Bryan Kadzban wrote:
> Bruce Dubbs wrote:
>> The section 7.2.1. Creating stable names for network interfaces
>> failed.
>
> Argh. Again.
>
>> Running manually gives:
>>
>> <...>
>>
>> This program is for debugging only, it does not run any program,
>> specified by a RUN key. <...>
>
> I wonder if they finally made this true. This sentence has been in the
> output for a long time now, but I believe it was false; if it's now
> true, then that might break the ... no, wait. That's not it. The rule
> uses IMPORT{program} instead.
>
> But this is it:
>
>> # ignore KVM virtual interfaces
>> ENV{MATCHADDR}=="52:54:00:*", GOTO="persistent_net_generator_end"
>
> (From /lib/udev/rules.d/75-persistent-net-generator.rules.)
Well that makes running jhalfs from a kvm system barf. I had to
manually remove the line:
cat /etc/udev/rules.d/70-persistent-net.rules
I suppose we can change that line in the xml to <screen role="nodump">
We probably need to mention that the file doesn't get created in kvm,
but I don't think it's needed either.
> Let me ensure this works on my real-hardware machine (running kernel
> 3.2.2 and udev-180 now, but still a pretty old glibc, so maybe not a
> perfect test, but hopefully close)...
>
> Yeah, it works. Moving the existing 70-persistent-net.rules file out of
> the way (to root's home directory) and running a manual "udevadm test
> --action=add /sys/class/net/eth0" generates a file with almost the same
> rules in it. (The new file matches on ATTR{dev_id} while the old one
> did not.)
Good.
> Of course, the comments are totally wrong; it thinks the device ID is
> the devpath, since network devices have a "device" symlink, whereas the
> sysfs attribute containing the device ID for PCI devices is also named
> "device", and udev matches on the child first -- I don't see how to fix
> this in the rules. :-(
>
> Also, I see this output in the middle:
>
> udev_rules_apply_to_event: IMPORT builtin 'pci-db'
> /lib/udev/rules.d/75-net-description.rules:11
> Failed to open database file 'no': No such file or directory
>
> Looks like either the --with-pci-ids-path=no flag isn't doing what we
> expect, or the 75-net-description.rules file is not being correctly
> suppressed. It doesn't hurt anything, but it'd be nice to be able to
> kill this. (And it seems that the net-description rules file pulls in a
> lot more than just the PCI and USB device databases, as well, so we
> can't just delete it. Looks like the default udev rules do something
> similar for TTYs (probably usb-serial) and sound cards, as well. Fun fun.)
Sigh. I suppose a message to linux-hotplug may be in order.
Thanks for the research.
-- Bruce
--
http://linuxfromscratch.org/mailman/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page
