Re: [PATCH v8 02/18] soc: ti: k3: add navss ringacc driver

[Vinod Koul]

On 15-01-20, 11:44, Peter Ujfalusi wrote:
> 
> 
> On 14/01/2020 20.06, santosh.shilimkar@xxxxxxxxxx wrote:
> >>>>> Can you please giver your Acked-by for the ringacc patches if they are
> >>>>> still OK from your point of view as you had offered to take them
> >>>>> before
> >>>>> I got comments from Lokesh.
> >>>>>
> >>>> Sure. But you really need to split the series so that dma engine and
> >>>> soc driver patches can be applied independently.
> >>>
> >>> The ringacc is a build and runtime dependency for the DMA. I have hoped
> >>> that all of them can go via DMAengine (hence asking for your ACK on the
> >>> drivers/soc/ti/ patches) for 5.6.
> >>>
> >>>> Can you please do that?
> >>>
> >>> This late in the merge window that would really mean that I will miss
> >>> another release for the KS3 DMA...
> >>> I can live with that if you can pick the ringacc for 5.6 and if Vinod
> >>> takes the DMAengine core changes as well.
> >>>
> >>> That would leave only the DMA drivers for 5.7 and we can also queue up
> >>> changes for 5.7 which depends on the DMAengine API (ASoC changes, UART,
> >>> sa2ul, etc).
> >>>
> >>> If they go independently and nothing makes it to 5.6 then 5.8 is the
> >>> realistic target for the DMA support for the KS3 family of devices...
> >>
> >> Thats too many kernel versions to get this important piece in.
> >>
> >> Santosh, if you do not have anything else queued up that clashes with
> >> this, can the whole series be picked up by Vinod with your ack on the
> >> drivers/soc/ti/ pieces?
> >>
> > I would prefer driver patches to go via driver tree.
> 
> Not sure what you mean by 'driver patches'...
> The series to enable DMA support on TI's K3 platform consists:
> Patch 1-2: Ring Accelerator _driver_ (drivers/soc/ti/)
> Patch 3-6: DMAengine core patches to add new features needed for k3-udma
> Patch 7-11: DMA _driver_ patches for K3 (drivers/dma/ti/)
> 
> Patch 10 depends on the ringacc and the DMAengine core patches
> Patch 11 depends on patch 10
> 
> I kept it as a single series in hope that they will go via one subsystem
> tree to avoid build dependency issues and will have a good amount of
> time in linux-next for testing.
> 
> >> Vinod could also perhaps setup an immutable branch based on v5.5-rc1
> >> with just the drivers/soc/ti parts applied so you can merge that branch
> >> in case you end up having to send up anything that conflicts.
> >>
> > As suggested on other email to Peter, these DMA engine related patches
> > should be queued up since they don't have any dependency. Based on
> > the status of that patchset, will take care of pulling in the driver
> > patches either for this merge window or early part of next merge window.
> 
> OK, I'll send the two patch for ringacc as a separate series.
> 
> Vinod: Would it be possible for you to pick up the DMAengine core
> patches (patch 3-6)?
> The UDMA driver patches have hard dependency on DMAengine core and
> ringacc, not sure how they are going to go in...

Since they have build dependency, the usual method for this is:

1. Santosh acks the dependent patches and I will apply the rest (nice
and simple)

2. Santosh picks up ring driver patches, provides a signed immutable tag
which I will pull in and apply the rest, i.e., dmaengine updates and new
dmaengine driver

That way we are all okay and changes get merged now.. there is a 3rd
option

3. Santosh picks ring driver patches, and Vinod picks rest after next
rc1 (provided they make to linus in merge window)

I would love to see either 1 or 2 whichever way folks are comfortable
to deal with :)

-- 
~Vinod