]> git.ipfire.org Git - thirdparty/kernel/linux.git/commit
KVM: riscv: Rely on common MMU notifier locking
authorJinyu Tang <tjytimi@163.com>
Sun, 17 May 2026 15:34:23 +0000 (23:34 +0800)
committerAnup Patel <anup@brainfault.org>
Wed, 3 Jun 2026 12:49:53 +0000 (18:19 +0530)
commit9090ba2e7cf8bf8a54879182db5665452d515bb0
treeee5806e9331fa414a9e96ab8db3e8997e88c6400
parentc70faabb6ddd06327ae778bf4f9b5b977749b315
KVM: riscv: Rely on common MMU notifier locking

The common KVM invalidation paths call kvm_unmap_gfn_range() with
mmu_lock already held for write.

For the standard MMU notifier path, the call chain is:

  kvm_mmu_notifier_invalidate_range_start()
    kvm_handle_hva_range()
      kvm_unmap_gfn_range()

kvm_mmu_notifier_invalidate_range_start() leaves range.lockless clear.
kvm_handle_hva_range() therefore takes KVM_MMU_LOCK(kvm) before invoking
the handler.

The guest_memfd path has the same locking contract:

  __kvm_gmem_invalidate_begin()
    kvm_mmu_unmap_gfn_range()
      kvm_unmap_gfn_range()

__kvm_gmem_invalidate_begin() explicitly takes KVM_MMU_LOCK(kvm) before
calling kvm_mmu_unmap_gfn_range().

So remove the local trylock and make the common locking contract explicit
with lockdep_assert_held_write() like x86.

Signed-off-by: Jinyu Tang <tjytimi@163.com>
Reviewed-by: Anup Patel <anup@brainfault.org>
Link: https://lore.kernel.org/r/20260517153427.94889-2-tjytimi@163.com
Signed-off-by: Anup Patel <anup@brainfault.org>
arch/riscv/kvm/mmu.c