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 S965435AbcCOKDT (ORCPT <rfc822;w@1wt.eu>);
	Tue, 15 Mar 2016 06:03:19 -0400
Received: from mx2.suse.de ([195.135.220.15]:43031 "EHLO mx2.suse.de"
	rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP
	id S932585AbcCOKDO (ORCPT <rfc822;linux-kernel@vger.kernel.org>);
	Tue, 15 Mar 2016 06:03:14 -0400
Date: Tue, 15 Mar 2016 11:03:23 +0100
From: Jan Kara <jack@suse.cz>
To: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Cc: Andrew Morton <akpm@linux-foundation.org>, Jan Kara <jack@suse.com>,
        Petr Mladek <pmladek@suse.com>, Tejun Heo <tj@kernel.org>,
        Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>,
        linux-kernel@vger.kernel.org, Byungchul Park <byungchul.park@lge.com>,
        Sergey Senozhatsky <sergey.senozhatsky.work@gmail.com>,
        Jan Kara <jack@suse.cz>
Subject: Re: [RFC][PATCH v4 1/2] printk: Make printk() completely async
Message-ID: <20160315100323.GF17942@quack.suse.cz>
References: <1457964820-4642-1-git-send-email-sergey.senozhatsky@gmail.com>
 <1457964820-4642-2-git-send-email-sergey.senozhatsky@gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <1457964820-4642-2-git-send-email-sergey.senozhatsky@gmail.com>
User-Agent: Mutt/1.5.24 (2015-08-30)
Sender: linux-kernel-owner@vger.kernel.org
List-ID: <linux-kernel.vger.kernel.org>
X-Mailing-List: linux-kernel@vger.kernel.org

> +	if (!sync_print) {
> +		if (printk_thread && !in_panic) {
> +			/*
> +			 * This will wakeup the printing kthread and offload
> +			 * printing to a schedulable context.
> +			 */
> +			__this_cpu_or(printk_pending,
> +					PRINTK_PENDING_KTHREAD_OUTPUT);
> +			irq_work_queue(this_cpu_ptr(&wake_up_klogd_work));
> +		} else if (in_sched) {
> +			/*
> +			 * @in_sched messages may come too early, when we don't
> +			 * yet have @printk_thread. We can't print deferred
> +			 * messages directly, because this may deadlock, route
> +			 * them via IRQ context.
> +			 */
> +			__this_cpu_or(printk_pending,
> +					PRINTK_PENDING_IRQ_OUTPUT);
> +			irq_work_queue(this_cpu_ptr(&wake_up_klogd_work));
> +		} else {
> +			sync_print = true;
> +		}
> +	}

I'm a bit undecided whether we want to go through irq work even for the
common case of !in_sched messages or whether we want to directly call
wake_up() in that case. Maybe I'd do it like:

	if (!sync_print) {
		if (in_sched) {
			__this_cpu_or(printk_pending,
				      PRINTK_PENDING_IRQ_OUTPUT);
			irq_work_queue(this_cpu_ptr(&wake_up_klogd_work));
		} else if (printk_thread && !in_panic) {
			wake_up(&printing_wait);
		} else {
			sync_print = true;
		}
	}

and the wake_up_klogd_work_func() would look like:

static void wake_up_klogd_work_func(struct irq_work *irq_work)
{
	int pending = __this_cpu_xchg(printk_pending, 0);

	if (pending & PRINTK_PENDING_OUTPUT) {
		if (printk_thread) {
			wake_up(&printing_wait);
		} else {
			/*
			 * If trylock fails, someone else is doing the printing
			 */
			if (console_trylock())
				console_unlock();
		}
	}
	...

								Honza

-- 
Jan Kara <jack@suse.com>
SUSE Labs, CR