CVE-2025-21951

In the Linux kernel, the following vulnerability has been resolved:

bus: mhi: host: pci_generic: Use pci_try_reset_function() to avoid deadlock

There are multiple places from where the recovery work gets scheduled asynchronously. Also, there are multiple places where the caller waits synchronously for the recovery to be completed. One such place is during the PM shutdown() callback.

If the device is not alive during recovery_work, it will try to reset the device using pci_reset_function(). This function internally will take the device_lock() first before resetting the device. By this time, if the lock has already been acquired, then recovery_work will get stalled while waiting for the lock. And if the lock was already acquired by the caller which waits for the recovery_work to be completed, it will lead to deadlock.

This is what happened on the X1E80100 CRD device when the device died before shutdown() callback. Driver core calls the drivers shutdown() callback while holding the device_lock() leading to deadlock.

And this deadlock scenario can occur on other paths as well, like during the PM suspend() callback, where the driver core would hold the device_lock() before calling drivers suspend() callback. And if the recovery_work was already started, it could lead to deadlock. This is also observed on the X1E80100 CRD.

So to fix both issues, use pci_try_reset_function() in recovery_work. This function first checks for the availability of the device_lock() before trying to reset the device. If the lock is available, it will acquire it and reset the device. Otherwise, it will return -EAGAIN. If that happens, recovery_work will fail with the error message Recovery failed as not much could be done.

Weakness

The product does not properly acquire or release a lock on a resource, leading to unexpected resource state changes and behaviors.

Affected Software

Extended Description

Locking is a type of synchronization behavior that ensures that multiple independently-operating processes or threads do not interfere with each other when accessing the same resource. All processes/threads are expected to follow the same steps for locking. If these steps are not followed precisely - or if no locking is done at all - then another process/thread could modify the shared resource in a way that is not visible or predictable to the original process. This can lead to data or memory corruption, denial of service, etc.

Potential Mitigations

Related Attack Patterns

• https://cwe.mitre.org/data/definitions/25.html
• https://cwe.mitre.org/data/definitions/26.html
• https://cwe.mitre.org/data/definitions/27.html

References

• https://git.kernel.org/stable/c/1f9eb7078bc6b5fb5cbfbcb37c4bc01685332b95
• https://git.kernel.org/stable/c/62505657475c245c9cd46e42ac01026d1e61f027
• https://git.kernel.org/stable/c/7746f3bb8917fccb4571a576f3837d80fc513054
• https://git.kernel.org/stable/c/7a5ffadd54fe2662f5c99cdccf30144d060376f7
• https://git.kernel.org/stable/c/985d3cf56d8745ca637deee273929e01df449f85
• https://git.kernel.org/stable/c/a321d163de3d8aa38a6449ab2becf4b1581aed96