Skip to contentSkip to conversation

Kexec/Kdump: honour non-zero crashkernel offset.

DISCUSSION Horms 2 participants 7 messages Sep 20, 2006 – Dec 12, 2006

lore
Share & tools
lore

Horms

author
#0 lore raw

There seems to be a value in both allowing the kernel to determine the base offset of the crashkernel automatically and allowing users's to sepcify it.

The old behaviour on ia64, which is still the current behaviour on most architectures is for the user to always specify the address. Recently ia64 was changed so that it is always automatically determined.

With this patch the kernel automatically determines the offset if the supplied value is 0, otherwise it uses the value provided.

This should probably be backed by a documentation change.

Index: linux-2.6/arch/ia64/kernel/setup.c
=================================--- linux-2.6.orig/arch/ia64/kernel/setup.c 2006-09-20 10:56:28.000000000 +0900
+++ linux-2.6/arch/ia64/kernel/setup.c 2006-09-20 10:58:06.000000000 +0900
@@ -257,7 +257,7 @@

#ifdef CONFIG_KEXEC
/* crashkernel=size@offset specifies the size to reserve for a crash
- * kernel.(offset is ingored for keep compatibility with other archs)
+ * kernel. If offset is 0, then it is determined automatically.
* By reserving this memory we guarantee that linux
* never set's it up as a DMA target.
* Useful for holding code to do something appropriate
@@ -268,7 +268,7 @@
unsigned long base, size;
if (from) {
size = memparse(from + 12, &from);
- if (size) {
+ if (size && !base) {
sort_regions(rsvd_region, n); base = kdump_find_rsvd_region(size, rsvd_region, n);

Signature · 3 lines
Horms
  H: http://www.vergenet.net/~horms/
  W: http://www.valinux.co.jp/en/
#1 lore raw
Quoted 34 lines · Horms

There seems to be a value in both allowing the kernel to determine the base offset of the crashkernel automatically and allowing users's to sepcify it.

The old behaviour on ia64, which is still the current behaviour on most architectures is for the user to always specify the address. Recently ia64 was changed so that it is always automatically determined.

With this patch the kernel automatically determines the offset if the supplied value is 0, otherwise it uses the value provided.

This should probably be backed by a documentation change.

Index: linux-2.6/arch/ia64/kernel/setup.c
=================================> --- linux-2.6.orig/arch/ia64/kernel/setup.c 2006-09-20
10:56:28.000000000 +0900
+++ linux-2.6/arch/ia64/kernel/setup.c 2006-09-20 10:58:06.000000000 +0900

(quoted diff) +0 -0
@@ -257,7 +257,7 @@
 
 #ifdef CONFIG_KEXEC
        /* crashkernel=size@offset specifies the size to reserve for a

crash
- * kernel.(offset is ingored for keep compatibility with other
archs)
+ * kernel. If offset is 0, then it is determined
automatically.
* By reserving this memory we guarantee that linux
* never set's it up as a DMA target.
* Useful for holding code to do something appropriate

Review line 274

Quoted 8 lines
(quoted diff) +1 -1
@@ -268,7 +268,7 @@
                unsigned long base, size;
                if (from) {
                        size = memparse(from + 12, &from);
-                       if (size) {
+                       if (size && !base) {
                                sort_regions(rsvd_region, n);
                                base = kdump_find_rsvd_region(size,
                                rsvd_region, n);

This patch is incorrect.
I think you will get a compile warning of base is used before it is
aligned a value here...
Thanks
Zou Nan hai

Quoted 5 lines
Signature · 3 lines
Horms
  H: http://www.vergenet.net/~horms/
  W: http://www.valinux.co.jp/en/
author
#2 lore raw
In reply to Horms
Quoted 3 lines · Zou Nan hai

This patch is incorrect.
I think you will get a compile warning of base is used before it is
aligned a value here...

Sorry about that. Is this better?

Signature · 46 lines
Horms
  H: http://www.vergenet.net/~horms/
  W: http://www.valinux.co.jp/en/

[IA64] Kexec/Kdump: honour non-zero crashkernel offset.

There seems to be a value in both allowing the kernel to determine
the base offset of the crashkernel automatically and allowing
users's to sepcify it.

The old behaviour on ia64, which is still the current behaviour on
most architectures is for the user to always specify the address.
Recently ia64 was changed so that it is always automatically determined.

With this patch the kernel automatically determines the offset if
the supplied value is 0, otherwise it uses the value provided.

This should probably be backed by a documentation change.

Signed-Off-By: Simon Horman <[email protected]>

Index: linux-2.6-ia64/arch/ia64/kernel/setup.c
=================================--- linux-2.6-ia64.orig/arch/ia64/kernel/setup.c	2006-09-21 14:08:20.000000000 +0900
+++ linux-2.6-ia64/arch/ia64/kernel/setup.c	2006-09-21 14:15:36.000000000 +0900
@@ -257,7 +257,7 @@
 
 #ifdef CONFIG_KEXEC
 	/* crashkernel=size@offset specifies the size to reserve for a crash
-	 * kernel.(offset is ingored for keep compatibility with other archs)
+	 * kernel. If offset is 0, then it is determined automatically.
 	 * By reserving this memory we guarantee that linux
 	 * never set's it up as a DMA target.
 	 * Useful for holding code to do something appropriate
@@ -268,7 +268,11 @@
 		unsigned long base, size;
 		if (from) {
 			size = memparse(from + 12, &from);
-			if (size) {
+			if (*from = '@')
+				base = memparse(from+1, &from);
+			else
+				base = 0;
+			if (size && !base) {
 				sort_regions(rsvd_region, n);
 				base = kdump_find_rsvd_region(size,
 				rsvd_region, n);
#3 lore raw
In reply to Horms
Quoted 54 lines

-----Original Message-----
From: Horms [mailto:[email protected]]
Sent: 2006Äê9ÔÂ21ÈÕ 13:33
To: Zou, Nanhai
Cc: Luck, Tony; Linux-IA64; fastboot
Subject: Re: Kexec/Kdump: honour non-zero crashkernel offset.

Quoted 3 lines · Zou Nan hai

This patch is incorrect.
I think you will get a compile warning of base is used before it is
aligned a value here...

Sorry about that. Is this better?

--
Horms
H: http://www.vergenet.net/~horms/
W: http://www.valinux.co.jp/en/

[IA64] Kexec/Kdump: honour non-zero crashkernel offset.

There seems to be a value in both allowing the kernel to determine the base offset of the crashkernel automatically and allowing users's to sepcify it.

The old behaviour on ia64, which is still the current behaviour on most architectures is for the user to always specify the address. Recently ia64 was changed so that it is always automatically determined.

With this patch the kernel automatically determines the offset if the supplied value is 0, otherwise it uses the value provided.

This should probably be backed by a documentation change.

Index: linux-2.6-ia64/arch/ia64/kernel/setup.c
=================================> --- linux-2.6-ia64.orig/arch/ia64/kernel/setup.c 2006-09-21
14:08:20.000000000 +0900
+++ linux-2.6-ia64/arch/ia64/kernel/setup.c 2006-09-21
14:15:36.000000000 +0900

(quoted diff) +1 -1
@@ -257,7 +257,7 @@
 
 #ifdef CONFIG_KEXEC
 	/* crashkernel=size@offset specifies the size to reserve for a crash
-	 * kernel.(offset is ingored for keep compatibility with other archs)
+	 * kernel. If offset is 0, then it is determined automatically.
 	 * By reserving this memory we guarantee that linux
 	 * never set's it up as a DMA target.
 	 * Useful for holding code to do something appropriate
@@ -268,7 +268,11 @@
 		unsigned long base, size;
 		if (from) {

Review line 278

Quoted 10 lines
(quoted diff) +5 -1
 			size = memparse(from + 12, &from);
-			if (size) {
+			if (*from = '@')
+				base = memparse(from+1, &from);
+			else
+				base = 0;
+			if (size && !base) {
 				sort_regions(rsvd_region, n);
 				base = kdump_find_rsvd_region(size,
 				rsvd_region, n);

This patch is still incorrect. You will miss the rsvd_region code if base not equals 0.

Zou Nan hai

author
#4 lore raw
In reply to Horms
Quoted 1 line · Nanhai

This patch is still incorrect. You will miss the rsvd_region code if base not equals 0.

Thanks, I guess this is closer, though I'm assuming that calling sort_regions() isn't needed unless kdump_find_rsvd_region() is going to be called.

Signature · 52 lines
Horms
  H: http://www.vergenet.net/~horms/
  W: http://www.valinux.co.jp/en/

[IA64] Kexec/Kdump: honour non-zero crashkernel offset.

There seems to be a value in both allowing the kernel to determine
the base offset of the crashkernel automatically and allowing
users's to sepcify it.

The old behaviour on ia64, which is still the current behaviour on
most architectures is for the user to always specify the address.
Recently ia64 was changed so that it is always automatically determined.

With this patch the kernel automatically determines the offset if
the supplied value is 0, otherwise it uses the value provided.

This should probably be backed by a documentation change.

Signed-Off-By: Simon Horman <[email protected]>

Index: linux-2.6-ia64/arch/ia64/kernel/setup.c
=================================--- linux-2.6-ia64.orig/arch/ia64/kernel/setup.c	2006-09-21 14:08:20.000000000 +0900
+++ linux-2.6-ia64/arch/ia64/kernel/setup.c	2006-09-21 18:43:10.000000000 +0900
@@ -257,7 +257,7 @@
 
 #ifdef CONFIG_KEXEC
 	/* crashkernel=size@offset specifies the size to reserve for a crash
-	 * kernel.(offset is ingored for keep compatibility with other archs)
+	 * kernel. If offset is 0, then it is determined automatically.
 	 * By reserving this memory we guarantee that linux
 	 * never set's it up as a DMA target.
 	 * Useful for holding code to do something appropriate
@@ -268,10 +268,16 @@
 		unsigned long base, size;
 		if (from) {
 			size = memparse(from + 12, &from);
+			if (*from = '@')
+				base = memparse(from+1, &from);
+			else
+				base = 0;
 			if (size) {
-				sort_regions(rsvd_region, n);
-				base = kdump_find_rsvd_region(size,
-				rsvd_region, n);
+				if (!base) {
+					sort_regions(rsvd_region, n);
+					base = kdump_find_rsvd_region(size,
+								rsvd_region, n);
+				}
 				if (base != ~0UL) {
 					rsvd_region[n].start  						(unsigned long)__va(base);
author
#5 lore raw
In reply to Horms
Quoted 6 lines · Horms
Quoted 1 line · Nanhai

This patch is still incorrect. You will miss the rsvd_region code if base not equals 0.

Thanks, I guess this is closer, though I'm assuming that calling sort_regions() isn't needed unless kdump_find_rsvd_region() is going to be called.

Hi,

here is yet another version of this patch. This time I tested it and it actually seems to work (unlike previous itterations I sent).

Signature · 52 lines
Horms
  H: http://www.vergenet.net/~horms/
  W: http://www.valinux.co.jp/en/

[IA64] Kexec/Kdump: honour non-zero crashkernel offset.

There seems to be a value in both allowing the kernel to determine
the base offset of the crashkernel automatically and allowing
users's to sepcify it.

The old behaviour on ia64, which is still the current behaviour on
most architectures is for the user to always specify the address.
Recently ia64 was changed so that it is always automatically determined.

With this patch the kernel automatically determines the offset if
the supplied value is 0, otherwise it uses the value provided.

This should probably be backed by a documentation change.

Signed-Off-By: Simon Horman <[email protected]>

Index: kexec-ia64-2.6/arch/ia64/kernel/setup.c
=================================--- kexec-ia64-2.6.orig/arch/ia64/kernel/setup.c	2006-10-20 13:13:38.000000000 +0900
+++ kexec-ia64-2.6/arch/ia64/kernel/setup.c	2006-10-20 14:49:05.000000000 +0900
@@ -257,7 +257,7 @@
 
 #ifdef CONFIG_KEXEC
 	/* crashkernel=size@offset specifies the size to reserve for a crash
-	 * kernel.(offset is ingored for keep compatibility with other archs)
+	 * kernel. If offset is 0, then it is determined automatically.
 	 * By reserving this memory we guarantee that linux never set's it
 	 * up as a DMA target.Useful for holding code to do something
 	 * appropriate after a kernel panic.
@@ -267,10 +267,16 @@
 		unsigned long base, size;
 		if (from) {
 			size = memparse(from + 12, &from);
+			if (*from = '@')
+				base = memparse(from+1, &from);
+			else
+				base = 0;
 			if (size) {
-				sort_regions(rsvd_region, n);
-				base = kdump_find_rsvd_region(size,
-				rsvd_region, n);
+				if (!base) {
+					sort_regions(rsvd_region, n);
+					base = kdump_find_rsvd_region(size,
+							      	rsvd_region, n);
+					}
 				if (base != ~0UL) {
 					rsvd_region[n].start  						(unsigned long)__va(base);
author
#6 lore raw

Re: Kexec/Kdump: honour non-zero crashkernel offset

In reply to Horms

[IA64] Kexec/Kdump: honour non-zero crashkernel offset.

There seems to be a value in both allowing the kernel to determine the base offset of the crashkernel automatically and allowing users's to sepcify it.

The old behaviour on ia64, which is still the current behaviour on most architectures is for the user to always specify the address. Recently ia64 was changed so that it is always automatically determined.

With this patch the kernel automatically determines the offset if the supplied value is 0, otherwise it uses the value provided.

This should probably be backed by a documentation change.

Index: linux-2.6/arch/ia64/kernel/setup.c
=================================--- linux-2.6.orig/arch/ia64/kernel/setup.c 2006-10-23 17:34:24.000000000 +0900
+++ linux-2.6/arch/ia64/kernel/setup.c 2006-10-23 17:34:56.000000000 +0900
@@ -257,7 +257,7 @@

#ifdef CONFIG_KEXEC
/* crashkernel=size@offset specifies the size to reserve for a crash
- * kernel.(offset is ingored for keep compatibility with other archs)
+ * kernel. If offset is 0, then it is determined automatically.
* By reserving this memory we guarantee that linux never set's it
* up as a DMA target.Useful for holding code to do something
* appropriate after a kernel panic.
@@ -267,10 +267,16 @@
unsigned long base, size;
if (from) {
size = memparse(from + 12, &from);
+ if (*from = '@')
+ base = memparse(from+1, &from);
+ else
+ base = 0;
if (size) {
- sort_regions(rsvd_region, n);
- base = kdump_find_rsvd_region(size,
- rsvd_region, n);
+ if (!base) {
+ sort_regions(rsvd_region, n);
+ base = kdump_find_rsvd_region(size,
+ rsvd_region, n);
+ }
if (base != ~0UL) {
rsvd_region[n].start (unsigned long)__va(base);