resuming writing an unclosed DVD+R track

[Ivan Shmakov]

        Recently I’ve stumbled on my DVD drive failing mid-write of
        a pre-mastered image.  Noticing that the track is marked as
        ‘incomplete’ in the dvd+rw-mediainfo(1) output, and recalling
        that one of the features of the DVD+R specification was
        ‘easy resumption’ of recording, I did a Web search and
        followed the suggestion to use the -use-the-force-luke=seek=
        option to growisofs(1) to resume writing from [1].

[1] http://unix.stackexchange.com/questions/313840/

        The resulting medium had only two 2 KiB sectors (= 4 KiB)
        worth of lost data.  Given the availability of some 13 MiB
        of spare space left on disk, I’ve copied there (as a second
        track / session) a range of blocks covering the lost
        portion, as well as a chunk from the beginning for a good
        measure, and an ad-hoc text file describing the resulting
        layout of the medium.

        A cursory reading of xorriso(1) / xorrecord(1) didn’t reveal
        any feature similar to the growisofs(1) seek= option, though,
        making me wonder if there’s one?

        A more detailed description follows.

        (Portability note: I’ve relied on GNU/Linux tools and facilities
        below, such as losetup(8), GNU Awk, as well as the -i, -n
        options that seem to be specific to GNU cmp.)

        The growisofs(1) invocation failed as follows:

+ /usr/bin/time -- growisofs -dvd-compat -speed=6 -Z /dev/sr0=IMAGE.FILE
Executing 'builtin_dd if=IMAGE.FILE of=/dev/sr0 obs=32k seek=0'
/dev/sr0: "Current Write Speed" is 6.1x1352KBps.
          0/4700372992 ( 0.0%) @0x, remaining ??:?? RBU 100.0% UBU   0.0%
          0/4700372992 ( 0.0%) @0x, remaining ??:?? RBU 100.0% UBU   0.0%
          0/4700372992 ( 0.0%) @0x, remaining ??:?? RBU 100.0% UBU   0.0%
    4980736/4700372992 ( 0.1%) @1.1x, remaining 251:23 RBU 100.0% UBU -12.5%
   32768000/4700372992 ( 0.7%) @6.0x, remaining 45:06 RBU 100.0% UBU  98.4%

…

  484114432/4700372992 (10.3%) @5.4x, remaining 11:01 RBU 100.0% UBU  98.4%
:-[ WRITE@LBA=39b60h failed with SK=4h/LOGICAL UNIT COMMUNICATION TIMEOUT]: 
Input/output error
:-( write failed: Input/output error
/dev/sr0: flushing cache
:-[ FLUSH CACHE failed with SK=2h/LOGICAL UNIT NOT READY, LONG WRITE IN 
PROGRESS]: Resource temporarily unavailable
:-[ SYNCHRONOUS FLUSH CACHE failed with SK=2h/LOGICAL UNIT NOT READY, LONG 
WRITE IN PROGRESS]: Resource temporarily unavailable
Command exited with non-zero status 11
0.06user 2.31system 1:16.95elapsed 3%CPU (0avgtext+0avgdata 35836maxresident)k
1011312inputs+0outputs (1major+8500minor)pagefaults 0swaps

        (The reasons are as yet unidentified, but it wasn’t the only
        failure to happen that day; OTOH, I don’t recall such failures
        over the prior months of somewhat occasional use of the drive.)

        As par the course, I’ve checked the dvd+rw-mediainfo(1) output
        immediately; notable portion being:

READ DISC INFORMATION:
 Disc status:           appendable
 Number of Sessions:    1
 State of Last Session: incomplete
 "Next" Track:          1
 Number of Tracks:      1
READ TRACK INFORMATION[#1]:
 Track State:           invisible
 Track Start Address:   0*2KB
 Next Writable Address: 236400*2KB
 Free Blocks:           2058704*2KB
 Track Size:            2295104*2KB
 ROM Compatibility LBA: 270336
READ CAPACITY:          0*2048=0

        (Note that LBA  236400 refers to the DVD sector immediately
        after 0x39b60 = 236384 reported in the error above.)

        In order to write the latter part of IMAGE.FILE (i. e., to skip
        the already written part), I’ve created a loop device (no idea
        if growisofs(1) can be instructed to do such a skip by itself):

# chacl u::rw-,g::rw-,u:USERNAME:r--,m::rw-,o::--- \
      "$(losetup --show --find -r -o $((236400*2))K -- IMAGE.FILE)" 

        Then resumed writing (below, 14775 = 236400 / 16, for there’re
        16 2 KiB blocks in one 32 KiB DVD sector):

+ /usr/bin/time -- growisofs -dvd-compat -speed=6 
-use-the-force-luke=seek=236400 -Z /dev/sr0=/dev/loopXX
Executing 'builtin_dd if=/dev/loopXX of=/dev/sr0 obs=32k seek=14775'
/dev/sr0: "Current Write Speed" is 6.1x1352KBps.
:-? resuming track#1 from LBA#236400
  503087104/4700372992 ( 0.4%) @4.0x, remaining 22:09 RBU 100.0% UBU -12.5%
  528515072/4700372992 ( 1.1%) @5.5x, remaining 14:06 RBU 100.0% UBU  97.7%

…

 4616650752/4700372992 (98.0%) @6.0x, remaining 0:10 RBU 100.0% UBU  98.4%
 4644405248/4700372992 (98.7%) @6.0x, remaining 0:06 RBU 100.0% UBU  98.4%
 4672159744/4700372992 (99.3%) @6.0x, remaining 0:03 RBU  29.8% UBU  98.4%
builtin_dd: 2049808*2KB out @ average 5.9x1352KBps
/dev/sr0: flushing cache
/dev/sr0: closing track
/dev/sr0: closing session
/dev/sr0: reloading tray
0.60user 18.40system 9:03.17elapsed 3%CPU (0avgtext+0avgdata 35768maxresident)k
8199680inputs+0outputs (4major+8609minor)pagefaults 0swaps

        The relevant portion of the dvd+rw-mediainfo(1) output after
        the write:

READ DISC INFORMATION:
 Disc status:           appendable
 Number of Sessions:    2
 State of Last Session: empty
 "Next" Track:          2
 Number of Tracks:      2
READ TRACK INFORMATION[#1]:
 Track State:           invisible
 Track Start Address:   0*2KB
 Free Blocks:           0*2KB
 Track Size:            2286208*2KB
 ROM Compatibility LBA: 270336
READ TRACK INFORMATION[#2]:
 Track State:           blank
 Track Start Address:   2288256*2KB
 Next Writable Address: 2288256*2KB
 Free Blocks:           6848*2KB
 Track Size:            6848*2KB
 ROM Compatibility LBA: 270336
FABRICATED TOC:
 Track#1  :             17@0
 Track#AA :             17@2286208
 Multi-session Info:    #1@0
READ CAPACITY:          2286208*2048=4682153984

        Comparing the data on disk with the source IMAGE.FILE
        revealed two sectors of lost data:

$ (/usr/bin/time ionice -c 3 -- cmp -l -- IMAGE.FILE /dev/sr0 \
       |  gawk '! /^ *[0-9]+\s/ { print $0; next; }
                (c = rshift($1, 11)) > 1 + b { print a "-" b; a = c; }
                1 { b = c; } END { print a "-" b; }') \
              2>&1 | less -UF 
Command exited with non-zero status 1
1.84user 17.56system 5:31.47elapsed 5%CPU (0avgtext+0avgdata 1740maxresident)k
18289736inputs+0outputs (2major+165minor)pagefaults 0swaps
-
236398-236400
$ 

        (The expression above should’ve been c = rshift(-1 + $1, 11),
        for cmp -l counts bytes from 1, not 0; and the output should’ve
        been 236398-236399.)

        At this point I’ve decided to make use of the space still
        available and prepared the following map file:

## block-size: 2048
0 236398 payload+0
236398 2 bad
236400 2049808 payload+236400
## copies
2288256 384 payload+236160
2288640 6400 payload+0
2295040 64 map

        The file to write was created with shell code along the lines of:

#!/bin/sh

## Ivan Shmakov, 2022
## License: CC0-1.0

set -e 
set -C -u

## Hard-coded arguments:
set -- IMAGE.FILE MAP.FILE -2288256

f=${1} ; m=${2} ; adj=${3:-0} ; shift 3
exec < "$m" > "$(mktemp -- /tmp/pieces.XXXXXXXX)"

if test -t 1 ; then
    printf %s\\n "Refusing to write to a tty" >&2
    ## .
    exit 1
fi

pos=0
sed -e 1,/copies/d | while read -r -- o z src ; do
    o=$((adj + o))
    test "$o" = "$pos" \
        || exit 1
    case "$src" in
        (payload+*) dd bs=2048 count="$z" skip="${src##*+}" < "$f" ;;
        (map)
            sz=$(wc -c < "$m")
            sz=$((1 + (-1 + sz) / 2048))
            rz=${z}
            while test "$rz" -gt "$sz" ; do
                rz=$((rz - sz))
                dd bs=2048 count="$sz" conv=sync < "$m"
            done
            dd bs=2048 count="$rz" conv=sync < "$m"
            ;;
    esac
    pos=$((z + o))
done

        (Note that MAP.FILE is written over and over to fill up the
        allocated 64 2 KiB sectors.)

        The resulting pieces.-file was checked to contain the
        intended data, in particular:

$ /usr/bin/time -- sh -Cexuc '
cmp -n $((384 * 2))K -i $((236160 * 2))K:$(((2288256 - 2288256) * 2))K -- "$@"
cmp -n $((6400 * 2))K -i $((0 * 2))K:$(((2288640 - 2288256) * 2))K -- "$@"
' dummy.sh  IMAGE.FILE /tmp/pieces.Vm3nMUcT 

        Then written to the disk:

+ /usr/bin/time -- growisofs -dvd-compat -speed=6 -overburn 
-use-the-force-luke=seek=2288256 -Z /dev/sr0=/tmp/pieces.Vm3nMUcT
Executing 'builtin_dd if=/tmp/pieces.Vm3nMUcT of=/dev/sr0 obs=32k seek=143016'
/dev/sr0: "Current Write Speed" is 6.1x1352KBps.
 4695326720/4700372992 (64.0%) @1.9x, remaining 0:03 RBU  15.0% UBU  37.5%
builtin_dd: 6848*2KB out @ average 1.4x1352KBps
/dev/sr0: flushing cache
/dev/sr0: closing track
:-[ CLOSE TRACK failed with SK=5h/INVALID FIELD IN CDB]: Input/output error
/dev/sr0: closing session
/dev/sr0: reloading tray
0.00user 0.12system 0:24.28elapsed 0%CPU (0avgtext+0avgdata 35752maxresident)k
0inputs+0outputs (0major+8603minor)pagefaults 0swaps

        Bringing the dvd+rw-mediainfo(1) output to the following
        state.  (The CLOSE TRACK failure above being apparently benign.)

READ DISC INFORMATION:
 Disc status:           complete
 Number of Sessions:    2
 State of Last Session: complete
 Number of Tracks:      2
READ TRACK INFORMATION[#1]:

…

READ TRACK INFORMATION[#2]:
 Track State:           invisible
 Track Start Address:   2288256*2KB
 Free Blocks:           0*2KB
 Track Size:            6848*2KB
 ROM Compatibility LBA: 270336
FABRICATED TOC:
 Track#1  :             17@0
 Track#2  :             17@2288256
 Track#AA :             17@2295104
 Multi-session Info:    #2@2288256
READ CAPACITY:          2295104*2048=4700372992

        Finally, I’ve used dmsetup(1) to create a block device that
        maps the copy of the data from the tail of the medium over
        the miswritten portion:

bash# dmsetup create -- sr0.fixed \
          <(printf %s\\n "0 944640 linear /dev/sr0 0" \
                "944640 1536 linear /dev/sr0 9153024" \
                "946176 8198656 linear /dev/sr0 946176") 

        Comparing the resulting device to the original IMAGE.FILE
        revealed no differences.

-- 
FSF associate member #7257  http://am-1.org/~ivan/