Open exander77 opened 3 years ago
I don't think the sensing pin is connected, but I don't know for sure. I assume if they connected it they would have, like, done something with it...
I think USB-C sensing might be more possible. Earlier in the WeChat thread, they said that USB-C was controlled by TPS65987D. It has a SMBus interface (three, in fact!), but I'm not sure if that is connected to the EC or not. There are also a handful of GPIOs, and I'm not sure if and how they are connected, either. On I2C1, the address should be 7'b0100xxx (xxx depending on pin strap); on I2C2, the address should be 7'b0111000 (assuming it hasn't been changed by firmware in flash). It's possible to ask the TPS65987D not just whether it's powered from USB-C, but what the power contract it has negotiated is.
Looking in more detail at strings in the EC, some of the few strings that are there are related to USB-C. For instance, there's a string at 2bd68, "TUSB546Set Err:", which seems to allude to a USB / DP1.4 mux, a lower speed version of TUSB1046 shown in figure 9-6 of the TPS65987 datasheet. And there's a function at 22736 that seems to be called in response to a parameter read from the TPS65987, according to the strings; that callback is set up in what I call do_computed_smbus_query_maybe
in what I call the computed_bat_query
. (I've since renamed that to usbc_smbus_query
, but how that actually gets called is a really deep mystery to me.)
So... I think the EC is theoretically capable of doing something to talk to the USB-C PD chip. What it does, I... could not tell you. I dumped a bunch of the EC RAM associated with the USB-C PD code. All of it seems to be zeroes, which seems to be an indication that the code is never hit.
I added a new Ghidra project dump in 1c1aafb with my latest reversing work on that.
@jwise @xy-tech Already confirmed that the sensing pin is not connected. I would most likely need to connect it to some EC GPIO if I really wanted this to work.
I think sensing on USB-C and fixed value on the barrel would be a great compromise.
I dumped EC RAM on USB-C and barrel, didn't see anything interesting yet.
I think the next step is probably to manually trigger a bat_query
to read some SMBus parameter on the TPS65987 and see if we can successfully do that or not. I'll need to refresh my memory on how the bat_query
logic actually works, and when it's safe to trigger one. Maybe in a food coma this evening...
Looking into this. Doesn't it mean that if enable_disable_charger(1,0)
is called by usbc_smbus_query
it is charging from USB-C? If that is so, we can pass enable_disable_charger(2,0)
to differentiate between normal and USB-C charging.
We could then either statically set input current for USB-C and barrel or perform negotiation on SMBus etc.
Making some progress. I have a patch that I haven't tested yet for the smbus routines that adds a 16byte read opcode. And I wrote an atomic patchloader routine that can be used to safely jam larger patches into memory. Time to start really crashing my laptop tomorrow :)
Let me know if you guys need a guinea pig!
@xy-tech Yeah, We are ready to crash our ECs.
Well, good news: my I2C code seems to work. And the atomic patchloader seems to work. And I wrote a script to scan all four I2C interfaces on the EC. And I found some I2C peripherals that I don't know!
The bad news ... neither of the other I2C peripherals that I found seem to be the USB-C controller.
So I spent a lot of hours and learned a lot, but I think this might be a dead-end, unless we get some news from Shenzhen. @xy-tech I posted in the wechat thread with a question for the ODM about how the USB-C controller is hooked up. Hopefully when they wake up they will see it and be able to find an answer for me.
Charging amperage should be set based on the adapter supplied:
It has a sensing pin, but I don't know if it is connected:
It should use 1.5-2A for 65W, 3-3.5A for 90W and above.