Like other hardware, USB devices are interfaced to the operating system by OS drivers, and often there are several suitable drivers with different capabilities. In Solaris and related systems, this mapping is detailed in the /etc/driver_aliases file and properly managed by dedicated tools. By default, USB devices can be captured by the generic USB HID driver, or none at all; however an "UGEN" driver can behave better with the libusb library used on Solaris.

Connect the power device using its USB port to your computer.

Run prtconf -v | less to see the details of device connections, and search for its probable strings (vendor, model, serial number).

Two examples follow:

You can also check with cfgadm if the device is at least somehow visible (if not, there can be hardware issues in play). For example, if there is a physical link but no recognized driver was attached, the device would show up as "unconfigured":

:; cfgadm | grep usb-
usb8/1                 usb-input    connected    unconfigured ok

If you conclude that a change is needed, you would need to unload the existing copy of the "ugen" driver and set it up to handle the device patterns that you find in compatible values from prtconf. For example, to monitor the devices from listings above, you would run:

:; rem_drv ugen
:; add_drv -i '"usb463,ffff.100"' -m '* 0666 root sys' ugen

or

:; rem_drv ugen
:; add_drv -i '"usb665,5161.2"' -m '* 0666 root sys' ugen

Note that there are many patterns in the compatible line which allow for narrower or wider catchment. It is recommended to match with the narrowest fit, to avoid potential conflict with other devices from same vendor (especially if the declared identifiers are for a generic USB chipset).

Also note that the add_drv definition above lists the POSIX access metadata for the device node files that would be generated when the device is plugged in and detected. In the examples above, it would be owned by root:sys but accessible for reads and writes (0666) to anyone on the system. On shared systems you may want to constrain this access to the account that the NUT driver would run as.

After proper driver binding, cfgadm should expose the details:

# cfgadm -lv
...
usb8/1                         connected    configured   ok
    Mfg: EATON  Product: Eaton 9PX  NConfigs: 1  Config: 0  <no cfg str descr>
    unavailable  usb-input    n        /devices/pci@0,0/pci103c,1309@1d,2:1
...

Usually the driver mapping should set up the "friendly" device nodes under /dev/ tree as well (symlinks to real entries in /devices/) so for NUT drivers you would specify a port=/dev/usb/463.ffff/0 for your new driver section in ups.conf.

For some serial-to-USB converter chips however it was noted that while the device driver is attached, and the /device/... path is exposed in the dmesg output (saved to /var/adm/messages) the /dev/... symlinks are not created. In this case you can pass the low-level name of the character-device node as the "port" option, e.g.:

./mge-shut -s 9px-ser -DDDDD -d2 -u root \
    -x port=/devices/pci@0,0/pci103c,1309@1a,2/device@1:0

Until NUT release 2.7.4 the only option to build NUT drivers for USB connectivity was to use libusb-0.1 or a distribution’s variant of it; the original Sun Solaris releases and later related systems provided their customized version for example (packaged originally as SUNWlibusbugen, SUNWugen{,u} and SUNWusb{,s,u,vc}).

However, libusb-0.1 consuming programs had some stability issues reported when running with long-term connections to devices (such as an UPS), especially when using USB hubs and chips where hardware vendors had cut a few corners too many, which were addressed in a newer rewrite of the library as libusb-1.0.

Subsequently just as at least the illumos-based distributions evolved to include the new library and certain patches for it, and the library itself matured, the NUT project also added an ability to build with libusb-1.0 either directly or using its 0.1-compat API (available since NUT 2.8.0 release).

If your "standard" build of NUT has problems connecting to your USB UPS (libusb binary variant should be visible in driver debug messages), consider re-building NUT from source with the other option using the recent library build as available for your distribution.

In this context, note the OpenIndiana libusb-1 package pull requests with code which was successfully used when developing this documentation:

Binaries from builds made in OpenIndiana using the recipe from PR #5382 above were successfully directly used on contemporary OmniOS CE as well.

So… your setup worked nicely, and then one day you see the console flooded with messages like the following:

Broadcast Message from nut (???) on n54l Mon May  9 12:05:59...
Communications with UPS innotech@localhost lost

Broadcast Message from nut (???) on n54l Mon May  9 12:10:55...
UPS innotech@localhost is unavailable

Unfortunately, some devices "get stuck" on USB level (whether in the chips, in the OS driver layer, libusb or NUT driver) and their NUT drivers have to be restarted to regain monitoring, as opposed to intermittent losses of connectivity that software recovers from automatically.

As in all systems, you should stop all programs using the connection, including NUT driver instances that might have been started beside the wrapping service (SMF). It may be possible to just start the new driver instance at this point, but if it still does not see the device — you have to re-initialize the connection on the OS level.

As a symptom, attempts to start the NUT driver with elevated debug verbosity would not even see the device details:

   0.000606     [D1] Saving PID 5187 into /var/run/nut/nutdrv_qx-innotech.pid
   0.000727     [D1] upsdrv_initups...
   0.012065     [D2] Checking device 1 of 2 (0665/5161)
   0.012303     [D1] Failed to open device (0665/5161), skipping: Other error
   0.012394     [D2] Checking device 2 of 2 (099A/610A)
...
   0.020364     [D2] Trying to match device
   0.020586     [D3] match_function_regex: matching a device...
   0.020839     [D2] match_function_regex: failed match of VendorID:  99a
   0.021061     [D2] Device does not match - skipping
   0.021371     [D2] libusb1: No appropriate HID device found
Network UPS Tools - Generic Q* USB/Serial driver 0.32 (2.8.0-20-g535395363)
USB communication driver (libusb 1.0) 0.43
   0.021720     libusb1: Could not open any HID devices: insufficient permissions on everything
   0.021821     No supported devices found. Please check your device availability with 'lsusb'
and make sure you have an up-to-date version of NUT. If this does not help,
try running the driver with at least 'subdriver', 'vendorid' and 'productid'
options specified. Please refer to the man page for details about these options
(man 8 nutdrv_qx).

Driver failed to start (exit status=1)
Network UPS Tools - UPS driver controller 2.8.0-20-g535395363
[ May  9 03:10:01 Method "start" exited with status 1. ]

:; svcadm disable -ts nut-driver:innotech

:; ps -ef | grep -Ei 'nut|ups' ; svcs -p innotech
    root 10522     1   0   May 06 ?           0:00 /usr/sbin/upsmon
    root 16927     1   0   Feb 25 ?           1:20 /usr/lib/nut/bin/nutdrv_qx -a innotech
     nut 10257     1   0   May 06 ?           0:39 /usr/sbin/upsd
    root 16985 15379   0 11:27:36 pts/1       0:00 grep -Ei nut|ups
     nut 10524 10522   0   May 06 ?           0:25 /usr/sbin/upsmon
STATE          STIME    FMRI
offline         11:26:49 svc:/system/power/nut-driver:innotech

# In the ps listing above, a driver daemon is seen that was started as
# the root user beside the actual service. It has to be stopped too:
:; kill -9 16927

:; cfgadm -lv usb10/1

Ap_Id                          Receptacle   Occupant     Condition
Information
When         Type         Busy     Phys_Id

usb10/1                        connected    configured   ok
Mfg: INNO TECH  Product: USB to Serial  NConfigs: 1  Config: 0  : 20100826
unavailable  usb-input    n        /devices/pci@0,0/pci103c,1609@13:1

:; cfgadm -c disconnect usb10/1

Disconnect the device: /devices/pci@0,0/pci103c,1609@13:1
This operation will suspend activity on the USB device
Continue (yes/no)? yes
cfgadm: Hardware specific failure: Cannot issue devctl
  to ap_id: /devices/pci@0,0/pci103c,1609@13:1

:; cfgadm -c unconfigure usb10/1
Unconfigure the device: /devices/pci@0,0/pci103c,1609@13:1
This operation will suspend activity on the USB device
Continue (yes/no)? yes

:; cfgadm -c disconnect usb10/1
Disconnect the device: /devices/pci@0,0/pci103c,1609@13:1
This operation will suspend activity on the USB device
Continue (yes/no)? yes

:; cfgadm -lv usb10/1
Ap_Id                          Receptacle   Occupant     Condition  Information
When         Type         Busy     Phys_Id

usb10/1                        disconnected unconfigured ok
unavailable  unknown      n        /devices/pci@0,0/pci103c,1609@13:1

:; cfgadm -c configure usb10/1
cfgadm: Hardware specific failure: Cannot issue devctl
  to ap_id: /devices/pci@0,0/pci103c,1609@13:1

# Despite the error above, the device is seen now:
:; cfgadm -lv usb10/1
Ap_Id                          Receptacle   Occupant     Condition
Information
When         Type         Busy     Phys_Id

usb10/1                        connected    configured   ok
Mfg: INNO TECH  Product: USB to Serial  NConfigs: 1  Config: 0  : 20100826
unavailable  usb-input    n        /devices/pci@0,0/pci103c,1609@13:1

# ... and the driver can start:
:; svcadm enable innotech

Broadcast Message from nut (???) on n54l Mon May  9 12:12:30...
Communications with UPS innotech@localhost established

0,5,10,15,20,25,30,35,40,45,50,55 * * * * MODE=optional /etc/nut/reset-ups-usb-solaris.sh