The temporary file is made read-only before the rename. When the rename
fails, Windows refuses to remove a read-only file, so the temporary copy
stayed in the cache directory. Give the owner write permission back
before removing it. The failure test now clears its cache directory,
including read-only files, before and after it runs, so it can run
again in the same directory.
A guest that reopens `stdin`, `stdout` or `stderr` onto a directory in a read mode threw instead of
returning the stream, while the console answers that a directory opens in a read mode and the
debt entry already recorded the throw as the Windows difference.
The standard stream now goes through the same descriptor move a regular file uses: the directory
descriptor is opened, moved onto the standard descriptor, and libkernel's directory entry follows
it, so `fstat` of the guest descriptor reports a directory and the first read answers `EISDIR` with
the error indicator set, like every other directory stream. A failed open restores the previous
standard target.
`guest_freopen_host_standard` reopens `stdout` onto a directory, asserts `S_IFDIR` on the guest
descriptor, asserts the first read answers `EISDIR`, then reopens the same stream on a file path
and writes through it.
Writing and reading the cached CA_LIST.cer used the throwing filesystem
and stream calls. Tests that share the cache directory raced on the
rename over the cached file, and on Windows the resulting
filesystem_error cannot be delivered through libkernel frames, so the
test process died. Use the error_code overloads, read in chunks that
set the stream error state, treat a lost rename race as success when the
cached file already holds the bundle, and answer an unavailable list
with no file instead of caching the failure.
The new test blocks the cache directory and the cached file in turn and
expects the open to fail with ENOENT, then to succeed once unblocked.
The deferred-dependency start state in InitializeDeferredModule (the
resolving vector and the started-results map) were thread_local objects with
destructors. On Windows a thread_local destructor in a module runs after
emulated TLS has freed the object, and prx_thread_local_destructors fails on
main for it. Hold both in HostThreadLocal like the other prx modules; the
trivial depth counter stays thread_local. Behaviour is unchanged.
On macOS under Rosetta, Aps5StartGuest moves the main thread off the host
stack and releases the old range to the guest arena. dyld keeps its own
pointers into the process start block at the top of that range and offers
no call that redirects them, where argv and environ have one:
_NSGetExecutablePath copies the path string out of the block on every
call, so the first call after the move read unmapped memory and the
process died with KERN_PROTECTION_FAILURE. A title does not have to call
it: libxpc reaches it from xpc_connection_resume, and ShaderCacheDirectory
calls it for the shader disk cache.
Release the old stack only up to the pages that hold the block, and only
while argv still points into the stack. They are the top two pages, 8 KiB
of the 8372224-byte stack here, so a title that fixes a mapping over that
range now fails with a host region overlap; the rest is free as before.
guest_main_stack_move asked for a 64 KiB guest mapping aligned down from
a local of main, which falls inside the kept pages. It now probes the
64 KiB below them, requires the kept pages to be refused, and calls
_NSGetExecutablePath after the move, which segfaults without this change.
The test declared sceKernelInstallExceptionHandler, sceKernelRemoveExceptionHandler
and longjmp_nid_postfix itself, the last with a void* buffer. Include the API
headers instead and keep the jump buffer as std::uint64_t, as the canonical
longjmp_nid_postfix(std::uint64_t*, int) takes it.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
dladdr1 searches the object's symbol table, which made every host fault
chained past an installed handler about 8 us slower on Linux. glibc 2.35
and later answer the same question with _dl_find_object, which is also
async-signal-safe; older glibc keeps dladdr1.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
A handler installed for SIGILL, SIGFPE, SIGBUS or SIGSEGV now receives
the faults raised in title code instead of the process ending, with
mc_trapno, mc_addr and mc_err filled as FreeBSD 11's amd64 trap() does.
On Windows the crash reporter's vectored handler offers unclaimed faults
to libkernel, which redirects the thread to a stub that runs the handler
outside the exception dispatcher; on Linux chained host signal handlers
classify faults by REG_TRAPNO.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Dean