public inbox for linux-nfs@vger.kernel.org
 help / color / mirror / Atom feed
From: "J. Bruce Fields" <bfields@fieldses.org>
To: andros@netapp.com
Cc: linux-nfs@vger.kernel.org, pnfs@linux-nfs.org
Subject: Re: [pnfs] [PATCH 2/5] nfsd41: change from page to memory based drc limits
Date: Thu, 16 Jul 2009 14:50:51 -0400	[thread overview]
Message-ID: <20090716185051.GA2495@fieldses.org> (raw)
In-Reply-To: <20090716141748.GC30477@fieldses.org>

On Thu, Jul 16, 2009 at 10:17:48AM -0400, J. Bruce Fields wrote:
> On Wed, Jun 24, 2009 at 03:37:46PM -0400, andros@netapp.com wrote:
> > From: Andy Adamson <andros@netapp.com>
> > 
> > NFSD_SLOT_CACHE_SIZE is the size of all encoded operation responses (excluding
> > the sequence operation) that we want to cache.
> > 
> > Adjust NFSD_DRC_SIZE_SHIFT to reflect using 512 bytes instead of PAGE_SIZE.
> > 
> > Signed-off-by: Andy Adamson <andros@netapp.com>
> > ---
> >  fs/nfsd/nfs4state.c        |   30 ++++++++++++++----------------
> >  fs/nfsd/nfssvc.c           |   17 +++++++++--------
> >  include/linux/nfsd/nfsd.h  |    4 ++--
> >  include/linux/nfsd/state.h |    1 +
> >  4 files changed, 26 insertions(+), 26 deletions(-)
> > 
> > diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c
> > index 2e6a44e..b4a536d 100644
> > --- a/fs/nfsd/nfs4state.c
> > +++ b/fs/nfsd/nfs4state.c
> > @@ -414,35 +414,33 @@ gen_sessionid(struct nfsd4_session *ses)
> >  
> >  /*
> >   * Give the client the number of slots it requests bound by
> > - * NFSD_MAX_SLOTS_PER_SESSION and by sv_drc_max_pages.
> > + * NFSD_MAX_SLOTS_PER_SESSION and by nfsd_drc_max_mem.
> >   *
> > - * If we run out of pages (sv_drc_pages_used == sv_drc_max_pages) we
> > - * should (up to a point) re-negotiate active sessions and reduce their
> > - * slot usage to make rooom for new connections. For now we just fail the
> > - * create session.
> > + * If we run out of reserved DRC memory we should (up to a point) re-negotiate
> > + * active sessions and reduce their slot usage to make rooom for new
> > + * connections. For now we just fail the create session.
> >   */
> >  static int set_forechannel_maxreqs(struct nfsd4_channel_attrs *fchan)
> >  {
> > -	int status = 0, np = fchan->maxreqs * NFSD_PAGES_PER_SLOT;
> > +	int mem;
> >  
> >  	if (fchan->maxreqs < 1)
> >  		return nfserr_inval;
> >  	else if (fchan->maxreqs > NFSD_MAX_SLOTS_PER_SESSION)
> >  		fchan->maxreqs = NFSD_MAX_SLOTS_PER_SESSION;
> >  
> > +	mem = fchan->maxreqs * NFSD_SLOT_CACHE_SIZE;
> > +
> >  	spin_lock(&nfsd_drc_lock);
> > -	if (np + nfsd_drc_pages_used > nfsd_drc_max_pages)
> > -		np = nfsd_drc_max_pages - nfsd_drc_pages_used;
> > -	nfsd_drc_pages_used += np;
> > +	if (mem + nfsd_drc_mem_used > nfsd_drc_max_mem)
> > +		mem = nfsd_drc_max_mem - nfsd_drc_mem_used;
> > +	nfsd_drc_mem_used += mem;
> 
> There's a small bug here: in the case where
> 
> 	0 < mem < NFSD_PAGES_PER_SLOT,
> 
> we add mem into nfsd_drc_mem_used, but never subtract it back off.

(Actually this is true whenever there's roundoff error, since we're
adding in the un-rounded-off version but subtracting off (in the next
patch) the rounded-off version.)

--b.

> 
> >  	spin_unlock(&nfsd_drc_lock);
> >  
> > -	if (np <= 0) {
> > -		status = nfserr_resource;
> > -		fchan->maxreqs = 0;
> > -	} else
> > -		fchan->maxreqs = np / NFSD_PAGES_PER_SLOT;
> > -
> > -	return status;
> > +	fchan->maxreqs = mem / NFSD_SLOT_CACHE_SIZE;
> > +	if (fchan->maxreqs == 0)
> > +		return nfserr_resource;
> > +	return 0;
> >  }
> >  
> >  /*
> > diff --git a/fs/nfsd/nfssvc.c b/fs/nfsd/nfssvc.c
> > index 78d8fcd..1793dba 100644
> > --- a/fs/nfsd/nfssvc.c
> > +++ b/fs/nfsd/nfssvc.c
> > @@ -74,8 +74,8 @@ struct svc_serv 		*nfsd_serv;
> >   * nfsd_drc_pages_used tracks the current version 4.1 DRC memory usage.
> >   */
> >  spinlock_t	nfsd_drc_lock;
> > -unsigned int	nfsd_drc_max_pages;
> > -unsigned int	nfsd_drc_pages_used;
> > +unsigned int	nfsd_drc_max_mem;
> > +unsigned int	nfsd_drc_mem_used;
> >  
> >  #if defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL)
> >  static struct svc_stat	nfsd_acl_svcstats;
> > @@ -247,13 +247,14 @@ void nfsd_reset_versions(void)
> >  static void set_max_drc(void)
> >  {
> >  	/* The percent of nr_free_buffer_pages used by the V4.1 server DRC */
> > -	#define NFSD_DRC_SIZE_SHIFT	7
> > -	nfsd_drc_max_pages = nr_free_buffer_pages()
> > -						>> NFSD_DRC_SIZE_SHIFT;
> > -	nfsd_drc_pages_used = 0;
> > +	#define NFSD_DRC_SIZE_SHIFT	10
> > +	nfsd_drc_max_mem = (nr_free_buffer_pages()
> > +					>> NFSD_DRC_SIZE_SHIFT) * PAGE_SIZE;
> > +	nfsd_drc_mem_used = 0;
> >  	spin_lock_init(&nfsd_drc_lock);
> > -	dprintk("%s nfsd_drc_max_pages %u\n", __func__,
> > -		nfsd_drc_max_pages);
> > +	dprintk("%s nfsd_drc_max_mem %u [in pages %lu]\n", __func__,
> > +		nfsd_drc_max_mem,
> > +		nfsd_drc_max_mem / PAGE_SIZE);
> 
> The reader can do the "in pages" for him/herself if necessary.
> 
> >  }
> >  
> >  int nfsd_create_serv(void)
> > diff --git a/include/linux/nfsd/nfsd.h b/include/linux/nfsd/nfsd.h
> > index 2571f85..2812ed5 100644
> > --- a/include/linux/nfsd/nfsd.h
> > +++ b/include/linux/nfsd/nfsd.h
> > @@ -57,8 +57,8 @@ extern u32			nfsd_supported_minorversion;
> >  extern struct mutex		nfsd_mutex;
> >  extern struct svc_serv		*nfsd_serv;
> >  extern spinlock_t		nfsd_drc_lock;
> > -extern unsigned int		nfsd_drc_max_pages;
> > -extern unsigned int		nfsd_drc_pages_used;
> > +extern unsigned int		nfsd_drc_max_mem;
> > +extern unsigned int		nfsd_drc_mem_used;
> >  
> >  extern struct seq_operations nfs_exports_op;
> >  
> > diff --git a/include/linux/nfsd/state.h b/include/linux/nfsd/state.h
> > index f5a95fd..10b6b30 100644
> > --- a/include/linux/nfsd/state.h
> > +++ b/include/linux/nfsd/state.h
> > @@ -98,6 +98,7 @@ struct nfs4_cb_conn {
> >  /* Maximum number of pages per slot cache entry */
> >  #define NFSD_PAGES_PER_SLOT	1
> >  /* Maximum number of operations per session compound */
> > +#define NFSD_SLOT_CACHE_SIZE		512
> 
> Whoops--the above comment's misplaced now.
> 
> Also, this patch leaves the code in an inconsistent state where some
> places still uses page size, others don't.  (E.g. we promise to cache
> page_size, and still will, I think, and allocate that much, but we
> calculate the maximum requests by the new smaller 512 byte value.)
> 
> Either fix all that at once, or just keep NFSD_SLOT_CACHE_SIZE 4k for
> now.
> 
> --b.
> 
> >  #define NFSD_MAX_OPS_PER_COMPOUND	16
> >  
> >  struct nfsd4_cache_entry {
> > -- 
> > 1.5.4.3
> > 
> > _______________________________________________
> > pNFS mailing list
> > pNFS@linux-nfs.org
> > http://linux-nfs.org/cgi-bin/mailman/listinfo/pnfs
> --
> To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

  reply	other threads:[~2009-07-16 18:50 UTC|newest]

Thread overview: 26+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2009-06-24 19:37 [PATCH 0/14] NFSv4.1 Server DRC rewrite Version 4 andros
2009-06-24 19:37 ` [PATCH 1/5] nfsd41: use globals for DRC limits andros
2009-06-24 19:37   ` [PATCH 2/5] nfsd41: change from page to memory based drc limits andros
2009-06-24 19:37     ` [PATCH 3/5] nfsd41: reclaim DRC memory on session free andros
2009-06-24 19:37       ` [PATCH 4/5] nfsd41: set the session maximum response size cached andros
2009-06-24 19:37         ` [PATCH 5/5] nfsd41: remove redundant forechannel max requests check andros
2009-06-24 19:37           ` [PATCH 06/14] nfsd41: change check_slot_seqid parameters andros
2009-06-24 19:37             ` [PATCH 07/14] nfsd41: replace create_session DRC with xdr structure andros
2009-06-24 19:37               ` [PATCH 08/14] nfsd41: replace page based DRC with buffer based DRC andros
2009-06-24 19:37                 ` [PATCH 09/14] nfsd41: rename nfsd4_enc_uncached_replay andros
2009-06-24 19:37                   ` [PATCH 10/14] nfsd41: encode replay sequence from the slot values andros
2009-06-24 19:37                     ` [PATCH 11/14] nfsd41: fix nfsd4_replay_cache_entry comments andros
2009-06-24 19:37                       ` [PATCH 12/14] nfsd41: support 16 slots per session andros
2009-06-24 19:37                         ` [PATCH 13/14] nfsd41: add test for failed sequence operation andros
2009-06-24 19:37                           ` [PATCH 14/14] nfsd41: remove redundant failed sequence check andros
2009-07-01 23:09         ` [pnfs] [PATCH 4/5] nfsd41: set the session maximum response sizecached Alexandros Batsakis
2009-07-06 19:27           ` Andy Adamson
2009-07-06 21:48             ` J. Bruce Fields
2009-07-16 14:18       ` [pnfs] [PATCH 3/5] nfsd41: reclaim DRC memory on session free J. Bruce Fields
2009-07-16 15:49         ` William A. (Andy) Adamson
2009-07-27 22:57         ` J. Bruce Fields
2009-07-31  8:38           ` Andy Adamson
2009-07-16 14:17     ` [pnfs] [PATCH 2/5] nfsd41: change from page to memory based drc limits J. Bruce Fields
2009-07-16 18:50       ` J. Bruce Fields [this message]
2009-07-16 14:24     ` J. Bruce Fields
2009-06-28 16:01 ` [pnfs] [PATCH 0/14] NFSv4.1 Server DRC rewrite Version 4 Benny Halevy

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20090716185051.GA2495@fieldses.org \
    --to=bfields@fieldses.org \
    --cc=andros@netapp.com \
    --cc=linux-nfs@vger.kernel.org \
    --cc=pnfs@linux-nfs.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox