[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <alpine.LNX.2.00.1612100948260.22611@nippy.intranet>
Date: Sat, 10 Dec 2016 09:50:09 +1100 (AEDT)
From: Finn Thain <fthain@...egraphics.com.au>
To: Geert Uytterhoeven <geert@...ux-m68k.org>
cc: Greg Ungerer <gerg@...ux-m68k.org>, Sam Creasey <sammy@...my.net>,
Joshua Thompson <funaho@...ai.org>,
linux-m68k@...ts.linux-m68k.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH 18/22] m68k/mm: kmap - Modernize printing of kernel
messages
On Thu, 8 Dec 2016, I wrote:
>
> On Wed, 7 Dec 2016, Geert Uytterhoeven wrote:
>
> > - Convert from printk() to pr_*(),
> > - Add missing continuations,
> > - Remove #undef DEBUG.
> >
> > Note that "#ifdef DEBUG" is sometimes retained because pr_cont() is
> > not optimized away when debugging is disabled.
> >
>
> I think that argues for using printk(KERN_DEBUG ...) and print(KERN_CONT
> ...) inside #ifdef DEBUG, which would need no explanation.
>
> If instead you use a combination of pr_debug and pr_cont and #ifdef
> DEBUG, perhaps the explanation should be moved from the commit log to a
> comment in the code?
>
Perhaps a better solution than these alternatives would be,
#if defined(DEBUG)
#define pr_debug_cont pr_cont
#else
#define pr_debug_cont no_printk
#endif
But this API is still surprising and ugly. It doesn't work with
CONFIG_DYNAMIC_DEBUG but that's not so important.
IMO, a far better linux/printk.h would have provided us with these
definitions:
#define pr_emerg(fmt, ...) \
printk(KERN_EMERG pr_fmt(fmt), ##__VA_ARGS__)
/* ... */
#define pr_debug(fmt, ...) \
printk(KERN_DEBUG pr_fmt(fmt), ##__VA_ARGS__)
#define pr_cont(fmt, ...) \
printk(KERN_CONT fmt, ##__VA_ARGS__)
#if defined(CONFIG_DYNAMIC_DEBUG)
#define pr_debug_cond(fmt, ...) \
dynamic_pr_debug(fmt, ##__VA_ARGS__)
#define pr_cont_cond(fmt, ...) \
dynamic_pr_cont(fmt, ##__VA_ARGS__)
#elif defined(DEBUG)
#define pr_debug_cond(fmt, ...) \
printk(KERN_DEBUG pr_fmt(fmt), ##__VA_ARGS__)
#define pr_cont_cond(fmt, ...) \
no_printk(KERN_CONT pr_fmt(fmt), ##__VA_ARGS__)
#else
#define pr_debug_cond(fmt, ...) \
no_printk(KERN_DEBUG pr_fmt(fmt), ##__VA_ARGS__)
#define pr_cont_cond(fmt, ...) \
no_printk(KERN_CONT pr_fmt(fmt), ##__VA_ARGS__)
#endif
Which have the virtues of symmetry and least surprise.
--
Powered by blists - more mailing lists