On 5/18/2010 12:31 PM, Manuel Bouyer wrote:
Can you see what syn_cache_add+0x4ca points to in kernel sources ? Maybe try reverting tcp_input.c 1.304 and see if it helps ...
Reverting to tcp_input.c 1.303 seems to have solved my problem. ~5 hour max uptime with little USB activity has turned into 16 hours with significant umass activity (I did a dry-run of the system backup, which is what repeatedly killed the system before).
Thanks for the clue. I'm not cluefull enough to know WHY the changes in 1.304 are a problem, but at least now it's clear that it is the problem. :-)
ScottE