Re: [PATCH net-next,v3 0/4] flow_offload: update mangle action representation

[Edward Cree <ecree@solarflare.com>]

On 06/09/2019 11:56, Pablo Neira Ayuso wrote:
> On Fri, Sep 06, 2019 at 11:02:18AM +0100, Edward Cree wrote:
>> Still NAK for the same reasons as three versions ago (when it was called
>>  "netfilter: payload mangling offload support"), you've never managed to
>>  explain why this extra API complexity is useful.  (Reducing LOC does not
>>  mean you've reduced complexity.)
> Oh well...
>
> Patch 1) Mask is inverted for no reason, just because tc pedit needs
> this in this way. All drivers reverse this mask.
>
> Patch 2) All drivers mask out meaningless fields in the value field.
To be clear: I have no issue with these two patches; they look fine to me.
(Though I'd like to see some comments on struct flow_action_entry specifying
 the semantics of these fields, especially if they're going to differ from
 the corresponding fields in struct tc_pedit_key.)

> Patch 3) Without this patchset, offsets are on the 32-bit boundary.
> Drivers need to play with the 32-bit mask to infer what field they are
> supposed to mangle... eg. with 32-bit offset alignment, checking if
> the use want to alter the ttl/hop_limit need for helper structures to
> check the 32-bit mask. Mangling a IPv6 address comes with one single
> action...
Drivers are still going to need to handle multiple pedit actions, in
 case the original rule wanted to mangle two non-consecutive fields.
And you can't just coalesce all consecutive mangles, because if you
 mangle two consecutive fields (e.g. UDP sport and dport) the driver
 still needs to disentangle that if it works on a 'fields' (rather
 than 'u32s') level.
So either have the core convert things into named protocol fields
 (i.e. "set src IPv6 to 1234::5 and add 1 to UDP sport"), or leave
 the current sequence-of-u32-mangles as it is.  This in-between "we'll
 coalesce things together despite not knowing what fields they are" is
 neither fish nor fowl.

-Ed