I run an x500, and I have massive issues with the Web Interface crashing the whole unit otherwise its fine like the guy above said.

Sorry, I posted this at bed time…

Ok, to expand on your post....
I've read the forum links and asked for assitance from the posters - thankyou for the link too... :)

I have a WG FBII that has been upgraded to a AMDk6-200 with 512Mg Ram. I have the proper cable for a CF card adpater and different sizes of CF cards, (1G & up). I've also got different sizes of laptop 2.5" HDD's. I've tried the embedded images going back as far as 1.2 all the way up to 1.2.3 nanobsd versions. I've written them all with the "physdiskwrite" application both through the command line and the gui. I've used a USB card reader to write the images, I've also used the CD installer and put the CF card into the adapter and installed it into the laptop as a HDD and installed this way, selecting the embedded kernel at the finish - boots fine except only 1 NIC.

I have tried and followed the "NanoBSD on WRAP" link and all the other documents that I can come across, but I can't seem to get it to go. I've worked with various flavours of *nix for the last 10 or so years, so I'm not afriad to get in and play..

The end result is that the FBII won't boot. It comes up with an option of "F1 FreeBSD" and "F5 Drive 0" and then proceeds to show a line of ### forever. Every push of a key on the keyboard results in more ###'s.

Thanks for your direction and my apologies for the first post.
Regards
Karl

Search the forum and doc for tuning your WRAP. Basically you need to enable polling on your network card, so the cpu doesn't get hammered by interrupts from the network card.

MageMinds

pheleven: Thanks for the pciconf output. The four NICs are these devices:

none13@pci0:9:0:0:      class=0x020000 card=0xa02b8086 chip=0x10e88086 rev=0x01 hdr=0x00
none14@pci0:9:0:1:      class=0x020000 card=0xa02b8086 chip=0x10e88086 rev=0x01 hdr=0x00
none15@pci0:8:0:0:      class=0x020000 card=0xa02b8086 chip=0x10e88086 rev=0x01 hdr=0x00
none16@pci0:8:0:1:      class=0x020000 card=0xa02b8086 chip=0x10e88086 rev=0x01 hdr=0x00

The device id 0x10e8 (upper half of what pciconf lists as the chip value) is identified in the FreeBSD 8.0 sources as E1000_DEV_ID_82576_QUAD_COPPER in source file e1000/e1000_hw.h which is referenced in e1000/if_igb.c in a table of IDs for supported devices. This entry is missing from the equivalent table in the FreeBSD 7.2 sources.

The FreeBSD 7.2 table includes these 82576 devices: E1000_DEV_ID_82576, E1000_DEV_ID_82576_FIBER,  E1000_DEV_ID_82576_SERDES

The FreeBSD 8.0 table includes these 82576 devices: E1000_DEV_ID_82576, E1000_DEV_ID_82576_NS,  E1000_DEV_ID_82576_FIBER,  E1000_DEV_ID_82576_SERDES, E1000_DEV_ID_82576_SERDES_QUAD, E1000_DEV_ID_82576_QUAD_COPPER

You might also consider some kind of high performance IP router in front, and then multiple pfSense boxes on separate network segments behind to do filtering. For example, have 1/3 of your IP space routed to each of three pfSense boxes from whatever CPE is terminating the OC-48.

As far as I can tell your parts all look compatible. The E8400 looks like a good choice for the throughput you're looking for. Should mae for an interesting unit. Be sure to post some observations when you get it up and running.  :)

If they're on the accepted hardware list for BSD, it really won't matter too much.  The chipset of the card is what's going to make the difference though.  If any of them have a Broadcom or Intel chipset, then use that card(s).

Still remember that those brands are consumer based cards and not "server" cards.  So really, for your environment, you probably wouldn't be pushing the card's limitations anyway to notice any sort of a difference.  If you plan to use this in a production environment, then by all means, spend the money for Intel cards.

Perfect! thanks guys!

I was hoping I wouldn't have to buy a atom / core2duo and waste more power.

I can't wait to get started…...Thanks again!  :)

FYI
I have a Via C7 1GHz, Intel PRO/1000 GT and 512MB ram.

On my home network the max i have seen CPU usage peak at is 30-32%.
This is when downloading at 20Mb/s on my cable.

Neither the dbus output nor the pciconf output show any devices downstream of the cardbus bridge. Either your device is not fully engaged in the socket (and therefore its not visible to the hardware) or the cardbus bridge is not correctly initialised by the BIOS for FreeBSD or the bridge is broken or your card is broken. You should do some experiments to eliminate some of these possibilities.

I have come across some systems where the cardbus bridge is not correctly initialised by the BIOS for operation with FreeBSD and FreeBSD has not seen cardbus devices but other operating systems have seen cardbus devices on the same hardware. FreeBSD seems to be more commonly used on servers (which typically don't have cardbus devices) than on laptops (which do commonly have cardbus slots) and hence the cardbus support in FreeBSD probably doesn't get as much of a workout as it needs.

An alternative to the cardbus NIC might be a USB NIC. I don't know of any USB gigabit NICs but there are 10/100 USB ethernet NICs that have 480Mbps USB speed and are supported in FreeBSD. Warning: Many of the cheap 10/100 USB ethernet NICs have only 12Mbps USB speed (which is more than enough for some circumstances).

Another option might be to get an inexpensive (But Supported) Sata Raid card which supports Raid1 and put that in.  You could probably buy one for under $75.00.

Still humming along. It doesn't have a huge load on it. A T1 for the internet, two ipsec vpn's and 5 openvpn clients. 19 internal users. Perhaps overkill I don't know.

In the next month or so, I hope to get the Supermicro Atom mobo (I forget the model, but it has dual intel nics) and use it as a test

Thanks for the link. I re did the image and used the imbedded image written to a CF card via a usb reader. Took the card and inserted into a laptop - worked like a charm - booted and asked to setup the interfaces. Whislt I had it running, made the changes as advised in the link. Put the card into the WG and booted with a serial console. THe boot menu now comes up with three options, 1 freebsd, 2 freebsd, 5 other. After making the changes, at least it appears to want to boot, but just sits there with the cursor flashing. Redid the image, used the CD and did an embedded install, used the nano image and still the same thing. I thought after reading the link, this is it, this will fix it. Bummer, no.

Thanks for the idea.
Karl

On topic, my brother just retired a PD 3.0GHz. It was using 130W at the wall, idle. The replacement X2 240 system measures ~45W idle at the wall. Both using integrated ATI graphics. The newer C2D I would expect to use no more than that in a similar configuration, so yeah, quite a difference.

@jimp:

I thought that some USB ralink devices were supposed to work, but it may be picky about the chipset. IIRC the ural and rum drivers were both on 1.2.3, both of those are USB ralink drivers.

I've had the pfSense 1.2.3 rum driver recognise a DLink G122 and a TP-Link TL-WN321G USB sticks. I think only some revisions of the G122 have the Ralink chipset.

@vatson:

VLAN 2 is not actually used anymore on our network, so I'll probably keep it as a "ghost native vlan" for the pfSense uplink port.

Interesting.  I do that for all my Cisco switches as a matter of best practice, but I've never seen it required.

The AP is separate, a Ubiquiti radio. I don't think they support client isolation, but I could be wrong.

no i mean double port card not quad, why? it is cheaper :)
Why would riser make difference,
well since is one slot with riser that have 2 slots, and 2 nics… so i asked...

many thanks for answers!

I don't know if you're still working on this, but this seems similar behavior to my Nokia ip530.  I needed to enable "device polling" in the advanced setup (while the HDD was still in the host machine used to load pfSense).  Otherwise, ifconfig would show two NICs as active, but would not have an IP or be pingable and the other 2 NICs wouldn't come up at all. YMMV.
edit: these were Intel 21143 NICs.