Merge git://git.kernel.org/pub/scm/linux/kernel/git/cmetcalf/linux-tile
* git://git.kernel.org/pub/scm/linux/kernel/git/cmetcalf/linux-tile:
arch/tile: fix memchr() not to dereference memory for zero length
arch/tile: make glibc's sysconf(_SC_NPROCESSORS_CONF) work correctly
arch/tile: fix rwlock so would-be write lockers don't block new readers
diff --git a/arch/tile/kernel/setup.c b/arch/tile/kernel/setup.c
index fb0b3cb..f185736 100644
--- a/arch/tile/kernel/setup.c
+++ b/arch/tile/kernel/setup.c
@@ -840,7 +840,7 @@
for_each_online_node(i)
register_one_node(i);
- for_each_present_cpu(i)
+ for (i = 0; i < smp_height * smp_width; ++i)
register_cpu(&cpu_devices[i], i);
return 0;
diff --git a/arch/tile/lib/memchr_32.c b/arch/tile/lib/memchr_32.c
index 6235283..cc3d9ba 100644
--- a/arch/tile/lib/memchr_32.c
+++ b/arch/tile/lib/memchr_32.c
@@ -18,12 +18,24 @@
void *memchr(const void *s, int c, size_t n)
{
+ const uint32_t *last_word_ptr;
+ const uint32_t *p;
+ const char *last_byte_ptr;
+ uintptr_t s_int;
+ uint32_t goal, before_mask, v, bits;
+ char *ret;
+
+ if (__builtin_expect(n == 0, 0)) {
+ /* Don't dereference any memory if the array is empty. */
+ return NULL;
+ }
+
/* Get an aligned pointer. */
- const uintptr_t s_int = (uintptr_t) s;
- const uint32_t *p = (const uint32_t *)(s_int & -4);
+ s_int = (uintptr_t) s;
+ p = (const uint32_t *)(s_int & -4);
/* Create four copies of the byte for which we are looking. */
- const uint32_t goal = 0x01010101 * (uint8_t) c;
+ goal = 0x01010101 * (uint8_t) c;
/* Read the first word, but munge it so that bytes before the array
* will not match goal.
@@ -31,23 +43,14 @@
* Note that this shift count expression works because we know
* shift counts are taken mod 32.
*/
- const uint32_t before_mask = (1 << (s_int << 3)) - 1;
- uint32_t v = (*p | before_mask) ^ (goal & before_mask);
+ before_mask = (1 << (s_int << 3)) - 1;
+ v = (*p | before_mask) ^ (goal & before_mask);
/* Compute the address of the last byte. */
- const char *const last_byte_ptr = (const char *)s + n - 1;
+ last_byte_ptr = (const char *)s + n - 1;
/* Compute the address of the word containing the last byte. */
- const uint32_t *const last_word_ptr =
- (const uint32_t *)((uintptr_t) last_byte_ptr & -4);
-
- uint32_t bits;
- char *ret;
-
- if (__builtin_expect(n == 0, 0)) {
- /* Don't dereference any memory if the array is empty. */
- return NULL;
- }
+ last_word_ptr = (const uint32_t *)((uintptr_t) last_byte_ptr & -4);
while ((bits = __insn_seqb(v, goal)) == 0) {
if (__builtin_expect(p == last_word_ptr, 0)) {
diff --git a/arch/tile/lib/spinlock_32.c b/arch/tile/lib/spinlock_32.c
index 485e24d..5cd1c40 100644
--- a/arch/tile/lib/spinlock_32.c
+++ b/arch/tile/lib/spinlock_32.c
@@ -167,23 +167,30 @@
* when we compare them.
*/
u32 my_ticket_;
-
- /* Take out the next ticket; this will also stop would-be readers. */
- if (val & 1)
- val = get_rwlock(rwlock);
- rwlock->lock = __insn_addb(val, 1 << WR_NEXT_SHIFT);
-
- /* Extract my ticket value from the original word. */
- my_ticket_ = val >> WR_NEXT_SHIFT;
+ u32 iterations = 0;
/*
- * Wait until the "current" field matches our ticket, and
- * there are no remaining readers.
+ * Wait until there are no readers, then bump up the next
+ * field and capture the ticket value.
*/
for (;;) {
+ if (!(val & 1)) {
+ if ((val >> RD_COUNT_SHIFT) == 0)
+ break;
+ rwlock->lock = val;
+ }
+ delay_backoff(iterations++);
+ val = __insn_tns((int *)&rwlock->lock);
+ }
+
+ /* Take out the next ticket and extract my ticket value. */
+ rwlock->lock = __insn_addb(val, 1 << WR_NEXT_SHIFT);
+ my_ticket_ = val >> WR_NEXT_SHIFT;
+
+ /* Wait until the "current" field matches our ticket. */
+ for (;;) {
u32 curr_ = val >> WR_CURR_SHIFT;
- u32 readers = val >> RD_COUNT_SHIFT;
- u32 delta = ((my_ticket_ - curr_) & WR_MASK) + !!readers;
+ u32 delta = ((my_ticket_ - curr_) & WR_MASK);
if (likely(delta == 0))
break;
