* [Qemu-devel] [PATCH v3 00/10] qcow2: Fix image repairing
@ 2014-08-22 16:31 Max Reitz
2014-08-22 16:31 ` [Qemu-devel] [PATCH v3 01/10] qcow2: Fix leaks in dirty images Max Reitz
` (9 more replies)
0 siblings, 10 replies; 28+ messages in thread
From: Max Reitz @ 2014-08-22 16:31 UTC (permalink / raw)
To: qemu-devel; +Cc: Kevin Wolf, Benoît Canet, Stefan Hajnoczi, Max Reitz
As can be seen in the final patch of this series, there are certain
cases where the current repair implementation of qcow2 actually damages
the image further because it allocates new clusters for the refcount
structure which overlap with existing but according to the on-disk
refcounts (which are assumed to be wrong to begin with) unallocated
clusters.
This series fixes this by completely recreating the refcount structure
based on the in-memory information calculated during the check operation
if the possibility of damaging the image while repairing the refcount
structures in-place exists.
v3:
- Reorganised patch 2 and 3 for a smaller diff: [Benoît]
- Squashed 3 into 2 (makes patch 2 longer but avoids needless diff
blob resulting from the beginning of qcow2_check_refcounts())
- check_refblocks() is below calculate_refcounts() after patch 2, this
minimizes diff for patch 2; patch 3 pulls it up again
- Made a small functional difference from patch 2 into an own patch
(patch 4)
- Patch 2 is now mostly only code movement with additionally adding
dereferenciation to refcount_table and nb_clusters in
check_refblocks() and calculate_refcounts(), because those
parameters are (and must be) given by reference there; the same
applies to highest_cluster in compare_refcounts()
- Patch 6 (prev. 5): Removed a superfluous empty line and rebased on the
new patch 2 which no longer changes the g_realloc() (or g_renew() by
now) call to g_try_realloc() because this call is removed in this
patch anyway
- Patch 10 (prev. 9): Renamed test file from 101 to 104
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
001/10:[----] [--] 'qcow2: Fix leaks in dirty images'
002/10:[down] 'qcow2: Split qcow2_check_refcounts()'
003/10:[down] 'qcow2: Pull check_refblocks() up'
004/10:[down] 'qcow2: Reuse refcount table in calculate_refcounts()'
005/10:[----] [--] 'qcow2: Fix refcount blocks beyond image end'
006/10:[0013] [FC] 'qcow2: Do not perform potentially damaging repairs'
007/10:[----] [--] 'qcow2: Rebuild refcount structure during check'
008/10:[----] [--] 'qcow2: Clean up after refcount rebuild'
009/10:[----] [--] 'iotests: Fix test outputs'
010/10:[0004] [FC] 'iotests: Add test for potentially damaging repairs'
Max Reitz (10):
qcow2: Fix leaks in dirty images
qcow2: Split qcow2_check_refcounts()
qcow2: Pull check_refblocks() up
qcow2: Reuse refcount table in calculate_refcounts()
qcow2: Fix refcount blocks beyond image end
qcow2: Do not perform potentially damaging repairs
qcow2: Rebuild refcount structure during check
qcow2: Clean up after refcount rebuild
iotests: Fix test outputs
iotests: Add test for potentially damaging repairs
block/qcow2-refcount.c | 644 +++++++++++++++++++++++++++++++--------------
block/qcow2.c | 2 +-
tests/qemu-iotests/039.out | 10 +-
tests/qemu-iotests/060.out | 10 +-
tests/qemu-iotests/061.out | 18 +-
tests/qemu-iotests/104 | 98 +++++++
tests/qemu-iotests/104.out | 46 ++++
tests/qemu-iotests/group | 1 +
8 files changed, 623 insertions(+), 206 deletions(-)
create mode 100755 tests/qemu-iotests/104
create mode 100644 tests/qemu-iotests/104.out
--
2.0.4
^ permalink raw reply [flat|nested] 28+ messages in thread
* [Qemu-devel] [PATCH v3 01/10] qcow2: Fix leaks in dirty images
2014-08-22 16:31 [Qemu-devel] [PATCH v3 00/10] qcow2: Fix image repairing Max Reitz
@ 2014-08-22 16:31 ` Max Reitz
2014-08-22 18:44 ` Benoît Canet
2014-08-22 16:31 ` [Qemu-devel] [PATCH v3 02/10] qcow2: Split qcow2_check_refcounts() Max Reitz
` (8 subsequent siblings)
9 siblings, 1 reply; 28+ messages in thread
From: Max Reitz @ 2014-08-22 16:31 UTC (permalink / raw)
To: qemu-devel; +Cc: Kevin Wolf, Benoît Canet, Stefan Hajnoczi, Max Reitz
When opening dirty images, qcow2's repair function should not only
repair errors but leaks as well.
Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
---
block/qcow2.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/block/qcow2.c b/block/qcow2.c
index f9e045f..67d6368 100644
--- a/block/qcow2.c
+++ b/block/qcow2.c
@@ -898,7 +898,7 @@ static int qcow2_open(BlockDriverState *bs, QDict *options, int flags,
(s->incompatible_features & QCOW2_INCOMPAT_DIRTY)) {
BdrvCheckResult result = {0};
- ret = qcow2_check(bs, &result, BDRV_FIX_ERRORS);
+ ret = qcow2_check(bs, &result, BDRV_FIX_ERRORS | BDRV_FIX_LEAKS);
if (ret < 0) {
error_setg_errno(errp, -ret, "Could not repair dirty image");
goto fail;
--
2.0.4
^ permalink raw reply related [flat|nested] 28+ messages in thread
* [Qemu-devel] [PATCH v3 02/10] qcow2: Split qcow2_check_refcounts()
2014-08-22 16:31 [Qemu-devel] [PATCH v3 00/10] qcow2: Fix image repairing Max Reitz
2014-08-22 16:31 ` [Qemu-devel] [PATCH v3 01/10] qcow2: Fix leaks in dirty images Max Reitz
@ 2014-08-22 16:31 ` Max Reitz
2014-08-22 18:03 ` Benoît Canet
2014-08-22 16:31 ` [Qemu-devel] [PATCH v3 03/10] qcow2: Pull check_refblocks() up Max Reitz
` (7 subsequent siblings)
9 siblings, 1 reply; 28+ messages in thread
From: Max Reitz @ 2014-08-22 16:31 UTC (permalink / raw)
To: qemu-devel; +Cc: Kevin Wolf, Benoît Canet, Stefan Hajnoczi, Max Reitz
Put the code for calculating the reference counts and comparing them
during qemu-img check into own functions.
Signed-off-by: Max Reitz <mreitz@redhat.com>
---
block/qcow2-refcount.c | 153 ++++++++++++++++++++++++++++++++-----------------
1 file changed, 102 insertions(+), 51 deletions(-)
diff --git a/block/qcow2-refcount.c b/block/qcow2-refcount.c
index 43665b8..5f0920b 100644
--- a/block/qcow2-refcount.c
+++ b/block/qcow2-refcount.c
@@ -1496,71 +1496,70 @@ done:
return new_offset;
}
+static int check_refblocks(BlockDriverState *bs, BdrvCheckResult *res,
+ BdrvCheckMode fix, uint16_t **refcount_table,
+ int64_t *nb_clusters);
+
/*
- * Checks an image for refcount consistency.
- *
- * Returns 0 if no errors are found, the number of errors in case the image is
- * detected as corrupted, and -errno when an internal error occurred.
+ * Calculates an in-memory refcount table.
*/
-int qcow2_check_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
- BdrvCheckMode fix)
+static int calculate_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
+ BdrvCheckMode fix, uint16_t **refcount_table,
+ int64_t *nb_clusters)
{
BDRVQcowState *s = bs->opaque;
- int64_t size, i, highest_cluster, nb_clusters;
- int refcount1, refcount2;
+ int64_t i;
QCowSnapshot *sn;
- uint16_t *refcount_table;
int ret;
- size = bdrv_getlength(bs->file);
- if (size < 0) {
- res->check_errors++;
- return size;
- }
-
- nb_clusters = size_to_clusters(s, size);
- if (nb_clusters > INT_MAX) {
- res->check_errors++;
- return -EFBIG;
- }
-
- refcount_table = g_try_new0(uint16_t, nb_clusters);
- if (nb_clusters && refcount_table == NULL) {
+ *refcount_table = g_try_new0(uint16_t, *nb_clusters);
+ if (*nb_clusters && *refcount_table == NULL) {
res->check_errors++;
return -ENOMEM;
}
- res->bfi.total_clusters =
- size_to_clusters(s, bs->total_sectors * BDRV_SECTOR_SIZE);
-
/* header */
- inc_refcounts(bs, res, refcount_table, nb_clusters,
+ inc_refcounts(bs, res, *refcount_table, *nb_clusters,
0, s->cluster_size);
/* current L1 table */
- ret = check_refcounts_l1(bs, res, refcount_table, nb_clusters,
+ ret = check_refcounts_l1(bs, res, *refcount_table, *nb_clusters,
s->l1_table_offset, s->l1_size, CHECK_FRAG_INFO);
if (ret < 0) {
- goto fail;
+ return ret;
}
/* snapshots */
- for(i = 0; i < s->nb_snapshots; i++) {
+ for (i = 0; i < s->nb_snapshots; i++) {
sn = s->snapshots + i;
- ret = check_refcounts_l1(bs, res, refcount_table, nb_clusters,
+ ret = check_refcounts_l1(bs, res, *refcount_table, *nb_clusters,
sn->l1_table_offset, sn->l1_size, 0);
if (ret < 0) {
- goto fail;
+ return ret;
}
}
- inc_refcounts(bs, res, refcount_table, nb_clusters,
+ inc_refcounts(bs, res, *refcount_table, *nb_clusters,
s->snapshots_offset, s->snapshots_size);
/* refcount data */
- inc_refcounts(bs, res, refcount_table, nb_clusters,
+ inc_refcounts(bs, res, *refcount_table, *nb_clusters,
s->refcount_table_offset,
s->refcount_table_size * sizeof(uint64_t));
+ return check_refblocks(bs, res, fix, refcount_table, nb_clusters);
+}
+
+/*
+ * Checks consistency of refblocks and accounts for each refblock in
+ * *refcount_table.
+ */
+static int check_refblocks(BlockDriverState *bs, BdrvCheckResult *res,
+ BdrvCheckMode fix, uint16_t **refcount_table,
+ int64_t *nb_clusters)
+{
+ BDRVQcowState *s = bs->opaque;
+ int64_t i;
+
for(i = 0; i < s->refcount_table_size; i++) {
uint64_t offset, cluster;
offset = s->refcount_table[i];
@@ -1574,7 +1573,7 @@ int qcow2_check_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
continue;
}
- if (cluster >= nb_clusters) {
+ if (cluster >= *nb_clusters) {
fprintf(stderr, "ERROR refcount block %" PRId64
" is outside image\n", i);
res->corruptions++;
@@ -1582,14 +1581,14 @@ int qcow2_check_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
}
if (offset != 0) {
- inc_refcounts(bs, res, refcount_table, nb_clusters,
+ inc_refcounts(bs, res, *refcount_table, *nb_clusters,
offset, s->cluster_size);
- if (refcount_table[cluster] != 1) {
+ if ((*refcount_table)[cluster] != 1) {
fprintf(stderr, "%s refcount block %" PRId64
" refcount=%d\n",
fix & BDRV_FIX_ERRORS ? "Repairing" :
"ERROR",
- i, refcount_table[cluster]);
+ i, (*refcount_table)[cluster]);
if (fix & BDRV_FIX_ERRORS) {
int64_t new_offset;
@@ -1601,17 +1600,18 @@ int qcow2_check_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
}
/* update refcounts */
- if ((new_offset >> s->cluster_bits) >= nb_clusters) {
+ if ((new_offset >> s->cluster_bits) >= *nb_clusters) {
/* increase refcount_table size if necessary */
- int old_nb_clusters = nb_clusters;
- nb_clusters = (new_offset >> s->cluster_bits) + 1;
- refcount_table = g_renew(uint16_t, refcount_table,
- nb_clusters);
- memset(&refcount_table[old_nb_clusters], 0, (nb_clusters
- - old_nb_clusters) * sizeof(uint16_t));
+ int old_nb_clusters = *nb_clusters;
+ *nb_clusters = (new_offset >> s->cluster_bits) + 1;
+ *refcount_table = g_renew(uint16_t, *refcount_table,
+ *nb_clusters);
+ memset(&(*refcount_table)[old_nb_clusters], 0,
+ (*nb_clusters - old_nb_clusters) *
+ sizeof(uint16_t));
}
- refcount_table[cluster]--;
- inc_refcounts(bs, res, refcount_table, nb_clusters,
+ (*refcount_table)[cluster]--;
+ inc_refcounts(bs, res, *refcount_table, *nb_clusters,
new_offset, s->cluster_size);
res->corruptions_fixed++;
@@ -1622,8 +1622,22 @@ int qcow2_check_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
}
}
- /* compare ref counts */
- for (i = 0, highest_cluster = 0; i < nb_clusters; i++) {
+ return 0;
+}
+
+/*
+ * Compares the actual reference count for each cluster in the image against the
+ * refcount as reported by the refcount structures on-disk.
+ */
+static void compare_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
+ BdrvCheckMode fix, int64_t *highest_cluster,
+ uint16_t *refcount_table, int64_t nb_clusters)
+{
+ BDRVQcowState *s = bs->opaque;
+ int64_t i;
+ int refcount1, refcount2, ret;
+
+ for (i = 0, *highest_cluster = 0; i < nb_clusters; i++) {
refcount1 = get_refcount(bs, i);
if (refcount1 < 0) {
fprintf(stderr, "Can't get refcount for cluster %" PRId64 ": %s\n",
@@ -1635,11 +1649,10 @@ int qcow2_check_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
refcount2 = refcount_table[i];
if (refcount1 > 0 || refcount2 > 0) {
- highest_cluster = i;
+ *highest_cluster = i;
}
if (refcount1 != refcount2) {
-
/* Check if we're allowed to fix the mismatch */
int *num_fixed = NULL;
if (refcount1 > refcount2 && (fix & BDRV_FIX_LEAKS)) {
@@ -1672,6 +1685,44 @@ int qcow2_check_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
}
}
}
+}
+
+/*
+ * Checks an image for refcount consistency.
+ *
+ * Returns 0 if no errors are found, the number of errors in case the image is
+ * detected as corrupted, and -errno when an internal error occurred.
+ */
+int qcow2_check_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
+ BdrvCheckMode fix)
+{
+ BDRVQcowState *s = bs->opaque;
+ int64_t size, highest_cluster, nb_clusters;
+ uint16_t *refcount_table;
+ int ret;
+
+ size = bdrv_getlength(bs->file);
+ if (size < 0) {
+ res->check_errors++;
+ return size;
+ }
+
+ nb_clusters = size_to_clusters(s, size);
+ if (nb_clusters > INT_MAX) {
+ res->check_errors++;
+ return -EFBIG;
+ }
+
+ res->bfi.total_clusters =
+ size_to_clusters(s, bs->total_sectors * BDRV_SECTOR_SIZE);
+
+ ret = calculate_refcounts(bs, res, fix, &refcount_table, &nb_clusters);
+ if (ret < 0) {
+ goto fail;
+ }
+
+ compare_refcounts(bs, res, fix, &highest_cluster, refcount_table,
+ nb_clusters);
/* check OFLAG_COPIED */
ret = check_oflag_copied(bs, res, fix);
--
2.0.4
^ permalink raw reply related [flat|nested] 28+ messages in thread
* [Qemu-devel] [PATCH v3 03/10] qcow2: Pull check_refblocks() up
2014-08-22 16:31 [Qemu-devel] [PATCH v3 00/10] qcow2: Fix image repairing Max Reitz
2014-08-22 16:31 ` [Qemu-devel] [PATCH v3 01/10] qcow2: Fix leaks in dirty images Max Reitz
2014-08-22 16:31 ` [Qemu-devel] [PATCH v3 02/10] qcow2: Split qcow2_check_refcounts() Max Reitz
@ 2014-08-22 16:31 ` Max Reitz
2014-08-22 18:04 ` Benoît Canet
2014-08-22 16:31 ` [Qemu-devel] [PATCH v3 04/10] qcow2: Reuse refcount table in calculate_refcounts() Max Reitz
` (6 subsequent siblings)
9 siblings, 1 reply; 28+ messages in thread
From: Max Reitz @ 2014-08-22 16:31 UTC (permalink / raw)
To: qemu-devel; +Cc: Kevin Wolf, Benoît Canet, Stefan Hajnoczi, Max Reitz
Pull check_refblocks() before calculate_refcounts() so we can drop its
static declaration.
Signed-off-by: Max Reitz <mreitz@redhat.com>
---
block/qcow2-refcount.c | 102 ++++++++++++++++++++++++-------------------------
1 file changed, 49 insertions(+), 53 deletions(-)
diff --git a/block/qcow2-refcount.c b/block/qcow2-refcount.c
index 5f0920b..2b728ef 100644
--- a/block/qcow2-refcount.c
+++ b/block/qcow2-refcount.c
@@ -1496,59 +1496,6 @@ done:
return new_offset;
}
-static int check_refblocks(BlockDriverState *bs, BdrvCheckResult *res,
- BdrvCheckMode fix, uint16_t **refcount_table,
- int64_t *nb_clusters);
-
-/*
- * Calculates an in-memory refcount table.
- */
-static int calculate_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
- BdrvCheckMode fix, uint16_t **refcount_table,
- int64_t *nb_clusters)
-{
- BDRVQcowState *s = bs->opaque;
- int64_t i;
- QCowSnapshot *sn;
- int ret;
-
- *refcount_table = g_try_new0(uint16_t, *nb_clusters);
- if (*nb_clusters && *refcount_table == NULL) {
- res->check_errors++;
- return -ENOMEM;
- }
-
- /* header */
- inc_refcounts(bs, res, *refcount_table, *nb_clusters,
- 0, s->cluster_size);
-
- /* current L1 table */
- ret = check_refcounts_l1(bs, res, *refcount_table, *nb_clusters,
- s->l1_table_offset, s->l1_size, CHECK_FRAG_INFO);
- if (ret < 0) {
- return ret;
- }
-
- /* snapshots */
- for (i = 0; i < s->nb_snapshots; i++) {
- sn = s->snapshots + i;
- ret = check_refcounts_l1(bs, res, *refcount_table, *nb_clusters,
- sn->l1_table_offset, sn->l1_size, 0);
- if (ret < 0) {
- return ret;
- }
- }
- inc_refcounts(bs, res, *refcount_table, *nb_clusters,
- s->snapshots_offset, s->snapshots_size);
-
- /* refcount data */
- inc_refcounts(bs, res, *refcount_table, *nb_clusters,
- s->refcount_table_offset,
- s->refcount_table_size * sizeof(uint64_t));
-
- return check_refblocks(bs, res, fix, refcount_table, nb_clusters);
-}
-
/*
* Checks consistency of refblocks and accounts for each refblock in
* *refcount_table.
@@ -1626,6 +1573,55 @@ static int check_refblocks(BlockDriverState *bs, BdrvCheckResult *res,
}
/*
+ * Calculates an in-memory refcount table.
+ */
+static int calculate_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
+ BdrvCheckMode fix, uint16_t **refcount_table,
+ int64_t *nb_clusters)
+{
+ BDRVQcowState *s = bs->opaque;
+ int64_t i;
+ QCowSnapshot *sn;
+ int ret;
+
+ *refcount_table = g_try_new0(uint16_t, *nb_clusters);
+ if (*nb_clusters && *refcount_table == NULL) {
+ res->check_errors++;
+ return -ENOMEM;
+ }
+
+ /* header */
+ inc_refcounts(bs, res, *refcount_table, *nb_clusters,
+ 0, s->cluster_size);
+
+ /* current L1 table */
+ ret = check_refcounts_l1(bs, res, *refcount_table, *nb_clusters,
+ s->l1_table_offset, s->l1_size, CHECK_FRAG_INFO);
+ if (ret < 0) {
+ return ret;
+ }
+
+ /* snapshots */
+ for (i = 0; i < s->nb_snapshots; i++) {
+ sn = s->snapshots + i;
+ ret = check_refcounts_l1(bs, res, *refcount_table, *nb_clusters,
+ sn->l1_table_offset, sn->l1_size, 0);
+ if (ret < 0) {
+ return ret;
+ }
+ }
+ inc_refcounts(bs, res, *refcount_table, *nb_clusters,
+ s->snapshots_offset, s->snapshots_size);
+
+ /* refcount data */
+ inc_refcounts(bs, res, *refcount_table, *nb_clusters,
+ s->refcount_table_offset,
+ s->refcount_table_size * sizeof(uint64_t));
+
+ return check_refblocks(bs, res, fix, refcount_table, nb_clusters);
+}
+
+/*
* Compares the actual reference count for each cluster in the image against the
* refcount as reported by the refcount structures on-disk.
*/
--
2.0.4
^ permalink raw reply related [flat|nested] 28+ messages in thread
* [Qemu-devel] [PATCH v3 04/10] qcow2: Reuse refcount table in calculate_refcounts()
2014-08-22 16:31 [Qemu-devel] [PATCH v3 00/10] qcow2: Fix image repairing Max Reitz
` (2 preceding siblings ...)
2014-08-22 16:31 ` [Qemu-devel] [PATCH v3 03/10] qcow2: Pull check_refblocks() up Max Reitz
@ 2014-08-22 16:31 ` Max Reitz
2014-08-22 18:07 ` Benoît Canet
2014-08-22 16:31 ` [Qemu-devel] [PATCH v3 05/10] qcow2: Fix refcount blocks beyond image end Max Reitz
` (5 subsequent siblings)
9 siblings, 1 reply; 28+ messages in thread
From: Max Reitz @ 2014-08-22 16:31 UTC (permalink / raw)
To: qemu-devel; +Cc: Kevin Wolf, Benoît Canet, Stefan Hajnoczi, Max Reitz
We will later call calculate_refcounts multiple times, so reuse the
refcount table if possible.
Signed-off-by: Max Reitz <mreitz@redhat.com>
---
block/qcow2-refcount.c | 12 +++++++-----
1 file changed, 7 insertions(+), 5 deletions(-)
diff --git a/block/qcow2-refcount.c b/block/qcow2-refcount.c
index 2b728ef..babe6cb 100644
--- a/block/qcow2-refcount.c
+++ b/block/qcow2-refcount.c
@@ -1584,10 +1584,12 @@ static int calculate_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
QCowSnapshot *sn;
int ret;
- *refcount_table = g_try_new0(uint16_t, *nb_clusters);
- if (*nb_clusters && *refcount_table == NULL) {
- res->check_errors++;
- return -ENOMEM;
+ if (!*refcount_table) {
+ *refcount_table = g_try_new0(uint16_t, *nb_clusters);
+ if (*nb_clusters && *refcount_table == NULL) {
+ res->check_errors++;
+ return -ENOMEM;
+ }
}
/* header */
@@ -1694,7 +1696,7 @@ int qcow2_check_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
{
BDRVQcowState *s = bs->opaque;
int64_t size, highest_cluster, nb_clusters;
- uint16_t *refcount_table;
+ uint16_t *refcount_table = NULL;
int ret;
size = bdrv_getlength(bs->file);
--
2.0.4
^ permalink raw reply related [flat|nested] 28+ messages in thread
* [Qemu-devel] [PATCH v3 05/10] qcow2: Fix refcount blocks beyond image end
2014-08-22 16:31 [Qemu-devel] [PATCH v3 00/10] qcow2: Fix image repairing Max Reitz
` (3 preceding siblings ...)
2014-08-22 16:31 ` [Qemu-devel] [PATCH v3 04/10] qcow2: Reuse refcount table in calculate_refcounts() Max Reitz
@ 2014-08-22 16:31 ` Max Reitz
2014-08-22 18:20 ` Benoît Canet
2014-08-26 13:07 ` Eric Blake
2014-08-22 16:31 ` [Qemu-devel] [PATCH v3 06/10] qcow2: Do not perform potentially damaging repairs Max Reitz
` (4 subsequent siblings)
9 siblings, 2 replies; 28+ messages in thread
From: Max Reitz @ 2014-08-22 16:31 UTC (permalink / raw)
To: qemu-devel; +Cc: Kevin Wolf, Benoît Canet, Stefan Hajnoczi, Max Reitz
If the qcow2 check function detects a refcount block located beyond the
image end, grow the image appropriately. This cannot break anything and
is the logical fix for such a case.
Signed-off-by: Max Reitz <mreitz@redhat.com>
---
block/qcow2-refcount.c | 50 ++++++++++++++++++++++++++++++++++++++++++++++----
1 file changed, 46 insertions(+), 4 deletions(-)
diff --git a/block/qcow2-refcount.c b/block/qcow2-refcount.c
index babe6cb..1f0f44e 100644
--- a/block/qcow2-refcount.c
+++ b/block/qcow2-refcount.c
@@ -1505,7 +1505,8 @@ static int check_refblocks(BlockDriverState *bs, BdrvCheckResult *res,
int64_t *nb_clusters)
{
BDRVQcowState *s = bs->opaque;
- int64_t i;
+ int64_t i, size;
+ int ret;
for(i = 0; i < s->refcount_table_size; i++) {
uint64_t offset, cluster;
@@ -1521,9 +1522,50 @@ static int check_refblocks(BlockDriverState *bs, BdrvCheckResult *res,
}
if (cluster >= *nb_clusters) {
- fprintf(stderr, "ERROR refcount block %" PRId64
- " is outside image\n", i);
- res->corruptions++;
+ fprintf(stderr, "%s refcount block %" PRId64 " is outside image\n",
+ fix & BDRV_FIX_ERRORS ? "Repairing" : "ERROR", i);
+
+ if (fix & BDRV_FIX_ERRORS) {
+ int64_t old_nb_clusters = *nb_clusters;
+
+ ret = bdrv_truncate(bs->file, offset + s->cluster_size);
+ if (ret < 0) {
+ goto resize_fail;
+ }
+ size = bdrv_getlength(bs->file);
+ if (size < 0) {
+ ret = size;
+ goto resize_fail;
+ }
+
+ *nb_clusters = size_to_clusters(s, size);
+ assert(*nb_clusters >= old_nb_clusters);
+
+ *refcount_table = g_try_realloc(*refcount_table,
+ *nb_clusters * sizeof(uint16_t));
+ if (!*refcount_table) {
+ res->check_errors++;
+ return -ENOMEM;
+ }
+
+ memset(*refcount_table + old_nb_clusters, 0,
+ (*nb_clusters - old_nb_clusters) * sizeof(uint16_t));
+
+ if (cluster >= *nb_clusters) {
+ ret = -EINVAL;
+ goto resize_fail;
+ }
+
+ res->corruptions_fixed++;
+ continue;
+
+resize_fail:
+ res->corruptions++;
+ fprintf(stderr, "ERROR could not resize image: %s\n",
+ strerror(-ret));
+ } else {
+ res->corruptions++;
+ }
continue;
}
--
2.0.4
^ permalink raw reply related [flat|nested] 28+ messages in thread
* [Qemu-devel] [PATCH v3 06/10] qcow2: Do not perform potentially damaging repairs
2014-08-22 16:31 [Qemu-devel] [PATCH v3 00/10] qcow2: Fix image repairing Max Reitz
` (4 preceding siblings ...)
2014-08-22 16:31 ` [Qemu-devel] [PATCH v3 05/10] qcow2: Fix refcount blocks beyond image end Max Reitz
@ 2014-08-22 16:31 ` Max Reitz
2014-08-22 18:25 ` Benoît Canet
2014-08-22 16:31 ` [Qemu-devel] [PATCH v3 07/10] qcow2: Rebuild refcount structure during check Max Reitz
` (3 subsequent siblings)
9 siblings, 1 reply; 28+ messages in thread
From: Max Reitz @ 2014-08-22 16:31 UTC (permalink / raw)
To: qemu-devel; +Cc: Kevin Wolf, Benoît Canet, Stefan Hajnoczi, Max Reitz
If a referenced cluster has a refcount of 0, increasing its refcount may
result in clusters being allocated for the refcount structures. This may
overwrite the referenced cluster, therefore we cannot simply increase
the refcount then.
In such cases, we can either try to replicate all the refcount
operations solely for the check operation, basing the allocations on the
in-memory refcount table; or we can simply rebuild the whole refcount
structure based on the in-memory refcount table. Since the latter will
be much easier, do that.
To prepare for this, introduce a "rebuild" boolean which should be set
to true whenever a fix is rather dangerous or too complicated using the
current refcount structures. Another example for this is refcount blocks
being referenced more than once.
Signed-off-by: Max Reitz <mreitz@redhat.com>
---
block/qcow2-refcount.c | 185 ++++++++-----------------------------------------
1 file changed, 27 insertions(+), 158 deletions(-)
diff --git a/block/qcow2-refcount.c b/block/qcow2-refcount.c
index 1f0f44e..242a20c 100644
--- a/block/qcow2-refcount.c
+++ b/block/qcow2-refcount.c
@@ -1382,127 +1382,12 @@ fail:
}
/*
- * Writes one sector of the refcount table to the disk
- */
-#define RT_ENTRIES_PER_SECTOR (512 / sizeof(uint64_t))
-static int write_reftable_entry(BlockDriverState *bs, int rt_index)
-{
- BDRVQcowState *s = bs->opaque;
- uint64_t buf[RT_ENTRIES_PER_SECTOR];
- int rt_start_index;
- int i, ret;
-
- rt_start_index = rt_index & ~(RT_ENTRIES_PER_SECTOR - 1);
- for (i = 0; i < RT_ENTRIES_PER_SECTOR; i++) {
- buf[i] = cpu_to_be64(s->refcount_table[rt_start_index + i]);
- }
-
- ret = qcow2_pre_write_overlap_check(bs, QCOW2_OL_REFCOUNT_TABLE,
- s->refcount_table_offset + rt_start_index * sizeof(uint64_t),
- sizeof(buf));
- if (ret < 0) {
- return ret;
- }
-
- BLKDBG_EVENT(bs->file, BLKDBG_REFTABLE_UPDATE);
- ret = bdrv_pwrite_sync(bs->file, s->refcount_table_offset +
- rt_start_index * sizeof(uint64_t), buf, sizeof(buf));
- if (ret < 0) {
- return ret;
- }
-
- return 0;
-}
-
-/*
- * Allocates a new cluster for the given refcount block (represented by its
- * offset in the image file) and copies the current content there. This function
- * does _not_ decrement the reference count for the currently occupied cluster.
- *
- * This function prints an informative message to stderr on error (and returns
- * -errno); on success, the offset of the newly allocated cluster is returned.
- */
-static int64_t realloc_refcount_block(BlockDriverState *bs, int reftable_index,
- uint64_t offset)
-{
- BDRVQcowState *s = bs->opaque;
- int64_t new_offset = 0;
- void *refcount_block = NULL;
- int ret;
-
- /* allocate new refcount block */
- new_offset = qcow2_alloc_clusters(bs, s->cluster_size);
- if (new_offset < 0) {
- fprintf(stderr, "Could not allocate new cluster: %s\n",
- strerror(-new_offset));
- ret = new_offset;
- goto done;
- }
-
- /* fetch current refcount block content */
- ret = qcow2_cache_get(bs, s->refcount_block_cache, offset, &refcount_block);
- if (ret < 0) {
- fprintf(stderr, "Could not fetch refcount block: %s\n", strerror(-ret));
- goto fail_free_cluster;
- }
-
- /* new block has not yet been entered into refcount table, therefore it is
- * no refcount block yet (regarding this check) */
- ret = qcow2_pre_write_overlap_check(bs, 0, new_offset, s->cluster_size);
- if (ret < 0) {
- fprintf(stderr, "Could not write refcount block; metadata overlap "
- "check failed: %s\n", strerror(-ret));
- /* the image will be marked corrupt, so don't even attempt on freeing
- * the cluster */
- goto done;
- }
-
- /* write to new block */
- ret = bdrv_write(bs->file, new_offset / BDRV_SECTOR_SIZE, refcount_block,
- s->cluster_sectors);
- if (ret < 0) {
- fprintf(stderr, "Could not write refcount block: %s\n", strerror(-ret));
- goto fail_free_cluster;
- }
-
- /* update refcount table */
- assert(!offset_into_cluster(s, new_offset));
- s->refcount_table[reftable_index] = new_offset;
- ret = write_reftable_entry(bs, reftable_index);
- if (ret < 0) {
- fprintf(stderr, "Could not update refcount table: %s\n",
- strerror(-ret));
- goto fail_free_cluster;
- }
-
- goto done;
-
-fail_free_cluster:
- qcow2_free_clusters(bs, new_offset, s->cluster_size, QCOW2_DISCARD_OTHER);
-
-done:
- if (refcount_block) {
- /* This should never fail, as it would only do so if the given refcount
- * block cannot be found in the cache. As this is impossible as long as
- * there are no bugs, assert the success. */
- int tmp = qcow2_cache_put(bs, s->refcount_block_cache, &refcount_block);
- assert(tmp == 0);
- }
-
- if (ret < 0) {
- return ret;
- }
-
- return new_offset;
-}
-
-/*
* Checks consistency of refblocks and accounts for each refblock in
* *refcount_table.
*/
static int check_refblocks(BlockDriverState *bs, BdrvCheckResult *res,
- BdrvCheckMode fix, uint16_t **refcount_table,
- int64_t *nb_clusters)
+ BdrvCheckMode fix, bool *rebuild,
+ uint16_t **refcount_table, int64_t *nb_clusters)
{
BDRVQcowState *s = bs->opaque;
int64_t i, size;
@@ -1518,6 +1403,7 @@ static int check_refblocks(BlockDriverState *bs, BdrvCheckResult *res,
fprintf(stderr, "ERROR refcount block %" PRId64 " is not "
"cluster aligned; refcount table entry corrupted\n", i);
res->corruptions++;
+ *rebuild = true;
continue;
}
@@ -1561,6 +1447,7 @@ static int check_refblocks(BlockDriverState *bs, BdrvCheckResult *res,
resize_fail:
res->corruptions++;
+ *rebuild = true;
fprintf(stderr, "ERROR could not resize image: %s\n",
strerror(-ret));
} else {
@@ -1573,40 +1460,10 @@ resize_fail:
inc_refcounts(bs, res, *refcount_table, *nb_clusters,
offset, s->cluster_size);
if ((*refcount_table)[cluster] != 1) {
- fprintf(stderr, "%s refcount block %" PRId64
- " refcount=%d\n",
- fix & BDRV_FIX_ERRORS ? "Repairing" :
- "ERROR",
- i, (*refcount_table)[cluster]);
-
- if (fix & BDRV_FIX_ERRORS) {
- int64_t new_offset;
-
- new_offset = realloc_refcount_block(bs, i, offset);
- if (new_offset < 0) {
- res->corruptions++;
- continue;
- }
-
- /* update refcounts */
- if ((new_offset >> s->cluster_bits) >= *nb_clusters) {
- /* increase refcount_table size if necessary */
- int old_nb_clusters = *nb_clusters;
- *nb_clusters = (new_offset >> s->cluster_bits) + 1;
- *refcount_table = g_renew(uint16_t, *refcount_table,
- *nb_clusters);
- memset(&(*refcount_table)[old_nb_clusters], 0,
- (*nb_clusters - old_nb_clusters) *
- sizeof(uint16_t));
- }
- (*refcount_table)[cluster]--;
- inc_refcounts(bs, res, *refcount_table, *nb_clusters,
- new_offset, s->cluster_size);
-
- res->corruptions_fixed++;
- } else {
- res->corruptions++;
- }
+ fprintf(stderr, "ERROR refcount block %" PRId64
+ " refcount=%d\n", i, (*refcount_table)[cluster]);
+ res->corruptions++;
+ *rebuild = true;
}
}
}
@@ -1618,8 +1475,8 @@ resize_fail:
* Calculates an in-memory refcount table.
*/
static int calculate_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
- BdrvCheckMode fix, uint16_t **refcount_table,
- int64_t *nb_clusters)
+ BdrvCheckMode fix, bool *rebuild,
+ uint16_t **refcount_table, int64_t *nb_clusters)
{
BDRVQcowState *s = bs->opaque;
int64_t i;
@@ -1662,7 +1519,7 @@ static int calculate_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
s->refcount_table_offset,
s->refcount_table_size * sizeof(uint64_t));
- return check_refblocks(bs, res, fix, refcount_table, nb_clusters);
+ return check_refblocks(bs, res, fix, rebuild, refcount_table, nb_clusters);
}
/*
@@ -1670,7 +1527,8 @@ static int calculate_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
* refcount as reported by the refcount structures on-disk.
*/
static void compare_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
- BdrvCheckMode fix, int64_t *highest_cluster,
+ BdrvCheckMode fix, bool *rebuild,
+ int64_t *highest_cluster,
uint16_t *refcount_table, int64_t nb_clusters)
{
BDRVQcowState *s = bs->opaque;
@@ -1697,10 +1555,15 @@ static void compare_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
int *num_fixed = NULL;
if (refcount1 > refcount2 && (fix & BDRV_FIX_LEAKS)) {
num_fixed = &res->leaks_fixed;
- } else if (refcount1 < refcount2 && (fix & BDRV_FIX_ERRORS)) {
+ } else if (refcount1 > 0 && refcount1 < refcount2 &&
+ (fix & BDRV_FIX_ERRORS)) {
num_fixed = &res->corruptions_fixed;
}
+ if (refcount1 == 0) {
+ *rebuild = true;
+ }
+
fprintf(stderr, "%s cluster %" PRId64 " refcount=%d reference=%d\n",
num_fixed != NULL ? "Repairing" :
refcount1 < refcount2 ? "ERROR" :
@@ -1739,6 +1602,7 @@ int qcow2_check_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
BDRVQcowState *s = bs->opaque;
int64_t size, highest_cluster, nb_clusters;
uint16_t *refcount_table = NULL;
+ bool rebuild = false;
int ret;
size = bdrv_getlength(bs->file);
@@ -1756,14 +1620,19 @@ int qcow2_check_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
res->bfi.total_clusters =
size_to_clusters(s, bs->total_sectors * BDRV_SECTOR_SIZE);
- ret = calculate_refcounts(bs, res, fix, &refcount_table, &nb_clusters);
+ ret = calculate_refcounts(bs, res, fix, &rebuild, &refcount_table,
+ &nb_clusters);
if (ret < 0) {
goto fail;
}
- compare_refcounts(bs, res, fix, &highest_cluster, refcount_table,
+ compare_refcounts(bs, res, fix, &rebuild, &highest_cluster, refcount_table,
nb_clusters);
+ if (rebuild) {
+ fprintf(stderr, "ERROR need to rebuild refcount structures\n");
+ }
+
/* check OFLAG_COPIED */
ret = check_oflag_copied(bs, res, fix);
if (ret < 0) {
--
2.0.4
^ permalink raw reply related [flat|nested] 28+ messages in thread
* [Qemu-devel] [PATCH v3 07/10] qcow2: Rebuild refcount structure during check
2014-08-22 16:31 [Qemu-devel] [PATCH v3 00/10] qcow2: Fix image repairing Max Reitz
` (5 preceding siblings ...)
2014-08-22 16:31 ` [Qemu-devel] [PATCH v3 06/10] qcow2: Do not perform potentially damaging repairs Max Reitz
@ 2014-08-22 16:31 ` Max Reitz
2014-08-25 17:40 ` Benoît Canet
2014-08-22 16:31 ` [Qemu-devel] [PATCH v3 08/10] qcow2: Clean up after refcount rebuild Max Reitz
` (2 subsequent siblings)
9 siblings, 1 reply; 28+ messages in thread
From: Max Reitz @ 2014-08-22 16:31 UTC (permalink / raw)
To: qemu-devel; +Cc: Kevin Wolf, Benoît Canet, Stefan Hajnoczi, Max Reitz
The previous commit introduced the "rebuild" variable to qcow2's
implementation of the image consistency check. Now make use of this by
adding a function which creates a completely new refcount structure
based solely on the in-memory information gathered before.
The old refcount structure will be leaked, however.
Signed-off-by: Max Reitz <mreitz@redhat.com>
---
block/qcow2-refcount.c | 265 ++++++++++++++++++++++++++++++++++++++++++++++++-
1 file changed, 262 insertions(+), 3 deletions(-)
diff --git a/block/qcow2-refcount.c b/block/qcow2-refcount.c
index 242a20c..59cab65 100644
--- a/block/qcow2-refcount.c
+++ b/block/qcow2-refcount.c
@@ -1591,6 +1591,245 @@ static void compare_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
}
/*
+ * Allocates a cluster using an in-memory refcount table (IMRT) in contrast to
+ * the on-disk refcount structures.
+ *
+ * *first_free_cluster does not necessarily point to the first free cluster, but
+ * may point to one cluster as close as possible before it. The offset returned
+ * will never be before that cluster.
+ *
+ * Note that *first_free_cluster is a cluster index whereas the return value is
+ * an offset.
+ */
+static int64_t alloc_clusters_imrt(BlockDriverState *bs,
+ int cluster_count,
+ uint16_t **refcount_table,
+ int64_t *nb_clusters,
+ int64_t *first_free_cluster)
+{
+ BDRVQcowState *s = bs->opaque;
+ int64_t cluster = *first_free_cluster, i;
+ bool first_gap = true;
+ int contiguous_clusters;
+
+ for (contiguous_clusters = 0;
+ cluster < *nb_clusters && contiguous_clusters < cluster_count;
+ cluster++)
+ {
+ if (!(*refcount_table)[cluster]) {
+ contiguous_clusters++;
+ if (first_gap) {
+ *first_free_cluster = cluster;
+ first_gap = false;
+ }
+ } else if (contiguous_clusters) {
+ contiguous_clusters = 0;
+ }
+ }
+
+ if (contiguous_clusters < cluster_count) {
+ int64_t old_nb_clusters = *nb_clusters;
+
+ *nb_clusters = cluster + cluster_count - contiguous_clusters;
+ *refcount_table = g_try_realloc(*refcount_table,
+ *nb_clusters * sizeof(uint16_t));
+ if (!*refcount_table) {
+ return -ENOMEM;
+ }
+
+ memset(*refcount_table + old_nb_clusters, 0,
+ (*nb_clusters - old_nb_clusters) * sizeof(uint16_t));
+ }
+
+ cluster -= contiguous_clusters;
+ for (i = 0; i < cluster_count; i++) {
+ (*refcount_table)[cluster + i] = 1;
+ }
+
+ return cluster << s->cluster_bits;
+}
+
+/*
+ * Creates a new refcount structure based solely on the in-memory information
+ * given through *refcount_table. All necessary allocations will be reflected
+ * in that array.
+ *
+ * On success, the old refcount structure is leaked (it will be covered by the
+ * new refcount structure).
+ */
+static int rebuild_refcount_structure(BlockDriverState *bs,
+ BdrvCheckResult *res,
+ uint16_t **refcount_table,
+ int64_t *nb_clusters)
+{
+ BDRVQcowState *s = bs->opaque;
+ int64_t first_free_cluster = 0, rt_ofs = -1, cluster = 0;
+ int64_t rb_ofs, rb_start, rb_index;
+ uint32_t reftable_size = 0;
+ uint64_t *reftable = NULL;
+ uint16_t *on_disk_rb;
+ uint8_t rt_offset_and_clusters[sizeof(uint64_t) + sizeof(uint32_t)];
+ int i, ret = 0;
+
+ qcow2_cache_empty(bs, s->refcount_block_cache);
+
+write_refblocks:
+ for (; cluster < *nb_clusters; cluster++) {
+ if (!(*refcount_table)[cluster]) {
+ continue;
+ }
+
+ rb_index = cluster >> (s->cluster_bits - 1);
+ rb_start = rb_index << (s->cluster_bits - 1);
+
+ /* Don't allocate a cluster in a refblock already written to disk */
+ if (first_free_cluster < rb_start) {
+ first_free_cluster = rb_start;
+ }
+ rb_ofs = alloc_clusters_imrt(bs, 1, refcount_table, nb_clusters,
+ &first_free_cluster);
+ if (rb_ofs < 0) {
+ fprintf(stderr, "ERROR allocating refblock: %s\n", strerror(-ret));
+ res->check_errors++;
+ ret = rb_ofs;
+ goto fail;
+ }
+
+ if (reftable_size <= rb_index) {
+ uint32_t old_rt_size = reftable_size;
+ reftable_size = ROUND_UP((rb_index + 1) * sizeof(uint64_t),
+ s->cluster_size) / sizeof(uint64_t);
+ reftable = g_try_realloc(reftable,
+ reftable_size * sizeof(uint64_t));
+ if (!reftable) {
+ res->check_errors++;
+ ret = -ENOMEM;
+ goto fail;
+ }
+
+ memset(reftable + old_rt_size, 0,
+ (reftable_size - old_rt_size) * sizeof(uint64_t));
+
+ /* The offset we have for the reftable is now no longer valid;
+ * this will leak that range, but we can easily fix that by running
+ * a leak-fixing check after this rebuild operation */
+ rt_ofs = -1;
+ }
+ reftable[rb_index] = rb_ofs;
+
+ /* If this is apparently the last refblock (for now), try to squeeze the
+ * reftable in */
+ if (rb_index == (*nb_clusters - 1) >> (s->cluster_bits - 1) &&
+ rt_ofs < 0)
+ {
+ rt_ofs = alloc_clusters_imrt(bs, size_to_clusters(s, reftable_size *
+ sizeof(uint64_t)),
+ refcount_table, nb_clusters,
+ &first_free_cluster);
+ if (rt_ofs < 0) {
+ fprintf(stderr, "ERROR allocating reftable: %s\n",
+ strerror(-ret));
+ res->check_errors++;
+ ret = rt_ofs;
+ goto fail;
+ }
+ }
+
+ ret = qcow2_pre_write_overlap_check(bs, 0, rb_ofs, s->cluster_size);
+ if (ret < 0) {
+ fprintf(stderr, "ERROR writing refblock: %s\n", strerror(-ret));
+ goto fail;
+ }
+
+ on_disk_rb = g_malloc0(s->cluster_size);
+ for (i = 0; i < s->cluster_size / sizeof(uint16_t) &&
+ rb_start + i < *nb_clusters; i++)
+ {
+ on_disk_rb[i] = cpu_to_be16((*refcount_table)[rb_start + i]);
+ }
+
+ ret = bdrv_write(bs->file, rb_ofs / BDRV_SECTOR_SIZE,
+ (void *)on_disk_rb, s->cluster_sectors);
+ g_free(on_disk_rb);
+ if (ret < 0) {
+ fprintf(stderr, "ERROR writing refblock: %s\n", strerror(-ret));
+ goto fail;
+ }
+
+ /* Go to the end of this refblock */
+ cluster = rb_start + s->cluster_size / sizeof(uint16_t) - 1;
+ }
+
+ if (rt_ofs < 0) {
+ int64_t post_rb_start = ROUND_UP(*nb_clusters,
+ s->cluster_size / sizeof(uint16_t));
+
+ /* Not pretty but simple */
+ if (first_free_cluster < post_rb_start) {
+ first_free_cluster = post_rb_start;
+ }
+ rt_ofs = alloc_clusters_imrt(bs, size_to_clusters(s, reftable_size *
+ sizeof(uint64_t)),
+ refcount_table, nb_clusters,
+ &first_free_cluster);
+ if (rt_ofs < 0) {
+ fprintf(stderr, "ERROR allocating reftable: %s\n", strerror(-ret));
+ res->check_errors++;
+ ret = rt_ofs;
+ goto fail;
+ }
+
+ goto write_refblocks;
+ }
+
+ assert(reftable);
+
+ for (rb_index = 0; rb_index < reftable_size; rb_index++) {
+ cpu_to_be64s(&reftable[rb_index]);
+ }
+
+ ret = qcow2_pre_write_overlap_check(bs, 0, rt_ofs,
+ reftable_size * sizeof(uint64_t));
+ if (ret < 0) {
+ fprintf(stderr, "ERROR writing reftable: %s\n", strerror(-ret));
+ goto fail;
+ }
+
+ ret = bdrv_write(bs->file, rt_ofs / BDRV_SECTOR_SIZE, (void *)reftable,
+ reftable_size * sizeof(uint64_t) / BDRV_SECTOR_SIZE);
+ if (ret < 0) {
+ fprintf(stderr, "ERROR writing reftable: %s\n", strerror(-ret));
+ goto fail;
+ }
+
+ /* Enter new reftable into the image header */
+ cpu_to_be64w((uint64_t *)&rt_offset_and_clusters[0], rt_ofs);
+ cpu_to_be32w((uint32_t *)&rt_offset_and_clusters[sizeof(uint64_t)],
+ size_to_clusters(s, reftable_size * sizeof(uint64_t)));
+ ret = bdrv_pwrite_sync(bs->file, offsetof(QCowHeader,
+ refcount_table_offset),
+ rt_offset_and_clusters,
+ sizeof(rt_offset_and_clusters));
+ if (ret < 0) {
+ fprintf(stderr, "ERROR setting reftable: %s\n", strerror(-ret));
+ goto fail;
+ }
+
+ for (rb_index = 0; rb_index < reftable_size; rb_index++) {
+ be64_to_cpus(&reftable[rb_index]);
+ }
+ s->refcount_table = reftable;
+ s->refcount_table_offset = rt_ofs;
+ s->refcount_table_size = reftable_size;
+
+ return 0;
+
+fail:
+ g_free(reftable);
+ return ret;
+}
+
+/*
* Checks an image for refcount consistency.
*
* Returns 0 if no errors are found, the number of errors in case the image is
@@ -1600,6 +1839,7 @@ int qcow2_check_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
BdrvCheckMode fix)
{
BDRVQcowState *s = bs->opaque;
+ BdrvCheckResult pre_compare_res;
int64_t size, highest_cluster, nb_clusters;
uint16_t *refcount_table = NULL;
bool rebuild = false;
@@ -1626,11 +1866,30 @@ int qcow2_check_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
goto fail;
}
- compare_refcounts(bs, res, fix, &rebuild, &highest_cluster, refcount_table,
+ /* In case we don't need to rebuild the refcount structure (but want to fix
+ * something), this function is immediately called again, in which case the
+ * result should be ignored */
+ pre_compare_res = *res;
+ compare_refcounts(bs, res, 0, &rebuild, &highest_cluster, refcount_table,
nb_clusters);
- if (rebuild) {
- fprintf(stderr, "ERROR need to rebuild refcount structures\n");
+ if (rebuild && (fix & BDRV_FIX_ERRORS)) {
+ fprintf(stderr, "Rebuilding refcount structure\n");
+ ret = rebuild_refcount_structure(bs, res, &refcount_table,
+ &nb_clusters);
+ if (ret < 0) {
+ goto fail;
+ }
+ } else if (fix) {
+ if (rebuild) {
+ fprintf(stderr, "ERROR need to rebuild refcount structures\n");
+ }
+
+ if (res->leaks || res->corruptions) {
+ *res = pre_compare_res;
+ compare_refcounts(bs, res, fix, &rebuild, &highest_cluster,
+ refcount_table, nb_clusters);
+ }
}
/* check OFLAG_COPIED */
--
2.0.4
^ permalink raw reply related [flat|nested] 28+ messages in thread
* [Qemu-devel] [PATCH v3 08/10] qcow2: Clean up after refcount rebuild
2014-08-22 16:31 [Qemu-devel] [PATCH v3 00/10] qcow2: Fix image repairing Max Reitz
` (6 preceding siblings ...)
2014-08-22 16:31 ` [Qemu-devel] [PATCH v3 07/10] qcow2: Rebuild refcount structure during check Max Reitz
@ 2014-08-22 16:31 ` Max Reitz
2014-08-22 18:43 ` Benoît Canet
2014-08-22 16:31 ` [Qemu-devel] [PATCH v3 09/10] iotests: Fix test outputs Max Reitz
2014-08-22 16:31 ` [Qemu-devel] [PATCH v3 10/10] iotests: Add test for potentially damaging repairs Max Reitz
9 siblings, 1 reply; 28+ messages in thread
From: Max Reitz @ 2014-08-22 16:31 UTC (permalink / raw)
To: qemu-devel; +Cc: Kevin Wolf, Benoît Canet, Stefan Hajnoczi, Max Reitz
Because the old refcount structure will be leaked after having rebuilt
it, we need to recalculate the refcounts and run a leak-fixing operation
afterwards (if leaks should be fixed at all).
Signed-off-by: Max Reitz <mreitz@redhat.com>
---
block/qcow2-refcount.c | 35 +++++++++++++++++++++++++++++++++++
1 file changed, 35 insertions(+)
diff --git a/block/qcow2-refcount.c b/block/qcow2-refcount.c
index 59cab65..818facc 100644
--- a/block/qcow2-refcount.c
+++ b/block/qcow2-refcount.c
@@ -1874,12 +1874,47 @@ int qcow2_check_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
nb_clusters);
if (rebuild && (fix & BDRV_FIX_ERRORS)) {
+ BdrvCheckResult old_res = *res;
+
fprintf(stderr, "Rebuilding refcount structure\n");
ret = rebuild_refcount_structure(bs, res, &refcount_table,
&nb_clusters);
if (ret < 0) {
goto fail;
}
+
+ res->corruptions = 0;
+ res->leaks = 0;
+
+ /* Because the old reftable has been exchanged for a new one the
+ * references have to be recalculated */
+ rebuild = false;
+ memset(refcount_table, 0, nb_clusters * sizeof(uint16_t));
+ ret = calculate_refcounts(bs, res, 0, &rebuild, &refcount_table,
+ &nb_clusters);
+ if (ret < 0) {
+ goto fail;
+ }
+
+ if (fix & BDRV_FIX_LEAKS) {
+ /* The old refcount structures are now leaked, fix it; the result
+ * can be ignored */
+ pre_compare_res = *res;
+ compare_refcounts(bs, res, BDRV_FIX_LEAKS, &rebuild,
+ &highest_cluster, refcount_table, nb_clusters);
+ if (rebuild) {
+ fprintf(stderr, "ERROR rebuilt refcount structure is still "
+ "broken\n");
+ }
+ *res = pre_compare_res;
+ }
+
+ if (res->corruptions < old_res.corruptions) {
+ res->corruptions_fixed += old_res.corruptions - res->corruptions;
+ }
+ if (res->leaks < old_res.leaks) {
+ res->leaks_fixed += old_res.leaks - res->leaks;
+ }
} else if (fix) {
if (rebuild) {
fprintf(stderr, "ERROR need to rebuild refcount structures\n");
--
2.0.4
^ permalink raw reply related [flat|nested] 28+ messages in thread
* [Qemu-devel] [PATCH v3 09/10] iotests: Fix test outputs
2014-08-22 16:31 [Qemu-devel] [PATCH v3 00/10] qcow2: Fix image repairing Max Reitz
` (7 preceding siblings ...)
2014-08-22 16:31 ` [Qemu-devel] [PATCH v3 08/10] qcow2: Clean up after refcount rebuild Max Reitz
@ 2014-08-22 16:31 ` Max Reitz
2014-08-26 13:09 ` Eric Blake
2014-08-22 16:31 ` [Qemu-devel] [PATCH v3 10/10] iotests: Add test for potentially damaging repairs Max Reitz
9 siblings, 1 reply; 28+ messages in thread
From: Max Reitz @ 2014-08-22 16:31 UTC (permalink / raw)
To: qemu-devel; +Cc: Kevin Wolf, Benoît Canet, Stefan Hajnoczi, Max Reitz
039, 060 and 061 all create images with referenced clusters having a
refcount of 0. Because previous commits changed handling of such errors,
these tests now have a different output. Fix it.
Furthermore, 060 created a refblock with a refcount greater than one
which now results in having to rebuild the refcount structure as well.
Signed-off-by: Max Reitz <mreitz@redhat.com>
---
tests/qemu-iotests/039.out | 10 ++++++++--
tests/qemu-iotests/060.out | 10 ++++++++--
tests/qemu-iotests/061.out | 18 ++++++++++++------
3 files changed, 28 insertions(+), 10 deletions(-)
diff --git a/tests/qemu-iotests/039.out b/tests/qemu-iotests/039.out
index 67e7744..0adf153 100644
--- a/tests/qemu-iotests/039.out
+++ b/tests/qemu-iotests/039.out
@@ -25,7 +25,10 @@ read 512/512 bytes at offset 0
incompatible_features 0x1
== Repairing the image file must succeed ==
-Repairing cluster 5 refcount=0 reference=1
+ERROR cluster 5 refcount=0 reference=1
+Rebuilding refcount structure
+Repairing cluster 1 refcount=1 reference=0
+Repairing cluster 2 refcount=1 reference=0
The following inconsistencies were found and repaired:
0 leaked clusters
@@ -45,7 +48,10 @@ wrote 512/512 bytes at offset 0
512 bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
./039: Aborted ( ulimit -c 0; exec "$@" )
incompatible_features 0x1
-Repairing cluster 5 refcount=0 reference=1
+ERROR cluster 5 refcount=0 reference=1
+Rebuilding refcount structure
+Repairing cluster 1 refcount=1 reference=0
+Repairing cluster 2 refcount=1 reference=0
wrote 512/512 bytes at offset 0
512 bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
incompatible_features 0x0
diff --git a/tests/qemu-iotests/060.out b/tests/qemu-iotests/060.out
index c27c952..41ecbc7 100644
--- a/tests/qemu-iotests/060.out
+++ b/tests/qemu-iotests/060.out
@@ -27,11 +27,15 @@ incompatible_features 0x0
qcow2: Preventing invalid write on metadata (overlaps with refcount block); image marked as corrupt.
write failed: Input/output error
incompatible_features 0x2
-Repairing refcount block 0 refcount=2
+ERROR refcount block 0 refcount=2
+ERROR cluster 2 refcount=1 reference=2
+Rebuilding refcount structure
+Repairing cluster 1 refcount=1 reference=0
+Repairing cluster 2 refcount=2 reference=1
The following inconsistencies were found and repaired:
0 leaked clusters
- 1 corruptions
+ 2 corruptions
Double checking the fixed image now...
No errors were found on the image.
@@ -59,6 +63,8 @@ incompatible_features 0x0
qcow2: Preventing invalid write on metadata (overlaps with inactive L2 table); image marked as corrupt.
write failed: Input/output error
incompatible_features 0x2
+ERROR cluster 4 refcount=1 reference=2
+Leaked cluster 9 refcount=1 reference=0
Repairing cluster 4 refcount=1 reference=2
Repairing cluster 9 refcount=1 reference=0
Repairing OFLAG_COPIED data cluster: l2_entry=8000000000040000 refcount=2
diff --git a/tests/qemu-iotests/061.out b/tests/qemu-iotests/061.out
index e372470..4ae6472 100644
--- a/tests/qemu-iotests/061.out
+++ b/tests/qemu-iotests/061.out
@@ -76,8 +76,11 @@ autoclear_features 0x0
refcount_order 4
header_length 104
-Repairing cluster 5 refcount=0 reference=1
-Repairing cluster 6 refcount=0 reference=1
+ERROR cluster 5 refcount=0 reference=1
+ERROR cluster 6 refcount=0 reference=1
+Rebuilding refcount structure
+Repairing cluster 1 refcount=1 reference=0
+Repairing cluster 2 refcount=1 reference=0
magic 0x514649fb
version 2
backing_file_offset 0x0
@@ -87,7 +90,7 @@ size 67108864
crypt_method 0
l1_size 1
l1_table_offset 0x30000
-refcount_table_offset 0x10000
+refcount_table_offset 0x80000
refcount_table_clusters 1
nb_snapshots 0
snapshot_offset 0x0
@@ -230,8 +233,11 @@ autoclear_features 0x0
refcount_order 4
header_length 104
-Repairing cluster 5 refcount=0 reference=1
-Repairing cluster 6 refcount=0 reference=1
+ERROR cluster 5 refcount=0 reference=1
+ERROR cluster 6 refcount=0 reference=1
+Rebuilding refcount structure
+Repairing cluster 1 refcount=1 reference=0
+Repairing cluster 2 refcount=1 reference=0
magic 0x514649fb
version 3
backing_file_offset 0x0
@@ -241,7 +247,7 @@ size 67108864
crypt_method 0
l1_size 1
l1_table_offset 0x30000
-refcount_table_offset 0x10000
+refcount_table_offset 0x80000
refcount_table_clusters 1
nb_snapshots 0
snapshot_offset 0x0
--
2.0.4
^ permalink raw reply related [flat|nested] 28+ messages in thread
* [Qemu-devel] [PATCH v3 10/10] iotests: Add test for potentially damaging repairs
2014-08-22 16:31 [Qemu-devel] [PATCH v3 00/10] qcow2: Fix image repairing Max Reitz
` (8 preceding siblings ...)
2014-08-22 16:31 ` [Qemu-devel] [PATCH v3 09/10] iotests: Fix test outputs Max Reitz
@ 2014-08-22 16:31 ` Max Reitz
2014-08-22 18:50 ` Benoît Canet
2014-08-22 19:55 ` Eric Blake
9 siblings, 2 replies; 28+ messages in thread
From: Max Reitz @ 2014-08-22 16:31 UTC (permalink / raw)
To: qemu-devel; +Cc: Kevin Wolf, Benoît Canet, Stefan Hajnoczi, Max Reitz
There are certain cases where repairing a qcow2 image might actually
damage it further (or rather, where repairing it has in fact damaged it
further with the old qcow2 check implementation). This should not
happen, so add a test for these cases.
Signed-off-by: Max Reitz <mreitz@redhat.com>
---
tests/qemu-iotests/104 | 98 ++++++++++++++++++++++++++++++++++++++++++++++
tests/qemu-iotests/104.out | 46 ++++++++++++++++++++++
tests/qemu-iotests/group | 1 +
3 files changed, 145 insertions(+)
create mode 100755 tests/qemu-iotests/104
create mode 100644 tests/qemu-iotests/104.out
diff --git a/tests/qemu-iotests/104 b/tests/qemu-iotests/104
new file mode 100755
index 0000000..cc6d0b3
--- /dev/null
+++ b/tests/qemu-iotests/104
@@ -0,0 +1,98 @@
+#!/bin/bash
+#
+# Test case for repairing qcow2 images which cannot be repaired using
+# the on-disk refcount structures
+#
+# Copyright (C) 2013 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=mreitz@redhat.com
+
+seq="$(basename $0)"
+echo "QA output created by $seq"
+
+here="$PWD"
+tmp=/tmp/$$
+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
+
+# This tests qocw2-specific low-level functionality
+_supported_fmt qcow2
+_supported_proto file
+_supported_os Linux
+
+echo
+echo '=== Repairing an image without any refcount table ==='
+echo
+
+_make_test_img 64M
+# just write some data
+$QEMU_IO -c 'write -P 42 0 64k' "$TEST_IMG" | _filter_qemu_io
+
+# refcount_table_offset
+poke_file "$TEST_IMG" 48 "\x00\x00\x00\x00\x00\x00\x00\x00"
+# refcount_table_clusters
+poke_file "$TEST_IMG" 56 "\x00\x00\x00\x00"
+
+_check_test_img -r all
+
+$QEMU_IO -c 'read -P 42 0 64k' "$TEST_IMG" | _filter_qemu_io
+
+echo
+echo '=== Repairing unreferenced data cluster in new refblock area ==='
+echo
+
+IMGOPTS='cluster_size=512' _make_test_img 64M
+# Allocate the first 128 kB in the image (first refblock)
+$QEMU_IO -c 'write 0 111104' "$TEST_IMG" | _filter_qemu_io
+# should be 131072
+stat -c '%s' "$TEST_IMG"
+
+# Enter a cluster at 128 kB (0x20000)
+# XXX: This (0x1ccc8) should be the first free entry in the last L2 table, but
+# we cannot be sure
+poke_file "$TEST_IMG" 117960 "\x80\x00\x00\x00\x00\x02\x00\x00"
+
+# Fill the cluster
+truncate -s 131584 "$TEST_IMG"
+$QEMU_IO -c "open -o driver=raw $TEST_IMG" -c 'write -P 42 128k 512' \
+ | _filter_qemu_io
+
+# The data should now appear at this guest offset
+$QEMU_IO -c 'read -P 42 111104 512' "$TEST_IMG" | _filter_qemu_io
+
+# This cluster is unallocated; fix it
+_check_test_img -r all
+
+# This repair operation must have allocated a new refblock; and that refblock
+# should not overlap with the unallocated data cluster. If it does, the data
+# will be damaged, so check it.
+$QEMU_IO -c 'read -P 42 111104 512' "$TEST_IMG" | _filter_qemu_io
+
+# success, all done
+echo '*** done'
+rm -f $seq.full
+status=0
diff --git a/tests/qemu-iotests/104.out b/tests/qemu-iotests/104.out
new file mode 100644
index 0000000..742fe77
--- /dev/null
+++ b/tests/qemu-iotests/104.out
@@ -0,0 +1,46 @@
+QA output created by 104
+
+=== Repairing an image without any refcount table ===
+
+Formatting 'TEST_DIR/t.IMGFMT', fmt=IMGFMT size=67108864
+wrote 65536/65536 bytes at offset 0
+64 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
+ERROR cluster 0 refcount=0 reference=1
+ERROR cluster 3 refcount=0 reference=1
+ERROR cluster 4 refcount=0 reference=1
+ERROR cluster 5 refcount=0 reference=1
+Rebuilding refcount structure
+The following inconsistencies were found and repaired:
+
+ 0 leaked clusters
+ 4 corruptions
+
+Double checking the fixed image now...
+No errors were found on the image.
+read 65536/65536 bytes at offset 0
+64 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
+
+=== Repairing unreferenced data cluster in new refblock area ===
+
+Formatting 'TEST_DIR/t.IMGFMT', fmt=IMGFMT size=67108864
+wrote 111104/111104 bytes at offset 0
+108.500 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
+131072
+wrote 512/512 bytes at offset 131072
+512 bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
+read 512/512 bytes at offset 111104
+512 bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
+ERROR cluster 256 refcount=0 reference=1
+Rebuilding refcount structure
+Repairing cluster 1 refcount=1 reference=0
+Repairing cluster 2 refcount=1 reference=0
+The following inconsistencies were found and repaired:
+
+ 0 leaked clusters
+ 1 corruptions
+
+Double checking the fixed image now...
+No errors were found on the image.
+read 512/512 bytes at offset 111104
+512 bytes, 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 2803d68..5a3f9dd 100644
--- a/tests/qemu-iotests/group
+++ b/tests/qemu-iotests/group
@@ -103,3 +103,4 @@
099 rw auto quick
101 rw auto quick
103 rw auto quick
+104 rw auto quick
--
2.0.4
^ permalink raw reply related [flat|nested] 28+ messages in thread
* Re: [Qemu-devel] [PATCH v3 02/10] qcow2: Split qcow2_check_refcounts()
2014-08-22 16:31 ` [Qemu-devel] [PATCH v3 02/10] qcow2: Split qcow2_check_refcounts() Max Reitz
@ 2014-08-22 18:03 ` Benoît Canet
0 siblings, 0 replies; 28+ messages in thread
From: Benoît Canet @ 2014-08-22 18:03 UTC (permalink / raw)
To: Max Reitz; +Cc: Kevin Wolf, qemu-devel, Stefan Hajnoczi, Benoît Canet
On Fri, Aug 22, 2014 at 06:31:36PM +0200, Max Reitz wrote:
> Put the code for calculating the reference counts and comparing them
> during qemu-img check into own functions.
>
> Signed-off-by: Max Reitz <mreitz@redhat.com>
> ---
> block/qcow2-refcount.c | 153 ++++++++++++++++++++++++++++++++-----------------
> 1 file changed, 102 insertions(+), 51 deletions(-)
>
> diff --git a/block/qcow2-refcount.c b/block/qcow2-refcount.c
> index 43665b8..5f0920b 100644
> --- a/block/qcow2-refcount.c
> +++ b/block/qcow2-refcount.c
> @@ -1496,71 +1496,70 @@ done:
> return new_offset;
> }
>
> +static int check_refblocks(BlockDriverState *bs, BdrvCheckResult *res,
> + BdrvCheckMode fix, uint16_t **refcount_table,
> + int64_t *nb_clusters);
> +
> /*
> - * Checks an image for refcount consistency.
> - *
> - * Returns 0 if no errors are found, the number of errors in case the image is
> - * detected as corrupted, and -errno when an internal error occurred.
> + * Calculates an in-memory refcount table.
> */
> -int qcow2_check_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
> - BdrvCheckMode fix)
> +static int calculate_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
> + BdrvCheckMode fix, uint16_t **refcount_table,
> + int64_t *nb_clusters)
> {
> BDRVQcowState *s = bs->opaque;
> - int64_t size, i, highest_cluster, nb_clusters;
> - int refcount1, refcount2;
> + int64_t i;
> QCowSnapshot *sn;
> - uint16_t *refcount_table;
> int ret;
>
> - size = bdrv_getlength(bs->file);
> - if (size < 0) {
> - res->check_errors++;
> - return size;
> - }
> -
> - nb_clusters = size_to_clusters(s, size);
> - if (nb_clusters > INT_MAX) {
> - res->check_errors++;
> - return -EFBIG;
> - }
> -
> - refcount_table = g_try_new0(uint16_t, nb_clusters);
> - if (nb_clusters && refcount_table == NULL) {
> + *refcount_table = g_try_new0(uint16_t, *nb_clusters);
> + if (*nb_clusters && *refcount_table == NULL) {
> res->check_errors++;
> return -ENOMEM;
> }
>
> - res->bfi.total_clusters =
> - size_to_clusters(s, bs->total_sectors * BDRV_SECTOR_SIZE);
> -
> /* header */
> - inc_refcounts(bs, res, refcount_table, nb_clusters,
> + inc_refcounts(bs, res, *refcount_table, *nb_clusters,
> 0, s->cluster_size);
>
> /* current L1 table */
> - ret = check_refcounts_l1(bs, res, refcount_table, nb_clusters,
> + ret = check_refcounts_l1(bs, res, *refcount_table, *nb_clusters,
> s->l1_table_offset, s->l1_size, CHECK_FRAG_INFO);
> if (ret < 0) {
> - goto fail;
> + return ret;
> }
>
> /* snapshots */
> - for(i = 0; i < s->nb_snapshots; i++) {
> + for (i = 0; i < s->nb_snapshots; i++) {
> sn = s->snapshots + i;
> - ret = check_refcounts_l1(bs, res, refcount_table, nb_clusters,
> + ret = check_refcounts_l1(bs, res, *refcount_table, *nb_clusters,
> sn->l1_table_offset, sn->l1_size, 0);
> if (ret < 0) {
> - goto fail;
> + return ret;
> }
> }
> - inc_refcounts(bs, res, refcount_table, nb_clusters,
> + inc_refcounts(bs, res, *refcount_table, *nb_clusters,
> s->snapshots_offset, s->snapshots_size);
>
> /* refcount data */
> - inc_refcounts(bs, res, refcount_table, nb_clusters,
> + inc_refcounts(bs, res, *refcount_table, *nb_clusters,
> s->refcount_table_offset,
> s->refcount_table_size * sizeof(uint64_t));
>
> + return check_refblocks(bs, res, fix, refcount_table, nb_clusters);
> +}
> +
> +/*
> + * Checks consistency of refblocks and accounts for each refblock in
> + * *refcount_table.
> + */
> +static int check_refblocks(BlockDriverState *bs, BdrvCheckResult *res,
> + BdrvCheckMode fix, uint16_t **refcount_table,
> + int64_t *nb_clusters)
> +{
> + BDRVQcowState *s = bs->opaque;
> + int64_t i;
> +
> for(i = 0; i < s->refcount_table_size; i++) {
> uint64_t offset, cluster;
> offset = s->refcount_table[i];
> @@ -1574,7 +1573,7 @@ int qcow2_check_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
> continue;
> }
>
> - if (cluster >= nb_clusters) {
> + if (cluster >= *nb_clusters) {
> fprintf(stderr, "ERROR refcount block %" PRId64
> " is outside image\n", i);
> res->corruptions++;
> @@ -1582,14 +1581,14 @@ int qcow2_check_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
> }
>
> if (offset != 0) {
> - inc_refcounts(bs, res, refcount_table, nb_clusters,
> + inc_refcounts(bs, res, *refcount_table, *nb_clusters,
> offset, s->cluster_size);
> - if (refcount_table[cluster] != 1) {
> + if ((*refcount_table)[cluster] != 1) {
> fprintf(stderr, "%s refcount block %" PRId64
> " refcount=%d\n",
> fix & BDRV_FIX_ERRORS ? "Repairing" :
> "ERROR",
> - i, refcount_table[cluster]);
> + i, (*refcount_table)[cluster]);
>
> if (fix & BDRV_FIX_ERRORS) {
> int64_t new_offset;
> @@ -1601,17 +1600,18 @@ int qcow2_check_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
> }
>
> /* update refcounts */
> - if ((new_offset >> s->cluster_bits) >= nb_clusters) {
> + if ((new_offset >> s->cluster_bits) >= *nb_clusters) {
> /* increase refcount_table size if necessary */
> - int old_nb_clusters = nb_clusters;
> - nb_clusters = (new_offset >> s->cluster_bits) + 1;
> - refcount_table = g_renew(uint16_t, refcount_table,
> - nb_clusters);
> - memset(&refcount_table[old_nb_clusters], 0, (nb_clusters
> - - old_nb_clusters) * sizeof(uint16_t));
> + int old_nb_clusters = *nb_clusters;
> + *nb_clusters = (new_offset >> s->cluster_bits) + 1;
> + *refcount_table = g_renew(uint16_t, *refcount_table,
> + *nb_clusters);
> + memset(&(*refcount_table)[old_nb_clusters], 0,
> + (*nb_clusters - old_nb_clusters) *
> + sizeof(uint16_t));
> }
> - refcount_table[cluster]--;
> - inc_refcounts(bs, res, refcount_table, nb_clusters,
> + (*refcount_table)[cluster]--;
> + inc_refcounts(bs, res, *refcount_table, *nb_clusters,
> new_offset, s->cluster_size);
>
> res->corruptions_fixed++;
> @@ -1622,8 +1622,22 @@ int qcow2_check_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
> }
> }
>
> - /* compare ref counts */
> - for (i = 0, highest_cluster = 0; i < nb_clusters; i++) {
> + return 0;
> +}
> +
> +/*
> + * Compares the actual reference count for each cluster in the image against the
> + * refcount as reported by the refcount structures on-disk.
> + */
> +static void compare_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
> + BdrvCheckMode fix, int64_t *highest_cluster,
> + uint16_t *refcount_table, int64_t nb_clusters)
> +{
> + BDRVQcowState *s = bs->opaque;
> + int64_t i;
> + int refcount1, refcount2, ret;
> +
> + for (i = 0, *highest_cluster = 0; i < nb_clusters; i++) {
> refcount1 = get_refcount(bs, i);
> if (refcount1 < 0) {
> fprintf(stderr, "Can't get refcount for cluster %" PRId64 ": %s\n",
> @@ -1635,11 +1649,10 @@ int qcow2_check_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
> refcount2 = refcount_table[i];
>
> if (refcount1 > 0 || refcount2 > 0) {
> - highest_cluster = i;
> + *highest_cluster = i;
> }
>
> if (refcount1 != refcount2) {
> -
> /* Check if we're allowed to fix the mismatch */
> int *num_fixed = NULL;
> if (refcount1 > refcount2 && (fix & BDRV_FIX_LEAKS)) {
> @@ -1672,6 +1685,44 @@ int qcow2_check_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
> }
> }
> }
> +}
> +
> +/*
> + * Checks an image for refcount consistency.
> + *
> + * Returns 0 if no errors are found, the number of errors in case the image is
> + * detected as corrupted, and -errno when an internal error occurred.
> + */
> +int qcow2_check_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
> + BdrvCheckMode fix)
> +{
> + BDRVQcowState *s = bs->opaque;
> + int64_t size, highest_cluster, nb_clusters;
> + uint16_t *refcount_table;
> + int ret;
> +
> + size = bdrv_getlength(bs->file);
> + if (size < 0) {
> + res->check_errors++;
> + return size;
> + }
> +
> + nb_clusters = size_to_clusters(s, size);
> + if (nb_clusters > INT_MAX) {
> + res->check_errors++;
> + return -EFBIG;
> + }
> +
> + res->bfi.total_clusters =
> + size_to_clusters(s, bs->total_sectors * BDRV_SECTOR_SIZE);
> +
> + ret = calculate_refcounts(bs, res, fix, &refcount_table, &nb_clusters);
> + if (ret < 0) {
> + goto fail;
> + }
> +
> + compare_refcounts(bs, res, fix, &highest_cluster, refcount_table,
> + nb_clusters);
>
> /* check OFLAG_COPIED */
> ret = check_oflag_copied(bs, res, fix);
> --
> 2.0.4
>
Thanks it's nicer to read.
My opinion changed from "It must be good Max took care of it" to "I did not find any bug" :)
Reviewed-by: Benoît Canet <benoit.canet@nodalink.com>
^ permalink raw reply [flat|nested] 28+ messages in thread
* Re: [Qemu-devel] [PATCH v3 03/10] qcow2: Pull check_refblocks() up
2014-08-22 16:31 ` [Qemu-devel] [PATCH v3 03/10] qcow2: Pull check_refblocks() up Max Reitz
@ 2014-08-22 18:04 ` Benoît Canet
0 siblings, 0 replies; 28+ messages in thread
From: Benoît Canet @ 2014-08-22 18:04 UTC (permalink / raw)
To: Max Reitz; +Cc: Kevin Wolf, qemu-devel, Stefan Hajnoczi, Benoît Canet
On Fri, Aug 22, 2014 at 06:31:37PM +0200, Max Reitz wrote:
> Pull check_refblocks() before calculate_refcounts() so we can drop its
> static declaration.
>
> Signed-off-by: Max Reitz <mreitz@redhat.com>
> ---
> block/qcow2-refcount.c | 102 ++++++++++++++++++++++++-------------------------
> 1 file changed, 49 insertions(+), 53 deletions(-)
>
> diff --git a/block/qcow2-refcount.c b/block/qcow2-refcount.c
> index 5f0920b..2b728ef 100644
> --- a/block/qcow2-refcount.c
> +++ b/block/qcow2-refcount.c
> @@ -1496,59 +1496,6 @@ done:
> return new_offset;
> }
>
> -static int check_refblocks(BlockDriverState *bs, BdrvCheckResult *res,
> - BdrvCheckMode fix, uint16_t **refcount_table,
> - int64_t *nb_clusters);
> -
> -/*
> - * Calculates an in-memory refcount table.
> - */
> -static int calculate_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
> - BdrvCheckMode fix, uint16_t **refcount_table,
> - int64_t *nb_clusters)
> -{
> - BDRVQcowState *s = bs->opaque;
> - int64_t i;
> - QCowSnapshot *sn;
> - int ret;
> -
> - *refcount_table = g_try_new0(uint16_t, *nb_clusters);
> - if (*nb_clusters && *refcount_table == NULL) {
> - res->check_errors++;
> - return -ENOMEM;
> - }
> -
> - /* header */
> - inc_refcounts(bs, res, *refcount_table, *nb_clusters,
> - 0, s->cluster_size);
> -
> - /* current L1 table */
> - ret = check_refcounts_l1(bs, res, *refcount_table, *nb_clusters,
> - s->l1_table_offset, s->l1_size, CHECK_FRAG_INFO);
> - if (ret < 0) {
> - return ret;
> - }
> -
> - /* snapshots */
> - for (i = 0; i < s->nb_snapshots; i++) {
> - sn = s->snapshots + i;
> - ret = check_refcounts_l1(bs, res, *refcount_table, *nb_clusters,
> - sn->l1_table_offset, sn->l1_size, 0);
> - if (ret < 0) {
> - return ret;
> - }
> - }
> - inc_refcounts(bs, res, *refcount_table, *nb_clusters,
> - s->snapshots_offset, s->snapshots_size);
> -
> - /* refcount data */
> - inc_refcounts(bs, res, *refcount_table, *nb_clusters,
> - s->refcount_table_offset,
> - s->refcount_table_size * sizeof(uint64_t));
> -
> - return check_refblocks(bs, res, fix, refcount_table, nb_clusters);
> -}
> -
> /*
> * Checks consistency of refblocks and accounts for each refblock in
> * *refcount_table.
> @@ -1626,6 +1573,55 @@ static int check_refblocks(BlockDriverState *bs, BdrvCheckResult *res,
> }
>
> /*
> + * Calculates an in-memory refcount table.
> + */
> +static int calculate_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
> + BdrvCheckMode fix, uint16_t **refcount_table,
> + int64_t *nb_clusters)
> +{
> + BDRVQcowState *s = bs->opaque;
> + int64_t i;
> + QCowSnapshot *sn;
> + int ret;
> +
> + *refcount_table = g_try_new0(uint16_t, *nb_clusters);
> + if (*nb_clusters && *refcount_table == NULL) {
> + res->check_errors++;
> + return -ENOMEM;
> + }
> +
> + /* header */
> + inc_refcounts(bs, res, *refcount_table, *nb_clusters,
> + 0, s->cluster_size);
> +
> + /* current L1 table */
> + ret = check_refcounts_l1(bs, res, *refcount_table, *nb_clusters,
> + s->l1_table_offset, s->l1_size, CHECK_FRAG_INFO);
> + if (ret < 0) {
> + return ret;
> + }
> +
> + /* snapshots */
> + for (i = 0; i < s->nb_snapshots; i++) {
> + sn = s->snapshots + i;
> + ret = check_refcounts_l1(bs, res, *refcount_table, *nb_clusters,
> + sn->l1_table_offset, sn->l1_size, 0);
> + if (ret < 0) {
> + return ret;
> + }
> + }
> + inc_refcounts(bs, res, *refcount_table, *nb_clusters,
> + s->snapshots_offset, s->snapshots_size);
> +
> + /* refcount data */
> + inc_refcounts(bs, res, *refcount_table, *nb_clusters,
> + s->refcount_table_offset,
> + s->refcount_table_size * sizeof(uint64_t));
> +
> + return check_refblocks(bs, res, fix, refcount_table, nb_clusters);
> +}
> +
> +/*
> * Compares the actual reference count for each cluster in the image against the
> * refcount as reported by the refcount structures on-disk.
> */
> --
> 2.0.4
>
Reviewed-by: Benoît Canet <benoit.canet@nodalink.com>
^ permalink raw reply [flat|nested] 28+ messages in thread
* Re: [Qemu-devel] [PATCH v3 04/10] qcow2: Reuse refcount table in calculate_refcounts()
2014-08-22 16:31 ` [Qemu-devel] [PATCH v3 04/10] qcow2: Reuse refcount table in calculate_refcounts() Max Reitz
@ 2014-08-22 18:07 ` Benoît Canet
0 siblings, 0 replies; 28+ messages in thread
From: Benoît Canet @ 2014-08-22 18:07 UTC (permalink / raw)
To: Max Reitz; +Cc: Kevin Wolf, qemu-devel, Stefan Hajnoczi, Benoît Canet
On Fri, Aug 22, 2014 at 06:31:38PM +0200, Max Reitz wrote:
> We will later call calculate_refcounts multiple times, so reuse the
> refcount table if possible.
>
> Signed-off-by: Max Reitz <mreitz@redhat.com>
> ---
> block/qcow2-refcount.c | 12 +++++++-----
> 1 file changed, 7 insertions(+), 5 deletions(-)
>
> diff --git a/block/qcow2-refcount.c b/block/qcow2-refcount.c
> index 2b728ef..babe6cb 100644
> --- a/block/qcow2-refcount.c
> +++ b/block/qcow2-refcount.c
> @@ -1584,10 +1584,12 @@ static int calculate_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
> QCowSnapshot *sn;
> int ret;
>
> - *refcount_table = g_try_new0(uint16_t, *nb_clusters);
> - if (*nb_clusters && *refcount_table == NULL) {
> - res->check_errors++;
> - return -ENOMEM;
> + if (!*refcount_table) {
> + *refcount_table = g_try_new0(uint16_t, *nb_clusters);
> + if (*nb_clusters && *refcount_table == NULL) {
> + res->check_errors++;
> + return -ENOMEM;
> + }
> }
>
> /* header */
> @@ -1694,7 +1696,7 @@ int qcow2_check_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
> {
> BDRVQcowState *s = bs->opaque;
> int64_t size, highest_cluster, nb_clusters;
> - uint16_t *refcount_table;
> + uint16_t *refcount_table = NULL;
> int ret;
>
> size = bdrv_getlength(bs->file);
> --
> 2.0.4
>
Reviewed-by: Benoît Canet <benoit.canet@nodalink.com>
^ permalink raw reply [flat|nested] 28+ messages in thread
* Re: [Qemu-devel] [PATCH v3 05/10] qcow2: Fix refcount blocks beyond image end
2014-08-22 16:31 ` [Qemu-devel] [PATCH v3 05/10] qcow2: Fix refcount blocks beyond image end Max Reitz
@ 2014-08-22 18:20 ` Benoît Canet
2014-08-26 13:07 ` Eric Blake
1 sibling, 0 replies; 28+ messages in thread
From: Benoît Canet @ 2014-08-22 18:20 UTC (permalink / raw)
To: Max Reitz; +Cc: Kevin Wolf, qemu-devel, Stefan Hajnoczi, Benoît Canet
On Fri, Aug 22, 2014 at 06:31:39PM +0200, Max Reitz wrote:
> If the qcow2 check function detects a refcount block located beyond the
> image end, grow the image appropriately. This cannot break anything and
> is the logical fix for such a case.
>
> Signed-off-by: Max Reitz <mreitz@redhat.com>
> ---
> block/qcow2-refcount.c | 50 ++++++++++++++++++++++++++++++++++++++++++++++----
> 1 file changed, 46 insertions(+), 4 deletions(-)
>
> diff --git a/block/qcow2-refcount.c b/block/qcow2-refcount.c
> index babe6cb..1f0f44e 100644
> --- a/block/qcow2-refcount.c
> +++ b/block/qcow2-refcount.c
> @@ -1505,7 +1505,8 @@ static int check_refblocks(BlockDriverState *bs, BdrvCheckResult *res,
> int64_t *nb_clusters)
> {
> BDRVQcowState *s = bs->opaque;
> - int64_t i;
> + int64_t i, size;
> + int ret;
>
> for(i = 0; i < s->refcount_table_size; i++) {
> uint64_t offset, cluster;
> @@ -1521,9 +1522,50 @@ static int check_refblocks(BlockDriverState *bs, BdrvCheckResult *res,
> }
>
> if (cluster >= *nb_clusters) {
> - fprintf(stderr, "ERROR refcount block %" PRId64
> - " is outside image\n", i);
> - res->corruptions++;
> + fprintf(stderr, "%s refcount block %" PRId64 " is outside image\n",
> + fix & BDRV_FIX_ERRORS ? "Repairing" : "ERROR", i);
> +
> + if (fix & BDRV_FIX_ERRORS) {
> + int64_t old_nb_clusters = *nb_clusters;
> +
> + ret = bdrv_truncate(bs->file, offset + s->cluster_size);
> + if (ret < 0) {
> + goto resize_fail;
> + }
> + size = bdrv_getlength(bs->file);
> + if (size < 0) {
> + ret = size;
> + goto resize_fail;
> + }
> +
> + *nb_clusters = size_to_clusters(s, size);
> + assert(*nb_clusters >= old_nb_clusters);
> +
> + *refcount_table = g_try_realloc(*refcount_table,
> + *nb_clusters * sizeof(uint16_t));
> + if (!*refcount_table) {
> + res->check_errors++;
> + return -ENOMEM;
> + }
> +
> + memset(*refcount_table + old_nb_clusters, 0,
> + (*nb_clusters - old_nb_clusters) * sizeof(uint16_t));
> +
> + if (cluster >= *nb_clusters) {
> + ret = -EINVAL;
> + goto resize_fail;
> + }
> +
> + res->corruptions_fixed++;
> + continue;
> +
> +resize_fail:
> + res->corruptions++;
> + fprintf(stderr, "ERROR could not resize image: %s\n",
> + strerror(-ret));
> + } else {
> + res->corruptions++;
> + }
> continue;
> }
>
> --
> 2.0.4
>
Reviewed-by: Benoît Canet <benoit.canet@nodalink.com>
^ permalink raw reply [flat|nested] 28+ messages in thread
* Re: [Qemu-devel] [PATCH v3 06/10] qcow2: Do not perform potentially damaging repairs
2014-08-22 16:31 ` [Qemu-devel] [PATCH v3 06/10] qcow2: Do not perform potentially damaging repairs Max Reitz
@ 2014-08-22 18:25 ` Benoît Canet
0 siblings, 0 replies; 28+ messages in thread
From: Benoît Canet @ 2014-08-22 18:25 UTC (permalink / raw)
To: Max Reitz; +Cc: Kevin Wolf, qemu-devel, Stefan Hajnoczi, Benoît Canet
On Fri, Aug 22, 2014 at 06:31:40PM +0200, Max Reitz wrote:
> If a referenced cluster has a refcount of 0, increasing its refcount may
> result in clusters being allocated for the refcount structures. This may
> overwrite the referenced cluster, therefore we cannot simply increase
> the refcount then.
>
> In such cases, we can either try to replicate all the refcount
> operations solely for the check operation, basing the allocations on the
> in-memory refcount table; or we can simply rebuild the whole refcount
> structure based on the in-memory refcount table. Since the latter will
> be much easier, do that.
>
> To prepare for this, introduce a "rebuild" boolean which should be set
> to true whenever a fix is rather dangerous or too complicated using the
> current refcount structures. Another example for this is refcount blocks
> being referenced more than once.
>
> Signed-off-by: Max Reitz <mreitz@redhat.com>
> ---
> block/qcow2-refcount.c | 185 ++++++++-----------------------------------------
> 1 file changed, 27 insertions(+), 158 deletions(-)
>
> diff --git a/block/qcow2-refcount.c b/block/qcow2-refcount.c
> index 1f0f44e..242a20c 100644
> --- a/block/qcow2-refcount.c
> +++ b/block/qcow2-refcount.c
> @@ -1382,127 +1382,12 @@ fail:
> }
>
> /*
> - * Writes one sector of the refcount table to the disk
> - */
> -#define RT_ENTRIES_PER_SECTOR (512 / sizeof(uint64_t))
> -static int write_reftable_entry(BlockDriverState *bs, int rt_index)
> -{
> - BDRVQcowState *s = bs->opaque;
> - uint64_t buf[RT_ENTRIES_PER_SECTOR];
> - int rt_start_index;
> - int i, ret;
> -
> - rt_start_index = rt_index & ~(RT_ENTRIES_PER_SECTOR - 1);
> - for (i = 0; i < RT_ENTRIES_PER_SECTOR; i++) {
> - buf[i] = cpu_to_be64(s->refcount_table[rt_start_index + i]);
> - }
> -
> - ret = qcow2_pre_write_overlap_check(bs, QCOW2_OL_REFCOUNT_TABLE,
> - s->refcount_table_offset + rt_start_index * sizeof(uint64_t),
> - sizeof(buf));
> - if (ret < 0) {
> - return ret;
> - }
> -
> - BLKDBG_EVENT(bs->file, BLKDBG_REFTABLE_UPDATE);
> - ret = bdrv_pwrite_sync(bs->file, s->refcount_table_offset +
> - rt_start_index * sizeof(uint64_t), buf, sizeof(buf));
> - if (ret < 0) {
> - return ret;
> - }
> -
> - return 0;
> -}
> -
> -/*
> - * Allocates a new cluster for the given refcount block (represented by its
> - * offset in the image file) and copies the current content there. This function
> - * does _not_ decrement the reference count for the currently occupied cluster.
> - *
> - * This function prints an informative message to stderr on error (and returns
> - * -errno); on success, the offset of the newly allocated cluster is returned.
> - */
> -static int64_t realloc_refcount_block(BlockDriverState *bs, int reftable_index,
> - uint64_t offset)
> -{
> - BDRVQcowState *s = bs->opaque;
> - int64_t new_offset = 0;
> - void *refcount_block = NULL;
> - int ret;
> -
> - /* allocate new refcount block */
> - new_offset = qcow2_alloc_clusters(bs, s->cluster_size);
> - if (new_offset < 0) {
> - fprintf(stderr, "Could not allocate new cluster: %s\n",
> - strerror(-new_offset));
> - ret = new_offset;
> - goto done;
> - }
> -
> - /* fetch current refcount block content */
> - ret = qcow2_cache_get(bs, s->refcount_block_cache, offset, &refcount_block);
> - if (ret < 0) {
> - fprintf(stderr, "Could not fetch refcount block: %s\n", strerror(-ret));
> - goto fail_free_cluster;
> - }
> -
> - /* new block has not yet been entered into refcount table, therefore it is
> - * no refcount block yet (regarding this check) */
> - ret = qcow2_pre_write_overlap_check(bs, 0, new_offset, s->cluster_size);
> - if (ret < 0) {
> - fprintf(stderr, "Could not write refcount block; metadata overlap "
> - "check failed: %s\n", strerror(-ret));
> - /* the image will be marked corrupt, so don't even attempt on freeing
> - * the cluster */
> - goto done;
> - }
> -
> - /* write to new block */
> - ret = bdrv_write(bs->file, new_offset / BDRV_SECTOR_SIZE, refcount_block,
> - s->cluster_sectors);
> - if (ret < 0) {
> - fprintf(stderr, "Could not write refcount block: %s\n", strerror(-ret));
> - goto fail_free_cluster;
> - }
> -
> - /* update refcount table */
> - assert(!offset_into_cluster(s, new_offset));
> - s->refcount_table[reftable_index] = new_offset;
> - ret = write_reftable_entry(bs, reftable_index);
> - if (ret < 0) {
> - fprintf(stderr, "Could not update refcount table: %s\n",
> - strerror(-ret));
> - goto fail_free_cluster;
> - }
> -
> - goto done;
> -
> -fail_free_cluster:
> - qcow2_free_clusters(bs, new_offset, s->cluster_size, QCOW2_DISCARD_OTHER);
> -
> -done:
> - if (refcount_block) {
> - /* This should never fail, as it would only do so if the given refcount
> - * block cannot be found in the cache. As this is impossible as long as
> - * there are no bugs, assert the success. */
> - int tmp = qcow2_cache_put(bs, s->refcount_block_cache, &refcount_block);
> - assert(tmp == 0);
> - }
> -
> - if (ret < 0) {
> - return ret;
> - }
> -
> - return new_offset;
> -}
> -
> -/*
> * Checks consistency of refblocks and accounts for each refblock in
> * *refcount_table.
> */
> static int check_refblocks(BlockDriverState *bs, BdrvCheckResult *res,
> - BdrvCheckMode fix, uint16_t **refcount_table,
> - int64_t *nb_clusters)
> + BdrvCheckMode fix, bool *rebuild,
> + uint16_t **refcount_table, int64_t *nb_clusters)
> {
> BDRVQcowState *s = bs->opaque;
> int64_t i, size;
> @@ -1518,6 +1403,7 @@ static int check_refblocks(BlockDriverState *bs, BdrvCheckResult *res,
> fprintf(stderr, "ERROR refcount block %" PRId64 " is not "
> "cluster aligned; refcount table entry corrupted\n", i);
> res->corruptions++;
> + *rebuild = true;
> continue;
> }
>
> @@ -1561,6 +1447,7 @@ static int check_refblocks(BlockDriverState *bs, BdrvCheckResult *res,
>
> resize_fail:
> res->corruptions++;
> + *rebuild = true;
> fprintf(stderr, "ERROR could not resize image: %s\n",
> strerror(-ret));
> } else {
> @@ -1573,40 +1460,10 @@ resize_fail:
> inc_refcounts(bs, res, *refcount_table, *nb_clusters,
> offset, s->cluster_size);
> if ((*refcount_table)[cluster] != 1) {
> - fprintf(stderr, "%s refcount block %" PRId64
> - " refcount=%d\n",
> - fix & BDRV_FIX_ERRORS ? "Repairing" :
> - "ERROR",
> - i, (*refcount_table)[cluster]);
> -
> - if (fix & BDRV_FIX_ERRORS) {
> - int64_t new_offset;
> -
> - new_offset = realloc_refcount_block(bs, i, offset);
> - if (new_offset < 0) {
> - res->corruptions++;
> - continue;
> - }
> -
> - /* update refcounts */
> - if ((new_offset >> s->cluster_bits) >= *nb_clusters) {
> - /* increase refcount_table size if necessary */
> - int old_nb_clusters = *nb_clusters;
> - *nb_clusters = (new_offset >> s->cluster_bits) + 1;
> - *refcount_table = g_renew(uint16_t, *refcount_table,
> - *nb_clusters);
> - memset(&(*refcount_table)[old_nb_clusters], 0,
> - (*nb_clusters - old_nb_clusters) *
> - sizeof(uint16_t));
> - }
> - (*refcount_table)[cluster]--;
> - inc_refcounts(bs, res, *refcount_table, *nb_clusters,
> - new_offset, s->cluster_size);
> -
> - res->corruptions_fixed++;
> - } else {
> - res->corruptions++;
> - }
> + fprintf(stderr, "ERROR refcount block %" PRId64
> + " refcount=%d\n", i, (*refcount_table)[cluster]);
> + res->corruptions++;
> + *rebuild = true;
> }
> }
> }
> @@ -1618,8 +1475,8 @@ resize_fail:
> * Calculates an in-memory refcount table.
> */
> static int calculate_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
> - BdrvCheckMode fix, uint16_t **refcount_table,
> - int64_t *nb_clusters)
> + BdrvCheckMode fix, bool *rebuild,
> + uint16_t **refcount_table, int64_t *nb_clusters)
> {
> BDRVQcowState *s = bs->opaque;
> int64_t i;
> @@ -1662,7 +1519,7 @@ static int calculate_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
> s->refcount_table_offset,
> s->refcount_table_size * sizeof(uint64_t));
>
> - return check_refblocks(bs, res, fix, refcount_table, nb_clusters);
> + return check_refblocks(bs, res, fix, rebuild, refcount_table, nb_clusters);
> }
>
> /*
> @@ -1670,7 +1527,8 @@ static int calculate_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
> * refcount as reported by the refcount structures on-disk.
> */
> static void compare_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
> - BdrvCheckMode fix, int64_t *highest_cluster,
> + BdrvCheckMode fix, bool *rebuild,
> + int64_t *highest_cluster,
> uint16_t *refcount_table, int64_t nb_clusters)
> {
> BDRVQcowState *s = bs->opaque;
> @@ -1697,10 +1555,15 @@ static void compare_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
> int *num_fixed = NULL;
> if (refcount1 > refcount2 && (fix & BDRV_FIX_LEAKS)) {
> num_fixed = &res->leaks_fixed;
> - } else if (refcount1 < refcount2 && (fix & BDRV_FIX_ERRORS)) {
> + } else if (refcount1 > 0 && refcount1 < refcount2 &&
> + (fix & BDRV_FIX_ERRORS)) {
> num_fixed = &res->corruptions_fixed;
> }
>
> + if (refcount1 == 0) {
> + *rebuild = true;
> + }
> +
> fprintf(stderr, "%s cluster %" PRId64 " refcount=%d reference=%d\n",
> num_fixed != NULL ? "Repairing" :
> refcount1 < refcount2 ? "ERROR" :
> @@ -1739,6 +1602,7 @@ int qcow2_check_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
> BDRVQcowState *s = bs->opaque;
> int64_t size, highest_cluster, nb_clusters;
> uint16_t *refcount_table = NULL;
> + bool rebuild = false;
> int ret;
>
> size = bdrv_getlength(bs->file);
> @@ -1756,14 +1620,19 @@ int qcow2_check_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
> res->bfi.total_clusters =
> size_to_clusters(s, bs->total_sectors * BDRV_SECTOR_SIZE);
>
> - ret = calculate_refcounts(bs, res, fix, &refcount_table, &nb_clusters);
> + ret = calculate_refcounts(bs, res, fix, &rebuild, &refcount_table,
> + &nb_clusters);
> if (ret < 0) {
> goto fail;
> }
>
> - compare_refcounts(bs, res, fix, &highest_cluster, refcount_table,
> + compare_refcounts(bs, res, fix, &rebuild, &highest_cluster, refcount_table,
> nb_clusters);
>
> + if (rebuild) {
> + fprintf(stderr, "ERROR need to rebuild refcount structures\n");
> + }
> +
> /* check OFLAG_COPIED */
> ret = check_oflag_copied(bs, res, fix);
> if (ret < 0) {
> --
> 2.0.4
>
Reviewed-by: Benoît Canet <benoit.canet@nodalink.com>
^ permalink raw reply [flat|nested] 28+ messages in thread
* Re: [Qemu-devel] [PATCH v3 08/10] qcow2: Clean up after refcount rebuild
2014-08-22 16:31 ` [Qemu-devel] [PATCH v3 08/10] qcow2: Clean up after refcount rebuild Max Reitz
@ 2014-08-22 18:43 ` Benoît Canet
0 siblings, 0 replies; 28+ messages in thread
From: Benoît Canet @ 2014-08-22 18:43 UTC (permalink / raw)
To: Max Reitz; +Cc: Kevin Wolf, qemu-devel, Stefan Hajnoczi, Benoît Canet
On Fri, Aug 22, 2014 at 06:31:42PM +0200, Max Reitz wrote:
> Because the old refcount structure will be leaked after having rebuilt
> it, we need to recalculate the refcounts and run a leak-fixing operation
> afterwards (if leaks should be fixed at all).
>
> Signed-off-by: Max Reitz <mreitz@redhat.com>
> ---
> block/qcow2-refcount.c | 35 +++++++++++++++++++++++++++++++++++
> 1 file changed, 35 insertions(+)
>
> diff --git a/block/qcow2-refcount.c b/block/qcow2-refcount.c
> index 59cab65..818facc 100644
> --- a/block/qcow2-refcount.c
> +++ b/block/qcow2-refcount.c
> @@ -1874,12 +1874,47 @@ int qcow2_check_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
> nb_clusters);
>
> if (rebuild && (fix & BDRV_FIX_ERRORS)) {
> + BdrvCheckResult old_res = *res;
> +
> fprintf(stderr, "Rebuilding refcount structure\n");
> ret = rebuild_refcount_structure(bs, res, &refcount_table,
> &nb_clusters);
> if (ret < 0) {
> goto fail;
> }
> +
> + res->corruptions = 0;
> + res->leaks = 0;
> +
> + /* Because the old reftable has been exchanged for a new one the
> + * references have to be recalculated */
> + rebuild = false;
> + memset(refcount_table, 0, nb_clusters * sizeof(uint16_t));
> + ret = calculate_refcounts(bs, res, 0, &rebuild, &refcount_table,
> + &nb_clusters);
> + if (ret < 0) {
> + goto fail;
> + }
> +
> + if (fix & BDRV_FIX_LEAKS) {
> + /* The old refcount structures are now leaked, fix it; the result
> + * can be ignored */
> + pre_compare_res = *res;
> + compare_refcounts(bs, res, BDRV_FIX_LEAKS, &rebuild,
> + &highest_cluster, refcount_table, nb_clusters);
> + if (rebuild) {
> + fprintf(stderr, "ERROR rebuilt refcount structure is still "
> + "broken\n");
> + }
> + *res = pre_compare_res;
> + }
> +
> + if (res->corruptions < old_res.corruptions) {
> + res->corruptions_fixed += old_res.corruptions - res->corruptions;
> + }
> + if (res->leaks < old_res.leaks) {
> + res->leaks_fixed += old_res.leaks - res->leaks;
> + }
> } else if (fix) {
> if (rebuild) {
> fprintf(stderr, "ERROR need to rebuild refcount structures\n");
> --
> 2.0.4
>
Reviewed-by: Benoît Canet <benoit.canet@nodalink.com>
^ permalink raw reply [flat|nested] 28+ messages in thread
* Re: [Qemu-devel] [PATCH v3 01/10] qcow2: Fix leaks in dirty images
2014-08-22 16:31 ` [Qemu-devel] [PATCH v3 01/10] qcow2: Fix leaks in dirty images Max Reitz
@ 2014-08-22 18:44 ` Benoît Canet
0 siblings, 0 replies; 28+ messages in thread
From: Benoît Canet @ 2014-08-22 18:44 UTC (permalink / raw)
To: Max Reitz; +Cc: Kevin Wolf, qemu-devel, Stefan Hajnoczi, Benoît Canet
On Fri, Aug 22, 2014 at 06:31:35PM +0200, Max Reitz wrote:
> When opening dirty images, qcow2's repair function should not only
> repair errors but leaks as well.
>
> Signed-off-by: Max Reitz <mreitz@redhat.com>
> Reviewed-by: Eric Blake <eblake@redhat.com>
> ---
> block/qcow2.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/block/qcow2.c b/block/qcow2.c
> index f9e045f..67d6368 100644
> --- a/block/qcow2.c
> +++ b/block/qcow2.c
> @@ -898,7 +898,7 @@ static int qcow2_open(BlockDriverState *bs, QDict *options, int flags,
> (s->incompatible_features & QCOW2_INCOMPAT_DIRTY)) {
> BdrvCheckResult result = {0};
>
> - ret = qcow2_check(bs, &result, BDRV_FIX_ERRORS);
> + ret = qcow2_check(bs, &result, BDRV_FIX_ERRORS | BDRV_FIX_LEAKS);
> if (ret < 0) {
> error_setg_errno(errp, -ret, "Could not repair dirty image");
> goto fail;
> --
> 2.0.4
>
Reviewed-by: Benoît Canet <benoit.canet@nodalink.com>
^ permalink raw reply [flat|nested] 28+ messages in thread
* Re: [Qemu-devel] [PATCH v3 10/10] iotests: Add test for potentially damaging repairs
2014-08-22 16:31 ` [Qemu-devel] [PATCH v3 10/10] iotests: Add test for potentially damaging repairs Max Reitz
@ 2014-08-22 18:50 ` Benoît Canet
2014-08-22 19:55 ` Eric Blake
1 sibling, 0 replies; 28+ messages in thread
From: Benoît Canet @ 2014-08-22 18:50 UTC (permalink / raw)
To: Max Reitz; +Cc: Kevin Wolf, qemu-devel, Stefan Hajnoczi, Benoît Canet
On Fri, Aug 22, 2014 at 06:31:44PM +0200, Max Reitz wrote:
> There are certain cases where repairing a qcow2 image might actually
> damage it further (or rather, where repairing it has in fact damaged it
> further with the old qcow2 check implementation). This should not
> happen, so add a test for these cases.
>
> Signed-off-by: Max Reitz <mreitz@redhat.com>
> ---
> tests/qemu-iotests/104 | 98 ++++++++++++++++++++++++++++++++++++++++++++++
> tests/qemu-iotests/104.out | 46 ++++++++++++++++++++++
> tests/qemu-iotests/group | 1 +
> 3 files changed, 145 insertions(+)
> create mode 100755 tests/qemu-iotests/104
> create mode 100644 tests/qemu-iotests/104.out
>
> diff --git a/tests/qemu-iotests/104 b/tests/qemu-iotests/104
> new file mode 100755
> index 0000000..cc6d0b3
> --- /dev/null
> +++ b/tests/qemu-iotests/104
> @@ -0,0 +1,98 @@
> +#!/bin/bash
> +#
> +# Test case for repairing qcow2 images which cannot be repaired using
> +# the on-disk refcount structures
> +#
> +# Copyright (C) 2013 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=mreitz@redhat.com
> +
> +seq="$(basename $0)"
> +echo "QA output created by $seq"
> +
> +here="$PWD"
> +tmp=/tmp/$$
> +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
> +
> +# This tests qocw2-specific low-level functionality
> +_supported_fmt qcow2
> +_supported_proto file
> +_supported_os Linux
> +
> +echo
> +echo '=== Repairing an image without any refcount table ==='
> +echo
> +
> +_make_test_img 64M
> +# just write some data
> +$QEMU_IO -c 'write -P 42 0 64k' "$TEST_IMG" | _filter_qemu_io
> +
> +# refcount_table_offset
> +poke_file "$TEST_IMG" 48 "\x00\x00\x00\x00\x00\x00\x00\x00"
> +# refcount_table_clusters
> +poke_file "$TEST_IMG" 56 "\x00\x00\x00\x00"
> +
> +_check_test_img -r all
> +
> +$QEMU_IO -c 'read -P 42 0 64k' "$TEST_IMG" | _filter_qemu_io
> +
> +echo
> +echo '=== Repairing unreferenced data cluster in new refblock area ==='
> +echo
> +
> +IMGOPTS='cluster_size=512' _make_test_img 64M
> +# Allocate the first 128 kB in the image (first refblock)
> +$QEMU_IO -c 'write 0 111104' "$TEST_IMG" | _filter_qemu_io
> +# should be 131072
> +stat -c '%s' "$TEST_IMG"
> +
> +# Enter a cluster at 128 kB (0x20000)
> +# XXX: This (0x1ccc8) should be the first free entry in the last L2 table, but
> +# we cannot be sure
> +poke_file "$TEST_IMG" 117960 "\x80\x00\x00\x00\x00\x02\x00\x00"
> +
> +# Fill the cluster
> +truncate -s 131584 "$TEST_IMG"
> +$QEMU_IO -c "open -o driver=raw $TEST_IMG" -c 'write -P 42 128k 512' \
> + | _filter_qemu_io
> +
> +# The data should now appear at this guest offset
> +$QEMU_IO -c 'read -P 42 111104 512' "$TEST_IMG" | _filter_qemu_io
> +
> +# This cluster is unallocated; fix it
> +_check_test_img -r all
> +
> +# This repair operation must have allocated a new refblock; and that refblock
> +# should not overlap with the unallocated data cluster. If it does, the data
> +# will be damaged, so check it.
> +$QEMU_IO -c 'read -P 42 111104 512' "$TEST_IMG" | _filter_qemu_io
> +
> +# success, all done
> +echo '*** done'
> +rm -f $seq.full
> +status=0
> diff --git a/tests/qemu-iotests/104.out b/tests/qemu-iotests/104.out
> new file mode 100644
> index 0000000..742fe77
> --- /dev/null
> +++ b/tests/qemu-iotests/104.out
> @@ -0,0 +1,46 @@
> +QA output created by 104
> +
> +=== Repairing an image without any refcount table ===
> +
> +Formatting 'TEST_DIR/t.IMGFMT', fmt=IMGFMT size=67108864
> +wrote 65536/65536 bytes at offset 0
> +64 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
> +ERROR cluster 0 refcount=0 reference=1
> +ERROR cluster 3 refcount=0 reference=1
> +ERROR cluster 4 refcount=0 reference=1
> +ERROR cluster 5 refcount=0 reference=1
> +Rebuilding refcount structure
> +The following inconsistencies were found and repaired:
> +
> + 0 leaked clusters
> + 4 corruptions
> +
> +Double checking the fixed image now...
> +No errors were found on the image.
> +read 65536/65536 bytes at offset 0
> +64 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
> +
> +=== Repairing unreferenced data cluster in new refblock area ===
> +
> +Formatting 'TEST_DIR/t.IMGFMT', fmt=IMGFMT size=67108864
> +wrote 111104/111104 bytes at offset 0
> +108.500 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
> +131072
> +wrote 512/512 bytes at offset 131072
> +512 bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
> +read 512/512 bytes at offset 111104
> +512 bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
> +ERROR cluster 256 refcount=0 reference=1
> +Rebuilding refcount structure
> +Repairing cluster 1 refcount=1 reference=0
> +Repairing cluster 2 refcount=1 reference=0
> +The following inconsistencies were found and repaired:
> +
> + 0 leaked clusters
> + 1 corruptions
> +
> +Double checking the fixed image now...
> +No errors were found on the image.
> +read 512/512 bytes at offset 111104
> +512 bytes, 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 2803d68..5a3f9dd 100644
> --- a/tests/qemu-iotests/group
> +++ b/tests/qemu-iotests/group
> @@ -103,3 +103,4 @@
> 099 rw auto quick
> 101 rw auto quick
> 103 rw auto quick
> +104 rw auto quick
> --
> 2.0.4
>
Reviewed-by: Benoît Canet <benoit.canet@nodalink.com>
^ permalink raw reply [flat|nested] 28+ messages in thread
* Re: [Qemu-devel] [PATCH v3 10/10] iotests: Add test for potentially damaging repairs
2014-08-22 16:31 ` [Qemu-devel] [PATCH v3 10/10] iotests: Add test for potentially damaging repairs Max Reitz
2014-08-22 18:50 ` Benoît Canet
@ 2014-08-22 19:55 ` Eric Blake
2014-08-22 20:55 ` Benoît Canet
2014-08-26 17:53 ` Max Reitz
1 sibling, 2 replies; 28+ messages in thread
From: Eric Blake @ 2014-08-22 19:55 UTC (permalink / raw)
To: Max Reitz, qemu-devel; +Cc: Kevin Wolf, Benoît Canet, Stefan Hajnoczi
[-- Attachment #1: Type: text/plain, Size: 1234 bytes --]
On 08/22/2014 10:31 AM, Max Reitz wrote:
> There are certain cases where repairing a qcow2 image might actually
> damage it further (or rather, where repairing it has in fact damaged it
> further with the old qcow2 check implementation). This should not
> happen, so add a test for these cases.
>
> Signed-off-by: Max Reitz <mreitz@redhat.com>
> ---
> tests/qemu-iotests/104 | 98 ++++++++++++++++++++++++++++++++++++++++++++++
> tests/qemu-iotests/104.out | 46 ++++++++++++++++++++++
> tests/qemu-iotests/group | 1 +
> 3 files changed, 145 insertions(+)
> create mode 100755 tests/qemu-iotests/104
> create mode 100644 tests/qemu-iotests/104.out
>
> diff --git a/tests/qemu-iotests/104 b/tests/qemu-iotests/104
> new file mode 100755
> index 0000000..cc6d0b3
> --- /dev/null
> +++ b/tests/qemu-iotests/104
> @@ -0,0 +1,98 @@
> +#!/bin/bash
> +#
> +# Test case for repairing qcow2 images which cannot be repaired using
> +# the on-disk refcount structures
> +#
> +# Copyright (C) 2013 Red Hat, Inc.
Welcome to 2014, where we are practically 2/3 through with the year
already :)
--
Eric Blake eblake redhat com +1-919-301-3266
Libvirt virtualization library http://libvirt.org
[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 539 bytes --]
^ permalink raw reply [flat|nested] 28+ messages in thread
* Re: [Qemu-devel] [PATCH v3 10/10] iotests: Add test for potentially damaging repairs
2014-08-22 19:55 ` Eric Blake
@ 2014-08-22 20:55 ` Benoît Canet
2014-08-26 17:53 ` Max Reitz
1 sibling, 0 replies; 28+ messages in thread
From: Benoît Canet @ 2014-08-22 20:55 UTC (permalink / raw)
To: Eric Blake
Cc: Kevin Wolf, Benoît Canet, qemu-devel, Stefan Hajnoczi, Max Reitz
On Fri, Aug 22, 2014 at 01:55:01PM -0600, Eric Blake wrote:
> On 08/22/2014 10:31 AM, Max Reitz wrote:
> > There are certain cases where repairing a qcow2 image might actually
> > damage it further (or rather, where repairing it has in fact damaged it
> > further with the old qcow2 check implementation). This should not
> > happen, so add a test for these cases.
> >
> > Signed-off-by: Max Reitz <mreitz@redhat.com>
> > ---
> > tests/qemu-iotests/104 | 98 ++++++++++++++++++++++++++++++++++++++++++++++
> > tests/qemu-iotests/104.out | 46 ++++++++++++++++++++++
> > tests/qemu-iotests/group | 1 +
> > 3 files changed, 145 insertions(+)
> > create mode 100755 tests/qemu-iotests/104
> > create mode 100644 tests/qemu-iotests/104.out
> >
> > diff --git a/tests/qemu-iotests/104 b/tests/qemu-iotests/104
> > new file mode 100755
> > index 0000000..cc6d0b3
> > --- /dev/null
> > +++ b/tests/qemu-iotests/104
> > @@ -0,0 +1,98 @@
> > +#!/bin/bash
> > +#
> > +# Test case for repairing qcow2 images which cannot be repaired using
> > +# the on-disk refcount structures
> > +#
> > +# Copyright (C) 2013 Red Hat, Inc.
>
> Welcome to 2014, where we are practically 2/3 through with the year
> already :)
It may have been copied from an existing template so maybe the initial copyright
is really from 2013.
Benoît
>
> --
> Eric Blake eblake redhat com +1-919-301-3266
> Libvirt virtualization library http://libvirt.org
>
^ permalink raw reply [flat|nested] 28+ messages in thread
* Re: [Qemu-devel] [PATCH v3 07/10] qcow2: Rebuild refcount structure during check
2014-08-22 16:31 ` [Qemu-devel] [PATCH v3 07/10] qcow2: Rebuild refcount structure during check Max Reitz
@ 2014-08-25 17:40 ` Benoît Canet
2014-08-27 18:37 ` Max Reitz
0 siblings, 1 reply; 28+ messages in thread
From: Benoît Canet @ 2014-08-25 17:40 UTC (permalink / raw)
To: Max Reitz; +Cc: Kevin Wolf, qemu-devel, Stefan Hajnoczi, Benoît Canet
On Fri, Aug 22, 2014 at 06:31:41PM +0200, Max Reitz wrote:
> The previous commit introduced the "rebuild" variable to qcow2's
> implementation of the image consistency check. Now make use of this by
> adding a function which creates a completely new refcount structure
> based solely on the in-memory information gathered before.
>
> The old refcount structure will be leaked, however.
>
> Signed-off-by: Max Reitz <mreitz@redhat.com>
> ---
> block/qcow2-refcount.c | 265 ++++++++++++++++++++++++++++++++++++++++++++++++-
> 1 file changed, 262 insertions(+), 3 deletions(-)
>
> diff --git a/block/qcow2-refcount.c b/block/qcow2-refcount.c
> index 242a20c..59cab65 100644
> --- a/block/qcow2-refcount.c
> +++ b/block/qcow2-refcount.c
> @@ -1591,6 +1591,245 @@ static void compare_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
> }
>
> /*
> + * Allocates a cluster using an in-memory refcount table (IMRT) in contrast to
> + * the on-disk refcount structures.
> + *
> + * *first_free_cluster does not necessarily point to the first free cluster, but
> + * may point to one cluster as close as possible before it. The offset returned
> + * will never be before that cluster.
> + *
> + * Note that *first_free_cluster is a cluster index whereas the return value is
> + * an offset.
> + */
> +static int64_t alloc_clusters_imrt(BlockDriverState *bs,
> + int cluster_count,
> + uint16_t **refcount_table,
> + int64_t *nb_clusters,
> + int64_t *first_free_cluster)
> +{
> + BDRVQcowState *s = bs->opaque;
> + int64_t cluster = *first_free_cluster, i;
> + bool first_gap = true;
> + int contiguous_clusters;
> +
> + for (contiguous_clusters = 0;
> + cluster < *nb_clusters && contiguous_clusters < cluster_count;
> + cluster++)
> + {
> + if (!(*refcount_table)[cluster]) {
> + contiguous_clusters++;
> + if (first_gap) {
> + *first_free_cluster = cluster;
> + first_gap = false;
> + }
> + } else if (contiguous_clusters) {
> + contiguous_clusters = 0;
> + }
> + }
s/contignuous_clusters/contiguous_free_clusters/
I think I understood the above block better than the first time but I it would
worth a comment on it's rather complex purpose.
> +
> + if (contiguous_clusters < cluster_count) {
> + int64_t old_nb_clusters = *nb_clusters;
> +
> + *nb_clusters = cluster + cluster_count - contiguous_clusters;
> + *refcount_table = g_try_realloc(*refcount_table,
> + *nb_clusters * sizeof(uint16_t));
> + if (!*refcount_table) {
> + return -ENOMEM;
> + }
> +
> + memset(*refcount_table + old_nb_clusters, 0,
> + (*nb_clusters - old_nb_clusters) * sizeof(uint16_t));
> + }
> +
> + cluster -= contiguous_clusters;
> + for (i = 0; i < cluster_count; i++) {
> + (*refcount_table)[cluster + i] = 1;
> + }
> +
> + return cluster << s->cluster_bits;
> +}
> +
> +/*
> + * Creates a new refcount structure based solely on the in-memory information
> + * given through *refcount_table. All necessary allocations will be reflected
> + * in that array.
> + *
> + * On success, the old refcount structure is leaked (it will be covered by the
> + * new refcount structure).
> + */
> +static int rebuild_refcount_structure(BlockDriverState *bs,
> + BdrvCheckResult *res,
> + uint16_t **refcount_table,
> + int64_t *nb_clusters)
> +{
> + BDRVQcowState *s = bs->opaque;
> + int64_t first_free_cluster = 0, rt_ofs = -1, cluster = 0;
/* rb stands for refcount block and rt stands for refcount table */
> + int64_t rb_ofs, rb_start, rb_index;
> + uint32_t reftable_size = 0;
> + uint64_t *reftable = NULL;
> + uint16_t *on_disk_rb;
> + uint8_t rt_offset_and_clusters[sizeof(uint64_t) + sizeof(uint32_t)];
> + int i, ret = 0;
> +
> + qcow2_cache_empty(bs, s->refcount_block_cache);
> +
> +write_refblocks:
> + for (; cluster < *nb_clusters; cluster++) {
> + if (!(*refcount_table)[cluster]) {
> + continue;
> + }
> +
> + rb_index = cluster >> (s->cluster_bits - 1);
> + rb_start = rb_index << (s->cluster_bits - 1);
> +
> + /* Don't allocate a cluster in a refblock already written to disk */
> + if (first_free_cluster < rb_start) {
> + first_free_cluster = rb_start;
> + }
> + rb_ofs = alloc_clusters_imrt(bs, 1, refcount_table, nb_clusters,
> + &first_free_cluster);
> + if (rb_ofs < 0) {
> + fprintf(stderr, "ERROR allocating refblock: %s\n", strerror(-ret));
> + res->check_errors++;
> + ret = rb_ofs;
> + goto fail;
> + }
> +
> + if (reftable_size <= rb_index) {
> + uint32_t old_rt_size = reftable_size;
> + reftable_size = ROUND_UP((rb_index + 1) * sizeof(uint64_t),
> + s->cluster_size) / sizeof(uint64_t);
> + reftable = g_try_realloc(reftable,
> + reftable_size * sizeof(uint64_t));
> + if (!reftable) {
> + res->check_errors++;
> + ret = -ENOMEM;
> + goto fail;
> + }
> +
> + memset(reftable + old_rt_size, 0,
> + (reftable_size - old_rt_size) * sizeof(uint64_t));
> +
> + /* The offset we have for the reftable is now no longer valid;
> + * this will leak that range, but we can easily fix that by running
> + * a leak-fixing check after this rebuild operation */
> + rt_ofs = -1;
> + }
> + reftable[rb_index] = rb_ofs;
> +
> + /* If this is apparently the last refblock (for now), try to squeeze the
> + * reftable in */
> + if (rb_index == (*nb_clusters - 1) >> (s->cluster_bits - 1)
^ I don't understand the - 1 here.
> &&
> + rt_ofs < 0)
> + {
> + rt_ofs = alloc_clusters_imrt(bs, size_to_clusters(s, reftable_size *
> + sizeof(uint64_t)),
> + refcount_table, nb_clusters,
> + &first_free_cluster);
> + if (rt_ofs < 0) {
> + fprintf(stderr, "ERROR allocating reftable: %s\n",
> + strerror(-ret));
> + res->check_errors++;
> + ret = rt_ofs;
> + goto fail;
> + }
> + }
> +
> + ret = qcow2_pre_write_overlap_check(bs, 0, rb_ofs, s->cluster_size);
> + if (ret < 0) {
> + fprintf(stderr, "ERROR writing refblock: %s\n", strerror(-ret));
> + goto fail;
> + }
> +
> + on_disk_rb = g_malloc0(s->cluster_size);
> + for (i = 0; i < s->cluster_size / sizeof(uint16_t) &&
> + rb_start + i < *nb_clusters; i++)
> + {
> + on_disk_rb[i] = cpu_to_be16((*refcount_table)[rb_start + i]);
> + }
> +
> + ret = bdrv_write(bs->file, rb_ofs / BDRV_SECTOR_SIZE,
> + (void *)on_disk_rb, s->cluster_sectors);
> + g_free(on_disk_rb);
> + if (ret < 0) {
> + fprintf(stderr, "ERROR writing refblock: %s\n", strerror(-ret));
> + goto fail;
> + }
> +
> + /* Go to the end of this refblock */
> + cluster = rb_start + s->cluster_size / sizeof(uint16_t) - 1;
> + }
> +
> + if (rt_ofs < 0) {
> + int64_t post_rb_start = ROUND_UP(*nb_clusters,
> + s->cluster_size / sizeof(uint16_t));
> +
> + /* Not pretty but simple */
> + if (first_free_cluster < post_rb_start) {
> + first_free_cluster = post_rb_start;
> + }
> + rt_ofs = alloc_clusters_imrt(bs, size_to_clusters(s, reftable_size *
> + sizeof(uint64_t)),
> + refcount_table, nb_clusters,
> + &first_free_cluster);
> + if (rt_ofs < 0) {
> + fprintf(stderr, "ERROR allocating reftable: %s\n", strerror(-ret));
> + res->check_errors++;
> + ret = rt_ofs;
> + goto fail;
> + }
> +
> + goto write_refblocks;
> + }
> +
> + assert(reftable);
> +
> + for (rb_index = 0; rb_index < reftable_size; rb_index++) {
> + cpu_to_be64s(&reftable[rb_index]);
> + }
> +
> + ret = qcow2_pre_write_overlap_check(bs, 0, rt_ofs,
> + reftable_size * sizeof(uint64_t));
> + if (ret < 0) {
> + fprintf(stderr, "ERROR writing reftable: %s\n", strerror(-ret));
> + goto fail;
> + }
> +
> + ret = bdrv_write(bs->file, rt_ofs / BDRV_SECTOR_SIZE, (void *)reftable,
> + reftable_size * sizeof(uint64_t) / BDRV_SECTOR_SIZE);
> + if (ret < 0) {
> + fprintf(stderr, "ERROR writing reftable: %s\n", strerror(-ret));
> + goto fail;
> + }
> +
> + /* Enter new reftable into the image header */
> + cpu_to_be64w((uint64_t *)&rt_offset_and_clusters[0], rt_ofs);
> + cpu_to_be32w((uint32_t *)&rt_offset_and_clusters[sizeof(uint64_t)],
> + size_to_clusters(s, reftable_size * sizeof(uint64_t)));
> + ret = bdrv_pwrite_sync(bs->file, offsetof(QCowHeader,
> + refcount_table_offset),
> + rt_offset_and_clusters,
> + sizeof(rt_offset_and_clusters));
> + if (ret < 0) {
> + fprintf(stderr, "ERROR setting reftable: %s\n", strerror(-ret));
> + goto fail;
> + }
> +
> + for (rb_index = 0; rb_index < reftable_size; rb_index++) {
> + be64_to_cpus(&reftable[rb_index]);
> + }
> + s->refcount_table = reftable;
> + s->refcount_table_offset = rt_ofs;
> + s->refcount_table_size = reftable_size;
> +
> + return 0;
> +
> +fail:
> + g_free(reftable);
> + return ret;
> +}
> +
> +/*
> * Checks an image for refcount consistency.
> *
> * Returns 0 if no errors are found, the number of errors in case the image is
> @@ -1600,6 +1839,7 @@ int qcow2_check_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
> BdrvCheckMode fix)
> {
> BDRVQcowState *s = bs->opaque;
> + BdrvCheckResult pre_compare_res;
> int64_t size, highest_cluster, nb_clusters;
> uint16_t *refcount_table = NULL;
> bool rebuild = false;
> @@ -1626,11 +1866,30 @@ int qcow2_check_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
> goto fail;
> }
>
> - compare_refcounts(bs, res, fix, &rebuild, &highest_cluster, refcount_table,
> + /* In case we don't need to rebuild the refcount structure (but want to fix
> + * something), this function is immediately called again, in which case the
> + * result should be ignored */
> + pre_compare_res = *res;
> + compare_refcounts(bs, res, 0, &rebuild, &highest_cluster, refcount_table,
> nb_clusters);
>
> - if (rebuild) {
> - fprintf(stderr, "ERROR need to rebuild refcount structures\n");
> + if (rebuild && (fix & BDRV_FIX_ERRORS)) {
> + fprintf(stderr, "Rebuilding refcount structure\n");
> + ret = rebuild_refcount_structure(bs, res, &refcount_table,
> + &nb_clusters);
> + if (ret < 0) {
> + goto fail;
> + }
> + } else if (fix) {
> + if (rebuild) {
> + fprintf(stderr, "ERROR need to rebuild refcount structures\n");
> + }
> +
> + if (res->leaks || res->corruptions) {
> + *res = pre_compare_res;
> + compare_refcounts(bs, res, fix, &rebuild, &highest_cluster,
> + refcount_table, nb_clusters);
> + }
> }
>
> /* check OFLAG_COPIED */
> --
> 2.0.4
>
^ permalink raw reply [flat|nested] 28+ messages in thread
* Re: [Qemu-devel] [PATCH v3 05/10] qcow2: Fix refcount blocks beyond image end
2014-08-22 16:31 ` [Qemu-devel] [PATCH v3 05/10] qcow2: Fix refcount blocks beyond image end Max Reitz
2014-08-22 18:20 ` Benoît Canet
@ 2014-08-26 13:07 ` Eric Blake
2014-08-26 18:06 ` Max Reitz
1 sibling, 1 reply; 28+ messages in thread
From: Eric Blake @ 2014-08-26 13:07 UTC (permalink / raw)
To: Max Reitz, qemu-devel; +Cc: Kevin Wolf, Benoît Canet, Stefan Hajnoczi
[-- Attachment #1: Type: text/plain, Size: 621 bytes --]
On 08/22/2014 10:31 AM, Max Reitz wrote:
> If the qcow2 check function detects a refcount block located beyond the
> image end, grow the image appropriately. This cannot break anything and
> is the logical fix for such a case.
Does the testsuite cover this one? I didn't see it mentioned in patch 10/10.
>
> Signed-off-by: Max Reitz <mreitz@redhat.com>
> ---
> block/qcow2-refcount.c | 50 ++++++++++++++++++++++++++++++++++++++++++++++----
> 1 file changed, 46 insertions(+), 4 deletions(-)
>
--
Eric Blake eblake redhat com +1-919-301-3266
Libvirt virtualization library http://libvirt.org
[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 539 bytes --]
^ permalink raw reply [flat|nested] 28+ messages in thread
* Re: [Qemu-devel] [PATCH v3 09/10] iotests: Fix test outputs
2014-08-22 16:31 ` [Qemu-devel] [PATCH v3 09/10] iotests: Fix test outputs Max Reitz
@ 2014-08-26 13:09 ` Eric Blake
2014-08-26 18:03 ` Max Reitz
0 siblings, 1 reply; 28+ messages in thread
From: Eric Blake @ 2014-08-26 13:09 UTC (permalink / raw)
To: Max Reitz, qemu-devel; +Cc: Kevin Wolf, Benoît Canet, Stefan Hajnoczi
[-- Attachment #1: Type: text/plain, Size: 678 bytes --]
On 08/22/2014 10:31 AM, Max Reitz wrote:
> 039, 060 and 061 all create images with referenced clusters having a
> refcount of 0. Because previous commits changed handling of such errors,
> these tests now have a different output. Fix it.
>
> Furthermore, 060 created a refblock with a refcount greater than one
> which now results in having to rebuild the refcount structure as well.
For bisection purposes, should these test tweaks be made in the same
patch that causes the improved behavior?
>
> Signed-off-by: Max Reitz <mreitz@redhat.com>
> ---
--
Eric Blake eblake redhat com +1-919-301-3266
Libvirt virtualization library http://libvirt.org
[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 539 bytes --]
^ permalink raw reply [flat|nested] 28+ messages in thread
* Re: [Qemu-devel] [PATCH v3 10/10] iotests: Add test for potentially damaging repairs
2014-08-22 19:55 ` Eric Blake
2014-08-22 20:55 ` Benoît Canet
@ 2014-08-26 17:53 ` Max Reitz
1 sibling, 0 replies; 28+ messages in thread
From: Max Reitz @ 2014-08-26 17:53 UTC (permalink / raw)
To: Eric Blake, qemu-devel; +Cc: Kevin Wolf, Benoît Canet, Stefan Hajnoczi
On 22.08.2014 21:55, Eric Blake wrote:
> On 08/22/2014 10:31 AM, Max Reitz wrote:
>> There are certain cases where repairing a qcow2 image might actually
>> damage it further (or rather, where repairing it has in fact damaged it
>> further with the old qcow2 check implementation). This should not
>> happen, so add a test for these cases.
>>
>> Signed-off-by: Max Reitz <mreitz@redhat.com>
>> ---
>> tests/qemu-iotests/104 | 98 ++++++++++++++++++++++++++++++++++++++++++++++
>> tests/qemu-iotests/104.out | 46 ++++++++++++++++++++++
>> tests/qemu-iotests/group | 1 +
>> 3 files changed, 145 insertions(+)
>> create mode 100755 tests/qemu-iotests/104
>> create mode 100644 tests/qemu-iotests/104.out
>>
>> diff --git a/tests/qemu-iotests/104 b/tests/qemu-iotests/104
>> new file mode 100755
>> index 0000000..cc6d0b3
>> --- /dev/null
>> +++ b/tests/qemu-iotests/104
>> @@ -0,0 +1,98 @@
>> +#!/bin/bash
>> +#
>> +# Test case for repairing qcow2 images which cannot be repaired using
>> +# the on-disk refcount structures
>> +#
>> +# Copyright (C) 2013 Red Hat, Inc.
> Welcome to 2014, where we are practically 2/3 through with the year
> already :)
Oops *g*
And I thought it was bad when I forgot which day of the week it was...
Max
^ permalink raw reply [flat|nested] 28+ messages in thread
* Re: [Qemu-devel] [PATCH v3 09/10] iotests: Fix test outputs
2014-08-26 13:09 ` Eric Blake
@ 2014-08-26 18:03 ` Max Reitz
0 siblings, 0 replies; 28+ messages in thread
From: Max Reitz @ 2014-08-26 18:03 UTC (permalink / raw)
To: Eric Blake, qemu-devel; +Cc: Kevin Wolf, Benoît Canet, Stefan Hajnoczi
On 26.08.2014 15:09, Eric Blake wrote:
> On 08/22/2014 10:31 AM, Max Reitz wrote:
>> 039, 060 and 061 all create images with referenced clusters having a
>> refcount of 0. Because previous commits changed handling of such errors,
>> these tests now have a different output. Fix it.
>>
>> Furthermore, 060 created a refblock with a refcount greater than one
>> which now results in having to rebuild the refcount structure as well.
> For bisection purposes, should these test tweaks be made in the same
> patch that causes the improved behavior?
We could do that, but patch 6 changed the behavior to "Report an error
instead of trying to fix it", 7 then changed it to "Fix it, but leave
leaks behind" and finally 8 changed it to "Fix it and fix the leaks". It
would be possible to add these changes to each of these patches, but it
would look pretty strange in my opinion.
Benoît commented on this in patch 5 in v1 (which is patch 6 here). I
replied to this in the cover letter of v2: I don't think it really
breaks bisects because the tests were kind of faulty before this series
anyway. They assumed qcow2's repair function would repair blocks with a
refcount of 0 by simply increasing the refcount through the standard
qcow2 functions, which, as this patch 10 shows, will not always work as
intended.
Max
^ permalink raw reply [flat|nested] 28+ messages in thread
* Re: [Qemu-devel] [PATCH v3 05/10] qcow2: Fix refcount blocks beyond image end
2014-08-26 13:07 ` Eric Blake
@ 2014-08-26 18:06 ` Max Reitz
0 siblings, 0 replies; 28+ messages in thread
From: Max Reitz @ 2014-08-26 18:06 UTC (permalink / raw)
To: Eric Blake, qemu-devel; +Cc: Kevin Wolf, Benoît Canet, Stefan Hajnoczi
On 26.08.2014 15:07, Eric Blake wrote:
> On 08/22/2014 10:31 AM, Max Reitz wrote:
>> If the qcow2 check function detects a refcount block located beyond the
>> image end, grow the image appropriately. This cannot break anything and
>> is the logical fix for such a case.
> Does the testsuite cover this one? I didn't see it mentioned in patch 10/10.
No, it doesn't. I'll add it.
Max
>> Signed-off-by: Max Reitz <mreitz@redhat.com>
>> ---
>> block/qcow2-refcount.c | 50 ++++++++++++++++++++++++++++++++++++++++++++++----
>> 1 file changed, 46 insertions(+), 4 deletions(-)
>>
^ permalink raw reply [flat|nested] 28+ messages in thread
* Re: [Qemu-devel] [PATCH v3 07/10] qcow2: Rebuild refcount structure during check
2014-08-25 17:40 ` Benoît Canet
@ 2014-08-27 18:37 ` Max Reitz
0 siblings, 0 replies; 28+ messages in thread
From: Max Reitz @ 2014-08-27 18:37 UTC (permalink / raw)
To: Benoît Canet; +Cc: Kevin Wolf, qemu-devel, Stefan Hajnoczi
On 25.08.2014 19:40, Benoît Canet wrote:
> On Fri, Aug 22, 2014 at 06:31:41PM +0200, Max Reitz wrote:
>> The previous commit introduced the "rebuild" variable to qcow2's
>> implementation of the image consistency check. Now make use of this by
>> adding a function which creates a completely new refcount structure
>> based solely on the in-memory information gathered before.
>>
>> The old refcount structure will be leaked, however.
>>
>> Signed-off-by: Max Reitz <mreitz@redhat.com>
>> ---
>> block/qcow2-refcount.c | 265 ++++++++++++++++++++++++++++++++++++++++++++++++-
>> 1 file changed, 262 insertions(+), 3 deletions(-)
>>
>> diff --git a/block/qcow2-refcount.c b/block/qcow2-refcount.c
>> index 242a20c..59cab65 100644
>> --- a/block/qcow2-refcount.c
>> +++ b/block/qcow2-refcount.c
>> @@ -1591,6 +1591,245 @@ static void compare_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
>> }
>>
>> /*
>> + * Allocates a cluster using an in-memory refcount table (IMRT) in contrast to
>> + * the on-disk refcount structures.
>> + *
>> + * *first_free_cluster does not necessarily point to the first free cluster, but
>> + * may point to one cluster as close as possible before it. The offset returned
>> + * will never be before that cluster.
>> + *
>> + * Note that *first_free_cluster is a cluster index whereas the return value is
>> + * an offset.
>> + */
>> +static int64_t alloc_clusters_imrt(BlockDriverState *bs,
>> + int cluster_count,
>> + uint16_t **refcount_table,
>> + int64_t *nb_clusters,
>> + int64_t *first_free_cluster)
>> +{
>> + BDRVQcowState *s = bs->opaque;
>> + int64_t cluster = *first_free_cluster, i;
>> + bool first_gap = true;
>> + int contiguous_clusters;
>> +
>
>> + for (contiguous_clusters = 0;
>> + cluster < *nb_clusters && contiguous_clusters < cluster_count;
>> + cluster++)
>> + {
>> + if (!(*refcount_table)[cluster]) {
>> + contiguous_clusters++;
>> + if (first_gap) {
>> + *first_free_cluster = cluster;
>> + first_gap = false;
>> + }
>> + } else if (contiguous_clusters) {
>> + contiguous_clusters = 0;
>> + }
>> + }
> s/contignuous_clusters/contiguous_free_clusters/
>
> I think I understood the above block better than the first time but I it would
> worth a comment on it's rather complex purpose.
>
>> +
>> + if (contiguous_clusters < cluster_count) {
>> + int64_t old_nb_clusters = *nb_clusters;
>> +
>> + *nb_clusters = cluster + cluster_count - contiguous_clusters;
>> + *refcount_table = g_try_realloc(*refcount_table,
>> + *nb_clusters * sizeof(uint16_t));
>> + if (!*refcount_table) {
>> + return -ENOMEM;
>> + }
>> +
>> + memset(*refcount_table + old_nb_clusters, 0,
>> + (*nb_clusters - old_nb_clusters) * sizeof(uint16_t));
>> + }
>> +
>> + cluster -= contiguous_clusters;
>> + for (i = 0; i < cluster_count; i++) {
>> + (*refcount_table)[cluster + i] = 1;
>> + }
>> +
>> + return cluster << s->cluster_bits;
>> +}
>> +
>> +/*
>> + * Creates a new refcount structure based solely on the in-memory information
>> + * given through *refcount_table. All necessary allocations will be reflected
>> + * in that array.
>> + *
>> + * On success, the old refcount structure is leaked (it will be covered by the
>> + * new refcount structure).
>> + */
>> +static int rebuild_refcount_structure(BlockDriverState *bs,
>> + BdrvCheckResult *res,
>> + uint16_t **refcount_table,
>> + int64_t *nb_clusters)
>> +{
>> + BDRVQcowState *s = bs->opaque;
>> + int64_t first_free_cluster = 0, rt_ofs = -1, cluster = 0;
> /* rb stands for refcount block and rt stands for refcount table */
>
>> + int64_t rb_ofs, rb_start, rb_index;
>> + uint32_t reftable_size = 0;
>> + uint64_t *reftable = NULL;
>> + uint16_t *on_disk_rb;
>> + uint8_t rt_offset_and_clusters[sizeof(uint64_t) + sizeof(uint32_t)];
>> + int i, ret = 0;
>> +
>> + qcow2_cache_empty(bs, s->refcount_block_cache);
>> +
>> +write_refblocks:
>> + for (; cluster < *nb_clusters; cluster++) {
>> + if (!(*refcount_table)[cluster]) {
>> + continue;
>> + }
>> +
>> + rb_index = cluster >> (s->cluster_bits - 1);
>> + rb_start = rb_index << (s->cluster_bits - 1);
>> +
>> + /* Don't allocate a cluster in a refblock already written to disk */
>> + if (first_free_cluster < rb_start) {
>> + first_free_cluster = rb_start;
>> + }
>> + rb_ofs = alloc_clusters_imrt(bs, 1, refcount_table, nb_clusters,
>> + &first_free_cluster);
>> + if (rb_ofs < 0) {
>> + fprintf(stderr, "ERROR allocating refblock: %s\n", strerror(-ret));
>> + res->check_errors++;
>> + ret = rb_ofs;
>> + goto fail;
>> + }
>> +
>> + if (reftable_size <= rb_index) {
>> + uint32_t old_rt_size = reftable_size;
>> + reftable_size = ROUND_UP((rb_index + 1) * sizeof(uint64_t),
>> + s->cluster_size) / sizeof(uint64_t);
>> + reftable = g_try_realloc(reftable,
>> + reftable_size * sizeof(uint64_t));
>> + if (!reftable) {
>> + res->check_errors++;
>> + ret = -ENOMEM;
>> + goto fail;
>> + }
>> +
>> + memset(reftable + old_rt_size, 0,
>> + (reftable_size - old_rt_size) * sizeof(uint64_t));
>> +
>> + /* The offset we have for the reftable is now no longer valid;
>> + * this will leak that range, but we can easily fix that by running
>> + * a leak-fixing check after this rebuild operation */
>> + rt_ofs = -1;
>> + }
>> + reftable[rb_index] = rb_ofs;
>> +
>> + /* If this is apparently the last refblock (for now), try to squeeze the
>> + * reftable in */
>> + if (rb_index == (*nb_clusters - 1) >> (s->cluster_bits - 1)
> ^ I don't understand the - 1 here.
>> &&
Oh, sorry, I missed your remark here. The 1 is log2(sizeof(uint16_t)),
that is, log2(size of a refcount block entry). I'll try to make it more
obvious in v4 (and I'll add comments addressing for alloc_clusters_imrt()).
Max
>> + rt_ofs < 0)
>> + {
>> + rt_ofs = alloc_clusters_imrt(bs, size_to_clusters(s, reftable_size *
>> + sizeof(uint64_t)),
>> + refcount_table, nb_clusters,
>> + &first_free_cluster);
>> + if (rt_ofs < 0) {
>> + fprintf(stderr, "ERROR allocating reftable: %s\n",
>> + strerror(-ret));
>> + res->check_errors++;
>> + ret = rt_ofs;
>> + goto fail;
>> + }
>> + }
>> +
>> + ret = qcow2_pre_write_overlap_check(bs, 0, rb_ofs, s->cluster_size);
>> + if (ret < 0) {
>> + fprintf(stderr, "ERROR writing refblock: %s\n", strerror(-ret));
>> + goto fail;
>> + }
>> +
>> + on_disk_rb = g_malloc0(s->cluster_size);
>> + for (i = 0; i < s->cluster_size / sizeof(uint16_t) &&
>> + rb_start + i < *nb_clusters; i++)
>> + {
>> + on_disk_rb[i] = cpu_to_be16((*refcount_table)[rb_start + i]);
>> + }
>> +
>> + ret = bdrv_write(bs->file, rb_ofs / BDRV_SECTOR_SIZE,
>> + (void *)on_disk_rb, s->cluster_sectors);
>> + g_free(on_disk_rb);
>> + if (ret < 0) {
>> + fprintf(stderr, "ERROR writing refblock: %s\n", strerror(-ret));
>> + goto fail;
>> + }
>> +
>> + /* Go to the end of this refblock */
>> + cluster = rb_start + s->cluster_size / sizeof(uint16_t) - 1;
>> + }
>> +
>> + if (rt_ofs < 0) {
>> + int64_t post_rb_start = ROUND_UP(*nb_clusters,
>> + s->cluster_size / sizeof(uint16_t));
>> +
>> + /* Not pretty but simple */
>> + if (first_free_cluster < post_rb_start) {
>> + first_free_cluster = post_rb_start;
>> + }
>> + rt_ofs = alloc_clusters_imrt(bs, size_to_clusters(s, reftable_size *
>> + sizeof(uint64_t)),
>> + refcount_table, nb_clusters,
>> + &first_free_cluster);
>> + if (rt_ofs < 0) {
>> + fprintf(stderr, "ERROR allocating reftable: %s\n", strerror(-ret));
>> + res->check_errors++;
>> + ret = rt_ofs;
>> + goto fail;
>> + }
>> +
>> + goto write_refblocks;
>> + }
>> +
>> + assert(reftable);
>> +
>> + for (rb_index = 0; rb_index < reftable_size; rb_index++) {
>> + cpu_to_be64s(&reftable[rb_index]);
>> + }
>> +
>> + ret = qcow2_pre_write_overlap_check(bs, 0, rt_ofs,
>> + reftable_size * sizeof(uint64_t));
>> + if (ret < 0) {
>> + fprintf(stderr, "ERROR writing reftable: %s\n", strerror(-ret));
>> + goto fail;
>> + }
>> +
>> + ret = bdrv_write(bs->file, rt_ofs / BDRV_SECTOR_SIZE, (void *)reftable,
>> + reftable_size * sizeof(uint64_t) / BDRV_SECTOR_SIZE);
>> + if (ret < 0) {
>> + fprintf(stderr, "ERROR writing reftable: %s\n", strerror(-ret));
>> + goto fail;
>> + }
>> +
>> + /* Enter new reftable into the image header */
>> + cpu_to_be64w((uint64_t *)&rt_offset_and_clusters[0], rt_ofs);
>> + cpu_to_be32w((uint32_t *)&rt_offset_and_clusters[sizeof(uint64_t)],
>> + size_to_clusters(s, reftable_size * sizeof(uint64_t)));
>> + ret = bdrv_pwrite_sync(bs->file, offsetof(QCowHeader,
>> + refcount_table_offset),
>> + rt_offset_and_clusters,
>> + sizeof(rt_offset_and_clusters));
>> + if (ret < 0) {
>> + fprintf(stderr, "ERROR setting reftable: %s\n", strerror(-ret));
>> + goto fail;
>> + }
>> +
>> + for (rb_index = 0; rb_index < reftable_size; rb_index++) {
>> + be64_to_cpus(&reftable[rb_index]);
>> + }
>> + s->refcount_table = reftable;
>> + s->refcount_table_offset = rt_ofs;
>> + s->refcount_table_size = reftable_size;
>> +
>> + return 0;
>> +
>> +fail:
>> + g_free(reftable);
>> + return ret;
>> +}
>> +
>> +/*
>> * Checks an image for refcount consistency.
>> *
>> * Returns 0 if no errors are found, the number of errors in case the image is
>> @@ -1600,6 +1839,7 @@ int qcow2_check_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
>> BdrvCheckMode fix)
>> {
>> BDRVQcowState *s = bs->opaque;
>> + BdrvCheckResult pre_compare_res;
>> int64_t size, highest_cluster, nb_clusters;
>> uint16_t *refcount_table = NULL;
>> bool rebuild = false;
>> @@ -1626,11 +1866,30 @@ int qcow2_check_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
>> goto fail;
>> }
>>
>> - compare_refcounts(bs, res, fix, &rebuild, &highest_cluster, refcount_table,
>> + /* In case we don't need to rebuild the refcount structure (but want to fix
>> + * something), this function is immediately called again, in which case the
>> + * result should be ignored */
>> + pre_compare_res = *res;
>> + compare_refcounts(bs, res, 0, &rebuild, &highest_cluster, refcount_table,
>> nb_clusters);
>>
>> - if (rebuild) {
>> - fprintf(stderr, "ERROR need to rebuild refcount structures\n");
>> + if (rebuild && (fix & BDRV_FIX_ERRORS)) {
>> + fprintf(stderr, "Rebuilding refcount structure\n");
>> + ret = rebuild_refcount_structure(bs, res, &refcount_table,
>> + &nb_clusters);
>> + if (ret < 0) {
>> + goto fail;
>> + }
>> + } else if (fix) {
>> + if (rebuild) {
>> + fprintf(stderr, "ERROR need to rebuild refcount structures\n");
>> + }
>> +
>> + if (res->leaks || res->corruptions) {
>> + *res = pre_compare_res;
>> + compare_refcounts(bs, res, fix, &rebuild, &highest_cluster,
>> + refcount_table, nb_clusters);
>> + }
>> }
>>
>> /* check OFLAG_COPIED */
>> --
>> 2.0.4
>>
^ permalink raw reply [flat|nested] 28+ messages in thread
end of thread, other threads:[~2014-08-27 18:37 UTC | newest]
Thread overview: 28+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2014-08-22 16:31 [Qemu-devel] [PATCH v3 00/10] qcow2: Fix image repairing Max Reitz
2014-08-22 16:31 ` [Qemu-devel] [PATCH v3 01/10] qcow2: Fix leaks in dirty images Max Reitz
2014-08-22 18:44 ` Benoît Canet
2014-08-22 16:31 ` [Qemu-devel] [PATCH v3 02/10] qcow2: Split qcow2_check_refcounts() Max Reitz
2014-08-22 18:03 ` Benoît Canet
2014-08-22 16:31 ` [Qemu-devel] [PATCH v3 03/10] qcow2: Pull check_refblocks() up Max Reitz
2014-08-22 18:04 ` Benoît Canet
2014-08-22 16:31 ` [Qemu-devel] [PATCH v3 04/10] qcow2: Reuse refcount table in calculate_refcounts() Max Reitz
2014-08-22 18:07 ` Benoît Canet
2014-08-22 16:31 ` [Qemu-devel] [PATCH v3 05/10] qcow2: Fix refcount blocks beyond image end Max Reitz
2014-08-22 18:20 ` Benoît Canet
2014-08-26 13:07 ` Eric Blake
2014-08-26 18:06 ` Max Reitz
2014-08-22 16:31 ` [Qemu-devel] [PATCH v3 06/10] qcow2: Do not perform potentially damaging repairs Max Reitz
2014-08-22 18:25 ` Benoît Canet
2014-08-22 16:31 ` [Qemu-devel] [PATCH v3 07/10] qcow2: Rebuild refcount structure during check Max Reitz
2014-08-25 17:40 ` Benoît Canet
2014-08-27 18:37 ` Max Reitz
2014-08-22 16:31 ` [Qemu-devel] [PATCH v3 08/10] qcow2: Clean up after refcount rebuild Max Reitz
2014-08-22 18:43 ` Benoît Canet
2014-08-22 16:31 ` [Qemu-devel] [PATCH v3 09/10] iotests: Fix test outputs Max Reitz
2014-08-26 13:09 ` Eric Blake
2014-08-26 18:03 ` Max Reitz
2014-08-22 16:31 ` [Qemu-devel] [PATCH v3 10/10] iotests: Add test for potentially damaging repairs Max Reitz
2014-08-22 18:50 ` Benoît Canet
2014-08-22 19:55 ` Eric Blake
2014-08-22 20:55 ` Benoît Canet
2014-08-26 17:53 ` Max Reitz
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.