POSIX_FADV_DONTNEED

POSIX_FADV_DONTNEED

[Trevor Woollacott [ MTN - Innovation Centre ]]

Hi all

Does anyone know if posix_fadvise() with POSIX_FADV_DONTNEED is
blocking/unblocking?

Thanks,
Trevor

NOTE: This e-mail message is subject to the MTN Group disclaimer see http://www.mtn.co.za/default.aspx?pid=34411 

Re: POSIX_FADV_DONTNEED

[Andrew Morton]

On Fri, 1 Aug 2008 17:05:39 +0200 "Trevor Woollacott [ MTN - Innovation Centre ]" <Woolla_T@mtn.co.za> wrote:

> 
> Hi all
> 
> Does anyone know if posix_fadvise() with POSIX_FADV_DONTNEED is
> blocking/unblocking?

It tries to be nonblocking and will mostly succeed in that.  It might
occasionally block on a metadata read or journal congestion or
something, so it isn't guaranteed.

RE: POSIX_FADV_DONTNEED

[Trevor Woollacott [ MTN - Innovation Centre ]]

Thanks for the response Andrew. Is there a way to check when the cached
data has been freed after fadvise has been called?

-----Original Message-----
From: Andrew Morton [mailto:akpm@linux-foundation.org] 
Sent: Thursday, 14 August 2008 11:17 AM
To: Trevor Woollacott [ MTN - Innovation Centre ]
Cc: linux-kernel@vger.kernel.org
Subject: Re: POSIX_FADV_DONTNEED

On Fri, 1 Aug 2008 17:05:39 +0200 "Trevor Woollacott [ MTN - Innovation
Centre ]" <Woolla_T@mtn.co.za> wrote:

> 
> Hi all
> 
> Does anyone know if posix_fadvise() with POSIX_FADV_DONTNEED is
> blocking/unblocking?

It tries to be nonblocking and will mostly succeed in that.  It might
occasionally block on a metadata read or journal congestion or
something, so it isn't guaranteed.


NOTE: This e-mail message is subject to the MTN Group disclaimer see http://www.mtn.co.za/default.aspx?pid=34411 

Re: POSIX_FADV_DONTNEED

[Andrew Morton]

On Thu, 14 Aug 2008 12:26:28 +0200 "Trevor Woollacott [ MTN - Innovation Centre ]" <Woolla_T@mtn.co.za> wrote:

> 
> > On Fri, 1 Aug 2008 17:05:39 +0200 "Trevor Woollacott [ MTN - Innovation
> > Centre ]" <Woolla_T@mtn.co.za> wrote:

(please don't top-post).

> > > 
> > > Hi all
> > > 
> > > Does anyone know if posix_fadvise() with POSIX_FADV_DONTNEED is
> > > blocking/unblocking?
> > 
> > It tries to be nonblocking and will mostly succeed in that.  It might
> > occasionally block on a metadata read or journal congestion or
> > something, so it isn't guaranteed.
> >
>
> Thanks for the response Andrew. Is there a way to check when the cached
> data has been freed after fadvise has been called?
> 

I assume you meant "whether", not "when".

Not directly.  I believe that the best way of obtaining that
information remains mmap()ing the file then querying the presence of
pages via mincore().

RE: POSIX_FADV_DONTNEED

[Trevor Woollacott [ MTN - Innovation Centre ]]

> On Thu, 14 Aug 2008 12:26:28 +0200 "Trevor Woollacott [ MTN -
Innovation Centre ]" <Woolla_T@mtn.co.za> wrote:
>
> > 
> > > On Fri, 1 Aug 2008 17:05:39 +0200 "Trevor Woollacott [ MTN - 
> > > Innovation Centre ]" <Woolla_T@mtn.co.za> wrote:
> 
> (please don't top-post).
> 
> > > > 
> > > > Hi all
> > > > 
> > > > Does anyone know if posix_fadvise() with POSIX_FADV_DONTNEED is 
> > > > blocking/unblocking?
> > > 
> > > It tries to be nonblocking and will mostly succeed in that.  It 
> > > might occasionally block on a metadata read or journal congestion 
> > > or something, so it isn't guaranteed.
> > >
> >
> > Thanks for the response Andrew. Is there a way to check when the 
> > cached data has been freed after fadvise has been called?
> > 
>
> I assume you meant "whether", not "when".
>
> Not directly.  I believe that the best way of obtaining that 
> information remains mmap()ing the file then querying the presence of 
> pages via mincore().

Thanks for the advice

NOTE: This e-mail message is subject to the MTN Group disclaimer see http://www.mtn.co.za/default.aspx?pid=34411