Translation of system calls

System call translation is the most important and complex part of proot-rs. This is because it can easily introduce bugs or behave inconsistently with the kernel.

We use unit tests and test_with_proot() to test if these system calls work correctly.

Current State of Translation

The following table describes the current state of translation of system calls. These system calls are organized by SyscallGroup, and system calls in the same SyscallGroup usually imply similar processing.

SyscallGroup	Syscall	Status	UnitTest Covered	Remarks
Execve	execve	✅	✅
Ptrace	ptrace	❌	❌
Wait	wait4	❌	❌
Brk	brk	❌	❌
GetCwd	getcwd	✅	✅
Chdir	fchdir, chdir	✅	✅
BindConnect	bind, connect	❌	❌
Accept	accept, accept4	❌	❌
GetSockOrPeerName	getsockname, getpeername	❌	❌
StandardSyscall	access, acct, chmod, chown, chroot, getxattr, listxattr, mknod, creat, removexattr, setxattr, stat, swapoff, swapon, truncate, umount2, uselib, utime, utimes	✅	✅	int syscall(const char *pathname, ...) follow symlink
Open	open	✅	✅	int syscall(const char *pathname, int flags, ...)
StatAt	fchownat, newfstatat, utimensat, name_to_handle_at, statx	✅	✅	int syscall(int dirfd, const char *pathname, ... , int flags, ...)
ChmodAccessMkNodAt	fchmodat, faccessat, futimesat, mknodat	✅	✅	int syscall(int dirfd, const char *pathname, ...) follow symlink
InotifyAddWatch	inotify_add_watch	❌	❌
DirLinkAttr	lchown, lgetxattr, llistxattr, lremovexattr, lsetxattr, lstat, unlink, rmdir, mkdir	✅	✅	int syscall(const char *pathname, ...) not follow symlink
PivotRoot	pivot_root	❌	❌
LinkAt	linkat	✅	✅
Mount	mount	❌	❌
OpenAt	openat	✅	✅
ReadLink	readlink	✅	✅
ReadLinkAt	readlinkat	✅	✅
UnlinkMkdirAt	unlinkat, mkdirat	✅	✅	int syscall(int dirfd, const char *pathname, ...) not follow symlink
Link	link	✅	✅
Rename	rename	✅	✅
RenameAt	renameat	✅	✅
SymLink	symlink	✅	✅
SymLinkAt	symlinkat	✅	✅
Uname	uname	❌	❌

Path Translation

One complication in path translation is whether we need to dereference/follow it if the last component of the path is a symbolic link.

In solving this problem, we found something worthy of reference in the kernel source code: https://elixir.bootlin.com/linux/latest/source/fs/namei.c#L97

/* [Feb-Apr 2000 AV] Complete rewrite. Rules for symlinks:
 *	inside the path - always follow.
 *	in the last component in creation/removal/renaming - never follow.
 *	if LOOKUP_FOLLOW passed - follow.
 *	if the pathname has trailing slashes - follow.
 *	otherwise - don't follow.
 * (applied in that order).
*/