This section describes what happens in the normal operation of a file system, what can go wrong, what problems fsck (the checking and repair utility) looks for, and how it corrects the inconsistencies it finds.
Every working day hundreds of files might be created, modified, and removed. Each time a file is modified, the operating system performs a series of file system updates. These updates, when written to the disk reliably, yield a consistent file system.
When a user program does an operation to change the file system, such as a write, the data to be written is first copied into an in-core buffer in the kernel. Normally, the disk update is handled asynchronously; the user process is allowed to proceed even though the data write might not happen until long after the write system call has returned. Thus at any given time, the file system, as it resides on the disk, lags behind the state of the file system represented by the in-core information.
The disk information is updated to reflect the in-core information when the buffer is required for another use or when the kernel automatically runs the fsflush daemon (at 30-second intervals). If the system is halted without writing out the in-core information, the file system on the disk might be in an inconsistent state.
A file system can develop inconsistencies in several ways. The most common causes are operator error and hardware failures.
Problems might result from an unclean shutdown, if a system is shut down improperly, or when a mounted file system is taken offline improperly. To prevent unclean shutdowns, the current state of the file systems must be written to disk (that is, "synchronized") before halting the CPU, physically taking a disk pack out of a drive, or taking a disk offline.
Inconsistencies can also result from defective hardware. Blocks can become damaged on a disk drive at any time, or a disk controller can stop functioning correctly.
This section describes the kinds of consistency checks that fsck applies to these UFS file system components: superblock, cylinder group blocks, inodes, indirect blocks, and data blocks.
The superblock stores summary information, which is the most commonly corrupted item in a UFS file system. Each change to the file system inodes or data blocks also modifies the superblock. If the CPU is halted and the last command is not a sync command, the superblock will almost certainly be corrupted.
The superblock is checked for inconsistencies in:
File system size
Number of inodes
The file system size must be larger than the number of blocks used by the superblock and the list of inodes. The number of inodes must be less than the maximum number allowed for the file system. The file system size and layout information are the most critical pieces of information for fsck. Although there is no way to actually check these sizes, because they are statically determined when the file system is created, fsck can check that the sizes are within reasonable bounds. All other file system checks require that these sizes be correct. If fsck detects corruption in the static parameters of the primary superblock, it requests the operator to specify the location of an alternate superblock.
Free blocks are stored in the cylinder group block maps. fsck checks that all the blocks marked as free are not claimed by any files. When all the blocks have been accounted for, fsck checks to see if the number of free blocks plus the number of blocks claimed by the inodes equal the total number of blocks in the file system. If anything is wrong with the block allocation maps, fsck rebuilds them, leaving out blocks already allocated.
The summary information in the superblock contains a count of the total number of free blocks within the file system. The fsck program compares this count to the number of free blocks it finds within the file system. If the counts do not agree, fsck replaces the count in the superblock with the actual free-block count.
The summary information in the superblock contains a count of the free inodes within the file system. The fsck program compares this count to the number of free inodes it finds within the file system. If the counts do not agree, fsck replaces the count in the superblock with the actual free inode count.
The list of inodes is checked sequentially starting with inode 2 (inode 0 and inode 1 are reserved). Each inode is checked for inconsistencies in:
Format and type
Bad block numbers
Shadow (used for ACLs)
When the file system is created, a fixed number of inodes are set aside, but they are not allocated until they are needed. An allocated inode is one that points to a file. An unallocated inode does not point to a file and, therefore, should be empty. The partially allocated state means that the inode is incorrectly formatted. An inode can get into this state if, for example, bad data is written into the inode list because of a hardware failure. The only corrective action fsck can take is to clear the inode.
Each inode contains a count of the number of directory entries linked to it. The fsck program verifies the link count of each inode by examining the entire directory structure, starting from the root directory, and calculating an actual link count for each inode.
Discrepancies between the link count stored in the inode and the actual link count as determined by fsck might be of three types:
The stored count is not 0 and the actual count is 0.
This condition can occur if no directory entry exists for the inode. In this case, fsck puts the disconnected file in the lost+found directory.
The stored count is not 0 and the actual count is not 0, but the counts are unequal.
This condition can occur if a directory entry has been added or removed but the inode has not been updated. In this case, fsck replaces the stored link count with the actual link count.
The stored count is 0 and the actual count is not 0.
In this case fsck changes the link count of the inode to the actual count.
Each inode contains a list, or pointers to lists (indirect blocks), of all the blocks claimed by the inode. Because indirect blocks are owned by an inode, inconsistencies in indirect blocks directly affect the inode that owns the indirect block.
The fsck program compares each block number claimed by an inode to a list of allocated blocks. If another inode already claims a block number, the block number is put on a list of duplicate blocks. Otherwise, the list of allocated blocks is updated to include the block number.
If there are any duplicate blocks, fsck makes a second pass of the inode list to find the other inode that claims each duplicate block. (A large number of duplicate blocks in an inode might be caused by an indirect block not being written to the file system.) It is not possible to determine with certainty which inode is in error. The fsck program prompts you to choose which inode should be kept and which should be cleared.
The fsck program checks each block number claimed by an inode to see that its value is higher than that of the first data block and lower than that of the last data block in the file system. If the block number is outside this range, it is considered a bad block number.
Bad block numbers in an inode might be caused by an indirect block not being written to the file system. The fsck program prompts you to clear the inode.
Each inode contains a count of the number of data blocks that it references. The number of actual data blocks is the sum of the allocated data blocks and the indirect blocks. fsck computes the number of data blocks and compares that block count against the number of blocks the inode claims. If an inode contains an incorrect count, fsck prompts you to fix it.
Each inode contains a 64-bit size field. This field shows the number of characters (data bytes) in the file associated with the inode. A rough check of the consistency of the size field of an inode is done by using the number of characters shown in the size field to calculate how many blocks should be associated with the inode, and then comparing that to the actual number of blocks claimed by the inode.
Blocks already claimed by another inode
Block numbers outside the range of the file system
The consistency checks listed above are also performed for indirect blocks.
Plain data blocks
Symbolic-link data blocks
Directory data blocks
Plain data blocks contain the information stored in a file. Symbolic-link data blocks contain the path name stored in a symbolic link. Directory data blocks contain directory entries. fsck can check the validity only of directory data blocks.
Directories are distinguished from regular files by an entry in the mode field of the inode. Data blocks associated with a directory contain the directory entries. Directory data blocks are checked for inconsistencies involving:
Directory inode numbers pointing to unallocated inodes
Directory inode numbers greater than the number of inodes in the file system
Incorrect directory inode numbers for "." and ".." directories
Directories disconnected from the file system
If the inode number in a directory data block points to an unallocated inode, fsck removes the directory entry. This condition can occur if the data blocks containing a new directory entry are modified and written out but the inode does not get written out. This condition can occur if the CPU is halted without warning.
The directory inode number entry for "." must be the first entry in the directory data block. It must reference itself; that is, its value must be equal to the inode number for the directory data block.
The directory inode number entry for ".." must be the second entry in the directory data block. Its value must be equal to the inode number of the parent directory (or the inode number of itself if the directory is the root directory).
If the directory inode numbers for "." and ".." are incorrect, fsck replaces them with the correct values. If there are multiple hard links to a directory, the first one found is considered the real parent to which ".." should point. In this case, fsck recommends you have it delete the other names.
The fsck program checks the general connectivity of the file system. If a directory is found that is not linked to the file system, fsck links the directory to the lost+found directory of the file system. (This condition can occur when inodes are written to the file system but the corresponding directory data blocks are not.)
When you run fsck interactively and it completes successfully, the following message is displayed:
# fsck /dev/rdsk/c0t0d0s7 ** /dev/rdsk/c0t0d0s7 ** Last Mounted on /export/home ** Phase 1 - Check Blocks and Sizes ** Phase 2 - Check Pathnames ** Phase 3 - Check Connectivity ** Phase 4 - Check Reference Counts ** Phase 5 - Check Cyl groups 2 files, 9 used, 2833540 free (20 frags, 354190 blocks, 0.0% fragmentation) #
The last line of fsck output describes the following information about the file system:
Number of inodes in use
Number of fragments in use
Number of unused fragments
Number of unused non-block fragments
Number of unused full blocks
Percentage of fragmentation, where: free fragments x 100 / total fragments in the file system