freebsd-dev/share/man/man4/sa.4
Kenneth D. Merry 93729c1796 Add support to physio(9) for devices that don't want I/O split and
configure sa(4) to request no I/O splitting by default.

For tape devices, the user needs to be able to clearly understand
what blocksize is actually being used when writing to a tape
device.  The previous behavior of physio(9) was that it would split
up any I/O that was too large for the device, or too large to fit
into MAXPHYS.  This means that if, for instance, the user wrote a
1MB block to a tape device, and MAXPHYS was 128KB, the 1MB write
would be split into 8 128K chunks.  This would be done without
informing the user.

This has suboptimal effects, especially when trying to communicate
status to the user.  In the event of an error writing to a tape
(e.g. physical end of tape) in the middle of a 1MB block that has
been split into 8 pieces, the user could have the first two 128K
pieces written successfully, the third returned with an error, and
the last 5 returned with 0 bytes written.  If the user is using
a standard write(2) system call, all he will see is the ENOSPC
error.  He won't have a clue how much actually got written.  (With
a writev(2) system call, he should be able to determine how much
got written in addition to the error.)

The solution is to prevent physio(9) from splitting the I/O.  The
new cdev flag, SI_NOSPLIT, tells physio that the driver does not
want I/O to be split beforehand.

Although the sa(4) driver now enables SI_NOSPLIT by default,
that can be disabled by two loader tunables for now.  It will not
be configurable starting in FreeBSD 11.0.  kern.cam.sa.allow_io_split
allows the user to configure I/O splitting for all sa(4) driver
instances.  kern.cam.sa.%d.allow_io_split allows the user to
configure I/O splitting for a specific sa(4) instance.

There are also now three sa(4) driver sysctl variables that let the
users see some sa(4) driver values.  kern.cam.sa.%d.allow_io_split
shows whether I/O splitting is turned on.  kern.cam.sa.%d.maxio shows
the maximum I/O size allowed by kernel configuration parameters
(e.g. MAXPHYS, DFLTPHYS) and the capabilities of the controller.
kern.cam.sa.%d.cpi_maxio shows the maximum I/O size supported by
the controller.

Note that a better long term solution would be to implement support
for chaining buffers, so that that MAXPHYS is no longer a limiting
factor for I/O size to tape and disk devices.  At that point, the
controller and the tape drive would become the limiting factors.

sys/conf.h:	Add a new cdev flag, SI_NOSPLIT, that allows a
		driver to tell physio not to split up I/O.

sys/param.h:	Bump __FreeBSD_version to 1000049 for the addition
		of the SI_NOSPLIT cdev flag.

kern_physio.c:	If the SI_NOSPLIT flag is set on the cdev, return
		any I/O that is larger than si_iosize_max or
		MAXPHYS, has more than one segment, or would have
		to be split because of misalignment with EFBIG.
		(File too large).

		In the event of an error, print a console message to
		give the user a clue about what happened.

scsi_sa.c:	Set the SI_NOSPLIT cdev flag on the devices created
		for the sa(4) driver by default.

		Add tunables to control whether we allow I/O splitting
		in physio(9).

		Explain in the comments that allowing I/O splitting
		will be deprecated for the sa(4) driver in FreeBSD
		11.0.

		Add sysctl variables to display the maximum I/O
		size we can do (which could be further limited by
		read block limits) and the maximum I/O size that
		the controller can do.

		Limit our maximum I/O size (recorded in the cdev's
		si_iosize_max) by MAXPHYS.  This isn't strictly
		necessary, because physio(9) will limit it to
		MAXPHYS, but it will provide some clarity for the
		application.

		Record the controller's maximum I/O size reported
		in the Path Inquiry CCB.

sa.4:		Document the block size behavior, and explain that
		the option of allowing physio(9) to split the I/O
		will disappear in FreeBSD 11.0.

Sponsored by:	Spectra Logic
2013-08-24 04:52:22 +00:00

312 lines
10 KiB
Groff

.\" Copyright (c) 1996
.\" Julian Elischer <julian@FreeBSD.org>. All rights reserved.
.\"
.\" Redistribution and use in source and binary forms, with or without
.\" modification, are permitted provided that the following conditions
.\" are met:
.\" 1. Redistributions of source code must retain the above copyright
.\" notice, this list of conditions and the following disclaimer.
.\"
.\" 2. Redistributions in binary form must reproduce the above copyright
.\" notice, this list of conditions and the following disclaimer in the
.\" documentation and/or other materials provided with the distribution.
.\"
.\" THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
.\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
.\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
.\" ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
.\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
.\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
.\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
.\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
.\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
.\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
.\" SUCH DAMAGE.
.\"
.\" $FreeBSD$
.\"
.Dd August 23, 2013
.Dt SA 4
.Os
.Sh NAME
.Nm sa
.Nd SCSI Sequential Access device driver
.Sh SYNOPSIS
.Cd device sa
.Sh DESCRIPTION
The
.Nm
driver provides support for all
.Tn SCSI
devices of the sequential access class that are attached to the system
through a supported
.Tn SCSI
Host Adapter.
The sequential access class includes tape and other linear access devices.
.Pp
A
.Tn SCSI
Host
adapter must also be separately configured into the system
before a
.Tn SCSI
sequential access device can be configured.
.Sh MOUNT SESSIONS
The
.Nm
driver is based around the concept of a
.Dq Em mount session ,
which is defined as the period between the time that a tape is
mounted, and the time when it is unmounted.
Any parameters set during
a mount session remain in effect for the remainder of the session or
until replaced.
The tape can be unmounted, bringing the session to a
close in several ways.
These include:
.Bl -enum
.It
Closing a `rewind device',
referred to as sub-mode 00 below.
An example is
.Pa /dev/sa0 .
.It
Using the MTOFFL
.Xr ioctl 2
command, reachable through the
.Sq Cm offline
command of
.Xr mt 1 .
.El
.Pp
It should be noted that tape devices are exclusive open devices, except in
the case where a control mode device is opened.
In the latter case, exclusive
access is only sought when needed (e.g., to set parameters).
.Sh SUB-MODES
Bits 0 and 1 of the minor number are interpreted as
.Sq sub-modes .
The sub-modes differ in the action taken when the device is closed:
.Bl -tag -width XXXX
.It 00
A close will rewind the device; if the tape has been
written, then a file mark will be written before the rewind is requested.
The device is unmounted.
.It 01
A close will leave the tape mounted.
If the tape was written to, a file mark will be written.
No other head positioning takes place.
Any further reads or writes will occur directly after the
last read, or the written file mark.
.It 10
A close will rewind the device.
If the tape has been
written, then a file mark will be written before the rewind is requested.
On completion of the rewind an unload command will be issued.
The device is unmounted.
.El
.Sh BLOCKING MODES
.Tn SCSI
tapes may run in either
.Sq Em variable
or
.Sq Em fixed
block-size modes.
Most
.Tn QIC Ns -type
devices run in fixed block-size mode, where most nine-track tapes and
many new cartridge formats allow variable block-size.
The difference between the two is as follows:
.Bl -inset
.It Variable block-size:
Each write made to the device results in a single logical record
written to the tape.
One can never read or write
.Em part
of a record from tape (though you may request a larger block and read
a smaller record); nor can one read multiple blocks.
Data from a single write is therefore read by a single read.
The block size used
may be any value supported by the device, the
.Tn SCSI
adapter and the system (usually between 1 byte and 64 Kbytes,
sometimes more).
.Pp
When reading a variable record/block from the tape, the head is
logically considered to be immediately after the last item read,
and before the next item after that.
If the next item is a file mark,
but it was never read, then the next
process to read will immediately hit the file mark and receive an end-of-file notification.
.It Fixed block-size:
Data written by the user is passed to the tape as a succession of
fixed size blocks.
It may be contiguous in memory, but it is
considered to be a series of independent blocks.
One may never write
an amount of data that is not an exact multiple of the blocksize.
One may read and write the same data as a different set of records.
In other words, blocks that were written together may be read separately,
and vice-versa.
.Pp
If one requests more blocks than remain in the file, the drive will
encounter the file mark.
As there is some data to return (unless
there were no records before the file mark), the read will succeed,
returning that data.
The next read will return immediately with a value
of 0.
(As above, if the file mark is never read, it remains for the next
process to read if in no-rewind mode.)
.El
.Sh BLOCK SIZES
By default, the driver will NOT accept reads or writes to a tape device that
are larger than may be written to or read from the mounted tape using a single
write or read request.
Because of this, the application author may have confidence that his wishes
are respected in terms of the block size written to tape.
For example, if the user tries to write a 256KB block to the tape, but the
controller can handle no more than 128KB, the write will fail.
The previous
.Fx
behavior, prior to
.Fx
10.0,
was to break up large reads or writes into smaller blocks when going to the
tape.
The problem with that behavior, though, is that it hides the actual on-tape
block size from the application writer, at least in variable block mode.
.Pp
If the user would like his large reads and writes broken up into separate
pieces, he may set the following loader tunables.
Note that these tunables WILL GO AWAY in
.Fx 11.0 .
They are provided for transition purposes only.
.Bl -tag -width 12
.It kern.cam.sa.allow_io_split
.Pp
This variable, when set to 1, will configure all
.Nm
devices to split large buffers into smaller pieces when needed.
.It kern.cam.sa.%d.allow_io_split
.Pp
This variable, when set to 1, will configure the given
.Nm
unit to split large buffers into multiple pieces.
This will override the global setting, if it exists.
.El
.Pp
There are several
.Xr sysctl 8
variables available to view block handling parameters:
.Bl -tag -width 12
.It kern.cam.sa.%d.allow_io_split
.Pp
This variable allows the user to see, but not modify, the current I/O split
setting.
The user is not permitted to modify this setting so that there is no chance
of behavior changing for the application while a tape is mounted.
.It kern.cam.sa.%d.maxio
.Pp
This variable shows the maximum I/O size in bytes that is allowed by the
combination of kernel tuning parameters (MAXPHYS, DFLTPHYS) and the
capabilities of the controller that is attached to the tape drive.
Applications may look at this value for a guide on how large an I/O may be
permitted, but should keep in mind that the actual maximum may be
restricted further by the tape drive via the
.Tn SCSI
READ BLOCK LIMITS command.
.It kern.cam.sa.%d.cpi_maxio
.Pp
This variable shows the maximum I/O size supported by the controller, in
bytes, that is reported via the CAM Path Inquiry CCB (XPT_PATH_INQ).
If this is 0, that means that the controller has not reported a maximum I/O
size.
.El
.Sh FILE MARK HANDLING
The handling of file marks on write is automatic.
If the user has
written to the tape, and has not done a read since the last write,
then a file mark will be written to the tape when the device is
closed.
If a rewind is requested after a write, then the driver
assumes that the last file on the tape has been written, and ensures
that there are two file marks written to the tape.
The exception to
this is that there seems to be a standard (which we follow, but do not
understand why) that certain types of tape do not actually write two
file marks to tape, but when read, report a `phantom' file mark when the
last file is read.
These devices include the QIC family of devices.
(It might be that this set of devices is the same set as that of fixed
block devices.
This has not been determined yet, and they are treated
as separate behaviors by the driver at this time.)
.Sh IOCTLS
The
.Nm
driver supports all of the ioctls of
.Xr mtio 4 .
.Sh FILES
.Bl -tag -width /dev/[n][e]sa[0-9] -compact
.It Pa /dev/[n][e]sa[0-9]
general form:
.It Pa /dev/sa0
Rewind on close
.It Pa /dev/nsa0
No rewind on close
.It Pa /dev/esa0
Eject on close (if capable)
.It Pa /dev/sa0.ctl
Control mode device (to examine state while another program is
accessing the device, e.g.).
.El
.Sh DIAGNOSTICS
None.
.Sh SEE ALSO
.Xr cam 4 ,
.Xr mt 1
.Sh AUTHORS
.An -nosplit
The
.Nm
driver was written for the
.Tn CAM
.Tn SCSI
subsystem by
.An Justin T. Gibbs
and
.An Kenneth Merry .
Many ideas were gleaned from the
.Nm st
device driver written and ported from
.Tn Mach
2.5
by
.An Julian Elischer .
.Pp
The current owner of record is
.An Matthew Jacob
who has suffered too many
years of breaking tape drivers.
.Sh BUGS
This driver lacks many of the hacks required to deal with older devices.
Many older
.Tn SCSI-1
devices may not work properly with this driver yet.
.Pp
Additionally, certain
tapes (QIC tapes mostly) that were written under
.Fx
2.X
are not automatically read correctly with this driver: you may need to
explicitly set variable block mode or set to the blocksize that works best
for your device in order to read tapes written under
.Fx
2.X.
.Pp
Fine grained density and compression mode support that is bound to specific
device names needs to be added.
.Pp
Support for fast indexing by use of partitions is missing.