Device_usb_oh1_57e_305 and Jit question devs only :)

memberTwo · September 6th, 2008

dev: First of all, impressive work

I'm investigating the game<->wiimote and some other jit's stuff for my own knowledge and fun but if I could help... later (who knows?

).
I'm far from the complete understanding of this emu. So, feel free to tell me if my points make sense or not.

I'm posting here because I don't want to bother irc stars live and it gives me some time to think. My english is "noisy" and ... hum... how can i say again... SLOW

But if anyone has some time to correct me:
I think I have figured out that the actual wm (wiimote) emulation wip:

1 - intends to parse wm HID commands (applicative layer) over L2CAP and over ACL, and finally over HCI (OS layers... maybe kernel). I might have misundertood the boring bluetooth spec here, so go ahead: yell at me

2 - the latest masken commit (r438) "seems" (I mean:I dont know where he finds those hci sequences. Really impressive job!

) to have setup the connection and handshake with the hard coded faked-wimote ready for ACL Tx/Rx.
By the way: To masken: I think have found a tiny datalen mistake (not quite sure) in "WII_IPC_HLE_Device_usb_oh1_57e_305::SendEventConn ectionComplete()". I didn't check everywhere.

Code:

--	pConnectionComplete->EventType = 0x03;
--	pConnectionComplete->PayloadLength = 13; // should be 11
++	pConnectionComplete->EventType = HCI_EVENT_CON_COMPL;
++	pConnectionComplete->PayloadLength =

sizeof(hci_con_compl_ep);
3 - if we simulate ACL incoming with something (not sure to have the right format but it doesn't matter yet) like this:

Code:

536: 0 WIIMOTE ( 804ab358 ) : Event: SendEventReadRemoteFeatures
537: 0 WIIMOTE ( 804ab358 ) :   Connection_Handle: 0x0100
538: 0 WIIMOTE ( 804ab358 ) :   features: bc:02:04:38:08:00:00:00
539: 0 WIIMOTE ( 804ab358 ) : USB_IOCTL_CTRLMSG(0x2)
540: 0 WIIMOTE ( 804ab358 ) : HCI_EVENT_ENDPOINT: 0x133e0200
541: 0 CONSOLE ( 804ab358 ) : siim-ote send 11 bytes : 0x0B 0x00 0x13 0x00 0x20 0x00 0x00 0x02 0x00 0x00 0x9D
542: 0 WIIMOTE ( 804ab358 ) : USB_IOCTL_CTRLMSG(0x1)
543: 0 WIIMOTE ( 804ab358 ) : ACL_DATA_ENDPOINT: 0x133e0080

The ACL incoming is well ack but no func seems to read the ACL buffer. Unless, again, I have missed something with the check-memory-read-breakpoint-system.

If I'm not wrong, at least the "ACL-received-so-push-it-to-HID-Application-stuff" func or something like that is missing and is on the TODO list. Am I right?

Now... hehehe, my question list:

a - Are wiimote (usb) interrupts handled between OSEnableInterrupts and OSDisableInterrupts? If not, where? I need an entry point.

b - Is there an updated "totaldb.dsy" with wpad's API?

c - where did you get all those wii specs (WII_IPC, ...)? If I can ask of course.

d - Just a thought, maybe plugin system is almost similar but : what do you think of a system that catches wellknown PPC asm block (corresponding to a func) that redirects func's args ( Ho ho!!!? Don't know PPC asm but I didn't see lots of PUSH 'n POP

) to the corresponding PC func. Which tweaks wii regs and mem.
My point is: why not bypass lots of syscall and more by catching and redirecting wii's "standards" API on the application side? "OSEnableInterrupts", for example, should be replaced by a short efficient regs tweaker PC func... No?
I want to try that with WPAD's API because, to start with dolphin, wm2wii protocols have been more nuts breaker... so far

Problem: I don't know those API's entry points. If anybody knows where I can find a "pseudo-original" *.elf WITH HIS *.MAP file which embedded wpad's lib, I'll buy the info

e - about jit. Is somebody is actually working on grabbing 2(or 4) successive similar instructions for SSE 128b handling (I mean with 2 or 4 successive load, store and paired instructions with data successive too)? Sounds fun.

Thanks for the help.

Masken rev 446 may answer some of my questions, masken rules

Last edited by memberTwo; September 6th, 2008 at 20:22.. Reason: clarity :D

**ector** · September 6th, 2008