Re: [PATCH 1/3] Add ratelimited printk for different alert levels

[Dave Chinner <david@fromorbit.com>]

On Tue, Sep 11, 2012 at 08:22:39PM -0700, Joe Perches wrote:
> On Wed, 2012-09-12 at 03:43 +0530, raghu.prabhu13@gmail.com wrote:
> > Ratelimited printk will be useful in printing xfs messages which are otherwise
> > not required to be printed always due to their high rate (to prevent kernel ring
> > buffer from overflowing), while at the same time required to be printed.
> []
> > diff --git a/fs/xfs/xfs_message.h b/fs/xfs/xfs_message.h
> []
> > @@ -30,6 +32,32 @@ void xfs_debug(const struct xfs_mount *mp, const char *fmt, ...)
> >  }
> >  #endif
> >  
> > +#define xfs_printk_ratelimited(xfs_printk, dev, fmt, ...)		\
> > +do {									\
> > +	static DEFINE_RATELIMIT_STATE(_rs,				\
> > +				      DEFAULT_RATELIMIT_INTERVAL,	\
> > +				      DEFAULT_RATELIMIT_BURST);		\
> > +	if (__ratelimit(&_rs))						\
> > +		xfs_printk(dev, fmt, ##__VA_ARGS__);			\
> > +} while (0)
> 
> It might be better to use an xfs singleton RATELIMIT_STATE
> 
> DEFINE_RATELIMIT_STATE(xfs_rs);
> ...
> #define xfs_printk_ratelimited(xfs_printk, dev, fmt, ...)		\
> do {									\
> 	if (__ratelimit(&xfs_rs))					\
> 		xfs_printk(dev, fmt, ##__VA_ARGS__);			\
> } while (0)

Which would then result in ratelimiting dropping potentially
important, unique messages. I think it's much better to guarantee
ratelimited messages get emitted at least once, especially as there
is the potential for multiple filesystems to emit messages
simultaneously.

I think per-location rate limiting is fine for the current usage -
ratelimiting is not widespread so there isn't a massive increase in
size as a result of this. If we do start to use ratelimiting in lots
of places in XFS, then we might have to revisit this, but it's OK
for now.

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com