#include "Log.h" #include "Pledge.h" #include "fs/Pipe.h" #include "fs/VFS.h" #include "memory/MemoryManager.h" #include "sys/Syscall.h" #include "thread/Scheduler.h" #include #include #include #include #include #include #include Result sys_read(Registers* regs, SyscallArgs args) { int fd = (int)args[0]; u8* buf = (u8*)args[1]; usize size = (usize)args[2]; if (!size) return 0; if (!MemoryManager::validate_user_write(buf, size)) return err(EFAULT); Thread* current = Scheduler::current(); TRY(check_pledge(current, Promise::p_stdio)); auto& descriptor = *TRY(current->resolve_fd(fd)); if (!descriptor.is_readable()) return err(EBADF); while (descriptor.inode()->will_block_if_read()) { if (descriptor.should_block()) kernel_sleep(10); else return err(EAGAIN); if (current->interrupted) { kdbgln("signal: read interrupted by signal"); if (current->will_ignore_pending_signal()) { current->process_pending_signals(regs); continue; } return err(EINTR); } } usize nread = TRY(descriptor.inode()->read(buf, descriptor.offset, size)); if (VFS::is_seekable(descriptor.inode())) descriptor.offset += nread; return nread; } Result sys_write(Registers*, SyscallArgs args) { int fd = (int)args[0]; const u8* buf = (const u8*)args[1]; usize size = (usize)args[2]; if (!size) return 0; if (!MemoryManager::validate_user_read(buf, size)) return err(EFAULT); Thread* current = Scheduler::current(); TRY(check_pledge(current, Promise::p_stdio)); auto& descriptor = *TRY(current->resolve_fd(fd)); if (!descriptor.is_writable()) return err(EBADF); if (descriptor.should_append() && VFS::is_seekable(descriptor.inode())) descriptor.offset = descriptor.inode()->metadata().size; usize nwritten = TRY(descriptor.inode()->write(buf, descriptor.offset, size)); if (VFS::is_seekable(descriptor.inode())) descriptor.offset += nwritten; return nwritten; } Result sys_lseek(Registers*, SyscallArgs args) { int fd = (int)args[0]; off_t offset = (long)args[1]; int whence = (int)args[2]; Thread* current = Scheduler::current(); TRY(check_pledge(current, Promise::p_stdio)); auto& descriptor = *TRY(current->resolve_fd(fd)); if (descriptor.inode()->type() == VFS::InodeType::FIFO) return err(ESPIPE); if (!VFS::is_seekable(descriptor.inode())) return descriptor.offset; off_t new_offset; switch (whence) { case SEEK_SET: new_offset = offset; break; case SEEK_CUR: new_offset = TRY(safe_add((long)descriptor.offset, offset)); break; case SEEK_END: new_offset = TRY(safe_add((long)descriptor.inode()->metadata().size, offset)); break; default: return err(EINVAL); } if (new_offset < 0) return err(EINVAL); descriptor.offset = (usize)new_offset; return (u64)new_offset; } Result sys_fcntl(Registers*, SyscallArgs args) { int fd = (int)args[0]; int cmd = (int)args[1]; Thread* current = Scheduler::current(); TRY(check_pledge(current, Promise::p_stdio)); auto& descriptor = *TRY(current->resolve_fd(fd)); bool is_cloexec = true; switch (cmd) { case F_DUPFD: is_cloexec = false; [[fallthrough]]; case F_DUPFD_CLOEXEC: { int arg = (int)args[2]; int new_fd = TRY(current->allocate_fd(arg)); current->fd_table[new_fd] = descriptor; if (is_cloexec) current->fd_table[new_fd]->flags |= O_CLOEXEC; else current->fd_table[new_fd]->flags &= ~O_CLOEXEC; return (u64)new_fd; } case F_GETFD: return (u64) !!(descriptor.flags & O_CLOEXEC); case F_SETFD: { int arg = (int)args[2]; if (arg == FD_CLOEXEC) descriptor.flags |= O_CLOEXEC; else descriptor.flags &= ~O_CLOEXEC; return 0; } case F_GETFL: return (u64)descriptor.status_flags(); case F_SETFL: { int arg = (int)args[2]; descriptor.status_flags() &= ~(O_APPEND | O_NONBLOCK); arg &= (O_APPEND | O_NONBLOCK); descriptor.status_flags() |= arg; return 0; } default: return err(EINVAL); } } Result sys_ioctl(Registers*, SyscallArgs args) { int fd = (int)args[0]; int request = (int)args[1]; void* arg = (void*)args[2]; Thread* current = Scheduler::current(); auto& descriptor = *TRY(current->resolve_fd(fd)); return descriptor.inode()->ioctl(request, arg); } Result sys_isatty(Registers*, SyscallArgs args) { int fd = (int)args[0]; Thread* current = Scheduler::current(); TRY(check_pledge(current, Promise::p_stdio)); auto& descriptor = *TRY(current->resolve_fd(fd)); return descriptor.inode()->isatty(); } Result sys_dup2(Registers*, SyscallArgs args) { int oldfd = (int)args[0]; int newfd = (int)args[1]; Thread* current = Scheduler::current(); TRY(check_pledge(current, Promise::p_stdio)); if (newfd < 0 || newfd >= FD_MAX) return err(EBADF); auto descriptor = *TRY(current->resolve_fd(oldfd)); if (newfd == oldfd) return (u64)newfd; current->fd_table[newfd] = descriptor; current->fd_table[newfd]->flags &= ~O_CLOEXEC; return (u64)newfd; } Result sys_pipe(Registers*, SyscallArgs args) { int* pfds = (int*)args[0]; Thread* current = Scheduler::current(); TRY(check_pledge(current, Promise::p_stdio)); int rfd = TRY(current->allocate_fd(0)); int wfd = TRY(current->allocate_fd(rfd + 1)); if (!MemoryManager::copy_to_user_typed(pfds, &rfd)) return err(EFAULT); if (!MemoryManager::copy_to_user_typed(pfds + 1, &wfd)) return err(EFAULT); SharedPtr rpipe; SharedPtr wpipe; TRY(Pipe::create(rpipe, wpipe)); current->fd_table[rfd] = FileDescriptor { TRY(make_shared(rpipe, O_RDONLY)), 0 }; current->fd_table[wfd] = FileDescriptor { TRY(make_shared(wpipe, O_WRONLY)), 0 }; return 0; } Result sys_umask(Registers*, SyscallArgs args) { mode_t new_umask = (mode_t)args[0]; auto* current = Scheduler::current(); TRY(check_pledge(current, Promise::p_stdio)); mode_t old_umask = current->umask; current->umask = new_umask & 0777; return old_umask; } Result sys_truncate(Registers*, SyscallArgs args) { auto path = TRY(MemoryManager::strdup_from_user(args[0])); size_t length = (size_t)args[1]; auto* current = Scheduler::current(); TRY(check_pledge(current, Promise::p_wpath)); auto inode = TRY(VFS::resolve_path(path.chars(), current->auth, current->current_directory)); if (!VFS::can_write(inode, current->auth)) return err(EACCES); TRY(inode->truncate(length)); return 0; } Result sys_ftruncate(Registers*, SyscallArgs args) { int fd = (int)args[0]; size_t length = (size_t)args[1]; auto* current = Scheduler::current(); TRY(check_pledge(current, Promise::p_stdio)); auto description = TRY(current->resolve_fd(fd))->description; if (!(description->flags & O_WRONLY)) return err(EBADF); TRY(description->inode->truncate(length)); return 0; } Result sys_utimensat(Registers*, SyscallArgs args) { int dirfd = (int)args[0]; auto path = TRY(MemoryManager::strdup_from_user(args[1])); const auto* times = (const struct timespec*)args[2]; int flags = (int)args[3]; auto* current = Scheduler::current(); TRY(check_pledge(current, Promise::p_fattr)); auto inode = TRY(current->resolve_atfile(dirfd, path, flags & AT_EMPTY_PATH, !(flags & AT_SYMLINK_NOFOLLOW))); struct timespec ktimes[2]; ktimes[0].tv_sec = ktimes[1].tv_sec = 0; ktimes[0].tv_nsec = ktimes[1].tv_nsec = UTIME_NOW; if (times && !MemoryManager::copy_from_user(times, ktimes, sizeof(ktimes))) return err(EFAULT); // No permission checks are performed, since no actual modification is done, but the above checks are still // performed. if (ktimes[0].tv_nsec == UTIME_OMIT && ktimes[1].tv_nsec == UTIME_OMIT) return 0; bool allow_write_access = ktimes[0].tv_nsec == UTIME_NOW && ktimes[1].tv_nsec == UTIME_NOW; if (allow_write_access) { if (!VFS::can_write(inode, current->auth) && current->auth.euid != inode->metadata().uid && current->auth.euid != 0) return err(EACCES); } else if (current->auth.euid != inode->metadata().uid && current->auth.euid != 0) return err(EPERM); auto metadata = inode->metadata(); if (ktimes[0].tv_nsec != UTIME_OMIT) { if (ktimes[0].tv_nsec == UTIME_NOW) metadata.atime = *Timer::realtime_clock(); else { if (ktimes[0].tv_nsec < 0 || ktimes[0].tv_nsec > 999'999'999) return err(EINVAL); metadata.atime = ktimes[0]; } } if (ktimes[1].tv_nsec != UTIME_OMIT) { if (ktimes[1].tv_nsec == UTIME_NOW) metadata.mtime = *Timer::realtime_clock(); else { if (ktimes[1].tv_nsec < 0 || ktimes[1].tv_nsec > 999'999'999) return err(EINVAL); metadata.mtime = ktimes[1]; } } TRY(inode->set_metadata(metadata)); return 0; }