Re: Print taint info in more places.

[Dave Jones <davej@redhat.com>]

On Thu, Dec 13, 2007 at 11:25:06PM -0800, Jeremy Fitzhardinge wrote:
 > Dave Jones wrote:
 > > On Fri, Dec 14, 2007 at 01:03:50AM +0100, Adrian Bunk wrote:
 > >
 > >  > >  #ifndef HAVE_ARCH_BUG
 > >  > >  #define BUG() do { \
 > >  > > -	printk("BUG: failure at %s:%d/%s()!\n", __FILE__, __LINE__, __FUNCTION__); \
 > >  > > +	printk(KERN_ERR "BUG: failure at %s:%d/%s()! (%s)\n",
 > >  > > +		__FILE__, __LINE__, __FUNCTION__, print_tainted()); \
 > >  > >  	panic("BUG!"); \
 > >  > >  } while (0)
 > >  > >  #endif
 > >  > >...
 > >  > 
 > >  > Note that this only changes a handful of architectures and most likely 
 > >  > not the ones you are interested in.
 > >
 > > Hmm, it appears that I was mistaken, and we never did patch x86.
 > > Which leaves me wondering if its worth it or not to  patch BUG()
 > > Anyways, here's the latest rev with the out-of-line changes as
 > > suggested by Andi.
 > >
 > > init/main.c may not be the best place for the ool variant. suggestions?
 > >   
 > 
 > lib/bug.c would be the place for architectures using
 > CONFIG_GENERIC_BUG.  x86 could be converted to use the BUG-trapping
 > mechanism for WARN_ON like Power does, so it would be inherently out of
 > line anyway.

The BUG()/WARN_ON() parts of my patch are becoming something of a rathole,
which I don't really have time to dig into right now, so I'm going to split
this up I think into mm/ additions, spinlockdebug additions, and bugon/warnon foo.
The three should be independant, and blocking the first two until I get
time to look at the third seems pointless.

	Dave

-- 
http://www.codemonkey.org.uk