/*      $NetBSD: kern_lock.c,v 1.188 2024/01/14 11:46:05 andvar Exp $   */

/*-
* Copyright (c) 2002, 2006, 2007, 2008, 2009, 2020, 2023
*     The NetBSD Foundation, Inc.
* All rights reserved.
*
* This code is derived from software contributed to The NetBSD Foundation
* by Jason R. Thorpe of the Numerical Aerospace Simulation Facility,
* NASA Ames Research Center, and by Andrew Doran.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
*    notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
*    notice, this list of conditions and the following disclaimer in the
*    documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
* ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
* TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
* PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
* BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
* POSSIBILITY OF SUCH DAMAGE.
*/

#include <sys/cdefs.h>
__KERNEL_RCSID(0, "$NetBSD: kern_lock.c,v 1.188 2024/01/14 11:46:05 andvar Exp $");

#ifdef _KERNEL_OPT
#include "opt_lockdebug.h"
#endif

#include <sys/param.h>
#include <sys/proc.h>
#include <sys/lock.h>
#include <sys/systm.h>
#include <sys/kernel.h>
#include <sys/lockdebug.h>
#include <sys/cpu.h>
#include <sys/syslog.h>
#include <sys/atomic.h>
#include <sys/lwp.h>
#include <sys/pserialize.h>

#if defined(DIAGNOSTIC) && !defined(LOCKDEBUG)
#include <sys/ksyms.h>
#endif

#include <machine/lock.h>

#include <dev/lockstat.h>

#define RETURN_ADDRESS  (uintptr_t)__builtin_return_address(0)

bool    kernel_lock_dodebug;

__cpu_simple_lock_t kernel_lock[CACHE_LINE_SIZE / sizeof(__cpu_simple_lock_t)]
   __cacheline_aligned;

void
assert_sleepable(void)
{
       const char *reason;
       long pctr;
       bool idle;

       if (__predict_false(panicstr != NULL)) {
               return;
       }

       LOCKDEBUG_BARRIER(kernel_lock, 1);

       /*
        * Avoid disabling/re-enabling preemption here since this
        * routine may be called in delicate situations.
        */
       do {
               pctr = lwp_pctr();
               idle = CURCPU_IDLE_P();
       } while (__predict_false(pctr != lwp_pctr()));

       reason = NULL;
       if (__predict_false(idle) && !cold) {
               reason = "idle";
               goto panic;
       }
       if (__predict_false(cpu_intr_p())) {
               reason = "interrupt";
               goto panic;
       }
       if (__predict_false(cpu_softintr_p())) {
               reason = "softint";
               goto panic;
       }
       if (__predict_false(!pserialize_not_in_read_section())) {
               reason = "pserialize";
               goto panic;
       }
       return;

panic:  panic("%s: %s caller=%p", __func__, reason, (void *)RETURN_ADDRESS);
}

/*
* Functions for manipulating the kernel_lock.  We put them here
* so that they show up in profiles.
*/

#define _KERNEL_LOCK_ABORT(msg)                                         \
   LOCKDEBUG_ABORT(__func__, __LINE__, kernel_lock, &_kernel_lock_ops, msg)

#ifdef LOCKDEBUG
#define _KERNEL_LOCK_ASSERT(cond)                                       \
do {                                                                    \
       if (!(cond))                                                    \
               _KERNEL_LOCK_ABORT("assertion failed: " #cond);         \
} while (/* CONSTCOND */ 0)
#else
#define _KERNEL_LOCK_ASSERT(cond)       /* nothing */
#endif

static void     _kernel_lock_dump(const volatile void *, lockop_printer_t);

lockops_t _kernel_lock_ops = {
       .lo_name = "Kernel lock",
       .lo_type = LOCKOPS_SPIN,
       .lo_dump = _kernel_lock_dump,
};

#ifdef LOCKDEBUG

#ifdef DDB
#include <ddb/ddb.h>
#endif

static void
kernel_lock_trace_ipi(void *cookie)
{

       printf("%s[%d %s]: hogging kernel lock\n", cpu_name(curcpu()),
           curlwp->l_lid,
           curlwp->l_name ? curlwp->l_name : curproc->p_comm);
#ifdef DDB
       db_stacktrace();
#endif
}

#endif

/*
* Initialize the kernel lock.
*/
void
kernel_lock_init(void)
{

       __cpu_simple_lock_init(kernel_lock);
       kernel_lock_dodebug = LOCKDEBUG_ALLOC(kernel_lock, &_kernel_lock_ops,
           RETURN_ADDRESS);
}
CTASSERT(CACHE_LINE_SIZE >= sizeof(__cpu_simple_lock_t));

/*
* Print debugging information about the kernel lock.
*/
static void
_kernel_lock_dump(const volatile void *junk, lockop_printer_t pr)
{
       struct cpu_info *ci = curcpu();

       (void)junk;

       pr("curcpu holds : %18d wanted by: %#018lx\n",
           ci->ci_biglock_count, (long)ci->ci_biglock_wanted);
}

/*
* Acquire 'nlocks' holds on the kernel lock.
*
* Although it may not look it, this is one of the most central, intricate
* routines in the kernel, and tons of code elsewhere depends on its exact
* behaviour.  If you change something in here, expect it to bite you in the
* rear.
*/
void
_kernel_lock(int nlocks)
{
       struct cpu_info *ci;
       LOCKSTAT_TIMER(spintime);
       LOCKSTAT_FLAG(lsflag);
       struct lwp *owant;
#ifdef LOCKDEBUG
       static struct cpu_info *kernel_lock_holder;
       u_int spins = 0;
       u_int starttime = getticks();
#endif
       int s;
       struct lwp *l = curlwp;

       _KERNEL_LOCK_ASSERT(nlocks > 0);

       s = splvm();
       ci = curcpu();
       if (ci->ci_biglock_count != 0) {
               _KERNEL_LOCK_ASSERT(__SIMPLELOCK_LOCKED_P(kernel_lock));
               ci->ci_biglock_count += nlocks;
               l->l_blcnt += nlocks;
               splx(s);
               return;
       }

       _KERNEL_LOCK_ASSERT(l->l_blcnt == 0);
       LOCKDEBUG_WANTLOCK(kernel_lock_dodebug, kernel_lock, RETURN_ADDRESS,
           0);

       if (__predict_true(__cpu_simple_lock_try(kernel_lock))) {
#ifdef LOCKDEBUG
               kernel_lock_holder = curcpu();
#endif
               ci->ci_biglock_count = nlocks;
               l->l_blcnt = nlocks;
               LOCKDEBUG_LOCKED(kernel_lock_dodebug, kernel_lock, NULL,
                   RETURN_ADDRESS, 0);
               splx(s);
               return;
       }

       /*
        * To remove the ordering constraint between adaptive mutexes
        * and kernel_lock we must make it appear as if this thread is
        * blocking.  For non-interlocked mutex release, a store fence
        * is required to ensure that the result of any mutex_exit()
        * by the current LWP becomes visible on the bus before the set
        * of ci->ci_biglock_wanted becomes visible.
        *
        * This membar_producer matches the membar_consumer in
        * mutex_vector_enter.
        *
        * That way, if l has just released a mutex, mutex_vector_enter
        * can't see this store ci->ci_biglock_wanted := l until it
        * will also see the mutex_exit store mtx->mtx_owner := 0 which
        * clears the has-waiters bit.
        */
       membar_producer();
       owant = ci->ci_biglock_wanted;
       atomic_store_relaxed(&ci->ci_biglock_wanted, l);
#if defined(DIAGNOSTIC) && !defined(LOCKDEBUG)
       l->l_ld_wanted = __builtin_return_address(0);
#endif

       /*
        * Spin until we acquire the lock.  Once we have it, record the
        * time spent with lockstat.
        */
       LOCKSTAT_ENTER(lsflag);
       LOCKSTAT_START_TIMER(lsflag, spintime);

       do {
               splx(s);
               while (__SIMPLELOCK_LOCKED_P(kernel_lock)) {
#ifdef LOCKDEBUG
                       if (SPINLOCK_SPINOUT(spins) && start_init_exec &&
                           (getticks() - starttime) > 10*hz) {
                               ipi_msg_t msg = {
                                       .func = kernel_lock_trace_ipi,
                               };
                               kpreempt_disable();
                               ipi_unicast(&msg, kernel_lock_holder);
                               ipi_wait(&msg);
                               kpreempt_enable();
                               _KERNEL_LOCK_ABORT("spinout");
                       }
#endif
                       SPINLOCK_BACKOFF_HOOK;
                       SPINLOCK_SPIN_HOOK;
               }
               s = splvm();
       } while (!__cpu_simple_lock_try(kernel_lock));

       ci->ci_biglock_count = nlocks;
       l->l_blcnt = nlocks;
       LOCKSTAT_STOP_TIMER(lsflag, spintime);
       LOCKDEBUG_LOCKED(kernel_lock_dodebug, kernel_lock, NULL,
           RETURN_ADDRESS, 0);
       if (owant == NULL) {
               LOCKSTAT_EVENT_RA(lsflag, kernel_lock,
                   LB_KERNEL_LOCK | LB_SPIN, 1, spintime, RETURN_ADDRESS);
       }
       LOCKSTAT_EXIT(lsflag);
       splx(s);

       /*
        * Now that we have kernel_lock, reset ci_biglock_wanted.  This
        * store must be visible on other CPUs before a mutex_exit() on
        * this CPU can test the has-waiters bit.
        *
        * This membar_enter matches the membar_enter in
        * mutex_vector_enter.  (Yes, not membar_exit -- the legacy
        * naming is confusing, but store-before-load usually pairs
        * with store-before-load, in the extremely rare cases where it
        * is used at all.)
        *
        * That way, mutex_vector_enter can't see this store
        * ci->ci_biglock_wanted := owant until it has set the
        * has-waiters bit.
        */
       (void)atomic_swap_ptr(&ci->ci_biglock_wanted, owant);
#ifndef __HAVE_ATOMIC_AS_MEMBAR
       membar_enter();
#endif

#ifdef LOCKDEBUG
       kernel_lock_holder = curcpu();
#endif
}

/*
* Release 'nlocks' holds on the kernel lock.  If 'nlocks' is zero, release
* all holds.
*/
void
_kernel_unlock(int nlocks, int *countp)
{
       struct cpu_info *ci;
       u_int olocks;
       int s;
       struct lwp *l = curlwp;

       _KERNEL_LOCK_ASSERT(nlocks < 2);

       olocks = l->l_blcnt;

       if (olocks == 0) {
               _KERNEL_LOCK_ASSERT(nlocks <= 0);
               if (countp != NULL)
                       *countp = 0;
               return;
       }

       _KERNEL_LOCK_ASSERT(__SIMPLELOCK_LOCKED_P(kernel_lock));

       if (nlocks == 0)
               nlocks = olocks;
       else if (nlocks == -1) {
               nlocks = 1;
               _KERNEL_LOCK_ASSERT(olocks == 1);
       }
       s = splvm();
       ci = curcpu();
       _KERNEL_LOCK_ASSERT(ci->ci_biglock_count >= l->l_blcnt);
       if (ci->ci_biglock_count == nlocks) {
               LOCKDEBUG_UNLOCKED(kernel_lock_dodebug, kernel_lock,
                   RETURN_ADDRESS, 0);
               ci->ci_biglock_count = 0;
               __cpu_simple_unlock(kernel_lock);
               l->l_blcnt -= nlocks;
               splx(s);
               if (l->l_dopreempt)
                       kpreempt(0);
       } else {
               ci->ci_biglock_count -= nlocks;
               l->l_blcnt -= nlocks;
               splx(s);
       }

       if (countp != NULL)
               *countp = olocks;
}

bool
_kernel_locked_p(void)
{
       return __SIMPLELOCK_LOCKED_P(kernel_lock);
}