cosmopolitan/test/libc/calls/pledge_test.c

693 lines
22 KiB
C
Raw Normal View History

2020-06-15 14:18:57 +00:00
/*-*- mode:c;indent-tabs-mode:nil;c-basic-offset:2;tab-width:8;coding:utf-8 -*-│
vi: set noet ft=c ts=2 sts=2 sw=2 fenc=utf-8 :vi
2020-06-15 14:18:57 +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"
#include "libc/calls/internal.h"
#include "libc/calls/pledge.internal.h"
#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"
#include "libc/calls/struct/seccomp.internal.h"
#include "libc/calls/struct/sigaction.h"
#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"
#include "libc/dce.h"
#include "libc/errno.h"
#include "libc/macros.internal.h"
#include "libc/mem/mem.h"
Make pledge() and unveil() work amazingly This change reconciles our pledge() implementation with the OpenBSD kernel source code. We now a polyfill that's much closer to OpenBSD's behavior. For example, it was discovered that "stdio" permits threads. There were a bunch of Linux system calls that needed to be added, like sched_yield(). The exec / execnative category division is now dropped. We're instead using OpenBSD's "prot_exec" promise for launching APE binaries and dynamic shared objects. We also now filter clone() flags. The pledge.com command has been greatly improved. It now does unveiling by default when Landlock is available. It's now smart enough to unveil a superset of paths that OpenBSD automatically unveils with pledge(), such as /etc/localtime. pledge.com also now checks if the executable being launched is a dynamic shared object, in which case it unveils libraries. These changes now make it possible to pledge curl on ubuntu 20.04 glibc: pledge.com -p 'stdio rpath prot_exec inet dns tty sendfd recvfd' \ curl -s https://justine.lol/hello.txt Here's what pledging curl on Alpine 3.16 with Musl Libc looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ curl -s https://justine.lol/hello.txt Here's what pledging curl.com w/ ape loader looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ o//examples/curl.com https://justine.lol/hello.txt The most secure sandbox, is curl.com converted to static ELF: o//tool/build/assimilate.com o//examples/curl.com pledge.com -p 'stdio rpath dns inet' \ o//examples/curl.com https://justine.lol/hello.txt A weird corner case needed to be handled when resolving symbolic links during the unveiling process, that's arguably a Landlock bug. It's not surprising since Musl and Glibc are also inconsistent here too.
2022-07-20 04:18:33 +00:00
#include "libc/runtime/internal.h"
#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"
#include "libc/stdio/internal.h"
Make pledge() and unveil() work amazingly This change reconciles our pledge() implementation with the OpenBSD kernel source code. We now a polyfill that's much closer to OpenBSD's behavior. For example, it was discovered that "stdio" permits threads. There were a bunch of Linux system calls that needed to be added, like sched_yield(). The exec / execnative category division is now dropped. We're instead using OpenBSD's "prot_exec" promise for launching APE binaries and dynamic shared objects. We also now filter clone() flags. The pledge.com command has been greatly improved. It now does unveiling by default when Landlock is available. It's now smart enough to unveil a superset of paths that OpenBSD automatically unveils with pledge(), such as /etc/localtime. pledge.com also now checks if the executable being launched is a dynamic shared object, in which case it unveils libraries. These changes now make it possible to pledge curl on ubuntu 20.04 glibc: pledge.com -p 'stdio rpath prot_exec inet dns tty sendfd recvfd' \ curl -s https://justine.lol/hello.txt Here's what pledging curl on Alpine 3.16 with Musl Libc looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ curl -s https://justine.lol/hello.txt Here's what pledging curl.com w/ ape loader looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ o//examples/curl.com https://justine.lol/hello.txt The most secure sandbox, is curl.com converted to static ELF: o//tool/build/assimilate.com o//examples/curl.com pledge.com -p 'stdio rpath dns inet' \ o//examples/curl.com https://justine.lol/hello.txt A weird corner case needed to be handled when resolving symbolic links during the unveiling process, that's arguably a Landlock bug. It's not surprising since Musl and Glibc are also inconsistent here too.
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"
#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"
#include "libc/sysv/consts/sig.h"
#include "libc/sysv/consts/so.h"
2022-06-27 20:01:58 +00:00
#include "libc/sysv/consts/sock.h"
#include "libc/sysv/consts/sol.h"
Make pledge() and unveil() work amazingly This change reconciles our pledge() implementation with the OpenBSD kernel source code. We now a polyfill that's much closer to OpenBSD's behavior. For example, it was discovered that "stdio" permits threads. There were a bunch of Linux system calls that needed to be added, like sched_yield(). The exec / execnative category division is now dropped. We're instead using OpenBSD's "prot_exec" promise for launching APE binaries and dynamic shared objects. We also now filter clone() flags. The pledge.com command has been greatly improved. It now does unveiling by default when Landlock is available. It's now smart enough to unveil a superset of paths that OpenBSD automatically unveils with pledge(), such as /etc/localtime. pledge.com also now checks if the executable being launched is a dynamic shared object, in which case it unveils libraries. These changes now make it possible to pledge curl on ubuntu 20.04 glibc: pledge.com -p 'stdio rpath prot_exec inet dns tty sendfd recvfd' \ curl -s https://justine.lol/hello.txt Here's what pledging curl on Alpine 3.16 with Musl Libc looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ curl -s https://justine.lol/hello.txt Here's what pledging curl.com w/ ape loader looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ o//examples/curl.com https://justine.lol/hello.txt The most secure sandbox, is curl.com converted to static ELF: o//tool/build/assimilate.com o//examples/curl.com pledge.com -p 'stdio rpath dns inet' \ o//examples/curl.com https://justine.lol/hello.txt A weird corner case needed to be handled when resolving symbolic links during the unveiling process, that's arguably a Landlock bug. It's not surprising since Musl and Glibc are also inconsistent here too.
2022-07-20 04:18:33 +00:00
#include "libc/testlib/ezbench.h"
#include "libc/testlib/subprocess.h"
#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"
#include "libc/thread/thread.h"
Make pledge() and unveil() work amazingly This change reconciles our pledge() implementation with the OpenBSD kernel source code. We now a polyfill that's much closer to OpenBSD's behavior. For example, it was discovered that "stdio" permits threads. There were a bunch of Linux system calls that needed to be added, like sched_yield(). The exec / execnative category division is now dropped. We're instead using OpenBSD's "prot_exec" promise for launching APE binaries and dynamic shared objects. We also now filter clone() flags. The pledge.com command has been greatly improved. It now does unveiling by default when Landlock is available. It's now smart enough to unveil a superset of paths that OpenBSD automatically unveils with pledge(), such as /etc/localtime. pledge.com also now checks if the executable being launched is a dynamic shared object, in which case it unveils libraries. These changes now make it possible to pledge curl on ubuntu 20.04 glibc: pledge.com -p 'stdio rpath prot_exec inet dns tty sendfd recvfd' \ curl -s https://justine.lol/hello.txt Here's what pledging curl on Alpine 3.16 with Musl Libc looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ curl -s https://justine.lol/hello.txt Here's what pledging curl.com w/ ape loader looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ o//examples/curl.com https://justine.lol/hello.txt The most secure sandbox, is curl.com converted to static ELF: o//tool/build/assimilate.com o//examples/curl.com pledge.com -p 'stdio rpath dns inet' \ o//examples/curl.com https://justine.lol/hello.txt A weird corner case needed to be handled when resolving symbolic links during the unveiling process, that's arguably a Landlock bug. It's not surprising since Musl and Glibc are also inconsistent here too.
2022-07-20 04:18:33 +00:00
#include "libc/time/time.h"
#include "libc/x/x.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
void SetUpOnce(void) {
testlib_enable_tmp_setup_teardown();
}
2022-06-27 20:01:58 +00:00
void OnSig(int sig) {
// do nothing
}
int sys_memfd_secret(unsigned int); // our ENOSYS threshold
Make pledge() and unveil() work amazingly This change reconciles our pledge() implementation with the OpenBSD kernel source code. We now a polyfill that's much closer to OpenBSD's behavior. For example, it was discovered that "stdio" permits threads. There were a bunch of Linux system calls that needed to be added, like sched_yield(). The exec / execnative category division is now dropped. We're instead using OpenBSD's "prot_exec" promise for launching APE binaries and dynamic shared objects. We also now filter clone() flags. The pledge.com command has been greatly improved. It now does unveiling by default when Landlock is available. It's now smart enough to unveil a superset of paths that OpenBSD automatically unveils with pledge(), such as /etc/localtime. pledge.com also now checks if the executable being launched is a dynamic shared object, in which case it unveils libraries. These changes now make it possible to pledge curl on ubuntu 20.04 glibc: pledge.com -p 'stdio rpath prot_exec inet dns tty sendfd recvfd' \ curl -s https://justine.lol/hello.txt Here's what pledging curl on Alpine 3.16 with Musl Libc looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ curl -s https://justine.lol/hello.txt Here's what pledging curl.com w/ ape loader looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ o//examples/curl.com https://justine.lol/hello.txt The most secure sandbox, is curl.com converted to static ELF: o//tool/build/assimilate.com o//examples/curl.com pledge.com -p 'stdio rpath dns inet' \ o//examples/curl.com https://justine.lol/hello.txt A weird corner case needed to be handled when resolving symbolic links during the unveiling process, that's arguably a Landlock bug. It's not surprising since Musl and Glibc are also inconsistent here too.
2022-07-20 04:18:33 +00:00
void SetUp(void) {
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");
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);
__pledge_mode = PLEDGE_PENALTY_RETURN_EPERM;
}
2020-06-15 14:18:57 +00:00
TEST(pledge, default_allowsExit) {
2022-07-14 11:32:33 +00:00
int *job;
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
if (!pid) {
ASSERT_SYS(0, 0, pledge("", 0));
2022-07-14 11:32:33 +00:00
job[0] = job[0] + job[1]; // do work
_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
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));
}
TEST(pledge, execpromises_notok) {
if (IsOpenbsd()) return; // b/c testing linux bpf
int ws, pid;
ASSERT_NE(-1, (pid = fork()));
if (!pid) {
__pledge_mode = PLEDGE_PENALTY_RETURN_EPERM;
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
}
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
}
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) {
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;
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));
EXITS(0);
}
TEST(pledge, stdio_forbidsOpeningPasswd1) {
if (!IsLinux()) return;
int ws, pid;
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(WIFEXITED(ws));
EXPECT_EQ(0, WEXITSTATUS(ws));
}
TEST(pledge, stdio_forbidsOpeningPasswd2) {
int ws, pid;
__pledge_mode = PLEDGE_PENALTY_KILL_PROCESS;
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));
}
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));
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));
_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) {
ASSERT_SYS(0, 0, pledge("stdio", "stdio"));
ASSERT_SYS(EPERM, -1, pledge("stdio unix", "stdio unix"));
_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);
}
EXPECT_NE(-1, wait(&ws));
EXPECT_TRUE(WIFEXITED(ws));
EXPECT_EQ(0, WEXITSTATUS(ws));
}
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) {
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));
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;
// 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;
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));
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));
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));
}
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
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));
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));
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));
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)}};
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)));
uint32_t az = sizeof(sin);
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));
}
TEST(pledge, anet_forbidsUdpSocketsAndConnect) {
if (IsOpenbsd()) return; // b/c testing linux bpf
int ws, pid;
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));
}
Make pledge() and unveil() work amazingly This change reconciles our pledge() implementation with the OpenBSD kernel source code. We now a polyfill that's much closer to OpenBSD's behavior. For example, it was discovered that "stdio" permits threads. There were a bunch of Linux system calls that needed to be added, like sched_yield(). The exec / execnative category division is now dropped. We're instead using OpenBSD's "prot_exec" promise for launching APE binaries and dynamic shared objects. We also now filter clone() flags. The pledge.com command has been greatly improved. It now does unveiling by default when Landlock is available. It's now smart enough to unveil a superset of paths that OpenBSD automatically unveils with pledge(), such as /etc/localtime. pledge.com also now checks if the executable being launched is a dynamic shared object, in which case it unveils libraries. These changes now make it possible to pledge curl on ubuntu 20.04 glibc: pledge.com -p 'stdio rpath prot_exec inet dns tty sendfd recvfd' \ curl -s https://justine.lol/hello.txt Here's what pledging curl on Alpine 3.16 with Musl Libc looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ curl -s https://justine.lol/hello.txt Here's what pledging curl.com w/ ape loader looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ o//examples/curl.com https://justine.lol/hello.txt The most secure sandbox, is curl.com converted to static ELF: o//tool/build/assimilate.com o//examples/curl.com pledge.com -p 'stdio rpath dns inet' \ o//examples/curl.com https://justine.lol/hello.txt A weird corner case needed to be handled when resolving symbolic links during the unveiling process, that's arguably a Landlock bug. It's not surprising since Musl and Glibc are also inconsistent here too.
2022-07-20 04:18:33 +00:00
TEST(pledge, mmapProtExec) {
if (IsOpenbsd()) return; // b/c testing linux bpf
char *p;
int ws, pid;
ASSERT_NE(-1, (pid = fork()));
if (!pid) {
Make pledge() and unveil() work amazingly This change reconciles our pledge() implementation with the OpenBSD kernel source code. We now a polyfill that's much closer to OpenBSD's behavior. For example, it was discovered that "stdio" permits threads. There were a bunch of Linux system calls that needed to be added, like sched_yield(). The exec / execnative category division is now dropped. We're instead using OpenBSD's "prot_exec" promise for launching APE binaries and dynamic shared objects. We also now filter clone() flags. The pledge.com command has been greatly improved. It now does unveiling by default when Landlock is available. It's now smart enough to unveil a superset of paths that OpenBSD automatically unveils with pledge(), such as /etc/localtime. pledge.com also now checks if the executable being launched is a dynamic shared object, in which case it unveils libraries. These changes now make it possible to pledge curl on ubuntu 20.04 glibc: pledge.com -p 'stdio rpath prot_exec inet dns tty sendfd recvfd' \ curl -s https://justine.lol/hello.txt Here's what pledging curl on Alpine 3.16 with Musl Libc looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ curl -s https://justine.lol/hello.txt Here's what pledging curl.com w/ ape loader looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ o//examples/curl.com https://justine.lol/hello.txt The most secure sandbox, is curl.com converted to static ELF: o//tool/build/assimilate.com o//examples/curl.com pledge.com -p 'stdio rpath dns inet' \ o//examples/curl.com https://justine.lol/hello.txt A weird corner case needed to be handled when resolving symbolic links during the unveiling process, that's arguably a Landlock bug. It's not surprising since Musl and Glibc are also inconsistent here too.
2022-07-20 04:18:33 +00:00
ASSERT_SYS(0, 0, pledge("stdio prot_exec", 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));
Make pledge() and unveil() work amazingly This change reconciles our pledge() implementation with the OpenBSD kernel source code. We now a polyfill that's much closer to OpenBSD's behavior. For example, it was discovered that "stdio" permits threads. There were a bunch of Linux system calls that needed to be added, like sched_yield(). The exec / execnative category division is now dropped. We're instead using OpenBSD's "prot_exec" promise for launching APE binaries and dynamic shared objects. We also now filter clone() flags. The pledge.com command has been greatly improved. It now does unveiling by default when Landlock is available. It's now smart enough to unveil a superset of paths that OpenBSD automatically unveils with pledge(), such as /etc/localtime. pledge.com also now checks if the executable being launched is a dynamic shared object, in which case it unveils libraries. These changes now make it possible to pledge curl on ubuntu 20.04 glibc: pledge.com -p 'stdio rpath prot_exec inet dns tty sendfd recvfd' \ curl -s https://justine.lol/hello.txt Here's what pledging curl on Alpine 3.16 with Musl Libc looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ curl -s https://justine.lol/hello.txt Here's what pledging curl.com w/ ape loader looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ o//examples/curl.com https://justine.lol/hello.txt The most secure sandbox, is curl.com converted to static ELF: o//tool/build/assimilate.com o//examples/curl.com pledge.com -p 'stdio rpath dns inet' \ o//examples/curl.com https://justine.lol/hello.txt A weird corner case needed to be handled when resolving symbolic links during the unveiling process, that's arguably a Landlock bug. It's not surprising since Musl and Glibc are also inconsistent here too.
2022-07-20 04:18:33 +00:00
ASSERT_SYS(0, 0, mprotect(p, FRAMESIZE, PROT_READ | PROT_EXEC));
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) {
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));
ASSERT_SYS(EINVAL, -1, open("foo", O_RDONLY | O_TRUNC));
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));
ASSERT_SYS(0, 3, open("foo", O_WRONLY | O_TRUNC));
ASSERT_SYS(0, 4, open("foo", O_RDWR));
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) {
unsigned omask = umask(022);
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);
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));
}
TEST(pledge, execpromises_ok) {
if (IsOpenbsd()) return; // b/c testing linux bpf
int ws, pid;
ASSERT_NE(-1, (pid = fork()));
if (!pid) {
Make pledge() and unveil() work amazingly This change reconciles our pledge() implementation with the OpenBSD kernel source code. We now a polyfill that's much closer to OpenBSD's behavior. For example, it was discovered that "stdio" permits threads. There were a bunch of Linux system calls that needed to be added, like sched_yield(). The exec / execnative category division is now dropped. We're instead using OpenBSD's "prot_exec" promise for launching APE binaries and dynamic shared objects. We also now filter clone() flags. The pledge.com command has been greatly improved. It now does unveiling by default when Landlock is available. It's now smart enough to unveil a superset of paths that OpenBSD automatically unveils with pledge(), such as /etc/localtime. pledge.com also now checks if the executable being launched is a dynamic shared object, in which case it unveils libraries. These changes now make it possible to pledge curl on ubuntu 20.04 glibc: pledge.com -p 'stdio rpath prot_exec inet dns tty sendfd recvfd' \ curl -s https://justine.lol/hello.txt Here's what pledging curl on Alpine 3.16 with Musl Libc looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ curl -s https://justine.lol/hello.txt Here's what pledging curl.com w/ ape loader looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ o//examples/curl.com https://justine.lol/hello.txt The most secure sandbox, is curl.com converted to static ELF: o//tool/build/assimilate.com o//examples/curl.com pledge.com -p 'stdio rpath dns inet' \ o//examples/curl.com https://justine.lol/hello.txt A weird corner case needed to be handled when resolving symbolic links during the unveiling process, that's arguably a Landlock bug. It's not surprising since Musl and Glibc are also inconsistent here too.
2022-07-20 04:18:33 +00:00
ASSERT_SYS(0, 0, pledge("stdio exec", "stdio"));
execl("life.elf", "life.elf", 0);
_Exit(127);
}
EXPECT_NE(-1, wait(&ws));
EXPECT_TRUE(WIFEXITED(ws));
EXPECT_EQ(42, WEXITSTATUS(ws));
}
TEST(pledge, execpromises_notok1) {
if (IsOpenbsd()) return; // b/c testing linux bpf
int ws, pid;
ASSERT_NE(-1, (pid = fork()));
if (!pid) {
Make pledge() and unveil() work amazingly This change reconciles our pledge() implementation with the OpenBSD kernel source code. We now a polyfill that's much closer to OpenBSD's behavior. For example, it was discovered that "stdio" permits threads. There were a bunch of Linux system calls that needed to be added, like sched_yield(). The exec / execnative category division is now dropped. We're instead using OpenBSD's "prot_exec" promise for launching APE binaries and dynamic shared objects. We also now filter clone() flags. The pledge.com command has been greatly improved. It now does unveiling by default when Landlock is available. It's now smart enough to unveil a superset of paths that OpenBSD automatically unveils with pledge(), such as /etc/localtime. pledge.com also now checks if the executable being launched is a dynamic shared object, in which case it unveils libraries. These changes now make it possible to pledge curl on ubuntu 20.04 glibc: pledge.com -p 'stdio rpath prot_exec inet dns tty sendfd recvfd' \ curl -s https://justine.lol/hello.txt Here's what pledging curl on Alpine 3.16 with Musl Libc looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ curl -s https://justine.lol/hello.txt Here's what pledging curl.com w/ ape loader looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ o//examples/curl.com https://justine.lol/hello.txt The most secure sandbox, is curl.com converted to static ELF: o//tool/build/assimilate.com o//examples/curl.com pledge.com -p 'stdio rpath dns inet' \ o//examples/curl.com https://justine.lol/hello.txt A weird corner case needed to be handled when resolving symbolic links during the unveiling process, that's arguably a Landlock bug. It's not surprising since Musl and Glibc are also inconsistent here too.
2022-07-20 04:18:33 +00:00
ASSERT_SYS(0, 0, pledge("stdio exec", "stdio"));
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) {
Make pledge() and unveil() work amazingly This change reconciles our pledge() implementation with the OpenBSD kernel source code. We now a polyfill that's much closer to OpenBSD's behavior. For example, it was discovered that "stdio" permits threads. There were a bunch of Linux system calls that needed to be added, like sched_yield(). The exec / execnative category division is now dropped. We're instead using OpenBSD's "prot_exec" promise for launching APE binaries and dynamic shared objects. We also now filter clone() flags. The pledge.com command has been greatly improved. It now does unveiling by default when Landlock is available. It's now smart enough to unveil a superset of paths that OpenBSD automatically unveils with pledge(), such as /etc/localtime. pledge.com also now checks if the executable being launched is a dynamic shared object, in which case it unveils libraries. These changes now make it possible to pledge curl on ubuntu 20.04 glibc: pledge.com -p 'stdio rpath prot_exec inet dns tty sendfd recvfd' \ curl -s https://justine.lol/hello.txt Here's what pledging curl on Alpine 3.16 with Musl Libc looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ curl -s https://justine.lol/hello.txt Here's what pledging curl.com w/ ape loader looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ o//examples/curl.com https://justine.lol/hello.txt The most secure sandbox, is curl.com converted to static ELF: o//tool/build/assimilate.com o//examples/curl.com pledge.com -p 'stdio rpath dns inet' \ o//examples/curl.com https://justine.lol/hello.txt A weird corner case needed to be handled when resolving symbolic links during the unveiling process, that's arguably a Landlock bug. It's not surprising since Musl and Glibc are also inconsistent here too.
2022-07-20 04:18:33 +00:00
ASSERT_SYS(0, 0, pledge("stdio inet tty exec", "stdio tty"));
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) {
Make pledge() and unveil() work amazingly This change reconciles our pledge() implementation with the OpenBSD kernel source code. We now a polyfill that's much closer to OpenBSD's behavior. For example, it was discovered that "stdio" permits threads. There were a bunch of Linux system calls that needed to be added, like sched_yield(). The exec / execnative category division is now dropped. We're instead using OpenBSD's "prot_exec" promise for launching APE binaries and dynamic shared objects. We also now filter clone() flags. The pledge.com command has been greatly improved. It now does unveiling by default when Landlock is available. It's now smart enough to unveil a superset of paths that OpenBSD automatically unveils with pledge(), such as /etc/localtime. pledge.com also now checks if the executable being launched is a dynamic shared object, in which case it unveils libraries. These changes now make it possible to pledge curl on ubuntu 20.04 glibc: pledge.com -p 'stdio rpath prot_exec inet dns tty sendfd recvfd' \ curl -s https://justine.lol/hello.txt Here's what pledging curl on Alpine 3.16 with Musl Libc looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ curl -s https://justine.lol/hello.txt Here's what pledging curl.com w/ ape loader looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ o//examples/curl.com https://justine.lol/hello.txt The most secure sandbox, is curl.com converted to static ELF: o//tool/build/assimilate.com o//examples/curl.com pledge.com -p 'stdio rpath dns inet' \ o//examples/curl.com https://justine.lol/hello.txt A weird corner case needed to be handled when resolving symbolic links during the unveiling process, that's arguably a Landlock bug. It's not surprising since Musl and Glibc are also inconsistent here too.
2022-07-20 04:18:33 +00:00
ASSERT_SYS(0, 0, pledge("stdio exec", 0));
execl("sock.elf", "sock.elf", 0);
_Exit(127);
}
EXPECT_NE(-1, wait(&ws));
EXPECT_TRUE(WIFEXITED(ws));
EXPECT_FALSE(WIFSIGNALED(ws));
EXPECT_EQ(0, WTERMSIG(ws));
EXPECT_EQ(3, WEXITSTATUS(ws));
}
TEST(pledge_openbsd, execpromisesIsSuperset_letsItDoAnything) {
if (!IsOpenbsd()) return;
int ws, pid;
ASSERT_NE(-1, (pid = fork()));
if (!pid) {
Make pledge() and unveil() work amazingly This change reconciles our pledge() implementation with the OpenBSD kernel source code. We now a polyfill that's much closer to OpenBSD's behavior. For example, it was discovered that "stdio" permits threads. There were a bunch of Linux system calls that needed to be added, like sched_yield(). The exec / execnative category division is now dropped. We're instead using OpenBSD's "prot_exec" promise for launching APE binaries and dynamic shared objects. We also now filter clone() flags. The pledge.com command has been greatly improved. It now does unveiling by default when Landlock is available. It's now smart enough to unveil a superset of paths that OpenBSD automatically unveils with pledge(), such as /etc/localtime. pledge.com also now checks if the executable being launched is a dynamic shared object, in which case it unveils libraries. These changes now make it possible to pledge curl on ubuntu 20.04 glibc: pledge.com -p 'stdio rpath prot_exec inet dns tty sendfd recvfd' \ curl -s https://justine.lol/hello.txt Here's what pledging curl on Alpine 3.16 with Musl Libc looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ curl -s https://justine.lol/hello.txt Here's what pledging curl.com w/ ape loader looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ o//examples/curl.com https://justine.lol/hello.txt The most secure sandbox, is curl.com converted to static ELF: o//tool/build/assimilate.com o//examples/curl.com pledge.com -p 'stdio rpath dns inet' \ o//examples/curl.com https://justine.lol/hello.txt A weird corner case needed to be handled when resolving symbolic links during the unveiling process, that's arguably a Landlock bug. It's not surprising since Musl and Glibc are also inconsistent here too.
2022-07-20 04:18:33 +00:00
ASSERT_SYS(0, 0, pledge("stdio rpath exec", "stdio rpath tty inet"));
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;
Make pledge() and unveil() work amazingly This change reconciles our pledge() implementation with the OpenBSD kernel source code. We now a polyfill that's much closer to OpenBSD's behavior. For example, it was discovered that "stdio" permits threads. There were a bunch of Linux system calls that needed to be added, like sched_yield(). The exec / execnative category division is now dropped. We're instead using OpenBSD's "prot_exec" promise for launching APE binaries and dynamic shared objects. We also now filter clone() flags. The pledge.com command has been greatly improved. It now does unveiling by default when Landlock is available. It's now smart enough to unveil a superset of paths that OpenBSD automatically unveils with pledge(), such as /etc/localtime. pledge.com also now checks if the executable being launched is a dynamic shared object, in which case it unveils libraries. These changes now make it possible to pledge curl on ubuntu 20.04 glibc: pledge.com -p 'stdio rpath prot_exec inet dns tty sendfd recvfd' \ curl -s https://justine.lol/hello.txt Here's what pledging curl on Alpine 3.16 with Musl Libc looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ curl -s https://justine.lol/hello.txt Here's what pledging curl.com w/ ape loader looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ o//examples/curl.com https://justine.lol/hello.txt The most secure sandbox, is curl.com converted to static ELF: o//tool/build/assimilate.com o//examples/curl.com pledge.com -p 'stdio rpath dns inet' \ o//examples/curl.com https://justine.lol/hello.txt A weird corner case needed to be handled when resolving symbolic links during the unveiling process, that's arguably a Landlock bug. It's not surprising since Musl and Glibc are also inconsistent here too.
2022-07-20 04:18:33 +00:00
ASSERT_SYS(EINVAL, -1, pledge("stdio exec", "stdio inet exec"));
}
TEST(pledge_openbsd, execpromises_notok) {
int ws, pid;
ASSERT_NE(-1, (pid = fork()));
if (!pid) {
Make pledge() and unveil() work amazingly This change reconciles our pledge() implementation with the OpenBSD kernel source code. We now a polyfill that's much closer to OpenBSD's behavior. For example, it was discovered that "stdio" permits threads. There were a bunch of Linux system calls that needed to be added, like sched_yield(). The exec / execnative category division is now dropped. We're instead using OpenBSD's "prot_exec" promise for launching APE binaries and dynamic shared objects. We also now filter clone() flags. The pledge.com command has been greatly improved. It now does unveiling by default when Landlock is available. It's now smart enough to unveil a superset of paths that OpenBSD automatically unveils with pledge(), such as /etc/localtime. pledge.com also now checks if the executable being launched is a dynamic shared object, in which case it unveils libraries. These changes now make it possible to pledge curl on ubuntu 20.04 glibc: pledge.com -p 'stdio rpath prot_exec inet dns tty sendfd recvfd' \ curl -s https://justine.lol/hello.txt Here's what pledging curl on Alpine 3.16 with Musl Libc looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ curl -s https://justine.lol/hello.txt Here's what pledging curl.com w/ ape loader looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ o//examples/curl.com https://justine.lol/hello.txt The most secure sandbox, is curl.com converted to static ELF: o//tool/build/assimilate.com o//examples/curl.com pledge.com -p 'stdio rpath dns inet' \ o//examples/curl.com https://justine.lol/hello.txt A weird corner case needed to be handled when resolving symbolic links during the unveiling process, that's arguably a Landlock bug. It's not surprising since Musl and Glibc are also inconsistent here too.
2022-07-20 04:18:33 +00:00
ASSERT_SYS(0, 0, pledge("stdio exec", "stdio"));
execl("sock.elf", "sock.elf", 0);
_Exit(127);
}
EXPECT_NE(-1, wait(&ws));
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));
}
}
Make pledge() and unveil() work amazingly This change reconciles our pledge() implementation with the OpenBSD kernel source code. We now a polyfill that's much closer to OpenBSD's behavior. For example, it was discovered that "stdio" permits threads. There were a bunch of Linux system calls that needed to be added, like sched_yield(). The exec / execnative category division is now dropped. We're instead using OpenBSD's "prot_exec" promise for launching APE binaries and dynamic shared objects. We also now filter clone() flags. The pledge.com command has been greatly improved. It now does unveiling by default when Landlock is available. It's now smart enough to unveil a superset of paths that OpenBSD automatically unveils with pledge(), such as /etc/localtime. pledge.com also now checks if the executable being launched is a dynamic shared object, in which case it unveils libraries. These changes now make it possible to pledge curl on ubuntu 20.04 glibc: pledge.com -p 'stdio rpath prot_exec inet dns tty sendfd recvfd' \ curl -s https://justine.lol/hello.txt Here's what pledging curl on Alpine 3.16 with Musl Libc looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ curl -s https://justine.lol/hello.txt Here's what pledging curl.com w/ ape loader looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ o//examples/curl.com https://justine.lol/hello.txt The most secure sandbox, is curl.com converted to static ELF: o//tool/build/assimilate.com o//examples/curl.com pledge.com -p 'stdio rpath dns inet' \ o//examples/curl.com https://justine.lol/hello.txt A weird corner case needed to be handled when resolving symbolic links during the unveiling process, that's arguably a Landlock bug. It's not surprising since Musl and Glibc are also inconsistent here too.
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));
ASSERT_SYS(ENOSYS, -1, sys_memfd_secret(0));
Make pledge() and unveil() work amazingly This change reconciles our pledge() implementation with the OpenBSD kernel source code. We now a polyfill that's much closer to OpenBSD's behavior. For example, it was discovered that "stdio" permits threads. There were a bunch of Linux system calls that needed to be added, like sched_yield(). The exec / execnative category division is now dropped. We're instead using OpenBSD's "prot_exec" promise for launching APE binaries and dynamic shared objects. We also now filter clone() flags. The pledge.com command has been greatly improved. It now does unveiling by default when Landlock is available. It's now smart enough to unveil a superset of paths that OpenBSD automatically unveils with pledge(), such as /etc/localtime. pledge.com also now checks if the executable being launched is a dynamic shared object, in which case it unveils libraries. These changes now make it possible to pledge curl on ubuntu 20.04 glibc: pledge.com -p 'stdio rpath prot_exec inet dns tty sendfd recvfd' \ curl -s https://justine.lol/hello.txt Here's what pledging curl on Alpine 3.16 with Musl Libc looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ curl -s https://justine.lol/hello.txt Here's what pledging curl.com w/ ape loader looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ o//examples/curl.com https://justine.lol/hello.txt The most secure sandbox, is curl.com converted to static ELF: o//tool/build/assimilate.com o//examples/curl.com pledge.com -p 'stdio rpath dns inet' \ o//examples/curl.com https://justine.lol/hello.txt A weird corner case needed to be handled when resolving symbolic links during the unveiling process, that's arguably a Landlock bug. It's not surprising since Musl and Glibc are also inconsistent here too.
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) {
Make pledge() and unveil() work amazingly This change reconciles our pledge() implementation with the OpenBSD kernel source code. We now a polyfill that's much closer to OpenBSD's behavior. For example, it was discovered that "stdio" permits threads. There were a bunch of Linux system calls that needed to be added, like sched_yield(). The exec / execnative category division is now dropped. We're instead using OpenBSD's "prot_exec" promise for launching APE binaries and dynamic shared objects. We also now filter clone() flags. The pledge.com command has been greatly improved. It now does unveiling by default when Landlock is available. It's now smart enough to unveil a superset of paths that OpenBSD automatically unveils with pledge(), such as /etc/localtime. pledge.com also now checks if the executable being launched is a dynamic shared object, in which case it unveils libraries. These changes now make it possible to pledge curl on ubuntu 20.04 glibc: pledge.com -p 'stdio rpath prot_exec inet dns tty sendfd recvfd' \ curl -s https://justine.lol/hello.txt Here's what pledging curl on Alpine 3.16 with Musl Libc looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ curl -s https://justine.lol/hello.txt Here's what pledging curl.com w/ ape loader looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ o//examples/curl.com https://justine.lol/hello.txt The most secure sandbox, is curl.com converted to static ELF: o//tool/build/assimilate.com o//examples/curl.com pledge.com -p 'stdio rpath dns inet' \ o//examples/curl.com https://justine.lol/hello.txt A weird corner case needed to be handled when resolving symbolic links during the unveiling process, that's arguably a Landlock bug. It's not surprising since Musl and Glibc are also inconsistent here too.
2022-07-20 04:18:33 +00:00
flockfile(stdout);
ASSERT_EQ(gettid(), stdout->lock._owner);
Make pledge() and unveil() work amazingly This change reconciles our pledge() implementation with the OpenBSD kernel source code. We now a polyfill that's much closer to OpenBSD's behavior. For example, it was discovered that "stdio" permits threads. There were a bunch of Linux system calls that needed to be added, like sched_yield(). The exec / execnative category division is now dropped. We're instead using OpenBSD's "prot_exec" promise for launching APE binaries and dynamic shared objects. We also now filter clone() flags. The pledge.com command has been greatly improved. It now does unveiling by default when Landlock is available. It's now smart enough to unveil a superset of paths that OpenBSD automatically unveils with pledge(), such as /etc/localtime. pledge.com also now checks if the executable being launched is a dynamic shared object, in which case it unveils libraries. These changes now make it possible to pledge curl on ubuntu 20.04 glibc: pledge.com -p 'stdio rpath prot_exec inet dns tty sendfd recvfd' \ curl -s https://justine.lol/hello.txt Here's what pledging curl on Alpine 3.16 with Musl Libc looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ curl -s https://justine.lol/hello.txt Here's what pledging curl.com w/ ape loader looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ o//examples/curl.com https://justine.lol/hello.txt The most secure sandbox, is curl.com converted to static ELF: o//tool/build/assimilate.com o//examples/curl.com pledge.com -p 'stdio rpath dns inet' \ o//examples/curl.com https://justine.lol/hello.txt A weird corner case needed to be handled when resolving symbolic links during the unveiling process, that's arguably a Landlock bug. It's not surprising since Musl and Glibc are also inconsistent here too.
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;
int ws;
Make pledge() and unveil() work amazingly This change reconciles our pledge() implementation with the OpenBSD kernel source code. We now a polyfill that's much closer to OpenBSD's behavior. For example, it was discovered that "stdio" permits threads. There were a bunch of Linux system calls that needed to be added, like sched_yield(). The exec / execnative category division is now dropped. We're instead using OpenBSD's "prot_exec" promise for launching APE binaries and dynamic shared objects. We also now filter clone() flags. The pledge.com command has been greatly improved. It now does unveiling by default when Landlock is available. It's now smart enough to unveil a superset of paths that OpenBSD automatically unveils with pledge(), such as /etc/localtime. pledge.com also now checks if the executable being launched is a dynamic shared object, in which case it unveils libraries. These changes now make it possible to pledge curl on ubuntu 20.04 glibc: pledge.com -p 'stdio rpath prot_exec inet dns tty sendfd recvfd' \ curl -s https://justine.lol/hello.txt Here's what pledging curl on Alpine 3.16 with Musl Libc looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ curl -s https://justine.lol/hello.txt Here's what pledging curl.com w/ ape loader looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ o//examples/curl.com https://justine.lol/hello.txt The most secure sandbox, is curl.com converted to static ELF: o//tool/build/assimilate.com o//examples/curl.com pledge.com -p 'stdio rpath dns inet' \ o//examples/curl.com https://justine.lol/hello.txt A weird corner case needed to be handled when resolving symbolic links during the unveiling process, that's arguably a Landlock bug. It's not surprising since Musl and Glibc are also inconsistent here too.
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));
Make pledge() and unveil() work amazingly This change reconciles our pledge() implementation with the OpenBSD kernel source code. We now a polyfill that's much closer to OpenBSD's behavior. For example, it was discovered that "stdio" permits threads. There were a bunch of Linux system calls that needed to be added, like sched_yield(). The exec / execnative category division is now dropped. We're instead using OpenBSD's "prot_exec" promise for launching APE binaries and dynamic shared objects. We also now filter clone() flags. The pledge.com command has been greatly improved. It now does unveiling by default when Landlock is available. It's now smart enough to unveil a superset of paths that OpenBSD automatically unveils with pledge(), such as /etc/localtime. pledge.com also now checks if the executable being launched is a dynamic shared object, in which case it unveils libraries. These changes now make it possible to pledge curl on ubuntu 20.04 glibc: pledge.com -p 'stdio rpath prot_exec inet dns tty sendfd recvfd' \ curl -s https://justine.lol/hello.txt Here's what pledging curl on Alpine 3.16 with Musl Libc looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ curl -s https://justine.lol/hello.txt Here's what pledging curl.com w/ ape loader looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ o//examples/curl.com https://justine.lol/hello.txt The most secure sandbox, is curl.com converted to static ELF: o//tool/build/assimilate.com o//examples/curl.com pledge.com -p 'stdio rpath dns inet' \ o//examples/curl.com https://justine.lol/hello.txt A weird corner case needed to be handled when resolving symbolic links during the unveiling process, that's arguably a Landlock bug. It's not surprising since Musl and Glibc are also inconsistent here too.
2022-07-20 04:18:33 +00:00
_Exit(0);
}
EXPECT_NE(-1, wait(&ws));
EXPECT_TRUE(WIFEXITED(ws));
EXPECT_EQ(0, WEXITSTATUS(ws));
}
TEST(pledge, everything) {
int ws;
if (!fork()) {
2022-08-20 19:32:51 +00:00
// contains 591 bpf instructions [2022-07-24]
ASSERT_SYS(0, 0,
pledge("stdio rpath wpath cpath dpath "
2023-06-04 17:57:11 +00:00
"flock fattr inet unix dns tty "
"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));
}
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));
}
}
Make pledge() and unveil() work amazingly This change reconciles our pledge() implementation with the OpenBSD kernel source code. We now a polyfill that's much closer to OpenBSD's behavior. For example, it was discovered that "stdio" permits threads. There were a bunch of Linux system calls that needed to be added, like sched_yield(). The exec / execnative category division is now dropped. We're instead using OpenBSD's "prot_exec" promise for launching APE binaries and dynamic shared objects. We also now filter clone() flags. The pledge.com command has been greatly improved. It now does unveiling by default when Landlock is available. It's now smart enough to unveil a superset of paths that OpenBSD automatically unveils with pledge(), such as /etc/localtime. pledge.com also now checks if the executable being launched is a dynamic shared object, in which case it unveils libraries. These changes now make it possible to pledge curl on ubuntu 20.04 glibc: pledge.com -p 'stdio rpath prot_exec inet dns tty sendfd recvfd' \ curl -s https://justine.lol/hello.txt Here's what pledging curl on Alpine 3.16 with Musl Libc looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ curl -s https://justine.lol/hello.txt Here's what pledging curl.com w/ ape loader looks like: pledge.com -p 'stdio rpath prot_exec dns inet' \ o//examples/curl.com https://justine.lol/hello.txt The most secure sandbox, is curl.com converted to static ELF: o//tool/build/assimilate.com o//examples/curl.com pledge.com -p 'stdio rpath dns inet' \ o//examples/curl.com https://justine.lol/hello.txt A weird corner case needed to be handled when resolving symbolic links during the unveiling process, that's arguably a Landlock bug. It's not surprising since Musl and Glibc are also inconsistent here too.
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);
}