[U-Boot] [PATCH v2 1/3] ARM: cache: add an empty stub function for invalidate/flush dcache

[Tom Rini]

On Tue, Jun 09, 2015 at 11:54:14AM +0800, Josh Wu wrote:
> Hi, Tom
> 
> On 6/8/2015 9:20 PM, Tom Rini wrote:
> >On Mon, Jun 08, 2015 at 10:26:29AM +0200, Marek Vasut wrote:
> >>On Monday, June 08, 2015 at 04:05:04 AM, Josh Wu wrote:
> >>>Hi, Marek
> >>>
> >>>On 6/5/2015 9:18 PM, Marek Vasut wrote:
> >>>>On Friday, June 05, 2015 at 09:47:30 AM, Josh Wu wrote:
> >>>>>Since some driver like ohci, lcd used dcache functions. But some ARM
> >>>>>cpu don't implement the invalidate_dcache_range()/flush_dcache_range()
> >>>>>functions.
> >>>>>
> >>>>>To avoid compiling errors this patch adds an weak empty stub function
> >>>>>for all ARM cpu.
> >>>>>
> >>>>>And each cpu can implement its own implementation. If not implemented
> >>>>>by default it will use an empty function.
> >>>>>
> >>>>>Signed-off-by: Josh Wu <josh.wu at atmel.com>
> >>>>>---
> >>>>>
> >>>>>Changes in v2: new added
> >>>>>
> >>>>>   arch/arm/lib/cache.c | 9 +++++++++
> >>>>>   1 file changed, 9 insertions(+)
> >>>>So, why exactly can't this be entirely common code , but a CPU-specific
> >>>>code ? :)
> >>>Do you mean to make those empty functions as common code for all arch
> >>>cpu to use?
> >>Yes.
> >>
> >>>It seems there is no place to put common code for all arch cpu.
> >>Isn't that what the common/ directory is for ? ;-)
> >No, this is making something common for the sake of making it common
> >rather than good architecutre of the code I think.  We aren't going to
> >share real cache functions just these dummy ones.
> 
> So it is the right place: arch/arm/lib/cache.c to put the dummy
> cache function.

Yes.

> >We may re-evaluate
> >our dummy ones at some point in the future when people try and use some
> >particular combination harder, see for example the PowerPC dummy cache
> >functions we just recently dropped.  Thanks!
> 
> If this patch set are ok for you. I will sent more patches to drop
> the dummy cache functions in <arch>/cpu/ folders.

Right, that's the best palce I think, baring an existing arch having a
location for weak cache functions already (like arm does).

> Also the flush_cache() in arch/arm/lib/cache.c can be dropped and
> just call flush_dcache_range() as the ARM1136, ARM926ejs alreay
> implemented their own flush_cache().

Nope, other things fail if you do that (vpac270_nor_256 and 24 others at
least so far in my test build).

-- 
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: Digital signature
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20150609/8212e1fc/attachment.sig>