x86/bpf: Avoid emitting LOCK prefix for XCHG atomic ops

The x86 XCHG instruction is implicitly locked when one of the
operands is a memory location, making an explicit LOCK prefix
unnecessary.

Stop emitting the LOCK prefix for BPF_XCHG in the JIT atomic
read-modify-write helpers. This avoids redundant instruction
prefixes while preserving correct atomic semantics.

No functional change for other atomic operations.

Signed-off-by: Uros Bizjak <[email protected]>
Cc: Alexei Starovoitov <[email protected]>
Cc: Daniel Borkmann <[email protected]>
Cc: Andrii Nakryiko <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Borislav Petkov <[email protected]>
Cc: Dave Hansen <[email protected]>
Cc: "H. Peter Anvin" <[email protected]>

Author: ubizjak

arch/x86/net/bpf_jit_comp.c

@@ -1305,7 +1305,8 @@ static int emit_atomic_rmw(u8 **pprog, u32 atomic_op,
 {
 	u8 *prog = *pprog;
 
-	EMIT1(0xF0); /* lock prefix */
+	if (atomic_op != BPF_XCHG)
+		EMIT1(0xF0); /* lock prefix */
 
 	maybe_emit_mod(&prog, dst_reg, src_reg, bpf_size == BPF_DW);
 
@@ -1347,7 +1348,9 @@ static int emit_atomic_rmw_index(u8 **pprog, u32 atomic_op, u32 size,
 {
 	u8 *prog = *pprog;
 
-	EMIT1(0xF0); /* lock prefix */
+	if (atomic_op != BPF_XCHG)
+		EMIT1(0xF0); /* lock prefix */
+
 	switch (size) {
 	case BPF_W:
 		EMIT1(add_3mod(0x40, dst_reg, src_reg, index_reg));