CVE-2023-53501

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> iommu/amd/iommu_v2: Fix pasid_state refcount dec hit 0 warning on pasid unbind<br /> <br /> When unbinding pasid - a race condition exists vs outstanding page faults.<br /> <br /> To prevent this, the pasid_state object contains a refcount.<br /> * set to 1 on pasid bind<br /> * incremented on each ppr notification start<br /> * decremented on each ppr notification done<br /> * decremented on pasid unbind<br /> <br /> Since refcount_dec assumes that refcount will never reach 0:<br /> the current implementation causes the following to be invoked on<br /> pasid unbind:<br /> REFCOUNT_WARN("decrement hit 0; leaking memory")<br /> <br /> Fix this issue by changing refcount_dec to refcount_dec_and_test<br /> to explicitly handle refcount=1.