mirrors/cosmopolitan
1
0
Fork 0
mirror of https://github.com/jart/cosmopolitan.git synced 2025-01-31 19:43:32 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions
cosmopolitan/libc/thread/pthread_cancel.c

410 lines
14 KiB
C
Raw Normal View History

Make mutex calling code 10x tinier Calls to lock/unlock functions are now NOPs by default. The first time clone() is called, they get turned into CALL instructions. Doing this caused funcctions like fputc() to shrink from 85 bytes to 45+4 bytes. Since the ANSI solution of `(__threaded && lock())` inlines os much superfluous binary content into functions all over the place.
2022-06-13 02:33:42 +00:00
/*-*- mode:c;indent-tabs-mode:nil;c-basic-offset:2;tab-width:8;coding:utf-8 -*-│
Fix vi modelines (#989) At least in neovim, `│vi:` is not recognized as a modeline because it has no preceding whitespace. After fixing this, opening a file yields an error because `net` is not an option. (`noet`, however, is.)
2023-12-05 22:37:54 +00:00
│ vi: set noet ft=c ts=2 sts=2 sw=2 fenc=utf-8 :vi │
Reduce makefile dependencies by 10% The includes in libc/calls/calls.h have now been refactored so that functions with struct parameters are declared in libc/calls/struct/
2022-06-09 03:01:28 +00:00
╞══════════════════════════════════════════════════════════════════════════════╡
│ Copyright 2022 Justine Alexandra Roberts Tunney │
│ │
│ Permission to use, copy, modify, and/or distribute this software for │
│ any purpose with or without fee is hereby granted, provided that the │
│ above copyright notice and this permission notice appear in all copies. │
│ │
│ THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL │
│ WARRANTIES WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED │
│ WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE │
│ AUTHOR BE LIABLE FOR ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL │
│ DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR │
│ PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER │
│ TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR │
│ PERFORMANCE OF THIS SOFTWARE. │
╚─────────────────────────────────────────────────────────────────────────────*/
Implement support for POSIX thread cancellations This change makes some miracle modifications to the System Five system call support, which lets us have safe, correct, and atomic handling of thread cancellations. It all turned out to be cheaper than anticipated because it wasn't necessary to modify the system call veneers. We were able to encode the cancellability of each system call into the magnums found in libc/sysv/syscalls.sh. Since cancellations are so waq, we are also supporting a lovely Musl Libc mask feature for raising ECANCELED.
2022-11-04 08:04:43 +00:00
#include "libc/assert.h"
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
#include "libc/atomic.h"
Implement support for POSIX thread cancellations This change makes some miracle modifications to the System Five system call support, which lets us have safe, correct, and atomic handling of thread cancellations. It all turned out to be cheaper than anticipated because it wasn't necessary to modify the system call veneers. We were able to encode the cancellability of each system call into the magnums found in libc/sysv/syscalls.sh. Since cancellations are so waq, we are also supporting a lovely Musl Libc mask feature for raising ECANCELED.
2022-11-04 08:04:43 +00:00
#include "libc/calls/calls.h"
#include "libc/calls/struct/sigaction.h"
#include "libc/calls/struct/siginfo.h"
#include "libc/calls/struct/sigset.h"
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
#include "libc/calls/struct/ucontext-freebsd.internal.h"
Implement thread cancellation for aarch64
2023-09-07 15:29:16 +00:00
#include "libc/calls/struct/ucontext.internal.h"
Implement support for POSIX thread cancellations This change makes some miracle modifications to the System Five system call support, which lets us have safe, correct, and atomic handling of thread cancellations. It all turned out to be cheaper than anticipated because it wasn't necessary to modify the system call veneers. We were able to encode the cancellability of each system call into the magnums found in libc/sysv/syscalls.sh. Since cancellations are so waq, we are also supporting a lovely Musl Libc mask feature for raising ECANCELED.
2022-11-04 08:04:43 +00:00
#include "libc/calls/ucontext.h"
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
#include "libc/cosmo.h"
Improve cancellations, randomness, and time - Exhaustively document cancellation points - Rename SIGCANCEL to SIGTHR just like BSDs - Further improve POSIX thread cancellations - Ensure asynchronous cancellations work correctly - Elevate the quality of getrandom() and getentropy() - Make futexes cancel correctly on OpenBSD 6.x and 7.x - Add reboot.com and shutdown.com to examples directory - Remove underscore prefix from awesome timespec_*() APIs - Create assertions that help verify our cancellation points - Remove bad timespec APIs (cmp generalizes eq/ne/gt/gte/lt/lte)
2022-11-06 02:49:41 +00:00
#include "libc/dce.h"
Implement support for POSIX thread cancellations This change makes some miracle modifications to the System Five system call support, which lets us have safe, correct, and atomic handling of thread cancellations. It all turned out to be cheaper than anticipated because it wasn't necessary to modify the system call veneers. We were able to encode the cancellability of each system call into the magnums found in libc/sysv/syscalls.sh. Since cancellations are so waq, we are also supporting a lovely Musl Libc mask feature for raising ECANCELED.
2022-11-04 08:04:43 +00:00
#include "libc/errno.h"
#include "libc/intrin/atomic.h"
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
#include "libc/intrin/describeflags.internal.h"
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
#include "libc/intrin/kprintf.h"
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
#include "libc/intrin/strace.internal.h"
Implement support for POSIX thread cancellations This change makes some miracle modifications to the System Five system call support, which lets us have safe, correct, and atomic handling of thread cancellations. It all turned out to be cheaper than anticipated because it wasn't necessary to modify the system call veneers. We were able to encode the cancellability of each system call into the magnums found in libc/sysv/syscalls.sh. Since cancellations are so waq, we are also supporting a lovely Musl Libc mask feature for raising ECANCELED.
2022-11-04 08:04:43 +00:00
#include "libc/str/str.h"
#include "libc/sysv/consts/sa.h"
#include "libc/sysv/consts/sig.h"
#include "libc/sysv/errfuns.h"
#include "libc/thread/posixthread.internal.h"
Tidy up the threading implementation The organization of the source files is now much more rational. Old experiments that didn't work out are now deleted. Naming of things like files is now more intuitive.
2022-09-10 09:56:25 +00:00
#include "libc/thread/thread.h"
Implement support for POSIX thread cancellations This change makes some miracle modifications to the System Five system call support, which lets us have safe, correct, and atomic handling of thread cancellations. It all turned out to be cheaper than anticipated because it wasn't necessary to modify the system call veneers. We were able to encode the cancellability of each system call into the magnums found in libc/sysv/syscalls.sh. Since cancellations are so waq, we are also supporting a lovely Musl Libc mask feature for raising ECANCELED.
2022-11-04 08:04:43 +00:00
#include "libc/thread/tls.h"
Get LIBC_MEM and LIBC_STDIO building with aarch64
2023-05-09 15:08:56 +00:00
Fix backtraces on cancellation points
2022-11-05 02:55:41 +00:00
int systemfive_cancel(void);
extern const char systemfive_cancellable[];
extern const char systemfive_cancellable_end[];
Implement support for POSIX thread cancellations This change makes some miracle modifications to the System Five system call support, which lets us have safe, correct, and atomic handling of thread cancellations. It all turned out to be cheaper than anticipated because it wasn't necessary to modify the system call veneers. We were able to encode the cancellability of each system call into the magnums found in libc/sysv/syscalls.sh. Since cancellations are so waq, we are also supporting a lovely Musl Libc mask feature for raising ECANCELED.
2022-11-04 08:04:43 +00:00
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
long _pthread_cancel_ack(void) {
Fix some issues
2023-10-10 03:18:48 +00:00
struct PosixThread *pt;
STRACE("_pthread_cancel_ack()");
pt = _pthread_self();
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
if (!(pt->pt_flags & (PT_NOCANCEL | PT_MASKED)) ||
(pt->pt_flags & PT_ASYNC)) {
Implement support for POSIX thread cancellations This change makes some miracle modifications to the System Five system call support, which lets us have safe, correct, and atomic handling of thread cancellations. It all turned out to be cheaper than anticipated because it wasn't necessary to modify the system call veneers. We were able to encode the cancellability of each system call into the magnums found in libc/sysv/syscalls.sh. Since cancellations are so waq, we are also supporting a lovely Musl Libc mask feature for raising ECANCELED.
2022-11-04 08:04:43 +00:00
pthread_exit(PTHREAD_CANCELED);
}
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
pt->pt_flags |= PT_NOCANCEL;
if (IsOpenbsd()) {
pt->pt_flags |= PT_OPENBSD_KLUDGE;
}
Implement support for POSIX thread cancellations This change makes some miracle modifications to the System Five system call support, which lets us have safe, correct, and atomic handling of thread cancellations. It all turned out to be cheaper than anticipated because it wasn't necessary to modify the system call veneers. We were able to encode the cancellability of each system call into the magnums found in libc/sysv/syscalls.sh. Since cancellations are so waq, we are also supporting a lovely Musl Libc mask feature for raising ECANCELED.
2022-11-04 08:04:43 +00:00
return ecanceled();
}
Fix some issues
2023-10-10 03:18:48 +00:00
// the purpose of this routine is to force a blocking system call to end
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
static void _pthread_cancel_sig(int sig, siginfo_t *si, void *arg) {
ucontext_t *ctx = arg;
// check thread runtime state is initialized and cancelled
Make more fixes and improvements - Remove PAGESIZE constant - Fix realloc() documentation - Fix ttyname_r() error reporting - Make forking more reliable on Windows - Make execvp() a few microseconds faster - Make system() a few microseconds faster - Tighten up the socket-related magic numbers - Loosen restrictions on mmap() offset alignment - Improve GetProgramExecutableName() with getenv("_") - Use mkstemp() as basis for mktemp(), tmpfile(), tmpfd() - Fix flakes in pthread_cancel_test, unix_test, fork_test - Fix recently introduced futex stack overflow regression - Let sockets be passed as stdio to subprocesses on Windows - Improve security of bind() on Windows w/ SO_EXCLUSIVEADDRUSE
2023-07-30 01:44:15 +00:00
struct PosixThread *pt;
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
if (!__tls_enabled) return;
if (!(pt = _pthread_self())) return;
if (pt->pt_flags & PT_NOCANCEL) return;
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
if (!atomic_load_explicit(&pt->pt_canceled, memory_order_acquire)) return;
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
Always use compiler builtin for offsetof() We have received multiple reports of GCC breaking builds when compiler flags like `-std=c11` were being passed. The workaround until the next release is to simply not define `__STRICT_ANSI__` which is a bad idea.
2023-12-07 22:20:12 +00:00
// in asynchronous mode the asynchronous signal calls exit
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
if (pt->pt_flags & PT_ASYNC) {
sigaddset(&ctx->uc_sigmask, SIGTHR);
pthread_sigmask(SIG_SETMASK, &ctx->uc_sigmask, 0);
pthread_exit(PTHREAD_CANCELED);
}
// prevent this handler from being called again by thread
sigaddset(&ctx->uc_sigmask, SIGTHR);
// check for race condition between pre-check and syscall
// rewrite the thread's execution state to acknowledge it
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
// sadly windows isn't able to be sophisticated like this
if (!IsWindows()) {
if (systemfive_cancellable <= (char *)ctx->uc_mcontext.PC &&
(char *)ctx->uc_mcontext.PC < systemfive_cancellable_end) {
ctx->uc_mcontext.PC = (intptr_t)systemfive_cancel;
return;
}
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
}
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
// punts cancelation to start of next cancellation point
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
// we ensure sigthr is a pending signal in case unblocked
Make improvements - Improved async signal safety of read() particularly for longjmp() - Started adding cancel cleanup handlers for locks / etc on Windows - Make /dev/tty work better particularly for uses like `foo | less` - Eagerly read console input into a linked list, so poll can signal - Fix some libc definitional bugs, which configure scripts detected
2023-09-21 14:30:39 +00:00
raise(sig);
Implement support for POSIX thread cancellations This change makes some miracle modifications to the System Five system call support, which lets us have safe, correct, and atomic handling of thread cancellations. It all turned out to be cheaper than anticipated because it wasn't necessary to modify the system call veneers. We were able to encode the cancellability of each system call into the magnums found in libc/sysv/syscalls.sh. Since cancellations are so waq, we are also supporting a lovely Musl Libc mask feature for raising ECANCELED.
2022-11-04 08:04:43 +00:00
}
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
static void _pthread_cancel_listen(void) {
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
struct sigaction sa = {
.sa_mask = -1,
.sa_flags = SA_SIGINFO,
.sa_sigaction = _pthread_cancel_sig,
};
sigaction(SIGTHR, &sa, 0);
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
}
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
static errno_t _pthread_cancel_single(struct PosixThread *pt) {
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
// install our special signal handler
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
static atomic_uint once;
cosmo_once(&once, _pthread_cancel_listen);
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
// check if thread is already dead
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
// we don't care about any further esrch checks upstream
switch (atomic_load_explicit(&pt->pt_status, memory_order_acquire)) {
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
case kPosixThreadZombie:
case kPosixThreadTerminated:
return ESRCH;
default:
break;
}
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
// erase this thread from the book of life
atomic_store_explicit(&pt->pt_canceled, 1, memory_order_release);
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
// does this thread want to cancel itself? just exit
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
if (pt == _pthread_self()) {
if (!(pt->pt_flags & (PT_NOCANCEL | PT_MASKED)) &&
(pt->pt_flags & PT_ASYNC)) {
pthread_exit(PTHREAD_CANCELED);
}
return 0;
}
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
// send the cancelation signal
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
errno_t err;
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
err = pthread_kill((pthread_t)pt, SIGTHR);
if (err == ESRCH) err = 0;
return err;
}
static errno_t _pthread_cancel_everyone(void) {
errno_t err;
struct Dll *e;
struct PosixThread *other;
err = ESRCH;
_pthread_lock();
for (e = dll_first(_pthread_list); e; e = dll_next(_pthread_list, e)) {
other = POSIXTHREAD_CONTAINER(e);
if (other != _pthread_self() &&
atomic_load_explicit(&other->pt_status, memory_order_acquire) <
kPosixThreadTerminated) {
_pthread_cancel_single(other);
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
err = 0;
}
}
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
_pthread_unlock();
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
return err;
Implement support for POSIX thread cancellations This change makes some miracle modifications to the System Five system call support, which lets us have safe, correct, and atomic handling of thread cancellations. It all turned out to be cheaper than anticipated because it wasn't necessary to modify the system call veneers. We were able to encode the cancellability of each system call into the magnums found in libc/sysv/syscalls.sh. Since cancellations are so waq, we are also supporting a lovely Musl Libc mask feature for raising ECANCELED.
2022-11-04 08:04:43 +00:00
}
Reduce makefile dependencies by 10% The includes in libc/calls/calls.h have now been refactored so that functions with struct parameters are declared in libc/calls/struct/
2022-06-09 03:01:28 +00:00
Perform some code maintenance - Change IDT code so kprintf() isn't mandatory dependency - Document current intentions around pthread_cancel() - Make _npassert() an _unassert() in MODE=tiny
2022-10-09 20:00:46 +00:00
/**
* Cancels thread.
*
Implement support for POSIX thread cancellations This change makes some miracle modifications to the System Five system call support, which lets us have safe, correct, and atomic handling of thread cancellations. It all turned out to be cheaper than anticipated because it wasn't necessary to modify the system call veneers. We were able to encode the cancellability of each system call into the magnums found in libc/sysv/syscalls.sh. Since cancellations are so waq, we are also supporting a lovely Musl Libc mask feature for raising ECANCELED.
2022-11-04 08:04:43 +00:00
* When a thread is cancelled, it'll interrupt blocking i/o calls,
* invoke any cleanup handlers that were pushed on the thread's stack
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
* before the cancelation occurred, in addition to destructing pthread
Improve cancellations, randomness, and time - Exhaustively document cancellation points - Rename SIGCANCEL to SIGTHR just like BSDs - Further improve POSIX thread cancellations - Ensure asynchronous cancellations work correctly - Elevate the quality of getrandom() and getentropy() - Make futexes cancel correctly on OpenBSD 6.x and 7.x - Add reboot.com and shutdown.com to examples directory - Remove underscore prefix from awesome timespec_*() APIs - Create assertions that help verify our cancellation points - Remove bad timespec APIs (cmp generalizes eq/ne/gt/gte/lt/lte)
2022-11-06 02:49:41 +00:00
* keys, before finally, the thread shall abruptly exit.
*
* By default, pthread_cancel() can only take effect when a thread
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
* reaches a cancelation point. Such functions are documented with
* `@cancelationpoint`. They check the cancellation state before the
Improve cancellations, randomness, and time - Exhaustively document cancellation points - Rename SIGCANCEL to SIGTHR just like BSDs - Further improve POSIX thread cancellations - Ensure asynchronous cancellations work correctly - Elevate the quality of getrandom() and getentropy() - Make futexes cancel correctly on OpenBSD 6.x and 7.x - Add reboot.com and shutdown.com to examples directory - Remove underscore prefix from awesome timespec_*() APIs - Create assertions that help verify our cancellation points - Remove bad timespec APIs (cmp generalizes eq/ne/gt/gte/lt/lte)
2022-11-06 02:49:41 +00:00
* underlying system call is issued. If the system call is issued and
* blocks, then pthread_cancel() will interrupt the operation in which
* case the syscall wrapper will check the cancelled state a second
* time, only if the raw system call returned EINTR.
*
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
* The following system calls are implemented as cancelation points.
Improve cancellations, randomness, and time - Exhaustively document cancellation points - Rename SIGCANCEL to SIGTHR just like BSDs - Further improve POSIX thread cancellations - Ensure asynchronous cancellations work correctly - Elevate the quality of getrandom() and getentropy() - Make futexes cancel correctly on OpenBSD 6.x and 7.x - Add reboot.com and shutdown.com to examples directory - Remove underscore prefix from awesome timespec_*() APIs - Create assertions that help verify our cancellation points - Remove bad timespec APIs (cmp generalizes eq/ne/gt/gte/lt/lte)
2022-11-06 02:49:41 +00:00
*
* - `accept4`
* - `accept`
* - `clock_nanosleep`
* - `connect`
* - `copy_file_range`
* - `creat`
* - `epoll_wait`
* - `fcntl(F_OFD_SETLKW)`
* - `fcntl(F_SETLKW)`
* - `fdatasync`
* - `flock`
* - `fstatfs`
* - `fsync`
* - `ftruncate`
* - `getrandom`
* - `msync`
* - `nanosleep`
* - `open`
* - `openat`
* - `pause`
* - `poll`
* - `ppoll`
* - `pread`
* - `preadv`
* - `pselect`
* - `pwrite`
* - `pwritev`
* - `read`
* - `readv`
* - `recvfrom`
* - `recvmsg`
* - `select`
* - `sendmsg`
* - `sendto`
* - `sigsuspend`
* - `sigtimedwait`
* - `sigwaitinfo`
* - `statfs`
* - `tcdrain`
* - `truncate`
* - `wait3`
* - `wait4`
* - `wait`
* - `waitpid`
* - `write`
* - `writev`
*
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
* The following library calls are implemented as cancelation points.
Improve cancellations, randomness, and time - Exhaustively document cancellation points - Rename SIGCANCEL to SIGTHR just like BSDs - Further improve POSIX thread cancellations - Ensure asynchronous cancellations work correctly - Elevate the quality of getrandom() and getentropy() - Make futexes cancel correctly on OpenBSD 6.x and 7.x - Add reboot.com and shutdown.com to examples directory - Remove underscore prefix from awesome timespec_*() APIs - Create assertions that help verify our cancellation points - Remove bad timespec APIs (cmp generalizes eq/ne/gt/gte/lt/lte)
2022-11-06 02:49:41 +00:00
*
* - `fopen`
* - `gzopen`, `gzread`, `gzwrite`, etc.
* - `lockf(F_LOCK)`
* - `nsync_cv_wait_with_deadline`
* - `nsync_cv_wait`
* - `opendir`
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
* - `openatemp`, 'mkstemp', etc.
Fix some issues
2023-10-10 03:18:48 +00:00
* - `sleep`, `usleep`, `nanosleep`, `timespec_sleep`, etc.
Improve cancellations, randomness, and time - Exhaustively document cancellation points - Rename SIGCANCEL to SIGTHR just like BSDs - Further improve POSIX thread cancellations - Ensure asynchronous cancellations work correctly - Elevate the quality of getrandom() and getentropy() - Make futexes cancel correctly on OpenBSD 6.x and 7.x - Add reboot.com and shutdown.com to examples directory - Remove underscore prefix from awesome timespec_*() APIs - Create assertions that help verify our cancellation points - Remove bad timespec APIs (cmp generalizes eq/ne/gt/gte/lt/lte)
2022-11-06 02:49:41 +00:00
* - `pclose`
* - `popen`
* - `fwrite`, `printf`, `fprintf`, `putc`, etc.
* - `pthread_cond_timedwait`
* - `pthread_cond_wait`
* - `pthread_join`
* - `sem_timedwait`
* - `sem_wait`
* - `sleep`
* - `timespec_sleep_until`
* - `tmpfd`
* - `tmpfile`
* - `usleep`
*
* Other userspace libraries provided by Cosmopolitan Libc that call the
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
* cancelation points above will block cancellations while running. The
* following are examples of functions that *aren't* cancelation points
Improve cancellations, randomness, and time - Exhaustively document cancellation points - Rename SIGCANCEL to SIGTHR just like BSDs - Further improve POSIX thread cancellations - Ensure asynchronous cancellations work correctly - Elevate the quality of getrandom() and getentropy() - Make futexes cancel correctly on OpenBSD 6.x and 7.x - Add reboot.com and shutdown.com to examples directory - Remove underscore prefix from awesome timespec_*() APIs - Create assertions that help verify our cancellation points - Remove bad timespec APIs (cmp generalizes eq/ne/gt/gte/lt/lte)
2022-11-06 02:49:41 +00:00
*
* - `INFOF()`, `WARNF()`, etc.
* - `getentropy`
* - `gmtime_r`
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
* - `kprintf` (by virtue of asm(syscall) and write_nocancel() on xnu)
Improve cancellations, randomness, and time - Exhaustively document cancellation points - Rename SIGCANCEL to SIGTHR just like BSDs - Further improve POSIX thread cancellations - Ensure asynchronous cancellations work correctly - Elevate the quality of getrandom() and getentropy() - Make futexes cancel correctly on OpenBSD 6.x and 7.x - Add reboot.com and shutdown.com to examples directory - Remove underscore prefix from awesome timespec_*() APIs - Create assertions that help verify our cancellation points - Remove bad timespec APIs (cmp generalizes eq/ne/gt/gte/lt/lte)
2022-11-06 02:49:41 +00:00
* - `localtime_r`
* - `nsync_mu_lock`
* - `nsync_mu_unlock`
* - `openpty`
* - `pthread_getname_np`
* - `pthread_mutex_lock`
* - `pthread_mutex_unlock`
* - `pthread_setname_np`
* - `sem_open`
* - `system`
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
* - `openatemp`, 'mkstemp', etc.
Improve cancellations, randomness, and time - Exhaustively document cancellation points - Rename SIGCANCEL to SIGTHR just like BSDs - Further improve POSIX thread cancellations - Ensure asynchronous cancellations work correctly - Elevate the quality of getrandom() and getentropy() - Make futexes cancel correctly on OpenBSD 6.x and 7.x - Add reboot.com and shutdown.com to examples directory - Remove underscore prefix from awesome timespec_*() APIs - Create assertions that help verify our cancellation points - Remove bad timespec APIs (cmp generalizes eq/ne/gt/gte/lt/lte)
2022-11-06 02:49:41 +00:00
* - `timespec_sleep`
* - `touch`
*
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
* The way to block cancelations temporarily is:
Improve cancellations, randomness, and time - Exhaustively document cancellation points - Rename SIGCANCEL to SIGTHR just like BSDs - Further improve POSIX thread cancellations - Ensure asynchronous cancellations work correctly - Elevate the quality of getrandom() and getentropy() - Make futexes cancel correctly on OpenBSD 6.x and 7.x - Add reboot.com and shutdown.com to examples directory - Remove underscore prefix from awesome timespec_*() APIs - Create assertions that help verify our cancellation points - Remove bad timespec APIs (cmp generalizes eq/ne/gt/gte/lt/lte)
2022-11-06 02:49:41 +00:00
*
* int cs;
* pthread_setcancelstate(PTHREAD_CANCEL_DISABLE, &cs);
* // ...
* pthread_setcancelstate(cs, 0);
*
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
* In order to support cancelations all your code needs to be rewritten
Improve cancellations, randomness, and time - Exhaustively document cancellation points - Rename SIGCANCEL to SIGTHR just like BSDs - Further improve POSIX thread cancellations - Ensure asynchronous cancellations work correctly - Elevate the quality of getrandom() and getentropy() - Make futexes cancel correctly on OpenBSD 6.x and 7.x - Add reboot.com and shutdown.com to examples directory - Remove underscore prefix from awesome timespec_*() APIs - Create assertions that help verify our cancellation points - Remove bad timespec APIs (cmp generalizes eq/ne/gt/gte/lt/lte)
2022-11-06 02:49:41 +00:00
* so that when resources such as file descriptors are managed they must
* have a cleanup crew pushed to the stack. For example even malloc() is
* technically unsafe w.r.t. leaks without doing something like this:
*
* void *p = malloc(123);
* pthread_cleanup_push(free, p);
* read(0, p, 123);
* pthread_cleanup_pop(1);
*
* Consider using Cosmopolitan Libc's garbage collector since it will be
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
* executed when a thread exits due to a cancelation.
Improve cancellations, randomness, and time - Exhaustively document cancellation points - Rename SIGCANCEL to SIGTHR just like BSDs - Further improve POSIX thread cancellations - Ensure asynchronous cancellations work correctly - Elevate the quality of getrandom() and getentropy() - Make futexes cancel correctly on OpenBSD 6.x and 7.x - Add reboot.com and shutdown.com to examples directory - Remove underscore prefix from awesome timespec_*() APIs - Create assertions that help verify our cancellation points - Remove bad timespec APIs (cmp generalizes eq/ne/gt/gte/lt/lte)
2022-11-06 02:49:41 +00:00
*
* void *p = _gc(malloc(123));
* read(0, p, 123);
Perform some code maintenance - Change IDT code so kprintf() isn't mandatory dependency - Document current intentions around pthread_cancel() - Make _npassert() an _unassert() in MODE=tiny
2022-10-09 20:00:46 +00:00
*
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
* It's possible to put a thread in asynchronous cancelation mode with
Improve cancellations, randomness, and time - Exhaustively document cancellation points - Rename SIGCANCEL to SIGTHR just like BSDs - Further improve POSIX thread cancellations - Ensure asynchronous cancellations work correctly - Elevate the quality of getrandom() and getentropy() - Make futexes cancel correctly on OpenBSD 6.x and 7.x - Add reboot.com and shutdown.com to examples directory - Remove underscore prefix from awesome timespec_*() APIs - Create assertions that help verify our cancellation points - Remove bad timespec APIs (cmp generalizes eq/ne/gt/gte/lt/lte)
2022-11-06 02:49:41 +00:00
*
* pthread_setcancelstate(PTHREAD_CANCEL_ASYNCHRONOUS, 0);
* for (;;) donothing;
*
* In which case a thread may be cancelled at any assembly opcode. This
* is useful for immediately halting threads that consume cpu and don't
* use any system calls. It shouldn't be used on threads that will call
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
* cancelation points since in that case asynchronous mode could cause
Improve cancellations, randomness, and time - Exhaustively document cancellation points - Rename SIGCANCEL to SIGTHR just like BSDs - Further improve POSIX thread cancellations - Ensure asynchronous cancellations work correctly - Elevate the quality of getrandom() and getentropy() - Make futexes cancel correctly on OpenBSD 6.x and 7.x - Add reboot.com and shutdown.com to examples directory - Remove underscore prefix from awesome timespec_*() APIs - Create assertions that help verify our cancellation points - Remove bad timespec APIs (cmp generalizes eq/ne/gt/gte/lt/lte)
2022-11-06 02:49:41 +00:00
* resource leaks to happen, in such a way that can't be worked around.
Perform some code maintenance - Change IDT code so kprintf() isn't mandatory dependency - Document current intentions around pthread_cancel() - Make _npassert() an _unassert() in MODE=tiny
2022-10-09 20:00:46 +00:00
*
Implement support for POSIX thread cancellations This change makes some miracle modifications to the System Five system call support, which lets us have safe, correct, and atomic handling of thread cancellations. It all turned out to be cheaper than anticipated because it wasn't necessary to modify the system call veneers. We were able to encode the cancellability of each system call into the magnums found in libc/sysv/syscalls.sh. Since cancellations are so waq, we are also supporting a lovely Musl Libc mask feature for raising ECANCELED.
2022-11-04 08:04:43 +00:00
* If none of the above options seem savory to you, then a third way is
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
* offered for doing cancelations. Cosmopolitan Libc supports the Musl
Implement support for POSIX thread cancellations This change makes some miracle modifications to the System Five system call support, which lets us have safe, correct, and atomic handling of thread cancellations. It all turned out to be cheaper than anticipated because it wasn't necessary to modify the system call veneers. We were able to encode the cancellability of each system call into the magnums found in libc/sysv/syscalls.sh. Since cancellations are so waq, we are also supporting a lovely Musl Libc mask feature for raising ECANCELED.
2022-11-04 08:04:43 +00:00
* Libc `PTHREAD_CANCEL_MASKED` non-POSIX extension. Any thread may pass
* this setting to pthread_setcancelstate(), in which case threads won't
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
* be abruptly destroyed upon cancelation and have their stack unwound;
* instead, cancelation points will simply raise an `ECANCELED` error,
Implement support for POSIX thread cancellations This change makes some miracle modifications to the System Five system call support, which lets us have safe, correct, and atomic handling of thread cancellations. It all turned out to be cheaper than anticipated because it wasn't necessary to modify the system call veneers. We were able to encode the cancellability of each system call into the magnums found in libc/sysv/syscalls.sh. Since cancellations are so waq, we are also supporting a lovely Musl Libc mask feature for raising ECANCELED.
2022-11-04 08:04:43 +00:00
* which can be more safely and intuitively handled for many use cases.
Improve cancellations, randomness, and time - Exhaustively document cancellation points - Rename SIGCANCEL to SIGTHR just like BSDs - Further improve POSIX thread cancellations - Ensure asynchronous cancellations work correctly - Elevate the quality of getrandom() and getentropy() - Make futexes cancel correctly on OpenBSD 6.x and 7.x - Add reboot.com and shutdown.com to examples directory - Remove underscore prefix from awesome timespec_*() APIs - Create assertions that help verify our cancellation points - Remove bad timespec APIs (cmp generalizes eq/ne/gt/gte/lt/lte)
2022-11-06 02:49:41 +00:00
* For example:
*
* pthread_setcancelstate(PTHREAD_CANCEL_MASKED, 0);
* void *p = malloc(123);
* int rc = read(0, p, 123);
* free(p);
* if (rc == ECANCELED) {
* pthread_exit(0);
* }
*
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
* Shows how the masked cancelations paradigm can be safely used. Note
* that it's so important that cancelation point error return codes be
Improve cancellations, randomness, and time - Exhaustively document cancellation points - Rename SIGCANCEL to SIGTHR just like BSDs - Further improve POSIX thread cancellations - Ensure asynchronous cancellations work correctly - Elevate the quality of getrandom() and getentropy() - Make futexes cancel correctly on OpenBSD 6.x and 7.x - Add reboot.com and shutdown.com to examples directory - Remove underscore prefix from awesome timespec_*() APIs - Create assertions that help verify our cancellation points - Remove bad timespec APIs (cmp generalizes eq/ne/gt/gte/lt/lte)
2022-11-06 02:49:41 +00:00
* checked. Code such as the following:
*
* pthread_setcancelstate(PTHREAD_CANCEL_MASKED, 0);
* void *p = malloc(123);
* write(2, "log\n", 4); // XXX: fails to check result
* int rc = read(0, p, 123);
* free(p);
* if (rc == ECANCELED) {
* pthread_exit(0); // XXX: not run if write() was cancelled
* }
*
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
* Isn't safe to use in masked mode. That's because if a cancelation
* occurs during the write() operation then cancelations are blocked
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
* while running read(). MASKED MODE DOESN'T HAVE SECOND CHANCES. You
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
* must rigorously check the results of each cancelation point call.
Perform some code maintenance - Change IDT code so kprintf() isn't mandatory dependency - Document current intentions around pthread_cancel() - Make _npassert() an _unassert() in MODE=tiny
2022-10-09 20:00:46 +00:00
*
Make more fixes and improvements - Remove PAGESIZE constant - Fix realloc() documentation - Fix ttyname_r() error reporting - Make forking more reliable on Windows - Make execvp() a few microseconds faster - Make system() a few microseconds faster - Tighten up the socket-related magic numbers - Loosen restrictions on mmap() offset alignment - Improve GetProgramExecutableName() with getenv("_") - Use mkstemp() as basis for mktemp(), tmpfile(), tmpfd() - Fix flakes in pthread_cancel_test, unix_test, fork_test - Fix recently introduced futex stack overflow regression - Let sockets be passed as stdio to subprocesses on Windows - Improve security of bind() on Windows w/ SO_EXCLUSIVEADDRUSE
2023-07-30 01:44:15 +00:00
* Unit tests should be able to safely ignore the return value, or at
* the very least be programmed to consider ESRCH a successful status
*
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
* @param thread may be 0 to cancel all threads except self
Implement support for POSIX thread cancellations This change makes some miracle modifications to the System Five system call support, which lets us have safe, correct, and atomic handling of thread cancellations. It all turned out to be cheaper than anticipated because it wasn't necessary to modify the system call veneers. We were able to encode the cancellability of each system call into the magnums found in libc/sysv/syscalls.sh. Since cancellations are so waq, we are also supporting a lovely Musl Libc mask feature for raising ECANCELED.
2022-11-04 08:04:43 +00:00
* @return 0 on success, or errno on error
Make more fixes and improvements - Remove PAGESIZE constant - Fix realloc() documentation - Fix ttyname_r() error reporting - Make forking more reliable on Windows - Make execvp() a few microseconds faster - Make system() a few microseconds faster - Tighten up the socket-related magic numbers - Loosen restrictions on mmap() offset alignment - Improve GetProgramExecutableName() with getenv("_") - Use mkstemp() as basis for mktemp(), tmpfile(), tmpfd() - Fix flakes in pthread_cancel_test, unix_test, fork_test - Fix recently introduced futex stack overflow regression - Let sockets be passed as stdio to subprocesses on Windows - Improve security of bind() on Windows w/ SO_EXCLUSIVEADDRUSE
2023-07-30 01:44:15 +00:00
* @raise ESRCH if system thread wasn't alive or we lost a race
Perform some code maintenance - Change IDT code so kprintf() isn't mandatory dependency - Document current intentions around pthread_cancel() - Make _npassert() an _unassert() in MODE=tiny
2022-10-09 20:00:46 +00:00
*/
Make futexes cancellable by pthreads
2022-11-05 01:19:05 +00:00
errno_t pthread_cancel(pthread_t thread) {
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
struct PosixThread *arg;
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
if ((arg = (struct PosixThread *)thread)) {
Fix some issues
2023-10-10 03:18:48 +00:00
return _pthread_cancel_single(arg);
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
} else {
Fix some issues
2023-10-10 03:18:48 +00:00
return _pthread_cancel_everyone();
Implement support for POSIX thread cancellations This change makes some miracle modifications to the System Five system call support, which lets us have safe, correct, and atomic handling of thread cancellations. It all turned out to be cheaper than anticipated because it wasn't necessary to modify the system call veneers. We were able to encode the cancellability of each system call into the magnums found in libc/sysv/syscalls.sh. Since cancellations are so waq, we are also supporting a lovely Musl Libc mask feature for raising ECANCELED.
2022-11-04 08:04:43 +00:00
}
}
/**
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
* Creates cancelation point in calling thread.
Implement support for POSIX thread cancellations This change makes some miracle modifications to the System Five system call support, which lets us have safe, correct, and atomic handling of thread cancellations. It all turned out to be cheaper than anticipated because it wasn't necessary to modify the system call veneers. We were able to encode the cancellability of each system call into the magnums found in libc/sysv/syscalls.sh. Since cancellations are so waq, we are also supporting a lovely Musl Libc mask feature for raising ECANCELED.
2022-11-04 08:04:43 +00:00
*
* This function can be used to force `PTHREAD_CANCEL_DEFERRED` threads
* to cancel without needing to invoke an interruptible system call. If
* the calling thread is in the `PTHREAD_CANCEL_DISABLE` then this will
* do nothing. If the calling thread hasn't yet been cancelled, this'll
* do nothing. In `PTHREAD_CANCEL_MASKED` mode, this also does nothing.
*
* @see pthread_testcancel_np()
*/
void pthread_testcancel(void) {
struct PosixThread *pt;
if (!__tls_enabled) return;
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
if (!(pt = _pthread_self())) return;
if (pt->pt_flags & PT_NOCANCEL) return;
if ((!(pt->pt_flags & PT_MASKED) || (pt->pt_flags & PT_ASYNC)) &&
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
atomic_load_explicit(&pt->pt_canceled, memory_order_acquire)) {
Implement support for POSIX thread cancellations This change makes some miracle modifications to the System Five system call support, which lets us have safe, correct, and atomic handling of thread cancellations. It all turned out to be cheaper than anticipated because it wasn't necessary to modify the system call veneers. We were able to encode the cancellability of each system call into the magnums found in libc/sysv/syscalls.sh. Since cancellations are so waq, we are also supporting a lovely Musl Libc mask feature for raising ECANCELED.
2022-11-04 08:04:43 +00:00
pthread_exit(PTHREAD_CANCELED);
}
}
/**
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
* Creates cancelation point in calling thread.
Implement support for POSIX thread cancellations This change makes some miracle modifications to the System Five system call support, which lets us have safe, correct, and atomic handling of thread cancellations. It all turned out to be cheaper than anticipated because it wasn't necessary to modify the system call veneers. We were able to encode the cancellability of each system call into the magnums found in libc/sysv/syscalls.sh. Since cancellations are so waq, we are also supporting a lovely Musl Libc mask feature for raising ECANCELED.
2022-11-04 08:04:43 +00:00
*
* This function can be used to force `PTHREAD_CANCEL_DEFERRED` threads
* to cancel without needing to invoke an interruptible system call. If
* the calling thread is in the `PTHREAD_CANCEL_DISABLE` then this will
* do nothing. If the calling thread hasn't yet been cancelled, this'll
* do nothing. If the calling thread uses `PTHREAD_CANCEL_MASKED`, then
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
* this function returns `ECANCELED` if a cancelation occurred, rather
Implement support for POSIX thread cancellations This change makes some miracle modifications to the System Five system call support, which lets us have safe, correct, and atomic handling of thread cancellations. It all turned out to be cheaper than anticipated because it wasn't necessary to modify the system call veneers. We were able to encode the cancellability of each system call into the magnums found in libc/sysv/syscalls.sh. Since cancellations are so waq, we are also supporting a lovely Musl Libc mask feature for raising ECANCELED.
2022-11-04 08:04:43 +00:00
* than the normal behavior which is to destroy and cleanup the thread.
* Any `ECANCELED` result must not be ignored, because the thread shall
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
* have cancelations disabled once it occurs.
Implement support for POSIX thread cancellations This change makes some miracle modifications to the System Five system call support, which lets us have safe, correct, and atomic handling of thread cancellations. It all turned out to be cheaper than anticipated because it wasn't necessary to modify the system call veneers. We were able to encode the cancellability of each system call into the magnums found in libc/sysv/syscalls.sh. Since cancellations are so waq, we are also supporting a lovely Musl Libc mask feature for raising ECANCELED.
2022-11-04 08:04:43 +00:00
*
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
* @return 0 if not cancelled or cancelation is blocked or `ECANCELED`
Implement support for POSIX thread cancellations This change makes some miracle modifications to the System Five system call support, which lets us have safe, correct, and atomic handling of thread cancellations. It all turned out to be cheaper than anticipated because it wasn't necessary to modify the system call veneers. We were able to encode the cancellability of each system call into the magnums found in libc/sysv/syscalls.sh. Since cancellations are so waq, we are also supporting a lovely Musl Libc mask feature for raising ECANCELED.
2022-11-04 08:04:43 +00:00
* in masked mode when the calling thread has been cancelled
*/
Make futexes cancellable by pthreads
2022-11-05 01:19:05 +00:00
errno_t pthread_testcancel_np(void) {
Implement support for POSIX thread cancellations This change makes some miracle modifications to the System Five system call support, which lets us have safe, correct, and atomic handling of thread cancellations. It all turned out to be cheaper than anticipated because it wasn't necessary to modify the system call veneers. We were able to encode the cancellability of each system call into the magnums found in libc/sysv/syscalls.sh. Since cancellations are so waq, we are also supporting a lovely Musl Libc mask feature for raising ECANCELED.
2022-11-04 08:04:43 +00:00
struct PosixThread *pt;
if (!__tls_enabled) return 0;
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
if (!(pt = _pthread_self())) return 0;
if (pt->pt_flags & PT_NOCANCEL) return 0;
Make improvements - We now serialize the file descriptor table when spawning / executing processes on Windows. This means you can now inherit more stuff than just standard i/o. It's needed by bash, which duplicates the console to file descriptor #255. We also now do a better job serializing the environment variables, so you're less likely to encounter E2BIG when using your bash shell. We also no longer coerce environ to uppercase - execve() on Windows now remotely controls its parent process to make them spawn a replacement for itself. Then it'll be able to terminate immediately once the spawn succeeds, without having to linger around for the lifetime as a shell process for proxying the exit code. When process worker thread running in the parent sees the child die, it's given a handle to the new child, to replace it in the process table. - execve() and posix_spawn() on Windows will now provide CreateProcess an explicit handle list. This allows us to remove handle locks which enables better fork/spawn concurrency, with seriously correct thread safety. Other codebases like Go use the same technique. On the other hand fork() still favors the conventional WIN32 inheritence approach which can be a little bit messy, but is *controlled* by guaranteeing perfectly clean slates at both the spawning and execution boundaries - sigset_t is now 64 bits. Having it be 128 bits was a mistake because there's no reason to use that and it's only supported by FreeBSD. By using the system word size, signal mask manipulation on Windows goes very fast. Furthermore @asyncsignalsafe funcs have been rewritten on Windows to take advantage of signal masking, now that it's much more pleasant to use. - All the overlapped i/o code on Windows has been rewritten for pretty good signal and cancelation safety. We're now able to ensure overlap data structures are cleaned up so long as you don't longjmp() out of out of a signal handler that interrupted an i/o operation. Latencies are also improved thanks to the removal of lots of "busy wait" code. Waits should be optimal for everything except poll(), which shall be the last and final demon we slay in the win32 i/o horror show. - getrusage() on Windows is now able to report RUSAGE_CHILDREN as well as RUSAGE_SELF, thanks to aggregation in the process manager thread.
2023-10-08 12:36:18 +00:00
if (!atomic_load_explicit(&pt->pt_canceled, memory_order_acquire)) return 0;
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
if (!(pt->pt_flags & PT_MASKED) || (pt->pt_flags & PT_ASYNC)) {
Implement support for POSIX thread cancellations This change makes some miracle modifications to the System Five system call support, which lets us have safe, correct, and atomic handling of thread cancellations. It all turned out to be cheaper than anticipated because it wasn't necessary to modify the system call veneers. We were able to encode the cancellability of each system call into the magnums found in libc/sysv/syscalls.sh. Since cancellations are so waq, we are also supporting a lovely Musl Libc mask feature for raising ECANCELED.
2022-11-04 08:04:43 +00:00
pthread_exit(PTHREAD_CANCELED);
} else {
Make improvements - Every unit test now passes on Apple Silicon. The final piece of this puzzle was porting our POSIX threads cancelation support, since that works differently on ARM64 XNU vs. AMD64. Our semaphore support on Apple Silicon is also superior now compared to AMD64, thanks to the grand central dispatch library which lets *NSYNC locks go faster. - The Cosmopolitan runtime is now more stable, particularly on Windows. To do this, thread local storage is mandatory at all runtime levels, and the innermost packages of the C library is no longer being built using ASAN. TLS is being bootstrapped with a 128-byte TIB during the process startup phase, and then later on the runtime re-allocates it either statically or dynamically to support code using _Thread_local. fork() and execve() now do a better job cooperating with threads. We can now check how much stack memory is left in the process or thread when functions like kprintf() / execve() etc. call alloca(), so that ENOMEM can be raised, reduce a buffer size, or just print a warning. - POSIX signal emulation is now implemented the same way kernels do it with pthread_kill() and raise(). Any thread can interrupt any other thread, regardless of what it's doing. If it's blocked on read/write then the killer thread will cancel its i/o operation so that EINTR can be returned in the mark thread immediately. If it's doing a tight CPU bound operation, then that's also interrupted by the signal delivery. Signal delivery works now by suspending a thread and pushing context data structures onto its stack, and redirecting its execution to a trampoline function, which calls SetThreadContext(GetCurrentThread()) when it's done. - We're now doing a better job managing locks and handles. On NetBSD we now close semaphore file descriptors in forked children. Semaphores on Windows can now be canceled immediately, which means mutexes/condition variables will now go faster. Apple Silicon semaphores can be canceled too. We're now using Apple's pthread_yield() funciton. Apple _nocancel syscalls are now used on XNU when appropriate to ensure pthread_cancel requests aren't lost. The MbedTLS library has been updated to support POSIX thread cancelations. See tool/build/runitd.c for an example of how it can be used for production multi-threaded tls servers. Handles on Windows now leak less often across processes. All i/o operations on Windows are now overlapped, which means file pointers can no longer be inherited across dup() and fork() for the time being. - We now spawn a thread on Windows to deliver SIGCHLD and wakeup wait4() which means, for example, that posix_spawn() now goes 3x faster. POSIX spawn is also now more correct. Like Musl, it's now able to report the failure code of execve() via a pipe although our approach favors using shared memory to do that on systems that have a true vfork() function. - We now spawn a thread to deliver SIGALRM to threads when setitimer() is used. This enables the most precise wakeups the OS makes possible. - The Cosmopolitan runtime now uses less memory. On NetBSD for example, it turned out the kernel would actually commit the PT_GNU_STACK size which caused RSS to be 6mb for every process. Now it's down to ~4kb. On Apple Silicon, we reduce the mandatory upstream thread size to the smallest possible size to reduce the memory overhead of Cosmo threads. The examples directory has a program called greenbean which can spawn a web server on Linux with 10,000 worker threads and have the memory usage of the process be ~77mb. The 1024 byte overhead of POSIX-style thread-local storage is now optional; it won't be allocated until the pthread_setspecific/getspecific functions are called. On Windows, the threads that get spawned which are internal to the libc implementation use reserve rather than commit memory, which shaves a few hundred kb. - sigaltstack() is now supported on Windows, however it's currently not able to be used to handle stack overflows, since crash signals are still generated by WIN32. However the crash handler will still switch to the alt stack, which is helpful in environments with tiny threads. - Test binaries are now smaller. Many of the mandatory dependencies of the test runner have been removed. This ensures many programs can do a better job only linking the the thing they're testing. This caused the test binaries for LIBC_FMT for example, to decrease from 200kb to 50kb - long double is no longer used in the implementation details of libc, except in the APIs that define it. The old code that used long double for time (instead of struct timespec) has now been thoroughly removed. - ShowCrashReports() is now much tinier in MODE=tiny. Instead of doing backtraces itself, it'll just print a command you can run on the shell using our new `cosmoaddr2line` program to view the backtrace. - Crash report signal handling now works in a much better way. Instead of terminating the process, it now relies on SA_RESETHAND so that the default SIG_IGN behavior can terminate the process if necessary. - Our pledge() functionality has now been fully ported to AARCH64 Linux.
2023-09-19 03:44:45 +00:00
pt->pt_flags |= PT_NOCANCEL;
Implement support for POSIX thread cancellations This change makes some miracle modifications to the System Five system call support, which lets us have safe, correct, and atomic handling of thread cancellations. It all turned out to be cheaper than anticipated because it wasn't necessary to modify the system call veneers. We were able to encode the cancellability of each system call into the magnums found in libc/sysv/syscalls.sh. Since cancellations are so waq, we are also supporting a lovely Musl Libc mask feature for raising ECANCELED.
2022-11-04 08:04:43 +00:00
return ECANCELED;
}
Make mutex calling code 10x tinier Calls to lock/unlock functions are now NOPs by default. The first time clone() is called, they get turned into CALL instructions. Doing this caused funcctions like fputc() to shrink from 85 bytes to 45+4 bytes. Since the ANSI solution of `(__threaded && lock())` inlines os much superfluous binary content into functions all over the place.
2022-06-13 02:33:42 +00:00
}
Reference in a new issue Copy permalink