Re: initialization of static per-cpu variables

[Vegard Nossum]

On Fri, May 23, 2008 at 4:29 PM, Adrian Bunk <bunk@xxxxxxxxxx> wrote:
> On Thu, May 22, 2008 at 06:20:06PM +1000, Rusty Russell wrote:
>> On Thursday 22 May 2008 04:28:02 Vegard Nossum wrote:
>> > Hi,
>> >
>> > I encountered this comment in kernel/softirq.c:
>> >
>> > /* Some compilers disobey section attribute on statics when not
>> >    initialized -- RR */
>> > static DEFINE_PER_CPU(struct tasklet_head, tasklet_vec) = { NULL };
>> > static DEFINE_PER_CPU(struct tasklet_head, tasklet_hi_vec) = { NULL };
>> >
>> > So I assume it's the combination of static and whatever section
>> > DFINE_PER_CPU puts the variable in which is the problem.
>> >
>> > However, there's a LOT of these "static DEFINE_PER_CPU" without any
>> > initializer in the rest of the code, e.g.:
>>
>> Yep, it was an old toolchain used by Sparc: DaveM found this one.  As you say,
>> it's ancient: I'm happy to queue a cleanup patch now everyone is on a modern
>> compiler.
>
> The commit says:
>
>    [PATCH] softirq.c per_cpu fix
>
>    GCC3.1 apparently gets confused about uninitialized sections
>
>
> We do still support gcc 3.2 (which is the same as 3.1 except for a C++
> ABI change) as a compiler for the kernel.

Ah, thanks for looking that up!

The question is then whether 3.1/3.2 will _ever_ do the right thing,
when you consider the number of places in the kernel that *don't*
initialize static per-cpu variables. (Or maybe: What makes this
variable so special that it breaks the kernel?)

Maybe it was fixed by accident in the kernel at some point later (e.g.
by change of the linker script) or maybe the commit message/original
reason was wrong and it was really something else?

Or maybe it is time to deprecate gcc 3.1/3.2 support?

Vegard

-- 
"The animistic metaphor of the bug that maliciously sneaked in while
the programmer was not looking is intellectually dishonest as it
disguises that the error is the programmer's own creation."
	-- E. W. Dijkstra, EWD1036
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/