Messages in this thread | | | Date | Thu, 4 Aug 2011 22:04:39 +0200 | From | Wim Van Sebroeck <> | Subject | Re: [RFC PATCH v2] watchdog: ep93xx: Use the WatchDog Timer Driver Core. |
| |
Hi All,
> > BTW, now that the driver uses watchdog core should we also select > > WATCHDOG_CORE in the Kconfig? If it is not enabled we get: > > > > drivers/built-in.o: In function `ep93xx_wdt_init': > > clkdev.c:(.init.text+0x2bd8): undefined reference to > > `watchdog_register_device' > > make: *** [.tmp_vmlinux1] Error 1 > > Of course. > > I wasn't sure how Wim wanted to handle this. I see two ways of doing it > in the Kconfig. > > 1) Each driver config option has a select WATCHDOG_CORE if it is required. > > Pro: all the watchdog drivers (that are available) are displayed > Con: each driver needs to add the select if needed > Side-effect: the WATCHDOG_CORE option could (should) be a hidden option > > 2) Separate the watchdog drivers into groups that need the WATCHDOG_CORE > and those that don't then put a if WATCHDOG_CORE/endif around the ones > that need it. > > Pro: keeps the Kconfig clean > Con: the drivers that use WATCHDOG_CORE are not displayed unless the > option is selected > Con: defconfigs probably will need to be updated to select WATCHDOG_CORE > > Wim, do you have any opinion on this?
I allready discussed this with Wolfram a few weeks ago. Since we don't want defconfigs to change, the best option is to: 1) use 'select WATCHDOG_CORE' during conversion of the drivers. 2) remove CONFIG_WATCHDOG_CORE and replace it by CONFIG_WATCHDOG in a later stage so that the core API becomes the user-space layer for all watchdog device drivers.This so that we can support multiple watchdogs in the future and to make sure that we keep having a consistent user-space interface.
Kind regards, Wim.
| |