Skip to contentSkip to conversation

shrink address space reserved for kmap

PATCH New Manfred Spraul 1 participant 1 message Jan 14, 2004

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 shrink address space reserved for kmap

If highmem is enabled, 32 MB at the top of the address space is reserved for kmap. This is not necessary: the kmap area is at most 4 MB long, reserving 8 MB is sufficient (the last few kB are needed for the fixmaps, and the kmap area must be 4 MB-aligned). Recovering these 24 MB is important - it's increases the size of the vmalloc pool. 2.6 contains a similar patch.
The patch also adds a safety check to detect a collision between fixmap and kmap mappings - just in case someone uses huge fixmaps.

Marcelo, could you add it to your tree?

--
Manfred

// $Header$
// Kernel Version:

//  VERSION = 2
//  PATCHLEVEL = 4
//  SUBLEVEL = 21
//  EXTRAVERSION = -pre2
include/asm-i386/highmem.h +1 -1
@@ -46,7 +46,7 @@
  * easily, subsequent pte tables have to be allocated in one physical
  * chunk of RAM.
  */
-#define PKMAP_BASE (0xfe000000UL)
+#define PKMAP_BASE (0xff800000UL)
 #ifdef CONFIG_X86_PAE
 #define LAST_PKMAP 512
 #else
arch/i386/mm/init.c +9 -1
@@ -510,7 +510,15 @@
 
 	if (!mem_map)
 		BUG();
-	
+#ifdef CONFIG_HIGHMEM
+	/* check that fixmap and pkmap do not overlap */
+	if (PKMAP_BASE+LAST_PKMAP*PAGE_SIZE >= FIXADDR_START) {
+		printk(KERN_ERR "fixmap and kmap areas overlap - this will crash\n");
+		printk(KERN_ERR "pkstart: %lxh pkend: %lxh fixstart %lxh\n",
+				PKMAP_BASE, PKMAP_BASE+LAST_PKMAP*PAGE_SIZE, FIXADDR_START);
+		BUG();
+	}
+#endif
 	set_max_mapnr_init();
 
 	high_memory = (void *) __va(max_low_pfn * PAGE_SIZE);