Re: [PATCH v2 0/7] hwrng: Add support for STMicroelectronics' RNG IP

From: Lee Jones <lee.jones@linaro.org>
To: Herbert Xu <herbert@gondor.apana.org.au>
Cc: "linux-arm-kernel@lists.infradead.org" 
	<linux-arm-kernel@lists.infradead.org>,
	"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
	"devicetree@vger.kernel.org" <devicetree@vger.kernel.org>,
	linux-crypto@vger.kernel.org,
	Peter Korsgaard <peter@korsgaard.com>,
	Fabio Estevam <festevam@gmail.com>,
	Kieran Bingham <kieranbingham@gmail.com>,
	"kernel@stlinux.com" <kernel@stlinux.com>,
	Linus Torvalds <torvalds@linux-foundation.org>
Subject: Re: [PATCH v2 0/7] hwrng: Add support for STMicroelectronics' RNG IP
Date: Wed, 30 Sep 2015 15:50:34 +0100	[thread overview]
Message-ID: <20150930145034.GE27197@x1> (raw)
In-Reply-To: <20150930142812.GA19039@gondor.apana.org.au>

On Wed, 30 Sep 2015, Herbert Xu wrote:
> On Wed, Sep 30, 2015 at 03:15:39PM +0100, Lee Jones wrote:
> >
> > > I prefer not to merge patches that cannot be tested.  Without
> > > the DT bits in patch 6 the other five patches are useless.  So
> > > I think patch 6 should be applied together with the other five
> > > which add the driver.
> > 
> > That's crazy talk.  If all subsystem maintainers abide by this rule
> > there would be chaos.  We'd either need to send pull-requests to each
> > other for every set which crossed a subsystems boundary, or 1000's of
> > merge conflicts would ensue at merge time.
> > 
> > The (sensible) rule we normally stick to is; as long as there isn't
> > a _build_ dependency, then the patches should filter though their
> > respective trees; _functional_ dependencies have nothing to do with
> > us as maintainers.  Another chaos preventing rule we abide by is; thou
> > shalt not apply patches belonging to other maintainer's subsystems
> > without the appropriate Ack-by and a subsequent "you may take this
> > though your tree" and/or "please send me an immutable pull-request".
> 
> So you want the series to be merged in two parts via two different
> trees where neither can be tested? That sounds crazy to me.

Who is going to checkout the HWRNG tree and run-test it on it's own on
all of the required hardware?  No one.  Agreed, subsystem trees should
be bisectably (new word? :D) buildable as per my first rule above, but
that's it.  Per-subsystem repos are not designed to be tested for
full-functionality orthogonally, that's the point of Stephen's -next
tree.

Please take my other points into consideration too.  The kernel would
be unmainatinable if we all stuck to your rule.  No-one else has that
rule, and for good reason.

-- 
Lee Jones
Linaro STMicroelectronics Landing Team Lead
Linaro.org │ Open source software for ARM SoCs
Follow Linaro: Facebook | Twitter | Blog