This function iterates over all snapshots of a qcow2 file in order to
expand all zero clusters, but it does not validate the snapshots' L1
tables first.
We now have a function to take care of this, so let's use it.
We can also take the opportunity to replace the sector-based
bdrv_read() with bdrv_pread().
Signed-off-by: Alberto Garcia <berto@igalia.com>
---
block/qcow2-cluster.c | 20 +++++++++++++-------
tests/qemu-iotests/080 | 2 ++
tests/qemu-iotests/080.out | 2 ++
3 files changed, 17 insertions(+), 7 deletions(-)
diff --git a/block/qcow2-cluster.c b/block/qcow2-cluster.c
index e406b0f3b9..40167ac09c 100644
--- a/block/qcow2-cluster.c
+++ b/block/qcow2-cluster.c
@@ -2092,11 +2092,18 @@ int qcow2_expand_zero_clusters(BlockDriverState *bs,
}
for (i = 0; i < s->nb_snapshots; i++) {
- int l1_sectors = DIV_ROUND_UP(s->snapshots[i].l1_size *
- sizeof(uint64_t), BDRV_SECTOR_SIZE);
+ int l1_size2;
+ uint64_t *new_l1_table;
- uint64_t *new_l1_table =
- g_try_realloc(l1_table, l1_sectors * BDRV_SECTOR_SIZE);
+ ret = qcow2_validate_table(bs, s->snapshots[i].l1_table_offset,
+ s->snapshots[i].l1_size, sizeof(uint64_t),
+ QCOW_MAX_L1_SIZE, "", NULL);
+ if (ret < 0) {
+ return ret;
+ }
+
+ l1_size2 = s->snapshots[i].l1_size * sizeof(uint64_t);
+ new_l1_table = g_try_realloc(l1_table, l1_size2);
if (!new_l1_table) {
ret = -ENOMEM;
@@ -2105,9 +2112,8 @@ int qcow2_expand_zero_clusters(BlockDriverState *bs,
l1_table = new_l1_table;
- ret = bdrv_read(bs->file,
- s->snapshots[i].l1_table_offset / BDRV_SECTOR_SIZE,
- (void *)l1_table, l1_sectors);
+ ret = bdrv_pread(bs->file, s->snapshots[i].l1_table_offset,
+ l1_table, l1_size2);
if (ret < 0) {
goto fail;
}
diff --git a/tests/qemu-iotests/080 b/tests/qemu-iotests/080
index 6a10e7defa..5622604f83 100755
--- a/tests/qemu-iotests/080
+++ b/tests/qemu-iotests/080
@@ -177,6 +177,7 @@ _make_test_img 64M
{ $QEMU_IMG snapshot -c test $TEST_IMG; } 2>&1 | _filter_testdir
poke_file "$TEST_IMG" "$offset_snap1_l1_offset" "\x00\x00\x00\x00\x00\x40\x02\x00"
{ $QEMU_IMG convert -s test $TEST_IMG $TEST_IMG.snap; } 2>&1 | _filter_testdir
+{ $QEMU_IMG amend -o compat=0.10 $TEST_IMG; } 2>&1 | _filter_testdir
echo
echo "== Invalid snapshot L1 table size =="
@@ -185,6 +186,7 @@ _make_test_img 64M
{ $QEMU_IMG snapshot -c test $TEST_IMG; } 2>&1 | _filter_testdir
poke_file "$TEST_IMG" "$offset_snap1_l1_size" "\x10\x00\x00\x00"
{ $QEMU_IMG convert -s test $TEST_IMG $TEST_IMG.snap; } 2>&1 | _filter_testdir
+{ $QEMU_IMG amend -o compat=0.10 $TEST_IMG; } 2>&1 | _filter_testdir
# success, all done
echo "*** done"
diff --git a/tests/qemu-iotests/080.out b/tests/qemu-iotests/080.out
index f0d9038d55..5d9030ab93 100644
--- a/tests/qemu-iotests/080.out
+++ b/tests/qemu-iotests/080.out
@@ -64,10 +64,12 @@ Formatting 'TEST_DIR/t.IMGFMT', fmt=IMGFMT size=67108864
wrote 512/512 bytes at offset 0
512 bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
qemu-img: Failed to load snapshot: Snapshot L1 table offset invalid
+qemu-img: Error while amending options: Invalid argument
== Invalid snapshot L1 table size ==
Formatting 'TEST_DIR/t.IMGFMT', fmt=IMGFMT size=67108864
wrote 512/512 bytes at offset 0
512 bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
qemu-img: Failed to load snapshot: Snapshot L1 table too large
+qemu-img: Error while amending options: File too large
*** done
--
2.11.0
On 03/01/2018 10:27 AM, Alberto Garcia wrote: > This function iterates over all snapshots of a qcow2 file in order to > expand all zero clusters, but it does not validate the snapshots' L1 > tables first. > > We now have a function to take care of this, so let's use it. > > We can also take the opportunity to replace the sector-based > bdrv_read() with bdrv_pread(). Doesn't my pending patch do that as well? https://lists.gnu.org/archive/html/qemu-devel/2018-02/msg06799.html I guess it remains to be seen in what order these patches are merged. > > Signed-off-by: Alberto Garcia <berto@igalia.com> > --- > block/qcow2-cluster.c | 20 +++++++++++++------- > tests/qemu-iotests/080 | 2 ++ > tests/qemu-iotests/080.out | 2 ++ > 3 files changed, 17 insertions(+), 7 deletions(-) Reviewed-by: Eric Blake <eblake@redhat.com> -- Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3266 Virtualization: qemu.org | libvirt.org
Am 01.03.2018 um 17:27 hat Alberto Garcia geschrieben:
> This function iterates over all snapshots of a qcow2 file in order to
> expand all zero clusters, but it does not validate the snapshots' L1
> tables first.
>
> We now have a function to take care of this, so let's use it.
>
> We can also take the opportunity to replace the sector-based
> bdrv_read() with bdrv_pread().
>
> Signed-off-by: Alberto Garcia <berto@igalia.com>
> ---
> block/qcow2-cluster.c | 20 +++++++++++++-------
> tests/qemu-iotests/080 | 2 ++
> tests/qemu-iotests/080.out | 2 ++
> 3 files changed, 17 insertions(+), 7 deletions(-)
>
> diff --git a/block/qcow2-cluster.c b/block/qcow2-cluster.c
> index e406b0f3b9..40167ac09c 100644
> --- a/block/qcow2-cluster.c
> +++ b/block/qcow2-cluster.c
> @@ -2092,11 +2092,18 @@ int qcow2_expand_zero_clusters(BlockDriverState *bs,
> }
>
> for (i = 0; i < s->nb_snapshots; i++) {
> - int l1_sectors = DIV_ROUND_UP(s->snapshots[i].l1_size *
> - sizeof(uint64_t), BDRV_SECTOR_SIZE);
> + int l1_size2;
> + uint64_t *new_l1_table;
>
> - uint64_t *new_l1_table =
> - g_try_realloc(l1_table, l1_sectors * BDRV_SECTOR_SIZE);
> + ret = qcow2_validate_table(bs, s->snapshots[i].l1_table_offset,
> + s->snapshots[i].l1_size, sizeof(uint64_t),
> + QCOW_MAX_L1_SIZE, "", NULL);
> + if (ret < 0) {
> + return ret;
Shouldn't this be goto fail?
Kevin
On Tue 06 Mar 2018 03:54:26 PM CET, Kevin Wolf wrote:
>> @@ -2092,11 +2092,18 @@ int qcow2_expand_zero_clusters(BlockDriverState *bs,
>> }
>>
>> for (i = 0; i < s->nb_snapshots; i++) {
>> - int l1_sectors = DIV_ROUND_UP(s->snapshots[i].l1_size *
>> - sizeof(uint64_t), BDRV_SECTOR_SIZE);
>> + int l1_size2;
>> + uint64_t *new_l1_table;
>>
>> - uint64_t *new_l1_table =
>> - g_try_realloc(l1_table, l1_sectors * BDRV_SECTOR_SIZE);
>> + ret = qcow2_validate_table(bs, s->snapshots[i].l1_table_offset,
>> + s->snapshots[i].l1_size, sizeof(uint64_t),
>> + QCOW_MAX_L1_SIZE, "", NULL);
>> + if (ret < 0) {
>> + return ret;
>
> Shouldn't this be goto fail?
You're right, this is a loop, and l1_table could have been initialized
in previous iterations.
I'll send a corrected version with this change, but first I'll wait a
bit in case you see anything else in the series.
Berto
Am 06.03.2018 um 16:01 hat Alberto Garcia geschrieben:
> On Tue 06 Mar 2018 03:54:26 PM CET, Kevin Wolf wrote:
> >> @@ -2092,11 +2092,18 @@ int qcow2_expand_zero_clusters(BlockDriverState *bs,
> >> }
> >>
> >> for (i = 0; i < s->nb_snapshots; i++) {
> >> - int l1_sectors = DIV_ROUND_UP(s->snapshots[i].l1_size *
> >> - sizeof(uint64_t), BDRV_SECTOR_SIZE);
> >> + int l1_size2;
> >> + uint64_t *new_l1_table;
> >>
> >> - uint64_t *new_l1_table =
> >> - g_try_realloc(l1_table, l1_sectors * BDRV_SECTOR_SIZE);
> >> + ret = qcow2_validate_table(bs, s->snapshots[i].l1_table_offset,
> >> + s->snapshots[i].l1_size, sizeof(uint64_t),
> >> + QCOW_MAX_L1_SIZE, "", NULL);
> >> + if (ret < 0) {
> >> + return ret;
> >
> > Shouldn't this be goto fail?
>
> You're right, this is a loop, and l1_table could have been initialized
> in previous iterations.
>
> I'll send a corrected version with this change, but first I'll wait a
> bit in case you see anything else in the series.
I've finished the review now, the rest looks correct.
The only other thing I wondered is about the cases where you pass a
NULL errp because the callers don't get an Error parameter, so they
can't pass it on. Some of these callers already use error_report(), so
it would be okay to use error_report_err() for an error returned by
qcow2_validate_table(), too. I think that would improve the messages.
Kevin
On Tue 06 Mar 2018 04:11:17 PM CET, Kevin Wolf wrote: > I've finished the review now, the rest looks correct. > > The only other thing I wondered is about the cases where you pass a > NULL errp because the callers don't get an Error parameter, so they > can't pass it on. Some of these callers already use error_report(), so > it would be okay to use error_report_err() for an error returned by > qcow2_validate_table(), too. I think that would improve the messages. Good idea, I'll change that and resend the series. Berto
© 2016 - 2026 Red Hat, Inc.