Re: [PATCH 1/2] mm: Avoid putting a bad page back on the LRU

[Russ Anderson <rja@sgi.com>]

On Tue, Apr 29, 2008 at 02:33:29PM -0700, Andrew Morton wrote:
> On Mon, 28 Apr 2008 14:23:58 -0500
> Russ Anderson <rja@sgi.com> wrote:
> 
> > --- test.orig/include/linux/kernel.h	2008-04-25 15:22:07.640501421 -0500
> > +++ test/include/linux/kernel.h	2008-04-25 15:42:02.826266036 -0500
> > @@ -412,6 +412,7 @@ struct sysinfo {
> >  	unsigned long totalhigh;	/* Total high memory size */
> >  	unsigned long freehigh;		/* Available high memory size */
> >  	unsigned int mem_unit;		/* Memory unit size in bytes */
> > +	unsigned int badpages;		/* Number of bad (PG_memerror) pages */
> >  	char _f[20-2*sizeof(long)-sizeof(int)];	/* Padding: libc5 uses this.. */
> >  };
> 
> eek.  sys_sysinfo() copies this structure out to userspace verbatim. 
> Increasing its size by four bytes will cause memory scribbles in userspace
> applications.
> 
> That's fixable by suitably adjusting sysinfo._f, but given the lack of
> version information in the sysinfo struct, I'm not sure how useful this
> change really is.

OK.  The only reason badpages was added was to print the number
of page pages in /proc/meminfo.  I'll go with plan B, a badmem driver 
for displaying bad page info. 

-- 
Russ Anderson, OS RAS/Partitioning Project Lead  
SGI - Silicon Graphics Inc          rja@sgi.com