From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <linux-kernel-owner@vger.kernel.org>
Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand
	id S1754488AbbCBXoT (ORCPT <rfc822;w@1wt.eu>);
	Mon, 2 Mar 2015 18:44:19 -0500
Received: from ozlabs.org ([103.22.144.67]:49643 "EHLO ozlabs.org"
	rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP
	id S1753832AbbCBXoR (ORCPT <rfc822;linux-kernel@vger.kernel.org>);
	Mon, 2 Mar 2015 18:44:17 -0500
From: Rusty Russell <rusty@rustcorp.com.au>
To: Oleg Drokin <green@linuxhacker.ru>
Cc: linux-kernel@vger.kernel.org
Subject: Re: [PATCH 05/16] staging/lustre: fix up obsolete cpu function usage.
In-Reply-To: <74B1B5A4-D3CE-4A7A-A09B-6B755623ADA4@linuxhacker.ru>
References: <1425296150-4722-1-git-send-email-rusty@rustcorp.com.au> <1425296150-4722-5-git-send-email-rusty@rustcorp.com.au> <74B1B5A4-D3CE-4A7A-A09B-6B755623ADA4@linuxhacker.ru>
User-Agent: Notmuch/0.17 (http://notmuchmail.org) Emacs/24.3.1 (x86_64-pc-linux-gnu)
Date: Tue, 03 Mar 2015 10:09:07 +1030
Message-ID: <87twy3j7fo.fsf@rustcorp.com.au>
MIME-Version: 1.0
Content-Type: text/plain
Sender: linux-kernel-owner@vger.kernel.org
List-ID: <linux-kernel.vger.kernel.org>
X-Mailing-List: linux-kernel@vger.kernel.org

Oleg Drokin <green@linuxhacker.ru> writes:
> Thanks!
> Seems there was a midair collsion with my own patch that was not as comprehensive
> wrt functions touched: https://lkml.org/lkml/2015/3/2/10

Yep, I posted this for completeness (and for your reference), but
figured you'd handle it.

> But on the other hand I also tried to clean up
> some of the NR_CPUS usage while I was at it and this raises 
> this question, from me, in the code like:
>
> for_each_cpu_mask(i, blah) {
>     blah
>     if (something)
>         break;
> }
> if (i == NR_CPUS)
>     blah;
>
> when we are replacing for_each_cpu_mask with for_each_cpu,
> what do we check the counter against now to see that the entire loop was executed
> and we did not exit prematurely? nr_cpu_ids?

You want >= nr_cpu_ids here.

> Also I assume we still want to get rid of direct cpumask assignments like
>> mask = *cpumask_of_node(cpu_to_node(index));

Yes, but this code is wrong anyway:

		mask = *cpumask_of_node(cpu_to_node(index));
		for (i = max; i < num_online_cpus(); i++)
			cpumask_clear_cpu(i, &mask);

*Never* iterate to num_online_cpus().  eg. if cpus 0 and 3 are online,
num_online_cpus() == 2.  I'm not sure what this code is doing, but it's
not doing it well :)

There are several issues here.  You need to handle cpus going offline
(during this routine, as well as after).  You need to use a
cpumask_var_t, like so:

        cpumask_var_t mask;

...
	case PDB_POLICY_NEIGHBOR:
                if (!alloc_cpumask_var(&mask, GFP_???)) {
                        rc = -ENOMEM;
                        break;
                }
                ...

Or get rid of the mask altogether, eg:

        pc->pc_npartners = -1;
        for_each_cpu(i, cpu_online_mask) {
                if (i < max)
                        pc->pc_npartners++;
        }
        ...

	pidx = 0;
        for_each_cpu(i, cpu_online_mask) {
                if (i >= max)
                        break;
		ppc = &ptlrpcds->pd_threads[i];
		pc->pc_partners[pidx++] = ppc;
		ppc->pc_partners[ppc->pc_npartners++] = pc;
        }

[ This is off the top of my head, no idea if it's right...]

Thanks,
Rusty.