Re: watchdog timer programming

On 9/24/2024 2:46 AM, Stephane Rochoy wrote:
>
> mike tancsa <mike@sentex.net> writes:
>
>> I am trying to get
>>
>> superio0: <ITE IT8613 SuperIO (revision 0x0c)> at port 0x2e-0x2f on isa0
>> itwd0: <Watchdog Timer on ITE SuperIO> at WDT ldn 0x07 on superio0
>> itwd0: Configured for system reset on timeout
>>
>> working on FreeBSD.  The driver seems to load / attach fine, but it does
>> not want to reboot the box.  Adding a #define DIAGNOSTICS 1
>>
>> shows
>>
>> itwd0: <Watchdog Timer on ITE SuperIO> at WDT ldn 0x07 on superio0
>> itwd0: Configured for system reset on timeout
>> itwd0: setting timeout to 4
>> itwd0: setting timeout to 4
>> itwd0: setting timeout to 4
>> itwd0: setting timeout to 4
>> itwd0: setting timeout to 4
>> itwd0: setting timeout to 4
>>
>> when I do
>>
>>  watchdogd -t 3
>>  killall -9 watchdogd
>>
>> but never a reboot :(
>>
>> Any idea how to get this hardware working ?
>
> Do you know if, at least, the pre-timeout is working?

Hi,

     How do I check that ?


>
> Glancing at the code, this chip can be configured to use a specific 
> IRQ (its
> default one is 0x40) and have an "NMI mode" (relying on an
> explicitly configured IRQ). Maybe playing with the dev.itwd.irq and .nwi
> tunables could get us some hints.
>
> Also this code from itwd's wd_func puzzle me a bit:
>
>    superio_write(dev, 0x73, val);
>    if (superio_read(dev, 0x73) != val)
>      superio_write(dev, 0x73, val);
>
> It let me think that we either are writing twice to the same I/O
> port by mistake or the corresponding chip's register enforce a
> specific protocol to modify it. If the later, the code don't seems
> to be able to detect a "protocol error".
>
> By the way, do you have the datasheet of the ITE chip? (I know I
> must be very optimistic to ask such a question ;))

I dont have the datasheet sadly. However, they do seem to be at 
http://www.ite.com.tw/  and are publicly available according to the 
Linux driver.

I did load up ubuntu on the box, and the watchdog does seem to work 
correctly.   sysctl doesnt seem to allow for setting those values

sysctl -A dev.itwd
dev.itwd.0.%parent: superio0
dev.itwd.0.%pnpinfo: type=WDT
dev.itwd.0.%location: ldn=0x07
dev.itwd.0.%driver: itwd
dev.itwd.0.%desc: Watchdog Timer on ITE SuperIO
dev.itwd.%parent:

On linux,
modprobe it87_wdt
] it87_wdt: Chip IT8613 revision 12 initialized. timeout=60 sec 
(nowayout=0 testmode=0)

wdctl
Device:        /dev/watchdog0
Identity:      IT87 WDT [version 1]
Timeout:       60 seconds
Pre-timeout:    0 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

The code is at 
https://github.com/torvalds/linux/blob/master/drivers/watchdog/it87_wdt.c 
if that helps shed any light.


i tried defining the irq and playing with the nmi at 1 or 0, and still 
no luck. e.g. setting it to 0x05 and nmi 0, kldload (with boot verbose) 
and then starting up watchdogd -t 8... wait 9 seconds,  and then killall 
-9 watchdogd, still does not reboot the box

itwd0: <Watchdog Timer on ITE SuperIO> at WDT ldn 0x07 on superio0
ioapic0: routing intpin 5 (ISA IRQ 5) to lapic 0 vector 56
itwd0: Using IRQ5 to signal timeout
itwd0: setting timeout to 8
itwd0: setting timeout to 8
itwd0: setting timeout to 8
itwd0: setting timeout to 8
itwd0: setting timeout to 8
itwd0: setting timeout to 8
itwd0: got interrupt, wdt status = 1

Same with irq at 0x0f with and without nmi=0 or 1.

itwd0: <Watchdog Timer on ITE SuperIO> at WDT ldn 0x07 on superio0
itwd0: Using IRQ15 to signal timeout
itwd0: setting timeout to 16
itwd0: setting timeout to 16
itwd0: setting timeout to 16
itwd0: setting timeout to 16
itwd0: setting timeout to 4
itwd0: setting timeout to 4
itwd0: setting timeout to 4
itwd0: setting timeout to 4
itwd0: setting timeout to 4
itwd0: setting timeout to 4
itwd0: setting timeout to 4
itwd0: setting timeout to 4


     ---Mike




>
> Regards,