bhargava wrote: [Fri Jul 28 2006, 10:24:53AM EDT]
+config MD_RAID5_IA64
+ bool "RAID 5 support for IA64"
+ depends on MD_RAID456 && IA64
+ default y
+ ---help---
+ Say "Y" if you plan on using RAID5 in IA64
+
This looks wrong to me. If somebody selects CONFIG_RAID456 on ia64
then clearly they're intending to use it on ia64.
One option would be to remove the quoted description ("RAID 5 support
for IA64") then it wouldn't show up in the menu and would be
automatically selected if the conditions were met. But it's still
adding an arch-specific depend in a generic Kconfig, so I don't think
it's the right answer.
diff -urNp linux-2.6.17.ia64.orig/arch/ia64/kernel/ia64_ksyms.c
linux-2.6.17.ia64/arch/ia64/kernel/ia64_ksyms.c
--- linux-2.6.17.ia64.orig/arch/ia64/kernel/ia64_ksyms.c 2006-07-27 09:20:01.000000000
-0400
+++ linux-2.6.17.ia64/arch/ia64/kernel/ia64_ksyms.c 2006-07-27 14:51:42.000000000 -0400
@@ -67,7 +68,7 @@ EXPORT_SYMBOL(__udivdi3);
EXPORT_SYMBOL(__moddi3);
EXPORT_SYMBOL(__umoddi3);
-#if defined(CONFIG_MD_RAID5) || defined(CONFIG_MD_RAID5_MODULE)
+#ifdef CONFIG_MD_RAID5_IA64
extern void xor_ia64_2(void);
extern void xor_ia64_3(void);
extern void xor_ia64_4(void);
diff -urNp linux-2.6.17.ia64.orig/arch/ia64/lib/Makefile
linux-2.6.17.ia64/arch/ia64/lib/Makefile
--- linux-2.6.17.ia64.orig/arch/ia64/lib/Makefile 2006-07-27 09:20:01.000000000 -0400
+++ linux-2.6.17.ia64/arch/ia64/lib/Makefile 2006-07-27 14:51:19.000000000 -0400
@@ -14,7 +14,7 @@ lib-y := __divsi3.o __udivsi3.o __modsi3
lib-$(CONFIG_ITANIUM) += copy_page.o copy_user.o memcpy.o
lib-$(CONFIG_MCKINLEY) += copy_page_mck.o memcpy_mck.o
lib-$(CONFIG_PERFMON) += carta_random.o
-lib-$(CONFIG_MD_RAID5) += xor.o
+lib-$(CONFIG_MD_RAID5_IA64) += xor.o
AFLAGS___divdi3.o =
AFLAGS___udivdi3.o = -DUNSIGNED
This looks wrong too. Since this is in arch/ia64, there's should be
no need to have an arch-specific lib-$(CONFIG_MD_RAID5_IA64), the old
invocation should work fine.
Could you explain what is wrong presently so we can find the right
solution?
Thanks,
Aron
3:21 p.m.
New subject: [RHEL5 PATCH 1/2] Fix RAID5 components
On Fri, Jul 28, 2006 at 03:31:38PM -0400, Aron Griffis wrote:
This looks wrong too. Since this is in arch/ia64, there's should
be
no need to have an arch-specific lib-$(CONFIG_MD_RAID5_IA64), the old
invocation should work fine.
Could you explain what is wrong presently so we can find the right
solution?
Probably this:
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=196903
Dave
--
http://www.codemonkey.org.uk
5:40 p.m.
New subject: [RHEL5 PATCH 1/2] Fix RAID5 components
Aron Griffis wrote:
Prarit Bhargava wrote: [Fri Jul 28 2006, 10:24:53AM EDT]
> +config MD_RAID5_IA64
> + bool "RAID 5 support for IA64"
> + depends on MD_RAID456 && IA64
> + default y
> + ---help---
> + Say "Y" if you plan on using RAID5 in IA64
> +
>
This looks wrong to me. If somebody selects CONFIG_RAID456 on ia64
then clearly they're intending to use it on ia64.
The issue is that CONFIG_RAID456 is tristate, and if it's config'd as a
module, then
#ifdef CONFIG_RAID456 is not true.
One option would be to remove the quoted description ("RAID 5
support
for IA64") then it wouldn't show up in the menu and would be
automatically selected if the conditions were met. But it's still
adding an arch-specific depend in a generic Kconfig, so I don't think
it's the right answer.
Yeah ... I was thinking about that too. Maybe the option does belong in
the arch specific stuff.
(In hindsight maybe I should have RFC'd this ;) )
> This looks wrong too. Since this is in arch/ia64, there's
should be
> no need to have an arch-specific lib-$(CONFIG_MD_RAID5_IA64), the old
> invocation should work fine.
>
>
See above comment about MD_RAID456 = m ... It definately does not work.
> Could you explain what is wrong presently so we can find the
right
> solution?
>
>
See https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=196903
> Thanks,
> Aron
>
9:55 p.m.
New subject: [RHEL5 PATCH 1/2] Fix RAID5 components
Prarit Bhargava wrote: [Fri Jul 28 2006, 06:40:47PM EDT]
The issue is that CONFIG_RAID456 is tristate, and if it's
config'd as a
module, then #ifdef CONFIG_RAID456 is not true.
Ah, thanks for the explanation and the pointer to the BZ. It appears
that #ifdef CONFIG_MD_RAID456_MODULE works in that case. This patch
seems to solve the issue. What do you think?
Regards,
Aron
kernel/ia64_ksyms.c | 2 +-
lib/Makefile | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
CONFIG_MD_RAID5 became CONFIG_MD_RAID456 in drivers/md/Kconfig. Make the same
change in arch/ia64
Signed-off-by: Aron Griffis <aron(a)hp.com>
diff -r 825f7ede8fb3 -r 18551d19b60d kernel/ia64_ksyms.c
--- a/kernel/ia64_ksyms.c Fri Jul 28 22:37:50 2006 -0400
+++ b/kernel/ia64_ksyms.c Fri Jul 28 22:41:01 2006 -0400
@@ -68,7 +68,7 @@ EXPORT_SYMBOL(__moddi3);
EXPORT_SYMBOL(__moddi3);
EXPORT_SYMBOL(__umoddi3);
-#if defined(CONFIG_MD_RAID5) || defined(CONFIG_MD_RAID5_MODULE)
+#if defined(CONFIG_MD_RAID456) || defined(CONFIG_MD_RAID456_MODULE)
extern void xor_ia64_2(void);
extern void xor_ia64_3(void);
extern void xor_ia64_4(void);
diff -r 825f7ede8fb3 -r 18551d19b60d lib/Makefile
--- a/lib/Makefile Fri Jul 28 22:37:50 2006 -0400
+++ b/lib/Makefile Fri Jul 28 22:41:01 2006 -0400
@@ -14,7 +14,7 @@ lib-$(CONFIG_ITANIUM) += copy_page.o cop
lib-$(CONFIG_ITANIUM) += copy_page.o copy_user.o memcpy.o
lib-$(CONFIG_MCKINLEY) += copy_page_mck.o memcpy_mck.o
lib-$(CONFIG_PERFMON) += carta_random.o
-lib-$(CONFIG_MD_RAID5) += xor.o
+lib-$(CONFIG_MD_RAID456) += xor.o
AFLAGS___divdi3.o =
AFLAGS___udivdi3.o = -DUNSIGNED
8:04 a.m.
New subject: [RHEL5 PATCH 1/2] Fix RAID5 components
Aron Griffis wrote:
Prarit Bhargava wrote: [Fri Jul 28 2006, 06:40:47PM EDT]
> The issue is that CONFIG_RAID456 is tristate, and if it's config'd as a
> module, then #ifdef CONFIG_RAID456 is not true.
>
Ah, thanks for the explanation and the pointer to the BZ. It appears
that #ifdef CONFIG_MD_RAID456_MODULE works in that case. This patch
seems to solve the issue. What do you think?
That's _exactly_ what I was looking for :)
I'll resubmit with -p1, etc., in a few minutes ...
P.
Regards,
Aron
kernel/ia64_ksyms.c | 2 +-
lib/Makefile | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
CONFIG_MD_RAID5 became CONFIG_MD_RAID456 in drivers/md/Kconfig. Make the same
change in arch/ia64
Signed-off-by: Aron Griffis <aron(a)hp.com>
diff -r 825f7ede8fb3 -r 18551d19b60d kernel/ia64_ksyms.c
--- a/kernel/ia64_ksyms.c Fri Jul 28 22:37:50 2006 -0400
+++ b/kernel/ia64_ksyms.c Fri Jul 28 22:41:01 2006 -0400
@@ -68,7 +68,7 @@ EXPORT_SYMBOL(__moddi3);
EXPORT_SYMBOL(__moddi3);
EXPORT_SYMBOL(__umoddi3);
-#if defined(CONFIG_MD_RAID5) || defined(CONFIG_MD_RAID5_MODULE)
+#if defined(CONFIG_MD_RAID456) || defined(CONFIG_MD_RAID456_MODULE)
extern void xor_ia64_2(void);
extern void xor_ia64_3(void);
extern void xor_ia64_4(void);
diff -r 825f7ede8fb3 -r 18551d19b60d lib/Makefile
--- a/lib/Makefile Fri Jul 28 22:37:50 2006 -0400
+++ b/lib/Makefile Fri Jul 28 22:41:01 2006 -0400
@@ -14,7 +14,7 @@ lib-$(CONFIG_ITANIUM) += copy_page.o cop
lib-$(CONFIG_ITANIUM) += copy_page.o copy_user.o memcpy.o
lib-$(CONFIG_MCKINLEY) += copy_page_mck.o memcpy_mck.o
lib-$(CONFIG_PERFMON) += carta_random.o
-lib-$(CONFIG_MD_RAID5) += xor.o
+lib-$(CONFIG_MD_RAID456) += xor.o
AFLAGS___divdi3.o =
AFLAGS___udivdi3.o = -DUNSIGNED
6479
days inactive
6482
days old
7 comments
3 participants
participants (3)
-
Aron Griffis -
Dave Jones -
Prarit Bhargava