Add Ext2 fs and basic bio layer

Cargo.lock

@@ -121,9 +121,12 @@ dependencies = [
  "aster-util",
  "bitflags 1.3.2",
  "component",
+ "int-to-c-enum",
  "lazy_static",
  "log",
+ "pod",
  "spin 0.9.8",
+ "static_assertions",
 ]
 
 [[package]]
@@ -270,10 +273,12 @@ dependencies = [
  "aster-util",
  "aster-virtio",
  "bitflags 1.3.2",
+ "bitvec",
  "controlled",
  "core2",
  "cpio-decoder",
  "getrandom",
+ "inherit-methods-macro",
  "int-to-c-enum",
  "intrusive-collections",
  "keyable-arc",
@@ -287,6 +292,7 @@ dependencies = [
  "ringbuf",
  "smoltcp",
  "spin 0.9.8",
+ "static_assertions",
  "tdx-guest",
  "time",
  "typeflags",

framework/aster-frame/src/vm/frame.rs

@@ -543,7 +543,7 @@ impl<'a> VmReader<'a> {
     /// Limits the length of remaining data.
     ///
     /// This method ensures the postcondition of `self.remain() <= max_remain`.
-    pub const fn limit(&mut self, max_remain: usize) -> &mut Self {
+    pub const fn limit(mut self, max_remain: usize) -> Self {
         if max_remain < self.remain() {
             // Safety: the new end is less than the old end.
             unsafe { self.end = self.cursor.add(max_remain) };
@@ -557,7 +557,7 @@ impl<'a> VmReader<'a> {
     /// # Panic
     ///
     /// If `nbytes` is greater than `self.remain()`, then the method panics.
-    pub fn skip(&mut self, nbytes: usize) -> &mut Self {
+    pub fn skip(mut self, nbytes: usize) -> Self {
         assert!(nbytes <= self.remain());
 
         // Safety: the new cursor is less than or equal to the end.
@@ -653,7 +653,7 @@ impl<'a> VmWriter<'a> {
     /// Limits the length of available space.
     ///
     /// This method ensures the postcondition of `self.avail() <= max_avail`.
-    pub const fn limit(&mut self, max_avail: usize) -> &mut Self {
+    pub const fn limit(mut self, max_avail: usize) -> Self {
         if max_avail < self.avail() {
             // Safety: the new end is less than the old end.
             unsafe { self.end = self.cursor.add(max_avail) };
@@ -667,7 +667,7 @@ impl<'a> VmWriter<'a> {
     /// # Panic
     ///
     /// If `nbytes` is greater than `self.avail()`, then the method panics.
-    pub fn skip(&mut self, nbytes: usize) -> &mut Self {
+    pub fn skip(mut self, nbytes: usize) -> Self {
         assert!(nbytes <= self.avail());
 
         // Safety: the new cursor is less than or equal to the end.

regression/Makefile

@@ -3,6 +3,7 @@ CUR_DIR := $(patsubst %/,%,$(dir $(MKFILE_PATH)))
 BUILD_DIR := $(CUR_DIR)/build
 INITRAMFS := $(BUILD_DIR)/initramfs
 INITRAMFS_IMAGE := $(BUILD_DIR)/initramfs.cpio.gz
+EXT2_IMAGE := $(BUILD_DIR)/ext2.img
 SHELL := /bin/bash
 INITRAMFS_EMPTY_DIRS := \
 	$(INITRAMFS)/sbin \
@@ -10,7 +11,8 @@ INITRAMFS_EMPTY_DIRS := \
 	$(INITRAMFS)/tmp \
 	$(INITRAMFS)/opt \
 	$(INITRAMFS)/proc \
-	$(INITRAMFS)/dev
+	$(INITRAMFS)/dev \
+	$(INITRAMFS)/ext2
 INITRAMFS_ALL_DIRS := \
 	$(INITRAMFS)/etc \
 	$(INITRAMFS)/lib/x86_64-linux-gnu \
@@ -90,7 +92,11 @@ endif
 	@echo "Generating the initramfs image..."
 	@(cd $(INITRAMFS); find . | cpio -o -H newc | gzip) > $@
 
-build: $(INITRAMFS_IMAGE)
+$(EXT2_IMAGE):
+	@dd if=/dev/zero of=$(EXT2_IMAGE) bs=2G count=1
+	@mke2fs $(EXT2_IMAGE)
+
+build: $(INITRAMFS_IMAGE) $(EXT2_IMAGE)
 
 clean:
 	@rm -rf $(BUILD_DIR)

runner/src/main.rs

@@ -14,7 +14,6 @@ pub mod gdb;
 pub mod machine;
 
 use std::{
-    fs::OpenOptions,
     path::{Path, PathBuf},
     process::Command,
 };
@@ -175,9 +174,11 @@ fn main() {
         qemu_cmd.args(qemu_grub_efi::NOIOMMU_DEVICE_ARGS);
     }
 
-    let fs_image = create_fs_image(args.path.as_path());
+    // TODO: Add arguments to the runner CLI tool so that the user can specify
+    //       a list of disk drives, each of which may be in a different FS format.
+    let ext2_image = get_fs_image(&PathBuf::from("regression/build/ext2.img"), 0);
     qemu_cmd.arg("-drive");
-    qemu_cmd.arg(fs_image);
+    qemu_cmd.arg(ext2_image);
 
     if args.boot_method == BootMethod::Microvm {
         let image = microvm::create_bootdev_image(args.path);
@@ -221,20 +222,14 @@ fn main() {
     }
 }
 
-pub fn create_fs_image(path: &Path) -> String {
-    let mut fs_img_path = path.parent().unwrap().to_str().unwrap().to_string();
-    fs_img_path.push_str("/fs.img");
-    let path = Path::new(fs_img_path.as_str());
-    if path.exists() {
-        return format!("file={},if=none,format=raw,id=x0", fs_img_path.as_str());
+pub fn get_fs_image(path: &Path, drive_id: u32) -> String {
+    if !path.exists() {
+        panic!("can not find the fs image")
     }
-    let f = OpenOptions::new()
-        .read(true)
-        .write(true)
-        .create(true)
-        .open(fs_img_path.as_str())
-        .unwrap();
-    // 32MiB
-    f.set_len(64 * 1024 * 1024).unwrap();
-    format!("file={},if=none,format=raw,id=x0", fs_img_path.as_str())
+
+    format!(
+        "file={},if=none,format=raw,id=x{}",
+        path.to_string_lossy(),
+        drive_id
+    )
 }

services/comps/block/Cargo.toml

@@ -8,10 +8,13 @@ edition = "2021"
 [dependencies]
 bitflags = "1.3"
 spin = "0.9.4"
+pod = { git = "https://github.com/asterinas/pod", rev = "d7dba56" }
 aster-frame = { path = "../../../framework/aster-frame" }
 aster-util = { path = "../../libs/aster-util" }
+int-to-c-enum = { path = "../../libs/int-to-c-enum" }
 component = { path = "../../libs/comp-sys/component" }
 log = "0.4"
+static_assertions = "1.1.0"
 
 [features]
 

services/comps/block/src/bio.rs

@@ -0,0 +1,475 @@
+use crate::prelude::*;
+
+use super::{id::Sid, BlockDevice};
+
+use aster_frame::{
+    sync::WaitQueue,
+    vm::{VmFrame, VmReader, VmSegment, VmWriter},
+};
+use int_to_c_enum::TryFromInt;
+
+/// The unit for block I/O.
+///
+/// Each `Bio` packs the following information:
+/// (1) The type of the I/O,
+/// (2) The target sectors on the device for doing I/O,
+/// (3) The memory locations (`BioSegment`) from/to which data are read/written,
+/// (4) The optional callback function that will be invoked when the I/O is completed.
+pub struct Bio(Arc<BioInner>);
+
+impl Bio {
+    /// Constructs a new `Bio`.
+    ///
+    /// The `type_` describes the type of the I/O.
+    /// The `start_sid` is the starting sector id on the device.
+    /// The `segments` describes the memory segments.
+    /// The `complete_fn` is the optional callback function.
+    pub fn new(
+        type_: BioType,
+        start_sid: Sid,
+        segments: Vec<BioSegment>,
+        complete_fn: Option<fn(&SubmittedBio)>,
+    ) -> Self {
+        let nsectors = segments
+            .iter()
+            .map(|segment| segment.nsectors().to_raw())
+            .sum();
+
+        let inner = Arc::new(BioInner {
+            type_,
+            sid_range: start_sid..start_sid + nsectors,
+            segments,
+            complete_fn,
+            status: AtomicU32::new(BioStatus::Init as u32),
+            wait_queue: WaitQueue::new(),
+        });
+        Self(inner)
+    }
+
+    /// Returns the type.
+    pub fn type_(&self) -> BioType {
+        self.0.type_()
+    }
+
+    /// Returns the range of target sectors on the device.
+    pub fn sid_range(&self) -> &Range<Sid> {
+        self.0.sid_range()
+    }
+
+    /// Returns the slice to the memory segments.
+    pub fn segments(&self) -> &[BioSegment] {
+        self.0.segments()
+    }
+
+    /// Returns the status.
+    pub fn status(&self) -> BioStatus {
+        self.0.status()
+    }
+
+    /// Submits self to the `block_device` asynchronously.
+    ///
+    /// Returns a `BioWaiter` to the caller to wait for its completion.
+    ///
+    /// # Panic
+    ///
+    /// The caller must not submit a `Bio` more than once. Otherwise, a panic shall be triggered.
+    pub fn submit(&self, block_device: &dyn BlockDevice) -> Result<BioWaiter, BioEnqueueError> {
+        // Change the status from "Init" to "Submit".
+        let result = self.0.status.compare_exchange(
+            BioStatus::Init as u32,
+            BioStatus::Submit as u32,
+            Ordering::Release,
+            Ordering::Relaxed,
+        );
+        assert!(result.is_ok());
+
+        if let Err(e) = block_device
+            .request_queue()
+            .enqueue(SubmittedBio(self.0.clone()))
+        {
+            // Fail to submit, revert the status.
+            let result = self.0.status.compare_exchange(
+                BioStatus::Submit as u32,
+                BioStatus::Init as u32,
+                Ordering::Release,
+                Ordering::Relaxed,
+            );
+            assert!(result.is_ok());
+            return Err(e);
+        }
+
+        Ok(BioWaiter {
+            bios: vec![self.0.clone()],
+        })
+    }
+
+    /// Submits self to the `block_device` and waits for the result synchronously.
+    ///
+    /// Returns the result status of the `Bio`.
+    ///
+    /// # Panic
+    ///
+    /// The caller must not submit a `Bio` more than once. Otherwise, a panic shall be triggered.
+    pub fn submit_sync(
+        &self,
+        block_device: &dyn BlockDevice,
+    ) -> Result<BioStatus, BioEnqueueError> {
+        let waiter = self.submit(block_device)?;
+        match waiter.wait() {
+            Some(status) => {
+                assert!(status == BioStatus::Complete);
+                Ok(status)
+            }
+            None => {
+                let status = self.status();
+                assert!(status != BioStatus::Complete);
+                Ok(status)
+            }
+        }
+    }
+}
+
+/// The error type returned when enqueueing the `Bio`.
+#[derive(Clone, Copy, PartialEq, Eq, Debug)]
+pub enum BioEnqueueError {
+    /// The request queue is full
+    IsFull,
+    /// Refuse to enqueue the bio
+    Refused,
+}
+
+impl From<BioEnqueueError> for aster_frame::Error {
+    fn from(_error: BioEnqueueError) -> Self {
+        aster_frame::Error::NotEnoughResources
+    }
+}
+
+/// A waiter for `Bio` submissions.
+///
+/// This structure holds a list of `Bio` requests and provides functionality to
+/// wait for their completion and retrieve their statuses.
+#[must_use]
+pub struct BioWaiter {
+    bios: Vec<Arc<BioInner>>,
+}
+
+impl BioWaiter {
+    /// Constructs a new `BioWaiter` instance with no `Bio` requests.
+    pub fn new() -> Self {
+        Self { bios: Vec::new() }
+    }
+
+    /// Returns the number of `Bio` requests associated with `self`.
+    pub fn nreqs(&self) -> usize {
+        self.bios.len()
+    }
+
+    /// Gets the `index`-th `Bio` request associated with `self`.
+    ///
+    /// # Panic
+    ///
+    /// If the `index` is out of bounds, this method will panic.
+    pub fn req(&self, index: usize) -> Bio {
+        Bio(self.bios[index].clone())
+    }
+
+    /// Returns the status of the `index`-th `Bio` request associated with `self`.
+    ///
+    /// # Panic
+    ///
+    /// If the `index` is out of bounds, this method will panic.
+    pub fn status(&self, index: usize) -> BioStatus {
+        self.bios[index].status()
+    }
+
+    /// Merges the `Bio` requests from another `BioWaiter` into this one.
+    ///
+    /// The another `BioWaiter`'s `Bio` requests are appended to the end of
+    /// the `Bio` list of `self`, effectively concatenating the two lists.
+    pub fn concat(&mut self, mut other: Self) {
+        self.bios.append(&mut other.bios);
+    }
+
+    /// Waits for the completion of all `Bio` requests.
+    ///
+    /// This method iterates through each `Bio` in the list, waiting for their
+    /// completion.
+    ///
+    /// The return value is an option indicating whether all the requests in the list
+    /// have successfully completed.
+    /// On success this value is guaranteed to be equal to `Some(BioStatus::Complete)`.
+    pub fn wait(&self) -> Option<BioStatus> {
+        let mut ret = Some(BioStatus::Complete);
+
+        for bio in self.bios.iter() {
+            let status = bio.wait_queue.wait_until(|| {
+                let status = bio.status();
+                if status != BioStatus::Submit {
+                    Some(status)
+                } else {
+                    None
+                }
+            });
+            if status != BioStatus::Complete && ret.is_some() {
+                ret = None;
+            }
+        }
+
+        ret
+    }
+}
+
+impl Default for BioWaiter {
+    fn default() -> Self {
+        Self::new()
+    }
+}
+
+/// A submitted `Bio` object.
+///
+/// The request queue of block device only accepts a `SubmittedBio` into the queue.
+pub struct SubmittedBio(Arc<BioInner>);
+
+impl SubmittedBio {
+    /// Returns the type.
+    pub fn type_(&self) -> BioType {
+        self.0.type_()
+    }
+
+    /// Returns the range of target sectors on the device.
+    pub fn sid_range(&self) -> &Range<Sid> {
+        self.0.sid_range()
+    }
+
+    /// Returns the slice to the memory segments.
+    pub fn segments(&self) -> &[BioSegment] {
+        self.0.segments()
+    }
+
+    /// Returns the status.
+    pub fn status(&self) -> BioStatus {
+        self.0.status()
+    }
+
+    /// Completes the `Bio` with the `status` and invokes the callback function.
+    ///
+    /// When the driver finishes the request for this `Bio`, it will call this method.
+    pub fn complete(&self, status: BioStatus) {
+        assert!(status != BioStatus::Init && status != BioStatus::Submit);
+
+        // Set the status.
+        let result = self.0.status.compare_exchange(
+            BioStatus::Submit as u32,
+            status as u32,
+            Ordering::Release,
+            Ordering::Relaxed,
+        );
+        assert!(result.is_ok());
+
+        self.0.wait_queue.wake_all();
+        if let Some(complete_fn) = self.0.complete_fn {
+            complete_fn(self);
+        }
+    }
+}
+
+/// The common inner part of `Bio`.
+struct BioInner {
+    /// The type of the I/O
+    type_: BioType,
+    /// The range of the sector id on device
+    sid_range: Range<Sid>,
+    /// The memory segments in this `Bio`
+    segments: Vec<BioSegment>,
+    /// The I/O completion method
+    complete_fn: Option<fn(&SubmittedBio)>,
+    /// The I/O status
+    status: AtomicU32,
+    /// The wait queue for I/O completion
+    wait_queue: WaitQueue,
+}
+
+impl BioInner {
+    pub fn type_(&self) -> BioType {
+        self.type_
+    }
+
+    pub fn sid_range(&self) -> &Range<Sid> {
+        &self.sid_range
+    }
+
+    pub fn segments(&self) -> &[BioSegment] {
+        &self.segments
+    }
+
+    pub fn status(&self) -> BioStatus {
+        BioStatus::try_from(self.status.load(Ordering::Relaxed)).unwrap()
+    }
+}
+
+/// The type of `Bio`.
+#[derive(Clone, Copy, Debug, PartialEq, TryFromInt)]
+#[repr(u8)]
+pub enum BioType {
+    /// Read sectors from the device.
+    Read = 0,
+    /// Write sectors into the device.
+    Write = 1,
+    /// Flush the volatile write cache.
+    Flush = 2,
+    /// Discard sectors.
+    Discard = 3,
+}
+
+/// The status of `Bio`.
+#[derive(Clone, Copy, PartialEq, Eq, Debug, TryFromInt)]
+#[repr(u32)]
+pub enum BioStatus {
+    /// The initial status for a newly created `Bio`.
+    Init = 0,
+    /// After a `Bio` is submitted, its status will be changed to "Submit".
+    Submit = 1,
+    /// The I/O operation has been successfully completed.
+    Complete = 2,
+    /// The I/O operation is not supported.
+    NotSupported = 3,
+    /// Insufficient space is available to perform the I/O operation.
+    NoSpace = 4,
+    /// An error occurred while doing I/O.
+    IoError = 5,
+}
+
+/// `BioSegment` is a smallest memory unit in block I/O.
+///
+/// It is a contiguous memory region that contains multiple sectors.
+#[derive(Debug, Clone)]
+pub struct BioSegment {
+    /// The contiguous pages on which this segment resides.
+    pages: Pages,
+    /// The offset (in bytes) relative to the first page.
+    offset: AlignedUsize<SECTOR_SIZE>,
+    // The length (in bytes), may cross pages.
+    len: AlignedUsize<SECTOR_SIZE>,
+}
+
+const SECTOR_SIZE: u16 = super::SECTOR_SIZE as u16;
+
+#[derive(Debug, Clone)]
+enum Pages {
+    Frame(VmFrame),
+    Segment(VmSegment),
+}
+
+impl<'a> BioSegment {
+    /// Constructs a new `BioSegment` from `VmSegment`.
+    pub fn from_segment(segment: VmSegment, offset: usize, len: usize) -> Self {
+        assert!(offset + len <= segment.nbytes());
+
+        Self {
+            pages: Pages::Segment(segment),
+            offset: AlignedUsize::<SECTOR_SIZE>::new(offset).unwrap(),
+            len: AlignedUsize::<SECTOR_SIZE>::new(len).unwrap(),
+        }
+    }
+
+    /// Constructs a new `BioSegment` from `VmFrame`.
+    pub fn from_frame(frame: VmFrame, offset: usize, len: usize) -> Self {
+        assert!(offset + len <= super::BLOCK_SIZE);
+
+        Self {
+            pages: Pages::Frame(frame),
+            offset: AlignedUsize::<SECTOR_SIZE>::new(offset).unwrap(),
+            len: AlignedUsize::<SECTOR_SIZE>::new(len).unwrap(),
+        }
+    }
+
+    /// Returns the number of sectors.
+    pub fn nsectors(&self) -> Sid {
+        Sid::from_offset(self.len.value())
+    }
+
+    /// Returns the number of bytes.
+    pub fn nbytes(&self) -> usize {
+        self.len.value()
+    }
+
+    /// Returns a reader to read data from it.
+    pub fn reader(&'a self) -> VmReader<'a> {
+        let reader = match &self.pages {
+            Pages::Segment(segment) => segment.reader(),
+            Pages::Frame(frame) => frame.reader(),
+        };
+        reader.skip(self.offset.value()).limit(self.len.value())
+    }
+
+    /// Returns a writer to write data into it.
+    pub fn writer(&'a self) -> VmWriter<'a> {
+        let writer = match &self.pages {
+            Pages::Segment(segment) => segment.writer(),
+            Pages::Frame(frame) => frame.writer(),
+        };
+        writer.skip(self.offset.value()).limit(self.len.value())
+    }
+}
+
+/// An aligned unsigned integer number.
+///
+/// An instance of `AlignedUsize<const N: u16>` is guaranteed to have a value that is a multiple
+/// of `N`, a predetermined const value. It is preferable to express an unsigned integer value
+/// in type `AlignedUsize<_>` instead of `usize` if the value must satisfy an alignment requirement.
+/// This helps readability and prevents bugs.
+///
+/// # Examples
+///
+/// ```rust
+/// const SECTOR_SIZE: u16 = 512;
+///
+/// let sector_num = 1234; // The 1234-th sector
+/// let sector_offset: AlignedUsize<SECTOR_SIZE> = {
+///     let sector_offset = sector_num * (SECTOR_SIZE as usize);
+///     AlignedUsize::<SECTOR_SIZE>::new(sector_offset).unwrap()
+/// };
+/// assert!(sector_offset.value() % sector_offset.align() == 0);
+/// ```
+///
+/// # Limitation
+///
+/// Currently, the alignment const value must be expressed in `u16`;
+/// it is not possible to use a larger or smaller type.
+/// This limitation is inherited from that of Rust's const generics:
+/// your code can be generic over the _value_ of a const, but not the _type_ of the const.
+/// We choose `u16` because it is reasonably large to represent any alignment value
+/// used in practice.
+#[derive(Debug, Clone)]
+pub struct AlignedUsize<const N: u16>(usize);
+
+impl<const N: u16> AlignedUsize<N> {
+    /// Constructs a new instance of aligned integer if the given value is aligned.
+    pub fn new(val: usize) -> Option<Self> {
+        if val % (N as usize) == 0 {
+            Some(Self(val))
+        } else {
+            None
+        }
+    }
+
+    /// Returns the value.
+    pub fn value(&self) -> usize {
+        self.0
+    }
+
+    /// Returns the corresponding ID.
+    ///
+    /// The so-called "ID" of an aligned integer is defined to be `self.value() / self.align()`.
+    /// This value is named ID because one common use case is using `Aligned` to express
+    /// the byte offset of a sector, block, or page. In this case, the `id` method returns
+    /// the ID of the corresponding sector, block, or page.
+    pub fn id(&self) -> usize {
+        self.value() / self.align()
+    }
+
+    /// Returns the alignment.
+    pub fn align(&self) -> usize {
+        N as usize
+    }
+}

services/comps/block/src/id.rs

@@ -0,0 +1,100 @@
+use core::{
+    iter::Step,
+    ops::{Add, Sub},
+};
+use pod::Pod;
+use static_assertions::const_assert;
+
+/// The block index used in the filesystem.
+pub type Bid = BlockId<BLOCK_SIZE>;
+/// The sector index used in the device.
+pub type Sid = BlockId<SECTOR_SIZE>;
+
+impl From<Bid> for Sid {
+    fn from(bid: Bid) -> Self {
+        Self::new(bid.to_raw() * (BLOCK_SIZE / SECTOR_SIZE) as u64)
+    }
+}
+
+const BLOCK_SIZE: u16 = super::BLOCK_SIZE as u16;
+const SECTOR_SIZE: u16 = super::SECTOR_SIZE as u16;
+const_assert!(BLOCK_SIZE / SECTOR_SIZE >= 1);
+
+/// An index of a block.
+///
+/// The `BlockId<const N: u16>` is a generic type that is parameterized by a constant `N`, which
+/// represents the size of each block in bytes. The `BlockId<_>` provides a type-safe way of handling
+/// block indices.
+/// An Instance of `BlockId<_>` is guaranteed to represent valid block index, derived from byte offset
+/// and the specified block size `N`.
+///
+/// # Examples
+///
+/// ```rust
+/// const BLOCK_SIZE: u16 = 512;
+///
+/// let bytes_offset = 2048;
+/// let block_id = BlockId<BLOCK_SIZE>::from_offset(bytes_offset);
+/// assert!(block_id == (bytes_offset / BLOCK_SIZE));
+/// ```
+///
+/// # Limitation
+///
+/// Currently, the block size is expressed in `u16`. We choose `u16` because
+/// it is reasonably large to represent the common block size used in practice.
+#[repr(C)]
+#[derive(Copy, Clone, Debug, Hash, PartialEq, Eq, PartialOrd, Ord, Pod)]
+pub struct BlockId<const N: u16>(u64);
+
+impl<const N: u16> BlockId<N> {
+    /// Constructs an id from a raw id.
+    pub const fn new(raw_id: u64) -> Self {
+        Self(raw_id)
+    }
+
+    /// Constructs an id from a byte offset.
+    pub const fn from_offset(offset: usize) -> Self {
+        Self((offset / (N as usize)) as _)
+    }
+
+    /// Converts to a byte offset.
+    pub fn to_offset(self) -> usize {
+        (self.0 as usize) * (N as usize)
+    }
+
+    /// Converts to raw id.
+    pub fn to_raw(self) -> u64 {
+        self.0
+    }
+}
+
+impl<const N: u16> Add<u64> for BlockId<N> {
+    type Output = Self;
+
+    fn add(self, other: u64) -> Self::Output {
+        Self(self.0 + other)
+    }
+}
+
+impl<const N: u16> Sub<u64> for BlockId<N> {
+    type Output = Self;
+
+    fn sub(self, other: u64) -> Self::Output {
+        Self(self.0 - other)
+    }
+}
+
+/// Implements the `Step` trait to iterate over `Range<Id>`.
+impl<const N: u16> Step for BlockId<N> {
+    fn steps_between(start: &Self, end: &Self) -> Option<usize> {
+        u64::steps_between(&start.0, &end.0)
+    }
+
+    fn forward_checked(start: Self, count: usize) -> Option<Self> {
+        u64::forward_checked(start.0, count).map(Self::new)
+    }
+
+    fn backward_checked(start: Self, count: usize) -> Option<Self> {
+        u64::backward_checked(start.0, count).map(Self::new)
+    }
+}

services/comps/block/src/impl_block_device.rs

@@ -0,0 +1,237 @@
+use crate::prelude::*;
+
+use super::{
+    bio::{Bio, BioEnqueueError, BioSegment, BioStatus, BioType, BioWaiter, SubmittedBio},
+    id::{Bid, Sid},
+    BlockDevice, BLOCK_SIZE, SECTOR_SIZE,
+};
+
+use aster_frame::vm::{VmAllocOptions, VmFrame, VmIo, VmSegment};
+
+/// Implements several commonly used APIs for the block device to conveniently
+/// read and write block(s).
+impl dyn BlockDevice {
+    /// Synchronously reads contiguous blocks starting from the `bid`.
+    pub fn read_blocks_sync(
+        &self,
+        bid: Bid,
+        segment: &VmSegment,
+    ) -> Result<BioStatus, BioEnqueueError> {
+        let bio = create_bio_from_segment(BioType::Read, bid, segment);
+        let status = bio.submit_sync(self)?;
+        Ok(status)
+    }
+
+    /// Asynchronously reads contiguous blocks starting from the `bid`.
+    pub fn read_blocks(&self, bid: Bid, segment: &VmSegment) -> Result<BioWaiter, BioEnqueueError> {
+        let bio = create_bio_from_segment(BioType::Read, bid, segment);
+        bio.submit(self)
+    }
+
+    /// Synchronously reads one block indicated by the `bid`.
+    pub fn read_block_sync(&self, bid: Bid, frame: &VmFrame) -> Result<BioStatus, BioEnqueueError> {
+        let bio = create_bio_from_frame(BioType::Read, bid, frame);
+        let status = bio.submit_sync(self)?;
+        Ok(status)
+    }
+
+    /// Asynchronously reads one block indicated by the `bid`.
+    pub fn read_block(&self, bid: Bid, frame: &VmFrame) -> Result<BioWaiter, BioEnqueueError> {
+        let bio = create_bio_from_frame(BioType::Read, bid, frame);
+        bio.submit(self)
+    }
+
+    /// Synchronously writes contiguous blocks starting from the `bid`.
+    pub fn write_blocks_sync(
+        &self,
+        bid: Bid,
+        segment: &VmSegment,
+    ) -> Result<BioStatus, BioEnqueueError> {
+        let bio = create_bio_from_segment(BioType::Write, bid, segment);
+        let status = bio.submit_sync(self)?;
+        Ok(status)
+    }
+
+    /// Asynchronously writes contiguous blocks starting from the `bid`.
+    pub fn write_blocks(
+        &self,
+        bid: Bid,
+        segment: &VmSegment,
+    ) -> Result<BioWaiter, BioEnqueueError> {
+        let bio = create_bio_from_segment(BioType::Write, bid, segment);
+        bio.submit(self)
+    }
+
+    /// Synchronously writes one block indicated by the `bid`.
+    pub fn write_block_sync(
+        &self,
+        bid: Bid,
+        frame: &VmFrame,
+    ) -> Result<BioStatus, BioEnqueueError> {
+        let bio = create_bio_from_frame(BioType::Write, bid, frame);
+        let status = bio.submit_sync(self)?;
+        Ok(status)
+    }
+
+    /// Asynchronously writes one block indicated by the `bid`.
+    pub fn write_block(&self, bid: Bid, frame: &VmFrame) -> Result<BioWaiter, BioEnqueueError> {
+        let bio = create_bio_from_frame(BioType::Write, bid, frame);
+        bio.submit(self)
+    }
+}
+
+impl VmIo for dyn BlockDevice {
+    /// Reads consecutive bytes of several sectors in size.
+    fn read_bytes(&self, offset: usize, buf: &mut [u8]) -> aster_frame::Result<()> {
+        if offset % SECTOR_SIZE != 0 || buf.len() % SECTOR_SIZE != 0 {
+            return Err(aster_frame::Error::InvalidArgs);
+        }
+        if buf.is_empty() {
+            return Ok(());
+        }
+
+        let (bio, bio_segment) = {
+            let num_blocks = {
+                let first = Bid::from_offset(offset).to_raw();
+                let last = Bid::from_offset(offset + buf.len() - 1).to_raw();
+                last - first + 1
+            };
+            let segment = VmAllocOptions::new(num_blocks as usize)
+                .uninit(true)
+                .is_contiguous(true)
+                .alloc_contiguous()?;
+            let bio_segment = BioSegment::from_segment(segment, offset % BLOCK_SIZE, buf.len());
+
+            (
+                Bio::new(
+                    BioType::Read,
+                    Sid::from_offset(offset),
+                    vec![bio_segment.clone()],
+                    None,
+                ),
+                bio_segment,
+            )
+        };
+
+        let status = bio.submit_sync(self)?;
+        match status {
+            BioStatus::Complete => {
+                let _ = bio_segment.reader().read(&mut buf.into());
+                Ok(())
+            }
+            _ => Err(aster_frame::Error::IoError),
+        }
+    }
+
+    /// Writes consecutive bytes of several sectors in size.
+    fn write_bytes(&self, offset: usize, buf: &[u8]) -> aster_frame::Result<()> {
+        if offset % SECTOR_SIZE != 0 || buf.len() % SECTOR_SIZE != 0 {
+            return Err(aster_frame::Error::InvalidArgs);
+        }
+        if buf.is_empty() {
+            return Ok(());
+        }
+
+        let bio = {
+            let num_blocks = {
+                let first = Bid::from_offset(offset).to_raw();
+                let last = Bid::from_offset(offset + buf.len() - 1).to_raw();
+                last - first + 1
+            };
+            let segment = VmAllocOptions::new(num_blocks as usize)
+                .uninit(true)
+                .is_contiguous(true)
+                .alloc_contiguous()?;
+            segment.write_bytes(offset % BLOCK_SIZE, buf)?;
+            let len = segment
+                .writer()
+                .skip(offset % BLOCK_SIZE)
+                .write(&mut buf.into());
+            let bio_segment = BioSegment::from_segment(segment, offset % BLOCK_SIZE, len);
+            Bio::new(
+                BioType::Write,
+                Sid::from_offset(offset),
+                vec![bio_segment],
+                None,
+            )
+        };
+
+        let status = bio.submit_sync(self)?;
+        match status {
+            BioStatus::Complete => Ok(()),
+            _ => Err(aster_frame::Error::IoError),
+        }
+    }
+}
+
+impl dyn BlockDevice {
+    /// Asynchronously writes consecutive bytes of several sectors in size.
+    pub fn write_bytes_async(&self, offset: usize, buf: &[u8]) -> aster_frame::Result<BioWaiter> {
+        if offset % SECTOR_SIZE != 0 || buf.len() % SECTOR_SIZE != 0 {
+            return Err(aster_frame::Error::InvalidArgs);
+        }
+        if buf.is_empty() {
+            return Ok(BioWaiter::new());
+        }
+
+        let bio = {
+            let num_blocks = {
+                let first = Bid::from_offset(offset).to_raw();
+                let last = Bid::from_offset(offset + buf.len() - 1).to_raw();
+                last - first + 1
+            };
+            let segment = VmAllocOptions::new(num_blocks as usize)
+                .uninit(true)
+                .is_contiguous(true)
+                .alloc_contiguous()?;
+            segment.write_bytes(offset % BLOCK_SIZE, buf)?;
+            let len = segment
+                .writer()
+                .skip(offset % BLOCK_SIZE)
+                .write(&mut buf.into());
+            let bio_segment = BioSegment::from_segment(segment, offset % BLOCK_SIZE, len);
+            Bio::new(
+                BioType::Write,
+                Sid::from_offset(offset),
+                vec![bio_segment],
+                Some(general_complete_fn),
+            )
+        };
+
+        let complete = bio.submit(self)?;
+        Ok(complete)
+    }
+}
+
+// TODO: Maybe we should have a builder for `Bio`.
+fn create_bio_from_segment(type_: BioType, bid: Bid, segment: &VmSegment) -> Bio {
+    let bio_segment = BioSegment::from_segment(segment.clone(), 0, segment.nbytes());
+    Bio::new(
+        type_,
+        Sid::from(bid),
+        vec![bio_segment],
+        Some(general_complete_fn),
+    )
+}
+
+// TODO: Maybe we should have a builder for `Bio`.
+fn create_bio_from_frame(type_: BioType, bid: Bid, frame: &VmFrame) -> Bio {
+    let bio_segment = BioSegment::from_frame(frame.clone(), 0, BLOCK_SIZE);
+    Bio::new(
+        type_,
+        Sid::from(bid),
+        vec![bio_segment],
+        Some(general_complete_fn),
+    )
+}
+
+fn general_complete_fn(bio: &SubmittedBio) {
+    match bio.status() {
+        BioStatus::Complete => (),
+        err_status => log::error!(
+            "faild to do {:?} on the device with error status: {:?}",
+            bio.type_(),
+            err_status
+        ),
+    }
+}

services/comps/block/src/lib.rs

@@ -1,32 +1,67 @@
 //! The block devices of Asterinas.
+//！
+//！This crate provides a number of base components for block devices, including
+//! an abstraction of block devices, as well as the registration and lookup of block devices.
+//!
+//! Block devices use a queue-based model for asynchronous I/O operations. It is necessary
+//! for a block device to maintain a queue to handle I/O requests. The users (e.g., fs)
+//! submit I/O requests to this queue and wait for their completion. Drivers implementing
+//! block devices can create their own queues as needed, with the possibility to reorder
+//! and merge requests within the queue.
+//!
+//! This crate also offers the `Bio` related data structures and APIs to accomplish
+//! safe and convenient block I/O operations, for exmaple:
+//!
+//! ```no_run
+//! // Creates a bio request.
+//! let bio = Bio::new(BioType::Write, sid, segments, None);
+//! // Submits to the block device.
+//! let bio_waiter = bio.submit(block_device)?;
+//! // Waits for the the completion.
+//! let Some(status) = bio_waiter.wait() else {
+//!     return Err(IoError);
+//! };
+//! assert!(status == BioStatus::Complete);
+//! ```
+//!
 #![no_std]
 #![forbid(unsafe_code)]
 #![feature(fn_traits)]
+#![feature(step_trait)]
+#![feature(trait_upcasting)]
+#![allow(dead_code)]
 
 extern crate alloc;
 
-use core::any::Any;
-use core::fmt::Debug;
+pub mod bio;
+pub mod id;
+mod impl_block_device;
+mod prelude;
+pub mod request_queue;
+
+use self::{prelude::*, request_queue::BioRequestQueue};
 
-use alloc::collections::BTreeMap;
-use alloc::string::String;
-use alloc::sync::Arc;
-use alloc::vec::Vec;
 use aster_frame::sync::SpinLock;
-use aster_frame::vm::VmReader;
-use aster_frame::vm::VmWriter;
 use component::init_component;
 use component::ComponentInitError;
+
 use spin::Once;
 
-pub const BLK_SIZE: usize = 512;
+pub const BLOCK_SIZE: usize = aster_frame::config::PAGE_SIZE;
+pub const SECTOR_SIZE: usize = 512;
 
 pub trait BlockDevice: Send + Sync + Any + Debug {
-    fn read_block(&self, block_id: usize, buf: &[VmWriter]);
-    fn write_block(&self, block_id: usize, buf: &[VmReader]);
+    /// Returns this block device's request queue, to which block I/O requests may be submitted.
+    fn request_queue(&self) -> &dyn BioRequestQueue;
     fn handle_irq(&self);
 }
 
+impl dyn BlockDevice {
+    pub fn downcast_ref<T: BlockDevice>(&self) -> Option<&T> {
+        (self as &dyn Any).downcast_ref::<T>()
+    }
+}
+
 pub fn register_device(name: String, device: Arc<dyn BlockDevice>) {
     COMPONENT
         .get()

services/comps/block/src/prelude.rs

@@ -0,0 +1,9 @@
+pub(crate) use alloc::collections::{BTreeMap, VecDeque};
+pub(crate) use alloc::string::String;
+pub(crate) use alloc::sync::Arc;
+pub(crate) use alloc::vec;
+pub(crate) use alloc::vec::Vec;
+pub(crate) use core::any::Any;
+pub(crate) use core::fmt::Debug;
+pub(crate) use core::ops::Range;
+pub(crate) use core::sync::atomic::{AtomicU32, Ordering};

services/comps/block/src/request_queue.rs

@@ -0,0 +1,93 @@
+use crate::prelude::*;
+
+use super::{
+    bio::{BioEnqueueError, BioType, SubmittedBio},
+    id::Sid,
+};
+
+/// Represents the software staging queue for the `BioRequest` objects.
+pub trait BioRequestQueue {
+    /// Enqueues a `SubmittedBio` to this queue.
+    ///
+    /// This `SubmittedBio` will be merged into an existing `BioRequest`, or a new
+    /// `BioRequest` will be created from the `SubmittedBio` before being placed
+    /// into the queue.
+    ///
+    /// This method will wake up the waiter if a new `BioRequest` is enqueued.
+    fn enqueue(&self, bio: SubmittedBio) -> Result<(), BioEnqueueError>;
+
+    /// Dequeues a `BioRequest` from this queue.
+    ///
+    /// This method will wait until one request can be retrieved.
+    fn dequeue(&self) -> BioRequest;
+}
+
+/// The block I/O request.
+pub struct BioRequest {
+    /// The type of the I/O
+    type_: BioType,
+    /// The range of target sectors on the device
+    sid_range: Range<Sid>,
+    /// The submitted bios
+    bios: VecDeque<SubmittedBio>,
+}
+
+impl BioRequest {
+    /// Returns the type of the I/O.
+    pub fn type_(&self) -> BioType {
+        self.type_
+    }
+
+    /// Returns the range of sector id on device.
+    pub fn sid_range(&self) -> &Range<Sid> {
+        &self.sid_range
+    }
+
+    /// Returns an iterator to the `SubmittedBio`s.
+    pub fn bios(&self) -> impl Iterator<Item = &SubmittedBio> {
+        self.bios.iter()
+    }
+
+    /// Returns `true` if can merge the `SubmittedBio`, `false` otherwise.
+    pub fn can_merge(&self, rq_bio: &SubmittedBio) -> bool {
+        if rq_bio.type_() != self.type_ {
+            return false;
+        }
+
+        rq_bio.sid_range().start == self.sid_range.end
+            || rq_bio.sid_range().end == self.sid_range.start
+    }
+
+    /// Merges the `SubmittedBio` into this request.
+    ///
+    /// The merged `SubmittedBio` can only be placed at the front or back.
+    ///
+    /// # Panic
+    ///
+    /// If the `SubmittedBio` can not be merged, this method will panic.
+    pub fn merge_bio(&mut self, rq_bio: SubmittedBio) {
+        assert!(self.can_merge(&rq_bio));
+
+        if rq_bio.sid_range().start == self.sid_range.end {
+            self.sid_range.end = rq_bio.sid_range().end;
+            self.bios.push_back(rq_bio);
+        } else {
+            self.sid_range.start = rq_bio.sid_range().start;
+            self.bios.push_front(rq_bio);
+        }
+    }
+}
+
+impl From<SubmittedBio> for BioRequest {
+    fn from(bio: SubmittedBio) -> Self {
+        Self {
+            type_: bio.type_(),
+            sid_range: bio.sid_range().clone(),
+            bios: {
+                let mut bios = VecDeque::with_capacity(1);
+                bios.push_front(bio);
+                bios
+            },
+        }
+    }
+}

services/comps/virtio/src/device/block/device.rs

@@ -1,28 +1,132 @@
-use core::{hint::spin_loop, mem::size_of};
+use core::{
+    fmt::Debug,
+    hint::spin_loop,
+    mem::size_of,
+    sync::atomic::{AtomicUsize, Ordering},
+};
 
-use alloc::{boxed::Box, string::ToString, sync::Arc};
-use alloc::{boxed::Box, string::ToString, sync::Arc, vec::Vec};
+use alloc::{boxed::Box, collections::VecDeque, string::ToString, sync::Arc, vec::Vec};
+use aster_block::{
+    bio::{BioEnqueueError, BioStatus, BioType, SubmittedBio},
+    id::Sid,
+    request_queue::{BioRequest, BioRequestQueue},
+};
 use aster_frame::{
     io_mem::IoMem,
     sync::SpinLock,
+    sync::{Mutex, WaitQueue},
     trap::TrapFrame,
     vm::{VmAllocOptions, VmFrame, VmIo, VmReader, VmWriter},
 };
 use aster_util::safe_ptr::SafePtr;
 use log::info;
+use pod::Pod;
 
 use crate::{
-    device::block::{BlkReq, BlkResp, ReqType, RespStatus},
+    device::block::{ReqType, RespStatus},
     device::VirtioDeviceError,
     queue::VirtQueue,
     transport::VirtioTransport,
 };
 
-use super::{BlkFeatures, VirtioBlkConfig};
+use super::{BlockFeatures, VirtioBlockConfig};
 
 #[derive(Debug)]
 pub struct BlockDevice {
-    config: SafePtr<VirtioBlkConfig, IoMem>,
+    device: DeviceInner,
+    /// The software staging queue.
+    queue: BioSingleQueue,
+}
+
+impl BlockDevice {
+    /// Creates a new VirtIO-Block driver and registers it.
+    pub(crate) fn init(transport: Box<dyn VirtioTransport>) -> Result<(), VirtioDeviceError> {
+        let block_device = {
+            let device = DeviceInner::init(transport)?;
+            Self {
+                device,
+                queue: BioSingleQueue::new(),
+            }
+        };
+        aster_block::register_device(super::DEVICE_NAME.to_string(), Arc::new(block_device));
+        Ok(())
+    }
+
+    /// Dequeues a `BioRequest` from the software staging queue and
+    /// processes the request.
+    ///
+    /// TODO: Current read and write operations are still synchronous，
+    /// it needs to be modified to use the queue-based asynchronous programming pattern.
+    pub fn handle_requests(&self) {
+        let request = self.queue.dequeue();
+        match request.type_() {
+            BioType::Read => self.do_read(&request),
+            BioType::Write => self.do_write(&request),
+            BioType::Flush | BioType::Discard => todo!(),
+        }
+    }
+
+    fn do_read(&self, request: &BioRequest) {
+        let start_sid = request.sid_range().start;
+
+        let writers = {
+            let mut writers = Vec::new();
+            for bio in request.bios() {
+                for segment in bio.segments() {
+                    writers.push(segment.writer());
+                }
+            }
+            writers
+        };
+
+        self.device.read(start_sid, writers.as_slice());
+
+        for bio in request.bios() {
+            bio.complete(BioStatus::Complete);
+        }
+    }
+
+    fn do_write(&self, request: &BioRequest) {
+        let start_sid = request.sid_range().start;
+
+        let readers = {
+            let mut readers = Vec::new();
+            for bio in request.bios() {
+                for segment in bio.segments() {
+                    readers.push(segment.reader());
+                }
+            }
+            readers
+        };
+
+        self.device.write(start_sid, readers.as_slice());
+
+        for bio in request.bios() {
+            bio.complete(BioStatus::Complete);
+        }
+    }
+
+    /// Negotiate features for the device specified bits 0~23
+    pub(crate) fn negotiate_features(features: u64) -> u64 {
+        let feature = BlockFeatures::from_bits(features).unwrap();
+        let support_features = BlockFeatures::from_bits(features).unwrap();
+        (feature & support_features).bits
+    }
+}
+
+impl aster_block::BlockDevice for BlockDevice {
+    fn request_queue(&self) -> &dyn BioRequestQueue {
+        &self.queue
+    }
+
+    fn handle_irq(&self) {
+        info!("Virtio block device handle irq");
+    }
+}
+
+#[derive(Debug)]
+struct DeviceInner {
+    config: SafePtr<VirtioBlockConfig, IoMem>,
     queue: SpinLock<VirtQueue>,
     transport: Box<dyn VirtioTransport>,
     /// Block requests, we use VmFrame to store the requests so that
@@ -34,109 +138,10 @@ pub struct BlockDevice {
     id_allocator: SpinLock<Vec<u8>>,
 }
 
-impl BlockDevice {
-    /// read data from block device, this function is blocking
-    /// FIEME: replace slice with a more secure data structure to use dma mapping.
-    pub fn read(&self, block_id: usize, buf: &[VmWriter]) {
-        // FIXME: Handling cases without id.
-        let id = self.id_allocator.lock().pop().unwrap() as usize;
-        let req = BlkReq {
-            type_: ReqType::In as _,
-            reserved: 0,
-            sector: block_id as u64,
-        };
-        let resp = BlkResp::default();
-        self.block_requests
-            .write_val(id * size_of::<BlkReq>(), &req)
-            .unwrap();
-        self.block_responses
-            .write_val(id * size_of::<BlkResp>(), &resp)
-            .unwrap();
-        let req = self
-            .block_requests
-            .reader()
-            .skip(id * size_of::<BlkReq>())
-            .limit(size_of::<BlkReq>());
-        let resp = self
-            .block_responses
-            .writer()
-            .skip(id * size_of::<BlkResp>())
-            .limit(size_of::<BlkResp>());
-
-        let mut outputs: Vec<&VmWriter<'_>> = buf.iter().collect();
-        outputs.push(&resp);
-        let mut queue = self.queue.lock_irq_disabled();
-        let token = queue
-            .add_vm(&[&req], outputs.as_slice())
-            .expect("add queue failed");
-        queue.notify();
-        while !queue.can_pop() {
-            spin_loop();
-        }
-        queue.pop_used_with_token(token).expect("pop used failed");
-        let resp: BlkResp = self
-            .block_responses
-            .read_val(id * size_of::<BlkResp>())
-            .unwrap();
-        self.id_allocator.lock().push(id as u8);
-        match RespStatus::try_from(resp.status).unwrap() {
-            RespStatus::Ok => {}
-            _ => panic!("io error in block device"),
-        };
-    }
-    /// write data to block device, this function is blocking
-    /// FIEME: replace slice with a more secure data structure to use dma mapping.
-    pub fn write(&self, block_id: usize, buf: &[VmReader]) {
-        // FIXME: Handling cases without id.
-        let id = self.id_allocator.lock().pop().unwrap() as usize;
-        let req = BlkReq {
-            type_: ReqType::Out as _,
-            reserved: 0,
-            sector: block_id as u64,
-        };
-        let resp = BlkResp::default();
-        self.block_requests
-            .write_val(id * size_of::<BlkReq>(), &req)
-            .unwrap();
-        self.block_responses
-            .write_val(id * size_of::<BlkResp>(), &resp)
-            .unwrap();
-        let req = self
-            .block_requests
-            .reader()
-            .skip(id * size_of::<BlkReq>())
-            .limit(size_of::<BlkReq>());
-        let resp = self
-            .block_responses
-            .writer()
-            .skip(id * size_of::<BlkResp>())
-            .limit(size_of::<BlkResp>());
-
-        let mut queue = self.queue.lock_irq_disabled();
-        let mut inputs: Vec<&VmReader<'_>> = buf.iter().collect();
-        inputs.insert(0, &req);
-        let token = queue
-            .add_vm(inputs.as_slice(), &[&resp])
-            .expect("add queue failed");
-        queue.notify();
-        while !queue.can_pop() {
-            spin_loop();
-        }
-        queue.pop_used_with_token(token).expect("pop used failed");
-        let resp: BlkResp = self
-            .block_responses
-            .read_val(id * size_of::<BlkResp>())
-            .unwrap();
-        self.id_allocator.lock().push(id as u8);
-        match RespStatus::try_from(resp.status).unwrap() {
-            RespStatus::Ok => {}
-            _ => panic!("io error in block device:{:?}", resp.status),
-        };
-    }
-
-    /// Create a new VirtIO-Block driver.
-    pub(crate) fn init(mut transport: Box<dyn VirtioTransport>) -> Result<(), VirtioDeviceError> {
-        let config = VirtioBlkConfig::new(transport.as_mut());
+impl DeviceInner {
+    /// Creates and inits the device.
+    pub fn init(mut transport: Box<dyn VirtioTransport>) -> Result<Self, VirtioDeviceError> {
+        let config = VirtioBlockConfig::new(transport.as_mut());
         let num_queues = transport.num_queues();
         if num_queues != 1 {
             return Err(VirtioDeviceError::QueuesAmountDoNotMatch(num_queues, 1));
@@ -170,30 +175,225 @@ impl BlockDevice {
             info!("Virtio block device config space change");
         }
         device.transport.finish_init();
+        Ok(device)
+    }
 
-        aster_block::register_device(super::DEVICE_NAME.to_string(), Arc::new(device));
+    /// Reads data from the block device, this function is blocking.
+    /// FIEME: replace slice with a more secure data structure to use dma mapping.
+    pub fn read(&self, sector_id: Sid, buf: &[VmWriter]) {
+        // FIXME: Handling cases without id.
+        let id = self.id_allocator.lock().pop().unwrap() as usize;
+        let req = BlockReq {
+            type_: ReqType::In as _,
+            reserved: 0,
+            sector: sector_id.to_raw(),
+        };
+        let resp = BlockResp::default();
+        self.block_requests
+            .write_val(id * size_of::<BlockReq>(), &req)
+            .unwrap();
+        self.block_responses
+            .write_val(id * size_of::<BlockResp>(), &resp)
+            .unwrap();
+        let req_reader = self
+            .block_requests
+            .reader()
+            .skip(id * size_of::<BlockReq>())
+            .limit(size_of::<BlockReq>());
+        let resp_writer = self
+            .block_responses
+            .writer()
+            .skip(id * size_of::<BlockResp>())
+            .limit(size_of::<BlockResp>());
 
+        let mut outputs: Vec<&VmWriter<'_>> = buf.iter().collect();
+        outputs.push(&resp_writer);
+        let mut queue = self.queue.lock_irq_disabled();
+        let token = queue
+            .add_vm(&[&req_reader], outputs.as_slice())
+            .expect("add queue failed");
+        queue.notify();
+        while !queue.can_pop() {
+            spin_loop();
+        }
+        queue.pop_used_with_token(token).expect("pop used failed");
+        let resp: BlockResp = self
+            .block_responses
+            .read_val(id * size_of::<BlockResp>())
+            .unwrap();
+        self.id_allocator.lock().push(id as u8);
+        match RespStatus::try_from(resp.status).unwrap() {
+            RespStatus::Ok => {}
+            _ => panic!("io error in block device"),
+        };
+    }
+
+    /// Writes data to the block device, this function is blocking.
+    /// FIEME: replace slice with a more secure data structure to use dma mapping.
+    pub fn write(&self, sector_id: Sid, buf: &[VmReader]) {
+        // FIXME: Handling cases without id.
+        let id = self.id_allocator.lock().pop().unwrap() as usize;
+        let req = BlockReq {
+            type_: ReqType::Out as _,
+            reserved: 0,
+            sector: sector_id.to_raw(),
+        };
+        let resp = BlockResp::default();
+        self.block_requests
+            .write_val(id * size_of::<BlockReq>(), &req)
+            .unwrap();
+        self.block_responses
+            .write_val(id * size_of::<BlockResp>(), &resp)
+            .unwrap();
+        let req_reader = self
+            .block_requests
+            .reader()
+            .skip(id * size_of::<BlockReq>())
+            .limit(size_of::<BlockReq>());
+        let resp_writer = self
+            .block_responses
+            .writer()
+            .skip(id * size_of::<BlockResp>())
+            .limit(size_of::<BlockResp>());
+
+        let mut queue = self.queue.lock_irq_disabled();
+        let mut inputs: Vec<&VmReader<'_>> = buf.iter().collect();
+        inputs.insert(0, &req_reader);
+        let token = queue
+            .add_vm(inputs.as_slice(), &[&resp_writer])
+            .expect("add queue failed");
+        queue.notify();
+        while !queue.can_pop() {
+            spin_loop();
+        }
+        queue.pop_used_with_token(token).expect("pop used failed");
+        let resp: BlockResp = self
+            .block_responses
+            .read_val(id * size_of::<BlockResp>())
+            .unwrap();
+        self.id_allocator.lock().push(id as u8);
+        match RespStatus::try_from(resp.status).unwrap() {
+            RespStatus::Ok => {}
+            _ => panic!("io error in block device:{:?}", resp.status),
+        };
+    }
+}
+
+#[repr(C)]
+#[derive(Debug, Copy, Clone, Pod)]
+struct BlockReq {
+    pub type_: u32,
+    pub reserved: u32,
+    pub sector: u64,
+}
+
+/// Response of a VirtIOBlock request.
+#[repr(C)]
+#[derive(Debug, Copy, Clone, Pod)]
+struct BlockResp {
+    pub status: u8,
+}
+
+impl Default for BlockResp {
+    fn default() -> Self {
+        Self {
+            status: RespStatus::_NotReady as _,
+        }
+    }
+}
+
+/// A simple block I/O request queue with a single queue.
+///
+/// It is a FIFO producer-consumer queue, where the producer (e.g., filesystem)
+/// submits requests to the queue, and the consumer (e.g., block device driver)
+/// continuously consumes and processes these requests from the queue.
+pub struct BioSingleQueue {
+    queue: Mutex<VecDeque<BioRequest>>,
+    num_requests: AtomicUsize,
+    wait_queue: WaitQueue,
+}
+
+impl BioSingleQueue {
+    /// Creates an empty queue.
+    pub fn new() -> Self {
+        Self {
+            queue: Mutex::new(VecDeque::new()),
+            num_requests: AtomicUsize::new(0),
+            wait_queue: WaitQueue::new(),
+        }
+    }
+
+    /// Returns the number of requests currently in this queue.
+    pub fn num_requests(&self) -> usize {
+        self.num_requests.load(Ordering::Relaxed)
+    }
+
+    fn dec_num_requests(&self) {
+        self.num_requests.fetch_sub(1, Ordering::Relaxed);
+    }
+
+    fn inc_num_requests(&self) {
+        self.num_requests.fetch_add(1, Ordering::Relaxed);
+    }
+}
+
+impl Default for BioSingleQueue {
+    fn default() -> Self {
+        Self::new()
+    }
+}
+
+impl Debug for BioSingleQueue {
+    fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result {
+        f.debug_struct("BioSingleQueue")
+            .field("num_requests", &self.num_requests())
+            .finish()
+    }
+}
+
+impl BioRequestQueue for BioSingleQueue {
+    /// Enqueues a `SubmittedBio` to this queue.
+    ///
+    /// When enqueueing the `SubmittedBio`, try to insert it into the last request if the
+    /// type is same and the sector range is contiguous.
+    /// Otherwise, creates and inserts a new request for the `SubmittedBio`.
+    fn enqueue(&self, bio: SubmittedBio) -> Result<(), BioEnqueueError> {
+        let mut queue = self.queue.lock();
+        if let Some(request) = queue.front_mut() {
+            if request.can_merge(&bio) {
+                request.merge_bio(bio);
+                return Ok(());
+            }
+        }
+
+        let new_request = BioRequest::from(bio);
+        queue.push_front(new_request);
+        drop(queue);
+        self.inc_num_requests();
+        self.wait_queue.wake_all();
         Ok(())
     }
 
-    /// Negotiate features for the device specified bits 0~23
-    pub(crate) fn negotiate_features(features: u64) -> u64 {
-        let feature = BlkFeatures::from_bits(features).unwrap();
-        let support_features = BlkFeatures::from_bits(features).unwrap();
-        (feature & support_features).bits
+    /// Dequeues a `BioRequest` from this queue.
+    fn dequeue(&self) -> BioRequest {
+        let mut num_requests = self.num_requests();
+
+        loop {
+            if num_requests > 0 {
+                if let Some(request) = self.queue.lock().pop_back() {
+                    self.dec_num_requests();
+                    return request;
                 }
             }
 
-impl aster_block::BlockDevice for BlockDevice {
-    fn read_block(&self, block_id: usize, buf: &[VmWriter]) {
-        self.read(block_id, buf);
+            num_requests = self.wait_queue.wait_until(|| {
+                let num_requests = self.num_requests();
+                if num_requests > 0 {
+                    Some(num_requests)
+                } else {
+                    None
+                }
+            });
         }
-
-    fn write_block(&self, block_id: usize, buf: &[VmReader]) {
-        self.write(block_id, buf);
-    }
-
-    fn handle_irq(&self) {
-        info!("Virtio block device handle irq");
     }
 }

services/comps/virtio/src/device/block/mod.rs

@@ -8,12 +8,11 @@ use pod::Pod;
 
 use crate::transport::VirtioTransport;
 
-pub const BLK_SIZE: usize = 512;
 pub static DEVICE_NAME: &str = "Virtio-Block";
 
 bitflags! {
     /// features for virtio block device
-    pub(crate) struct BlkFeatures : u64{
+    pub(crate) struct BlockFeatures : u64 {
         const BARRIER       = 1 << 0;
         const SIZE_MAX      = 1 << 1;
         const SEG_MAX       = 1 << 2;
@@ -29,29 +28,6 @@ bitflags! {
     }
 }
 
-#[repr(C)]
-#[derive(Debug, Copy, Clone, Pod)]
-pub struct BlkReq {
-    pub type_: u32,
-    pub reserved: u32,
-    pub sector: u64,
-}
-
-/// Response of a VirtIOBlk request.
-#[repr(C)]
-#[derive(Debug, Copy, Clone, Pod)]
-pub struct BlkResp {
-    pub status: u8,
-}
-
-impl Default for BlkResp {
-    fn default() -> Self {
-        BlkResp {
-            status: RespStatus::_NotReady as _,
-        }
-    }
-}
-
 #[repr(u32)]
 #[derive(Debug, Copy, Clone, TryFromInt)]
 pub enum ReqType {
@@ -77,12 +53,12 @@ pub enum RespStatus {
 
 #[derive(Debug, Copy, Clone, Pod)]
 #[repr(C)]
-pub struct VirtioBlkConfig {
+pub struct VirtioBlockConfig {
     capacity: u64,
     size_max: u64,
-    geometry: VirtioBlkGeometry,
+    geometry: VirtioBlockGeometry,
     blk_size: u32,
-    topology: VirtioBlkTopology,
+    topology: VirtioBlockTopology,
     writeback: u8,
     unused0: [u8; 3],
     max_discard_sectors: u32,
@@ -96,7 +72,7 @@ pub struct VirtioBlkConfig {
 
 #[derive(Debug, Copy, Clone, Pod)]
 #[repr(C)]
-pub struct VirtioBlkGeometry {
+pub struct VirtioBlockGeometry {
     cylinders: u16,
     heads: u8,
     sectors: u8,
@@ -104,14 +80,14 @@ pub struct VirtioBlkGeometry {
 
 #[derive(Debug, Copy, Clone, Pod)]
 #[repr(C)]
-pub struct VirtioBlkTopology {
+pub struct VirtioBlockTopology {
     physical_block_exp: u8,
     alignment_offset: u8,
     min_io_size: u16,
     opt_io_size: u32,
 }
 
-impl VirtioBlkConfig {
+impl VirtioBlockConfig {
     pub(self) fn new(transport: &dyn VirtioTransport) -> SafePtr<Self, IoMem> {
         let memory = transport.device_config_memory();
         SafePtr::new(memory, 0)

services/libs/aster-std/Cargo.toml

@@ -64,6 +64,9 @@ log = "0.4"
 getrandom = { version = "0.2.10", default-features = false, features = [
     "rdrand",
 ] }
+bitvec = { version = "1.0", default-features = false, features = ["alloc"] }
+static_assertions = "1.1.0"
+inherit-methods-macro = { git = "https://github.com/asterinas/inherit-methods-macro", rev = "98f7e3e" }
 
 [dependencies.lazy_static]
 version = "1.0"

services/libs/aster-std/src/driver/mod.rs

@@ -1,10 +1,3 @@
-use core::mem::size_of;
-
-use alloc::vec::Vec;
-use aster_frame::{
-    println,
-    vm::{VmAllocOptions, VmIo},
-};
 use log::info;
 
 pub fn init() {
@@ -13,43 +6,3 @@ pub fn init() {
         info!("Found Input device, name:{}", name);
     }
 }
-
-#[allow(unused)]
-fn block_device_test() {
-    for (_, device) in aster_block::all_devices() {
-        let write_frame = VmAllocOptions::new(1).alloc_single().unwrap();
-        let read_frame = VmAllocOptions::new(1).alloc_single().unwrap();
-        info!("write_buffer address:{:x}", write_frame.start_paddr());
-        info!("read_buffer address:{:x}", read_frame.start_paddr());
-
-        // init write frame
-        for i in 0..=8 {
-            let slice: [u8; 512] = [i; 512];
-            write_frame.write_slice(i as usize * 512, &slice);
-        }
-
-        // Test multiple Writer & Reader
-        let mut writers = Vec::with_capacity(8);
-        for i in 0..8 {
-            let writer = read_frame.writer().skip(i * 512).limit(512);
-            writers.push(writer);
-        }
-
-        let mut readers = Vec::with_capacity(8);
-        for i in 0..8 {
-            let reader = write_frame.reader().skip(i * 512).limit(512);
-            readers.push(reader);
-        }
-
-        device.write_block(0, readers.as_slice());
-        device.read_block(0, writers.as_slice());
-        let mut read_slice = [0u8; 512];
-        let mut write_slice = [0u8; 512];
-        for i in 0..8 {
-            read_frame.read_bytes(i * size_of::<[u8; 512]>(), &mut read_slice);
-            write_frame.read_bytes(i * size_of::<[u8; 512]>(), &mut write_slice);
-            assert_eq!(read_slice, write_slice);
-        }
-        println!("block device test passed!");
-    }
-}

services/libs/aster-std/src/error.rs

@@ -191,12 +191,48 @@ impl From<aster_frame::Error> for Error {
     }
 }
 
+impl From<aster_block::bio::BioEnqueueError> for Error {
+    fn from(error: aster_block::bio::BioEnqueueError) -> Self {
+        match error {
+            aster_block::bio::BioEnqueueError::IsFull => {
+                Error::with_message(Errno::EBUSY, "The request queue is full")
+            }
+            aster_block::bio::BioEnqueueError::Refused => {
+                Error::with_message(Errno::EBUSY, "Refuse to enqueue the bio")
+            }
+        }
+    }
+}
+
+impl From<aster_block::bio::BioStatus> for Error {
+    fn from(err_status: aster_block::bio::BioStatus) -> Self {
+        match err_status {
+            aster_block::bio::BioStatus::NotSupported => {
+                Error::with_message(Errno::EIO, "I/O operation is not supported")
+            }
+            aster_block::bio::BioStatus::NoSpace => {
+                Error::with_message(Errno::ENOSPC, "Insufficient space on device")
+            }
+            aster_block::bio::BioStatus::IoError => {
+                Error::with_message(Errno::EIO, "I/O operation fails")
+            }
+            status => panic!("Can not convert the status: {:?} to an error", status),
+        }
+    }
+}
+
 impl From<core::str::Utf8Error> for Error {
     fn from(_: core::str::Utf8Error) -> Self {
         Error::with_message(Errno::EINVAL, "Invalid utf-8 string")
     }
 }
 
+impl From<alloc::string::FromUtf8Error> for Error {
+    fn from(_: alloc::string::FromUtf8Error) -> Self {
+        Error::with_message(Errno::EINVAL, "Invalid utf-8 string")
+    }
+}
+
 impl From<core::ffi::FromBytesUntilNulError> for Error {
     fn from(_: core::ffi::FromBytesUntilNulError) -> Self {
         Error::with_message(Errno::E2BIG, "Cannot find null in cstring")

services/libs/aster-std/src/fs/devpts/mod.rs

@@ -6,7 +6,6 @@ use crate::fs::utils::{
 };
 use crate::prelude::*;
 
-use aster_frame::vm::VmFrame;
 use aster_util::{id_allocator::IdAlloc, slot_vec::SlotVec};
 use core::time::Duration;
 
@@ -140,12 +139,18 @@ impl Inode for RootInode {
         self.metadata.size
     }
 
-    fn resize(&self, new_size: usize) {}
+    fn resize(&self, new_size: usize) -> Result<()> {
+        Err(Error::new(Errno::EISDIR))
+    }
 
     fn metadata(&self) -> Metadata {
         self.metadata.clone()
     }
 
+    fn ino(&self) -> u64 {
+        self.metadata.ino as _
+    }
+
     fn type_(&self) -> InodeType {
         self.metadata.type_
     }

services/libs/aster-std/src/fs/devpts/ptmx.rs

@@ -65,12 +65,18 @@ impl Inode for Ptmx {
         self.metadata.size
     }
 
-    fn resize(&self, new_size: usize) {}
+    fn resize(&self, new_size: usize) -> Result<()> {
+        Ok(())
+    }
 
     fn metadata(&self) -> Metadata {
         self.metadata.clone()
     }
 
+    fn ino(&self) -> u64 {
+        self.metadata.ino as _
+    }
+
     fn type_(&self) -> InodeType {
         self.metadata.type_
     }
@@ -93,14 +99,6 @@ impl Inode for Ptmx {
 
     fn set_mtime(&self, time: Duration) {}
 
-    fn read_page(&self, idx: usize, frame: &VmFrame) -> Result<()> {
-        Ok(())
-    }
-
-    fn write_page(&self, idx: usize, frame: &VmFrame) -> Result<()> {
-        Ok(())
-    }
-
     fn read_at(&self, offset: usize, buf: &mut [u8]) -> Result<usize> {
         Ok(0)
     }

services/libs/aster-std/src/fs/devpts/slave.rs

@@ -45,12 +45,18 @@ impl Inode for PtySlaveInode {
         self.metadata.size
     }
 
-    fn resize(&self, new_size: usize) {}
+    fn resize(&self, new_size: usize) -> Result<()> {
+        Err(Error::new(Errno::EPERM))
+    }
 
     fn metadata(&self) -> Metadata {
         self.metadata.clone()
     }
 
+    fn ino(&self) -> u64 {
+        self.metadata.ino as _
+    }
+
     fn type_(&self) -> InodeType {
         self.metadata.type_
     }
@@ -73,14 +79,6 @@ impl Inode for PtySlaveInode {
 
     fn set_mtime(&self, time: Duration) {}
 
-    fn read_page(&self, idx: usize, frame: &VmFrame) -> Result<()> {
-        Ok(())
-    }
-
-    fn write_page(&self, idx: usize, frame: &VmFrame) -> Result<()> {
-        Ok(())
-    }
-
     fn read_at(&self, offset: usize, buf: &mut [u8]) -> Result<usize> {
         self.device.read(buf)
     }

services/libs/aster-std/src/fs/ext2/block_group.rs

@@ -0,0 +1,476 @@
+use super::fs::Ext2;
+use super::inode::{Inode, InodeDesc, RawInode};
+use super::prelude::*;
+use super::super_block::SuperBlock;
+
+use aster_util::id_allocator::IdAlloc;
+
+/// Blocks are clustered into block groups in order to reduce fragmentation and minimise
+/// the amount of head seeking when reading a large amount of consecutive data.
+pub(super) struct BlockGroup {
+    idx: usize,
+    bg_impl: Arc<BlockGroupImpl>,
+    raw_inodes_cache: PageCache,
+}
+
+struct BlockGroupImpl {
+    inode_table_bid: Bid,
+    raw_inodes_size: usize,
+    inner: RwMutex<Inner>,
+    fs: Weak<Ext2>,
+}
+
+impl BlockGroup {
+    /// Loads and constructs a block group.
+    pub fn load(
+        group_descriptors_segment: &VmSegment,
+        idx: usize,
+        block_device: &dyn BlockDevice,
+        super_block: &SuperBlock,
+        fs: Weak<Ext2>,
+    ) -> Result<Self> {
+        let raw_inodes_size = (super_block.inodes_per_group() as usize) * super_block.inode_size();
+
+        let bg_impl = {
+            let metadata = {
+                let descriptor = {
+                    // Read the block group descriptor
+                    // TODO: if the main is corrupted, should we load the backup?
+                    let offset = idx * core::mem::size_of::<RawGroupDescriptor>();
+                    let raw_descriptor = group_descriptors_segment
+                        .read_val::<RawGroupDescriptor>(offset)
+                        .unwrap();
+                    GroupDescriptor::from(raw_descriptor)
+                };
+
+                let get_bitmap = |bid: Bid, capacity: usize| -> Result<IdAlloc> {
+                    if capacity > BLOCK_SIZE * 8 {
+                        return_errno_with_message!(Errno::EINVAL, "bad bitmap");
+                    }
+                    let mut buf = vec![0u8; BLOCK_SIZE];
+                    block_device.read_bytes(bid.to_offset(), &mut buf)?;
+                    Ok(IdAlloc::from_bytes_with_capacity(&buf, capacity))
+                };
+
+                let block_bitmap = get_bitmap(
+                    descriptor.block_bitmap_bid,
+                    super_block.blocks_per_group() as usize,
+                )?;
+                let inode_bitmap = get_bitmap(
+                    descriptor.inode_bitmap_bid,
+                    super_block.inodes_per_group() as usize,
+                )?;
+
+                GroupMetadata {
+                    descriptor,
+                    block_bitmap,
+                    inode_bitmap,
+                }
+            };
+
+            Arc::new(BlockGroupImpl {
+                inode_table_bid: metadata.descriptor.inode_table_bid,
+                raw_inodes_size,
+                inner: RwMutex::new(Inner {
+                    metadata: Dirty::new(metadata),
+                    inode_cache: BTreeMap::new(),
+                }),
+                fs,
+            })
+        };
+
+        let raw_inodes_cache =
+            PageCache::with_capacity(raw_inodes_size, Arc::downgrade(&bg_impl) as _)?;
+
+        Ok(Self {
+            idx,
+            bg_impl,
+            raw_inodes_cache,
+        })
+    }
+
+    /// Finds and returns the inode.
+    pub fn lookup_inode(&self, inode_idx: u32) -> Result<Arc<Inode>> {
+        // The fast path
+        let inner = self.bg_impl.inner.read();
+        if !inner.metadata.is_inode_allocated(inode_idx) {
+            return_errno!(Errno::ENOENT);
+        }
+        if let Some(inode) = inner.inode_cache.get(&inode_idx) {
+            return Ok(inode.clone());
+        }
+
+        // The slow path
+        drop(inner);
+        let mut inner = self.bg_impl.inner.write();
+        if !inner.metadata.is_inode_allocated(inode_idx) {
+            return_errno!(Errno::ENOENT);
+        }
+        if let Some(inode) = inner.inode_cache.get(&inode_idx) {
+            return Ok(inode.clone());
+        }
+
+        // Loads the inode, then inserts it into the inode cache.
+        let inode = self.load_inode(inode_idx)?;
+        inner.inode_cache.insert(inode_idx, inode.clone());
+        Ok(inode)
+    }
+
+    /// Loads an existing inode.
+    ///
+    /// This method may load the raw inode metadata from block device.
+    fn load_inode(&self, inode_idx: u32) -> Result<Arc<Inode>> {
+        let fs = self.fs();
+        let raw_inode = {
+            let offset = (inode_idx as usize) * fs.inode_size();
+            self.raw_inodes_cache
+                .pages()
+                .read_val::<RawInode>(offset)
+                .unwrap()
+        };
+        let inode_desc = Dirty::new(InodeDesc::try_from(raw_inode)?);
+        let ino = inode_idx + self.idx as u32 * fs.inodes_per_group() + 1;
+
+        Ok(Inode::new(ino, self.idx, inode_desc, Arc::downgrade(&fs)))
+    }
+
+    /// Inserts the inode into the inode cache.
+    ///
+    /// # Panic
+    ///
+    /// If `inode_idx` has not been allocated before, then the method panics.
+    pub fn insert_cache(&self, inode_idx: u32, inode: Arc<Inode>) {
+        let mut inner = self.bg_impl.inner.write();
+        assert!(inner.metadata.is_inode_allocated(inode_idx));
+        inner.inode_cache.insert(inode_idx, inode);
+    }
+
+    /// Allocates and returns an inode index.
+    pub fn alloc_inode(&self, is_dir: bool) -> Option<u32> {
+        // The fast path
+        if self.bg_impl.inner.read().metadata.free_inodes_count() == 0 {
+            return None;
+        }
+
+        // The slow path
+        self.bg_impl.inner.write().metadata.alloc_inode(is_dir)
+    }
+
+    /// Frees the allocated inode idx.
+    ///
+    /// # Panic
+    ///
+    /// If `inode_idx` has not been allocated before, then the method panics.
+    pub fn free_inode(&self, inode_idx: u32, is_dir: bool) {
+        let mut inner = self.bg_impl.inner.write();
+        assert!(inner.metadata.is_inode_allocated(inode_idx));
+
+        inner.metadata.free_inode(inode_idx, is_dir);
+        inner.inode_cache.remove(&inode_idx);
+    }
+
+    /// Allocates and returns a block index.
+    pub fn alloc_block(&self) -> Option<u32> {
+        // The fast path
+        if self.bg_impl.inner.read().metadata.free_blocks_count() == 0 {
+            return None;
+        }
+
+        // The slow path
+        self.bg_impl.inner.write().metadata.alloc_block()
+    }
+
+    /// Frees the allocated block idx.
+    ///
+    /// # Panic
+    ///
+    /// If `block_idx` has not been allocated before, then the method panics.
+    pub fn free_block(&self, block_idx: u32) {
+        let mut inner = self.bg_impl.inner.write();
+        assert!(inner.metadata.is_block_allocated(block_idx));
+
+        inner.metadata.free_block(block_idx);
+    }
+
+    /// Writes back the raw inode metadata to the raw inode metadata cache.
+    pub fn sync_raw_inode(&self, inode_idx: u32, raw_inode: &RawInode) {
+        let offset = (inode_idx as usize) * self.fs().inode_size();
+        self.raw_inodes_cache
+            .pages()
+            .write_val(offset, raw_inode)
+            .unwrap();
+    }
+
+    /// Writes back the metadata of this group.
+    pub fn sync_metadata(&self, super_block: &SuperBlock) -> Result<()> {
+        if !self.bg_impl.inner.read().metadata.is_dirty() {
+            return Ok(());
+        }
+
+        let mut inner = self.bg_impl.inner.write();
+        let fs = self.fs();
+        // Writes back the descriptor.
+        let raw_descriptor = RawGroupDescriptor::from(&inner.metadata.descriptor);
+        self.fs().sync_group_descriptor(self.idx, &raw_descriptor)?;
+
+        let mut bio_waiter = BioWaiter::new();
+        // Writes back the inode bitmap.
+        let inode_bitmap_bid = inner.metadata.descriptor.inode_bitmap_bid;
+        bio_waiter.concat(fs.block_device().write_bytes_async(
+            inode_bitmap_bid.to_offset(),
+            inner.metadata.inode_bitmap.as_bytes(),
+        )?);
+
+        // Writes back the block bitmap.
+        let block_bitmap_bid = inner.metadata.descriptor.block_bitmap_bid;
+        bio_waiter.concat(fs.block_device().write_bytes_async(
+            block_bitmap_bid.to_offset(),
+            inner.metadata.block_bitmap.as_bytes(),
+        )?);
+
+        // Waits for the completion of all submitted bios.
+        bio_waiter.wait().ok_or_else(|| {
+            Error::with_message(Errno::EIO, "failed to sync metadata of block group")
+        })?;
+
+        inner.metadata.clear_dirty();
+        Ok(())
+    }
+
+    /// Writes back all of the cached inodes.
+    ///
+    /// The `sync_all` method of inode may modify the data of this block group,
+    /// so we should not hold the lock while syncing the inodes.
+    pub fn sync_all_inodes(&self) -> Result<()> {
+        // Removes the inodes that is unused from the inode cache.
+        let unused_inodes: Vec<Arc<Inode>> = self
+            .bg_impl
+            .inner
+            .write()
+            .inode_cache
+            .extract_if(|_, inode| Arc::strong_count(inode) == 1)
+            .map(|(_, inode)| inode)
+            .collect();
+
+        // Writes back the unused inodes.
+        for inode in unused_inodes.iter() {
+            inode.sync_all()?;
+        }
+        drop(unused_inodes);
+
+        // Writes back the remaining inodes in the inode cache.
+        let remaining_inodes: Vec<Arc<Inode>> = self
+            .bg_impl
+            .inner
+            .read()
+            .inode_cache
+            .values()
+            .cloned()
+            .collect();
+        for inode in remaining_inodes.iter() {
+            inode.sync_all()?;
+        }
+        drop(remaining_inodes);
+
+        // Writes back the raw inode metadata.
+        self.raw_inodes_cache
+            .pages()
+            .decommit(0..self.bg_impl.raw_inodes_size)?;
+        Ok(())
+    }
+
+    fn fs(&self) -> Arc<Ext2> {
+        self.bg_impl.fs.upgrade().unwrap()
+    }
+}
+
+impl Debug for BlockGroup {
+    fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result {
+        f.debug_struct("BlockGroup")
+            .field("idx", &self.idx)
+            .field("descriptor", &self.bg_impl.inner.read().metadata.descriptor)
+            .field(
+                "block_bitmap",
+                &self.bg_impl.inner.read().metadata.block_bitmap,
+            )
+            .field(
+                "inode_bitmap",
+                &self.bg_impl.inner.read().metadata.inode_bitmap,
+            )
+            .finish()
+    }
+}
+
+impl PageCacheBackend for BlockGroupImpl {
+    fn read_page(&self, idx: usize, frame: &VmFrame) -> Result<()> {
+        let bid = self.inode_table_bid + idx as u64;
+        self.fs.upgrade().unwrap().read_block(bid, frame)?;
+        Ok(())
+    }
+
+    fn write_page(&self, idx: usize, frame: &VmFrame) -> Result<()> {
+        let bid = self.inode_table_bid + idx as u64;
+        self.fs.upgrade().unwrap().write_block(bid, frame)?;
+        Ok(())
+    }
+
+    fn npages(&self) -> usize {
+        self.raw_inodes_size.div_ceil(BLOCK_SIZE)
+    }
+}
+
+#[derive(Debug)]
+struct Inner {
+    metadata: Dirty<GroupMetadata>,
+    inode_cache: BTreeMap<u32, Arc<Inode>>,
+}
+
+#[derive(Clone, Debug)]
+struct GroupMetadata {
+    descriptor: GroupDescriptor,
+    block_bitmap: IdAlloc,
+    inode_bitmap: IdAlloc,
+}
+
+impl GroupMetadata {
+    pub fn is_inode_allocated(&self, inode_idx: u32) -> bool {
+        self.inode_bitmap.is_allocated(inode_idx as usize)
+    }
+
+    pub fn alloc_inode(&mut self, is_dir: bool) -> Option<u32> {
+        let Some(inode_idx) = self.inode_bitmap.alloc() else {
+            return None;
+        };
+        self.dec_free_inodes();
+        if is_dir {
+            self.inc_dirs();
+        }
+        Some(inode_idx as u32)
+    }
+
+    pub fn free_inode(&mut self, inode_idx: u32, is_dir: bool) {
+        self.inode_bitmap.free(inode_idx as usize);
+        self.inc_free_inodes();
+        if is_dir {
+            self.dec_dirs();
+        }
+    }
+
+    pub fn is_block_allocated(&self, block_idx: u32) -> bool {
+        self.block_bitmap.is_allocated(block_idx as usize)
+    }
+
+    pub fn alloc_block(&mut self) -> Option<u32> {
+        let Some(block_idx) = self.block_bitmap.alloc() else {
+            return None;
+        };
+        self.dec_free_blocks();
+        Some(block_idx as u32)
+    }
+
+    pub fn free_block(&mut self, block_idx: u32) {
+        self.block_bitmap.free(block_idx as usize);
+        self.inc_free_blocks();
+    }
+
+    pub fn free_inodes_count(&self) -> u16 {
+        self.descriptor.free_inodes_count
+    }
+
+    pub fn free_blocks_count(&self) -> u16 {
+        self.descriptor.free_blocks_count
+    }
+
+    pub fn inc_free_inodes(&mut self) {
+        self.descriptor.free_inodes_count += 1;
+    }
+
+    pub fn dec_free_inodes(&mut self) {
+        debug_assert!(self.descriptor.free_inodes_count > 0);
+        self.descriptor.free_inodes_count -= 1;
+    }
+
+    pub fn inc_free_blocks(&mut self) {
+        self.descriptor.free_blocks_count += 1;
+    }
+
+    pub fn dec_free_blocks(&mut self) {
+        debug_assert!(self.descriptor.free_blocks_count > 0);
+        self.descriptor.free_blocks_count -= 1;
+    }
+
+    pub fn inc_dirs(&mut self) {
+        self.descriptor.dirs_count += 1;
+    }
+
+    pub fn dec_dirs(&mut self) {
+        debug_assert!(self.descriptor.dirs_count > 0);
+        self.descriptor.dirs_count -= 1;
+    }
+}
+
+/// The in-memory rust block group descriptor.
+///
+/// The block group descriptor contains information regarding where important data
+/// structures for that group are located.
+#[derive(Clone, Copy, Debug)]
+struct GroupDescriptor {
+    /// Blocks usage bitmap block
+    block_bitmap_bid: Bid,
+    /// Inodes usage bitmap block
+    inode_bitmap_bid: Bid,
+    /// Starting block of inode table
+    inode_table_bid: Bid,
+    /// Number of free blocks in group
+    free_blocks_count: u16,
+    /// Number of free inodes in group
+    free_inodes_count: u16,
+    /// Number of directories in group
+    dirs_count: u16,
+}
+
+impl From<RawGroupDescriptor> for GroupDescriptor {
+    fn from(desc: RawGroupDescriptor) -> Self {
+        Self {
+            block_bitmap_bid: Bid::new(desc.block_bitmap as _),
+            inode_bitmap_bid: Bid::new(desc.inode_bitmap as _),
+            inode_table_bid: Bid::new(desc.inode_table as _),
+            free_blocks_count: desc.free_blocks_count,
+            free_inodes_count: desc.free_inodes_count,
+            dirs_count: desc.dirs_count,
+        }
+    }
+}
+
+const_assert!(core::mem::size_of::<RawGroupDescriptor>() == 32);
+
+/// The raw block group descriptor.
+///
+/// The table starts on the first block following the superblock.
+#[repr(C)]
+#[derive(Clone, Copy, Debug, Pod)]
+pub(super) struct RawGroupDescriptor {
+    pub block_bitmap: u32,
+    pub inode_bitmap: u32,
+    pub inode_table: u32,
+    pub free_blocks_count: u16,
+    pub free_inodes_count: u16,
+    pub dirs_count: u16,
+    pad: u16,
+    reserved: [u32; 3],
+}
+
+impl From<&GroupDescriptor> for RawGroupDescriptor {
+    fn from(desc: &GroupDescriptor) -> Self {
+        Self {
+            block_bitmap: desc.block_bitmap_bid.to_raw() as _,
+            inode_bitmap: desc.inode_bitmap_bid.to_raw() as _,
+            inode_table: desc.inode_table_bid.to_raw() as _,
+            free_blocks_count: desc.free_blocks_count,
+            free_inodes_count: desc.free_inodes_count,
+            dirs_count: desc.dirs_count,
+            pad: 0u16,
+            reserved: [0u32; 3],
+        }
+    }
+}

services/libs/aster-std/src/fs/ext2/blocks_hole.rs

@@ -0,0 +1,56 @@
+use bitvec::prelude::BitVec;
+
+/// A blocks hole descriptor implemented by the `BitVec`.
+///
+/// The true bit implies that the block is a hole, and conversely.
+pub(super) struct BlocksHoleDesc(BitVec);
+
+impl BlocksHoleDesc {
+    /// Constructs a blocks hole descriptor with initial size.
+    ///
+    /// The `initial_size` usually is the number of blocks for a file.
+    pub fn new(initial_size: usize) -> Self {
+        let mut bit_vec = BitVec::with_capacity(initial_size);
+        bit_vec.resize(initial_size, false);
+        Self(bit_vec)
+    }
+
+    /// Returns the size.
+    pub fn size(&self) -> usize {
+        self.0.len()
+    }
+
+    /// Resizes the blocks hole to a new size.
+    ///
+    /// If `new_size` is greater than current size, the new blocks are all marked as hole.
+    pub fn resize(&mut self, new_size: usize) {
+        self.0.resize(new_size, true);
+    }
+
+    /// Returns if the block `idx` is a hole.
+    ///
+    /// # Panic
+    ///
+    /// If the `idx` is out of bounds, this method will panic.
+    pub fn is_hole(&self, idx: usize) -> bool {
+        self.0[idx]
+    }
+
+    /// Marks the block `idx` as a hole.
+    ///
+    /// # Panic
+    ///
+    /// If the `idx` is out of bounds, this method will panic.
+    pub fn set(&mut self, idx: usize) {
+        self.0.set(idx, true);
+    }
+
+    /// Unmarks the block `idx` as a hole.
+    ///
+    /// # Panic
+    ///
+    /// If the `idx` is out of bounds, this method will panic.
+    pub fn unset(&mut self, idx: usize) {
+        self.0.set(idx, false);
+    }
+}

services/libs/aster-std/src/fs/ext2/dir.rs

@@ -0,0 +1,321 @@
+use super::inode::{FileType, MAX_FNAME_LEN};
+use super::prelude::*;
+
+use core::iter::Iterator;
+
+/// The data structure in a directory's data block. It is stored in a linked list.
+///
+/// Each entry contains the name of the entry, the inode number, the file type,
+/// and the distance within the directory file to the next entry.
+#[derive(Clone, Debug)]
+pub struct DirEntry {
+    /// The header part.
+    header: DirEntryHeader,
+    /// Name of the entry, up to 255 bytes (excluding the null terminator).
+    name: CStr256,
+}
+
+impl DirEntry {
+    /// Constructs a new `DirEntry` object with the specified inode (`ino`),
+    /// name (`name`), and file type (`file_type`).
+    pub(super) fn new(ino: u32, name: &str, file_type: FileType) -> Self {
+        debug_assert!(name.len() <= MAX_FNAME_LEN);
+
+        let record_len = (Self::header_len() + name.len()).align_up(4) as u16;
+        Self {
+            header: DirEntryHeader {
+                ino,
+                record_len,
+                name_len: name.len() as u8,
+                file_type: DirEntryFileType::from(file_type) as _,
+            },
+            name: CStr256::from(name),
+        }
+    }
+
+    /// Constructs a `DirEntry` with the name "." and `self_ino` as its inode.
+    pub(super) fn self_entry(self_ino: u32) -> Self {
+        Self::new(self_ino, ".", FileType::Dir)
+    }
+
+    /// Constructs a `DirEntry` with the name ".." and `parent_ino` as its inode.
+    pub(super) fn parent_entry(parent_ino: u32) -> Self {
+        Self::new(parent_ino, "..", FileType::Dir)
+    }
+
+    /// Returns a reference to the header.
+    fn header(&self) -> &DirEntryHeader {
+        &self.header
+    }
+
+    /// Returns the length of the header.
+    fn header_len() -> usize {
+        core::mem::size_of::<DirEntryHeader>()
+    }
+
+    /// Returns the inode number.
+    pub fn ino(&self) -> u32 {
+        self.header.ino
+    }
+
+    /// Modifies the inode number.
+    pub fn set_ino(&mut self, ino: u32) {
+        self.header.ino = ino;
+    }
+
+    /// Returns the name.
+    pub fn name(&self) -> &str {
+        self.name.as_str().unwrap()
+    }
+
+    /// Returns the type.
+    pub fn type_(&self) -> FileType {
+        FileType::from(DirEntryFileType::try_from(self.header.file_type).unwrap())
+    }
+
+    /// Returns the distance to the next entry.
+    pub fn record_len(&self) -> usize {
+        self.header.record_len as _
+    }
+
+    /// Modifies the distance to the next entry.
+    pub(super) fn set_record_len(&mut self, record_len: usize) {
+        debug_assert!(record_len >= self.actual_len());
+        self.header.record_len = record_len as _;
+    }
+
+    /// Returns the actual length of the current entry.
+    pub(super) fn actual_len(&self) -> usize {
+        (Self::header_len() + self.name.len()).align_up(4)
+    }
+
+    /// Returns the length of the gap between the current entry and the next entry.
+    pub(super) fn gap_len(&self) -> usize {
+        self.record_len() - self.actual_len()
+    }
+}
+
+/// The header of `DirEntry`.
+#[repr(C)]
+#[derive(Clone, Copy, Debug, Pod)]
+struct DirEntryHeader {
+    /// Inode number
+    ino: u32,
+    /// Directory entry length
+    record_len: u16,
+    /// Name Length
+    name_len: u8,
+    /// Type indicator
+    file_type: u8,
+}
+
+/// The type indicator in the `DirEntry`.
+#[repr(u8)]
+#[derive(Copy, Clone, Debug, Eq, PartialEq, TryFromInt)]
+enum DirEntryFileType {
+    Unknown = 0,
+    File = 1,
+    Dir = 2,
+    Char = 3,
+    Block = 4,
+    Fifo = 5,
+    Socket = 6,
+    Symlink = 7,
+}
+
+impl From<FileType> for DirEntryFileType {
+    fn from(file_type: FileType) -> Self {
+        match file_type {
+            FileType::Fifo => Self::Fifo,
+            FileType::Char => Self::Char,
+            FileType::Dir => Self::Dir,
+            FileType::Block => Self::Block,
+            FileType::File => Self::File,
+            FileType::Symlink => Self::Symlink,
+            FileType::Socket => Self::Socket,
+        }
+    }
+}
+
+impl From<DirEntryFileType> for FileType {
+    fn from(file_type: DirEntryFileType) -> Self {
+        match file_type {
+            DirEntryFileType::Fifo => Self::Fifo,
+            DirEntryFileType::Char => Self::Char,
+            DirEntryFileType::Dir => Self::Dir,
+            DirEntryFileType::Block => Self::Block,
+            DirEntryFileType::File => Self::File,
+            DirEntryFileType::Symlink => Self::Symlink,
+            DirEntryFileType::Socket => Self::Socket,
+            DirEntryFileType::Unknown => panic!("unknown file type"),
+        }
+    }
+}
+
+/// A reader for reading `DirEntry` from the page cache.
+pub struct DirEntryReader<'a> {
+    page_cache: &'a PageCache,
+    offset: usize,
+}
+
+impl<'a> DirEntryReader<'a> {
+    /// Constructs a reader with the given page cache and offset.
+    pub(super) fn new(page_cache: &'a PageCache, from_offset: usize) -> Self {
+        Self {
+            page_cache,
+            offset: from_offset,
+        }
+    }
+
+    /// Reads one `DirEntry` from the current offset.
+    pub fn read_entry(&mut self) -> Result<DirEntry> {
+        let header = self
+            .page_cache
+            .pages()
+            .read_val::<DirEntryHeader>(self.offset)?;
+        if header.ino == 0 {
+            return_errno!(Errno::ENOENT);
+        }
+
+        let mut name = vec![0u8; header.name_len as _];
+        self.page_cache
+            .pages()
+            .read_bytes(self.offset + DirEntry::header_len(), &mut name)?;
+        let entry = DirEntry {
+            header,
+            name: CStr256::from(name.as_slice()),
+        };
+        self.offset += entry.record_len();
+
+        Ok(entry)
+    }
+}
+
+impl<'a> Iterator for DirEntryReader<'a> {
+    type Item = (usize, DirEntry);
+
+    fn next(&mut self) -> Option<Self::Item> {
+        let offset = self.offset;
+        let entry = match self.read_entry() {
+            Ok(entry) => entry,
+            Err(_) => {
+                return None;
+            }
+        };
+
+        Some((offset, entry))
+    }
+}
+
+/// A writer for modifying `DirEntry` of the page cache.
+pub struct DirEntryWriter<'a> {
+    page_cache: &'a PageCache,
+    offset: usize,
+}
+
+impl<'a> DirEntryWriter<'a> {
+    /// Constructs a writer with the given page cache and offset.
+    pub(super) fn new(page_cache: &'a PageCache, from_offset: usize) -> Self {
+        Self {
+            page_cache,
+            offset: from_offset,
+        }
+    }
+
+    /// Writes a `DirEntry` at the current offset.
+    pub fn write_entry(&mut self, entry: &DirEntry) -> Result<()> {
+        self.page_cache
+            .pages()
+            .write_val(self.offset, entry.header())?;
+        self.page_cache.pages().write_bytes(
+            self.offset + DirEntry::header_len(),
+            entry.name().as_bytes(),
+        )?;
+        self.offset += entry.record_len();
+        Ok(())
+    }
+
+    /// Appends a new `DirEntry` starting from the current offset.
+    ///
+    /// If there is a gap between existing entries, inserts the new entry into the gap；
+    /// If there is no available space, expands the size and appends the new entry at the end.
+    pub fn append_entry(&mut self, mut new_entry: DirEntry) -> Result<()> {
+        let Some((offset, mut entry)) = DirEntryReader::new(self.page_cache, self.offset)
+            .find(|(_, entry)| entry.gap_len() >= new_entry.record_len())
+        else {
+            // Resize and append it at the new block.
+            let old_size = self.page_cache.pages().size();
+            let new_size = old_size + BLOCK_SIZE;
+            self.page_cache.pages().resize(new_size)?;
+            new_entry.set_record_len(BLOCK_SIZE);
+            self.offset = old_size;
+            self.write_entry(&new_entry)?;
+            return Ok(());
+        };
+
+        // Write in the gap between existing entries.
+        new_entry.set_record_len(entry.gap_len());
+        entry.set_record_len(entry.actual_len());
+        self.offset = offset;
+        self.write_entry(&entry)?;
+        self.write_entry(&new_entry)?;
+        Ok(())
+    }
+
+    /// Removes and returns an existing `DirEntry` indicated by `name`.
+    pub fn remove_entry(&mut self, name: &str) -> Result<DirEntry> {
+        let self_entry_record_len = DirEntry::self_entry(0).record_len();
+        let reader = DirEntryReader::new(self.page_cache, 0);
+        let next_reader = DirEntryReader::new(self.page_cache, self_entry_record_len);
+        let Some(((pre_offset, mut pre_entry), (offset, entry))) = reader
+            .zip(next_reader)
+            .find(|((offset, _), (_, dir_entry))| dir_entry.name() == name)
+        else {
+            return_errno!(Errno::ENOENT);
+        };
+
+        if DirEntryReader::new(self.page_cache, offset)
+            .next()
+            .is_none()
+            && Bid::from_offset(pre_offset) != Bid::from_offset(offset)
+        {
+            // Shrink the size.
+            let new_size = pre_offset.align_up(BLOCK_SIZE);
+            self.page_cache.pages().resize(new_size)?;
+            pre_entry.set_record_len(new_size - pre_offset);
+            self.offset = pre_offset;
+            self.write_entry(&pre_entry)?;
+        } else {
+            // Update the previous entry.
+            pre_entry.set_record_len(pre_entry.record_len() + entry.record_len());
+            self.offset = pre_offset;
+            self.write_entry(&pre_entry)?;
+        }
+
+        Ok(entry)
+    }
+
+    /// Renames the `DirEntry` from `old_name` to the `new_name` from the current offset.
+    ///
+    /// It will moves the `DirEntry` to another position,
+    /// if the record length is not big enough.
+    pub fn rename_entry(&mut self, old_name: &str, new_name: &str) -> Result<()> {
+        let (offset, entry) = DirEntryReader::new(self.page_cache, self.offset)
+            .find(|(offset, entry)| entry.name() == old_name)
+            .ok_or(Error::new(Errno::ENOENT))?;
+
+        let mut new_entry = DirEntry::new(entry.ino(), new_name, entry.type_());
+        if new_entry.record_len() <= entry.record_len() {
+            // Just rename the entry.
+            new_entry.set_record_len(entry.record_len());
+            self.offset = offset;
+            self.write_entry(&new_entry)?;
+        } else {
+            // Move to another position.
+            self.remove_entry(old_name)?;
+            self.offset = 0;
+            self.append_entry(new_entry)?;
+        }
+        Ok(())
+    }
+}

services/libs/aster-std/src/fs/ext2/fs.rs

@@ -0,0 +1,366 @@
+use super::block_group::{BlockGroup, RawGroupDescriptor};
+use super::inode::{FilePerm, FileType, Inode, InodeDesc, RawInode};
+use super::prelude::*;
+use super::super_block::{RawSuperBlock, SuperBlock, SUPER_BLOCK_OFFSET};
+
+/// The root inode number.
+const ROOT_INO: u32 = 2;
+
+/// The Ext2 filesystem.
+#[derive(Debug)]
+pub struct Ext2 {
+    block_device: Arc<dyn BlockDevice>,
+    super_block: RwMutex<Dirty<SuperBlock>>,
+    block_groups: Vec<BlockGroup>,
+    inodes_per_group: u32,
+    blocks_per_group: u32,
+    inode_size: usize,
+    block_size: usize,
+    group_descriptors_segment: VmSegment,
+    self_ref: Weak<Self>,
+}
+
+impl Ext2 {
+    /// Opens and loads an Ext2 from the `block_device`.
+    pub fn open(block_device: Arc<dyn BlockDevice>) -> Result<Arc<Self>> {
+        // Load the superblock
+        // TODO: if the main superblock is corrupted, should we load the backup?
+        let super_block = {
+            let raw_super_block = block_device.read_val::<RawSuperBlock>(SUPER_BLOCK_OFFSET)?;
+            SuperBlock::try_from(raw_super_block)?
+        };
+        assert!(super_block.block_size() == BLOCK_SIZE);
+
+        let group_descriptors_segment = {
+            let npages = ((super_block.block_groups_count() as usize)
+                * core::mem::size_of::<RawGroupDescriptor>())
+            .div_ceil(BLOCK_SIZE);
+            let segment = VmAllocOptions::new(npages)
+                .uninit(true)
+                .is_contiguous(true)
+                .alloc_contiguous()?;
+            match block_device.read_blocks_sync(super_block.group_descriptors_bid(0), &segment)? {
+                BioStatus::Complete => (),
+                err_status => {
+                    return Err(Error::from(err_status));
+                }
+            }
+            segment
+        };
+
+        // Load the block groups information
+        let load_block_groups = |fs: Weak<Ext2>,
+                                 block_device: &dyn BlockDevice,
+                                 group_descriptors_segment: &VmSegment|
+         -> Result<Vec<BlockGroup>> {
+            let block_groups_count = super_block.block_groups_count() as usize;
+            let mut block_groups = Vec::with_capacity(block_groups_count);
+            for idx in 0..block_groups_count {
+                let block_group = BlockGroup::load(
+                    group_descriptors_segment,
+                    idx,
+                    block_device,
+                    &super_block,
+                    fs.clone(),
+                )?;
+                block_groups.push(block_group);
+            }
+            Ok(block_groups)
+        };
+
+        let ext2 = Arc::new_cyclic(|weak_ref| Self {
+            inodes_per_group: super_block.inodes_per_group(),
+            blocks_per_group: super_block.blocks_per_group(),
+            inode_size: super_block.inode_size(),
+            block_size: super_block.block_size(),
+            block_groups: load_block_groups(
+                weak_ref.clone(),
+                block_device.as_ref(),
+                &group_descriptors_segment,
+            )
+            .unwrap(),
+            block_device,
+            super_block: RwMutex::new(Dirty::new(super_block)),
+            group_descriptors_segment,
+            self_ref: weak_ref.clone(),
+        });
+        Ok(ext2)
+    }
+
+    /// Returns the block device.
+    pub fn block_device(&self) -> &dyn BlockDevice {
+        self.block_device.as_ref()
+    }
+
+    /// Returns the size of block.
+    pub fn block_size(&self) -> usize {
+        self.block_size
+    }
+
+    /// Returns the size of inode.
+    pub fn inode_size(&self) -> usize {
+        self.inode_size
+    }
+
+    /// Returns the number of inodes in each block group.
+    pub fn inodes_per_group(&self) -> u32 {
+        self.inodes_per_group
+    }
+
+    /// Returns the number of blocks in each block group.
+    pub fn blocks_per_group(&self) -> u32 {
+        self.blocks_per_group
+    }
+
+    /// Returns the super block.
+    pub fn super_block(&self) -> RwMutexReadGuard<'_, Dirty<SuperBlock>> {
+        self.super_block.read()
+    }
+
+    /// Returns the root inode.
+    pub fn root_inode(&self) -> Result<Arc<Inode>> {
+        self.lookup_inode(ROOT_INO)
+    }
+
+    /// Finds and returns the inode by `ino`.
+    pub(super) fn lookup_inode(&self, ino: u32) -> Result<Arc<Inode>> {
+        let (_, block_group) = self.block_group_of_ino(ino)?;
+        let inode_idx = self.inode_idx(ino);
+        block_group.lookup_inode(inode_idx)
+    }
+
+    /// Creates a new inode.
+    pub(super) fn create_inode(
+        &self,
+        dir_block_group_idx: usize,
+        file_type: FileType,
+        file_perm: FilePerm,
+    ) -> Result<Arc<Inode>> {
+        let (block_group_idx, ino) =
+            self.alloc_ino(dir_block_group_idx, file_type == FileType::Dir)?;
+        let inode = {
+            let inode_desc = InodeDesc::new(file_type, file_perm);
+            Inode::new(ino, block_group_idx, inode_desc, self.self_ref.clone())
+        };
+        let block_group = &self.block_groups[block_group_idx];
+        block_group.insert_cache(self.inode_idx(ino), inode.clone());
+        Ok(inode)
+    }
+
+    /// Allocates a new inode number, internally used by `new_inode`.
+    ///
+    /// Attempts to allocate from the `dir_block_group_idx` group first.
+    /// If allocation is not possible from this group, then search the remaining groups.
+    fn alloc_ino(&self, dir_block_group_idx: usize, is_dir: bool) -> Result<(usize, u32)> {
+        let mut block_group_idx = dir_block_group_idx;
+        if block_group_idx >= self.block_groups.len() {
+            return_errno_with_message!(Errno::EINVAL, "invalid block group idx");
+        }
+
+        for _ in 0..self.block_groups.len() {
+            if block_group_idx >= self.block_groups.len() {
+                block_group_idx = 0;
+            }
+            let block_group = &self.block_groups[block_group_idx];
+            if let Some(inode_idx) = block_group.alloc_inode(is_dir) {
+                let ino = block_group_idx as u32 * self.inodes_per_group + inode_idx + 1;
+                self.super_block.write().dec_free_inodes();
+                return Ok((block_group_idx, ino));
+            }
+            block_group_idx += 1;
+        }
+
+        return_errno_with_message!(Errno::ENOSPC, "no space on device");
+    }
+
+    /// Frees an inode.
+    pub(super) fn free_inode(&self, ino: u32, is_dir: bool) -> Result<()> {
+        let (_, block_group) = self.block_group_of_ino(ino)?;
+        let inode_idx = self.inode_idx(ino);
+        // In order to prevent value underflow, it is necessary to increment
+        // the free inode counter prior to freeing the inode.
+        self.super_block.write().inc_free_inodes();
+        block_group.free_inode(inode_idx, is_dir);
+        Ok(())
+    }
+
+    /// Writes back the metadata of inode.
+    pub(super) fn sync_inode(&self, ino: u32, inode: &InodeDesc) -> Result<()> {
+        let (_, block_group) = self.block_group_of_ino(ino)?;
+        let inode_idx = self.inode_idx(ino);
+        block_group.sync_raw_inode(inode_idx, &RawInode::from(inode));
+        Ok(())
+    }
+
+    /// Writes back the block group descriptor to the descriptors table.
+    pub(super) fn sync_group_descriptor(
+        &self,
+        block_group_idx: usize,
+        raw_descriptor: &RawGroupDescriptor,
+    ) -> Result<()> {
+        let offset = block_group_idx * core::mem::size_of::<RawGroupDescriptor>();
+        self.group_descriptors_segment
+            .write_val(offset, raw_descriptor)?;
+        Ok(())
+    }
+
+    /// Allocates a new block.
+    ///
+    /// Attempts to allocate from the `block_group_idx` group first.
+    /// If allocation is not possible from this group, then search the remaining groups.
+    pub(super) fn alloc_block(&self, block_group_idx: usize) -> Result<Bid> {
+        let mut block_group_idx = block_group_idx;
+        if block_group_idx >= self.block_groups.len() {
+            return_errno_with_message!(Errno::EINVAL, "invalid block group idx");
+        }
+
+        for _ in 0..self.block_groups.len() {
+            if block_group_idx >= self.block_groups.len() {
+                block_group_idx = 0;
+            }
+            let block_group = &self.block_groups[block_group_idx];
+            if let Some(block_idx) = block_group.alloc_block() {
+                let bid = block_group_idx as u32 * self.blocks_per_group + block_idx;
+                self.super_block.write().dec_free_blocks();
+                return Ok(Bid::new(bid as _));
+            }
+            block_group_idx += 1;
+        }
+
+        return_errno_with_message!(Errno::ENOSPC, "no space on device");
+    }
+
+    /// Frees a block.
+    pub(super) fn free_block(&self, bid: Bid) -> Result<()> {
+        let (_, block_group) = self.block_group_of_bid(bid)?;
+        let block_idx = self.block_idx(bid);
+        // In order to prevent value underflow, it is necessary to increment
+        // the free block counter prior to freeing the block.
+        self.super_block.write().inc_free_blocks();
+        block_group.free_block(block_idx);
+        Ok(())
+    }
+
+    /// Reads contiguous blocks starting from the `bid` synchronously.
+    pub(super) fn read_blocks(&self, bid: Bid, segment: &VmSegment) -> Result<()> {
+        let status = self.block_device.read_blocks_sync(bid, segment)?;
+        match status {
+            BioStatus::Complete => Ok(()),
+            err_status => Err(Error::from(err_status)),
+        }
+    }
+
+    /// Reads one block indicated by the `bid` synchronously.
+    pub(super) fn read_block(&self, bid: Bid, frame: &VmFrame) -> Result<()> {
+        let status = self.block_device.read_block_sync(bid, frame)?;
+        match status {
+            BioStatus::Complete => Ok(()),
+            err_status => Err(Error::from(err_status)),
+        }
+    }
+
+    /// Writes contiguous blocks starting from the `bid` synchronously.
+    pub(super) fn write_blocks(&self, bid: Bid, segment: &VmSegment) -> Result<()> {
+        let status = self.block_device.write_blocks_sync(bid, segment)?;
+        match status {
+            BioStatus::Complete => Ok(()),
+            err_status => Err(Error::from(err_status)),
+        }
+    }
+
+    /// Writes one block indicated by the `bid` synchronously.
+    pub(super) fn write_block(&self, bid: Bid, frame: &VmFrame) -> Result<()> {
+        let status = self.block_device.write_block_sync(bid, frame)?;
+        match status {
+            BioStatus::Complete => Ok(()),
+            err_status => Err(Error::from(err_status)),
+        }
+    }
+
+    /// Writes back the metadata to the block device.
+    pub fn sync_metadata(&self) -> Result<()> {
+        // If the superblock is clean, the block groups must be clean.
+        if !self.super_block.read().is_dirty() {
+            return Ok(());
+        }
+
+        let mut super_block = self.super_block.write();
+        // Writes back the metadata of block groups
+        for block_group in &self.block_groups {
+            block_group.sync_metadata(&super_block)?;
+        }
+
+        let mut bio_waiter = BioWaiter::new();
+        // Writes back the main superblock and group descriptor table.
+        let raw_super_block = RawSuperBlock::from((*super_block).deref());
+        bio_waiter.concat(
+            self.block_device
+                .write_bytes_async(SUPER_BLOCK_OFFSET, raw_super_block.as_bytes())?,
+        );
+        bio_waiter.concat(self.block_device.write_blocks(
+            super_block.group_descriptors_bid(0),
+            &self.group_descriptors_segment,
+        )?);
+
+        // Writes back the backups of superblock and group descriptor table.
+        let mut raw_super_block_backup = raw_super_block;
+        for idx in 1..super_block.block_groups_count() {
+            if super_block.is_backup_group(idx as usize) {
+                raw_super_block_backup.block_group_idx = idx as u16;
+                bio_waiter.concat(self.block_device.write_bytes_async(
+                    super_block.bid(idx as usize).to_offset(),
+                    raw_super_block_backup.as_bytes(),
+                )?);
+                bio_waiter.concat(self.block_device.write_blocks(
+                    super_block.group_descriptors_bid(idx as usize),
+                    &self.group_descriptors_segment,
+                )?);
+            }
+        }
+
+        // Waits for the completion of all submitted bios.
+        bio_waiter
+            .wait()
+            .ok_or_else(|| Error::with_message(Errno::EIO, "failed to sync metadata of fs"))?;
+
+        // Reset to clean.
+        super_block.clear_dirty();
+        Ok(())
+    }
+
+    /// Writes back all the cached inodes to the block device.
+    pub fn sync_all_inodes(&self) -> Result<()> {
+        for block_group in &self.block_groups {
+            block_group.sync_all_inodes()?;
+        }
+        Ok(())
+    }
+
+    #[inline]
+    fn block_group_of_bid(&self, bid: Bid) -> Result<(usize, &BlockGroup)> {
+        let block_group_idx = (bid.to_raw() / (self.blocks_per_group as u64)) as usize;
+        if block_group_idx >= self.block_groups.len() {
+            return_errno!(Errno::ENOENT);
+        }
+        Ok((block_group_idx, &self.block_groups[block_group_idx]))
+    }
+
+    #[inline]
+    fn block_group_of_ino(&self, ino: u32) -> Result<(usize, &BlockGroup)> {
+        let block_group_idx = ((ino - 1) / self.inodes_per_group) as usize;
+        if block_group_idx >= self.block_groups.len() {
+            return_errno!(Errno::ENOENT);
+        }
+        Ok((block_group_idx, &self.block_groups[block_group_idx]))
+    }
+
+    #[inline]
+    fn inode_idx(&self, ino: u32) -> u32 {
+        (ino - 1) % self.inodes_per_group
+    }
+
+    #[inline]
+    fn block_idx(&self, bid: Bid) -> u32 {
+        (bid.to_raw() as u32) % self.blocks_per_group
+    }
+}

services/libs/aster-std/src/fs/ext2/impl_for_vfs/fs.rs

@@ -0,0 +1,43 @@
+use crate::fs::ext2::{utils::Dirty, Ext2, SuperBlock as Ext2SuperBlock, MAGIC_NUM as EXT2_MAGIC};
+use crate::fs::utils::{FileSystem, FsFlags, Inode, SuperBlock, NAME_MAX};
+use crate::prelude::*;
+
+use aster_frame::sync::RwMutexReadGuard;
+
+impl FileSystem for Ext2 {
+    fn sync(&self) -> Result<()> {
+        self.sync_all_inodes()?;
+        self.sync_metadata()?;
+        Ok(())
+    }
+
+    fn root_inode(&self) -> Arc<dyn Inode> {
+        self.root_inode().unwrap()
+    }
+
+    fn sb(&self) -> SuperBlock {
+        SuperBlock::from(self.super_block())
+    }
+
+    fn flags(&self) -> FsFlags {
+        FsFlags::empty()
+    }
+}
+
+impl From<RwMutexReadGuard<'_, Dirty<Ext2SuperBlock>>> for SuperBlock {
+    fn from(ext2_sb: RwMutexReadGuard<Dirty<Ext2SuperBlock>>) -> Self {
+        Self {
+            magic: EXT2_MAGIC as _,
+            bsize: ext2_sb.block_size(),
+            blocks: ext2_sb.total_blocks() as _,
+            bfree: ext2_sb.free_blocks() as _,
+            bavail: ext2_sb.free_blocks() as _,
+            files: ext2_sb.total_inodes() as _,
+            ffree: ext2_sb.free_inodes() as _,
+            fsid: 0, // TODO
+            namelen: NAME_MAX,
+            frsize: ext2_sb.fragment_size(),
+            flags: 0, // TODO
+        }
+    }
+}

services/libs/aster-std/src/fs/ext2/impl_for_vfs/inode.rs

@@ -0,0 +1,175 @@
+use crate::fs::device::Device;
+use crate::fs::ext2::{FilePerm, FileType, Inode as Ext2Inode};
+use crate::fs::utils::{
+    DirentVisitor, FileSystem, Inode, InodeMode, InodeType, IoctlCmd, Metadata,
+};
+use crate::prelude::*;
+use crate::vm::vmo::Vmo;
+
+use aster_rights::Full;
+use core::time::Duration;
+
+impl Inode for Ext2Inode {
+    fn len(&self) -> usize {
+        self.file_size() as _
+    }
+
+    fn resize(&self, new_size: usize) -> Result<()> {
+        self.resize(new_size)
+    }
+
+    fn metadata(&self) -> Metadata {
+        Metadata {
+            dev: 0, // TODO: ID of block device
+            ino: self.ino() as _,
+            size: self.file_size() as _,
+            blk_size: self.fs().super_block().block_size(),
+            blocks: self.blocks_count() as _,
+            atime: self.atime(),
+            mtime: self.mtime(),
+            ctime: self.ctime(),
+            type_: InodeType::from(self.file_type()),
+            mode: InodeMode::from(self.file_perm()),
+            nlinks: self.hard_links() as _,
+            uid: self.uid() as _,
+            gid: self.gid() as _,
+            rdev: self.device_id(),
+        }
+    }
+
+    fn atime(&self) -> Duration {
+        self.atime()
+    }
+
+    fn set_atime(&self, time: Duration) {
+        self.set_atime(time)
+    }
+
+    fn mtime(&self) -> Duration {
+        self.mtime()
+    }
+
+    fn set_mtime(&self, time: Duration) {
+        self.set_mtime(time)
+    }
+
+    fn ino(&self) -> u64 {
+        self.ino() as _
+    }
+
+    fn type_(&self) -> InodeType {
+        InodeType::from(self.file_type())
+    }
+
+    fn mode(&self) -> InodeMode {
+        InodeMode::from(self.file_perm())
+    }
+
+    fn set_mode(&self, mode: InodeMode) {
+        self.set_file_perm(mode.into());
+    }
+
+    fn page_cache(&self) -> Option<Vmo<Full>> {
+        Some(self.page_cache())
+    }
+
+    fn read_at(&self, offset: usize, buf: &mut [u8]) -> Result<usize> {
+        self.read_at(offset, buf)
+    }
+
+    fn read_direct_at(&self, offset: usize, buf: &mut [u8]) -> Result<usize> {
+        self.read_direct_at(offset, buf)
+    }
+
+    fn write_at(&self, offset: usize, buf: &[u8]) -> Result<usize> {
+        self.write_at(offset, buf)
+    }
+
+    fn write_direct_at(&self, offset: usize, buf: &[u8]) -> Result<usize> {
+        self.write_direct_at(offset, buf)
+    }
+
+    fn create(&self, name: &str, type_: InodeType, mode: InodeMode) -> Result<Arc<dyn Inode>> {
+        Ok(self.create(name, type_.into(), mode.into())?)
+    }
+
+    fn mknod(&self, name: &str, mode: InodeMode, dev: Arc<dyn Device>) -> Result<Arc<dyn Inode>> {
+        let inode = self.create(name, InodeType::from(dev.type_()).into(), mode.into())?;
+        inode.set_device_id(dev.id().into()).unwrap();
+        Ok(inode)
+    }
+
+    fn lookup(&self, name: &str) -> Result<Arc<dyn Inode>> {
+        Ok(self.lookup(name)?)
+    }
+
+    fn readdir_at(&self, offset: usize, visitor: &mut dyn DirentVisitor) -> Result<usize> {
+        self.readdir_at(offset, visitor)
+    }
+
+    fn link(&self, old: &Arc<dyn Inode>, name: &str) -> Result<()> {
+        let old = old
+            .downcast_ref::<Ext2Inode>()
+            .ok_or_else(|| Error::with_message(Errno::EXDEV, "not same fs"))?;
+        self.link(old, name)
+    }
+
+    fn unlink(&self, name: &str) -> Result<()> {
+        self.unlink(name)
+    }
+
+    fn rmdir(&self, name: &str) -> Result<()> {
+        self.rmdir(name)
+    }
+
+    fn rename(&self, old_name: &str, target: &Arc<dyn Inode>, new_name: &str) -> Result<()> {
+        let target = target
+            .downcast_ref::<Ext2Inode>()
+            .ok_or_else(|| Error::with_message(Errno::EXDEV, "not same fs"))?;
+        self.rename(old_name, target, new_name)
+    }
+
+    fn read_link(&self) -> Result<String> {
+        self.read_link()
+    }
+
+    fn write_link(&self, target: &str) -> Result<()> {
+        self.write_link(target)
+    }
+
+    fn ioctl(&self, cmd: IoctlCmd, arg: usize) -> Result<i32> {
+        Err(Error::new(Errno::EINVAL))
+    }
+
+    fn sync(&self) -> Result<()> {
+        self.sync_all()
+    }
+
+    fn fs(&self) -> Arc<dyn FileSystem> {
+        self.fs()
+    }
+}
+
+impl From<FilePerm> for InodeMode {
+    fn from(perm: FilePerm) -> Self {
+        Self::from_bits_truncate(perm.bits() as _)
+    }
+}
+
+impl From<InodeMode> for FilePerm {
+    fn from(mode: InodeMode) -> Self {
+        Self::from_bits_truncate(mode.bits() as _)
+    }
+}
+
+impl From<FileType> for InodeType {
+    fn from(type_: FileType) -> Self {
+        Self::try_from(type_ as u32).unwrap()
+    }
+}
+
+impl From<InodeType> for FileType {
+    fn from(type_: InodeType) -> Self {
+        Self::try_from(type_ as u16).unwrap()
+    }
+}

services/libs/aster-std/src/fs/ext2/impl_for_vfs/mod.rs

@@ -0,0 +1,2 @@
+mod fs;
+mod inode;

services/libs/aster-std/src/fs/ext2/mod.rs

@@ -0,0 +1,50 @@
+//! A safe Rust Ext2 filesystem.
+//!
+//! The Second Extended File System(Ext2) is a major rewrite of the Ext filesystem.
+//! It is the predominant filesystem in use by Linux from the early 1990s to the early 2000s.
+//! The structures of Ext3 and Ext4 are based on Ext2 and add some additional options
+//! such as journaling.
+//!
+//! The features of this version of Ext2 are as follows:
+//! 1. No unsafe Rust. The filesystem is written is Rust without any unsafe code,
+//!    ensuring that there are no memory safety issues in the code.
+//! 2. Deep integration with PageCache. The data and metadata of the filesystem are
+//!    stored in PageCache, which accelerates the performance of data access.
+//! 3. Compatible with queue-based block device. The filesystem can submits multiple
+//!    BIO requests to be block device at once, thereby enhancing I/O performance.
+//!
+//! # Example
+//!
+//! ```no_run
+//! // Opens an Ext2 from the block device.
+//! let ext2 = Ext2::open(block_device)?;
+//! // Lookup the root inode.
+//! let root = ext2.root_inode()?;
+//! // Create a file inside root directory.
+//! let file = root.create("file", FileType::File, FilePerm::from_bits_truncate(0o666))?;
+//! // Write data into the file.
+//! const WRITE_DATA: &[u8] = b"Hello, World";
+//! let len = file.write_at(0, WRITE_DATA)?;
+//! assert!(len == WRITE_DATA.len());
+//! ```
+//!
+//! # Limitation
+//!
+//! Here we summarizes the features that need to be implemented in the future.
+//! 1. Supports large file.
+//! 2. Supports merging small read/write operations.
+//! 3. Handles the intermediate failure status correctly.
+
+pub use fs::Ext2;
+pub use inode::{FilePerm, FileType, Inode};
+pub use super_block::{SuperBlock, MAGIC_NUM};
+
+mod block_group;
+mod blocks_hole;
+mod dir;
+mod fs;
+mod impl_for_vfs;
+mod inode;
+mod prelude;
+mod super_block;
+mod utils;

services/libs/aster-std/src/fs/ext2/prelude.rs

@@ -0,0 +1,23 @@
+pub(super) use super::utils::{Dirty, IsPowerOf};
+
+pub(super) use crate::fs::utils::{
+    CStr256, DirentVisitor, InodeType, PageCache, PageCacheBackend, Str16, Str64,
+};
+pub(super) use crate::prelude::*;
+pub(super) use crate::time::UnixTime;
+pub(super) use crate::vm::vmo::Vmo;
+
+pub(super) use align_ext::AlignExt;
+pub(super) use aster_block::{
+    bio::{BioStatus, BioWaiter},
+    id::Bid,
+    BlockDevice, BLOCK_SIZE,
+};
+pub(super) use aster_frame::sync::{RwMutex, RwMutexReadGuard};
+pub(super) use aster_frame::vm::VmAllocOptions;
+pub(super) use aster_frame::vm::VmIo;
+pub(super) use aster_frame::vm::{VmFrame, VmSegment};
+pub(super) use aster_rights::Full;
+pub(super) use core::ops::{Deref, DerefMut};
+pub(super) use core::time::Duration;
+pub(super) use static_assertions::const_assert;

services/libs/aster-std/src/fs/ext2/super_block.rs

@@ -0,0 +1,542 @@
+use super::inode::RawInode;
+use super::prelude::*;
+
+/// The magic number of Ext2.
+pub const MAGIC_NUM: u16 = 0xef53;
+
+/// The main superblock is located at byte 1024 from the beginning of the device.
+pub const SUPER_BLOCK_OFFSET: usize = 1024;
+
+const SUPER_BLOCK_SIZE: usize = 1024;
+
+/// The in-memory rust superblock.
+///
+/// It contains all information about the layout of the Ext2.
+#[derive(Clone, Copy, Debug)]
+pub struct SuperBlock {
+    /// Total number of inodes.
+    inodes_count: u32,
+    /// Total number of blocks.
+    blocks_count: u32,
+    /// Total number of reserved blocks.
+    reserved_blocks_count: u32,
+    /// Total number of free blocks.
+    free_blocks_count: u32,
+    /// Total number of free inodes.
+    free_inodes_count: u32,
+    /// First data block.
+    first_data_block: Bid,
+    /// Block size.
+    block_size: usize,
+    /// Fragment size.
+    frag_size: usize,
+    /// Number of blocks in each block group.
+    blocks_per_group: u32,
+    /// Number of fragments in each block group.
+    frags_per_group: u32,
+    /// Number of inodes in each block group.
+    inodes_per_group: u32,
+    /// Mount time.
+    mtime: UnixTime,
+    /// Write time.
+    wtime: UnixTime,
+    /// Mount count.
+    mnt_count: u16,
+    /// Maximal mount count.
+    max_mnt_count: u16,
+    /// Magic signature.
+    magic: u16,
+    /// Filesystem state.
+    state: FsState,
+    /// Behaviour when detecting errors.
+    errors_behaviour: ErrorsBehaviour,
+    /// Time of last check.
+    last_check_time: UnixTime,
+    /// Interval between checks.
+    check_interval: Duration,
+    /// Creator OS ID.
+    creator_os: OsId,
+    /// Revision level.
+    rev_level: RevLevel,
+    /// Default uid for reserved blocks.
+    def_resuid: u32,
+    /// Default gid for reserved blocks.
+    def_resgid: u32,
+    //
+    // These fields are valid for RevLevel::Dynamic only.
+    //
+    /// First non-reserved inode number.
+    first_ino: u32,
+    /// Size of inode structure.
+    inode_size: usize,
+    /// Block group that this superblock is part of (if backup copy).
+    block_group_idx: usize,
+    /// Compatible feature set.
+    feature_compat: FeatureCompatSet,
+    /// Incompatible feature set.
+    feature_incompat: FeatureInCompatSet,
+    /// Readonly-compatible feature set.
+    feature_ro_compat: FeatureRoCompatSet,
+    /// 128-bit uuid for volume.
+    uuid: [u8; 16],
+    /// Volume name.
+    volume_name: Str16,
+    /// Directory where last mounted.
+    last_mounted_dir: Str64,
+    ///
+    /// This fields are valid if the FeatureCompatSet::DIR_PREALLOC is set.
+    ///
+    /// Number of blocks to preallocate for files.
+    prealloc_file_blocks: u8,
+    /// Number of blocks to preallocate for directories.
+    prealloc_dir_blocks: u8,
+}
+
+impl TryFrom<RawSuperBlock> for SuperBlock {
+    type Error = crate::error::Error;
+
+    fn try_from(sb: RawSuperBlock) -> Result<Self> {
+        Ok(Self {
+            inodes_count: sb.inodes_count,
+            blocks_count: sb.blocks_count,
+            reserved_blocks_count: sb.reserved_blocks_count,
+            free_blocks_count: sb.free_blocks_count,
+            free_inodes_count: sb.free_inodes_count,
+            first_data_block: Bid::new(sb.first_data_block as _),
+            block_size: 1024 << sb.log_block_size,
+            frag_size: 1024 << sb.log_frag_size,
+            blocks_per_group: sb.blocks_per_group,
+            frags_per_group: sb.frags_per_group,
+            inodes_per_group: sb.inodes_per_group,
+            mtime: sb.mtime,
+            wtime: sb.wtime,
+            mnt_count: sb.mnt_count,
+            max_mnt_count: sb.max_mnt_count,
+            magic: {
+                if sb.magic != MAGIC_NUM {
+                    return_errno_with_message!(Errno::EINVAL, "bad ext2 magic number");
+                }
+                MAGIC_NUM
+            },
+            state: {
+                let state = FsState::try_from(sb.state)
+                    .map_err(|_| Error::with_message(Errno::EINVAL, "invalid fs state"))?;
+                if state == FsState::Corrupted {
+                    return_errno_with_message!(Errno::EUCLEAN, "fs is corrupted");
+                }
+                state
+            },
+            errors_behaviour: ErrorsBehaviour::try_from(sb.errors)
+                .map_err(|_| Error::with_message(Errno::EINVAL, "invalid errors behaviour"))?,
+            last_check_time: sb.last_check_time,
+            check_interval: Duration::from_secs(sb.check_interval as _),
+            creator_os: {
+                let os_id = OsId::try_from(sb.creator_os)
+                    .map_err(|_| Error::with_message(Errno::EINVAL, "invalid creater os"))?;
+                if os_id != OsId::Linux {
+                    return_errno_with_message!(Errno::EINVAL, "not supported os id");
+                }
+                OsId::Linux
+            },
+            rev_level: {
+                let rev_level = RevLevel::try_from(sb.rev_level)
+                    .map_err(|_| Error::with_message(Errno::EINVAL, "invalid revision level"))?;
+                if rev_level != RevLevel::Dynamic {
+                    return_errno_with_message!(Errno::EINVAL, "not supported rev level");
+                }
+                RevLevel::Dynamic
+            },
+            def_resuid: sb.def_resuid as _,
+            def_resgid: sb.def_resgid as _,
+            first_ino: sb.first_ino,
+            inode_size: {
+                let inode_size = sb.inode_size as _;
+                if inode_size < core::mem::size_of::<RawInode>() {
+                    return_errno_with_message!(Errno::EINVAL, "inode size is too small");
+                }
+                inode_size
+            },
+            block_group_idx: sb.block_group_idx as _,
+            feature_compat: FeatureCompatSet::from_bits(sb.feature_compat).ok_or(
+                Error::with_message(Errno::EINVAL, "invalid feature compat set"),
+            )?,
+            feature_incompat: FeatureInCompatSet::from_bits(sb.feature_incompat).ok_or(
+                Error::with_message(Errno::EINVAL, "invalid feature incompat set"),
+            )?,
+            feature_ro_compat: FeatureRoCompatSet::from_bits(sb.feature_ro_compat).ok_or(
+                Error::with_message(Errno::EINVAL, "invalid feature ro compat set"),
+            )?,
+            uuid: sb.uuid,
+            volume_name: sb.volume_name,
+            last_mounted_dir: sb.last_mounted_dir,
+            prealloc_file_blocks: sb.prealloc_file_blocks,
+            prealloc_dir_blocks: sb.prealloc_dir_blocks,
+        })
+    }
+}
+
+impl SuperBlock {
+    /// Returns the block size.
+    pub fn block_size(&self) -> usize {
+        self.block_size
+    }
+
+    /// Returns the size of inode structure.
+    pub fn inode_size(&self) -> usize {
+        self.inode_size
+    }
+
+    /// Returns the fragment size.
+    pub fn fragment_size(&self) -> usize {
+        self.frag_size
+    }
+
+    /// Returns total number of inodes.
+    pub fn total_inodes(&self) -> u32 {
+        self.inodes_count
+    }
+
+    /// Returns total number of blocks.
+    pub fn total_blocks(&self) -> u32 {
+        self.blocks_count
+    }
+
+    /// Returns the number of blocks in each block group.
+    pub fn blocks_per_group(&self) -> u32 {
+        self.blocks_per_group
+    }
+
+    /// Returns the number of inodes in each block group.
+    pub fn inodes_per_group(&self) -> u32 {
+        self.inodes_per_group
+    }
+
+    /// Returns the number of block groups.
+    pub fn block_groups_count(&self) -> u32 {
+        self.blocks_count / self.blocks_per_group
+    }
+
+    /// Returns the filesystem state.
+    pub fn state(&self) -> FsState {
+        self.state
+    }
+
+    /// Returns the revision level.
+    pub fn rev_level(&self) -> RevLevel {
+        self.rev_level
+    }
+
+    /// Returns the compatible feature set.
+    pub fn feature_compat(&self) -> FeatureCompatSet {
+        self.feature_compat
+    }
+
+    /// Returns the incompatible feature set.
+    pub fn feature_incompat(&self) -> FeatureInCompatSet {
+        self.feature_incompat
+    }
+
+    /// Returns the readonly-compatible feature set.
+    pub fn feature_ro_compat(&self) -> FeatureRoCompatSet {
+        self.feature_ro_compat
+    }
+
+    /// Returns the number of free blocks.
+    pub fn free_blocks(&self) -> u32 {
+        self.free_blocks_count
+    }
+
+    /// Increase the number of free blocks.
+    pub(super) fn inc_free_blocks(&mut self) {
+        self.free_blocks_count += 1;
+    }
+
+    /// Decrease the number of free blocks.
+    pub(super) fn dec_free_blocks(&mut self) {
+        debug_assert!(self.free_blocks_count > 0);
+        self.free_blocks_count -= 1;
+    }
+
+    /// Returns the number of free inodes.
+    pub fn free_inodes(&self) -> u32 {
+        self.free_inodes_count
+    }
+
+    /// Increase the number of free inodes.
+    pub(super) fn inc_free_inodes(&mut self) {
+        self.free_inodes_count += 1;
+    }
+
+    /// Decrease the number of free inodes.
+    pub(super) fn dec_free_inodes(&mut self) {
+        debug_assert!(self.free_inodes_count > 0);
+        self.free_inodes_count -= 1;
+    }
+
+    /// Checks if the block group will backup the super block.
+    pub(super) fn is_backup_group(&self, block_group_idx: usize) -> bool {
+        if block_group_idx == 0 {
+            false
+        } else if self
+            .feature_ro_compat
+            .contains(FeatureRoCompatSet::SPARSE_SUPER)
+        {
+            // The backup groups chosen are 1 and powers of 3, 5 and 7.
+            block_group_idx == 1
+                || block_group_idx.is_power_of(3)
+                || block_group_idx.is_power_of(5)
+                || block_group_idx.is_power_of(7)
+        } else {
+            true
+        }
+    }
+
+    /// Returns the starting block id of the super block
+    /// inside the block group pointed by `block_group_idx`.
+    ///
+    /// # Panic
+    ///
+    /// If `block_group_idx` is neither 0 nor a backup block group index,
+    /// then the method panics.
+    pub(super) fn bid(&self, block_group_idx: usize) -> Bid {
+        if block_group_idx == 0 {
+            let bid = (SUPER_BLOCK_OFFSET / self.block_size) as u64;
+            return Bid::new(bid);
+        }
+
+        assert!(self.is_backup_group(block_group_idx));
+        let super_block_bid = block_group_idx * (self.blocks_per_group as usize);
+        Bid::new(super_block_bid as u64)
+    }
+
+    /// Returns the starting block id of the block group descripter table
+    /// inside the block group pointed by `block_group_idx`.
+    ///
+    /// # Panic
+    ///
+    /// If `block_group_idx` is neither 0 nor a backup block group index,
+    /// then the method panics.
+    pub(super) fn group_descriptors_bid(&self, block_group_idx: usize) -> Bid {
+        let super_block_bid = self.bid(block_group_idx);
+        super_block_bid + (SUPER_BLOCK_SIZE.div_ceil(self.block_size) as u64)
+    }
+}
+
+bitflags! {
+    /// Compatible feature set.
+    pub struct FeatureCompatSet: u32 {
+        /// Preallocate some number of blocks to a directory when creating a new one
+        const DIR_PREALLOC = 1 << 0;
+        /// AFS server inodes exist
+        const IMAGIC_INODES = 1 << 1;
+        /// File system has a journal
+        const HAS_JOURNAL = 1 << 2;
+        /// Inodes have extended attributes
+        const EXT_ATTR = 1 << 3;
+        /// File system can resize itself for larger partitions
+        const RESIZE_INO = 1 << 4;
+        /// Directories use hash index
+        const DIR_INDEX = 1 << 5;
+    }
+}
+
+bitflags! {
+    /// Incompatible feature set.
+    pub struct FeatureInCompatSet: u32 {
+        /// Compression is used
+        const COMPRESSION = 1 << 0;
+        /// Directory entries contain a type field
+        const FILETYPE = 1 << 1;
+        /// File system needs to replay its journal
+        const RECOVER = 1 << 2;
+        /// File system uses a journal device
+        const JOURNAL_DEV = 1 << 3;
+        /// Metablock block group
+        const META_BG = 1 << 4;
+    }
+}
+
+bitflags! {
+    /// Readonly-compatible feature set.
+    pub struct FeatureRoCompatSet: u32 {
+        /// Sparse superblocks and group descriptor tables
+        const SPARSE_SUPER = 1 << 0;
+        /// File system uses a 64-bit file size
+        const LARGE_FILE = 1 << 1;
+        /// Directory contents are stored in the form of a Binary Tree
+        const BTREE_DIR = 1 << 2;
+    }
+}
+
+#[repr(u16)]
+#[derive(Copy, Clone, Debug, Eq, PartialEq, TryFromInt)]
+pub enum FsState {
+    /// Unmounted cleanly
+    Valid = 1,
+    /// Errors detected
+    Err = 2,
+    /// Filesystem is corrupted (EUCLEAN)
+    Corrupted = 117,
+}
+
+#[repr(u16)]
+#[derive(Copy, Clone, Debug, Eq, PartialEq, TryFromInt)]
+pub enum ErrorsBehaviour {
+    /// Continue execution
+    Continue = 1,
+    // Remount fs read-only
+    RemountReadonly = 2,
+    // Should panic
+    Panic = 3,
+}
+
+impl Default for ErrorsBehaviour {
+    fn default() -> Self {
+        Self::Continue
+    }
+}
+
+#[repr(u32)]
+#[derive(Copy, Clone, Debug, Eq, PartialEq, TryFromInt)]
+pub enum OsId {
+    Linux = 0,
+    Hurd = 1,
+    Masix = 2,
+    FreeBSD = 3,
+    Lites = 4,
+}
+
+#[repr(u32)]
+#[derive(Copy, Clone, Debug, Eq, PartialEq, TryFromInt)]
+pub enum RevLevel {
+    /// The good old (original) format.
+    GoodOld = 0,
+    /// V2 format with dynamic inode size.
+    Dynamic = 1,
+}
+
+const_assert!(core::mem::size_of::<RawSuperBlock>() == SUPER_BLOCK_SIZE);
+
+/// The raw superblock, it must be exactly 1024 bytes in length.
+#[repr(C)]
+#[derive(Clone, Copy, Debug, Pod, Default)]
+pub(super) struct RawSuperBlock {
+    pub inodes_count: u32,
+    pub blocks_count: u32,
+    pub reserved_blocks_count: u32,
+    pub free_blocks_count: u32,
+    pub free_inodes_count: u32,
+    pub first_data_block: u32,
+    /// The number to left-shift 1024 to obtain the block size.
+    pub log_block_size: u32,
+    /// The number to left-shift 1024 to obtain the fragment size.
+    pub log_frag_size: u32,
+    pub blocks_per_group: u32,
+    pub frags_per_group: u32,
+    pub inodes_per_group: u32,
+    /// Mount time.
+    pub mtime: UnixTime,
+    /// Write time.
+    pub wtime: UnixTime,
+    pub mnt_count: u16,
+    pub max_mnt_count: u16,
+    pub magic: u16,
+    pub state: u16,
+    pub errors: u16,
+    pub min_rev_level: u16,
+    /// Time of last check.
+    pub last_check_time: UnixTime,
+    pub check_interval: u32,
+    pub creator_os: u32,
+    pub rev_level: u32,
+    pub def_resuid: u16,
+    pub def_resgid: u16,
+    pub first_ino: u32,
+    pub inode_size: u16,
+    pub block_group_idx: u16,
+    pub feature_compat: u32,
+    pub feature_incompat: u32,
+    pub feature_ro_compat: u32,
+    pub uuid: [u8; 16],
+    pub volume_name: Str16,
+    pub last_mounted_dir: Str64,
+    pub algorithm_usage_bitmap: u32,
+    pub prealloc_file_blocks: u8,
+    pub prealloc_dir_blocks: u8,
+    padding1: u16,
+    ///
+    /// This fileds are for journaling support in Ext3.
+    ///
+    /// Uuid of journal superblock.
+    pub journal_uuid: [u8; 16],
+    /// Inode number of journal file.
+    pub journal_ino: u32,
+    /// Device number of journal file.
+    pub journal_dev: u32,
+    /// Start of list of inodes to delete.
+    pub last_orphan: u32,
+    /// HTREE hash seed.
+    pub hash_seed: [u32; 4],
+    /// Default hash version to use
+    pub def_hash_version: u8,
+    reserved_char_pad: u8,
+    reserved_word_pad: u16,
+    /// Default mount options.
+    pub default_mount_opts: u32,
+    /// First metablock block group.
+    pub first_meta_bg: u32,
+    reserved: Reserved,
+}
+
+impl From<&SuperBlock> for RawSuperBlock {
+    fn from(sb: &SuperBlock) -> Self {
+        Self {
+            inodes_count: sb.inodes_count,
+            blocks_count: sb.blocks_count,
+            reserved_blocks_count: sb.reserved_blocks_count,
+            free_blocks_count: sb.free_blocks_count,
+            free_inodes_count: sb.free_inodes_count,
+            first_data_block: sb.first_data_block.to_raw() as u32,
+            log_block_size: (sb.block_size >> 11) as u32,
+            log_frag_size: (sb.frag_size >> 11) as u32,
+            blocks_per_group: sb.blocks_per_group,
+            frags_per_group: sb.frags_per_group,
+            inodes_per_group: sb.inodes_per_group,
+            mtime: sb.mtime,
+            wtime: sb.wtime,
+            mnt_count: sb.mnt_count,
+            max_mnt_count: sb.max_mnt_count,
+            magic: sb.magic,
+            state: sb.state as u16,
+            errors: sb.errors_behaviour as u16,
+            last_check_time: sb.last_check_time,
+            check_interval: sb.check_interval.as_secs() as u32,
+            creator_os: sb.creator_os as u32,
+            rev_level: sb.rev_level as u32,
+            def_resuid: sb.def_resuid as u16,
+            def_resgid: sb.def_resgid as u16,
+            first_ino: sb.first_ino,
+            inode_size: sb.inode_size as u16,
+            block_group_idx: sb.block_group_idx as u16,
+            feature_compat: sb.feature_compat.bits(),
+            feature_incompat: sb.feature_incompat.bits(),
+            feature_ro_compat: sb.feature_ro_compat.bits(),
+            uuid: sb.uuid,
+            volume_name: sb.volume_name,
+            last_mounted_dir: sb.last_mounted_dir,
+            prealloc_file_blocks: sb.prealloc_file_blocks,
+            prealloc_dir_blocks: sb.prealloc_dir_blocks,
+            ..Default::default()
+        }
+    }
+}
+
+#[repr(C)]
+#[derive(Clone, Copy, Debug, Pod)]
+struct Reserved([u32; 190]);
+
+impl Default for Reserved {
+    fn default() -> Self {
+        Self([0u32; 190])
+    }
+}

services/libs/aster-std/src/fs/ext2/utils.rs

@@ -0,0 +1,93 @@
+use super::prelude::*;
+
+use core::ops::MulAssign;
+
+pub trait IsPowerOf: Copy + Sized + MulAssign + PartialOrd {
+    /// Returns true if and only if `self == x^k` for some `k` where `k > 0`.
+    ///
+    /// The `x` must be a positive value.
+    fn is_power_of(&self, x: Self) -> bool {
+        let mut power = x;
+        while power < *self {
+            power *= x;
+        }
+
+        power == *self
+    }
+}
+
+macro_rules! impl_ipo_for {
+    ($($ipo_ty:ty),*) => {
+        $(impl IsPowerOf for $ipo_ty {})*
+    };
+}
+
+impl_ipo_for!(u8, u16, u32, u64, u128, i8, i16, i32, i64, i128, isize, usize);
+
+/// The `Dirty` wraps a value of type `T` with functions similar to that of a rw-lock,
+/// but simply sets a dirty flag on `write()`.
+pub struct Dirty<T: Debug> {
+    value: T,
+    dirty: bool,
+}
+
+impl<T: Debug> Dirty<T> {
+    /// Creates a new Dirty without setting the dirty flag.
+    pub fn new(val: T) -> Dirty<T> {
+        Dirty {
+            value: val,
+            dirty: false,
+        }
+    }
+
+    /// Creates a new Dirty with setting the dirty flag.
+    pub fn new_dirty(val: T) -> Dirty<T> {
+        Dirty {
+            value: val,
+            dirty: true,
+        }
+    }
+
+    /// Returns true if dirty, false otherwise.
+    pub fn is_dirty(&self) -> bool {
+        self.dirty
+    }
+
+    /// Clears the dirty flag.
+    pub fn clear_dirty(&mut self) {
+        self.dirty = false;
+    }
+}
+
+impl<T: Debug> Deref for Dirty<T> {
+    type Target = T;
+
+    /// Returns the imutable value.
+    fn deref(&self) -> &T {
+        &self.value
+    }
+}
+
+impl<T: Debug> DerefMut for Dirty<T> {
+    /// Returns the mutable value, sets the dirty flag.
+    fn deref_mut(&mut self) -> &mut T {
+        self.dirty = true;
+        &mut self.value
+    }
+}
+
+impl<T: Debug> Drop for Dirty<T> {
+    /// Guards if it is dirty when dropping.
+    fn drop(&mut self) {
+        if self.is_dirty() {
+            warn!("[{:?}] is dirty then dropping", self.value);
+        }
+    }
+}
+
+impl<T: Debug> Debug for Dirty<T> {
+    fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result {
+        let tag = if self.dirty { "Dirty" } else { "Clean" };
+        write!(f, "[{}] {:?}", tag, self.value)
+    }
+}

services/libs/aster-std/src/fs/mod.rs

@@ -1,6 +1,7 @@
 pub mod device;
 pub mod devpts;
 pub mod epoll;
+pub mod ext2;
 pub mod file_handle;
 pub mod file_table;
 pub mod fs_resolver;
@@ -10,3 +11,28 @@ pub mod procfs;
 pub mod ramfs;
 pub mod rootfs;
 pub mod utils;
+
+use crate::fs::{ext2::Ext2, fs_resolver::FsPath};
+use crate::prelude::*;
+use crate::thread::kernel_thread::KernelThreadExt;
+use aster_virtio::device::block::device::BlockDevice as VirtIoBlockDevice;
+use aster_virtio::device::block::DEVICE_NAME as VIRTIO_BLOCK_NAME;
+
+pub fn lazy_init() {
+    let block_device = aster_block::get_device(VIRTIO_BLOCK_NAME).unwrap();
+    let cloned_block_device = block_device.clone();
+
+    let task_fn = move || {
+        info!("spawn the virt-io-block thread");
+        let virtio_block_device = block_device.downcast_ref::<VirtIoBlockDevice>().unwrap();
+        loop {
+            virtio_block_device.handle_requests();
+        }
+    };
+    crate::Thread::spawn_kernel_thread(crate::ThreadOptions::new(task_fn));
+
+    let ext2_fs = Ext2::open(cloned_block_device).unwrap();
+    let target_path = FsPath::try_from("/ext2").unwrap();
+    println!("[kernel] Mount Ext2 fs at {:?} ", target_path);
+    self::rootfs::mount_fs_at(ext2_fs, &target_path).unwrap();
+}

services/libs/aster-std/src/fs/procfs/template/dir.rs

@@ -58,12 +58,18 @@ impl<D: DirOps + 'static> Inode for ProcDir<D> {
         self.info.size()
     }
 
-    fn resize(&self, _new_size: usize) {}
+    fn resize(&self, _new_size: usize) -> Result<()> {
+        Err(Error::new(Errno::EISDIR))
+    }
 
     fn metadata(&self) -> Metadata {
         self.info.metadata()
     }
 
+    fn ino(&self) -> u64 {
+        self.info.ino()
+    }
+
     fn type_(&self) -> InodeType {
         InodeType::Dir
     }

services/libs/aster-std/src/fs/procfs/template/file.rs

@@ -1,4 +1,3 @@
-use aster_frame::vm::VmFrame;
 use core::time::Duration;
 
 use crate::fs::utils::{FileSystem, Inode, InodeMode, InodeType, IoctlCmd, Metadata};
@@ -31,12 +30,18 @@ impl<F: FileOps + 'static> Inode for ProcFile<F> {
         self.info.size()
     }
 
-    fn resize(&self, _new_size: usize) {}
+    fn resize(&self, _new_size: usize) -> Result<()> {
+        Err(Error::new(Errno::EPERM))
+    }
 
     fn metadata(&self) -> Metadata {
         self.info.metadata()
     }
 
+    fn ino(&self) -> u64 {
+        self.info.ino()
+    }
+
     fn type_(&self) -> InodeType {
         InodeType::File
     }
@@ -65,14 +70,6 @@ impl<F: FileOps + 'static> Inode for ProcFile<F> {
         self.info.set_mtime(time)
     }
 
-    fn read_page(&self, _idx: usize, _frame: &VmFrame) -> Result<()> {
-        unreachable!()
-    }
-
-    fn write_page(&self, _idx: usize, _frame: &VmFrame) -> Result<()> {
-        unreachable!()
-    }
-
     fn read_at(&self, offset: usize, buf: &mut [u8]) -> Result<usize> {
         let data = self.inner.data()?;
         let start = data.len().min(offset);

services/libs/aster-std/src/fs/procfs/template/mod.rs

@@ -38,6 +38,10 @@ impl ProcInodeInfo {
         self.metadata.read().clone()
     }
 
+    pub fn ino(&self) -> u64 {
+        self.metadata.read().ino as _
+    }
+
     pub fn size(&self) -> usize {
         self.metadata.read().size
     }

services/libs/aster-std/src/fs/procfs/template/sym.rs

@@ -1,4 +1,3 @@
-use aster_frame::vm::VmFrame;
 use core::time::Duration;
 
 use crate::fs::utils::{FileSystem, Inode, InodeMode, InodeType, IoctlCmd, Metadata};
@@ -31,12 +30,18 @@ impl<S: SymOps + 'static> Inode for ProcSym<S> {
         self.info.size()
     }
 
-    fn resize(&self, _new_size: usize) {}
+    fn resize(&self, _new_size: usize) -> Result<()> {
+        Err(Error::new(Errno::EPERM))
+    }
 
     fn metadata(&self) -> Metadata {
         self.info.metadata()
     }
 
+    fn ino(&self) -> u64 {
+        self.info.ino()
+    }
+
     fn type_(&self) -> InodeType {
         InodeType::SymLink
     }
@@ -65,14 +70,6 @@ impl<S: SymOps + 'static> Inode for ProcSym<S> {
         self.info.set_mtime(time)
     }
 
-    fn read_page(&self, _idx: usize, _frame: &VmFrame) -> Result<()> {
-        Err(Error::new(Errno::EPERM))
-    }
-
-    fn write_page(&self, _idx: usize, _frame: &VmFrame) -> Result<()> {
-        Err(Error::new(Errno::EPERM))
-    }
-
     fn read_at(&self, _offset: usize, _buf: &mut [u8]) -> Result<usize> {
         Err(Error::new(Errno::EPERM))
     }

services/libs/aster-std/src/fs/ramfs/fs.rs

@@ -1,6 +1,6 @@
-use alloc::str;
-use aster_frame::sync::{RwLock, RwLockWriteGuard};
-use aster_frame::vm::{VmFrame, VmIo};
+use aster_frame::sync::RwLockWriteGuard;
+use aster_frame::vm::VmFrame;
+use aster_frame::vm::VmIo;
 use aster_rights::Full;
 use aster_util::slot_vec::SlotVec;
 use core::sync::atomic::{AtomicUsize, Ordering};
@@ -10,8 +10,8 @@ use super::*;
 use crate::events::IoEvents;
 use crate::fs::device::Device;
 use crate::fs::utils::{
-    DirentVisitor, FileSystem, FsFlags, Inode, InodeMode, InodeType, IoctlCmd, Metadata, PageCache,
-    SuperBlock,
+    CStr256, DirentVisitor, FileSystem, FsFlags, Inode, InodeMode, InodeType, IoctlCmd, Metadata,
+    PageCache, PageCacheBackend, SuperBlock,
 };
 use crate::prelude::*;
 use crate::process::signal::Poller;
@@ -219,7 +219,7 @@ impl Inner {
 }
 
 struct DirEntry {
-    children: SlotVec<(Str256, Arc<RamInode>)>,
+    children: SlotVec<(CStr256, Arc<RamInode>)>,
     this: Weak<RamInode>,
     parent: Weak<RamInode>,
 }
@@ -248,7 +248,7 @@ impl DirEntry {
         } else {
             self.children
                 .iter()
-                .any(|(child, _)| child.as_ref() == name)
+                .any(|(child, _)| child.as_str().unwrap() == name)
         }
     }
 
@@ -260,16 +260,16 @@ impl DirEntry {
         } else {
             self.children
                 .idxes_and_items()
-                .find(|(_, (child, _))| child.as_ref() == name)
+                .find(|(_, (child, _))| child.as_str().unwrap() == name)
                 .map(|(idx, (_, inode))| (idx + 2, inode.clone()))
         }
     }
 
     fn append_entry(&mut self, name: &str, inode: Arc<RamInode>) -> usize {
-        self.children.put((Str256::from(name), inode))
+        self.children.put((CStr256::from(name), inode))
     }
 
-    fn remove_entry(&mut self, idx: usize) -> Option<(Str256, Arc<RamInode>)> {
+    fn remove_entry(&mut self, idx: usize) -> Option<(CStr256, Arc<RamInode>)> {
         assert!(idx >= 2);
         self.children.remove(idx - 2)
     }
@@ -277,8 +277,8 @@ impl DirEntry {
     fn substitute_entry(
         &mut self,
         idx: usize,
-        new_entry: (Str256, Arc<RamInode>),
-    ) -> Option<(Str256, Arc<RamInode>)> {
+        new_entry: (CStr256, Arc<RamInode>),
+    ) -> Option<(CStr256, Arc<RamInode>)> {
         assert!(idx >= 2);
         self.children.put_at(idx - 2, new_entry)
     }
@@ -315,7 +315,7 @@ impl DirEntry {
                 .skip_while(|(offset, _)| offset < &start_idx)
             {
                 visitor.visit(
-                    name.as_ref(),
+                    name.as_str().unwrap(),
                     child.metadata().ino as u64,
                     child.metadata().type_,
                     offset,
@@ -337,36 +337,6 @@ impl DirEntry {
     }
 }
 
-#[repr(C)]
-#[derive(Clone, PartialEq, PartialOrd, Eq, Ord)]
-pub struct Str256([u8; 256]);
-
-impl AsRef<str> for Str256 {
-    fn as_ref(&self) -> &str {
-        let len = self.0.iter().enumerate().find(|(_, &b)| b == 0).unwrap().0;
-        str::from_utf8(&self.0[0..len]).unwrap()
-    }
-}
-
-impl<'a> From<&'a str> for Str256 {
-    fn from(s: &'a str) -> Self {
-        let mut inner = [0u8; 256];
-        let len = if s.len() > NAME_MAX {
-            NAME_MAX
-        } else {
-            s.len()
-        };
-        inner[0..len].copy_from_slice(&s.as_bytes()[0..len]);
-        Self(inner)
-    }
-}
-
-impl core::fmt::Debug for Str256 {
-    fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result {
-        write!(f, "{}", self.as_ref())
-    }
-}
-
 impl RamInode {
     fn new_dir(fs: &Arc<RamFS>, mode: InodeMode, parent: &Weak<Self>) -> Arc<Self> {
         Arc::new_cyclic(|weak_self| {
@@ -439,7 +409,7 @@ impl RamInode {
     }
 }
 
-impl Inode for RamInode {
+impl PageCacheBackend for RamInode {
     fn read_page(&self, _idx: usize, _frame: &VmFrame) -> Result<()> {
         // do nothing
         Ok(())
@@ -450,6 +420,12 @@ impl Inode for RamInode {
         Ok(())
     }
 
+    fn npages(&self) -> usize {
+        self.0.read().metadata.blocks
+    }
+}
+
+impl Inode for RamInode {
     fn page_cache(&self) -> Option<Vmo<Full>> {
         self.0
             .read()
@@ -515,8 +491,9 @@ impl Inode for RamInode {
         self.0.read().metadata.size
     }
 
-    fn resize(&self, new_size: usize) {
-        self.0.write().resize(new_size)
+    fn resize(&self, new_size: usize) -> Result<()> {
+        self.0.write().resize(new_size);
+        Ok(())
     }
 
     fn atime(&self) -> Duration {
@@ -535,6 +512,10 @@ impl Inode for RamInode {
         self.0.write().metadata.mtime = time;
     }
 
+    fn ino(&self) -> u64 {
+        self.0.read().metadata.ino as _
+    }
+
     fn type_(&self) -> InodeType {
         self.0.read().metadata.type_
     }
@@ -780,7 +761,7 @@ impl Inode for RamInode {
             let (idx, inode) = self_dir
                 .get_entry(old_name)
                 .ok_or(Error::new(Errno::ENOENT))?;
-            self_dir.substitute_entry(idx, (Str256::from(new_name), inode));
+            self_dir.substitute_entry(idx, (CStr256::from(new_name), inode));
         } else {
             let (mut self_inode, mut target_inode) = write_lock_two_inodes(self, target);
             let self_dir = self_inode.inner.as_direntry_mut().unwrap();

services/libs/aster-std/src/fs/rootfs.rs

@@ -3,7 +3,7 @@ use crate::prelude::*;
 use super::fs_resolver::{FsPath, FsResolver};
 use super::procfs::ProcFS;
 use super::ramfs::RamFS;
-use super::utils::{InodeMode, InodeType, MountNode};
+use super::utils::{FileSystem, InodeMode, InodeType, MountNode};
 
 use cpio_decoder::{CpioDecoder, FileType};
 use lending_iterator::LendingIterator;
@@ -77,11 +77,18 @@ pub fn init(initramfs_buf: &[u8]) -> Result<()> {
     // Mount DevFS
     let dev_dentry = fs.lookup(&FsPath::try_from("/dev")?)?;
     dev_dentry.mount(RamFS::new())?;
+
     println!("[kernel] rootfs is ready");
 
     Ok(())
 }
 
+pub fn mount_fs_at(fs: Arc<dyn FileSystem>, fs_path: &FsPath) -> Result<()> {
+    let target_dentry = FsResolver::new().lookup(fs_path)?;
+    target_dentry.mount(fs)?;
+    Ok(())
+}
+
 static ROOT_MOUNT: Once<Arc<MountNode>> = Once::new();
 
 pub fn init_root_mount() {

services/libs/aster-std/src/fs/utils/inode.rs

@@ -1,4 +1,3 @@
-use aster_frame::vm::VmFrame;
 use aster_rights::Full;
 use core::time::Duration;
 use core2::io::{Error as IoError, ErrorKind as IoErrorKind, Result as IoResult, Write};
@@ -232,10 +231,12 @@ pub trait Inode: Any + Sync + Send {
         self.len() == 0
     }
 
-    fn resize(&self, new_size: usize);
+    fn resize(&self, new_size: usize) -> Result<()>;
 
     fn metadata(&self) -> Metadata;
 
+    fn ino(&self) -> u64;
+
     fn type_(&self) -> InodeType;
 
     fn mode(&self) -> InodeMode;
@@ -250,14 +251,6 @@ pub trait Inode: Any + Sync + Send {
 
     fn set_mtime(&self, time: Duration);
 
-    fn read_page(&self, idx: usize, frame: &VmFrame) -> Result<()> {
-        Err(Error::new(Errno::EISDIR))
-    }
-
-    fn write_page(&self, idx: usize, frame: &VmFrame) -> Result<()> {
-        Err(Error::new(Errno::EISDIR))
-    }
-
     fn page_cache(&self) -> Option<Vmo<Full>> {
         None
     }

services/libs/aster-std/src/fs/utils/mod.rs

@@ -11,7 +11,7 @@ pub use fs::{FileSystem, FsFlags, SuperBlock};
 pub use inode::{Inode, InodeMode, InodeType, Metadata};
 pub use ioctl::IoctlCmd;
 pub use mount::MountNode;
-pub use page_cache::PageCache;
+pub use page_cache::{PageCache, PageCacheBackend};
 pub use status_flags::StatusFlags;
 
 mod access_mode;
@@ -28,6 +28,8 @@ mod mount;
 mod page_cache;
 mod status_flags;
 
+use crate::prelude::*;
+
 #[derive(Copy, PartialEq, Eq, Clone, Debug)]
 pub enum SeekFrom {
     Start(usize),
@@ -43,3 +45,152 @@ pub const NAME_MAX: usize = 255;
 
 /// The upper limit for resolving symbolic links
 pub const SYMLINKS_MAX: usize = 40;
+
+pub type CStr256 = FixedCStr<256>;
+pub type Str16 = FixedStr<16>;
+pub type Str64 = FixedStr<64>;
+
+/// An owned C-compatible string with a fixed capacity of `N`.
+///
+/// The string is terminated with a null byte.
+#[repr(C)]
+#[derive(Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Pod)]
+pub struct FixedCStr<const N: usize>([u8; N]);
+
+impl<const N: usize> FixedCStr<N> {
+    pub fn len(&self) -> usize {
+        self.0.iter().position(|&b| b == 0).unwrap()
+    }
+
+    pub fn is_empty(&self) -> bool {
+        self.len() == 0
+    }
+
+    pub fn as_str(&self) -> Result<&str> {
+        Ok(alloc::str::from_utf8(self.as_bytes())?)
+    }
+
+    pub fn as_cstr(&self) -> Result<&CStr> {
+        Ok(CStr::from_bytes_with_nul(self.as_bytes_with_nul())?)
+    }
+
+    pub fn as_bytes(&self) -> &[u8] {
+        &self.0[0..self.len()]
+    }
+
+    pub fn as_bytes_with_nul(&self) -> &[u8] {
+        &self.0[0..=self.len()]
+    }
+}
+
+impl<'a, const N: usize> From<&'a [u8]> for FixedCStr<N> {
+    fn from(bytes: &'a [u8]) -> Self {
+        assert!(N > 0);
+
+        let mut inner = [0u8; N];
+        let len = {
+            let mut nul_byte_idx = match bytes.iter().position(|&b| b == 0) {
+                Some(idx) => idx,
+                None => bytes.len(),
+            };
+            if nul_byte_idx >= N {
+                nul_byte_idx = N - 1;
+            }
+            nul_byte_idx
+        };
+        inner[0..len].copy_from_slice(&bytes[0..len]);
+        Self(inner)
+    }
+}
+
+impl<'a, const N: usize> From<&'a str> for FixedCStr<N> {
+    fn from(string: &'a str) -> Self {
+        let bytes = string.as_bytes();
+        Self::from(bytes)
+    }
+}
+
+impl<'a, const N: usize> From<&'a CStr> for FixedCStr<N> {
+    fn from(cstr: &'a CStr) -> Self {
+        let bytes = cstr.to_bytes_with_nul();
+        Self::from(bytes)
+    }
+}
+
+impl<const N: usize> Default for FixedCStr<N> {
+    fn default() -> Self {
+        Self([0u8; N])
+    }
+}
+
+impl<const N: usize> Debug for FixedCStr<N> {
+    fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result {
+        match self.as_cstr() {
+            Ok(cstr) => write!(f, "{:?}", cstr),
+            Err(_) => write!(f, "{:?}", self.as_bytes()),
+        }
+    }
+}
+
+/// An owned string with a fixed capacity of `N`.
+#[repr(C)]
+#[derive(Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Pod)]
+pub struct FixedStr<const N: usize>([u8; N]);
+
+impl<const N: usize> FixedStr<N> {
+    pub fn len(&self) -> usize {
+        self.0.iter().position(|&b| b == 0).unwrap_or(N)
+    }
+
+    pub fn is_empty(&self) -> bool {
+        self.len() == 0
+    }
+
+    pub fn as_str(&self) -> Result<&str> {
+        Ok(alloc::str::from_utf8(self.as_bytes())?)
+    }
+
+    pub fn as_bytes(&self) -> &[u8] {
+        &self.0[0..self.len()]
+    }
+}
+
+impl<'a, const N: usize> From<&'a [u8]> for FixedStr<N> {
+    fn from(bytes: &'a [u8]) -> Self {
+        let mut inner = [0u8; N];
+        let len = {
+            let mut nul_byte_idx = match bytes.iter().position(|&b| b == 0) {
+                Some(idx) => idx,
+                None => bytes.len(),
+            };
+            if nul_byte_idx > N {
+                nul_byte_idx = N;
+            }
+            nul_byte_idx
+        };
+        inner[0..len].copy_from_slice(&bytes[0..len]);
+        Self(inner)
+    }
+}
+
+impl<'a, const N: usize> From<&'a str> for FixedStr<N> {
+    fn from(string: &'a str) -> Self {
+        let bytes = string.as_bytes();
+        Self::from(bytes)
+    }
+}
+
+impl<const N: usize> Default for FixedStr<N> {
+    fn default() -> Self {
+        Self([0u8; N])
+    }
+}
+
+impl<const N: usize> Debug for FixedStr<N> {
+    fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result {
+        match self.as_str() {
+            Ok(string) => write!(f, "{}", string),
+            Err(_) => write!(f, "{:?}", self.as_bytes()),
+        }
+    }
+}

services/libs/aster-std/src/fs/utils/page_cache.rs

@@ -1,4 +1,3 @@
-use super::Inode;
 use crate::prelude::*;
 use crate::vm::vmo::{get_page_idx_range, Pager, Vmo, VmoFlags, VmoOptions};
 use aster_rights::Full;
@@ -13,9 +12,9 @@ pub struct PageCache {
 }
 
 impl PageCache {
-    /// Creates an empty size page cache associated with a new inode.
-    pub fn new(backed_inode: Weak<dyn Inode>) -> Result<Self> {
-        let manager = Arc::new(PageCacheManager::new(backed_inode));
+    /// Creates an empty size page cache associated with a new backend.
+    pub fn new(backend: Weak<dyn PageCacheBackend>) -> Result<Self> {
+        let manager = Arc::new(PageCacheManager::new(backend));
         let pages = VmoOptions::<Full>::new(0)
             .flags(VmoFlags::RESIZABLE)
             .pager(manager.clone())
@@ -23,12 +22,12 @@ impl PageCache {
         Ok(Self { pages, manager })
     }
 
-    /// Creates a page cache associated with an existing inode.
+    /// Creates a page cache associated with an existing backend.
     ///
-    /// The `capacity` is the initial cache size required by the inode.
-    /// It is usually used the same size as the inode.
-    pub fn with_capacity(capacity: usize, backed_inode: Weak<dyn Inode>) -> Result<Self> {
-        let manager = Arc::new(PageCacheManager::new(backed_inode));
+    /// The `capacity` is the initial cache size required by the backend.
+    /// This size usually corresponds to the size of the backend.
+    pub fn with_capacity(capacity: usize, backend: Weak<dyn PageCacheBackend>) -> Result<Self> {
+        let manager = Arc::new(PageCacheManager::new(backend));
         let pages = VmoOptions::<Full>::new(capacity)
             .flags(VmoFlags::RESIZABLE)
             .pager(manager.clone())
@@ -36,7 +35,7 @@ impl PageCache {
         Ok(Self { pages, manager })
     }
 
-    /// Returns the Vmo object backed by inode.
+    /// Returns the Vmo object.
     // TODO: The capability is too high，restrict it to eliminate the possibility of misuse.
     //       For example, the `resize` api should be forbidded.
     pub fn pages(&self) -> Vmo<Full> {
@@ -44,10 +43,15 @@ impl PageCache {
     }
 
     /// Evict the data within a specified range from the page cache and persist
-    /// them to the disk.
+    /// them to the backend.
     pub fn evict_range(&self, range: Range<usize>) -> Result<()> {
         self.manager.evict_range(range)
     }
+
+    /// Returns the backend.
+    pub fn backend(&self) -> Arc<dyn PageCacheBackend> {
+        self.manager.backend()
+    }
 }
 
 impl Debug for PageCache {
@@ -61,33 +65,36 @@ impl Debug for PageCache {
 
 struct PageCacheManager {
     pages: Mutex<LruCache<usize, Page>>,
-    backed_inode: Weak<dyn Inode>,
+    backend: Weak<dyn PageCacheBackend>,
 }
 
 impl PageCacheManager {
-    pub fn new(backed_inode: Weak<dyn Inode>) -> Self {
+    pub fn new(backend: Weak<dyn PageCacheBackend>) -> Self {
         Self {
             pages: Mutex::new(LruCache::unbounded()),
-            backed_inode,
+            backend,
         }
     }
 
+    pub fn backend(&self) -> Arc<dyn PageCacheBackend> {
+        self.backend.upgrade().unwrap()
+    }
+
     pub fn evict_range(&self, range: Range<usize>) -> Result<()> {
         let page_idx_range = get_page_idx_range(&range);
         let mut pages = self.pages.lock();
-        for page_idx in page_idx_range {
-            if let Some(page) = pages.get_mut(&page_idx) {
+        for idx in page_idx_range {
+            if let Some(page) = pages.get_mut(&idx) {
                 if let PageState::Dirty = page.state() {
-                    self.backed_inode
-                        .upgrade()
-                        .unwrap()
-                        .write_page(page_idx, page.frame())?;
+                    let backend = self.backend();
+                    if idx < backend.npages() {
+                        backend.write_page(idx, page.frame())?;
                         page.set_state(PageState::UpToDate);
                     }
-            } else {
-                warn!("page {} is not in page cache, do nothing", page_idx);
                 }
             }
+        }
+
         Ok(())
     }
 }
@@ -101,53 +108,50 @@ impl Debug for PageCacheManager {
 }
 
 impl Pager for PageCacheManager {
-    fn commit_page(&self, offset: usize) -> Result<VmFrame> {
-        let page_idx = offset / PAGE_SIZE;
+    fn commit_page(&self, idx: usize) -> Result<VmFrame> {
         let mut pages = self.pages.lock();
-        let frame = if let Some(page) = pages.get(&page_idx) {
+        let frame = if let Some(page) = pages.get(&idx) {
             page.frame().clone()
         } else {
-            let backed_inode = self.backed_inode.upgrade().unwrap();
-            let page = if offset < backed_inode.len() {
+            let backend = self.backend();
+            let page = if idx < backend.npages() {
                 let mut page = Page::alloc()?;
-                backed_inode.read_page(page_idx, page.frame())?;
+                backend.read_page(idx, page.frame())?;
                 page.set_state(PageState::UpToDate);
                 page
             } else {
                 Page::alloc_zero()?
             };
             let frame = page.frame().clone();
-            pages.put(page_idx, page);
+            pages.put(idx, page);
             frame
         };
+
         Ok(frame)
     }
 
-    fn update_page(&self, offset: usize) -> Result<()> {
-        let page_idx = offset / PAGE_SIZE;
+    fn update_page(&self, idx: usize) -> Result<()> {
         let mut pages = self.pages.lock();
-        if let Some(page) = pages.get_mut(&page_idx) {
+        if let Some(page) = pages.get_mut(&idx) {
             page.set_state(PageState::Dirty);
         } else {
-            error!("page {} is not in page cache", page_idx);
-            panic!();
+            warn!("The page {} is not in page cache", idx);
         }
+
         Ok(())
     }
 
-    fn decommit_page(&self, offset: usize) -> Result<()> {
-        let page_idx = offset / PAGE_SIZE;
+    fn decommit_page(&self, idx: usize) -> Result<()> {
         let mut pages = self.pages.lock();
-        if let Some(page) = pages.pop(&page_idx) {
+        if let Some(page) = pages.pop(&idx) {
             if let PageState::Dirty = page.state() {
-                self.backed_inode
-                    .upgrade()
-                    .unwrap()
-                    .write_page(page_idx, page.frame())?
+                let backend = self.backend();
+                if idx < backend.npages() {
+                    backend.write_page(idx, page.frame())?;
                 }
-        } else {
-            warn!("page {} is not in page cache, do nothing", page_idx);
             }
+        }
+
         Ok(())
     }
 }
@@ -200,3 +204,13 @@ enum PageState {
     /// The page is available to read and write.
     Dirty,
 }
+
+/// This trait represents the backend for the page cache.
+pub trait PageCacheBackend: Sync + Send {
+    /// Reads a page from the backend.
+    fn read_page(&self, idx: usize, frame: &VmFrame) -> Result<()>;
+    /// Writes a page to the backend.
+    fn write_page(&self, idx: usize, frame: &VmFrame) -> Result<()>;
+    /// Returns the number of pages in the backend.
+    fn npages(&self) -> usize;
+}

services/libs/aster-std/src/lib.rs

@@ -19,6 +19,8 @@
 #![feature(register_tool)]
 #![feature(trait_upcasting)]
 #![feature(format_args_nl)]
+#![feature(int_roundings)]
+#![feature(step_trait)]
 #![register_tool(component_access_control)]
 
 use crate::{
@@ -73,6 +75,7 @@ fn init_thread() {
         current_thread!().tid()
     );
     net::lazy_init();
+    fs::lazy_init();
     // driver::pci::virtio::block::block_device_test();
     let thread = Thread::spawn_kernel_thread(ThreadOptions::new(|| {
         println!("[kernel] Hello world from kernel!");

services/libs/aster-std/src/prelude.rs

@@ -13,7 +13,7 @@ pub(crate) use alloc::sync::Weak;
 pub(crate) use alloc::vec;
 pub(crate) use alloc::vec::Vec;
 pub(crate) use aster_frame::config::PAGE_SIZE;
-pub(crate) use aster_frame::sync::{Mutex, MutexGuard, RwLock, SpinLock, SpinLockGuard};
+pub(crate) use aster_frame::sync::{Mutex, MutexGuard, RwLock, RwMutex, SpinLock, SpinLockGuard};
 pub(crate) use aster_frame::vm::Vaddr;
 pub(crate) use bitflags::bitflags;
 pub(crate) use core::any::Any;

services/libs/aster-std/src/process/clone.rs

@@ -379,13 +379,13 @@ fn clone_cpu_context(
 }
 
 fn clone_fs(
-    parent_fs: &Arc<RwLock<FsResolver>>,
+    parent_fs: &Arc<RwMutex<FsResolver>>,
     clone_flags: CloneFlags,
-) -> Arc<RwLock<FsResolver>> {
+) -> Arc<RwMutex<FsResolver>> {
     if clone_flags.contains(CloneFlags::CLONE_FS) {
         parent_fs.clone()
     } else {
-        Arc::new(RwLock::new(parent_fs.read().clone()))
+        Arc::new(RwMutex::new(parent_fs.read().clone()))
     }
 }
 

services/libs/aster-std/src/process/process/builder.rs

@@ -23,7 +23,7 @@ pub struct ProcessBuilder<'a> {
     envp: Option<Vec<CString>>,
     process_vm: Option<ProcessVm>,
     file_table: Option<Arc<Mutex<FileTable>>>,
-    fs: Option<Arc<RwLock<FsResolver>>>,
+    fs: Option<Arc<RwMutex<FsResolver>>>,
     umask: Option<Arc<RwLock<FileCreationMask>>>,
     resource_limits: Option<ResourceLimits>,
     sig_dispositions: Option<Arc<Mutex<SigDispositions>>>,
@@ -64,7 +64,7 @@ impl<'a> ProcessBuilder<'a> {
         self
     }
 
-    pub fn fs(&mut self, fs: Arc<RwLock<FsResolver>>) -> &mut Self {
+    pub fn fs(&mut self, fs: Arc<RwMutex<FsResolver>>) -> &mut Self {
         self.fs = Some(fs);
         self
     }
@@ -142,7 +142,7 @@ impl<'a> ProcessBuilder<'a> {
             .unwrap();
 
         let fs = fs
-            .or_else(|| Some(Arc::new(RwLock::new(FsResolver::new()))))
+            .or_else(|| Some(Arc::new(RwMutex::new(FsResolver::new()))))
             .unwrap();
 
         let umask = umask

services/libs/aster-std/src/process/process/mod.rs

@@ -64,7 +64,7 @@ pub struct Process {
     /// File table
     file_table: Arc<Mutex<FileTable>>,
     /// FsResolver
-    fs: Arc<RwLock<FsResolver>>,
+    fs: Arc<RwMutex<FsResolver>>,
     /// umask
     umask: Arc<RwLock<FileCreationMask>>,
     /// resource limits
@@ -84,7 +84,7 @@ impl Process {
         executable_path: String,
         process_vm: ProcessVm,
         file_table: Arc<Mutex<FileTable>>,
-        fs: Arc<RwLock<FsResolver>>,
+        fs: Arc<RwMutex<FsResolver>>,
         umask: Arc<RwLock<FileCreationMask>>,
         sig_dispositions: Arc<Mutex<SigDispositions>>,
         resource_limits: ResourceLimits,
@@ -496,7 +496,7 @@ impl Process {
         &self.file_table
     }
 
-    pub fn fs(&self) -> &Arc<RwLock<FsResolver>> {
+    pub fn fs(&self) -> &Arc<RwMutex<FsResolver>> {
         &self.fs
     }
 
@@ -595,7 +595,7 @@ mod test {
             String::new(),
             ProcessVm::alloc(),
             Arc::new(Mutex::new(FileTable::new())),
-            Arc::new(RwLock::new(FsResolver::new())),
+            Arc::new(RwMutex::new(FsResolver::new())),
             Arc::new(RwLock::new(FileCreationMask::default())),
             Arc::new(Mutex::new(SigDispositions::default())),
             ResourceLimits::default(),

services/libs/aster-std/src/time/mod.rs

@@ -104,3 +104,25 @@ pub fn now_as_duration(clock_id: &ClockID) -> Result<Duration> {
         }
     }
 }
+
+/// Unix time measures time by the number of seconds that have elapsed since
+/// the Unix epoch, without adjustments made due to leap seconds.
+#[repr(C)]
+#[derive(Debug, Default, Copy, Clone, Pod)]
+pub struct UnixTime {
+    sec: u32,
+}
+
+impl From<Duration> for UnixTime {
+    fn from(duration: Duration) -> Self {
+        Self {
+            sec: duration.as_secs() as u32,
+        }
+    }
+}
+
+impl From<UnixTime> for Duration {
+    fn from(time: UnixTime) -> Self {
+        Duration::from_secs(time.sec as _)
+    }
+}

services/libs/aster-std/src/vm/vmo/mod.rs

@@ -154,7 +154,7 @@ impl VmoInner {
         }
         let frame = match &self.pager {
             None => VmAllocOptions::new(1).alloc_single()?,
-            Some(pager) => pager.commit_page(offset)?,
+            Some(pager) => pager.commit_page(page_idx)?,
         };
         self.insert_frame(page_idx, frame);
         Ok(())
@@ -164,7 +164,7 @@ impl VmoInner {
         let page_idx = offset / PAGE_SIZE;
         if self.committed_pages.remove(&page_idx).is_some() {
             if let Some(pager) = &self.pager {
-                pager.decommit_page(offset)?;
+                pager.decommit_page(page_idx)?;
             }
         }
         Ok(())
@@ -302,7 +302,7 @@ impl Vmo_ {
         if let Some(pager) = &self.inner.lock().pager {
             let page_idx_range = get_page_idx_range(&write_range);
             for page_idx in page_idx_range {
-                pager.update_page(page_idx * PAGE_SIZE)?;
+                pager.update_page(page_idx)?;
             }
         }
         Ok(())

services/libs/aster-std/src/vm/vmo/pager.rs

@@ -11,7 +11,7 @@ use aster_frame::vm::VmFrame;
 /// Finally, when a frame is no longer needed (i.e., on decommits),
 /// the frame pager will also be notified.
 pub trait Pager: Send + Sync {
-    /// Ask the pager to provide a frame at a specified offset (in bytes).
+    /// Ask the pager to provide a frame at a specified index.
     ///
     /// After a page of a VMO is committed, the VMO shall not call this method
     /// again until the page is decommitted. But a robust implementation of
@@ -22,13 +22,10 @@ pub trait Pager: Send + Sync {
     /// and is to be committed again, then the pager is free to return
     /// whatever frame that may or may not be the same as the last time.
     ///
-    /// It is up to the pager to decide the range of valid offsets.
-    ///
-    /// The offset will be rounded down to page boundary.
-    fn commit_page(&self, offset: usize) -> Result<VmFrame>;
+    /// It is up to the pager to decide the range of valid indices.
+    fn commit_page(&self, idx: usize) -> Result<VmFrame>;
 
-    /// Notify the pager that the frame at a specified offset (in bytes)
-    /// has been updated.
+    /// Notify the pager that the frame at a specified index has been updated.
     ///
     /// Being aware of the updates allow the pager (e.g., an inode) to
     /// know which pages are dirty and only write back the _dirty_ pages back
@@ -38,12 +35,9 @@ pub trait Pager: Send + Sync {
     /// But a robust implementation of `Pager` should not make
     /// such an assumption for its correctness; instead, it should simply ignore the
     /// call or return an error.
-    ///
-    /// The offset will be rounded down to page boundary.
-    fn update_page(&self, offset: usize) -> Result<()>;
+    fn update_page(&self, idx: usize) -> Result<()>;
 
-    /// Notify the pager that the frame at the specified offset (in bytes)
-    /// has been decommitted.
+    /// Notify the pager that the frame at the specified index has been decommitted.
     ///
     /// Knowing that a frame is no longer needed, the pager (e.g., an inode)
     /// can free the frame after writing back its data to the disk.
@@ -52,7 +46,5 @@ pub trait Pager: Send + Sync {
     /// But a robust implementation of `Pager` should not make
     /// such an assumption for its correctness; instead, it should simply ignore the
     /// call or return an error.
-    ///
-    /// The offset will be rounded down to page boundary.
-    fn decommit_page(&self, offset: usize) -> Result<()>;
+    fn decommit_page(&self, idx: usize) -> Result<()>;
 }

services/libs/aster-util/src/id_allocator.rs

@@ -1,15 +1,16 @@
 use bitvec::prelude::BitVec;
+use core::fmt::Debug;
 
-/// An id allocator with BitVec.
-/// The true bit means the id is allocated and vice versa.
-#[derive(Clone, Debug)]
+/// An id allocator implemented by the bitmap.
+/// The true bit implies that the id is allocated, and vice versa.
+#[derive(Clone)]
 pub struct IdAlloc {
-    bitset: BitVec,
+    bitset: BitVec<u8>,
     first_available_id: usize,
 }
 
 impl IdAlloc {
-    /// Constructs a new id allocator with the maximum capacity.
+    /// Constructs a new id allocator with a maximum capacity.
     pub fn with_capacity(capacity: usize) -> Self {
         let mut bitset = BitVec::with_capacity(capacity);
         bitset.resize(capacity, false);
@@ -19,9 +20,33 @@ impl IdAlloc {
         }
     }
 
-    /// Allocates and returns an id.
+    /// Constructs a new id allocator from a slice of `u8` bytes and a maximum capacity.
     ///
-    /// Returns None if can not allocate.
+    /// The slice of `u8` bytes is the raw data of a bitmap.
+    pub fn from_bytes_with_capacity(slice: &[u8], capacity: usize) -> Self {
+        let bitset = if capacity > slice.len() * 8 {
+            let mut bitset = BitVec::from_slice(slice);
+            bitset.resize(capacity, false);
+            bitset
+        } else {
+            let mut bitset = BitVec::from_slice(&slice[..capacity.div_ceil(8)]);
+            bitset.truncate(capacity);
+            bitset
+        };
+
+        let first_available_id = (0..bitset.len())
+            .find(|&i| !bitset[i])
+            .map_or(bitset.len(), |i| i);
+
+        Self {
+            bitset,
+            first_available_id,
+        }
+    }
+
+    /// Allocates and returns a new `id`.
+    ///
+    /// If allocation is not possible, it returns `None`.
     pub fn alloc(&mut self) -> Option<usize> {
         if self.first_available_id < self.bitset.len() {
             let id = self.first_available_id;
@@ -35,9 +60,11 @@ impl IdAlloc {
         }
     }
 
-    /// Frees the allocated id.
+    /// Releases the allocated `id`.
     ///
-    /// This panics if the id is out of bounds.
+    /// # Panic
+    ///
+    /// If the `id` is out of bounds, this method will panic.
     pub fn free(&mut self, id: usize) {
         debug_assert!(self.is_allocated(id));
 
@@ -47,10 +74,26 @@ impl IdAlloc {
         }
     }
 
-    /// Returns true is the id is allocated.
+    /// Returns true if the `id` is allocated.
     ///
-    /// This panics if the id is out of bounds.
+    /// # Panic
+    ///
+    /// If the `id` is out of bounds, this method will panic.
     pub fn is_allocated(&self, id: usize) -> bool {
         self.bitset[id]
     }
+
+    /// Views the id allocator as a slice of `u8` bytes.
+    pub fn as_bytes(&self) -> &[u8] {
+        self.bitset.as_raw_slice()
+    }
+}
+
+impl Debug for IdAlloc {
+    fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result {
+        f.debug_struct("IdAlloc")
+            .field("len", &self.bitset.len())
+            .field("first_available_id", &self.first_available_id)
+            .finish()
+    }
 }

services/libs/aster-util/src/lib.rs

@@ -1,6 +1,7 @@
 //! The util of Asterinas.
 #![no_std]
 #![forbid(unsafe_code)]
+#![feature(int_roundings)]
 
 extern crate alloc;