Skip to contentSkip to conversation

netfilter: nf_tables: ignore -EOPNOTSUPP on flowtable device offload setup

RFC In review Felix Fietkau 2 participants 5 messages Aug 31, 2023 – Sep 1, 2023

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

RFC 1/1 netfilter: nf_tables: ignore -EOPNOTSUPP on flowtable device offload setup

On many embedded devices, it is common to configure flowtable offloading for a mix of different devices, some of which have hardware offload support and some of which don't.
The current code limits the ability of user space to properly set up such a configuration by only allowing adding devices with hardware offload support to a offload-enabled flowtable.
Given that offload-enabled flowtables also imply fallback to pure software offloading, this limitation makes little sense.
Fix it by not bailing out when the offload setup returns -EOPNOTSUPP

---
 net/netfilter/nf_tables_api.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
net/netfilter/nf_tables_api.c +1 -1
index 41b826dff6f5..dfa2ea98088b 100644
@@ -8103,7 +8103,7 @@ static int nft_register_flowtable_net_hooks(struct net *net,
 		err = flowtable->data.type->setup(&flowtable->data,
 						  hook->ops.dev,
 						  FLOW_BLOCK_BIND);
-		if (err < 0)
+		if (err < 0 && err != -EOPNOTSUPP)
 			goto err_unregister_net_hooks;
 
 		err = nf_register_net_hook(net, &hook->ops);
Signature · 1 line
2.41.0
#1 lore raw

Hi Felix,

Quoted 9 lines · Felix Fietkau

On many embedded devices, it is common to configure flowtable offloading for a mix of different devices, some of which have hardware offload support and some of which don't.
The current code limits the ability of user space to properly set up such a configuration by only allowing adding devices with hardware offload support to a offload-enabled flowtable.
Given that offload-enabled flowtables also imply fallback to pure software offloading, this limitation makes little sense.
Fix it by not bailing out when the offload setup returns -EOPNOTSUPP

Would you send a v2 to untoggle the offload flag when listing the ruleset if EOPNOTSUPP is reported? Thus, the user knows that no hardware offload is being used.

Quoted 21 lines
---
 net/netfilter/nf_tables_api.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
net/netfilter/nf_tables_api.c +1 -1
index 41b826dff6f5..dfa2ea98088b 100644
@@ -8103,7 +8103,7 @@ static int nft_register_flowtable_net_hooks(struct net *net,
 		err = flowtable->data.type->setup(&flowtable->data,
 						  hook->ops.dev,
 						  FLOW_BLOCK_BIND);
-		if (err < 0)
+		if (err < 0 && err != -EOPNOTSUPP)
 			goto err_unregister_net_hooks;
 
 		err = nf_register_net_hook(net, &hook->ops);
Signature · 1 line
2.41.0
#2 lore raw
Quoted 13 lines · Pablo Neira Ayuso

Hi Felix,

Quoted 9 lines · Felix Fietkau

On many embedded devices, it is common to configure flowtable offloading for a mix of different devices, some of which have hardware offload support and some of which don't.
The current code limits the ability of user space to properly set up such a configuration by only allowing adding devices with hardware offload support to a offload-enabled flowtable.
Given that offload-enabled flowtables also imply fallback to pure software offloading, this limitation makes little sense.
Fix it by not bailing out when the offload setup returns -EOPNOTSUPP

Quoted 3 lines

Would you send a v2 to untoggle the offload flag when listing the ruleset if EOPNOTSUPP is reported? Thus, the user knows that no hardware offload is being used.

Wouldn't that mess up further updates to the flowtable? From what I can tell, when updating a flow table, changing its offload flag is not supported.

- Felix

#3 lore raw
Quoted 14 lines · Felix Fietkau
Quoted 13 lines · Pablo Neira Ayuso

Hi Felix,

Quoted 9 lines · Felix Fietkau

On many embedded devices, it is common to configure flowtable offloading for a mix of different devices, some of which have hardware offload support and some of which don't.
The current code limits the ability of user space to properly set up such a configuration by only allowing adding devices with hardware offload support to a offload-enabled flowtable.
Given that offload-enabled flowtables also imply fallback to pure software offloading, this limitation makes little sense.
Fix it by not bailing out when the offload setup returns -EOPNOTSUPP

Quoted 7 lines
Quoted 3 lines

Would you send a v2 to untoggle the offload flag when listing the ruleset if EOPNOTSUPP is reported? Thus, the user knows that no hardware offload is being used.

Wouldn't that mess up further updates to the flowtable? From what I can tell, when updating a flow table, changing its offload flag is not supported.

The flag would be untoggled if hardware offload is not supported. What problematic scenario are you having in mind that might break?

In any case, there is a need to provide a way to tell the user if the hardware offload is actually happening or not, if not what I suggest, then propose a different way. But user really needs to know if it runs software or hardware plane to debug issues.

#4 lore raw
Quoted 19 lines · Pablo Neira Ayuso
Quoted 18 lines · Felix Fietkau
Quoted 16 lines · Pablo Neira Ayuso

Hi Felix,

Quoted 9 lines · Felix Fietkau

On many embedded devices, it is common to configure flowtable offloading for a mix of different devices, some of which have hardware offload support and some of which don't.
The current code limits the ability of user space to properly set up such a configuration by only allowing adding devices with hardware offload support to a offload-enabled flowtable.
Given that offload-enabled flowtables also imply fallback to pure software offloading, this limitation makes little sense.
Fix it by not bailing out when the offload setup returns -EOPNOTSUPP

Would you send a v2 to untoggle the offload flag when listing the ruleset if EOPNOTSUPP is reported? Thus, the user knows that no hardware offload is being used.

Quoted 6 lines
Quoted 3 lines

Wouldn't that mess up further updates to the flowtable? From what I can tell, when updating a flow table, changing its offload flag is not supported.

The flag would be untoggled if hardware offload is not supported. What problematic scenario are you having in mind that might break?

The scenario I'm thinking about is this:
Initially, the flowtable is created with a set of devices which don't support offload.
Afterwards, the flowtable gets updated with the intention of adding an extra device which *does* support hw offload to the existing flowtable. If the flag was cleared after initially creating the table, I think the update would fail. Or did I misread the code?

Quoted 4 lines

In any case, there is a need to provide a way to tell the user if the hardware offload is actually happening or not, if not what I suggest, then propose a different way. But user really needs to know if it runs software or hardware plane to debug issues.

In my opinion, a single flag indication for the flow table is mostly useless. Much more useful would be if you could query which of the devices that were added to the flowtable support hw offload and which ones don't. That requires some API changes though, and I don't think that should be done in this patch.

- Felix