Hello all, I have a mini-itx machine that acts as a router for a small office with a couple of servers behind it, and it runs openbsd.
I previously had issues with this sf card, and while I'm very tempted to replace it, I'm trying to hold off for at least the summer, when I'm going to be able to replace the entire thing with something newer/faster. The sf0: device timeout errors were gone for a while, but have returned since upgrading the machine to 4.4 just recently. However, with that said, a new error has creeped up, usually around the same time as the "device timeout" errors, and so while I have no proof to say they're related, it's a fair assumption. When the error noted below occurs, all new connections are refused, and only active ones keep going, and this on all of the four interfaces (sf0-3), even though it's always sf0 that has the timeout error. This has begun since upgrading to 4.4, and the machine is running exactly the same services (samba as a pdc, pf, dns, dhcp, nothing else) as before the upgrade. By "upgrade" I mean wiping the disk of 4.3, and doing a fresh 4.4 install. I did not simply dump the backed up /etc into the new installation, in fact, so far, all I've done is put back dhcpd.conf (with changes reflecting removal of dhcp.interfaces), my pf.conf, /etc/samba/smb.conf and the dns stuff in /var/named. Oh, and I placed the old rc.conf as rc.conf.local. Jan 30 17:58:09 cerberus /bsd: WARNING: mclpool limit reached; increase kern.maxclusters Jan 30 18:43:43 cerberus /bsd: WARNING: mclpool limit reached; increase kern.maxclusters Jan 30 18:43:48 cerberus /bsd: sf0: device timeout Upping the kern.maxclusters using sysctl temporarily solves the problem, it's unclear yet whether it'll just max out again though later on. However, I've run openbsd on this hardware from 3.8 and this is the first time it complains about that sysctl setting (which from what I can see has been 6144 all along). Here is the dmesg for my machine. I'm hoping that at the very least, someone can point me in the right direction for determining what to try and log more thoroughly to hopefully get more answers. Thanks in advance for any help provided, Sandro
