ST(4D) | Devices | ST(4D) |
st - driver for SCSI tape devices
st@target,lun:l,m,h,c,ubn
The st device driver provides a standard interface to various SCSI tape devices. See mtio(4I) for details.
To determine if the st device driver supports your tape device, SPARC users should enter the following on a command line:
% strings /kernel/drv/sparcv9/st | grep -i <tape device name>
x86 users can do the following to determine if the st device driver supports a particular tape device:
% strings /kernel/drv/st | grep -i <tape device name>
The driver can be opened with either rewind on close or no rewind on close options. It can also be opened with the O_NDELAY (see open(2)) option when there is no tape inserted in the drive. A maximum of four tape formats per device are supported (see FILES below). The tape format is specified using the device name. (Tape format is also referred to as tape density).
Following are a list of SCSI commands that can be executed while another host reserves the tape drive. The commands are:
SCMD_TEST_UNIT_READY SCMD_REQUEST_SENSE SCMD_READ_BLKLIM SCMD_INQUIRY SCMD_RESERVE SCMD_RELEASE SCMD_DOORLOCK SCMD_REPORT_DENSITIES SCMD_LOG_SENSE_G1 SCMD_PERSISTENT_RESERVE_IN SCMD_PERSISTENT_RESERVE_OUT SCMD_REPORT_LUNS
In multi-initiator environments, the driver will not reserve the tape drive if above commands are issued. For other SCSI commands, the driver reserves the tape drive and releases the drive at close if it has been reserved. Refer to the MTIOCRESERVE and MTIOCRELEASE ioctls in mtio(4I) for information about how to allow a tape drive to remain reserved upon close. See the flag options below for information about disabling this feature.
If a SCSI-3 persistent reservation is done through the driver, the driver disables all existing SCSI-2 reservations.
If the tape drive is opened in O_NDELAY mode, no reservation occurs during the open, as per the POSIX standard (see standards(7)). However, if a command not found in the above list is used, a reservation will occur to provide reserve/release functionality before the command is issued.
The st driver now supports persistent errors (see mtio(4I) and asynchronous tape operations (see mtio(4I), aioread(3C), and aiowrite(3C)).
If the driver is opened for reading in a different format than the tape is written in, the driver overrides the user-selected format. For example, if a 1/4" cartridge tape is written in QIC-24 format and opened for reading in QIC-150, the driver detects a read failure on the first read and automatically switches to QIC-24 to read the data.
Note that if the low density format is used, no indication is given that the driver has overridden the format you selected. Other formats issue a warning message to inform you of an overridden format selection. Some devices automatically perform this function and do not require driver support (1/2" reel tape drive, for example).
Writing from the beginning of tape is performed in the user-specified format. The original tape format is used for appending onto previously written tapes.
The st driver has a built-in configuration table for most Sun-supported tape drives. For those tape drives that are not in the table, the st driver tries to read the configuration from the tape drive through optional SCSI-3 commands. To support the addition of third party tape devices which are not in the built-in configuration table or not able to report their configuration, device information can be supplied in st.conf as global properties that apply to each node, or as properties that are applicable to one node only. By supplying the information in st.conf, the built-in configuration is overridden and the st driver will not query the configuration from tape drives. The st driver looks for the property called tape-config-list. The value of this property is a list of triplets, where each triplet consists of three strings.
The formal syntax is:
tape-config-list = <triplet> [, <triplet> *];
where
<triplet> := <vid+pid>, <pretty print>, <data-property-name>
and
<data-property-name> = <version>, <type>, <bsize>,
<options>, <number of densities>,
<density> [, <density>*], <default-density>;
or
<data-property-name> = <version 2>, <type>, <bsize>, <options>, <number of densities>, <density> [, <density>*], <default-density>, <non-motion time-out>, <I/O time-out>, <rewind time-out>, <space time-out>, <load time-out>, <unload time-out>, <erase time-out>;
A semicolon (;) is used to terminate a prototype devinfo node specification. Individual elements listed within the specification should not be separated by a semicolon. (Refer to driver.conf(5) for more information.)
<vid+pid> is the string that is returned by the tape device on a SCSI inquiry command. This string may contain any character in the range 0x20-0x7e. Characters such as " " " (double quote) or " ' " (single quote), which are not permitted in property value strings, are represented by their octal equivalent (for example, \042 and \047). Trailing spaces may be truncated.
<pretty print> is used to report the device on the console. This string may have zero length, in which case the <vid+pid> will be used to report the device.
<data-property-name> is the name of the property which contains all the tape configuration values (such as <type>, <bsize>, etc.) corresponding for the tape drive for the specified <vid+pid>.
<version> is a version number and should be 1 or 2. In the future, higher version numbers may be used to allow for changes in the syntax of the <data-property-name> value list.
<type> is a type field. Valid types are defined in /usr/include/sys/mtio.h. For third party tape configuration, the following generic types are recommended:
MT_ISQIC | 0x32 |
MT_ISREEL | 0x33 |
MT_ISDAT | 0x34 |
MT_IS8MM | 0x35 |
MT_ISOTHER | 0x36 |
MT_ISTAND25G | 0x37 |
MT_ISDLT | 0x38 |
MT_ISSTK9840 | 0x39 |
MT_ISBMDLT1 | 0x3A |
MT_LTO | 0x3B |
<bsize> is the preferred block size of the tape device. The value should be 0 for variable block size devices.
<options> is a bit pattern representing the devices, as defined in /usr/include/sys/scsi/targets/stdef.h. Valid flags for tape configuration are shown in the following table. Note that this table does not list flags that are non-configurable in st.conf (including ST_KNOWS_MEDIA which uses the media type reported from the mode select data to select the correct density code).
ST_VARIABLE | 0x0001 |
ST_QIC | 0x0002 |
ST_REEL | 0x0004 |
ST_BSF | 0x0008 |
ST_BSR | 0x0010 |
ST_LONG_ERASE | 0x0020 |
ST_AUTODEN_OVERRIDE | 0x0040 |
ST_NOBUF | 0x0080 |
ST_KNOWS_EOD | 0x0200 |
ST_UNLOADABLE | 0x0400 |
ST_SOFT_ERROR_REPORTING | 0x0800 |
ST_LONG_TIMEOUTS | 0x1000 |
ST_NO_RECSIZE_LIMIT | 0x8000 |
ST_MODE_SEL_COMP | 0x10000 |
ST_NO_RESERVE_RELEASE | 0x20000 |
ST_READ_IGNORE_ILI | 0x40000 |
ST_READ_IGNORE_EOFS | 0x80000 |
ST_SHORT_FILEMARKS | 0x100000 |
ST_EJECT_TAPE_ON_CHANGER_FAILURE | 0x200000 |
ST_RETRY_ON_RECOVERED_DEFERRED_ERROR | 0x400000 |
ST_WORMABLE | 0x1000000 |
ST_VARIABLE
ST_QIC
ST_REEL
ST_BSF
ST_BSR
ST_LONG_ERASE
ST_AUTODEN_OVERRIDE
ST_NOBUF
ST_KNOWS_EOD
ST_UNLOADABLE
ST_SOFT_ERROR_REPORTING
ST_LONG_TIMEOUTS
ST_NO_RECSIZE_LIMIT
ST_MODE_SEL_COMP
ST_NO_RESERVE_RELEASE
ST_READ_IGNORE_ILI
ST_READ_IGNORE_EOFS
ST_SHORT_FILEMARKS
ST_EJECT_TAPE_ON_CHANGER_FAILURE
The following ASC/ASCQ keys are defined to the reasons for causing tape ejection if ST_EJECT_TAPE_ON_CHANGER_FAILURE option is set to 0x200000:
Sense ASC/ASCQ Description
Key
4 15/01 Mechanical Failure
4 44/00 Internal Target Failure
2 53/00 Media Load or Eject Failed
4 53/00 Media Load or Eject Failed
4 53/01 Unload Tape Failure
ST_RETRY_ON_RECOVERED_DEFERRED_ERROR
ST_WORMABLE
<number of densities> is the number of densities specified. Each tape drive can support up to four densities. The value entered should therefore be between 1 and 4; if less than 4, the remaining densities will be assigned a value of 0x0.
<density> is a single-byte hexadecimal number. It can either be found in the device specification manual or be obtained from the device vendor.
<default-density> has a value between 0 and (<number of densities> - 1).
<non-motion time-out> Time in seconds that the drive should be able to perform any SCSI command that doesn't require tape to be moved. This includes mode sense, mode select, reserve, release, read block limits, and test unit ready.
<I/O time-out> Time in seconds to perform data transfer I/O to or from tape including worst case error recovery.
<rewind time-out> Time in seconds to rewind from anywhere on tape to BOT including worst case recovery forcing buffered write data to tape.
<space time-out> Time in seconds to space to any file, block or end of data on tape. Including worst case when any form of cataloging is invalid.
<load time-out> Time in seconds to load tape and be ready to transfer first block. This should include worst case recovery reading tape catalog or drive specific operations done at load.
<unload time-out> Time in seconds to unload tape. Should include worst case time to write to catalog, unthread, and tape cartridge unloading. Also should include worst case time for any drive specific operations that are preformed at unload. Should not include rewind time as the driver rewinds tape before issuing the unload.
<erase time-out> Time in seconds to preform a full (BOT to EOT) erase of longest medium with worst case error recovery.
Each device maintains I/O statistics both for the device and for each partition allocated on that device. For each device/partition, the driver accumulates reads, writes, bytes read, and bytes written. The driver also takes hi-resolution time stamps at queue entry and exit points, which facilitates monitoring the residence time and cumulative residence-length product for each queue.
Each device also has error statistics associated with it. These must include counters for hard errors, soft errors and transport errors. Other data may be implemented as required.
The behavior of SCSI tape positioning ioctls is the same across all devices which support them. (Refer to mtio(4I).) However, not all devices support all ioctls. The driver returns an ENOTTY error on unsupported ioctls.
The retension ioctl only applies to 1/4" cartridge tape devices. It is used to restore tape tension, thus improving the tape's soft error rate after extensive start-stop operations or long-term storage.
In order to increase performance of variable-length tape devices (particularly when they are used to read/write small record sizes), two operations in the MTIOCTOP ioctl, MTSRSZ and MTGRSZ, can be used to set and get fixed record lengths. The ioctl also works with fixed-length tape drives which allow multiple record sizes. The min/max limits of record size allowed on a driver are found by using a SCSI-2 READ BLOCK LIMITS command to the device. If this command fails, the default min/max record sizes allowed are 1 byte and 63k bytes. An application that needs to use a different record size opens the device, sets the size with the MTSRSZ ioctl, and then continues with I/O. The scope of the change in record size remains until the device is closed. The next open to the device resets the record size to the default record size (retrieved from st.conf).
Note that the error status is reset by the MTIOCGET get status ioctl call or by the next read, write, or other ioctl operation. If no error has occurred (sense key is 0), the current file and record position is returned.
EACCES
EBUSY
EINVAL
EIO
ENOTTY
ENXIO
ENOMEM
Example 1 Global tape-config list property
The following is an example of a global tape-config-list property:
tape-config-list = "Magic DAT", "Magic 4mm Helical Scan", "magic-data", "Major Appliance", "Major Appliance Tape", "major-tape"; magic-data = 1,0x34,1024,0x1639,4,0,0x8c,0x8c,0x8c,3; major-tape = 2,0x3c,0,0x18619,4,0x0,0x0,0x0,0x0,
3,0,0,30,120,0,0,36000; name="st" class="scsi"
target=0 lun=0; name="st" class="scsi"
target=1 lun=0; name="st" class="scsi"
target=2 lun=0;
.
.
. name="st" class="scsi"
target=6 lun=0;
Example 2 Tape-config-list property applicable to target 2 only
The following is an example of a tape-config-list property applicable to target 2 only:
name="st" class="scsi"
target=0 lun=0; name="st" class="scsi"
target=1 lun=0; name="st" class="scsi"
target=2 lun=0
tape-config-list =
"Magic DAT", "Magic 4mm Helical Scan", "magic-data"
magic-data = 1,0x34,1024,0x1639,4,0,0x8c,0x8c,0x8c,3; name="st" class="scsi"
target=3 lun=0;
.
.
. name="st" class="scsi"
target=6 lun=0;
To support applications such as seismic programs that require large record sizes, the flag ST_NO_RECSIZE_LIMIT must be set in drive option in the configuration entry. A SCSI tape drive that needs to transfer large records should OR this flag with other flags in the 'options' field in st.conf. (Refer to Tape Configuration.) By default, this flag is set for the built-in config entries of Archive DAT and Exabyte drives.
If this flag is set, the st driver issues a SCSI-2 READ BLOCK LIMITS command to the device to determine the maximum record size allowed by it. If the command fails, st continues to use the maximum record sizes mentioned in the mtio(4I) man page.
If the command succeeds, st restricts the maximum transfer size of a variable-length device to the minimum of that record size and the maximum DMA size that the host adapter can handle. Fixed-length devices are bound by the maximum DMA size allocated by the machine. Note that tapes created with a large record size may not be readable by earlier releases or on other platforms.
(Refer to the WARNINGS section for more information.)
The Emulex drives have only a physical end of tape (PEOT); thus it is not possible to write past EOT. All other drives have a logical end of tape (LEOT) before PEOT to guarantee flushing the data onto the tape. The amount of storage between LEOT and PEOT varies from less than 1 Mbyte to about 20 Mbyte, depending on the tape drive.
If EOT is encountered while writing an Emulex, no error is reported but the number of bytes transferred is 0 and no further writing is allowed. On all other drives, the first write that encounters EOT will return a short count or 0. If a short count is returned, then the next write will return 0. After a zero count is returned, the next write returns a full count or short count. A following write returns 0 again. It is important that the number and size of trailer records be kept as small as possible to prevent data loss. Therefore, writing after EOT is not recommended.
Reading past EOT is transparent to the user. Reading is stopped only by reading EOF's. For 1/2" reel devices, it is possible to read off the end of the reel if one reads past the two file marks which mark the end of recorded media.
/kernel/drv/st.conf
/usr/include/sys/mtio.h
/usr/include/sys/scsi/targets/stdef.h
/dev/rmt/[0− 127][l,m,h,u,c][b][n]
For 1/2" reel tape devices (HP-88780), the densities are:
l | 800 BPI density |
m | 1600 BPI density |
h | 6250 BPI density |
c | data compression |
(not supported on all modules) |
For 8mm tape devices (Exabyte 8200/8500/8505):
l | Standard 2 Gbyte format |
m | 5 Gbyte format (8500, 8505 only) |
h,c | 5 Gbyte compressed format (8505 only) |
For 4mm DAT tape devices (Archive Python):
l | Standard format |
m,h,c | data compression |
For all QIC (other than QIC-24) tape devices:
l,m,h,c | density of the tape cartridge type |
(not all devices can read and | |
write all formats) |
For QIC-24 tape devices (Emulex MT−02):
l | QIC-11 Format |
m,h,c | QIC-24 Format |
mt(1), open(2), read(2), write(2), aioread(3C), aiowrite(3C), kstat(3KSTAT), esp(4D), isp(4D), mtio(4I), driver.conf(5), scsi(5), standards(7), modload(8), modunload(8), ioctl(9E)
The st driver diagnostics may be printed to the console or messages file.
Each diagnostic is dependent on the value of the system variable st_error_level. st_error_level may be set in the /etc/system file. The default setting for st_error_level is 4 (SCSI_ERR_RETRYABLE) which is suitable for most configurations since only actual fault diagnostics are printed. Settings range from values 0 (SCSI_ERR_ALL) which is most verbose, to 6 (SCSI_ERR_NONE) which is least verbose. See stdef.h for the full list of error-levels. SCSI_ERR_ALL level the amount of diagnostic information is likely to be excessive and unnecessary.
The st driver diagnostics are described below:
Error for Command: <scsi_cmd_name()> Error Level:<error_class> Requested Block: <blkno> Error Block: <err_blkno> Vendor: <name>: Serial Number: <inq_serial> Sense Key: <es_key> ASC: 0x<es_add_code> (scsi_asc_ascq_name()>), ASCQ: 0x<es_qual_code>, FRU: 0x<ex_fru_code>
where <error_class> may be any one of the following: "All," "Unknown," "Informational," "Recovered," "Retryable," "Fatal"
The command indicated by <scsi_cmd_name> failed. Requested Block represents the block where the transfer started. Error Block represents the block that caused the error. Sense Key, ASC, ASCQ and FRU information is returned by the target in response to a request sense command. See SCSI protocol documentation for description of Sense Key, ASC, ASCQ, FRU.
The st driver attempts to validate entries in the st.conf file. Each field in the entry is checked for upper and lower limits and invalid bits set. The fields are named as follows in config string order:
conf version
drive type
block size
options
number of densities
density code
default density
non motion timeout
I/O timeout
space timeout
load timeout
unload timeout
erase timeout
The st.conf diagnostics are described below:
<con-name> <field-in-err> <problem-with-field>
where <con-name> is the name of the config string. Where <field-in-err> is the field containing invalid entries and where <problem-with-field> describes the nature of the invalid entry.
Write/read: not modulo <n> block size
The request size for fixed record size devices must be a multiple of the specified block size.
Recovery by resets failed
After a transport error, the driver attempted to recover by issuing a device reset and then a bus reset if device reset failed. These recoveries failed.
Periodic head cleaning required
The driver reported that periodic head cleaning is now required. This diagnostic is generated either due to a threshold number of retries, or due to the device communicating to the driver that head cleaning is required.
Soft error rate (<n>%) during writing/reading was too high
The soft error rate has exceeded the threshold specified by the vendor.
SCSI transport failed: reason 'xxxx': {retrying|giving up}
The Host Bus Adapter (HBA) has failed to transport a command to the target for the reason stated. The driver will either retry the command or, ultimately, give up.
Tape not inserted in drive
A media access command was attempted while there was no tape inserted into the specified drive. In this case, the drive returns sense key of DRIVE NOT READY.
Transport rejected
The Host Bus Adapter (HBA) driver is not accepting commands after failing to successfully transport a scsi packet to the target. The actual status received by the st driver from the underlying HBA driver was either TRAN_FATAL_ERROR or TRAN_BADPKT.
Retrying command
The st driver failed to complete a command. However the command is retryable and will be retried.
Giving up
The st driver has exhausted retries or otherwise is unable to retry the command and so is giving up.
No target struct for st%d
The st driver failed to obtain state information because the requested state structure was not allocated. The specified device was probably not attached.
File mark detected
The operation detected an end of file mark. (File marks signify the end of a file on the tape media).
End-of-media detected
The operation reached the end of the tape media.
Exabyte soft error reporting failed. DAT soft error reporting failed
The st driver was unable to determine if the soft error threshold had been exceeded because it did not successfully read the data it requires or did not obtain enough data. This data is retrieved using the log sense command.
Log sense parameter code does not make sense
The log sense command retrieves hardware statistics that are stored on the drive (for example, soft error counts and retries.) If the data retrieved from the drive is invalid, this message is printed and the data is not used.
Restoring tape position at fileno=%x, blkno=%lx....
The st driver is positioning to the specified file and block. This occurs on an open.
Failed to restore the last <file/block> position: In this state, tape will be loaded at BOT during next open
The st driver could not position to the specified location and will revert to the beginning of the tape when the next open is attempted.
Device does not support compression
The compression facility of the device was requested. However the device does not have a hardware compression capability.
DAT soft error reset failed
After DAT soft error reporting, the counters within the device that accumulate this sense data need to be re-set. This operation failed.
Errors after pkt alloc (b_flags=0x%x, b_error=0x%x)
Memory allocation for a scsi packet failed.
Incorrect length indicator set
The drive reported the length of data requested in a READ operation, was incorrect. Incorrect Length Indicator (ILI) is a very commonly used facility in SCSI tape protocol and should not be seen as an error per-se. Applications typically probe a new tape with a read of any length, using the returned length to the read system call for future reads. Along with this operation, an underlying ILI error is received. ILI errors are therefore informational only and are masked at the default st_error_level.
Data property (%s) has no value Data property (%s) incomplete Version # for data property (%s) greater than 1
These diagnostics indicate problems in retrieving the values of the various property settings. The st driver is in the process of setting the property/parameter values for the tape drive using information from either the built-in table within the driver or from uncommented entries in the st.conf file. The effect on the system may be that the tape drive may be set with default or generic driver settings which may not be appropriate for the actual type of tape drive being used.
st_attach-RESUME: tape failure tape position will be lost
On a resume after a power management suspend, the previously known tape position is no longer valid. This can occur if the tape was changed while the system was in power management suspend. The operation will not be retried.
Write Data Buffering has been deprecated. Your applications should continue to work normally. However, they should be ported to use Asynchronous I/O.
Indicates that buffering has been removed from Solaris.
Cannot detach: fileno=%x, blkno=%lx
The st driver cannot unload because the tape is not positioned at BOT (beginning of tape). May indicate hardware problems with the tape drive.
Variable record length I/O Fixed record length (%d byte blocks) I/O
Tape-drives can use either Fixed or Variable record length. If the drive uses Fixed length records, then the built in property table or the st.conf file will contain a non-zero record-length property. Most DAT, Exabyte and DLT drives support Variable record lengths. Many QIC format tape drives have historically been of Fixed record length.
Command will be retried
un_ncmds: %d can't retry cmd
These diagnostics are only seen with tape drives with the ST_RETRY_ON_RECOVERED_DEFERRED_ERROR bit set. See stdef.h for explanation of the specific usage of this setting.
Effective with Solaris 2.4, the ST_NO_RECSIZE_LIMIT flag is set for the built-in config entries of the Archive DAT and Exabyte drivers by default. (Refer to Large Record Sizes.) Tapes written with large block sizes prior to Solaris 2.4 may cause some applications to fail if the number of bytes returned by a read request is less than the requested block size (for example, asking for 128 Kbytes and receiving less than 64 Kbytes).
The ST_NO_RECSIZE_LIMIT flag can be disabled in the config entry for the device as a work-around. (Refer to Tape Configuration.) This action disables the ability to read and write with large block sizes and allows the reading of tapes written prior to Solaris 2.4 with large block sizes.
(Refer to mtio(4I) for a description of maximum record sizes.)
Tape devices that do not return a BUSY status during tape loading prevent user commands from being held until the device is ready. The user must delay issuing any tape operations until the tape device is ready. This is not a problem for tape devices supplied by Sun Microsystems.
Tape devices that do not report a blank check error at the end of recorded media may cause file positioning operations to fail. Some tape drives, for example, mistakenly report media error instead of blank check error.
April 9, 2016 | OmniOS |