Introduce fallocate support

fallocate() allows for reserving disk space for a particular inode and
fd. Hence, a client can be sure that he won't see a ENOSPC (eventually
a 507 HTTP response) later during writes. Swift's object server has
had fallocate() support from a long time.

P.S: Older versions of glusterfs (<3.6) did not support fallocate
because FUSE did not support it earlier.
http://review.gluster.org/4969
http://fuse.996288.n3.nabble.com/fallocate-support-in-FUSE-td10668.html

Change-Id: Ida4b16357901707d624f92bf1b2dc8f07da4f1ad
Signed-off-by: Prashanth Pai <ppai@redhat.com>

etc/object-server.conf-swiftonfile

@@ -20,8 +20,13 @@ max_clients = 1024
 # If not doing the above, setting this value initially to match the number of
 # CPUs is a good starting point for determining the right value.
 workers = 1
-# Override swift's default behaviour for fallocate.
+# You can set disable_fallocate to true to turn off usage of fallocate()
-disable_fallocate = true
+# sys call to reserve space during a PUT operation.
+# disable_fallocate = false
+#
+# You can set fallocate_reserve to the number of bytes you'd like fallocate to
+# reserve, whether there is space for the given file size or not.
+# fallocate_reserve = 0
 
 [pipeline:main]
 pipeline = object-server

swiftonfile/swift/obj/diskfile.py

@@ -28,7 +28,8 @@ from eventlet import sleep
 from contextlib import contextmanager
 from swiftonfile.swift.common.exceptions import AlreadyExistsAsFile, \
     AlreadyExistsAsDir
-from swift.common.utils import ThreadPool, hash_path, normalize_timestamp
+from swift.common.utils import ThreadPool, hash_path, \
+    normalize_timestamp, fallocate
 from swift.common.exceptions import DiskFileNotExist, DiskFileError, \
     DiskFileNoSpace, DiskFileDeviceUnavailable, DiskFileNotOpen, \
     DiskFileExpired
@@ -931,14 +932,19 @@ class DiskFile(object):
                 break
         dw = None
         try:
+            if size is not None and size > 0:
+                try:
+                    fallocate(fd, size)
+                except OSError as err:
+                    if err.errno in (errno.ENOSPC, errno.EDQUOT):
+                        raise DiskFileNoSpace()
+                    raise
             # Ensure it is properly owned before we make it available.
             do_fchown(fd, self._uid, self._gid)
-            # NOTE: we do not perform the fallocate() call at all. We ignore
-            # it completely since at the time of this writing FUSE does not
-            # support it.
             dw = DiskFileWriter(fd, tmppath, self)
             yield dw
         finally:
+            if dw:
                 dw.close()
                 if dw._tmppath:
                     do_unlink(dw._tmppath)

test/unit/obj/test_diskfile.py

@@ -89,6 +89,10 @@ def _mock_do_fsync(fd):
     return
 
 
+def _mock_fallocate(fd, size):
+    return
+
+
 class MockRenamerCalled(Exception):
     pass
 
@@ -134,6 +138,8 @@ class TestDiskFile(unittest.TestCase):
         swiftonfile.swift.common.utils.read_metadata = _mock_read_metadata
         self._saved_do_fsync = swiftonfile.swift.obj.diskfile.do_fsync
         swiftonfile.swift.obj.diskfile.do_fsync = _mock_do_fsync
+        self._saved_fallocate = swiftonfile.swift.obj.diskfile.fallocate
+        swiftonfile.swift.obj.diskfile.fallocate = _mock_fallocate
         self.td = tempfile.mkdtemp()
         self.conf = dict(devices=self.td, mb_per_sync=2,
                          keep_cache_size=(1024 * 1024), mount_check=False)
@@ -148,6 +154,7 @@ class TestDiskFile(unittest.TestCase):
         swiftonfile.swift.common.utils.write_metadata = self._saved_ut_wm
         swiftonfile.swift.common.utils.read_metadata = self._saved_ut_rm
         swiftonfile.swift.obj.diskfile.do_fsync = self._saved_do_fsync
+        swiftonfile.swift.obj.diskfile.fallocate = self._saved_fallocate
         shutil.rmtree(self.td)
 
     def _get_diskfile(self, d, p, a, c, o, **kwargs):