From e2af0daf4a52cd4b313ced4e0a8f6ae4b4479683 Mon Sep 17 00:00:00 2001 From: Michael Kerrisk Date: Tue, 5 Apr 2005 12:20:49 +0000 Subject: [PATCH] Updates from Kai Makisara , the author of the st driver. --- man4/st.4 | 162 +++++++++++++++++++++++++++++++++++------------------- 1 file changed, 106 insertions(+), 56 deletions(-) diff --git a/man4/st.4 b/man4/st.4 index bf5095485..c8cc202d2 100644 --- a/man4/st.4 +++ b/man4/st.4 @@ -1,5 +1,5 @@ .\" Copyright 1995 Robert K. Nichols (Robert.K.Nichols@att.com) -.\" Copyright 1999 Kai Mäkisara (Kai.Makisara@metla.fi) +.\" Copyright 1999-2005 Kai Mäkisara (Kai.Makisara@kolumbus.fi) .\" .\" Permission is granted to make and distribute verbatim copies of this .\" manual provided the copyright notice and this permission notice are @@ -9,7 +9,7 @@ .\" manual under the conditions for verbatim copying, provided that the .\" entire resulting derived work is distributed under the terms of a .\" permission notice identical to this one. -.\" +.\" .\" Since the Linux kernel and libraries are constantly changing, this .\" manual page may be incorrect or out-of-date. The author(s) assume no .\" responsibility for errors or omissions, or for damages resulting from @@ -17,10 +17,10 @@ .\" have taken the same level of care in the production of this manual, .\" which is licensed free of charge, as they might when working .\" professionally. -.\" +.\" .\" Formatted or processed versions of this manual, if unaccompanied by .\" the source, must acknowledge the copyright and authors of this work. -.TH ST 4 1999-01-18 "Linux 2.0 - 2.2" "Linux Programmer's Manual" +.TH ST 4 2005-03-13 "Linux 2.0 - 2.6" "Linux Programmer's Manual" .SH NAME st \- SCSI tape device .SH SYNOPSIS @@ -44,10 +44,12 @@ driver uses major device number 9. .PP Each device uses eight minor device numbers. The lower-most five bits in the minor numbers are assigned sequentially in the order of -detection. The minor numbers can be grouped into +detection. In the 2.6 kernel, the bits above the eight lower-most bits are +concatenated to the five lower-most bits to form the tape number. +The minor numbers can be grouped into two sets of four numbers: the principal (auto-rewind) minor device numbers, .IR n , -and a \(lqno-rewind\(rq device numbers, +and the \(lqno-rewind\(rq device numbers, .IR "" ( n "+ 128)." Devices opened using the principal device number will be sent a \s-1REWIND\s+1 command when they are closed. @@ -88,8 +90,8 @@ The driver uses an internal buffer that has to be large enough to hold at least one tape block. In kernels before 2.1.121, the buffer is allocated as one contiguous block. This limits the block size to the largest contiguous block of memory the kernel allocator can provide. -The limit is currently 128 kB for the 32-bit architectures and -256 kB for the 64-bit architectures. In newer kernels the driver +The limit is currently 128 kB for 32-bit architectures and +256 kB for 64-bit architectures. In newer kernels the driver allocates the buffer in several parts if necessary. By default, the maximum number of parts is 16. This means that the maximum block size is very large (2 MB if allocation of 16 blocks of 128 kB succeeds). @@ -104,8 +106,9 @@ too much on dynamic buffer allocation with kernels older than 2.1.121 .PP The driver does not specifically support any tape drive brand or model. After system start-up the tape device options are defined by -the drive firmware. For example, if the drive firmware selects fixed -block mode, the tape device uses fixed block mode. The options can +the drive firmware. +For example, if the drive firmware selects fixed-block mode, +the tape device uses fixed-block mode. The options can be changed with explicit .B ioctl() calls and remain in effect when the device is closed and reopened. @@ -115,8 +118,8 @@ device. Different options can be specified for the different devices within the subgroup of four. The options take effect when the device is opened. For example, the system administrator can define -one device that writes in fixed block mode with a certain block size, -and one which writes in variable block mode (if the drive supports +one device that writes in fixed-block mode with a certain block size, +and one which writes in variable-block mode (if the drive supports both modes). .PP The driver supports @@ -137,31 +140,44 @@ Device .B /dev/tape is usually created as a hard or soft link to the default tape device on the system. +.PP +Starting from kernel 2.6.2, the driver exports in the sysfs directory +.BR /sys/class/scsi_tape +the attached devices and some parameters assigned to the devices. .SH "DATA TRANSFER" -The driver supports operation in both fixed block mode and variable -block mode (if supported by the drive). In fixed block mode the drive +The driver supports operation in both fixed-block mode and +variable-block mode (if supported by the drive). +In fixed-block mode the drive writes blocks of the specified size and the block size is not -dependent on the byte counts of the write system calls. In variable -block mode one tape block is written for each write call and the byte +dependent on the byte counts of the write system calls. +In variable-block mode one tape block is written for each write call +and the byte count determines the size of the corresponding tape block. Note that -the blocks on the tape do don't contain any information about the +the blocks on the tape don't contain any information about the writing mode: when reading, the only important thing is to use commands that accept the block sizes on the tape. .PP -In variable block mode the read byte count does not have to match +In variable-block mode the read byte count does not have to match the tape block size exactly. If the byte count is larger than the next block on tape, the driver returns the data and the function returns the actual block size. If the block size is larger than the byte count, the requested amount of data from the start of the block is returned and the rest of the block is discarded. .PP -In fixed block mode the read byte counts can be arbitrary if +In fixed-block mode the read byte counts can be arbitrary if buffering is enabled, or a multiple of the tape block size if buffering is disabled. Kernels before 2.1.121 allow writes with arbitrary byte count if buffering is enabled. In all other cases (kernel before 2.1.121 with buffering disabled or newer kernel) the write byte count must be a multiple of the tape block size. .PP +In the 2.6 kernel, the driver tries to use direct transfers between the user +buffer and the device. If this is not possible, the driver's internal buffer +is used. The reasons for not using direct transfers include improper alignment +of the user buffer (default is 512 bytes but this can be changed by the HBA +driver), one of more pages of the user buffer not reachable by the +SCSI adapter, etc. +.PP A filemark is automatically written to tape if the last tape operation before close was a write. .PP @@ -191,8 +207,8 @@ The driver returns an EIO error if the drive rejects an operation. .ta +.4i +.7i +1i /* Structure for \s-1MTIOCTOP\s+1 \- mag tape op command: */ struct mtop { - short mt_op; /* operations defined below */ - int mt_count; /* how many of them */ + short mt_op; /* operations defined below */ + int mt_count; /* how many of them */ }; .fi .PP @@ -224,7 +240,8 @@ is zero. This command uses the MODE page 15 supported by most DATs. .IP MTEOM Go to the end of the recorded media (for appending files). .IP MTERASE -Erase tape. +Erase tape. With 2.6 kernel, short erase (mark tape empty) is performed if the +argument is zero. Otherwise long erase (erase all) is done. .IP MTFSF Forward space over .B mt_count @@ -290,7 +307,7 @@ The density codes supported by a drive can be found from the drive documentation. .IP MTSETPART The active partition is switched to -.B mt_count . +.BR mt_count . The partitions are numbered from zero. This command is not allowed for a drive unless the partition support is enabled for the drive (see MT_ST_CAN_PARTITIONS below). @@ -299,7 +316,7 @@ Execute the SCSI unload command (does not eject the tape). .IP MTUNLOCK Unlock the tape drive door. .IP MTWEOF -Write +Write .B mt_count filemarks. .IP MTWSM @@ -313,7 +330,7 @@ Magnetic Tape operations for setting of device options (by the superuser): .IP MTSETDRVBUFFER 8 Set various drive and driver options according to bits encoded in .BR mt_count . -These consist of the drive's buffering mode, 13 Boolean driver +These consist of the drive's buffering mode, a set of Boolean driver options, the buffer write threshold, defaults for the block size and density, and timeouts (only in kernels >= 2.1). A single operation can affect only one item in the list above (the @@ -353,9 +370,9 @@ above). To set and clear the Boolean options the value in .B mt_count -must include one the constants \s-1MT_ST_BOOLEANS\s+1, +must include one of the constants \s-1MT_ST_BOOLEANS\s+1, \s-1MT_ST_SETBOOLEANS\s+1, \s-1MT_ST_CLEARBOOLEANS\s+1, or -\s-1MT_ST_DEFBOOLEANS\s+1 logically ORed with +\s-1MT_ST_DEFBOOLEANS\s+1 logically or'ed with whatever combination of the following options is desired. Using \s-1MT_ST_BOOLEANS\s+1 the options can be set to the values defined in the corresponding bits. With \s-1MT_ST_SETBOOLEANS\s+1 the @@ -373,12 +390,12 @@ The Boolean options are: .PD 0 .RS .IP "\s-1MT_ST_BUFFER_WRITES\s+1 (Default: true)" -Buffer all write operations in fixed block mode. +Buffer all write operations in fixed-block mode. If this option is false and the drive uses a fixed block size, then all write operations must be for a multiple of the block size. This option must be set false to write reliable multi-volume archives. .IP "\s-1MT_ST_ASYNC_WRITES\s+1 (Default: true)" -When this options is true write operations return immediately without +When this option is true, write operations return immediately without waiting for the data to be transferred to the drive if the data fits into the driver's buffer. The write threshold determines how full the buffer must be before a @@ -388,7 +405,7 @@ operation. This option must be set false to write reliable multi-volume archives. .IP "\s-1MT_ST_READ_AHEAD\s+1 (Default: true)" This option causes the driver to provide read buffering and -read-ahead in fixed block mode. +read-ahead in fixed-block mode. If this option is false and the drive uses a fixed block size, then all read operations must be for a multiple of the block size. .IP "\s-1MT_ST_TWO_FM\s+1 (Default: false)" @@ -439,8 +456,9 @@ correct position when the device is closed and the SCSI command to space backwards over records is used for this purpose. Some older drives can't process this command reliably and this option can be used to instruct the driver not to use the command. The end result is that, -with read-ahead and fixed block mode, the tape may not be correctly -positioned within a file when the device is closed. +with read-ahead and fixed-block mode, the tape may not be correctly +positioned within a file when the device is closed. With 2.6 kernel, the +default is true for drives supporting SCSI-3. .IP "\s-1MT_ST_NO_BLKLIMS\s+1 (Default: false)" Some drives don't accept the READ BLOCK LIMITS SCSI command. If this is used, the driver does not use the command. The drawback is @@ -464,14 +482,17 @@ difference between the semantics is what happens when a device used for reading is closed: in SYSV semantics the tape is spaced forward past the next filemark if this has not happened while using the device. In BSD semantics the tape position is not changed. +.IP "\s-1MT_NO_WAIT\s+1 (Default: false)" +Enables immediate mode (i.e., don't wait for the command to finish) for some +commands (e.g., rewind). .IP \s-1EXAMPLE\s+1 .nf .ta +.4i +.7i +1i .BI "struct mtop " mt_cmd ; .IB "mt_cmd.mt_op" " = \s-1MTSETDRVBUFFER\s+1;" .IB "mt_cmd.mt_count" " = \s-1MT_ST_BOOLEANS\s+1 |" -.B " \s-1MT_ST_BUFFER_WRITES\s+1 |" -.B " \s-1MT_ST_ASYNC_WRITES\s+1;" +.B " \s-1MT_ST_BUFFER_WRITES\s+1 |" +.B " \s-1MT_ST_ASYNC_WRITES\s+1;" .BI "ioctl(" fd ", \s-1MTIOCTOP\s+1, &" mt_cmd ");" .fi .RE @@ -479,11 +500,11 @@ device. In BSD semantics the tape position is not changed. .IP "" The default block size for a device can be set with \s-1MT_ST_DEF_BLKSIZE\s+1 and the default density code can be set with -\s-1MT_ST_DEFDENSITY\s+1. The values for the parameters are ORed +\s-1MT_ST_DEFDENSITY\s+1. The values for the parameters are or'ed with the operation code. .IP "" With kernels 2.1.x and later, the timeout values can be set with the -subcommand \s-1MT_ST_SET_TIMEOUT\s+1 or'ed with the timeout in seconds. +subcommand \s-1MT_ST_SET_TIMEOUT\s+1 ORed with the timeout in seconds. The long timeout (used for rewinds and other commands that may take a long time) can be set with \s-1MT_ST_SET_LONG_TIMEOUT\s+1. The kernel defaults are very long to @@ -492,6 +513,20 @@ drive. Because of this the driver may seem stuck even if it is only waiting for the timeout. These commands can be used to set more practical values for a specific drive. The timeouts set for one device apply for all devices linked to the same drive. +.IP "" +Starting from kernels 2.4.19 and 2.5.43, the driver supports a status +bit which indicates whether the drive requests cleaning. +The method used by the +drive to return cleaning information is set using the +\s-1MT_ST_SEL_CLN\s+1 subcommand. If the value is zero, the cleaning +bit is always zero. If the value is one, the TapeAlert data defined +in the SCSI-3 standard is used (not yet implemented). Values 2-17 are +reserved. If the lowest eight bits are >= 18, bits from the extended +sense data are used. The bits 9-16 specify a mask to select the bits +to look at and the bits 17-23 specify the bit pattern to look for. +If the bit pattern is zero, one or more bits under the mask indicate +the cleaning request. If the pattern is non-zero, the pattern must match +the masked sense data byte. .SS "\s-1MTIOCGET\s+1 \- Get status" .PP This request takes an argument of type @@ -500,15 +535,15 @@ This request takes an argument of type .nf /* structure for \s-1MTIOCGET\s+1 - mag tape get status command */ struct mtget { - long mt_type; - long mt_resid; - /* the following registers are device dependent */ - long mt_dsreg; - long mt_gstat; - long mt_erreg; - /* The next two fields are not always used */ - daddr_t mt_fileno; - daddr_t mt_blkno; + long mt_type; + long mt_resid; + /* the following registers are device dependent */ + long mt_dsreg; + long mt_gstat; + long mt_erreg; + /* The next two fields are not always used */ + daddr_t mt_fileno; + daddr_t mt_blkno; }; .fi .IP \fBmt_type\fP 11 @@ -569,6 +604,9 @@ Immediate report mode. This bit is set if there are no guarantees that the data has been physically written to the tape when the write call returns. It is set zero only when the driver does not buffer data and the drive is set not to buffer data. +.HP +\s-1GMT_CLN(\s+1\fIx\fP\s-1)\s+1: +The drive has requested cleaning. Implemented in kernels >= 2.4.19 and 2.5.43. .RE .IP \fBmt_erreg\fP The only field defined in @@ -604,8 +642,8 @@ Viper, Wangtek, ... ). .PP .nf /* structure for \s-1MTIOCPOS\s+1 - mag tape get position command */ -struct mtpos { - long mt_blkno; /* current block number */ +struct mtpos { + long mt_blkno; /* current block number */ }; .fi @@ -615,6 +653,10 @@ The requested operation could not be completed. .IP ENOSPC A write operation could not be completed because the tape reached end-of-medium. +.IP ENOMEM +The byte count in read() is smaller than the next physical block on +the tape. (Before 2.2.18 and 2.4.0-test6 the extra bytes have been +silently ignored.) .IP EACCES An attempt was made to write or erase a write-protected tape. (This error is not detected during @@ -651,25 +693,25 @@ people have also contributed to the driver. .SH "SEE ALSO" .BR mt (1) .PP -The file README.st in the kernel sources contains the most recent -information about the driver and its configuration +The file README.st or st.txt (kernel >= 2.6) in the kernel sources contains +the most recent information about the driver and its configuration possibilities. .SH NOTES 1. When exchanging data between systems, both systems have to agree on the physical tape block size. The parameters of a drive after startup are often not the ones most operating systems use with these -devices. Most systems use drives in variable block mode if the drive +devices. Most systems use drives in variable-block mode if the drive supports that mode. This applies to most modern drives, including -DATs, 8mm helical scan drives, DLTs, etc. It may be advisable use -these drives in variable block mode also in Linux (i.e., use MTSETBLK +DATs, 8mm helical scan drives, DLTs, etc. It may be advisable to use +these drives in variable-block mode also in Linux (i.e., use MTSETBLK or MTSETDEFBLK at system startup to set the mode), at least when -exchanging data with foreign system. The drawback of +exchanging data with a foreign system. The drawback of this is that a fairly large tape block size has to be used to get acceptable data transfer rates on the SCSI bus. .PP 2. Many programs (e.g., tar) allow the user to specify the blocking -factor on command line. Note that this determines the physical block -size on tape only in variable block mode. +factor on the command line. Note that this determines the physical block +size on tape only in variable-block mode. .PP 3. In order to use SCSI tape drives, the basic SCSI driver, a SCSI-adapter driver and the SCSI tape driver must be either @@ -680,10 +722,18 @@ described in this page is not available. 4. The driver writes error messages to the console/log. The SENSE codes written into some messages are automatically translated to text if verbose SCSI messages are enabled in kernel configuration. +.PP +5. The driver's internal buffering allows good throughput in fixed-block +mode also with small read() and write() byte counts. With direct transfers +this is not possible and may cause a surprise when moving to the 2.6 +kernel. +The solution is to tell the software to use larger transfers (often +telling it to use larger blocks). +If this is not possible, direct transfers can be disabled. .SH COPYRIGHT Copyright \(co 1995 Robert K. Nichols. .br -Copyright \(co 1999 Kai M\(:akisara. +Copyright \(co 1999-2005 Kai M\(:akisara. .PP Permission is granted to make and distribute verbatim copies of this manual provided the copyright notice and this permission notice are