On Sat, 15 Dec 2018, at 7:50 AM, H. Nikolaus Schaller wrote:
Hi Tony,
Am 13.12.2018 um 19:14 schrieb Tony Lindgren tony@atomide.com:
- H. Nikolaus Schaller hns@goldelico.com [181208 21:35]:
Hi Tony, I had the OMAP5 running very very hot by a kernel hang (most likely a spinlock). And I wondered if there is no safety mechanism (watchdog timer).
I did a little research and there seems to be no WDT driver for the Palmas in mainline.
There was something:
https://lore.kernel.org/patchwork/patch/367777/
but it seems it did not arrive upstream. Any reasons or problems with the driver? Or a newer version sitting in some git?
Sorry no idea. Maybe check if it's similar to the drivers/watchdog/twl4030_wdt.c?
Yes, it is. Code structure is almost the same, because it is based on twl4030 code :)
Differences:
- uses enable/disable instead of start/stop (probably older WDT core API
and easy to fix)
- different compatible list (obviously necessary)
- uses palmas_write() from include/linux/mfd/palmas.h instead of
twl_i2c_write_u8() from include/linux/mfd/twl.h
- (have not checked): probably different register offsets
- has no remove/suspend/resume handlers (probably older WDT core API and
easy to copy&paste)
But there seems to be a bug in palmas_wdt_write(). IMHO it ignores the "data" argument when passing down to palmas_write()...
So it may be that the driver did not work as proposed.
I am not sure but I think I had spotted a v10 of the patch set.
Ah, yes:
https://lore.kernel.org/patchwork/patch/367819/
There are already several review comments (including confirmation of the data argument issue) but I could not find a v11 or v12.
Note that some devices just use drivers/watchdog/omap_wdt.c that is wired to the PMIC used. So maybe the Palmas WDT was just forgotten if it was not needed?
Maybe it was simply forgotten to be completed and resubmitted.
I am not sure if Graeme is still actively working on this topic. So I put him on CC.
TImageddon ended my access to the documentation on that chip as I was contracted for it. I never saw the actual release version of the chip :-(
Graeme