Hi Shawn,
It fixes the return value of funciont early_init_dt_scan_memory on the success return path.
[In general, the changelog should explain why you're making this change, not just re-iterate what the patch does. Does this fix a problem you were seeing?]
With regards to this specific patch - I don't think this is correct; if we return 1 here, we'll abort the of_scan_flat_dt loop after successfully parsing one memory node, whereas machines may have multiple nodes. This change will break booting on those machines.
Cheers,
Jeremy