* int32-modifiers
GCN VOP3 abs/neg modifier bits always operate on the sign bit (bit 31)
regardless of instruction type. For integer operands this means:
abs = clear bit 31 (x & 0x7FFFFFFF)
neg = toggle bit 31 (x ^ 0x80000000)
* int64-modifiers
Previously GetSrc64<IR::U64> completely ignored input modifiers
for integer operands. Now unpacks to two U32s, modifies the high
dword's bit 31 (= bit 63 of the 64-bit value), and repacks.
* V_MUL_LEGACY_F32
GCN V_MUL_LEGACY_F32: if either source is zero, result is +0.0
regardless of the other operand (even NaN or Inf). Standard IEEE
multiply produces NaN for 0*Inf. The fix adds a zero-check select
before the multiply.
* fallback for IMAGE_STORE_MIP when not natively supported
* Lod should be treated as absolute, independent of sharp's base_level (judging by other implemented instructions)
* fix descriptor set layouts
* dumb error
* force fallback for testing
* treat Lod as relative to base_level
* optimization when lod index is constant
* Test
* More robust logic for storing and freeing dummy data
Anything heap allocated is invalidated when the function returns. Use malloc to allocate the string instead, and make sure to free those allocations in sceSslFreeCaCerts.
* Fix game arguments.
Tested with Crash Team Racing Nitro Fueled
* Fix the fix
This callback runs unconditionally, so only perform erase if we actually place anything in gameArgs
Now that dynamic HLE loads happen after the eboot loads, HLEs for most "preload" modules wouldn't detect if you didn't have libSceRtc dumped. This was because, while we stored all the new symbols from the HLE lib, we weren't relocating loaded modules to use these symbols.
* To implement ImageAtomicCmpSwap
...but it doesn't work, so here it shall stay.
* a fix
* Clang
* Add to MayHaveSideEffects
I missed this while digging through IR code.
* improve signal emulation
* make the sce function use the new posix ones
* ifdefing away the issues
* fix me being very tired yesterday night
* let macOS handle SIGRT signals with the native sigaction call instead of an early error return
* windows still has no clue what the fuck is going on
* the loathsome clang-formatter
* fix oact
* return the guest handler, not the host one
* Clear any existing signal mask for game threads.
* don't rely on implementation specific things
* Fix Windows support and sceKernelRaiseException bug
* Review suggestions
@kalaposfos13 suggested I push these.
---------
Co-authored-by: Stephen Miller <56742918+StevenMiller123@users.noreply.github.com>
* Replace sysmodule enums with table
Dumped this from the 12.52 module, using a script I created.
* Better documentation
* Separate from system libraries
That system folder is going to be getting quite large if I left all the sysmodule stuff in there.
* More arrays from library
* Found another preload list
Ghidra really hates decompiling libSceSysmodule, so I didn't notice this one at first.
Also documented specific versions tied to each preload list.
* Start work on implementation
* Some basic implementations
* Initial stub for module loading
Just enough to see if the logic functions appropriately.
* Clang
* sceSysmoduleLoadModule
Now I need to get sceSysmodulePreloadModuleForLibkernel done so that we don't have bugs from not loading internal LLEs.
* sceSysmoduleLoadModuleInternal
* sceSysmodulePreloadModuleForLibkernel
I have successfully broken quite literally everything.
I shall debug this tomorrow.
* Slight fix
* Maybe fix?
* Change log
Enjoy the log spam 😄
* Increased defined stub count
Now that libc and libSceLibcInternal loads later, all the auto stubs are getting consumed by it.
* sceSysmoduleUnloadModule stub
Also a couple fixes. Sysmodule does pass argc and argv to game modules, but only after loading them once to check binaries.
Shouldn't matter for the most part.
* Clang
* Less stubs
2 thousand is seemingly enough.
* sceSysmoduleLoadModuleInternalWithArg
Doesn't hurt to have, since Apex Legends calls it.
* Oops
* Oops 2
* Rename isModuleLoaded to getModuleHandle
Review comment
* Remove debug game module loads
These cases only trigger when specific sceRegMgr key values are set, and for our purposes, we can treat that case as false.
* Allow preloading to fail
For kalaposfos
* Clang
* stencil fixes hope it fixes driveclub
* revert image copy to the one that had driveclub worked
* reverted texture cache change
* some more fixes and reverts
* added logging for overlap again
Cyberpunk's issue seems to actually come from the incrementing in the loop. It wasn't clear while debugging, but the problem is that the pattern the game supplies causes match to fail when str_wild_it hits the end, and then tries iterating past end due to the loop condition.
Our pattern matching code seems broken for the case Cyberpunk triggers, but since I'm not aware of the intricacies of how real hardware behaves, best to just revert the loop condition change and instead break the loop before the broken iteration.
* Force resolver errors when not connected to network
Error values are based on real hardware testing.
sceNetResolverGetError is based on libSceNet decompilation.
* Update net_resolver.h
* Remove dead code from EqueueInternal::WaitForEvents
No longer necessary now that we avoid using small timers when falling back on equeue logic.
* Refactor type names
Might as well
* Properly define OrbisKernelEqueue as a handle
Most of the functions using an "OrbisKernelEqueue" call directly into kevent. Therefore, OrbisKernelEqueue should be a equeue handle.
* Clang
* Widen OrbisKernelEqueue type
On real hardware, it's some value that contains the handle, as opposed to just the handle itself.
* kqueue implementation
The easy part
* Hardware-accurate timer data
Needed to make kevent simpler for these uses.
* Move callback scheduling to EqueueInternal::AddEvent
kevent would become excessively bloated if I needed to deal with that in there.
* posix_kevent
kevent is a bit of a pain, for now I've implemented as much as libkernel actually uses for it's wrappers, and left error logs to skip behavior when necessary.
* Log calls
* Apple, why are you calling fstat on an equeue?
* Change thread pausing to use SIGTRMIN on UNIX
* Allow handling of the rest of the signals
* Add orbis-native signal number conversion and fix a few bugs
* ifdefing away the issues
* add check for mac for the signal that's used for thread pausing there
* Add a few more registers
* Don't break HLE memory tracking
Now, if a guest app installs a handler for SIGSEGV/SIGBUS/SIGILL, that'll be handled by keeping the original signal handler, and if we can't handle the signal ourselves (as in it didn't come from HLE memory tracking), we pass it on to the guest
* copyright 2026
* +
* Ignore enabled flag on critical log entries
This ensures critical errors (asserts and unreachables) are logged when the log file exceeds 100MB, or when logging is disabled.
* I apparently need sleep
* Fix vblank event data
* Various logical fixes for timer events
Store timer timeouts in nanoseconds now, properly handle event "replacement", avoid employing small timers when adding events to equeues, fix stored timer data
* Clang
* fixed image copy to ignore stencil aspect
* Added logging for "Encountered unresolvable image overlap with equal memory address."
* fixed overlap issues with different pitch , added more detailed logging for rest of overlap issues
* improved log error
* maybe mipmaps ?
* array layers or different mip map range
* rewrote case new image has fewer mip levels than cached image
* array with 2 layers?
* last case
* improved
* no it didn't work
* Hardcoded limit to pending flips
Real hardware has a fixed-size queue, and doesn't depend on the number of registered buffers.
While the kernel supposedly uses an array of 18 elements, my tests suggest the cap is 16 pending flips.
* Assert on trying to flip unregistered buffer
I haven't seen anything do this intentionally yet, but I do have cases where games do this unintentionally (do to unimplemented functions).
* There is a mountain of evidence suggesting that flip_arg for these functions should be a 64-bit integer.
This fixes "memory" errors in some Unity titles.
* oops
* Fix sceVideoOutGetEventData
This bug went unnoticed for a while because the selection of Unity games I had at the time didn't actually care.
This + the prior fix is needed for Unity titles.
* Improve stack clearing logic in ExecuteGuest
Added a check for fiber stacks before clearing the stack in ExecuteGuest.
That fixes Gravity Rush 2 crash on Windows.
* Refactor ExecuteGuest to simplify stack clearing logic
This enough for GR2
* Recover thread initialization in ExecuteGuest function
* Enhance null check for thread control block
* Fix condition to check tcb before clearing stack
* Don't clear events that don't need clearing
Unless the event has the clear flag, it will be returned multiple times after triggering.
* Fix event flags
As older code suggests, the PS4 kernel does append the clear flag to various event types internally. This is visible when observing the returned event data from sceKernelWaitEqueue (or kevent, if you're feeling ambitious)
Add flag is also removed from events internally.
Just a typical day of me pushing something a month ago, nobody testing/reviewing it, then finding out it's broken when that code inevitably makes it into production.
* Initial definitions
* internal__Fofind
* Libcinternal threads
fopen stores a valid pthread mutex in the FILE struct. Since this is exposed to the game/app, we need to handle this accurately.
* internal__Foprep (and various other functions called in it)
* Actual fopen implementation
At long last, an actual function I'm supposed to implement.
* fflush + compile fixes
* fseek implementation
Comes with functions fseek calls, aside from fflush which I pushed earlier.
* fread, _Frprep
Also changed some parameter names a tad to match how I named things in my decomp.
And fixed some bugs with how I was handling the weird offseted mode thing
* fclose, _Fofree
Not confident on this one, but we'll see I guess.
* Bug fixing
No more crashes at least, fread seems to be broken though.
* fopen bugfixes
Behavior now matches LLE, at least in how LLE font seems to use it.
* Fix _Frprep
Seems like everything works now?
* Logging
Probably going to need to swap lseek and read logs to debug/trace later but this is for debugging.
* Remove alignment check
Seems I must've misinterpreted some of what Ghidra spat out, since libSceNgs2 is calling with size 1, nmemb 4.
* Reduce fseek, fread logs to trace
* Clang
* twos
* Fixes
Still can't test properly, but this seems to hide system libs, which I'm pretty sure is the necessary difference here.
* Clang
* Extra export for sceKernelGetModuleInfo2
* Fix assert
Just a typical day of me pushing something a month ago, nobody testing/reviewing it, then finding out it's broken when that code inevitably makes it into production.
* Remove unreachable in posix_pthread_mutex_timedlock
It's apparently something that was added during pthreads rewrite, but the actual code for this function seems to be fully implemented?
Just a typical day of me pushing something a month ago, nobody testing/reviewing it, then finding out it's broken when that code inevitably makes it into production.
* added function parameters
* added logging
* more logging
* added error codes file
* sceNpWebApiCreateExtdPushEventFilter some re
* added np_web_api_internal
* more np_web_api_internal definations
* Initial types cleanup
* Basic library context handling.
Followed decomp closely, using standard library classes where possible to simplify code.
* Fix params to sceNpWebApiCreateContext
* Context logic fixes
* User contexts
* Clang
* sceNpWebApiVshInitialize
* Better initialization
* Request creation logic
* Some cleanup
* sceNpWebApiAbortRequest, sceNpWebApiDeleteRequest
* SendRequest functions
* formatting
* Update terminateContext and deleteUserContext
Addressing some unimplemented bits now that I have requests and user contexts here.
* Copyright
* sceNpWebApiCreateHandle, sceNpWebApiDeleteHandle, sceNpWebApiAbortHandle
also some bugfixing
* Extended push event filter
* abort handles in terminateContext
* Other push event filter types
* Register callbacks
* unregister callbacks
* oops
* One final update to deleteContext
* Logging changes
* Bug fixes
Fixes memory leaks, pretty sure these are the only places where that was an issue.
* sceNpWebApiCheckTimeout
* Handle and request timeouts
* Oops
* Push event filter parameters
Tested with Assassin's Creed Unity, seems to be correct.
* Service push event filter parameters
Tested again with Assassin's Creed Unity, seems to work fine.
Also fixed some code bugs I noticed, and removed an unnecessary part of my internal structs
* Stub implementation for createUserContextWithOnlineId
Might need a PSN check to be properly accurate, not sure.
* added sceNpWebApiGetHttpStatusCode
* opps
* opss 2
* sceNpWebApiReadData
* clang
* Fix context ids, user context ids, and request ids
Overlooked how these ids are actually calculated.
* Additional PSN checks
Seems creating any form of push event filter with an np service name fails when you're not connected to PSN.
Not sure of the actual cause yet, but given the error code, it's related to sceNpManagerIntGetUserList.
* compile fix
---------
Co-authored-by: Stephen Miller <millerste004@gmail.com>
Co-authored-by: Stephen Miller <56742918+StevenMiller123@users.noreply.github.com>
* improved sdl backend
* small cleanups
* misc
* adjustments and new definations
* cleanups
* more debuging
* rewrote sceAudioOut calls
* fixed a trace
* fixed audio3d port
* small debug fixes
* small additions
* using shared_ptr
* compile fixes
* make macOS happy
* using shared mutex
* implemented audio input backend
* fixed port construction based on decompile
* implemented partially sceAudioInGetSilentState to return correct code if mic device is null
* improved sdl volume handling
* dynamic volume update
* this one is for @UltraDaCat
---------
Co-authored-by: kalaposfos13 <153381648+kalaposfos13@users.noreply.github.com>
* Fix bug with "phys_addr_to_search" logic
By improperly updating the variable, some games would mark the same dmem area as mapped in different parts of the vma, and the dmem map wouldn't properly reflect the state of the vma's phys areas.
* Clang
* Oops
* Fixed sceCompanionUtilGetEvent stub
Previously we effectively stubbed with ORBIS_COMPANION_UTIL_INVALID_POINTER, which makes no sense and caused issues in games.
* Check for null issuer_id in libSceNpAuth's GetAuthorizationCode
Comes up in Mirror's Edge Catalyst, according to some debugging done by a community member.
Given the library didn't have any null checks for that value, this is probably allowed.
SetThreadName gets passed an std::string's c_str whose pointer gets invalidated by the assignment of g_curthread->name, resulting in broken thread names further down the line
* PngEnc hle
* format
* formatting + fix scePngEncDelete
* fix cmake + misc improvements
i think the setjmp is right according to the libpng manual, works fine from my testing
* fixes
fix an issue with how alpha was handled, and PngEncode() now properly sets the processed_height in outputInfo.
* format
* Update pngenc.cpp
* set outputInfo->processed_height during png write
i assume some games will use this for error handling
* Avoid nullptr dereference on GetSocket
Was gonna include this in my socket PR, but that got merged before I could push this.
* Lock unmap mutex in PoolExpand and Allocate
PAYDAY 2 has a rare race condition involving dmem releases.
I'm not certain this commit will fix it, but this would cause a race condition that could cause asserts like what PAYDAY 2 can hit, so I'll just pray this does the job until I can prove it doesn't.
This issue would cause memory leaks in some EA titles, also just generally makes it harder to debug stuff when the fd table is flooded with closed sockets and epolls.
* no
no
* Adjust locking strategy
Use a separate mutex for the initial error checks + GPU unmap instead of using the reader lock. Make sure all writers lock this separate mutex, and for those that don't perform GPU unmaps, lock the writer lock immediately too.
This gets around every race condition I've envisioned so far, and hopefully does the trick?
* Clang
* Always GPU unmap
GPU unmaps have logic built-in to only run on mapped areas.
Not sure if userfaultfd would work with this, but since that's already broken anyway, I'll let reviewers decide that.
Without doing this, I'd need to do an extra pass through VMAs to find what all needs to be GPU modified before I can unmap from GPU, then perform remaining unmap work. Especially for places like MapMemory, that's a lot of code bloat.
* Fixups
* Update memory.cpp
* Rename mutex
It's really just a mutex for the sole purpose of dealing with GPU unmaps, so unmap_mutex is a bit more fitting than transition_mutex
* Optimizations
Microsoft allows you to coalesce multiple free placeholders in one VirtualFreeEx call, so we can perform the VirtualFreeEx after coalescing with neighboring regions to eliminate a VirtualFreeEx call in some situations.
* Remove unnecessary VirtualProtect call
As far as I can tell, this call wastes a bunch of time, and is completely unnecessary.
With our current codebase, simply supplying prot to MapViewOfFile3 works properly.
* Properly handle file mmaps with offsets
Pretty easy fix to perform while I'm here, so I might as well include it.
* Oops
Leftover stuff from local things + clang
* Disable tracy memory tracking
Tracy's memory tracking is built around a typical malloc/free API, so each individual alloc must correspond to a free.
Moving these to address space would fix issues on Windows, but Linux/Mac would have the same issues with our current code.
Disabling VMA merging is technically a fix, but since that's hardware-accurate behavior, I'd rather not disable it.
I'm sure there's a simple solution I'm missing, but unless other devs have a better idea of how this should be handled, the best I can do is disable it so we can keep using Tracy to trace performance.
* Update address_space.cpp
* Debug logging
Should give a decent idea of how nasty these AddressSpace calls are in games that lost perf.
* test removing thread safety
Just for testing, will revert afterwards.
* Check name before merging
Fixes a regression in Apex Legends
* Revert "test removing thread safety"
This reverts commit ab897f4b1c.
* Move mutex locks before IsValidMapping calls
These aren't thread safe, this fixes a rare race condition that I ran into with Apex Legends.
* Revert "Debug logging"
This reverts commit eb2b12a46c.
* Proper VMA splitting in ProtectBytes, SetDirectMemoryType, and NameVirtualRange
Also slight optimization by eliminating AddressSpace protect calls when requested prot matches the previous prot.
Fixes a regression in God of War: Ragnarok
* Clang
* Fixes to SetDirectMemoryType logic
Fixes some regressions in Marvel's Spider-Man that occurred with my previous commits to this PR.
* Fix Genshin Impact again
* Assert on out-of-bounds protect calls
Our page tracking code is prone to causing this.
* test mutex again
This time, remember all mutex stuff
* Revert hack
I'll work on a better way to deal with mutexes in a bit, first I'm pushing up some extra fixes
* Proper logic for checked ReleaseDirectMemory, added bounds checks
Should help some games.
* Better logging for ReleaseDirectMemory errors.
* Only perform region coalescing after all unmap operations.
A small optimization for unmapping multiple regions. Since Microsoft lets you coalesce multiple placeholders at once, we can save doing any VirtualFreeEx calls for coalescing until after we unmap everything in the requested range.
* Separate VMA creation logic into a separate method, update MapFile to use it
MapFile is technically another "emulation" of MapMemory, both should follow similar logic.
To avoid duplicating code, move shared logic to a different function that both MapMemory and MapFile can call.
This fixes memory asserts in a couple of online-only apps I have.
* Clang
* Fix TryWriteBacking
This fixes a lot of regressions that got misattributed
Co-Authored-By: TheTurtle <47210458+raphaelthegreat@users.noreply.github.com>
* Fix again
Fixes device lost crashes with some games after my last commit.
* Oops
* Mutex cleanup
Avoided changing anything in MapMemory, UnmapMemory, PoolCommit, or PoolDecommit since those all need a little extra granularity to prevent GPU deadlocking.
Everything else now uses standard library locks to make things a little simpler.
* Swap MapMemory and PoolCommit to use scoped lock
GPU maps are safe, so this is fine. Unmaps are the primary issue.
---------
Co-authored-by: TheTurtle <47210458+raphaelthegreat@users.noreply.github.com>
* Initial work
* Bug fixing
deadlocks and broken unmaps
* Fix more bugs
broken memory pools
* More bug fixing
Still plenty more to fix though
* Even more bug fixing
Finally got Final Fantasy XV back to running, haven't found anymore bugs yet.
* More bugfixing
* Update memory.cpp
* Rewrite start
* Fix for oversized unmaps
* Oops
* Update address_space.cpp
* Clang
* Mac fix?
* Track VMA physical areas based on start in VMA
Allows me to simplify some logic, and should (finally) allow merging VMAs in memory code.
* Merge VMAs, fix some bugs
Finally possible thanks to address space + phys tracking changes
* Clang
* Oops
* Oops2
* Oops3
* Bugfixing
* SDK check for coalescing
Just to rule out any issues from games that wouldn't see coalescing in the first place.
* More ReleaseDirectMemory fixes
I really suck at logic some days
* Merge physical areas within VMAs
In games that perform a lot of similar mappings, you can wind up with 1000+ phys areas in one vma.
This should reduce some of the overhead that might cause.
* Hopefully fix Mac compile
Why must their uint64_t be different?
* Mac pt.2
Oops
* Introducing key_manager for storing encryption keys . Currently only trophy key is neccesary
* keep gcc happy?
* addded logging to keymanager
* revert file
* added npbind file format and rewrote part of trp file format
When walking the users of special constants which form LDS
addresses:
-ignore when a user contributes to the wrong operand of an LDS inst, for
example the data operand of WriteShared* instead of the address operand.
This can mistakenly happen due to phi nodes.
-don't use flags to stash temp info about phis, since flags may already
be in use. Use a separate map.
* added libSceAudiodec to lle modules list
* crappy float resample , use it at your own risk
* clang
* adjustments to aac
---------
Co-authored-by: Vladislav Mikhalin <mikhalinvlad@gmail.com>
* OrbisAudio3dOpenParameters struct fix
Not sure why we have the extra filler, but decomp suggests it shouldn't exist.
This fixes stack_chk_fail issues in audio3d using titles.
* Bring back filler, only copy 0x20 bytes.
The library accepts variations on struct size, with the maximum size being the 0x28 size our current struct has.
This fixes the issue without potentially breaking the struct.
* Fix memcpy
Prevent OOB read
* Run libSceRtc LLE
The more we've used our HLE, the more issues we've had with it.
While we debug these bugs, re-enabling LLE will address any regressions the swap caused.
* libSceJpegDec LLE
Needed for Trackmania until we implement HLE for this library
* libScePngEnc LLE
Needed for Minecraft until we implement HLE for this library
* Update documentation appropriately
* libSceJpegEnc LLE
By @georgemoralis's request
On real hardware, each pool commit can have multiple physical memory allocations, always allocating the lowest free physical addresses first.
Currently, since we currently only support one physical mapping per VMA, I create a separate VMA representing each physical allocation we perform.
* lseek for directories behaves correctly when final index is smaller than 0 (EINVAL)
Backported and improved dirents from QFS
Normal directory dirents update on change
* PFS moves pointer to end when last dirent is returned
* Correct entry type in PFS directory
* the bare minimum (this won't even compile on windows yet)
* well I guess this is redundant now
* Windows GetThreadName
* Move function to common/thread and add full guest name where applicable
* the loathsome clang-formatter
* do stuff first ask for opinions later
* copyright 2026
* remove unused header
* copyright 2024-2026
---------
Co-authored-by: Stephen Miller <56742918+StevenMiller123@users.noreply.github.com>
* Remove SceKernelInternalMemory mapping
Contrary to my initial beliefs, this is very much a hack.
* Unreachable for unpatched code
This will always infinitely loop, making logs extremely large.
* Update linker.cpp
* Patch stack checks done using fs:[0x28]
Additionally adds support for multiple patches per instruction, since this makes two separate patches we need to conditionally perform for mov instructions.
* Missing include
* Disable patches for Apple
Mac can use their native FS segment directly, so these patches aren't needed
* Oops
* VectorFpRound64 decode table
Also fixed definition for V_TRUNC_F64, though I doubt that would change anything important.
* V_FLOOR_F64 implementation
Used by Just Cause 4
* Oops
Never forget your 64s
* Some mprotect fixes
The biggest thing here is preventing mprotect on memory that isn't mapped in address space. This would cause exceptions before, but succeeds on real hardware.
I've also included a couple other minor fixes, mostly based around some tests I recently performed.
Note: All changes to memory pools in this PR are assumed. I have not yet tested memory pools with any of this logic, but I do at least want to prevent mprotect on pool reserved memory to avoid crashes.
* Update memory.cpp
* clang
* Avoid storing the Tcb pointer on the stack
* Just return the already stored pointer in GetTcbBase
* Replace uses of GetTcbBase with g_curthread->tcb
* copyright 2025
* sir clang offnir, the all-formatting
The bitfield in the struct is padded, which produces uninitialized memory on initialization.
To avoid modifying the struct while making our GraphicsPipelineKey struct properly hashable, set values directly instead of re-initializing.
This fixes pipeline compile spam, and the subsequent poor performance, on certain setups.
* Earlier initialization of elf info.
Everything used for elf info initialization comes from the param.sfo, so we can initialize this earlier to have this information accessible during memory init.
* Extract compiled SDK version from pubtoolinfo string
Up until now, we've been using the game's reported "firmware version" as our compiled SDK version. This behavior is inaccurate, and is something that has come up in my hardware tests before.
For the actual compiled SDK version, we should use the SDK version in the PUBTOOLINFO string of the param.sfo, only falling back on the firmware version when that the sdk_ver component isn't present.
* Store compiled SDK version in ElfInfo
* Limit address space for compiled SDK version at or above FW 3
Sony placed a hard cap at 0xfc00000000, with a slight extension for stack mappings. For now, though stack mappings aren't implemented, there's no harm in keeping a slightly extended address space (since this cap is lower than our old user max).
Limiting the max through address space is necessary for Windows due to performance issues, in the future I plan to properly implement checks in memory manager code to properly handle this behavior for all platforms.
* Use compiled SDK version for sceKernelGetCompiledSdkVersion
I think this is pretty self explanatory.
* Log SDK version
Since this value is what most internal firmware version checks are against, logging the value will help with debugging.
* Update address_space.cpp
* Update emulator.cpp
* Backwards compatible logging
Because that's apparently an issue now
* Define latest released firmware version, use that for sceKernelGetSystemSwVersion
I feel this is less hacky and error-prone than just returning the game firmware.
* sceKernelGetAllowedSdkVersionOnSystem
* sceKernelHasNeoMode
* sceKernelGetAppInfo stub
* sceKernelGetCurrentCpu
* fixups
* sceKernelGetMainSocId
Used by libSceAvPlayer to determine if console is a pro or not.
* Update process.cpp
* Set has_param_sfo to true
* Clang
* Swap write access mode for read write
Opening with access mode w will erase the opened file. We do not want this.
* Create mode
Opening with write access was previously the only way to create a file through open, so add a separate FileAccessMode that uses the write access mode to create files.
* Update file_system.cpp
Remove a hack added to posix_rename to bypass the file clearing behaviors of FileAccessMode::Write
* Check access mode in read functions
Write-only files cause the EBADF return on the various read functions. Now that we're opening files differently, properly handling this is necessary.
* Separate appends into proper modes
Fixes a potential regression from one of my prior PRs, and ensures the Write | Append flag combo also behaves properly in read-related functions.
* Move IsWriteOnly check after device/socket reads
file->f is only valid for files, so checking this before checking for sockets/devices will cause access violations.
* Fix issues
Now that Write is identical to ReadWrite, internal uses of Write need to be swapped to my new Create mode
* Fix remaining uses of FileAccessMode write to create files
Missed these before.
* Fix rebase
* Add stubbed get_authinfo (#3722)
* mostly stubbed get_authinfo
* Return value observed on console if get_authinfo was called for the current thread, esrch otherwise
---------
Co-authored-by: kalaposfos13 <153381648+kalaposfos13@users.noreply.github.com>
Co-authored-by: georgemoralis <giorgosmrls@gmail.com>
These are used by LLE libSceVideodec.
From decompiling the two GnmDriver libraries, it seems like sceGnmDrawInitToDefaultContextStateInternalCommand inlines a call to sceGnmDrawInitToDefaultContextState, so I've replaced that with an actual call to the function for readability.
sceGnmDrawInitToDefaultContextStateInternalSize is one to one with decomp.
* buffer_cache: smaller regions
this was a change back between v0.9.0 to v0.10.0
9f37ede336
reverting the TRACKER_HIGHER_PAGE_BITS from 24 to 22 gives a notable increase to performance.
* Update region_definitions.h
updated copyright
- Reserve an extra space for the terminating character, resolving an issue in GE2 where the last character did not appear when input reached the maximum length.
Co-authored-by: w1naenator <valdis.bogdans@hotmail.com>
* Changes
-Added support for OrbisImeParamExtended (extended IME parameters) in ImeHandler, ImeState, and ImeUi
-Updated all relevant constructors and logic to propagate and store the extended parameter
- Now fully supports passing extended options from sceImeOpen to the IME UI and backend
* Potential CUSA00434 [Debug] <Critical> assert.cpp:30 assert_fail_debug_msg: Assertion Failed!
buf_len + 1 <= buf_size && "Is your input buffer properly zero-terminated?" at C:/VS/shadPS4-ime-fixes/externals/dear_imgui/imgui_widgets.cpp:4601 fix
* Attempting to resolve an assertion failure in Diablo III:
- Adjusted buffer sizes
- Updated the calculation of text‑length values
* ime-lib another hotfix
Fixed incorrect param->title validation, which caused the IME dialog to fail to appear in Stardew Valley. Need to be checked.
* Clang fix
* FF9 ImeDialog Hotfix
* Removed the validation that disallowed null text and null placeholder, since using null values is valid in `ImeDialog`.
* Added additional debug logs to aid troubleshooting.
* IME Fixes
- Add missing flags to `OrbisImeExtOption`
- Improve debug logging
- Resolve nonstop `sceImeKeyboardOpen` calls in Stardew Valley (MonoGame engine) for `userId = 254`
* IME: guard null params for CUSA04909
- Add null checks in IME constructors to prevent crashes seen in CUSA04909.
- Leave a clear note about deferring keyboard event dispatch until guest-space translation is ready.
* Some improvements
- Added debug logs so every IME event and host callback (text/caret updates) shows what the guest sent back.
- Updated ImeState to respect the guest’s text-length limit, keep buffers in sync, and record caret/text changes without duplicates.
- Fixed shutdown by actually destroying the handler on close and letting sceImeUpdate exit quietly once the IME is gone.
* CLang
* IME: simplify handlers, add param checks, fix caret index
- Unify ImeHandler init; support optional OrbisImeParamExtended; drop userId from keyboard handler.
- Add basic null checks for work and inputTextBuffer; early error logging.
- Fixed incorrect caret position. Make caret and text area indices 1-based in ImeUi::InputTextCallback.
- Set default user_id to ORBIS_USER_SERVICE_USER_ID_INVALID in sceImeParamInit.
- Reduce noisy debug logs; promote key calls to LOG_INFO.
- Remove unused extended fields from ImeState; minor cleanups.
* IME: text/caret sync fixes; add Enter payload
- Sync UI input and work buffers on UpdateText
- Sync caret position on mouse click by emiting multiple UpdateCaret events for jumps (loop over delta)
- Add text payload to PressEnter (and Close); fixes IME in God Eater 2
- Queue initial Open event after open
- Fix UTF-8 → UTF-16 conversion bounds
- Add debug logs for all queued events
* CLang
* fixed accidental copy / paste replacement in text update event that broke text deletion.
* IME: Add code-point limited InputText and use in IME UI
- Add InputTextExLimited helper to cap Unicode code points and forward callbacks
- Switch IME input to InputTextExLimited with ime_param->maxTextLength and CallbackAlways
---------
Co-authored-by: w1naenator <valdis.bogdans@hotmail.com>
Reserved memory counts here, so we need to use !IsFree instead of IsMapped.
I swear this is like the 10th time I've messed this sorta thing up. Seems like it's the last case of this type of mistake in our current code though.
* Some structs and function definitions
* Fill in remaining function definitions and structs
The original variants of GetIdToken and GetAuthorizationCode use an online id instead of user id.
The V3 functions use the same internal function, but with a different flag. Unless games show me something different, they likely use the same structs, and definitely use the same parameters.
* Some errors
* Minor formatting change
* Some more errors
* GetIdToken error cases
* Remaining error cases
Just need to tackle request-related logic now.
* Basic request handling
Seems to internally behave similarly to libSceNpManager, but the actual data stored in libSceNpAuth requests appears to be different, so I've kept everything separated.
* NpAuthRequest usage
Again, behavior mirrors libSceNpManager request behavior, though it appears to be a separate implementation.
The only time libSceNpAuth uses libSceNpManager is to actually send the requests, where the act of sending a request involves creating a completely separate NpManager request, using NpManager functions to retrieve the desired data, then deleting the underlying NpManager request. All of this would happen inside GetAuthorizationCode and GetIdToken.
* Oops
* Missing mutexes
* Default output variables to zero
Not sure what all games might check for here, but setting the outputs to zero is probably safe.
* Implement sceNpCheckPlus
* Rework request storage
We'll need to store more data to "fake" async requests.
* sceNpAbortRequest
Pretty simple to add, so might as well.
* Formatting changes
* Async request logic
There's probably some things I'm getting wrong for cases where PSN is connected, but for reasons that should be pretty obvious, learning how that all works is a little more involved than the PSN disconnected results.
* Add missing error check
* Update np_manager.cpp
* Add a mutex to prevent concurrent reads/writes to g_requests
I imagine multi-threading is a lot more commonly used with the async functions, though I haven't tested enough to know.
* Update np_manager.h
* Move request creation to separate internal function
* Oops
Not sure how that got missed, but good thing I spotted it
* Oops
For now, I've included up to Windows 11 22H2 in the workaround.
I've only personally seen reports of issues on Windows 11 21H2, but better safe than sorry (considering Windows 10 22H2 has issues).
Based on the games I've checked, it seems like the service label parameter of sceNpTrophyCreateContext is what determines the trophy list.
Since we're already storing the service label in our contexts, and we're already extracting all trophy lists, all that needs doing is using the service label to select the right trophy list.
* SearchFree adjustments
* Robust address validation
I've adjusted IsValidAddress to take in a size, and check whether the whole range is contained in vma map.
If no size is provided, the function reverts to the old form of address validation instead.
* Map around gaps
As is, this should work mostly.
Only remaining issue is adding logic to pass the "mapped regions" to the guest vma map (and make such logic cross-platform).
* Initialize vma_map using gaps
This should allow memory code to catch any issues from address space gaps, and prevent non-fixed mappings from jumping to a location that isn't actually available.
* Clang
* Fix compile
* Clang
* Fix compile again
* Set system_managed_base and system_managed_size based on
Many places in our code use system_managed_base as the minimum mappable address, ensure this fact remains the same on Windows to prevent potential bugs.
* Reduce address validation in SearchFree
Allows SearchFree to function when a certain Windows GPU driver goes and reserves the whole system managed area.
Since SearchFree is only called on flexible addresses, allowing this particular case, where addresses are in bounds, but there's not enough space to map, should be safe enough.
* Bump address space size further
To handle Madden NFL 16 (and any games like it)
* More thorough logging of available memory regions
Should help with spotting weirdness.
* Formatting fixes
* Clang
* Slight reduction of user space
Still large enough to handle EA's shenanigans, but small enough that Linux doesn't break.
* Assert on VirtualQuery failure
* Extra debugging information
* Further reduce user space
This will unfix most of EA's titles, but UFC will still work.
Older windows versions support the high addresses, but trying to actually use them causes significant performance issues.
* Extra debugging info
Just in case other testers still run into issues.
* Remove debug logging
* Revert user space increases
Technically this constant is still higher than before, but weird side effects of our old logic resulted in a max address somewhere around this in main.
* address_space: Support expanded virtual memory space on macOS.
Co-Authored-By: squidbus <175574877+squidbus@users.noreply.github.com>
* Move address space constants to address_space.cpp
This ensures that all code must use the calculated address space memory values instead of the constants, since the calculated values can differ based on the platform.
This does require slight modification to thread state and gnmdriver code, since both were already using these constants directly.
* Workaround Windows 10 limitations
If a Windows 10 device is detected, use a lower value for USER_MAX to prevent system-wide hangs in VirtualAlloc2 calls.
* Fix compile for Windows-Qt
* Move tessellation_factors_ring_addr initialization to sceGnmGetTheTessellationFactorRingBufferBaseAddress
* Set image base address on Windows
This seems to work fine on Windows 11, needs testing from Windows 10 due to the previously discussed bugs.
* Set Linux executable base to 0x700000000000
This allows Linux to map the full user space without any workarounds.
Co-Authored-By: Marcin Mikołajczyk <2052578+mikusp@users.noreply.github.com>
* Basic formatting changes
* Reduce USER_MAX on Linux
Seems like finding a reliable way to move shadPS4's position in memory is difficult, for now limit the user size so we aren't trying to overwrite ourselves.
* Move memory and address_space variables.
---------
Co-authored-by: squidbus <175574877+squidbus@users.noreply.github.com>
Co-authored-by: Marcin Mikołajczyk <2052578+mikusp@users.noreply.github.com>
* Changes
-Added support for OrbisImeParamExtended (extended IME parameters) in ImeHandler, ImeState, and ImeUi
-Updated all relevant constructors and logic to propagate and store the extended parameter
- Now fully supports passing extended options from sceImeOpen to the IME UI and backend
* Potential CUSA00434 [Debug] <Critical> assert.cpp:30 assert_fail_debug_msg: Assertion Failed!
buf_len + 1 <= buf_size && "Is your input buffer properly zero-terminated?" at C:/VS/shadPS4-ime-fixes/externals/dear_imgui/imgui_widgets.cpp:4601 fix
* Attempting to resolve an assertion failure in Diablo III:
- Adjusted buffer sizes
- Updated the calculation of text‑length values
* ime-lib another hotfix
Fixed incorrect param->title validation, which caused the IME dialog to fail to appear in Stardew Valley. Need to be checked.
* Clang fix
* FF9 ImeDialog Hotfix
* Removed the validation that disallowed null text and null placeholder, since using null values is valid in `ImeDialog`.
* Added additional debug logs to aid troubleshooting.
* IME Fixes
- Add missing flags to `OrbisImeExtOption`
- Improve debug logging
- Resolve nonstop `sceImeKeyboardOpen` calls in Stardew Valley (MonoGame engine) for `userId = 254`
* IME: guard null params for CUSA04909
- Add null checks in IME constructors to prevent crashes seen in CUSA04909.
- Leave a clear note about deferring keyboard event dispatch until guest-space translation is ready.
* Some improvements
- Added debug logs so every IME event and host callback (text/caret updates) shows what the guest sent back.
- Updated ImeState to respect the guest’s text-length limit, keep buffers in sync, and record caret/text changes without duplicates.
- Fixed shutdown by actually destroying the handler on close and letting sceImeUpdate exit quietly once the IME is gone.
* CLang
---------
Co-authored-by: w1naenator <valdis.bogdans@hotmail.com>
Module relocation is not thread safe, games calling LoadAndStartModule on multiple threads can invalidate pointers while iterating through m_modules.
This fixes crashes in some apps on Windows.
It was reported this resulted in a noticeable 10+ fps drop in Driveclub so bring it back just with check to avoid it if the source range is an image alias
* Add RNG device
* rng device implementation
Tailored around libSceSsl2's usage, and based on fpPS4's implementation.
* Device file function types and log fixups
* Updated creates
Updates device file create functions to be the same as the directory file create functions.
* Fix compile
* Includes cleanup
Generally preferred to have full paths. Also removed some unused imports too.
* Fix buffer size
* Bring back cstdlib imports
Needed for Mac OS.
* video_core: Refactor render target bind to allow disabling MSAA
* video_core: Implement swapping of backing samples
* clang format
* video_core: Better implementation
Instead of downgrading to 1 sample, always try to match depth samples. This avoids needing to copy depth-stencil attachment and copying multisampled stencil is not possible on some vendors
* video_core: Small bugfixes
* image: Add null check
* vk_rasterizer: Swap backing samples on resolve dst
* vk_presenter: Reset backing samples before present
* video_core: Small refactor to make this implementation better
* reinterpret: Fix channel check for degamma
Seems this was simpler than I thought, hardware doesn't apply degamma on the W channel regardless of swizzle
* image: Add missing end rendering call
* blit_helper: Fix bug in old reinterpret path
* blit_helper: Remove unused layer vertex
Should be used in the future if copying many layers is needed
* vk_rasterizer: Apply suggestion
* vk_rasterizer: More bind refactor
* vk_instance: Re-enable extensions
* Add configurable extra memory
* lowercase getter and setter
* Refactor memory setup to configure maximum memory limits at runtime
* sir clang offnir, the all-formatting
* Correctly update BackingSize on W*ndows too
* small format change
* remove total_memory_to_use from the header
* i have no idea how to name this commit
"addressing review comments" is a good name i guess
* Do not include extraDmem in the general config
* Fix flag handling on Windows
Fixes a weird homebrew kalaposfos made
* Fix backing protects
Windows requires that protections on areas committed through MapViewOfFile functions are less than the original mapping.
The best way to make sure everything works is to VirtualProtect the code area with the requested protection instead of applying prot directly.
* Fix error code for sceKernelMapDirectMemory2
Real hardware returns EINVAL instead of EACCES here
* Fix prot setting in ProtectBytes
* Handle some extra protection-related edge cases.
Real hardware treats read and write as separate perms, but appends read if you call with write-only (this is visible in VirtualQuery calls)
Additionally, execute permissions are ignored when protecting dmem mappings.
* Properly handle exec permission behavior for memory pools
Calling sceKernelMemoryPoolCommit with executable permissions returns EINVAL, mprotect on pooled mappings ignores the exec protection.
* Clang
* Allow execution protection for direct memory
Further hardware tests show that the dmem area is actually executable, this permission is just hidden from the end user.
* Clang
* More descriptive assert message
* Align address and size in mmap
Like most POSIX functions, mmap aligns address down to the nearest page boundary, and aligns address up to the nearest page boundary.
Since mmap is the only memory mapping function that doesn't error early on misaligned length or size, handle the alignment in the libkernel code.
* Clang
* Fix valid flags
After changing the value, games that specify just CpuWrite would hit the error return.
* Fix prot conversion functions
The True(bool) function returns true whenever value is greater than 0. While this rarely manifested before because of our wrongly defined CpuReadWrite prot, it's now causing trouble with the corrected values.
Technically this could've also caused trouble with games mapping GpuRead permissions, but that seems to be a rare enough use case that I guess it never happened?
I've also added a warning for the case where `write & !read`, since we don't properly handle write-only permissions, and I'm not entirely sure what it would take to deal with that.
* Fix some lingering dmem issues
ReleaseDirectMemory was always unmapping with the size parameter, which could cause it to unmap too much. Since multiple mappings can reference the same dmem area, I've calculated how much of each VMA we're supposed to unmap.
Additionally, I've adjusted the logic for carving out the free dmem area to properly work if ReleaseDirectMemory is called over multiple dmem areas.
Finally, I've patched a bug with my code in UnmapMemory.
* Remove mapped dmem type
Since physical addresses can be mapped multiple times, tracking mapped pages is not necessary.
This also allows me to significantly simplify the MapMemory physical address validation logic.
* Proper implementation for sceKernelMtypeprotect
I've rewritten SetDirectMemoryType to use virtual addresses instead of physical addresses, allowing it to be used in sceKernelMtypeprotect.
To accommodate this change, I've also moved address and size alignment out of MemoryManager::Protect
* Apply memory type in sceKernelMemoryPoolCommit
* Organization
Some potentially important missing mutexes, removed some unnecessary mutexes, moved some mutexes after early error returns, and updated copyright dates
* Iterator logic cleanup
Missing end check in ClampRangeSize, and adjusted VirtualQuery and DirectMemoryQuery.
* Clang
* Adjustments
* Properly account for behavior differences in MapDirectMemory2
Undid the changes to direct memory areas, added more robust logic for changing dma types, and fixed DirectMemoryQuery to return hardware-accurate direct memory information in cases where dmas split here, but not on real hardware.
I've also changed MapMemory's is_exec flag to a validate_dmem flag, used to handle alternate behavior in MapDirectMemory2. is_exec is now determined by the use of MemoryProt::CpuExec instead.
* Clang
* Add execute permissions to physical backing
Needed for executable mappings to work properly on Windows, fixes regression in RE2 with prior commit.
* Minor variable cleanup
* Update memory.h
* Prohibit direct memory mappings with exec protections
Did a quick hardware test to confirm, only seems to be prohibited for dmem mappings though.
* Update memory.cpp
* Add support for restarting the emulator with new configurations
- Implement `Restart` function in `Emulator` to enable process relaunch with updated parameters.
- Modify `sceSystemServiceLoadExec` to use the restart functionality.
* Add logging for emulator restart and system service load execution
* Add IPC emulator PID output command
Impl `PID` output command to return the emulator process ID
- required for launches supporting emulator restart
* Add log file append mode support (used after restarting to keep the same log file)
* Keep game root between restarts
* add --wait-for-debugger option flag
* add --wait-for-pid flag
used for sync between parent & child process during restart
* impl restart via ipc
* fix override game root
* add qt flags to allow restart
