[PATCH] x86_64 kernel mapping fix
authorKeith Mannthey <kmannth@us.ibm.com>
Tue, 26 Sep 2006 08:52:36 +0000 (10:52 +0200)
committerAndi Kleen <andi@basil.nowhere.org>
Tue, 26 Sep 2006 08:52:36 +0000 (10:52 +0200)
Fix for the x86_64 kernel mapping code.  Without this patch the update path
only inits one pmd_page worth of memory and tramples any entries on it.  now
the calling convention to phys_pmd_init and phys_init is to always pass a
[pmd/pud] page not an offset within a page.

Signed-off-by: Keith Mannthey<kmannth@us.ibm.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
arch/x86_64/mm/init.c

index d40134bd639928d370b32fd1bd9cd28e600e8147..984155b75e4c03f8ee14e32f57a9b03d226a610c 100644 (file)
@@ -250,12 +250,13 @@ __init void early_iounmap(void *addr, unsigned long size)
 }
 
 static void __meminit
-phys_pmd_init(pmd_t *pmd, unsigned long address, unsigned long end)
+phys_pmd_init(pmd_t *pmd_page, unsigned long address, unsigned long end)
 {
-       int i;
+       int i = pmd_index(address);
 
-       for (i = 0; i < PTRS_PER_PMD; pmd++, i++, address += PMD_SIZE) {
+       for (; i < PTRS_PER_PMD; i++, address += PMD_SIZE) {
                unsigned long entry;
+               pmd_t *pmd = pmd_page + pmd_index(address);
 
                if (address >= end) {
                        if (!after_bootmem)
@@ -263,6 +264,10 @@ phys_pmd_init(pmd_t *pmd, unsigned long address, unsigned long end)
                                        set_pmd(pmd, __pmd(0));
                        break;
                }
+
+               if (pmd_val(*pmd))
+                       continue;
+
                entry = _PAGE_NX|_PAGE_PSE|_KERNPG_TABLE|_PAGE_GLOBAL|address;
                entry &= __supported_pte_mask;
                set_pmd(pmd, __pmd(entry));
@@ -272,45 +277,41 @@ phys_pmd_init(pmd_t *pmd, unsigned long address, unsigned long end)
 static void __meminit
 phys_pmd_update(pud_t *pud, unsigned long address, unsigned long end)
 {
-       pmd_t *pmd = pmd_offset(pud, (unsigned long)__va(address));
-
-       if (pmd_none(*pmd)) {
-               spin_lock(&init_mm.page_table_lock);
-               phys_pmd_init(pmd, address, end);
-               spin_unlock(&init_mm.page_table_lock);
-               __flush_tlb_all();
-       }
+       pmd_t *pmd = pmd_offset(pud,0);
+       spin_lock(&init_mm.page_table_lock);
+       phys_pmd_init(pmd, address, end);
+       spin_unlock(&init_mm.page_table_lock);
+       __flush_tlb_all();
 }
 
-static void __meminit phys_pud_init(pud_t *pud, unsigned long address, unsigned long end)
+static void __meminit phys_pud_init(pud_t *pud_page, unsigned long addr, unsigned long end)
 { 
-       long i = pud_index(address);
-
-       pud = pud + i;
+       int i = pud_index(addr);
 
-       if (after_bootmem && pud_val(*pud)) {
-               phys_pmd_update(pud, address, end);
-               return;
-       }
 
-       for (; i < PTRS_PER_PUD; pud++, i++) {
+       for (; i < PTRS_PER_PUD; i++, addr = (addr & PUD_MASK) + PUD_SIZE ) {
                int map; 
-               unsigned long paddr, pmd_phys;
+               unsigned long pmd_phys;
+               pud_t *pud = pud_page + pud_index(addr);
                pmd_t *pmd;
 
-               paddr = (address & PGDIR_MASK) + i*PUD_SIZE;
-               if (paddr >= end)
+               if (addr >= end)
                        break;
 
-               if (!after_bootmem && !e820_any_mapped(paddr, paddr+PUD_SIZE, 0)) {
+               if (!after_bootmem && !e820_any_mapped(addr,addr+PUD_SIZE,0)) {
                        set_pud(pud, __pud(0)); 
                        continue;
                } 
 
+               if (pud_val(*pud)) {
+                       phys_pmd_update(pud, addr, end);
+                       continue;
+               }
+
                pmd = alloc_low_page(&map, &pmd_phys);
                spin_lock(&init_mm.page_table_lock);
                set_pud(pud, __pud(pmd_phys | _KERNPG_TABLE));
-               phys_pmd_init(pmd, paddr, end);
+               phys_pmd_init(pmd, addr, end);
                spin_unlock(&init_mm.page_table_lock);
                unmap_low_page(map);
        }