Drop might_sleep() calls from get_user() & co. This should fix the issue

in http://www.linux-mips.org/cgi-bin/mesg.cgi?a=linux-mips&i=200508171321.20094.Joshua.Wise%40sicortex.com and it's the right thing to do anyway because
it was inflating those functions way too much.
Signed-off-by: NRalf Baechle <ralf@linux-mips.org>

include/asm-mips/uaccess.h

@@ -213,7 +213,6 @@ struct __large_struct { unsigned long buf[100]; };
 	__typeof(*(ptr)) __gu_val =  (__typeof(*(ptr))) 0;		\
 	long __gu_err = 0;						\
 									\
-	might_sleep();							\
 	switch (size) {							\
 	case 1: __get_user_asm("lb", ptr); break;			\
 	case 2: __get_user_asm("lh", ptr); break;			\
@@ -300,7 +299,6 @@ extern void __get_user_unknown(void);
 	__typeof__(*(ptr)) __pu_val;					\
 	long __pu_err = 0;						\
 									\
-	might_sleep();							\
 	__pu_val = (x);							\
 	switch (size) {							\
 	case 1: __put_user_asm("sb", ptr); break;			\