Re: [PATCH 5.10 113/290] net: dsa: implement a central TX reallocation procedure

From: Vladimir Oltean
Date: Tue Mar 16 2021 - 12:24:30 EST


On Tue, Mar 16, 2021 at 05:05:11PM +0100, gregkh@xxxxxxxxxxxxxxxxxxx wrote:
> On Tue, Mar 16, 2021 at 09:54:01AM -0400, Sasha Levin wrote:
> > On Tue, Mar 16, 2021 at 06:46:10AM +0100, gregkh@xxxxxxxxxxxxxxxxxxx wrote:
> > > On Mon, Mar 15, 2021 at 07:56:02PM +0000, Vladimir Oltean wrote:
> > > > +Andrew, Vivien,
> > > >
> > > > On Mon, Mar 15, 2021 at 02:53:26PM +0100, gregkh@xxxxxxxxxxxxxxxxxxx wrote:
> > > > > From: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx>
> > > > >
> > > > > From: Vladimir Oltean <vladimir.oltean@xxxxxxx>
> > > > >
> > > > > [ Upstream commit a3b0b6479700a5b0af2c631cb2ec0fb7a0d978f2 ]
> > > > >
> > > > > At the moment, taggers are left with the task of ensuring that the skb
> > > > > headers are writable (which they aren't, if the frames were cloned for
> > > > > TX timestamping, for flooding by the bridge, etc), and that there is
> > > > > enough space in the skb data area for the DSA tag to be pushed.
> > > > >
> > > > > Moreover, the life of tail taggers is even harder, because they need to
> > > > > ensure that short frames have enough padding, a problem that normal
> > > > > taggers don't have.
> > > > >
> > > > > The principle of the DSA framework is that everything except for the
> > > > > most intimate hardware specifics (like in this case, the actual packing
> > > > > of the DSA tag bits) should be done inside the core, to avoid having
> > > > > code paths that are very rarely tested.
> > > > >
> > > > > So provide a TX reallocation procedure that should cover the known needs
> > > > > of DSA today.
> > > > >
> > > > > Note that this patch also gives the network stack a good hint about the
> > > > > headroom/tailroom it's going to need. Up till now it wasn't doing that.
> > > > > So the reallocation procedure should really be there only for the
> > > > > exceptional cases, and for cloned packets which need to be unshared.
> > > > >
> > > > > Signed-off-by: Vladimir Oltean <vladimir.oltean@xxxxxxx>
> > > > > Tested-by: Christian Eggers <ceggers@xxxxxxx> # For tail taggers only
> > > > > Tested-by: Kurt Kanzenbach <kurt@xxxxxxxxxxxxx>
> > > > > Reviewed-by: Florian Fainelli <f.fainelli@xxxxxxxxx>
> > > > > Signed-off-by: Jakub Kicinski <kuba@xxxxxxxxxx>
> > > > > Signed-off-by: Sasha Levin <sashal@xxxxxxxxxx>
> > > > > ---
> > > >
> > > > For context, Sasha explains here:
> > > > https://www.spinics.net/lists/stable-commits/msg190151.html
> > > > (the conversation is somewhat truncated, unfortunately, because
> > > > stable-commits@xxxxxxxxxxxxxxx ate my replies)
> > > > that 13 patches were backported to get the unrelated commit 9200f515c41f
> > > > ("net: dsa: tag_mtk: fix 802.1ad VLAN egress") to apply cleanly with git-am.
> > > >
> > > > I am not strictly against this, even though I would have liked to know
> > > > that the maintainers were explicitly informed about it.
> > > >
> > > > Greg, could you make your stable backporting emails include the output
> > > > of ./get_maintainer.pl into the list of recipients? Thanks.
> > >
> > > I cc: everyone on the signed-off-by list on the patch, why would we need
> > > to add more? A maintainer should always be on that list automatically.
> >
> > Oh, hm, could this be an issue with subsystems that have a shared
> > maintainership model? In that scenario not all maintainers will sign-off
> > on a commit.
>
> So a shared maintainer trusts their co-maintainer for reviewing patches
> for Linus's tree and all future kernels, but NOT into an old backported
> stable tree? I doubt that, trust should be the same for both.

Greg, the problem is that we have the following maintainership layout:

General networking maintainers (David Miller && Jakub Kicinski)
-> DSA framework maintainers
-> DSA hardware driver maintainers

But there is a single tree with mandatory sign-off from a single
maintainer, and that would be David or Jakub. And in rare cases it may
happen for patches to get accepted without the written ACK of any of the
sub-maintainers.

If the question is whether I trust David or Jakub to pay attention on
why are 13 patches that don't fix anything being backported to stable,
and then take the time to check/test whether anything is going to be
broken in subtle ways because code was backported in places it was never
meant to belong, then yeah, sorry, but no.

In this case, things could have gone a lot worse: the model you're
following makes it possible to backport a breaking change into a
subsystem and the maintainer can never find out until there'a a bug
report.