Fully take over init/teardown so the LLE userService never sets up its IPMI client (whose login round-trip to a non-existent system daemon spins forev

Haptics on the touch-gamepad build would need a JNI call to the Vibrator service; no-op until that's wired through the NativeActivity.

* output = HMAC-SHA-1( hmac key, input buffer ) */

* SHA1 HMAC context reset */

Pack a literal string into SPIR-V words (LSB-first, null-terminated, zero padded to a word boundary).

The real syscall terminates the process with a diagnostic. We log the message and continue rather than killing boot.

We don't model async IO. Failing with eOPNOTSUPP makes guests fall back to synchronous IO. Every AIO entry point funnels here, so the log can't name w

Applies a list of dmem map/unmap/protect ops in one call. We don't model the direct-memory pool, so accept and report all entries processed. Logged so

Carve a mapping out of a block pool. We don't model pools, so back it with a plain anonymous RW region of the requested length. Logged once because th

sys_blockpool_open (653): allocates a "block pool" used by the flexible-memory allocator and returns a descriptor. We don't model block pools yet; the

Budget objects gate flexible memory / resource pools. We don't enforce them, but the guest stores the id and passes it back, so hand out a fixed non-z

The libkernel_sys wrappers (stub at 0x25c0) forward the args straight to the syscall; the kernel fills a caller-provided buffer that no libkernel func

Map any (pid, which) request onto a single cpu-time clock id. 2 corresponds to CLOCK_PROCESS_CPUTIME_ID on FreeBSD; callers only ever feed the result

Report a 1ns clock resolution. The host clocks we back these with are far finer than anything the guest schedules on, so advertising 1ns avoids the gu

PS4 timespec is {int64 tv_sec; int64 tv_nsec}, same layout as the host's on x86-64. Map the guest clock id onto a host clock; unknown ids fall back to

All guest threads run on the host scheduler with no pinning, so cpuset operations are accepted and ignored.

sceKernelGet/SetDirectMemoryContainer: -1 queries the current container id, any other value selects it and returns the previous one. We don't enforce

sys_dynlib_do_copy_relocations (596): processes R_X86_64_COPY relocations for the main executable. Our loader already resolves data relocations when i