[PATCH v3] aarch64: vdso: Wire up getrandom() vDSO implementation

[Mark Rutland]

On Mon, Sep 02, 2024 at 04:41:28PM +0200, Jason A. Donenfeld wrote:
> On Mon, Sep 02, 2024 at 03:31:43PM +0100, Mark Rutland wrote:
> > ssize_t __kernel_getrandom(void *buffer, size_t len, unsigned int flags,
> > 			   void *opaque_state, size_t opaque_len)
> > {
> > 	if (alternative_has_cap_likely(ARM64_HAS_FPSIMD)) {
> > 		return __cvdso_getrandom(buffer, len, flags,
> > 					 opaque_state, opaque_len);
> > 	}
> > 
> > 	if (unlikely(opaque_len == ~0UL && !buffer && !len && !flags))
> > 		return -ENOSYS;
> > 	
> > 	return getrandom_syscall(buffer, len, flags);
> > }
> > 
> > ... though the conditions for returning -ENOSYS look very odd to me; why
> > do we care about fast-pathing that specific case rather than forwarding
> > that to the kernel, and does __cvdso_getrandom() handle that correctly?
> 
> Adhemerval's code here is fine and correct. The opaque_len==~0UL thing
> is a special vDSO case for getting the param struct back, not something
> related to the kernel. See __cvdso_getrandom_data() for details.

Ok, so this is to say "we cannot provide a vgetrandom_opaque_params".

Is the syscall fallback just for the CRIU case mentioned in
__cvdso_getrandom_data()? The comment above __cvdso_getrandom_data()
says:

  If @buffer, @len, and @flags are 0, and @opaque_len is ~0UL, then
  @opaque_state is populated with a struct vgetrandom_opaque_params and the
  function returns 0; if it does not return 0, this function should not be
  used.

... so presumably the caller shouldn't bother to call again if it got
-ENOSYS above.

Mark.