On Tue, Feb 19, 2019 at 11:24:00AM +0100, Jiri Pirko wrote:

Mon, Feb 18, 2019 at 07:22:24PM CET, mkubecek@suse.cz wrote:

quoted

Add information about permanent hadware address of a device (as provided by
ETHTOOL_GPERMADDR ioctl command) in GET_INFO reply if ETH_INFO_IM_PERMADDR
flag is set in the request.

There is no separate attribute for hardware address length as nla_len gives
this information. The reply also provides address type (net_device::type).

Signed-off-by: Michal Kubecek <redacted>
---
Documentation/networking/ethtool-netlink.txt |  9 ++++-
include/uapi/linux/ethtool_netlink.h         | 12 +++++-
net/ethtool/info.c                           | 39 ++++++++++++++++++++
3 files changed, 58 insertions(+), 2 deletions(-)

diff --git a/Documentation/networking/ethtool-netlink.txt b/Documentation/networking/ethtool-netlink.txt
index b6999a2167e8..1e615e111262 100644
--- a/Documentation/networking/ethtool-netlink.txt
+++ b/Documentation/networking/ethtool-netlink.txt

@@ -239,6 +239,9 @@ Kernel response contents:
        ETHA_DRVINFO_FWVERSION		(string)	firmware version
        ETHA_DRVINFO_BUSINFO		(string)	device bus address
        ETHA_DRVINFO_EROM_VER		(string)	expansion ROM version
+    ETHA_INFO_PERMADDR		(nested)
+        ETHA_PERMADDR_ADDRESS		(binary)	permanent HW address

I think this is a nice example of thing that should not be exposed with
ethtool but rather via rtnetlink, alongside with the actual hw address.

[...]

I guess you are right. As we don't have to query the driver and just
read the information from struct net_device, rtnetlink does indeed seem
more appropriate.

Michal