1 /* $NetBSD: kern_lock.c,v 1.172 2020/12/22 01:57:29 ad Exp $ */ 2 3 /*- 4 * Copyright (c) 2002, 2006, 2007, 2008, 2009, 2020 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 1.172 2020/12/22 01:57:29 ad Exp $"); 35 36 #ifdef _KERNEL_OPT 37 #include "opt_lockdebug.h" 38 #endif 39 40 #include <sys/param.h> 41 #include <sys/proc.h> 42 #include <sys/lock.h> 43 #include <sys/systm.h> 44 #include <sys/kernel.h> 45 #include <sys/lockdebug.h> 46 #include <sys/cpu.h> 47 #include <sys/syslog.h> 48 #include <sys/atomic.h> 49 #include <sys/lwp.h> 50 #include <sys/pserialize.h> 51 52 #if defined(DIAGNOSTIC) && !defined(LOCKDEBUG) 53 #include <sys/ksyms.h> 54 #endif 55 56 #include <machine/lock.h> 57 58 #include <dev/lockstat.h> 59 60 #define RETURN_ADDRESS (uintptr_t)__builtin_return_address(0) 61 62 bool kernel_lock_dodebug; 63 64 __cpu_simple_lock_t kernel_lock[CACHE_LINE_SIZE / sizeof(__cpu_simple_lock_t)] 65 __cacheline_aligned; 66 67 void 68 assert_sleepable(void) 69 { 70 const char *reason; 71 uint64_t pctr; 72 bool idle; 73 74 if (panicstr != NULL) { 75 return; 76 } 77 78 LOCKDEBUG_BARRIER(kernel_lock, 1); 79 80 /* 81 * Avoid disabling/re-enabling preemption here since this 82 * routine may be called in delicate situations. 83 */ 84 do { 85 pctr = lwp_pctr(); 86 __insn_barrier(); 87 idle = CURCPU_IDLE_P(); 88 __insn_barrier(); 89 } while (pctr != lwp_pctr()); 90 91 reason = NULL; 92 if (idle && !cold && 93 kcpuset_isset(kcpuset_running, cpu_index(curcpu()))) { 94 reason = "idle"; 95 } 96 if (cpu_intr_p()) { 97 reason = "interrupt"; 98 } 99 if (cpu_softintr_p()) { 100 reason = "softint"; 101 } 102 if (!pserialize_not_in_read_section()) { 103 reason = "pserialize"; 104 } 105 106 if (reason) { 107 panic("%s: %s caller=%p", __func__, reason, 108 (void *)RETURN_ADDRESS); 109 } 110 } 111 112 /* 113 * Functions for manipulating the kernel_lock. We put them here 114 * so that they show up in profiles. 115 */ 116 117 #define _KERNEL_LOCK_ABORT(msg) \ 118 LOCKDEBUG_ABORT(__func__, __LINE__, kernel_lock, &_kernel_lock_ops, msg) 119 120 #ifdef LOCKDEBUG 121 #define _KERNEL_LOCK_ASSERT(cond) \ 122 do { \ 123 if (!(cond)) \ 124 _KERNEL_LOCK_ABORT("assertion failed: " #cond); \ 125 } while (/* CONSTCOND */ 0) 126 #else 127 #define _KERNEL_LOCK_ASSERT(cond) /* nothing */ 128 #endif 129 130 static void _kernel_lock_dump(const volatile void *, lockop_printer_t); 131 132 lockops_t _kernel_lock_ops = { 133 .lo_name = "Kernel lock", 134 .lo_type = LOCKOPS_SPIN, 135 .lo_dump = _kernel_lock_dump, 136 }; 137 138 /* 139 * Initialize the kernel lock. 140 */ 141 void 142 kernel_lock_init(void) 143 { 144 145 __cpu_simple_lock_init(kernel_lock); 146 kernel_lock_dodebug = LOCKDEBUG_ALLOC(kernel_lock, &_kernel_lock_ops, 147 RETURN_ADDRESS); 148 } 149 CTASSERT(CACHE_LINE_SIZE >= sizeof(__cpu_simple_lock_t)); 150 151 /* 152 * Print debugging information about the kernel lock. 153 */ 154 static void 155 _kernel_lock_dump(const volatile void *junk, lockop_printer_t pr) 156 { 157 struct cpu_info *ci = curcpu(); 158 159 (void)junk; 160 161 pr("curcpu holds : %18d wanted by: %#018lx\n", 162 ci->ci_biglock_count, (long)ci->ci_biglock_wanted); 163 } 164 165 /* 166 * Acquire 'nlocks' holds on the kernel lock. 167 * 168 * Although it may not look it, this is one of the most central, intricate 169 * routines in the kernel, and tons of code elsewhere depends on its exact 170 * behaviour. If you change something in here, expect it to bite you in the 171 * rear. 172 */ 173 void 174 _kernel_lock(int nlocks) 175 { 176 struct cpu_info *ci; 177 LOCKSTAT_TIMER(spintime); 178 LOCKSTAT_FLAG(lsflag); 179 struct lwp *owant; 180 #ifdef LOCKDEBUG 181 u_int spins = 0; 182 #endif 183 int s; 184 struct lwp *l = curlwp; 185 186 _KERNEL_LOCK_ASSERT(nlocks > 0); 187 188 s = splvm(); 189 ci = curcpu(); 190 if (ci->ci_biglock_count != 0) { 191 _KERNEL_LOCK_ASSERT(__SIMPLELOCK_LOCKED_P(kernel_lock)); 192 ci->ci_biglock_count += nlocks; 193 l->l_blcnt += nlocks; 194 splx(s); 195 return; 196 } 197 198 _KERNEL_LOCK_ASSERT(l->l_blcnt == 0); 199 LOCKDEBUG_WANTLOCK(kernel_lock_dodebug, kernel_lock, RETURN_ADDRESS, 200 0); 201 202 if (__predict_true(__cpu_simple_lock_try(kernel_lock))) { 203 ci->ci_biglock_count = nlocks; 204 l->l_blcnt = nlocks; 205 LOCKDEBUG_LOCKED(kernel_lock_dodebug, kernel_lock, NULL, 206 RETURN_ADDRESS, 0); 207 splx(s); 208 return; 209 } 210 211 /* 212 * To remove the ordering constraint between adaptive mutexes 213 * and kernel_lock we must make it appear as if this thread is 214 * blocking. For non-interlocked mutex release, a store fence 215 * is required to ensure that the result of any mutex_exit() 216 * by the current LWP becomes visible on the bus before the set 217 * of ci->ci_biglock_wanted becomes visible. 218 */ 219 membar_producer(); 220 owant = ci->ci_biglock_wanted; 221 ci->ci_biglock_wanted = l; 222 #if defined(DIAGNOSTIC) && !defined(LOCKDEBUG) 223 l->l_ld_wanted = __builtin_return_address(0); 224 #endif 225 226 /* 227 * Spin until we acquire the lock. Once we have it, record the 228 * time spent with lockstat. 229 */ 230 LOCKSTAT_ENTER(lsflag); 231 LOCKSTAT_START_TIMER(lsflag, spintime); 232 233 do { 234 splx(s); 235 while (__SIMPLELOCK_LOCKED_P(kernel_lock)) { 236 #ifdef LOCKDEBUG 237 if (SPINLOCK_SPINOUT(spins)) { 238 extern int start_init_exec; 239 if (start_init_exec) 240 _KERNEL_LOCK_ABORT("spinout"); 241 } 242 SPINLOCK_BACKOFF_HOOK; 243 SPINLOCK_SPIN_HOOK; 244 #endif 245 } 246 s = splvm(); 247 } while (!__cpu_simple_lock_try(kernel_lock)); 248 249 ci->ci_biglock_count = nlocks; 250 l->l_blcnt = nlocks; 251 LOCKSTAT_STOP_TIMER(lsflag, spintime); 252 LOCKDEBUG_LOCKED(kernel_lock_dodebug, kernel_lock, NULL, 253 RETURN_ADDRESS, 0); 254 if (owant == NULL) { 255 LOCKSTAT_EVENT_RA(lsflag, kernel_lock, 256 LB_KERNEL_LOCK | LB_SPIN, 1, spintime, RETURN_ADDRESS); 257 } 258 LOCKSTAT_EXIT(lsflag); 259 splx(s); 260 261 /* 262 * Now that we have kernel_lock, reset ci_biglock_wanted. This 263 * store must be unbuffered (immediately visible on the bus) in 264 * order for non-interlocked mutex release to work correctly. 265 * It must be visible before a mutex_exit() can execute on this 266 * processor. 267 * 268 * Note: only where CAS is available in hardware will this be 269 * an unbuffered write, but non-interlocked release cannot be 270 * done on CPUs without CAS in hardware. 271 */ 272 (void)atomic_swap_ptr(&ci->ci_biglock_wanted, owant); 273 274 /* 275 * Issue a memory barrier as we have acquired a lock. This also 276 * prevents stores from a following mutex_exit() being reordered 277 * to occur before our store to ci_biglock_wanted above. 278 */ 279 #ifndef __HAVE_ATOMIC_AS_MEMBAR 280 membar_enter(); 281 #endif 282 } 283 284 /* 285 * Release 'nlocks' holds on the kernel lock. If 'nlocks' is zero, release 286 * all holds. 287 */ 288 void 289 _kernel_unlock(int nlocks, int *countp) 290 { 291 struct cpu_info *ci; 292 u_int olocks; 293 int s; 294 struct lwp *l = curlwp; 295 296 _KERNEL_LOCK_ASSERT(nlocks < 2); 297 298 olocks = l->l_blcnt; 299 300 if (olocks == 0) { 301 _KERNEL_LOCK_ASSERT(nlocks <= 0); 302 if (countp != NULL) 303 *countp = 0; 304 return; 305 } 306 307 _KERNEL_LOCK_ASSERT(__SIMPLELOCK_LOCKED_P(kernel_lock)); 308 309 if (nlocks == 0) 310 nlocks = olocks; 311 else if (nlocks == -1) { 312 nlocks = 1; 313 _KERNEL_LOCK_ASSERT(olocks == 1); 314 } 315 s = splvm(); 316 ci = curcpu(); 317 _KERNEL_LOCK_ASSERT(ci->ci_biglock_count >= l->l_blcnt); 318 if (ci->ci_biglock_count == nlocks) { 319 LOCKDEBUG_UNLOCKED(kernel_lock_dodebug, kernel_lock, 320 RETURN_ADDRESS, 0); 321 ci->ci_biglock_count = 0; 322 __cpu_simple_unlock(kernel_lock); 323 l->l_blcnt -= nlocks; 324 splx(s); 325 if (l->l_dopreempt) 326 kpreempt(0); 327 } else { 328 ci->ci_biglock_count -= nlocks; 329 l->l_blcnt -= nlocks; 330 splx(s); 331 } 332 333 if (countp != NULL) 334 *countp = olocks; 335 } 336 337 bool 338 _kernel_locked_p(void) 339 { 340 return __SIMPLELOCK_LOCKED_P(kernel_lock); 341 } 342