On 29-Nov-01 Peter Pentchev wrote:
> On Wed, Nov 21, 2001 at 11:38:35AM -0800, John Baldwin wrote:
>>
>> On 21-Nov-01 David Xu wrote:
>> > 4.4-stable, file sys/i386/include/cpufunc.h,
>> >
>> > --- cpufunc.h.orig Wed Nov 21 13:35:36 2001
>> > +++ cpufunc.h Wed Nov 21 15:00:12 2001
>> > @@ -72,7 +72,7 @@
>> > {
>> > u_int result;
>> >
>> > - __asm __volatile("bsfl %0,%0" : "=r" (result) : "0" (mask));
>> > + __asm __volatile("bsfl %0,%0" : "=r" (result) : "0" (mask) :
>> > "cc");
>> > return (result);
>> > }
>> >
>> > @@ -81,7 +81,7 @@
>> > {
>> > u_int result;
>> >
>> > - __asm __volatile("bsrl %0,%0" : "=r" (result) : "0" (mask));
>> > + __asm __volatile("bsrl %0,%0" : "=r" (result) : "0" (mask) :
>> > "cc");
>> > return (result);
>> > }
>> >
>> > @@ -305,7 +305,7 @@
>> > u_int result;
>> >
>> > __asm __volatile("xorl %0,%0; xchgl %1,%0"
>> > - : "=&r" (result) : "m" (*addr));
>> > + : "=&r" (result) : "m" (*addr) : "cc");
>> > return (result);
>> > }
>> >
>>
>> Have you had actual bugs as a result of "cc" not being in the constraints?
>>
>> If so, there's a _lot_ more places that need this. All the atomic ops, for
>> example.
>
> How about PR gnu/32365? I know next to nothing about assembler opcodes,
> if "cc" would fix the PR, then the PR is a demonstration of an actual bug.
>
> G'luck,
> Peter
That PR is a bug in the actual compiler itself. I do have a rather largish
patch of asm constraint fixes all across the sys/i386 tree in the kernel at
www.freebsd.org/~jhb/patches/i386_asm.patch that includes these patches among
other things.
--
John Baldwin <[EMAIL PROTECTED]> <>< http://www.FreeBSD.org/~jhb/
"Power Users Use the Power to Serve!" - http://www.FreeBSD.org/
To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-hackers" in the body of the message
