I couldn't agree more.
The results we got so far were by using Sarahn's kernel patches which basically enable forcing the software statemachine of OTG driver into arbitrary states. Added on top of that on my suggestion was using FORCE_HOST musb core testmode, some patches (http://talk.maemo.org/showpost.php?p...&postcount=894 http://talk.maemo.org/showpost.php?p...&postcount=908) MohammadAG applied to a particular detail of ENUM (just commenting out the power negotiation bits, which I never suggested for a valid patch but just for testing purposes), plus a sequence of echo commands which somewhat were along the lines of what's been suggested in http://talk.maemo.org/showthread.php...914#post685914, in several other posts in this thread, and in IRC.
The result was a somewhat non-reproducible fake hostmode as of http://talk.maemo.org/showthread.php...115#post696115, which is fine to verify we face no hardware obstacles for hostmode, but it still suffers from the driver not knowing about generic hostmode and thus doing all sorts of OTG stuff, like e.g. detecting VBUS.

Alas since this legendary ""[2010-06-02 18:25:11] <MohammadAG> WOHOOOOOOOOOO"" everybody is caught in a mace of trying to get this kernel as described above into a state where we can reproduce this OTG tricking, by improving/fixing Sarahn's patches, and arbitrarily shuffling the sequence and timing of issuing the "echo X >foo" commands which force the OTG statemachine from one state to another.


This isn't really helping in inplementing a clean working generic hostmode as described here: http://talk.maemo.org/showthread.php...914#post685914
We need to implement a new state for the OTG statemachine, which is basically suspending the whole statemachine so it won't leave that state by e.g peripheral disconnect, drop / rise of VBUS, or any other of the events normal OTG deals with. This state needs to be entered on "echo host >mode", then should do all that's needed to put musb core, the interfaces and their power states, and PHY into hostmode.

Then, when in this state, all OTG related stuff like e.g. the power negotiating MohammadAG needed to comment out, white/blacklist, VBUS handling, etc, all can just be skipped by a "if !STATE_GENERIC_HOSTMODE { ... }"


As I understood from some comment from HCM, he's considering to implement such a clean generic hostmode state using FORCE_HOST from scratch, along the line suggested in http://talk.maemo.org/showthread.php...914#post685914, which I appreciate much and think is the right direction to reach the goal of a flawlessly working hostmode.

All fiddling with the patched testkernel based on OTG tricking/forcing is most probably a waste of time and energy

cheers
jOERG