[U-Boot] [PATCH] ARM1136: add cache flush and invalidate operations

[Marek Vasut]

Dear Stefano Babic,

> On 30/03/2012 16:35, Anatolij Gustschin wrote:
> > Hi Stefano,
> > 
> > On Fri, 30 Mar 2012 16:20:19 +0200
> > 
> > Stefano Babic <sbabic at denx.de> wrote:
> >> On 30/03/2012 16:02, Anatolij Gustschin wrote:
> >>> Since commit 5c1ad3e6f8ae578bbe30e09652f1531e9bc22031
> >> 
> >> Hi Antolji,
> >> 
> >>> (net: fec_mxc: allow use with cache enabled) the FEC_MXC
> >>> driver uses flush_dcache_range() and invalidate_dcache_range()
> >>> functions. This driver is also configured for ARM1136 based
> >>> 'flea3' and 'mx35pdk' boards which currently do not build
> >>> as there are no ARM1136 specific flush_dcache_range() and
> >> 
> >> The issue is known - that is one reason why I marked the cache patches
> >> for the -next. I do not know if we can run enough tests before release.
> >> 
> >> Patches for M28 / MX5 / MX6 are not part of u-boot-imx, neither are yet
> >> merged into u-boot mainline by Wolfgang. On which tree have you seen
> >> that the patch was already merged ?
> > 
> > I pulled u-boot-arm.git master for build tests and see this
> > change on the FEC driver in resulting tree.
> 
> However, Albert has sent a report
>  http://www.mail-archive.com/u-boot@lists.denx.de/msg80566.html
> 
> a none of these boards was broken. But I see now that other boards are
> affected (the mx28evk does not compile due to missing CONFIG_APBH_DMA).

Fabio, can you fix please? This is trivial.

> Albert, are these patches part of your pull-request to Wolfgang ?

I believe the pullRQ isn't cooked yet. The fix for this issue right now would be 
to merge a patch that implements blank dcache-management functions for arm1136 
-- like is in AG's patch. So I'm all for merging AG's patch into AA's tree.

It's a good thing this stirred a wave of response including patches. We now know 
very well which boards are maintained ;-)

Also, once any such breaking patch lands into mainline, we'll know in 
_less_than_24_hours_ that something got broken. (this is handled by DENX CI 
machine).

Finally, we can't really run physical (HW) tests indeed, but did we ever run 
physical tests with each and every patch? (and to conclude this -- these patches 
were tested on M28 and MX6Q-board)

> 
> Stefano

Best regards,
Marek Vasut