From 3d230880aed85737365deafe3c9a32c67da2a79e Mon Sep 17 00:00:00 2001
From: Susant Palai <spalai@redhat.com>
Date: Mon, 4 May 2020 19:09:00 +0530
Subject: [PATCH 370/375] dht: Do opendir selectively in gf_defrag_process_dir

Currently opendir is done from the cluster view. Hence, even if
one opendir is successful, the opendir operation as a whole is considered
successful.

But since in gf_defrag_get_entry we fetch entries selectively from
local_subvols, we need to opendir individually on those local subvols
and keep track of fds separately. Otherwise it is possible that opendir
failed on one of the subvol and we wind readdirp call on the fd to the
corresponding subvol, which will ultimately result in EINVAL error.

> fixes: #1218
> Change-Id: I50dd88b9597852a15579f4ee325918979417f570
> Signed-off-by: Susant Palai <spalai@redhat.com>
(Backport of https://review.gluster.org/#/c/glusterfs/+/24404/)

BUG: 1831403
Change-Id: I96e19fdd630279c3ef44f361c1d1fc5c1c429821
Signed-off-by: Susant Palai <spalai@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/200306
Tested-by: RHGS Build Bot <nigelb@redhat.com>
Reviewed-by: Sunil Kumar Heggodu Gopala Acharya <sheggodu@redhat.com>
---
 xlators/cluster/dht/src/dht-common.h    |  2 +
 xlators/cluster/dht/src/dht-rebalance.c | 74 +++++++++++++++++++++++----------
 2 files changed, 54 insertions(+), 22 deletions(-)

diff --git a/xlators/cluster/dht/src/dht-common.h b/xlators/cluster/dht/src/dht-common.h
index 4d2aae6..8e65111 100644
--- a/xlators/cluster/dht/src/dht-common.h
+++ b/xlators/cluster/dht/src/dht-common.h
@@ -742,6 +742,8 @@ struct dir_dfmeta {
     struct list_head **head;
     struct list_head **iterator;
     int *fetch_entries;
+    /* fds corresponding to local subvols only */
+    fd_t **lfd;
 };
 
 typedef struct dht_migrate_info {
diff --git a/xlators/cluster/dht/src/dht-rebalance.c b/xlators/cluster/dht/src/dht-rebalance.c
index 33cacfe..c692119 100644
--- a/xlators/cluster/dht/src/dht-rebalance.c
+++ b/xlators/cluster/dht/src/dht-rebalance.c
@@ -48,6 +48,8 @@ gf_defrag_free_dir_dfmeta(struct dir_dfmeta *meta, int local_subvols_cnt)
     if (meta) {
         for (i = 0; i < local_subvols_cnt; i++) {
             gf_dirent_free(&meta->equeue[i]);
+            if (meta->lfd && meta->lfd[i])
+                fd_unref(meta->lfd[i]);
         }
 
         GF_FREE(meta->equeue);
@@ -55,6 +57,7 @@ gf_defrag_free_dir_dfmeta(struct dir_dfmeta *meta, int local_subvols_cnt)
         GF_FREE(meta->iterator);
         GF_FREE(meta->offset_var);
         GF_FREE(meta->fetch_entries);
+        GF_FREE(meta->lfd);
         GF_FREE(meta);
     }
 }
@@ -3095,7 +3098,7 @@ int static gf_defrag_get_entry(xlator_t *this, int i,
                                struct dir_dfmeta *dir_dfmeta, dict_t *xattr_req,
                                int *should_commit_hash, int *perrno)
 {
-    int ret = -1;
+    int ret = 0;
     char is_linkfile = 0;
     gf_dirent_t *df_entry = NULL;
     struct dht_container *tmp_container = NULL;
@@ -3111,6 +3114,13 @@ int static gf_defrag_get_entry(xlator_t *this, int i,
     }
 
     if (dir_dfmeta->fetch_entries[i] == 1) {
+        if (!fd) {
+            dir_dfmeta->fetch_entries[i] = 0;
+            dir_dfmeta->offset_var[i].readdir_done = 1;
+            ret = 0;
+            goto out;
+        }
+
         ret = syncop_readdirp(conf->local_subvols[i], fd, 131072,
                               dir_dfmeta->offset_var[i].offset,
                               &(dir_dfmeta->equeue[i]), xattr_req, NULL);
@@ -3270,7 +3280,6 @@ gf_defrag_process_dir(xlator_t *this, gf_defrag_info_t *defrag, loc_t *loc,
                       dict_t *migrate_data, int *perrno)
 {
     int ret = -1;
-    fd_t *fd = NULL;
     dht_conf_t *conf = NULL;
     gf_dirent_t entries;
     dict_t *xattr_req = NULL;
@@ -3304,28 +3313,49 @@ gf_defrag_process_dir(xlator_t *this, gf_defrag_info_t *defrag, loc_t *loc,
         goto out;
     }
 
-    fd = fd_create(loc->inode, defrag->pid);
-    if (!fd) {
-        gf_log(this->name, GF_LOG_ERROR, "Failed to create fd");
+    dir_dfmeta = GF_CALLOC(1, sizeof(*dir_dfmeta), gf_common_mt_pointer);
+    if (!dir_dfmeta) {
+        gf_log(this->name, GF_LOG_ERROR, "dir_dfmeta is NULL");
         ret = -1;
         goto out;
     }
 
-    ret = syncop_opendir(this, loc, fd, NULL, NULL);
-    if (ret) {
-        gf_msg(this->name, GF_LOG_WARNING, -ret, DHT_MSG_MIGRATE_DATA_FAILED,
-               "Migrate data failed: Failed to open dir %s", loc->path);
-        *perrno = -ret;
+    dir_dfmeta->lfd = GF_CALLOC(local_subvols_cnt, sizeof(fd_t *),
+                                gf_common_mt_pointer);
+    if (!dir_dfmeta->lfd) {
+        gf_msg(this->name, GF_LOG_ERROR, ENOMEM, 0,
+               "could not allocate memory for dir_dfmeta");
         ret = -1;
+        *perrno = ENOMEM;
         goto out;
     }
 
-    fd_bind(fd);
-    dir_dfmeta = GF_CALLOC(1, sizeof(*dir_dfmeta), gf_common_mt_pointer);
-    if (!dir_dfmeta) {
-        gf_log(this->name, GF_LOG_ERROR, "dir_dfmeta is NULL");
-        ret = -1;
-        goto out;
+    for (i = 0; i < local_subvols_cnt; i++) {
+        dir_dfmeta->lfd[i] = fd_create(loc->inode, defrag->pid);
+        if (!dir_dfmeta->lfd[i]) {
+            gf_msg(this->name, GF_LOG_ERROR, ENOMEM, 0, "failed to create fd");
+            *perrno = ENOMEM;
+            ret = -1;
+            goto out;
+        }
+
+        ret = syncop_opendir(conf->local_subvols[i], loc, dir_dfmeta->lfd[i],
+                             NULL, NULL);
+        if (ret) {
+            fd_unref(dir_dfmeta->lfd[i]);
+            dir_dfmeta->lfd[i] = NULL;
+            gf_smsg(this->name, GF_LOG_WARNING, 0, 0,
+                    "failed to open dir: %s subvol: %s", loc->path,
+                    conf->local_subvols[i]->name);
+
+            if (conf->decommission_in_progress) {
+                *perrno = -ret;
+                ret = -1;
+                goto out;
+            }
+        } else {
+            fd_bind(dir_dfmeta->lfd[i]);
+        }
     }
 
     dir_dfmeta->head = GF_CALLOC(local_subvols_cnt, sizeof(*(dir_dfmeta->head)),
@@ -3360,6 +3390,7 @@ gf_defrag_process_dir(xlator_t *this, gf_defrag_info_t *defrag, loc_t *loc,
         ret = -1;
         goto out;
     }
+
     ret = gf_defrag_ctx_subvols_init(dir_dfmeta->offset_var, this);
     if (ret) {
         gf_log(this->name, GF_LOG_ERROR,
@@ -3372,7 +3403,8 @@ gf_defrag_process_dir(xlator_t *this, gf_defrag_info_t *defrag, loc_t *loc,
     dir_dfmeta->fetch_entries = GF_CALLOC(local_subvols_cnt, sizeof(int),
                                           gf_common_mt_int);
     if (!dir_dfmeta->fetch_entries) {
-        gf_log(this->name, GF_LOG_ERROR, "dir_dfmeta->fetch_entries is NULL");
+        gf_msg(this->name, GF_LOG_ERROR, ENOMEM, 0,
+               "could not allocate memory for dir_dfmeta->fetch_entries");
         ret = -1;
         goto out;
     }
@@ -3442,8 +3474,9 @@ gf_defrag_process_dir(xlator_t *this, gf_defrag_info_t *defrag, loc_t *loc,
             ldfq_count <= MAX_MIGRATE_QUEUE_COUNT &&
             !dht_dfreaddirp_done(dir_dfmeta->offset_var, local_subvols_cnt)) {
             ret = gf_defrag_get_entry(this, dfc_index, &container, loc, conf,
-                                      defrag, fd, migrate_data, dir_dfmeta,
-                                      xattr_req, &should_commit_hash, perrno);
+                                      defrag, dir_dfmeta->lfd[dfc_index],
+                                      migrate_data, dir_dfmeta, xattr_req,
+                                      &should_commit_hash, perrno);
 
             if (ret) {
                 gf_log(this->name, GF_LOG_WARNING,
@@ -3497,9 +3530,6 @@ out:
     if (xattr_req)
         dict_unref(xattr_req);
 
-    if (fd)
-        fd_unref(fd);
-
     if (ret == 0 && should_commit_hash == 0) {
         ret = 2;
     }
-- 
1.8.3.1