Change in coreboot[master]: soc/mediatek/mt8195: Reserve memory to store PCIe timestamp

[Hung-Te Lin (Code Review)]

Attention is currently required from: Paul Menzel, Rex-BC Chen, Angel Pons, Yu-Ping Wu, Jianjun Wang.
Hung-Te Lin has posted comments on this change. ( https://review.coreboot.org/c/coreboot/+/62934 )

Change subject: soc/mediatek/mt8195: Reserve memory to store PCIe timestamp
......................................................................

Patch Set 1:

(1 comment)

Patchset:

PS1: 
In comparison to a single region with 8 bytes only, I'd rather make it a generic feature for passing data across sessions, especially for early init. This can be useful if someday we have to build up a general framework for eMMC, NVMe, and UFS.

What about

struct mtk_early_init {
   u64 pcie_timestamp;
 } *mtk_early_init;

And REGION(MTK_EARLY_INIT)

In bootblock, initialize that to all empty;

struct mtk_early_init *find_early_init() {
   static_assert(sizeof(struct mtk_early_init) <= REGION_SIZE(mtk_early_init));
   return _mtk_early_init;
 }

void reset_early_init() {
  struct mtk_early_init *p = find_early_init();
  if (!p)
     return;
  memset(mtk_early_init, 0, sizeof(*mtk_early_init));
 }

There can even be a general early_init.c in common for reset_early_init(); , in pcie_preinit do find_early_init() then update pcie_timestamp to a timestamp, and in NVMe driver, find_early_init(), check if pcie_timestamp is non-zero (if zero, sleep enough time assuming the pre_init was never done), and sleep() by calculating pcie_timestamp.