st.txt

linux 内核源代码

This file contains brief information about the SCSI tape driver.
The driver is currently maintained by Kai Mäkisara (email
[e-mail])

Last modified: Mon Mar  7 21:14:44 2005 by kai.makisara


BASICS

The driver is generic, i.e., it does not contain any code tailored
to any specific tape drive. The tape parameters can be specified with
one of the following three methods:

1. Each user can specify the tape parameters he/she wants to use
directly with ioctls. This is administratively a very simple and
flexible method and applicable to single-user workstations. However,
in a multiuser environment the next user finds the tape parameters in
state the previous user left them.

2. The system manager (root) can define default values for some tape
parameters, like block size and density using the MTSETDRVBUFFER ioctl.
These parameters can be programmed to come into effect either when a
new tape is loaded into the drive or if writing begins at the
beginning of the tape. The second method is applicable if the tape
drive performs auto-detection of the tape format well (like some
QIC-drives). The result is that any tape can be read, writing can be
continued using existing format, and the default format is used if
the tape is rewritten from the beginning (or a new tape is written
for the first time). The first method is applicable if the drive
does not perform auto-detection well enough and there is a single
"sensible" mode for the device. An example is a DAT drive that is
used only in variable block mode (I don't know if this is sensible
or not :-).

The user can override the parameters defined by the system
manager. The changes persist until the defaults again come into
effect.

3. By default, up to four modes can be defined and selected using the minor
number (bits 5 and 6). The number of modes can be changed by changing
ST_NBR_MODE_BITS in st.h. Mode 0 corresponds to the defaults discussed
above. Additional modes are dormant until they are defined by the
system manager (root). When specification of a new mode is started,
the configuration of mode 0 is used to provide a starting point for
definition of the new mode.

Using the modes allows the system manager to give the users choices
over some of the buffering parameters not directly accessible to the
users (buffered and asynchronous writes). The modes also allow choices
between formats in multi-tape operations (the explicitly overridden
parameters are reset when a new tape is loaded).

If more than one mode is used, all modes should contain definitions
for the same set of parameters.

Many Unices contain internal tables that associate different modes to
supported devices. The Linux SCSI tape driver does not contain such
tables (and will not do that in future). Instead of that, a utility
program can be made that fetches the inquiry data sent by the device,
scans its database, and sets up the modes using the ioctls. Another
alternative is to make a small script that uses mt to set the defaults
tailored to the system.

The driver supports fixed and variable block size (within buffer
limits). Both the auto-rewind (minor equals device number) and
non-rewind devices (minor is 128 + device number) are implemented.

In variable block mode, the byte count in write() determines the size
of the physical block on tape. When reading, the drive reads the next
tape block and returns to the user the data if the read() byte count
is at least the block size. Otherwise, error ENOMEM is returned.

In fixed block mode, the data transfer between the drive and the
driver is in multiples of the block size. The write() byte count must
be a multiple of the block size. This is not required when reading but
may be advisable for portability.

Support is provided for changing the tape partition and partitioning
of the tape with one or two partitions. By default support for
partitioned tape is disabled for each driver and it can be enabled
with the ioctl MTSETDRVBUFFER.

By default the driver writes one filemark when the device is closed after
writing and the last operation has been a write. Two filemarks can be
optionally written. In both cases end of data is signified by
returning zero bytes for two consecutive reads.

If rewind, offline, bsf, or seek is done and previous tape operation was
write, a filemark is written before moving tape.

The compile options are defined in the file linux/drivers/scsi/st_options.h.

4. If the open option O_NONBLOCK is used, open succeeds even if the
drive is not ready. If O_NONBLOCK is not used, the driver waits for
the drive to become ready. If this does not happen in ST_BLOCK_SECONDS
seconds, open fails with the errno value EIO. With O_NONBLOCK the
device can be opened for writing even if there is a write protected
tape in the drive (commands trying to write something return error if
attempted).


MINOR NUMBERS

The tape driver currently supports 128 drives by default. This number
can be increased by editing st.h and recompiling the driver if
necessary. The upper limit is 2^17 drives if 4 modes for each drive
are used.

The minor numbers consist of the following bit fields:

dev_upper non-rew mode dev-lower
  20 -  8     7    6 5  4      0
The non-rewind bit is always bit 7 (the uppermost bit in the lowermost
byte). The bits defining the mode are below the non-rewind bit. The
remaining bits define the tape device number. This numbering is
backward compatible with the numbering used when the minor number was
only 8 bits wide.


SYSFS SUPPORT

The driver creates the directory /sys/class/scsi_tape and populates it with
directories corresponding to the existing tape devices. There are autorewind
and non-rewind entries for each mode. The names are stxy and nstxy, where x
is the tape number and y a character corresponding to the mode (none, l, m,
a). For example, the directories for the first tape device are (assuming four
modes): st0  nst0  st0l  nst0l  st0m  nst0m  st0a  nst0a.

Each directory contains the entries: default_blksize  default_compression
default_density  defined  dev  device  driver. The file 'defined' contains 1
if the mode is defined and zero if not defined. The files 'default_*' contain
the defaults set by the user. The value -1 means the default is not set. The
file 'dev' contains the device numbers corresponding to this device. The links
'device' and 'driver' point to the SCSI device and driver entries.

A link named 'tape' is made from the SCSI device directory to the class
directory corresponding to the mode 0 auto-rewind device (e.g., st0). 


BSD AND SYS V SEMANTICS

The user can choose between these two behaviours of the tape driver by
defining the value of the symbol ST_SYSV. The semantics differ when a
file being read is closed. The BSD semantics leaves the tape where it
currently is whereas the SYS V semantics moves the tape past the next
filemark unless the filemark has just been crossed.

The default is BSD semantics.


BUFFERING

The driver tries to do transfers directly to/from user space. If this
is not possible, a driver buffer allocated at run-time is used. If
direct i/o is not possible for the whole transfer, the driver buffer
is used (i.e., bounce buffers for individual pages are not
used). Direct i/o can be impossible because of several reasons, e.g.:
- one or more pages are at addresses not reachable by the HBA
- the number of pages in the transfer exceeds the number of
  scatter/gather segments permitted by the HBA
- one or more pages can't be locked into memory (should not happen in
  any reasonable situation)

The size of the driver buffers is always at least one tape block. In fixed
block mode, the minimum buffer size is defined (in 1024 byte units) by
ST_FIXED_BUFFER_BLOCKS. With small block size this allows buffering of
several blocks and using one SCSI read or write to transfer all of the
blocks. Buffering of data across write calls in fixed block mode is
allowed if ST_BUFFER_WRITES is non-zero and direct i/o is not used.
Buffer allocation uses chunks of memory having sizes 2^n * (page
size). Because of this the actual buffer size may be larger than the
minimum allowable buffer size.

NOTE that if direct i/o is used, the small writes are not buffered. This may
cause a surprise when moving from 2.4. There small writes (e.g., tar without
-b option) may have had good throughput but this is not true any more with
2.6. Direct i/o can be turned off to solve this problem but a better solution
is to use bigger write() byte counts (e.g., tar -b 64).

Asynchronous writing. Writing the buffer contents to the tape is
started and the write call returns immediately. The status is checked
at the next tape operation. Asynchronous writes are not done with
direct i/o and not in fixed block mode.

Buffered writes and asynchronous writes may in some rare cases cause
problems in multivolume operations if there is not enough space on the
tape after the early-warning mark to flush the driver buffer.

Read ahead for fixed block mode (ST_READ_AHEAD). Filling the buffer is
attempted even if the user does not want to get all of the data at
this read command. Should be disabled for those drives that don't like
a filemark to truncate a read request or that don't like backspacing.

Scatter/gather buffers (buffers that consist of chunks non-contiguous
in the physical memory) are used if contiguous buffers can't be
allocated. To support all SCSI adapters (including those not
supporting scatter/gather), buffer allocation is using the following
three kinds of chunks:
1. The initial segment that is used for all SCSI adapters including
those not supporting scatter/gather. The size of this buffer will be
(PAGE_SIZE << ST_FIRST_ORDER) bytes if the system can give a chunk of
this size (and it is not larger than the buffer size specified by
ST_BUFFER_BLOCKS). If this size is not available, the driver halves
the size and tries again until the size of one page. The default
settings in st_options.h make the driver to try to allocate all of the
buffer as one chunk.
2. The scatter/gather segments to fill the specified buffer size are
allocated so that as many segments as possible are used but the number
of segments does not exceed ST_FIRST_SG.
3. The remaining segments between ST_MAX_SG (or the module parameter
max_sg_segs) and the number of segments used in phases 1 and 2
are used to extend the buffer at run-time if this is necessary. The
number of scatter/gather segments allowed for the SCSI adapter is not
exceeded if it is smaller than the maximum number of scatter/gather
segments specified. If the maximum number allowed for the SCSI adapter
is smaller than the number of segments used in phases 1 and 2,
extending the buffer will always fail.


EOM BEHAVIOUR WHEN WRITING

When the end of medium early warning is encountered, the current write
is finished and the number of bytes is returned. The next write
returns -1 and errno is set to ENOSPC. To enable writing a trailer,
the next write is allowed to proceed and, if successful, the number of
bytes is returned. After this, -1 and the number of bytes are
alternately returned until the physical end of medium (or some other
error) is encountered.


MODULE PARAMETERS

The buffer size, write threshold, and the maximum number of allocated buffers
are configurable when the driver is loaded as a module. The keywords are:

buffer_kbs=xxx             the buffer size for fixed block mode is set
			   to xxx kilobytes
write_threshold_kbs=xxx    the write threshold in kilobytes set to xxx
max_sg_segs=xxx		   the maximum number of scatter/gather
			   segments
try_direct_io=x		   try direct transfer between user buffer and
			   tape drive if this is non-zero

Note that if the buffer size is changed but the write threshold is not
set, the write threshold is set to the new buffer size - 2 kB.


BOOT TIME CONFIGURATION

If the driver is compiled into the kernel, the same parameters can be