Re: [patch] xen udev rule interfering with openvpn

[Teck Choon Giam <giamteckchoon@gmail.com>]

On Fri, Apr 20, 2012 at 11:38 PM, Ian Campbell <Ian.Campbell@citrix.com> wrote:
> On Fri, 2012-04-20 at 16:26 +0100, Teck Choon Giam wrote:
>> On Fri, Apr 20, 2012 at 9:21 PM, Ian Campbell <Ian.Campbell@citrix.com> wrote:
>> > On Fri, 2012-04-20 at 12:04 +0100, Ian Jackson wrote:
>> >> Ian Campbell writes ("Re: [Xen-devel] [patch] xen udev rule interfering with openvpn"):
>> >> > On Fri, 2012-04-20 at 11:55 +0100, Ian Jackson wrote:
>> >> > > I'm not quite up to speed with all the context here but is the reason
>> >> > > that you're not suggesting "xen-" is that that's already used for
>> >> > > something else ?
>> >> >
>> >> > This is to distinguish the vif device from the associated tap device,
>> >> > which would otherwise both be called whatever the user gave as "vifname"
>> >> > in their config, so for vifname=foo you would get foo (the PV one) and
>> >> > xen-foo (the EMU one) which does the job but doesn't really distinguish
>> >> > them.
>> >>
>> >> Ah, I see.  This sounds like more a job for a suffix than a prefix so
>> >> if we can spare 4 chars I would suggest foo-emu.
>>
>> So what is the final prefix/suffix here?  Is it "emu-"?  Sorry, need
>> to counter check :p
>
> I think we have agreed on a "-emu" suffix.

Thanks and I will update my backport to reflect this "-emu" instead of
"xentap-" as well then test ;)

>
>> Question... vifname is limited to 16 characters?  If so, does the
>> configuration for xm/xl check for its allowable length?  I mean if a
>> user set vifname more than its allowable length in the domU
>> configuration, would xm/xl show error?
>
> I can't see any existing check for the length of the vif name.
>
> It's a bit hard for us to do, consider a network driver domain running a
> kernel which has a different, or no, limit here. libxl might not have
> any idea what that name limit should be. We could pick an arbitrary
> limit which is the smallest we know of, e.g. 12 chars (to allow +4), I
> suppose.
>
> BTW, the failure if your name is too long will be that the vif hotplug
> script will fail to rename the device.

Ok, noted.

>
> This limit has long existed and I don't recall ever having seen a report
> about it, maybe the failure case is so obvious that people just fix it
> and move on. I also expect that setting such a long name is rare...

Yes, rare I agree but you know some _rare_ users... ... As long as the
error is obvious then I think this shouldn't be an issue.

Thanks for taking time to reply.

Kindest regards,
Giam Teck Choon


>
>>
>> Sorry for asking as it isn't clear in manual... ...
>> http://xenbits.xen.org/docs/unstable/misc/xl-network-configuration.html
>> ================================
>> vifname
>>
>> This keyword is valid for HVM guest devices with type=ioemu only.
>>
>> Specifies the backend device name for an emulated device. The default
>> is tapDOMID.DEVID where DOMID is the guest domain ID and DEVID is the
>> device number.
>> ================================
>>
>> >
>> > I agree.
>> >
>> > This patch interacts a bit with Roger's hotplug series, I'll rebase on
>> > top of his with this change when he reposts it.
>>
>> Looking forward for your reports ;)
>>
>> >
>> > Ian.
>> >
>>
>> Thanks.
>>
>> Kindest regards,
>> Giam Teck Choon
>
>