[<prev] [next>] [day] [month] [year] [list]
Message-ID: <20250228081824.4640-1-johan+linaro@kernel.org>
Date: Fri, 28 Feb 2025 09:18:24 +0100
From: Johan Hovold <johan+linaro@...nel.org>
To: Greg Kroah-Hartman <gregkh@...uxfoundation.org>
Cc: "Rafael J . Wysocki " <rafael@...nel.org>,
Danilo Krummrich <dakr@...nel.org>,
dri-devel@...ts.freedesktop.org,
linux-kernel@...r.kernel.org,
Johan Hovold <johan+linaro@...nel.org>
Subject: [PATCH] component: do not try to unbind unbound components
Error handling is apparently hard and driver authors often get it wrong.
Continue to warn but do not try to unbind components that have never
been bound in order to avoid crashing systems where such a buggy
teardown path is hit.
Signed-off-by: Johan Hovold <johan+linaro@...nel.org>
---
I've been sitting on this one for too long now after I ran into such an
issue while fixing up the MDM DRM driver which triggered all sorts
problems on probe deferral.
Here's a recent example of where a return here would have a avoided a
double free:
https://lore.kernel.org/lkml/6089322f3d5f2e56f4d7a5899d70da2bc45978f7.1702509741.git.soyer@irl.hu/
Arguably, crashing the system is more noticeable, but not very nice for
users.
Johan
drivers/base/component.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/drivers/base/component.c b/drivers/base/component.c
index 741497324d78..3faa92d26be3 100644
--- a/drivers/base/component.c
+++ b/drivers/base/component.c
@@ -572,7 +572,8 @@ EXPORT_SYMBOL_GPL(component_master_del);
static void component_unbind(struct component *component,
struct aggregate_device *adev, void *data)
{
- WARN_ON(!component->bound);
+ if (WARN_ON(!component->bound))
+ return;
if (component->ops && component->ops->unbind)
component->ops->unbind(component->dev, adev->parent, data);
--
2.45.3
Powered by blists - more mailing lists