Re: [PATCH/RFC net-next 3/9] nfp: provide infrastructure for offloading flower based TC filters

On Tue, Jun 27, 2017 at 11:13:06PM -0700, Jakub Kicinski wrote:

On Wed, 28 Jun 2017 01:21:43 +0200, Simon Horman wrote:

quoted

From: Pieter Jansen van Vuuren <redacted>

Adds a flower based TC offload handler for representor devices, this
is in addition to the bpf based offload handler. The changes in this
patch will be used in a follow-up patch to add tc flower offload to
the NFP.

The flower app enables tc offloads on representors by default.

Signed-off-by: Pieter Jansen van Vuuren <redacted>
Signed-off-by: Simon Horman <redacted>

quoted

diff --git a/drivers/net/ethernet/netronome/nfp/flower/main.c b/drivers/net/ethernet/netronome/nfp/flower/main.c
index ab68a8f58862..7b27871f489c 100644
--- a/drivers/net/ethernet/netronome/nfp/flower/main.c
+++ b/drivers/net/ethernet/netronome/nfp/flower/main.c

@@ -37,6 +37,7 @@
 #include <net/devlink.h>
 #include <net/dst_metadata.h>
 
+#include "main.h"
 #include "../nfpcore/nfp_cpp.h"
 #include "../nfpcore/nfp_nsp.h"
 #include "../nfp_app.h"

@@ -303,8 +304,14 @@ static int nfp_flower_vnic_init(struct nfp_app *app, struct nfp_net *nn,
 	eth_hw_addr_random(nn->dp.netdev);
 	netif_keep_dst(nn->dp.netdev);
 
+	if (nfp_flower_repr_init(app))
+		goto err_free_priv;

Please make the contents of nfp_flower_repr_init() part of app's .init
callback.

Thanks, I will fix this up and other comments relating to
nfp_flower_repr_init()

quoted

 	return 0;
 
+err_free_priv:
+	kfree(app->priv);
+	app->priv = NULL;

This doesn't belong here after my recent series.  priv init was moved
to app's init callback.

quoted

 err_invalid_port:
 	nn->port = nfp_port_alloc(app, NFP_PORT_INVALID, nn->dp.netdev);
 	return PTR_ERR_OR_ZERO(nn->port);

quoted

+int nfp_flower_repr_init(struct nfp_app *app)
+{
+	u64 version;
+	int err;
+
+	version = nfp_rtsym_read_le(app->pf->rtbl, "hw_flower_version", &err);
+	if (err)
+		return -EINVAL;

Nit: this could return err directly.  Also I think it's worth printing
an error message.

quoted

+	/* We need to ensure hardware has enough flower capabilities. */
+	if (version != NFP_FLOWER_ALLOWED_VER)
+		return -EINVAL;

Here we should definitely tell the user what went wrong.

quoted

+	return 0;
+}

quoted

diff --git a/drivers/net/ethernet/netronome/nfp/nfp_net_repr.c b/drivers/net/ethernet/netronome/nfp/nfp_net_repr.c
index bc9108071e5b..a73b311c1f75 100644
--- a/drivers/net/ethernet/netronome/nfp/nfp_net_repr.c
+++ b/drivers/net/ethernet/netronome/nfp/nfp_net_repr.c

@@ -250,6 +250,18 @@ static int nfp_repr_open(struct net_device *netdev)
 	return nfp_app_repr_open(repr->app, repr);
 }
 
+static int
+nfp_repr_setup_tc(struct net_device *netdev, u32 handle, u32 chain_index,
+		  __be16 proto, struct tc_to_netdev *tc)
+{
+	struct nfp_repr *repr = netdev_priv(netdev);
+
+	if (chain_index)
+		return -EOPNOTSUPP;
+
+	return nfp_app_setup_tc(repr->app, netdev, handle, proto, tc);
+}

Just a reminder that this could be a nfp_port function.

Sure, will do.