scsiha(1m) scsiha(1m)
scsiha - probe and control scsi buses
scsiha [-lprtw] [-D debuglevel] [-L targetID|all] {bus_number | device} ...
scsiha is used to perform operations on fibrechannel loops of SCSI
devices and parallel SCSI buses.
The following options apply to both parallel SCSI and fibrechannel SCSI:
-p Probe for devices on the bus or loop. On parallel SCSI buses, this
option causes the driver to look for devices at every ID on the SCSI
bus. On fibrechannel SCSI loops, this option causes the driver to
look for devices at every ID that was reported in use at the most
recent loop initialization. On parallel SCSI buses, when a device
is found at lun 0, all other luns are then probed. When devices or
luns are not valid, a warning message may be printed on the console.
With fibrechannel, the Report Luns SCSI command is used to determine
which LUNs are valid. If the Report Luns command returns error,
only LUN 0 is assumed to be valid.
After probing, new devices will have a controller number of "-1"
until ioconfig -f /hw is run to update the hardware graph.
-r Perform a SCSI bus reset and/or reset host adapter.
This affects all devices on that SCSI bus, causing current commands
to be aborted, and specially set parameters to be reset, in many
cases. Tape drive state will be reset after this, and any tape
operations in progress will be aborted.
This option should therefore be used with some caution. In very
rare cases, it may cause a device to become unusable without a
powercycle. There are times when it is desired to reset devices or
adapters that appear to be hung, however, so this function is
provided. Some host adapter drivers will reset the adapter as part
of this ioctl. Fibrechannel adapters will perform loop
initialization, making this operation similar to the -l option,
except that the adapter will also be reset.
-D debuglevel
Set the debug level in the host adapter driver. Where supported,
this option changes the verbosity of a given host adapter driver.
In general, 0 is least verbose, with no specific upward limit.
The following option only applies to some parallel SCSI adapters:
-t Return interesting SCSI bus and attached target parameters. Bus
parameters include whether the bus is single-ended or differential,
the SCSI host ID and the selection timeout. The target parameters
Page 1
scsiha(1m) scsiha(1m)
include whether the device is capable of operating in narrow or wide
mode, whether it's capable of doing synchronous transfer and if so,
what the period and offset is.
The following options only apply to Fibrechannel SCSI adapters:
-l Perform loop initialization. During loop initialization, all
devices on the loop will attempt to acquire a target ID. SGI
devices will attempt to get the ID they have been selected for,
either by settings on and position within an enclosure, or by
dials/switches on the device. If multiple devices have the same ID
selection, only one device will get that ID. Other devices will
attempt to acquire IDs not in already in use.
This option may have a side effect of aborting current commands
outstanding on the adapter, so it should be used with care. The
disk driver will typically reissue the aborted commands, but there
will be a performance impact while commands are reissued.
-L targetID
Send LIPRST to the specified device and then perform loop
initialization. This option is much like -l above, except that it
also requests that the device resets itself. Not all devices will
honor the request. RAID controllers typically don't while disk
drives typically do. This option can be used in rare cases where a
device is not functioning properly, but is still able to complete
loop initialization and acquire a target ID. For example, if a
drive gets timeouts on attempts to probe (using the -p option
above), then this option may bring it back to working condition. In
this respect, the -L option can be used for reasons similar to -r on
parallel SCSI buses.
-L all
This option is like the -L option directly above, except that it
directs all devices to reset themselves. Only devices that
implement FC-AL-2 or later will respond to this. This option can be
used if a loop is malfunctioning, and it is suspected that one or
more devices may be causing the malfunction, but it isn't known
which. RAID controllers typically don't honor the "self-reset"
function of the LIPRST
The -L options may have a side effect of aborting current commands
outstanding on the adapter, and it if the device honors the reset
function, outstanding commands to it will be dropped. The disk
driver will typically reissue the aborted commands, but there will
be a performance impact while commands are reissued and while a
device resets.
Page 2
scsiha(1m) scsiha(1m)
-w This option will print the portname (sometimes known as the World
Wide Name) of the fibrechannel adapter.
To reset parallel scsi bus 2, use one of these commands:
scsiha -r 2
scsiha -r /hw/scsi_ctlr/2/bus
To probe scsi bus 0, use one of these commands:
scsiha -p 0
scsiha -p /hw/scsi_ctlr/0/bus
To reset target 8 and then perform loop initialization on fibrechannel
bus 6, do the following:
scsiha -L 8 6
scsiha -L 8 /hw/scsi_ctlr/6/bus
/hw/scsi_ctlr/*/bus the scsi bus operation devices
ioconfig(1M), scsicontrol(1M), scsiha(7M).
This command is normally usable only by the superuser, because the
standard permissions of the devices in /hw/scsi_ctlr/* restrict access to
other users.
With Fibrechannel adapters, if a loop is broken after a system has been
operational, numerous attempts will be made to reinitialize the loop,
after which the host adapter driver will give up on the loop. At this
point, if the hardware condition causing the broken loop is fixed (box of
drives powered back on, cable reinserted, etc.), the -r option should be
used to re-establish loop operation.
The source to this program is also shipped as sample source code in the
irix_dev.gifts.scsi subsystem as part of IDO.
The Performance Co-Pilot (see PCP(1)) may be used to monitor SCSI
devices, however configuration changes by scsiha may not be
asynchronously detected. If SCSI devices are being monitored,
libirixpmda(5) may be notified and forced to reset using the command:
> pmstore irix.xbow.switch 1
PPPPaaaaggggeeee 3333 [ Back ]
|
