Re: [RFC PATCH] Tracepoints: fix section alignment using pointer array

[Mathieu Desnoyers <mathieu.desnoyers@efficios.com>]

* Rusty Russell (rusty@rustcorp.com.au) wrote:
> On Thu, 27 Jan 2011 08:56:22 am Mathieu Desnoyers wrote:
> > Make the tracepoints more robust, making them solid enough to handle compiler
> > changes by not relying on anything based on compiler-specific behavior with
> > respect to structure alignment. Implement an approach proposed by David Miller:
> > use an array of const pointers to refer to the individual structures, and export
> > this pointer array through the linker script rather than the structures per se.
> > It will consume 32 extra bytes per tracepoint (24 for structure padding and 8
> > for the pointers), but are less likely to break due to compiler changes.
> > 
> > History:
> > 
> > commit 7e066fb870fcd1025ec3ba7bbde5d541094f4ce1 added the aligned(32) type and
> > variable attribute to the tracepoint structures to deal with gcc happily
> > aligning statically defined structures on 32-byte multiples.
> > 
> > commit 15e3540ce2159705f18fad6147ffedf04445ad64 tried to use a 8-byte alignment
> > for tracepoint structures by applying both the variable and type attribute to
> > tracepoint structures definitions and declarations. It worked fine with gcc
> > 4.5.1, but broke with gcc 4.4.4 and 4.4.5.
> > 
> > The reason is that the "aligned" attribute only specify the _minimum_ alignment
> > for a structure, leaving both the compiler and the linker free to align on
> > larger multiples. Because tracepoint.c expects the structures to be placed as an
> > array within each section, up-alignment cause NULL-pointer exceptions due to the
> > extra unexpected padding.
> 
> Hmm, that assumption is used in module parameters too, so we already rely on 
> the toolchain not to over-pad.
> 
> Perhaps we should fix that too, or wait until it explodes?

Hrm, yeah, struct kernel_param seems to fit into the same category. I'd
recommend to fix it too. On 64-bit, this structure size is 28 bytes, but its
alignment is specified by:

__attribute__ ((unused,__section__ ("__param"),aligned(sizeof(void *))))

So AFAIU, if you declare __param sections in multiple different objects, that
you later link together to generate a module (or the kernel core), you might end
up with a whole caused by the realignment on 32-byte done by the linker.

In the past, tracepoints were 8-byte aligned, and I had to bump their structure
to a 32-byte alignment because of compiler behavior changes. I would personally
prefer not to wait for other things to break before introducing this fix for
struct kernel_param too.

Thanks,

Mathieu

-- 
Mathieu Desnoyers
Operating System Efficiency R&D Consultant
EfficiOS Inc.
http://www.efficios.com