Update pyproject.toml

Statx32/Statx64 (and their StatxTimestamp structs) were defined only as
little-endian ctypes.Structure, and ql_syscall_statx used them regardless
of guest endianness. Every other stat-family struct already has an explicit
big-endian variant selected by ql.arch.endian; statx was the lone gap.

On a big-endian guest (e.g. MIPS/MIPS64 EB) this byte-swaps every statx
field. In particular stx_mode loses its S_IFDIR bit, so a directory looks
like a plain file and `busybox ls /` just echoes the argument instead of
listing the directory.

Add Statx{32,64}EB / StatxTimestamp{32,64}EB (ctypes.BigEndianStructure,
reusing the little-endian field layouts) and pick them when the guest is
big-endian. The little-endian path is byte-for-byte unchanged.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Drives ql_syscall_statx on a big-endian MIPS context and checks that
stx_mode, read back in the guest's byte order, still reports a directory
(S_IFDIR). Before the fix the statx struct was emitted little-endian
regardless of guest endianness, so the type bits were byte-swapped and a
directory looked like a regular file. Self-contained (no rootfs binary,
runs on stock unicorn).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

renew docs

Fix statx byte order on big-endian guests

ARM and ARM64 Linux only hooked the svc syscall trap (exception 2), so any
other CPU exception fell through to the unhandled-interrupt dispatcher and
raised QlErrorCoreHook, crashing the framework.

This surfaces in examples/shellcode_run.py: since execve was patched to
return -1 for paths outside the rootfs, rootfs-less shellcode no longer
stops at execve and instead runs into its trailing data. On ARM64 that data
decodes as an undefined instruction (exception 1), which a real kernel would
deliver as SIGILL and terminate the process.

Hook EXCP.UDEF and emulate that termination by stopping cleanly, and replace
the magic syscall-trap number with the named EXCP.SWI constant.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Same class of crash as the ARM/ARM64 fix: MIPS Linux only hooked the
syscall exception, so any other CPU exception reached the unhandled-interrupt
dispatcher and raised QlErrorCoreHook.

In examples/shellcode_run.py the rootfs-less MIPS shellcode runs past its
denied execve into the trailing "/bin/sh" string, which decodes as a
reserved instruction (exception 20) -- delivered as SIGILL on a real kernel.

Hook the reserved-instruction exception and route it through the existing
hook_cpu_exception handler. Add a small EXCP enum to mips_const so the
syscall and reserved-instruction codes are named rather than magic numbers.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

The existing shellcode tests end emulation from their execve hook, so they
never reach the trailing '/bin/sh' string and never execute it as code. These
two deliberately let execve return and fall through into it, which raises an
undefined/reserved instruction exception on ARM64 and MIPS.

Without the preceding commits both fail with
QlErrorCoreHook: _hook_intr_cb : not handled.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

refine ARCHITECTURE file

Terminate the emulated process on an illegal-instruction CPU exception

Dev

Up ARCHITECTURE docs