Cleanup Redox repo, update Rust, remove old target

kernel/src/syscall/driver.rs

@@ -0,0 +1,130 @@
+use arch;
+use arch::memory::{allocate_frames, deallocate_frames, Frame};
+use arch::paging::{entry, ActivePageTable, PhysicalAddress, VirtualAddress};
+use context;
+use context::memory::Grant;
+use syscall::error::{Error, EFAULT, ENOMEM, EPERM, ESRCH, Result};
+use syscall::flag::{MAP_WRITE, MAP_WRITE_COMBINE};
+
+fn enforce_root() -> Result<()> {
+    let contexts = context::contexts();
+    let context_lock = contexts.current().ok_or(Error::new(ESRCH))?;
+    let context = context_lock.read();
+    if context.euid == 0 {
+        Ok(())
+    } else {
+        Err(Error::new(EPERM))
+    }
+}
+
+pub fn iopl(_level: usize, _stack_base: usize) -> Result<usize> {
+    enforce_root()?;
+
+    //TODO
+    Ok(0)
+}
+
+pub fn physalloc(size: usize) -> Result<usize> {
+    enforce_root()?;
+
+    allocate_frames((size + 4095)/4096).ok_or(Error::new(ENOMEM)).map(|frame| frame.start_address().get())
+}
+
+pub fn physfree(physical_address: usize, size: usize) -> Result<usize> {
+    enforce_root()?;
+
+    deallocate_frames(Frame::containing_address(PhysicalAddress::new(physical_address)), (size + 4095)/4096);
+    //TODO: Check that no double free occured
+    Ok(0)
+}
+
+//TODO: verify exlusive access to physical memory
+pub fn physmap(physical_address: usize, size: usize, flags: usize) -> Result<usize> {
+    enforce_root()?;
+
+    if size == 0 {
+        Ok(0)
+    } else {
+        let contexts = context::contexts();
+        let context_lock = contexts.current().ok_or(Error::new(ESRCH))?;
+        let context = context_lock.read();
+
+        let mut grants = context.grants.lock();
+
+        let from_address = (physical_address/4096) * 4096;
+        let offset = physical_address - from_address;
+        let full_size = ((offset + size + 4095)/4096) * 4096;
+        let mut to_address = arch::USER_GRANT_OFFSET;
+
+        let mut entry_flags = entry::PRESENT | entry::NO_EXECUTE | entry::USER_ACCESSIBLE;
+        if flags & MAP_WRITE == MAP_WRITE {
+            entry_flags |= entry::WRITABLE;
+        }
+        if flags & MAP_WRITE_COMBINE == MAP_WRITE_COMBINE {
+            entry_flags |= entry::HUGE_PAGE;
+        }
+
+        for i in 0 .. grants.len() {
+            let start = grants[i].start_address().get();
+            if to_address + full_size < start {
+                grants.insert(i, Grant::physmap(
+                    PhysicalAddress::new(from_address),
+                    VirtualAddress::new(to_address),
+                    full_size,
+                    entry_flags
+                ));
+
+                return Ok(to_address + offset);
+            } else {
+                let pages = (grants[i].size() + 4095) / 4096;
+                let end = start + pages * 4096;
+                to_address = end;
+            }
+        }
+
+        grants.push(Grant::physmap(
+            PhysicalAddress::new(from_address),
+            VirtualAddress::new(to_address),
+            full_size,
+            entry_flags
+        ));
+
+        Ok(to_address + offset)
+    }
+}
+
+pub fn physunmap(virtual_address: usize) -> Result<usize> {
+    enforce_root()?;
+
+    if virtual_address == 0 {
+        Ok(0)
+    } else {
+        let contexts = context::contexts();
+        let context_lock = contexts.current().ok_or(Error::new(ESRCH))?;
+        let context = context_lock.read();
+
+        let mut grants = context.grants.lock();
+
+        for i in 0 .. grants.len() {
+            let start = grants[i].start_address().get();
+            let end = start + grants[i].size();
+            if virtual_address >= start && virtual_address < end {
+                grants.remove(i).unmap();
+
+                return Ok(0);
+            }
+        }
+
+        Err(Error::new(EFAULT))
+    }
+}
+
+pub fn virttophys(virtual_address: usize) -> Result<usize> {
+    enforce_root()?;
+
+    let active_table = unsafe { ActivePageTable::new() };
+    match active_table.translate(VirtualAddress::new(virtual_address)) {
+        Some(physical_address) => Ok(physical_address.get()),
+        None => Err(Error::new(EFAULT))
+    }
+}

kernel/src/syscall/fs.rs

@@ -0,0 +1,356 @@
+//! Filesystem syscalls
+use core::sync::atomic::Ordering;
+
+use context;
+use scheme::{self, FileHandle};
+use syscall;
+use syscall::data::{Packet, Stat};
+use syscall::error::*;
+use syscall::flag::{MODE_DIR, MODE_FILE};
+
+pub fn file_op(a: usize, fd: FileHandle, c: usize, d: usize) -> Result<usize> {
+    let (file, pid, uid, gid) = {
+        let contexts = context::contexts();
+        let context_lock = contexts.current().ok_or(Error::new(ESRCH))?;
+        let context = context_lock.read();
+        let file = context.get_file(fd).ok_or(Error::new(EBADF))?;
+        (file, context.id, context.euid, context.egid)
+    };
+
+    let scheme = {
+        let schemes = scheme::schemes();
+        let scheme = schemes.get(file.scheme).ok_or(Error::new(EBADF))?;
+        scheme.clone()
+    };
+
+    let mut packet = Packet {
+        id: 0,
+        pid: pid.into(),
+        uid: uid,
+        gid: gid,
+        a: a,
+        b: file.number,
+        c: c,
+        d: d
+    };
+
+    scheme.handle(&mut packet);
+
+    Error::demux(packet.a)
+}
+
+pub fn file_op_slice(a: usize, fd: FileHandle, slice: &[u8]) -> Result<usize> {
+    file_op(a, fd, slice.as_ptr() as usize, slice.len())
+}
+
+pub fn file_op_mut_slice(a: usize, fd: FileHandle, slice: &mut [u8]) -> Result<usize> {
+    file_op(a, fd, slice.as_mut_ptr() as usize, slice.len())
+}
+
+/// Change the current working directory
+pub fn chdir(path: &[u8]) -> Result<usize> {
+    let fd = open(path, syscall::flag::O_RDONLY | syscall::flag::O_DIRECTORY)?;
+    let mut stat = Stat::default();
+    let stat_res = file_op_mut_slice(syscall::number::SYS_FSTAT, fd, &mut stat);
+    let _ = close(fd);
+    stat_res?;
+    if stat.st_mode & (MODE_FILE | MODE_DIR) == MODE_DIR {
+        let contexts = context::contexts();
+        let context_lock = contexts.current().ok_or(Error::new(ESRCH))?;
+        let context = context_lock.read();
+        let canonical = context.canonicalize(path);
+        *context.cwd.lock() = canonical;
+        Ok(0)
+    } else {
+        Err(Error::new(ENOTDIR))
+    }
+}
+
+/// Get the current working directory
+pub fn getcwd(buf: &mut [u8]) -> Result<usize> {
+    let contexts = context::contexts();
+    let context_lock = contexts.current().ok_or(Error::new(ESRCH))?;
+    let context = context_lock.read();
+    let cwd = context.cwd.lock();
+    let mut i = 0;
+    while i < buf.len() && i < cwd.len() {
+        buf[i] = cwd[i];
+        i += 1;
+    }
+    Ok(i)
+}
+
+/// Open syscall
+pub fn open(path: &[u8], flags: usize) -> Result<FileHandle> {
+    let (path_canon, uid, gid, scheme_ns) = {
+        let contexts = context::contexts();
+        let context_lock = contexts.current().ok_or(Error::new(ESRCH))?;
+        let context = context_lock.read();
+        (context.canonicalize(path), context.euid, context.egid, context.ens)
+    };
+
+    //println!("open {}", unsafe { ::core::str::from_utf8_unchecked(&path_canon) });
+
+    let mut parts = path_canon.splitn(2, |&b| b == b':');
+    let scheme_name_opt = parts.next();
+    let reference_opt = parts.next();
+
+    let (scheme_id, file_id) = {
+        let scheme_name = scheme_name_opt.ok_or(Error::new(ENODEV))?;
+        let (scheme_id, scheme) = {
+            let schemes = scheme::schemes();
+            let (scheme_id, scheme) = schemes.get_name(scheme_ns, scheme_name).ok_or(Error::new(ENODEV))?;
+            (scheme_id, scheme.clone())
+        };
+        let file_id = scheme.open(reference_opt.unwrap_or(b""), flags, uid, gid)?;
+        (scheme_id, file_id)
+    };
+
+    let contexts = context::contexts();
+    let context_lock = contexts.current().ok_or(Error::new(ESRCH))?;
+    let context = context_lock.read();
+    context.add_file(::context::file::File {
+        scheme: scheme_id,
+        number: file_id,
+        event: None,
+    }).ok_or(Error::new(EMFILE))
+}
+
+pub fn pipe2(fds: &mut [usize], flags: usize) -> Result<usize> {
+    if fds.len() >= 2 {
+        let scheme_id = ::scheme::pipe::PIPE_SCHEME_ID.load(Ordering::SeqCst);
+        let (read_id, write_id) = ::scheme::pipe::pipe(flags);
+
+        let contexts = context::contexts();
+        let context_lock = contexts.current().ok_or(Error::new(ESRCH))?;
+        let context = context_lock.read();
+
+        let read_fd = context.add_file(::context::file::File {
+            scheme: scheme_id,
+            number: read_id,
+            event: None,
+        }).ok_or(Error::new(EMFILE))?;
+
+        let write_fd = context.add_file(::context::file::File {
+            scheme: scheme_id,
+            number: write_id,
+            event: None,
+        }).ok_or(Error::new(EMFILE))?;
+
+        fds[0] = read_fd.into();
+        fds[1] = write_fd.into();
+
+        Ok(0)
+    } else {
+        Err(Error::new(EFAULT))
+    }
+}
+
+/// chmod syscall
+pub fn chmod(path: &[u8], mode: u16) -> Result<usize> {
+    let (path_canon, uid, gid, scheme_ns) = {
+        let contexts = context::contexts();
+        let context_lock = contexts.current().ok_or(Error::new(ESRCH))?;
+        let context = context_lock.read();
+        (context.canonicalize(path), context.euid, context.egid, context.ens)
+    };
+
+    let mut parts = path_canon.splitn(2, |&b| b == b':');
+    let scheme_name_opt = parts.next();
+    let reference_opt = parts.next();
+
+    let scheme_name = scheme_name_opt.ok_or(Error::new(ENODEV))?;
+    let scheme = {
+        let schemes = scheme::schemes();
+        let (_scheme_id, scheme) = schemes.get_name(scheme_ns, scheme_name).ok_or(Error::new(ENODEV))?;
+        scheme.clone()
+    };
+    scheme.chmod(reference_opt.unwrap_or(b""), mode, uid, gid)
+}
+
+/// rmdir syscall
+pub fn rmdir(path: &[u8]) -> Result<usize> {
+    let (path_canon, uid, gid, scheme_ns) = {
+        let contexts = context::contexts();
+        let context_lock = contexts.current().ok_or(Error::new(ESRCH))?;
+        let context = context_lock.read();
+        (context.canonicalize(path), context.euid, context.egid, context.ens)
+    };
+
+    let mut parts = path_canon.splitn(2, |&b| b == b':');
+    let scheme_name_opt = parts.next();
+    let reference_opt = parts.next();
+
+    let scheme_name = scheme_name_opt.ok_or(Error::new(ENODEV))?;
+    let scheme = {
+        let schemes = scheme::schemes();
+        let (_scheme_id, scheme) = schemes.get_name(scheme_ns, scheme_name).ok_or(Error::new(ENODEV))?;
+        scheme.clone()
+    };
+    scheme.rmdir(reference_opt.unwrap_or(b""), uid, gid)
+}
+
+/// Unlink syscall
+pub fn unlink(path: &[u8]) -> Result<usize> {
+    let (path_canon, uid, gid, scheme_ns) = {
+        let contexts = context::contexts();
+        let context_lock = contexts.current().ok_or(Error::new(ESRCH))?;
+        let context = context_lock.read();
+        (context.canonicalize(path), context.euid, context.egid, context.ens)
+    };
+
+    let mut parts = path_canon.splitn(2, |&b| b == b':');
+    let scheme_name_opt = parts.next();
+    let reference_opt = parts.next();
+
+    let scheme_name = scheme_name_opt.ok_or(Error::new(ENODEV))?;
+    let scheme = {
+        let schemes = scheme::schemes();
+        let (_scheme_id, scheme) = schemes.get_name(scheme_ns, scheme_name).ok_or(Error::new(ENODEV))?;
+        scheme.clone()
+    };
+    scheme.unlink(reference_opt.unwrap_or(b""), uid, gid)
+}
+
+/// Close syscall
+pub fn close(fd: FileHandle) -> Result<usize> {
+    let file = {
+        let contexts = context::contexts();
+        let context_lock = contexts.current().ok_or(Error::new(ESRCH))?;
+        let context = context_lock.read();
+        let file = context.remove_file(fd).ok_or(Error::new(EBADF))?;
+        file
+    };
+
+    if let Some(event_id) = file.event {
+        context::event::unregister(fd, file.scheme, event_id);
+    }
+
+    let scheme = {
+        let schemes = scheme::schemes();
+        let scheme = schemes.get(file.scheme).ok_or(Error::new(EBADF))?;
+        scheme.clone()
+    };
+    scheme.close(file.number)
+}
+
+/// Duplicate file descriptor
+pub fn dup(fd: FileHandle, buf: &[u8]) -> Result<FileHandle> {
+    let file = {
+        let contexts = context::contexts();
+        let context_lock = contexts.current().ok_or(Error::new(ESRCH))?;
+        let context = context_lock.read();
+        let file = context.get_file(fd).ok_or(Error::new(EBADF))?;
+        file
+    };
+
+    let new_id = {
+        let scheme = {
+            let schemes = scheme::schemes();
+            let scheme = schemes.get(file.scheme).ok_or(Error::new(EBADF))?;
+            scheme.clone()
+        };
+        scheme.dup(file.number, buf)?
+    };
+
+    let contexts = context::contexts();
+    let context_lock = contexts.current().ok_or(Error::new(ESRCH))?;
+    let context = context_lock.read();
+    context.add_file(::context::file::File {
+        scheme: file.scheme,
+        number: new_id,
+        event: None,
+    }).ok_or(Error::new(EMFILE))
+}
+
+/// Duplicate file descriptor, replacing another
+pub fn dup2(fd: FileHandle, new_fd: FileHandle, buf: &[u8]) -> Result<FileHandle> {
+    if fd == new_fd {
+        Ok(new_fd)
+    } else {
+        let _ = close(new_fd)?;
+
+        let file = {
+            let contexts = context::contexts();
+            let context_lock = contexts.current().ok_or(Error::new(ESRCH))?;
+            let context = context_lock.read();
+            let file = context.get_file(fd).ok_or(Error::new(EBADF))?;
+            file
+        };
+
+        let new_id = {
+            let scheme = {
+                let schemes = scheme::schemes();
+                let scheme = schemes.get(file.scheme).ok_or(Error::new(EBADF))?;
+                scheme.clone()
+            };
+            scheme.dup(file.number, buf)?
+        };
+
+        let contexts = context::contexts();
+        let context_lock = contexts.current().ok_or(Error::new(ESRCH))?;
+        let context = context_lock.read();
+        context.insert_file(new_fd, ::context::file::File {
+            scheme: file.scheme,
+            number: new_id,
+            event: None,
+        }).ok_or(Error::new(EBADF))
+    }
+}
+
+/// Register events for file
+pub fn fevent(fd: FileHandle, flags: usize) -> Result<usize> {
+    let file = {
+        let contexts = context::contexts();
+        let context_lock = contexts.current().ok_or(Error::new(ESRCH))?;
+        let context = context_lock.read();
+        let mut files = context.files.lock();
+        let mut file = files.get_mut(fd.into()).ok_or(Error::new(EBADF))?.ok_or(Error::new(EBADF))?;
+        if let Some(event_id) = file.event.take() {
+            println!("{:?}: {:?}:{}: events already registered: {}", fd, file.scheme, file.number, event_id);
+            context::event::unregister(fd, file.scheme, event_id);
+        }
+        file.clone()
+    };
+
+    let scheme = {
+        let schemes = scheme::schemes();
+        let scheme = schemes.get(file.scheme).ok_or(Error::new(EBADF))?;
+        scheme.clone()
+    };
+    let event_id = scheme.fevent(file.number, flags)?;
+    {
+        let contexts = context::contexts();
+        let context_lock = contexts.current().ok_or(Error::new(ESRCH))?;
+        let context = context_lock.read();
+        let mut files = context.files.lock();
+        let mut file = files.get_mut(fd.into()).ok_or(Error::new(EBADF))?.ok_or(Error::new(EBADF))?;
+        file.event = Some(event_id);
+    }
+    context::event::register(fd, file.scheme, event_id);
+    Ok(0)
+}
+
+pub fn funmap(virtual_address: usize) -> Result<usize> {
+    if virtual_address == 0 {
+        Ok(0)
+    } else {
+        let contexts = context::contexts();
+        let context_lock = contexts.current().ok_or(Error::new(ESRCH))?;
+        let context = context_lock.read();
+
+        let mut grants = context.grants.lock();
+
+        for i in 0 .. grants.len() {
+            let start = grants[i].start_address().get();
+            let end = start + grants[i].size();
+            if virtual_address >= start && virtual_address < end {
+                grants.remove(i).unmap();
+
+                return Ok(0);
+            }
+        }
+
+        Err(Error::new(EFAULT))
+    }
+}

kernel/src/syscall/futex.rs

@@ -0,0 +1,110 @@
+use alloc::arc::Arc;
+use collections::VecDeque;
+use core::intrinsics;
+use spin::{Once, RwLock, RwLockReadGuard, RwLockWriteGuard};
+
+use context::{self, Context};
+use syscall::error::{Error, Result, ESRCH, EAGAIN, EINVAL};
+use syscall::flag::{FUTEX_WAIT, FUTEX_WAKE, FUTEX_REQUEUE};
+use syscall::validate::validate_slice_mut;
+
+type FutexList = VecDeque<(usize, Arc<RwLock<Context>>)>;
+
+/// Fast userspace mutex list
+static FUTEXES: Once<RwLock<FutexList>> = Once::new();
+
+/// Initialize futexes, called if needed
+fn init_futexes() -> RwLock<FutexList> {
+    RwLock::new(VecDeque::new())
+}
+
+/// Get the global futexes list, const
+pub fn futexes() -> RwLockReadGuard<'static, FutexList> {
+    FUTEXES.call_once(init_futexes).read()
+}
+
+/// Get the global futexes list, mutable
+pub fn futexes_mut() -> RwLockWriteGuard<'static, FutexList> {
+    FUTEXES.call_once(init_futexes).write()
+}
+
+pub fn futex(addr: &mut i32, op: usize, val: i32, val2: usize, addr2: *mut i32) -> Result<usize> {
+    match op {
+        FUTEX_WAIT => {
+            {
+                let mut futexes = futexes_mut();
+
+                let context_lock = {
+                    let contexts = context::contexts();
+                    let context_lock = contexts.current().ok_or(Error::new(ESRCH))?;
+                    context_lock.clone()
+                };
+
+                if unsafe { intrinsics::atomic_load(addr) != val } {
+                    return Err(Error::new(EAGAIN));
+                }
+
+                context_lock.write().block();
+
+                futexes.push_back((addr as *mut i32 as usize, context_lock));
+            }
+
+            unsafe { context::switch(); }
+
+            Ok(0)
+        },
+        FUTEX_WAKE => {
+            let mut woken = 0;
+
+            {
+                let mut futexes = futexes_mut();
+
+                let mut i = 0;
+                while i < futexes.len() && (woken as i32) < val {
+                    if futexes[i].0 == addr as *mut i32 as usize {
+                        if let Some(futex) = futexes.swap_remove_back(i) {
+                            futex.1.write().unblock();
+                            woken += 1;
+                        }
+                    } else {
+                        i += 1;
+                    }
+                }
+            }
+
+            Ok(woken)
+        },
+        FUTEX_REQUEUE => {
+            let addr2_safe = validate_slice_mut(addr2, 1).map(|addr2_safe| &mut addr2_safe[0])?;
+
+            let mut woken = 0;
+            let mut requeued = 0;
+
+            {
+                let mut futexes = futexes_mut();
+
+                let mut i = 0;
+                while i < futexes.len() && (woken as i32) < val {
+                    if futexes[i].0 == addr as *mut i32 as usize {
+                        if let Some(futex) = futexes.swap_remove_back(i) {
+                            futex.1.write().unblock();
+                            woken += 1;
+                        }
+                    } else {
+                        i += 1;
+                    }
+                }
+                while i < futexes.len() && requeued < val2 {
+                    if futexes[i].0 == addr as *mut i32 as usize {
+                        futexes[i].0 = addr2_safe as *mut i32 as usize;
+                        requeued += 1;
+                    }
+                    i += 1;
+                }
+            }
+
+            Ok(woken)
+        },
+        _ => Err(Error::new(EINVAL))
+    }
+}

kernel/src/syscall/mod.rs

@@ -0,0 +1,121 @@
+///! Syscall handlers
+
+extern crate syscall;
+
+pub use self::syscall::{data, error, flag, number, scheme};
+
+pub use self::driver::*;
+pub use self::fs::*;
+pub use self::futex::futex;
+pub use self::privilege::*;
+pub use self::process::*;
+pub use self::time::*;
+pub use self::validate::*;
+
+use self::data::TimeSpec;
+use self::error::{Error, Result, ENOSYS};
+use self::number::*;
+
+use context::ContextId;
+use scheme::{FileHandle, SchemeNamespace};
+
+/// Driver syscalls
+pub mod driver;
+
+/// Filesystem syscalls
+pub mod fs;
+
+/// Fast userspace mutex
+pub mod futex;
+
+/// Privilege syscalls
+pub mod privilege;
+
+/// Process syscalls
+pub mod process;
+
+/// Time syscalls
+pub mod time;
+
+/// Validate input
+pub mod validate;
+
+#[no_mangle]
+pub extern fn syscall(a: usize, b: usize, c: usize, d: usize, e: usize, f: usize, stack: usize) -> usize {
+    #[inline(always)]
+    fn inner(a: usize, b: usize, c: usize, d: usize, e: usize, f: usize, stack: usize) -> Result<usize> {
+        match a & SYS_CLASS {
+            SYS_CLASS_FILE => {
+                let fd = FileHandle::from(b);
+                match a & SYS_ARG {
+                    SYS_ARG_SLICE => file_op_slice(a, fd, validate_slice(c as *const u8, d)?),
+                    SYS_ARG_MSLICE => file_op_mut_slice(a, fd, validate_slice_mut(c as *mut u8, d)?),
+                    _ => match a {
+                        SYS_CLOSE => close(fd),
+                        SYS_DUP => dup(fd, validate_slice(c as *const u8, d)?).map(FileHandle::into),
+                        SYS_DUP2 => dup2(fd, FileHandle::from(c), validate_slice(d as *const u8, e)?).map(FileHandle::into),
+                        SYS_FEVENT => fevent(fd, c),
+                        SYS_FUNMAP => funmap(b),
+                        _ => file_op(a, fd, c, d)
+                    }
+                }
+            },
+            SYS_CLASS_PATH => match a {
+                SYS_OPEN => open(validate_slice(b as *const u8, c)?, d).map(FileHandle::into),
+                SYS_CHMOD => chmod(validate_slice(b as *const u8, c)?, d as u16),
+                SYS_RMDIR => rmdir(validate_slice(b as *const u8, c)?),
+                SYS_UNLINK => unlink(validate_slice(b as *const u8, c)?),
+                _ => unreachable!()
+            },
+            _ => match a {
+                SYS_YIELD => sched_yield(),
+                SYS_NANOSLEEP => nanosleep(validate_slice(b as *const TimeSpec, 1).map(|req| &req[0])?, validate_slice_mut(c as *mut TimeSpec, 1).ok().map(|rem| &mut rem[0])),
+                SYS_CLOCK_GETTIME => clock_gettime(b, validate_slice_mut(c as *mut TimeSpec, 1).map(|time| &mut time[0])?),
+                SYS_FUTEX => futex(validate_slice_mut(b as *mut i32, 1).map(|uaddr| &mut uaddr[0])?, c, d as i32, e, f as *mut i32),
+                SYS_BRK => brk(b),
+                SYS_GETPID => getpid().map(ContextId::into),
+                SYS_CLONE => clone(b, stack).map(ContextId::into),
+                SYS_EXIT => exit((b & 0xFF) << 8),
+                SYS_KILL => kill(ContextId::from(b), c),
+                SYS_WAITPID => waitpid(ContextId::from(b), c, d).map(ContextId::into),
+                SYS_CHDIR => chdir(validate_slice(b as *const u8, c)?),
+                SYS_EXECVE => exec(validate_slice(b as *const u8, c)?, validate_slice(d as *const [usize; 2], e)?),
+                SYS_IOPL => iopl(b, stack),
+                SYS_GETCWD => getcwd(validate_slice_mut(b as *mut u8, c)?),
+                SYS_GETEGID => getegid(),
+                SYS_GETENS => getens(),
+                SYS_GETEUID => geteuid(),
+                SYS_GETGID => getgid(),
+                SYS_GETNS => getns(),
+                SYS_GETUID => getuid(),
+                SYS_MKNS => mkns(validate_slice(b as *const [usize; 2], c)?),
+                SYS_SETREUID => setreuid(b as u32, c as u32),
+                SYS_SETRENS => setrens(SchemeNamespace::from(b), SchemeNamespace::from(c)),
+                SYS_SETREGID => setregid(b as u32, c as u32),
+                SYS_PIPE2 => pipe2(validate_slice_mut(b as *mut usize, 2)?, c),
+                SYS_PHYSALLOC => physalloc(b),
+                SYS_PHYSFREE => physfree(b, c),
+                SYS_PHYSMAP => physmap(b, c, d),
+                SYS_PHYSUNMAP => physunmap(b),
+                SYS_VIRTTOPHYS => virttophys(b),
+                _ => Err(Error::new(ENOSYS))
+            }
+        }
+    }
+
+    let result = inner(a, b, c, d, e, f, stack);
+
+    /*
+    if let Err(ref err) = result {
+        let contexts = ::context::contexts();
+        if let Some(context_lock) = contexts.current() {
+            let context = context_lock.read();
+            print!("{}: {}: ", unsafe { ::core::str::from_utf8_unchecked(&context.name.lock()) }, context.id.into());
+        }
+
+        println!("{:X}, {:X}, {:X}, {:X}: {}", a, b, c, d, err);
+    }
+    */
+
+    Error::mux(result)
+}

kernel/src/syscall/privilege.rs

@@ -0,0 +1,147 @@
+use collections::Vec;
+
+use context;
+use scheme::{self, SchemeNamespace};
+use syscall::error::*;
+use syscall::validate::validate_slice;
+
+pub fn getegid() -> Result<usize> {
+    let contexts = context::contexts();
+    let context_lock = contexts.current().ok_or(Error::new(ESRCH))?;
+    let context = context_lock.read();
+    Ok(context.egid as usize)
+}
+
+pub fn getens() -> Result<usize> {
+    let contexts = context::contexts();
+    let context_lock = contexts.current().ok_or(Error::new(ESRCH))?;
+    let context = context_lock.read();
+    Ok(context.ens.into())
+}
+
+pub fn geteuid() -> Result<usize> {
+    let contexts = context::contexts();
+    let context_lock = contexts.current().ok_or(Error::new(ESRCH))?;
+    let context = context_lock.read();
+    Ok(context.euid as usize)
+}
+
+pub fn getgid() -> Result<usize> {
+    let contexts = context::contexts();
+    let context_lock = contexts.current().ok_or(Error::new(ESRCH))?;
+    let context = context_lock.read();
+    Ok(context.rgid as usize)
+}
+
+pub fn getns() -> Result<usize> {
+    let contexts = context::contexts();
+    let context_lock = contexts.current().ok_or(Error::new(ESRCH))?;
+    let context = context_lock.read();
+    Ok(context.rns.into())
+}
+
+pub fn getuid() -> Result<usize> {
+    let contexts = context::contexts();
+    let context_lock = contexts.current().ok_or(Error::new(ESRCH))?;
+    let context = context_lock.read();
+    Ok(context.ruid as usize)
+}
+
+pub fn mkns(name_ptrs: &[[usize; 2]]) -> Result<usize> {
+    let mut names = Vec::new();
+    for name_ptr in name_ptrs {
+        names.push(validate_slice(name_ptr[0] as *const u8, name_ptr[1])?);
+    }
+
+    let (uid, from) = {
+        let contexts = context::contexts();
+        let context_lock = contexts.current().ok_or(Error::new(ESRCH))?;
+        let context = context_lock.read();
+        (context.euid, context.ens)
+    };
+
+    if uid == 0 {
+        let to = scheme::schemes_mut().make_ns(from, &names)?;
+        Ok(to.into())
+    } else {
+        Err(Error::new(EACCES))
+    }
+}
+
+pub fn setregid(rgid: u32, egid: u32) -> Result<usize> {
+    let contexts = context::contexts();
+    let context_lock = contexts.current().ok_or(Error::new(ESRCH))?;
+    let mut context = context_lock.write();
+
+    if (context.euid == 0
+    || rgid as i32 == -1
+    || rgid == context.egid
+    || rgid == context.rgid)
+    && (context.euid == 0
+    || egid as i32 == -1
+    || egid == context.egid
+    || egid == context.rgid)
+    {
+        if rgid as i32 != -1 {
+            context.rgid = rgid;
+        }
+        if egid as i32 != -1 {
+            context.egid = egid;
+        }
+        Ok(0)
+    } else {
+        Err(Error::new(EPERM))
+    }
+}
+
+pub fn setrens(rns: SchemeNamespace, ens: SchemeNamespace) -> Result<usize> {
+    let contexts = context::contexts();
+    let context_lock = contexts.current().ok_or(Error::new(ESRCH))?;
+    let mut context = context_lock.write();
+
+    if (context.euid == 0
+    || rns.into() as isize == -1
+    || rns == context.ens
+    || rns == context.rns)
+    && (context.euid == 0
+    || ens.into() as isize == -1
+    || ens == context.ens
+    || ens == context.rns)
+    {
+        if rns.into() as isize != -1 {
+            context.rns = rns;
+        }
+        if ens.into() as isize != -1 {
+            context.ens = ens;
+        }
+        Ok(0)
+    } else {
+        Err(Error::new(EPERM))
+    }
+}
+
+pub fn setreuid(ruid: u32, euid: u32) -> Result<usize> {
+    let contexts = context::contexts();
+    let context_lock = contexts.current().ok_or(Error::new(ESRCH))?;
+    let mut context = context_lock.write();
+
+    if (context.euid == 0
+    || ruid as i32 == -1
+    || ruid == context.euid
+    || ruid == context.ruid)
+    && (context.euid == 0
+    || euid as i32 == -1
+    || euid == context.euid
+    || euid == context.ruid)
+    {
+        if ruid as i32 != -1 {
+            context.ruid = ruid;
+        }
+        if euid as i32 != -1 {
+            context.euid = euid;
+        }
+        Ok(0)
+    } else {
+        Err(Error::new(EPERM))
+    }
+}

kernel/src/syscall/process.rs

@@ -0,0 +1,965 @@
+///! Process syscalls
+use alloc::arc::Arc;
+use alloc::boxed::Box;
+use collections::{BTreeMap, Vec};
+use core::{intrinsics, mem, str};
+use core::ops::DerefMut;
+use spin::Mutex;
+
+use arch;
+use arch::memory::allocate_frame;
+use arch::paging::{ActivePageTable, InactivePageTable, Page, VirtualAddress, entry};
+use arch::paging::temporary_page::TemporaryPage;
+use arch::start::usermode;
+use context;
+use context::ContextId;
+use elf::{self, program_header};
+use scheme::{self, FileHandle};
+use syscall;
+use syscall::data::Stat;
+use syscall::error::*;
+use syscall::flag::{CLONE_VFORK, CLONE_VM, CLONE_FS, CLONE_FILES, WNOHANG};
+use syscall::validate::{validate_slice, validate_slice_mut};
+
+pub fn brk(address: usize) -> Result<usize> {
+    let contexts = context::contexts();
+    let context_lock = contexts.current().ok_or(Error::new(ESRCH))?;
+    let context = context_lock.read();
+
+    //println!("{}: {}: BRK {:X}", unsafe { ::core::str::from_utf8_unchecked(&context.name.lock()) },
+    //                             context.id.into(), address);
+
+    let current = if let Some(ref heap_shared) = context.heap {
+        heap_shared.with(|heap| {
+            heap.start_address().get() + heap.size()
+        })
+    } else {
+        panic!("user heap not initialized");
+    };
+
+    if address == 0 {
+        //println!("Brk query {:X}", current);
+        Ok(current)
+    } else if address >= arch::USER_HEAP_OFFSET {
+        //TODO: out of memory errors
+        if let Some(ref heap_shared) = context.heap {
+            heap_shared.with(|heap| {
+                heap.resize(address - arch::USER_HEAP_OFFSET, true);
+            });
+        } else {
+            panic!("user heap not initialized");
+        }
+
+        //println!("Brk resize {:X}", address);
+        Ok(address)
+    } else {
+        //println!("Brk no mem");
+        Err(Error::new(ENOMEM))
+    }
+}
+
+pub fn clone(flags: usize, stack_base: usize) -> Result<ContextId> {
+    let ppid;
+    let pid;
+    {
+        let ruid;
+        let rgid;
+        let rns;
+        let euid;
+        let egid;
+        let ens;
+        let mut cpu_id = None;
+        let arch;
+        let vfork;
+        let mut kfx_option = None;
+        let mut kstack_option = None;
+        let mut offset = 0;
+        let mut image = vec![];
+        let mut heap_option = None;
+        let mut stack_option = None;
+        let mut tls_option = None;
+        let grants;
+        let name;
+        let cwd;
+        let env;
+        let files;
+
+        // Copy from old process
+        {
+            let contexts = context::contexts();
+            let context_lock = contexts.current().ok_or(Error::new(ESRCH))?;
+            let context = context_lock.read();
+
+            ppid = context.id;
+            ruid = context.ruid;
+            rgid = context.rgid;
+            rns = context.rns;
+            euid = context.euid;
+            egid = context.egid;
+            ens = context.ens;
+
+            if flags & CLONE_VM == CLONE_VM {
+                cpu_id = context.cpu_id;
+            }
+
+            arch = context.arch.clone();
+
+            if let Some(ref fx) = context.kfx {
+                let mut new_fx = unsafe { Box::from_raw(::alloc::heap::allocate(512, 16) as *mut [u8; 512]) };
+                for (new_b, b) in new_fx.iter_mut().zip(fx.iter()) {
+                    *new_b = *b;
+                }
+                kfx_option = Some(new_fx);
+            }
+
+            if let Some(ref stack) = context.kstack {
+                offset = stack_base - stack.as_ptr() as usize - mem::size_of::<usize>(); // Add clone ret
+                let mut new_stack = stack.clone();
+
+                unsafe {
+                    let func_ptr = new_stack.as_mut_ptr().offset(offset as isize);
+                    *(func_ptr as *mut usize) = arch::interrupt::syscall::clone_ret as usize;
+                }
+
+                kstack_option = Some(new_stack);
+            }
+
+            if flags & CLONE_VM == CLONE_VM {
+                for memory_shared in context.image.iter() {
+                    image.push(memory_shared.clone());
+                }
+
+                if let Some(ref heap_shared) = context.heap {
+                    heap_option = Some(heap_shared.clone());
+                }
+            } else {
+                for memory_shared in context.image.iter() {
+                    memory_shared.with(|memory| {
+                        let mut new_memory = context::memory::Memory::new(
+                            VirtualAddress::new(memory.start_address().get() + arch::USER_TMP_OFFSET),
+                            memory.size(),
+                            entry::PRESENT | entry::NO_EXECUTE | entry::WRITABLE,
+                            false
+                        );
+
+                        unsafe {
+                            intrinsics::copy(memory.start_address().get() as *const u8,
+                                            new_memory.start_address().get() as *mut u8,
+                                            memory.size());
+                        }
+
+                        new_memory.remap(memory.flags());
+                        image.push(new_memory.to_shared());
+                    });
+                }
+
+                if let Some(ref heap_shared) = context.heap {
+                    heap_shared.with(|heap| {
+                        let mut new_heap = context::memory::Memory::new(
+                            VirtualAddress::new(arch::USER_TMP_HEAP_OFFSET),
+                            heap.size(),
+                            entry::PRESENT | entry::NO_EXECUTE | entry::WRITABLE,
+                            false
+                        );
+
+                        unsafe {
+                            intrinsics::copy(heap.start_address().get() as *const u8,
+                                            new_heap.start_address().get() as *mut u8,
+                                            heap.size());
+                        }
+
+                        new_heap.remap(heap.flags());
+                        heap_option = Some(new_heap.to_shared());
+                    });
+                }
+            }
+
+            if let Some(ref stack) = context.stack {
+                let mut new_stack = context::memory::Memory::new(
+                    VirtualAddress::new(arch::USER_TMP_STACK_OFFSET),
+                    stack.size(),
+                    entry::PRESENT | entry::NO_EXECUTE | entry::WRITABLE,
+                    false
+                );
+
+                unsafe {
+                    intrinsics::copy(stack.start_address().get() as *const u8,
+                                    new_stack.start_address().get() as *mut u8,
+                                    stack.size());
+                }
+
+                new_stack.remap(stack.flags());
+                stack_option = Some(new_stack);
+            }
+
+            if let Some(ref tls) = context.tls {
+                let mut new_tls = context::memory::Tls {
+                    master: tls.master,
+                    file_size: tls.file_size,
+                    mem: context::memory::Memory::new(
+                        VirtualAddress::new(arch::USER_TMP_TLS_OFFSET),
+                        tls.mem.size(),
+                        entry::PRESENT | entry::NO_EXECUTE | entry::WRITABLE,
+                        true
+                    )
+                };
+
+                unsafe {
+                    intrinsics::copy(tls.master.get() as *const u8,
+                                    new_tls.mem.start_address().get() as *mut u8,
+                                    tls.file_size);
+                }
+
+                new_tls.mem.remap(tls.mem.flags());
+                tls_option = Some(new_tls);
+            }
+
+            if flags & CLONE_VM == CLONE_VM {
+                grants = context.grants.clone();
+            } else {
+                grants = Arc::new(Mutex::new(Vec::new()));
+            }
+
+            if flags & CLONE_VM == CLONE_VM {
+                name = context.name.clone();
+            } else {
+                name = Arc::new(Mutex::new(context.name.lock().clone()));
+            }
+
+            if flags & CLONE_FS == CLONE_FS {
+                cwd = context.cwd.clone();
+            } else {
+                cwd = Arc::new(Mutex::new(context.cwd.lock().clone()));
+            }
+
+            if flags & CLONE_VM == CLONE_VM {
+                env = context.env.clone();
+            } else {
+                let mut new_env = BTreeMap::new();
+                for item in context.env.lock().iter() {
+                    new_env.insert(item.0.clone(), Arc::new(Mutex::new(item.1.lock().clone())));
+                }
+                env = Arc::new(Mutex::new(new_env));
+            }
+
+            if flags & CLONE_FILES == CLONE_FILES {
+                files = context.files.clone();
+            } else {
+                files = Arc::new(Mutex::new(context.files.lock().clone()));
+            }
+        }
+
+        // If not cloning files, dup to get a new number from scheme
+        // This has to be done outside the context lock to prevent deadlocks
+        if flags & CLONE_FILES == 0 {
+            for (_fd, mut file_option) in files.lock().iter_mut().enumerate() {
+                let new_file_option = if let Some(file) = *file_option {
+                    let result = {
+                        let scheme = {
+                            let schemes = scheme::schemes();
+                            let scheme = schemes.get(file.scheme).ok_or(Error::new(EBADF))?;
+                            scheme.clone()
+                        };
+                        let result = scheme.dup(file.number, b"clone");
+                        result
+                    };
+                    match result {
+                        Ok(new_number) => {
+                            Some(context::file::File {
+                                scheme: file.scheme,
+                                number: new_number,
+                                event: None,
+                            })
+                        },
+                        Err(_err) => {
+                            None
+                        }
+                    }
+                } else {
+                    None
+                };
+
+                *file_option = new_file_option;
+            }
+        }
+
+        // If vfork, block the current process
+        // This has to be done after the operations that may require context switches
+        if flags & CLONE_VFORK == CLONE_VFORK {
+            let contexts = context::contexts();
+            let context_lock = contexts.current().ok_or(Error::new(ESRCH))?;
+            let mut context = context_lock.write();
+            context.block();
+            vfork = true;
+        } else {
+            vfork = false;
+        }
+
+        // Set up new process
+        {
+            let mut contexts = context::contexts_mut();
+            let context_lock = contexts.new_context()?;
+            let mut context = context_lock.write();
+
+            pid = context.id;
+
+            context.ppid = ppid;
+            context.ruid = ruid;
+            context.rgid = rgid;
+            context.rns = rns;
+            context.euid = euid;
+            context.egid = egid;
+            context.ens = ens;
+
+            context.cpu_id = cpu_id;
+
+            context.status = context::Status::Runnable;
+
+            context.vfork = vfork;
+
+            context.arch = arch;
+
+            let mut active_table = unsafe { ActivePageTable::new() };
+
+            let mut temporary_page = TemporaryPage::new(Page::containing_address(VirtualAddress::new(arch::USER_TMP_MISC_OFFSET)));
+
+            let mut new_table = {
+                let frame = allocate_frame().expect("no more frames in syscall::clone new_table");
+                InactivePageTable::new(frame, &mut active_table, &mut temporary_page)
+            };
+
+            context.arch.set_page_table(unsafe { new_table.address() });
+
+            // Copy kernel mapping
+            {
+                let frame = active_table.p4()[510].pointed_frame().expect("kernel table not mapped");
+                let flags = active_table.p4()[510].flags();
+                active_table.with(&mut new_table, &mut temporary_page, |mapper| {
+                    mapper.p4_mut()[510].set(frame, flags);
+                });
+            }
+
+            if let Some(fx) = kfx_option.take() {
+                context.arch.set_fx(fx.as_ptr() as usize);
+                context.kfx = Some(fx);
+            }
+
+            // Set kernel stack
+            if let Some(stack) = kstack_option.take() {
+                context.arch.set_stack(stack.as_ptr() as usize + offset);
+                context.kstack = Some(stack);
+            }
+
+            // Setup heap
+            if flags & CLONE_VM == CLONE_VM {
+                // Copy user image mapping, if found
+                if ! image.is_empty() {
+                    let frame = active_table.p4()[0].pointed_frame().expect("user image not mapped");
+                    let flags = active_table.p4()[0].flags();
+                    active_table.with(&mut new_table, &mut temporary_page, |mapper| {
+                        mapper.p4_mut()[0].set(frame, flags);
+                    });
+                }
+                context.image = image;
+
+                // Copy user heap mapping, if found
+                if let Some(heap_shared) = heap_option {
+                    let frame = active_table.p4()[1].pointed_frame().expect("user heap not mapped");
+                    let flags = active_table.p4()[1].flags();
+                    active_table.with(&mut new_table, &mut temporary_page, |mapper| {
+                        mapper.p4_mut()[1].set(frame, flags);
+                    });
+                    context.heap = Some(heap_shared);
+                }
+
+                // Copy grant mapping
+                if ! grants.lock().is_empty() {
+                    let frame = active_table.p4()[2].pointed_frame().expect("user grants not mapped");
+                    let flags = active_table.p4()[2].flags();
+                    active_table.with(&mut new_table, &mut temporary_page, |mapper| {
+                        mapper.p4_mut()[2].set(frame, flags);
+                    });
+                }
+                context.grants = grants;
+            } else {
+                // Copy percpu mapping
+                for cpu_id in 0..::cpu_count() {
+                    extern {
+                        /// The starting byte of the thread data segment
+                        static mut __tdata_start: u8;
+                        /// The ending byte of the thread BSS segment
+                        static mut __tbss_end: u8;
+                    }
+
+                    let size = unsafe { & __tbss_end as *const _ as usize - & __tdata_start as *const _ as usize };
+
+                    let start = arch::KERNEL_PERCPU_OFFSET + arch::KERNEL_PERCPU_SIZE * cpu_id;
+                    let end = start + size;
+
+                    let start_page = Page::containing_address(VirtualAddress::new(start));
+                    let end_page = Page::containing_address(VirtualAddress::new(end - 1));
+                    for page in Page::range_inclusive(start_page, end_page) {
+                        let frame = active_table.translate_page(page).expect("kernel percpu not mapped");
+                        active_table.with(&mut new_table, &mut temporary_page, |mapper| {
+                            let result = mapper.map_to(page, frame, entry::PRESENT | entry::NO_EXECUTE | entry::WRITABLE);
+                            // Ignore result due to operating on inactive table
+                            unsafe { result.ignore(); }
+                        });
+                    }
+                }
+
+                // Move copy of image
+                for memory_shared in image.iter_mut() {
+                    memory_shared.with(|memory| {
+                        let start = VirtualAddress::new(memory.start_address().get() - arch::USER_TMP_OFFSET + arch::USER_OFFSET);
+                        memory.move_to(start, &mut new_table, &mut temporary_page);
+                    });
+                }
+                context.image = image;
+
+                // Move copy of heap
+                if let Some(heap_shared) = heap_option {
+                    heap_shared.with(|heap| {
+                        heap.move_to(VirtualAddress::new(arch::USER_HEAP_OFFSET), &mut new_table, &mut temporary_page);
+                    });
+                    context.heap = Some(heap_shared);
+                }
+            }
+
+            // Setup user stack
+            if let Some(mut stack) = stack_option {
+                stack.move_to(VirtualAddress::new(arch::USER_STACK_OFFSET), &mut new_table, &mut temporary_page);
+                context.stack = Some(stack);
+            }
+
+            // Setup user TLS
+            if let Some(mut tls) = tls_option {
+                tls.mem.move_to(VirtualAddress::new(arch::USER_TLS_OFFSET), &mut new_table, &mut temporary_page);
+                context.tls = Some(tls);
+            }
+
+            context.name = name;
+
+            context.cwd = cwd;
+
+            context.env = env;
+
+            context.files = files;
+        }
+    }
+
+    unsafe { context::switch(); }
+
+    Ok(pid)
+}
+
+fn empty(context: &mut context::Context, reaping: bool) {
+    if reaping {
+        // Memory should already be unmapped
+        assert!(context.image.is_empty());
+        assert!(context.heap.is_none());
+        assert!(context.stack.is_none());
+        assert!(context.tls.is_none());
+    } else {
+        // Unmap previous image, heap, grants, stack, and tls
+        context.image.clear();
+        drop(context.heap.take());
+        drop(context.stack.take());
+        drop(context.tls.take());
+    }
+
+    // FIXME: Looks like a race condition.
+    // Is it possible for Arc::strong_count to return 1 to two contexts that exit at the
+    // same time, or return 2 to both, thus either double freeing or leaking the grants?
+    if Arc::strong_count(&context.grants) == 1 {
+        let mut grants = context.grants.lock();
+        for grant in grants.drain(..) {
+            if reaping {
+                println!("{}: {}: Grant should not exist: {:?}", context.id.into(), unsafe { ::core::str::from_utf8_unchecked(&context.name.lock()) }, grant);
+
+                let mut new_table = unsafe { InactivePageTable::from_address(context.arch.get_page_table()) };
+                let mut temporary_page = TemporaryPage::new(Page::containing_address(VirtualAddress::new(arch::USER_TMP_GRANT_OFFSET)));
+
+                grant.unmap_inactive(&mut new_table, &mut temporary_page);
+            } else {
+                grant.unmap();
+            }
+        }
+    }
+}
+
+pub fn exec(path: &[u8], arg_ptrs: &[[usize; 2]]) -> Result<usize> {
+    let entry;
+    let mut sp = arch::USER_STACK_OFFSET + arch::USER_STACK_SIZE - 256;
+
+    {
+        let mut args = Vec::new();
+        for arg_ptr in arg_ptrs {
+            let arg = validate_slice(arg_ptr[0] as *const u8, arg_ptr[1])?;
+            args.push(arg.to_vec()); // Must be moved into kernel space before exec unmaps all memory
+        }
+
+        let (uid, gid, canonical) = {
+            let contexts = context::contexts();
+            let context_lock = contexts.current().ok_or(Error::new(ESRCH))?;
+            let context = context_lock.read();
+            (context.euid, context.egid, context.canonicalize(path))
+        };
+
+        let file = syscall::open(&canonical, syscall::flag::O_RDONLY)?;
+        let mut stat = Stat::default();
+        syscall::file_op_mut_slice(syscall::number::SYS_FSTAT, file, &mut stat)?;
+
+        let mut perm = stat.st_mode & 0o7;
+        if stat.st_uid == uid {
+            perm |= (stat.st_mode >> 6) & 0o7;
+        }
+        if stat.st_gid == gid {
+            perm |= (stat.st_mode >> 3) & 0o7;
+        }
+        if uid == 0 {
+            perm |= 0o7;
+        }
+
+        if perm & 0o1 != 0o1 {
+            let _ = syscall::close(file);
+            return Err(Error::new(EACCES));
+        }
+
+        //TODO: Only read elf header, not entire file. Then read required segments
+        let mut data = vec![0; stat.st_size as usize];
+        syscall::file_op_mut_slice(syscall::number::SYS_READ, file, &mut data)?;
+        let _ = syscall::close(file);
+
+        match elf::Elf::from(&data) {
+            Ok(elf) => {
+                entry = elf.entry();
+
+                drop(path); // Drop so that usage is not allowed after unmapping context
+                drop(arg_ptrs); // Drop so that usage is not allowed after unmapping context
+
+                let contexts = context::contexts();
+                let (vfork, ppid, files) = {
+                    let context_lock = contexts.current().ok_or(Error::new(ESRCH))?;
+                    let mut context = context_lock.write();
+
+                    // Set name
+                    context.name = Arc::new(Mutex::new(canonical));
+
+                    empty(&mut context, false);
+
+                    if stat.st_mode & syscall::flag::MODE_SETUID == syscall::flag::MODE_SETUID {
+                        context.euid = stat.st_uid;
+                    }
+
+                    if stat.st_mode & syscall::flag::MODE_SETGID == syscall::flag::MODE_SETGID {
+                        context.egid = stat.st_gid;
+                    }
+
+                    // Map and copy new segments
+                    let mut tls_option = None;
+                    for segment in elf.segments() {
+                        if segment.p_type == program_header::PT_LOAD {
+                            let mut memory = context::memory::Memory::new(
+                                VirtualAddress::new(segment.p_vaddr as usize),
+                                segment.p_memsz as usize,
+                                entry::NO_EXECUTE | entry::WRITABLE,
+                                true
+                            );
+
+                            unsafe {
+                                // Copy file data
+                                intrinsics::copy((elf.data.as_ptr() as usize + segment.p_offset as usize) as *const u8,
+                                                segment.p_vaddr as *mut u8,
+                                                segment.p_filesz as usize);
+                            }
+
+                            let mut flags = entry::NO_EXECUTE | entry::USER_ACCESSIBLE;
+
+                            if segment.p_flags & program_header::PF_R == program_header::PF_R {
+                                flags.insert(entry::PRESENT);
+                            }
+
+                            // W ^ X. If it is executable, do not allow it to be writable, even if requested
+                            if segment.p_flags & program_header::PF_X == program_header::PF_X {
+                                flags.remove(entry::NO_EXECUTE);
+                            } else if segment.p_flags & program_header::PF_W == program_header::PF_W {
+                                flags.insert(entry::WRITABLE);
+                            }
+
+                            memory.remap(flags);
+
+                            context.image.push(memory.to_shared());
+                        } else if segment.p_type == program_header::PT_TLS {
+                            let memory = context::memory::Memory::new(
+                                VirtualAddress::new(arch::USER_TCB_OFFSET),
+                                4096,
+                                entry::NO_EXECUTE | entry::WRITABLE | entry::USER_ACCESSIBLE,
+                                true
+                            );
+
+                            unsafe { *(arch::USER_TCB_OFFSET as *mut usize) = arch::USER_TLS_OFFSET + segment.p_memsz as usize; }
+
+                            context.image.push(memory.to_shared());
+
+                            tls_option = Some((
+                                VirtualAddress::new(segment.p_vaddr as usize),
+                                segment.p_filesz as usize,
+                                segment.p_memsz as usize
+                            ));
+                        }
+                    }
+
+                    // Map heap
+                    context.heap = Some(context::memory::Memory::new(
+                        VirtualAddress::new(arch::USER_HEAP_OFFSET),
+                        0,
+                        entry::NO_EXECUTE | entry::WRITABLE | entry::USER_ACCESSIBLE,
+                        true
+                    ).to_shared());
+
+                    // Map stack
+                    context.stack = Some(context::memory::Memory::new(
+                        VirtualAddress::new(arch::USER_STACK_OFFSET),
+                        arch::USER_STACK_SIZE,
+                        entry::NO_EXECUTE | entry::WRITABLE | entry::USER_ACCESSIBLE,
+                        true
+                    ));
+
+                    // Map TLS
+                    if let Some((master, file_size, size)) = tls_option {
+                        let tls = context::memory::Tls {
+                            master: master,
+                            file_size: file_size,
+                            mem: context::memory::Memory::new(
+                                VirtualAddress::new(arch::USER_TLS_OFFSET),
+                                size,
+                                entry::NO_EXECUTE | entry::WRITABLE | entry::USER_ACCESSIBLE,
+                                true
+                            )
+                        };
+
+                        unsafe {
+                            // Copy file data
+                            intrinsics::copy(master.get() as *const u8,
+                                    tls.mem.start_address().get() as *mut u8,
+                                    file_size);
+                        }
+
+                        context.tls = Some(tls);
+                    }
+
+                    // Push arguments
+                    let mut arg_size = 0;
+                    for arg in args.iter().rev() {
+                        sp -= mem::size_of::<usize>();
+                        unsafe { *(sp as *mut usize) = arch::USER_ARG_OFFSET + arg_size; }
+                        sp -= mem::size_of::<usize>();
+                        unsafe { *(sp as *mut usize) = arg.len(); }
+
+                        arg_size += arg.len();
+                    }
+
+                    sp -= mem::size_of::<usize>();
+                    unsafe { *(sp as *mut usize) = args.len(); }
+
+                    if arg_size > 0 {
+                        let mut memory = context::memory::Memory::new(
+                            VirtualAddress::new(arch::USER_ARG_OFFSET),
+                            arg_size,
+                            entry::NO_EXECUTE | entry::WRITABLE,
+                            true
+                        );
+
+                        let mut arg_offset = 0;
+                        for arg in args.iter().rev() {
+                            unsafe {
+                                intrinsics::copy(arg.as_ptr(),
+                                       (arch::USER_ARG_OFFSET + arg_offset) as *mut u8,
+                                       arg.len());
+                            }
+
+                            arg_offset += arg.len();
+                        }
+
+                        memory.remap(entry::NO_EXECUTE | entry::USER_ACCESSIBLE);
+
+                        context.image.push(memory.to_shared());
+                    }
+
+                    let files = Arc::new(Mutex::new(context.files.lock().clone()));
+                    context.files = files.clone();
+
+                    let vfork = context.vfork;
+                    context.vfork = false;
+                    (vfork, context.ppid, files)
+                };
+
+                // Duplicate current files using b"exec", close previous
+                for (fd, mut file_option) in files.lock().iter_mut().enumerate() {
+                    let new_file_option = if let Some(file) = *file_option {
+                        // Duplicate
+                        let result = {
+                            let scheme_option = {
+                                let schemes = scheme::schemes();
+                                schemes.get(file.scheme).map(|scheme| scheme.clone())
+                            };
+                            if let Some(scheme) = scheme_option {
+                                let result = scheme.dup(file.number, b"exec");
+                                result
+                            } else {
+                                Err(Error::new(EBADF))
+                            }
+                        };
+
+                        // Close
+                        {
+                            if let Some(event_id) = file.event {
+                                context::event::unregister(FileHandle::from(fd), file.scheme, event_id);
+                            }
+
+                            let scheme_option = {
+                                let schemes = scheme::schemes();
+                                schemes.get(file.scheme).map(|scheme| scheme.clone())
+                            };
+                            if let Some(scheme) = scheme_option {
+                                let _ = scheme.close(file.number);
+                            }
+                        }
+
+                        // Return new descriptor
+                        match result {
+                            Ok(new_number) => {
+                                Some(context::file::File {
+                                    scheme: file.scheme,
+                                    number: new_number,
+                                    event: None,
+                                })
+                            },
+                            Err(_err) => {
+                                None
+                            }
+                        }
+                    } else {
+                        None
+                    };
+
+                    *file_option = new_file_option;
+                }
+
+                if vfork {
+                    if let Some(context_lock) = contexts.get(ppid) {
+                        let mut context = context_lock.write();
+                        if ! context.unblock() {
+                            println!("{:?} not blocked for exec vfork unblock", ppid);
+                        }
+                    } else {
+                        println!("{:?} not found for exec vfork unblock", ppid);
+                    }
+                }
+            },
+            Err(err) => {
+                println!("failed to execute {}: {}", unsafe { str::from_utf8_unchecked(path) }, err);
+                return Err(Error::new(ENOEXEC));
+            }
+        }
+    }
+
+    // Go to usermode
+    unsafe { usermode(entry, sp); }
+}
+
+pub fn exit(status: usize) -> ! {
+    {
+        let context_lock = {
+            let contexts = context::contexts();
+            let context_lock = contexts.current().ok_or(Error::new(ESRCH)).expect("exit failed to find context");
+            context_lock.clone()
+        };
+
+        let mut close_files = Vec::new();
+        let (pid, ppid) = {
+            let mut context = context_lock.write();
+            // FIXME: Looks like a race condition.
+            // Is it possible for Arc::strong_count to return 1 to two contexts that exit at the
+            // same time, or return 2 to both, thus either double closing or leaking the files?
+            if Arc::strong_count(&context.files) == 1 {
+                mem::swap(context.files.lock().deref_mut(), &mut close_files);
+            }
+            context.files = Arc::new(Mutex::new(Vec::new()));
+            (context.id, context.ppid)
+        };
+
+        /// Files must be closed while context is valid so that messages can be passed
+        for (fd, file_option) in close_files.drain(..).enumerate() {
+            if let Some(file) = file_option {
+                if let Some(event_id) = file.event {
+                    context::event::unregister(FileHandle::from(fd), file.scheme, event_id);
+                }
+
+                let scheme_option = {
+                    let schemes = scheme::schemes();
+                    schemes.get(file.scheme).map(|scheme| scheme.clone())
+                };
+                if let Some(scheme) = scheme_option {
+                    let _ = scheme.close(file.number);
+                }
+            }
+        }
+
+        /// Transfer child processes to parent
+        {
+            let contexts = context::contexts();
+            for (_id, context_lock) in contexts.iter() {
+                let mut context = context_lock.write();
+                if context.ppid == pid {
+                    context.ppid = ppid;
+                    context.vfork = false;
+                }
+            }
+        }
+
+        let (vfork, children) = {
+            let mut context = context_lock.write();
+
+            empty(&mut context, false);
+
+            let vfork = context.vfork;
+            context.vfork = false;
+
+            context.status = context::Status::Exited(status);
+
+            let children = context.waitpid.receive_all();
+
+            (vfork, children)
+        };
+
+        {
+            let contexts = context::contexts();
+            if let Some(parent_lock) = contexts.get(ppid) {
+                let waitpid = {
+                    let mut parent = parent_lock.write();
+                    if vfork {
+                        if ! parent.unblock() {
+                            println!("{:?} not blocked for exit vfork unblock", ppid);
+                        }
+                    }
+                    parent.waitpid.clone()
+                };
+
+                for (c_pid, c_status) in children {
+                    waitpid.send(c_pid, c_status);
+                }
+                waitpid.send(pid, status);
+            } else {
+                println!("{:?} not found for exit vfork unblock", ppid);
+            }
+        }
+    }
+
+    unsafe { context::switch(); }
+
+    unreachable!();
+}
+
+pub fn getpid() -> Result<ContextId> {
+    let contexts = context::contexts();
+    let context_lock = contexts.current().ok_or(Error::new(ESRCH))?;
+    let context = context_lock.read();
+    Ok(context.id)
+}
+
+pub fn kill(pid: ContextId, sig: usize) -> Result<usize> {
+    let (ruid, euid) = {
+        let contexts = context::contexts();
+        let context_lock = contexts.current().ok_or(Error::new(ESRCH))?;
+        let context = context_lock.read();
+        (context.ruid, context.euid)
+    };
+
+    if sig > 0 && sig <= 0x7F {
+        let contexts = context::contexts();
+        let context_lock = contexts.get(pid).ok_or(Error::new(ESRCH))?;
+        let mut context = context_lock.write();
+        if euid == 0
+        || euid == context.ruid
+        || ruid == context.ruid
+        {
+            context.pending.push_back(sig as u8);
+            Ok(0)
+        } else {
+            Err(Error::new(EPERM))
+        }
+    } else {
+        Err(Error::new(EINVAL))
+    }
+}
+
+fn reap(pid: ContextId) -> Result<ContextId> {
+    // Spin until not running
+    let mut running = false;
+    while running {
+        {
+            let contexts = context::contexts();
+            let context_lock = contexts.get(pid).ok_or(Error::new(ESRCH))?;
+            let context = context_lock.read();
+            running = context.running;
+        }
+
+        arch::interrupt::pause();
+    }
+
+    let mut contexts = context::contexts_mut();
+    let context_lock = contexts.remove(pid).ok_or(Error::new(ESRCH))?;
+    {
+        let mut context = context_lock.write();
+        empty(&mut context, true);
+    }
+    drop(context_lock);
+
+    Ok(pid)
+}
+
+pub fn waitpid(pid: ContextId, status_ptr: usize, flags: usize) -> Result<ContextId> {
+    let waitpid = {
+        let contexts = context::contexts();
+        let context_lock = contexts.current().ok_or(Error::new(ESRCH))?;
+        let context = context_lock.read();
+        context.waitpid.clone()
+    };
+
+    let mut tmp = [0];
+    let status_slice = if status_ptr != 0 {
+        validate_slice_mut(status_ptr as *mut usize, 1)?
+    } else {
+        &mut tmp
+    };
+
+    if pid.into() == 0 {
+        if flags & WNOHANG == WNOHANG {
+            if let Some((w_pid, status)) = waitpid.receive_any_nonblock() {
+                status_slice[0] = status;
+                reap(w_pid)
+            } else {
+                Ok(ContextId::from(0))
+            }
+        } else {
+            let (w_pid, status) = waitpid.receive_any();
+            status_slice[0] = status;
+            reap(w_pid)
+        }
+    } else {
+        if flags & WNOHANG == WNOHANG {
+            if let Some(status) = waitpid.receive_nonblock(&pid) {
+                status_slice[0] = status;
+                reap(pid)
+            } else {
+                Ok(ContextId::from(0))
+            }
+        } else {
+            let status = waitpid.receive(&pid);
+            status_slice[0] = status;
+            reap(pid)
+        }
+    }
+}

kernel/src/syscall/time.rs

@@ -0,0 +1,53 @@
+use arch;
+use context;
+use syscall::data::TimeSpec;
+use syscall::error::*;
+use syscall::flag::{CLOCK_REALTIME, CLOCK_MONOTONIC};
+
+pub fn clock_gettime(clock: usize, time: &mut TimeSpec) -> Result<usize> {
+    match clock {
+        CLOCK_REALTIME => {
+            let arch_time = arch::time::realtime();
+            time.tv_sec = arch_time.0 as i64;
+            time.tv_nsec = arch_time.1 as i32;
+            Ok(0)
+        },
+        CLOCK_MONOTONIC => {
+            let arch_time = arch::time::monotonic();
+            time.tv_sec = arch_time.0 as i64;
+            time.tv_nsec = arch_time.1 as i32;
+            Ok(0)
+        },
+        _ => Err(Error::new(EINVAL))
+    }
+}
+
+pub fn nanosleep(req: &TimeSpec, rem_opt: Option<&mut TimeSpec>) -> Result<usize> {
+    let start = arch::time::monotonic();
+    let sum = start.1 + req.tv_nsec as u64;
+    let end = (start.0 + req.tv_sec as u64 + sum / 1000000000, sum % 1000000000);
+
+    {
+        let contexts = context::contexts();
+        let context_lock = contexts.current().ok_or(Error::new(ESRCH))?;
+        let mut context = context_lock.write();
+
+        context.wake = Some(end);
+        context.block();
+    }
+
+    unsafe { context::switch(); }
+
+    if let Some(mut rem) = rem_opt {
+        //TODO let current = arch::time::monotonic();
+        rem.tv_sec = 0;
+        rem.tv_nsec = 0;
+    }
+
+    Ok(0)
+}
+
+pub fn sched_yield() -> Result<usize> {
+    unsafe { context::switch(); }
+    Ok(0)
+}

kernel/src/syscall/validate.rs

@@ -0,0 +1,44 @@
+use core::{mem, slice};
+
+use arch::paging::{ActivePageTable, Page, VirtualAddress, entry};
+use syscall::error::*;
+
+fn validate(address: usize, size: usize, flags: entry::EntryFlags) -> Result<()> {
+    let active_table = unsafe { ActivePageTable::new() };
+
+    let start_page = Page::containing_address(VirtualAddress::new(address));
+    let end_page = Page::containing_address(VirtualAddress::new(address + size - 1));
+    for page in Page::range_inclusive(start_page, end_page) {
+        if let Some(page_flags) = active_table.translate_page_flags(page) {
+            if ! page_flags.contains(flags) {
+                //println!("{:X}: Not {:?}", page.start_address().get(), flags);
+                return Err(Error::new(EFAULT));
+            }
+        } else {
+            //println!("{:X}: Not found", page.start_address().get());
+            return Err(Error::new(EFAULT));
+        }
+    }
+
+    Ok(())
+}
+
+/// Convert a pointer and length to slice, if valid
+pub fn validate_slice<T>(ptr: *const T, len: usize) -> Result<&'static [T]> {
+    if len == 0 {
+        Ok(&[])
+    } else {
+        validate(ptr as usize, len * mem::size_of::<T>(), entry::PRESENT /* TODO | entry::USER_ACCESSIBLE */)?;
+        Ok(unsafe { slice::from_raw_parts(ptr, len) })
+    }
+}
+
+/// Convert a pointer and length to slice, if valid
+pub fn validate_slice_mut<T>(ptr: *mut T, len: usize) -> Result<&'static mut [T]> {
+    if len == 0 {
+        Ok(&mut [])
+    } else {
+        validate(ptr as usize, len * mem::size_of::<T>(), entry::PRESENT | entry::WRITABLE /* TODO | entry::USER_ACCESSIBLE */)?;
+        Ok(unsafe { slice::from_raw_parts_mut(ptr, len) })
+    }
+}