Cryptsetup with sata drive and cryptoroot

[Reinhard Tartler]

Ilkka Tuohela <hile at ner.dy.fi> writes:

> Reinhard Tartler kirjoitti:
>> Scott, Ian,
>> 
>> After reading and chatting with Scott in IRC, it seems that the below
>> patch should not be necessary anymore. Could you please comment on the
>> patch below? How does this integrate into
>> https://wiki.ubuntu.com/UdevDeviceMapper?
>
> So what's status of including my small patches for feisty package?
>
> I'm talking about:
> - patch to wait for the crypto device to appear (not LVM), this patch
>   also cleaned up anything LVM and EVMS-related parts from the scripts
> - patch to add documentation of LVMROOT-ON-LUKS to the README in package
>
> I have been running my own, patched version with the first patch since
> we last talked about this, and it's working just perfectly. The patch
> for waiting for device to appear is really required, as long as we don't
> change whole cryptsetup package to asynchronous one, which does not seem
> feasible to do with feisty (no way to handle asynchronous scripts with
> input from user). Without the device waiting patch you just don't get a
> working system in many cases, because the cryptsetup script is executed
> before kernel has the devices ready.

Hm, the problem is that I cannot conviniently test your packages. Could
you perhaps provide sources+binary packages somewhere and ask on the
mailing lists ubuntu-devel@ and ububntu-motu@ for having them tested?
I'd feel much better if at least another developer could confirm that
this actually works for them.

It's not that I wouldn't trust you. It's rather that I know from my own
experience, that it's pretty easy to produce packages, which only work
in your particular setting, and having not really documented how to
setup that setting. The 2nd reviwer ensures that the setting you like to
fix is actually reproducable outside your environment.

> I would love to see completely asynchornous cryptsetup handling, but we
> really should fix the waiting problem NOW, not waiting for someone to
> actually implement required asynchornous input tasks.

I agree.

> Oh, btw, same 'wait for root device' problem exists in edgy as well, and
> same 'wait for device' fix works there. The maximum wait time is maybe
> silly (360*0.5 seconds, i.e. 3 minutes), I think maximum of 30 seconds
> should be really fine.

This can be adjusted with the option rootdelay=20 as kernel parameter.

-- 
Gruesse/greetings,
Reinhard Tartler, KeyID 945348A4