Re: [patch net-next rfc 0/7] net: introduce alternative names for network interfaces

[Jiri Pirko <jiri@resnulli.us>]

Fri, Jul 19, 2019 at 06:31:35PM CEST, stephen@networkplumber.org wrote:
>On Fri, 19 Jul 2019 13:00:22 +0200
>Jiri Pirko <jiri@resnulli.us> wrote:
>
>> From: Jiri Pirko <jiri@mellanox.com>
>> 
>> In the past, there was repeatedly discussed the IFNAMSIZ (16) limit for
>> netdevice name length. Now when we have PF and VF representors
>> with port names like "pfXvfY", it became quite common to hit this limit:
>> 0123456789012345
>> enp131s0f1npf0vf6
>> enp131s0f1npf0vf22
>> 
>> Udev cannot rename these interfaces out-of-the-box and user needs to
>> create custom rules to handle them.
>> 
>> Also, udev has multiple schemes of netdev names. From udev code:
>>  * Type of names:
>>  *   b<number>                             - BCMA bus core number
>>  *   c<bus_id>                             - bus id of a grouped CCW or CCW device,
>>  *                                           with all leading zeros stripped [s390]
>>  *   o<index>[n<phys_port_name>|d<dev_port>]
>>  *                                         - on-board device index number
>>  *   s<slot>[f<function>][n<phys_port_name>|d<dev_port>]
>>  *                                         - hotplug slot index number
>>  *   x<MAC>                                - MAC address
>>  *   [P<domain>]p<bus>s<slot>[f<function>][n<phys_port_name>|d<dev_port>]
>>  *                                         - PCI geographical location
>>  *   [P<domain>]p<bus>s<slot>[f<function>][u<port>][..][c<config>][i<interface>]
>>  *                                         - USB port number chain
>>  *   v<slot>                               - VIO slot number (IBM PowerVM)
>>  *   a<vendor><model>i<instance>           - Platform bus ACPI instance id
>>  *   i<addr>n<phys_port_name>              - Netdevsim bus address and port name
>> 
>> One device can be often renamed by multiple patterns at the
>> same time (e.g. pci address/mac).
>> 
>> This patchset introduces alternative names for network interfaces.
>> Main goal is to:
>> 1) Overcome the IFNAMSIZ limitation
>> 2) Allow to have multiple names at the same time (multiple udev patterns)
>> 3) Allow to use alternative names as handle for commands
>> 
>> See following examples.
>> 
>> $ ip link
>> 1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN mode DEFAULT group default qlen 1000
>>     link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
>> 3: dummy0: <BROADCAST,NOARP> mtu 1500 qdisc noop state DOWN mode DEFAULT group default qlen 1000
>>     link/ether 7e:a2:d4:b8:91:7a brd ff:ff:ff:ff:ff:ff
>> 
>> -> Add alternative names for dummy0:  
>> 
>> $ ip link altname add dummy0 name someothername
>> $ ip link altname add dummy0 name someotherveryveryveryverylongname
>> $ ip link
>> 1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN mode DEFAULT group default qlen 1000
>>     link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
>> 3: dummy0: <BROADCAST,NOARP> mtu 1500 qdisc noop state DOWN mode DEFAULT group default qlen 1000
>>     link/ether 7e:a2:d4:b8:91:7a brd ff:ff:ff:ff:ff:ff
>>     altname someothername
>>     altname someotherveryveryveryverylongname
>>   
>> -> Add bridge brx, add it's alternative name and use alternative names to  
>>    do enslavement.
>> 
>> $ ip link add name brx type bridge
>> $ ip link altname add brx name mypersonalsuperspecialbridge
>> $ ip link set someotherveryveryveryverylongname master mypersonalsuperspecialbridge
>> $ ip link
>> 1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN mode DEFAULT group default qlen 1000
>>     link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
>> 3: dummy0: <BROADCAST,NOARP> mtu 1500 qdisc noop master brx state DOWN mode DEFAULT group default qlen 1000
>>     link/ether 7e:a2:d4:b8:91:7a brd ff:ff:ff:ff:ff:ff
>>     altname someothername
>>     altname someotherveryveryveryverylongname
>> 4: brx: <BROADCAST,MULTICAST> mtu 1500 qdisc noop state DOWN mode DEFAULT group default qlen 1000
>>     link/ether 7e:a2:d4:b8:91:7a brd ff:ff:ff:ff:ff:ff
>>     altname mypersonalsuperspecialbridge
>> 
>> -> Add ipv4 address to the bridge using alternative name:  
>>     
>> $ ip addr add 192.168.0.1/24 dev mypersonalsuperspecialbridge
>> $ ip addr show mypersonalsuperspecialbridge     
>> 4: brx: <BROADCAST,MULTICAST> mtu 1500 qdisc noop state DOWN group default qlen 1000
>>     link/ether 7e:a2:d4:b8:91:7a brd ff:ff:ff:ff:ff:ff
>>     altname mypersonalsuperspecialbridge
>>     inet 192.168.0.1/24 scope global brx
>>        valid_lft forever preferred_lft forever
>> 
>> -> Delete one of dummy0 alternative names:  
>> 
>> $ ip link altname del someotherveryveryveryverylongname    
>> $ ip link
>> 1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN mode DEFAULT group default qlen 1000
>>     link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
>> 3: dummy0: <BROADCAST,NOARP> mtu 1500 qdisc noop master brx state DOWN mode DEFAULT group default qlen 1000
>>     link/ether 7e:a2:d4:b8:91:7a brd ff:ff:ff:ff:ff:ff
>>     altname someothername
>> 4: brx: <BROADCAST,MULTICAST> mtu 1500 qdisc noop state DOWN mode DEFAULT group default qlen 1000
>>     link/ether 7e:a2:d4:b8:91:7a brd ff:ff:ff:ff:ff:ff
>>     altname mypersonalsuperspecialbridge
>> 
>> TODO:
>> - notifications for alternative names add/removal
>> - sanitization of add/del cmds (similar to get link)
>> - test more usecases and write selftests
>> - extend support for other netlink ifaces (ovs for example)
>> - add sysfs symlink altname->basename?
>> 
>> Jiri Pirko (7):
>>   net: procfs: use index hashlist instead of name hashlist
>>   net: introduce name_node struct to be used in hashlist
>>   net: rtnetlink: add commands to add and delete alternative ifnames
>>   net: rtnetlink: put alternative names to getlink message
>>   net: rtnetlink: unify the code in __rtnl_newlink get dev with the rest
>>   net: rtnetlink: introduce helper to get net_device instance by ifname
>>   net: rtnetlink: add possibility to use alternative names as message
>>     handle
>> 
>>  include/linux/netdevice.h      |  14 ++-
>>  include/uapi/linux/if.h        |   1 +
>>  include/uapi/linux/if_link.h   |   3 +
>>  include/uapi/linux/rtnetlink.h |   7 ++
>>  net/core/dev.c                 | 152 ++++++++++++++++++++++----
>>  net/core/net-procfs.c          |   4 +-
>>  net/core/rtnetlink.c           | 192 +++++++++++++++++++++++++++++----
>>  security/selinux/nlmsgtab.c    |   4 +-
>>  8 files changed, 334 insertions(+), 43 deletions(-)
>> 
>
>You might want to add altname/ object property in sysfs?
>I.e
>   /sys/class/net/eth0/altname/

Sysfs representation is one point on my TODO list.
I didn't look much into this, but yeah, that would work.
I also want to have symlinks altname->basename.

>