[hugodearaujo]

What

sched_get_priority_max returned 256 and sched_get_priority_min 767 for policies 1 to 3, and other policies threw. TechnicalDebt said the console's answer was unverified. I called the console's exports from a homebrew payload on a PS5 (firmware 13.60):

Policy	Console max / min	main max / min
1 (SCHED_FIFO)	767 / 256	256 / 767
3 (SCHED_RR)	767 / 256	256 / 767
2 (SCHED_OTHER)	-1, errno 22 / -1, errno 22	256 / 767
-1, 0, 4, 5	-1, errno 22 / -1, errno 22	throws

So the range was swapped, and every policy without a range fails with EINVAL instead of a host exception. A second check agrees: on the console a mutex attribute set to PTHREAD_PRIO_PROTECT reports a default ceiling of 767, the highest SCHED_RR priority in FreeBSD libthr. pthread_mutexattr_setprotocol here takes that ceiling from sched_get_priority_max(SCHED_RR), so on main it is 256; this change fixes it too.

Title: Propagation: Paradise Hotel (PPSA15878) sets its thread priorities this way (eboot.bin 0x4b75d60):

pthread_getschedparam(thread, &policy, &param)
max = sched_get_priority_max(policy); min = sched_get_priority_min(policy)
priority = max == min ? 0 : min + roundf((4 - level) * (max - min) * 0.25f)
pthread_setschedparam(thread, policy, &param)

Threads report SCHED_FIFO both on the console and here, so with the swapped range each level got the mirrored priority: level 4 gave 767 instead of 256.

The SCHED_OTHER range was accepted for SDL titles (#1908). SDL now gets -1 and EINVAL, as on the console, instead of an exception or a range.

The TechnicalDebt entry gives the measurements. It also notes what is still different and left for a separate change: pthread_setschedparam ignores the policy and does not check the priority, where the console rejects SCHED_OTHER and priorities below 256 with EINVAL.

Tested

• PS5 firmware 13.60: the table, the mutex ceiling, and pthread_getschedparam on the main thread and on a created thread (both SCHED_FIFO).
• Linux: full CTest 906/906. guest_sched_priority now checks both ranges and the EINVAL cases; it aborts on main. guest_pthread_mutexattr already compares the ceiling with sched_get_priority_max(SCHED_RR) and passes.
• Windows Server 2025 VM, WinLibs GCC 15.2, CI configure flags: current main does not build on Windows (HostPoll in Socket.cpp, fixed by fix(libkernel): restore Windows select build for Demons Souls testing #4021), so I built this branch with fix(libkernel): restore Windows select build for Demons Souls testing #4021 merged: 903/905, guest_sched_priority and guest_pthread_mutexattr pass. agc_write_tracking and guest_kernel_stack_protect fail the same way on main with fix(libkernel): restore Windows select build for Demons Souls testing #4021 in this VM.

Overlap: #3654 adds its getpriority TechnicalDebt entry right below the entry this PR rewrites, so whichever lands second needs a one-line rebase. #3654 already conflicts with main elsewhere.

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 in TechnicalDebt
• 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
• AI-assisted: yes

🤖 Generated with Claude Code