Skip to contentSkip to conversation

qemu-img: Allow rebase with no input base

PATCH In review Max Reitz 2 participants 5 messages May 9, 2019 – May 10, 2019

lore
Share & tools
lore

Max Reitz

author
#0 lore raw

v3 cover qemu-img: Allow rebase with no input base

This series allows using qemu-img rebase (without -u) on images that do not have a backing file. Right now, this fails with the rather cryptic error message:

$ qemu-img rebase -b base.qcow2 foo.qcow2
qemu-img: Could not open old backing file '': The 'file' block driver requires a file name

Yeah, well, OK.

With how rebase currently works, this would lead to the overlay being filled with zeroes, however. This is where patch 2 comes in and instead makes rebase use blk_pwrite_zeroes() whenever it handles an area past the input’s backing file’s EOF.

(Note that additionally we could try to punch holes in the overlay whenever it matches the new backing file, but that’s something I’ll put off for later. (We don’t even have a reliable method for punching holes into an overlay yet, although I would like to have such because it could make active commit more efficient.))

And patch 3 adds the usual test.

v3:
- Patch 3: Move test to own file so it doesn’t run for qed or qcow2 v2 (because it can’t, it requires zero clusters)

git backport-diff against v2:

Key:
[----] : patches are identical
[####] : number of functional differences between upstream/downstream patch [down] : patch is downstream-only
The flags [FC] indicate (F)unctional and (C)ontextual differences, respectively

001/3:[----] [--] 'qemu-img: Allow rebase with no input base' 002/3:[----] [--] 'qemu-img: Use zero writes after source backing EOF' 003/3:[0059] [FC] 'iotests: Add test for rebase without input base'

v2:
- Patch 1: Added example to the commit message, and made a note that this is about rebase without -u [Eric]

git backport-diff against v1:

Key:
[----] : patches are identical
[####] : number of functional differences between upstream/downstream patch [down] : patch is downstream-only
The flags [FC] indicate (F)unctional and (C)ontextual differences, respectively

001/3:[----] [--] 'qemu-img: Allow rebase with no input base' 002/3:[----] [--] 'qemu-img: Use zero writes after source backing EOF' 003/3:[----] [--] 'iotests: Add test for rebase without input base'

Max Reitz (3):
qemu-img: Allow rebase with no input base
qemu-img: Use zero writes after source backing EOF
iotests: Add test for rebase without input base

 qemu-img.c                 |  72 ++++++++++++---------
 tests/qemu-iotests/252     | 124 +++++++++++++++++++++++++++++++++++++
 tests/qemu-iotests/252.out |  39 ++++++++++++
 tests/qemu-iotests/group   |   1 +
 4 files changed, 207 insertions(+), 29 deletions(-)
 create mode 100755 tests/qemu-iotests/252
 create mode 100644 tests/qemu-iotests/252.out
Signature · 1 line
2.21.0
author
#1 lore raw

PATCH v3 1/3 qemu-img: Allow rebase with no input base

Currently, without -u, you cannot add a backing file to an image when it currently has none:

$ qemu-img rebase -b base.qcow2 foo.qcow2
qemu-img: Could not open old backing file '': The 'file' block driver
requires a file name

It is really simple to allow this, though (effectively by setting old_backing_size to 0), so this patch does just that.

---
 qemu-img.c | 61 ++++++++++++++++++++++++++++++------------------------
 1 file changed, 34 insertions(+), 27 deletions(-)
qemu-img.c +34 -27
index 71c92f142a..cfa44b4153 100644
@@ -3312,26 +3312,30 @@ static int img_rebase(int argc, char **argv)
         char backing_name[PATH_MAX];
         QDict *options = NULL;
 
-        if (bs->backing_format[0] != '\0') {
-            options = qdict_new();
-            qdict_put_str(options, "driver", bs->backing_format);
-        }
-
-        if (force_share) {
-            if (!options) {
+        if (bs->backing) {
+            if (bs->backing_format[0] != '\0') {
                 options = qdict_new();
+                qdict_put_str(options, "driver", bs->backing_format);
             }
-            qdict_put_bool(options, BDRV_OPT_FORCE_SHARE, true);
-        }
-        bdrv_get_backing_filename(bs, backing_name, sizeof(backing_name));
-        blk_old_backing = blk_new_open(backing_name, NULL,
-                                       options, src_flags, &local_err);
-        if (!blk_old_backing) {
-            error_reportf_err(local_err,
-                              "Could not open old backing file '%s': ",
-                              backing_name);
-            ret = -1;
-            goto out;
+
+            if (force_share) {
+                if (!options) {
+                    options = qdict_new();
+                }
+                qdict_put_bool(options, BDRV_OPT_FORCE_SHARE, true);
+            }
+            bdrv_get_backing_filename(bs, backing_name, sizeof(backing_name));
+            blk_old_backing = blk_new_open(backing_name, NULL,
+                                           options, src_flags, &local_err);
+            if (!blk_old_backing) {
+                error_reportf_err(local_err,
+                                  "Could not open old backing file '%s': ",
+                                  backing_name);
+                ret = -1;
+                goto out;
+            }
+        } else {
+            blk_old_backing = NULL;
         }
 
         if (out_baseimg[0]) {
@@ -3384,7 +3388,7 @@ static int img_rebase(int argc, char **argv)
      */
     if (!unsafe) {
         int64_t size;
-        int64_t old_backing_size;
+        int64_t old_backing_size = 0;
         int64_t new_backing_size = 0;
         uint64_t offset;
         int64_t n;
@@ -3400,15 +3404,18 @@ static int img_rebase(int argc, char **argv)
             ret = -1;
             goto out;
         }
-        old_backing_size = blk_getlength(blk_old_backing);
-        if (old_backing_size < 0) {
-            char backing_name[PATH_MAX];
+        if (blk_old_backing) {
+            old_backing_size = blk_getlength(blk_old_backing);
+            if (old_backing_size < 0) {
+                char backing_name[PATH_MAX];
 
-            bdrv_get_backing_filename(bs, backing_name, sizeof(backing_name));
-            error_report("Could not get size of '%s': %s",
-                         backing_name, strerror(-old_backing_size));
-            ret = -1;
-            goto out;
+                bdrv_get_backing_filename(bs, backing_name,
+                                          sizeof(backing_name));
+                error_report("Could not get size of '%s': %s",
+                             backing_name, strerror(-old_backing_size));
+                ret = -1;
+                goto out;
+            }
         }
         if (blk_new_backing) {
             new_backing_size = blk_getlength(blk_new_backing);
Signature · 1 line
2.21.0
author
#2 lore raw

PATCH v3 2/3 qemu-img: Use zero writes after source backing EOF

In reply to Max Reitz

Past the end of the source backing file, we memset() buf_old to zero, so it is clearly easy to use blk_pwrite_zeroes() instead of blk_pwrite() then.

---
 qemu-img.c | 11 +++++++++--
 1 file changed, 9 insertions(+), 2 deletions(-)
qemu-img.c +9 -2
index cfa44b4153..28fba1e7a7 100644
@@ -3432,6 +3432,8 @@ static int img_rebase(int argc, char **argv)
         }
 
         for (offset = 0; offset < size; offset += n) {
+            bool buf_old_is_zero = false;
+
             /* How many bytes can we handle with the next read? */
             n = MIN(IO_BUF_SIZE, size - offset);
 
@@ -3452,6 +3454,7 @@ static int img_rebase(int argc, char **argv)
              */
             if (offset >= old_backing_size) {
                 memset(buf_old, 0, n);
+                buf_old_is_zero = true;
             } else {
                 if (offset + n > old_backing_size) {
                     n = old_backing_size - offset;
@@ -3487,8 +3490,12 @@ static int img_rebase(int argc, char **argv)
                 if (compare_buffers(buf_old + written, buf_new + written,
                                     n - written, &pnum))
                 {
-                    ret = blk_pwrite(blk, offset + written,
-                                     buf_old + written, pnum, 0);
+                    if (buf_old_is_zero) {
+                        ret = blk_pwrite_zeroes(blk, offset + written, pnum, 0);
+                    } else {
+                        ret = blk_pwrite(blk, offset + written,
+                                         buf_old + written, pnum, 0);
+                    }
                     if (ret < 0) {
                         error_report("Error while writing to COW image: %s",
                             strerror(-ret));
Signature · 1 line
2.21.0
author
#3 lore raw

PATCH v3 3/3 iotests: Add test for rebase without input base

In reply to Max Reitz

This patch adds a test for rebasing an image that currently does not have a backing file.

---
 tests/qemu-iotests/252     | 124 +++++++++++++++++++++++++++++++++++++
 tests/qemu-iotests/252.out |  39 ++++++++++++
 tests/qemu-iotests/group   |   1 +
 3 files changed, 164 insertions(+)
 create mode 100755 tests/qemu-iotests/252
 create mode 100644 tests/qemu-iotests/252.out
tests/qemu-iotests/252 new file +124 -0
new file mode 100755
index 0000000000..f6c8f71444
@@ -0,0 +1,124 @@
+#!/usr/bin/env bash
+#
+# Tests for rebasing COW images that require zero cluster support
+#
+# Copyright (C) 2019 Red Hat, Inc.
+#
+# This program is free software; you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation; either version 2 of the License, or
+# (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program.  If not, see <http://www.gnu.org/licenses/>.
+#
+
+# creator
[email protected]
+
+seq=$(basename $0)
+echo "QA output created by $seq"
+
+status=1	# failure is the default!
+
+_cleanup()
+{
+    _cleanup_test_img
+    rm -f "$TEST_IMG.base_new"
+}
+trap "_cleanup; exit \$status" 0 1 2 3 15
+
+# get standard environment, filters and checks
+. ./common.rc
+. ./common.filter
+. ./common.pattern
+
+# Currently only qcow2 and qed support rebasing, and only qcow2 v3 has
+# zero cluster support
+_supported_fmt qcow2
+_unsupported_imgopts 'compat=0.10'
+_supported_proto file
+_supported_os Linux
+
+CLUSTER_SIZE=65536
+
+echo
+echo "=== Test rebase without input base ==="
+echo
+
+# Cluster allocations to be tested:
+#
+# Backing (new) 11 -- 11 -- 11 --
+# COW image     22 22 11 11 -- --
+#
+# Expected result:
+#
+# COW image     22 22 11 11 00 --
+#
+# (Cluster 2 might be "--" after the rebase, too, but rebase just
+#  compares the new backing file to the old one and disregards the
+#  overlay.  Therefore, it will never discard overlay clusters.)
+
+_make_test_img $((6 * CLUSTER_SIZE))
+TEST_IMG="$TEST_IMG.base_new" _make_test_img $((6 * CLUSTER_SIZE))
+
+echo
+
+$QEMU_IO "$TEST_IMG" \
+    -c "write -P 0x22 $((0 * CLUSTER_SIZE)) $((2 * CLUSTER_SIZE))" \
+    -c "write -P 0x11 $((2 * CLUSTER_SIZE)) $((2 * CLUSTER_SIZE))" \
+    | _filter_qemu_io
+
+$QEMU_IO "$TEST_IMG.base_new" \
+    -c "write -P 0x11 $((0 * CLUSTER_SIZE)) $CLUSTER_SIZE" \
+    -c "write -P 0x11 $((2 * CLUSTER_SIZE)) $CLUSTER_SIZE" \
+    -c "write -P 0x11 $((4 * CLUSTER_SIZE)) $CLUSTER_SIZE" \
+    | _filter_qemu_io
+
+echo
+
+# This should be a no-op
+$QEMU_IMG rebase -b "" "$TEST_IMG"
+
+# Verify the data is correct
+$QEMU_IO "$TEST_IMG" \
+    -c "read -P 0x22 $((0 * CLUSTER_SIZE)) $((2 * CLUSTER_SIZE))" \
+    -c "read -P 0x11 $((2 * CLUSTER_SIZE)) $((2 * CLUSTER_SIZE))" \
+    -c "read -P 0x00 $((4 * CLUSTER_SIZE)) $((2 * CLUSTER_SIZE))" \
+    | _filter_qemu_io
+
+echo
+
+# Verify the allocation status (first four cluster should be allocated
+# in TEST_IMG, clusters 4 and 5 should be unallocated (marked as zero
+# clusters here because there is no backing file))
+$QEMU_IMG map --output=json "$TEST_IMG" | _filter_qemu_img_map
+
+echo
+
+$QEMU_IMG rebase -b "$TEST_IMG.base_new" "$TEST_IMG"
+
+# Verify the data is correct
+$QEMU_IO "$TEST_IMG" \
+    -c "read -P 0x22 $((0 * CLUSTER_SIZE)) $((2 * CLUSTER_SIZE))" \
+    -c "read -P 0x11 $((2 * CLUSTER_SIZE)) $((2 * CLUSTER_SIZE))" \
+    -c "read -P 0x00 $((4 * CLUSTER_SIZE)) $((2 * CLUSTER_SIZE))" \
+    | _filter_qemu_io
+
+echo
+
+# Verify the allocation status (first four cluster should be allocated
+# in TEST_IMG, cluster 4 should be zero, and cluster 5 should be
+# unallocated (signified by '"depth": 1'))
+$QEMU_IMG map --output=json "$TEST_IMG" | _filter_qemu_img_map
+
+
+# success, all done
+echo "*** done"
+rm -f $seq.full
+status=0
tests/qemu-iotests/252.out new file +39 -0
new file mode 100644
index 0000000000..12dce889f8
@@ -0,0 +1,39 @@
+QA output created by 252
+
+=== Test rebase without input base ===
+
+Formatting 'TEST_DIR/t.IMGFMT', fmt=IMGFMT size=393216
+Formatting 'TEST_DIR/t.IMGFMT.base_new', fmt=IMGFMT size=393216
+
+wrote 131072/131072 bytes at offset 0
+128 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
+wrote 131072/131072 bytes at offset 131072
+128 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
+wrote 65536/65536 bytes at offset 0
+64 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
+wrote 65536/65536 bytes at offset 131072
+64 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
+wrote 65536/65536 bytes at offset 262144
+64 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
+
+read 131072/131072 bytes at offset 0
+128 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
+read 131072/131072 bytes at offset 131072
+128 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
+read 131072/131072 bytes at offset 262144
+128 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
+
+[{ "start": 0, "length": 262144, "depth": 0, "zero": false, "data": true, "offset": OFFSET},
+{ "start": 262144, "length": 131072, "depth": 0, "zero": true, "data": false}]
+
+read 131072/131072 bytes at offset 0
+128 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
+read 131072/131072 bytes at offset 131072
+128 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
+read 131072/131072 bytes at offset 262144
+128 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
+
+[{ "start": 0, "length": 262144, "depth": 0, "zero": false, "data": true, "offset": OFFSET},
+{ "start": 262144, "length": 65536, "depth": 0, "zero": true, "data": false},
+{ "start": 327680, "length": 65536, "depth": 1, "zero": true, "data": false}]
+*** done
tests/qemu-iotests/group +1 -0
index 7ac9a5ea4a..00e474ab0a 100644
@@ -249,3 +249,4 @@
 247 rw auto quick
 248 rw auto quick
 249 rw auto quick
+252 rw auto backing quick
Signature · 1 line
2.21.0
#4 lore raw
In reply to Max Reitz

Am 09.05.2019 um 19:52 hat Max Reitz geschrieben:

Quoted 21 lines

This series allows using qemu-img rebase (without -u) on images that do not have a backing file. Right now, this fails with the rather cryptic error message:

$ qemu-img rebase -b base.qcow2 foo.qcow2
qemu-img: Could not open old backing file '': The 'file' block driver requires a file name

Yeah, well, OK.

With how rebase currently works, this would lead to the overlay being filled with zeroes, however. This is where patch 2 comes in and instead makes rebase use blk_pwrite_zeroes() whenever it handles an area past the input’s backing file’s EOF.

(Note that additionally we could try to punch holes in the overlay whenever it matches the new backing file, but that’s something I’ll put off for later. (We don’t even have a reliable method for punching holes into an overlay yet, although I would like to have such because it could make active commit more efficient.))

Quoted 6 lines

And patch 3 adds the usual test.

v3:
- Patch 3: Move test to own file so it doesn’t run for qed or qcow2 v2 (because it can’t, it requires zero clusters)

Thanks, updated in the block branch.

Kevin