Re: [PATCH net-next 1/4] net/flow_dissector: add support for dissection of misc ip header fields

On Tue, May 30, 2017 at 6:10 AM, Jiri Pirko [off-list ref] wrote:

Sat, May 27, 2017 at 07:18:45PM CEST, tom@herbertland.com wrote:

quoted

On Sat, May 27, 2017 at 9:31 AM, Or Gerlitz [off-list ref] wrote:

quoted

On Thu, May 25, 2017 at 7:22 PM, Tom Herbert [off-list ref] wrote:

quoted

On Thu, May 25, 2017 at 6:24 AM, Or Gerlitz [off-list ref] wrote:

quoted

Add support for dissection of ip tos and ttl and ipv6 traffic-class
and hoplimit. Both are dissected into the same struct.

quoted

quoted

Uses similar call to ip dissection function as with tcp, arp and others.

quoted

quoted

+/**
+ * struct flow_dissector_key_ip:
+ * @tos: tos
+ * @ttl: ttl
+ */
+struct flow_dissector_key_ip {
+       __u8    tos;
+       __u8    ttl;
+};

--- a/net/core/flow_dissector.c
+++ b/net/core/flow_dissector.c

quoted

quoted

+static void
+__skb_flow_dissect_ipv4(const struct sk_buff *skb,
+                       struct flow_dissector *flow_dissector,
+                       void *target_container, void *data, const struct iphdr *iph)
+{
+       struct flow_dissector_key_ip *key_ip;
+
+       if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP))
+               return;
+
+       key_ip = skb_flow_dissector_target(flow_dissector,
+                                          FLOW_DISSECTOR_KEY_IP,
+                                          target_container);
+       key_ip->tos = iph->tos;
+       key_ip->ttl = iph->ttl;

In an encapsulation this returns the tos and ttl of the encapsulated
packet. Is that really useful to the caller? Seems more likely that
they need the outer tos and ttl for forwarding.

In what we are dealing with, classification is carried after the
packet is decapsulated by the shared tunnel device. So even today,e.g
for the src/dst IP, the dissection is carried on what were the inner
fields before decap.

Or,

I think the problem is I don't know what you're dealing with. The only
thing I can derive from the commit log is that tos and ttl are being
extracted, but I don't know why they are needed. I do know this is
adding complexity to an already overly complex function, and this
introduces new conditionals and code into the primary use case of
flow_dissector which is to create a key for deriving skb->hash. I
don't see that the cost of this patch has been justified.

Tom, we have been over this multiple times. The decision DaveM made at
the time I was pushing cls_flower was to have one shared dissection code
(I originally had a separate dissector inside cls_flower). And I
agree with that decision. It was a bit painful to work out the
flow_dissector in a generic way, but it was worth the efford.
is i
So when we need to dissect something new for cls_flower, we put it here.
flow_dissector is now miles away from being just a plain
"creator of the key to derive skb->hash".

Jiri,

The commit log does not mention cls_flower. There are no test results
given for the impact of the patch. There is no description why tos and
ttl are being added as opposed to id, flags, or IP options for that
matter.

Tom

Jiří