summaryrefslogtreecommitdiff
path: root/arch/parisc/include
diff options
context:
space:
mode:
Diffstat (limited to 'arch/parisc/include')
-rw-r--r--arch/parisc/include/asm/atomic.h63
-rw-r--r--arch/parisc/include/asm/hash.h146
-rw-r--r--arch/parisc/include/asm/spinlock.h9
-rw-r--r--arch/parisc/include/asm/uaccess.h23
4 files changed, 221 insertions, 20 deletions
diff --git a/arch/parisc/include/asm/atomic.h b/arch/parisc/include/asm/atomic.h
index 1d109990a..5394b9c5f 100644
--- a/arch/parisc/include/asm/atomic.h
+++ b/arch/parisc/include/asm/atomic.h
@@ -121,16 +121,39 @@ static __inline__ int atomic_##op##_return(int i, atomic_t *v) \
return ret; \
}
-#define ATOMIC_OPS(op, c_op) ATOMIC_OP(op, c_op) ATOMIC_OP_RETURN(op, c_op)
+#define ATOMIC_FETCH_OP(op, c_op) \
+static __inline__ int atomic_fetch_##op(int i, atomic_t *v) \
+{ \
+ unsigned long flags; \
+ int ret; \
+ \
+ _atomic_spin_lock_irqsave(v, flags); \
+ ret = v->counter; \
+ v->counter c_op i; \
+ _atomic_spin_unlock_irqrestore(v, flags); \
+ \
+ return ret; \
+}
+
+#define ATOMIC_OPS(op, c_op) \
+ ATOMIC_OP(op, c_op) \
+ ATOMIC_OP_RETURN(op, c_op) \
+ ATOMIC_FETCH_OP(op, c_op)
ATOMIC_OPS(add, +=)
ATOMIC_OPS(sub, -=)
-ATOMIC_OP(and, &=)
-ATOMIC_OP(or, |=)
-ATOMIC_OP(xor, ^=)
+#undef ATOMIC_OPS
+#define ATOMIC_OPS(op, c_op) \
+ ATOMIC_OP(op, c_op) \
+ ATOMIC_FETCH_OP(op, c_op)
+
+ATOMIC_OPS(and, &=)
+ATOMIC_OPS(or, |=)
+ATOMIC_OPS(xor, ^=)
#undef ATOMIC_OPS
+#undef ATOMIC_FETCH_OP
#undef ATOMIC_OP_RETURN
#undef ATOMIC_OP
@@ -185,15 +208,39 @@ static __inline__ s64 atomic64_##op##_return(s64 i, atomic64_t *v) \
return ret; \
}
-#define ATOMIC64_OPS(op, c_op) ATOMIC64_OP(op, c_op) ATOMIC64_OP_RETURN(op, c_op)
+#define ATOMIC64_FETCH_OP(op, c_op) \
+static __inline__ s64 atomic64_fetch_##op(s64 i, atomic64_t *v) \
+{ \
+ unsigned long flags; \
+ s64 ret; \
+ \
+ _atomic_spin_lock_irqsave(v, flags); \
+ ret = v->counter; \
+ v->counter c_op i; \
+ _atomic_spin_unlock_irqrestore(v, flags); \
+ \
+ return ret; \
+}
+
+#define ATOMIC64_OPS(op, c_op) \
+ ATOMIC64_OP(op, c_op) \
+ ATOMIC64_OP_RETURN(op, c_op) \
+ ATOMIC64_FETCH_OP(op, c_op)
ATOMIC64_OPS(add, +=)
ATOMIC64_OPS(sub, -=)
-ATOMIC64_OP(and, &=)
-ATOMIC64_OP(or, |=)
-ATOMIC64_OP(xor, ^=)
#undef ATOMIC64_OPS
+#define ATOMIC64_OPS(op, c_op) \
+ ATOMIC64_OP(op, c_op) \
+ ATOMIC64_FETCH_OP(op, c_op)
+
+ATOMIC64_OPS(and, &=)
+ATOMIC64_OPS(or, |=)
+ATOMIC64_OPS(xor, ^=)
+
+#undef ATOMIC64_OPS
+#undef ATOMIC64_FETCH_OP
#undef ATOMIC64_OP_RETURN
#undef ATOMIC64_OP
diff --git a/arch/parisc/include/asm/hash.h b/arch/parisc/include/asm/hash.h
new file mode 100644
index 000000000..dbe93311a
--- /dev/null
+++ b/arch/parisc/include/asm/hash.h
@@ -0,0 +1,146 @@
+#ifndef _ASM_HASH_H
+#define _ASM_HASH_H
+
+/*
+ * HP-PA only implements integer multiply in the FPU. However, for
+ * integer multiplies by constant, it has a number of shift-and-add
+ * (but no shift-and-subtract, sigh!) instructions that a compiler
+ * can synthesize a code sequence with.
+ *
+ * Unfortunately, GCC isn't very efficient at using them. For example
+ * it uses three instructions for "x *= 21" when only two are needed.
+ * But we can find a sequence manually.
+ */
+
+#define HAVE_ARCH__HASH_32 1
+
+/*
+ * This is a multiply by GOLDEN_RATIO_32 = 0x61C88647 optimized for the
+ * PA7100 pairing rules. This is an in-order 2-way superscalar processor.
+ * Only one instruction in a pair may be a shift (by more than 3 bits),
+ * but other than that, simple ALU ops (including shift-and-add by up
+ * to 3 bits) may be paired arbitrarily.
+ *
+ * PA8xxx processors also dual-issue ALU instructions, although with
+ * fewer constraints, so this schedule is good for them, too.
+ *
+ * This 6-step sequence was found by Yevgen Voronenko's implementation
+ * of the Hcub algorithm at http://spiral.ece.cmu.edu/mcm/gen.html.
+ */
+static inline u32 __attribute_const__ __hash_32(u32 x)
+{
+ u32 a, b, c;
+
+ /*
+ * Phase 1: Compute a = (x << 19) + x,
+ * b = (x << 9) + a, c = (x << 23) + b.
+ */
+ a = x << 19; /* Two shifts can't be paired */
+ b = x << 9; a += x;
+ c = x << 23; b += a;
+ c += b;
+ /* Phase 2: Return (b<<11) + (c<<6) + (a<<3) - c */
+ b <<= 11;
+ a += c << 3; b -= c;
+ return (a << 3) + b;
+}
+
+#if BITS_PER_LONG == 64
+
+#define HAVE_ARCH_HASH_64 1
+
+/*
+ * Finding a good shift-and-add chain for GOLDEN_RATIO_64 is tricky,
+ * because available software for the purpose chokes on constants this
+ * large. (It's mostly designed for compiling FIR filter coefficients
+ * into FPGAs.)
+ *
+ * However, Jason Thong pointed out a work-around. The Hcub software
+ * (http://spiral.ece.cmu.edu/mcm/gen.html) is designed for *multiple*
+ * constant multiplication, and is good at finding shift-and-add chains
+ * which share common terms.
+ *
+ * Looking at 0x0x61C8864680B583EB in binary:
+ * 0110000111001000100001100100011010000000101101011000001111101011
+ * \______________/ \__________/ \_______/ \________/
+ * \____________________________/ \____________________/
+ * you can see the non-zero bits are divided into several well-separated
+ * blocks. Hcub can find algorithms for those terms separately, which
+ * can then be shifted and added together.
+ *
+ * Dividing the input into 2, 3 or 4 blocks, Hcub can find solutions
+ * with 10, 9 or 8 adds, respectively, making a total of 11 for the
+ * whole number.
+ *
+ * Using just two large blocks, 0xC3910C8D << 31 in the high bits,
+ * and 0xB583EB in the low bits, produces as good an algorithm as any,
+ * and with one more small shift than alternatives.
+ *
+ * The high bits are a larger number and more work to compute, as well
+ * as needing one extra cycle to shift left 31 bits before the final
+ * addition, so they are the critical path for scheduling. The low bits
+ * can fit into the scheduling slots left over.
+ */
+
+
+/*
+ * This _ASSIGN(dst, src) macro performs "dst = src", but prevents GCC
+ * from inferring anything about the value assigned to "dest".
+ *
+ * This prevents it from mis-optimizing certain sequences.
+ * In particular, gcc is annoyingly eager to combine consecutive shifts.
+ * Given "x <<= 19; y += x; z += x << 1;", GCC will turn this into
+ * "y += x << 19; z += x << 20;" even though the latter sequence needs
+ * an additional instruction and temporary register.
+ *
+ * Because no actual assembly code is generated, this construct is
+ * usefully portable across all GCC platforms, and so can be test-compiled
+ * on non-PA systems.
+ *
+ * In two places, additional unused input dependencies are added. This
+ * forces GCC's scheduling so it does not rearrange instructions too much.
+ * Because the PA-8xxx is out of order, I'm not sure how much this matters,
+ * but why make it more difficult for the processor than necessary?
+ */
+#define _ASSIGN(dst, src, ...) asm("" : "=r" (dst) : "0" (src), ##__VA_ARGS__)
+
+/*
+ * Multiply by GOLDEN_RATIO_64 = 0x0x61C8864680B583EB using a heavily
+ * optimized shift-and-add sequence.
+ *
+ * Without the final shift, the multiply proper is 19 instructions,
+ * 10 cycles and uses only 4 temporaries. Whew!
+ *
+ * You are not expected to understand this.
+ */
+static __always_inline u32 __attribute_const__
+hash_64(u64 a, unsigned int bits)
+{
+ u64 b, c, d;
+
+ /*
+ * Encourage GCC to move a dynamic shift to %sar early,
+ * thereby freeing up an additional temporary register.
+ */
+ if (!__builtin_constant_p(bits))
+ asm("" : "=q" (bits) : "0" (64 - bits));
+ else
+ bits = 64 - bits;
+
+ _ASSIGN(b, a*5); c = a << 13;
+ b = (b << 2) + a; _ASSIGN(d, a << 17);
+ a = b + (a << 1); c += d;
+ d = a << 10; _ASSIGN(a, a << 19);
+ d = a - d; _ASSIGN(a, a << 4, "X" (d));
+ c += b; a += b;
+ d -= c; c += a << 1;
+ a += c << 3; _ASSIGN(b, b << (7+31), "X" (c), "X" (d));
+ a <<= 31; b += d;
+ a += b;
+ return a >> bits;
+}
+#undef _ASSIGN /* We're a widely-used header file, so don't litter! */
+
+#endif /* BITS_PER_LONG == 64 */
+
+#endif /* _ASM_HASH_H */
diff --git a/arch/parisc/include/asm/spinlock.h b/arch/parisc/include/asm/spinlock.h
index 64f2992e4..e32936cd7 100644
--- a/arch/parisc/include/asm/spinlock.h
+++ b/arch/parisc/include/asm/spinlock.h
@@ -13,8 +13,13 @@ static inline int arch_spin_is_locked(arch_spinlock_t *x)
}
#define arch_spin_lock(lock) arch_spin_lock_flags(lock, 0)
-#define arch_spin_unlock_wait(x) \
- do { cpu_relax(); } while (arch_spin_is_locked(x))
+
+static inline void arch_spin_unlock_wait(arch_spinlock_t *x)
+{
+ volatile unsigned int *a = __ldcw_align(x);
+
+ smp_cond_load_acquire(a, VAL);
+}
static inline void arch_spin_lock_flags(arch_spinlock_t *x,
unsigned long flags)
diff --git a/arch/parisc/include/asm/uaccess.h b/arch/parisc/include/asm/uaccess.h
index 37a1bee96..482847865 100644
--- a/arch/parisc/include/asm/uaccess.h
+++ b/arch/parisc/include/asm/uaccess.h
@@ -209,13 +209,13 @@ unsigned long copy_in_user(void __user *dst, const void __user *src, unsigned lo
#define __copy_to_user_inatomic __copy_to_user
#define __copy_from_user_inatomic __copy_from_user
-extern void copy_from_user_overflow(void)
-#ifdef CONFIG_DEBUG_STRICT_USER_COPY_CHECKS
- __compiletime_error("copy_from_user() buffer size is not provably correct")
-#else
- __compiletime_warning("copy_from_user() buffer size is not provably correct")
-#endif
-;
+extern void __compiletime_error("usercopy buffer size is too small")
+__bad_copy_user(void);
+
+static inline void copy_user_overflow(int size, unsigned long count)
+{
+ WARN(1, "Buffer overflow detected (%d < %lu)!\n", size, count);
+}
static inline unsigned long __must_check copy_from_user(void *to,
const void __user *from,
@@ -224,10 +224,13 @@ static inline unsigned long __must_check copy_from_user(void *to,
int sz = __compiletime_object_size(to);
unsigned long ret = n;
- if (likely(sz == -1 || !__builtin_constant_p(n) || sz >= n))
+ if (likely(sz == -1 || sz >= n))
ret = __copy_from_user(to, from, n);
- else
- copy_from_user_overflow();
+ else if (!__builtin_constant_p(n))
+ copy_user_overflow(sz, n);
+ else
+ __bad_copy_user();
+
if (unlikely(ret))
memset(to + (n - ret), 0, ret);
return ret;