Re-implement the client side of rpc.lockd in the kernel. This implementation

provides the correct semantics for flock(2) style locks which are used by the
lockf(1) command line tool and the pidfile(3) library. It also implements
recovery from server restarts and ensures that dirty cache blocks are written
to the server before obtaining locks (allowing multiple clients to use file
locking to safely share data).

Sponsored by:	Isilon Systems
PR:		94256
MFC after:	2 weeks

6 files changed, 28 insertions, 2 deletions

diff --git a/sys/nfsclient/nfs.h b/sys/nfsclient/nfs.h
index 29fb332..9e52420 100644
--- a/sys/nfsclient/nfs.h
+++ b/sys/nfsclient/nfs.h
@@ -93,6 +93,7 @@
 #define	NFSSTA_SNDLOCK		0x01000000  /* Send socket lock */
 #define	NFSSTA_WANTSND		0x02000000  /* Want above */
 #define	NFSSTA_TIMEO		0x10000000  /* Experiencing a timeout */
+#define	NFSSTA_LOCKTIMEO	0x20000000  /* Experiencing a lockd timeout */
 
 
 /*
diff --git a/sys/nfsclient/nfs_node.c b/sys/nfsclient/nfs_node.c
index 7876d32..03e672a 100644
--- a/sys/nfsclient/nfs_node.c
+++ b/sys/nfsclient/nfs_node.c
@@ -234,6 +234,13 @@ nfs_reclaim(struct vop_reclaim_args *ap)
 		vprint("nfs_reclaim: pushing active", vp);
 
 	/*
+	 * If the NLM is running, give it a chance to abort pending
+	 * locks.
+	 */
+	if (nfs_reclaim_p)
+		nfs_reclaim_p(ap);
+
+	/*
 	 * Destroy the vm object and flush associated pages.
 	 */
 	vnode_destroy_vobject(vp);
diff --git a/sys/nfsclient/nfs_vfsops.c b/sys/nfsclient/nfs_vfsops.c
index f342211..17536491 100644
--- a/sys/nfsclient/nfs_vfsops.c
+++ b/sys/nfsclient/nfs_vfsops.c
@@ -495,6 +495,7 @@ nfs_mountroot(struct mount *mp, struct thread *td)
 		(l >> 24) & 0xff, (l >> 16) & 0xff,
 		(l >>  8) & 0xff, (l >>  0) & 0xff, nd->root_hostnam);
 	printf("NFS ROOT: %s\n", buf);
+	nd->root_args.hostname = buf;
 	if ((error = nfs_mountdiskless(buf,
 	    &nd->root_saddr, &nd->root_args, td, &vp, mp)) != 0) {
 		return (error);
@@ -540,6 +541,7 @@ nfs_decode_args(struct mount *mp, struct nfsmount *nmp, struct nfs_args *argp)
 	int s;
 	int adjsock;
 	int maxio;
+	char *p;
 
 	s = splnet();
 
@@ -699,6 +701,11 @@ nfs_decode_args(struct mount *mp, struct nfsmount *nmp, struct nfs_args *argp)
 				(void) tsleep((caddr_t)&lbolt, PSOCK, "nfscon", 0);
 			}
 	}
+
+	strlcpy(nmp->nm_hostname, argp->hostname, sizeof(nmp->nm_hostname));
+	p = strchr(nmp->nm_hostname, ':');
+	if (p)
+		*p = '\0';
 }
 
 static const char *nfs_opts[] = { "from", "nfs_args",
diff --git a/sys/nfsclient/nfs_vnops.c b/sys/nfsclient/nfs_vnops.c
index 524c564..3711165 100644
--- a/sys/nfsclient/nfs_vnops.c
+++ b/sys/nfsclient/nfs_vnops.c
@@ -198,6 +198,8 @@ struct mtx 	nfs_iod_mtx;
 struct proc	*nfs_iodwant[NFS_MAXASYNCDAEMON];
 struct nfsmount *nfs_iodmount[NFS_MAXASYNCDAEMON];
 int		 nfs_numasync = 0;
+vop_advlock_t	*nfs_advlock_p = nfs_dolock;
+vop_reclaim_t	*nfs_reclaim_p = NULL;
 #define	DIRHDSIZ	(sizeof (struct dirent) - (MAXNAMLEN + 1))
 
 SYSCTL_DECL(_vfs_nfs);
@@ -3051,8 +3053,13 @@ nfs_advlock(struct vop_advlock_args *ap)
 		size = VTONFS(vp)->n_size;
 		VOP_UNLOCK(vp, 0);
 		error = lf_advlock(ap, &(vp->v_lockf), size);
-	} else
-		error = nfs_dolock(ap);
+	} else {
+		if (nfs_advlock_p)
+			error = nfs_advlock_p(ap);
+		else
+			error = ENOLCK;
+	}
+
 	return (error);
 }
 
diff --git a/sys/nfsclient/nfsmount.h b/sys/nfsclient/nfsmount.h
index 8615846..6fa7f8b 100644
--- a/sys/nfsclient/nfsmount.h
+++ b/sys/nfsclient/nfsmount.h
@@ -91,6 +91,7 @@ struct	nfsmount {
 	int	nm_tprintf_initial_delay;	/* initial delay */
 	int	nm_tprintf_delay;		/* interval for messages */
 	struct nfs_tcp_mountstate nm_nfstcpstate;
+	char	nm_hostname[MNAMELEN];	 /* server's name */
 
 	/* NFSv4 */
 	uint64_t nm_clientid;
diff --git a/sys/nfsclient/nfsnode.h b/sys/nfsclient/nfsnode.h
index f227361..03e5e7f 100644
--- a/sys/nfsclient/nfsnode.h
+++ b/sys/nfsclient/nfsnode.h
@@ -187,6 +187,9 @@ extern	struct vop_vector	nfs4_vnodeops;
 extern struct buf_ops buf_ops_nfs;
 extern struct buf_ops buf_ops_nfs4;
 
+extern vop_advlock_t *nfs_advlock_p;
+extern vop_reclaim_t *nfs_reclaim_p;
+
 /*
  * Prototypes for NFS vnode operations
  */