gfapi/handleops: Introducing glfs_xreaddirplus_r() fop for handleops

Its known that readdirplus operation fetches stat as well for each of the
dirents. But often applications may need extra information, like for eg.,
NFS-Ganesha which operates on handles needs handles for each of those
dirents returned. So this would require extra calls to the backend, in this
case LOOKUP (which is very expensive operation) resulting in very low
readdir performance.

To address that introducing this new API using which applications can
make request for any extra information to be returned as part of
readdirplus response.

Currently this new api returns stat and handles as demanded by application.
The synopsis of the API is noted in glfs.h.

@todo:
* Enhance test script using this new API

Below were the perf results on single brick volume with and without
these changes -

Dataset used -
10*100 directories and each directory containing 100 empty files.

I used NFS-Ganesha application to test these changes -
>for i in {1..5}; do systemctl restart nfs-ganesha; sleep 10; mount -t nfs -o vers=4 localhost:/brick_vol /mnt; cd /mnt; echo "ITERATION$i"; date; find . > tmp-nfs.log; date; cd /; umount /mnt; sleep 2; done;

Without these changes -
ITERATION1
Mon Mar 20 17:22:26 IST 2017
Mon Mar 20 17:23:18 IST 2017
ITERATION2
Mon Mar 20 17:23:39 IST 2017
Mon Mar 20 17:24:28 IST 2017
ITERATION3
Mon Mar 20 17:24:49 IST 2017
Mon Mar 20 17:25:36 IST 2017
ITERATION4
Mon Mar 20 17:30:57 IST 2017
Mon Mar 20 17:31:37 IST 2017
ITERATION5
Mon Mar 20 17:31:57 IST 2017
Mon Mar 20 17:32:40 IST 2017
[root@dhcp35-197 /]#

On an average ~46.2 sec

With these changes applied -
ITERATION1
Mon Mar 20 17:35:03 IST 2017
Mon Mar 20 17:35:15 IST 2017
ITERATION2
Mon Mar 20 17:35:36 IST 2017
Mon Mar 20 17:35:46 IST 2017
ITERATION3
Mon Mar 20 17:36:06 IST 2017
Mon Mar 20 17:36:17 IST 2017
ITERATION4
Mon Mar 20 17:41:38 IST 2017
Mon Mar 20 17:41:49 IST 2017
ITERATION5
Mon Mar 20 17:42:10 IST 2017
Mon Mar 20 17:42:20 IST 2017

On an average ~10.8 sec

Updates #174
BUG: 1442950
Change-Id: I0f74f74dc62085ca4c4a23c38e3edc84bd850876
Signed-off-by: Soumya Koduri <skoduri@redhat.com>
Reviewed-on: https://review.gluster.org/15663
Smoke: Gluster Build System <jenkins@build.gluster.org>
NetBSD-regression: NetBSD Build System <jenkins@build.gluster.org>
Reviewed-by: Niels de Vos <ndevos@redhat.com>
CentOS-regression: Gluster Build System <jenkins@build.gluster.org>

10 files changed, 303 insertions, 4 deletions

diff --git a/api/src/gfapi-messages.h b/api/src/gfapi-messages.h
index b4ecefbeb56..6fcbd9d392e 100644
--- a/api/src/gfapi-messages.h
+++ b/api/src/gfapi-messages.h
@@ -40,7 +40,7 @@
  */
 
 #define GLFS_GFAPI_BASE             GLFS_MSGID_COMP_API
-#define GLFS_NUM_MESSAGES           49
+#define GLFS_NUM_MESSAGES           50
 #define GLFS_MSGID_END              (GLFS_GFAPI_BASE + GLFS_NUM_MESSAGES + 1)
 /* Messages with message IDs */
 #define glfs_msg_start_x GLFS_GFAPI_BASE, "Invalid: Start of messages"
@@ -95,6 +95,7 @@
 #define API_MSG_CREATE_HANDLE_FAILED            (GLFS_GFAPI_BASE + 47)
 #define API_MSG_INODE_LINK_FAILED               (GLFS_GFAPI_BASE + 48)
 #define API_MSG_STATEDUMP_FAILED                (GLFS_GFAPI_BASE + 49)
+#define API_MSG_XREADDIRP_R_FAILED              (GLFS_GFAPI_BASE + 50)
 
 /*------------*/
 #define glfs_msg_end_x GLFS_MSGID_END, "Invalid: End of messages"
diff --git a/api/src/gfapi.aliases b/api/src/gfapi.aliases
index fed2fd82b1e..f3726a33484 100644
--- a/api/src/gfapi.aliases
+++ b/api/src/gfapi.aliases
@@ -158,4 +158,9 @@ _pub_glfs_realpath _glfs_realpath$GFAPI_3.7.17
 
 _pub_glfs_sysrq _glfs_sysrq$GFAPI_3.10.0
 
+_pub_glfs_xreaddirplus_r _glfs_xreaddirplus_r$GFAPI_3.11.0
+_pub_glfs_xreaddirplus_r_get_stat _glfs_xreaddirplus_r_get_stat$GFAPI_3.11.0
+_pub_glfs_xreaddirplus_r_get_object _glfs_xreaddirplus_r_get_object$GFAPI_3.11.0
+_pub_glfs_object_copy _glfs_object_copy$GFAPI_3.11.0
+
 _pub_glfs_ipc _glfs_ipc$GFAPI_4.0.0
diff --git a/api/src/gfapi.map b/api/src/gfapi.map
index 3f7d2bb56a6..5e0c877077d 100644
--- a/api/src/gfapi.map
+++ b/api/src/gfapi.map
@@ -199,7 +199,14 @@ GFAPI_3.10.0 {
 		glfs_sysrq;
 } GFAPI_3.7.17;
 
+GFAPI_3.11.0 {
+		glfs_xreaddirplus_r;
+		glfs_xreaddirplus_r_get_stat;
+		glfs_xreaddirplus_r_get_object;
+                glfs_object_copy;
+} GFAPI_3.10.0;
+
 GFAPI_4.0.0 {
 	global:
 		glfs_ipc;
-} GFAPI_3.10.0;
+} GFAPI_3.11.0;
diff --git a/api/src/glfs-fops.c b/api/src/glfs-fops.c
index 0e5b206b98c..1bb958fc1b7 100644
--- a/api/src/glfs-fops.c
+++ b/api/src/glfs-fops.c
@@ -2792,7 +2792,7 @@ glfd_entry_next (struct glfs_fd *glfd, int plus)
 }
 
 
-static struct dirent *
+struct dirent *
 glfs_readdirbuf_get (struct glfs_fd *glfd)
 {
         struct dirent *buf = NULL;
@@ -4608,3 +4608,138 @@ out:
 invalid_fs:
         return ret;
 }
+
+/*
+ * Given glfd of a directory, this function does readdirp and returns
+ * xstat along with dirents.
+ */
+int
+pub_glfs_xreaddirplus_r (struct glfs_fd *glfd, uint32_t flags,
+                         struct glfs_xreaddirp_stat **xstat_p,
+                         struct dirent *ext,
+                         struct dirent **res)
+{
+        int              ret = -1;
+        gf_dirent_t     *entry = NULL;
+        struct dirent   *buf = NULL;
+        struct glfs_xreaddirp_stat *xstat = NULL;
+
+        DECLARE_OLD_THIS;
+        __GLFS_ENTRY_VALIDATE_FD (glfd, invalid_fs);
+
+        GF_REF_GET (glfd);
+
+        GF_VALIDATE_OR_GOTO (THIS->name, xstat_p, out);
+        GF_VALIDATE_OR_GOTO (THIS->name, res, out);
+
+        errno = 0;
+
+        if (ext)
+                buf = ext;
+        else
+                buf = glfs_readdirbuf_get (glfd);
+
+        if (!buf)
+                goto out;
+
+        xstat = GF_CALLOC(1, sizeof(struct glfs_xreaddirp_stat),
+                          glfs_mt_xreaddirp_stat_t);
+
+        if (!xstat)
+                goto out;
+
+        /* this is readdirplus operation */
+        entry = glfd_entry_next (glfd, 1);
+
+        /* XXX: Ideally when we reach EOD, errno should have been
+         * set to ENOENT. But that doesn't seem to be the case.
+         *
+         * The only way to confirm if its EOD at this point is that
+         * errno == 0 and entry == NULL
+         */
+        if (errno)
+                goto out;
+
+        if (!entry) {
+                /* reached EOD, ret = 0  */
+                ret = 0;
+                *res = NULL;
+                goto out;
+        }
+
+        *res = buf;
+        gf_dirent_to_dirent (entry, buf);
+
+        if (flags & GFAPI_XREADDIRP_STAT) {
+                glfs_iatt_to_stat (glfd->fs, &entry->d_stat, &xstat->st);
+                xstat->flags_handled |= GFAPI_XREADDIRP_STAT;
+        }
+
+        if ((flags & GFAPI_XREADDIRP_HANDLE) &&
+                /* skip . and .. */
+                strcmp(buf->d_name, ".")
+                     && strcmp(buf->d_name, "..")) {
+
+                /* Now create object.
+                 * We can use "glfs_h_find_handle" as well as inodes would have
+                 * already got linked as part of 'gf_link_inodes_from_dirent' */
+                xstat->object = glfs_h_create_from_handle (glfd->fs,
+                                                        entry->d_stat.ia_gfid,
+                                                        GFAPI_HANDLE_LENGTH,
+                                                        NULL);
+
+                if (xstat->object) { /* success */
+                        /* note: xstat->object->inode->ref is taken
+                         * This shall be unref'ed when application does
+                         * glfs_free(xstat) */
+                        xstat->flags_handled |= GFAPI_XREADDIRP_HANDLE;
+                }
+        }
+
+        ret = xstat->flags_handled;
+        *xstat_p = xstat;
+
+out:
+        gf_msg_debug (THIS->name, 0,
+                      "xreaddirp- requested_flags (%x) , processed_flags (%x)",
+                      flags, xstat->flags_handled);
+
+        GF_REF_PUT (glfd);
+
+        if (ret < 0) {
+                gf_msg (THIS->name, GF_LOG_WARNING, errno,
+                        API_MSG_XREADDIRP_R_FAILED,
+                        "glfs_x_readdirp_r failed - reason (%s)",
+                        strerror(errno));
+
+                if (xstat)
+                        glfs_free (xstat);
+        }
+
+        __GLFS_EXIT_FS;
+
+        return ret;
+
+invalid_fs:
+        return -1;
+}
+GFAPI_SYMVER_PUBLIC_DEFAULT(glfs_xreaddirplus_r, 3.11.0);
+
+struct stat*
+pub_glfs_xreaddirplus_get_stat (struct glfs_xreaddirp_stat *xstat)
+{
+        GF_VALIDATE_OR_GOTO ("glfs_xreaddirplus_get_stat", xstat, out);
+
+        if (!xstat->flags_handled & GFAPI_XREADDIRP_STAT)
+                gf_msg (THIS->name, GF_LOG_ERROR, errno,
+                        LG_MSG_INVALID_ARG,
+                        "GFAPI_XREADDIRP_STAT is not set. Flags"
+                        "handled for xstat(%p) are (%x)",
+                        xstat, xstat->flags_handled);
+        return &xstat->st;
+
+out:
+        return NULL;
+}
+GFAPI_SYMVER_PUBLIC_DEFAULT(glfs_xreaddirplus_get_stat, 3.11.0);
+
diff --git a/api/src/glfs-handleops.c b/api/src/glfs-handleops.c
index 69c542e5886..de0a6deb87f 100644
--- a/api/src/glfs-handleops.c
+++ b/api/src/glfs-handleops.c
@@ -2379,3 +2379,48 @@ pub_glfs_h_anonymous_write (struct glfs *fs, struct glfs_object *object,
 }
 
 GFAPI_SYMVER_PUBLIC_DEFAULT(glfs_h_anonymous_write, 3.7.0);
+
+struct glfs_object*
+pub_glfs_object_copy (struct glfs_object *src)
+{
+        struct glfs_object *object = NULL;
+
+        GF_VALIDATE_OR_GOTO ("glfs_dup_object", src, out);
+
+        object = GF_CALLOC (1, sizeof(struct glfs_object),
+                            glfs_mt_glfs_object_t);
+        if (object == NULL) {
+                errno = ENOMEM;
+                gf_msg (THIS->name, GF_LOG_WARNING, errno,
+                         API_MSG_CREATE_HANDLE_FAILED,
+                        "glfs_dup_object for gfid-%s failed",
+                        uuid_utoa (src->inode->gfid));
+                return NULL;
+        }
+
+        object->inode = inode_ref (src->inode);
+        gf_uuid_copy (object->gfid, src->inode->gfid);
+
+out:
+        return object;
+}
+GFAPI_SYMVER_PUBLIC_DEFAULT(glfs_object_copy, 3.11.0);
+
+struct glfs_object*
+pub_glfs_xreaddirplus_get_object (struct glfs_xreaddirp_stat *xstat)
+{
+        GF_VALIDATE_OR_GOTO ("glfs_xreaddirplus_get_object", xstat, out);
+
+        if (!(xstat->flags_handled & GFAPI_XREADDIRP_HANDLE))
+                gf_msg (THIS->name, GF_LOG_ERROR, errno,
+                        LG_MSG_INVALID_ARG,
+                        "GFAPI_XREADDIRP_HANDLE is not set. Flags"
+                        "handled for xstat(%p) are (%x)",
+                        xstat, xstat->flags_handled);
+
+        return xstat->object;
+
+out:
+        return NULL;
+}
+GFAPI_SYMVER_PUBLIC_DEFAULT(glfs_xreaddirplus_get_object, 3.11.0);
diff --git a/api/src/glfs-handles.h b/api/src/glfs-handles.h
index ef994aaf88c..9583fab069a 100644
--- a/api/src/glfs-handles.h
+++ b/api/src/glfs-handles.h
@@ -355,6 +355,22 @@ glfs_h_anonymous_read (struct glfs *fs, struct glfs_object *object,
                       const void *buf, size_t count, off_t offset) __THROW
         GFAPI_PUBLIC(glfs_h_anonymous_read, 3.7.0);
 
+/*
+ * Caution: The object returned by this object gets freed as part
+ * of 'glfs_free(xstat)'. Make sure to have a copy using 'glfs_object_copy()'
+ * to use post that.
+ */
+struct glfs_object*
+glfs_xreaddirplus_get_object (struct glfs_xreaddirp_stat *xstat) __THROW
+        GFAPI_PUBLIC(glfs_xreaddirplus_get_object, 3.11.0);
+
+/* Applications should close the object returned by this routine
+ * explicitly using 'glfs_h_close()'
+ */
+struct glfs_object*
+glfs_object_copy (struct glfs_object *src);
+        GFAPI_PUBLIC(glfs_object_copy, 3.11.0);
+
 __END_DECLS
 
 #endif /* !_GLFS_HANDLES_H */
diff --git a/api/src/glfs-internal.h b/api/src/glfs-internal.h
index a42822420ea..6b964b57304 100644
--- a/api/src/glfs-internal.h
+++ b/api/src/glfs-internal.h
@@ -245,6 +245,12 @@ struct glfs_upcall_inode {
         struct stat           oldp_buf; /* Latest stat of old parent dir handle */
 };
 
+struct glfs_xreaddirp_stat {
+        struct stat st; /* Stat for that dirent - corresponds to GFAPI_XREADDIRP_STAT */
+        struct glfs_object *object; /* handled for GFAPI_XREADDIRP_HANDLE */
+        uint32_t flags_handled; /* final set of flags successfulyy handled */
+};
+
 #define DEFAULT_EVENT_POOL_SIZE           16384
 #define GF_MEMPOOL_COUNT_OF_DICT_T        4096
 #define GF_MEMPOOL_COUNT_OF_DATA_T        (GF_MEMPOOL_COUNT_OF_DICT_T * 4)
@@ -445,7 +451,6 @@ glfs_anonymous_pwritev (struct glfs *fs, struct glfs_object *object,
 struct glfs_object *
 glfs_h_resolve_symlink (struct glfs *fs, struct glfs_object *object);
 
-
 /* Deprecated structures that were passed to client applications, replaced by
  * accessor functions. Do not use these in new applications, and update older
  * usage.
@@ -475,5 +480,13 @@ struct glfs_callback_inode_arg {
         struct stat             oldp_buf; /* Latest stat of old parent
                                            * dir handle */
 };
+struct dirent *
+glfs_readdirbuf_get (struct glfs_fd *glfd);
+
+gf_dirent_t *
+glfd_entry_next (struct glfs_fd *glfd, int plus);
+
+void
+gf_dirent_to_dirent (gf_dirent_t *gf_dirent, struct dirent *dirent);
 
 #endif /* !_GLFS_INTERNAL_H */
diff --git a/api/src/glfs-mem-types.h b/api/src/glfs-mem-types.h
index 52033360853..4179138e65e 100644
--- a/api/src/glfs-mem-types.h
+++ b/api/src/glfs-mem-types.h
@@ -29,6 +29,7 @@ enum glfs_mem_types_ {
 	glfs_mt_acl_t,
         glfs_mt_upcall_inode_t,
         glfs_mt_realpath_t,
+        glfs_mt_xreaddirp_stat_t,
 	glfs_mt_end
 };
 #endif
diff --git a/api/src/glfs.c b/api/src/glfs.c
index 10af6c78e17..d9f07603b0f 100644
--- a/api/src/glfs.c
+++ b/api/src/glfs.c
@@ -1375,6 +1375,16 @@ pub_glfs_free (void *ptr)
                 GF_FREE (ptr);
                 break;
         }
+        case glfs_mt_xreaddirp_stat_t:
+        {
+                struct glfs_xreaddirp_stat *to_free = ptr;
+
+                if (to_free->object)
+                        glfs_h_close (to_free->object);
+
+                GF_FREE (ptr);
+                break;
+        }
         default:
                 GF_FREE (ptr);
         }
diff --git a/api/src/glfs.h b/api/src/glfs.h
index 68fcf4ddcd9..4e535bda168 100644
--- a/api/src/glfs.h
+++ b/api/src/glfs.h
@@ -41,6 +41,7 @@
 #include <sys/cdefs.h>
 #include <dirent.h>
 #include <sys/statvfs.h>
+#include <inttypes.h>
 
 #if defined(HAVE_SYS_ACL_H) || (defined(USE_POSIX_ACLS) && USE_POSIX_ACLS)
 #include <sys/acl.h>
@@ -790,6 +791,71 @@ int glfs_sysrq (glfs_t *fs, char sysrq) __THROW
 
 
 /*
+ * Structure returned as part of xreaddirplus
+ */
+struct glfs_xreaddirp_stat;
+
+/* Request flags to be used in XREADDIRP operation */
+#define GFAPI_XREADDIRP_NULL    0x00000000 /* by default, no stat will be fetched */
+#define GFAPI_XREADDIRP_STAT    0x00000001 /* Get stat */
+#define GFAPI_XREADDIRP_HANDLE  0x00000002 /* Get object handle */
+
+/*
+ * This stat structure returned gets freed as part of glfs_free(xstat)
+ */
+struct stat*
+glfs_xreaddirplus_get_stat (struct glfs_xreaddirp_stat *xstat) __THROW
+        GFAPI_PUBLIC(glfs_xreaddirplus_get_stat, 3.11.0);
+
+/*
+ * SYNOPSIS
+ *
+ * glfs_xreaddirplus_r: Extended Readirplus operation
+ *
+ * DESCRIPTION
+ *
+ * This API does readdirplus operation, but along with stat it can fetch other
+ * extra information like object handles etc for each of the dirents returned
+ * based on requested flags. On success it returns the set of flags successfully
+ * processed.
+ *
+ * Note that there are chances that some of the requested information may not be
+ * available or returned (for example if reached EOD). Ensure to validate the
+ * returned value to determine what flags have been successfully processed
+ * & set.
+ *
+ * PARAMETERS
+ *
+ * INPUT:
+ * @glfd: GFAPI file descriptor of the directory
+ * @flags: Flags determining xreaddirp_stat requested
+ *         Current available values are:
+ *              GFAPI_XREADDIRP_NULL
+ *              GFAPI_XREADDIRP_STAT
+ *              GFAPI_XREADDIRP_HANDLE
+ * @ext: Dirent struture to copy the values to
+ *       (though optional recommended to be allocated by application
+ *        esp., in multi-threaded environement)
+ *
+ * OUTPUT:
+ * @res: to store the next dirent value. If NULL and return value is '0',
+ *       it means it reached end of the directory.
+ * @xstat_p: Pointer to contain all the requested data returned
+ *           for that dirent. Application should make use of glfs_free() API
+ *           to free this pointer and the variables returned by
+ *           glfs_xreaddirplus_get_*() APIs.
+ *
+ * RETURN VALUE:
+ * >=0: SUCCESS (value contains the flags successfully processed)
+ *  -1: FAILURE
+ */
+int
+glfs_xreaddirplus_r (struct glfs_fd *glfd, uint32_t flags,
+                     struct glfs_xreaddirp_stat **xstat_p,
+                     struct dirent *ext, struct dirent **res);
+        GFAPI_PUBLIC(glfs_xreaddirplus_r, 3.11.0);
+
+/*
  * Nobody needs this call at all yet except for the test script.
  */
 int glfs_ipc (glfs_fd_t *fd, int cmd,  void *xd_in, void **xd_out) __THROW