Return to filsys.5 CVS log

Up to [Research Unix] / researchv10dc / man / man5

researchv10 Dan Cross

.TH FILSYS 5
.CT 2 sa
.SH NAME
filsys, flblk, ino \- format of a disk file system 
.SH SYNOPSIS
.B #include <sys/types.h>
.br
.B #include <sys/fblk.h>
.br
.B #include <sys/filsys.h>
.br
.B #include <sys/ino.h>
.SH DESCRIPTION
Every
file system is divided into a certain number
of blocks of 1K or 4K bytes, as determined by
the predicate
.L BITFS()
applied to the minor device number
where the file system is mounted.
Block 0 is unused and is available to contain
a bootstrap program, pack label, or other information.
.PP
Block 1 is the
`super-block'.
Its layout is defined in
.LR <sys/filsys.h> :
.PP
.EX
.ta \w'#define 'u +\w's_cylsize 'u +\w'123 'u +\w'daddr_t 'u
struct filsys {
	unsigned short s_isize;
	daddr_t	s_fsize;
	short  	s_ninode;
	ino_t  	s_inode[NICINOD];
	char   	s_flock;
	char   	s_ilock;
	char   	s_fmod;
	char   	s_ronly;
	time_t 	s_time;
	daddr_t	s_tfree;
	ino_t  	s_tinode;
	short	s_dinfo[2];
#define	s_m	s_dinfo[0]
#define	s_n	s_dinfo[1]
#define	s_cylsize	s_dinfo[0]
#define	s_aspace	s_dinfo[1]
	char   	s_fsmnt[14];
	ino_t	s_lasti;
	ino_t	s_nbehind;
	union {
		struct {
			short  	S_nfree;
			daddr_t	S_free[NICFREE];
		} R;
		struct {
			char	S_valid;
#define	BITMAP	961
			long	S_bfree[BITMAP];
		} B;
		struct {
			char	S_valid;
			char	S_flag;  /* 1 means bitmap not in S_bfree */
			long	S_bsize; /* size of bitmap blocks */
			struct buf *	S_blk[BITMAP-1];
		} N;
	} U;
};
#define	s_nfree	U.R.S_nfree
#define	s_free	U.R.S_free
#define	s_valid	U.B.S_valid
#define	s_bfree	U.B.S_bfree
.EE
.TF s_isize
.TP
.B s_isize
The address of the first block after the i-list,
which starts in block 2.
Thus the i-list is
.LR s_isize-2
blocks long.
.PD
.TP
.B s_fsize
The address of the first block not in the file system.
.TP
.B s_inode
Array of free inode numbers.
.TP
.B  s_ninode
The number of free i-numbers in the
.L s_inode
array.
Inodes are placed in the list in LIFO order.
If the list underflows, it is replenished by
searching the i-list
to obtain the numbers of free inodes.
When the list is full,
freed inodes are not recorded in
.LR s_inode .
.TP
.B s_lasti
Where the last search for free inodes ended.
.TP
.B s_nbehind
Number of free inodes before
.L s_lasti
that are not listed in
.LR s_inode .
The system will search forward for free inodes from
.L s_lasti
for more inodes unless
.L s_nbehind
is sufficiently large, in which case it will search the
i-list from the beginning.
.TP
.B s_flock
.br
.ns
.TP
.B s_ilock
Flags maintained in the core
copy of the super-block while the file system
while it is mounted.
The values on disk are immaterial.
.TP
.B s_fmod
Flag to indicate that the super-block has
changed and should be copied to
the disk during the next periodic update of file
system information.
The value on disk is immaterial.
.TP
.B s_ronly
Flag for read-only file system.
The value on disk is immaterial.
.TP
.B  s_time
Time of the last change to the super block.
.TP
.B s_dinfo
Disk interleave information:
.BR s_cylsize =
blocks per cylinder,
.BR s_aspace =
blocks to skip; see
.IR fsck (8).
.TP
.B s_fsmnt
Unused.
.TP
.B s_tfree
.br
.ns
.TP
.B s_tinode
Numbers of free blocks and free inodes.
Maintained for the benefit of
.IR df
(see
.IR du (1)),
these values are otherwise irrelevant.
.PD
.PP
Different data are used to manage free space in 1K and 4K file systems.
These fields are for 1K file systems:
.TF s_isize
.TP
.B s_free
An array of free block numbers.
.LR s_free[0]
is the block address of the next
in a chain of blocks constituting the free list.
The layout of these blocks is defined in 
.LR <sys/fblk.h> :
.PD
.LP
.EX
struct fblk {
	int	df_nfree;
	daddr_t	df_free[NICFREE];
}
.EE
.ns
.IP
where 
.L df_nfree
and
.L df_free
are exactly like
.L s_nfree
and
.L s_free.
.TP
.B s_nfree
Blocks given in
.L s_free[1]
through
.L s_free[s_nfree-1]
are available for allocation.
Blocks are allocated in LIFO fashion from this list.
If freeing would cause
the array to overflow,
it is cleared by copying into the newly freed block,
which is pushed onto the free chain.
If allocation would cause underflow, 
the array is replenished from the next block on the chain.
.PP
These are for 4K file systems:
.TF s_isize
.TP
.B s_bfree
a bit array specifying the free blocks of a 4K file system.
The bit
.LR (s_bfree[i/w]>>(i%w))&1 ,
where
.I w
is the bit size of a long,
is nonzero if the
.IR i th
data block is free.  If the file system is too large for the bitmap
to fit here, then it is stored at the end of the file system, and
locked into memory when the file system is mounted.  The
.B N
variant of the union is used by the kernel in this case.
.PD
.TP
.B s_valid
The bitmap of a mounted file system is maintained only in main memory;
the bitmap on the medium is marked invalid by setting
.L s_valid
to zero.
Unmounting
updates the medium copy and sets
.L s_valid
to 1.
A file system with invalid bitmap may be mounted
read-only; its bitmap can be corrected by
.IR chuck (8).
.PP
I-numbers begin at 1, and the storage for inodes
begins in block 2.
Inodes are 64 bytes long.
Inode 2 is reserved for the root directory of the file
system, but no other i-number has a built-in
meaning.
Each inode represents one file.
.PP
The layout of an inode is defined in
.LR <sys/ino.h> :
.PP
.EX
.ta \w'struct 'u +\w'unsigned 'u
struct	dinode {
	unsigned short di_mode;
	short	di_nlink;
	short	di_uid;
	short	di_gid;
	off_t	di_size;
	char	di_addr[40];
	time_t	di_atime;
	time_t	di_mtime;
	time_t	di_ctime;
};
.EE
.TF di_nlink
.TP
.B di_mode
The kind of file; it
is encoded as
.L st_mode field of
.IR stat (2),
and is 0 for a free inode.
.PD
.TP
.B di_nlink
The number of directory entries
(links) that refer to this inode
.TP
.B di_uid
Owner's userid.
.TP
.B di_gid
Owner's groupid.
.TP
.B di_size
Number of bytes in the file.
.TP
.B di_atime
Time of last access; see
.IR times (2).
.TP
.B di_mtime
Time of last modification.
.TP
.B di_ctime
Time of last change to inode or contents.
.TP
.B di_addr
For special files the first two bytes of
.L di_addr
contain the device code; see
.IR intro (4)
and
.IR types (5).
.PD
.IP
For plain files and directories
.L di_addr
contains block numbers packed into 3 bytes each.
The first 10 numbers specify device blocks directly.
The last 3 are singly, doubly, and triply
indirect and point to blocks of block pointers of type
.L daddr_t
(see
.IR types (5)).
A zero pointer indicates a `hole'
where no data has been written.
Holes read as if they contained all zeros.
.PP
A symbolic link is, aside from mode,
a plain file whose sole content is the name of the file linked to.
.SH "SEE ALSO"
.IR chuck (8), 
.IR fsck (8),
.IR icheck (8),
.IR dir (5), 
.IR mount (8), 
.IR stat (2), 
.IR types (5),
.IR l3tol (3)

This archive runs on limited infrastructure. Preserving old code on modern bandwidth. Automated agents are requested to crawl responsibly.