On Mon, 21 Jan 2013 13:51:02 +0100, "Edgar E. Iglesias" <edgar.igles...@gmail.com> wrote: > On Sat, Jan 19, 2013 at 06:28:10PM -0400, Grant Likely wrote: > > The etraxfs Ethernet model implements quite a nice bitbanging core. The > > change splits it out into a separate .c file. There are no functional > > changes here, just movement of code and reformatting to match qemu > > coding standards. > > Hi Grant! > > It makes sense to move it out. > > A few comments: > * Maybe we could separate the phy models from the bus model/infrastructure. > I.e, the qemu_mdio parts into one file and the qemu_phy into another one. > That way we can add other phy models and keep things separated.
That makes sense too, but it should be done as a follow on patch. That keeps the factoring out as a single logical block of code. > * Not sure bitbang_ makes sense in the naming, the abstraction supports > both the bit twiddeling model with MDIO cycles but also a higher level > transactional version (mdio_write_req etc) that bypass the cycle accuracy. > The latter is useful for modeling NICs that have MDIO support in hw, e.g > like the hw/xilinx_axienet.c while beeing able to reuse the PHY models. axienet and etraxfs implementations are subtely different (as you know). I originally was going to do both, but I didn't want to mix functionality changes in with the factoring out. Doing both at the same time would have required (slight) behavour changes in one of the models, so instead I decided to do only one to keep it simple with the other being a follow-on. > I realize I should have done this split when doing the axienet model, sorry.. Can you help me out with a patch that migrates axienet to the new common code? Thanks, g.
