On Sat, Jul 22, 2023 at 05:26:07PM +0200, Andreas Schwab wrote:
On Jul 22 2023, Matthew Wilcox wrote:
On Sat, Jul 22, 2023 at 08:24:34AM +0200, Andreas Schwab wrote:
On Jul 21 2023, Michael Schmitz wrote:
static inline bool clear_bit_unlock_is_negative_byte(unsigned int nr,
volatile unsigned long *p)
{
unsigned char *cp = (unsigned char *) p;
char result;
char mask = 1 << nr; /* nr guaranteed to be < 7 */
__asm__ __volatile__ ("eori.b %1, %2; smi %0"
: "=d" (result)
: "i" (mask), "o" (*(cp+3))
That should use "id" as constraint, so that the compiler can share the
constant with other insns. Also, the third operand is modified, so it
needs to be marked as in/out.
The "o" constraint is shared with bfset_mem_set_bit,
bfclr_mem_test_and_clear_bit and a few other functions. Should they
all be changed?
They use a memory clobber, because the "o" constraint cannot accurately
describe the situation there.
My understanding is that clear_bit_unlock_is_negative_byte() also
needs the "memory" clobber because it has "release" semantics (ie
all previous stores must be visible before this store is visible).
static inline bool clear_bit_unlock_is_negative_byte(unsigned int nr,
volatile unsigned long *p)
{
char result;
char mask = 1 << nr; /* nr guaranteed to be < 7 */
char *cp = (char *)p + 3; /* m68k is big-endian */
__asm__ __volatile__ ("eori.b %1, %2; smi %0"
: "=d" (result)
: "di" (mask), "o" (*cp)
: "memory");
return result;
}
is what I currently have.
