System locks up after watchdog timeout

[Zeschg, Thomas]

Hi,

I encountered the problem, that the system gets stuck after a watchdog timeout and no reboot/power cycle is performed. It seems like the efibootguard initializes the Intel TCO correctly, obviously the timeout value is the one I configured. I use a Fujitsu Server with the Intel C612 chipset.

A second problem is the iTCO_wdt kernel module. I get an error message when loading the module: "failed to reset NO_REBOOT flag, reboot disabled by hardware/BIOS". Maybe the two problems are related.

Does anyone have a suggestion?

With best regards,
Thomas Zeschg

Siemens AG, Corporate Technology, CT RDA ITS SES-DE
Otto-Hahn-Ring 6, 81739 Muenchen, Germany

[Jan Kiszka]

On 2018-02-16 10:10, Zeschg, Thomas (CT RDA ITS SES-DE) wrote:
> Hi,
>
> I encountered the problem, that the system gets stuck after a watchdog timeout and no reboot/power cycle is performed. It seems like the efibootguard initializes the Intel TCO correctly, obviously the timeout value is the one I configured. I use a Fujitsu Server with the Intel C612 chipset.

Still need to look into that, namely the reference to
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/watchdog/iTCO_vendor_support.c
you mentioned offlist. We are likely missing those quirks...

What is the PCI ID of your board? Would Linux select a quirk?

>
> A second problem is the iTCO_wdt kernel module. I get an error message when loading the module: "failed to reset NO_REBOOT flag, reboot disabled by hardware/BIOS". Maybe the two problems are related.

If you boot without watchdog support (watchdog=0), does Linux still
complain?

Jan

--
Siemens AG, Corporate Technology, CT RDA IOT SES-DE
Corporate Competence Center Embedded Linux

[Zeschg, Thomas]

Hi,

The PCI ID of our board is 0x8d44. Linux selects the vendor specific support mode of 0 (none), which is the default value. The whole output of dmesg regarding iTCO is:
[ 12.574996] iTCO_vendor_support: vendor-support=0
[ 12.575764] iTCO_wdt: Intel TCO WatchDog Timer Driver v1.11
[ 12.575844] iTCO_wdt: unable to reset NO_REBOOT flag, device disabled by hardware/BIOS

I get the same result from Linux, when I use Grub to boot.

Best regards
Thomas

[Jan Kiszka]

On 2018-02-19 08:26, Zeschg, Thomas (CT RDA ITS SES-DE) wrote:
> Hi,
>
> The PCI ID of our board is 0x8d44. Linux selects the vendor specific support mode of 0 (none), which is the default value. The whole output of dmesg regarding iTCO is:
> [ 12.574996] iTCO_vendor_support: vendor-support=0
> [ 12.575764] iTCO_wdt: Intel TCO WatchDog Timer Driver v1.11
> [ 12.575844] iTCO_wdt: unable to reset NO_REBOOT flag, device disabled by hardware/BIOS
>
> I get the same result from Linux, when I use Grub to boot.
>

Then the BIOS is preventing the use of the watchdog in this machine. Did
you check if there are any configuration options in the BIOS that may
change this?

[Zeschg, Thomas]

Yes, I checked this, but there is nothing. Nevertheless I don't understand why the system gets stuck after the watchdog timeout.

[Jan Kiszka]

On 2018-02-19 08:48, Zeschg, Thomas (CT RDA ITS SES-DE) wrote:
> Yes, I checked this, but there is nothing. Nevertheless I don't understand why the system gets stuck after the watchdog timeout.

BIOS is broken? That's way too common to be surprised. If there is no
update available, switch the hardware - sorry.