Re: [RFC PATCH 2/2] Kbuild: avoid partial linking of drivers/built-in.o

From: Dave Martin <Dave.Martin@arm.com>
To: Nicolas Pitre <nicolas.pitre@linaro.org>
Cc: Michal Marek <mmarek@suse.cz>,
	Russell King - ARM Linux <linux@arm.linux.org.uk>,
	Arnd Bergmann <arnd@arndb.de>,
	linux-kbuild@vger.kernel.org,
	Ard Biesheuvel <ard.biesheuvel@linaro.org>,
	"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
	"linux-arm-kernel@lists.infradead.org" 
	<linux-arm-kernel@lists.infradead.org>
Subject: Re: [RFC PATCH 2/2] Kbuild: avoid partial linking of drivers/built-in.o
Date: Wed, 1 Apr 2015 10:04:57 +0100	[thread overview]
Message-ID: <20150401090449.GA3602@e103592.cambridge.arm.com> (raw)
In-Reply-To: <alpine.LFD.2.11.1503311429240.25806@knanqh.ubzr>

On Tue, Mar 31, 2015 at 02:42:24PM -0400, Nicolas Pitre wrote:
> On Tue, 31 Mar 2015, Dave Martin wrote:

[...]

> > We'd need to avoid pruning needed code that has no explicit caller,
> > and functions that are part of the kernel/module ABI but not used
> > within vmlinux.
> 
> Those are usually located in special sections, like the initcall table, 
> the CPU entry table, etc. The linker allows for those sections to be 
> marked with KEEP() not to prune them.

True.  See what breaks, I guess.

> > The GCC docs suggest that -ffunction-sections may impact performance
> > and/or increase code size, but I don't know by how much.  Maybe it
> > interferes with inling.
> 
> It doesn't interfere with inlining. However it impose a section 
> alignment on every function. That still can be overriden though. Also, I 
> suppose that gcc may not assume that calls to a global function that 
> happens to be located in the same C file will be close by anymore, 
> however I don't see this having any impact on ARM code generation.

More veneers may be introduced during linking, and it's likely that
cache/TLB locality would be affected, but I've no idea whether the
impact would be significant.

Cheers
---Dave