1 /* $NetBSD: kern_lock.c,v 1.150 2009/12/20 20:42:23 mrg Exp $ */ 2 3 /*- 4 * Copyright (c) 2002, 2006, 2007, 2008, 2009 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Jason R. Thorpe of the Numerical Aerospace Simulation Facility, 9 * NASA Ames Research Center, and by Andrew Doran. 10 * 11 * Redistribution and use in source and binary forms, with or without 12 * modification, are permitted provided that the following conditions 13 * are met: 14 * 1. Redistributions of source code must retain the above copyright 15 * notice, this list of conditions and the following disclaimer. 16 * 2. Redistributions in binary form must reproduce the above copyright 17 * notice, this list of conditions and the following disclaimer in the 18 * documentation and/or other materials provided with the distribution. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 21 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 22 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 23 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 24 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 25 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 26 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 27 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 28 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 29 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 30 * POSSIBILITY OF SUCH DAMAGE. 31 */ 32 33 #include <sys/cdefs.h> 34 __KERNEL_RCSID(0, "$NetBSD: kern_lock.c,v 1.150 2009/12/20 20:42:23 mrg Exp $"); 35 36 #include <sys/param.h> 37 #include <sys/proc.h> 38 #include <sys/lock.h> 39 #include <sys/systm.h> 40 #include <sys/kernel.h> 41 #include <sys/lockdebug.h> 42 #include <sys/cpu.h> 43 #include <sys/syslog.h> 44 #include <sys/atomic.h> 45 #include <sys/lwp.h> 46 47 #include <machine/stdarg.h> 48 #include <machine/lock.h> 49 50 #include <dev/lockstat.h> 51 52 #define RETURN_ADDRESS (uintptr_t)__builtin_return_address(0) 53 54 bool kernel_lock_dodebug; 55 56 __cpu_simple_lock_t kernel_lock[CACHE_LINE_SIZE / sizeof(__cpu_simple_lock_t)] 57 __aligned(CACHE_LINE_SIZE); 58 59 void 60 assert_sleepable(void) 61 { 62 const char *reason; 63 uint64_t pctr; 64 bool idle; 65 66 if (panicstr != NULL) { 67 return; 68 } 69 70 LOCKDEBUG_BARRIER(kernel_lock, 1); 71 72 /* 73 * Avoid disabling/re-enabling preemption here since this 74 * routine may be called in delicate situations. 75 */ 76 do { 77 pctr = lwp_pctr(); 78 idle = CURCPU_IDLE_P(); 79 } while (pctr != lwp_pctr()); 80 81 reason = NULL; 82 if (idle && !cold) { 83 reason = "idle"; 84 } 85 if (cpu_intr_p()) { 86 reason = "interrupt"; 87 } 88 if (cpu_softintr_p()) { 89 reason = "softint"; 90 } 91 92 if (reason) { 93 panic("%s: %s caller=%p", __func__, reason, 94 (void *)RETURN_ADDRESS); 95 } 96 } 97 98 /* 99 * Functions for manipulating the kernel_lock. We put them here 100 * so that they show up in profiles. 101 */ 102 103 #define _KERNEL_LOCK_ABORT(msg) \ 104 LOCKDEBUG_ABORT(kernel_lock, &_kernel_lock_ops, __func__, msg) 105 106 #ifdef LOCKDEBUG 107 #define _KERNEL_LOCK_ASSERT(cond) \ 108 do { \ 109 if (!(cond)) \ 110 _KERNEL_LOCK_ABORT("assertion failed: " #cond); \ 111 } while (/* CONSTCOND */ 0) 112 #else 113 #define _KERNEL_LOCK_ASSERT(cond) /* nothing */ 114 #endif 115 116 void _kernel_lock_dump(volatile void *); 117 118 lockops_t _kernel_lock_ops = { 119 "Kernel lock", 120 LOCKOPS_SPIN, 121 _kernel_lock_dump 122 }; 123 124 /* 125 * Initialize the kernel lock. 126 */ 127 void 128 kernel_lock_init(void) 129 { 130 131 CTASSERT(CACHE_LINE_SIZE >= sizeof(__cpu_simple_lock_t)); 132 __cpu_simple_lock_init(kernel_lock); 133 kernel_lock_dodebug = LOCKDEBUG_ALLOC(kernel_lock, &_kernel_lock_ops, 134 RETURN_ADDRESS); 135 } 136 137 /* 138 * Print debugging information about the kernel lock. 139 */ 140 void 141 _kernel_lock_dump(volatile void *junk) 142 { 143 struct cpu_info *ci = curcpu(); 144 145 (void)junk; 146 147 printf_nolog("curcpu holds : %18d wanted by: %#018lx\n", 148 ci->ci_biglock_count, (long)ci->ci_biglock_wanted); 149 } 150 151 /* 152 * Acquire 'nlocks' holds on the kernel lock. 153 */ 154 void 155 _kernel_lock(int nlocks) 156 { 157 struct cpu_info *ci; 158 LOCKSTAT_TIMER(spintime); 159 LOCKSTAT_FLAG(lsflag); 160 struct lwp *owant; 161 u_int spins; 162 int s; 163 struct lwp *l = curlwp; 164 165 _KERNEL_LOCK_ASSERT(nlocks > 0); 166 167 s = splvm(); 168 ci = curcpu(); 169 if (ci->ci_biglock_count != 0) { 170 _KERNEL_LOCK_ASSERT(__SIMPLELOCK_LOCKED_P(kernel_lock)); 171 ci->ci_biglock_count += nlocks; 172 l->l_blcnt += nlocks; 173 splx(s); 174 return; 175 } 176 177 _KERNEL_LOCK_ASSERT(l->l_blcnt == 0); 178 LOCKDEBUG_WANTLOCK(kernel_lock_dodebug, kernel_lock, RETURN_ADDRESS, 179 false, false); 180 181 if (__cpu_simple_lock_try(kernel_lock)) { 182 ci->ci_biglock_count = nlocks; 183 l->l_blcnt = nlocks; 184 LOCKDEBUG_LOCKED(kernel_lock_dodebug, kernel_lock, NULL, 185 RETURN_ADDRESS, 0); 186 splx(s); 187 return; 188 } 189 190 /* 191 * To remove the ordering constraint between adaptive mutexes 192 * and kernel_lock we must make it appear as if this thread is 193 * blocking. For non-interlocked mutex release, a store fence 194 * is required to ensure that the result of any mutex_exit() 195 * by the current LWP becomes visible on the bus before the set 196 * of ci->ci_biglock_wanted becomes visible. 197 */ 198 membar_producer(); 199 owant = ci->ci_biglock_wanted; 200 ci->ci_biglock_wanted = l; 201 202 /* 203 * Spin until we acquire the lock. Once we have it, record the 204 * time spent with lockstat. 205 */ 206 LOCKSTAT_ENTER(lsflag); 207 LOCKSTAT_START_TIMER(lsflag, spintime); 208 209 spins = 0; 210 do { 211 splx(s); 212 while (__SIMPLELOCK_LOCKED_P(kernel_lock)) { 213 if (SPINLOCK_SPINOUT(spins)) { 214 extern int start_init_exec; 215 if (!start_init_exec) 216 _KERNEL_LOCK_ABORT("spinout"); 217 } 218 SPINLOCK_BACKOFF_HOOK; 219 SPINLOCK_SPIN_HOOK; 220 } 221 s = splvm(); 222 } while (!__cpu_simple_lock_try(kernel_lock)); 223 224 ci->ci_biglock_count = nlocks; 225 l->l_blcnt = nlocks; 226 LOCKSTAT_STOP_TIMER(lsflag, spintime); 227 LOCKDEBUG_LOCKED(kernel_lock_dodebug, kernel_lock, NULL, 228 RETURN_ADDRESS, 0); 229 if (owant == NULL) { 230 LOCKSTAT_EVENT_RA(lsflag, kernel_lock, 231 LB_KERNEL_LOCK | LB_SPIN, 1, spintime, RETURN_ADDRESS); 232 } 233 LOCKSTAT_EXIT(lsflag); 234 splx(s); 235 236 /* 237 * Now that we have kernel_lock, reset ci_biglock_wanted. This 238 * store must be unbuffered (immediately visible on the bus) in 239 * order for non-interlocked mutex release to work correctly. 240 * It must be visible before a mutex_exit() can execute on this 241 * processor. 242 * 243 * Note: only where CAS is available in hardware will this be 244 * an unbuffered write, but non-interlocked release cannot be 245 * done on CPUs without CAS in hardware. 246 */ 247 (void)atomic_swap_ptr(&ci->ci_biglock_wanted, owant); 248 249 /* 250 * Issue a memory barrier as we have acquired a lock. This also 251 * prevents stores from a following mutex_exit() being reordered 252 * to occur before our store to ci_biglock_wanted above. 253 */ 254 membar_enter(); 255 } 256 257 /* 258 * Release 'nlocks' holds on the kernel lock. If 'nlocks' is zero, release 259 * all holds. 260 */ 261 void 262 _kernel_unlock(int nlocks, int *countp) 263 { 264 struct cpu_info *ci; 265 u_int olocks; 266 int s; 267 struct lwp *l = curlwp; 268 269 _KERNEL_LOCK_ASSERT(nlocks < 2); 270 271 olocks = l->l_blcnt; 272 273 if (olocks == 0) { 274 _KERNEL_LOCK_ASSERT(nlocks <= 0); 275 if (countp != NULL) 276 *countp = 0; 277 return; 278 } 279 280 _KERNEL_LOCK_ASSERT(__SIMPLELOCK_LOCKED_P(kernel_lock)); 281 282 if (nlocks == 0) 283 nlocks = olocks; 284 else if (nlocks == -1) { 285 nlocks = 1; 286 _KERNEL_LOCK_ASSERT(olocks == 1); 287 } 288 s = splvm(); 289 ci = curcpu(); 290 _KERNEL_LOCK_ASSERT(ci->ci_biglock_count >= l->l_blcnt); 291 if (ci->ci_biglock_count == nlocks) { 292 LOCKDEBUG_UNLOCKED(kernel_lock_dodebug, kernel_lock, 293 RETURN_ADDRESS, 0); 294 ci->ci_biglock_count = 0; 295 __cpu_simple_unlock(kernel_lock); 296 l->l_blcnt -= nlocks; 297 splx(s); 298 if (l->l_dopreempt) 299 kpreempt(0); 300 } else { 301 ci->ci_biglock_count -= nlocks; 302 l->l_blcnt -= nlocks; 303 splx(s); 304 } 305 306 if (countp != NULL) 307 *countp = olocks; 308 } 309