When error happens in dev_xdp_attach(), it should have a way to tell
users the error message like the netlink approach.

To avoid breaking uapi, adding a tracepoint in bpf_xdp_link_attach() is
an appropriate way to notify users the error message.

Hence, bpf libraries are able to retrieve the error message by this
tracepoint, and then report the error message to users.

Signed-off-by: Leon Hwang <hffilwlqm@gmail.com>
---
 include/trace/events/xdp.h | 17 +++++++++++++++++
 net/core/dev.c             |  5 ++++-
 2 files changed, 21 insertions(+), 1 deletion(-)

diff --git a/include/trace/events/xdp.h b/include/trace/events/xdp.h
index c40fc97f94171..cd89f1d5ce7b8 100644
--- a/include/trace/events/xdp.h
+++ b/include/trace/events/xdp.h
@@ -404,6 +404,23 @@ TRACE_EVENT(mem_return_failed,
 	)
 );
 
+TRACE_EVENT(bpf_xdp_link_attach_failed,
+
+	TP_PROTO(const char *msg),
+
+	TP_ARGS(msg),
+
+	TP_STRUCT__entry(
+		__string(msg, msg)
+	),
+
+	TP_fast_assign(
+		__assign_str(msg, msg);
+	),
+
+	TP_printk("errmsg=%s", __get_str(msg))
+);
+
 #endif /* _TRACE_XDP_H */
 
 #include <trace/define_trace.h>
diff --git a/net/core/dev.c b/net/core/dev.c
index 8e7d0cb540cdb..002fec07de739 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -133,6 +133,7 @@
 #include <trace/events/net.h>
 #include <trace/events/skb.h>
 #include <trace/events/qdisc.h>
+#include <trace/events/xdp.h>
 #include <linux/inetdevice.h>
 #include <linux/cpu_rmap.h>
 #include <linux/static_key.h>
@@ -9470,6 +9471,7 @@ int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog)
 {
 	struct net *net = current->nsproxy->net_ns;
 	struct bpf_link_primer link_primer;
+	struct netlink_ext_ack extack = {};
 	struct bpf_xdp_link *link;
 	struct net_device *dev;
 	int err, fd;
@@ -9497,12 +9499,13 @@ int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog)
 		goto unlock;
 	}
 
-	err = dev_xdp_attach_link(dev, NULL, link);
+	err = dev_xdp_attach_link(dev, &extack, link);
 	rtnl_unlock();
 
 	if (err) {
 		link->dev = NULL;
 		bpf_link_cleanup(&link_primer);
+		trace_bpf_xdp_link_attach_failed(extack._msg);
 		goto out_put_dev;
 	}
 
-- 
2.41.0

On Tue,  1 Aug 2023 22:26:20 +0800 Leon Hwang wrote:
> When error happens in dev_xdp_attach(), it should have a way to tell
> users the error message like the netlink approach.
> 
> To avoid breaking uapi, adding a tracepoint in bpf_xdp_link_attach() is
> an appropriate way to notify users the error message.
> 
> Hence, bpf libraries are able to retrieve the error message by this
> tracepoint, and then report the error message to users.

Whatevered-by: Jakub Kicinski <kuba@kernel.org> ?

On 2/8/23 06:08, Jakub Kicinski wrote:
> On Tue,  1 Aug 2023 22:26:20 +0800 Leon Hwang wrote:
>> When error happens in dev_xdp_attach(), it should have a way to tell
>> users the error message like the netlink approach.
>>
>> To avoid breaking uapi, adding a tracepoint in bpf_xdp_link_attach() is
>> an appropriate way to notify users the error message.
>>
>> Hence, bpf libraries are able to retrieve the error message by this
>> tracepoint, and then report the error message to users.
> 
> Whatevered-by: Jakub Kicinski <kuba@kernel.org> ?

Oh, it's not responsible for libraries to retrieve the error message. It
should be users' responsibility.

Sorry for the misguiding text.

In the previous patch[0], which changes uapi, it's able to back-propagate
the error message from dev_xdp_attach() to userspace through BPF syscall.
Then, tracepoint idea is suggested to avoid changing uapi. I do agree to
implement a tracepoint to avoid changing uapi.

[0] bpf: Introduce user log
https://lore.kernel.org/bpf/20230708040750.72570-1-hffilwlqm@gmail.com/

Thanks,
Leon