[PATCH] Fix /proc/kcore for i386

[Andi Kleen]

The recent IA64 changes for /proc/kcore broke the access on i386.
Currently no notes are written for the direct mapped or vmalloced
memory, which makes gdb reject it.

This patch fixes it. Other ports probably need to do the same changes.

There is still another problem in /proc/kcore for which I am submitting a 
separate patch.

-Andi

diff -u linux-2.5.72-work/arch/i386/mm/init.c-o linux-2.5.72-work/arch/i386/mm/init.c
--- linux-2.5.72-work/arch/i386/mm/init.c-o	2003-05-27 03:00:45.000000000 +0200
+++ linux-2.5.72-work/arch/i386/mm/init.c	2003-06-17 15:55:03.000000000 +0200
@@ -27,6 +27,7 @@
 #include <linux/pagemap.h>
 #include <linux/bootmem.h>
 #include <linux/slab.h>
+#include <linux/proc_fs.h>
 
 #include <asm/processor.h>
 #include <asm/system.h>
@@ -425,6 +426,8 @@
 extern void set_max_mapnr_init(void);
 #endif /* !CONFIG_DISCONTIGMEM */
 
+static struct kcore_list kcore_mem, kcore_vmalloc; 
+
 void __init mem_init(void)
 {
 	extern int ppro_with_ram_bug(void);
@@ -477,6 +480,10 @@
 	datasize =  (unsigned long) &_edata - (unsigned long) &_etext;
 	initsize =  (unsigned long) &__init_end - (unsigned long) &__init_begin;
 
+	kclist_add(&kcore_mem, __va(0), max_low_pfn << PAGE_SHIFT); 
+	kclist_add(&kcore_vmalloc, (void *)VMALLOC_START, 
+		   VMALLOC_END-VMALLOC_START);
+
 	printk(KERN_INFO "Memory: %luk/%luk available (%dk kernel code, %dk reserved, %dk data, %dk init, %ldk highmem)\n",
 		(unsigned long) nr_free_pages() << (PAGE_SHIFT-10),
 		num_physpages << (PAGE_SHIFT-10),