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/*
22 * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
23 * Use is subject to license terms.
24 */
25
26#pragma ident	"%Z%%M%	%I%	%E% SMI"
27
28#include <sys/types.h>
29#include <sys/param.h>
30#include <sys/stat.h>
31#include <sys/open.h>
32#include <sys/file.h>
33#include <sys/conf.h>
34#include <sys/modctl.h>
35#include <sys/cmn_err.h>
36#include <sys/bitmap.h>
37#include <sys/debug.h>
38#include <sys/kmem.h>
39#include <sys/errno.h>
40#include <sys/sysmacros.h>
41#include <sys/lockstat.h>
42#include <sys/atomic.h>
43#include <sys/dtrace.h>
44
45#include <sys/ddi.h>
46#include <sys/sunddi.h>
47
48typedef struct lockstat_probe {
49	const char	*lsp_func;
50	const char	*lsp_name;
51	int		lsp_probe;
52	dtrace_id_t	lsp_id;
53} lockstat_probe_t;
54
55lockstat_probe_t lockstat_probes[] =
56{
57	{ LS_MUTEX_ENTER,	LSA_ACQUIRE,	LS_MUTEX_ENTER_ACQUIRE },
58	{ LS_MUTEX_ENTER,	LSA_BLOCK,	LS_MUTEX_ENTER_BLOCK },
59	{ LS_MUTEX_ENTER,	LSA_SPIN,	LS_MUTEX_ENTER_SPIN },
60	{ LS_MUTEX_EXIT,	LSA_RELEASE,	LS_MUTEX_EXIT_RELEASE },
61	{ LS_MUTEX_DESTROY,	LSA_RELEASE,	LS_MUTEX_DESTROY_RELEASE },
62	{ LS_MUTEX_TRYENTER,	LSA_ACQUIRE,	LS_MUTEX_TRYENTER_ACQUIRE },
63	{ LS_LOCK_SET,		LSS_ACQUIRE,	LS_LOCK_SET_ACQUIRE },
64	{ LS_LOCK_SET,		LSS_SPIN,	LS_LOCK_SET_SPIN },
65	{ LS_LOCK_SET_SPL,	LSS_ACQUIRE,	LS_LOCK_SET_SPL_ACQUIRE },
66	{ LS_LOCK_SET_SPL,	LSS_SPIN,	LS_LOCK_SET_SPL_SPIN },
67	{ LS_LOCK_TRY,		LSS_ACQUIRE,	LS_LOCK_TRY_ACQUIRE },
68	{ LS_LOCK_CLEAR,	LSS_RELEASE,	LS_LOCK_CLEAR_RELEASE },
69	{ LS_LOCK_CLEAR_SPLX,	LSS_RELEASE,	LS_LOCK_CLEAR_SPLX_RELEASE },
70	{ LS_CLOCK_UNLOCK,	LSS_RELEASE,	LS_CLOCK_UNLOCK_RELEASE },
71	{ LS_RW_ENTER,		LSR_ACQUIRE,	LS_RW_ENTER_ACQUIRE },
72	{ LS_RW_ENTER,		LSR_BLOCK,	LS_RW_ENTER_BLOCK },
73	{ LS_RW_EXIT,		LSR_RELEASE,	LS_RW_EXIT_RELEASE },
74	{ LS_RW_TRYENTER,	LSR_ACQUIRE,	LS_RW_TRYENTER_ACQUIRE },
75	{ LS_RW_TRYUPGRADE,	LSR_UPGRADE,	LS_RW_TRYUPGRADE_UPGRADE },
76	{ LS_RW_DOWNGRADE,	LSR_DOWNGRADE,	LS_RW_DOWNGRADE_DOWNGRADE },
77	{ LS_THREAD_LOCK,	LST_SPIN,	LS_THREAD_LOCK_SPIN },
78	{ LS_THREAD_LOCK_HIGH,	LST_SPIN,	LS_THREAD_LOCK_HIGH_SPIN },
79	{ NULL }
80};
81
82static dev_info_t	*lockstat_devi;	/* saved in xxattach() for xxinfo() */
83static kmutex_t		lockstat_test;	/* for testing purposes only */
84static dtrace_provider_id_t lockstat_id;
85
86/*ARGSUSED*/
87static void
88lockstat_enable(void *arg, dtrace_id_t id, void *parg)
89{
90	lockstat_probe_t *probe = parg;
91
92	ASSERT(!lockstat_probemap[probe->lsp_probe]);
93
94	lockstat_probemap[probe->lsp_probe] = id;
95	membar_producer();
96
97	lockstat_hot_patch();
98	membar_producer();
99
100	/*
101	 * Immediately generate a record for the lockstat_test mutex
102	 * to verify that the mutex hot-patch code worked as expected.
103	 */
104	mutex_enter(&lockstat_test);
105	mutex_exit(&lockstat_test);
106}
107
108/*ARGSUSED*/
109static void
110lockstat_disable(void *arg, dtrace_id_t id, void *parg)
111{
112	lockstat_probe_t *probe = parg;
113	int i;
114
115	ASSERT(lockstat_probemap[probe->lsp_probe]);
116
117	lockstat_probemap[probe->lsp_probe] = 0;
118	lockstat_hot_patch();
119	membar_producer();
120
121	/*
122	 * See if we have any probes left enabled.
123	 */
124	for (i = 0; i < LS_NPROBES; i++) {
125		if (lockstat_probemap[i]) {
126			/*
127			 * This probe is still enabled.  We don't need to deal
128			 * with waiting for all threads to be out of the
129			 * lockstat critical sections; just return.
130			 */
131			return;
132		}
133	}
134
135	/*
136	 * The delay() here isn't as cheesy as you might think.  We don't
137	 * want to busy-loop in the kernel, so we have to give up the
138	 * CPU between calls to lockstat_active_threads(); that much is
139	 * obvious.  But the reason it's a do..while loop rather than a
140	 * while loop is subtle.  The memory barrier above guarantees that
141	 * no threads will enter the lockstat code from this point forward.
142	 * However, another thread could already be executing lockstat code
143	 * without our knowledge if the update to its t_lockstat field hasn't
144	 * cleared its CPU's store buffer.  Delaying for one clock tick
145	 * guarantees that either (1) the thread will have *ample* time to
146	 * complete its work, or (2) the thread will be preempted, in which
147	 * case it will have to grab and release a dispatcher lock, which
148	 * will flush that CPU's store buffer.  Either way we're covered.
149	 */
150	do {
151		delay(1);
152	} while (lockstat_active_threads());
153}
154
155/*ARGSUSED*/
156static int
157lockstat_open(dev_t *devp, int flag, int otyp, cred_t *cred_p)
158{
159	return (0);
160}
161
162/* ARGSUSED */
163static int
164lockstat_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
165{
166	int error;
167
168	switch (infocmd) {
169	case DDI_INFO_DEVT2DEVINFO:
170		*result = (void *) lockstat_devi;
171		error = DDI_SUCCESS;
172		break;
173	case DDI_INFO_DEVT2INSTANCE:
174		*result = (void *)0;
175		error = DDI_SUCCESS;
176		break;
177	default:
178		error = DDI_FAILURE;
179	}
180	return (error);
181}
182
183/*ARGSUSED*/
184static void
185lockstat_provide(void *arg, const dtrace_probedesc_t *desc)
186{
187	int i = 0;
188
189	for (i = 0; lockstat_probes[i].lsp_func != NULL; i++) {
190		lockstat_probe_t *probe = &lockstat_probes[i];
191
192		if (dtrace_probe_lookup(lockstat_id, "genunix",
193		    probe->lsp_func, probe->lsp_name) != 0)
194			continue;
195
196		ASSERT(!probe->lsp_id);
197		probe->lsp_id = dtrace_probe_create(lockstat_id,
198		    "genunix", probe->lsp_func, probe->lsp_name,
199		    1, probe);
200	}
201}
202
203/*ARGSUSED*/
204static void
205lockstat_destroy(void *arg, dtrace_id_t id, void *parg)
206{
207	lockstat_probe_t *probe = parg;
208
209	ASSERT(!lockstat_probemap[probe->lsp_probe]);
210	probe->lsp_id = 0;
211}
212
213static dtrace_pattr_t lockstat_attr = {
214{ DTRACE_STABILITY_EVOLVING, DTRACE_STABILITY_EVOLVING, DTRACE_CLASS_COMMON },
215{ DTRACE_STABILITY_PRIVATE, DTRACE_STABILITY_PRIVATE, DTRACE_CLASS_UNKNOWN },
216{ DTRACE_STABILITY_PRIVATE, DTRACE_STABILITY_PRIVATE, DTRACE_CLASS_UNKNOWN },
217{ DTRACE_STABILITY_EVOLVING, DTRACE_STABILITY_EVOLVING, DTRACE_CLASS_COMMON },
218{ DTRACE_STABILITY_EVOLVING, DTRACE_STABILITY_EVOLVING, DTRACE_CLASS_COMMON },
219};
220
221static dtrace_pops_t lockstat_pops = {
222	lockstat_provide,
223	NULL,
224	lockstat_enable,
225	lockstat_disable,
226	NULL,
227	NULL,
228	NULL,
229	NULL,
230	NULL,
231	lockstat_destroy
232};
233
234static int
235lockstat_attach(dev_info_t *devi, ddi_attach_cmd_t cmd)
236{
237	switch (cmd) {
238	case DDI_ATTACH:
239		break;
240	case DDI_RESUME:
241		return (DDI_SUCCESS);
242	default:
243		return (DDI_FAILURE);
244	}
245
246	if (ddi_create_minor_node(devi, "lockstat", S_IFCHR, 0,
247	    DDI_PSEUDO, 0) == DDI_FAILURE ||
248	    dtrace_register("lockstat", &lockstat_attr, DTRACE_PRIV_KERNEL,
249	    NULL, &lockstat_pops, NULL, &lockstat_id) != 0) {
250		ddi_remove_minor_node(devi, NULL);
251		return (DDI_FAILURE);
252	}
253
254	lockstat_probe = dtrace_probe;
255	membar_producer();
256
257	ddi_report_dev(devi);
258	lockstat_devi = devi;
259	return (DDI_SUCCESS);
260}
261
262static int
263lockstat_detach(dev_info_t *devi, ddi_detach_cmd_t cmd)
264{
265	switch (cmd) {
266	case DDI_DETACH:
267		break;
268	case DDI_SUSPEND:
269		return (DDI_SUCCESS);
270	default:
271		return (DDI_FAILURE);
272	}
273
274	if (dtrace_unregister(lockstat_id) != 0)
275		return (DDI_FAILURE);
276
277	ddi_remove_minor_node(devi, NULL);
278	return (DDI_SUCCESS);
279}
280
281/*
282 * Configuration data structures
283 */
284static struct cb_ops lockstat_cb_ops = {
285	lockstat_open,		/* open */
286	nodev,			/* close */
287	nulldev,		/* strategy */
288	nulldev,		/* print */
289	nodev,			/* dump */
290	nodev,			/* read */
291	nodev,			/* write */
292	nodev,			/* ioctl */
293	nodev,			/* devmap */
294	nodev,			/* mmap */
295	nodev,			/* segmap */
296	nochpoll,		/* poll */
297	ddi_prop_op,		/* cb_prop_op */
298	0,			/* streamtab */
299	D_MP | D_NEW		/* Driver compatibility flag */
300};
301
302static struct dev_ops lockstat_ops = {
303	DEVO_REV,		/* devo_rev, */
304	0,			/* refcnt */
305	lockstat_info,		/* getinfo */
306	nulldev,		/* identify */
307	nulldev,		/* probe */
308	lockstat_attach,	/* attach */
309	lockstat_detach,	/* detach */
310	nulldev,		/* reset */
311	&lockstat_cb_ops,	/* cb_ops */
312	NULL,			/* bus_ops */
313};
314
315static struct modldrv modldrv = {
316	&mod_driverops,		/* Type of module.  This one is a driver */
317	"Lock Statistics %I%",	/* name of module */
318	&lockstat_ops,		/* driver ops */
319};
320
321static struct modlinkage modlinkage = {
322	MODREV_1, (void *)&modldrv, NULL
323};
324
325int
326_init(void)
327{
328	return (mod_install(&modlinkage));
329}
330
331int
332_fini(void)
333{
334	return (mod_remove(&modlinkage));
335}
336
337int
338_info(struct modinfo *modinfop)
339{
340	return (mod_info(&modlinkage, modinfop));
341}
342