Need driver for hardware watchdog

clarknova

My Supermicro X7SPA-H BIOS has an option to turn on a hardware watchdog. When I turn it on (last tried while running pfsense 2.0 beta 3 or 4) the machine resets at a regular interval, so I guess it's working. Is there a way to get a driver for this in pfsense 2.0? I did google around but found nothing useful, except suggestions that the driver might already be included (if so, it's not doing a very good lob, last I tried).

Cino

I noticed the same thing a few months ago on the X7SPA-HF board. I may have to try this again and see. Your using the watchdog with the built-in 5min timer? I think it was under CPU or chipset config(cant remember). Since the HF has IPMI, it has a watchdog under the IPMI BMC setup in the bios. This setting allows me to 10sec/30sec/1min/5min timers I believe.

clarknova

It's been a while since I looked at the bios on this board, but I don't believe there is an option to set the watchdog timeout. It was on about a 5 minute interval as I recall. I thought I read that pfsense included a watchdog driver, so maybe SM has a specific driver for this board that I would have to download, assuming there's a FreeBSD flavour?

Cino

I checked SM website real quick and didn't see anything for FreeBSD :-( There is a BIOS update 1.1a for the boards tho…

stephenw10

Isn't the watchdog part of the ICH9 southbridge on this board?
In which case have you tried the ichwd driver?

Steve

Edit: Looks like it should work, see here.

Another edit: It looks like the watchdog set in the bios is probably on the Winbond SuperIO chip for which no drivers exist, yet. You could write your own code to 'pat' it though.

_igor_

I have the same X7SPA-HF and entered an ipmi_load="YES" at /boot/loader.conf.local.
Now dmesg shows that:

ipmi0: IPMI device rev. 1, firmware rev. 2.2, version 2.0
ipmi0: Number of channels 2
ipmi0: Attached watchdog

Tested without getting the watchdog resets. But the ugly thing is that you have at several points watchdog-entries in the BIOS. I have to look back where i activated the watchdog here…
So that will only work with the HF boards i think.

ichwd seems not to work on my board, a "kldload ichwd" gave this:
kldload: can't load ichwd: No such file or directory

But by the way, does anyone have kvm AND console redirected via com1? Say KVM or console via serial at the same time?

stephenw10

I am still using beta5 but the module exists for me in /boot/kernel.


[2.0-BETA5][root@pfSense.localdomain]/boot/kernel(29): kldload ichwd
ichwd module loaded
ichwd0: <intel ich6m="" watchdog="" timer="">on isa0
ichwd0: unable to reserve SMI registers
device_attach: ichwd0 attach returned 6</intel>

Steve

Cino

I have the HF board. I created the loader.conf.local file and now i see this in dmeg :-)

ipmi0: <ipmi system="" interface="">on isa0
ipmi0: KCS mode found at io 0xca2 alignment 0x1 on isa
ipmi0: IPMI device rev. 1, firmware rev. 2.2, version 2.0
ipmi0: Number of channels 2
ipmi0: Attached watchdog</ipmi>

I'm thinking this watchdog is under the IPMI settings. I can't access the BIOS right now but will later and see if this works.

For the ichwd driver:
I'm running 2.0-RC1 (i386) built on Thu Mar 3 17:43:05 EST 2011 and don't have ichwd driver… I would be looking for ichwd.ko, under /boot/kernel?

stephenw10

Hmm, OK.
Well here are mine from beta5. Obviously rename by removing the .png extension.
Don't know if you need ichwd.ko.symbols or not.

Steve

ichwd.ko.png
ichwd.ko.symbols.png

Cino

@stephenw10:

Hmm, OK.
Well here are mine from beta5. Obviously rename by removing the .png extension.
Don't know if you need ichwd.ko.symbols or not.

Steve

Thanks Steve!! I copied them over to my box and it loaded… Now i'll have to change the bios settings... Which beta5 snapshot did these come from? i looked at a couple of the upgrades and didn't see them...

stephenw10

Now I think about it I may have copied them onto my box from a FreeBSD 8.1_rel install when I was playing with ICH drivers a while ago. :P

As I said above, from reading the blog post it looks likely that the bios settings will be for the SuperIO watchdog not the ICH one. If that is the case you should leave everything turned off in the bios if you're experimenting with ichwd and watchdogd.

Steve

Cino

thanks again Steve.. I did some experimenting this evening based on the blog post and it seems to work. I added '/etc/rc.d/watchdogd forcestart' as a shellcmd… rebooted the box... left it for an hour then ran 'killall -9 watchdogd' less then a minute later, the box rebooted on its own.

igor, If you remember what you did, please share.. I'm curious about the ipmi watchdog..

_igor_

@cino: yep. I'm outside this week, but then i will report back. Got a new bios and that will give me time to test it.

clarknova

@stephenw10:

Well here are mine from beta5.


kldload ichwd
kldload: can't load ichwd: Exec format error

I'm guessing yours is a 32-bit system? I'm running amd64.

stephenw10

Indeed, 32bit. Sorry.

Just to set the record straight they came from a standard install of FreeBSD 8.1_rel I use for compiling stuff for pfSense.

Steve

Cino

@clarknova, if you don't have FreeBSD AMD64 installed on a PC.. Go to PC-BSD website and download their DVD ISO. You can run it LIVE without installing it to a harddrive. If the files are there, copy them to a USB stick then transfer them to your pfsense box. Its long shot, but it worked for me on a different issue. I needed cdcontrol to get my 3G USB Modem working.. Worked like a charm… I used there 8.0 release if I remember correctly.

MrKoen

I got myself a SuperMicro X7SPA-HF-D525 board for my pfSense 2 RC1 AMD64 server. It tends to hang every now and then so I would like to use the watchdog feature so it will reboot. I found this topic and did as told:

1. In the BIOS I went to Advanced -> IPMI configuration and set BMC Watch Dog Timer Action to "Reset System" and BMC WatchDog TimeOut to 5 minutes
2. I booted pfSense and updated /boot/loader.conf.local and added ipmi_load="YES"
3. I downloaded ichwd.ko and ichwd.ko.symbols using fetch in the console to /boot/kernel on my pfSense machine

I do see the ipmi0 entires in dmesg. It still reboots every 5 minutes though. Did you guys chmod the ichwd.ko and ichwd.ko.symbols files with +x in /boot/kernel? I'm getting the message "Unsupported file layout. Exec format error." when doing a kldload ichwd in the console.

I'm using pfSense build 2.0-RC1-IPv6 (amd64)built on Thu Apr 28 03:47:19 EDT 2011.

Can anyone provide me with tips what to try?

wallabybob

You probably need to invoke a program to "tickle" the watchdog "often enough".

See the FreeBSD man page for ichwd (http://www.freebsd.org/cgi/man.cgi?query=ichwd&apropos=0&sektion=0&manpath=FreeBSD+8.2-RELEASE&format=html ) and the links in the See Also section.

stephenw10

There are at least two watchdogs on this board.
As I said in a previous message the watchdog invoked by the bios is likely the one on the Winbond SuperIO chip and NOT the ich watchdog. Thus the ICH driver will not 'pat' it.
The drivers I posted are for 32bit only so you will need some for 64bit, unless you've already done that. ;)

Cino's report of success involved loading the ich drivers and starting the watchdog daemon. This will then set the watchdog and keep 'patting' it. You should leave the ipmi watchdog off in the bios.

Good luck! :)

Steve

MrKoen

Thanks for your pointers guys. I'm fairly new to the worlds of Unix and Linux, so let me rephrase your comments to see if I understand them correctly.

1. This particular board probably has two Watchdogs of which one of them is controlled by the BIOS and one can be controlled by the right driver/software
2. I should not use the BIOS setting since the ichwd driver/software can not tickle that dog to keep it awake and thus it will reboot
3. I should install a new server with FreeBSD 8.1 AMD64 and retrieve the ichwd.ko and ichwd.ko.symbols files from the /boot/kernel folder and copy them to my pfSense installation. After that I should try the kldload ichwd.ko again and I can kill the FreeBSD 8.1 server again.

Makes me wonder.. what sets the timeout on the non-BIOS watchdog? Does it work like when the ichwd software sends the non-BIOS watchdog out of the door to guard the garden and whistle at it at some interval to keep it awake? And when it stops whistling at it, a chip on the mainboard activated by ichwd once it started detects the watchdog has fallen asleep because of the lack of whistles and reboots the system?