| Summary: | eth1394 grabs eth0 since coldplug went away | ||
|---|---|---|---|
| Product: | Gentoo Linux | Reporter: | Roger Binns <gentoo> |
| Component: | [OLD] baselayout | Assignee: | Gentoo's Team for Core System packages <base-system> |
| Status: | RESOLVED DUPLICATE | ||
| Severity: | normal | CC: | stefan-r-bz |
| Priority: | High | ||
| Version: | 2006.1 | ||
| Hardware: | x86 | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Package list: | Runtime testing required: | --- | |
|
Description
Roger Binns
2006-12-04 01:31:07 UTC
all the naming comes from the kernel so if you want to control that, you should be messing with kernel parameters so that you tell the kernel which interface should be what this is not a userspace problem by any means Don't rely on non-deterministic module load ordering; if you want persistent device names, that's what udev rules are for. (And, if you don't want ethernet over firewire, then why did you compile that module in the first place?) *** This bug has been marked as a duplicate of 156893 *** I don't want eth1394 and it gets included by default with firewire. I am using genkernel. I am surprised at the responses in this and in 156893 as well as on the mailing lists. The users keep being blamed for behaviour that has now changed suddenly for the first time in the 13 years I have been using Linux. Sure the module loading order determines names, but that behaviour has always been deterministic in the past. Now it is random. The install guide does mention that your network interfaces have random names, or how to make them fixed. Upstream bug http://bugzilla.kernel.org/show_bug.cgi?id=7793 was closed, fix went into mainline for 2.6.22-rc1. Note, after that fix ieee1394 will still generate a hotplug event to let userspace load eth1394 if it detects an external IP-over-1394 capable device. Re comment #3: I'd say module load order has always been semi-deterministic, now it is semi-random. The old philosophy was "all devices are fixed, except only some are hotpluggable", the current philosophy is "all devices are hotpluggable from the drivers' point of view, it's just that the hardware of some devices doesn't feature hotplug-ability". And the reality is, new bus architectures tend to be hotpluggable or are otherwise highly reconfigurable or feature more complex discovery protocols. The approach to rely on module load order and each subsystem's device enumeration habits is simply not future proof. The only future-proof way is to rely on persistent unique device properties, like e.g. MACs. All bus architectures of the more flexible kind feature such persistent unique device properties, in one way or another. Re comment #1: The kernel only offers some simply enumerated default names. Persistent naming schemes are the responsibility of userspace tools. |