From patchwork Sat Jan 27 03:57:37 2018 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Palmer Dabbelt X-Patchwork-Id: 25601 Received: (qmail 105845 invoked by alias); 27 Jan 2018 03:58:18 -0000 Mailing-List: contact libc-alpha-help@sourceware.org; run by ezmlm Precedence: bulk List-Id: List-Unsubscribe: List-Subscribe: List-Archive: List-Post: List-Help: , Sender: libc-alpha-owner@sourceware.org Delivered-To: mailing list libc-alpha@sourceware.org Received: (qmail 105736 invoked by uid 89); 27 Jan 2018 03:58:17 -0000 Authentication-Results: sourceware.org; auth=none X-Virus-Found: No X-Spam-SWARE-Status: No, score=-26.1 required=5.0 tests=AWL, BAYES_00, GIT_PATCH_0, GIT_PATCH_1, GIT_PATCH_2, GIT_PATCH_3, KAM_SHORT, RCVD_IN_DNSWL_NONE, SPF_PASS autolearn=ham version=3.3.2 spammy= X-HELO: mail-pl0-f65.google.com X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20161025; h=x-gm-message-state:subject:date:message-id:in-reply-to:references :cc:from:to; bh=avyI81ezIdimK21HfIRdjKbgTVMCXMQJXgVPZotiD68=; b=WsEX4BU697TJgaUtawv51l8WsHZlmOmWX6ZhFR6uCVZzmKeBi6UkX656IwiYuII265 dvHcGySTNR1+LS8DKKgLw3udKA0LarnFdabh6J3lI5Okyz+a2xjjHqPn/SMerWjhAMgd kqKWucPBxp6BhxohNgnWEpzTKZnD+Nxo8fpUhBaHXqLp3YkOZcumWBBQxzc4vA0ZtFBU xYUh5SHQV3NmmRkWBaEOdp/WzOIOk1+S8tPzYCiX55gAJAZKhP8ikYGg6/DU1L2gJ8Aq 7pqnYw2NdiuLTb1Rcwc+AF5+Nq7C9u/u5chO7DZtDMVumo+nWyvrHu2ct+G/eNMqIeMK Q7xA== X-Gm-Message-State: AKwxyteX+bTvHs0zdv/LgmMfF0ZqaLekPeBrukJe9FvUU27vD4STDCe3 TIgVxyOEVfhLUqshLmxsgsmCTg== X-Google-Smtp-Source: AH8x224odXfjuwFKR3SH6WFbA8Ki4LOpLpuhFsUGP6MbkdNibMCJrr8CPEwp0s3vIEkIUy0mq/hy1Q== X-Received: by 2002:a17:902:581a:: with SMTP id m26-v6mr15844194pli.158.1517025491689; Fri, 26 Jan 2018 19:58:11 -0800 (PST) Subject: [PATCH v7 10/17] RISC-V: Atomic and Locking Routines Date: Fri, 26 Jan 2018 19:57:37 -0800 Message-Id: <20180127035744.10058-11-palmer@dabbelt.com> In-Reply-To: <20180127035744.10058-1-palmer@dabbelt.com> References: <20180127035744.10058-1-palmer@dabbelt.com> Cc: Palmer Dabbelt From: Palmer Dabbelt To: libc-alpha@sourceware.org, joseph@codesourcery.com This patch implements various atomic and locking routines on RISC-V. We mandate the A extension on Linux-capable RISC-V systems, so this can rely on always having the various atomic instructions availiable. 2018-01-26 Palmer Dabbelt * sysdeps/riscv/nptl/bits/pthreadtypes-arch.h: New file. * sysdeps/riscv/nptl/bits/semaphore.h: Likewise. * sysdeps/riscv/nptl/libc-lowlevellock.c: Likewise. * sysdeps/unix/sysv/linux/riscv/atomic-machine.h: Likewise. --- sysdeps/riscv/nptl/bits/pthreadtypes-arch.h | 68 +++++++++ sysdeps/riscv/nptl/bits/semaphore.h | 33 +++++ sysdeps/riscv/nptl/libc-lowlevellock.c | 8 + sysdeps/unix/sysv/linux/riscv/atomic-machine.h | 194 +++++++++++++++++++++++++ 4 files changed, 303 insertions(+) create mode 100644 sysdeps/riscv/nptl/bits/pthreadtypes-arch.h create mode 100644 sysdeps/riscv/nptl/bits/semaphore.h create mode 100644 sysdeps/riscv/nptl/libc-lowlevellock.c create mode 100644 sysdeps/unix/sysv/linux/riscv/atomic-machine.h diff --git a/sysdeps/riscv/nptl/bits/pthreadtypes-arch.h b/sysdeps/riscv/nptl/bits/pthreadtypes-arch.h new file mode 100644 index 000000000000..f6a55b6233a4 --- /dev/null +++ b/sysdeps/riscv/nptl/bits/pthreadtypes-arch.h @@ -0,0 +1,68 @@ +/* Machine-specific pthread type layouts. RISC-V version. + Copyright (C) 2011-2018 Free Software Foundation, Inc. + This file is part of the GNU C Library. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library 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 + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library. If not, see + . */ + +#ifndef _BITS_PTHREADTYPES_ARCH_H +#define _BITS_PTHREADTYPES_ARCH_H 1 + +#include + +#if __riscv_xlen == 64 +# define __SIZEOF_PTHREAD_ATTR_T 56 +# define __SIZEOF_PTHREAD_MUTEX_T 40 +# define __SIZEOF_PTHREAD_MUTEXATTR_T 4 +# define __SIZEOF_PTHREAD_COND_T 48 +# define __SIZEOF_PTHREAD_CONDATTR_T 4 +# define __SIZEOF_PTHREAD_RWLOCK_T 56 +# define __SIZEOF_PTHREAD_RWLOCKATTR_T 8 +# define __SIZEOF_PTHREAD_BARRIER_T 32 +# define __SIZEOF_PTHREAD_BARRIERATTR_T 4 +#else +# error "rv32i-based systems are not supported" +#endif + +#define __PTHREAD_COMPAT_PADDING_MID +#define __PTHREAD_COMPAT_PADDING_END +#define __PTHREAD_MUTEX_LOCK_ELISION 0 +#define __PTHREAD_MUTEX_USE_UNION 0 +#define __PTHREAD_MUTEX_NUSERS_AFTER_KIND 0 + +#define __LOCK_ALIGNMENT +#define __ONCE_ALIGNMENT + +/* There is a lot of padding in this structure. While it's not strictly + necessary on RISC-V, we're going to leave it in to be on the safe side in + case it's needed in the future. Most other architectures have the padding, + so this gives us the same extensibility as everyone else has. */ +struct __pthread_rwlock_arch_t +{ + unsigned int __readers; + unsigned int __writers; + unsigned int __wrphase_futex; + unsigned int __writers_futex; + unsigned int __pad3; + unsigned int __pad4; + int __cur_writer; + int __shared; + unsigned long int __pad1; + unsigned long int __pad2; + unsigned int __flags; +}; + +#define __PTHREAD_RWLOCK_ELISION_EXTRA 0 + +#endif /* bits/pthreadtypes.h */ diff --git a/sysdeps/riscv/nptl/bits/semaphore.h b/sysdeps/riscv/nptl/bits/semaphore.h new file mode 100644 index 000000000000..2fed17d0e530 --- /dev/null +++ b/sysdeps/riscv/nptl/bits/semaphore.h @@ -0,0 +1,33 @@ +/* Machine-specific POSIX semaphore type layouts. RISC-V version. + Copyright (C) 2002-2018 Free Software Foundation, Inc. + This file is part of the GNU C Library. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library 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 + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library. If not, see + . */ + +#ifndef _SEMAPHORE_H +# error "Never use directly; include instead." +#endif + +#define __SIZEOF_SEM_T (4 * __SIZEOF_POINTER__) + +/* Value returned if `sem_open' failed. */ +#define SEM_FAILED ((sem_t *) 0) + + +typedef union +{ + char __size[__SIZEOF_SEM_T]; + long int __align; +} sem_t; diff --git a/sysdeps/riscv/nptl/libc-lowlevellock.c b/sysdeps/riscv/nptl/libc-lowlevellock.c new file mode 100644 index 000000000000..9523fb46a596 --- /dev/null +++ b/sysdeps/riscv/nptl/libc-lowlevellock.c @@ -0,0 +1,8 @@ +/* This kludge works around a libpthread static linking problem: + https://sourceware.org/bugzilla/show_bug.cgi?id=15648. */ + +#ifndef SHARED +# define __lll_lock_wait_private weak_function __lll_lock_wait_private +#endif + +#include diff --git a/sysdeps/unix/sysv/linux/riscv/atomic-machine.h b/sysdeps/unix/sysv/linux/riscv/atomic-machine.h new file mode 100644 index 000000000000..3d967d3b08f0 --- /dev/null +++ b/sysdeps/unix/sysv/linux/riscv/atomic-machine.h @@ -0,0 +1,194 @@ +/* Low-level functions for atomic operations. RISC-V version. + Copyright (C) 2014-2018 Free Software Foundation, Inc. + This file is part of the GNU C Library. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library 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 + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library. If not, see + . */ + +#ifndef _LINUX_RISCV_BITS_ATOMIC_H +#define _LINUX_RISCV_BITS_ATOMIC_H 1 + +#include + +typedef int32_t atomic32_t; +typedef uint32_t uatomic32_t; + +typedef int64_t atomic64_t; +typedef uint64_t uatomic64_t; + +typedef intptr_t atomicptr_t; +typedef uintptr_t uatomicptr_t; +typedef intmax_t atomic_max_t; +typedef uintmax_t uatomic_max_t; + +#define atomic_full_barrier() __sync_synchronize () + +#ifdef __riscv_atomic + +# define __HAVE_64B_ATOMICS (__riscv_xlen >= 64) +# define USE_ATOMIC_COMPILER_BUILTINS 1 +# define ATOMIC_EXCHANGE_USES_CAS 0 + +/* Compare and exchange. + For all "bool" routines, we return FALSE if exchange succesful. */ + +# define __arch_compare_and_exchange_bool_8_int(mem, newval, oldval, model) \ + ({ \ + typeof (*mem) __oldval = (oldval); \ + !__atomic_compare_exchange_n (mem, (void *) &__oldval, newval, 0, \ + model, __ATOMIC_RELAXED); \ + }) + +# define __arch_compare_and_exchange_bool_16_int(mem, newval, oldval, model) \ + ({ \ + typeof (*mem) __oldval = (oldval); \ + !__atomic_compare_exchange_n (mem, (void *) &__oldval, newval, 0, \ + model, __ATOMIC_RELAXED); \ + }) + +# define __arch_compare_and_exchange_bool_32_int(mem, newval, oldval, model) \ + ({ \ + typeof (*mem) __oldval = (oldval); \ + !__atomic_compare_exchange_n (mem, (void *) &__oldval, newval, 0, \ + model, __ATOMIC_RELAXED); \ + }) + +# define __arch_compare_and_exchange_bool_64_int(mem, newval, oldval, model) \ + ({ \ + typeof (*mem) __oldval = (oldval); \ + !__atomic_compare_exchange_n (mem, (void *) &__oldval, newval, 0, \ + model, __ATOMIC_RELAXED); \ + }) + +# define __arch_compare_and_exchange_val_8_int(mem, newval, oldval, model) \ + ({ \ + typeof (*mem) __oldval = (oldval); \ + __atomic_compare_exchange_n (mem, (void *) &__oldval, newval, 0, \ + model, __ATOMIC_RELAXED); \ + __oldval; \ + }) + +# define __arch_compare_and_exchange_val_16_int(mem, newval, oldval, model) \ + ({ \ + typeof (*mem) __oldval = (oldval); \ + __atomic_compare_exchange_n (mem, (void *) &__oldval, newval, 0, \ + model, __ATOMIC_RELAXED); \ + __oldval; \ + }) + +# define __arch_compare_and_exchange_val_32_int(mem, newval, oldval, model) \ + ({ \ + typeof (*mem) __oldval = (oldval); \ + __atomic_compare_exchange_n (mem, (void *) &__oldval, newval, 0, \ + model, __ATOMIC_RELAXED); \ + __oldval; \ + }) + +# define __arch_compare_and_exchange_val_64_int(mem, newval, oldval, model) \ + ({ \ + typeof (*mem) __oldval = (oldval); \ + __atomic_compare_exchange_n (mem, (void *) &__oldval, newval, 0, \ + model, __ATOMIC_RELAXED); \ + __oldval; \ + }) + +/* Atomic compare and exchange. */ + +# define atomic_compare_and_exchange_bool_acq(mem, new, old) \ + __atomic_bool_bysize (__arch_compare_and_exchange_bool, int, \ + mem, new, old, __ATOMIC_ACQUIRE) + +# define atomic_compare_and_exchange_val_acq(mem, new, old) \ + __atomic_val_bysize (__arch_compare_and_exchange_val, int, \ + mem, new, old, __ATOMIC_ACQUIRE) + +# define atomic_compare_and_exchange_val_rel(mem, new, old) \ + __atomic_val_bysize (__arch_compare_and_exchange_val, int, \ + mem, new, old, __ATOMIC_RELEASE) + +/* Atomic exchange (without compare). */ + +# define __arch_exchange_8_int(mem, newval, model) \ + __atomic_exchange_n (mem, newval, model) + +# define __arch_exchange_16_int(mem, newval, model) \ + __atomic_exchange_n (mem, newval, model) + +# define __arch_exchange_32_int(mem, newval, model) \ + __atomic_exchange_n (mem, newval, model) + +# define __arch_exchange_64_int(mem, newval, model) \ + __atomic_exchange_n (mem, newval, model) + +# define atomic_exchange_acq(mem, value) \ + __atomic_val_bysize (__arch_exchange, int, mem, value, __ATOMIC_ACQUIRE) + +# define atomic_exchange_rel(mem, value) \ + __atomic_val_bysize (__arch_exchange, int, mem, value, __ATOMIC_RELEASE) + +/* Atomically add value and return the previous (unincremented) value. */ + +# define __arch_exchange_and_add_8_int(mem, value, model) \ + __atomic_fetch_add (mem, value, model) + +# define __arch_exchange_and_add_16_int(mem, value, model) \ + __atomic_fetch_add (mem, value, model) + +# define __arch_exchange_and_add_32_int(mem, value, model) \ + __atomic_fetch_add (mem, value, model) + +# define __arch_exchange_and_add_64_int(mem, value, model) \ + __atomic_fetch_add (mem, value, model) + +# define atomic_exchange_and_add_acq(mem, value) \ + __atomic_val_bysize (__arch_exchange_and_add, int, mem, value, \ + __ATOMIC_ACQUIRE) + +# define atomic_exchange_and_add_rel(mem, value) \ + __atomic_val_bysize (__arch_exchange_and_add, int, mem, value, \ + __ATOMIC_RELEASE) + +/* Miscellaneous. */ + +# define asm_amo(which, ordering, mem, value) ({ \ + __atomic_check_size (mem); \ + typeof (*mem) __tmp; \ + if (sizeof (__tmp) == 4) \ + asm volatile (which ".w" ordering "\t%0, %z2, %1" \ + : "=r" (__tmp), "+A" (* (mem)) \ + : "rJ" (value)); \ + else if (sizeof (__tmp) == 8) \ + asm volatile (which ".d" ordering "\t%0, %z2, %1" \ + : "=r" (__tmp), "+A" (* (mem)) \ + : "rJ" (value)); \ + else \ + abort (); \ + __tmp; }) + +# define atomic_max(mem, value) asm_amo ("amomaxu", ".aq", mem, value) +# define atomic_min(mem, value) asm_amo ("amominu", ".aq", mem, value) + +# define atomic_bit_test_set(mem, bit) \ + ({ typeof (*mem) __mask = (typeof (*mem))1 << (bit); \ + asm_amo ("amoor", ".aq", mem, __mask) & __mask; }) + +# define catomic_exchange_and_add(mem, value) \ + atomic_exchange_and_add (mem, value) +# define catomic_max(mem, value) atomic_max (mem, value) + +#else /* __riscv_atomic */ +# error "ISAs that do not subsume the A extension are not supported" +#endif /* !__riscv_atomic */ + +#endif /* bits/atomic.h */