[PATCH] virtio-dmabuf: Ensure UUID persistence for hash table insertion

[PATCH] virtio-dmabuf: Ensure UUID persistence for hash table insertion

[Dorinda Bassey]

In `virtio_add_resource` function, the UUID used as a key for
`g_hash_table_insert` was temporary, which could lead to
invalid lookups when accessed later. This patch ensures that
the UUID remains valid by duplicating it into a newly allocated
memory space (persistent_uuid). The value is then inserted into
the hash table with this persistent UUID key to ensure that the
key stored in the hash table remains valid as long as the hash
table entry exists.

Signed-off-by: Dorinda Bassey <dbassey@redhat.com>
---
 hw/display/virtio-dmabuf.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/hw/display/virtio-dmabuf.c b/hw/display/virtio-dmabuf.c
index 3dba4577ca7..4353970bc87 100644
--- a/hw/display/virtio-dmabuf.c
+++ b/hw/display/virtio-dmabuf.c
@@ -39,7 +39,12 @@ static bool virtio_add_resource(QemuUUID *uuid, VirtioSharedObject *value)
                                                g_free);
     }
     if (g_hash_table_lookup(resource_uuids, uuid) == NULL) {
-        g_hash_table_insert(resource_uuids, uuid, value);
+        QemuUUID *persistent_uuid = g_memdup2(uuid, sizeof(QemuUUID));
+        if (persistent_uuid == NULL) {
+            result = false;
+        } else {
+            g_hash_table_insert(resource_uuids, persistent_uuid, value);
+        }
     } else {
         result = false;
     }
-- 
2.47.0

Re: [PATCH] virtio-dmabuf: Ensure UUID persistence for hash table insertion

[Albert Esteve]

On Thu, Nov 7, 2024 at 1:52 PM Dorinda Bassey <dbassey@redhat.com> wrote:
>
> In `virtio_add_resource` function, the UUID used as a key for
> `g_hash_table_insert` was temporary, which could lead to
> invalid lookups when accessed later. This patch ensures that
> the UUID remains valid by duplicating it into a newly allocated
> memory space (persistent_uuid). The value is then inserted into
> the hash table with this persistent UUID key to ensure that the
> key stored in the hash table remains valid as long as the hash
> table entry exists.
>
> Signed-off-by: Dorinda Bassey <dbassey@redhat.com>
> ---
>  hw/display/virtio-dmabuf.c | 7 ++++++-
>  1 file changed, 6 insertions(+), 1 deletion(-)
>
> diff --git a/hw/display/virtio-dmabuf.c b/hw/display/virtio-dmabuf.c
> index 3dba4577ca7..4353970bc87 100644
> --- a/hw/display/virtio-dmabuf.c
> +++ b/hw/display/virtio-dmabuf.c
> @@ -39,7 +39,12 @@ static bool virtio_add_resource(QemuUUID *uuid, VirtioSharedObject *value)
>                                                 g_free);
>      }
>      if (g_hash_table_lookup(resource_uuids, uuid) == NULL) {
> -        g_hash_table_insert(resource_uuids, uuid, value);
> +        QemuUUID *persistent_uuid = g_memdup2(uuid, sizeof(QemuUUID));
> +        if (persistent_uuid == NULL) {
> +            result = false;
> +        } else {
> +            g_hash_table_insert(resource_uuids, persistent_uuid, value);
> +        }

Reviewed-by: Albert Esteve <aesteve@redhat.com>

The description of `virtio_add_*` functions in the header state:

"""
* Return: true if the UUID did not exist and the resource has been added,
* false if another resource with the same UUID already existed.
"""

I think it may be worth clarifying that it can also return false if
the resource failed
to be added.


>      } else {
>          result = false;
>      }
> --
> 2.47.0
>

Re: [PATCH] virtio-dmabuf: Ensure UUID persistence for hash table insertion

[Stefano Garzarella]

On Thu, Nov 07, 2024 at 01:52:01PM +0100, Dorinda Bassey wrote:
>In `virtio_add_resource` function, the UUID used as a key for
>`g_hash_table_insert` was temporary, which could lead to
>invalid lookups when accessed later. This patch ensures that
>the UUID remains valid by duplicating it into a newly allocated
>memory space (persistent_uuid). The value is then inserted into
>the hash table with this persistent UUID key to ensure that the
>key stored in the hash table remains valid as long as the hash
>table entry exists.

It's a fix right, so maybe better to add a Fixes tag:

Fixes: faefdba847 ("hw/display: introduce virtio-dmabuf")

>
>Signed-off-by: Dorinda Bassey <dbassey@redhat.com>
>---
> hw/display/virtio-dmabuf.c | 7 ++++++-
> 1 file changed, 6 insertions(+), 1 deletion(-)
>
>diff --git a/hw/display/virtio-dmabuf.c b/hw/display/virtio-dmabuf.c
>index 3dba4577ca7..4353970bc87 100644
>--- a/hw/display/virtio-dmabuf.c
>+++ b/hw/display/virtio-dmabuf.c
>@@ -39,7 +39,12 @@ static bool virtio_add_resource(QemuUUID *uuid, VirtioSharedObject *value)
>                                                g_free);
>     }
>     if (g_hash_table_lookup(resource_uuids, uuid) == NULL) {
>-        g_hash_table_insert(resource_uuids, uuid, value);
>+        QemuUUID *persistent_uuid = g_memdup2(uuid, sizeof(QemuUUID));

Since now we allocate memory for the key, we should provide the
`key_destroy_func` when calling g_hash_table_new_full(), otherwise
this new memory will not be de-allocated.

>+        if (persistent_uuid == NULL) {

IIUC it can be null, only if `uuid` was null since glib memory
API usually terminates the application if memory allocation fails,
see https://docs.gtk.org/glib/memory.html

So maybe we can just do:
           g_hash_table_insert(resource_uuids, g_memdup2(uuid, sizeof(QemuUUID)),
                               value);

>+            result = false;
>+        } else {
>+            g_hash_table_insert(resource_uuids, persistent_uuid, value);
>+        }
>     } else {
>         result = false;
>     }
>-- 
>2.47.0
>

Re: [PATCH] virtio-dmabuf: Ensure UUID persistence for hash table insertion

[Stefano Garzarella]

On Thu, Nov 7, 2024 at 2:43 PM Stefano Garzarella <sgarzare@redhat.com> wrote:
>
> On Thu, Nov 07, 2024 at 01:52:01PM +0100, Dorinda Bassey wrote:
> >In `virtio_add_resource` function, the UUID used as a key for
> >`g_hash_table_insert` was temporary, which could lead to
> >invalid lookups when accessed later. This patch ensures that
> >the UUID remains valid by duplicating it into a newly allocated
> >memory space (persistent_uuid). The value is then inserted into
> >the hash table with this persistent UUID key to ensure that the
> >key stored in the hash table remains valid as long as the hash
> >table entry exists.
>
> It's a fix right, so maybe better to add a Fixes tag:
>
> Fixes: faefdba847 ("hw/display: introduce virtio-dmabuf")
>
> >
> >Signed-off-by: Dorinda Bassey <dbassey@redhat.com>
> >---
> > hw/display/virtio-dmabuf.c | 7 ++++++-
> > 1 file changed, 6 insertions(+), 1 deletion(-)
> >
> >diff --git a/hw/display/virtio-dmabuf.c b/hw/display/virtio-dmabuf.c
> >index 3dba4577ca7..4353970bc87 100644
> >--- a/hw/display/virtio-dmabuf.c
> >+++ b/hw/display/virtio-dmabuf.c
> >@@ -39,7 +39,12 @@ static bool virtio_add_resource(QemuUUID *uuid, VirtioSharedObject *value)
> >                                                g_free);
> >     }
> >     if (g_hash_table_lookup(resource_uuids, uuid) == NULL) {
> >-        g_hash_table_insert(resource_uuids, uuid, value);
> >+        QemuUUID *persistent_uuid = g_memdup2(uuid, sizeof(QemuUUID));
>
> Since now we allocate memory for the key, we should provide the
> `key_destroy_func` when calling g_hash_table_new_full(), otherwise
> this new memory will not be de-allocated.
>
> >+        if (persistent_uuid == NULL) {
>
> IIUC it can be null, only if `uuid` was null since glib memory
> API usually terminates the application if memory allocation fails,
> see https://docs.gtk.org/glib/memory.html
>
> So maybe we can just do:
>            g_hash_table_insert(resource_uuids, g_memdup2(uuid, sizeof(QemuUUID)),
>                                value);

Or even better:
             g_hash_table_insert(resource_uuids, g_memdup2(uuid, sizeof(*uuid)),
                                 value);

Thanks,
Stefano

>
> >+            result = false;
> >+        } else {
> >+            g_hash_table_insert(resource_uuids, persistent_uuid, value);
> >+        }
> >     } else {
> >         result = false;
> >     }
> >--
> >2.47.0
> >

Re: [PATCH] virtio-dmabuf: Ensure UUID persistence for hash table insertion

[Dorinda Bassey]

[-- Attachment #1: Type: text/plain, Size: 3570 bytes --]

Hi,

Since now we allocate memory for the key, we should provide the
> `key_destroy_func` when calling g_hash_table_new_full(), otherwise
> this new memory will not be de-allocated.
>

Ack

IIUC it can be null, only if `uuid` was null since glib memory
> API usually terminates the application if memory allocation fails,
> see https://docs.gtk.org/glib/memory.html


Ok, but the documentation doesn't explicitly state what `g_memdup2` returns
on failure, I checked it manually and it terminates the application on
failure. So indeed this check is not necessary.

I think it may be worth clarifying that it can also return false if
> the resource failed
> to be added.


Hence, I think this is not needed since `g_memdup2` will terminate the
application if memory allocation fails and no return value is provided then.

Thanks,
Dorinda.

On Thu, Nov 7, 2024 at 2:47 PM Stefano Garzarella <sgarzare@redhat.com>
wrote:

> On Thu, Nov 7, 2024 at 2:43 PM Stefano Garzarella <sgarzare@redhat.com>
> wrote:
> >
> > On Thu, Nov 07, 2024 at 01:52:01PM +0100, Dorinda Bassey wrote:
> > >In `virtio_add_resource` function, the UUID used as a key for
> > >`g_hash_table_insert` was temporary, which could lead to
> > >invalid lookups when accessed later. This patch ensures that
> > >the UUID remains valid by duplicating it into a newly allocated
> > >memory space (persistent_uuid). The value is then inserted into
> > >the hash table with this persistent UUID key to ensure that the
> > >key stored in the hash table remains valid as long as the hash
> > >table entry exists.
> >
> > It's a fix right, so maybe better to add a Fixes tag:
> >
> > Fixes: faefdba847 ("hw/display: introduce virtio-dmabuf")
> >
> > >
> > >Signed-off-by: Dorinda Bassey <dbassey@redhat.com>
> > >---
> > > hw/display/virtio-dmabuf.c | 7 ++++++-
> > > 1 file changed, 6 insertions(+), 1 deletion(-)
> > >
> > >diff --git a/hw/display/virtio-dmabuf.c b/hw/display/virtio-dmabuf.c
> > >index 3dba4577ca7..4353970bc87 100644
> > >--- a/hw/display/virtio-dmabuf.c
> > >+++ b/hw/display/virtio-dmabuf.c
> > >@@ -39,7 +39,12 @@ static bool virtio_add_resource(QemuUUID *uuid,
> VirtioSharedObject *value)
> > >                                                g_free);
> > >     }
> > >     if (g_hash_table_lookup(resource_uuids, uuid) == NULL) {
> > >-        g_hash_table_insert(resource_uuids, uuid, value);
> > >+        QemuUUID *persistent_uuid = g_memdup2(uuid, sizeof(QemuUUID));
> >
> > Since now we allocate memory for the key, we should provide the
> > `key_destroy_func` when calling g_hash_table_new_full(), otherwise
> > this new memory will not be de-allocated.
> >
> > >+        if (persistent_uuid == NULL) {
> >
> > IIUC it can be null, only if `uuid` was null since glib memory
> > API usually terminates the application if memory allocation fails,
> > see https://docs.gtk.org/glib/memory.html
> >
> > So maybe we can just do:
> >            g_hash_table_insert(resource_uuids, g_memdup2(uuid,
> sizeof(QemuUUID)),
> >                                value);
>
> Or even better:
>              g_hash_table_insert(resource_uuids, g_memdup2(uuid,
> sizeof(*uuid)),
>                                  value);
>
> Thanks,
> Stefano
>
> >
> > >+            result = false;
> > >+        } else {
> > >+            g_hash_table_insert(resource_uuids, persistent_uuid,
> value);
> > >+        }
> > >     } else {
> > >         result = false;
> > >     }
> > >--
> > >2.47.0
> > >
>
>

[-- Attachment #2: Type: text/html, Size: 5255 bytes --]