iousagetop not working due to net_if_receive_skb

[thburghout]

The current project uses net_dev_xmit and netif_receive_skb to trace network usage. However; my machine (4.19.0-6-amd64 #1 SMP Debian 4.19.67-2+deb10u1 (2019-09-20) x86_64 GNU/Linux) reports net_if_receive_skb.

I've solved this for my self by replacing all string occurrences of netif_receive_skb by net_if_receive_skb.

diff --git a/lttnganalyses/core/io.py b/lttnganalyses/core/io.py
index 1337ab9..bb7297a 100644
--- a/lttnganalyses/core/io.py
+++ b/lttnganalyses/core/io.py
@@ -36,7 +36,7 @@ class IoAnalysis(Analysis):
     def __init__(self, state, conf):
         notification_cbs = {
             'net_dev_xmit': self._process_net_dev_xmit,
-            'netif_receive_skb': self._process_netif_receive_skb,
+            'net_if_receive_skb': self._process_netif_receive_skb,
             'block_rq_complete': self._process_block_rq_complete,
             'io_rq_exit': self._process_io_rq_exit,
             'create_fd': self._process_create_fd,
diff --git a/lttnganalyses/linuxautomaton/net.py b/lttnganalyses/linuxautomaton/net.py
index 9517422..6249159 100644
--- a/lttnganalyses/linuxautomaton/net.py
+++ b/lttnganalyses/linuxautomaton/net.py
@@ -28,7 +28,7 @@ class NetStateProvider(sp.StateProvider):
     def __init__(self, state):
         cbs = {
             'net_dev_xmit': self._process_net_dev_xmit,
-            'netif_receive_skb': self._process_netif_receive_skb,
+            'net_if_receive_skb': self._process_netif_receive_skb,
         }
 
         super().__init__(state, cbs)
@@ -63,7 +63,7 @@ class NetStateProvider(sp.StateProvider):
                 proc.fds[fd].fd_type = sv.FDType.maybe_net
 
     def _process_netif_receive_skb(self, event):
-        self._state.send_notification_cb('netif_receive_skb',
+        self._state.send_notification_cb('net_if_receive_skb',
                                          iface_name=event['name'],
                                          recv_bytes=event['len'],
                                          cpu_id=event['cpu_id'])
diff --git a/tests/integration/trace_writer.py b/tests/integration/trace_writer.py
index dc69051..a949d9b 100644
--- a/tests/integration/trace_writer.py
+++ b/tests/integration/trace_writer.py
@@ -261,7 +261,7 @@ class TraceWriter():
         self.add_event(self.net_dev_xmit)
 
     def define_netif_receive_skb(self):
-        self.netif_receive_skb = CTFWriter.EventClass("netif_receive_skb")
+        self.netif_receive_skb = CTFWriter.EventClass("net_if_receive_skb")
         self.netif_receive_skb.add_field(self.uint64_type, "_skbaddr")
         self.netif_receive_skb.add_field(self.uint32_type, "_len")
         self.netif_receive_skb.add_field(self.string_type, "_name")

I think it would be desired to fix this in this repo as well. Is this solution sufficient? Would all functions need to be renamed as well?

[jgalar]

I'm not entirely sure what the policy is to handle those types of changes.

The notification_cbs should probably target both event names to preserve compatibility with older kernels.

As for the function names themselves, I don't think renaming them is necessary if both event names are targeted.

Maybe @eepp and/or @jdesfossez would care to comment here.