Subversion Repositories or1k

\documentclass{article}

\def\version{$Id: cdrom-standard.tex,v 1.1 2005-12-20 10:05:54 jcastillo Exp $}

\evensidemargin=0pt

\oddsidemargin=0pt

\topmargin=-\headheight \advance\topmargin by -\headsep

\textwidth=15.99cm \textheight=24.62cm % normal A4, 1'' margin

\def\linux{{\sc Linux}}

\def\cdrom{{\sc CDrom}}

\def\cdromc{{\tt cdrom.c}}

\def\cdromh{{\tt cdrom.h}}

\def\ucdrom{{\tt ucdrom.h}}

\def\fo{\sl}

\everymath{\it} \everydisplay{\it}

\catcode `\_=\active \def_{\_\penalty100 }

\catcode`\<=\active \def<#1>{{\langle\hbox{\rm#1}\rangle}}

\begin{document}

\title{A \linux\ \cdrom\ standard}

\author{David van Leeuwen\\{\normalsize\tt david@tm.tno.nl}}

\maketitle

\section{Introduction}

\linux\ is probably the Unix-like operating system that supports the widest

variety of hardware devices. The reasons for this are presumably

\begin{itemize}

\item The large list of different hardware devices available for the popular

IBM PC-architecture,

\item The open design of the operating system, such that everybody can

write a driver for Linux (source code examples).

\end{itemize}

The vast choice and openness has lead not only to a wide support of

hardware devices, but also to a certain divergence in behavior.

Especially for \cdrom\ devices, the way a particular drive reacts to a

`standard' $ioctl()$ call varies a lot from one brand to another;

however, the \linux\ \cdrom\ driver writers kept away from wilderness

by the practice of evolving a new driver by copying, understanding and

changing an existing one.

Since the beginning of the \cdrom, many different interfaces

developed.  Some of them had their own proprietary design (Sony,

Mitsumi, Panasonic, Philips), other manufacturers adopted an existing

electrical interface and changed the functionality

(CreativeLabs/SoundBlaster, Teac, Funai) or simply adapted their

drives to one or more of the already existing electrical interfaces

(Aztech, Sanyo, Funai, Vertos, Longshine, Optics Storage and most of

the `NoName' manufacturers).  In cases where a new drive really

brought his own interface or used his own command set and flow control

scheme, either a separate driver had to be written, or an existing

driver had to get enhanced.

Nowadays, almost all new \cdrom\ types are either ATAPI/IDE or SCSI;

it is very unlikely that any manufacturer still will create a new

interface, and new drives for the existing proprietary interfaces are

getting rare.  But history has delivered us \cdrom\ support for many

different interfaces.

When (in the 1.3.70's) I looked at the existing interface which is

expressed through \cdromh\ it appeared to be a rather wild set of

commands and data formats.\footnote{I cannot recollect what kernel

  version I looked at, then, presumably 1.2.13 and 1.3.34---the latest

  kernel that I was indirectly involved in.} It seemed that many

features of the interface have been added to include certain specific

capabilities of certain drives, in an {\fo ad hoc\/} manner. More

importantly, it appeared that actual execution of the commands is

different for most of the different drivers: e.g., some drivers close

the tray if an $open()$ call occurs while the tray is unloaded, while

others do not. Some drivers lock the door upon opening the device, to

prevent an incoherent file system, but others don't, to allow software

ejection.  Undoubtedly, the capabilities of the different drives vary,

but even when two drives have the same capability the driver behavior

may be different.

I decided to start a discussion on how to improve uniformity,

addressing all \cdrom-driver developers found in the various driver

files. The reactions encouraged me to write a uniform (compatible)

software level \cdromc\ to which this document is the documentation.

In the mean time, the data structure definitions in \cdromh\ had been

cleaned up a lot---which was very helpful for the new code.

\begin{quote}

\small

[Apparently, not all \cdrom\ developers support this initiative.

They---mainly those who used the already existing drivers not only as

a coding example, but also as a `user interface' reference during

their own development---have taken care that \cdromh\ reflects a

software interface to `user programs' which is unique between all

drivers as much as possible; these driver writers will continue to

refine the existing \cdromh\ where it seems necessary, and they tend

to look if any newly requested functionality isn't already there

before they are ready to define new structures. The {\tt sbpcd} driver

gives an example that it is possible to let a robot arm play juke

box---either with audio or with data CDs---and that it is possible to

let the juke box work on even if a disk has fallen upon the floor and

the drive door has closed without having a disk inside; without any

new software layer or any structures which are not already present in

\cdromh.  This `other' group of \linux\ \cdrom\ driver writers

explicitly does {\em not\/} support the idea to define an additional

software layer between driver and user program.]\parfillskip=0pt

\end{quote}

The effort (\cdromc) of which this is the documentation is {\em not\/}

meant to drive a wedge between two groups of driver developers, but

rather to enable sharing of `strategic code' among drivers. The code

should {\em not\/} be viewed as a new interface to user-level

programs, but rather as a new interface between driver code and

kernel.

Care is taken that 100\,\% compatibility exists with the data

structures and programmer's interface defined in \cdromh, and in order

not to interfere with ongoing development in \cdromh, any `new' data

structures have been put in a separate header file called \ucdrom.

Because the data structures of \cdromh\ are fully supported, this

documentation may also be of help to the programmers using the

interface defined in \cdromh, but this guide is primarily written to

help \cdrom\ driver developers adapt their code to use the `common

\cdrom' code in \cdromc.

Personally, I think that the most important hardware interfaces will

be the IDE/ATAPI drives and of course the SCSI drives, but as prices

of hardware drop continuously, it is not unlikely that people will

have more than one \cdrom\ drive, possibly of mixed types. It is

important that these drives behave in the same way. (In December 1994,

one of the cheapest \cdrom\ drives was a Philips cm206, a double-speed

proprietary drive. In the months that I was busy writing a \linux\

driver for it, proprietary drives became old-fashioned and IDE/ATAPI

drives became standard. At the time of writing (April 1996) the

cheapest double speed drive is IDE and at one fifth of the price of

its predecessor. Eight speed drives are available now.)

This document defines (in pre-release versions: proposes) the various

$ioctl$s, and the way the drivers should implement this.

\section{Standardizing through another software level}

\label{cdrom.c}

At the time this document is written, all drivers directly implement

the $ioctl()$ calls through their own routines, with the danger of

forgetting calls to $verify_area()$ and the risk of divergence in

implementation.

For this reason, we\footnote{The writing style is such that `we' is

used when (at least part of) the \cdrom-device driver authors support

the idea, an `I' is used for personal opinions} propose to define

another software-level, that separates the $ioctl()$ and $open()$

implementation from the actual hardware implementation. Note that we

do not wish to alter the existing application interface defined in

\cdromh, but rather want to re-root the hardware-implementation through

some common code.

We believe that \cdrom\ drives are specific enough (i.e., different

from other block-devices such as floppy or hard disc drives), to

define a set of {\em \cdrom\ device operations},

$<cdrom-device>_dops$. These are of a different nature than the

classical block-device file operations $<block-device>_fops$.

The extra interfacing level routines are implemented in a file

\cdromc, and a low-level \cdrom\ driver hands over the interfacing to

the kernel by registering the following general $struct\ file_operations$:

$$

\halign{$#$\ \hfil&$#$\ \hfil&$/*$ \rm# $*/$\hfil\cr

struct& file_operations\ cdrom_fops = \{\hidewidth\cr

        &NULL,                  & lseek \cr

        &block_read,            & read---general\ block-dev\ read \cr

        &block_write,           & write---general block-dev write \cr

        &NULL,                  & readdir \cr

        &NULL,                  & select \cr

        &cdrom_ioctl,           & ioctl \cr

        &NULL,                  & mmap \cr

        &cdrom_open,            & open \cr

        &cdrom_release,         & release \cr

        &NULL,                  & fsync \cr

        &NULL,                  & fasync \cr

        &cdrom_media_changed,   & media_change \cr

        &NULL                   & revalidate \cr

\};\cr

}

$$

Every active \cdrom\ device shares this $struct$. The routines declared

above are all implemented in \cdromc, and this is the place where the

{\em behavior\/} of all \cdrom-devices is defined, and hence

standardized. The implementation of the interfacing to the various

types of hardware still is done by the various \cdrom-device drivers,

but these routines only implement certain {\em capabilities\/} that

are typical to \cdrom\ (removable-media) devices.

Registration of the \cdrom\ device driver should now be to the general

routines in \cdromc, not to the VFS any more. This is done though the

call

$$register_cdrom(int\ major, char * name,

  struct\ cdrom_device_ops\ device_options)

$$

The device operations structure lists the implemented routines for

interfacing to the hardware, and some specifications of capabilities

of the device, such as the maximum head-transfer rate.  [It is

impossible to come up with a complete list of all capabilities of

(future) \cdrom\ drives, as the developments in technology follow-up

at an incredible rate. Maybe write-operation (WORM devices) will

become very popular in the future.]  The list now is:

$$

\halign{$#$\ \hfil&$#$\ \hfil&\hbox to 10em{$#$\hss}&

  $/*$ \rm# $*/$\hfil\cr

struct& cdrom_device_ops\ \{ \hidewidth\cr

  &int& (* open)(kdev_t, int)\cr

  &void& (* release)(kdev_t);\cr

  &int& (* open_files)(kdev_t);  \cr

  &int& (* drive_status)(kdev_t);\cr

  &int& (* disc_status)(kdev_t);\cr

  &int& (* media_changed)(kdev_t);\cr

  &int& (* tray_move)(kdev_t, int);\cr

  &int& (* lock_door)(kdev_t, int);\cr

  &int& (* select_speed)(kdev_t, int);\cr

  &int& (* select_disc)(kdev_t, int);\cr

  &int& (* get_last_session) (kdev_t, struct\ cdrom_multisession *{});\cr

  &int& (* get_mcn)(kdev_t, struct\ cdrom_mcn *{});\cr

  &int& (* reset)(kdev_t);\cr

  &int& (* audio_ioctl)(kdev_t, unsigned\ int, void *{});\cr

  &int& (* dev_ioctl)(kdev_t, unsigned\ int, unsigned\ long);\cr

\noalign{\medskip}

  &\llap{const\ }int& capability;&  capability flags \cr

  &int& mask;& mask of capability: disables them \cr

  &\llap{$const\ $}int& speed;&  maximum speed for reading data \cr

  &\llap{$const\ $}int& minors;& number of supported minor devices \cr

  &\llap{$const\ $}int& capacity;& number of discs in jukebox \cr

\noalign{\medskip}

  &int& options;& options flags \cr

  &long& mc_flags;& media-change buffer flags ($2\times16$) \cr

\}\cr

}

$$ The \cdrom-driver should simply implement (some of) these

functions, and register the functions to the global \cdrom\ driver,

which performs interfacing with the Virtual File System and system

$ioctl$s. The flags $capability$ specify the hardware-capabilities on

registration of the device, the flags $mask$ can be used to mask some

of those capabilities (for one reason or another). The value $minors$

should be a positive value indicating the number of minor devices that

are supported by the driver, normally~1.  (They are supposed to be

numbered from 0 upwards). The value $capacity$ should be the number of

discs the drive can hold simultaneously, if it is designed as a

juke-box, or otherwise~1.

Two registers contain variables local to the \cdrom\ device. The flags

$options$ are used to specify how the general \cdrom\ routines

should behave. These various flags registers should provide enough

flexibility to adapt to the different user's wishes (and {\em not\/}

the `arbitrary' wishes of the author of the low-level device driver,

as is the case in the old scheme). The register $mc_flags$ is used to

buffer the information from $media_changed()$ to two separate queues.

Note that most functions have fewer parameters than their

$blkdev_fops$ counterparts. This is because very little of the

information in the structures $inode$ and $file$ are used, the main

parameter is the device $dev$, from which the minor-number can be

extracted. (Most low-level \cdrom\ drivers don't even look at that value

as only one device is supported.)

The intermediate software layer that \cdromc\ forms will performs some

additional bookkeeping. The minor number of the device is checked

against the maximum registered in $<device>_dops$. The function

$cdrom_ioctl()$ will verify the appropriate user-memory regions for

read and write, and in case a location on the CD is transferred, it

will `sanitize' the format by making requests to the low-level drivers

in a standard format, and translating all formats between the

user-software and low level drivers. This relieves much of the drivers

memory checking and format checking and translation. Also, the

necessary structures will be declared on the program stack.

The implementation of the functions should be as defined in the

following sections. Three functions {\em must\/} be implemented,

namely $open()$, $release()$ and $open_files()$. Other functions may

be omitted, their corresponding capability flags will be cleared upon

registration. Generally, a function returns zero on success and

negative on error. A function call should return only after the

command has completed, but of course waiting for the device should not

use processor time.

\subsection{$Open(kdev_t\ dev, int\ purpose)$}

$Open()$ should try to open the device for a specific $purpose$, which

can be either:

\begin{itemize}

\item[0] Open for data read, as is used by {\tt mount()} (2), or the

user commands {\tt dd} or {\tt cat}.

\item[1] Open for $ioctl$ commanding, as is used for audio-CD playing

programs mostly.

\end{itemize}

In this routine, a static counter should be updated, reflecting the

number of times the specific device is successfully opened (and in

case the driver supports modules, the call $MOD_INC_USE_COUNT$

should be performed exactly once, if successful). The return value is

negative on error, and zero on success. The open-for-ioctl call can

only fail if there is no hardware.

Notice that any strategic code (closing tray upon $open()$, etc.)\ is

done by the calling routine in \cdromc, so the low-level routine

should only be concerned with proper initialization and device-use

count.

\subsection{$Release(kdev_t\ dev)$}

The use-count of the device $dev$ should be decreased by 1, and a

single call $MOD_DEC_USE_COUNT$ should be coded here.  Possibly other

device-specific actions should be taken such as spinning down the

device. However, strategic actions such as ejection of the tray, or

unlocking the door, should be left over to the general routine

$cdrom_release()$. Also, the invalidation of the allocated buffers in

the VFS is taken care of by the routine in \cdromc.

\subsection{$Open_files(kdev_t\ dev)$}

This function should return the internal variable use-count of the

device $dev$. The use-count is not implemented in the routines in

\cdromc\ itself, because there may be many minor devices connected to

a single low-level driver.

\subsection{$Drive_status(kdev_t\ dev)$}

\label{drive status}

The function $drive_status$, if implemented, should provide

information of the status of the drive (not the status of the disc,

which may or may not be in the drive). In \ucdrom\ the possibilities

are listed:

$$

\halign{$#$\ \hfil&$/*$ \rm# $*/$\hfil\cr

CDS_NO_INFO& no information available\cr

CDS_NO_DISC& no disc is inserted, tray is closed\cr

CDS_TRAY_OPEN& tray is opened\cr

CDS_DRIVE_NOT_READY& something is wrong, tray is moving?\cr

CDS_DISC_OK& a disc is loaded and everything is fine\cr

}

$$

\subsection{$Disc_status(kdev_t\ dev)$}

\label{disc status}

As a complement to $drive_status()$, this function can provide the

general \cdrom-routines with information about the current disc that is

inserted in the drive represented by $dev$. The history of development

of the CD's use as a carrier medium for various digital information

has lead to many different disc types, hence this function can return:

$$

\halign{$#$\ \hfil&$/*$ \rm# $*/$\hfil\cr

CDS_NO_INFO& no information available\cr

CDS_NO_DISC& no disc is inserted, or tray is opened\cr

CDS_AUDIO& Audio disc (2352 audio bytes/frame)\cr

CDS_DATA_1& data disc, mode 1 (2048 user bytes/frame)\cr

CDS_DATA_2& data disc, mode 2 (2336 user bytes/frame)\cr

CDS_XA_2_1& mixed data (XA), mode 2, form 1 (2048 user bytes)\cr

CDS_XA_2_2& mixed data (XA), mode 2, form 1 (2324  user bytes)\cr

}

$$

As far as I know, \cdrom s are always of type $CDS_DATA_1$. For

some information concerning frame layout of the various disc types, see

a recent version of {\tt cdrom.h}.

\subsection{$Media_changed(dev\_t\ dev)$}

This function is very similar to the original function in $struct\

file_operations$. It returns 1 if the medium of the device $dev$ has

changed since the last call, and 0 otherwise. Note that by `re-routing'

this function through $cdrom_media_changed()$, we can implement

separate queues for the VFS and a new $ioctl()$ function that can

report device changes to software (e.g., an auto-mounting daemon).

\subsection{$Tray_move(kdev_t\ dev, int\ position)$}

This function, if implemented, should control the tray movement. (No

other function should control this.) The parameter $position$ controls

the desired direction of movement:

\begin{itemize}

\item[0] Close tray

\item[1] Open tray

\end{itemize}

This function returns 0 upon success, and a non-zero value upon

error. Note that if the tray is already in the desired position, no

action need be taken, and the return value should be 0.

\subsection{$Lock_door(kdev_t\ dev, int\ lock)$}

This function (and no other code) controls locking of the door, if the

drive allows this. The value of $lock$ controls the desired locking

state:

\begin{itemize}

\item[0] Unlock door, manual opening is allowed

\item[1] Lock door, tray cannot be ejected manually

\end{itemize}

Return values are as for $tray_move()$.

\subsection{$Select_speed(kdev_t\ dev, int\ speed)$}

Although none of the drivers has implemented this function so far,

some drives are capable of head-speed selection, and hence this is a

capability that should be standardized through a function in the

device-operations structure. This function should select the speed at

which data is read or audio is played back. The special value `0'

means `auto-selection', i.e., maximum data-rate or real-time audio

rate. If the drive doesn't have this `auto-selection' capability, the

decision should be made on the current disc loaded and the return

value should be positive. A negative return value indicates an

error. (Although the audio-low-pass filters probably aren't designed

for it, more than real-time playback of audio might be used for

high-speed copying of audio tracks). Badly pressed \cdrom s may

benefit from less-than-maximum head rate.

\subsection{$Select_disc(kdev_t\ dev, int\ number)$}

If the drive can store multiple discs (a juke-box), it is likely that

a disc selection can be made by software. This function should perform

disc selection. It should return the number of the selected disc on

success, a negative value on error. Currently, none of the \linux\

\cdrom\ drivers appears to support such functionality, but it is defined

here for future purposes.

\subsection{$Get_last_session(kdev_t\ dev, struct\ cdrom_multisession *

ms_info)$}

This function should implement the old corresponding $ioctl()$. For

device $dev$, the start of the last session of the current disc should

be returned in the pointer argument $ms_info$. Note that routines in \cdromc\ have sanitized this argument: its

requested format will {\em always\/} be of the type $CDROM_LBA$

(linear block addressing mode), whatever the calling software

requested. But sanitization goes even further: the low-level

implementation may return the requested information in $CDROM_MSF$

format if it wishes so (setting the $ms_info\rightarrow addr_format$

field appropriately, of course) and the routines in \cdromc\ will make

the transform if necessary. The return value is 0 upon success.

\subsection{$Get_mcn(kdev_t\ dev, struct\ cdrom_mcn * mcn)$}

Some discs carry a `Media Catalog Number' (MCN), also called

`Universal Product Code' (UPC). This number should reflect the number that

is generally found in the bar-code on the product. Unfortunately, the

few discs that carry such a number on the disc don't even use the same

format. The return argument to this function is a pointer to a

pre-declared memory region of type $struct\ cdrom_mcn$. The MCN is

expected as a 13-character string, terminated by a null-character.

\subsection{$Reset(kdev_t dev)$}

This call should implement hard-resetting the drive (although in

circumstances that a hard-reset is necessary, a drive may very well

not listen to commands anymore). Preferably, control is returned to the

caller only after the drive has finished resetting.

\subsection{$Audio_ioctl(kdev_t\ dev, unsigned\ int\ cmd, void *

arg)$}

Some of the \cdrom-$ioctl$s defined in {\tt cdrom.h} can be

implemented by the routines described above, and hence the function

$cdrom_ioctl$ will use those. However, most $ioctl$s deal with

audio-control. We have decided to leave these accessed through a

single function, repeating the arguments $cmd$ and $arg$. Note that

the latter is of type $void*{}$, rather than $unsigned\ long\

int$. The routine $cdrom_ioctl()$ does do some useful things,

though. It sanitizes the address format type to $CDROM_MSF$ (Minutes,

Seconds, Frames) for all audio calls. It also verifies the memory

location of $arg$, and reserves stack-memory for the argument. This

makes implementation of the $audio_ioctl()$ much simpler than in the

old driver scheme. For an example you may look up the function

$cm206_audio_ioctl()$ in {\tt cm206.c} that should be updated with

this documentation.

An unimplemented ioctl should return $-EINVAL$, but a harmless request

(e.g., $CDROMSTART$) may be ignored by returning 0 (success). Other

errors should be according to the standards, whatever they are. (We

may decide to sanitize the return value in $cdrom_ioctl()$, in order

to guarantee a uniform interface to the audio-player software.)

\subsection{$Dev_ioctl(kdev_t\ dev, unsigned\ int\ cmd, unsigned\ long\

arg)$}

Some $ioctl$s seem to be specific to certain \cdrom\ drives. That is,

they are introduced to service some capabilities of certain drives. In

fact, there are 6 different $ioctl$s for reading data, either in some

particular kind of format, or audio data. Not many drives support

reading audio tracks as data, I believe this is because of protection

of copyrights of artists. Moreover, I think that if audio-tracks are

supported, it should be done through the VFS and not via $ioctl$s. A

problem here could be the fact that audio-frames are 2352 bytes long,

so either the audio-file-system should ask for 75264 bytes at once

(the least common multiple of 512 and 2352), or the drivers should

bend their backs to cope with this incoherence (to which I would be

opposed). Once this question is resolved, this code should be standardized in

\cdromc.

Because there are so many $ioctl$s that seem to be introduced to

satisfy certain drivers,\footnote{Is there software around that actually uses

these? I'd be interested!} any `non-standard' $ioctl$s are routed through

the call $dev_ioctl()$. In principle, `private' $ioctl$s should be

numbered after the device's major number, and not the general \cdrom\

$ioctl$ number, {\tt 0x53}. Currently the non-supported $ioctl$s are:

{\it CDROMREADMODE1, CDROMREADMODE2, CDROMREADAUDIO, CDROMREADRAW,

CDROMREADCOOKED, CDROMSEEK, CDROMPLAY\-BLK and CDROMREADALL}.

\subsection{\cdrom\ capabilities}

Instead of just implementing some $ioctl$ calls, the interface in

\cdromc\ supplies the possibility to indicate the {\em capabilities\/}

of a \cdrom\ drive. This can be done by ORing any number of

capability-constants that are defined in \ucdrom\ at the registration

phase. Currently, the capabilities are any of:

$$

\halign{$#$\ \hfil&$/*$ \rm# $*/$\hfil\cr

CDC_CLOSE_TRAY& can close tray by software control\cr

CDC_OPEN_TRAY& can open tray\cr

CDC_LOCK& can lock and unlock the door\cr

CDC_SELECT_SPEED& can select speed, in units of $\sim$150\,kB/s\cr

CDC_SELECT_DISC& drive is juke-box\cr

CDC_MULTI_SESSION& can read sessions $>\rm1$\cr

CDC_MCN& can read Medium Catalog Number\cr

CDC_MEDIA_CHANGED& can report if disc has changed\cr

CDC_PLAY_AUDIO& can perform audio-functions (play, pause, etc)\cr

}

$$

The capability flag is declared $const$, to prevent drivers from

accidentally tampering with the contents. However, upon registration,

some (claimed) capability flags may be cleared if the supporting

function has not been implemented (see $register_cdrom()$ in

\cdromc).

If you want to disable any of the capabilities, there is a special

flag register $<device>_dops.mask$ that may (temporarily) disable

certain capabilities. In the file \cdromc\ you will encounter many

constructions of the type

$$\it

if\ (cdo\rightarrow capability \mathrel\& \mathord{\sim} cdo\rightarrow mask

   \mathrel{\&} CDC_<capability>) \ldots

$$

The $mask$ could be set in the low-level driver code to disable

certain capabilities for special brands of the device that can't

perform the actions.  However, there is not (yet) an $ioctl$ to set

the mask\dots The reason is that I think it is better to control the

{\em behavior\/} rather than the {\em capabilities}.

\subsection{Options}

A final flag register controls the {\em behavior\/} of the \cdrom\

drives, in order to satisfy different users' wishes, hopefully

independently of the ideas of the respective author who happened to

have made the drive's support available to the \linux\ community. The

current behavior options are:

$$

\halign{$#$\ \hfil&$/*$ \rm# $*/$\hfil\cr

CDO_AUTO_CLOSE& try to close tray upon device $open()$\cr

CDO_AUTO_EJECT& try to open tray on last device $close()$\cr

CDO_USE_FFLAGS& use $file_pointer\rightarrow f_flags$ to indicate

 purpose for $open()$\cr

CDO_LOCK& try to lock door if device is opened\cr

CDO_CHECK_TYPE& ensure disc type is data if opened for data\cr

}

$$

The initial value of this register is $CDO_AUTO_CLOSE \mathrel|

CDO_USE_FFLAGS \mathrel| CDO_LOCK$, reflecting my own view on user

interface and software standards. Before you protest, there are two

new $ioctl$s implemented in \cdromc, that allow you to control the

behavior by software. These are:

$$

\halign{$#$\ \hfil&$/*$ \rm# $*/$\hfil\cr

CDROM_SET_OPTIONS& set options specified in $(int)\ arg$\cr

CDROM_CLEAR_OPTIONS& clear options specified in $(int)\ arg$\cr

}

$$

One option needs some more explanation: $CDO_USE_FFLAGS$. In the next

section we explain what the need for this option is.

\section{The need to know the purpose of opening}

Traditionally, Unix devices can be used in two different `modes',

either by reading/writing to the device file, or by issuing

controlling commands to the device, by the device's $ioctl()$

call. The problem with \cdrom\ drives, is that they can be used for

two entirely different purposes. One is to mount removable

file systems, \cdrom s, the other is to play audio CD's. Audio commands

are implemented entirely through $ioctl$s, presumably because the

first implementation (SUN?) has been such. In principle there is

nothing wrong with this, but a good control of the `CD player' demands

that the device can {\em always\/} be opened in order to give the

$ioctl$ commands, regardless of the state the drive is in.

On the other hand, when used as a removable-media disc drive (what the

original purpose of \cdrom s is) we would like to make sure that the

disc drive is ready for operation upon opening the device. In the old

scheme, some \cdrom\ drivers don't do any integrity checking, resulting

in a number of i/o errors reported by the VFS to the kernel when an

attempt for mounting a \cdrom\ on an empty drive occurs. This is not a

particularly elegant way to find out that there is no \cdrom\ inserted;

it more-or-less looks like the old IBM-PC trying to read an empty floppy

drive for a couple of seconds, after which the system complains it

can't read from it. Nowadays we can {\em sense\/} the existence of a

removable medium in a drive, and we believe we should exploit that

fact. An integrity check on opening of the device, that verifies the

availability of a \cdrom\ and its correct type (data), would be

desirable.

These two ways of using a \cdrom\ drive, principally for data and

secondarily for playing audio discs, have different demands for the

behavior of the $open()$ call. Audio use simply wants to open the

device in order to get a file handle which is needed for issuing

$ioctl$ commands, while data use wants to open for correct and

reliable data transfer. The only way user programs can indicate what

their {\em purpose\/} of opening the device is, is through the $flags$

parameter (see {\tt open(2)}). For \cdrom\ devices, these flags aren't

implemented (some drivers implement checking for write-related flags,

but this is not strictly necessary if the device file has correct

permission flags). Most option flags simply don't make sense to

\cdrom\ devices: $O_CREAT$, $O_NOCTTY$, $O_TRUNC$, $O_APPEND$, and

$O_SYNC$ have no meaning to a \cdrom.

We therefore propose to use the flag $O_NONBLOCK$ to indicate

that the device is opened just for issuing $ioctl$

commands. Strictly, the meaning of $O_NONBLOCK$ is that opening and

subsequent calls to the device don't cause the calling process to

wait. We could interpret this as ``don't wait until someone has

inserted some valid data-\cdrom.'' Thus, our proposal of the

implementation for the $open()$ call for \cdrom s is:

\begin{itemize}

\item If no other flags are set than $O_RDONLY$, the device is opened

for data transfer, and the return value will be 0 only upon successful

initialization of the transfer. The call may even induce some actions

on the \cdrom, such as closing the tray.

\item If the option flag $O_NONBLOCK$ is set, opening will always be

successful, unless the whole device doesn't exist. The drive will take

no actions whatsoever.

\end{itemize}

\subsection{And what about standards?}

You might hesitate to accept this proposal as it comes from the

\linux\ community, and not from some standardizing institute. What

about SUN, SGI, HP and all those other Unix and hardware vendors?

Well, these companies are in the lucky position that they generally

control both the hardware and software of their supported products,

and are large enough to set their own standard. They do not have to

deal with a dozen or more different, competing hardware

configurations.\footnote{Personally, I think that SUN's approach to

mounting \cdrom s is very good in origin: under Solaris a

volume-daemon automatically mounts a newly inserted \cdrom\ under {\tt

/cdrom/$<volume-name>$/}. In my opinion they should have pushed this

further and have {\em every\/} \cdrom\ on the local area network be

mounted at the similar location, i.e., no matter in which particular

machine you insert a \cdrom, it will always appear at the same

position in the directory tree, on every system. When I wanted to

implement such a user-program for \linux, I came across the

differences in behavior of the various drivers, and the need for an

$ioctl$ informing about media changes.}

We believe that using $O_NONBLOCK$ to indicate that a device is being opened

for $ioctl$ commands only can be easily introduced in the \linux\

community. All the CD-player authors will have to be informed, we can

even send in our own patches to the programs. The use of $O_NONBLOCK$

has most likely no influence on the behavior of the CD-players on

other operating systems than \linux. Finally, a user can always revert

to old behavior by a call to $ioctl(file_descriptor, CDROM_CLEAR_OPTIONS,

CDO_USE_FFLAGS)$.

\subsection{The preferred strategy of $open()$}

The routines in \cdromc\ are designed in such a way that a run-time

configuration of the behavior of \cdrom\ devices (of {\em any\/} type)

can be carried out, by the $CDROM_SET/CLEAR_OPTIONS$ $ioctls$. Thus, various

modes of operation can be set:

\begin{description}

\item[$CDO_AUTO_CLOSE \mathrel| CDO_USE_FFLAGS \mathrel| CDO_LOCK$]

This is the default setting. (With $CDO_CHECK_TYPE$ it will be better,

in the future.) If the device is not yet opened by any other process,

and it is opened for data ($O_NONBLOCK$ is not set) and the tray is

found open, an attempt to close the tray is made. Then, it is verified

that a disc is in the drive and, if $CDO_CHECK_TYPE$ is set, that its

type is `data mode 1.' Only if all tests are passed, the return value

is zero. The door is locked to prevent file system corruption. If

opened for audio ($O_NONBLOCK$ is set), no actions are taken and a

value of 0 will be returned.

\item[0] $Open()$ will always be successful, the option flags are

ignored. Neither actions are undertaken, nor any integrity checks are

made.

\item[$CDO_AUTO_CLOSE \mathrel| CDO_AUTO_EJECT \mathrel| CDO_LOCK$]

This mimics the behavior of the current sbpcd-driver. The option flags

are ignored, the tray is closed on the first open, if

necessary. Similarly, the tray is opened on the last release, i.e., if

a \cdrom\ is unmounted, it is automatically ejected, such that the

user can replace it.

\end{description}

We hope that these option can convince everybody (both driver

maintainers and user program developers) to adapt to the new \cdrom\

driver scheme and option flag interpretation.

\section{Description of routines in \cdromc}

Only a few routines in \cdromc\ are exported to the drivers. In this

section we will treat these, as well as the functioning of the routines

that `take over' the interface to the kernel. The header file

belonging to \cdromc\ is called \ucdrom, but may be included in {\tt

cdrom.h} in the future.

\subsection{$struct\ file_operations\ cdrom_fops$}

The contents of this structure has been described in

section~\ref{cdrom.c}, and this structure should be used in

registering the block device to the kernel:

$$

register_blkdev(major, <name>, \&cdrom_fops);

$$

\subsection{$Int\ register_cdrom(int\ major, char * name, struct\

cdrom_device_ops\ * cdo)$}

Similar to registering $cdrom_fops$ to the kernel, the device

operations structure, as described in section~\ref{cdrom.c}, should be

registered to the general \cdrom\ interface:

$$

register_cdrom(major, <name>, \&<device>_dops);

$$

This function returns zero upon success, and non-zero upon failure.

\subsection{$Int\ unregister_cdrom(int\ major, char * name)$}

Unregistering device $name$ with major number $major$ disconnects the

registered device-operation routines from the \cdrom\ interface.

This function returns zero upon success, and non-zero upon failure.

\subsection{$Int\ cdrom_open(struct\ inode * ip, struct\ file * fp)$}

This function is not called directly by the low-level drivers, it is

listed in the standard $cdrom_fops$. If the VFS opens a file, this

function becomes active. A strategy logic is implemented in this

routine, taking care of all capabilities and options that are set in

the $cdrom_device_ops$ connected to the device. Then, the program flow is

transferred to the device_dependent $open()$ call.

\subsection{$Void\ cdrom_release(struct\ inode *ip, struct\ file

*fp)$}

This function implements the reverse-logic of $cdrom_open()$, and then

calls the device-dependent $release()$ routine.  When the use-count

has reached 0, the allocated buffers in the are flushed by calls to

$sync_dev(dev)$ and $invalidate_buffers(dev)$.

\subsection{$Int\ cdrom_ioctl(struct\ inode *ip, struct\ file *fp,

                       unsigned\ int\ cmd, unsigned\ long\ arg)$}

\label{cdrom-ioctl}

This function handles all $ioctl$ requests for \cdrom\ devices in a

uniform way. The different calls fall into three categories: $ioctl$s

that can be directly implemented by device operations, ones that are

routed through the call $audio_ioctl()$, and the remaining ones, that

are presumable device-dependent. Generally, a negative return value

indicates an error.

\subsubsection{Directly implemented $ioctl$s}

\label{ioctl-direct}

The following `old' \cdrom-$ioctl$s are implemented by directly

calling device-operations in $cdrom_device_ops$, if implemented and

not masked:

\begin{description}

\item[CDROMMULTISESSION] Requests the last session on a \cdrom.

\item[CDROMEJECT] Open tray.

\item[CDROMCLOSETRAY] Close tray.

\item[CDROMEJECT_SW] If $arg\not=0$, set behavior to auto-close (close

tray on first open) and auto-eject (eject on last release), otherwise

set behavior to non-moving on $open()$ and $release()$ calls.

\item[CDROM_GET_MCN or CDROM_GET_UPC] Get the Medium Catalog Number from a CD.

\end{description}

\subsubsection{$Ioctl$s rooted through $audio_ioctl()$}

\label{ioctl-audio}

The following set of $ioctl$s are all implemented through a call to

the $cdrom_fops$ function $audio_ioctl()$. Memory checks and

allocation are performed in $cdrom_ioctl()$, and also sanitization of

address format ($CDROM_LBA$/$CDROM_MSF$) is done.

\begin{description}

\item[CDROMSUBCHNL] Get sub-channel data in argument $arg$ of type $struct\

cdrom_subchnl *{}$.

\item[CDROMREADTOCHDR] Read Table of Contents header, in $arg$ of type

$struct\ cdrom_tochdr *{}$.

\item[CDROMREADTOCENTRY] Read a Table of Contents entry in $arg$ and

specified by $arg$ of type $struct\ cdrom_tocentry *{}$.

\item[CDROMPLAYMSF] Play audio fragment specified in Minute, Second,

Frame format, delimited by $arg$ of type $struct\ cdrom_msf *{}$.

\item[CDROMPLAYTRKIND] Play audio fragment in track-index format

delimited by $arg$ of type $struct cdrom_ti *{}$.

\item[CDROMVOLCTRL] Set volume specified by $arg$ of type $struct\

cdrom_volctrl *{}$.

\item[CDROMVOLREAD] Read volume into by $arg$ of type $struct\

cdrom_volctrl *{}$.

\item[CDROMSTART] Spin up disc.

\item[CDROMSTOP] Stop playback of audio fragment.

\item[CDROMPAUSE] Pause playback of audio fragment.

\item[CDROMRESUME] Resume playing.

\end{description}

\subsubsection{New $ioctl$s in \cdromc}

The following $ioctl$s have been introduced to allow user programs to

control the behavior of individual \cdrom\ devices. New $ioctl$

commands can be identified by the underscores in their names.

\begin{description}

\item[CDROM_SET_OPTIONS] Set options specified by $arg$. Returns the

option flag register after modification. Use  $arg = \rm0$ for reading

the current flags.

\item[CDROM_CLEAR_OPTIONS] Clear options specified by $arg$. Returns

  the option flag register after modification.

\item[CDROM_SELECT_SPEED] Select head-rate speed of disc specified as

  by $arg$. The value 0 means `auto-select', i.e., play audio discs at

  real time and data disc at maximum speed. The value $arg$ is

  checked against the maximum head rate of the drive found in

  the $cdrom_dops$.

\item[CDROM_SELECT_DISC] Select disc numbered $arg$ from a juke-box.

  First disc is numbered 0. The number $arg$ is checked against the

  maximum number of discs in the juke-box found in the $cdrom_dops$.

\item[CDROM_MEDIA_CHANGED] Returns 1 if a disc has been changed since

  the last call. Note that calls to $cdrom_media_changed$ by the VFS

  are treated by an independent queue, so both mechanisms will detect

  a media change once. Currently, \cdromc\ implements maximum 16 minors

  per major device.

\item[CDROM_DRIVE_STATUS] Returns the status of the drive by a call to

  $drive_status()$. Return values are as defined in section~\ref{drive

    status}. Note that this call doesn't return information on the

  current playing activity of the drive; this can be polled through an

  $ioctl$ call to $CDROMSUBCHNL$.

\item[CDROM_DISC_STATUS] Returns the type of the disc currently in the

  drive by a call to $disc_status()$. Return values are as defined in

  section~\ref{disc status}.

\end{description}

\subsubsection{Device dependent $ioct$s}

Finally, all other $ioctl$s are passed to the function $dev_ioctl()$,

if implemented. No memory allocation or verification is carried out.

\subsection{How to update your driver}

\begin{enumerate}

\item Make a backup of your current driver.

\item Get hold of the files \cdromc\ and \ucdrom, they should be in

the directory tree that came with this documentation.

\item Include {\tt \char`\<linux/ucdrom.h>} just after {\tt cdrom.h}.

\item Change the 3rd argument of $register_blkdev$ from

$\&<your-drive>_fops$ to $\&cdrom_fops$.

\item Just after that line, add a line to register to the \cdrom\

routines:

$$register_cdrom(major, <name>, <your-drive>_dops);$$

Similarly, add a call to $unregister_cdrom()$.

\item Copy an example of the device-operations $struct$ to your source,

e.g., from {\tt cm206.c} $cm206_dops$, and change all entries to names

corresponding to your driver, or names you just happen to like. If

your driver doesn't support a certain function, make the entry

$NULL$. At the entry $capability$ you should list all capabilities

your drive could support, in principle. If your drive has a capability

that is not listed, please send me a message.

\item Implement all functions in your $<device>_dops$ structure,

according to prototypes listed in \ucdrom, and specifications given in

section~\ref{cdrom.c}. Most likely you have already implemented

the code in a large part, and you may just have to adapt the prototype

and return values.

\item Rename your $<device>_ioctl()$ function to $audio_ioctl$ and

change the prototype a little. Remove entries listed in the first part

in section~\ref{cdrom-ioctl}, if your code was OK, these are just calls

to the routines you adapted in the previous step.

\item You may remove all remaining memory checking code in the

$audio_ioctl()$ function that deals with audio commands (these are

listed in the second part of section~\ref{cdrom-ioctl}). There is no

need for memory allocation either, so most $case$s in the $switch$

statement look similar to:

$$

case\ CDROMREADTOCENTRY\colon

get_toc_entry\bigl((struct\ cdrom_tocentry *{})\ arg\bigr);

$$

\item All remaining $ioctl$ cases must be moved to a separate

function, $<device>_ioctl$, the device-dependent $ioctl$s. Note that

memory checking and allocation must be kept in this code!

\item Change the prototypes of $<device>_open()$ and

$<device>_release()$, and remove any strategic code (i.e., tray

movement, door locking, etc.).

\item Try to recompile the drivers. We advice you to use modules, both

for {\tt cdrom.o} and your driver, as debugging is much easier this

way.

\end{enumerate}

\section{Thanks}

Thanks to all the people involved. Thanks to Thomas Quinot, Jon Tombs,

Ken Pizzini, Eberhard M\"onkeberg and Andrew Kroll, the \linux\

\cdrom\ device driver developers who were kind enough to give

suggestions and criticisms during the writing. Finally of course, I

want to thank Linus Torvalds for making this possible in the first

place.

\end{document}