On 07.05.2009 18:25, FENG Yu Ning wrote:
Carl-Daniel wrote:
Until the ICH SPI driver can handle preopcodes as standalone opcodes ...
I don't like ICH's preop design either. However, I think we shall write software following the hardware design's policy. Trying to compensate defect of hardware design costs and probably not gains more.
I beg to differ. Since we want to keep as much code as possible shared for every flasher (SPI master), working around design bugs is more efficient than changing flashrom to conform to every weird piece of hardware out there.
ICH SPI tries to be "smarter" than other chipsets. We may consider transfer control to ICH SPI driver as early as possible, leaving more decision to be done by the ICH SPI driver instead of the generic SPI driver layer.
Right now, drivers can override many SPI generic operations as needed. If the abstraction is now powerful enough, I'm all for improving it. Adding support for external programmers will certainly entail new challenges.
Regards, Carl-Daniel