[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

[dm-devel] Re: [PATCH 07/18] Replicator: suspend/deactivate replicator



On Wed, Nov 04 2009 at  5:05am -0500,
Zdenek Kabelac <zkabelac redhat com> wrote:

> Dne 3.11.2009 17:46, Mike Snitzer napsal(a):
> > On Mon, Nov 02 2009 at  9:20am -0500,
> > Zdenek Kabelac <zkabelac redhat com> wrote:
> > 
> >> Introducing dm_tree_set_replicator_suspend() to suspend
> >> replicator control device before actual deactivation of replicator-dev
> >> head device.
> >>
> >>
> >>  	int activation_priority;	/* 0 gets activated first */
> >> +	int replicator_suspend;		/* 1 gets suspend first */
> >>  
> >>  	uint16_t udev_flags;		/* Udev control flags */
> >>  
> > 
> > I think it would be wise to make this more generic,
> > e.g. "suspend_priority".
> > 
> > It could be that other future devices would like to prioritize the
> > suspend sequence too.  Having a means to do so (without using a
> > seemingly replicator-specific node attribute) would be good.
> > 
> > So this really just amounts to: s/replicator_suspend/suspend_priority/
> 
> 
>   I think using 'priority' would be more challenging here - for replicator it
> is only needs to check parental node - while with priority you would probably
> expect full tree traversal to see whether there is node which should be
> suspend in front of current device - thought the code could be probably
> extended in a way to specify recursive traversal depth - for replicator tree
> depth 1 is enough.
> 
>   Previous implementation was probably more generic in this, but required API
> changes - the use case for current implementation is rather focused on the
> replicator's needs with the advantage, it will not influence anything else.
> 
>   I think at this moment this API is static internal and might be probably
> easily changed/modified once some similar target would need to use it ?
> (It's not easy to predict future use case)

Fair enough, and maybe my naming suggestion of "suspend_priority" wasn't
quite right but I was just thinking there isn't a need to make this flag
replicator specific.  But you're right, if something else needs it in
the future we can change it then.

Mike


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]