Re: [PATCH v2] loongarch/bpf: Skip speculation barrier opcode, which caused ltp testcase bpf_prog02 to fail

From: WANG Xuerui
Date: Tue Mar 28 2023 - 03:53:00 EST


On 2023/3/28 15:22, Daniel Borkmann wrote:
On 3/28/23 9:13 AM, George Guo wrote:
Here just skip the opcode(BPF_ST | BPF_NOSPEC) that has no couterpart to the loongarch.

<snip>

diff --git a/arch/loongarch/net/bpf_jit.c b/arch/loongarch/net/bpf_jit.c
index 288003a9f0ca..d3c6b1c4ccbb 100644
--- a/arch/loongarch/net/bpf_jit.c
+++ b/arch/loongarch/net/bpf_jit.c
@@ -1022,6 +1022,11 @@ static int build_insn(const struct bpf_insn *insn, struct jit_ctx *ctx, bool ext
          emit_atomic(insn, ctx);
          break;
+    /* Speculation barrier */
+    case BPF_ST | BPF_NOSPEC:
+        pr_info_once("bpf_jit: skip speculation barrier opcode %0x2x\n", code);
+        break;

Thanks that looks better. Question to LoongArch folks (Cc): There is no equivalent
to a speculation barrier here, correct? Either way, I think the pr_info_once() can
just be removed given there is little value for a users to have this in the kernel
log. I can take care of this while applying, that's fine.

I can confirm there's currently no speculation barrier equivalent on lonogarch. (Loongson says there are builtin mitigations for Spectre-V1 and V2 on their chips, and AFAIK efforts to port the exploits to mips/loongarch have all failed a few years ago.)

And yes I'd agree with removing the warning altogether. Thanks for the reviews!

Acked-by: WANG Xuerui <git@xxxxxxxxxx>


      default:
          pr_err("bpf_jit: unknown opcode %02x\n", code);
          return -EINVAL;



--
WANG "xen0n" Xuerui

Linux/LoongArch mailing list: https://lore.kernel.org/loongarch/