Subversion Repositories or1k

/* The paper
 
     Duncan, Roy
     Design goals and implementation of the new High Performance File System
     Microsoft Systems Journal  Sept 1989  v4 n5 p1(13)
 
   describes what HPFS looked like when it was new, and it is the source
   of most of the information given here.  The rest is conjecture.
 
   For definitive information on the Duncan paper, see it, not this file.
   For definitive information on HPFS, ask somebody else -- this is guesswork.
   There are certain to be many mistakes. */
 
/* Notation */
 
typedef unsigned secno;			/* sector number, partition relative */
 
typedef secno dnode_secno;		/* sector number of a dnode */
typedef secno fnode_secno;		/* sector number of an fnode */
typedef secno anode_secno;		/* sector number of an anode */
 
/* sector 0 */
 
/* The boot block is very like a FAT boot block, except that the
   29h signature byte is 28h instead, and the ID string is "HPFS". */
 
struct hpfs_boot_block
{
  unsigned char jmp[3];
  unsigned char oem_id[8];
  unsigned char bytes_per_sector[2];	/* 512 */
  unsigned char sectors_per_cluster;
  unsigned char n_reserved_sectors[2];
  unsigned char n_fats;
  unsigned char n_rootdir_entries[2];
  unsigned char n_sectors_s[2];
  unsigned char media_byte;
  unsigned short sectors_per_fat;
  unsigned short sectors_per_track;
  unsigned short heads_per_cyl;
  unsigned int n_hidden_sectors;
  unsigned int n_sectors_l;		/* size of partition */
  unsigned char drive_number;
  unsigned char mbz;
  unsigned char sig_28h;		/* 28h */
  unsigned char vol_serno[4];
  unsigned char vol_label[11];
  unsigned char sig_hpfs[8];		/* "HPFS    " */
  unsigned char pad[448];
  unsigned short magic;			/* aa55 */
};
 
 
/* sector 16 */
 
/* The super block has the pointer to the root directory. */
 
#define SB_MAGIC 0xf995e849
 
struct hpfs_super_block
{
  unsigned magic;			/* f995 e849 */
  unsigned magic1;			/* fa53 e9c5, more magic? */
  unsigned huh202;			/* ?? 202 = N. of B. in 1.00390625 S.*/
  fnode_secno root;			/* fnode of root directory */
  secno n_sectors;			/* size of filesystem */
  unsigned n_badblocks;			/* number of bad blocks */
  secno bitmaps;			/* pointers to free space bit maps */
  unsigned zero1;			/* 0 */
  secno badblocks;			/* bad block list */
  unsigned zero3;			/* 0 */
  time_t last_chkdsk;			/* date last checked, 0 if never */
  unsigned zero4;			/* 0 */
  secno n_dir_band;			/* number of sectors in dir band */
  secno dir_band_start;			/* first sector in dir band */
  secno dir_band_end;			/* last sector in dir band */
  secno dir_band_bitmap;		/* free space map, 1 dnode per bit */
  unsigned zero5[8];			/* 0 */
  secno scratch_dnodes;			/* ?? 8 preallocated sectors near dir
					   band, 4-aligned. */
  unsigned zero6[103];			/* 0 */
};
 
 
/* sector 17 */
 
/* The spare block has pointers to spare sectors.  */
 
#define SP_MAGIC 0xf9911849
 
struct hpfs_spare_block
{
  unsigned magic;			/* f991 1849 */
  unsigned magic1;			/* fa52 29c5, more magic? */
 
  unsigned dirty: 1;			/* 0 clean, 1 "improperly stopped" */
  unsigned flag1234: 4;			/* unknown flags */
  unsigned fast: 1;			/* partition was fast formatted */
  unsigned flag6to31: 26;		/* unknown flags */
 
  secno hotfix_map;			/* info about remapped bad sectors */
  unsigned n_spares_used;		/* number of hotfixes */
  unsigned n_spares;			/* number of spares in hotfix map */
  unsigned n_dnode_spares_free;		/* spare dnodes unused */
  unsigned n_dnode_spares;		/* length of spare_dnodes[] list,
					   follows in this block*/
  secno code_page_dir;			/* code page directory block */
  unsigned n_code_pages;		/* number of code pages */
  unsigned large_numbers[2];		/* ?? */
  unsigned zero1[15];
  dnode_secno spare_dnodes[20];		/* emergency free dnode list */
  unsigned zero2[81];			/* room for more? */
};
 
/* The bad block list is 4 sectors long.  The first word must be zero,
   the remaining words give n_badblocks bad block numbers.
   I bet you can see it coming... */
 
#define BAD_MAGIC 0
 
/* The hotfix map is 4 sectors long.  It looks like
 
       secno from[n_spares];
       secno to[n_spares];
 
   The to[] list is initialized to point to n_spares preallocated empty
   sectors.  The from[] list contains the sector numbers of bad blocks
   which have been remapped to corresponding sectors in the to[] list.
   n_spares_used gives the length of the from[] list. */
 
 
/* Sectors 18 and 19 are preallocated and unused.
   Maybe they're spares for 16 and 17, but simple substitution fails. */
 
 
/* The code page info pointed to by the spare block consists of an index
   block and blocks containing uppercasing tables.  I don't know what
   these are for (CHKDSK, maybe?) -- OS/2 does not seem to use them
   itself.  Linux doesn't use them either. */
 
/* block pointed to by spareblock->code_page_dir */
 
#define CP_DIR_MAGIC 0x494521f7
 
struct code_page_directory
{
  unsigned magic;			/* 4945 21f7 */
  unsigned n_code_pages;		/* number of pointers following */
  unsigned zero1[2];
  struct {
    unsigned short ix;			/* index */
    unsigned short code_page_number;	/* code page number */
    unsigned bounds;			/* matches corresponding word
					   in data block */
    secno code_page_data;		/* sector number of a code_page_data
					   containing c.p. array */
    unsigned index;			/* index in c.p. array in that sector*/
  } array[31];				/* unknown length */
};
 
/* blocks pointed to by code_page_directory */
 
#define CP_DATA_MAGIC 0x894521f7
 
struct code_page_data
{
  unsigned magic;			/* 8945 21f7 */
  unsigned n_used;			/* # elements used in c_p_data[] */
  unsigned bounds[3];			/* looks a bit like
					     (beg1,end1), (beg2,end2)
					   one byte each */
  unsigned short offs[3];		/* offsets from start of sector
					   to start of c_p_data[ix] */
  struct {
    unsigned short ix;			/* index */
    unsigned short code_page_number;	/* code page number */
    unsigned short zero1;
    unsigned char map[128];		/* upcase table for chars 80..ff */
    unsigned short zero2;
  } code_page[3];
  unsigned char incognita[78];
};
 
 
/* Free space bitmaps are 4 sectors long, which is 16384 bits.
   16384 sectors is 8 meg, and each 8 meg band has a 4-sector bitmap.
   Bit order in the maps is little-endian.  0 means taken, 1 means free.
 
   Bit map sectors are marked allocated in the bit maps, and so are sectors 
   off the end of the partition.
 
   Band 0 is sectors 0-3fff, its map is in sectors 18-1b.
   Band 1 is 4000-7fff, its map is in 7ffc-7fff.
   Band 2 is 8000-ffff, its map is in 8000-8003.
   The remaining bands have maps in their first (even) or last (odd) 4 sectors
     -- if the last, partial, band is odd its map is in its last 4 sectors.
 
   The bitmap locations are given in a table pointed to by the super block.
   No doubt they aren't constrained to be at 18, 7ffc, 8000, ...; that is
   just where they usually are.
 
   The "directory band" is a bunch of sectors preallocated for dnodes.
   It has a 4-sector free space bitmap of its own.  Each bit in the map
   corresponds to one 4-sector dnode, bit 0 of the map corresponding to
   the first 4 sectors of the directory band.  The entire band is marked
   allocated in the main bitmap.   The super block gives the locations
   of the directory band and its bitmap.  ("band" doesn't mean it is
   8 meg long; it isn't.)  */
 
 
/* dnode: directory.  4 sectors long */
 
/* A directory is a tree of dnodes.  The fnode for a directory
   contains one pointer, to the root dnode of the tree.  The fnode
   never moves, the dnodes do the B-tree thing, splitting and merging
   as files are added and removed.  */
 
#define DNODE_MAGIC   0x77e40aae
 
struct dnode {
  unsigned magic;			/* 77e4 0aae */
  unsigned first_free;			/* offset from start of dnode to
					   first free dir entry */
  unsigned increment_me;		/* some kind of activity counter?
					   Neither HPFS.IFS nor CHKDSK cares
					   if you change this word */
  secno up;				/* (root dnode) directory's fnode
					   (nonroot) parent dnode */
  dnode_secno self;			/* pointer to this dnode */
  unsigned char dirent[2028];		/* one or more dirents */
};
 
struct hpfs_dirent {
  unsigned short length;		/* offset to next dirent */
  unsigned first: 1;			/* set on phony ^A^A (".") entry */
  unsigned flag1: 1;
  unsigned down: 1;			/* down pointer present (after name) */
  unsigned last: 1;			/* set on phony \377 entry */
  unsigned flag4: 1;
  unsigned flag5: 1;
  unsigned flag6: 1;
  unsigned has_needea: 1;		/* ?? some EA has NEEDEA set
					   I have no idea why this is
					   interesting in a dir entry */
  unsigned read_only: 1;		/* dos attrib */
  unsigned hidden: 1;			/* dos attrib */
  unsigned system: 1;			/* dos attrib */
  unsigned flag11: 1;			/* would be volume label dos attrib */
  unsigned directory: 1;		/* dos attrib */
  unsigned archive: 1;			/* dos attrib */
  unsigned not_8x3: 1;			/* name is not 8.3 */
  unsigned flag15: 1;
  fnode_secno fnode;			/* fnode giving allocation info */
  time_t write_date;			/* mtime */
  unsigned file_size;			/* file length, bytes */
  time_t read_date;			/* atime */
  time_t creation_date;			/* ctime */
  unsigned ea_size;			/* total EA length, bytes */
  unsigned char zero1;
  unsigned char ix;			/* code page index (of filename), see
					   struct code_page_data */
  unsigned char namelen, name[1];	/* file name */
  /* dnode_secno down;	  btree down pointer, if present,
     			  follows name on next word boundary, or maybe it
			  precedes next dirent, which is on a word boundary. */
};
 
/* The b-tree down pointer from a dir entry */
 
static inline dnode_secno de_down_pointer (struct hpfs_dirent *de)
{
  return *(dnode_secno *) ((void *) de + de->length - 4);
}
 
/* The first dir entry in a dnode */
 
static inline struct hpfs_dirent *dnode_first_de (struct dnode *dnode)
{
  return (void *) dnode->dirent;
}
 
/* The end+1 of the dir entries */
 
static inline struct hpfs_dirent *dnode_end_de (struct dnode *dnode)
{
  return (void *) dnode + dnode->first_free;
}
 
/* The dir entry after dir entry de */
 
static inline struct hpfs_dirent *de_next_de (struct hpfs_dirent *de)
{
  return (void *) de + de->length;
}
 
 
/* B+ tree: allocation info in fnodes and anodes */
 
/* dnodes point to fnodes which are responsible for listing the sectors
   assigned to the file.  This is done with trees of (length,address)
   pairs.  (Actually triples, of (length, file-address, disk-address)
   which can represent holes.  Find out if HPFS does that.)
   At any rate, fnodes contain a small tree; if subtrees are needed
   they occupy essentially a full block in anodes.  A leaf-level tree node
   has 3-word entries giving sector runs, a non-leaf node has 2-word
   entries giving subtree pointers.  A flag in the header says which. */
 
struct bplus_leaf_node
{
  unsigned file_secno;			/* first file sector in extent */
  unsigned length;			/* length, sectors */
  secno disk_secno;			/* first corresponding disk sector */
};
 
struct bplus_internal_node
{
  unsigned file_secno;			/* subtree maps sectors < this  */
  anode_secno down;			/* pointer to subtree */
};
 
struct bplus_header
{
  unsigned flag0: 1;
  unsigned flag1: 1;
  unsigned flag2: 1;
  unsigned flag3: 1;
  unsigned flag4: 1;
  unsigned fnode_parent: 1;		/* ? we're pointed to by an fnode,
					   the data btree or some ea or the
					   main ea bootage pointer ea_secno */
					/* also can get set in fnodes, which
					   may be a chkdsk glitch or may mean
					   this bit is irrelevant in fnodes,
					   or this interpretation is all wet */
  unsigned flag6: 1;
  unsigned internal: 1;			/* 1 -> (internal) tree of anodes
					   0 -> (leaf) list of extents */
  unsigned char fill[3];
  unsigned char n_free_nodes;		/* free nodes in following array */
  unsigned char n_used_nodes;		/* used nodes in following array */
  unsigned short first_free;		/* offset from start of header to
					   first free node in array */
  union {
    struct bplus_internal_node internal[0]; /* (internal) 2-word entries giving
					       subtree pointers */
    struct bplus_leaf_node external[0];	    /* (external) 3-word entries giving
					       sector runs */
  } u;
};
 
/* fnode: root of allocation b+ tree, and EA's */
 
/* Every file and every directory has one fnode, pointed to by the directory
   entry and pointing to the file's sectors or directory's root dnode.  EA's
   are also stored here, and there are said to be ACL's somewhere here too. */
 
#define FNODE_MAGIC 0xf7e40aae
 
struct fnode
{
  unsigned magic;			/* f7e4 0aae */
  unsigned zero1[2];
  unsigned char len, name[15];		/* true length, truncated name */
  fnode_secno up;			/* pointer to file's directory fnode */
  unsigned zero2[3];
  unsigned ea_size_l;			/* length of disk-resident ea's */
  secno ea_secno;			/* first sector of disk-resident ea's*/
  unsigned short ea_size_s;		/* length of fnode-resident ea's */
 
  unsigned flag0: 1;
  unsigned ea_anode: 1;			/* 1 -> ea_secno is an anode */
  unsigned flag2: 1;
  unsigned flag3: 1;
  unsigned flag4: 1;
  unsigned flag5: 1;
  unsigned flag6: 1;
  unsigned flag7: 1;
  unsigned dirflag: 1;			/* 1 -> directory.  first & only extent
					   points to dnode. */
  unsigned flag9: 1;
  unsigned flag10: 1;
  unsigned flag11: 1;
  unsigned flag12: 1;
  unsigned flag13: 1;
  unsigned flag14: 1;
  unsigned flag15: 1;
 
  struct bplus_header btree;		/* b+ tree, 8 extents or 12 subtrees */
  union {
    struct bplus_leaf_node external[8];
    struct bplus_internal_node internal[12];
  } u;
 
  unsigned file_size;			/* file length, bytes */
  unsigned n_needea;			/* number of EA's with NEEDEA set */
  unsigned zero4[4];
  unsigned ea_offs;			/* offset from start of fnode
					   to first fnode-resident ea */
  unsigned zero5[2];
  unsigned char ea[316];		/* zero or more EA's, packed together
					   with no alignment padding.
					   (Do not use this name, get here
					   via fnode + ea_offs. I think.) */
};
 
 
/* anode: 99.44% pure allocation tree */
 
#define ANODE_MAGIC 0x37e40aae
 
struct anode
{
  unsigned magic;			/* 37e4 0aae */
  anode_secno self;			/* pointer to this anode */
  secno up;				/* parent anode or fnode */
 
  struct bplus_header btree;		/* b+tree, 40 extents or 60 subtrees */
  union {
    struct bplus_leaf_node external[40];
    struct bplus_internal_node internal[60];
  } u;
 
  unsigned fill[3];			/* unused */
};
 
 
/* extended attributes.
 
   A file's EA info is stored as a list of (name,value) pairs.  It is
   usually in the fnode, but (if it's large) it is moved to a single
   sector run outside the fnode, or to multiple runs with an anode tree
   that points to them.
 
   The value of a single EA is stored along with the name, or (if large)
   it is moved to a single sector run, or multiple runs pointed to by an
   anode tree, pointed to by the value field of the (name,value) pair.
 
   Flags in the EA tell whether the value is immediate, in a single sector
   run, or in multiple runs.  Flags in the fnode tell whether the EA list
   is immediate, in a single run, or in multiple runs. */
 
struct extended_attribute
{
  unsigned indirect: 1;			/* 1 -> value gives sector number
					   where real value starts */
  unsigned anode: 1;			/* 1 -> sector is an anode
					   that points to fragmented value */
  unsigned flag2: 1;
  unsigned flag3: 1;
  unsigned flag4: 1;
  unsigned flag5: 1;
  unsigned flag6: 1;
  unsigned needea: 1;			/* required ea */
  unsigned char namelen;		/* length of name, bytes */
  unsigned short valuelen;		/* length of value, bytes */
  /*
    unsigned char name[namelen];	ascii attrib name
    unsigned char nul;			terminating '\0', not counted
    unsigned char value[valuelen];	value, arbitrary
      if this.indirect, valuelen is 8 and the value is
        unsigned length;		real length of value, bytes
        secno secno;			sector address where it starts
      if this.anode, the above sector number is the root of an anode tree
        which points to the value.
  */
};
 
static inline unsigned char *ea_name (struct extended_attribute *ea)
{
  return (void *) ea + sizeof *ea;
}
 
static inline unsigned char *ea_value (struct extended_attribute *ea)
{
  return (void *) ea + sizeof *ea + ea->namelen + 1;
}
 
static inline struct extended_attribute *
    ea_next_ea (struct extended_attribute *ea)
{
  return (void *) ea + sizeof *ea + ea->namelen + 1 + ea->valuelen;
}
 
static inline unsigned ea_indirect_length (struct extended_attribute *ea)
{
  unsigned *v = (void *) ea_value (ea);
  return v[0];
}
 
static inline secno ea_indirect_secno (struct extended_attribute *ea)
{
  unsigned *v = (void *) ea_value (ea);
  return v[1];
}
 
/*
   Local Variables:
   comment-column: 40
   End:
*/