Openmoko Bug #2056: /etc/resolv.conf is empty on first boot
Openmoko Public Trac
bugs at docs.openmoko.org
Wed Oct 8 01:03:33 CEST 2008
#2056: /etc/resolv.conf is empty on first boot
------------------------+---------------------------------------------------
Reporter: rwhitby | Owner: julian_chu
Type: defect | Status: new
Priority: normal | Milestone:
Component: Distro | Version:
Severity: normal | Resolution:
Keywords: | Blockedby:
Reproducible: always | Blocking:
------------------------+---------------------------------------------------
Comment(by newkirk):
Sorry, I didn't understand you were speaking to the case where NO valid
network connection exists, but route and DNS via usb0 were still active.
I ran some tests with USB unplugged (Raster's latest image with FSO feed,
though I expect 2007.x and 2008.x would be very similar) and I get these
times with 'opkg update' (with two 'dead' feeds where I've not removed
unused 'my-distribution.org' entries):
160 secs - usb0 up, dns pointing out usb0
40 secs - usb0 up, /etc/resolv.conf empty
5 secs - usb0 down with or without /etc/resolv.conf populated
So generally speaking, emptying resolv.conf leads to 25% the wait,
bringing usb0 down leads to 3%. (oh, and of that 5 secs about 4 is opkg
setting up before it tries to communicate - same for all scenarios - it
fails each connection attempt faster than it can print the status to the
screen with no route available)
The flipside is that if the default is to have nothing in /etc/resolv.conf
when only usb0 is up, then there will be a stream of complaints and bug
reports that USB networking is broken, etc. We can't require manual
insertion of DNS each time a user plugs into a host computer through which
they wish to route.
So the 'real' solution as I see it is to have resolv.conf populated
'correctly' for usb0, but bring usb0 down when no usb cable is present.
(though I don't see a simple way to achieve that automatically ATM, only
manually with a widget or icon) Without that, I don't see a way to "make
both sides happy"... :(
Another possibility, addressing the specific opkg/assassin case, is to add
a test to opkg_download() that tries to ping the default gateway, and dies
if it's unreachable. Which seems sensible to me regardless. ;)
j
--
Ticket URL: <https://docs.openmoko.org/trac/ticket/2056#comment:5>
docs.openmoko.org <http://docs.openmoko.org/trac/>
openmoko trac
More information about the buglog
mailing list