Server backend storage hang should not cause the mount point to hang.

Submitted-by: Krishna Srinivas <krishna@gluster.com>
NOTE: fixed compilation issues in posix.c introduced while merging

storage/posix polls for FS/kernel being functional by issuing
statvfs() call. In case statvfs expires the timer, storage/posix will
send CHILD_DOWN to upper translator. Ultimately this will cause
protocol/server to disconnect all clients connected and also cleans up
the data structures. Hence if soft lockup or other kernel bug causes
backend FS to hang, the clients will not be hung.

Signed-off-by: Krishna Srinivas <krishna@gluster.com>
Signed-off-by: Anand V. Avati <avati@blackhole.gluster.com>
Signed-off-by: Anand V. Avati <avati@dev.gluster.com>

BUG: 272 (Server backend storage hang should not cause the mount point to hang)
URL: http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=272

2 files changed, 147 insertions, 5 deletions

diff --git a/xlators/storage/posix/src/posix.c b/xlators/storage/posix/src/posix.c
index 1ff9a06f9fe..c9342ac4d69 100644
--- a/xlators/storage/posix/src/posix.c
+++ b/xlators/storage/posix/src/posix.c
@@ -4703,6 +4703,119 @@ posix_inode (xlator_t *this)
         return 0;
 }
 
+void 
+posix_fsping_timer_expired (void *data)
+{
+        xlator_t *this = NULL;
+        struct posix_private *priv = NULL;
+
+        this = data;
+        priv = this->private;
+
+        pthread_mutex_lock (&priv->mutex);
+        {
+                if (priv->fsping_timer) {
+                        gf_timer_call_cancel (this->ctx,
+                                              priv->fsping_timer);
+                        priv->fsping_timer = NULL;
+                }
+
+                if (priv->fs_state) {
+                        priv->fs_state = 0;
+                        default_notify (this, GF_EVENT_CHILD_DOWN, NULL);
+                }
+        }
+        pthread_mutex_unlock (&priv->mutex);
+}
+
+void
+posix_fsping (void *arg);
+
+void *
+posix_fsping_statvfs (void *arg)
+{
+        int ret = -1;
+        xlator_t *this = NULL;
+        char *root_path = NULL;
+        struct statvfs buf = {0, };
+        struct posix_private *priv = NULL;
+        struct timeval delta = {0, };
+
+        this = arg;
+        priv = this->private;
+        root_path = POSIX_BASE_PATH (this);
+
+        ret = statvfs (root_path, &buf);
+
+        pthread_mutex_lock (&priv->mutex);
+        {
+                if (priv->fsping_timer) {
+                        gf_timer_call_cancel (this->ctx,
+                                              priv->fsping_timer);
+                        priv->fsping_timer = NULL;
+                }
+                if (ret == 0) {
+                        if (priv->fs_state == 0) {
+                                priv->fs_state = 1;
+                                default_notify (this, GF_EVENT_CHILD_UP,
+                                                NULL);
+                        }
+                } else {
+                        if (priv->fs_state) {
+                                priv->fs_state = 0;
+                                default_notify (this, GF_EVENT_CHILD_DOWN,
+                                                NULL);
+                        }
+                }
+        }
+        pthread_mutex_unlock (&priv->mutex);
+
+        delta.tv_sec = POSIX_FSPING_SLEEP_TIME;
+        priv->fsping_timer =
+                gf_timer_call_after (this->ctx,
+                                     delta,
+                                     posix_fsping,
+                                     (void *) this);
+        if (priv->fsping_timer == NULL) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "unable to register timer");
+        }
+        return NULL;
+}
+
+void
+posix_fsping (void *arg)
+{
+        xlator_t *this = NULL;
+        struct posix_private *priv = NULL;
+        struct timeval delta = {0, };
+
+        this = arg;
+        priv = this->private;
+
+        delta.tv_sec = priv->fsping_timeout;
+        delta.tv_usec = 0;
+
+        if (priv->fsping_timer) {
+                gf_timer_call_cancel (this->ctx,
+                                      priv->fsping_timer);
+        }
+        priv->fsping_timer =
+                gf_timer_call_after (this->ctx,
+                                     delta,
+                                     posix_fsping_timer_expired,
+                                     (void *) this);
+
+        if (priv->fsping_timer == NULL) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "unable to register timer");
+                /*FIXME: handle error*/
+        }
+        pthread_create (&priv->fsping,
+                        NULL,
+                        posix_fsping_statvfs,
+                        this);
+}
 
 int32_t
 posix_rchecksum (call_frame_t *frame, xlator_t *this,
@@ -4780,12 +4893,15 @@ notify (xlator_t *this,
         void *data,
         ...)
 {
+	struct posix_private *priv = NULL;
+
+	priv = this->private;
+
         switch (event)
                 {
                 case GF_EVENT_PARENT_UP:
                         {
-                                /* Tell the parent that posix xlator is up */
-                                default_notify (this, GF_EVENT_CHILD_UP, data);
+                                posix_fsping ((void *)this);
                         }
                         break;
                 default:
@@ -4809,9 +4925,9 @@ init (xlator_t *this)
         data_t *               dir_data = NULL;
 	data_t *               tmp_data = NULL;
         uint64_t               time64   = 0;
-
-        int dict_ret = 0;
-        int32_t janitor_sleep;
+        int                    dict_ret = -1;
+        int                    fsping_timeout = -1;
+        int32_t                janitor_sleep;
 
         dir_data = dict_get (this->options, "directory");
 
@@ -4908,6 +5024,7 @@ init (xlator_t *this)
         strcat (_private->trash_path, "/" GF_REPLICATE_TRASH_DIR);
 
         LOCK_INIT (&_private->lock);
+        pthread_mutex_init (&_private->mutex, NULL);
 
         ret = gethostname (_private->hostname, 256);
         if (ret < 0) {
@@ -4923,6 +5040,17 @@ init (xlator_t *this)
                 _private->max_write = 1;
         }
 
+        _private->fsping_timeout = POSIX_FSPING_TIMEOUT;
+        dict_ret = dict_get_int32 (this->options,
+                                   "fsping-timeout",
+                                   &fsping_timeout);
+
+        if (dict_ret == 0) {
+                _private->fsping_timeout = fsping_timeout;
+        }
+        gf_log (this->name, GF_LOG_DEBUG,
+                "fsping-timeout set to %d", _private->fsping_timeout);
+
         _private->export_statfs = 1;
         tmp_data = dict_get (this->options, "export-statfs-size");
         if (tmp_data) {
@@ -5056,6 +5184,7 @@ fini (xlator_t *this)
 {
         struct posix_private *priv = this->private;
         sys_lremovexattr (priv->base_path, "trusted.glusterfs.test");
+        pthread_mutex_destroy (&priv->mutex);
         FREE (priv);
         return;
 }
@@ -5132,6 +5261,8 @@ struct volume_options options[] = {
 	  .type = GF_OPTION_TYPE_BOOL },
 	{ .key  = {"span-devices"},
 	  .type = GF_OPTION_TYPE_INT },
+	{ .key  = {"fsping-timeout"},
+	  .type = GF_OPTION_TYPE_INT },
         { .key  = {"background-unlink"},
           .type = GF_OPTION_TYPE_BOOL },
         { .key  = {"janitor-sleep-duration"},
diff --git a/xlators/storage/posix/src/posix.h b/xlators/storage/posix/src/posix.h
index f92e256fbc0..92fe8e2515c 100644
--- a/xlators/storage/posix/src/posix.h
+++ b/xlators/storage/posix/src/posix.h
@@ -29,6 +29,7 @@
 #include <unistd.h>
 #include <sys/types.h>
 #include <dirent.h>
+#include <pthread.h>
 #include <time.h>
 
 #ifdef linux
@@ -50,6 +51,7 @@
 #include "xlator.h"
 #include "inode.h"
 #include "compat.h"
+#include "timer.h"
 
 /**
  * posix_fd - internal structure common to file and directory fd's
@@ -70,6 +72,7 @@ struct posix_private {
 	int32_t base_path_length;
 
         gf_lock_t lock;
+        pthread_mutex_t mutex;
 
         char   hostname[256];
         /* Statistics, provides activity of the server */
@@ -117,6 +120,11 @@ struct posix_private {
         int             num_devices_to_span;
         dev_t          *st_device;
 
+        pthread_t       fsping;
+	gf_timer_t     *fsping_timer;
+	int             fsping_timeout;
+	int             fs_state;
+
 /* a global generation number sequence is used to assign generation numbers in 
    sequence.
 */
@@ -129,6 +137,9 @@ struct posix_private {
         char *          trash_path;
 };
 
+#define POSIX_FSPING_SLEEP_TIME 10
+#define POSIX_FSPING_TIMEOUT 10
+
 #define POSIX_BASE_PATH(this) (((struct posix_private *)this->private)->base_path)
 
 #define POSIX_BASE_PATH_LEN(this) (((struct posix_private *)this->private)->base_path_length)