Date: Sat, 6 Sep 2003 22:25:49 +0100 From: Bruce M Simpson <bms@spc.org> To: Colin Watson <sb.mailinglist@lambdabroadband.com> Cc: freebsd-net@freebsd.org Subject: Re: Packet loss problem Message-ID: <20030906212549.GP1417@spc.org> In-Reply-To: <002201c3749d$c8cf4460$0b4e1151@blackbox> References: <002201c3749d$c8cf4460$0b4e1151@blackbox>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sat, Sep 06, 2003 at 06:39:12PM +0100, Colin Watson wrote: > / 33104 broadcast/multicast datagrams dropped due to no socket " which seems an inordinatly high amount. There are no drops due to full socket buffers, although I have recompiled the kernel with nmbclusters=8192 and Maxusers=1024 (to increase number of available sockets - kern.ipc.maxsockets to 9391), still the loss occurs. Any suggestions, and could someone explicitly explain what "broadcast/multicast datagrams dropped due to no socket" means. As Pete points out this unusually high 'no socket' figure could be due to packets being received for a service you don't run. Beefing up the number of PCBs you can have won't have any effect there if that is the case. Also it could be due to a socket application not binding to INADDR_ANY to pick up undirected broadcasts (they don't get delivered to interface bound ports). Of course this is probably not likely to affect your application if you're explicitly unicasting traffic. But if you intend to receive multicasts this could be the case. Are you filtering anything, or could your upstream be filtering anything? BMS
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20030906212549.GP1417>