[U-Boot] [PATCH] mmc:dcache: Cache line size aligned internal MMC buffers

[Marek Vasut <marek.vasut@gmail.com>]

On Monday, August 22, 2011 06:42:06 PM Anton Staaf wrote:
> On Mon, Aug 22, 2011 at 9:08 AM, Mike Frysinger <vapier@gentoo.org> wrote:
> > On Monday, August 22, 2011 03:29:52 Lukasz Majewski wrote:
> >> On Fri, 19 Aug 2011 11:35:50 -0400 Mike Frysinger wrote:
> >> > On Friday, August 19, 2011 11:28:18 Lukasz Majewski wrote:
> >> > > On Fri, 19 Aug 2011 09:57:10 -0400 Mike Frysinger wrote:
> >> > > > On Friday, August 19, 2011 05:25:13 Lukasz Majewski wrote:
> >> > > > also, what is the code size increase with your patch ?
> >> > > 
> >> > > Code size overhead (s5p_goni target):
> >> > > Without proposed changes: 167928 B (u-boot.bin)
> >> > > With changes: 168208 B (u-boot.bin)
> >> > > 
> >> > > Delta: 280 B
> >> > 
> >> > np if it gives significant (more than system noise) speedups.  any
> >> > details on that ?
> >> 
> >> No tests performed yet. The goal of those patches is to preserve the
> >> MMC subsystem functionality when dcache is enabled (the ext_csd[512]
> >> corruption is observed with d-cache enabled).
> > 
> > so you're papering over a bug in some controller's cache handling ?
> >  shouldnt you fix the controller in question by having it flush its
> > caches ?  aligning random buffers to make cache issues "go away" isnt
> > the right way for anything. -mike
> 
> No, this isn't something that can be fixed in the controller driver
> code.  This is a fundamental problem with buffers in U-Boot that needs
> to be resolved by aligning all buffers used for DMA.  The main problem
> is that invalidating a non-cache line aligned buffer is not a safe
> operation.  There have been a number of threads discussing this.  The
> general consensus was to make attempting to invalidate an unaligned
> buffer an error and then to clean up the unaligned buffers as we find
> them.
> 
> Lukasz, I also have been using memalign to clean up accesses in local
> patches, so you've got my vote there. I am curious as to whether we
> should provide a single block allocation API or if each subsection
> should add lazy memalign allocations to create aligned buffers when
> they are needed...

Maybe some dma_allocate_aligned() would be cool. And probably control the 
alignment with some #define CONFIG PLATFORM_ALIGNMENT_SIZE ?

Cheers
> 
> Thanks,
>     Anton
> 
> > _______________________________________________
> > U-Boot mailing list
> > U-Boot at lists.denx.de
> > http://lists.denx.de/mailman/listinfo/u-boot
> 
> _______________________________________________
> U-Boot mailing list
> U-Boot at lists.denx.de
> http://lists.denx.de/mailman/listinfo/u-boot