When a "mount -u" switches an NFS mount point from TCP to UDP,

any thread doing an I/O RPC with a transfer size greater than
NFS_UDPMAXDATA will be hung indefinitely, retrying the RPC.
After a discussion on freebsd-fs@, I decided to add a warning
message for this case, as suggested by Jeremy Chadwick.

Suggested by:	freebsd at jdc.parodius.com (Jeremy Chadwick)
MFC after:	2 weeks

1 files changed, 14 insertions, 0 deletions

diff --git a/sys/nfsclient/nfs_vfsops.c b/sys/nfsclient/nfs_vfsops.c
index 82891f9e29f8..58384e86a1bf 100644
--- a/sys/nfsclient/nfs_vfsops.c
+++ b/sys/nfsclient/nfs_vfsops.c
@@ -56,6 +56,7 @@ __FBSDID("$FreeBSD$");
 #include <sys/socketvar.h>
 #include <sys/sockio.h>
 #include <sys/sysctl.h>
+#include <sys/syslog.h>
 #include <sys/vnode.h>
 #include <sys/signalvar.h>
 
@@ -1116,6 +1117,19 @@ nfs_mount(struct mount *mp)
 			error = EIO;
 			goto out;
 		}
+
+		/*
+		 * If a change from TCP->UDP is done and there are thread(s)
+		 * that have I/O RPC(s) in progress with a tranfer size
+		 * greater than NFS_MAXDGRAMDATA, those thread(s) will be
+		 * hung, retrying the RPC(s) forever. Usually these threads
+		 * will be seen doing an uninterruptible sleep on wait channel
+		 * "newnfsreq" (truncated to "newnfsre" by procstat).
+		 */
+		if (args.sotype == SOCK_DGRAM && nmp->nm_sotype == SOCK_STREAM)
+			tprintf(curthread->td_proc, LOG_WARNING,
+	"Warning: mount -u that changes TCP->UDP can result in hung threads\n");
+
 		/*
 		 * When doing an update, we can't change from or to
 		 * v3, switch lockd strategies or change cookie translation