On 11/5/26 12:46, Peter Maydell wrote:
> This email originated from an IP that might not be authorized by the domain
> it was sent from.
> Do not click links or open attachments unless it is an email you expected to
> receive.
> In set_drive_helper(), we call visit_type_str() to get the value
> being set as a string. This is an allocated string, and we're
> supposed to free it, but not all paths out of this function do so.
> In particular, the "block backend already exists, change the attached
> node" code path does not free the string. This shows up as a memory
> leak under the newly added virtio-scsi-tests/replaced-node-size
> qtest:
>
> $
> UBSAN_OPTIONS=halt_on_error=1:abort_on_error=1:print_summary=1:print_stacktrace=1
> \
> QTEST_QEMU_BINARY=./qemu-system-i386 \
>
> ASAN_OPTIONS=halt_on_error=1:abort_on_error=1:print_summary=1:print_stacktrace=1
> \
> ./tests/qtest/qos-test --tap -k \
> -p
> /i386/pc/i440FX-pcihost/pci-bus-pc/pci-bus/virtio-scsi-pci/virtio-scsi/virtio-scsi-tests/replaced-node-size
> [...]
> Direct leak of 4 byte(s) in 1 object(s) allocated from:
> #0 0x55f7744155e3 in malloc
> (/home/pm215/qemu/build/x86-tgt-san-dbg/qemu-system-i386+0x20115e3) (BuildId:
> 271ffb42d2e322de7813813d8c6cc85351e48a4f)
> #1 0x713dbd345b79 in g_malloc
> /usr/src/glib2.0-2.80.0-6ubuntu3.10/debian/build/deb/../../../glib/gmem.c:100:13
> #2 0x713dbd35b988 in g_strdup
> /usr/src/glib2.0-2.80.0-6ubuntu3.10/debian/build/deb/../../../glib/gstrfuncs.c:323:17
> #3 0x55f777208021 in g_strdup_inline
> /usr/include/glib-2.0/glib/gstrfuncs.h:321:10
> #4 0x55f777208021 in qobject_input_type_str
> /home/pm215/qemu/build/x86-tgt-san-dbg/../../qapi/qobject-input-visitor.c:542:12
> #5 0x55f777204d1c in visit_type_str
> /home/pm215/qemu/build/x86-tgt-san-dbg/../../qapi/qapi-visit-core.c:349:10
> #6 0x55f774df80d1 in set_drive_helper
> /home/pm215/qemu/build/x86-tgt-san-dbg/../../hw/core/qdev-properties-system.c:129:10
> #7 0x55f774dee9fa in set_drive_iothread
> /home/pm215/qemu/build/x86-tgt-san-dbg/../../hw/core/qdev-properties-system.c:224:5
> #8 0x55f7767cc5c8 in field_prop_set
> /home/pm215/qemu/build/x86-tgt-san-dbg/../../hw/core/qdev-properties.c:89:12
> #9 0x55f7767ec12b in object_property_set
> /home/pm215/qemu/build/x86-tgt-san-dbg/../../qom/object.c:1562:5
>
> Switch to using g_autofree so that the string is properly freed
> all all exit paths.
>
> Cc: [email protected]
> Fixes: d1a58c176a11b ("qdev: allow setting drive property for realized
> device")
> Signed-off-by: Peter Maydell <[email protected]>
> ---
> Sorry for forgetting qemu-devel on first send...
>
> hw/core/qdev-properties-system.c | 5 +----
> 1 file changed, 1 insertion(+), 4 deletions(-)
>
> diff --git a/hw/core/qdev-properties-system.c
> b/hw/core/qdev-properties-system.c
> index 4d605c20561..d5b227dbecb 100644
> --- a/hw/core/qdev-properties-system.c
> +++ b/hw/core/qdev-properties-system.c
> @@ -120,7 +120,7 @@ static void set_drive_helper(Object *obj, Visitor *v,
> const char *name,
> DeviceState *dev = DEVICE(obj);
> const Property *prop = opaque;
> void **ptr = object_field_prop_ptr(obj, prop);
> - char *str;
> + g_autofree char *str = NULL;
> BlockBackend *blk;
> bool blk_created = false;
> int ret;
> @@ -155,7 +155,6 @@ static void set_drive_helper(Object *obj, Visitor *v,
> const char *name,
> }
>
> if (!*str) {
> - g_free(str);
> *ptr = NULL;
> return;
> }
> @@ -209,8 +208,6 @@ fail:
> /* If we need to keep a reference, blk_attach_dev() took it */
> blk_unref(blk);
> }
> -
> - g_free(str);
> }
>
> static void set_drive(Object *obj, Visitor *v, const char *name, void
> *opaque,
Reviewed-by: Denis V. Lunev <[email protected]>