simple OS kernel

• physical page allocator
• operations on 2-level page table
• kernel part of page directory, UPAGES, kernel stack, kernel base

In JOS, we use environment denote the concept of process
In JOS, individual environments do not have their own kernel stacks, there can be only one JOS environment active in the kernel at a time, so JOS needs only a single kernel stack.

• create user environment, load ELF binary and run it
• set idt with trap handler, the trap handler function take control after cs:eip get new value.
• set up page fault(right now just destroy the environment that cause the fault)
• set up syscall with four sys function
• set up breakpoint exception(triggered by panic() and run monitor() func)

• multiprocessor support

• set pgdir for Memory-Mapped I/O
• set up MP entry code, kernel stack and TSS for different CPUs
• use a big kernel lock that only one environment can enter kernel mode
• implement Round-Robin scheduling
• add system calls for environment creation

• Copy-on-Write Fork
  • on fork() the kernel only copy the address space mappings not its contents and mark writable and COW pages as COW pages.
• Clock interrupt and IPC

• implement clock interrupts. In JOS, external device interrupts are always disabled when in the kernel and enabled when in user space. controlled by FL_IF flag bit of %eflags register.
• implement Inter-Process communication
  • two sys function sys_ipc_recv and sys_ipc_try_send. To receive a msg, an process calls sys_ipc_recv, the current process will be de-scheduled and doesn't run again until a msg has been received. Now any other process can send it a msg by calling sys_ipc_try_send. a 32-bit int and an optional page mapping can be transferred between two processes.

• File system Most UNIX file systems divide disk space into inode regions and data regions. In JOS, we don't have the concept of inode which means we also won't support hard links, symbolic links. JOS will simply store all of a file's meta-data within the (one and only) directory entry describing that file.

• enable disk access for fs env
• set up a 3GB block cache for fs env in a form of demand paging.
• add block operations
• link block number with virtual memory address
• implement file system interface

     Regular env           FS env
      +---------------+   +---------------+
      |      read     |   |   file_read   |
      |   (lib/fd.c)  |   |   (fs/fs.c)   |
   ...|.......|.......|...|.......^.......|...............
      |       v       |   |       |       | RPC mechanism
      |  devfile_read |   |  serve_read   |
      |  (lib/file.c) |   |  (fs/serv.c)  |
      |       |       |   |       ^       |
      |       v       |   |       |       |
      |     fsipc     |   |     serve     |
      |  (lib/file.c) |   |  (fs/serv.c)  |
      |       |       |   |       ^       |
      |       v       |   |       |       |
      |   ipc_send    |   |   ipc_recv    |
      |       |       |   |       ^       |
      +-------|-------+   +-------|-------+
          |                   |
          +-------------------+

• implement spawn processes and sharing ability across fork and spawn