Re: [PATCH 0/9] Some OMAP2+ legacy hwmod data cleanup

["H. Nikolaus Schaller" <hns@goldelico.com>]

Hi,

> Am 13.02.2018 um 16:46 schrieb Tony Lindgren <tony@atomide.com>:
> 
> * Suman Anna <s-anna@ti.com> [180213 01:33]:
>> Hi Tony,
>> 
>> The following series cleans up some legacy dev_attr data associated
>> with OMAP DMTimer, I2C, McBSP, McSPI and GPIO modules and removes
>> roughly ~500 lines. Patches are based on v4.16-rc1. I have build tested
>> for both omap2plus and omap1 defconfigs, and booted on a few OMAP4+
>> boards that I have on my desk.
> 
> Hey that's great :)
> 
>> Here is the patch summary:
>> 1. First two patches fixes header inclusion in omap-dmaengine.h and
>>    hsmmc-omap.h as I am getting build errors on subsequent patches when
>>    I removed some of the platform_data header files.
>> 2. Patches 3 through 7 remove the .dev_attr data for the above modules
>>    from various hwmod data files and some macros from McSPI & McBSP
>>    platform data header files
>> 3. Patch 8 moves the plat-omap/include/plat/i2c.h into mach-omap1
>>    folder as there is no common i2c code in plat-omap folder anymore.
>> 4. Last patch is a minor cleanup of gpio headers from couple of files.
>> 
>> Some important notes/comments:
>> - The DMTimer patch cleanup will conflict with Keerthy's current dmtimer->
>>   clocksource series, but ideally he wouldn't need to convert the plat/dmtimer.h
>>   into a clocksource header with the removal of the plat/dmtimer.h inclusion
>>   from the hwmod data files.
> 
> OK. It seems that we should do the clean up first if it just means
> leaving out unused header files.
> 
>> - I found some gaps w.r.t DMTimers on DM814x/DM816x platforms, and haven't
>>   fixed them. Both seems to have DMTimer1 through 8, but only DMTimers 1, 2
>>   and 3 are defined in dm814x.dtsi, while dm816x.dtsi is missing DMTimer8.
>>   DMTimer3 hwmod is not in the dm814x hwmod list, while DMTimer8 is left out
>>   altogther. The timer capabilities are also different between the hwmod
>>   data files (lists always-on) which is missing from corresponding dts files.
>>   Not sure which one is correct.
> 
> OK we should fix them. The capabilities can be seen by looking at
> the hardware and clockdomains. The always-on timers are of sysc
> type 1 (ti,sysc-omap2-timer) and in a always on power domain. The
> others are of sysc type 2 (ti,sysc-omap4-timer). We should get
> warnings for those from ti,sysc driver eventually if they don't
> match :)
> 
>> This leaves the hsmmc-omap and omap-dma pieces still using the .dev_attr. The
>> former is only used by one OMAP3 board, so can be cleaned up if we limit the
>> init code just for that one board.
> 
> I think Nikolaus had patches coming for removing the last user
> of mmc pdata?

Well, the idea was to convert the wl1251 driver to DT (like e.g. wl183x)
and get rid of the pdata. But I didn't find any time to work on this,
especially since the driver itself is broken for a while (I think up
to 4.11 the wl1251 was found and working, after that it wasn't working
and after 4.14 or so it wasn't even found any more). Probably some
simple things, but as usual time consuming to dig into.

> 
> I think all we need to do is have pandora_wl1251_init_card
> initialized in the MMC driver based on a compatible value..

Yes, looks so.

BR and thanks,
Nikolaus