osfmk/kern/lock_group.c

/*
 * Copyright (c) 2000-2021 Apple Inc. All rights reserved.
 *
 * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
 *
 * This file contains Original Code and/or Modifications of Original Code
 * as defined in and that are subject to the Apple Public Source License
 * Version 2.0 (the 'License'). You may not use this file except in
 * compliance with the License. The rights granted to you under the License
 * may not be used to create, or enable the creation or redistribution of,
 * unlawful or unlicensed copies of an Apple operating system, or to
 * circumvent, violate, or enable the circumvention or violation of, any
 * terms of an Apple operating system software license agreement.
 *
 * Please obtain a copy of the License at
 * http://www.opensource.apple.com/apsl/ and read it before using this file.
 *
 * The Original Code and all software distributed under the License are
 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
 * Please see the License for the specific language governing rights and
 * limitations under the License.
 *
 * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
 */
/*
 * @OSF_COPYRIGHT@
 */
/*
 * Mach Operating System
 * Copyright (c) 1991,1990,1989,1988,1987 Carnegie Mellon University
 * All Rights Reserved.
 *
 * Permission to use, copy, modify and distribute this software and its
 * documentation is hereby granted, provided that both the copyright
 * notice and this permission notice appear in all copies of the
 * software, derivative works or modified versions, and any portions
 * thereof, and that both notices appear in supporting documentation.
 *
 * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
 * CONDITION.  CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND FOR
 * ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
 *
 * Carnegie Mellon requests users of this software to return to
 *
 *  Software Distribution Coordinator  or  [email protected]
 *  School of Computer Science
 *  Carnegie Mellon University
 *  Pittsburgh PA 15213-3890
 *
 * any improvements or extensions that they make and grant Carnegie Mellon
 * the rights to redistribute these changes.
 */

#define LOCK_PRIVATE 1

#include <mach_ldebug.h>
#include <debug.h>

#include <mach/mach_host_server.h>
#include <mach_debug/lockgroup_info.h>

#if __x86_64__
#include <i386/tsc.h>
#endif

#include <kern/compact_id.h>
#include <kern/kalloc.h>
#include <kern/lock_stat.h>
#include <kern/locks.h>

#include <os/atomic_private.h>

static KALLOC_TYPE_DEFINE(KT_LCK_GRP_ATTR, lck_grp_attr_t, KT_PRIV_ACCT);
static KALLOC_TYPE_DEFINE(KT_LCK_GRP, lck_grp_t, KT_PRIV_ACCT);
static KALLOC_TYPE_DEFINE(KT_LCK_ATTR, lck_attr_t, KT_PRIV_ACCT);

SECURITY_READ_ONLY_LATE(lck_attr_t) lck_attr_default;
static SECURITY_READ_ONLY_LATE(lck_grp_attr_t) lck_grp_attr_default;
static lck_grp_t lck_grp_compat_grp;
COMPACT_ID_TABLE_DEFINE(static, lck_grp_table);
struct lck_debug_state lck_debug_state;

#pragma mark lock group attributes

lck_grp_attr_t  *
lck_grp_attr_alloc_init(void)
{
	lck_grp_attr_t  *attr;

	attr = zalloc(KT_LCK_GRP_ATTR);
	lck_grp_attr_setdefault(attr);
	return attr;
}

void
lck_grp_attr_setdefault(lck_grp_attr_t *attr)
{
	attr->grp_attr_val = lck_grp_attr_default.grp_attr_val;
}

void
lck_grp_attr_setstat(lck_grp_attr_t *attr __unused)
{
	attr->grp_attr_val |= LCK_GRP_ATTR_STAT;
}


void
lck_grp_attr_free(lck_grp_attr_t *attr)
{
	zfree(KT_LCK_GRP_ATTR, attr);
}

#pragma mark lock groups

__startup_func
static void
lck_group_init(void)
{
	if (LcksOpts & enaLkStat) {
		lck_grp_attr_default.grp_attr_val |= LCK_GRP_ATTR_STAT;
	}
	if (LcksOpts & enaLkTimeStat) {
		lck_grp_attr_default.grp_attr_val |= LCK_GRP_ATTR_TIME_STAT;
	}
	if (LcksOpts & enaLkDeb) {
		lck_grp_attr_default.grp_attr_val |= LCK_GRP_ATTR_DEBUG;
	}

	if (LcksOpts & enaLkDeb) {
		lck_attr_default.lck_attr_val = LCK_ATTR_DEBUG;
	} else {
		lck_attr_default.lck_attr_val = LCK_ATTR_NONE;
	}

	/*
	 * This is a little gross, this allows us to use the table before
	 * compact_table_init() is called on it, but we have a chicken
	 * and egg problem otherwise.
	 *
	 * compact_table_init() really only inits the ticket lock
	 * with the proper lock group
	 */
	lck_grp_init(&lck_grp_compat_grp, "Compatibility APIs",
	    &lck_grp_attr_default);
	*compact_id_resolve(&lck_grp_table, 0) = LCK_GRP_NULL;
}
STARTUP(LOCKS, STARTUP_RANK_FIRST, lck_group_init);

__startup_func
void
lck_grp_startup_init(struct lck_grp_spec *sp)
{
	lck_grp_init_flags(sp->grp, sp->grp_name, sp->grp_flags |
	    lck_grp_attr_default.grp_attr_val);
}

bool
lck_grp_has_stats(lck_grp_t *grp)
{
	return grp->lck_grp_attr_id & LCK_GRP_ATTR_STAT;
}

lck_grp_t *
lck_grp_alloc_init(const char *grp_name, lck_grp_attr_t *attr)
{
	lck_grp_t *grp;

	if (attr == LCK_GRP_ATTR_NULL) {
		attr = &lck_grp_attr_default;
	}
	grp = zalloc(KT_LCK_GRP);
	lck_grp_init_flags(grp, grp_name,
	    attr->grp_attr_val | LCK_GRP_ATTR_ALLOCATED);
	return grp;
}

void
lck_grp_init(lck_grp_t *grp, const char *grp_name, lck_grp_attr_t *attr)
{
	if (attr == LCK_GRP_ATTR_NULL) {
		attr = &lck_grp_attr_default;
	}
	lck_grp_init_flags(grp, grp_name, attr->grp_attr_val);
}

lck_grp_t *
lck_grp_init_flags(lck_grp_t *grp, const char *grp_name, lck_grp_options_t flags)
{
	bzero(grp, sizeof(lck_grp_t));
	os_ref_init_raw(&grp->lck_grp_refcnt, NULL);
	(void)strlcpy(grp->lck_grp_name, grp_name, LCK_GRP_MAX_NAME);

#if CONFIG_DTRACE
	lck_grp_stats_t *stats = &grp->lck_grp_stats;

	if (flags & LCK_GRP_ATTR_STAT) {
		lck_grp_stat_enable(&stats->lgss_spin_held);
		lck_grp_stat_enable(&stats->lgss_spin_miss);

		lck_grp_stat_enable(&stats->lgss_ticket_held);
		lck_grp_stat_enable(&stats->lgss_ticket_miss);

		lck_grp_stat_enable(&stats->lgss_mtx_held);
		lck_grp_stat_enable(&stats->lgss_mtx_direct_wait);
		lck_grp_stat_enable(&stats->lgss_mtx_miss);
		lck_grp_stat_enable(&stats->lgss_mtx_wait);
	}
	if (flags & LCK_GRP_ATTR_TIME_STAT) {
		lck_grp_stat_enable(&stats->lgss_spin_spin);
		lck_grp_stat_enable(&stats->lgss_ticket_spin);
	}
#endif /* CONFIG_DTRACE */

	/* must be last as it publishes the group */
	if (startup_phase > STARTUP_SUB_LOCKS) {
		compact_id_table_lock(&lck_grp_table);
	}
	flags |= compact_id_get_locked(&lck_grp_table, LCK_GRP_ATTR_ID_MASK, grp);
	grp->lck_grp_attr_id = flags;
	if (startup_phase > STARTUP_SUB_LOCKS) {
		compact_id_table_unlock(&lck_grp_table);
	}

	return grp;
}

lck_grp_t *
lck_grp_resolve(uint32_t grp_attr_id)
{
	grp_attr_id &= LCK_GRP_ATTR_ID_MASK;
	return *compact_id_resolve(&lck_grp_table, grp_attr_id);
}

__abortlike
static void
__lck_grp_assert_id_panic(lck_grp_t *grp, uint32_t grp_attr_id)
{
	panic("lck_grp_t %p has ID %d, but %d was expected", grp,
	    grp->lck_grp_attr_id & LCK_GRP_ATTR_ID_MASK,
	    grp_attr_id & LCK_GRP_ATTR_ID_MASK);
}

__attribute__((always_inline))
void
lck_grp_assert_id(lck_grp_t *grp, uint32_t grp_attr_id)
{
	if ((grp->lck_grp_attr_id ^ grp_attr_id) & LCK_GRP_ATTR_ID_MASK) {
		__lck_grp_assert_id_panic(grp, grp_attr_id);
	}
}

static void
lck_grp_destroy(lck_grp_t *grp)
{
	compact_id_put(&lck_grp_table,
	    grp->lck_grp_attr_id & LCK_GRP_ATTR_ID_MASK);
	zfree(KT_LCK_GRP, grp);
}

void
lck_grp_free(lck_grp_t *grp)
{
	lck_grp_deallocate(grp, NULL);
}

void
lck_grp_reference(lck_grp_t *grp, uint32_t *cnt)
{
	if (cnt) {
		os_atomic_inc(cnt, relaxed);
	}
	if (grp->lck_grp_attr_id & LCK_GRP_ATTR_ALLOCATED) {
		os_ref_retain_raw(&grp->lck_grp_refcnt, NULL);
	}
}

void
lck_grp_deallocate(lck_grp_t *grp, uint32_t *cnt)
{
	if (cnt) {
		os_atomic_dec(cnt, relaxed);
	}
	if ((grp->lck_grp_attr_id & LCK_GRP_ATTR_ALLOCATED) &&
	    os_ref_release_raw(&grp->lck_grp_refcnt, 0) == 0) {
		lck_grp_destroy(grp);
	}
}

void
lck_grp_foreach(bool (^block)(lck_grp_t *))
{
	compact_id_for_each(&lck_grp_table, 64, (bool (^)(void *))block);
}

void
lck_grp_enable_feature(lck_debug_feature_t feat)
{
	uint32_t bit = 1u << feat;

	compact_id_table_lock(&lck_grp_table);
	if (lck_debug_state.lds_counts[feat]++ == 0) {
		os_atomic_or(&lck_debug_state.lds_value, bit, relaxed);
	}
	compact_id_table_unlock(&lck_grp_table);
}

void
lck_grp_disable_feature(lck_debug_feature_t feat)
{
	uint32_t bit = 1u << feat;
	long v;

	compact_id_table_lock(&lck_grp_table);
	v = --lck_debug_state.lds_counts[feat];
	if (v < 0) {
		panic("lck_debug_state: feature %d imbalance", feat);
	}
	if (v == 0) {
		os_atomic_andnot(&lck_debug_state.lds_value, bit, relaxed);
	}
	compact_id_table_unlock(&lck_grp_table);
}

kern_return_t
host_lockgroup_info(
	host_t                   host,
	lockgroup_info_array_t  *lockgroup_infop,
	mach_msg_type_number_t  *lockgroup_infoCntp)
{
	lockgroup_info_t *info;
	vm_offset_t       addr;
	vm_size_t         size, used;
	vm_size_t         vmsize, vmused;
	uint32_t          needed;
	__block uint32_t  count = 0;
	vm_map_copy_t     copy;
	kern_return_t     kr;

	if (host == HOST_NULL) {
		return KERN_INVALID_HOST;
	}

	/*
	 * Give about 10% of slop here, lock groups are mostly allocated
	 * during boot or kext loads, and is extremely unlikely to grow
	 * rapidly.
	 */
	needed  = os_atomic_load(&lck_grp_table.cidt_count, relaxed);
	needed += needed / 8;
	size    = needed * sizeof(lockgroup_info_t);
	vmsize = vm_map_round_page(size, VM_MAP_PAGE_MASK(ipc_kernel_map));
	kr     = kmem_alloc(ipc_kernel_map, &addr, vmsize,
	    KMA_DATA | KMA_ZERO, VM_KERN_MEMORY_IPC);
	if (kr != KERN_SUCCESS) {
		return kr;
	}

	info = (lockgroup_info_t *)addr;

	lck_grp_foreach(^bool (lck_grp_t *grp) {
		info[count].lock_spin_cnt = grp->lck_grp_spincnt;
		info[count].lock_rw_cnt   = grp->lck_grp_rwcnt;
		info[count].lock_mtx_cnt  = grp->lck_grp_mtxcnt;

#if CONFIG_DTRACE
		info[count].lock_spin_held_cnt = grp->lck_grp_stats.lgss_spin_held.lgs_count;
		info[count].lock_spin_miss_cnt = grp->lck_grp_stats.lgss_spin_miss.lgs_count;

		// Historically on x86, held was used for "direct wait" and util for "held"
		info[count].lock_mtx_util_cnt = grp->lck_grp_stats.lgss_mtx_held.lgs_count;
		info[count].lock_mtx_held_cnt = grp->lck_grp_stats.lgss_mtx_direct_wait.lgs_count;
		info[count].lock_mtx_miss_cnt = grp->lck_grp_stats.lgss_mtx_miss.lgs_count;
		info[count].lock_mtx_wait_cnt = grp->lck_grp_stats.lgss_mtx_wait.lgs_count;
#endif /* CONFIG_DTRACE */

		memcpy(info[count].lockgroup_name, grp->lck_grp_name, LOCKGROUP_MAX_NAME);

		return ++count >= needed ? false : true;
	});

	/*
	 * We might have found less groups than `needed`
	 * get rid of the excess now:
	 * - [0, used) is what we want to return
	 * - [0, size) is what we allocated
	 */
	used   = count * sizeof(lockgroup_info_t);
	vmused = vm_map_round_page(used, VM_MAP_PAGE_MASK(ipc_kernel_map));

	if (vmused < vmsize) {
		kmem_free(ipc_kernel_map, addr + vmused, vmsize - vmused);
	}

	kr = vm_map_unwire(ipc_kernel_map, addr, addr + vmused, FALSE);
	assert(kr == KERN_SUCCESS);

	kr = vm_map_copyin(ipc_kernel_map, addr, used, TRUE, &copy);
	assert(kr == KERN_SUCCESS);

	*lockgroup_infop = (lockgroup_info_t *)copy;
	*lockgroup_infoCntp = count;

	return KERN_SUCCESS;
}

#pragma mark lock attributes

__startup_func
void
lck_attr_startup_init(struct lck_attr_startup_spec *sp)
{
	lck_attr_t *attr = sp->lck_attr;
	lck_attr_setdefault(attr);
	attr->lck_attr_val |= sp->lck_attr_set_flags;
	attr->lck_attr_val &= ~sp->lck_attr_clear_flags;
}

lck_attr_t *
lck_attr_alloc_init(void)
{
	lck_attr_t      *attr;

	attr = zalloc(KT_LCK_ATTR);
	lck_attr_setdefault(attr);
	return attr;
}


void
lck_attr_setdefault(lck_attr_t *attr)
{
	attr->lck_attr_val = lck_attr_default.lck_attr_val;
}


void
lck_attr_setdebug(lck_attr_t *attr)
{
	os_atomic_or(&attr->lck_attr_val, LCK_ATTR_DEBUG, relaxed);
}

void
lck_attr_cleardebug(lck_attr_t *attr)
{
	os_atomic_andnot(&attr->lck_attr_val, LCK_ATTR_DEBUG, relaxed);
}

void
lck_attr_rw_shared_priority(lck_attr_t *attr)
{
	os_atomic_or(&attr->lck_attr_val, LCK_ATTR_RW_SHARED_PRIORITY, relaxed);
}


void
lck_attr_free(lck_attr_t *attr)
{
	zfree(KT_LCK_ATTR, attr);
}

#pragma mark lock stat
#if CONFIG_DTRACE

void
lck_grp_stat_enable(lck_grp_stat_t *stat)
{
	/* callers ensure this is properly synchronized */
	stat->lgs_enablings++;
}

void
lck_grp_stat_disable(lck_grp_stat_t *stat)
{
	stat->lgs_enablings--;
}

bool
lck_grp_stat_enabled(lck_grp_stat_t *stat)
{
	return stat->lgs_enablings != 0;
}


__attribute__((always_inline))
void
lck_grp_stat_inc(lck_grp_t *grp, lck_grp_stat_t *stat, bool always)
{
#pragma unused(grp)
	if (always || lck_grp_stat_enabled(stat)) {
		__unused uint64_t val = os_atomic_inc_orig(&stat->lgs_count, relaxed);
		if (__improbable(stat->lgs_limit && (val % (stat->lgs_limit)) == 0)) {
			lockprof_probe(grp, stat, val);
		}
	}
}

#if LOCK_STATS

static inline void
lck_grp_inc_time_stats(lck_grp_t *grp, lck_grp_stat_t *stat, uint64_t time)
{
	if (lck_grp_stat_enabled(stat)) {
		__unused uint64_t val = os_atomic_add_orig(&stat->lgs_count, time, relaxed);
		if (__improbable(stat->lgs_limit)) {
			while (__improbable(time > stat->lgs_limit)) {
				time -= stat->lgs_limit;
				lockprof_probe(grp, stat, val);
			}
			if (__improbable(((val % stat->lgs_limit) + time) > stat->lgs_limit)) {
				lockprof_probe(grp, stat, val);
			}
		}
	}
}

void
__lck_grp_spin_update_held(lck_grp_t *grp)
{
	if (grp) {
		lck_grp_stat_inc(grp, &grp->lck_grp_stats.lgss_spin_held, false);
	}
}

void
__lck_grp_spin_update_miss(lck_grp_t *grp)
{
	if (grp) {
		lck_grp_stat_inc(grp, &grp->lck_grp_stats.lgss_spin_miss, false);
	}
}

void
__lck_grp_spin_update_spin(lck_grp_t *grp, uint64_t time)
{
	if (grp) {
		lck_grp_stat_t *stat = &grp->lck_grp_stats.lgss_spin_spin;
		lck_grp_inc_time_stats(grp, stat, time);
	}
}

void
__lck_grp_ticket_update_held(lck_grp_t *grp)
{
	if (grp) {
		lck_grp_stat_inc(grp, &grp->lck_grp_stats.lgss_ticket_held, false);
	}
}

void
__lck_grp_ticket_update_miss(lck_grp_t *grp)
{
	if (grp) {
		lck_grp_stat_inc(grp, &grp->lck_grp_stats.lgss_ticket_miss, false);
	}
}

void
__lck_grp_ticket_update_spin(lck_grp_t *grp, uint64_t time)
{
	if (grp) {
		lck_grp_stat_t *stat = &grp->lck_grp_stats.lgss_ticket_spin;
		lck_grp_inc_time_stats(grp, stat, time);
	}
}

#endif /* LOCK_STATS */

void
lck_mtx_time_stat_record(
	enum lockstat_probe_id  pid,
	lck_mtx_t              *mtx,
	uint32_t                grp_attr_id,
	uint64_t                start)
{
	uint32_t id = lockstat_probemap[pid];

	if (__improbable(start && id)) {
		uint64_t delta = ml_get_timebase() - start;
		lck_grp_t *grp = lck_grp_resolve(grp_attr_id);

#if __x86_64__
		delta = tmrCvt(delta, tscFCvtt2n);
#endif
		dtrace_probe(id, (uintptr_t)mtx, delta, (uintptr_t)grp, 0, 0);
	}
}

#endif /* CONFIG_DTRACE */