In the Linux kernel, the following vulnerability has been resolved:
rpmsg: glink: fix deadlock in endpoint destroy during driver detach
During driver detach, the device core holds the device mutex throughout
the driver’s remove callback chain. When the rpmsg endpoint is
destroyed as part of that teardown, the GLINK endpoint destroy
implementation attempts to unregister the underlying rpmsg device.
That unregistration calls device_del(), which tries to re-acquire the
same device mutex already held higher up the stack, causing rmmod to
hang indefinitely.
The deadlock manifests with the following call chain:
[] device_del+0x44/0x414 <- tries to acquire same mutex
[] device_unregister+0x18/0x34
[] rpmsg_unregister_device+0x28/0x4c
[] qcom_glink_remove_rpmsg_device+0x70/0xc0
[] qcom_glink_destroy_ept+0x58/0xbc
[] rpmsg_dev_remove+0x50/0x60
[] device_remove+0x4c/0x80
[] device_release_driver_internal+0x1cc/0x228 <- acquires device mutex
[] driver_detach+0x4c/0x98
[] bus_remove_driver+0x6c/0xbc
[] driver_unregister+0x30/0x60
[] unregister_rpmsg_driver+0x10/0x1c
[] fastrpc_exit+0x28/0x38 [fastrpc]
[] __arm64_sys_delete_module+0x1b8/0x294
[] invoke_syscall+0x48/0x10c
[] el0_svc_common.constprop.0+0xc0/0xe0
[] do_el0_svc+0x1c/0x28
[] el0_svc+0x34/0x108
[] el0t_64_sync_handler+0xa0/0xe4
[] el0t_64_sync+0x198/0x19c
The rpmsg device unregistration inside endpoint destroy is redundant.
In both contexts where endpoint destruction is triggered:
– Driver detach path: the driver core already tears down the rpmsg
device.
– Channel close path: the rpmsg device is already unregistered before
endpoint destruction is reached.
Remove the redundant unregistration to fix the deadlock.
