1 (* 'df' command for virtual domains.
2 (C) Copyright 2007 Richard W.M. Jones, Red Hat Inc.
5 This program is free software; you can redistribute it and/or modify
6 it under the terms of the GNU General Public License as published by
7 the Free Software Foundation; either version 2 of the License, or
8 (at your option) any later version.
10 This program is distributed in the hope that it will be useful,
11 but WITHOUT ANY WARRANTY; without even the implied warranty of
12 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 GNU General Public License for more details.
15 You should have received a copy of the GNU General Public License
16 along with this program; if not, write to the Free Software
17 Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
30 (* Type of the private data, basically all the metadata that we
31 * read from the NTFS volume.
34 ntfs_dev : device; (* Device. *)
35 ntfs_blocksize : int63; (* Blocksize (cluster size) *)
36 ntfs_mft_lcn : int63; (* MFT location (bytes) *)
37 ntfs_mft_size : int63; (* MFT size (bytes) *)
38 ntfs_mft_records : ntfs_mft_record list; (* Files in MFT *)
40 and ntfs_mft_record = {
41 ntfs_filename : ntfs_filename option; (* Filename, if present. *)
42 ntfs_info : ntfs_info option; (* Standard information, if present. *)
43 ntfs_data : ntfs_data option; (* $Data stream, if present. *)
46 ntfs_name : string; (* Filename (UTF-8 encoded). *)
49 ntfs_creation_time : int64;
50 ntfs_last_data_change_time : int64;
51 ntfs_last_mft_change_time : int64;
52 ntfs_last_access_time : int64;
55 ntfs_data_size : int63; (* Actual size of data. *)
56 ntfs_runlist : ntfs_runentry list; (* Runlist. *)
59 (* VCN start,size => LCN / None if sparse hole *)
60 (int63 * int63) * int63 option
62 (* Private data functions. *)
63 let attach_private_data, get_private_data =
64 private_data_functions (fun {fs_cb = {fs_cb_uq = u}} -> u)
66 (* Probe for an NTFS filesystem on this device. *)
68 let fs = probe_superblock dev in
71 and probe_superblock dev =
72 (* Load the boot sector / superblock. *)
73 let bits = dev#read_bitstring ~^0 ~^512 in
75 (* Most of this data comes from ntfsprogs' layout.h header file. *)
77 | { _ : 24; (* Jump to boot up code. *)
78 "NTFS " : 64 : string; (* NTFS OEM ID (magic). *)
79 bytes_per_sector : 16 : littleendian;
80 sectors_per_cluster : 8 : littleendian;
81 _ : 16; (* Reserved sectors - unused. *)
82 _ : 8; (* FATs - unused. *)
83 _ : 16; (* Root entries - unused. *)
84 _ : 16; (* Sectors - unused. *)
85 _ : 8; (* Media type, probably 'f8' = HDD *)
86 _ : 16; (* Sectors per FAT - unused. *)
87 _ : 16; (* Sectors per track. *)
89 _ : 32; (* Hidden sectors. *)
90 _ : 32; (* Large sectors. *)
91 _ : 8; (* Physical drive, 0 = FDD, 0x80 = HDD*)
92 _ : 8; (* Current head. *)
93 _ : 8; (* Extended boot signature. *)
94 _ : 8; (* Reserved. *)
95 number_of_sectors : 64 : littleendian;
96 mft_lcn : 64 : littleendian; (* MFT location in clusters. *)
97 mftmirr_lcn : 64 : littleendian; (* MFT mirror location. *)
98 clusters_per_mft_record : 8;
100 clusters_per_index_record : 8;
102 volume_serial_number : 64 : littleendian;
103 checksum : 32 : littleendian; (* Boot sector checksum. *)
104 code : 8 * 426 : bitstring; (* Boot code. *)
105 0x55AA : 16 } -> (* End of bootsector magic. *)
107 let blocksize = bytes_per_sector * sectors_per_cluster in
110 eprintf "%s: NTFS boot sector with blocksize = %d, serial = %Lx\n%!"
111 dev#name blocksize volume_serial_number;
113 let blocksize = Int63.of_int blocksize in
114 let number_of_sectors = Int63.of_int64 number_of_sectors in
116 (* The blocksize of the filesystem is likely to be quite different
117 * from that of the underlying device, so create an overlay device
118 * with the natural filesystem blocksize.
120 let fs_dev = new blocksize_overlay blocksize dev in
122 (* Get the location and size of the Master File Table. *)
123 let mft_lcn = Int63.of_int64 mft_lcn *^ blocksize in
124 let mft_size = Int63.of_int clusters_per_mft_record *^ blocksize in
126 let mft = parse_mft dev mft_lcn mft_size in
130 ntfs_blocksize = blocksize;
131 ntfs_mft_lcn = mft_lcn;
132 ntfs_mft_size = mft_size;
133 ntfs_mft_records = mft
136 (* Query free space. I cannot find any metadata in the NTFS
137 * structures which records free space directly, so instead we
138 * need to read the $Bitmap::$Data (bitmap of allocated LCNs).
140 let blocks_used, blocks_avail = parse_bitmap_freespace ntfs in
142 (* Create a filesystem structure. *)
144 fs_cb = callbacks ();
146 fs_blocksize = blocksize;
147 fs_blocks_total = number_of_sectors *^ bytes_per_sector /^ blocksize;
149 fs_blocks_reserved = ~^0; (* XXX MFT, bitmap are "reserved" *)
150 fs_blocks_avail = blocks_avail;
151 fs_blocks_used = blocks_used;
152 fs_inodes_total = ~^0; (* XXX MFT records are like inodes *)
153 fs_inodes_reserved = ~^0;
154 fs_inodes_avail = ~^0;
155 fs_inodes_used = ~^0;
158 attach_private_data fs ntfs;
161 | { _ } -> raise Not_found (* Not an NTFS boot sector. *)
163 and parse_mft dev mft_lcn mft_size =
164 (* Read the whole of the MFT (which is an array of MFT records) ... *)
165 let bits = dev#read_bitstring mft_lcn mft_size in
167 (* ... and turn the MFT into records. *)
168 let records = parse_mft_records bits in
171 and parse_mft_records bits =
173 | { "FILE" : 32 : string;
174 (* Assume 3 USAs starting at offset 0x30. XXX? *)
175 0x30 : 16 : littleendian;
176 0x03 : 16 : littleendian;
178 _ : 16; (* sequence_number *)
179 _ : 16; (* link_count *)
180 _ : 16; (* attrs_offset *)
181 _ : 16; (* MFT_RECORD_FLAGS *)
182 bytes_in_use : 32 : littleendian;
183 record_size : 32 : littleendian;
184 _ : 64; (* base_mft_record *)
185 _ : 16; (* next_attr_instance *)
186 _ : 16; (* reserved *)
187 _ : 32; (* mft_record_number *)
188 _ : 64; (* USN, 3 * USAs -- see above. *)
190 (* The attributes. Subtract header size (0x30 bytes)
191 * and space for the USN/USAs (8 bytes).
193 attrs : (Int32.to_int record_size - 0x30 - 8)*8 : bitstring;
195 (* Subsequent MFT records: *)
196 rest : -1 : bitstring } ->
199 eprintf "got an MFT record, now parsing attributes ...\n%!";
202 ntfs_filename = None;
206 let mft_record = parse_attrs attrs mft_record in
208 mft_record :: parse_mft_records rest (* loop rest of MFT records *)
210 (* Just assume that the end of the list of MFT records
211 * is marked by all zeroes. This seems to be the
212 * case, but not sure if it is generally true.
215 | { 0x00000000_l : 32 } -> []
219 and parse_attrs attrs mft_record =
220 (* Parse the MFT record attributes. *)
222 | { 0xFFFFFFFF_l : 32 : littleendian } -> (* AT_END *)
224 eprintf "found AT_END, end of attributes\n%!";
227 | { attr_type : 32 : littleendian;
228 attr_size : 32 : littleendian;
229 0 : 8; (* means attribute is resident *)
230 pad : 24*8 - 8 - 64 : bitstring; (* actually meaningful *)
231 attr : (Int32.to_int attr_size - 24) * 8 : bitstring;
232 rest : -1 : bitstring } ->
234 let mft_record = parse_resident_attr attr_type attr mft_record in
235 parse_attrs rest mft_record
237 | { attr_type : 32 : littleendian;
238 attr_size : 32 : littleendian;
239 1 : 8; (* non-resident attribute *)
240 0 : 8; (* name length, assume unnamed *)
241 _ : 16; (* name offset *)
243 _ : 16; (* instance number *)
244 0L : 64 : littleendian; (* lowest VCN, assume single extent *)
245 highest_vcn : 64 : littleendian; (* size in clusters - 1 *)
246 0x40 : 16 : littleendian; (* mapping pairs offset *)
247 0 : 8; (* assume not compressed *)
248 pad : 40 : bitstring; (* padding *)
249 allocated_size : 64 : littleendian; (* allocate size on disk *)
250 data_size : 64 : littleendian; (* byte size of the attribute *)
251 initialized_size : 64 : littleendian;
253 (* Table of virtual clusters to logical clusters. *)
254 mapping_pairs : (Int32.to_int attr_size - 0x40) * 8 : bitstring;
256 rest : -1 : bitstring } ->
258 let data_size = Int63.of_int64 data_size in
261 parse_nonresident_attr attr_type highest_vcn
262 allocated_size data_size initialized_size
263 mapping_pairs mft_record in
265 parse_attrs rest mft_record
267 (* Not matched above, so we don't know how to parse this attribute, but
268 * there is still enough information to skip to the next one.
270 | { attr_type : 32 : littleendian;
271 attr_size : 32 : littleendian;
272 pad : (Int32.to_int attr_size - 8) * 8 : bitstring;
273 rest : -1 : bitstring } ->
276 eprintf "cannot parse MFT attribute entry, attr_type = %lx\n%!"
279 parse_attrs rest mft_record
281 (* Otherwise unparsable & unskippable attribute entry. *)
284 eprintf "corrupt MFT attribute entry\n%!";
287 and parse_resident_attr attr_type attr mft_record =
289 | 0x10_l -> (* AT_STANDARD_INFORMATION *)
291 | { creation_time : 64;
292 last_data_change_time : 64;
293 last_mft_change_time : 64;
294 last_access_time : 64
295 (* other stuff follows, just ignore it *) } ->
298 ntfs_creation_time = creation_time;
299 ntfs_last_data_change_time = last_data_change_time;
300 ntfs_last_mft_change_time = last_mft_change_time;
301 ntfs_last_access_time = last_access_time
303 { mft_record with ntfs_info = Some info }
307 eprintf "cannot parse AT_STANDARD_INFORMATION\n%!";
311 | 0x30_l -> (* AT_FILE_NAME *)
313 | { _ : 64; (* parent directory ref *)
314 _ : 64; (* creation time *)
315 _ : 64; (* last change time *)
316 _ : 64; (* last MFT change time *)
317 _ : 64; (* last access time *)
318 _ : 64; (* allocated size *)
319 _ : 64; (* data size *)
324 name : name_len*16 : string } ->
326 let name = ucs2_to_utf8 name name_len in
330 { mft_record with ntfs_filename = Some filename }
334 eprintf "cannot parse AT_FILE_NAME\n%!";
338 | _ -> (* unknown attribute - just ignore *)
340 eprintf "unknown resident attribute %lx\n%!" attr_type;
343 and parse_nonresident_attr attr_type highest_vcn
344 allocated_size data_size initialized_size
345 mapping_pairs mft_record =
347 | 0x80_l -> (* AT_DATA, ie. the $Data stream *)
348 let lowest_vcn = ~^0 (* see assumption above *) in
349 let runlist = parse_runlist lowest_vcn ~^0 mapping_pairs in
351 eprintf "AT_DATA: runlist is:\n";
354 | ((vcn, deltavcn), Some lcn) ->
355 eprintf "\tVCNs %s..%s -> LCN %s\n"
356 (Int63.to_string vcn) (Int63.to_string (vcn +^ deltavcn -^ ~^1))
357 (Int63.to_string lcn)
358 | ((vcn, deltavcn), None) ->
359 eprintf "\tVCNs %s..%s -> sparse hole\n"
360 (Int63.to_string vcn) (Int63.to_string (vcn +^ deltavcn -^ ~^1))
365 ntfs_data_size = data_size;
366 ntfs_runlist = runlist
368 { mft_record with ntfs_data = Some data }
372 eprintf "unknown non-resident attribute %lx\n%!" attr_type;
375 (* mapping_pairs is not straightforward and not documented well. See
376 * ntfsprogs libntfs/runlist.c:ntfs_mapping_pairs_decompress
378 and parse_runlist vcn lcn bits =
380 | { 0 : 8 } -> (* end of table *)
385 deltavcn : vcnlen * 8 : littleendian;
386 rest : -1 : bitstring
387 } when vcnlen >= 1 && vcnlen <= 4 ->
389 let deltavcn = Int63.of_int64 deltavcn in
391 (* This is a sparse file hole. *)
392 ((vcn, deltavcn), None) ::
393 parse_runlist (vcn +^ deltavcn) lcn rest
395 | { (* Really these fields are signed, but we'll just limit it to
396 * sensible values in the when clause instead.
400 deltavcn : vcnlen * 8 : littleendian;
401 deltalcn : lcnlen * 8 : littleendian;
402 rest : -1 : bitstring
403 } when (vcnlen >= 1 && vcnlen <= 4) && (lcnlen >= 1 || lcnlen <= 4) ->
405 let deltavcn = Int63.of_int64 deltavcn in
406 let deltalcn = Int63.of_int64 deltalcn in (* XXX signed *)
408 let lcn = lcn +^ deltalcn in
410 ((vcn, deltavcn), Some lcn) ::
411 parse_runlist (vcn +^ deltavcn) lcn rest
415 eprintf "unknown field in the runlist\n%!";
416 Bitmatch.hexdump_bitstring Pervasives.stderr bits
420 (* Poor man's little-endian UCS-2 to UTF-8 conversion.
421 * XXX Should use Camomile.
423 and ucs2_to_utf8 name len =
424 (* Calculate length of final string. *)
425 let outlen = ref 0 in
427 for i = 0 to len-1 do
430 let c0 = Char.code name.[j'] and c1 = Char.code name.[j'+1] in
431 let c = c0 + c1 * 256 in
432 if c < 128 then incr outlen
433 else if c < 0x800 then outlen := !outlen + 2
434 else outlen := !outlen + 3
436 let outstr = String.create !outlen in
438 for i = 0 to len-1 do
441 let c0 = Char.code name.[j'] and c1 = Char.code name.[j'+1] in
442 let c = c0 + c1 * 256 in
444 outstr.[!outlen] <- Char.chr c;
446 ) else if c < 0x800 then (
447 outstr.[!outlen] <- Char.chr (0b11000000 lor (c lsr 6));
448 outstr.[!outlen+1] <- Char.chr (0b10000000 lor (c land 0b00111111));
449 outlen := !outlen + 2
451 outstr.[!outlen] <- Char.chr (0b11100000 lor (c lsr 12));
452 outstr.[!outlen+1] <- Char.chr (0b10000000 lor ((c lsr 6) lor 0b00111111));
453 outstr.[!outlen+2] <- Char.chr (0b10000000 lor (c land 0b00111111));
454 outlen := !outlen + 3
459 (* Parse $Bitmap::$Data to get free/used. Returns (used, free) blocks. *)
460 and parse_bitmap_freespace ntfs =
461 (* Can throw Not_found - allow that to escape because we don't
462 * expect an NTFS filesystem without this magic file.
464 let file = find_system_file ntfs "$Bitmap" in
466 (* Count used/free bits. *)
467 let used = ref ~^0 and free = ref ~^0 in
468 iter_blocks ntfs file (
470 for i = 0 to String.length data - 1 do
471 let c = Char.code data.[i] in
472 if c = 0 then (* common cases *)
474 else if c = 0xff then
476 else ( (* uncommon case: count the bits *)
479 if c land !m <> 0 then
482 free := !free +^ ~^1;
490 and find_system_file { ntfs_mft_records = mft_records } fname =
493 | [] -> raise Not_found
494 | ({ ntfs_filename = Some { ntfs_name = name } } as file) :: _
497 | _ :: rest -> loop rest
501 and iter_blocks { ntfs_blocksize = blocksize; ntfs_dev = dev }
502 { ntfs_data = data } f =
504 | None -> () (* No $Data attribute. *)
505 | Some { ntfs_data_size = data_size; ntfs_runlist = runlist } ->
506 let rec loop data_size = function
509 (* Run of vcnsize clusters. *)
510 | ((vcnstart, vcnsize), Some lcn) :: rest ->
511 let data_size = ref data_size in
513 let vcn = ref vcnstart in
514 let vcnsize = ref vcnsize in
515 while !vcnsize > ~^0 && !data_size > ~^0 do
516 let size = min blocksize !data_size in
517 let data = dev#read (!lcn *^ blocksize) size in
518 f (Some !lcn) !vcn data;
521 vcnsize := !vcnsize -^ ~^1;
522 data_size := !data_size -^ size
527 | ((vcnstart, vcnsize), None) :: rest ->
528 let data_size = ref data_size in
529 let vcn = ref vcnstart in
530 let vcnsize = ref vcnsize in
531 while !vcnsize > ~^0 && !data_size > ~^0 do
532 let size = min blocksize !data_size in
533 let data = String.make size '\000' in
536 vcnsize := !vcnsize -^ ~^1;
537 data_size := !data_size -^ size
541 loop data_size runlist
543 (* This is a bit limited at the moment because it can only read from
544 * a contiguous part of the file. System files are usually contiguous
545 * so this is OK for us.
547 and read_file { ntfs_blocksize = blocksize; ntfs_dev = dev }
548 { ntfs_data = data } offset size =
550 | None -> raise Not_found (* No $Data attribute. *)
551 | Some { ntfs_data_size = data_size; ntfs_runlist = runlist } ->
552 if offset < ~^0 || size < ~^0 || offset +^ size >= data_size then
553 invalid_arg "ntfs: read_file: tried to read outside file";
555 (* Get the first and last VCNs containing the data. *)
556 let vcn = offset /^ blocksize in
557 let vcnoffset = offset %^ blocksize in
558 let vcnend = (offset +^ size -^ ~^1) /^ blocksize in
560 (* Find the run containing this VCN. *)
561 let rec find = function
562 | [] -> raise Not_found
563 | ((vcnstart, vcnsize), lcn) :: _
564 when vcnstart <= vcn && vcn < vcnstart +^ vcnsize &&
565 vcnstart <= vcnend && vcnend < vcnstart +^ vcnsize ->
567 | _ :: rest -> find rest
569 let lcn = find runlist in
574 | Some lcn -> dev#read (lcn *^ blocksize +^ vcnoffset) size
575 | None -> String.make size '\000' (* sparse hole *) in
578 (* This is easy: just look at the bitmap. *)
579 and offset_is_free fs offset =
581 let ntfs = get_private_data fs in
582 let blocksize = ntfs.ntfs_blocksize in
584 (* Get the $Bitmap file. *)
585 let file = find_system_file ntfs "$Bitmap" in
587 let lcn = offset /^ blocksize in
589 (* Read the byte in the bitmap corresponding to this LCN. *)
590 let byteoffset = lcn >^> 3 and bitoffset = lcn &^ ~^7 in
591 let byte = read_file ntfs file byteoffset ~^1 in
592 let byte = Char.code byte.[0] in
593 let bit = byte >^> (~^0x80 >^> (Int63.to_int bitoffset)) in
597 Not_found -> false (* play it safe *)
602 fs_cb_uq = (incr i; !i);
604 fs_cb_printable_name = "Windows NTFS";
605 fs_cb_offset_is_free = offset_is_free;
608 (* Register the plugin. *)
609 let () = register_plugin ~filesystem:probe id