Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 16 Oct 2009 19:30:49 +0000 (UTC)
From:      John Baldwin <jhb@FreeBSD.org>
To:        src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org
Subject:   svn commit: r198174 - head/sys/nfsclient
Message-ID:  <200910161930.n9GJUnHD011723@svn.freebsd.org>

next in thread | raw e-mail | index | archive | help
Author: jhb
Date: Fri Oct 16 19:30:48 2009
New Revision: 198174
URL: http://svn.freebsd.org/changeset/base/198174

Log:
  Close a race with caching of -ve name lookups in the NFS client.
  Specifically, clients only trust -ve cache entries while the directory
  remains unchanged and discard any -ve cache entries for a directory when
  they notice that the modification time of a directory entry changes.  The
  race involves two concurrent lookups as follows:
  - Thread A does a lookup for file 'foo' which sends a lookup RPC to the
    server.  The lookup fails and the server replies.
  - The 'foo' file is created (either by the same client or a different
    client) updating the modification time on the parent directory of 'foo'.
  - Thread B does a lookup for a different file 'bar' which updates the
    cached attributes of the parent directory of 'foo' to reflect the new
    modification time after 'foo' was created.
  - Thread A finally resumes execution to parse the reply from the NFS
    server.  It adds a -ve cache entry and sets the cached value of the
    directory's modification time that is used for invalidating -ve cached
    lookups to the new modification time set by thread B.
  
  At this point, future lookups of 'foo' will honor the -ve cached entry
  until the cached entry is pushed out of the name cache's LRU or the
  modification time of the parent directory is changed again by some other
  change.  The fix is to read the directory's modification time before
  sending the lookup RPC and use that cached modification time when setting
  the directory's cached modification time.  Also, we do not add a -ve cache
  entry if another thread has added -ve cache entry that set the directory's
  cached modification time to a newer value than the value we read before
  sending the lookup RPC.
  
  Reviewed by:	rmacklem
  MFC after:	1 week

Modified:
  head/sys/nfsclient/nfs_vnops.c

Modified: head/sys/nfsclient/nfs_vnops.c
==============================================================================
--- head/sys/nfsclient/nfs_vnops.c	Fri Oct 16 16:17:57 2009	(r198173)
+++ head/sys/nfsclient/nfs_vnops.c	Fri Oct 16 19:30:48 2009	(r198174)
@@ -924,6 +924,7 @@ nfs_lookup(struct vop_lookup_args *ap)
 	struct vnode **vpp = ap->a_vpp;
 	struct mount *mp = dvp->v_mount;
 	struct vattr vattr;
+	time_t dmtime;
 	int flags = cnp->cn_flags;
 	struct vnode *newvp;
 	struct nfsmount *nmp;
@@ -935,7 +936,7 @@ nfs_lookup(struct vop_lookup_args *ap)
 	int error = 0, attrflag, fhsize, ltype;
 	int v3 = NFS_ISV3(dvp);
 	struct thread *td = cnp->cn_thread;
-	
+
 	*vpp = NULLVP;
 	if ((flags & ISLASTCN) && (mp->mnt_flag & MNT_RDONLY) &&
 	    (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME))
@@ -992,6 +993,19 @@ nfs_lookup(struct vop_lookup_args *ap)
 		np->n_dmtime = 0;
 		mtx_unlock(&np->n_mtx);
 	}
+
+	/*
+	 * Cache the modification time of the parent directory in case
+	 * the lookup fails and results in adding the first negative
+	 * name cache entry for the directory.  Since this is reading
+	 * a single time_t, don't bother with locking.  The
+	 * modification time may be a bit stale, but it must be read
+	 * before performing the lookup RPC to prevent a race where
+	 * another lookup updates the timestamp on the directory after
+	 * the lookup RPC has been performed on the server but before
+	 * n_dmtime is set at the end of this function.
+	 */
+	dmtime = np->n_vattr.va_mtime.tv_sec;
 	error = 0;
 	newvp = NULLVP;
 	nfsstats.lookupcache_misses++;
@@ -1130,13 +1144,25 @@ nfsmout:
 			 * Maintain n_dmtime as the modification time
 			 * of the parent directory when the oldest -ve
 			 * name cache entry for this directory was
-			 * added.
+			 * added.  If a -ve cache entry has already
+			 * been added with a newer modification time
+			 * by a concurrent lookup, then don't bother
+			 * adding a cache entry.  The modification
+			 * time of the directory might have changed
+			 * due to the file this lookup failed to find
+			 * being created.  In that case a subsequent
+			 * lookup would incorrectly use the entry
+			 * added here instead of doing an extra
+			 * lookup.
 			 */
 			mtx_lock(&np->n_mtx);
-			if (np->n_dmtime == 0)
-				np->n_dmtime = np->n_vattr.va_mtime.tv_sec;
-			mtx_unlock(&np->n_mtx);
-			cache_enter(dvp, NULL, cnp);
+			if (np->n_dmtime <= dmtime) {
+				if (np->n_dmtime == 0)
+					np->n_dmtime = dmtime;
+				mtx_unlock(&np->n_mtx);
+				cache_enter(dvp, NULL, cnp);
+			} else
+				mtx_unlock(&np->n_mtx);
 		}
 		return (ENOENT);
 	}



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200910161930.n9GJUnHD011723>