I'll be following with two patches; one to implement a serprog_map that just returns the phys_addr for all valid cases and reports a warning if a mapping is attempted that is not in the 24-bit address space of serprog* and does what the current code would do - return NULL.
* effectively being the upper 16MB of the 4GB address space of x86.
The other one is to fix two stupid serial buffer usage checks that used to check if we havent already gone over by the amount the next message would be (ugh...), also has some opbuf check robusting.
Given the closeness of 0.9.8 i thought it'd be nice to suggest these in for that, although i'm not holding my breath.
The effect of these two should be localized to serprog parallel/lpc/fwh - which doesnt seem to be getting much use given that lpc/fwh is currently broken - also note that current SPI in-tree effectively doesn't use opbuf much - only for delays, and does all operations synchronously - thus serbuf checks dont affect it either...
Thus i think these are safe to include in that nobody should be worse off _even if this somehow broke something_, which i think it doesn't.
(Note: I'm trying to thread these, but ... gmail, so i dunno what how this will go.)