2020-06-15 14:18:57 +00:00
|
|
|
/*-*- mode:c;indent-tabs-mode:nil;c-basic-offset:2;tab-width:8;coding:utf-8 -*-│
|
2023-12-05 22:37:54 +00:00
|
|
|
│ vi: set noet ft=c ts=2 sts=2 sw=2 fenc=utf-8 :vi │
|
2020-06-15 14:18:57 +00:00
|
|
|
╞══════════════════════════════════════════════════════════════════════════════╡
|
2022-04-28 16:42:36 +00:00
|
|
|
│ Copyright 2022 Justine Alexandra Roberts Tunney │
|
2020-06-15 14:18:57 +00:00
|
|
|
│ │
|
2020-12-28 01:18:44 +00:00
|
|
|
│ 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. │
|
2020-06-15 14:18:57 +00:00
|
|
|
│ │
|
2020-12-28 01:18:44 +00:00
|
|
|
│ 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. │
|
2020-06-15 14:18:57 +00:00
|
|
|
╚─────────────────────────────────────────────────────────────────────────────*/
|
|
|
|
#include "libc/calls/calls.h"
|
2022-04-28 16:42:36 +00:00
|
|
|
#include "libc/calls/internal.h"
|
2022-07-25 02:40:32 +00:00
|
|
|
#include "libc/calls/pledge.internal.h"
|
2023-06-10 01:02:06 +00:00
|
|
|
#include "libc/calls/struct/bpf.internal.h"
|
|
|
|
#include "libc/calls/struct/filter.internal.h"
|
2022-06-27 20:01:58 +00:00
|
|
|
#include "libc/calls/struct/flock.h"
|
2023-06-10 01:02:06 +00:00
|
|
|
#include "libc/calls/struct/seccomp.internal.h"
|
2022-07-08 13:29:24 +00:00
|
|
|
#include "libc/calls/struct/sigaction.h"
|
2022-09-19 22:31:16 +00:00
|
|
|
#include "libc/calls/struct/sigset.h"
|
2022-06-27 20:01:58 +00:00
|
|
|
#include "libc/calls/struct/stat.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/calls/syscall-sysv.internal.h"
|
2022-05-23 22:06:11 +00:00
|
|
|
#include "libc/calls/syscall_support-sysv.internal.h"
|
2022-04-28 16:42:36 +00:00
|
|
|
#include "libc/dce.h"
|
|
|
|
#include "libc/errno.h"
|
2022-07-08 13:29:24 +00:00
|
|
|
#include "libc/macros.internal.h"
|
2022-04-28 16:42:36 +00:00
|
|
|
#include "libc/mem/mem.h"
|
2022-07-20 04:18:33 +00:00
|
|
|
#include "libc/runtime/internal.h"
|
2021-01-25 21:08:05 +00:00
|
|
|
#include "libc/runtime/runtime.h"
|
2022-06-27 20:01:58 +00:00
|
|
|
#include "libc/sock/sock.h"
|
|
|
|
#include "libc/sock/struct/sockaddr.h"
|
2023-06-09 13:41:34 +00:00
|
|
|
#include "libc/sock/struct/sockaddr6.h"
|
2023-08-14 03:31:27 +00:00
|
|
|
#include "libc/stdio/internal.h"
|
2022-07-20 04:18:33 +00:00
|
|
|
#include "libc/stdio/stdio.h"
|
2022-06-27 20:01:58 +00:00
|
|
|
#include "libc/sysv/consts/af.h"
|
|
|
|
#include "libc/sysv/consts/at.h"
|
|
|
|
#include "libc/sysv/consts/f.h"
|
|
|
|
#include "libc/sysv/consts/fio.h"
|
|
|
|
#include "libc/sysv/consts/ipproto.h"
|
|
|
|
#include "libc/sysv/consts/map.h"
|
|
|
|
#include "libc/sysv/consts/nrlinux.h"
|
2022-04-28 16:42:36 +00:00
|
|
|
#include "libc/sysv/consts/o.h"
|
2022-06-27 20:01:58 +00:00
|
|
|
#include "libc/sysv/consts/pr.h"
|
|
|
|
#include "libc/sysv/consts/prot.h"
|
2022-06-09 03:01:28 +00:00
|
|
|
#include "libc/sysv/consts/sig.h"
|
2022-09-18 10:56:52 +00:00
|
|
|
#include "libc/sysv/consts/so.h"
|
2022-06-27 20:01:58 +00:00
|
|
|
#include "libc/sysv/consts/sock.h"
|
2022-09-18 10:56:52 +00:00
|
|
|
#include "libc/sysv/consts/sol.h"
|
2022-07-20 04:18:33 +00:00
|
|
|
#include "libc/testlib/ezbench.h"
|
2022-09-19 22:31:16 +00:00
|
|
|
#include "libc/testlib/subprocess.h"
|
2021-01-25 21:08:05 +00:00
|
|
|
#include "libc/testlib/testlib.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/thread/posixthread.internal.h"
|
2022-10-09 06:54:05 +00:00
|
|
|
#include "libc/thread/thread.h"
|
2022-07-20 04:18:33 +00:00
|
|
|
#include "libc/time/time.h"
|
2022-07-25 06:40:49 +00:00
|
|
|
#include "libc/x/x.h"
|
2021-01-25 21:08:05 +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
|
|
|
void SetUpOnce(void) {
|
|
|
|
testlib_enable_tmp_setup_teardown();
|
|
|
|
}
|
2022-06-27 20:01:58 +00:00
|
|
|
|
2022-07-08 13:29:24 +00:00
|
|
|
void OnSig(int sig) {
|
|
|
|
// do nothing
|
|
|
|
}
|
|
|
|
|
2022-09-13 18:20:35 +00:00
|
|
|
int sys_memfd_secret(unsigned int); // our ENOSYS threshold
|
2022-07-20 04:18:33 +00:00
|
|
|
|
2022-07-18 14:23:15 +00:00
|
|
|
void SetUp(void) {
|
2023-06-03 15:12:13 +00:00
|
|
|
if (pledge(0, 0) == -1) {
|
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
|
|
|
fprintf(stderr, "warning: pledge() not supported on this system %m\n");
|
2023-06-03 15:12:13 +00:00
|
|
|
exit(0);
|
|
|
|
}
|
2022-10-02 14:03:12 +00:00
|
|
|
testlib_extract("/zip/life.elf", "life.elf", 0755);
|
|
|
|
testlib_extract("/zip/sock.elf", "sock.elf", 0755);
|
2022-08-11 18:27:25 +00:00
|
|
|
__pledge_mode = PLEDGE_PENALTY_RETURN_EPERM;
|
2021-01-25 21:08:05 +00:00
|
|
|
}
|
2020-06-15 14:18:57 +00:00
|
|
|
|
2022-04-28 16:42:36 +00:00
|
|
|
TEST(pledge, default_allowsExit) {
|
2022-07-14 11:32:33 +00:00
|
|
|
int *job;
|
2021-01-25 21:08:05 +00:00
|
|
|
int ws, pid;
|
2022-07-14 11:32:33 +00:00
|
|
|
// create small shared memory region
|
|
|
|
ASSERT_NE(-1, (job = mmap(0, FRAMESIZE, PROT_READ | PROT_WRITE,
|
|
|
|
MAP_SHARED | MAP_ANONYMOUS, -1, 0)));
|
|
|
|
job[0] = 2; // create workload
|
|
|
|
job[1] = 2;
|
|
|
|
ASSERT_NE(-1, (pid = fork())); // create enclaved worker
|
2021-01-25 21:08:05 +00:00
|
|
|
if (!pid) {
|
2022-04-28 16:42:36 +00:00
|
|
|
ASSERT_SYS(0, 0, pledge("", 0));
|
2022-07-14 11:32:33 +00:00
|
|
|
job[0] = job[0] + job[1]; // do work
|
2022-04-28 16:42:36 +00:00
|
|
|
_Exit(0);
|
2020-06-15 14:18:57 +00:00
|
|
|
}
|
2022-07-14 11:32:33 +00:00
|
|
|
EXPECT_NE(-1, wait(&ws)); // wait for worker
|
2022-04-28 16:42:36 +00:00
|
|
|
EXPECT_TRUE(WIFEXITED(ws));
|
|
|
|
EXPECT_EQ(0, WEXITSTATUS(ws));
|
2022-07-14 11:32:33 +00:00
|
|
|
EXPECT_EQ(4, job[0]); // check result
|
|
|
|
EXPECT_SYS(0, 0, munmap(job, FRAMESIZE));
|
|
|
|
}
|
|
|
|
|
2022-08-07 23:18:33 +00:00
|
|
|
TEST(pledge, execpromises_notok) {
|
|
|
|
if (IsOpenbsd()) return; // b/c testing linux bpf
|
|
|
|
int ws, pid;
|
|
|
|
ASSERT_NE(-1, (pid = fork()));
|
|
|
|
if (!pid) {
|
2022-08-11 18:27:25 +00:00
|
|
|
__pledge_mode = PLEDGE_PENALTY_RETURN_EPERM;
|
2022-08-07 23:18:33 +00:00
|
|
|
ASSERT_SYS(0, 0, pledge("stdio rpath exec", "stdio"));
|
|
|
|
execl("sock.elf", "sock.elf", 0);
|
|
|
|
_Exit(127);
|
|
|
|
}
|
|
|
|
EXPECT_NE(-1, wait(&ws));
|
|
|
|
EXPECT_TRUE(WIFEXITED(ws));
|
|
|
|
EXPECT_EQ(129, WEXITSTATUS(ws));
|
|
|
|
}
|
|
|
|
|
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
|
|
|
void *Enclave(void *arg) {
|
|
|
|
sigset_t ss;
|
|
|
|
sigfillset(&ss);
|
|
|
|
sigprocmask(SIG_BLOCK, &ss, 0);
|
2022-07-14 11:32:33 +00:00
|
|
|
ASSERT_SYS(0, 0, pledge("", 0));
|
|
|
|
int *job = arg; // get job
|
|
|
|
job[0] = job[0] + job[1]; // do work
|
|
|
|
return 0; // exit
|
|
|
|
}
|
|
|
|
|
2023-10-12 03:26:28 +00:00
|
|
|
TEST(pledge, tester) {
|
|
|
|
SPAWN(fork);
|
|
|
|
ASSERT_EQ(0, pledge("stdio rpath wpath cpath proc exec", NULL));
|
|
|
|
EXITS(0);
|
|
|
|
}
|
|
|
|
|
2022-07-14 11:32:33 +00:00
|
|
|
TEST(pledge, withThreadMemory) {
|
|
|
|
if (IsOpenbsd()) return; // openbsd doesn't allow it, wisely
|
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
|
|
|
pthread_t worker;
|
|
|
|
int job[2] = {2, 2}; // create workload
|
|
|
|
ASSERT_EQ(0, pthread_create(&worker, 0, Enclave, job)); // create worker
|
|
|
|
ASSERT_EQ(0, pthread_join(worker, 0)); // wait for exit
|
|
|
|
EXPECT_EQ(4, job[0]); // check result
|
2020-06-15 14:18:57 +00:00
|
|
|
}
|
2022-04-20 16:56:53 +00:00
|
|
|
|
2022-09-19 22:31:16 +00:00
|
|
|
bool gotusr1;
|
|
|
|
|
|
|
|
void OnUsr1(int sig) {
|
|
|
|
gotusr1 = true;
|
|
|
|
}
|
|
|
|
|
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
|
|
|
void *TgkillWorker(void *arg) {
|
2022-09-19 22:31:16 +00:00
|
|
|
sigset_t mask;
|
|
|
|
signal(SIGUSR1, OnUsr1);
|
|
|
|
sigemptyset(&mask);
|
|
|
|
ASSERT_SYS(EINTR, -1, sigsuspend(&mask));
|
|
|
|
ASSERT_TRUE(gotusr1);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST(pledge, tgkill) {
|
|
|
|
// https://github.com/jart/cosmopolitan/issues/628
|
|
|
|
if (!IsLinux()) return;
|
|
|
|
sigset_t mask;
|
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
|
|
|
pthread_t worker;
|
2022-09-19 22:31:16 +00:00
|
|
|
SPAWN(fork);
|
|
|
|
sigemptyset(&mask);
|
|
|
|
sigaddset(&mask, SIGUSR1);
|
|
|
|
sigprocmask(SIG_BLOCK, &mask, 0);
|
|
|
|
ASSERT_SYS(0, 0, pledge("stdio", 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
|
|
|
ASSERT_SYS(0, 0, pthread_create(&worker, 0, TgkillWorker, 0));
|
|
|
|
ASSERT_SYS(0, 0,
|
|
|
|
sys_tgkill(getpid(), _pthread_tid((struct PosixThread *)worker),
|
|
|
|
SIGUSR1));
|
|
|
|
ASSERT_SYS(0, 0, pthread_join(worker, 0));
|
2022-09-19 22:31:16 +00:00
|
|
|
EXITS(0);
|
|
|
|
}
|
|
|
|
|
2022-07-08 13:29:24 +00:00
|
|
|
TEST(pledge, stdio_forbidsOpeningPasswd1) {
|
|
|
|
if (!IsLinux()) return;
|
2022-04-28 16:42:36 +00:00
|
|
|
int ws, pid;
|
2022-04-20 16:56:53 +00:00
|
|
|
ASSERT_NE(-1, (pid = fork()));
|
2022-04-28 16:42:36 +00:00
|
|
|
if (!pid) {
|
|
|
|
ASSERT_SYS(0, 0, pledge("stdio", 0));
|
|
|
|
ASSERT_SYS(EPERM, -1, open("/etc/passwd", O_RDWR));
|
|
|
|
_Exit(0);
|
|
|
|
}
|
|
|
|
EXPECT_NE(-1, wait(&ws));
|
2022-07-08 13:29:24 +00:00
|
|
|
EXPECT_TRUE(WIFEXITED(ws));
|
|
|
|
EXPECT_EQ(0, WEXITSTATUS(ws));
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST(pledge, stdio_forbidsOpeningPasswd2) {
|
|
|
|
int ws, pid;
|
2022-08-11 18:27:25 +00:00
|
|
|
__pledge_mode = PLEDGE_PENALTY_KILL_PROCESS;
|
2022-07-08 13:29:24 +00:00
|
|
|
ASSERT_NE(-1, (pid = fork()));
|
|
|
|
if (!pid) {
|
|
|
|
ASSERT_SYS(0, 0, pledge("stdio", 0));
|
|
|
|
ASSERT_SYS(EPERM, -1, open("/etc/passwd", O_RDWR));
|
|
|
|
_Exit(0);
|
|
|
|
}
|
|
|
|
EXPECT_NE(-1, wait(&ws));
|
|
|
|
EXPECT_TRUE(WIFSIGNALED(ws));
|
2022-08-20 19:32:51 +00:00
|
|
|
EXPECT_EQ(IsOpenbsd() ? SIGABRT : SIGSYS, WTERMSIG(ws));
|
2022-07-08 13:29:24 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
TEST(pledge, multipleCalls_canOnlyBecomeMoreRestrictive1) {
|
|
|
|
if (IsOpenbsd()) return;
|
|
|
|
int ws, pid;
|
|
|
|
ASSERT_NE(-1, (pid = fork()));
|
|
|
|
if (!pid) {
|
|
|
|
ASSERT_SYS(0, 0, pledge("stdio", 0));
|
|
|
|
ASSERT_SYS(0, 0, pledge("stdio unix", 0));
|
|
|
|
ASSERT_SYS(0, 3, dup(2));
|
|
|
|
ASSERT_SYS(EPERM, -1, socket(AF_UNIX, SOCK_STREAM, 0));
|
2022-07-13 10:08:16 +00:00
|
|
|
ASSERT_SYS(0, 2, prctl(PR_GET_SECCOMP, 0, 0));
|
|
|
|
ASSERT_SYS(0, 0, prctl(PR_SET_NO_NEW_PRIVS, 1, 0, 0, 0));
|
|
|
|
ASSERT_SYS(EINVAL, -1, prctl(PR_SET_NO_NEW_PRIVS, 0, 0, 0, 0));
|
2022-07-08 13:29:24 +00:00
|
|
|
_Exit(0);
|
|
|
|
}
|
|
|
|
EXPECT_NE(-1, wait(&ws));
|
|
|
|
EXPECT_TRUE(WIFEXITED(ws));
|
|
|
|
EXPECT_EQ(0, WEXITSTATUS(ws));
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST(pledge, multipleCalls_canOnlyBecomeMoreRestrictive2) {
|
|
|
|
if (!IsOpenbsd()) return;
|
|
|
|
int ws, pid;
|
|
|
|
ASSERT_NE(-1, (pid = fork()));
|
|
|
|
if (!pid) {
|
2022-07-18 14:23:15 +00:00
|
|
|
ASSERT_SYS(0, 0, pledge("stdio", "stdio"));
|
|
|
|
ASSERT_SYS(EPERM, -1, pledge("stdio unix", "stdio unix"));
|
2022-07-08 13:29:24 +00:00
|
|
|
_Exit(0);
|
|
|
|
}
|
|
|
|
EXPECT_NE(-1, wait(&ws));
|
|
|
|
EXPECT_TRUE(WIFEXITED(ws));
|
|
|
|
EXPECT_EQ(0, WEXITSTATUS(ws));
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST(pledge, multipleCalls_canOnlyBecomeMoreRestrictive3) {
|
|
|
|
if (!IsOpenbsd()) return;
|
|
|
|
int ws, pid;
|
|
|
|
ASSERT_NE(-1, (pid = fork()));
|
|
|
|
if (!pid) {
|
|
|
|
ASSERT_SYS(0, 0, pledge("stdio unix", 0));
|
|
|
|
ASSERT_SYS(0, 0, pledge("stdio", 0));
|
|
|
|
ASSERT_SYS(0, 3, dup(2));
|
|
|
|
_Exit(0);
|
2022-04-28 16:42:36 +00:00
|
|
|
}
|
2022-07-08 13:29:24 +00:00
|
|
|
EXPECT_NE(-1, wait(&ws));
|
|
|
|
EXPECT_TRUE(WIFEXITED(ws));
|
|
|
|
EXPECT_EQ(0, WEXITSTATUS(ws));
|
2022-04-20 16:56:53 +00:00
|
|
|
}
|
2022-06-27 20:01:58 +00:00
|
|
|
|
|
|
|
TEST(pledge, stdio_fcntl_allowsSomeFirstArgs) {
|
|
|
|
if (IsOpenbsd()) return; // b/c testing linux bpf
|
|
|
|
int ws, pid;
|
|
|
|
struct flock lk;
|
|
|
|
ASSERT_NE(-1, (pid = fork()));
|
|
|
|
if (!pid) {
|
2022-07-08 13:29:24 +00:00
|
|
|
ASSERT_SYS(0, 0, pledge("stdio inet", 0));
|
2022-06-27 20:01:58 +00:00
|
|
|
ASSERT_SYS(0, 0, pledge("stdio", 0));
|
|
|
|
ASSERT_NE(-1, fcntl(0, F_GETFL));
|
|
|
|
ASSERT_SYS(0, 0, fcntl(0, F_GETFD));
|
|
|
|
ASSERT_SYS(0, 3, fcntl(2, F_DUPFD_CLOEXEC, 3));
|
|
|
|
ASSERT_SYS(0, 0, ioctl(0, FIOCLEX, 0));
|
|
|
|
ASSERT_SYS(EPERM, 0, isatty(0));
|
2022-10-13 20:44:41 +00:00
|
|
|
ASSERT_SYS(EPERM, -1, fcntl(0, 777));
|
2022-06-27 20:01:58 +00:00
|
|
|
ASSERT_SYS(EPERM, -1, fcntl(0, F_GETLK, &lk));
|
|
|
|
ASSERT_SYS(EPERM, -1, fcntl(0, F_NOTIFY));
|
|
|
|
ASSERT_SYS(EPERM, -1, socket(AF_INET, SOCK_STREAM, IPPROTO_TCP));
|
|
|
|
_Exit(0);
|
|
|
|
}
|
|
|
|
EXPECT_NE(-1, wait(&ws));
|
|
|
|
EXPECT_TRUE(WIFEXITED(ws) && !WEXITSTATUS(ws));
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST(pledge, stdioTty_sendtoRestricted_requiresNullAddr) {
|
|
|
|
if (IsOpenbsd()) return; // b/c testing linux bpf
|
|
|
|
int ws, pid, sv[2];
|
|
|
|
ASSERT_SYS(0, 0, socketpair(AF_UNIX, SOCK_STREAM, 0, sv));
|
|
|
|
ASSERT_NE(-1, (pid = fork()));
|
|
|
|
if (!pid) {
|
|
|
|
ASSERT_SYS(0, 0, pledge("stdio tty", 0));
|
|
|
|
ASSERT_SYS(0, 5, send(sv[0], "hello", 5, 0));
|
|
|
|
ASSERT_SYS(0, 5, sendto(sv[0], "hello", 5, 0, 0, 0));
|
|
|
|
isatty(0);
|
|
|
|
ASSERT_NE(EPERM, errno);
|
|
|
|
errno = 0;
|
2022-07-08 13:29:24 +00:00
|
|
|
// set lower 32-bit word of pointer to zero lool
|
|
|
|
struct sockaddr_in *sin =
|
|
|
|
mmap((void *)0x300000000000, FRAMESIZE, PROT_READ | PROT_WRITE,
|
|
|
|
MAP_ANONYMOUS | MAP_PRIVATE, -1, 0);
|
|
|
|
sin->sin_family = AF_INET;
|
2022-10-11 00:52:41 +00:00
|
|
|
ASSERT_SYS(
|
|
|
|
EPERM, -1,
|
|
|
|
sendto(sv[0], "hello", 5, 0, (struct sockaddr *)sin, sizeof(*sin)));
|
2022-06-27 20:01:58 +00:00
|
|
|
_Exit(0);
|
|
|
|
}
|
|
|
|
close(sv[0]);
|
|
|
|
close(sv[1]);
|
|
|
|
EXPECT_NE(-1, wait(&ws));
|
|
|
|
EXPECT_TRUE(WIFEXITED(ws) && !WEXITSTATUS(ws));
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST(pledge, unix_forbidsInetSockets) {
|
|
|
|
if (IsOpenbsd()) return; // b/c testing linux bpf
|
|
|
|
int ws, pid;
|
|
|
|
ASSERT_NE(-1, (pid = fork()));
|
|
|
|
if (!pid) {
|
|
|
|
ASSERT_SYS(0, 0, pledge("stdio unix", 0));
|
|
|
|
ASSERT_SYS(0, 3, socket(AF_UNIX, SOCK_STREAM, 0));
|
2022-07-08 13:29:24 +00:00
|
|
|
ASSERT_SYS(0, 4, socket(AF_UNIX, SOCK_STREAM | SOCK_CLOEXEC, 0));
|
2022-06-27 20:01:58 +00:00
|
|
|
ASSERT_SYS(EPERM, -1, socket(AF_INET, SOCK_STREAM, IPPROTO_TCP));
|
|
|
|
ASSERT_SYS(EPERM, -1, socket(AF_INET6, SOCK_STREAM, IPPROTO_TCP));
|
2022-07-08 13:29:24 +00:00
|
|
|
ASSERT_SYS(EPERM, -1, socket(AF_UNIX, SOCK_RAW, 0));
|
|
|
|
ASSERT_SYS(EPERM, -1, socket(AF_UNIX, SOCK_STREAM, 1));
|
2022-06-27 20:01:58 +00:00
|
|
|
_Exit(0);
|
|
|
|
}
|
|
|
|
EXPECT_NE(-1, wait(&ws));
|
|
|
|
EXPECT_TRUE(WIFEXITED(ws) && !WEXITSTATUS(ws));
|
|
|
|
}
|
|
|
|
|
2022-07-25 06:40:49 +00:00
|
|
|
TEST(pledge, wpath_doesNotImplyRpath) {
|
|
|
|
int ws, pid;
|
|
|
|
bool *gotsome;
|
|
|
|
ASSERT_NE(-1, (gotsome = _mapshared(FRAMESIZE)));
|
|
|
|
ASSERT_SYS(0, 0, touch("foo", 0644));
|
|
|
|
ASSERT_NE(-1, (pid = fork()));
|
|
|
|
if (!pid) {
|
|
|
|
ASSERT_SYS(0, 0, pledge("stdio wpath", 0));
|
|
|
|
ASSERT_SYS(0, 3, open("foo", O_WRONLY));
|
|
|
|
*gotsome = true;
|
|
|
|
ASSERT_SYS(EPERM, -1, open("foo", O_RDONLY));
|
|
|
|
_Exit(0);
|
|
|
|
}
|
|
|
|
EXPECT_NE(-1, wait(&ws));
|
|
|
|
ASSERT_TRUE(*gotsome);
|
|
|
|
if (IsOpenbsd()) {
|
|
|
|
ASSERT_TRUE(WIFSIGNALED(ws));
|
|
|
|
ASSERT_EQ(SIGABRT, WTERMSIG(ws));
|
|
|
|
} else {
|
|
|
|
ASSERT_TRUE(WIFEXITED(ws));
|
|
|
|
ASSERT_EQ(0, WEXITSTATUS(ws));
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2022-06-27 20:01:58 +00:00
|
|
|
TEST(pledge, inet_forbidsOtherSockets) {
|
|
|
|
if (IsOpenbsd()) return; // b/c testing linux bpf
|
2022-09-18 10:56:52 +00:00
|
|
|
int ws, pid, yes = 1;
|
2022-06-27 20:01:58 +00:00
|
|
|
ASSERT_NE(-1, (pid = fork()));
|
|
|
|
if (!pid) {
|
|
|
|
ASSERT_SYS(0, 0, pledge("stdio inet", 0));
|
2022-07-08 13:29:24 +00:00
|
|
|
ASSERT_SYS(0, 3, socket(AF_INET, SOCK_DGRAM, IPPROTO_UDP));
|
|
|
|
ASSERT_SYS(0, 4, socket(AF_INET, SOCK_DGRAM, IPPROTO_UDP));
|
|
|
|
ASSERT_SYS(0, 5, socket(AF_INET, SOCK_STREAM, IPPROTO_TCP));
|
|
|
|
ASSERT_SYS(0, 6, socket(AF_INET6, SOCK_DGRAM, IPPROTO_UDP));
|
|
|
|
ASSERT_SYS(0, 7, socket(AF_INET6, SOCK_DGRAM | SOCK_NONBLOCK, IPPROTO_UDP));
|
2022-09-18 10:56:52 +00:00
|
|
|
ASSERT_SYS(0, 0, setsockopt(3, SOL_SOCKET, SO_BROADCAST, &yes, 4));
|
2022-06-27 20:01:58 +00:00
|
|
|
ASSERT_SYS(EPERM, -1, socket(AF_UNIX, SOCK_STREAM, 0));
|
2022-07-08 13:29:24 +00:00
|
|
|
ASSERT_SYS(EPERM, -1, socket(AF_BLUETOOTH, SOCK_DGRAM, IPPROTO_UDP));
|
|
|
|
ASSERT_SYS(EPERM, -1, socket(AF_INET, SOCK_RAW, IPPROTO_UDP));
|
|
|
|
ASSERT_SYS(EPERM, -1, socket(AF_INET, SOCK_DGRAM, IPPROTO_RAW));
|
|
|
|
struct sockaddr_in sin = {AF_INET, 0, {htonl(0x7f000001)}};
|
2022-10-11 00:52:41 +00:00
|
|
|
ASSERT_SYS(0, 0, bind(4, (struct sockaddr *)&sin, sizeof(sin)));
|
2023-06-09 13:41:34 +00:00
|
|
|
struct sockaddr_in6 sin6 = {.sin6_family = AF_INET6,
|
|
|
|
.sin6_addr.s6_addr[15] = 1};
|
|
|
|
ASSERT_SYS(0, 0, bind(6, (struct sockaddr *)&sin6, sizeof(sin6)));
|
2022-07-08 13:29:24 +00:00
|
|
|
uint32_t az = sizeof(sin);
|
2022-10-11 00:52:41 +00:00
|
|
|
ASSERT_SYS(0, 0, getsockname(4, (struct sockaddr *)&sin, &az));
|
|
|
|
ASSERT_SYS(0, 5,
|
|
|
|
sendto(3, "hello", 5, 0, (struct sockaddr *)&sin, sizeof(sin)));
|
2022-06-27 20:01:58 +00:00
|
|
|
_Exit(0);
|
|
|
|
}
|
|
|
|
EXPECT_NE(-1, wait(&ws));
|
|
|
|
EXPECT_TRUE(WIFEXITED(ws) && !WEXITSTATUS(ws));
|
|
|
|
}
|
|
|
|
|
2023-06-19 01:06:47 +00:00
|
|
|
TEST(pledge, anet_forbidsUdpSocketsAndConnect) {
|
|
|
|
if (IsOpenbsd()) return; // b/c testing linux bpf
|
2023-09-02 03:49:13 +00:00
|
|
|
int ws, pid;
|
2023-06-19 01:06:47 +00:00
|
|
|
ASSERT_NE(-1, (pid = fork()));
|
|
|
|
if (!pid) {
|
|
|
|
ASSERT_SYS(0, 0, pledge("stdio anet", 0));
|
|
|
|
ASSERT_SYS(0, 3, socket(AF_INET, SOCK_STREAM, IPPROTO_TCP));
|
|
|
|
ASSERT_SYS(EPERM, -1, socket(AF_INET, SOCK_DGRAM, IPPROTO_UDP));
|
|
|
|
struct sockaddr_in sin = {AF_INET, 0, {htonl(0x7f000001)}};
|
|
|
|
ASSERT_SYS(EPERM, -1, connect(4, (struct sockaddr *)&sin, sizeof(sin)));
|
|
|
|
_Exit(0);
|
|
|
|
}
|
|
|
|
EXPECT_NE(-1, wait(&ws));
|
|
|
|
EXPECT_EQ(0, ws);
|
|
|
|
}
|
|
|
|
|
2022-06-27 20:01:58 +00:00
|
|
|
TEST(pledge, mmap) {
|
|
|
|
if (IsOpenbsd()) return; // b/c testing linux bpf
|
|
|
|
char *p;
|
|
|
|
int ws, pid;
|
|
|
|
ASSERT_NE(-1, (pid = fork()));
|
|
|
|
if (!pid) {
|
|
|
|
ASSERT_SYS(0, 0, pledge("stdio", 0));
|
|
|
|
ASSERT_NE(MAP_FAILED, (p = mmap(0, FRAMESIZE, PROT_READ | PROT_WRITE,
|
|
|
|
MAP_ANONYMOUS | MAP_PRIVATE, -1, 0)));
|
|
|
|
ASSERT_SYS(0, 0, mprotect(p, FRAMESIZE, PROT_READ));
|
|
|
|
ASSERT_SYS(EPERM, MAP_FAILED,
|
|
|
|
mprotect(p, FRAMESIZE, PROT_READ | PROT_EXEC));
|
|
|
|
ASSERT_SYS(EPERM, MAP_FAILED,
|
|
|
|
mmap(0, FRAMESIZE, PROT_EXEC | PROT_READ,
|
|
|
|
MAP_ANONYMOUS | MAP_PRIVATE, -1, 0));
|
|
|
|
_Exit(0);
|
|
|
|
}
|
|
|
|
EXPECT_NE(-1, wait(&ws));
|
|
|
|
EXPECT_TRUE(WIFEXITED(ws) && !WEXITSTATUS(ws));
|
|
|
|
}
|
|
|
|
|
2022-07-20 04:18:33 +00:00
|
|
|
TEST(pledge, mmapProtExec) {
|
2022-07-08 13:29:24 +00:00
|
|
|
if (IsOpenbsd()) return; // b/c testing linux bpf
|
|
|
|
char *p;
|
|
|
|
int ws, pid;
|
|
|
|
ASSERT_NE(-1, (pid = fork()));
|
|
|
|
if (!pid) {
|
2022-07-20 04:18:33 +00:00
|
|
|
ASSERT_SYS(0, 0, pledge("stdio prot_exec", 0));
|
2022-07-08 13:29:24 +00:00
|
|
|
ASSERT_NE(MAP_FAILED, (p = mmap(0, FRAMESIZE, PROT_READ | PROT_WRITE,
|
|
|
|
MAP_ANONYMOUS | MAP_PRIVATE, -1, 0)));
|
|
|
|
ASSERT_SYS(0, 0, mprotect(p, FRAMESIZE, PROT_READ));
|
2022-07-20 04:18:33 +00:00
|
|
|
ASSERT_SYS(0, 0, mprotect(p, FRAMESIZE, PROT_READ | PROT_EXEC));
|
2022-07-08 13:29:24 +00:00
|
|
|
ASSERT_NE(MAP_FAILED, mmap(0, FRAMESIZE, PROT_EXEC | PROT_READ,
|
|
|
|
MAP_ANONYMOUS | MAP_PRIVATE, -1, 0));
|
|
|
|
_Exit(0);
|
|
|
|
}
|
|
|
|
EXPECT_NE(-1, wait(&ws));
|
|
|
|
EXPECT_TRUE(WIFEXITED(ws) && !WEXITSTATUS(ws));
|
|
|
|
}
|
|
|
|
|
2022-06-27 20:01:58 +00:00
|
|
|
TEST(pledge, chmod_ignoresDangerBits) {
|
|
|
|
if (IsOpenbsd()) return; // b/c testing linux bpf
|
|
|
|
int ws, pid;
|
|
|
|
ASSERT_SYS(0, 3, creat("foo", 0644));
|
|
|
|
ASSERT_NE(-1, (pid = fork()));
|
|
|
|
if (!pid) {
|
2022-07-08 13:29:24 +00:00
|
|
|
ASSERT_SYS(0, 0, pledge("stdio rpath wpath", 0));
|
2022-06-27 20:01:58 +00:00
|
|
|
ASSERT_SYS(0, 0, fchmod(3, 00700));
|
|
|
|
ASSERT_SYS(0, 0, chmod("foo", 00700));
|
|
|
|
ASSERT_SYS(0, 0, fchmodat(AT_FDCWD, "foo", 00700, 0));
|
|
|
|
ASSERT_SYS(EPERM, -1, fchmod(3, 07700));
|
|
|
|
ASSERT_SYS(EPERM, -1, chmod("foo", 04700));
|
|
|
|
ASSERT_SYS(EPERM, -1, fchmodat(AT_FDCWD, "foo", 02700, 0));
|
|
|
|
_Exit(0);
|
|
|
|
}
|
|
|
|
EXPECT_NE(-1, wait(&ws));
|
|
|
|
EXPECT_TRUE(WIFEXITED(ws) && !WEXITSTATUS(ws));
|
|
|
|
close(3);
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST(pledge, open_rpath) {
|
|
|
|
if (IsOpenbsd()) return; // b/c testing linux bpf
|
|
|
|
int ws, pid;
|
|
|
|
ASSERT_SYS(0, 0, touch("foo", 0644));
|
|
|
|
ASSERT_NE(-1, (pid = fork()));
|
|
|
|
if (!pid) {
|
|
|
|
ASSERT_SYS(0, 0, pledge("stdio rpath", 0));
|
|
|
|
ASSERT_SYS(0, 3, open("foo", O_RDONLY));
|
2023-08-21 09:28:24 +00:00
|
|
|
ASSERT_SYS(EINVAL, -1, open("foo", O_RDONLY | O_TRUNC));
|
2022-07-25 06:40:49 +00:00
|
|
|
ASSERT_SYS(EPERM, -1, open("foo", O_RDONLY | O_TMPFILE));
|
2022-06-27 20:01:58 +00:00
|
|
|
ASSERT_SYS(EPERM, -1, open("foo", O_RDWR | O_TRUNC | O_CREAT, 0644));
|
|
|
|
ASSERT_SYS(EPERM, -1, open("foo", O_WRONLY | O_TRUNC | O_CREAT, 0644));
|
|
|
|
_Exit(0);
|
|
|
|
}
|
|
|
|
EXPECT_NE(-1, wait(&ws));
|
|
|
|
EXPECT_TRUE(WIFEXITED(ws) && !WEXITSTATUS(ws));
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST(pledge, open_wpath) {
|
|
|
|
if (IsOpenbsd()) return; // b/c testing linux bpf
|
|
|
|
int ws, pid;
|
|
|
|
ASSERT_SYS(0, 0, touch("foo", 0644));
|
|
|
|
ASSERT_NE(-1, (pid = fork()));
|
|
|
|
if (!pid) {
|
|
|
|
ASSERT_SYS(0, 0, pledge("stdio wpath", 0));
|
|
|
|
ASSERT_SYS(EPERM, -1, open(".", O_RDWR | O_TMPFILE, 07644));
|
2022-07-25 06:40:49 +00:00
|
|
|
ASSERT_SYS(0, 3, open("foo", O_WRONLY | O_TRUNC));
|
|
|
|
ASSERT_SYS(0, 4, open("foo", O_RDWR));
|
2022-07-13 10:08:16 +00:00
|
|
|
ASSERT_SYS(EPERM, -1, open("foo", O_WRONLY | O_TRUNC | O_CREAT, 0644));
|
|
|
|
_Exit(0);
|
|
|
|
}
|
|
|
|
EXPECT_NE(-1, wait(&ws));
|
|
|
|
EXPECT_TRUE(WIFEXITED(ws) && !WEXITSTATUS(ws));
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST(pledge, open_cpath) {
|
|
|
|
if (IsOpenbsd()) return; // b/c testing linux bpf
|
|
|
|
int ws, pid;
|
|
|
|
struct stat st;
|
|
|
|
ASSERT_SYS(0, 0, touch("foo", 0644));
|
|
|
|
ASSERT_NE(-1, (pid = fork()));
|
|
|
|
if (!pid) {
|
2023-11-16 08:47:21 +00:00
|
|
|
unsigned omask = umask(022);
|
2022-07-13 10:08:16 +00:00
|
|
|
ASSERT_SYS(0, 0, pledge("stdio cpath", 0));
|
|
|
|
ASSERT_SYS(0, 3, open("foo", O_WRONLY | O_TRUNC | O_CREAT, 0644));
|
2022-07-16 03:47:20 +00:00
|
|
|
ASSERT_SYS(0, 0, fstat(3, &st));
|
|
|
|
ASSERT_EQ(0100644, st.st_mode);
|
2023-11-16 08:47:21 +00:00
|
|
|
umask(omask);
|
2022-07-16 03:47:20 +00:00
|
|
|
// make sure open() can't apply the setuid bit
|
|
|
|
ASSERT_SYS(EPERM, -1, open("bar", O_WRONLY | O_TRUNC | O_CREAT, 04644));
|
2022-06-27 20:01:58 +00:00
|
|
|
_Exit(0);
|
|
|
|
}
|
|
|
|
EXPECT_NE(-1, wait(&ws));
|
|
|
|
EXPECT_TRUE(WIFEXITED(ws) && !WEXITSTATUS(ws));
|
2022-07-08 13:29:24 +00:00
|
|
|
}
|
|
|
|
|
2022-07-18 14:23:15 +00:00
|
|
|
TEST(pledge, execpromises_ok) {
|
|
|
|
if (IsOpenbsd()) return; // b/c testing linux bpf
|
|
|
|
int ws, pid;
|
|
|
|
ASSERT_NE(-1, (pid = fork()));
|
|
|
|
if (!pid) {
|
2022-07-20 04:18:33 +00:00
|
|
|
ASSERT_SYS(0, 0, pledge("stdio exec", "stdio"));
|
2022-07-18 14:23:15 +00:00
|
|
|
execl("life.elf", "life.elf", 0);
|
|
|
|
_Exit(127);
|
|
|
|
}
|
|
|
|
EXPECT_NE(-1, wait(&ws));
|
|
|
|
EXPECT_TRUE(WIFEXITED(ws));
|
|
|
|
EXPECT_EQ(42, WEXITSTATUS(ws));
|
|
|
|
}
|
|
|
|
|
2022-08-07 23:18:33 +00:00
|
|
|
TEST(pledge, execpromises_notok1) {
|
2022-07-18 14:23:15 +00:00
|
|
|
if (IsOpenbsd()) return; // b/c testing linux bpf
|
|
|
|
int ws, pid;
|
|
|
|
ASSERT_NE(-1, (pid = fork()));
|
|
|
|
if (!pid) {
|
2022-07-20 04:18:33 +00:00
|
|
|
ASSERT_SYS(0, 0, pledge("stdio exec", "stdio"));
|
2022-07-18 14:23:15 +00:00
|
|
|
execl("sock.elf", "sock.elf", 0);
|
|
|
|
_Exit(127);
|
|
|
|
}
|
|
|
|
EXPECT_NE(-1, wait(&ws));
|
|
|
|
EXPECT_TRUE(WIFEXITED(ws));
|
|
|
|
EXPECT_EQ(128 + EPERM, WEXITSTATUS(ws));
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST(pledge, execpromises_reducesAtExecOnLinux) {
|
|
|
|
if (IsOpenbsd()) return; // b/c testing linux bpf
|
|
|
|
int ws, pid;
|
|
|
|
ASSERT_NE(-1, (pid = fork()));
|
|
|
|
if (!pid) {
|
2022-07-20 04:18:33 +00:00
|
|
|
ASSERT_SYS(0, 0, pledge("stdio inet tty exec", "stdio tty"));
|
2022-07-18 14:23:15 +00:00
|
|
|
execl("sock.elf", "sock.elf", 0);
|
|
|
|
_Exit(127);
|
|
|
|
}
|
|
|
|
EXPECT_NE(-1, wait(&ws));
|
|
|
|
EXPECT_TRUE(WIFEXITED(ws));
|
|
|
|
EXPECT_EQ(128 + EPERM, WEXITSTATUS(ws));
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST(pledge_openbsd, execpromisesIsNull_letsItDoAnything) {
|
|
|
|
if (!IsOpenbsd()) return;
|
|
|
|
int ws, pid;
|
|
|
|
ASSERT_NE(-1, (pid = fork()));
|
|
|
|
if (!pid) {
|
2022-07-20 04:18:33 +00:00
|
|
|
ASSERT_SYS(0, 0, pledge("stdio exec", 0));
|
2022-07-18 14:23:15 +00:00
|
|
|
execl("sock.elf", "sock.elf", 0);
|
|
|
|
_Exit(127);
|
|
|
|
}
|
|
|
|
EXPECT_NE(-1, wait(&ws));
|
|
|
|
EXPECT_TRUE(WIFEXITED(ws));
|
2023-10-04 14:07:43 +00:00
|
|
|
EXPECT_FALSE(WIFSIGNALED(ws));
|
|
|
|
EXPECT_EQ(0, WTERMSIG(ws));
|
2022-07-18 14:23:15 +00:00
|
|
|
EXPECT_EQ(3, WEXITSTATUS(ws));
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST(pledge_openbsd, execpromisesIsSuperset_letsItDoAnything) {
|
|
|
|
if (!IsOpenbsd()) return;
|
|
|
|
int ws, pid;
|
|
|
|
ASSERT_NE(-1, (pid = fork()));
|
|
|
|
if (!pid) {
|
2022-07-20 04:18:33 +00:00
|
|
|
ASSERT_SYS(0, 0, pledge("stdio rpath exec", "stdio rpath tty inet"));
|
2022-07-18 14:23:15 +00:00
|
|
|
execl("sock.elf", "sock.elf", 0);
|
|
|
|
_Exit(127);
|
|
|
|
}
|
|
|
|
EXPECT_NE(-1, wait(&ws));
|
|
|
|
EXPECT_TRUE(WIFEXITED(ws));
|
|
|
|
EXPECT_EQ(3, WEXITSTATUS(ws));
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST(pledge_linux, execpromisesIsSuperset_notPossible) {
|
|
|
|
if (IsOpenbsd()) return;
|
2022-07-20 04:18:33 +00:00
|
|
|
ASSERT_SYS(EINVAL, -1, pledge("stdio exec", "stdio inet exec"));
|
2022-07-18 14:23:15 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
TEST(pledge_openbsd, execpromises_notok) {
|
|
|
|
int ws, pid;
|
|
|
|
ASSERT_NE(-1, (pid = fork()));
|
|
|
|
if (!pid) {
|
2022-07-20 04:18:33 +00:00
|
|
|
ASSERT_SYS(0, 0, pledge("stdio exec", "stdio"));
|
2022-07-18 14:23:15 +00:00
|
|
|
execl("sock.elf", "sock.elf", 0);
|
|
|
|
_Exit(127);
|
|
|
|
}
|
|
|
|
EXPECT_NE(-1, wait(&ws));
|
2022-08-08 18:41:08 +00:00
|
|
|
if (IsOpenbsd()) {
|
|
|
|
EXPECT_TRUE(WIFSIGNALED(ws));
|
|
|
|
EXPECT_EQ(SIGABRT, WTERMSIG(ws));
|
|
|
|
} else {
|
|
|
|
// linux can't be consistent here since we pledged exec
|
|
|
|
// so we return EPERM instead and sock.elf passes along
|
|
|
|
EXPECT_TRUE(WIFEXITED(ws));
|
|
|
|
EXPECT_EQ(128 + EPERM, WEXITSTATUS(ws));
|
|
|
|
}
|
2022-07-18 14:23:15 +00:00
|
|
|
}
|
2022-07-20 04:18:33 +00:00
|
|
|
|
|
|
|
TEST(pledge_openbsd, bigSyscalls) {
|
|
|
|
if (IsOpenbsd()) return; // testing lunix
|
|
|
|
int ws, pid;
|
|
|
|
ASSERT_NE(-1, (pid = fork()));
|
|
|
|
if (!pid) {
|
|
|
|
ASSERT_SYS(0, 0, pledge("stdio", 0));
|
2022-09-13 18:20:35 +00:00
|
|
|
ASSERT_SYS(ENOSYS, -1, sys_memfd_secret(0));
|
2022-07-20 04:18:33 +00:00
|
|
|
ASSERT_SYS(ENOSYS, -1, sys_bogus());
|
|
|
|
_Exit(0);
|
|
|
|
}
|
|
|
|
EXPECT_NE(-1, wait(&ws));
|
|
|
|
EXPECT_TRUE(WIFEXITED(ws));
|
|
|
|
EXPECT_EQ(0, WEXITSTATUS(ws));
|
|
|
|
}
|
|
|
|
|
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
|
|
|
void *LockWorker(void *arg) {
|
2022-07-20 04:18:33 +00:00
|
|
|
flockfile(stdout);
|
2023-08-14 03:31:27 +00:00
|
|
|
ASSERT_EQ(gettid(), stdout->lock._owner);
|
2022-07-20 04:18:33 +00:00
|
|
|
funlockfile(stdout);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST(pledge, threadWithLocks_canCodeMorph) {
|
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
|
|
|
pthread_t worker;
|
2023-09-02 03:49:13 +00:00
|
|
|
int ws;
|
2022-07-20 04:18:33 +00:00
|
|
|
// not sure how this works on OpenBSD but it works!
|
|
|
|
if (!fork()) {
|
2022-08-10 15:11:05 +00:00
|
|
|
ASSERT_SYS(0, 0, pledge("stdio", 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
|
|
|
ASSERT_EQ(0, pthread_create(&worker, 0, LockWorker, 0));
|
|
|
|
ASSERT_EQ(0, pthread_join(worker, 0));
|
2022-07-20 04:18:33 +00:00
|
|
|
_Exit(0);
|
|
|
|
}
|
|
|
|
EXPECT_NE(-1, wait(&ws));
|
|
|
|
EXPECT_TRUE(WIFEXITED(ws));
|
|
|
|
EXPECT_EQ(0, WEXITSTATUS(ws));
|
|
|
|
}
|
|
|
|
|
2022-07-25 02:40:32 +00:00
|
|
|
TEST(pledge, everything) {
|
2023-09-02 03:49:13 +00:00
|
|
|
int ws;
|
2022-07-25 02:40:32 +00:00
|
|
|
if (!fork()) {
|
2022-08-20 19:32:51 +00:00
|
|
|
// contains 591 bpf instructions [2022-07-24]
|
2022-07-25 02:40:32 +00:00
|
|
|
ASSERT_SYS(0, 0,
|
|
|
|
pledge("stdio rpath wpath cpath dpath "
|
2023-06-04 17:57:11 +00:00
|
|
|
"flock fattr inet unix dns tty "
|
2022-07-25 02:40:32 +00:00
|
|
|
"recvfd sendfd proc exec id "
|
|
|
|
"unveil settime prot_exec "
|
|
|
|
"vminfo tmppath",
|
|
|
|
"stdio rpath wpath cpath dpath "
|
|
|
|
"flock fattr inet unix dns tty "
|
|
|
|
"recvfd sendfd proc exec id "
|
|
|
|
"unveil settime prot_exec "
|
|
|
|
"vminfo tmppath"));
|
|
|
|
_Exit(0);
|
|
|
|
}
|
|
|
|
EXPECT_NE(-1, wait(&ws));
|
|
|
|
EXPECT_TRUE(WIFEXITED(ws));
|
|
|
|
EXPECT_EQ(0, WEXITSTATUS(ws));
|
|
|
|
}
|
|
|
|
|
2022-07-24 09:56:03 +00:00
|
|
|
TEST(pledge, execWithoutRpath) {
|
|
|
|
int ws, pid;
|
|
|
|
ASSERT_SYS(0, 0, touch("foo", 0644));
|
|
|
|
ASSERT_NE(-1, (pid = fork()));
|
|
|
|
if (!pid) {
|
|
|
|
ASSERT_SYS(0, 0, pledge("stdio prot_exec exec", "stdio prot_exec exec"));
|
|
|
|
ASSERT_SYS(EPERM, -1, open("foo", O_RDONLY));
|
|
|
|
_Exit(0);
|
|
|
|
}
|
|
|
|
EXPECT_NE(-1, wait(&ws));
|
|
|
|
if (IsOpenbsd()) {
|
|
|
|
EXPECT_TRUE(WIFSIGNALED(ws));
|
|
|
|
EXPECT_EQ(SIGABRT, WTERMSIG(ws));
|
|
|
|
} else {
|
|
|
|
EXPECT_TRUE(WIFEXITED(ws));
|
|
|
|
EXPECT_EQ(0, WEXITSTATUS(ws));
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2022-07-20 04:18:33 +00:00
|
|
|
BENCH(pledge, bench) {
|
|
|
|
if (!fork()) {
|
|
|
|
ASSERT_SYS(0, 0, pledge("stdio", 0));
|
|
|
|
EZBENCH2("sched_yield", donothing, sched_yield());
|
|
|
|
_Exit(0);
|
|
|
|
}
|
|
|
|
wait(0);
|
|
|
|
}
|