------------------------------------------------------------------------------- Note: This is about how libburn operates optical drives. Not about how to operate libburn. The libburn API is described in libburn/libburn.h ------------------------------------------------------------------------------- libburnia-project.org Optical Media Rotisserie Recipes as of January 2010 Content: - TAO Multi-Session CD Cookbook (CD-R, CD-RW) - SAO CD Cookbook (CD-R, CD-RW, pure audio or pure data only) - Overwriteable DVD Cookbook (DVD-RAM, DVD+RW, DVD-RW, BD-RE) - Sequential DVD-R[W] Cookbook - DVD+R[/DL] Cookbook - BD-R Cookbook ------------------------------------------------------------------------------- TAO Multi-Session CD Cookbook ------------------------------------------------------------------------------- Guided by reading mmc-r10a.pdf , O.8 "Write a Track" from http://www.t10.org/ftp/t10/drafts/mmc/ backed by reading mmc5r03c.pdf from http://www.t10.org/ftp/t10/drafts/mmc5/ by reading mmc3r10g.pdf from http://www.t10.org/ftp/t10/drafts/mmc3/ by reading spc3r23.pdf from http://www.t10.org/ftp/t10/drafts/spc3/ by reading libburn/* from http://icculus.org/burn and by experiments with drives NEC ND-4570A, LG GSA-4082B, LITE-ON LTR48125S which used in part code from http://icculus.org/burn. For libburnia-project.org by Thomas Schmitt ------------------------------------------------------------------------------- Media type can be recognized by Current Profile from 46h GET CONFIGURATION. (mmc5r03c.pdf 6.6.2.1) CD-R 0009h CD-RW 000ah The following topics are covered in this text: - About blank, appendable and finalized CD media - Writing a session to CD in TAO mode - Obtaining CD multi-session info for extending ISO-9660 filesystems - Obtaining a Table Of Content from CD ------------------------------------------------------------------------------- About blank, appendable and finalized CD media : CD media have to be blank or appendable in order to be writeable in TAO mode. The according status may be inquired by 51h READ DISC INFORMATION requesting Data Type 000b Standard Disc Information, where reply value Disc Status indicates: 00b blank 01b appendable 10b finalized 11b others (unsuitable for this recipe) (mmc5r03c.pdf 6.22.3.1.4) CD-RW which are finalized or appendable may be blanked by command A1h BLANK with blanking types 000b "Blank the disc" or 001b "Minimally blank the disc". The Start Address/Track Number will be ignored so it may well be 0. Because the operation is long running it is advised to set the Immed bit and to watch the progress by commands 00h TEST UNIT READY and 03h REQUEST SENSE with DESC bit set to 0 for fixed format reply. It is done when 00h succeeds and 03h reports 0 in PROGRESS INDICATION (byte 1+2 in Table 22 = byte 16+17 SENSE KEY SPECIFIC in table 26). (mmc5r03c.pdf 6.2 BLANK) (spc3r23.pdf 4.5.2.4.4 table 22, 4.5.3 table 26, 6.27 REQUEST SENSE, 6.33 TEST UNIT READY) ------------------------------------------------------------------------------- Writing a session to CD in TAO mode : The writing method for blank or appendable media is the same. A new session will get created automatically by the first track when it is written. If the media is blank then the new session will be the first and only one in the table of content. If the media is appendable then a new session will be appended to the existing sessions. In any case the new track will be the first one in the new session. Speed may be set by BBh SET CD SPEED parameter Drive Write Speed. Note that kbytes/sec means 1000 bytes/sec and not 1024/sec. Rotational control should be set to 00b. 1x CD speed is 176.4 kbytes/sec. Speed is usually set to the next lower possible value by the drive. So it is helpful to add a few kbytes/sec just in case the drive has rounding problems. (mmc5r03c.pdf 6.37) Before writing can occur, a Write Parameters mode page 05h has to be composed and transmitted via 55h MODE SELECT. Mode page 05h describes several burn parameters: BUFE Buffer Underrun protection 0=off, 1=on Test Write -dummy mode for writing 0=off, 1=on Write Type Packet/TAO/SAO/RAW 01h = TAO Multi-session Whether to keep appendable 00b = finalize 11b = keep appendable Track Mode Describes frame type 4 for data , 0 for audio Data Block Type Layout of payload blocks 8 for 2048 byte data blocks 0 for 2352 byte audio blocks Audio Pause Length 150 = 2 seconds Any other parameters may be set to 0. Mode page data as of MMC-5 table 644 are preceded by a Mode Parameter Header as of SPC-3 table 240. This 8-byte header may be filled with zeros. (mmc5r03c.pdf 7.5.4 The Mode Page, 4.2.3.4 Table 17 CONTROL = Track Mode) (spc3r23.pdf 6.8 MODE SELECT, 7.4.3 Mode parameter header formats) Writing has to begin at the address returned by 52h READ TRACK INFORMATION with Address/Number Type set to 01b and Logical Block Address/Track/Session Number set to FFh. The Next Writeable Address as of table 500 is the number to start writing with. (mmc5r03c.pdf 6.27 ) Writing is performed by one or more 2Ah WRITE transactions with the Logical Block Address counted up from the initial number in sync with the number of blocks written. I.e the Transfer Length of the previous 2Ah WRITE has to be added to the Logical Block Address for the next 2Ah WRITE. Only full blocks can be written. (mmc5r03c.pdf, 6.44) When writing is done, it is mandatory to force the drive's buffer to media by 35h SYNCHRONIZE CACHE. (mmc5r03c.pdf, 6.41) A track must at least contain 300 payload blocks: 4 seconds of audio or 600 KiB of data. (mmc5r03c.pdf 6.3.3.1.2) Up to december 2009 the track was closed by 5Bh CLOSE TRACK SESSION Close Function 001b. Older MMC specifies a valid Logical Track Number FFh to depict the open track. MMC-5 is quite silent about this. FFh worked for my drives. (mmc5r03c.pdf 6.3.3.1.2) This is omitted since libburn-0.7.4, relying entirely on 35h SYNCHRONIZE CACHE. First appeared a drive where CLOSE TRACK fails in simulation mode, later another one produced error replies even with real burning. After that, a new track may be written beginning with sending the mode page 05h again. It is not tested whether 05h can be omitted if Track Mode and Data Block Type are the same as with the previous track. The new track will be added to the session which was opened by the first track. After the last track of a session, 5Bh CLOSE TRACK SESSION Close Function 010b with Logical Track Number 0 closes the session. It depends on the Multi-Session value in mode page 05h whether the disc is finalized or stays appendable. (mmc5r03c.pdf 6.3.3.1.3) ------------------------------------------------------------------------------- Obtaining CD multi-session info for extending ISO-9660 filesystems : Program mkisofs expects two numbers with its option -C which describe the situation on an appendable CD which already contains a ISO-9660 filesystem in the first track of the last session. The first number is the Logical Block Address of that track containing the existing ISO-9660 filesystem image. This number is needed for mkisofs option -M to connect to the existing image. The new image will refer to files in the previously existing image. mkisofs option -M needs read access to the CD or a blockwise copy of it on hard disk. The number is gained by 43h READ TOC/PMA/ATIP. (mmc5r03c.pdf 6.26) Untested is Format 0001b which in table 478 promises quick access via Start Address Of First Track In Last Session. (mmc5r03c.pdf 6.26.2.5 table 478, 6.26.3.3.1) libburn gets the number from its Table Of Content model which is obtained by 43h READ TOC/PMA/ATIP, Format 0010b. See below. The second number is an exact prediction of the Logical Block Address of the new track which will contain the newly generated ISO-9660 image. Even without mkisofs option -M this second number is still needed to make the inner block address pointers of the image match the Logical Block Addresses on CD. For that one may inquire 52h READ TRACK INFORMATION with Address/Number Type set to 01b and Logical Block Address/Track/Session Number set to FFh. The Next Writeable Address as of table 500 is the number to use. (mmc5r03c.pdf 6.27 ) ------------------------------------------------------------------------------- Obtaining a Table Of Content from CD : The structure of a CD is comprised of sessions. Each session contains one or more tracks and is followed by a lead-out. A track has an address and a length. Table of content information is gained by 43h READ TOC/PMA/ATIP, Format 0010b. (mmc5r03c.pdf 6.26.2.5 table 478) The number of sessions is given by Last Complete Session Number. The number of TOC Track descriptors is: (TOC Data Length - 2)/11 . Each TOC Track Descriptor contains a Session Number. If POINT is >= 1 and <= 99 (63h) then the descriptor is about the track of which POINT tells the number. The start address of this track can be read from PMIN, PSEC, PFRAME where it is encoded in MSF format: If M is smaller than 90: LBA = (M * 60 + S) * 75 + F - 150 Else : LBA = (M * 60 + S) * 75 + F - 450150 The length of the track is given by MIN,SEC,FRAME in the same format. If POINT = A0h then the descriptor tells in PMIN the first track number of its session. POINT = A1h tells in PMIN the last track number of its session. POINT = A2h describes in PMIN, PSEC, PFRAME the lead-out of a session, i.e the first address after the session's end. (Next writeable address typically is lead-out + 11400 after the first session, lead-out + 6900 after further sessions.) POINT = B0h tells in MIN,SEC,FRAME this next writeable address or FFh,FFh,FFh for finalized disc. (mmc5r03c.pdf 6.26.3.4 table 489, 4.2.3.7 Mode-1 Q, Mode-5 Q) In libburn the address of the first track in the last session is obtained from the last session's POINT = A0h and from the track descriptor with the POINT value matching the PMIN value of the A0h descriptor. Untested is whether POINT = B0h and 52h READ TRACK INFORMATION are always in sync. libburn uses the info provided by 52h READ TRACK INFORMATION. ------------------------------------------------------------------------------- ------------------------------------------------------------------------------- SAO CD Cookbook ------------------------------------------------------------------------------- Guided by reading libburn/* from http://icculus.org/burn backed by reading mmc5r03c.pdf from http://www.t10.org/ftp/t10/drafts/mmc5/ and by experiments with drives NEC ND-4570A, LG GSA-4082B, LITE-ON LTR48125S which used in part code from http://icculus.org/burn. For libburnia-project.org by Thomas Schmitt ------------------------------------------------------------------------------- Recognition of media type and state (blank, appendable, finalized) is as described in the TAO Multi-Session CD Cookbook. See there. The MMC specs do not give much hint about the combination of SAO and multi-session. My drives refused not only on a few experiments which i did in libburn but also failed with cdrecord -sao on an appendable CD. So for now only blank CD seem to be suitable for SAO writing. Different from TAO mode, the whole session layout is announced to the drive by sending a Cue Sheet. This implies that the sizes of the tracks have to be known in advance, which is a heavy drawback when dealing with track data sources like stdin, named pipes or sockets. Nevertheless, SAO seems to be best writing mode for audio purposes, as our audio expert Lorenzo Taylor found out. A SAO session in libburn may either consist entirely of audio tracks or entirely of data tracks. For mixed sessions, only TAO is usable yet. - Composing a SAO CD Cue Sheet (either audio or data, but not mixed) - Writing the prepared SAO CD session - What is known about mixed mode sessions ------------------------------------------------------------------------------- Composing a Cue Sheet (either audio or data, but not mixed) : The Cue Sheet will get submitted to the drive by 5Dh SEND CUE SHEET. Each entry of the sheet is of 8 bytes size. Its fields are named CTL|ADR, TNO, INDEX, DATA FORM, SCMS, MIN, SEC, FRAME . (mmc5r03c.pdf 6.33) CTL is 40h for data and 00h for audio. (mmc5r03c.pdf 6.33.3.4) ADR is always 01h. TNO is the track number (1 to 99). INDEX is a subaddress within tracks. This recipe uses only INDEX 01h within tracks. (mmc5r03c.pdf 4.2.3.5.2) DATA FORM is 00h for audio payload, 01h for audio pause, 10h for data, 41h for CD-TEXT in Lead-in. (mmc5r03c.pdf 6.33.3.11 CD-DA Data Form, 6.33.3.12 CD-ROM mode 1 Form) SCMS is always 00h. MIN, SEC, FRAME give the MSF address where the described data entity starts. LBA = frames - 150, 75 frames = 1 sec , 60 sec = 1 min. This address must increase from entry to entry (or at least stay equal). The first entry describes the Lead-in. Its content is (CTL|ADR ,00h,00h, DATA FORM ,00h,00h,00h,00h) With the CTL|ADR for the first track: 41h for data, 01h for audio. DATA FORM is 41h if CD-TEXT shall be stored in Lean-in. Else it is 01h. The LBA for the first write is negative: -150. This corresponds to MSF address 00h:00h:00h. All addresses are to be given in MSF format. The first information track on disc is preceded by a pause encoding of 2 sec: (CTL|ADR,01h,00h, DATA FORM ,00h,00h,00h,00h) with DATA FORM = 00h for audio and 10h for data. By those 2 seconds the MSF address increases to 00h:02h:00h = LBA 0. Each track is represented by an entry (CTL|ADR, TNO ,01h,DATA FORM,00h, MIN , SEC , FRAME) TNO gives the track number. MIN, SEC, FRAME give the MSF address which becomes the start address of the track. The MSF address is then increased by the size of the track (to be used with next track or with lead-out). A track must at least contain 300 payload blocks: 4 seconds of audio or 600 KiB of data. (mmc5r03c.pdf 6.33.3.6) At the end of the session there is a lead-out entry (CTL|ADR,AAh,01h,01h,00h,MIN,SEC,FRAME) marking the end of the last track. (With libburn CTL is as of the last track.) ------------------------------------------------------------------------------- Writing the prepared session : Speed may be set by BBh SET CD SPEED parameter Drive Write Speed. Note that kbytes/sec means 1000 bytes/sec and not 1024/sec. Rotational control should be set to 00b. 1x CD speed is 176.4 kbytes/sec. Speed is usually set to the next lower possible value by the drive. So it is helpful to add a few kbytes/sec just in case the drive has rounding problems. (mmc5r03c.pdf 6.37) If CD-TEXT shall be written into Lead-in, then it is necessary to obtain the Start Time of Lead-in by 43h READ TOC/PMA/ATIP Format 0100b. It is an MFS address which varies from media manufacturer to media manufacturer. Minute will be >= 90. Therefore this conversion applies: LBA = (M * 60 + S) * 75 + F - 450150 A Write Parameters mode page 05h has to be composed and transmitted via 55h MODE SELECT. This page describes the following parameters: BUFE Buffer Underrun protection 0=off, 1=on Test Write -dummy mode for writing 0=off, 1=on Write Type Packet/TAO/SAO/RAW 02h = SAO Multi-session Whether to keep appendable 00b = finalize 11b = keep appendable Track Mode Describes frame type 0 (is ignored) Data Block Type Layout of payload blocks 0 (is ignored) Audio Pause Length 150 = 2 seconds (ignored ?) Any other parameters may be set to 0. Mode page data as of MMC-5 table 644 are preceded by a Mode Parameter Header as of SPC-3 table 240. This 8-byte header may be filled with zeros. (mmc5r03c.pdf 7.5.4 The Mode Page, 4.2.3.4 Table 17 CONTROL = Track Mode) (spc3r23.pdf 6.8 MODE SELECT, 7.4.3 Mode parameter header formats) The Cue Sheet is submitted to the drive by 5Dh SEND CUE SHEET. Cue Sheet Size is 8 times the number of entries. (mmc5r03c.pdf 6.33) Writing is performed by multiple 2Ah WRITE transactions with the Logical Block Address counted up from the initial number in sync with the number of blocks written. I.e the Transfer Length of the previous 2Ah WRITE has to be added to the Logical Block Address for the next 2Ah WRITE. Only full blocks can be written. (mmc5r03c.pdf, 6.44) Block addresses may be negative for areas before the normally readable data. Data representation of addresses is 4-byte, big-endian, two's-complement. E.g: -150 = FFh FFh FFh 6Ah. This is the natural form found with about any 32-bit processor, so only the endianness has to be taken into respect when converting a 32-bit integer into a LBA for command 2Ah WRITE. If CD-TEXT shall be written into Lead-in, then writing begins at the start address of Lead-in, which was obtained above. The 18 bytes of each text pack have to be split up to 24 bytes with only the lowest six bits used in each byte. E.g. text pack 8F 00 2A 00 01 01 03 00 06 05 04 05 07 06 01 02 48 65 becomes 23 30 00 2A 00 00 04 01 00 30 00 06 01 10 10 05 01 30 18 01 00 24 21 25 4 of these 24 byte packs form a block of DATA FORM 41h. I.e. only 96 bytes payload per block. The whole range from Lead-in start to LBA -150 has to be filled with blocks of this form. Therefore it is necessary to write the list of given packs in repeated cycles. A typical Lead-in start address is -11635 = FFh FFh D2h 8Dh. Writing without CD-TEXT begins at LBA -150 = FFh FFh FFh 6Ah. In both cases, the mandatory pause preceding the first track has to be written as 150 blocks of the matching sector size: 2048 for data, 2352 for audio. By this, the LBA increases from -150 to 0. Next the tracks' payload is sent. For each track exactly the number of blocks has to be transmitted as is announced in the Cue Sheet by the difference of the track's own start address and the start address of the next entry in the Cue Sheet. After each write the LBA for the next write has to be increased by the number of blocks transmitted. Just like with TAO writing. There is no separator between the tracks of a pure mode SAO session. (If the session was mixed mode, there would be extended Pre-gaps and Post-gaps between data mode tracks and audio mode tracks.) (libburn sends its own buffer to the drive at the end of each track but does not sync the drive's chache. It is unclear whether this separation of tracks on the level of 2Ah WRITE is necessary with a pure mode session. It does not harm in any case and would probably be unavoidable if audio and data tracks were mixed.) When writing of all tracks is done, it is mandatory to force the drive's buffer to media by 35h SYNCHRONIZE CACHE. (mmc5r03c.pdf, 6.41) No further finalization is necessary. (I.e. no 5Bh CLOSE TRACK SESSION.) ------------------------------------------------------------------------------- Obtaining CD-TEXT from Lead-in : Audio CDs may contain CD-TEXT information in their Lead-in. It is gained by 43h READ TOC/PMA/ATIP, Format 0101b. The reply consists of 4 bytes header, of which the first two bytes give the number of following bytes as big-endian 16 bit number. The other two bytes are 0. Following are text packs of 18 bytes each. (mmc5r03c.pdf 6.26.3.7.1 table 495) The format of a text pack is explained in (mmc3r10g.pdf, appendix J). Each pack of a 4-bytes are header, 12 byte pieces of 0-terminated texts or binary data, and 2 bytes of CRC. The first byte of each pack tells the pack type (text meaning): 0x80 = Title 0x81 = Names of performers 0x82 = Songwriters 0x83 = Composers, 0x84 = Arrangers 0x85 = Messages 0x86 = text-and-binary: Disc Identification 0x87 = text-and-binary: Genre Identification 0x88 = binary: Table of Content information 0x89 = binary: Second Table of Content information 0x8e = UPC/EAN code of the album and ISRC code of each track 0x8f = binary: Size Information of the Block The second byte tells the track number to which the first text piece in a pack is associated. Number 0 means the whole album. Higher numbers are valid for types 0x80 to 0x85, and 0x8e. With these types, there should be one text for the disc and one for each track. The third byte is a sequential counter. The fourth byte is the Block Number and Character Position Indicator. It consists of three bit fields: bit7 = Double Bytes Character Code (0= single byte characters) bit4-6 = Block Number (groups text packs in language blocks) bit0-3 = Character position. Either the number of characters which the current text inherited from the previous pack, or 15 if the current text started before the previous pack. A text may span over several packs. Unused characters in a pack are used for the next text of the same pack type. If no text of the same type follows, then the remaining text bytes are set to 0. The CRC algorithm is known as CRC-16-CCITT with divisor 0x11021. MMC-3 says: "All bits shall be inverted." ---------------------------------------------------------------------------- What is known about mixed mode sessions : For now, SAO sessions with a mix of data and audio are not supported in libburn. Here are the reasons why. Obviously the code of http://icculus.org/burn is incomplete in this aspect. In mmc5r03c.pdf comparison of table 555 and 6.33.3.18 seems self-contradicting. (The second Pre-gap in table 555 does not match any of the criteria of 6.33.3.18. Also, there is no Post-gap shown in table 555 although 6.33.3.19 would prescribe some.) If a data track follows an audio track then the data track gets a preceding extended Pre-gap: (CTL|ADR, TNO ,01h,DATA FORM,00h, MIN , SEC , FRAME) with TNO already the number of the data track. The MSF address is to be increased by 3 seconds. The first second of the extended Pre-gap needs to be written in the audio track's mode and the other 2 seconds are to be written in the data track's mode. (libburn compares DATA FORM rather than burn_track.mode . Wrong ?) (libburn currently does only 2 seconds and the second part of Pre-gap. There is an issue with burn_track.pregap1 about this. Seems libburn mistakes the pause preceding track 1 for a part 2 of an extended Pre-gap.) If a data track is followed by an audio track then it gets a Post-gap of at least two seconds. No example of Post-gap is given for Cue Sheet. Maybe it is to be added to the track, or maybe it gets an own Cue Sheet entry ... who knows ? (libburn contains write code for pregap1, pregap2 and postgap. But only pregap2 ever gets activated. Once hackingly for the first 2 second pause, once incompletely for a change of DATA FORM.) Seems nobody ever tested this. Libburnia simply knows no use case where the correctness of Pre-gap and Post-gap would become evident. ------------------------------------------------------------------------------- ------------------------------------------------------------------------------- Overwriteable DVD Cookbook ------------------------------------------------------------------------------- Inspired by Andy Polyakov's http://fy.chalmers.se/~appro/linux/DVD+RW/tools , backed by reading mmc5r03c.pdf from http://www.t10.org/ftp/t10/drafts/mmc5/ by own experiments with drives NEC ND-4570A, LG GSA-4082B, PHILIPS SPD3300L, LG GGW H20L, and by BD-RE experiments done by Giulio Orsero on LG BE06LU10. For libburnia-project.org by Thomas Schmitt ------------------------------------------------------------------------------- Media type can be recognized by Current Profile from 46h GET CONFIGURATION. (mmc5r03c.pdf 6.6.2.1) DVD-RAM 0012h DVD-RW Restricted Overwrite 0013h DVD-RW Sequential Recording 0014h (i.e. unformatted) DVD+RW 001Ah BD-RE 0043h A short compilation of the write model: - Overwriting in general The recipes described here are depending on formatting state: - DVD-RAM, fully formatted DVD+RW, DVD-RW, BD-RE - Unformatted DVD+RW - Partly formatted DVD+RW - Unformatted DVD-RW - Partly formatted DVD-RW - Intermediate state DVD-RW - DVD-RAM and BD-RE formatting - DVD-RAM and BD-RE speed tuning Slightly off topic are - ISO 9660 multi-session emulation on overwriteable media - ISO 9660 based TOC emulation on overwriteable media ------------------------------------------------------------------------------- Overwriting in general : Depending on media type, some kind of formatting has to have happened before data can be written. Formatting may happen separately from writing or simultaneously. See the particular recipes below. No Write Parameters mode page 05h is to be sent. Speed can be influenced by B6h SET STREAMING , speed capabilities can be inquired by ACh GET PERFORMANCE. It is advised to set only speeds and sizes which are returned by ACh. (mmc5r03c.pdf 6.39 SET STREAMING, 6.8 GET PERFORMANCE) Optimal performance is promised without any speed setting. But my experiments showed that SET STREAMING values persist after media change. In the formatted area of the media, coarse random access is possible. For DVD-RAM, DVD+RW, BD-RE write addresses and data size need to be aligned to 2 KiB. For DVD-RW alignment has to be 32 KiB. Within these limitations the write address is at the discretion of the sending program. Just use 2Ah WRITE to send data. (mmc5r03c.pdf, 6.44) When writing is done, it is mandatory to force the drive's buffer to media by 35h SYNCHRONIZE CACHE. (mmc5r03c.pdf, 6.41) The size of the formatted area can be inquired by 23h READ FORMAT CAPACITIES. The Number Of Blocks value in the Current/Maximum Capacity Descriptor gives this size in 2 KiB blocks. But this is true only if Descriptor Type is 10b ("Formatted Media"). (mmc5r03c.pdf, 6.24.3.2.1, 6.24.3.2.3) Not yet formatted areas may be completely forbidden or they may be allowed for sequential writing (DVD-RW Intermediate state) or they may be allowed for random access only after the necessary waiting time for formatting to reach the desired address (DVD+RW with background formatting active). Already written areas can be overwritten without special precaution. Blanking a DVD-RW actually destroys its formatting. Most of the concepts usually expressed in Write Parameters mode page 05h do not apply to the recipes here: Test-Write, Buffer Underrun protection, Multi-session, Write Type, Block Type, Track Mode, ... There are hints for multi-session formats with DVD-RW but both of my drives do not offer "Add Session" Format Types 12h or 14h. (mmc5r03c.pdf 6.5.4.2.7 , 6.5.4.2.9) Caution: Drive and media compatibility seems still to be quite an adventure. If you experience problems, especially problems with readability, then try different drives and media brands. Failure does not necessarily mean that the software did anything wrong. ------------------------------------------------------------------------------- DVD-RAM, fully formatted DVD+RW, DVD-RW, BD-RE : Full format is the natural state of DVD-RAM. BD-RE are sold unformatted and need to be fully formatted first. See paragraph about DVD-RAM and BD-RE formatting below. DVD+RW reaches this state if Background Formatting is allowed to finish without being stopped by 5Bh CLOSE TRACK SESSION. (mmc5r03c.pdf, 6.5 FORMAT UNIT, 6.5.4.2.14 Format Type = 26h) The formatting state of a DVD+RW may be inquired by 51h READ DISC INFORMATION requesting Data Type 000b "Standard Disc Information". In the reply, BG Format 3 indicates fully formatted media. (mmc5r03c.pdf 6.22.3.1.13) DVD-RW reaches this state either by Format Type 00h (or 10h) with maximum size given as Number Of Blocks, or by writing sequentially until the disc is completely full into an intermediate session opened by format 15h resp. 13h. (mmc5r03c.pdf, 6.5 FORMAT UNIT, 6.5.4.2.1, 6.5.4.2.10, 6.5.4.2.8) A fully formatted DVD-RW can be recognized by 23h READ FORMAT CAPACITIES. The Descriptor Type of the Current/Maximum Capacity Descriptor is 10b ("Formatted Media") and 0 blocks are offered with Format Types 13h or 11h. (mmc5r03c.pdf, 6.24.3.2.1, 6.24.3.3) See also discussion of unformatted or partially formatted DVD-RW below. In fully formatted state there is no need for any formatting before writing nor for any finalizing other than forcing the drive's buffer to media by 35h SYNCHRONIZE CACHE (which is mandatory for writing, anyway). (mmc5r03c.pdf, 6.41) (It seems to do no harm to send to DVD+RW or DVD-RW a 5Bh CLOSE TRACK SESSION with Close Function 010b despite there is no session open in this scenario.) ------------------------------------------------------------------------------- Unformatted DVD+RW : This is the state of previously unused DVD+RW media. The formatting state of a DVD+RW may be inquired by 51h READ DISC INFORMATION requiring Data Type 000b "Standard Disc Information". In the reply, BG Format 0 indicates unformatted media (or unsuitable media). (mmc5r03c.pdf 6.22.3.1.13) Formatting has to be started by command 04h FORMAT UNIT, Format Type 26h. Different from other format types, 26h allows to send a fantasy size of 0xffffffff blocks and does not require the caller to know the exact maximum size offered with that format. (mmc5r03c.pdf, 6.5 FORMAT UNIT, 6.5.4.2.14 Format Type = 26h) As its name suggests, one has not to wait for background formatting to end but may very soon start writing as on formatted media. Random access to yet unformatted areas can last long, though. If backup formatting has been started at the beginning of the session, then it may get stopped after the final cache sync by 5Bh CLOSE TRACK SESSION with Close Function 010b. (mmc5r03c.pdf 6.3.3.6) Formatting of DVD+RW is called "de-icing" because unformatted areas offer no hold for random access addressing and are thus slippery like ice. One can also see a color change from shiny unformatted to more dull formatted media. ------------------------------------------------------------------------------- Partly formatted DVD+RW : This state is achieved by stopping background formatting before the media was completely formmatted. The formatting state of a DVD+RW is obtained by 51h READ DISC INFORMATION requiring Data Type 000b "Standard Disc Information". In the reply, BG Format 1 indicates partly formatted media. (mmc5r03c.pdf 6.22.3.1.13) If the data of the session surely fit into the formatted area, then it would be unnecessary to restart background formatting. But in order to make the DVD+RW surely accept its maximum number of bytes, formatting may be restarted by command 04h FORMAT UNIT, Format Type 26h, with the Restart bit set and Number of Blocks set to 0xffffffff. (mmc5r03c.pdf, 6.5 FORMAT UNIT, 6.5.4.2.14 Format Type = 26h) From then on, the same rules apply as for previously unformatted DVD+RW. ------------------------------------------------------------------------------- Unformatted DVD-RW (media profile is 0014h) : This state is present with previously unused media. It is also present with media blanked by programs cdrecord, wodim or dvd+rw-format and with media which were sequentially written from blank state. Profile transition from formatted 0013h to unformatted 0014h is done by A1h BLANK. (mmc5r03c.pdf, 6.2) Experiments on my drives indicate that only Blanking Type 000b "Blank the disc" achieves neat media. Media blanked via type 001b offer no feature 0021h and stall cdrecord or libburn already when those media get examined. growisofs can burn them - but only via DAO (feature 002Fh which prescribes Write Type 2). (mmc5r03c.pdf 5.3.11, 5.3.25) For becoming overwriteable such media need to be treated by command 04h FORMAT UNIT. (mmc5r03c.pdf, 6.5) The Format Type has to be chosen from the list replied by 23h READ FORMAT CAPACITIES. Suitable are Format Types 00h, 10h, 15h. (mmc5r03c.pdf 6.24) Format Types 00h and 10h provide a writeable area of a size given by Number of Blocks. Type 00h seems to be the most traditional and complete one. It needs no closing of a session at the end of writing. The Number Of Blocks may be at most the value reported by 23h READ FORMAT CAPACITIES in the entry for the desired format type. Full format is achieved by sending exactly the reported value. (mmc5r03c.pdf, 6.5.4.2.1 Format Type = 00h, 6.5.4.2.5 Format Type = 10h) Format Type 15h provides a writeable area of given size too, but this area can be expanded by sequential writing and afterwards marked as overwriteable by closing the session. It is even allowed to format with size 0 and to leave the size claim entirely to a sequential write process beginning at LBA 0. (mmc5r03c.pdf, 6.5.4.2.10 Format Type = 15h) When writing is done and cache is synced, one should send 5Bh CLOSE TRACK SESSION with Close Function 010b in order to bring the session out of Intermediate state. (mmc5r03c.pdf 6.3.3.2.3) If not written up to the last 32 KiB block, the DVD-RW is only partly formatted after that. ------------------------------------------------------------------------------- Partly formatted DVD-RW (media profile is 0013h) : This state is achieved by formatting a DVD-RW with a number of blocks which is less than offered for the Format Type by the drive's reply to 23h READ FORMAT CAPACITIES. If the media was most recently formatted by Format Types 015h or 013h then it must have got written some bytes and afterwards treated by 5Bh CLOSE TRACK SESSION, 010b in order to be partly formatted. (mmc5r03c.pdf 6.3.3.2.3 CLOSE TRACK SESSION 010b, 6.24 READ FORMAT CAPACITIES) Elsewise the media is in Intermediate state. See below. A partly formatted DVD-RW can be recognized by 23h READ FORMAT CAPACITIES. The Descriptor Type of the Current/Maximum Capacity Descriptor is 10b ("Formatted Media") and the Number Of Blocks with formats 00h, 10h or 15h is larger than the currently formatted size, resp. more than 0 blocks are offered with Format Types 13h or 11h. (mmc5r03c.pdf, 6.24.3.2.1, 6.24.3.3) If the data of the session surely fit into the formatted area, then it would be unnecessary to do any further formatting. But in order to make the DVD-RW surely accept its maximum number of bytes, partial formatting may be expanded by command 04h FORMAT UNIT, Format Type 13h, which is supposed to be offered by the drive in this state. This brings the session again into Intermediate state and thus allows expansion by sequential writing. As with Format Type 15h it is ok to set Number Of Blocks to 0, so that no fixed size formatting work is done and writing can begin soon after. (mmc5r03c.pdf, 6.5.4.2.8 Format Type = 13h) When writing is done and cache is synced, one should send 5Bh CLOSE TRACK SESSION with Close Function 010b in order to bring the session out of Intermediate state. (mmc5r03c.pdf 6.3.3.2.3) If not written up to the last 32 KiB block, the DVD-RW is only partly formatted after that. Format Type 13h has been tested only with expanding sessions formatted by 15h. Nevertheless it is offered with sessions from 00h and 10h, too. According to the specs, Format Type 11h would expand a session by a fixed size. This has not been tested yet because it is less appealing than 13h. (mmc5r03c.pdf, 6.5.4.2.6 Format Type = 11h) ------------------------------------------------------------------------------- Intermediate state DVD-RW (media profile is 0013h) : This state is achieved by formatting a DVD-RW with Format Type 15h or 13h without subsequentially writing data and sending 5Bh CLOSE TRACK SESSION with Close Function 010b. Such media behave very unpleasing with my DVD-ROM drive under Linux 2.4 ide-cd. One should therefore better avoid to release media in this state. This state can be recognized by 23h READ FORMAT CAPACITIES. The Descriptor Type of the Current/Maximum Capacity Descriptor is 11b ("Unknown Capacity") and no formats 13h or 11h are offered. (mmc5r03c.pdf, 6.24.3.2.1, 6.24.3.3) One may treat such media as if Format Type 15h or 13h had been freshly applied. I.e. sequential writing from LBA 0. After cache sync bring the session out of Intermediate state by 5Bh CLOSE TRACK SESSION with Close Function 010b. (mmc5r03c.pdf 6.3.3.2.3) ------------------------------------------------------------------------------- DVD-RAM and BD-RE formatting : Although DVD-RAM usually are sold formatted, there may still arise the wish to adjust formatting. BD-RE are sold unformatted and need to be formatted prior to usage. Two format types are relevant for DVD-RAM : 00h and 01h. 00h offers the default size format and usually a maximum payload size format. Even with that maximum size payload there is hardware defect management. (mmc5r03c.pdf 6.5.4.2.1.2) 01h allows to convert payload capacity into spare blocks for defect managment. There is no way to increase payload capacity by format 01h. (mmc5r03c.pdf 6.5.4.2.2.1) With BD-RE there are three format types : 00h, 30h and 31h. 00h offers the default size format. This may be the only fast formatting mode that is offered by the drive. Feature 0023h tells whether format 31h and certain 30h subtypes are available. (mmc5r03c.pdf 5.3.13) 30h offers several sizes with defect management. Usually there are three sizes given: #1: default size, #2: maximum spare area, #3: minimal spare. One may demand any spare size between maximum and minimum. There may be quick certification and full certification. See feature 0023h. 31h offers a single size and disables defect management. This has the side effect to speed up writing to nominal speed. (mmc5r03c.pdf 6.5.4.2.15, 6.24.3.3, Table 472) Only format sizes from the list of format descriptors are permissible for DVD-RAM. The format list can be obtained by 23h READ FORMAT CAPACITIES. It also includes a description of the current formatting state. (mmc5r03c.pdf 6.24, 6.24.3.2, 6.24.3.3) Formatting is done by command 04h FORMAT UNIT. Its data payload consists of a Format List Header and a Format Descriptor. It is advisable to set the Immed bit and the FOV bit in header byte number 1. The descriptor should be a copy of a descriptor from 23h READ FORMAT CAPACITIES. (mmc5r03c.pdf 6.5, 6.5.3.2, 6.5.3.3) With nearly all formats Sub-type should be set to 0. But with BD-RE formats 30h and 31h the Sub-type selects the certification mode. Usable with 30h seem 10b Full Certification and 11b Quick Certification. Usable with 31h seem also 00b Quick Reformat and 01b No Certification. (mmc5r03c.pdf 6.5.4.2.15.1) Other format types have their certification intensity controlled by a pair of bits: CmpList and DCRT. CmpList resides in CDB byte 1 as bit 3. DCRT resides in the payload byte 1 as bit 5. Both together should request a quick size change without lengthy certification but maintaining the list of known defects. (mmc5r03c.pdf 6.5, table 249, 6.5.3.2) With DVD-RAM on my PHILIPS SPD3300L drive they prevent any format size change though. The TSSTcorp SH-S203B works properly. With BD-RE format 00h, the default is specified to be Quick Reformat, and with 00h in general certification can only be disabled not enabled. (mmc5r03c.pdf 6.5.4.2.1.7) ------------------------------------------------------------------------------- DVD-RAM and BD-RE tuning : A special aspect of DVD-RAM and BD-RE is their low speed with write operations, which usually is only half than the nominal speed of media and drive. This is blamed to the automatic checkreading which happens for managing eventual defects. Defect management of BD-RE can be disabled by format type 31h. See above. There is no drive known yet which would apply command 2Ah WRITE10 to DVD-RAM with full speed. The only known way to get full speed from DVD-RAM or BD-RE with enabled defect management is the use of AAh WRITE12 with Streaming Bit set to 1. (mmc5r03c.pdf 6.45) With some DVD-RAM drives this fails if a write buffer is not full 32 kB. With the tested BD-RE one must write full 64 kB buffers, or else writing might not get into effect at all. Although it seems not optimal, this is specified not only to disable the cumbersome checkread but also to ignore known defects and to write data to these defective addresses. (mmc5r03c.pdf 4.8.5) So the speed-up is only advisable as long as the media are free of incorrectable errors. Caveat: MMC-5 does not guarantee AAh WRITE12 to work on DVD-RAM or BD-RE at all. None of the features of profiles 0012h and 0043h promises the existence of AAh WRITE12. (mmc5r03c.pdf 5.4.13, 6.45) Nevertheless it worked on all tested drives if proper alignment and block size was observed. ------------------------------------------------------------------------------- ISO 9660 multi-session emulation on overwriteable media : Overwriteable media provide a single overwriteable track which may grow up to the full media capacity. There is no builtin table-of-content which records the history of write sessions. mount -t iso9660 will use sbsector=0 as default. The term "superblock" shall depict the first 64 KiB after the sbsector address. ISO 9660 multi-session depends on typical TOC information in two ways: It needs the superblock address MSC1 of the most recently recorded session and it needs the Next Writeable Address NWA for which to prepare the adress offset. The following is learned from growisofs and from ECMA-119: http://www.ecma-international.org/publications/files/ECMA-ST/Ecma-119.pdf ISO 9660 filesystems provide information about the number of sectors which is also the lowest unused block address and thus a suitable NWA. This block address is stored in the Primary Volume Descriptor which is supposed to be stored in block 16 (eventually relative to MSC1). The bytes 0 to 5 of a PVD block are 0x01 'C' 'D' '0' '0' '1' The sector count can then be read from byte 80 to 83 sectors= pvd[80] | (pvd[81] << 8) | (pvd[82] << 16) | (pvd[83] << 24); (Ecma-119.pdf 8.4) To support CD, DVD and BD media alike, it is advisable to round the NWA to the next multiple of 32 (= 64 KiB). So one can use 0 as MSC1 and prepare a new ISO session for the computed NWA. After writing the session it is necessary to patch the PVD at LBA 16. The minimal change would be to update the number of image sectors. It is stored in both notations LSB and MSB: for(i= 0; i < 4; i++) pvd[87 - i]= pvd[80 + i]= (sectors >> (8 * i)) & 0xff; cdrskin --grow_overwriteable_iso not only patches the sector fields of the PVD block but also the blocks up to LBA 31 which begin with 0xff 'C' 'D' '0' '0' '1' libisoburn submits 64 KiB data buffer to libisofs before image generation and afterwards writes these 64 KiB as new superblock to LBA 0. ------------------------------------------------------------------------------- ISO 9660 based TOC emulation on overwriteable media : Above method of multi-session emulation yields a single session image after each add-on session. No reliable session history can be read because the sector size of the existing session gets overwritten by the new one. A TOC with session history is nevertheless desirable with incremental backups in order to access older backup states by mounting older superblocks at the start addresses of older sessions. All usual ISO 9660 formatter programs write a complete superblock to the start of each session. With a uniform NWA rounding rule it is possible to compute the address of superblock N+1 as the NWA after session N. The only problem is N=1 because it gets overwritten by later sessions. libisoburn preserves the information of session 1 by writing the first session to LBA 32 rather than LBA 0. Afterwards it writes the overall superblock to LBA 0 (up to 31). So with all further add-on sessions the superblock at LBA 0 will enclose the overall image, while the superblocks of the sessions form a chain beginning at LBA 32. Each session superblock points to the next one by its sector count rounded up to 32. The chain end is marked by the overall image size. This chain gives the start addresses of sessions. The sector count minus start address gives the size of a particular session. ECMA-119 explains how to retrieve more info from the PVD (e.g. the volume id). See also the multi-session example in libisofs/doc/checksums.txt. ------------------------------------------------------------------------------- ------------------------------------------------------------------------------- Sequential DVD-R[W] Cookbook ------------------------------------------------------------------------------- Inspired by Andy Polyakov's http://fy.chalmers.se/~appro/linux/DVD+RW/tools , backed by reading mmc5r03c.pdf from http://www.t10.org/ftp/t10/drafts/mmc5/ and by experiments with drives NEC ND-4570A and LG GSA-4082B. For libburnia-project.org by Thomas Schmitt ------------------------------------------------------------------------------- Media type can be recognized by Current Profile from 46h GET CONFIGURATION. (mmc5r03c.pdf 6.6.2.1) DVD-R 0011h DVD-RW Restricted Overwrite 0013h DVD-RW Sequential Recording 0014h (DVD-R/DL Sequential Recording 0015h untested, might be single-session only) There are two approaches for writing to sequential DVD-R[W]: DAO and Incremental. Not all media and drives offer Incremental which allows multi-session as with CD media and does not demand a predicted track size. DAO seems to be the older method. It allows only one single session and track and it demands an exactly predicted track size. - About overwriteable, blank, appendable and finalized DVD-R[W] media - Incremental writing - DAO writing - Obtaining DVD-R[W] multi-session info for extending ISO-9660 filesystems - Obtaining a Table Of Content from DVD-R[W] - Hearsay about DVD-R/DL (Dual Layer) ------------------------------------------------------------------------------- About overwriteable, blank, appendable and finalized DVD-R[W] media : DVD-RW can be either in formatted state Restricted Overwrite or in unformatted state Sequential Recording. Sequential media can be either blank, appendable or finalized. Only blank and appendable media are sequentially writeable. For overwriteable DVD-RW see the Overwriteable DVD Cookbook. Overwriteable DVD-RW can be detected by their profile number 0013h in contrast to profile number 0014h for sequential DVD-RW. The status of sequential media can be inquired like with CD by 51h READ DISC INFORMATION requesting Data Type 000b Standard Disc Information, where reply value Disc Status indicates: 00b blank 01b appendable 10b finalized 11b others (unsuitable for this recipe) (mmc5r03c.pdf 6.22.3.1.4) Finalized, appendable or overwriteable DVD-RW can be brought into blank sequential state by command A1h BLANK with blanking type 000b "Blank the disc". See TAO Multi-Session CD Cookbook for details about blanking. After minimal blanking (type 001b) DVD-RW my two drives do not offer the Incremental Streaming feature 0021h the media any more. Full blanking (000b) brings back this feature. (mmc5r03c.pdf 6.2) ------------------------------------------------------------------------------- Incremental writing : Incremental writing allows to produce multi-session DVDs. It is indicated by feature 0021h being marked current in the reply of 46h GET CONFIGURATION. growisofs inquires 0021h by setting Starting Feature Number to 0x21 and Allocation Length to 16 in order to get only this one. The feature descriptor begins at byte 8 of the reply. Its availability is indicated by the Current Bit. libburn obtains the full feature list for this and other info. (mmc5r03c.pdf 5.2.2. Feature Descriptor format, 5.3.11 Feature 0021h, 6.2 46h GET CONFIGURATION, ) In mode page 05h this method is selected by Write Type 00h. Speed can be influenced by B6h SET STREAMING , speed capabilities can be inquired by ACh GET PERFORMANCE. It is advised to set only speeds and sizes which are returned by ACh. (mmc5r03c.pdf 6.39 SET STREAMING, 6.8 GET PERFORMANCE) growisofs fetches a mode page 05h template by MODE SENSE and inserts its own parameters. It sets Multi-session to 11b, unless dvd_compat is nonzero. libburn composes its mode page 05h from zero and allows control of Multi-Session by the application. BUFE Buffer Underrun protection 0=off, 1=on LS_V Link size valid 1=true Test Write -dummy mode for writing 0=off, 1=on Write Type Packet/TAO/SAO/RAW 00h = Incremental (Packet) Multi-session Whether to keep appendable 00b = finalize 11b = keep appendable Track Mode Describes frame type 5 [*1] Data Block Type Layout of payload blocks 8 [*2] Link Size ??? 16 [*3] FP Fixed Packet Size Bit 1 Packet Size 16 [*4] (mmc5r03c.pdf 7.5.4 The Mode Page, 4.2.3.4 Table 17 CONTROL = Track Mode) (spc3r23.pdf 6.8 MODE SELECT, 7.4.3 Mode parameter header formats) [*1:] growisofs takes the Track Mode from 52h READ TRACK INFORMATION, Address/Number Type 1, Track 1, Track Information Block byte 5 & 0xf. (mmc5r03.pdf 6.27) The specs predict that this will be Track Mode 4 (6.27.3.8) and also state that default is 5 (7.5.4.12). 4 means: uninterrupted, do not copy. 5 means increment, do not copy. [*2:] 8 means: 2048 byte data blocks. growisofs sets this value if Data Mode from above 52h READ TRACK INFORMATION is 1 or Fh, which is predicted by the specs to be always true. (mmc5r03.pdf 6.27.3.10) [*3:] growisofs (transport.hxx) sets Link Size to 16 for profiles 0011h and 0014h. libburn now records the first link size from feature 0021h in its burn_drive structure. If another link size item is 16, then 16 is used. [*4:] growisofs takes Packet Size from 52h. Specs predict it will be 16 (= 32 kiB). (mmc5r03.pdf 7.5.4.16) The writing process is much like in "Writing a session to CD in TAO mode" : Next Writeable Address is fetched from the reply of 52h READ TRACK INFORMATION. libburn writes full 32 kiB buffers via 2Ah WRITE. (mmc5r03c.pdf, 6.27 READ TRACK INFORMATION, 6.44 WRITE) When writing is done, it is mandatory to force the drive's buffer to media by 35h SYNCHRONIZE CACHE. (mmc5r03c.pdf, 6.41) The track has to be closed by 5Bh CLOSE TRACK SESSION Close Function 001b. growisofs uses the logical track number for that and not FFh like libburn does with TAO CD. So libburn obtains the Last Track Number in Last Session from the reply of 51h READ DISC INFORMATION requesting Data Type 000b "Standard Disc Information". (mmc5r03c.pdf 6.3.3.2.2 CLOSE TRACK, 6.22.3.1.) Multiple tracks are permissible in a single session. After all of them have been written, 5Bh CLOSE TRACK SESSION Close Function 010b with Logical Track Number 0 closes the session. It depends on the Multi-Session value in mode page 05h whether the disc is finalized or stays appendable. (mmc5r03c.pdf 6.3.3.2.3) ------------------------------------------------------------------------------- DAO writing : DAO is the mode described by feature 002Fh. This feature also gives information about capabilities for Burnfree (BUF), Test Write and DVD-RW. (mmc5r03c.pdf 5.3.25) Experiments with growisofs showed that the track size needs to be predicted and may not be exceeded during the write process. (growisofs ran into SCSI errors with piped non-ISO-9660 images and with piped ISO-9660 which have trailing data.) Speed can be influenced by B6h SET STREAMING , speed capabilities can be inquired by ACh GET PERFORMANCE. It is advised to set only speeds and sizes which are returned by ACh. (mmc5r03c.pdf 6.39 SET STREAMING, 6.8 GET PERFORMANCE) The mode page 05h to be sent : BUFE Buffer Underrun protection 0=off, 1=on LS_V Link size valid 0=false [*3] Test Write -dummy mode for writing 0=off, 1=on Write Type Packet/TAO/SAO/RAW 02h = DAO (same code as SAO) Multi-session Whether to keep appendable 00b = finalize Track Mode Describes frame type 5 [*1] Data Block Type Layout of payload blocks 8 [*2] Link Size ??? 0 [*3] FP Fixed Packet Size Bit 0 [*3] Packet Size 0 [*3] (mmc5r03c.pdf 7.5.4 The Mode Page, 4.2.3.4 Table 17 CONTROL = Track Mode) (spc3r23.pdf 6.8 MODE SELECT, 7.4.3 Mode parameter header formats) [*1:] growisofs takes the Track Mode from 52h READ TRACK INFORMATION, Address/Number Type 1, Track 1, Track Information Block byte 5 & 0xf. (mmc5r03.pdf 6.27) [*2:] 8 means: 2048 byte data blocks. growisofs sets this value if Data Mode from above 52h READ TRACK INFORMATION is 1 or Fh, which is predicted by the specs to be always true. (If not: growisofs aborts.) (mmc5r03.pdf 6.27.3.10) [*3:] Link Size, Packet Size and their companions only apply to Write Type 00h. The session layout must be described by 53h RESERVE TRACK, RMZ=ARSV=0. Reservation size should better already be aligned to 32 KiB. It has not been tested yet, what happens if not enough data get written. (mmc5r03c.pdf 6.31) Next Writeable Address is fetched from the reply of 52h READ TRACK INFORMATION. The reply is supposed to be 0. libburn writes full 32 kiB buffers via 2Ah WRITE. (mmc5r03c.pdf, 6.27 READ TRACK INFORMATION, 6.44 WRITE) If the track source delivers less than the announced size then libburn pads up by zeros. When writing is done, it is mandatory to force the drive's buffer to media by 35h SYNCHRONIZE CACHE. (mmc5r03c.pdf, 6.41) No further finalization is necessary. (I.e. no 5Bh CLOSE TRACK SESSION.) ------------------------------------------------------------------------------- Obtaining DVD-R[W] multi-session info for extending ISO-9660 filesystems : (valid for DVD+R too) Like with CD it is necessary to obtain the two numbers for mkisofs option -C in order to prepare a ISO-9660 filesystem image which by its inner pointers matches the block addresses of the future location on media. These are the start address of the first track in the last complete session and the predicted start address of the track which will host the new image. See TAO Multi-Session CD Cookbook for some more info about mkisofs aspects. The first number may be gained by 43h READ TOC/PMA/ATIP Format 0001b which in table 478 promises quick access via Start Address Of First Track In Last Session. (mmc5r03c.pdf 6.26.2.5 table 478, 6.26.3.3.1) Regrettably the MMC-5 specs still define a useless reply for non-CD media which obviously stems from MMC-3 times when no multi-session was possible with non-CD. (mmc5r03c.pdf 6.26.3.3.3) Both my drives do give a useful reply with the correct number for appendable DVD-RW. But not being backed by the specs this method appears unappealing . Another approach would be a formatted Table of Content, obtained by 43h READ TOC/PMA/ATIP Format 0000b. The specs do not totally outrule that this returns useful data with non-CD but they define a crippled TOC for multi-session. (mmc5r03c.pdf 6.26.3.2.4) My LG drive returns a more detailed TOC, my NEC drive stays with the rather suboptimal specs. So one would get different TOCs on different drives. Nevertheless, the MMC-5 compliant TOC would return the desired number in the Track Start address of the track with the highest number before AAh. Most stable seems the approach to obtain the desired number from the reply of 52h READ TRACK INFORMATION, Address/Number Type 01b. The field Logical Block Address/Track/Session has to bear the track number of the first track in the last complete session. To determine this number one has to determine the number of the last session and the number of the last track from 51h READ DISC INFORMATION and to iterate over the tracknumber by 52h READ TRACK INFORMATION until the first track with the desired session number appears and reveils its start address. (mmc5r03c.pdf 6.22 51h DISC, 6.27 52h TRACK) This method is very near to fabricating an own TOC. So libburn does this when inspecting the media. If the first number for -C is needed, libburn inquires its TOC model for the address of the first track in the last complete session. See below for a detailed description of TOC fabrication. The second -C number is the exact prediction of future track start address. It is gained like with CD by 52h READ TRACK INFORMATION Type 01b. Different from CD one may not use track number FFh but has to use the Last Track in Last Session from 51h READ DISC INFORMATION. (mmc5r03c.pdf 6.22 51h DISC, 6.27 52h TRACK) ------------------------------------------------------------------------------- Obtaining a Table Of Content from DVD-R[W]: (valid for DVD+R too) The raw TOC entries from 43h READ TOC/PMA/ATIP Format 0010b as described with CD media are not available with non-CD. There is a Format 0000b "Formatted TOC" but this is with non-CD a fictional information much at the discretion of the drive. Two drives with the same disc may well return different Formatted TOC. They are supposed to be consistent only about the last complete session and even there the MMC-5 specification 6.26.3.2.5 seems to prescribe a structure which does not match the true structure of incremental writing to sequential DVD-R[W]. (mmc5r03c.pdf 6.26.3.2) So i prefer not to use this method of getting a TOC. The alternative is to produce an own TOC from information gained by 51h READ DISC INFORMATION and by 52h READ TRACK INFORMATION which reveil a CD-like structure of sessions and 1:n related tracks. 51h READ DISC INFORMATION Data Type 000b, fields Number of Sessions (Least Significant Byte) and Number of Sessions (Most Significant Byte) give the number of sessions. The last complete session number of an appendable disc is one less because there is an incomplete session at its end. libburn only records complete sessions in its TOC model. libburn uses Last Track in Last Session as a hint for the range of track numbers. (mmc5r03c.pdf 6.22) Next step is to iterate from 1 up to the last track number and to obtain the according track info by 52h READ TRACK INFORMATION. Each track tells its Session Number (LSB at byte 2, MSB at 33), its Logical Track Start Address, its Logical Track Size, and much more which is not needed for a fake CD TOC. One may analyze the track info more finely but for this special purpose it is enough to discard the tracks which do not belong to complete sessions. (mmc5r03c.pdf 6.27) At the end of each session libburn inserts fake leadout entries into its TOC model. Their start address is computed from the start and size of the last track of the session. ------------------------------------------------------------------------------- Hearsay about DVD-R/DL (Dual Layer) : DVD-R/DL can assume profile 0015h DVD-R Dual Layer Sequential which is supposed to behave like DVD-R or 0016h DVD-R Dual Layer Jump which has no counterpart with DVD-R. A half-sentence in mmc5r03c.pdf 6.3.3.3.3 might indicate that closing a session by 5Bh CLOSE TRACK SESSION Close Function 010b overrides the multi-session bits in mode page 05h. growisofs applies this function in case of not DAO, though. A comment in growisofs_mmc.cpp states: "// DVD-R DL Seq has no notion of multi-session". I am not reading this from the specs - but not explicitely the contrary either. For now libburn will close the session but there is a macro prepared in libburn/write.c Libburn_dvd_r_dl_multi_no_close_sessioN which will suppress close session if multi-session is demanded. ------------------------------------------------------------------------------- ------------------------------------------------------------------------------- DVD+R[/DL] Cookbook ------------------------------------------------------------------------------- Inspired by reading mmc5r03c.pdf from http://www.t10.org/ftp/t10/drafts/mmc5/ backed by Andy Polyakov's http://fy.chalmers.se/~appro/linux/DVD+RW/tools , For libburnia-project.org by Thomas Schmitt ------------------------------------------------------------------------------- Media type can be recognized by Current Profile from 46h GET CONFIGURATION. (mmc5r03c.pdf 6.6.2.1) DVD+R 001bh DVD+R/DL 002bh - About empty, appendable and finalized DVD+R - Writing a Pseudo Session to DVD+R - DVD+R/DL (Dual Layer The following two chapters of the Sequential DVD-R[W] Cookbook are valid for DVD+R media too: - Obtaining DVD-R[W] multi-session info for extending ISO-9660 filesystems - Obtaining a Table Of Content from DVD-R[W] ------------------------------------------------------------------------------- About blank, appendable and finalized DVD+R : In the beginning a DVD+R holds an empty session and the Incomplete Fragment. From these one may spawn reserved fragments or one may write directly to the incomplete fragment. As soon as this is done the empty session becomes the open session which finally needs to get closed. By closing fragments and session a new empty session with empty Incomplete Fragment gets spawned. So the disc stays appendable. A DVD+R may hold 153 closed sessions with a single track each. The open session may hold up to 15 open fragments. But on closure of the session those fragments together form a single logical track. So one will usually only use a single fragment for sequential writing. (mmc5r03c.pdf 4.3.6.2) The disc may get finalized by another close command so that no more data can be written. (mmc5r03c.pdf 6.3.3.4.4) ------------------------------------------------------------------------------- Writing a Pseudo Session to DVD+R : Session writing has to be pseudo because only one logical track per session can be distinguished. So actually there have to be written multiple sessions to mark multiple tracks. The pseudo session cannot get marked on disc and thus the tracks of a pseudo session cannot be grouped accordingly in a TOC. Speed can be influenced by B6h SET STREAMING , speed capabilities can be inquired by ACh GET PERFORMANCE. It is advised to set only speeds and sizes which are returned by ACh. (mmc5r03c.pdf 6.39 SET STREAMING, 6.8 GET PERFORMANCE) No mode page 05h is to be sent. growisofs sends a page but the specs clearly state that one shall not do. (mmc5r03c.pdf 7.5.3) It is optional whether a track size is reserved in advance or not. Eventually this is done by 53h RESERVE TRACK, RMZ=ARSV=0. Reservation size should better already be aligned to 32 KiB. (mmc5r03c.pdf 6.31) The specs promise to pad up the track if not enough data get written. (mmc5r03c.pdf 6.3.3.4.2) Next Writeable Address is fetched from the reply of 52h READ TRACK INFORMATION with track number FFh. (mmc5r03c.pdf 6.27) Since the fixely set write type is 16-block packet, full 32 kiB buffers have to be transmitted via 2Ah WRITE. (mmc5r03c.pdf 4.3.6.2.2) When writing is done, it is mandatory to force the drive's buffer to media by 35h SYNCHRONIZE CACHE. (mmc5r03c.pdf 6.41) The written fragment (i.e. track-to-be) has to be closed by 5Bh CLOSE TRACK SESSION Close Function 001b. (mmc5r03c.pdf 6.3.3.4.2) libburn obtains the necessary logical track number from Last Track Number in Last Session from the reply of 51h READ DISC INFORMATION requesting Data Type 000b. (mmc5r03c.pdf 6.22) After each track 5Bh CLOSE TRACK SESSION Close Function 010b with Logical Track Number 0 closes the DVD+R session but keeps the media appendable. (mmc5r03c.pdf 6.3.3.4.3) If the media shall not stay appendable then the last DVD+R session is to be closed by Close Function 101b rather than 010b. This finalizes the media "with minimal radius". (mmc5r03c.pdf 6.3.3.4.4) Note: growisofs has code for that gesture but explicitly avoids to use it, if the media was appendable before writing began. Instead it recommends to fill up the media with zeros. This gesture nevertheless caused error replies from the drives in my own experiments. The reason given by Andy Polyakov is that some DVD-ROM drives get mislead by the lead-out information of (formerly) appendable media unless the media is fully written. (http://fy.chalmers.se/~appro/linux/DVD+RW/ , "Compatibility: caveat lector") Own experiments showed no such problems with PC attached PATA DVD-ROM drives. For best DVD-ROM compatibility one should avoid appendable media at all by closing them already after the first session. ------------------------------------------------------------------------------- DVD+R/DL (Dual Layer) : libburn treats DL media just like their single layer equivalents. This seems to work fine for DVD+R/DL, according to a report by nightmorph in http://libburnia-project.org/ticket/13 . ------------------------------------------------------------------------------- ------------------------------------------------------------------------------- BD-R Cookbook ------------------------------------------------------------------------------- Inspired by reading mmc5r03c.pdf from http://www.t10.org/ftp/t10/drafts/mmc5/ backed by experiments iwith drive LG GGW H20L. For libburnia-project.org by Thomas Schmitt ------------------------------------------------------------------------------- Media type can be recognized by Current Profile from 46h GET CONFIGURATION. (mmc5r03c.pdf 6.6.2.1) BD-R 0042h There are two basic recording modes defined: Sequential Recording Mode SRM and Random Recording Mode RRM. The latter is optional and for now not topic of this text. (mmc5r03c.pdf 4.5.3.5) - SRM Formatting - Writing a session in SRM-POW (- Pseudo-OverWrite SRM+POW) ------------------------------------------------------------------------------- SRM Formatting: Despite being write-once media BD-R can optionally carry some formatting. SRM has a disc structure model with tracks and sessions. Several tracks may be open at the same time, each having its own NWA. (mmc5r03c.pdf 4.5.3.5.2.2) This structure is formatted onto blank media automatically as soon as the first serious write attempt occurs. (mmc5r03c.pdf 4.5.3.5) Before such a write attempt, blank media may be explicitely formatted with spares, which provide defect management. (mmc5r03c.pdf 4.5.3.5.3) Tracks get created from other tracks via RESERVE TRACK splitting. (mmc5r03c.pdf 4.5.3.5.2.5) On top of defect management there may be Pseudo-OverWrite SRM+POW, a costly way to write several times to the same LBA. See below. Fully sequential states are called SRM-POW. (mmc5r03c.pdf 4.5.3.5.4) Explicite formatting is done by 04h FORMAT UNIT. Its data payload consists of a Format List Header and a Format Descriptor. It is advisable to set the Immed bit and the FOV bit in header byte number 1. The descriptor should be a copy of a descriptor from 23h READ FORMAT CAPACITIES but the size may be adjusted within a certain range. (mmc5r03c.pdf 6.5, 6.5.3.2, 6.5.3.3) Format type 00h creates SRM layouts with a default number of spares (or eventually RRM) chosen by the format sub-type: 00b = SRM with support for POW 01b = SRM without POW (but with some spares for defect management) 10b = (RRM) (mmc5r03c.pdf 6.5.4.2.1.6) Format type 32h uses the same sub-types but allows to allocate non-default amounts of spares. Similar to BD-RE format 31h, three format descriptors are offered: #1: default size, #2: maximum spare area, #3: minimal spare. The size may be chosen within that range. The sense behind the Type Dependent Parameters is obscure to me. Best will be to set ISA_V and TDMA_V to 0. (mmc5r03c.pdf 6.5.4.2.1.17) ------------------------------------------------------------------------------- Writing a session in SRM: The procedure and constraints for writing BD-R SRM-POW are very similar to DVD+R. libburn flatly re-uses its DVD+R code except the Close Function for finalizing a disc. In short: If all written sessions are closed, then there is exactly one NWA. In the beginning there is an empty session and track. A new track can be written either with pre-announced size (by RESERVE TRACK) or open-end by simply starting to write to the NWA. When done the track gets closed by close function 001b. Then either session or disc gets closed depending on the Close Function used: - Close Function 010b closes the session and keeps the media appendable (same as with DVD+R) - Close Function 110b finalizes the media and makes it read-only. (differs from libburn DVD+R procedure which uses 101b) ------------------------------------------------------------------------------- Pseudo-OverWrite POW: (no used yet by libburn) This enhancement of SRM emulates overwriting of existing data blocks. (mmc5r03c.pdf 4.5.3.5.4) POW establishes a virtual vLBA space on top of the real address space rLBA. All read and write commands deal with vLBA. It seems that track NWAs are assumed to be identical in vLBA space and in rLBA space. It is not clear whether one may write to vLBA blocks which are neither written yet nor at one of the track NWAs. Probably not, or else one could make NWAs run into vLBAs which are associated with older rLBAs. Replacing invalidated blocks consumes addresses in rLBA space at the NWA of some track. I.e. no spares are consumed by POW. Nevertheless it is costly by a special map called Orphanage. It covers rLBA which have been consumed by differing vLBAs. It never shrinks and can grow with each write to remapped addresses. To avoid heavy Orphanage growth it is advised to write mostly to vLBA which still coincide with their rLBA. E.g. those addresses which have neither been written as rLBA nor as vLBA yet. So one should begin the vLBA of new sessions at the NWA of a sufficiently sized track. (mmc5r03c.pdf 4.5.3.5.4.2 , 4.5.3.6.9) -------------------------------------------------------------------------------