Re: [PATCH net-next 2/4] ixgbe: Remove driver specific fdb handlers.

On 03/01/2013 12:54 AM, John Fastabend wrote:

On 2/28/2013 3:13 PM, Vlad Yasevich wrote:

quoted

Remove driver specific fdb handlers since they are the same
as the default ones.

CC: John Fastabend <redacted>
CC: CC: Gregory Rose <redacted>
Signed-off-by: Vlad Yasevich <redacted>
---
  drivers/net/ethernet/intel/ixgbe/ixgbe_main.c |   81
-------------------------
  1 files changed, 0 insertions(+), 81 deletions(-)

diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c

b/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
index 68478d6..212b74e 100644

--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c

@@ -6998,84 +6998,6 @@ static int ixgbe_set_features(struct net_device

*netdev,
      return 0;
  }

-static int ixgbe_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
-                 struct net_device *dev,
-                 const unsigned char *addr,
-                 u16 flags)
-{

[...]

quoted

-    if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) {
-        u32 rar_uc_entries = IXGBE_MAX_PF_MACVLANS;
-
-        if (netdev_uc_count(dev) < rar_uc_entries)
-            err = dev_uc_add_excl(dev, addr);
-        else
-            err = -ENOMEM;

OK after talking with Greg I realized there is a problem with
just removing the add routine.

The problem with this is we lose the ability to indicate the
hardware table is full and return ENOMEM. Although our set_rx_mode
routine will put the device in promisc mode this won't forward
packets from a VF to the PF in the SR-IOV case.

So we either need to leave ixgbe_ndo_fdb_add() or let set_rx_mode
return an -ENOMEM error. I believe in the first case we can still
remove the dump and del cases.

Sorry for the hassle took me a second review to catch it.

No problem.  That was the reason I left it in the first version (because
of different error handling).  I'll restore the add function.

-vlad

.John