Collection of issues with Atom JNC96FL setup - RC2 05/Sept/13 i386 - update 0909

splat1

I will continue to update this post as i find new issues and solutions.

Motherboard: JNC96FL-525-LF
Ram: 4GB
Hdd: 500gb SATA
Network: RTL8111DL + PCI Intel 21143 quad port (Don't have the exact card model to hand)

Full HDD install.

Previous version working 100% fine 2.0.3
Current Snapshot RC2 05/Sept/13 i386

WAN: Re0 - PPPOE to Vigor 120 modem.
LAN: dc3 - 192.168.150.3 - DHCPD enabled.
WLAN: dc2 - 192.168.1.1 - DHCPD enabled.
dc1 & dc0 unused.

[2.1-RC2][root@hostnamewithheld]/root(1): vmstat -i
interrupt total rate
irq4: uart0 958 0
irq16: dc0 uhci3 14 0
irq17: dc1 atapci0 16 0
irq18: dc2 atapci1+ 934701 2
irq19: dc3 uhci1+ 33319168 106
irq23: uhci0 ehci0 1130 0
cpu0: timer 623206855 1999
irq256: re0 49411182 158
cpu1: timer 623186454 1999
cpu3: timer 623187281 1999
cpu2: timer 623187076 1999
Total 2576434835 8268

Just yell for any additional info, logs, kernel output.

Hey all,

First a hard drive oddity
The HDD on this box is plugged into the second sata port on the board and was set for IDE compatible mode.
Should be on the Intel NM10 chipset and not the JMB363.
Under 2.0.3 this worked fine and came up as ad2s1a, under 2.1-rc2 it fails to detect completely under GOM.
Switching the Sata mode to AHCI it now shows up as ad14s1a.

Annoying but far from the end of the world, solved with a quick edit to fstab.

Second and fairly big problem.

The Intel 21143 quad port is detected fine and the ports receive data but can't send anything.
First noticed this looking at the logs, DHCPD sees a request on the port and replies.
Yet a packet capture of the port shows no such reply.
Tried to ping a machine with a fixed ip on the lan and got, sendto: No buffer space.

Unplugged the cable, waited 30 seconds and plugged it back in. port works fine with traffic in both directions.

Sadly this happens on every reboot and I could do with solving.

Any help would be appreciated,
Splat.

eri--

Probably disable msix interrupts on the driver.
Mostly a driver issue there is suspect.

wallabybob

@ermal:

Probably disable msix interrupts on the driver.

I would be surprised if the 21143 was even MSI capable but lets check:

splat1: Please post the output of pfSense shell command```
vmstat -i


@splat1:

> Under 2.0.3 this worked fine and came up as ad2sa1, under 2.1-rc2 it fails to detect completely under GOM.
> Switching the Sata mode to AHCI it now shows up as ad14sa1.

I wouldn't expect disk names like those! Did you mean **ad2s1a** and **ad14s1a**?

splat1

[2.1-RC2][root@hostnamewithheld]/root(1): vmstat -i
interrupt total rate
irq4: uart0 958 0
irq16: dc0 uhci3 14 0
irq17: dc1 atapci0 16 0
irq18: dc2 atapci1+ 934701 2
irq19: dc3 uhci1+ 33319168 106
irq23: uhci0 ehci0 1130 0
cpu0: timer 623206855 1999
irq256: re0 49411182 158
cpu1: timer 623186454 1999
cpu3: timer 623187281 1999
cpu2: timer 623187076 1999
Total 2576434835 8268

Yep sorry that's a mistype, ad2s1a and ad14s1a are correct.

wallabybob

The vmstat output shows dc0, dc1, dc2 and dc3 use irps 16, 7, 18 and 19 respectively. In FreeBSD msi interrupts start at 256. Your dcx NICs are not using MSI.

On rereading this post
@splat1:

The Intel 21143 quad port is detected fine and the ports receive data but can't send anything.
First noticed this looking at the logs, DHCPD sees a request on the port and replies.
Yet a packet capture of the port shows no such reply.

I realized I didn't understand what you meant by "packet capture of the port". Do you mean packet capture run on the pfSense box OR on the box issuing the DHCP requests? Are you concluding dhcpd replies from a reading of the pfSense dhcp log file?

splat1

I ran a packet capture on the pfsense box and the box issuing the DHCP request, both identical.
yes, my first stop was the log and that shows its making an offer, so then I ran a packet cap at both ends.