[Nut-upsuser] nut and eaton 3s over usb, lot of logs

[Arnaud Quette]

Hi,

2011/12/15 bonecollector <bonecollector at hot.ee>

>  Hi,
>
>
>   Dec 12 13:49:31 my-server klogd: [11382.642584] uhci_hcd 0000:00:1d.1:
>> reserve dev 2 ep81-INT, period 16, phase 8, 118 us
>> Dec 12 13:49:31 my-server klogd: [11382.894706] uhci_hcd 0000:00:1d.1:
>> release dev 2 ep81-INT, period 16, phase 8, 118 us
>> ...
>>
> these messages from the kernel, related to USB interrupt (possibly
> timeout):
> we get data from the devices through explicit requests (polling) and
> interrupt. Interrupts are notifications generated by the device, upon data
> change, without requiring any explicit request.
>
> there is nothing we can do in NUT. The only thing for you would be to
> disable interrupt, by adding "pollonly" to your ups section in ups.conf:
>
> [Eaton3s]
>    driver = usbhid-ups
>    port = auto
>    pollonly
>    desc = "Serveri UPS"
>
> the should remove the above log message (if it's really related to
> timeout), but will decrease reactivity WRT status changes.
> Another option may be to look at the kernel logging level.
>
>
> Thank you! Adding "pollonly" helped, and no more messages in logs.
>
> Little question still stays - even if exclusive access is used, why so
> many messages (there are lots of USB devices around nowdays, don't know
> about exclusive access used to connect to them, but I have never seen such
> logging before.).
>

there is only the first one that was related to detaching the kernel driver
to the profit of NUT driver.
Others were related to timeout on the interrupt pipe.

IMHO, NUT is the only one that has:
- persistance as a daemon,
- possible support in kernel (USB/HID UPS only), that needs to be removed
before NUT can talk to the device.
Note that this more a switch of driver (from hiddev to usbfs), but the end
result for the kernel is the same: log detachment.

Other system that are somehow similar are either pure kernel or pure
userland, or without persistance (printer, scanner, webcams).

A final note on Unix kernel drivers: RW support always requires exclusive
access.
Multipoint drivers are userland, based on the kernel drivers, like NUT
drivers (which are in reality daemons).

cheers
Arnaud
-- 
Linux / Unix Expert R&D - Eaton - http://powerquality.eaton.com
Network UPS Tools (NUT) Project Leader - http://www.networkupstools.org/
Debian Developer - http://www.debian.org
Free Software Developer - http://arnaud.quette.free.fr/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.alioth.debian.org/pipermail/nut-upsuser/attachments/20111215/5450a57f/attachment.html>