Merge branch 'dev' into geminilab

README.md

@@ -8,7 +8,7 @@ Going to be the next generation teaching operating system.
 
 Supported architectures: x86_64, RISCV32/64, AArch64, MIPS32
 
-Tested boards: QEMU, HiFive Unleashed, x86_64 PC (i5/i7), Raspberry Pi 3B+
+Tested boards: QEMU, HiFive Unleashed, x86_64 PC (i5/i7), Raspberry Pi 3B+, Kendryte K210 and FPGA running Rocket Chip
 
 ![demo](./docs/2_OSLab/os2atc/demo.png)
 

crate/memory/src/memory_set/handler/byframe.rs

@@ -22,6 +22,20 @@ impl<T: FrameAllocator> MemoryHandler for ByFrame<T> {
         pt.unmap(addr);
     }
 
+    fn clone_map(
+        &self,
+        pt: &mut PageTable,
+        with: &Fn(&mut FnMut()),
+        addr: VirtAddr,
+        attr: &MemoryAttr,
+    ) {
+        let data = Vec::from(pt.get_page_slice_mut(addr));
+        with(&mut || {
+            self.map(pt, addr, attr);
+            pt.get_page_slice_mut(addr).copy_from_slice(&data);
+        });
+    }
+
     fn handle_page_fault(&self, _pt: &mut PageTable, _addr: VirtAddr) -> bool {
         false
     }

crate/memory/src/memory_set/handler/delay.rs

@@ -16,13 +16,6 @@ impl<T: FrameAllocator> MemoryHandler for Delay<T> {
         attr.apply(entry);
     }
 
-    fn map_eager(&self, pt: &mut PageTable, addr: VirtAddr, attr: &MemoryAttr) {
-        let target = self.allocator.alloc().expect("failed to alloc frame");
-        let entry = pt.map(addr, target);
-        entry.set_present(true);
-        attr.apply(entry);
-    }
-
     fn unmap(&self, pt: &mut PageTable, addr: VirtAddr) {
         let entry = pt.get_entry(addr).expect("failed to get entry");
         if entry.present() {
@@ -34,6 +27,30 @@ impl<T: FrameAllocator> MemoryHandler for Delay<T> {
         pt.unmap(addr);
     }
 
+    fn clone_map(
+        &self,
+        pt: &mut PageTable,
+        with: &Fn(&mut FnMut()),
+        addr: VirtAddr,
+        attr: &MemoryAttr,
+    ) {
+        let entry = pt.get_entry(addr).expect("failed to get entry");
+        if entry.present() {
+            // eager map and copy data
+            let data = Vec::from(pt.get_page_slice_mut(addr));
+            with(&mut || {
+                let target = self.allocator.alloc().expect("failed to alloc frame");
+                let target_data = pt.get_page_slice_mut(addr);
+                let entry = pt.map(addr, target);
+                target_data.copy_from_slice(&data);
+                attr.apply(entry);
+            });
+        } else {
+            // delay map
+            with(&mut || self.map(pt, addr, attr));
+        }
+    }
+
     fn handle_page_fault(&self, pt: &mut PageTable, addr: VirtAddr) -> bool {
         let entry = pt.get_entry(addr).expect("failed to get entry");
         if entry.present() {
@@ -44,6 +61,11 @@ impl<T: FrameAllocator> MemoryHandler for Delay<T> {
         entry.set_target(frame);
         entry.set_present(true);
         entry.update();
+        //init with zero for delay mmap mode
+        let data = pt.get_page_slice_mut(addr);
+        for x in data {
+            *x = 0;
+        }
         true
     }
 }

crate/memory/src/memory_set/handler/file.rs

@@ -0,0 +1,108 @@
+use super::*;
+
+/// Delay mapping a page to an area of a file.
+#[derive(Clone)]
+pub struct File<F, T> {
+    pub file: F,
+    pub mem_start: usize,
+    pub file_start: usize,
+    pub file_end: usize,
+    pub allocator: T,
+}
+
+pub trait Read: Clone + Send + Sync + 'static {
+    fn read_at(&self, offset: usize, buf: &mut [u8]) -> usize;
+}
+
+impl<F: Read, T: FrameAllocator> MemoryHandler for File<F, T> {
+    fn box_clone(&self) -> Box<MemoryHandler> {
+        Box::new(self.clone())
+    }
+
+    fn map(&self, pt: &mut PageTable, addr: usize, attr: &MemoryAttr) {
+        let entry = pt.map(addr, 0);
+        entry.set_present(false);
+        attr.apply(entry);
+    }
+
+    fn unmap(&self, pt: &mut PageTable, addr: usize) {
+        let entry = pt.get_entry(addr).expect("failed to get entry");
+        if entry.present() {
+            self.allocator.dealloc(entry.target());
+        }
+
+        // PageTable::unmap requires page to be present
+        entry.set_present(true);
+        pt.unmap(addr);
+    }
+
+    fn clone_map(
+        &self,
+        pt: &mut PageTable,
+        with: &Fn(&mut FnMut()),
+        addr: usize,
+        attr: &MemoryAttr,
+    ) {
+        let entry = pt.get_entry(addr).expect("failed to get entry");
+        if entry.present() && !attr.readonly {
+            // eager map and copy data
+            let data = Vec::from(pt.get_page_slice_mut(addr));
+            with(&mut || {
+                let target = self.allocator.alloc().expect("failed to alloc frame");
+                let target_data = pt.get_page_slice_mut(addr);
+                let entry = pt.map(addr, target);
+                target_data.copy_from_slice(&data);
+                attr.apply(entry);
+            });
+        } else {
+            // delay map
+            with(&mut || self.map(pt, addr, attr));
+        }
+    }
+
+    fn handle_page_fault(&self, pt: &mut PageTable, addr: usize) -> bool {
+        let addr = addr & !(PAGE_SIZE - 1);
+        let entry = pt.get_entry(addr).expect("failed to get entry");
+        if entry.present() {
+            return false;
+        }
+        let frame = self.allocator.alloc().expect("failed to alloc frame");
+        entry.set_target(frame);
+        entry.set_present(true);
+        let writable = entry.writable();
+        entry.set_writable(true);
+        entry.update();
+
+        self.fill_data(pt, addr);
+
+        let entry = pt.get_entry(addr).expect("failed to get entry");
+        entry.set_writable(writable);
+        entry.update();
+
+        true
+    }
+}
+
+impl<F: Read, T: FrameAllocator> File<F, T> {
+    fn fill_data(&self, pt: &mut PageTable, addr: VirtAddr) {
+        let data = pt.get_page_slice_mut(addr);
+        let file_offset = addr + self.file_start - self.mem_start;
+        let read_size = (self.file_end as isize - file_offset as isize)
+            .min(PAGE_SIZE as isize)
+            .max(0) as usize;
+        let read_size = self.file.read_at(file_offset, &mut data[..read_size]);
+        if read_size != PAGE_SIZE {
+            data[read_size..].iter_mut().for_each(|x| *x = 0);
+        }
+    }
+}
+
+impl<F, T> Debug for File<F, T> {
+    fn fmt(&self, f: &mut Formatter<'_>) -> Result<(), Error> {
+        f.debug_struct("FileHandler")
+            .field("mem_start", &self.mem_start)
+            .field("file_start", &self.file_start)
+            .field("file_end", &self.file_end)
+            .finish()
+    }
+}

crate/memory/src/memory_set/handler/linear.rs

@@ -20,6 +20,16 @@ impl MemoryHandler for Linear {
         pt.unmap(addr);
     }
 
+    fn clone_map(
+        &self,
+        pt: &mut PageTable,
+        with: &Fn(&mut FnMut()),
+        addr: VirtAddr,
+        attr: &MemoryAttr,
+    ) {
+        with(&mut || self.map(pt, addr, attr));
+    }
+
     fn handle_page_fault(&self, _pt: &mut PageTable, _addr: VirtAddr) -> bool {
         false
     }

crate/memory/src/memory_set/handler/mod.rs

@@ -1,23 +1,28 @@
 use super::*;
 
 // here may be a interesting part for lab
-pub trait MemoryHandler: Debug + 'static {
+pub trait MemoryHandler: Debug + Send + Sync + 'static {
     fn box_clone(&self) -> Box<MemoryHandler>;
 
     /// Map `addr` in the page table
     /// Should set page flags here instead of in page_fault_handler
     fn map(&self, pt: &mut PageTable, addr: VirtAddr, attr: &MemoryAttr);
 
-    /// Map `addr` in the page table eagerly (i.e. no delay allocation)
-    /// Should set page flags here instead of in page_fault_handler
-    fn map_eager(&self, pt: &mut PageTable, addr: VirtAddr, attr: &MemoryAttr) {
-        // override this when pages are allocated lazily
-        self.map(pt, addr, attr);
-    }
-
     /// Unmap `addr` in the page table
     fn unmap(&self, pt: &mut PageTable, addr: VirtAddr);
 
+    /// Clone map `addr` from one page table to another.
+    /// `pt` is the current active page table.
+    /// `with` is the `InactivePageTable::with` function.
+    /// Call `with` then use `pt` as target page table inside.
+    fn clone_map(
+        &self,
+        pt: &mut PageTable,
+        with: &Fn(&mut FnMut()),
+        addr: VirtAddr,
+        attr: &MemoryAttr,
+    );
+
     /// Handle page fault on `addr`
     /// Return true if success, false if error
     fn handle_page_fault(&self, pt: &mut PageTable, addr: VirtAddr) -> bool;
@@ -29,16 +34,18 @@ impl Clone for Box<MemoryHandler> {
     }
 }
 
-pub trait FrameAllocator: Debug + Clone + 'static {
+pub trait FrameAllocator: Debug + Clone + Send + Sync + 'static {
     fn alloc(&self) -> Option<PhysAddr>;
     fn dealloc(&self, target: PhysAddr);
 }
 
 mod byframe;
 mod delay;
+mod file;
 mod linear;
 //mod swap;
 
 pub use self::byframe::ByFrame;
 pub use self::delay::Delay;
+pub use self::file::{File, Read};
 pub use self::linear::Linear;

crate/memory/src/memory_set/mod.rs

@@ -3,6 +3,7 @@
 
 use alloc::{boxed::Box, string::String, vec::Vec};
 use core::fmt::{Debug, Error, Formatter};
+use core::mem::size_of;
 
 use crate::paging::*;
 
@@ -23,8 +24,6 @@ pub struct MemoryArea {
     name: &'static str,
 }
 
-unsafe impl Send for MemoryArea {}
-
 impl MemoryArea {
     /*
      **  @brief  get slice of the content in the memory area
@@ -54,15 +53,27 @@ impl MemoryArea {
     pub fn contains(&self, addr: VirtAddr) -> bool {
         addr >= self.start_addr && addr < self.end_addr
     }
-    /// Check the array is within the readable memory
-    fn check_read_array<S>(&self, ptr: *const S, count: usize) -> bool {
+    /// Check the array is within the readable memory.
+    /// Return the size of space covered in the area.
+    fn check_read_array<S>(&self, ptr: *const S, count: usize) -> usize {
         // page align
-        ptr as usize >= Page::of_addr(self.start_addr).start_address()
-            && unsafe { ptr.add(count) as usize } < Page::of_addr(self.end_addr + PAGE_SIZE - 1).start_address()
+        let min_bound = (ptr as usize).max(Page::of_addr(self.start_addr).start_address());
+        let max_bound = unsafe { ptr.add(count) as usize }
+            .min(Page::of_addr(self.end_addr + PAGE_SIZE - 1).start_address());
+        if max_bound >= min_bound {
+            max_bound - min_bound
+        } else {
+            0
+        }
     }
-    /// Check the array is within the writable memory
-    fn check_write_array<S>(&self, ptr: *mut S, count: usize) -> bool {
-        !self.attr.readonly && self.check_read_array(ptr, count)
+    /// Check the array is within the writable memory.
+    /// Return the size of space covered in the area.
+    fn check_write_array<S>(&self, ptr: *mut S, count: usize) -> usize {
+        if self.attr.readonly {
+            0
+        } else {
+            self.check_read_array(ptr, count)
+        }
     }
     /// Check the null-end C string is within the readable memory, and is valid.
     /// If so, clone it to a String.
@@ -86,31 +97,13 @@ impl MemoryArea {
         let p3 = Page::of_addr(end_addr - 1) + 1;
         !(p1 <= p2 || p0 >= p3)
     }
-    /*
-     **  @brief  map the memory area to the physice address in a page table
-     **  @param  pt: &mut T::Active   the page table to use
-     **  @retval none
-     */
+    /// Map all pages in the area to page table `pt`
     fn map(&self, pt: &mut PageTable) {
         for page in Page::range_of(self.start_addr, self.end_addr) {
             self.handler.map(pt, page.start_address(), &self.attr);
         }
     }
-    /*
-     **  @brief  map the memory area to the physice address in a page table eagerly
-     **  @param  pt: &mut T::Active   the page table to use
-     **  @retval none
-     */
-    fn map_eager(&self, pt: &mut PageTable) {
-        for page in Page::range_of(self.start_addr, self.end_addr) {
-            self.handler.map_eager(pt, page.start_address(), &self.attr);
-        }
-    }
-    /*
-     **  @brief  unmap the memory area from the physice address in a page table
-     **  @param  pt: &mut T::Active   the page table to use
-     **  @retval none
-     */
+    /// Unmap all pages in the area from page table `pt`
     fn unmap(&self, pt: &mut PageTable) {
         for page in Page::range_of(self.start_addr, self.end_addr) {
             self.handler.unmap(pt, page.start_address());
@@ -204,28 +197,60 @@ impl<T: InactivePageTable> MemorySet<T> {
         }
     }
     /// Check the pointer is within the readable memory
-    pub fn check_read_ptr<S>(&self, ptr: *const S) -> VMResult<()> {
-        self.check_read_array(ptr, 1)
+    pub unsafe fn check_read_ptr<S>(&self, ptr: *const S) -> VMResult<&'static S> {
+        self.check_read_array(ptr, 1).map(|s| &s[0])
     }
     /// Check the pointer is within the writable memory
-    pub fn check_write_ptr<S>(&self, ptr: *mut S) -> VMResult<()> {
-        self.check_write_array(ptr, 1)
+    pub unsafe fn check_write_ptr<S>(&self, ptr: *mut S) -> VMResult<&'static mut S> {
+        self.check_write_array(ptr, 1).map(|s| &mut s[0])
     }
     /// Check the array is within the readable memory
-    pub fn check_read_array<S>(&self, ptr: *const S, count: usize) -> VMResult<()> {
-        self.areas
-            .iter()
-            .find(|area| area.check_read_array(ptr, count))
-            .map(|_| ())
-            .ok_or(VMError::InvalidPtr)
+    pub unsafe fn check_read_array<S>(
+        &self,
+        ptr: *const S,
+        count: usize,
+    ) -> VMResult<&'static [S]> {
+        let mut valid_size = 0;
+        for area in self.areas.iter() {
+            valid_size += area.check_read_array(ptr, count);
+            if valid_size == size_of::<S>() * count {
+                return Ok(core::slice::from_raw_parts(ptr, count));
+            }
+        }
+        Err(VMError::InvalidPtr)
     }
     /// Check the array is within the writable memory
-    pub fn check_write_array<S>(&self, ptr: *mut S, count: usize) -> VMResult<()> {
-        self.areas
-            .iter()
-            .find(|area| area.check_write_array(ptr, count))
-            .map(|_| ())
-            .ok_or(VMError::InvalidPtr)
+    pub unsafe fn check_write_array<S>(
+        &self,
+        ptr: *mut S,
+        count: usize,
+    ) -> VMResult<&'static mut [S]> {
+        let mut valid_size = 0;
+        for area in self.areas.iter() {
+            valid_size += area.check_write_array(ptr, count);
+            if valid_size == size_of::<S>() * count {
+                return Ok(core::slice::from_raw_parts_mut(ptr, count));
+            }
+        }
+        Err(VMError::InvalidPtr)
+    }
+    /// Check the null-end C string pointer array
+    /// Used for getting argv & envp
+    pub unsafe fn check_and_clone_cstr_array(
+        &self,
+        mut argv: *const *const u8,
+    ) -> VMResult<Vec<String>> {
+        let mut args = Vec::new();
+        loop {
+            let cstr = *self.check_read_ptr(argv)?;
+            if cstr.is_null() {
+                break;
+            }
+            let arg = self.check_and_clone_cstr(cstr)?;
+            args.push(arg);
+            argv = argv.add(1);
+        }
+        Ok(args)
     }
     /// Check the null-end C string is within the readable memory, and is valid.
     /// If so, clone it to a String.
@@ -283,7 +308,15 @@ impl<T: InactivePageTable> MemorySet<T> {
             name,
         };
         self.page_table.edit(|pt| area.map(pt));
-        self.areas.push(area);
+        // keep order by start address
+        let idx = self
+            .areas
+            .iter()
+            .enumerate()
+            .find(|(_, other)| start_addr < other.start_addr)
+            .map(|(i, _)| i)
+            .unwrap_or(self.areas.len());
+        self.areas.insert(idx, area);
     }
 
     /*
@@ -474,20 +507,29 @@ impl<T: InactivePageTable> MemorySet<T> {
             None => false,
         }
     }
-}
 
-impl<T: InactivePageTable> Clone for MemorySet<T> {
-    fn clone(&self) -> Self {
-        let mut page_table = T::new();
+    pub fn clone(&mut self) -> Self {
+        let new_page_table = T::new();
+        let Self {
+            ref mut page_table,
+            ref areas,
+            ..
+        } = self;
         page_table.edit(|pt| {
-            // without CoW, we should allocate the pages eagerly
-            for area in self.areas.iter() {
-                area.map_eager(pt);
+            for area in areas.iter() {
+                for page in Page::range_of(area.start_addr, area.end_addr) {
+                    area.handler.clone_map(
+                        pt,
+                        &|f| unsafe { new_page_table.with(f) },
+                        page.start_address(),
+                        &area.attr,
+                    );
+                }
             }
         });
         MemorySet {
-            areas: self.areas.clone(),
-            page_table,
+            areas: areas.clone(),
+            page_table: new_page_table,
         }
     }
 }

kernel/Cargo.lock

@@ -111,7 +111,7 @@ dependencies = [
 
 [[package]]
 name = "buddy_system_allocator"
-version = "0.1.2"
+version = "0.3.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 dependencies = [
  "spin 0.5.0 (registry+https://github.com/rust-lang/crates.io-index)",
@@ -271,11 +271,6 @@ name = "nodrop"
 version = "0.1.13"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 
-[[package]]
-name = "once"
-version = "0.3.3"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-
 [[package]]
 name = "os_bootinfo"
 version = "0.2.1"
@@ -395,7 +390,7 @@ dependencies = [
  "bitmap-allocator 0.1.0 (git+https://github.com/rcore-os/bitmap-allocator)",
  "bitvec 0.11.0 (git+https://github.com/myrrlyn/bitvec.git?rev=ed2aec38bfb5b1116e3585b1574c50655b9c85ec)",
  "bootloader 0.4.0 (git+https://github.com/rcore-os/bootloader)",
- "buddy_system_allocator 0.1.2 (registry+https://github.com/rust-lang/crates.io-index)",
+ "buddy_system_allocator 0.3.4 (registry+https://github.com/rust-lang/crates.io-index)",
  "cc 1.0.36 (registry+https://github.com/rust-lang/crates.io-index)",
  "console-traits 0.3.0 (registry+https://github.com/rust-lang/crates.io-index)",
  "device_tree 1.0.3 (git+https://github.com/rcore-os/device_tree-rs)",
@@ -404,7 +399,6 @@ dependencies = [
  "lazy_static 1.3.0 (registry+https://github.com/rust-lang/crates.io-index)",
  "log 0.4.6 (registry+https://github.com/rust-lang/crates.io-index)",
  "mips 0.2.0 (registry+https://github.com/rust-lang/crates.io-index)",
- "once 0.3.3 (registry+https://github.com/rust-lang/crates.io-index)",
  "paste 0.1.5 (registry+https://github.com/rust-lang/crates.io-index)",
  "pc-keyboard 0.5.0 (registry+https://github.com/rust-lang/crates.io-index)",
  "pci 0.0.1 (git+https://github.com/rcore-os/pci-rs)",
@@ -699,7 +693,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 "checksum bitvec 0.11.0 (git+https://github.com/myrrlyn/bitvec.git?rev=ed2aec38bfb5b1116e3585b1574c50655b9c85ec)" = "<none>"
 "checksum bitvec 0.9.0 (registry+https://github.com/rust-lang/crates.io-index)" = "cfadef5c4e2c2e64067b9ecc061179f12ac7ec65ba613b1f60f3972bbada1f5b"
 "checksum bootloader 0.4.0 (git+https://github.com/rcore-os/bootloader)" = "<none>"
-"checksum buddy_system_allocator 0.1.2 (registry+https://github.com/rust-lang/crates.io-index)" = "2ed828f1e227d6e32b998d6375b67fd63ac5389d50b23f258ce151d22b6cc595"
+"checksum buddy_system_allocator 0.3.4 (registry+https://github.com/rust-lang/crates.io-index)" = "94a6c0143a07fea0db2f4b43cb9540dcc7c17af8a7beafdf2184e5e4e35aae91"
 "checksum byteorder 1.3.1 (registry+https://github.com/rust-lang/crates.io-index)" = "a019b10a2a7cdeb292db131fc8113e57ea2a908f6e7894b0c3c671893b65dbeb"
 "checksum cast 0.2.2 (registry+https://github.com/rust-lang/crates.io-index)" = "926013f2860c46252efceabb19f4a6b308197505082c609025aa6706c011d427"
 "checksum cc 1.0.36 (registry+https://github.com/rust-lang/crates.io-index)" = "a0c56216487bb80eec9c4516337b2588a4f2a2290d72a1416d930e4dcdb0c90d"
@@ -723,7 +717,6 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 "checksum managed 0.7.1 (registry+https://github.com/rust-lang/crates.io-index)" = "fdcec5e97041c7f0f1c5b7d93f12e57293c831c646f4cc7a5db59460c7ea8de6"
 "checksum mips 0.2.0 (registry+https://github.com/rust-lang/crates.io-index)" = "4cbf449a63e4db77af9f662d6b42068c0925e779a3a7c70ad02f191cf1e6c802"
 "checksum nodrop 0.1.13 (registry+https://github.com/rust-lang/crates.io-index)" = "2f9667ddcc6cc8a43afc9b7917599d7216aa09c463919ea32c59ed6cac8bc945"
-"checksum once 0.3.3 (registry+https://github.com/rust-lang/crates.io-index)" = "931fb7a4cf34610cf6cbe58d52a8ca5ef4c726d4e2e178abd0dc13a6551c6d73"
 "checksum os_bootinfo 0.2.1 (registry+https://github.com/rust-lang/crates.io-index)" = "66481dbeb5e773e7bd85b63cd6042c30786f834338288c5ec4f3742673db360a"
 "checksum paste 0.1.5 (registry+https://github.com/rust-lang/crates.io-index)" = "1f4a4a1c555c6505821f9d58b8779d0f630a6b7e4e1be24ba718610acf01fa79"
 "checksum paste-impl 0.1.5 (registry+https://github.com/rust-lang/crates.io-index)" = "26e796e623b8b257215f27e6c80a5478856cae305f5b59810ff9acdaa34570e6"

kernel/Cargo.toml

@@ -23,6 +23,8 @@ default = ["sv39"]
 # Page table sv39 or sv48 (for riscv64)
 sv39 = []
 board_u540 = ["sv39", "link_user"]
+board_k210 = ["sv39", "link_user"]
+board_rocket_chip = ["sv39", "link_user"]
 # (for aarch64 RaspberryPi3)
 nographic = []
 board_raspi3 = ["bcm2837", "link_user"]
@@ -38,6 +40,8 @@ board_pc = ["link_user"]
 link_user = []
 # Run cmdline instead of user shell, useful for automatic testing
 run_cmdline = []
+# Add performance profiling
+profile = []
 
 [profile.dev]
 # MUST >= 2 : Enable RVO to avoid stack overflow
@@ -46,7 +50,6 @@ opt-level = 2
 [dependencies]
 log = "0.4"
 spin = "0.5"
-once = "0.3"
 xmas-elf = "0.6"
 bitflags = "1.0"
 bit_field = "0.9"
@@ -54,7 +57,7 @@ volatile = "0.2"
 heapless = "0.4"
 bitvec = { git = "https://github.com/myrrlyn/bitvec.git", rev = "ed2aec38bfb5b1116e3585b1574c50655b9c85ec", default-features = false, features = ["alloc"] }
 console-traits = "0.3"
-buddy_system_allocator = "0.1"
+buddy_system_allocator = "0.3"
 pci = { git = "https://github.com/rcore-os/pci-rs" }
 device_tree = { git = "https://github.com/rcore-os/device_tree-rs" }
 isomorphic_drivers = { git = "https://github.com/rcore-os/isomorphic_drivers" }

kernel/Makefile

@@ -23,13 +23,16 @@
 #   smp     = 1 | 2 | ...        SMP core number
 #   graphic = on | off           Enable/disable qemu graphical output
 #   board   = none               Running on QEMU
-#         | pc                  Only available on x86_64, run on real pc
+#         | pc                   Only available on x86_64, run on real pc
 #         | u540                 Only available on riscv64, run on HiFive U540, use Sv39
+#         | k210                 Only available on riscv64, run on K210, use Sv39
+#         | rocket_chip          Only available on riscv64, run on Rocket Chip, use Sv39
 #         | raspi3               Only available on aarch64, run on Raspberry Pi 3 Model B/B+
 #   pci_passthru = 0000:00:00.1  Only available on x86_64, passthrough the specified PCI device
 #   init = /bin/ls               Only available on riscv64, run specified program instead of user shell
 #   extra_nic = on | off         Only available on x86_64, add an additional e1000 nic
 #   u_boot = /path/to/u-boot.bin Only available on aarch64, use u-boot to boot rcore
+# . extra_features = profile | ... Add additional features
 
 arch ?= riscv64
 board ?= none
@@ -59,7 +62,7 @@ ifeq ($(arch), $(filter $(arch), aarch64 mipsel))
 export SFSIMG = $(user_dir)/build/$(arch).img
 else
 # board is pc or qemu?
-ifeq ($(board), pc)
+ifeq ($(board), $(filter $(board), pc u540 k210 rocket_chip))
 #link user img, so use original image
 export SFSIMG = $(user_dir)/build/$(arch).img
 else
@@ -124,6 +127,7 @@ endif
 else ifeq ($(arch), riscv32)
 qemu_opts += \
 	-machine virt \
+	-serial mon:stdio \
 	-kernel ../tools/opensbi/virt_rv32.elf \
 	-device loader,addr=0x80400000,file=$(kernel_img) \
 	-drive file=$(SFSIMG),format=qcow2,id=sfs \
@@ -136,11 +140,13 @@ else ifeq ($(arch), riscv64)
 ifeq ($(board), u540)
 qemu_opts += \
 	-machine virt \
+	-serial mon:stdio \
 	-kernel ../tools/opensbi/fu540.elf \
 	-device loader,addr=0x80200000,file=$(kernel_img)
 else
 qemu_opts += \
 	-machine virt \
+	-serial mon:stdio \
 	-kernel ../tools/opensbi/virt_rv64.elf \
 	-device loader,addr=0x80200000,file=$(kernel_img) \
 	-drive file=$(SFSIMG),format=qcow2,id=sfs \
@@ -201,15 +207,12 @@ features += raspi3_use_generic_timer
 endif
 endif
 
-ifeq ($(board), u540)
-features += sv39
-riscv_pk_args += --enable-sv39
-endif
-
 ifneq ($(board), none)
 features += board_$(board)
 endif
 
+features += $(extra_features)
+
 build_args := --target targets/$(target).json --features "$(features)"
 
 ifeq ($(mode), release)
@@ -270,7 +273,7 @@ justrunui: build
 		-device virtio-mouse-device
 
 justruntest: build
-	@qemu-system-$(arch) $(qemu_opts) --append $(init) -serial file:../tests/stdout -monitor null
+	@qemu-system-$(arch) $(filter-out -serial mon:stdio, $(qemu_opts)) --append $(init) -serial file:../tests/stdout -monitor null
 
 debug: $(kernel) $(kernel_img)
 	@qemu-system-$(arch) $(qemu_opts) -s -S &
@@ -325,6 +328,11 @@ ifeq ($(arch), x86_64)
 	@bootimage build $(build_args)
 	@mv target/x86_64/bootimage.bin $(bootimage)
 else ifeq ($(arch), $(filter $(arch), riscv32 riscv64))
+ifeq ($(board), k210)
+	@cp src/arch/riscv32/board/k210/linker.ld src/arch/riscv32/boot/linker64.ld
+else
+	@cp src/arch/riscv32/board/u540/linker.ld src/arch/riscv32/boot/linker64.ld
+endif
 	@-patch -p0 -N -b \
 		$(shell rustc --print sysroot)/lib/rustlib/src/rust/src/libcore/sync/atomic.rs \
 		src/arch/riscv32/atomic.patch
@@ -367,10 +375,18 @@ ifeq ($(board), u540)
 .PHONY:
 install: $(kernel_img)
 	@$(objcopy) -S -O binary ../tools/opensbi/fu540.elf $(build_path)/bin
-	@dd if=$< of=$(build_path)/bin bs=131072 seek=16
-	@../tools/u540/mkimg.sh $(build_path)/bin $(build_path)/sd.img
+	@dd if=$< of=$(build_path)/bin bs=0x20000 seek=16
+	@../tools/u540/mkimg.sh $(build_path)/bin $(build_path)/u540.img
+endif
+
+ifeq ($(board), k210)
+.PHONY:
+install: $(kernel_img)
+	@$(objcopy) -S -O binary ../tools/opensbi/k210.elf $(build_path)/k210.img
+	@dd if=$< of=$(build_path)/k210.img bs=0x10000 seek=1
+	@python3 ../tools/k210/kflash.py -b 600000 $(build_path)/k210.img
 endif
 
 .PHONY:
 addr2line:
-	@python3 ../tools/addr2line.py $(prefix)addr2line $(arch) $(mode)
+	@python3.7 ../tools/addr2line.py $(prefix)addr2line $(arch) $(mode)

kernel/src/arch/aarch64/board/raspi3/mod.rs

@@ -2,7 +2,6 @@
 
 use alloc::string::String;
 use bcm2837::atags::Atags;
-use once::*;
 
 #[path = "../../../../drivers/gpu/fb.rs"]
 pub mod fb;
@@ -18,10 +17,7 @@ pub const IO_REMAP_END: usize = bcm2837::consts::KERNEL_OFFSET + 0x4000_1000;
 
 /// Initialize serial port before other initializations.
 pub fn init_serial_early() {
-    assert_has_not_been_called!("board::init must be called only once");
-
     serial::init();
-
     println!("Hello Raspberry Pi!");
 }
 

kernel/src/arch/aarch64/board/raspi3/serial.rs

@@ -1,7 +1,6 @@
 use bcm2837::mini_uart::{MiniUart, MiniUartInterruptId};
 use core::fmt;
 use lazy_static::lazy_static;
-use once::*;
 use spin::Mutex;
 
 /// Struct to get a global SerialPort interface
@@ -23,8 +22,6 @@ impl SerialPort {
 
     /// Init a newly created SerialPort, can only be called once.
     fn init(&mut self) {
-        assert_has_not_been_called!("SerialPort::init must be called only once");
-
         self.mu.init();
         super::irq::register_irq(super::irq::Interrupt::Aux, handle_serial_irq);
     }

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

@@ -1,7 +1,6 @@
 //! ARM64 drivers
 
 use super::board;
-use once::*;
 
 pub use self::board::fb;
 pub use self::board::serial;
@@ -10,8 +9,6 @@ pub mod console;
 
 /// Initialize ARM64 common drivers
 pub fn init() {
-    assert_has_not_been_called!("driver::init must be called only once");
-
     board::init_driver();
     console::init();
 }

kernel/src/arch/aarch64/interrupt/context.rs

@@ -32,7 +32,7 @@ impl TrapFrame {
         tf.spsr = 0b1101_00_0101; // To EL 1, enable IRQ
         tf
     }
-    fn new_user_thread(entry_addr: usize, sp: usize) -> Self {
+    pub fn new_user_thread(entry_addr: usize, sp: usize) -> Self {
         use core::mem::zeroed;
         let mut tf: Self = unsafe { zeroed() };
         tf.sp = sp;
@@ -40,9 +40,6 @@ impl TrapFrame {
         tf.spsr = 0b1101_00_0000; // To EL 0, enable IRQ
         tf
     }
-    pub fn is_user(&self) -> bool {
-        unimplemented!()
-    }
 }
 
 /// 新线程的内核栈初始内容
@@ -201,11 +198,6 @@ impl Context {
         }
         .push_at(kstack_top, ttbr)
     }
-    /// Called at a new user context
-    /// To get the init TrapFrame in sys_exec
-    pub unsafe fn get_init_tf(&self) -> TrapFrame {
-        (*(self.stack_top as *const InitStack)).tf.clone()
-    }
 }
 
 const ASID_MASK: u16 = 0xffff;

kernel/src/arch/mipsel/board/malta/mod.rs

@@ -1,7 +1,6 @@
 use crate::drivers::bus::pci;
 use alloc::string::String;
 use mips::registers::cp0;
-use once::*;
 
 #[path = "../../../../drivers/console/mod.rs"]
 pub mod console;
@@ -17,7 +16,6 @@ use fb::FramebufferInfo;
 
 /// Initialize serial port first
 pub fn init_serial_early() {
-    assert_has_not_been_called!("board::init must be called only once");
     // initialize serial driver
     serial::init(0xbf000900);
     // Enable serial interrupt

kernel/src/arch/mipsel/board/mipssim/mod.rs

@@ -1,5 +1,4 @@
 use alloc::string::String;
-use once::*;
 
 #[path = "../../../../drivers/console/mod.rs"]
 pub mod console;
@@ -11,7 +10,6 @@ pub mod serial;
 
 /// Initialize serial port first
 pub fn init_serial_early() {
-    assert_has_not_been_called!("board::init must be called only once");
     serial::init(0xbfd003f8);
     println!("Hello QEMU MIPSSIM!");
 }

kernel/src/arch/mipsel/board/thinpad/mod.rs

@@ -1,5 +1,4 @@
 use alloc::string::String;
-use once::*;
 
 #[path = "../../../../drivers/console/mod.rs"]
 pub mod console;
@@ -14,7 +13,6 @@ use fb::FramebufferResult;
 
 /// Initialize serial port first
 pub fn init_serial_early() {
-    assert_has_not_been_called!("board::init must be called only once");
     serial::init(0xa3000000);
     println!("Hello ThinPad!");
 }

kernel/src/arch/mipsel/context.rs

@@ -76,7 +76,7 @@ impl TrapFrame {
     ///
     /// The new thread starts at `entry_addr`.
     /// The stack pointer will be set to `sp`.
-    fn new_user_thread(entry_addr: usize, sp: usize) -> Self {
+    pub fn new_user_thread(entry_addr: usize, sp: usize) -> Self {
         use core::mem::zeroed;
         let mut tf: Self = unsafe { zeroed() };
         tf.sp = sp;
@@ -269,9 +269,4 @@ impl Context {
         }
         .push_at(kstack_top)
     }
-
-    /// Used for getting the init TrapFrame of a new user context in `sys_exec`.
-    pub unsafe fn get_init_tf(&self) -> TrapFrame {
-        (*(self.sp as *const InitStack)).tf.clone()
-    }
 }

kernel/src/arch/mipsel/driver/mod.rs

@@ -1,7 +1,6 @@
 //! mipsel drivers
 
 use super::board;
-use once::*;
 
 pub use self::board::fb;
 pub use self::board::serial;
@@ -10,7 +9,6 @@ pub mod console;
 
 /// Initialize common drivers
 pub fn init() {
-    assert_has_not_been_called!("driver::init must be called only once");
     board::init_driver();
     console::init();
 }

kernel/src/arch/riscv32/board/k210/linker.ld

@@ -0,0 +1,49 @@
+/* Copy from bbl-ucore : https://ring00.github.io/bbl-ucore      */
+
+/* Simple linker script for the ucore kernel.
+   See the GNU ld 'info' manual ("info ld") to learn the syntax. */
+
+OUTPUT_ARCH(riscv)
+ENTRY(_start)
+
+BASE_ADDRESS = 0xffffffffc0010000;
+
+SECTIONS
+{
+    /* Load the kernel at this address: "." means the current address */
+    . = BASE_ADDRESS;
+    start = .;
+
+    .text : {
+        stext = .;
+        *(.text.entry)
+        *(.text .text.*)
+        . = ALIGN(4K);
+        etext = .;
+    }
+
+    .rodata : {
+        srodata = .;
+        *(.rodata .rodata.*)
+        . = ALIGN(4K);
+        erodata = .;
+    }
+
+    .data : {
+        sdata = .;
+        *(.data .data.*)
+        edata = .;
+    }
+
+    .stack : {
+        *(.bss.stack)
+    }
+
+    .bss : {
+        sbss = .;
+        *(.bss .bss.*)
+        ebss = .;
+    }
+
+    PROVIDE(end = .);
+}

kernel/src/arch/riscv32/board/u540/linker.ld

@@ -0,0 +1,49 @@
+/* Copy from bbl-ucore : https://ring00.github.io/bbl-ucore      */
+
+/* Simple linker script for the ucore kernel.
+   See the GNU ld 'info' manual ("info ld") to learn the syntax. */
+
+OUTPUT_ARCH(riscv)
+ENTRY(_start)
+
+BASE_ADDRESS = 0xffffffffc0200000;
+
+SECTIONS
+{
+    /* Load the kernel at this address: "." means the current address */
+    . = BASE_ADDRESS;
+    start = .;
+
+    .text : {
+        stext = .;
+        *(.text.entry)
+        *(.text .text.*)
+        . = ALIGN(4K);
+        etext = .;
+    }
+
+    .rodata : {
+        srodata = .;
+        *(.rodata .rodata.*)
+        . = ALIGN(4K);
+        erodata = .;
+    }
+
+    .data : {
+        sdata = .;
+        *(.data .data.*)
+        edata = .;
+    }
+
+    .stack : {
+        *(.bss.stack)
+    }
+
+    .bss : {
+        sbss = .;
+        *(.bss .bss.*)
+        ebss = .;
+    }
+
+    PROVIDE(end = .);
+}

kernel/src/arch/riscv32/boot/entry32.asm

@@ -8,7 +8,7 @@ _start:
     # 1. set sp
     # sp = bootstack + (hartid + 1) * 0x10000
     add     t0, a0, 1
-    slli    t0, t0, 16
+    slli    t0, t0, 14
     lui     sp, %hi(bootstack)
     add     sp, sp, t0
 
@@ -32,7 +32,7 @@ _start:
     .align 12   # page align
     .global bootstack
 bootstack:
-    .space 4096 * 16 * 8
+    .space 4096 * 4 * 8
     .global bootstacktop
 bootstacktop:
 

kernel/src/arch/riscv32/boot/entry64.asm

@@ -8,7 +8,7 @@ _start:
     # 1. set sp
     # sp = bootstack + (hartid + 1) * 0x10000
     add     t0, a0, 1
-    slli    t0, t0, 16
+    slli    t0, t0, 14
     lui     sp, %hi(bootstack)
     add     sp, sp, t0
 
@@ -32,7 +32,7 @@ _start:
     .align 12   # page align
     .global bootstack
 bootstack:
-    .space 4096 * 16 * 8
+    .space 4096 * 4 * 8
     .global bootstacktop
 bootstacktop:
 

kernel/src/arch/riscv32/boot/entry_k210.asm

@@ -0,0 +1,30 @@
+    .section .text.entry
+    .globl _start
+_start:
+    # a0 == hartid
+    # pc == 0x80010000
+    # sp == 0x8000xxxx
+
+    # 1. set sp
+    # sp = bootstack + (hartid + 1) * 0x10000
+    add     t0, a0, 1
+    slli    t0, t0, 14
+    lui     sp, %hi(bootstack)
+    add     sp, sp, t0
+
+    # 1.1 set device tree paddr
+    # OpenSBI give me 0 ???
+    li      a1, 0x800003b0
+
+    # 2. jump to rust_main (absolute address)
+    lui     t0, %hi(rust_main)
+    addi    t0, t0, %lo(rust_main)
+    jr      t0
+
+    .section .bss.stack
+    .align 12   # page align
+    .global bootstack
+bootstack:
+    .space 4096 * 4 * 2
+    .global bootstacktop
+bootstacktop:

kernel/src/arch/riscv32/consts.rs

@@ -22,10 +22,16 @@ pub const KERNEL_P2_INDEX: usize = (KERNEL_OFFSET >> 12 >> 10) & 0x3ff;
 #[cfg(target_arch = "riscv64")]
 pub const KERNEL_P4_INDEX: usize = (KERNEL_OFFSET >> 12 >> 9 >> 9 >> 9) & 0o777;
 
+#[cfg(feature = "board_k210")]
+pub const KERNEL_HEAP_SIZE: usize = 0x0020_0000;
+#[cfg(not(feature = "board_k210"))]
 pub const KERNEL_HEAP_SIZE: usize = 0x0080_0000;
 
 pub const MEMORY_OFFSET: usize = 0x8000_0000;
 // TODO: get memory end from device tree
+#[cfg(feature = "board_k210")]
+pub const MEMORY_END: usize = 0x8060_0000;
+#[cfg(not(feature = "board_k210"))]
 pub const MEMORY_END: usize = 0x8800_0000;
 
 // FIXME: rv64 `sh` and `ls` will crash if stack top > 0x80000000 ???

kernel/src/arch/riscv32/context.rs

@@ -40,7 +40,7 @@ impl TrapFrame {
     ///
     /// The new thread starts at `entry_addr`.
     /// The stack pointer will be set to `sp`.
-    fn new_user_thread(entry_addr: usize, sp: usize) -> Self {
+    pub fn new_user_thread(entry_addr: usize, sp: usize) -> Self {
         use core::mem::zeroed;
         let mut tf: Self = unsafe { zeroed() };
         tf.x[2] = sp;
@@ -289,9 +289,4 @@ impl Context {
         }
         .push_at(kstack_top)
     }
-
-    /// Used for getting the init TrapFrame of a new user context in `sys_exec`.
-    pub unsafe fn get_init_tf(&self) -> TrapFrame {
-        (*(self.sp as *const InitStack)).tf.clone()
-    }
 }

kernel/src/arch/riscv32/memory.rs

@@ -9,6 +9,8 @@ use riscv::{addr::*, register::sstatus};
 /// Initialize the memory management module
 pub fn init(dtb: usize) {
     // allow user memory access
+    // NOTE: In K210 priv v1.9.1, sstatus.SUM is PUM which has opposite meaning!
+    #[cfg(not(feature = "board_k210"))]
     unsafe {
         sstatus::set_sum();
     }
@@ -90,6 +92,8 @@ fn remap_the_kernel(dtb: usize) {
         Linear::new(offset),
         "bss",
     );
+    // TODO: dtb on rocket chip
+    #[cfg(not(feature = "board_rocket_chip"))]
     ms.push(
         dtb,
         dtb + super::consts::MAX_DTB_SIZE,

kernel/src/arch/riscv32/mod.rs

@@ -53,9 +53,11 @@ pub extern "C" fn rust_main(hartid: usize, device_tree_paddr: usize) -> ! {
     memory::init(device_tree_vaddr);
     timer::init();
     // FIXME: init driver on u540
-    #[cfg(not(feature = "board_u540"))]
+    #[cfg(not(any(feature = "board_u540", feature = "board_rocket_chip")))]
     crate::drivers::init(device_tree_vaddr);
+    #[cfg(not(feature = "board_k210"))]
     unsafe {
+        #[cfg(not(feature = "board_rocket_chip"))]
         board::enable_serial_interrupt();
         board::init_external_interrupt();
     }
@@ -108,6 +110,8 @@ global_asm!(
 );
 #[cfg(target_arch = "riscv32")]
 global_asm!(include_str!("boot/entry32.asm"));
-#[cfg(target_arch = "riscv64")]
+#[cfg(all(target_arch = "riscv64", not(feature = "board_k210")))]
 global_asm!(include_str!("boot/entry64.asm"));
+#[cfg(feature = "board_k210")]
+global_asm!(include_str!("boot/entry_k210.asm"));
 global_asm!(include_str!("boot/trap.asm"));

kernel/src/arch/x86_64/consts.rs

@@ -1,96 +1,6 @@
-// Copy from Redox consts.rs:
-
-// Because the memory map is so important to not be aliased, it is defined here, in one place
-// The lower 256 PML4 entries are reserved for userspace
-// Each PML4 entry references up to 512 GB of memory
-// The top (511) PML4 is reserved for recursive mapping
-// The second from the top (510) PML4 is reserved for the kernel
-/// The size of a single PML4
-pub const PML4_SIZE: usize = 0x0000_0080_0000_0000;
-pub const PML4_MASK: usize = 0x0000_ff80_0000_0000;
-
-/// Offset of recursive paging
-pub const RECURSIVE_PAGE_OFFSET: usize = (-(PML4_SIZE as isize)) as usize;
-pub const RECURSIVE_PAGE_PML4: usize = (RECURSIVE_PAGE_OFFSET & PML4_MASK) / PML4_SIZE;
-
-/// Offset of kernel
-pub const KERNEL_OFFSET: usize = RECURSIVE_PAGE_OFFSET - PML4_SIZE;
-pub const KERNEL_PML4: usize = (KERNEL_OFFSET & PML4_MASK) / PML4_SIZE;
-
-pub const KERNEL_SIZE: usize = PML4_SIZE;
-
-/// Offset to kernel heap
-pub const KERNEL_HEAP_OFFSET: usize = KERNEL_OFFSET - PML4_SIZE;
-pub const KERNEL_HEAP_PML4: usize = (KERNEL_HEAP_OFFSET & PML4_MASK) / PML4_SIZE;
-/// Size of kernel heap
-pub const KERNEL_HEAP_SIZE: usize = 32 * 1024 * 1024; // 32 MB
-
 pub const MEMORY_OFFSET: usize = 0;
+pub const KERNEL_OFFSET: usize = 0xffffff00_00000000;
+pub const KERNEL_HEAP_SIZE: usize = 8 * 1024 * 1024; // 8 MB
 
-/// Offset to kernel percpu variables
-//TODO: Use 64-bit fs offset to enable this pub const KERNEL_PERCPU_OFFSET: usize = KERNEL_HEAP_OFFSET - PML4_SIZE;
-pub const KERNEL_PERCPU_OFFSET: usize = 0xC000_0000;
-/// Size of kernel percpu variables
-pub const KERNEL_PERCPU_SIZE: usize = 64 * 1024; // 64 KB
-
-/// Offset to user image
-pub const USER_OFFSET: usize = 0;
-pub const USER_PML4: usize = (USER_OFFSET & PML4_MASK) / PML4_SIZE;
-
-/// Offset to user TCB
-pub const USER_TCB_OFFSET: usize = 0xB000_0000;
-
-/// Offset to user arguments
-pub const USER_ARG_OFFSET: usize = USER_OFFSET + PML4_SIZE / 2;
-
-/// Offset to user heap
-pub const USER_HEAP_OFFSET: usize = USER_OFFSET + PML4_SIZE;
-pub const USER_HEAP_PML4: usize = (USER_HEAP_OFFSET & PML4_MASK) / PML4_SIZE;
-
-/// Offset to user grants
-pub const USER_GRANT_OFFSET: usize = USER_HEAP_OFFSET + PML4_SIZE;
-pub const USER_GRANT_PML4: usize = (USER_GRANT_OFFSET & PML4_MASK) / PML4_SIZE;
-
-/// Offset to user stack
-pub const USER_STACK_OFFSET: usize = USER_GRANT_OFFSET + PML4_SIZE;
-pub const USER_STACK_PML4: usize = (USER_STACK_OFFSET & PML4_MASK) / PML4_SIZE;
-/// Size of user stack
+pub const USER_STACK_OFFSET: usize = 0x00008000_00000000 - USER_STACK_SIZE;
 pub const USER_STACK_SIZE: usize = 8 * 1024 * 1024; // 8 MB, the default config of Linux
-
-/// Offset to user sigstack
-pub const USER_SIGSTACK_OFFSET: usize = USER_STACK_OFFSET + PML4_SIZE;
-pub const USER_SIGSTACK_PML4: usize = (USER_SIGSTACK_OFFSET & PML4_MASK) / PML4_SIZE;
-/// Size of user sigstack
-pub const USER_SIGSTACK_SIZE: usize = 256 * 1024; // 256 KB
-
-/// Offset to user TLS
-pub const USER_TLS_OFFSET: usize = USER_SIGSTACK_OFFSET + PML4_SIZE;
-pub const USER_TLS_PML4: usize = (USER_TLS_OFFSET & PML4_MASK) / PML4_SIZE;
-
-/// Offset to user temporary image (used when cloning)
-pub const USER_TMP_OFFSET: usize = USER_TLS_OFFSET + PML4_SIZE;
-pub const USER_TMP_PML4: usize = (USER_TMP_OFFSET & PML4_MASK) / PML4_SIZE;
-
-/// Offset to user temporary heap (used when cloning)
-pub const USER_TMP_HEAP_OFFSET: usize = USER_TMP_OFFSET + PML4_SIZE;
-pub const USER_TMP_HEAP_PML4: usize = (USER_TMP_HEAP_OFFSET & PML4_MASK) / PML4_SIZE;
-
-/// Offset to user temporary page for grants
-pub const USER_TMP_GRANT_OFFSET: usize = USER_TMP_HEAP_OFFSET + PML4_SIZE;
-pub const USER_TMP_GRANT_PML4: usize = (USER_TMP_GRANT_OFFSET & PML4_MASK) / PML4_SIZE;
-
-/// Offset to user temporary stack (used when cloning)
-pub const USER_TMP_STACK_OFFSET: usize = USER_TMP_GRANT_OFFSET + PML4_SIZE;
-pub const USER_TMP_STACK_PML4: usize = (USER_TMP_STACK_OFFSET & PML4_MASK) / PML4_SIZE;
-
-/// Offset to user temporary sigstack (used when cloning)
-pub const USER_TMP_SIGSTACK_OFFSET: usize = USER_TMP_STACK_OFFSET + PML4_SIZE;
-pub const USER_TMP_SIGSTACK_PML4: usize = (USER_TMP_SIGSTACK_OFFSET & PML4_MASK) / PML4_SIZE;
-
-/// Offset to user temporary tls (used when cloning)
-pub const USER_TMP_TLS_OFFSET: usize = USER_TMP_SIGSTACK_OFFSET + PML4_SIZE;
-pub const USER_TMP_TLS_PML4: usize = (USER_TMP_TLS_OFFSET & PML4_MASK) / PML4_SIZE;
-
-/// Offset for usage in other temporary pages
-pub const USER_TMP_MISC_OFFSET: usize = USER_TMP_TLS_OFFSET + PML4_SIZE;
-pub const USER_TMP_MISC_PML4: usize = (USER_TMP_MISC_OFFSET & PML4_MASK) / PML4_SIZE;

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

@@ -1,5 +1,3 @@
-use once::*;
-
 pub mod ide;
 pub mod keyboard;
 pub mod pic;
@@ -9,8 +7,6 @@ pub mod serial;
 pub mod vga;
 
 pub fn init() {
-    assert_has_not_been_called!();
-
     // Use IOAPIC instead of PIC
     pic::disable();
 

kernel/src/arch/x86_64/driver/pic.rs

@@ -1,7 +1,6 @@
 // Copy from Redox
 
 use log::*;
-use once::*;
 use spin::Mutex;
 use x86_64::instructions::port::Port;
 
@@ -18,8 +17,6 @@ pub fn disable() {
 }
 
 pub unsafe fn init() {
-    assert_has_not_been_called!("pic::init must be called only once");
-
     let mut master = MASTER.lock();
     let mut slave = SLAVE.lock();
 

kernel/src/arch/x86_64/driver/pit.rs

@@ -1,9 +1,7 @@
 use log::*;
-use once::*;
 use x86_64::instructions::port::Port;
 
 pub fn init() {
-    assert_has_not_been_called!("pit::init must be called only once");
     Pit::new(0x40).init(100);
     info!("pit: init end");
 }

kernel/src/arch/x86_64/driver/serial.rs

@@ -1,4 +1,3 @@
-use once::*;
 use spin::Mutex;
 use uart_16550::SerialPort;
 use x86_64::instructions::port::Port;
@@ -9,8 +8,6 @@ pub static COM1: Mutex<SerialPort> = Mutex::new(unsafe { SerialPort::new(0x3F8)
 pub static COM2: Mutex<SerialPort> = Mutex::new(unsafe { SerialPort::new(0x2F8) });
 
 pub fn init() {
-    assert_has_not_been_called!("serial::init must be called only once");
-
     COM1.lock().init();
     COM2.lock().init();
     enable_irq(consts::COM1);

kernel/src/arch/x86_64/gdt.rs

@@ -39,10 +39,6 @@ pub struct Cpu {
 }
 
 impl Cpu {
-    pub fn current() -> &'static mut Self {
-        unsafe { CPUS[super::cpu::id()].as_mut().unwrap() }
-    }
-
     fn new() -> Self {
         Cpu {
             gdt: GlobalDescriptorTable::new(),
@@ -72,18 +68,9 @@ impl Cpu {
         set_cs(KCODE_SELECTOR);
         // load TSS
         load_tss(TSS_SELECTOR);
-        // for fast syscall:
-        // store address of TSS to kernel_gsbase
-        let mut kernel_gsbase = Msr::new(0xC0000102);
-        kernel_gsbase.write(&self.tss as *const _ as u64);
-    }
-
-    /// 设置从Ring3跳到Ring0时，自动切换栈的地址
-    ///
-    /// 每次进入用户态前，都要调用此函数，才能保证正确返回内核态
-    pub fn set_ring0_rsp(&mut self, rsp: usize) {
-        trace!("gdt.set_ring0_rsp: {:#x}", rsp);
-        self.tss.privilege_stack_table[0] = VirtAddr::new(rsp as u64);
+        // store address of TSS to GSBase
+        let mut gsbase = Msr::new(0xC0000101);
+        gsbase.write(&self.tss as *const _ as u64);
     }
 }
 

kernel/src/arch/x86_64/interrupt/fast_syscall.rs

@@ -10,9 +10,9 @@ pub fn init() {
             *flags |= EferFlags::SYSTEM_CALL_EXTENSIONS;
         });
 
-        let mut star = Msr::new(0xC0000081);
-        let mut lstar = Msr::new(0xC0000082);
-        let mut sfmask = Msr::new(0xC0000084);
+        let mut star = Msr::new(0xC0000081); // legacy mode SYSCALL target
+        let mut lstar = Msr::new(0xC0000082); // long mode SYSCALL target
+        let mut sfmask = Msr::new(0xC0000084); // EFLAGS mask for syscall
 
         // flags to clear on syscall
         // copy from Linux 5.0

kernel/src/arch/x86_64/interrupt/handler.rs

@@ -213,9 +213,3 @@ fn invalid_opcode(tf: &mut TrapFrame) {
 fn error(tf: &TrapFrame) {
     crate::trap::error(tf);
 }
-
-#[no_mangle]
-pub unsafe extern "C" fn set_return_rsp(tf: *const TrapFrame) {
-    use crate::arch::gdt::Cpu;
-    Cpu::current().set_ring0_rsp(tf.add(1) as usize);
-}

kernel/src/arch/x86_64/interrupt/trap.asm

@@ -49,8 +49,10 @@ __alltraps:
 .global trap_ret
 trap_ret:
 
+    # store kernel rsp -> TSS.sp0
     mov rdi, rsp
-    call set_return_rsp
+    add rdi, 720
+    mov gs:[4], rdi
 
     # pop fp state offset
     pop rcx
@@ -104,8 +106,6 @@ syscall_entry:
     # - store rip -> rcx
     # - load rip
 
-    # swap in kernel gs
-    swapgs
     # store user rsp -> scratch at TSS.sp1
     mov gs:[12], rsp
     # load kernel rsp <- TSS.sp0
@@ -119,11 +119,8 @@ syscall_entry:
     push 0          # error_code (dummy)
     push 0          # trap_num (dummy)
 
-    # swap out kernel gs
-    swapgs
-
     # enable interrupt
-    # sti
+    sti
 
     push rax
     push rcx
@@ -173,8 +170,10 @@ syscall_return:
     # disable interrupt
     cli
 
+    # store kernel rsp -> TSS.sp0
     mov rdi, rsp
-    call set_return_rsp
+    add rdi, 720
+    mov gs:[4], rdi
 
     # pop fp state offset
     pop rcx

kernel/src/arch/x86_64/interrupt/trapframe.rs

@@ -73,7 +73,7 @@ impl TrapFrame {
         tf.fpstate_offset = 16; // skip restoring for first time
         tf
     }
-    fn new_user_thread(entry_addr: usize, rsp: usize) -> Self {
+    pub fn new_user_thread(entry_addr: usize, rsp: usize) -> Self {
         use crate::arch::gdt;
         let mut tf = TrapFrame::default();
         tf.cs = gdt::UCODE_SELECTOR.0 as usize;
@@ -234,9 +234,4 @@ impl Context {
         }
         .push_at(kstack_top)
     }
-    /// Called at a new user context
-    /// To get the init TrapFrame in sys_exec
-    pub unsafe fn get_init_tf(&self) -> TrapFrame {
-        (*(self.0 as *const InitStack)).tf.clone()
-    }
 }

kernel/src/arch/x86_64/memory.rs

@@ -2,20 +2,15 @@ use crate::consts::KERNEL_OFFSET;
 use bitmap_allocator::BitAlloc;
 // Depends on kernel
 use super::{BootInfo, MemoryRegionType};
-use crate::memory::{active_table, alloc_frame, init_heap, FRAME_ALLOCATOR};
-use crate::HEAP_ALLOCATOR;
-use alloc::vec::Vec;
+use crate::memory::{active_table, init_heap, FRAME_ALLOCATOR};
 use log::*;
-use once::*;
 use rcore_memory::paging::*;
 use rcore_memory::PAGE_SIZE;
 
 pub fn init(boot_info: &BootInfo) {
-    assert_has_not_been_called!("memory::init must be called only once");
     init_frame_allocator(boot_info);
     init_device_vm_map();
     init_heap();
-    enlarge_heap();
     info!("memory: init end");
 }
 
@@ -42,30 +37,3 @@ fn init_device_vm_map() {
         .map(KERNEL_OFFSET + 0xfee00000, 0xfee00000)
         .update();
 }
-
-fn enlarge_heap() {
-    let mut page_table = active_table();
-    let mut addrs = Vec::new();
-    let va_offset = KERNEL_OFFSET + 0xe0000000;
-    for i in 0..16384 {
-        let page = alloc_frame().unwrap();
-        let va = KERNEL_OFFSET + 0xe0000000 + page;
-        if let Some((ref mut addr, ref mut len)) = addrs.last_mut() {
-            if *addr - PAGE_SIZE == va {
-                *len += PAGE_SIZE;
-                *addr -= PAGE_SIZE;
-                continue;
-            }
-        }
-        addrs.push((va, PAGE_SIZE));
-    }
-    for (addr, len) in addrs.into_iter() {
-        for va in (addr..(addr + len)).step_by(PAGE_SIZE) {
-            page_table.map(va, va - va_offset).update();
-        }
-        info!("Adding {:#X} {:#X} to heap", addr, len);
-        unsafe {
-            HEAP_ALLOCATOR.lock().init(addr, len);
-        }
-    }
-}

kernel/src/arch/x86_64/mod.rs

@@ -15,7 +15,7 @@ pub mod rand;
 pub mod syscall;
 pub mod timer;
 
-static AP_CAN_INIT: AtomicBool = ATOMIC_BOOL_INIT;
+static AP_CAN_INIT: AtomicBool = AtomicBool::new(false);
 
 /// The entry point of kernel
 #[no_mangle] // don't mangle the name of this function
@@ -40,26 +40,33 @@ pub extern "C" fn _start(boot_info: &'static BootInfo) -> ! {
     memory::init(boot_info);
 
     // Now heap is available
+
+    // Init GDT
     gdt::init();
-
+    //get local apic id of cpu
     cpu::init();
-
+    // Use IOAPIC instead of PIC, use APIC Timer instead of PIT, init serial&keyboard in x86_64
     driver::init();
-
+    // init pci/bus-based devices ,e.g. Intel 10Gb NIC, ...
     crate::drivers::init();
-
+    // init cpu scheduler and process manager, and add user shell app in process manager
     crate::process::init();
-
+    //wake up other CPUs
     AP_CAN_INIT.store(true, Ordering::Relaxed);
-
+    //call the first main function in kernel.
     crate::kmain();
 }
 
 /// The entry point for other processors
 fn other_start() -> ! {
+    // Init trap handling.
     idt::init();
+    // init gdt
     gdt::init();
+    // init local apic
     cpu::init();
+    // setup fast syscall in xv6-64
     interrupt::fast_syscall::init();
+    //call the first main function in kernel.
     crate::kmain();
 }

kernel/src/arch/x86_64/paging.rs

@@ -47,23 +47,21 @@ impl PageTable for ActivePageTable {
     fn map(&mut self, addr: usize, target: usize) -> &mut Entry {
         let flags = EF::PRESENT | EF::WRITABLE | EF::NO_EXECUTE;
         unsafe {
-            if let Ok(flush) = self.0.map_to(
-                Page::of_addr(addr),
-                Frame::of_addr(target),
-                flags,
-                &mut FrameAllocatorForX86,
-            ) {
-                flush.flush();
-            }
+            self.0
+                .map_to(
+                    Page::of_addr(addr),
+                    Frame::of_addr(target),
+                    flags,
+                    &mut FrameAllocatorForX86,
+                )
+                .unwrap()
+                .flush();
         }
         unsafe { &mut *(get_entry_ptr(addr, 1)) }
     }
 
     fn unmap(&mut self, addr: usize) {
-        // unmap and flush if it is mapped
-        if let Ok((_, flush)) = self.0.unmap(Page::of_addr(addr)) {
-            flush.flush();
-        }
+        self.0.unmap(Page::of_addr(addr)).unwrap().1.flush();
     }
 
     fn get_entry(&mut self, addr: usize) -> Option<&mut Entry> {
@@ -238,6 +236,9 @@ impl InactivePageTable for InactivePageTable0 {
 
     fn edit<T>(&mut self, f: impl FnOnce(&mut Self::Active) -> T) -> T {
         let target = Cr3::read().0.start_address().as_u64() as usize;
+        if self.p4_frame == Cr3::read().0 {
+            return f(&mut active_table());
+        }
         active_table().with_temporary_map(target, |active_table, p4_table: &mut x86PageTable| {
             let backup = p4_table[0o777].clone();
 

kernel/src/drivers/block/ahci.rs

@@ -48,8 +48,6 @@ pub fn init(_irq: Option<u32>, header: usize, size: usize) -> Arc<AHCIDriver> {
     let ahci = AHCI::new(header, size);
     let driver = Arc::new(AHCIDriver(Mutex::new(ahci)));
     DRIVERS.write().push(driver.clone());
-    BLK_DRIVERS
-        .write()
-        .push(Arc::new(BlockDriver(driver.clone())));
+    BLK_DRIVERS.write().push(driver.clone());
     driver
 }

kernel/src/drivers/block/virtio_blk.rs

@@ -12,8 +12,6 @@ use rcore_memory::paging::PageTable;
 use rcore_memory::PAGE_SIZE;
 use volatile::Volatile;
 
-use rcore_fs::dev::BlockDevice;
-
 use crate::drivers::BlockDriver;
 use crate::memory::active_table;
 use crate::sync::SpinNoIrqLock as Mutex;
@@ -224,5 +222,5 @@ pub fn virtio_blk_init(node: &Node) {
 
     let driver = Arc::new(driver);
     DRIVERS.write().push(driver.clone());
-    BLK_DRIVERS.write().push(Arc::new(BlockDriver(driver)));
+    BLK_DRIVERS.write().push(driver);
 }

kernel/src/drivers/bus/pci.rs

@@ -5,7 +5,6 @@ use crate::drivers::{Driver, DRIVERS, NET_DRIVERS};
 use crate::memory::active_table;
 use alloc::collections::BTreeMap;
 use alloc::sync::Arc;
-use core::cmp::Ordering;
 use pci::*;
 use rcore_memory::{paging::PageTable, PAGE_SIZE};
 use spin::Mutex;
@@ -201,7 +200,7 @@ pub fn detach_driver(loc: &Location) -> bool {
 }
 
 pub fn init() {
-    let mut pci_iter = unsafe { scan_bus(&PortOpsImpl, CSpaceAccessMethod::IO) };
+    let pci_iter = unsafe { scan_bus(&PortOpsImpl, CSpaceAccessMethod::IO) };
     for dev in pci_iter {
         info!(
             "pci: {:02x}:{:02x}.{} {:#x} {:#x} ({} {}) irq: {}:{:?}",
@@ -220,7 +219,7 @@ pub fn init() {
 }
 
 pub fn find_device(vendor: u16, product: u16) -> Option<Location> {
-    let mut pci_iter = unsafe { scan_bus(&PortOpsImpl, CSpaceAccessMethod::IO) };
+    let pci_iter = unsafe { scan_bus(&PortOpsImpl, CSpaceAccessMethod::IO) };
     for dev in pci_iter {
         if dev.id.vendor_id == vendor && dev.id.device_id == product {
             return Some(dev.loc);

kernel/src/drivers/gpu/fb.rs

@@ -4,7 +4,6 @@ use alloc::string::String;
 use core::fmt;
 use lazy_static::lazy_static;
 use log::*;
-use once::*;
 use spin::Mutex;
 
 /// Framebuffer information
@@ -134,8 +133,6 @@ impl fmt::Debug for Framebuffer {
 
 impl Framebuffer {
     fn new(width: u32, height: u32, depth: u32) -> Result<Framebuffer, String> {
-        assert_has_not_been_called!("Framebuffer::new must be called only once");
-
         let probed_info = super::probe_fb_info(width, height, depth);
 
         match probed_info {

kernel/src/drivers/mod.rs

@@ -7,7 +7,7 @@ use smoltcp::wire::{EthernetAddress, IpAddress, IpCidr, Ipv4Address};
 use spin::RwLock;
 
 use crate::sync::Condvar;
-use rcore_fs::dev::BlockDevice;
+use rcore_fs::dev::{self, BlockDevice, DevError};
 
 #[allow(dead_code)]
 pub mod block;
@@ -72,21 +72,21 @@ pub trait Driver: Send + Sync {
     }
 
     // send an ethernet frame, only use it when necessary
-    fn send(&self, data: &[u8]) -> Option<usize> {
+    fn send(&self, _data: &[u8]) -> Option<usize> {
         unimplemented!("not a net driver")
     }
 
     // get mac address from ip address in arp table
-    fn get_arp(&self, ip: IpAddress) -> Option<EthernetAddress> {
+    fn get_arp(&self, _ip: IpAddress) -> Option<EthernetAddress> {
         unimplemented!("not a net driver")
     }
 
     // block related drivers should implement these
-    fn read_block(&self, block_id: usize, buf: &mut [u8]) -> bool {
+    fn read_block(&self, _block_id: usize, _buf: &mut [u8]) -> bool {
         unimplemented!("not a block driver")
     }
 
-    fn write_block(&self, block_id: usize, buf: &[u8]) -> bool {
+    fn write_block(&self, _block_id: usize, _buf: &[u8]) -> bool {
         unimplemented!("not a block driver")
     }
 }
@@ -95,19 +95,29 @@ lazy_static! {
     // NOTE: RwLock only write when initializing drivers
     pub static ref DRIVERS: RwLock<Vec<Arc<Driver>>> = RwLock::new(Vec::new());
     pub static ref NET_DRIVERS: RwLock<Vec<Arc<Driver>>> = RwLock::new(Vec::new());
-    pub static ref BLK_DRIVERS: RwLock<Vec<Arc<BlockDriver>>> = RwLock::new(Vec::new());
+    pub static ref BLK_DRIVERS: RwLock<Vec<Arc<Driver>>> = RwLock::new(Vec::new());
 }
 
-pub struct BlockDriver(Arc<Driver>);
+pub struct BlockDriver(pub Arc<Driver>);
 
 impl BlockDevice for BlockDriver {
     const BLOCK_SIZE_LOG2: u8 = 9; // 512
-    fn read_at(&self, block_id: usize, buf: &mut [u8]) -> bool {
-        self.0.read_block(block_id, buf)
+    fn read_at(&self, block_id: usize, buf: &mut [u8]) -> dev::Result<()> {
+        match self.0.read_block(block_id, buf) {
+            true => Ok(()),
+            false => Err(DevError),
+        }
     }
 
-    fn write_at(&self, block_id: usize, buf: &[u8]) -> bool {
-        self.0.write_block(block_id, buf)
+    fn write_at(&self, block_id: usize, buf: &[u8]) -> dev::Result<()> {
+        match self.0.write_block(block_id, buf) {
+            true => Ok(()),
+            false => Err(DevError),
+        }
+    }
+
+    fn sync(&self) -> dev::Result<()> {
+        Ok(())
     }
 }
 

kernel/src/drivers/net/ixgbe.rs

@@ -15,7 +15,6 @@ use smoltcp::wire::EthernetAddress;
 use smoltcp::wire::*;
 use smoltcp::Result;
 
-use crate::memory::active_table;
 use crate::net::SOCKETS;
 use crate::sync::FlagsGuard;
 use crate::sync::SpinNoIrqLock as Mutex;
@@ -203,7 +202,7 @@ pub fn ixgbe_init(
     let ip_addrs = [IpCidr::new(IpAddress::v4(10, 0, index as u8, 2), 24)];
     let neighbor_cache = NeighborCache::new(BTreeMap::new());
     let routes = Routes::new(BTreeMap::new());
-    let mut iface = EthernetInterfaceBuilder::new(net_driver.clone())
+    let iface = EthernetInterfaceBuilder::new(net_driver.clone())
         .ethernet_addr(ethernet_addr)
         .ip_addrs(ip_addrs)
         .neighbor_cache(neighbor_cache)

kernel/src/drivers/net/virtio_net.rs

@@ -2,7 +2,6 @@ use alloc::alloc::{GlobalAlloc, Layout};
 use alloc::format;
 use alloc::string::String;
 use alloc::sync::Arc;
-use alloc::vec::Vec;
 use core::mem::size_of;
 use core::slice;
 

kernel/src/fs/device.rs

@@ -19,34 +19,42 @@ impl MemBuf {
 }
 
 impl Device for MemBuf {
-    fn read_at(&self, offset: usize, buf: &mut [u8]) -> Option<usize> {
+    fn read_at(&self, offset: usize, buf: &mut [u8]) -> Result<usize> {
         let slice = self.0.read();
         let len = buf.len().min(slice.len() - offset);
         buf[..len].copy_from_slice(&slice[offset..offset + len]);
-        Some(len)
+        Ok(len)
     }
-    fn write_at(&self, offset: usize, buf: &[u8]) -> Option<usize> {
+    fn write_at(&self, offset: usize, buf: &[u8]) -> Result<usize> {
         let mut slice = self.0.write();
         let len = buf.len().min(slice.len() - offset);
         slice[offset..offset + len].copy_from_slice(&buf[..len]);
-        Some(len)
+        Ok(len)
+    }
+    fn sync(&self) -> Result<()> {
+        Ok(())
     }
 }
 
 #[cfg(target_arch = "x86_64")]
 impl BlockDevice for ide::IDE {
     const BLOCK_SIZE_LOG2: u8 = 9;
-    fn read_at(&self, block_id: usize, buf: &mut [u8]) -> bool {
+    fn read_at(&self, block_id: usize, buf: &mut [u8]) -> Result<()> {
         use core::slice;
         assert!(buf.len() >= ide::BLOCK_SIZE);
         let buf =
             unsafe { slice::from_raw_parts_mut(buf.as_ptr() as *mut u32, ide::BLOCK_SIZE / 4) };
-        self.read(block_id as u64, 1, buf).is_ok()
+        self.read(block_id as u64, 1, buf).map_err(|_| DevError)?;
+        Ok(())
     }
-    fn write_at(&self, block_id: usize, buf: &[u8]) -> bool {
+    fn write_at(&self, block_id: usize, buf: &[u8]) -> Result<()> {
         use core::slice;
         assert!(buf.len() >= ide::BLOCK_SIZE);
         let buf = unsafe { slice::from_raw_parts(buf.as_ptr() as *mut u32, ide::BLOCK_SIZE / 4) };
-        self.write(block_id as u64, 1, buf).is_ok()
+        self.write(block_id as u64, 1, buf).map_err(|_| DevError)?;
+        Ok(())
+    }
+    fn sync(&self) -> Result<()> {
+        Ok(())
     }
 }

kernel/src/fs/file.rs

@@ -1,6 +1,7 @@
 //! File handle for process
 
 use alloc::{string::String, sync::Arc};
+use core::fmt;
 
 use rcore_fs::vfs::{FsError, INode, Metadata, PollStatus, Result};
 
@@ -9,6 +10,7 @@ pub struct FileHandle {
     inode: Arc<INode>,
     offset: u64,
     options: OpenOptions,
+    pub path: String,
 }
 
 #[derive(Debug, Clone)]
@@ -27,12 +29,13 @@ pub enum SeekFrom {
 }
 
 impl FileHandle {
-    pub fn new(inode: Arc<INode>, options: OpenOptions) -> Self {
-        FileHandle {
+    pub fn new(inode: Arc<INode>, options: OpenOptions, path: String) -> Self {
+        return FileHandle {
             inode,
             offset: 0,
             options,
-        }
+            path,
+        };
     }
 
     pub fn read(&mut self, buf: &mut [u8]) -> Result<usize> {
@@ -116,4 +119,19 @@ impl FileHandle {
     pub fn io_control(&self, cmd: u32, arg: usize) -> Result<()> {
         self.inode.io_control(cmd, arg)
     }
+
+    pub fn inode(&self) -> Arc<INode> {
+        self.inode.clone()
+    }
+}
+
+impl fmt::Debug for FileHandle {
+    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+        return f
+            .debug_struct("FileHandle")
+            .field("offset", &self.offset)
+            .field("options", &self.options)
+            .field("path", &self.path)
+            .finish();
+    }
 }

kernel/src/fs/file_like.rs

@@ -1,5 +1,6 @@
 use core::fmt;
 
+use super::ioctl::*;
 use super::FileHandle;
 use crate::net::Socket;
 use crate::syscall::{SysError, SysResult};
@@ -30,13 +31,19 @@ impl FileLike {
         Ok(len)
     }
     pub fn ioctl(&mut self, request: usize, arg1: usize, arg2: usize, arg3: usize) -> SysResult {
-        match self {
-            FileLike::File(file) => file.io_control(request as u32, arg1)?,
-            FileLike::Socket(socket) => {
-                socket.ioctl(request, arg1, arg2, arg3)?;
+        match request {
+            // TODO: place flags & path in FileLike in stead of FileHandle/Socket
+            FIOCLEX => Ok(0),
+            _ => {
+                match self {
+                    FileLike::File(file) => file.io_control(request as u32, arg1)?,
+                    FileLike::Socket(socket) => {
+                        socket.ioctl(request, arg1, arg2, arg3)?;
+                    }
+                }
+                Ok(0)
             }
         }
-        Ok(0)
     }
     pub fn poll(&self) -> Result<PollStatus, SysError> {
         let status = match self {
@@ -53,8 +60,8 @@ impl FileLike {
 impl fmt::Debug for FileLike {
     fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
         match self {
-            FileLike::File(_) => write!(f, "File"),
-            FileLike::Socket(_) => write!(f, "Socket"),
+            FileLike::File(file) => write!(f, "File({:?})", file),
+            FileLike::Socket(socket) => write!(f, "Socket({:?})", socket),
         }
     }
 }

kernel/src/fs/ioctl.rs

@@ -0,0 +1,36 @@
+// for IOR and IOW:
+// 32bits total, command in lower 16bits, size of the parameter structure in the lower 14 bits of the upper 16 bits
+// higher 2 bits: 01 = write, 10 = read
+
+#[cfg(not(target_arch = "mips"))]
+pub const TCGETS: usize = 0x5401;
+#[cfg(target_arch = "mips")]
+pub const TCGETS: usize = 0x540D;
+
+#[cfg(not(target_arch = "mips"))]
+pub const TIOCGPGRP: usize = 0x540F;
+// _IOR('t', 119, int)
+#[cfg(target_arch = "mips")]
+pub const TIOCGPGRP: usize = 0x4_004_74_77;
+
+#[cfg(not(target_arch = "mips"))]
+pub const TIOCSPGRP: usize = 0x5410;
+// _IOW('t', 118, int)
+#[cfg(target_arch = "mips")]
+pub const TIOCSPGRP: usize = 0x8_004_74_76;
+
+#[cfg(not(target_arch = "mips"))]
+pub const TIOCGWINSZ: usize = 0x5413;
+// _IOR('t', 104, struct winsize)
+#[cfg(target_arch = "mips")]
+pub const TIOCGWINSZ: usize = 0x4_008_74_68;
+
+#[cfg(not(target_arch = "mips"))]
+pub const FIONCLEX: usize = 0x5450;
+#[cfg(target_arch = "mips")]
+pub const FIONCLEX: usize = 0x6602;
+
+#[cfg(not(target_arch = "mips"))]
+pub const FIOCLEX: usize = 0x5451;
+#[cfg(target_arch = "mips")]
+pub const FIOCLEX: usize = 0x6601;

kernel/src/fs/mod.rs

@@ -1,20 +1,23 @@
 use alloc::{sync::Arc, vec::Vec};
 
+use rcore_fs::dev::block_cache::BlockCache;
 use rcore_fs::vfs::*;
 use rcore_fs_sfs::SimpleFileSystem;
 
-#[cfg(target_arch = "x86_64")]
-use crate::arch::driver::ide;
+use crate::drivers::BlockDriver;
 
 pub use self::file::*;
 pub use self::file_like::*;
 pub use self::pipe::Pipe;
+pub use self::pseudo::*;
 pub use self::stdio::{STDIN, STDOUT};
 
 mod device;
 mod file;
 mod file_like;
+mod ioctl;
 mod pipe;
+mod pseudo;
 mod stdio;
 
 /// Hard link user programs
@@ -39,9 +42,15 @@ lazy_static! {
         let device = {
             #[cfg(any(target_arch = "riscv32", target_arch = "riscv64", target_arch = "x86_64"))]
             {
-                crate::drivers::BLK_DRIVERS.read().iter()
-                    .next().expect("Block device not found")
-                    .clone()
+                let driver = BlockDriver(
+                    crate::drivers::BLK_DRIVERS
+                        .read().iter()
+                        .next().expect("Block device not found")
+                        .clone()
+                );
+                // enable block cache
+                Arc::new(BlockCache::new(driver, 0x100))
+                // Arc::new(driver)
             }
             #[cfg(target_arch = "aarch64")]
             {

kernel/src/fs/pipe.rs

@@ -57,7 +57,6 @@ impl Pipe {
 // TODO: better way to provide default impl?
 macro_rules! impl_inode {
     () => {
-        fn poll(&self) -> Result<PollStatus> { Err(FsError::NotSupported) }
         fn metadata(&self) -> Result<Metadata> { Err(FsError::NotSupported) }
         fn set_metadata(&self, _metadata: &Metadata) -> Result<()> { Ok(()) }
         fn sync_all(&self) -> Result<()> { Ok(()) }
@@ -104,5 +103,41 @@ impl INode for Pipe {
             Ok(0)
         }
     }
+
+    fn poll(&self) -> Result<PollStatus> {
+        let data = self.data.lock();
+        match self.direction {
+            PipeEnd::Read => {
+                if data.buf.len() > 0 {
+                    Ok(PollStatus {
+                        read: true,
+                        write: false,
+                        error: false,
+                    })
+                } else {
+                    Ok(PollStatus {
+                        read: false,
+                        write: false,
+                        error: false,
+                    })
+                }
+            }
+            PipeEnd::Write => {
+                if data.buf.len() > 0 {
+                    Ok(PollStatus {
+                        read: false,
+                        write: true,
+                        error: false,
+                    })
+                } else {
+                    Ok(PollStatus {
+                        read: false,
+                        write: false,
+                        error: false,
+                    })
+                }
+            }
+        }
+    }
     impl_inode!();
 }

kernel/src/fs/pseudo.rs

@@ -0,0 +1,78 @@
+//! Pseudo file system INode
+
+use alloc::{string::String, sync::Arc, vec::Vec};
+use core::any::Any;
+
+use rcore_fs::vfs::*;
+
+pub struct Pseudo {
+    content: Vec<u8>,
+    type_: FileType,
+}
+
+impl Pseudo {
+    pub fn new(s: &str, type_: FileType) -> Self {
+        Pseudo {
+            content: Vec::from(s.as_bytes()),
+            type_,
+        }
+    }
+}
+
+// TODO: better way to provide default impl?
+macro_rules! impl_inode {
+    () => {
+        fn set_metadata(&self, _metadata: &Metadata) -> Result<()> { Ok(()) }
+        fn sync_all(&self) -> Result<()> { Ok(()) }
+        fn sync_data(&self) -> Result<()> { Ok(()) }
+        fn resize(&self, _len: usize) -> Result<()> { Err(FsError::NotSupported) }
+        fn create(&self, _name: &str, _type_: FileType, _mode: u32) -> Result<Arc<INode>> { Err(FsError::NotDir) }
+        fn unlink(&self, _name: &str) -> Result<()> { Err(FsError::NotDir) }
+        fn link(&self, _name: &str, _other: &Arc<INode>) -> Result<()> { Err(FsError::NotDir) }
+        fn move_(&self, _old_name: &str, _target: &Arc<INode>, _new_name: &str) -> Result<()> { Err(FsError::NotDir) }
+        fn find(&self, _name: &str) -> Result<Arc<INode>> { Err(FsError::NotDir) }
+        fn get_entry(&self, _id: usize) -> Result<String> { Err(FsError::NotDir) }
+        fn io_control(&self, cmd: u32, data: usize) -> Result<()> { Err(FsError::NotSupported) }
+        fn fs(&self) -> Arc<FileSystem> { unimplemented!() }
+        fn as_any_ref(&self) -> &Any { self }
+    };
+}
+
+impl INode for Pseudo {
+    fn read_at(&self, offset: usize, buf: &mut [u8]) -> Result<usize> {
+        if offset >= self.content.len() {
+            return Ok(0);
+        }
+        let len = (self.content.len() - offset).min(buf.len());
+        buf[..len].copy_from_slice(&self.content[offset..offset + len]);
+        Ok(len)
+    }
+    fn write_at(&self, _offset: usize, _buf: &[u8]) -> Result<usize> {
+        Err(FsError::NotSupported)
+    }
+    fn poll(&self) -> Result<PollStatus> {
+        Ok(PollStatus {
+            read: true,
+            write: false,
+            error: false,
+        })
+    }
+    fn metadata(&self) -> Result<Metadata> {
+        Ok(Metadata {
+            dev: 0,
+            inode: 0,
+            size: self.content.len(),
+            blk_size: 0,
+            blocks: 0,
+            atime: Timespec { sec: 0, nsec: 0 },
+            mtime: Timespec { sec: 0, nsec: 0 },
+            ctime: Timespec { sec: 0, nsec: 0 },
+            type_: self.type_,
+            mode: 0,
+            nlinks: 0,
+            uid: 0,
+            gid: 0,
+        })
+    }
+    impl_inode!();
+}

kernel/src/fs/stdio.rs

@@ -5,6 +5,7 @@ use core::any::Any;
 
 use rcore_fs::vfs::*;
 
+use super::ioctl::*;
 use crate::sync::Condvar;
 use crate::sync::SpinNoIrqLock as Mutex;
 
@@ -20,16 +21,39 @@ impl Stdin {
         self.pushed.notify_one();
     }
     pub fn pop(&self) -> char {
+        #[cfg(feature = "board_k210")]
         loop {
-            let ret = self.buf.lock().pop_front();
-            match ret {
+            // polling
+            let c = crate::arch::io::getchar();
+            if c != '\0' {
+                return c;
+            }
+        }
+        #[cfg(feature = "board_rocket_chip")]
+        loop {
+            let c = crate::arch::io::getchar();
+            if c != '\0' && c as u8 != 254 {
+                return c;
+            }
+        }
+        #[cfg(not(any(feature = "board_k210", feature = "board_rocket_chip")))]
+        loop {
+            let mut buf_lock = self.buf.lock();
+            match buf_lock.pop_front() {
                 Some(c) => return c,
-                None => self.pushed._wait(),
+                None => {
+                    self.pushed.wait(buf_lock);
+                }
             }
         }
     }
     pub fn can_read(&self) -> bool {
-        self.buf.lock().len() > 0
+        // Currently, rocket-chip implementation rely on htif interface, the serial interrupt DO
+        // NOT work, so return true always
+        #[cfg(feature = "board_rocket_chip")]
+        return true;
+        #[cfg(not(feature = "board_rocket_chip"))]
+        return self.buf.lock().len() > 0;
     }
 }
 
@@ -41,32 +65,6 @@ lazy_static! {
     pub static ref STDOUT: Arc<Stdout> = Arc::new(Stdout::default());
 }
 
-// 32bits total, command in lower 16bits, size of the parameter structure in the lower 14 bits of the upper 16 bits
-// higher 2 bits: 01 = write, 10 = read
-
-#[cfg(not(target_arch = "mips"))]
-const TCGETS: u32 = 0x5401;
-#[cfg(target_arch = "mips")]
-const TCGETS: u32 = 0x540D;
-
-#[cfg(not(target_arch = "mips"))]
-const TIOCGPGRP: u32 = 0x540F;
-// _IOR('t', 119, int)
-#[cfg(target_arch = "mips")]
-const TIOCGPGRP: u32 = 0x4_004_74_77;
-
-#[cfg(not(target_arch = "mips"))]
-const TIOCSPGRP: u32 = 0x5410;
-// _IOW('t', 118, int)
-#[cfg(target_arch = "mips")]
-const TIOCSPGRP: u32 = 0x8_004_74_76;
-
-#[cfg(not(target_arch = "mips"))]
-const TIOCGWINSZ: u32 = 0x5413;
-// _IOR('t', 104, struct winsize)
-#[cfg(target_arch = "mips")]
-const TIOCGWINSZ: u32 = 0x4_008_74_68;
-
 // TODO: better way to provide default impl?
 macro_rules! impl_inode {
     () => {
@@ -82,7 +80,7 @@ macro_rules! impl_inode {
         fn find(&self, _name: &str) -> Result<Arc<INode>> { Err(FsError::NotDir) }
         fn get_entry(&self, _id: usize) -> Result<String> { Err(FsError::NotDir) }
         fn io_control(&self, cmd: u32, data: usize) -> Result<()> {
-            match cmd {
+            match cmd as usize {
                 TCGETS | TIOCGWINSZ | TIOCSPGRP => {
                     // pretend to be tty
                     Ok(())

kernel/src/lib.rs

@@ -18,7 +18,7 @@ extern crate log;
 extern crate lazy_static;
 
 pub use crate::process::{new_kernel_context, processor};
-use buddy_system_allocator::LockedHeap;
+use buddy_system_allocator::LockedHeapWithRescue;
 use rcore_thread::std_thread as thread;
 
 #[macro_use] // print!
@@ -65,4 +65,5 @@ pub fn kmain() -> ! {
 ///
 /// It should be defined in memory mod, but in Rust `global_allocator` must be in root mod.
 #[global_allocator]
-static HEAP_ALLOCATOR: LockedHeap = LockedHeap::empty();
+static HEAP_ALLOCATOR: LockedHeapWithRescue =
+    LockedHeapWithRescue::new(crate::memory::enlarge_heap);

kernel/src/memory.rs

@@ -1,14 +1,30 @@
+//! Define the FrameAllocator for physical memory
+//! x86_64      --  64GB
+//! AARCH64/MIPS/RV --  1GB
+//! K210(rv64)  --  8MB
+//! NOTICE:
+//! type FrameAlloc = bitmap_allocator::BitAllocXXX
+//! KSTACK_SIZE         -- 16KB
+//!
+//! KERNEL_HEAP_SIZE:
+//! x86-64              -- 32MB
+//! AARCH64/RV64        -- 8MB
+//! MIPS/RV32           -- 2MB
+//! mipssim/malta(MIPS) -- 10MB
+
 use super::HEAP_ALLOCATOR;
 pub use crate::arch::paging::*;
-use crate::consts::MEMORY_OFFSET;
-use crate::process::process_unsafe;
+use crate::consts::{KERNEL_OFFSET, MEMORY_OFFSET};
 use crate::sync::SpinNoIrqLock;
+use alloc::boxed::Box;
 use bitmap_allocator::BitAlloc;
-use buddy_system_allocator::LockedHeap;
+use buddy_system_allocator::Heap;
 use lazy_static::*;
 use log::*;
 pub use rcore_memory::memory_set::{handler::*, MemoryArea, MemoryAttr};
+use rcore_memory::paging::PageTable;
 use rcore_memory::*;
+use crate::process::current_thread;
 
 pub type MemorySet = rcore_memory::memory_set::MemorySet<InactivePageTable0>;
 
@@ -16,13 +32,21 @@ pub type MemorySet = rcore_memory::memory_set::MemorySet<InactivePageTable0>;
 #[cfg(target_arch = "x86_64")]
 pub type FrameAlloc = bitmap_allocator::BitAlloc16M;
 
-// RISCV has 1G memory
-#[cfg(any(target_arch = "riscv32", target_arch = "riscv64"))]
+// RISCV, ARM, MIPS has 1G memory
+#[cfg(all(
+    any(
+        target_arch = "riscv32",
+        target_arch = "riscv64",
+        target_arch = "aarch64",
+        target_arch = "mips"
+    ),
+    not(feature = "board_k210")
+))]
 pub type FrameAlloc = bitmap_allocator::BitAlloc1M;
 
-// Raspberry Pi 3 has 1G memory
-#[cfg(any(target_arch = "aarch64", target_arch = "mips"))]
-pub type FrameAlloc = bitmap_allocator::BitAlloc1M;
+// K210 has 8M memory
+#[cfg(feature = "board_k210")]
+pub type FrameAlloc = bitmap_allocator::BitAlloc4K;
 
 lazy_static! {
     pub static ref FRAME_ALLOCATOR: SpinNoIrqLock<FrameAlloc> =
@@ -77,17 +101,17 @@ pub fn dealloc_frame(target: usize) {
 }
 
 pub struct KernelStack(usize);
-const STACK_SIZE: usize = 0x8000;
+const KSTACK_SIZE: usize = 0x4000; //16KB
 
 impl KernelStack {
     pub fn new() -> Self {
         use alloc::alloc::{alloc, Layout};
         let bottom =
-            unsafe { alloc(Layout::from_size_align(STACK_SIZE, STACK_SIZE).unwrap()) } as usize;
+            unsafe { alloc(Layout::from_size_align(KSTACK_SIZE, KSTACK_SIZE).unwrap()) } as usize;
         KernelStack(bottom)
     }
     pub fn top(&self) -> usize {
-        self.0 + STACK_SIZE
+        self.0 + KSTACK_SIZE
     }
 }
 
@@ -97,7 +121,7 @@ impl Drop for KernelStack {
         unsafe {
             dealloc(
                 self.0 as _,
-                Layout::from_size_align(STACK_SIZE, STACK_SIZE).unwrap(),
+                Layout::from_size_align(KSTACK_SIZE, KSTACK_SIZE).unwrap(),
             );
         }
     }
@@ -108,8 +132,8 @@ impl Drop for KernelStack {
 pub fn handle_page_fault(addr: usize) -> bool {
     // debug!("page fault @ {:#x}", addr);
 
-    // This is safe as long as page fault never happens in page fault handler
-    unsafe { process_unsafe().vm.handle_page_fault(addr) }
+    let thread = unsafe { current_thread() };
+    thread.vm.lock().handle_page_fault(addr)
 }
 
 pub fn init_heap() {
@@ -123,5 +147,37 @@ pub fn init_heap() {
     info!("heap init end");
 }
 
-/// Allocator for the rest memory space on NO-MMU case.
-pub static MEMORY_ALLOCATOR: LockedHeap = LockedHeap::empty();
+pub fn enlarge_heap(heap: &mut Heap) {
+    info!("Enlarging heap to avoid oom");
+
+    let mut page_table = active_table();
+    let mut addrs = [(0, 0); 32];
+    let mut addr_len = 0;
+    #[cfg(target_arch = "x86_64")]
+    let va_offset = KERNEL_OFFSET + 0xe0000000;
+    #[cfg(not(target_arch = "x86_64"))]
+    let va_offset = KERNEL_OFFSET + 0x00e00000;
+    for i in 0..16384 {
+        let page = alloc_frame().unwrap();
+        let va = va_offset + page;
+        if addr_len > 0 {
+            let (ref mut addr, ref mut len) = addrs[addr_len - 1];
+            if *addr - PAGE_SIZE == va {
+                *len += PAGE_SIZE;
+                *addr -= PAGE_SIZE;
+                continue;
+            }
+        }
+        addrs[addr_len] = (va, PAGE_SIZE);
+        addr_len += 1;
+    }
+    for (addr, len) in addrs[..addr_len].into_iter() {
+        for va in (*addr..(*addr + *len)).step_by(PAGE_SIZE) {
+            page_table.map(va, va - va_offset).update();
+        }
+        info!("Adding {:#X} {:#X} to heap", addr, len);
+        unsafe {
+            heap.init(*addr, *len);
+        }
+    }
+}

kernel/src/net/structs.rs

@@ -4,6 +4,7 @@ use crate::sync::SpinNoIrqLock as Mutex;
 use crate::syscall::*;
 use crate::util;
 use alloc::boxed::Box;
+use alloc::fmt::Debug;
 use alloc::sync::Arc;
 use alloc::vec::Vec;
 use bitflags::*;
@@ -50,12 +51,12 @@ pub enum Endpoint {
 }
 
 /// Common methods that a socket must have
-pub trait Socket: Send + Sync {
+pub trait Socket: Send + Sync + Debug {
     fn read(&self, data: &mut [u8]) -> (SysResult, Endpoint);
     fn write(&self, data: &[u8], sendto_endpoint: Option<Endpoint>) -> SysResult;
     fn poll(&self) -> (bool, bool, bool); // (in, out, err)
     fn connect(&mut self, endpoint: Endpoint) -> SysResult;
-    fn bind(&mut self, endpoint: Endpoint) -> SysResult {
+    fn bind(&mut self, _endpoint: Endpoint) -> SysResult {
         Err(SysError::EINVAL)
     }
     fn listen(&mut self) -> SysResult {
@@ -73,11 +74,11 @@ pub trait Socket: Send + Sync {
     fn remote_endpoint(&self) -> Option<Endpoint> {
         None
     }
-    fn setsockopt(&mut self, level: usize, opt: usize, data: &[u8]) -> SysResult {
+    fn setsockopt(&mut self, _level: usize, _opt: usize, _data: &[u8]) -> SysResult {
         warn!("setsockopt is unimplemented");
         Ok(0)
     }
-    fn ioctl(&mut self, request: usize, arg1: usize, arg2: usize, arg3: usize) -> SysResult {
+    fn ioctl(&mut self, _request: usize, _arg1: usize, _arg2: usize, _arg3: usize) -> SysResult {
         warn!("ioctl is unimplemented for this socket");
         Ok(0)
     }
@@ -265,9 +266,8 @@ impl Socket for TcpSocketState {
                             TcpState::SynSent => {
                                 // still connecting
                                 drop(socket);
-                                drop(sockets);
                                 debug!("poll for connection wait");
-                                SOCKET_ACTIVITY._wait();
+                                SOCKET_ACTIVITY.wait(sockets);
                             }
                             TcpState::Established => {
                                 break Ok(0);
@@ -285,7 +285,7 @@ impl Socket for TcpSocketState {
         }
     }
 
-    fn bind(&mut self, mut endpoint: Endpoint) -> SysResult {
+    fn bind(&mut self, endpoint: Endpoint) -> SysResult {
         if let Endpoint::Ip(mut ip) = endpoint {
             if ip.port == 0 {
                 ip.port = get_ephemeral_port();
@@ -357,10 +357,8 @@ impl Socket for TcpSocketState {
                 return Ok((new_socket, Endpoint::Ip(remote_endpoint)));
             }
 
-            // avoid deadlock
             drop(socket);
-            drop(sockets);
-            SOCKET_ACTIVITY._wait();
+            SOCKET_ACTIVITY.wait(sockets);
         }
     }
 
@@ -447,9 +445,8 @@ impl Socket for UdpSocketState {
                 );
             }
 
-            // avoid deadlock
             drop(socket);
-            SOCKET_ACTIVITY._wait()
+            SOCKET_ACTIVITY.wait(sockets);
         }
     }
 
@@ -626,10 +623,8 @@ impl Socket for RawSocketState {
                 );
             }
 
-            // avoid deadlock
             drop(socket);
-            drop(sockets);
-            SOCKET_ACTIVITY._wait()
+            SOCKET_ACTIVITY.wait(sockets);
         }
     }
 
@@ -941,7 +936,7 @@ impl Socket for NetlinkSocketState {
                 let ifaces = NET_DRIVERS.read();
                 for i in 0..ifaces.len() {
                     let mut msg = Vec::new();
-                    let mut new_header = NetlinkMessageHeader {
+                    let new_header = NetlinkMessageHeader {
                         nlmsg_len: 0, // to be determined later
                         nlmsg_type: NetlinkMessageType::NewLink.into(),
                         nlmsg_flags: NetlinkMessageFlags::MULTI,
@@ -999,7 +994,7 @@ impl Socket for NetlinkSocketState {
                     let ip_addrs = ifaces[i].get_ip_addresses();
                     for j in 0..ip_addrs.len() {
                         let mut msg = Vec::new();
-                        let mut new_header = NetlinkMessageHeader {
+                        let new_header = NetlinkMessageHeader {
                             nlmsg_len: 0, // to be determined later
                             nlmsg_type: NetlinkMessageType::NewAddr.into(),
                             nlmsg_flags: NetlinkMessageFlags::MULTI,
@@ -1045,7 +1040,7 @@ impl Socket for NetlinkSocketState {
             _ => {}
         }
         let mut msg = Vec::new();
-        let mut new_header = NetlinkMessageHeader {
+        let new_header = NetlinkMessageHeader {
             nlmsg_len: 0, // to be determined later
             nlmsg_type: NetlinkMessageType::Done.into(),
             nlmsg_flags: NetlinkMessageFlags::MULTI,

kernel/src/process/mod.rs

@@ -20,7 +20,7 @@ pub fn init() {
         }
     }
 
-    crate::shell::run_user_shell();
+    crate::shell::add_user_shell();
 
     info!("process: init end");
 }
@@ -109,25 +109,13 @@ static PROCESSORS: [Processor; MAX_CPU_NUM] = [
     //    Processor::new(),    Processor::new(),    Processor::new(),    Processor::new(),
 ];
 
-/// Get current process
-pub fn process() -> MutexGuard<'static, Process, SpinNoIrq> {
-    current_thread().proc.lock()
-}
-
-/// Get current process, ignoring its lock
-/// Only use this when necessary
-pub unsafe fn process_unsafe() -> MutexGuard<'static, Process, SpinNoIrq> {
-    let thread = current_thread();
-    thread.proc.force_unlock();
-    thread.proc.lock()
-}
-
 /// Get current thread
 ///
-/// FIXME: It's obviously unsafe to get &mut !
-pub fn current_thread() -> &'static mut Thread {
-    use core::mem::transmute;
-    let (process, _): (&mut Thread, *const ()) = unsafe { transmute(processor().context()) };
+/// `Thread` is a thread-local object.
+/// It is safe to call this once, and pass `&mut Thread` as a function argument.
+pub unsafe fn current_thread() -> &'static mut Thread {
+    // trick: force downcast from trait object
+    let (process, _): (&mut Thread, *const ()) = core::mem::transmute(processor().context());
     process
 }
 

kernel/src/process/structs.rs

@@ -13,68 +13,55 @@ use xmas_elf::{
 };
 
 use crate::arch::interrupt::{Context, TrapFrame};
-use crate::fs::{FileHandle, FileLike, INodeExt, OpenOptions, FOLLOW_MAX_DEPTH};
-use crate::memory::{ByFrame, GlobalFrameAlloc, KernelStack, MemoryAttr, MemorySet};
-use crate::net::SOCKETS;
+use crate::fs::{FileHandle, FileLike, OpenOptions, FOLLOW_MAX_DEPTH};
+use crate::memory::{
+    ByFrame, Delay, File, GlobalFrameAlloc, KernelStack, MemoryAttr, MemorySet, Read,
+};
 use crate::sync::{Condvar, SpinNoIrqLock as Mutex};
 
 use super::abi::{self, ProcInitInfo};
+use core::mem::uninitialized;
+use rcore_fs::vfs::INode;
 
-// TODO: avoid pub
 pub struct Thread {
-    pub context: Context,
-    pub kstack: KernelStack,
+    context: Context,
+    kstack: KernelStack,
     /// Kernel performs futex wake when thread exits.
     /// Ref: [http://man7.org/linux/man-pages/man2/set_tid_address.2.html]
     pub clear_child_tid: usize,
+    // This is same as `proc.vm`
+    pub vm: Arc<Mutex<MemorySet>>,
     pub proc: Arc<Mutex<Process>>,
 }
 
 /// Pid type
 /// For strong type separation
-#[derive(Clone, PartialEq, Eq, PartialOrd, Ord)]
-pub struct Pid(Option<usize>);
+#[derive(Debug, Clone, PartialEq, Eq, PartialOrd, Ord)]
+pub struct Pid(usize);
 
 impl Pid {
-    pub fn uninitialized() -> Self {
-        Pid(None)
-    }
-
-    /// Return if it was uninitialized before this call
-    /// When returning true, it usually means this is the first thread
-    pub fn set_if_uninitialized(&mut self, tid: Tid) -> bool {
-        if self.0 == None {
-            self.0 = Some(tid as usize);
-            true
-        } else {
-            false
-        }
-    }
-
     pub fn get(&self) -> usize {
-        self.0.unwrap()
+        self.0
     }
 
     /// Return whether this pid represents the init process
     pub fn is_init(&self) -> bool {
-        self.0 == Some(0)
+        self.0 == 0
     }
 }
 
 impl fmt::Display for Pid {
     fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
-        match self.0 {
-            Some(pid) => write!(f, "{}", pid),
-            None => write!(f, "None"),
-        }
+        write!(f, "{}", self.0)
     }
 }
 
 pub struct Process {
     // resources
-    pub vm: MemorySet,
+    pub vm: Arc<Mutex<MemorySet>>,
     pub files: BTreeMap<usize, FileLike>,
     pub cwd: String,
+    pub exec_path: String,
     futexes: BTreeMap<usize, Arc<Condvar>>,
 
     // relationship
@@ -103,21 +90,9 @@ impl rcore_thread::Context for Thread {
     }
 
     fn set_tid(&mut self, tid: Tid) {
-        // set pid=tid if unspecified
         let mut proc = self.proc.lock();
-        if proc.pid.set_if_uninitialized(tid) {
-            // first thread in the process
-            // link to its ppid
-            if let Some(parent) = &proc.parent {
-                let mut parent = parent.lock();
-                parent.children.push(Arc::downgrade(&self.proc));
-            }
-        }
         // add it to threads
         proc.threads.push(tid);
-        PROCESSES
-            .write()
-            .insert(proc.pid.get(), Arc::downgrade(&self.proc));
     }
 }
 
@@ -126,52 +101,63 @@ impl Thread {
     pub unsafe fn new_init() -> Box<Thread> {
         Box::new(Thread {
             context: Context::null(),
-            kstack: KernelStack::new(),
-            clear_child_tid: 0,
-            // safety: this field will never be used
-            proc: core::mem::uninitialized(),
+            // safety: other fields will never be used
+            ..core::mem::uninitialized()
         })
     }
 
     /// Make a new kernel thread starting from `entry` with `arg`
     pub fn new_kernel(entry: extern "C" fn(usize) -> !, arg: usize) -> Box<Thread> {
         let vm = MemorySet::new();
+        let vm_token = vm.token();
+        let vm = Arc::new(Mutex::new(vm));
         let kstack = KernelStack::new();
         Box::new(Thread {
-            context: unsafe { Context::new_kernel_thread(entry, arg, kstack.top(), vm.token()) },
+            context: unsafe { Context::new_kernel_thread(entry, arg, kstack.top(), vm_token) },
             kstack,
             clear_child_tid: 0,
+            vm: vm.clone(),
             // TODO: kernel thread should not have a process
-            proc: Arc::new(Mutex::new(Process {
+            proc: Process {
                 vm,
                 files: BTreeMap::default(),
                 cwd: String::from("/"),
+                exec_path: String::new(),
                 futexes: BTreeMap::default(),
-                pid: Pid::uninitialized(),
+                pid: Pid(0),
                 parent: None,
                 children: Vec::new(),
                 threads: Vec::new(),
                 child_exit: Arc::new(Condvar::new()),
                 child_exit_code: BTreeMap::new(),
-            })),
+            }
+            .add_to_table(),
         })
     }
 
-    /// Make a new user process from ELF `data`
-    pub fn new_user(
-        data: &[u8],
+    /// Construct virtual memory of a new user process from ELF `data`.
+    /// Return `(MemorySet, entry_point, ustack_top)`
+    pub fn new_user_vm(
+        inode: &Arc<INode>,
         exec_path: &str,
         mut args: Vec<String>,
         envs: Vec<String>,
-    ) -> Box<Thread> {
+    ) -> Result<(MemorySet, usize, usize), &'static str> {
+        // Read ELF header
+        // 0x3c0: magic number from ld-musl.so
+        let mut data: [u8; 0x3c0] = unsafe { uninitialized() };
+        inode
+            .read_at(0, &mut data)
+            .map_err(|_| "failed to read from INode")?;
+
         // Parse ELF
-        let elf = ElfFile::new(data).expect("failed to read elf");
+        let elf = ElfFile::new(&data)?;
 
         // Check ELF type
         match elf.header.pt2.type_().as_type() {
             header::Type::Executable => {}
             header::Type::SharedObject => {}
-            _ => panic!("ELF is not executable or shared object"),
+            _ => return Err("ELF is not executable or shared object"),
         }
 
         // Check ELF arch
@@ -184,31 +170,25 @@ impl Thread {
             header::Machine::Other(243) => {}
             #[cfg(target_arch = "mips")]
             header::Machine::Mips => {}
-            machine @ _ => panic!("invalid elf arch: {:?}", machine),
+            _ => return Err("invalid ELF arch"),
         }
 
         // Check interpreter (for dynamic link)
         if let Ok(loader_path) = elf.get_interpreter() {
             // assuming absolute path
-            if let Ok(inode) = crate::fs::ROOT_INODE.lookup_follow(loader_path, FOLLOW_MAX_DEPTH) {
-                if let Ok(buf) = inode.read_as_vec() {
-                    // Elf loader should not have INTERP
-                    // No infinite loop
-                    args.insert(0, loader_path.into());
-                    args.insert(1, exec_path.into());
-                    args.remove(2);
-                    info!("loader args: {:?}", args);
-                    return Thread::new_user(buf.as_slice(), exec_path, args, envs);
-                } else {
-                    warn!("loader specified as {} but failed to read", &loader_path);
-                }
-            } else {
-                warn!("loader specified as {} but not found", &loader_path);
-            }
+            let inode = crate::fs::ROOT_INODE
+                .lookup_follow(loader_path, FOLLOW_MAX_DEPTH)
+                .map_err(|_| "interpreter not found")?;
+            // modify args for loader
+            args[0] = exec_path.into();
+            args.insert(0, loader_path.into());
+            // Elf loader should not have INTERP
+            // No infinite loop
+            return Thread::new_user_vm(&inode, exec_path, args, envs);
         }
 
         // Make page table
-        let mut vm = elf.make_memory_set();
+        let mut vm = elf.make_memory_set(inode);
 
         // User stack
         use crate::consts::{USER_STACK_OFFSET, USER_STACK_SIZE};
@@ -217,6 +197,14 @@ impl Thread {
             let ustack_top = USER_STACK_OFFSET + USER_STACK_SIZE;
             vm.push(
                 ustack_buttom,
+                ustack_top - PAGE_SIZE * 4,
+                MemoryAttr::default().user(),
+                Delay::new(GlobalFrameAlloc),
+                "user_stack_delay",
+            );
+            // We are going to write init info now. So map the last 4 pages eagerly.
+            vm.push(
+                ustack_top - PAGE_SIZE * 4,
                 ustack_top,
                 MemoryAttr::default().user(),
                 ByFrame::new(GlobalFrameAlloc),
@@ -246,6 +234,21 @@ impl Thread {
 
         trace!("{:#x?}", vm);
 
+        let entry_addr = elf.header.pt2.entry_point() as usize;
+        Ok((vm, entry_addr, ustack_top))
+    }
+
+    /// Make a new user process from ELF `data`
+    pub fn new_user(
+        inode: &Arc<INode>,
+        exec_path: &str,
+        args: Vec<String>,
+        envs: Vec<String>,
+    ) -> Box<Thread> {
+        let (vm, entry_addr, ustack_top) = Self::new_user_vm(inode, exec_path, args, envs).unwrap();
+
+        let vm_token = vm.token();
+        let vm = Arc::new(Mutex::new(vm));
         let kstack = KernelStack::new();
 
         let mut files = BTreeMap::new();
@@ -258,6 +261,7 @@ impl Thread {
                     write: false,
                     append: false,
                 },
+                String::from("stdin"),
             )),
         );
         files.insert(
@@ -269,6 +273,7 @@ impl Thread {
                     write: true,
                     append: false,
                 },
+                String::from("stdout"),
             )),
         );
         files.insert(
@@ -280,69 +285,66 @@ impl Thread {
                     write: true,
                     append: false,
                 },
+                String::from("stderr"),
             )),
         );
 
-        let entry_addr = elf.header.pt2.entry_point() as usize;
-
         Box::new(Thread {
             context: unsafe {
-                Context::new_user_thread(entry_addr, ustack_top, kstack.top(), vm.token())
+                Context::new_user_thread(entry_addr, ustack_top, kstack.top(), vm_token)
             },
             kstack,
             clear_child_tid: 0,
-            proc: Arc::new(Mutex::new(Process {
+            vm: vm.clone(),
+            proc: Process {
                 vm,
                 files,
                 cwd: String::from("/"),
+                exec_path: String::from(exec_path),
                 futexes: BTreeMap::default(),
-                pid: Pid::uninitialized(),
+                pid: Pid(0),
                 parent: None,
                 children: Vec::new(),
                 threads: Vec::new(),
                 child_exit: Arc::new(Condvar::new()),
                 child_exit_code: BTreeMap::new(),
-            })),
+            }
+            .add_to_table(),
         })
     }
 
     /// Fork a new process from current one
     pub fn fork(&self, tf: &TrapFrame) -> Box<Thread> {
-        // Clone memory set, make a new page table
-        let proc = self.proc.lock();
-        let vm = proc.vm.clone();
-        let files = proc.files.clone();
-        let cwd = proc.cwd.clone();
-        drop(proc);
-        let parent = Some(self.proc.clone());
-        debug!("fork: finish clone MemorySet");
-
-        // MMU:   copy data to the new space
-        // NoMMU: coping data has been done in `vm.clone()`
-        for area in vm.iter() {
-            let data = Vec::<u8>::from(unsafe { area.as_slice() });
-            unsafe { vm.with(|| area.as_slice_mut().copy_from_slice(data.as_slice())) }
-        }
-
-        debug!("fork: temporary copy data!");
         let kstack = KernelStack::new();
+        let vm = self.vm.lock().clone();
+        let vm_token = vm.token();
+        let vm = Arc::new(Mutex::new(vm));
+        let context = unsafe { Context::new_fork(tf, kstack.top(), vm_token) };
+
+        let mut proc = self.proc.lock();
+        let new_proc = Process {
+            vm: vm.clone(),
+            files: proc.files.clone(),
+            cwd: proc.cwd.clone(),
+            exec_path: proc.exec_path.clone(),
+            futexes: BTreeMap::default(),
+            pid: Pid(0),
+            parent: Some(self.proc.clone()),
+            children: Vec::new(),
+            threads: Vec::new(),
+            child_exit: Arc::new(Condvar::new()),
+            child_exit_code: BTreeMap::new(),
+        }
+        .add_to_table();
+        // link to parent
+        proc.children.push(Arc::downgrade(&new_proc));
 
         Box::new(Thread {
-            context: unsafe { Context::new_fork(tf, kstack.top(), vm.token()) },
+            context,
             kstack,
             clear_child_tid: 0,
-            proc: Arc::new(Mutex::new(Process {
-                vm,
-                files,
-                cwd,
-                futexes: BTreeMap::default(),
-                pid: Pid::uninitialized(),
-                parent,
-                children: Vec::new(),
-                threads: Vec::new(),
-                child_exit: Arc::new(Condvar::new()),
-                child_exit_code: BTreeMap::new(),
-            })),
+            vm,
+            proc: new_proc,
         })
     }
 
@@ -355,33 +357,52 @@ impl Thread {
         clear_child_tid: usize,
     ) -> Box<Thread> {
         let kstack = KernelStack::new();
-        let token = self.proc.lock().vm.token();
+        let vm_token = self.vm.lock().token();
         Box::new(Thread {
-            context: unsafe { Context::new_clone(tf, stack_top, kstack.top(), token, tls) },
+            context: unsafe { Context::new_clone(tf, stack_top, kstack.top(), vm_token, tls) },
             kstack,
             clear_child_tid,
+            vm: self.vm.clone(),
             proc: self.proc.clone(),
         })
     }
 }
 
 impl Process {
-    pub fn get_free_fd(&self) -> usize {
+    /// Assign a pid and put itself to global process table.
+    fn add_to_table(mut self) -> Arc<Mutex<Self>> {
+        let mut process_table = PROCESSES.write();
+
+        // assign pid
+        let pid = (0..)
+            .find(|i| match process_table.get(i) {
+                Some(p) if p.upgrade().is_some() => false,
+                _ => true,
+            })
+            .unwrap();
+        self.pid = Pid(pid);
+
+        // put to process table
+        let self_ref = Arc::new(Mutex::new(self));
+        process_table.insert(pid, Arc::downgrade(&self_ref));
+
+        self_ref
+    }
+    fn get_free_fd(&self) -> usize {
         (0..).find(|i| !self.files.contains_key(i)).unwrap()
     }
+    /// Add a file to the process, return its fd.
+    pub fn add_file(&mut self, file_like: FileLike) -> usize {
+        let fd = self.get_free_fd();
+        self.files.insert(fd, file_like);
+        fd
+    }
     pub fn get_futex(&mut self, uaddr: usize) -> Arc<Condvar> {
         if !self.futexes.contains_key(&uaddr) {
             self.futexes.insert(uaddr, Arc::new(Condvar::new()));
         }
         self.futexes.get(&uaddr).unwrap().clone()
     }
-    pub fn clone_for_exec(&mut self, other: &Self) {
-        self.files = other.files.clone();
-        self.cwd = other.cwd.clone();
-        self.pid = other.pid.clone();
-        self.parent = other.parent.clone();
-        self.threads = other.threads.clone();
-    }
 }
 
 trait ToMemoryAttr {
@@ -391,10 +412,12 @@ trait ToMemoryAttr {
 impl ToMemoryAttr for Flags {
     fn to_attr(&self) -> MemoryAttr {
         let mut flags = MemoryAttr::default().user();
-        // FIXME: handle readonly
         if self.is_execute() {
             flags = flags.execute();
         }
+        if !self.is_write() {
+            flags = flags.readonly();
+        }
         flags
     }
 }
@@ -402,7 +425,7 @@ impl ToMemoryAttr for Flags {
 /// Helper functions to process ELF file
 trait ElfExt {
     /// Generate a MemorySet according to the ELF file.
-    fn make_memory_set(&self) -> MemorySet;
+    fn make_memory_set(&self, inode: &Arc<INode>) -> MemorySet;
 
     /// Get interpreter string if it has.
     fn get_interpreter(&self) -> Result<&str, &str>;
@@ -412,7 +435,7 @@ trait ElfExt {
 }
 
 impl ElfExt for ElfFile<'_> {
-    fn make_memory_set(&self) -> MemorySet {
+    fn make_memory_set(&self, inode: &Arc<INode>) -> MemorySet {
         debug!("creating MemorySet from ELF");
         let mut ms = MemorySet::new();
 
@@ -420,33 +443,19 @@ impl ElfExt for ElfFile<'_> {
             if ph.get_type() != Ok(Type::Load) {
                 continue;
             }
-            let virt_addr = ph.virtual_addr() as usize;
-            let mem_size = ph.mem_size() as usize;
-            let data = match ph.get_data(self).unwrap() {
-                SegmentData::Undefined(data) => data,
-                _ => unreachable!(),
-            };
-
-            // Get target slice
-            let target = {
-                ms.push(
-                    virt_addr,
-                    virt_addr + mem_size,
-                    ph.flags().to_attr(),
-                    ByFrame::new(GlobalFrameAlloc),
-                    "elf",
-                );
-                unsafe { ::core::slice::from_raw_parts_mut(virt_addr as *mut u8, mem_size) }
-            };
-            // Copy data
-            unsafe {
-                ms.with(|| {
-                    if data.len() != 0 {
-                        target[..data.len()].copy_from_slice(data);
-                    }
-                    target[data.len()..].iter_mut().for_each(|x| *x = 0);
-                });
-            }
+            ms.push(
+                ph.virtual_addr() as usize,
+                ph.virtual_addr() as usize + ph.mem_size() as usize,
+                ph.flags().to_attr(),
+                File {
+                    file: INodeForMap(inode.clone()),
+                    mem_start: ph.virtual_addr() as usize,
+                    file_start: ph.offset() as usize,
+                    file_end: ph.offset() as usize + ph.file_size() as usize,
+                    allocator: GlobalFrameAlloc,
+                },
+                "elf",
+            );
         }
         ms
     }
@@ -488,3 +497,12 @@ impl ElfExt for ElfFile<'_> {
         }
     }
 }
+
+#[derive(Clone)]
+pub struct INodeForMap(pub Arc<INode>);
+
+impl Read for INodeForMap {
+    fn read_at(&self, offset: usize, buf: &mut [u8]) -> usize {
+        self.0.read_at(offset, buf).unwrap()
+    }
+}

kernel/src/shell.rs

@@ -1,33 +1,51 @@
 //! Kernel shell
 
-use crate::drivers::CMDLINE;
-use crate::fs::{INodeExt, ROOT_INODE};
+use crate::arch::io;
+use crate::fs::ROOT_INODE;
 use crate::process::*;
 use alloc::string::String;
 use alloc::vec::Vec;
 
 #[cfg(not(feature = "run_cmdline"))]
-pub fn run_user_shell() {
-    if let Ok(inode) = ROOT_INODE.lookup("busybox") {
-        let data = inode.read_as_vec().unwrap();
-        processor().manager().add(Thread::new_user(
-            data.as_slice(),
-            "busybox",
-            vec!["busybox".into(), "sh".into()],
-            Vec::new(),
-        ));
+pub fn add_user_shell() {
+    // the busybox of alpine linux can not transfer env vars into child process
+    // Now we use busybox from
+    // https://raw.githubusercontent.com/docker-library/busybox/82bc0333a9ae148fbb4246bcbff1487b3fc0c510/musl/busybox.tar.xz -O busybox.tar.xz
+    // This one can transfer env vars!
+    // Why???
+
+    //    #[cfg(target_arch = "x86_64")]
+    //        let init_shell="/bin/busybox"; // from alpine linux
+    //
+    //    #[cfg(not(target_arch = "x86_64"))]
+    let init_shell = "/busybox"; //from docker-library
+
+    #[cfg(target_arch = "x86_64")]
+    let init_envs =
+        vec!["PATH=/usr/sbin:/usr/bin:/sbin:/bin:/usr/x86_64-alpine-linux-musl/bin".into()];
+
+    #[cfg(not(target_arch = "x86_64"))]
+    let init_envs = Vec::new();
+
+    let init_args = vec!["busybox".into(), "ash".into()];
+
+    if let Ok(inode) = ROOT_INODE.lookup(init_shell) {
+        processor()
+            .manager()
+            .add(Thread::new_user(&inode, init_shell, init_args, init_envs));
     } else {
         processor().manager().add(Thread::new_kernel(shell, 0));
     }
 }
 
 #[cfg(feature = "run_cmdline")]
-pub fn run_user_shell() {
+pub fn add_user_shell() {
+    use crate::drivers::CMDLINE;
     let cmdline = CMDLINE.read();
     let inode = ROOT_INODE.lookup(&cmdline).unwrap();
-    let data = inode.read_as_vec().unwrap();
     processor().manager().add(Thread::new_user(
-        data.as_slice(),
+        &inode,
+        &cmdline,
         cmdline.split(' ').map(|s| s.into()).collect(),
         Vec::new(),
     ));
@@ -45,16 +63,14 @@ pub extern "C" fn shell(_arg: usize) -> ! {
             continue;
         }
         let name = cmd.trim().split(' ').next().unwrap();
-        if let Ok(file) = ROOT_INODE.lookup(name) {
-            let data = file.read_as_vec().unwrap();
-            let _pid = processor().manager().add(Thread::new_user(
-                data.as_slice(),
+        if let Ok(inode) = ROOT_INODE.lookup(name) {
+            let _tid = processor().manager().add(Thread::new_user(
+                &inode,
                 &cmd,
                 cmd.split(' ').map(|s| s.into()).collect(),
                 Vec::new(),
             ));
         // TODO: wait until process exits, or use user land shell completely
-        //unsafe { thread::JoinHandle::<()>::_of(pid) }.join().unwrap();
         } else {
             println!("Program not exist");
         }

kernel/src/sync/condvar.rs

@@ -15,6 +15,7 @@ impl Condvar {
     }
 
     /// Park current thread and wait for this condvar to be notified.
+    #[deprecated(note = "this may leads to lost wakeup problem. please use `wait` instead.")]
     pub fn _wait(&self) {
         // The condvar might be notified between adding to queue and thread parking.
         // So park current thread before wait queue lock is freed.
@@ -25,6 +26,7 @@ impl Condvar {
         });
     }
 
+    #[deprecated(note = "this may leads to lost wakeup problem. please use `wait` instead.")]
     pub fn wait_any(condvars: &[&Condvar]) {
         let token = Arc::new(thread::current());
         // Avoid racing in the same way as the function above
@@ -40,19 +42,23 @@ impl Condvar {
         });
     }
 
-    pub fn add_to_wait_queue(&self) -> MutexGuard<VecDeque<Arc<thread::Thread>>, SpinNoIrq> {
+    fn add_to_wait_queue(&self) -> MutexGuard<VecDeque<Arc<thread::Thread>>, SpinNoIrq> {
         let mut lock = self.wait_queue.lock();
         lock.push_back(Arc::new(thread::current()));
         return lock;
     }
 
+    /// Park current thread and wait for this condvar to be notified.
     pub fn wait<'a, T, S>(&self, guard: MutexGuard<'a, T, S>) -> MutexGuard<'a, T, S>
     where
         S: MutexSupport,
     {
         let mutex = guard.mutex;
-        drop(guard);
-        self._wait();
+        let lock = self.add_to_wait_queue();
+        thread::park_action(move || {
+            drop(lock);
+            drop(guard);
+        });
         mutex.lock()
     }
 
@@ -80,7 +86,4 @@ impl Condvar {
         }
         count
     }
-    pub fn _clear(&self) {
-        self.wait_queue.lock().clear();
-    }
 }

kernel/src/syscall/custom.rs

@@ -3,53 +3,56 @@ use super::*;
 use rcore_memory::memory_set::handler::Linear;
 use rcore_memory::memory_set::MemoryAttr;
 
-/// Allocate this PCI device to user space
-/// The kernel driver using the PCI device will be unloaded
-#[cfg(target_arch = "x86_64")]
-pub fn sys_map_pci_device(vendor: usize, product: usize) -> SysResult {
-    use crate::drivers::bus::pci;
-    info!(
-        "map_pci_device: vendor: {:x}, product: {:x}",
-        vendor, product
-    );
+impl Syscall<'_> {
+    /// Allocate this PCI device to user space
+    /// The kernel driver using the PCI device will be unloaded
+    #[cfg(target_arch = "x86_64")]
+    pub fn sys_map_pci_device(&mut self, vendor: usize, product: usize) -> SysResult {
+        use crate::drivers::bus::pci;
+        info!(
+            "map_pci_device: vendor: {:x}, product: {:x}",
+            vendor, product
+        );
 
-    let tag = pci::find_device(vendor as u16, product as u16).ok_or(SysError::ENOENT)?;
-    if pci::detach_driver(&tag) {
-        info!("Kernel driver detached");
+        let tag = pci::find_device(vendor as u16, product as u16).ok_or(SysError::ENOENT)?;
+        if pci::detach_driver(&tag) {
+            info!("Kernel driver detached");
+        }
+
+        // Get BAR0 memory
+        let (base, len) = pci::get_bar0_mem(tag).ok_or(SysError::ENOENT)?;
+
+        let virt_addr = self.vm().find_free_area(0, len);
+        let attr = MemoryAttr::default().user();
+        self.vm().push(
+            virt_addr,
+            virt_addr + len,
+            attr,
+            Linear::new(base as isize - virt_addr as isize),
+            "pci",
+        );
+        Ok(virt_addr)
     }
 
-    // Get BAR0 memory
-    let (base, len) = pci::get_bar0_mem(tag).ok_or(SysError::ENOENT)?;
-
-    let mut proc = process();
-    let virt_addr = proc.vm.find_free_area(0, len);
-    let attr = MemoryAttr::default().user();
-    proc.vm.push(
-        virt_addr,
-        virt_addr + len,
-        attr,
-        Linear::new(base as isize - virt_addr as isize),
-        "pci",
-    );
-    Ok(virt_addr)
-}
-
-#[cfg(not(target_arch = "x86_64"))]
-pub fn sys_map_pci_device(vendor: usize, product: usize) -> SysResult {
-    Err(SysError::ENOSYS)
-}
-
-/// Get start physical addresses of frames
-/// mapped to a list of virtual addresses.
-pub fn sys_get_paddr(vaddrs: *const u64, paddrs: *mut u64, count: usize) -> SysResult {
-    let mut proc = process();
-    proc.vm.check_read_array(vaddrs, count)?;
-    proc.vm.check_write_array(paddrs, count)?;
-    let vaddrs = unsafe { slice::from_raw_parts(vaddrs, count) };
-    let paddrs = unsafe { slice::from_raw_parts_mut(paddrs, count) };
-    for i in 0..count {
-        let paddr = proc.vm.translate(vaddrs[i] as usize).unwrap_or(0);
-        paddrs[i] = paddr as u64;
+    #[cfg(not(target_arch = "x86_64"))]
+    pub fn sys_map_pci_device(&mut self, vendor: usize, product: usize) -> SysResult {
+        Err(SysError::ENOSYS)
+    }
+
+    /// Get start physical addresses of frames
+    /// mapped to a list of virtual addresses.
+    pub fn sys_get_paddr(
+        &mut self,
+        vaddrs: *const u64,
+        paddrs: *mut u64,
+        count: usize,
+    ) -> SysResult {
+        let vaddrs = unsafe { self.vm().check_read_array(vaddrs, count)? };
+        let paddrs = unsafe { self.vm().check_write_array(paddrs, count)? };
+        for i in 0..count {
+            let paddr = self.vm().translate(vaddrs[i] as usize).unwrap_or(0);
+            paddrs[i] = paddr as u64;
+        }
+        Ok(0)
     }
-    Ok(0)
 }

kernel/src/syscall/mem.rs

@@ -1,4 +1,4 @@
-use rcore_memory::memory_set::handler::{ByFrame, Delay};
+use rcore_memory::memory_set::handler::{Delay, File};
 use rcore_memory::memory_set::MemoryAttr;
 use rcore_memory::paging::PageTable;
 use rcore_memory::Page;
@@ -8,108 +8,95 @@ use crate::memory::GlobalFrameAlloc;
 
 use super::*;
 
-pub fn sys_mmap(
-    mut addr: usize,
-    len: usize,
-    prot: usize,
-    flags: usize,
-    fd: usize,
-    offset: usize,
-) -> SysResult {
-    let prot = MmapProt::from_bits_truncate(prot);
-    let flags = MmapFlags::from_bits_truncate(flags);
-    info!(
-        "mmap: addr={:#x}, size={:#x}, prot={:?}, flags={:?}, fd={}, offset={:#x}",
-        addr, len, prot, flags, fd, offset
-    );
-
-    let mut proc = process();
-    if addr == 0 {
-        // although NULL can be a valid address
-        // but in C, NULL is regarded as allocation failure
-        // so just skip it
-        addr = PAGE_SIZE;
-    }
-
-    if flags.contains(MmapFlags::FIXED) {
-        // we have to map it to addr, so remove the old mapping first
-        proc.vm.pop_with_split(addr, addr + len);
-    } else {
-        addr = proc.vm.find_free_area(addr, len);
-    }
-
-    if flags.contains(MmapFlags::ANONYMOUS) {
-        if flags.contains(MmapFlags::SHARED) {
-            return Err(SysError::EINVAL);
-        }
-        proc.vm.push(
-            addr,
-            addr + len,
-            prot.to_attr(),
-            Delay::new(GlobalFrameAlloc),
-            "mmap_anon",
+impl Syscall<'_> {
+    pub fn sys_mmap(
+        &mut self,
+        mut addr: usize,
+        len: usize,
+        prot: usize,
+        flags: usize,
+        fd: usize,
+        offset: usize,
+    ) -> SysResult {
+        let prot = MmapProt::from_bits_truncate(prot);
+        let flags = MmapFlags::from_bits_truncate(flags);
+        info!(
+            "mmap: addr={:#x}, size={:#x}, prot={:?}, flags={:?}, fd={}, offset={:#x}",
+            addr, len, prot, flags, fd, offset
         );
-        return Ok(addr);
-    } else {
-        // only check
-        let _ = proc.get_file(fd)?;
 
-        // TODO: delay mmap file
-        proc.vm.push(
-            addr,
-            addr + len,
-            prot.to_attr(),
-            ByFrame::new(GlobalFrameAlloc),
-            "mmap_file",
+        let mut proc = self.process();
+        if addr == 0 {
+            // although NULL can be a valid address
+            // but in C, NULL is regarded as allocation failure
+            // so just skip it
+            addr = PAGE_SIZE;
+        }
+
+        if flags.contains(MmapFlags::FIXED) {
+            // we have to map it to addr, so remove the old mapping first
+            self.vm().pop_with_split(addr, addr + len);
+        } else {
+            addr = self.vm().find_free_area(addr, len);
+        }
+
+        if flags.contains(MmapFlags::ANONYMOUS) {
+            if flags.contains(MmapFlags::SHARED) {
+                return Err(SysError::EINVAL);
+            }
+            self.vm().push(
+                addr,
+                addr + len,
+                prot.to_attr(),
+                Delay::new(GlobalFrameAlloc),
+                "mmap_anon",
+            );
+            return Ok(addr);
+        } else {
+            let inode = proc.get_file(fd)?.inode();
+            self.vm().push(
+                addr,
+                addr + len,
+                prot.to_attr(),
+                File {
+                    file: INodeForMap(inode),
+                    mem_start: addr,
+                    file_start: offset,
+                    file_end: offset + len,
+                    allocator: GlobalFrameAlloc,
+                },
+                "mmap_file",
+            );
+            return Ok(addr);
+        }
+    }
+
+    pub fn sys_mprotect(&mut self, addr: usize, len: usize, prot: usize) -> SysResult {
+        let prot = MmapProt::from_bits_truncate(prot);
+        info!(
+            "mprotect: addr={:#x}, size={:#x}, prot={:?}",
+            addr, len, prot
         );
-        let data = unsafe { slice::from_raw_parts_mut(addr as *mut u8, len) };
-        let file = proc.get_file(fd)?;
-        let read_len = file.read_at(offset, data)?;
-        if read_len != data.len() {
-            // use count() to consume the iterator
-            data[read_len..].iter_mut().map(|x| *x = 0).count();
+        let attr = prot.to_attr();
+
+        // FIXME: properly set the attribute of the area
+        //        now some mut ptr check is fault
+        let vm = self.vm();
+        let memory_area = vm
+            .iter()
+            .find(|area| area.is_overlap_with(addr, addr + len));
+        if memory_area.is_none() {
+            return Err(SysError::ENOMEM);
         }
-        return Ok(addr);
+        Ok(0)
+    }
+
+    pub fn sys_munmap(&mut self, addr: usize, len: usize) -> SysResult {
+        info!("munmap addr={:#x}, size={:#x}", addr, len);
+        self.vm().pop_with_split(addr, addr + len);
+        Ok(0)
     }
 }
-
-pub fn sys_mprotect(addr: usize, len: usize, prot: usize) -> SysResult {
-    let prot = MmapProt::from_bits_truncate(prot);
-    info!(
-        "mprotect: addr={:#x}, size={:#x}, prot={:?}",
-        addr, len, prot
-    );
-
-    let mut proc = process();
-    let attr = prot.to_attr();
-
-    // FIXME: properly set the attribute of the area
-    //        now some mut ptr check is fault
-    let memory_area = proc
-        .vm
-        .iter()
-        .find(|area| area.is_overlap_with(addr, addr + len));
-    if memory_area.is_none() {
-        return Err(SysError::ENOMEM);
-    }
-    proc.vm.edit(|pt| {
-        for page in Page::range_of(addr, addr + len) {
-            let entry = pt
-                .get_entry(page.start_address())
-                .expect("failed to get entry");
-            attr.apply(entry);
-        }
-    });
-    Ok(0)
-}
-
-pub fn sys_munmap(addr: usize, len: usize) -> SysResult {
-    info!("munmap addr={:#x}, size={:#x}", addr, len);
-    let mut proc = process();
-    proc.vm.pop_with_split(addr, addr + len);
-    Ok(0)
-}
-
 bitflags! {
     pub struct MmapProt: usize {
         /// Data cannot be accessed

kernel/src/syscall/misc.rs

@@ -4,121 +4,185 @@ use crate::consts::USER_STACK_SIZE;
 use core::mem::size_of;
 use core::sync::atomic::{AtomicI32, Ordering};
 
-pub fn sys_arch_prctl(code: i32, addr: usize, tf: &mut TrapFrame) -> SysResult {
-    const ARCH_SET_FS: i32 = 0x1002;
-    match code {
-        #[cfg(target_arch = "x86_64")]
-        ARCH_SET_FS => {
-            info!("sys_arch_prctl: set FS to {:#x}", addr);
-            tf.fsbase = addr;
-            Ok(0)
-        }
-        _ => Err(SysError::EINVAL),
-    }
-}
-
-pub fn sys_uname(buf: *mut u8) -> SysResult {
-    info!("sched_uname: buf: {:?}", buf);
-
-    let offset = 65;
-    let strings = ["rCore", "orz", "0.1.0", "1", "machine", "domain"];
-    let proc = process();
-    proc.vm.check_write_array(buf, strings.len() * offset)?;
-
-    for i in 0..strings.len() {
-        unsafe {
-            util::write_cstr(buf.add(i * offset), &strings[i]);
-        }
-    }
-    Ok(0)
-}
-
-pub fn sys_sched_getaffinity(pid: usize, size: usize, mask: *mut u32) -> SysResult {
-    info!(
-        "sched_getaffinity: pid: {}, size: {}, mask: {:?}",
-        pid, size, mask
-    );
-    let proc = process();
-    proc.vm.check_write_array(mask, size / size_of::<u32>())?;
-
-    // we only have 4 cpu at most.
-    // so just set it.
-    unsafe {
-        *mask = 0b1111;
-    }
-    Ok(0)
-}
-
-pub fn sys_sysinfo(sys_info: *mut SysInfo) -> SysResult {
-    let proc = process();
-    proc.vm.check_write_ptr(sys_info)?;
-
-    let sysinfo = SysInfo::default();
-    unsafe { *sys_info = sysinfo };
-    Ok(0)
-}
-
-pub fn sys_futex(uaddr: usize, op: u32, val: i32, timeout: *const TimeSpec) -> SysResult {
-    info!(
-        "futex: [{}] uaddr: {:#x}, op: {:#x}, val: {}, timeout_ptr: {:?}",
-        thread::current().id(),
-        uaddr,
-        op,
-        val,
-        timeout
-    );
-    //    if op & OP_PRIVATE == 0 {
-    //        unimplemented!("futex only support process-private");
-    //        return Err(SysError::ENOSYS);
-    //    }
-    if uaddr % size_of::<u32>() != 0 {
-        return Err(SysError::EINVAL);
-    }
-    process().vm.check_write_ptr(uaddr as *mut AtomicI32)?;
-    let atomic = unsafe { &mut *(uaddr as *mut AtomicI32) };
-    let _timeout = if timeout.is_null() {
-        None
-    } else {
-        process().vm.check_read_ptr(timeout)?;
-        Some(unsafe { *timeout })
-    };
-
-    const OP_WAIT: u32 = 0;
-    const OP_WAKE: u32 = 1;
-    const OP_PRIVATE: u32 = 128;
-
-    let queue = process().get_futex(uaddr);
-
-    match op & 0xf {
-        OP_WAIT => {
-            if atomic.load(Ordering::Acquire) != val {
-                return Err(SysError::EAGAIN);
+impl Syscall<'_> {
+    #[cfg(target_arch = "x86_64")]
+    pub fn sys_arch_prctl(&mut self, code: i32, addr: usize) -> SysResult {
+        const ARCH_SET_FS: i32 = 0x1002;
+        match code {
+            ARCH_SET_FS => {
+                info!("sys_arch_prctl: set FSBASE to {:#x}", addr);
+                self.tf.fsbase = addr;
+                Ok(0)
             }
-            // FIXME: support timeout
-            queue._wait();
-            Ok(0)
+            _ => Err(SysError::EINVAL),
         }
-        OP_WAKE => {
-            let woken_up_count = queue.notify_n(val as usize);
-            Ok(woken_up_count)
+    }
+
+    pub fn sys_uname(&mut self, buf: *mut u8) -> SysResult {
+        info!("uname: buf: {:?}", buf);
+
+        let offset = 65;
+        let strings = ["rCore", "orz", "0.1.0", "1", "machine", "domain"];
+        let buf = unsafe { self.vm().check_write_array(buf, strings.len() * offset)? };
+
+        for i in 0..strings.len() {
+            unsafe {
+                util::write_cstr(&mut buf[i * offset], &strings[i]);
+            }
         }
-        _ => {
-            warn!("unsupported futex operation: {}", op);
-            Err(SysError::ENOSYS)
+        Ok(0)
+    }
+
+    pub fn sys_sched_getaffinity(&mut self, pid: usize, size: usize, mask: *mut u32) -> SysResult {
+        info!(
+            "sched_getaffinity: pid: {}, size: {}, mask: {:?}",
+            pid, size, mask
+        );
+        let mask = unsafe { self.vm().check_write_array(mask, size / size_of::<u32>())? };
+
+        // we only have 4 cpu at most.
+        // so just set it.
+        mask[0] = 0b1111;
+        Ok(0)
+    }
+
+    pub fn sys_sysinfo(&mut self, sys_info: *mut SysInfo) -> SysResult {
+        let sys_info = unsafe { self.vm().check_write_ptr(sys_info)? };
+
+        let sysinfo = SysInfo::default();
+        *sys_info = sysinfo;
+        Ok(0)
+    }
+
+    pub fn sys_futex(
+        &mut self,
+        uaddr: usize,
+        op: u32,
+        val: i32,
+        timeout: *const TimeSpec,
+    ) -> SysResult {
+        info!(
+            "futex: [{}] uaddr: {:#x}, op: {:#x}, val: {}, timeout_ptr: {:?}",
+            thread::current().id(),
+            uaddr,
+            op,
+            val,
+            timeout
+        );
+        //    if op & OP_PRIVATE == 0 {
+        //        unimplemented!("futex only support process-private");
+        //        return Err(SysError::ENOSYS);
+        //    }
+        if uaddr % size_of::<u32>() != 0 {
+            return Err(SysError::EINVAL);
         }
+        let atomic = unsafe { self.vm().check_write_ptr(uaddr as *mut AtomicI32)? };
+        let _timeout = if timeout.is_null() {
+            None
+        } else {
+            Some(unsafe { *self.vm().check_read_ptr(timeout)? })
+        };
+
+        const OP_WAIT: u32 = 0;
+        const OP_WAKE: u32 = 1;
+        const OP_PRIVATE: u32 = 128;
+
+        let mut proc = self.process();
+        let queue = proc.get_futex(uaddr);
+
+        match op & 0xf {
+            OP_WAIT => {
+                if atomic.load(Ordering::Acquire) != val {
+                    return Err(SysError::EAGAIN);
+                }
+                // FIXME: support timeout
+                queue.wait(proc);
+                Ok(0)
+            }
+            OP_WAKE => {
+                let woken_up_count = queue.notify_n(val as usize);
+                Ok(woken_up_count)
+            }
+            _ => {
+                warn!("unsupported futex operation: {}", op);
+                Err(SysError::ENOSYS)
+            }
+        }
+    }
+
+    pub fn sys_reboot(&mut self, _magic: u32, _magic2: u32, cmd: u32, _arg: *const u8) -> SysResult {
+        // we will skip verifying magic
+        if cmd == LINUX_REBOOT_CMD_HALT {
+            unsafe {
+                cpu::exit_in_qemu(1);
+            }
+        }
+        Ok(0)
+    }
+
+    pub fn sys_prlimit64(
+        &mut self,
+        pid: usize,
+        resource: usize,
+        new_limit: *const RLimit,
+        old_limit: *mut RLimit,
+    ) -> SysResult {
+        info!(
+            "prlimit64: pid: {}, resource: {}, new_limit: {:x?}, old_limit: {:x?}",
+            pid, resource, new_limit, old_limit
+        );
+        match resource {
+            RLIMIT_STACK => {
+                if !old_limit.is_null() {
+                    let old_limit = unsafe { self.vm().check_write_ptr(old_limit)? };
+                    *old_limit = RLimit {
+                        cur: USER_STACK_SIZE as u64,
+                        max: USER_STACK_SIZE as u64,
+                    };
+                }
+                Ok(0)
+            }
+            RLIMIT_NOFILE => {
+                if !old_limit.is_null() {
+                    let old_limit = unsafe { self.vm().check_write_ptr(old_limit)? };
+                    *old_limit = RLimit {
+                        cur: 1024,
+                        max: 1024,
+                    };
+                }
+                Ok(0)
+            }
+            RLIMIT_RSS | RLIMIT_AS => {
+                if !old_limit.is_null() {
+                    let old_limit = unsafe { self.vm().check_write_ptr(old_limit)? };
+                    // 1GB
+                    *old_limit = RLimit {
+                        cur: 1024 * 1024 * 1024,
+                        max: 1024 * 1024 * 1024,
+                    };
+                }
+                Ok(0)
+            }
+            _ => Err(SysError::ENOSYS),
+        }
+    }
+
+    pub fn sys_getrandom(&mut self, buf: *mut u8, len: usize, _flag: u32) -> SysResult {
+        //info!("getrandom: buf: {:?}, len: {:?}, falg {:?}", buf, len,flag);
+        let slice = unsafe { self.vm().check_write_array(buf, len)? };
+        let mut i = 0;
+        for elm in slice {
+            unsafe {
+                *elm = i + crate::trap::TICK as u8;
+            }
+            i += 1;
+        }
+
+        Ok(len)
     }
 }
 
 const LINUX_REBOOT_CMD_HALT: u32 = 0xcdef0123;
-pub fn sys_reboot(_magic: u32, magic2: u32, cmd: u32, _arg: *const u8) -> SysResult {
-    // we will skip verifying magic
-    if cmd == LINUX_REBOOT_CMD_HALT {
-        unsafe {
-            cpu::exit_in_qemu(1);
-        }
-    }
-    Ok(0)
-}
 
 #[repr(C)]
 #[derive(Debug, Default)]
@@ -142,59 +206,6 @@ const RLIMIT_RSS: usize = 5;
 const RLIMIT_NOFILE: usize = 7;
 const RLIMIT_AS: usize = 9;
 
-pub fn sys_prlimit64(
-    pid: usize,
-    resource: usize,
-    new_limit: *const RLimit,
-    old_limit: *mut RLimit,
-) -> SysResult {
-    let proc = process();
-    info!(
-        "prlimit64: pid: {}, resource: {}, new_limit: {:x?}, old_limit: {:x?}",
-        pid, resource, new_limit, old_limit
-    );
-    match resource {
-        RLIMIT_STACK => {
-            if !old_limit.is_null() {
-                proc.vm.check_write_ptr(old_limit)?;
-                unsafe {
-                    *old_limit = RLimit {
-                        cur: USER_STACK_SIZE as u64,
-                        max: USER_STACK_SIZE as u64,
-                    };
-                }
-            }
-            Ok(0)
-        }
-        RLIMIT_NOFILE => {
-            if !old_limit.is_null() {
-                proc.vm.check_write_ptr(old_limit)?;
-                unsafe {
-                    *old_limit = RLimit {
-                        cur: 1024,
-                        max: 1024,
-                    };
-                }
-            }
-            Ok(0)
-        }
-        RLIMIT_RSS | RLIMIT_AS => {
-            if !old_limit.is_null() {
-                proc.vm.check_write_ptr(old_limit)?;
-                unsafe {
-                    // 1GB
-                    *old_limit = RLimit {
-                        cur: 1024 * 1024 * 1024,
-                        max: 1024 * 1024 * 1024,
-                    };
-                }
-            }
-            Ok(0)
-        }
-        _ => Err(SysError::ENOSYS),
-    }
-}
-
 #[repr(C)]
 #[derive(Debug, Default)]
 pub struct RLimit {

kernel/src/syscall/mod.rs

@@ -10,8 +10,9 @@ use rcore_memory::VMError;
 use crate::arch::cpu;
 use crate::arch::interrupt::TrapFrame;
 use crate::arch::syscall::*;
+use crate::memory::MemorySet;
 use crate::process::*;
-use crate::sync::Condvar;
+use crate::sync::{Condvar, MutexGuard, SpinNoIrq};
 use crate::thread;
 use crate::util;
 
@@ -31,428 +32,382 @@ mod net;
 mod proc;
 mod time;
 
+use alloc::collections::BTreeMap;
+use spin::Mutex;
+
+#[cfg(feature = "profile")]
+lazy_static! {
+    static ref SYSCALL_TIMING: Mutex<BTreeMap<usize, i64>> = Mutex::new(BTreeMap::new());
+}
+
 /// System call dispatcher
-// This #[deny(unreachable_patterns)] checks if each match arm is defined
-// See discussion in https://github.com/oscourse-tsinghua/rcore_plus/commit/17e644e54e494835f1a49b34b80c2c4f15ed0dbe.
-#[deny(unreachable_patterns)]
 pub fn syscall(id: usize, args: [usize; 6], tf: &mut TrapFrame) -> isize {
-    let cid = cpu::id();
-    let pid = { process().pid.clone() };
-    let tid = processor().tid();
-    if !pid.is_init() {
-        // we trust pid 0 process
-        debug!("{}:{}:{} syscall id {} begin", cid, pid, tid, id);
-    }
-
-    // use platform-specific syscal numbers
-    // See https://filippo.io/linux-syscall-table/
-    // And https://fedora.juszkiewicz.com.pl/syscalls.html.
-    let ret = match id {
-        // 0
-        SYS_READ => sys_read(args[0], args[1] as *mut u8, args[2]),
-        SYS_WRITE => sys_write(args[0], args[1] as *const u8, args[2]),
-        SYS_CLOSE => sys_close(args[0]),
-        SYS_FSTAT => sys_fstat(args[0], args[1] as *mut Stat),
-        SYS_LSEEK => sys_lseek(args[0], args[1] as i64, args[2] as u8),
-        SYS_MMAP => sys_mmap(args[0], args[1], args[2], args[3], args[4], args[5]),
-        // 10
-        SYS_MPROTECT => sys_mprotect(args[0], args[1], args[2]),
-        SYS_MUNMAP => sys_munmap(args[0], args[1]),
-        SYS_BRK => {
-            warn!("sys_brk is unimplemented, return -1");
-            Err(SysError::ENOMEM)
-        }
-        SYS_RT_SIGACTION => {
-            warn!("sys_sigaction is unimplemented");
-            Ok(0)
-        }
-        SYS_RT_SIGPROCMASK => {
-            warn!("sys_sigprocmask is unimplemented");
-            Ok(0)
-        }
-        SYS_IOCTL => sys_ioctl(args[0], args[1], args[2], args[3], args[4]),
-        SYS_PREAD64 => sys_pread(args[0], args[1] as *mut u8, args[2], args[3]),
-        SYS_PWRITE64 => sys_pwrite(args[0], args[1] as *const u8, args[2], args[3]),
-        SYS_READV => sys_readv(args[0], args[1] as *const IoVec, args[2]),
-        // 20
-        SYS_WRITEV => sys_writev(args[0], args[1] as *const IoVec, args[2]),
-        SYS_SCHED_YIELD => sys_yield(),
-        SYS_MADVISE => {
-            warn!("sys_madvise is unimplemented");
-            Ok(0)
-        }
-        SYS_NANOSLEEP => sys_nanosleep(args[0] as *const TimeSpec),
-        SYS_SETITIMER => {
-            warn!("sys_setitimer is unimplemented");
-            Ok(0)
-        }
-        SYS_GETPID => sys_getpid(),
-        // 40
-        SYS_SENDFILE => sys_sendfile(args[0], args[1], args[2] as *mut usize, args[3]),
-        SYS_SOCKET => sys_socket(args[0], args[1], args[2]),
-        SYS_CONNECT => sys_connect(args[0], args[1] as *const SockAddr, args[2]),
-        SYS_ACCEPT => sys_accept(args[0], args[1] as *mut SockAddr, args[2] as *mut u32),
-        SYS_SENDTO => sys_sendto(
-            args[0],
-            args[1] as *const u8,
-            args[2],
-            args[3],
-            args[4] as *const SockAddr,
-            args[5],
-        ),
-        SYS_RECVFROM => sys_recvfrom(
-            args[0],
-            args[1] as *mut u8,
-            args[2],
-            args[3],
-            args[4] as *mut SockAddr,
-            args[5] as *mut u32,
-        ),
-        //        SYS_SENDMSG => sys_sendmsg(),
-        SYS_RECVMSG => sys_recvmsg(args[0], args[1] as *mut MsgHdr, args[2]),
-        SYS_SHUTDOWN => sys_shutdown(args[0], args[1]),
-        SYS_BIND => sys_bind(args[0], args[1] as *const SockAddr, args[2]),
-        // 50
-        SYS_LISTEN => sys_listen(args[0], args[1]),
-        SYS_GETSOCKNAME => sys_getsockname(args[0], args[1] as *mut SockAddr, args[2] as *mut u32),
-        SYS_GETPEERNAME => sys_getpeername(args[0], args[1] as *mut SockAddr, args[2] as *mut u32),
-        SYS_SETSOCKOPT => sys_setsockopt(args[0], args[1], args[2], args[3] as *const u8, args[4]),
-        SYS_GETSOCKOPT => sys_getsockopt(
-            args[0],
-            args[1],
-            args[2],
-            args[3] as *mut u8,
-            args[4] as *mut u32,
-        ),
-        SYS_CLONE => sys_clone(
-            args[0],
-            args[1],
-            args[2] as *mut u32,
-            args[3] as *mut u32,
-            args[4],
-            tf,
-        ),
-        SYS_EXECVE => sys_exec(
-            args[0] as *const u8,
-            args[1] as *const *const u8,
-            args[2] as *const *const u8,
-            tf,
-        ),
-        // 60
-        SYS_EXIT => sys_exit(args[0] as usize),
-        SYS_WAIT4 => sys_wait4(args[0] as isize, args[1] as *mut i32), // TODO: wait4
-        SYS_KILL => sys_kill(args[0], args[1]),
-        SYS_UNAME => sys_uname(args[0] as *mut u8),
-        SYS_FCNTL => {
-            warn!("sys_fcntl is unimplemented");
-            Ok(0)
-        }
-        SYS_FLOCK => {
-            warn!("sys_flock is unimplemented");
-            Ok(0)
-        }
-        SYS_FSYNC => sys_fsync(args[0]),
-        SYS_FDATASYNC => sys_fdatasync(args[0]),
-        SYS_TRUNCATE => sys_truncate(args[0] as *const u8, args[1]),
-        SYS_FTRUNCATE => sys_ftruncate(args[0], args[1]),
-        SYS_GETCWD => sys_getcwd(args[0] as *mut u8, args[1]),
-        // 80
-        SYS_CHDIR => sys_chdir(args[0] as *const u8),
-        SYS_FCHMOD => {
-            warn!("sys_fchmod is unimplemented");
-            Ok(0)
-        }
-        SYS_FCHOWN => {
-            warn!("sys_fchown is unimplemented");
-            Ok(0)
-        }
-        SYS_UMASK => {
-            warn!("sys_umask is unimplemented");
-            Ok(0o777)
-        }
-        SYS_GETTIMEOFDAY => sys_gettimeofday(args[0] as *mut TimeVal, args[1] as *const u8),
-        //        SYS_GETRLIMIT => sys_getrlimit(),
-        SYS_GETRUSAGE => sys_getrusage(args[0], args[1] as *mut RUsage),
-        SYS_SYSINFO => sys_sysinfo(args[0] as *mut SysInfo),
-        SYS_TIMES => sys_times(args[0] as *mut Tms),
-        SYS_GETUID => {
-            warn!("sys_getuid is unimplemented");
-            Ok(0)
-        }
-        SYS_GETGID => {
-            warn!("sys_getgid is unimplemented");
-            Ok(0)
-        }
-        SYS_SETUID => {
-            warn!("sys_setuid is unimplemented");
-            Ok(0)
-        }
-        SYS_GETEUID => {
-            warn!("sys_geteuid is unimplemented");
-            Ok(0)
-        }
-        SYS_GETEGID => {
-            warn!("sys_getegid is unimplemented");
-            Ok(0)
-        }
-        SYS_SETPGID => {
-            warn!("sys_setpgid is unimplemented");
-            Ok(0)
-        }
-        // 110
-        SYS_GETPPID => sys_getppid(),
-        SYS_SETSID => {
-            warn!("sys_setsid is unimplemented");
-            Ok(0)
-        }
-        SYS_GETPGID => {
-            warn!("sys_getpgid is unimplemented");
-            Ok(0)
-        }
-        SYS_GETGROUPS => {
-            warn!("sys_getgroups is unimplemented");
-            Ok(0)
-        }
-        SYS_SETGROUPS => {
-            warn!("sys_setgroups is unimplemented");
-            Ok(0)
-        }
-        SYS_SIGALTSTACK => {
-            warn!("sys_sigaltstack is unimplemented");
-            Ok(0)
-        }
-        SYS_STATFS => {
-            warn!("statfs is unimplemented");
-            Err(SysError::EACCES)
-        }
-        SYS_FSTATFS => {
-            warn!("fstatfs is unimplemented");
-            Err(SysError::EACCES)
-        }
-        SYS_SETPRIORITY => sys_set_priority(args[0]),
-        SYS_PRCTL => {
-            warn!("prctl is unimplemented");
-            Ok(0)
-        }
-        //        SYS_SETRLIMIT => sys_setrlimit(),
-        SYS_SYNC => sys_sync(),
-        SYS_MOUNT => {
-            warn!("mount is unimplemented");
-            Err(SysError::EACCES)
-        }
-        SYS_UMOUNT2 => {
-            warn!("umount2 is unimplemented");
-            Err(SysError::EACCES)
-        }
-        SYS_REBOOT => sys_reboot(
-            args[0] as u32,
-            args[1] as u32,
-            args[2] as u32,
-            args[3] as *const u8,
-        ),
-        SYS_GETTID => sys_gettid(),
-        SYS_FUTEX => sys_futex(
-            args[0],
-            args[1] as u32,
-            args[2] as i32,
-            args[3] as *const TimeSpec,
-        ),
-        SYS_SCHED_GETAFFINITY => sys_sched_getaffinity(args[0], args[1], args[2] as *mut u32),
-        SYS_GETDENTS64 => sys_getdents64(args[0], args[1] as *mut LinuxDirent64, args[2]),
-        SYS_SET_TID_ADDRESS => {
-            warn!("sys_set_tid_address is unimplemented");
-            Ok(thread::current().id())
-        }
-        SYS_CLOCK_GETTIME => sys_clock_gettime(args[0], args[1] as *mut TimeSpec),
-        SYS_EXIT_GROUP => sys_exit_group(args[0]),
-        SYS_OPENAT => sys_openat(args[0], args[1] as *const u8, args[2], args[3]),
-        SYS_MKDIRAT => sys_mkdirat(args[0], args[1] as *const u8, args[2]),
-        //        SYS_MKNODAT => sys_mknod(),
-        // 260
-        SYS_FCHOWNAT => {
-            warn!("sys_fchownat is unimplemented");
-            Ok(0)
-        }
-        SYS_NEWFSTATAT => sys_fstatat(args[0], args[1] as *const u8, args[2] as *mut Stat, args[3]),
-        SYS_UNLINKAT => sys_unlinkat(args[0], args[1] as *const u8, args[2]),
-        SYS_RENAMEAT => sys_renameat(args[0], args[1] as *const u8, args[2], args[3] as *const u8),
-        SYS_LINKAT => sys_linkat(
-            args[0],
-            args[1] as *const u8,
-            args[2],
-            args[3] as *const u8,
-            args[4],
-        ),
-        SYS_SYMLINKAT => Err(SysError::EACCES),
-        SYS_READLINKAT => {
-            sys_readlinkat(args[0], args[1] as *const u8, args[2] as *mut u8, args[3])
-        }
-        SYS_FCHMODAT => {
-            warn!("sys_fchmodat is unimplemented");
-            Ok(0)
-        }
-        SYS_FACCESSAT => sys_faccessat(args[0], args[1] as *const u8, args[2], args[3]),
-        SYS_PPOLL => sys_ppoll(args[0] as *mut PollFd, args[1], args[2] as *const TimeSpec), // ignore sigmask
-        // 280
-        SYS_UTIMENSAT => {
-            warn!("sys_utimensat is unimplemented");
-            Ok(0)
-        }
-        SYS_ACCEPT4 => sys_accept(args[0], args[1] as *mut SockAddr, args[2] as *mut u32), // use accept for accept4
-        SYS_EPOLL_CREATE1 => {
-            warn!("sys_epoll_create1 is unimplemented");
-            Err(SysError::ENOSYS)
-        }
-        SYS_DUP3 => sys_dup2(args[0], args[1]), // TODO: handle `flags`
-        SYS_PIPE2 => sys_pipe(args[0] as *mut u32), // TODO: handle `flags`
-        SYS_PRLIMIT64 => sys_prlimit64(
-            args[0],
-            args[1],
-            args[2] as *const RLimit,
-            args[3] as *mut RLimit,
-        ),
-        // custom temporary syscall
-        SYS_MAP_PCI_DEVICE => sys_map_pci_device(args[0], args[1]),
-        SYS_GET_PADDR => sys_get_paddr(args[0] as *const u64, args[1] as *mut u64, args[2]),
-
-        _ => {
-            #[cfg(target_arch = "x86_64")]
-            let x86_64_ret = x86_64_syscall(id, args, tf);
-            #[cfg(not(target_arch = "x86_64"))]
-            let x86_64_ret = None;
-
-            #[cfg(target_arch = "mips")]
-            let mips_ret = mips_syscall(id, args, tf);
-            #[cfg(not(target_arch = "mips"))]
-            let mips_ret = None;
-            if let Some(ret) = x86_64_ret {
-                ret
-            } else if let Some(ret) = mips_ret {
-                ret
-            } else {
-                error!("unknown syscall id: {}, args: {:x?}", id, args);
-                crate::trap::error(tf);
-            }
-        }
-    };
-    if !pid.is_init() {
-        // we trust pid 0 process
-        debug!(
-            "{}:{}:{} syscall id {} ret with {:x?}",
-            cid, pid, tid, id, ret
-        );
-    }
-    match ret {
-        Ok(code) => code as isize,
-        Err(err) => -(err as isize),
-    }
+    let thread = unsafe { current_thread() };
+    let mut syscall = Syscall { thread, tf };
+    syscall.syscall(id, args)
 }
 
-#[cfg(target_arch = "mips")]
-fn mips_syscall(id: usize, args: [usize; 6], tf: &mut TrapFrame) -> Option<SysResult> {
-    let ret = match id {
-        SYS_OPEN => sys_open(args[0] as *const u8, args[1], args[2]),
-        SYS_POLL => sys_poll(args[0] as *mut PollFd, args[1], args[2]),
-        SYS_DUP2 => sys_dup2(args[0], args[1]),
-        SYS_FORK => sys_fork(tf),
-        SYS_MMAP2 => sys_mmap(args[0], args[1], args[2], args[3], args[4], args[5] * 4096),
-        SYS_FSTAT64 => sys_fstat(args[0], args[1] as *mut Stat),
-        SYS_LSTAT64 => sys_lstat(args[0] as *const u8, args[1] as *mut Stat),
-        SYS_STAT64 => sys_stat(args[0] as *const u8, args[1] as *mut Stat),
-        SYS_PIPE => {
-            let fd_ptr = args[0] as *mut u32;
-            match sys_pipe(fd_ptr) {
-                Ok(code) => {
-                    unsafe {
-                        tf.v0 = *fd_ptr as usize;
-                        tf.v1 = *(fd_ptr.add(1)) as usize;
-                    }
-                    Ok(tf.v0)
+/// All context needed for syscall
+struct Syscall<'a> {
+    thread: &'a mut Thread,
+    tf: &'a mut TrapFrame,
+}
+
+impl Syscall<'_> {
+    /// Get current process
+    pub fn process(&self) -> MutexGuard<'_, Process, SpinNoIrq> {
+        self.thread.proc.lock()
+    }
+
+    /// Get current virtual memory
+    pub fn vm(&self) -> MutexGuard<'_, MemorySet, SpinNoIrq> {
+        self.thread.vm.lock()
+    }
+
+    /// System call dispatcher
+    // This #[deny(unreachable_patterns)] checks if each match arm is defined
+    // See discussion in https://github.com/oscourse-tsinghua/rcore_plus/commit/17e644e54e494835f1a49b34b80c2c4f15ed0dbe.
+    #[deny(unreachable_patterns)]
+    fn syscall(&mut self, id: usize, args: [usize; 6]) -> isize {
+        #[cfg(feature = "profile")]
+        let begin_time = unsafe { core::arch::x86_64::_rdtsc() };
+        let cid = cpu::id();
+        let pid = self.process().pid.clone();
+        let tid = processor().tid();
+        if !pid.is_init() {
+            // we trust pid 0 process
+            debug!("{}:{}:{} syscall id {} begin", cid, pid, tid, id);
+        }
+
+        // use platform-specific syscal numbers
+        // See https://filippo.io/linux-syscall-table/
+        // And https://fedora.juszkiewicz.com.pl/syscalls.html.
+        let ret = match id {
+            // file
+            SYS_READ => self.sys_read(args[0], args[1] as *mut u8, args[2]),
+            SYS_WRITE => self.sys_write(args[0], args[1] as *const u8, args[2]),
+            SYS_OPENAT => self.sys_openat(args[0], args[1] as *const u8, args[2], args[3]),
+            SYS_CLOSE => self.sys_close(args[0]),
+            SYS_FSTAT => self.sys_fstat(args[0], args[1] as *mut Stat),
+            SYS_NEWFSTATAT => {
+                self.sys_fstatat(args[0], args[1] as *const u8, args[2] as *mut Stat, args[3])
+            }
+            SYS_LSEEK => self.sys_lseek(args[0], args[1] as i64, args[2] as u8),
+            SYS_IOCTL => self.sys_ioctl(args[0], args[1], args[2], args[3], args[4]),
+            SYS_PREAD64 => self.sys_pread(args[0], args[1] as *mut u8, args[2], args[3]),
+            SYS_PWRITE64 => self.sys_pwrite(args[0], args[1] as *const u8, args[2], args[3]),
+            SYS_READV => self.sys_readv(args[0], args[1] as *const IoVec, args[2]),
+            SYS_WRITEV => self.sys_writev(args[0], args[1] as *const IoVec, args[2]),
+            SYS_SENDFILE => self.sys_sendfile(args[0], args[1], args[2] as *mut usize, args[3]),
+            SYS_FCNTL => self.unimplemented("fcntl", Ok(0)),
+            SYS_FLOCK => self.unimplemented("flock", Ok(0)),
+            SYS_FSYNC => self.sys_fsync(args[0]),
+            SYS_FDATASYNC => self.sys_fdatasync(args[0]),
+            SYS_TRUNCATE => self.sys_truncate(args[0] as *const u8, args[1]),
+            SYS_FTRUNCATE => self.sys_ftruncate(args[0], args[1]),
+            SYS_GETDENTS64 => self.sys_getdents64(args[0], args[1] as *mut LinuxDirent64, args[2]),
+            SYS_GETCWD => self.sys_getcwd(args[0] as *mut u8, args[1]),
+            SYS_CHDIR => self.sys_chdir(args[0] as *const u8),
+            SYS_RENAMEAT => {
+                self.sys_renameat(args[0], args[1] as *const u8, args[2], args[3] as *const u8)
+            }
+            SYS_MKDIRAT => self.sys_mkdirat(args[0], args[1] as *const u8, args[2]),
+            SYS_LINKAT => self.sys_linkat(
+                args[0],
+                args[1] as *const u8,
+                args[2],
+                args[3] as *const u8,
+                args[4],
+            ),
+            SYS_UNLINKAT => self.sys_unlinkat(args[0], args[1] as *const u8, args[2]),
+            SYS_SYMLINKAT => self.unimplemented("symlinkat", Err(SysError::EACCES)),
+            SYS_READLINKAT => {
+                self.sys_readlinkat(args[0], args[1] as *const u8, args[2] as *mut u8, args[3])
+            }
+            SYS_FCHMOD => self.unimplemented("fchmod", Ok(0)),
+            SYS_FCHMODAT => self.unimplemented("fchmodat", Ok(0)),
+            SYS_FCHOWN => self.unimplemented("fchown", Ok(0)),
+            SYS_FCHOWNAT => self.unimplemented("fchownat", Ok(0)),
+            SYS_FACCESSAT => self.sys_faccessat(args[0], args[1] as *const u8, args[2], args[3]),
+            SYS_DUP3 => self.sys_dup2(args[0], args[1]), // TODO: handle `flags`
+            SYS_PIPE2 => self.sys_pipe(args[0] as *mut u32), // TODO: handle `flags`
+            SYS_UTIMENSAT => self.unimplemented("utimensat", Ok(0)),
+
+            // io multiplexing
+            SYS_PPOLL => {
+                self.sys_ppoll(args[0] as *mut PollFd, args[1], args[2] as *const TimeSpec)
+            } // ignore sigmask
+            SYS_EPOLL_CREATE1 => self.unimplemented("epoll_create1", Err(SysError::ENOSYS)),
+
+            // file system
+            SYS_STATFS => self.unimplemented("statfs", Err(SysError::EACCES)),
+            SYS_FSTATFS => self.unimplemented("fstatfs", Err(SysError::EACCES)),
+            SYS_SYNC => self.sys_sync(),
+            SYS_MOUNT => self.unimplemented("mount", Err(SysError::EACCES)),
+            SYS_UMOUNT2 => self.unimplemented("umount2", Err(SysError::EACCES)),
+
+            // memory
+            SYS_BRK => self.unimplemented("brk", Err(SysError::ENOMEM)),
+            SYS_MMAP => self.sys_mmap(args[0], args[1], args[2], args[3], args[4], args[5]),
+            SYS_MPROTECT => self.sys_mprotect(args[0], args[1], args[2]),
+            SYS_MUNMAP => self.sys_munmap(args[0], args[1]),
+            SYS_MADVISE => self.unimplemented("madvise", Ok(0)),
+
+            // signal
+            SYS_RT_SIGACTION => self.unimplemented("sigaction", Ok(0)),
+            SYS_RT_SIGPROCMASK => self.unimplemented("sigprocmask", Ok(0)),
+            SYS_SIGALTSTACK => self.unimplemented("sigaltstack", Ok(0)),
+            SYS_KILL => self.sys_kill(args[0], args[1]),
+
+            // schedule
+            SYS_SCHED_YIELD => self.sys_yield(),
+            SYS_SCHED_GETAFFINITY => {
+                self.sys_sched_getaffinity(args[0], args[1], args[2] as *mut u32)
+            }
+
+            // socket
+            SYS_SOCKET => self.sys_socket(args[0], args[1], args[2]),
+            SYS_CONNECT => self.sys_connect(args[0], args[1] as *const SockAddr, args[2]),
+            SYS_ACCEPT => self.sys_accept(args[0], args[1] as *mut SockAddr, args[2] as *mut u32),
+            SYS_ACCEPT4 => self.sys_accept(args[0], args[1] as *mut SockAddr, args[2] as *mut u32), // use accept for accept4
+            SYS_SENDTO => self.sys_sendto(
+                args[0],
+                args[1] as *const u8,
+                args[2],
+                args[3],
+                args[4] as *const SockAddr,
+                args[5],
+            ),
+            SYS_RECVFROM => self.sys_recvfrom(
+                args[0],
+                args[1] as *mut u8,
+                args[2],
+                args[3],
+                args[4] as *mut SockAddr,
+                args[5] as *mut u32,
+            ),
+            //        SYS_SENDMSG => self.sys_sendmsg(),
+            SYS_RECVMSG => self.sys_recvmsg(args[0], args[1] as *mut MsgHdr, args[2]),
+            SYS_SHUTDOWN => self.sys_shutdown(args[0], args[1]),
+            SYS_BIND => self.sys_bind(args[0], args[1] as *const SockAddr, args[2]),
+            SYS_LISTEN => self.sys_listen(args[0], args[1]),
+            SYS_GETSOCKNAME => {
+                self.sys_getsockname(args[0], args[1] as *mut SockAddr, args[2] as *mut u32)
+            }
+            SYS_GETPEERNAME => {
+                self.sys_getpeername(args[0], args[1] as *mut SockAddr, args[2] as *mut u32)
+            }
+            SYS_SETSOCKOPT => {
+                self.sys_setsockopt(args[0], args[1], args[2], args[3] as *const u8, args[4])
+            }
+            SYS_GETSOCKOPT => self.sys_getsockopt(
+                args[0],
+                args[1],
+                args[2],
+                args[3] as *mut u8,
+                args[4] as *mut u32,
+            ),
+
+            // process
+            SYS_CLONE => self.sys_clone(
+                args[0],
+                args[1],
+                args[2] as *mut u32,
+                args[3] as *mut u32,
+                args[4],
+            ),
+            SYS_EXECVE => self.sys_exec(
+                args[0] as *const u8,
+                args[1] as *const *const u8,
+                args[2] as *const *const u8,
+            ),
+            SYS_EXIT => self.sys_exit(args[0] as usize),
+            SYS_EXIT_GROUP => self.sys_exit_group(args[0]),
+            SYS_WAIT4 => self.sys_wait4(args[0] as isize, args[1] as *mut i32), // TODO: wait4
+            SYS_SET_TID_ADDRESS => self.sys_set_tid_address(args[0] as *mut u32),
+            SYS_FUTEX => self.sys_futex(
+                args[0],
+                args[1] as u32,
+                args[2] as i32,
+                args[3] as *const TimeSpec,
+            ),
+
+            // time
+            SYS_NANOSLEEP => self.sys_nanosleep(args[0] as *const TimeSpec),
+            SYS_SETITIMER => self.unimplemented("setitimer", Ok(0)),
+            SYS_GETTIMEOFDAY => {
+                self.sys_gettimeofday(args[0] as *mut TimeVal, args[1] as *const u8)
+            }
+            SYS_CLOCK_GETTIME => self.sys_clock_gettime(args[0], args[1] as *mut TimeSpec),
+
+            // system
+            SYS_GETPID => self.sys_getpid(),
+            SYS_GETTID => self.sys_gettid(),
+            SYS_UNAME => self.sys_uname(args[0] as *mut u8),
+            SYS_UMASK => self.unimplemented("umask", Ok(0o777)),
+            //        SYS_GETRLIMIT => self.sys_getrlimit(),
+            //        SYS_SETRLIMIT => self.sys_setrlimit(),
+            SYS_GETRUSAGE => self.sys_getrusage(args[0], args[1] as *mut RUsage),
+            SYS_SYSINFO => self.sys_sysinfo(args[0] as *mut SysInfo),
+            SYS_TIMES => self.sys_times(args[0] as *mut Tms),
+            SYS_GETUID => self.unimplemented("getuid", Ok(0)),
+            SYS_GETGID => self.unimplemented("getgid", Ok(0)),
+            SYS_SETUID => self.unimplemented("setuid", Ok(0)),
+            SYS_GETEUID => self.unimplemented("geteuid", Ok(0)),
+            SYS_GETEGID => self.unimplemented("getegid", Ok(0)),
+            SYS_SETPGID => self.unimplemented("setpgid", Ok(0)),
+            SYS_GETPPID => self.sys_getppid(),
+            SYS_SETSID => self.unimplemented("setsid", Ok(0)),
+            SYS_GETPGID => self.unimplemented("getpgid", Ok(0)),
+            SYS_GETGROUPS => self.unimplemented("getgroups", Ok(0)),
+            SYS_SETGROUPS => self.unimplemented("setgroups", Ok(0)),
+            SYS_SETPRIORITY => self.sys_set_priority(args[0]),
+            SYS_PRCTL => self.unimplemented("prctl", Ok(0)),
+            SYS_PRLIMIT64 => self.sys_prlimit64(
+                args[0],
+                args[1],
+                args[2] as *const RLimit,
+                args[3] as *mut RLimit,
+            ),
+            SYS_REBOOT => self.sys_reboot(
+                args[0] as u32,
+                args[1] as u32,
+                args[2] as u32,
+                args[3] as *const u8,
+            ),
+
+            // custom
+            SYS_MAP_PCI_DEVICE => self.sys_map_pci_device(args[0], args[1]),
+            SYS_GET_PADDR => {
+                self.sys_get_paddr(args[0] as *const u64, args[1] as *mut u64, args[2])
+            }
+            //SYS_GETRANDOM => self.unimplemented("getrandom", Err(SysError::EINVAL)),
+            SYS_GETRANDOM => {
+                self.sys_getrandom(args[0] as *mut u8, args[1] as usize, args[2] as u32)
+            }
+            SYS_TKILL => self.unimplemented("tkill", Ok(0)),
+            _ => {
+                let ret = match () {
+                    #[cfg(target_arch = "x86_64")]
+                    () => self.x86_64_syscall(id, args),
+                    #[cfg(target_arch = "mips")]
+                    () => self.mips_syscall(id, args),
+                    #[cfg(all(not(target_arch = "x86_64"), not(target_arch = "mips")))]
+                    () => None,
+                };
+                if let Some(ret) = ret {
+                    ret
+                } else {
+                    error!("unknown syscall id: {}, args: {:x?}", id, args);
+                    crate::trap::error(self.tf);
+                }
+            }
+        };
+        if !pid.is_init() {
+            // we trust pid 0 process
+            info!("=> {:x?}", ret);
+        }
+        #[cfg(feature = "profile")]
+        {
+            let end_time = unsafe { core::arch::x86_64::_rdtsc() };
+            *SYSCALL_TIMING.lock().entry(id).or_insert(0) += end_time - begin_time;
+            if end_time % 1000 == 0 {
+                let timing = SYSCALL_TIMING.lock();
+                let mut count_vec: Vec<(&usize, &i64)> = timing.iter().collect();
+                count_vec.sort_by(|a, b| b.1.cmp(a.1));
+                for (id, time) in count_vec.iter().take(5) {
+                    warn!("timing {:03} time {:012}", id, time);
                 }
-                Err(err) => Err(err),
             }
         }
-        SYS_GETPGID => {
-            warn!("sys_getpgid is unimplemented");
-            Ok(0)
+        match ret {
+            Ok(code) => code as isize,
+            Err(err) => -(err as isize),
         }
-        SYS_SETPGID => {
-            warn!("sys_setpgid is unimplemented");
-            Ok(0)
-        }
-        SYS_FCNTL64 => {
-            warn!("sys_fcntl64 is unimplemented");
-            Ok(0)
-        }
-        SYS_SET_THREAD_AREA => {
-            info!("set_thread_area: tls: 0x{:x}", args[0]);
-            extern "C" {
-                fn _cur_tls();
-            }
+    }
 
-            unsafe {
-                asm!("mtc0 $0, $$4, 2": :"r"(args[0]));
-                *(_cur_tls as *mut usize) = args[0];
-            }
-            Ok(0)
-        }
-        _ => {
-            return None;
-        }
-    };
-    Some(ret)
-}
+    fn unimplemented(&self, name: &str, ret: SysResult) -> SysResult {
+        warn!("{} is unimplemented", name);
+        ret
+    }
 
-#[cfg(target_arch = "x86_64")]
-fn x86_64_syscall(id: usize, args: [usize; 6], tf: &mut TrapFrame) -> Option<SysResult> {
-    let ret = match id {
-        SYS_OPEN => sys_open(args[0] as *const u8, args[1], args[2]),
-        SYS_STAT => sys_stat(args[0] as *const u8, args[1] as *mut Stat),
-        SYS_LSTAT => sys_lstat(args[0] as *const u8, args[1] as *mut Stat),
-        SYS_POLL => sys_poll(args[0] as *mut PollFd, args[1], args[2]),
-        SYS_ACCESS => sys_access(args[0] as *const u8, args[1]),
-        SYS_PIPE => sys_pipe(args[0] as *mut u32),
-        SYS_SELECT => sys_select(
-            args[0],
-            args[1] as *mut u32,
-            args[2] as *mut u32,
-            args[3] as *mut u32,
-            args[4] as *const TimeVal,
-        ),
-        SYS_DUP2 => sys_dup2(args[0], args[1]),
-        SYS_ALARM => {
-            warn!("sys_alarm is unimplemented");
-            Ok(0)
-        }
-        SYS_FORK => sys_fork(tf),
-        // use fork for vfork
-        SYS_VFORK => sys_fork(tf),
-        SYS_RENAME => sys_rename(args[0] as *const u8, args[1] as *const u8),
-        SYS_MKDIR => sys_mkdir(args[0] as *const u8, args[1]),
-        SYS_RMDIR => sys_rmdir(args[0] as *const u8),
-        SYS_LINK => sys_link(args[0] as *const u8, args[1] as *const u8),
-        SYS_UNLINK => sys_unlink(args[0] as *const u8),
-        SYS_READLINK => sys_readlink(args[0] as *const u8, args[1] as *mut u8, args[2]),
-        // 90
-        SYS_CHMOD => {
-            warn!("sys_chmod is unimplemented");
-            Ok(0)
-        }
-        SYS_CHOWN => {
-            warn!("sys_chown is unimplemented");
-            Ok(0)
-        }
-        SYS_ARCH_PRCTL => sys_arch_prctl(args[0] as i32, args[1], tf),
-        SYS_TIME => sys_time(args[0] as *mut u64),
-        SYS_EPOLL_CREATE => {
-            warn!("sys_epoll_create is unimplemented");
-            Err(SysError::ENOSYS)
-        }
-        _ => {
-            return None;
-        }
-    };
-    Some(ret)
+    #[cfg(target_arch = "mips")]
+    fn mips_syscall(&mut self, id: usize, args: [usize; 6]) -> Option<SysResult> {
+        let ret = match id {
+            SYS_OPEN => self.sys_open(args[0] as *const u8, args[1], args[2]),
+            SYS_POLL => self.sys_poll(args[0] as *mut PollFd, args[1], args[2]),
+            SYS_DUP2 => self.sys_dup2(args[0], args[1]),
+            SYS_FORK => self.sys_fork(),
+            SYS_MMAP2 => self.sys_mmap(args[0], args[1], args[2], args[3], args[4], args[5] * 4096),
+            SYS_FSTAT64 => self.sys_fstat(args[0], args[1] as *mut Stat),
+            SYS_LSTAT64 => self.sys_lstat(args[0] as *const u8, args[1] as *mut Stat),
+            SYS_STAT64 => self.sys_stat(args[0] as *const u8, args[1] as *mut Stat),
+            SYS_PIPE => {
+                let fd_ptr = args[0] as *mut u32;
+                match self.sys_pipe(fd_ptr) {
+                    Ok(code) => {
+                        unsafe {
+                            tf.v0 = *fd_ptr as usize;
+                            tf.v1 = *(fd_ptr.add(1)) as usize;
+                        }
+                        Ok(tf.v0)
+                    }
+                    Err(err) => Err(err),
+                }
+            }
+            SYS_FCNTL64 => self.unimplemented("fcntl64", Ok(0)),
+            SYS_SET_THREAD_AREA => {
+                info!("set_thread_area: tls: 0x{:x}", args[0]);
+                extern "C" {
+                    fn _cur_tls();
+                }
+
+                unsafe {
+                    asm!("mtc0 $0, $$4, 2": :"r"(args[0]));
+                    *(_cur_tls as *mut usize) = args[0];
+                }
+                Ok(0)
+            }
+            _ => return None,
+        };
+        Some(ret)
+    }
+
+    #[cfg(target_arch = "x86_64")]
+    fn x86_64_syscall(&mut self, id: usize, args: [usize; 6]) -> Option<SysResult> {
+        let ret = match id {
+            SYS_OPEN => self.sys_open(args[0] as *const u8, args[1], args[2]),
+            SYS_STAT => self.sys_stat(args[0] as *const u8, args[1] as *mut Stat),
+            SYS_LSTAT => self.sys_lstat(args[0] as *const u8, args[1] as *mut Stat),
+            SYS_POLL => self.sys_poll(args[0] as *mut PollFd, args[1], args[2]),
+            SYS_ACCESS => self.sys_access(args[0] as *const u8, args[1]),
+            SYS_PIPE => self.sys_pipe(args[0] as *mut u32),
+            SYS_SELECT => self.sys_select(
+                args[0],
+                args[1] as *mut u32,
+                args[2] as *mut u32,
+                args[3] as *mut u32,
+                args[4] as *const TimeVal,
+            ),
+            SYS_DUP2 => self.sys_dup2(args[0], args[1]),
+            SYS_ALARM => self.unimplemented("alarm", Ok(0)),
+            SYS_FORK => self.sys_fork(),
+            SYS_VFORK => self.sys_vfork(),
+            SYS_RENAME => self.sys_rename(args[0] as *const u8, args[1] as *const u8),
+            SYS_MKDIR => self.sys_mkdir(args[0] as *const u8, args[1]),
+            SYS_RMDIR => self.sys_rmdir(args[0] as *const u8),
+            SYS_LINK => self.sys_link(args[0] as *const u8, args[1] as *const u8),
+            SYS_UNLINK => self.sys_unlink(args[0] as *const u8),
+            SYS_READLINK => self.sys_readlink(args[0] as *const u8, args[1] as *mut u8, args[2]),
+            SYS_CHMOD => self.unimplemented("chmod", Ok(0)),
+            SYS_CHOWN => self.unimplemented("chown", Ok(0)),
+            SYS_ARCH_PRCTL => self.sys_arch_prctl(args[0] as i32, args[1]),
+            SYS_TIME => self.sys_time(args[0] as *mut u64),
+            SYS_EPOLL_CREATE => self.unimplemented("epoll_create", Err(SysError::ENOSYS)),
+            _ => return None,
+        };
+        Some(ret)
+    }
 }
 
 pub type SysResult = Result<usize, SysError>;

kernel/src/syscall/net.rs

@@ -2,295 +2,300 @@
 
 use super::fs::IoVecs;
 use super::*;
-use crate::drivers::SOCKET_ACTIVITY;
 use crate::fs::FileLike;
+use crate::memory::MemorySet;
 use crate::net::{
     Endpoint, LinkLevelEndpoint, NetlinkEndpoint, NetlinkSocketState, PacketSocketState,
     RawSocketState, Socket, TcpSocketState, UdpSocketState, SOCKETS,
 };
-use crate::sync::{MutexGuard, SpinNoIrq, SpinNoIrqLock as Mutex};
 use alloc::boxed::Box;
 use core::cmp::min;
 use core::mem::size_of;
 use smoltcp::wire::*;
 
-pub fn sys_socket(domain: usize, socket_type: usize, protocol: usize) -> SysResult {
-    let domain = AddressFamily::from(domain as u16);
-    let socket_type = SocketType::from(socket_type as u8 & SOCK_TYPE_MASK);
-    info!(
-        "socket: domain: {:?}, socket_type: {:?}, protocol: {}",
-        domain, socket_type, protocol
-    );
-    let mut proc = process();
-    let socket: Box<dyn Socket> = match domain {
-        AddressFamily::Internet | AddressFamily::Unix => match socket_type {
-            SocketType::Stream => Box::new(TcpSocketState::new()),
-            SocketType::Datagram => Box::new(UdpSocketState::new()),
-            SocketType::Raw => Box::new(RawSocketState::new(protocol as u8)),
-            _ => return Err(SysError::EINVAL),
-        },
-        AddressFamily::Packet => match socket_type {
-            SocketType::Raw => Box::new(PacketSocketState::new()),
-            _ => return Err(SysError::EINVAL),
-        },
-        AddressFamily::Netlink => match socket_type {
-            SocketType::Raw => Box::new(NetlinkSocketState::new()),
-            _ => return Err(SysError::EINVAL),
-        },
-        _ => return Err(SysError::EAFNOSUPPORT),
-    };
-    let fd = proc.get_free_fd();
-    proc.files.insert(fd, FileLike::Socket(socket));
-    Ok(fd)
-}
-
-pub fn sys_setsockopt(
-    fd: usize,
-    level: usize,
-    optname: usize,
-    optval: *const u8,
-    optlen: usize,
-) -> SysResult {
-    info!(
-        "setsockopt: fd: {}, level: {}, optname: {}",
-        fd, level, optname
-    );
-    let mut proc = process();
-    proc.vm.check_read_array(optval, optlen)?;
-    let data = unsafe { slice::from_raw_parts(optval, optlen) };
-    let socket = proc.get_socket(fd)?;
-    socket.setsockopt(level, optname, data)
-}
-
-pub fn sys_getsockopt(
-    fd: usize,
-    level: usize,
-    optname: usize,
-    optval: *mut u8,
-    optlen: *mut u32,
-) -> SysResult {
-    info!(
-        "getsockopt: fd: {}, level: {}, optname: {} optval: {:?} optlen: {:?}",
-        fd, level, optname, optval, optlen
-    );
-    let proc = process();
-    proc.vm.check_write_ptr(optlen)?;
-    match level {
-        SOL_SOCKET => match optname {
-            SO_SNDBUF => {
-                proc.vm.check_write_array(optval, 4)?;
-                unsafe {
-                    *(optval as *mut u32) = crate::net::TCP_SENDBUF as u32;
-                    *optlen = 4;
-                }
-                Ok(0)
-            }
-            SO_RCVBUF => {
-                proc.vm.check_write_array(optval, 4)?;
-                unsafe {
-                    *(optval as *mut u32) = crate::net::TCP_RECVBUF as u32;
-                    *optlen = 4;
-                }
-                Ok(0)
-            }
-            _ => Err(SysError::ENOPROTOOPT),
-        },
-        IPPROTO_TCP => match optname {
-            TCP_CONGESTION => Ok(0),
-            _ => Err(SysError::ENOPROTOOPT),
-        },
-        _ => Err(SysError::ENOPROTOOPT),
+impl Syscall<'_> {
+    pub fn sys_socket(&mut self, domain: usize, socket_type: usize, protocol: usize) -> SysResult {
+        let domain = AddressFamily::from(domain as u16);
+        let socket_type = SocketType::from(socket_type as u8 & SOCK_TYPE_MASK);
+        info!(
+            "socket: domain: {:?}, socket_type: {:?}, protocol: {}",
+            domain, socket_type, protocol
+        );
+        let mut proc = self.process();
+        let socket: Box<dyn Socket> = match domain {
+            AddressFamily::Internet | AddressFamily::Unix => match socket_type {
+                SocketType::Stream => Box::new(TcpSocketState::new()),
+                SocketType::Datagram => Box::new(UdpSocketState::new()),
+                SocketType::Raw => Box::new(RawSocketState::new(protocol as u8)),
+                _ => return Err(SysError::EINVAL),
+            },
+            AddressFamily::Packet => match socket_type {
+                SocketType::Raw => Box::new(PacketSocketState::new()),
+                _ => return Err(SysError::EINVAL),
+            },
+            AddressFamily::Netlink => match socket_type {
+                SocketType::Raw => Box::new(NetlinkSocketState::new()),
+                _ => return Err(SysError::EINVAL),
+            },
+            _ => return Err(SysError::EAFNOSUPPORT),
+        };
+        let fd = proc.add_file(FileLike::Socket(socket));
+        Ok(fd)
     }
-}
 
-pub fn sys_connect(fd: usize, addr: *const SockAddr, addr_len: usize) -> SysResult {
-    info!(
-        "sys_connect: fd: {}, addr: {:?}, addr_len: {}",
-        fd, addr, addr_len
-    );
+    pub fn sys_setsockopt(
+        &mut self,
+        fd: usize,
+        level: usize,
+        optname: usize,
+        optval: *const u8,
+        optlen: usize,
+    ) -> SysResult {
+        info!(
+            "setsockopt: fd: {}, level: {}, optname: {}",
+            fd, level, optname
+        );
+        let mut proc = self.process();
+        let data = unsafe { self.vm().check_read_array(optval, optlen)? };
+        let socket = proc.get_socket(fd)?;
+        socket.setsockopt(level, optname, data)
+    }
 
-    let mut proc = process();
-    let endpoint = sockaddr_to_endpoint(&mut proc, addr, addr_len)?;
-    let socket = proc.get_socket(fd)?;
-    socket.connect(endpoint)?;
-    Ok(0)
-}
-
-pub fn sys_sendto(
-    fd: usize,
-    base: *const u8,
-    len: usize,
-    _flags: usize,
-    addr: *const SockAddr,
-    addr_len: usize,
-) -> SysResult {
-    info!(
-        "sys_sendto: fd: {} base: {:?} len: {} addr: {:?} addr_len: {}",
-        fd, base, len, addr, addr_len
-    );
-
-    let mut proc = process();
-    proc.vm.check_read_array(base, len)?;
-
-    let slice = unsafe { slice::from_raw_parts(base, len) };
-    let endpoint = if addr.is_null() {
-        None
-    } else {
-        let endpoint = sockaddr_to_endpoint(&mut proc, addr, addr_len)?;
-        info!("sys_sendto: sending to endpoint {:?}", endpoint);
-        Some(endpoint)
-    };
-    let socket = proc.get_socket(fd)?;
-    socket.write(&slice, endpoint)
-}
-
-pub fn sys_recvfrom(
-    fd: usize,
-    base: *mut u8,
-    len: usize,
-    flags: usize,
-    addr: *mut SockAddr,
-    addr_len: *mut u32,
-) -> SysResult {
-    info!(
-        "sys_recvfrom: fd: {} base: {:?} len: {} flags: {} addr: {:?} addr_len: {:?}",
-        fd, base, len, flags, addr, addr_len
-    );
-
-    let mut proc = process();
-    proc.vm.check_write_array(base, len)?;
-
-    let socket = proc.get_socket(fd)?;
-    let mut slice = unsafe { slice::from_raw_parts_mut(base, len) };
-    let (result, endpoint) = socket.read(&mut slice);
-
-    if result.is_ok() && !addr.is_null() {
-        let sockaddr_in = SockAddr::from(endpoint);
-        unsafe {
-            sockaddr_in.write_to(&mut proc, addr, addr_len)?;
+    pub fn sys_getsockopt(
+        &mut self,
+        fd: usize,
+        level: usize,
+        optname: usize,
+        optval: *mut u8,
+        optlen: *mut u32,
+    ) -> SysResult {
+        info!(
+            "getsockopt: fd: {}, level: {}, optname: {} optval: {:?} optlen: {:?}",
+            fd, level, optname, optval, optlen
+        );
+        let optlen = unsafe { self.vm().check_write_ptr(optlen)? };
+        match level {
+            SOL_SOCKET => match optname {
+                SO_SNDBUF => {
+                    let optval = unsafe { self.vm().check_write_ptr(optval as *mut u32)? };
+                    *optval = crate::net::TCP_SENDBUF as u32;
+                    *optlen = 4;
+                    Ok(0)
+                }
+                SO_RCVBUF => {
+                    let optval = unsafe { self.vm().check_write_ptr(optval as *mut u32)? };
+                    *optval = crate::net::TCP_RECVBUF as u32;
+                    *optlen = 4;
+                    Ok(0)
+                }
+                _ => Err(SysError::ENOPROTOOPT),
+            },
+            IPPROTO_TCP => match optname {
+                TCP_CONGESTION => Ok(0),
+                _ => Err(SysError::ENOPROTOOPT),
+            },
+            _ => Err(SysError::ENOPROTOOPT),
         }
     }
 
-    result
-}
+    pub fn sys_connect(&mut self, fd: usize, addr: *const SockAddr, addr_len: usize) -> SysResult {
+        info!(
+            "sys_connect: fd: {}, addr: {:?}, addr_len: {}",
+            fd, addr, addr_len
+        );
 
-pub fn sys_recvmsg(fd: usize, msg: *mut MsgHdr, flags: usize) -> SysResult {
-    info!("recvmsg: fd: {}, msg: {:?}, flags: {}", fd, msg, flags);
-    let mut proc = process();
-    proc.vm.check_read_ptr(msg)?;
-    let hdr = unsafe { &mut *msg };
-    let mut iovs = IoVecs::check_and_new(hdr.msg_iov, hdr.msg_iovlen, &proc.vm, true)?;
+        let mut proc = self.process();
+        let endpoint = sockaddr_to_endpoint(&mut self.vm(), addr, addr_len)?;
+        let socket = proc.get_socket(fd)?;
+        socket.connect(endpoint)?;
+        Ok(0)
+    }
 
-    let mut buf = iovs.new_buf(true);
-    let socket = proc.get_socket(fd)?;
-    let (result, endpoint) = socket.read(&mut buf);
+    pub fn sys_sendto(
+        &mut self,
+        fd: usize,
+        base: *const u8,
+        len: usize,
+        _flags: usize,
+        addr: *const SockAddr,
+        addr_len: usize,
+    ) -> SysResult {
+        info!(
+            "sys_sendto: fd: {} base: {:?} len: {} addr: {:?} addr_len: {}",
+            fd, base, len, addr, addr_len
+        );
 
-    if let Ok(len) = result {
-        // copy data to user
-        iovs.write_all_from_slice(&buf[..len]);
+        let mut proc = self.process();
+
+        let slice = unsafe { self.vm().check_read_array(base, len)? };
+        let endpoint = if addr.is_null() {
+            None
+        } else {
+            let endpoint = sockaddr_to_endpoint(&mut self.vm(), addr, addr_len)?;
+            info!("sys_sendto: sending to endpoint {:?}", endpoint);
+            Some(endpoint)
+        };
+        let socket = proc.get_socket(fd)?;
+        socket.write(&slice, endpoint)
+    }
+
+    pub fn sys_recvfrom(
+        &mut self,
+        fd: usize,
+        base: *mut u8,
+        len: usize,
+        flags: usize,
+        addr: *mut SockAddr,
+        addr_len: *mut u32,
+    ) -> SysResult {
+        info!(
+            "sys_recvfrom: fd: {} base: {:?} len: {} flags: {} addr: {:?} addr_len: {:?}",
+            fd, base, len, flags, addr, addr_len
+        );
+
+        let mut proc = self.process();
+
+        let mut slice = unsafe { self.vm().check_write_array(base, len)? };
+        let socket = proc.get_socket(fd)?;
+        let (result, endpoint) = socket.read(&mut slice);
+
+        if result.is_ok() && !addr.is_null() {
+            let sockaddr_in = SockAddr::from(endpoint);
+            unsafe {
+                sockaddr_in.write_to(&mut self.vm(), addr, addr_len)?;
+            }
+        }
+
+        result
+    }
+
+    pub fn sys_recvmsg(&mut self, fd: usize, msg: *mut MsgHdr, flags: usize) -> SysResult {
+        info!("recvmsg: fd: {}, msg: {:?}, flags: {}", fd, msg, flags);
+        let mut proc = self.process();
+        let hdr = unsafe { self.vm().check_write_ptr(msg)? };
+        let mut iovs =
+            unsafe { IoVecs::check_and_new(hdr.msg_iov, hdr.msg_iovlen, &self.vm(), true)? };
+
+        let mut buf = iovs.new_buf(true);
+        let socket = proc.get_socket(fd)?;
+        let (result, endpoint) = socket.read(&mut buf);
+
+        if let Ok(len) = result {
+            // copy data to user
+            iovs.write_all_from_slice(&buf[..len]);
+            let sockaddr_in = SockAddr::from(endpoint);
+            unsafe {
+                sockaddr_in.write_to(&mut self.vm(), hdr.msg_name, &mut hdr.msg_namelen as *mut u32)?;
+            }
+        }
+        result
+    }
+
+    pub fn sys_bind(&mut self, fd: usize, addr: *const SockAddr, addr_len: usize) -> SysResult {
+        info!("sys_bind: fd: {} addr: {:?} len: {}", fd, addr, addr_len);
+        let mut proc = self.process();
+
+        let endpoint = sockaddr_to_endpoint(&mut self.vm(), addr, addr_len)?;
+        info!("sys_bind: fd: {} bind to {:?}", fd, endpoint);
+
+        let socket = proc.get_socket(fd)?;
+        socket.bind(endpoint)
+    }
+
+    pub fn sys_listen(&mut self, fd: usize, backlog: usize) -> SysResult {
+        info!("sys_listen: fd: {} backlog: {}", fd, backlog);
+        // smoltcp tcp sockets do not support backlog
+        // open multiple sockets for each connection
+        let mut proc = self.process();
+
+        let socket = proc.get_socket(fd)?;
+        socket.listen()
+    }
+
+    pub fn sys_shutdown(&mut self, fd: usize, how: usize) -> SysResult {
+        info!("sys_shutdown: fd: {} how: {}", fd, how);
+        let mut proc = self.process();
+
+        let socket = proc.get_socket(fd)?;
+        socket.shutdown()
+    }
+
+    pub fn sys_accept(&mut self, fd: usize, addr: *mut SockAddr, addr_len: *mut u32) -> SysResult {
+        info!(
+            "sys_accept: fd: {} addr: {:?} addr_len: {:?}",
+            fd, addr, addr_len
+        );
+        // smoltcp tcp sockets do not support backlog
+        // open multiple sockets for each connection
+        let mut proc = self.process();
+
+        let socket = proc.get_socket(fd)?;
+        let (new_socket, remote_endpoint) = socket.accept()?;
+
+        let new_fd = proc.add_file(FileLike::Socket(new_socket));
+
+        if !addr.is_null() {
+            let sockaddr_in = SockAddr::from(remote_endpoint);
+            unsafe {
+                sockaddr_in.write_to(&mut self.vm(), addr, addr_len)?;
+            }
+        }
+        Ok(new_fd)
+    }
+
+    pub fn sys_getsockname(
+        &mut self,
+        fd: usize,
+        addr: *mut SockAddr,
+        addr_len: *mut u32,
+    ) -> SysResult {
+        info!(
+            "sys_getsockname: fd: {} addr: {:?} addr_len: {:?}",
+            fd, addr, addr_len
+        );
+
+        let mut proc = self.process();
+
+        if addr.is_null() {
+            return Err(SysError::EINVAL);
+        }
+
+        let socket = proc.get_socket(fd)?;
+        let endpoint = socket.endpoint().ok_or(SysError::EINVAL)?;
         let sockaddr_in = SockAddr::from(endpoint);
         unsafe {
-            sockaddr_in.write_to(&mut proc, hdr.msg_name, &mut hdr.msg_namelen as *mut u32)?;
+            sockaddr_in.write_to(&mut self.vm(), addr, addr_len)?;
         }
+        Ok(0)
     }
-    result
-}
 
-pub fn sys_bind(fd: usize, addr: *const SockAddr, addr_len: usize) -> SysResult {
-    info!("sys_bind: fd: {} addr: {:?} len: {}", fd, addr, addr_len);
-    let mut proc = process();
+    pub fn sys_getpeername(
+        &mut self,
+        fd: usize,
+        addr: *mut SockAddr,
+        addr_len: *mut u32,
+    ) -> SysResult {
+        info!(
+            "sys_getpeername: fd: {} addr: {:?} addr_len: {:?}",
+            fd, addr, addr_len
+        );
 
-    let mut endpoint = sockaddr_to_endpoint(&mut proc, addr, addr_len)?;
-    info!("sys_bind: fd: {} bind to {:?}", fd, endpoint);
+        // smoltcp tcp sockets do not support backlog
+        // open multiple sockets for each connection
+        let mut proc = self.process();
 
-    let socket = proc.get_socket(fd)?;
-    socket.bind(endpoint)
-}
+        if addr as usize == 0 {
+            return Err(SysError::EINVAL);
+        }
 
-pub fn sys_listen(fd: usize, backlog: usize) -> SysResult {
-    info!("sys_listen: fd: {} backlog: {}", fd, backlog);
-    // smoltcp tcp sockets do not support backlog
-    // open multiple sockets for each connection
-    let mut proc = process();
-
-    let socket = proc.get_socket(fd)?;
-    socket.listen()
-}
-
-pub fn sys_shutdown(fd: usize, how: usize) -> SysResult {
-    info!("sys_shutdown: fd: {} how: {}", fd, how);
-    let mut proc = process();
-
-    let socket = proc.get_socket(fd)?;
-    socket.shutdown()
-}
-
-pub fn sys_accept(fd: usize, addr: *mut SockAddr, addr_len: *mut u32) -> SysResult {
-    info!(
-        "sys_accept: fd: {} addr: {:?} addr_len: {:?}",
-        fd, addr, addr_len
-    );
-    // smoltcp tcp sockets do not support backlog
-    // open multiple sockets for each connection
-    let mut proc = process();
-
-    let socket = proc.get_socket(fd)?;
-    let (new_socket, remote_endpoint) = socket.accept()?;
-
-    let new_fd = proc.get_free_fd();
-    proc.files.insert(new_fd, FileLike::Socket(new_socket));
-
-    if !addr.is_null() {
+        let socket = proc.get_socket(fd)?;
+        let remote_endpoint = socket.remote_endpoint().ok_or(SysError::EINVAL)?;
         let sockaddr_in = SockAddr::from(remote_endpoint);
         unsafe {
-            sockaddr_in.write_to(&mut proc, addr, addr_len)?;
+            sockaddr_in.write_to(&mut self.vm(), addr, addr_len)?;
         }
+        Ok(0)
     }
-    Ok(new_fd)
-}
-
-pub fn sys_getsockname(fd: usize, addr: *mut SockAddr, addr_len: *mut u32) -> SysResult {
-    info!(
-        "sys_getsockname: fd: {} addr: {:?} addr_len: {:?}",
-        fd, addr, addr_len
-    );
-
-    let mut proc = process();
-
-    if addr.is_null() {
-        return Err(SysError::EINVAL);
-    }
-
-    let socket = proc.get_socket(fd)?;
-    let endpoint = socket.endpoint().ok_or(SysError::EINVAL)?;
-    let sockaddr_in = SockAddr::from(endpoint);
-    unsafe {
-        sockaddr_in.write_to(&mut proc, addr, addr_len)?;
-    }
-    Ok(0)
-}
-
-pub fn sys_getpeername(fd: usize, addr: *mut SockAddr, addr_len: *mut u32) -> SysResult {
-    info!(
-        "sys_getpeername: fd: {} addr: {:?} addr_len: {:?}",
-        fd, addr, addr_len
-    );
-
-    // smoltcp tcp sockets do not support backlog
-    // open multiple sockets for each connection
-    let mut proc = process();
-
-    if addr as usize == 0 {
-        return Err(SysError::EINVAL);
-    }
-
-    let socket = proc.get_socket(fd)?;
-    let remote_endpoint = socket.remote_endpoint().ok_or(SysError::EINVAL)?;
-    let sockaddr_in = SockAddr::from(remote_endpoint);
-    unsafe {
-        sockaddr_in.write_to(&mut proc, addr, addr_len)?;
-    }
-    Ok(0)
 }
 
 impl Process {
@@ -401,42 +406,36 @@ impl From<Endpoint> for SockAddr {
 /// Convert sockaddr to endpoint
 // Check len is long enough
 fn sockaddr_to_endpoint(
-    proc: &mut Process,
+    vm: &MemorySet,
     addr: *const SockAddr,
     len: usize,
 ) -> Result<Endpoint, SysError> {
     if len < size_of::<u16>() {
         return Err(SysError::EINVAL);
     }
-    proc.vm.check_read_array(addr as *const u8, len)?;
+    let addr = unsafe { vm.check_read_ptr(addr)? };
+    if len < addr.len()? {
+        return Err(SysError::EINVAL);
+    }
     unsafe {
-        match AddressFamily::from((*addr).family) {
+        match AddressFamily::from(addr.family) {
             AddressFamily::Internet => {
-                if len < size_of::<SockAddrIn>() {
-                    return Err(SysError::EINVAL);
-                }
-                let port = u16::from_be((*addr).addr_in.sin_port);
+                let port = u16::from_be(addr.addr_in.sin_port);
                 let addr = IpAddress::from(Ipv4Address::from_bytes(
-                    &u32::from_be((*addr).addr_in.sin_addr).to_be_bytes()[..],
+                    &u32::from_be(addr.addr_in.sin_addr).to_be_bytes()[..],
                 ));
                 Ok(Endpoint::Ip((addr, port).into()))
             }
             AddressFamily::Unix => Err(SysError::EINVAL),
             AddressFamily::Packet => {
-                if len < size_of::<SockAddrLl>() {
-                    return Err(SysError::EINVAL);
-                }
                 Ok(Endpoint::LinkLevel(LinkLevelEndpoint::new(
-                    (*addr).addr_ll.sll_ifindex as usize,
+                    addr.addr_ll.sll_ifindex as usize,
                 )))
             }
             AddressFamily::Netlink => {
-                if len < size_of::<SockAddrNl>() {
-                    return Err(SysError::EINVAL);
-                }
                 Ok(Endpoint::Netlink(NetlinkEndpoint::new(
-                    (*addr).addr_nl.nl_pid,
-                    (*addr).addr_nl.nl_groups,
+                    addr.addr_nl.nl_pid,
+                    addr.addr_nl.nl_groups,
                 )))
             }
             _ => Err(SysError::EINVAL),
@@ -445,11 +444,21 @@ fn sockaddr_to_endpoint(
 }
 
 impl SockAddr {
+    fn len(&self) -> Result<usize, SysError> {
+        match AddressFamily::from(unsafe { self.family }) {
+            AddressFamily::Internet => Ok(size_of::<SockAddrIn>()),
+            AddressFamily::Packet => Ok(size_of::<SockAddrLl>()),
+            AddressFamily::Netlink => Ok(size_of::<SockAddrNl>()),
+            AddressFamily::Unix => Err(SysError::EINVAL),
+            _ => Err(SysError::EINVAL),
+        }
+    }
+
     /// Write to user sockaddr
     /// Check mutability for user
     unsafe fn write_to(
         self,
-        proc: &mut Process,
+        vm: &MemorySet,
         addr: *mut SockAddr,
         addr_len: *mut u32,
     ) -> SysResult {
@@ -458,24 +467,17 @@ impl SockAddr {
             return Ok(0);
         }
 
-        proc.vm.check_write_ptr(addr_len)?;
+        let addr_len = unsafe { vm.check_write_ptr(addr_len)? };
         let max_addr_len = *addr_len as usize;
-        let full_len = match AddressFamily::from(self.family) {
-            AddressFamily::Internet => size_of::<SockAddrIn>(),
-            AddressFamily::Packet => size_of::<SockAddrLl>(),
-            AddressFamily::Netlink => size_of::<SockAddrNl>(),
-            AddressFamily::Unix => return Err(SysError::EINVAL),
-            _ => return Err(SysError::EINVAL),
-        };
+        let full_len = self.len()?;
 
         let written_len = min(max_addr_len, full_len);
         if written_len > 0 {
-            proc.vm.check_write_array(addr as *mut u8, written_len)?;
+            let target = unsafe { vm.check_write_array(addr as *mut u8, written_len)? };
             let source = slice::from_raw_parts(&self as *const SockAddr as *const u8, written_len);
-            let target = slice::from_raw_parts_mut(addr as *mut u8, written_len);
             target.copy_from_slice(source);
         }
-        addr_len.write(full_len as u32);
+        *addr_len = full_len as u32;
         return Ok(0);
     }
 }

kernel/src/syscall/proc.rs

@@ -3,343 +3,349 @@
 use super::*;
 use crate::fs::INodeExt;
 
-/// Fork the current process. Return the child's PID.
-pub fn sys_fork(tf: &TrapFrame) -> SysResult {
-    let new_thread = current_thread().fork(tf);
-    let pid = processor().manager().add(new_thread);
-    info!("fork: {} -> {}", thread::current().id(), pid);
-    Ok(pid)
-}
+impl Syscall<'_> {
+    /// Fork the current process. Return the child's PID.
+    pub fn sys_fork(&mut self) -> SysResult {
+        let new_thread = self.thread.fork(self.tf);
+        let pid = new_thread.proc.lock().pid.get();
+        let tid = processor().manager().add(new_thread);
+        processor().manager().detach(tid);
+        info!("fork: {} -> {}", thread::current().id(), pid);
+        Ok(pid)
+    }
 
-/// Create a new thread in the current process.
-/// The new thread's stack pointer will be set to `newsp`,
-/// and thread pointer will be set to `newtls`.
-/// The child tid will be stored at both `parent_tid` and `child_tid`.
-/// This is partially implemented for musl only.
-pub fn sys_clone(
-    flags: usize,
-    newsp: usize,
-    parent_tid: *mut u32,
-    child_tid: *mut u32,
-    newtls: usize,
-    tf: &TrapFrame,
-) -> SysResult {
-    let clone_flags = CloneFlags::from_bits_truncate(flags);
-    info!(
-        "clone: flags: {:?} == {:#x}, newsp: {:#x}, parent_tid: {:?}, child_tid: {:?}, newtls: {:#x}",
-        clone_flags, flags, newsp, parent_tid, child_tid, newtls
-    );
-    if flags == 0x4111 || flags == 0x11 {
-        warn!("sys_clone is calling sys_fork instead, ignoring other args");
-        return sys_fork(tf);
+    pub fn sys_vfork(&mut self) -> SysResult {
+        self.sys_fork()
     }
-    if (flags != 0x7d0f00) && (flags != 0x5d0f00) {
-        //0x5d0f00 is the args from gcc of alpine linux
-        //warn!("sys_clone only support musl pthread_create");
-        panic!(
-            "sys_clone only support sys_fork OR musl pthread_create without flags{:x}",
-            flags
-        );
-        //return Err(SysError::ENOSYS);
-    }
-    {
-        let proc = process();
-        proc.vm.check_write_ptr(parent_tid)?;
-        proc.vm.check_write_ptr(child_tid)?;
-    }
-    let new_thread = current_thread().clone(tf, newsp, newtls, child_tid as usize);
-    // FIXME: parent pid
-    let tid = processor().manager().add(new_thread);
-    info!("clone: {} -> {}", thread::current().id(), tid);
-    unsafe {
-        parent_tid.write(tid as u32);
-        child_tid.write(tid as u32);
-    }
-    Ok(tid)
-}
 
-/// Wait for the process exit.
-/// Return the PID. Store exit code to `wstatus` if it's not null.
-pub fn sys_wait4(pid: isize, wstatus: *mut i32) -> SysResult {
-    info!("wait4: pid: {}, code: {:?}", pid, wstatus);
-    if !wstatus.is_null() {
-        process().vm.check_write_ptr(wstatus)?;
-    }
-    #[derive(Debug)]
-    enum WaitFor {
-        AnyChild,
-        Pid(usize),
-    }
-    let target = match pid {
-        -1 | 0 => WaitFor::AnyChild,
-        p if p > 0 => WaitFor::Pid(p as usize),
-        _ => unimplemented!(),
-    };
-    loop {
-        let mut proc = process();
-        // check child_exit_code
-        let find = match target {
-            WaitFor::AnyChild => proc
-                .child_exit_code
-                .iter()
-                .next()
-                .map(|(&pid, &code)| (pid, code)),
-            WaitFor::Pid(pid) => proc.child_exit_code.get(&pid).map(|&code| (pid, code)),
-        };
-        // if found, return
-        if let Some((pid, exit_code)) = find {
-            proc.child_exit_code.remove(&pid);
-            if !wstatus.is_null() {
-                unsafe {
-                    wstatus.write(exit_code as i32);
-                }
-            }
-            return Ok(pid);
-        }
-        // if not, check pid
-        let children: Vec<_> = proc
-            .children
-            .iter()
-            .filter_map(|weak| weak.upgrade())
-            .collect();
-        let invalid = match target {
-            WaitFor::AnyChild => children.len() == 0,
-            WaitFor::Pid(pid) => children
-                .iter()
-                .find(|p| p.lock().pid.get() == pid)
-                .is_none(),
-        };
-        if invalid {
-            return Err(SysError::ECHILD);
-        }
+    /// Create a new thread in the current process.
+    /// The new thread's stack pointer will be set to `newsp`,
+    /// and thread pointer will be set to `newtls`.
+    /// The child tid will be stored at both `parent_tid` and `child_tid`.
+    /// This is partially implemented for musl only.
+    pub fn sys_clone(
+        &mut self,
+        flags: usize,
+        newsp: usize,
+        parent_tid: *mut u32,
+        child_tid: *mut u32,
+        newtls: usize,
+    ) -> SysResult {
+        let clone_flags = CloneFlags::from_bits_truncate(flags);
         info!(
-            "wait: thread {} -> {:?}, sleep",
-            thread::current().id(),
-            target
+            "clone: flags: {:?} == {:#x}, newsp: {:#x}, parent_tid: {:?}, child_tid: {:?}, newtls: {:#x}",
+            clone_flags, flags, newsp, parent_tid, child_tid, newtls
         );
-        let condvar = proc.child_exit.clone();
-        drop(proc); // must release lock of current process
-        condvar._wait();
-    }
-}
-
-pub fn sys_exec(
-    name: *const u8,
-    argv: *const *const u8,
-    envp: *const *const u8,
-    tf: &mut TrapFrame,
-) -> SysResult {
-    info!("exec: name: {:?}, argv: {:?}, envp: {:?}", name, argv, envp);
-    let proc = process();
-    let exec_name = if name.is_null() {
-        String::from("")
-    } else {
-        unsafe { proc.vm.check_and_clone_cstr(name)? }
-    };
-
-    if argv.is_null() {
-        return Err(SysError::EINVAL);
-    }
-    // Check and copy args to kernel
-    let mut args = Vec::new();
-    unsafe {
-        let mut current_argv = argv as *const *const u8;
-        proc.vm.check_read_ptr(current_argv)?;
-        while !(*current_argv).is_null() {
-            let arg = proc.vm.check_and_clone_cstr(*current_argv)?;
-            args.push(arg);
-            current_argv = current_argv.add(1);
+        if flags == 0x4111 || flags == 0x11 {
+            warn!("sys_clone is calling sys_fork instead, ignoring other args");
+            return self.sys_fork();
         }
-    }
-    // Check and copy envs to kernel
-    let mut envs = Vec::new();
-    unsafe {
-        let mut current_env = envp as *const *const u8;
-        if !current_env.is_null() {
-            proc.vm.check_read_ptr(current_env)?;
-            while !(*current_env).is_null() {
-                let env = proc.vm.check_and_clone_cstr(*current_env)?;
-                envs.push(env);
-                current_env = current_env.add(1);
-            }
+        if (flags != 0x7d0f00) && (flags != 0x5d0f00) {
+            //0x5d0f00 is the args from gcc of alpine linux
+            //warn!("sys_clone only support musl pthread_create");
+            panic!(
+                "sys_clone only support sys_fork OR musl pthread_create without flags{:x}",
+                flags
+            );
+            //return Err(SysError::ENOSYS);
         }
+        let parent_tid_ref = unsafe { self.vm().check_write_ptr(parent_tid)? };
+        let child_tid_ref = unsafe { self.vm().check_write_ptr(child_tid)? };
+        let new_thread = self
+            .thread
+            .clone(self.tf, newsp, newtls, child_tid as usize);
+        // FIXME: parent pid
+        let tid = processor().manager().add(new_thread);
+        processor().manager().detach(tid);
+        info!("clone: {} -> {}", thread::current().id(), tid);
+        *parent_tid_ref = tid as u32;
+        *child_tid_ref = tid as u32;
+        Ok(tid)
     }
 
-    if args.is_empty() {
-        return Err(SysError::EINVAL);
-    }
-
-    info!(
-        "exec: name: {:?}, args: {:?}, envp: {:?}",
-        exec_name, args, envs
-    );
-
-    // Read program file
-    //let path = args[0].as_str();
-    let exec_path = exec_name.as_str();
-    let inode = proc.lookup_inode(exec_path)?;
-    let buf = inode.read_as_vec()?;
-
-    // Make new Thread
-    let mut thread = Thread::new_user(buf.as_slice(), exec_path, args, envs);
-    thread.proc.lock().clone_for_exec(&proc);
-
-    // Activate new page table
-    unsafe {
-        thread.proc.lock().vm.activate();
-    }
-
-    // Modify the TrapFrame
-    *tf = unsafe { thread.context.get_init_tf() };
-
-    // Swap Context but keep KStack
-    ::core::mem::swap(&mut current_thread().kstack, &mut thread.kstack);
-    ::core::mem::swap(current_thread(), &mut *thread);
-
-    Ok(0)
-}
-
-pub fn sys_yield() -> SysResult {
-    thread::yield_now();
-    Ok(0)
-}
-
-/// Kill the process
-pub fn sys_kill(pid: usize, sig: usize) -> SysResult {
-    info!(
-        "kill: {} killed: {} with sig {}",
-        thread::current().id(),
-        pid,
-        sig
-    );
-    let current_pid = process().pid.get().clone();
-    if current_pid == pid {
-        // killing myself
-        sys_exit_group(sig);
-    } else {
-        if let Some(proc_arc) = PROCESSES.read().get(&pid).and_then(|weak| weak.upgrade()) {
-            let proc = proc_arc.lock();
-            // quit all threads
-            for tid in proc.threads.iter() {
-                processor().manager().exit(*tid, sig);
-            }
-            // notify parent and fill exit code
-            // avoid deadlock
-            let proc_parent = proc.parent.clone();
-            let pid = proc.pid.get();
-            drop(proc);
-            if let Some(parent) = proc_parent {
-                let mut parent = parent.lock();
-                parent.child_exit_code.insert(pid, sig);
-                parent.child_exit.notify_one();
-            }
-            Ok(0)
+    /// Wait for the process exit.
+    /// Return the PID. Store exit code to `wstatus` if it's not null.
+    pub fn sys_wait4(&mut self, pid: isize, wstatus: *mut i32) -> SysResult {
+        //info!("wait4: pid: {}, code: {:?}", pid, wstatus);
+        let wstatus = if !wstatus.is_null() {
+            Some(unsafe { self.vm().check_write_ptr(wstatus)? })
         } else {
-            Err(SysError::EINVAL)
+            None
+        };
+        #[derive(Debug)]
+        enum WaitFor {
+            AnyChild,
+            Pid(usize),
+        }
+        let target = match pid {
+            -1 | 0 => WaitFor::AnyChild,
+            p if p > 0 => WaitFor::Pid(p as usize),
+            _ => unimplemented!(),
+        };
+        loop {
+            let mut proc = self.process();
+            // check child_exit_code
+            let find = match target {
+                WaitFor::AnyChild => proc
+                    .child_exit_code
+                    .iter()
+                    .next()
+                    .map(|(&pid, &code)| (pid, code)),
+                WaitFor::Pid(pid) => proc.child_exit_code.get(&pid).map(|&code| (pid, code)),
+            };
+            // if found, return
+            if let Some((pid, exit_code)) = find {
+                proc.child_exit_code.remove(&pid);
+                if let Some(wstatus) = wstatus {
+                    *wstatus = exit_code as i32;
+                }
+                return Ok(pid);
+            }
+            // if not, check pid
+            let children: Vec<_> = proc
+                .children
+                .iter()
+                .filter_map(|weak| weak.upgrade())
+                .collect();
+            let invalid = match target {
+                WaitFor::AnyChild => children.len() == 0,
+                WaitFor::Pid(pid) => children
+                    .iter()
+                    .find(|p| p.lock().pid.get() == pid)
+                    .is_none(),
+            };
+            if invalid {
+                return Err(SysError::ECHILD);
+            }
+            info!(
+                "wait: thread {} -> {:?}, sleep",
+                thread::current().id(),
+                target
+            );
+            let condvar = proc.child_exit.clone();
+            condvar.wait(proc);
         }
     }
-}
 
-/// Get the current process id
-pub fn sys_getpid() -> SysResult {
-    info!("getpid");
-    Ok(process().pid.get())
-}
+    /// Replaces the current ** process ** with a new process image
+    ///
+    /// `argv` is an array of argument strings passed to the new program.
+    /// `envp` is an array of strings, conventionally of the form `key=value`,
+    /// which are passed as environment to the new program.
+    ///
+    /// NOTICE: `argv` & `envp` can not be NULL (different from Linux)
+    ///
+    /// NOTICE: for multi-thread programs
+    /// A call to any exec function from a process with more than one thread
+    /// shall result in all threads being terminated and the new executable image
+    /// being loaded and executed.
+    pub fn sys_exec(
+        &mut self,
+        path: *const u8,
+        argv: *const *const u8,
+        envp: *const *const u8,
+    ) -> SysResult {
+        info!(
+            "exec:BEG: path: {:?}, argv: {:?}, envp: {:?}",
+            path, argv, envp
+        );
+        let mut proc = self.process();
+        let path = unsafe { self.vm().check_and_clone_cstr(path)? };
+        let args = unsafe { self.vm().check_and_clone_cstr_array(argv)? };
+        let envs = unsafe { self.vm().check_and_clone_cstr_array(envp)? };
 
-/// Get the current thread id
-pub fn sys_gettid() -> SysResult {
-    info!("gettid");
-    // use pid as tid for now
-    Ok(thread::current().id())
-}
+        if args.is_empty() {
+            error!("exec: args is null");
+            return Err(SysError::EINVAL);
+        }
 
-/// Get the parent process id
-pub fn sys_getppid() -> SysResult {
-    if let Some(ref parent) = process().parent.as_ref() {
-        Ok(parent.lock().pid.get())
-    } else {
+        info!(
+            "exec:STEP2: path: {:?}, args: {:?}, envs: {:?}",
+            path, args, envs
+        );
+
+        // Kill other threads
+        proc.threads.retain(|&tid| {
+            if tid != processor().tid() {
+                processor().manager().exit(tid, 1);
+            }
+            tid == processor().tid()
+        });
+
+        // Read program file
+        let inode = proc.lookup_inode(&path)?;
+
+        // Make new Thread
+        let (mut vm, entry_addr, ustack_top) =
+            Thread::new_user_vm(&inode, &path, args, envs).map_err(|_| SysError::EINVAL)?;
+
+        // Activate new page table
+        core::mem::swap(&mut *self.vm(), &mut vm);
+        unsafe {
+            self.vm().activate();
+        }
+
+        // Modify exec path
+        proc.exec_path = path.clone();
+        drop(proc);
+
+        // Modify the TrapFrame
+        *self.tf = TrapFrame::new_user_thread(entry_addr, ustack_top);
+
+        info!("exec:END: path: {:?}", path);
         Ok(0)
     }
-}
 
-/// Exit the current thread
-pub fn sys_exit(exit_code: usize) -> ! {
-    let tid = thread::current().id();
-    info!("exit: {}, code: {}", tid, exit_code);
-    let mut proc = process();
-    proc.threads.retain(|&id| id != tid);
+    pub fn sys_yield(&mut self) -> SysResult {
+        thread::yield_now();
+        Ok(0)
+    }
 
-    // for last thread,
-    // notify parent and fill exit code
-    // avoid deadlock
-    let exit = proc.threads.len() == 0;
-    let proc_parent = proc.parent.clone();
-    let pid = proc.pid.get();
-    drop(proc);
-    if exit {
+    /// Kill the process
+    pub fn sys_kill(&mut self, pid: usize, sig: usize) -> SysResult {
+        info!(
+            "kill: {} killed: {} with sig {}",
+            thread::current().id(),
+            pid,
+            sig
+        );
+        let current_pid = self.process().pid.get().clone();
+        if current_pid == pid {
+            // killing myself
+            self.sys_exit_group(sig);
+        } else {
+            if let Some(proc_arc) = PROCESSES.read().get(&pid).and_then(|weak| weak.upgrade()) {
+                let proc = proc_arc.lock();
+                // quit all threads
+                for tid in proc.threads.iter() {
+                    processor().manager().exit(*tid, sig);
+                }
+                // notify parent and fill exit code
+                // avoid deadlock
+                let proc_parent = proc.parent.clone();
+                let pid = proc.pid.get();
+                drop(proc);
+                if let Some(parent) = proc_parent {
+                    let mut parent = parent.lock();
+                    parent.child_exit_code.insert(pid, sig);
+                    parent.child_exit.notify_one();
+                }
+                Ok(0)
+            } else {
+                Err(SysError::EINVAL)
+            }
+        }
+    }
+
+    /// Get the current process id
+    pub fn sys_getpid(&mut self) -> SysResult {
+        info!("getpid");
+        Ok(self.process().pid.get())
+    }
+
+    /// Get the current thread id
+    pub fn sys_gettid(&mut self) -> SysResult {
+        info!("gettid");
+        // use pid as tid for now
+        Ok(thread::current().id())
+    }
+
+    /// Get the parent process id
+    pub fn sys_getppid(&mut self) -> SysResult {
+        if let Some(parent) = self.process().parent.as_ref() {
+            Ok(parent.lock().pid.get())
+        } else {
+            Ok(0)
+        }
+    }
+
+    /// Exit the current thread
+    pub fn sys_exit(&mut self, exit_code: usize) -> ! {
+        let tid = thread::current().id();
+        info!("exit: {}, code: {}", tid, exit_code);
+        let mut proc = self.process();
+        proc.threads.retain(|&id| id != tid);
+
+        // for last thread,
+        // notify parent and fill exit code
+        // avoid deadlock
+        let exit = proc.threads.len() == 0;
+        let proc_parent = proc.parent.clone();
+        let pid = proc.pid.get();
+        drop(proc);
+        if exit {
+            if let Some(parent) = proc_parent {
+                let mut parent = parent.lock();
+                parent.child_exit_code.insert(pid, exit_code);
+                parent.child_exit.notify_one();
+            }
+        }
+
+        // perform futex wake 1
+        // ref: http://man7.org/linux/man-pages/man2/set_tid_address.2.html
+        // FIXME: do it in all possible ways a thread can exit
+        //        it has memory access so we can't move it to Thread::drop?
+        let mut proc = self.process();
+        let clear_child_tid = self.thread.clear_child_tid as *mut u32;
+        if !clear_child_tid.is_null() {
+            info!("exit: futex {:#?} wake 1", clear_child_tid);
+            if let Ok(clear_child_tid_ref) = unsafe { self.vm().check_write_ptr(clear_child_tid) } {
+                *clear_child_tid_ref = 0;
+                let queue = proc.get_futex(clear_child_tid as usize);
+                queue.notify_one();
+            }
+        }
+        drop(proc);
+
+        processor().manager().exit(tid, exit_code as usize);
+        processor().yield_now();
+        unreachable!();
+    }
+
+    /// Exit the current thread group (i.e. process)
+    pub fn sys_exit_group(&mut self, exit_code: usize) -> ! {
+        let proc = self.process();
+        info!("exit_group: {}, code: {}", proc.pid, exit_code);
+
+        // quit all threads
+        for tid in proc.threads.iter() {
+            processor().manager().exit(*tid, exit_code);
+        }
+
+        // notify parent and fill exit code
+        // avoid deadlock
+        let proc_parent = proc.parent.clone();
+        let pid = proc.pid.get();
+        drop(proc);
         if let Some(parent) = proc_parent {
             let mut parent = parent.lock();
             parent.child_exit_code.insert(pid, exit_code);
             parent.child_exit.notify_one();
         }
+
+        processor().yield_now();
+        unreachable!();
     }
 
-    // perform futex wake 1
-    // ref: http://man7.org/linux/man-pages/man2/set_tid_address.2.html
-    // FIXME: do it in all possible ways a thread can exit
-    //        it has memory access so we can't move it to Thread::drop?
-    let clear_child_tid = current_thread().clear_child_tid;
-    if clear_child_tid != 0 {
-        unsafe {
-            (clear_child_tid as *mut u32).write(0);
-        }
-        let queue = process().get_futex(clear_child_tid);
-        queue.notify_one();
+    pub fn sys_nanosleep(&mut self, req: *const TimeSpec) -> SysResult {
+        let time = unsafe { *self.vm().check_read_ptr(req)? };
+        info!("nanosleep: time: {:#?}", time);
+        // TODO: handle spurious wakeup
+        thread::sleep(time.to_duration());
+        Ok(0)
     }
 
-    processor().manager().exit(tid, exit_code as usize);
-    processor().yield_now();
-    unreachable!();
-}
-
-/// Exit the current thread group (i.e. process)
-pub fn sys_exit_group(exit_code: usize) -> ! {
-    let proc = process();
-    info!("exit_group: {}, code: {}", proc.pid, exit_code);
-
-    // quit all threads
-    for tid in proc.threads.iter() {
-        processor().manager().exit(*tid, exit_code);
+    pub fn sys_set_priority(&mut self, priority: usize) -> SysResult {
+        let pid = thread::current().id();
+        processor().manager().set_priority(pid, priority as u8);
+        Ok(0)
     }
 
-    // notify parent and fill exit code
-    // avoid deadlock
-    let proc_parent = proc.parent.clone();
-    let pid = proc.pid.get();
-    drop(proc);
-    if let Some(parent) = proc_parent {
-        let mut parent = parent.lock();
-        parent.child_exit_code.insert(pid, exit_code);
-        parent.child_exit.notify_one();
+    pub fn sys_set_tid_address(&mut self, tidptr: *mut u32) -> SysResult {
+        info!("set_tid_address: {:?}", tidptr);
+        self.thread.clear_child_tid = tidptr as usize;
+        Ok(thread::current().id())
     }
-
-    processor().yield_now();
-    unreachable!();
-}
-
-pub fn sys_nanosleep(req: *const TimeSpec) -> SysResult {
-    process().vm.check_read_ptr(req)?;
-    let time = unsafe { req.read() };
-    info!("nanosleep: time: {:#?}", time);
-    // TODO: handle spurious wakeup
-    thread::sleep(time.to_duration());
-    Ok(0)
-}
-
-pub fn sys_set_priority(priority: usize) -> SysResult {
-    let pid = thread::current().id();
-    processor().manager().set_priority(pid, priority as u8);
-    Ok(0)
 }
 
 bitflags! {

kernel/src/syscall/time.rs

@@ -5,6 +5,80 @@ use crate::consts::USEC_PER_TICK;
 use core::time::Duration;
 use lazy_static::lazy_static;
 
+impl Syscall<'_> {
+    pub fn sys_gettimeofday(&mut self, tv: *mut TimeVal, tz: *const u8) -> SysResult {
+        info!("gettimeofday: tv: {:?}, tz: {:?}", tv, tz);
+        if tz as usize != 0 {
+            return Err(SysError::EINVAL);
+        }
+
+        let tv = unsafe { self.vm().check_write_ptr(tv)? };
+
+        let timeval = TimeVal::get_epoch();
+        *tv = timeval;
+        Ok(0)
+    }
+
+    pub fn sys_clock_gettime(&mut self, clock: usize, ts: *mut TimeSpec) -> SysResult {
+        info!("clock_gettime: clock: {:?}, ts: {:?}", clock, ts);
+
+        let ts = unsafe { self.vm().check_write_ptr(ts)? };
+
+        let timespec = TimeSpec::get_epoch();
+        *ts = timespec;
+        Ok(0)
+    }
+
+    pub fn sys_time(&mut self, time: *mut u64) -> SysResult {
+        let sec = get_epoch_usec() / USEC_PER_SEC;
+        if time as usize != 0 {
+            let time = unsafe { self.vm().check_write_ptr(time)? };
+            *time = sec as u64;
+        }
+        Ok(sec as usize)
+    }
+
+    pub fn sys_getrusage(&mut self, who: usize, rusage: *mut RUsage) -> SysResult {
+        info!("getrusage: who: {}, rusage: {:?}", who, rusage);
+        let rusage = unsafe { self.vm().check_write_ptr(rusage)? };
+
+        let tick_base = *TICK_BASE;
+        let tick = unsafe { crate::trap::TICK as u64 };
+
+        let usec = (tick - tick_base) * USEC_PER_TICK as u64;
+        let new_rusage = RUsage {
+            utime: TimeVal {
+                sec: (usec / USEC_PER_SEC) as usize,
+                usec: (usec % USEC_PER_SEC) as usize,
+            },
+            stime: TimeVal {
+                sec: (usec / USEC_PER_SEC) as usize,
+                usec: (usec % USEC_PER_SEC) as usize,
+            },
+        };
+        *rusage = new_rusage;
+        Ok(0)
+    }
+
+    pub fn sys_times(&mut self, buf: *mut Tms) -> SysResult {
+        info!("times: buf: {:?}", buf);
+        let buf = unsafe { self.vm().check_write_ptr(buf)? };
+
+        let tick_base = *TICK_BASE;
+        let tick = unsafe { crate::trap::TICK as u64 };
+
+        let new_buf = Tms {
+            tms_utime: 0,
+            tms_stime: 0,
+            tms_cutime: 0,
+            tms_cstime: 0,
+        };
+
+        *buf = new_buf;
+        Ok(tick as usize)
+    }
+}
+
 /// should be initialized together
 lazy_static! {
     pub static ref EPOCH_BASE: u64 = crate::arch::timer::read_epoch();
@@ -76,47 +150,6 @@ impl TimeSpec {
     }
 }
 
-pub fn sys_gettimeofday(tv: *mut TimeVal, tz: *const u8) -> SysResult {
-    info!("gettimeofday: tv: {:?}, tz: {:?}", tv, tz);
-    if tz as usize != 0 {
-        return Err(SysError::EINVAL);
-    }
-
-    let proc = process();
-    proc.vm.check_write_ptr(tv)?;
-
-    let timeval = TimeVal::get_epoch();
-    unsafe {
-        *tv = timeval;
-    }
-    Ok(0)
-}
-
-pub fn sys_clock_gettime(clock: usize, ts: *mut TimeSpec) -> SysResult {
-    info!("clock_gettime: clock: {:?}, ts: {:?}", clock, ts);
-
-    let proc = process();
-    proc.vm.check_write_ptr(ts)?;
-
-    let timespec = TimeSpec::get_epoch();
-    unsafe {
-        *ts = timespec;
-    }
-    Ok(0)
-}
-
-pub fn sys_time(time: *mut u64) -> SysResult {
-    let sec = get_epoch_usec() / USEC_PER_SEC;
-    if time as usize != 0 {
-        let proc = process();
-        proc.vm.check_write_ptr(time)?;
-        unsafe {
-            time.write(sec as u64);
-        }
-    }
-    Ok(sec as usize)
-}
-
 // ignore other fields for now
 #[repr(C)]
 pub struct RUsage {
@@ -124,27 +157,13 @@ pub struct RUsage {
     stime: TimeVal,
 }
 
-pub fn sys_getrusage(who: usize, rusage: *mut RUsage) -> SysResult {
-    info!("getrusage: who: {}, rusage: {:?}", who, rusage);
-    let proc = process();
-    proc.vm.check_write_ptr(rusage)?;
-
-    let tick_base = *TICK_BASE;
-    let tick = unsafe { crate::trap::TICK as u64 };
-
-    let usec = (tick - tick_base) * USEC_PER_TICK as u64;
-    let new_rusage = RUsage {
-        utime: TimeVal {
-            sec: (usec / USEC_PER_SEC) as usize,
-            usec: (usec % USEC_PER_SEC) as usize,
-        },
-        stime: TimeVal {
-            sec: (usec / USEC_PER_SEC) as usize,
-            usec: (usec % USEC_PER_SEC) as usize,
-        },
-    };
-    unsafe { *rusage = new_rusage };
-    Ok(0)
+#[repr(C)]
+#[derive(Debug, Copy, Clone)]
+pub struct Tms {
+    tms_utime: u64,  /* user time */
+    tms_stime: u64,  /* system time */
+    tms_cutime: u64, /* user time of children */
+    tms_cstime: u64, /* system time of children */
 }
 
 

kernel/src/util/color.rs

@@ -1,3 +1,4 @@
+#[allow(dead_code)]
 #[derive(Debug, Clone, Copy, Eq, PartialEq)]
 pub enum ConsoleColor {
     Black,

kernel/src/util/escape_parser.rs

@@ -1,6 +1,8 @@
 //! ANSI escape sequences parser
 //! (ref: https://en.wikipedia.org/wiki/ANSI_escape_code)
 
+#![allow(dead_code)]
+
 use super::color::ConsoleColor;
 use heapless::consts::U8;
 use heapless::Vec;

tests/.gitignore

@@ -1 +1,2 @@
 stdout
+stdout.new

tests/test.sh

@@ -13,7 +13,9 @@ do
 
     wait $pid
 
-    diff -I 'bbl loader' -I 'Hello RISCV! in hart' -u ${f%.cmd}.out stdout || { echo 'testing failed for' $f; exit 1; }
+    awk 'NR > 25 { print }' < stdout > stdout.new
+
+    diff -u ${f%.cmd}.out stdout.new || { echo 'testing failed for' $f; exit 1; }
 
     echo testing $f pass
 done

tools/opensbi/README.md

@@ -2,8 +2,12 @@
 
 These are binary release of OpenSBI on this [commit](https://github.com/riscv/opensbi/tree/194dbbe5a13dff2255411c26d249f3ad4ef42c0b) at 2019.04.15.
 
-- fu540.elf: opensbi-0.3-rv64-bin/platform/sifive/fu540/firmware/fw_jump.elf
 - virt_rv32.elf: opensbi-0.3-rv32-bin/platform/qemu/virt/firmware/fw_jump.elf
 - virt_rv64.elf: opensbi-0.3-rv64-bin/platform/qemu/virt/firmware/fw_jump.elf
 
-NOTE: The [official v0.3 release](https://github.com/riscv/opensbi/releases/tag/v0.3) has bug on serial interrupt.
+NOTE: The [official v0.3 release](https://github.com/riscv/opensbi/releases/tag/v0.3) has bug on serial interrupt. Also, Rocket-Chip based CPUs (including SiFive Unleashed) seem to have unintended behavior on
+
+For K210 & SiFive Unleashed: It needs some modification. The binary is from this [commit](https://github.com/rcore-os/opensbi/commit/a9638d092756975ceb50073d736a17cef439c7b6).
+
+* k210.elf: build/platform/kendryte/k210/firmware/fw_payload.elf
+* fu540.elf: build/platform/sifive/fu540/firmware/fw_jump.elf

user

@@ -1 +1 @@
-Subproject commit 8dbc0edb935a62d748aaac39258d4a985de0ae17
+Subproject commit bb73d6ecce1ab0e6fae692c51e4335772b0335d4