[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20101007184440.GA2999@libnet-test.oslab.blr.amer.dell.com>
Date: Thu, 7 Oct 2010 11:44:57 -0700
From: <Narendra_K@...l.com>
To: <kay.sievers@...y.org>
CC: <greg@...ah.com>, <Matt_Domsch@...l.com>, <netdev@...r.kernel.org>,
<linux-hotplug@...r.kernel.org>, <linux-pci@...r.kernel.org>,
<Jordan_Hargrave@...l.com>, <Vijay_Nijhawan@...l.com>,
<Charles_Rose@...l.com>
Subject: Re: [PATCH V2] Use firmware provided index to register a network
interface
On Thu, Oct 07, 2010 at 10:35:14PM +0530, Kay Sievers wrote:
> On Thu, Oct 7, 2010 at 18:48, Greg KH <greg@...ah.com> wrote:
> > On Thu, Oct 07, 2010 at 11:31:13AM -0500, Matt Domsch wrote:
> >> 1) SMBIOS type 41 method. Windows does not use this today, and I
> >> can't speak to their future plans. Narendra's kernel patch does,
> >> as has biosdevname, the udev helper we first wrote for this
> >> purpose, for several years.
> >
> > Then stick with that udev helper please :)
>
> What about just exporting this information in sysfs, and not touch the
> naming?
>
> Anyway, I'm pretty sure all of this naming of onboard devices should
> happen only at install time, or from a system management tool and not
> at hotplug time.
>
> We should not get confused by the way the (very simple)
> automatic-rule-creater for persistent netdev naming in udev works.
> This is really just a tool for the common case, and works fine for the
> majority of people.
Right. It works as the automatic rule creator saves the snapshot of the
registered network interfaces at run time.
>
> I'm not sure, if we should put all these special use cases in the
> hotplug path. I mean it's not that people add and remove 4 port
> network cards with special BIOS all the time, and expect proper naming
> on the first bootup, right? The installer, or the system management
> tool could just create/edit udev rules to provide proper device naming
> on whatever property is available at a specific hardware, be it the
> MAC address or some other persistent match?
>
The proposal made is not expecting deterministic naming when an add-in
card with 'N' ports is plugged in/out. It is specific to onboard devices
only. Expectation is onboard devices have deterministic naming at first
bootup. And no special BIOS required as SMBIOS tables are in use for
sometime now.
I did explore using rules based on the exported attribute ATTRS{index}
on a system with 4 Onboard devices and two add-in devices.(where add-in
device becomes eth0 and eth1)
# PCI device 0x14e4:0x164c (bnx2) (custom name provided by external
# tool) SUBSYSTEM=="net", ACTION=="add", DRIVERS=="?*",
# ATTRS{index}=="1", ATTR{type}=="1", KERNEL=="eth*", NAME="eth0"
(and similary for eth1..eth3)
And for add-in devices.
# PCI device 0x8086:0x10c9 (igb) (custom name provided by external tool)
# SUBSYSTEM=="net", ACTION=="add", DRIVERS=="?*",
# ATTR{address}=="00:1b:21:54:33:3c", ATTR{type}=="1", KERNEL=="eth*",
# NAME="eth4" (and similar for eth5 as they do not have an index)
This works as i edited the file manually.
If this has to be done on a large number of systems where an image
based deployment is preferred, getting onboard device names as expected
is an issue and is important.
--
With regards,
Narendra K
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Powered by blists - more mailing lists