The hardware we are discussing complies to ACPI specifications, as such I beleive the following link should answer your questions
I took a close look at the documentation that you have suggested.
This is not what I'm looking for.
I want to use the TCO watchdog that is a part of the intel 100 series chipset PCH.
I use Intel datasheets 100-series-chipset-datasheet-vol-1.pdf & 100-series-chipset-datasheet-vol-2.pdf
We also are having this same issue running Ubuntu 18.04LTS on NUC7i3BNK. Everything looks fine in dmesg etc., but the counter just never starts, so the server never reboots when the watchdog should tell it to. Not having much luck solving it... any help gratefully received!
Any movement on this as the problem; It is still there from what I can tell.
I'm using the 4.18.16, 4.20.16, and 5.1.11 kernels on a fedora 29 distro. This watchdog hasn't worked since moving to Skylake devices (working in Sandybridge, Ivybridge and before that). It's broken on Cannon Lake and Coffee lake devices as well. There are no idicatations of an error other than that fact that the timer countdown will not start (Timeleft) and the watchdog does not reset the device.
[ 4.206354] i801_smbus 0000:00:1f.4: enabling device (0001 -> 0003)
[ 4.206825] i801_smbus 0000:00:1f.4: SPD Write Disable is set
[ 4.206858] i801_smbus 0000:00:1f.4: SMBus using PCI interrupt
[ 168.218443] iTCO_vendor_support: vendor-support=0
[ 168.221443] iTCO_wdt: Intel TCO WatchDog Timer Driver v1.11
[ 168.221510] iTCO_wdt: Found a Intel PCH TCO device (Version=4, TCOBASE=0x0400)
[ 168.221985] iTCO_wdt: initialized. heartbeat=30 sec (nowayout=0)
~# wdctl /dev/watchdog0
wdctl: write failed: Invalid argument
Identity: iTCO_wdt [version 0]
Timeout: 30 seconds
Pre-timeout: 0 seconds
Timeleft: 30 seconds
FLAG DESCRIPTION STATUS BOOT-STATUS
KEEPALIVEPING Keep alive ping reply 1 0
MAGICCLOSE Supports magic close char 0 0
SETTIMEOUT Set timeout (in seconds) 0 0
Any help understanding why this is so would be greatly appreciated. Are there any known actual workarounds. It's hard to believe this issue isn't already known well in the embedded linux community.
are there any updates on this topic?
We have the same problem (dmesg output looks fine but watchdog is not counting down) on different Kaby Lake CPUs (i3, i5 and i7 NUC7 models) and different kernels (4.15, 4.19, 5.3, 5.4 rc5)
I ran into this same issue during a large deployment of NUC7i5BNK units.
I worked through this with my Intel rep and it was escalated to the BIOS team, and they have fixed the issue in BIOS 0083 which goes live tonight. Hopefully this is helpful.
I still see the same issue with iTCO watchdog on intel C236 chipset (Xeon E3-1275 V5 processor).
From the above comment it seems that a fix has been identified. Could you please share more details about the fix and how different boards with TCO watchdog can get the fix?
It will be very helpful. Looking forward for your quick response. Thank you!
We had the exact same problem with a NUC7 device. With the latest BIOS the problem gone.
But with a NUC8 (NUC8I5BEB), even with the latest BIOS, the problem persists. Is there a known problem with the watchdog ?