The vulnerability in the Linux kernel arm64 TLB handling occurred because when huge_pmd_unshare() unshared a PMD table, the function tlb_unshare_pmd_ptdesc() set tlb->unshared_tables=true, but the aarch64 tlb_flush() function only checked tlb->freed_tables to decide the type of TLB invalidation instruction to use. This caused the stale PMD page table entries to remain cached in the walk cache, potentially leading to incorrect page table walks. The fix involved including the unshared_tables flag in the check so that a full invalidation (TLBF_NONE) is used, properly clearing the walk cache during unsharing.

The stale PMD page table entries remaining in the walk cache could cause incorrect page table walks, which may lead to memory management inconsistencies on affected arm64 Linux kernel systems. The exact impact is limited to the kernel's memory subsystem behavior and does not specify direct exploitation or system compromise. No known exploits in the wild have been reported.

A fix has been implemented in the Linux kernel to address this issue by modifying the TLB flush logic to properly invalidate the walk cache when unsharing PMD tables. Since no specific patch links or vendor advisories are provided, users should apply the latest Linux kernel updates that include this fix. Patch status is not yet confirmed from vendor advisories; check the official Linux kernel release notes or security advisories for the relevant fixed versions and apply updates accordingly.