watchdog: iTCO_wdt: failed to load

watchdog: iTCO_wdt: failed to load

[Vignesh Raman]

Hi,

I'm seeing an issue where the iTCO_wdt module doesn't load with the below error,

[    3.827599] iTCO_wdt: Intel TCO WatchDog Timer Driver v1.11
[    3.828440] iTCO_wdt iTCO_wdt: can't request region for resource
[mem 0x00c5fffc-0x00c5ffff]
[    3.828803] iTCO_wdt: probe of iTCO_wdt failed with error -16

CPU and logs:
cpu: Intel(R) Core(TM) i3-8100 CPU @ 3.60GHz
dmesg logs: http://paste.debian.net/1092702/
lsmod output: https://paste.debian.net/1092703/
root@localhost:~# uname -a
Linux localhost.localdomain 4.14.103-rt55 #1 SMP PREEMPT RT Tue Jul 23
03:53:28 UTC 2019 x86_64 GNU/Linux

Please let me know if the error (can't request region for resource) is
related to the patches below,
https://patchwork.kernel.org/patch/9627085/
https://lore.kernel.org/patchwork/patch/770990/

I tested with the patch
https://lore.kernel.org/patchwork/patch/770990/ and it fixes the
issue.
root@localhost:/home/vignesh# dmesg | grep wdt
[    5.766470] iTCO_wdt: Intel TCO WatchDog Timer Driver v1.11
[    5.767104] iTCO_wdt: Found a Intel PCH TCO device (Version=4,
TCOBASE=0x0400)
[    5.767865] iTCO_wdt: initialized. heartbeat=30 sec (nowayout=0)

But the patch was not merged upstream.

When testing the hardware watchdog, it doesn't trigger a reset even
after triggering a sysrq crash.
The system freezes and there is no reboot.

I noticed the timeleft value is not getting updated.
root@localhost:~# cat /sys/class/watchdog/watchdog0/timeleft
15
root@localhost:~# cat /sys/class/watchdog/watchdog0/timeleft
15
root@localhost:~# cat /sys/class/watchdog/watchdog0/timeleft
15

This issue has been discussed in below email thread,
http://lkml.iu.edu/hypermail/linux/kernel/1609.1/01979.html
https://lkml.org/lkml/2016/9/8/641

In another platform the hardware watchdog works.
root@debian:~# dmesg | grep wdt
[    3.901842] iTCO_wdt: Intel TCO WatchDog Timer Driver v1.11
[    3.903415] iTCO_wdt: Found a Intel PCH TCO device (Version=4,
TCOBASE=0x0400)
[    3.937816] iTCO_wdt: initialized. heartbeat=30 sec (nowayout=0)

CPU and logs
cpu: Intel(R) Core(TM) i7-6700 CPU @ 3.40GHz
dmesg logs: http://paste.debian.net/1092845/
lsmod output: http://paste.debian.net/1092846/
root@debian:~# uname -a
Linux debian 4.14.103-rt55 #1 SMP PREEMPT RT Tue Jul 23 03:53:28 UTC
2019 x86_64 GNU/Linux

Hardware watchdog triggers a reset and the watchdog timer counter is
started and the value gets updated.
root@debian:~# cat /sys/class/watchdog/watchdog0/timeleft
13
root@debian:~# cat /sys/class/watchdog/watchdog0/timeleft
12
root@debian:~# cat /sys/class/watchdog/watchdog0/timeleft
11
root@debian:~#

My questions:
1. Please let me know if it is okay to take
https://lore.kernel.org/patchwork/patch/770990/ patch or will there be
an update to the original patch?
2. Could the hardware watchdog not triggering a reset (timeleft value
not getting updated) could be a hardware problem?

Thanks.

Regards,
Vignesh

Re: watchdog: iTCO_wdt: failed to load

[Guenter Roeck]

On 7/24/19 12:34 AM, Vignesh Raman wrote:
> Hi,
> 
> I'm seeing an issue where the iTCO_wdt module doesn't load with the below error,
> 
> [    3.827599] iTCO_wdt: Intel TCO WatchDog Timer Driver v1.11
> [    3.828440] iTCO_wdt iTCO_wdt: can't request region for resource
> [mem 0x00c5fffc-0x00c5ffff]
> [    3.828803] iTCO_wdt: probe of iTCO_wdt failed with error -16
> 
> CPU and logs:
> cpu: Intel(R) Core(TM) i3-8100 CPU @ 3.60GHz
> dmesg logs: http://paste.debian.net/1092702/
> lsmod output: https://paste.debian.net/1092703/
> root@localhost:~# uname -a
> Linux localhost.localdomain 4.14.103-rt55 #1 SMP PREEMPT RT Tue Jul 23
> 03:53:28 UTC 2019 x86_64 GNU/Linux
> 
> Please let me know if the error (can't request region for resource) is
> related to the patches below,
> https://patchwork.kernel.org/patch/9627085/
> https://lore.kernel.org/patchwork/patch/770990/
> 
> I tested with the patch
> https://lore.kernel.org/patchwork/patch/770990/ and it fixes the
> issue.
> root@localhost:/home/vignesh# dmesg | grep wdt
> [    5.766470] iTCO_wdt: Intel TCO WatchDog Timer Driver v1.11
> [    5.767104] iTCO_wdt: Found a Intel PCH TCO device (Version=4,
> TCOBASE=0x0400)
> [    5.767865] iTCO_wdt: initialized. heartbeat=30 sec (nowayout=0)
> 
> But the patch was not merged upstream.
> 
> When testing the hardware watchdog, it doesn't trigger a reset even
> after triggering a sysrq crash.
> The system freezes and there is no reboot.
> 
> I noticed the timeleft value is not getting updated.
> root@localhost:~# cat /sys/class/watchdog/watchdog0/timeleft
> 15
> root@localhost:~# cat /sys/class/watchdog/watchdog0/timeleft
> 15
> root@localhost:~# cat /sys/class/watchdog/watchdog0/timeleft
> 15
> 
> This issue has been discussed in below email thread,
> http://lkml.iu.edu/hypermail/linux/kernel/1609.1/01979.html
> https://lkml.org/lkml/2016/9/8/641
> 
> In another platform the hardware watchdog works.
> root@debian:~# dmesg | grep wdt
> [    3.901842] iTCO_wdt: Intel TCO WatchDog Timer Driver v1.11
> [    3.903415] iTCO_wdt: Found a Intel PCH TCO device (Version=4,
> TCOBASE=0x0400)
> [    3.937816] iTCO_wdt: initialized. heartbeat=30 sec (nowayout=0)
> 
> CPU and logs
> cpu: Intel(R) Core(TM) i7-6700 CPU @ 3.40GHz
> dmesg logs: http://paste.debian.net/1092845/
> lsmod output: http://paste.debian.net/1092846/
> root@debian:~# uname -a
> Linux debian 4.14.103-rt55 #1 SMP PREEMPT RT Tue Jul 23 03:53:28 UTC
> 2019 x86_64 GNU/Linux
> 
> Hardware watchdog triggers a reset and the watchdog timer counter is
> started and the value gets updated.
> root@debian:~# cat /sys/class/watchdog/watchdog0/timeleft
> 13
> root@debian:~# cat /sys/class/watchdog/watchdog0/timeleft
> 12
> root@debian:~# cat /sys/class/watchdog/watchdog0/timeleft
> 11
> root@debian:~#
> 
> My questions:
> 1. Please let me know if it is okay to take
> https://lore.kernel.org/patchwork/patch/770990/ patch or will there be
> an update to the original patch?

Not into the upstream kernel. As outlined in the discussion, it would
cause failures on various platforms (if the driver is instantiated
from lpc_ipc.c).

> 2. Could the hardware watchdog not triggering a reset (timeleft value
> not getting updated) could be a hardware problem?
> 
Sounds more like the watchdog isn't running. I don't know about that system;
it may well be that the driver doesn't fully support that hardware.

Guenter

Re: watchdog: iTCO_wdt: failed to load

[Guenter Roeck]

On 7/24/19 12:34 AM, Vignesh Raman wrote:
> Hi,
> 
> I'm seeing an issue where the iTCO_wdt module doesn't load with the below error,
> 
> [    3.827599] iTCO_wdt: Intel TCO WatchDog Timer Driver v1.11
> [    3.828440] iTCO_wdt iTCO_wdt: can't request region for resource
> [mem 0x00c5fffc-0x00c5ffff]
> [    3.828803] iTCO_wdt: probe of iTCO_wdt failed with error -16
> 
> CPU and logs:
> cpu: Intel(R) Core(TM) i3-8100 CPU @ 3.60GHz
> dmesg logs: http://paste.debian.net/1092702/
> lsmod output: https://paste.debian.net/1092703/
> root@localhost:~# uname -a
> Linux localhost.localdomain 4.14.103-rt55 #1 SMP PREEMPT RT Tue Jul 23
> 03:53:28 UTC 2019 x86_64 GNU/Linux
> 
> Please let me know if the error (can't request region for resource) is
> related to the patches below,
> https://patchwork.kernel.org/patch/9627085/
> https://lore.kernel.org/patchwork/patch/770990/
> 
> I tested with the patch
> https://lore.kernel.org/patchwork/patch/770990/ and it fixes the
> issue.
> root@localhost:/home/vignesh# dmesg | grep wdt
> [    5.766470] iTCO_wdt: Intel TCO WatchDog Timer Driver v1.11
> [    5.767104] iTCO_wdt: Found a Intel PCH TCO device (Version=4,
> TCOBASE=0x0400)
> [    5.767865] iTCO_wdt: initialized. heartbeat=30 sec (nowayout=0)
> 
> But the patch was not merged upstream.
> 
> When testing the hardware watchdog, it doesn't trigger a reset even
> after triggering a sysrq crash.
> The system freezes and there is no reboot.
> 
> I noticed the timeleft value is not getting updated.
> root@localhost:~# cat /sys/class/watchdog/watchdog0/timeleft
> 15
> root@localhost:~# cat /sys/class/watchdog/watchdog0/timeleft
> 15
> root@localhost:~# cat /sys/class/watchdog/watchdog0/timeleft
> 15
> 
> This issue has been discussed in below email thread,
> http://lkml.iu.edu/hypermail/linux/kernel/1609.1/01979.html
> https://lkml.org/lkml/2016/9/8/641
> 
> In another platform the hardware watchdog works.
> root@debian:~# dmesg | grep wdt
> [    3.901842] iTCO_wdt: Intel TCO WatchDog Timer Driver v1.11
> [    3.903415] iTCO_wdt: Found a Intel PCH TCO device (Version=4,
> TCOBASE=0x0400)
> [    3.937816] iTCO_wdt: initialized. heartbeat=30 sec (nowayout=0)
> 
> CPU and logs
> cpu: Intel(R) Core(TM) i7-6700 CPU @ 3.40GHz
> dmesg logs: http://paste.debian.net/1092845/
> lsmod output: http://paste.debian.net/1092846/
> root@debian:~# uname -a
> Linux debian 4.14.103-rt55 #1 SMP PREEMPT RT Tue Jul 23 03:53:28 UTC
> 2019 x86_64 GNU/Linux
> 
> Hardware watchdog triggers a reset and the watchdog timer counter is
> started and the value gets updated.
> root@debian:~# cat /sys/class/watchdog/watchdog0/timeleft
> 13
> root@debian:~# cat /sys/class/watchdog/watchdog0/timeleft
> 12
> root@debian:~# cat /sys/class/watchdog/watchdog0/timeleft
> 11
> root@debian:~#
> 
> My questions:
> 1. Please let me know if it is okay to take
> https://lore.kernel.org/patchwork/patch/770990/ patch or will there be
> an update to the original patch?
> 2. Could the hardware watchdog not triggering a reset (timeleft value
> not getting updated) could be a hardware problem?
> 

What is the output of /proc/iomem, what PCI devices does it have, and what are
the ACPI devices ? Reason for asking is that I don't immediately see the ACPI
or PCI devices associated with above patch in your dmesg. If not, the patch
might actually cause the watchdog in your system not to work.

Also, did this ever work in your system ? If it did work, did the failure
start after a kernel update or after a BIOS update ?

Thanks,
Guenter

Re: watchdog: iTCO_wdt: failed to load

[Andy Shevchenko]

+Cc: Mika, who is supporting watchdog driver(s).

On Wed, Jul 24, 2019 at 01:04:45PM +0530, Vignesh Raman wrote:
> Hi,
> 
> I'm seeing an issue where the iTCO_wdt module doesn't load with the below error,
> 
> [    3.827599] iTCO_wdt: Intel TCO WatchDog Timer Driver v1.11
> [    3.828440] iTCO_wdt iTCO_wdt: can't request region for resource
> [mem 0x00c5fffc-0x00c5ffff]
> [    3.828803] iTCO_wdt: probe of iTCO_wdt failed with error -16
> 
> CPU and logs:
> cpu: Intel(R) Core(TM) i3-8100 CPU @ 3.60GHz
> dmesg logs: http://paste.debian.net/1092702/
> lsmod output: https://paste.debian.net/1092703/
> root@localhost:~# uname -a
> Linux localhost.localdomain 4.14.103-rt55 #1 SMP PREEMPT RT Tue Jul 23
> 03:53:28 UTC 2019 x86_64 GNU/Linux
> 
> Please let me know if the error (can't request region for resource) is
> related to the patches below,
> https://patchwork.kernel.org/patch/9627085/
> https://lore.kernel.org/patchwork/patch/770990/
> 
> I tested with the patch
> https://lore.kernel.org/patchwork/patch/770990/ and it fixes the
> issue.
> root@localhost:/home/vignesh# dmesg | grep wdt
> [    5.766470] iTCO_wdt: Intel TCO WatchDog Timer Driver v1.11
> [    5.767104] iTCO_wdt: Found a Intel PCH TCO device (Version=4,
> TCOBASE=0x0400)
> [    5.767865] iTCO_wdt: initialized. heartbeat=30 sec (nowayout=0)
> 
> But the patch was not merged upstream.
> 
> When testing the hardware watchdog, it doesn't trigger a reset even
> after triggering a sysrq crash.
> The system freezes and there is no reboot.
> 
> I noticed the timeleft value is not getting updated.
> root@localhost:~# cat /sys/class/watchdog/watchdog0/timeleft
> 15
> root@localhost:~# cat /sys/class/watchdog/watchdog0/timeleft
> 15
> root@localhost:~# cat /sys/class/watchdog/watchdog0/timeleft
> 15
> 
> This issue has been discussed in below email thread,
> http://lkml.iu.edu/hypermail/linux/kernel/1609.1/01979.html
> https://lkml.org/lkml/2016/9/8/641
> 
> In another platform the hardware watchdog works.
> root@debian:~# dmesg | grep wdt
> [    3.901842] iTCO_wdt: Intel TCO WatchDog Timer Driver v1.11
> [    3.903415] iTCO_wdt: Found a Intel PCH TCO device (Version=4,
> TCOBASE=0x0400)
> [    3.937816] iTCO_wdt: initialized. heartbeat=30 sec (nowayout=0)
> 
> CPU and logs
> cpu: Intel(R) Core(TM) i7-6700 CPU @ 3.40GHz
> dmesg logs: http://paste.debian.net/1092845/
> lsmod output: http://paste.debian.net/1092846/
> root@debian:~# uname -a
> Linux debian 4.14.103-rt55 #1 SMP PREEMPT RT Tue Jul 23 03:53:28 UTC
> 2019 x86_64 GNU/Linux
> 
> Hardware watchdog triggers a reset and the watchdog timer counter is
> started and the value gets updated.
> root@debian:~# cat /sys/class/watchdog/watchdog0/timeleft
> 13
> root@debian:~# cat /sys/class/watchdog/watchdog0/timeleft
> 12
> root@debian:~# cat /sys/class/watchdog/watchdog0/timeleft
> 11
> root@debian:~#
> 
> My questions:
> 1. Please let me know if it is okay to take
> https://lore.kernel.org/patchwork/patch/770990/ patch or will there be
> an update to the original patch?
> 2. Could the hardware watchdog not triggering a reset (timeleft value
> not getting updated) could be a hardware problem?
> 
> Thanks.
> 
> Regards,
> Vignesh

-- 
With Best Regards,
Andy Shevchenko

Re: watchdog: iTCO_wdt: failed to load

[Vignesh Raman]

Hi Guenter,

On Wed, Jul 24, 2019 at 6:57 PM Guenter Roeck <linux@roeck-us.net> wrote:
>
> What is the output of /proc/iomem, what PCI devices does it have, and what are
> the ACPI devices ? Reason for asking is that I don't immediately see the ACPI
> or PCI devices associated with above patch in your dmesg. If not, the patch
> might actually cause the watchdog in your system not to work.

/proc/iomem  - http://paste.debian.net/1092996/
PCI devices  - http://paste.debian.net/1092997/
ACPI devices - http://paste.debian.net/1093003/

> Also, did this ever work in your system ? If it did work, did the failure
> start after a kernel update or after a BIOS update ?

The hardware watchdog has never worked in the system. Without the
patch https://lore.kernel.org/patchwork/patch/770990/ the probe error
is seen,
[    3.828440] iTCO_wdt iTCO_wdt: can't request region for resource
[mem 0x00c5fffc-0x00c5ffff]
[    3.828803] iTCO_wdt: probe of iTCO_wdt failed with error -16

With the patch the module loads fine, but the hardware watchdog does not work.

Regards,
Vignesh

Re: watchdog: iTCO_wdt: failed to load

[Guenter Roeck]

On 7/24/19 10:12 PM, Vignesh Raman wrote:
> Hi Guenter,
> 
> On Wed, Jul 24, 2019 at 6:57 PM Guenter Roeck <linux@roeck-us.net> wrote:
>>
>> What is the output of /proc/iomem, what PCI devices does it have, and what are
>> the ACPI devices ? Reason for asking is that I don't immediately see the ACPI
>> or PCI devices associated with above patch in your dmesg. If not, the patch
>> might actually cause the watchdog in your system not to work.
> 
> /proc/iomem  - http://paste.debian.net/1092996/

Here is your problem:

00100000-5e431017 : System RAM
   00c5fffc-00c5ffff : iTCO_wdt

The address range requested by the watchdog is marked by the BIOS as system memory.

The patch referenced below tries to address a different problem and does not apply
to your system.

Guenter

> PCI devices  - http://paste.debian.net/1092997/
> ACPI devices - http://paste.debian.net/1093003/
> 
>> Also, did this ever work in your system ? If it did work, did the failure
>> start after a kernel update or after a BIOS update ?
> 
> The hardware watchdog has never worked in the system. Without the
> patch https://lore.kernel.org/patchwork/patch/770990/ the probe error
> is seen,
> [    3.828440] iTCO_wdt iTCO_wdt: can't request region for resource
> [mem 0x00c5fffc-0x00c5ffff]
> [    3.828803] iTCO_wdt: probe of iTCO_wdt failed with error -16
> 
> With the patch the module loads fine, but the hardware watchdog does not work.
> 
> Regards,
> Vignesh
> 

Re: watchdog: iTCO_wdt: failed to load

[Vignesh Raman]

Hi Guenter,

On Thu, Jul 25, 2019 at 6:40 PM Guenter Roeck <linux@roeck-us.net> wrote:
>
> Here is your problem:
>
> 00100000-5e431017 : System RAM
>    00c5fffc-00c5ffff : iTCO_wdt
>
> The address range requested by the watchdog is marked by the BIOS as system memory.
>
> The patch referenced below tries to address a different problem and does not apply
> to your system.
>

Thanks for catching that. In working case the output of /proc/iomem is,

fd000000-fe7fffff : PCI Bus 0000:00
  fd000000-fdabffff : pnp 00:09
  fdac0000-fdacffff : pnp 00:0b
  fdad0000-fdadffff : pnp 00:09
  fdae0000-fdaeffff : pnp 00:0b
  fdaf0000-fdafffff : pnp 00:0b
  fdb00000-fdffffff : pnp 00:09
    fdc6000c-fdc6000f : iTCO_wdt
      fdc6000c-fdc6000f : iTCO_wdt

This explains why it works in another system (i7-6700).

Regards,
Vignesh

Re: watchdog: iTCO_wdt: failed to load

[Guenter Roeck]

On 7/31/19 9:18 PM, Vignesh Raman wrote:
> Hi Guenter,
> 
> On Thu, Jul 25, 2019 at 6:40 PM Guenter Roeck <linux@roeck-us.net> wrote:
>>
>> Here is your problem:
>>
>> 00100000-5e431017 : System RAM
>>     00c5fffc-00c5ffff : iTCO_wdt
>>
>> The address range requested by the watchdog is marked by the BIOS as system memory.
>>
>> The patch referenced below tries to address a different problem and does not apply
>> to your system.
>>
> 
> Thanks for catching that. In working case the output of /proc/iomem is,
> 

Wasn't me. Someone else had the same or a similar problem, and Jean Delvare
noticed the overlap with system memory.

Guenter

> fd000000-fe7fffff : PCI Bus 0000:00
>    fd000000-fdabffff : pnp 00:09
>    fdac0000-fdacffff : pnp 00:0b
>    fdad0000-fdadffff : pnp 00:09
>    fdae0000-fdaeffff : pnp 00:0b
>    fdaf0000-fdafffff : pnp 00:0b
>    fdb00000-fdffffff : pnp 00:09
>      fdc6000c-fdc6000f : iTCO_wdt
>        fdc6000c-fdc6000f : iTCO_wdt
> 
> This explains why it works in another system (i7-6700).
> 
> Regards,
> Vignesh
>