213 lines
9.3 KiB

cdrskin Wiki - plain text copy
cdrskin is the cdrecord compatibility middleware of libburn.
Its paragon, cdrecord, is a powerful GPL'ed burn program included in Joerg
Schilling's cdrtools. cdrskin strives to be a second source for the services
traditionally provided by cdrecord. Currently it does CD-R and CD-RW.
Its future ability to burn DVD media depends on the development of libburn.
cdrskin does not contain any bytes copied from cdrecord's sources.
Many bytes have been copied from the message output of cdrecord
runs, though. The most comprehensive technical overview of cdrskin
can be found in cdrskin/README .
cdrskin with CD media fails to match its paragon cdrecord on three major
fields: convenient TAO burn mode, multi session, audio features.
Audio features are the only topic where cdrskin did not yet exploit current
libburn to the maximum. This is due to my own lack of audiophile motivation
and due to the lack of sincere users who provide me with cdrecord use cases,
help me to explore the original cdrecord behavior and serve as dedicated
testers for eventual newly introduced cdrskin audio features.
cdrskin does not provide DVD burning yet. See advise to use dvd+rw-tools
at the end of this text.
About the command line options of cdrskin:
There are two families of options: cdrecord-compatible ones and options
which are specific to cdrskin. The latter are mostly used to configure
cdrskin for its task to emulate cdrecord. There are some, nevertheless,
which provide rather exotic unique features of cdrskin.
The cdrecord-compatible options are listed in the output of
cdrskin -help
where the option "help" has *one* dash.
For these options you may expect program behavior that is roughly the
same as described in original man 1 cdrecord .
The cdrskin-specific options are listed by
cdrskin --help
where the option "help" has *two* dashes.
Those have no man page yet. Some are very experimental and should only be
used in coordination with the libburn developer team.
Some are of general user interest, though:
--devices allows the sysadmin to scan the system for possible drives
and displays their detected properties.
The drives are listed one per line, with fields:
libburn-drive-number sysadmin-device-file permissions : vendor type
0 dev='/dev/sg0' rwrw-- : 'HL-DT-ST' 'DVDRAM GSA-4082B'
This feature is valuable since cdrskin -scanbus will not give you
the device file name and its current permissions.
cdrskin will accept of course the proposed dev= option as address
for any usage of the drive.
Different from cdrecord, cdrskin is intended to be run without special
privileges, i.e. no superuser setuid. It is intended that the sysadmin
controls drive accessability by rw-permissions of the drive rather than
by x-permission of the burn binary. To be usable with cdrskin, the drive
has to offer both, r- and w-permission.
eject_device=<path> is needed to work around yet broken tray ejection of
drives. cdrskin makes a bold shell call to program "eject" and regrettably
this program does not like our addresses for SCSI devices.
/dev/hdX work fine and /dev/sg0 is quite safely guess-translated to
/dev/sr0 . /dev/sg1 need the user's help. <path> must work with eject.
dev_translation=<sep><from><sep><to> is needed to foist cdrskin to frontend
programs of cdrecord which do *not* ask cdrecord -scanbus but which make
own assumptions and guesses about cdrecord's device addresses.
cdrskin project - at least for now - refuses to try to provide a similar
guesswork but uses own cdrecord style addresses which have a mere
semi-automatic text mapping to real libburn addresses. See cdrskin/README,
"Pseudo-SCSI Adresses".
If you need to foist cdrskin under a frontend then you may be lucky and
both ideas of an address coincide. Especially if the frontend has the
decency to ask its "cdrecord" via option -scanbus for a list of drives.
If not, look into the error protocol of the frontend, look at the output
of a run of cdrskin --devices and give cdrskin the necessary hint.
If your frontend insists in using "0,0,0" and --devices reported
dev='/dev/sg0' resp. cdrskin -scanbus reported "1,0,0" then this
would be the appropriate translation
The "+" character is a separator to be choosen by you.
Currently i am not aware of the need to choose any other than "+"
unless you get playful with custom translations like
for an illustrated example with K3b 0.10 .
--fifo_start_empty is a throughput enhancer for unsteady data streams
like they are produced by a compressing archiver program when piping to
CD on-the-fly. It makes better use of the general property of a FIFO
buffer to transport surplus bandwidth into the future. Yep. A time machine.
One-way, i fear.
FIFO originally was introduced by cdrecord's author Joerg Schilling in order
to protect mediocre burner hardware from suffering buffer underruns
and thus producing misburns (at 1x speed on CD-R media at the price of a
DVD-RAM nowadays). This purpose would not justify a fifo any more -
given the limited life time of burners and the seamless underrun protection
of contemporary consumer drives.
With an unsteady data stream the task of the buffer is to soak up peak
performance and to release it steadily at the drive's maximum speed.
The larger the buffer the more reserves can be built up and the longer
input drought can be compensated.
Original cdrecord has the historical property, though, to first wait until
the buffer is completely filled. Best practice for fighting drive
underruns, of course.
With a very fat fs=# buffer (128 MB for 12x CD is not unrealistic) this
can cause a big delay until burning finally starts and takes its due time.
--fifo_start_empty makes cdrskin start burning without waiting for the
FIFO to be full resp. the data stream to end. It can make use of the
seconds spend with drive preparation and lead-in, it risks a few drive
buffer underruns at the beginning of burn - but modern drives stand this.
Note: no FIFO can give you better average throughput than the average
throughput of the data source and the throughput of the burner.
It can be used, though, to bring the effective throughput very close
to the theoretical limit. Especially with high speed media.
--no_rc allows you to surely ban influence from systemwide or user specific
default settings of cdrskin. Possible locations for such settings:
tao_to_sao_tsize=<num> allows the - actually unsupported - cdrecord option
-tao and defines a default track size to be used if - as custom with -tao -
no option tsize=# is given.
As in general with cdrskin tsize=# the data source does not have to provide
the full annouced amount of data. Missing data will be padded up by 0-bytes.
Surplus data is supposed to cause an error, though. The burn will then
be a failure in any way.
DVD advise:
For burning of DVD media the cdrskin project currently advises to use
Andy Polyakov's dvd+rw-tools which despite their historic name burn
for me on above burner: DVD+RW, DVD+R, DVD-RW, DVD-R .
They are not compatible or related to cdrecord resp. cdrecord-ProDVD
(now obsoleted by original source cdrtools cdrecord with identical
capabilities besides the license key).
If there is sincere and well motivated interest, the cdrskin project could try
to employ growisofs as DVD burning engine. The cdrskin project would prefer to
wait for DVD support being included in libburn, though.
A very limited and specialized cdrecord-compatibility wrapper for growisofs
serves in my project scdbackup. It is not overly hard to make one that serves
some very few fixed use cases.
To my knowledge, Linux kernels 2.6 do write to DVD+RW via block devices as
they would write to a traditional tape device. Try old tape archiver
commands with addresses like /dev/sr0 or /dev/hdc rather than /dev/mt0 .
I have heard rumors that DVD-RW in mode "restricted overwrite" would be
block device ready, too. My burner is not a real friend of DVD-RW and
in an experiment the burn worked fine - but the result was not identical
to the stream sent to the device. I had similar failure with DVD-RAM, too.
Beware of the impact of a slow block device on overall system i/o buffering.
It is wise to curb its input to a speed which it is able to deliver to media.
Else your i/o dedicated RAM might buffer a big amount of stream data.