[Qemu-devel] [PATCH v7 0/3] Fix qcow2+luks corruption introduced by commit 8ac0f15f335

[Qemu-devel] [PATCH v7 0/3] Fix qcow2+luks corruption introduced by commit 8ac0f15f335

[Maxim Levitsky]

Commit 8ac0f15f335 accidently broke the COW of non changed areas
of newly allocated clusters, when the write spans multiple clusters,
and needs COW both prior and after the write.
This results in 'after' COW area being encrypted with wrong
sector address, which render it corrupted.

Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=1745922

CC: qemu-stable <qemu-stable@nongnu.org>

V2: grammar, spelling and code style fixes.
V3: more fixes after the review.
V4: addressed review comments from Max Reitz,
    and futher refactored the qcow2_co_encrypt to just take full host and guest offset
    which simplifies everything.

V5: reworked the patches so one of them fixes the bug
    only and other one is just refactoring

V6: removed do_perform_cow_encrypt

V7: removed do_perform_cow_encrypt take two, this
    time I hopefully did that correctly :-)
    Also updated commit names and messages a bit

Best regards,
	Maxim Levitsky

Maxim Levitsky (3):
  block/qcow2: Fix corruption introduced by commit 8ac0f15f335
  block/qcow2: refactor encryption code
  qemu-iotests: Add test for bz #1745922

 block/qcow2-cluster.c      | 40 ++++++-----------
 block/qcow2-threads.c      | 63 ++++++++++++++++++++------
 block/qcow2.c              |  5 ++-
 block/qcow2.h              |  8 ++--
 tests/qemu-iotests/263     | 91 ++++++++++++++++++++++++++++++++++++++
 tests/qemu-iotests/263.out | 40 +++++++++++++++++
 tests/qemu-iotests/group   |  1 +
 7 files changed, 202 insertions(+), 46 deletions(-)
 create mode 100755 tests/qemu-iotests/263
 create mode 100644 tests/qemu-iotests/263.out

-- 
2.17.2

[Qemu-devel] [PATCH v7 1/3] block/qcow2: Fix corruption introduced by commit 8ac0f15f335

[Maxim Levitsky]

This fixes subtle corruption introduced by luks threaded encryption
in commit 8ac0f15f335

Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=1745922

The corruption happens when we do a write that
   * writes to two or more unallocated clusters at once
   * doesn't fully cover the first sector
   * doesn't fully cover the last sector
   * uses luks encryption

In this case, when allocating the new clusters we COW both areas
prior to the write and after the write, and we encrypt them.

The above mentioned commit accidentally made it so we encrypt the
second COW area using the physical cluster offset of the first area.

The problem is that offset_in_cluster in do_perform_cow_encrypt
can be larger that the cluster size, thus cluster_offset
will no longer point to the start of the cluster at which encrypted
area starts.

Next patch in this series will refactor the code to avoid all these
assumptions.

In the bugreport that was triggered by rebasing a luks image to new,
zero filled base, which lot of such writes, and causes some files
with zero areas to contain garbage there instead.
But as described above it can happen elsewhere as well


Signed-off-by: Maxim Levitsky <mlevitsk@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
---
 block/qcow2-cluster.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/block/qcow2-cluster.c b/block/qcow2-cluster.c
index dcacd3c450..bfeb0241d7 100644
--- a/block/qcow2-cluster.c
+++ b/block/qcow2-cluster.c
@@ -474,9 +474,10 @@ static bool coroutine_fn do_perform_cow_encrypt(BlockDriverState *bs,
         assert((offset_in_cluster & ~BDRV_SECTOR_MASK) == 0);
         assert((bytes & ~BDRV_SECTOR_MASK) == 0);
         assert(s->crypto);
-        if (qcow2_co_encrypt(bs, cluster_offset,
-                             src_cluster_offset + offset_in_cluster,
-                             buffer, bytes) < 0) {
+        if (qcow2_co_encrypt(bs,
+                start_of_cluster(s, cluster_offset + offset_in_cluster),
+                src_cluster_offset + offset_in_cluster,
+                buffer, bytes) < 0) {
             return false;
         }
     }
-- 
2.17.2

[Qemu-devel] [PATCH v7 2/3] block/qcow2: refactor encryption code

[Maxim Levitsky]

* Change the qcow2_co_{encrypt|decrypt} to just receive full host and
  guest offsets and use this function directly instead of calling
  do_perform_cow_encrypt (which is removed by that patch).

* Adjust qcow2_co_encdec to take full host and guest offsets as well.

* Document the qcow2_co_{encrypt|decrypt} arguments
  to prevent the bug fixed in former commit from hopefully
  happening again.

Signed-off-by: Maxim Levitsky <mlevitsk@redhat.com>
---
 block/qcow2-cluster.c | 41 ++++++++++------------------
 block/qcow2-threads.c | 63 +++++++++++++++++++++++++++++++++----------
 block/qcow2.c         |  5 ++--
 block/qcow2.h         |  8 +++---
 4 files changed, 70 insertions(+), 47 deletions(-)

diff --git a/block/qcow2-cluster.c b/block/qcow2-cluster.c
index bfeb0241d7..a2d4909024 100644
--- a/block/qcow2-cluster.c
+++ b/block/qcow2-cluster.c
@@ -462,28 +462,6 @@ static int coroutine_fn do_perform_cow_read(BlockDriverState *bs,
     return 0;
 }
 
-static bool coroutine_fn do_perform_cow_encrypt(BlockDriverState *bs,
-                                                uint64_t src_cluster_offset,
-                                                uint64_t cluster_offset,
-                                                unsigned offset_in_cluster,
-                                                uint8_t *buffer,
-                                                unsigned bytes)
-{
-    if (bytes && bs->encrypted) {
-        BDRVQcow2State *s = bs->opaque;
-        assert((offset_in_cluster & ~BDRV_SECTOR_MASK) == 0);
-        assert((bytes & ~BDRV_SECTOR_MASK) == 0);
-        assert(s->crypto);
-        if (qcow2_co_encrypt(bs,
-                start_of_cluster(s, cluster_offset + offset_in_cluster),
-                src_cluster_offset + offset_in_cluster,
-                buffer, bytes) < 0) {
-            return false;
-        }
-    }
-    return true;
-}
-
 static int coroutine_fn do_perform_cow_write(BlockDriverState *bs,
                                              uint64_t cluster_offset,
                                              unsigned offset_in_cluster,
@@ -891,11 +869,20 @@ static int perform_cow(BlockDriverState *bs, QCowL2Meta *m)
 
     /* Encrypt the data if necessary before writing it */
     if (bs->encrypted) {
-        if (!do_perform_cow_encrypt(bs, m->offset, m->alloc_offset,
-                                    start->offset, start_buffer,
-                                    start->nb_bytes) ||
-            !do_perform_cow_encrypt(bs, m->offset, m->alloc_offset,
-                                    end->offset, end_buffer, end->nb_bytes)) {
+        ret = qcow2_co_encrypt(bs,
+                               m->alloc_offset + start->offset,
+                               m->offset + start->offset,
+                               start_buffer, start->nb_bytes);
+        if (ret < 0) {
+            ret = -EIO;
+            goto fail;
+        }
+
+        ret = qcow2_co_encrypt(bs,
+                               m->alloc_offset + end->offset,
+                               m->offset + end->offset,
+                               end_buffer, end->nb_bytes);
+        if (ret < 0) {
             ret = -EIO;
             goto fail;
         }
diff --git a/block/qcow2-threads.c b/block/qcow2-threads.c
index 3b1e63fe41..8f5a0d1ebe 100644
--- a/block/qcow2-threads.c
+++ b/block/qcow2-threads.c
@@ -234,35 +234,70 @@ static int qcow2_encdec_pool_func(void *opaque)
 }
 
 static int coroutine_fn
-qcow2_co_encdec(BlockDriverState *bs, uint64_t file_cluster_offset,
-                  uint64_t offset, void *buf, size_t len, Qcow2EncDecFunc func)
+qcow2_co_encdec(BlockDriverState *bs, uint64_t host_offset,
+                uint64_t guest_offset, void *buf, size_t len,
+                Qcow2EncDecFunc func)
 {
     BDRVQcow2State *s = bs->opaque;
     Qcow2EncDecData arg = {
         .block = s->crypto,
-        .offset = s->crypt_physical_offset ?
-                      file_cluster_offset + offset_into_cluster(s, offset) :
-                      offset,
+        .offset = s->crypt_physical_offset ? host_offset : guest_offset,
         .buf = buf,
         .len = len,
         .func = func,
     };
 
-    return qcow2_co_process(bs, qcow2_encdec_pool_func, &arg);
+    assert(QEMU_IS_ALIGNED(guest_offset, BDRV_SECTOR_SIZE));
+    assert(QEMU_IS_ALIGNED(host_offset, BDRV_SECTOR_SIZE));
+    assert(QEMU_IS_ALIGNED(len, BDRV_SECTOR_SIZE));
+    assert(s->crypto);
+
+    return len == 0 ? 0 : qcow2_co_process(bs, qcow2_encdec_pool_func, &arg);
 }
 
+/*
+ * qcow2_co_encrypt()
+ *
+ * Encrypts one or more contiguous aligned sectors
+ *
+ * @host_offset - underlying storage offset of the first sector of the
+ * data to be encrypted
+ *
+ * @guest_offset - guest (virtual) offset of the first sector of the
+ * data to be encrypted
+ *
+ * @buf - buffer with the data to encrypt, that after encryption
+ *        will be written to the underlying storage device at
+ *        @host_offset
+ *
+ * @len - length of the buffer (must be a BDRV_SECTOR_SIZE multiple)
+ *
+ * Depending on the encryption method, @host_offset and/or @guest_offset
+ * may be used for generating the initialization vector for
+ * encryption.
+ *
+ * Note that while the whole range must be aligned on sectors, it
+ * does not have to be aligned on clusters and can also cross cluster
+ * boundaries
+ */
 int coroutine_fn
-qcow2_co_encrypt(BlockDriverState *bs, uint64_t file_cluster_offset,
-                 uint64_t offset, void *buf, size_t len)
+qcow2_co_encrypt(BlockDriverState *bs, uint64_t host_offset,
+                 uint64_t guest_offset, void *buf, size_t len)
 {
-    return qcow2_co_encdec(bs, file_cluster_offset, offset, buf, len,
-                             qcrypto_block_encrypt);
+    return qcow2_co_encdec(bs, host_offset, guest_offset, buf, len,
+                           qcrypto_block_encrypt);
 }
 
+/*
+ * qcow2_co_decrypt()
+ *
+ * Decrypts one or more contiguous aligned sectors
+ * Similar to qcow2_co_encrypt
+ */
 int coroutine_fn
-qcow2_co_decrypt(BlockDriverState *bs, uint64_t file_cluster_offset,
-                 uint64_t offset, void *buf, size_t len)
+qcow2_co_decrypt(BlockDriverState *bs, uint64_t host_offset,
+                 uint64_t guest_offset, void *buf, size_t len)
 {
-    return qcow2_co_encdec(bs, file_cluster_offset, offset, buf, len,
-                             qcrypto_block_decrypt);
+    return qcow2_co_encdec(bs, host_offset, guest_offset, buf, len,
+                           qcrypto_block_decrypt);
 }
diff --git a/block/qcow2.c b/block/qcow2.c
index 57734f20cf..ac768092bb 100644
--- a/block/qcow2.c
+++ b/block/qcow2.c
@@ -2069,7 +2069,8 @@ static coroutine_fn int qcow2_co_preadv_part(BlockDriverState *bs,
 
                 assert((offset & (BDRV_SECTOR_SIZE - 1)) == 0);
                 assert((cur_bytes & (BDRV_SECTOR_SIZE - 1)) == 0);
-                if (qcow2_co_decrypt(bs, cluster_offset, offset,
+                if (qcow2_co_decrypt(bs, cluster_offset + offset_in_cluster,
+                                     offset,
                                      cluster_data, cur_bytes) < 0) {
                     ret = -EIO;
                     goto fail;
@@ -2288,7 +2289,7 @@ static coroutine_fn int qcow2_co_pwritev_part(
             qemu_iovec_to_buf(qiov, qiov_offset + bytes_done,
                               cluster_data, cur_bytes);
 
-            if (qcow2_co_encrypt(bs, cluster_offset, offset,
+            if (qcow2_co_encrypt(bs, cluster_offset + offset_in_cluster, offset,
                                  cluster_data, cur_bytes) < 0) {
                 ret = -EIO;
                 goto out_unlocked;
diff --git a/block/qcow2.h b/block/qcow2.h
index 998bcdaef1..a488d761ff 100644
--- a/block/qcow2.h
+++ b/block/qcow2.h
@@ -758,10 +758,10 @@ ssize_t coroutine_fn
 qcow2_co_decompress(BlockDriverState *bs, void *dest, size_t dest_size,
                     const void *src, size_t src_size);
 int coroutine_fn
-qcow2_co_encrypt(BlockDriverState *bs, uint64_t file_cluster_offset,
-                 uint64_t offset, void *buf, size_t len);
+qcow2_co_encrypt(BlockDriverState *bs, uint64_t host_offset,
+                 uint64_t guest_offset, void *buf, size_t len);
 int coroutine_fn
-qcow2_co_decrypt(BlockDriverState *bs, uint64_t file_cluster_offset,
-                 uint64_t offset, void *buf, size_t len);
+qcow2_co_decrypt(BlockDriverState *bs, uint64_t host_offset,
+                 uint64_t guest_offset, void *buf, size_t len);
 
 #endif
-- 
2.17.2

[Qemu-devel] [PATCH v7 3/3] qemu-iotests: Add test for bz #1745922

[Maxim Levitsky]

Signed-off-by: Maxim Levitsky <mlevitsk@redhat.com>
Tested-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
---
 tests/qemu-iotests/263     | 91 ++++++++++++++++++++++++++++++++++++++
 tests/qemu-iotests/263.out | 40 +++++++++++++++++
 tests/qemu-iotests/group   |  1 +
 3 files changed, 132 insertions(+)
 create mode 100755 tests/qemu-iotests/263
 create mode 100644 tests/qemu-iotests/263.out

diff --git a/tests/qemu-iotests/263 b/tests/qemu-iotests/263
new file mode 100755
index 0000000000..d2c030fae9
--- /dev/null
+++ b/tests/qemu-iotests/263
@@ -0,0 +1,91 @@
+#!/usr/bin/env bash
+#
+# Test encrypted write that crosses cluster boundary of two unallocated clusters
+# Based on 188
+#
+# 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
+owner=mlevitsk@redhat.com
+
+seq=`basename $0`
+echo "QA output created by $seq"
+
+status=1	# failure is the default!
+
+_cleanup()
+{
+	_cleanup_test_img
+}
+trap "_cleanup; exit \$status" 0 1 2 3 15
+
+# get standard environment, filters and checks
+. ./common.rc
+. ./common.filter
+
+_supported_fmt qcow2
+_supported_proto generic
+_supported_os Linux
+
+
+size=1M
+
+SECRET="secret,id=sec0,data=astrochicken"
+QEMU_IO_OPTIONS=$QEMU_IO_OPTIONS_NO_FMT
+
+
+_run_test()
+{
+	echo "== reading the whole image =="
+	$QEMU_IO --object $SECRET -c "read -P 0 0 $size" --image-opts "$1" | _filter_qemu_io | _filter_testdir
+
+	echo
+	echo "== write two 512 byte sectors on a cluster boundary =="
+	$QEMU_IO --object $SECRET -c "write -P 0xAA 0xFE00 0x400" --image-opts "$1" | _filter_qemu_io | _filter_testdir
+
+	echo
+	echo "== verify that the rest of the image is not changed =="
+	$QEMU_IO --object $SECRET -c "read -P 0x00 0x00000 0xFE00" --image-opts "$1" | _filter_qemu_io | _filter_testdir
+	$QEMU_IO --object $SECRET -c "read -P 0xAA 0x0FE00 0x400" --image-opts "$1" | _filter_qemu_io | _filter_testdir
+	$QEMU_IO --object $SECRET -c "read -P 0x00 0x10200 0xEFE00" --image-opts "$1" | _filter_qemu_io | _filter_testdir
+
+}
+
+
+echo
+echo "testing LUKS qcow2 encryption"
+echo
+
+_make_test_img --object $SECRET -o "encrypt.format=luks,encrypt.key-secret=sec0,encrypt.iter-time=10,cluster_size=64K" $size
+_run_test "driver=$IMGFMT,encrypt.key-secret=sec0,file.filename=$TEST_IMG"
+_cleanup_test_img
+
+echo
+echo "testing legacy AES qcow2 encryption"
+echo
+
+
+_make_test_img --object $SECRET -o "encrypt.format=aes,encrypt.key-secret=sec0,cluster_size=64K" $size
+_run_test "driver=$IMGFMT,encrypt.key-secret=sec0,file.filename=$TEST_IMG"
+_cleanup_test_img
+
+
+
+# success, all done
+echo "*** done"
+rm -f $seq.full
+status=0
diff --git a/tests/qemu-iotests/263.out b/tests/qemu-iotests/263.out
new file mode 100644
index 0000000000..0c982c55cb
--- /dev/null
+++ b/tests/qemu-iotests/263.out
@@ -0,0 +1,40 @@
+QA output created by 263
+
+testing LUKS qcow2 encryption
+
+Formatting 'TEST_DIR/t.IMGFMT', fmt=IMGFMT size=1048576 encrypt.format=luks encrypt.key-secret=sec0 encrypt.iter-time=10
+== reading the whole image ==
+read 1048576/1048576 bytes at offset 0
+1 MiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
+
+== write two 512 byte sectors on a cluster boundary ==
+wrote 1024/1024 bytes at offset 65024
+1 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
+
+== verify that the rest of the image is not changed ==
+read 65024/65024 bytes at offset 0
+63.500 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
+read 1024/1024 bytes at offset 65024
+1 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
+read 982528/982528 bytes at offset 66048
+959.500 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
+
+testing legacy AES qcow2 encryption
+
+Formatting 'TEST_DIR/t.IMGFMT', fmt=IMGFMT size=1048576 encrypt.format=aes encrypt.key-secret=sec0
+== reading the whole image ==
+read 1048576/1048576 bytes at offset 0
+1 MiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
+
+== write two 512 byte sectors on a cluster boundary ==
+wrote 1024/1024 bytes at offset 65024
+1 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
+
+== verify that the rest of the image is not changed ==
+read 65024/65024 bytes at offset 0
+63.500 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
+read 1024/1024 bytes at offset 65024
+1 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
+read 982528/982528 bytes at offset 66048
+959.500 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
+*** done
diff --git a/tests/qemu-iotests/group b/tests/qemu-iotests/group
index 6082c74806..5d3da937e4 100644
--- a/tests/qemu-iotests/group
+++ b/tests/qemu-iotests/group
@@ -274,5 +274,6 @@
 257 rw
 258 rw quick
 262 rw quick migration
+263 rw quick
 265 rw auto quick
 266 rw quick
-- 
2.17.2

Re: [Qemu-devel] [PATCH v7 2/3] block/qcow2: refactor encryption code

[Vladimir Sementsov-Ogievskiy]

15.09.2019 23:36, Maxim Levitsky wrote:
> * Change the qcow2_co_{encrypt|decrypt} to just receive full host and
>    guest offsets and use this function directly instead of calling
>    do_perform_cow_encrypt (which is removed by that patch).
> 
> * Adjust qcow2_co_encdec to take full host and guest offsets as well.
> 
> * Document the qcow2_co_{encrypt|decrypt} arguments
>    to prevent the bug fixed in former commit from hopefully
>    happening again.
> 
> Signed-off-by: Maxim Levitsky <mlevitsk@redhat.com>
> ---
>   block/qcow2-cluster.c | 41 ++++++++++------------------
>   block/qcow2-threads.c | 63 +++++++++++++++++++++++++++++++++----------
>   block/qcow2.c         |  5 ++--
>   block/qcow2.h         |  8 +++---
>   4 files changed, 70 insertions(+), 47 deletions(-)
> 
> diff --git a/block/qcow2-cluster.c b/block/qcow2-cluster.c
> index bfeb0241d7..a2d4909024 100644
> --- a/block/qcow2-cluster.c
> +++ b/block/qcow2-cluster.c

[..]

>   static int coroutine_fn do_perform_cow_write(BlockDriverState *bs,
>                                                uint64_t cluster_offset,
>                                                unsigned offset_in_cluster,
> @@ -891,11 +869,20 @@ static int perform_cow(BlockDriverState *bs, QCowL2Meta *m)
>   
>       /* Encrypt the data if necessary before writing it */
>       if (bs->encrypted) {
> -        if (!do_perform_cow_encrypt(bs, m->offset, m->alloc_offset,
> -                                    start->offset, start_buffer,
> -                                    start->nb_bytes) ||
> -            !do_perform_cow_encrypt(bs, m->offset, m->alloc_offset,
> -                                    end->offset, end_buffer, end->nb_bytes)) {
> +        ret = qcow2_co_encrypt(bs,
> +                               m->alloc_offset + start->offset,
> +                               m->offset + start->offset,
> +                               start_buffer, start->nb_bytes);
> +        if (ret < 0) {
> +            ret = -EIO;
> +            goto fail;

Just go to fail I think, don't reassign ret variable.

> +        }
> +
> +        ret = qcow2_co_encrypt(bs,
> +                               m->alloc_offset + end->offset,
> +                               m->offset + end->offset,
> +                               end_buffer, end->nb_bytes);
> +        if (ret < 0) {
>               ret = -EIO;

and here.

with these two places fixed:
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>

(I think, these simple changes may be applied while queuing, so you don't
need to resend, if there no other reasons)




-- 
Best regards,
Vladimir

Re: [Qemu-devel] [PATCH v7 1/3] block/qcow2: Fix corruption introduced by commit 8ac0f15f335

[Max Reitz]

[-- Attachment #1.1: Type: text/plain, Size: 1564 bytes --]

On 15.09.19 22:36, Maxim Levitsky wrote:
> This fixes subtle corruption introduced by luks threaded encryption
> in commit 8ac0f15f335
> 
> Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=1745922
> 
> The corruption happens when we do a write that
>    * writes to two or more unallocated clusters at once
>    * doesn't fully cover the first sector
>    * doesn't fully cover the last sector
>    * uses luks encryption
> 
> In this case, when allocating the new clusters we COW both areas
> prior to the write and after the write, and we encrypt them.
> 
> The above mentioned commit accidentally made it so we encrypt the
> second COW area using the physical cluster offset of the first area.
> 
> The problem is that offset_in_cluster in do_perform_cow_encrypt
> can be larger that the cluster size, thus cluster_offset
> will no longer point to the start of the cluster at which encrypted
> area starts.
> 
> Next patch in this series will refactor the code to avoid all these
> assumptions.
> 
> In the bugreport that was triggered by rebasing a luks image to new,
> zero filled base, which lot of such writes, and causes some files
> with zero areas to contain garbage there instead.
> But as described above it can happen elsewhere as well
> 
> 
> Signed-off-by: Maxim Levitsky <mlevitsk@redhat.com>
> Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
> ---
>  block/qcow2-cluster.c | 7 ++++---
>  1 file changed, 4 insertions(+), 3 deletions(-)

Reviewed-by: Max Reitz <mreitz@redhat.com>


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: [Qemu-devel] [PATCH v7 2/3] block/qcow2: refactor encryption code

[Max Reitz]

[-- Attachment #1.1: Type: text/plain, Size: 1386 bytes --]

On 15.09.19 22:36, Maxim Levitsky wrote:
> * Change the qcow2_co_{encrypt|decrypt} to just receive full host and
>   guest offsets and use this function directly instead of calling
>   do_perform_cow_encrypt (which is removed by that patch).
> 
> * Adjust qcow2_co_encdec to take full host and guest offsets as well.
> 
> * Document the qcow2_co_{encrypt|decrypt} arguments
>   to prevent the bug fixed in former commit from hopefully
>   happening again.
> 
> Signed-off-by: Maxim Levitsky <mlevitsk@redhat.com>
> ---
>  block/qcow2-cluster.c | 41 ++++++++++------------------
>  block/qcow2-threads.c | 63 +++++++++++++++++++++++++++++++++----------
>  block/qcow2.c         |  5 ++--
>  block/qcow2.h         |  8 +++---
>  4 files changed, 70 insertions(+), 47 deletions(-)

Looks good to me, but there are conflicts with two series I’ve already
taken to my branch: Namely “qcow2: async handling of fragmented io” by
Vladimir and “Alignment checks cleanup​” by Nir.

Unfortunately, those conflicts (while not difficult to resolve) are not
quite trivial, so I’d rather not resolve them myself.

I’ll send a pull request today, but until they are in master, you could
rebase on my branch (either of
- https://git.xanclic.moe/XanClic/qemu.git block
- https://github.com/XanClic/qemu.git block
, whichever you prefer that works).

Max


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: [Qemu-devel] [PATCH v7 3/3] qemu-iotests: Add test for bz #1745922

[Max Reitz]

[-- Attachment #1.1: Type: text/plain, Size: 524 bytes --]

On 15.09.19 22:36, Maxim Levitsky wrote:
> Signed-off-by: Maxim Levitsky <mlevitsk@redhat.com>
> Tested-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
> ---
>  tests/qemu-iotests/263     | 91 ++++++++++++++++++++++++++++++++++++++
>  tests/qemu-iotests/263.out | 40 +++++++++++++++++
>  tests/qemu-iotests/group   |  1 +
>  3 files changed, 132 insertions(+)
>  create mode 100755 tests/qemu-iotests/263
>  create mode 100644 tests/qemu-iotests/263.out

Reviewed-by: Max Reitz <mreitz@redhat.com>


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: [Qemu-devel] [PATCH v7 0/3] Fix qcow2+luks corruption introduced by commit 8ac0f15f335

[Max Reitz]

[-- Attachment #1.1: Type: text/plain, Size: 1448 bytes --]

On 15.09.19 22:36, Maxim Levitsky wrote:
> Commit 8ac0f15f335 accidently broke the COW of non changed areas
> of newly allocated clusters, when the write spans multiple clusters,
> and needs COW both prior and after the write.
> This results in 'after' COW area being encrypted with wrong
> sector address, which render it corrupted.
> 
> Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=1745922
> 
> CC: qemu-stable <qemu-stable@nongnu.org>
> 
> V2: grammar, spelling and code style fixes.
> V3: more fixes after the review.
> V4: addressed review comments from Max Reitz,
>     and futher refactored the qcow2_co_encrypt to just take full host and guest offset
>     which simplifies everything.
> 
> V5: reworked the patches so one of them fixes the bug
>     only and other one is just refactoring
> 
> V6: removed do_perform_cow_encrypt
> 
> V7: removed do_perform_cow_encrypt take two, this
>     time I hopefully did that correctly :-)
>     Also updated commit names and messages a bit

Luckily for you (maybe), Vladimir’s series doesn‘t quite pass the
iotests for me, so unfortunately (I find it unfortunate) I had to remove
it from my branch.  Thus, the conflicts are much more tame and I felt
comfortable taking the series to my branch (with the remaining trivial
conflicts resolved, and with Vladimir’s suggestion applied):

https://git.xanclic.moe/XanClic/qemu/commits/branch/block

Max


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: [Qemu-devel] [PATCH v7 0/3] Fix qcow2+luks corruption introduced by commit 8ac0f15f335

[Maxim Levitsky]

On Mon, 2019-09-16 at 15:39 +0200, Max Reitz wrote:
> On 15.09.19 22:36, Maxim Levitsky wrote:
> > Commit 8ac0f15f335 accidently broke the COW of non changed areas
> > of newly allocated clusters, when the write spans multiple clusters,
> > and needs COW both prior and after the write.
> > This results in 'after' COW area being encrypted with wrong
> > sector address, which render it corrupted.
> > 
> > Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=1745922
> > 
> > CC: qemu-stable <qemu-stable@nongnu.org>
> > 
> > V2: grammar, spelling and code style fixes.
> > V3: more fixes after the review.
> > V4: addressed review comments from Max Reitz,
> >     and futher refactored the qcow2_co_encrypt to just take full host and guest offset
> >     which simplifies everything.
> > 
> > V5: reworked the patches so one of them fixes the bug
> >     only and other one is just refactoring
> > 
> > V6: removed do_perform_cow_encrypt
> > 
> > V7: removed do_perform_cow_encrypt take two, this
> >     time I hopefully did that correctly :-)
> >     Also updated commit names and messages a bit
> 
> Luckily for you (maybe), Vladimir’s series doesn‘t quite pass the
> iotests for me, so unfortunately (I find it unfortunate) I had to remove
> it from my branch.  Thus, the conflicts are much more tame and I felt
> comfortable taking the series to my branch (with the remaining trivial
> conflicts resolved, and with Vladimir’s suggestion applied):
> 
> https://git.xanclic.moe/XanClic/qemu/commits/branch/block


First of all, Thanks!

I don't know if this is luckily for me since I already rebased my series on top of  
https://git.xanclic.moe/XanClic/qemu.git,
and run all qcow2 iotests, and only tests 
162 169 194 196 234 262 failed, and I know that 162 always fails
due to that kernel change I talked about here few days ago,
and rest for the AF_UNIX path len, which I need to do something
about in the long term. I sometimes do a separate build in 
directory which path doesn't trigger this, and sometimes,
when I know that I haven't done significant changes to the patches,
I just let these tests fail. In long term, maybe even in a few days
I'll allocate some time to rethink the build environment here to
fix that permanently.

Now I am rerunning the iotests just for fun, in short enough directory
to see if I can reproduce the failure that you had. After looking
in your report, that iotest 026 fails, it does pass here, but
then I am only running these iotests on my laptop so I probably
don't trigger the race you were able to.

So thanks again!
Best regards,
	Maxim Levitsky

Re: [Qemu-devel] [PATCH v7 0/3] Fix qcow2+luks corruption introduced by commit 8ac0f15f335

[Maxim Levitsky]

On Mon, 2019-09-16 at 16:59 +0300, Maxim Levitsky wrote:
> On Mon, 2019-09-16 at 15:39 +0200, Max Reitz wrote:
> > On 15.09.19 22:36, Maxim Levitsky wrote:
> > > Commit 8ac0f15f335 accidently broke the COW of non changed areas
> > > of newly allocated clusters, when the write spans multiple clusters,
> > > and needs COW both prior and after the write.
> > > This results in 'after' COW area being encrypted with wrong
> > > sector address, which render it corrupted.
> > > 
> > > Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=1745922
> > > 
> > > CC: qemu-stable <qemu-stable@nongnu.org>
> > > 
> > > V2: grammar, spelling and code style fixes.
> > > V3: more fixes after the review.
> > > V4: addressed review comments from Max Reitz,
> > >     and futher refactored the qcow2_co_encrypt to just take full host and guest offset
> > >     which simplifies everything.
> > > 
> > > V5: reworked the patches so one of them fixes the bug
> > >     only and other one is just refactoring
> > > 
> > > V6: removed do_perform_cow_encrypt
> > > 
> > > V7: removed do_perform_cow_encrypt take two, this
> > >     time I hopefully did that correctly :-)
> > >     Also updated commit names and messages a bit
> > 
> > Luckily for you (maybe), Vladimir’s series doesn‘t quite pass the
> > iotests for me, so unfortunately (I find it unfortunate) I had to remove
> > it from my branch.  Thus, the conflicts are much more tame and I felt
> > comfortable taking the series to my branch (with the remaining trivial
> > conflicts resolved, and with Vladimir’s suggestion applied):
> > 
> > https://git.xanclic.moe/XanClic/qemu/commits/branch/block
> 
> 
> First of all, Thanks!
> 
> I don't know if this is luckily for me since I already rebased my series on top of  
> https://git.xanclic.moe/XanClic/qemu.git,
> and run all qcow2 iotests, and only tests 
> 162 169 194 196 234 262 failed, and I know that 162 always fails
> due to that kernel change I talked about here few days ago,
> and rest for the AF_UNIX path len, which I need to do something
> about in the long term. I sometimes do a separate build in 
> directory which path doesn't trigger this, and sometimes,
> when I know that I haven't done significant changes to the patches,
> I just let these tests fail. In long term, maybe even in a few days
> I'll allocate some time to rethink the build environment here to
> fix that permanently.
> 
> Now I am rerunning the iotests just for fun, in short enough directory
> to see if I can reproduce the failure that you had. After looking
> in your report, that iotest 026 fails, it does pass here, but
> then I am only running these iotests on my laptop so I probably
> don't trigger the race you were able to.

Passed all the tests but 162, not that it matters much to be honest.
Best regards,
	Maxim Levitsky