exportfs: Stop racing exportfs on clusters

This problem can occur when multiple cluster services fail over
at the same time, causing missing high-available exports.
Having a lot of nfs-exports will trigger this issue easier.

https://bugzilla.linux-nfs.org/show_bug.cgi?id=224

Signed-off-by: Steve Dickson <steved@redhat.com>

diff --git a/support/include/exportfs.h b/support/include/exportfs.h
index 01e87ddc289252e818f50ab375d9fc25969a8c9f..99916e515af13691cbada636b10664673b7600dd 100644 (file)
--- a/support/include/exportfs.h
+++ b/support/include/exportfs.h
@@ -32,6 +32,10 @@ enum {
        FSLOC_STUB
 };
 
+#ifndef EXP_LOCKFILE
+#define EXP_LOCKFILE "/var/lib/nfs/export-lock"
+#endif
+
 typedef struct mclient {
        struct mclient *        m_next;
        char *                  m_hostname;

diff --git a/utils/exportfs/exportfs.c b/utils/exportfs/exportfs.c
index 7432a6592c8d0e69c60dfc69ececec56e92b1f34..7326802552e35005cdbbbdb316417895c8677983 100644 (file)
--- a/utils/exportfs/exportfs.c
+++ b/utils/exportfs/exportfs.c
@@ -16,6 +16,7 @@
 #include <sys/stat.h>
 #include <sys/vfs.h>
 #include <sys/stat.h>
+#include <sys/file.h>
 #include <unistd.h>
 #include <stdbool.h>
 #include <stdlib.h>
@@ -44,6 +45,39 @@ static void  validate_export(nfs_export *exp);
 static int     matchhostname(const char *hostname1, const char *hostname2);
 static void    export_d_read(const char *dname);
 
+static const char *lockfile = EXP_LOCKFILE;
+static int _lockfd = -1;
+
+/*
+ * If we aren't careful, changes made by exportfs can be lost
+ * when multiple exports process run at once:
+ *
+ *     exportfs process 1      exportfs process 2
+ *     ------------------------------------------
+ *     reads etab version A    reads etab version A
+ *     adds new export B       adds new export C
+ *     writes A+B              writes A+C
+ *
+ * The locking in support/export/xtab.c will prevent mountd from
+ * seeing a partially written version of etab, and will prevent 
+ * the two writers above from writing simultaneously and
+ * corrupting etab, but to prevent problems like the above we
+ * need these additional lockfile() routines.
+ */
+void 
+grab_lockfile()
+{
+       _lockfd = open(lockfile, O_CREAT|O_RDWR, 0666);
+       if (_lockfd != -1) 
+               lockf(_lockfd, F_LOCK, 0);
+}
+void 
+release_lockfile()
+{
+       if (_lockfd != -1)
+               lockf(_lockfd, F_ULOCK, 0);
+}
+
 int
 main(int argc, char **argv)
 {
@@ -129,6 +163,13 @@ main(int argc, char **argv)
                        return 0;
                }
        }
+
+       /*
+        * Serialize things as best we can
+        */
+       grab_lockfile();
+       atexit(release_lockfile);
+
        if (f_export && ! f_ignore) {
                export_read(_PATH_EXPORTS);
                export_d_read(_PATH_EXPORTS_D);