Can the grammar of the concatenation of two lists with common elements be written as an LALR grammar without any collisions

Question

Can the following ABNF rules be written as a bison LALR grammar without any conflicts (shift/reduce or reduce/reduce)? Among them, lowercase letters are non-terminal symbols, and uppercase letters are terminal symbols.

s = [*(A / D) D] *(A / B)

The key constraints are: A can appear anywhere, but any D must appear before any B (if any)

I wrote the following bison code, but it reports shift/reduce conflict

%token A
%token B
%token D

%start s

%%
s : %empty
  | a_d_d
  | s A
  | s B;

a_d_d : D
    | A a_d_d
    | D a_d_d;

I would like to be able to write a conflict-free LALR grammar, is this possible?