md-cache: Register the list of xattrs with cache-invalidation

Issue:
md-cache caches a specified list of xattrs, and when cache invalidation
is enabled, it makes sense to recieve invalidation only when those xattrs
are modified by other clients. But the current implementation of upcall
is that, it will send invalidation when any of the on-disk xattrs is modified.

Solution:
md-cache sends a list of xattrs that it is interested in, to upcall by
issuing an ipc(). The challenge here is to make sure everytime a brick
goes offline and comes back up, the ipc() needs to be issued to the
bricks. Hence ipc() is sent from md-cache every time there is a
CHILD_UP/CHILD_MODIFIED event.

TODO:
There will be patches following, in cluster xlators, to implement ipc fop.

Change-Id: I6efcf3df474f5ce6eabd3d6694c00c7bd89bc25d
BUG: 1211863
Signed-off-by: Poornima G <pgurusid@redhat.com>
Reviewed-on: http://review.gluster.org/15002
Smoke: Gluster Build System <jenkins@build.gluster.org>
CentOS-regression: Gluster Build System <jenkins@build.gluster.org>
Reviewed-by: Rajesh Joseph <rjoseph@redhat.com>
NetBSD-regression: NetBSD Build System <jenkins@build.gluster.org>
Reviewed-by: Prashanth Pai <ppai@redhat.com>
Reviewed-by: Raghavendra G <rgowdapp@redhat.com>

3 files changed, 45 insertions, 0 deletions

diff --git a/libglusterfs/src/dict.c b/libglusterfs/src/dict.c
index 6f7adb51589..96cb9e94bda 100644
--- a/libglusterfs/src/dict.c
+++ b/libglusterfs/src/dict.c
@@ -473,6 +473,26 @@ dict_get (dict_t *this, char *key)
         return NULL;
 }
 
+int
+dict_key_count (dict_t *this)
+{
+        int ret = -1;
+
+        if (!this) {
+                gf_msg_callingfn ("dict", GF_LOG_WARNING, EINVAL,
+                                  LG_MSG_INVALID_ARG, "dict passed is NULL");
+                return ret;
+        }
+
+        LOCK (&this->lock);
+        {
+                ret = this->count;
+        }
+        UNLOCK (&this->lock);
+
+        return ret;
+}
+
 void
 dict_del (dict_t *this, char *key)
 {
diff --git a/libglusterfs/src/dict.h b/libglusterfs/src/dict.h
index 04f0ed9b164..a7fb6c78425 100644
--- a/libglusterfs/src/dict.h
+++ b/libglusterfs/src/dict.h
@@ -107,6 +107,8 @@ data_t *dict_get (dict_t *this, char *key);
 void dict_del (dict_t *this, char *key);
 int dict_reset (dict_t *dict);
 
+int dict_key_count (dict_t *this);
+
 int32_t dict_serialized_length (dict_t *dict);
 int32_t dict_serialize (dict_t *dict, char *buf);
 int32_t dict_unserialize (char *buf, int32_t size, dict_t **fill);
diff --git a/libglusterfs/src/syncop.h b/libglusterfs/src/syncop.h
index c2387e62cd2..0d0da58f4cf 100644
--- a/libglusterfs/src/syncop.h
+++ b/libglusterfs/src/syncop.h
@@ -258,6 +258,29 @@ struct syncopctx {
         } while (0)
 
 
+/*
+ * syncop_xxx() calls are executed in two ways, one is inside a synctask where
+ * the executing function will do 'swapcontext' and the other is without
+ * synctask where the executing thread is made to wait using pthread_cond_wait.
+ * Executing thread may change when syncop_xxx() is executed inside a synctask.
+ * This leads to errno_location change i.e. errno may give errno of
+ * non-executing thread. So errno is not touched inside a synctask execution.
+ * All gfapi calls are executed using the second way of executing syncop_xxx()
+ * where the executing thread waits using pthread_cond_wait so it is ok to set
+ * errno in these cases. The following macro makes syncop_xxx() behave just
+ * like a system call, where -1 is returned and errno is set when a failure
+ * occurs.
+ */
+#define DECODE_SYNCOP_ERR(ret) do {  \
+        if (ret < 0) {          \
+                errno = -ret;   \
+                ret = -1;       \
+        } else {                \
+                errno = 0;      \
+        }                       \
+        } while (0)
+
+
 #define SYNCENV_DEFAULT_STACKSIZE (2 * 1024 * 1024)
 
 struct syncenv * syncenv_new (size_t stacksize, int procmin, int procmax);