/* $OpenBSD: kern_smr.c,v 1.9 2020/12/25 12:49:31 visa Exp $ */ /* * Copyright (c) 2019-2020 Visa Hankala * * Permission to use, copy, modify, and distribute this software for any * purpose with or without fee is hereby granted, provided that the above * copyright notice and this permission notice appear in all copies. * * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. */ #include #include #include #include #include #include #include #include #include #include #include #define SMR_PAUSE 100 /* pause between rounds in msec */ void smr_dispatch(struct schedstate_percpu *); void smr_grace_wait(void); void smr_thread(void *); void smr_wakeup(void *); struct mutex smr_lock = MUTEX_INITIALIZER(IPL_HIGH); struct smr_entry_list smr_deferred; struct timeout smr_wakeup_tmo; unsigned int smr_expedite; unsigned int smr_ndeferred; unsigned char smr_grace_period; #ifdef WITNESS static const char smr_lock_name[] = "smr"; struct lock_object smr_lock_obj = { .lo_name = smr_lock_name, .lo_flags = LO_WITNESS | LO_INITIALIZED | LO_SLEEPABLE | (LO_CLASS_RWLOCK << LO_CLASSSHIFT) }; struct lock_type smr_lock_type = { .lt_name = smr_lock_name }; #endif static inline int smr_cpu_is_idle(struct cpu_info *ci) { return ci->ci_curproc == ci->ci_schedstate.spc_idleproc; } void smr_startup(void) { SIMPLEQ_INIT(&smr_deferred); WITNESS_INIT(&smr_lock_obj, &smr_lock_type); timeout_set(&smr_wakeup_tmo, smr_wakeup, NULL); } void smr_startup_thread(void) { if (kthread_create(smr_thread, NULL, NULL, "smr") != 0) panic("could not create smr thread"); } struct timeval smr_logintvl = { 300, 0 }; void smr_thread(void *arg) { struct timeval elapsed, end, loglast, start; struct smr_entry_list deferred; struct smr_entry *smr; KERNEL_ASSERT_LOCKED(); KERNEL_UNLOCK(); memset(&loglast, 0, sizeof(loglast)); SIMPLEQ_INIT(&deferred); for (;;) { mtx_enter(&smr_lock); if (smr_ndeferred == 0) { while (smr_ndeferred == 0) msleep_nsec(&smr_ndeferred, &smr_lock, PVM, "bored", INFSLP); } else { if (smr_expedite == 0) msleep_nsec(&smr_ndeferred, &smr_lock, PVM, "pause", MSEC_TO_NSEC(SMR_PAUSE)); } SIMPLEQ_CONCAT(&deferred, &smr_deferred); smr_ndeferred = 0; smr_expedite = 0; mtx_leave(&smr_lock); getmicrouptime(&start); smr_grace_wait(); WITNESS_CHECKORDER(&smr_lock_obj, LOP_NEWORDER, NULL); WITNESS_LOCK(&smr_lock_obj, 0); while ((smr = SIMPLEQ_FIRST(&deferred)) != NULL) { SIMPLEQ_REMOVE_HEAD(&deferred, smr_list); smr->smr_func(smr->smr_arg); } WITNESS_UNLOCK(&smr_lock_obj, 0); getmicrouptime(&end); timersub(&end, &start, &elapsed); if (elapsed.tv_sec >= 5 && ratecheck(&loglast, &smr_logintvl)) printf("smr: dispatch took %ld seconds\n", (long)elapsed.tv_sec); } } /* * Announce next grace period and wait until all CPUs have entered it * by crossing quiescent state. */ void smr_grace_wait(void) { #ifdef MULTIPROCESSOR CPU_INFO_ITERATOR cii; struct cpu_info *ci; unsigned char smrgp; smrgp = READ_ONCE(smr_grace_period) + 1; WRITE_ONCE(smr_grace_period, smrgp); curcpu()->ci_schedstate.spc_smrgp = smrgp; CPU_INFO_FOREACH(cii, ci) { if (READ_ONCE(ci->ci_schedstate.spc_smrgp) == smrgp) continue; sched_peg_curproc(ci); KASSERT(ci->ci_schedstate.spc_smrgp == smrgp); } atomic_clearbits_int(&curproc->p_flag, P_CPUPEG); #endif /* MULTIPROCESSOR */ } void smr_wakeup(void *arg) { wakeup(&smr_ndeferred); } void smr_read_enter(void) { #ifdef DIAGNOSTIC struct schedstate_percpu *spc = &curcpu()->ci_schedstate; spc->spc_smrdepth++; #endif } void smr_read_leave(void) { #ifdef DIAGNOSTIC struct schedstate_percpu *spc = &curcpu()->ci_schedstate; KASSERT(spc->spc_smrdepth > 0); spc->spc_smrdepth--; #endif } /* * Move SMR entries from the local queue to the system-wide queue. */ void smr_dispatch(struct schedstate_percpu *spc) { int expedite = 0, wake = 0; mtx_enter(&smr_lock); if (smr_ndeferred == 0) wake = 1; SIMPLEQ_CONCAT(&smr_deferred, &spc->spc_deferred); smr_ndeferred += spc->spc_ndeferred; spc->spc_ndeferred = 0; smr_expedite |= spc->spc_smrexpedite; spc->spc_smrexpedite = 0; expedite = smr_expedite; mtx_leave(&smr_lock); if (expedite) wakeup(&smr_ndeferred); else if (wake) timeout_add_msec(&smr_wakeup_tmo, SMR_PAUSE); } /* * Signal that the current CPU is in quiescent state. */ void smr_idle(void) { struct schedstate_percpu *spc = &curcpu()->ci_schedstate; unsigned char smrgp; SMR_ASSERT_NONCRITICAL(); if (spc->spc_ndeferred > 0) smr_dispatch(spc); /* * Update this CPU's view of the system's grace period. * The update must become visible after any preceding reads * of SMR-protected data. */ smrgp = READ_ONCE(smr_grace_period); if (__predict_false(spc->spc_smrgp != smrgp)) { membar_exit(); WRITE_ONCE(spc->spc_smrgp, smrgp); } } void smr_call_impl(struct smr_entry *smr, void (*func)(void *), void *arg, int expedite) { struct cpu_info *ci = curcpu(); struct schedstate_percpu *spc = &ci->ci_schedstate; int s; KASSERT(smr->smr_func == NULL); smr->smr_func = func; smr->smr_arg = arg; s = splhigh(); SIMPLEQ_INSERT_TAIL(&spc->spc_deferred, smr, smr_list); spc->spc_ndeferred++; spc->spc_smrexpedite |= expedite; splx(s); /* * If this call was made from an interrupt context that * preempted idle state, dispatch the local queue to the shared * queue immediately. * The entries would linger in the local queue long if the CPU * went to sleep without calling smr_idle(). */ if (smr_cpu_is_idle(ci)) smr_dispatch(spc); } void smr_barrier_func(void *arg) { struct cond *c = arg; cond_signal(c); } void smr_barrier_impl(int expedite) { struct cond c = COND_INITIALIZER(); struct smr_entry smr; if (panicstr != NULL || db_active) return; WITNESS_CHECKORDER(&smr_lock_obj, LOP_NEWORDER, NULL); smr_init(&smr); smr_call_impl(&smr, smr_barrier_func, &c, expedite); cond_wait(&c, "smrbar"); }