Hi,
On 18 January 2016 at 02:39, Sjoerd Simons sjoerd.simons@collabora.co.uk wrote:
On Thu, 2016-01-14 at 08:51 -0700, Simon Glass wrote:
Hi Eddie,
On 14 January 2016 at 05:47, Eddie Cai eddie.cai.kernel@gmail.com wrote:
Hi Simon
I think the best way to reduce SPL size is to jump back to boot rom. Which don't require eMMC, SD card driver in SPL any more. Even clock and pinctrl driver is not required. All we need is DDR initialization. We should do as little thing as possible in SPL and let U-boot take care others.
Is it possible to use the ROM to load U-Boot and then get control back before jumping to it? If so, then I agree this would be best.
Ideally we just need a routine we can call which can load from eMMC/SD.
Incidentally I've been playing with this a bit over the last week. Mostly because I wanted to load u-boot over USB OTG to make it simpler to quickly test things (e.g. to test my networking series more easily).
I've got some work in progress code to jump back to the bootrom after the SPL which allows me to load u-boot via USB OTG as well, which is great start.
Unfortunately it seems that loading a kernel after that will cause the kernel to wedge during early boot. Initial analysis shows it doesn't get interrupts from the architecture timer, but it might well be interrupts don't work at all...
Eddie, any ideas what the maskrom code might do different with the interrupt setup if loading things over OTG vs. from MMC?
It would certainly be great if we can use the boot ROM code. This is what we do on exynos and it makes a big difference to the SPL size.
Regards, Simon