[Qemu-devel] [PATCH v8 00/10] qcow2: Allow refcount_bits amendment

[Qemu-devel] [PATCH v8 00/10] qcow2: Allow refcount_bits amendment

[Max Reitz]

This series adds support to qemu for changing the refcount_bits option
of an existing qcow2 file through the qemu-img amend command.

Originally (up until v7), this series was called
"qcow2: Support refcount orders != 4", but by now that support has
already been merged so all that is left is support for qemu-img amend.


v8:
- Rebase onto master (one contextual conflict in an iotest reference
  output, but other than that this only means that the first 15 patches
  are dropped)


git-backport-diff against v7:

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:[----] [--] 'progress: Allow regressing progress'
002/10:[----] [--] 'block: Add opaque value to the amend CB'
003/10:[----] [-C] 'qcow2: Use error_report() in qcow2_amend_options()'
004/10:[----] [--] 'qcow2: Use abort() instead of assert(false)'
005/10:[----] [--] 'qcow2: Split upgrade/downgrade paths for amend'
006/10:[----] [--] 'qcow2: Use intermediate helper CB for amend'
007/10:[----] [--] 'qcow2: Add function for refcount order amendment'
008/10:[----] [--] 'qcow2: Invoke refcount order amendment function'
009/10:[----] [--] 'qcow2: Point to amend function in check'
010/10:[----] [--] 'iotests: Extend test 112 for qemu-img amend'


Max Reitz (10):
  progress: Allow regressing progress
  block: Add opaque value to the amend CB
  qcow2: Use error_report() in qcow2_amend_options()
  qcow2: Use abort() instead of assert(false)
  qcow2: Split upgrade/downgrade paths for amend
  qcow2: Use intermediate helper CB for amend
  qcow2: Add function for refcount order amendment
  qcow2: Invoke refcount order amendment function
  qcow2: Point to amend function in check
  iotests: Extend test 112 for qemu-img amend

 block.c                    |   4 +-
 block/qcow2-cluster.c      |  14 +-
 block/qcow2-refcount.c     | 455 +++++++++++++++++++++++++++++++++++++++++++++
 block/qcow2.c              | 178 ++++++++++++++----
 block/qcow2.h              |   7 +-
 include/block/block.h      |   4 +-
 include/block/block_int.h  |   3 +-
 qemu-img.c                 |   5 +-
 tests/qemu-iotests/061.out |  14 +-
 tests/qemu-iotests/112     | 109 +++++++++++
 tests/qemu-iotests/112.out |  71 +++++++
 util/qemu-progress.c       |   3 +-
 12 files changed, 808 insertions(+), 59 deletions(-)

-- 
2.1.0

[Qemu-devel] [PATCH v8 01/10] progress: Allow regressing progress

[Max Reitz]

Progress may regress; this should be displayed correctly by
qemu_progress_print().

While touching that area of code, drop the redundant parentheses in the
same condition.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
---
 util/qemu-progress.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/util/qemu-progress.c b/util/qemu-progress.c
index 4ee5cd0..532333e 100644
--- a/util/qemu-progress.c
+++ b/util/qemu-progress.c
@@ -152,7 +152,8 @@ void qemu_progress_print(float delta, int max)
     state.current = current;
 
     if (current > (state.last_print + state.min_skip) ||
-        (current == 100) || (current == 0)) {
+        current < (state.last_print - state.min_skip) ||
+        current == 100 || current == 0) {
         state.last_print = state.current;
         state.print();
     }
-- 
2.1.0

[Qemu-devel] [PATCH v8 02/10] block: Add opaque value to the amend CB

[Max Reitz]

Add an opaque value which is to be passed to the bdrv_amend_options()
status callback.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
---
 block.c                   |  4 ++--
 block/qcow2-cluster.c     | 14 ++++++++------
 block/qcow2.c             |  9 +++++----
 block/qcow2.h             |  3 ++-
 include/block/block.h     |  4 ++--
 include/block/block_int.h |  3 ++-
 qemu-img.c                |  5 +++--
 7 files changed, 24 insertions(+), 18 deletions(-)

diff --git a/block.c b/block.c
index 0fe97de..46693d1 100644
--- a/block.c
+++ b/block.c
@@ -5927,12 +5927,12 @@ void bdrv_add_before_write_notifier(BlockDriverState *bs,
 }
 
 int bdrv_amend_options(BlockDriverState *bs, QemuOpts *opts,
-                       BlockDriverAmendStatusCB *status_cb)
+                       BlockDriverAmendStatusCB *status_cb, void *cb_opaque)
 {
     if (!bs->drv->bdrv_amend_options) {
         return -ENOTSUP;
     }
-    return bs->drv->bdrv_amend_options(bs, opts, status_cb);
+    return bs->drv->bdrv_amend_options(bs, opts, status_cb, cb_opaque);
 }
 
 /* This function will be called by the bdrv_recurse_is_first_non_filter method
diff --git a/block/qcow2-cluster.c b/block/qcow2-cluster.c
index ed2b44d..56363dd 100644
--- a/block/qcow2-cluster.c
+++ b/block/qcow2-cluster.c
@@ -1620,7 +1620,8 @@ fail:
 static int expand_zero_clusters_in_l1(BlockDriverState *bs, uint64_t *l1_table,
                                       int l1_size, int64_t *visited_l1_entries,
                                       int64_t l1_entries,
-                                      BlockDriverAmendStatusCB *status_cb)
+                                      BlockDriverAmendStatusCB *status_cb,
+                                      void *cb_opaque)
 {
     BDRVQcowState *s = bs->opaque;
     bool is_active_l1 = (l1_table == s->l1_table);
@@ -1646,7 +1647,7 @@ static int expand_zero_clusters_in_l1(BlockDriverState *bs, uint64_t *l1_table,
             /* unallocated */
             (*visited_l1_entries)++;
             if (status_cb) {
-                status_cb(bs, *visited_l1_entries, l1_entries);
+                status_cb(bs, *visited_l1_entries, l1_entries, cb_opaque);
             }
             continue;
         }
@@ -1787,7 +1788,7 @@ static int expand_zero_clusters_in_l1(BlockDriverState *bs, uint64_t *l1_table,
 
         (*visited_l1_entries)++;
         if (status_cb) {
-            status_cb(bs, *visited_l1_entries, l1_entries);
+            status_cb(bs, *visited_l1_entries, l1_entries, cb_opaque);
         }
     }
 
@@ -1816,7 +1817,8 @@ fail:
  * qcow2 version which doesn't yet support metadata zero clusters.
  */
 int qcow2_expand_zero_clusters(BlockDriverState *bs,
-                               BlockDriverAmendStatusCB *status_cb)
+                               BlockDriverAmendStatusCB *status_cb,
+                               void *cb_opaque)
 {
     BDRVQcowState *s = bs->opaque;
     uint64_t *l1_table = NULL;
@@ -1833,7 +1835,7 @@ int qcow2_expand_zero_clusters(BlockDriverState *bs,
 
     ret = expand_zero_clusters_in_l1(bs, s->l1_table, s->l1_size,
                                      &visited_l1_entries, l1_entries,
-                                     status_cb);
+                                     status_cb, cb_opaque);
     if (ret < 0) {
         goto fail;
     }
@@ -1868,7 +1870,7 @@ int qcow2_expand_zero_clusters(BlockDriverState *bs,
 
         ret = expand_zero_clusters_in_l1(bs, l1_table, s->snapshots[i].l1_size,
                                          &visited_l1_entries, l1_entries,
-                                         status_cb);
+                                         status_cb, cb_opaque);
         if (ret < 0) {
             goto fail;
         }
diff --git a/block/qcow2.c b/block/qcow2.c
index 32bdf75..fd66bf2 100644
--- a/block/qcow2.c
+++ b/block/qcow2.c
@@ -2597,7 +2597,7 @@ static int qcow2_load_vmstate(BlockDriverState *bs, uint8_t *buf,
  * have to be removed.
  */
 static int qcow2_downgrade(BlockDriverState *bs, int target_version,
-                           BlockDriverAmendStatusCB *status_cb)
+                           BlockDriverAmendStatusCB *status_cb, void *cb_opaque)
 {
     BDRVQcowState *s = bs->opaque;
     int current_version = s->qcow_version;
@@ -2646,7 +2646,7 @@ static int qcow2_downgrade(BlockDriverState *bs, int target_version,
     /* clearing autoclear features is trivial */
     s->autoclear_features = 0;
 
-    ret = qcow2_expand_zero_clusters(bs, status_cb);
+    ret = qcow2_expand_zero_clusters(bs, status_cb, cb_opaque);
     if (ret < 0) {
         return ret;
     }
@@ -2661,7 +2661,8 @@ static int qcow2_downgrade(BlockDriverState *bs, int target_version,
 }
 
 static int qcow2_amend_options(BlockDriverState *bs, QemuOpts *opts,
-                               BlockDriverAmendStatusCB *status_cb)
+                               BlockDriverAmendStatusCB *status_cb,
+                               void *cb_opaque)
 {
     BDRVQcowState *s = bs->opaque;
     int old_version = s->qcow_version, new_version = old_version;
@@ -2743,7 +2744,7 @@ static int qcow2_amend_options(BlockDriverState *bs, QemuOpts *opts,
                 return ret;
             }
         } else {
-            ret = qcow2_downgrade(bs, new_version, status_cb);
+            ret = qcow2_downgrade(bs, new_version, status_cb, cb_opaque);
             if (ret < 0) {
                 return ret;
             }
diff --git a/block/qcow2.h b/block/qcow2.h
index aa6d367..eed26db 100644
--- a/block/qcow2.h
+++ b/block/qcow2.h
@@ -546,7 +546,8 @@ int qcow2_discard_clusters(BlockDriverState *bs, uint64_t offset,
 int qcow2_zero_clusters(BlockDriverState *bs, uint64_t offset, int nb_sectors);
 
 int qcow2_expand_zero_clusters(BlockDriverState *bs,
-                               BlockDriverAmendStatusCB *status_cb);
+                               BlockDriverAmendStatusCB *status_cb,
+                               void *cb_opaque);
 
 /* qcow2-snapshot.c functions */
 int qcow2_snapshot_create(BlockDriverState *bs, QEMUSnapshotInfo *sn_info);
diff --git a/include/block/block.h b/include/block/block.h
index 4c57d63..aaa31a0 100644
--- a/include/block/block.h
+++ b/include/block/block.h
@@ -290,9 +290,9 @@ int bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, BdrvCheckMode fix);
  * block driver; total_work_size may change during the course of the amendment
  * operation */
 typedef void BlockDriverAmendStatusCB(BlockDriverState *bs, int64_t offset,
-                                      int64_t total_work_size);
+                                      int64_t total_work_size, void *opaque);
 int bdrv_amend_options(BlockDriverState *bs_new, QemuOpts *opts,
-                       BlockDriverAmendStatusCB *status_cb);
+                       BlockDriverAmendStatusCB *status_cb, void *cb_opaque);
 
 /* external snapshots */
 bool bdrv_recurse_is_first_non_filter(BlockDriverState *bs,
diff --git a/include/block/block_int.h b/include/block/block_int.h
index dccb092..615e880 100644
--- a/include/block/block_int.h
+++ b/include/block/block_int.h
@@ -237,7 +237,8 @@ struct BlockDriver {
         BdrvCheckMode fix);
 
     int (*bdrv_amend_options)(BlockDriverState *bs, QemuOpts *opts,
-                              BlockDriverAmendStatusCB *status_cb);
+                              BlockDriverAmendStatusCB *status_cb,
+                              void *cb_opaque);
 
     void (*bdrv_debug_event)(BlockDriverState *bs, BlkDebugEvent event);
 
diff --git a/qemu-img.c b/qemu-img.c
index 9dddfbe..e638f2b 100644
--- a/qemu-img.c
+++ b/qemu-img.c
@@ -2882,7 +2882,8 @@ out:
 }
 
 static void amend_status_cb(BlockDriverState *bs,
-                            int64_t offset, int64_t total_work_size)
+                            int64_t offset, int64_t total_work_size,
+                            void *opaque)
 {
     qemu_progress_print(100.f * offset / total_work_size, 0);
 }
@@ -3006,7 +3007,7 @@ static int img_amend(int argc, char **argv)
 
     /* In case the driver does not call amend_status_cb() */
     qemu_progress_print(0.f, 0);
-    ret = bdrv_amend_options(bs, opts, &amend_status_cb);
+    ret = bdrv_amend_options(bs, opts, &amend_status_cb, NULL);
     qemu_progress_print(100.f, 0);
     if (ret < 0) {
         error_report("Error while amending options: %s", strerror(-ret));
-- 
2.1.0

[Qemu-devel] [PATCH v8 03/10] qcow2: Use error_report() in qcow2_amend_options()

[Max Reitz]

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
---
 block/qcow2.c              | 14 ++++++--------
 tests/qemu-iotests/061.out | 14 +++++++-------
 2 files changed, 13 insertions(+), 15 deletions(-)

diff --git a/block/qcow2.c b/block/qcow2.c
index fd66bf2..d1a0e53 100644
--- a/block/qcow2.c
+++ b/block/qcow2.c
@@ -2691,11 +2691,11 @@ static int qcow2_amend_options(BlockDriverState *bs, QemuOpts *opts,
             } else if (!strcmp(compat, "1.1")) {
                 new_version = 3;
             } else {
-                fprintf(stderr, "Unknown compatibility level %s.\n", compat);
+                error_report("Unknown compatibility level %s", compat);
                 return -EINVAL;
             }
         } else if (!strcmp(desc->name, BLOCK_OPT_PREALLOC)) {
-            fprintf(stderr, "Cannot change preallocation mode.\n");
+            error_report("Cannot change preallocation mode");
             return -ENOTSUP;
         } else if (!strcmp(desc->name, BLOCK_OPT_SIZE)) {
             new_size = qemu_opt_get_size(opts, BLOCK_OPT_SIZE, 0);
@@ -2707,16 +2707,14 @@ static int qcow2_amend_options(BlockDriverState *bs, QemuOpts *opts,
             encrypt = qemu_opt_get_bool(opts, BLOCK_OPT_ENCRYPT,
                                         s->crypt_method);
             if (encrypt != !!s->crypt_method) {
-                fprintf(stderr, "Changing the encryption flag is not "
-                        "supported.\n");
+                error_report("Changing the encryption flag is not supported");
                 return -ENOTSUP;
             }
         } else if (!strcmp(desc->name, BLOCK_OPT_CLUSTER_SIZE)) {
             cluster_size = qemu_opt_get_size(opts, BLOCK_OPT_CLUSTER_SIZE,
                                              cluster_size);
             if (cluster_size != s->cluster_size) {
-                fprintf(stderr, "Changing the cluster size is not "
-                        "supported.\n");
+                error_report("Changing the cluster size is not supported");
                 return -ENOTSUP;
             }
         } else if (!strcmp(desc->name, BLOCK_OPT_LAZY_REFCOUNTS)) {
@@ -2762,8 +2760,8 @@ static int qcow2_amend_options(BlockDriverState *bs, QemuOpts *opts,
     if (s->use_lazy_refcounts != lazy_refcounts) {
         if (lazy_refcounts) {
             if (s->qcow_version < 3) {
-                fprintf(stderr, "Lazy refcounts only supported with compatibility "
-                        "level 1.1 and above (use compat=1.1 or greater)\n");
+                error_report("Lazy refcounts only supported with compatibility "
+                             "level 1.1 and above (use compat=1.1 or greater)");
                 return -EINVAL;
             }
             s->compatible_features |= QCOW2_COMPAT_LAZY_REFCOUNTS;
diff --git a/tests/qemu-iotests/061.out b/tests/qemu-iotests/061.out
index 5ec248f..713ca6b 100644
--- a/tests/qemu-iotests/061.out
+++ b/tests/qemu-iotests/061.out
@@ -281,18 +281,18 @@ No errors were found on the image.
 === Testing invalid configurations ===
 
 Formatting 'TEST_DIR/t.IMGFMT', fmt=IMGFMT size=67108864
-Lazy refcounts only supported with compatibility level 1.1 and above (use compat=1.1 or greater)
+qemu-img: Lazy refcounts only supported with compatibility level 1.1 and above (use compat=1.1 or greater)
 qemu-img: Error while amending options: Invalid argument
-Lazy refcounts only supported with compatibility level 1.1 and above (use compat=1.1 or greater)
+qemu-img: Lazy refcounts only supported with compatibility level 1.1 and above (use compat=1.1 or greater)
 qemu-img: Error while amending options: Invalid argument
-Unknown compatibility level 0.42.
+qemu-img: Unknown compatibility level 0.42
 qemu-img: Error while amending options: Invalid argument
 qemu-img: Invalid parameter 'foo'
-Changing the cluster size is not supported.
+qemu-img: Changing the cluster size is not supported
 qemu-img: Error while amending options: Operation not supported
-Changing the encryption flag is not supported.
+qemu-img: Changing the encryption flag is not supported
 qemu-img: Error while amending options: Operation not supported
-Cannot change preallocation mode.
+qemu-img: Cannot change preallocation mode
 qemu-img: Error while amending options: Operation not supported
 
 === Testing correct handling of unset value ===
@@ -300,7 +300,7 @@ qemu-img: Error while amending options: Operation not supported
 Formatting 'TEST_DIR/t.IMGFMT', fmt=IMGFMT size=67108864
 Should work:
 Should not work:
-Changing the cluster size is not supported.
+qemu-img: Changing the cluster size is not supported
 qemu-img: Error while amending options: Operation not supported
 
 === Testing zero expansion on inactive clusters ===
-- 
2.1.0

[Qemu-devel] [PATCH v8 04/10] qcow2: Use abort() instead of assert(false)

[Max Reitz]

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
---
 block/qcow2.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/block/qcow2.c b/block/qcow2.c
index d1a0e53..da7795c 100644
--- a/block/qcow2.c
+++ b/block/qcow2.c
@@ -2724,9 +2724,9 @@ static int qcow2_amend_options(BlockDriverState *bs, QemuOpts *opts,
             error_report("Cannot change refcount entry width");
             return -ENOTSUP;
         } else {
-            /* if this assertion fails, this probably means a new option was
+            /* if this point is reached, this probably means a new option was
              * added without having it covered here */
-            assert(false);
+            abort();
         }
 
         desc++;
-- 
2.1.0

[Qemu-devel] [PATCH v8 05/10] qcow2: Split upgrade/downgrade paths for amend

[Max Reitz]

If the image version should be upgraded, that is the first we should do;
if it should be downgraded, that is the last we should do. So split the
version change block into an upgrade part at the start and a downgrade
part at the end.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
---
 block/qcow2.c | 31 ++++++++++++++++---------------
 1 file changed, 16 insertions(+), 15 deletions(-)

diff --git a/block/qcow2.c b/block/qcow2.c
index da7795c..35b91cd 100644
--- a/block/qcow2.c
+++ b/block/qcow2.c
@@ -2732,20 +2732,13 @@ static int qcow2_amend_options(BlockDriverState *bs, QemuOpts *opts,
         desc++;
     }
 
-    if (new_version != old_version) {
-        if (new_version > old_version) {
-            /* Upgrade */
-            s->qcow_version = new_version;
-            ret = qcow2_update_header(bs);
-            if (ret < 0) {
-                s->qcow_version = old_version;
-                return ret;
-            }
-        } else {
-            ret = qcow2_downgrade(bs, new_version, status_cb, cb_opaque);
-            if (ret < 0) {
-                return ret;
-            }
+    /* Upgrade first (some features may require compat=1.1) */
+    if (new_version > old_version) {
+        s->qcow_version = new_version;
+        ret = qcow2_update_header(bs);
+        if (ret < 0) {
+            s->qcow_version = old_version;
+            return ret;
         }
     }
 
@@ -2759,7 +2752,7 @@ static int qcow2_amend_options(BlockDriverState *bs, QemuOpts *opts,
 
     if (s->use_lazy_refcounts != lazy_refcounts) {
         if (lazy_refcounts) {
-            if (s->qcow_version < 3) {
+            if (new_version < 3) {
                 error_report("Lazy refcounts only supported with compatibility "
                              "level 1.1 and above (use compat=1.1 or greater)");
                 return -EINVAL;
@@ -2795,6 +2788,14 @@ static int qcow2_amend_options(BlockDriverState *bs, QemuOpts *opts,
         }
     }
 
+    /* Downgrade last (so unsupported features can be removed before) */
+    if (new_version < old_version) {
+        ret = qcow2_downgrade(bs, new_version, status_cb, cb_opaque);
+        if (ret < 0) {
+            return ret;
+        }
+    }
+
     return 0;
 }
 
-- 
2.1.0

[Qemu-devel] [PATCH v8 06/10] qcow2: Use intermediate helper CB for amend

[Max Reitz]

If there is more than one time-consuming operation to be performed for
qcow2_amend_options(), we need an intermediate CB which coordinates the
progress of the individual operations and passes the result to the
original status callback.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
---
 block/qcow2.c | 80 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 79 insertions(+), 1 deletion(-)

diff --git a/block/qcow2.c b/block/qcow2.c
index 35b91cd..f311159 100644
--- a/block/qcow2.c
+++ b/block/qcow2.c
@@ -2660,6 +2660,75 @@ static int qcow2_downgrade(BlockDriverState *bs, int target_version,
     return 0;
 }
 
+typedef enum Qcow2AmendOperation {
+    /* This is the value Qcow2AmendHelperCBInfo::last_operation will be
+     * statically initialized to so that the helper CB can discern the first
+     * invocation from an operation change */
+    QCOW2_NO_OPERATION = 0,
+
+    QCOW2_DOWNGRADING,
+} Qcow2AmendOperation;
+
+typedef struct Qcow2AmendHelperCBInfo {
+    /* The code coordinating the amend operations should only modify
+     * these four fields; the rest will be managed by the CB */
+    BlockDriverAmendStatusCB *original_status_cb;
+    void *original_cb_opaque;
+
+    Qcow2AmendOperation current_operation;
+
+    /* Total number of operations to perform (only set once) */
+    int total_operations;
+
+    /* The following fields are managed by the CB */
+
+    /* Number of operations completed */
+    int operations_completed;
+
+    /* Cumulative offset of all completed operations */
+    int64_t offset_completed;
+
+    Qcow2AmendOperation last_operation;
+    int64_t last_work_size;
+} Qcow2AmendHelperCBInfo;
+
+static void qcow2_amend_helper_cb(BlockDriverState *bs,
+                                  int64_t operation_offset,
+                                  int64_t operation_work_size, void *opaque)
+{
+    Qcow2AmendHelperCBInfo *info = opaque;
+    int64_t current_work_size;
+    int64_t projected_work_size;
+
+    if (info->current_operation != info->last_operation) {
+        if (info->last_operation != QCOW2_NO_OPERATION) {
+            info->offset_completed += info->last_work_size;
+            info->operations_completed++;
+        }
+
+        info->last_operation = info->current_operation;
+    }
+
+    assert(info->total_operations > 0);
+    assert(info->operations_completed < info->total_operations);
+
+    info->last_work_size = operation_work_size;
+
+    current_work_size = info->offset_completed + operation_work_size;
+
+    /* current_work_size is the total work size for (operations_completed + 1)
+     * operations (which includes this one), so multiply it by the number of
+     * operations not covered and divide it by the number of operations
+     * covered to get a projection for the operations not covered */
+    projected_work_size = current_work_size * (info->total_operations -
+                                               info->operations_completed - 1)
+                                            / (info->operations_completed + 1);
+
+    info->original_status_cb(bs, info->offset_completed + operation_offset,
+                             current_work_size + projected_work_size,
+                             info->original_cb_opaque);
+}
+
 static int qcow2_amend_options(BlockDriverState *bs, QemuOpts *opts,
                                BlockDriverAmendStatusCB *status_cb,
                                void *cb_opaque)
@@ -2674,6 +2743,7 @@ static int qcow2_amend_options(BlockDriverState *bs, QemuOpts *opts,
     bool encrypt;
     int ret;
     QemuOptDesc *desc = opts->list->desc;
+    Qcow2AmendHelperCBInfo helper_cb_info;
 
     while (desc && desc->name) {
         if (!qemu_opt_find(opts, desc->name)) {
@@ -2732,6 +2802,12 @@ static int qcow2_amend_options(BlockDriverState *bs, QemuOpts *opts,
         desc++;
     }
 
+    helper_cb_info = (Qcow2AmendHelperCBInfo){
+        .original_status_cb = status_cb,
+        .original_cb_opaque = cb_opaque,
+        .total_operations = (new_version < old_version)
+    };
+
     /* Upgrade first (some features may require compat=1.1) */
     if (new_version > old_version) {
         s->qcow_version = new_version;
@@ -2790,7 +2866,9 @@ static int qcow2_amend_options(BlockDriverState *bs, QemuOpts *opts,
 
     /* Downgrade last (so unsupported features can be removed before) */
     if (new_version < old_version) {
-        ret = qcow2_downgrade(bs, new_version, status_cb, cb_opaque);
+        helper_cb_info.current_operation = QCOW2_DOWNGRADING;
+        ret = qcow2_downgrade(bs, new_version, &qcow2_amend_helper_cb,
+                              &helper_cb_info);
         if (ret < 0) {
             return ret;
         }
-- 
2.1.0

[Qemu-devel] [PATCH v8 07/10] qcow2: Add function for refcount order amendment

[Max Reitz]

Add a function qcow2_change_refcount_order() which allows changing the
refcount order of a qcow2 image.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
---
 block/qcow2-refcount.c | 452 +++++++++++++++++++++++++++++++++++++++++++++++++
 block/qcow2.h          |   4 +
 2 files changed, 456 insertions(+)

diff --git a/block/qcow2-refcount.c b/block/qcow2-refcount.c
index 6cbae1d..f7e5b5e 100644
--- a/block/qcow2-refcount.c
+++ b/block/qcow2-refcount.c
@@ -2461,3 +2461,455 @@ int qcow2_pre_write_overlap_check(BlockDriverState *bs, int ign, int64_t offset,
 
     return 0;
 }
+
+/* A pointer to a function of this type is given to walk_over_reftable(). That
+ * function will create refblocks and pass them to a RefblockFinishOp once they
+ * are completed (@refblock). @refblock_empty is set if the refblock is
+ * completely empty.
+ *
+ * Along with the refblock, a corresponding reftable entry is passed, in the
+ * reftable @reftable (which may be reallocated) at @reftable_index.
+ *
+ * @allocated should be set to true if a new cluster has been allocated.
+ */
+typedef int (RefblockFinishOp)(BlockDriverState *bs, uint64_t **reftable,
+                               uint64_t reftable_index, uint64_t *reftable_size,
+                               void *refblock, bool refblock_empty,
+                               bool *allocated, Error **errp);
+
+/**
+ * This "operation" for walk_over_reftable() allocates the refblock on disk (if
+ * it is not empty) and inserts its offset into the new reftable. The size of
+ * this new reftable is increased as required.
+ */
+static int alloc_refblock(BlockDriverState *bs, uint64_t **reftable,
+                          uint64_t reftable_index, uint64_t *reftable_size,
+                          void *refblock, bool refblock_empty, bool *allocated,
+                          Error **errp)
+{
+    BDRVQcowState *s = bs->opaque;
+    int64_t offset;
+
+    if (!refblock_empty && reftable_index >= *reftable_size) {
+        uint64_t *new_reftable;
+        uint64_t new_reftable_size;
+
+        new_reftable_size = ROUND_UP(reftable_index + 1,
+                                     s->cluster_size / sizeof(uint64_t));
+        if (new_reftable_size > QCOW_MAX_REFTABLE_SIZE / sizeof(uint64_t)) {
+            error_setg(errp,
+                       "This operation would make the refcount table grow "
+                       "beyond the maximum size supported by QEMU, aborting");
+            return -ENOTSUP;
+        }
+
+        new_reftable = g_try_realloc(*reftable, new_reftable_size *
+                                                sizeof(uint64_t));
+        if (!new_reftable) {
+            error_setg(errp, "Failed to increase reftable buffer size");
+            return -ENOMEM;
+        }
+
+        memset(new_reftable + *reftable_size, 0,
+               (new_reftable_size - *reftable_size) * sizeof(uint64_t));
+
+        *reftable      = new_reftable;
+        *reftable_size = new_reftable_size;
+    }
+
+    if (!refblock_empty && !(*reftable)[reftable_index]) {
+        offset = qcow2_alloc_clusters(bs, s->cluster_size);
+        if (offset < 0) {
+            error_setg_errno(errp, -offset, "Failed to allocate refblock");
+            return offset;
+        }
+        (*reftable)[reftable_index] = offset;
+        *allocated = true;
+    }
+
+    return 0;
+}
+
+/**
+ * This "operation" for walk_over_reftable() writes the refblock to disk at the
+ * offset specified by the new reftable's entry. It does not modify the new
+ * reftable or change any refcounts.
+ */
+static int flush_refblock(BlockDriverState *bs, uint64_t **reftable,
+                          uint64_t reftable_index, uint64_t *reftable_size,
+                          void *refblock, bool refblock_empty, bool *allocated,
+                          Error **errp)
+{
+    BDRVQcowState *s = bs->opaque;
+    int64_t offset;
+    int ret;
+
+    if (reftable_index < *reftable_size && (*reftable)[reftable_index]) {
+        offset = (*reftable)[reftable_index];
+
+        ret = qcow2_pre_write_overlap_check(bs, 0, offset, s->cluster_size);
+        if (ret < 0) {
+            error_setg_errno(errp, -ret, "Overlap check failed");
+            return ret;
+        }
+
+        ret = bdrv_pwrite(bs->file, offset, refblock, s->cluster_size);
+        if (ret < 0) {
+            error_setg_errno(errp, -ret, "Failed to write refblock");
+            return ret;
+        }
+    } else {
+        assert(refblock_empty);
+    }
+
+    return 0;
+}
+
+/**
+ * This function walks over the existing reftable and every referenced refblock;
+ * if @new_set_refcount is non-NULL, it is called for every refcount entry to
+ * create an equal new entry in the passed @new_refblock. Once that
+ * @new_refblock is completely filled, @operation will be called.
+ *
+ * @status_cb and @cb_opaque are used for the amend operation's status callback.
+ * @index is the index of the walk_over_reftable() calls and @total is the total
+ * number of walk_over_reftable() calls per amend operation. Both are used for
+ * calculating the parameters for the status callback.
+ *
+ * @allocated is set to true if a new cluster has been allocated.
+ */
+static int walk_over_reftable(BlockDriverState *bs, uint64_t **new_reftable,
+                              uint64_t *new_reftable_index,
+                              uint64_t *new_reftable_size,
+                              void *new_refblock, int new_refblock_size,
+                              int new_refcount_bits,
+                              RefblockFinishOp *operation, bool *allocated,
+                              Qcow2SetRefcountFunc *new_set_refcount,
+                              BlockDriverAmendStatusCB *status_cb,
+                              void *cb_opaque, int index, int total,
+                              Error **errp)
+{
+    BDRVQcowState *s = bs->opaque;
+    uint64_t reftable_index;
+    bool new_refblock_empty = true;
+    int refblock_index;
+    int new_refblock_index = 0;
+    int ret;
+
+    for (reftable_index = 0; reftable_index < s->refcount_table_size;
+         reftable_index++)
+    {
+        uint64_t refblock_offset = s->refcount_table[reftable_index]
+                                 & REFT_OFFSET_MASK;
+
+        status_cb(bs, (uint64_t)index * s->refcount_table_size + reftable_index,
+                  (uint64_t)total * s->refcount_table_size, cb_opaque);
+
+        if (refblock_offset) {
+            void *refblock;
+
+            if (offset_into_cluster(s, refblock_offset)) {
+                qcow2_signal_corruption(bs, true, -1, -1, "Refblock offset %#"
+                                        PRIx64 " unaligned (reftable index: %#"
+                                        PRIx64 ")", refblock_offset,
+                                        reftable_index);
+                error_setg(errp,
+                           "Image is corrupt (unaligned refblock offset)");
+                return -EIO;
+            }
+
+            ret = qcow2_cache_get(bs, s->refcount_block_cache, refblock_offset,
+                                  &refblock);
+            if (ret < 0) {
+                error_setg_errno(errp, -ret, "Failed to retrieve refblock");
+                return ret;
+            }
+
+            for (refblock_index = 0; refblock_index < s->refcount_block_size;
+                 refblock_index++)
+            {
+                uint64_t refcount;
+
+                if (new_refblock_index >= new_refblock_size) {
+                    /* new_refblock is now complete */
+                    ret = operation(bs, new_reftable, *new_reftable_index,
+                                    new_reftable_size, new_refblock,
+                                    new_refblock_empty, allocated, errp);
+                    if (ret < 0) {
+                        qcow2_cache_put(bs, s->refcount_block_cache, &refblock);
+                        return ret;
+                    }
+
+                    (*new_reftable_index)++;
+                    new_refblock_index = 0;
+                    new_refblock_empty = true;
+                }
+
+                refcount = s->get_refcount(refblock, refblock_index);
+                if (new_refcount_bits < 64 && refcount >> new_refcount_bits) {
+                    uint64_t offset;
+
+                    qcow2_cache_put(bs, s->refcount_block_cache, &refblock);
+
+                    offset = ((reftable_index << s->refcount_block_bits)
+                              + refblock_index) << s->cluster_bits;
+
+                    error_setg(errp, "Cannot decrease refcount entry width to "
+                               "%i bits: Cluster at offset %#" PRIx64 " has a "
+                               "refcount of %" PRIu64, new_refcount_bits,
+                               offset, refcount);
+                    return -EINVAL;
+                }
+
+                if (new_set_refcount) {
+                    new_set_refcount(new_refblock, new_refblock_index++,
+                                     refcount);
+                } else {
+                    new_refblock_index++;
+                }
+                new_refblock_empty = new_refblock_empty && refcount == 0;
+            }
+
+            ret = qcow2_cache_put(bs, s->refcount_block_cache, &refblock);
+            if (ret < 0) {
+                error_setg_errno(errp, -ret, "Failed to put refblock back into "
+                                 "the cache");
+                return ret;
+            }
+        } else {
+            /* No refblock means every refcount is 0 */
+            for (refblock_index = 0; refblock_index < s->refcount_block_size;
+                 refblock_index++)
+            {
+                if (new_refblock_index >= new_refblock_size) {
+                    /* new_refblock is now complete */
+                    ret = operation(bs, new_reftable, *new_reftable_index,
+                                    new_reftable_size, new_refblock,
+                                    new_refblock_empty, allocated, errp);
+                    if (ret < 0) {
+                        return ret;
+                    }
+
+                    (*new_reftable_index)++;
+                    new_refblock_index = 0;
+                    new_refblock_empty = true;
+                }
+
+                if (new_set_refcount) {
+                    new_set_refcount(new_refblock, new_refblock_index++, 0);
+                } else {
+                    new_refblock_index++;
+                }
+            }
+        }
+    }
+
+    if (new_refblock_index > 0) {
+        /* Complete the potentially existing partially filled final refblock */
+        if (new_set_refcount) {
+            for (; new_refblock_index < new_refblock_size;
+                 new_refblock_index++)
+            {
+                new_set_refcount(new_refblock, new_refblock_index, 0);
+            }
+        }
+
+        ret = operation(bs, new_reftable, *new_reftable_index,
+                        new_reftable_size, new_refblock, new_refblock_empty,
+                        allocated, errp);
+        if (ret < 0) {
+            return ret;
+        }
+
+        (*new_reftable_index)++;
+    }
+
+    status_cb(bs, (uint64_t)(index + 1) * s->refcount_table_size,
+              (uint64_t)total * s->refcount_table_size, cb_opaque);
+
+    return 0;
+}
+
+int qcow2_change_refcount_order(BlockDriverState *bs, int refcount_order,
+                                BlockDriverAmendStatusCB *status_cb,
+                                void *cb_opaque, Error **errp)
+{
+    BDRVQcowState *s = bs->opaque;
+    Qcow2GetRefcountFunc *new_get_refcount;
+    Qcow2SetRefcountFunc *new_set_refcount;
+    void *new_refblock = qemu_blockalign(bs->file, s->cluster_size);
+    uint64_t *new_reftable = NULL, new_reftable_size = 0;
+    uint64_t *old_reftable, old_reftable_size, old_reftable_offset;
+    uint64_t new_reftable_index = 0;
+    uint64_t i;
+    int64_t new_reftable_offset = 0, allocated_reftable_size = 0;
+    int new_refblock_size, new_refcount_bits = 1 << refcount_order;
+    int old_refcount_order;
+    int walk_index = 0;
+    int ret;
+    bool new_allocation;
+
+    assert(s->qcow_version >= 3);
+    assert(refcount_order >= 0 && refcount_order <= 6);
+
+    /* see qcow2_open() */
+    new_refblock_size = 1 << (s->cluster_bits - (refcount_order - 3));
+
+    new_get_refcount = get_refcount_funcs[refcount_order];
+    new_set_refcount = set_refcount_funcs[refcount_order];
+
+
+    do {
+        int total_walks;
+
+        new_allocation = false;
+
+        /* At least we have to do this walk and the one which writes the
+         * refblocks; also, at least we have to do this loop here at least
+         * twice (normally), first to do the allocations, and second to
+         * determine that everything is correctly allocated, this then makes
+         * three walks in total */
+        total_walks = MAX(walk_index + 2, 3);
+
+        /* First, allocate the structures so they are present in the refcount
+         * structures */
+        ret = walk_over_reftable(bs, &new_reftable, &new_reftable_index,
+                                 &new_reftable_size, NULL, new_refblock_size,
+                                 new_refcount_bits, &alloc_refblock,
+                                 &new_allocation, NULL, status_cb, cb_opaque,
+                                 walk_index++, total_walks, errp);
+        if (ret < 0) {
+            goto done;
+        }
+
+        new_reftable_index = 0;
+
+        if (new_allocation) {
+            if (new_reftable_offset) {
+                qcow2_free_clusters(bs, new_reftable_offset,
+                                    allocated_reftable_size * sizeof(uint64_t),
+                                    QCOW2_DISCARD_NEVER);
+            }
+
+            new_reftable_offset = qcow2_alloc_clusters(bs, new_reftable_size *
+                                                           sizeof(uint64_t));
+            if (new_reftable_offset < 0) {
+                error_setg_errno(errp, -new_reftable_offset,
+                                 "Failed to allocate the new reftable");
+                ret = new_reftable_offset;
+                goto done;
+            }
+            allocated_reftable_size = new_reftable_size;
+        }
+    } while (new_allocation);
+
+    /* Second, write the new refblocks */
+    ret = walk_over_reftable(bs, &new_reftable, &new_reftable_index,
+                             &new_reftable_size, new_refblock,
+                             new_refblock_size, new_refcount_bits,
+                             &flush_refblock, &new_allocation, new_set_refcount,
+                             status_cb, cb_opaque, walk_index, walk_index + 1,
+                             errp);
+    if (ret < 0) {
+        goto done;
+    }
+    assert(!new_allocation);
+
+
+    /* Write the new reftable */
+    ret = qcow2_pre_write_overlap_check(bs, 0, new_reftable_offset,
+                                        new_reftable_size * sizeof(uint64_t));
+    if (ret < 0) {
+        error_setg_errno(errp, -ret, "Overlap check failed");
+        goto done;
+    }
+
+    for (i = 0; i < new_reftable_size; i++) {
+        cpu_to_be64s(&new_reftable[i]);
+    }
+
+    ret = bdrv_pwrite(bs->file, new_reftable_offset, new_reftable,
+                      new_reftable_size * sizeof(uint64_t));
+
+    for (i = 0; i < new_reftable_size; i++) {
+        be64_to_cpus(&new_reftable[i]);
+    }
+
+    if (ret < 0) {
+        error_setg_errno(errp, -ret, "Failed to write the new reftable");
+        goto done;
+    }
+
+
+    /* Empty the refcount cache */
+    ret = qcow2_cache_flush(bs, s->refcount_block_cache);
+    if (ret < 0) {
+        error_setg_errno(errp, -ret, "Failed to flush the refblock cache");
+        goto done;
+    }
+
+    /* Update the image header to point to the new reftable; this only updates
+     * the fields which are relevant to qcow2_update_header(); other fields
+     * such as s->refcount_table or s->refcount_bits stay stale for now
+     * (because we have to restore everything if qcow2_update_header() fails) */
+    old_refcount_order  = s->refcount_order;
+    old_reftable_size   = s->refcount_table_size;
+    old_reftable_offset = s->refcount_table_offset;
+
+    s->refcount_order        = refcount_order;
+    s->refcount_table_size   = new_reftable_size;
+    s->refcount_table_offset = new_reftable_offset;
+
+    ret = qcow2_update_header(bs);
+    if (ret < 0) {
+        s->refcount_order        = old_refcount_order;
+        s->refcount_table_size   = old_reftable_size;
+        s->refcount_table_offset = old_reftable_offset;
+        error_setg_errno(errp, -ret, "Failed to update the qcow2 header");
+        goto done;
+    }
+
+    /* Now update the rest of the in-memory information */
+    old_reftable = s->refcount_table;
+    s->refcount_table = new_reftable;
+
+    s->refcount_bits = 1 << refcount_order;
+    s->refcount_max = UINT64_C(1) << (s->refcount_bits - 1);
+    s->refcount_max += s->refcount_max - 1;
+
+    s->refcount_block_bits = s->cluster_bits - (refcount_order - 3);
+    s->refcount_block_size = 1 << s->refcount_block_bits;
+
+    s->get_refcount = new_get_refcount;
+    s->set_refcount = new_set_refcount;
+
+    /* For cleaning up all old refblocks and the old reftable below the "done"
+     * label */
+    new_reftable        = old_reftable;
+    new_reftable_size   = old_reftable_size;
+    new_reftable_offset = old_reftable_offset;
+
+done:
+    if (new_reftable) {
+        /* On success, new_reftable actually points to the old reftable (and
+         * new_reftable_size is the old reftable's size); but that is just
+         * fine */
+        for (i = 0; i < new_reftable_size; i++) {
+            uint64_t offset = new_reftable[i] & REFT_OFFSET_MASK;
+            if (offset) {
+                qcow2_free_clusters(bs, offset, s->cluster_size,
+                                    QCOW2_DISCARD_OTHER);
+            }
+        }
+        g_free(new_reftable);
+
+        if (new_reftable_offset > 0) {
+            qcow2_free_clusters(bs, new_reftable_offset,
+                                new_reftable_size * sizeof(uint64_t),
+                                QCOW2_DISCARD_OTHER);
+        }
+    }
+
+    qemu_vfree(new_refblock);
+    return ret;
+}
diff --git a/block/qcow2.h b/block/qcow2.h
index eed26db..8f34a5c 100644
--- a/block/qcow2.h
+++ b/block/qcow2.h
@@ -521,6 +521,10 @@ int qcow2_check_metadata_overlap(BlockDriverState *bs, int ign, int64_t offset,
 int qcow2_pre_write_overlap_check(BlockDriverState *bs, int ign, int64_t offset,
                                   int64_t size);
 
+int qcow2_change_refcount_order(BlockDriverState *bs, int refcount_order,
+                                BlockDriverAmendStatusCB *status_cb,
+                                void *cb_opaque, Error **errp);
+
 /* qcow2-cluster.c functions */
 int qcow2_grow_l1_table(BlockDriverState *bs, uint64_t min_size,
                         bool exact_size);
-- 
2.1.0

[Qemu-devel] [PATCH v8 08/10] qcow2: Invoke refcount order amendment function

[Max Reitz]

Make use of qcow2_change_refcount_order() to support changing the
refcount order with qemu-img amend.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
---
 block/qcow2.c | 44 +++++++++++++++++++++++++++++++++++---------
 1 file changed, 35 insertions(+), 9 deletions(-)

diff --git a/block/qcow2.c b/block/qcow2.c
index f311159..3814361 100644
--- a/block/qcow2.c
+++ b/block/qcow2.c
@@ -2612,13 +2612,7 @@ static int qcow2_downgrade(BlockDriverState *bs, int target_version,
     }
 
     if (s->refcount_order != 4) {
-        /* we would have to convert the image to a refcount_order == 4 image
-         * here; however, since qemu (at the time of writing this) does not
-         * support anything different than 4 anyway, there is no point in doing
-         * so right now; however, we should error out (if qemu supports this in
-         * the future and this code has not been adapted) */
-        error_report("qcow2_downgrade: Image refcount orders other than 4 are "
-                     "currently not supported.");
+        error_report("compat=0.10 requires refcount_bits=16");
         return -ENOTSUP;
     }
 
@@ -2666,6 +2660,7 @@ typedef enum Qcow2AmendOperation {
      * invocation from an operation change */
     QCOW2_NO_OPERATION = 0,
 
+    QCOW2_CHANGING_REFCOUNT_ORDER,
     QCOW2_DOWNGRADING,
 } Qcow2AmendOperation;
 
@@ -2741,6 +2736,7 @@ static int qcow2_amend_options(BlockDriverState *bs, QemuOpts *opts,
     const char *compat = NULL;
     uint64_t cluster_size = s->cluster_size;
     bool encrypt;
+    int refcount_bits = s->refcount_bits;
     int ret;
     QemuOptDesc *desc = opts->list->desc;
     Qcow2AmendHelperCBInfo helper_cb_info;
@@ -2791,8 +2787,16 @@ static int qcow2_amend_options(BlockDriverState *bs, QemuOpts *opts,
             lazy_refcounts = qemu_opt_get_bool(opts, BLOCK_OPT_LAZY_REFCOUNTS,
                                                lazy_refcounts);
         } else if (!strcmp(desc->name, BLOCK_OPT_REFCOUNT_BITS)) {
-            error_report("Cannot change refcount entry width");
-            return -ENOTSUP;
+            refcount_bits = qemu_opt_get_number(opts, BLOCK_OPT_REFCOUNT_BITS,
+                                                refcount_bits);
+
+            if (refcount_bits <= 0 || refcount_bits > 64 ||
+                !is_power_of_2(refcount_bits))
+            {
+                error_report("Refcount width must be a power of two and may "
+                             "not exceed 64 bits");
+                return -EINVAL;
+            }
         } else {
             /* if this point is reached, this probably means a new option was
              * added without having it covered here */
@@ -2806,6 +2810,7 @@ static int qcow2_amend_options(BlockDriverState *bs, QemuOpts *opts,
         .original_status_cb = status_cb,
         .original_cb_opaque = cb_opaque,
         .total_operations = (new_version < old_version)
+                          + (s->refcount_bits != refcount_bits)
     };
 
     /* Upgrade first (some features may require compat=1.1) */
@@ -2818,6 +2823,27 @@ static int qcow2_amend_options(BlockDriverState *bs, QemuOpts *opts,
         }
     }
 
+    if (s->refcount_bits != refcount_bits) {
+        int refcount_order = ffs(refcount_bits) - 1;
+        Error *local_error = NULL;
+
+        if (new_version < 3 && refcount_bits != 16) {
+            error_report("Different refcount widths than 16 bits require "
+                         "compatibility level 1.1 or above (use compat=1.1 or "
+                         "greater)");
+            return -EINVAL;
+        }
+
+        helper_cb_info.current_operation = QCOW2_CHANGING_REFCOUNT_ORDER;
+        ret = qcow2_change_refcount_order(bs, refcount_order,
+                                          &qcow2_amend_helper_cb,
+                                          &helper_cb_info, &local_error);
+        if (ret < 0) {
+            qerror_report_err(local_error);
+            return ret;
+        }
+    }
+
     if (backing_file || backing_format) {
         ret = qcow2_change_backing_file(bs, backing_file ?: bs->backing_file,
                                         backing_format ?: bs->backing_format);
-- 
2.1.0

[Qemu-devel] [PATCH v8 09/10] qcow2: Point to amend function in check

[Max Reitz]

If a reference count is not representable with the current refcount
order, the image check should point to qemu-img amend for increasing the
refcount order. However, qemu-img amend needs write access to the image
which cannot be provided if the image is marked corrupt; and the image
check will not mark the image consistent unless everything actually is
consistent.

Therefore, if an image is marked corrupt and the image check encounters
a reference count overflow, it cannot be fixed by using qemu-img amend
to increase the refcount order. Instead, one has to use qemu-img convert
to create a completely new copy of the image in this case.

Alternatively, we may want to give the user a way of manually removing
the corrupt flag, maybe through qemu-img amend, but this is not part of
this patch.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
---
 block/qcow2-refcount.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/block/qcow2-refcount.c b/block/qcow2-refcount.c
index f7e5b5e..46f999c 100644
--- a/block/qcow2-refcount.c
+++ b/block/qcow2-refcount.c
@@ -1340,6 +1340,9 @@ static int inc_refcounts(BlockDriverState *bs,
         if (refcount == s->refcount_max) {
             fprintf(stderr, "ERROR: overflow cluster offset=0x%" PRIx64
                     "\n", cluster_offset);
+            fprintf(stderr, "Use qemu-img amend to increase the refcount entry "
+                    "width or qemu-img convert to create a clean copy if the "
+                    "image cannot be opened for writing\n");
             res->corruptions++;
             continue;
         }
-- 
2.1.0

[Qemu-devel] [PATCH v8 10/10] iotests: Extend test 112 for qemu-img amend

[Max Reitz]

Add tests for conversion between different refcount widths.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
---
 tests/qemu-iotests/112     | 109 +++++++++++++++++++++++++++++++++++++++++++++
 tests/qemu-iotests/112.out |  71 +++++++++++++++++++++++++++++
 2 files changed, 180 insertions(+)

diff --git a/tests/qemu-iotests/112 b/tests/qemu-iotests/112
index 3f054a3..34ba06a 100755
--- a/tests/qemu-iotests/112
+++ b/tests/qemu-iotests/112
@@ -180,6 +180,115 @@ $QEMU_IMG snapshot -c foo "$TEST_IMG"
 # leaked (refcount=UINT64_MAX reference=1)
 _check_test_img
 
+echo
+echo '=== Amend from refcount_bits=16 to refcount_bits=1 ==='
+echo
+
+_make_test_img 64M
+print_refcount_bits
+
+$QEMU_IO -c 'write 16M 32M' "$TEST_IMG" | _filter_qemu_io
+$QEMU_IMG amend -o refcount_bits=1 "$TEST_IMG"
+_check_test_img
+print_refcount_bits
+
+echo
+echo '=== Amend from refcount_bits=1 to refcount_bits=64 ==='
+echo
+
+$QEMU_IMG amend -o refcount_bits=64 "$TEST_IMG"
+_check_test_img
+print_refcount_bits
+
+echo
+echo '=== Amend to compat=0.10 ==='
+echo
+
+# Should not work because refcount_bits needs to be 16 for compat=0.10
+$QEMU_IMG amend -o compat=0.10 "$TEST_IMG"
+print_refcount_bits
+# Should work
+$QEMU_IMG amend -o compat=0.10,refcount_bits=16 "$TEST_IMG"
+_check_test_img
+print_refcount_bits
+
+# Get back to compat=1.1 and refcount_bits=16
+$QEMU_IMG amend -o compat=1.1 "$TEST_IMG"
+print_refcount_bits
+# Should not work
+$QEMU_IMG amend -o refcount_bits=32,compat=0.10 "$TEST_IMG"
+print_refcount_bits
+
+echo
+echo '=== Amend with snapshot ==='
+echo
+
+$QEMU_IMG snapshot -c foo "$TEST_IMG"
+# Just to have different refcounts across the image
+$QEMU_IO -c 'write 0 16M' "$TEST_IMG" | _filter_qemu_io
+
+# Should not work (may work in the future by first decreasing all refcounts so
+# they fit into the target range by copying them)
+$QEMU_IMG amend -o refcount_bits=1 "$TEST_IMG"
+_check_test_img
+print_refcount_bits
+
+# Should work
+$QEMU_IMG amend -o refcount_bits=2 "$TEST_IMG"
+_check_test_img
+print_refcount_bits
+
+echo
+echo '=== Testing too many references for check ==='
+echo
+
+IMGOPTS="$IMGOPTS,refcount_bits=1" _make_test_img 64M
+print_refcount_bits
+
+# This cluster should be created at 0x50000
+$QEMU_IO -c 'write 0 64k' "$TEST_IMG" | _filter_qemu_io
+# Now make the second L2 entry (the L2 table should be at 0x40000) point to that
+# cluster, so we have two references
+poke_file "$TEST_IMG" $((0x40008)) "\x80\x00\x00\x00\x00\x05\x00\x00"
+
+# This should say "please use amend"
+_check_test_img -r all
+
+# So we do that
+$QEMU_IMG amend -o refcount_bits=2 "$TEST_IMG"
+print_refcount_bits
+
+# And try again
+_check_test_img -r all
+
+echo
+echo '=== Multiple walks necessary during amend ==='
+echo
+
+IMGOPTS="$IMGOPTS,refcount_bits=1,cluster_size=512" _make_test_img 64k
+
+# Cluster 0 is the image header, clusters 1 to 4 are used by the L1 table, a
+# single L2 table, the reftable and a single refblock. This creates 58 data
+# clusters (actually, the L2 table is created here, too), so in total there are
+# then 63 used clusters in the image. With a refcount width of 64, one refblock
+# describes 64 clusters (512 bytes / 64 bits/entry = 64 entries), so this will
+# make the first refblock in the amended image have exactly one free entry.
+$QEMU_IO -c "write 0 $((58 * 512))" "$TEST_IMG" | _filter_qemu_io
+
+# Now change the refcount width; since the first new refblock will have exactly
+# one free entry, that entry will be used to store its own reference. No other
+# refblocks are needed, so then the new reftable will be allocated; since the
+# first new refblock is completely filled up, this will require a new refblock
+# which is why the refcount width changing function will need to run through
+# everything one more time until the allocations are stable.
+# Having more walks than usual should be visible as regressing progress (from
+# 66.67 % (2/3 walks) to 50.00 % (2/4 walks)).
+$QEMU_IMG amend -o refcount_bits=64 -p "$TEST_IMG" | tr '\r' '\n' \
+                                                   | grep -A 1 '66.67'
+print_refcount_bits
+
+_check_test_img
+
 
 # success, all done
 echo '*** done'
diff --git a/tests/qemu-iotests/112.out b/tests/qemu-iotests/112.out
index 9a98633..4438076 100644
--- a/tests/qemu-iotests/112.out
+++ b/tests/qemu-iotests/112.out
@@ -81,4 +81,75 @@ Leaked cluster 6 refcount=1 reference=0
 
 2 leaked clusters were found on the image.
 This means waste of disk space, but no harm to data.
+
+=== Amend from refcount_bits=16 to refcount_bits=1 ===
+
+Formatting 'TEST_DIR/t.IMGFMT', fmt=IMGFMT size=67108864
+refcount bits: 16
+wrote 33554432/33554432 bytes at offset 16777216
+32 MiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
+No errors were found on the image.
+refcount bits: 1
+
+=== Amend from refcount_bits=1 to refcount_bits=64 ===
+
+No errors were found on the image.
+refcount bits: 64
+
+=== Amend to compat=0.10 ===
+
+qemu-img: compat=0.10 requires refcount_bits=16
+qemu-img: Error while amending options: Operation not supported
+refcount bits: 64
+No errors were found on the image.
+refcount bits: 16
+refcount bits: 16
+qemu-img: Different refcount widths than 16 bits require compatibility level 1.1 or above (use compat=1.1 or greater)
+qemu-img: Error while amending options: Invalid argument
+refcount bits: 16
+
+=== Amend with snapshot ===
+
+wrote 16777216/16777216 bytes at offset 0
+16 MiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
+qemu-img: Cannot decrease refcount entry width to 1 bits: Cluster at offset 0x50000 has a refcount of 2
+qemu-img: Error while amending options: Invalid argument
+No errors were found on the image.
+refcount bits: 16
+No errors were found on the image.
+refcount bits: 2
+
+=== Testing too many references for check ===
+
+Formatting 'TEST_DIR/t.IMGFMT', fmt=IMGFMT size=67108864
+refcount bits: 1
+wrote 65536/65536 bytes at offset 0
+64 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
+ERROR: overflow cluster offset=0x50000
+Use qemu-img amend to increase the refcount entry width or qemu-img convert to create a clean copy if the image cannot be opened for writing
+
+1 errors were found on the image.
+Data may be corrupted, or further writes to the image may corrupt it.
+refcount bits: 2
+ERROR cluster 5 refcount=1 reference=2
+Repairing cluster 5 refcount=1 reference=2
+Repairing OFLAG_COPIED data cluster: l2_entry=8000000000050000 refcount=2
+Repairing OFLAG_COPIED data cluster: l2_entry=8000000000050000 refcount=2
+The following inconsistencies were found and repaired:
+
+    0 leaked clusters
+    3 corruptions
+
+Double checking the fixed image now...
+No errors were found on the image.
+
+=== Multiple walks necessary during amend ===
+
+Formatting 'TEST_DIR/t.IMGFMT', fmt=IMGFMT size=65536
+wrote 29696/29696 bytes at offset 0
+29 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
+    (66.67/100%)
+    (50.00/100%)
+refcount bits: 64
+No errors were found on the image.
 *** done
-- 
2.1.0

Re: [Qemu-devel] [PATCH v8 00/10] qcow2: Allow refcount_bits amendment

[Max Reitz]

On 18.03.2015 21:56, Max Reitz wrote:
> This series adds support to qemu for changing the refcount_bits option
> of an existing qcow2 file through the qemu-img amend command.
>
> Originally (up until v7), this series was called
> "qcow2: Support refcount orders != 4", but by now that support has
> already been merged so all that is left is support for qemu-img amend.
>
>
> v8:
> - Rebase onto master (one contextual conflict in an iotest reference
>    output, but other than that this only means that the first 15 patches
>    are dropped)

Ping; there is a contextual conflict in patch 8 which I found pretty 
simple to resolve, but if you want me to send a rebase, so be it.

Max

Re: [Qemu-devel] [Qemu-block] [PATCH v8 10/10] iotests: Extend test 112 for qemu-img amend

[Stefan Hajnoczi]

[-- Attachment #1: Type: text/plain, Size: 472 bytes --]

On Wed, Mar 18, 2015 at 04:56:28PM -0400, Max Reitz wrote:
> Add tests for conversion between different refcount widths.
> 
> Signed-off-by: Max Reitz <mreitz@redhat.com>
> Reviewed-by: Eric Blake <eblake@redhat.com>
> ---
>  tests/qemu-iotests/112     | 109 +++++++++++++++++++++++++++++++++++++++++++++
>  tests/qemu-iotests/112.out |  71 +++++++++++++++++++++++++++++
>  2 files changed, 180 insertions(+)

Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>

[-- Attachment #2: Type: application/pgp-signature, Size: 473 bytes --]

Re: [Qemu-devel] [Qemu-block] [PATCH v8 00/10] qcow2: Allow refcount_bits amendment

[Stefan Hajnoczi]

[-- Attachment #1: Type: text/plain, Size: 853 bytes --]

On Fri, Apr 24, 2015 at 05:00:18PM +0200, Max Reitz wrote:
> On 18.03.2015 21:56, Max Reitz wrote:
> >This series adds support to qemu for changing the refcount_bits option
> >of an existing qcow2 file through the qemu-img amend command.
> >
> >Originally (up until v7), this series was called
> >"qcow2: Support refcount orders != 4", but by now that support has
> >already been merged so all that is left is support for qemu-img amend.
> >
> >
> >v8:
> >- Rebase onto master (one contextual conflict in an iotest reference
> >   output, but other than that this only means that the first 15 patches
> >   are dropped)
> 
> Ping; there is a contextual conflict in patch 8 which I found pretty simple
> to resolve, but if you want me to send a rebase, so be it.

All patches have my R-b.

This series should go via Kevin.

Stefan

[-- Attachment #2: Type: application/pgp-signature, Size: 473 bytes --]

[Qemu-devel] [PATCH v8 06/10] qcow2: Use intermediate helper CB for amend

[Max Reitz]

If there is more than one time-consuming operation to be performed for
qcow2_amend_options(), we need an intermediate CB which coordinates the
progress of the individual operations and passes the result to the
original status callback.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
---
 block/qcow2.c | 80 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 79 insertions(+), 1 deletion(-)

diff --git a/block/qcow2.c b/block/qcow2.c
index a1fa348..62a6cb2 100644
--- a/block/qcow2.c
+++ b/block/qcow2.c
@@ -2680,6 +2680,75 @@ static int qcow2_downgrade(BlockDriverState *bs, int target_version,
     return 0;
 }
 
+typedef enum Qcow2AmendOperation {
+    /* This is the value Qcow2AmendHelperCBInfo::last_operation will be
+     * statically initialized to so that the helper CB can discern the first
+     * invocation from an operation change */
+    QCOW2_NO_OPERATION = 0,
+
+    QCOW2_DOWNGRADING,
+} Qcow2AmendOperation;
+
+typedef struct Qcow2AmendHelperCBInfo {
+    /* The code coordinating the amend operations should only modify
+     * these four fields; the rest will be managed by the CB */
+    BlockDriverAmendStatusCB *original_status_cb;
+    void *original_cb_opaque;
+
+    Qcow2AmendOperation current_operation;
+
+    /* Total number of operations to perform (only set once) */
+    int total_operations;
+
+    /* The following fields are managed by the CB */
+
+    /* Number of operations completed */
+    int operations_completed;
+
+    /* Cumulative offset of all completed operations */
+    int64_t offset_completed;
+
+    Qcow2AmendOperation last_operation;
+    int64_t last_work_size;
+} Qcow2AmendHelperCBInfo;
+
+static void qcow2_amend_helper_cb(BlockDriverState *bs,
+                                  int64_t operation_offset,
+                                  int64_t operation_work_size, void *opaque)
+{
+    Qcow2AmendHelperCBInfo *info = opaque;
+    int64_t current_work_size;
+    int64_t projected_work_size;
+
+    if (info->current_operation != info->last_operation) {
+        if (info->last_operation != QCOW2_NO_OPERATION) {
+            info->offset_completed += info->last_work_size;
+            info->operations_completed++;
+        }
+
+        info->last_operation = info->current_operation;
+    }
+
+    assert(info->total_operations > 0);
+    assert(info->operations_completed < info->total_operations);
+
+    info->last_work_size = operation_work_size;
+
+    current_work_size = info->offset_completed + operation_work_size;
+
+    /* current_work_size is the total work size for (operations_completed + 1)
+     * operations (which includes this one), so multiply it by the number of
+     * operations not covered and divide it by the number of operations
+     * covered to get a projection for the operations not covered */
+    projected_work_size = current_work_size * (info->total_operations -
+                                               info->operations_completed - 1)
+                                            / (info->operations_completed + 1);
+
+    info->original_status_cb(bs, info->offset_completed + operation_offset,
+                             current_work_size + projected_work_size,
+                             info->original_cb_opaque);
+}
+
 static int qcow2_amend_options(BlockDriverState *bs, QemuOpts *opts,
                                BlockDriverAmendStatusCB *status_cb,
                                void *cb_opaque)
@@ -2694,6 +2763,7 @@ static int qcow2_amend_options(BlockDriverState *bs, QemuOpts *opts,
     bool encrypt;
     int ret;
     QemuOptDesc *desc = opts->list->desc;
+    Qcow2AmendHelperCBInfo helper_cb_info;
 
     while (desc && desc->name) {
         if (!qemu_opt_find(opts, desc->name)) {
@@ -2752,6 +2822,12 @@ static int qcow2_amend_options(BlockDriverState *bs, QemuOpts *opts,
         desc++;
     }
 
+    helper_cb_info = (Qcow2AmendHelperCBInfo){
+        .original_status_cb = status_cb,
+        .original_cb_opaque = cb_opaque,
+        .total_operations = (new_version < old_version)
+    };
+
     /* Upgrade first (some features may require compat=1.1) */
     if (new_version > old_version) {
         s->qcow_version = new_version;
@@ -2811,7 +2887,9 @@ static int qcow2_amend_options(BlockDriverState *bs, QemuOpts *opts,
 
     /* Downgrade last (so unsupported features can be removed before) */
     if (new_version < old_version) {
-        ret = qcow2_downgrade(bs, new_version, status_cb, cb_opaque);
+        helper_cb_info.current_operation = QCOW2_DOWNGRADING;
+        ret = qcow2_downgrade(bs, new_version, &qcow2_amend_helper_cb,
+                              &helper_cb_info);
         if (ret < 0) {
             return ret;
         }
-- 
2.4.1