[U-Boot] [PATCH] ARM: keystone: Pass SPI MTD partition table via kernel command line

[Vignesh R <vigneshr@ti.com>]

On 3/16/2017 2:18 AM, Tom Rini wrote:
> On Tue, Mar 14, 2017 at 05:11:06PM +0530, Vignesh R wrote:
>> [...]
>>
>> On Friday 10 March 2017 11:32 PM, Tom Rini wrote:
>>>> Yes, I agree that initial DT layout of 512K may not have been good
>>>> design, but it would be good to have an agreeable way of fixing up MTD
>>>> partitions when there is overflow. So, is fdt_fixup_mtdparts() preferred
>>>> approach?
>>> You make a good point about fdt_fixup_mtdparts() being non-trivial to
>>> have happen correctly in all cases above, so OK, lets put that aside.
>>> I'll also accept that previous best wisdom of not shoving tons of stuff
>>> into the cmdline, rather than passing it in the device tree, isn't
>>> correct anymore.
>>>
>>> But the big, un-tackled problem is that the old DT layout is failing
>>> because we're constantly increasing the number of full linux DTB files
>>> we're including in an image and thus increasing the size of our blob
>>> every time.  We need to stop and think and maybe design things
>>> differently.  Perhaps it's time for more platforms to have a spot on
>>> their storage where the DT is supposed to be, and we only use a fall
>>> back one that's included in U-Boot if it's not found?  Franklin already
>>> posted a patch to have something kind-of similar be able to happen
>>> (which is to say, go from a generic DTB to the correct-for-the-HW one).
>> I agree that DTB files are making u-boot image bulky. But it does not
>> seem to be problem due to addition of DT alone. For example SPI boot
>> image on K2 platform is two stage SPL+U-Boot combined into one single
>> image u-boot-spi.gph which is about 555K. General boot image u-boot.bin
>> is about 491K and u-boot-nodtb.bin is 432K. So even w/o dtbs SPI image
>> may overflow and its because of new code/framework changes.
> Which platform exactly?  I don't see anything today that's quite that
> large.  
Sorry, I had few debug prints enabled when I collected the stats. On
vanila u-boot for k2hk here are the stats:
u-boot-spi.gph 512K
u-boot.bin 448K
u-boot-nodtb.bin 420K

Still at least for k2 platforms DTBs alone are not to be blamed for
image size increase.
> And can we not move towards the "normal" method of SPL loading
> the u-boot.img (or FIT) from?  I guess the current architecture here is
> confusing me.
This has been same for all k2 platforms. I guess we have single image so
that user don't have to bother flashing multiple images  for spi boot
given the fact that all other boot modes have single image.
> Regardless, I still see the DT problem as the bigger one long term, and
> dra7xx shows that.  And I agree we need to re-size how the flash is
> partitioned.
True.
>> There is similar issue with dra7xx where flash partition for SPL is
>> running out due to addition of new code.
> The DRA flash partition is, and should be fine because we have the
> ROM-mandated limits and don't need to include U-Boot with the SPL image.
> The main U-Boot image however is growing and that is a DTB problem.  The
> difference here between -nodtb and the .img (FIT) with all of the DTBs
> is over 300KiB.  And that's mostly linear growth when compared with the
> single-DTB case.
I agree DTBs are adding to image size on DRA. But even SPL is growing
and is bound to exceed its 64K limit[1].

[1] https://patchwork.kernel.org/patch/9515551/