Re: [PATCH] kernel parameter vmalloc size fix

["Dave Young" <hidave.darkstar@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 1544 bytes --]

On Mon, Jun 16, 2008 at 4:01 PM, Ingo Molnar <mingo@elte.hu> wrote:
>
> * Dave Young <hidave.darkstar@gmail.com> wrote:
>
>> booting kernel with vmalloc=[any size<=16m] will oops.
>>
>> It's due to the vm area hole.
>>
>> In include/asm-x86/pgtable_32.h:
>> #define VMALLOC_OFFSET        (8 * 1024 * 1024)
>> #define VMALLOC_START (((unsigned long)high_memory + 2 * VMALLOC_OFFSET - 1) \
>>                        & ~(VMALLOC_OFFSET - 1))
>>
>> BUG_ON in arch/x86/mm/init_32.c will be triggered:
>> BUG_ON((unsigned long)high_memory             > VMALLOC_START);
>>
>> Fixed by return -EINVAL for invalid parameter
>
> hm. Why dont we instead add the size of the hole to the
> __VMALLOC_RESERVE value instead? There's nothing inherently bad about
> using vmalloc=16m. The VM area hole is really a kernel-internal
> abstraction that should not be visible in the usage of the parameter.

I do some test about this last weekend, there's some questions,  could
you help to fix it?

1. MAXMEM :
 (-__PAGE_OFFSET - __VMALLOC_RESERVE).
The space after VMALLOC_END is included as well, seting it to
(VMALLOC_END - PAGE_OFFSET - __VMALLOC_RESERVE), is it right?

2. VMALLOC_OFFSET is not considered in __VMALLOC_RESERVE
Should fixed by adding VMALLOC_OFFSET to it.

3. VMALLOC_START :
 (((unsigned long)high_memory + 2 * VMALLOC_OFFSET - 1) & ~(VMALLOC_OFFSET - 1))
So it's not always 8M, bigger than 8M possible.
Set it to ((unsigned long)high_memory + VMALLOC_OFFSET), is it right?

Attached the proposed patch. please give some advice.

Regards
dave

[-- Attachment #2: diff.vmalloc --]
[-- Type: application/octet-stream, Size: 1949 bytes --]

diff -upr linux/arch/x86/kernel/setup_32.c linux.new/arch/x86/kernel/setup_32.c
--- linux/arch/x86/kernel/setup_32.c	2008-06-24 10:16:10.000000000 +0800
+++ linux.new/arch/x86/kernel/setup_32.c	2008-06-24 10:20:28.000000000 +0800
@@ -313,7 +313,8 @@ static int __init parse_vmalloc(char *ar
 	if (!arg)
 		return -EINVAL;
 
-	__VMALLOC_RESERVE = memparse(arg, &arg);
+	/* Add VMALLOC_OFFSET to the parsed value due to vm area guard hole*/
+	__VMALLOC_RESERVE = memparse(arg, &arg) + VMALLOC_OFFSET;
 	return 0;
 }
 early_param("vmalloc", parse_vmalloc);
diff -upr linux/include/asm-x86/page_32.h linux.new/include/asm-x86/page_32.h
--- linux/include/asm-x86/page_32.h	2008-06-24 10:16:34.000000000 +0800
+++ linux.new/include/asm-x86/page_32.h	2008-06-24 10:17:49.000000000 +0800
@@ -82,7 +82,6 @@ extern unsigned int __VMALLOC_RESERVE;
 extern int sysctl_legacy_va_layout;
 
 #define VMALLOC_RESERVE		((unsigned long)__VMALLOC_RESERVE)
-#define MAXMEM			(-__PAGE_OFFSET - __VMALLOC_RESERVE)
 
 #ifdef CONFIG_X86_USE_3DNOW
 #include <asm/mmx.h>
diff -upr linux/include/asm-x86/pgtable_32.h linux.new/include/asm-x86/pgtable_32.h
--- linux/include/asm-x86/pgtable_32.h	2008-06-24 10:16:42.000000000 +0800
+++ linux.new/include/asm-x86/pgtable_32.h	2008-06-24 11:46:39.000000000 +0800
@@ -56,8 +56,7 @@ void paging_init(void);
  * area for the same reason. ;)
  */
 #define VMALLOC_OFFSET	(8 * 1024 * 1024)
-#define VMALLOC_START	(((unsigned long)high_memory + 2 * VMALLOC_OFFSET - 1) \
-			 & ~(VMALLOC_OFFSET - 1))
+#define VMALLOC_START	((unsigned long)high_memory + VMALLOC_OFFSET)
 #ifdef CONFIG_X86_PAE
 #define LAST_PKMAP 512
 #else
@@ -73,6 +72,8 @@ void paging_init(void);
 # define VMALLOC_END	(FIXADDR_START - 2 * PAGE_SIZE)
 #endif
 
+#define MAXMEM	(VMALLOC_END - PAGE_OFFSET - __VMALLOC_RESERVE)
+
 /*
  * Define this if things work differently on an i386 and an i486:
  * it will (on an i486) warn about kernel memory accesses that are