[PATCH for-11.1 0/3] dmg: add missing input validation (CVE-2026-65929 & CVE-2026-65928)

[PATCH for-11.1 0/3] dmg: add missing input validation (CVE-2026-65929 & CVE-2026-65928)

[Stefan Hajnoczi]

The read-only dmg block driver can crash or leak information when it encounters
a malicious DMG file. The severity of these issues is low because the block
driver is used in qemu-img convert workflows where there is no sensitive
information to leak and crashing is a self-DoS.

See the individual patches for descriptions of the bugs.

Stefan Hajnoczi (3):
  dmg: fix out-of-bounds load in search_chunk() (CVE-2026-65929)
  dmg: refuse to open files with no chunks
  dmg: reject inconsistent UDRW chunk sector count and length
    (CVE-2026-65928)

 block/dmg.c | 36 +++++++++++++++++++++++++++++++++++-
 1 file changed, 35 insertions(+), 1 deletion(-)

-- 
2.55.0

[PATCH for-11.1 1/3] dmg: fix out-of-bounds load in search_chunk() (CVE-2026-65929)

[Stefan Hajnoczi]

The binary search in search_chunk() uses s->n_chunks as the (inclusive)
upper bound. Chunk indices are in the right-open interval [0,
s->n_chunks) so it is wrong to search all the way up to s->n_chunks
rather than s->n_chunks - 1.

The worst case security scenario I can see is convincing a victim to
hotplug a malicious DMG file to a running guest, potentially causing
QEMU to crash when loading from memory beyond the end of s->sectors[] or
s->sectorscounts[]. This could be a denial of service.

Fixes: CVE-2026-65929
Resolves: https://gitlab.com/qemu-project/qemu/-/work_items/3844
Reported-by: boy juju <agx1657748706@gmail.com>
Reported-by: Tristan Madani <tristan@talencesecurity.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
---
 block/dmg.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/block/dmg.c b/block/dmg.c
index 33dcb3a3498..e325127d144 100644
--- a/block/dmg.c
+++ b/block/dmg.c
@@ -609,7 +609,10 @@ static inline int is_sector_in_chunk(BDRVDMGState *s,
 static inline uint32_t search_chunk(BDRVDMGState *s, uint64_t sector_num)
 {
     /* binary search */
-    uint32_t chunk1 = 0, chunk2 = s->n_chunks, chunk3;
+    uint32_t chunk1 = 0, chunk2 = s->n_chunks - 1, chunk3;
+    if (s->n_chunks == 0) {
+        goto err; /* should never happen */
+    }
     while (chunk1 <= chunk2) {
         chunk3 = (chunk1 + chunk2) / 2;
         if (s->sectors[chunk3] > sector_num) {
-- 
2.55.0

Re: [PATCH for-11.1 1/3] dmg: fix out-of-bounds load in search_chunk() (CVE-2026-65929)

[Philippe Mathieu-Daudé]

On 23/7/26 16:45, Stefan Hajnoczi wrote:
> The binary search in search_chunk() uses s->n_chunks as the (inclusive)
> upper bound. Chunk indices are in the right-open interval [0,
> s->n_chunks) so it is wrong to search all the way up to s->n_chunks
> rather than s->n_chunks - 1.
> 
> The worst case security scenario I can see is convincing a victim to
> hotplug a malicious DMG file to a running guest, potentially causing
> QEMU to crash when loading from memory beyond the end of s->sectors[] or
> s->sectorscounts[]. This could be a denial of service.
> 
> Fixes: CVE-2026-65929
> Resolves: https://gitlab.com/qemu-project/qemu/-/work_items/3844
> Reported-by: boy juju <agx1657748706@gmail.com>
> Reported-by: Tristan Madani <tristan@talencesecurity.com>
> Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
> ---
>   block/dmg.c | 5 ++++-
>   1 file changed, 4 insertions(+), 1 deletion(-)
> 
> diff --git a/block/dmg.c b/block/dmg.c
> index 33dcb3a3498..e325127d144 100644
> --- a/block/dmg.c
> +++ b/block/dmg.c
> @@ -609,7 +609,10 @@ static inline int is_sector_in_chunk(BDRVDMGState *s,
>   static inline uint32_t search_chunk(BDRVDMGState *s, uint64_t sector_num)
>   {
>       /* binary search */
> -    uint32_t chunk1 = 0, chunk2 = s->n_chunks, chunk3;
> +    uint32_t chunk1 = 0, chunk2 = s->n_chunks - 1, chunk3;
> +    if (s->n_chunks == 0) {
> +        goto err; /* should never happen */
> +    }
>       while (chunk1 <= chunk2) {
>           chunk3 = (chunk1 + chunk2) / 2;
>           if (s->sectors[chunk3] > sector_num) {

Reviewed-by: Philippe Mathieu-Daudé <philmd@oss.qualcomm.com>

Alternatively avoiding the assignations:

-- >8 --
diff --git a/block/dmg.c b/block/dmg.c
index 33dcb3a3498..85edfae145b 100644
--- a/block/dmg.c
+++ b/block/dmg.c
@@ -609,8 +609,12 @@ static inline int is_sector_in_chunk(BDRVDMGState *s,
  static inline uint32_t search_chunk(BDRVDMGState *s, uint64_t sector_num)
  {
      /* binary search */
-    uint32_t chunk1 = 0, chunk2 = s->n_chunks, chunk3;
-    while (chunk1 <= chunk2) {
+    uint32_t chunk1, chunk2, chunk3;
+
+    if (s->n_chunks == 0) {
+        goto err; /* should never happen */
+    }
+    for (chunk1 = 0, chunk2 = s->n_chunks - 1; chunk1 <= chunk2; ) {
          chunk3 = (chunk1 + chunk2) / 2;
          if (s->sectors[chunk3] > sector_num) {
              if (chunk3 == 0) {
---

Re: [PATCH for-11.1 1/3] dmg: fix out-of-bounds load in search_chunk() (CVE-2026-65929)

[Kevin Wolf]

Am 23.07.2026 um 16:45 hat Stefan Hajnoczi geschrieben:
> The binary search in search_chunk() uses s->n_chunks as the (inclusive)
> upper bound. Chunk indices are in the right-open interval [0,
> s->n_chunks) so it is wrong to search all the way up to s->n_chunks
> rather than s->n_chunks - 1.
> 
> The worst case security scenario I can see is convincing a victim to
> hotplug a malicious DMG file to a running guest, potentially causing
> QEMU to crash when loading from memory beyond the end of s->sectors[] or
> s->sectorscounts[]. This could be a denial of service.
> 
> Fixes: CVE-2026-65929
> Resolves: https://gitlab.com/qemu-project/qemu/-/work_items/3844
> Reported-by: boy juju <agx1657748706@gmail.com>
> Reported-by: Tristan Madani <tristan@talencesecurity.com>
> Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
> ---
>  block/dmg.c | 5 ++++-
>  1 file changed, 4 insertions(+), 1 deletion(-)
> 
> diff --git a/block/dmg.c b/block/dmg.c
> index 33dcb3a3498..e325127d144 100644
> --- a/block/dmg.c
> +++ b/block/dmg.c
> @@ -609,7 +609,10 @@ static inline int is_sector_in_chunk(BDRVDMGState *s,
>  static inline uint32_t search_chunk(BDRVDMGState *s, uint64_t sector_num)
>  {
>      /* binary search */
> -    uint32_t chunk1 = 0, chunk2 = s->n_chunks, chunk3;
> +    uint32_t chunk1 = 0, chunk2 = s->n_chunks - 1, chunk3;
> +    if (s->n_chunks == 0) {
> +        goto err; /* should never happen */
> +    }
>      while (chunk1 <= chunk2) {
>          chunk3 = (chunk1 + chunk2) / 2;
>          if (s->sectors[chunk3] > sector_num) {

Fair enough to keep the fix minimal here, but is there any justification
for manually implementing binary search instead of using bsearch()? If
not, this could be a follow-up cleanup.

Kevin

[PATCH for-11.1 2/3] dmg: refuse to open files with no chunks

[Stefan Hajnoczi]

The dmg block driver expects the disk image file to contain at least one
chunk. Refuse to open such files. This ensures that dmg block driver
state always has non-NULL s->sectors[] and related fields.

Note that the previous commit fixed the only known way to trigger a
crash. This patch is just for defense - let's avoid opening the file and
having NULL pointers in dmg block driver state.

Resolves: https://gitlab.com/qemu-project/qemu/-/work_items/4021
Reported-by: Tristan Madani <tristan@talencesecurity.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
---
 block/dmg.c | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/block/dmg.c b/block/dmg.c
index e325127d144..6f8120e0338 100644
--- a/block/dmg.c
+++ b/block/dmg.c
@@ -559,6 +559,12 @@ static int dmg_open(BlockDriverState *bs, QDict *options, int flags,
         goto fail;
     }
 
+    /* There must be at least one chunk */
+    if (s->n_chunks == 0) {
+        ret = -EINVAL;
+        goto fail;
+    }
+
     /* initialize zlib engine */
     s->compressed_chunk = qemu_try_blockalign(bs->file->bs,
                                               ds.max_compressed_size + 1);
-- 
2.55.0

Re: [PATCH for-11.1 2/3] dmg: refuse to open files with no chunks

[Philippe Mathieu-Daudé]

On 23/7/26 16:45, Stefan Hajnoczi wrote:
> The dmg block driver expects the disk image file to contain at least one
> chunk. Refuse to open such files. This ensures that dmg block driver
> state always has non-NULL s->sectors[] and related fields.
> 
> Note that the previous commit fixed the only known way to trigger a
> crash. This patch is just for defense - let's avoid opening the file and
> having NULL pointers in dmg block driver state.
> 
> Resolves: https://gitlab.com/qemu-project/qemu/-/work_items/4021
> Reported-by: Tristan Madani <tristan@talencesecurity.com>
> Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
> ---
>   block/dmg.c | 6 ++++++
>   1 file changed, 6 insertions(+)
> 
> diff --git a/block/dmg.c b/block/dmg.c
> index e325127d144..6f8120e0338 100644
> --- a/block/dmg.c
> +++ b/block/dmg.c
> @@ -559,6 +559,12 @@ static int dmg_open(BlockDriverState *bs, QDict *options, int flags,
>           goto fail;
>       }
>   
> +    /* There must be at least one chunk */

dmg_read_mish_block indirectly updates s->n_chunks via the
dmg_read_resource_fork / dmg_read_plist_xml calls. Both
access BDRVDMGState @s via bs->opaque. OK then.

Reviewed-by: Philippe Mathieu-Daudé <philmd@oss.qualcomm.com>

> +    if (s->n_chunks == 0) {
> +        ret = -EINVAL;
> +        goto fail;
> +    }
> +
>       /* initialize zlib engine */
>       s->compressed_chunk = qemu_try_blockalign(bs->file->bs,
>                                                 ds.max_compressed_size + 1);

[PATCH for-11.1 3/3] dmg: reject inconsistent UDRW chunk sector count and length (CVE-2026-65928)

[Stefan Hajnoczi]

The chunk metadata contains both:
- Sector count: number of 512-byte sectors in the virtual disk
- Length: number of bytes in the image file

The UDRW chunk type indicates uncompressed data that can be accessed
directly. The code is missing input validation to verify that sector
count is consistent with length.

If sector count is larger than length, then read requests can access
beyond the end of the s->uncompressed_chunk buffer. This is an
out-of-bounds heap access that could lead to a crash or an information
leak.

While we're at it, also zero the end of the last sector when length is
unaligned. This prevents information leaks from the
s->uncompressed_chunk buffer.

Fixes: CVE-2026-65928
Resolves: https://gitlab.com/qemu-project/qemu/-/work_items/3846
Reported-by: boy juju <agx1657748706@gmail.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
---
 block/dmg.c | 25 +++++++++++++++++++++++++
 1 file changed, 25 insertions(+)

diff --git a/block/dmg.c b/block/dmg.c
index 6f8120e0338..5d7d3b8901f 100644
--- a/block/dmg.c
+++ b/block/dmg.c
@@ -312,6 +312,21 @@ static int dmg_read_mish_block(BDRVDMGState *s, DmgHeaderState *ds,
             goto fail;
         }
 
+        /*
+         * Uncompressed chunk length must match sector count. Compressed chunks
+         * are validated during dmg_read_chunk() since the uncompressed size is
+         * not known ahead of time.
+         */
+        if (s->types[i] == UDRW) {
+            if (s->sectorcounts[i] != DIV_ROUND_UP(s->lengths[i], 512)) {
+                error_report("length %" PRIu64 " for chunk %" PRIu32
+                             " is inconsistent with sector count %" PRIu64,
+                             s->lengths[i], i, s->sectorcounts[i]);
+                ret = -EINVAL;
+                goto fail;
+            }
+        }
+
         update_max_chunk_size(s, i, &ds->max_compressed_size,
                               &ds->max_sectors_per_chunk);
         offset += 40;
@@ -722,6 +737,16 @@ dmg_read_chunk(BlockDriverState *bs, uint64_t sector_num)
             if (ret < 0) {
                 return -1;
             }
+
+            /*
+             * Zero the unread part of the last sector when chunk length is
+             * unaligned to avoid exposing uninitialized memory. Valid image
+             * files may never hit this case, but cover it to be safe.
+             */
+            if (s->lengths[chunk] & 511) {
+                size_t trailing_bytes = 512 - (s->lengths[chunk] & 511);
+                memset(s->uncompressed_chunk + s->lengths[chunk], 0, trailing_bytes);
+            }
             break;
         case UDZE: /* zeros */
         case UDIG: /* ignore */
-- 
2.55.0

Re: [PATCH for-11.1 0/3] dmg: add missing input validation (CVE-2026-65929 & CVE-2026-65928)

[Kevin Wolf]

Am 23.07.2026 um 16:45 hat Stefan Hajnoczi geschrieben:
> The read-only dmg block driver can crash or leak information when it encounters
> a malicious DMG file. The severity of these issues is low because the block
> driver is used in qemu-img convert workflows where there is no sensitive
> information to leak and crashing is a self-DoS.
> 
> See the individual patches for descriptions of the bugs.

Thanks, applied to the block branch.

Kevin