[Lindamea1]

What

No library exported scePthreadAttrGetschedpolicy (NID NMyIQ9WgWbU) or scePthreadAttrGetinheritsched (lpMP8HhkBbg). libkernel exports their setters, scePthreadAttrSetschedpolicy (4+h9EzwKF4I) and scePthreadAttrSetinheritsched (eXbUSpEaTsA), and the POSIX getters. The relinker emits a title's NID imports as global symbols unless the title marks them weak, and Linux output binds them at load unless --lazy-binding is given. So a title that imports either getter stops before it starts: on Linux the loader fails with symbol lookup error: ... undefined symbol: NMyIQ9WgWbU, and on Windows the entry stub prints FAIL: unresolved ELF import for it and stops with 0xc0000139 (WindowsEntryStubBuilder.cpp).

Dragon Ball Xenoverse 2 (PPSA08587) imports both. In #6591 (Windows, AnyPS5 31cbb55), LuisAguilarDev lists them under "Local changes, not yet submitted": "libkernel: scePthreadAttrGetschedpolicy and scePthreadAttrGetinheritsched. These are imported but missing. The title has not called them yet." This PR adds them. On main the title stops earlier, at the module #6592 adds ("Without it the title fails to load: Failed to load module: libSceNpTus.prx GetLastError: 126"). Of the other imports the thread lists as missing, sceNpSessionSignalingGetGroupInfo is on main since #7125, and sceNetCtlGetInfoV6, still missing on main, is declared by #7612. With #6592, this PR and #7612, every library and import the thread lists as missing is provided. With #6592 and those local changes, the thread reports that the title reaches its logos and then stops in AGC (stencil clear requires a writable stencil plane).

FreeBSD 11.4 libthr: _pthread_attr_getinheritsched returns EINVAL when attr or *attr is null and otherwise stores the attribute's sched_inherit; it does not check the output pointer. _pthread_attr_getschedpolicy returns EINVAL when attr, *attr or policy is null and otherwise stores sched_policy. libkernel's sce calls return 0x80020000 | errno, so EINVAL is 0x80020016. #4153 measured the neighbouring getters on a PS5 (firmware 13.60): a null or uninitialized attribute returns EINVAL, and so does a null output of scePthreadAttrGetstacksize, scePthreadAttrGetdetachstate and scePthreadAttrGetschedparam, whose libthr twins check it (L332, L182, L240). The two calls #4153 found to crash the console, a null attribute pointer to scePthreadAttrInit and a null output of scePthreadAttrGetaffinity, are ones where libthr writes through that pointer unchecked (L361, L649-L652), so every null case measured there follows libthr. The two getters added here were not measured on a console.

This PR adds both getters to Pthread/src/Attr.cpp, next to the other getters, and they follow libthr. They return 0x80020016 for a null or uninitialized attribute, and scePthreadAttrGetschedpolicy also returns it for a null output. Otherwise they store the attribute's policy or inheritance and return 0. A new attribute reports policy 1 (SCHED_FIFO) and inheritance 4 (PTHREAD_INHERIT_SCHED), the values scePthreadAttrInit stores. The values the setters accept (policies 1 and 3, inheritances 0 and 4) read back as set.

Where this still differs from FreeBSD:

• A null output of scePthreadAttrGetinheritsched throws instead of faulting, since libthr writes through it unchecked. pthread_attr_getinheritsched (technical debt) and scePthreadAttrGetsolosched throw the same way, and so do the two calls that crashed the console (technical debt). This PR adds to that entry that the two getters were not measured and follow libthr, and that this case throws.
• scePthreadAttrGet still fills in policy 1 and inheritance 4 for every thread (Attr.cpp L84-L85). FreeBSD's pthread_attr_get_np instead copies the thread's own attribute, which pthread_create took from the creation attribute. So after scePthreadAttrGet, these getters report 1 and 4 for a thread created with explicit inheritance and SCHED_RR, which scePthreadGetschedparam reports as 3. That belongs to a different function, and fix(libkernel): report thread guard size from attribute queries #6893 changes the same function, so it is left for a separate PR.

Tested

Gentoo Linux, kernel 7.2.8, GCC 16.2.1, glibc 2.43, Release, BUILD_TESTING=ON, on main 889cee5.

• guest_pthread_attr_errors, extended:
  • A new attribute reports policy 1 and inheritance 4.
  • After each rejected scePthreadAttrSetschedpolicy (-1, 0, 2, 4) and scePthreadAttrSetinheritsched (-1, 1, 2, 8), the getters still report 1 and 4. After each accepted set (policies 3 then 1, inheritances 0 then 4), they report the value just set.
  • scePthreadAttrGetschedpolicy with a null output returns 0x80020016, and scePthreadAttrGetinheritsched with a null output throws std::runtime_error. Both return 0x80020016 for a null and for an uninitialized attribute.
  • With only Attr.cpp reverted to main, the test does not link (undefined reference to 'scePthreadAttrGetschedpolicy' and 'scePthreadAttrGetinheritsched'). With the change it passes.
• nm -D on the NID-patched libraries: with the change, libkernel.prx exports NMyIQ9WgWbU and lpMP8HhkBbg, at the addresses of scePthreadAttrGetschedpolicy and scePthreadAttrGetinheritsched in the unpatched build. None of the other 126 NID-patched .prx files in build/core/libs/libs exports either NID. On main, libkernel.prx exports neither.
• A load probe shaped like the relinker's Linux output (a global FUNC import of the NID, DT_NEEDED libkernel.prx, BIND_NOW), run against the built libkernel.prx. On main, importing NMyIQ9WgWbU stops in the loader with symbol lookup error: ./load_NMyIQ9WgWbU: undefined symbol: NMyIQ9WgWbU (exit 127), and importing lpMP8HhkBbg fails the same way; importing the setter's eXbUSpEaTsA reaches main. With the change, all three reach main.
• A probe with the change:
  • The sce and POSIX getters report the same values, after the sce setters and after the POSIX setters. A policy of 2 set with pthread_attr_setschedpolicy reads back as 2.
  • On a thread created with explicit inheritance and policy 3, scePthreadGetschedparam reports 3, while scePthreadAttrGet followed by the new getters reports 1 and 4 (the difference described above).
• Tests the change reaches, selected by tools/select_tests.py as CI selects them: 1202 tests, 0 failed, 13 skipped. Full ctest -j6 --timeout 300: 1452 tests, 0 failed, 13 skipped (12 VideoOut tests and agc_driver_stencil_ref_export, which skip on main on this machine too).
• Windows: Attr.cpp and GuestPthreadAttrErrors.cpp compile with x86_64-w64-mingw32-g++ 16.2 (-fsyntax-only -Wall -Wextra). This is a syntax check, not a Windows build or run.
• tools/check_conventions.py passes against main. tools/pr_overlap.py --branch reports no overlapping lines with other pull requests. It could not compare 20 pull requests that conflict with main and share a file with this branch: test: organize library tests and label CTest suites #6044 (GuestPthreadAttrErrors.cpp) and 19 that change docs/dev/debt/libkernel.md. Merged with this branch, each has exactly the conflicts it has with main. Of the files in those conflicts, this branch changes only docs/dev/debt/libkernel.md, and its line there merges cleanly. test: organize library tests and label CTest suites #6044 moves GuestPthreadAttrErrors.cpp to core/libs/tests/threads/, and this branch's test changes are carried into the moved file. fix(libkernel): report thread guard size from attribute queries #6893 adds a line to scePthreadAttrGet in Attr.cpp and merges with this branch without a conflict.

Not run on Windows, macOS or a console.

Checklist

• Based on current main; no other open PR implements the same functions
• One topic per PR; follow-ups go in a new PR
• No comments in code except technical debt
• Unimplemented paths throw (NotImplemented_nid_no_patch); silent stubs are listed under Silent stubs in their library's technical debt file
• No notes, investigation .md files or images added to the repository (attach them to this PR)
• New third-party code is a submodule built from source
• Generic behaviour, not specific to one title
• Depends on: none
• Fixes: none
• Unblocks: PPSA08587 Dragon Ball Xenoverse 2 (PPSA08587) on Windows #6591 (with feat(np-tus): add libSceNpTus with offline title user storage #6592 and feat(libSceNetCtl): declare sceNetCtlGetInfoV6 #7612)
• GPUs tested: n/a
• AI-assisted: yes

🤖 Generated with Claude Code

[github-actions]

Progress report (e1338ff → 2514ead)

📈 System libraries: 90.25% (+0.01%, +2 implemented functions)

✅ 2 implemented

Library	Name
libkernel	scePthreadAttrGetinheritsched
libkernel	scePthreadAttrGetschedpolicy

[github-actions]

Some comparisons could not be completed because a pull request conflicts with the base branch. Review these shared files manually before concluding that there is no overlap.

Pull request	Shared files not compared
PR 5021	docs/dev/debt/libkernel.md
PR 6044	core/libs/tests/GuestPthreadAttrErrors.cpp
PR 6149	docs/dev/debt/libkernel.md
PR 6446	docs/dev/debt/libkernel.md
PR 6469	docs/dev/debt/libkernel.md
PR 6519	docs/dev/debt/libkernel.md
PR 6740	docs/dev/debt/libkernel.md
PR 6977	docs/dev/debt/libkernel.md
PR 7049	docs/dev/debt/libkernel.md
PR 7088	docs/dev/debt/libkernel.md
PR 7105	docs/dev/debt/libkernel.md
PR 7121	docs/dev/debt/libkernel.md
PR 7149	docs/dev/debt/libkernel.md
PR 7332	docs/dev/debt/libkernel.md
PR 7509	docs/dev/debt/libkernel.md
PR 7554	docs/dev/debt/libkernel.md
PR 7556	docs/dev/debt/libkernel.md
PR 7605	docs/dev/debt/libkernel.md

[LeandroLP23]

Windows run, since the description lists Windows as not run. Radeon 860M (RDNA 3.5, driver 32.0.13058.8001), Windows 11 26300, WinLibs GCC 15.2.0 r7, Release, BUILD_TESTING=ON. Build: main c3591adf9 with #7623, #7612, #7611, #7603, #7579, #7563 and #7541 merged together (each merge clean, built once).

The build is clean and guest_pthread_attr_errors passes 3 runs of 3.

AI-assisted

[LuisAguilarDev]

Supporting this one. Dragon Ball Xenoverse 2 (PPSA08587, #6591) needs both getters to load.

I had the same two getters on my fork since 2026-10-15 (3cb3fef): the getters return the values the existing setters store and reject a null attribute or output with EINVAL, as the other attribute getters do. With it, the title loads and reaches the Conton City hub (Windows 11, RTX 3070 Ti). I won't open a duplicate.

[adriadam10]

Merged: checked diff against libthr 11.4 refs, setters' fields, test pins values; Windows run by @LeandroLP23 passes; merged.