userdlm.c

ocfs1.2.7 源码

/* -*- mode: c; c-basic-offset: 8; -*-
 * vim: noexpandtab sw=8 ts=8 sts=0:
 *
 * userdlm.c
 *
 * Code which implements the kernel side of a minimal userspace
 * interface to our DLM.
 *
 * Many of the functions here are pared down versions of dlmglue.c
 * functions.
 *
 * Copyright (C) 2003, 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 <asm/signal.h>

#include <linux/module.h>
#include <linux/fs.h>
#include <linux/types.h>
#include <linux/crc32.h>


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

#include "dlmapi.h"

#include "userdlm.h"

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

static inline int user_check_wait_flag(struct user_lock_res *lockres,
				       int flag)
{
	int ret;

	spin_lock(&lockres->l_lock);
	ret = lockres->l_flags & flag;
	spin_unlock(&lockres->l_lock);

	return ret;
}

static inline void user_wait_on_busy_lock(struct user_lock_res *lockres)

{
	wait_event(lockres->l_event,
		   !user_check_wait_flag(lockres, USER_LOCK_BUSY));
}

static inline void user_wait_on_blocked_lock(struct user_lock_res *lockres)

{
	wait_event(lockres->l_event,
		   !user_check_wait_flag(lockres, USER_LOCK_BLOCKED));
}

/* I heart container_of... */
static inline struct dlm_ctxt *
dlm_ctxt_from_user_lockres(struct user_lock_res *lockres)
{
	struct dlmfs_inode_private *ip;

	ip = container_of(lockres,
			  struct dlmfs_inode_private,
			  ip_lockres);
	return ip->ip_dlm;
}

static struct inode *
user_dlm_inode_from_user_lockres(struct user_lock_res *lockres)
{
	struct dlmfs_inode_private *ip;

	ip = container_of(lockres,
			  struct dlmfs_inode_private,
			  ip_lockres);
	return &ip->ip_vfs_inode;
}

static inline void user_recover_from_dlm_error(struct user_lock_res *lockres)
{
	spin_lock(&lockres->l_lock);
	lockres->l_flags &= ~USER_LOCK_BUSY;
	spin_unlock(&lockres->l_lock);
}

#define user_log_dlm_error(_func, _stat, _lockres) do {		\
	mlog(ML_ERROR, "Dlm error \"%s\" while calling %s on "	\
		"resource %s: %s\n", dlm_errname(_stat), _func,	\
		_lockres->l_name, dlm_errmsg(_stat));		\
} while (0)

/* WARNING: This function lives in a world where the only three lock
 * levels are EX, PR, and NL. It *will* have to be adjusted when more
 * lock types are added. */
static inline int user_highest_compat_lock_level(int level)
{
	int new_level = LKM_EXMODE;

	if (level == LKM_EXMODE)
		new_level = LKM_NLMODE;
	else if (level == LKM_PRMODE)
		new_level = LKM_PRMODE;
	return new_level;
}

static void user_ast(void *opaque)
{
	struct user_lock_res *lockres = opaque;
	struct dlm_lockstatus *lksb;

	mlog(0, "AST fired for lockres %s\n", lockres->l_name);

	spin_lock(&lockres->l_lock);

	lksb = &(lockres->l_lksb);
	if (lksb->status != DLM_NORMAL) {
		mlog(ML_ERROR, "lksb status value of %u on lockres %s\n",
		     lksb->status, lockres->l_name);
		spin_unlock(&lockres->l_lock);
		return;
	}

	mlog_bug_on_msg(lockres->l_requested == LKM_IVMODE,
			"Lockres %s, requested ivmode. flags 0x%x\n",
			lockres->l_name, lockres->l_flags);

	/* we're downconverting. */
	if (lockres->l_requested < lockres->l_level) {
		if (lockres->l_requested <=
		    user_highest_compat_lock_level(lockres->l_blocking)) {
			lockres->l_blocking = LKM_NLMODE;
			lockres->l_flags &= ~USER_LOCK_BLOCKED;
		}
	}

	lockres->l_level = lockres->l_requested;
	lockres->l_requested = LKM_IVMODE;
	lockres->l_flags |= USER_LOCK_ATTACHED;
	lockres->l_flags &= ~USER_LOCK_BUSY;

	spin_unlock(&lockres->l_lock);

	wake_up(&lockres->l_event);
}

static inline void user_dlm_grab_inode_ref(struct user_lock_res *lockres)
{
	struct inode *inode;
	inode = user_dlm_inode_from_user_lockres(lockres);
	if (!igrab(inode))
		BUG();
}

static void user_dlm_unblock_lock(void *opaque);

static void __user_dlm_queue_lockres(struct user_lock_res *lockres)
{
	if (!(lockres->l_flags & USER_LOCK_QUEUED)) {
		user_dlm_grab_inode_ref(lockres);

		INIT_WORK(&lockres->l_work, user_dlm_unblock_lock,
			  lockres);

		queue_work(user_dlm_worker, &lockres->l_work);
		lockres->l_flags |= USER_LOCK_QUEUED;
	}
}

static void __user_dlm_cond_queue_lockres(struct user_lock_res *lockres)
{
	int queue = 0;

	if (!(lockres->l_flags & USER_LOCK_BLOCKED))
		return;

	switch (lockres->l_blocking) {
	case LKM_EXMODE:
		if (!lockres->l_ex_holders && !lockres->l_ro_holders)
			queue = 1;
		break;
	case LKM_PRMODE:
		if (!lockres->l_ex_holders)
			queue = 1;
		break;
	default:
		BUG();
	}

	if (queue)
		__user_dlm_queue_lockres(lockres);
}

static void user_bast(void *opaque, int level)
{
	struct user_lock_res *lockres = opaque;

	mlog(0, "Blocking AST fired for lockres %s. Blocking level %d\n",
		lockres->l_name, level);

	spin_lock(&lockres->l_lock);
	lockres->l_flags |= USER_LOCK_BLOCKED;
	if (level > lockres->l_blocking)
		lockres->l_blocking = level;

	__user_dlm_queue_lockres(lockres);
	spin_unlock(&lockres->l_lock);

	wake_up(&lockres->l_event);
}

static void user_unlock_ast(void *opaque, enum dlm_status status)
{
	struct user_lock_res *lockres = opaque;

	mlog(0, "UNLOCK AST called on lock %s\n", lockres->l_name);

	if (status != DLM_NORMAL && status != DLM_CANCELGRANT)
		mlog(ML_ERROR, "Dlm returns status %d\n", status);

	spin_lock(&lockres->l_lock);
	/* The teardown flag gets set early during the unlock process,
	 * so test the cancel flag to make sure that this ast isn't
	 * for a concurrent cancel. */
	if (lockres->l_flags & USER_LOCK_IN_TEARDOWN
	    && !(lockres->l_flags & USER_LOCK_IN_CANCEL)) {
		lockres->l_level = LKM_IVMODE;
	} else if (status == DLM_CANCELGRANT) {
		mlog(0, "Lock %s, cancel fails, flags 0x%x\n",
		     lockres->l_name, lockres->l_flags);
		/* We tried to cancel a convert request, but it was
		 * already granted. Don't clear the busy flag - the
		 * ast should've done this already. */
		BUG_ON(!(lockres->l_flags & USER_LOCK_IN_CANCEL));
		lockres->l_flags &= ~USER_LOCK_IN_CANCEL;
		goto out_noclear;
	} else {
		BUG_ON(!(lockres->l_flags & USER_LOCK_IN_CANCEL));
		/* Cancel succeeded, we want to re-queue */
		mlog(0, "Lock %s, cancel succeeds, flags 0x%x\n",
		     lockres->l_name, lockres->l_flags);
		lockres->l_requested = LKM_IVMODE; /* cancel an
						    * upconvert
						    * request. */
		lockres->l_flags &= ~USER_LOCK_IN_CANCEL;
		/* we want the unblock thread to look at it again
		 * now. */
		if (lockres->l_flags & USER_LOCK_BLOCKED)
			__user_dlm_queue_lockres(lockres);
	}

	lockres->l_flags &= ~USER_LOCK_BUSY;
out_noclear:
	spin_unlock(&lockres->l_lock);

	wake_up(&lockres->l_event);
}

static inline void user_dlm_drop_inode_ref(struct user_lock_res *lockres)
{
	struct inode *inode;
	inode = user_dlm_inode_from_user_lockres(lockres);
	iput(inode);
}

static void user_dlm_unblock_lock(void *opaque)
{
	int new_level, status;
	struct user_lock_res *lockres = (struct user_lock_res *) opaque;
	struct dlm_ctxt *dlm = dlm_ctxt_from_user_lockres(lockres);

	mlog(0, "processing lockres %s\n", lockres->l_name);

	spin_lock(&lockres->l_lock);

	mlog_bug_on_msg(!(lockres->l_flags & USER_LOCK_QUEUED),
			"Lockres %s, flags 0x%x\n",
			lockres->l_name, lockres->l_flags);

	/* notice that we don't clear USER_LOCK_BLOCKED here. If it's
	 * set, we want user_ast clear it. */
	lockres->l_flags &= ~USER_LOCK_QUEUED;

	/* It's valid to get here and no longer be blocked - if we get
	 * several basts in a row, we might be queued by the first
	 * one, the unblock thread might run and clear the queued
	 * flag, and finally we might get another bast which re-queues
	 * us before our ast for the downconvert is called. */
	if (!(lockres->l_flags & USER_LOCK_BLOCKED)) {
		mlog(0, "Lockres %s, flags 0x%x: queued but not blocking\n",
			lockres->l_name, lockres->l_flags);
		spin_unlock(&lockres->l_lock);
		goto drop_ref;
	}

	if (lockres->l_flags & USER_LOCK_IN_TEARDOWN) {
		mlog(0, "lock is in teardown so we do nothing\n");
		spin_unlock(&lockres->l_lock);
		goto drop_ref;
	}

	if (lockres->l_flags & USER_LOCK_BUSY) {
		mlog(0, "Cancel lock %s, flags 0x%x\n",
		     lockres->l_name, lockres->l_flags);

		if (lockres->l_flags & USER_LOCK_IN_CANCEL) {
			spin_unlock(&lockres->l_lock);
			goto drop_ref;
		}

		lockres->l_flags |= USER_LOCK_IN_CANCEL;
		spin_unlock(&lockres->l_lock);

		status = dlmunlock(dlm,
				   &lockres->l_lksb,
				   LKM_CANCEL,
				   user_unlock_ast,
				   lockres);
		if (status != DLM_NORMAL)
			user_log_dlm_error("dlmunlock", status, lockres);
		goto drop_ref;
	}

	/* If there are still incompat holders, we can exit safely
	 * without worrying about re-queueing this lock as that will
	 * happen on the last call to user_cluster_unlock. */
	if ((lockres->l_blocking == LKM_EXMODE)
	    && (lockres->l_ex_holders || lockres->l_ro_holders)) {
		spin_unlock(&lockres->l_lock);
		mlog(0, "can't downconvert for ex: ro = %u, ex = %u\n",
			lockres->l_ro_holders, lockres->l_ex_holders);