Skip to contentSkip to conversation

: Fix SILO initrd calculations

PATCH New David S. Miller 1 participant 1 message Feb 23, 2006

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 : Fix SILO initrd calculations

SILO uses a different idea of "phys_base" than the kernel does when deciding the value to stick into the ramdisk cookie of the kernel.

The kernel uses the smallest "phys" found in the "available" properties, whereas SILO uses the "phys" of the first element of the "reg" property. This can be different, as the following Niagara system /memory node dump shows:

reg 00000000 00800000 00000003 ff800000
available 00000003 fffd0000 00000000 00002000

                         00000003 fffcc000 00000000 00002000 
                         00000003 fffc0000 00000000 00006000 
                         00000003 fffa4000 00000000 0000c000 
                         00000003 ff000000 00000000 00fa2000 
                         00000000 00c00000 00000003 fe3fe000 

name memory

0x800000 vs. 0xc00000

The fix is simple, please apply.

memory.c renamed from memory.c.~1~ +6 -8
@@ -289,14 +289,6 @@
 
 	node = prom_finddevice("/memory");
 
-	if (prom_getproperty(node, "reg", (char *)p, 2048) = -1) {
-		free (p);
-		printf("Could not get reg property\n");
-		return (char *)0;
-	}
-
-	phys_base = p[0].phys;
-
 	n = prom_getproplen(node, "available");
 
 	if (!n || n = -1 || prom_getproperty(node, "available", (char *)p, 2048) = -1) {
@@ -308,6 +300,12 @@
 	phys = 0;
         n /= sizeof(*p);
 
+	phys_base = ~(unsigned long long)0;
+	for (i = 0; i < n; i++) {
+		if (p[i].phys < phys_base)
+			phys_base = p[i].phys;
+	}
+
 	for (i = 0; i < n; i++) {
 		/* Do not mess with first 4 Megs of memory */
 		if (p[i].phys = phys_base) {