1.. SPDX-License-Identifier: GPL-2.0
2
3================================
4Optimized MPEG Filesystem (OMFS)
5================================
6
7Overview
8========
9
10OMFS is a filesystem created by SonicBlue for use in the ReplayTV DVR
11and Rio Karma MP3 player.  The filesystem is extent-based, utilizing
12block sizes from 2k to 8k, with hash-based directories.  This
13filesystem driver may be used to read and write disks from these
14devices.
15
16Note, it is not recommended that this FS be used in place of a general
17filesystem for your own streaming media device.  Native Linux filesystems
18will likely perform better.
19
20More information is available at:
21
22    http://linux-karma.sf.net/
23
24Various utilities, including mkomfs and omfsck, are included with
25omfsprogs, available at:
26
27    https://bobcopeland.com/karma/
28
29Instructions are included in its README.
30
31Options
32=======
33
34OMFS supports the following mount-time options:
35
36    ============   ========================================
37    uid=n          make all files owned by specified user
38    gid=n          make all files owned by specified group
39    umask=xxx      set permission umask to xxx
40    fmask=xxx      set umask to xxx for files
41    dmask=xxx      set umask to xxx for directories
42    ============   ========================================
43
44Disk format
45===========
46
47OMFS discriminates between "sysblocks" and normal data blocks.  The sysblock
48group consists of super block information, file metadata, directory structures,
49and extents.  Each sysblock has a header containing CRCs of the entire
50sysblock, and may be mirrored in successive blocks on the disk.  A sysblock may
51have a smaller size than a data block, but since they are both addressed by the
52same 64-bit block number, any remaining space in the smaller sysblock is
53unused.
54
55Sysblock header information::
56
57    struct omfs_header {
58	    __be64 h_self;                  /* FS block where this is located */
59	    __be32 h_body_size;             /* size of useful data after header */
60	    __be16 h_crc;                   /* crc-ccitt of body_size bytes */
61	    char h_fill1[2];
62	    u8 h_version;                   /* version, always 1 */
63	    char h_type;                    /* OMFS_INODE_X */
64	    u8 h_magic;                     /* OMFS_IMAGIC */
65	    u8 h_check_xor;                 /* XOR of header bytes before this */
66	    __be32 h_fill2;
67    };
68
69Files and directories are both represented by omfs_inode::
70
71    struct omfs_inode {
72	    struct omfs_header i_head;      /* header */
73	    __be64 i_parent;                /* parent containing this inode */
74	    __be64 i_sibling;               /* next inode in hash bucket */
75	    __be64 i_ctime;                 /* ctime, in milliseconds */
76	    char i_fill1[35];
77	    char i_type;                    /* OMFS_[DIR,FILE] */
78	    __be32 i_fill2;
79	    char i_fill3[64];
80	    char i_name[OMFS_NAMELEN];      /* filename */
81	    __be64 i_size;                  /* size of file, in bytes */
82    };
83
84Directories in OMFS are implemented as a large hash table.  Filenames are
85hashed then prepended into the bucket list beginning at OMFS_DIR_START.
86Lookup requires hashing the filename, then seeking across i_sibling pointers
87until a match is found on i_name.  Empty buckets are represented by block
88pointers with all-1s (~0).
89
90A file is an omfs_inode structure followed by an extent table beginning at
91OMFS_EXTENT_START::
92
93    struct omfs_extent_entry {
94	    __be64 e_cluster;               /* start location of a set of blocks */
95	    __be64 e_blocks;                /* number of blocks after e_cluster */
96    };
97
98    struct omfs_extent {
99	    __be64 e_next;                  /* next extent table location */
100	    __be32 e_extent_count;          /* total # extents in this table */
101	    __be32 e_fill;
102	    struct omfs_extent_entry e_entry;       /* start of extent entries */
103    };
104
105Each extent holds the block offset followed by number of blocks allocated to
106the extent.  The final extent in each table is a terminator with e_cluster
107being ~0 and e_blocks being ones'-complement of the total number of blocks
108in the table.
109
110If this table overflows, a continuation inode is written and pointed to by
111e_next.  These have a header but lack the rest of the inode structure.
112
113