In message <A81E58D7-27EC-4D00-9C05-858E983935C8@avernus.com>, Geoff Adams writ
es:
>Is there any hope of getting this fix pulled up to the 2.0 branch? I  
>have applied the relevant part of the fix in sys/netinet/tcp_input.c  
>1.213 to the 2.0 branch, and this fixes the problem for me. (Before  
>applying the fix, my mail server would panic every day or so. This  
>server talks, naturally, to hosts around the world. Mostly spammers,  
>of course. And some of those hosts triggered the bug. Since applying  
>the fix nearly two months ago, I have had no panics.)
>
>I've attached my patch against the 2.0 branch. I certainly imagine  
>this must affect more folks than just me. Otherwise, I'd be happy to  
>just keep going with my locally modified sources.

I have turned it from a kassert into a diagnostic printf, and I hit it
pretty often; often five times in a day.

# dmesg | grep to_tick
to_ticks: -350
to_ticks: -350
to_ticks: -300
to_ticks: -300
to_ticks: -300
to_ticks: -300
to_ticks: -300
to_ticks: -250
to_ticks: -250
to_ticks: -250
to_ticks: -250
to_ticks: -250
to_ticks: -300
to_ticks: -300
to_ticks: -300
to_ticks: -250
to_ticks: -250
to_ticks: -250

-s