Am Donnerstag, den 01.04.2010, 12:56 +0200 schrieb Carl-Daniel Hailfinger:
As I really like the rolling reflash stuff, I don't feel like ripping it out. On the other hand, you are right. This code is broken (as the FWH unification applied it to non-uniform-sized chips) right now, and that should be fixed before release. Having it work correctly is more important than the rolling reflash stuff, so I can resubmit a version that rips out the blockwise erase/write.
It should solve all non-uniform sector stuff.
Would a patch removing the rolling reflash be accepted before 0.9.2?
If you can get it tested on one previously supported 82802ab-style chip and on one 82802ab-style chip with non-uniform sector sizes, sure.
I don't have access to a previously supported 82802ab chip (with uniform sector size). Any testers? Hint: Chips that are OK for testing are: Intel 28F004S5 (the 'S' is important here) Intel 82802AB/AC ST M50FLW040A/B ST M50FLW080A/B ST M50FW016 ST M50FW040 ST M50FW080
If I get my Thinkpad T20 to boot, I can try with the 28F004 in it. It's a non-uniform flashchip supported by this patch.
Some people might complain about a feature regression, but to be honest this code is broken for some cases right now and sometimes radical surgery is initially painful but crucial for long-term viability. We're definitely shooting for long-term support and development of flashrom, and everything which makes continued development easier will get a thumbs up from me.
OK, I get you right that I should resubmit the patch without the incremental write and with "<< shifted", and have your Ack? Of course without the bogus "i28f00x" line in chipdrivers.h
Regards, Michael Karcher