Re: 2.6.18 perfmon new code base + libpfm + pfmon

[Andrew Morton <akpm@osdl.org>]

On Wed, 27 Sep 2006 23:49:49 -0700
Stephane Eranian <eranian@hpl.hp.com> wrote:

> Andrew,
> 
> On Wed, Sep 27, 2006 at 04:31:00PM -0700, Andrew Morton wrote:
> > > 
> > > Here is the summary of the various point raised by your review and the current
> > > status. I am hoping to close all points by next release.
> > > 
> > > ...
> > > 
> > > [akpm]: use fget_light() in some place instead of fget()
> > > 	- not sure understand when to use one versus the other
> > >
> > 
> > They are always interchangeable.  fget_light() is simply an optimised,
> > messier-to-use version.
> 
> What are exactly the assumptions of fget_light()?

Just the ones in the comment, really - it assumes that the caller has a
single ref on the file.  ie: we're within a system call which passed in an
fd, or within a pagefault and this file* is current->mm->some_vma->vm_file.

You _might_ get into trouble if using it against some file* which belongs
to a different process, and if you don't already have a ref on that file*. 
But that's usually a bug anyway, unless the code is aware of and is
exploiting the RCU management of these things.

There are lots of example usages around the place.

> > 
> > >
> > > ..
> > >
> > > [akpm]: carta_random32() should be in another header file
> > > 	- yes, I know. Should I create a specific header file? I don't think random.h
> > > 	  is meant for this.
> > 
> > I suppose so.  Or just stick the declaration into kernel.h.
> > 
> > I had a patch go past the other day which had a hand-rolled
> > fast-but-not-very-good pseudo random number generator in it.  I couldn't
> > remember where I'd seen one, and now I can't remember what patch it was
> > that needed it.  Sigh.
> > 
> > Anyway, a standalone patch which adds that function into lib/whatever.c
> > would be nice.
> 
> I will post a standalone patch for carta random. I can provide a standalone header
> file in include/linux/carta_random.h. 

Thanks.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/