diff options
Diffstat (limited to 'lib/mlibc/sysdeps/linux/generic')
-rw-r--r-- | lib/mlibc/sysdeps/linux/generic/cxx-syscall.hpp | 118 | ||||
-rw-r--r-- | lib/mlibc/sysdeps/linux/generic/entry.cpp | 38 | ||||
-rw-r--r-- | lib/mlibc/sysdeps/linux/generic/sysdeps.cpp | 2002 | ||||
-rw-r--r-- | lib/mlibc/sysdeps/linux/generic/thread.cpp | 60 |
4 files changed, 2218 insertions, 0 deletions
diff --git a/lib/mlibc/sysdeps/linux/generic/cxx-syscall.hpp b/lib/mlibc/sysdeps/linux/generic/cxx-syscall.hpp new file mode 100644 index 0000000..aa7d17c --- /dev/null +++ b/lib/mlibc/sysdeps/linux/generic/cxx-syscall.hpp @@ -0,0 +1,118 @@ +#pragma once + +#include <errno.h> +#include <mlibc/tcb.hpp> +#include <mlibc/thread.hpp> +#include <mlibc-config.h> +#include <utility> + +#include <sys/syscall.h> +#include <bits/syscall.h> + +using sc_word_t = __sc_word_t; + +extern "C" { + extern sc_word_t __mlibc_do_asm_cp_syscall(int sc, sc_word_t arg1, sc_word_t arg2, + sc_word_t arg3, sc_word_t arg4, sc_word_t arg5, sc_word_t arg6); + + extern void __mlibc_do_cancel(); +} + +namespace mlibc { + // C++ wrappers for the extern "C" functions. + inline sc_word_t do_nargs_syscall(int sc) { + return __do_syscall0(sc); + } + inline sc_word_t do_nargs_syscall(int sc, sc_word_t arg1) { + return __do_syscall1(sc, arg1); + } + inline sc_word_t do_nargs_syscall(int sc, sc_word_t arg1, sc_word_t arg2) { + return __do_syscall2(sc, arg1, arg2); + } + inline sc_word_t do_nargs_syscall(int sc, sc_word_t arg1, sc_word_t arg2, sc_word_t arg3) { + return __do_syscall3(sc, arg1, arg2, arg3); + } + inline sc_word_t do_nargs_syscall(int sc, sc_word_t arg1, sc_word_t arg2, sc_word_t arg3, + sc_word_t arg4) { + return __do_syscall4(sc, arg1, arg2, arg3, arg4); + } + inline sc_word_t do_nargs_syscall(int sc, sc_word_t arg1, sc_word_t arg2, sc_word_t arg3, + sc_word_t arg4, sc_word_t arg5) { + return __do_syscall5(sc, arg1, arg2, arg3, arg4, arg5); + } + inline sc_word_t do_nargs_syscall(int sc, sc_word_t arg1, sc_word_t arg2, sc_word_t arg3, + sc_word_t arg4, sc_word_t arg5, sc_word_t arg6) { + return __do_syscall6(sc, arg1, arg2, arg3, arg4, arg5, arg6); + } + + inline sc_word_t do_nargs_cp_syscall(int sc, sc_word_t arg1) { + return __mlibc_do_asm_cp_syscall(sc, arg1, 0, 0, 0, 0, 0); + } + inline sc_word_t do_nargs_cp_syscall(int sc, sc_word_t arg1, sc_word_t arg2) { + return __mlibc_do_asm_cp_syscall(sc, arg1, arg2, 0, 0, 0, 0); + } + inline sc_word_t do_nargs_cp_syscall(int sc, sc_word_t arg1, sc_word_t arg2, + sc_word_t arg3) { + return __mlibc_do_asm_cp_syscall(sc, arg1, arg2, arg3, 0, 0, 0); + } + inline sc_word_t do_nargs_cp_syscall(int sc, sc_word_t arg1, sc_word_t arg2, sc_word_t arg3, + sc_word_t arg4) { + return __mlibc_do_asm_cp_syscall(sc, arg1, arg2, arg3, arg4, 0, 0); + } + inline sc_word_t do_nargs_cp_syscall(int sc, sc_word_t arg1, sc_word_t arg2, sc_word_t arg3, + sc_word_t arg4, sc_word_t arg5) { + return __mlibc_do_asm_cp_syscall(sc, arg1, arg2, arg3, arg4, arg5, 0); + } + inline sc_word_t do_nargs_cp_syscall(int sc, sc_word_t arg1, sc_word_t arg2, sc_word_t arg3, + sc_word_t arg4, sc_word_t arg5, sc_word_t arg6) { + return __mlibc_do_asm_cp_syscall(sc, arg1, arg2, arg3, arg4, arg5, arg6); + } + + // Type-safe syscall result type. + enum class sc_result_t : sc_word_t { }; + + // Cast to the argument type of the extern "C" functions. + inline sc_word_t sc_cast(long x) { return x; } + inline sc_word_t sc_cast(const void *x) { return reinterpret_cast<sc_word_t>(x); } + + template<typename... T> + sc_result_t do_syscall(int sc, T... args) { + return static_cast<sc_result_t>(do_nargs_syscall(sc, sc_cast(args)...)); + } + + inline int sc_error(sc_result_t ret) { + auto v = static_cast<sc_word_t>(ret); + if(static_cast<unsigned long>(v) > -4096UL) + return -v; + return 0; + } + + template<typename... T> + sc_result_t do_cp_syscall(int sc, T... args) { +#if __MLIBC_POSIX_OPTION && !MLIBC_BUILDING_RTDL + auto result = static_cast<sc_result_t>(do_nargs_cp_syscall(sc, sc_cast(args)...)); + if (int e = sc_error(result); e) { + auto tcb = reinterpret_cast<Tcb*>(get_current_tcb()); + if (tcb_cancelled(tcb->cancelBits) && e == EINTR) { + __mlibc_do_cancel(); + __builtin_unreachable(); + } + } + return result; +#else + return do_syscall(sc, std::forward<T>(args)...); +#endif // __MLIBC_POSIX_OPTION || !MLIBC_BUILDING_RTDL + } + // Cast from the syscall result type. + template<typename T> + T sc_int_result(sc_result_t ret) { + auto v = static_cast<sc_word_t>(ret); + return v; + } + + template<typename T> + T *sc_ptr_result(sc_result_t ret) { + auto v = static_cast<sc_word_t>(ret); + return reinterpret_cast<T *>(v); + } +} diff --git a/lib/mlibc/sysdeps/linux/generic/entry.cpp b/lib/mlibc/sysdeps/linux/generic/entry.cpp new file mode 100644 index 0000000..aa049ce --- /dev/null +++ b/lib/mlibc/sysdeps/linux/generic/entry.cpp @@ -0,0 +1,38 @@ +#include <stdint.h> +#include <stdlib.h> +#include <bits/ensure.h> +#include <mlibc/elf/startup.h> +#include <sys/auxv.h> + +// defined by the POSIX library +void __mlibc_initLocale(); + +extern "C" uintptr_t *__dlapi_entrystack(); +extern "C" void __dlapi_enter(uintptr_t *); + +extern char **environ; +static mlibc::exec_stack_data __mlibc_stack_data; + +size_t __hwcap; + +struct LibraryGuard { + LibraryGuard(); +}; + +static LibraryGuard guard; + +LibraryGuard::LibraryGuard() { + __mlibc_initLocale(); + + // Parse the exec() stack. + mlibc::parse_exec_stack(__dlapi_entrystack(), &__mlibc_stack_data); + mlibc::set_startup_data(__mlibc_stack_data.argc, __mlibc_stack_data.argv, + __mlibc_stack_data.envp); +} + +extern "C" void __mlibc_entry(uintptr_t *entry_stack, int (*main_fn)(int argc, char *argv[], char *env[])) { + __dlapi_enter(entry_stack); + __hwcap = getauxval(AT_HWCAP); + auto result = main_fn(__mlibc_stack_data.argc, __mlibc_stack_data.argv, environ); + exit(result); +} diff --git a/lib/mlibc/sysdeps/linux/generic/sysdeps.cpp b/lib/mlibc/sysdeps/linux/generic/sysdeps.cpp new file mode 100644 index 0000000..bc33a9e --- /dev/null +++ b/lib/mlibc/sysdeps/linux/generic/sysdeps.cpp @@ -0,0 +1,2002 @@ +#include <asm/ioctls.h> +#include <errno.h> +#include <limits.h> + +#include <type_traits> + +#include <mlibc-config.h> +#include <bits/ensure.h> +#include <abi-bits/fcntl.h> +#include <abi-bits/socklen_t.h> +#include <mlibc/allocator.hpp> +#include <mlibc/debug.hpp> +#include <mlibc/all-sysdeps.hpp> +#include <mlibc/thread-entry.hpp> +#include <limits.h> +#include <sys/syscall.h> +#include "cxx-syscall.hpp" + +#define STUB_ONLY { __ensure(!"STUB_ONLY function was called"); __builtin_unreachable(); } +#define UNUSED(x) (void)(x); + +#ifndef MLIBC_BUILDING_RTDL +extern "C" long __do_syscall_ret(unsigned long ret) { + if(ret > -4096UL) { + errno = -ret; + return -1; + } + return ret; +} +#endif + +namespace mlibc { + +void sys_libc_log(const char *message) { + size_t n = 0; + while(message[n]) + n++; + do_syscall(SYS_write, 2, message, n); + char lf = '\n'; + do_syscall(SYS_write, 2, &lf, 1); +} + +void sys_libc_panic() { + __builtin_trap(); +} + +#if defined(__i386__) + +struct user_desc { + unsigned int entry_number; + unsigned long base_addr; + unsigned int limit; + unsigned int seg_32bit: 1; + unsigned int contents: 2; + unsigned int read_exec_only: 1; + unsigned int limit_in_pages: 1; + unsigned int seg_not_present: 1; + unsigned int useable: 1; +}; + +#endif + +int sys_tcb_set(void *pointer) { +#if defined(__x86_64__) + auto ret = do_syscall(SYS_arch_prctl, 0x1002 /* ARCH_SET_FS */, pointer); + if(int e = sc_error(ret); e) + return e; +#elif defined(__i386__) + struct user_desc desc = { + .entry_number = static_cast<unsigned int>(-1), + .base_addr = uintptr_t(pointer), + .limit = 0xfffff, + .seg_32bit = 1, + .contents = 0, + .read_exec_only = 0, + .limit_in_pages = 1, + .seg_not_present = 0, + .useable = 1, + }; + auto ret = do_syscall(SYS_set_thread_area, &desc); + __ensure(!sc_error(ret)); + asm volatile ("movw %w0, %%gs" : : "q"(desc.entry_number * 8 + 3) :); +#elif defined(__riscv) + uintptr_t thread_data = reinterpret_cast<uintptr_t>(pointer) + sizeof(Tcb); + asm volatile ("mv tp, %0" :: "r"(thread_data)); +#elif defined (__aarch64__) + uintptr_t thread_data = reinterpret_cast<uintptr_t>(pointer) + sizeof(Tcb) - 0x10; + asm volatile ("msr tpidr_el0, %0" :: "r"(thread_data)); +#else +#error "Missing architecture specific code." +#endif + return 0; +} + +int sys_anon_allocate(size_t size, void **pointer) { + return sys_vm_map(nullptr, size, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, + -1, 0, pointer); +} +int sys_anon_free(void *pointer, size_t size) { + return sys_vm_unmap(pointer, size); +} + +int sys_fadvise(int fd, off_t offset, off_t length, int advice) { + auto ret = do_syscall(SYS_fadvise64, fd, offset, length, advice); + if(int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_open(const char *path, int flags, mode_t mode, int *fd) { + auto ret = do_cp_syscall(SYS_openat, AT_FDCWD, path, flags, mode); + if(int e = sc_error(ret); e) + return e; + *fd = sc_int_result<int>(ret); + return 0; +} + +int sys_openat(int dirfd, const char *path, int flags, mode_t mode, int *fd) { + auto ret = do_syscall(SYS_openat, dirfd, path, flags, mode); + if (int e = sc_error(ret); e) + return e; + *fd = sc_int_result<int>(ret); + return 0; +} + +int sys_close(int fd) { + auto ret = do_cp_syscall(SYS_close, fd); + if(int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_dup2(int fd, int flags, int newfd) { + auto ret = do_cp_syscall(SYS_dup3, fd, newfd, flags); + if(int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_read(int fd, void *buffer, size_t size, ssize_t *bytes_read) { + auto ret = do_cp_syscall(SYS_read, fd, buffer, size); + if(int e = sc_error(ret); e) + return e; + *bytes_read = sc_int_result<ssize_t>(ret); + return 0; +} + +int sys_readv(int fd, const struct iovec *iovs, int iovc, ssize_t *bytes_read) { + auto ret = do_cp_syscall(SYS_readv, fd, iovs, iovc); + if(int e = sc_error(ret); e) + return e; + *bytes_read = sc_int_result<ssize_t>(ret); + return 0; +} + +int sys_write(int fd, const void *buffer, size_t size, ssize_t *bytes_written) { + auto ret = do_cp_syscall(SYS_write, fd, buffer, size); + if(int e = sc_error(ret); e) + return e; + if(bytes_written) + *bytes_written = sc_int_result<ssize_t>(ret); + return 0; +} + +int sys_seek(int fd, off_t offset, int whence, off_t *new_offset) { + auto ret = do_syscall(SYS_lseek, fd, offset, whence); + if(int e = sc_error(ret); e) + return e; + *new_offset = sc_int_result<off_t>(ret); + return 0; +} + +int sys_chmod(const char *pathname, mode_t mode) { + auto ret = do_cp_syscall(SYS_fchmodat, AT_FDCWD, pathname, mode); + if(int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_fchmod(int fd, mode_t mode) { + auto ret = do_cp_syscall(SYS_fchmod, fd, mode); + if(int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_fchmodat(int fd, const char *pathname, mode_t mode, int flags) { + auto ret = do_cp_syscall(SYS_fchmodat, fd, pathname, mode, flags); + if(int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_fchownat(int dirfd, const char *pathname, uid_t owner, gid_t group, int flags) { + auto ret = do_cp_syscall(SYS_fchownat, dirfd, pathname, owner, group, flags); + if(int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_utimensat(int dirfd, const char *pathname, const struct timespec times[2], int flags) { + auto ret = do_cp_syscall(SYS_utimensat, dirfd, pathname, times, flags); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_vm_map(void *hint, size_t size, int prot, int flags, + int fd, off_t offset, void **window) { + if(offset % 4096) + return EINVAL; + if(size >= PTRDIFF_MAX) + return ENOMEM; +#if defined(SYS_mmap2) + auto ret = do_syscall(SYS_mmap2, hint, size, prot, flags, fd, offset/4096); +#else + auto ret = do_syscall(SYS_mmap, hint, size, prot, flags, fd, offset); +#endif + // TODO: musl fixes up EPERM errors from the kernel. + if(int e = sc_error(ret); e) + return e; + *window = sc_ptr_result<void>(ret); + return 0; +} + +int sys_vm_unmap(void *pointer, size_t size) { + auto ret = do_syscall(SYS_munmap, pointer, size); + if(int e = sc_error(ret); e) + return e; + return 0; +} + +// All remaining functions are disabled in ldso. +#ifndef MLIBC_BUILDING_RTDL + +int sys_clock_get(int clock, time_t *secs, long *nanos) { + struct timespec tp = {}; + auto ret = do_syscall(SYS_clock_gettime, clock, &tp); + if (int e = sc_error(ret); e) + return e; + *secs = tp.tv_sec; + *nanos = tp.tv_nsec; + return 0; +} + +int sys_clock_getres(int clock, time_t *secs, long *nanos) { + struct timespec tp = {}; + auto ret = do_syscall(SYS_clock_getres, clock, &tp); + if (int e = sc_error(ret); e) + return e; + *secs = tp.tv_sec; + *nanos = tp.tv_nsec; + return 0; +} + +int sys_stat(fsfd_target fsfdt, int fd, const char *path, int flags, struct stat *statbuf) { + if (fsfdt == fsfd_target::path) + fd = AT_FDCWD; + else if (fsfdt == fsfd_target::fd) + flags |= AT_EMPTY_PATH; + else + __ensure(fsfdt == fsfd_target::fd_path); + +#if defined(SYS_newfstatat) + auto ret = do_cp_syscall(SYS_newfstatat, fd, path, statbuf, flags); +#else + auto ret = do_cp_syscall(SYS_fstatat64, fd, path, statbuf, flags); +#endif + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_statfs(const char *path, struct statfs *buf) { + auto ret = do_cp_syscall(SYS_statfs, path, buf); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_fstatfs(int fd, struct statfs *buf) { + auto ret = do_cp_syscall(SYS_fstatfs, fd, buf); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +extern "C" void __mlibc_signal_restore(void); +extern "C" void __mlibc_signal_restore_rt(void); + +int sys_sigaction(int signum, const struct sigaction *act, + struct sigaction *oldact) { + struct ksigaction { + void (*handler)(int); + unsigned long flags; + void (*restorer)(void); + sigset_t mask; + }; + + struct ksigaction kernel_act, kernel_oldact; + if (act) { + kernel_act.handler = act->sa_handler; + kernel_act.flags = act->sa_flags | SA_RESTORER; + kernel_act.restorer = (act->sa_flags & SA_SIGINFO) ? __mlibc_signal_restore_rt : __mlibc_signal_restore; + kernel_act.mask = act->sa_mask; + } + + static_assert(sizeof(sigset_t) == 8); + + auto ret = do_syscall(SYS_rt_sigaction, signum, act ? + &kernel_act : NULL, oldact ? + &kernel_oldact : NULL, sizeof(sigset_t)); + if (int e = sc_error(ret); e) + return e; + + if (oldact) { + oldact->sa_handler = kernel_oldact.handler; + oldact->sa_flags = kernel_oldact.flags; + oldact->sa_restorer = kernel_oldact.restorer; + oldact->sa_mask = kernel_oldact.mask; + } + return 0; +} + +int sys_socket(int domain, int type, int protocol, int *fd) { + auto ret = do_syscall(SYS_socket, domain, type, protocol); + if (int e = sc_error(ret); e) + return e; + *fd = sc_int_result<int>(ret); + return 0; +} + +int sys_msg_send(int sockfd, const struct msghdr *msg, int flags, ssize_t *length) { + auto ret = do_cp_syscall(SYS_sendmsg, sockfd, msg, flags); + if (int e = sc_error(ret); e) + return e; + *length = sc_int_result<ssize_t>(ret); + return 0; +} + +ssize_t sys_sendto(int fd, const void *buffer, size_t size, int flags, const struct sockaddr *sock_addr, socklen_t addr_length, ssize_t *length) { + auto ret = do_cp_syscall(SYS_sendto, fd, buffer, size, flags, sock_addr, addr_length); + if(int e = sc_error(ret); e) { + return e; + } + *length = sc_int_result<ssize_t>(ret); + return 0; +} + +ssize_t sys_recvfrom(int fd, void *buffer, size_t size, int flags, struct sockaddr *sock_addr, socklen_t *addr_length, ssize_t *length) { + auto ret = do_cp_syscall(SYS_recvfrom, fd, buffer, size, flags, sock_addr, addr_length); + if(int e = sc_error(ret); e) { + return e; + } + *length = sc_int_result<ssize_t>(ret); + return 0; +} + +int sys_msg_recv(int sockfd, struct msghdr *msg, int flags, ssize_t *length) { + auto ret = do_cp_syscall(SYS_recvmsg, sockfd, msg, flags); + if (int e = sc_error(ret); e) + return e; + *length = sc_int_result<ssize_t>(ret); + return 0; +} + +int sys_fcntl(int fd, int cmd, va_list args, int *result) { + auto arg = va_arg(args, unsigned long); + // TODO: the api for linux differs for each command so fcntl()s might fail with -EINVAL + // we should implement all the different fcntl()s + // TODO(geert): only some fcntl()s can fail with -EINTR, making do_cp_syscall useless + // on most fcntls(). Another reason to handle different fcntl()s seperately. + auto ret = do_cp_syscall(SYS_fcntl, fd, cmd, arg); + if (int e = sc_error(ret); e) + return e; + *result = sc_int_result<int>(ret); + return 0; +} + +int sys_getcwd(char *buf, size_t size) { + auto ret = do_syscall(SYS_getcwd, buf, size); + if (int e = sc_error(ret); e) { + return e; + } + return 0; +} + +int sys_unlinkat(int dfd, const char *path, int flags) { + auto ret = do_syscall(SYS_unlinkat, dfd, path, flags); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_sleep(time_t *secs, long *nanos) { + struct timespec req = { + .tv_sec = *secs, + .tv_nsec = *nanos + }; + struct timespec rem = {}; + + auto ret = do_cp_syscall(SYS_nanosleep, &req, &rem); + if (int e = sc_error(ret); e) + return e; + + *secs = rem.tv_sec; + *nanos = rem.tv_nsec; + return 0; +} + +int sys_isatty(int fd) { + unsigned short winsizeHack[4]; + auto ret = do_syscall(SYS_ioctl, fd, 0x5413 /* TIOCGWINSZ */, &winsizeHack); + if (int e = sc_error(ret); e) + return e; + auto res = sc_int_result<unsigned long>(ret); + if(!res) return 0; + return 1; +} + +#if __MLIBC_POSIX_OPTION + +#include <net/if.h> +#include <sys/ioctl.h> +#include <sys/ipc.h> +#include <sys/user.h> +#include <sys/utsname.h> +#include <sys/stat.h> +#include <sys/sysinfo.h> +#include <unistd.h> +#include <stdio.h> +#include <stdlib.h> +#include <sched.h> +#include <fcntl.h> +#include <pthread.h> + +int sys_ioctl(int fd, unsigned long request, void *arg, int *result) { + auto ret = do_syscall(SYS_ioctl, fd, request, arg); + if (int e = sc_error(ret); e) + return e; + if (result) + *result = sc_int_result<unsigned long>(ret); + return 0; +} + +int sys_connect(int sockfd, const struct sockaddr *addr, socklen_t addrlen) { + auto ret = do_cp_syscall(SYS_connect, sockfd, addr, addrlen); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_pselect(int nfds, fd_set *readfds, fd_set *writefds, + fd_set *exceptfds, const struct timespec *timeout, const sigset_t *sigmask, int *num_events) { + // The Linux kernel really wants 7 arguments, even tho this is not supported + // To fix that issue, they use a struct as the last argument. + // See the man page of pselect and the glibc source code + struct { + sigset_t ss; + size_t ss_len; + } data; + data.ss = (uintptr_t)sigmask; + data.ss_len = NSIG / 8; + + auto ret = do_cp_syscall(SYS_pselect6, nfds, readfds, writefds, + exceptfds, timeout, &data); + if (int e = sc_error(ret); e) + return e; + *num_events = sc_int_result<int>(ret); + return 0; +} + +int sys_pipe(int *fds, int flags) { + if(flags) { + auto ret = do_syscall(SYS_pipe2, fds, flags); + if (int e = sc_error(ret); e) + return e; + return 0; + } else { + auto ret = do_syscall(SYS_pipe2, fds, 0); + if (int e = sc_error(ret); e) + return e; + return 0; + } +} + +int sys_fork(pid_t *child) { + auto ret = do_syscall(SYS_clone, SIGCHLD, 0); + if (int e = sc_error(ret); e) + return e; + *child = sc_int_result<int>(ret); + return 0; +} + +int sys_waitpid(pid_t pid, int *status, int flags, struct rusage *ru, pid_t *ret_pid) { + auto ret = do_syscall(SYS_wait4, pid, status, flags, ru); + if (int e = sc_error(ret); e) + return e; + *ret_pid = sc_int_result<pid_t>(ret); + return 0; +} + +int sys_execve(const char *path, char *const argv[], char *const envp[]) { + auto ret = do_syscall(SYS_execve, path, argv, envp); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_sigprocmask(int how, const sigset_t *set, sigset_t *old) { + auto ret = do_syscall(SYS_rt_sigprocmask, how, set, old, NSIG / 8); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_setresuid(uid_t ruid, uid_t euid, uid_t suid) { + auto ret = do_syscall(SYS_setresuid, ruid, euid, suid); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid) { + auto ret = do_syscall(SYS_setresgid, rgid, egid, sgid); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_getresuid(uid_t *ruid, uid_t *euid, uid_t *suid) { + auto ret = do_syscall(SYS_getresuid, &ruid, &euid, &suid); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_getresgid(gid_t *rgid, gid_t *egid, gid_t *sgid) { + auto ret = do_syscall(SYS_getresgid, &rgid, &egid, &sgid); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_setreuid(uid_t ruid, uid_t euid) { + auto ret = do_syscall(SYS_setreuid, ruid, euid); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_setregid(gid_t rgid, gid_t egid) { + auto ret = do_syscall(SYS_setregid, rgid, egid); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_sysinfo(struct sysinfo *info) { + auto ret = do_syscall(SYS_sysinfo, info); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +void sys_yield() { + do_syscall(SYS_sched_yield); +} + +int sys_clone(void *tcb, pid_t *pid_out, void *stack) { + unsigned long flags = CLONE_VM | CLONE_FS | CLONE_FILES | CLONE_SIGHAND + | CLONE_THREAD | CLONE_SYSVSEM | CLONE_SETTLS + | CLONE_PARENT_SETTID; + +#if defined(__riscv) + // TP should point to the address immediately after the TCB. + // TODO: We should change the sysdep so that we don't need to do this. + auto tls = reinterpret_cast<char *>(tcb) + sizeof(Tcb); + tcb = reinterpret_cast<void *>(tls); +#elif defined(__aarch64__) + // TP should point to the address 16 bytes before the end of the TCB. + // TODO: We should change the sysdep so that we don't need to do this. + auto tp = reinterpret_cast<char *>(tcb) + sizeof(Tcb) - 0x10; + tcb = reinterpret_cast<void *>(tp); +#elif defined(__i386__) + /* get the entry number, as we don't request a new one here */ + uint32_t gs; + asm volatile("movw %%gs, %w0" : "=q"(gs)); + + auto user_desc = reinterpret_cast<struct user_desc *>(getAllocator().allocate(sizeof(struct user_desc))); + + user_desc->entry_number = (gs & 0xffff) >> 3; + user_desc->base_addr = uintptr_t(tcb); + user_desc->limit = 0xfffff; + user_desc->seg_32bit = 1; + user_desc->contents = 0; + user_desc->read_exec_only = 0; + user_desc->limit_in_pages = 1; + user_desc->seg_not_present = 0; + user_desc->useable = 1; + + tcb = reinterpret_cast<void *>(user_desc); +#endif + + auto ret = __mlibc_spawn_thread(flags, stack, pid_out, NULL, tcb); + if (ret < 0) + return ret; + + return 0; +} + +extern "C" const char __mlibc_syscall_begin[1]; +extern "C" const char __mlibc_syscall_end[1]; + +#if defined(__riscv) +// Disable UBSan here to work around qemu-user misaligning ucontext_t. +// https://github.com/qemu/qemu/blob/2bf40d0841b942e7ba12953d515e62a436f0af84/linux-user/riscv/signal.c#L68-L69 +[[gnu::no_sanitize("undefined")]] +#endif +int sys_before_cancellable_syscall(ucontext_t *uct) { +#if defined(__x86_64__) + auto pc = reinterpret_cast<void*>(uct->uc_mcontext.gregs[REG_RIP]); +#elif defined(__i386__) + auto pc = reinterpret_cast<void*>(uct->uc_mcontext.gregs[REG_EIP]); +#elif defined(__riscv) + auto pc = reinterpret_cast<void*>(uct->uc_mcontext.gregs[REG_PC]); +#elif defined(__aarch64__) + auto pc = reinterpret_cast<void*>(uct->uc_mcontext.pc); +#else +#error "Missing architecture specific code." +#endif + if (pc < __mlibc_syscall_begin || pc > __mlibc_syscall_end) + return 0; + return 1; +} + +int sys_tgkill(int tgid, int tid, int sig) { + auto ret = do_syscall(SYS_tgkill, tgid, tid, sig); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_tcgetattr(int fd, struct termios *attr) { + auto ret = do_syscall(SYS_ioctl, fd, TCGETS, attr); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_tcsetattr(int fd, int optional_action, const struct termios *attr) { + int req; + + switch (optional_action) { + case TCSANOW: req = TCSETS; break; + case TCSADRAIN: req = TCSETSW; break; + case TCSAFLUSH: req = TCSETSF; break; + default: return EINVAL; + } + + auto ret = do_syscall(SYS_ioctl, fd, req, attr); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_tcflush(int fd, int queue) { + auto ret = do_syscall(SYS_ioctl, fd, TCFLSH, queue); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_tcdrain(int fd) { + auto ret = do_syscall(SYS_ioctl, fd, TCSBRK, 1); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_tcflow(int fd, int action) { + auto ret = do_syscall(SYS_ioctl, fd, TCXONC, action); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_access(const char *path, int mode) { + auto ret = do_syscall(SYS_faccessat, AT_FDCWD, path, mode, 0); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_faccessat(int dirfd, const char *pathname, int mode, int flags) { + auto ret = do_syscall(SYS_faccessat, dirfd, pathname, mode, flags); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_accept(int fd, int *newfd, struct sockaddr *addr_ptr, socklen_t *addr_length, int flags) { + auto ret = do_syscall(SYS_accept4, fd, addr_ptr, addr_length, flags); + if (int e = sc_error(ret); e) + return e; + *newfd = sc_int_result<int>(ret); + return 0; +} + +int sys_bind(int fd, const struct sockaddr *addr_ptr, socklen_t addr_length) { + auto ret = do_syscall(SYS_bind, fd, addr_ptr, addr_length, 0, 0, 0); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_setsockopt(int fd, int layer, int number, const void *buffer, socklen_t size) { + auto ret = do_syscall(SYS_setsockopt, fd, layer, number, buffer, size, 0); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_sockname(int fd, struct sockaddr *addr_ptr, socklen_t max_addr_length, + socklen_t *actual_length) { + auto ret = do_syscall(SYS_getsockname, fd, addr_ptr, &max_addr_length); + if (int e = sc_error(ret); e) + return e; + *actual_length = max_addr_length; + return 0; +} + +int sys_peername(int fd, struct sockaddr *addr_ptr, socklen_t max_addr_length, + socklen_t *actual_length) { + auto ret = do_syscall(SYS_getpeername, fd, addr_ptr, &max_addr_length); + if (int e = sc_error(ret); e) + return e; + *actual_length = max_addr_length; + return 0; +} + +int sys_listen(int fd, int backlog) { + auto ret = do_syscall(SYS_listen, fd, backlog, 0, 0, 0, 0); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_shutdown(int sockfd, int how) { + auto ret = do_syscall(SYS_shutdown, sockfd, how); + if (int e = sc_error(ret); e) { + return e; + } + return 0; +} + +int sys_getpriority(int which, id_t who, int *value) { + auto ret = do_syscall(SYS_getpriority, which, who); + if (int e = sc_error(ret); e) { + return e; + } + *value = 20 - sc_int_result<int>(ret); + return 0; +} + +int sys_setpriority(int which, id_t who, int prio) { + auto ret = do_syscall(SYS_setpriority, which, who, prio); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_setitimer(int which, const struct itimerval *new_value, struct itimerval *old_value) { + auto ret = do_syscall(SYS_setitimer, which, new_value, old_value); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +/* Linux' uapi does some weird union stuff in place of `sigev_tid`, which we conveniently ignore */ +struct linux_uapi_sigevent { + union sigval sigev_value; + int sigev_signo; + int sigev_notify; + int sigev_tid; +}; + +int sys_timer_create(clockid_t clk, struct sigevent *__restrict evp, timer_t *__restrict res) { + struct linux_uapi_sigevent ksev; + struct linux_uapi_sigevent *ksevp = 0; + int timer_id; + + switch(evp ? evp->sigev_notify : SIGEV_SIGNAL) { + case SIGEV_NONE: + case SIGEV_SIGNAL: { + if(evp) { + ksev.sigev_value = evp->sigev_value; + ksev.sigev_signo = evp->sigev_signo; + ksev.sigev_notify = evp->sigev_notify; + ksev.sigev_tid = 0; + ksevp = &ksev; + } + + auto ret = do_syscall(SYS_timer_create, clk, ksevp, &timer_id); + if (int e = sc_error(ret); e) { + return e; + } + *res = (void *) (intptr_t) timer_id; + break; + } + case SIGEV_THREAD: + __ensure(!"sys_timer_create with evp->sigev_notify == SIGEV_THREAD is unimplemented"); + [[fallthrough]]; + default: + return EINVAL; + } + + return 0; +} + +int sys_timer_settime(timer_t t, int flags, const struct itimerspec *__restrict val, struct itimerspec *__restrict old) { + auto ret = do_syscall(SYS_timer_settime, t, flags, val, old); + if (int e = sc_error(ret); e) { + return e; + } + return 0; +} + +int sys_timer_delete(timer_t t) { + __ensure((intptr_t) t >= 0); + auto ret = do_syscall(SYS_timer_delete, t); + if (int e = sc_error(ret); e) { + return e; + } + return 0; +} + +int sys_ptrace(long req, pid_t pid, void *addr, void *data, long *out) { + auto ret = do_syscall(SYS_ptrace, req, pid, addr, data); + if (int e = sc_error(ret); e) + return e; + *out = sc_int_result<long>(ret); + return 0; +} + +int sys_open_dir(const char *path, int *fd) { + return sys_open(path, O_DIRECTORY, 0, fd); +} + +int sys_read_entries(int handle, void *buffer, size_t max_size, size_t *bytes_read) { + auto ret = do_syscall(SYS_getdents64, handle, buffer, max_size); + if(int e = sc_error(ret); e) + return e; + *bytes_read = sc_int_result<int>(ret); + return 0; +} + +int sys_prctl(int op, va_list ap, int *out) { + unsigned long x[4]; + for(int i = 0; i < 4; i++) + x[i] = va_arg(ap, unsigned long); + + auto ret = do_syscall(SYS_prctl, op, x[0], x[1], x[2], x[3]); + if (int e = sc_error(ret); e) + return e; + *out = sc_int_result<int>(ret); + return 0; +} + +int sys_uname(struct utsname *buf) { + auto ret = do_syscall(SYS_uname, buf); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_gethostname(char *buf, size_t bufsize) { + struct utsname uname_buf; + if (auto e = sys_uname(&uname_buf); e) + return e; + + auto node_len = strlen(uname_buf.nodename); + if (node_len >= bufsize) + return ENAMETOOLONG; + + memcpy(buf, uname_buf.nodename, node_len); + buf[node_len] = '\0'; + return 0; +} + +int sys_pread(int fd, void *buf, size_t n, off_t off, ssize_t *bytes_read) { + auto ret = do_syscall(SYS_pread64, fd, buf, n, off); + if (int e = sc_error(ret); e) + return e; + *bytes_read = sc_int_result<ssize_t>(ret); + return 0; +} + +int sys_pwrite(int fd, const void *buf, size_t n, off_t off, ssize_t *bytes_written) { + auto ret = do_syscall(SYS_pwrite64, fd, buf, n, off); + if (int e = sc_error(ret); e) + return e; + *bytes_written = sc_int_result<ssize_t>(ret); + return 0; +} + +int sys_poll(struct pollfd *fds, nfds_t count, int timeout, int *num_events) { + struct timespec tm; + tm.tv_sec = timeout / 1000; + tm.tv_nsec = timeout % 1000 * 1000000; + auto ret = do_syscall(SYS_ppoll, fds, count, timeout >= 0 ? &tm : nullptr, 0, NSIG / 8); + if (int e = sc_error(ret); e) + return e; + *num_events = sc_int_result<int>(ret); + return 0; +} + +int sys_getrusage(int scope, struct rusage *usage) { + auto ret = do_syscall(SYS_getrusage, scope, usage); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_madvise(void *addr, size_t length, int advice) { + auto ret = do_syscall(SYS_madvise, addr, length, advice); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_msync(void *addr, size_t length, int flags) { + auto ret = do_syscall(SYS_msync, addr, length, flags); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_swapon(const char *path, int flags) { + auto ret = do_syscall(SYS_swapon, path, flags); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_swapoff(const char *path) { + auto ret = do_syscall(SYS_swapoff, path); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_getaffinity(pid_t pid, size_t cpusetsize, cpu_set_t *mask) { + auto ret = do_syscall(SYS_sched_getaffinity, pid, cpusetsize, mask); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_mount(const char *source, const char *target, + const char *fstype, unsigned long flags, const void *data) { + auto ret = do_syscall(SYS_mount, source, target, fstype, flags, data); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_umount2(const char *target, int flags) { + auto ret = do_syscall(SYS_umount2, target, flags); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_sethostname(const char *buffer, size_t bufsize) { + auto ret = do_syscall(SYS_sethostname, buffer, bufsize); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_epoll_create(int flags, int *fd) { + auto ret = do_syscall(SYS_epoll_create1, flags); + if (int e = sc_error(ret); e) + return e; + *fd = sc_int_result<int>(ret); + return 0; +} + +int sys_epoll_ctl(int epfd, int mode, int fd, struct epoll_event *ev) { + auto ret = do_syscall(SYS_epoll_ctl, epfd, mode, fd, ev); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_epoll_pwait(int epfd, struct epoll_event *ev, int n, int timeout, const sigset_t *sigmask, int *raised) { + auto ret = do_syscall(SYS_epoll_pwait, epfd, ev, n, timeout, sigmask, NSIG / 8); + if (int e = sc_error(ret); e) + return e; + *raised = sc_int_result<int>(ret); + return 0; +} + +int sys_eventfd_create(unsigned int initval, int flags, int *fd) { + auto ret = do_syscall(SYS_eventfd2, initval, flags); + if (int e = sc_error(ret); e) + return e; + *fd = sc_int_result<int>(ret); + return 0; +} + +int sys_signalfd_create(const sigset_t *masks, int flags, int *fd) { + auto ret = do_syscall(SYS_signalfd4, *fd, masks, sizeof(sigset_t), flags); + if (int e = sc_error(ret); e) + return e; + *fd = sc_int_result<int>(ret); + return 0; +} + +int sys_timerfd_create(int clockid, int flags, int *fd) { + auto ret = do_syscall(SYS_timerfd_create, clockid, flags); + if (int e = sc_error(ret); e) + return e; + *fd = sc_int_result<int>(ret); + return 0; +} + +int sys_timerfd_settime(int fd, int flags, const struct itimerspec *value, struct itimerspec *oldvalue) { + auto ret = do_syscall(SYS_timerfd_settime, fd, flags, value, oldvalue); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_inotify_create(int flags, int *fd) { + auto ret = do_syscall(SYS_inotify_init1, flags); + if (int e = sc_error(ret); e) + return e; + *fd = sc_int_result<int>(ret); + return 0; +} + +int sys_init_module(void *module, unsigned long length, const char *args) { + auto ret = do_syscall(SYS_init_module, module, length, args); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_delete_module(const char *name, unsigned flags) { + auto ret = do_syscall(SYS_delete_module, name, flags); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_klogctl(int type, char *bufp, int len, int *out) { + auto ret = do_syscall(SYS_syslog, type, bufp, len); + if (int e = sc_error(ret); e) + return e; + *out = sc_int_result<int>(ret); + return 0; +} + +int sys_getcpu(int *cpu) { + auto ret = do_syscall(SYS_getcpu, cpu, NULL, NULL); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_socketpair(int domain, int type_and_flags, int proto, int *fds) { + auto ret = do_syscall(SYS_socketpair, domain, type_and_flags, proto, fds, 0, 0); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_getsockopt(int fd, int layer, int number, void *__restrict buffer, socklen_t *__restrict size) { + auto ret = do_syscall(SYS_getsockopt, fd, layer, number, buffer, size, 0); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_inotify_add_watch(int ifd, const char *path, uint32_t mask, int *wd) { + auto ret = do_syscall(SYS_inotify_add_watch, ifd, path, mask); + if (int e = sc_error(ret); e) + return e; + *wd = sc_int_result<int>(ret); + return 0; +} + +int sys_inotify_rm_watch(int ifd, int wd) { + auto ret = do_syscall(SYS_inotify_rm_watch, ifd, wd); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_ttyname(int fd, char *buf, size_t size) { + if (!isatty(fd)) + return errno; + + char *procname; + if(int e = asprintf(&procname, "/proc/self/fd/%i", fd); e) + return ENOMEM; + __ensure(procname); + + ssize_t l = readlink(procname, buf, size); + free(procname); + + if (l < 0) + return errno; + else if ((size_t)l >= size) + return ERANGE; + + buf[l] = '\0'; + struct stat st1; + struct stat st2; + + if (stat(buf, &st1) || fstat(fd, &st2)) + return errno; + if (st1.st_dev != st2.st_dev || st1.st_ino != st2.st_ino) + return ENODEV; + + return 0; +} + +int sys_pause() { +#ifdef SYS_pause + auto ret = do_syscall(SYS_pause); +#else + auto ret = do_syscall(SYS_ppoll, 0, 0, 0, 0); +#endif + if (int e = sc_error(ret); e) + return e; + return EINTR; +} + +int sys_mlockall(int flags) { + auto ret = do_syscall(SYS_mlockall, flags); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_get_min_priority(int policy, int *out) { + auto ret = do_syscall(SYS_sched_get_priority_min, policy); + if (int e = sc_error(ret); e) + return e; + *out = sc_int_result<int>(ret); + + return 0; +} + +int sys_getschedparam(void *tcb, int *policy, struct sched_param *param) { + auto t = reinterpret_cast<Tcb *>(tcb); + + if(!t->tid) { + return ESRCH; + } + + auto ret_param = do_syscall(SYS_sched_getparam, t->tid, param); + if (int e = sc_error(ret_param); e) + return e; + + auto ret_sched = do_syscall(SYS_sched_getscheduler, t->tid, param); + if (int e = sc_error(ret_sched); e) + return e; + *policy = sc_int_result<int>(ret_sched); + + return 0; +} + +int sys_setschedparam(void *tcb, int policy, const struct sched_param *param) { + auto t = reinterpret_cast<Tcb *>(tcb); + + if(!t->tid) { + return ESRCH; + } + + auto ret = do_syscall(SYS_sched_setscheduler, t->tid, policy, param); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_if_indextoname(unsigned int index, char *name) { + int fd = 0; + int r = sys_socket(AF_UNIX, SOCK_DGRAM | SOCK_CLOEXEC, AF_UNSPEC, &fd); + + if(r) + return r; + + struct ifreq ifr; + ifr.ifr_ifindex = index; + + int ret = sys_ioctl(fd, SIOCGIFNAME, &ifr, NULL); + close(fd); + + if(ret) { + if(ret == ENODEV) + return ENXIO; + return ret; + } + + strncpy(name, ifr.ifr_name, IF_NAMESIZE); + + return 0; +} + +int sys_if_nametoindex(const char *name, unsigned int *ret) { + int fd = 0; + int r = sys_socket(AF_UNIX, SOCK_DGRAM | SOCK_CLOEXEC, AF_UNSPEC, &fd); + + if(r) + return r; + + struct ifreq ifr; + strncpy(ifr.ifr_name, name, sizeof ifr.ifr_name); + + r = sys_ioctl(fd, SIOCGIFINDEX, &ifr, NULL); + close(fd); + + if(r) { + return r; + } + + *ret = ifr.ifr_ifindex; + + return 0; +} + +int sys_ptsname(int fd, char *buffer, size_t length) { + int index; + if(int e = sys_ioctl(fd, TIOCGPTN, &index, NULL); e) + return e; + if((size_t)snprintf(buffer, length, "/dev/pts/%d", index) >= length) { + return ERANGE; + } + return 0; +} + +int sys_unlockpt(int fd) { + int unlock = 0; + + if(int e = sys_ioctl(fd, TIOCSPTLCK, &unlock, NULL); e) + return e; + + return 0; +} + +int sys_thread_setname(void *tcb, const char *name) { + if(strlen(name) > 15) { + return ERANGE; + } + + auto t = reinterpret_cast<Tcb *>(tcb); + char *path; + int cs = 0; + + if(asprintf(&path, "/proc/self/task/%d/comm", t->tid) < 0) { + return ENOMEM; + } + + pthread_setcancelstate(PTHREAD_CANCEL_DISABLE, &cs); + + int fd; + if(int e = sys_open(path, O_WRONLY, 0, &fd); e) { + return e; + } + + if(int e = sys_write(fd, name, strlen(name) + 1, NULL)) { + return e; + } + + sys_close(fd); + + pthread_setcancelstate(cs, 0); + + return 0; +} + +int sys_thread_getname(void *tcb, char *name, size_t size) { + auto t = reinterpret_cast<Tcb *>(tcb); + char *path; + int cs = 0; + ssize_t real_size = 0; + + if(asprintf(&path, "/proc/self/task/%d/comm", t->tid) < 0) { + return ENOMEM; + } + + pthread_setcancelstate(PTHREAD_CANCEL_DISABLE, &cs); + + int fd; + if(int e = sys_open(path, O_RDONLY | O_CLOEXEC, 0, &fd); e) { + return e; + } + + if(int e = sys_read(fd, name, size, &real_size)) { + return e; + } + + name[real_size - 1] = 0; + sys_close(fd); + + pthread_setcancelstate(cs, 0); + + if(static_cast<ssize_t>(size) <= real_size) { + return ERANGE; + } + + return 0; +} + +int sys_mlock(const void *addr, size_t length) { + auto ret = do_syscall(SYS_mlock, addr, length); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_munlock(const void *addr, size_t length) { + auto ret = do_syscall(SYS_munlock, addr, length); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_munlockall(void) { + auto ret = do_syscall(SYS_munlockall); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_mincore(void *addr, size_t length, unsigned char *vec) { + auto ret = do_syscall(SYS_mincore, addr, length, vec); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_memfd_create(const char *name, int flags, int *fd) { + auto ret = do_syscall(SYS_memfd_create, name, flags); + if (int e = sc_error(ret); e) + return e; + *fd = sc_int_result<int>(ret); + return 0; +} + +int sys_fallocate(int fd, off_t offset, size_t size) { + auto ret = do_syscall(SYS_fallocate, fd, 0, offset, size); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_flock(int fd, int options) { + auto ret = do_syscall(SYS_flock, fd, options); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_seteuid(uid_t euid) { + return sys_setresuid(-1, euid, -1); +} + +int sys_vm_remap(void *pointer, size_t size, size_t new_size, void **window) { + auto ret = do_syscall(SYS_mremap, pointer, size, new_size, MREMAP_MAYMOVE); + // TODO: musl fixes up EPERM errors from the kernel. + if(int e = sc_error(ret); e) + return e; + *window = sc_ptr_result<void>(ret); + return 0; +} + +int sys_link(const char *old_path, const char *new_path) { +#ifdef SYS_link + auto ret = do_syscall(SYS_link, old_path, new_path); + if (int e = sc_error(ret); e) + return e; + return 0; +#else + auto ret = do_syscall(SYS_linkat, AT_FDCWD, old_path, AT_FDCWD, new_path, 0); + if (int e = sc_error(ret); e) + return e; + return 0; +#endif +} + +// Inspired by musl (src/stat/statvfs.c:28 fixup function) +static void statfs_to_statvfs(struct statfs *from, struct statvfs *to) { + *to = { + .f_bsize = from->f_bsize, + .f_frsize = from->f_frsize ? from->f_frsize : from->f_bsize, + .f_blocks = from->f_blocks, + .f_bfree = from->f_bfree, + .f_bavail = from->f_bavail, + .f_files = from->f_files, + .f_ffree = from->f_ffree, + .f_favail = from->f_ffree, + .f_fsid = (unsigned long) from->f_fsid.__val[0], + .f_flag = from->f_flags, + .f_namemax = from->f_namelen, + }; +} + +int sys_statvfs(const char *path, struct statvfs *out) { + struct statfs buf; + if(auto ret = sys_statfs(path, &buf); ret != 0) { + return ret; + } + statfs_to_statvfs(&buf, out); + return 0; +} + +int sys_fstatvfs(int fd, struct statvfs *out) { + struct statfs buf; + if(auto ret = sys_fstatfs(fd, &buf); ret != 0) { + return ret; + } + statfs_to_statvfs(&buf, out); + return 0; +} + +int sys_sysconf(int num, long *ret) { + switch(num) { + case _SC_OPEN_MAX: { + struct rlimit ru; + if(int e = sys_getrlimit(RLIMIT_NOFILE, &ru); e) { + return e; + } + *ret = (ru.rlim_cur == RLIM_INFINITY) ? -1 : ru.rlim_cur; + break; + } + case _SC_NPROCESSORS_ONLN: { + cpu_set_t set; + CPU_ZERO(&set); + if(int e = sys_getaffinity(0, sizeof(set), &set); e) { + return e; + } + *ret = CPU_COUNT(&set); + break; + } + case _SC_PHYS_PAGES: { + struct sysinfo info; + if(int e = sys_sysinfo(&info); e) { + return e; + } + unsigned unit = (info.mem_unit) ? info.mem_unit : 1; + *ret = std::min(long((info.totalram * unit) / PAGE_SIZE), LONG_MAX); + break; + } + case _SC_CHILD_MAX: { + struct rlimit ru; + if(int e = sys_getrlimit(RLIMIT_NPROC, &ru); e) { + return e; + } + *ret = (ru.rlim_cur == RLIM_INFINITY) ? -1 : ru.rlim_cur; + break; + } + case _SC_LINE_MAX: { + *ret = -1; + break; + } + default: { + return EINVAL; + } + } + + return 0; +} + +int sys_semget(key_t key, int n, int fl, int *id) { + auto ret = do_syscall(SYS_semget, key, n, fl); + if(int e = sc_error(ret); e) + return e; + *id = sc_int_result<int>(ret); + return 0; +} + +int sys_semctl(int semid, int semnum, int cmd, void *semun, int *out) { + auto ret = do_syscall(SYS_semctl, semid, semnum, cmd | IPC_64, semun); + if(int e = sc_error(ret); e) + return e; + *out = sc_int_result<int>(ret); + return 0; +} + +int sys_waitid(idtype_t idtype, id_t id, siginfo_t *info, int options) { + auto ret = do_syscall(SYS_waitid, idtype, id, info, options, 0); + if(int e = sc_error(ret); e) + return e; + return sc_int_result<int>(ret); +} + +#endif // __MLIBC_POSIX_OPTION + +#if __MLIBC_LINUX_OPTION + +#include <linux/reboot.h> + +int sys_reboot(int cmd) { + auto ret = do_syscall(SYS_reboot, LINUX_REBOOT_MAGIC1, LINUX_REBOOT_MAGIC2, cmd, nullptr); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +#endif // __MLIBC_LINUX_OPTION + +int sys_times(struct tms *tms, clock_t *out) { + auto ret = do_syscall(SYS_times, tms); + if (int e = sc_error(ret); e) + return e; + *out = sc_int_result<long>(ret); + return 0; +} + +pid_t sys_getpid() { + auto ret = do_syscall(SYS_getpid); + // getpid() always succeeds. + return sc_int_result<pid_t>(ret); +} + +pid_t sys_gettid() { + auto ret = do_syscall(SYS_gettid); + // gettid() always succeeds. + return sc_int_result<pid_t>(ret); +} + +uid_t sys_getuid() { + auto ret = do_syscall(SYS_getuid); + // getuid() always succeeds. + return sc_int_result<pid_t>(ret); +} + +uid_t sys_geteuid() { + auto ret = do_syscall(SYS_geteuid); + // geteuid() always succeeds. + return sc_int_result<pid_t>(ret); +} + +gid_t sys_getgid() { + auto ret = do_syscall(SYS_getgid); + // getgid() always succeeds. + return sc_int_result<pid_t>(ret); +} + +gid_t sys_getegid() { + auto ret = do_syscall(SYS_getegid); + // getegid() always succeeds. + return sc_int_result<pid_t>(ret); +} + +int sys_kill(int pid, int sig) { + auto ret = do_syscall(SYS_kill, pid, sig); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_vm_protect(void *pointer, size_t size, int prot) { + auto ret = do_syscall(SYS_mprotect, pointer, size, prot); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +void sys_thread_exit() { + do_syscall(SYS_exit, 0); + __builtin_trap(); +} + +void sys_exit(int status) { + do_syscall(SYS_exit_group, status); + __builtin_trap(); +} + +#endif // MLIBC_BUILDING_RTDL + +#define FUTEX_WAIT 0 +#define FUTEX_WAKE 1 + +int sys_futex_tid() { + auto ret = do_syscall(SYS_gettid); + // gettid() always succeeds. + return sc_int_result<pid_t>(ret); +} + +int sys_futex_wait(int *pointer, int expected, const struct timespec *time) { + auto ret = do_cp_syscall(SYS_futex, pointer, FUTEX_WAIT, expected, time); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_futex_wake(int *pointer) { + auto ret = do_syscall(SYS_futex, pointer, FUTEX_WAKE, INT_MAX); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_sigsuspend(const sigset_t *set) { + auto ret = do_syscall(SYS_rt_sigsuspend, set, NSIG / 8); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_sigaltstack(const stack_t *ss, stack_t *oss) { + auto ret = do_syscall(SYS_sigaltstack, ss, oss); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_mkdir(const char *path, mode_t mode) { + auto ret = do_syscall(SYS_mkdirat, AT_FDCWD, path, mode); + if (int e = sc_error(ret); e) + return e; + return 0; +} + + +int sys_mkdirat(int dirfd, const char *path, mode_t mode) { + auto ret = do_syscall(SYS_mkdirat, dirfd, path, mode); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_mknodat(int dirfd, const char *path, int mode, int dev) { + auto ret = do_syscall(SYS_mknodat, dirfd, path, mode, dev); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_mkfifoat(int dirfd, const char *path, int mode) { + return sys_mknodat(dirfd, path, mode | S_IFIFO, 0); +} + +int sys_symlink(const char *target_path, const char *link_path) { + auto ret = do_syscall(SYS_symlinkat, target_path, AT_FDCWD, link_path); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_symlinkat(const char *target_path, int dirfd, const char *link_path) { + auto ret = do_syscall(SYS_symlinkat, target_path, dirfd, link_path); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_umask(mode_t mode, mode_t *old) { + auto ret = do_syscall(SYS_umask, mode); + if (int e = sc_error(ret); e) + return e; + *old = sc_int_result<mode_t>(ret); + return 0; +} + +int sys_chdir(const char *path) { + auto ret = do_syscall(SYS_chdir, path); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_fchdir(int fd) { + auto ret = do_syscall(SYS_fchdir, fd); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_rename(const char *old_path, const char *new_path) { + return sys_renameat(AT_FDCWD, old_path, AT_FDCWD, new_path); +} + +int sys_renameat(int old_dirfd, const char *old_path, int new_dirfd, const char *new_path) { +#ifdef SYS_renameat2 + auto ret = do_syscall(SYS_renameat2, old_dirfd, old_path, new_dirfd, new_path, 0); +#else + auto ret = do_syscall(SYS_renameat, old_dirfd, old_path, new_dirfd, new_path); +#endif /* defined(SYS_renameat2) */ + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_rmdir(const char *path) { + auto ret = do_syscall(SYS_unlinkat, AT_FDCWD, path, AT_REMOVEDIR); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_ftruncate(int fd, size_t size) { + auto ret = do_syscall(SYS_ftruncate, fd, size); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_readlink(const char *path, void *buf, size_t bufsiz, ssize_t *len) { + auto ret = do_syscall(SYS_readlinkat, AT_FDCWD, path, buf, bufsiz); + if (int e = sc_error(ret); e) + return e; + *len = sc_int_result<ssize_t>(ret); + return 0; +} + +int sys_getrlimit(int resource, struct rlimit *limit) { + auto ret = do_syscall(SYS_getrlimit, resource, limit); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_setrlimit(int resource, const struct rlimit *limit) { + auto ret = do_syscall(SYS_setrlimit, resource, limit); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +pid_t sys_getppid() { + auto ret = do_syscall(SYS_getppid); + // getppid() always succeeds. + return sc_int_result<pid_t>(ret); +} + +int sys_setpgid(pid_t pid, pid_t pgid) { + auto ret = do_syscall(SYS_setpgid, pid, pgid); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_getsid(pid_t pid, pid_t *sid) { + auto ret = do_syscall(SYS_getsid, pid); + if (int e = sc_error(ret); e) + return e; + *sid = sc_int_result<pid_t>(ret); + return 0; +} + +int sys_setsid(pid_t *sid) { + auto ret = do_syscall(SYS_setsid); + if (int e = sc_error(ret); e) + return e; + *sid = sc_int_result<pid_t>(ret); + return 0; +} + +int sys_setuid(uid_t uid) { + auto ret = do_syscall(SYS_setuid, uid); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_setgid(gid_t gid) { + auto ret = do_syscall(SYS_setgid, gid); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_getpgid(pid_t pid, pid_t *out) { + auto ret = do_syscall(SYS_getpgid, pid); + if (int e = sc_error(ret); e) + return e; + *out = sc_int_result<pid_t>(ret); + return 0; +} + +int sys_getgroups(size_t size, const gid_t *list, int *retval) { + auto ret = do_syscall(SYS_getgroups, size, list); + if (int e = sc_error(ret); e) + return e; + *retval = sc_int_result<int>(ret); + return 0; +} + +int sys_dup(int fd, int flags, int *newfd) { + __ensure(!flags); + auto ret = do_cp_syscall(SYS_dup, fd); + if (int e = sc_error(ret); e) + return e; + *newfd = sc_int_result<int>(ret); + return 0; +} + +void sys_sync() { + do_syscall(SYS_sync); +} + +int sys_fsync(int fd) { + auto ret = do_syscall(SYS_fsync, fd); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_fdatasync(int fd) { + auto ret = do_syscall(SYS_fdatasync, fd); + if (int e = sc_error(ret); e) + return e; + return 0; +} + +int sys_getrandom(void *buffer, size_t length, int flags, ssize_t *bytes_written) { + auto ret = do_syscall(SYS_getrandom, buffer, length, flags); + if (int e = sc_error(ret); e) + return e; + *bytes_written = sc_int_result<ssize_t>(ret); + return 0; +} + +int sys_getentropy(void *buffer, size_t length) { + ssize_t written; + return sys_getrandom(buffer, length, 0, &written); +} + +int sys_setxattr(const char *path, const char *name, const void *val, + size_t size, int flags) { + auto ret = do_syscall(SYS_setxattr, path, name, val, size, flags); + return sc_error(ret); +} + +int sys_lsetxattr(const char *path, const char *name, const void *val, + size_t size, int flags) { + auto ret = do_syscall(SYS_lsetxattr, path, name, val, size, flags); + return sc_error(ret); +} + +int sys_fsetxattr(int fd, const char *name, const void *val, + size_t size, int flags) { + auto ret = do_syscall(SYS_fsetxattr, fd, name, val, size, flags); + return sc_error(ret); +} + +int sys_getxattr(const char *path, const char *name, void *val, size_t size, + ssize_t *nread) { + auto ret = do_syscall(SYS_getxattr, path, name, val, size); + if (int e = sc_error(ret); e) { + return e; + } + + *nread = sc_int_result<ssize_t>(ret); + return 0; +} + +int sys_lgetxattr(const char *path, const char *name, void *val, size_t size, + ssize_t *nread) { + auto ret = do_syscall(SYS_lgetxattr, path, name, val, size); + if (int e = sc_error(ret); e) { + return e; + } + + *nread = sc_int_result<ssize_t>(ret); + return 0; +} + +int sys_fgetxattr(int fd, const char *name, void *val, size_t size, + ssize_t *nread) { + auto ret = do_syscall(SYS_fgetxattr, fd, name, val, size); + if (int e = sc_error(ret); e) { + return e; + } + + *nread = sc_int_result<ssize_t>(ret); + return 0; +} + +int sys_removexattr(const char *path, const char *name) { + auto ret = do_syscall(SYS_removexattr, path, name); + return sc_error(ret); +} + +int sys_lremovexattr(const char *path, const char *name) { + auto ret = do_syscall(SYS_lremovexattr, path, name); + return sc_error(ret); +} + +int sys_fremovexattr(int fd, const char *name) { + auto ret = do_syscall(SYS_fremovexattr, fd, name); + return sc_error(ret); +} + +int sys_listxattr(const char *path, char *list, size_t size, ssize_t *nread) { + auto ret = do_syscall(SYS_listxattr, path, list, size); + if (int e = sc_error(ret); e) { + return e; + } + + *nread = sc_int_result<ssize_t>(ret); + return 0; +} + +int sys_llistxattr(const char *path, char *list, size_t size, ssize_t *nread) { + auto ret = do_syscall(SYS_llistxattr, path, list, size); + if (int e = sc_error(ret); e) { + return e; + } + + *nread = sc_int_result<ssize_t>(ret); + return 0; +} + +int sys_flistxattr(int fd, char *list, size_t size, ssize_t *nread) { + auto ret = do_syscall(SYS_flistxattr, fd, list, size); + if (int e = sc_error(ret); e) { + return e; + } + + *nread = sc_int_result<ssize_t>(ret); + return 0; +} + +int sys_sigtimedwait(const sigset_t *__restrict set, siginfo_t *__restrict info, const struct timespec *__restrict timeout, int *out_signal) { + auto ret = do_syscall(SYS_rt_sigtimedwait, set, info, timeout, NSIG / 8); + + if (int e = sc_error(ret); e) + return e; + + *out_signal = sc_int_result<int>(ret); + + return 0; +} + +#if __MLIBC_BSD_OPTION +int sys_brk(void **out) { + auto ret = do_syscall(SYS_brk, 0); + if(int e = sc_error(ret); e) { + return e; + } + + *out = (void *) sc_int_result<uintptr_t>(ret); + return 0; +} +#endif // __MLIBC_BSD_OPTION + +#if __MLIBC_GLIBC_OPTION + +int sys_personality(unsigned long persona, int *out) { + auto ret = do_syscall(SYS_personality, persona); + + if(int e = sc_error(ret); e) { + return e; + } + + *out = sc_int_result<int>(ret); + return 0; +} + +int sys_ioperm(unsigned long int from, unsigned long int num, int turn_on) { +#if defined(SYS_ioperm) + auto ret = do_syscall(SYS_ioperm, from, num, turn_on); + + if(int e = sc_error(ret); e) { + return e; + } + + return 0; +#else + (void) from; + (void) num; + (void) turn_on; + return ENOSYS; +#endif +} + +int sys_iopl(int level) { +#if defined(SYS_iopl) + auto ret = do_syscall(SYS_iopl, level); + + if(int e = sc_error(ret); e) { + return e; + } + + return 0; +#else + (void) level; + return ENOSYS; +#endif +} + +#endif // __MLIBC_GLIBC_OPTION + +} // namespace mlibc diff --git a/lib/mlibc/sysdeps/linux/generic/thread.cpp b/lib/mlibc/sysdeps/linux/generic/thread.cpp new file mode 100644 index 0000000..e413e4f --- /dev/null +++ b/lib/mlibc/sysdeps/linux/generic/thread.cpp @@ -0,0 +1,60 @@ +#include <mlibc/thread-entry.hpp> +#include <mlibc/all-sysdeps.hpp> +#include <mlibc/thread.hpp> +#include <bits/ensure.h> +#include <sys/mman.h> +#include <stdint.h> +#include <stddef.h> +#include <errno.h> + +extern "C" void __mlibc_enter_thread(void *entry, void *user_arg) { + // The linux kernel already sets the TCB in sys_clone(). + auto tcb = mlibc::get_current_tcb(); + + // Wait until our parent sets up the TID. + while(!__atomic_load_n(&tcb->tid, __ATOMIC_RELAXED)) + mlibc::sys_futex_wait(&tcb->tid, 0, nullptr); + + tcb->invokeThreadFunc(entry, user_arg); + + __atomic_store_n(&tcb->didExit, 1, __ATOMIC_RELEASE); + mlibc::sys_futex_wake(&tcb->didExit); + + mlibc::sys_thread_exit(); +} + +namespace mlibc { + +static constexpr size_t default_stacksize = 0x200000; + +int sys_prepare_stack(void **stack, void *entry, void *user_arg, void *tcb, size_t *stack_size, size_t *guard_size, void **stack_base) { + (void)tcb; + if (!*stack_size) + *stack_size = default_stacksize; + + uintptr_t map; + if (*stack) { + map = reinterpret_cast<uintptr_t>(*stack); + *guard_size = 0; + } else { + map = reinterpret_cast<uintptr_t>( + mmap(nullptr, *stack_size + *guard_size, + PROT_NONE, + MAP_PRIVATE | MAP_ANONYMOUS, -1, 0) + ); + if (reinterpret_cast<void*>(map) == MAP_FAILED) + return EAGAIN; + int ret = mprotect(reinterpret_cast<void*>(map + *guard_size), *stack_size, + PROT_READ | PROT_WRITE); + if(ret) + return EAGAIN; + } + + *stack_base = reinterpret_cast<void*>(map); + auto sp = reinterpret_cast<uintptr_t*>(map + *guard_size + *stack_size); + *--sp = reinterpret_cast<uintptr_t>(user_arg); + *--sp = reinterpret_cast<uintptr_t>(entry); + *stack = reinterpret_cast<void*>(sp); + return 0; +} +} // namespace mlibc |