On Fri, 2014-05-30 at 11:00 +0200, Hans de Goede wrote:
> Hi,
> 
> On 05/30/2014 10:54 AM, Ian Campbell wrote:
> > On Thu, 2014-05-29 at 23:51 +0200, Hans de Goede wrote:
> >> The DMA code in sunxi_mmc.c is broken. mmc_trans_data_by_dma() allocates 
> >> the
> >> dma descriptors on the stack, and then exits while the dma transfer is in
> >> progress, so the dma engine is reading stack memory which at that point may
> >> be re-used. So far we've gotten away with this by luck, but recent u-boot
> >> changes have shifted the stack start address by 16 bytes, which combined
> >> with dma alignment now exposes this problem.
> >>
> >> Since we end up just busy waiting for the dma engine anyway, this commit
> >> fixes things by simply removing the dma code, resulting in smaller bug-free
> >> code.
> > 
> > Should/couldn't this go to uboot upstream, which has the mmc stuff now? 
> 
> Yes, I'm literary minutes away from sending a patch-set upstream which
> includes this now. Actually trying to get sun4i / sun5i to work with
> the latest upstream u-boot-arm/master code is what made me hit this bug
> (and cost me a day to debug).

Great, thanks! (and see it now too)

> 
> Regards,
> 
> Hans
> 


-- 
You received this message because you are subscribed to the Google Groups 
"linux-sunxi" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
For more options, visit https://groups.google.com/d/optout.

Reply via email to