sbpcd

《嵌入式系统设计与实例开发实验教材二源码》Linux内核移植与编译实验

This README belongs to release 4.2 or newer of the SoundBlaster Pro
(Matsushita, Kotobuki, Panasonic, CreativeLabs, Longshine and Teac)
CD-ROM driver for Linux.

sbpcd really, really is NOT for ANY IDE/ATAPI drive!
Not even if you have an "original" SoundBlaster card with an IDE interface!
So, you'd better have a look into README.ide if your port address is 0x1F0,
0x170, 0x1E8, 0x168 or similar.
I get tons of mails from IDE/ATAPI drive users - I really can't continue
any more to answer them all. So, if your drive/interface information sheets
mention "IDE" (primary, secondary, tertiary, quaternary) and the DOS driver
invoking line within your CONFIG.SYS is using an address below 0x230:
DON'T ROB MY LAST NERVE - jumper your interface to address 0x170 and IRQ 15
(that is the "secondary IDE" configuration), set your drive to "master" and
use ide-cd as your driver. If you do not have a second IDE hard disk, use the
LILO commands
   hdb=noprobe hdc=cdrom
and get lucky.
To make it fully clear to you: if you mail me about IDE/ATAPI drive problems,
my answer is above, and I simply will discard your mail, hoping to stop the
flood and to find time to lead my 12-year old son towards happy computing.

The driver is able to drive the whole family of "traditional" AT-style (that
is NOT the new "Enhanced IDE" or "ATAPI" drive standard) Matsushita,
Kotobuki, Panasonic drives, sometimes labelled as "CreativeLabs". The
well-known drives are CR-521, CR-522, CR-523, CR-562, CR-563.
CR-574 is an IDE/ATAPI drive.

The Longshine LCS-7260 is a double-speed drive which uses the "old"
Matsushita command set. It is supported - with help by Serge Robyns.
Vertos ("Elitegroup Computer Systems", ECS) has a similar drive - support
has started; get in contact if you have such a "Vertos 100" or "ECS-AT"
drive.

There exists an "IBM External ISA CD-ROM Drive" which in fact is a CR-563
with a special controller board. This drive is supported (the interface is
of the "LaserMate" type), and it is possibly the best buy today (cheaper than
an internal drive, and you can use it as an internal, too - e.g. plug it into
a soundcard).

CreativeLabs has a new drive "CD200" and a similar drive "CD200F". The latter
is made by Funai and sometimes named "E2550UA", newer models may be named
"MK4015". The CD200F drives should fully work.
CD200 drives without "F" are still giving problems: drive detection and
playing audio should work, data access will result in errors. I need qualified
feedback about the bugs within the data functions or a drive (I never saw a
CD200).

The quad-speed Teac CD-55A drive is supported, but still does not reach "full
speed". The data rate already reaches 500 kB/sec if you set SBP_BUFFER_FRAMES
to 64 (it is not recommended to do that for normal "file access" usage, but it
can speed up things a lot if you use something like "dd" to read from the
drive; I use it for verifying self-written CDs this way).
The drive itself is able to deliver 600 kB/sec, so this needs
work; with the normal setup, the performance currently is not even as good as
double-speed.

This driver is NOT for Mitsumi or Sony or Aztech or Philips or XXX drives,
and again: this driver is in no way usable for any IDE/ATAPI drive. If you 
think your drive should work and it doesn't: send me the DOS driver for your
beast (gzipped + uuencoded) and your CONFIG.SYS if you want to ask me for help,
and include an original log message excerpt, and try to give all information
a complete idiot needs to understand your hassle already with your first
mail. And if you want to say "as I have mailed you before", be sure that I
don't remember your "case" by such remarks; at the moment, I have some 
hundreds of open correspondences about Linux CDROM questions (hope to reduce if
the IDE/ATAPI user questions disappear). 


This driver will work with the soundcard interfaces (SB Pro, SB 16, Galaxy,
SoundFX, Mozart, MAD16 ...) and with the "no-sound" cards (Panasonic CI-101P,
LaserMate, WDH-7001C, Longshine LCS-6853, Teac ...).

It works with the "configurable" interface "Sequoia S-1000", too, which is 
used on the Spea Media FX and Ensonic Soundscape sound cards. You have to
specify the type "SBPRO 2" and the true CDROM port address with it, not the
"configuration port" address.

If you have a sound card which needs a "configuration driver" instead of
jumpers for interface types and addresses (like Mozart cards) - those
drivers get invoked before the DOS CDROM driver in your CONFIG.SYS, typical
names are "cdsetup.sys" and "mztinit.sys" - let the sound driver do the
CDROM port configuration (the leading comments in linux/drivers/sound/mad16.c
are just for you!). Hannu Savolainen's mad16.c code is able to set up my
Mozart card - I simply had to add
   #define MAD16_CONF 0x06
   #define MAD16_CDSEL 0x03
to configure the CDROM interface for type "Panasonic" (LaserMate) and address
0x340.

The interface type has to get configured in linux/drivers/cdrom/sbpcd.h, 
because the register layout is different between the "SoundBlaster" and the
"LaserMate" type.

I got a report that the Teac interface card "I/F E117098" is of type
"SoundBlaster" (i.e. you have to set SBPRO to 1) even with the addresses
0x300 and above. This is unusual, and it can't get covered by the auto
probing scheme.
The Teac 16-bit interface cards (like P/N E950228-00A, default address 0x2C0)
need the SBPRO 3 setup.

If auto-probing found the drive, the address is correct. The reported type
may be wrong. A "mount" will give success only if the interface type is set
right. Playing audio should work with a wrong set interface type, too.

With some Teac and some CD200 drives I have seen interface cards which seem
to lack the "drive select" lines; always drive 0 gets addressed. To avoid
"mirror drives" (four drives detected where you only have one) with such
interface cards, set MAX_DRIVES to 1 and jumper your drive to ID 0 (if
possible).


Up to 4 drives per interface card, and up to 4 interface cards are supported.
All supported drive families can be mixed, but the CR-521 drives are 
hard-wired to drive ID 0. The drives have to use different drive IDs, and each
drive has to get a unique minor number (0...3), corresponding indirectly to 
its drive ID.
The drive IDs may be selected freely from 0 to 3 - they do not have to be in
consecutive order.

As Don Carroll, [e-mail] or FIDO 1:382/14, told me, it is possible
to change old drives to any ID, too. He writes in this sense:
   "In order to be able to use more than one single speed drive
   (they do not have the ID jumpers) you must add a DIP switch
   and two resistors. The pads are already on the board next to
   the power connector. You will see the silkscreen for the
   switch if you remove the top cover.
                    1 2 3 4
             ID 0 = x F F x             O = "on"
             ID 1 = x O F x             F = "off"
             ID 2 = x F O x             x = "don't care"
             ID 3 = x O O x
   Next to the switch are the positions for R76 (7k) and R78
   (12k). I had to play around with the resistor values - ID 3
   did not work with other values. If the values are not good,
   ID 3 behaves like ID 0."

To use more than 4 drives, you simply need a second controller card at a 
different address and a second cable.

The driver supports reading of data from the CD and playing of audio tracks.
The audio part should run with WorkMan, xcdplayer, with the "non-X11" products
CDplayer and WorkBone - tell me if it is not compatible with other software.
The only accepted measure for correctness with the audio functions is the
"cdtester" utility (appended) - most audio player programmers seem to be
better musicians than programmers. ;-)

With the CR-56x and the CD200 drives, the reading of audio frames is possible.
This is implemented by an IOCTL function which reads READ_AUDIO frames of
2352 bytes at once (configurable with the "READ_AUDIO" define, default is 0).
Reading the same frame a second time gives different data; the frame data 
start at a different position, but all read bytes are valid, and we always
read 98 consecutive chunks (of 24 Bytes) as a frame. Reading more than 1 frame
at once possibly misses some chunks at each frame boundary. This lack has to
get corrected by external, "higher level" software which reads the same frame 
again and tries to find and eliminate overlapping chunks (24-byte-pieces).

The transfer rate with reading audio (1-frame-pieces) currently is very slow.
This can be better reading bigger chunks, but the "missing" chunks possibly
occur at the beginning of each single frame.
The software interface possibly may change a bit the day the SCSI driver
supports it too.

With all but the CR-52x drives, MultiSession is supported.
Photo CDs work (the "old" drives like CR-521 can access only the first
session of a photoCD).
At ftp.gwdg.de:/pub/linux/hpcdtoppm/ you will find Hadmut Danisch's package to
convert photo CD image files and Gerd Knorr's viewing utility.

The transfer rate will reach 150 kB/sec with CR-52x drives, 300 kB/sec with
CR-56x drives, and currently not more than 500 kB/sec (usually less than
250 kB/sec) with the Teac quad speed drives.
XA (PhotoCD) disks with "old" drives give only 50 kB/sec.

This release consists of
- this README file
- the driver file linux/drivers/cdrom/sbpcd.c
- the stub files linux/drivers/cdrom/sbpcd[234].c
- the header file linux/drivers/cdrom/sbpcd.h.


To install:
-----------

1. Setup your hardware parameters. Though the driver does "auto-probing" at a
   lot of (not all possible!) addresses, this step is recommended for
   everyday use. You should let sbpcd auto-probe once and use the reported
   address if a drive got found. The reported type may be incorrect; it is
   correct if you can mount a data CD. There is no choice for you with the
   type; only one is right, the others are deadly wrong.

   a. Go into /usr/src/linux/drivers/cdrom/sbpcd.h and configure it for your
      hardware (near the beginning):
      a1. Set it up for the appropriate type of interface board.
          "Original" CreativeLabs sound cards need "SBPRO 1".
          Most "compatible" sound cards (almost all "non-CreativeLabs" cards)
          need "SBPRO 0".
          The "no-sound" board from OmniCd needs the "SBPRO 1" setup.
          The Teac 8-bit "no-sound" boards need the "SBPRO 1" setup.
          The Teac 16-bit "no-sound" boards need the "SBPRO 3" setup.
          All other "no-sound" boards need the "SBPRO 0" setup.
          The Spea Media FX and Ensoniq SoundScape cards need "SBPRO 2".
          sbpcd.c holds some examples in its auto-probe list.
          If you configure "SBPRO" wrong, the playing of audio CDs will work,
          but you will not be able to mount a data CD.
      a2. Tell the address of your CDROM_PORT (not of the sound port).
      a3. If 4 drives get found, but you have only one, set MAX_DRIVES to 1.
      a4. Set DISTRIBUTION to 0.
   b. Additionally for 2.a1 and 2.a2, the setup may be done during
      boot time (via the "kernel command line" or "LILO option"):
          sbpcd=0x320,LaserMate
      or
          sbpcd=0x230,SoundBlaster
      or
          sbpcd=0x338,SoundScape
      or
          sbpcd=0x2C0,Teac16bit
      This is especially useful if you install a fresh distribution.
      If the second parameter is a number, it gets taken as the type
      setting; 0 is "LaserMate", 1 is "SoundBlaster", 2 is "SoundScape",
      3 is "Teac16bit".
      So, for example
          sbpcd=0x230,1
      is equivalent to
          sbpcd=0x230,SoundBlaster

2. "cd /usr/src/linux" and do a "make config" and select "y" for Matsushita
   CD-ROM support and for ISO9660 FileSystem support. If you do not have a
   second, third, or fourth controller installed, do not say "y" to the 
   secondary Matsushita CD-ROM questions.

3. Then do a "make dep", then make the kernel image ("make zlilo" or similar).

4. Make the device file(s). This step usually already has been done by the
   MAKEDEV script.
   The driver uses MAJOR 25, so, if necessary, do
        mknod /dev/sbpcd  b 25 0       (if you have only one drive)
   and/or
        mknod /dev/sbpcd0 b 25 0
        mknod /dev/sbpcd1 b 25 1
        mknod /dev/sbpcd2 b 25 2
        mknod /dev/sbpcd3 b 25 3
   to make the node(s).

   The "first found" drive gets MINOR 0 (regardless of its jumpered ID), the
   "next found" (at the same cable) gets MINOR 1, ...
   
   For a second interface board, you have to make nodes like
        mknod /dev/sbpcd4 b 26 0
        mknod /dev/sbpcd5 b 26 1
   and so on. Use the MAJORs 26, 27, 28.

   If you further make a link like
        ln -s sbpcd /dev/cdrom
   you can use the name /dev/cdrom, too.

5. Reboot with the new kernel.

You should now be able to do
              mkdir /CD
and 
              mount -rt iso9660 /dev/sbpcd /CD
or
              mount -rt iso9660 -o block=2048 /dev/sbpcd /CD
and see the contents of your CD in the /CD directory.
To use audio CDs, a mounting is not recommended (and it would fail if the
first track is not a data track).


Using sbpcd as a "loadable module":
-----------------------------------

If you do NOT select "Matsushita/Panasonic CDROM driver support" during the
"make config" of your kernel, you can build the "loadable module" sbpcd.o.
Read /usr/src/linux/Documentation/modules.txt on this.

If sbpcd gets used as a module, the support of more than one interface
card (i.e. drives 4...15) is disabled.

You can specify interface address and type with the "insmod" command like:
 # insmod /usr/src/linux/modules/sbpcd.o sbpcd=0x340,0
or
 # insmod /usr/src/linux/modules/sbpcd.o sbpcd=0x230,1
or
 # insmod /usr/src/linux/modules/sbpcd.o sbpcd=0x338,2
where the last number represents the SBPRO setting (no strings allowed here).


Things of interest:
-------------------

The driver is configured to try the LaserMate type of interface at I/O port
0x0340 first. If this is not appropriate, sbpcd.h should get changed
(you will find the right place - just at the beginning).

No DMA and no IRQ is used.

To reduce or increase the amount of kernel messages, edit sbpcd.c and play
with the "DBG_xxx" switches (initialization of the variable "sbpcd_debug").
Don't forget to reflect on what you do; enabling all DBG_xxx switches at once
may crash your system, and each message line is accompanied by a delay.

The driver uses the "variable BLOCK_SIZE" feature. To use it, you have to
specify "block=2048" as a mount option. Doing this will disable the direct
execution of a binary from the CD; you have to copy it to a device with the
standard BLOCK_SIZE (1024) first. So, do not use this if your system is
directly "running from the CDROM" (like some of Yggdrasil's installation
variants). There are CDs on the market (like the German "unifix" Linux
distribution) which MUST get handled with a block_size of 1024. Generally,
one can say all the CDs which hold files of the name YMTRANS.TBL are defective;
do not use block=2048 with those.

Within sbpcd.h, you will find some "#define"s (e.g. EJECT and JUKEBOX). With
these, you can configure the driver for some special things.
You can use the appended program "cdtester" to set the auto-eject feature
during runtime. Jeff Tranter's "eject" utility can do this, too (and more)
for you.

There is an ioctl CDROMMULTISESSION to obtain with a user program if
the CD is an XA disk and - if it is - where the last session starts. The
"cdtester" program illustrates how to call it.


Auto-probing at boot time:
--------------------------

The driver does auto-probing at many well-known interface card addresses,
but not all:
Some probings can cause a hang if an NE2000 ethernet card gets touched, because
SBPCD's auto-probing happens before the initialization of the net drivers.
Those "hazardous" addresses are excluded from auto-probing; the "kernel 
command line" feature has to be used during installation if you have your 
drive at those addresses. The "module" version is allowed to probe at those
addresses, too.

The auto-probing looks first at the configured address resp. the address
submitted by the kernel command line. With this, it is possible to use this
driver within installation boot floppies, and for any non-standard address,
too.

Auto-probing will make an assumption about the interface type ("SBPRO" or not),
based upon the address. That assumption may be wrong (initialization will be
o.k., but you will get I/O errors during mount). In that case, use the "kernel
command line" feature and specify address & type at boot time to find out the
right setup.

For everyday use, address and type should get configured within sbpcd.h. That
will stop the auto-probing due to success with the first try.

The kernel command "sbpcd=0" suppresses each auto-probing and causes
the driver not to find any drive; it is meant for people who love sbpcd
so much that they do not want to miss it, even if they miss the drives. ;-)  

If you configure "#define CDROM_PORT 0" in sbpcd.h, the auto-probing is
initially disabled and needs an explicit kernel command to get activated.