We maintain our own patches here as well to get around this problem. IMHO, it is far better to have some applications "waste" a meg or two of buffer space than to hamstring any high-performance bpf app that runs on a FreeBSD box.
This is most likely a trivial code fix -- how hard would it be to get
this committed?
--eli
In reply to "Guy Helmer" <[EMAIL PROTECTED]> :
> Petri Helenius wrote:
> > Guy Helmer wrote:
> > >I use "sysctl debug.dbf_bufsize=131072" on my appliances to increase the
> > >BPF buffer size to something more reasonable without having to directly
> > >modify libpcap.
> > >
> > Hope you're not disappointed to find out that modifying that parameter has
> > no effect when using applications which use libpcap since libpcap always
> > sets the buffer size to 32768. (which is exactly the problem I'm
> complaining about)
>
> You are right - I misremembered how the BIOCSBLEN ioctl worked. My
> appliances do have a private copy of libpcap with a larger buffer size
> because of this problem.
>
> IMHO, it would be better for the libpcap code to query the default BPF
> buffer size (BIOCGLEN) and use it if it is larger than the libpcap default
> size (32768). Then libpcap would obey the buffer size set by the sysctl.
>
> Guy
>
>
> To Unsubscribe: send mail to [EMAIL PROTECTED]
> with "unsubscribe freebsd-net" in the body of the message
msg07872/pgp00000.pgp
Description: PGP signature
