Merge branch 'feat/thread' into feat/smp

Cargo.lock

@@ -35,6 +35,17 @@ version = "0.1.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "b6419a5c75e40011b9fe0174db3fe24006ab122fbe1b7e9cc5974b338a755c76"
 
+[[package]]
+name = "enum-repr"
+version = "0.2.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "bad30c9c0fa1aaf1ae5010dab11f1117b15d35faf62cda4bbbc53b9987950f18"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn",
+]
+
 [[package]]
 name = "fallible-iterator"
 version = "0.2.0"
@@ -58,7 +69,7 @@ checksum = "99a40cabc11c8258822a593f5c51f2d9f4923e715ca9e2a0630cf77ae15f390b"
 dependencies = [
  "endian-type-rs",
  "fallible-iterator",
- "memoffset",
+ "memoffset 0.5.6",
  "num-derive",
  "num-traits",
  "rustc_version",
@@ -111,6 +122,7 @@ name = "libsys"
 version = "0.1.0"
 dependencies = [
  "bitflags",
+ "enum-repr",
 ]
 
 [[package]]
@@ -119,6 +131,7 @@ version = "0.1.0"
 dependencies = [
  "lazy_static",
  "libsys",
+ "memoffset 0.6.4",
 ]
 
 [[package]]
@@ -139,6 +152,15 @@ dependencies = [
  "autocfg",
 ]
 
+[[package]]
+name = "memoffset"
+version = "0.6.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "59accc507f1338036a0477ef61afdae33cde60840f4dfe481319ce3ad116ddf9"
+dependencies = [
+ "autocfg",
+]
+
 [[package]]
 name = "num-derive"
 version = "0.3.3"
@@ -250,6 +272,8 @@ checksum = "1230ec65f13e0f9b28d789da20d2d419511893ea9dac2c1f4ef67b8b14e5da80"
 name = "user"
 version = "0.1.0"
 dependencies = [
+ "lazy_static",
+ "libsys",
  "libusr",
 ]
 

Makefile

@@ -34,6 +34,7 @@ else
 ifeq ($(MACH),qemu)
 QEMU_OPTS+=-kernel $(O)/kernel.bin \
 		   -initrd $(O)/initrd.img \
+		   -smp cpus=4 \
 		   -M virt,virtualization=on \
 		   -cpu cortex-a72 \
 		   -m 512 \
@@ -68,7 +69,7 @@ endif
 all: kernel initrd
 
 kernel:
-	cd kernel && cargo build $(CARGO_BUILD_OPTS)
+	cd kernel && ARCH=$(ARCH) MACH=$(MACH) cargo build $(CARGO_BUILD_OPTS)
 ifeq ($(ARCH),aarch64)
 	$(LLVM_BASE)/llvm-strip -o $(O)/kernel.strip $(O)/kernel
 	$(LLVM_BASE)/llvm-size $(O)/kernel.strip
@@ -92,9 +93,15 @@ initrd:
 		--target=../etc/$(ARCH)-osdev5.json \
 		-Z build-std=core,alloc,compiler_builtins \
 		$(CARGO_COMMON_OPTS)
-	mkdir -p $(O)/rootfs/bin
+	mkdir -p $(O)/rootfs/bin $(O)/rootfs/sbin $(O)/rootfs/dev
+	touch $(O)/rootfs/dev/.do_no_remove
 	cp target/$(ARCH)-osdev5/$(PROFILE)/init $(O)/rootfs/init
 	cp target/$(ARCH)-osdev5/$(PROFILE)/shell $(O)/rootfs/bin
+	cp target/$(ARCH)-osdev5/$(PROFILE)/fuzzy $(O)/rootfs/bin
+	cp target/$(ARCH)-osdev5/$(PROFILE)/ls $(O)/rootfs/bin
+	cp target/$(ARCH)-osdev5/$(PROFILE)/cat $(O)/rootfs/bin
+	cp target/$(ARCH)-osdev5/$(PROFILE)/hexd $(O)/rootfs/bin
+	cp target/$(ARCH)-osdev5/$(PROFILE)/login $(O)/rootfs/sbin
 	cd $(O)/rootfs && tar cf ../initrd.img `find -type f -printf "%P\n"`
 ifeq ($(MACH),orangepi3)
 	$(MKIMAGE) \

fs/macros/src/lib.rs

@@ -31,8 +31,8 @@ fn impl_inode_fn<T: ToTokens>(name: &str, behavior: T) -> ImplItem {
             }
         },
         "stat" => quote! {
-            fn stat(&mut self, _at: VnodeRef, _stat: &mut libsys::stat::Stat) ->
-                Result<(), libsys::error::Errno>
+            fn stat(&mut self, _at: VnodeRef) ->
+                Result<libsys::stat::Stat, libsys::error::Errno>
             {
                 #behavior
             }
@@ -94,6 +94,18 @@ fn impl_inode_fn<T: ToTokens>(name: &str, behavior: T) -> ImplItem {
                 #behavior
             }
         },
+        "readdir" => quote! {
+            fn readdir(
+                &mut self,
+                _node: VnodeRef,
+                _pos: usize,
+                _entries: &mut [libsys::stat::DirectoryEntry]
+            ) ->
+                Result<usize, libsys::error::Errno>
+            {
+                #behavior
+            }
+        },
         _ => panic!("TODO implement {:?}", name),
     })
 }
@@ -126,6 +138,7 @@ pub fn auto_inode(attr: TokenStream, input: TokenStream) -> TokenStream {
     missing.insert("size".to_string());
     missing.insert("ioctl".to_string());
     missing.insert("is_ready".to_string());
+    missing.insert("readdir".to_string());
 
     for item in &impl_item.items {
         match item {

fs/memfs/src/dir.rs

@@ -1,6 +1,6 @@
 use crate::{BlockAllocator, Bvec, FileInode};
 use alloc::boxed::Box;
-use libsys::error::Errno;
+use libsys::{error::Errno, stat::Stat};
 use vfs::{Vnode, VnodeImpl, VnodeKind, VnodeRef};
 
 pub struct DirInode<A: BlockAllocator + Copy + 'static> {
@@ -15,7 +15,7 @@ impl<A: BlockAllocator + Copy + 'static> VnodeImpl for DirInode<A> {
         name: &str,
         kind: VnodeKind,
     ) -> Result<VnodeRef, Errno> {
-        let vnode = Vnode::new(name, kind, Vnode::SEEKABLE);
+        let vnode = Vnode::new(name, kind, Vnode::SEEKABLE | Vnode::CACHE_READDIR);
         match kind {
             VnodeKind::Directory => vnode.set_data(Box::new(DirInode { alloc: self.alloc })),
             VnodeKind::Regular => vnode.set_data(Box::new(FileInode::new(Bvec::new(self.alloc)))),
@@ -31,6 +31,15 @@ impl<A: BlockAllocator + Copy + 'static> VnodeImpl for DirInode<A> {
     fn remove(&mut self, _parent: VnodeRef, _name: &str) -> Result<(), Errno> {
         Ok(())
     }
+
+    fn stat(&mut self, node: VnodeRef) -> Result<Stat, Errno> {
+        let props = node.props();
+        Ok(Stat {
+            size: 0,
+            blksize: 4096,
+            mode: props.mode,
+        })
+    }
 }
 
 impl<A: BlockAllocator + Copy + 'static> DirInode<A> {

fs/memfs/src/file.rs

@@ -35,11 +35,13 @@ impl<'a, A: BlockAllocator + Copy + 'static> VnodeImpl for FileInode<'a, A> {
         Ok(self.data.size())
     }
 
-    fn stat(&mut self, _node: VnodeRef, stat: &mut Stat) -> Result<(), Errno> {
-        stat.size = self.data.size() as u64;
-        stat.blksize = 4096;
-        stat.mode = 0o755;
-        Ok(())
+    fn stat(&mut self, node: VnodeRef) -> Result<Stat, Errno> {
+        let props = node.props();
+        Ok(Stat {
+            size: self.data.size() as u64,
+            blksize: 4096,
+            mode: props.mode
+        })
     }
 }
 

fs/memfs/src/lib.rs

@@ -29,7 +29,7 @@ pub use block::{BlockAllocator, BlockRef};
 mod bvec;
 use bvec::Bvec;
 mod tar;
-use tar::TarIterator;
+use tar::{TarIterator, Tar};
 mod file;
 use file::FileInode;
 mod dir;
@@ -67,8 +67,10 @@ impl<A: BlockAllocator + Copy + 'static> Ramfs<A> {
         Ok(res)
     }
 
-    fn create_node_initial(self: Rc<Self>, name: &str, kind: VnodeKind) -> VnodeRef {
-        let node = Vnode::new(name, kind, Vnode::SEEKABLE);
+    fn create_node_initial(self: Rc<Self>, name: &str, tar: &Tar) -> VnodeRef {
+        let kind = tar.node_kind();
+        let node = Vnode::new(name, kind, Vnode::SEEKABLE | Vnode::CACHE_READDIR);
+        node.props_mut().mode = tar.mode();
         node.set_fs(self.clone());
         match kind {
             VnodeKind::Directory => node.set_data(Box::new(DirInode::new(self.alloc))),
@@ -111,7 +113,10 @@ impl<A: BlockAllocator + Copy + 'static> Ramfs<A> {
     }
 
     unsafe fn load_tar(self: Rc<Self>, base: *const u8, size: usize) -> Result<VnodeRef, Errno> {
-        let root = self.clone().create_node_initial("", VnodeKind::Directory);
+        let root = Vnode::new("", VnodeKind::Directory, Vnode::SEEKABLE | Vnode::CACHE_READDIR);
+        root.set_fs(self.clone());
+        root.set_data(Box::new(DirInode::new(self.alloc)));
+        root.props_mut().mode = FileMode::default_dir();
 
         // 1. Create all the paths in TAR
         for block in TarIterator::new(base, base.add(size)) {
@@ -120,7 +125,7 @@ impl<A: BlockAllocator + Copy + 'static> Ramfs<A> {
             let parent = self.clone().make_path(root.clone(), dirname, true)?;
             let node = self
                 .clone()
-                .create_node_initial(basename, block.node_kind());
+                .create_node_initial(basename, block);
             assert_eq!(node.kind(), block.node_kind());
             parent.attach(node);
         }

fs/memfs/src/tar.rs

@@ -1,4 +1,4 @@
-use libsys::error::Errno;
+use libsys::{error::Errno, stat::FileMode};
 use vfs::VnodeKind;
 
 #[repr(packed)]
@@ -81,6 +81,15 @@ impl Tar {
         }
     }
 
+    pub fn mode(&self) -> FileMode {
+        let t = match self.node_kind() {
+            VnodeKind::Regular => FileMode::S_IFREG,
+            VnodeKind::Directory => FileMode::S_IFDIR,
+            _ => todo!()
+        };
+        FileMode::from_bits(from_octal(&self.mode) as u32).unwrap() | t
+    }
+
     pub fn data(&self) -> &[u8] {
         unsafe {
             core::slice::from_raw_parts(

fs/vfs/src/char.rs

@@ -19,6 +19,7 @@ pub trait CharDevice {
     /// Performs a TTY control request
     fn ioctl(&self, cmd: IoctlCmd, ptr: usize, lim: usize) -> Result<usize, Errno>;
 
+    /// Returns `true` if the device is ready for an operation
     fn is_ready(&self, write: bool) -> Result<bool, Errno>;
 }
 

fs/vfs/src/file.rs

@@ -1,9 +1,10 @@
-use crate::{VnodeKind, VnodeRef};
+use crate::{VnodeKind, VnodeRef, Vnode};
 use alloc::rc::Rc;
 use core::cell::RefCell;
 use core::cmp::min;
 use libsys::{
     error::Errno,
+    stat::DirectoryEntry,
     traits::{Read, Seek, SeekDir, Write},
 };
 
@@ -97,6 +98,11 @@ impl File {
     /// File has to be closed on execve() calls
     pub const CLOEXEC: u32 = 1 << 2;
 
+    /// Special position for cache-readdir: "." entry
+    pub const POS_CACHE_DOT: usize = usize::MAX - 1;
+    /// Special position for cache-readdir: ".." entry
+    pub const POS_CACHE_DOT_DOT: usize = usize::MAX;
+
     /// Constructs a new file handle for a regular file
     pub fn normal(vnode: VnodeRef, pos: usize, flags: u32) -> FileRef {
         Rc::new(RefCell::new(Self {
@@ -119,12 +125,68 @@ impl File {
         self.flags & Self::CLOEXEC != 0
     }
 
+    /// Returns `true` if the file is ready for an operation
     pub fn is_ready(&self, write: bool) -> Result<bool, Errno> {
         match &self.inner {
             FileInner::Normal(inner) => inner.vnode.is_ready(write),
             _ => todo!(),
         }
     }
+
+    fn cache_readdir(inner: &mut NormalFile, entries: &mut [DirectoryEntry]) -> Result<usize, Errno> {
+        let mut count = entries.len();
+        let mut offset = 0usize;
+
+        if inner.pos == Self::POS_CACHE_DOT {
+            if count == 0 {
+                return Ok(offset);
+            }
+
+            entries[offset] = DirectoryEntry::from_str(".");
+            inner.pos = Self::POS_CACHE_DOT_DOT;
+
+            offset += 1;
+            count -= 1;
+        }
+
+        if inner.pos == Self::POS_CACHE_DOT_DOT {
+            if count == 0 {
+                return Ok(offset);
+            }
+
+            entries[offset] = DirectoryEntry::from_str("..");
+            inner.pos = 0;
+
+            offset += 1;
+            count -= 1;
+        }
+
+        if count == 0 {
+            return Ok(offset);
+        }
+
+        let count = inner.vnode.for_each_entry(inner.pos, count, |i, e| {
+            entries[offset + i] = DirectoryEntry::from_str(e.name());
+        });
+        inner.pos += count;
+        Ok(offset + count)
+    }
+
+    /// Reads directory entries into the target buffer
+    pub fn readdir(&mut self, entries: &mut [DirectoryEntry]) -> Result<usize, Errno> {
+        match &mut self.inner {
+            FileInner::Normal(inner) => {
+                assert_eq!(inner.vnode.kind(), VnodeKind::Directory);
+
+                if inner.vnode.flags() & Vnode::CACHE_READDIR != 0 {
+                    Self::cache_readdir(inner, entries)
+                } else {
+                    todo!();
+                }
+            },
+            _ => todo!(),
+        }
+    }
 }
 
 impl Drop for File {

fs/vfs/src/ioctx.rs

@@ -1,8 +1,8 @@
 use crate::{FileRef, VnodeKind, VnodeRef};
 use libsys::{
     error::Errno,
-    stat::{OpenFlags, FileMode},
     path::{path_component_left, path_component_right},
+    stat::{FileMode, GroupId, OpenFlags, UserId},
 };
 
 /// I/O context structure
@@ -10,13 +10,19 @@ use libsys::{
 pub struct Ioctx {
     root: VnodeRef,
     cwd: VnodeRef,
+    /// Process user ID
+    pub uid: UserId,
+    /// Process group ID
+    pub gid: GroupId,
 }
 
 impl Ioctx {
     /// Creates a new I/O context with given root node
-    pub fn new(root: VnodeRef) -> Self {
+    pub fn new(root: VnodeRef, uid: UserId, gid: GroupId) -> Self {
         Self {
             cwd: root.clone(),
+            uid,
+            gid,
             root,
         }
     }
@@ -41,6 +47,11 @@ impl Ioctx {
             }
         }
 
+        while let Some(target) = at.target() {
+            assert!(at.kind() == VnodeKind::Directory);
+            at = target;
+        }
+
         if element.is_empty() && rest.is_empty() {
             return Ok(at);
         }
@@ -113,6 +124,16 @@ impl Ioctx {
 
         node.open(opts)
     }
+
+    /// Changes current working directory of the process
+    pub fn chdir(&mut self, path: &str) -> Result<(), Errno> {
+        let node = self.find(None, path, true)?;
+        if !node.is_directory() {
+            return Err(Errno::NotADirectory);
+        }
+        self.cwd = node;
+        Ok(())
+    }
 }
 
 #[cfg(test)]

fs/vfs/src/node.rs

@@ -1,11 +1,11 @@
-use crate::{File, FileRef, Filesystem};
+use crate::{File, FileRef, Filesystem, Ioctx};
 use alloc::{borrow::ToOwned, boxed::Box, rc::Rc, string::String, vec::Vec};
-use core::cell::{RefCell, RefMut};
+use core::cell::{RefCell, RefMut, Ref};
 use core::fmt;
 use libsys::{
     error::Errno,
     ioctl::IoctlCmd,
-    stat::{FileMode, OpenFlags, Stat},
+    stat::{AccessMode, DirectoryEntry, FileMode, OpenFlags, Stat},
 };
 
 /// Convenience type alias for [Rc<Vnode>]
@@ -31,7 +31,8 @@ pub(crate) struct TreeNode {
 
 /// File property cache struct
 pub struct VnodeProps {
-    mode: FileMode,
+    /// Node permissions and type
+    pub mode: FileMode,
 }
 
 /// Virtual filesystem node struct, generalizes access to
@@ -74,12 +75,21 @@ pub trait VnodeImpl {
     /// Resizes the file storage if necessary.
     fn write(&mut self, node: VnodeRef, pos: usize, data: &[u8]) -> Result<usize, Errno>;
 
+    /// Read directory entries into target buffer
+    fn readdir(
+        &mut self,
+        node: VnodeRef,
+        pos: usize,
+        data: &mut [DirectoryEntry],
+    ) -> Result<usize, Errno>;
+
     /// Retrieves file status
-    fn stat(&mut self, node: VnodeRef, stat: &mut Stat) -> Result<(), Errno>;
+    fn stat(&mut self, node: VnodeRef) -> Result<Stat, Errno>;
 
     /// Reports the size of this filesystem object in bytes
     fn size(&mut self, node: VnodeRef) -> Result<usize, Errno>;
 
+    /// Returns `true` if node is ready for an operation
     fn is_ready(&mut self, node: VnodeRef, write: bool) -> Result<bool, Errno>;
 
     /// Performs filetype-specific request
@@ -97,6 +107,11 @@ impl Vnode {
     /// be seeked to arbitrary offsets
     pub const SEEKABLE: u32 = 1 << 0;
 
+    /// If set, readdir() uses only in-memory node tree
+    pub const CACHE_READDIR: u32 = 1 << 1;
+    /// If set, stat() uses only in-memory stat data
+    pub const CACHE_STAT: u32 = 1 << 2;
+
     /// Constructs a new [Vnode], wrapping it in [Rc]. The resulting node
     /// then needs to have [Vnode::set_data()] called on it to be usable.
     pub fn new(name: &str, kind: VnodeKind, flags: u32) -> VnodeRef {
@@ -122,6 +137,16 @@ impl Vnode {
         &self.name
     }
 
+    /// Returns a borrowed reference to cached file properties
+    pub fn props_mut(&self) -> RefMut<VnodeProps> {
+        self.props.borrow_mut()
+    }
+
+    /// Returns a borrowed reference to cached file properties
+    pub fn props(&self) -> Ref<VnodeProps> {
+        self.props.borrow()
+    }
+
     /// Sets an associated [VnodeImpl] for the [Vnode]
     pub fn set_data(&self, data: Box<dyn VnodeImpl>) {
         *self.data.borrow_mut() = Some(data);
@@ -158,6 +183,12 @@ impl Vnode {
         self.kind
     }
 
+    /// Returns flags of the vnode
+    #[inline(always)]
+    pub const fn flags(&self) -> u32 {
+        self.flags
+    }
+
     // Tree operations
 
     /// Attaches `child` vnode to `self` in in-memory tree. NOTE: does not
@@ -230,6 +261,29 @@ impl Vnode {
             .cloned()
     }
 
+    pub(crate) fn for_each_entry<F: FnMut(usize, &VnodeRef)>(
+        &self,
+        offset: usize,
+        limit: usize,
+        mut f: F,
+    ) -> usize {
+        assert!(self.is_directory());
+        let mut count = 0;
+        for (index, item) in self
+            .tree
+            .borrow()
+            .children
+            .iter()
+            .skip(offset)
+            .take(limit)
+            .enumerate()
+        {
+            f(index, item);
+            count += 1;
+        }
+        count
+    }
+
     /// Looks up a child `name` in `self`. Will first try looking up a cached
     /// vnode and will load it from disk if it's missing.
     pub fn lookup_or_load(self: &VnodeRef, name: &str) -> Result<VnodeRef, Errno> {
@@ -303,35 +357,55 @@ impl Vnode {
 
     /// Opens a vnode for access
     pub fn open(self: &VnodeRef, flags: OpenFlags) -> Result<FileRef, Errno> {
-        if self.kind == VnodeKind::Directory {
-            return Err(Errno::IsADirectory);
+        let mut open_flags = 0;
+        if flags.contains(OpenFlags::O_DIRECTORY) {
+            if self.kind != VnodeKind::Directory {
+                return Err(Errno::NotADirectory);
+            }
+            if flags & OpenFlags::O_ACCESS != OpenFlags::O_RDONLY {
+                return Err(Errno::IsADirectory);
+            }
+
+            open_flags = File::READ;
+        } else {
+            if self.kind == VnodeKind::Directory {
+                return Err(Errno::IsADirectory);
+            }
+
+            match flags & OpenFlags::O_ACCESS {
+                OpenFlags::O_RDONLY => open_flags |= File::READ,
+                OpenFlags::O_WRONLY => open_flags |= File::WRITE,
+                OpenFlags::O_RDWR => open_flags |= File::READ | File::WRITE,
+                _ => unimplemented!(),
+            }
         }
 
-        let mut open_flags = 0;
-        match flags & OpenFlags::O_ACCESS {
-            OpenFlags::O_RDONLY => open_flags |= File::READ,
-            OpenFlags::O_WRONLY => open_flags |= File::WRITE,
-            OpenFlags::O_RDWR => open_flags |= File::READ | File::WRITE,
-            _ => unimplemented!(),
-        }
         if flags.contains(OpenFlags::O_CLOEXEC) {
             open_flags |= File::CLOEXEC;
         }
 
-        if let Some(ref mut data) = *self.data() {
-            let pos = data.open(self.clone(), flags)?;
-            Ok(File::normal(self.clone(), pos, open_flags))
+        if self.kind == VnodeKind::Directory && self.flags & Vnode::CACHE_READDIR != 0 {
+            Ok(File::normal(self.clone(), File::POS_CACHE_DOT, open_flags))
         } else {
-            Err(Errno::NotImplemented)
+            if let Some(ref mut data) = *self.data() {
+                let pos = data.open(self.clone(), flags)?;
+                Ok(File::normal(self.clone(), pos, open_flags))
+            } else {
+                Err(Errno::NotImplemented)
+            }
         }
     }
 
     /// Closes a vnode
     pub fn close(self: &VnodeRef) -> Result<(), Errno> {
-        if let Some(ref mut data) = *self.data() {
-            data.close(self.clone())
+        if self.kind == VnodeKind::Directory && self.flags & Vnode::CACHE_READDIR != 0 {
+            Ok(())
         } else {
-            Err(Errno::NotImplemented)
+            if let Some(ref mut data) = *self.data() {
+                data.close(self.clone())
+            } else {
+                Err(Errno::NotImplemented)
+            }
         }
     }
 
@@ -384,9 +458,16 @@ impl Vnode {
     }
 
     /// Reports file status
-    pub fn stat(self: &VnodeRef, stat: &mut Stat) -> Result<(), Errno> {
-        if let Some(ref mut data) = *self.data() {
-            data.stat(self.clone(), stat)
+    pub fn stat(self: &VnodeRef) -> Result<Stat, Errno> {
+        if self.flags & Self::CACHE_STAT != 0 {
+            let props = self.props();
+            Ok(Stat {
+                blksize: 0,
+                size: 0,
+                mode: props.mode
+            })
+        } else if let Some(ref mut data) = *self.data() {
+            data.stat(self.clone())
         } else {
             Err(Errno::NotImplemented)
         }
@@ -401,6 +482,7 @@ impl Vnode {
         }
     }
 
+    /// Returns `true` if the node is ready for operation
     pub fn is_ready(self: &VnodeRef, write: bool) -> Result<bool, Errno> {
         if let Some(ref mut data) = *self.data() {
             data.is_ready(self.clone(), write)
@@ -408,6 +490,39 @@ impl Vnode {
             Err(Errno::NotImplemented)
         }
     }
+
+    /// Checks if given [Ioctx] has `access` permissions to the vnode
+    pub fn check_access(&self, _ioctx: &Ioctx, access: AccessMode) -> Result<(), Errno> {
+        let props = self.props.borrow();
+        let mode = props.mode;
+
+        if access.contains(AccessMode::F_OK) {
+            if access.intersects(AccessMode::R_OK | AccessMode::W_OK | AccessMode::X_OK) {
+                return Err(Errno::InvalidArgument);
+            }
+            return Ok(());
+        } else {
+            if access.contains(AccessMode::F_OK) {
+                return Err(Errno::InvalidArgument);
+            }
+
+            // Check user
+            if access.contains(AccessMode::R_OK) && !mode.contains(FileMode::USER_READ) {
+                return Err(Errno::PermissionDenied);
+            }
+            if access.contains(AccessMode::W_OK) && !mode.contains(FileMode::USER_WRITE) {
+                return Err(Errno::PermissionDenied);
+            }
+            if access.contains(AccessMode::X_OK) && !mode.contains(FileMode::USER_EXEC) {
+                return Err(Errno::PermissionDenied);
+            }
+
+            // TODO check group
+            // TODO check other
+
+            return Ok(());
+        }
+    }
 }
 
 impl fmt::Debug for Vnode {

init/src/main.rs

@@ -0,0 +1,17 @@
+#![feature(asm)]
+#![no_std]
+#![no_main]
+
+#[macro_use]
+extern crate libusr;
+
+#[no_mangle]
+fn main() -> i32 {
+    loop {
+        trace!("Hello from userspace");
+        for _ in 0..100000 {
+            unsafe { asm!("nop"); }
+        }
+    }
+    123
+}

kernel/Cargo.toml

@@ -23,9 +23,11 @@ kernel-macros = { path = "macros" }
 cortex-a = { version = "6.x.x" }
 
 [features]
+default = ["aggressive_syscall"]
 pl011 = []
 pl031 = []
 verbose = []
+aggressive_syscall = []
 
 mach_qemu = ["pl011", "pl031"]
 mach_orangepi3 = []

kernel/build.rs

@@ -0,0 +1,12 @@
+use std::env;
+
+fn main() -> Result<(), i32> {
+    let arch = env::var("ARCH").expect("$ARCH is not set");
+    let mach = if arch == "aarch64" {
+        env::var("MACH").expect("$MACH is not set")
+    } else {
+        "none".to_owned()
+    };
+    println!("cargo:rerun-if-changed=../etc/{}-{}.ld", arch, mach);
+    Ok(())
+}

kernel/src/arch/aarch64/boot/mod.rs

@@ -4,12 +4,15 @@ use crate::arch::{
     aarch64::reg::{CNTKCTL_EL1, CPACR_EL1},
     machine,
 };
-use crate::config::{ConfigKey, CONFIG};
-use crate::dev::{
-    fdt::{find_prop, DeviceTree},
-    irq::IntSource,
-    Device,
+use crate::arch::{
+    aarch64::{
+        cpu,
+        smp,
+    },
 };
+use crate::config::{ConfigKey, CONFIG};
+use crate::dev::fdt::find_prop;
+use crate::dev::{fdt::DeviceTree, irq::IntSource, Device};
 use crate::fs::devfs;
 use libsys::error::Errno;
 //use crate::debug::Level;
@@ -20,17 +23,17 @@ use crate::mem::{
 };
 use crate::proc;
 use cortex_a::asm::barrier::{self, dsb, isb};
-use cortex_a::registers::{SCTLR_EL1, VBAR_EL1};
-use tock_registers::interfaces::{ReadWriteable, Writeable};
+use cortex_a::registers::{MPIDR_EL1, SCTLR_EL1, VBAR_EL1};
+use tock_registers::interfaces::{ReadWriteable, Readable, Writeable};
 
-fn init_device_tree(fdt_base_phys: usize) -> Result<(), Errno> {
+fn init_device_tree(fdt_base_phys: usize) -> Result<Option<DeviceTree>, Errno> {
     use fdt_rs::prelude::*;
 
     let fdt = if fdt_base_phys != 0 {
         DeviceTree::from_phys(fdt_base_phys + 0xFFFFFF8000000000)?
     } else {
         warnln!("No FDT present");
-        return Ok(());
+        return Ok(None);
     };
 
     #[cfg(feature = "verbose")]
@@ -56,11 +59,10 @@ fn init_device_tree(fdt_base_phys: usize) -> Result<(), Errno> {
         }
     }
 
-    Ok(())
+    Ok(Some(fdt))
 }
 
-#[no_mangle]
-extern "C" fn __aa64_bsp_main(fdt_base: usize) -> ! {
+fn cpu_setup_common() {
     // Disable FP instruction trapping
     CPACR_EL1.modify(CPACR_EL1::FPEN::TrapNone);
 
@@ -84,10 +86,34 @@ extern "C" fn __aa64_bsp_main(fdt_base: usize) -> ! {
         isb(barrier::SY);
     }
 
+    unsafe {
+        cpu::init_self();
+    }
+}
+
+#[no_mangle]
+extern "C" fn __aa64_secondary_main() -> ! {
+    cpu_setup_common();
+
+    unsafe {
+        use crate::dev::irq::IntController;
+        machine::local_timer().enable().unwrap();
+        machine::intc().enable_secondary();
+        machine::intc().enable_irq(machine::IrqNumber::new(30));
+
+        proc::enter(false);
+    }
+}
+
+#[no_mangle]
+extern "C" fn __aa64_bsp_main(fdt_base: usize) -> ! {
+    // Boot CPU is MPDIR_EL1 = 0
+    cpu_setup_common();
+
     // Enable MMU
     virt::enable().expect("Failed to initialize virtual memory");
 
-    init_device_tree(fdt_base).expect("Device tree init failed");
+    let fdt = init_device_tree(fdt_base).expect("Device tree init failed");
 
     // Most basic machine init: initialize proper debug output
     // physical memory
@@ -102,16 +128,21 @@ extern "C" fn __aa64_bsp_main(fdt_base: usize) -> ! {
     }
 
     devfs::init();
-
     machine::init_board().unwrap();
 
+    if let Some(fdt) = &fdt {
+        unsafe {
+            smp::enable_secondary_cpus(fdt);
+        }
+    }
+
     infoln!("Machine init finished");
 
     unsafe {
         machine::local_timer().enable().unwrap();
         machine::local_timer().init_irqs().unwrap();
 
-        proc::enter();
+        proc::enter(true);
     }
 }
 

kernel/src/arch/aarch64/boot/uboot.S

@@ -15,6 +15,22 @@
 _entry:
     mov x8, x0
 
+    // Zero .bss
+    ADR_ABS x0, __bss_start_phys
+    ADR_ABS x1, __bss_end_phys
+
+1:
+    cmp x0, x1
+    beq 2f
+
+    str xzr, [x0], #8
+
+    b 1b
+2:
+    ADR_ABS x9, __aa64_entry_upper
+    ADR_REL x10, __aa64_enter_upper
+
+_entry_common:
     // Test for EL2
     mrs x0, CurrentEL
     lsr x0, x0, #2
@@ -46,32 +62,34 @@ _entry:
     dsb sy
     isb
 
-    // Zero .bss
-    ADR_ABS x0, __bss_start_phys
-    ADR_ABS x1, __bss_end_phys
-1:
-    cmp x0, x1
-    beq 2f
-
-    str xzr, [x0], #8
-
-    b 1b
-2:
-
-    ADR_ABS x9, __aa64_entry_upper
-    b __aa64_enter_upper
+    br x10
 
 .section .text._entry_upper
 __aa64_entry_upper:
     // x0 -- fdt address
 
-    ADR_REL x1, bsp_stack_top
+    ADR_ABS x1, bsp_stack_top
     mov sp, x1
 
     mov lr, xzr
     bl __aa64_bsp_main
     b .
 
+__aa64_entry_upper_secondary:
+    // x0 -- stack
+    mov sp, x0
+    mov lr, xzr
+    bl __aa64_secondary_main
+    b .
+
+.section .text._entry_secondary
+.global _entry_secondary
+_entry_secondary:
+    mov x8, x0
+    ADR_ABS x9, __aa64_entry_upper_secondary
+    ADR_ABS x10, __aa64_enter_upper_secondary
+    b _entry_common
+
 .section .bss
 .p2align 12
 bsp_stack_bottom:

kernel/src/arch/aarch64/boot/upper.S

@@ -61,7 +61,13 @@ __aa64_enter_upper:
 
     cbnz x2, 1b
 
-.init_mmu_regs:
+__aa64_enter_upper_secondary:
+    ADR_ABS x5, KERNEL_TTBR1
+    ADR_ABS x6, KERNEL_OFFSET
+
+    // x5 = KERNEL_TTBR1 physical address
+    sub x5, x5, x6
+
     mov x0, #(MAIR_EL1_Attr0_Normal_Outer_NC | MAIR_EL1_Attr0_Normal_Inner_NC | MAIR_EL1_Attr1_Device | MAIR_EL1_Attr1_Device_nGnRE)
     msr mair_el1, x0
 

kernel/src/arch/aarch64/context.S

@@ -32,27 +32,6 @@ __aa64_ctx_enter_kernel:
     eret
 
 __aa64_ctx_enter_from_fork:
-    // stack.push(frame.x[18]);
-    // stack.push(frame.x[17]);
-    // stack.push(frame.x[16]);
-    // stack.push(frame.x[15]);
-    // stack.push(frame.x[14]);
-    // stack.push(frame.x[13]);
-    // stack.push(frame.x[12]);
-    // stack.push(frame.x[11]);
-    // stack.push(frame.x[10]);
-    // stack.push(frame.x[9]);
-    // stack.push(frame.x[8]);
-    // stack.push(frame.x[7]);
-    // stack.push(frame.x[6]);
-    // stack.push(frame.x[5]);
-    // stack.push(frame.x[4]);
-    // stack.push(frame.x[3]);
-    // stack.push(frame.x[2]);
-    // stack.push(frame.x[1]);
-
-    // stack.push(frame.elr_el1 as usize);
-    // stack.push(frame.sp_el0 as usize);
     ldp x0,  x1,  [sp, #16 * 0]
     msr sp_el0, x0
     msr elr_el1, x1
@@ -82,7 +61,8 @@ __aa64_ctx_switch:
     stp x27, x28, [sp, #16 * 4]
     stp x29, x30, [sp, #16 * 5]
     mrs x19, TTBR0_EL1
-    stp x19, xzr, [sp, #16 * 6]
+    mrs x20, TPIDR_EL0
+    stp x19, x20, [sp, #16 * 6]
 
     mov x19, sp
     str x19, [x1]
@@ -90,8 +70,9 @@ __aa64_ctx_switch_to:
     ldr x0, [x0]
     mov sp, x0
 
-    ldp x19, xzr, [sp, #16 * 6]
+    ldp x19, x20, [sp, #16 * 6]
     msr TTBR0_EL1, x19
+    msr TPIDR_EL0, x20
     ldp x19, x20, [sp, #16 * 0]
     ldp x21, x22, [sp, #16 * 1]
     ldp x23, x24, [sp, #16 * 2]

kernel/src/arch/aarch64/context.rs

@@ -67,7 +67,7 @@ impl Context {
         stack.push(frame.sp_el0 as usize);
 
         // Setup common
-        stack.push(0);
+        stack.push(0);          // tpidr_el0
         stack.push(ttbr0);
         stack.push(__aa64_ctx_enter_from_fork as usize); // x30/lr
         stack.push(frame.x[29]); // x29
@@ -96,7 +96,7 @@ impl Context {
 
         stack.push(entry);
         stack.push(arg);
-        stack.push(/* ttbr0 */ 0);
+        stack.push(0);
         stack.push(ustack);
 
         stack.setup_common(__aa64_ctx_enter_user as usize, ttbr0);
@@ -176,20 +176,20 @@ impl Stack {
     }
 
     pub fn setup_common(&mut self, entry: usize, ttbr: usize) {
-        self.push(0);
+        self.push(0);       // tpidr_el0
         self.push(ttbr);
-        self.push(entry); // x30/lr
-        self.push(0); // x29
-        self.push(0); // x28
-        self.push(0); // x27
-        self.push(0); // x26
-        self.push(0); // x25
-        self.push(0); // x24
-        self.push(0); // x23
-        self.push(0); // x22
-        self.push(0); // x21
-        self.push(0); // x20
-        self.push(0); // x19
+        self.push(entry);   // x30/lr
+        self.push(0);       // x29
+        self.push(0);       // x28
+        self.push(0);       // x27
+        self.push(0);       // x26
+        self.push(0);       // x25
+        self.push(0);       // x24
+        self.push(0);       // x23
+        self.push(0);       // x22
+        self.push(0);       // x21
+        self.push(0);       // x20
+        self.push(0);       // x19
     }
 
     pub fn push(&mut self, value: usize) {

kernel/src/arch/aarch64/cpu.rs

@@ -0,0 +1,72 @@
+#![allow(missing_docs)]
+
+use crate::proc::Scheduler;
+use crate::util::InitOnce;
+use core::mem::MaybeUninit;
+use core::ptr::null_mut;
+use core::sync::atomic::{AtomicUsize, Ordering};
+use cortex_a::registers::{MPIDR_EL1, TPIDR_EL1};
+use tock_registers::interfaces::{Readable, Writeable};
+
+#[repr(C)]
+pub struct Cpu {
+    counter: AtomicUsize, // 0x08
+
+    id: usize,
+    scheduler: Scheduler,
+}
+
+impl Cpu {
+    pub fn new(id: usize) -> Self {
+        Self {
+            counter: AtomicUsize::new(0),
+
+            id,
+            scheduler: Scheduler::new(),
+        }
+    }
+
+    pub fn id(&self) -> usize {
+        self.id
+    }
+
+    pub fn scheduler(&mut self) -> &Scheduler {
+        &self.scheduler
+    }
+
+    pub unsafe fn set(&mut self) {
+        TPIDR_EL1.set(self as *mut _ as u64);
+    }
+
+    pub unsafe fn get() -> &'static mut Self {
+        &mut *(TPIDR_EL1.get() as *mut Self)
+    }
+}
+
+pub unsafe fn cpus() -> impl Iterator<Item = &'static mut Cpu> {
+    CPUS[..CPU_COUNT.load(Ordering::Acquire)]
+        .iter_mut()
+        .map(|c| c.assume_init_mut())
+}
+
+pub unsafe fn by_index(idx: usize) -> &'static mut Cpu {
+    assert!(idx < CPU_COUNT.load(Ordering::Acquire));
+    CPUS[idx].assume_init_mut()
+}
+
+pub fn count() -> usize {
+    CPU_COUNT.load(Ordering::Acquire)
+}
+
+static CPU_COUNT: AtomicUsize = AtomicUsize::new(0);
+static mut CPUS: [MaybeUninit<Cpu>; 8] = MaybeUninit::uninit_array();
+
+pub unsafe fn init_self() {
+    let cpu_index = CPU_COUNT.load(Ordering::Acquire);
+    let mpidr_id = (MPIDR_EL1.get() & 0xF) as usize;
+
+    CPUS[cpu_index].write(Cpu::new(mpidr_id));
+    CPUS[cpu_index].assume_init_mut().set();
+
+    CPU_COUNT.store(cpu_index + 1, Ordering::Release);
+}

kernel/src/arch/aarch64/exception.rs

@@ -4,10 +4,10 @@ use crate::arch::machine;
 use crate::debug::Level;
 use crate::dev::irq::{IntController, IrqContext};
 use crate::mem;
-use crate::proc::{sched, Process};
+use crate::proc::{sched, Thread};
 use crate::syscall;
 use cortex_a::registers::{ESR_EL1, FAR_EL1};
-use libsys::{abi, signal::Signal};
+use libsys::{abi::SystemCall, signal::Signal};
 use tock_registers::interfaces::Readable;
 
 /// Trapped SIMD/FP functionality
@@ -88,17 +88,21 @@ extern "C" fn __aa64_exc_sync_handler(exc: &mut ExceptionFrame) {
     match err_code {
         EC_DATA_ABORT_EL0 | EC_DATA_ABORT_ELX => {
             let far = FAR_EL1.get() as usize;
+            let iss = esr & 0x1FFFFFF;
 
-            if far < mem::KERNEL_OFFSET && sched::is_ready() {
-                let proc = Process::current();
+            // TODO handle scenarios when sheduler is not yet initialized
+            if iss & (1 << 6) != 0 && far < mem::KERNEL_OFFSET {
+                let thread = Thread::current();
+                let proc = thread.owner().unwrap();
 
                 if proc
                     .manipulate_space(|space| space.try_cow_copy(far))
                     .is_err()
                 {
                     // Kill program
+                    errorln!("Data abort from {:#x}", exc.elr_el1);
                     dump_data_abort(Level::Error, esr, far as u64);
-                    proc.enter_signal(Signal::SegmentationFault);
+                    proc.enter_fault_signal(thread, Signal::SegmentationFault);
                 }
 
                 unsafe {
@@ -112,32 +116,47 @@ extern "C" fn __aa64_exc_sync_handler(exc: &mut ExceptionFrame) {
             }
 
             errorln!("Unresolved data abort");
+            errorln!("Data abort from {:#x}", exc.elr_el1);
             dump_data_abort(Level::Error, esr, far as u64);
         }
         EC_SVC_AA64 => {
-            unsafe {
-                if exc.x[8] == abi::SYS_FORK {
-                    match syscall::sys_fork(exc) {
-                        Ok(pid) => exc.x[0] = pid.value() as usize,
-                        Err(err) => {
-                            exc.x[0] = err.to_negative_isize() as usize;
-                        }
-                    }
-                    return;
-                }
+            let num = SystemCall::from_repr(exc.x[8]);
+            if num.is_none() {
+                todo!();
+            }
+            let num = num.unwrap();
 
-                match syscall::syscall(exc.x[8], &exc.x[..6]) {
-                    Ok(val) => exc.x[0] = val,
+            if num == SystemCall::Fork {
+                match unsafe { syscall::sys_fork(exc) } {
+                    Ok(pid) => exc.x[0] = pid.value() as usize,
                     Err(err) => {
                         exc.x[0] = err.to_negative_isize() as usize;
                     }
                 }
+                return;
             }
+
+            match syscall::syscall(num, &exc.x[..6]) {
+                Ok(val) => exc.x[0] = val,
+                Err(err) => {
+                    exc.x[0] = err.to_negative_isize() as usize;
+                }
+            }
+
             return;
         }
         _ => {}
     }
 
+    // if sched::is_ready() {
+    //     let thread = Thread::current();
+    //     errorln!(
+    //         "Unhandled exception in thread {}, {:?}",
+    //         thread.id(),
+    //         thread.owner().map(|e| e.id())
+    //     );
+    // }
+
     errorln!(
         "Unhandled exception at ELR={:#018x}, ESR={:#010x}",
         exc.elr_el1,

kernel/src/arch/aarch64/irq/gic/gicd.rs

@@ -1,7 +1,7 @@
 use crate::mem::virt::DeviceMemoryIo;
 use crate::sync::IrqSafeSpinLock;
 use tock_registers::interfaces::{Readable, Writeable};
-use tock_registers::registers::{ReadOnly, ReadWrite};
+use tock_registers::registers::{ReadOnly, WriteOnly, ReadWrite};
 use tock_registers::{register_bitfields, register_structs};
 
 register_bitfields! {
@@ -31,7 +31,9 @@ register_structs! {
         (0x820 => ITARGETSR: [ReadWrite<u32, ITARGETSR::Register>; 248]),
         (0xC00 => _res2),
         (0xC08 => ICFGR: [ReadWrite<u32>; 62]),
-        (0xC0C => @END),
+        (0xD00 => _res3),
+        (0xF00 => SGIR: WriteOnly<u32>),
+        (0xF04 => @END),
     }
 }
 
@@ -103,6 +105,14 @@ impl Gicd {
         }
     }
 
+    pub fn set_sgir(&self, filter: bool, mask: u32, intid: u32) {
+        let mut value = (mask << 16) | intid;
+        if filter {
+            value |= 1 << 24;
+        }
+        self.shared_regs.lock().SGIR.set(value);
+    }
+
     pub fn enable_irq(&self, irq: super::IrqNumber) {
         let irq = irq.get();
 

kernel/src/arch/aarch64/irq/gic/mod.rs

@@ -1,7 +1,7 @@
 //! ARM Generic Interrupt Controller
 
 use crate::dev::{
-    irq::{IntController, IntSource, IrqContext},
+    irq::{IntController, IntSource, IrqContext, IpiSender},
     Device,
 };
 use crate::mem::virt::{DeviceMemory, DeviceMemoryIo};
@@ -17,6 +17,8 @@ use gicd::Gicd;
 /// Maximum available IRQ number
 pub const MAX_IRQ: usize = 300;
 
+const SGI_IRQ: u32 = 2;
+
 /// Range-checked IRQ number type
 #[repr(transparent)]
 #[derive(Copy, Clone)]
@@ -28,7 +30,6 @@ pub struct Gic {
     gicd: InitOnce<Gicd>,
     gicd_base: usize,
     gicc_base: usize,
-    scheduler_irq: IrqNumber,
     table: IrqSafeSpinLock<[Option<&'static (dyn IntSource + Sync)>; MAX_IRQ]>,
 }
 
@@ -62,11 +63,12 @@ impl Device for Gic {
         let gicc = Gicc::new(gicc_mmio);
 
         gicd.enable();
-        gicc.enable();
 
         self.gicd.init(gicd);
         self.gicc.init(gicc);
 
+        self.enable_secondary();
+
         Ok(())
     }
 }
@@ -86,11 +88,30 @@ impl IntController for Gic {
             return;
         }
 
-        if self.scheduler_irq.0 == irq_number {
+//<<<<<<< HEAD
+        if irq_number == 1 {
             gicc.clear_irq(irq_number as u32, ic);
+            debugln!("Received IPI");
+            loop {}
         }
+//
+//        if self.scheduler_irq.0 == irq_number {
+//            use crate::proc::sched;
+//            use cortex_a::registers::{CNTP_TVAL_EL0, CNTP_CTL_EL0};
+//            use tock_registers::interfaces::Writeable;
+//            use crate::arch::platform::cpu::Cpu;
+//            gicc.clear_irq(irq_number as u32, ic);
+//            CNTP_TVAL_EL0.set(1000000);
+//            CNTP_CTL_EL0.write(CNTP_CTL_EL0::ENABLE::SET);
+//            sched::switch(false);
+//            return;
+//        }
+//=======
+        gicc.clear_irq(irq_number as u32, ic);
+//>>>>>>> feat/thread
 
         {
+            // TODO make timer interrupt a special case and drop table lock
             let table = self.table.lock();
             match table[irq_number] {
                 None => panic!("No handler registered for irq{}", irq_number),
@@ -100,8 +121,6 @@ impl IntController for Gic {
                 }
             }
         }
-
-        gicc.clear_irq(irq_number as u32, ic);
     }
 
     fn register_handler(
@@ -122,19 +141,29 @@ impl IntController for Gic {
     }
 }
 
+impl IpiSender for Gic {
+    fn send_to_mask(&self, exclude_self: bool, target: u32, data: u64) {
+        self.gicd.get().set_sgir(exclude_self, target, 1);
+    }
+}
+
 impl Gic {
+    ///
+    pub unsafe fn enable_secondary(&self) {
+        self.gicc.get().enable();
+    }
+
     /// Constructs an instance of GICv2.
     ///
     /// # Safety
     ///
     /// Does not perform `gicd_base` and `gicc_base` validation.
-    pub const unsafe fn new(gicd_base: usize, gicc_base: usize, scheduler_irq: IrqNumber) -> Self {
+    pub const unsafe fn new(gicd_base: usize, gicc_base: usize) -> Self {
         Self {
             gicc: InitOnce::new(),
             gicd: InitOnce::new(),
             gicd_base,
             gicc_base,
-            scheduler_irq,
             table: IrqSafeSpinLock::new([None; MAX_IRQ]),
         }
     }

kernel/src/arch/aarch64/mach_qemu/mod.rs

@@ -72,12 +72,18 @@ pub fn local_timer() -> &'static GenericTimer {
 
 /// Returns CPU's interrupt controller device
 #[inline]
-pub fn intc() -> &'static impl IntController<IrqNumber = IrqNumber> {
+pub fn intc() -> &'static Gic {
+    &GIC
+}
+
+/// Returns CPU's IPI sender device
+#[inline]
+pub fn ipi_sender() -> &'static Gic {
     &GIC
 }
 
 static UART0: Pl011 = unsafe { Pl011::new(UART0_BASE, UART0_IRQ) };
 static RTC: Pl031 = unsafe { Pl031::new(RTC_BASE, RTC_IRQ) };
-static GIC: Gic = unsafe { Gic::new(GICD_BASE, GICC_BASE, LOCAL_TIMER_IRQ) };
+static GIC: Gic = unsafe { Gic::new(GICD_BASE, GICC_BASE) };
 static PCIE: GenericPcieHost = unsafe { GenericPcieHost::new(ECAM_BASE, 8) };
 static LOCAL_TIMER: GenericTimer = GenericTimer::new(LOCAL_TIMER_IRQ);

kernel/src/arch/aarch64/mod.rs

@@ -5,9 +5,11 @@ use tock_registers::interfaces::{Readable, Writeable};
 
 pub mod boot;
 pub mod context;
+pub mod cpu;
 pub mod exception;
 pub mod irq;
 pub mod reg;
+pub mod smp;
 pub mod timer;
 
 cfg_if! {

kernel/src/arch/aarch64/smp.rs

@@ -0,0 +1,123 @@
+#![allow(missing_docs)]
+
+use crate::arch::{aarch64::cpu, machine};
+use crate::dev::{
+    fdt::{self, DeviceTree},
+    irq::IpiSender,
+};
+use crate::mem::{
+    self,
+    phys::{self, PageUsage},
+};
+use cortex_a::registers::MPIDR_EL1;
+use libsys::error::Errno;
+use fdt_rs::prelude::*;
+use tock_registers::interfaces::Readable;
+
+pub type NodeAddress = u32;
+
+#[derive(Clone, Copy, Debug)]
+pub enum PsciError {
+    NotSupported,
+    InvalidParameters,
+    Denied,
+    AlreadyOn,
+    OnPending,
+    InternalFailure,
+    NotPresent,
+    Disabled,
+    InvalidAddress,
+}
+
+struct Psci {
+    use_smc: bool,
+}
+
+impl Psci {
+    const PSCI_VERSION: usize = 0x84000000;
+    const PSCI_CPU_OFF: usize = 0x84000002;
+    const PSCI_CPU_ON: usize = 0xC4000003;
+
+    pub const fn new() -> Self {
+        Self { use_smc: true }
+    }
+
+    unsafe fn call(&self, x0: usize, x1: usize, x2: usize, x3: usize) -> usize {
+        if self.use_smc {
+            call_smc(x0, x1, x2, x3)
+        } else {
+            todo!()
+        }
+    }
+
+    pub unsafe fn cpu_on(
+        &self,
+        target_cpu: usize,
+        entry_point_address: usize,
+        context_id: usize,
+    ) -> Result<(), PsciError> {
+        wrap_psci_ok(self.call(
+            Self::PSCI_CPU_ON,
+            target_cpu,
+            entry_point_address,
+            context_id,
+        ))
+    }
+}
+
+const SECONDARY_STACK_PAGES: usize = 4;
+
+unsafe fn call_smc(mut x0: usize, x1: usize, x2: usize, x3: usize) -> usize {
+    asm!("smc #0", inout("x0") x0, in("x1") x1, in("x2") x2, in("x3") x3);
+    x0
+}
+
+pub unsafe fn send_ipi(exclude_self: bool, target_mask: u32, data: u64) {
+    machine::ipi_sender().send_to_mask(exclude_self, target_mask, data);
+}
+
+fn wrap_psci_ok(a: usize) -> Result<(), PsciError> {
+    const NOT_SUPPORTED: isize = -1;
+    const INVALID_PARAMETERS: isize = -2;
+    const DENIED: isize = -3;
+    const ALREADY_ON: isize = -4;
+
+    match a as isize {
+        0 => Ok(()),
+        NOT_SUPPORTED => Err(PsciError::NotSupported),
+        INVALID_PARAMETERS => Err(PsciError::InvalidParameters),
+        DENIED => Err(PsciError::Denied),
+        ALREADY_ON => Err(PsciError::AlreadyOn),
+        _ => unimplemented!(),
+    }
+}
+
+pub unsafe fn enable_secondary_cpus(dt: &DeviceTree) {
+    extern "C" {
+        fn _entry_secondary();
+    }
+
+    let cpus = dt.node_by_path("/cpus").unwrap();
+    let psci = Psci::new();
+
+    for cpu_node in cpus.children() {
+        let reg = fdt::find_prop(cpu_node, "reg").unwrap().u32(0).unwrap();
+        if reg == 0 {
+            continue;
+        }
+        infoln!("Enabling cpu{}", reg);
+        let stack_pages =
+            phys::alloc_contiguous_pages(PageUsage::Kernel, SECONDARY_STACK_PAGES).unwrap();
+        let count_old = cpu::count();
+        psci.cpu_on(
+            reg as usize,
+            _entry_secondary as usize - mem::KERNEL_OFFSET,
+            mem::virtualize(stack_pages + SECONDARY_STACK_PAGES * mem::PAGE_SIZE),
+        )
+        .unwrap();
+        while cpu::count() == count_old {
+            cortex_a::asm::wfe();
+        }
+        debugln!("Done");
+    }
+}

kernel/src/arch/aarch64/timer.rs

@@ -34,9 +34,9 @@ impl IntSource for GenericTimer {
     fn handle_irq(&self) -> Result<(), Errno> {
         CNTP_TVAL_EL0.set(TIMER_TICK);
         CNTP_CTL_EL0.write(CNTP_CTL_EL0::ENABLE::SET);
-        use crate::proc;
-        proc::wait::tick();
-        proc::switch();
+        use crate::proc::{wait, sched};
+        wait::tick();
+        sched::switch(false);
         Ok(())
     }
 

kernel/src/arch/aarch64/vectors.S

@@ -74,8 +74,18 @@ __aa\bits\()_el\el\ht\()_\kind:
 
     EXC_SAVE_STATE
     mov x0, sp
+.if \el == 0
+    sub sp, sp, #16
+    stp fp, lr, [sp, #0]
+.else
     mov lr, xzr
+.endif
     bl __aa64_exc_\kind\()_handler
+
+.if \el == 0
+    add sp, sp, #16
+.endif
+
     EXC_RESTORE_STATE
     eret
 .endm

kernel/src/debug.rs

@@ -12,6 +12,8 @@
 //! * [errorln!]
 
 use crate::dev::serial::SerialDevice;
+use crate::sync::IrqSafeSpinLock;
+use libsys::debug::TraceLevel;
 use core::fmt;
 
 /// Kernel logging levels
@@ -27,6 +29,18 @@ pub enum Level {
     Error,
 }
 
+impl From<TraceLevel> for Level {
+    #[inline(always)]
+    fn from(l: TraceLevel) -> Self {
+        match l {
+            TraceLevel::Debug => Self::Debug,
+            TraceLevel::Info => Self::Info,
+            TraceLevel::Warn => Self::Warn,
+            TraceLevel::Error => Self::Error,
+        }
+    }
+}
+
 struct SerialOutput<T: 'static + SerialDevice> {
     inner: &'static T,
 }
@@ -102,9 +116,11 @@ macro_rules! errorln {
 
 #[doc(hidden)]
 pub fn _debug(_level: Level, args: fmt::Arguments) {
+    static LOCK: IrqSafeSpinLock<()> = IrqSafeSpinLock::new(());
     use crate::arch::machine;
     use fmt::Write;
 
+    let _lock = LOCK.lock();
     SerialOutput {
         inner: machine::console(),
     }

kernel/src/dev/fdt.rs

@@ -3,7 +3,9 @@ use crate::debug::Level;
 use fdt_rs::prelude::*;
 use fdt_rs::{
     base::DevTree,
-    index::{DevTreeIndex, DevTreeIndexNode, DevTreeIndexProp},
+    index::{
+        iters::DevTreeIndexCompatibleNodeIter, DevTreeIndex, DevTreeIndexNode, DevTreeIndexProp,
+    },
 };
 use libsys::{error::Errno, path::path_component_left};
 
@@ -44,7 +46,7 @@ fn dump_node(level: Level, node: &INode, depth: usize) {
         print!(level, "{:?} = ", name);
 
         match name {
-            "compatible" => print!(level, "{:?}", prop.str().unwrap()),
+            "compatible" | "enable-method" => print!(level, "{:?}", prop.str().unwrap()),
             "#address-cells" | "#size-cells" => print!(level, "{}", prop.u32(0).unwrap()),
             "reg" => {
                 print!(level, "<");
@@ -115,6 +117,20 @@ impl DeviceTree {
 
     /// Loads a device tree from physical `base` address and
     /// creates an index for it
+    pub fn compatible<'a, 's>(&'a self, compat: &'s str) -> DevTreeIndexCompatibleNodeIter<'s, 'a, 'a, 'a> {
+        self.index.compatible_nodes(compat)
+    }
+
+    pub fn initrd(&self) -> Option<(usize, usize)> {
+        let chosen = self.node_by_path("/chosen")?;
+        let initrd_start = find_prop(chosen.clone(), "linux,initrd-start")?
+            .u32(0)
+            .ok()?;
+        let initrd_end = find_prop(chosen, "linux,initrd-end")?.u32(0).ok()?;
+
+        Some((initrd_start as usize, initrd_end as usize))
+    }
+
     pub fn from_phys(base: usize) -> Result<DeviceTree, Errno> {
         // TODO virtualize address
         let tree = unsafe { DevTree::from_raw_pointer(base as *const _) }

kernel/src/dev/irq.rs

@@ -1,4 +1,5 @@
 //! Interrupt controller and handler interfaces
+use crate::arch::platform::smp::NodeAddress;
 use crate::dev::Device;
 use core::marker::PhantomData;
 use libsys::error::Errno;
@@ -27,6 +28,12 @@ pub trait IntController: Device {
     fn handle_pending_irqs<'irq_context>(&'irq_context self, ic: &IrqContext<'irq_context>);
 }
 
+/// Inter-processor interrupt delivery method
+pub trait IpiSender: Device {
+    /// Raise an IPI for the target CPU mask, optionally excluding source CPU
+    fn send_to_mask(&self, except_self: bool, target: u32, data: u64);
+}
+
 /// Interface for peripherals capable of emitting IRQs
 pub trait IntSource: Device {
     /// Handles pending IRQs, if any, of this [IntSource].

kernel/src/dev/serial/pl011.rs

@@ -131,7 +131,6 @@ impl IntSource for Pl011 {
     fn handle_irq(&self) -> Result<(), Errno> {
         let inner = self.inner.get().lock();
         inner.regs.ICR.write(ICR::ALL::CLEAR);
-
         let byte = inner.regs.DR.get();
         drop(inner);
 

kernel/src/dev/tty.rs

@@ -1,14 +1,16 @@
 //! Teletype (TTY) device facilities
 use crate::dev::serial::SerialDevice;
-use crate::proc::wait::{Wait, WAIT_SELECT};
+use crate::proc::{Process, wait::{Wait, WAIT_SELECT}};
 use crate::sync::IrqSafeSpinLock;
 use libsys::error::Errno;
 use libsys::{
     termios::{Termios, TermiosIflag, TermiosLflag, TermiosOflag},
+    proc::Pid,
+    signal::Signal,
     ioctl::IoctlCmd
 };
 use core::mem::size_of;
-use crate::syscall::arg::validate_user_ptr_struct;
+use crate::syscall::arg;
 
 #[derive(Debug)]
 struct CharRingInner<const N: usize> {
@@ -16,6 +18,7 @@ struct CharRingInner<const N: usize> {
     wr: usize,
     data: [u8; N],
     flags: u8,
+    fg_pgid: Option<Pid>,
 }
 
 /// Ring buffer for TTYs
@@ -31,6 +34,7 @@ pub trait TtyDevice<const N: usize>: SerialDevice {
     /// Returns a reference to character device's ring buffer
     fn ring(&self) -> &CharRing<N>;
 
+    /// Returns `true` if the TTY is ready for an operation
     fn is_ready(&self, write: bool) -> Result<bool, Errno> {
         let ring = self.ring();
         if write {
@@ -45,15 +49,20 @@ pub trait TtyDevice<const N: usize>: SerialDevice {
         match cmd {
             IoctlCmd::TtyGetAttributes => {
                 // TODO validate size
-                let res = validate_user_ptr_struct::<Termios>(ptr)?;
+                let res = arg::struct_mut::<Termios>(ptr)?;
                 *res = self.ring().config.lock().clone();
                 Ok(size_of::<Termios>())
             },
             IoctlCmd::TtySetAttributes => {
-                let src = validate_user_ptr_struct::<Termios>(ptr)?;
+                let src = arg::struct_ref::<Termios>(ptr)?;
                 *self.ring().config.lock() = src.clone();
                 Ok(size_of::<Termios>())
             },
+            IoctlCmd::TtySetPgrp => {
+                let src = arg::struct_ref::<u32>(ptr)?;
+                self.ring().inner.lock().fg_pgid = Some(unsafe { Pid::from_raw(*src) });
+                Ok(0)
+            },
             _ => Err(Errno::InvalidArgument)
         }
     }
@@ -110,6 +119,19 @@ pub trait TtyDevice<const N: usize>: SerialDevice {
             }
         }
 
+        if byte == 0x3 && config.lflag.contains(TermiosLflag::ISIG) {
+            drop(config);
+            let pgid = ring.inner.lock().fg_pgid;
+            if let Some(pgid) = pgid {
+                // TODO send to pgid
+                let proc = Process::get(pgid);
+                if let Some(proc) = proc {
+                    proc.set_signal(Signal::Interrupt);
+                }
+            }
+            return;
+        }
+
         self.ring().putc(byte, false).ok();
     }
 
@@ -232,17 +254,19 @@ impl<const N: usize> CharRing<N> {
     pub const fn new() -> Self {
         Self {
             inner: IrqSafeSpinLock::new(CharRingInner {
+                fg_pgid: None,
                 rd: 0,
                 wr: 0,
                 data: [0; N],
                 flags: 0,
             }),
             config: IrqSafeSpinLock::new(Termios::new()),
-            wait_read: Wait::new(),
-            wait_write: Wait::new(),
+            wait_read: Wait::new("tty_read"),
+            wait_write: Wait::new("tty_write"),
         }
     }
 
+    /// Returns `true` if a character/line is available for reception
     pub fn is_readable(&self) -> bool {
         let inner = self.inner.lock();
         let config = self.config.lock();

kernel/src/fs/devfs.rs

@@ -2,7 +2,7 @@
 use crate::util::InitOnce;
 use alloc::boxed::Box;
 use core::sync::atomic::{AtomicUsize, Ordering};
-use libsys::error::Errno;
+use libsys::{stat::FileMode, error::Errno};
 use vfs::{CharDevice, CharDeviceWrapper, Vnode, VnodeKind, VnodeRef};
 
 /// Possible character device kinds
@@ -16,7 +16,9 @@ static DEVFS_ROOT: InitOnce<VnodeRef> = InitOnce::new();
 
 /// Initializes devfs
 pub fn init() {
-    DEVFS_ROOT.init(Vnode::new("", VnodeKind::Directory, 0));
+    let node = Vnode::new("", VnodeKind::Directory, Vnode::CACHE_READDIR | Vnode::CACHE_STAT);
+    node.props_mut().mode = FileMode::default_dir();
+    DEVFS_ROOT.init(node);
 }
 
 /// Returns devfs root node reference
@@ -27,7 +29,8 @@ pub fn root() -> &'static VnodeRef {
 fn _add_char_device(dev: &'static dyn CharDevice, name: &str) -> Result<(), Errno> {
     infoln!("Add char device: {}", name);
 
-    let node = Vnode::new(name, VnodeKind::Char, 0);
+    let node = Vnode::new(name, VnodeKind::Char, Vnode::CACHE_STAT);
+    node.props_mut().mode = FileMode::from_bits(0o600).unwrap() | FileMode::S_IFCHR;
     node.set_data(Box::new(CharDeviceWrapper::new(dev)));
 
     DEVFS_ROOT.get().attach(node);

kernel/src/fs/mod.rs

@@ -3,6 +3,8 @@ use crate::mem::{
     self,
     phys::{self, PageUsage},
 };
+use libsys::{error::Errno, stat::MountOptions};
+use vfs::VnodeRef;
 use memfs::BlockAllocator;
 
 pub mod devfs;
@@ -25,3 +27,14 @@ unsafe impl BlockAllocator for MemfsBlockAlloc {
         phys::free_page(phys).unwrap();
     }
 }
+
+/// Creates a filesystem instance based on `options`
+pub fn create_filesystem(options: &MountOptions) -> Result<VnodeRef, Errno> {
+    let fs_name = options.fs.unwrap();
+
+    if fs_name == "devfs" {
+        Ok(devfs::root().clone())
+    } else {
+        todo!();
+    }
+}

kernel/src/init.rs

@@ -4,7 +4,7 @@ use crate::config::{ConfigKey, CONFIG};
 use crate::fs::{devfs, MemfsBlockAlloc};
 use crate::mem;
 use crate::proc::{elf, Process};
-use libsys::stat::{FileDescriptor, OpenFlags};
+use libsys::stat::{FileDescriptor, OpenFlags, UserId, GroupId};
 use memfs::Ramfs;
 use vfs::{Filesystem, Ioctx};
 
@@ -29,7 +29,7 @@ pub extern "C" fn init_fn(_arg: usize) -> ! {
         unsafe { Ramfs::open(initrd_start as *mut u8, initrd_size, MemfsBlockAlloc {}).unwrap() };
     let root = fs.root().unwrap();
 
-    let ioctx = Ioctx::new(root);
+    let ioctx = Ioctx::new(root, UserId::root(), GroupId::root());
 
     let node = ioctx.find(None, "/init", true).unwrap();
     let file = node.open(OpenFlags::O_RDONLY | OpenFlags::O_EXEC).unwrap();
@@ -54,10 +54,11 @@ pub extern "C" fn init_fn(_arg: usize) -> ! {
         io.set_file(FileDescriptor::STDIN, stdin).unwrap();
         io.set_file(FileDescriptor::STDOUT, stdout).unwrap();
         io.set_file(FileDescriptor::STDERR, stderr).unwrap();
+        io.set_ctty(tty_node);
     }
 
     drop(cfg);
 
-    Process::execve(|space| elf::load_elf(space, file), 0).unwrap();
+    Process::execve(|space| elf::load_elf(space, file), &["/init"]).unwrap();
     panic!("Unreachable");
 }

kernel/src/main.rs

@@ -12,7 +12,8 @@
     panic_info_message,
     alloc_error_handler,
     linked_list_cursors,
-    const_btree_new
+    const_btree_new,
+    maybe_uninit_uninit_array
 )]
 #![no_std]
 #![no_main]
@@ -44,9 +45,19 @@ pub mod util;
 fn panic_handler(pi: &core::panic::PanicInfo) -> ! {
     unsafe {
         asm!("msr daifset, #2");
+        use crate::arch::platform::cpu::{self, Cpu};
+
+        crate::arch::platform::smp::send_ipi(true, (1 << cpu::count()) - 1, 0);
     }
 
-    errorln!("Panic: {:?}", pi);
+    use cortex_a::registers::MPIDR_EL1;
+    use tock_registers::interfaces::Readable;
+
+    errorln!("Panic on node{}: {:?}", MPIDR_EL1.get() & 0xF, pi);
     // TODO
-    loop {}
+    loop {
+        unsafe {
+            asm!("wfe");
+        }
+    }
 }

kernel/src/mem/phys/manager.rs

@@ -17,6 +17,7 @@ pub struct SimpleManager {
     pages: &'static mut [PageInfo],
     stats: PageStatistics,
     base_index: usize,
+    last_index: usize,
 }
 impl SimpleManager {
     pub(super) unsafe fn initialize(base: usize, at: usize, count: usize) -> Self {
@@ -34,6 +35,7 @@ impl SimpleManager {
         }
         Self {
             base_index: base / PAGE_SIZE,
+            last_index: 0,
             stats: PageStatistics {
                 available: 0,
                 kernel: 0,
@@ -57,11 +59,21 @@ impl SimpleManager {
     }
 
     fn alloc_single_index(&mut self, pu: PageUsage) -> Result<usize, Errno> {
-        for index in 0..self.pages.len() {
+        for index in self.last_index..self.pages.len() {
             let page = &mut self.pages[index];
             if page.usage == PageUsage::Available {
                 page.usage = pu;
                 page.refcount = 1;
+                self.last_index = index;
+                return Ok(index);
+            }
+        }
+        for index in 0..self.last_index {
+            let page = &mut self.pages[index];
+            if page.usage == PageUsage::Available {
+                page.usage = pu;
+                page.refcount = 1;
+                self.last_index = index;
                 return Ok(index);
             }
         }
@@ -81,18 +93,18 @@ impl SimpleManager {
         self.stats.available -= count;
     }
 
-    fn update_stats_free(&mut self, pu: PageUsage, count: usize) {
-        let field = match pu {
-            PageUsage::Kernel => &mut self.stats.kernel,
-            PageUsage::KernelHeap => &mut self.stats.kernel_heap,
-            PageUsage::Paging => &mut self.stats.paging,
-            PageUsage::UserPrivate => &mut self.stats.user_private,
-            PageUsage::Filesystem => &mut self.stats.filesystem,
-            _ => panic!("TODO {:?}", pu),
-        };
-        *field -= count;
-        self.stats.available += count;
-    }
+    // fn update_stats_free(&mut self, pu: PageUsage, count: usize) {
+    //     let field = match pu {
+    //         PageUsage::Kernel => &mut self.stats.kernel,
+    //         PageUsage::KernelHeap => &mut self.stats.kernel_heap,
+    //         PageUsage::Paging => &mut self.stats.paging,
+    //         PageUsage::UserPrivate => &mut self.stats.user_private,
+    //         PageUsage::Filesystem => &mut self.stats.filesystem,
+    //         _ => panic!("TODO {:?}", pu),
+    //     };
+    //     *field -= count;
+    //     self.stats.available += count;
+    // }
 }
 unsafe impl Manager for SimpleManager {
     fn alloc_page(&mut self, pu: PageUsage) -> Result<usize, Errno> {
@@ -122,23 +134,21 @@ unsafe impl Manager for SimpleManager {
         Err(Errno::OutOfMemory)
     }
     fn free_page(&mut self, addr: usize) -> Result<(), Errno> {
-        let usage = {
-            let index = self.page_index(addr);
-            let page = &mut self.pages[index];
+        let index = self.page_index(addr);
+        let page = &mut self.pages[index];
 
-            let usage = page.usage;
-            assert!(page.usage != PageUsage::Reserved && page.usage != PageUsage::Available);
+        assert!(page.usage != PageUsage::Reserved && page.usage != PageUsage::Available);
 
-            if page.refcount > 1 {
-                page.refcount -= 1;
-            } else {
-                assert_eq!(page.refcount, 1);
-                page.usage = PageUsage::Available;
-                page.refcount = 0;
-            }
+        if page.refcount > 1 {
+            page.refcount -= 1;
+        } else {
+            assert_eq!(page.refcount, 1);
+            page.usage = PageUsage::Available;
+            page.refcount = 0;
+
+            self.last_index = index;
+        }
 
-            usage
-        };
         // FIXME
         // self.update_stats_free(usage, 1);
 

kernel/src/mem/virt/table.rs

@@ -210,6 +210,26 @@ impl Space {
         }
     }
 
+    /// Translates a virtual address into a corresponding physical one.
+    ///
+    /// Only works for 4K pages atm.
+    // TODO extract attributes
+    pub fn translate(&mut self, virt: usize) -> Result<usize, Errno> {
+        let l0i = virt >> 30;
+        let l1i = (virt >> 21) & 0x1FF;
+        let l2i = (virt >> 12) & 0x1FF;
+
+        let l1_table = self.0.next_level_table(l0i).ok_or(Errno::DoesNotExist)?;
+        let l2_table = l1_table.next_level_table(l1i).ok_or(Errno::DoesNotExist)?;
+
+        let entry = l2_table[l2i];
+        if entry.is_present() {
+            Ok(unsafe { entry.address_unchecked() })
+        } else {
+            Err(Errno::DoesNotExist)
+        }
+    }
+
     /// Attempts to resolve a page fault at `virt` address by copying the
     /// underlying Copy-on-Write mapping (if any is present)
     pub fn try_cow_copy(&mut self, virt: usize) -> Result<(), Errno> {
@@ -247,6 +267,71 @@ impl Space {
         Ok(())
     }
 
+    /// Allocates a contiguous region from the address space and maps
+    /// physical pages to it
+    pub fn allocate(
+        &mut self,
+        start: usize,
+        end: usize,
+        len: usize,
+        flags: MapAttributes,
+        usage: PageUsage,
+    ) -> Result<usize, Errno> {
+        'l0: for page in (start..end).step_by(0x1000) {
+            for i in 0..len {
+                if self.translate(page + i * 0x1000).is_ok() {
+                    continue 'l0;
+                }
+            }
+
+            for i in 0..len {
+                let phys = phys::alloc_page(usage).unwrap();
+                self.map(page + i * 0x1000, phys, flags).unwrap();
+            }
+            return Ok(page);
+        }
+        Err(Errno::OutOfMemory)
+    }
+
+    /// Removes a single 4K page mapping from the table and
+    /// releases the underlying physical memory
+    pub fn unmap_single(&mut self, page: usize) -> Result<(), Errno> {
+        let l0i = page >> 30;
+        let l1i = (page >> 21) & 0x1FF;
+        let l2i = (page >> 12) & 0x1FF;
+
+        let l1_table = self.0.next_level_table(l0i).ok_or(Errno::DoesNotExist)?;
+        let l2_table = l1_table.next_level_table(l1i).ok_or(Errno::DoesNotExist)?;
+
+        let entry = l2_table[l2i];
+
+        if !entry.is_present() {
+            return Err(Errno::DoesNotExist);
+        }
+
+        let phys = unsafe { entry.address_unchecked() };
+        unsafe {
+            phys::free_page(phys)?;
+        }
+        l2_table[l2i] = Entry::invalid();
+
+        unsafe {
+            asm!("tlbi vaae1, {}", in(reg) page);
+        }
+
+        // TODO release paging structure memory
+
+        Ok(())
+    }
+
+    /// Releases a range of virtual pages and their corresponding physical pages
+    pub fn free(&mut self, start: usize, len: usize) -> Result<(), Errno> {
+        for i in 0..len {
+            self.unmap_single(start + i * 0x1000)?;
+        }
+        Ok(())
+    }
+
     /// Performs a copy of the address space, cloning data owned by it
     pub fn fork(&mut self) -> Result<&'static mut Self, Errno> {
         let res = Self::alloc_empty()?;
@@ -271,12 +356,19 @@ impl Space {
                                 todo!();
                                 // res.map(virt_addr, dst_phys, flags)?;
                             } else {
-                                // TODO only apply CoW to writable pages
-                                flags |= MapAttributes::AP_BOTH_READONLY | MapAttributes::EX_COW;
-                                l2_table[l2i].set_cow();
-                                unsafe {
-                                    asm!("tlbi vaae1, {}", in(reg) virt_addr);
+                                let writable = flags & MapAttributes::AP_BOTH_READONLY
+                                    == MapAttributes::AP_BOTH_READWRITE;
+
+                                if writable {
+                                    flags |=
+                                        MapAttributes::AP_BOTH_READONLY | MapAttributes::EX_COW;
+                                    l2_table[l2i].set_cow();
+
+                                    unsafe {
+                                        asm!("tlbi vaae1, {}", in(reg) virt_addr);
+                                    }
                                 }
+
                                 res.map(virt_addr, dst_phys, flags)?;
                             }
                         }

kernel/src/proc/io.rs

@@ -1,12 +1,13 @@
 //! Process file descriptors and I/O context
 use alloc::collections::BTreeMap;
-use libsys::{error::Errno, stat::FileDescriptor};
-use vfs::{FileRef, Ioctx};
+use libsys::{error::Errno, stat::{FileDescriptor, UserId, GroupId}};
+use vfs::{FileRef, Ioctx, VnodeRef, VnodeKind};
 
 /// Process I/O context. Contains file tables, root/cwd info etc.
 pub struct ProcessIo {
     ioctx: Option<Ioctx>,
     files: BTreeMap<u32, FileRef>,
+    ctty: Option<VnodeRef>,
 }
 
 impl ProcessIo {
@@ -21,6 +22,73 @@ impl ProcessIo {
         Ok(dst)
     }
 
+    /// Sets controlling terminal for the process
+    pub fn set_ctty(&mut self, node: VnodeRef) {
+        assert_eq!(node.kind(), VnodeKind::Char);
+        self.ctty = Some(node);
+    }
+
+    /// Returns current controlling terminal of the process
+    pub fn ctty(&mut self) -> Option<VnodeRef> {
+        self.ctty.clone()
+    }
+
+    /// Returns user ID of the process
+    #[inline(always)]
+    pub fn uid(&self) -> UserId {
+        self.ioctx.as_ref().unwrap().uid
+    }
+
+    /// Returns group ID of the process
+    #[inline(always)]
+    pub fn gid(&self) -> GroupId {
+        self.ioctx.as_ref().unwrap().gid
+    }
+
+    /// Changes (if permitted) user ID of the process
+    #[inline(always)]
+    pub fn set_uid(&mut self, uid: UserId) -> Result<(), Errno> {
+        let old_uid = self.uid();
+        if old_uid == uid {
+            Ok(())
+        } else if !old_uid.is_root() {
+            Err(Errno::PermissionDenied)
+        } else {
+            self.ioctx.as_mut().unwrap().uid = uid;
+            Ok(())
+        }
+    }
+
+    /// Changes (if permitted) group ID of the process
+    #[inline(always)]
+    pub fn set_gid(&mut self, gid: GroupId) -> Result<(), Errno> {
+        let old_gid = self.gid();
+        if old_gid == gid {
+            Ok(())
+        } else if !old_gid.is_root() {
+            Err(Errno::PermissionDenied)
+        } else {
+            self.ioctx.as_mut().unwrap().gid = gid;
+            Ok(())
+        }
+    }
+
+    /// Clones a file descriptor into an available slot or, if specified, requested one
+    pub fn duplicate_file(&mut self, src: FileDescriptor, dst: Option<FileDescriptor>) -> Result<FileDescriptor, Errno> {
+        let file_ref = self.file(src)?;
+        if let Some(dst) = dst {
+            let idx = u32::from(dst);
+            if self.files.get(&idx).is_some() {
+                return Err(Errno::AlreadyExists);
+            }
+
+            self.files.insert(idx, file_ref);
+            Ok(dst)
+        } else {
+            self.place_file(file_ref)
+        }
+    }
+
     /// Returns [File] struct referred to by file descriptor `idx`
     pub fn file(&mut self, fd: FileDescriptor) -> Result<FileRef, Errno> {
         self.files.get(&u32::from(fd)).cloned().ok_or(Errno::InvalidFile)
@@ -54,6 +122,7 @@ impl ProcessIo {
         Self {
             files: BTreeMap::new(),
             ioctx: None,
+            ctty: None,
         }
     }
 

kernel/src/proc/mod.rs

@@ -2,12 +2,21 @@
 
 use crate::init;
 use crate::sync::IrqSafeSpinLock;
-use alloc::collections::BTreeMap;
+use crate::mem;
+use alloc::{
+    boxed::Box,
+    collections::{BTreeMap},
+};
+use core::sync::atomic::{AtomicUsize, Ordering};
+use crate::arch::platform::cpu::{self, Cpu};
 use libsys::proc::Pid;
 
 pub mod elf;
+pub mod thread;
+pub use thread::{Thread, ThreadRef, State as ThreadState};
+pub(self) use thread::Context;
 pub mod process;
-pub use process::{Process, ProcessRef, State as ProcessState};
+pub use process::{Process, ProcessRef, ProcessState};
 pub mod io;
 pub use io::ProcessIo;
 
@@ -15,8 +24,10 @@ pub mod wait;
 
 pub mod sched;
 pub use sched::Scheduler;
-pub(self) use sched::SCHED;
+//pub(self) use sched::SCHED;
 
+//<<<<<<< HEAD
+// <<<<<<< HEAD
 // macro_rules! spawn {
 //     (fn ($dst_arg:ident : usize) $body:block, $src_arg:expr) => {{
 //         #[inline(never)]
@@ -36,31 +47,68 @@ pub(self) use sched::SCHED;
 //     (fn () $body:block) => (spawn!(fn (_arg: usize) $body, 0usize))
 // }
 
-/// Performs a task switch.
-///
-/// See [Scheduler::switch]
-pub fn switch() {
-    SCHED.switch(false);
+///// Performs a task switch.
+/////
+///// See [Scheduler::switch]
+//pub fn switch() {
+//    SCHED.switch(false);
+//}
+
+// ///
+// pub fn process(id: Pid) -> ProcessRef {
+//     PROCESSES.lock().get(&id).unwrap().clone()
+// }
+
+macro_rules! spawn {
+    (fn ($dst_arg:ident : usize) $body:block, $src_arg:expr) => {{
+        #[inline(never)]
+        extern "C" fn __inner_func($dst_arg : usize) -> ! {
+            let __res = $body;
+            {
+                todo!();
+                // #![allow(unreachable_code)]
+                // SCHED.current_process().exit(__res);
+                panic!();
+            }
+        }
+
+        let __proc = $crate::proc::Process::new_kernel(__inner_func, $src_arg).unwrap();
+        $crate::proc::sched::enqueue(__proc.id());
+    }};
+
+    (fn () $body:block) => (spawn!(fn (_arg: usize) $body, 0usize))
 }
 
-///
-pub fn process(id: Pid) -> ProcessRef {
-    PROCESSES.lock().get(&id).unwrap().clone()
-}
+// /// Global list of all processes in the system
+// // =======
+// /// Performs a task switch.
+// ///
+// /// See [Scheduler::switch]
+// pub fn switch() {
+//     SCHED.switch(false);
+// }
 
-/// Global list of all processes in the system
+// >>>>>>> feat/thread
 pub(self) static PROCESSES: IrqSafeSpinLock<BTreeMap<Pid, ProcessRef>> =
     IrqSafeSpinLock::new(BTreeMap::new());
 
-/// Sets up initial process and enters it.
-///
-/// See [Scheduler::enter]
-///
-/// # Safety
-///
-/// Unsafe: May only be called once.
-pub unsafe fn enter() -> ! {
-    SCHED.init();
-    SCHED.enqueue(Process::new_kernel(init::init_fn, 0).unwrap().id());
-    SCHED.enter();
+pub(self) static THREADS: IrqSafeSpinLock<BTreeMap<u32, ThreadRef>> =
+    IrqSafeSpinLock::new(BTreeMap::new());
+
+pub unsafe fn enter(is_bsp: bool) -> ! {
+    static COUNTER: AtomicUsize = AtomicUsize::new(0);
+
+    let sched = Cpu::get().scheduler();
+    sched.init();
+
+    COUNTER.fetch_add(1, Ordering::Release);
+    while COUNTER.load(Ordering::Acquire) != cpu::count() {
+        cortex_a::asm::nop();
+    }
+
+    if is_bsp {
+        Process::new_kernel(init::init_fn, 0).unwrap().enqueue();
+    }
+
+    sched.enter();
 }

kernel/src/proc/process.rs

@@ -5,50 +5,49 @@ use crate::mem::{
     phys::{self, PageUsage},
     virt::{MapAttributes, Space},
 };
-use crate::proc::{wait::Wait, ProcessIo, PROCESSES, SCHED};
+use crate::proc::{
+    wait::Wait, Context, ProcessIo, Thread, ThreadRef, ThreadState, PROCESSES, sched,
+};
 use crate::sync::IrqSafeSpinLock;
-use alloc::rc::Rc;
-use core::cell::UnsafeCell;
+use alloc::{rc::Rc, vec::Vec};
 use core::sync::atomic::{AtomicU32, Ordering};
-use libsys::{error::Errno, signal::Signal, proc::{ExitCode, Pid}};
-
-pub use crate::arch::platform::context::{self, Context};
+use libsys::{
+    error::Errno,
+    mem::memcpy,
+    proc::{ExitCode, Pid},
+    signal::Signal,
+    ProgramArgs,
+};
 
 /// Wrapper type for a process struct reference
 pub type ProcessRef = Rc<Process>;
 
 /// List of possible process states
 #[derive(Clone, Copy, Debug, PartialEq)]
-pub enum State {
-    /// Process is ready to be executed and/or is scheduled for it
-    Ready,
-    /// Process is currently running or is in system call/interrupt handler
-    Running,
+pub enum ProcessState {
+    /// Process is alive
+    Active,
     /// Process has finished execution and is waiting to be reaped
     Finished,
-    /// Process is waiting for some external event
-    Waiting,
 }
 
 struct ProcessInner {
     space: Option<&'static mut Space>,
-    state: State,
+    state: ProcessState,
     id: Pid,
-    wait_flag: bool,
+    pgid: Pid,
+    ppid: Option<Pid>,
+    sid: Pid,
     exit: Option<ExitCode>,
-    signal_entry: usize,
-    signal_stack: usize,
+    threads: Vec<u32>,
 }
 
 /// Structure describing an operating system process
 #[allow(dead_code)]
 pub struct Process {
-    ctx: UnsafeCell<Context>,
-    signal_ctx: UnsafeCell<Context>,
     inner: IrqSafeSpinLock<ProcessInner>,
     exit_wait: Wait,
     signal_state: AtomicU32,
-    signal_pending: AtomicU32,
     /// Process I/O context
     pub io: IrqSafeSpinLock<ProcessIo>,
 }
@@ -57,9 +56,89 @@ impl Process {
     const USTACK_VIRT_TOP: usize = 0x100000000;
     const USTACK_PAGES: usize = 4;
 
-    /// Returns currently executing process
+    /// Returns the process ID
+    #[inline]
+    pub fn id(&self) -> Pid {
+        self.inner.lock().id
+    }
+
+    /// Returns the process session ID
+    #[inline]
+    pub fn sid(&self) -> Pid {
+        self.inner.lock().sid
+    }
+
+    /// Returns parent's [Pid]
+    #[inline]
+    pub fn pgid(&self) -> Pid {
+        self.inner.lock().pgid
+    }
+
+    /// Returns parent's [Pid]
+    #[inline]
+    pub fn ppid(&self) -> Option<Pid> {
+        self.inner.lock().ppid
+    }
+
+    /// Sets a new group id for the process
+    pub fn set_pgid(&self, pgid: Pid) {
+        self.inner.lock().pgid = pgid;
+    }
+
+    /// Sets a new session id for the process
+    pub fn set_sid(&self, sid: Pid) {
+        self.inner.lock().sid = sid;
+    }
+
+    /// Returns [Rc]-reference to current process
+    #[inline]
     pub fn current() -> ProcessRef {
-        SCHED.current_process()
+        Thread::current().owner().unwrap()
+    }
+
+    /// Executes a closure performing manipulations on the process address space
+    #[inline]
+    pub fn manipulate_space<R, F>(&self, f: F) -> R
+    where
+        F: FnOnce(&mut Space) -> R,
+    {
+        f(self.inner.lock().space.as_mut().unwrap())
+    }
+
+    /// Creates a new kernel process
+    pub fn new_kernel(entry: extern "C" fn(usize) -> !, arg: usize) -> Result<ProcessRef, Errno> {
+        let id = new_kernel_pid();
+        let thread = Thread::new_kernel(Some(id), entry, arg)?;
+        let mut inner = ProcessInner {
+            threads: Vec::new(),
+            id,
+            pgid: id,
+            ppid: None,
+            sid: id,
+            exit: None,
+            space: None,
+            state: ProcessState::Active,
+        };
+        inner.threads.push(thread.id());
+
+        let res = Rc::new(Self {
+            exit_wait: Wait::new("process_exit"),
+            io: IrqSafeSpinLock::new(ProcessIo::new()),
+            signal_state: AtomicU32::new(0),
+            inner: IrqSafeSpinLock::new(inner),
+        });
+        debugln!("New kernel process: {:?}", id);
+        let prev = PROCESSES.lock().insert(id, res.clone());
+        assert!(prev.is_none());
+        Ok(res)
+    }
+
+    /// Adds all of the process threads to scheduler queue
+    pub fn enqueue(&self) {
+        let inner = self.inner.lock();
+        for &tid in inner.threads.iter() {
+            sched::enqueue(tid);
+        }
     }
 
     /// Returns process (if any) to which `pid` refers
@@ -67,204 +146,149 @@ impl Process {
         PROCESSES.lock().get(&pid).cloned()
     }
 
+// <<<<<<< HEAD
+    // /// Schedules an initial thread for execution
+    // ///
+    // /// # Safety
+    // ///
+    // /// Unsafe: only allowed to be called once, repeated calls
+    // ///         will generate undefined behavior
+    // pub unsafe fn enter(cpu: u32, proc: ProcessRef) -> ! {
+    //     // FIXME use some global lock to guarantee atomicity of thread entry?
+    //     proc.inner.lock().state = State::Running;
+    //     proc.cpu.store(cpu, Ordering::SeqCst);
+    //     let ctx = proc.ctx.get();
+
+    //     // I don't think this is bad: process can't be dropped fully unless
+    //     // it's been reaped (and this function won't run for such process)
+    //     // drop(proc);
+    //     (&mut *ctx).enter()
+    // }
+// =======
     /// Sets a pending signal for a process
     pub fn set_signal(&self, signal: Signal) {
-        let lock = self.inner.lock();
+        todo!();
+        // let mut lock = self.inner.lock();
+        // let ttbr0 = lock.space.as_mut().unwrap().address_phys() | ((lock.id.asid() as usize) << 48);
+        // let main_thread = Thread::get(lock.threads[0]).unwrap();
+        // drop(lock);
 
-        match lock.state {
-            State::Running => {
-                drop(lock);
-                self.enter_signal(signal);
-            }
-            State::Waiting => {
-                // TODO abort whatever the process is waiting for
-                todo!()
-            }
-            State::Ready => {
-                todo!()
-            }
-            State::Finished => {
-                // TODO report error back
-                todo!()
-            }
-        }
+        // // TODO check that `signal` is not a fault signal
+        // //      it is illegal to call this function with
+        // //      fault signals
+        // match main_thread.state() {
+        //     ThreadState::Running => {
+        //         main_thread.enter_signal(signal, ttbr0);
+        //     }
+        //     ThreadState::Waiting => {
+        //         main_thread.clone().setup_signal(signal, ttbr0);
+        //         main_thread.interrupt_wait(true);
+        //     }
+        //     ThreadState::Ready => {
+        //         main_thread.clone().setup_signal(signal, ttbr0);
+        //         main_thread.interrupt_wait(false);
+        //     }
+        //     ThreadState::Finished => {
+        //         // TODO report error back
+        //         todo!()
+        //     }
+        // }
     }
 
-    /// Switches current thread back from signal handler
-    pub fn return_from_signal(&self) {
-        if self.signal_pending.load(Ordering::Acquire) == 0 {
-            panic!("TODO handle cases when returning from no signal");
-        }
-        self.signal_pending.store(0, Ordering::Release);
-
-        let src_ctx = self.signal_ctx.get();
-        let dst_ctx = self.ctx.get();
-
-        assert_eq!(self.inner.lock().state, State::Running);
-
-        unsafe {
-            (&mut *src_ctx).switch(&mut *dst_ctx);
-        }
+    /// Immediately delivers a signal to requested thread
+    pub fn enter_fault_signal(&self, thread: ThreadRef, signal: Signal) {
+        todo!();
+        // let mut lock = self.inner.lock();
+        // let ttbr0 = lock.space.as_mut().unwrap().address_phys() | ((lock.id.asid() as usize) << 48);
+        // thread.enter_signal(signal, ttbr0);
     }
 
-    /// Switches current thread to a signal handler
-    pub fn enter_signal(&self, signal: Signal) {
-        if self
-            .signal_pending
-            .compare_exchange_weak(0, signal as u32, Ordering::SeqCst, Ordering::Relaxed)
-            .is_err()
-        {
-            panic!("Already handling a signal (maybe handle this case)");
-        }
+    // /// Schedules a next thread for execution
+    // ///
+    // /// # Safety
+    // ///
+    // /// Unsafe:
+    // ///
+    // /// * Does not ensure src and dst threads are not the same thread
+    // /// * Does not ensure src is actually current context
+    // pub unsafe fn switch(cpu: u32, src: ProcessRef, dst: ProcessRef, discard: bool) {
+    //     {
+    //         let mut src_lock = src.inner.lock();
+    //         let mut dst_lock = dst.inner.lock();
 
+    //         if !discard {
+    //             assert_eq!(src_lock.state, State::Running);
+    //             src_lock.state = State::Ready;
+    //         }
+    //         assert!(dst_lock.state == State::Ready || dst_lock.state == State::Waiting);
+    //         dst_lock.state = State::Running;
+
+    //         src.cpu.store(Self::CPU_NONE, Ordering::SeqCst);
+    //         dst.cpu.store(cpu, Ordering::SeqCst);
+    //     }
+
+    //     let src_ctx = src.ctx.get();
+    //     let dst_ctx = dst.ctx.get();
+
+    //     // See "drop" note in Process::enter()
+    //     // drop(src);
+    //     // drop(dst);
+
+    //     (&mut *src_ctx).switch(&mut *dst_ctx);
+    // }
+
+    // /// Suspends current process with a "waiting" status
+    // pub fn enter_wait(&self) {
+    //     let drop = {
+    //         let mut lock = self.inner.lock();
+    //         let drop = lock.state == State::Running;
+    //         lock.state = State::Waiting;
+    //         sched::dequeue(lock.id);
+    //         // SCHED.dequeue(lock.id);
+    //         drop
+    //     };
+    //     if drop {
+    //         sched::switch(true);
+    //         // todo!();
+    //         // SCHED.switch(true);
+    //     }
+    // }
+
+    /// Crates a new thread in the process
+    pub fn new_user_thread(&self, entry: usize, stack: usize, arg: usize) -> Result<u32, Errno> {
         let mut lock = self.inner.lock();
-        let signal_ctx = unsafe { &mut *self.signal_ctx.get() };
 
-        let dst_id = lock.id;
-        let dst_space_phys = lock.space.as_mut().unwrap().address_phys();
-        let dst_ttbr0 = dst_space_phys | ((dst_id.asid() as usize) << 48);
+        let space_phys = lock.space.as_mut().unwrap().address_phys();
+        let ttbr0 = space_phys | ((lock.id.asid() as usize) << 48);
 
-        debugln!(
-            "Signal entry: pc={:#x}, sp={:#x}, ttbr0={:#x}",
-            lock.signal_entry,
-            lock.signal_stack,
-            dst_ttbr0
-        );
-        assert_eq!(lock.state, State::Running);
+        let thread = Thread::new_user(lock.id, entry, stack, arg, ttbr0)?;
+        let tid = thread.id();
+        lock.threads.push(tid);
+        sched::enqueue(tid);
 
-        unsafe {
-            signal_ctx.setup_signal_entry(
-                lock.signal_entry,
-                signal as usize,
-                dst_ttbr0,
-                lock.signal_stack,
-            );
-        }
-        let src_ctx = self.ctx.get();
-        drop(lock);
-
-        unsafe {
-            (&mut *src_ctx).switch(signal_ctx);
-        }
+        Ok(tid)
     }
 
-    /// Sets up values needed for signal entry
-    pub fn setup_signal_context(&self, entry: usize, stack: usize) {
-        let mut lock = self.inner.lock();
-        lock.signal_entry = entry;
-        lock.signal_stack = stack;
-    }
-
-    /// Schedules an initial thread for execution
-    ///
-    /// # Safety
-    ///
-    /// Unsafe: only allowed to be called once, repeated calls
-    ///         will generate undefined behavior
-    pub unsafe fn enter(proc: ProcessRef) -> ! {
-        // FIXME use some global lock to guarantee atomicity of thread entry?
-        proc.inner.lock().state = State::Running;
-        proc.current_context().enter()
-    }
-
-    /// Executes a function allowing mutation of the process address space
-    #[inline]
-    pub fn manipulate_space<F: FnOnce(&mut Space) -> Result<(), Errno>>(
-        &self,
-        f: F,
-    ) -> Result<(), Errno> {
-        f(self.inner.lock().space.as_mut().unwrap())
-    }
-
-    #[allow(clippy::mut_from_ref)]
-    fn current_context(&self) -> &mut Context {
-        if self.signal_pending.load(Ordering::Acquire) != 0 {
-            unsafe { &mut *self.signal_ctx.get() }
-        } else {
-            unsafe { &mut *self.ctx.get() }
-        }
-    }
-
-    /// Schedules a next thread for execution
-    ///
-    /// # Safety
-    ///
-    /// Unsafe:
-    ///
-    /// * Does not ensure src and dst threads are not the same thread
-    /// * Does not ensure src is actually current context
-    pub unsafe fn switch(src: ProcessRef, dst: ProcessRef, discard: bool) {
-        {
-            let mut src_lock = src.inner.lock();
-            let mut dst_lock = dst.inner.lock();
-
-            if !discard {
-                assert_eq!(src_lock.state, State::Running);
-                src_lock.state = State::Ready;
-            }
-            assert!(dst_lock.state == State::Ready || dst_lock.state == State::Waiting);
-            dst_lock.state = State::Running;
-        }
-
-        let src_ctx = src.current_context();
-        let dst_ctx = dst.current_context();
-
-        (&mut *src_ctx).switch(&mut *dst_ctx);
-    }
-
-    /// Suspends current process with a "waiting" status
-    pub fn enter_wait(&self) {
-        let drop = {
-            let mut lock = self.inner.lock();
-            let drop = lock.state == State::Running;
-            lock.state = State::Waiting;
-            SCHED.dequeue(lock.id);
-            drop
-        };
-        if drop {
-            SCHED.switch(true);
-        }
-    }
-
-    /// Changes process wait condition status
-    pub fn set_wait_flag(&self, v: bool) {
-        self.inner.lock().wait_flag = v;
-    }
-
-    /// Returns `true` if process wait condition has not been reached
-    pub fn wait_flag(&self) -> bool {
-        self.inner.lock().wait_flag
-    }
-
-    /// Returns the process ID
-    pub fn id(&self) -> Pid {
-        self.inner.lock().id
-    }
-
-    /// Creates a new kernel process
-    pub fn new_kernel(entry: extern "C" fn(usize) -> !, arg: usize) -> Result<ProcessRef, Errno> {
-        let id = new_kernel_pid();
-        let res = Rc::new(Self {
-            ctx: UnsafeCell::new(Context::kernel(entry as usize, arg)),
-            signal_ctx: UnsafeCell::new(Context::empty()),
-            io: IrqSafeSpinLock::new(ProcessIo::new()),
-            exit_wait: Wait::new(),
-            signal_state: AtomicU32::new(0),
-            signal_pending: AtomicU32::new(0),
-            inner: IrqSafeSpinLock::new(ProcessInner {
-                signal_entry: 0,
-                signal_stack: 0,
-                id,
-                exit: None,
-                space: None,
-                wait_flag: false,
-                state: State::Ready,
-            }),
-        });
-        debugln!("New kernel process: {:?}", id);
-        assert!(PROCESSES.lock().insert(id, res.clone()).is_none());
-        Ok(res)
-    }
+    // /// Creates a new kernel process
+    // pub fn new_kernel(entry: extern "C" fn(usize) -> !, arg: usize) -> Result<ProcessRef, Errno> {
+    //     let id = Pid::new_kernel();
+    //     let res = Rc::new(Self {
+    //         ctx: UnsafeCell::new(Context::kernel(entry as usize, arg)),
+    //         io: IrqSafeSpinLock::new(ProcessIo::new()),
+    //         exit_wait: Wait::new(),
+    //         inner: IrqSafeSpinLock::new(ProcessInner {
+    //             id,
+    //             exit: None,
+    //             space: None,
+    //             wait_flag: false,
+    //             state: State::Ready,
+    //         }),
+    //         cpu: AtomicU32::new(Self::CPU_NONE),
+    //     });
+    //     debugln!("New kernel process: {}", id);
+    //     assert!(PROCESSES.lock().insert(id, res.clone()).is_none());
+    //     Ok(res)
+    // }
 
     /// Creates a "fork" of the process, cloning its address space and
     /// resources
@@ -277,63 +301,114 @@ impl Process {
         let dst_space_phys = (dst_space as *mut _ as usize) - mem::KERNEL_OFFSET;
         let dst_ttbr0 = dst_space_phys | ((dst_id.asid() as usize) << 48);
 
+        let mut threads = Vec::new();
+        let tid = Thread::fork(Some(dst_id), frame, dst_ttbr0)?.id();
+        threads.push(tid);
+
         let dst = Rc::new(Self {
-            ctx: UnsafeCell::new(Context::fork(frame, dst_ttbr0)),
-            signal_ctx: UnsafeCell::new(Context::empty()),
+            exit_wait: Wait::new("process_exit"),
             io: IrqSafeSpinLock::new(src_io.fork()?),
-            exit_wait: Wait::new(),
             signal_state: AtomicU32::new(0),
-            signal_pending: AtomicU32::new(0),
             inner: IrqSafeSpinLock::new(ProcessInner {
-                signal_entry: 0,
-                signal_stack: 0,
-                id: dst_id,
+                threads,
                 exit: None,
                 space: Some(dst_space),
-                state: State::Ready,
-                wait_flag: false,
+                state: ProcessState::Active,
+                id: dst_id,
+                pgid: src_inner.pgid,
+                ppid: Some(src_inner.id),
+                sid: src_inner.sid,
             }),
         });
+
         debugln!("Process {:?} forked into {:?}", src_inner.id, dst_id);
         assert!(PROCESSES.lock().insert(dst_id, dst).is_none());
-        SCHED.enqueue(dst_id);
+
+        sched::enqueue(tid);
+        // SCHED.enqueue(dst_id);
 
         Ok(dst_id)
     }
 
     /// Terminates a process.
-    pub fn exit<I: Into<ExitCode>>(&self, status: I) {
-        let status = status.into();
-        let drop = {
-            let mut lock = self.inner.lock();
-            let drop = lock.state == State::Running;
-            infoln!("Process {:?} is exiting: {:?}", lock.id, status);
-            assert!(lock.exit.is_none());
-            lock.exit = Some(status);
-            lock.state = State::Finished;
+    pub fn exit(self: ProcessRef, status: ExitCode) {
+        let thread = Thread::current();
+        let mut lock = self.inner.lock();
+        let is_running = thread.owner_id().map(|e| e == lock.id).unwrap_or(false);
 
-            if let Some(space) = lock.space.take() {
-                unsafe {
-                    Space::release(space);
-                    asm!("tlbi aside1, {}", in(reg) ((lock.id.asid() as usize) << 48));
-                }
+        infoln!("Process {:?} is exiting: {:?}", lock.id, status);
+        assert!(lock.exit.is_none());
+        lock.exit = Some(status);
+        lock.state = ProcessState::Finished;
+
+        for &tid in lock.threads.iter() {
+            Thread::get(tid).unwrap().terminate(status);
+            sched::dequeue(tid);
+            // SCHED.dequeue(tid);
+        }
+
+        if let Some(space) = lock.space.take() {
+            unsafe {
+                Space::release(space);
+                asm!("tlbi aside1, {}", in(reg) ((lock.id.asid() as usize) << 48));
+            }
+        }
+
+        // TODO when exiting from signal handler interrupting an IO operation
+        //      deadlock is achieved
+        self.io.lock().handle_exit();
+
+        drop(lock);
+
+        self.exit_wait.wakeup_all();
+
+        if is_running {
+            sched::switch(true);
+            panic!("This code should never run");
+        }
+    }
+
+    /// Terminates a thread of the process. If the thread is the only
+    /// one remaining, process itself is exited (see [Process::exit])
+    pub fn exit_thread(thread: ThreadRef, status: ExitCode) {
+        let switch = {
+            let switch = thread.state() == ThreadState::Running;
+            let process = thread.owner().unwrap();
+            let mut lock = process.inner.lock();
+            let tid = thread.id();
+
+            if lock.threads.len() == 1 {
+                // TODO call Process::exit instead?
+                drop(lock);
+                process.exit(status);
+                return;
             }
 
-            self.io.lock().handle_exit();
+            lock.threads.retain(|&e| e != tid);
 
-            SCHED.dequeue(lock.id);
-            drop
+            thread.terminate(status);
+            todo!();
+            // SCHED.dequeue(tid);
+            debugln!("Thread {} terminated", tid);
+
+            switch
         };
-        self.exit_wait.wakeup_all();
-        if drop {
-            SCHED.switch(true);
-            panic!("This code should never run");
+
+        if switch {
+            // TODO retain thread ID in process "finished" list and
+            //      drop it when process finishes
+            // SCHED.switch(true);
+            todo!();
+            panic!("This code should not run");
+        } else {
+            // Can drop this thread: it's not running
+            todo!();
         }
     }
 
     fn collect(&self) -> Option<ExitCode> {
         let lock = self.inner.lock();
-        if lock.state == State::Finished {
+        if lock.state == ProcessState::Finished {
             lock.exit
         } else {
             None
@@ -352,7 +427,6 @@ impl Process {
             if let Some(r) = proc.collect() {
                 // TODO drop the process struct itself
                 PROCESSES.lock().remove(&proc.id());
-                debugln!("pid {:?} has {} refs", proc.id(), Rc::strong_count(&proc));
                 return Ok(r);
             }
 
@@ -360,42 +434,160 @@ impl Process {
         }
     }
 
-    /// Loads a new program into current process address space
-    pub fn execve<F: FnOnce(&mut Space) -> Result<usize, Errno>>(
-        loader: F,
-        arg: usize,
-    ) -> Result<(), Errno> {
-        unsafe {
-            // Run with interrupts disabled
-            asm!("msr daifset, #2");
+    fn write_paged<T>(space: &mut Space, dst: usize, src: T) -> Result<(), Errno> {
+        let size = core::mem::size_of::<T>();
+        if (size + (dst % 4096)) > 4096 {
+            todo!("Object crossed page boundary");
         }
 
-        let proc = SCHED.current_process();
-        let mut lock = proc.inner.lock();
-        if lock.id.is_kernel() {
-            let mut proc_lock = PROCESSES.lock();
-            let old_pid = lock.id;
-            assert!(
-                proc_lock.remove(&old_pid).is_some(),
-                "Failed to downgrade kernel process (remove kernel pid)"
-            );
-            lock.id = new_user_pid();
-            debugln!(
-                "Process downgrades from kernel to user: {:?} -> {:?}",
-                old_pid,
-                lock.id
-            );
-            assert!(proc_lock.insert(lock.id, proc.clone()).is_none());
-            unsafe {
-                SCHED.hack_current_pid(lock.id);
-            }
+        let page_virt = dst & !4095;
+        let page_phys = if let Ok(phys) = space.translate(dst) {
+            phys
         } else {
-            // Invalidate user ASID
-            let input = (lock.id.asid() as usize) << 48;
-            unsafe {
-                asm!("tlbi aside1, {}", in(reg) input);
-            }
+            let page = phys::alloc_page(PageUsage::UserPrivate)?;
+            let flags = MapAttributes::SH_OUTER
+                | MapAttributes::NOT_GLOBAL
+                | MapAttributes::UXN
+                | MapAttributes::PXN
+                | MapAttributes::AP_BOTH_READONLY;
+            space.map(page_virt, page, flags)?;
+            page
+        };
+
+        unsafe {
+            core::ptr::write((mem::virtualize(page_phys) + (dst % 4096)) as *mut T, src);
         }
+        Ok(())
+    }
+
+    fn write_paged_bytes(space: &mut Space, dst: usize, src: &[u8]) -> Result<(), Errno> {
+        if (src.len() + (dst % 4096)) > 4096 {
+            todo!("Object crossed page boundary");
+        }
+        let page_virt = dst & !4095;
+        let page_phys = if let Ok(phys) = space.translate(dst) {
+            phys
+        } else {
+            let page = phys::alloc_page(PageUsage::UserPrivate)?;
+            let flags = MapAttributes::SH_OUTER
+                | MapAttributes::NOT_GLOBAL
+                | MapAttributes::UXN
+                | MapAttributes::PXN
+                | MapAttributes::AP_BOTH_READONLY;
+            space.map(page_virt, page, flags)?;
+            page
+        };
+
+        unsafe {
+            memcpy(
+                (mem::virtualize(page_phys) + (dst % 4096)) as *mut u8,
+                src.as_ptr(),
+                src.len(),
+            );
+        }
+        Ok(())
+    }
+
+    fn store_arguments(space: &mut Space, argv: &[&str]) -> Result<usize, Errno> {
+        let mut offset = 0usize;
+        // TODO vmalloc?
+        let base = 0x60000000;
+
+        // 1. Store program argument string bytes
+        for arg in argv.iter() {
+            Self::write_paged_bytes(space, base + offset, arg.as_bytes())?;
+            offset += arg.len();
+        }
+        // Align
+        offset = (offset + 15) & !15;
+        let argv_offset = offset;
+
+        // 2. Store arg pointers
+        let mut data_offset = 0usize;
+        for arg in argv.iter() {
+            // XXX this is really unsafe and I am not really sure ABI will stay like this XXX
+            Self::write_paged(space, base + offset + 0, base + data_offset)?;
+            Self::write_paged(space, base + offset + 8, arg.len())?;
+            offset += 16;
+            data_offset += arg.len();
+        }
+
+        // 3. Store ProgramArgs
+        let data = ProgramArgs {
+            argc: argv.len(),
+            argv: base + argv_offset,
+            storage: base,
+            size: offset + core::mem::size_of::<ProgramArgs>(),
+        };
+        Self::write_paged(space, base + offset, data)?;
+
+        Ok(base + offset)
+    }
+
+   /// Loads a new program into current process address space
+   pub fn execve<F: FnOnce(&mut Space) -> Result<usize, Errno>>(
+       loader: F,
+       argv: &[&str],
+   ) -> Result<(), Errno> {
+       unsafe {
+           // Run with interrupts disabled
+           asm!("msr daifset, #2");
+       }
+
+// <<<<<<< HEAD
+//         let proc = sched::current_process();
+//         let mut lock = proc.inner.lock();
+//         if lock.id.is_kernel() {
+//             let mut proc_lock = PROCESSES.lock();
+//             let old_pid = lock.id;
+//             assert!(
+//                 proc_lock.remove(&old_pid).is_some(),
+//                 "Failed to downgrade kernel process (remove kernel pid)"
+//             );
+//             lock.id = Pid::new_user();
+//             debugln!(
+//                 "Process downgrades from kernel to user: {} -> {}",
+//                 old_pid,
+//                 lock.id
+//             );
+//             assert!(proc_lock.insert(lock.id, proc.clone()).is_none());
+//             unsafe {
+//                 use crate::arch::platform::cpu::Cpu;
+//                 Cpu::get().scheduler().hack_current_pid(lock.id);
+//             }
+//         } else {
+//             // Invalidate user ASID
+//             let input = (lock.id.asid() as usize) << 48;
+//             unsafe {
+//                 asm!("tlbi aside1, {}", in(reg) input);
+//             }
+// =======
+        let proc = Process::current();
+        let mut process_lock = proc.inner.lock();
+
+        if process_lock.threads.len() != 1 {
+            todo!();
+// >>>>>>> feat/thread
+        }
+
+        let thread = Thread::get(process_lock.threads[0]).unwrap();
+
+        if process_lock.id.is_kernel() {
+            let mut processes = PROCESSES.lock();
+            let old_pid = process_lock.id;
+            let new_pid = new_user_pid();
+            debugln!("Downgrading process {:?} -> {:?}", old_pid, new_pid);
+
+            let r = processes.remove(&old_pid);
+            assert!(r.is_some());
+            process_lock.id = new_pid;
+            process_lock.pgid = new_pid;
+            process_lock.sid = new_pid;
+            let r = processes.insert(new_pid, proc.clone());
+            assert!(r.is_none());
+        }
+
+        thread.set_owner(process_lock.id);
 
         proc.io.lock().handle_cloexec();
 
@@ -416,25 +608,26 @@ impl Process {
         }
 
         let entry = loader(new_space)?;
+        let arg = Self::store_arguments(new_space, argv)?;
 
         debugln!("Will now enter at {:#x}", entry);
         // TODO drop old address space
-        lock.space = Some(new_space);
+        process_lock.space = Some(new_space);
 
         unsafe {
             // TODO drop old context
-            let ctx = proc.ctx.get();
+            let ctx = thread.ctx.get();
+            let asid = (process_lock.id.asid() as usize) << 48;
+            asm!("tlbi aside1, {}", in(reg) asid);
 
             ctx.write(Context::user(
                 entry,
                 arg,
-                new_space_phys | ((lock.id.asid() as usize) << 48),
+                new_space_phys | asid,
                 Self::USTACK_VIRT_TOP,
             ));
 
-            assert_eq!(lock.state, State::Running);
-
-            drop(lock);
+            drop(process_lock);
 
             (*ctx).enter();
         }

kernel/src/proc/sched.rs

@@ -1,13 +1,17 @@
 //!
-use crate::proc::{Pid, Process, ProcessRef, PROCESSES};
-use crate::sync::IrqSafeSpinLock;
+use crate::proc::{Thread, ThreadRef, THREADS};
 use crate::util::InitOnce;
 use alloc::{collections::VecDeque, rc::Rc};
+use crate::sync::{IrqSafeSpinLock, IrqSafeSpinLockGuard};
+use crate::arch::platform::cpu::{self, Cpu};
+use cortex_a::registers::{MPIDR_EL1, DAIF};
+use core::ops::Deref;
+use tock_registers::interfaces::Readable;
 
 struct SchedulerInner {
-    queue: VecDeque<Pid>,
-    idle: Option<Pid>,
-    current: Option<Pid>,
+    queue: VecDeque<u32>,
+    idle: Option<u32>,
+    current: Option<u32>,
 }
 
 /// Process scheduler state and queues
@@ -23,13 +27,31 @@ impl SchedulerInner {
             current: None,
         };
 
-        this.idle = Some(Process::new_kernel(idle_fn, 0).unwrap().id());
+        this.idle = Some(Thread::new_kernel(None, idle_fn, 0).unwrap().id());
 
         this
     }
 }
 
 impl Scheduler {
+    ///
+    pub const fn new() -> Self {
+        Self {
+            inner: InitOnce::new()
+        }
+    }
+
+    ///
+    pub fn queue_size(&self) -> usize {
+        let lock = self.inner.get().lock();
+        let c = if lock.current.is_some() {
+            1
+        } else {
+            0
+        };
+        lock.queue.len() + c
+    }
+
     /// Initializes inner data structure:
     ///
     /// * idle thread
@@ -39,13 +61,13 @@ impl Scheduler {
     }
 
     /// Schedules a thread for execution
-    pub fn enqueue(&self, pid: Pid) {
-        self.inner.get().lock().queue.push_back(pid);
+    pub fn enqueue(&self, tid: u32) {
+        self.inner.get().lock().queue.push_back(tid);
     }
 
-    /// Removes given `pid` from execution queue
-    pub fn dequeue(&self, pid: Pid) {
-        self.inner.get().lock().queue.retain(|&p| p != pid)
+    /// Removes given `tid` from execution queue
+    pub fn dequeue(&self, tid: u32) {
+        self.inner.get().lock().queue.retain(|&p| p != tid)
     }
 
     /// Performs initial process entry.
@@ -63,11 +85,11 @@ impl Scheduler {
             };
 
             inner.current = Some(id);
-            PROCESSES.lock().get(&id).unwrap().clone()
+            THREADS.lock().get(&id).unwrap().clone()
         };
 
         asm!("msr daifset, #2");
-        Process::enter(thread)
+        Thread::enter((MPIDR_EL1.get() & 0xF) as u32, thread)
     }
 
     /// This hack is required to be called from execve() when downgrading current
@@ -76,20 +98,32 @@ impl Scheduler {
     /// # Safety
     ///
     /// Unsafe: only allowed to be called from Process::execve()
-    pub unsafe fn hack_current_pid(&self, new: Pid) {
-        self.inner.get().lock().current = Some(new);
+    pub unsafe fn hack_current_tid(&self, old: u32, new: u32) {
+        let mut lock = self.inner.get().lock();
+        match lock.current {
+            Some(t) if t == old => {
+                lock.current = Some(new);
+            }
+            _ => {}
+        }
     }
 
     /// Switches to the next task scheduled for execution. If there're
     /// none present in the queue, switches to the idle task.
-    pub fn switch(&self, discard: bool) {
+    pub fn switch(&self, discard: bool, sched_lock: IrqSafeSpinLockGuard<()>) {
         let (from, to) = {
             let mut inner = self.inner.get().lock();
             let current = inner.current.unwrap();
 
-            if !discard && current != Pid::IDLE {
+//<<<<<<< HEAD
+            if !discard && current != inner.idle.unwrap() {
+//=======
+//            if !discard && current != 0 {
+//>>>>>>> feat/thread
                 // Put the process into the back of the queue
-                inner.queue.push_back(current);
+                if !enqueue_somewhere_else((MPIDR_EL1.get() & 0xF) as usize, current, &sched_lock) {
+                    inner.queue.push_back(current);
+                }
             }
 
             let next = if inner.queue.is_empty() {
@@ -100,7 +134,7 @@ impl Scheduler {
 
             inner.current = Some(next);
             let (from, to) = {
-                let lock = PROCESSES.lock();
+                let lock = THREADS.lock();
                 (
                     lock.get(&current).unwrap().clone(),
                     lock.get(&next).unwrap().clone(),
@@ -112,24 +146,42 @@ impl Scheduler {
 
         if !Rc::ptr_eq(&from, &to) {
             unsafe {
+//<<<<<<< HEAD
+                drop(sched_lock);
+//                Process::switch((MPIDR_EL1.get() & 0xF) as u32, from, to, discard);
+//=======
                 asm!("msr daifset, #2");
-                Process::switch(from, to, discard);
+                Thread::switch((MPIDR_EL1.get() & 0xF) as u32, from, to, discard);
+//>>>>>>> feat/thread
             }
         }
     }
 
-    /// Returns a Rc-reference to currently running process
-    pub fn current_process(&self) -> ProcessRef {
+    /// Returns a [Rc]-reference to currently running Thread
+    pub fn current_thread(&self) -> ThreadRef {
         let inner = self.inner.get().lock();
-        let current = inner.current.unwrap();
-        PROCESSES.lock().get(&current).unwrap().clone()
+        let id = inner.current.unwrap();
+        THREADS.lock().get(&id).unwrap().clone()
     }
+
+    // /// Returns a Rc-reference to currently running process
+    // pub fn current_process(&self) -> ProcessRef {
+    //     let inner = self.inner.get().lock();
+    //     let current = inner.current.unwrap();
+    //     PROCESSES.lock().get(&current).unwrap().clone()
+    // }
 }
 
-/// Returns `true` if the scheduler has been initialized
-pub fn is_ready() -> bool {
-    SCHED.inner.is_initialized()
-}
+// <<<<<<< HEAD
+// // pub fn is_ready() -> bool {
+// //     SCHED.inner.is_initialized()
+// // }
+// =======
+// /// Returns `true` if the scheduler has been initialized
+// pub fn is_ready() -> bool {
+//     SCHED.inner.is_initialized()
+// }
+// >>>>>>> feat/thread
 
 #[inline(never)]
 extern "C" fn idle_fn(_a: usize) -> ! {
@@ -138,8 +190,94 @@ extern "C" fn idle_fn(_a: usize) -> ! {
     }
 }
 
+pub fn current_thread() -> ThreadRef {
+    let guard = SCHED_LOCK.lock();
+    unsafe { Cpu::get().scheduler().current_thread() }
+}
+
+/// Performs a task switch.
+///
+/// See [Scheduler::switch]
+pub fn switch(discard: bool) {
+    assert!(DAIF.matches_all(DAIF::I::SET));
+    let guard = SCHED_LOCK.lock();
+    unsafe { Cpu::get().scheduler().switch(discard, guard); }
+}
+
+///
+pub fn enqueue_to(cpu: usize, tid: u32) {
+    todo!()
+    //let _lock = SCHED_LOCK.lock();
+    //debugln!("Queue {} to cpu{}", pid, cpu);
+    //unsafe {
+    //    cpu::by_index(cpu).scheduler().enqueue(pid)
+    //}
+}
+
+///
+pub fn enqueue(tid: u32) {
+    let _lock = SCHED_LOCK.lock();
+    let mut min_idx = 0;
+    let mut min_cnt = usize::MAX;
+    for (i, cpu) in unsafe { cpu::cpus() }.enumerate() {
+        let size = cpu.scheduler().queue_size();
+        if size < min_cnt {
+            min_cnt = size;
+            min_idx = i;
+        }
+    }
+
+    // debugln!("Queue {} to cpu{}", pid, min_idx);
+    unsafe {
+        cpu::by_index(min_idx).scheduler().enqueue(tid)
+    }
+}
+
+///
+pub fn enqueue_somewhere_else(ignore: usize, tid: u32, _guard: &IrqSafeSpinLockGuard<()>) -> bool {
+    let mut min_idx = 0;
+    //let mut min_cnt = usize::MAX;
+    static mut LAST: usize = 0;
+    //for (i, cpu) in unsafe { cpu::cpus() }.enumerate() {
+    //for (i, cpu) in wacky_cpu_iterate() {
+    //    if i == ignore {
+    //        continue;
+    //    }
+    //    let size = cpu.scheduler().queue_size();
+    //    if size < min_cnt {
+    //        min_cnt = size;
+    //        min_idx = i;
+    //    }
+    //}
+    unsafe {
+        LAST = (LAST + 1) % cpu::count();
+        min_idx = LAST;
+    }
+
+    if min_idx == ignore {
+        false
+    } else {
+        unsafe {
+            cpu::by_index(min_idx).scheduler().enqueue(tid)
+        }
+        true
+    }
+}
+
+///
+pub fn dequeue(tid: u32) {
+    // TODO process can be rescheduled to other CPU between scheduler locks
+    let lock = SCHED_LOCK.lock();
+    let cpu_id = Thread::get(tid).unwrap().cpu();
+    unsafe {
+        cpu::by_index(cpu_id as usize).scheduler().dequeue(tid);
+    }
+}
+
+static SCHED_LOCK: IrqSafeSpinLock<()> = IrqSafeSpinLock::new(());
+
 // TODO maybe move this into a per-CPU struct
-/// Global scheduler struct
-pub static SCHED: Scheduler = Scheduler {
-    inner: InitOnce::new(),
-};
+// /// Global scheduler struct
+// pub static SCHED: Scheduler = Scheduler {
+//     inner: InitOnce::new(),
+// };

kernel/src/proc/thread.rs

@@ -0,0 +1,423 @@
+//! Facilities for controlling threads - smallest units of
+//! execution in the operating system
+use crate::arch::aarch64::exception::ExceptionFrame;
+use crate::proc::{
+    wait::{Wait, WaitStatus},
+    Process, ProcessRef, sched, THREADS,
+};
+use crate::sync::IrqSafeSpinLock;
+use crate::util::InitOnce;
+use alloc::rc::Rc;
+use core::cell::UnsafeCell;
+use core::sync::atomic::{AtomicU32, Ordering};
+use libsys::{
+    error::Errno,
+    proc::{ExitCode, Pid},
+    signal::Signal,
+};
+
+pub use crate::arch::platform::context::{self, Context};
+
+/// Convenience wrapper for [Thread] references
+pub type ThreadRef = Rc<Thread>;
+
+/// List of possible process states
+#[derive(Clone, Copy, Debug, PartialEq)]
+pub enum State {
+    /// Process is ready to be executed and/or is scheduled for it
+    Ready,
+    /// Process is currently running or is in system call/interrupt handler
+    Running,
+    /// Process has finished execution and is waiting to be reaped
+    Finished,
+    /// Process is waiting for some external event
+    Waiting,
+}
+
+struct ThreadInner {
+    id: u32,
+    state: State,
+    owner: Option<Pid>,
+    pending_wait: Option<&'static Wait>,
+    wait_status: WaitStatus,
+    signal_entry: usize,
+    signal_stack: usize,
+}
+
+/// Thread control data
+pub struct Thread {
+    inner: IrqSafeSpinLock<ThreadInner>,
+    exit_wait: Wait,
+    exit_status: InitOnce<ExitCode>,
+    pub(super) ctx: UnsafeCell<Context>,
+    signal_ctx: UnsafeCell<Context>,
+    signal_pending: AtomicU32,
+    cpu: AtomicU32,
+}
+
+impl Thread {
+    const CPU_NONE: u32 = u32::MAX;
+
+    /// Returns currently active thread [Rc]-reference
+    #[inline]
+    pub fn current() -> ThreadRef {
+        sched::current_thread()
+    }
+
+    /// Returns a reference to thread `tid`, if it exists
+    #[inline]
+    pub fn get(tid: u32) -> Option<ThreadRef> {
+        THREADS.lock().get(&tid).cloned()
+    }
+
+    /// Returns the owner process
+    #[inline]
+    pub fn owner(&self) -> Option<ProcessRef> {
+        self.inner.lock().owner.and_then(Process::get)
+    }
+
+    /// Returns [Pid] of the owner process
+    pub fn owner_id(&self) -> Option<Pid> {
+        self.inner.lock().owner
+    }
+
+    /// Creates a new kernel process
+    pub fn new_kernel(
+        owner: Option<Pid>,
+        entry: extern "C" fn(usize) -> !,
+        arg: usize,
+    ) -> Result<ThreadRef, Errno> {
+        let id = new_tid();
+
+        let res = Rc::new(Self {
+            cpu: AtomicU32::new(Self::CPU_NONE),
+            ctx: UnsafeCell::new(Context::kernel(entry as usize, arg)),
+            signal_ctx: UnsafeCell::new(Context::empty()),
+            signal_pending: AtomicU32::new(0),
+            exit_wait: Wait::new("thread_exit"),
+            exit_status: InitOnce::new(),
+            inner: IrqSafeSpinLock::new(ThreadInner {
+                signal_entry: 0,
+                signal_stack: 0,
+                id,
+                owner,
+                pending_wait: None,
+                wait_status: WaitStatus::Done,
+                state: State::Ready,
+            }),
+        });
+        debugln!("New kernel thread: {:?}", id);
+        assert!(THREADS.lock().insert(id, res.clone()).is_none());
+        Ok(res)
+    }
+
+    /// Creates a new userspace process
+    pub fn new_user(
+        owner: Pid,
+        entry: usize,
+        stack: usize,
+        arg: usize,
+        ttbr0: usize,
+    ) -> Result<ThreadRef, Errno> {
+        let id = new_tid();
+
+        let res = Rc::new(Self {
+            cpu: AtomicU32::new(Self::CPU_NONE),
+            ctx: UnsafeCell::new(Context::user(entry, arg, ttbr0, stack)),
+            signal_ctx: UnsafeCell::new(Context::empty()),
+            signal_pending: AtomicU32::new(0),
+            exit_wait: Wait::new("thread_exit"),
+            exit_status: InitOnce::new(),
+            inner: IrqSafeSpinLock::new(ThreadInner {
+                signal_entry: 0,
+                signal_stack: 0,
+                id,
+                owner: Some(owner),
+                pending_wait: None,
+                wait_status: WaitStatus::Done,
+                state: State::Ready,
+            }),
+        });
+        debugln!("New userspace thread: {:?}", id);
+        assert!(THREADS.lock().insert(id, res.clone()).is_none());
+        Ok(res)
+    }
+
+    /// Creates a fork thread cloning `frame` context
+    pub fn fork(
+        owner: Option<Pid>,
+        frame: &ExceptionFrame,
+        ttbr0: usize,
+    ) -> Result<ThreadRef, Errno> {
+        let id = new_tid();
+
+        let res = Rc::new(Self {
+            cpu: AtomicU32::new(Self::CPU_NONE),
+            ctx: UnsafeCell::new(Context::fork(frame, ttbr0)),
+            signal_ctx: UnsafeCell::new(Context::empty()),
+            signal_pending: AtomicU32::new(0),
+            exit_wait: Wait::new("thread_exit"),
+            exit_status: InitOnce::new(),
+            inner: IrqSafeSpinLock::new(ThreadInner {
+                signal_entry: 0,
+                signal_stack: 0,
+                id,
+                owner,
+                pending_wait: None,
+                wait_status: WaitStatus::Done,
+                state: State::Ready,
+            }),
+        });
+        debugln!("Forked new user thread: {:?}", id);
+        assert!(THREADS.lock().insert(id, res.clone()).is_none());
+        Ok(res)
+    }
+
+    /// Returns the thread ID
+    #[inline]
+    pub fn id(&self) -> u32 {
+        self.inner.lock().id
+    }
+
+    /// Schedules an initial thread for execution
+    ///
+    /// # Safety
+    ///
+    /// Unsafe: only allowed to be called once, repeated calls
+    ///         will generate undefined behavior
+    pub unsafe fn enter(cpu: u32, thread: ThreadRef) -> ! {
+        // FIXME use some global lock to guarantee atomicity of thread entry?
+        thread.inner.lock().state = State::Running;
+        thread.cpu.store(cpu, Ordering::SeqCst);
+        thread.current_context().enter()
+    }
+
+    ///
+    pub fn cpu(&self) -> u32 {
+        self.cpu.load(Ordering::SeqCst)
+    }
+
+    /// Schedules a next thread for execution
+    ///
+    /// # Safety
+    ///
+    /// Unsafe:
+    ///
+    /// * Does not ensure src and dst threads are not the same thread
+    /// * Does not ensure src is actually current context
+    pub unsafe fn switch(cpu: u32, src: ThreadRef, dst: ThreadRef, discard: bool) {
+        {
+            let mut src_lock = src.inner.lock();
+            let mut dst_lock = dst.inner.lock();
+
+            if !discard {
+                assert_eq!(src_lock.state, State::Running);
+                src_lock.state = State::Ready;
+            }
+            // assert!(dst_lock.state == State::Ready || dst_lock.state == State::Waiting);
+            dst_lock.state = State::Running;
+
+            src.cpu.store(Self::CPU_NONE, Ordering::SeqCst);
+            dst.cpu.store(cpu, Ordering::SeqCst);
+        }
+
+        let src_ctx = src.current_context();
+        let dst_ctx = dst.current_context();
+
+        (&mut *src_ctx).switch(&mut *dst_ctx);
+    }
+
+    #[allow(clippy::mut_from_ref)]
+    fn current_context(&self) -> &mut Context {
+        if self.signal_pending.load(Ordering::Acquire) != 0 {
+            unsafe { &mut *self.signal_ctx.get() }
+        } else {
+            unsafe { &mut *self.ctx.get() }
+        }
+    }
+
+    /// Suspends current process with a "waiting" status
+    pub fn enter_wait(&self) {
+        let drop = {
+            let mut lock = self.inner.lock();
+            let drop = lock.state == State::Running;
+            lock.state = State::Waiting;
+            sched::dequeue(lock.id);
+            // SCHED.dequeue(lock.id);
+            drop
+        };
+        if drop {
+            sched::switch(true);
+            // SCHED.switch(true);
+        }
+    }
+
+    /// Changes process wait condition status
+    pub fn setup_wait(&self, wait: *const Wait) {
+        let mut lock = self.inner.lock();
+        // FIXME this is not cool
+        lock.pending_wait = Some(unsafe { &*wait });
+        lock.wait_status = WaitStatus::Pending;
+    }
+
+    /// Suspends current thread until thread `tid` terminates
+    pub fn waittid(tid: u32) -> Result<(), Errno> {
+        loop {
+            let thread = THREADS
+                .lock()
+                .get(&tid)
+                .cloned()
+                .ok_or(Errno::DoesNotExist)?;
+
+            if thread.state() == State::Finished {
+                // TODO remove thread from its parent?
+                return Ok(());
+            }
+
+            thread.exit_wait.wait(None)?;
+        }
+    }
+
+    /// Updates pending wait status
+    pub fn set_wait_status(&self, status: WaitStatus) {
+        let mut lock = self.inner.lock();
+        lock.wait_status = status;
+    }
+
+    /// Resets wait channel back to initial state
+    pub fn reset_wait(&self) {
+        let mut lock = self.inner.lock();
+        lock.pending_wait = None;
+        lock.wait_status = WaitStatus::Done;
+    }
+
+    /// Returns status of the thread's pending wait
+    pub fn wait_status(&self) -> WaitStatus {
+        self.inner.lock().wait_status
+    }
+
+    /// Switches current thread back from signal handler
+    pub fn return_from_signal(&self) {
+        if self.signal_pending.load(Ordering::Acquire) == 0 {
+            panic!("TODO handle cases when returning from no signal");
+        }
+        self.signal_pending.store(0, Ordering::Release);
+
+        let src_ctx = self.signal_ctx.get();
+        let dst_ctx = self.ctx.get();
+
+        assert_eq!(self.inner.lock().state, State::Running);
+
+        unsafe {
+            (&mut *src_ctx).switch(&mut *dst_ctx);
+        }
+    }
+
+    /// Returns the thread state
+    #[inline]
+    pub fn state(&self) -> State {
+        self.inner.lock().state
+    }
+
+    /// Sets the thread's owner process ID
+    pub fn set_owner(&self, pid: Pid) {
+        self.inner.lock().owner = Some(pid);
+    }
+
+    /// Sets up values needed for signal entry
+    pub fn set_signal_entry(&self, entry: usize, stack: usize) {
+        let mut lock = self.inner.lock();
+        lock.signal_entry = entry;
+        lock.signal_stack = stack;
+    }
+
+    /// Sets up a context for signal handler
+    pub fn setup_signal(self: ThreadRef, signal: Signal, ttbr0: usize) {
+        if self
+            .signal_pending
+            .compare_exchange_weak(0, signal as u32, Ordering::SeqCst, Ordering::Relaxed)
+            .is_err()
+        {
+            panic!("Already handling a signal (maybe handle this case)");
+        }
+
+        let lock = self.inner.lock();
+        if lock.signal_entry == 0 || lock.signal_stack == 0 {
+            drop(lock);
+            Process::exit_thread(self, ExitCode::from(-1));
+            return;
+        }
+
+        let signal_ctx = unsafe { &mut *self.signal_ctx.get() };
+
+        debugln!(
+            "Signal entry: tid={}, pc={:#x}, sp={:#x}, ttbr0={:#x}",
+            lock.id,
+            lock.signal_entry,
+            lock.signal_stack,
+            ttbr0
+        );
+
+        unsafe {
+            signal_ctx.setup_signal_entry(
+                lock.signal_entry,
+                signal as usize,
+                ttbr0,
+                lock.signal_stack,
+            );
+        }
+    }
+
+    /// Switches process main thread to a signal handler
+    pub fn enter_signal(self: ThreadRef, signal: Signal, ttbr0: usize) {
+        let src_ctx = self.ctx.get();
+        let signal_ctx = unsafe { &mut *self.signal_ctx.get() };
+
+        assert_eq!(self.state(), State::Running);
+        self.setup_signal(signal, ttbr0);
+
+        unsafe {
+            (&mut *src_ctx).switch(signal_ctx);
+        }
+    }
+
+    /// Interrupts pending wait (from signal routines)
+    pub fn interrupt_wait(&self, enqueue: bool) {
+        let mut lock = self.inner.lock();
+        let tid = lock.id;
+        let wait = lock.pending_wait.take();
+        drop(lock);
+        if let Some(wait) = wait {
+            wait.abort(tid, enqueue);
+        }
+    }
+
+    /// Cleans up any resources of the thread and aborts
+    /// pending wait, if any
+    pub fn terminate(&self, status: ExitCode) {
+        let mut lock = self.inner.lock();
+        lock.state = State::Finished;
+        let tid = lock.id;
+        let wait = lock.pending_wait.take();
+        drop(lock);
+        if let Some(wait) = wait {
+            wait.abort(tid, false);
+        }
+        self.exit_status.init(status);
+        self.exit_wait.wakeup_all();
+    }
+}
+
+impl Drop for Thread {
+    fn drop(&mut self) {
+        debugln!("Dropping process {:?}", self.id());
+    }
+}
+
+/// Allocates a new thread ID
+pub fn new_tid() -> u32 {
+    static LAST: AtomicU32 = AtomicU32::new(1);
+    let id = LAST.fetch_add(1, Ordering::Relaxed);
+    assert!(id < 256, "Out of user TIDs");
+    id
+}

kernel/src/proc/wait.rs

@@ -2,25 +2,40 @@
 
 use crate::arch::machine;
 use crate::dev::timer::TimestampSource;
-use crate::proc::{self, sched::SCHED, Process, ProcessRef};
+use crate::proc::{self, sched, Thread, ThreadRef};
 use crate::sync::IrqSafeSpinLock;
 use alloc::collections::LinkedList;
 use core::time::Duration;
-use libsys::{error::Errno, stat::FdSet, proc::Pid};
+use libsys::{error::Errno, stat::FdSet};
 
 /// Wait channel structure. Contains a queue of processes
 /// waiting for some event to happen.
 pub struct Wait {
-    queue: IrqSafeSpinLock<LinkedList<Pid>>,
+    queue: IrqSafeSpinLock<LinkedList<u32>>,
+    #[allow(dead_code)]
+    name: &'static str
+}
+
+/// Status of a (possibly) pending wait
+#[derive(PartialEq, Eq, Copy, Clone, Debug)]
+pub enum WaitStatus {
+    /// In progress
+    Pending,
+    /// Wait was interrupted by a signal
+    Interrupted,
+    /// Channel reported data available
+    Done,
 }
 
 struct Timeout {
-    pid: Pid,
+    tid: u32,
     deadline: Duration,
 }
 
 static TICK_LIST: IrqSafeSpinLock<LinkedList<Timeout>> = IrqSafeSpinLock::new(LinkedList::new());
-pub static WAIT_SELECT: Wait = Wait::new();
+/// Global wait channel for blocking on select. Gets notified
+/// of ANY I/O operations available, so not very efficient.
+pub static WAIT_SELECT: Wait = Wait::new("select");
 
 /// Checks for any timed out wait channels and interrupts them
 pub fn tick() {
@@ -30,9 +45,14 @@ pub fn tick() {
 
     while let Some(item) = cursor.current() {
         if time > item.deadline {
-            let pid = item.pid;
+            let tid = item.tid;
             cursor.remove_current();
-            SCHED.enqueue(pid);
+            todo!();
+//<<<<<<< HEAD
+//            sched::enqueue(pid);
+//=======
+//            SCHED.enqueue(tid);
+//>>>>>>> feat/thread
         } else {
             cursor.move_next();
         }
@@ -42,7 +62,7 @@ pub fn tick() {
 /// Suspends current process for given duration
 pub fn sleep(timeout: Duration, remaining: &mut Duration) -> Result<(), Errno> {
     // Dummy wait descriptor which will never receive notifications
-    static SLEEP_NOTIFY: Wait = Wait::new();
+    static SLEEP_NOTIFY: Wait = Wait::new("sleep");
     let deadline = machine::local_timer().timestamp()? + timeout;
     match SLEEP_NOTIFY.wait(Some(deadline)) {
         Err(Errno::Interrupt) => {
@@ -55,14 +75,15 @@ pub fn sleep(timeout: Duration, remaining: &mut Duration) -> Result<(), Errno> {
     }
 }
 
+/// Suspends current process until some file descriptor
+/// signals data available
 pub fn select(
-    proc: ProcessRef,
+    thread: ThreadRef,
     mut rfds: Option<&mut FdSet>,
     mut wfds: Option<&mut FdSet>,
     timeout: Option<Duration>,
 ) -> Result<usize, Errno> {
-    // TODO support wfds
-    if wfds.is_some() || rfds.is_none() {
+    if wfds.is_none() && rfds.is_none() {
         todo!();
     }
     let read = rfds.as_deref().map(FdSet::clone);
@@ -71,6 +92,7 @@ pub fn select(
     wfds.as_deref_mut().map(FdSet::reset);
 
     let deadline = timeout.map(|v| v + machine::local_timer().timestamp().unwrap());
+    let proc = thread.owner().unwrap();
     let mut io = proc.io.lock();
 
     loop {
@@ -104,9 +126,41 @@ pub fn select(
 
 impl Wait {
     /// Constructs a new wait channel
-    pub const fn new() -> Self {
+    pub const fn new(name: &'static str) -> Self {
         Self {
             queue: IrqSafeSpinLock::new(LinkedList::new()),
+            name
+        }
+    }
+
+    /// Interrupt wait pending on the channel
+    pub fn abort(&self, tid: u32, enqueue: bool) {
+        let mut queue = self.queue.lock();
+        let mut tick_lock = TICK_LIST.lock();
+        let mut cursor = tick_lock.cursor_front_mut();
+        while let Some(item) = cursor.current() {
+            if tid == item.tid {
+                cursor.remove_current();
+                break;
+            } else {
+                cursor.move_next();
+            }
+        }
+
+        let mut cursor = queue.cursor_front_mut();
+        while let Some(item) = cursor.current() {
+            if tid == *item {
+                cursor.remove_current();
+                let thread = Thread::get(tid).unwrap();
+                thread.set_wait_status(WaitStatus::Interrupted);
+                if enqueue {
+                    sched::enqueue(tid);
+                    // SCHED.enqueue(tid);
+                }
+                break;
+            } else {
+                cursor.move_next();
+            }
         }
     }
 
@@ -115,12 +169,12 @@ impl Wait {
         let mut queue = self.queue.lock();
         let mut count = 0;
         while limit != 0 && !queue.is_empty() {
-            let pid = queue.pop_front();
-            if let Some(pid) = pid {
+            let tid = queue.pop_front();
+            if let Some(tid) = tid {
                 let mut tick_lock = TICK_LIST.lock();
                 let mut cursor = tick_lock.cursor_front_mut();
                 while let Some(item) = cursor.current() {
-                    if pid == item.pid {
+                    if tid == item.tid {
                         cursor.remove_current();
                         break;
                     } else {
@@ -129,8 +183,8 @@ impl Wait {
                 }
                 drop(tick_lock);
 
-                proc::process(pid).set_wait_flag(false);
-                SCHED.enqueue(pid);
+                Thread::get(tid).unwrap().set_wait_status(WaitStatus::Done);
+                sched::enqueue(tid);
             }
 
             limit -= 1;
@@ -152,26 +206,33 @@ impl Wait {
     /// Suspends current process until event is signalled or
     /// (optional) deadline is reached
     pub fn wait(&self, deadline: Option<Duration>) -> Result<(), Errno> {
-        let proc = Process::current();
+        let thread = Thread::current();
         //let deadline = timeout.map(|t| machine::local_timer().timestamp().unwrap() + t);
         let mut queue_lock = self.queue.lock();
 
-        queue_lock.push_back(proc.id());
-        proc.set_wait_flag(true);
+        queue_lock.push_back(thread.id());
+        thread.setup_wait(self);
+
         if let Some(deadline) = deadline {
             TICK_LIST.lock().push_back(Timeout {
-                pid: proc.id(),
+                tid: thread.id(),
                 deadline,
             });
         }
 
         loop {
-            if !proc.wait_flag() {
-                return Ok(());
-            }
+            match thread.wait_status() {
+                WaitStatus::Pending => {}
+                WaitStatus::Done => {
+                    return Ok(());
+                }
+                WaitStatus::Interrupted => {
+                    return Err(Errno::Interrupt);
+                }
+            };
 
             drop(queue_lock);
-            proc.enter_wait();
+            thread.enter_wait();
             queue_lock = self.queue.lock();
 
             if let Some(deadline) = deadline {
@@ -179,7 +240,7 @@ impl Wait {
                     let mut cursor = queue_lock.cursor_front_mut();
 
                     while let Some(&mut item) = cursor.current() {
-                        if proc.id() == item {
+                        if thread.id() == item {
                             cursor.remove_current();
                             break;
                         } else {

kernel/src/sync.rs

@@ -4,12 +4,14 @@ use crate::arch::platform::{irq_mask_save, irq_restore};
 use core::cell::UnsafeCell;
 use core::fmt;
 use core::ops::{Deref, DerefMut};
-use core::sync::atomic::{AtomicBool, Ordering};
+use core::sync::atomic::{AtomicUsize, Ordering};
+use cortex_a::registers::MPIDR_EL1;
+use tock_registers::interfaces::Readable;
 
 /// Lock structure ensuring IRQs are disabled when inner value is accessed
 pub struct IrqSafeSpinLock<T> {
     value: UnsafeCell<T>,
-    state: AtomicBool,
+    state: AtomicUsize,
 }
 
 /// Guard-structure wrapping a reference to value owned by [IrqSafeSpinLock].
@@ -25,28 +27,25 @@ impl<T> IrqSafeSpinLock<T> {
     pub const fn new(value: T) -> Self {
         Self {
             value: UnsafeCell::new(value),
-            state: AtomicBool::new(false),
+            state: AtomicUsize::new(usize::MAX),
         }
     }
 
-    #[inline(always)]
-    fn try_lock(&self) -> Result<bool, bool> {
-        self.state
-            .compare_exchange_weak(false, true, Ordering::Acquire, Ordering::Relaxed)
-    }
-
-    #[inline(always)]
-    unsafe fn force_release(&self) {
-        self.state.store(false, Ordering::Release);
-        cortex_a::asm::sev();
-    }
-
     /// Returns [IrqSafeSpinLockGuard] for this lock
     #[inline]
     pub fn lock(&self) -> IrqSafeSpinLockGuard<T> {
         let irq_state = unsafe { irq_mask_save() };
+        let id = MPIDR_EL1.get() & 0xF;
 
-        while self.try_lock().is_err() {
+        while let Err(e) = self.state.compare_exchange_weak(
+            usize::MAX,
+            id as usize,
+            Ordering::Acquire,
+            Ordering::Relaxed,
+        ) {
+            // if e == id as usize {
+            //     break;
+            // }
             cortex_a::asm::wfe();
         }
 
@@ -55,11 +54,17 @@ impl<T> IrqSafeSpinLock<T> {
             irq_state,
         }
     }
+
+    pub unsafe fn force_release(&self) {
+        self.state.store(usize::MAX, Ordering::Release);
+        cortex_a::asm::sev();
+    }
 }
 
 impl<T> Deref for IrqSafeSpinLockGuard<'_, T> {
     type Target = T;
 
+    #[inline(always)]
     fn deref(&self) -> &Self::Target {
         unsafe { &*self.lock.value.get() }
     }

kernel/src/syscall/arg.rs

@@ -1,73 +1,190 @@
 //! System call argument ABI helpers
 
 use crate::mem;
-use core::mem::size_of;
+use core::alloc::Layout;
 use libsys::error::Errno;
+use crate::proc::Process;
 
-fn translate(virt: usize) -> Option<usize> {
+// TODO _mut() versions checking whether pages are actually writable
+
+macro_rules! invalid_memory {
+    ($($args:tt)+) => {
+        warnln!($($args)+);
+        #[cfg(feature = "aggressive_syscall")]
+        {
+            use libsys::signal::Signal;
+            use crate::proc::Thread;
+
+            let thread = Thread::current();
+            let proc = thread.owner().unwrap();
+            proc.enter_fault_signal(thread, Signal::SegmentationFault);
+        }
+        return Err(Errno::InvalidArgument);
+    }
+}
+
+#[inline(always)]
+fn is_el0_accessible(virt: usize, write: bool) -> bool {
     let mut res: usize;
     unsafe {
-        asm!("at s1e1r, {}; mrs {}, par_el1", in(reg) virt, out(reg) res);
-    }
-    if res & 1 == 0 {
-        Some(res & !(0xFFF | (0xFF << 56)))
-    } else {
-        None
+        if write {
+            asm!("at s1e0w, {}; mrs {}, par_el1", in(reg) virt, out(reg) res);
+        } else {
+            asm!("at s1e0r, {}; mrs {}, par_el1", in(reg) virt, out(reg) res);
+        }
     }
+    res & 1 == 0
 }
 
-/// Unwraps a slim structure pointer
-pub fn validate_user_ptr_struct<'a, T>(base: usize) -> Result<&'a mut T, Errno> {
-    validate_user_ptr_struct_option(base).and_then(|e| e.ok_or(Errno::InvalidArgument))
+/// Checks given argument and interprets it as a `T` reference
+pub fn struct_ref<'a, T>(base: usize) -> Result<&'a T, Errno> {
+    let layout = Layout::new::<T>();
+    if base % layout.align() != 0 {
+        invalid_memory!(
+            "Structure pointer is misaligned: base={:#x}, expected {:?}",
+            base,
+            layout
+        );
+    }
+    let bytes = buf_ref(base, layout.size())?;
+    Ok(unsafe { &*(bytes.as_ptr() as *const T) })
 }
 
-pub fn validate_user_ptr_struct_option<'a, T>(base: usize) -> Result<Option<&'a mut T>, Errno> {
+/// Checks given argument and interprets it as a `T` mutable reference
+pub fn struct_mut<'a, T>(base: usize) -> Result<&'a mut T, Errno> {
+    let layout = Layout::new::<T>();
+    if base % layout.align() != 0 {
+        invalid_memory!(
+            "Structure pointer is misaligned: base={:#x}, expected {:?}",
+            base,
+            layout
+        );
+    }
+    let bytes = buf_mut(base, layout.size())?;
+    Ok(unsafe { &mut *(bytes.as_mut_ptr() as *mut T) })
+}
+
+/// Checks given argument and interprets it as a `T` array buffer of size `count`
+pub fn struct_buf_ref<'a, T>(base: usize, count: usize) -> Result<&'a [T], Errno> {
+    let layout = Layout::array::<T>(count).unwrap();
+    if base % layout.align() != 0 {
+        invalid_memory!(
+            "Structure pointer is misaligned: base={:#x}, expected {:?}",
+            base,
+            layout
+        );
+    }
+    let bytes = buf_ref(base, layout.size())?;
+    Ok(unsafe { core::slice::from_raw_parts(bytes.as_ptr() as *const T, count) })
+}
+
+/// Checks given argument and interprets it as a `T` array buffer of size `count`
+pub fn struct_buf_mut<'a, T>(base: usize, count: usize) -> Result<&'a mut [T], Errno> {
+    let layout = Layout::array::<T>(count).unwrap();
+    if base % layout.align() != 0 {
+        invalid_memory!(
+            "Structure pointer is misaligned: base={:#x}, expected {:?}",
+            base,
+            layout
+        );
+    }
+    let bytes = buf_mut(base, layout.size())?;
+    Ok(unsafe { core::slice::from_raw_parts_mut(bytes.as_mut_ptr() as *mut T, count) })
+}
+
+/// Checks given argument and interprets it as a `Option<&'a T>`
+pub fn option_struct_ref<'a, T>(base: usize) -> Result<Option<&'a T>, Errno> {
     if base == 0 {
         Ok(None)
     } else {
-        let bytes = validate_user_ptr(base, size_of::<T>())?;
-        Ok(Some(unsafe { &mut *(bytes.as_mut_ptr() as *mut T) }))
+        struct_ref(base).map(Some)
     }
 }
 
-/// Unwraps an user buffer reference
-pub fn validate_user_ptr<'a>(base: usize, len: usize) -> Result<&'a mut [u8], Errno> {
+/// Checks given argument and interprets it as a `Option<&'a mut T>`
+pub fn option_struct_mut<'a, T>(base: usize) -> Result<Option<&'a mut T>, Errno> {
+    if base == 0 {
+        Ok(None)
+    } else {
+        struct_mut(base).map(Some)
+    }
+}
+
+/// Validates that the argument pointer is accessible for requested operation
+/// for current process
+pub fn validate_ptr(base: usize, len: usize, write: bool) -> Result<(), Errno> {
     if base > mem::KERNEL_OFFSET || base + len > mem::KERNEL_OFFSET {
-        warnln!(
+        invalid_memory!(
             "User region refers to kernel memory: base={:#x}, len={:#x}",
             base,
             len
         );
-        return Err(Errno::InvalidArgument);
     }
 
+    let process = Process::current();
+
     for i in (base / mem::PAGE_SIZE)..((base + len + mem::PAGE_SIZE - 1) / mem::PAGE_SIZE) {
-        if translate(i * mem::PAGE_SIZE).is_none() {
-            warnln!(
-                "User region refers to unmapped memory: base={:#x}, len={:#x} (page {:#x})",
+        if !is_el0_accessible(i * mem::PAGE_SIZE, write) {
+            // It's possible a CoW page hasn't yet been cloned when trying
+            // a write access
+            let res = if write {
+                process.manipulate_space(|space| {
+                    space.try_cow_copy(i * mem::PAGE_SIZE)
+                })
+            } else {
+                Err(Errno::DoesNotExist)
+            };
+
+            if res.is_ok() {
+                continue;
+            }
+
+            invalid_memory!(
+                "User region refers to inaccessible/unmapped memory: base={:#x}, len={:#x} (page {:#x}, write={})",
                 base,
                 len,
-                i * mem::PAGE_SIZE
+                i * mem::PAGE_SIZE,
+                write
             );
-            return Err(Errno::InvalidArgument);
         }
     }
 
+    Ok(())
+}
+
+/// Checks given argument and interprets it as a byte buffer
+pub fn buf_ref<'a>(base: usize, len: usize) -> Result<&'a [u8], Errno> {
+    validate_ptr(base, len, false)?;
+    Ok(unsafe { core::slice::from_raw_parts(base as *const u8, len) })
+}
+
+/// Checks given argument and interprets it as a mutable byte buffer
+pub fn buf_mut<'a>(base: usize, len: usize) -> Result<&'a mut [u8], Errno> {
+    validate_ptr(base, len, true)?;
     Ok(unsafe { core::slice::from_raw_parts_mut(base as *mut u8, len) })
 }
 
-/// Unwraps a nullable user buffer reference
-pub fn validate_user_ptr_null<'a>(base: usize, len: usize) -> Result<Option<&'a mut [u8]>, Errno> {
+/// Checks possibly NULL given argument and interprets it as a byte buffer
+pub fn option_buf_ref<'a>(base: usize, len: usize) -> Result<Option<&'a [u8]>, Errno> {
     if base == 0 {
         Ok(None)
     } else {
-        validate_user_ptr(base, len).map(Some)
+        buf_ref(base, len).map(Some)
+    }
+}
+
+/// Checks possibly NULL given argument and interprets it as a mutable byte buffer
+pub fn option_buf_mut<'a>(base: usize, len: usize) -> Result<Option<&'a mut [u8]>, Errno> {
+    if base == 0 {
+        Ok(None)
+    } else {
+        buf_mut(base, len).map(Some)
     }
 }
 
 /// Unwraps user string argument
-pub fn validate_user_str<'a>(base: usize, len: usize) -> Result<&'a str, Errno> {
-    let bytes = validate_user_ptr(base, len)?;
+pub fn str_ref<'a>(base: usize, len: usize) -> Result<&'a str, Errno> {
+    let bytes = buf_ref(base, len)?;
     core::str::from_utf8(bytes).map_err(|_| {
         warnln!(
             "User string contains invalid UTF-8 characters: base={:#x}, len={:#x}",
@@ -77,49 +194,3 @@ pub fn validate_user_str<'a>(base: usize, len: usize) -> Result<&'a str, Errno>
         Errno::InvalidArgument
     })
 }
-//     if base > mem::KERNEL_OFFSET {
-//         warnln!("User string refers to kernel memory: base={:#x}", base);
-//         return Err(Errno::InvalidArgument);
-//     }
-//
-//     let base_ptr = base as *const u8;
-//     let mut len = 0;
-//     let mut page_valid = false;
-//     loop {
-//         if len == limit {
-//             warnln!("User string exceeded limit: base={:#x}", base);
-//             return Err(Errno::InvalidArgument);
-//         }
-//
-//         if (base + len) % mem::PAGE_SIZE == 0 {
-//             page_valid = false;
-//         }
-//
-//         if !page_valid && translate((base + len) & !0xFFF).is_none() {
-//             warnln!(
-//                 "User string refers to unmapped memory: base={:#x}, off={:#x}",
-//                 base,
-//                 len
-//             );
-//             return Err(Errno::InvalidArgument);
-//         }
-//
-//         page_valid = true;
-//
-//         let byte = unsafe { *base_ptr.add(len) };
-//         if byte == 0 {
-//             break;
-//         }
-//
-//         len += 1;
-//     }
-//
-//     let slice = unsafe { core::slice::from_raw_parts(base_ptr, len) };
-//     core::str::from_utf8(slice).map_err(|_| {
-//         warnln!(
-//             "User string contains invalid UTF-8 characters: base={:#x}",
-//             base
-//         );
-//         Errno::InvalidArgument
-//     })
-// }

kernel/src/syscall/mod.rs

@@ -1,24 +1,30 @@
 //! System call implementation
 
-use crate::arch::platform::exception::ExceptionFrame;
+use crate::arch::{machine, platform::exception::ExceptionFrame};
+use crate::mem::{virt::MapAttributes, phys::PageUsage};
 use crate::debug::Level;
-use crate::proc::{elf, wait, Process, ProcessIo};
+use crate::dev::timer::TimestampSource;
+use crate::fs::create_filesystem;
+use crate::proc::{self, sched, elf, wait, Process, ProcessIo, Thread};
 use core::mem::size_of;
 use core::ops::DerefMut;
 use core::time::Duration;
 use libsys::{
-    abi,
+    abi::SystemCall,
+    debug::TraceLevel,
     error::Errno,
     ioctl::IoctlCmd,
-    proc::Pid,
+    proc::{ExitCode, Pid, MemoryAccess},
     signal::{Signal, SignalDestination},
-    stat::{FdSet, FileDescriptor, FileMode, OpenFlags, Stat, AT_EMPTY_PATH},
+    stat::{
+        AccessMode, DirectoryEntry, FdSet, FileDescriptor, FileMode, GroupId, MountOptions,
+        OpenFlags, Stat, UserId, AT_EMPTY_PATH,
+    },
     traits::{Read, Write},
 };
 use vfs::VnodeRef;
 
 pub mod arg;
-pub use arg::*;
 
 /// Creates a "fork" process from current one using its register frame.
 /// See [Process::fork()].
@@ -51,18 +57,28 @@ fn find_at_node<T: DerefMut<Target = ProcessIo>>(
 }
 
 /// Main system call dispatcher function
-pub fn syscall(num: usize, args: &[usize]) -> Result<usize, Errno> {
+pub fn syscall(num: SystemCall, args: &[usize]) -> Result<usize, Errno> {
     match num {
-        // Process management system calls
-        abi::SYS_EXIT => {
-            Process::current().exit(args[0] as i32);
-            unreachable!();
-        }
+        // I/O
+        SystemCall::Read => {
+            let proc = Process::current();
+            let fd = FileDescriptor::from(args[0] as u32);
+            let mut io = proc.io.lock();
+            let buf = arg::buf_mut(args[1], args[2])?;
 
-        // I/O system calls
-        abi::SYS_OPENAT => {
+            io.file(fd)?.borrow_mut().read(buf)
+        }
+        SystemCall::Write => {
+            let proc = Process::current();
+            let fd = FileDescriptor::from(args[0] as u32);
+            let mut io = proc.io.lock();
+            let buf = arg::buf_ref(args[1], args[2])?;
+
+            io.file(fd)?.borrow_mut().write(buf)
+        }
+        SystemCall::Open => {
             let at_fd = FileDescriptor::from_i32(args[0] as i32)?;
-            let path = validate_user_str(args[1], args[2])?;
+            let path = arg::str_ref(args[1], args[2])?;
             let mode = FileMode::from_bits(args[3] as u32).ok_or(Errno::InvalidArgument)?;
             let opts = OpenFlags::from_bits(args[4] as u32).ok_or(Errno::InvalidArgument)?;
 
@@ -78,34 +94,7 @@ pub fn syscall(num: usize, args: &[usize]) -> Result<usize, Errno> {
             let file = io.ioctx().open(at, path, mode, opts)?;
             Ok(u32::from(io.place_file(file)?) as usize)
         }
-        abi::SYS_READ => {
-            let proc = Process::current();
-            let fd = FileDescriptor::from(args[0] as u32);
-            let mut io = proc.io.lock();
-            let buf = validate_user_ptr(args[1], args[2])?;
-
-            io.file(fd)?.borrow_mut().read(buf)
-        }
-        abi::SYS_WRITE => {
-            let proc = Process::current();
-            let fd = FileDescriptor::from(args[0] as u32);
-            let mut io = proc.io.lock();
-            let buf = validate_user_ptr(args[1], args[2])?;
-
-            io.file(fd)?.borrow_mut().write(buf)
-        }
-        abi::SYS_FSTATAT => {
-            let at_fd = FileDescriptor::from_i32(args[0] as i32)?;
-            let filename = validate_user_str(args[1], args[2])?;
-            let buf = validate_user_ptr_struct::<Stat>(args[3])?;
-            let flags = args[4] as u32;
-
-            let proc = Process::current();
-            let mut io = proc.io.lock();
-            find_at_node(&mut io, at_fd, filename, flags & AT_EMPTY_PATH != 0)?.stat(buf)?;
-            Ok(0)
-        }
-        abi::SYS_CLOSE => {
+        SystemCall::Close => {
             let proc = Process::current();
             let mut io = proc.io.lock();
             let fd = FileDescriptor::from(args[0] as u32);
@@ -113,34 +102,20 @@ pub fn syscall(num: usize, args: &[usize]) -> Result<usize, Errno> {
             io.close_file(fd)?;
             Ok(0)
         }
-        abi::SYS_EXECVE => {
-            let node = {
-                let proc = Process::current();
-                let mut io = proc.io.lock();
-                let filename = validate_user_str(args[0], args[1])?;
-                // TODO argv, envp array passing ABI?
-                let node = io.ioctx().find(None, filename, true)?;
-                drop(io);
-                node
-            };
-            let file = node.open(OpenFlags::O_RDONLY)?;
-            Process::execve(|space| elf::load_elf(space, file), 0).unwrap();
-            panic!();
-        }
-        abi::SYS_WAITPID => {
-            // TODO special "pid" values
-            let pid = unsafe { Pid::from_raw(args[0] as u32) };
-            let status = validate_user_ptr_struct::<i32>(args[1])?;
+        SystemCall::FileStatus => {
+            let at_fd = FileDescriptor::from_i32(args[0] as i32)?;
+            let filename = arg::str_ref(args[1], args[2])?;
+            let buf = arg::struct_mut::<Stat>(args[3])?;
+            let flags = args[4] as u32;
 
-            match Process::waitpid(pid) {
-                Ok(exit) => {
-                    *status = i32::from(exit);
-                    Ok(0)
-                }
-                _ => todo!(),
-            }
+            let proc = Process::current();
+            let mut io = proc.io.lock();
+            let stat =
+                find_at_node(&mut io, at_fd, filename, flags & AT_EMPTY_PATH != 0)?.stat()?;
+            *buf = stat;
+            Ok(0)
         }
-        abi::SYS_IOCTL => {
+        SystemCall::Ioctl => {
             let fd = FileDescriptor::from(args[0] as u32);
             let cmd = IoctlCmd::try_from(args[1] as u32)?;
 
@@ -150,19 +125,197 @@ pub fn syscall(num: usize, args: &[usize]) -> Result<usize, Errno> {
             let node = io.file(fd)?.borrow().node().ok_or(Errno::InvalidFile)?;
             node.ioctl(cmd, args[2], args[3])
         }
+        SystemCall::Select => {
+            let rfds = arg::option_struct_mut::<FdSet>(args[0])?;
+            let wfds = arg::option_struct_mut::<FdSet>(args[1])?;
+            let timeout = if args[2] == 0 {
+                None
+            } else {
+                Some(Duration::from_nanos(args[2] as u64))
+            };
 
-        // Extra system calls
-        abi::SYS_EX_DEBUG_TRACE => {
-            let buf = validate_user_ptr(args[0], args[1])?;
-            print!(Level::Debug, "[trace] ");
-            for &byte in buf.iter() {
-                print!(Level::Debug, "{}", byte as char);
-            }
-            println!(Level::Debug, "");
-            Ok(args[1])
+            wait::select(Thread::current(), rfds, wfds, timeout)
         }
-        abi::SYS_EX_NANOSLEEP => {
-            let rem_buf = validate_user_ptr_null(args[1], size_of::<u64>() * 2)?;
+        SystemCall::Access => {
+            let at_fd = FileDescriptor::from_i32(args[0] as i32)?;
+            let path = arg::str_ref(args[1], args[2])?;
+            let mode = AccessMode::from_bits(args[3] as u32).ok_or(Errno::InvalidArgument)?;
+            let flags = args[4] as u32;
+
+            let proc = Process::current();
+            let mut io = proc.io.lock();
+
+            find_at_node(&mut io, at_fd, path, flags & AT_EMPTY_PATH != 0)?
+                .check_access(io.ioctx(), mode)?;
+            Ok(0)
+        }
+        SystemCall::ReadDirectory => {
+            let proc = Process::current();
+            let fd = FileDescriptor::from(args[0] as u32);
+            let mut io = proc.io.lock();
+            let buf = arg::struct_buf_mut::<DirectoryEntry>(args[1], args[2])?;
+
+            io.file(fd)?.borrow_mut().readdir(buf)
+        }
+        SystemCall::GetUserId => {
+            let proc = Process::current();
+            let uid = proc.io.lock().uid();
+            Ok(u32::from(uid) as usize)
+        }
+        SystemCall::GetGroupId => {
+            let proc = Process::current();
+            let gid = proc.io.lock().gid();
+            Ok(u32::from(gid) as usize)
+        }
+        SystemCall::DuplicateFd => {
+            let src = FileDescriptor::from(args[0] as u32);
+            let dst = FileDescriptor::from_i32(args[1] as i32)?;
+
+            let proc = Process::current();
+            let mut io = proc.io.lock();
+
+            let res = io.duplicate_file(src, dst)?;
+
+            Ok(u32::from(res) as usize)
+        }
+        SystemCall::SetUserId => {
+            let uid = UserId::from(args[0] as u32);
+            let proc = Process::current();
+            proc.io.lock().set_uid(uid)?;
+            Ok(0)
+        }
+        SystemCall::SetGroupId => {
+            let gid = GroupId::from(args[0] as u32);
+            let proc = Process::current();
+            proc.io.lock().set_gid(gid)?;
+            Ok(0)
+        }
+        SystemCall::SetCurrentDirectory => {
+            let path = arg::str_ref(args[0], args[1])?;
+            let proc = Process::current();
+            proc.io.lock().ioctx().chdir(path)?;
+            Ok(0)
+        }
+        SystemCall::GetCurrentDirectory => {
+            todo!()
+        }
+        SystemCall::Seek => {
+            todo!()
+        }
+        SystemCall::MapMemory => {
+            let len = args[1];
+            if len == 0 || (len & 0xFFF) != 0 {
+                return Err(Errno::InvalidArgument);
+            }
+            let acc = MemoryAccess::from_bits(args[2] as u32).ok_or(Errno::InvalidArgument)?;
+            let _flags = MemoryAccess::from_bits(args[3] as u32).ok_or(Errno::InvalidArgument)?;
+
+            let mut attrs = MapAttributes::NOT_GLOBAL | MapAttributes::SH_OUTER | MapAttributes::PXN;
+            if !acc.contains(MemoryAccess::READ) {
+                return Err(Errno::NotImplemented);
+            }
+            if acc.contains(MemoryAccess::WRITE) {
+                if acc.contains(MemoryAccess::EXEC) {
+                    return Err(Errno::PermissionDenied);
+                }
+                attrs |= MapAttributes::AP_BOTH_READWRITE;
+            } else {
+                attrs |= MapAttributes::AP_BOTH_READONLY;
+            }
+            if !acc.contains(MemoryAccess::EXEC) {
+                attrs |= MapAttributes::UXN;
+            }
+
+            // TODO don't ignore flags
+            let usage = PageUsage::UserPrivate;
+
+            let proc = Process::current();
+
+            proc.manipulate_space(move |space| {
+                space.allocate(0x100000000, 0xF00000000, len / 4096, attrs, usage)
+            })
+        }
+        SystemCall::UnmapMemory => {
+            let addr = args[0];
+            let len = args[1];
+
+            if addr == 0 || len == 0 || addr & 0xFFF != 0 || len & 0xFFF != 0 {
+                return Err(Errno::InvalidArgument);
+            }
+
+            let proc = Process::current();
+            proc.manipulate_space(move |space| {
+                space.free(addr, len / 4096)
+            })?;
+            Ok(0)
+        }
+
+        // Process
+        SystemCall::Clone => {
+            let entry = args[0];
+            let stack = args[1];
+            let arg = args[2];
+
+            Process::current()
+                .new_user_thread(entry, stack, arg)
+                .map(|e| e as usize)
+        }
+        SystemCall::Exec => {
+            let filename = arg::str_ref(args[0], args[1])?;
+            let argv = arg::struct_buf_ref::<&str>(args[2], args[3])?;
+            // Validate each argument as well
+            for item in argv.iter() {
+                arg::validate_ptr(item.as_ptr() as usize, item.len(), false)?;
+            }
+            let node = {
+                let proc = Process::current();
+                let mut io = proc.io.lock();
+                // TODO argv, envp array passing ABI?
+                let node = io.ioctx().find(None, filename, true)?;
+                drop(io);
+                node
+            };
+            let file = node.open(OpenFlags::O_RDONLY)?;
+            Process::execve(move |space| elf::load_elf(space, file), argv).unwrap();
+            panic!();
+        }
+        SystemCall::Exit => {
+            let status = ExitCode::from(args[0] as i32);
+            let flags = args[1];
+
+            if flags & (1 << 0) != 0 {
+                Process::exit_thread(Thread::current(), status);
+            } else {
+                Process::current().exit(status);
+            }
+
+            unreachable!();
+        }
+        SystemCall::WaitPid => {
+            // TODO special "pid" values
+            let pid = unsafe { Pid::from_raw(args[0] as u32) };
+            let status = arg::struct_mut::<i32>(args[1])?;
+
+            match Process::waitpid(pid) {
+                Ok(exit) => {
+                    *status = i32::from(exit);
+                    Ok(0)
+                }
+                e => e.map(|e| i32::from(e) as usize),
+            }
+        }
+        SystemCall::WaitTid => {
+            let tid = args[0] as u32;
+
+            match Thread::waittid(tid) {
+                Ok(_) => Ok(0),
+                _ => todo!(),
+            }
+        }
+        SystemCall::GetPid => Ok(Process::current().id().value() as usize),
+        SystemCall::GetTid => Ok(Thread::current().id() as usize),
+        SystemCall::Sleep => {
+            let rem_buf = arg::option_buf_ref(args[1], size_of::<u64>() * 2)?;
             let mut rem = Duration::new(0, 0);
             let res = wait::sleep(Duration::from_nanos(args[0] as u64), &mut rem);
             if res == Err(Errno::Interrupt) {
@@ -173,17 +326,15 @@ pub fn syscall(num: usize, args: &[usize]) -> Result<usize, Errno> {
             }
             res.map(|_| 0)
         }
-        abi::SYS_EX_SIGNAL => {
-            let proc = Process::current();
-            proc.setup_signal_context(args[0], args[1]);
+        SystemCall::SetSignalEntry => {
+            Thread::current().set_signal_entry(args[0], args[1]);
             Ok(0)
         }
-        abi::SYS_EX_SIGRETURN => {
-            let proc = Process::current();
-            proc.return_from_signal();
-            panic!("This code won't run");
+        SystemCall::SignalReturn => {
+            Thread::current().return_from_signal();
+            unreachable!();
         }
-        abi::SYS_EX_KILL => {
+        SystemCall::SendSignal => {
             let target = SignalDestination::from(args[0] as isize);
             let signal = Signal::try_from(args[1] as u32)?;
 
@@ -196,25 +347,104 @@ pub fn syscall(num: usize, args: &[usize]) -> Result<usize, Errno> {
             };
             Ok(0)
         }
-
-        abi::SYS_SELECT => {
-            let rfds = validate_user_ptr_struct_option::<FdSet>(args[0])?;
-            let wfds = validate_user_ptr_struct_option::<FdSet>(args[1])?;
-            let timeout = if args[2] == 0 {
-                None
+        SystemCall::Yield => {
+            sched::switch(false);
+            Ok(0)
+        }
+        SystemCall::GetSid => {
+            // TODO handle kernel processes here?
+            let pid = args[0] as u32;
+            let current = Process::current();
+            let proc = if pid == 0 {
+                current
             } else {
-                Some(Duration::from_nanos(args[2] as u64))
+                let pid = unsafe { Pid::from_raw(pid) };
+                let proc = Process::get(pid).ok_or(Errno::DoesNotExist)?;
+                if proc.sid() != current.sid() {
+                    return Err(Errno::PermissionDenied);
+                }
+                proc
             };
 
+            Ok(proc.sid().value() as usize)
+        }
+        SystemCall::GetPgid => {
+            // TODO handle kernel processes here?
+            let pid = args[0] as u32;
+            let current = Process::current();
+            let proc = if pid == 0 {
+                current
+            } else {
+                let pid = unsafe { Pid::from_raw(pid) };
+                Process::get(pid).ok_or(Errno::DoesNotExist)?
+            };
+
+            Ok(proc.pgid().value() as usize)
+        }
+        SystemCall::GetPpid => Ok(Process::current().ppid().unwrap().value() as usize),
+        SystemCall::SetSid => {
             let proc = Process::current();
-            wait::select(proc, rfds, wfds, timeout)
+            let mut io = proc.io.lock();
+
+            if let Some(_ctty) = io.ctty() {
+                todo!();
+            }
+
+            let id = proc.id();
+            proc.set_sid(id);
+            Ok(id.value() as usize)
+        }
+        SystemCall::SetPgid => {
+            let pid = args[0] as u32;
+            let pgid = args[1] as u32;
+
+            let current = Process::current();
+            let proc = if pid == 0 { current } else { todo!() };
+
+            if pgid == 0 {
+                proc.set_pgid(proc.id());
+            } else {
+                todo!();
+            }
+
+            Ok(proc.pgid().value() as usize)
         }
 
-        _ => {
-            let proc = Process::current();
-            errorln!("Undefined system call: {}", num);
-            proc.enter_signal(Signal::InvalidSystemCall);
-            todo!()
+        // System
+        SystemCall::GetCpuTime => {
+            let time = machine::local_timer().timestamp()?;
+            Ok(time.as_nanos() as usize)
         }
+        SystemCall::Mount => {
+            let target = arg::str_ref(args[0], args[1])?;
+            let options = arg::struct_ref::<MountOptions>(args[2])?;
+
+            let proc = Process::current();
+            let mut io = proc.io.lock();
+
+            debugln!("mount(target={:?}, options={:#x?})", target, options);
+
+            let target_node = io.ioctx().find(None, target, true)?;
+            let root = create_filesystem(options)?;
+
+            target_node.mount(root)?;
+
+            Ok(0)
+        }
+
+        // Debugging
+        SystemCall::DebugTrace => {
+            let level = TraceLevel::from_repr(args[0])
+                .map(Level::from)
+                .ok_or(Errno::InvalidArgument)?;
+            let buf = arg::str_ref(args[1], args[2])?;
+            let thread = Thread::current();
+            let proc = thread.owner().unwrap();
+            println!(level, "[trace {:?}:{}] {}", proc.id(), thread.id(), buf);
+            Ok(args[1])
+        }
+
+        // Handled elsewhere
+        SystemCall::Fork => unreachable!(),
     }
 }

libsys/Cargo.toml

@@ -7,6 +7,7 @@ edition = "2021"
 
 [dependencies]
 bitflags = "^1.3.0"
+enum-repr = "^0.2.6"
 
 [features]
 user = []

libsys/src/abi.rs

@@ -1,18 +1,51 @@
-pub const SYS_EX_DEBUG_TRACE: usize = 128;
-pub const SYS_EX_NANOSLEEP: usize = 129;
+use enum_repr::EnumRepr;
 
-pub const SYS_EX_SIGNAL: usize = 130;
-pub const SYS_EX_SIGRETURN: usize = 131;
-pub const SYS_EX_KILL: usize = 132;
+#[EnumRepr(type = "usize")]
+#[derive(Clone, Copy, Debug, PartialEq, Eq, PartialOrd, Ord)]
+pub enum SystemCall {
+    // I/O
+    Read = 1,
+    Write = 2,
+    Open = 3,
+    Close = 4,
+    FileStatus = 5,
+    Ioctl = 6,
+    Select = 7,
+    Access = 8,
+    ReadDirectory = 9,
+    GetUserId = 10,
+    GetGroupId = 11,
+    DuplicateFd = 12,
+    SetUserId = 13,
+    SetGroupId = 14,
+    SetCurrentDirectory = 15,
+    GetCurrentDirectory = 16,
+    Seek = 17,
+    MapMemory = 18,
+    UnmapMemory = 19,
 
-pub const SYS_EXIT: usize = 1;
-pub const SYS_READ: usize = 2;
-pub const SYS_WRITE: usize = 3;
-pub const SYS_OPENAT: usize = 4;
-pub const SYS_FSTATAT: usize = 5;
-pub const SYS_CLOSE: usize = 6;
-pub const SYS_FORK: usize = 7;
-pub const SYS_EXECVE: usize = 8;
-pub const SYS_WAITPID: usize = 9;
-pub const SYS_IOCTL: usize = 10;
-pub const SYS_SELECT: usize = 11;
+    // Process manipulation
+    Fork = 32,
+    Clone = 33,
+    Exec = 34,
+    Exit = 35,
+    WaitPid = 36,
+    WaitTid = 37,
+    GetPid = 38,
+    GetTid = 39,
+    Sleep = 40,
+    SetSignalEntry = 41,
+    SignalReturn = 42,
+    SendSignal = 43,
+    Yield = 44,
+    GetSid = 45,
+    GetPgid = 46,
+    GetPpid = 47,
+    SetSid = 48,
+    SetPgid = 49,
+    // System
+    GetCpuTime = 64,
+    Mount = 65,
+    // Debugging
+    DebugTrace = 128
+}

libsys/src/calls.rs

@@ -1,53 +1,59 @@
-use crate::abi;
+use crate::abi::SystemCall;
 use crate::{
+    debug::TraceLevel,
     error::Errno,
     ioctl::IoctlCmd,
-    proc::{ExitCode, Pid},
+    proc::{ExitCode, MemoryAccess, MemoryMap, Pid},
     signal::{Signal, SignalDestination},
-    stat::{FdSet, FileDescriptor, FileMode, OpenFlags, Stat},
+    stat::{
+        AccessMode, DirectoryEntry, FdSet, FileDescriptor, FileMode, GroupId, MountOptions,
+        OpenFlags, Stat, UserId,
+    },
 };
+use core::time::Duration;
 
 // TODO document the syscall ABI
 
+// TODO move this to libusr
 macro_rules! syscall {
     ($num:expr) => {{
         let mut res: usize;
-        asm!("svc #0", out("x0") res, in("x8") $num, options(nostack));
+        asm!("svc #0", out("x0") res, in("x8") $num.repr(), options(nostack));
         res
     }};
     ($num:expr, $a0:expr) => {{
         let mut res: usize = $a0;
         asm!("svc #0",
              inout("x0") res,
-             in("x8") $num, options(nostack));
+             in("x8") $num.repr(), options(nostack));
         res
     }};
     ($num:expr, $a0:expr, $a1:expr) => {{
         let mut res: usize = $a0;
         asm!("svc #0",
              inout("x0") res, in("x1") $a1,
-             in("x8") $num, options(nostack));
+             in("x8") $num.repr(), options(nostack));
         res
     }};
     ($num:expr, $a0:expr, $a1:expr, $a2:expr) => {{
         let mut res: usize = $a0;
         asm!("svc #0",
              inout("x0") res, in("x1") $a1, in("x2") $a2,
-             in("x8") $num, options(nostack));
+             in("x8") $num.repr(), options(nostack));
         res
     }};
     ($num:expr, $a0:expr, $a1:expr, $a2:expr, $a3:expr) => {{
         let mut res: usize = $a0;
         asm!("svc #0",
              inout("x0") res, in("x1") $a1, in("x2") $a2,
-             in("x3") $a3, in("x8") $num, options(nostack));
+             in("x3") $a3, in("x8") $num.repr(), options(nostack));
         res
     }};
     ($num:expr, $a0:expr, $a1:expr, $a2:expr, $a3:expr, $a4:expr) => {{
         let mut res: usize = $a0;
         asm!("svc #0",
              inout("x0") res, in("x1") $a1, in("x2") $a2,
-             in("x3") $a3, in("x4") $a4, in("x8") $num, options(nostack));
+             in("x3") $a3, in("x4") $a4, in("x8") $num.repr(), options(nostack));
         res
     }};
 }
@@ -75,7 +81,7 @@ macro_rules! argp {
 #[inline(always)]
 pub fn sys_exit(code: ExitCode) -> ! {
     unsafe {
-        syscall!(abi::SYS_EXIT, argn!(i32::from(code)));
+        syscall!(SystemCall::Exit, argn!(i32::from(code)), argn!(0));
     }
     unreachable!();
 }
@@ -85,7 +91,7 @@ pub fn sys_exit(code: ExitCode) -> ! {
 /// System call
 #[inline(always)]
 pub fn sys_close(fd: FileDescriptor) -> Result<(), Errno> {
-    Errno::from_syscall_unit(unsafe { syscall!(abi::SYS_CLOSE, argn!(u32::from(fd))) })
+    Errno::from_syscall_unit(unsafe { syscall!(SystemCall::Close, argn!(u32::from(fd))) })
 }
 
 /// # Safety
@@ -94,7 +100,7 @@ pub fn sys_close(fd: FileDescriptor) -> Result<(), Errno> {
 #[inline(always)]
 pub fn sys_ex_nanosleep(ns: u64, rem: &mut [u64; 2]) -> Result<(), Errno> {
     Errno::from_syscall_unit(unsafe {
-        syscall!(abi::SYS_EX_NANOSLEEP, argn!(ns), argp!(rem.as_mut_ptr()))
+        syscall!(SystemCall::Sleep, argn!(ns), argp!(rem.as_mut_ptr()))
     })
 }
 
@@ -102,10 +108,11 @@ pub fn sys_ex_nanosleep(ns: u64, rem: &mut [u64; 2]) -> Result<(), Errno> {
 ///
 /// System call
 #[inline(always)]
-pub fn sys_ex_debug_trace(msg: &[u8]) -> Result<(), Errno> {
+pub fn sys_ex_debug_trace(level: TraceLevel, msg: &[u8]) -> Result<(), Errno> {
     Errno::from_syscall_unit(unsafe {
         syscall!(
-            abi::SYS_EX_DEBUG_TRACE,
+            SystemCall::DebugTrace,
+            argn!(level.repr()),
             argp!(msg.as_ptr()),
             argn!(msg.len())
         )
@@ -124,7 +131,7 @@ pub fn sys_openat(
 ) -> Result<FileDescriptor, Errno> {
     Errno::from_syscall(unsafe {
         syscall!(
-            abi::SYS_OPENAT,
+            SystemCall::Open,
             argn!(FileDescriptor::into_i32(at)),
             argp!(pathname.as_ptr()),
             argn!(pathname.len()),
@@ -142,7 +149,7 @@ pub fn sys_openat(
 pub fn sys_read(fd: FileDescriptor, data: &mut [u8]) -> Result<usize, Errno> {
     Errno::from_syscall(unsafe {
         syscall!(
-            abi::SYS_READ,
+            SystemCall::Read,
             argn!(u32::from(fd)),
             argp!(data.as_mut_ptr()),
             argn!(data.len())
@@ -154,7 +161,7 @@ pub fn sys_read(fd: FileDescriptor, data: &mut [u8]) -> Result<usize, Errno> {
 pub fn sys_write(fd: FileDescriptor, data: &[u8]) -> Result<usize, Errno> {
     Errno::from_syscall(unsafe {
         syscall!(
-            abi::SYS_WRITE,
+            SystemCall::Write,
             argn!(u32::from(fd)),
             argp!(data.as_ptr()),
             argn!(data.len())
@@ -174,7 +181,7 @@ pub fn sys_fstatat(
 ) -> Result<(), Errno> {
     Errno::from_syscall_unit(unsafe {
         syscall!(
-            abi::SYS_FSTATAT,
+            SystemCall::FileStatus,
             argn!(FileDescriptor::into_i32(at)),
             argp!(pathname.as_ptr()),
             argn!(pathname.len()),
@@ -188,8 +195,8 @@ pub fn sys_fstatat(
 ///
 /// System call
 #[inline(always)]
-pub fn sys_fork() -> Result<Option<Pid>, Errno> {
-    Errno::from_syscall(unsafe { syscall!(abi::SYS_FORK) }).map(|res| {
+pub unsafe fn sys_fork() -> Result<Option<Pid>, Errno> {
+    Errno::from_syscall(syscall!(SystemCall::Fork)).map(|res| {
         if res != 0 {
             Some(unsafe { Pid::from_raw(res as u32) })
         } else {
@@ -202,12 +209,14 @@ pub fn sys_fork() -> Result<Option<Pid>, Errno> {
 ///
 /// System call
 #[inline(always)]
-pub fn sys_execve(pathname: &str) -> Result<(), Errno> {
+pub fn sys_execve(pathname: &str, argv: &[&str]) -> Result<(), Errno> {
     Errno::from_syscall_unit(unsafe {
         syscall!(
-            abi::SYS_EXECVE,
+            SystemCall::Exec,
             argp!(pathname.as_ptr()),
-            argn!(pathname.len())
+            argn!(pathname.len()),
+            argp!(argv.as_ptr()),
+            argn!(argv.len())
         )
     })
 }
@@ -219,7 +228,7 @@ pub fn sys_execve(pathname: &str) -> Result<(), Errno> {
 pub fn sys_waitpid(pid: Pid, status: &mut i32) -> Result<(), Errno> {
     Errno::from_syscall_unit(unsafe {
         syscall!(
-            abi::SYS_WAITPID,
+            SystemCall::WaitPid,
             argn!(pid.value()),
             argp!(status as *mut i32)
         )
@@ -238,7 +247,7 @@ pub fn sys_ioctl(
 ) -> Result<usize, Errno> {
     Errno::from_syscall(unsafe {
         syscall!(
-            abi::SYS_IOCTL,
+            SystemCall::Ioctl,
             argn!(u32::from(fd)),
             argn!(cmd),
             argn!(ptr),
@@ -247,15 +256,23 @@ pub fn sys_ioctl(
     })
 }
 
+#[inline(always)]
+pub fn sys_ex_getcputime() -> Result<Duration, Errno> {
+    Errno::from_syscall(unsafe { syscall!(SystemCall::GetCpuTime) })
+        .map(|e| Duration::from_nanos(e as u64))
+}
+
 #[inline(always)]
 pub fn sys_ex_signal(entry: usize, stack: usize) -> Result<(), Errno> {
-    Errno::from_syscall_unit(unsafe { syscall!(abi::SYS_EX_SIGNAL, argn!(entry), argn!(stack)) })
+    Errno::from_syscall_unit(unsafe {
+        syscall!(SystemCall::SetSignalEntry, argn!(entry), argn!(stack))
+    })
 }
 
 #[inline(always)]
 pub fn sys_ex_sigreturn() -> ! {
     unsafe {
-        syscall!(abi::SYS_EX_SIGRETURN);
+        syscall!(SystemCall::SignalReturn);
     }
     unreachable!();
 }
@@ -264,13 +281,41 @@ pub fn sys_ex_sigreturn() -> ! {
 pub fn sys_ex_kill(pid: SignalDestination, signum: Signal) -> Result<(), Errno> {
     Errno::from_syscall_unit(unsafe {
         syscall!(
-            abi::SYS_EX_KILL,
+            SystemCall::SendSignal,
             argn!(isize::from(pid)),
             argn!(signum as u32)
         )
     })
 }
 
+#[inline(always)]
+pub fn sys_ex_clone(entry: usize, stack: usize, arg: usize) -> Result<usize, Errno> {
+    Errno::from_syscall(unsafe {
+        syscall!(SystemCall::Clone, argn!(entry), argn!(stack), argn!(arg))
+    })
+}
+
+#[inline(always)]
+pub fn sys_ex_thread_exit(status: ExitCode) -> ! {
+    unsafe {
+        syscall!(SystemCall::Exit, argn!(i32::from(status)), argn!(1));
+    }
+    unreachable!();
+}
+
+#[inline(always)]
+pub fn sys_ex_thread_wait(tid: u32) -> Result<ExitCode, Errno> {
+    Errno::from_syscall(unsafe { syscall!(SystemCall::WaitTid, argn!(tid)) })
+        .map(|_| ExitCode::from(0))
+}
+
+#[inline(always)]
+pub fn sys_ex_yield() {
+    unsafe {
+        syscall!(SystemCall::Yield);
+    }
+}
+
 #[inline(always)]
 pub fn sys_select(
     read_fds: Option<&mut FdSet>,
@@ -279,7 +324,7 @@ pub fn sys_select(
 ) -> Result<usize, Errno> {
     Errno::from_syscall(unsafe {
         syscall!(
-            abi::SYS_SELECT,
+            SystemCall::Select,
             argp!(read_fds
                 .map(|e| e as *mut _)
                 .unwrap_or(core::ptr::null_mut())),
@@ -290,3 +335,142 @@ pub fn sys_select(
         )
     })
 }
+
+#[inline(always)]
+pub fn sys_faccessat(
+    fd: Option<FileDescriptor>,
+    name: &str,
+    mode: AccessMode,
+    flags: u32,
+) -> Result<(), Errno> {
+    Errno::from_syscall_unit(unsafe {
+        syscall!(
+            SystemCall::Access,
+            argn!(FileDescriptor::into_i32(fd)),
+            argp!(name.as_ptr()),
+            argn!(name.len()),
+            argn!(mode.bits()),
+            argn!(flags)
+        )
+    })
+}
+
+#[inline(always)]
+pub fn sys_ex_gettid() -> u32 {
+    unsafe { syscall!(SystemCall::GetTid) as u32 }
+}
+
+#[inline(always)]
+pub fn sys_getpid() -> Pid {
+    unsafe { Pid::from_raw(syscall!(SystemCall::GetPid) as u32) }
+}
+
+#[inline(always)]
+pub fn sys_getpgid(pid: Pid) -> Result<Pid, Errno> {
+    Errno::from_syscall(unsafe { syscall!(SystemCall::GetPgid, argn!(pid.value())) })
+        .map(|e| unsafe { Pid::from_raw(e as u32) })
+}
+
+#[inline(always)]
+pub fn sys_setpgid(pid: Pid, pgid: Pid) -> Result<Pid, Errno> {
+    Errno::from_syscall(unsafe {
+        syscall!(SystemCall::SetPgid, argn!(pid.value()), argn!(pgid.value()))
+    })
+    .map(|e| unsafe { Pid::from_raw(e as u32) })
+}
+
+#[inline(always)]
+pub fn sys_readdir(fd: FileDescriptor, buf: &mut [DirectoryEntry]) -> Result<usize, Errno> {
+    Errno::from_syscall(unsafe {
+        syscall!(
+            SystemCall::ReadDirectory,
+            argn!(u32::from(fd)),
+            argp!(buf.as_mut_ptr()),
+            argn!(buf.len())
+        )
+    })
+}
+
+#[inline(always)]
+pub fn sys_getuid() -> UserId {
+    UserId::from(unsafe { syscall!(SystemCall::GetUserId) as u32 })
+}
+
+#[inline(always)]
+pub fn sys_getgid() -> GroupId {
+    GroupId::from(unsafe { syscall!(SystemCall::GetGroupId) as u32 })
+}
+
+#[inline(always)]
+pub fn sys_setsid() -> Result<Pid, Errno> {
+    Errno::from_syscall(unsafe { syscall!(SystemCall::SetSid) })
+        .map(|e| unsafe { Pid::from_raw(e as u32) })
+}
+
+#[inline(always)]
+pub fn sys_mount(target: &str, options: &MountOptions) -> Result<(), Errno> {
+    Errno::from_syscall_unit(unsafe {
+        syscall!(
+            SystemCall::Mount,
+            argp!(target.as_ptr()),
+            argn!(target.len()),
+            argp!(options as *const _)
+        )
+    })
+}
+
+#[inline(always)]
+pub fn sys_dup(src: FileDescriptor, dst: Option<FileDescriptor>) -> Result<FileDescriptor, Errno> {
+    Errno::from_syscall(unsafe {
+        syscall!(
+            SystemCall::DuplicateFd,
+            argn!(u32::from(src)),
+            argn!(FileDescriptor::into_i32(dst))
+        )
+    })
+    .map(|e| FileDescriptor::from(e as u32))
+}
+
+#[inline(always)]
+pub fn sys_setuid(uid: UserId) -> Result<(), Errno> {
+    Errno::from_syscall_unit(unsafe { syscall!(SystemCall::SetUserId, u32::from(uid) as usize) })
+}
+
+#[inline(always)]
+pub fn sys_setgid(gid: GroupId) -> Result<(), Errno> {
+    Errno::from_syscall_unit(unsafe { syscall!(SystemCall::SetGroupId, u32::from(gid) as usize) })
+}
+
+#[inline(always)]
+pub fn sys_chdir(path: &str) -> Result<(), Errno> {
+    Errno::from_syscall_unit(unsafe {
+        syscall!(
+            SystemCall::SetCurrentDirectory,
+            argp!(path.as_ptr()),
+            argn!(path.len())
+        )
+    })
+}
+
+#[inline(always)]
+pub fn sys_mmap(
+    hint: usize,
+    len: usize,
+    acc: MemoryAccess,
+    flags: MemoryMap,
+) -> Result<usize, Errno> {
+    Errno::from_syscall(unsafe {
+        syscall!(
+            SystemCall::MapMemory,
+            argn!(hint),
+            argn!(len),
+            argn!(acc.bits()),
+            argn!(flags.bits())
+        )
+    })
+}
+
+#[inline(always)]
+pub unsafe fn sys_munmap(addr: usize, len: usize) -> Result<(), Errno> {
+    Errno::from_syscall_unit(syscall!(SystemCall::UnmapMemory, argn!(addr), argn!(len)))
+}

libsys/src/debug.rs

@@ -0,0 +1,10 @@
+use enum_repr::EnumRepr;
+
+#[EnumRepr(type = "usize")]
+#[derive(Clone, Copy, Debug, PartialEq, Eq, PartialOrd, Ord)]
+pub enum TraceLevel {
+    Debug = 1,
+    Info = 2,
+    Warn = 3,
+    Error = 4,
+}

libsys/src/error.rs

@@ -15,6 +15,7 @@ pub enum Errno {
     NotADirectory,
     NotImplemented,
     OutOfMemory,
+    PermissionDenied,
     ReadOnly,
     TimedOut,
     TooManyDescriptors,
@@ -47,6 +48,6 @@ impl Errno {
 
 impl From<usize> for Errno {
     fn from(u: usize) -> Errno {
-        todo!()
+        unsafe { core::mem::transmute(u as u32) }
     }
 }

libsys/src/ioctl.rs

@@ -7,6 +7,7 @@ use crate::error::Errno;
 pub enum IoctlCmd {
     TtySetAttributes = 1,
     TtyGetAttributes = 2,
+    TtySetPgrp = 3,
 }
 
 impl TryFrom<u32> for IoctlCmd {
@@ -17,6 +18,7 @@ impl TryFrom<u32> for IoctlCmd {
         match u {
             1 => Ok(Self::TtySetAttributes),
             2 => Ok(Self::TtyGetAttributes),
+            3 => Ok(Self::TtySetPgrp),
             _ => Err(Errno::InvalidArgument)
         }
     }

libsys/src/lib.rs

@@ -5,6 +5,7 @@
 extern crate bitflags;
 
 pub mod abi;
+pub mod debug;
 pub mod error;
 pub mod ioctl;
 pub mod mem;
@@ -15,6 +16,14 @@ pub mod stat;
 pub mod termios;
 pub mod traits;
 
+#[derive(Debug)]
+pub struct ProgramArgs {
+    pub argv: usize,
+    pub argc: usize,
+    pub storage: usize,
+    pub size: usize
+}
+
 #[cfg(feature = "user")]
 pub mod calls;
 #[cfg(feature = "user")]

libsys/src/mem.rs

@@ -13,7 +13,7 @@ pub fn read_le16(src: &[u8]) -> u16 {
 /// Unsafe: writes to arbitrary memory locations, performs no pointer
 /// validation.
 #[no_mangle]
-pub unsafe extern "C" fn memcpy(dst: *mut u8, src: *mut u8, mut len: usize) -> *mut u8 {
+pub unsafe extern "C" fn memcpy(dst: *mut u8, src: *const u8, mut len: usize) -> *mut u8 {
     while len != 0 {
         len -= 1;
         *dst.add(len) = *src.add(len);
@@ -28,15 +28,16 @@ pub unsafe extern "C" fn memcpy(dst: *mut u8, src: *mut u8, mut len: usize) -> *
 /// Unsafe: performs reads from arbitrary memory locations, performs no
 /// pointer validation.
 #[no_mangle]
-pub unsafe extern "C" fn memcmp(a: *mut u8, b: *mut u8, mut len: usize) -> isize {
-    while len != 0 {
-        len -= 1;
-        if *a.add(len) < *b.add(len) {
+pub unsafe extern "C" fn memcmp(a: *mut u8, b: *mut u8, len: usize) -> isize {
+    let mut off = 0;
+    while off != len {
+        if *a.add(off) < *b.add(off) {
             return -1;
         }
-        if *a.add(len) > *b.add(len) {
+        if *a.add(off) > *b.add(off) {
             return 1;
         }
+        off += 1;
     }
     0
 }

libsys/src/proc.rs

@@ -16,6 +16,24 @@ pub struct Pid(u32);
 #[repr(transparent)]
 pub struct Pgid(u32);
 
+bitflags! {
+    pub struct MemoryAccess: u32 {
+        const READ = 1 << 0;
+        const WRITE = 1 << 1;
+        const EXEC = 1 << 2;
+    }
+}
+
+bitflags! {
+    pub struct MemoryMap: u32 {
+        const BACKEND = 0x3 << 0;
+        const ANONYMOUS = 1 << 0;
+
+        const SHARING = 0x3 << 2;
+        const PRIVATE = 1 << 2;
+    }
+}
+
 impl From<i32> for ExitCode {
     fn from(f: i32) -> Self {
         Self(f)

libsys/src/stat.rs

@@ -1,5 +1,6 @@
-use core::fmt;
+// TODO split up this file
 use crate::error::Errno;
+use core::fmt;
 
 const AT_FDCWD: i32 = -2;
 pub const AT_EMPTY_PATH: u32 = 1 << 16;
@@ -14,11 +15,18 @@ bitflags! {
         const O_CREAT =     1 << 4;
         const O_EXEC =      1 << 5;
         const O_CLOEXEC =   1 << 6;
+        const O_DIRECTORY = 1 << 7;
+        const O_CTTY =      1 << 8;
     }
 }
 
 bitflags! {
     pub struct FileMode: u32 {
+        const FILE_TYPE = 0xF << 12;
+        const S_IFREG = 0x8 << 12;
+        const S_IFDIR = 0x4 << 12;
+        const S_IFCHR = 0x2 << 12;
+
         const USER_READ = 1 << 8;
         const USER_WRITE = 1 << 7;
         const USER_EXEC = 1 << 6;
@@ -31,28 +39,131 @@ bitflags! {
     }
 }
 
+bitflags! {
+    pub struct AccessMode: u32 {
+        const R_OK = 1 << 0;
+        const W_OK = 1 << 1;
+        const X_OK = 1 << 2;
+        const F_OK = 1 << 3;
+    }
+}
+
+#[derive(Clone, Debug)]
+pub struct MountOptions<'a> {
+    pub device: Option<&'a str>,
+    pub fs: Option<&'a str>,
+    // TODO flags etc.
+}
+
+#[derive(Clone, Copy, Debug, PartialEq, Eq, PartialOrd, Ord)]
+#[repr(transparent)]
+pub struct UserId(u32);
+
+#[derive(Clone, Copy, Debug, PartialEq, Eq, PartialOrd, Ord)]
+#[repr(transparent)]
+pub struct GroupId(u32);
+
+impl UserId {
+    pub const fn root() -> Self {
+        Self(0)
+    }
+
+    pub const fn is_root(self) -> bool {
+        self.0 == 0
+    }
+}
+
+impl From<u32> for UserId {
+    #[inline(always)]
+    fn from(v: u32) -> Self {
+        Self(v)
+    }
+}
+
+impl From<UserId> for u32 {
+    #[inline(always)]
+    fn from(v: UserId) -> u32 {
+        v.0
+    }
+}
+
+impl GroupId {
+    pub const fn root() -> Self {
+        Self(0)
+    }
+
+    pub const fn is_root(self) -> bool {
+        self.0 == 0
+    }
+}
+
+impl From<u32> for GroupId {
+    #[inline(always)]
+    fn from(v: u32) -> Self {
+        Self(v)
+    }
+}
+
+impl From<GroupId> for u32 {
+    #[inline(always)]
+    fn from(v: GroupId) -> u32 {
+        v.0
+    }
+}
+
 #[derive(Clone, Default)]
 pub struct FdSet {
-    bits: [u64; 2]
+    bits: [u64; 2],
 }
 
 #[derive(Clone, Copy, Debug)]
 #[repr(transparent)]
 pub struct FileDescriptor(u32);
 
+#[derive(Clone, Copy)]
+pub struct DirectoryEntry {
+    name: [u8; 64],
+}
+
 struct FdSetIter<'a> {
     idx: u32,
-    set: &'a FdSet
+    set: &'a FdSet,
 }
 
 #[derive(Clone, Copy, Debug, Default)]
 #[repr(C)]
 pub struct Stat {
-    pub mode: u32,
+    pub mode: FileMode,
     pub size: u64,
     pub blksize: u32,
 }
 
+impl DirectoryEntry {
+    pub const fn empty() -> Self {
+        Self { name: [0; 64] }
+    }
+
+    pub fn from_str(i: &str) -> DirectoryEntry {
+        let mut res = DirectoryEntry { name: [0; 64] };
+        let bytes = i.as_bytes();
+        res.name[..bytes.len()].copy_from_slice(bytes);
+        res
+    }
+
+    pub fn as_str(&self) -> &str {
+        let zero = self.name.iter().position(|&c| c == 0).unwrap();
+        core::str::from_utf8(&self.name[..zero]).unwrap()
+    }
+}
+
+impl fmt::Debug for DirectoryEntry {
+    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+        f.debug_struct("DirectoryEntry")
+            .field("name", &self.as_str())
+            .finish()
+    }
+}
+
 impl FdSet {
     pub const fn empty() -> Self {
         Self { bits: [0; 2] }
@@ -84,10 +195,7 @@ impl FdSet {
     }
 
     pub fn iter(&self) -> impl Iterator<Item = FileDescriptor> + '_ {
-        FdSetIter {
-            idx: 0,
-            set: self
-        }
+        FdSetIter { idx: 0, set: self }
     }
 }
 
@@ -122,13 +230,51 @@ impl fmt::Debug for FdSet {
 
 impl FileMode {
     /// Returns default permission set for directories
-    pub const fn default_dir() -> Self {
-        unsafe { Self::from_bits_unchecked(0o755) }
+    pub fn default_dir() -> Self {
+        unsafe { Self::from_bits_unchecked(0o755) | Self::S_IFDIR }
     }
 
     /// Returns default permission set for regular files
-    pub const fn default_reg() -> Self {
-        unsafe { Self::from_bits_unchecked(0o644) }
+    pub fn default_reg() -> Self {
+        unsafe { Self::from_bits_unchecked(0o644) | Self::S_IFREG }
+    }
+}
+
+fn choose<T>(q: bool, a: T, b: T) -> T {
+    if q { a } else { b }
+}
+
+impl Default for FileMode {
+    fn default() -> Self {
+        unsafe { Self::from_bits_unchecked(0) }
+    }
+}
+
+impl fmt::Display for FileMode {
+    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+        write!(
+            f,
+            "{}{}{}{}{}{}{}{}{}{}",
+            // File type
+            match *self & Self::FILE_TYPE {
+                Self::S_IFCHR => 'c',
+                Self::S_IFDIR => 'd',
+                Self::S_IFREG => '-',
+                _ => '?'
+            },
+            // User
+            choose(self.contains(Self::USER_READ), 'r', '-'),
+            choose(self.contains(Self::USER_WRITE), 'w', '-'),
+            choose(self.contains(Self::USER_EXEC), 'x', '-'),
+            // Group
+            choose(self.contains(Self::GROUP_READ), 'r', '-'),
+            choose(self.contains(Self::GROUP_WRITE), 'w', '-'),
+            choose(self.contains(Self::GROUP_EXEC), 'x', '-'),
+            // Other
+            choose(self.contains(Self::OTHER_READ), 'r', '-'),
+            choose(self.contains(Self::OTHER_WRITE), 'w', '-'),
+            choose(self.contains(Self::OTHER_EXEC), 'x', '-'),
+        )
     }
 }
 

libusr/Cargo.toml

@@ -8,3 +8,4 @@ edition = "2021"
 [dependencies]
 libsys = { path = "../libsys", features = ["user"] }
 lazy_static = { version = "^1.4.0", features = ["spin_no_std"] }
+memoffset = "^0.6.4"

libusr/src/allocator.rs

@@ -0,0 +1,262 @@
+use core::alloc::{GlobalAlloc, Layout};
+use core::mem::{size_of, MaybeUninit};
+use core::ptr::null_mut;
+use libsys::{
+    calls::{sys_mmap, sys_munmap},
+    error::Errno,
+    proc::{MemoryAccess, MemoryMap},
+};
+use memoffset::offset_of;
+
+use crate::trace_debug;
+
+struct Allocator;
+
+const BLOCK_MAGIC: u32 = 0xBADB10C0;
+const BLOCK_MAGIC_MASK: u32 = 0xFFFFFFF0;
+const BLOCK_ALLOC: u32 = 1 << 0;
+const SMALL_ZONE_ELEM: usize = 256;
+const SMALL_ZONE_SIZE: usize = 6 * 0x1000;
+const MID_ZONE_ELEM: usize = 2048;
+const MID_ZONE_SIZE: usize = 24 * 0x1000;
+const LARGE_ZONE_ELEM: usize = 8192;
+const LARGE_ZONE_SIZE: usize = 48 * 0x1000;
+
+struct ZoneList {
+    prev: *mut ZoneList,
+    next: *mut ZoneList,
+}
+
+#[repr(C)]
+struct Zone {
+    size: usize,
+    list: ZoneList,
+}
+
+#[repr(C)]
+struct Block {
+    prev: *mut Block,
+    next: *mut Block,
+    flags: u32,
+    size: u32,
+}
+
+static mut SMALL_ZONE_LIST: MaybeUninit<ZoneList> = MaybeUninit::uninit();
+static mut MID_ZONE_LIST: MaybeUninit<ZoneList> = MaybeUninit::uninit();
+static mut LARGE_ZONE_LIST: MaybeUninit<ZoneList> = MaybeUninit::uninit();
+
+impl ZoneList {
+    fn init(&mut self) {
+        self.prev = self;
+        self.next = self;
+    }
+
+    unsafe fn init_uninit(list: &mut MaybeUninit<Self>) {
+        list.assume_init_mut().init()
+    }
+
+    fn add(&mut self, new: *mut ZoneList) {
+        let new = unsafe { &mut *new };
+        let next = unsafe { &mut *self.next };
+
+        next.prev = new;
+        new.next = next;
+        new.prev = self;
+        self.next = new;
+    }
+
+    fn del(&mut self) {
+        let prev = unsafe { &mut *self.prev };
+        let next = unsafe { &mut *self.next };
+
+        next.prev = prev;
+        prev.next = next;
+    }
+}
+
+impl Zone {
+    fn alloc(size: usize) -> Result<*mut Self, Errno> {
+        let pages = sys_mmap(
+            0,
+            size,
+            MemoryAccess::READ | MemoryAccess::WRITE,
+            MemoryMap::ANONYMOUS | MemoryMap::PRIVATE,
+        )?;
+        trace_debug!("Zone::alloc({}) => {:#x}", size, pages);
+
+        let zone_ptr = pages as *mut Zone;
+        let head_ptr = (pages + size_of::<Zone>()) as *mut Block;
+
+        let zone = unsafe { &mut *zone_ptr };
+        let head = unsafe { &mut *head_ptr };
+        zone.list.init();
+        zone.size = size - size_of::<Zone>();
+
+        head.size = (size - (size_of::<Zone>() + size_of::<Block>())) as u32;
+        head.flags = BLOCK_MAGIC;
+        head.prev = null_mut();
+        head.next = null_mut();
+
+        Ok(zone)
+    }
+
+    unsafe fn free(zone: *mut Self) {
+        trace_debug!("Zone::free({:p})", zone);
+        sys_munmap(zone as usize, (&*zone).size + size_of::<Zone>())
+            .expect("Failed to unmap heap pages");
+    }
+
+    fn get(item: *mut ZoneList) -> *mut Zone {
+        ((item as usize) - offset_of!(Zone, list)) as *mut Zone
+    }
+}
+
+unsafe fn zone_alloc(zone: &mut Zone, size: usize) -> *mut u8 {
+    assert_eq!(size & 15, 0);
+
+    let begin = ((zone as *mut _ as usize) + size_of::<Zone>()) as *mut Block;
+
+    let mut block = begin;
+    while !block.is_null() {
+        let block_ref = &mut *block;
+        if block_ref.flags & BLOCK_ALLOC != 0 {
+            block = block_ref.next;
+            continue;
+        }
+
+        if size == block_ref.size as usize {
+            block_ref.flags |= BLOCK_ALLOC;
+            let ptr = block.add(1) as *mut u8;
+            // TODO fill with zeros
+            return ptr;
+        } else if block_ref.size as usize >= size + size_of::<Block>() {
+            let cur_next = block_ref.next;
+            let cur_next_ref = &mut *cur_next;
+            let new_block = ((block as usize) + size_of::<Block>() + size) as *mut Block;
+            let new_block_ref = &mut *new_block;
+
+            if !cur_next.is_null() {
+                cur_next_ref.prev = new_block;
+            }
+            new_block_ref.next = cur_next;
+            new_block_ref.prev = block;
+            new_block_ref.size = ((block_ref.size as usize) - size_of::<Block>() - size) as u32;
+            new_block_ref.flags = BLOCK_MAGIC;
+            block_ref.next = new_block;
+            block_ref.size = size as u32;
+            block_ref.flags |= BLOCK_ALLOC;
+
+            let ptr = block.add(1) as *mut u8;
+            // TODO fill with zeros
+            return ptr;
+        }
+
+        block = block_ref.next;
+    }
+
+    null_mut()
+}
+
+unsafe fn alloc_from(list: &mut ZoneList, zone_size: usize, size: usize) -> *mut u8 {
+    loop {
+        let mut zone = list.next;
+        while zone != list {
+            let ptr = zone_alloc(&mut *Zone::get(zone), size);
+            if !ptr.is_null() {
+                return ptr;
+            }
+            zone = (&mut *zone).next;
+        }
+
+        let zone = match Zone::alloc(zone_size) {
+            Ok(zone) => zone,
+            Err(e) => {
+                trace_debug!("Zone alloc failed: {:?}", e);
+                return null_mut();
+            }
+        };
+        list.add(&mut (&mut *zone).list);
+    }
+}
+
+unsafe impl GlobalAlloc for Allocator {
+    unsafe fn alloc(&self, layout: Layout) -> *mut u8 {
+        assert!(layout.align() < 16);
+        let size = (layout.size() + 15) & !15;
+        trace_debug!("alloc({:?})", layout);
+        if size <= SMALL_ZONE_ELEM {
+            alloc_from(SMALL_ZONE_LIST.assume_init_mut(), SMALL_ZONE_SIZE, size)
+        } else if size <= MID_ZONE_ELEM {
+            alloc_from(MID_ZONE_LIST.assume_init_mut(), MID_ZONE_SIZE, size)
+        } else if size <= LARGE_ZONE_ELEM {
+            alloc_from(LARGE_ZONE_LIST.assume_init_mut(), LARGE_ZONE_SIZE, size)
+        } else {
+            todo!();
+        }
+    }
+
+    unsafe fn dealloc(&self, ptr: *mut u8, layout: Layout) {
+        trace_debug!("free({:p}, {:?})", ptr, layout);
+        assert!(!ptr.is_null());
+        let mut block = ptr.sub(size_of::<Block>()) as *mut Block;
+        let mut block_ref = &mut *block;
+
+        if block_ref.flags & BLOCK_MAGIC_MASK != BLOCK_MAGIC {
+            panic!("Heap block is malformed: block={:p}, ptr={:p}", block, ptr);
+        }
+        if block_ref.flags & BLOCK_ALLOC == 0 {
+            panic!(
+                "Double free error in heap: block={:p}, ptr={:p}",
+                block, ptr
+            );
+        }
+
+        block_ref.flags &= !BLOCK_ALLOC;
+        let prev = block_ref.prev;
+        let next = block_ref.next;
+        let prev_ref = &mut *prev;
+        let next_ref = &mut *next;
+
+        if !prev.is_null() && prev_ref.flags & BLOCK_ALLOC == 0 {
+            block_ref.flags = 0;
+            prev_ref.next = next;
+            if !next.is_null() {
+                next_ref.prev = prev;
+            }
+            prev_ref.size += (block_ref.size as usize + size_of::<Block>()) as u32;
+
+            block = prev;
+            block_ref = &mut *block;
+        }
+
+        if !next.is_null() && next_ref.flags & BLOCK_ALLOC == 0 {
+            next_ref.flags = 0;
+            if !next_ref.next.is_null() {
+                (&mut *(next_ref.next)).prev = block;
+            }
+            block_ref.next = next_ref.next;
+            block_ref.size += (next_ref.size as usize + size_of::<Block>()) as u32;
+        }
+
+        if block_ref.prev.is_null() && block_ref.next.is_null() {
+            let zone = (block as usize - size_of::<Zone>()) as *mut Zone;
+            assert_eq!((zone as usize) & 0xFFF, 0);
+            (&mut *zone).list.del();
+            Zone::free(zone);
+        }
+    }
+}
+
+#[alloc_error_handler]
+fn alloc_error_handler(_layout: Layout) -> ! {
+    loop {}
+}
+
+#[global_allocator]
+static ALLOC: Allocator = Allocator;
+
+pub unsafe fn init() {
+    ZoneList::init_uninit(&mut SMALL_ZONE_LIST);
+    ZoneList::init_uninit(&mut MID_ZONE_LIST);
+    ZoneList::init_uninit(&mut LARGE_ZONE_LIST);
+}

libusr/src/env.rs

@@ -0,0 +1,21 @@
+use libsys::{debug::TraceLevel, ProgramArgs};
+use alloc::vec::Vec;
+use crate::trace;
+
+static mut PROGRAM_ARGS: Vec<&'static str> = Vec::new();
+
+pub fn args() -> &'static [&'static str] {
+    unsafe { &PROGRAM_ARGS }
+}
+
+pub(crate) unsafe fn setup_env(arg: &ProgramArgs) {
+    for i in 0..arg.argc {
+        let base = core::ptr::read((arg.argv + i * 16 + 0) as *const *const u8);
+        let len = core::ptr::read((arg.argv + i * 16 + 8) as *const usize);
+
+        let string = core::str::from_utf8(core::slice::from_raw_parts(base, len)).unwrap();
+        PROGRAM_ARGS.push(string);
+    }
+
+    trace!(TraceLevel::Debug, "args = {:?}", PROGRAM_ARGS);
+}

libusr/src/file.rs

@@ -1,7 +1,8 @@
-use crate::io::{AsRawFd, Error};
-use crate::os;
-use crate::trace;
-use libsys::stat::FileDescriptor;
+use crate::io::{AsRawFd, Error, Read};
+use libsys::{
+    calls::{sys_openat, sys_read, sys_close},
+    stat::{FileDescriptor, FileMode, OpenFlags},
+};
 
 pub struct File {
     fd: FileDescriptor,
@@ -9,7 +10,9 @@ pub struct File {
 
 impl File {
     pub fn open(path: &str) -> Result<File, Error> {
-        todo!()
+        let fd = sys_openat(None, path, FileMode::default_reg(), OpenFlags::O_RDONLY)
+            .map_err(Error::from)?;
+        Ok(File { fd })
     }
 }
 
@@ -21,6 +24,12 @@ impl AsRawFd for File {
 
 impl Drop for File {
     fn drop(&mut self) {
-        todo!();
+        sys_close(self.fd).ok();
+    }
+}
+
+impl Read for File {
+    fn read(&mut self, bytes: &mut [u8]) -> Result<usize, Error> {
+        sys_read(self.fd, bytes).map_err(Error::from)
     }
 }

libusr/src/io/error.rs

@@ -2,6 +2,7 @@ use libsys::error::Errno;
 
 #[derive(Debug)]
 pub struct Error {
+    #[allow(dead_code)]
     repr: Repr,
 }
 

libusr/src/io/mod.rs

@@ -1,7 +1,11 @@
 use libsys::{
-    calls::sys_fstatat,
+    calls::{sys_fstatat, sys_ioctl},
     stat::{FileDescriptor, Stat},
+    ioctl::IoctlCmd,
+    error::Errno,
+    proc::Pid
 };
+use core::mem::size_of;
 use core::fmt;
 
 mod error;
@@ -24,9 +28,17 @@ pub trait AsRawFd {
     fn as_raw_fd(&self) -> FileDescriptor;
 }
 
+pub fn tcgetpgrp(_fd: FileDescriptor) -> Result<Pid, Errno> {
+    todo!()
+}
+
+pub fn tcsetpgrp(fd: FileDescriptor, pgid: Pid) -> Result<(), Errno> {
+    sys_ioctl(fd, IoctlCmd::TtySetPgrp, &pgid as *const _ as usize, size_of::<Pid>()).map(|_| ())
+}
+
 pub fn stat(pathname: &str) -> Result<Stat, Error> {
     let mut buf = Stat::default();
     // TODO error handling
-    let res = sys_fstatat(None, pathname, &mut buf, 0).unwrap();
+    sys_fstatat(None, pathname, &mut buf, 0).unwrap();
     Ok(buf)
 }

libusr/src/io/stdio.rs

@@ -1,40 +1,40 @@
-use libsys::{
-    stat::FileDescriptor,
-    calls::{sys_read, sys_write}
-};
-use crate::io::{Read, Write, Error};
-use crate::sync::{Mutex, MutexGuard};
+use crate::io::{Error, Read, Write};
+use crate::sync::Mutex;
 use core::fmt;
+use libsys::{
+    calls::{sys_read, sys_write},
+    stat::FileDescriptor,
+};
 
 struct InputInner {
-    fd: FileDescriptor
+    fd: FileDescriptor,
 }
 struct OutputInner {
-    fd: FileDescriptor
+    fd: FileDescriptor,
 }
 
-pub struct StdinLock<'a> {
-    lock: MutexGuard<'a, InputInner>
-}
-
-pub struct StdoutLock<'a> {
-    lock: MutexGuard<'a, OutputInner>
-}
-
-pub struct StderrLock<'a> {
-    lock: MutexGuard<'a, OutputInner>
-}
+//pub struct StdinLock<'a> {
+//    lock: MutexGuard<'a, InputInner>
+//}
+//
+//pub struct StdoutLock<'a> {
+//    lock: MutexGuard<'a, OutputInner>
+//}
+//
+//pub struct StderrLock<'a> {
+//    lock: MutexGuard<'a, OutputInner>
+//}
 
 pub struct Stdin {
     inner: &'static Mutex<InputInner>,
 }
 
 pub struct Stdout {
-    inner: &'static Mutex<OutputInner>
+    inner: &'static Mutex<OutputInner>,
 }
 
 pub struct Stderr {
-    inner: &'static Mutex<OutputInner>
+    inner: &'static Mutex<OutputInner>,
 }
 
 // STDIN
@@ -51,6 +51,14 @@ impl Read for Stdin {
     }
 }
 
+// impl Stdin {
+//     pub fn lock(&self) -> StdinLock {
+//         StdinLock {
+//             lock: self.inner.lock()
+//         }
+//     }
+// }
+
 // STDOUT/STDERR
 
 impl fmt::Write for OutputInner {
@@ -89,21 +97,21 @@ impl Write for Stderr {
     }
 }
 
-impl Stdout {
-    pub fn lock(&self) -> StdoutLock {
-        StdoutLock {
-            lock: self.inner.lock()
-        }
-    }
-}
-
-impl Stderr {
-    pub fn lock(&self) -> StderrLock {
-        StderrLock {
-            lock: self.inner.lock()
-        }
-    }
-}
+// impl Stdout {
+//     pub fn lock(&self) -> StdoutLock {
+//         StdoutLock {
+//             lock: self.inner.lock()
+//         }
+//     }
+// }
+//
+// impl Stderr {
+//     pub fn lock(&self) -> StderrLock {
+//         StderrLock {
+//             lock: self.inner.lock()
+//         }
+//     }
+// }
 
 lazy_static! {
     static ref STDIN: Mutex<InputInner> = Mutex::new(InputInner {

libusr/src/io/writer.rs

@@ -1,5 +1,5 @@
 use core::fmt;
-use crate::io::{self, Write};
+use crate::io::Write;
 
 #[macro_export]
 macro_rules! print {

libusr/src/lib.rs

@@ -1,39 +1,35 @@
 #![feature(asm, alloc_error_handler)]
 #![no_std]
 
-use core::panic::PanicInfo;
-use libsys::proc::ExitCode;
-
 #[macro_use]
 extern crate lazy_static;
 
+extern crate alloc;
+
+use core::panic::PanicInfo;
+use libsys::{debug::TraceLevel, ProgramArgs, proc::ExitCode};
+
+mod allocator;
+pub mod env;
 pub mod file;
 pub mod io;
 pub mod os;
 pub mod sys;
 pub mod sync;
-
-#[inline(never)]
-extern "C" fn _signal_handler(arg: sys::Signal) -> ! {
-    trace!("Entered signal handler: arg={:?}", arg);
-    sys::sys_ex_sigreturn();
-}
-
-static mut SIGNAL_STACK: [u8; 4096] = [0; 4096];
+pub mod thread;
+pub mod signal;
 
 #[link_section = ".text._start"]
 #[no_mangle]
-extern "C" fn _start(_arg: usize) -> ! {
+extern "C" fn _start(arg: &'static ProgramArgs) -> ! {
     extern "Rust" {
         fn main() -> i32;
     }
 
     unsafe {
-        sys::sys_ex_signal(
-            _signal_handler as usize,
-            SIGNAL_STACK.as_ptr() as usize + 4096,
-        )
-        .unwrap();
+        allocator::init();
+        thread::init_main();
+        env::setup_env(arg);
     }
 
     let res = unsafe { main() };
@@ -42,7 +38,9 @@ extern "C" fn _start(_arg: usize) -> ! {
 
 #[panic_handler]
 fn panic_handler(pi: &PanicInfo) -> ! {
+    // TODO unwind to send panic argument back to parent thread
     // TODO print to stdout/stderr (if available)
-    trace!("Panic ocurred: {}", pi);
+    let thread = thread::current();
+    trace!(TraceLevel::Error, "{:?} panicked: {:?}", thread, pi);
     sys::sys_exit(ExitCode::from(-1));
 }

libusr/src/os.rs

@@ -1,11 +1,16 @@
+use libsys::debug::TraceLevel;
 use crate::sys;
 use core::fmt;
-use core::mem::{size_of, MaybeUninit};
-use libsys::{ioctl::IoctlCmd, stat::FileDescriptor, termios::Termios};
 
 #[macro_export]
 macro_rules! trace {
-    ($($args:tt)+) => ($crate::os::_trace(format_args!($($args)+)))
+    ($level:expr, $($args:tt)+) => ($crate::os::_trace($level, format_args!($($args)+)))
+}
+
+
+#[macro_export]
+macro_rules! trace_debug {
+    ($($args:tt)+) => ($crate::os::_trace($crate::sys::debug::TraceLevel::Debug, format_args!($($args)+)))
 }
 
 struct BufferWriter<'a> {
@@ -23,7 +28,7 @@ impl fmt::Write for BufferWriter<'_> {
     }
 }
 
-pub fn _trace(args: fmt::Arguments) {
+pub fn _trace(level: TraceLevel, args: fmt::Arguments) {
     use core::fmt::Write;
     static mut BUFFER: [u8; 4096] = [0; 4096];
     let mut writer = BufferWriter {
@@ -31,5 +36,5 @@ pub fn _trace(args: fmt::Arguments) {
         pos: 0,
     };
     writer.write_fmt(args).ok();
-    sys::sys_ex_debug_trace(unsafe { &BUFFER[..writer.pos] }).ok();
+    sys::sys_ex_debug_trace(level, unsafe { &BUFFER[..writer.pos] }).ok();
 }

libusr/src/signal.rs

@@ -0,0 +1,42 @@
+use crate::trace;
+use libsys::{
+    debug::TraceLevel,
+    calls::{sys_ex_sigreturn, sys_exit},
+    proc::ExitCode,
+    signal::Signal,
+};
+
+#[derive(Clone, Copy)]
+pub enum SignalHandler {
+    Func(fn(Signal) -> ()),
+    Ignore,
+    Terminate,
+}
+
+// TODO per-thread signal handler table
+static mut SIGNAL_HANDLERS: [SignalHandler; 32] = [SignalHandler::Terminate; 32];
+
+pub fn set_handler(sig: Signal, handler: SignalHandler) -> SignalHandler {
+    unsafe {
+        let old = SIGNAL_HANDLERS[sig as usize];
+        SIGNAL_HANDLERS[sig as usize] = handler;
+        old
+    }
+}
+
+#[inline(never)]
+pub(crate) extern "C" fn signal_handler(arg: Signal) -> ! {
+    // TODO tpidr_el0 is invalidated when entering signal context
+    trace!(TraceLevel::Debug, "Entered signal handler: arg={:?}", arg);
+    let no = arg as usize;
+    if no >= 32 {
+        panic!("Undefined signal number: {}", no);
+    }
+    match unsafe { SIGNAL_HANDLERS[no] } {
+        SignalHandler::Func(f) => f(arg),
+        SignalHandler::Ignore => (),
+        SignalHandler::Terminate => sys_exit(ExitCode::from(-1)),
+    }
+
+    sys_ex_sigreturn();
+}

libusr/src/sys/mod.rs

@@ -1,7 +1,11 @@
 pub use libsys::signal::{Signal, SignalDestination};
+pub use libsys::proc::{self, ExitCode};
 pub use libsys::termios;
+pub use libsys::abi;
 pub use libsys::calls::*;
-pub use libsys::stat::{self, FileDescriptor};
+pub use libsys::stat::{self, AccessMode, FileDescriptor};
+pub use libsys::error::Errno;
+pub use libsys::debug;
 
 use core::sync::atomic::{Ordering, AtomicBool};
 
@@ -20,15 +24,10 @@ impl RawMutex {
         self.inner.compare_exchange_weak(false, true, Ordering::Acquire, Ordering::Relaxed).is_ok()
     }
 
-    #[inline]
-    unsafe fn is_locked(&self) -> bool {
-        self.inner.load(Ordering::Acquire)
-    }
-
     #[inline]
     pub unsafe fn lock(&self) {
         while !self.try_lock() {
-            asm!("nop");
+            sys_ex_yield();
         }
     }
 

libusr/src/thread.rs

@@ -0,0 +1,143 @@
+use crate::signal;
+use alloc::{boxed::Box, sync::Arc, vec};
+use core::any::Any;
+use core::cell::UnsafeCell;
+use core::fmt;
+use core::mem::MaybeUninit;
+use libsys::{
+    calls::{sys_ex_clone, sys_ex_gettid, sys_ex_signal, sys_ex_thread_exit, sys_ex_thread_wait},
+    proc::ExitCode,
+};
+
+struct NativeData<F, T>
+where
+    F: FnOnce() -> T,
+    F: Send + 'static,
+    T: Send + 'static,
+{
+    closure: F,
+    result: ThreadPacket<T>,
+    stack: usize,
+}
+
+#[derive(Clone)]
+pub struct Thread {
+    id: u32,
+}
+
+pub type ThreadResult<T> = Result<T, Box<dyn Any + Send + Sync>>;
+pub type ThreadPacket<T> = Arc<UnsafeCell<MaybeUninit<ThreadResult<T>>>>;
+
+pub struct JoinHandle<T> {
+    native: u32,
+    result: ThreadPacket<T>,
+}
+
+impl Thread {
+    pub const fn id(&self) -> u32 {
+        self.id
+    }
+}
+
+impl fmt::Debug for Thread {
+    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+        f.debug_struct("Thread")
+            .field("id", &self.id)
+            .finish_non_exhaustive()
+    }
+}
+
+impl<T> JoinHandle<T> {
+    pub fn join(self) -> ThreadResult<T> {
+        sys_ex_thread_wait(self.native).unwrap();
+        unsafe {
+            Arc::try_unwrap(self.result)
+                .unwrap()
+                .into_inner()
+                .assume_init()
+        }
+    }
+}
+
+unsafe fn init_common(signal_stack_pointer: *mut u8) {
+    let tid = sys_ex_gettid() as u64;
+    asm!("msr tpidr_el0, {:x}", in(reg) tid);
+
+    // thread::current() should be valid at this point
+
+    sys_ex_signal(
+        signal::signal_handler as usize,
+        signal_stack_pointer as usize,
+    )
+    .unwrap();
+}
+
+pub(crate) unsafe fn init_main() {
+    #[repr(align(16))]
+    struct StackWrapper {
+        data: [u8; 8192],
+    }
+    static mut STACK: StackWrapper = StackWrapper { data: [0; 8192] };
+    init_common(STACK.data.as_mut_ptr().add(8192))
+}
+
+pub fn current() -> Thread {
+    let mut id: u64;
+    unsafe {
+        asm!("mrs {:x}, tpidr_el0", out(reg) id);
+    }
+
+    Thread { id: id as u32 }
+}
+
+pub fn spawn<F, T>(f: F) -> JoinHandle<T>
+where
+    F: FnOnce() -> T,
+    F: Send + 'static,
+    T: Send + 'static,
+{
+    let stack = vec![0u8; 8192].leak();
+    let result = Arc::new(UnsafeCell::new(MaybeUninit::uninit()));
+
+    #[inline(never)]
+    extern "C" fn thread_entry<F, T>(data: *mut NativeData<F, T>) -> !
+    where
+        F: FnOnce() -> T,
+        F: Send + 'static,
+        T: Send + 'static,
+    {
+        let (_stack, _len) = {
+            // Setup signal handling
+            let mut signal_stack = vec![0u8; 8192];
+
+            unsafe {
+                init_common(signal_stack.as_mut_ptr().add(signal_stack.len()));
+            }
+
+            let data: Box<NativeData<F, T>> = unsafe { Box::from_raw(data) };
+            let res = (data.closure)();
+
+            unsafe {
+                (&mut *data.result.get()).write(Ok(res));
+            }
+
+            (data.stack, 8192)
+        };
+
+        // TODO free stack
+        sys_ex_thread_exit(ExitCode::from(0));
+    }
+
+    let native = {
+        let stack = stack.as_mut_ptr() as usize + stack.len();
+        let data: *mut NativeData<F, T> = Box::into_raw(Box::new(NativeData {
+            closure: f,
+            stack,
+            result: result.clone(),
+        }));
+
+        sys_ex_clone(thread_entry::<F, T> as usize, stack, data as usize).unwrap() as u32
+    };
+
+    JoinHandle { native, result }
+}

user/Cargo.toml

@@ -11,7 +11,29 @@ path = "src/init/main.rs"
 
 [[bin]]
 name = "shell"
-path = "src/shell/main.rs"
+path = "src/bin/shell.rs"
+
+[[bin]]
+name = "fuzzy"
+path = "src/bin/fuzzy.rs"
+
+[[bin]]
+name = "ls"
+path = "src/bin/ls.rs"
+
+[[bin]]
+name = "cat"
+path = "src/bin/cat.rs"
+
+[[bin]]
+name = "hexd"
+path = "src/bin/hexd.rs"
+
+[[bin]]
+name = "login"
+path = "src/sbin/login.rs"
 
 [dependencies]
 libusr = { path = "../libusr" }
+libsys = { path = "../libsys" }
+lazy_static = { version = "*", features = ["spin_no_std"] }

user/src/bin/cat.rs

@@ -0,0 +1,46 @@
+#![no_std]
+#![no_main]
+
+#[macro_use]
+extern crate libusr;
+
+use libusr::io::{self, Read, Write};
+use libusr::file::File;
+
+fn do_cat<F: Read>(mut fd: F) -> Result<(), io::Error> {
+    let mut buf = [0; 4096];
+    let mut out = io::stdout();
+
+    loop {
+        let count = fd.read(&mut buf)?;
+        if count == 0 {
+            break;
+        }
+
+        out.write(&buf[..count])?;
+    }
+
+    Ok(())
+}
+
+#[no_mangle]
+fn main() -> i32 {
+    let args = libusr::env::args();
+    let mut res = 0;
+
+    if args.len() == 1 {
+        if let Err(e) = do_cat(io::stdin()) {
+            eprintln!("{}: {:?}", ".", e);
+            res = -1;
+        }
+    } else {
+        for arg in &args[1..] {
+            if let Err(e) = File::open(arg).map(do_cat) {
+                eprintln!("{}: {:?}", arg, e);
+                res = -1;
+            }
+        }
+    }
+
+    res
+}

user/src/bin/fuzzy.rs

@@ -0,0 +1,139 @@
+#![feature(asm)]
+#![no_std]
+#![no_main]
+
+#![allow(unused_macros)]
+#![allow(dead_code)]
+
+#[macro_use]
+extern crate libusr;
+
+use libusr::sys::{abi::SystemCall, stat::Stat};
+
+static mut STATE: u64 = 0;
+
+macro_rules! syscall {
+    ($num:expr) => {{
+        let mut res: usize;
+        asm!("svc #0", out("x0") res, in("x8") $num, options(nostack));
+        res
+    }};
+    ($num:expr, $a0:expr) => {{
+        let mut res: usize = $a0;
+        asm!("svc #0",
+             inout("x0") res,
+             in("x8") $num, options(nostack));
+        res
+    }};
+    ($num:expr, $a0:expr, $a1:expr) => {{
+        let mut res: usize = $a0;
+        asm!("svc #0",
+             inout("x0") res, in("x1") $a1,
+             in("x8") $num, options(nostack));
+        res
+    }};
+    ($num:expr, $a0:expr, $a1:expr, $a2:expr) => {{
+        let mut res: usize = $a0;
+        asm!("svc #0",
+             inout("x0") res, in("x1") $a1, in("x2") $a2,
+             in("x8") $num, options(nostack));
+        res
+    }};
+    ($num:expr, $a0:expr, $a1:expr, $a2:expr, $a3:expr) => {{
+        let mut res: usize = $a0;
+        asm!("svc #0",
+             inout("x0") res, in("x1") $a1, in("x2") $a2,
+             in("x3") $a3, in("x8") $num, options(nostack));
+        res
+    }};
+    ($num:expr, $a0:expr, $a1:expr, $a2:expr, $a3:expr, $a4:expr) => {{
+        let mut res: usize = $a0;
+        asm!("svc #0",
+             inout("x0") res, in("x1") $a1, in("x2") $a2,
+             in("x3") $a3, in("x4") $a4, in("x8") $num, options(nostack));
+        res
+    }};
+}
+
+/// Integer/size argument
+macro_rules! argn {
+    ($a:expr) => {
+        $a as usize
+    };
+}
+/// Pointer/base argument
+macro_rules! argp {
+    ($a:expr) => {
+        $a as usize
+    };
+}
+
+fn random_set_seed(seed: u64) {
+    unsafe { STATE = seed; }
+}
+
+fn random_u64() -> u64 {
+    let mut x = unsafe { STATE };
+    x ^= x << 13;
+    x ^= x >> 7;
+    x ^= x << 17;
+    unsafe {
+        STATE = x;
+    }
+    x
+}
+
+fn random_ascii_char() -> u8 {
+    ((random_u64() % (0x7F - 0x20)) as u8) + 0x20
+}
+
+fn random_str_range(buf: &mut [u8], min: usize, max: usize) -> &str {
+    let max = core::cmp::min(buf.len(), max);
+    assert!(max > min);
+    let len = ((random_u64() as usize) % (max - min)) + min;
+    for c in buf[..len].iter_mut() {
+        *c = random_ascii_char();
+    }
+    core::str::from_utf8(&buf[..len]).unwrap()
+}
+
+fn random_str(buf: &mut [u8]) -> &str {
+    random_str_range(buf, 0, buf.len())
+}
+
+fn random_bytes(buf: &mut [u8]) {
+    for byte in buf.iter_mut() {
+        *byte = (random_u64() & 0xFF) as u8;
+    }
+}
+
+#[no_mangle]
+fn main() -> i32 {
+    let seed = libusr::sys::sys_ex_getcputime().unwrap().as_nanos() as u64 / 13;
+    println!("Using seed: {:#x}", seed);
+    random_set_seed(seed);
+
+    let mut buf = [0; 256];
+
+    // Test sys_ex_getcputime()
+    let mut prev_time = libusr::sys::sys_ex_getcputime().unwrap().as_nanos();
+    for _ in 0..1000 {
+        let t = libusr::sys::sys_ex_getcputime().unwrap().as_nanos();
+        assert!(t >= prev_time);
+        prev_time = t;
+    }
+
+    // Test non-utf8 input fed into syscalls expecting strings
+    // let old_signal = signal::set_handler(Signal::InvalidSystemCall, SignalHandler::Ignore);
+    for _ in 0..10000 {
+        random_bytes(&mut buf);
+        let mut stat = Stat::default();
+
+        unsafe {
+            syscall!(SystemCall::FileStatus.repr(), (-2i32) as usize, buf.as_mut_ptr() as usize, buf.len(), (&mut stat) as *mut _ as usize);
+        }
+    }
+    // signal::set_handler(Signal::InvalidSystemCall, old_signal);
+
+    0
+}

user/src/bin/hexd.rs

@@ -0,0 +1,69 @@
+#![no_std]
+#![no_main]
+
+#[macro_use]
+extern crate libusr;
+
+use libusr::io::{self, Read};
+use libusr::file::File;
+
+fn line_print(off: usize, line: &[u8]) {
+    print!("{:08x}: ", off);
+    for i in 0..16 {
+        if i < line.len() {
+            print!("{:02x}", line[i]);
+        } else {
+            print!("  ");
+        }
+        if i % 2 != 0 {
+            print!(" ");
+        }
+    }
+    print!("| ");
+    for &b in line.iter() {
+        if b.is_ascii() && !b.is_ascii_control() {
+            print!("{}", b as char);
+        } else {
+            print!(".");
+        }
+    }
+    println!("");
+}
+
+fn do_hexd<F: Read>(mut fd: F) -> Result<(), io::Error> {
+    let mut buf = [0; 16];
+    let mut off = 0;
+    loop {
+        let count = fd.read(&mut buf)?;
+        if count == 0 {
+            break;
+        }
+
+        line_print(off, &buf[..count]);
+        off += count;
+    }
+
+    Ok(())
+}
+
+#[no_mangle]
+fn main() -> i32 {
+    let args = libusr::env::args();
+    let mut res = 0;
+
+    if args.len() == 1 {
+        if let Err(e) = do_hexd(io::stdin()) {
+            eprintln!("{}: {:?}", ".", e);
+            res = -1;
+        }
+    } else {
+        for arg in &args[1..] {
+            if let Err(e) = File::open(arg).map(do_hexd) {
+                eprintln!("{}: {:?}", arg, e);
+                res = -1;
+            }
+        }
+    }
+
+    res
+}

user/src/bin/ls.rs

@@ -0,0 +1,73 @@
+#![no_std]
+#![no_main]
+
+#[macro_use]
+extern crate libusr;
+#[macro_use]
+extern crate alloc;
+
+use alloc::borrow::ToOwned;
+use libusr::sys::{
+    stat::{DirectoryEntry, FileMode, OpenFlags, Stat},
+    sys_close, sys_fstatat, sys_openat, sys_readdir, Errno,
+};
+
+fn list_directory(path: &str) -> Result<(), Errno> {
+    let mut buffer = [DirectoryEntry::empty(); 8];
+    let mut stat = Stat::default();
+    let mut data = vec![];
+
+    let fd = sys_openat(
+        None,
+        path,
+        FileMode::default_dir(),
+        OpenFlags::O_DIRECTORY | OpenFlags::O_RDONLY,
+    )?;
+
+    loop {
+        let count = sys_readdir(fd, &mut buffer)?;
+        if count == 0 {
+            break;
+        }
+
+        buffer.iter().take(count).for_each(|e| {
+            data.push(e.as_str().to_owned());
+        });
+    }
+
+    data.sort();
+
+    data.iter().for_each(|item| {
+        let stat = sys_fstatat(Some(fd), item, &mut stat, 0).map(|_| &stat);
+        if let Ok(stat) = stat {
+            print!("{} ", stat.mode);
+        } else {
+            print!("?????????? ");
+        }
+        println!("{}", item);
+    });
+
+    sys_close(fd)
+}
+
+#[no_mangle]
+fn main() -> i32 {
+    let args = libusr::env::args();
+    let mut res = 0;
+
+    if args.len() == 1 {
+        if let Err(e) = list_directory(".") {
+            eprintln!("{}: {:?}", ".", e);
+            res = -1;
+        }
+    } else {
+        for arg in &args[1..] {
+            if let Err(e) = list_directory(arg) {
+                eprintln!("{}: {:?}", arg, e);
+                res = -1;
+            }
+        }
+    }
+
+    res
+}

user/src/bin/shell.rs

@@ -0,0 +1,113 @@
+#![no_std]
+#![no_main]
+
+#[macro_use]
+extern crate libusr;
+extern crate alloc;
+
+use alloc::{borrow::ToOwned, vec::Vec};
+use libusr::io::{self, Read};
+use libusr::signal::{self, SignalHandler};
+use libusr::sys::{
+    proc::Pid, sys_chdir, sys_execve, sys_exit, sys_faccessat, sys_fork, sys_getpgid, sys_setpgid,
+    sys_waitpid, AccessMode, Errno, ExitCode, FileDescriptor, Signal,
+};
+
+struct Builtin {
+    func: fn(&[&str]) -> ExitCode,
+    name: &'static str,
+}
+
+fn cmd_cd(args: &[&str]) -> ExitCode {
+    if args.len() != 2 {
+        eprintln!("Usage: cd DIR");
+        ExitCode::from(-1)
+    } else {
+        if let Err(err) = sys_chdir(args[1]) {
+            eprintln!("{}: {:?}", args[1], err);
+            ExitCode::from(-1)
+        } else {
+            ExitCode::from(0)
+        }
+    }
+}
+
+static BUILTINS: [Builtin; 1] = [Builtin {
+    name: "cd",
+    func: cmd_cd,
+}];
+
+fn readline<'a, F: Read>(f: &mut F, bytes: &'a mut [u8]) -> Result<Option<&'a str>, io::Error> {
+    let size = f.read(bytes)?;
+    Ok(if size == 0 {
+        None
+    } else {
+        Some(
+            core::str::from_utf8(&bytes[..size])
+                .unwrap()
+                .trim_end_matches('\n'),
+        )
+    })
+}
+
+fn execute(line: &str) -> Result<ExitCode, Errno> {
+    // TODO proper arg handling
+    let args: Vec<&str> = line.split(' ').collect();
+    let cmd = args[0];
+
+    for item in BUILTINS.iter() {
+        if item.name == cmd {
+            return Ok((item.func)(&args));
+        }
+    }
+
+    let filename = "/bin/".to_owned() + cmd;
+    sys_faccessat(None, &filename, AccessMode::X_OK, 0)?;
+
+    if let Some(pid) = unsafe { sys_fork()? } {
+        let mut status = 0;
+        sys_waitpid(pid, &mut status)?;
+        let pgid = sys_getpgid(unsafe { Pid::from_raw(0) }).unwrap();
+        io::tcsetpgrp(FileDescriptor::STDIN, pgid).unwrap();
+        Ok(ExitCode::from(status))
+    } else {
+        let pgid = sys_setpgid(unsafe { Pid::from_raw(0) }, unsafe { Pid::from_raw(0) }).unwrap();
+        io::tcsetpgrp(FileDescriptor::STDIN, pgid).unwrap();
+        sys_execve(&filename, &args).unwrap();
+        sys_exit(ExitCode::from(-1));
+    }
+}
+
+#[no_mangle]
+fn main() -> i32 {
+    let mut buf = [0; 256];
+    let mut stdin = io::stdin();
+
+    signal::set_handler(Signal::Interrupt, SignalHandler::Ignore);
+    let pgid = sys_setpgid(unsafe { Pid::from_raw(0) }, unsafe { Pid::from_raw(0) }).unwrap();
+    io::tcsetpgrp(FileDescriptor::STDIN, pgid).unwrap();
+
+    loop {
+        print!("> ");
+        match readline(&mut stdin, &mut buf) {
+            Ok(line) => {
+                if line.is_none() {
+                    break;
+                }
+                let line = line.unwrap().trim_start_matches(' ');
+                if line.is_empty() {
+                    continue;
+                }
+
+                if let Err(e) = execute(line) {
+                    eprintln!("{}: {:?}", line.split(' ').next().unwrap(), e);
+                }
+            }
+            Err(_) => {
+                println!("Interrupt!");
+                continue;
+            }
+        }
+    }
+    0
+}

user/src/init/main.rs

@@ -5,13 +5,22 @@
 #[macro_use]
 extern crate libusr;
 
+use libusr::sys::{stat::MountOptions, sys_execve, sys_fork, sys_mount, sys_waitpid};
+
 #[no_mangle]
 fn main() -> i32 {
-    let pid = libusr::sys::sys_fork().unwrap();
+    sys_mount(
+        "/dev",
+        &MountOptions {
+            device: None,
+            fs: Some("devfs"),
+        },
+    )
+    .expect("Failed to mount devfs");
 
-    if let Some(pid) = pid {
+    if let Some(pid) = unsafe { sys_fork().unwrap() } {
         let mut status = 0;
-        libusr::sys::sys_waitpid(pid, &mut status).unwrap();
+        sys_waitpid(pid, &mut status).unwrap();
         println!("Process {:?} exited with status {}", pid, status);
 
         loop {
@@ -20,7 +29,7 @@ fn main() -> i32 {
             }
         }
     } else {
-        libusr::sys::sys_execve("/bin/shell").unwrap();
+        sys_execve("/sbin/login", &["/sbin/login", "/dev/ttyS0"]).unwrap();
         loop {}
     }
 }

user/src/sbin/login.rs

@@ -0,0 +1,147 @@
+#![no_std]
+#![no_main]
+
+#[macro_use]
+extern crate libusr;
+
+use libsys::{
+    calls::{
+        sys_close, sys_dup, sys_fork, sys_getgid, sys_getpgid, sys_getuid, sys_ioctl, sys_openat,
+        sys_read, sys_setgid, sys_setpgid, sys_setsid, sys_setuid, sys_waitpid, sys_execve
+    },
+    error::Errno,
+    ioctl::IoctlCmd,
+    proc::Pid,
+    stat::{FileDescriptor, FileMode, GroupId, OpenFlags, UserId},
+    termios::{Termios, TermiosLflag},
+};
+use libusr::{env, io};
+
+struct HiddenInput {
+    fd: FileDescriptor,
+    termios: Termios,
+}
+
+impl HiddenInput {
+    fn open(fd: FileDescriptor) -> Result<Self, Errno> {
+        use core::mem::{size_of, MaybeUninit};
+        let mut termios: MaybeUninit<Termios> = MaybeUninit::uninit();
+        sys_ioctl(
+            fd,
+            IoctlCmd::TtyGetAttributes,
+            termios.as_mut_ptr() as usize,
+            size_of::<Termios>(),
+        )?;
+        let termios = unsafe { termios.assume_init() };
+
+        let mut new_termios = termios.clone();
+        new_termios.lflag &= !(TermiosLflag::ECHO | TermiosLflag::ECHOK | TermiosLflag::ECHOE);
+        sys_ioctl(
+            fd,
+            IoctlCmd::TtySetAttributes,
+            &new_termios as *const _ as usize,
+            size_of::<Termios>(),
+        )?;
+
+        Ok(Self { fd, termios })
+    }
+
+    fn readline<'a>(&mut self, buf: &'a mut [u8]) -> Result<&'a str, Errno> {
+        readline(self.fd, buf)
+    }
+}
+
+impl Drop for HiddenInput {
+    fn drop(&mut self) {
+        use core::mem::size_of;
+        sys_ioctl(
+            self.fd,
+            IoctlCmd::TtySetAttributes,
+            &self.termios as *const _ as usize,
+            size_of::<Termios>(),
+        )
+        .ok();
+    }
+}
+
+fn readline(fd: FileDescriptor, buf: &mut [u8]) -> Result<&str, Errno> {
+    let len = sys_read(fd, buf)?;
+
+    if len == 0 {
+        Ok("")
+    } else {
+        Ok(core::str::from_utf8(&buf[..len - 1]).unwrap())
+    }
+}
+
+fn login_as(uid: UserId, gid: GroupId, shell: &str) -> Result<(), Errno> {
+    if let Some(pid) = unsafe { sys_fork() }? {
+        let mut status = 0;
+        sys_waitpid(pid, &mut status).ok();
+        let pgid = sys_getpgid(unsafe { Pid::from_raw(0) }).unwrap();
+        io::tcsetpgrp(FileDescriptor::STDIN, pgid).unwrap();
+        Ok(())
+    } else {
+        sys_setuid(uid).expect("setuid failed");
+        sys_setgid(gid).expect("setgid failed");
+        let pgid = sys_setpgid(unsafe { Pid::from_raw(0) }, unsafe { Pid::from_raw(0) }).unwrap();
+        io::tcsetpgrp(FileDescriptor::STDIN, pgid).unwrap();
+        sys_execve(shell, &[shell]).expect("execve() failed");
+        panic!();
+    }
+}
+
+// TODO baud rate and misc port settings
+#[no_mangle]
+fn main() -> i32 {
+    if !sys_getuid().is_root() || !sys_getgid().is_root() {
+        panic!("This program must be run as root");
+    }
+
+    let args = env::args();
+    if args.len() != 2 {
+        panic!("Usage: {} TTY", args[0]);
+    }
+
+    sys_setsid().expect("setsid() failed");
+
+    // Close controlling terminal
+    // NOTE this will invalidate rust-side Stdin, Stdout, Stderr
+    //      until replacement is re-opened using the specified TTY
+    sys_close(FileDescriptor::STDERR).unwrap();
+    sys_close(FileDescriptor::STDOUT).unwrap();
+    sys_close(FileDescriptor::STDIN).unwrap();
+
+    sys_openat(
+        None,
+        args[1],
+        FileMode::default_reg(),
+        OpenFlags::O_RDONLY | OpenFlags::O_CTTY,
+    )
+    .expect("Failed to open stdin");
+    sys_openat(
+        None,
+        args[1],
+        FileMode::default_reg(),
+        OpenFlags::O_WRONLY | OpenFlags::O_CTTY,
+    )
+    .expect("Failed to open stdout");
+    sys_dup(FileDescriptor::STDOUT, Some(FileDescriptor::STDERR)).expect("Failed to open stderr");
+
+    let mut user_buf = [0; 128];
+    let mut password_buf = [0; 128];
+    loop {
+        print!("login: ");
+        let username = readline(FileDescriptor::STDIN, &mut user_buf).expect("Login read failed");
+        print!("password: ");
+        let password = {
+            let mut input = HiddenInput::open(FileDescriptor::STDIN).unwrap();
+            input.readline(&mut password_buf)
+        }
+        .expect("Password read failed");
+
+        if username == "root" && password == "toor" {
+            login_as(UserId::from(0), GroupId::from(0), "/bin/shell").unwrap();
+        }
+    }
+}

user/src/shell/main.rs

@@ -1,26 +0,0 @@
-#![no_std]
-#![no_main]
-
-#[macro_use]
-extern crate libusr;
-
-use libusr::io::{self, Read};
-
-#[no_mangle]
-fn main() -> i32 {
-    let mut buf = [0; 512];
-    let mut stdin = io::stdin();
-
-    eprintln!("stderr test");
-
-    loop {
-        let count = stdin.read(&mut buf).unwrap();
-        if count == 0 {
-            break;
-        }
-        let line = core::str::from_utf8(&buf[..count]).unwrap();
-        println!("{:?}", line);
-    }
-
-    0
-}