Date: Tue, 12 Sep 2006 08:46:30 -0700 From: "Jack Vogel" <jfvogel@gmail.com> To: "Andre Oppermann" <andre@freebsd.org> Cc: Ian FREISLICH <if@hetzner.co.za>, current@freebsd.org Subject: Re: TSO, SMP and the em driver. Message-ID: <2a41acea0609120846i32fd682wd834e0373c23d557@mail.gmail.com> In-Reply-To: <4506BE2C.1050903@freebsd.org> References: <E1GN5FA-00085C-CJ@hetzner.co.za> <4506BE2C.1050903@freebsd.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On 9/12/06, Andre Oppermann <andre@freebsd.org> wrote: > Ian FREISLICH wrote: > > Hi > > > > I've noticed that after an apparently random period of time something > > goes wierd with my networking. A tcpdump shows traffic is recieved, > > but trussing a daemon (sshd for instance) shows the packets are not > > making it that far. Also, no packets are transmitted either. > > > > Here's an ifconfig of the interface while it doesn't transmit or > > recieve packets, note the OACTIVE: > > OACTIVE means the interface is stuck. This looks like a bug in the > em(4) driver. I've copied Jack Vogel, the driver maintainer from Intel, > into this email. > > -- > Andre > > > em0: flags=8c43<UP,BROADCAST,RUNNING,OACTIVE,SIMPLEX,MULTICAST> mtu 1500 > > options=3cb<RXCSUM,TXCSUM,VLAN_MTU,POLLING,VLAN_HWCSUM,TSO4,TSO6> > > inet 41.204.193.40 netmask 0xfffffff0 broadcast 41.204.193.47 > > ether 00:04:23:d4:12:2e > > media: Ethernet autoselect (100baseTX <full-duplex>) > > status: active > > > > If I disable and re-enable TSO, the interface cames back, but > > probably because of an interface reset. It comes back without the > > OACTIVE flag and then all works well for some time. > > > > Here's the pciconf output for the ethernet device. > > > > em0@pci3:4:0: class=0x020000 card=0x10798086 chip=0x10798086 rev=0x03 hdr=0x00 > > vendor = 'Intel Corporation' > > device = '82546EB Dual Port Gigabit Ethernet Controller' > > class = network > > subclass = ethernet > > > > So far, it hasn't happened again with TSO disabled over a period > > where it happened twice. Let me know if there's any debugging I > > can do for you. Interestingly, I have 4 more of these dual NICs > > running on the same version of the source without ony problems, but > > the working cards are on UP machines. Hmm, so if memory serves Linux disallows TSO when you are only at 100Mb, I dont recall why, I will have to look into that and perhaps have that same limitation in our driver. Any other data you can give me? Cheers, Jack
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?2a41acea0609120846i32fd682wd834e0373c23d557>