Re: [PATCH 1/3] stack overflow safe kdump (2.6.18-rc1-i386) - safe_smp_processor_id

[ebiederm@xmission.com (Eric W. Biederman)]

James Bottomley <James.Bottomley@SteelEye.com> writes:

> On Mon, 2006-07-10 at 21:42 -0600, Eric W. Biederman wrote:
>> But I do agree the subarch header files are clean.
>> And no this case except for the fact no one realized that the
>> code doesn't even compile on voyager does not show how brittle
>> the x86 subarch code is.    Except for the fact that it seems
>> obvious that kernel/smp.c is generic code that every smp subarch
>> would use.
>
> OK ... that's the mistaken assumption.  kernel/smp.c is not subarch
> generic, it's APIC specific.  So all apic using subarchs, which is
> pretty much everything except voyager, use it.  Since voyager uses
> vic/qic based smp harness, it has its own version of this file (in fact
> voyager has a completely separate SMP HAL).

Yep.  My point is that with the current subarch structure on x86 it is
really easy to make mistaken assumptions like kernel/smp.c applies to
all x86 subarchitectures, because the lines are not clear.  The
architectures where I have seen that the lines are clear generally
allow for building a single kernel that can boot on any subarch.

My hope is that we can recognized how non-obvious the x86 subarch code
is so that future work will be able to improve the situation.

To give credit I do think the division of labor between the subarch's
appears sound.  I just don't like how the subarches are glued together
into the x86 arch.

Eric