This commit adds all the changes made to the 3GX plugin loader fork of Luma3DS. The most important features are:
- Add 3GX plugin loader support. New service added to rosalina: plg:ldr
- Add svcControlProcess, svcControlMemoryUnsafe and improve svcMapProcessMemoryEx (breaking change)
- Allow applications to override certain configurations depending on their needs:
- Disable core2 thread redirection
- Disable game patching for the next app
- Force New 3DS speedup
- Force next application in a specific memory mode
- Block the opening of the Rosalina menu
- Add GDB commands to list all process handles and catch all SVC (latter is for IDA Pro as gdb client supports it)
- Other changes necessary for plugins to work properly. Please check changed files in this PR for more details.
---------
Co-authored-by: PabloMK7 <hackyglitch@gmail.com>
Co-authored-by: Nanquitas <nath.doidi@gmail.com>
Co-authored-by: TuxSH <1922548+TuxSH@users.noreply.github.com>
And also move the location of IPS/BPS patches for (non-KIP) sysmodules
to /luma/sysmodules/<titleId>.<ips/bps> and only require "Load ext
firm/modules" option for it.
This is a leftover from the Gateway era that has long lost its utility
and that unnecessarily complicated the code base.
Please just load an external FIRM from SD card if you need something
similar.
Also refactor the multi-EmuNAND code and CTRNAND mounting code.
This was a long-standing bug since Luma3DS v8.0.
If you changed the homebrew title and didn't immediately reboot, then
the ExHeader during termination would not match the ExHeader that was
was used for loading the process, and thus sysmodule process refcounts
would get all messed up.
The obvious solution to this is to ensure no application is running
while changing the hbldr titleID (hence the changes in custom PM).
This was quite possibly one of the cause of homebrew failing to load
when using N3DS H&S.
... even when HPs aren't inserted.
This is aimed at people using Bluetooth audio hw mods.
There are some software limitations, please read the INI/option
description.
Closes#1828.
N3DS-only option. See config.ini for details and context. This should
make very demanding games (those which already use 3x clockrate and L2C
by default, like Pokémon US/UM) between 5 to 15% faster. Not widely
tested, and can potentially break some games and other apps.
When "load external firms and modules" is enabled, Loader will load the
sysmodule from /luma/sysmodule/<titleid>.cxi (all uppercase, and with
the N3DS title ID bit if relevant) and skip patching. Note that this is
a title ID here, not a process name (unlike what we do for KIPs).
While this is aimed at enabling people to easily load replacements for
official sysmodules, you can load your own custom sysmodules that don't
correspond to anything installed. You can use gdb to do so:
set remote exec-file <tid>
run
Not tested (except for HM which is known to work).
And restrict code.bin and exheader.bin to when *both* "external firm
and modules" and "game patching" are enabled.
This commits adds an "Advanced configuration" option to screen filters,
where CCT, color gamma, contrast and "brightness" can be fine-tuned,
instead of just using a preset.
Persistence in config.ini TBD in a later commit.
- transform some error codes into "NAND damaged" or "Gamecard removed"
like official errdisp does
- fix bug in SetUserString
- add more info to logfile (/luma/errdisp.txt)
For stuff like testing PASLR, *hax2x, khc3ds, etc.
Also fix a corner-case bug when changing the 3dsx app from itself to
itself, if its TID corresponds to the default.
Add config option to autoboot into 3DS and DSi homebrew menu, without
going through Home Menu (nor launching it).
For 3DS homebrew, this requires homebrew built with libctru v2.0.0 or
later (v2.0.0 was released 2.5y ago).
We simulate a "reboot into title" to achieve this. This being said, when
launching stuff like Pokemon US/UM on O3DS, Home Menu reboots into
itself and not the game directly. This will cause Home Menu to crash if
you use this feature and configure it to use a non-default memory layout
(but if you don't, Home Menu will work just fine).
Caching exheader for applications launched through Home Menu has very
little benefit, as UnregisterProgram is called a lot, and, for 3DSX,
boot.3dsx/target 3DSX can be changed at any time.
Faithfully implement the PASLR algorithm official Loader
uses (it's not very good). Physical address space layout
randomization means that the layout of the pages in physical
memory are randomized, but doens't randomize the virtual addresses.
Also refactor some parts of our Loader impl a little more.
Let's not pretend in 2022 that it needed things from rosalina sysmodule
- it did not.
This moves 3DSX loading from Rosalina to Loader, and also removes all
the dependencies Loader had to other Luma3DS components (if kernel ext.
is missing, a default config will be used). This means that, as long as
you replace Loader to the one in here, you will be able to properly load
3DSX files.
Changes:
- hb:ldr is now hosted in loader
- hb:ldr LoadProcess, PatchExHeaderInfo, DebugNextApplicationByForce:
all removed
- fix a bug where some malformed 3DSX files were not rejected
- grant access to CONFIG11 registers to 3DSX homebrew
- move dirty homebrew chainload (when HM. isn't loaded nor loadable) to
pm
- pm:dbg (ext.) PrepareToChainloadHomebrew: removed
This is mostly a libctru bug (well, sort of).
This can only happen to built-in sysmodules, and to
processes waiting for err:f, that fail to obtain
handles through svcConnectToPort first try; and only
prior to 11.0.
Prior to fw 11.0, kernel didn't zero-initialize output
handles, and thus the output handle gets filled with
junk (leaked kernel stack data) in case of failure.
Libctru does not account for this, and closes such
handles anyway (in srvInit, errfInit, and possibly
more).
The problem is that, in our case, that garbage was
equal to 0x8000, actually a valid handle, in fact
the first handle to be created (and not closed) in
a process... a handle to KAddressArbiter.
Accidentally closing this handle resulted in one or
more KIPs spin-waiting and starving core1, resulting
in an inability to boot.
We fix this simply by replicating what recent k11 does,
in kext (for svcConnectToPort). For srvGetServiceHandle,
add two layers of safety.
