super.c

ocfs1.4.1 oracle分布式文件系统

}

static int ocfs2_sb_probe(struct super_block *sb,
			  struct buffer_head **bh,
			  int *sector_size)
{
	int status, tmpstat;
	struct ocfs1_vol_disk_hdr *hdr;
	struct ocfs2_dinode *di;
	int blksize;

	*bh = NULL;

	/* may be > 512 */
	*sector_size = bdev_hardsect_size(sb->s_bdev);
	if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
		mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
		     *sector_size, OCFS2_MAX_BLOCKSIZE);
		status = -EINVAL;
		goto bail;
	}

	/* Can this really happen? */
	if (*sector_size < OCFS2_MIN_BLOCKSIZE)
		*sector_size = OCFS2_MIN_BLOCKSIZE;

	/* check block zero for old format */
	status = ocfs2_get_sector(sb, bh, 0, *sector_size);
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}
	hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
	if (hdr->major_version == OCFS1_MAJOR_VERSION) {
		mlog(ML_ERROR, "incompatible version: %u.%u\n",
		     hdr->major_version, hdr->minor_version);
		status = -EINVAL;
	}
	if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
		   strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
		mlog(ML_ERROR, "incompatible volume signature: %8s\n",
		     hdr->signature);
		status = -EINVAL;
	}
	brelse(*bh);
	*bh = NULL;
	if (status < 0) {
		mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
		     "upgraded before mounting with ocfs v2\n");
		goto bail;
	}

	/*
	 * Now check at magic offset for 512, 1024, 2048, 4096
	 * blocksizes.  4096 is the maximum blocksize because it is
	 * the minimum clustersize.
	 */
	status = -EINVAL;
	for (blksize = *sector_size;
	     blksize <= OCFS2_MAX_BLOCKSIZE;
	     blksize <<= 1) {
		tmpstat = ocfs2_get_sector(sb, bh,
					   OCFS2_SUPER_BLOCK_BLKNO,
					   blksize);
		if (tmpstat < 0) {
			status = tmpstat;
			mlog_errno(status);
			goto bail;
		}
		di = (struct ocfs2_dinode *) (*bh)->b_data;
		status = ocfs2_verify_volume(di, *bh, blksize);
		if (status >= 0)
			goto bail;
		brelse(*bh);
		*bh = NULL;
		if (status != -EAGAIN)
			break;
	}

bail:
	return status;
}

static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
{
	if (ocfs2_mount_local(osb)) {
		if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
			mlog(ML_ERROR, "Cannot heartbeat on a locally "
			     "mounted device.\n");
			return -EINVAL;
		}
	}

	if (!(osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
		if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb)) {
			mlog(ML_ERROR, "Heartbeat has to be started to mount "
			     "a read-write clustered device.\n");
			return -EINVAL;
		}
	}

	return 0;
}

static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
{
	struct dentry *root;
	int status, sector_size;
	struct mount_options parsed_options;
	struct inode *inode = NULL;
	struct ocfs2_super *osb = NULL;
	struct buffer_head *bh = NULL;
	char nodestr[8];

	mlog_entry("%p, %p, %i", sb, data, silent);

	if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
		status = -EINVAL;
		goto read_super_error;
	}

	/* for now we only have one cluster/node, make sure we see it
	 * in the heartbeat universe */
	if (parsed_options.mount_opt & OCFS2_MOUNT_HB_LOCAL) {
		if (!o2hb_check_local_node_heartbeating()) {
			status = -EINVAL;
			goto read_super_error;
		}
	}

	/* probe for superblock */
	status = ocfs2_sb_probe(sb, &bh, &sector_size);
	if (status < 0) {
		mlog(ML_ERROR, "superblock probe failed!\n");
		goto read_super_error;
	}

	status = ocfs2_initialize_super(sb, bh, sector_size);
	osb = OCFS2_SB(sb);
	if (status < 0) {
		mlog_errno(status);
		goto read_super_error;
	}
	brelse(bh);
	bh = NULL;
	osb->s_mount_opt = parsed_options.mount_opt;
	osb->s_atime_quantum = parsed_options.atime_quantum;
	osb->preferred_slot = parsed_options.slot;
	osb->osb_commit_interval = parsed_options.commit_interval;
	osb->local_alloc_size = parsed_options.localalloc_opt;

#ifdef OCFS2_ORACORE_WORKAROUNDS
	if (osb->s_mount_opt & OCFS2_MOUNT_COMPAT_OCFS)
		sb->s_magic = OCFS_SUPER_MAGIC;
	else
#endif
		sb->s_magic = OCFS2_SUPER_MAGIC;

	/* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
	 * heartbeat=none */
	if (bdev_read_only(sb->s_bdev)) {
		if (!(sb->s_flags & MS_RDONLY)) {
			status = -EACCES;
			mlog(ML_ERROR, "Readonly device detected but readonly "
			     "mount was not specified.\n");
			goto read_super_error;
		}

		/* You should not be able to start a local heartbeat
		 * on a readonly device. */
		if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
			status = -EROFS;
			mlog(ML_ERROR, "Local heartbeat specified on readonly "
			     "device.\n");
			goto read_super_error;
		}

		status = ocfs2_check_journals_nolocks(osb);
		if (status < 0) {
			if (status == -EROFS)
				mlog(ML_ERROR, "Recovery required on readonly "
				     "file system, but write access is "
				     "unavailable.\n");
			else
				mlog_errno(status);			
			goto read_super_error;
		}

		ocfs2_set_ro_flag(osb, 1);

		printk(KERN_NOTICE "Readonly device detected. No cluster "
		       "services will be utilized for this mount. Recovery "
		       "will be skipped.\n");
	}

	if (!ocfs2_is_hard_readonly(osb)) {
		if (sb->s_flags & MS_RDONLY)
			ocfs2_set_ro_flag(osb, 0);
	}

	status = ocfs2_verify_heartbeat(osb);
	if (status < 0) {
		mlog_errno(status);
		goto read_super_error;
	}

	osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
						 ocfs2_debugfs_root);
	if (!osb->osb_debug_root) {
		status = -EINVAL;
		mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
		goto read_super_error;
	}

	status = ocfs2_mount_volume(sb);
	if (osb->root_inode)
		inode = igrab(osb->root_inode);

	if (status < 0)
		goto read_super_error;

	if (!inode) {
		status = -EIO;
		mlog_errno(status);
		goto read_super_error;
	}

	root = d_alloc_root(inode);
	if (!root) {
		status = -ENOMEM;
		mlog_errno(status);
		goto read_super_error;
	}

	sb->s_root = root;

	ocfs2_complete_mount_recovery(osb);

	if (ocfs2_mount_local(osb))
		snprintf(nodestr, sizeof(nodestr), "local");
	else
		snprintf(nodestr, sizeof(nodestr), "%d", osb->node_num);

	printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
	       "with %s data mode.\n",
	       osb->dev_str, nodestr, osb->slot_num,
	       osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
	       "ordered");

	atomic_set(&osb->vol_state, VOLUME_MOUNTED);
	wake_up(&osb->osb_mount_event);

#ifdef HAS_MS_LOOP_NO_AOPS
	/* OCFS2 1.4 doesn't provide prepare and commit write operations */
	sb->s_flags |= MS_LOOP_NO_AOPS;
#endif
	mlog_exit(status);
	return status;

read_super_error:
	if (bh != NULL)
		brelse(bh);

	if (inode)
		iput(inode);

	if (osb) {
		atomic_set(&osb->vol_state, VOLUME_DISABLED);
		wake_up(&osb->osb_mount_event);
		ocfs2_dismount_volume(sb, 1);
	}

	mlog_exit(status);
	return status;
}

#ifdef NO_VFSMOUNT_IN_GET_SB_BDEV
static struct super_block *ocfs2_get_sb(struct file_system_type *fs_type,
					int flags,
					const char *dev_name,
					void *data)
{
	return get_sb_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super);
}
#else
static int ocfs2_get_sb(struct file_system_type *fs_type,
			int flags,
			const char *dev_name,
			void *data,
			struct vfsmount *mnt)
{
	return get_sb_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super,
			   mnt);
}
#endif

static struct file_system_type ocfs2_fs_type = {
	.owner          = THIS_MODULE,
	.name           = "ocfs2",
	.get_sb         = ocfs2_get_sb, /* is this called when we mount
					* the fs? */
	.kill_sb        = kill_block_super, /* set to the generic one
					     * right now, but do we
					     * need to change that? */
	.fs_flags       = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
	.next           = NULL
};

static int ocfs2_parse_options(struct super_block *sb,
			       char *options,
			       struct mount_options *mopt,
			       int is_remount)
{
	int status;
	char *p;

	mlog_entry("remount: %d, options: \"%s\"\n", is_remount,
		   options ? options : "(none)");

	mopt->commit_interval = 0;
	mopt->mount_opt = 0;
	mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
	mopt->slot = OCFS2_INVALID_SLOT;
	mopt->localalloc_opt = OCFS2_DEFAULT_LOCAL_ALLOC_SIZE;

	if (!options) {
		status = 1;
		goto bail;
	}

	while ((p = strsep(&options, ",")) != NULL) {
		int token, option;
		substring_t args[MAX_OPT_ARGS];

		if (!*p)
			continue;

		token = match_token(p, tokens, args);
		switch (token) {
		case Opt_hb_local:
			mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
			break;
		case Opt_hb_none:
			mopt->mount_opt &= ~OCFS2_MOUNT_HB_LOCAL;
			break;
		case Opt_barrier:
			if (match_int(&args[0], &option)) {
				status = 0;
				goto bail;
			}
			if (option)
				mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
			else
				mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
			break;
		case Opt_intr:
			mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
			break;
		case Opt_nointr:
			mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
			break;
		case Opt_err_panic:
			mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
			break;
		case Opt_err_ro:
			mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
			break;
		case Opt_data_ordered:
			mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
			break;
		case Opt_data_writeback:
			mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
			break;
#ifdef OCFS2_ORACORE_WORKAROUNDS
		case Opt_datavolume:
			if (is_remount) {
				mlog(ML_ERROR, "Cannot specify datavolume "
				     "on remount.\n");
				status = 0;
				goto bail;
			}
			mopt->mount_opt |= OCFS2_MOUNT_COMPAT_OCFS;
			break;
#endif
		case Opt_atime_quantum:
			if (match_int(&args[0], &option)) {
				status = 0;
				goto bail;
			}
			if (option >= 0)
				mopt->atime_quantum = option;
			break;
		case Opt_slot:
			option = 0;
			if (match_int(&args[0], &option)) {
				status = 0;
				goto bail;
			}
			if (option)
				mopt->slot = (s16)option;
			break;
		case Opt_commit:
			option = 0;
			if (match_int(&args[0], &option)) {
				status = 0;
				goto bail;
			}
			if (option < 0)
				return 0;
			if (option == 0)
				option = JBD_DEFAULT_MAX_COMMIT_AGE;
			mopt->commit_interval = HZ * option;
			break;
		case Opt_localalloc:
			option = 0;
			if (match_int(&args[0], &option)) {
				status = 0;
				goto bail;
			}
			if (option >= 0 && (option <= ocfs2_local_alloc_size(sb) * 8))
				mopt->localalloc_opt = option;
			break;
		case Opt_localflocks:
			/*
			 * Changing this during remount could race
			 * flock() requests, or "unbalance" existing
			 * ones (e.g., a lock is taken in one mode but
			 * dropped in the other). If users care enough
			 * to flip locking modes during remount, we
			 * could add a "local" flag to individual
			 * flock structures for proper tracking of
			 * state.
			 */
			if (!is_remount)
				mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
			break;
		default:
			mlog(ML_ERROR,
			     "Unrecognized mount option \"%s\" "
			     "or missing value\n", p);
			status = 0;
			goto bail;
		}
	}

	status = 1;

bail:
	mlog_exit(status);
	return status;
}

static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt)
{
	struct ocfs2_super *osb = OCFS2_SB(mnt->mnt_sb);
	unsigned long opts = osb->s_mount_opt;

	if (opts & OCFS2_MOUNT_HB_LOCAL)
		seq_printf(s, ",_netdev,heartbeat=local");
	else
		seq_printf(s, ",heartbeat=none");

	if (opts & OCFS2_MOUNT_NOINTR)
		seq_printf(s, ",nointr");

	if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
		seq_printf(s, ",data=writeback");
	else
		seq_printf(s, ",data=ordered");

	if (opts & OCFS2_MOUNT_BARRIER)
		seq_printf(s, ",barrier=1");

	if (opts & OCFS2_MOUNT_ERRORS_PANIC)
		seq_printf(s, ",errors=panic");
	else
		seq_printf(s, ",errors=remount-ro");