powerpc/mm/thp: Make page table walk safe against thp split/collapse

We can disable a THP split or a hugepage collapse by disabling irq.
We do send IPI to all the cpus in the early part of split/collapse,
and disabling local irq ensure we don't make progress with
split/collapse. If the THP is getting split we return NULL from
find_linux_pte_or_hugepte(). For all the current callers it should be ok.
We need to be careful if we want to use returned pte_t pointer outside
the irq disabled region. W.r.t to THP split, the pfn remains the same,
but then a hugepage collapse will result in a pfn change. There are
few steps we can take to avoid a hugepage collapse.One way is to take page
reference inside the irq disable region. Other option is to take
mmap_sem so that a parallel collapse will not happen. We can also
disable collapse by taking pmd_lock. Another method used by kvm
subsystem is to check whether we had a mmu_notifer update in between
using mmu_notifier_retry().

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

3 files changed, 38 insertions, 11 deletions

diff --git a/arch/powerpc/kvm/book3s_64_mmu_hv.c b/arch/powerpc/kvm/book3s_64_mmu_hv.c
index 534acb3..26df386 100644
--- a/arch/powerpc/kvm/book3s_64_mmu_hv.c
+++ b/arch/powerpc/kvm/book3s_64_mmu_hv.c
@@ -539,12 +539,13 @@ int kvmppc_book3s_hv_page_fault(struct kvm_run *run, struct kvm_vcpu *vcpu,
 		if (!writing && hpte_is_writable(r)) {
 			unsigned int hugepage_shift;
 			pte_t *ptep, pte;
+			unsigned long flags;
 
 			/*
 			 * We need to protect against page table destruction
 			 * while looking up and updating the pte.
 			 */
-			rcu_read_lock_sched();
+			local_irq_save(flags);
 			ptep = find_linux_pte_or_hugepte(current->mm->pgd,
 							 hva, &hugepage_shift);
 			if (ptep) {
@@ -553,7 +554,7 @@ int kvmppc_book3s_hv_page_fault(struct kvm_run *run, struct kvm_vcpu *vcpu,
 				if (pte_write(pte))
 					write_ok = 1;
 			}
-			rcu_read_unlock_sched();
+			local_irq_restore(flags);
 		}
 	}
 
diff --git a/arch/powerpc/kvm/book3s_hv_rm_mmu.c b/arch/powerpc/kvm/book3s_hv_rm_mmu.c
index 73e083c..f559b25 100644
--- a/arch/powerpc/kvm/book3s_hv_rm_mmu.c
+++ b/arch/powerpc/kvm/book3s_hv_rm_mmu.c
@@ -26,11 +26,14 @@ static void *real_vmalloc_addr(void *x)
 {
 	unsigned long addr = (unsigned long) x;
 	pte_t *p;
-
-	p = find_linux_pte_or_hugepte(swapper_pg_dir, addr, NULL);
+	/*
+	 * assume we don't have huge pages in vmalloc space...
+	 * So don't worry about THP collapse/split. Called
+	 * Only in realmode, hence won't need irq_save/restore.
+	 */
+	p = __find_linux_pte_or_hugepte(swapper_pg_dir, addr, NULL);
 	if (!p || !pte_present(*p))
 		return NULL;
-	/* assume we don't have huge pages in vmalloc space... */
 	addr = (pte_pfn(*p) << PAGE_SHIFT) | (addr & ~PAGE_MASK);
 	return __va(addr);
 }
@@ -153,7 +156,7 @@ long kvmppc_do_h_enter(struct kvm *kvm, unsigned long flags,
 	pte_t *ptep;
 	unsigned int writing;
 	unsigned long mmu_seq;
-	unsigned long rcbits;
+	unsigned long rcbits, irq_flags = 0;
 
 	psize = hpte_page_size(pteh, ptel);
 	if (!psize)
@@ -189,7 +192,16 @@ long kvmppc_do_h_enter(struct kvm *kvm, unsigned long flags,
 
 	/* Translate to host virtual address */
 	hva = __gfn_to_hva_memslot(memslot, gfn);
-	ptep = find_linux_pte_or_hugepte(pgdir, hva, &hpage_shift);
+	/*
+	 * If we had a page table table change after lookup, we would
+	 * retry via mmu_notifier_retry.
+	 */
+	if (realmode)
+		ptep = __find_linux_pte_or_hugepte(pgdir, hva, &hpage_shift);
+	else {
+		local_irq_save(irq_flags);
+		ptep = find_linux_pte_or_hugepte(pgdir, hva, &hpage_shift);
+	}
 	if (ptep) {
 		pte_t pte;
 		unsigned int host_pte_size;
@@ -202,9 +214,11 @@ long kvmppc_do_h_enter(struct kvm *kvm, unsigned long flags,
 		 * We should always find the guest page size
 		 * to <= host page size, if host is using hugepage
 		 */
-		if (host_pte_size < psize)
+		if (host_pte_size < psize) {
+			if (!realmode)
+				local_irq_restore(flags);
 			return H_PARAMETER;
-
+		}
 		pte = kvmppc_read_update_linux_pte(ptep, writing, hpage_shift);
 		if (pte_present(pte) && !pte_protnone(pte)) {
 			if (writing && !pte_write(pte))
@@ -216,6 +230,8 @@ long kvmppc_do_h_enter(struct kvm *kvm, unsigned long flags,
 			pa |= gpa & ~PAGE_MASK;
 		}
 	}
+	if (!realmode)
+		local_irq_restore(irq_flags);
 
 	ptel &= ~(HPTE_R_PP0 - psize);
 	ptel |= pa;
diff --git a/arch/powerpc/kvm/e500_mmu_host.c b/arch/powerpc/kvm/e500_mmu_host.c
index a1f5b0d4..4d33e19 100644
--- a/arch/powerpc/kvm/e500_mmu_host.c
+++ b/arch/powerpc/kvm/e500_mmu_host.c
@@ -338,6 +338,7 @@ static inline int kvmppc_e500_shadow_map(struct kvmppc_vcpu_e500 *vcpu_e500,
 	pte_t *ptep;
 	unsigned int wimg = 0;
 	pgd_t *pgdir;
+	unsigned long flags;
 
 	/* used to check for invalidations in progress */
 	mmu_seq = kvm->mmu_notifier_seq;
@@ -468,14 +469,23 @@ static inline int kvmppc_e500_shadow_map(struct kvmppc_vcpu_e500 *vcpu_e500,
 
 
 	pgdir = vcpu_e500->vcpu.arch.pgdir;
+	/*
+	 * We are just looking at the wimg bits, so we don't
+	 * care much about the trans splitting bit.
+	 * We are holding kvm->mmu_lock so a notifier invalidate
+	 * can't run hence pfn won't change.
+	 */
+	local_irq_save(flags);
 	ptep = find_linux_pte_or_hugepte(pgdir, hva, NULL);
 	if (ptep) {
 		pte_t pte = READ_ONCE(*ptep);
 
-		if (pte_present(pte))
+		if (pte_present(pte)) {
 			wimg = (pte_val(pte) >> PTE_WIMGE_SHIFT) &
 				MAS2_WIMGE_MASK;
-		else {
+			local_irq_restore(flags);
+		} else {
+			local_irq_restore(flags);
 			pr_err_ratelimited("%s: pte not present: gfn %lx,pfn %lx\n",
 					   __func__, (long)gfn, pfn);
 			ret = -EINVAL;