Skip to contentSkip to conversation

qcow2: Support refcount orders != 4

PATCH In review Max Reitz 2 participants 22 messages Feb 18, 2015 – Feb 19, 2015 Block layer

lore
Share & tools
lore

Max Reitz

author
#0 lore raw

v7 cover qcow2: Support refcount orders != 4

As of version 3, the qcow2 file format supports different widths for refcount entries, ranging from 1 to 64 bit (only powers of two). Currently, qemu only supports 16 bit, which is the only width supported by version 2 (compat=0.10) images.

This series adds support to qemu for all other valid refcount orders. This is mainly done by adding two function pointers into the BDRVQcowState structure for reading and writing refcount values independently of the current refcount entry width; all in-memory refcount arrays (mostly cached refcount blocks) now are void pointers and are accessed through these functions alone.

Thanks to previous work of making the qemu code agnostic of e.g. the number of refcount entries per refcount block, the rest is fairly trivial. The most complex patch in this series is patch 22 which implements changing the refcount order through qemu-img amend.

To test different refcount widths, simply invoke the qemu-iotests check program with -o refcount_bits=${your_desired_width}. The final test in this series adds some tests for operations which do not work with certain refcount orders and for refcount order amendment.

Patches 1 to 10 are already merged to Kevin's block branch, so I omitted them in this version.

v7:
- Patch 11: Make refcount_bits a uint64_t so it won't overflow in patch 14 [Kevin]
- Patch 14: refcount_bits now being a uint64_t, I dropped the "refcount_bits <= 0" check ("== 0" is checked by is_power_of_2())
- Split the old patch 24 into patches 15 and 25 [Kevin]

git-backport-diff against v6:

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

011/25:[0003] [FC] 'qcow2: refcount_order parameter for qcow2_create2' 012/25:[----] [--] 'qcow2: Use symbolic macros in qcow2_amend_options' 013/25:[----] [--] 'iotests: Prepare for refcount_bits option' 014/25:[0004] [FC] 'qcow2: Allow creation with refcount order != 4' 015/25:[0180] [FC] 'iotests: Add test for different refcount widths' 016/25:[----] [--] 'progress: Allow regressing progress' 017/25:[----] [--] 'block: Add opaque value to the amend CB' 018/25:[----] [--] 'qcow2: Use error_report() in qcow2_amend_options()' 019/25:[----] [--] 'qcow2: Use abort() instead of assert(false)' 020/25:[----] [--] 'qcow2: Split upgrade/downgrade paths for amend' 021/25:[----] [--] 'qcow2: Use intermediate helper CB for amend' 022/25:[----] [--] 'qcow2: Add function for refcount order amendment' 023/25:[----] [--] 'qcow2: Invoke refcount order amendment function' 024/25:[----] [--] 'qcow2: Point to amend function in check' 025/25:[down] 'iotests: Extend test 112 for qemu-img amend'

Max Reitz (15):
qcow2: refcount_order parameter for qcow2_create2
qcow2: Use symbolic macros in qcow2_amend_options
iotests: Prepare for refcount_bits option
qcow2: Allow creation with refcount order != 4
iotests: Add test for different refcount widths
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                    | 270 +++++++++++++++++------
 block/qcow2.h                    |   7 +-
 include/block/block.h            |   4 +-
 include/block/block_int.h        |   4 +-
 qemu-img.c                       |   5 +-
 tests/qemu-iotests/007           |   3 +
 tests/qemu-iotests/015           |   2 +
 tests/qemu-iotests/026           |   7 +
 tests/qemu-iotests/029           |   2 +
 tests/qemu-iotests/049.out       | 108 +++++-----
 tests/qemu-iotests/051           |   3 +
 tests/qemu-iotests/058           |   2 +
 tests/qemu-iotests/061.out       |  14 +-
 tests/qemu-iotests/067           |   2 +
 tests/qemu-iotests/079           |  10 +-
 tests/qemu-iotests/079.out       |  38 +---
 tests/qemu-iotests/080           |   2 +
 tests/qemu-iotests/082.out       |  41 +++-
 tests/qemu-iotests/085.out       |  38 ++--
 tests/qemu-iotests/089           |   2 +
 tests/qemu-iotests/108           |   2 +
 tests/qemu-iotests/112           | 296 +++++++++++++++++++++++++
 tests/qemu-iotests/112.out       | 155 +++++++++++++
 tests/qemu-iotests/common.filter |   3 +-
 tests/qemu-iotests/group         |   1 +
 util/qemu-progress.c             |   3 +-
 29 files changed, 1297 insertions(+), 200 deletions(-)
 create mode 100755 tests/qemu-iotests/112
 create mode 100644 tests/qemu-iotests/112.out
Signature · 1 line
2.1.0
author
#1 lore raw

PATCH v7 11/25 qcow2: refcount_order parameter for qcow2_create2

Add a refcount_order parameter to qcow2_create2(), use that value for the image header and for calculating the size required for preallocation.

For now, always pass 4.

This addition requires changes to the calculation of the file size for the "full" and "falloc" preallocation modes. That in turn is a nice opportunity to add a comment about that calculation not necessarily being exact (and that being intentional).

---
 block/qcow2.c | 47 ++++++++++++++++++++++++++++++++++++-----------
 1 file changed, 36 insertions(+), 11 deletions(-)
block/qcow2.c +36 -11
index e563a30..df2cc0e 100644
@@ -1783,7 +1783,7 @@ static int preallocate(BlockDriverState *bs)
 static int qcow2_create2(const char *filename, int64_t total_size,
                          const char *backing_file, const char *backing_format,
                          int flags, size_t cluster_size, PreallocMode prealloc,
-                         QemuOpts *opts, int version,
+                         QemuOpts *opts, int version, int refcount_order,
                          Error **errp)
 {
     /* Calculate cluster_bits */
@@ -1816,9 +1816,21 @@ static int qcow2_create2(const char *filename, int64_t total_size,
     int ret;
 
     if (prealloc == PREALLOC_MODE_FULL || prealloc == PREALLOC_MODE_FALLOC) {
+        /* Note: The following calculation does not need to be exact; if it is a
+         * bit off, either some bytes will be "leaked" (which is fine) or we
+         * will need to increase the file size by some bytes (which is fine,
+         * too, as long as the bulk is allocated here). Therefore, using
+         * floating point arithmetic is fine. */
         int64_t meta_size = 0;
         uint64_t nreftablee, nrefblocke, nl1e, nl2e;
         int64_t aligned_total_size = align_offset(total_size, cluster_size);
+        int refblock_bits, refblock_size;
+        /* refcount entry size in bytes */
+        double rces = (1 << refcount_order) / 8.;
+
+        /* see qcow2_open() */
+        refblock_bits = cluster_bits - (refcount_order - 3);
+        refblock_size = 1 << refblock_bits;
 
         /* header: 1 cluster */
         meta_size += cluster_size;
@@ -1843,20 +1855,20 @@ static int qcow2_create2(const char *filename, int64_t total_size,
          *   c = cluster size
          *   y1 = number of refcount blocks entries
          *   y2 = meta size including everything
+         *   rces = refcount entry size in bytes
          * then,
          *   y1 = (y2 + a)/c
-         *   y2 = y1 * sizeof(u16) + y1 * sizeof(u16) * sizeof(u64) / c + m
+         *   y2 = y1 * rces + y1 * rces * sizeof(u64) / c + m
          * we can get y1:
-         *   y1 = (a + m) / (c - sizeof(u16) - sizeof(u16) * sizeof(u64) / c)
+         *   y1 = (a + m) / (c - rces - rces * sizeof(u64) / c)
          */
-        nrefblocke = (aligned_total_size + meta_size + cluster_size) /
-            (cluster_size - sizeof(uint16_t) -
-             1.0 * sizeof(uint16_t) * sizeof(uint64_t) / cluster_size);
-        nrefblocke = align_offset(nrefblocke, cluster_size / sizeof(uint16_t));
-        meta_size += nrefblocke * sizeof(uint16_t);
+        nrefblocke = (aligned_total_size + meta_size + cluster_size)
+                   / (cluster_size - rces - rces * sizeof(uint64_t)
+                                                 / cluster_size);
+        meta_size += DIV_ROUND_UP(nrefblocke, refblock_size) * cluster_size;
 
         /* total size of refcount tables */
-        nreftablee = nrefblocke * sizeof(uint16_t) / cluster_size;
+        nreftablee = nrefblocke / refblock_size;
         nreftablee = align_offset(nreftablee, cluster_size / sizeof(uint64_t));
         meta_size += nreftablee * sizeof(uint64_t);
 
@@ -1891,7 +1903,7 @@ static int qcow2_create2(const char *filename, int64_t total_size,
         .l1_size                    = cpu_to_be32(0),
         .refcount_table_offset      = cpu_to_be64(cluster_size),
         .refcount_table_clusters    = cpu_to_be32(1),
-        .refcount_order             = cpu_to_be32(4),
+        .refcount_order             = cpu_to_be32(refcount_order),
         .header_length              = cpu_to_be32(sizeof(*header)),
     };
 
@@ -2010,6 +2022,8 @@ static int qcow2_create(const char *filename, QemuOpts *opts, Error **errp)
     size_t cluster_size = DEFAULT_CLUSTER_SIZE;
     PreallocMode prealloc;
     int version = 3;
+    uint64_t refcount_bits = 16;
+    int refcount_order;
     Error *local_err = NULL;
     int ret;
 
@@ -2064,8 +2078,19 @@ static int qcow2_create(const char *filename, QemuOpts *opts, Error **errp)
         goto finish;
     }
 
+    if (version < 3 && refcount_bits != 16) {
+        error_setg(errp, "Different refcount widths than 16 bits require "
+                   "compatibility level 1.1 or above (use compat=1.1 or "
+                   "greater)");
+        ret = -EINVAL;
+        goto finish;
+    }
+
+    refcount_order = ffs(refcount_bits) - 1;
+
     ret = qcow2_create2(filename, size, backing_file, backing_fmt, flags,
-                        cluster_size, prealloc, opts, version, &local_err);
+                        cluster_size, prealloc, opts, version, refcount_order,
+                        &local_err);
     if (local_err) {
         error_propagate(errp, local_err);
     }
Signature · 1 line
2.1.0
Reviewed-by on v7 11/25
#2 lore raw

Re: [Qemu-devel] [PATCH v7 11/25] qcow2: refcount_order parameter for qcow2_create2

Quoted 29 lines · Max Reitz

Add a refcount_order parameter to qcow2_create2(), use that value for the image header and for calculating the size required for preallocation.

For now, always pass 4.

This addition requires changes to the calculation of the file size for the "full" and "falloc" preallocation modes. That in turn is a nice opportunity to add a comment about that calculation not necessarily being exact (and that being intentional).

---
 block/qcow2.c | 47 ++++++++++++++++++++++++++++++++++++-----------
 1 file changed, 36 insertions(+), 11 deletions(-)
block/qcow2.c +2 -0
@@ -2010,6 +2022,8 @@ static int qcow2_create(const char *filename, QemuOpts *opts, Error **errp)
     size_t cluster_size = DEFAULT_CLUSTER_SIZE;
     PreallocMode prealloc;
     int version = 3;
+    uint64_t refcount_bits = 16;
+    int refcount_order;
     Error *local_err = NULL;
     int ret;
 
@@ -2064,8 +2078,19 @@ static int qcow2_create(const char *filename, QemuOpts *opts, Error **errp)
         goto finish;
     }

block/qcow2.c static int qcow2_create(const char *filename, QemuOpts *opts, Error **errp) · line 2089

Quoted 10 lines
block/qcow2.c +9 -0
 
+    if (version < 3 && refcount_bits != 16) {
+        error_setg(errp, "Different refcount widths than 16 bits require "
+                   "compatibility level 1.1 or above (use compat=1.1 or "
+                   "greater)");
+        ret = -EINVAL;
+        goto finish;
+    }
+
+    refcount_order = ffs(refcount_bits) - 1;

ffs() doesn't work on uint64_t (it gives the wrong answer for 0x100000000, for example); you want to use ffsll(). But ffsll() isn't portable. But we have include/qemu/host-utils.h that gives us ctz64() which is what we want (where ctz==ffs-1 other than for the special case of 0). So this should be:

refcount_order = ctz64(refcount_bits);

With that change,

(Hmm, that means that at least qcow2.c:qcow2_create2() has a bug for calling ffs(size_t), and we probably ought to eradicate other uses of ffs from the tree - but as a separate followup).

Signature · 2 lines
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org
author
#3 lore raw

Re: [Qemu-devel] [PATCH v7 11/25] qcow2: refcount_order parameter for qcow2_create2

Quoted 45 lines · Eric Blake
Quoted 28 lines · Max Reitz

Add a refcount_order parameter to qcow2_create2(), use that value for the image header and for calculating the size required for preallocation.

For now, always pass 4.

This addition requires changes to the calculation of the file size for the "full" and "falloc" preallocation modes. That in turn is a nice opportunity to add a comment about that calculation not necessarily being exact (and that being intentional).

---
  block/qcow2.c | 47 ++++++++++++++++++++++++++++++++++++-----------
  1 file changed, 36 insertions(+), 11 deletions(-)
block/qcow2.c +2 -0
@@ -2010,6 +2022,8 @@ static int qcow2_create(const char *filename, QemuOpts *opts, Error **errp)
      size_t cluster_size = DEFAULT_CLUSTER_SIZE;
      PreallocMode prealloc;
      int version = 3;
+    uint64_t refcount_bits = 16;
+    int refcount_order;
      Error *local_err = NULL;
      int ret;
 
@@ -2064,8 +2078,19 @@ static int qcow2_create(const char *filename, QemuOpts *opts, Error **errp)
          goto finish;
      }

block/qcow2.c static int qcow2_create(const char *filename, QemuOpts *opts, Error **errp) · line 2089

Quoted 10 lines
block/qcow2.c +9 -0
 
+    if (version < 3 && refcount_bits != 16) {
+        error_setg(errp, "Different refcount widths than 16 bits require "
+                   "compatibility level 1.1 or above (use compat=1.1 or "
+                   "greater)");
+        ret = -EINVAL;
+        goto finish;
+    }
+
+    refcount_order = ffs(refcount_bits) - 1;

ffs() doesn't work on uint64_t (it gives the wrong answer for 0x100000000, for example); you want to use ffsll(). But ffsll() isn't portable. But we have include/qemu/host-utils.h that gives us ctz64() which is what we want (where ctz==ffs-1 other than for the special case of 0). So this should be:

Quoted 4 lines

refcount_order = ctz64(refcount_bits);

With that change,

I intentionally left the ffs() here, because after this patch, refcount_bits is guaranteed to be 16, and after patch 14, it's guaranteed to be 64 or less (it's more obvious after patch 14 than here). So I would be fine with ctz64(), but in my opinion, ffs() is just fine, too.

Max

Quoted 4 lines

(Hmm, that means that at least qcow2.c:qcow2_create2() has a bug for calling ffs(size_t), and we probably ought to eradicate other uses of ffs from the tree - but as a separate followup).

#4 lore raw

Re: [Qemu-devel] [PATCH v7 11/25] qcow2: refcount_order parameter for qcow2_create2

Quoted 18 lines · Max Reitz
Quoted 18 lines

block/qcow2.c static int qcow2_create(const char *filename,

Quoted 10 lines
block/qcow2.c +3 -0
@@ -2010,6 +2022,8 @@ static int qcow2_create(const char *filename,
 QemuOpts *opts, Error **errp)
      size_t cluster_size = DEFAULT_CLUSTER_SIZE;
      PreallocMode prealloc;
      int version = 3;
+    uint64_t refcount_bits = 16;
+    int refcount_order;
      Error *local_err = NULL;
 
+    refcount_order = ffs(refcount_bits) - 1;

ffs() doesn't work on uint64_t (it gives the wrong answer for 0x100000000, for example); you want to use ffsll(). But ffsll() isn't portable. But we have include/qemu/host-utils.h that gives us ctz64() which is what we want (where ctz==ffs-1 other than for the special case of 0). So this should be:

refcount_order = ctz64(refcount_bits);

Quoted 8 lines
Quoted 2 lines

With that change,

I intentionally left the ffs() here, because after this patch, refcount_bits is guaranteed to be 16, and after patch 14, it's guaranteed to be 64 or less (it's more obvious after patch 14 than here). So I would be fine with ctz64(), but in my opinion, ffs() is just fine, too.

Ah, you are right - here, we are calling ffs(constant), and in patch 14, the ffs() is quite obviously bounded by a range check, so what you have works, even if it is not as short as possible, and even though it caused me to think about integer wrapping.

I'll leave it up to Kevin whether to switch to ctz64() or stick with ffs().

Quoted 6 lines

Max

Quoted 3 lines

(Hmm, that means that at least qcow2.c:qcow2_create2() has a bug for calling ffs(size_t), and we probably ought to eradicate other uses of ffs from the tree - but as a separate followup).

This still remains true - we should scrub all uses of ffs() on size_t or uint64_t and make sure they can't suffer from wraparound issues, but doing that shouldn't hold up this series.

Signature · 2 lines
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org
author
#5 lore raw

PATCH v7 12/25 qcow2: Use symbolic macros in qcow2_amend_options

In reply to Max Reitz

qcow2_amend_options() should not compare options against some inline strings but rather use the symbolic macros available for each of the creation options.

---
 block/qcow2.c | 31 ++++++++++++++++---------------
 1 file changed, 16 insertions(+), 15 deletions(-)
block/qcow2.c +16 -15
index df2cc0e..a07f134 100644
@@ -2677,8 +2677,8 @@ static int qcow2_amend_options(BlockDriverState *bs, QemuOpts *opts,
             continue;
         }
 
-        if (!strcmp(desc->name, "compat")) {
-            compat = qemu_opt_get(opts, "compat");
+        if (!strcmp(desc->name, BLOCK_OPT_COMPAT_LEVEL)) {
+            compat = qemu_opt_get(opts, BLOCK_OPT_COMPAT_LEVEL);
             if (!compat) {
                 /* preserve default */
             } else if (!strcmp(compat, "0.10")) {
@@ -2689,32 +2689,33 @@ static int qcow2_amend_options(BlockDriverState *bs, QemuOpts *opts,
                 fprintf(stderr, "Unknown compatibility level %s.\n", compat);
                 return -EINVAL;
             }
-        } else if (!strcmp(desc->name, "preallocation")) {
+        } else if (!strcmp(desc->name, BLOCK_OPT_PREALLOC)) {
             fprintf(stderr, "Cannot change preallocation mode.\n");
             return -ENOTSUP;
-        } else if (!strcmp(desc->name, "size")) {
-            new_size = qemu_opt_get_size(opts, "size", 0);
-        } else if (!strcmp(desc->name, "backing_file")) {
-            backing_file = qemu_opt_get(opts, "backing_file");
-        } else if (!strcmp(desc->name, "backing_fmt")) {
-            backing_format = qemu_opt_get(opts, "backing_fmt");
-        } else if (!strcmp(desc->name, "encryption")) {
-            encrypt = qemu_opt_get_bool(opts, "encryption", s->crypt_method);
+        } else if (!strcmp(desc->name, BLOCK_OPT_SIZE)) {
+            new_size = qemu_opt_get_size(opts, BLOCK_OPT_SIZE, 0);
+        } else if (!strcmp(desc->name, BLOCK_OPT_BACKING_FILE)) {
+            backing_file = qemu_opt_get(opts, BLOCK_OPT_BACKING_FILE);
+        } else if (!strcmp(desc->name, BLOCK_OPT_BACKING_FMT)) {
+            backing_format = qemu_opt_get(opts, BLOCK_OPT_BACKING_FMT);
+        } else if (!strcmp(desc->name, BLOCK_OPT_ENCRYPT)) {
+            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");
                 return -ENOTSUP;
             }
-        } else if (!strcmp(desc->name, "cluster_size")) {
-            cluster_size = qemu_opt_get_size(opts, "cluster_size",
+        } 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");
                 return -ENOTSUP;
             }
-        } else if (!strcmp(desc->name, "lazy_refcounts")) {
-            lazy_refcounts = qemu_opt_get_bool(opts, "lazy_refcounts",
+        } else if (!strcmp(desc->name, BLOCK_OPT_LAZY_REFCOUNTS)) {
+            lazy_refcounts = qemu_opt_get_bool(opts, BLOCK_OPT_LAZY_REFCOUNTS,
                                                lazy_refcounts);
         } else {
             /* if this assertion fails, this probably means a new option was
Signature · 1 line
2.1.0
author
#6 lore raw

PATCH v7 13/25 iotests: Prepare for refcount_bits option

In reply to Max Reitz

Some tests do not work well with certain refcount widths (i.e. you cannot create internal snapshots with refcount_bits=1), so make those widths unsupported.

Furthermore, add another filter to _filter_img_create in common.filter which filters out the refcount_bits value.

This is necessary for test 079, which does actually work with any refcount width, but invoking qemu-img directly leads to the refcount_bits value being visible in the output; use _make_test_img instead which will filter it out.

---
 tests/qemu-iotests/007           |  3 +++
 tests/qemu-iotests/015           |  2 ++
 tests/qemu-iotests/026           |  7 +++++++
 tests/qemu-iotests/029           |  2 ++
 tests/qemu-iotests/051           |  3 +++
 tests/qemu-iotests/058           |  2 ++
 tests/qemu-iotests/067           |  2 ++
 tests/qemu-iotests/079           | 10 ++--------
 tests/qemu-iotests/079.out       | 38 ++++++++++----------------------------
 tests/qemu-iotests/080           |  2 ++
 tests/qemu-iotests/089           |  2 ++
 tests/qemu-iotests/108           |  2 ++
 tests/qemu-iotests/common.filter |  3 ++-
 13 files changed, 41 insertions(+), 37 deletions(-)
tests/qemu-iotests/007 +3 -0
index fe1a743..7b5aff5 100755
@@ -43,6 +43,9 @@ trap "_cleanup; exit \$status" 0 1 2 3 15
 _supported_fmt qcow2
 _supported_proto generic
 _supported_os Linux
+# refcount_bits must be at least 4 so we can create ten internal snapshots
+# (1 bit supports none, 2 bits support two, 4 bits support 14)
+_unsupported_imgopts 'refcount_bits=\(1\|2\)[^0-9]'
 
 echo
 echo "creating image"
tests/qemu-iotests/015 +2 -0
index 099d757..6f26095 100755
@@ -43,6 +43,8 @@ trap "_cleanup; exit \$status" 0 1 2 3 15
 _supported_fmt qcow2
 _supported_proto generic
 _supported_os Linux
+# Internal snapshots are (currently) impossible with refcount_bits=1
+_unsupported_imgopts 'refcount_bits=1[^0-9]'
 
 echo
 echo "creating image"
tests/qemu-iotests/026 +7 -0
index df2884b..0fc3244 100755
@@ -46,6 +46,13 @@ _supported_proto file
 _supported_os Linux
 _default_cache_mode "writethrough"
 _supported_cache_modes "writethrough" "none"
+# The refcount table tests expect a certain minimum width for refcount entries
+# (so that the refcount table actually needs to grow); that minimum is 16 bits,
+# being the default refcount entry width.
+# 32 and 64 bits do not work either, however, due to different leaked cluster
+# count on error.
+# Thus, the only remaining option is refcount_bits=16.
+_unsupported_imgopts 'refcount_bits=\([^1]\|.\([^6]\|$\)\)'
 
 echo "Errors while writing 128 kB"
 echo
tests/qemu-iotests/029 +2 -0
index fa46ace..b9cd826 100755
@@ -44,6 +44,8 @@ trap "_cleanup; exit \$status" 0 1 2 3 15
 _supported_fmt qcow2
 _supported_proto generic
 _supported_os Linux
+# Internal snapshots are (currently) impossible with refcount_bits=1
+_unsupported_imgopts 'refcount_bits=1[^0-9]'
 
 offset_size=24
 offset_l1_size=36
tests/qemu-iotests/051 +3 -0
index 11c858f..1677592 100755
@@ -41,6 +41,9 @@ trap "_cleanup; exit \$status" 0 1 2 3 15
 _supported_fmt qcow2
 _supported_proto file
 _supported_os Linux
+# A compat=0.10 image is created in this test which does not support anything
+# other than refcount_bits=16
+_unsupported_imgopts 'refcount_bits=\([^1]\|.\([^6]\|$\)\)'
 
 function do_run_qemu()
 {
tests/qemu-iotests/058 +2 -0
index a60b34b..f2bdd0b 100755
@@ -89,6 +89,8 @@ _supported_fmt qcow2
 _supported_proto file
 _supported_os Linux
 _require_command QEMU_NBD
+# Internal snapshots are (currently) impossible with refcount_bits=1
+_unsupported_imgopts 'refcount_bits=1[^0-9]'
 
 # Use -f raw instead of -f $IMGFMT for the NBD connection
 QEMU_IO_NBD="$QEMU_IO -f raw --cache=$CACHEMODE"
tests/qemu-iotests/067 +2 -0
index 0508c69..83eefa3 100755
@@ -35,6 +35,8 @@ status=1	# failure is the default!
 _supported_fmt qcow2
 _supported_proto file
 _supported_os Linux
+# Because anything other than 16 would change the output of query-block
+_unsupported_imgopts 'refcount_bits=\([^1]\|.\([^6]\|$\)\)'
 
 function do_run_qemu()
 {
tests/qemu-iotests/079 +2 -8
index 6613cfb..ade6efa 100755
@@ -42,19 +42,13 @@ _supported_fmt qcow2
 _supported_proto file nfs
 _supported_os Linux
 
-function test_qemu_img()
-{
-    echo qemu-img "$@" | _filter_testdir
-    $QEMU_IMG "$@" 2>&1 | _filter_testdir
-    echo
-}
-
 echo "=== Check option preallocation and cluster_size ==="
 echo
 cluster_sizes="16384 32768 65536 131072 262144 524288 1048576 2097152 4194304"
 
 for s in $cluster_sizes; do
-    test_qemu_img create -f $IMGFMT -o preallocation=metadata,cluster_size=$s "$TEST_IMG" 4G
+    IMGOPTS=$(_optstr_add "$IMGOPTS" "preallocation=metadata,cluster_size=$s") \
+        _make_test_img 4G
 done
 
 # success, all done
tests/qemu-iotests/079.out +10 -28
index ef4b8c9..6dc5d57 100644
@@ -1,32 +1,14 @@
 QA output created by 079
 === Check option preallocation and cluster_size ===
 
-qemu-img create -f qcow2 -o preallocation=metadata,cluster_size=16384 TEST_DIR/t.qcow2 4G
-Formatting 'TEST_DIR/t.qcow2', fmt=qcow2 size=4294967296 encryption=off cluster_size=16384 preallocation='metadata' lazy_refcounts=off
-
-qemu-img create -f qcow2 -o preallocation=metadata,cluster_size=32768 TEST_DIR/t.qcow2 4G
-Formatting 'TEST_DIR/t.qcow2', fmt=qcow2 size=4294967296 encryption=off cluster_size=32768 preallocation='metadata' lazy_refcounts=off
-
-qemu-img create -f qcow2 -o preallocation=metadata,cluster_size=65536 TEST_DIR/t.qcow2 4G
-Formatting 'TEST_DIR/t.qcow2', fmt=qcow2 size=4294967296 encryption=off cluster_size=65536 preallocation='metadata' lazy_refcounts=off
-
-qemu-img create -f qcow2 -o preallocation=metadata,cluster_size=131072 TEST_DIR/t.qcow2 4G
-Formatting 'TEST_DIR/t.qcow2', fmt=qcow2 size=4294967296 encryption=off cluster_size=131072 preallocation='metadata' lazy_refcounts=off
-
-qemu-img create -f qcow2 -o preallocation=metadata,cluster_size=262144 TEST_DIR/t.qcow2 4G
-Formatting 'TEST_DIR/t.qcow2', fmt=qcow2 size=4294967296 encryption=off cluster_size=262144 preallocation='metadata' lazy_refcounts=off
-
-qemu-img create -f qcow2 -o preallocation=metadata,cluster_size=524288 TEST_DIR/t.qcow2 4G
-Formatting 'TEST_DIR/t.qcow2', fmt=qcow2 size=4294967296 encryption=off cluster_size=524288 preallocation='metadata' lazy_refcounts=off
-
-qemu-img create -f qcow2 -o preallocation=metadata,cluster_size=1048576 TEST_DIR/t.qcow2 4G
-Formatting 'TEST_DIR/t.qcow2', fmt=qcow2 size=4294967296 encryption=off cluster_size=1048576 preallocation='metadata' lazy_refcounts=off
-
-qemu-img create -f qcow2 -o preallocation=metadata,cluster_size=2097152 TEST_DIR/t.qcow2 4G
-Formatting 'TEST_DIR/t.qcow2', fmt=qcow2 size=4294967296 encryption=off cluster_size=2097152 preallocation='metadata' lazy_refcounts=off
-
-qemu-img create -f qcow2 -o preallocation=metadata,cluster_size=4194304 TEST_DIR/t.qcow2 4G
-qemu-img: TEST_DIR/t.qcow2: Cluster size must be a power of two between 512 and 2048k
-Formatting 'TEST_DIR/t.qcow2', fmt=qcow2 size=4294967296 encryption=off cluster_size=4194304 preallocation='metadata' lazy_refcounts=off
-
+Formatting 'TEST_DIR/t.IMGFMT', fmt=IMGFMT size=4294967296 preallocation='metadata'
+Formatting 'TEST_DIR/t.IMGFMT', fmt=IMGFMT size=4294967296 preallocation='metadata'
+Formatting 'TEST_DIR/t.IMGFMT', fmt=IMGFMT size=4294967296 preallocation='metadata'
+Formatting 'TEST_DIR/t.IMGFMT', fmt=IMGFMT size=4294967296 preallocation='metadata'
+Formatting 'TEST_DIR/t.IMGFMT', fmt=IMGFMT size=4294967296 preallocation='metadata'
+Formatting 'TEST_DIR/t.IMGFMT', fmt=IMGFMT size=4294967296 preallocation='metadata'
+Formatting 'TEST_DIR/t.IMGFMT', fmt=IMGFMT size=4294967296 preallocation='metadata'
+Formatting 'TEST_DIR/t.IMGFMT', fmt=IMGFMT size=4294967296 preallocation='metadata'
+qemu-img: TEST_DIR/t.IMGFMT: Cluster size must be a power of two between 512 and 2048k
+Formatting 'TEST_DIR/t.IMGFMT', fmt=IMGFMT size=4294967296 preallocation='metadata'
 *** done
tests/qemu-iotests/080 +2 -0
index 73795f1..a2c58ae 100755
@@ -42,6 +42,8 @@ trap "_cleanup; exit \$status" 0 1 2 3 15
 _supported_fmt qcow2
 _supported_proto file
 _supported_os Linux
+# Internal snapshots are (currently) impossible with refcount_bits=1
+_unsupported_imgopts 'refcount_bits=1[^0-9]'
 
 header_size=104
 
tests/qemu-iotests/089 +2 -0
index 6f74cec..3e0038d 100755
@@ -41,6 +41,8 @@ trap "_cleanup; exit \$status" 0 1 2 3 15
 _supported_fmt qcow2
 _supported_proto file
 _supported_os Linux
+# Because anything other than 16 would change the output of qemu_io -c info
+_unsupported_imgopts 'refcount_bits=\([^1]\|.\([^6]\|$\)\)'
 
 # Using an image filename containing quotation marks will render the JSON data
 # below invalid. In that case, we have little choice but simply not to run this
tests/qemu-iotests/108 +2 -0
index 12fc92a..ce44749 100755
@@ -43,6 +43,8 @@ trap "_cleanup; exit \$status" 0 1 2 3 15
 _supported_fmt qcow2
 _supported_proto file
 _supported_os Linux
+# This test directly modifies a refblock so it relies on refcount_bits being 16
+_unsupported_imgopts 'refcount_bits=\([^1]\|.\([^6]\|$\)\)'
 
 echo
 echo '=== Repairing an image without any refcount table ==='
tests/qemu-iotests/common.filter +2 -1
index 06e1bb0..012a812 100644
@@ -192,7 +192,8 @@ _filter_img_create()
         -e "s# block_size=[0-9]\\+##g" \
         -e "s# block_state_zero=\\(on\\|off\\)##g" \
         -e "s# log_size=[0-9]\\+##g" \
-        -e "s/archipelago:a/TEST_DIR\//g"
+        -e "s/archipelago:a/TEST_DIR\//g" \
+        -e "s# refcount_bits=[0-9]\\+##g"
 }
 
 _filter_img_info()
Signature · 1 line
2.1.0
author
#7 lore raw

PATCH v7 14/25 qcow2: Allow creation with refcount order != 4

In reply to Max Reitz

Add a creation option to qcow2 for setting the refcount order of images to be created, and respect that option's value.

This breaks some test outputs, fix them.

---
 block/qcow2.c              |  18 ++++++++
 include/block/block_int.h  |   1 +
 tests/qemu-iotests/049.out | 108 ++++++++++++++++++++++-----------------------
 tests/qemu-iotests/082.out |  41 ++++++++++++++---
 tests/qemu-iotests/085.out |  38 ++++++++--------
 5 files changed, 126 insertions(+), 80 deletions(-)
block/qcow2.c +18 -0
index a07f134..4cdbd7e 100644
@@ -2078,6 +2078,15 @@ static int qcow2_create(const char *filename, QemuOpts *opts, Error **errp)
         goto finish;
     }
 
+    refcount_bits = qemu_opt_get_number_del(opts, BLOCK_OPT_REFCOUNT_BITS,
+                                            refcount_bits);
+    if (refcount_bits > 64 || !is_power_of_2(refcount_bits)) {
+        error_setg(errp, "Refcount width must be a power of two and may not "
+                   "exceed 64 bits");
+        ret = -EINVAL;
+        goto finish;
+    }
+
     if (version < 3 && refcount_bits != 16) {
         error_setg(errp, "Different refcount widths than 16 bits require "
                    "compatibility level 1.1 or above (use compat=1.1 or "
@@ -2717,6 +2726,9 @@ static int qcow2_amend_options(BlockDriverState *bs, QemuOpts *opts,
         } else if (!strcmp(desc->name, BLOCK_OPT_LAZY_REFCOUNTS)) {
             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;
         } else {
             /* if this assertion fails, this probably means a new option was
              * added without having it covered here */
@@ -2886,6 +2898,12 @@ static QemuOptsList qcow2_create_opts = {
             .help = "Postpone refcount updates",
             .def_value_str = "off"
         },
+        {
+            .name = BLOCK_OPT_REFCOUNT_BITS,
+            .type = QEMU_OPT_NUMBER,
+            .help = "Width of a reference count entry in bits",
+            .def_value_str = "16"
+        },
         { /* end of list */ }
     }
 };
include/block/block_int.h +1 -0
index 5e718a3..b661b43 100644
@@ -57,6 +57,7 @@
 #define BLOCK_OPT_REDUNDANCY        "redundancy"
 #define BLOCK_OPT_NOCOW             "nocow"
 #define BLOCK_OPT_OBJECT_SIZE       "object_size"
+#define BLOCK_OPT_REFCOUNT_BITS     "refcount_bits"
 
 #define BLOCK_PROBE_BUF_SIZE        512
 
tests/qemu-iotests/049.out +54 -54

249 lines not on this page · raw message

tests/qemu-iotests/082.out +34 -7

238 lines not on this page · raw message

tests/qemu-iotests/085.out +19 -19
index 0f2b17f..5eb8b94 100644
@@ -11,7 +11,7 @@ Formatting 'TEST_DIR/t.IMGFMT', fmt=IMGFMT size=134217728
 
 === Create a single snapshot on virtio0 ===
 
-Formatting 'TEST_DIR/1-snapshot-v0.qcow2', fmt=qcow2 size=134217728 backing_file='TEST_DIR/t.qcow2.orig' backing_fmt='qcow2' encryption=off cluster_size=65536 lazy_refcounts=off
+Formatting 'TEST_DIR/1-snapshot-v0.qcow2', fmt=qcow2 size=134217728 backing_file='TEST_DIR/t.qcow2.orig' backing_fmt='qcow2' encryption=off cluster_size=65536 lazy_refcounts=off refcount_bits=16
 {"return": {}}
 
 === Invalid command - missing device and nodename ===
@@ -25,31 +25,31 @@ Formatting 'TEST_DIR/1-snapshot-v0.qcow2', fmt=qcow2 size=134217728 backing_file
 
 === Create several transactional group snapshots ===
 
-Formatting 'TEST_DIR/2-snapshot-v0.qcow2', fmt=qcow2 size=134217728 backing_file='TEST_DIR/1-snapshot-v0.qcow2' backing_fmt='qcow2' encryption=off cluster_size=65536 lazy_refcounts=off
-Formatting 'TEST_DIR/2-snapshot-v1.qcow2', fmt=qcow2 size=134217728 backing_file='TEST_DIR/t.qcow2' backing_fmt='qcow2' encryption=off cluster_size=65536 lazy_refcounts=off
+Formatting 'TEST_DIR/2-snapshot-v0.qcow2', fmt=qcow2 size=134217728 backing_file='TEST_DIR/1-snapshot-v0.qcow2' backing_fmt='qcow2' encryption=off cluster_size=65536 lazy_refcounts=off refcount_bits=16
+Formatting 'TEST_DIR/2-snapshot-v1.qcow2', fmt=qcow2 size=134217728 backing_file='TEST_DIR/t.qcow2' backing_fmt='qcow2' encryption=off cluster_size=65536 lazy_refcounts=off refcount_bits=16
 {"return": {}}
-Formatting 'TEST_DIR/3-snapshot-v0.qcow2', fmt=qcow2 size=134217728 backing_file='TEST_DIR/2-snapshot-v0.qcow2' backing_fmt='qcow2' encryption=off cluster_size=65536 lazy_refcounts=off
-Formatting 'TEST_DIR/3-snapshot-v1.qcow2', fmt=qcow2 size=134217728 backing_file='TEST_DIR/2-snapshot-v1.qcow2' backing_fmt='qcow2' encryption=off cluster_size=65536 lazy_refcounts=off
+Formatting 'TEST_DIR/3-snapshot-v0.qcow2', fmt=qcow2 size=134217728 backing_file='TEST_DIR/2-snapshot-v0.qcow2' backing_fmt='qcow2' encryption=off cluster_size=65536 lazy_refcounts=off refcount_bits=16
+Formatting 'TEST_DIR/3-snapshot-v1.qcow2', fmt=qcow2 size=134217728 backing_file='TEST_DIR/2-snapshot-v1.qcow2' backing_fmt='qcow2' encryption=off cluster_size=65536 lazy_refcounts=off refcount_bits=16
 {"return": {}}
-Formatting 'TEST_DIR/4-snapshot-v0.qcow2', fmt=qcow2 size=134217728 backing_file='TEST_DIR/3-snapshot-v0.qcow2' backing_fmt='qcow2' encryption=off cluster_size=65536 lazy_refcounts=off
-Formatting 'TEST_DIR/4-snapshot-v1.qcow2', fmt=qcow2 size=134217728 backing_file='TEST_DIR/3-snapshot-v1.qcow2' backing_fmt='qcow2' encryption=off cluster_size=65536 lazy_refcounts=off
+Formatting 'TEST_DIR/4-snapshot-v0.qcow2', fmt=qcow2 size=134217728 backing_file='TEST_DIR/3-snapshot-v0.qcow2' backing_fmt='qcow2' encryption=off cluster_size=65536 lazy_refcounts=off refcount_bits=16
+Formatting 'TEST_DIR/4-snapshot-v1.qcow2', fmt=qcow2 size=134217728 backing_file='TEST_DIR/3-snapshot-v1.qcow2' backing_fmt='qcow2' encryption=off cluster_size=65536 lazy_refcounts=off refcount_bits=16
 {"return": {}}
-Formatting 'TEST_DIR/5-snapshot-v0.qcow2', fmt=qcow2 size=134217728 backing_file='TEST_DIR/4-snapshot-v0.qcow2' backing_fmt='qcow2' encryption=off cluster_size=65536 lazy_refcounts=off
-Formatting 'TEST_DIR/5-snapshot-v1.qcow2', fmt=qcow2 size=134217728 backing_file='TEST_DIR/4-snapshot-v1.qcow2' backing_fmt='qcow2' encryption=off cluster_size=65536 lazy_refcounts=off
+Formatting 'TEST_DIR/5-snapshot-v0.qcow2', fmt=qcow2 size=134217728 backing_file='TEST_DIR/4-snapshot-v0.qcow2' backing_fmt='qcow2' encryption=off cluster_size=65536 lazy_refcounts=off refcount_bits=16
+Formatting 'TEST_DIR/5-snapshot-v1.qcow2', fmt=qcow2 size=134217728 backing_file='TEST_DIR/4-snapshot-v1.qcow2' backing_fmt='qcow2' encryption=off cluster_size=65536 lazy_refcounts=off refcount_bits=16
 {"return": {}}
-Formatting 'TEST_DIR/6-snapshot-v0.qcow2', fmt=qcow2 size=134217728 backing_file='TEST_DIR/5-snapshot-v0.qcow2' backing_fmt='qcow2' encryption=off cluster_size=65536 lazy_refcounts=off
-Formatting 'TEST_DIR/6-snapshot-v1.qcow2', fmt=qcow2 size=134217728 backing_file='TEST_DIR/5-snapshot-v1.qcow2' backing_fmt='qcow2' encryption=off cluster_size=65536 lazy_refcounts=off
+Formatting 'TEST_DIR/6-snapshot-v0.qcow2', fmt=qcow2 size=134217728 backing_file='TEST_DIR/5-snapshot-v0.qcow2' backing_fmt='qcow2' encryption=off cluster_size=65536 lazy_refcounts=off refcount_bits=16
+Formatting 'TEST_DIR/6-snapshot-v1.qcow2', fmt=qcow2 size=134217728 backing_file='TEST_DIR/5-snapshot-v1.qcow2' backing_fmt='qcow2' encryption=off cluster_size=65536 lazy_refcounts=off refcount_bits=16
 {"return": {}}
-Formatting 'TEST_DIR/7-snapshot-v0.qcow2', fmt=qcow2 size=134217728 backing_file='TEST_DIR/6-snapshot-v0.qcow2' backing_fmt='qcow2' encryption=off cluster_size=65536 lazy_refcounts=off
-Formatting 'TEST_DIR/7-snapshot-v1.qcow2', fmt=qcow2 size=134217728 backing_file='TEST_DIR/6-snapshot-v1.qcow2' backing_fmt='qcow2' encryption=off cluster_size=65536 lazy_refcounts=off
+Formatting 'TEST_DIR/7-snapshot-v0.qcow2', fmt=qcow2 size=134217728 backing_file='TEST_DIR/6-snapshot-v0.qcow2' backing_fmt='qcow2' encryption=off cluster_size=65536 lazy_refcounts=off refcount_bits=16
+Formatting 'TEST_DIR/7-snapshot-v1.qcow2', fmt=qcow2 size=134217728 backing_file='TEST_DIR/6-snapshot-v1.qcow2' backing_fmt='qcow2' encryption=off cluster_size=65536 lazy_refcounts=off refcount_bits=16
 {"return": {}}
-Formatting 'TEST_DIR/8-snapshot-v0.qcow2', fmt=qcow2 size=134217728 backing_file='TEST_DIR/7-snapshot-v0.qcow2' backing_fmt='qcow2' encryption=off cluster_size=65536 lazy_refcounts=off
-Formatting 'TEST_DIR/8-snapshot-v1.qcow2', fmt=qcow2 size=134217728 backing_file='TEST_DIR/7-snapshot-v1.qcow2' backing_fmt='qcow2' encryption=off cluster_size=65536 lazy_refcounts=off
+Formatting 'TEST_DIR/8-snapshot-v0.qcow2', fmt=qcow2 size=134217728 backing_file='TEST_DIR/7-snapshot-v0.qcow2' backing_fmt='qcow2' encryption=off cluster_size=65536 lazy_refcounts=off refcount_bits=16
+Formatting 'TEST_DIR/8-snapshot-v1.qcow2', fmt=qcow2 size=134217728 backing_file='TEST_DIR/7-snapshot-v1.qcow2' backing_fmt='qcow2' encryption=off cluster_size=65536 lazy_refcounts=off refcount_bits=16
 {"return": {}}
-Formatting 'TEST_DIR/9-snapshot-v0.qcow2', fmt=qcow2 size=134217728 backing_file='TEST_DIR/8-snapshot-v0.qcow2' backing_fmt='qcow2' encryption=off cluster_size=65536 lazy_refcounts=off
-Formatting 'TEST_DIR/9-snapshot-v1.qcow2', fmt=qcow2 size=134217728 backing_file='TEST_DIR/8-snapshot-v1.qcow2' backing_fmt='qcow2' encryption=off cluster_size=65536 lazy_refcounts=off
+Formatting 'TEST_DIR/9-snapshot-v0.qcow2', fmt=qcow2 size=134217728 backing_file='TEST_DIR/8-snapshot-v0.qcow2' backing_fmt='qcow2' encryption=off cluster_size=65536 lazy_refcounts=off refcount_bits=16
+Formatting 'TEST_DIR/9-snapshot-v1.qcow2', fmt=qcow2 size=134217728 backing_file='TEST_DIR/8-snapshot-v1.qcow2' backing_fmt='qcow2' encryption=off cluster_size=65536 lazy_refcounts=off refcount_bits=16
 {"return": {}}
-Formatting 'TEST_DIR/10-snapshot-v0.qcow2', fmt=qcow2 size=134217728 backing_file='TEST_DIR/9-snapshot-v0.qcow2' backing_fmt='qcow2' encryption=off cluster_size=65536 lazy_refcounts=off
-Formatting 'TEST_DIR/10-snapshot-v1.qcow2', fmt=qcow2 size=134217728 backing_file='TEST_DIR/9-snapshot-v1.qcow2' backing_fmt='qcow2' encryption=off cluster_size=65536 lazy_refcounts=off
+Formatting 'TEST_DIR/10-snapshot-v0.qcow2', fmt=qcow2 size=134217728 backing_file='TEST_DIR/9-snapshot-v0.qcow2' backing_fmt='qcow2' encryption=off cluster_size=65536 lazy_refcounts=off refcount_bits=16
+Formatting 'TEST_DIR/10-snapshot-v1.qcow2', fmt=qcow2 size=134217728 backing_file='TEST_DIR/9-snapshot-v1.qcow2' backing_fmt='qcow2' encryption=off cluster_size=65536 lazy_refcounts=off refcount_bits=16
 {"return": {}}
 *** done
Signature · 1 line
2.1.0
Reviewed-by on v7 14/25
#8 lore raw

Re: [Qemu-devel] [PATCH v7 14/25] qcow2: Allow creation with refcount order != 4

Quoted 24 lines · Max Reitz

Add a creation option to qcow2 for setting the refcount order of images to be created, and respect that option's value.

This breaks some test outputs, fix them.

---
 block/qcow2.c              |  18 ++++++++
 include/block/block_int.h  |   1 +
 tests/qemu-iotests/049.out | 108 ++++++++++++++++++++++-----------------------
 tests/qemu-iotests/082.out |  41 ++++++++++++++---
 tests/qemu-iotests/085.out |  38 ++++++++--------
 5 files changed, 126 insertions(+), 80 deletions(-)
block/qcow2.c +0 -0
index a07f134..4cdbd7e 100644
@@ -2078,6 +2078,15 @@ static int qcow2_create(const char *filename, QemuOpts *opts, Error **errp)
         goto finish;
     }
 

block/qcow2.c static int qcow2_create(const char *filename, QemuOpts *opts, Error **errp) · line 2090

Quoted 10 lines
block/qcow2.c +9 -0
+    refcount_bits = qemu_opt_get_number_del(opts, BLOCK_OPT_REFCOUNT_BITS,
+                                            refcount_bits);
+    if (refcount_bits > 64 || !is_power_of_2(refcount_bits)) {
+        error_setg(errp, "Refcount width must be a power of two and may not "
+                   "exceed 64 bits");
+        ret = -EINVAL;
+        goto finish;
+    }
+
     if (version < 3 && refcount_bits != 16) {

The ffs() I complained about in 11/25 occurs shortly after this point, and is therefore safe even though it is called on a 64-bit number, because we have capped the number to be in range even after truncation to int.

Signature · 2 lines
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org
author
#9 lore raw

PATCH v7 15/25 iotests: Add test for different refcount widths

In reply to Max Reitz

Add a test for errors specific to certain widths (i.e. snapshots with refcount_bits=1).

---
 tests/qemu-iotests/112     | 187 +++++++++++++++++++++++++++++++++++++++++++++
 tests/qemu-iotests/112.out |  84 ++++++++++++++++++++
 tests/qemu-iotests/group   |   1 +
 3 files changed, 272 insertions(+)
 create mode 100755 tests/qemu-iotests/112
 create mode 100644 tests/qemu-iotests/112.out
tests/qemu-iotests/112 new file +187 -0
new file mode 100755
index 0000000..3f054a3
@@ -0,0 +1,187 @@
+#!/bin/bash
+#
+# Test cases for different refcount_bits values
+#
+# Copyright (C) 2015 Red Hat, Inc.
+#
+# This program is free software; you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation; either version 2 of the License, or
+# (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program.  If not, see <http://www.gnu.org/licenses/>.
+#
+
+# creator
[email protected]
+
+seq="$(basename $0)"
+echo "QA output created by $seq"
+
+here="$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 qcow2-specific low-level functionality
+_supported_fmt qcow2
+_supported_proto file
+_supported_os Linux
+# This test will set refcount_bits on its own which would conflict with the
+# manual setting; compat will be overridden as well
+_unsupported_imgopts refcount_bits 'compat=0.10'
+
+function print_refcount_bits()
+{
+    $QEMU_IMG info "$TEST_IMG" | sed -n '/refcount bits:/ s/^ *//p'
+}
+
+echo
+echo '=== refcount_bits limits ==='
+echo
+
+# Must be positive (non-zero)
+IMGOPTS="$IMGOPTS,refcount_bits=0" _make_test_img 64M
+# Must be positive (non-negative)
+IMGOPTS="$IMGOPTS,refcount_bits=-1" _make_test_img 64M
+# May not exceed 64
+IMGOPTS="$IMGOPTS,refcount_bits=128" _make_test_img 64M
+# Must be a power of two
+IMGOPTS="$IMGOPTS,refcount_bits=42" _make_test_img 64M
+
+# 1 is the minimum
+IMGOPTS="$IMGOPTS,refcount_bits=1" _make_test_img 64M
+print_refcount_bits
+
+# 64 is the maximum
+IMGOPTS="$IMGOPTS,refcount_bits=64" _make_test_img 64M
+print_refcount_bits
+
+# 16 is the default
+_make_test_img 64M
+print_refcount_bits
+
+echo
+echo '=== refcount_bits and compat=0.10 ==='
+echo
+
+# Should work
+IMGOPTS="$IMGOPTS,compat=0.10,refcount_bits=16" _make_test_img 64M
+print_refcount_bits
+
+# Should not work
+IMGOPTS="$IMGOPTS,compat=0.10,refcount_bits=1" _make_test_img 64M
+IMGOPTS="$IMGOPTS,compat=0.10,refcount_bits=64" _make_test_img 64M
+
+
+echo
+echo '=== Snapshot limit on refcount_bits=1 ==='
+echo
+
+IMGOPTS="$IMGOPTS,refcount_bits=1" _make_test_img 64M
+print_refcount_bits
+
+$QEMU_IO -c 'write 0 512' "$TEST_IMG" | _filter_qemu_io
+
+# Should fail for now; in the future, this might be supported by automatically
+# copying all clusters with overflowing refcount
+$QEMU_IMG snapshot -c foo "$TEST_IMG"
+
+# The new L1 table could/should be leaked
+_check_test_img
+
+echo
+echo '=== Snapshot limit on refcount_bits=2 ==='
+echo
+
+IMGOPTS="$IMGOPTS,refcount_bits=2" _make_test_img 64M
+print_refcount_bits
+
+$QEMU_IO -c 'write 0 512' "$TEST_IMG" | _filter_qemu_io
+
+# Should succeed
+$QEMU_IMG snapshot -c foo "$TEST_IMG"
+$QEMU_IMG snapshot -c bar "$TEST_IMG"
+# Should fail (4th reference)
+$QEMU_IMG snapshot -c baz "$TEST_IMG"
+
+# The new L1 table could/should be leaked
+_check_test_img
+
+echo
+echo '=== Compressed clusters with refcount_bits=1 ==='
+echo
+
+IMGOPTS="$IMGOPTS,refcount_bits=1" _make_test_img 64M
+print_refcount_bits
+
+# Both should fit into a single host cluster; instead of failing to increase the
+# refcount of that cluster, qemu should just allocate a new cluster and make
+# this operation succeed
+$QEMU_IO -c 'write -P 0 -c  0  64k' \
+         -c 'write -P 1 -c 64k 64k' \
+         "$TEST_IMG" | _filter_qemu_io
+
+_check_test_img
+
+echo
+echo '=== MSb set in 64 bit refcount ==='
+echo
+
+IMGOPTS="$IMGOPTS,refcount_bits=64" _make_test_img 64M
+print_refcount_bits
+
+$QEMU_IO -c 'write 0 512' "$TEST_IMG" | _filter_qemu_io
+
+# Set the MSb in the refblock entry of the data cluster
+poke_file "$TEST_IMG" $((0x20028)) "\x80\x00\x00\x00\x00\x00\x00\x00"
+
+# Clear OFLAG_COPIED in the L2 entry of the data cluster
+poke_file "$TEST_IMG" $((0x40000)) "\x00\x00\x00\x00\x00\x05\x00\x00"
+
+# Try to write to that cluster (should work, even though the MSb is set)
+$QEMU_IO -c 'write 0 512' "$TEST_IMG" | _filter_qemu_io
+
+echo
+echo '=== Snapshot on maximum 64 bit refcount value ==='
+echo
+
+IMGOPTS="$IMGOPTS,refcount_bits=64" _make_test_img 64M
+print_refcount_bits
+
+$QEMU_IO -c 'write 0 512' "$TEST_IMG" | _filter_qemu_io
+
+# Set the refblock entry to the maximum value possible
+poke_file "$TEST_IMG" $((0x20028)) "\xff\xff\xff\xff\xff\xff\xff\xff"
+
+# Clear OFLAG_COPIED in the L2 entry of the data cluster
+poke_file "$TEST_IMG" $((0x40000)) "\x00\x00\x00\x00\x00\x05\x00\x00"
+
+# Try a snapshot (should correctly identify the overflow; may work in the future
+# by falling back to COW)
+$QEMU_IMG snapshot -c foo "$TEST_IMG"
+
+# The new L1 table could/should be leaked; and obviously the data cluster is
+# leaked (refcount=UINT64_MAX reference=1)
+_check_test_img
+
+
+# success, all done
+echo '*** done'
+rm -f $seq.full
+status=0
tests/qemu-iotests/112.out new file +84 -0

84 lines not on this page · raw message

tests/qemu-iotests/group +1 -0
index 7b74c27..8d24dea 100644
@@ -114,6 +114,7 @@
 109 rw auto
 110 rw auto backing quick
 111 rw auto quick
+112 rw auto
 113 rw auto quick
 114 rw auto quick
 116 rw auto quick
Signature · 1 line
2.1.0
Reviewed-by on v7 15/25
#10 lore raw

Re: [Qemu-devel] [PATCH v7 15/25] iotests: Add test for different refcount widths

Quoted 12 lines · Max Reitz

Add a test for errors specific to certain widths (i.e. snapshots with refcount_bits=1).

---
 tests/qemu-iotests/112     | 187 +++++++++++++++++++++++++++++++++++++++++++++
 tests/qemu-iotests/112.out |  84 ++++++++++++++++++++
 tests/qemu-iotests/group   |   1 +
 3 files changed, 272 insertions(+)
 create mode 100755 tests/qemu-iotests/112
 create mode 100644 tests/qemu-iotests/112.out
Signature · 2 lines
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org
author
#11 lore raw

PATCH v7 16/25 progress: Allow regressing progress

In reply to 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.

---
 util/qemu-progress.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
util/qemu-progress.c +2 -1
index 4ee5cd0..532333e 100644
@@ -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();
     }
Signature · 1 line
2.1.0
author
#12 lore raw

PATCH v7 17/25 block: Add opaque value to the amend CB

In reply to Max Reitz

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

---
 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(-)
block.c +2 -2
index 210fd5f..555ad9c 100644
@@ -5887,12 +5887,12 @@ 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
block/qcow2-cluster.c +8 -6
index ed2b44d..56363dd 100644
@@ -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;
         }
block/qcow2.c +5 -4
index 4cdbd7e..3ee1466 100644
@@ -2602,7 +2602,7 @@ 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;
@@ -2651,7 +2651,7 @@ 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;
     }
@@ -2666,7 +2666,8 @@ 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;
@@ -2748,7 +2749,7 @@ 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;
             }
block/qcow2.h +2 -1
index aa6d367..eed26db 100644
@@ -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);
include/block/block.h +2 -2
index 321295e..9a5e122 100644
@@ -278,9 +278,9 @@ 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,
include/block/block_int.h +2 -1
index b661b43..6abbea4 100644
@@ -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);
 
qemu-img.c +3 -2
index e148af8..c9daf9a 100644
@@ -2883,7 +2883,8 @@ 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);
 }
@@ -3003,7 +3004,7 @@ 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));
Signature · 1 line
2.1.0
author
#13 lore raw

PATCH v7 18/25 qcow2: Use error_report() in qcow2_amend_options()

In reply to Max Reitz
---
 block/qcow2.c              | 14 ++++++--------
 tests/qemu-iotests/061.out | 14 +++++++-------
 2 files changed, 13 insertions(+), 15 deletions(-)
block/qcow2.c +6 -8
index 3ee1466..0548248 100644
@@ -2696,11 +2696,11 @@ 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);
@@ -2712,16 +2712,14 @@ 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)) {
@@ -2767,8 +2765,8 @@ 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;
tests/qemu-iotests/061.out +7 -7
index e70f983..2cd9eee 100644
@@ -281,19 +281,19 @@ 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'
 qemu-img: Invalid options for file format 'qcow2'
-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 ===
@@ -301,7 +301,7 @@ 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 ===
Signature · 1 line
2.1.0
author
#14 lore raw

PATCH v7 19/25 qcow2: Use abort() instead of assert(false)

In reply to Max Reitz
---
 block/qcow2.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
block/qcow2.c +2 -2
index 0548248..6283a1c 100644
@@ -2729,9 +2729,9 @@ 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++;
Signature · 1 line
2.1.0
author
#15 lore raw

PATCH v7 20/25 qcow2: Split upgrade/downgrade paths for amend

In reply to 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.

---
 block/qcow2.c | 31 ++++++++++++++++---------------
 1 file changed, 16 insertions(+), 15 deletions(-)
block/qcow2.c +16 -15
index 6283a1c..ef68540 100644
@@ -2737,20 +2737,13 @@ 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;
         }
     }
 
@@ -2764,7 +2757,7 @@ 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;
@@ -2800,6 +2793,14 @@ 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;
 }
 
Signature · 1 line
2.1.0
author
#16 lore raw

PATCH v7 21/25 qcow2: Use intermediate helper CB for amend

In reply to 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.

---
 block/qcow2.c | 80 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 79 insertions(+), 1 deletion(-)
block/qcow2.c +79 -1
index ef68540..8931132 100644
@@ -2665,6 +2665,75 @@ 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)
@@ -2679,6 +2748,7 @@ 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)) {
@@ -2737,6 +2807,12 @@ 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;
@@ -2795,7 +2871,9 @@ 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;
         }
Signature · 1 line
2.1.0
author
#17 lore raw

PATCH v7 22/25 qcow2: Add function for refcount order amendment

In reply to Max Reitz

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

---
 block/qcow2-refcount.c | 452 +++++++++++++++++++++++++++++++++++++++++++++++++
 block/qcow2.h          |   4 +
 2 files changed, 456 insertions(+)
block/qcow2-refcount.c +452 -0
index c9f9f4f..ddacdcc 100644
@@ -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;
+}
block/qcow2.h +4 -0
index eed26db..8f34a5c 100644
@@ -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);
Signature · 1 line
2.1.0
author
#18 lore raw

PATCH v7 23/25 qcow2: Invoke refcount order amendment function

In reply to Max Reitz

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

---
 block/qcow2.c | 44 +++++++++++++++++++++++++++++++++++---------
 1 file changed, 35 insertions(+), 9 deletions(-)
block/qcow2.c +35 -9
index 8931132..42df2b5 100644
@@ -2617,13 +2617,7 @@ 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;
     }
 
@@ -2671,6 +2665,7 @@ typedef enum Qcow2AmendOperation {
      * invocation from an operation change */
     QCOW2_NO_OPERATION = 0,
 
+    QCOW2_CHANGING_REFCOUNT_ORDER,
     QCOW2_DOWNGRADING,
 } Qcow2AmendOperation;
 
@@ -2746,6 +2741,7 @@ 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;
@@ -2796,8 +2792,16 @@ 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 */
@@ -2811,6 +2815,7 @@ 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) */
@@ -2823,6 +2828,27 @@ 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);
Signature · 1 line
2.1.0
author
#19 lore raw

PATCH v7 24/25 qcow2: Point to amend function in check

In reply to 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.

---
 block/qcow2-refcount.c | 3 +++
 1 file changed, 3 insertions(+)
block/qcow2-refcount.c +3 -0
index ddacdcc..0821912 100644
@@ -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;
         }
Signature · 1 line
2.1.0
author
#20 lore raw

PATCH v7 25/25 iotests: Extend test 112 for qemu-img amend

In reply to Max Reitz

Add tests for conversion between different refcount widths.

---
 tests/qemu-iotests/112     | 109 +++++++++++++++++++++++++++++++++++++++++++++
 tests/qemu-iotests/112.out |  71 +++++++++++++++++++++++++++++
 2 files changed, 180 insertions(+)
tests/qemu-iotests/112 +109 -0
index 3f054a3..34ba06a 100755
@@ -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'
tests/qemu-iotests/112.out +71 -0

75 lines not on this page · raw message

Signature · 1 line
2.1.0
Reviewed-by on v7 25/25
#21 lore raw

Re: [Qemu-devel] [PATCH v7 25/25] iotests: Extend test 112 for qemu-img amend

Quoted 8 lines · Max Reitz

Add tests for conversion between different refcount widths.

---
 tests/qemu-iotests/112     | 109 +++++++++++++++++++++++++++++++++++++++++++++
 tests/qemu-iotests/112.out |  71 +++++++++++++++++++++++++++++
 2 files changed, 180 insertions(+)
Signature · 2 lines
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org