Arceos posix api

[coolyjg]

• Add arceos posix_api layer to provide syscall api. This PR finishes work including:
  • Refactor framework: ulib/axlibc -> arceos-posix-api-> arceos modules/crates. Remove axstd dependency in axlibc.
  • Provide syscalls
  • ulib/axlibc should call syscalls to access kernel modules. And re-implemented some libc functions in pure Rust.
• Syscall IDs are similar to Linux syscall IDs. HOWEVER there are some arceos-specific syscalls (syscall ids exceed 500),
  which are not standard syscall, such as:
  • sys_pthread create (should be sys_clone)
  • sys_open (should be sys_openat)
  • sys_pthread self (should not be a syscall, but it accesses kernel struct).
• Nonstandard syscalls, like syscalls mentioned above, will be discussed and reconstructed in future work. This PR aims to refactor the framework of current ARCEOS.
• Things remains TODO:
  • Framework check: is this framework acceptable?
  • More precise debug information: axlibc layer vs arceos_posix_api layer
  • Re-implement more libc functions in Rust.
  • Redundant c to rust binding file in both axlibc and arceos_posix_api.
  • Where to put STDIN/STDOUT and how to handle fd_table with fd = 0, 1, 2.

[equation314]

1. I think syscall ID is unnecessary, arceos_posix_api is API compatible rather than ABI compatible. e.g., call sys_pthread_create in pthread_create directly instead of sys_clone.
2. We should not put any debug info in axlibc (Do you see any debug output from libc?).
3. May be in next PR.
4. We can export C structures in arceos_posix_api so that axlibc can access them without C to Rust bindings.
5. Put stdin/stdout into arceos_posix_api and make sys_read/sys_write with fd 0/1/2 work fine.

[coolyjg]

Changes:

1. Remove syscall/syscall_id related code.
2. Remove c to rust bingdings in axlibc.
3. Move stdin/stdout to arceos-posix-api layer, remove print_str (call write), keep println_str (use Mutex)
4. Why change write to ax_write again: I found that if I keep a function named write in rust packages, this package cannot pass cargo test (caused signal: 11, SIGSEGV: invalid memory reference), which fails the unit-test. And relibc encounters the same thing as well. Maybe cargo test uses this write to print messages?

[equation314]

4. Why change write to ax_write again: I found that if I keep a function named write in rust packages, this package cannot pass cargo test (caused signal: 11, SIGSEGV: invalid memory reference), which fails the unit-test. And relibc encounters the same thing as well. Maybe cargo test uses this write to print messages?

You can add #[cfg(not(test))] above pub unsafe extern "C" fn write to resolve it.

[equation314]

Good job! Please organize commits.

Thanks for your contribution~