dlmglue.c

ocfs1.4.1 oracle分布式文件系统

				       struct ocfs2_lock_res *res,
				       enum ocfs2_lock_type type,
				       struct ocfs2_lock_res_ops *ops,
				       void *priv)
{
	res->l_type          = type;
	res->l_ops           = ops;
	res->l_priv          = priv;

	res->l_level         = LKM_IVMODE;
	res->l_requested     = LKM_IVMODE;
	res->l_blocking      = LKM_IVMODE;
	res->l_action        = OCFS2_AST_INVALID;
	res->l_unlock_action = OCFS2_UNLOCK_INVALID;

	res->l_flags         = OCFS2_LOCK_INITIALIZED;

	ocfs2_add_lockres_tracking(res, osb->osb_dlm_debug);

	ocfs2_init_lock_stats(res);
}

void ocfs2_lock_res_init_once(struct ocfs2_lock_res *res)
{
	/* This also clears out the lock status block */
	memset(res, 0, sizeof(struct ocfs2_lock_res));
	spin_lock_init(&res->l_lock);
	init_waitqueue_head(&res->l_event);
	INIT_LIST_HEAD(&res->l_blocked_list);
	INIT_LIST_HEAD(&res->l_mask_waiters);
}

void ocfs2_inode_lock_res_init(struct ocfs2_lock_res *res,
			       enum ocfs2_lock_type type,
			       unsigned int generation,
			       struct inode *inode)
{
	struct ocfs2_lock_res_ops *ops;

	switch(type) {
		case OCFS2_LOCK_TYPE_RW:
			ops = &ocfs2_inode_rw_lops;
			break;
		case OCFS2_LOCK_TYPE_META:
			ops = &ocfs2_inode_inode_lops;
			break;
		case OCFS2_LOCK_TYPE_OPEN:
			ops = &ocfs2_inode_open_lops;
			break;
		default:
			mlog_bug_on_msg(1, "type: %d\n", type);
			ops = NULL; /* thanks, gcc */
			break;
	};

	ocfs2_build_lock_name(type, OCFS2_I(inode)->ip_blkno,
			      generation, res->l_name);
	ocfs2_lock_res_init_common(OCFS2_SB(inode->i_sb), res, type, ops, inode);
}

static struct ocfs2_super *ocfs2_get_inode_osb(struct ocfs2_lock_res *lockres)
{
	struct inode *inode = ocfs2_lock_res_inode(lockres);

	return OCFS2_SB(inode->i_sb);
}

static struct ocfs2_super *ocfs2_get_file_osb(struct ocfs2_lock_res *lockres)
{
	struct ocfs2_file_private *fp = lockres->l_priv;

	return OCFS2_SB(fp->fp_file->f_mapping->host->i_sb);
}

static __u64 ocfs2_get_dentry_lock_ino(struct ocfs2_lock_res *lockres)
{
	__be64 inode_blkno_be;

	memcpy(&inode_blkno_be, &lockres->l_name[OCFS2_DENTRY_LOCK_INO_START],
	       sizeof(__be64));

	return be64_to_cpu(inode_blkno_be);
}

static struct ocfs2_super *ocfs2_get_dentry_osb(struct ocfs2_lock_res *lockres)
{
	struct ocfs2_dentry_lock *dl = lockres->l_priv;

	return OCFS2_SB(dl->dl_inode->i_sb);
}

void ocfs2_dentry_lock_res_init(struct ocfs2_dentry_lock *dl,
				u64 parent, struct inode *inode)
{
	int len;
	u64 inode_blkno = OCFS2_I(inode)->ip_blkno;
	__be64 inode_blkno_be = cpu_to_be64(inode_blkno);
	struct ocfs2_lock_res *lockres = &dl->dl_lockres;

	ocfs2_lock_res_init_once(lockres);

	/*
	 * Unfortunately, the standard lock naming scheme won't work
	 * here because we have two 16 byte values to use. Instead,
	 * we'll stuff the inode number as a binary value. We still
	 * want error prints to show something without garbling the
	 * display, so drop a null byte in there before the inode
	 * number. A future version of OCFS2 will likely use all
	 * binary lock names. The stringified names have been a
	 * tremendous aid in debugging, but now that the debugfs
	 * interface exists, we can mangle things there if need be.
	 *
	 * NOTE: We also drop the standard "pad" value (the total lock
	 * name size stays the same though - the last part is all
	 * zeros due to the memset in ocfs2_lock_res_init_once()
	 */
	len = snprintf(lockres->l_name, OCFS2_DENTRY_LOCK_INO_START,
		       "%c%016llx",
		       ocfs2_lock_type_char(OCFS2_LOCK_TYPE_DENTRY),
		       (long long)parent);

	BUG_ON(len != (OCFS2_DENTRY_LOCK_INO_START - 1));

	memcpy(&lockres->l_name[OCFS2_DENTRY_LOCK_INO_START], &inode_blkno_be,
	       sizeof(__be64));

	ocfs2_lock_res_init_common(OCFS2_SB(inode->i_sb), lockres,
				   OCFS2_LOCK_TYPE_DENTRY, &ocfs2_dentry_lops,
				   dl);
}

static void ocfs2_super_lock_res_init(struct ocfs2_lock_res *res,
				      struct ocfs2_super *osb)
{
	/* Superblock lockres doesn't come from a slab so we call init
	 * once on it manually.  */
	ocfs2_lock_res_init_once(res);
	ocfs2_build_lock_name(OCFS2_LOCK_TYPE_SUPER, OCFS2_SUPER_BLOCK_BLKNO,
			      0, res->l_name);
	ocfs2_lock_res_init_common(osb, res, OCFS2_LOCK_TYPE_SUPER,
				   &ocfs2_super_lops, osb);
}

static void ocfs2_rename_lock_res_init(struct ocfs2_lock_res *res,
				       struct ocfs2_super *osb)
{
	/* Rename lockres doesn't come from a slab so we call init
	 * once on it manually.  */
	ocfs2_lock_res_init_once(res);
	ocfs2_build_lock_name(OCFS2_LOCK_TYPE_RENAME, 0, 0, res->l_name);
	ocfs2_lock_res_init_common(osb, res, OCFS2_LOCK_TYPE_RENAME,
				   &ocfs2_rename_lops, osb);
}

void ocfs2_file_lock_res_init(struct ocfs2_lock_res *lockres,
			      struct ocfs2_file_private *fp)
{
	struct inode *inode = fp->fp_file->f_mapping->host;
	struct ocfs2_inode_info *oi = OCFS2_I(inode);

	ocfs2_lock_res_init_once(lockres);
	ocfs2_build_lock_name(OCFS2_LOCK_TYPE_FLOCK, oi->ip_blkno,
			      inode->i_generation, lockres->l_name);
	ocfs2_lock_res_init_common(OCFS2_SB(inode->i_sb), lockres,
				   OCFS2_LOCK_TYPE_FLOCK, &ocfs2_flock_lops,
				   fp);
	lockres->l_flags |= OCFS2_LOCK_NOCACHE;
}

void ocfs2_lock_res_free(struct ocfs2_lock_res *res)
{
	mlog_entry_void();

	if (!(res->l_flags & OCFS2_LOCK_INITIALIZED))
		return;

	ocfs2_remove_lockres_tracking(res);

	mlog_bug_on_msg(!list_empty(&res->l_blocked_list),
			"Lockres %s is on the blocked list\n",
			res->l_name);
	mlog_bug_on_msg(!list_empty(&res->l_mask_waiters),
			"Lockres %s has mask waiters pending\n",
			res->l_name);
	mlog_bug_on_msg(spin_is_locked(&res->l_lock),
			"Lockres %s is locked\n",
			res->l_name);
	mlog_bug_on_msg(res->l_ro_holders,
			"Lockres %s has %u ro holders\n",
			res->l_name, res->l_ro_holders);
	mlog_bug_on_msg(res->l_ex_holders,
			"Lockres %s has %u ex holders\n",
			res->l_name, res->l_ex_holders);

	/* Need to clear out the lock status block for the dlm */
	memset(&res->l_lksb, 0, sizeof(res->l_lksb));

	res->l_flags = 0UL;
	mlog_exit_void();
}

static inline void ocfs2_inc_holders(struct ocfs2_lock_res *lockres,
				     int level)
{
	mlog_entry_void();

	BUG_ON(!lockres);

	switch(level) {
	case LKM_EXMODE:
		lockres->l_ex_holders++;
		break;
	case LKM_PRMODE:
		lockres->l_ro_holders++;
		break;
	default:
		BUG();
	}

	mlog_exit_void();
}

static inline void ocfs2_dec_holders(struct ocfs2_lock_res *lockres,
				     int level)
{
	mlog_entry_void();

	BUG_ON(!lockres);

	switch(level) {
	case LKM_EXMODE:
		BUG_ON(!lockres->l_ex_holders);
		lockres->l_ex_holders--;
		break;
	case LKM_PRMODE:
		BUG_ON(!lockres->l_ro_holders);
		lockres->l_ro_holders--;
		break;
	default:
		BUG();
	}
	mlog_exit_void();
}

/* 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 ocfs2_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 lockres_set_flags(struct ocfs2_lock_res *lockres,
			      unsigned long newflags)
{
	struct ocfs2_mask_waiter *mw, *tmp;

 	assert_spin_locked(&lockres->l_lock);

	lockres->l_flags = newflags;

	list_for_each_entry_safe(mw, tmp, &lockres->l_mask_waiters, mw_item) {
		if ((lockres->l_flags & mw->mw_mask) != mw->mw_goal)
			continue;

		list_del_init(&mw->mw_item);
		mw->mw_status = 0;
		complete(&mw->mw_complete);
	}
}
static void lockres_or_flags(struct ocfs2_lock_res *lockres, unsigned long or)
{
	lockres_set_flags(lockres, lockres->l_flags | or);
}
static void lockres_clear_flags(struct ocfs2_lock_res *lockres,
				unsigned long clear)
{
	lockres_set_flags(lockres, lockres->l_flags & ~clear);
}

static inline void ocfs2_generic_handle_downconvert_action(struct ocfs2_lock_res *lockres)
{
	mlog_entry_void();

	BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BUSY));
	BUG_ON(!(lockres->l_flags & OCFS2_LOCK_ATTACHED));
	BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BLOCKED));
	BUG_ON(lockres->l_blocking <= LKM_NLMODE);

	lockres->l_level = lockres->l_requested;
	if (lockres->l_level <=
	    ocfs2_highest_compat_lock_level(lockres->l_blocking)) {
		lockres->l_blocking = LKM_NLMODE;
		lockres_clear_flags(lockres, OCFS2_LOCK_BLOCKED);
	}
	lockres_clear_flags(lockres, OCFS2_LOCK_BUSY);

	mlog_exit_void();
}

static inline void ocfs2_generic_handle_convert_action(struct ocfs2_lock_res *lockres)
{
	mlog_entry_void();

	BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BUSY));
	BUG_ON(!(lockres->l_flags & OCFS2_LOCK_ATTACHED));

	/* Convert from RO to EX doesn't really need anything as our
	 * information is already up to data. Convert from NL to
	 * *anything* however should mark ourselves as needing an
	 * update */
	if (lockres->l_level == LKM_NLMODE &&
	    lockres->l_ops->flags & LOCK_TYPE_REQUIRES_REFRESH)
		lockres_or_flags(lockres, OCFS2_LOCK_NEEDS_REFRESH);

	lockres->l_level = lockres->l_requested;
	lockres_clear_flags(lockres, OCFS2_LOCK_BUSY);

	mlog_exit_void();
}

static inline void ocfs2_generic_handle_attach_action(struct ocfs2_lock_res *lockres)
{
	mlog_entry_void();

	BUG_ON((!(lockres->l_flags & OCFS2_LOCK_BUSY)));
	BUG_ON(lockres->l_flags & OCFS2_LOCK_ATTACHED);

	if (lockres->l_requested > LKM_NLMODE &&
	    !(lockres->l_flags & OCFS2_LOCK_LOCAL) &&
	    lockres->l_ops->flags & LOCK_TYPE_REQUIRES_REFRESH)
		lockres_or_flags(lockres, OCFS2_LOCK_NEEDS_REFRESH);

	lockres->l_level = lockres->l_requested;
	lockres_or_flags(lockres, OCFS2_LOCK_ATTACHED);
	lockres_clear_flags(lockres, OCFS2_LOCK_BUSY);

	mlog_exit_void();
}

static int ocfs2_generic_handle_bast(struct ocfs2_lock_res *lockres,
				     int level)
{
	int needs_downconvert = 0;
	mlog_entry_void();

	assert_spin_locked(&lockres->l_lock);

	lockres_or_flags(lockres, OCFS2_LOCK_BLOCKED);

	if (level > lockres->l_blocking) {
		/* only schedule a downconvert if we haven't already scheduled
		 * one that goes low enough to satisfy the level we're
		 * blocking.  this also catches the case where we get
		 * duplicate BASTs */
		if (ocfs2_highest_compat_lock_level(level) <
		    ocfs2_highest_compat_lock_level(lockres->l_blocking))
			needs_downconvert = 1;

		lockres->l_blocking = level;
	}

	mlog_exit(needs_downconvert);
	return needs_downconvert;
}

static void ocfs2_blocking_ast(void *opaque, int level)
{
	struct ocfs2_lock_res *lockres = opaque;
	struct ocfs2_super *osb = ocfs2_get_lockres_osb(lockres);
	int needs_downconvert;
	unsigned long flags;

	BUG_ON(level <= LKM_NLMODE);

	mlog(0, "BAST fired for lockres %s, blocking %d, level %d type %s\n",
	     lockres->l_name, level, lockres->l_level,
	     ocfs2_lock_type_string(lockres->l_type));

	/*
	 * We can skip the bast for locks which don't enable caching -
	 * they'll be dropped at the earliest possible time anyway.
	 */
	if (lockres->l_flags & OCFS2_LOCK_NOCACHE)
		return;

	spin_lock_irqsave(&lockres->l_lock, flags);
	needs_downconvert = ocfs2_generic_handle_bast(lockres, level);
	if (needs_downconvert)
		ocfs2_schedule_blocked_lock(osb, lockres);
	spin_unlock_irqrestore(&lockres->l_lock, flags);

	wake_up(&lockres->l_event);

	ocfs2_wake_downconvert_thread(osb);
}

static void ocfs2_locking_ast(void *opaque)
{
	struct ocfs2_lock_res *lockres = opaque;
	struct dlm_lockstatus *lksb = &lockres->l_lksb;
	unsigned long flags;

	spin_lock_irqsave(&lockres->l_lock, flags);

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

	switch(lockres->l_action) {
	case OCFS2_AST_ATTACH:
		ocfs2_generic_handle_attach_action(lockres);
		lockres_clear_flags(lockres, OCFS2_LOCK_LOCAL);
		break;
	case OCFS2_AST_CONVERT:
		ocfs2_generic_handle_convert_action(lockres);
		break;
	case OCFS2_AST_DOWNCONVERT:
		ocfs2_generic_handle_downconvert_action(lockres);
		break;
	default:
		mlog(ML_ERROR, "lockres %s: ast fired with invalid action: %u "
		     "lockres flags = 0x%lx, unlock action: %u\n",
		     lockres->l_name, lockres->l_action, lockres->l_flags,
		     lockres->l_unlock_action);
		BUG();
	}

	/* set it to something invalid so if we get called again we
	 * can catch it. */
	lockres->l_action = OCFS2_AST_INVALID;

	wake_up(&lockres->l_event);
	spin_unlock_irqrestore(&lockres->l_lock, flags);
}

static inline void ocfs2_recover_from_dlm_error(struct ocfs2_lock_res *lockres,
						int convert)
{
	unsigned long flags;

	mlog_entry_void();
	spin_lock_irqsave(&lockres->l_lock, flags);
	lockres_clear_flags(lockres, OCFS2_LOCK_BUSY);
	if (convert)
		lockres->l_action = OCFS2_AST_INVALID;
	else
		lockres->l_unlock_action = OCFS2_UNLOCK_INVALID;
	spin_unlock_irqrestore(&lockres->l_lock, flags);

	wake_up(&lockres->l_event);
	mlog_exit_void();
}

/* Note: If we detect another process working on the lock (i.e.,
 * OCFS2_LOCK_BUSY), we'll bail out returning 0. It's up to the caller
 * to do the right thing in that case.
 */
static int ocfs2_lock_create(struct ocfs2_super *osb,
			     struct ocfs2_lock_res *lockres,
			     int level,
			     int dlm_flags)
{