[apple/xnu.git] / bsd / dev / dtrace / lockstat.c

/*
 * CDDL HEADER START
 *
 * The contents of this file are subject to the terms of the
 * Common Development and Distribution License (the "License").
 * You may not use this file except in compliance with the License.
 *
 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
 * or http://www.opensolaris.org/os/licensing.
 * See the License for the specific language governing permissions
 * and limitations under the License.
 *
 * When distributing Covered Code, include this CDDL HEADER in each
 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
 * If applicable, add the following below this CDDL HEADER, with the
 * fields enclosed by brackets "[]" replaced with your own identifying
 * information: Portions Copyright [yyyy] [name of copyright owner]
 *
 * CDDL HEADER END
 */
/*
 * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
 * Use is subject to license terms.
 */

/* #pragma ident	"@(#)lockstat.c	1.12	08/01/16 SMI" */


#ifdef KERNEL
#ifndef _KERNEL
#define _KERNEL /* Solaris vs. Darwin */
#endif
#endif

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/errno.h>
#include <sys/stat.h>
#include <sys/ioctl.h>
#include <sys/conf.h>
#include <sys/fcntl.h>
#include <miscfs/devfs/devfs.h>

#include <sys/dtrace.h>
#include <sys/dtrace_impl.h>

#include <sys/dtrace_glue.h>

#include <sys/lockstat.h>

#include <kern/processor.h>

#define membar_producer dtrace_membar_producer

/*
 * Hot patch values, x86
 */
#if defined(__x86_64__)
#define	NOP	0x90
#define	RET	0xc3
#define LOCKSTAT_AFRAMES 1
#else
#error "not ported to this architecture"
#endif

typedef struct lockstat_probe {
	const char	*lsp_func;
	const char	*lsp_name;
	int		lsp_probe;
	dtrace_id_t	lsp_id;
} lockstat_probe_t;

lockstat_probe_t lockstat_probes[] =
{
#if defined(__x86_64__)
	/* Only provide implemented probes for each architecture  */
	{ LS_LCK_MTX_LOCK,	LSA_ACQUIRE,	LS_LCK_MTX_LOCK_ACQUIRE, DTRACE_IDNONE },
	{ LS_LCK_MTX_LOCK,	LSA_SPIN,	LS_LCK_MTX_LOCK_SPIN, DTRACE_IDNONE },
	{ LS_LCK_MTX_LOCK,	LSA_BLOCK,	LS_LCK_MTX_LOCK_BLOCK, DTRACE_IDNONE },	
	{ LS_LCK_MTX_TRY_LOCK,	LSA_ACQUIRE,	LS_LCK_MTX_TRY_LOCK_ACQUIRE, DTRACE_IDNONE },
	{ LS_LCK_MTX_TRY_SPIN_LOCK, LSA_ACQUIRE, LS_LCK_MTX_TRY_SPIN_LOCK_ACQUIRE, DTRACE_IDNONE },
	{ LS_LCK_MTX_UNLOCK,	LSA_RELEASE,	LS_LCK_MTX_UNLOCK_RELEASE, DTRACE_IDNONE },
	{ LS_LCK_MTX_EXT_LOCK,	LSA_ACQUIRE,	LS_LCK_MTX_EXT_LOCK_ACQUIRE, DTRACE_IDNONE },
	{ LS_LCK_MTX_EXT_LOCK,	LSA_SPIN,	LS_LCK_MTX_EXT_LOCK_SPIN, DTRACE_IDNONE },
	{ LS_LCK_MTX_EXT_LOCK,	LSA_BLOCK,	LS_LCK_MTX_EXT_LOCK_BLOCK, DTRACE_IDNONE },
//	{ LS_LCK_MTX_EXT_TRY_LOCK, LSA_ACQUIRE,	LS_LCK_MTX_TRY_EXT_LOCK_ACQUIRE, DTRACE_IDNONE },	
	{ LS_LCK_MTX_EXT_UNLOCK,   LSA_RELEASE,	LS_LCK_MTX_EXT_UNLOCK_RELEASE, DTRACE_IDNONE },
	{ LS_LCK_MTX_LOCK_SPIN_LOCK,	LSA_ACQUIRE,	LS_LCK_MTX_LOCK_SPIN_ACQUIRE, DTRACE_IDNONE },
	{ LS_LCK_RW_LOCK_SHARED,	LSR_ACQUIRE,	LS_LCK_RW_LOCK_SHARED_ACQUIRE, DTRACE_IDNONE },
	{ LS_LCK_RW_LOCK_SHARED,	LSR_BLOCK,	LS_LCK_RW_LOCK_SHARED_BLOCK, DTRACE_IDNONE },
	{ LS_LCK_RW_LOCK_SHARED,	LSR_SPIN,	LS_LCK_RW_LOCK_SHARED_SPIN, DTRACE_IDNONE },
	{ LS_LCK_RW_LOCK_EXCL,		LSR_ACQUIRE,	LS_LCK_RW_LOCK_EXCL_ACQUIRE, DTRACE_IDNONE },
	{ LS_LCK_RW_LOCK_EXCL,		LSR_BLOCK,	LS_LCK_RW_LOCK_EXCL_BLOCK, DTRACE_IDNONE },
	{ LS_LCK_RW_LOCK_EXCL,		LSR_SPIN,	LS_LCK_RW_LOCK_EXCL_SPIN, DTRACE_IDNONE },
	{ LS_LCK_RW_DONE,		LSR_RELEASE,	LS_LCK_RW_DONE_RELEASE, DTRACE_IDNONE },
	{ LS_LCK_RW_TRY_LOCK_SHARED,	LSR_ACQUIRE,	LS_LCK_RW_TRY_LOCK_SHARED_ACQUIRE, DTRACE_IDNONE },
	{ LS_LCK_RW_TRY_LOCK_EXCL,	LSR_ACQUIRE,	LS_LCK_RW_TRY_LOCK_EXCL_ACQUIRE, DTRACE_IDNONE },
	{ LS_LCK_RW_LOCK_SHARED_TO_EXCL, LSR_UPGRADE,	LS_LCK_RW_LOCK_SHARED_TO_EXCL_UPGRADE, DTRACE_IDNONE },
	{ LS_LCK_RW_LOCK_SHARED_TO_EXCL,	LSR_SPIN,	LS_LCK_RW_LOCK_SHARED_TO_EXCL_SPIN, DTRACE_IDNONE },
	{ LS_LCK_RW_LOCK_SHARED_TO_EXCL,	LSR_BLOCK,	LS_LCK_RW_LOCK_SHARED_TO_EXCL_BLOCK, DTRACE_IDNONE },	
	{ LS_LCK_RW_LOCK_EXCL_TO_SHARED,	LSR_DOWNGRADE,	LS_LCK_RW_LOCK_EXCL_TO_SHARED_DOWNGRADE, DTRACE_IDNONE },
#endif
#ifdef	LATER
	/* Interlock and spinlock measurements would be nice, but later */
	{ LS_LCK_SPIN_LOCK,	LSS_ACQUIRE,	LS_LCK_SPIN_LOCK_ACQUIRE, DTRACE_IDNONE },
	{ LS_LCK_SPIN_LOCK,	LSS_SPIN,	LS_LCK_SPIN_LOCK_SPIN, DTRACE_IDNONE },
	{ LS_LCK_SPIN_UNLOCK,	LSS_RELEASE,	LS_LCK_SPIN_UNLOCK_RELEASE, DTRACE_IDNONE },

	{ LS_LCK_RW_LOCK_EXCL_TO_SHARED,	LSA_ILK_SPIN,	LS_LCK_RW_LOCK_EXCL_TO_SHARED_ILK_SPIN, DTRACE_IDNONE },
	{ LS_LCK_MTX_LOCK,	LSA_ILK_SPIN,	LS_LCK_MTX_LOCK_ILK_SPIN, DTRACE_IDNONE },
	{ LS_LCK_MTX_EXT_LOCK,	LSA_ILK_SPIN,	LS_LCK_MTX_EXT_LOCK_ILK_SPIN, DTRACE_IDNONE },
	{ LS_LCK_RW_TRY_LOCK_EXCL,	LSA_ILK_SPIN,	LS_LCK_RW_TRY_LOCK_EXCL_ILK_SPIN, DTRACE_IDNONE },
	{ LS_LCK_RW_TRY_LOCK_SHARED,	LSA_SPIN,	LS_LCK_RW_TRY_LOCK_SHARED_SPIN, DTRACE_IDNONE },
#endif

	{ NULL, NULL, 0, 0 }
};

dtrace_id_t lockstat_probemap[LS_NPROBES];

#if CONFIG_DTRACE
#if defined(__x86_64__)
extern void lck_mtx_lock_lockstat_patch_point(void);
extern void lck_mtx_try_lock_lockstat_patch_point(void);
extern void lck_mtx_try_lock_spin_lockstat_patch_point(void);
extern void lck_mtx_unlock_lockstat_patch_point(void);
extern void lck_mtx_lock_ext_lockstat_patch_point(void);
extern void lck_mtx_ext_unlock_lockstat_patch_point(void);
extern void lck_rw_lock_shared_lockstat_patch_point(void);
extern void lck_rw_lock_exclusive_lockstat_patch_point(void);
extern void lck_rw_lock_shared_to_exclusive_lockstat_patch_point(void);
extern void lck_rw_try_lock_shared_lockstat_patch_point(void);
extern void lck_rw_try_lock_exclusive_lockstat_patch_point(void);
extern void lck_mtx_lock_spin_lockstat_patch_point(void);
#endif

#endif /* CONFIG_DTRACE */

typedef struct lockstat_assembly_probe {
	int lsap_probe;
	vm_offset_t * lsap_patch_point;
} lockstat_assembly_probe_t;
	

	lockstat_assembly_probe_t assembly_probes[] =
	{
#if CONFIG_DTRACE
#if defined(__x86_64__)
		/*
		 * On x86 these points are better done via hot patches, which ensure
		 * there is zero overhead when not in use.  On x86 these patch points
		 * are swapped between the return instruction and a no-op, with the
		 * Dtrace call following the return.
		 */ 
		{ LS_LCK_MTX_LOCK_ACQUIRE,		(vm_offset_t *) lck_mtx_lock_lockstat_patch_point },
		{ LS_LCK_MTX_TRY_LOCK_ACQUIRE,		(vm_offset_t *) lck_mtx_try_lock_lockstat_patch_point },
		{ LS_LCK_MTX_TRY_SPIN_LOCK_ACQUIRE,	(vm_offset_t *) lck_mtx_try_lock_spin_lockstat_patch_point },
		{ LS_LCK_MTX_UNLOCK_RELEASE,		(vm_offset_t *) lck_mtx_unlock_lockstat_patch_point },
		{ LS_LCK_MTX_EXT_LOCK_ACQUIRE,		(vm_offset_t *) lck_mtx_lock_ext_lockstat_patch_point },
		{ LS_LCK_MTX_EXT_UNLOCK_RELEASE,	(vm_offset_t *) lck_mtx_ext_unlock_lockstat_patch_point },
		{ LS_LCK_RW_LOCK_SHARED_ACQUIRE,	(vm_offset_t *) lck_rw_lock_shared_lockstat_patch_point },
		{ LS_LCK_RW_LOCK_EXCL_ACQUIRE,		(vm_offset_t *) lck_rw_lock_exclusive_lockstat_patch_point },
		{ LS_LCK_RW_LOCK_SHARED_TO_EXCL_UPGRADE,(vm_offset_t *) lck_rw_lock_shared_to_exclusive_lockstat_patch_point },
		{ LS_LCK_RW_TRY_LOCK_SHARED_ACQUIRE,	(vm_offset_t *) lck_rw_try_lock_shared_lockstat_patch_point },
		{ LS_LCK_RW_TRY_LOCK_EXCL_ACQUIRE,	(vm_offset_t *) lck_rw_try_lock_exclusive_lockstat_patch_point },
		{ LS_LCK_MTX_LOCK_SPIN_ACQUIRE,		(vm_offset_t *) lck_mtx_lock_spin_lockstat_patch_point },
#endif
#endif /* CONFIG_DTRACE */
		{ LS_LCK_INVALID, NULL }
};


/*
 * Hot patch switches back and forth the probe points between NOP and RET.
 * The active argument indicates whether the probe point will turn on or off.
 *	on == plant a NOP and thus fall through to the probe call
 *     off == plant a RET and thus avoid the probe call completely
 * The lsap_probe identifies which probe we will patch.
 */
#if defined(__APPLE__)
static
void lockstat_hot_patch(boolean_t active, int ls_probe)
{
#pragma unused(active)
	int i;

	/*
	 * Loop through entire table, in case there are
	 * multiple patch points per probe. 
	 */
	for (i = 0; assembly_probes[i].lsap_patch_point; i++) {
		if (ls_probe == assembly_probes[i].lsap_probe)
#if defined(__x86_64__)
		{			
			uint8_t instr;
			instr = (active ? NOP : RET );
			(void) ml_nofault_copy( (vm_offset_t)&instr, *(assembly_probes[i].lsap_patch_point), 
								sizeof(instr));
		}
#endif
	} /* for */
}
#endif /* __APPLE__*/


void (*lockstat_probe)(dtrace_id_t, uint64_t, uint64_t,
				    uint64_t, uint64_t, uint64_t);

#if defined(__APPLE__)
/* This wrapper is used by arm assembler hot patched probes */
void
lockstat_probe_wrapper(int probe, uintptr_t lp, int rwflag)
{
	dtrace_id_t id;
	id = lockstat_probemap[probe];
	if (id != 0)
	{
		(*lockstat_probe)(id, (uintptr_t)lp, (uint64_t)rwflag, 0,0,0);
	}
}
#endif /* __APPLE__ */
    

static dev_info_t	*lockstat_devi;	/* saved in xxattach() for xxinfo() */
static dtrace_provider_id_t lockstat_id;

/*ARGSUSED*/
static int
lockstat_enable(void *arg, dtrace_id_t id, void *parg)
{
#pragma unused(arg) /* __APPLE__ */
    
	lockstat_probe_t *probe = parg;

	ASSERT(!lockstat_probemap[probe->lsp_probe]);

	lockstat_probemap[probe->lsp_probe] = id;
	membar_producer();

	lockstat_hot_patch(TRUE, probe->lsp_probe);
	membar_producer();
	return(0);

}

/*ARGSUSED*/
static void
lockstat_disable(void *arg, dtrace_id_t id, void *parg)
{
#pragma unused(arg, id) /* __APPLE__ */

	lockstat_probe_t *probe = parg;
	int i;

	ASSERT(lockstat_probemap[probe->lsp_probe]);

	lockstat_probemap[probe->lsp_probe] = 0;
	lockstat_hot_patch(FALSE, probe->lsp_probe);
	membar_producer();

	/*
	 * See if we have any probes left enabled.
	 */
	for (i = 0; i < LS_NPROBES; i++) {
		if (lockstat_probemap[i]) {
			/*
			 * This probe is still enabled.  We don't need to deal
			 * with waiting for all threads to be out of the
			 * lockstat critical sections; just return.
			 */
			return;
		}
	}

}

/*ARGSUSED*/
static void
lockstat_provide(void *arg, const dtrace_probedesc_t *desc)
{
#pragma unused(arg, desc) /* __APPLE__ */
    
	int i = 0;

	for (i = 0; lockstat_probes[i].lsp_func != NULL; i++) {
		lockstat_probe_t *probe = &lockstat_probes[i];

		if (dtrace_probe_lookup(lockstat_id, "mach_kernel",
		    probe->lsp_func, probe->lsp_name) != 0)
			continue;

		ASSERT(!probe->lsp_id);
		probe->lsp_id = dtrace_probe_create(lockstat_id,
		    "mach_kernel", probe->lsp_func, probe->lsp_name,
		    LOCKSTAT_AFRAMES, probe);
	}
}


/*ARGSUSED*/
static void
lockstat_destroy(void *arg, dtrace_id_t id, void *parg)
{
#pragma unused(arg, id) /* __APPLE__ */
    
	lockstat_probe_t *probe = parg;

	ASSERT(!lockstat_probemap[probe->lsp_probe]);
	probe->lsp_id = 0;
}

static dtrace_pattr_t lockstat_attr = {
{ DTRACE_STABILITY_EVOLVING, DTRACE_STABILITY_EVOLVING, DTRACE_CLASS_COMMON },
{ DTRACE_STABILITY_PRIVATE, DTRACE_STABILITY_PRIVATE, DTRACE_CLASS_UNKNOWN },
{ DTRACE_STABILITY_PRIVATE, DTRACE_STABILITY_PRIVATE, DTRACE_CLASS_UNKNOWN },
{ DTRACE_STABILITY_EVOLVING, DTRACE_STABILITY_EVOLVING, DTRACE_CLASS_COMMON },
{ DTRACE_STABILITY_EVOLVING, DTRACE_STABILITY_EVOLVING, DTRACE_CLASS_COMMON },
};

static dtrace_pops_t lockstat_pops = {
	lockstat_provide,
	NULL,
	lockstat_enable,
	lockstat_disable,
	NULL,
	NULL,
	NULL,
	NULL,
	NULL,
	lockstat_destroy
};

static int
lockstat_attach(dev_info_t *devi, ddi_attach_cmd_t cmd)
{
	switch (cmd) {
	case DDI_ATTACH:
		break;
	case DDI_RESUME:
		return (DDI_SUCCESS);
	default:
		return (DDI_FAILURE);
	}

	if (ddi_create_minor_node(devi, "lockstat", S_IFCHR, 0,
	    DDI_PSEUDO, 0) == DDI_FAILURE ||
	    dtrace_register("lockstat", &lockstat_attr, DTRACE_PRIV_KERNEL,
	    NULL, &lockstat_pops, NULL, &lockstat_id) != 0) {
		ddi_remove_minor_node(devi, NULL);
		return (DDI_FAILURE);
	}

	lockstat_probe = dtrace_probe;
	membar_producer();

	ddi_report_dev(devi);
	lockstat_devi = devi;
	return (DDI_SUCCESS);
}

d_open_t _lockstat_open;

int _lockstat_open(dev_t dev, int flags, int devtype, struct proc *p)
{
#pragma unused(dev,flags,devtype,p)
	return 0;
}

#define LOCKSTAT_MAJOR  -24 /* let the kernel pick the device number */

/*
 * A struct describing which functions will get invoked for certain
 * actions.
 */
static struct cdevsw lockstat_cdevsw =
{
	_lockstat_open,		/* open */
	eno_opcl,			/* close */
	eno_rdwrt,			/* read */
	eno_rdwrt,			/* write */
	eno_ioctl,			/* ioctl */
	(stop_fcn_t *)nulldev, /* stop */
	(reset_fcn_t *)nulldev, /* reset */
	NULL,				/* tty's */
	eno_select,			/* select */
	eno_mmap,			/* mmap */
	eno_strat,			/* strategy */
	eno_getc,			/* getc */
	eno_putc,			/* putc */
	0					/* type */
};

static int gLockstatInited = 0;

void lockstat_init( void );

void lockstat_init( void )
{
	if (0 == gLockstatInited)
	{
		int majdevno = cdevsw_add(LOCKSTAT_MAJOR, &lockstat_cdevsw);
		
		if (majdevno < 0) {
			printf("lockstat_init: failed to allocate a major number!\n");
			gLockstatInited = 0;
			return;
		}

		lockstat_attach( (dev_info_t	*)(uintptr_t)majdevno, DDI_ATTACH );
		gLockstatInited = 1;
	} else
		panic("lockstat_init: called twice!\n");
}
#undef LOCKSTAT_MAJOR
Commit	Line	Data
2d21ac55 A	1	/*
	2	* CDDL HEADER START
	3	*
	4	* The contents of this file are subject to the terms of the
	5	* Common Development and Distribution License (the "License").
	6	* You may not use this file except in compliance with the License.
	7	*
	8	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	9	* or http://www.opensolaris.org/os/licensing.
	10	* See the License for the specific language governing permissions
	11	* and limitations under the License.
	12	*
	13	* When distributing Covered Code, include this CDDL HEADER in each
	14	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	15	* If applicable, add the following below this CDDL HEADER, with the
	16	* fields enclosed by brackets "[]" replaced with your own identifying
	17	* information: Portions Copyright [yyyy] [name of copyright owner]
	18	*
	19	* CDDL HEADER END
	20	*/
	21	/*
6d2010ae	22	* Copyright 2009 Sun Microsystems, Inc. All rights reserved.
2d21ac55 A	23	* Use is subject to license terms.
	24	*/
	25
b0d623f7	26	/* #pragma ident "@(#)lockstat.c 1.12 08/01/16 SMI" */
2d21ac55 A	27
	28
	29	#ifdef KERNEL
	30	#ifndef _KERNEL
	31	#define _KERNEL /* Solaris vs. Darwin */
	32	#endif
	33	#endif
	34
	35	#include <sys/param.h>
	36	#include <sys/systm.h>
	37	#include <sys/errno.h>
	38	#include <sys/stat.h>
	39	#include <sys/ioctl.h>
	40	#include <sys/conf.h>
	41	#include <sys/fcntl.h>
	42	#include <miscfs/devfs/devfs.h>
	43
	44	#include <sys/dtrace.h>
	45	#include <sys/dtrace_impl.h>
	46
	47	#include <sys/dtrace_glue.h>
	48
	49	#include <sys/lockstat.h>
	50
	51	#include <kern/processor.h>
	52
	53	#define membar_producer dtrace_membar_producer
	54
	55	/*
	56	* Hot patch values, x86
	57	*/
39236c6e	58	#if defined(__x86_64__)
2d21ac55 A	59	#define NOP 0x90
	60	#define RET 0xc3
	61	#define LOCKSTAT_AFRAMES 1
2d21ac55 A	62	#else
	63	#error "not ported to this architecture"
	64	#endif
	65
2d21ac55 A	66	typedef struct lockstat_probe {
	67	const char *lsp_func;
	68	const char *lsp_name;
	69	int lsp_probe;
	70	dtrace_id_t lsp_id;
	71	} lockstat_probe_t;
	72
	73	lockstat_probe_t lockstat_probes[] =
	74	{
39236c6e	75	#if defined(__x86_64__)
316670eb	76	/* Only provide implemented probes for each architecture */
2d21ac55 A	77	{ LS_LCK_MTX_LOCK, LSA_ACQUIRE, LS_LCK_MTX_LOCK_ACQUIRE, DTRACE_IDNONE },
2d21ac55 A	78	{ LS_LCK_MTX_LOCK, LSA_SPIN, LS_LCK_MTX_LOCK_SPIN, DTRACE_IDNONE },
316670eb	79	{ LS_LCK_MTX_LOCK, LSA_BLOCK, LS_LCK_MTX_LOCK_BLOCK, DTRACE_IDNONE },
2d21ac55 A	80	{ LS_LCK_MTX_TRY_LOCK, LSA_ACQUIRE, LS_LCK_MTX_TRY_LOCK_ACQUIRE, DTRACE_IDNONE },
	81	{ LS_LCK_MTX_TRY_SPIN_LOCK, LSA_ACQUIRE, LS_LCK_MTX_TRY_SPIN_LOCK_ACQUIRE, DTRACE_IDNONE },
	82	{ LS_LCK_MTX_UNLOCK, LSA_RELEASE, LS_LCK_MTX_UNLOCK_RELEASE, DTRACE_IDNONE },
	83	{ LS_LCK_MTX_EXT_LOCK, LSA_ACQUIRE, LS_LCK_MTX_EXT_LOCK_ACQUIRE, DTRACE_IDNONE },
	84	{ LS_LCK_MTX_EXT_LOCK, LSA_SPIN, LS_LCK_MTX_EXT_LOCK_SPIN, DTRACE_IDNONE },
2d21ac55	85	{ LS_LCK_MTX_EXT_LOCK, LSA_BLOCK, LS_LCK_MTX_EXT_LOCK_BLOCK, DTRACE_IDNONE },
316670eb A	86	// { LS_LCK_MTX_EXT_TRY_LOCK, LSA_ACQUIRE, LS_LCK_MTX_TRY_EXT_LOCK_ACQUIRE, DTRACE_IDNONE },
	87	{ LS_LCK_MTX_EXT_UNLOCK, LSA_RELEASE, LS_LCK_MTX_EXT_UNLOCK_RELEASE, DTRACE_IDNONE },
	88	{ LS_LCK_MTX_LOCK_SPIN_LOCK, LSA_ACQUIRE, LS_LCK_MTX_LOCK_SPIN_ACQUIRE, DTRACE_IDNONE },
2d21ac55 A	89	{ LS_LCK_RW_LOCK_SHARED, LSR_ACQUIRE, LS_LCK_RW_LOCK_SHARED_ACQUIRE, DTRACE_IDNONE },
	90	{ LS_LCK_RW_LOCK_SHARED, LSR_BLOCK, LS_LCK_RW_LOCK_SHARED_BLOCK, DTRACE_IDNONE },
	91	{ LS_LCK_RW_LOCK_SHARED, LSR_SPIN, LS_LCK_RW_LOCK_SHARED_SPIN, DTRACE_IDNONE },
	92	{ LS_LCK_RW_LOCK_EXCL, LSR_ACQUIRE, LS_LCK_RW_LOCK_EXCL_ACQUIRE, DTRACE_IDNONE },
	93	{ LS_LCK_RW_LOCK_EXCL, LSR_BLOCK, LS_LCK_RW_LOCK_EXCL_BLOCK, DTRACE_IDNONE },
	94	{ LS_LCK_RW_LOCK_EXCL, LSR_SPIN, LS_LCK_RW_LOCK_EXCL_SPIN, DTRACE_IDNONE },
	95	{ LS_LCK_RW_DONE, LSR_RELEASE, LS_LCK_RW_DONE_RELEASE, DTRACE_IDNONE },
	96	{ LS_LCK_RW_TRY_LOCK_SHARED, LSR_ACQUIRE, LS_LCK_RW_TRY_LOCK_SHARED_ACQUIRE, DTRACE_IDNONE },
	97	{ LS_LCK_RW_TRY_LOCK_EXCL, LSR_ACQUIRE, LS_LCK_RW_TRY_LOCK_EXCL_ACQUIRE, DTRACE_IDNONE },
	98	{ LS_LCK_RW_LOCK_SHARED_TO_EXCL, LSR_UPGRADE, LS_LCK_RW_LOCK_SHARED_TO_EXCL_UPGRADE, DTRACE_IDNONE },
2d21ac55	99	{ LS_LCK_RW_LOCK_SHARED_TO_EXCL, LSR_SPIN, LS_LCK_RW_LOCK_SHARED_TO_EXCL_SPIN, DTRACE_IDNONE },
316670eb	100	{ LS_LCK_RW_LOCK_SHARED_TO_EXCL, LSR_BLOCK, LS_LCK_RW_LOCK_SHARED_TO_EXCL_BLOCK, DTRACE_IDNONE },
2d21ac55	101	{ LS_LCK_RW_LOCK_EXCL_TO_SHARED, LSR_DOWNGRADE, LS_LCK_RW_LOCK_EXCL_TO_SHARED_DOWNGRADE, DTRACE_IDNONE },
316670eb	102	#endif
2d21ac55 A	103	#ifdef LATER
	104	/* Interlock and spinlock measurements would be nice, but later */
	105	{ LS_LCK_SPIN_LOCK, LSS_ACQUIRE, LS_LCK_SPIN_LOCK_ACQUIRE, DTRACE_IDNONE },
	106	{ LS_LCK_SPIN_LOCK, LSS_SPIN, LS_LCK_SPIN_LOCK_SPIN, DTRACE_IDNONE },
	107	{ LS_LCK_SPIN_UNLOCK, LSS_RELEASE, LS_LCK_SPIN_UNLOCK_RELEASE, DTRACE_IDNONE },
	108
	109	{ LS_LCK_RW_LOCK_EXCL_TO_SHARED, LSA_ILK_SPIN, LS_LCK_RW_LOCK_EXCL_TO_SHARED_ILK_SPIN, DTRACE_IDNONE },
	110	{ LS_LCK_MTX_LOCK, LSA_ILK_SPIN, LS_LCK_MTX_LOCK_ILK_SPIN, DTRACE_IDNONE },
	111	{ LS_LCK_MTX_EXT_LOCK, LSA_ILK_SPIN, LS_LCK_MTX_EXT_LOCK_ILK_SPIN, DTRACE_IDNONE },
	112	{ LS_LCK_RW_TRY_LOCK_EXCL, LSA_ILK_SPIN, LS_LCK_RW_TRY_LOCK_EXCL_ILK_SPIN, DTRACE_IDNONE },
	113	{ LS_LCK_RW_TRY_LOCK_SHARED, LSA_SPIN, LS_LCK_RW_TRY_LOCK_SHARED_SPIN, DTRACE_IDNONE },
	114	#endif
	115
b0d623f7	116	{ NULL, NULL, 0, 0 }
2d21ac55 A	117	};
	118
	119	dtrace_id_t lockstat_probemap[LS_NPROBES];
	120
b0d623f7	121	#if CONFIG_DTRACE
39236c6e	122	#if defined(__x86_64__)
b0d623f7 A	123	extern void lck_mtx_lock_lockstat_patch_point(void);
	124	extern void lck_mtx_try_lock_lockstat_patch_point(void);
	125	extern void lck_mtx_try_lock_spin_lockstat_patch_point(void);
	126	extern void lck_mtx_unlock_lockstat_patch_point(void);
	127	extern void lck_mtx_lock_ext_lockstat_patch_point(void);
	128	extern void lck_mtx_ext_unlock_lockstat_patch_point(void);
b0d623f7 A	129	extern void lck_rw_lock_shared_lockstat_patch_point(void);
	130	extern void lck_rw_lock_exclusive_lockstat_patch_point(void);
	131	extern void lck_rw_lock_shared_to_exclusive_lockstat_patch_point(void);
	132	extern void lck_rw_try_lock_shared_lockstat_patch_point(void);
	133	extern void lck_rw_try_lock_exclusive_lockstat_patch_point(void);
	134	extern void lck_mtx_lock_spin_lockstat_patch_point(void);
39236c6e A	135	#endif
39236c6e A	136
b0d623f7	137	#endif /* CONFIG_DTRACE */
2d21ac55	138
316670eb A	139	typedef struct lockstat_assembly_probe {
	140	int lsap_probe;
	141	vm_offset_t * lsap_patch_point;
	142	} lockstat_assembly_probe_t;
	143
	144
	145	lockstat_assembly_probe_t assembly_probes[] =
	146	{
b0d623f7	147	#if CONFIG_DTRACE
39236c6e	148	#if defined(__x86_64__)
316670eb A	149	/*
	150	* On x86 these points are better done via hot patches, which ensure
	151	* there is zero overhead when not in use. On x86 these patch points
	152	* are swapped between the return instruction and a no-op, with the
	153	* Dtrace call following the return.
	154	*/
	155	{ LS_LCK_MTX_LOCK_ACQUIRE, (vm_offset_t *) lck_mtx_lock_lockstat_patch_point },
	156	{ LS_LCK_MTX_TRY_LOCK_ACQUIRE, (vm_offset_t *) lck_mtx_try_lock_lockstat_patch_point },
	157	{ LS_LCK_MTX_TRY_SPIN_LOCK_ACQUIRE, (vm_offset_t *) lck_mtx_try_lock_spin_lockstat_patch_point },
	158	{ LS_LCK_MTX_UNLOCK_RELEASE, (vm_offset_t *) lck_mtx_unlock_lockstat_patch_point },
	159	{ LS_LCK_MTX_EXT_LOCK_ACQUIRE, (vm_offset_t *) lck_mtx_lock_ext_lockstat_patch_point },
	160	{ LS_LCK_MTX_EXT_UNLOCK_RELEASE, (vm_offset_t *) lck_mtx_ext_unlock_lockstat_patch_point },
	161	{ LS_LCK_RW_LOCK_SHARED_ACQUIRE, (vm_offset_t *) lck_rw_lock_shared_lockstat_patch_point },
	162	{ LS_LCK_RW_LOCK_EXCL_ACQUIRE, (vm_offset_t *) lck_rw_lock_exclusive_lockstat_patch_point },
	163	{ LS_LCK_RW_LOCK_SHARED_TO_EXCL_UPGRADE,(vm_offset_t *) lck_rw_lock_shared_to_exclusive_lockstat_patch_point },
	164	{ LS_LCK_RW_TRY_LOCK_SHARED_ACQUIRE, (vm_offset_t *) lck_rw_try_lock_shared_lockstat_patch_point },
	165	{ LS_LCK_RW_TRY_LOCK_EXCL_ACQUIRE, (vm_offset_t *) lck_rw_try_lock_exclusive_lockstat_patch_point },
	166	{ LS_LCK_MTX_LOCK_SPIN_ACQUIRE, (vm_offset_t *) lck_mtx_lock_spin_lockstat_patch_point },
b0d623f7 A	167	#endif
b0d623f7 A	168	#endif /* CONFIG_DTRACE */
316670eb	169	{ LS_LCK_INVALID, NULL }
2d21ac55	170	};
39236c6e A	171
39236c6e A	172
2d21ac55 A	173	/*
2d21ac55 A	174	* Hot patch switches back and forth the probe points between NOP and RET.
316670eb A	175	* The active argument indicates whether the probe point will turn on or off.
	176	* on == plant a NOP and thus fall through to the probe call
	177	* off == plant a RET and thus avoid the probe call completely
	178	* The lsap_probe identifies which probe we will patch.
2d21ac55	179	*/
b0d623f7 A	180	#if defined(__APPLE__)
b0d623f7 A	181	static
316670eb	182	void lockstat_hot_patch(boolean_t active, int ls_probe)
2d21ac55	183	{
c910b4d9	184	#pragma unused(active)
2d21ac55 A	185	int i;
2d21ac55 A	186
316670eb A	187	/*
	188	* Loop through entire table, in case there are
	189	* multiple patch points per probe.
	190	*/
	191	for (i = 0; assembly_probes[i].lsap_patch_point; i++) {
	192	if (ls_probe == assembly_probes[i].lsap_probe)
39236c6e	193	#if defined(__x86_64__)
316670eb A	194	{
	195	uint8_t instr;
	196	instr = (active ? NOP : RET );
	197	(void) ml_nofault_copy( (vm_offset_t)&instr, *(assembly_probes[i].lsap_patch_point),
2d21ac55	198	sizeof(instr));
316670eb	199	}
2d21ac55	200	#endif
316670eb	201	} /* for */
2d21ac55	202	}
316670eb	203	#endif /* __APPLE__*/
2d21ac55 A	204
	205
	206	void (*lockstat_probe)(dtrace_id_t, uint64_t, uint64_t,
	207	uint64_t, uint64_t, uint64_t);
	208
316670eb A	209	#if defined(__APPLE__)
	210	/* This wrapper is used by arm assembler hot patched probes */
	211	void
	212	lockstat_probe_wrapper(int probe, uintptr_t lp, int rwflag)
	213	{
	214	dtrace_id_t id;
	215	id = lockstat_probemap[probe];
	216	if (id != 0)
	217	{
	218	(*lockstat_probe)(id, (uintptr_t)lp, (uint64_t)rwflag, 0,0,0);
	219	}
	220	}
	221	#endif /* __APPLE__ */
	222
	223
2d21ac55 A	224	static dev_info_t lockstat_devi; / saved in xxattach() for xxinfo() */
	225	static dtrace_provider_id_t lockstat_id;
	226
	227	/ARGSUSED/
6d2010ae	228	static int
2d21ac55 A	229	lockstat_enable(void arg, dtrace_id_t id, void parg)
2d21ac55 A	230	{
b0d623f7 A	231	#pragma unused(arg) /* __APPLE__ */
b0d623f7 A	232
2d21ac55 A	233	lockstat_probe_t *probe = parg;
	234
	235	ASSERT(!lockstat_probemap[probe->lsp_probe]);
	236
	237	lockstat_probemap[probe->lsp_probe] = id;
	238	membar_producer();
	239
316670eb	240	lockstat_hot_patch(TRUE, probe->lsp_probe);
2d21ac55	241	membar_producer();
6d2010ae	242	return(0);
2d21ac55 A	243
	244	}
	245
	246	/ARGSUSED/
	247	static void
	248	lockstat_disable(void arg, dtrace_id_t id, void parg)
	249	{
b0d623f7 A	250	#pragma unused(arg, id) /* __APPLE__ */
b0d623f7 A	251
2d21ac55 A	252	lockstat_probe_t *probe = parg;
	253	int i;
	254
	255	ASSERT(lockstat_probemap[probe->lsp_probe]);
	256
	257	lockstat_probemap[probe->lsp_probe] = 0;
316670eb	258	lockstat_hot_patch(FALSE, probe->lsp_probe);
2d21ac55 A	259	membar_producer();
	260
	261	/*
	262	* See if we have any probes left enabled.
	263	*/
	264	for (i = 0; i < LS_NPROBES; i++) {
	265	if (lockstat_probemap[i]) {
	266	/*
	267	* This probe is still enabled. We don't need to deal
	268	* with waiting for all threads to be out of the
	269	* lockstat critical sections; just return.
	270	*/
	271	return;
	272	}
	273	}
	274
	275	}
	276
	277	/ARGSUSED/
	278	static void
	279	lockstat_provide(void arg, const dtrace_probedesc_t desc)
	280	{
b0d623f7 A	281	#pragma unused(arg, desc) /* __APPLE__ */
b0d623f7 A	282
2d21ac55 A	283	int i = 0;
	284
	285	for (i = 0; lockstat_probes[i].lsp_func != NULL; i++) {
	286	lockstat_probe_t *probe = &lockstat_probes[i];
	287
	288	if (dtrace_probe_lookup(lockstat_id, "mach_kernel",
	289	probe->lsp_func, probe->lsp_name) != 0)
	290	continue;
	291
	292	ASSERT(!probe->lsp_id);
	293	probe->lsp_id = dtrace_probe_create(lockstat_id,
	294	"mach_kernel", probe->lsp_func, probe->lsp_name,
	295	LOCKSTAT_AFRAMES, probe);
	296	}
	297	}
	298
	299
	300	/ARGSUSED/
	301	static void
	302	lockstat_destroy(void arg, dtrace_id_t id, void parg)
	303	{
b0d623f7 A	304	#pragma unused(arg, id) /* __APPLE__ */
b0d623f7 A	305
2d21ac55 A	306	lockstat_probe_t *probe = parg;
	307
	308	ASSERT(!lockstat_probemap[probe->lsp_probe]);
	309	probe->lsp_id = 0;
	310	}
	311
	312	static dtrace_pattr_t lockstat_attr = {
	313	{ DTRACE_STABILITY_EVOLVING, DTRACE_STABILITY_EVOLVING, DTRACE_CLASS_COMMON },
	314	{ DTRACE_STABILITY_PRIVATE, DTRACE_STABILITY_PRIVATE, DTRACE_CLASS_UNKNOWN },
	315	{ DTRACE_STABILITY_PRIVATE, DTRACE_STABILITY_PRIVATE, DTRACE_CLASS_UNKNOWN },
	316	{ DTRACE_STABILITY_EVOLVING, DTRACE_STABILITY_EVOLVING, DTRACE_CLASS_COMMON },
	317	{ DTRACE_STABILITY_EVOLVING, DTRACE_STABILITY_EVOLVING, DTRACE_CLASS_COMMON },
	318	};
	319
	320	static dtrace_pops_t lockstat_pops = {
	321	lockstat_provide,
	322	NULL,
	323	lockstat_enable,
	324	lockstat_disable,
	325	NULL,
	326	NULL,
	327	NULL,
	328	NULL,
	329	NULL,
	330	lockstat_destroy
	331	};
	332
	333	static int
	334	lockstat_attach(dev_info_t *devi, ddi_attach_cmd_t cmd)
	335	{
	336	switch (cmd) {
	337	case DDI_ATTACH:
	338	break;
	339	case DDI_RESUME:
	340	return (DDI_SUCCESS);
	341	default:
	342	return (DDI_FAILURE);
	343	}
	344
	345	if (ddi_create_minor_node(devi, "lockstat", S_IFCHR, 0,
	346	DDI_PSEUDO, 0) == DDI_FAILURE \|\|
	347	dtrace_register("lockstat", &lockstat_attr, DTRACE_PRIV_KERNEL,
	348	NULL, &lockstat_pops, NULL, &lockstat_id) != 0) {
	349	ddi_remove_minor_node(devi, NULL);
	350	return (DDI_FAILURE);
	351	}
	352
b0d623f7 A	353	lockstat_probe = dtrace_probe;
	354	membar_producer();
	355
2d21ac55 A	356	ddi_report_dev(devi);
2d21ac55 A	357	lockstat_devi = devi;
2d21ac55 A	358	return (DDI_SUCCESS);
	359	}
	360
	361	d_open_t _lockstat_open;
	362
	363	int _lockstat_open(dev_t dev, int flags, int devtype, struct proc *p)
	364	{
	365	#pragma unused(dev,flags,devtype,p)
	366	return 0;
	367	}
	368
	369	#define LOCKSTAT_MAJOR -24 /* let the kernel pick the device number */
	370
	371	/*
	372	* A struct describing which functions will get invoked for certain
	373	* actions.
	374	*/
	375	static struct cdevsw lockstat_cdevsw =
	376	{
	377	_lockstat_open, /* open */
	378	eno_opcl, /* close */
	379	eno_rdwrt, /* read */
	380	eno_rdwrt, /* write */
	381	eno_ioctl, /* ioctl */
	382	(stop_fcn_t )nulldev, / stop */
	383	(reset_fcn_t )nulldev, / reset */
	384	NULL, /* tty's */
	385	eno_select, /* select */
	386	eno_mmap, /* mmap */
	387	eno_strat, /* strategy */
	388	eno_getc, /* getc */
	389	eno_putc, /* putc */
	390	0 /* type */
	391	};
	392
	393	static int gLockstatInited = 0;
	394
	395	void lockstat_init( void );
	396
	397	void lockstat_init( void )
	398	{
	399	if (0 == gLockstatInited)
	400	{
	401	int majdevno = cdevsw_add(LOCKSTAT_MAJOR, &lockstat_cdevsw);
	402
	403	if (majdevno < 0) {
	404	printf("lockstat_init: failed to allocate a major number!\n");
	405	gLockstatInited = 0;
	406	return;
	407	}
	408
b0d623f7	409	lockstat_attach( (dev_info_t *)(uintptr_t)majdevno, DDI_ATTACH );
2d21ac55 A	410	gLockstatInited = 1;
	411	} else
	412	panic("lockstat_init: called twice!\n");
	413	}
	414	#undef LOCKSTAT_MAJOR