sh: Make is_valid_bugaddr() more intelligent on nommu.
authorPaul Mundt <lethal@linux-sh.org>
Mon, 19 May 2008 10:32:07 +0000 (19:32 +0900)
committerPaul Mundt <lethal@linux-sh.org>
Mon, 19 May 2008 10:32:07 +0000 (19:32 +0900)
Currently is_valid_bugaddr() is true for anything >= PAGE_OFFSET, which
happens to be 0 on nommu configurations. Make this a bit smarter by just
reading in the opcode and comparing it against the trap type that we
already know. Follows the logic from avr32.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
arch/sh/kernel/traps.c

index a3bdc68ef02c75adcdbfaca4b5d61c3218c60521..438f1ebcc45330cc0a5d320fee82a4b6840bcf1d 100644 (file)
@@ -4,6 +4,7 @@
 #include <linux/kdebug.h>
 #include <linux/signal.h>
 #include <linux/sched.h>
+#include <linux/uaccess.h>
 #include <asm/system.h>
 
 #ifdef CONFIG_BUG
@@ -21,7 +22,14 @@ static void handle_BUG(struct pt_regs *regs)
 
 int is_valid_bugaddr(unsigned long addr)
 {
-       return addr >= PAGE_OFFSET;
+       unsigned short opcode;
+
+       if (addr < PAGE_OFFSET)
+               return 0;
+       if (probe_kernel_address((u16 *)addr, opcode))
+               return 0;
+
+       return opcode == TRAPA_BUG_OPCODE;
 }
 #endif