linux-kernel - Re: [RFC][PATCH v8 1/2] printk: Make printk() completely async

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20160401011428.GB501@swordfish>
Date:	Fri, 1 Apr 2016 10:14:28 +0900
From:	Sergey Senozhatsky <sergey.senozhatsky.work@...il.com>
To:	Jan Kara <jack@...e.cz>
Cc:	Sergey Senozhatsky <sergey.senozhatsky@...il.com>,
	Andrew Morton <akpm@...ux-foundation.org>,
	Jan Kara <jack@...e.com>, Petr Mladek <pmladek@...e.com>,
	Tejun Heo <tj@...nel.org>,
	Tetsuo Handa <penguin-kernel@...ove.SAKURA.ne.jp>,
	linux-kernel@...r.kernel.org,
	Byungchul Park <byungchul.park@....com>,
	Sergey Senozhatsky <sergey.senozhatsky.work@...il.com>
Subject: Re: [RFC][PATCH v8 1/2] printk: Make printk() completely async

Hello Jan,

On (03/29/16 11:14), Jan Kara wrote:
> On Fri 25-03-16 00:43:22, Sergey Senozhatsky wrote:
> > @@ -1655,6 +1670,14 @@ asmlinkage int vprintk_emit(int facility, int level,
> >  	raw_spin_lock(&logbuf_lock);
> >  	logbuf_cpu = this_cpu;
> >  
> > +	/*
> > +	 * Set printing kthread sleep condition early, under the
> > +	 * logbuf_lock, so it (if RUNNING) will go to console_lock()
> > +	 * and spin on logbuf_lock.
> > +	 */
> > +	if (!in_panic && printk_kthread && !need_flush_console)
> > +		need_flush_console = true;
> > +
> >  	if (unlikely(recursion_bug)) {
> >  		static const char recursion_msg[] =
> >  			"BUG: recent printk recursion!";
> 
> I like the cleaned up need_flush_console handling! Just a suggestion: It
> may be more logical to handle need_flush_console setting under logbuf_lock
> but after we actually store the message in the buffer, not before. Doesn't
> matter for correctness now but may be more future-proof.

I wanted to set it asap (when we know for sure that we will log_store()),
to have a bit more chances to avoid schedule() in printing thread. well,
that's just several log_store()-s, not a big deal, so I guess I can move
it to:

+	if (...)
+		need_flush_console = true;
	logbuf_cpu = UINT_MAX;
	raw_spin_unlock(&logbuf_lock);
	lockdep_on();
	local_irq_restore(flags);

	-ss