lkml.org 
[lkml]   [2006]   [Sep]   [8]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectRe: [PATCH 2.6.18-rc5] PCI: sort device lists breadth-first
From
Date
On Fri, 2006-09-08 at 11:18 -0500, Matt Domsch wrote:
> On Fri, Sep 08, 2006 at 11:56:39AM -0400, Dave Jones wrote:
> > On Thu, Sep 07, 2006 at 10:14:22PM -0500, Matt Domsch wrote:
> > > Problem:
> > > New Dell PowerEdge servers have 2 embedded ethernet ports, which are
> > > labeled NIC1 and NIC2 on the chassis, in the BIOS setup screens, and
> > > in the printed documentation. Assuming no other add-in ethernet ports
> > > in the system, Linux 2.4 kernels name these eth0 and eth1
> > > respectively. Many people have come to expect this naming. Linux 2.6
> > > kernels name these eth1 and eth0 respectively (backwards from
> > > expectations). I also have reports that various Sun and HP servers
> > > have similar behavior.
> >
> > This came up years back when 2.6 was something new, and the answer
> > then was 'bind the interface to the MAC address'.
>
> Both Red Hat-based distros and openSuSE-based distros do something
> like this with configuration files automatically. Red Hat's
> anaconda/kudzu puts the HWADDR lines in the generated
> /etc/sysconfig/network-scripts/ifcfg-* files. openSuSE's udev rules
> puts lines in /etc/udev/rules.d/30-net_persistent_names.rules the
> first time it discovers a new interface. Both methods are intended to
> maintain a persistent name for each NIC, after being set up the first
> time. Neither deals well with replacing one NIC with another - you
> must edit the config files.
>
> This works pretty well post-install. It doesn't work well at install
> time, all the installers use the kernel's original names, and then
> those names become the persistent names in the config files.
>
>
> > Whilst your patch will fix the case that's currently broken (2.4->2.6),
> > doesn't it offer equal possibility to break existing setups when people move
> > from <=2.6.18 -> 2.6.19 ?
>
> If they're using config files / udev rules as suggested, it shouldn't
> break them.
>
> If they're not, then yes, this could. Debian's
> /etc/network/interfaces file allows use of hwaddr fields, though by
> default it doesn't appear anything sets it up.
>
> I'm open to suggestions on how *not* to break setups that don't use
> the MAC addresses.

to be honest I really don't like the PCI ordering change thing for this.
It's just too fragile altogether to cause a fixed ordering as you want.

Maybe the kernel's initial ordering should do a numeric sort by mac
address or something.. (or userspace should)


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2006-09-08 18:59    [W:0.045 / U:25.848 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site