Re: Linker Issues with built-in.o on my machine for x86

[Steven Rostedt <rostedt@goodmis.org>]

On Sun, 11 Jan 2015 15:39:18 -0500
nick <xerofoify@gmail.com> wrote:

> Greetings x86 Maintainers.
> I am getting a linker error as of today on a up to date git tree of linus's on my x86 machine.
> Below this message is the linker message I am getting.
> WARNING: arch/x86/kernel/built-in.o(.text.unlikely+0x1204): Section mismatch in reference from the function cpumask_empty.constprop.3() to the variable .init.data:nmi_ipi_mask
> The function cpumask_empty.constprop.3() references
> the variable __initdata nmi_ipi_mask.
> This is often because cpumask_empty.constprop.3 lacks a __initdata 

Hmm, I bet it had something to do with some sort of gcc optimization,
and is mostly harmless.

> annotation or the annotation of nmi_ipi_mask is wrong.

What gcc are you using and can you send your .config.

remote_ipi() (which is annotated as __init), calls cpumask_empty()
(a static inlined function) with to_cpumask(nmi_ipi_mask). There's
nothing wrong here. I don't see anyway that cpumask_empty() could use
nmi_ipi_mask after boot up.

Seeing that it's not cpumask_empty(), but cpumask_empty.constprop.3
tells me that gcc probably created some function from the static inline
function of cpumask_empty() and that causes the section mismatch test
to give a false error.

If anything should be fixed here, it is to find a way to fix the section
mismatch test from reporting bogus errors like this.

-- Steve