RISC-V: Atomic and Locking Code
authorPalmer Dabbelt <palmer@dabbelt.com>
Tue, 11 Jul 2017 01:02:19 +0000 (18:02 -0700)
committerPalmer Dabbelt <palmer@dabbelt.com>
Tue, 26 Sep 2017 22:26:45 +0000 (15:26 -0700)
This contains all the code that directly interfaces with the RISC-V
memory model.  While this code corforms to the current RISC-V ISA
specifications (user 2.2 and priv 1.10), the memory model is somewhat
underspecified in those documents.  There is a working group that hopes
to produce a formal memory model by the end of the year, but my
understanding is that the basic definitions we're relying on here won't
change significantly.

Reviewed-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Palmer Dabbelt <palmer@dabbelt.com>
arch/riscv/include/asm/atomic.h [new file with mode: 0644]
arch/riscv/include/asm/barrier.h [new file with mode: 0644]
arch/riscv/include/asm/bitops.h [new file with mode: 0644]
arch/riscv/include/asm/cacheflush.h [new file with mode: 0644]
arch/riscv/include/asm/cmpxchg.h [new file with mode: 0644]
arch/riscv/include/asm/io.h [new file with mode: 0644]
arch/riscv/include/asm/spinlock.h [new file with mode: 0644]
arch/riscv/include/asm/spinlock_types.h [new file with mode: 0644]
arch/riscv/include/asm/tlb.h [new file with mode: 0644]
arch/riscv/include/asm/tlbflush.h [new file with mode: 0644]

diff --git a/arch/riscv/include/asm/atomic.h b/arch/riscv/include/asm/atomic.h
new file mode 100644 (file)
index 0000000..e2e37c5
--- /dev/null
@@ -0,0 +1,375 @@
+/*
+ * Copyright (C) 2007 Red Hat, Inc. All Rights Reserved.
+ * Copyright (C) 2012 Regents of the University of California
+ * Copyright (C) 2017 SiFive
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public Licence
+ * as published by the Free Software Foundation; either version
+ * 2 of the Licence, or (at your option) any later version.
+ */
+
+#ifndef _ASM_RISCV_ATOMIC_H
+#define _ASM_RISCV_ATOMIC_H
+
+#ifdef CONFIG_GENERIC_ATOMIC64
+# include <asm-generic/atomic64.h>
+#else
+# if (__riscv_xlen < 64)
+#  error "64-bit atomics require XLEN to be at least 64"
+# endif
+#endif
+
+#include <asm/cmpxchg.h>
+#include <asm/barrier.h>
+
+#define ATOMIC_INIT(i) { (i) }
+static __always_inline int atomic_read(const atomic_t *v)
+{
+       return READ_ONCE(v->counter);
+}
+static __always_inline void atomic_set(atomic_t *v, int i)
+{
+       WRITE_ONCE(v->counter, i);
+}
+
+#ifndef CONFIG_GENERIC_ATOMIC64
+#define ATOMIC64_INIT(i) { (i) }
+static __always_inline long atomic64_read(const atomic64_t *v)
+{
+       return READ_ONCE(v->counter);
+}
+static __always_inline void atomic64_set(atomic64_t *v, long i)
+{
+       WRITE_ONCE(v->counter, i);
+}
+#endif
+
+/*
+ * First, the atomic ops that have no ordering constraints and therefor don't
+ * have the AQ or RL bits set.  These don't return anything, so there's only
+ * one version to worry about.
+ */
+#define ATOMIC_OP(op, asm_op, c_op, I, asm_type, c_type, prefix)                               \
+static __always_inline void atomic##prefix##_##op(c_type i, atomic##prefix##_t *v)             \
+{                                                                                              \
+       __asm__ __volatile__ (                                                                  \
+               "amo" #asm_op "." #asm_type " zero, %1, %0"                                     \
+               : "+A" (v->counter)                                                             \
+               : "r" (I)                                                                       \
+               : "memory");                                                                    \
+}
+
+#ifdef CONFIG_GENERIC_ATOMIC64
+#define ATOMIC_OPS(op, asm_op, c_op, I)                        \
+        ATOMIC_OP (op, asm_op, c_op, I, w,  int,   )
+#else
+#define ATOMIC_OPS(op, asm_op, c_op, I)                        \
+        ATOMIC_OP (op, asm_op, c_op, I, w,  int,   )   \
+        ATOMIC_OP (op, asm_op, c_op, I, d, long, 64)
+#endif
+
+ATOMIC_OPS(add, add, +,  i)
+ATOMIC_OPS(sub, add, +, -i)
+ATOMIC_OPS(and, and, &,  i)
+ATOMIC_OPS( or,  or, |,  i)
+ATOMIC_OPS(xor, xor, ^,  i)
+
+#undef ATOMIC_OP
+#undef ATOMIC_OPS
+
+/*
+ * Atomic ops that have ordered, relaxed, acquire, and relese variants.
+ * There's two flavors of these: the arithmatic ops have both fetch and return
+ * versions, while the logical ops only have fetch versions.
+ */
+#define ATOMIC_FETCH_OP(op, asm_op, c_op, I, asm_or, c_or, asm_type, c_type, prefix)                   \
+static __always_inline c_type atomic##prefix##_fetch_##op##c_or(c_type i, atomic##prefix##_t *v)       \
+{                                                                                                      \
+       register c_type ret;                                                                            \
+       __asm__ __volatile__ (                                                                          \
+               "amo" #asm_op "." #asm_type #asm_or " %1, %2, %0"                                       \
+               : "+A" (v->counter), "=r" (ret)                                                         \
+               : "r" (I)                                                                               \
+               : "memory");                                                                            \
+       return ret;                                                                                     \
+}
+
+#define ATOMIC_OP_RETURN(op, asm_op, c_op, I, asm_or, c_or, asm_type, c_type, prefix)                  \
+static __always_inline c_type atomic##prefix##_##op##_return##c_or(c_type i, atomic##prefix##_t *v)    \
+{                                                                                                      \
+        return atomic##prefix##_fetch_##op##c_or(i, v) c_op I;                                         \
+}
+
+#ifdef CONFIG_GENERIC_ATOMIC64
+#define ATOMIC_OPS(op, asm_op, c_op, I, asm_or, c_or)                          \
+        ATOMIC_FETCH_OP (op, asm_op, c_op, I, asm_or, c_or, w,  int,   )       \
+        ATOMIC_OP_RETURN(op, asm_op, c_op, I, asm_or, c_or, w,  int,   )
+#else
+#define ATOMIC_OPS(op, asm_op, c_op, I, asm_or, c_or)                          \
+        ATOMIC_FETCH_OP (op, asm_op, c_op, I, asm_or, c_or, w,  int,   )       \
+        ATOMIC_OP_RETURN(op, asm_op, c_op, I, asm_or, c_or, w,  int,   )       \
+        ATOMIC_FETCH_OP (op, asm_op, c_op, I, asm_or, c_or, d, long, 64)       \
+        ATOMIC_OP_RETURN(op, asm_op, c_op, I, asm_or, c_or, d, long, 64)
+#endif
+
+ATOMIC_OPS(add, add, +,  i,      , _relaxed)
+ATOMIC_OPS(add, add, +,  i, .aq  , _acquire)
+ATOMIC_OPS(add, add, +,  i, .rl  , _release)
+ATOMIC_OPS(add, add, +,  i, .aqrl,         )
+
+ATOMIC_OPS(sub, add, +, -i,      , _relaxed)
+ATOMIC_OPS(sub, add, +, -i, .aq  , _acquire)
+ATOMIC_OPS(sub, add, +, -i, .rl  , _release)
+ATOMIC_OPS(sub, add, +, -i, .aqrl,         )
+
+#undef ATOMIC_OPS
+
+#ifdef CONFIG_GENERIC_ATOMIC64
+#define ATOMIC_OPS(op, asm_op, c_op, I, asm_or, c_or)                          \
+        ATOMIC_FETCH_OP(op, asm_op, c_op, I, asm_or, c_or, w,  int,   )
+#else
+#define ATOMIC_OPS(op, asm_op, c_op, I, asm_or, c_or)                          \
+        ATOMIC_FETCH_OP(op, asm_op, c_op, I, asm_or, c_or, w,  int,   )                \
+        ATOMIC_FETCH_OP(op, asm_op, c_op, I, asm_or, c_or, d, long, 64)
+#endif
+
+ATOMIC_OPS(and, and, &,  i,      , _relaxed)
+ATOMIC_OPS(and, and, &,  i, .aq  , _acquire)
+ATOMIC_OPS(and, and, &,  i, .rl  , _release)
+ATOMIC_OPS(and, and, &,  i, .aqrl,         )
+
+ATOMIC_OPS( or,  or, |,  i,      , _relaxed)
+ATOMIC_OPS( or,  or, |,  i, .aq  , _acquire)
+ATOMIC_OPS( or,  or, |,  i, .rl  , _release)
+ATOMIC_OPS( or,  or, |,  i, .aqrl,         )
+
+ATOMIC_OPS(xor, xor, ^,  i,      , _relaxed)
+ATOMIC_OPS(xor, xor, ^,  i, .aq  , _acquire)
+ATOMIC_OPS(xor, xor, ^,  i, .rl  , _release)
+ATOMIC_OPS(xor, xor, ^,  i, .aqrl,         )
+
+#undef ATOMIC_OPS
+
+#undef ATOMIC_FETCH_OP
+#undef ATOMIC_OP_RETURN
+
+/*
+ * The extra atomic operations that are constructed from one of the core
+ * AMO-based operations above (aside from sub, which is easier to fit above).
+ * These are required to perform a barrier, but they're OK this way because
+ * atomic_*_return is also required to perform a barrier.
+ */
+#define ATOMIC_OP(op, func_op, comp_op, I, c_type, prefix)                     \
+static __always_inline bool atomic##prefix##_##op(c_type i, atomic##prefix##_t *v) \
+{                                                                              \
+       return atomic##prefix##_##func_op##_return(i, v) comp_op I;             \
+}
+
+#ifdef CONFIG_GENERIC_ATOMIC64
+#define ATOMIC_OPS(op, func_op, comp_op, I)                    \
+        ATOMIC_OP (op, func_op, comp_op, I,  int,   )
+#else
+#define ATOMIC_OPS(op, func_op, comp_op, I)                    \
+        ATOMIC_OP (op, func_op, comp_op, I,  int,   )          \
+        ATOMIC_OP (op, func_op, comp_op, I, long, 64)
+#endif
+
+ATOMIC_OPS(add_and_test, add, ==, 0)
+ATOMIC_OPS(sub_and_test, sub, ==, 0)
+ATOMIC_OPS(add_negative, add,  <, 0)
+
+#undef ATOMIC_OP
+#undef ATOMIC_OPS
+
+#define ATOMIC_OP(op, func_op, c_op, I, c_type, prefix)                                \
+static __always_inline void atomic##prefix##_##op(atomic##prefix##_t *v)       \
+{                                                                              \
+       atomic##prefix##_##func_op(I, v);                                       \
+}
+
+#define ATOMIC_FETCH_OP(op, func_op, c_op, I, c_type, prefix)                          \
+static __always_inline c_type atomic##prefix##_fetch_##op(atomic##prefix##_t *v)       \
+{                                                                                      \
+       return atomic##prefix##_fetch_##func_op(I, v);                                  \
+}
+
+#define ATOMIC_OP_RETURN(op, asm_op, c_op, I, c_type, prefix)                          \
+static __always_inline c_type atomic##prefix##_##op##_return(atomic##prefix##_t *v)    \
+{                                                                                      \
+        return atomic##prefix##_fetch_##op(v) c_op I;                                  \
+}
+
+#ifdef CONFIG_GENERIC_ATOMIC64
+#define ATOMIC_OPS(op, asm_op, c_op, I)                                                \
+        ATOMIC_OP       (op, asm_op, c_op, I,  int,   )                                \
+        ATOMIC_FETCH_OP (op, asm_op, c_op, I,  int,   )                                \
+        ATOMIC_OP_RETURN(op, asm_op, c_op, I,  int,   )
+#else
+#define ATOMIC_OPS(op, asm_op, c_op, I)                                                \
+        ATOMIC_OP       (op, asm_op, c_op, I,  int,   )                                \
+        ATOMIC_FETCH_OP (op, asm_op, c_op, I,  int,   )                                \
+        ATOMIC_OP_RETURN(op, asm_op, c_op, I,  int,   )                                \
+        ATOMIC_OP       (op, asm_op, c_op, I, long, 64)                                \
+        ATOMIC_FETCH_OP (op, asm_op, c_op, I, long, 64)                                \
+        ATOMIC_OP_RETURN(op, asm_op, c_op, I, long, 64)
+#endif
+
+ATOMIC_OPS(inc, add, +,  1)
+ATOMIC_OPS(dec, add, +, -1)
+
+#undef ATOMIC_OPS
+#undef ATOMIC_OP
+#undef ATOMIC_FETCH_OP
+#undef ATOMIC_OP_RETURN
+
+#define ATOMIC_OP(op, func_op, comp_op, I, prefix)                             \
+static __always_inline bool atomic##prefix##_##op(atomic##prefix##_t *v)       \
+{                                                                              \
+       return atomic##prefix##_##func_op##_return(v) comp_op I;                \
+}
+
+ATOMIC_OP(inc_and_test, inc, ==, 0,   )
+ATOMIC_OP(dec_and_test, dec, ==, 0,   )
+#ifndef CONFIG_GENERIC_ATOMIC64
+ATOMIC_OP(inc_and_test, inc, ==, 0, 64)
+ATOMIC_OP(dec_and_test, dec, ==, 0, 64)
+#endif
+
+#undef ATOMIC_OP
+
+/* This is required to provide a barrier on success. */
+static __always_inline int __atomic_add_unless(atomic_t *v, int a, int u)
+{
+       int prev, rc;
+
+       __asm__ __volatile__ (
+               "0:\n\t"
+               "lr.w.aqrl  %[p],  %[c]\n\t"
+               "beq        %[p],  %[u], 1f\n\t"
+               "add       %[rc],  %[p], %[a]\n\t"
+               "sc.w.aqrl %[rc], %[rc], %[c]\n\t"
+               "bnez      %[rc], 0b\n\t"
+               "1:"
+               : [p]"=&r" (prev), [rc]"=&r" (rc), [c]"+A" (v->counter)
+               : [a]"r" (a), [u]"r" (u)
+               : "memory");
+       return prev;
+}
+
+#ifndef CONFIG_GENERIC_ATOMIC64
+static __always_inline long __atomic64_add_unless(atomic64_t *v, long a, long u)
+{
+       long prev, rc;
+
+       __asm__ __volatile__ (
+               "0:\n\t"
+               "lr.d.aqrl  %[p],  %[c]\n\t"
+               "beq        %[p],  %[u], 1f\n\t"
+               "add       %[rc],  %[p], %[a]\n\t"
+               "sc.d.aqrl %[rc], %[rc], %[c]\n\t"
+               "bnez      %[rc], 0b\n\t"
+               "1:"
+               : [p]"=&r" (prev), [rc]"=&r" (rc), [c]"+A" (v->counter)
+               : [a]"r" (a), [u]"r" (u)
+               : "memory");
+       return prev;
+}
+
+static __always_inline int atomic64_add_unless(atomic64_t *v, long a, long u)
+{
+       return __atomic64_add_unless(v, a, u) != u;
+}
+#endif
+
+/*
+ * The extra atomic operations that are constructed from one of the core
+ * LR/SC-based operations above.
+ */
+static __always_inline int atomic_inc_not_zero(atomic_t *v)
+{
+        return __atomic_add_unless(v, 1, 0);
+}
+
+#ifndef CONFIG_GENERIC_ATOMIC64
+static __always_inline long atomic64_inc_not_zero(atomic64_t *v)
+{
+        return atomic64_add_unless(v, 1, 0);
+}
+#endif
+
+/*
+ * atomic_{cmp,}xchg is required to have exactly the same ordering semantics as
+ * {cmp,}xchg and the operations that return, so they need a barrier.  We just
+ * use the other implementations directly.
+ */
+#define ATOMIC_OP(c_t, prefix, c_or, size, asm_or)                                             \
+static __always_inline c_t atomic##prefix##_cmpxchg##c_or(atomic##prefix##_t *v, c_t o, c_t n)         \
+{                                                                                              \
+       return __cmpxchg(&(v->counter), o, n, size, asm_or, asm_or);                            \
+}                                                                                              \
+static __always_inline c_t atomic##prefix##_xchg##c_or(atomic##prefix##_t *v, c_t n)           \
+{                                                                                              \
+       return __xchg(n, &(v->counter), size, asm_or);                                          \
+}
+
+#ifdef CONFIG_GENERIC_ATOMIC64
+#define ATOMIC_OPS(c_or, asm_or)                       \
+       ATOMIC_OP( int,   , c_or, 4, asm_or)
+#else
+#define ATOMIC_OPS(c_or, asm_or)                       \
+       ATOMIC_OP( int,   , c_or, 4, asm_or)            \
+       ATOMIC_OP(long, 64, c_or, 8, asm_or)
+#endif
+
+ATOMIC_OPS(        , .aqrl)
+ATOMIC_OPS(_acquire,   .aq)
+ATOMIC_OPS(_release,   .rl)
+ATOMIC_OPS(_relaxed,      )
+
+#undef ATOMIC_OPS
+#undef ATOMIC_OP
+
+static __always_inline int atomic_sub_if_positive(atomic_t *v, int offset)
+{
+       int prev, rc;
+
+       __asm__ __volatile__ (
+               "0:\n\t"
+               "lr.w.aqrl  %[p],  %[c]\n\t"
+               "sub       %[rc],  %[p], %[o]\n\t"
+               "bltz      %[rc],    1f\n\t"
+               "sc.w.aqrl %[rc], %[rc], %[c]\n\t"
+               "bnez      %[rc],    0b\n\t"
+               "1:"
+               : [p]"=&r" (prev), [rc]"=&r" (rc), [c]"+A" (v->counter)
+               : [o]"r" (offset)
+               : "memory");
+       return prev - offset;
+}
+
+#define atomic_dec_if_positive(v)      atomic_sub_if_positive(v, 1)
+
+#ifndef CONFIG_GENERIC_ATOMIC64
+static __always_inline long atomic64_sub_if_positive(atomic64_t *v, int offset)
+{
+       long prev, rc;
+
+       __asm__ __volatile__ (
+               "0:\n\t"
+               "lr.d.aqrl  %[p],  %[c]\n\t"
+               "sub       %[rc],  %[p], %[o]\n\t"
+               "bltz      %[rc],    1f\n\t"
+               "sc.d.aqrl %[rc], %[rc], %[c]\n\t"
+               "bnez      %[rc],    0b\n\t"
+               "1:"
+               : [p]"=&r" (prev), [rc]"=&r" (rc), [c]"+A" (v->counter)
+               : [o]"r" (offset)
+               : "memory");
+       return prev - offset;
+}
+
+#define atomic64_dec_if_positive(v)    atomic64_sub_if_positive(v, 1)
+#endif
+
+#endif /* _ASM_RISCV_ATOMIC_H */
diff --git a/arch/riscv/include/asm/barrier.h b/arch/riscv/include/asm/barrier.h
new file mode 100644 (file)
index 0000000..183534b
--- /dev/null
@@ -0,0 +1,68 @@
+/*
+ * Based on arch/arm/include/asm/barrier.h
+ *
+ * Copyright (C) 2012 ARM Ltd.
+ * Copyright (C) 2013 Regents of the University of California
+ * Copyright (C) 2017 SiFive
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef _ASM_RISCV_BARRIER_H
+#define _ASM_RISCV_BARRIER_H
+
+#ifndef __ASSEMBLY__
+
+#define nop()          __asm__ __volatile__ ("nop")
+
+#define RISCV_FENCE(p, s) \
+       __asm__ __volatile__ ("fence " #p "," #s : : : "memory")
+
+/* These barriers need to enforce ordering on both devices or memory. */
+#define mb()           RISCV_FENCE(iorw,iorw)
+#define rmb()          RISCV_FENCE(ir,ir)
+#define wmb()          RISCV_FENCE(ow,ow)
+
+/* These barriers do not need to enforce ordering on devices, just memory. */
+#define smp_mb()       RISCV_FENCE(rw,rw)
+#define smp_rmb()      RISCV_FENCE(r,r)
+#define smp_wmb()      RISCV_FENCE(w,w)
+
+/*
+ * These fences exist to enforce ordering around the relaxed AMOs.  The
+ * documentation defines that
+ * "
+ *     atomic_fetch_add();
+ *   is equivalent to:
+ *     smp_mb__before_atomic();
+ *     atomic_fetch_add_relaxed();
+ *     smp_mb__after_atomic();
+ * "
+ * So we emit full fences on both sides.
+ */
+#define __smb_mb__before_atomic()      smp_mb()
+#define __smb_mb__after_atomic()       smp_mb()
+
+/*
+ * These barriers prevent accesses performed outside a spinlock from being moved
+ * inside a spinlock.  Since RISC-V sets the aq/rl bits on our spinlock only
+ * enforce release consistency, we need full fences here.
+ */
+#define smb_mb__before_spinlock()      smp_mb()
+#define smb_mb__after_spinlock()       smp_mb()
+
+#include <asm-generic/barrier.h>
+
+#endif /* __ASSEMBLY__ */
+
+#endif /* _ASM_RISCV_BARRIER_H */
diff --git a/arch/riscv/include/asm/bitops.h b/arch/riscv/include/asm/bitops.h
new file mode 100644 (file)
index 0000000..7c281ef
--- /dev/null
@@ -0,0 +1,218 @@
+/*
+ * Copyright (C) 2012 Regents of the University of California
+ *
+ *   This program is free software; you can redistribute it and/or
+ *   modify it under the terms of the GNU General Public License
+ *   as published by the Free Software Foundation, version 2.
+ *
+ *   This program is distributed in the hope that it will be useful,
+ *   but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *   GNU General Public License for more details.
+ */
+
+#ifndef _ASM_RISCV_BITOPS_H
+#define _ASM_RISCV_BITOPS_H
+
+#ifndef _LINUX_BITOPS_H
+#error "Only <linux/bitops.h> can be included directly"
+#endif /* _LINUX_BITOPS_H */
+
+#include <linux/compiler.h>
+#include <linux/irqflags.h>
+#include <asm/barrier.h>
+#include <asm/bitsperlong.h>
+
+#ifndef smp_mb__before_clear_bit
+#define smp_mb__before_clear_bit()  smp_mb()
+#define smp_mb__after_clear_bit()   smp_mb()
+#endif /* smp_mb__before_clear_bit */
+
+#include <asm-generic/bitops/__ffs.h>
+#include <asm-generic/bitops/ffz.h>
+#include <asm-generic/bitops/fls.h>
+#include <asm-generic/bitops/__fls.h>
+#include <asm-generic/bitops/fls64.h>
+#include <asm-generic/bitops/find.h>
+#include <asm-generic/bitops/sched.h>
+#include <asm-generic/bitops/ffs.h>
+
+#include <asm-generic/bitops/hweight.h>
+
+#if (BITS_PER_LONG == 64)
+#define __AMO(op)      "amo" #op ".d"
+#elif (BITS_PER_LONG == 32)
+#define __AMO(op)      "amo" #op ".w"
+#else
+#error "Unexpected BITS_PER_LONG"
+#endif
+
+#define __test_and_op_bit_ord(op, mod, nr, addr, ord)          \
+({                                                             \
+       unsigned long __res, __mask;                            \
+       __mask = BIT_MASK(nr);                                  \
+       __asm__ __volatile__ (                                  \
+               __AMO(op) #ord " %0, %2, %1"                    \
+               : "=r" (__res), "+A" (addr[BIT_WORD(nr)])       \
+               : "r" (mod(__mask))                             \
+               : "memory");                                    \
+       ((__res & __mask) != 0);                                \
+})
+
+#define __op_bit_ord(op, mod, nr, addr, ord)                   \
+       __asm__ __volatile__ (                                  \
+               __AMO(op) #ord " zero, %1, %0"                  \
+               : "+A" (addr[BIT_WORD(nr)])                     \
+               : "r" (mod(BIT_MASK(nr)))                       \
+               : "memory");
+
+#define __test_and_op_bit(op, mod, nr, addr)                   \
+       __test_and_op_bit_ord(op, mod, nr, addr, )
+#define __op_bit(op, mod, nr, addr)                            \
+       __op_bit_ord(op, mod, nr, addr, )
+
+/* Bitmask modifiers */
+#define __NOP(x)       (x)
+#define __NOT(x)       (~(x))
+
+/**
+ * test_and_set_bit - Set a bit and return its old value
+ * @nr: Bit to set
+ * @addr: Address to count from
+ *
+ * This operation may be reordered on other architectures than x86.
+ */
+static inline int test_and_set_bit(int nr, volatile unsigned long *addr)
+{
+       return __test_and_op_bit(or, __NOP, nr, addr);
+}
+
+/**
+ * test_and_clear_bit - Clear a bit and return its old value
+ * @nr: Bit to clear
+ * @addr: Address to count from
+ *
+ * This operation can be reordered on other architectures other than x86.
+ */
+static inline int test_and_clear_bit(int nr, volatile unsigned long *addr)
+{
+       return __test_and_op_bit(and, __NOT, nr, addr);
+}
+
+/**
+ * test_and_change_bit - Change a bit and return its old value
+ * @nr: Bit to change
+ * @addr: Address to count from
+ *
+ * This operation is atomic and cannot be reordered.
+ * It also implies a memory barrier.
+ */
+static inline int test_and_change_bit(int nr, volatile unsigned long *addr)
+{
+       return __test_and_op_bit(xor, __NOP, nr, addr);
+}
+
+/**
+ * set_bit - Atomically set a bit in memory
+ * @nr: the bit to set
+ * @addr: the address to start counting from
+ *
+ * Note: there are no guarantees that this function will not be reordered
+ * on non x86 architectures, so if you are writing portable code,
+ * make sure not to rely on its reordering guarantees.
+ *
+ * Note that @nr may be almost arbitrarily large; this function is not
+ * restricted to acting on a single-word quantity.
+ */
+static inline void set_bit(int nr, volatile unsigned long *addr)
+{
+       __op_bit(or, __NOP, nr, addr);
+}
+
+/**
+ * clear_bit - Clears a bit in memory
+ * @nr: Bit to clear
+ * @addr: Address to start counting from
+ *
+ * Note: there are no guarantees that this function will not be reordered
+ * on non x86 architectures, so if you are writing portable code,
+ * make sure not to rely on its reordering guarantees.
+ */
+static inline void clear_bit(int nr, volatile unsigned long *addr)
+{
+       __op_bit(and, __NOT, nr, addr);
+}
+
+/**
+ * change_bit - Toggle a bit in memory
+ * @nr: Bit to change
+ * @addr: Address to start counting from
+ *
+ * change_bit()  may be reordered on other architectures than x86.
+ * Note that @nr may be almost arbitrarily large; this function is not
+ * restricted to acting on a single-word quantity.
+ */
+static inline void change_bit(int nr, volatile unsigned long *addr)
+{
+       __op_bit(xor, __NOP, nr, addr);
+}
+
+/**
+ * test_and_set_bit_lock - Set a bit and return its old value, for lock
+ * @nr: Bit to set
+ * @addr: Address to count from
+ *
+ * This operation is atomic and provides acquire barrier semantics.
+ * It can be used to implement bit locks.
+ */
+static inline int test_and_set_bit_lock(
+       unsigned long nr, volatile unsigned long *addr)
+{
+       return __test_and_op_bit_ord(or, __NOP, nr, addr, .aq);
+}
+
+/**
+ * clear_bit_unlock - Clear a bit in memory, for unlock
+ * @nr: the bit to set
+ * @addr: the address to start counting from
+ *
+ * This operation is atomic and provides release barrier semantics.
+ */
+static inline void clear_bit_unlock(
+       unsigned long nr, volatile unsigned long *addr)
+{
+       __op_bit_ord(and, __NOT, nr, addr, .rl);
+}
+
+/**
+ * __clear_bit_unlock - Clear a bit in memory, for unlock
+ * @nr: the bit to set
+ * @addr: the address to start counting from
+ *
+ * This operation is like clear_bit_unlock, however it is not atomic.
+ * It does provide release barrier semantics so it can be used to unlock
+ * a bit lock, however it would only be used if no other CPU can modify
+ * any bits in the memory until the lock is released (a good example is
+ * if the bit lock itself protects access to the other bits in the word).
+ *
+ * On RISC-V systems there seems to be no benefit to taking advantage of the
+ * non-atomic property here: it's a lot more instructions and we still have to
+ * provide release semantics anyway.
+ */
+static inline void __clear_bit_unlock(
+       unsigned long nr, volatile unsigned long *addr)
+{
+       clear_bit_unlock(nr, addr);
+}
+
+#undef __test_and_op_bit
+#undef __op_bit
+#undef __NOP
+#undef __NOT
+#undef __AMO
+
+#include <asm-generic/bitops/non-atomic.h>
+#include <asm-generic/bitops/le.h>
+#include <asm-generic/bitops/ext2-atomic.h>
+
+#endif /* _ASM_RISCV_BITOPS_H */
diff --git a/arch/riscv/include/asm/cacheflush.h b/arch/riscv/include/asm/cacheflush.h
new file mode 100644 (file)
index 0000000..0595585
--- /dev/null
@@ -0,0 +1,39 @@
+/*
+ * Copyright (C) 2015 Regents of the University of California
+ *
+ *   This program is free software; you can redistribute it and/or
+ *   modify it under the terms of the GNU General Public License
+ *   as published by the Free Software Foundation, version 2.
+ *
+ *   This program is distributed in the hope that it will be useful,
+ *   but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *   GNU General Public License for more details.
+ */
+
+#ifndef _ASM_RISCV_CACHEFLUSH_H
+#define _ASM_RISCV_CACHEFLUSH_H
+
+#include <asm-generic/cacheflush.h>
+
+#undef flush_icache_range
+#undef flush_icache_user_range
+
+static inline void local_flush_icache_all(void)
+{
+       asm volatile ("fence.i" ::: "memory");
+}
+
+#ifndef CONFIG_SMP
+
+#define flush_icache_range(start, end) local_flush_icache_all()
+#define flush_icache_user_range(vma, pg, addr, len) local_flush_icache_all()
+
+#else /* CONFIG_SMP */
+
+#define flush_icache_range(start, end) sbi_remote_fence_i(0)
+#define flush_icache_user_range(vma, pg, addr, len) sbi_remote_fence_i(0)
+
+#endif /* CONFIG_SMP */
+
+#endif /* _ASM_RISCV_CACHEFLUSH_H */
diff --git a/arch/riscv/include/asm/cmpxchg.h b/arch/riscv/include/asm/cmpxchg.h
new file mode 100644 (file)
index 0000000..db249db
--- /dev/null
@@ -0,0 +1,134 @@
+/*
+ * Copyright (C) 2014 Regents of the University of California
+ *
+ *   This program is free software; you can redistribute it and/or
+ *   modify it under the terms of the GNU General Public License
+ *   as published by the Free Software Foundation, version 2.
+ *
+ *   This program is distributed in the hope that it will be useful,
+ *   but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *   GNU General Public License for more details.
+ */
+
+#ifndef _ASM_RISCV_CMPXCHG_H
+#define _ASM_RISCV_CMPXCHG_H
+
+#include <linux/bug.h>
+
+#include <asm/barrier.h>
+
+#define __xchg(new, ptr, size, asm_or)                         \
+({                                                             \
+       __typeof__(ptr) __ptr = (ptr);                          \
+       __typeof__(new) __new = (new);                          \
+       __typeof__(*(ptr)) __ret;                               \
+       switch (size) {                                         \
+       case 4:                                                 \
+               __asm__ __volatile__ (                          \
+                       "amoswap.w" #asm_or " %0, %2, %1"       \
+                       : "=r" (__ret), "+A" (*__ptr)           \
+                       : "r" (__new)                           \
+                       : "memory");                            \
+               break;                                          \
+       case 8:                                                 \
+               __asm__ __volatile__ (                          \
+                       "amoswap.d" #asm_or " %0, %2, %1"       \
+                       : "=r" (__ret), "+A" (*__ptr)           \
+                       : "r" (__new)                           \
+                       : "memory");                            \
+               break;                                          \
+       default:                                                \
+               BUILD_BUG();                                    \
+       }                                                       \
+       __ret;                                                  \
+})
+
+#define xchg(ptr, x)    (__xchg((x), (ptr), sizeof(*(ptr)), .aqrl))
+
+#define xchg32(ptr, x)                         \
+({                                             \
+       BUILD_BUG_ON(sizeof(*(ptr)) != 4);      \
+       xchg((ptr), (x));                       \
+})
+
+#define xchg64(ptr, x)                         \
+({                                             \
+       BUILD_BUG_ON(sizeof(*(ptr)) != 8);      \
+       xchg((ptr), (x));                       \
+})
+
+/*
+ * Atomic compare and exchange.  Compare OLD with MEM, if identical,
+ * store NEW in MEM.  Return the initial value in MEM.  Success is
+ * indicated by comparing RETURN with OLD.
+ */
+#define __cmpxchg(ptr, old, new, size, lrb, scb)                       \
+({                                                                     \
+       __typeof__(ptr) __ptr = (ptr);                                  \
+       __typeof__(*(ptr)) __old = (old);                               \
+       __typeof__(*(ptr)) __new = (new);                               \
+       __typeof__(*(ptr)) __ret;                                       \
+       register unsigned int __rc;                                     \
+       switch (size) {                                                 \
+       case 4:                                                         \
+               __asm__ __volatile__ (                                  \
+               "0:"                                                    \
+                       "lr.w" #scb " %0, %2\n"                         \
+                       "bne         %0, %z3, 1f\n"                     \
+                       "sc.w" #lrb " %1, %z4, %2\n"                    \
+                       "bnez        %1, 0b\n"                          \
+               "1:"                                                    \
+                       : "=&r" (__ret), "=&r" (__rc), "+A" (*__ptr)    \
+                       : "rJ" (__old), "rJ" (__new)                    \
+                       : "memory");                                    \
+               break;                                                  \
+       case 8:                                                         \
+               __asm__ __volatile__ (                                  \
+               "0:"                                                    \
+                       "lr.d" #scb " %0, %2\n"                         \
+                       "bne         %0, %z3, 1f\n"                     \
+                       "sc.d" #lrb " %1, %z4, %2\n"                    \
+                       "bnez        %1, 0b\n"                          \
+               "1:"                                                    \
+                       : "=&r" (__ret), "=&r" (__rc), "+A" (*__ptr)    \
+                       : "rJ" (__old), "rJ" (__new)                    \
+                       : "memory");                                    \
+               break;                                                  \
+       default:                                                        \
+               BUILD_BUG();                                            \
+       }                                                               \
+       __ret;                                                          \
+})
+
+#define cmpxchg(ptr, o, n) \
+       (__cmpxchg((ptr), (o), (n), sizeof(*(ptr)), .aqrl, .aqrl))
+
+#define cmpxchg_local(ptr, o, n) \
+       (__cmpxchg((ptr), (o), (n), sizeof(*(ptr)), , ))
+
+#define cmpxchg32(ptr, o, n)                   \
+({                                             \
+       BUILD_BUG_ON(sizeof(*(ptr)) != 4);      \
+       cmpxchg((ptr), (o), (n));               \
+})
+
+#define cmpxchg32_local(ptr, o, n)             \
+({                                             \
+       BUILD_BUG_ON(sizeof(*(ptr)) != 4);      \
+       cmpxchg_local((ptr), (o), (n));         \
+})
+
+#define cmpxchg64(ptr, o, n)                   \
+({                                             \
+       BUILD_BUG_ON(sizeof(*(ptr)) != 8);      \
+       cmpxchg((ptr), (o), (n));               \
+})
+
+#define cmpxchg64_local(ptr, o, n)             \
+({                                             \
+       BUILD_BUG_ON(sizeof(*(ptr)) != 8);      \
+       cmpxchg_local((ptr), (o), (n));         \
+})
+
+#endif /* _ASM_RISCV_CMPXCHG_H */
diff --git a/arch/riscv/include/asm/io.h b/arch/riscv/include/asm/io.h
new file mode 100644 (file)
index 0000000..c1f32cf
--- /dev/null
@@ -0,0 +1,303 @@
+/*
+ * {read,write}{b,w,l,q} based on arch/arm64/include/asm/io.h
+ *   which was based on arch/arm/include/io.h
+ *
+ * Copyright (C) 1996-2000 Russell King
+ * Copyright (C) 2012 ARM Ltd.
+ * Copyright (C) 2014 Regents of the University of California
+ *
+ *   This program is free software; you can redistribute it and/or
+ *   modify it under the terms of the GNU General Public License
+ *   as published by the Free Software Foundation, version 2.
+ *
+ *   This program is distributed in the hope that it will be useful,
+ *   but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *   GNU General Public License for more details.
+ */
+
+#ifndef _ASM_RISCV_IO_H
+#define _ASM_RISCV_IO_H
+
+#ifdef CONFIG_MMU
+
+extern void __iomem *ioremap(phys_addr_t offset, unsigned long size);
+
+/*
+ * The RISC-V ISA doesn't yet specify how to query or modify PMAs, so we can't
+ * change the properties of memory regions.  This should be fixed by the
+ * upcoming platform spec.
+ */
+#define ioremap_nocache(addr, size) ioremap((addr), (size))
+#define ioremap_wc(addr, size) ioremap((addr), (size))
+#define ioremap_wt(addr, size) ioremap((addr), (size))
+
+extern void iounmap(void __iomem *addr);
+
+#endif /* CONFIG_MMU */
+
+/* Generic IO read/write.  These perform native-endian accesses. */
+#define __raw_writeb __raw_writeb
+static inline void __raw_writeb(u8 val, volatile void __iomem *addr)
+{
+       asm volatile("sb %0, 0(%1)" : : "r" (val), "r" (addr));
+}
+
+#define __raw_writew __raw_writew
+static inline void __raw_writew(u16 val, volatile void __iomem *addr)
+{
+       asm volatile("sh %0, 0(%1)" : : "r" (val), "r" (addr));
+}
+
+#define __raw_writel __raw_writel
+static inline void __raw_writel(u32 val, volatile void __iomem *addr)
+{
+       asm volatile("sw %0, 0(%1)" : : "r" (val), "r" (addr));
+}
+
+#ifdef CONFIG_64BIT
+#define __raw_writeq __raw_writeq
+static inline void __raw_writeq(u64 val, volatile void __iomem *addr)
+{
+       asm volatile("sd %0, 0(%1)" : : "r" (val), "r" (addr));
+}
+#endif
+
+#define __raw_readb __raw_readb
+static inline u8 __raw_readb(const volatile void __iomem *addr)
+{
+       u8 val;
+
+       asm volatile("lb %0, 0(%1)" : "=r" (val) : "r" (addr));
+       return val;
+}
+
+#define __raw_readw __raw_readw
+static inline u16 __raw_readw(const volatile void __iomem *addr)
+{
+       u16 val;
+
+       asm volatile("lh %0, 0(%1)" : "=r" (val) : "r" (addr));
+       return val;
+}
+
+#define __raw_readl __raw_readl
+static inline u32 __raw_readl(const volatile void __iomem *addr)
+{
+       u32 val;
+
+       asm volatile("lw %0, 0(%1)" : "=r" (val) : "r" (addr));
+       return val;
+}
+
+#ifdef CONFIG_64BIT
+#define __raw_readq __raw_readq
+static inline u64 __raw_readq(const volatile void __iomem *addr)
+{
+       u64 val;
+
+       asm volatile("ld %0, 0(%1)" : "=r" (val) : "r" (addr));
+       return val;
+}
+#endif
+
+/*
+ * FIXME: I'm flip-flopping on whether or not we should keep this or enforce
+ * the ordering with I/O on spinlocks like PowerPC does.  The worry is that
+ * drivers won't get this correct, but I also don't want to introduce a fence
+ * into the lock code that otherwise only uses AMOs (and is essentially defined
+ * by the ISA to be correct).   For now I'm leaving this here: "o,w" is
+ * sufficient to ensure that all writes to the device have completed before the
+ * write to the spinlock is allowed to commit.  I surmised this from reading
+ * "ACQUIRES VS I/O ACCESSES" in memory-barriers.txt.
+ */
+#define mmiowb()       __asm__ __volatile__ ("fence o,w" : : : "memory");
+
+/*
+ * Unordered I/O memory access primitives.  These are even more relaxed than
+ * the relaxed versions, as they don't even order accesses between successive
+ * operations to the I/O regions.
+ */
+#define readb_cpu(c)           ({ u8  __r = __raw_readb(c); __r; })
+#define readw_cpu(c)           ({ u16 __r = le16_to_cpu((__force __le16)__raw_readw(c)); __r; })
+#define readl_cpu(c)           ({ u32 __r = le32_to_cpu((__force __le32)__raw_readl(c)); __r; })
+
+#define writeb_cpu(v,c)                ((void)__raw_writeb((v),(c)))
+#define writew_cpu(v,c)                ((void)__raw_writew((__force u16)cpu_to_le16(v),(c)))
+#define writel_cpu(v,c)                ((void)__raw_writel((__force u32)cpu_to_le32(v),(c)))
+
+#ifdef CONFIG_64BIT
+#define readq_cpu(c)           ({ u64 __r = le64_to_cpu((__force __le64)__raw_readq(c)); __r; })
+#define writeq_cpu(v,c)                ((void)__raw_writeq((__force u64)cpu_to_le64(v),(c)))
+#endif
+
+/*
+ * Relaxed I/O memory access primitives. These follow the Device memory
+ * ordering rules but do not guarantee any ordering relative to Normal memory
+ * accesses.  These are defined to order the indicated access (either a read or
+ * write) with all other I/O memory accesses. Since the platform specification
+ * defines that all I/O regions are strongly ordered on channel 2, no explicit
+ * fences are required to enforce this ordering.
+ */
+/* FIXME: These are now the same as asm-generic */
+#define __io_rbr()             do {} while (0)
+#define __io_rar()             do {} while (0)
+#define __io_rbw()             do {} while (0)
+#define __io_raw()             do {} while (0)
+
+#define readb_relaxed(c)       ({ u8  __v; __io_rbr(); __v = readb_cpu(c); __io_rar(); __v; })
+#define readw_relaxed(c)       ({ u16 __v; __io_rbr(); __v = readw_cpu(c); __io_rar(); __v; })
+#define readl_relaxed(c)       ({ u32 __v; __io_rbr(); __v = readl_cpu(c); __io_rar(); __v; })
+
+#define writeb_relaxed(v,c)    ({ __io_rbw(); writeb_cpu((v),(c)); __io_raw(); })
+#define writew_relaxed(v,c)    ({ __io_rbw(); writew_cpu((v),(c)); __io_raw(); })
+#define writel_relaxed(v,c)    ({ __io_rbw(); writel_cpu((v),(c)); __io_raw(); })
+
+#ifdef CONFIG_64BIT
+#define readq_relaxed(c)       ({ u64 __v; __io_rbr(); __v = readq_cpu(c); __io_rar(); __v; })
+#define writeq_relaxed(v,c)    ({ __io_rbw(); writeq_cpu((v),(c)); __io_raw(); })
+#endif
+
+/*
+ * I/O memory access primitives. Reads are ordered relative to any
+ * following Normal memory access. Writes are ordered relative to any prior
+ * Normal memory access.  The memory barriers here are necessary as RISC-V
+ * doesn't define any ordering between the memory space and the I/O space.
+ */
+#define __io_br()      do {} while (0)
+#define __io_ar()      __asm__ __volatile__ ("fence i,r" : : : "memory");
+#define __io_bw()      __asm__ __volatile__ ("fence w,o" : : : "memory");
+#define __io_aw()      do {} while (0)
+
+#define readb(c)       ({ u8  __v; __io_br(); __v = readb_cpu(c); __io_ar(); __v; })
+#define readw(c)       ({ u16 __v; __io_br(); __v = readw_cpu(c); __io_ar(); __v; })
+#define readl(c)       ({ u32 __v; __io_br(); __v = readl_cpu(c); __io_ar(); __v; })
+
+#define writeb(v,c)    ({ __io_bw(); writeb_cpu((v),(c)); __io_aw(); })
+#define writew(v,c)    ({ __io_bw(); writew_cpu((v),(c)); __io_aw(); })
+#define writel(v,c)    ({ __io_bw(); writel_cpu((v),(c)); __io_aw(); })
+
+#ifdef CONFIG_64BIT
+#define readq(c)       ({ u64 __v; __io_br(); __v = readq_cpu(c); __io_ar(); __v; })
+#define writeq(v,c)    ({ __io_bw(); writeq_cpu((v),(c)); __io_aw(); })
+#endif
+
+/*
+ * Emulation routines for the port-mapped IO space used by some PCI drivers.
+ * These are defined as being "fully synchronous", but also "not guaranteed to
+ * be fully ordered with respect to other memory and I/O operations".  We're
+ * going to be on the safe side here and just make them:
+ *  - Fully ordered WRT each other, by bracketing them with two fences.  The
+ *    outer set contains both I/O so inX is ordered with outX, while the inner just
+ *    needs the type of the access (I for inX and O for outX).
+ *  - Ordered in the same manner as readX/writeX WRT memory by subsuming their
+ *    fences.
+ *  - Ordered WRT timer reads, so udelay and friends don't get elided by the
+ *    implementation.
+ * Note that there is no way to actually enforce that outX is a non-posted
+ * operation on RISC-V, but hopefully the timer ordering constraint is
+ * sufficient to ensure this works sanely on controllers that support I/O
+ * writes.
+ */
+#define __io_pbr()     __asm__ __volatile__ ("fence io,i"  : : : "memory");
+#define __io_par()     __asm__ __volatile__ ("fence i,ior" : : : "memory");
+#define __io_pbw()     __asm__ __volatile__ ("fence iow,o" : : : "memory");
+#define __io_paw()     __asm__ __volatile__ ("fence o,io"  : : : "memory");
+
+#define inb(c)         ({ u8  __v; __io_pbr(); __v = readb_cpu((void*)(PCI_IOBASE + (c))); __io_par(); __v; })
+#define inw(c)         ({ u16 __v; __io_pbr(); __v = readw_cpu((void*)(PCI_IOBASE + (c))); __io_par(); __v; })
+#define inl(c)         ({ u32 __v; __io_pbr(); __v = readl_cpu((void*)(PCI_IOBASE + (c))); __io_par(); __v; })
+
+#define outb(v,c)      ({ __io_pbw(); writeb_cpu((v),(void*)(PCI_IOBASE + (c))); __io_paw(); })
+#define outw(v,c)      ({ __io_pbw(); writew_cpu((v),(void*)(PCI_IOBASE + (c))); __io_paw(); })
+#define outl(v,c)      ({ __io_pbw(); writel_cpu((v),(void*)(PCI_IOBASE + (c))); __io_paw(); })
+
+#ifdef CONFIG_64BIT
+#define inq(c)         ({ u64 __v; __io_pbr(); __v = readq_cpu((void*)(c)); __io_par(); __v; })
+#define outq(v,c)      ({ __io_pbw(); writeq_cpu((v),(void*)(c)); __io_paw(); })
+#endif
+
+/*
+ * Accesses from a single hart to a single I/O address must be ordered.  This
+ * allows us to use the raw read macros, but we still need to fence before and
+ * after the block to ensure ordering WRT other macros.  These are defined to
+ * perform host-endian accesses so we use __raw instead of __cpu.
+ */
+#define __io_reads_ins(port, ctype, len, bfence, afence)                       \
+       static inline void __ ## port ## len(const volatile void __iomem *addr, \
+                                            void *buffer,                      \
+                                            unsigned int count)                \
+       {                                                                       \
+               bfence;                                                         \
+               if (count) {                                                    \
+                       ctype *buf = buffer;                                    \
+                                                                               \
+                       do {                                                    \
+                               ctype x = __raw_read ## len(addr);              \
+                               *buf++ = x;                                     \
+                       } while (--count);                                      \
+               }                                                               \
+               afence;                                                         \
+       }
+
+#define __io_writes_outs(port, ctype, len, bfence, afence)                     \
+       static inline void __ ## port ## len(volatile void __iomem *addr,       \
+                                            const void *buffer,                \
+                                            unsigned int count)                \
+       {                                                                       \
+               bfence;                                                         \
+               if (count) {                                                    \
+                       const ctype *buf = buffer;                              \
+                                                                               \
+                       do {                                                    \
+                               __raw_writeq(*buf++, addr);                     \
+                       } while (--count);                                      \
+               }                                                               \
+               afence;                                                         \
+       }
+
+__io_reads_ins(reads,  u8, b, __io_br(), __io_ar())
+__io_reads_ins(reads, u16, w, __io_br(), __io_ar())
+__io_reads_ins(reads, u32, l, __io_br(), __io_ar())
+#define readsb(addr, buffer, count) __readsb(addr, buffer, count)
+#define readsw(addr, buffer, count) __readsw(addr, buffer, count)
+#define readsl(addr, buffer, count) __readsl(addr, buffer, count)
+
+__io_reads_ins(ins,  u8, b, __io_pbr(), __io_par())
+__io_reads_ins(ins, u16, w, __io_pbr(), __io_par())
+__io_reads_ins(ins, u32, l, __io_pbr(), __io_par())
+#define insb(addr, buffer, count) __insb((void __iomem *)addr, buffer, count)
+#define insw(addr, buffer, count) __insw((void __iomem *)addr, buffer, count)
+#define insl(addr, buffer, count) __insl((void __iomem *)addr, buffer, count)
+
+__io_writes_outs(writes,  u8, b, __io_bw(), __io_aw())
+__io_writes_outs(writes, u16, w, __io_bw(), __io_aw())
+__io_writes_outs(writes, u32, l, __io_bw(), __io_aw())
+#define writesb(addr, buffer, count) __writesb(addr, buffer, count)
+#define writesw(addr, buffer, count) __writesw(addr, buffer, count)
+#define writesl(addr, buffer, count) __writesl(addr, buffer, count)
+
+__io_writes_outs(outs,  u8, b, __io_pbw(), __io_paw())
+__io_writes_outs(outs, u16, w, __io_pbw(), __io_paw())
+__io_writes_outs(outs, u32, l, __io_pbw(), __io_paw())
+#define outsb(addr, buffer, count) __outsb((void __iomem *)addr, buffer, count)
+#define outsw(addr, buffer, count) __outsw((void __iomem *)addr, buffer, count)
+#define outsl(addr, buffer, count) __outsl((void __iomem *)addr, buffer, count)
+
+#ifdef CONFIG_64BIT
+__io_reads_ins(reads, u64, q, __io_br(), __io_ar())
+#define readsq(addr, buffer, count) __readsq(addr, buffer, count)
+
+__io_reads_ins(ins, u64, q, __io_pbr(), __io_par())
+#define insq(addr, buffer, count) __insq((void __iomem *)addr, buffer, count)
+
+__io_writes_outs(writes, u64, q, __io_bw(), __io_aw())
+#define writesq(addr, buffer, count) __writesq(addr, buffer, count)
+
+__io_writes_outs(outs, u64, q, __io_pbr(), __io_paw())
+#define outsq(addr, buffer, count) __outsq((void __iomem *)addr, buffer, count)
+#endif
+
+#include <asm-generic/io.h>
+
+#endif /* _ASM_RISCV_IO_H */
diff --git a/arch/riscv/include/asm/spinlock.h b/arch/riscv/include/asm/spinlock.h
new file mode 100644 (file)
index 0000000..b3b394f
--- /dev/null
@@ -0,0 +1,165 @@
+/*
+ * Copyright (C) 2015 Regents of the University of California
+ * Copyright (C) 2017 SiFive
+ *
+ *   This program is free software; you can redistribute it and/or
+ *   modify it under the terms of the GNU General Public License
+ *   as published by the Free Software Foundation, version 2.
+ *
+ *   This program is distributed in the hope that it will be useful,
+ *   but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *   GNU General Public License for more details.
+ */
+
+#ifndef _ASM_RISCV_SPINLOCK_H
+#define _ASM_RISCV_SPINLOCK_H
+
+#include <linux/kernel.h>
+#include <asm/current.h>
+
+/*
+ * Simple spin lock operations.  These provide no fairness guarantees.
+ */
+
+/* FIXME: Replace this with a ticket lock, like MIPS. */
+
+#define arch_spin_lock_flags(lock, flags) arch_spin_lock(lock)
+#define arch_spin_is_locked(x) ((x)->lock != 0)
+
+static inline void arch_spin_unlock(arch_spinlock_t *lock)
+{
+       __asm__ __volatile__ (
+               "amoswap.w.rl x0, x0, %0"
+               : "=A" (lock->lock)
+               :: "memory");
+}
+
+static inline int arch_spin_trylock(arch_spinlock_t *lock)
+{
+       int tmp = 1, busy;
+
+       __asm__ __volatile__ (
+               "amoswap.w.aq %0, %2, %1"
+               : "=r" (busy), "+A" (lock->lock)
+               : "r" (tmp)
+               : "memory");
+
+       return !busy;
+}
+
+static inline void arch_spin_lock(arch_spinlock_t *lock)
+{
+       while (1) {
+               if (arch_spin_is_locked(lock))
+                       continue;
+
+               if (arch_spin_trylock(lock))
+                       break;
+       }
+}
+
+static inline void arch_spin_unlock_wait(arch_spinlock_t *lock)
+{
+       smp_rmb();
+       do {
+               cpu_relax();
+       } while (arch_spin_is_locked(lock));
+       smp_acquire__after_ctrl_dep();
+}
+
+/***********************************************************/
+
+static inline int arch_read_can_lock(arch_rwlock_t *lock)
+{
+       return lock->lock >= 0;
+}
+
+static inline int arch_write_can_lock(arch_rwlock_t *lock)
+{
+       return lock->lock == 0;
+}
+
+static inline void arch_read_lock(arch_rwlock_t *lock)
+{
+       int tmp;
+
+       __asm__ __volatile__(
+               "1:     lr.w    %1, %0\n"
+               "       bltz    %1, 1b\n"
+               "       addi    %1, %1, 1\n"
+               "       sc.w.aq %1, %1, %0\n"
+               "       bnez    %1, 1b\n"
+               : "+A" (lock->lock), "=&r" (tmp)
+               :: "memory");
+}
+
+static inline void arch_write_lock(arch_rwlock_t *lock)
+{
+       int tmp;
+
+       __asm__ __volatile__(
+               "1:     lr.w    %1, %0\n"
+               "       bnez    %1, 1b\n"
+               "       li      %1, -1\n"
+               "       sc.w.aq %1, %1, %0\n"
+               "       bnez    %1, 1b\n"
+               : "+A" (lock->lock), "=&r" (tmp)
+               :: "memory");
+}
+
+static inline int arch_read_trylock(arch_rwlock_t *lock)
+{
+       int busy;
+
+       __asm__ __volatile__(
+               "1:     lr.w    %1, %0\n"
+               "       bltz    %1, 1f\n"
+               "       addi    %1, %1, 1\n"
+               "       sc.w.aq %1, %1, %0\n"
+               "       bnez    %1, 1b\n"
+               "1:\n"
+               : "+A" (lock->lock), "=&r" (busy)
+               :: "memory");
+
+       return !busy;
+}
+
+static inline int arch_write_trylock(arch_rwlock_t *lock)
+{
+       int busy;
+
+       __asm__ __volatile__(
+               "1:     lr.w    %1, %0\n"
+               "       bnez    %1, 1f\n"
+               "       li      %1, -1\n"
+               "       sc.w.aq %1, %1, %0\n"
+               "       bnez    %1, 1b\n"
+               "1:\n"
+               : "+A" (lock->lock), "=&r" (busy)
+               :: "memory");
+
+       return !busy;
+}
+
+static inline void arch_read_unlock(arch_rwlock_t *lock)
+{
+       __asm__ __volatile__(
+               "amoadd.w.rl x0, %1, %0"
+               : "+A" (lock->lock)
+               : "r" (-1)
+               : "memory");
+}
+
+static inline void arch_write_unlock(arch_rwlock_t *lock)
+{
+       __asm__ __volatile__ (
+               "amoswap.w.rl x0, x0, %0"
+               : "=A" (lock->lock)
+               :: "memory");
+}
+
+#define arch_read_lock_flags(lock, flags) arch_read_lock(lock)
+#define arch_write_lock_flags(lock, flags) arch_write_lock(lock)
+
+#endif /* _ASM_RISCV_SPINLOCK_H */
diff --git a/arch/riscv/include/asm/spinlock_types.h b/arch/riscv/include/asm/spinlock_types.h
new file mode 100644 (file)
index 0000000..83ac4ac
--- /dev/null
@@ -0,0 +1,33 @@
+/*
+ * Copyright (C) 2015 Regents of the University of California
+ *
+ *   This program is free software; you can redistribute it and/or
+ *   modify it under the terms of the GNU General Public License
+ *   as published by the Free Software Foundation, version 2.
+ *
+ *   This program is distributed in the hope that it will be useful,
+ *   but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *   GNU General Public License for more details.
+ */
+
+#ifndef _ASM_RISCV_SPINLOCK_TYPES_H
+#define _ASM_RISCV_SPINLOCK_TYPES_H
+
+#ifndef __LINUX_SPINLOCK_TYPES_H
+# error "please don't include this file directly"
+#endif
+
+typedef struct {
+       volatile unsigned int lock;
+} arch_spinlock_t;
+
+#define __ARCH_SPIN_LOCK_UNLOCKED      { 0 }
+
+typedef struct {
+       volatile unsigned int lock;
+} arch_rwlock_t;
+
+#define __ARCH_RW_LOCK_UNLOCKED                { 0 }
+
+#endif
diff --git a/arch/riscv/include/asm/tlb.h b/arch/riscv/include/asm/tlb.h
new file mode 100644 (file)
index 0000000..c229509
--- /dev/null
@@ -0,0 +1,24 @@
+/*
+ * Copyright (C) 2012 Regents of the University of California
+ *
+ *   This program is free software; you can redistribute it and/or
+ *   modify it under the terms of the GNU General Public License
+ *   as published by the Free Software Foundation, version 2.
+ *
+ *   This program is distributed in the hope that it will be useful,
+ *   but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *   GNU General Public License for more details.
+ */
+
+#ifndef _ASM_RISCV_TLB_H
+#define _ASM_RISCV_TLB_H
+
+#include <asm-generic/tlb.h>
+
+static inline void tlb_flush(struct mmu_gather *tlb)
+{
+       flush_tlb_mm(tlb->mm);
+}
+
+#endif /* _ASM_RISCV_TLB_H */
diff --git a/arch/riscv/include/asm/tlbflush.h b/arch/riscv/include/asm/tlbflush.h
new file mode 100644 (file)
index 0000000..5ee4ae3
--- /dev/null
@@ -0,0 +1,64 @@
+/*
+ * Copyright (C) 2009 Chen Liqin <liqin.chen@sunplusct.com>
+ * Copyright (C) 2012 Regents of the University of California
+ *
+ *   This program is free software; you can redistribute it and/or
+ *   modify it under the terms of the GNU General Public License
+ *   as published by the Free Software Foundation, version 2.
+ *
+ *   This program is distributed in the hope that it will be useful,
+ *   but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *   GNU General Public License for more details.
+ */
+
+#ifndef _ASM_RISCV_TLBFLUSH_H
+#define _ASM_RISCV_TLBFLUSH_H
+
+#ifdef CONFIG_MMU
+
+/* Flush entire local TLB */
+static inline void local_flush_tlb_all(void)
+{
+       __asm__ __volatile__ ("sfence.vma" : : : "memory");
+}
+
+/* Flush one page from local TLB */
+static inline void local_flush_tlb_page(unsigned long addr)
+{
+       __asm__ __volatile__ ("sfence.vma %0" : : "r" (addr) : "memory");
+}
+
+#ifndef CONFIG_SMP
+
+#define flush_tlb_all() local_flush_tlb_all()
+#define flush_tlb_page(vma, addr) local_flush_tlb_page(addr)
+#define flush_tlb_range(vma, start, end) local_flush_tlb_all()
+
+#else /* CONFIG_SMP */
+
+#include <asm/sbi.h>
+
+#define flush_tlb_all() sbi_remote_sfence_vma(0, 0, -1)
+#define flush_tlb_page(vma, addr) flush_tlb_range(vma, addr, 0)
+#define flush_tlb_range(vma, start, end) \
+       sbi_remote_sfence_vma(0, start, (end) - (start))
+
+#endif /* CONFIG_SMP */
+
+/* Flush the TLB entries of the specified mm context */
+static inline void flush_tlb_mm(struct mm_struct *mm)
+{
+       flush_tlb_all();
+}
+
+/* Flush a range of kernel pages */
+static inline void flush_tlb_kernel_range(unsigned long start,
+       unsigned long end)
+{
+       flush_tlb_all();
+}
+
+#endif /* CONFIG_MMU */
+
+#endif /* _ASM_RISCV_TLBFLUSH_H */