path: root/0260-ctime-Set-mdata-xattr-on-legacy-files.patch

From fc0903de1f7565e06db9d41e6dfd62221a745d24 Mon Sep 17 00:00:00 2001
From: Kotresh HR <khiremat@redhat.com>
Date: Mon, 24 Jun 2019 13:06:49 +0530
Subject: [PATCH 260/261] ctime: Set mdata xattr on legacy files

Problem:
The files which were created before ctime enabled would not
have "trusted.glusterfs.mdata"(stores time attributes) xattr.
Upon fops which modifies either ctime or mtime, the xattr
gets created with latest ctime, mtime and atime, which is
incorrect. It should update only the corresponding time
attribute and rest from backend

Solution:
Creating xattr with values from brick is not possible as
each brick of replica set would have different times.
So create the xattr upon successful lookup if the xattr
is not created

Note To Reviewers:
The time attributes used to set xattr is got from successful
lookup. Instead of sending the whole iatt over the wire via
setxattr, a structure called mdata_iatt is sent. The mdata_iatt
contains only time attributes.

Backport of
 > Patch: https://review.gluster.org/22936
 > Change-Id: I5e535631ddef04195361ae0364336410a2895dd4
 > fixes: bz#1593542

Change-Id: I5e535631ddef04195361ae0364336410a2895dd4
BUG: 1715422
Signed-off-by: Kotresh HR <khiremat@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/176725
Tested-by: RHGS Build Bot <nigelb@redhat.com>
Reviewed-by: Amar Tumballi Suryanarayan <amarts@redhat.com>
Reviewed-by: Atin Mukherjee <amukherj@redhat.com>
---
 libglusterfs/src/dict.c                        |  59 ++++++++++
 libglusterfs/src/glusterfs/dict.h              |   5 +
 libglusterfs/src/glusterfs/glusterfs.h         |   3 +
 libglusterfs/src/glusterfs/iatt.h              |  20 ++++
 libglusterfs/src/libglusterfs.sym              |   3 +
 rpc/xdr/src/glusterfs-fops.x                   |   1 +
 rpc/xdr/src/glusterfs3.h                       |  59 ++++++++++
 rpc/xdr/src/glusterfs4-xdr.x                   |  12 ++
 rpc/xdr/src/libgfxdr.sym                       |   3 +-
 tests/basic/ctime/ctime-mdata-legacy-files.t   |  83 +++++++++++++
 xlators/features/utime/src/utime-messages.h    |   3 +-
 xlators/features/utime/src/utime.c             | 154 ++++++++++++++++++++++---
 xlators/storage/posix/src/posix-inode-fd-ops.c |  17 +++
 xlators/storage/posix/src/posix-messages.h     |   3 +-
 xlators/storage/posix/src/posix-metadata.c     | 103 ++++++++++-------
 xlators/storage/posix/src/posix-metadata.h     |   4 +
 16 files changed, 475 insertions(+), 57 deletions(-)
 create mode 100644 tests/basic/ctime/ctime-mdata-legacy-files.t

diff --git a/libglusterfs/src/dict.c b/libglusterfs/src/dict.c
index 6917df9..d8cdda4 100644
--- a/libglusterfs/src/dict.c
+++ b/libglusterfs/src/dict.c
@@ -124,6 +124,7 @@ int32_t
 is_data_equal(data_t *one, data_t *two)
 {
     struct iatt *iatt1, *iatt2;
+    struct mdata_iatt *mdata_iatt1, *mdata_iatt2;
 
     if (!one || !two || !one->data || !two->data) {
         gf_msg_callingfn("dict", GF_LOG_ERROR, EINVAL, LG_MSG_INVALID_ARG,
@@ -188,6 +189,24 @@ is_data_equal(data_t *one, data_t *two)
         */
         return 1;
     }
+    if (one->data_type == GF_DATA_TYPE_MDATA) {
+        if ((one->len < sizeof(struct mdata_iatt)) ||
+            (two->len < sizeof(struct mdata_iatt))) {
+            return 0;
+        }
+        mdata_iatt1 = (struct mdata_iatt *)one->data;
+        mdata_iatt2 = (struct mdata_iatt *)two->data;
+
+        if (mdata_iatt1->ia_atime != mdata_iatt2->ia_atime ||
+            mdata_iatt1->ia_mtime != mdata_iatt2->ia_mtime ||
+            mdata_iatt1->ia_ctime != mdata_iatt2->ia_ctime ||
+            mdata_iatt1->ia_atime_nsec != mdata_iatt2->ia_atime_nsec ||
+            mdata_iatt1->ia_mtime_nsec != mdata_iatt2->ia_mtime_nsec ||
+            mdata_iatt1->ia_ctime_nsec != mdata_iatt2->ia_ctime_nsec) {
+            return 0;
+        }
+        return 1;
+    }
 
     if (one->len != two->len)
         return 0;
@@ -1078,6 +1097,7 @@ static char *data_type_name[GF_DATA_TYPE_MAX] = {
     [GF_DATA_TYPE_PTR] = "pointer",
     [GF_DATA_TYPE_GFUUID] = "gf-uuid",
     [GF_DATA_TYPE_IATT] = "iatt",
+    [GF_DATA_TYPE_MDATA] = "mdata",
 };
 
 int64_t
@@ -2666,6 +2686,45 @@ err:
 }
 
 int
+dict_set_mdata(dict_t *this, char *key, struct mdata_iatt *mdata,
+               bool is_static)
+{
+    return dict_set_bin_common(this, key, mdata, sizeof(struct mdata_iatt),
+                               is_static, GF_DATA_TYPE_MDATA);
+}
+
+int
+dict_get_mdata(dict_t *this, char *key, struct mdata_iatt *mdata)
+{
+    data_t *data = NULL;
+    int ret = -EINVAL;
+
+    if (!this || !key || !mdata) {
+        goto err;
+    }
+    ret = dict_get_with_ref(this, key, &data);
+    if (ret < 0) {
+        goto err;
+    }
+
+    VALIDATE_DATA_AND_LOG(data, GF_DATA_TYPE_MDATA, key, -EINVAL);
+    if (data->len < sizeof(struct mdata_iatt)) {
+        gf_msg("glusterfs", GF_LOG_ERROR, ENOBUFS, LG_MSG_UNDERSIZED_BUF,
+               "data value for '%s' is smaller than expected", key);
+        ret = -ENOBUFS;
+        goto err;
+    }
+
+    memcpy(mdata, data->data, min(data->len, sizeof(struct mdata_iatt)));
+
+err:
+    if (data)
+        data_unref(data);
+
+    return ret;
+}
+
+int
 dict_set_iatt(dict_t *this, char *key, struct iatt *iatt, bool is_static)
 {
     return dict_set_bin_common(this, key, iatt, sizeof(struct iatt), is_static,
diff --git a/libglusterfs/src/glusterfs/dict.h b/libglusterfs/src/glusterfs/dict.h
index 022f564..8239c7a 100644
--- a/libglusterfs/src/glusterfs/dict.h
+++ b/libglusterfs/src/glusterfs/dict.h
@@ -392,6 +392,11 @@ GF_MUST_CHECK int
 dict_set_iatt(dict_t *this, char *key, struct iatt *iatt, bool is_static);
 GF_MUST_CHECK int
 dict_get_iatt(dict_t *this, char *key, struct iatt *iatt);
+GF_MUST_CHECK int
+dict_set_mdata(dict_t *this, char *key, struct mdata_iatt *mdata,
+               bool is_static);
+GF_MUST_CHECK int
+dict_get_mdata(dict_t *this, char *key, struct mdata_iatt *mdata);
 
 void
 dict_dump_to_statedump(dict_t *dict, char *dict_name, char *domain);
diff --git a/libglusterfs/src/glusterfs/glusterfs.h b/libglusterfs/src/glusterfs/glusterfs.h
index 2cedf1a..79c93ae 100644
--- a/libglusterfs/src/glusterfs/glusterfs.h
+++ b/libglusterfs/src/glusterfs/glusterfs.h
@@ -229,6 +229,9 @@ enum gf_internal_fop_indicator {
 #define VIRTUAL_QUOTA_XATTR_CLEANUP_KEY "glusterfs.quota-xattr-cleanup"
 #define QUOTA_READ_ONLY_KEY "trusted.glusterfs.quota.read-only"
 
+/* ctime related */
+#define CTIME_MDATA_XDATA_KEY "set-ctime-mdata"
+
 /* afr related */
 #define AFR_XATTR_PREFIX "trusted.afr"
 
diff --git a/libglusterfs/src/glusterfs/iatt.h b/libglusterfs/src/glusterfs/iatt.h
index bee7a0a..f03d68b 100644
--- a/libglusterfs/src/glusterfs/iatt.h
+++ b/libglusterfs/src/glusterfs/iatt.h
@@ -92,6 +92,15 @@ struct old_iatt {
     uint32_t ia_ctime_nsec;
 };
 
+struct mdata_iatt {
+    int64_t ia_atime; /* last access time */
+    int64_t ia_mtime; /* last modification time */
+    int64_t ia_ctime; /* last status change time */
+    uint32_t ia_atime_nsec;
+    uint32_t ia_mtime_nsec;
+    uint32_t ia_ctime_nsec;
+};
+
 /* 64-bit mask for valid members in struct iatt. */
 #define IATT_TYPE 0x0000000000000001U
 #define IATT_MODE 0x0000000000000002U
@@ -313,6 +322,17 @@ st_mode_from_ia(ia_prot_t prot, ia_type_t type)
     return st_mode;
 }
 
+static inline void
+iatt_to_mdata(struct mdata_iatt *mdata, struct iatt *iatt)
+{
+    mdata->ia_atime = iatt->ia_atime;
+    mdata->ia_atime_nsec = iatt->ia_atime_nsec;
+    mdata->ia_mtime = iatt->ia_mtime;
+    mdata->ia_mtime_nsec = iatt->ia_mtime_nsec;
+    mdata->ia_ctime = iatt->ia_ctime;
+    mdata->ia_ctime_nsec = iatt->ia_ctime_nsec;
+}
+
 static inline int
 iatt_from_stat(struct iatt *iatt, struct stat *stat)
 {
diff --git a/libglusterfs/src/libglusterfs.sym b/libglusterfs/src/libglusterfs.sym
index 4dca7de..b161380 100644
--- a/libglusterfs/src/libglusterfs.sym
+++ b/libglusterfs/src/libglusterfs.sym
@@ -380,6 +380,7 @@ dict_get_bin
 dict_get_double
 dict_get_gfuuid
 dict_get_iatt
+dict_get_mdata
 dict_get_int16
 dict_get_int32
 dict_get_int32n
@@ -417,6 +418,7 @@ dict_set_dynstrn
 dict_set_dynstr_with_alloc
 dict_set_gfuuid
 dict_set_iatt
+dict_set_mdata
 dict_set_int16
 dict_set_int32
 dict_set_int32n
@@ -509,6 +511,7 @@ fop_lease_stub
 fop_link_stub
 fop_lk_stub
 fop_log_level
+fop_lookup_cbk_stub
 fop_lookup_stub
 fop_mkdir_stub
 fop_mknod_stub
diff --git a/rpc/xdr/src/glusterfs-fops.x b/rpc/xdr/src/glusterfs-fops.x
index bacf0773..651f8de 100644
--- a/rpc/xdr/src/glusterfs-fops.x
+++ b/rpc/xdr/src/glusterfs-fops.x
@@ -245,5 +245,6 @@ enum gf_dict_data_type_t {
         GF_DATA_TYPE_PTR,
         GF_DATA_TYPE_GFUUID,
         GF_DATA_TYPE_IATT,
+        GF_DATA_TYPE_MDATA,
         GF_DATA_TYPE_MAX
 };
diff --git a/rpc/xdr/src/glusterfs3.h b/rpc/xdr/src/glusterfs3.h
index 5521f4d..86b3a4c 100644
--- a/rpc/xdr/src/glusterfs3.h
+++ b/rpc/xdr/src/glusterfs3.h
@@ -585,6 +585,34 @@ out:
 }
 
 static inline void
+gfx_mdata_iatt_to_mdata_iatt(struct gfx_mdata_iatt *gf_mdata_iatt,
+                             struct mdata_iatt *mdata_iatt)
+{
+    if (!mdata_iatt || !gf_mdata_iatt)
+        return;
+    mdata_iatt->ia_atime = gf_mdata_iatt->ia_atime;
+    mdata_iatt->ia_atime_nsec = gf_mdata_iatt->ia_atime_nsec;
+    mdata_iatt->ia_mtime = gf_mdata_iatt->ia_mtime;
+    mdata_iatt->ia_mtime_nsec = gf_mdata_iatt->ia_mtime_nsec;
+    mdata_iatt->ia_ctime = gf_mdata_iatt->ia_ctime;
+    mdata_iatt->ia_ctime_nsec = gf_mdata_iatt->ia_ctime_nsec;
+}
+
+static inline void
+gfx_mdata_iatt_from_mdata_iatt(struct gfx_mdata_iatt *gf_mdata_iatt,
+                               struct mdata_iatt *mdata_iatt)
+{
+    if (!mdata_iatt || !gf_mdata_iatt)
+        return;
+    gf_mdata_iatt->ia_atime = mdata_iatt->ia_atime;
+    gf_mdata_iatt->ia_atime_nsec = mdata_iatt->ia_atime_nsec;
+    gf_mdata_iatt->ia_mtime = mdata_iatt->ia_mtime;
+    gf_mdata_iatt->ia_mtime_nsec = mdata_iatt->ia_mtime_nsec;
+    gf_mdata_iatt->ia_ctime = mdata_iatt->ia_ctime;
+    gf_mdata_iatt->ia_ctime_nsec = mdata_iatt->ia_ctime_nsec;
+}
+
+static inline void
 gfx_stat_to_iattx(struct gfx_iattx *gf_stat, struct iatt *iatt)
 {
     if (!iatt || !gf_stat)
@@ -721,6 +749,12 @@ dict_to_xdr(dict_t *this, gfx_dict *dict)
                 gfx_stat_from_iattx(&xpair->value.gfx_value_u.iatt,
                                     (struct iatt *)dpair->value->data);
                 break;
+            case GF_DATA_TYPE_MDATA:
+                index++;
+                gfx_mdata_iatt_from_mdata_iatt(
+                    &xpair->value.gfx_value_u.mdata_iatt,
+                    (struct mdata_iatt *)dpair->value->data);
+                break;
             case GF_DATA_TYPE_GFUUID:
                 index++;
                 memcpy(&xpair->value.gfx_value_u.uuid, dpair->value->data,
@@ -787,6 +821,7 @@ xdr_to_dict(gfx_dict *dict, dict_t **to)
     dict_t *this = NULL;
     unsigned char *uuid = NULL;
     struct iatt *iatt = NULL;
+    struct mdata_iatt *mdata_iatt = NULL;
 
     if (!to || !dict)
         goto out;
@@ -854,6 +889,30 @@ xdr_to_dict(gfx_dict *dict, dict_t **to)
                 gfx_stat_to_iattx(&xpair->value.gfx_value_u.iatt, iatt);
                 ret = dict_set_iatt(this, key, iatt, false);
                 break;
+            case GF_DATA_TYPE_MDATA:
+                mdata_iatt = GF_CALLOC(1, sizeof(struct mdata_iatt),
+                                       gf_common_mt_char);
+                if (!mdata_iatt) {
+                    errno = ENOMEM;
+                    gf_msg(THIS->name, GF_LOG_ERROR, ENOMEM, LG_MSG_NO_MEMORY,
+                           "failed to allocate memory. key: %s", key);
+                    ret = -1;
+                    goto out;
+                }
+                gfx_mdata_iatt_to_mdata_iatt(
+                    &xpair->value.gfx_value_u.mdata_iatt, mdata_iatt);
+                ret = dict_set_mdata(this, key, mdata_iatt, false);
+                if (ret != 0) {
+                    GF_FREE(mdata_iatt);
+                    gf_msg(THIS->name, GF_LOG_ERROR, ENOMEM,
+                           LG_MSG_DICT_SET_FAILED,
+                           "failed to set the key (%s)"
+                           " into dict",
+                           key);
+                    ret = -1;
+                    goto out;
+                }
+                break;
             case GF_DATA_TYPE_PTR:
             case GF_DATA_TYPE_STR_OLD:
                 value = GF_MALLOC(xpair->value.gfx_value_u.other.other_len + 1,
diff --git a/rpc/xdr/src/glusterfs4-xdr.x b/rpc/xdr/src/glusterfs4-xdr.x
index bec0872..6f92b70 100644
--- a/rpc/xdr/src/glusterfs4-xdr.x
+++ b/rpc/xdr/src/glusterfs4-xdr.x
@@ -46,6 +46,16 @@ struct gfx_iattx {
         unsigned int     mode;          /* type of file and rwx mode */
 };
 
+struct gfx_mdata_iatt {
+        hyper      ia_atime;      /* last access time */
+        hyper      ia_mtime;      /* last modification time */
+        hyper      ia_ctime;      /* last status change time */
+
+        unsigned int     ia_atime_nsec;
+        unsigned int     ia_mtime_nsec;
+        unsigned int     ia_ctime_nsec;
+};
+
 union gfx_value switch (gf_dict_data_type_t type) {
         case GF_DATA_TYPE_INT:
                 hyper value_int;
@@ -62,6 +72,8 @@ union gfx_value switch (gf_dict_data_type_t type) {
         case GF_DATA_TYPE_PTR:
         case GF_DATA_TYPE_STR_OLD:
                 opaque other<>;
+        case GF_DATA_TYPE_MDATA:
+                gfx_mdata_iatt mdata_iatt;
 };
 
 /* AUTH */
diff --git a/rpc/xdr/src/libgfxdr.sym b/rpc/xdr/src/libgfxdr.sym
index 22cdf30..dd4ac85 100644
--- a/rpc/xdr/src/libgfxdr.sym
+++ b/rpc/xdr/src/libgfxdr.sym
@@ -251,6 +251,7 @@ xdr_to_write3args
 xdr_vector_round_up
 xdr_gfx_read_rsp
 xdr_gfx_iattx
+xdr_gfx_mdata_iatt
 xdr_gfx_value
 xdr_gfx_dict_pair
 xdr_gfx_dict
@@ -344,4 +345,4 @@ xdr_compound_req_v2
 xdr_gfx_compound_req
 xdr_compound_rsp_v2
 xdr_gfx_compound_rsp
-xdr_gfx_copy_file_range_req
\ No newline at end of file
+xdr_gfx_copy_file_range_req
diff --git a/tests/basic/ctime/ctime-mdata-legacy-files.t b/tests/basic/ctime/ctime-mdata-legacy-files.t
new file mode 100644
index 0000000..2e782d5
--- /dev/null
+++ b/tests/basic/ctime/ctime-mdata-legacy-files.t
@@ -0,0 +1,83 @@
+#!/bin/bash
+. $(dirname $0)/../../include.rc
+. $(dirname $0)/../../volume.rc
+. $(dirname $0)/../../afr.rc
+cleanup;
+
+###############################################################################
+#Replica volume
+
+TEST glusterd
+TEST pidof glusterd
+TEST $CLI volume create $V0 replica 3 $H0:$B0/${V0}{0,1,2}
+TEST $CLI volume set $V0 performance.stat-prefetch off
+TEST $CLI volume start $V0
+
+TEST glusterfs --volfile-id=$V0 --volfile-server=$H0 --entry-timeout=0 $M0;
+
+#Disable ctime and create file, file doesn't contain "trusted.glusterfs.mdata" xattr
+TEST $CLI volume set $V0 ctime off
+
+TEST "mkdir $M0/DIR"
+TEST "echo hello_world > $M0/DIR/FILE"
+
+#Verify absence of xattr
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT "" check_for_xattr 'trusted.glusterfs.mdata' "$B0/${V0}0/DIR"
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT "" check_for_xattr 'trusted.glusterfs.mdata' "$B0/${V0}0/DIR/FILE"
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT "" check_for_xattr 'trusted.glusterfs.mdata' "$B0/${V0}1/DIR"
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT "" check_for_xattr 'trusted.glusterfs.mdata' "$B0/${V0}1/DIR/FILE"
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT "" check_for_xattr 'trusted.glusterfs.mdata' "$B0/${V0}2/DIR"
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT "" check_for_xattr 'trusted.glusterfs.mdata' "$B0/${V0}2/DIR/FILE"
+
+#Enable ctime
+TEST $CLI volume set $V0 ctime on
+sleep 3
+TEST stat $M0/DIR/FILE
+
+#Verify presence "trusted.glusterfs.mdata" xattr on backend
+#The lookup above should have created xattr
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT 'trusted.glusterfs.mdata' check_for_xattr 'trusted.glusterfs.mdata' "$B0/${V0}0/DIR"
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT 'trusted.glusterfs.mdata' check_for_xattr 'trusted.glusterfs.mdata' "$B0/${V0}0/DIR/FILE"
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT 'trusted.glusterfs.mdata' check_for_xattr 'trusted.glusterfs.mdata' "$B0/${V0}1/DIR"
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT 'trusted.glusterfs.mdata' check_for_xattr 'trusted.glusterfs.mdata' "$B0/${V0}1/DIR/FILE"
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT 'trusted.glusterfs.mdata' check_for_xattr 'trusted.glusterfs.mdata' "$B0/${V0}2/DIR"
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT 'trusted.glusterfs.mdata' check_for_xattr 'trusted.glusterfs.mdata' "$B0/${V0}2/DIR/FILE"
+
+###############################################################################
+#Disperse Volume
+
+TEST $CLI volume create $V1 disperse 3 redundancy 1  $H0:$B0/${V1}{0,1,2}
+TEST $CLI volume set $V1 performance.stat-prefetch off
+TEST $CLI volume start $V1
+
+TEST glusterfs --volfile-id=$V1 --volfile-server=$H0 --entry-timeout=0 $M1;
+
+#Disable ctime and create file, file doesn't contain "trusted.glusterfs.mdata" xattr
+TEST $CLI volume set $V1 ctime off
+TEST "mkdir $M1/DIR"
+TEST "echo hello_world > $M1/DIR/FILE"
+
+#Verify absence of xattr
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT "" check_for_xattr 'trusted.glusterfs.mdata' "$B0/${V1}0/DIR"
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT "" check_for_xattr 'trusted.glusterfs.mdata' "$B0/${V1}0/DIR/FILE"
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT "" check_for_xattr 'trusted.glusterfs.mdata' "$B0/${V1}1/DIR"
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT "" check_for_xattr 'trusted.glusterfs.mdata' "$B0/${V1}1/DIR/FILE"
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT "" check_for_xattr 'trusted.glusterfs.mdata' "$B0/${V1}2/DIR"
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT "" check_for_xattr 'trusted.glusterfs.mdata' "$B0/${V1}2/DIR/FILE"
+
+#Enable ctime
+TEST $CLI volume set $V1 ctime on
+sleep 3
+TEST stat $M1/DIR/FILE
+
+#Verify presence "trusted.glusterfs.mdata" xattr on backend
+#The lookup above should have created xattr
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT 'trusted.glusterfs.mdata' check_for_xattr 'trusted.glusterfs.mdata' "$B0/${V1}0/DIR"
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT 'trusted.glusterfs.mdata' check_for_xattr 'trusted.glusterfs.mdata' "$B0/${V1}0/DIR/FILE"
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT 'trusted.glusterfs.mdata' check_for_xattr 'trusted.glusterfs.mdata' "$B0/${V1}1/DIR"
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT 'trusted.glusterfs.mdata' check_for_xattr 'trusted.glusterfs.mdata' "$B0/${V1}1/DIR/FILE"
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT 'trusted.glusterfs.mdata' check_for_xattr 'trusted.glusterfs.mdata' "$B0/${V1}2/DIR"
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT 'trusted.glusterfs.mdata' check_for_xattr 'trusted.glusterfs.mdata' "$B0/${V1}2/DIR/FILE"
+
+cleanup;
+###############################################################################
diff --git a/xlators/features/utime/src/utime-messages.h b/xlators/features/utime/src/utime-messages.h
index bac18ab..bd40265 100644
--- a/xlators/features/utime/src/utime-messages.h
+++ b/xlators/features/utime/src/utime-messages.h
@@ -23,6 +23,7 @@
  * glfs-message-id.h.
  */
 
-GLFS_MSGID(UTIME, UTIME_MSG_NO_MEMORY);
+GLFS_MSGID(UTIME, UTIME_MSG_NO_MEMORY, UTIME_MSG_SET_MDATA_FAILED,
+           UTIME_MSG_DICT_SET_FAILED);
 
 #endif /* __UTIME_MESSAGES_H__ */
diff --git a/xlators/features/utime/src/utime.c b/xlators/features/utime/src/utime.c
index 877c751..2a986e7 100644
--- a/xlators/features/utime/src/utime.c
+++ b/xlators/features/utime/src/utime.c
@@ -9,8 +9,10 @@
 */
 
 #include "utime.h"
+#include "utime-helpers.h"
 #include "utime-messages.h"
 #include "utime-mem-types.h"
+#include <glusterfs/call-stub.h>
 
 int32_t
 gf_utime_invalidate(xlator_t *this, inode_t *inode)
@@ -133,6 +135,124 @@ mem_acct_init(xlator_t *this)
 }
 
 int32_t
+gf_utime_set_mdata_setxattr_cbk(call_frame_t *frame, void *cookie,
+                                xlator_t *this, int op_ret, int op_errno,
+                                dict_t *xdata)
+{
+    /* Don't fail lookup if mdata setxattr fails */
+    if (op_ret) {
+        gf_msg(this->name, GF_LOG_ERROR, op_errno, UTIME_MSG_SET_MDATA_FAILED,
+               "dict set of key for set-ctime-mdata failed");
+    }
+    call_resume(frame->local);
+    return 0;
+}
+
+int32_t
+gf_utime_set_mdata_lookup_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+                              int32_t op_ret, int32_t op_errno, inode_t *inode,
+                              struct iatt *stbuf, dict_t *xdata,
+                              struct iatt *postparent)
+{
+    dict_t *dict = NULL;
+    struct mdata_iatt *mdata = NULL;
+    int ret = 0;
+    loc_t loc = {
+        0,
+    };
+
+    if (!op_ret && dict_get(xdata, GF_XATTR_MDATA_KEY) == NULL) {
+        dict = dict_new();
+        if (!dict) {
+            op_errno = ENOMEM;
+            goto err;
+        }
+        mdata = GF_MALLOC(sizeof(struct mdata_iatt), gf_common_mt_char);
+        if (mdata == NULL) {
+            op_errno = ENOMEM;
+            goto err;
+        }
+        iatt_to_mdata(mdata, stbuf);
+        ret = dict_set_mdata(dict, CTIME_MDATA_XDATA_KEY, mdata, _gf_false);
+        if (ret < 0) {
+            gf_msg(this->name, GF_LOG_WARNING, ENOMEM, UTIME_MSG_NO_MEMORY,
+                   "dict set of key for set-ctime-mdata failed");
+            goto err;
+        }
+        frame->local = fop_lookup_cbk_stub(frame, default_lookup_cbk, op_ret,
+                                           op_errno, inode, stbuf, xdata,
+                                           postparent);
+        if (!frame->local) {
+            gf_msg(this->name, GF_LOG_WARNING, ENOMEM, UTIME_MSG_NO_MEMORY,
+                   "lookup_cbk stub allocation failed");
+            goto stub_err;
+        }
+
+        loc.inode = inode_ref(inode);
+        gf_uuid_copy(loc.gfid, stbuf->ia_gfid);
+        STACK_WIND(frame, gf_utime_set_mdata_setxattr_cbk, FIRST_CHILD(this),
+                   FIRST_CHILD(this)->fops->setxattr, &loc, dict, 0, NULL);
+
+        dict_unref(dict);
+        inode_unref(loc.inode);
+        return 0;
+    }
+
+    STACK_UNWIND_STRICT(lookup, frame, op_ret, op_errno, inode, stbuf, xdata,
+                        postparent);
+    return 0;
+
+err:
+    if (mdata) {
+        GF_FREE(mdata);
+    }
+stub_err:
+    if (dict) {
+        dict_unref(dict);
+    }
+    STACK_UNWIND_STRICT(lookup, frame, -1, op_errno, NULL, NULL, NULL, NULL);
+    return 0;
+}
+
+int
+gf_utime_lookup(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+{
+    int op_errno = -1;
+    int ret = -1;
+
+    VALIDATE_OR_GOTO(frame, err);
+    VALIDATE_OR_GOTO(this, err);
+    VALIDATE_OR_GOTO(loc, err);
+    VALIDATE_OR_GOTO(loc->inode, err);
+
+    xdata = xdata ? dict_ref(xdata) : dict_new();
+    if (!xdata) {
+        op_errno = ENOMEM;
+        goto err;
+    }
+
+    ret = dict_set_int8(xdata, GF_XATTR_MDATA_KEY, 1);
+    if (ret < 0) {
+        gf_msg(this->name, GF_LOG_WARNING, -ret, UTIME_MSG_DICT_SET_FAILED,
+               "%s: Unable to set dict value for %s", loc->path,
+               GF_XATTR_MDATA_KEY);
+        op_errno = -ret;
+        goto free_dict;
+    }
+
+    STACK_WIND(frame, gf_utime_set_mdata_lookup_cbk, FIRST_CHILD(this),
+               FIRST_CHILD(this)->fops->lookup, loc, xdata);
+    dict_unref(xdata);
+    return 0;
+
+free_dict:
+    dict_unref(xdata);
+err:
+    STACK_UNWIND_STRICT(lookup, frame, -1, op_errno, NULL, NULL, NULL, NULL);
+    return 0;
+}
+
+int32_t
 init(xlator_t *this)
 {
     utime_priv_t *utime = NULL;
@@ -182,19 +302,27 @@ notify(xlator_t *this, int event, void *data, ...)
 }
 
 struct xlator_fops fops = {
-    /* TODO: Need to go through other fops and
-     *       check if they modify time attributes
-     */
-    .rename = gf_utime_rename,       .mknod = gf_utime_mknod,
-    .readv = gf_utime_readv,         .fremovexattr = gf_utime_fremovexattr,
-    .open = gf_utime_open,           .create = gf_utime_create,
-    .mkdir = gf_utime_mkdir,         .writev = gf_utime_writev,
-    .rmdir = gf_utime_rmdir,         .fallocate = gf_utime_fallocate,
-    .truncate = gf_utime_truncate,   .symlink = gf_utime_symlink,
-    .zerofill = gf_utime_zerofill,   .link = gf_utime_link,
-    .ftruncate = gf_utime_ftruncate, .unlink = gf_utime_unlink,
-    .setattr = gf_utime_setattr,     .fsetattr = gf_utime_fsetattr,
-    .opendir = gf_utime_opendir,     .removexattr = gf_utime_removexattr,
+    .rename = gf_utime_rename,
+    .mknod = gf_utime_mknod,
+    .readv = gf_utime_readv,
+    .fremovexattr = gf_utime_fremovexattr,
+    .open = gf_utime_open,
+    .create = gf_utime_create,
+    .mkdir = gf_utime_mkdir,
+    .writev = gf_utime_writev,
+    .rmdir = gf_utime_rmdir,
+    .fallocate = gf_utime_fallocate,
+    .truncate = gf_utime_truncate,
+    .symlink = gf_utime_symlink,
+    .zerofill = gf_utime_zerofill,
+    .link = gf_utime_link,
+    .ftruncate = gf_utime_ftruncate,
+    .unlink = gf_utime_unlink,
+    .setattr = gf_utime_setattr,
+    .fsetattr = gf_utime_fsetattr,
+    .opendir = gf_utime_opendir,
+    .removexattr = gf_utime_removexattr,
+    .lookup = gf_utime_lookup,
 };
 struct xlator_cbks cbks = {
     .invalidate = gf_utime_invalidate,
diff --git a/xlators/storage/posix/src/posix-inode-fd-ops.c b/xlators/storage/posix/src/posix-inode-fd-ops.c
index ea3b69c..d22bbc2 100644
--- a/xlators/storage/posix/src/posix-inode-fd-ops.c
+++ b/xlators/storage/posix/src/posix-inode-fd-ops.c
@@ -2625,6 +2625,9 @@ posix_setxattr(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
     gf_cs_obj_state state = -1;
     int i = 0;
     int len;
+    struct mdata_iatt mdata_iatt = {
+        0,
+    };
 
     DECLARE_OLD_FS_ID_VAR;
     SET_FS_ID(frame->root->uid, frame->root->gid);
@@ -2638,6 +2641,20 @@ posix_setxattr(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
     priv = this->private;
     DISK_SPACE_CHECK_AND_GOTO(frame, priv, xdata, op_ret, op_errno, out);
 
+    ret = dict_get_mdata(dict, CTIME_MDATA_XDATA_KEY, &mdata_iatt);
+    if (ret == 0) {
+        /* This is initiated by lookup when ctime feature is enabled to create
+         * "trusted.glusterfs.mdata" xattr if not present. These are the files
+         * which were created when ctime feature is disabled.
+         */
+        ret = posix_set_mdata_xattr_legacy_files(this, loc->inode, &mdata_iatt,
+                                                 &op_errno);
+        if (ret != 0) {
+            op_ret = -1;
+        }
+        goto out;
+    }
+
     MAKE_INODE_HANDLE(real_path, this, loc, NULL);
     if (!real_path) {
         op_ret = -1;
diff --git a/xlators/storage/posix/src/posix-messages.h b/xlators/storage/posix/src/posix-messages.h
index 3229275..15e23ff 100644
--- a/xlators/storage/posix/src/posix-messages.h
+++ b/xlators/storage/posix/src/posix-messages.h
@@ -68,6 +68,7 @@ GLFS_MSGID(POSIX, P_MSG_XATTR_FAILED, P_MSG_NULL_GFID, P_MSG_FCNTL_FAILED,
            P_MSG_FALLOCATE_FAILED, P_MSG_STOREMDATA_FAILED,
            P_MSG_FETCHMDATA_FAILED, P_MSG_GETMDATA_FAILED,
            P_MSG_SETMDATA_FAILED, P_MSG_FRESHFILE, P_MSG_MUTEX_FAILED,
-           P_MSG_COPY_FILE_RANGE_FAILED, P_MSG_TIMER_DELETE_FAILED);
+           P_MSG_COPY_FILE_RANGE_FAILED, P_MSG_TIMER_DELETE_FAILED,
+           P_MSG_NOMEM);
 
 #endif /* !_GLUSTERD_MESSAGES_H_ */
diff --git a/xlators/storage/posix/src/posix-metadata.c b/xlators/storage/posix/src/posix-metadata.c
index 5a5e6cd..647c0bb 100644
--- a/xlators/storage/posix/src/posix-metadata.c
+++ b/xlators/storage/posix/src/posix-metadata.c
@@ -245,6 +245,10 @@ __posix_get_mdata_xattr(xlator_t *this, const char *real_path, int _fd,
     if (ret == -1 || !mdata) {
         mdata = GF_CALLOC(1, sizeof(posix_mdata_t), gf_posix_mt_mdata_attr);
         if (!mdata) {
+            gf_msg(this->name, GF_LOG_ERROR, ENOMEM, P_MSG_NOMEM,
+                   "Could not allocate mdata. file: %s: gfid: %s",
+                   real_path ? real_path : "null",
+                   inode ? uuid_utoa(inode->gfid) : "null");
             ret = -1;
             goto out;
         }
@@ -262,18 +266,8 @@ __posix_get_mdata_xattr(xlator_t *this, const char *real_path, int _fd,
             }
         } else {
             /* Failed to get mdata from disk, xattr missing.
-             * This happens on two cases.
-             * 1. File is created before ctime is enabled.
-             * 2. On new file creation.
-             *
-             * Do nothing, just return success. It is as
-             * good as ctime feature is not enabled for this
-             * file. For files created before ctime is enabled,
-             * time attributes gets updated into ctime structure
-             * once the metadata modification fop happens and
-             * time attributes become consistent eventually.
-             * For new files, it would obviously get updated
-             * before the fop completion.
+             * This happens when the file is created before
+             * ctime is enabled.
              */
             if (stbuf && op_errno != ENOENT) {
                 ret = 0;
@@ -345,6 +339,54 @@ posix_compare_timespec(struct timespec *first, struct timespec *second)
         return first->tv_sec - second->tv_sec;
 }
 
+int
+posix_set_mdata_xattr_legacy_files(xlator_t *this, inode_t *inode,
+                                   struct mdata_iatt *mdata_iatt, int *op_errno)
+{
+    posix_mdata_t *mdata = NULL;
+    int ret = 0;
+
+    GF_VALIDATE_OR_GOTO("posix", this, out);
+    GF_VALIDATE_OR_GOTO(this->name, inode, out);
+
+    LOCK(&inode->lock);
+    {
+        mdata = GF_CALLOC(1, sizeof(posix_mdata_t), gf_posix_mt_mdata_attr);
+        if (!mdata) {
+            gf_msg(this->name, GF_LOG_ERROR, ENOMEM, P_MSG_NOMEM,
+                   "Could not allocate mdata. gfid: %s",
+                   uuid_utoa(inode->gfid));
+            ret = -1;
+            *op_errno = ENOMEM;
+            goto unlock;
+        }
+
+        mdata->version = 1;
+        mdata->flags = 0;
+        mdata->ctime.tv_sec = mdata_iatt->ia_ctime;
+        mdata->ctime.tv_nsec = mdata_iatt->ia_ctime_nsec;
+        mdata->atime.tv_sec = mdata_iatt->ia_atime;
+        mdata->atime.tv_nsec = mdata_iatt->ia_atime_nsec;
+        mdata->mtime.tv_sec = mdata_iatt->ia_mtime;
+        mdata->mtime.tv_nsec = mdata_iatt->ia_mtime_nsec;
+
+        __inode_ctx_set1(inode, this, (uint64_t *)&mdata);
+
+        ret = posix_store_mdata_xattr(this, NULL, -1, inode, mdata);
+        if (ret) {
+            gf_msg(this->name, GF_LOG_ERROR, errno, P_MSG_STOREMDATA_FAILED,
+                   "gfid: %s key:%s ", uuid_utoa(inode->gfid),
+                   GF_XATTR_MDATA_KEY);
+            *op_errno = errno;
+            goto unlock;
+        }
+    }
+unlock:
+    UNLOCK(&inode->lock);
+out:
+    return ret;
+}
+
 /* posix_set_mdata_xattr updates the posix_mdata_t based on the flag
  * in inode context and stores it on disk
  */
@@ -372,6 +414,9 @@ posix_set_mdata_xattr(xlator_t *this, const char *real_path, int fd,
              */
             mdata = GF_CALLOC(1, sizeof(posix_mdata_t), gf_posix_mt_mdata_attr);
             if (!mdata) {
+                gf_msg(this->name, GF_LOG_ERROR, ENOMEM, P_MSG_NOMEM,
+                       "Could not allocate mdata. file: %s: gfid: %s",
+                       real_path ? real_path : "null", uuid_utoa(inode->gfid));
                 ret = -1;
                 goto unlock;
             }
@@ -386,35 +431,11 @@ posix_set_mdata_xattr(xlator_t *this, const char *real_path, int fd,
                 __inode_ctx_set1(inode, this, (uint64_t *)&mdata);
             } else {
                 /*
-                 * This is the first time creating the time
-                 * attr. This happens when you activate this
-                 * feature, and the legacy file will not have
-                 * any xattr set.
-                 *
-                 * New files will create extended attributes.
-                 */
-
-                /*
-                 * TODO: This is wrong approach, because before
-                 * creating fresh xattr, we should consult
-                 * to all replica and/or distribution set.
-                 *
-                 * We should contact the time management
-                 * xlators, and ask them to create an xattr.
-                 */
-                /* We should not be relying on backend file's
-                 * time attributes to load the initial ctime
-                 * time attribute structure. This is incorrect
-                 * as each replica set would have witnessed the
-                 * file creation at different times.
-                 *
-                 * For new file creation, ctime, atime and mtime
-                 * should be same, hence initiate the ctime
-                 * structure with the time from the frame. But
-                 * for the files which were created before ctime
-                 * feature is enabled, this is not accurate but
-                 * still fine as the times would get eventually
-                 * accurate.
+                 * This is the first time creating the time attr. This happens
+                 * when you activate this feature. On this code path, only new
+                 * files will create mdata xattr. The legacy files (files
+                 * created before ctime enabled) will not have any xattr set.
+                 * The xattr on legacy file will be set via lookup.
                  */
 
                 /* Don't create xattr with utimes/utimensat, only update if
diff --git a/xlators/storage/posix/src/posix-metadata.h b/xlators/storage/posix/src/posix-metadata.h
index 3416148..dc25e59 100644
--- a/xlators/storage/posix/src/posix-metadata.h
+++ b/xlators/storage/posix/src/posix-metadata.h
@@ -53,5 +53,9 @@ posix_set_ctime_cfr(call_frame_t *frame, xlator_t *this,
                     const char *real_path_in, int fd_in, inode_t *inode_in,
                     struct iatt *stbuf_in, const char *read_path_put,
                     int fd_out, inode_t *inode_out, struct iatt *stbuf_out);
+int
+posix_set_mdata_xattr_legacy_files(xlator_t *this, inode_t *inode,
+                                   struct mdata_iatt *mdata_iatt,
+                                   int *op_errno);
 
 #endif /* _POSIX_METADATA_H */
-- 
1.8.3.1