Re: convert_fxsr_from_user

[Andrew Morton]

Dave Jones <[e-mail]> wrote:
>
> Andrew,
>  A while back you optimised this routine to not do lots of memory
> copies.  I've noticed it does no checking on the validity of the
> addresses it dereferences from userspace.

It never has performed those checks.   The check is in the caller,
arch/i386/kernel/signal.c:restore_i387.

Bless you for merging Jon's uaccess.h documentation patch. 

My __get_user()'s are arse-about.

diff -puN arch/i386/kernel/i387.c~convert_fxsr_from_user-get_user-fixes arch/i386/kernel/i387.c
--- 25/arch/i386/kernel/i387.c~convert_fxsr_from_user-get_user-fixes	Tue Apr  8 16:47:08 2003
+++ 25-akpm/arch/i386/kernel/i387.c	Tue Apr  8 16:48:13 2003
@@ -275,9 +275,9 @@ static int convert_fxsr_from_user( struc
 		unsigned long *t = (unsigned long *)to;
 		unsigned long *f = (unsigned long *)from;
 
-		if (__get_user(*f, t) ||
-				__get_user(*(f + 1), t + 1) ||
-				__get_user(from->exponent, &to->exponent))
+		if (__get_user(*t, f) ||
+				__get_user(*(t + 1), f + 1) ||
+				__get_user(to->exponent, &from->exponent))
 			return 1;
 	}
 	return 0;

_