netdev - Re: [PATCH net-next 5/5] sfc: add offloading of 'foreign' TC (decap) rules

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <b3ec1f69-eeb0-9c5c-7fdc-e520bd63385f@gmail.com>
Date:   Wed, 22 Mar 2023 22:35:44 +0000
From:   Edward Cree <ecree.xilinx@...il.com>
To:     Michal Swiatkowski <michal.swiatkowski@...ux.intel.com>,
        edward.cree@....com
Cc:     linux-net-drivers@....com, davem@...emloft.net, kuba@...nel.org,
        pabeni@...hat.com, edumazet@...gle.com, netdev@...r.kernel.org,
        habetsm.xilinx@...il.com
Subject: Re: [PATCH net-next 5/5] sfc: add offloading of 'foreign' TC (decap)
 rules

On 15/03/2023 14:43, Edward Cree wrote:
> On 15/03/2023 10:11, Michal Swiatkowski wrote:
>> On Tue, Mar 14, 2023 at 05:35:25PM +0000, edward.cree@....com wrote:
>>> +enum efx_encap_type efx_tc_indr_netdev_type(struct net_device *net_dev)
>>> +{
>>> +	if (netif_is_vxlan(net_dev))
>>> +		return EFX_ENCAP_TYPE_VXLAN;
>>> +	if (netif_is_geneve(net_dev))
>>> +		return EFX_ENCAP_TYPE_GENEVE;
>> netif_is_gretap or NVGRE isn't supported?
> 
> It should be supported, the hardware can handle it.
> I'll add it in v2, and test to make sure it actually works ;)

Fun discovery: while the hardware supports NVGRE, it can *only*
 match on the VSID field, not the whole GRE Key.
TC flower, meanwhile, neither knows nor cares about NVGRE; gretap
 decap rules expect to match on the full 32-bit Key field, and you
 can't even mask them (there's no TCA_FLOWER_KEY_ENC_KEY_ID_MASK
 in the uAPI), meaning the driver can't just require the FlowID is
 masked out and shift the rest.

So enabling this support is nontrivial; I've decided to leave it
 out of the series and just remove all mention of NVGRE for now.