[U-Boot] mmc changes in v2014.04-rc3 cause problems when using mmc in SPL

[U-Boot] mmc changes in v2014.04-rc3 cause problems when using mmc in SPL

[Hans de Goede]

Hi,

While merging v2014.04-rc3 into our sunxi u-boot branch, I hit a problem
where mmc would no longer work in the SPL.

The problem is that the new mmc_create call forces use of HEAP memory,
while in the SPL there may be no HEAP.

For now I've created a small 2048 bytes HEAP in our SPL, but if possible
I would like to get rid of the need for a HEAP though, as we only have
24K of RAM for our SPL.

Regards,

Hans

[U-Boot] mmc changes in v2014.04-rc3 cause problems when using mmc in SPL

[Tom Rini]

On Wed, Apr 09, 2014 at 12:05:04AM +0200, Hans de Goede wrote:

> Hi,
> 
> While merging v2014.04-rc3 into our sunxi u-boot branch, I hit a problem
> where mmc would no longer work in the SPL.
> 
> The problem is that the new mmc_create call forces use of HEAP memory,
> while in the SPL there may be no HEAP.
> 
> For now I've created a small 2048 bytes HEAP in our SPL, but if possible
> I would like to get rid of the need for a HEAP though, as we only have
> 24K of RAM for our SPL.

24K should still be enough to have a small malloc pool and the rest of
SPL, especially since it sounds like you already didn't have FAT support
in there (which has required malloc for ages and is why Pantelis didn't
spot this as a problem I think).  How close to the limit are we today in
the sunxi platforms?  Thanks!

-- 
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 836 bytes
Desc: Digital signature
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20140408/c664ad1b/attachment.pgp>

[U-Boot] mmc changes in v2014.04-rc3 cause problems when using mmc in SPL

[Hans de Goede]

Hi,

On 04/09/2014 01:07 AM, Tom Rini wrote:
> On Wed, Apr 09, 2014 at 12:05:04AM +0200, Hans de Goede wrote:
> 
>> Hi,
>>
>> While merging v2014.04-rc3 into our sunxi u-boot branch, I hit a problem
>> where mmc would no longer work in the SPL.
>>
>> The problem is that the new mmc_create call forces use of HEAP memory,
>> while in the SPL there may be no HEAP.
>>
>> For now I've created a small 2048 bytes HEAP in our SPL, but if possible
>> I would like to get rid of the need for a HEAP though, as we only have
>> 24K of RAM for our SPL.
> 
> 24K should still be enough to have a small malloc pool and the rest of
> SPL, especially since it sounds like you already didn't have FAT support
> in there (which has required malloc for ages and is why Pantelis didn't
> spot this as a problem I think).  How close to the limit are we today in
> the sunxi platforms?

We're close to the limit, but in the mean time I've realized that we don't
need the HEAP until after we've brought up the DRAM (and the base SPL code
does not init the HEAP until after DRAM init, and once we've DRAM we no
longer have any space constraints. So things are fine as is. Sorry for the
noise.

Regards,

Hans