// SPDX-License-Identifier: MPL-2.0 use core::{num::NonZeroU64, sync::atomic::Ordering}; use ostd::{cpu::context::UserContext, sync::RwArc, task::Task, user::UserContextApi}; use super::{ posix_thread::{AsPosixThread, PosixThreadBuilder, ThreadName}, process_table, process_vm::ProcessVm, rlimit::ResourceLimits, signal::{constants::SIGCHLD, sig_disposition::SigDispositions, sig_num::SigNum}, Credentials, Pid, Process, }; use crate::{ cpu::LinuxAbi, current_userspace, fs::{ file_table::{FdFlags, FileTable}, thread_info::ThreadFsInfo, }, prelude::*, process::{pid_file::PidFile, posix_thread::allocate_posix_tid}, sched::Nice, thread::{AsThread, Tid}, }; bitflags! { #[derive(Default)] pub struct CloneFlags: u32 { const CLONE_VM = 0x00000100; /* Set if VM shared between processes. */ const CLONE_FS = 0x00000200; /* Set if fs info shared between processes. */ const CLONE_FILES = 0x00000400; /* Set if open files shared between processes. */ const CLONE_SIGHAND = 0x00000800; /* Set if signal handlers shared. */ const CLONE_PIDFD = 0x00001000; /* Set if a pidfd should be placed in parent. */ const CLONE_PTRACE = 0x00002000; /* Set if tracing continues on the child. */ const CLONE_VFORK = 0x00004000; /* Set if the parent wants the child to wake it up on mm_release. */ const CLONE_PARENT = 0x00008000; /* Set if we want to have the same parent as the cloner. */ const CLONE_THREAD = 0x00010000; /* Set to add to same thread group. */ const CLONE_NEWNS = 0x00020000; /* Set to create new namespace. */ const CLONE_SYSVSEM = 0x00040000; /* Set to shared SVID SEM_UNDO semantics. */ const CLONE_SETTLS = 0x00080000; /* Set TLS info. */ const CLONE_PARENT_SETTID = 0x00100000; /* Store TID in userlevel buffer before MM copy. */ const CLONE_CHILD_CLEARTID = 0x00200000;/* Register exit futex and memory location to clear. */ const CLONE_DETACHED = 0x00400000; /* Create clone detached. */ const CLONE_UNTRACED = 0x00800000; /* Set if the tracing process can't force CLONE_PTRACE on this clone. */ const CLONE_CHILD_SETTID = 0x01000000; /* Store TID in userlevel buffer in the child. */ const CLONE_NEWCGROUP = 0x02000000; /* New cgroup namespace. */ const CLONE_NEWUTS = 0x04000000; /* New utsname group. */ const CLONE_NEWIPC = 0x08000000; /* New ipcs. */ const CLONE_NEWUSER = 0x10000000; /* New user namespace. */ const CLONE_NEWPID = 0x20000000; /* New pid namespace. */ const CLONE_NEWNET = 0x40000000; /* New network namespace. */ const CLONE_IO = 0x80000000; /* Clone I/O context. */ } } /// An internal structure to homogenize the arguments for `clone` and /// `clone3`. /// /// From the clone(2) man page: /// /// ``` /// The following table shows the equivalence between the arguments /// of clone() and the fields in the clone_args argument supplied to /// clone3(): /// clone() clone3() Notes /// cl_args field /// flags & ~0xff flags For most flags; details /// below /// parent_tid pidfd See CLONE_PIDFD /// child_tid child_tid See CLONE_CHILD_SETTID /// parent_tid parent_tid See CLONE_PARENT_SETTID /// flags & 0xff exit_signal /// stack stack /// --- stack_size /// tls tls See CLONE_SETTLS /// --- set_tid See below for details /// --- set_tid_size /// --- cgroup See CLONE_INTO_CGROUP /// ``` #[derive(Debug, Clone, Copy, Default)] pub struct CloneArgs { pub flags: CloneFlags, pub pidfd: Option, pub child_tid: Vaddr, pub parent_tid: Option, pub exit_signal: Option, pub stack: u64, pub stack_size: Option, pub tls: u64, pub _set_tid: Option, pub _set_tid_size: Option, pub _cgroup: Option, } impl CloneArgs { /// Prepares a new [`CloneArgs`] based on the arguments for clone(2). pub fn for_clone( raw_flags: u64, parent_tid: Vaddr, child_tid: Vaddr, tls: u64, stack: u64, ) -> Result { const FLAG_MASK: u64 = 0xff; let flags = CloneFlags::from(raw_flags & !FLAG_MASK); let exit_signal = raw_flags & FLAG_MASK; // Disambiguate the `parent_tid` parameter. The field is used // both for `CLONE_PIDFD` and `CLONE_PARENT_SETTID`, so at // most only one can be specified. let (pidfd, parent_tid) = match ( flags.contains(CloneFlags::CLONE_PIDFD), flags.contains(CloneFlags::CLONE_PARENT_SETTID), ) { (false, false) => (None, None), (true, false) => (Some(parent_tid), None), (false, true) => (None, Some(parent_tid)), (true, true) => { return_errno_with_message!( Errno::EINVAL, "CLONE_PIDFD was specified with CLONE_PARENT_SETTID" ); } }; Ok(Self { flags, pidfd, child_tid, parent_tid, exit_signal: (exit_signal != 0).then(|| SigNum::from_u8(exit_signal as u8)), stack, tls, ..Default::default() }) } pub fn for_fork() -> Self { Self { exit_signal: Some(SIGCHLD), ..Default::default() } } pub fn for_vfork() -> Self { Self { flags: CloneFlags::CLONE_VFORK | CloneFlags::CLONE_VM, exit_signal: Some(SIGCHLD), ..Default::default() } } } impl From for CloneFlags { fn from(flags: u64) -> Self { // We use the lower 32 bits let clone_flags = (flags & 0xffff_ffff) as u32; CloneFlags::from_bits_truncate(clone_flags) } } impl CloneFlags { fn check_unsupported_flags(&self) -> Result<()> { let supported_flags = CloneFlags::CLONE_VM | CloneFlags::CLONE_FS | CloneFlags::CLONE_FILES | CloneFlags::CLONE_SIGHAND | CloneFlags::CLONE_PIDFD | CloneFlags::CLONE_THREAD | CloneFlags::CLONE_SYSVSEM | CloneFlags::CLONE_SETTLS | CloneFlags::CLONE_PARENT_SETTID | CloneFlags::CLONE_CHILD_SETTID | CloneFlags::CLONE_CHILD_CLEARTID | CloneFlags::CLONE_VFORK; let unsupported_flags = *self - supported_flags; if !unsupported_flags.is_empty() { warn!("contains unsupported clone flags: {:?}", unsupported_flags); } Ok(()) } } /// Clone a child thread or child process. /// /// FIXME: currently, the child process or thread will be scheduled to run at once, /// but this may not be the expected behavior. pub fn clone_child( ctx: &Context, parent_context: &UserContext, clone_args: CloneArgs, ) -> Result { clone_args.flags.check_unsupported_flags()?; if clone_args.flags.contains(CloneFlags::CLONE_THREAD) { let child_task = clone_child_task(ctx, parent_context, clone_args)?; let child_thread = child_task.as_thread().unwrap(); child_thread.run(); let child_tid = child_thread.as_posix_thread().unwrap().tid(); Ok(child_tid) } else { let child_process = clone_child_process(ctx, parent_context, clone_args)?; if clone_args.flags.contains(CloneFlags::CLONE_VFORK) { child_process.status().set_vfork_child(true); } child_process.run(); if child_process.status().is_vfork_child() { let cond = || (!child_process.status().is_vfork_child()).then_some(()); let current = ctx.process; current.children_wait_queue().wait_until(cond); } let child_pid = child_process.pid(); Ok(child_pid) } } fn clone_child_task( ctx: &Context, parent_context: &UserContext, clone_args: CloneArgs, ) -> Result> { let clone_flags = clone_args.flags; // This combination is not valid, according to the Linux man pages. See // . if !clone_flags.contains(CloneFlags::CLONE_VM | CloneFlags::CLONE_SIGHAND) { return_errno_with_message!( Errno::EINVAL, "`CLONE_THREAD` without `CLONE_VM` and `CLONE_SIGHAND` is not valid" ); } if clone_flags.contains(CloneFlags::CLONE_PIDFD) { return_errno_with_message!( Errno::EINVAL, "`CLONE_THREAD` cannot be used together with `CLONE_PIDFD`" ); } let Context { process, thread_local, posix_thread, .. } = ctx; // Clone system V semaphore clone_sysvsem(clone_flags)?; // Clone file table let child_file_table = clone_files(thread_local.borrow_file_table().unwrap(), clone_flags); // Clone fs let child_fs = clone_fs(posix_thread.fs(), clone_flags); // Clone FPU context let child_fpu_context = thread_local.fpu().clone_context(); let child_user_ctx = Arc::new(clone_user_ctx( parent_context, clone_args.stack, clone_args.stack_size, clone_args.tls, clone_flags, )); // Inherit sigmask from current thread let sig_mask = posix_thread.sig_mask().load(Ordering::Relaxed).into(); // Inherit the thread name. let thread_name = posix_thread.thread_name().lock().as_ref().cloned(); let child_tid = allocate_posix_tid(); let child_task = { let credentials = { let credentials = ctx.posix_thread.credentials(); Credentials::new_from(&credentials) }; let mut thread_builder = PosixThreadBuilder::new(child_tid, child_user_ctx, credentials) .process(posix_thread.weak_process()) .thread_name(thread_name) .sig_mask(sig_mask) .file_table(child_file_table) .fs(child_fs) .fpu_context(child_fpu_context); // Deal with SETTID/CLEARTID flags clone_parent_settid(child_tid, clone_args.parent_tid, clone_flags)?; thread_builder = clone_child_cleartid(thread_builder, clone_args.child_tid, clone_flags); thread_builder = clone_child_settid(thread_builder, clone_args.child_tid, clone_flags); thread_builder.build() }; process .tasks() .lock() .insert(child_task.clone()) .map_err(|_| Error::with_message(Errno::EINTR, "the process has exited"))?; Ok(child_task) } fn clone_child_process( ctx: &Context, parent_context: &UserContext, clone_args: CloneArgs, ) -> Result> { let Context { process, thread_local, posix_thread, .. } = ctx; let clone_flags = clone_args.flags; // Clone the virtual memory space let child_process_vm = { let parent_process_vm = process.vm(); clone_vm(parent_process_vm, clone_flags)? }; // Clone the user context let child_user_ctx = Arc::new(clone_user_ctx( parent_context, clone_args.stack, clone_args.stack_size, clone_args.tls, clone_flags, )); // Clone the file table let child_file_table = clone_files(thread_local.borrow_file_table().unwrap(), clone_flags); // Clone the filesystem information let child_fs = clone_fs(posix_thread.fs(), clone_flags); // Clone signal dispositions let child_sig_dispositions = clone_sighand(process.sig_dispositions(), clone_flags); // Clone System V semaphore clone_sysvsem(clone_flags)?; // Clone FPU context let child_fpu_context = thread_local.fpu().clone_context(); // Inherit the parent's signal mask let child_sig_mask = posix_thread.sig_mask().load(Ordering::Relaxed).into(); // Inherit the parent's resource limits let child_resource_limits = process.resource_limits().clone(); // Inherit the parent's nice value let child_nice = process.nice().load(Ordering::Relaxed); let child_tid = allocate_posix_tid(); let child = { let child_elf_path = process.executable_path(); let mut child_thread_builder = { let child_thread_name = ThreadName::new_from_executable_path(&child_elf_path)?; let credentials = { let credentials = ctx.posix_thread.credentials(); Credentials::new_from(&credentials) }; PosixThreadBuilder::new(child_tid, child_user_ctx, credentials) .thread_name(Some(child_thread_name)) .sig_mask(child_sig_mask) .file_table(child_file_table) .fs(child_fs) .fpu_context(child_fpu_context) }; // Deal with SETTID/CLEARTID flags clone_parent_settid(child_tid, clone_args.parent_tid, clone_flags)?; child_thread_builder = clone_child_cleartid(child_thread_builder, clone_args.child_tid, clone_flags); child_thread_builder = clone_child_settid(child_thread_builder, clone_args.child_tid, clone_flags); create_child_process( child_tid, posix_thread.weak_process(), &child_elf_path, child_process_vm, child_resource_limits, child_nice, child_sig_dispositions, child_thread_builder, ) }; clone_pidfd(ctx, &child, clone_flags, clone_args.pidfd)?; if let Some(sig) = clone_args.exit_signal { child.set_exit_signal(sig); }; // Sets parent process and group for child process. set_parent_and_group(process, &child); // Updates `has_child_subreaper` for the child process after inserting // it to its parent's children to make sure the `has_child_subreaper` // state of the child process will be consistent with its parent. if process.has_child_subreaper.load(Ordering::Relaxed) { child.has_child_subreaper.store(true, Ordering::Relaxed); } Ok(child) } fn clone_child_cleartid( child_builder: PosixThreadBuilder, child_tidptr: Vaddr, clone_flags: CloneFlags, ) -> PosixThreadBuilder { if clone_flags.contains(CloneFlags::CLONE_CHILD_CLEARTID) { child_builder.clear_child_tid(child_tidptr) } else { child_builder } } fn clone_child_settid( child_builder: PosixThreadBuilder, child_tidptr: Vaddr, clone_flags: CloneFlags, ) -> PosixThreadBuilder { if clone_flags.contains(CloneFlags::CLONE_CHILD_SETTID) { child_builder.set_child_tid(child_tidptr) } else { child_builder } } fn clone_parent_settid( child_tid: Tid, parent_tidptr: Option, clone_flags: CloneFlags, ) -> Result<()> { if let Some(addr) = parent_tidptr.filter(|_| clone_flags.contains(CloneFlags::CLONE_PARENT_SETTID)) { current_userspace!().write_val(addr, &child_tid)?; } Ok(()) } /// Clone child process vm. If CLONE_VM is set, both threads share the same root vmar. /// Otherwise, fork a new copy-on-write vmar. fn clone_vm(parent_process_vm: &ProcessVm, clone_flags: CloneFlags) -> Result { if clone_flags.contains(CloneFlags::CLONE_VM) { Ok(parent_process_vm.clone()) } else { ProcessVm::fork_from(parent_process_vm) } } fn clone_user_ctx( parent_context: &UserContext, new_sp: u64, stack_size: Option, tls: u64, clone_flags: CloneFlags, ) -> UserContext { let mut child_context = parent_context.clone(); // The return value of child thread is zero child_context.set_syscall_ret(0); if clone_flags.contains(CloneFlags::CLONE_VM) && !clone_flags.contains(CloneFlags::CLONE_VFORK) { // If parent and child shares the same address space and not in vfork situation, // a new stack must be specified. debug_assert!(new_sp != 0); } if new_sp != 0 { // If stack size is not 0, the `new_sp` points to the BOTTOMMOST byte of stack. if let Some(size) = stack_size { child_context.set_stack_pointer((new_sp + size.get()) as usize) } // If stack size is 0, the new_sp points to the TOPMOST byte of stack. else { child_context.set_stack_pointer(new_sp as usize); } } if clone_flags.contains(CloneFlags::CLONE_SETTLS) { child_context.set_tls_pointer(tls as usize); } child_context } fn clone_fs(parent_fs: &Arc, clone_flags: CloneFlags) -> Arc { if clone_flags.contains(CloneFlags::CLONE_FS) { parent_fs.clone() } else { Arc::new(parent_fs.as_ref().clone()) } } fn clone_files(parent_file_table: &RwArc, clone_flags: CloneFlags) -> RwArc { // if CLONE_FILES is set, the child and parent shares the same file table // Otherwise, the child will deep copy a new file table. // FIXME: the clone may not be deep copy. if clone_flags.contains(CloneFlags::CLONE_FILES) { parent_file_table.clone() } else { RwArc::new(parent_file_table.read().clone()) } } fn clone_sighand( parent_sig_dispositions: &Arc>, clone_flags: CloneFlags, ) -> Arc> { // similar to CLONE_FILES if clone_flags.contains(CloneFlags::CLONE_SIGHAND) { parent_sig_dispositions.clone() } else { Arc::new(Mutex::new(*parent_sig_dispositions.lock())) } } fn clone_sysvsem(clone_flags: CloneFlags) -> Result<()> { if clone_flags.contains(CloneFlags::CLONE_SYSVSEM) { warn!("CLONE_SYSVSEM is not supported now"); } Ok(()) } fn clone_pidfd( ctx: &Context, child: &Arc, clone_flags: CloneFlags, pidfd_addr: Option, ) -> Result<()> { if !clone_flags.contains(CloneFlags::CLONE_PIDFD) { return Ok(()); } let pidfd_addr = pidfd_addr.unwrap(); let fd = { let pid_file = PidFile::new(child.clone(), false); let file_table = ctx.thread_local.borrow_file_table(); let mut file_table_locked = file_table.unwrap().write(); file_table_locked.insert(Arc::new(pid_file), FdFlags::CLOEXEC) }; // Since `write_val` may sleep, we cannot hold the file table lock during its execution. // FIXME: Should we remove the file from the file table if the write operation fails? match ctx.user_space().write_val(pidfd_addr, &fd) { Ok(()) => Ok(()), Err(e) => { let file_table = ctx.thread_local.borrow_file_table(); let mut file_table_locked = file_table.unwrap().write(); file_table_locked.close_file(fd); Err(e) } } } #[expect(clippy::too_many_arguments)] fn create_child_process( pid: Pid, parent: Weak, executable_path: &str, process_vm: ProcessVm, resource_limits: ResourceLimits, nice: Nice, sig_dispositions: Arc>, thread_builder: PosixThreadBuilder, ) -> Arc { let child_proc = Process::new( pid, parent, executable_path.to_string(), process_vm, resource_limits, nice, sig_dispositions, ); let child_task = thread_builder.process(Arc::downgrade(&child_proc)).build(); child_proc.tasks().lock().insert(child_task).unwrap(); child_proc } fn set_parent_and_group(parent: &Process, child: &Arc) { // Lock order: children of process -> process table -> group of process // -> group inner -> session inner let mut children_mut = parent.children().lock(); let mut process_table_mut = process_table::process_table_mut(); let process_group_mut = parent.process_group.lock(); let process_group = process_group_mut.upgrade().unwrap(); let mut process_group_inner = process_group.lock(); // Put the child process in the parent's process group process_group_inner.insert_process(child.clone()); *child.process_group.lock() = Arc::downgrade(&process_group); // Put the child process in the parent's `children` field children_mut.insert(child.pid(), child.clone()); // Put the child process in the global table process_table_mut.insert(child.pid(), child.clone()); }