[PATCH 1/22] lockd: make NLM_HOST_MAX tunable

[PATCH 1/22] lockd: make NLM_HOST_MAX tunable

[Olaf Kirch]

From: Olaf Kirch <okir@suse.de>
Subject: lockd: make NLM_HOST_MAX tunable
Patch-mainline: not-yet

  This patch makes NLM_HOST_MAX tunable, and exports the
  new variable as a sysctl.
  The old value (64 clients) just wasn't adequate anymore.

Signed-off-by: Olaf Kirch <okir@suse.de>

 fs/lockd/host.c             |    7 +++----
 fs/lockd/svc.c              |   10 ++++++++++
 include/linux/lockd/lockd.h |    1 +
 3 files changed, 14 insertions(+), 4 deletions(-)

Index: build/fs/lockd/svc.c
===================================================================
--- build.orig/fs/lockd/svc.c
+++ build/fs/lockd/svc.c
@@ -59,6 +59,7 @@ static DECLARE_WAIT_QUEUE_HEAD(lockd_exi
 static unsigned long		nlm_grace_period;
 static unsigned long		nlm_timeout = LOCKD_DFLT_TIMEO;
 static int			nlm_udpport, nlm_tcpport;
+int				nlm_max_hosts = 256;
 
 /*
  * Constants needed for the sysctl interface.
@@ -367,6 +368,15 @@ static ctl_table nlm_sysctls[] = {
 		.extra1		= (int *) &nlm_port_min,
 		.extra2		= (int *) &nlm_port_max,
 	},
+	{
+		.ctl_name	= CTL_UNNUMBERED,
+		.procname	= "nlm_max_hosts",
+		.data		= &nlm_max_hosts,
+		.maxlen		= sizeof(int),
+		.mode		= 0644,
+		.proc_handler	= &proc_dointvec,
+	},
+
 	{ .ctl_name = 0 }
 };
 
Index: build/include/linux/lockd/lockd.h
===================================================================
--- build.orig/include/linux/lockd/lockd.h
+++ build/include/linux/lockd/lockd.h
@@ -144,6 +144,7 @@ extern struct svc_procedure	nlmsvc_proce
 #endif
 extern int			nlmsvc_grace_period;
 extern unsigned long		nlmsvc_timeout;
+extern int			nlm_max_hosts;
 
 /*
  * Lockd client functions
Index: build/fs/lockd/host.c
===================================================================
--- build.orig/fs/lockd/host.c
+++ build/fs/lockd/host.c
@@ -20,12 +20,11 @@
 
 
 #define NLMDBG_FACILITY		NLMDBG_HOSTCACHE
-#define NLM_HOST_MAX		64
 #define NLM_HOST_NRHASH		32
 #define NLM_ADDRHASH(addr)	(ntohl(addr) & (NLM_HOST_NRHASH-1))
 #define NLM_HOST_REBIND		(60 * HZ)
-#define NLM_HOST_EXPIRE		((nrhosts > NLM_HOST_MAX)? 300 * HZ : 120 * HZ)
-#define NLM_HOST_COLLECT	((nrhosts > NLM_HOST_MAX)? 120 * HZ :  60 * HZ)
+#define NLM_HOST_EXPIRE		((nrhosts > nlm_max_hosts)? 300 * HZ : 120 * HZ)
+#define NLM_HOST_COLLECT	((nrhosts > nlm_max_hosts)? 120 * HZ :  60 * HZ)
 #define NLM_HOST_ADDR(sv)	(&(sv)->s_nlmclnt->cl_xprt->addr)
 
 static struct nlm_host *	nlm_hosts[NLM_HOST_NRHASH];
@@ -128,7 +127,7 @@ nlm_lookup_host(int server, struct socka
 	INIT_LIST_HEAD(&host->h_granted);
 	INIT_LIST_HEAD(&host->h_reclaim);
 
-	if (++nrhosts > NLM_HOST_MAX)
+	if (++nrhosts > nlm_max_hosts)
 		next_gc = 0;
 
 nohost:

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys -- and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs

Re: [PATCH 1/22] lockd: make NLM_HOST_MAX tunable

[Neil Brown]

On Saturday August 5, okir@suse.de wrote:
> From: Olaf Kirch <okir@suse.de>
> Subject: lockd: make NLM_HOST_MAX tunable
> Patch-mainline: not-yet
> 
>   This patch makes NLM_HOST_MAX tunable, and exports the
>   new variable as a sysctl.
>   The old value (64 clients) just wasn't adequate anymore.

Why do we need NLM_HOST_MAX at all?  And if we do, we need
a better name, because it isn't really a maximum.

It is more of a threshold.
While there are fewer than this mean, we garbage collect every 2
minutes discarding hosts more than 5 minutes old.
When we exceed this threshold, we garbage collect every minutes (or
every time a new host is allocated). and discard hosts older than
2 minutes.

It seems like we are picking an arbitrary point at which to switch
from one arbitrary algorithm to another.  Do we really want to do
this?

I cannot see any good reason not to just pick one arbitrary algorithm
and stick with that the whole time.
The only tunable might make sense is the max age before garbage
collection removes hosts.

Am I missing something important?

NeilBrown

-------------------------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs