We have a fairly simple filter driver (at least for now), however there
are some strange things happening. This filter driver matches on a
device with a proper volume label, and then creates an IOMedia device
with a special "dev-name" property, so that we get a custom /dev/ entry.
This works well enough on the initial appearance of the devices. It also
correctly goes away when the device disappears. The problem occurs when
the device disappears for a period of time, and then reappears. We see
erratic behaviour, sometimes all the devices reappear, and sometimes all
from one controller only, and mostly, none reappear.
Even if I completely remove the filter driver from the equation, I see
the same behaviour.
Some info:
Mac OS 10.4.8 (however I have seen this on 10.4.9 and also eariler
10.4.x... I haven't updated to 10.4.10 yet as there is no source available)
Apple HBA
Testing methodology:
1. Boot system, check /dev/rdiskN entries, run fibreconfig -l to confirm
we can see the LUNs
2. Fence ports via switch to simulate a connectivity failure (eg cable pull)
3. Wait for diskarb to notice it
4. check /dev/rdiskN and run firbeconfig -l to confirm LUNs have disappeared
5. Un fence ports (eg cable plugged back in)
6. Wait
7. check /dev/rdiskN and run firbeconfig -l to confirm LUNs have reappeared
It looks like this might be a bug with the HBA driver? Is there anything
I can do to ensure that the devices reappearing are noticed? Running
fibreconfig -r doesn't help (although if the device appears for the
first time, it does help).
Any help would be appreciated,
thanks
Nick
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Darwin-kernel mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
http://lists.apple.com/mailman/options/darwin-kernel/email@hidden
