lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date:   Mon, 12 Aug 2019 00:10:27 +0200
From:   Michal Kubecek <mkubecek@...e.cz>
To:     netdev@...r.kernel.org
Cc:     Roopa Prabhu <roopa@...ulusnetworks.com>,
        Jiri Pirko <jiri@...nulli.us>,
        David Miller <davem@...emloft.net>,
        Jakub Kicinski <jakub.kicinski@...ronome.com>,
        Stephen Hemminger <sthemmin@...rosoft.com>,
        David Ahern <dsahern@...il.com>, dcbw@...hat.com,
        Andrew Lunn <andrew@...n.ch>, parav@...lanox.com,
        Saeed Mahameed <saeedm@...lanox.com>,
        mlxsw <mlxsw@...lanox.com>
Subject: Re: [patch net-next rfc 3/7] net: rtnetlink: add commands to add and
 delete alternative ifnames

On Sat, Aug 10, 2019 at 12:39:31PM -0700, Roopa Prabhu wrote:
> On Sat, Aug 10, 2019 at 8:50 AM Michal Kubecek <mkubecek@...e.cz> wrote:
> >
> > On Sat, Aug 10, 2019 at 06:46:57AM -0700, Roopa Prabhu wrote:
> > > On Fri, Aug 9, 2019 at 8:46 AM Michal Kubecek <mkubecek@...e.cz> wrote:
> > > >
> > > > On Fri, Aug 09, 2019 at 08:40:25AM -0700, Roopa Prabhu wrote:
> > > > > to that point, I am also not sure why we have a new API For multiple
> > > > > names. I mean why support more than two names  (existing old name and
> > > > > a new name to remove the length limitation) ?
> > > >
> > > > One use case is to allow "predictable names" from udev/systemd to work
> > > > the way do for e.g. block devices, see
> > > >
> > > >   http://lkml.kernel.org/r/20190628162716.GF29149@unicorn.suse.cz
> > > >
> > >
> > > thanks for the link. don't know the details about alternate block
> > > device names. Does user-space generate multiple and assign them to a
> > > kernel object as proposed in this series ?. is there a limit to number
> > > of names ?. my understanding of 'predictable names' was still a single
> > > name but predictable structure to the name.
> >
> > It is a single name but IMHO mostly because we can only have one name.
> > For block devices, udev uses symlinks to create multiple aliases based
> > on different naming schemes, e.g.
> >
> > mike@...n:~> find -L /dev/disk/ -samefile /dev/sda2 -exec ls -l {} +
> > lrwxrwxrwx 1 root root 10 srp  5 21:47 /dev/disk/by-id/ata-WDC_WD30EFRX-68AX9N0_WD-WMC1T3114933-part2 -> ../../sda2
> > lrwxrwxrwx 1 root root 10 srp  5 21:47 /dev/disk/by-id/scsi-SATA_WDC_WD30EFRX-68A_WD-WMC1T3114933-part2 -> ../../sda2
> > lrwxrwxrwx 1 root root 10 srp  5 21:47 /dev/disk/by-id/scsi-SATA_WDC_WD30EFRX-68_WD-WMC1T3114933-part2 -> ../../sda2
> > lrwxrwxrwx 1 root root 10 srp  5 21:47 /dev/disk/by-id/scsi-0ATA_WDC_WD30EFRX-68A_WD-WMC1T3114933-part2 -> ../../sda2
> > lrwxrwxrwx 1 root root 10 srp  5 21:47 /dev/disk/by-id/scsi-1ATA_WDC_WD30EFRX-68AX9N0_WD-WMC1T3114933-part2 -> ../../sda2
> > lrwxrwxrwx 1 root root 10 srp  5 21:47 /dev/disk/by-id/scsi-350014ee6589cfea0-part2 -> ../../sda2
> > lrwxrwxrwx 1 root root 10 srp  5 21:47 /dev/disk/by-id/wwn-0x50014ee6589cfea0-part2 -> ../../sda2
> > lrwxrwxrwx 1 root root 10 srp  5 21:47 /dev/disk/by-partlabel/root2 -> ../../sda2
> > lrwxrwxrwx 1 root root 10 srp  5 21:47 /dev/disk/by-partuuid/71affb47-a93b-40fd-8986-d2e227e1b39d -> ../../sda2
> > lrwxrwxrwx 1 root root 10 srp  5 21:47 /dev/disk/by-path/pci-0000:00:11.0-ata-1-part2 -> ../../sda2
> > lrwxrwxrwx 1 root root 10 srp  5 21:47 /dev/disk/by-path/pci-0000:00:11.0-scsi-0:0:0:0-part2 -> ../../sda2
> >
> > Few years ago, udev even dropped support for renaming block and
> > character devices (NAME="...") so that it now keeps kernel name and only
> > creates symlinks to it. Recent versions only allow NAME="..." for
> > network devices.
> 
> ok thanks for the details. This looks like names that are structured
> on hardware info which could fall into devlinks scope and they point
> to a single name.
> We should think about keeping them under devlink (by-id, by-mac etc).
> It already can recognize network interfaces by id.

Not all of them are hardware based, there are also links based on
filesystem label or UUID. But my point is rather that udev creates
multiple links so that any of them can be used in any place where
a block device is to be identified.

As network devices can have only one name, udev drops kernel provided
name completely and replaces it with name following one naming scheme.
Thus we have to know which naming scheme is going to be used and make
sure it does not change. With multiple alternative names, we could also
have all udev provided names at once (and also the original one from
kernel).

Michal

Powered by blists - more mailing lists