[deliverable/linux.git] / drivers / staging / lustre / lustre / ptlrpc / sec_gc.c

/*
 * GPL HEADER START
 *
 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License version 2 only,
 * as published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it will be useful, but
 * WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * General Public License version 2 for more details (a copy is included
 * in the LICENSE file that accompanied this code).
 *
 * You should have received a copy of the GNU General Public License
 * version 2 along with this program; If not, see
 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
 *
 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
 * CA 95054 USA or visit www.sun.com if you need additional information or
 * have any questions.
 *
 * GPL HEADER END
 */
/*
 * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
 * Use is subject to license terms.
 *
 * Copyright (c) 2011, 2012, Intel Corporation.
 */
/*
 * This file is part of Lustre, http://www.lustre.org/
 * Lustre is a trademark of Sun Microsystems, Inc.
 *
 * lustre/ptlrpc/sec_gc.c
 *
 * Author: Eric Mei <ericm@clusterfs.com>
 */

#define DEBUG_SUBSYSTEM S_SEC

#include "../../include/linux/libcfs/libcfs.h"

#include <obd_support.h>
#include <obd_class.h>
#include <lustre_net.h>
#include <lustre_sec.h>

#define SEC_GC_INTERVAL (30 * 60)


static struct mutex sec_gc_mutex;
static LIST_HEAD(sec_gc_list);
static spinlock_t sec_gc_list_lock;

static LIST_HEAD(sec_gc_ctx_list);
static spinlock_t sec_gc_ctx_list_lock;

static struct ptlrpc_thread sec_gc_thread;
static atomic_t sec_gc_wait_del = ATOMIC_INIT(0);


void sptlrpc_gc_add_sec(struct ptlrpc_sec *sec)
{
	LASSERT(sec->ps_policy->sp_cops->gc_ctx);
	LASSERT(sec->ps_gc_interval > 0);
	LASSERT(list_empty(&sec->ps_gc_list));

	sec->ps_gc_next = cfs_time_current_sec() + sec->ps_gc_interval;

	spin_lock(&sec_gc_list_lock);
	list_add_tail(&sec_gc_list, &sec->ps_gc_list);
	spin_unlock(&sec_gc_list_lock);

	CDEBUG(D_SEC, "added sec %p(%s)\n", sec, sec->ps_policy->sp_name);
}
EXPORT_SYMBOL(sptlrpc_gc_add_sec);

void sptlrpc_gc_del_sec(struct ptlrpc_sec *sec)
{
	if (list_empty(&sec->ps_gc_list))
		return;

	might_sleep();

	/* signal before list_del to make iteration in gc thread safe */
	atomic_inc(&sec_gc_wait_del);

	spin_lock(&sec_gc_list_lock);
	list_del_init(&sec->ps_gc_list);
	spin_unlock(&sec_gc_list_lock);

	/* barrier */
	mutex_lock(&sec_gc_mutex);
	mutex_unlock(&sec_gc_mutex);

	atomic_dec(&sec_gc_wait_del);

	CDEBUG(D_SEC, "del sec %p(%s)\n", sec, sec->ps_policy->sp_name);
}
EXPORT_SYMBOL(sptlrpc_gc_del_sec);

void sptlrpc_gc_add_ctx(struct ptlrpc_cli_ctx *ctx)
{
	LASSERT(list_empty(&ctx->cc_gc_chain));

	CDEBUG(D_SEC, "hand over ctx %p(%u->%s)\n",
	       ctx, ctx->cc_vcred.vc_uid, sec2target_str(ctx->cc_sec));
	spin_lock(&sec_gc_ctx_list_lock);
	list_add(&ctx->cc_gc_chain, &sec_gc_ctx_list);
	spin_unlock(&sec_gc_ctx_list_lock);

	thread_add_flags(&sec_gc_thread, SVC_SIGNAL);
	wake_up(&sec_gc_thread.t_ctl_waitq);
}
EXPORT_SYMBOL(sptlrpc_gc_add_ctx);

static void sec_process_ctx_list(void)
{
	struct ptlrpc_cli_ctx *ctx;

	spin_lock(&sec_gc_ctx_list_lock);

	while (!list_empty(&sec_gc_ctx_list)) {
		ctx = list_entry(sec_gc_ctx_list.next,
				     struct ptlrpc_cli_ctx, cc_gc_chain);
		list_del_init(&ctx->cc_gc_chain);
		spin_unlock(&sec_gc_ctx_list_lock);

		LASSERT(ctx->cc_sec);
		LASSERT(atomic_read(&ctx->cc_refcount) == 1);
		CDEBUG(D_SEC, "gc pick up ctx %p(%u->%s)\n",
		       ctx, ctx->cc_vcred.vc_uid, sec2target_str(ctx->cc_sec));
		sptlrpc_cli_ctx_put(ctx, 1);

		spin_lock(&sec_gc_ctx_list_lock);
	}

	spin_unlock(&sec_gc_ctx_list_lock);
}

static void sec_do_gc(struct ptlrpc_sec *sec)
{
	LASSERT(sec->ps_policy->sp_cops->gc_ctx);

	if (unlikely(sec->ps_gc_next == 0)) {
		CDEBUG(D_SEC, "sec %p(%s) has 0 gc time\n",
		      sec, sec->ps_policy->sp_name);
		return;
	}

	CDEBUG(D_SEC, "check on sec %p(%s)\n", sec, sec->ps_policy->sp_name);

	if (cfs_time_after(sec->ps_gc_next, cfs_time_current_sec()))
		return;

	sec->ps_policy->sp_cops->gc_ctx(sec);
	sec->ps_gc_next = cfs_time_current_sec() + sec->ps_gc_interval;
}

static int sec_gc_main(void *arg)
{
	struct ptlrpc_thread *thread = (struct ptlrpc_thread *) arg;
	struct l_wait_info    lwi;

	unshare_fs_struct();

	/* Record that the thread is running */
	thread_set_flags(thread, SVC_RUNNING);
	wake_up(&thread->t_ctl_waitq);

	while (1) {
		struct ptlrpc_sec *sec;

		thread_clear_flags(thread, SVC_SIGNAL);
		sec_process_ctx_list();
again:
		/* go through sec list do gc.
		 * FIXME here we iterate through the whole list each time which
		 * is not optimal. we perhaps want to use balanced binary tree
		 * to trace each sec as order of expiry time.
		 * another issue here is we wakeup as fixed interval instead of
		 * according to each sec's expiry time */
		mutex_lock(&sec_gc_mutex);
		list_for_each_entry(sec, &sec_gc_list, ps_gc_list) {
			/* if someone is waiting to be deleted, let it
			 * proceed as soon as possible. */
			if (atomic_read(&sec_gc_wait_del)) {
				CDEBUG(D_SEC, "deletion pending, start over\n");
				mutex_unlock(&sec_gc_mutex);
				goto again;
			}

			sec_do_gc(sec);
		}
		mutex_unlock(&sec_gc_mutex);

		/* check ctx list again before sleep */
		sec_process_ctx_list();

		lwi = LWI_TIMEOUT(SEC_GC_INTERVAL * HZ, NULL, NULL);
		l_wait_event(thread->t_ctl_waitq,
			     thread_is_stopping(thread) ||
			     thread_is_signal(thread),
			     &lwi);

		if (thread_test_and_clear_flags(thread, SVC_STOPPING))
			break;
	}

	thread_set_flags(thread, SVC_STOPPED);
	wake_up(&thread->t_ctl_waitq);
	return 0;
}

int sptlrpc_gc_init(void)
{
	struct l_wait_info lwi = { 0 };
	struct task_struct *task;

	mutex_init(&sec_gc_mutex);
	spin_lock_init(&sec_gc_list_lock);
	spin_lock_init(&sec_gc_ctx_list_lock);

	/* initialize thread control */
	memset(&sec_gc_thread, 0, sizeof(sec_gc_thread));
	init_waitqueue_head(&sec_gc_thread.t_ctl_waitq);

	task = kthread_run(sec_gc_main, &sec_gc_thread, "sptlrpc_gc");
	if (IS_ERR(task)) {
		CERROR("can't start gc thread: %ld\n", PTR_ERR(task));
		return PTR_ERR(task);
	}

	l_wait_event(sec_gc_thread.t_ctl_waitq,
		     thread_is_running(&sec_gc_thread), &lwi);
	return 0;
}

void sptlrpc_gc_fini(void)
{
	struct l_wait_info lwi = { 0 };

	thread_set_flags(&sec_gc_thread, SVC_STOPPING);
	wake_up(&sec_gc_thread.t_ctl_waitq);

	l_wait_event(sec_gc_thread.t_ctl_waitq,
		     thread_is_stopped(&sec_gc_thread), &lwi);
}
Commit	Line	Data
d7e09d03 PT	1	/*
	2	* GPL HEADER START
	3	*
	4	* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
	5	*
	6	* This program is free software; you can redistribute it and/or modify
	7	* it under the terms of the GNU General Public License version 2 only,
	8	* as published by the Free Software Foundation.
	9	*
	10	* This program is distributed in the hope that it will be useful, but
	11	* WITHOUT ANY WARRANTY; without even the implied warranty of
	12	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
	13	* General Public License version 2 for more details (a copy is included
	14	* in the LICENSE file that accompanied this code).
	15	*
	16	* You should have received a copy of the GNU General Public License
	17	* version 2 along with this program; If not, see
	18	* http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
	19	*
	20	* Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
	21	* CA 95054 USA or visit www.sun.com if you need additional information or
	22	* have any questions.
	23	*
	24	* GPL HEADER END
	25	*/
	26	/*
	27	* Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
	28	* Use is subject to license terms.
	29	*
	30	* Copyright (c) 2011, 2012, Intel Corporation.
	31	*/
	32	/*
	33	* This file is part of Lustre, http://www.lustre.org/
	34	* Lustre is a trademark of Sun Microsystems, Inc.
	35	*
	36	* lustre/ptlrpc/sec_gc.c
	37	*
	38	* Author: Eric Mei <ericm@clusterfs.com>
	39	*/
	40
	41	#define DEBUG_SUBSYSTEM S_SEC
	42
9fdaf8c0	43	#include "../../include/linux/libcfs/libcfs.h"
d7e09d03 PT	44
	45	#include <obd_support.h>
	46	#include <obd_class.h>
	47	#include <lustre_net.h>
	48	#include <lustre_sec.h>
	49
	50	#define SEC_GC_INTERVAL (30 * 60)
	51
	52
	53	static struct mutex sec_gc_mutex;
	54	static LIST_HEAD(sec_gc_list);
	55	static spinlock_t sec_gc_list_lock;
	56
	57	static LIST_HEAD(sec_gc_ctx_list);
	58	static spinlock_t sec_gc_ctx_list_lock;
	59
	60	static struct ptlrpc_thread sec_gc_thread;
	61	static atomic_t sec_gc_wait_del = ATOMIC_INIT(0);
	62
	63
	64	void sptlrpc_gc_add_sec(struct ptlrpc_sec *sec)
	65	{
	66	LASSERT(sec->ps_policy->sp_cops->gc_ctx);
	67	LASSERT(sec->ps_gc_interval > 0);
	68	LASSERT(list_empty(&sec->ps_gc_list));
	69
	70	sec->ps_gc_next = cfs_time_current_sec() + sec->ps_gc_interval;
	71
	72	spin_lock(&sec_gc_list_lock);
	73	list_add_tail(&sec_gc_list, &sec->ps_gc_list);
	74	spin_unlock(&sec_gc_list_lock);
	75
	76	CDEBUG(D_SEC, "added sec %p(%s)\n", sec, sec->ps_policy->sp_name);
	77	}
	78	EXPORT_SYMBOL(sptlrpc_gc_add_sec);
	79
	80	void sptlrpc_gc_del_sec(struct ptlrpc_sec *sec)
	81	{
	82	if (list_empty(&sec->ps_gc_list))
	83	return;
	84
	85	might_sleep();
	86
	87	/* signal before list_del to make iteration in gc thread safe */
	88	atomic_inc(&sec_gc_wait_del);
	89
	90	spin_lock(&sec_gc_list_lock);
	91	list_del_init(&sec->ps_gc_list);
	92	spin_unlock(&sec_gc_list_lock);
	93
	94	/* barrier */
	95	mutex_lock(&sec_gc_mutex);
	96	mutex_unlock(&sec_gc_mutex);
	97
	98	atomic_dec(&sec_gc_wait_del);
	99
	100	CDEBUG(D_SEC, "del sec %p(%s)\n", sec, sec->ps_policy->sp_name);
	101	}
	102	EXPORT_SYMBOL(sptlrpc_gc_del_sec);
	103
	104	void sptlrpc_gc_add_ctx(struct ptlrpc_cli_ctx *ctx)
	105	{
	106	LASSERT(list_empty(&ctx->cc_gc_chain));
	107
108	CDEBUG(D_SEC, "hand over ctx %p(%u->%s)\n",
109	ctx, ctx->cc_vcred.vc_uid, sec2target_str(ctx->cc_sec));
110	spin_lock(&sec_gc_ctx_list_lock);
111	list_add(&ctx->cc_gc_chain, &sec_gc_ctx_list);
112	spin_unlock(&sec_gc_ctx_list_lock);
113
114	thread_add_flags(&sec_gc_thread, SVC_SIGNAL);
115	wake_up(&sec_gc_thread.t_ctl_waitq);
116	}
117	EXPORT_SYMBOL(sptlrpc_gc_add_ctx);
118
119	static void sec_process_ctx_list(void)
120	{
121	struct ptlrpc_cli_ctx *ctx;
122
123	spin_lock(&sec_gc_ctx_list_lock);
124
125	while (!list_empty(&sec_gc_ctx_list)) {
126	ctx = list_entry(sec_gc_ctx_list.next,
127	struct ptlrpc_cli_ctx, cc_gc_chain);
128	list_del_init(&ctx->cc_gc_chain);
129	spin_unlock(&sec_gc_ctx_list_lock);
130
131	LASSERT(ctx->cc_sec);
132	LASSERT(atomic_read(&ctx->cc_refcount) == 1);
133	CDEBUG(D_SEC, "gc pick up ctx %p(%u->%s)\n",
134	ctx, ctx->cc_vcred.vc_uid, sec2target_str(ctx->cc_sec));
135	sptlrpc_cli_ctx_put(ctx, 1);
136
137	spin_lock(&sec_gc_ctx_list_lock);
138	}
139
140	spin_unlock(&sec_gc_ctx_list_lock);
141	}
142
143	static void sec_do_gc(struct ptlrpc_sec *sec)
144	{
145	LASSERT(sec->ps_policy->sp_cops->gc_ctx);
146
147	if (unlikely(sec->ps_gc_next == 0)) {
148	CDEBUG(D_SEC, "sec %p(%s) has 0 gc time\n",
149	sec, sec->ps_policy->sp_name);
150	return;
151	}
152
153	CDEBUG(D_SEC, "check on sec %p(%s)\n", sec, sec->ps_policy->sp_name);
154
155	if (cfs_time_after(sec->ps_gc_next, cfs_time_current_sec()))
156	return;
157
158	sec->ps_policy->sp_cops->gc_ctx(sec);
159	sec->ps_gc_next = cfs_time_current_sec() + sec->ps_gc_interval;
160	}
161
162	static int sec_gc_main(void *arg)
163	{
164	struct ptlrpc_thread thread = (struct ptlrpc_thread ) arg;
165	struct l_wait_info lwi;
166
167	unshare_fs_struct();
168
169	/* Record that the thread is running */
170	thread_set_flags(thread, SVC_RUNNING);
171	wake_up(&thread->t_ctl_waitq);
172
173	while (1) {
174	struct ptlrpc_sec *sec;
175
176	thread_clear_flags(thread, SVC_SIGNAL);
177	sec_process_ctx_list();
178	again:
179	/* go through sec list do gc.
180	* FIXME here we iterate through the whole list each time which
181	* is not optimal. we perhaps want to use balanced binary tree
182	* to trace each sec as order of expiry time.
183	* another issue here is we wakeup as fixed interval instead of
184	* according to each sec's expiry time */
185	mutex_lock(&sec_gc_mutex);
186	list_for_each_entry(sec, &sec_gc_list, ps_gc_list) {
187	/* if someone is waiting to be deleted, let it
188	* proceed as soon as possible. */
189	if (atomic_read(&sec_gc_wait_del)) {
190	CDEBUG(D_SEC, "deletion pending, start over\n");
191	mutex_unlock(&sec_gc_mutex);
192	goto again;
193	}
194
195	sec_do_gc(sec);
196	}
197	mutex_unlock(&sec_gc_mutex);
198
199	/* check ctx list again before sleep */
200	sec_process_ctx_list();
201
202	lwi = LWI_TIMEOUT(SEC_GC_INTERVAL * HZ, NULL, NULL);
203	l_wait_event(thread->t_ctl_waitq,
204	thread_is_stopping(thread) \|\|
205	thread_is_signal(thread),
206	&lwi);
207
208	if (thread_test_and_clear_flags(thread, SVC_STOPPING))
209	break;
210	}
211
212	thread_set_flags(thread, SVC_STOPPED);
213	wake_up(&thread->t_ctl_waitq);
214	return 0;
215	}
216
217	int sptlrpc_gc_init(void)
218	{
219	struct l_wait_info lwi = { 0 };
68b636b6	220	struct task_struct *task;
d7e09d03 PT	221
	222	mutex_init(&sec_gc_mutex);
	223	spin_lock_init(&sec_gc_list_lock);
	224	spin_lock_init(&sec_gc_ctx_list_lock);
	225
	226	/* initialize thread control */
	227	memset(&sec_gc_thread, 0, sizeof(sec_gc_thread));
	228	init_waitqueue_head(&sec_gc_thread.t_ctl_waitq);
	229
	230	task = kthread_run(sec_gc_main, &sec_gc_thread, "sptlrpc_gc");
	231	if (IS_ERR(task)) {
	232	CERROR("can't start gc thread: %ld\n", PTR_ERR(task));
	233	return PTR_ERR(task);
	234	}
	235
	236	l_wait_event(sec_gc_thread.t_ctl_waitq,
	237	thread_is_running(&sec_gc_thread), &lwi);
	238	return 0;
	239	}
	240
	241	void sptlrpc_gc_fini(void)
	242	{
	243	struct l_wait_info lwi = { 0 };
	244
	245	thread_set_flags(&sec_gc_thread, SVC_STOPPING);
	246	wake_up(&sec_gc_thread.t_ctl_waitq);
	247
	248	l_wait_event(sec_gc_thread.t_ctl_waitq,
	249	thread_is_stopped(&sec_gc_thread), &lwi);
	250	}