Move the initialization of parent->mutex into
hda_component_manager_init() so that it is always valid.

In hda_component_manager_bind() do not clear the parent information.
Only zero-fill the per-component data ready for it to be filled in
by the components as they bind.

Previously parent->mutex was being initialized only in
hda_component_manager_bind(). This meant that it was only
initialized if all components appeared and there was a bind callback.
If there wasn't a bind the mutex object was not valid when the
Realtek driver called any of the other functions.

Signed-off-by: Richard Fitzgerald <rf@opensource.cirrus.com>
Fixes: 047b9cbbaa8e ("ALSA: hda: hda_component: Protect shared data with a mutex")
---
 sound/pci/hda/hda_component.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/sound/pci/hda/hda_component.c b/sound/pci/hda/hda_component.c
index 7b19cb38b4e0..b7dfdb10d156 100644
--- a/sound/pci/hda/hda_component.c
+++ b/sound/pci/hda/hda_component.c
@@ -141,8 +141,7 @@ int hda_component_manager_bind(struct hda_codec *cdc,
 	int ret;
 
 	/* Init shared and component specific data */
-	memset(parent, 0, sizeof(*parent));
-	mutex_init(&parent->mutex);
+	memset(parent->comps, 0, sizeof(parent->comps));
 	parent->codec = cdc;
 
 	mutex_lock(&parent->mutex);
@@ -164,6 +163,8 @@ int hda_component_manager_init(struct hda_codec *cdc,
 	struct hda_scodec_match *sm;
 	int ret, i;
 
+	mutex_init(&parent->mutex);
+
 	for (i = 0; i < count; i++) {
 		sm = devm_kmalloc(dev, sizeof(*sm), GFP_KERNEL);
 		if (!sm)
-- 
2.39.2

On Mon, 26 Aug 2024 11:49:40 +0200,
Richard Fitzgerald wrote:
> 
> Move the initialization of parent->mutex into
> hda_component_manager_init() so that it is always valid.
> 
> In hda_component_manager_bind() do not clear the parent information.
> Only zero-fill the per-component data ready for it to be filled in
> by the components as they bind.
> 
> Previously parent->mutex was being initialized only in
> hda_component_manager_bind(). This meant that it was only
> initialized if all components appeared and there was a bind callback.
> If there wasn't a bind the mutex object was not valid when the
> Realtek driver called any of the other functions.
> 
> Signed-off-by: Richard Fitzgerald <rf@opensource.cirrus.com>
> Fixes: 047b9cbbaa8e ("ALSA: hda: hda_component: Protect shared data with a mutex")

Thanks, applied now to for-linus branch.


Takashi