Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
From:   Peter Xu <peterx@redhat.com>
To:     linux-mm@kvack.org, linux-kernel@vger.kernel.org
Cc:     James Houghton <jthoughton@google.com>,
        Mike Kravetz <mike.kravetz@oracle.com>,
        David Hildenbrand <david@redhat.com>,
        Andrea Arcangeli <aarcange@redhat.com>,
        Rik van Riel <riel@surriel.com>, peterx@redhat.com,
        Andrew Morton <akpm@linux-foundation.org>,
        Muchun Song <songmuchun@bytedance.com>,
        Miaohe Lin <linmiaohe@huawei.com>,
        Nadav Amit <nadav.amit@gmail.com>
Subject: [PATCH RFC 10/10] mm/hugetlb: Comment at rest huge_pte_offset() places
Date:   Sun, 30 Oct 2022 17:30:45 -0400
Message-Id: <20221030213045.335680-1-peterx@redhat.com>
In-Reply-To: <20221030212929.335473-1-peterx@redhat.com>
References: <20221030212929.335473-1-peterx@redhat.com>
MIME-Version: 1.0
Content-type: text/plain
Content-Transfer-Encoding: 8bit
Precedence: bulk

This makes sure that we're covering all the existing huge_pte_offset()
callers and mention why they are safe regarding to pmd unsharing.

Signed-off-by: Peter Xu <peterx@redhat.com>
---
 mm/hugetlb.c | 14 ++++++++++++++
 1 file changed, 14 insertions(+)

diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index 6d336d286394..270bfc578115 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -4822,6 +4822,7 @@ int copy_hugetlb_page_range(struct mm_struct *dst, struct mm_struct *src,
 	last_addr_mask = hugetlb_mask_last_page(h);
 	for (addr = src_vma->vm_start; addr < src_vma->vm_end; addr += sz) {
 		spinlock_t *src_ptl, *dst_ptl;
+		/* With vma lock held, safe without RCU */
 		src_pte = huge_pte_offset(src, addr, sz);
 		if (!src_pte) {
 			addr |= last_addr_mask;
@@ -5026,6 +5027,7 @@ int move_hugetlb_page_tables(struct vm_area_struct *vma,
 	hugetlb_vma_lock_write(vma);
 	i_mmap_lock_write(mapping);
 	for (; old_addr < old_end; old_addr += sz, new_addr += sz) {
+		/* With vma lock held, safe without RCU */
 		src_pte = huge_pte_offset(mm, old_addr, sz);
 		if (!src_pte) {
 			old_addr |= last_addr_mask;
@@ -5097,6 +5099,7 @@ static void __unmap_hugepage_range(struct mmu_gather *tlb, struct vm_area_struct
 	last_addr_mask = hugetlb_mask_last_page(h);
 	address = start;
 	for (; address < end; address += sz) {
+		/* With vma lock held, safe without RCU */
 		ptep = huge_pte_offset(mm, address, sz);
 		if (!ptep) {
 			address |= last_addr_mask;
@@ -5402,6 +5405,7 @@ static vm_fault_t hugetlb_wp(struct mm_struct *mm, struct vm_area_struct *vma,
 			mutex_lock(&hugetlb_fault_mutex_table[hash]);
 			hugetlb_vma_lock_read(vma);
 			spin_lock(ptl);
+			/* With vma lock held, safe without RCU */
 			ptep = huge_pte_offset(mm, haddr, huge_page_size(h));
 			if (likely(ptep &&
 				   pte_same(huge_ptep_get(ptep), pte)))
@@ -5440,6 +5444,7 @@ static vm_fault_t hugetlb_wp(struct mm_struct *mm, struct vm_area_struct *vma,
 	 * before the page tables are altered
 	 */
 	spin_lock(ptl);
+	/* With vma lock (and even pgtable lock) held, safe without RCU */
 	ptep = huge_pte_offset(mm, haddr, huge_page_size(h));
 	if (likely(ptep && pte_same(huge_ptep_get(ptep), pte))) {
 		/* Break COW or unshare */
@@ -6511,6 +6516,7 @@ unsigned long hugetlb_change_protection(struct vm_area_struct *vma,
 	last_addr_mask = hugetlb_mask_last_page(h);
 	for (; address < end; address += psize) {
 		spinlock_t *ptl;
+		/* With vma lock held, safe without RCU */
 		ptep = huge_pte_offset(mm, address, psize);
 		if (!ptep) {
 			address |= last_addr_mask;
@@ -7060,7 +7066,14 @@ pte_t *huge_pmd_share(struct mm_struct *mm, struct vm_area_struct *vma,
 
 		saddr = page_table_shareable(svma, vma, addr, idx);
 		if (saddr) {
+			/*
+			 * huge_pmd_share() (or say its solo caller,
+			 * huge_pte_alloc()) always takes the hugetlb vma
+			 * lock, so it's always safe to walk the pgtable of
+			 * the process, even without RCU.
+			 */
 			spte = huge_pte_offset(svma->vm_mm, saddr,
+
 					       vma_mmu_pagesize(svma));
 			if (spte) {
 				get_page(virt_to_page(spte));
@@ -7420,6 +7433,7 @@ void hugetlb_unshare_all_pmds(struct vm_area_struct *vma)
 	hugetlb_vma_lock_write(vma);
 	i_mmap_lock_write(vma->vm_file->f_mapping);
 	for (address = start; address < end; address += PUD_SIZE) {
+		/* With vma lock held, safe without RCU */
 		ptep = huge_pte_offset(mm, address, sz);
 		if (!ptep)
 			continue;
-- 
2.37.3