BACKGROUND

[0001] Computer storage systems store data on a computer readable medium. The data may include a set of files and related metadata that together make up a file system. The metadata may include the name, location, organization and/or other information relating to the files. A logical storage volume may be divided into blocks of storage space. The metadata may include a storage bit map (SBM) that indicates which blocks are currently free and which bits are used to store data. When writing to the storage medium, a storage allocation request may be made, and space is assigned from among the free blocks indicated in the SBM.

[0002] File system metadata may become corrupted or contain errors.

Therefore, a file system check may be made to verify the accuracy and consistency of the file system metadata. For instance, a file system check may verify whether the free blocks indicated in the SBM are actually free. The file system check corrects any errors it finds in the metadata. In this way, a file system check helps to prevent corruption of the file system, or errors such as files, or data blocks being accidentally overwritten.

[0003] A file system check may, for example, be carried out in response to a user request, or automatically carried out periodically, or automatically carried out in response to the file system detecting an error in the metadata. A file system check is usually carried out offline, when the file system is un-mounted. In this way the user cannot make read or write requests that might interfere with the offline file system check. However, as the amount of data stored increases, file system checks may take a long time. Thus, an offline file system check, may result in a file system being unavailable for an extended period of time. BRIEF DESCRIPTION OF THE DRAWINGS

[0004] Examples of the disclosure will now be described, by way of non- limiting example only, with reference to the accompanying drawings, in which:

Figure 1 is a schematic example of a storage system;

Figure 2 is a flow diagram showing an example method of allocating storage space;

Figure 3 shows an example structure of a file system;

Figure 4 shows an example of a file system at block level;

Figure 5 shows an example method of mounting a file system in online file system check mode;

Figure 6 shows an example method of enlarging a soft-reserved zone;

Figure 7 shows an example method of re-assigning storage space; and

Figure 8 shows an example apparatus for implementing a file system.

DETAILED DESCRIPTION

[0005] In order to increase availability, some storage systems allow a file system check to be carried out online, i.e. while the file system is mounted. The file system may be read by a user while the online file system check is in progress. However, writing to the file system during an online file system check is more problematic, due to the risk that new writes may conflict with changes made to the data, or metadata, as a result of the file system check.

[0006] One way to allow writes to a storage system during a file system check is to define a hard reserved partition. For example, an entire volume may be designated as a hard reserved volume. The hard reserved volume is not available during normal operation of the file system, but may be written to during an online file system check. Typically the hard reserved volume is not visible to the user. After the file system check is completed, the data may be migrated from the hard reserved volume to one of the normal volumes. However, this approach means that a large amount of storage space is hard reserved and not available during normal use. Also, if the hard reserved volume runs out of space, then no further writing to the file system is possible until the file system check has been completed. Furthermore, the migration of data from the hard reserved volume to a normal volume takes additional overhead.

[0007] The present disclosure proposes a storage volume that includes a non-reserved zone and a soft-reserved zone. "Soft-reserved" means that the reservation is not permanent and storage space from the soft reserved zone may be re-assigned to the non-reserved zone in certain circumstances. For example, during normal operation, file space may be allocated from the non- reserved zone, but during an online file system check storage space may be allocated from the soft-reserved zone. The storage space allocated from the soft-reserved zone is re-assigned to the non-reserved zone and the data is written to the allocated storage space. This allows writing to the storage system during an online file system check, but does not require migration of the data written during the online file system check. It also allows flexibility in the size of the soft-reserved zone. In this way storage space may be more effectively utilized.

[0008] Figure 1 shows a schematic diagram of a storage system. The storage system includes a first logical storage volume 10 and second logical storage volume 20. The first volume includes a soft-reserved zone 12 and a non-reserved zone 14. Likewise, the second volume includes a soft-reserved zone 22 and a non-reserved zone 24. The logical volumes, and their respective zones, are mapped to a physical computer readable storage medium 30. The storage medium 30 may for example comprise a plurality of hard disks.

[0009] Each logical volume may be mapped to a single physical disk, or may be mapped to a plurality of physical disks. In the illustrated example, first volume 10 is mapped to storage space on physical disks 1 -3 and second volume 20 is mapped to storage space on physicals disks 4-6. However, in other examples, first volume 10 and second volume 20 could be mapped to non-overlapping portions of all of disks 1 -6.

[0010] Figure 2 shows a method of allocating storage space in the file system. At block 100 a storage allocation request is received, for example when a user instructs data to be written to a volume of the file system. For instance this may occur when a new file is created, or a when a user saves changes to an existing file. As indicated by block 1 10, the way in which storage is allocated depends upon whether or not a file system check is in progress. If a file system check is not in progress then storage space is allocated from the non-reserved zone at 120. The data may then be written to the allocated storage space. On the other hand, if a file system check is in progress, then storage space is allocated from the soft-reserved zone at block 130. In that case the allocated storage space may be re-assigned from the soft-reserved zone to the non- reserved zone, before writing the data to the allocated storage space.

[001 1] As the allocated storage space is re-assigned, it is not necessary to migrate the data from a reserved volume to a non-reserved volume of the file system. Compared to some conventional systems that migrate data, this approach may save overhead and bandwidth. Further, in the above example, it can be seen that each volume of the file system has a non-reserved zone and a soft-reserved zone. Therefore, even if a file system check is being carried out on for instance the first volume, it is still possible to write data to the first volume by allocating space from the soft-reserved zone of the first volume.

[0012] The file system check may be a full file system check applied to all volumes in the file system, or a partial file system check applied to some, but not all, volumes in the file system. However, unlike some previous known storage systems, the above example allows data to be written to the first volume when a full or partial file system check is applied to the first volume, by writing the data to storage space allocated from the soft-reserved zone of the first volume. Thus, it is not necessary to write the data to a different volume, when the first volume is undergoing a file system check.

[0013] In one example the allocated storage space is re-assigned from the soft-reserved zone to the non-reserved zone by updating the file system metadata.

[0014] Figure 3 is a schematic example of one possible structure of a file system 200. The file system 300 includes a reserved metadata section 210, which is used to store certain types of reserved metadata and a files section 202 that is used to store files and file metadata. The reserved metadata includes a Superblock 212, a Storage Bitmap (SBM) 214 and an Inode Table 216. It may also include a system log and other types of reserved metadata.

[0015] The Superblock 212 indicates where certain types of reserved metadata can be found. Thus the Superblock can be described as metadata of the metadata. For example the Superblock may include descriptors of reserved metadata files. Each descriptor may include a pointer to the metadata which it refers to. In the illustrated example the Superblock 212 includes descriptors of the Superblock itself, descriptors of the SBM, descriptors of the Inode Table and descriptors of the soft-reserved zone. Each descriptor may for example include a pointer to a block and an extent indicating the number of contiguous blocks included in the descriptor. The soft-reserved zone may be defined by one or more descriptors. In one example the soft-reserved zone may be defined by a linked list of descriptors. Thus by reading the Superblock the file system is able to determine where to find certain types of metadata and which blocks belong to the soft-reserved zone.

[0016] The Storage Bitmap (SBM) is a bitmap that indicates for each block in a logical volume, whether the block is free or is already used to store data. For instance, in one example each bit of the bitmap corresponds to a block of storage space, and a 0 indicates that the block is free, while a 1 indicates that the block is used.

[0017] The Inode Table indicates where to find the file metadata for each file. For instance, it may include a list of file names and descriptors indicating where to find each file. The files may be stored in the files section 220. The illustrated example, just shows four files, but in practice there may be fewer files or many more files.

[0018] In the case of an AdvFS file system, the Superblock may be the Reserved Bitmap Table (RBMT) and the Inode Table may be the Bitfile Map Table (BMT).

[0019] When file space is allocated from the soft-reserved zone, the metadata may be updated to reflect this. For instance the descriptors relating to the allocated blocks may be deleted from the part of the metadata that defines the soft-reserved zone. The SBM may be updated to indicate that the allocated blocks are used, and descriptors may be added to the Inode Table to indicate any blocks that are now used for storing file metadata. In this example, once the metadata is updated, the data is written to the allocated blocks.

[0020] Figure 4 shows an example of a file system and methods of writing to data to the file system. The part indicated by 300 shows the file system as it appears to a user, with various directories and files etc. The part indicated generally by 310 is a representation of a volume of the file system at the physical storage block level. At this level the volume comprises blocks of storage for instance as shown by 31 1 , 312, 313etc. In the diagram, the empty blocks such as 31 1 , 1 13 are free blocks that are not currently used to store data, while the filled blocks such as 312 are used blocks that are being used to store data. The volume is divided into a non-reserved zone indicated generally by 320 and a soft-reserved zone indicated generally by 330. The extent of the soft-reserved zone may be defined in metadata of the file system as discussed above.

[0021] Generally when the file system is first set up the soft-reserved zone will be a set fraction of the total size of the volume. For instance it may be equal to 1 % of the size of the volume. In any case, the soft-reserved zone is usually significantly smaller than the non-reserved zone. In this example, the soft- reserved zone consists entirely of free blocks of storage. Thus the soft-reserved zone may be described as a zone of consistent free space. The soft-reserved zone is not used in normal operation of the storage system, it may be guaranteed with a high degree of confidence that it comprises consistent free space. Consistent free space means that the metadata describing the zone is accurate and the blocks are indeed free storage blocks.

[0022] In order for a file system to be available to a user it should be mounted. Once a file system is mounted it is online and may accessed by a user for reading or writing data. The file system may be mounted in a normal mode or in an online file system check (OFSCK) mode.

[0023] In the normal mode, a file system check may not be carried out while the file system is online. In that case, to conduct a file system check, the file system should be un-mounted first. Thus, any file system check should be carried out, before issuing an instruction to mount the file system. In the online file system check mode, an online file system check may be carried out while the system is mounted and online. If it is desired to switch from normal mode to online file system cache mode or vice versa, then file system is un-mounted before re-mounting in the other mode.

[0024] Mounting the file system in online file system check mode may, for example, be conducted as shown in Figure 5. At 700 an instruction is received to mount the file system in an online file check mode. At 710 the consistency of the soft-reserved zone is verified. For instance, the consistency of the soft- reserved zone may be verified by comparing the descriptors of the soft-reserved zone with the SMB to make sure that all blocks indicated as being in the soft- reserved zone are indicated as free blocks by the SMB. It may also be checked that the soft-reserved zone does not overlap with the extent of any files described in the file metadata. At 720 the file system is mounted and thus made available to users. Due to the verification process at 710, it may be ensured with a high degree of confidence that the blocks in the soft-reserved zone are indeed free blocks.

[0025] Referring back to Figure 4, when a write instruction is received, for instance due to creating or amending a file, a storage space allocation request is made to the file system. The way in which the storage space is allocated in response to this request depends upon whether the file system is mounted in a normal mode, or an online file system check mode.

[0026] If the file system is mounted in a normal mode, then as shown by arrow 340, the free blocks of storage space are allocated from the non-reserved zone. If the file system is mounted in an online file system check mode, then blocks should not be allocated from the non-reserved zone which is having its consistency checked, therefore free blocks are allocated from the soft-reserved zone, as shown by arrow 350.

[0027] When allocating free blocks to satisfy storage allocation requests, the storage system may refer to a file system free space cache stored in memory of the computer storage system, rather than reading the metadata directly from the storage volume. Reading the file system free space cache may be quicker than reading the storage volume.

[0028] If the file system is mounted in normal mode, then the file system free space cache may be populated with at least some of the free space blocks from the non-reserved zone. The free blocks may for instance be found from the SBM, but checked against the soft-reserved zone descriptors to make sure that they do not belong to the soft-reserved zone. Upon receiving a storage allocation request, free blocks are allocated based on the cache and then removed from the cache, the SBM metadata on the storage volume updated and data written to the allocated blocks. If the free space cache is empty or reduced to a certain threshold, then it may be repopulated by further free space blocks from the non-reserved zone. [0029] If the file system is mounted in online file system check mode, then the file system free space cache may be populated with at least some of the free blocks from the soft-reserved zone. This may be done based on consulting the descriptors defining the soft-reserved zone. Upon receiving a storage allocation request, free blocks are allocated based on the cache and then removed from the cache, the soft-reserved zone metadata on the storage volume is updated and data is written to the allocated blocks. If the free space cache is empty, or reduced to a certain threshold, then it may be repopulated by further free space blocks from the soft-reserved zone.

[0030] Referring to another part of Figure 4, 360 shows an example in which a number of blocks 325 from the soft-reserved zone are allocated to satisfy the storage allocation request. This allocated blocks 325 are re-assigned, for instance by updating file system metadata, so that they belong to the non- reserved zone 320. Data may then be written to the allocated blocks. As a result the size of the soft-reserved zone 330 is reduced as shown in the diagram.

[0031] If the size of the soft-reserved zone falls below a certain threshold, action may be taken to increase its size again. An example is shown at 370. In this case the non-reserved zone comprises the portions indicate by 320, 320A, 320B, 320C and 320D. The soft-reserved zone has been reduced to the portion indicated by 330. Therefore the file system enters a scavenging mode to find blocks of free space that may be added to the soft-reserved zone to increase the size of the soft-reserve zone should be in normal mode. The scavenging process finds the free blocks in portions 320C and 320A and re-assigns them to the soft-reserved zone. For instance the file system metadata is updated so that these free blocks now belong to the soft-reserved zone. In this way the soft- reserved zone is enlarged. It will be noted that in this example the soft-reserved zone now comprises a plurality of non-contiguous portions of free storage space 330, 320C and 320A linked as shown by the arrows in the diagram. The soft- reserved zone may be defined by descriptors in a linked list in metadata. [0032] This is a flexible system and allows the soft-reserved zone to use of relatively small portions of free space, wherever they occur in the disk. It may also reduce the need for disk level de-fragmentation. The file system may specify a minimum size of free space, e.g. a minimum number of contiguous free blocks, to be added to the soft-reserved zone in this way. E.g. it can be seen in the example of 370, that minimum size is four blocks of free space, and so the single block of free space 313 is not added to the soft-reserved zone. This minimum size may be determined heuristically depending on

circumstances.

[0033] Figure 6 is a flow chart showing an example method of enlarging the soft-reserved zone. At 500 it is determined that the size of the soft-reserved zone is below a threshold. At 510 free blocks of storage space are found in the non-reserved zone. At 520 the free blocks of storage space that were found at 510 are re-assigned to the soft-reserved zone, for instance by updating the metadata. This process should be carried out when the file system is in the normal mode, not when the file system is in the online file system check mode.

[0034] Referring back to Figure 4, 380 shows an example in which the non- reserved zone is full and the file system is mounted in normal mode. Thus storage space should be allocated from the non-reserved zone. However, there are no free blocks left in the non-reserved zone. If a situation like this occurs, or if the non-reserved zone is almost full, or if a large file allocation request cannot be satisfied by the available free space in the non-reserved zone, then free space may be allocated from the soft-reserved zone. In the illustrated example blocks of free space 325 are re-assigned from the soft-reserved zone to the non-reserved zone, in a manger similar to that described for 360. After the metadata has been updated these free blocks may be written to. In this way storage space is not wasted.

[0035] Figure 7 shows an example method of growing the non-reserved zone. At 600 it is determined that there is insufficient space in the non-reserved zone. E.g. the non-reserved zone is full, almost full or has insufficient space to satisfy a file space allocation request as explained above. At 610 storage space from the soft-reserved zone is re-assigned to the non-reserved zone, for instance by updating the metadata. The non-reserved zone is thus enlarged and has extra free space that may be allocated and written to.

[0036] Figure 8 is a schematic diagram showing an example of a computer system 700 for implementing the file system of the present disclosure. The computer system includes a processor 710, a memory 720 and a

communication interface 740 that may be connected by bus 730. The memory 720 stores machine readable instructions that are executable by the processor. The machine readable instructions may include the following modules. A file system creator 722 to set up the file system and define a soft-reserved zone. A file system mount 724 to mount the file system in a normal mode or an online file system check mode. The file system mount 724 may for example mount the file system in an online file system check mode according to the method shown in Figure 5. A write controller 726 to handle write requests and allocate storage space in the file system. The write controller may allocate write requests according to the methods shown in Figures 2 and 4. A space scavenger 727 to find free blocks in the non-reserved zone and re-assign the found blocks to the soft-reserved zone when the size of the soft reserved zone falls below a certain threshold. The space scavenger 727 may operate according to the method shown in Figure 6 for example. A file system check application 728 to carry out a file system check. The memory may have a cache as well as discussed above with reference to Figure 4.

[0037] All of the features disclosed in this specification (including any accompanying claims, abstract and drawings), and/or all of the steps of any method or process so disclosed, may be combined in any combination, except combinations where at least some of such features and/or steps are mutually exclusive.

[0038] Each feature disclosed in this specification (including any

accompanying claims, abstract and drawings), may be replaced by alternative features serving the same, equivalent or similar purpose, unless expressly stated otherwise. Thus, unless expressly stated otherwise, each feature disclosed is one example only of a generic series of equivalent or similar features.