Merge tag 'sched_urgent_for_v5.15_rc1' of git://git.kernel.org/pub/scm/linux/kernel...
[linux-2.6-microblaze.git] / arch / arc / include / asm / atomic.h
index 7a36d79..52ee51e 100644 (file)
 #define arch_atomic_read(v)  READ_ONCE((v)->counter)
 
 #ifdef CONFIG_ARC_HAS_LLSC
-
-#define arch_atomic_set(v, i) WRITE_ONCE(((v)->counter), (i))
-
-#define ATOMIC_OP(op, c_op, asm_op)                                    \
-static inline void arch_atomic_##op(int i, atomic_t *v)                        \
-{                                                                      \
-       unsigned int val;                                               \
-                                                                       \
-       __asm__ __volatile__(                                           \
-       "1:     llock   %[val], [%[ctr]]                \n"             \
-       "       " #asm_op " %[val], %[val], %[i]        \n"             \
-       "       scond   %[val], [%[ctr]]                \n"             \
-       "       bnz     1b                              \n"             \
-       : [val] "=&r"   (val) /* Early clobber to prevent reg reuse */  \
-       : [ctr] "r"     (&v->counter), /* Not "m": llock only supports reg direct addr mode */  \
-         [i]   "ir"    (i)                                             \
-       : "cc");                                                        \
-}                                                                      \
-
-#define ATOMIC_OP_RETURN(op, c_op, asm_op)                             \
-static inline int arch_atomic_##op##_return(int i, atomic_t *v)                \
-{                                                                      \
-       unsigned int val;                                               \
-                                                                       \
-       /*                                                              \
-        * Explicit full memory barrier needed before/after as          \
-        * LLOCK/SCOND themselves don't provide any such semantics      \
-        */                                                             \
-       smp_mb();                                                       \
-                                                                       \
-       __asm__ __volatile__(                                           \
-       "1:     llock   %[val], [%[ctr]]                \n"             \
-       "       " #asm_op " %[val], %[val], %[i]        \n"             \
-       "       scond   %[val], [%[ctr]]                \n"             \
-       "       bnz     1b                              \n"             \
-       : [val] "=&r"   (val)                                           \
-       : [ctr] "r"     (&v->counter),                                  \
-         [i]   "ir"    (i)                                             \
-       : "cc");                                                        \
-                                                                       \
-       smp_mb();                                                       \
-                                                                       \
-       return val;                                                     \
-}
-
-#define ATOMIC_FETCH_OP(op, c_op, asm_op)                              \
-static inline int arch_atomic_fetch_##op(int i, atomic_t *v)           \
-{                                                                      \
-       unsigned int val, orig;                                         \
-                                                                       \
-       /*                                                              \
-        * Explicit full memory barrier needed before/after as          \
-        * LLOCK/SCOND themselves don't provide any such semantics      \
-        */                                                             \
-       smp_mb();                                                       \
-                                                                       \
-       __asm__ __volatile__(                                           \
-       "1:     llock   %[orig], [%[ctr]]               \n"             \
-       "       " #asm_op " %[val], %[orig], %[i]       \n"             \
-       "       scond   %[val], [%[ctr]]                \n"             \
-       "       bnz     1b                              \n"             \
-       : [val] "=&r"   (val),                                          \
-         [orig] "=&r" (orig)                                           \
-       : [ctr] "r"     (&v->counter),                                  \
-         [i]   "ir"    (i)                                             \
-       : "cc");                                                        \
-                                                                       \
-       smp_mb();                                                       \
-                                                                       \
-       return orig;                                                    \
-}
-
-#else  /* !CONFIG_ARC_HAS_LLSC */
-
-#ifndef CONFIG_SMP
-
- /* violating atomic_xxx API locking protocol in UP for optimization sake */
-#define arch_atomic_set(v, i) WRITE_ONCE(((v)->counter), (i))
-
+#include <asm/atomic-llsc.h>
 #else
+#include <asm/atomic-spinlock.h>
+#endif
 
-static inline void arch_atomic_set(atomic_t *v, int i)
-{
-       /*
-        * Independent of hardware support, all of the atomic_xxx() APIs need
-        * to follow the same locking rules to make sure that a "hardware"
-        * atomic insn (e.g. LD) doesn't clobber an "emulated" atomic insn
-        * sequence
-        *
-        * Thus atomic_set() despite being 1 insn (and seemingly atomic)
-        * requires the locking.
-        */
-       unsigned long flags;
+#define arch_atomic_cmpxchg(v, o, n)                                   \
+({                                                                     \
+       arch_cmpxchg(&((v)->counter), (o), (n));                        \
+})
 
-       atomic_ops_lock(flags);
-       WRITE_ONCE(v->counter, i);
-       atomic_ops_unlock(flags);
-}
+#ifdef arch_cmpxchg_relaxed
+#define arch_atomic_cmpxchg_relaxed(v, o, n)                           \
+({                                                                     \
+       arch_cmpxchg_relaxed(&((v)->counter), (o), (n));                \
+})
+#endif
 
-#define arch_atomic_set_release(v, i)  arch_atomic_set((v), (i))
+#define arch_atomic_xchg(v, n)                                         \
+({                                                                     \
+       arch_xchg(&((v)->counter), (n));                                \
+})
 
+#ifdef arch_xchg_relaxed
+#define arch_atomic_xchg_relaxed(v, n)                                 \
+({                                                                     \
+       arch_xchg_relaxed(&((v)->counter), (n));                        \
+})
 #endif
 
 /*
- * Non hardware assisted Atomic-R-M-W
- * Locking would change to irq-disabling only (UP) and spinlocks (SMP)
+ * 64-bit atomics
  */
-
-#define ATOMIC_OP(op, c_op, asm_op)                                    \
-static inline void arch_atomic_##op(int i, atomic_t *v)                        \
-{                                                                      \
-       unsigned long flags;                                            \
-                                                                       \
-       atomic_ops_lock(flags);                                         \
-       v->counter c_op i;                                              \
-       atomic_ops_unlock(flags);                                       \
-}
-
-#define ATOMIC_OP_RETURN(op, c_op, asm_op)                             \
-static inline int arch_atomic_##op##_return(int i, atomic_t *v)                \
-{                                                                      \
-       unsigned long flags;                                            \
-       unsigned long temp;                                             \
-                                                                       \
-       /*                                                              \
-        * spin lock/unlock provides the needed smp_mb() before/after   \
-        */                                                             \
-       atomic_ops_lock(flags);                                         \
-       temp = v->counter;                                              \
-       temp c_op i;                                                    \
-       v->counter = temp;                                              \
-       atomic_ops_unlock(flags);                                       \
-                                                                       \
-       return temp;                                                    \
-}
-
-#define ATOMIC_FETCH_OP(op, c_op, asm_op)                              \
-static inline int arch_atomic_fetch_##op(int i, atomic_t *v)           \
-{                                                                      \
-       unsigned long flags;                                            \
-       unsigned long orig;                                             \
-                                                                       \
-       /*                                                              \
-        * spin lock/unlock provides the needed smp_mb() before/after   \
-        */                                                             \
-       atomic_ops_lock(flags);                                         \
-       orig = v->counter;                                              \
-       v->counter c_op i;                                              \
-       atomic_ops_unlock(flags);                                       \
-                                                                       \
-       return orig;                                                    \
-}
-
-#endif /* !CONFIG_ARC_HAS_LLSC */
-
-#define ATOMIC_OPS(op, c_op, asm_op)                                   \
-       ATOMIC_OP(op, c_op, asm_op)                                     \
-       ATOMIC_OP_RETURN(op, c_op, asm_op)                              \
-       ATOMIC_FETCH_OP(op, c_op, asm_op)
-
-ATOMIC_OPS(add, +=, add)
-ATOMIC_OPS(sub, -=, sub)
-
-#undef ATOMIC_OPS
-#define ATOMIC_OPS(op, c_op, asm_op)                                   \
-       ATOMIC_OP(op, c_op, asm_op)                                     \
-       ATOMIC_FETCH_OP(op, c_op, asm_op)
-
-ATOMIC_OPS(and, &=, and)
-ATOMIC_OPS(andnot, &= ~, bic)
-ATOMIC_OPS(or, |=, or)
-ATOMIC_OPS(xor, ^=, xor)
-
-#define arch_atomic_andnot             arch_atomic_andnot
-#define arch_atomic_fetch_andnot       arch_atomic_fetch_andnot
-
-#undef ATOMIC_OPS
-#undef ATOMIC_FETCH_OP
-#undef ATOMIC_OP_RETURN
-#undef ATOMIC_OP
-
 #ifdef CONFIG_GENERIC_ATOMIC64
-
 #include <asm-generic/atomic64.h>
-
-#else  /* Kconfig ensures this is only enabled with needed h/w assist */
-
-/*
- * ARCv2 supports 64-bit exclusive load (LLOCKD) / store (SCONDD)
- *  - The address HAS to be 64-bit aligned
- *  - There are 2 semantics involved here:
- *    = exclusive implies no interim update between load/store to same addr
- *    = both words are observed/updated together: this is guaranteed even
- *      for regular 64-bit load (LDD) / store (STD). Thus atomic64_set()
- *      is NOT required to use LLOCKD+SCONDD, STD suffices
- */
-
-typedef struct {
-       s64 __aligned(8) counter;
-} atomic64_t;
-
-#define ATOMIC64_INIT(a) { (a) }
-
-static inline s64 arch_atomic64_read(const atomic64_t *v)
-{
-       s64 val;
-
-       __asm__ __volatile__(
-       "       ldd   %0, [%1]  \n"
-       : "=r"(val)
-       : "r"(&v->counter));
-
-       return val;
-}
-
-static inline void arch_atomic64_set(atomic64_t *v, s64 a)
-{
-       /*
-        * This could have been a simple assignment in "C" but would need
-        * explicit volatile. Otherwise gcc optimizers could elide the store
-        * which borked atomic64 self-test
-        * In the inline asm version, memory clobber needed for exact same
-        * reason, to tell gcc about the store.
-        *
-        * This however is not needed for sibling atomic64_add() etc since both
-        * load/store are explicitly done in inline asm. As long as API is used
-        * for each access, gcc has no way to optimize away any load/store
-        */
-       __asm__ __volatile__(
-       "       std   %0, [%1]  \n"
-       :
-       : "r"(a), "r"(&v->counter)
-       : "memory");
-}
-
-#define ATOMIC64_OP(op, op1, op2)                                      \
-static inline void arch_atomic64_##op(s64 a, atomic64_t *v)            \
-{                                                                      \
-       s64 val;                                                        \
-                                                                       \
-       __asm__ __volatile__(                                           \
-       "1:                             \n"                             \
-       "       llockd  %0, [%1]        \n"                             \
-       "       " #op1 " %L0, %L0, %L2  \n"                             \
-       "       " #op2 " %H0, %H0, %H2  \n"                             \
-       "       scondd   %0, [%1]       \n"                             \
-       "       bnz     1b              \n"                             \
-       : "=&r"(val)                                                    \
-       : "r"(&v->counter), "ir"(a)                                     \
-       : "cc");                                                        \
-}                                                                      \
-
-#define ATOMIC64_OP_RETURN(op, op1, op2)                               \
-static inline s64 arch_atomic64_##op##_return(s64 a, atomic64_t *v)    \
-{                                                                      \
-       s64 val;                                                        \
-                                                                       \
-       smp_mb();                                                       \
-                                                                       \
-       __asm__ __volatile__(                                           \
-       "1:                             \n"                             \
-       "       llockd   %0, [%1]       \n"                             \
-       "       " #op1 " %L0, %L0, %L2  \n"                             \
-       "       " #op2 " %H0, %H0, %H2  \n"                             \
-       "       scondd   %0, [%1]       \n"                             \
-       "       bnz     1b              \n"                             \
-       : [val] "=&r"(val)                                              \
-       : "r"(&v->counter), "ir"(a)                                     \
-       : "cc");        /* memory clobber comes from smp_mb() */        \
-                                                                       \
-       smp_mb();                                                       \
-                                                                       \
-       return val;                                                     \
-}
-
-#define ATOMIC64_FETCH_OP(op, op1, op2)                                        \
-static inline s64 arch_atomic64_fetch_##op(s64 a, atomic64_t *v)       \
-{                                                                      \
-       s64 val, orig;                                                  \
-                                                                       \
-       smp_mb();                                                       \
-                                                                       \
-       __asm__ __volatile__(                                           \
-       "1:                             \n"                             \
-       "       llockd   %0, [%2]       \n"                             \
-       "       " #op1 " %L1, %L0, %L3  \n"                             \
-       "       " #op2 " %H1, %H0, %H3  \n"                             \
-       "       scondd   %1, [%2]       \n"                             \
-       "       bnz     1b              \n"                             \
-       : "=&r"(orig), "=&r"(val)                                       \
-       : "r"(&v->counter), "ir"(a)                                     \
-       : "cc");        /* memory clobber comes from smp_mb() */        \
-                                                                       \
-       smp_mb();                                                       \
-                                                                       \
-       return orig;                                                    \
-}
-
-#define ATOMIC64_OPS(op, op1, op2)                                     \
-       ATOMIC64_OP(op, op1, op2)                                       \
-       ATOMIC64_OP_RETURN(op, op1, op2)                                \
-       ATOMIC64_FETCH_OP(op, op1, op2)
-
-ATOMIC64_OPS(add, add.f, adc)
-ATOMIC64_OPS(sub, sub.f, sbc)
-ATOMIC64_OPS(and, and, and)
-ATOMIC64_OPS(andnot, bic, bic)
-ATOMIC64_OPS(or, or, or)
-ATOMIC64_OPS(xor, xor, xor)
-
-#define arch_atomic64_andnot           arch_atomic64_andnot
-#define arch_atomic64_fetch_andnot     arch_atomic64_fetch_andnot
-
-#undef ATOMIC64_OPS
-#undef ATOMIC64_FETCH_OP
-#undef ATOMIC64_OP_RETURN
-#undef ATOMIC64_OP
-
-static inline s64
-arch_atomic64_cmpxchg(atomic64_t *ptr, s64 expected, s64 new)
-{
-       s64 prev;
-
-       smp_mb();
-
-       __asm__ __volatile__(
-       "1:     llockd  %0, [%1]        \n"
-       "       brne    %L0, %L2, 2f    \n"
-       "       brne    %H0, %H2, 2f    \n"
-       "       scondd  %3, [%1]        \n"
-       "       bnz     1b              \n"
-       "2:                             \n"
-       : "=&r"(prev)
-       : "r"(ptr), "ir"(expected), "r"(new)
-       : "cc");        /* memory clobber comes from smp_mb() */
-
-       smp_mb();
-
-       return prev;
-}
-
-static inline s64 arch_atomic64_xchg(atomic64_t *ptr, s64 new)
-{
-       s64 prev;
-
-       smp_mb();
-
-       __asm__ __volatile__(
-       "1:     llockd  %0, [%1]        \n"
-       "       scondd  %2, [%1]        \n"
-       "       bnz     1b              \n"
-       "2:                             \n"
-       : "=&r"(prev)
-       : "r"(ptr), "r"(new)
-       : "cc");        /* memory clobber comes from smp_mb() */
-
-       smp_mb();
-
-       return prev;
-}
-
-/**
- * arch_atomic64_dec_if_positive - decrement by 1 if old value positive
- * @v: pointer of type atomic64_t
- *
- * The function returns the old value of *v minus 1, even if
- * the atomic variable, v, was not decremented.
- */
-
-static inline s64 arch_atomic64_dec_if_positive(atomic64_t *v)
-{
-       s64 val;
-
-       smp_mb();
-
-       __asm__ __volatile__(
-       "1:     llockd  %0, [%1]        \n"
-       "       sub.f   %L0, %L0, 1     # w0 - 1, set C on borrow\n"
-       "       sub.c   %H0, %H0, 1     # if C set, w1 - 1\n"
-       "       brlt    %H0, 0, 2f      \n"
-       "       scondd  %0, [%1]        \n"
-       "       bnz     1b              \n"
-       "2:                             \n"
-       : "=&r"(val)
-       : "r"(&v->counter)
-       : "cc");        /* memory clobber comes from smp_mb() */
-
-       smp_mb();
-
-       return val;
-}
-#define arch_atomic64_dec_if_positive arch_atomic64_dec_if_positive
-
-/**
- * arch_atomic64_fetch_add_unless - add unless the number is a given value
- * @v: pointer of type atomic64_t
- * @a: the amount to add to v...
- * @u: ...unless v is equal to u.
- *
- * Atomically adds @a to @v, if it was not @u.
- * Returns the old value of @v
- */
-static inline s64 arch_atomic64_fetch_add_unless(atomic64_t *v, s64 a, s64 u)
-{
-       s64 old, temp;
-
-       smp_mb();
-
-       __asm__ __volatile__(
-       "1:     llockd  %0, [%2]        \n"
-       "       brne    %L0, %L4, 2f    # continue to add since v != u \n"
-       "       breq.d  %H0, %H4, 3f    # return since v == u \n"
-       "2:                             \n"
-       "       add.f   %L1, %L0, %L3   \n"
-       "       adc     %H1, %H0, %H3   \n"
-       "       scondd  %1, [%2]        \n"
-       "       bnz     1b              \n"
-       "3:                             \n"
-       : "=&r"(old), "=&r" (temp)
-       : "r"(&v->counter), "r"(a), "r"(u)
-       : "cc");        /* memory clobber comes from smp_mb() */
-
-       smp_mb();
-
-       return old;
-}
-#define arch_atomic64_fetch_add_unless arch_atomic64_fetch_add_unless
-
-#endif /* !CONFIG_GENERIC_ATOMIC64 */
+#else
+#include <asm/atomic64-arcv2.h>
+#endif
 
 #endif /* !__ASSEMBLY__ */