Re: [PATCH] Put the BUG __FILE__ and __LINE__ info out of line

[Hugh Dickins <hugh@veritas.com>]

On Thu, 28 Sep 2006, Andi Kleen wrote:
> On Thu, Sep 28, 2006 at 03:30:12AM -0700, Jeremy Fitzhardinge wrote:
> > Andi Kleen wrote:
> > >But no out of line section. So overall it's smaller, although the cache 
> > >footprint
> > >is 2 bytes larger. But then is 2 bytes larger really an issue? We don't 
> > >have
> > >_that_ many BUGs anyways.
> > >  
> > 
> > I think the out of line section is a feature; no point in crufting up 
> > the icache with BUG gunk, especially since a number of them are on 
> > fairly hot paths.
> 
> It's 10 bytes per BUG. 

Or 9 bytes per BUG: I protested about the disassembly problem back
when the minimized BUG() first went in, and have been using "ljmp"
in my i386 builds ever since:

--- 2.6.18-git9/arch/i386/kernel/traps.c	2006-09-28 12:03:47.000000000 +0100
+++ linux/arch/i386/kernel/traps.c	2006-09-28 16:23:56.000000000 +0100
@@ -422,10 +422,10 @@ static void handle_BUG(struct pt_regs *r
 		char *file;
 		char c;
 
-		if (probe_kernel_address((unsigned short __user *)(eip + 2),
+		if (probe_kernel_address((unsigned short __user *)(eip + 7),
 					line))
 			break;
-		if (__get_user(file, (char * __user *)(eip + 4)) ||
+		if (__get_user(file, (char * __user *)(eip + 3)) ||
 		    (unsigned long)file < PAGE_OFFSET || __get_user(c, file))
 			file = "<bad filename>";
 
--- 2.6.18-git9/include/asm-i386/bug.h	2006-09-20 04:42:06.000000000 +0100
+++ linux/include/asm-i386/bug.h	2006-09-28 16:23:56.000000000 +0100
@@ -13,9 +13,11 @@
 #ifdef CONFIG_DEBUG_BUGVERBOSE
 #define BUG()				\
  __asm__ __volatile__(	"ud2\n"		\
-			"\t.word %c0\n"	\
-			"\t.long %c1\n"	\
-			 : : "i" (__LINE__), "i" (__FILE__))
+			"\t.byte 0xea\n" \
+			"\t.long %c0\n"	\
+			"\t.word %c1\n"	\
+			 : : "i" (__FILE__), "i" (__LINE__))
+	/* "ljmp long short" so disassemblers can make sense of it */
 #else
 #define BUG() __asm__ __volatile__("ud2\n")
 #endif