id - disk driver for IPI disk controllers

---

CONFIG -- SUN-4, SPARCsystem 600MP SERIES

controller	idc0	at ipi ? csr 0x0000ff priority 2   # channel 0 slave 0
disk		id 0x000 at idc0 drive 0 	# facility 0
disk		id 0x001 at idc0 drive 1
disk		id 0x002 at idc0 drive 2
disk		id 0x003 at idc0 drive 3
disk		id 0x004 at idc0 drive 4
disk		id 0x005 at idc0 drive 5
disk		id 0x006 at idc0 drive 6
disk		id 0x007 at idc0 drive 7

controller	idc1	at ipi ? csr 0x0001ff priority 2 # channel 0 slave 1
disk		id 0x010 at idc1 drive 0 	# facility 0
disk		id 0x011 at idc1 drive 1
disk		id 0x012 at idc1 drive 2
disk		id 0x013 at idc1 drive 3
disk		id 0x014 at idc1 drive 4
disk		id 0x015 at idc1 drive 5
disk		id 0x016 at idc1 drive 6
disk		id 0x017 at idc1 drive 7

controller	idc2	at ipi ? csr 0x0002ff priority 2 # channel 0 slave 2
disk		id 0x020 at idc2 drive 0 	# facility 0
disk		id 0x021 at idc2 drive 1
disk		id 0x022 at idc2 drive 2
disk		id 0x023 at idc2 drive 3
disk		id 0x024 at idc2 drive 4
disk		id 0x025 at idc2 drive 5
disk		id 0x026 at idc2 drive 6
disk		id 0x027 at idc2 drive 7

controller	idc3	at ipi ? csr 0x0003ff priority 2 # channel 0 slave 3
disk		id 0x030 at idc3 drive 0 	# facility 0
disk		id 0x031 at idc3 drive 1
disk		id 0x032 at idc3 drive 2
disk		id 0x033 at idc3 drive 3
disk		id 0x034 at idc3 drive 4
disk		id 0x035 at idc3 drive 5
disk		id 0x036 at idc3 drive 6
disk		id 0x037 at idc3 drive 7

controller	idc4	at ipi ? csr 0x0004ff priority 2  # channel 0 slave 4
disk		id 0x040 at idc4 drive 0 	# facility 0
disk		id 0x041 at idc4 drive 1
disk		id 0x042 at idc4 drive 2
disk		id 0x043 at idc4 drive 3
disk		id 0x044 at idc4 drive 4
disk		id 0x045 at idc4 drive 5
disk		id 0x046 at idc4 drive 6
disk		id 0x047 at idc4 drive 7

The first four controller lines given in the synopsis section above correspond to the first, second, third, and fourth IPI disk controllers in a Sun system; the fifth controller is currently available on SPARCsystem 600MP series machines only. These controllers are on IPI channels or combined on integrated channel/disk-controller cards such as the ISP-80 IPI disk controller. The csr value gives the channel, slave, and facility address of the controller. The facility address should always be 0xff. See is.4s

DESCRIPTION

In order to accommodate a large number of disk drives, a multiple major number device addressing scheme is used. The minor number is formed using the low order 5 bits of the unit number and the 3-bit partition number. The low-order 3 bits of the minor number give the partition number on the drive. The high order bits of the unit number are added to the first major number for id devices to give the major number of the particular device. The unit number itself is formed as follows: the 4-bit facility number forms the low order 4-bits, next the 3-bit slave number, and then the channel number.

The standard device names begin with id followed by three hex digits giving the channel number, slave number, and facility number, and then a letter a-h, for partitions 0-7 respectively.

The block files access the disk using the system's normal buffering mechanism and may be read and written without regard to physical disk records. There is also a raw interface which provides for direct transmission between the disk and the user's read or write buffer. A single read or write call usually results in only one I/O operation; therefore raw I/O is considerably more efficient when many words are transmitted. The names of the raw files conventionally begin with an extra `r'.

In raw I/O, counts should be a multiple of 512 bytes (a disk sector). Likewise directory.3 calls should specify a multiple of 512 bytes. Depending on the channel adaptor, the buffer for raw reads or writes may be required to be on a 2-byte or 4-byte boundary.

The ioctl() interface described in dkio.4s is supported by this driver. The DKIOCSCMD ioctl can be used to issue certain IPI commands to the drive. The argument structure is:

struct dk_cmd {
	u_short	dkc_cmd;	/* command to be executed */
	int	dkc_flags;	/* execution flags */
	daddr_t	dkc_blkno;	/* disk address for command */
	int	dkc_secnt;	/* sector count for command */
	caddr_t	dkc_bufaddr;	/* user's buffer address */
	u_int	dkc_buflen;	/* size of user's buffer */
};

The lower 8-bits of the dkc_cmd field indicate one of the supported commands listed below. The upper 8-bits indicate the IPI Opcode modifier. These commands are defined in /usr/include/sundev/ipi3.h. Block numbers are not remapped by the partition map when these commands are used. The supported commands are:

IP_READ

IP_WRITE

Read or write data. The addressing is always by logical block (ignoring [a-h] logical partition information); the Opcode modifier is ignored.

IP_READ_DEFLIST

IP_WRITE_DEFLIST

Read or write one of the defect lists. The defect list is selected by the Opcode modifier in bits <15:8> of the dkc_cmd.

IP_FORMAT

Format a range of cylinders. For this command, the block number and sector count fields must both be a multiple of the number of blocks per cylinder. The dk_buflen field must be zero for this command.

IP_REALLOC

Reallocate a block. The controller attempts to recover the data from the old block being reallocated. If the old data cannot be recovered, a conditional success status is presented and a message may be printed. The dk_buflen field must be zero for this command.

DISK SUPPORT

This driver handles all supported IPI drives by reading controller attributes and a label from sector 0 of the drive which describes the disk geometry and partitioning.

The id000a partition is normally used for the root file system on a disk, the id0??b partition as a paging area, and the id???c partition for pack-pack copying (it normally maps the entire disk). The id???g partition is generally a partition suitable for holding the /usr file system. The rest of the disk is normally the id???h partition.

FILES

/dev/id[0-7][0-7][0-7][a-h]

block files

/dev/rid[0-7][0-7][0-7][a-h]

raw files

SEE ALSO

DIAGNOSTICS

The following messages are usually preceded by either idcn, indicating a message associated with controller n, or idcsf, indicating a message associated with the drive csf, where csf is the drive's channel, slave, and facility address.

idcn: configured with bad IPI address addr

The IPI address of the controller should end with `ff'.

idcn: unknown ctlr vendor id: manuf 'name' model 'model'

The controller type is unsupported. The driver has not been tested with the controller being used.

idcn: unknown string ctlr type

The controller type is unsupported. The driver has not been tested with the controller being used.

idcsf: out of memory for label info

The disk could not be initialized because the driver could not allocate kernel memory for the label.

idcsf: warning: starting block address nonzero: addr

The disk attributes indicate a non-zero first block number. None of the supported controllers do this.

idcsf: zero blocks per track

The disk attributes indicate zero blocks per track. None of the supported controllers do this.

idcsf: inconsistent geometry: blk/cyl num blk/trk num head num

The disk attributes indicate different number of blocks per cylinder than would be expected from the blocks per track and the number of tracks (heads).

idcsf: inconsistent geometry: tot blks num blks/cyl num pcyl num

The disk attributes indicate different number of total blocks than would be expected from the blocks per cylinder and the number of cylinders.

idcsf: invalid logical block size num

The disk logical block size, given by the logical block size attribute parameter, is not a power of two less than 2**31.

idintr: bad rupt NULL q_dev

The interrupt routine was called with a request that cannot be identified.

idcn: ctlr message: 'message-text'

The firmware in the disk controller issued the message message-text. The controller manual should describe the message.

idcn: ctlr failure fru num message: 'message-text'

The firmware in the disk controller issued the failure message message-text, and indicates a likely field-replaceable unit number (printed in hex). The controller manual should describe the message and FRU number.

idcsf: corrupt label

The label checksum mismatched.

idcsf: <ASCII-label>

The label contains an ASCII string which is printed after the label is read. The ASCII string should give the disk type and number of cylinders, heads, and sectors.

idcsf: warning: label doesn't match IPI attribute info

idcsf: label ncyl num pcyl num acyl num head num sect num rpm num

idcsf: geom ncyl num pcyl num acyl num head num sect num rpm num

Some of the geometry information from the disk label does not match the geometry information from the controller attributes. This may or may not be a problem, but indicates a non-standard label.

idcsf: id_rdwr: block num past end of disk

An ioctl() read or write was issued to a block in a cylinder with a number greater than or equal to the number of physical cylinders indicated by the disk geometry from the label or attributes.

idcsf: id_format failed. result num

A format operation failed. The code number printed, num, is defined in /usr/include/sundev/ipi_driver.h.

idcsf: id_rdwr_deflist failed. result num

A read or write defect list operation failed. The code number printed, num, is defined in /usr/include/sundev/ipi_driver.h.

idcsf: id_realloc failed. result num

A reallocation operation failed. The code number printed, num, is defined in /usr/include/sundev/ipi_driver.h.

idcsf[a-h]: opcode block rel-block (block abs) idintr: bad rupt NULL qdev

An interrupt occurred for which the associated disk could not be located.

idcsf[a-h]: opcode block rel-block (block abs) idintr: neg cmd_q_len

The driver's count of commands outstanding to the drive became less than zero. System operation should continue normally, but this message indicates a minor driver problem.

idcsf[a-h]: opcode block rel-block (block abs) poll timeout

A command hung or took longer than the driver expected.

idcsf[a-h]: opcode block rel-block (block abs) id_cmd_intr: ctlr cmd_q_len negative

The driver's count of commands outstanding to the controller became less than zero. System operation should continue normally, but this message indicates a minor driver problem.

idcsf[a-h]: opcode block rel-block (block abs) id_error: slave IML not supported

An error response from the controller indicates that it requires an initial microcode load. If this message occurs, controllers which issue this error are not supported.

idcsf[a-h]: opcode block rel-block (block abs) id_error: slave reset not supported

An error response from the controller indicates that it requires a reset. If this message occurs, controllers which issue this error are not supported.

idcsf[a-h]: opcode block rel-block (block abs) id_error: full queue: not handled

An error response from the controller indicates that its internal command queue is full, and the command cannot be accepted. This message should not occur because the driver respects the queue limits determined from the controller attributes.

idcsf[a-h]: opcode block rel-block (block abs) id_error: log read not supported

The controller presented status indicating that its internal log has overflowed and should be read by the host. If this message occurs, controllers which issue this status are not supported.

idcsf[a-h]: opcode block rel-block (block abs) id_error: aborted command not supported

An interrupt indicated that the associated command had been terminated by abort. If this message occurs, the driver does not issue aborts or expect this status.

idcsf[a-h]: opcode block rel-block (block abs) id_error_parse: no response where expected

An interrupt occurred which should have been accompanied by a response packet, but none was found.

idcsf[a-h]: opcode block rel-block (block abs) missing interrupt - attempting recovery

The command took much longer than expected, indicating a possible problem with the controller or drive. Recovery of the controller will be started, and the command will be retried if possible.

idcsf[a-h]: opcode block rel-block (block abs) missing interrupt - recovery in progress

The command took much longer than expected, indicating a possible problem with the controller or drive. Since recovery was already in progress, the command was added to the list or commands to be retried after recovery.

idcsf[a-h]: opcode block rel-block (block abs) no memory for label

The driver could not obtain a temporary memory area for the disk label.

idcsf[a-h]: opcode block rel-block (block abs) cannot abort yet

The recovery code cannot abort specific commands. It will assume the command was cleared by a reset and simply reissue the command.

idcsf: id_recover state-number state-name

idcnum: id_recover state-number state-name

idcsf: id_recover_intr state-number state-name result

idcnum: id_recover_intr state-number state-name result

These messages trace the recovery action during missing interrupt handling and controller recovery. These are for diagnosing problems in the event that recovery is unsuccessful.

idcnum: synchronous command setup failed

During recovery or initialization, the setup for a command failed, probably due to an internal error or lack of free memory.

idcsf[a-h]: write check disabled

Write checking for the partition has been disabled via the s-1IDKIOCWCHK ioctl() function. See dkio.4s and dkctl.8 This message is issued whether or not write checking was previously disabled for that partition.

idcsf[a-h]: write check enabled

Write checking for the partition has been enabled via the s-1IDKIOCWCHK ioctl() function, possibly by dkctl.8 See dkio.4s and dkctl.8 This message is issued whether or not write checking was previously enabled for that partition. Write checking, when enabled, causes the driver to perform a read after every write to the partition to verify that the write succeeded. If the read fails, the write will be retried.

IPI DIAGNOSTICS

The following diagnostics come directly from controller status. The messages may occur in combinations determined by the controller. In most cases, more information about the conditions under which the status is issued will be found in the controller manual.

In the examples shown, the message for the response parameter is shown first, then the message for a single bit in the response parameter. In practice, there will be a response parameter message followed by one or more messages for the individual bits, all on the same line.

The messages will be preceded by the disk or controller number. The message descriptions refer to the appropriate disk or controller as the addressee, since it was addressed in the command or response.

Normally, these messages are not printed when the condition described arises. They will only be printed if the condition is ``interesting.'' Certain conditions print all of the messages associated with a response, even though the individual messages might not be especially interesting.

The description of the message meaning is excerpted from the IPI-3 standards document. Refer to it for a more complete explanation.

Message/Microcode Exception. Message.

The slave has included a message within Extended Substatus for the master.

Message/Microcode Exception. Failure Message.

The slave encountered a failure condition which resulted in the identification of a FRU (Field Replaceable Unit).

Message/Microcode Exception. Port Disable Pending.

The addressee has received a manual or programmed Port Disable command that will take effect when the Disable conditions are met.

Message/Microcode Exception. Port Response.

A port has executed a Port Response command.

Intervention Required. Not P-Available.

The selected addressee is not powered on or is not installed.

Intervention Required. Not Ready.

The selected addressee cannot execute its intended functions. The addressee's Not Ready condition may be cleared by operator intervention.

Intervention Required. Not P-Available Transition.

This is presented by the controller to advise the driver that a facility has become Not P-Available since the time that a command addressed to it was accepted. Note: if the transition had occurred before the command packet was accepted, the status would have been Not P-Available.

Intervention Required. Not Ready Transition.

This is presented by the controller to advise the driver that a facility has dropped ready since the time that a command addressed to it was accepted. Note: if the transition had occurred before the command packet was accepted, the status would have been Not Ready.

Intervention Required. Attribute Table may be corrupted.

The controller has encountered a condition under which it is possible that the Attributes table has been corrupted, and it is not prepared to continue operation without master intervention.

Intervention Required. Busy.

The command cannot be executed because the addressee has been Busy for a time determined by the Facility Timeout Value specified in Attributes.

Alternate Port Exception. Priority Reserve Issued.

The addressee has been instructed to release allegiance to this port because of a Priority Reserve from an alternate port.

Alternate Port Exception. Attributes Updated.

An Attributes command has been issued from an alternate port which has changed the addressee's attributes.

Alternate Port Exception. Initialization Completed.

The addressee has completed an initialization procedure which may have affected this port, and was originated by a Reset from an alternate port.

Alternate Port Exception. Format Completed.

The addressee has completed a FORMAT command from an alternate port.

Alternate Port Exception. Facility switched to other port.

The slave has determined that the facility is switched to another port.

Machine Exception. No longer busy.

The addressee is notifying the master that it is no longer busy.

Machine Exception. P-Available.

This is presented asynchronously by the slave to advise the master that a facility which was previously Not P-Available has become P-Available.

Machine Exception. Ready.

This is presented asynchronously by the slave to advise the master that a facility which was not previously ready has become Ready.

Machine Exception. Operation Timeout.

There has been a failure condition in the addressee which has been detected by an internal timeout mechanism.

Machine Exception. Physical Interface Check.

The slave detected a check condition on the Physical Interface -- for example, an invalid sequence generated by the "state machine" or parity error on the bus(es).

Machine Exception. Slave Initiated Reset.

An internal condition caused the slave to initiate a reset; the master shall assume all outstanding commands and buffer contents are either lost or suspect.

Machine Exception. Environmental Error.

Some condition internal or external to the addressee has been detected which may cause failure condition(s) -- for example, a temperature sensor alert.

Machine Exception. Power Fail Alert.

The addressee has detected an impending power failure condition.

Machine Exception. Data Check (on Raw Data).

The master has requested raw data and the addressee has detected a data error.

Machine Exception. Uncorrectable Data Check.

The slave detected a data error which has persisted after the slave has exhausted any possible recovery actions. On write operations, the malfunction may have caused invalid data to be recorded.

Machine Exception. Fatal Error.

The addressee detected an internal machine error that precludes execution or continuation of the current command.

Machine Exception. Hardware Write Protected.

An attempt was made to write on a drive that was protected against writing by something physical -- for example, a switch on the drive.

Machine Exception. Queue Full.

The command queue for the addressee is full.

Machine Exception. Command Failure.

The command in execution encountered a condition which caused it to complete correctly but unsuccessfully; for example, a COMPARE of two files detected a discrepancy.

Machine Exception. Read Access Violation.

An attempt was made to read on an addressee which had been protected using Access Permits.

Machine Exception. Write Access Violation.

An attempt was made to write on an addressee which had been protected using Access Permits.

Machine Exception. Data Overrun.

This can occur during direct data transfer, or if the slave has a buffer which was not adequate, and the buffer overran during a read or a write operation.

Machine Exception. Reallocation space exhausted.

Space required for reallocation of data due to media defects is not available; that is, all space assigned for that purpose has been exhausted.

Machine Exception. Unexpected Master Action.

The slave has encountered an unexpected action by the master; for example, the Master Status at the Physical Interface does not correlate to the anticipated status, no status was expected and some was presented by the master, the master did not respond with Data I/O or Control of Bus following a Transfer Notification packet.

Machine Exception. Error Log Full.

The Error Log capacity has been exceeded.

Machine Exception. Defect Directory Full.

The Defect Directory capacity has been exceeded and no more blocks can be re-allocated.

Command Exception. Invalid Packet Length.

The packet length is invalid; for example, the length of the parameter list plus the basic packet does not equal the packet length.

Command Exception. Invalid Command Reference Number.

The Command Reference Number is invalid; for example, it duplicates one in a command that is currently active.

Command Exception. Invalid Slave Address.

The Slave Address in the command packet is invalid; for example, it does not match the selected slave's address.

Command Exception. Invalid Facility Address.

The Facility Address in the command packet is invalid.

Command Exception. Invalid Selection Address.

The facility selected at the Physical Interface does not match the facility address supplied in the command packet.

Command Exception. Invalid Opcode.

The command packet contained an invalid or unsupported Opcode.

Command Exception. Invalid Modifier.

The Modifier was invalid or is not supported for the Opcode specified.

Command Exception. Invalid Extent.

The Data Address (for example, the block number) plus the Count specified in an Extent parameter is not valid for the addressee.

Command Exception. Out of Context.

The slave has encountered a situation in which it cannot process the command because it considers it out of context -- for example, a RESUME command without a previous COPY.

Command Exception. Invalid Parameter.

One or more of the parameters in the command packet was invalid.

Command Exception. Missing Parameter.

One or more of the parameters required in the command is not present.

Command Exception. Reserved Value nonzero.

A reserved value defined by the standard does not contain zero.

Command Exception. Invalid Combination.

The addressee has detected that two valid but mutually exclusive options have been selected by the master.

Command Aborted. Command Aborted.

The command this response packet is related to was ABORTed by the master.

Command Aborted. Command Sequence Terminated.

Command Sequencing was terminated because this command failed to complete successfully.

Command Aborted. Unexecuted Command from Terminated Sequence.

The command related to this response packet was not executed but was terminated because a prior command which was sequenced to it failed to complete successfully.

Command Aborted. Command Chain Terminated.

Command Chaining was terminated because this command failed to complete successfully.

Command Aborted. Unexecuted Command from Terminated Chain.

The command related to this response packet was not executed but was terminated because a prior command which was chained to it failed to complete successfully.

Command Aborted. Command Order Terminated.

Command Ordering was terminated because this command failed to complete successfully.

Command Aborted. Unexecuted Command from Terminated Order.

The command related to this response packet was not executed but was terminated because a prior command which was ordered to it failed to complete successfully.

Conditional Success. Logging Data Appended.

The slave has appended information in this response which the slave is advising the master is relevant to be logged.

Conditional Success. Abort Received - no Command Active.

An ABORT command was issued to an addressee in the L-Available condition but the referenced command could not be found.

Conditional Success. Abort Received - Status Pending.

An ABORT command was issued to an addressee which has the response status for the referenced command pending; for example, the command has been completed.

Conditional Success. Abort Received - Not Operational.

An ABORT command was issued to a facility which is Not Operational.

Conditional Success. Anticipated Error.

The addressee has detected a condition which may result in future error conditions; for example, on disc seek retries were needed.

Conditional Success. Anticipated Data Error.

The addressee has detected a condition which may result in future data errors; for example, successive retries were needed for reading disk data.

Conditional Success. Re-allocation Required.

The addressee has detected a data error condition which requires reallocation action -- for example, an unrecoverable read error.

Conditional Success. Re-allocation Discontinuity.

The slave has automatically reallocated a block which contained a data error and the reallocated data is no longer in close proximity to the blocks previously contiguous to it.

Conditional Success. Defect Directory Threshold Exceeded.

The threshold within the addressee's Defect Directory has been exceeded, indicating that there is a limited number of entries remaining for adding more defects.

Conditional Success. Error Retry Performed.

The addressee has completed the command but error retry had to be invoked. Note: Error Retry does not include actions associated with data transfer.

Conditional Success. Data Retry Performed.

The addressee has completed the command but data retry had to be invoked -- for example, a physical re-read. Data Retry includes all actions associated with the transfer of data.

Conditional Success. Motion Retry Performed.

The addressee has completed the command but motion retry had to be invoked.

Conditional Success. Data Correction Performed.

The addressee has completed the command but data correction had to be applied.

Conditional Success. Soft Error.

The slave detected an internal machine error that did not preclude execution or continuation of the current command.

Conditional Success. Release of Unreserved Addressee.

The addressee has received a release command for which there is no reservation.

Conditional Success. Request Diagnostic Control Command.

As a result of executing a diagnostic command which provided more information than can be returned by a Response, the addressee is requesting that the master issue a Diagnostic Read Command.

Conditional Success. Error Log Request.

The master is requested to capture the contents of the Error Log (which contains manufacturer dependent information) because the threshold has been exceeded.

Conditional Success. Statistics Update Requested.

There has been a change in meaningful statistics during the execution of this command, and the master is requested to update its Statistics Table (if any).

Conditional Success. Parameter Update Requested.

There has been a change in meaningful device parameters during the execution of this command, and the master is requested to update its Statistics Table (if any).

Conditional Success. Asynchronous Event Occurrence.

An asynchronous event has occurred which may be described further in Extended Status.

Conditional Success. Master Terminated Transfer.

The previous Information Transfer which had a Master Termination Parameter, was terminated by the master.

Incomplete. Command May be Resumed.

This status is used to advise the master that an otherwise successful command did not complete. The incomplete command remains on the slave's queue, and its Command Reference Number shall remain valid, until the command is resumed or aborted.

Incomplete. Response Packet Truncated.

The maximum Information Transfer Length specified by the attributes was exceeded by the response packet, which was truncated at that size, and the response is considered complete by the slave.

Incomplete. Data Length Difference.

The addressee has not transferred all the information specified in a transfer command.

Created by unroff & hp-tools. © somebody (See intro for details). All Rights Reserved. Last modified 11/5/97