[systemd-devel] Possible net_id incorrect bus and/or slot

[Kay Sievers]

On Thu, Jan 9, 2014 at 7:55 PM, Robert Milasan <rmilasan at suse.com> wrote:
> On Thu, 9 Jan 2014 11:07:40 +0100
> "Tomasz Torcz" <tomek at pipebreaker.pl> wrote:
>
>> On Thu, Jan 09, 2014 at 10:59:34AM +0100, Robert Milasan wrote:
>> > Hello, just notice that my network card is named enp0s25, but when
>> > I do:
>> >
>> > # readlink -f /sys/class/net/enp0s25
>> > /sys/devices/pci0000:00/0000:00:19.0/net/enp0s25
>> >
>> > # lspci |grep Network
>> > 00:19.0 Ethernet controller: Intel Corporation 82579LM Gigabit
>> > Network Connection (rev 04)
>> >
>> > enp0s25 seems wrong to me, or the slot is based on what?
>> >
>> > Please clarify how to get this.
>>
>>   0x19 = 25.  It's hexadecimal vs decimal.
>>
>
> OK, now because that is clear, there is another issue:
>
> Example:
>
> user has an on board network adapter named enp3s0.
> user adds 2 additinal network adapters on pcie and they will be named
> enp6s0 and enp5s0, but enp3s0 also got renamed to enp4s0.
>
> Now I'm not sure if this is normal, but in my own limited logic, it
> doesn't sounds or look normal at all. The setup or the naming is not
> very stable.
>
> The on board nic didn't change or wasn't placed someplace else, well
> it's on board, so it's a bit difficult to place on some other slot.
> Also the users hasn't touch the system's BIOS.
>
> My question is: is this expected or normal?

No, but it happens on boxes where the firmware does magic things.

Re-configuring hardware, or BIOS/firmware updates can always change
the visible PCI geography, also for otherwise unrelated devices.

We can only be as good with the predictable names as the stuff we rely
on is. In some cases the MAC addresses need to be used because the
physical properties are not stable enough; but that is nothing we
could find out automatically.

The majority of devices, especially the one with sane firmware, do not
have these problems.

Kay