[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Thu, 12 Dec 2019 10:23:19 +0900
From: Sergey Senozhatsky <sergey.senozhatsky.work@...il.com>
To: Petr Mladek <pmladek@...e.com>
Cc: Benjamin Herrenschmidt <benh@...nel.crashing.org>,
linux-kernel@...r.kernel.org,
Sergey Senozhatsky <sergey.senozhatsky@...il.com>,
Steven Rostedt <rostedt@...dmis.org>,
Linus Torvalds <torvalds@...ux-foundation.org>,
AlekseyMakarov <aleksey.makarov@...aro.org>
Subject: Re: [RFC/PATCH] printk: Fix preferred console selection with
multiple matches
On (19/12/11 10:17), Petr Mladek wrote:
> And this would exactly happen as pointed out by Sergey. match() does
> also some setup operations. I would be scared to call them twice.
>
> > This looks indeed a lot more invasive. Is there any reason why what I
> > propose wouldn't work as a first patch that can easily be backported ?
>
> Your solution is not acceptable because it might cause calling
> match() more times.
>
> The reverse search of list of console does not work for ttySX
> consoles because the number is omitted when matching. And the messages
> will appear only on the first matched serial console. There is
> a paragraph about this in the commit message of my patch.
A fast path for preferred_console match()/setup() sounds like
something that may work. There won't be double setup()-s and
we scan console list in the same direction. We just have
separate match()/setup() for preferred_console. Do you think
this won't do the trick?
-ss
Powered by blists - more mailing lists