Linux: change syscalls?

Question

Syscalls is an interface that accesses kernel user space. The user process usually does not call them directly, but uses libc to do this. libc or simply provides a thin wrapper around the script, or performs some work, as in the case fork()andexec()

So my question is: has the syscall kernel interface ever changed between kernel versions in non-backward compatible mode? Or is it once installed syscall, will it never change?

Answer 1

A syscall in terms of application code is an atomic atomic operation (this is a “virtual machine instruction”). See also syscalls (2) .

ABI determines how this happens. For x86-64 you can find here . See also x86 calling conventions .

For some system calls, the latest kernels provide vdso (7) to speed up their execution.

The Linux kernel is trying to maintain binary compatibility. Rumor has it that the 15-year-old (statically linked) ELF executable should work unchanged in the latest kernels.

( , . Linux howto). libc (, GNU libc MUSL libc). libc.so. - ( , libc, ).

Answer 2

:

• libc, libc - libc, .
• Syscalls , libc.

libc fork() exec() , , . , system(), fork-s exec .

syscalls , Linus , , syscalls, _reiser4..., .

, sys_open, sys_close .., .

---

, , , , . ancients linux/i386 int 0x80, syscals eax. / , , asm-opcode, SYSENTER, . , , ​​ , syscalls. cat /proc/$$/maps, [vdso] [vsyscall], . (, 2.4?) .