Re: [PATCH net-next v5 13/22] ethtool: provide driver/device information in GET_INFO request

[Jiri Pirko]

Wed, Mar 27, 2019 at 11:25:54PM CET, mkubecek@xxxxxxx wrote:
>On Wed, Mar 27, 2019 at 09:14:11PM +0100, Jiri Pirko wrote:
>> Mon, Mar 25, 2019 at 06:08:33PM CET, mkubecek@xxxxxxx wrote:
>> >+
>> >+Kernel response contents:
>> >+
>> >+    ETHA_INFO_DEV		(nested)	device identification
>> >+    ETHA_INFO_DRVINFO		(nested)	driver information
>> >+        ETHA_DRVINFO_DRIVER		(string)	driver name
>> >+        ETHA_DRVINFO_FWVERSION		(string)	firmware version
>> >+        ETHA_DRVINFO_BUSINFO		(string)	device bus address
>> >+        ETHA_DRVINFO_EROM_VER		(string)	expansion ROM version
>> 
>> These are already very nicely supported in devlink. No need to duplicate
>> here.
>
>They are supported by devlink as an interface. But devlink itself is
>only supported by few NIC drivers at the moment:
>
>mike@unicorn:~/work/git/net-next> grep -r devlink_ops drivers/net/
>drivers/net/ethernet/mellanox/mlx4/main.c:static const struct devlink_ops mlx4_devlink_ops = {
>drivers/net/ethernet/mellanox/mlx4/main.c:      devlink = devlink_alloc(&mlx4_devlink_ops, sizeof(*priv));
>drivers/net/ethernet/mellanox/mlxsw/core.c:static const struct devlink_ops mlxsw_devlink_ops = {
>drivers/net/ethernet/mellanox/mlxsw/core.c:             devlink = devlink_alloc(&mlxsw_devlink_ops, alloc_size);
>drivers/net/ethernet/mellanox/mlx5/core/main.c:static const struct devlink_ops mlx5_devlink_ops = {
>drivers/net/ethernet/mellanox/mlx5/core/main.c: devlink = devlink_alloc(&mlx5_devlink_ops, sizeof(*dev));
>drivers/net/ethernet/cavium/liquidio/lio_main.c:static const struct devlink_ops liquidio_devlink_ops = {
>drivers/net/ethernet/cavium/liquidio/lio_main.c:        devlink = devlink_alloc(&liquidio_devlink_ops,
>drivers/net/ethernet/broadcom/bnxt/bnxt_devlink.c:static const struct devlink_ops bnxt_dl_ops = {
>drivers/net/ethernet/netronome/nfp/nfp_main.c:  devlink = devlink_alloc(&nfp_devlink_ops, sizeof(*pf));
>drivers/net/ethernet/netronome/nfp/nfp_main.h:extern const struct devlink_ops nfp_devlink_ops;
>drivers/net/ethernet/netronome/nfp/nfp_devlink.c:const struct devlink_ops nfp_devlink_ops = {
>drivers/net/netdevsim/devlink.c:static const struct devlink_ops nsim_devlink_ops = {
>drivers/net/netdevsim/devlink.c:        devlink = devlink_alloc(&nsim_devlink_ops, 0);
>
>That's 6 drivers from 4 vendors (if I don't count netdevsim). And I did
>not check if all of them do actually provide the information shown
>above. On the other hand:
>
>mike@unicorn:~/work/git/net-next> egrep -r '\.get_drvinfo' drivers/net/ | wc -l
>240
>
>Some of these 240 lines assign the same handler but not enough to make
>me optimistic about being able to implement "ethtool -i <dev>" using
>devlink interface in near future (say few months or one year).
>
>I'm all for implementing new features which are are related to physical
>device (ASIC) rather than network interface only in devlink (at the
>level of kernel-userspace interface). But for features already provided
>by ethtool (userspace utility) I can't help seeing the state of devlink
>support in NIC drivers as a serious blocker.

What I'm thinking about at for some time now would be en explicit
default devlink and devlink_port registration for every netdev for
drivers that does not support devlink themselves. I need to think about
it more, but it seems doable. Then we can hang appropriate things there
and make the ethtoolnl/devlink separation clear. I believe we need to do
it.


>
>Michal