[GitHub/mt8127/android_kernel_alcatel_ttab.git] / fs / ocfs2 / dlm / dlmast.c

/* -*- mode: c; c-basic-offset: 8; -*-
 * vim: noexpandtab sw=8 ts=8 sts=0:
 *
 * dlmast.c
 *
 * AST and BAST functionality for local and remote nodes
 *
 * Copyright (C) 2004 Oracle.  All rights reserved.
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public
 * License as published by the Free Software Foundation; either
 * version 2 of the License, or (at your option) any later version.
 *
 * 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 for more details.
 *
 * You should have received a copy of the GNU General Public
 * License along with this program; if not, write to the
 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
 * Boston, MA 021110-1307, USA.
 *
 */


#include <linux/module.h>
#include <linux/fs.h>
#include <linux/types.h>
#include <linux/highmem.h>
#include <linux/init.h>
#include <linux/sysctl.h>
#include <linux/random.h>
#include <linux/blkdev.h>
#include <linux/socket.h>
#include <linux/inet.h>
#include <linux/spinlock.h>


#include "cluster/heartbeat.h"
#include "cluster/nodemanager.h"
#include "cluster/tcp.h"

#include "dlmapi.h"
#include "dlmcommon.h"

#define MLOG_MASK_PREFIX ML_DLM
#include "cluster/masklog.h"

static void dlm_update_lvb(struct dlm_ctxt *dlm, struct dlm_lock_resource *res,
			   struct dlm_lock *lock);
static int dlm_should_cancel_bast(struct dlm_ctxt *dlm, struct dlm_lock *lock);

/* Should be called as an ast gets queued to see if the new
 * lock level will obsolete a pending bast.
 * For example, if dlm_thread queued a bast for an EX lock that
 * was blocking another EX, but before sending the bast the
 * lock owner downconverted to NL, the bast is now obsolete.
 * Only the ast should be sent.
 * This is needed because the lock and convert paths can queue
 * asts out-of-band (not waiting for dlm_thread) in order to
 * allow for LKM_NOQUEUE to get immediate responses. */
static int dlm_should_cancel_bast(struct dlm_ctxt *dlm, struct dlm_lock *lock)
{
	assert_spin_locked(&dlm->ast_lock);
	assert_spin_locked(&lock->spinlock);

	if (lock->ml.highest_blocked == LKM_IVMODE)
		return 0;
	BUG_ON(lock->ml.highest_blocked == LKM_NLMODE);

	if (lock->bast_pending &&
	    list_empty(&lock->bast_list))
		/* old bast already sent, ok */
		return 0;

	if (lock->ml.type == LKM_EXMODE)
		/* EX blocks anything left, any bast still valid */
		return 0;
	else if (lock->ml.type == LKM_NLMODE)
		/* NL blocks nothing, no reason to send any bast, cancel it */
		return 1;
	else if (lock->ml.highest_blocked != LKM_EXMODE)
		/* PR only blocks EX */
		return 1;

	return 0;
}

void __dlm_queue_ast(struct dlm_ctxt *dlm, struct dlm_lock *lock)
{
	struct dlm_lock_resource *res;

	BUG_ON(!dlm);
	BUG_ON(!lock);

	res = lock->lockres;

	assert_spin_locked(&dlm->ast_lock);

	if (!list_empty(&lock->ast_list)) {
		mlog(ML_ERROR, "%s: res %.*s, lock %u:%llu, "
		     "AST list not empty, pending %d, newlevel %d\n",
		     dlm->name, res->lockname.len, res->lockname.name,
		     dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)),
		     dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)),
		     lock->ast_pending, lock->ml.type);
		BUG();
	}
	if (lock->ast_pending)
		mlog(0, "%s: res %.*s, lock %u:%llu, AST getting flushed\n",
		     dlm->name, res->lockname.len, res->lockname.name,
		     dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)),
		     dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)));

	/* putting lock on list, add a ref */
	dlm_lock_get(lock);
	spin_lock(&lock->spinlock);

	/* check to see if this ast obsoletes the bast */
	if (dlm_should_cancel_bast(dlm, lock)) {
		mlog(0, "%s: res %.*s, lock %u:%llu, Cancelling BAST\n",
		     dlm->name, res->lockname.len, res->lockname.name,
		     dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)),
		     dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)));
		lock->bast_pending = 0;
		list_del_init(&lock->bast_list);
		lock->ml.highest_blocked = LKM_IVMODE;
		/* removing lock from list, remove a ref.  guaranteed
		 * this won't be the last ref because of the get above,
		 * so res->spinlock will not be taken here */
		dlm_lock_put(lock);
		/* free up the reserved bast that we are cancelling.
		 * guaranteed that this will not be the last reserved
		 * ast because *both* an ast and a bast were reserved
		 * to get to this point.  the res->spinlock will not be
		 * taken here */
		dlm_lockres_release_ast(dlm, res);
	}
	list_add_tail(&lock->ast_list, &dlm->pending_asts);
	lock->ast_pending = 1;
	spin_unlock(&lock->spinlock);
}

void dlm_queue_ast(struct dlm_ctxt *dlm, struct dlm_lock *lock)
{
	BUG_ON(!dlm);
	BUG_ON(!lock);

	spin_lock(&dlm->ast_lock);
	__dlm_queue_ast(dlm, lock);
	spin_unlock(&dlm->ast_lock);
}


void __dlm_queue_bast(struct dlm_ctxt *dlm, struct dlm_lock *lock)
{
	struct dlm_lock_resource *res;

	BUG_ON(!dlm);
	BUG_ON(!lock);

	assert_spin_locked(&dlm->ast_lock);

	res = lock->lockres;

	BUG_ON(!list_empty(&lock->bast_list));
	if (lock->bast_pending)
		mlog(0, "%s: res %.*s, lock %u:%llu, BAST getting flushed\n",
		     dlm->name, res->lockname.len, res->lockname.name,
		     dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)),
		     dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)));

	/* putting lock on list, add a ref */
	dlm_lock_get(lock);
	spin_lock(&lock->spinlock);
	list_add_tail(&lock->bast_list, &dlm->pending_basts);
	lock->bast_pending = 1;
	spin_unlock(&lock->spinlock);
}

void dlm_queue_bast(struct dlm_ctxt *dlm, struct dlm_lock *lock)
{
	BUG_ON(!dlm);
	BUG_ON(!lock);

	spin_lock(&dlm->ast_lock);
	__dlm_queue_bast(dlm, lock);
	spin_unlock(&dlm->ast_lock);
}

static void dlm_update_lvb(struct dlm_ctxt *dlm, struct dlm_lock_resource *res,
			   struct dlm_lock *lock)
{
	struct dlm_lockstatus *lksb = lock->lksb;
	BUG_ON(!lksb);

	/* only updates if this node masters the lockres */
	spin_lock(&res->spinlock);
	if (res->owner == dlm->node_num) {
		/* check the lksb flags for the direction */
		if (lksb->flags & DLM_LKSB_GET_LVB) {
			mlog(0, "getting lvb from lockres for %s node\n",
				  lock->ml.node == dlm->node_num ? "master" :
				  "remote");
			memcpy(lksb->lvb, res->lvb, DLM_LVB_LEN);
		}
		/* Do nothing for lvb put requests - they should be done in
 		 * place when the lock is downconverted - otherwise we risk
 		 * racing gets and puts which could result in old lvb data
 		 * being propagated. We leave the put flag set and clear it
 		 * here. In the future we might want to clear it at the time
 		 * the put is actually done.
		 */
	}
	spin_unlock(&res->spinlock);

	/* reset any lvb flags on the lksb */
	lksb->flags &= ~(DLM_LKSB_PUT_LVB|DLM_LKSB_GET_LVB);
}

void dlm_do_local_ast(struct dlm_ctxt *dlm, struct dlm_lock_resource *res,
		      struct dlm_lock *lock)
{
	dlm_astlockfunc_t *fn;
	struct dlm_lockstatus *lksb;

	mlog(0, "%s: res %.*s, lock %u:%llu, Local AST\n", dlm->name,
	     res->lockname.len, res->lockname.name,
	     dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)),
	     dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)));

	lksb = lock->lksb;
	fn = lock->ast;
	BUG_ON(lock->ml.node != dlm->node_num);

	dlm_update_lvb(dlm, res, lock);
	(*fn)(lock->astdata);
}


int dlm_do_remote_ast(struct dlm_ctxt *dlm, struct dlm_lock_resource *res,
		      struct dlm_lock *lock)
{
	int ret;
	struct dlm_lockstatus *lksb;
	int lksbflags;

	mlog(0, "%s: res %.*s, lock %u:%llu, Remote AST\n", dlm->name,
	     res->lockname.len, res->lockname.name,
	     dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)),
	     dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)));

	lksb = lock->lksb;
	BUG_ON(lock->ml.node == dlm->node_num);

	lksbflags = lksb->flags;
	dlm_update_lvb(dlm, res, lock);

	/* lock request came from another node
	 * go do the ast over there */
	ret = dlm_send_proxy_ast(dlm, res, lock, lksbflags);
	return ret;
}

void dlm_do_local_bast(struct dlm_ctxt *dlm, struct dlm_lock_resource *res,
		       struct dlm_lock *lock, int blocked_type)
{
	dlm_bastlockfunc_t *fn = lock->bast;

	BUG_ON(lock->ml.node != dlm->node_num);

	mlog(0, "%s: res %.*s, lock %u:%llu, Local BAST, blocked %d\n",
	     dlm->name, res->lockname.len, res->lockname.name,
	     dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)),
	     dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)),
	     blocked_type);

	(*fn)(lock->astdata, blocked_type);
}


int dlm_proxy_ast_handler(struct o2net_msg *msg, u32 len, void *data,
			  void **ret_data)
{
	int ret;
	unsigned int locklen;
	struct dlm_ctxt *dlm = data;
	struct dlm_lock_resource *res = NULL;
	struct dlm_lock *lock = NULL;
	struct dlm_proxy_ast *past = (struct dlm_proxy_ast *) msg->buf;
	char *name;
	struct list_head *iter, *head=NULL;
	u64 cookie;
	u32 flags;
	u8 node;

	if (!dlm_grab(dlm)) {
		dlm_error(DLM_REJECTED);
		return DLM_REJECTED;
	}

	mlog_bug_on_msg(!dlm_domain_fully_joined(dlm),
			"Domain %s not fully joined!\n", dlm->name);

	name = past->name;
	locklen = past->namelen;
	cookie = past->cookie;
	flags = be32_to_cpu(past->flags);
	node = past->node_idx;

	if (locklen > DLM_LOCKID_NAME_MAX) {
		ret = DLM_IVBUFLEN;
		mlog(ML_ERROR, "Invalid name length (%d) in proxy ast "
		     "handler!\n", locklen);
		goto leave;
	}

	if ((flags & (LKM_PUT_LVB|LKM_GET_LVB)) ==
	     (LKM_PUT_LVB|LKM_GET_LVB)) {
		mlog(ML_ERROR, "Both PUT and GET lvb specified, (0x%x)\n",
		     flags);
		ret = DLM_BADARGS;
		goto leave;
	}

	mlog(0, "lvb: %s\n", flags & LKM_PUT_LVB ? "put lvb" :
		  (flags & LKM_GET_LVB ? "get lvb" : "none"));

	mlog(0, "type=%d, blocked_type=%d\n", past->type, past->blocked_type);

	if (past->type != DLM_AST &&
	    past->type != DLM_BAST) {
		mlog(ML_ERROR, "Unknown ast type! %d, cookie=%u:%llu"
		     "name=%.*s, node=%u\n", past->type,
		     dlm_get_lock_cookie_node(be64_to_cpu(cookie)),
		     dlm_get_lock_cookie_seq(be64_to_cpu(cookie)),
		     locklen, name, node);
		ret = DLM_IVLOCKID;
		goto leave;
	}

	res = dlm_lookup_lockres(dlm, name, locklen);
	if (!res) {
		mlog(0, "Got %sast for unknown lockres! cookie=%u:%llu, "
		     "name=%.*s, node=%u\n", (past->type == DLM_AST ? "" : "b"),
		     dlm_get_lock_cookie_node(be64_to_cpu(cookie)),
		     dlm_get_lock_cookie_seq(be64_to_cpu(cookie)),
		     locklen, name, node);
		ret = DLM_IVLOCKID;
		goto leave;
	}

	/* cannot get a proxy ast message if this node owns it */
	BUG_ON(res->owner == dlm->node_num);

	mlog(0, "%s: res %.*s\n", dlm->name, res->lockname.len,
	     res->lockname.name);

	spin_lock(&res->spinlock);
	if (res->state & DLM_LOCK_RES_RECOVERING) {
		mlog(0, "Responding with DLM_RECOVERING!\n");
		ret = DLM_RECOVERING;
		goto unlock_out;
	}
	if (res->state & DLM_LOCK_RES_MIGRATING) {
		mlog(0, "Responding with DLM_MIGRATING!\n");
		ret = DLM_MIGRATING;
		goto unlock_out;
	}
	/* try convert queue for both ast/bast */
	head = &res->converting;
	lock = NULL;
	list_for_each(iter, head) {
		lock = list_entry (iter, struct dlm_lock, list);
		if (lock->ml.cookie == cookie)
			goto do_ast;
	}

	/* if not on convert, try blocked for ast, granted for bast */
	if (past->type == DLM_AST)
		head = &res->blocked;
	else
		head = &res->granted;

	list_for_each(iter, head) {
		lock = list_entry (iter, struct dlm_lock, list);
		if (lock->ml.cookie == cookie)
			goto do_ast;
	}

	mlog(0, "Got %sast for unknown lock! cookie=%u:%llu, name=%.*s, "
	     "node=%u\n", past->type == DLM_AST ? "" : "b",
	     dlm_get_lock_cookie_node(be64_to_cpu(cookie)),
	     dlm_get_lock_cookie_seq(be64_to_cpu(cookie)),
	     locklen, name, node);

	ret = DLM_NORMAL;
unlock_out:
	spin_unlock(&res->spinlock);
	goto leave;

do_ast:
	ret = DLM_NORMAL;
	if (past->type == DLM_AST) {
		/* do not alter lock refcount.  switching lists. */
		list_move_tail(&lock->list, &res->granted);
		mlog(0, "%s: res %.*s, lock %u:%llu, Granted type %d => %d\n",
		     dlm->name, res->lockname.len, res->lockname.name,
		     dlm_get_lock_cookie_node(be64_to_cpu(cookie)),
		     dlm_get_lock_cookie_seq(be64_to_cpu(cookie)),
		     lock->ml.type, lock->ml.convert_type);

		if (lock->ml.convert_type != LKM_IVMODE) {
			lock->ml.type = lock->ml.convert_type;
			lock->ml.convert_type = LKM_IVMODE;
		} else {
			// should already be there....
		}

		lock->lksb->status = DLM_NORMAL;

		/* if we requested the lvb, fetch it into our lksb now */
		if (flags & LKM_GET_LVB) {
			BUG_ON(!(lock->lksb->flags & DLM_LKSB_GET_LVB));
			memcpy(lock->lksb->lvb, past->lvb, DLM_LVB_LEN);
		}
	}
	spin_unlock(&res->spinlock);

	if (past->type == DLM_AST)
		dlm_do_local_ast(dlm, res, lock);
	else
		dlm_do_local_bast(dlm, res, lock, past->blocked_type);

leave:
	if (res)
		dlm_lockres_put(res);

	dlm_put(dlm);
	return ret;
}


int dlm_send_proxy_ast_msg(struct dlm_ctxt *dlm, struct dlm_lock_resource *res,
			   struct dlm_lock *lock, int msg_type,
			   int blocked_type, int flags)
{
	int ret = 0;
	struct dlm_proxy_ast past;
	struct kvec vec[2];
	size_t veclen = 1;
	int status;

	mlog(0, "%s: res %.*s, to %u, type %d, blocked_type %d\n", dlm->name,
	     res->lockname.len, res->lockname.name, lock->ml.node, msg_type,
	     blocked_type);

	memset(&past, 0, sizeof(struct dlm_proxy_ast));
	past.node_idx = dlm->node_num;
	past.type = msg_type;
	past.blocked_type = blocked_type;
	past.namelen = res->lockname.len;
	memcpy(past.name, res->lockname.name, past.namelen);
	past.cookie = lock->ml.cookie;

	vec[0].iov_len = sizeof(struct dlm_proxy_ast);
	vec[0].iov_base = &past;
	if (flags & DLM_LKSB_GET_LVB) {
		be32_add_cpu(&past.flags, LKM_GET_LVB);
		vec[1].iov_len = DLM_LVB_LEN;
		vec[1].iov_base = lock->lksb->lvb;
		veclen++;
	}

	ret = o2net_send_message_vec(DLM_PROXY_AST_MSG, dlm->key, vec, veclen,
				     lock->ml.node, &status);
	if (ret < 0)
		mlog(ML_ERROR, "%s: res %.*s, error %d send AST to node %u\n",
		     dlm->name, res->lockname.len, res->lockname.name, ret,
		     lock->ml.node);
	else {
		if (status == DLM_RECOVERING) {
			mlog(ML_ERROR, "sent AST to node %u, it thinks this "
			     "node is dead!\n", lock->ml.node);
			BUG();
		} else if (status == DLM_MIGRATING) {
			mlog(ML_ERROR, "sent AST to node %u, it returned "
			     "DLM_MIGRATING!\n", lock->ml.node);
			BUG();
		} else if (status != DLM_NORMAL && status != DLM_IVLOCKID) {
			mlog(ML_ERROR, "AST to node %u returned %d!\n",
			     lock->ml.node, status);
			/* ignore it */
		}
		ret = 0;
	}
	return ret;
}
Commit	Line	Data
6714d8e8 KH	1	/* -- mode: c; c-basic-offset: 8; --
	2	* vim: noexpandtab sw=8 ts=8 sts=0:
	3	*
	4	* dlmast.c
	5	*
	6	* AST and BAST functionality for local and remote nodes
	7	*
	8	* Copyright (C) 2004 Oracle. All rights reserved.
	9	*
	10	* This program is free software; you can redistribute it and/or
	11	* modify it under the terms of the GNU General Public
	12	* License as published by the Free Software Foundation; either
	13	* version 2 of the License, or (at your option) any later version.
	14	*
	15	* This program is distributed in the hope that it will be useful,
	16	* but WITHOUT ANY WARRANTY; without even the implied warranty of
	17	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
	18	* General Public License for more details.
	19	*
	20	* You should have received a copy of the GNU General Public
	21	* License along with this program; if not, write to the
	22	* Free Software Foundation, Inc., 59 Temple Place - Suite 330,
	23	* Boston, MA 021110-1307, USA.
	24	*
	25	*/
	26
	27
	28	#include <linux/module.h>
	29	#include <linux/fs.h>
	30	#include <linux/types.h>
6714d8e8	31	#include <linux/highmem.h>
6714d8e8 KH	32	#include <linux/init.h>
	33	#include <linux/sysctl.h>
	34	#include <linux/random.h>
	35	#include <linux/blkdev.h>
	36	#include <linux/socket.h>
	37	#include <linux/inet.h>
	38	#include <linux/spinlock.h>
	39
	40
	41	#include "cluster/heartbeat.h"
	42	#include "cluster/nodemanager.h"
	43	#include "cluster/tcp.h"
6714d8e8 KH	44
	45	#include "dlmapi.h"
	46	#include "dlmcommon.h"
	47
	48	#define MLOG_MASK_PREFIX ML_DLM
	49	#include "cluster/masklog.h"
	50
	51	static void dlm_update_lvb(struct dlm_ctxt dlm, struct dlm_lock_resource res,
	52	struct dlm_lock *lock);
	53	static int dlm_should_cancel_bast(struct dlm_ctxt dlm, struct dlm_lock lock);
	54
	55	/* Should be called as an ast gets queued to see if the new
	56	* lock level will obsolete a pending bast.
	57	* For example, if dlm_thread queued a bast for an EX lock that
	58	* was blocking another EX, but before sending the bast the
	59	* lock owner downconverted to NL, the bast is now obsolete.
	60	* Only the ast should be sent.
	61	* This is needed because the lock and convert paths can queue
	62	* asts out-of-band (not waiting for dlm_thread) in order to
	63	* allow for LKM_NOQUEUE to get immediate responses. */
	64	static int dlm_should_cancel_bast(struct dlm_ctxt dlm, struct dlm_lock lock)
	65	{
	66	assert_spin_locked(&dlm->ast_lock);
	67	assert_spin_locked(&lock->spinlock);
	68
	69	if (lock->ml.highest_blocked == LKM_IVMODE)
	70	return 0;
	71	BUG_ON(lock->ml.highest_blocked == LKM_NLMODE);
	72
	73	if (lock->bast_pending &&
	74	list_empty(&lock->bast_list))
	75	/* old bast already sent, ok */
	76	return 0;
	77
	78	if (lock->ml.type == LKM_EXMODE)
	79	/* EX blocks anything left, any bast still valid */
	80	return 0;
	81	else if (lock->ml.type == LKM_NLMODE)
	82	/* NL blocks nothing, no reason to send any bast, cancel it */
	83	return 1;
	84	else if (lock->ml.highest_blocked != LKM_EXMODE)
	85	/* PR only blocks EX */
	86	return 1;
	87
	88	return 0;
	89	}
	90
d9ef7522	91	void __dlm_queue_ast(struct dlm_ctxt dlm, struct dlm_lock lock)
6714d8e8	92	{
8e17d16f	93	struct dlm_lock_resource *res;
6714d8e8 KH	94
	95	BUG_ON(!dlm);
	96	BUG_ON(!lock);
	97
8e17d16f SM	98	res = lock->lockres;
8e17d16f SM	99
6714d8e8	100	assert_spin_locked(&dlm->ast_lock);
8e17d16f	101
6714d8e8	102	if (!list_empty(&lock->ast_list)) {
8e17d16f SM	103	mlog(ML_ERROR, "%s: res %.*s, lock %u:%llu, "
	104	"AST list not empty, pending %d, newlevel %d\n",
	105	dlm->name, res->lockname.len, res->lockname.name,
	106	dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)),
	107	dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)),
6714d8e8 KH	108	lock->ast_pending, lock->ml.type);
	109	BUG();
	110	}
6714d8e8	111	if (lock->ast_pending)
8e17d16f SM	112	mlog(0, "%s: res %.*s, lock %u:%llu, AST getting flushed\n",
	113	dlm->name, res->lockname.len, res->lockname.name,
	114	dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)),
	115	dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)));
6714d8e8 KH	116
	117	/* putting lock on list, add a ref */
	118	dlm_lock_get(lock);
	119	spin_lock(&lock->spinlock);
	120
	121	/* check to see if this ast obsoletes the bast */
	122	if (dlm_should_cancel_bast(dlm, lock)) {
8e17d16f SM	123	mlog(0, "%s: res %.*s, lock %u:%llu, Cancelling BAST\n",
	124	dlm->name, res->lockname.len, res->lockname.name,
	125	dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)),
	126	dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)));
6714d8e8 KH	127	lock->bast_pending = 0;
	128	list_del_init(&lock->bast_list);
	129	lock->ml.highest_blocked = LKM_IVMODE;
	130	/* removing lock from list, remove a ref. guaranteed
	131	* this won't be the last ref because of the get above,
	132	* so res->spinlock will not be taken here */
	133	dlm_lock_put(lock);
	134	/* free up the reserved bast that we are cancelling.
	135	* guaranteed that this will not be the last reserved
2bd63216	136	* ast because both an ast and a bast were reserved
6714d8e8 KH	137	* to get to this point. the res->spinlock will not be
	138	* taken here */
	139	dlm_lockres_release_ast(dlm, res);
	140	}
	141	list_add_tail(&lock->ast_list, &dlm->pending_asts);
	142	lock->ast_pending = 1;
	143	spin_unlock(&lock->spinlock);
	144	}
	145
	146	void dlm_queue_ast(struct dlm_ctxt dlm, struct dlm_lock lock)
	147	{
6714d8e8 KH	148	BUG_ON(!dlm);
	149	BUG_ON(!lock);
	150
	151	spin_lock(&dlm->ast_lock);
	152	__dlm_queue_ast(dlm, lock);
	153	spin_unlock(&dlm->ast_lock);
	154	}
	155
	156
d9ef7522	157	void __dlm_queue_bast(struct dlm_ctxt dlm, struct dlm_lock lock)
6714d8e8	158	{
8e17d16f	159	struct dlm_lock_resource *res;
6714d8e8 KH	160
	161	BUG_ON(!dlm);
	162	BUG_ON(!lock);
8e17d16f	163
6714d8e8 KH	164	assert_spin_locked(&dlm->ast_lock);
6714d8e8 KH	165
8e17d16f SM	166	res = lock->lockres;
8e17d16f SM	167
6714d8e8 KH	168	BUG_ON(!list_empty(&lock->bast_list));
6714d8e8 KH	169	if (lock->bast_pending)
8e17d16f SM	170	mlog(0, "%s: res %.*s, lock %u:%llu, BAST getting flushed\n",
	171	dlm->name, res->lockname.len, res->lockname.name,
	172	dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)),
	173	dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)));
6714d8e8 KH	174
	175	/* putting lock on list, add a ref */
	176	dlm_lock_get(lock);
	177	spin_lock(&lock->spinlock);
	178	list_add_tail(&lock->bast_list, &dlm->pending_basts);
	179	lock->bast_pending = 1;
	180	spin_unlock(&lock->spinlock);
	181	}
	182
	183	void dlm_queue_bast(struct dlm_ctxt dlm, struct dlm_lock lock)
	184	{
6714d8e8 KH	185	BUG_ON(!dlm);
	186	BUG_ON(!lock);
	187
	188	spin_lock(&dlm->ast_lock);
	189	__dlm_queue_bast(dlm, lock);
	190	spin_unlock(&dlm->ast_lock);
	191	}
	192
	193	static void dlm_update_lvb(struct dlm_ctxt dlm, struct dlm_lock_resource res,
	194	struct dlm_lock *lock)
	195	{
	196	struct dlm_lockstatus *lksb = lock->lksb;
	197	BUG_ON(!lksb);
	198
	199	/* only updates if this node masters the lockres */
428257f8	200	spin_lock(&res->spinlock);
6714d8e8	201	if (res->owner == dlm->node_num) {
6714d8e8 KH	202	/* check the lksb flags for the direction */
	203	if (lksb->flags & DLM_LKSB_GET_LVB) {
	204	mlog(0, "getting lvb from lockres for %s node\n",
	205	lock->ml.node == dlm->node_num ? "master" :
	206	"remote");
	207	memcpy(lksb->lvb, res->lvb, DLM_LVB_LEN);
6714d8e8	208	}
c0a8520c MF	209	/* Do nothing for lvb put requests - they should be done in
	210	* place when the lock is downconverted - otherwise we risk
	211	* racing gets and puts which could result in old lvb data
	212	* being propagated. We leave the put flag set and clear it
	213	* here. In the future we might want to clear it at the time
	214	* the put is actually done.
	215	*/
6714d8e8	216	}
428257f8	217	spin_unlock(&res->spinlock);
6714d8e8 KH	218
	219	/* reset any lvb flags on the lksb */
	220	lksb->flags &= ~(DLM_LKSB_PUT_LVB\|DLM_LKSB_GET_LVB);
	221	}
	222
	223	void dlm_do_local_ast(struct dlm_ctxt dlm, struct dlm_lock_resource res,
	224	struct dlm_lock *lock)
	225	{
	226	dlm_astlockfunc_t *fn;
	227	struct dlm_lockstatus *lksb;
	228
8e17d16f SM	229	mlog(0, "%s: res %.*s, lock %u:%llu, Local AST\n", dlm->name,
	230	res->lockname.len, res->lockname.name,
	231	dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)),
	232	dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)));
6714d8e8 KH	233
	234	lksb = lock->lksb;
	235	fn = lock->ast;
	236	BUG_ON(lock->ml.node != dlm->node_num);
	237
	238	dlm_update_lvb(dlm, res, lock);
	239	(*fn)(lock->astdata);
	240	}
	241
	242
	243	int dlm_do_remote_ast(struct dlm_ctxt dlm, struct dlm_lock_resource res,
	244	struct dlm_lock *lock)
	245	{
	246	int ret;
	247	struct dlm_lockstatus *lksb;
	248	int lksbflags;
	249
8e17d16f SM	250	mlog(0, "%s: res %.*s, lock %u:%llu, Remote AST\n", dlm->name,
	251	res->lockname.len, res->lockname.name,
	252	dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)),
	253	dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)));
6714d8e8 KH	254
	255	lksb = lock->lksb;
	256	BUG_ON(lock->ml.node == dlm->node_num);
	257
	258	lksbflags = lksb->flags;
	259	dlm_update_lvb(dlm, res, lock);
	260
	261	/* lock request came from another node
	262	* go do the ast over there */
	263	ret = dlm_send_proxy_ast(dlm, res, lock, lksbflags);
	264	return ret;
	265	}
	266
	267	void dlm_do_local_bast(struct dlm_ctxt dlm, struct dlm_lock_resource res,
	268	struct dlm_lock *lock, int blocked_type)
	269	{
	270	dlm_bastlockfunc_t *fn = lock->bast;
	271
6714d8e8 KH	272	BUG_ON(lock->ml.node != dlm->node_num);
6714d8e8 KH	273
8e17d16f SM	274	mlog(0, "%s: res %.*s, lock %u:%llu, Local BAST, blocked %d\n",
	275	dlm->name, res->lockname.len, res->lockname.name,
	276	dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)),
	277	dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)),
	278	blocked_type);
	279
6714d8e8 KH	280	(*fn)(lock->astdata, blocked_type);
	281	}
	282
	283
	284
d74c9803 KH	285	int dlm_proxy_ast_handler(struct o2net_msg msg, u32 len, void data,
d74c9803 KH	286	void **ret_data)
6714d8e8 KH	287	{
	288	int ret;
	289	unsigned int locklen;
	290	struct dlm_ctxt *dlm = data;
	291	struct dlm_lock_resource *res = NULL;
	292	struct dlm_lock *lock = NULL;
	293	struct dlm_proxy_ast past = (struct dlm_proxy_ast ) msg->buf;
	294	char *name;
	295	struct list_head iter, head=NULL;
	296	u64 cookie;
	297	u32 flags;
57dff267	298	u8 node;
6714d8e8 KH	299
	300	if (!dlm_grab(dlm)) {
	301	dlm_error(DLM_REJECTED);
	302	return DLM_REJECTED;
	303	}
	304
	305	mlog_bug_on_msg(!dlm_domain_fully_joined(dlm),
	306	"Domain %s not fully joined!\n", dlm->name);
	307
	308	name = past->name;
	309	locklen = past->namelen;
57dff267	310	cookie = past->cookie;
6714d8e8	311	flags = be32_to_cpu(past->flags);
57dff267	312	node = past->node_idx;
6714d8e8 KH	313
	314	if (locklen > DLM_LOCKID_NAME_MAX) {
	315	ret = DLM_IVBUFLEN;
57dff267 SM	316	mlog(ML_ERROR, "Invalid name length (%d) in proxy ast "
57dff267 SM	317	"handler!\n", locklen);
6714d8e8 KH	318	goto leave;
	319	}
	320
	321	if ((flags & (LKM_PUT_LVB\|LKM_GET_LVB)) ==
	322	(LKM_PUT_LVB\|LKM_GET_LVB)) {
57dff267 SM	323	mlog(ML_ERROR, "Both PUT and GET lvb specified, (0x%x)\n",
57dff267 SM	324	flags);
6714d8e8 KH	325	ret = DLM_BADARGS;
	326	goto leave;
	327	}
	328
	329	mlog(0, "lvb: %s\n", flags & LKM_PUT_LVB ? "put lvb" :
	330	(flags & LKM_GET_LVB ? "get lvb" : "none"));
	331
	332	mlog(0, "type=%d, blocked_type=%d\n", past->type, past->blocked_type);
	333
	334	if (past->type != DLM_AST &&
	335	past->type != DLM_BAST) {
29004858	336	mlog(ML_ERROR, "Unknown ast type! %d, cookie=%u:%llu"
57dff267 SM	337	"name=%.*s, node=%u\n", past->type,
	338	dlm_get_lock_cookie_node(be64_to_cpu(cookie)),
	339	dlm_get_lock_cookie_seq(be64_to_cpu(cookie)),
	340	locklen, name, node);
6714d8e8 KH	341	ret = DLM_IVLOCKID;
	342	goto leave;
	343	}
	344
	345	res = dlm_lookup_lockres(dlm, name, locklen);
	346	if (!res) {
57dff267 SM	347	mlog(0, "Got %sast for unknown lockres! cookie=%u:%llu, "
	348	"name=%.*s, node=%u\n", (past->type == DLM_AST ? "" : "b"),
	349	dlm_get_lock_cookie_node(be64_to_cpu(cookie)),
	350	dlm_get_lock_cookie_seq(be64_to_cpu(cookie)),
	351	locklen, name, node);
6714d8e8 KH	352	ret = DLM_IVLOCKID;
	353	goto leave;
	354	}
	355
	356	/* cannot get a proxy ast message if this node owns it */
	357	BUG_ON(res->owner == dlm->node_num);
	358
8e17d16f SM	359	mlog(0, "%s: res %.*s\n", dlm->name, res->lockname.len,
8e17d16f SM	360	res->lockname.name);
6714d8e8 KH	361
	362	spin_lock(&res->spinlock);
	363	if (res->state & DLM_LOCK_RES_RECOVERING) {
57dff267	364	mlog(0, "Responding with DLM_RECOVERING!\n");
6714d8e8 KH	365	ret = DLM_RECOVERING;
	366	goto unlock_out;
	367	}
	368	if (res->state & DLM_LOCK_RES_MIGRATING) {
57dff267	369	mlog(0, "Responding with DLM_MIGRATING!\n");
6714d8e8 KH	370	ret = DLM_MIGRATING;
	371	goto unlock_out;
	372	}
	373	/* try convert queue for both ast/bast */
	374	head = &res->converting;
	375	lock = NULL;
	376	list_for_each(iter, head) {
	377	lock = list_entry (iter, struct dlm_lock, list);
57dff267	378	if (lock->ml.cookie == cookie)
6714d8e8 KH	379	goto do_ast;
	380	}
	381
	382	/* if not on convert, try blocked for ast, granted for bast */
	383	if (past->type == DLM_AST)
	384	head = &res->blocked;
	385	else
	386	head = &res->granted;
	387
	388	list_for_each(iter, head) {
	389	lock = list_entry (iter, struct dlm_lock, list);
57dff267	390	if (lock->ml.cookie == cookie)
6714d8e8 KH	391	goto do_ast;
	392	}
	393
57dff267 SM	394	mlog(0, "Got %sast for unknown lock! cookie=%u:%llu, name=%.*s, "
	395	"node=%u\n", past->type == DLM_AST ? "" : "b",
	396	dlm_get_lock_cookie_node(be64_to_cpu(cookie)),
	397	dlm_get_lock_cookie_seq(be64_to_cpu(cookie)),
	398	locklen, name, node);
6714d8e8 KH	399
	400	ret = DLM_NORMAL;
	401	unlock_out:
	402	spin_unlock(&res->spinlock);
	403	goto leave;
	404
	405	do_ast:
	406	ret = DLM_NORMAL;
	407	if (past->type == DLM_AST) {
	408	/* do not alter lock refcount. switching lists. */
f116629d	409	list_move_tail(&lock->list, &res->granted);
8e17d16f SM	410	mlog(0, "%s: res %.*s, lock %u:%llu, Granted type %d => %d\n",
	411	dlm->name, res->lockname.len, res->lockname.name,
	412	dlm_get_lock_cookie_node(be64_to_cpu(cookie)),
	413	dlm_get_lock_cookie_seq(be64_to_cpu(cookie)),
	414	lock->ml.type, lock->ml.convert_type);
	415
6714d8e8 KH	416	if (lock->ml.convert_type != LKM_IVMODE) {
	417	lock->ml.type = lock->ml.convert_type;
	418	lock->ml.convert_type = LKM_IVMODE;
	419	} else {
	420	// should already be there....
	421	}
	422
	423	lock->lksb->status = DLM_NORMAL;
	424
	425	/* if we requested the lvb, fetch it into our lksb now */
	426	if (flags & LKM_GET_LVB) {
	427	BUG_ON(!(lock->lksb->flags & DLM_LKSB_GET_LVB));
	428	memcpy(lock->lksb->lvb, past->lvb, DLM_LVB_LEN);
	429	}
	430	}
	431	spin_unlock(&res->spinlock);
	432
	433	if (past->type == DLM_AST)
	434	dlm_do_local_ast(dlm, res, lock);
	435	else
	436	dlm_do_local_bast(dlm, res, lock, past->blocked_type);
	437
	438	leave:
6714d8e8 KH	439	if (res)
	440	dlm_lockres_put(res);
	441
	442	dlm_put(dlm);
	443	return ret;
	444	}
	445
	446
	447
	448	int dlm_send_proxy_ast_msg(struct dlm_ctxt dlm, struct dlm_lock_resource res,
	449	struct dlm_lock *lock, int msg_type,
	450	int blocked_type, int flags)
	451	{
	452	int ret = 0;
	453	struct dlm_proxy_ast past;
	454	struct kvec vec[2];
	455	size_t veclen = 1;
	456	int status;
	457
8e17d16f SM	458	mlog(0, "%s: res %.*s, to %u, type %d, blocked_type %d\n", dlm->name,
	459	res->lockname.len, res->lockname.name, lock->ml.node, msg_type,
	460	blocked_type);
6714d8e8 KH	461
	462	memset(&past, 0, sizeof(struct dlm_proxy_ast));
	463	past.node_idx = dlm->node_num;
	464	past.type = msg_type;
	465	past.blocked_type = blocked_type;
	466	past.namelen = res->lockname.len;
	467	memcpy(past.name, res->lockname.name, past.namelen);
	468	past.cookie = lock->ml.cookie;
	469
	470	vec[0].iov_len = sizeof(struct dlm_proxy_ast);
	471	vec[0].iov_base = &past;
	472	if (flags & DLM_LKSB_GET_LVB) {
6714d8e8 KH	473	be32_add_cpu(&past.flags, LKM_GET_LVB);
	474	vec[1].iov_len = DLM_LVB_LEN;
	475	vec[1].iov_base = lock->lksb->lvb;
	476	veclen++;
	477	}
	478
	479	ret = o2net_send_message_vec(DLM_PROXY_AST_MSG, dlm->key, vec, veclen,
	480	lock->ml.node, &status);
	481	if (ret < 0)
8e17d16f SM	482	mlog(ML_ERROR, "%s: res %.*s, error %d send AST to node %u\n",
8e17d16f SM	483	dlm->name, res->lockname.len, res->lockname.name, ret,
a5196ec5	484	lock->ml.node);
6714d8e8 KH	485	else {
	486	if (status == DLM_RECOVERING) {
	487	mlog(ML_ERROR, "sent AST to node %u, it thinks this "
	488	"node is dead!\n", lock->ml.node);
	489	BUG();
	490	} else if (status == DLM_MIGRATING) {
	491	mlog(ML_ERROR, "sent AST to node %u, it returned "
	492	"DLM_MIGRATING!\n", lock->ml.node);
	493	BUG();
e2c73698	494	} else if (status != DLM_NORMAL && status != DLM_IVLOCKID) {
6714d8e8 KH	495	mlog(ML_ERROR, "AST to node %u returned %d!\n",
	496	lock->ml.node, status);
	497	/* ignore it */
	498	}
	499	ret = 0;
	500	}
	501	return ret;
	502	}