Keep both EVFILT_PROC (main) and EVFILT_SIGNAL (this PR) in Kqueue.cpp: the constants, the two knote functions, the filter check and the dispatch. Debt doc: main's EVFILT_PROC and pipe wording with the PR's signal text.
ProsperoWeb (PRWB00001) 0.2.1 imports the three from libkernel. Only
libkernel_sys exported them, and a Linux run of the title does not load
libkernel_sys.prx, so the loader stopped with "undefined symbol:
LxIacrWiJ3w" before the title started.
Cpuset.cpp moves to libkernel whole, because cpuset and cpuset_setid
share its set table with cpuset_getid; its test and debt entry follow.
rfork_thread and sceKernelRdup move with the prototypes the title uses
and still throw.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The blocked mask of sigprocmask, _sigprocmask, pthread_sigmask and
sigsuspend was one process-wide set, so a thread changing its mask changed
every thread's, and scePthreadCreate did not hand the creating thread's mask
to the new thread. Each guest thread now has its own blocked mask and handler
holds, and a new thread starts with its creator's mask. The pending set stays
the process-wide one, so a signal that raise leaves pending is still taken by
sigwait and sigtimedwait on any thread and delivered by the first thread that
unblocks it. A SIGALRM expiry checks the mask of the thread that armed the
timer.
Fixes#2879.
Claude-Session: https://claude.ai/code/session_012SLQbhAiexxW6TtjYToW8a
With APS5_MIMALLOC set, libkernel restarts the title with
libmimalloc.so from libs/ first in LD_PRELOAD, as it does on macOS
with DYLD_INSERT_LIBRARIES. The macOS restart becomes one
implementation with platform pieces (MimallocRestart.cpp): on Linux
the executable is the target of /proc/self/exe and the arguments
those of /proc/self/cmdline. The restart is a priority 101
constructor, so it runs before libkernel's other initializers. It
gives the standard streams back from an ANYPS5_LOG_FILE copy before
execv, which also fixes the macOS restart under that variable, and
refuses a path the preload list cannot hold.
The build copies libmimalloc.so beside the libraries except in ASan
and TSan builds, the Linux release archives ship it, and the restart
tests run on Linux too, with a new case for ANYPS5_LOG_FILE.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_015W3M6NbYpmwWkzkL8YNLaF
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]>
Resolves docs/dev/debt/libkernel.md: main's sigaction entry (SIGPIPE outside Windows) and this branch's signal entry, which points to libSceLibcInternal for siginterrupt.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Resolves docs/dev/debt/libc.md: main's updated _Thrd entry, and siginterrupt stays documented in libSceLibcInternal.md.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
vm.overcommit is a read-write sysctl without CTLFLAG_PRISON, so FreeBSD 11.4 sysctl_root asks priv_check for PRIV_SYSCTL_WRITE before the handler runs, and prison_priv_check denies it to a jailed root. Under the credential model agreed on #11216 a write through sysctlbyname now fails with EPERM, whatever the value or its length, instead of throwing.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
ExpireAlarm and the pthread_kill path for another live thread called KqueueSignalSent before reading the signal's disposition. A thread woken by the EVFILT_SIGNAL knote could restore SIG_DFL for SIGALRM first, so the expiring alarm then terminated the process, as in the Windows guest_kqueue failure on #14886. Both now read the disposition first and count the signal in kqueues afterwards, as raise already did.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
An EV_ADD of EVFILT_SIGNAL is now accepted instead of throwing, as FreeBSD 11.4 filt_sigattach accepts it: the knote gets EV_CLEAR, and data counts the signals with that number sent since the knote was last reported, ignored ones included, as tdsendsignal counts them before its ignore check. Signals sent by raise, kill, killpg and pthread_kill and the ITIMER_REAL expiry are counted. EV_ENABLE, EV_DISABLE and EV_DELETE act on the knote as on other knotes.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
Every committed copy and compare on Windows went through
ReadProcessMemory, so a page released during the read counts as
uncommitted instead of faulting. Revalidation issues that system call
for each compare, and it dominates the committer thread.
Ranges the guest arena owns are now read by leaf copy and compare
routines. A fault on one of their source loads resumes at a failure
return, and the read falls back to the existing page by page path, so
the result for released pages is unchanged. Host regions and addresses
outside the arena keep ReadProcessMemory.
Hosts with AVX2 take 64 bytes per pass, with 16-byte and byte tails;
others keep the 16-byte loop.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
sysctl serves {1, 14, 8} since #13532, but sysctlbyname and sysctlnametomib still returned ENOENT for kern.proc.proc. Map the name to the same MIB, as FreeBSD's name lookup does.
QueueForDeliveryDrain ran after ResumeThread with no coordination with
the target's final ReapDeliveryDrain: a target could exit between resume
and enqueue, leaving the post in a box nothing would drain again and
retaining the QueuedDelivery reference.
Admission is now decided under drainBoxesLock against the target's
_finished flag, which is stored before the final reap in program order;
denied posts run their callback inline so the entry and the reference
are resolved exactly once on either path.
The GuestRaiseException regression arms a post hook, holds the producer
between resume and enqueue, lets the target finish and reap, then
completes the post and requires the drain to stay empty; the churn
rounds also assert the post hook observed mailbox posts alongside the
delivered raises.
A title is jailed (agreed on #11216). FreeBSD 11.4 prison_priv_check denies
PRIV_SCHED_RTPRIO, PRIV_SCHED_SET and PRIV_SCHED_SETPRIORITY to a jailed caller.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
NevesHenry