Almost forking, UB

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",
@@ -82,9 +93,11 @@ dependencies = [
  "cfg-if",
  "cortex-a",
  "fdt-rs",
+ "fs-macros",
  "kernel-macros",
  "libsys",
  "memfs",
+ "multiboot2",
  "tock-registers",
  "vfs",
 ]
@@ -111,6 +124,7 @@ name = "libsys"
 version = "0.1.0"
 dependencies = [
  "bitflags",
+ "enum-repr",
 ]
 
 [[package]]
@@ -119,6 +133,7 @@ version = "0.1.0"
 dependencies = [
  "lazy_static",
  "libsys",
+ "memoffset 0.6.4",
 ]
 
 [[package]]
@@ -139,6 +154,23 @@ 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 = "multiboot2"
+version = "0.12.2"
+source = "git+https://github.com/alnyan/multiboot2?branch=expose-extra-traits-for-iters#7e86b55fa5ab82e54978021f8022068a1591166b"
+dependencies = [
+ "bitflags",
+]
+
 [[package]]
 name = "num-derive"
 version = "0.3.3"
@@ -250,6 +282,8 @@ checksum = "1230ec65f13e0f9b28d789da20d2d419511893ea9dac2c1f4ef67b8b14e5da80"
 name = "user"
 version = "0.1.0"
 dependencies = [
+ "lazy_static",
+ "libsys",
  "libusr",
 ]
 

Makefile

@@ -17,19 +17,24 @@ CARGO_COMMON_OPTS=
 ifeq ($(PROFILE),release)
 CARGO_COMMON_OPTS+=--release
 endif
+ifeq ($(VERBOSE),1)
+CARGO_COMMON_OPTS+=--features verbose
+endif
 
 CARGO_BUILD_OPTS=$(CARGO_COMMON_OPTS) \
 				 --target=../etc/$(ARCH)-$(MACH).json
-ifeq ($(VERBOSE),1)
-CARGO_BUILD_OPTS+=--features verbose
-endif
 ifneq ($(MACH),)
 CARGO_BUILD_OPTS+=--features mach_$(MACH)
 endif
 
 QEMU_OPTS=-s
 ifeq ($(ARCH),x86_64)
-$(error TODO)
+MACH=none
+QEMU_OPTS+=-cdrom $(O)/image.iso \
+		   -M q35 \
+		   -m 512 \
+		   -serial mon:stdio \
+		   -net none
 else
 ifeq ($(MACH),qemu)
 QEMU_OPTS+=-kernel $(O)/kernel.bin \
@@ -65,7 +70,7 @@ endif
 
 .PHONY: address error etc kernel src
 
-all: kernel initrd
+all: image
 
 kernel:
 	cd kernel && cargo build $(CARGO_BUILD_OPTS)
@@ -87,14 +92,32 @@ ifeq ($(MACH),orangepi3)
 		$(O)/uImage
 endif
 
+image: kernel initrd
+ifeq ($(ARCH),x86_64)
+	mkdir -p $(O)/image/boot/grub
+	cp etc/x86_64-none.grub $(O)/image/boot/grub/grub.cfg
+	cp $(O)/kernel $(O)/image/boot/kernel
+	cp $(O)/initrd.img $(O)/image/boot/initrd.img
+
+	grub-mkrescue -o $(O)/image.iso $(O)/image
+endif
+
 initrd:
 	cd user && cargo build \
 		--target=../etc/$(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 $(O)/rootfs/etc $(O)/rootfs/sys
+	cp etc/initrd/passwd $(O)/rootfs/etc
+	cp etc/initrd/shadow $(O)/rootfs/etc
+	touch $(O)/rootfs/dev/.do_not_remove
+	touch $(O)/rootfs/sys/.do_not_remove
 	cp target/$(ARCH)-osdev5/$(PROFILE)/init $(O)/rootfs/init
 	cp target/$(ARCH)-osdev5/$(PROFILE)/shell $(O)/rootfs/bin
+	cp target/$(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) \

etc/initrd/passwd

@@ -0,0 +1,2 @@
+root:0:0:root:/:/bin/shell
+alnyan:1000:1000:alnyan:/:/bin/shell

etc/initrd/shadow

@@ -0,0 +1,2 @@
+root:toor
+alnyan:

etc/x86_64-none.grub

@@ -1,3 +1,4 @@
 menuentry "OS" {
-    multiboot2 /boot/kernel.elf
+    multiboot2 /boot/kernel
+    module2 /boot/initrd.img
 }

etc/x86_64-none.ld

@@ -6,7 +6,8 @@ SECTIONS {
     . = 0x400000 + KERNEL_OFFSET;
 
     .text : AT(. - KERNEL_OFFSET) {
-        KEEP(*(.multiboot))
+        KEEP(*(.text._multiboot))
+        *(.text._entry)
         *(.text*)
     }
 
@@ -21,5 +22,8 @@ SECTIONS {
     .bss : AT(. - KERNEL_OFFSET) {
         *(COMMON)
         *(.bss*)
+        . = ALIGN(4K);
     }
+
+    PROVIDE(__kernel_end = .);
 }

etc/x86_64-osdev5.json

@@ -0,0 +1,21 @@
+{
+  "arch": "x86_64",
+  "cpu": "x86-64",
+  "data-layout": "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128",
+
+  "disable-redzone": true,
+  "executables": true,
+  "panic-strategy": "abort",
+
+  "linker": "rust-lld",
+  "linker-flavor": "ld.lld",
+
+  "llvm-target": "x86_64-unknown-linux-gnu",
+  "max-atomic-width": 64,
+  "target-pointer-width": "64",
+
+  "os": "none",
+  "pre-link-args": {
+    "ld.lld": [ "-Tetc/x86_64-osdev5.ld" ]
+  }
+}

etc/x86_64-osdev5.ld

@@ -0,0 +1,32 @@
+ENTRY(_start);
+
+PHDRS {
+    text PT_LOAD ;
+    rodata PT_LOAD ;
+    data PT_LOAD ;
+}
+
+SECTIONS {
+    . = 0x400000;
+
+    .text : {
+        *(.text._start)
+        *(.text*)
+        *(.eh_frame*)
+    } :text
+
+    . = ALIGN(0x1000);
+    .rodata : {
+        *(.rodata*)
+    } :rodata
+
+    . = ALIGN(0x1000);
+    .data : {
+        *(.data*)
+    } :data
+
+    .bss : {
+        *(COMMON)
+        *(.bss*)
+    } :data
+}

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 +1,11 @@
-use crate::{VnodeKind, VnodeRef};
+use crate::{VnodeKind, VnodeRef, Vnode};
 use alloc::rc::Rc;
 use core::cell::RefCell;
 use core::cmp::min;
+use core::str::FromStr;
 use libsys::{
     error::Errno,
+    stat::DirectoryEntry,
     traits::{Read, Seek, SeekDir, Write},
 };
 
@@ -97,6 +99,11 @@ 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 +126,68 @@ 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(".").unwrap();
+            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("..").unwrap();
+            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()).unwrap();
+        });
+        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::{Ref, RefCell, RefMut};
 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
@@ -230,14 +261,35 @@ 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> {
         if let Some(node) = self.lookup(name) {
-            return Ok(node);
-        }
-
-        if let Some(ref mut data) = *self.data() {
+            Ok(node)
+        } else if let Some(ref mut data) = *self.data() {
             let vnode = data.lookup(self.clone(), name)?;
             if let Some(fs) = self.fs() {
                 vnode.set_fs(fs);
@@ -303,22 +355,36 @@ 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() {
+        if self.kind == VnodeKind::Directory && self.flags & Vnode::CACHE_READDIR != 0 {
+            Ok(File::normal(self.clone(), File::POS_CACHE_DOT, open_flags))
+        } else if let Some(ref mut data) = *self.data() {
             let pos = data.open(self.clone(), flags)?;
             Ok(File::normal(self.clone(), pos, open_flags))
         } else {
@@ -328,7 +394,9 @@ impl Vnode {
 
     /// Closes a vnode
     pub fn close(self: &VnodeRef) -> Result<(), Errno> {
-        if let Some(ref mut data) = *self.data() {
+        if self.kind == VnodeKind::Directory && self.flags & Vnode::CACHE_READDIR != 0 {
+            Ok(())
+        } else if let Some(ref mut data) = *self.data() {
             data.close(self.clone())
         } else {
             Err(Errno::NotImplemented)
@@ -338,10 +406,8 @@ impl Vnode {
     /// Reads data from offset `pos` into `buf`
     pub fn read(self: &VnodeRef, pos: usize, buf: &mut [u8]) -> Result<usize, Errno> {
         if self.kind == VnodeKind::Directory {
-            return Err(Errno::IsADirectory);
-        }
-
-        if let Some(ref mut data) = *self.data() {
+            Err(Errno::IsADirectory)
+        } else if let Some(ref mut data) = *self.data() {
             data.read(self.clone(), pos, buf)
         } else {
             Err(Errno::NotImplemented)
@@ -351,10 +417,8 @@ impl Vnode {
     /// Writes data from `buf` to offset `pos`
     pub fn write(self: &VnodeRef, pos: usize, buf: &[u8]) -> Result<usize, Errno> {
         if self.kind == VnodeKind::Directory {
-            return Err(Errno::IsADirectory);
-        }
-
-        if let Some(ref mut data) = *self.data() {
+            Err(Errno::IsADirectory)
+        } else if let Some(ref mut data) = *self.data() {
             data.write(self.clone(), pos, buf)
         } else {
             Err(Errno::NotImplemented)
@@ -364,10 +428,8 @@ impl Vnode {
     /// Resizes the vnode data
     pub fn truncate(self: &VnodeRef, size: usize) -> Result<(), Errno> {
         if self.kind != VnodeKind::Regular {
-            return Err(Errno::IsADirectory);
-        }
-
-        if let Some(ref mut data) = *self.data() {
+            Err(Errno::IsADirectory)
+        } else if let Some(ref mut data) = *self.data() {
             data.truncate(self.clone(), size)
         } else {
             Err(Errno::NotImplemented)
@@ -384,9 +446,16 @@ 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 +470,7 @@ 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 +478,38 @@ 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);
+            }
+        } 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
+        }
+
+        Ok(())
+    }
 }
 
 impl fmt::Debug for Vnode {

kernel/Cargo.toml

@@ -15,17 +15,23 @@ memfs = { path = "../fs/memfs" }
 libsys = { path = "../libsys" }
 cfg-if = "1.x.x"
 tock-registers = "0.7.x"
-fdt-rs = { version = "0.x.x", default-features = false }
 bitflags = "^1.3.0"
 kernel-macros = { path = "macros" }
+fs-macros = { path = "../fs/macros" }
+
+[target.'cfg(target_arch = "x86_64")'.dependencies]
+multiboot2 = { git = "https://github.com/alnyan/multiboot2", branch = "expose-extra-traits-for-iters" }
 
 [target.'cfg(target_arch = "aarch64")'.dependencies]
 cortex-a = { version = "6.x.x" }
+fdt-rs = { version = "0.x.x", default-features = false }
 
 [features]
+default = ["aggressive_syscall"]
 pl011 = []
 pl031 = []
 verbose = []
+aggressive_syscall = []
 
 mach_qemu = ["pl011", "pl031"]
 mach_orangepi3 = []

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

@@ -10,7 +10,8 @@ use crate::dev::{
     irq::IntSource,
     Device,
 };
-use crate::fs::devfs;
+use crate::fs::{devfs, sysfs};
+use crate::dev::pseudo;
 use libsys::error::Errno;
 //use crate::debug::Level;
 use crate::mem::{
@@ -23,22 +24,16 @@ use cortex_a::asm::barrier::{self, dsb, isb};
 use cortex_a::registers::{SCTLR_EL1, VBAR_EL1};
 use tock_registers::interfaces::{ReadWriteable, 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")]
-    {
-        use crate::debug::Level;
-        fdt.dump(Level::Debug);
-    }
-
     let mut cfg = CONFIG.lock();
 
     if let Some(chosen) = fdt.node_by_path("/chosen") {
@@ -56,7 +51,7 @@ fn init_device_tree(fdt_base_phys: usize) -> Result<(), Errno> {
         }
     }
 
-    Ok(())
+    Ok(Some(fdt))
 }
 
 #[no_mangle]
@@ -87,7 +82,7 @@ extern "C" fn __aa64_bsp_main(fdt_base: usize) -> ! {
     // 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
@@ -102,9 +97,19 @@ extern "C" fn __aa64_bsp_main(fdt_base: usize) -> ! {
     }
 
     devfs::init();
+    sysfs::init();
 
     machine::init_board().unwrap();
 
+    #[cfg(feature = "verbose")]
+    if let Some(fdt) = fdt {
+        use crate::debug::Level;
+        fdt.dump(Level::Debug);
+    }
+
+    devfs::add_named_char_device(&pseudo::ZERO, "zero").unwrap();
+    devfs::add_named_char_device(&pseudo::RANDOM, "random").unwrap();
+
     infoln!("Machine init finished");
 
     unsafe {

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/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, Process, Thread};
 use crate::syscall;
 use cortex_a::registers::{ESR_EL1, FAR_EL1};
-use libsys::{abi, signal::Signal};
+use libsys::{abi::SystemCall, signal::Signal, error::Errno};
 use tock_registers::interfaces::Readable;
 
 /// Trapped SIMD/FP functionality
@@ -88,56 +88,71 @@ 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();
+            if iss & (1 << 6) != 0 && far < mem::KERNEL_OFFSET && sched::is_ready() {
+                let thread = Thread::current();
+                let proc = thread.owner().unwrap();
+                let asid = proc.asid();
 
-                if proc
-                    .manipulate_space(|space| space.try_cow_copy(far))
-                    .is_err()
-                {
+                let res = proc.manipulate_space(|space| {
+                    space.try_cow_copy(far)?;
+                    Process::invalidate_asid(asid);
+                    Result::<(), Errno>::Ok(())
+                });
+
+                if res.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);
-                }
-
-                unsafe {
-                    use cortex_a::registers::TTBR0_EL1;
-                    let ttbr = TTBR0_EL1.get() as usize;
-                    let asid = (ttbr >> 48) & 0xFF;
-                    asm!("tlbi aside1, {}", in(reg) (asid << 48));
+                    proc.enter_fault_signal(thread, Signal::SegmentationFault);
                 }
 
                 return;
             }
 
             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] = u32::from(pid) 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/mod.rs

@@ -28,7 +28,6 @@ pub struct Gic {
     gicd: InitOnce<Gicd>,
     gicd_base: usize,
     gicc_base: usize,
-    scheduler_irq: IrqNumber,
     table: IrqSafeSpinLock<[Option<&'static (dyn IntSource + Sync)>; MAX_IRQ]>,
 }
 
@@ -86,9 +85,7 @@ impl IntController for Gic {
             return;
         }
 
-        if self.scheduler_irq.0 == irq_number {
-            gicc.clear_irq(irq_number as u32, ic);
-        }
+        gicc.clear_irq(irq_number as u32, ic);
 
         {
             let table = self.table.lock();
@@ -100,8 +97,6 @@ impl IntController for Gic {
                 }
             }
         }
-
-        gicc.clear_irq(irq_number as u32, ic);
     }
 
     fn register_handler(
@@ -128,13 +123,12 @@ impl Gic {
     /// # 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_orangepi3/mod.rs

@@ -96,4 +96,4 @@ static UART0: Uart = unsafe { Uart::new(UART0_BASE, IrqNumber::new(32)) };
 static LOCAL_TIMER: GenericTimer = GenericTimer::new(LOCAL_TIMER_IRQ);
 pub(super) static GPIO: Gpio = unsafe { Gpio::new(PIO_BASE) };
 static RTC: Rtc = unsafe { Rtc::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) };

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

@@ -78,6 +78,6 @@ pub fn intc() -> &'static impl IntController<IrqNumber = IrqNumber> {
 
 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/table.rs

@@ -0,0 +1,422 @@
+//
+// #[no_mangle]
+// static mut KERNEL_TTBR1: FixedTableGroup = FixedTableGroup::empty();
+
+/// Transparent wrapper structure representing a single
+/// translation table entry
+#[derive(Clone, Copy)]
+#[repr(transparent)]
+pub struct Entry(u64);
+
+/// Structure describing a single level of translation mappings
+#[repr(C, align(0x1000))]
+pub struct Table {
+    entries: [Entry; 512],
+}
+
+/// Wrapper for top-most level of address translation tables
+#[repr(transparent)]
+pub struct Space(Table);
+
+bitflags! {
+    /// Attributes attached to each translation [Entry]
+    pub struct MapAttributes: u64 {
+        // TODO use 2 lower bits to determine mapping size?
+        /// nG bit -- determines whether a TLB entry associated with this mapping
+        ///           applies only to current ASID or all ASIDs.
+        const NOT_GLOBAL = 1 << 11;
+        /// AF bit -- must be set by software, otherwise Access Error exception is
+        ///           generated when the page is accessed
+        const ACCESS = 1 << 10;
+        /// The memory region is outer-shareable
+        const SH_OUTER = 2 << 8;
+        /// This page is used for device-MMIO mapping and uses MAIR attribute #1
+        const DEVICE = 1 << 2;
+
+        /// Pages marked with this bit are Copy-on-Write
+        const EX_COW = 1 << 55;
+
+        /// UXN bit -- if set, page may not be used for instruction fetching from EL0
+        const UXN = 1 << 54;
+        /// PXN bit -- if set, page may not be used for instruction fetching from EL1
+        const PXN = 1 << 53;
+
+        // AP field
+        // Default behavior is: read-write for EL1, no access for EL0
+        /// If set, the page referred to by this entry is read-only for both EL0/EL1
+        const AP_BOTH_READONLY = 3 << 6;
+        /// If set, the page referred to by this entry is read-write for both EL0/EL1
+        const AP_BOTH_READWRITE = 1 << 6;
+    }
+}
+
+impl Table {
+    /// Returns next-level translation table reference for `index`, if one is present.
+    /// If `index` represents a `Block`-type mapping, will return an error.
+    /// If `index` does not map to any translation table, will try to allocate, init and
+    /// map a new one, returning it after doing so.
+    pub fn next_level_table_or_alloc(&mut self, index: usize) -> Result<&'static mut Table, Errno> {
+        let entry = self[index];
+        if entry.is_present() {
+            if !entry.is_table() {
+                return Err(Errno::InvalidArgument);
+            }
+
+            Ok(unsafe { &mut *(mem::virtualize(entry.address_unchecked()) as *mut _) })
+        } else {
+            let phys = phys::alloc_page(PageUsage::Paging)?;
+            let res = unsafe { &mut *(mem::virtualize(phys) as *mut Self) };
+            self[index] = Entry::table(phys, MapAttributes::empty());
+            res.entries.fill(Entry::invalid());
+            Ok(res)
+        }
+    }
+
+    /// Returns next-level translation table reference for `index`, if one is present.
+    /// Same as [next_level_table_or_alloc], but returns `None` if no table is mapped.
+    pub fn next_level_table(&mut self, index: usize) -> Option<&'static mut Table> {
+        let entry = self[index];
+        if entry.is_present() {
+            if !entry.is_table() {
+                panic!("Entry is not a table: idx={}", index);
+            }
+
+            Some(unsafe { &mut *(mem::virtualize(entry.address_unchecked()) as *mut _) })
+        } else {
+            None
+        }
+    }
+
+    /// Constructs and fills a [Table] with non-present mappings
+    pub const fn empty() -> Table {
+        Table {
+            entries: [Entry::invalid(); 512],
+        }
+    }
+}
+
+impl Index<usize> for Table {
+    type Output = Entry;
+
+    fn index(&self, index: usize) -> &Self::Output {
+        &self.entries[index]
+    }
+}
+
+impl IndexMut<usize> for Table {
+    fn index_mut(&mut self, index: usize) -> &mut Self::Output {
+        &mut self.entries[index]
+    }
+}
+
+impl Entry {
+    const PRESENT: u64 = 1 << 0;
+    const TABLE: u64 = 1 << 1;
+    const PHYS_MASK: u64 = 0x0000FFFFFFFFF000;
+
+    /// Constructs a single non-present mapping
+    pub const fn invalid() -> Self {
+        Self(0)
+    }
+
+    /// Constructs a `Block`-type memory mapping
+    pub const fn block(phys: usize, attrs: MapAttributes) -> Self {
+        Self((phys as u64 & Self::PHYS_MASK) | attrs.bits() | Self::PRESENT)
+    }
+
+    /// Constructs a `Table` or `Page`-type mapping depending on translation level
+    /// this entry is used at
+    pub const fn table(phys: usize, attrs: MapAttributes) -> Self {
+        Self((phys as u64 & Self::PHYS_MASK) | attrs.bits() | Self::PRESENT | Self::TABLE)
+    }
+
+    /// Returns `true` if this entry is not invalid
+    pub const fn is_present(self) -> bool {
+        self.0 & Self::PRESENT != 0
+    }
+
+    /// Returns `true` if this entry is a `Table` or `Page`-type mapping
+    pub const fn is_table(self) -> bool {
+        self.0 & Self::TABLE != 0
+    }
+
+    /// Returns the target address of this translation entry.
+    ///
+    /// # Safety
+    ///
+    /// Does not check if the entry is actually valid.
+    pub const unsafe fn address_unchecked(self) -> usize {
+        (self.0 & Self::PHYS_MASK) as usize
+    }
+
+    unsafe fn set_address(&mut self, address: usize) {
+        self.0 &= !Self::PHYS_MASK;
+        self.0 |= (address as u64) & Self::PHYS_MASK;
+    }
+
+    unsafe fn fork_flags(self) -> MapAttributes {
+        MapAttributes::from_bits_unchecked(self.0 & !Self::PHYS_MASK)
+    }
+
+    fn set_cow(&mut self) {
+        self.0 |= (MapAttributes::AP_BOTH_READONLY | MapAttributes::EX_COW).bits();
+    }
+
+    fn clear_cow(&mut self) {
+        self.0 &= !(MapAttributes::AP_BOTH_READONLY | MapAttributes::EX_COW).bits();
+        self.0 |= MapAttributes::AP_BOTH_READWRITE.bits();
+    }
+
+    #[inline]
+    fn is_cow(self) -> bool {
+        let attrs = (MapAttributes::AP_BOTH_READONLY | MapAttributes::EX_COW).bits();
+        self.0 & attrs == attrs
+    }
+}
+
+impl Space {
+    /// Creates a new virtual address space and fills it with [Entry::invalid()]
+    /// mappings. Does physical memory page allocation.
+    pub fn alloc_empty() -> Result<&'static mut Self, Errno> {
+        let phys = phys::alloc_page(PageUsage::Paging)?;
+        let res = unsafe { &mut *(mem::virtualize(phys) as *mut Self) };
+        res.0.entries.fill(Entry::invalid());
+        Ok(res)
+    }
+
+    /// Inserts a single `virt` -> `phys` translation entry to this address space.
+    ///
+    /// TODO: only works with 4K-sized pages at this moment.
+    pub fn map(&mut self, virt: usize, phys: usize, flags: MapAttributes) -> Result<(), Errno> {
+        let l0i = virt >> 30;
+        let l1i = (virt >> 21) & 0x1FF;
+        let l2i = (virt >> 12) & 0x1FF;
+
+        let l1_table = self.0.next_level_table_or_alloc(l0i)?;
+        let l2_table = l1_table.next_level_table_or_alloc(l1i)?;
+
+        if l2_table[l2i].is_present() {
+            Err(Errno::AlreadyExists)
+        } else {
+            l2_table[l2i] = Entry::table(phys, flags | MapAttributes::ACCESS);
+            #[cfg(feature = "verbose")]
+            debugln!("{:#p} Map {:#x} -> {:#x}, {:?}", self, virt, phys, flags);
+            Ok(())
+        }
+    }
+
+    /// 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> {
+        let virt = virt & !0xFFF;
+        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() {
+            warnln!("Entry is not present: {:#x}", virt);
+            return Err(Errno::DoesNotExist);
+        }
+
+        let src_phys = unsafe { entry.address_unchecked() };
+        if !entry.is_cow() {
+            warnln!(
+                "Entry is not marked as CoW: {:#x}, points to {:#x}",
+                virt,
+                src_phys
+            );
+            return Err(Errno::DoesNotExist);
+        }
+
+        let dst_phys = unsafe { phys::copy_cow_page(src_phys)? };
+        unsafe {
+            l2_table[l2i].set_address(dst_phys);
+        }
+        l2_table[l2i].clear_cow();
+
+        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()?;
+        for l0i in 0..512 {
+            if let Some(l1_table) = self.0.next_level_table(l0i) {
+                for l1i in 0..512 {
+                    if let Some(l2_table) = l1_table.next_level_table(l1i) {
+                        for l2i in 0..512 {
+                            let entry = l2_table[l2i];
+
+                            if !entry.is_present() {
+                                continue;
+                            }
+
+                            assert!(entry.is_table());
+                            let src_phys = unsafe { entry.address_unchecked() };
+                            let virt_addr = (l0i << 30) | (l1i << 21) | (l2i << 12);
+                            let dst_phys = unsafe { phys::fork_page(src_phys)? };
+
+                            let mut flags = unsafe { entry.fork_flags() };
+                            if dst_phys != src_phys {
+                                todo!();
+                                // res.map(virt_addr, dst_phys, flags)?;
+                            } else {
+                                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)?;
+                            }
+                        }
+                    }
+                }
+            }
+        }
+        Ok(res)
+    }
+
+    /// Releases all the mappings from the address space. Frees all
+    /// memory pages referenced by this space as well as those used for
+    /// its paging tables.
+    ///
+    /// # Safety
+    ///
+    /// Unsafe: may invalidate currently active address space
+    pub unsafe fn release(space: &mut Self) {
+        for l0i in 0..512 {
+            let l0_entry = space.0[l0i];
+            if !l0_entry.is_present() {
+                continue;
+            }
+
+            assert!(l0_entry.is_table());
+            let l1_table = &mut *(mem::virtualize(l0_entry.address_unchecked()) as *mut Table);
+
+            for l1i in 0..512 {
+                let l1_entry = l1_table[l1i];
+                if !l1_entry.is_present() {
+                    continue;
+                }
+                assert!(l1_entry.is_table());
+                let l2_table = &mut *(mem::virtualize(l1_entry.address_unchecked()) as *mut Table);
+
+                for l2i in 0..512 {
+                    let entry = l2_table[l2i];
+                    if !entry.is_present() {
+                        continue;
+                    }
+
+                    assert!(entry.is_table());
+                    phys::free_page(entry.address_unchecked()).unwrap();
+                }
+                phys::free_page(l1_entry.address_unchecked()).unwrap();
+            }
+            phys::free_page(l0_entry.address_unchecked()).unwrap();
+        }
+        memset(space as *mut Space as *mut u8, 0, 4096);
+    }
+
+    /// Returns the physical address of this structure
+    pub fn address_phys(&mut self) -> usize {
+        (self as *mut _ as usize) - mem::KERNEL_OFFSET
+    }
+}

kernel/src/arch/aarch64/timer.rs

@@ -1,7 +1,9 @@
 //! ARM generic timer implementation
 
 use crate::arch::machine::{self, IrqNumber};
+use crate::proc;
 use crate::dev::{
+    pseudo,
     irq::{IntController, IntSource},
     timer::TimestampSource,
     Device,
@@ -34,9 +36,9 @@ 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();
+        pseudo::RANDOM.set_state(CNTPCT_EL0.get() as u32);
         Ok(())
     }
 
@@ -50,9 +52,11 @@ impl IntSource for GenericTimer {
 
 impl TimestampSource for GenericTimer {
     fn timestamp(&self) -> Result<Duration, Errno> {
-        let cnt = CNTPCT_EL0.get() * 1_000_000_000;
-        let frq = CNTFRQ_EL0.get();
-        Ok(Duration::from_nanos(cnt / frq))
+        let cnt = (CNTPCT_EL0.get() as u128) * 1_000_000_000u128;
+        let frq = CNTFRQ_EL0.get() as u128;
+        let secs = ((cnt / frq) / 1_000_000_000) as u64;
+        let nanos = ((cnt / frq) % 1_000_000_000) as u32;
+        Ok(Duration::new(secs, nanos))
     }
 }
 

kernel/src/arch/mod.rs

@@ -15,6 +15,11 @@ cfg_if! {
 
         pub use aarch64 as platform;
         pub use aarch64::machine;
+    } else if #[cfg(target_arch = "x86_64")] {
+        pub mod x86_64;
+
+        pub use x86_64 as platform;
+        pub use x86_64 as machine;
     }
 }
 

kernel/src/arch/x86_64/boot/entry.S

@@ -0,0 +1,53 @@
+.section .text._multiboot
+.set MAGIC,     0xE85250D6
+.set ARCH,      0x0
+.set HDRLEN,    16
+.set CHKSUM,    (-(MAGIC + ARCH + HDRLEN)) & 0xFFFFFFFF
+
+.long MAGIC
+.long ARCH
+.long HDRLEN
+.long CHKSUM
+
+.short 5
+.short 0
+.long 20
+.long 800
+.long 600
+.long 32
+
+.short 0
+.long 8
+
+.section .text._entry
+.global _entry
+_entry:
+.code32
+    cli
+    lea (multiboot_registers - KERNEL_OFFSET), %edi
+    mov %eax, 0(%edi)
+    mov %ebx, 4(%edi)
+
+    // Setup paging tables
+    lea (_entry_upper - KERNEL_OFFSET), %ebx
+    jmp __x86_64_enter_upper
+
+.code64
+_entry_upper:
+    movabsq $1f, %rax
+    jmp *%rax
+1:
+    lea bsp_stack_top(%rip), %rax
+    mov %rax, %rsp
+
+    mov multiboot_registers(%rip), %edi
+    mov (4 + multiboot_registers)(%rip), %esi
+    call __x86_64_bsp_main
+
+.section .bss
+    .align 16
+bsp_stack_bottom:
+    .skip 65536
+bsp_stack_top:
+multiboot_registers:
+    .skip 8

kernel/src/arch/x86_64/boot/macros.S

@@ -0,0 +1,7 @@
+.set KERNEL_OFFSET,     0xFFFFFF8000000000
+
+.set PTE_PRESENT,   1 << 0
+.set PTE_WRITABLE,  1 << 1
+.set PTE_USERSPACE, 1 << 2
+.set PTE_BLOCK,     1 << 7
+

kernel/src/arch/x86_64/boot/mod.rs

@@ -0,0 +1,114 @@
+use crate::arch::x86_64::{
+    self, gdt, idt, intc,
+    reg::{CR0, CR4},
+    syscall,
+};
+use crate::config::{ConfigKey, CONFIG};
+use crate::debug;
+use crate::dev::{display::FramebufferInfo, pseudo, Device};
+use crate::font;
+use crate::fs::{devfs::{self, CharDeviceType}, sysfs};
+use crate::mem::{
+    self, heap,
+    phys::{self, MemoryRegion, PageUsage, ReservedRegion},
+    virt,
+};
+use crate::proc;
+use core::arch::{asm, global_asm};
+use core::mem::MaybeUninit;
+use multiboot2::{BootInformation, MemoryArea};
+use tock_registers::interfaces::ReadWriteable;
+
+static mut RESERVED_REGION_MB2: MaybeUninit<ReservedRegion> = MaybeUninit::uninit();
+
+#[no_mangle]
+extern "C" fn __x86_64_bsp_main(mb_checksum: u32, mb_info_ptr: u32) -> ! {
+    CR4.modify(CR4::OSXMMEXCPT::SET + CR4::OSFXSR::SET);
+    CR0.modify(CR0::EM::CLEAR + CR0::MP::SET);
+
+    unsafe {
+        // Setup a proper GDT
+        gdt::init();
+        idt::init(intc::map_isr_entries);
+    }
+
+    virt::enable().expect("Failed to initialize virtual memory");
+
+    let mb_info = unsafe {
+        multiboot2::load_with_offset(mb_info_ptr as usize, mem::KERNEL_OFFSET)
+            .expect("Failed to load multiboot info structure")
+    };
+
+    unsafe {
+        let mb_info_page = (mb_info_ptr & !0xFFF) as usize;
+        RESERVED_REGION_MB2.write(ReservedRegion::new(
+            mb_info_page,
+            mb_info_page + ((mb_info.total_size() + 0xFFF) & !0xFFF),
+        ));
+        phys::reserve("multiboot2", RESERVED_REGION_MB2.as_mut_ptr());
+
+        phys::init_from_iter(
+            mb_info
+                .memory_map_tag()
+                .unwrap()
+                .memory_areas()
+                .map(|entry| MemoryRegion {
+                    start: ((entry.start_address() + 0xFFF) & !0xFFF) as usize,
+                    end: (entry.end_address() & !0xFFF) as usize,
+                }),
+        );
+    }
+
+    // Setup a heap
+    unsafe {
+        let heap_base_phys = phys::alloc_contiguous_pages(PageUsage::KernelHeap, 4096)
+            .expect("Failed to allocate memory for heap");
+        let heap_base_virt = mem::virtualize(heap_base_phys);
+        heap::init(heap_base_virt, 16 * 1024 * 1024);
+    }
+
+    let initrd_info = mb_info.module_tags().next().unwrap();
+    {
+        let mut cfg = CONFIG.lock();
+
+        cfg.set_usize(ConfigKey::InitrdBase, initrd_info.start_address() as usize);
+        cfg.set_usize(ConfigKey::InitrdSize, initrd_info.module_size() as usize);
+    }
+
+    // Setup hardware
+    unsafe {
+        x86_64::INTC.enable().ok();
+    }
+
+    let fb_info = mb_info.framebuffer_tag().unwrap();
+    let virt = mem::virtualize(fb_info.address as usize);
+    debugln!(
+        "Framebuffer base: phys={:#x}, virt={:#x}",
+        fb_info.address,
+        virt
+    );
+    x86_64::DISPLAY.set_framebuffer(FramebufferInfo {
+        width: fb_info.width as usize,
+        height: fb_info.height as usize,
+        phys_base: fb_info.address as usize,
+        virt_base: virt,
+    });
+    font::init();
+    debug::set_display(&x86_64::DISPLAY);
+    syscall::init();
+
+    devfs::init();
+    sysfs::init();
+
+    devfs::add_char_device(&x86_64::COM1, CharDeviceType::TtySerial).unwrap();
+    devfs::add_named_char_device(&pseudo::ZERO, "zero").unwrap();
+    devfs::add_named_char_device(&pseudo::RANDOM, "random").unwrap();
+
+    unsafe {
+        proc::enter();
+    }
+}
+
+global_asm!(include_str!("macros.S"), options(att_syntax));
+global_asm!(include_str!("entry.S"), options(att_syntax));
+global_asm!(include_str!("upper.S"), options(att_syntax));

kernel/src/arch/x86_64/boot/upper.S

@@ -0,0 +1,94 @@
+.code32
+
+.section .text._entry
+__x86_64_enter_upper:
+    mov $(PTE_PRESENT | PTE_WRITABLE | PTE_USERSPACE), %edx
+    // Setup PML4
+    lea (KERNEL_FIXED - KERNEL_OFFSET), %edi
+    lea (KERNEL_FIXED + 4096 - KERNEL_OFFSET), %esi
+    mov %edx, %eax
+    or %esi, %eax
+    //  pml4[0] = %eax
+    mov %eax, (%edi)
+    //  pml4[511] = %eax
+    mov %eax, 4088(%edi)
+
+    // Setup PDPT
+    mov %esi, %edi
+    lea (KERNEL_FIXED + 8192 - KERNEL_OFFSET), %esi
+    xor %ecx, %ecx
+1:
+    // %eax = &table[%ecx] | attrs
+    mov %esi, %eax
+    or %edx, %eax
+    mov %eax, (%edi, %ecx, 8)
+
+    add $4096, %esi
+    inc %ecx
+    cmp $16, %ecx
+    jne 1b
+
+    // Setup PDs
+    lea (KERNEL_FIXED + 8192 - KERNEL_OFFSET), %edi
+    mov $(PTE_PRESENT | PTE_BLOCK | PTE_WRITABLE), %edx
+    mov $(512 * 16), %ecx
+1:
+    dec %ecx
+
+    // %eax = attrs | (i << 21)
+    mov %ecx, %eax
+    shl $21, %eax
+    or %edx, %eax
+
+    mov %eax, (%edi, %ecx, 8)
+
+    test %ecx, %ecx
+    jnz 1b
+
+    // Enable PAE/PSE
+    mov %cr4, %eax
+    or $((1 << 5) | (1 << 4)), %eax
+    mov %eax, %cr4
+
+    // Enable EFER.LME
+    mov $0xC0000080, %ecx
+    rdmsr
+    or $(1 << 8), %eax
+    wrmsr
+
+    // Set CR3
+    lea (KERNEL_FIXED - KERNEL_OFFSET), %edi
+    mov %edi, %cr3
+
+    // Enable paging
+    mov %cr0, %eax
+    or $(1 << 31), %eax
+    mov %eax, %cr0
+
+    lgdt (gdtr64 - KERNEL_OFFSET)
+    ljmp $0x08, $(1f - KERNEL_OFFSET)
+1:
+    mov $0x10, %ax
+    mov %ax, %ds
+    mov %ax, %es
+    mov %ax, %fs
+    mov %ax, %gs
+    mov %ax, %ss
+
+.code64
+    mov $KERNEL_OFFSET, %rax
+    add %rax, %rbx
+    jmp *%rbx
+
+.section .rodata
+.code32
+.align 16
+gdt64:
+    .quad 0
+    .quad 0x00209A0000000000
+    .quad 0x0000920000000000
+gdt_end64:
+.align 16
+gdtr64:
+    .short gdt_end64 - gdt64 - 1
+    .long gdt64 - KERNEL_OFFSET

kernel/src/arch/x86_64/context.S

@@ -0,0 +1,78 @@
+.section .text
+.global __x86_64_ctx_switch
+.global __x86_64_ctx_switch_to
+.global __x86_64_ctx_enter_kernel
+.global __x86_64_ctx_enter_from_fork
+
+__x86_64_ctx_enter_user:
+    pop %rcx
+    pop %rdi
+    pop %rdi
+    pop %rdx
+
+    push $0x1B
+    push %rcx
+
+    push $0x200
+
+    push $0x23
+    push %rdx
+
+    iretq
+
+__x86_64_ctx_enter_kernel:
+    pop %rdi
+    pop %rdx
+    mov %rsp, %rcx
+
+    push $0x10
+    push %rcx
+
+    push $0x200
+
+    push $0x08
+    push %rdx
+
+    iretq
+
+__x86_64_ctx_enter_from_fork:
+    jmp .
+
+__x86_64_ctx_switch:
+    // %rsi -- src ctx ptr
+    // %rdi -- dst ctx ptr
+
+    push %r15
+    push %r14
+    push %r13
+    push %r12
+    push %rbx
+    push %rbp
+
+    mov %cr3, %rax
+    push %rax
+    // TODO save gs_base
+    mov (4 + TSS)(%rip), %rax
+    push %rax
+
+    mov %rsp, (%rsi)
+__x86_64_ctx_switch_to:
+    mov (%rdi), %rsp
+
+    pop %rbp
+    pop %rbx
+    pop %r12
+    pop %r13
+    pop %r14
+    pop %r15
+
+    pop %rax
+    test %rax, %rax
+    jz 1f
+    mov %rax, %cr3
+1:
+    pop %rax
+    mov %rax, (4 + TSS)(%rip)
+    // TODO set gs_base = rax
+
+    ret

kernel/src/arch/x86_64/context.rs

@@ -0,0 +1,151 @@
+use crate::mem::{
+    self,
+    phys::{self, PageUsage},
+};
+use core::mem::size_of;
+use core::arch::global_asm;
+
+struct Stack {
+    bp: usize,
+    sp: usize,
+}
+
+/// Structure representing thread context
+#[repr(C)]
+pub struct Context {
+    /// Thread's kernel stack pointer
+    pub k_sp: usize, // 0x00
+
+    stack_base: usize,
+    stack_page_count: usize,
+}
+
+impl Context {
+    /// Constructs a new kernel-space thread context
+    pub fn kernel(entry: usize, arg: usize) -> Self {
+        let mut stack = Stack::new(8);
+
+        stack.push(entry);
+        stack.push(arg);
+
+        stack.setup_common(__x86_64_ctx_enter_kernel as usize, 0, 0);
+
+        Self {
+            k_sp: stack.sp,
+
+            stack_base: stack.bp,
+            stack_page_count: 8,
+        }
+    }
+
+    /// Constructs a new user-space thread context
+    pub fn user(entry: usize, arg: usize, cr3: usize, ustack: usize) -> Self {
+        let cr3 = cr3 & 0xFFFFFFFF;
+        let mut stack = Stack::new(8);
+        let stack_top = stack.sp;
+
+        stack.push(entry);
+        stack.push(arg);
+        stack.push(0);
+        stack.push(ustack);
+
+        stack.setup_common(__x86_64_ctx_enter_user as usize, cr3, stack_top);
+
+        Self {
+            k_sp: stack.sp,
+
+            stack_base: stack.bp,
+            stack_page_count: 8,
+        }
+    }
+
+    /// Constructs an uninitialized thread context
+    pub fn empty() -> Self {
+        let stack = Stack::new(8);
+        Self {
+            k_sp: stack.sp,
+            stack_base: stack.bp,
+            stack_page_count: 8
+        }
+    }
+
+    /// Sets up a context for signal entry
+    ///
+    /// # Safety
+    ///
+    /// Unsafe: may clobber an already active context
+    pub unsafe fn setup_signal_entry(&mut self, entry: usize, arg: usize, cr3: usize, ustack: usize) {
+        todo!()
+    }
+
+    /// Performs initial thread entry
+    ///
+    /// # Safety
+    ///
+    /// Unsafe: does not check if any context has already been activated
+    /// before, so must only be called once.
+    pub unsafe extern "C" fn enter(&mut self) -> ! {
+        __x86_64_ctx_switch_to(self);
+        panic!("This code should not run");
+    }
+
+    /// Performs context switch from `self` to `to`.
+    ///
+    /// # Safety
+    ///
+    /// Unsafe: does not check if `self` is actually an active context.
+    pub unsafe extern "C" fn switch(&mut self, to: &mut Context) {
+        __x86_64_ctx_switch(to, self);
+    }
+}
+
+impl Stack {
+    pub fn new(page_count: usize) -> Stack {
+        let phys = phys::alloc_contiguous_pages(PageUsage::Kernel, page_count).unwrap();
+        let bp = mem::virtualize(phys);
+        Stack {
+            bp,
+            sp: bp + page_count * mem::PAGE_SIZE,
+        }
+    }
+
+    pub unsafe fn from_base_size(bp: usize, page_count: usize) -> Stack {
+        Stack {
+            bp,
+            sp: bp + page_count * mem::PAGE_SIZE
+        }
+    }
+
+    pub fn setup_common(&mut self, entry: usize, cr3: usize, tss_rsp0: usize) {
+        self.push(entry);   // return address
+        self.push(tss_rsp0);       // gs_base
+        self.push(cr3);
+        self.push(0);       // r15
+        self.push(0);       // r14
+        self.push(0);       // r13
+        self.push(0);       // r12
+        self.push(0);       // rbx
+        self.push(0);       // rbp
+    }
+
+    pub fn push(&mut self, value: usize) {
+        if self.bp == self.sp {
+            panic!("Stack overflow");
+        }
+
+        self.sp -= size_of::<usize>();
+        unsafe {
+            *(self.sp as *mut usize) = value;
+        }
+    }
+}
+
+extern "C" {
+    fn __x86_64_ctx_enter_from_fork();
+    fn __x86_64_ctx_enter_kernel();
+    fn __x86_64_ctx_enter_user();
+    fn __x86_64_ctx_switch(dst: *mut Context, src: *mut Context);
+    fn __x86_64_ctx_switch_to(dst: *mut Context);
+}
+
+global_asm!(include_str!("context.S"), options(att_syntax));

kernel/src/arch/x86_64/exception.rs

@@ -0,0 +1,86 @@
+use crate::arch::x86_64;
+use crate::debug::Level;
+use crate::dev::irq::{IntController, IrqContext};
+use core::arch::{asm, global_asm};
+
+#[derive(Debug)]
+struct ExceptionFrame {
+    r15: u64,
+    r14: u64,
+    r13: u64,
+    r12: u64,
+    r11: u64,
+    r10: u64,
+    r9: u64,
+    r8: u64,
+    rdi: u64,
+    rsi: u64,
+    rbp: u64,
+    rbx: u64,
+    rdx: u64,
+    rcx: u64,
+    rax: u64,
+
+    err_no: u64,
+    err_code: u64,
+    rip: u64,
+    cs: u64,
+    rflags: u64,
+    rsp: u64,
+    ss: u64,
+}
+
+fn pfault_read_cr2() -> u64 {
+    let mut res;
+    unsafe {
+        asm!("mov %cr2, {}", out(reg) res, options(att_syntax));
+    }
+    res
+}
+
+fn pfault_access_type(code: u64) -> &'static str {
+    if code & (1 << 4) != 0 {
+        "INSTRUCTION FETCH"
+    } else if code & (1 << 1) != 0 {
+        "WRITE"
+    } else {
+        "READ"
+    }
+}
+
+fn pfault_dump(level: Level, frame: &ExceptionFrame, cr2: u64) {
+    println!(level, "\x1B[41;1mPage fault:");
+    println!(
+        level,
+        "  Illegal {} at {:#018x}\x1B[0m",
+        pfault_access_type(frame.err_code),
+        cr2
+    );
+}
+
+#[no_mangle]
+extern "C" fn __x86_64_exception_handler(frame: &mut ExceptionFrame) {
+    if frame.err_no == 14 {
+        // TODO userspace page faults
+        let cr2 = pfault_read_cr2();
+        pfault_dump(Level::Error, frame, cr2);
+    }
+
+    errorln!(
+        "Exception occurred: err_no={}, err_code={:#x}",
+        frame.err_no,
+        frame.err_code,
+    );
+    errorln!("cs:rip = {:02x}:{:#x}", frame.cs, frame.rip);
+    errorln!("ss:rsp = {:02x}:{:#x}", frame.ss, frame.rsp);
+
+    panic!("Unhandled exception");
+}
+
+#[no_mangle]
+extern "C" fn __x86_64_irq_handler(frame: &mut ExceptionFrame) {
+    unsafe {
+        let ic = IrqContext::new(frame.err_no as usize);
+        x86_64::intc().handle_pending_irqs(&ic);
+    }
+}

kernel/src/arch/x86_64/gdt.rs

@@ -0,0 +1,147 @@
+use core::mem::size_of_val;
+use core::arch::asm;
+
+#[repr(packed)]
+struct Entry {
+    limit_lo: u16,
+    base_lo: u16,
+    base_mi: u8,
+    access: u8,
+    flags: u8,
+    base_hi: u8,
+}
+
+#[repr(packed)]
+struct Tss {
+    __res0: u32,
+    rsp0: u64,
+    rsp1: u64,
+    rsp2: u64,
+    __res1: u32,
+    ist1: u64,
+    ist2: u64,
+    ist3: u64,
+    ist4: u64,
+    ist5: u64,
+    ist6: u64,
+    ist7: u64,
+    __res2: u64,
+    __res3: u16,
+    iopb_base: u16,
+}
+
+#[repr(packed)]
+struct Pointer {
+    size: u16,
+    offset: usize,
+}
+
+impl Entry {
+    const FLAG_LONG: u8 = 1 << 5;
+    const ACC_PRESENT: u8 = 1 << 7;
+    const ACC_SYSTEM: u8 = 1 << 4;
+    const ACC_EXECUTE: u8 = 1 << 3;
+    const ACC_WRITE: u8 = 1 << 1;
+    const ACC_RING3: u8 = 3 << 5;
+    const ACC_ACCESS: u8 = 1 << 0;
+
+    const fn new(base: u32, limit: u32, flags: u8, access: u8) -> Self {
+        Self {
+            base_lo: (base & 0xFFFF) as u16,
+            base_mi: ((base >> 16) & 0xFF) as u8,
+            base_hi: ((base >> 24) & 0xFF) as u8,
+            access,
+            flags: (flags & 0xF0) | (((limit >> 16) & 0xF) as u8),
+            limit_lo: (limit & 0xFFFF) as u16,
+        }
+    }
+
+    const fn null() -> Self {
+        Self {
+            base_lo: 0,
+            base_mi: 0,
+            base_hi: 0,
+            access: 0,
+            flags: 0,
+            limit_lo: 0,
+        }
+    }
+}
+
+impl Tss {
+    const fn new() -> Self {
+        Self {
+            __res0: 0,
+            rsp0: 0,
+            rsp1: 0,
+            rsp2: 0,
+            __res1: 0,
+            ist1: 0,
+            ist2: 0,
+            ist3: 0,
+            ist4: 0,
+            ist5: 0,
+            ist6: 0,
+            ist7: 0,
+            __res2: 0,
+            __res3: 0,
+            iopb_base: 0,
+        }
+    }
+}
+
+const SIZE: usize = 7;
+#[no_mangle]
+static mut TSS: Tss = Tss::new();
+static mut GDT: [Entry; SIZE] = [
+    Entry::null(),
+    Entry::new(
+        0,
+        0,
+        Entry::FLAG_LONG,
+        Entry::ACC_PRESENT | Entry::ACC_SYSTEM | Entry::ACC_EXECUTE,
+    ),
+    Entry::new(
+        0,
+        0,
+        0,
+        Entry::ACC_PRESENT | Entry::ACC_SYSTEM | Entry::ACC_WRITE,
+    ),
+    Entry::new(
+        0,
+        0,
+        0,
+        Entry::ACC_PRESENT | Entry::ACC_SYSTEM | Entry::ACC_RING3 | Entry::ACC_WRITE,
+    ),
+    Entry::new(
+        0,
+        0,
+        Entry::FLAG_LONG,
+        Entry::ACC_PRESENT | Entry::ACC_SYSTEM | Entry::ACC_RING3 | Entry::ACC_EXECUTE,
+    ),
+    Entry::null(),
+    Entry::null(),
+];
+
+pub unsafe fn init() {
+    let tss_addr = &TSS as *const _ as usize;
+
+    GDT[5] = Entry::new(
+        (tss_addr & 0xFFFFFFFF) as u32,
+        size_of_val(&TSS) as u32 - 1,
+        Entry::FLAG_LONG,
+        Entry::ACC_ACCESS | Entry::ACC_PRESENT | Entry::ACC_EXECUTE,
+    );
+    core::ptr::write(&mut GDT[6] as *mut _ as *mut u64, (tss_addr >> 32) as u64);
+
+    let gdtr = Pointer {
+        size: size_of_val(&GDT) as u16 - 1,
+        offset: &GDT as *const _ as usize,
+    };
+    asm!(r#"
+        lgdt ({})
+
+        mov $0x28, %ax
+        ltr %ax
+    "#, in(reg) &gdtr, options(att_syntax));
+}

kernel/src/arch/x86_64/idt.S

@@ -0,0 +1,110 @@
+.macro isr_nerr, n
+exc_isr_\n:
+    cli
+    pushq $0
+    pushq $\n
+    jmp __x86_64_isr_common
+.endm
+
+// ISR for exception with an error code
+.macro isr_yerr, n
+exc_isr_\n:
+    cli
+    pushq $\n
+    jmp __x86_64_isr_common
+.endm
+
+.section .text
+__x86_64_isr_common:
+    push %rax
+    push %rcx
+    push %rdx
+    push %rbx
+    push %rbp
+    push %rsi
+    push %rdi
+    push %r8
+    push %r9
+    push %r10
+    push %r11
+    push %r12
+    push %r13
+    push %r14
+    push %r15
+
+    mov %rsp, %rdi
+    call __x86_64_exception_handler
+
+1:
+    cli
+    hlt
+    jmp 1b
+
+isr_nerr 0
+isr_nerr 1
+isr_nerr 2
+isr_nerr 3
+isr_nerr 4
+isr_nerr 5
+isr_nerr 6
+isr_nerr 7
+isr_yerr 8
+isr_nerr 9
+isr_yerr 10
+isr_yerr 11
+isr_yerr 12
+isr_yerr 13
+isr_yerr 14
+isr_nerr 15
+isr_nerr 16
+isr_yerr 17
+isr_nerr 18
+isr_nerr 19
+isr_nerr 20
+isr_nerr 21
+isr_nerr 22
+isr_nerr 23
+isr_nerr 24
+isr_nerr 25
+isr_nerr 26
+isr_nerr 27
+isr_nerr 28
+isr_nerr 29
+isr_yerr 30
+isr_nerr 31
+
+.section .rodata
+.global __x86_64_exception_vectors
+__x86_64_exception_vectors:
+    .quad exc_isr_0
+    .quad exc_isr_1
+    .quad exc_isr_2
+    .quad exc_isr_3
+    .quad exc_isr_4
+    .quad exc_isr_5
+    .quad exc_isr_6
+    .quad exc_isr_7
+    .quad exc_isr_8
+    .quad exc_isr_9
+    .quad exc_isr_10
+    .quad exc_isr_11
+    .quad exc_isr_12
+    .quad exc_isr_13
+    .quad exc_isr_14
+    .quad exc_isr_15
+    .quad exc_isr_16
+    .quad exc_isr_17
+    .quad exc_isr_18
+    .quad exc_isr_19
+    .quad exc_isr_20
+    .quad exc_isr_21
+    .quad exc_isr_22
+    .quad exc_isr_23
+    .quad exc_isr_24
+    .quad exc_isr_25
+    .quad exc_isr_26
+    .quad exc_isr_27
+    .quad exc_isr_28
+    .quad exc_isr_29
+    .quad exc_isr_30
+    .quad exc_isr_31

kernel/src/arch/x86_64/idt.rs

@@ -0,0 +1,73 @@
+use core::arch::{asm, global_asm};
+use core::mem::size_of_val;
+
+#[derive(Clone, Copy)]
+#[repr(packed)]
+pub struct Entry {
+    base_lo: u16,
+    selector: u16,
+    __res0: u8,
+    flags: u8,
+    base_hi: u16,
+    base_ex: u32,
+    __res1: u32,
+}
+
+#[repr(packed)]
+struct Pointer {
+    limit: u16,
+    offset: usize,
+}
+
+pub const SIZE: usize = 256;
+
+impl Entry {
+    pub const PRESENT: u8 = 1 << 7;
+    pub const INT32: u8 = 0xE;
+
+    pub const fn new(base: usize, selector: u16, flags: u8) -> Self {
+        Self {
+            base_lo: (base & 0xFFFF) as u16,
+            base_hi: ((base >> 16) & 0xFFFF) as u16,
+            base_ex: (base >> 32) as u32,
+            selector,
+            flags,
+            __res0: 0,
+            __res1: 0,
+        }
+    }
+
+    const fn empty() -> Self {
+        Self {
+            base_lo: 0,
+            base_hi: 0,
+            base_ex: 0,
+            selector: 0,
+            flags: 0,
+            __res0: 0,
+            __res1: 0,
+        }
+    }
+}
+
+static mut IDT: [Entry; SIZE] = [Entry::empty(); SIZE];
+
+pub unsafe fn init<F: FnOnce(&mut [Entry; SIZE]) -> ()>(f: F) {
+    extern "C" {
+        static __x86_64_exception_vectors: [usize; 32];
+    }
+
+    for (i, &entry) in __x86_64_exception_vectors.iter().enumerate() {
+        IDT[i] = Entry::new(entry, 0x08, Entry::PRESENT | Entry::INT32);
+    }
+
+    f(&mut IDT);
+
+    let idtr = Pointer {
+        limit: size_of_val(&IDT) as u16 - 1,
+        offset: &IDT as *const _ as usize,
+    };
+    asm!("lidt ({})", in(reg) &idtr, options(att_syntax));
+}
+
+global_asm!(include_str!("idt.S"), options(att_syntax));

kernel/src/arch/x86_64/intc.rs

@@ -0,0 +1,145 @@
+use crate::arch::x86_64::{
+    idt::{Entry as IdtEntry, SIZE as IDT_SIZE},
+    PortIo,
+};
+use crate::dev::{
+    irq::{IntController, IntSource, IrqContext},
+    Device,
+};
+use crate::sync::IrqSafeSpinLock;
+use libsys::error::Errno;
+use core::arch::global_asm;
+
+const ICW1_INIT: u8 = 0x10;
+const ICW1_ICW4: u8 = 0x01;
+
+const ICW4_8086: u8 = 0x01;
+
+pub(super) struct I8259 {
+    cmd_a: PortIo<u8>,
+    cmd_b: PortIo<u8>,
+    data_a: PortIo<u8>,
+    data_b: PortIo<u8>,
+
+    table: IrqSafeSpinLock<[Option<&'static (dyn IntSource + Sync)>; 15]>,
+}
+
+#[derive(Clone, Copy, Debug)]
+#[repr(transparent)]
+pub struct IrqNumber(u32);
+
+impl IrqNumber {
+    pub const MAX: u32 = 16;
+
+    pub const fn new(u: u32) -> Self {
+        if u > Self::MAX {
+            panic!();
+        }
+        Self(u)
+    }
+}
+
+impl Device for I8259 {
+    fn name(&self) -> &'static str {
+        "i8259-compatible IRQ controller"
+    }
+
+    unsafe fn enable(&self) -> Result<(), Errno> {
+        self.cmd_a.write(ICW1_INIT | ICW1_ICW4);
+        self.cmd_b.write(ICW1_INIT | ICW1_ICW4);
+        self.data_a.write(32);
+        self.data_b.write(32 + 8);
+        self.data_a.write(4);
+        self.data_b.write(2);
+
+        self.data_a.write(ICW4_8086);
+        self.data_b.write(ICW4_8086);
+
+        self.data_a.write(0xFE);
+        self.data_b.write(0xFF);
+
+        Ok(())
+    }
+}
+
+impl IntController for I8259 {
+    type IrqNumber = IrqNumber;
+
+    fn register_handler(
+        &self,
+        irq: Self::IrqNumber,
+        handler: &'static (dyn IntSource + Sync),
+    ) -> Result<(), Errno> {
+        if irq.0 == 0 {
+            return Err(Errno::InvalidArgument);
+        }
+
+        let index = (irq.0 - 1) as usize;
+        let mut lock = self.table.lock();
+        if lock[index].is_some() {
+            return Err(Errno::AlreadyExists);
+        }
+
+        lock[index] = Some(handler);
+        Ok(())
+    }
+
+    fn enable_irq(&self, irq: Self::IrqNumber) -> Result<(), Errno> {
+        let port = if irq.0 < 8 {
+            &self.data_a
+        } else {
+            &self.data_b
+        };
+
+        let mask = port.read() & !(1 << (irq.0 & 0x7));
+        port.write(mask);
+        Ok(())
+    }
+
+    fn handle_pending_irqs<'irq_context>(&'irq_context self, ic: &IrqContext<'irq_context>) {
+        let irq_number = ic.token();
+        assert!(irq_number > 0);
+
+        if irq_number > 8 {
+            self.cmd_b.write(0x20);
+        }
+        self.cmd_a.write(0x20);
+
+        {
+            let table = self.table.lock();
+            match table[irq_number - 1] {
+                None => panic!("No handler registered for irq{}", irq_number),
+                Some(handler) => {
+                    drop(table);
+                    handler.handle_irq().expect("irq handler failed")
+                }
+            }
+        }
+    }
+}
+
+impl I8259 {
+    pub const fn new() -> Self {
+        unsafe {
+            Self {
+                cmd_a: PortIo::new(0x20),
+                data_a: PortIo::new(0x21),
+                cmd_b: PortIo::new(0xA0),
+                data_b: PortIo::new(0xA1),
+                table: IrqSafeSpinLock::new([None; 15]),
+            }
+        }
+    }
+}
+
+pub fn map_isr_entries(entries: &mut [IdtEntry; IDT_SIZE]) {
+    extern "C" {
+        static __x86_64_irq_vectors: [usize; 16];
+    }
+
+    for (i, &entry) in unsafe { __x86_64_irq_vectors.iter().enumerate() } {
+        entries[i + 32] = IdtEntry::new(entry, 0x08, IdtEntry::PRESENT | IdtEntry::INT32);
+    }
+}
+
+global_asm!(include_str!("irq_vectors.S"), options(att_syntax));

kernel/src/arch/x86_64/intrin.rs

@@ -0,0 +1,16 @@
+use core::arch::asm;
+
+#[inline(always)]
+pub unsafe fn rdmsr(a: u32) -> u64 {
+    let mut eax: u32;
+    let mut edx: u32;
+    asm!("rdmsr", in("ecx") a, out("eax") eax, out("edx") edx);
+    (eax as u64) | ((edx as u64) << 32)
+}
+
+#[inline(always)]
+pub unsafe fn wrmsr(a: u32, b: u64) {
+    let eax = b as u32;
+    let edx = (b >> 32) as u32;
+    asm!("wrmsr", in("ecx") a, in("eax") eax, in("edx") edx);
+}

kernel/src/arch/x86_64/io.rs

@@ -0,0 +1,29 @@
+use core::marker::PhantomData;
+use core::arch::asm;
+
+pub struct PortIo<T> {
+    port: u16,
+    _pd: PhantomData<T>
+}
+
+impl<T> PortIo<T> {
+    pub const unsafe fn new(port: u16) -> Self {
+        Self { port, _pd: PhantomData }
+    }
+}
+
+impl PortIo<u8> {
+    pub fn read(&self) -> u8 {
+        let mut res: u8;
+        unsafe {
+            asm!("inb %dx, %al", in("dx") self.port, out("al") res, options(att_syntax));
+        }
+        res
+    }
+
+    pub fn write(&self, value: u8) {
+        unsafe {
+            asm!("outb %al, %dx", in("dx") self.port, in("al") value, options(att_syntax));
+        }
+    }
+}

kernel/src/arch/x86_64/irq_vectors.S

@@ -0,0 +1,115 @@
+.macro irq_entry no
+__x86_64_irq_\no:
+    cli
+    pushq $0
+    pushq $\no
+
+    push %rax
+    push %rcx
+    push %rdx
+    push %rbx
+    push %rbp
+    push %rsi
+    push %rdi
+
+    push %r8
+    push %r9
+    push %r10
+    push %r11
+    push %r12
+    push %r13
+    push %r14
+    push %r15
+
+    mov %rsp, %rdi
+    call __x86_64_irq_handler
+
+    jmp .
+.endm
+
+.section .text
+
+__x86_64_irq_0:
+    cli
+
+    push %rax
+    push %rcx
+    push %rdx
+    push %rbx
+    push %rbp
+    push %rsi
+    push %rdi
+
+    push %r8
+    push %r9
+    push %r10
+    push %r11
+    push %r12
+    push %r13
+    push %r14
+    push %r15
+
+    mov $0x3F8, %dx
+    mov $'T', %al
+    outb %al, %dx
+
+    mov $0x20, %al
+    mov $0x20, %dx
+    outb %al, %dx
+
+    call sched_yield
+
+    pop %r15
+    pop %r14
+    pop %r13
+    pop %r12
+    pop %r11
+    pop %r10
+    pop %r9
+    pop %r8
+
+    pop %rdi
+    pop %rsi
+    pop %rbp
+    pop %rbx
+    pop %rdx
+    pop %rcx
+    pop %rax
+
+    iretq
+
+irq_entry 1
+irq_entry 2
+irq_entry 3
+irq_entry 4
+irq_entry 5
+irq_entry 6
+irq_entry 7
+irq_entry 8
+irq_entry 9
+irq_entry 10
+irq_entry 11
+irq_entry 12
+irq_entry 13
+irq_entry 14
+irq_entry 15
+
+.section .rodata
+.global __x86_64_irq_vectors
+__x86_64_irq_vectors:
+    .quad __x86_64_irq_0
+    .quad __x86_64_irq_1
+    .quad __x86_64_irq_2
+    .quad __x86_64_irq_3
+    .quad __x86_64_irq_4
+    .quad __x86_64_irq_5
+    .quad __x86_64_irq_6
+    .quad __x86_64_irq_7
+    .quad __x86_64_irq_8
+    .quad __x86_64_irq_9
+    .quad __x86_64_irq_10
+    .quad __x86_64_irq_11
+    .quad __x86_64_irq_12
+    .quad __x86_64_irq_13
+    .quad __x86_64_irq_14
+    .quad __x86_64_irq_15

kernel/src/arch/x86_64/mod.rs

@@ -0,0 +1,59 @@
+use crate::dev::{serial::SerialDevice, display::StaticFramebuffer, irq::IntController};
+use core::arch::asm;
+
+mod uart;
+use uart::Uart;
+mod intc;
+use intc::I8259;
+
+mod io;
+pub(self) use io::PortIo;
+
+pub mod boot;
+pub mod table;
+pub mod context;
+pub mod intrin;
+pub mod reg;
+pub(self) mod syscall;
+pub(self) mod gdt;
+pub(self) mod idt;
+pub(self) mod exception;
+
+pub use syscall::SyscallFrame as ForkFrame;
+
+/// Masks IRQs and returns previous IRQ mask state
+///
+/// # Safety
+///
+/// Unsafe: disables IRQ handling temporarily
+#[inline(always)]
+pub unsafe fn irq_mask_save() -> u64 {
+    let mut res;
+    asm!("pushf; cli; pop {}", out(reg) res, options(att_syntax));
+    res
+}
+
+/// Restores IRQ mask state
+///
+/// # Safety
+///
+/// Unsafe: modifies interrupt behavior. Must only be used in
+/// conjunction with [irq_mask_save]
+#[inline(always)]
+pub unsafe fn irq_restore(state: u64) {
+    if state & (1 << 9) != 0 {
+        asm!("sti");
+    }
+}
+
+pub fn intc() -> &'static impl IntController {
+    &INTC
+}
+
+pub fn console() -> &'static impl SerialDevice {
+    &COM1
+}
+
+static COM1: Uart = unsafe { Uart::new(0x3F8) };
+static INTC: I8259 = I8259::new();
+pub(self) static DISPLAY: StaticFramebuffer = StaticFramebuffer::uninit();

kernel/src/arch/x86_64/reg/mod.rs

@@ -0,0 +1,135 @@
+macro_rules! wrap_msr {
+    ($struct_name:ident, $name:ident, $address:expr, $fields:tt) => {
+        register_bitfields! {
+            u64,
+            pub $name $fields
+        }
+
+        pub struct $struct_name;
+
+        impl Readable for $struct_name {
+            type T = u64;
+            type R = $name::Register;
+
+            #[inline(always)]
+            fn get(&self) -> Self::T {
+                unsafe {
+                    rdmsr($address)
+                }
+            }
+        }
+
+        impl Writeable for $struct_name {
+            type T = u64;
+            type R = $name::Register;
+
+            #[inline(always)]
+            fn set(&self, value: Self::T) {
+                unsafe {
+                    wrmsr($address, value);
+                }
+            }
+        }
+
+        pub const $name: $struct_name = $struct_name;
+    }
+}
+
+use tock_registers::{
+    interfaces::{Readable, Writeable},
+    register_bitfields,
+};
+use core::arch::asm;
+use crate::arch::x86_64::intrin::{rdmsr, wrmsr};
+
+// CRn registers
+register_bitfields! {
+    u64,
+    pub CR4 [
+        OSFXSR OFFSET(9) NUMBITS(1) [],
+        OSXMMEXCPT OFFSET(10) NUMBITS(1) []
+    ]
+}
+
+register_bitfields! {
+    u64,
+    pub CR0 [
+        EM OFFSET(2) NUMBITS(1) [],
+        MP OFFSET(1) NUMBITS(1) []
+    ]
+}
+
+pub struct Cr4;
+pub struct Cr0;
+
+impl Readable for Cr4 {
+    type T = u64;
+    type R = CR4::Register;
+
+    #[inline(always)]
+    fn get(&self) -> Self::T {
+        let mut res: u64;
+        unsafe {
+            asm!("mov %cr4, {}", out(reg) res, options(att_syntax))
+        }
+        res
+    }
+}
+
+impl Writeable for Cr4 {
+    type T = u64;
+    type R = CR4::Register;
+
+    #[inline(always)]
+    fn set(&self, value: Self::T) {
+        unsafe {
+            asm!("mov {}, %cr4", in(reg) value, options(att_syntax));
+        }
+    }
+}
+
+impl Readable for Cr0 {
+    type T = u64;
+    type R = CR0::Register;
+
+    #[inline(always)]
+    fn get(&self) -> Self::T {
+        let mut res: u64;
+        unsafe {
+            asm!("mov %cr0, {}", out(reg) res, options(att_syntax))
+        }
+        res
+    }
+}
+
+impl Writeable for Cr0 {
+    type T = u64;
+    type R = CR0::Register;
+
+    #[inline(always)]
+    fn set(&self, value: Self::T) {
+        unsafe {
+            asm!("mov {}, %cr0", in(reg) value, options(att_syntax));
+        }
+    }
+}
+
+pub const CR4: Cr4 = Cr4;
+pub const CR0: Cr0 = Cr0;
+
+wrap_msr!(MsrIa32Efer, MSR_IA32_EFER, 0xC0000080, [
+    SCE OFFSET(0) NUMBITS(1) [],
+    LME OFFSET(8) NUMBITS(1) [],
+    LMA OFFSET(10) NUMBITS(1) [],
+    NXE OFFSET(11) NUMBITS(1) []
+]);
+wrap_msr!(MsrIa32Lstar, MSR_IA32_LSTAR, 0xC0000082, [
+    VALUE OFFSET(0) NUMBITS(64) []
+]);
+wrap_msr!(MsrIa32Star, MSR_IA32_STAR, 0xC0000081, [
+    SYSCALL_CS_SS OFFSET(32) NUMBITS(8) [],
+    SYSRET_CS_SS OFFSET(48) NUMBITS(8) []
+]);
+wrap_msr!(MsrIa32Sfmask, MSR_IA32_SFMASK, 0xC0000084, [
+    IF OFFSET(9) NUMBITS(1) []
+]);

kernel/src/arch/x86_64/reg/msr_ia32_efer.rs

@@ -0,0 +1,13 @@
+use tock_registers::{
+    interfaces::{Readable, Writeable},
+    register_bitfields,
+};
+use crate::arch::x86_64::intrin::{rdmsr, wrmsr};
+
+register_bitfields! {
+    u64,
+    pub MSR_IA32_EFER [
+    ]
+}
+
+wrap_msr!(MSR_IA32_EFER, 0xC0000080);

kernel/src/arch/x86_64/syscall.S

@@ -0,0 +1,60 @@
+.section .text
+.global __x86_64_syscall_entry
+__x86_64_syscall_entry:
+    // Syscalls only happen from user space, so
+    // relying on TSS.RSP0 is safe here I guess
+    mov %rsp, scratch(%rip)
+    mov (4 + TSS)(%rip), %rsp
+
+    // Now on kernel stack
+    // Push the whole state
+    push %rcx   // saved %rip
+    push %r11   // saved %rflags
+    mov scratch(%rip), %r11
+    push %r11   // saved %rsp
+
+    push %r12
+    push %r13
+    push %r14
+    push %r15
+    push %rbp
+    push %rbx
+
+    push %rax
+
+    push %r9
+    push %r8
+    push %r10
+    push %rdx
+    push %rsi
+    push %rdi
+
+    mov %rsp, %rdi
+    call __x86_64_syscall
+
+    pop %rdi
+    pop %rsi
+    pop %rdx
+    pop %r10
+    pop %r8
+    pop %r9
+
+    pop %rax
+
+    pop %rbx
+    pop %rbp
+    pop %r15
+    pop %r14
+    pop %r13
+    pop %r12
+
+    pop %rdi
+    pop %r11
+    pop %rcx
+    mov %rdi, %rsp
+
+    sysretq
+
+.section .bss
+scratch:
+    .skip 8

kernel/src/arch/x86_64/syscall.rs

@@ -0,0 +1,53 @@
+use crate::arch::x86_64::reg::{MSR_IA32_EFER, MSR_IA32_LSTAR, MSR_IA32_SFMASK, MSR_IA32_STAR};
+use core::arch::global_asm;
+use tock_registers::interfaces::{ReadWriteable, Writeable};
+use libsys::abi::SystemCall;
+use crate::syscall;
+
+#[derive(Clone, Debug)]
+pub struct SyscallFrame {
+    x: [usize; 13],
+
+    saved_rsp: usize,
+    saved_rflags: usize,
+    saved_rip: usize,
+}
+
+pub(super) fn init() {
+    extern "C" {
+        fn __x86_64_syscall_entry();
+    }
+
+    MSR_IA32_SFMASK.write(MSR_IA32_SFMASK::IF::SET);
+    MSR_IA32_LSTAR.set(__x86_64_syscall_entry as u64);
+    MSR_IA32_STAR
+        .write(MSR_IA32_STAR::SYSRET_CS_SS.val(0x1B - 8) + MSR_IA32_STAR::SYSCALL_CS_SS.val(0x08));
+    MSR_IA32_EFER.modify(MSR_IA32_EFER::SCE::SET);
+}
+
+#[no_mangle]
+extern "C" fn __x86_64_syscall(frame: &mut SyscallFrame) {
+    let num = SystemCall::from_repr(frame.x[6]);
+    if num.is_none() {
+        todo!();
+    }
+    let num = num.unwrap();
+    if num == SystemCall::Fork {
+        match unsafe { syscall::sys_fork(frame) } {
+            Ok(pid) => frame.x[6] = u32::from(pid) as usize,
+            Err(err) => {
+                frame.x[6] = err.to_negative_isize() as usize;
+            }
+        }
+        return;
+    }
+
+    match syscall::syscall(num, &frame.x[..6]) {
+        Ok(val) => frame.x[6] = val,
+        Err(err) => {
+            frame.x[6] = err.to_negative_isize() as usize;
+        }
+    }
+}
+
+global_asm!(include_str!("syscall.S"), options(att_syntax));

kernel/src/arch/x86_64/table.rs

@@ -0,0 +1,286 @@
+use crate::mem::{
+    self,
+    virt::{AddressSpace, table::{MapAttributes, Entry as AbstractEntry}},
+    phys::{self, PageUsage}
+};
+use core::ops::{Index, IndexMut};
+use libsys::error::Errno;
+use core::arch::asm;
+
+#[derive(Clone, Copy)]
+#[repr(transparent)]
+pub struct Entry(u64);
+
+#[derive(Clone, Copy)]
+#[repr(C, align(0x1000))]
+pub struct Table {
+    entries: [Entry; 512]
+}
+
+#[repr(C, align(0x1000))]
+pub struct FixedTableGroup {
+    pml4: Table,
+    pdpt: Table,
+    pd: [Table; 16]
+}
+
+#[repr(transparent)]
+pub struct Space(Table);
+
+bitflags! {
+    /// Attributes attached to each translation [Entry]
+    pub struct RawAttributes: u64 {
+        const PRESENT = 1 << 0;
+        const WRITE = 1 << 1;
+        const USER = 1 << 2;
+        const BLOCK = 1 << 7;
+        const GLOBAL = 1 << 8;
+    }
+}
+
+// Upper mappings
+#[no_mangle]
+static mut KERNEL_FIXED: FixedTableGroup = FixedTableGroup {
+    pml4: Table::empty(),
+    pdpt: Table::empty(),
+    pd: [Table::empty(); 16]
+};
+
+impl TryFrom<MapAttributes> for RawAttributes {
+    type Error = Errno;
+
+    fn try_from(i: MapAttributes) -> Result<Self, Errno> {
+        let mut res = RawAttributes::empty();
+
+        if i.contains(MapAttributes::USER_READ) {
+            res |= RawAttributes::USER;
+        }
+        if i.contains(MapAttributes::USER_WRITE) || i.contains(MapAttributes::KERNEL_WRITE) {
+            res |= RawAttributes::WRITE;
+        }
+
+        Ok(res)
+    }
+}
+
+impl AbstractEntry for Entry {
+    fn from_parts(phys: usize, attrs: MapAttributes) -> Self {
+        let attrs = RawAttributes::try_from(attrs).unwrap();
+        Self((phys as u64) | attrs.bits() | 1)
+    }
+
+    fn is_present(self) -> bool {
+        self.0 & (1 << 0) != 0
+    }
+
+    fn is_table(self) -> bool {
+        self.0 & (1 << 7) == 0
+    }
+
+    fn target(self) -> usize {
+        (self.0 & !0xFFF) as usize
+    }
+}
+
+impl Entry {
+    const fn invalid() -> Self {
+        Self(0)
+    }
+}
+
+impl Table {
+    const fn empty() -> Self {
+        Self {
+            entries: [Entry::invalid(); 512]
+        }
+    }
+
+    /// Returns next-level translation table reference for `index`, if one is present.
+    /// If `index` represents a `Block`-type mapping, will return an error.
+    /// If `index` does not map to any translation table, will try to allocate, init and
+    /// map a new one, returning it after doing so.
+    pub fn next_level_table_or_alloc(&mut self, index: usize) -> Result<&'static mut Table, Errno> {
+        let entry = self[index];
+        if entry.is_present() {
+            if !entry.is_table() {
+                return Err(Errno::InvalidArgument);
+            }
+
+            Ok(unsafe { &mut *(mem::virtualize(entry.target()) as *mut _) })
+        } else {
+            let phys = phys::alloc_page(PageUsage::Paging)?;
+            let res = unsafe { &mut *(mem::virtualize(phys) as *mut Self) };
+            self[index] = Entry::from_parts(phys, MapAttributes::USER_WRITE | MapAttributes::USER_READ | MapAttributes::NOT_GLOBAL);
+            res.entries.fill(Entry::invalid());
+            Ok(res)
+        }
+    }
+
+    /// Returns next-level translation table reference for `index`, if one is present.
+    /// Same as [next_level_table_or_alloc], but returns `None` if no table is mapped.
+    pub fn next_level_table(&mut self, index: usize) -> Option<&'static mut Table> {
+        let entry = self[index];
+        if entry.is_present() {
+            if !entry.is_table() {
+                panic!("Entry is not a table: idx={}", index);
+            }
+
+            Some(unsafe { &mut *(mem::virtualize(entry.target()) as *mut _) })
+        } else {
+            None
+        }
+    }
+}
+
+impl Space {
+    const fn empty() -> Self {
+        Self(Table::empty())
+    }
+}
+
+impl AddressSpace for Space {
+    type Entry = Entry;
+
+    fn alloc_empty() -> Result<&'static mut Self, Errno> {
+        let pdpt_phys = unsafe {
+            &KERNEL_FIXED.pdpt as *const _ as usize - mem::KERNEL_OFFSET
+        };
+        let page = phys::alloc_page(PageUsage::Paging)?;
+        let res = unsafe { &mut *(mem::virtualize(page) as *mut Self) };
+        res.0.entries[..511].fill(Entry::invalid());
+        res.0.entries[511] = Entry::from_parts(pdpt_phys, MapAttributes::SHARE_OUTER | MapAttributes::KERNEL_EXEC | MapAttributes::KERNEL_WRITE | MapAttributes::NOT_GLOBAL);
+        Ok(res)
+    }
+
+    fn release(space: &mut Self) {
+        todo!()
+    }
+
+    fn address_phys(&mut self) -> usize {
+        todo!();
+    }
+
+    fn read_last_level_entry(&mut self, virt: usize) -> Result<Entry, Errno> {
+        let l0i = virt >> 39;
+        let l1i = (virt >> 30) & 0x1FF;
+        let l2i = (virt >> 21) & 0x1FF;
+        let l3i = (virt >> 12) & 0x1FF;
+
+        let l0_table = self.0.next_level_table(l0i).ok_or(Errno::DoesNotExist)?;
+        let l1_table = l0_table.next_level_table(l1i).ok_or(Errno::DoesNotExist)?;
+        let l2_table = l1_table.next_level_table(l2i).ok_or(Errno::DoesNotExist)?;
+
+        let entry = l2_table[l3i];
+        if entry.is_present() {
+            Ok(entry)
+        } else {
+            Err(Errno::DoesNotExist)
+        }
+    }
+
+    fn write_last_level_entry(
+        &mut self,
+        virt: usize,
+        entry: Entry,
+        map_intermediate: bool,
+    ) -> Result<(), Errno> {
+        let l0i = virt >> 39;
+        let l1i = (virt >> 30) & 0x1FF;
+        let l2i = (virt >> 21) & 0x1FF;
+        let l3i = (virt >> 12) & 0x1FF;
+
+        let l0_table = self.0.next_level_table_or_alloc(l0i)?;
+        let l1_table = l0_table.next_level_table_or_alloc(l1i)?;
+        let l2_table = l1_table.next_level_table_or_alloc(l2i)?;
+
+        if l2_table[l3i].is_present() {
+            warnln!("Entry already exists for address: virt={:#x}, prev={:#x}, new={:#x}", virt, l2_table[l3i].target(), entry.target());
+            Err(Errno::AlreadyExists)
+        } else {
+            l2_table[l3i] = entry;
+            unsafe {
+                core::arch::asm!("invlpg ({})", in(reg) virt, options(att_syntax));
+            }
+            #[cfg(feature = "verbose")]
+            debugln!("{:#p} Map {:#x} -> {:#x}", self, virt, entry.target());
+            Ok(())
+        }
+    }
+
+    /// Performs a copy of the address space, cloning data owned by it
+    fn fork(&mut self) -> Result<&'static mut Self, Errno> {
+        let res = Self::alloc_empty()?;
+        let pdpt0 = self.0.next_level_table(0).unwrap();
+
+        for pdpti in 0..512 {
+            if let Some(pd) = pdpt0.next_level_table(pdpti) {
+                for pdi in 0..512 {
+                    if let Some(pt) = pd.next_level_table(pdi) {
+                        for pti in 0..512 {
+                            let entry = pt[pti];
+
+                            if !entry.is_present() {
+                                continue;
+                            }
+
+                            assert!(entry.is_table());
+
+                            todo!();
+                            // let src_phys = unsafe { entry.address_unchecked() };
+                            // let virt_addr = (l0i << 30) | (l1i << 21) | (l2i << 12);
+                            // let dst_phys = unsafe { phys::fork_page(src_phys)? };
+
+                            // let mut flags = unsafe { entry.fork_flags() };
+                            // if dst_phys != src_phys {
+                            //     todo!();
+                            //     // res.map(virt_addr, dst_phys, flags)?;
+                            // } else {
+                            //     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)?;
+                            // }
+                        }
+                    }
+                }
+            }
+        }
+        Ok(res)
+    }
+
+}
+
+impl Index<usize> for Table {
+    type Output = Entry;
+
+    fn index(&self, index: usize) -> &Self::Output {
+        &self.entries[index]
+    }
+}
+
+impl IndexMut<usize> for Table {
+    fn index_mut(&mut self, index: usize) -> &mut Self::Output {
+        &mut self.entries[index]
+    }
+}
+
+pub fn enable() -> Result<(), Errno> {
+    unsafe {
+        // Remove the lower mapping
+        KERNEL_FIXED.pml4.entries[0] = Entry::invalid();
+
+        // Flush the TLB by reloading cr3
+        asm!("mov %cr3, %rax; mov %rax, %cr3", options(att_syntax));
+    }
+    Ok(())
+}

kernel/src/arch/x86_64/uart.rs

@@ -0,0 +1,50 @@
+use crate::arch::x86_64::PortIo;
+use libsys::error::Errno;
+use crate::dev::{
+    tty::{CharRing, TtyDevice},
+    irq::{IntController, IntSource},
+    serial::SerialDevice,
+    Device,
+};
+
+#[derive(TtyCharDevice)]
+pub(super) struct Uart {
+    dr: PortIo<u8>,
+    ring: CharRing<16>
+}
+
+impl Device for Uart {
+    fn name(&self) -> &'static str {
+        "x86 COM-port"
+    }
+
+    unsafe fn enable(&self) -> Result<(), Errno> {
+        Ok(())
+    }
+}
+
+impl TtyDevice<16> for Uart {
+    fn ring(&self) -> &CharRing<16> {
+        &self.ring
+    }
+}
+
+impl SerialDevice for Uart {
+    fn send(&self, byte: u8) -> Result<(), Errno> {
+        self.dr.write(byte);
+        Ok(())
+    }
+
+    fn recv(&self, _blocking: bool) -> Result<u8, Errno> {
+        todo!()
+    }
+}
+
+impl Uart {
+    pub const unsafe fn new(base: u16) -> Self {
+        Self {
+            dr: PortIo::new(base),
+            ring: CharRing::new()
+        }
+    }
+}

kernel/src/debug.rs

@@ -11,20 +11,108 @@
 //! * [warnln!]
 //! * [errorln!]
 
-use crate::dev::serial::SerialDevice;
+use crate::dev::{
+    display::{Display, FramebufferInfo},
+    serial::SerialDevice,
+};
+use crate::font;
+use crate::sync::IrqSafeSpinLock;
+use core::convert::TryFrom;
 use core::fmt;
+use libsys::{debug::TraceLevel, error::Errno, mem::memcpy};
+
+pub static LEVEL: Level = Level::Debug;
+static COLOR_MAP: [u32; 16] = [
+    0x000000, 0x0000AA, 0x00AA00, 0x00AAAA, 0xAA0000, 0xAA00AA, 0xAA5500, 0xAAAAAA, 0x555555,
+    0x5555FF, 0x55FF55, 0x55FFFF, 0xFF5555, 0xFF55FF, 0xFFFF55, 0xFFFFFF,
+];
+static ATTR_MAP: [usize; 10] = [0, 4, 2, 6, 1, 5, 3, 7, 7, 7];
+static DISPLAY: IrqSafeSpinLock<FramebufferOutput> = IrqSafeSpinLock::new(FramebufferOutput {
+    display: None,
+    col: 0,
+    row: 0,
+    fg: 0xBBBBBB,
+    bg: 0x000000,
+    esc: EscapeState::None,
+    esc_argv: [0; 8],
+    esc_argc: 0,
+});
+
+enum EscapeState {
+    None,
+    Esc,
+    Data,
+}
+
+struct FramebufferOutput {
+    display: Option<&'static dyn Display>,
+    row: usize,
+    col: usize,
+    fg: u32,
+    bg: u32,
+    esc: EscapeState,
+    esc_argv: [usize; 8],
+    esc_argc: usize,
+}
+
+impl fmt::Write for FramebufferOutput {
+    fn write_str(&mut self, s: &str) -> fmt::Result {
+        if self.display.is_none() {
+            return Ok(());
+        }
+        let fb = self.display.unwrap().framebuffer().unwrap();
+
+        for ch in s.chars() {
+            self.putc(&fb, ch);
+        }
+
+        Ok(())
+    }
+}
+
+pub fn set_display(disp: &'static dyn Display) {
+    DISPLAY.lock().display = Some(disp);
+}
 
 /// Kernel logging levels
-#[derive(Clone, Copy, PartialEq)]
+#[derive(Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Debug)]
+#[repr(u32)]
 pub enum Level {
     /// Debugging information
-    Debug,
+    Debug = 1,
     /// General informational messages
-    Info,
+    Info = 2,
     /// Non-critical warnings
-    Warn,
+    Warn = 3,
     /// Critical errors
-    Error,
+    Error = 4,
+}
+
+impl TryFrom<u32> for Level {
+    type Error = Errno;
+
+    #[inline(always)]
+    fn try_from(l: u32) -> Result<Level, Errno> {
+        match l {
+            1 => Ok(Level::Debug),
+            2 => Ok(Level::Info),
+            3 => Ok(Level::Warn),
+            4 => Ok(Level::Error),
+            _ => Err(Errno::InvalidArgument),
+        }
+    }
+}
+
+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> {
@@ -101,13 +189,117 @@ macro_rules! errorln {
 }
 
 #[doc(hidden)]
-pub fn _debug(_level: Level, args: fmt::Arguments) {
+pub fn _debug(level: Level, args: fmt::Arguments) {
     use crate::arch::machine;
     use fmt::Write;
 
-    SerialOutput {
-        inner: machine::console(),
+    if level > Level::Debug {
+        DISPLAY.lock().write_fmt(args).ok();
+    }
+
+    if level >= LEVEL {
+        SerialOutput {
+            inner: machine::console(),
+        }
+        .write_fmt(args)
+        .ok();
+    }
+}
+
+impl FramebufferOutput {
+    const CW: usize = 8;
+    const CH: usize = 12;
+
+    pub fn set_char(&mut self, fb: &FramebufferInfo, x: usize, y: usize, ch: char) {
+        if (x + 1) * Self::CW >= fb.width || (y + 1) * Self::CH >= fb.height {
+            return;
+        }
+        font::get().draw(fb, x * Self::CW, y * Self::CH, ch, self.fg, self.bg);
+    }
+
+    pub fn scroll(&mut self, fb: &FramebufferInfo) {
+        let stride = 4 * Self::CH * fb.width;
+        let h = fb.height / Self::CH - 1;
+        if self.row == h {
+            for y in 0..(h - 1) {
+                unsafe {
+                    memcpy(
+                        (fb.virt_base + stride * y) as *mut u8,
+                        (fb.virt_base + (y + 1) * stride) as *const u8,
+                        stride,
+                    );
+                }
+            }
+            self.row = h - 1;
+        }
+    }
+
+    pub fn putc(&mut self, fb: &FramebufferInfo, ch: char) {
+        match self.esc {
+            EscapeState::None => {
+                match ch {
+                    '\x1B' => {
+                        self.esc = EscapeState::Esc;
+                        self.esc_argv.fill(0);
+                        self.esc_argc = 0;
+                    }
+                    ' '..='\x7E' => {
+                        self.set_char(fb, self.col, self.row, ch);
+
+                        // Advance the cursor
+                        self.col += 1;
+                        if (self.col + 1) * Self::CW >= fb.width {
+                            self.col = 0;
+                            self.row += 1;
+                        }
+                    }
+                    '\n' => {
+                        self.col = 0;
+                        self.row += 1;
+                    }
+                    _ => {}
+                }
+            }
+            EscapeState::Esc => match ch {
+                '[' => {
+                    self.esc = EscapeState::Data;
+                }
+                _ => {
+                    self.esc = EscapeState::None;
+                }
+            },
+            EscapeState::Data => {
+                match ch {
+                    '0'..='9' => {
+                        self.esc_argv[self.esc_argc] *= 10;
+                        self.esc_argv[self.esc_argc] += (ch as u8 - b'0') as usize;
+                    }
+                    ';' => {
+                        self.esc_argc += 1;
+                    }
+                    _ => {
+                        self.esc_argc += 1;
+                        self.esc = EscapeState::None;
+                    }
+                }
+
+                match ch {
+                    'm' => {
+                        for i in 0..self.esc_argc {
+                            let item = self.esc_argv[i];
+                            if item / 10 == 4 {
+                                self.bg = COLOR_MAP[ATTR_MAP[(item % 10) as usize]];
+                            }
+                            if item / 10 == 3 {
+                                self.fg = COLOR_MAP[ATTR_MAP[(item % 10) as usize]];
+                            }
+                        }
+                    }
+                    _ => {}
+                }
+            }
+        };
+
+        self.scroll(fb);
     }
-    .write_fmt(args)
-    .ok();
 }

kernel/src/dev/display.rs

@@ -0,0 +1,58 @@
+use crate::dev::Device;
+use libsys::error::Errno;
+use crate::util::InitOnce;
+
+pub struct FramebufferInfo {
+    pub width: usize,
+    pub height: usize,
+    pub phys_base: usize,
+    pub virt_base: usize
+}
+
+pub trait Display: Device {
+    fn set_mode(&self, mode: DisplayMode) -> Result<(), Errno>;
+    fn framebuffer<'a>(&'a self) -> Result<&'a FramebufferInfo, Errno>;
+}
+
+pub struct DisplayMode {
+    width: u16,
+    height: u16,
+}
+
+pub struct StaticFramebuffer {
+    framebuffer: InitOnce<FramebufferInfo>
+}
+
+impl Device for StaticFramebuffer {
+    fn name(&self) -> &'static str {
+        "Generic framebuffer device"
+    }
+
+    unsafe fn enable(&self) -> Result<(), Errno> {
+        Ok(())
+    }
+}
+
+impl Display for StaticFramebuffer {
+    fn set_mode(&self, mode: DisplayMode) -> Result<(), Errno> {
+        Err(Errno::InvalidOperation)
+    }
+
+    fn framebuffer(&self) -> Result<&FramebufferInfo, Errno> {
+        if let Some(fb) = self.framebuffer.as_ref_option() {
+            Ok(fb)
+        } else {
+            Err(Errno::InvalidOperation)
+        }
+    }
+}
+
+impl StaticFramebuffer {
+    pub const fn uninit() -> Self {
+        Self { framebuffer: InitOnce::new() }
+    }
+
+    pub fn set_framebuffer(&self, framebuffer: FramebufferInfo) {
+        self.framebuffer.init(framebuffer);
+    }
+}

kernel/src/dev/irq.rs

@@ -5,6 +5,7 @@ use libsys::error::Errno;
 
 /// Token to indicate the local core is running in IRQ context
 pub struct IrqContext<'irq_context> {
+    token: usize,
     _0: PhantomData<&'irq_context ()>,
 }
 
@@ -45,7 +46,11 @@ impl<'q> IrqContext<'q> {
     ///
     /// Only allowed to be constructed in top-level IRQ handlers
     #[inline(always)]
-    pub unsafe fn new() -> Self {
-        Self { _0: PhantomData }
+    pub unsafe fn new(token: usize) -> Self {
+        Self { token, _0: PhantomData }
+    }
+
+    pub const fn token(&self) -> usize {
+        self.token
     }
 }

kernel/src/dev/mod.rs

@@ -3,14 +3,16 @@
 use libsys::error::Errno;
 
 // Device classes
-pub mod fdt;
-pub mod gpio;
+// pub mod fdt;
+// pub mod gpio;
 pub mod irq;
-pub mod pci;
-pub mod rtc;
-pub mod sd;
+pub mod display;
+// pub mod pci;
+// pub mod rtc;
+// pub mod sd;
 pub mod serial;
 pub mod timer;
+pub mod pseudo;
 pub mod tty;
 
 /// Generic device trait

kernel/src/dev/pseudo.rs

@@ -0,0 +1,99 @@
+use crate::dev::{
+    serial::SerialDevice,
+    Device,
+};
+use crate::mem::virt::DeviceMemoryIo;
+use crate::sync::IrqSafeSpinLock;
+use crate::util::InitOnce;
+use libsys::{error::Errno, ioctl::IoctlCmd};
+use core::sync::atomic::{AtomicU32, Ordering};
+use tock_registers::{
+    interfaces::{ReadWriteable, Readable, Writeable},
+    register_bitfields, register_structs,
+    registers::{ReadOnly, ReadWrite, WriteOnly},
+};
+use vfs::CharDevice;
+
+pub struct Random {
+    state: AtomicU32
+}
+pub struct Zero;
+
+impl Device for Random {
+    fn name(&self) -> &'static str {
+        "Pseudo-random device"
+    }
+
+    unsafe fn enable(&self) -> Result<(), Errno> {
+        Ok(())
+    }
+}
+
+impl CharDevice for Random {
+    fn read(&self, _blocking: bool, data: &mut [u8]) -> Result<usize, Errno> {
+        for byte in data.iter_mut() {
+            *byte = self.read_single() as u8;
+        }
+        Ok(data.len())
+    }
+
+    fn write(&self, _blocking: bool, _data: &[u8]) -> Result<usize, Errno> {
+        Ok(0)
+    }
+
+    fn is_ready(&self, _write: bool) -> Result<bool, Errno> {
+        Ok(true)
+    }
+
+    fn ioctl(&self, _cmd: IoctlCmd, _ptr: usize, _lim: usize) -> Result<usize, Errno> {
+        Err(Errno::InvalidArgument)
+    }
+}
+
+
+impl Device for Zero {
+    fn name(&self) -> &'static str {
+        "Zero device"
+    }
+
+    unsafe fn enable(&self) -> Result<(), Errno> {
+        Ok(())
+    }
+}
+
+impl CharDevice for Zero {
+    fn read(&self, _blocking: bool, data: &mut [u8]) -> Result<usize, Errno> {
+        data.fill(0);
+        Ok(data.len())
+    }
+
+    fn write(&self, _blocking: bool, _data: &[u8]) -> Result<usize, Errno> {
+        Ok(0)
+    }
+
+    fn is_ready(&self, _write: bool) -> Result<bool, Errno> {
+        Ok(true)
+    }
+
+    fn ioctl(&self, _cmd: IoctlCmd, _ptr: usize, _lim: usize) -> Result<usize, Errno> {
+        Err(Errno::InvalidArgument)
+    }
+}
+
+impl Random {
+    pub fn set_state(&self, state: u32) {
+        self.state.store(state, Ordering::Release);
+    }
+
+    pub fn read_single(&self) -> u32 {
+        let mut x = self.state.load(Ordering::Acquire);
+        x ^= x << 13;
+        x ^= x >> 7;
+        x ^= x << 17;
+        self.state.store(x, Ordering::Release);
+        x
+    }
+}
+
+pub static RANDOM: Random = Random { state: AtomicU32::new(0) };
+pub static ZERO: Zero = Zero;

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(Pid::try_from(*src)?);
+                Ok(0)
+            },
             _ => Err(Errno::InvalidArgument)
         }
     }
@@ -110,6 +119,20 @@ 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 {
+                    // TODO
+                    // proc.set_signal(Signal::Interrupt);
+                }
+            }
+            return;
+        }
+
         self.ring().putc(byte, false).ok();
     }
 
@@ -232,17 +255,19 @@ 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/font.rs

@@ -0,0 +1,49 @@
+use crate::util::InitOnce;
+use libsys::mem::read_le32;
+use crate::dev::display::FramebufferInfo;
+
+static FONT_DATA: &[u8] = include_bytes!("../../etc/default8x16.psfu");
+static FONT: InitOnce<Font> = InitOnce::new();
+
+pub struct Font {
+    char_width: usize,
+    char_height: usize,
+    bytes_per_glyph: usize,
+    data: &'static [u8],
+}
+
+impl Font {
+    pub fn draw(&self, fb: &FramebufferInfo, bx: usize, by: usize, ch: char, fg: u32, bg: u32) {
+        if ch >= ' ' && ch < '\x7B' {
+            let char_data = &self.data[ch as usize * self.bytes_per_glyph..];
+
+            for iy in 0..self.char_height {
+                for ix in 0..self.char_width {
+                    let cx = self.char_width - ix - 1;
+                    let ptr = fb.virt_base + (ix + bx + (iy + by) * fb.width) * 4;
+                    let value = if char_data[iy + (cx) / 8] & (1 << (cx & 0x7)) != 0 {
+                        fg
+                    } else {
+                        bg
+                    };
+                    unsafe { core::ptr::write_volatile(ptr as *mut u32, value) }
+                }
+            }
+        }
+    }
+}
+
+pub fn init() {
+    assert_eq!(read_le32(&FONT_DATA[..]), 0x864ab572);
+
+    FONT.init(Font {
+        char_width: read_le32(&FONT_DATA[28..]) as usize,
+        char_height: read_le32(&FONT_DATA[24..]) as usize,
+        bytes_per_glyph: read_le32(&FONT_DATA[20..]) as usize,
+        data: &FONT_DATA[32..]
+    });
+}
+
+pub fn get() -> &'static Font {
+    FONT.get()
+}

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
@@ -24,10 +26,11 @@ pub fn root() -> &'static VnodeRef {
     DEVFS_ROOT.get()
 }
 
-fn _add_char_device(dev: &'static dyn CharDevice, name: &str) -> Result<(), Errno> {
+pub fn add_named_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);
@@ -53,5 +56,5 @@ pub fn add_char_device(dev: &'static dyn CharDevice, kind: CharDeviceType) -> Re
 
     let name = core::str::from_utf8(&buf[..=prefix.len()]).map_err(|_| Errno::InvalidArgument)?;
 
-    _add_char_device(dev, name)
+    add_named_char_device(dev, name)
 }

kernel/src/fs/mod.rs

@@ -3,9 +3,12 @@ use crate::mem::{
     self,
     phys::{self, PageUsage},
 };
+use libsys::{error::Errno, stat::MountOptions};
+use vfs::VnodeRef;
 use memfs::BlockAllocator;
 
 pub mod devfs;
+pub mod sysfs;
 
 /// Allocator implementation for memfs
 #[derive(Clone, Copy)]
@@ -25,3 +28,14 @@ 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();
+
+    match fs_name {
+        "devfs" => Ok(devfs::root().clone()),
+        "sysfs" => Ok(sysfs::root().clone()),
+        _ => todo!()
+    }
+}

kernel/src/fs/sysfs.rs

@@ -0,0 +1,169 @@
+use crate::util::InitOnce;
+use alloc::boxed::Box;
+use core::sync::atomic::{AtomicUsize, Ordering};
+use fs_macros::auto_inode;
+use libsys::{
+    error::Errno,
+    stat::{FileMode, OpenFlags, Stat},
+};
+use vfs::{CharDevice, CharDeviceWrapper, Vnode, VnodeImpl, VnodeKind, VnodeRef};
+use core::fmt::{self, Write};
+use core::str::FromStr;
+use crate::debug::{self, Level};
+
+struct NodeData<
+    R: Fn(&mut [u8]) -> Result<usize, Errno>,
+    W: Fn(&[u8]) -> Result<usize, Errno>,
+> {
+    read_func: R,
+    write_func: W,
+}
+
+struct BufferWriter<'a> {
+    dst: &'a mut [u8],
+    pos: usize
+}
+
+impl<'a> fmt::Write for BufferWriter<'a> {
+    fn write_str(&mut self, s: &str) -> fmt::Result {
+        for byte in s.bytes() {
+            if self.pos == self.dst.len() {
+                todo!();
+            }
+            self.dst[self.pos] = byte;
+            self.pos += 1;
+        }
+        Ok(())
+    }
+}
+
+impl<'a> BufferWriter<'a> {
+    pub const fn new(dst: &'a mut [u8]) -> Self {
+        Self { dst, pos: 0 }
+    }
+
+    pub const fn count(&self) -> usize {
+        self.pos
+    }
+}
+
+#[auto_inode]
+impl<
+        R: Fn(&mut [u8]) -> Result<usize, Errno>,
+        W: Fn(&[u8]) -> Result<usize, Errno>,
+    > VnodeImpl for NodeData<R, W>
+{
+    fn open(&mut self, _node: VnodeRef, _mode: OpenFlags) -> Result<usize, Errno> {
+        Ok(0)
+    }
+
+    fn close(&mut self, _node: VnodeRef) -> Result<(), Errno> {
+        Ok(())
+    }
+
+    fn read(&mut self, _node: VnodeRef, pos: usize, data: &mut [u8]) -> Result<usize, Errno> {
+        if pos != 0 {
+            // TODO handle this
+            Ok(0)
+        } else {
+            (self.read_func)(data)
+        }
+    }
+
+    fn write(&mut self, _node: VnodeRef, pos: usize, data: &[u8]) -> Result<usize, Errno> {
+        if pos != 0 {
+            todo!();
+        }
+        (self.write_func)(data)
+    }
+}
+
+impl<
+        R: Fn(&mut [u8]) -> Result<usize, Errno>,
+        W: Fn(&[u8]) -> Result<usize, Errno>,
+    > NodeData<R, W>
+{
+    pub const fn new(read_func: R, write_func: W) -> Self {
+        Self {
+            read_func,
+            write_func,
+        }
+    }
+}
+
+static SYSFS_ROOT: InitOnce<VnodeRef> = InitOnce::new();
+static TEST_COUNTER: AtomicUsize = AtomicUsize::new(0);
+
+// TODO subdirs
+fn add_generic_node<R, W>(parent: Option<VnodeRef>, name: &str, mode: FileMode, read: R, write: W)
+where
+    R: Fn(&mut [u8]) -> Result<usize, Errno> + 'static,
+    W: Fn(&[u8]) -> Result<usize, Errno> + 'static,
+{
+    let node = Vnode::new(name, VnodeKind::Regular, Vnode::CACHE_STAT);
+    node.props_mut().mode = mode | FileMode::S_IFREG;
+    node.set_data(Box::new(NodeData::new(read, write)));
+
+    if let Some(parent) = parent {
+        parent.attach(node);
+    } else {
+        SYSFS_ROOT.get().attach(node);
+    }
+}
+
+pub fn add_read_write_node<R, W>(parent: Option<VnodeRef>, name: &str, read: R, write: W)
+where
+    R: Fn(&mut [u8]) -> Result<usize, Errno> + 'static,
+    W: Fn(&[u8]) -> Result<usize, Errno> + 'static,
+{
+    add_generic_node(parent, name, FileMode::from_bits(0o600).unwrap(), read, write)
+}
+
+pub fn add_read_node<R>(parent: Option<VnodeRef>, name: &str, read: R) where R: Fn(&mut [u8]) -> Result<usize, Errno> + 'static {
+    add_generic_node(parent, name, FileMode::from_bits(0o400).unwrap(), read, |_| Err(Errno::ReadOnly))
+}
+
+pub fn add_directory(parent: Option<VnodeRef>, name: &str) -> Result<VnodeRef, Errno> {
+    let node = Vnode::new(name, VnodeKind::Directory, Vnode::CACHE_READDIR | Vnode::CACHE_STAT);
+    node.props_mut().mode = FileMode::from_bits(0o500).unwrap() | FileMode::S_IFDIR;
+
+    if let Some(parent) = parent {
+        parent.attach(node.clone());
+    } else {
+        SYSFS_ROOT.get().attach(node.clone());
+    }
+
+    Ok(node)
+}
+
+pub fn root() -> &'static VnodeRef {
+    SYSFS_ROOT.get()
+}
+
+pub fn init() {
+    let node = Vnode::new("", VnodeKind::Directory, Vnode::CACHE_READDIR | Vnode::CACHE_STAT);
+    node.props_mut().mode = FileMode::default_dir();
+    SYSFS_ROOT.init(node);
+
+    let debug_dir = add_directory(None, "debug").unwrap();
+
+    add_read_write_node(Some(debug_dir.clone()), "level", |buf| {
+        let mut writer = BufferWriter::new(buf);
+        write!(&mut writer, "{}\n", debug::LEVEL as u32).map_err(|_| Errno::InvalidArgument)?;
+        Ok(writer.count())
+    }, |buf| {
+        let s = core::str::from_utf8(buf).map_err(|_| Errno::InvalidArgument)?;
+        let value = u32::from_str(s).map_err(|_| Errno::InvalidArgument).and_then(Level::try_from)?;
+        todo!()
+    });
+
+    add_read_node(None, "uptime", |buf| {
+        use crate::arch::machine;
+        use crate::dev::timer::TimestampSource;
+
+        let mut writer = BufferWriter::new(buf);
+        // let time = machine::local_timer().timestamp()?;
+        // write!(&mut writer, "{} {}\n", time.as_secs(), time.subsec_nanos()).map_err(|_| Errno::InvalidArgument)?;
+        Ok(writer.count())
+    });
+}

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 +12,9 @@
     panic_info_message,
     alloc_error_handler,
     linked_list_cursors,
-    const_btree_new
+    const_btree_new,
+    core_intrinsics,
+    const_generics_defaults,
 )]
 #![no_std]
 #![no_main]
@@ -28,11 +30,12 @@ extern crate alloc;
 
 #[macro_use]
 pub mod debug;
-
+//
 pub mod arch;
 pub mod config;
 pub mod dev;
 pub mod fs;
+pub mod font;
 pub mod init;
 pub mod mem;
 pub mod proc;
@@ -42,9 +45,9 @@ pub mod util;
 
 #[panic_handler]
 fn panic_handler(pi: &core::panic::PanicInfo) -> ! {
-    unsafe {
-        asm!("msr daifset, #2");
-    }
+    // unsafe {
+    //     asm!("msr daifset, #2");
+    // }
 
     errorln!("Panic: {:?}", pi);
     // TODO

kernel/src/mem/heap.rs

@@ -59,7 +59,7 @@ static HEAP: InitOnce<IrqSafeSpinLock<Heap>> = InitOnce::new();
 pub unsafe fn init(base: usize, size: usize) {
     let heap = Heap { base, size, ptr: 0 };
 
-    infoln!("Kernel heap: {:#x}..{:#x}", base, base + size);
+    // infoln!("Kernel heap: {:#x}..{:#x}", base, base + size);
 
     HEAP.init(IrqSafeSpinLock::new(heap));
 }

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/phys/mod.rs

@@ -9,7 +9,7 @@ mod manager;
 mod reserved;
 
 use manager::{Manager, SimpleManager, MANAGER};
-pub use reserved::ReservedRegion;
+pub use reserved::{ReservedRegion, reserve};
 
 type ManagerImpl = SimpleManager;
 
@@ -213,7 +213,10 @@ pub unsafe fn init_from_iter<T: Iterator<Item = MemoryRegion> + Clone>(iter: T)
     // Step 1. Count available memory
     let mut total_pages = 0usize;
     for reg in iter.clone() {
-        total_pages += (reg.end - reg.start) / PAGE_SIZE;
+        let upper = (reg.end - mem_base) / PAGE_SIZE;
+        if upper > total_pages {
+            total_pages = upper;
+        }
     }
     // TODO maybe instead of size_of::<...> use Layout?
     let need_pages = ((total_pages * size_of::<PageInfo>()) + 0xFFF) / 0x1000;

kernel/src/mem/virt/mod.rs

@@ -2,18 +2,18 @@
 
 use core::marker::PhantomData;
 use core::ops::Deref;
-use cortex_a::asm::barrier::{self, dsb, isb};
-use cortex_a::registers::TTBR0_EL1;
+// use cortex_a::asm::barrier::{self, dsb, isb};
+// use cortex_a::registers::TTBR0_EL1;
 use libsys::error::Errno;
-use tock_registers::interfaces::Writeable;
-
+// use tock_registers::interfaces::Writeable;
+//
 pub mod table;
-pub use table::{Entry, MapAttributes, Space, Table};
-pub mod fixed;
-pub use fixed::FixedTableGroup;
+pub use table::{AddressSpace, Space};
 
-#[no_mangle]
-static mut KERNEL_TTBR1: FixedTableGroup = FixedTableGroup::empty();
+// pub use table::{Entry, MapAttributes, Space, Table};
+// pub mod fixed;
+// pub use fixed::FixedTableGroup;
+use crate::arch::platform::table as plat_table;
 
 /// Structure representing a region of memory used for MMIO/device access
 // TODO: this shouldn't be trivially-cloneable and should instead incorporate
@@ -45,15 +45,16 @@ impl DeviceMemory {
     ///
     /// See [FixedTableGroup::map_region]
     pub fn map(name: &'static str, phys: usize, count: usize) -> Result<Self, Errno> {
-        let base = unsafe { KERNEL_TTBR1.map_region(phys, count) }?;
-        debugln!(
-            "Mapping {:#x}..{:#x} -> {:#x} for {:?}",
-            base,
-            base + count * 0x1000,
-            phys,
-            name
-        );
-        Ok(Self { name, base, count })
+        todo!();
+        // let base = unsafe { KERNEL_TTBR1.map_region(phys, count) }?;
+        // debugln!(
+        //     "Mapping {:#x}..{:#x} -> {:#x} for {:?}",
+        //     base,
+        //     base + count * 0x1000,
+        //     phys,
+        //     name
+        // );
+        // Ok(Self { name, base, count })
     }
 }
 
@@ -90,16 +91,18 @@ impl<T> Deref for DeviceMemoryIo<T> {
 /// Sets up device mapping tables and disable lower-half
 /// identity-mapped translation
 pub fn enable() -> Result<(), Errno> {
-    unsafe {
-        KERNEL_TTBR1.init_device_map();
-
-        dsb(barrier::ISH);
-        isb(barrier::SY);
-    }
-
-    // Disable lower-half translation
-    TTBR0_EL1.set(0);
-    //TCR_EL1.modify(TCR_EL1::EPD0::SET);
-
-    Ok(())
+    plat_table::enable()
 }
+//     unsafe {
+//         KERNEL_TTBR1.init_device_map();
+//
+//         dsb(barrier::ISH);
+//         isb(barrier::SY);
+//     }
+//
+//     // Disable lower-half translation
+//     TTBR0_EL1.set(0);
+//     //TCR_EL1.modify(TCR_EL1::EPD0::SET);
+//
+//     Ok(())
+// }

kernel/src/mem/virt/table.rs

@@ -7,329 +7,82 @@ use crate::mem::{
 use core::ops::{Index, IndexMut};
 use libsys::{error::Errno, mem::memset};
 
-/// Transparent wrapper structure representing a single
-/// translation table entry
-#[derive(Clone, Copy)]
-#[repr(transparent)]
-pub struct Entry(u64);
+use crate::arch::platform::table;
 
-/// Structure describing a single level of translation mappings
-#[repr(C, align(0x1000))]
-pub struct Table {
-    entries: [Entry; 512],
-}
-
-/// Wrapper for top-most level of address translation tables
-#[repr(transparent)]
-pub struct Space(Table);
+pub use table::{Space, Table};
 
 bitflags! {
     /// Attributes attached to each translation [Entry]
     pub struct MapAttributes: u64 {
-        // TODO use 2 lower bits to determine mapping size?
-        /// nG bit -- determines whether a TLB entry associated with this mapping
-        ///           applies only to current ASID or all ASIDs.
-        const NOT_GLOBAL = 1 << 11;
-        /// AF bit -- must be set by software, otherwise Access Error exception is
-        ///           generated when the page is accessed
-        const ACCESS = 1 << 10;
-        /// The memory region is outer-shareable
-        const SH_OUTER = 2 << 8;
-        /// This page is used for device-MMIO mapping and uses MAIR attribute #1
-        const DEVICE = 1 << 2;
+        const USER_READ = 1 << 1;
+        const USER_WRITE = 1 << 2;
+        const USER_EXEC = 1 << 3;
+        const KERNEL_WRITE = 1 << 4;
+        const KERNEL_EXEC = 1 << 5;
 
-        /// Pages marked with this bit are Copy-on-Write
-        const EX_COW = 1 << 55;
+        const SHARE_OUTER = 1 << 6;
+        const SHARE_INNER = 2 << 6;
 
-        /// UXN bit -- if set, page may not be used for instruction fetching from EL0
-        const UXN = 1 << 54;
-        /// PXN bit -- if set, page may not be used for instruction fetching from EL1
-        const PXN = 1 << 53;
-
-        // AP field
-        // Default behavior is: read-write for EL1, no access for EL0
-        /// If set, the page referred to by this entry is read-only for both EL0/EL1
-        const AP_BOTH_READONLY = 3 << 6;
-        /// If set, the page referred to by this entry is read-write for both EL0/EL1
-        const AP_BOTH_READWRITE = 1 << 6;
+        const NOT_GLOBAL = 1 << 8;
     }
 }
 
-impl Table {
-    /// Returns next-level translation table reference for `index`, if one is present.
-    /// If `index` represents a `Block`-type mapping, will return an error.
-    /// If `index` does not map to any translation table, will try to allocate, init and
-    /// map a new one, returning it after doing so.
-    pub fn next_level_table_or_alloc(&mut self, index: usize) -> Result<&'static mut Table, Errno> {
-        let entry = self[index];
-        if entry.is_present() {
-            if !entry.is_table() {
-                return Err(Errno::InvalidArgument);
-            }
-
-            Ok(unsafe { &mut *(mem::virtualize(entry.address_unchecked()) as *mut _) })
-        } else {
-            let phys = phys::alloc_page(PageUsage::Paging)?;
-            let res = unsafe { &mut *(mem::virtualize(phys) as *mut Self) };
-            self[index] = Entry::table(phys, MapAttributes::empty());
-            res.entries.fill(Entry::invalid());
-            Ok(res)
-        }
-    }
-
-    /// Returns next-level translation table reference for `index`, if one is present.
-    /// Same as [next_level_table_or_alloc], but returns `None` if no table is mapped.
-    pub fn next_level_table(&mut self, index: usize) -> Option<&'static mut Table> {
-        let entry = self[index];
-        if entry.is_present() {
-            if !entry.is_table() {
-                panic!("Entry is not a table: idx={}", index);
-            }
-
-            Some(unsafe { &mut *(mem::virtualize(entry.address_unchecked()) as *mut _) })
-        } else {
-            None
-        }
-    }
-
-    /// Constructs and fills a [Table] with non-present mappings
-    pub const fn empty() -> Table {
-        Table {
-            entries: [Entry::invalid(); 512],
-        }
-    }
+pub trait Entry: Clone + Copy {
+    fn from_parts(phys: usize, attrs: MapAttributes) -> Self;
+    fn target(self) -> usize;
+    fn is_present(self) -> bool;
+    fn is_table(self) -> bool;
 }
 
-impl Index<usize> for Table {
-    type Output = Entry;
+pub trait AddressSpace {
+    type Entry: Entry;
 
-    fn index(&self, index: usize) -> &Self::Output {
-        &self.entries[index]
-    }
-}
+    fn alloc_empty() -> Result<&'static mut Self, Errno>;
+    fn fork(&mut self) -> Result<&'static mut Self, Errno>;
+    fn release(space: &mut Self);
+    fn address_phys(&mut self) -> usize;
+    fn read_last_level_entry(&mut self, virt: usize) -> Result<Self::Entry, Errno>;
+    fn write_last_level_entry(
+        &mut self,
+        virt: usize,
+        entry: Self::Entry,
+        map_intermediate: bool,
+    ) -> Result<(), Errno>;
 
-impl IndexMut<usize> for Table {
-    fn index_mut(&mut self, index: usize) -> &mut Self::Output {
-        &mut self.entries[index]
-    }
-}
-
-impl Entry {
-    const PRESENT: u64 = 1 << 0;
-    const TABLE: u64 = 1 << 1;
-    const PHYS_MASK: u64 = 0x0000FFFFFFFFF000;
-
-    /// Constructs a single non-present mapping
-    pub const fn invalid() -> Self {
-        Self(0)
+    #[inline(always)]
+    fn map(&mut self, virt: usize, phys: usize, attrs: MapAttributes) -> Result<(), Errno> {
+        let entry = Entry::from_parts(phys, attrs);
+        self.write_last_level_entry(virt, entry, true).map(|_| ())
     }
 
-    /// Constructs a `Block`-type memory mapping
-    pub const fn block(phys: usize, attrs: MapAttributes) -> Self {
-        Self((phys as u64 & Self::PHYS_MASK) | attrs.bits() | Self::PRESENT)
+    #[inline(always)]
+    fn translate(&mut self, virt: usize) -> Result<usize, Errno> {
+        self.read_last_level_entry(virt).map(Entry::target)
     }
 
-    /// Constructs a `Table` or `Page`-type mapping depending on translation level
-    /// this entry is used at
-    pub const fn table(phys: usize, attrs: MapAttributes) -> Self {
-        Self((phys as u64 & Self::PHYS_MASK) | attrs.bits() | Self::PRESENT | Self::TABLE)
-    }
-
-    /// Returns `true` if this entry is not invalid
-    pub const fn is_present(self) -> bool {
-        self.0 & Self::PRESENT != 0
-    }
-
-    /// Returns `true` if this entry is a `Table` or `Page`-type mapping
-    pub const fn is_table(self) -> bool {
-        self.0 & Self::TABLE != 0
-    }
-
-    /// Returns the target address of this translation entry.
-    ///
-    /// # Safety
-    ///
-    /// Does not check if the entry is actually valid.
-    pub const unsafe fn address_unchecked(self) -> usize {
-        (self.0 & Self::PHYS_MASK) as usize
-    }
-
-    unsafe fn set_address(&mut self, address: usize) {
-        self.0 &= !Self::PHYS_MASK;
-        self.0 |= (address as u64) & Self::PHYS_MASK;
-    }
-
-    unsafe fn fork_flags(self) -> MapAttributes {
-        MapAttributes::from_bits_unchecked(self.0 & !Self::PHYS_MASK)
-    }
-
-    fn set_cow(&mut self) {
-        self.0 |= (MapAttributes::AP_BOTH_READONLY | MapAttributes::EX_COW).bits();
-    }
-
-    fn clear_cow(&mut self) {
-        self.0 &= !(MapAttributes::AP_BOTH_READONLY | MapAttributes::EX_COW).bits();
-        self.0 |= MapAttributes::AP_BOTH_READWRITE.bits();
-    }
-
-    #[inline]
-    fn is_cow(self) -> bool {
-        let attrs = (MapAttributes::AP_BOTH_READONLY | MapAttributes::EX_COW).bits();
-        self.0 & attrs == attrs
-    }
-}
-
-impl Space {
-    /// Creates a new virtual address space and fills it with [Entry::invalid()]
-    /// mappings. Does physical memory page allocation.
-    pub fn alloc_empty() -> Result<&'static mut Self, Errno> {
-        let phys = phys::alloc_page(PageUsage::Paging)?;
-        let res = unsafe { &mut *(mem::virtualize(phys) as *mut Self) };
-        res.0.entries.fill(Entry::invalid());
-        Ok(res)
-    }
-
-    /// Inserts a single `virt` -> `phys` translation entry to this address space.
-    ///
-    /// TODO: only works with 4K-sized pages at this moment.
-    pub fn map(&mut self, virt: usize, phys: usize, flags: MapAttributes) -> Result<(), Errno> {
-        let l0i = virt >> 30;
-        let l1i = (virt >> 21) & 0x1FF;
-        let l2i = (virt >> 12) & 0x1FF;
-
-        let l1_table = self.0.next_level_table_or_alloc(l0i)?;
-        let l2_table = l1_table.next_level_table_or_alloc(l1i)?;
-
-        if l2_table[l2i].is_present() {
-            Err(Errno::AlreadyExists)
-        } else {
-            l2_table[l2i] = Entry::table(phys, flags | MapAttributes::ACCESS);
-            #[cfg(feature = "verbose")]
-            debugln!("{:#p} Map {:#x} -> {:#x}, {:?}", self, virt, phys, flags);
-            Ok(())
-        }
-    }
-
-    /// 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> {
-        let virt = virt & !0xFFF;
-        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() {
-            warnln!("Entry is not present: {:#x}", virt);
-            return Err(Errno::DoesNotExist);
-        }
-
-        let src_phys = unsafe { entry.address_unchecked() };
-        if !entry.is_cow() {
-            warnln!(
-                "Entry is not marked as CoW: {:#x}, points to {:#x}",
-                virt,
-                src_phys
-            );
-            return Err(Errno::DoesNotExist);
-        }
-
-        let dst_phys = unsafe { phys::copy_cow_page(src_phys)? };
-        unsafe {
-            l2_table[l2i].set_address(dst_phys);
-        }
-        l2_table[l2i].clear_cow();
-
-        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()?;
-        for l0i in 0..512 {
-            if let Some(l1_table) = self.0.next_level_table(l0i) {
-                for l1i in 0..512 {
-                    if let Some(l2_table) = l1_table.next_level_table(l1i) {
-                        for l2i in 0..512 {
-                            let entry = l2_table[l2i];
-
-                            if !entry.is_present() {
-                                continue;
-                            }
-
-                            assert!(entry.is_table());
-                            let src_phys = unsafe { entry.address_unchecked() };
-                            let virt_addr = (l0i << 30) | (l1i << 21) | (l2i << 12);
-                            let dst_phys = unsafe { phys::fork_page(src_phys)? };
-
-                            let mut flags = unsafe { entry.fork_flags() };
-                            if dst_phys != src_phys {
-                                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);
-                                }
-                                res.map(virt_addr, dst_phys, flags)?;
-                            }
-                        }
-                    }
+    /// Allocates a contiguous region from the address space and maps
+    /// physical pages to it
+    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;
                 }
             }
-        }
-        Ok(res)
-    }
 
-    /// Releases all the mappings from the address space. Frees all
-    /// memory pages referenced by this space as well as those used for
-    /// its paging tables.
-    ///
-    /// # Safety
-    ///
-    /// Unsafe: may invalidate currently active address space
-    pub unsafe fn release(space: &mut Self) {
-        for l0i in 0..512 {
-            let l0_entry = space.0[l0i];
-            if !l0_entry.is_present() {
-                continue;
+            for i in 0..len {
+                let phys = phys::alloc_page(usage).unwrap();
+                self.map(page + i * 0x1000, phys, flags).unwrap();
             }
-
-            assert!(l0_entry.is_table());
-            let l1_table = &mut *(mem::virtualize(l0_entry.address_unchecked()) as *mut Table);
-
-            for l1i in 0..512 {
-                let l1_entry = l1_table[l1i];
-                if !l1_entry.is_present() {
-                    continue;
-                }
-                assert!(l1_entry.is_table());
-                let l2_table = &mut *(mem::virtualize(l1_entry.address_unchecked()) as *mut Table);
-
-                for l2i in 0..512 {
-                    let entry = l2_table[l2i];
-                    if !entry.is_present() {
-                        continue;
-                    }
-
-                    assert!(entry.is_table());
-                    phys::free_page(entry.address_unchecked()).unwrap();
-                }
-                phys::free_page(l1_entry.address_unchecked()).unwrap();
-            }
-            phys::free_page(l0_entry.address_unchecked()).unwrap();
+            return Ok(page);
         }
-        memset(space as *mut Space as *mut u8, 0, 4096);
-    }
-
-    /// Returns the physical address of this structure
-    pub fn address_phys(&mut self) -> usize {
-        (self as *mut _ as usize) - mem::KERNEL_OFFSET
+        Err(Errno::OutOfMemory)
     }
 }

kernel/src/proc/elf.rs

@@ -2,7 +2,7 @@
 use crate::mem::{
     self,
     phys::{self, PageUsage},
-    virt::{MapAttributes, Space},
+    virt::{Space, table::{AddressSpace, MapAttributes}},
 };
 use core::mem::{size_of, MaybeUninit};
 use libsys::{
@@ -66,10 +66,10 @@ struct Phdr<E: Elf> {
 }
 
 fn map_flags(elf_flags: usize) -> MapAttributes {
-    let mut dst_flags = MapAttributes::NOT_GLOBAL | MapAttributes::SH_OUTER;
+    let mut dst_flags = MapAttributes::SHARE_OUTER | MapAttributes::NOT_GLOBAL;
 
-    if elf_flags & (1 << 0) /* PF_X */ == 0 {
-        dst_flags |= MapAttributes::UXN | MapAttributes::PXN;
+    if elf_flags & (1 << 0) /* PF_X */ != 0 {
+        dst_flags |= MapAttributes::USER_EXEC | MapAttributes::KERNEL_EXEC;
     }
 
     match (elf_flags & (3 << 1)) >> 1 {
@@ -78,9 +78,9 @@ fn map_flags(elf_flags: usize) -> MapAttributes {
         // Write-only: not sure if such mapping should exist at all
         1 => todo!(),
         // Read-only
-        2 => dst_flags |= MapAttributes::AP_BOTH_READONLY,
+        2 => dst_flags |= MapAttributes::USER_READ,
         // Read+Write
-        3 => dst_flags |= MapAttributes::AP_BOTH_READWRITE,
+        3 => dst_flags |= MapAttributes::USER_READ | MapAttributes::USER_WRITE | MapAttributes::KERNEL_WRITE,
         _ => unreachable!(),
     };
 

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

@@ -3,11 +3,14 @@
 use crate::init;
 use crate::sync::IrqSafeSpinLock;
 use alloc::collections::BTreeMap;
-use libsys::proc::Pid;
+use libsys::proc::{Tid, 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;
 
@@ -17,25 +20,6 @@ pub mod sched;
 pub use sched::Scheduler;
 pub(self) use sched::SCHED;
 
-// 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;
-//             {
-//                 #![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))
-// }
-
 /// Performs a task switch.
 ///
 /// See [Scheduler::switch]
@@ -43,15 +27,17 @@ pub fn switch() {
     SCHED.switch(false);
 }
 
-///
-pub fn process(id: Pid) -> ProcessRef {
-    PROCESSES.lock().get(&id).unwrap().clone()
+#[no_mangle]
+extern "C" fn sched_yield() {
+    SCHED.switch(false);
 }
 
-/// Global list of all processes in the system
 pub(self) static PROCESSES: IrqSafeSpinLock<BTreeMap<Pid, ProcessRef>> =
     IrqSafeSpinLock::new(BTreeMap::new());
 
+pub(self) static THREADS: IrqSafeSpinLock<BTreeMap<Tid, ThreadRef>> =
+    IrqSafeSpinLock::new(BTreeMap::new());
+
 /// Sets up initial process and enters it.
 ///
 /// See [Scheduler::enter]
@@ -61,6 +47,6 @@ pub(self) static PROCESSES: IrqSafeSpinLock<BTreeMap<Pid, ProcessRef>> =
 /// Unsafe: May only be called once.
 pub unsafe fn enter() -> ! {
     SCHED.init();
-    SCHED.enqueue(Process::new_kernel(init::init_fn, 0).unwrap().id());
+    Process::new_kernel(init::init_fn, 0).unwrap().enqueue();
     SCHED.enter();
 }

kernel/src/proc/process.rs

@@ -1,65 +1,145 @@
 //! Process data and control
-use crate::arch::aarch64::exception::ExceptionFrame;
+use crate::arch::platform::ForkFrame;
 use crate::mem::{
     self,
     phys::{self, PageUsage},
-    virt::{MapAttributes, Space},
+    virt::{table::{MapAttributes, AddressSpace}, Space},
 };
-use crate::proc::{wait::Wait, ProcessIo, PROCESSES, SCHED};
-use crate::sync::IrqSafeSpinLock;
-use alloc::rc::Rc;
-use core::cell::UnsafeCell;
+use crate::proc::{
+    wait::Wait, Context, ProcessIo, Thread, ThreadRef, ThreadState, PROCESSES, SCHED, Tid,
+};
+use crate::sync::{IrqSafeSpinLock, IrqSafeSpinLockGuard};
+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,
+};
+use core::arch::asm;
 
 /// 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<Tid>,
 }
 
 /// 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>,
 }
 
 impl Process {
     const USTACK_VIRT_TOP: usize = 0x100000000;
-    const USTACK_PAGES: usize = 4;
+    const USTACK_PAGES: usize = 8;
 
-    /// 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
@@ -67,273 +147,214 @@ impl Process {
         PROCESSES.lock().get(&pid).cloned()
     }
 
-    /// Sets a pending signal for a process
-    pub fn set_signal(&self, signal: Signal) {
-        let lock = self.inner.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!()
+    fn find1(a: u32) -> Option<usize> {
+        for i in 0..32 {
+            if a & (1 << i) != 0 {
+                return Some(i);
             }
         }
+        None
     }
 
-    /// 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);
-        }
+    fn space_phys(lock: &mut IrqSafeSpinLockGuard<ProcessInner>) -> usize {
+        lock.space.as_mut().unwrap().address_phys() | ((lock.id.asid() as usize) << 48)
     }
 
-    /// 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)");
-        }
+    // /// Handles all pending signals (when returning from aborted syscall)
+    // pub fn handle_pending_signals(&self) {
+    //     let mut lock = self.inner.lock();
+    //     let table = Self::space_phys(&lock);
+    //     let main_thread = Thread::get(lock.threads[0]).unwrap();
+    //     drop(lock);
 
-        let mut lock = self.inner.lock();
-        let signal_ctx = unsafe { &mut *self.signal_ctx.get() };
+    //     loop {
+    //         let state = self.signal_state.load(Ordering::Acquire);
+    //         if let Some(signal) = Self::find1(state).map(|e| Signal::try_from(e as u32).unwrap()) {
+    //             self.signal_state.fetch_and(!(1 << (signal as u32)), Ordering::Release);
+    //             main_thread.clone().enter_signal(signal, table);
+    //         } else {
+    //             break;
+    //         }
+    //     }
+    // }
 
-        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);
+    // /// Sets a pending signal for a process
+    // pub fn set_signal(&self, signal: Signal) {
+    //     let mut lock = self.inner.lock();
+    //     let table = Self::space_phys(&lock);
+    //     let main_thread = Thread::get(lock.threads[0]).unwrap();
+    //     drop(lock);
 
-        debugln!(
-            "Signal entry: pc={:#x}, sp={:#x}, ttbr0={:#x}",
-            lock.signal_entry,
-            lock.signal_stack,
-            dst_ttbr0
-        );
-        assert_eq!(lock.state, State::Running);
+    //     // TODO check that `signal` is not a fault signal
+    //     //      it is illegal to call this function with
+    //     //      fault signals
 
-        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);
+    //     match main_thread.state() {
+    //         ThreadState::Running => {
+    //             main_thread.enter_signal(signal, table);
+    //         }
+    //         ThreadState::Waiting => {
+    //             self.signal_state.fetch_or(1 << (signal as u32), Ordering::Release);
+    //             main_thread.interrupt_wait(true);
+    //         }
+    //         ThreadState::Ready => {
+    //             main_thread.clone().setup_signal(signal, table);
+    //             main_thread.interrupt_wait(false);
+    //         }
+    //         ThreadState::Finished => {
+    //             // TODO report error back
+    //             todo!()
+    //         }
+    //     }
+    // }
 
-        unsafe {
-            (&mut *src_ctx).switch(signal_ctx);
-        }
-    }
+    // /// Immediately delivers a signal to requested thread
+    // pub fn enter_fault_signal(&self, thread: ThreadRef, signal: Signal) {
+    //     let mut lock = self.inner.lock();
+    //     let table = Self::space_phys(&lock);
+    //     drop(lock);
+    //     thread.enter_signal(signal, table);
+    // }
 
-    /// 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;
-    }
+    // /// Crates a new thread in the process
+    // pub fn new_user_thread(&self, entry: usize, stack: usize, arg: usize) -> Result<Tid, Errno> {
+    //     let mut lock = self.inner.lock();
 
-    /// 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()
-    }
+    //     let table = Self::space_phys(&lock);
+    //     let thread = Thread::new_user(lock.id, entry, stack, arg, table)?;
+    //     let tid = thread.id();
+    //     lock.threads.push(tid);
+    //     SCHED.enqueue(tid);
 
-    /// 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)
-    }
+    //     Ok(tid)
+    // }
 
     /// Creates a "fork" of the process, cloning its address space and
     /// resources
-    pub fn fork(&self, frame: &mut ExceptionFrame) -> Result<Pid, Errno> {
+    pub fn fork(&self, frame: &mut ForkFrame) -> Result<Pid, Errno> {
+        todo!();
         let src_io = self.io.lock();
         let mut src_inner = self.inner.lock();
 
         let dst_id = new_user_pid();
         let dst_space = src_inner.space.as_mut().unwrap().fork()?;
-        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 dst = Rc::new(Self {
-            ctx: UnsafeCell::new(Context::fork(frame, dst_ttbr0)),
-            signal_ctx: UnsafeCell::new(Context::empty()),
-            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,
-                exit: None,
-                space: Some(dst_space),
-                state: State::Ready,
-                wait_flag: false,
-            }),
-        });
-        debugln!("Process {:?} forked into {:?}", src_inner.id, dst_id);
-        assert!(PROCESSES.lock().insert(dst_id, dst).is_none());
-        SCHED.enqueue(dst_id);
+        todo!()
 
-        Ok(dst_id)
+        // 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 {
+        //     exit_wait: Wait::new("process_exit"),
+        //     io: IrqSafeSpinLock::new(src_io.fork()?),
+        //     signal_state: AtomicU32::new(0),
+        //     inner: IrqSafeSpinLock::new(ProcessInner {
+        //         threads,
+        //         exit: None,
+        //         space: Some(dst_space),
+        //         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(tid);
+
+        // 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) {
+        todo!()
+        // 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;
 
-            self.io.lock().handle_exit();
+        // for &tid in lock.threads.iter() {
+        //     let thread = Thread::get(tid).unwrap();
+        //     if thread.state() == ThreadState::Waiting {
+        //         todo!()
+        //     }
+        //     thread.terminate(status);
+        //     SCHED.dequeue(tid);
+        // }
 
-            SCHED.dequeue(lock.id);
-            drop
-        };
-        self.exit_wait.wakeup_all();
-        if drop {
-            SCHED.switch(true);
-            panic!("This code should never run");
-        }
+        // if let Some(space) = lock.space.take() {
+        //     unsafe {
+        //         Space::release(space);
+        //         Process::invalidate_asid((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) {
+        todo!()
+
+        // 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;
+        //     }
+
+        //     lock.threads.retain(|&e| e != tid);
+
+        //     thread.terminate(status);
+        //     SCHED.dequeue(tid);
+        //     debugln!("Thread {:?} terminated", tid);
+
+        //     switch
+        // };
+
+        // if switch {
+        //     // TODO retain thread ID in process "finished" list and
+        //     //      drop it when process finishes
+        //     SCHED.switch(true);
+        //     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 +373,6 @@ 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);
             }
 
@@ -360,43 +380,144 @@ impl Process {
         }
     }
 
+    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 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::SHARE_OUTER
+                | MapAttributes::NOT_GLOBAL
+                | MapAttributes::USER_READ;
+            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::SHARE_OUTER
+                | MapAttributes::NOT_GLOBAL
+                | MapAttributes::USER_READ;
+            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, 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)
+    }
+
+    pub fn asid(&self) -> usize {
+        (self.id().asid() as usize) << 48
+    }
+
+    // pub fn invalidate_tlb(&self) {
+    //     Process::invalidate_asid(self.asid());
+    // }
+
+    // #[inline]
+    // pub fn invalidate_asid(asid: usize) {
+    //     unsafe {
+    //         asm!("tlbi aside1, {}", in(reg) asid);
+    //     }
+    // }
+
     /// Loads a new program into current process address space
     pub fn execve<F: FnOnce(&mut Space) -> Result<usize, Errno>>(
         loader: F,
-        arg: usize,
+        argv: &[&str],
     ) -> Result<(), Errno> {
         unsafe {
             // Run with interrupts disabled
-            asm!("msr daifset, #2");
+            // asm!("msr daifset, #2");
+            asm!("cli");
         }
 
-        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);
-            }
-        } 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!();
         }
 
+        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();
 
         let new_space = Space::alloc_empty()?;
@@ -405,36 +526,36 @@ impl Process {
         let ustack_virt_bottom = Self::USTACK_VIRT_TOP - Self::USTACK_PAGES * mem::PAGE_SIZE;
         for i in 0..Self::USTACK_PAGES {
             let page = phys::alloc_page(PageUsage::UserPrivate).unwrap();
-            let flags = MapAttributes::SH_OUTER
+            let flags = MapAttributes::SHARE_OUTER
                 | MapAttributes::NOT_GLOBAL
-                | MapAttributes::UXN
-                | MapAttributes::PXN
-                | MapAttributes::AP_BOTH_READWRITE;
+                | MapAttributes::USER_WRITE
+                | MapAttributes::USER_READ
+                | MapAttributes::KERNEL_WRITE;
             new_space
                 .map(ustack_virt_bottom + i * mem::PAGE_SIZE, page, flags)
                 .unwrap();
         }
 
         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;
+            // Process::invalidate_asid(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 +1,15 @@
 //!
-use crate::proc::{Pid, Process, ProcessRef, PROCESSES};
+use crate::proc::{Thread, ThreadRef, THREADS};
 use crate::sync::IrqSafeSpinLock;
 use crate::util::InitOnce;
+use libsys::proc::Tid;
 use alloc::{collections::VecDeque, rc::Rc};
+use core::arch::asm;
 
 struct SchedulerInner {
-    queue: VecDeque<Pid>,
-    idle: Option<Pid>,
-    current: Option<Pid>,
+    queue: VecDeque<Tid>,
+    idle: Option<Tid>,
+    current: Option<Tid>,
 }
 
 /// Process scheduler state and queues
@@ -23,7 +25,9 @@ impl SchedulerInner {
             current: None,
         };
 
-        this.idle = Some(Process::new_kernel(idle_fn, 0).unwrap().id());
+        let idle = Thread::new_kernel(None, idle_fn, 0).unwrap().id();
+        assert_eq!(idle, Tid::IDLE);
+        this.idle = Some(idle);
 
         this
     }
@@ -39,13 +43,13 @@ 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: Tid) {
+        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: Tid) {
+        self.inner.get().lock().queue.retain(|&p| p != tid)
     }
 
     /// Performs initial process entry.
@@ -63,11 +67,12 @@ 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)
+        asm!("cli");
+        // asm!("msr daifset, #2");
+        Thread::enter(thread)
     }
 
     /// This hack is required to be called from execve() when downgrading current
@@ -76,8 +81,14 @@ 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: Tid, new: Tid) {
+        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
@@ -87,7 +98,7 @@ impl Scheduler {
             let mut inner = self.inner.get().lock();
             let current = inner.current.unwrap();
 
-            if !discard && current != Pid::IDLE {
+            if !discard && current != Tid::IDLE {
                 // Put the process into the back of the queue
                 inner.queue.push_back(current);
             }
@@ -100,7 +111,7 @@ 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(),
@@ -112,18 +123,26 @@ impl Scheduler {
 
         if !Rc::ptr_eq(&from, &to) {
             unsafe {
-                asm!("msr daifset, #2");
-                Process::switch(from, to, discard);
+                asm!("cli");
+                // asm!("msr daifset, #2");
+                Thread::switch(from, to, discard);
             }
         }
     }
 
-    /// 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
@@ -134,7 +153,7 @@ pub fn is_ready() -> bool {
 #[inline(never)]
 extern "C" fn idle_fn(_a: usize) -> ! {
     loop {
-        cortex_a::asm::wfi();
+        // cortex_a::asm::wfi();
     }
 }
 

kernel/src/proc/thread.rs

@@ -0,0 +1,411 @@
+//! 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, Tid},
+    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: Tid,
+    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,
+}
+
+impl Thread {
+    /// Returns currently active thread [Rc]-reference
+    #[inline]
+    pub fn current() -> ThreadRef {
+        SCHED.current_thread()
+    }
+
+    /// Returns `true` if the thread is currently executing a signal handler context
+    pub fn is_handling_signal(&self) -> bool {
+        self.signal_pending.load(Ordering::Acquire) != 0
+    }
+
+    /// Returns a reference to thread `tid`, if it exists
+    #[inline]
+    pub fn get(tid: Tid) -> 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 {
+            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 {
+            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 {
+    //         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) -> Tid {
+        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(thread: ThreadRef) -> ! {
+        // FIXME use some global lock to guarantee atomicity of thread entry?
+        thread.inner.lock().state = State::Running;
+        thread.current_context().enter()
+    }
+
+    /// 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: 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;
+        }
+
+        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);
+            drop
+        };
+        if drop {
+            SCHED.switch(true);
+        }
+    }
+
+    /// Changes process wait condition status
+    pub fn setup_wait(&self, wait: *const Wait) {
+        #![allow(clippy::not_unsafe_ptr_arg_deref)]
+        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: Tid) -> 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() -> Tid {
+    static LAST: AtomicU32 = AtomicU32::new(0);
+    let id = LAST.fetch_add(1, Ordering::Relaxed);
+    Tid::from(id)
+}

kernel/src/proc/wait.rs

@@ -2,141 +2,201 @@
 
 use crate::arch::machine;
 use crate::dev::timer::TimestampSource;
-use crate::proc::{self, sched::SCHED, Process, ProcessRef};
+use crate::proc::{sched::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, proc::Tid, 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<Tid>>,
+    #[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: Tid,
     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() {
-    let time = machine::local_timer().timestamp().unwrap();
-    let mut list = TICK_LIST.lock();
-    let mut cursor = list.cursor_front_mut();
+    todo!();
+    // let time = machine::local_timer().timestamp().unwrap();
+    // let mut list = TICK_LIST.lock();
+    // let mut cursor = list.cursor_front_mut();
 
-    while let Some(item) = cursor.current() {
-        if time > item.deadline {
-            let pid = item.pid;
-            cursor.remove_current();
-            SCHED.enqueue(pid);
-        } else {
-            cursor.move_next();
-        }
-    }
+    // while let Some(item) = cursor.current() {
+    //     if time > item.deadline {
+    //         let tid = item.tid;
+    //         cursor.remove_current();
+    //         SCHED.enqueue(tid);
+    //     } else {
+    //         cursor.move_next();
+    //     }
+    // }
 }
 
 /// 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();
-    let deadline = machine::local_timer().timestamp()? + timeout;
-    match SLEEP_NOTIFY.wait(Some(deadline)) {
-        Err(Errno::Interrupt) => {
-            *remaining = deadline - machine::local_timer().timestamp()?;
-            Err(Errno::Interrupt)
-        }
-        Err(Errno::TimedOut) => Ok(()),
-        Ok(_) => panic!("Impossible result"),
-        res => res,
-    }
+    todo!()
+    // // Dummy wait descriptor which will never receive notifications
+    // static SLEEP_NOTIFY: Wait = Wait::new("sleep");
+    // let deadline = machine::local_timer().timestamp()? + timeout;
+    // match SLEEP_NOTIFY.wait(Some(deadline)) {
+    //     Err(Errno::Interrupt) => {
+    //         *remaining = deadline - machine::local_timer().timestamp()?;
+    //         Err(Errno::Interrupt)
+    //     }
+    //     Err(Errno::TimedOut) => Ok(()),
+    //     Ok(_) => panic!("Impossible result"),
+    //     res => res,
+    // }
 }
 
+/// 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() {
-        todo!();
-    }
-    let read = rfds.as_deref().map(FdSet::clone);
-    let write = wfds.as_deref().map(FdSet::clone);
-    rfds.as_deref_mut().map(FdSet::reset);
-    wfds.as_deref_mut().map(FdSet::reset);
+    todo!();
 
-    let deadline = timeout.map(|v| v + machine::local_timer().timestamp().unwrap());
-    let mut io = proc.io.lock();
+    // if wfds.is_none() && rfds.is_none() {
+    //     todo!();
+    // }
+    // let read = rfds.as_deref().map(FdSet::clone);
+    // let write = wfds.as_deref().map(FdSet::clone);
+    // if let Some(rfds) = &mut rfds {
+    //     rfds.reset();
+    // }
+    // if let Some(wfds) = &mut wfds {
+    //     wfds.reset();
+    // }
 
-    loop {
-        if let Some(read) = &read {
-            for fd in read.iter() {
-                let file = io.file(fd)?;
-                if file.borrow().is_ready(false)? {
-                    rfds.as_mut().unwrap().set(fd);
-                    return Ok(1);
-                }
-            }
-        }
-        if let Some(write) = &write {
-            for fd in write.iter() {
-                let file = io.file(fd)?;
-                if file.borrow().is_ready(true)? {
-                    wfds.as_mut().unwrap().set(fd);
-                    return Ok(1);
-                }
-            }
-        }
+    // let deadline = timeout.map(|v| v + machine::local_timer().timestamp().unwrap());
+    // let proc = thread.owner().unwrap();
+    // let mut io = proc.io.lock();
 
-        // Suspend
-        match WAIT_SELECT.wait(deadline) {
-            Err(Errno::TimedOut) => return Ok(0),
-            Err(e) => return Err(e),
-            Ok(_) => {}
-        }
-    }
+    // loop {
+    //     if let Some(read) = &read {
+    //         for fd in read.iter() {
+    //             let file = io.file(fd)?;
+    //             if file.borrow().is_ready(false)? {
+    //                 rfds.as_mut().unwrap().set(fd);
+    //                 return Ok(1);
+    //             }
+    //         }
+    //     }
+    //     if let Some(write) = &write {
+    //         for fd in write.iter() {
+    //             let file = io.file(fd)?;
+    //             if file.borrow().is_ready(true)? {
+    //                 wfds.as_mut().unwrap().set(fd);
+    //                 return Ok(1);
+    //             }
+    //         }
+    //     }
+
+    //     // Suspend
+    //     match WAIT_SELECT.wait(deadline) {
+    //         Err(Errno::TimedOut) => return Ok(0),
+    //         Err(e) => return Err(e),
+    //         Ok(_) => {}
+    //     }
+    // }
 }
 
 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: Tid, enqueue: bool) {
+        todo!();
+
+        // 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);
+        //         }
+        //         break;
+        //     } else {
+        //         cursor.move_next();
+        //     }
+        // }
+    }
+
     fn wakeup_some(&self, mut limit: usize) -> usize {
-        // No IRQs will arrive now == safe to manipulate tick list
-        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 mut tick_lock = TICK_LIST.lock();
-                let mut cursor = tick_lock.cursor_front_mut();
-                while let Some(item) = cursor.current() {
-                    if pid == item.pid {
-                        cursor.remove_current();
-                        break;
-                    } else {
-                        cursor.move_next();
-                    }
-                }
-                drop(tick_lock);
+        todo!();
 
-                proc::process(pid).set_wait_flag(false);
-                SCHED.enqueue(pid);
-            }
+        // // No IRQs will arrive now == safe to manipulate tick list
+        // let mut queue = self.queue.lock();
+        // let mut count = 0;
+        // while limit != 0 && !queue.is_empty() {
+        //     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 tid == item.tid {
+        //                 cursor.remove_current();
+        //                 break;
+        //             } else {
+        //                 cursor.move_next();
+        //             }
+        //         }
+        //         drop(tick_lock);
 
-            limit -= 1;
-            count += 1;
-        }
-        count
+        //         Thread::get(tid).unwrap().set_wait_status(WaitStatus::Done);
+        //         SCHED.enqueue(tid);
+        //     }
+
+        //     limit -= 1;
+        //     count += 1;
+        // }
+        // count
     }
 
     /// Notifies all processes waiting for this event
@@ -152,44 +212,53 @@ 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 deadline = timeout.map(|t| machine::local_timer().timestamp().unwrap() + t);
-        let mut queue_lock = self.queue.lock();
+        todo!();
 
-        queue_lock.push_back(proc.id());
-        proc.set_wait_flag(true);
-        if let Some(deadline) = deadline {
-            TICK_LIST.lock().push_back(Timeout {
-                pid: proc.id(),
-                deadline,
-            });
-        }
+        // let thread = Thread::current();
+        // //let deadline = timeout.map(|t| machine::local_timer().timestamp().unwrap() + t);
+        // let mut queue_lock = self.queue.lock();
 
-        loop {
-            if !proc.wait_flag() {
-                return Ok(());
-            }
+        // queue_lock.push_back(thread.id());
+        // thread.setup_wait(self);
 
-            drop(queue_lock);
-            proc.enter_wait();
-            queue_lock = self.queue.lock();
+        // if let Some(deadline) = deadline {
+        //     TICK_LIST.lock().push_back(Timeout {
+        //         tid: thread.id(),
+        //         deadline,
+        //     });
+        // }
 
-            if let Some(deadline) = deadline {
-                if machine::local_timer().timestamp()? > deadline {
-                    let mut cursor = queue_lock.cursor_front_mut();
+        // loop {
+        //     match thread.wait_status() {
+        //         WaitStatus::Pending => {}
+        //         WaitStatus::Done => {
+        //             return Ok(());
+        //         }
+        //         WaitStatus::Interrupted => {
+        //             return Err(Errno::Interrupt);
+        //         }
+        //     };
 
-                    while let Some(&mut item) = cursor.current() {
-                        if proc.id() == item {
-                            cursor.remove_current();
-                            break;
-                        } else {
-                            cursor.move_next();
-                        }
-                    }
+        //     drop(queue_lock);
+        //     thread.enter_wait();
+        //     queue_lock = self.queue.lock();
 
-                    return Err(Errno::TimedOut);
-                }
-            }
-        }
+        //     if let Some(deadline) = deadline {
+        //         if machine::local_timer().timestamp()? > deadline {
+        //             let mut cursor = queue_lock.cursor_front_mut();
+
+        //             while let Some(&mut item) = cursor.current() {
+        //                 if thread.id() == item {
+        //                     cursor.remove_current();
+        //                     break;
+        //                 } else {
+        //                     cursor.move_next();
+        //                 }
+        //             }
+
+        //             return Err(Errno::TimedOut);
+        //         }
+        //     }
+        // }
     }
 }

kernel/src/sync.rs

@@ -38,7 +38,7 @@ impl<T> IrqSafeSpinLock<T> {
     #[inline(always)]
     unsafe fn force_release(&self) {
         self.state.store(false, Ordering::Release);
-        cortex_a::asm::sev();
+        // cortex_a::asm::sev();
     }
 
     /// Returns [IrqSafeSpinLockGuard] for this lock
@@ -47,7 +47,7 @@ impl<T> IrqSafeSpinLock<T> {
         let irq_state = unsafe { irq_mask_save() };
 
         while self.try_lock().is_err() {
-            cortex_a::asm::wfe();
+            // cortex_a::asm::wfe();
         }
 
         IrqSafeSpinLockGuard {

kernel/src/syscall/arg.rs

@@ -1,73 +1,205 @@
 //! 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> {
-    let mut res: usize;
-    unsafe {
-        asm!("at s1e1r, {}; mrs {}, par_el1", in(reg) virt, out(reg) res);
+// TODO _mut() versions checking whether pages are actually writable
+
+macro_rules! invalid_memory {
+    ($($args:tt)+) => {
+        warnln!($($args)+);
+        #[cfg(feature = "aggressive_syscall")]
+        {
+            todo!()
+            // 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);
     }
-    if res & 1 == 0 {
-        Some(res & !(0xFFF | (0xFF << 56)))
+}
+
+cfg_if! {
+    if #[cfg(target_arch = "aarch64")] {
+        #[inline(always)]
+        fn is_el0_accessible(virt: usize, write: bool) -> bool {
+            let mut res: usize;
+            unsafe {
+                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
+        }
     } else {
-        None
+        #[inline(always)]
+        fn is_el0_accessible(virt: usize, write: bool) -> bool {
+            // TODO implement this
+            true
+        }
     }
 }
 
-/// 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();
+    let asid = process.asid();
+
     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 {
+                todo!();
+                process.manipulate_space(|space| {
+                    // space.try_cow_copy(i * mem::PAGE_SIZE)?;
+                    // Process::invalidate_asid(asid);
+                    Ok(())
+                })
+            } 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 +209,3 @@ 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 +1,34 @@
 //! System call implementation
 
-use crate::arch::platform::exception::ExceptionFrame;
+// use crate::arch::{machine, platform::exception::ExceptionFrame};
+use crate::arch::platform::ForkFrame;
 use crate::debug::Level;
-use crate::proc::{elf, wait, Process, ProcessIo};
-use core::mem::size_of;
+// use crate::dev::timer::TimestampSource;
+use crate::fs::create_filesystem;
+use crate::mem::{
+    phys::PageUsage,
+    virt::table::{AddressSpace, MapAttributes},
+};
+use crate::proc::{self, elf, wait, Process, ProcessIo, Thread};
+// use core::mem::size_of;
 use core::ops::DerefMut;
-use core::time::Duration;
+// use core::time::Duration;
 use libsys::{
-    abi,
+    abi::SystemCall,
+    debug::TraceLevel,
     error::Errno,
     ioctl::IoctlCmd,
-    proc::Pid,
+    proc::{ExitCode, MemoryAccess, Pid, Tid},
     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()].
@@ -27,7 +37,7 @@ pub use arg::*;
 ///
 /// Unsafe: accepts and clones process states. Only legal to call
 /// from exception handlers.
-pub unsafe fn sys_fork(regs: &mut ExceptionFrame) -> Result<Pid, Errno> {
+pub unsafe fn sys_fork(regs: &mut ForkFrame) -> Result<Pid, Errno> {
     Process::current().fork(regs)
 }
 
@@ -50,171 +60,407 @@ fn find_at_node<T: DerefMut<Target = ProcessIo>>(
     }
 }
 
-/// Main system call dispatcher function
-pub fn syscall(num: usize, args: &[usize]) -> Result<usize, Errno> {
+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 system calls
-        abi::SYS_OPENAT => {
-            let at_fd = FileDescriptor::from_i32(args[0] as i32)?;
-            let path = validate_user_str(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)?;
-
-            let proc = Process::current();
-            let mut io = proc.io.lock();
-
-            let at = if let Some(fd) = at_fd {
-                io.file(fd)?.borrow().node()
-            } else {
-                None
-            };
-
-            let file = io.ioctx().open(at, path, mode, opts)?;
-            Ok(u32::from(io.place_file(file)?) as usize)
-        }
-        abi::SYS_READ => {
+        // I/O
+        SystemCall::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])?;
+            let buf = arg::buf_mut(args[1], args[2])?;
 
             io.file(fd)?.borrow_mut().read(buf)
         }
-        abi::SYS_WRITE => {
+        SystemCall::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])?;
+            let buf = arg::buf_ref(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;
+        //         SystemCall::Open => {
+        //             let at_fd = FileDescriptor::from_i32(args[0] as i32)?;
+        //             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)?;
+        //
+        //             let proc = Process::current();
+        //             let mut io = proc.io.lock();
+        //
+        //             let at = if let Some(fd) = at_fd {
+        //                 io.file(fd)?.borrow().node()
+        //             } else {
+        //                 None
+        //             };
+        //
+        //             let file = io.ioctx().open(at, path, mode, opts)?;
+        //             Ok(u32::from(io.place_file(file)?) as usize)
+        //         }
+        //         SystemCall::Close => {
+        //             let proc = Process::current();
+        //             let mut io = proc.io.lock();
+        //             let fd = FileDescriptor::from(args[0] as u32);
+        //
+        //             io.close_file(fd)?;
+        //             Ok(0)
+        //         }
+        //         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;
+        //
+        //             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)
+        //         }
+        //         SystemCall::Ioctl => {
+        //             let fd = FileDescriptor::from(args[0] as u32);
+        //             let cmd = IoctlCmd::try_from(args[1] as u32)?;
+        //
+        //             let proc = Process::current();
+        //             let mut io = proc.io.lock();
+        //
+        //             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))
+        //             };
+        //
+        //             wait::select(Thread::current(), rfds, wfds, timeout)
+        //         }
+        //         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 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 => {
-            let proc = Process::current();
-            let mut io = proc.io.lock();
-            let fd = FileDescriptor::from(args[0] as u32);
-
-            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])?;
-
-            match Process::waitpid(pid) {
-                Ok(exit) => {
-                    *status = i32::from(exit);
-                    Ok(0)
+            let mut attrs =
+                MapAttributes::NOT_GLOBAL | MapAttributes::SHARE_OUTER | MapAttributes::USER_READ;
+            if !acc.contains(MemoryAccess::READ) {
+                return Err(Errno::NotImplemented);
+            }
+            if acc.contains(MemoryAccess::WRITE) {
+                if acc.contains(MemoryAccess::EXEC) {
+                    return Err(Errno::PermissionDenied);
                 }
-                _ => todo!(),
+                attrs |= MapAttributes::USER_WRITE | MapAttributes::KERNEL_WRITE;
+            } else {
+                attrs |= MapAttributes::USER_READ;
             }
+            if !acc.contains(MemoryAccess::EXEC) {
+                attrs |= MapAttributes::USER_EXEC;
+            }
+
+            // 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)
+            })
         }
-        abi::SYS_IOCTL => {
-            let fd = FileDescriptor::from(args[0] as u32);
-            let cmd = IoctlCmd::try_from(args[1] as u32)?;
+        //         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| u32::from(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 = Pid::try_from(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 = Tid::from(args[0] as u32);
+        //
+        //             match Thread::waittid(tid) {
+        //                 Ok(_) => Ok(0),
+        //                 _ => todo!(),
+        //             }
+        //         }
+        //         SystemCall::GetPid => Ok(u32::from(Process::current().id()) as usize),
+        //         SystemCall::GetTid => Ok(u32::from(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) {
+        //                 warnln!("Sleep interrupted, {:?} remaining", rem);
+        //                 if rem_buf.is_some() {
+        //                     todo!()
+        //                 }
+        //             }
+        //             res.map(|_| 0)
+        //         }
+        //         SystemCall::SetSignalEntry => {
+        //             Thread::current().set_signal_entry(args[0], args[1]);
+        //             Ok(0)
+        //         }
+        //         SystemCall::SignalReturn => {
+        //             Thread::current().return_from_signal();
+        //             unreachable!();
+        //         }
+        //         SystemCall::SendSignal => {
+        //             let target = SignalDestination::from(args[0] as isize);
+        //             let signal = Signal::try_from(args[1] as u32)?;
+        //
+        //             match target {
+        //                 SignalDestination::This => Process::current().set_signal(signal),
+        //                 SignalDestination::Process(pid) => Process::get(pid)
+        //                     .ok_or(Errno::DoesNotExist)?
+        //                     .set_signal(signal),
+        //                 _ => todo!(),
+        //             };
+        //             Ok(0)
+        //         }
+        //         SystemCall::Yield => {
+        //             proc::switch();
+        //             Ok(0)
+        //         }
+        //         SystemCall::GetSid => {
+        //             // TODO handle kernel processes here?
+        //             let pid = Pid::to_option(args[0] as u32);
+        //             let current = Process::current();
+        //             let proc = if let Some(pid) = pid {
+        //                 let proc = Process::get(pid).ok_or(Errno::DoesNotExist)?;
+        //                 if proc.sid() != current.sid() {
+        //                     return Err(Errno::PermissionDenied);
+        //                 }
+        //                 proc
+        //             } else {
+        //                 current
+        //             };
+        //
+        //             Ok(u32::from(proc.sid()) as usize)
+        //         }
+        //         SystemCall::GetPgid => {
+        //             // TODO handle kernel processes here?
+        //             let pid = Pid::to_option(args[0] as u32);
+        //             let current = Process::current();
+        //             let proc = if let Some(pid) = pid {
+        //                 Process::get(pid).ok_or(Errno::DoesNotExist)?
+        //             } else {
+        //                 current
+        //             };
+        //
+        //             Ok(u32::from(proc.pgid()) as usize)
+        //         }
+        //         SystemCall::GetPpid => Ok(u32::from(Process::current().ppid().unwrap()) as usize),
+        //         SystemCall::SetSid => {
+        //             let proc = Process::current();
+        //             let mut io = proc.io.lock();
+        //
+        //             if let Some(_ctty) = io.ctty() {
+        //                 todo!();
+        //             }
+        //
+        //             let id = proc.id();
+        //             proc.set_sid(id);
+        //             Ok(u32::from(id) as usize)
+        //         }
+        //         SystemCall::SetPgid => {
+        //             let pid = Pid::to_option(args[0] as u32);
+        //             let pgid = Pid::to_option(args[1] as u32);
+        //
+        //             let current = Process::current();
+        //             let proc = if let Some(_pid) = pid {
+        //                 todo!()
+        //             } else {
+        //                 current
+        //             };
+        //
+        //             if let Some(_pgid) = pgid {
+        //                 todo!();
+        //             } else {
+        //                 proc.set_pgid(proc.id());
+        //             }
+        //
+        //             Ok(u32::from(proc.pgid()) as usize)
+        //         }
+        //
+        //         // 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();
 
-            let node = io.file(fd)?.borrow().node().ok_or(Errno::InvalidFile)?;
-            node.ioctl(cmd, args[2], args[3])
-        }
+            debugln!("mount(target={:?}, options={:#x?})", target, options);
 
-        // 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, "");
+            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])
         }
-        abi::SYS_EX_NANOSLEEP => {
-            let rem_buf = validate_user_ptr_null(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) {
-                warnln!("Sleep interrupted, {:?} remaining", rem);
-                if rem_buf.is_some() {
-                    todo!()
-                }
-            }
-            res.map(|_| 0)
-        }
-        abi::SYS_EX_SIGNAL => {
-            let proc = Process::current();
-            proc.setup_signal_context(args[0], args[1]);
-            Ok(0)
-        }
-        abi::SYS_EX_SIGRETURN => {
-            let proc = Process::current();
-            proc.return_from_signal();
-            panic!("This code won't run");
-        }
-        abi::SYS_EX_KILL => {
-            let target = SignalDestination::from(args[0] as isize);
-            let signal = Signal::try_from(args[1] as u32)?;
 
-            match target {
-                SignalDestination::This => Process::current().set_signal(signal),
-                SignalDestination::Process(pid) => Process::get(pid)
-                    .ok_or(Errno::DoesNotExist)?
-                    .set_signal(signal),
-                _ => todo!(),
-            };
-            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
-            } else {
-                Some(Duration::from_nanos(args[2] as u64))
-            };
-
-            let proc = Process::current();
-            wait::select(proc, rfds, wfds, timeout)
-        }
-
-        _ => {
-            let proc = Process::current();
-            errorln!("Undefined system call: {}", num);
-            proc.enter_signal(Signal::InvalidSystemCall);
-            todo!()
-        }
+        //         // Handled elsewhere
+        //         SystemCall::Fork => unreachable!(),
+        _ => panic!("Unimplemented: {:?}", num),
     }
 }
+
+/// Main system call dispatcher function
+pub fn syscall(num: SystemCall, args: &[usize]) -> Result<usize, Errno> {
+    let thread = Thread::current();
+    let process = thread.owner().unwrap();
+    let result = _syscall(num, args);
+    if !thread.is_handling_signal() {
+        // process.handle_pending_signals();
+    }
+    result
+}

kernel/src/util.rs

@@ -20,6 +20,14 @@ impl<T> InitOnce<T> {
         }
     }
 
+    pub fn as_ref_option(&self) -> Option<&T> {
+        if self.is_initialized() {
+            Some(self.get())
+        } else {
+            None
+        }
+    }
+
     /// Returns `true` if this [InitOnce<T>] can be used
     #[inline(always)]
     pub fn is_initialized(&self) -> bool {
@@ -29,7 +37,7 @@ impl<T> InitOnce<T> {
     /// Returns the initialized value. Will panic if the value has not
     /// yet been initialized.
     #[allow(clippy::mut_from_ref)]
-    pub fn get(&self) -> &mut T {
+    pub fn get(&self) -> &T {
         assert!(self.is_initialized(), "Access to uninitialized InitOnce<T>");
         unsafe { (*self.inner.get()).assume_init_mut() }
     }

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 +1,72 @@
-use crate::abi;
+use crate::abi::SystemCall;
 use crate::{
+    debug::TraceLevel,
     error::Errno,
     ioctl::IoctlCmd,
-    proc::{ExitCode, Pid},
+    proc::{ExitCode, MemoryAccess, MemoryMap, Pid, Tid},
     signal::{Signal, SignalDestination},
-    stat::{FdSet, FileDescriptor, FileMode, OpenFlags, Stat},
+    stat::{
+        AccessMode, DirectoryEntry, FdSet, FileDescriptor, FileMode, GroupId, MountOptions,
+        OpenFlags, Stat, UserId,
+    },
 };
+use core::time::Duration;
+use core::arch::asm;
 
 // TODO document the syscall ABI
 
+// rax
+//
+// rdi
+// rsi
+// rdx
+// r10
+// r8
+// r9
+
+// TODO move this to libusr
+#[cfg(target_arch = "x86_64")]
 macro_rules! syscall {
     ($num:expr) => {{
-        let mut res: usize;
-        asm!("svc #0", out("x0") res, in("x8") $num, options(nostack));
+        let mut res: usize = $num.repr();
+        asm!("syscall",
+             inout("rax") res,
+             options(nostack));
         res
     }};
     ($num:expr, $a0:expr) => {{
-        let mut res: usize = $a0;
-        asm!("svc #0",
-             inout("x0") res,
-             in("x8") $num, options(nostack));
+        let mut res: usize = $num.repr();
+        asm!("syscall",
+             inout("rax") res, in("rdi") $a0,
+             options(nostack));
         res
     }};
     ($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));
+        let mut res: usize = $num.repr();
+        asm!("syscall",
+             inout("rax") res, in("rdi") $a0, in("rsi") $a1,
+             options(nostack));
         res
     }};
     ($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));
+        let mut res: usize = $num.repr();
+        asm!("syscall",
+             inout("rax") res, in("rdi") $a0, in("rsi") $a1,
+             in("rdx") $a2, options(nostack));
         res
     }};
     ($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));
+        let mut res: usize = $num.repr();
+        asm!("syscall",
+             inout("rax") res, in("rdi") $a0, in("rsi") $a1,
+             in("rdx") $a2, in("r10") $a3, options(nostack));
         res
     }};
     ($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));
+        let mut res: usize = $num.repr();
+        asm!("syscall",
+             inout("rax") res, in("rdi") $a0, in("rsi") $a1,
+             in("rdx") $a2, in("r10") $a3, in("r8") $a4, options(nostack));
         res
     }};
 }
@@ -69,52 +88,38 @@ macro_rules! argp {
 //     ($a:expr) => ($a as *const core::ffi::c_void as usize)
 // }
 
-/// # Safety
-///
-/// System call
 #[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!();
 }
 
-/// # Safety
-///
-/// 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
-///
-/// System call
 #[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()))
     })
 }
 
-/// # Safety
-///
-/// 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())
         )
     })
 }
 
-/// # Safety
-///
-/// System call
 #[inline(always)]
 pub fn sys_openat(
     at: Option<FileDescriptor>,
@@ -124,7 +129,7 @@ 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()),
@@ -135,14 +140,11 @@ pub fn sys_openat(
     .map(|e| FileDescriptor::from(e as u32))
 }
 
-/// # Safety
-///
-/// System call
 #[inline(always)]
 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 +156,7 @@ 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())
@@ -162,9 +164,6 @@ pub fn sys_write(fd: FileDescriptor, data: &[u8]) -> Result<usize, Errno> {
     })
 }
 
-/// # Safety
-///
-/// System call
 #[inline(always)]
 pub fn sys_fstatat(
     at: Option<FileDescriptor>,
@@ -174,7 +173,7 @@ 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()),
@@ -188,47 +187,40 @@ 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)).and_then(|res| {
         if res != 0 {
-            Some(unsafe { Pid::from_raw(res as u32) })
+            Pid::try_from(res as u32).map(Some)
         } else {
-            None
+            Ok(None)
         }
     })
 }
 
-/// # Safety
-///
-/// 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())
         )
     })
 }
 
-/// # Safety
-///
-/// System call
 #[inline(always)]
 pub fn sys_waitpid(pid: Pid, status: &mut i32) -> Result<(), Errno> {
     Errno::from_syscall_unit(unsafe {
         syscall!(
-            abi::SYS_WAITPID,
-            argn!(pid.value()),
+            SystemCall::WaitPid,
+            argn!(u32::from(pid)),
             argp!(status as *mut i32)
         )
     })
 }
 
-/// # Safety
-///
-/// System call
 #[inline(always)]
 pub fn sys_ioctl(
     fd: FileDescriptor,
@@ -238,7 +230,7 @@ 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 +239,23 @@ 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 +264,41 @@ 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<Tid, Errno> {
+    Errno::from_syscall(unsafe {
+        syscall!(SystemCall::Clone, argn!(entry), argn!(stack), argn!(arg))
+    }).map(|e| Tid::from(e as u32))
+}
+
+#[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: Tid) -> Result<ExitCode, Errno> {
+    Errno::from_syscall(unsafe { syscall!(SystemCall::WaitTid, argn!(u32::from(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 +307,7 @@ 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 +318,150 @@ 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() -> Tid {
+    Tid::from(unsafe { syscall!(SystemCall::GetTid) as u32 })
+}
+
+#[inline(always)]
+pub fn sys_getpid() -> Pid {
+    Pid::try_from(unsafe { syscall!(SystemCall::GetPid) as u32 }).unwrap()
+}
+
+#[inline(always)]
+pub fn sys_getpgid(pid: Option<Pid>) -> Result<Pid, Errno> {
+    Errno::from_syscall(unsafe { syscall!(SystemCall::GetPgid, argn!(Pid::from_option(pid))) })
+        .and_then(|e| Pid::try_from(e as u32))
+}
+
+#[inline(always)]
+pub fn sys_setpgid(pid: Option<Pid>, pgid: Option<Pid>) -> Result<Pid, Errno> {
+    Errno::from_syscall(unsafe {
+        syscall!(
+            SystemCall::SetPgid,
+            argn!(Pid::from_option(pid)),
+            argn!(Pid::from_option(pgid))
+        )
+    })
+    .and_then(|e| Pid::try_from(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) })
+        .and_then(|e| Pid::try_from(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())
+        )
+    })
+}
+
+
+/// # Safety
+///
+/// System call
+#[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 +16,63 @@ 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
+}
+
+// TODO utils
+use core::fmt;
+
+#[derive(Clone, Copy)]
+pub struct FixedStr<const N: usize> {
+    len: usize,
+    data: [u8; N],
+}
+
+impl<const N: usize> FixedStr<N> {
+    pub const fn empty() -> Self {
+        Self {
+            len: 0,
+            data: [0; N]
+        }
+    }
+
+    pub fn copy_from_str(&mut self, src: &str) {
+        if src.len() > self.data.len() {
+            panic!("copy_from_str: src len > data len");
+        }
+        self.len = src.len();
+        self.data[..self.len].copy_from_slice(src.as_bytes());
+    }
+
+    pub fn as_str(&self) -> &str {
+        unsafe {
+            core::str::from_utf8_unchecked(&self.data[..self.len])
+        }
+    }
+}
+
+impl<const N: usize> fmt::Debug for FixedStr<N> {
+    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+        write!(f, "\"")?;
+        fmt::Display::fmt(self, f)?;
+        write!(f, "\"")
+    }
+}
+
+impl<const N: usize> fmt::Display for FixedStr<N> {
+    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+        for &byte in &self.data[..self.len] {
+            write!(f, "{}", byte as char)?;
+        }
+        Ok(())
+    }
+}
+
 #[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

@@ -12,10 +12,33 @@ pub struct ExitCode(i32);
 #[repr(transparent)]
 pub struct Pid(u32);
 
+/// Wrapper type for thread ID
+#[derive(Clone, Copy, PartialOrd, Ord, PartialEq, Eq)]
+#[repr(transparent)]
+pub struct Tid(u32);
+
 #[derive(Clone, Copy, PartialOrd, Ord, PartialEq, Eq, Debug)]
 #[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;
+        const ANONYMOUS = 1;
+
+        const SHARING = 0x3 << 2;
+        const PRIVATE = 1 << 2;
+    }
+}
+
 impl From<i32> for ExitCode {
     fn from(f: i32) -> Self {
         Self(f)
@@ -35,14 +58,15 @@ impl From<ExitCode> for i32 {
 }
 
 impl Pid {
-    /// Kernel idle process always has PID of zero
-    pub const IDLE: Self = Self(Self::KERNEL_BIT);
-
     const KERNEL_BIT: u32 = 1 << 31;
+    const USER_MAX: u32 = 256;
 
     /// Constructs an instance of user-space PID
     pub const fn user(id: u32) -> Self {
-        assert!(id < 256, "PID is too high");
+        assert!(id < Self::USER_MAX, "PID is too high");
+        if id == 0 {
+            panic!("User PID cannot be zero");
+        }
         Self(id)
     }
 
@@ -65,18 +89,20 @@ impl Pid {
         self.0 as u8
     }
 
-    /// Returns bit value of this pid
-    pub const fn value(self) -> u32 {
-        self.0
+    pub fn from_option(m: Option<Self>) -> u32 {
+        if let Some(pid) = m {
+            u32::from(pid)
+        } else {
+            0
+        }
     }
 
-    /// Constructs [Pid] from raw [u32] value
-    ///
-    /// # Safety
-    ///
-    /// Unsafe: does not check `num`
-    pub const unsafe fn from_raw(num: u32) -> Self {
-        Self(num)
+    pub fn to_option(m: u32) -> Option<Self> {
+        if m != 0 {
+            Some(Self::try_from(m).unwrap())
+        } else {
+            None
+        }
     }
 }
 
@@ -91,6 +117,26 @@ impl fmt::Debug for Pid {
     }
 }
 
+impl TryFrom<u32> for Pid {
+    type Error = Errno;
+
+    fn try_from(raw: u32) -> Result<Pid, Errno> {
+        if raw & Self::KERNEL_BIT != 0 {
+            Ok(Pid::kernel(raw & !Self::KERNEL_BIT))
+        } else if raw != 0 && raw < Self::USER_MAX {
+            Ok(Pid::user(raw))
+        } else {
+            Err(Errno::InvalidArgument)
+        }
+    }
+}
+
+impl From<Pid> for u32 {
+    fn from(pid: Pid) -> u32 {
+        pid.0
+    }
+}
+
 impl TryFrom<Pid> for Pgid {
     type Error = Errno;
 
@@ -114,3 +160,25 @@ impl From<Pgid> for u32 {
         p.0
     }
 }
+
+impl Tid {
+    pub const IDLE: Tid = Tid(0);
+}
+
+impl fmt::Debug for Tid {
+    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+        write!(f, "Tid(#{})", self.0)
+    }
+}
+
+impl From<u32> for Tid {
+    fn from(p: u32) -> Tid {
+        Self(p)
+    }
+}
+
+impl From<Tid> for u32 {
+    fn from(p: Tid) -> u32 {
+        p.0
+    }
+}

libsys/src/signal.rs

@@ -37,7 +37,7 @@ impl From<isize> for SignalDestination {
 impl From<SignalDestination> for isize {
     fn from(p: SignalDestination) -> isize {
         match p {
-            SignalDestination::Process(pid) => pid.value() as isize,
+            SignalDestination::Process(pid) => u32::from(pid) as isize,
             SignalDestination::Group(pgid) => -(u32::from(pgid) as isize),
             SignalDestination::This => 0,
             SignalDestination::All => -1

libsys/src/stat.rs

@@ -1,5 +1,7 @@
-use core::fmt;
+// TODO split up this file
 use crate::error::Errno;
+use core::str::FromStr;
+use core::fmt;
 
 const AT_FDCWD: i32 = -2;
 pub const AT_EMPTY_PATH: u32 = 1 << 16;
@@ -14,11 +16,18 @@ 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 +40,135 @@ 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 as_str(&self) -> &str {
+        let zero = self.name.iter().position(|&c| c == 0).unwrap();
+        core::str::from_utf8(&self.name[..zero]).unwrap()
+    }
+}
+
+impl FromStr for DirectoryEntry {
+    type Err = Errno;
+
+    fn from_str(i: &str) -> Result<Self, Errno> {
+        let mut res = DirectoryEntry { name: [0; 64] };
+        let bytes = i.as_bytes();
+        res.name[..bytes.len()].copy_from_slice(bytes);
+        Ok(res)
+    }
+}
+
+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 +200,7 @@ impl FdSet {
     }
 
     pub fn iter(&self) -> impl Iterator<Item = FileDescriptor> + '_ {
-        FdSetIter {
-            idx: 0,
-            set: self
-        }
+        FdSetIter { idx: 0, set: self }
     }
 }
 
@@ -122,13 +235,51 @@ 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 +8,7 @@ edition = "2021"
 [dependencies]
 libsys = { path = "../libsys", features = ["user"] }
 lazy_static = { version = "^1.4.0", features = ["spin_no_std"] }
+memoffset = "^0.6.4"
+
+[features]
+verbose = []

libusr/src/allocator.rs

@@ -0,0 +1,266 @@
+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,
+        )?;
+        #[cfg(feature = "verbose")]
+        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) {
+        #[cfg(feature = "verbose")]
+        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 = (*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 (*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;
+        #[cfg(feature = "verbose")]
+        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) {
+        #[cfg(feature = "verbose")]
+        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() {
+                (*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);
+            (*zone).list.del();
+            Zone::free(zone);
+        }
+    }
+}
+
+#[alloc_error_handler]
+fn alloc_error_handler(layout: Layout) -> ! {
+    panic!("Allocation failed: {:?}", layout);
+}
+
+#[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/mod.rs

@@ -0,0 +1,30 @@
+use crate::trace;
+use alloc::vec::Vec;
+use libsys::{
+    debug::TraceLevel,
+    ProgramArgs,
+};
+
+mod passwd;
+pub use passwd::UserInfo;
+mod shadow;
+pub use shadow::UserShadow;
+
+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) 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);
+    }
+
+    #[cfg(feature = "verbose")]
+    trace!(TraceLevel::Debug, "args = {:?}", PROGRAM_ARGS);
+}

libusr/src/env/passwd.rs

@@ -0,0 +1,99 @@
+use crate::io::{Read, read_line};
+use core::str::FromStr;
+use core::fmt;
+use crate::trace_debug;
+use crate::file::File;
+use libsys::{FixedStr, stat::{UserId, GroupId}};
+
+#[derive(Debug, Clone, Copy)]
+pub struct UserInfo {
+    name: FixedStr<32>,
+    uid: UserId,
+    gid: GroupId,
+    home: FixedStr<64>,
+    shell: FixedStr<64>,
+}
+
+impl UserInfo {
+    pub fn name(&self) -> &str {
+        self.name.as_str()
+    }
+
+    pub fn home(&self) -> &str {
+        self.home.as_str()
+    }
+
+    pub fn shell(&self) -> &str {
+        self.shell.as_str()
+    }
+
+    pub fn uid(&self) -> UserId {
+        self.uid
+    }
+
+    pub fn gid(&self) -> GroupId {
+        self.gid
+    }
+
+    pub fn find<F: Fn(&Self) -> bool>(pred: F) -> Result<Self, ()> {
+        let mut file = File::open("/etc/passwd").map_err(|_| ())?;
+        let mut buf = [0; 128];
+        loop {
+            let line = read_line(&mut file, &mut buf).map_err(|_| ())?;
+            if let Some(line) = line {
+                let ent = UserInfo::from_str(line)?;
+                if pred(&ent) {
+                    return Ok(ent);
+                }
+            } else {
+                break;
+            }
+        }
+        Err(())
+    }
+
+    pub fn by_name(name: &str) -> Result<Self, ()> {
+        Self::find(|ent| ent.name() == name)
+    }
+}
+
+impl FromStr for UserInfo {
+    type Err = ();
+
+    fn from_str(s: &str) -> Result<Self, ()> {
+        let mut iter = s.split(":");
+
+        let name = iter.next().ok_or(())?;
+        let uid = iter
+            .next()
+            .ok_or(())
+            .and_then(|e| u32::from_str(e).map_err(|_| ()))
+            .map(UserId::from)?;
+        let gid = iter
+            .next()
+            .ok_or(())
+            .and_then(|e| u32::from_str(e).map_err(|_| ()))
+            .map(GroupId::from)?;
+        let comment = iter.next().ok_or(())?;
+        let home = iter.next().ok_or(())?;
+        let shell = iter.next().ok_or(())?;
+
+        if iter.next().is_some() {
+            return Err(());
+        }
+
+        let mut res = Self {
+            uid,
+            gid,
+            name: FixedStr::empty(),
+            home: FixedStr::empty(),
+            shell: FixedStr::empty(),
+        };
+
+        res.name.copy_from_str(&name);
+        res.home.copy_from_str(&home);
+        res.shell.copy_from_str(&shell);
+
+        Ok(res)
+    }
+}

libusr/src/env/shadow.rs

@@ -0,0 +1,67 @@
+use crate::file::File;
+use crate::io::{Read, read_line};
+use core::str::FromStr;
+use libsys::FixedStr;
+
+#[derive(Debug, Clone, Copy)]
+pub struct UserShadow {
+    name: FixedStr<32>,
+    password: FixedStr<64>,
+}
+
+impl UserShadow {
+    pub fn name(&self) -> &str {
+        self.name.as_str()
+    }
+
+    pub fn password(&self) -> &str {
+        self.password.as_str()
+    }
+
+
+    pub fn find<F: Fn(&Self) -> bool>(pred: F) -> Result<Self, ()> {
+        let mut file = File::open("/etc/shadow").map_err(|_| ())?;
+        let mut buf = [0; 128];
+        loop {
+            let line = read_line(&mut file, &mut buf).map_err(|_| ())?;
+            if let Some(line) = line {
+                let ent = UserShadow::from_str(line)?;
+                if pred(&ent) {
+                    return Ok(ent);
+                }
+            } else {
+                break;
+            }
+        }
+        Err(())
+    }
+
+    pub fn by_name(name: &str) -> Result<Self, ()> {
+        Self::find(|ent| ent.name() == name)
+    }
+}
+
+impl FromStr for UserShadow {
+    type Err = ();
+
+    fn from_str(s: &str) -> Result<Self, ()> {
+        let mut iter = s.split(':');
+
+        let name = iter.next().ok_or(())?;
+        let password = iter.next().ok_or(())?;
+
+        if iter.next().is_some() {
+            return Err(());
+        }
+
+        let mut res = Self {
+            name: FixedStr::empty(),
+            password: FixedStr::empty(),
+        };
+
+        res.name.copy_from_str(name);
+        res.password.copy_from_str(password);
+
+        Ok(res)
+    }
+}

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 +28,41 @@ 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)
 }
+
+// TODO use BufRead instead once it's implemented
+pub(crate) fn read_line<'a, F: Read>(f: &mut F, buf: &'a mut [u8]) -> Result<Option<&'a str>, ()> {
+    let mut pos = 0;
+    loop {
+        if pos == buf.len() {
+            return Err(());
+        }
+
+        let count = f.read(&mut buf[pos..=pos]).map_err(|_| ())?;
+        if count == 0 {
+            if pos == 0 {
+                return Ok(None);
+            }
+            break;
+        }
+        if buf[pos] == b'\n' {
+            break;
+        }
+
+        pos += 1;
+    }
+    core::str::from_utf8(&buf[..pos]).map_err(|_| ()).map(Some)
+}

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 {
@@ -8,7 +8,8 @@ macro_rules! print {
 
 #[macro_export]
 macro_rules! println {
-    ($($args:tt)+) => (print!("{}\n", format_args!($($args)+)))
+    ($($args:tt)+) => (print!("{}\n", format_args!($($args)+)));
+    () => (print!("\n"));
 }
 
 #[macro_export]
@@ -18,7 +19,8 @@ macro_rules! eprint {
 
 #[macro_export]
 macro_rules! eprintln {
-    ($($args:tt)+) => (eprint!("{}\n", format_args!($($args)+)))
+    ($($args:tt)+) => (eprint!("{}\n", format_args!($($args)+)));
+    () => (eprint!("\n"));
 }
 
 pub fn _print<T: Write>(out: fn() -> T, args: fmt::Arguments) {