[dm-devel] [RFC PATCH v2] dm mpath: add a queue_if_no_path timeout

[Mike Snitzer]

On Fri, Oct 18 2013 at  6:53pm -0400,
Mike Snitzer <snitzer at redhat.com> wrote:

> On Fri, Oct 18 2013 at  4:51pm -0400,
> Frank Mayhar <fmayhar at google.com> wrote:
> 
> > On Thu, 2013-10-17 at 17:13 -0400, Mike Snitzer wrote:
> > > Cannot say that argument wins me over but I will say that if you intend
> > > to take the approach to have the kernel have a timeout; please pursue
> > > the approach Hannes offered:
> > > 
> > > https://patchwork.kernel.org/patch/2953231/
> > > 
> > > It is much cleaner and if it works for your needs we can see about
> > > getting a tested version upstream.
> > 
> > Unfortunately his patch doesn't work as-is; it turns out that it tries
> > to set the timeout only if the target is request-based but at the time
> > he tries to set it the table type hasn't yet been set.
> > 
> > I'm looking into fixing it.
> 
> Ouch, yeah, can't access the DM device's queue from .ctr()
> There were other issues with Hannes RFC patch, wouldn't compile.
> 
> Anyway, looks like we need a new target_type hook (e.g. .init_queue)
> that is called from dm_init_request_based_queue().
> 
> Request-based DM only allows a single DM target per device so we don't
> need the usual multi DM-target iterators.
> 
> But, unfortunately, at the time we call dm_init_request_based_queue()
> the mapped_device isn't yet connected to the inactive table that is
> being loaded (but the table is connected to the mapped_device).
> 
> In dm-ioctl.c:table_load(), the inactive table could be passed directly
> into dm_setup_md_queue().
> 
> Please give the following revised patch a try, if it works we can clean
> it up further (think multipath_status needs updating, we also may want
> to constrain .init_queue to only being called if the target is a
> singleton, which dm-mpath should be, but isn't flagged as such yet).
> 
> It compiles, but I haven't tested it...

Frank,

Any interest in this or should I just table it for >= v3.14?