features/shard: Implement zerofill FOP

Change-Id: Iabb89a502cc560edd7fe3755623c74a2ccb853d5
BUG: 1261841
Signed-off-by: Krutika Dhananjay <kdhananj@redhat.com>
Reviewed-on: http://review.gluster.org/13234
Smoke: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Pranith Kumar Karampuri <pkarampu@redhat.com>
Tested-by: Pranith Kumar Karampuri <pkarampu@redhat.com>
NetBSD-regression: NetBSD Build System <jenkins@build.gluster.org>
CentOS-regression: Gluster Build System <jenkins@build.gluster.com>

4 files changed, 164 insertions, 26 deletions

diff --git a/tests/bugs/shard/bug-shard-zerofill.c b/tests/bugs/shard/bug-shard-zerofill.c
new file mode 100644
index 00000000000..2ec411921cf
--- /dev/null
+++ b/tests/bugs/shard/bug-shard-zerofill.c
@@ -0,0 +1,59 @@
+#include <stdio.h>
+#include "glfs.h"
+#include "glfs-handles.h"
+
+int
+main (int argc, char *argv[])
+{
+        glfs_t    *fs = NULL;
+        glfs_fd_t *fd = NULL;
+        int        ret = 1;
+
+        if (argc != 4) {
+                fprintf (stderr, "Syntax: %s <host> <volname> <file-path> \n", argv[0]);
+                return 1;
+        }
+
+        fs = glfs_new (argv[2]);
+        if (!fs) {
+                fprintf (stderr, "glfs_new: returned NULL\n");
+                return 1;
+        }
+
+        ret = glfs_set_volfile_server (fs, "tcp", argv[1], 24007);
+        if (ret != 0) {
+                fprintf (stderr, "glfs_set_volfile_server: retuned %d\n", ret);
+                goto out;
+        }
+        ret = glfs_set_logging (fs, "/var/log/glusterfs/glfs-dis.log", 7);
+        if (ret != 0) {
+                fprintf (stderr, "glfs_set_logging: returned %d\n", ret);
+                goto out;
+        }
+        ret = glfs_init (fs);
+        if (ret != 0) {
+                fprintf (stderr, "glfs_init: returned %d\n", ret);
+                goto out;
+        }
+
+        fd = glfs_open (fs, argv[3], O_RDWR);
+        if (fd == NULL) {
+                fprintf (stderr, "glfs_open: returned NULL\n");
+                goto out;
+        }
+
+        /* Zero-fill "foo" with 10MB of data */
+        ret = glfs_zerofill (fd, 0, 10485760);
+        if (ret <= 0) {
+                fprintf (stderr, "glfs_zerofill: returned %d\n", ret);
+                goto out;
+        }
+
+        glfs_close(fd);
+
+        ret = 0;
+
+out:
+        glfs_fini (fs);
+        return ret;
+}
diff --git a/tests/bugs/shard/bug-shard-zerofill.t b/tests/bugs/shard/bug-shard-zerofill.t
new file mode 100644
index 00000000000..d5113001b69
--- /dev/null
+++ b/tests/bugs/shard/bug-shard-zerofill.t
@@ -0,0 +1,50 @@
+#!/bin/bash
+
+. $(dirname $0)/../../include.rc
+. $(dirname $0)/../../volume.rc
+
+function file_all_zeroes {
+        < $1 tr -d '\0' | read -n 1 || echo 1
+}
+
+cleanup
+
+TEST glusterd
+TEST pidof glusterd
+TEST $CLI volume create $V0 replica 2 $H0:$B0/${V0}{0..3}
+TEST $CLI volume set $V0 features.shard on
+TEST $CLI volume start $V0
+
+TEST $GFS --volfile-id=$V0 --volfile-server=$H0 $M0
+
+# Create a file.
+TEST touch $M0/foo
+
+gfid_foo=`getfattr -n glusterfs.gfid.string $M0/foo 2>/dev/null \
+          | grep glusterfs.gfid.string | cut -d '"' -f 2`
+
+TEST gcc -Wall -O2 -I api/src -o $(dirname $0)/bug-shard-zerofill $(dirname $0)/bug-shard-zerofill.c -lgfapi
+TEST $(dirname $0)/bug-shard-zerofill $H0 $V0 /foo
+
+# This should ensure /.shard is created on the bricks.
+TEST stat $B0/${V0}0/.shard
+TEST stat $B0/${V0}1/.shard
+TEST stat $B0/${V0}2/.shard
+TEST stat $B0/${V0}3/.shard
+
+EXPECT "4194304" echo `find $B0 -name $gfid_foo.1 | xargs stat -c %s`
+EXPECT "2097152" echo `find $B0 -name $gfid_foo.2 | xargs stat -c %s`
+
+EXPECT "1" file_all_zeroes $M0/foo
+
+TEST `echo "abc" >> $M0/foo`
+
+EXPECT_NOT "1" file_all_zeroes $M0/foo
+
+TEST umount $M0
+TEST $CLI volume stop $V0
+TEST $CLI volume delete $V0
+
+TEST rm -f $(dirname $0)/bug-shard-zerofill
+
+cleanup
diff --git a/xlators/features/shard/src/shard.c b/xlators/features/shard/src/shard.c
index 89a47ef2ce2..e46d4dc33ec 100644
--- a/xlators/features/shard/src/shard.c
+++ b/xlators/features/shard/src/shard.c
@@ -498,6 +498,10 @@ shard_common_inode_write_failure_unwind (glusterfs_fop_t fop,
                 SHARD_STACK_UNWIND (fallocate, frame, op_ret, op_errno,
                                     NULL, NULL, NULL);
                 break;
+        case GF_FOP_ZEROFILL:
+                SHARD_STACK_UNWIND (zerofill, frame, op_ret, op_errno,
+                                    NULL, NULL, NULL);
+                break;
         default:
                 gf_msg (THIS->name, GF_LOG_WARNING, 0, SHARD_MSG_INVALID_FOP,
                         "Invalid fop id = %d", fop);
@@ -523,6 +527,10 @@ shard_common_inode_write_success_unwind (glusterfs_fop_t fop,
                 SHARD_STACK_UNWIND (fallocate, frame, op_ret, 0, &local->prebuf,
                                     &local->postbuf, local->xattr_rsp);
                 break;
+        case GF_FOP_ZEROFILL:
+                SHARD_STACK_UNWIND (zerofill, frame, op_ret, 0, &local->prebuf,
+                                    &local->postbuf, local->xattr_rsp);
+                break;
         default:
                 gf_msg (THIS->name, GF_LOG_WARNING, 0, SHARD_MSG_INVALID_FOP,
                         "Invalid fop id = %d", fop);
@@ -3520,7 +3528,7 @@ shard_common_inode_write_do_cbk (call_frame_t *frame, void *cookie,
 
 int
 shard_common_inode_write_wind (call_frame_t *frame, xlator_t *this,
-                               fd_t *anon_fd, struct iovec *vec, int count,
+                               fd_t *fd, struct iovec *vec, int count,
                                off_t shard_offset, size_t size)
 {
         shard_local_t *local = NULL;
@@ -3529,19 +3537,25 @@ shard_common_inode_write_wind (call_frame_t *frame, xlator_t *this,
 
         switch (local->fop) {
         case GF_FOP_WRITE:
-                STACK_WIND_COOKIE (frame, shard_common_inode_write_do_cbk,
-                                   anon_fd, FIRST_CHILD(this),
-                                   FIRST_CHILD(this)->fops->writev, anon_fd,
-                                   vec, count, shard_offset, local->flags,
+                STACK_WIND_COOKIE (frame, shard_common_inode_write_do_cbk, fd,
+                                   FIRST_CHILD(this),
+                                   FIRST_CHILD(this)->fops->writev, fd, vec,
+                                   count, shard_offset, local->flags,
                                    local->iobref, local->xattr_req);
                 break;
         case GF_FOP_FALLOCATE:
-                STACK_WIND_COOKIE (frame, shard_common_inode_write_do_cbk,
-                                   anon_fd, FIRST_CHILD(this),
-                                   FIRST_CHILD(this)->fops->fallocate, anon_fd,
+                STACK_WIND_COOKIE (frame, shard_common_inode_write_do_cbk, fd,
+                                   FIRST_CHILD(this),
+                                   FIRST_CHILD(this)->fops->fallocate, fd,
                                    local->flags, shard_offset, size,
                                    local->xattr_req);
                 break;
+        case GF_FOP_ZEROFILL:
+                STACK_WIND_COOKIE (frame, shard_common_inode_write_do_cbk, fd,
+                                   FIRST_CHILD(this),
+                                   FIRST_CHILD(this)->fops->zerofill, fd,
+                                   shard_offset, size, local->xattr_req);
+                break;
         default:
                 gf_msg (this->name, GF_LOG_WARNING, 0, SHARD_MSG_INVALID_FOP,
                         "Invalid fop id = %d", local->fop);
@@ -4479,6 +4493,11 @@ shard_common_inode_write_begin (call_frame_t *frame, xlator_t *this,
                                          FIRST_CHILD(this)->fops->fallocate, fd,
                                          flags, offset, len, xdata);
                         break;
+                case GF_FOP_ZEROFILL:
+                        STACK_WIND_TAIL (frame, FIRST_CHILD(this),
+                                         FIRST_CHILD(this)->fops->zerofill,
+                                         fd, offset, len, xdata);
+                        break;
                 default:
                 gf_msg (this->name, GF_LOG_WARNING, 0, SHARD_MSG_INVALID_FOP,
                         "Invalid fop id = %d", fop);
@@ -4573,24 +4592,22 @@ shard_fallocate (call_frame_t *frame, xlator_t *this, fd_t *fd,
 }
 
 int
-shard_discard (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
-              size_t len, dict_t *xdata)
+shard_zerofill (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+                off_t len, dict_t *xdata)
 {
-        /* TBD */
-        gf_msg (this->name, GF_LOG_INFO, ENOTSUP, SHARD_MSG_FOP_NOT_SUPPORTED,
-                "discard called on %s.", uuid_utoa (fd->inode->gfid));
-        SHARD_STACK_UNWIND (discard, frame, -1, ENOTSUP, NULL, NULL, NULL);
+        shard_common_inode_write_begin (frame, this, GF_FOP_ZEROFILL, fd, NULL,
+                                        0, offset, 0, len, NULL, xdata);
         return 0;
 }
 
 int
-shard_zerofill (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
-                off_t len, dict_t *xdata)
+shard_discard (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+              size_t len, dict_t *xdata)
 {
         /* TBD */
         gf_msg (this->name, GF_LOG_INFO, ENOTSUP, SHARD_MSG_FOP_NOT_SUPPORTED,
-                "zerofill called on %s.", uuid_utoa (fd->inode->gfid));
-        SHARD_STACK_UNWIND (zerofill, frame, -1, ENOTSUP, NULL, NULL, NULL);
+                "discard called on %s.", uuid_utoa (fd->inode->gfid));
+        SHARD_STACK_UNWIND (discard, frame, -1, ENOTSUP, NULL, NULL, NULL);
         return 0;
 }
 
diff --git a/xlators/storage/posix/src/posix.c b/xlators/storage/posix/src/posix.c
index e834fc09752..b0c23716a64 100644
--- a/xlators/storage/posix/src/posix.c
+++ b/xlators/storage/posix/src/posix.c
@@ -809,12 +809,13 @@ err:
 }
 
 static int32_t
-posix_do_zerofill(call_frame_t *frame, xlator_t *this, fd_t *fd,
-                  off_t offset, off_t len, struct iatt *statpre,
-                  struct iatt *statpost)
+posix_do_zerofill (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+                   off_t len, struct iatt *statpre, struct iatt *statpost,
+                   dict_t *xdata)
 {
-        struct posix_fd *pfd       = NULL;
-        int32_t          ret       = -1;
+        int32_t            ret       = -1;
+        struct posix_fd   *pfd       = NULL;
+        gf_boolean_t       locked    = _gf_false;
 
         DECLARE_OLD_FS_ID_VAR;
 
@@ -830,6 +831,11 @@ posix_do_zerofill(call_frame_t *frame, xlator_t *this, fd_t *fd,
                 goto out;
         }
 
+        if (dict_get (xdata, GLUSTERFS_WRITE_UPDATE_ATOMIC)) {
+                locked = _gf_true;
+                LOCK(&fd->inode->lock);
+        }
+
         ret = posix_fdstat (this, pfd->fd, statpre);
         if (ret == -1) {
                 ret = -errno;
@@ -837,7 +843,8 @@ posix_do_zerofill(call_frame_t *frame, xlator_t *this, fd_t *fd,
                         "pre-operation fstat failed on fd = %p", fd);
                 goto out;
         }
-        ret = _posix_do_zerofill(pfd->fd, offset, len, pfd->flags & O_DIRECT);
+
+        ret = _posix_do_zerofill (pfd->fd, offset, len, pfd->flags & O_DIRECT);
         if (ret < 0) {
                 ret = -errno;
                 gf_msg (this->name, GF_LOG_ERROR, errno, P_MSG_ZEROFILL_FAILED,
@@ -845,6 +852,7 @@ posix_do_zerofill(call_frame_t *frame, xlator_t *this, fd_t *fd,
                         pfd->fd, len);
                 goto out;
         }
+
         if (pfd->flags & (O_SYNC|O_DSYNC)) {
                 ret = sys_fsync (pfd->fd);
                 if (ret) {
@@ -865,6 +873,10 @@ posix_do_zerofill(call_frame_t *frame, xlator_t *this, fd_t *fd,
         }
 
 out:
+	if (locked) {
+		UNLOCK (&fd->inode->lock);
+		locked = _gf_false;
+	}
         SET_TO_OLD_FS_ID ();
 
         return ret;
@@ -932,8 +944,8 @@ posix_zerofill(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
         struct  iatt statpre             = {0,};
         struct  iatt statpost            = {0,};
 
-        ret = posix_do_zerofill(frame, this, fd, offset, len,
-                                 &statpre, &statpost);
+        ret = posix_do_zerofill (frame, this, fd, offset, len,
+                                 &statpre, &statpost, xdata);
         if (ret < 0)
                 goto err;