MFC r260229, r260258, r260367, r260390, r260459, r260648:

Rework NFS Duplicate Request Cache cleanup logic.

 - Introduce additional hash to group requests by hash of sockref.  This
allows to process TCP acknowledgements without looping though all the cache,
and as result allows to do it every time.
 - Indroduce additional callbacks to notify application layer about sockets
disconnection.  Without this last few requests processed just before socket
disconnection never processed their ACKs and stuck in cache for many hours.
 - Implement transport-specific method for tracking reply acknowledgements.
New implementation does not cross multiple stack layers to get the data and
does not have race conditions that previously made some requests stuck
in cache.  This could be done more efficiently at sockbuf layer, but that
would broke some KBIs, while I don't know other consumers for it aside NFS.
 - Instead of traversing all DRC twice per request, run cleaning only once
per request, and except in some conditions traverse only single hash slot
at a time.

Together this limits NFS DRC growth only to situations of real connectivity
problems.  If network is working well, and so all replies are acknowledged,
cache remains almost empty even after hours of heavy load.  Without this
change on the same test cache was growing to many thousand requests even
with perfectly working local network.

As another result this reduces CPU time spent on the DRC handling during
SPEC NFS benchmark from about 10% to 0.5%.

Sponsored by:   iXsystems, Inc.

2 files changed, 13 insertions, 8 deletions

diff --git a/sys/fs/nfs/nfs_var.h b/sys/fs/nfs/nfs_var.h
index 7fbabaa..895829d 100644
--- a/sys/fs/nfs/nfs_var.h
+++ b/sys/fs/nfs/nfs_var.h
@@ -218,14 +218,14 @@ void nfsrvd_dorpc(struct nfsrv_descript *, int, NFSPROC_T *);
 
 /* nfs_nfsdcache.c */
 void nfsrvd_initcache(void);
-int nfsrvd_getcache(struct nfsrv_descript *, struct socket *);
-struct nfsrvcache *nfsrvd_updatecache(struct nfsrv_descript *,
-    struct socket *);
-void nfsrvd_sentcache(struct nfsrvcache *, struct socket *, int);
+int nfsrvd_getcache(struct nfsrv_descript *);
+struct nfsrvcache *nfsrvd_updatecache(struct nfsrv_descript *);
+void nfsrvd_sentcache(struct nfsrvcache *, int, uint32_t);
 void nfsrvd_cleancache(void);
 void nfsrvd_refcache(struct nfsrvcache *);
 void nfsrvd_derefcache(struct nfsrvcache *);
 void nfsrvd_delcache(struct nfsrvcache *);
+void nfsrc_trimcache(uint64_t, uint32_t, int);
 
 /* nfs_commonsubs.c */
 void newnfs_init(void);
@@ -327,9 +327,6 @@ int nfsd_checkrootexp(struct nfsrv_descript *);
 void nfscl_retopts(struct nfsmount *, char *, size_t);
 
 /* nfs_commonport.c */
-int nfsrv_checksockseqnum(struct socket *, tcp_seq);
-int nfsrv_getsockseqnum(struct socket *, tcp_seq *);
-int nfsrv_getsocksndseq(struct socket *, tcp_seq *, tcp_seq *);
 int nfsrv_lookupfilename(struct nameidata *, char *, NFSPROC_T *);
 void nfsrv_object_create(vnode_t, NFSPROC_T *);
 int nfsrv_mallocmget_limit(void);
diff --git a/sys/fs/nfs/nfsrvcache.h b/sys/fs/nfs/nfsrvcache.h
index 5c9dc57..7db3035 100644
--- a/sys/fs/nfs/nfsrvcache.h
+++ b/sys/fs/nfs/nfsrvcache.h
@@ -46,6 +46,7 @@
 /* Cache table entry. */
 struct nfsrvcache {
 	LIST_ENTRY(nfsrvcache) rc_hash;		/* Hash chain */
+	LIST_ENTRY(nfsrvcache) rc_ahash;	/* ACK hash chain */
 	TAILQ_ENTRY(nfsrvcache)	rc_lru;		/* UDP lru chain */
 	u_int32_t	rc_xid;			/* rpc id number */
 	time_t		rc_timestamp;		/* Time done */
@@ -64,6 +65,7 @@ struct nfsrvcache {
 			int16_t		refcnt;
 			u_int16_t	cksum;
 			time_t		cachetime;
+			int		acked;
 		} ot;
 	} rc_un2;
 	u_int16_t	rc_proc;		/* rpc proc number */
@@ -81,6 +83,13 @@ struct nfsrvcache {
 #define	rc_reqlen	rc_un2.ot.len
 #define	rc_cksum	rc_un2.ot.cksum
 #define	rc_cachetime	rc_un2.ot.cachetime
+#define	rc_acked	rc_un2.ot.acked
+
+/* TCP ACK values */
+#define	RC_NO_SEQ		0
+#define	RC_NO_ACK		1
+#define	RC_ACK			2
+#define	RC_NACK			3
 
 /* Return values */
 #define	RC_DROPIT		0
@@ -95,7 +104,6 @@ struct nfsrvcache {
 #define	RC_UDP		0x0010
 #define	RC_INETIPV6	0x0020
 #define	RC_INPROG	0x0040
-#define	RC_TCPSEQ	0x0080
 #define	RC_NFSV2	0x0100
 #define	RC_NFSV3	0x0200
 #define	RC_NFSV4	0x0400