On Tue, Dec 21, 2010 at 10:40:48AM +0100, Filip Aben wrote: > In case you have USB selective suspend enabled, can you disable it and > run your test again ? I disabled it (power/control set to 'on' for all modems) and made some other tests. After digging deeper in logs I noticed that there is more XactErrs, but most of them recover in < 15 tries (does not depend on suspend setting). As far as I can check this, it doesn't depend on external environmental conditions. For all the modems I run the software, that, in test conditions, mostly does repetitive polling of the modem (i.e. issues AT+CREG?, then AT+CMGL=4, then 15 seconds sleep). When this software is run for 2 modems there are no XactErrs at all. If I run it for 16 modems I get XactErrs. I don't know the cause. Maybe it's bad USB controller (Intel 82801G), but I tested on NEC controller before, and it wasn't perfect either. I increased the number of XactErr retries and if it would recover automatically I could probably live with that, but I'm afraid that hso crashes after reattachment (one visible at the end of dmesg I sent previously) can leave system in unstable state.
Attachment:
pgpw1LxOIc1Vh.pgp
Description: PGP signature
