hal and gmirror, gjournal

[Andriy Gapon]

on 28/02/2008 23:35 Joe Marcus Clarke said the following:
> Andriy Gapon wrote:
>> on 28/02/2008 23:02 Joe Marcus Clarke said the following:
>>> Andriy Gapon wrote:
>>>> It seems that hal gets confused by some "g" things.
>>>> For example, I have two slices on my hdd: s1 and s2.
>>>> s1 contains "primary" filesystems, s2 contains gmirror counter-part for
>>>> s1a and gjournal journals for s1d and s1e.
>>>> In the end, it's hard to say to which slice belongs a gmirror using s1a
>>>> and s2a as underlying providers. Probably to none. But hal insists on
>>>> rooting everything into a strict hierarchy and it seems that it does a
>>>> poor job at qualifying such GEOM configurations.
>>>> Please find lshal output here:
>>>> http://www.icyb.net.ua/~avg/lshal.txt
>>>> I used hal-device-manager to visualize the topology.
>>>>
>>>> P.S.
>>>> exact info:
>>>> ad4s1 and ad4s2a are combined into gmirror named rootfs
>>>> ad4s1d is data provide and ad4s2d is journal provider for ad4s1d.journal
>>>> ditto for 'e'
>>>>
>>>> And, for example, hal qualifies ad4s2d (journal provider) as a parent of
>>>> ad4s1d.journal (the 'gjournal-ed' partition).
>>> In the future, always include everything listed at 
>>> http://www.freebsd.org/gnome/docs/faq2.html#q19 when reporting hal problems.
>> Thank you for the pointer to this extensive information.
>> Additional data is attached.
>>
>> P.S. the instructions have --verbose-yes for hald command, it should
>> '=yes' ('-' vs '=').
>>
>> P.P.S. this info should be also applicable to the other problem that I
>> described (dvd-ram+fat32).
> 
> According to GEOM the .journal volume is a child of the main volume. 
> That is, ad4s2d is the parent of ad4s2d.journal.  So hal is reporting 
> this correctly (and this is what I see on my home machine which is using 
> gjournal).  I'm not sure why you think this is a problem since you mount 
> the .journal node in fstab.

It just seemed to me that it would be more natural to be a child of data
provider (ad4s1d) or of both. But if that is what GEOM says, then be it.
Or maybe I should bark at a different tree :-)

> As for the mirror problem, I'm not sure how to fix that.  The mirror is 
> a child of two nodes which violates the tree structure.  I'm open to 
> suggestions or patches.

No suggestions at this moment. I've been acting like a mere user,
complaining about things that didn't fit into 'intuitive' world-view.

-- 
Andriy Gapon