Home » Mailing list archives » linux-kernel » 2008 » March » 3

Re: [PATCH 0/52] First attempt at smp integration

!MAILaRCHIVE_VOTE_RePLACE
Previous message: [thread] [date] [author]
Next message: [thread] [date] [author]
[view in full thread]
From: Andi Kleen <andi@...>
To: Glauber Costa <gcosta@...>
Cc: <linux-kernel@...>, <akpm@...>, <glommer@...>, <mingo@...>, <tglx@...>
Subject: Re: [PATCH 0/52] First attempt at smp integration
Date: Monday, March 3, 2008 - 1:23 pm

Glauber Costa <gcosta@redhat.com> writes:

The main difference between i386 and x86-64 smp boot up (among a lot
of quirks and workarounds in i386 that x86-64 doesn't have/need and
some trivialities) is that x86-64 follows the standard hotplug state
machine and i386 doesn't.

I would suggest that if you want to unify you convert i386 over to the new 
state machine first because that is the key difference. Everything else
after that is relatively simple. You can probably pattern that after
the original changeset who did this for x86-64.

The reason I never attempted this myself is that i was too worried about
regressions on old machines for i386 (this code is partly very fragile)
You'll likely encounter that problem. Just make sure to keep a stable
email address for some years to handle the fallout.

-Andi
--
Previous message: [thread] [date] [author]
Next message: [thread] [date] [author]
Messages in current thread:
[PATCH 0/52] First attempt at smp integration, Glauber Costa, (Mon Mar 3, 1:12 pm)
Re: [PATCH 0/52] First attempt at smp integration, Andi Kleen, (Mon Mar 3, 1:23 pm)
Re: [PATCH 0/52] First attempt at smp integration, Glauber Costa, (Mon Mar 3, 1:38 pm)
Re: [PATCH 0/52] First attempt at smp integration, Ingo Molnar, (Mon Mar 3, 1:22 pm)
Re: [PATCH 0/52] First attempt at smp integration, Glauber Costa, (Mon Mar 3, 1:35 pm)
[PATCH 01/52] commonize smp.h, Glauber Costa, (Mon Mar 3, 1:12 pm)
[PATCH 02/52] merge extern function definitions, Glauber Costa, (Mon Mar 3, 1:12 pm)
[PATCH 03/52] merge extern variables definitions, Glauber Costa, (Mon Mar 3, 1:12 pm)
[PATCH 04/52] define smp_ops in common header, Glauber Costa, (Mon Mar 3, 1:12 pm)
[PATCH 05/52] move smp_ops extern declaration to common header, Glauber Costa, (Mon Mar 3, 1:12 pm)
[PATCH 06/52] merge smp_send_reschedule, Glauber Costa, (Mon Mar 3, 1:12 pm)
[PATCH 07/52] unify smp_call_function_mask, Glauber Costa, (Mon Mar 3, 1:12 pm)
[PATCH 08/52] unify __cpu_up., Glauber Costa, (Mon Mar 3, 1:12 pm)
[PATCH 09/52] unify prepare_boot_cpu, Glauber Costa, (Mon Mar 3, 1:12 pm)
[PATCH 10/52] unify smp_prepare_cpus, Glauber Costa, (Mon Mar 3, 1:12 pm)
[PATCH 11/52] unify smp_cpus_done, Glauber Costa, (Mon Mar 3, 1:12 pm)
[PATCH 12/52] move disabled_cpus to common header, Glauber Costa, (Mon Mar 3, 1:12 pm)
[PATCH 13/52] use disabled_cpus in i386, Glauber Costa, (Mon Mar 3, 1:12 pm)
[PATCH 14/52] move prefill_possible_map to common file, Glauber Costa, (Mon Mar 3, 1:12 pm)
[PATCH 15/52] remove export for smp_call_function_mask., Glauber Costa, (Mon Mar 3, 1:12 pm)
[PATCH 16/52] remove irqs disabled warning., Glauber Costa, (Mon Mar 3, 1:12 pm)
[PATCH 17/52] create smpcommon.c, Glauber Costa, (Mon Mar 3, 1:12 pm)
[PATCH 18/52] provide __smp_call_function, Glauber Costa, (Mon Mar 3, 1:12 pm)
[PATCH 19/52] change x86_64 smp_call_function_mask to look a..., Glauber Costa, (Mon Mar 3, 1:12 pm)
[PATCH 20/52] provide hlt_works function., Glauber Costa, (Mon Mar 3, 1:12 pm)
[PATCH 21/52] make stop_this_cpu looks exactly equal in both..., Glauber Costa, (Mon Mar 3, 1:12 pm)
[PATCH 22/52] add reboot_force test to native_smp_send_stop, Glauber Costa, (Mon Mar 3, 1:12 pm)
[PATCH 23/52] unify smp_send_stop, Glauber Costa, (Mon Mar 3, 1:12 pm)
[PATCH 24/52] create smp.c, Glauber Costa, (Mon Mar 3, 1:12 pm)
[PATCH 25/52] create ipi.c, Glauber Costa, (Mon Mar 3, 1:12 pm)
[PATCH 26/52] create tlb files, Glauber Costa, (Mon Mar 3, 1:12 pm)