After yesterdays post I couldn't help but look into removing RTC again.
It's actually quite easy to get rid of it, just lie!
The boot1 pseudo-code on WiiUBrew is a little hard to understand but here's what I think is happening:
The way the IOSU knows what to do (ECO, normal, reboot from vWii/HAI, etc) is a thing called "PRSH". I think boot1 does it instead of IOSU because of vWii but im not sure.
Anyway, on a stock system boot1 reads the RTC registers, specifically
RTC_CONTROL0/
RTC_CONTROL1 and stores the flags to a field called
boot_state in the
boot_info struct in PRSH, then writes it to MEM2. It also reads the SMC but it's not a problem for us.
On a system with isfshax this is mostly the same, with the difference that we have early code exec. I'm not exactly sure when in boot1 isfshax is triggered though. minute_minute being a boot1 replacement for de_Fuse does identical to boot1: reads the registers, sets the flags, writes PRSH to MEM2...
The problem is when the RTC is removed, basically every flag is set and the console thinks it's in standby mode.
But we can change minute to our hearts content!
So we lie about what flags are actually set after the initial read. minute_minute still sees the true flags but IOSU doesn't.
(ignore the function name I wasn't expecting it to work)
What this is doing is overwriting the
boot_state field with what we want.
PON_SMC probably isn't needed but it works and I'm scared to change it.
I call it just before launching IOSU:
I've published the edited code on Github:
https://github.com/Lazr1026/minute_minute
Now for the hardware end of this:
RTC does PoR, so the first thing to knock out is that. I just used a U10 from a Wii board, but anything similar to it should be fine.
Obviously it would be on the inside of the trim line on a trim, but for testing this was a convenient spot to put it.
Now the RTC can be removed!
Everything was done on stock boards with stock regulators, so a few things are required to get everything functional. On a trim these shouldn't be a worry because either they are done anyway or not relevant.
Tying the PFET gate to GND is necessary. The cut here looks ugly and I'm not even sure if breaking the connection is required with RTC gone lol.
3v3 is not provided to the WiFi module, so a jumper is required. Pin 3 on the 1v15 controller needs to be tied high. If you break the connection between ABLIC and the 5v controller that solder blob isn't necessary as it's pulled up elsewhere.
Enable the 1v5 controller. Should use a 10k resistor instead of a wire but I got lazy.
And it boots! I've done this on 3 consoles so far just to make sure it can be replicated.
Now I don't want to make any crazy claims like I did 2 years ago, so let me just say that I don't know the full implications that removing the SMC and RTC have. Idk if any games rely on the RTC being present. In theory every Wii U game should be compatible since all of the hardware goes through CafeOS/IOSU so if they're happy everything should be, but testing each game would be required.
I also want to create a little addendum to my previous post about SMC removal.
The I2C pullups (R221 and R223) need to be removed. On the original setup I cut the traces by the SoC before even trying to remove the chip so I wasn't aware they would cause issues.
Here's what shorting TP173 to 3v3 and R250 should look like:
EDIT (like 5 hours after initially posting):
The custom minute_minute is no longer required.
tldr; grounding EXI0 MISO (TP176) causes every bit in boot_state to be 0's
(this is sounding very de_Fuse-y lol)
After initially posting this I was wondering how tf the code even worked and investigate why it wasn't working in the past, so I looked in my past DMs with Yveltal when we were trying to get it working in minute.
In one of the screenshots had a really weird
boot_state (
ffe20000) and thought "hm did I connect the pico to EXI then?". For reference, the
boot_state I was getting here was
07e00000.
Uh, no. I only checked the pico setup pic after I wired up EXI to my current test boards pico but EXI was NOT wired.
I noticed something strange here though, the
boot_state was all 0s? I guess the pico was holding MISO low? I was under the impression it was holding it high.
This planted a theory in my head: Surely just grounding TP176 wouldn't just work, right..? right....?
With TP176 tied to ground, it booted (Well in this pic its R65 but it's still MISO).
I even
double checked this and replaced my custom minute_minute with one straight off the releases on Github, and it still booted.
I still don't know why the code worked this time. Maybe something in minute changed? Idk. I guess I don't have to bother trying to figure it out though.