Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Date:   Thu, 24 Sep 2020 18:06:53 +0900
From:   Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
To:     Petr Mladek <pmladek@suse.com>
Cc:     Sergey Senozhatsky <sergey.senozhatsky@gmail.com>,
        John Ogness <john.ogness@linutronix.de>,
        Sergey Senozhatsky <sergey.senozhatsky.work@gmail.com>,
        Steven Rostedt <rostedt@goodmis.org>,
        Linus Torvalds <torvalds@linux-foundation.org>,
        Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
        Thomas Gleixner <tglx@linutronix.de>,
        linux-kernel@vger.kernel.org
Subject: Re: [PATCH printk 3/5] printk: use buffer pool for sprint buffers
Message-ID: <20200924090653.GC541@jagdpanzerIV.localdomain>
References: <20200922153816.5883-1-john.ogness@linutronix.de>
 <20200922153816.5883-4-john.ogness@linutronix.de>
 <20200924061746.GF577@jagdpanzerIV.localdomain>
 <20200924085445.GK6442@alley>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20200924085445.GK6442@alley>
Precedence: bulk

On (20/09/24 10:54), Petr Mladek wrote:
> > Just a question:
> > 
> > Can dynamic_textbuf be a PER_CPU array of five textbuf[1024] buffers
> > (for normal printk, nmi, hard irq, soft irq and one extra buffer for
> > recursive printk calls)?
> 
> That would be my preferred fallback when the approach with
> vsprintf(NULL, ) is not acceptable for some reasons.

OK.

> But I still think that calling vsprintf(NULL, ) is the most trivial
> and good enough solution.

It's probably good enough.

> IMHO, the solution with per-CPU buffers is not that trivial, for
> example:
> 
>   What if recursive printk() is interrupted by NMI and it causes
>   yet another recursion?
> 
>   Is one level of recursion enough?

We might try the current approach - when we, for example, have
recursion in printk_safe() we just end up writing data to the
same per-CPU buffer. We need to limit the depth of recursion
one way or another. With per-CPU counter we will just bail out
of "deeply recursive printk" without attempting to store its
messages; with the buffers approach we will write the data to
a static buffer and see how badly it will be overlapped at the
end. Just a thought.

	-ss