path: root/0227-glusterd-svc-update-pid-of-mux-volumes-from-the-shd-.patch

From b0815b8a84a07d17a1215c55afc38888ee9fc37c Mon Sep 17 00:00:00 2001
From: Mohammed Rafi KC <rkavunga@redhat.com>
Date: Mon, 24 Jun 2019 12:00:20 +0530
Subject: [PATCH 227/255] glusterd/svc: update pid of mux volumes from the shd
 process

For a normal volume, we are updating the pid from a the
process while we do a daemonization or at the end of the
init if it is no-daemon mode. Along with updating the pid
we also lock the file, to make sure that the process is
running fine.

With brick mux, we were updating the pidfile from gluterd
after an attach/detach request.

There are two problems with this approach.
1) We are not holding a pidlock for any file other than parent
   process.
2) There is a chance for possible race conditions with attach/detach.
   For example, shd start and a volume stop could race. Let's say
   we are starting an shd and it is attached to a volume.
   While we trying to link the pid file to the running process,
   this would have deleted by the thread that doing a volume stop.

> upstream patch : https://review.gluster.org/#/c/glusterfs/+/22935/

>Change-Id: I29a00352102877ce09ea3f376ca52affceb5cf1a
>Updates: bz#1722541
>Signed-off-by: Mohammed Rafi KC <rkavunga@redhat.com>

Change-Id: I29a00352102877ce09ea3f376ca52affceb5cf1a
BUG: 1721802
Signed-off-by: Mohammed Rafi KC <rkavunga@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/175723
Tested-by: RHGS Build Bot <nigelb@redhat.com>
Reviewed-by: Atin Mukherjee <amukherj@redhat.com>
---
 glusterfsd/src/gf_attach.c                         |   2 +
 glusterfsd/src/glusterfsd-mgmt.c                   |  66 +++++++--
 libglusterfs/src/glusterfs/glusterfs.h             |   2 +-
 libglusterfs/src/glusterfs/libglusterfs-messages.h |   3 +-
 libglusterfs/src/graph.c                           | 154 ++++++++++++++++++++-
 rpc/xdr/src/glusterd1-xdr.x                        |   1 +
 xlators/mgmt/glusterd/src/glusterd-handler.c       |   2 +
 xlators/mgmt/glusterd/src/glusterd-handshake.c     |  42 +++++-
 xlators/mgmt/glusterd/src/glusterd-op-sm.c         |   4 +
 .../mgmt/glusterd/src/glusterd-shd-svc-helper.c    |  25 ++++
 .../mgmt/glusterd/src/glusterd-shd-svc-helper.h    |   3 +
 xlators/mgmt/glusterd/src/glusterd-shd-svc.c       |   8 +-
 xlators/mgmt/glusterd/src/glusterd-svc-helper.c    |  57 ++++----
 xlators/mgmt/glusterd/src/glusterd-syncop.c        |   2 +
 xlators/mgmt/glusterd/src/glusterd-utils.c         |   6 +-
 15 files changed, 325 insertions(+), 52 deletions(-)

diff --git a/glusterfsd/src/gf_attach.c b/glusterfsd/src/gf_attach.c
index 6293b9b..1bff854 100644
--- a/glusterfsd/src/gf_attach.c
+++ b/glusterfsd/src/gf_attach.c
@@ -65,6 +65,8 @@ send_brick_req(xlator_t *this, struct rpc_clnt *rpc, char *path, int op)
     brick_req.name = path;
     brick_req.input.input_val = NULL;
     brick_req.input.input_len = 0;
+    brick_req.dict.dict_val = NULL;
+    brick_req.dict.dict_len = 0;
 
     req_size = xdr_sizeof((xdrproc_t)xdr_gd1_mgmt_brick_op_req, req);
     iobuf = iobuf_get2(rpc->ctx->iobuf_pool, req_size);
diff --git a/glusterfsd/src/glusterfsd-mgmt.c b/glusterfsd/src/glusterfsd-mgmt.c
index 1d2cd1a..f930e0a 100644
--- a/glusterfsd/src/glusterfsd-mgmt.c
+++ b/glusterfsd/src/glusterfsd-mgmt.c
@@ -50,13 +50,16 @@ int
 emancipate(glusterfs_ctx_t *ctx, int ret);
 int
 glusterfs_process_svc_attach_volfp(glusterfs_ctx_t *ctx, FILE *fp,
-                                   char *volfile_id, char *checksum);
+                                   char *volfile_id, char *checksum,
+                                   dict_t *dict);
 int
 glusterfs_mux_volfile_reconfigure(FILE *newvolfile_fp, glusterfs_ctx_t *ctx,
-                                  gf_volfile_t *volfile_obj, char *checksum);
+                                  gf_volfile_t *volfile_obj, char *checksum,
+                                  dict_t *dict);
 int
 glusterfs_process_svc_attach_volfp(glusterfs_ctx_t *ctx, FILE *fp,
-                                   char *volfile_id, char *checksum);
+                                   char *volfile_id, char *checksum,
+                                   dict_t *dict);
 int
 glusterfs_process_svc_detach(glusterfs_ctx_t *ctx, gf_volfile_t *volfile_obj);
 
@@ -75,7 +78,8 @@ mgmt_cbk_spec(struct rpc_clnt *rpc, void *mydata, void *data)
 }
 
 int
-mgmt_process_volfile(const char *volfile, ssize_t size, char *volfile_id)
+mgmt_process_volfile(const char *volfile, ssize_t size, char *volfile_id,
+                     dict_t *dict)
 {
     glusterfs_ctx_t *ctx = NULL;
     int ret = 0;
@@ -145,11 +149,11 @@ mgmt_process_volfile(const char *volfile, ssize_t size, char *volfile_id)
              * the volfile
              */
             ret = glusterfs_process_svc_attach_volfp(ctx, tmpfp, volfile_id,
-                                                     sha256_hash);
+                                                     sha256_hash, dict);
             goto unlock;
         }
         ret = glusterfs_mux_volfile_reconfigure(tmpfp, ctx, volfile_obj,
-                                                sha256_hash);
+                                                sha256_hash, dict);
         if (ret < 0) {
             gf_msg_debug("glusterfsd-mgmt", EINVAL, "Reconfigure failed !!");
         }
@@ -387,6 +391,8 @@ err:
         UNLOCK(&ctx->volfile_lock);
     if (xlator_req.input.input_val)
         free(xlator_req.input.input_val);
+    if (xlator_req.dict.dict_val)
+        free(xlator_req.dict.dict_val);
     free(xlator_req.name);
     xlator_req.name = NULL;
     return 0;
@@ -561,6 +567,8 @@ out:
 
     free(xlator_req.name);
     free(xlator_req.input.input_val);
+    if (xlator_req.dict.dict_val)
+        free(xlator_req.dict.dict_val);
     if (output)
         dict_unref(output);
     if (dict)
@@ -982,6 +990,8 @@ out:
     if (input)
         dict_unref(input);
     free(xlator_req.input.input_val); /*malloced by xdr*/
+    if (xlator_req.dict.dict_val)
+        free(xlator_req.dict.dict_val);
     if (output)
         dict_unref(output);
     free(xlator_req.name);
@@ -1062,6 +1072,8 @@ glusterfs_handle_attach(rpcsvc_request_t *req)
     out:
         UNLOCK(&ctx->volfile_lock);
     }
+    if (xlator_req.dict.dict_val)
+        free(xlator_req.dict.dict_val);
     free(xlator_req.input.input_val);
     free(xlator_req.name);
 
@@ -1077,6 +1089,7 @@ glusterfs_handle_svc_attach(rpcsvc_request_t *req)
     };
     xlator_t *this = NULL;
     glusterfs_ctx_t *ctx = NULL;
+    dict_t *dict = NULL;
 
     GF_ASSERT(req);
     this = THIS;
@@ -1091,20 +1104,41 @@ glusterfs_handle_svc_attach(rpcsvc_request_t *req)
         req->rpc_err = GARBAGE_ARGS;
         goto out;
     }
+
     gf_msg(THIS->name, GF_LOG_INFO, 0, glusterfsd_msg_41,
            "received attach "
            "request for volfile-id=%s",
            xlator_req.name);
+
+    dict = dict_new();
+    if (!dict) {
+        ret = -1;
+        errno = ENOMEM;
+        goto out;
+    }
+
+    ret = dict_unserialize(xlator_req.dict.dict_val, xlator_req.dict.dict_len,
+                           &dict);
+    if (ret) {
+        gf_msg(this->name, GF_LOG_WARNING, EINVAL, glusterfsd_msg_42,
+               "failed to unserialize xdata to dictionary");
+        goto out;
+    }
+    dict->extra_stdfree = xlator_req.dict.dict_val;
+
     ret = 0;
 
     if (ctx->active) {
         ret = mgmt_process_volfile(xlator_req.input.input_val,
-                                   xlator_req.input.input_len, xlator_req.name);
+                                   xlator_req.input.input_len, xlator_req.name,
+                                   dict);
     } else {
         gf_msg(this->name, GF_LOG_WARNING, EINVAL, glusterfsd_msg_42,
                "got attach for %s but no active graph", xlator_req.name);
     }
 out:
+    if (dict)
+        dict_unref(dict);
     if (xlator_req.input.input_val)
         free(xlator_req.input.input_val);
     if (xlator_req.name)
@@ -1241,6 +1275,8 @@ out:
     GF_FREE(filepath);
     if (xlator_req.input.input_val)
         free(xlator_req.input.input_val);
+    if (xlator_req.dict.dict_val)
+        free(xlator_req.dict.dict_val);
 
     return ret;
 }
@@ -1313,6 +1349,8 @@ out:
     if (dict)
         dict_unref(dict);
     free(xlator_req.input.input_val);  // malloced by xdr
+    if (xlator_req.dict.dict_val)
+        free(xlator_req.dict.dict_val);
     if (output)
         dict_unref(output);
     free(xlator_req.name);  // malloced by xdr
@@ -1461,6 +1499,8 @@ out:
     if (output)
         dict_unref(output);
     free(brick_req.input.input_val);
+    if (brick_req.dict.dict_val)
+        free(brick_req.dict.dict_val);
     free(brick_req.name);
     GF_FREE(xname);
     GF_FREE(msg);
@@ -1654,6 +1694,8 @@ out:
     if (dict)
         dict_unref(dict);
     free(node_req.input.input_val);
+    if (node_req.dict.dict_val)
+        free(node_req.dict.dict_val);
     GF_FREE(msg);
     GF_FREE(rsp.output.output_val);
     GF_FREE(node_name);
@@ -1757,6 +1799,8 @@ glusterfs_handle_nfs_profile(rpcsvc_request_t *req)
 
 out:
     free(nfs_req.input.input_val);
+    if (nfs_req.dict.dict_val)
+        free(nfs_req.dict.dict_val);
     if (dict)
         dict_unref(dict);
     if (output)
@@ -1835,6 +1879,8 @@ out:
     if (dict)
         dict_unref(dict);
     free(xlator_req.input.input_val);  // malloced by xdr
+    if (xlator_req.dict.dict_val)
+        free(xlator_req.dict.dict_val);
     if (output)
         dict_unref(output);
     free(xlator_req.name);  // malloced by xdr
@@ -1963,7 +2009,8 @@ out:
     if (dict)
         dict_unref(dict);
     free(brick_req.input.input_val);
-
+    if (brick_req.dict.dict_val)
+        free(brick_req.dict.dict_val);
     gf_log(THIS->name, GF_LOG_DEBUG, "Returning %d", ret);
     return ret;
 }
@@ -2213,7 +2260,8 @@ volfile:
     size = rsp.op_ret;
     volfile_id = frame->local;
     if (mgmt_is_multiplexed_daemon(ctx->cmd_args.process_name)) {
-        ret = mgmt_process_volfile((const char *)rsp.spec, size, volfile_id);
+        ret = mgmt_process_volfile((const char *)rsp.spec, size, volfile_id,
+                                   dict);
         goto post_graph_mgmt;
     }
 
diff --git a/libglusterfs/src/glusterfs/glusterfs.h b/libglusterfs/src/glusterfs/glusterfs.h
index 9ec2365..b6327b8 100644
--- a/libglusterfs/src/glusterfs/glusterfs.h
+++ b/libglusterfs/src/glusterfs/glusterfs.h
@@ -744,7 +744,7 @@ typedef struct {
     char vol_id[NAME_MAX + 1];
     struct list_head volfile_list;
     glusterfs_graph_t *graph;
-
+    FILE *pidfp;
 } gf_volfile_t;
 
 glusterfs_ctx_t *
diff --git a/libglusterfs/src/glusterfs/libglusterfs-messages.h b/libglusterfs/src/glusterfs/libglusterfs-messages.h
index ea2aa60..7e0eebb 100644
--- a/libglusterfs/src/glusterfs/libglusterfs-messages.h
+++ b/libglusterfs/src/glusterfs/libglusterfs-messages.h
@@ -111,6 +111,7 @@ GLFS_MSGID(
     LG_MSG_PTHREAD_NAMING_FAILED, LG_MSG_SYSCALL_RETURNS_WRONG,
     LG_MSG_XXH64_TO_GFID_FAILED, LG_MSG_ASYNC_WARNING, LG_MSG_ASYNC_FAILURE,
     LG_MSG_GRAPH_CLEANUP_FAILED, LG_MSG_GRAPH_SETUP_FAILED,
-    LG_MSG_GRAPH_DETACH_STARTED, LG_MSG_GRAPH_ATTACH_FAILED);
+    LG_MSG_GRAPH_DETACH_STARTED, LG_MSG_GRAPH_ATTACH_FAILED,
+    LG_MSG_GRAPH_ATTACH_PID_FILE_UPDATED);
 
 #endif /* !_LG_MESSAGES_H_ */
diff --git a/libglusterfs/src/graph.c b/libglusterfs/src/graph.c
index 172dc61..05f76bf 100644
--- a/libglusterfs/src/graph.c
+++ b/libglusterfs/src/graph.c
@@ -1467,6 +1467,21 @@ out:
 }
 
 int
+glusterfs_svc_mux_pidfile_cleanup(gf_volfile_t *volfile_obj)
+{
+    if (!volfile_obj || !volfile_obj->pidfp)
+        return 0;
+
+    gf_msg_trace("glusterfsd", 0, "pidfile %s cleanup", volfile_obj->vol_id);
+
+    lockf(fileno(volfile_obj->pidfp), F_ULOCK, 0);
+    fclose(volfile_obj->pidfp);
+    volfile_obj->pidfp = NULL;
+
+    return 0;
+}
+
+int
 glusterfs_process_svc_detach(glusterfs_ctx_t *ctx, gf_volfile_t *volfile_obj)
 {
     xlator_t *last_xl = NULL;
@@ -1502,6 +1517,7 @@ glusterfs_process_svc_detach(glusterfs_ctx_t *ctx, gf_volfile_t *volfile_obj)
 
         list_del_init(&volfile_obj->volfile_list);
         glusterfs_mux_xlator_unlink(parent_graph->top, xl);
+        glusterfs_svc_mux_pidfile_cleanup(volfile_obj);
         parent_graph->last_xl = glusterfs_get_last_xlator(parent_graph);
         parent_graph->xl_count -= graph->xl_count;
         parent_graph->leaf_count -= graph->leaf_count;
@@ -1531,8 +1547,126 @@ out:
 }
 
 int
+glusterfs_svc_mux_pidfile_setup(gf_volfile_t *volfile_obj, const char *pid_file)
+{
+    int ret = -1;
+    FILE *pidfp = NULL;
+
+    if (!pid_file || !volfile_obj)
+        goto out;
+
+    if (volfile_obj->pidfp) {
+        ret = 0;
+        goto out;
+    }
+    pidfp = fopen(pid_file, "a+");
+    if (!pidfp) {
+        goto out;
+    }
+    volfile_obj->pidfp = pidfp;
+
+    ret = lockf(fileno(pidfp), F_TLOCK, 0);
+    if (ret) {
+        ret = 0;
+        goto out;
+    }
+out:
+    return ret;
+}
+
+int
+glusterfs_svc_mux_pidfile_update(gf_volfile_t *volfile_obj,
+                                 const char *pid_file, pid_t pid)
+{
+    int ret = 0;
+    FILE *pidfp = NULL;
+    int old_pid;
+
+    if (!volfile_obj->pidfp) {
+        ret = glusterfs_svc_mux_pidfile_setup(volfile_obj, pid_file);
+        if (ret == -1)
+            goto out;
+    }
+    pidfp = volfile_obj->pidfp;
+    ret = fscanf(pidfp, "%d", &old_pid);
+    if (ret <= 0) {
+        goto update;
+    }
+    if (old_pid == pid) {
+        ret = 0;
+        goto out;
+    } else {
+        gf_msg("mgmt", GF_LOG_INFO, 0, LG_MSG_GRAPH_ATTACH_PID_FILE_UPDATED,
+               "Old pid=%d found in pidfile %s. Cleaning the old pid and "
+               "Updating new pid=%d",
+               old_pid, pid_file, pid);
+    }
+update:
+    ret = sys_ftruncate(fileno(pidfp), 0);
+    if (ret) {
+        gf_msg("glusterfsd", GF_LOG_ERROR, errno,
+               LG_MSG_GRAPH_ATTACH_PID_FILE_UPDATED,
+               "pidfile %s truncation failed", pid_file);
+        goto out;
+    }
+
+    ret = fprintf(pidfp, "%d\n", pid);
+    if (ret <= 0) {
+        gf_msg("glusterfsd", GF_LOG_ERROR, errno,
+               LG_MSG_GRAPH_ATTACH_PID_FILE_UPDATED, "pidfile %s write failed",
+               pid_file);
+        goto out;
+    }
+
+    ret = fflush(pidfp);
+    if (ret) {
+        gf_msg("glusterfsd", GF_LOG_ERROR, errno,
+               LG_MSG_GRAPH_ATTACH_PID_FILE_UPDATED, "pidfile %s write failed",
+               pid_file);
+        goto out;
+    }
+out:
+    return ret;
+}
+
+int
+glusterfs_update_mux_pid(dict_t *dict, gf_volfile_t *volfile_obj)
+{
+    char *file = NULL;
+    int ret = -1;
+
+    GF_VALIDATE_OR_GOTO("graph", dict, out);
+    GF_VALIDATE_OR_GOTO("graph", volfile_obj, out);
+
+    ret = dict_get_str(dict, "pidfile", &file);
+    if (ret < 0) {
+        gf_msg("mgmt", GF_LOG_ERROR, EINVAL, LG_MSG_GRAPH_SETUP_FAILED,
+               "Failed to get pidfile from dict for  volfile_id=%s",
+               volfile_obj->vol_id);
+    }
+
+    ret = glusterfs_svc_mux_pidfile_update(volfile_obj, file, getpid());
+    if (ret < 0) {
+        ret = -1;
+        gf_msg("mgmt", GF_LOG_ERROR, EINVAL, LG_MSG_GRAPH_SETUP_FAILED,
+               "Failed to update "
+               "the pidfile for volfile_id=%s",
+               volfile_obj->vol_id);
+
+        goto out;
+    }
+
+    if (ret == 1)
+        gf_msg("mgmt", GF_LOG_INFO, 0, LG_MSG_GRAPH_ATTACH_PID_FILE_UPDATED,
+               "PID %d updated in pidfile=%s", getpid(), file);
+    ret = 0;
+out:
+    return ret;
+}
+int
 glusterfs_process_svc_attach_volfp(glusterfs_ctx_t *ctx, FILE *fp,
-                                   char *volfile_id, char *checksum)
+                                   char *volfile_id, char *checksum,
+                                   dict_t *dict)
 {
     glusterfs_graph_t *graph = NULL;
     glusterfs_graph_t *parent_graph = NULL;
@@ -1615,18 +1749,25 @@ glusterfs_process_svc_attach_volfp(glusterfs_ctx_t *ctx, FILE *fp,
         ret = -1;
         goto out;
     }
+    volfile_obj->pidfp = NULL;
+    snprintf(volfile_obj->vol_id, sizeof(volfile_obj->vol_id), "%s",
+             volfile_id);
+
+    if (strcmp(ctx->cmd_args.process_name, "glustershd") == 0) {
+        ret = glusterfs_update_mux_pid(dict, volfile_obj);
+        if (ret == -1) {
+            goto out;
+        }
+    }
 
     graph->used = 1;
     parent_graph->id++;
     list_add(&graph->list, &ctx->graphs);
     INIT_LIST_HEAD(&volfile_obj->volfile_list);
     volfile_obj->graph = graph;
-    snprintf(volfile_obj->vol_id, sizeof(volfile_obj->vol_id), "%s",
-             volfile_id);
     memcpy(volfile_obj->volfile_checksum, checksum,
            sizeof(volfile_obj->volfile_checksum));
     list_add_tail(&volfile_obj->volfile_list, &ctx->volfile_list);
-
     gf_log_dump_graph(fp, graph);
     graph = NULL;
 
@@ -1654,7 +1795,8 @@ out:
 
 int
 glusterfs_mux_volfile_reconfigure(FILE *newvolfile_fp, glusterfs_ctx_t *ctx,
-                                  gf_volfile_t *volfile_obj, char *checksum)
+                                  gf_volfile_t *volfile_obj, char *checksum,
+                                  dict_t *dict)
 {
     glusterfs_graph_t *oldvolfile_graph = NULL;
     glusterfs_graph_t *newvolfile_graph = NULL;
@@ -1703,7 +1845,7 @@ glusterfs_mux_volfile_reconfigure(FILE *newvolfile_fp, glusterfs_ctx_t *ctx,
         }
         volfile_obj = NULL;
         ret = glusterfs_process_svc_attach_volfp(ctx, newvolfile_fp, vol_id,
-                                                 checksum);
+                                                 checksum, dict);
         goto out;
     }
 
diff --git a/rpc/xdr/src/glusterd1-xdr.x b/rpc/xdr/src/glusterd1-xdr.x
index 9b36d34..02ebec2 100644
--- a/rpc/xdr/src/glusterd1-xdr.x
+++ b/rpc/xdr/src/glusterd1-xdr.x
@@ -132,6 +132,7 @@ struct gd1_mgmt_brick_op_req {
         string  name<>;
         int     op;
         opaque  input<>;
+        opaque  dict<>;
 } ;
 
 struct gd1_mgmt_brick_op_rsp {
diff --git a/xlators/mgmt/glusterd/src/glusterd-handler.c b/xlators/mgmt/glusterd/src/glusterd-handler.c
index af8a8a4..cc1f1df 100644
--- a/xlators/mgmt/glusterd/src/glusterd-handler.c
+++ b/xlators/mgmt/glusterd/src/glusterd-handler.c
@@ -5423,6 +5423,8 @@ glusterd_print_client_details(FILE *fp, dict_t *dict,
 
     brick_req->op = GLUSTERD_BRICK_STATUS;
     brick_req->name = "";
+    brick_req->dict.dict_val = NULL;
+    brick_req->dict.dict_len = 0;
 
     ret = dict_set_strn(dict, "brick-name", SLEN("brick-name"),
                         brickinfo->path);
diff --git a/xlators/mgmt/glusterd/src/glusterd-handshake.c b/xlators/mgmt/glusterd/src/glusterd-handshake.c
index 1ba58c3..86dec82 100644
--- a/xlators/mgmt/glusterd/src/glusterd-handshake.c
+++ b/xlators/mgmt/glusterd/src/glusterd-handshake.c
@@ -203,7 +203,7 @@ out:
 
 size_t
 build_volfile_path(char *volume_id, char *path, size_t path_len,
-                   char *trusted_str)
+                   char *trusted_str, dict_t *dict)
 {
     struct stat stbuf = {
         0,
@@ -340,11 +340,19 @@ build_volfile_path(char *volume_id, char *path, size_t path_len,
 
         ret = glusterd_volinfo_find(volid_ptr, &volinfo);
         if (ret == -1) {
-            gf_log(this->name, GF_LOG_ERROR, "Couldn't find volinfo");
+            gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_VOLINFO_GET_FAIL,
+                   "Couldn't find volinfo for volid=%s", volid_ptr);
             goto out;
         }
 
         glusterd_svc_build_shd_volfile_path(volinfo, path, path_len);
+
+        ret = glusterd_svc_set_shd_pidfile(volinfo, dict);
+        if (ret == -1) {
+            gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_DICT_SET_FAILED,
+                   "Couldn't set pidfile in dict for volid=%s", volid_ptr);
+            goto out;
+        }
         ret = 0;
         goto out;
     }
@@ -919,6 +927,7 @@ __server_getspec(rpcsvc_request_t *req)
     char addrstr[RPCSVC_PEER_STRLEN] = {0};
     peer_info_t *peerinfo = NULL;
     xlator_t *this = NULL;
+    dict_t *dict = NULL;
 
     this = THIS;
     GF_ASSERT(this);
@@ -971,6 +980,12 @@ __server_getspec(rpcsvc_request_t *req)
         goto fail;
     }
 
+    dict = dict_new();
+    if (!dict) {
+        ret = -ENOMEM;
+        goto fail;
+    }
+
     trans = req->trans;
     /* addrstr will be empty for cli socket connections */
     ret = rpcsvc_transport_peername(trans, (char *)&addrstr, sizeof(addrstr));
@@ -989,12 +1004,26 @@ __server_getspec(rpcsvc_request_t *req)
      */
     if (strlen(addrstr) == 0 || gf_is_local_addr(addrstr)) {
         ret = build_volfile_path(volume, filename, sizeof(filename),
-                                 TRUSTED_PREFIX);
+                                 TRUSTED_PREFIX, dict);
     } else {
-        ret = build_volfile_path(volume, filename, sizeof(filename), NULL);
+        ret = build_volfile_path(volume, filename, sizeof(filename), NULL,
+                                 dict);
     }
 
     if (ret == 0) {
+        if (dict->count > 0) {
+            ret = dict_allocate_and_serialize(dict, &rsp.xdata.xdata_val,
+                                              &rsp.xdata.xdata_len);
+            if (ret) {
+                gf_msg(this->name, GF_LOG_ERROR, 0,
+                       GD_MSG_DICT_SERL_LENGTH_GET_FAIL,
+                       "Failed to serialize dict "
+                       "to request buffer");
+                goto fail;
+            }
+            dict->extra_free = rsp.xdata.xdata_val;
+        }
+
         /* to allocate the proper buffer to hold the file data */
         ret = sys_stat(filename, &stbuf);
         if (ret < 0) {
@@ -1036,7 +1065,6 @@ __server_getspec(rpcsvc_request_t *req)
             goto fail;
         }
     }
-
     /* convert to XDR */
 fail:
     if (spec_fd >= 0)
@@ -1056,6 +1084,10 @@ fail:
                           (xdrproc_t)xdr_gf_getspec_rsp);
     free(args.key);  // malloced by xdr
     free(rsp.spec);
+
+    if (dict)
+        dict_unref(dict);
+
     if (args.xdata.xdata_val)
         free(args.xdata.xdata_val);
 
diff --git a/xlators/mgmt/glusterd/src/glusterd-op-sm.c b/xlators/mgmt/glusterd/src/glusterd-op-sm.c
index 9ea695e..454877b 100644
--- a/xlators/mgmt/glusterd/src/glusterd-op-sm.c
+++ b/xlators/mgmt/glusterd/src/glusterd-op-sm.c
@@ -655,6 +655,8 @@ glusterd_brick_op_build_payload(glusterd_op_t op,
             break;
     }
 
+    brick_req->dict.dict_len = 0;
+    brick_req->dict.dict_val = NULL;
     ret = dict_allocate_and_serialize(dict, &brick_req->input.input_val,
                                       &brick_req->input.input_len);
     if (ret)
@@ -723,6 +725,8 @@ glusterd_node_op_build_payload(glusterd_op_t op, gd1_mgmt_brick_op_req **req,
             goto out;
     }
 
+    brick_req->dict.dict_len = 0;
+    brick_req->dict.dict_val = NULL;
     ret = dict_allocate_and_serialize(dict, &brick_req->input.input_val,
                                       &brick_req->input.input_len);
 
diff --git a/xlators/mgmt/glusterd/src/glusterd-shd-svc-helper.c b/xlators/mgmt/glusterd/src/glusterd-shd-svc-helper.c
index 57ceda9..5661e39 100644
--- a/xlators/mgmt/glusterd/src/glusterd-shd-svc-helper.c
+++ b/xlators/mgmt/glusterd/src/glusterd-shd-svc-helper.c
@@ -126,3 +126,28 @@ glusterd_shd_svcproc_cleanup(glusterd_shdsvc_t *shd)
 out:
     return;
 }
+
+int
+glusterd_svc_set_shd_pidfile(glusterd_volinfo_t *volinfo, dict_t *dict)
+{
+    int ret = -1;
+    glusterd_svc_t *svc = NULL;
+    xlator_t *this = NULL;
+
+    this = THIS;
+    GF_VALIDATE_OR_GOTO("glusterd", this, out);
+    GF_VALIDATE_OR_GOTO(this->name, volinfo, out);
+    GF_VALIDATE_OR_GOTO(this->name, dict, out);
+
+    svc = &(volinfo->shd.svc);
+
+    ret = dict_set_dynstr_with_alloc(dict, "pidfile", svc->proc.pidfile);
+    if (ret) {
+        gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_DICT_SET_FAILED,
+               "Failed to set pidfile %s in dict", svc->proc.pidfile);
+        goto out;
+    }
+    ret = 0;
+out:
+    return ret;
+}
diff --git a/xlators/mgmt/glusterd/src/glusterd-shd-svc-helper.h b/xlators/mgmt/glusterd/src/glusterd-shd-svc-helper.h
index 59466ec..1f0984b 100644
--- a/xlators/mgmt/glusterd/src/glusterd-shd-svc-helper.h
+++ b/xlators/mgmt/glusterd/src/glusterd-shd-svc-helper.h
@@ -36,4 +36,7 @@ glusterd_recover_shd_attach_failure(glusterd_volinfo_t *volinfo,
 int
 glusterd_shdsvc_create_volfile(glusterd_volinfo_t *volinfo);
 
+int
+glusterd_svc_set_shd_pidfile(glusterd_volinfo_t *volinfo, dict_t *dict);
+
 #endif
diff --git a/xlators/mgmt/glusterd/src/glusterd-shd-svc.c b/xlators/mgmt/glusterd/src/glusterd-shd-svc.c
index 8ad90a9..590169f 100644
--- a/xlators/mgmt/glusterd/src/glusterd-shd-svc.c
+++ b/xlators/mgmt/glusterd/src/glusterd-shd-svc.c
@@ -258,14 +258,20 @@ glusterd_shdsvc_manager(glusterd_svc_t *svc, void *data, int flags)
     gf_boolean_t shd_restart = _gf_false;
 
     conf = THIS->private;
-    volinfo = data;
     GF_VALIDATE_OR_GOTO("glusterd", conf, out);
     GF_VALIDATE_OR_GOTO("glusterd", svc, out);
+    volinfo = data;
     GF_VALIDATE_OR_GOTO("glusterd", volinfo, out);
 
     if (volinfo)
         glusterd_volinfo_ref(volinfo);
 
+    if (volinfo->is_snap_volume) {
+        /* healing of a snap volume is not supported yet*/
+        ret = 0;
+        goto out;
+    }
+
     while (conf->restart_shd) {
         synclock_unlock(&conf->big_lock);
         sleep(2);
diff --git a/xlators/mgmt/glusterd/src/glusterd-svc-helper.c b/xlators/mgmt/glusterd/src/glusterd-svc-helper.c
index 400826f..e106111 100644
--- a/xlators/mgmt/glusterd/src/glusterd-svc-helper.c
+++ b/xlators/mgmt/glusterd/src/glusterd-svc-helper.c
@@ -519,7 +519,7 @@ glusterd_shd_svc_mux_init(glusterd_volinfo_t *volinfo, glusterd_svc_t *svc)
                 /* Take first entry from the process */
                 parent_svc = cds_list_entry(mux_proc->svcs.next, glusterd_svc_t,
                                             mux_svc);
-                sys_link(parent_svc->proc.pidfile, svc->proc.pidfile);
+                glusterd_copy_file(parent_svc->proc.pidfile, svc->proc.pidfile);
                 mux_conn = &parent_svc->conn;
                 if (volinfo)
                     volinfo->shd.attached = _gf_true;
@@ -623,12 +623,9 @@ glusterd_svc_attach_cbk(struct rpc_req *req, struct iovec *iov, int count,
     glusterd_volinfo_t *volinfo = NULL;
     glusterd_shdsvc_t *shd = NULL;
     glusterd_svc_t *svc = frame->cookie;
-    glusterd_svc_t *parent_svc = NULL;
-    glusterd_svc_proc_t *mux_proc = NULL;
     glusterd_conf_t *conf = NULL;
     int *flag = (int *)frame->local;
     xlator_t *this = THIS;
-    int pid = -1;
     int ret = -1;
     gf_getspec_rsp rsp = {
         0,
@@ -679,27 +676,7 @@ glusterd_svc_attach_cbk(struct rpc_req *req, struct iovec *iov, int count,
     }
 
     if (rsp.op_ret == 0) {
-        pthread_mutex_lock(&conf->attach_lock);
-        {
-            if (!strcmp(svc->name, "glustershd")) {
-                mux_proc = svc->svc_proc;
-                if (mux_proc &&
-                    !gf_is_service_running(svc->proc.pidfile, &pid)) {
-                    /*
-                     * When svc's are restarting, there is a chance that the
-                     * attached svc might not have updated it's pid. Because
-                     * it was at connection stage. So in that case, we need
-                     * to retry the pid file copy.
-                     */
-                    parent_svc = cds_list_entry(mux_proc->svcs.next,
-                                                glusterd_svc_t, mux_svc);
-                    if (parent_svc)
-                        sys_link(parent_svc->proc.pidfile, svc->proc.pidfile);
-                }
-            }
-            svc->online = _gf_true;
-        }
-        pthread_mutex_unlock(&conf->attach_lock);
+        svc->online = _gf_true;
         gf_msg(this->name, GF_LOG_INFO, 0, GD_MSG_SVC_ATTACH_FAIL,
                "svc %s of volume %s attached successfully to pid %d", svc->name,
                volinfo->volname, glusterd_proc_get_pid(&svc->proc));
@@ -726,7 +703,7 @@ out:
 
 extern size_t
 build_volfile_path(char *volume_id, char *path, size_t path_len,
-                   char *trusted_str);
+                   char *trusted_str, dict_t *dict);
 
 int
 __glusterd_send_svc_configure_req(glusterd_svc_t *svc, int flags,
@@ -751,6 +728,7 @@ __glusterd_send_svc_configure_req(glusterd_svc_t *svc, int flags,
     ssize_t req_size = 0;
     call_frame_t *frame = NULL;
     gd1_mgmt_brick_op_req brick_req;
+    dict_t *dict = NULL;
     void *req = &brick_req;
     void *errlbl = &&err;
     struct rpc_clnt_connection *conn;
@@ -776,6 +754,8 @@ __glusterd_send_svc_configure_req(glusterd_svc_t *svc, int flags,
     brick_req.name = volfile_id;
     brick_req.input.input_val = NULL;
     brick_req.input.input_len = 0;
+    brick_req.dict.dict_val = NULL;
+    brick_req.dict.dict_len = 0;
 
     frame = create_frame(this, this->ctx->pool);
     if (!frame) {
@@ -783,7 +763,13 @@ __glusterd_send_svc_configure_req(glusterd_svc_t *svc, int flags,
     }
 
     if (op == GLUSTERD_SVC_ATTACH) {
-        (void)build_volfile_path(volfile_id, path, sizeof(path), NULL);
+        dict = dict_new();
+        if (!dict) {
+            ret = -ENOMEM;
+            goto *errlbl;
+        }
+
+        (void)build_volfile_path(volfile_id, path, sizeof(path), NULL, dict);
 
         ret = sys_stat(path, &stbuf);
         if (ret < 0) {
@@ -818,6 +804,18 @@ __glusterd_send_svc_configure_req(glusterd_svc_t *svc, int flags,
             ret = -EIO;
             goto *errlbl;
         }
+        if (dict->count > 0) {
+            ret = dict_allocate_and_serialize(dict, &brick_req.dict.dict_val,
+                                              &brick_req.dict.dict_len);
+            if (ret) {
+                gf_msg(this->name, GF_LOG_ERROR, 0,
+                       GD_MSG_DICT_SERL_LENGTH_GET_FAIL,
+                       "Failed to serialize dict "
+                       "to request buffer");
+                goto *errlbl;
+            }
+            dict->extra_free = brick_req.dict.dict_val;
+        }
 
         frame->cookie = svc;
         frame->local = GF_CALLOC(1, sizeof(int), gf_gld_mt_int);
@@ -862,6 +860,8 @@ __glusterd_send_svc_configure_req(glusterd_svc_t *svc, int flags,
     GF_ATOMIC_INC(conf->blockers);
     ret = rpc_clnt_submit(rpc, &gd_brick_prog, op, cbkfn, &iov, 1, NULL, 0,
                           iobref, frame, NULL, 0, NULL, 0, NULL);
+    if (dict)
+        dict_unref(dict);
     GF_FREE(volfile_content);
     if (spec_fd >= 0)
         sys_close(spec_fd);
@@ -874,6 +874,9 @@ maybe_free_iobuf:
         iobuf_unref(iobuf);
     }
 err:
+    if (dict)
+        dict_unref(dict);
+
     GF_FREE(volfile_content);
     if (spec_fd >= 0)
         sys_close(spec_fd);
diff --git a/xlators/mgmt/glusterd/src/glusterd-syncop.c b/xlators/mgmt/glusterd/src/glusterd-syncop.c
index 618d8bc..a8098df 100644
--- a/xlators/mgmt/glusterd/src/glusterd-syncop.c
+++ b/xlators/mgmt/glusterd/src/glusterd-syncop.c
@@ -143,6 +143,8 @@ gd_brick_op_req_free(gd1_mgmt_brick_op_req *req)
     if (!req)
         return;
 
+    if (req->dict.dict_val)
+        GF_FREE(req->dict.dict_val);
     GF_FREE(req->input.input_val);
     GF_FREE(req);
 }
diff --git a/xlators/mgmt/glusterd/src/glusterd-utils.c b/xlators/mgmt/glusterd/src/glusterd-utils.c
index 4c487d0..2eb5116 100644
--- a/xlators/mgmt/glusterd/src/glusterd-utils.c
+++ b/xlators/mgmt/glusterd/src/glusterd-utils.c
@@ -5914,6 +5914,8 @@ send_attach_req(xlator_t *this, struct rpc_clnt *rpc, char *path,
     brick_req.name = path;
     brick_req.input.input_val = NULL;
     brick_req.input.input_len = 0;
+    brick_req.dict.dict_val = NULL;
+    brick_req.dict.dict_len = 0;
 
     req_size = xdr_sizeof((xdrproc_t)xdr_gd1_mgmt_brick_op_req, req);
     iobuf = iobuf_get2(rpc->ctx->iobuf_pool, req_size);
@@ -5977,7 +5979,7 @@ err:
 
 extern size_t
 build_volfile_path(char *volume_id, char *path, size_t path_len,
-                   char *trusted_str);
+                   char *trusted_str, dict_t *dict);
 
 static int
 attach_brick(xlator_t *this, glusterd_brickinfo_t *brickinfo,
@@ -6022,7 +6024,7 @@ attach_brick(xlator_t *this, glusterd_brickinfo_t *brickinfo,
         goto out;
     }
 
-    (void)build_volfile_path(full_id, path, sizeof(path), NULL);
+    (void)build_volfile_path(full_id, path, sizeof(path), NULL, NULL);
 
     for (tries = 15; tries > 0; --tries) {
         rpc = rpc_clnt_ref(other_brick->rpc);
-- 
1.8.3.1