2026-08-31 - TSC-frequency band + more export candidates (D274)
The hardware-vs-orbistoun diff showed
sceKernelGetTscFrequency "refuted" in 139-exports, and the
cause was ours: the check did an exact == 0x5f259b8e where
the console answered 0x5f259bdd (~79 Hz of boot-calibration
drift). Replaced the equality with a tight band
(0x5f259000..0x5f25a000) - strong enough to be a signature,
loose enough to survive recalibration. Added four more export candidates
(getsid, sceKernelReadTsc, sceKernelGetProcessTimeCounter,
sceKernelGetProcessTimeCounterFrequency) so the next payload run
confirms more libkernel offsets for orbistoun's firmware table to
consume. Reached by base+vaddr, so no platform.h/imports.c churn. host +
check green.
Surprise worth keeping: the struct-layout data orbistoun most needs
is largely already here. 130-layout/system-software-version
already dumps the SwVersion bytes, and the last hardware run shows the
function writes the string at offset 8 and the version int at 0x24 but
leaves offset 0 (the caller's size) untouched - which is a correction
for orbistoun, whose new implementation writes that size field. And
sceKernelGetModuleInfo cannot be dumped on the current path
at all: it refuses in ps4-mode (0x8002_0016), so the
module-info layout waits on a native-mode run, not a new probe. The
attribute objects (mutexattr, thread attr) are void *
platform allocations of unknown size, so dumping them would risk the
over-read D008 exists to forbid - left alone.