On 04/06/18 14:44, Rob Herring wrote:
On Fri, Jun 1, 2018 at 7:03 PM, Srinivas Kandagatla
<[email protected]> wrote:
Immediately after the platform_device_unregister() the device will be cleaned 
up.
Accessing the freed pointer immediately after that will crash the system.

Found this bug when kernel is built with CONFIG_PAGE_POISONING and testing
loading/unloading audio drivers in a loop on Qcom platforms.

Curious, does the unittest not catch this too?

Not sure!

Fix this by removing accessing the dev pointer.
Below is the carsh trace:

s/carsh/crash/
Yep.

[...]

diff --git a/drivers/of/platform.c b/drivers/of/platform.c
index c00d81dfac0b..84c5c899187b 100644
--- a/drivers/of/platform.c
+++ b/drivers/of/platform.c
@@ -529,10 +529,13 @@ arch_initcall_sync(of_platform_default_populate_init);

  int of_platform_device_destroy(struct device *dev, void *data)
  {
+       struct device_node *np;
+
         /* Do not touch devices not populated from the device tree */
         if (!dev->of_node || !of_node_check_flag(dev->of_node, OF_POPULATED))
                 return 0;

+       np = dev->of_node;
         /* Recurse for any nodes that were treated as busses */
         if (of_node_check_flag(dev->of_node, OF_POPULATED_BUS))
                 device_for_each_child(dev, NULL, of_platform_device_destroy);
@@ -544,8 +547,8 @@ int of_platform_device_destroy(struct device *dev, void 
*data)
                 amba_device_unregister(to_amba_device(dev));
  #endif

-       of_node_clear_flag(dev->of_node, OF_POPULATED);
-       of_node_clear_flag(dev->of_node, OF_POPULATED_BUS);

Just move these 2 lines to before unregister calls.
Make sense.. I will do that in v3.

thanks,
srini

+       of_node_clear_flag(np, OF_POPULATED);
+       of_node_clear_flag(np, OF_POPULATED_BUS);
         return 0;
  }
  EXPORT_SYMBOL_GPL(of_platform_device_destroy);

--
2.16.2

Reply via email to