PM4 type 2 in acb

[mailwl]

this PR passes CUSA00625 - Battlefield™ Hardline through main menu

[StevenMiller123]

This PR progresses updated versions of Call of Duty®: Advanced Warfare (CUSA00803) further, though they still crash before reaching ingame.
CUSA00803 v1.24 PR.zip
CUSA00803 v1.24 Main.txt

[Randomuser8219]

shad_log.txt
Knack now gives another garbage PM4 without debug filter.

[squidbus]

Last I remember, seeing errors for PM4 like this are usually other issues with processing the command buffer and not actually unimplemented packets.

[StevenMiller123]

Type 0 case definitely feels like a hack, though I'm pretty sure type 2 is valid.
Call of Duty®: Advanced Warfare hits the type 2 error, and I think if it were a bad packet, then implementing type 2 properly wouldn't fix it.

[mailwl]

Wrong command size, return as been in first commit, now error:

[Debug] <Critical> vk_presenter.cpp:646 operator(): Assertion Failed!
Device lost during waiting for a frame

in knack, and all screen blink black. Something really broken.

Battlefield still softlock on game start

[mailwl]

@StevenMiller123, do you think, something wrong with offsets, and PM4t2 - is part (data) of other command?

[Randomuser8219]

Wrong command size, return as been in first commit, now error:

[Debug] <Critical> vk_presenter.cpp:646 operator(): Assertion Failed!
Device lost during waiting for a frame

in knack, and all screen blink black. Something really broken.

Battlefield still softlock on game start

Knack has a pretty bad race condition going on still. This PR won't fix it.

[StevenMiller123]

There are a lot of reasons why games might be triggering the PM4 type 0 unreachable, and it will need to be looked into at some point. It's pretty clear in most games that the PM4 type 0 error is caused by some race condition. No amount of properly handling type 0 packets will fix these crashes in games like Knack or The Last of Us Remastered.
The type 2 packet handling is fine, if there was an issue there then Call of Duty wouldn't work on my end.

[georgemoralis]

ok maybe we can keep type 2 for this pr?

[mailwl]

i leave the possibility to skip Type0, for developers

[squidbus]

I don't think we should have this kind of thing committed into the main branch source

[mailwl]

why? this code not really executed without editing code
and source code full of stubs, which cannot be reversed from fw ATM

[squidbus]

Because it’s clutter, and this isn’t even a stub, it’s an invalid packet. This is just hiding another issue. Hacks don’t belong in mainline.

[mailwl]

this code is NOT executing in mainline at all, it crash early on UNREACHABLE_MSG()
it's just for devs, to see, how long a game goes on w/o this crash.
After fixing reasons, why games go to this case, the code probably never reach this point (but, as it depricated, may still be used)

[squidbus]

Yes, and I think even commented out we should not be committing incorrect hacks into mainline. There’s nothing stopping someone from making this change for debugging or having a branch but I won’t approve on this as-is.