Re: [PATCH net-next 2/2] bpf: let bpf_warn_invalid_xdp_action() report more info

On Mon, 2021-11-22 at 12:32 +0100, Toke Høiland-Jørgensen wrote:

You still have:

quoted

diff --git a/kernel/bpf/cpumap.c b/kernel/bpf/cpumap.c
index 585b2b77ccc4..f7359bcb8fa3 100644
--- a/kernel/bpf/cpumap.c
+++ b/kernel/bpf/cpumap.c

@@ -195,7 +195,7 @@ static void cpu_map_bpf_prog_run_skb(struct bpf_cpu_map_entry *rcpu,
 			}
 			return;
 		default:
-			bpf_warn_invalid_xdp_action(act);
+			bpf_warn_invalid_xdp_action(skb->dev, rcpu->prog, act);
 			fallthrough;
 		case XDP_ABORTED:
 			trace_xdp_exception(skb->dev, rcpu->prog, act);

@@ -254,7 +254,7 @@ static int cpu_map_bpf_prog_run_xdp(struct bpf_cpu_map_entry *rcpu,
 			}
 			break;
 		default:
-			bpf_warn_invalid_xdp_action(act);
+			bpf_warn_invalid_xdp_action(xdpf->dev_rx, rcpu->prog, act);
 			fallthrough;
 		case XDP_DROP:
 			xdp_return_frame(xdpf);

diff --git a/kernel/bpf/devmap.c b/kernel/bpf/devmap.c
index f02d04540c0c..79bcf2169881 100644
--- a/kernel/bpf/devmap.c
+++ b/kernel/bpf/devmap.c

@@ -348,7 +348,7 @@ static int dev_map_bpf_prog_run(struct bpf_prog *xdp_prog,
 				frames[nframes++] = xdpf;
 			break;
 		default:
-			bpf_warn_invalid_xdp_action(act);
+			bpf_warn_invalid_xdp_action(dev, xdp_prog, act);
 			fallthrough;
 		case XDP_ABORTED:
 			trace_xdp_exception(dev, xdp_prog, act);

@@ -507,7 +507,7 @@ static u32 dev_map_bpf_prog_run_skb(struct sk_buff *skb, struct bpf_dtab_netdev
 		__skb_push(skb, skb->mac_len);
 		break;
 	default:
-		bpf_warn_invalid_xdp_action(act);
+		bpf_warn_invalid_xdp_action(dst->dev, dst->xdp_prog, act);
 		fallthrough;
 	case XDP_ABORTED:
 		trace_xdp_exception(dst->dev, dst->xdp_prog, act);

... and ...

quoted

diff --git a/net/core/filter.c b/net/core/filter.c
index 3ba584bb23f8..658f7a84d9bc 100644
--- a/net/core/filter.c
+++ b/net/core/filter.c

@@ -8179,13 +8179,13 @@ static bool xdp_is_valid_access(int off, int size,
 	return __is_valid_xdp_access(off, size);
 }
 
-void bpf_warn_invalid_xdp_action(u32 act)
+void bpf_warn_invalid_xdp_action(struct net_device *dev, struct bpf_prog *prog, u32 act)
 {
 	const u32 act_max = XDP_REDIRECT;
 
-	pr_warn_once("%s XDP return value %u, expect packet loss!\n",
+	pr_warn_once("%s XDP return value %u on prog %s (id %d) dev %s, expect packet loss!\n",
 		     act > act_max ? "Illegal" : "Driver unsupported",
-		     act);
+		     act, prog->aux->name, prog->aux->id, dev->name);
 }
 EXPORT_SYMBOL_GPL(bpf_warn_invalid_xdp_action);

This will still print the dev name for cpumap and devmap programs, which
is misleading - it will have people looking at the drivers when the
problem is somewhere else.

I pointed this out multiple times in comments on your last revision so
I'm a bit puzzled as to why you're still doing this? :/

I'm sorry, quite a bit of relevant incoming messages keep landing in
the spam folder due to corporate/gmail filters.

Just pass NULL as the dev from cpumap/devmap and don't print the dev
name in this case...

No opposition on my side, but I think that is a bit conflicting with
Alexei suggestion of keeping the message handling as simple as
possible?!?

@Alexei, would something like the following:

	pr_warn_once("%s XDP return value %u on prog %s (id %d) dev %s, expect packet loss!\n",
  		     act > act_max ? "Illegal" : "Driver unsupported",
		     act, prog->aux->name, prog->aux->id, dev ? dev->name, "");

fit you?

Thanks!

Paolo