On Wed, May 15, 2019 at 01:03:31AM +0200, Pablo Neira Ayuso wrote:
> On Tue, May 14, 2019 at 07:01:08PM +0200, Jiri Pirko wrote:
> > Thu, May 09, 2019 at 06:39:51PM CEST, pablo@xxxxxxxxxxxxx wrote:
> > >This patch adds hardware offload support for nftables through the
> > >existing netdev_ops->ndo_setup_tc() interface, the TC_SETUP_CLSFLOWER
> > >classifier and the flow rule API. This hardware offload support is
> > >available for the NFPROTO_NETDEV family and the ingress hook.
> > >
> > >Each nftables expression has a new ->offload interface, that is used to
> > >populate the flow rule object that is attached to the transaction
> > >object.
> > >
> > >There is a new per-table NFT_TABLE_F_HW flag, that is set on to offload
> > >an entire table, including all of its chains.
> > >
> > >This patch supports for basic metadata (layer 3 and 4 protocol numbers),
> > >5-tuple payload matching and the accept/drop actions; this also includes
> > >basechain hardware offload only.
> > >
> > >Signed-off-by: Pablo Neira Ayuso <pablo@xxxxxxxxxxxxx>
> >
> > [...]
> >
> > >+static int nft_flow_offload_chain(struct nft_trans *trans,
> > >+ enum flow_block_command cmd)
> > >+{
> > >+ struct nft_chain *chain = trans->ctx.chain;
> > >+ struct netlink_ext_ack extack = {};
> > >+ struct flow_block_offload bo = {};
> > >+ struct nft_base_chain *basechain;
> > >+ struct net_device *dev;
> > >+ int err;
> > >+
> > >+ if (!nft_is_base_chain(chain))
> > >+ return -EOPNOTSUPP;
> > >+
> > >+ basechain = nft_base_chain(chain);
> > >+ dev = basechain->ops.dev;
> > >+ if (!dev)
> > >+ return -EOPNOTSUPP;
> > >+
> > >+ bo.command = cmd;
> > >+ bo.binder_type = TCF_BLOCK_BINDER_TYPE_CLSACT_INGRESS;
> > >+ bo.block_index = (u32)trans->ctx.chain->handle;
> > >+ bo.extack = &extack;
> > >+ INIT_LIST_HEAD(&bo.cb_list);
> > >+
> > >+ err = dev->netdev_ops->ndo_setup_tc(dev, TC_SETUP_BLOCK, &bo);
> >
> > Okay, so you pretend to be clsact-ingress-flower. That looks fine.
> > But how do you ensure that the real one does not bind a block on the
> > same device too?
>
> I could store the interface index in the block_cb object, then use the
> tuple [ cb, cb_ident, ifindex ] to check if the block is already bound
> by when flow_block_cb_alloc() is called.
Actually cb_ident would be sufficient. One possibility would be to
extend flow_block_cb_alloc() to check for an existing binding.
diff --git a/net/core/flow_offload.c b/net/core/flow_offload.c
index cf984ef05609..44172014cebe 100644
--- a/net/core/flow_offload.c
+++ b/net/core/flow_offload.c
@@ -193,9 +193,15 @@ struct flow_block_cb *flow_block_cb_alloc(u32 block_index, tc_setup_cb_t *cb,
{
struct flow_block_cb *block_cb;
+ list_for_each_entry(block_cb, &flow_block_cb_list, list) {
+ if (block_cb->cb == cb &&
+ block_cb->cb_ident == cb_ident)
+ return ERR_PTR(-EBUSY);
+ }
+
block_cb = kzalloc(sizeof(*block_cb), GFP_KERNEL);
if (!block_cb)
- return NULL;
+ return ERR_PTR(-ENOMEM);
block_cb->cb = cb;
block_cb->cb_ident = cb_ident;
Thanks.
