Re: [PATCH 0/8] FPGA DFL Changes for 5.12

[Greg KH]

On Mon, Jan 11, 2021 at 11:46:03AM -0800, Tom Rix wrote:
> 
> On 1/11/21 10:21 AM, Greg KH wrote:
> > On Mon, Jan 11, 2021 at 08:43:15AM -0800, Tom Rix wrote:
> >> On 1/11/21 8:09 AM, Greg KH wrote:
> >>> On Mon, Jan 11, 2021 at 07:55:24AM -0800, Tom Rix wrote:
> >>>> On 1/11/21 6:54 AM, Greg KH wrote:
> >>>>> On Mon, Jan 11, 2021 at 06:40:24AM -0800, Tom Rix wrote:
> >>>>>> On 1/10/21 10:57 PM, Greg KH wrote:
> >>>>>>> On Sun, Jan 10, 2021 at 11:43:54AM -0800, Tom Rix wrote:
> >>>>>>>> On 1/10/21 9:05 AM, Moritz Fischer wrote:
> >>>>>>>>> Tom,
> >>>>>>>>>
> >>>>>>>>> On Sun, Jan 10, 2021 at 07:46:29AM -0800, Tom Rix wrote:
> >>>>>>>>>> On 1/7/21 8:09 AM, Tom Rix wrote:
> >>>>>>>>>>> On 1/6/21 8:37 PM, Moritz Fischer wrote:
> >>>>>>>>>>>> This is a resend of the previous (unfortunately late) patchset of
> >>>>>>>>>>>> changes for FPGA DFL.
> >>>>>>>>>>> Is there something I can do to help ?
> >>>>>>>>>>>
> >>>>>>>>>>> I am paid to look after linux-fpga, so i have plenty of time.
> >>>>>>>>>>>
> >>>>>>>>>>> Some ideas of what i am doing now privately i can do publicly.
> >>>>>>>>>>>
> >>>>>>>>>>> 1. keep linux-fpga sync-ed to greg's branch so linux-fpga is normally in a pullable state.
> >>>>>>>>> Is it not? It currently points to v5.11-rc1. If I start applying patches
> >>>>>>>>> that require the changes that went into Greg's branch I can merge.
> >>>>>>>> I mean the window between when we have staged patches and when they go into Greg's branch.
> >>>>>>>>
> >>>>>>>> We don't have any now, maybe those two trival ones.
> >>>>>>>>
> >>>>>>>> Since Greg's branch moves much faster than ours, our staging branch needs to be rebased regularly until its merge.
> >>>>>>> Ick, no!  NEVER rebase a public branch.  Why does it matter the speed of
> >>>>>>> my branch vs. anyone elses?  Git handles merges very well.
> >>>>>>>
> >>>>>>> Just like Linus's branches move much faster than mine, and I don't
> >>>>>>> rebase my branches, you shouldn't rebase yours.
> >>>>>>>
> >>>>>>> Becides, I'm only taking _PATCHES_ for fpga changes at the moment, no
> >>>>>>> git pulls, so why does it matter at all for any of this?
> >>>>>>>
> >>>>>>> What is the problem you are trying to solve here?
> >>>>>> This 5.12 fpga patchset not making it into 5.11.
> >>>>> Ok, but isn't it the responsibility of the submitter to make sure they
> >>>>> apply properly when sending them out?
> >>>>>
> >>>>>> At some point before the 5.11 window, I tried it on next and it failed to merge.
> >>>>>>
> >>>>>> This points to needing some c/i so it does not happen again.
> >>>>> "again"?  Merges and the like are a totally normal thing and happen all
> >>>>> the time, I still fail to understand what you are trying to "solve" for
> >>>>> here...
> >>>> What can I do to help make your merges as easy as possible ?
> >>> I have not had any problems with merges, I've only had "problems"
> >>> rejecting patches for their content.
> >>>
> >>> Try helping out with patch reviews if you want, finding and fixing
> >>> things before I review them is usually a good idea :)
> >> ok.
> >>>> Does the patchwork infra Moritz was speaking of earlier need fixing help?
> >>> No idea, I don't use it.
> >>>
> >>>> Any other things ?
> >>> What problems are you trying to solve here?  What's wrong with how this
> >>> subsystem is working that you are feeling needs to be addressed?
> >> I do not believe the issue I raised in 5.10 has made any progress.
> > What issue?
> >
> >> If you look at the content in 5.11 we have actually regressed.
> > What bugs regressed?
> >
> >> https://lore.kernel.org/linux-fpga/3295710c-5e82-7b97-43de-99b9870a8c8c@xxxxxxxxxx/
> > I don't see the problem here, other than a low-quality of patches that
> > need reworking for some patchsets, and others are just fine.  Just like
> > all kernel subsystems, I don't see anything odd here.
> >
> >> Over the last two releases, I have shown i have the time and interest to maintain this subsystem.
> > That's not how any of this works :)
> >
> >> So I am asking for
> >>
> >> diff --git a/MAINTAINERS b/MAINTAINERS
> >> index 11b38acb4c08..269cd08f4969 100644
> >> --- a/MAINTAINERS
> >> +++ b/MAINTAINERS
> >> @@ -6951,7 +6951,7 @@ F:        drivers/net/ethernet/nvidia/*
> >>  
> >>  FPGA DFL DRIVERS
> >>  M:     Wu Hao <hao.wu@xxxxxxxxx>
> >> -R:     Tom Rix <trix@xxxxxxxxxx>
> >> +M:     Tom Rix <trix@xxxxxxxxxx>
> > That's generous, but how about doing review first, the maintainership of
> > this subsystem does not feel like any sort of bottleneck to me.  I
> > personally have no problems with Moritz's interactions with the
> > community, his reviewing of patches, and forwarding on to me.
> >
> > Of course we all have delays as we have other work to do than just this,
> > that's just part of normal development.  I don't see anything stalled at
> > the moment, nor anything that having another maintainer would have
> > helped out with at all, so this feels like it is not needed from my end.
> >
> > Again, it feels like the developers need more reviews, and good ones, so
> > please continue to help out with that, as that's the best thing I can
> > see to do here.
> 
> I have been doing the first review in a couple of days after every patch landing.
> 
> I see some pretty good response from the developers to fix the issues raised. 
> 
> But I do not see Moritz picking up the review until weeks later.
> 
> This consistent delay in timely reviews is a bottleneck.

What is being prevented from merging at this moment because of this?

There is always ebbs and flows with patch reviews, what do you consider
"manditory" time to review a patch from a maintainer?

Remember, we are almost all doing this not as our paid job, but as part
of what we do.

And to be frank, I personally have been putting the fpga patchset at the
_end_ of my reviews because of the quality of them in the past.  While
it is slowly getting better, you all have a big hole to dig out from,
and maybe you need to step up your reviews to raise that bar so I don't
dread seeing them in my review queue :)

> It would be good if the big first reviews could be done in parallel.

parallel with what?

All I'm getting here is "I wish patches were accepted faster!" from you.
When personally, I think they are being accepted at a very nice rate for
such a tiny subsystem.

Good luck with your reviews,

greg k-h