mm: teach core mm about pte markers

This patch still does not use pte marker in any way, however it teaches
the core mm about the pte marker idea.

For example, handle_pte_marker() is introduced that will parse and handle
all the pte marker faults.

Many of the places are more about commenting it up - so that we know
there's the possibility of pte marker showing up, and why we don't need
special code for the cases.

[peterx@redhat.com: userfaultfd.c needs swapops.h]
  Link: https://lkml.kernel.org/r/YmRlVj3cdizYJsr0@xz-m1.local
Link: https://lkml.kernel.org/r/20220405014833.14015-1-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: "Kirill A . Shutemov" <kirill@shutemov.name>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

6 files changed, 40 insertions, 4 deletions

diff --git a/mm/filemap.c b/mm/filemap.c
index 9a1eef6c5d35..de34c6c10384 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -3376,6 +3376,11 @@ again:
 		vmf->pte += xas.xa_index - last_pgoff;
 		last_pgoff = xas.xa_index;
 
+		/*
+		 * NOTE: If there're PTE markers, we'll leave them to be
+		 * handled in the specific fault path, and it'll prohibit the
+		 * fault-around logic.
+		 */
 		if (!pte_none(*vmf->pte))
 			goto unlock;
 
diff --git a/mm/hmm.c b/mm/hmm.c
index af71aac3140e..3fd3242c5e50 100644
--- a/mm/hmm.c
+++ b/mm/hmm.c
@@ -239,7 +239,7 @@ static int hmm_vma_handle_pte(struct mm_walk *walk, unsigned long addr,
 	pte_t pte = *ptep;
 	uint64_t pfn_req_flags = *hmm_pfn;
 
-	if (pte_none(pte)) {
+	if (pte_none_mostly(pte)) {
 		required_fault =
 			hmm_pte_need_fault(hmm_vma_walk, pfn_req_flags, 0);
 		if (required_fault)
diff --git a/mm/memcontrol.c b/mm/memcontrol.c
index 235c85e17c2d..6e74ca98c862 100644
--- a/mm/memcontrol.c
+++ b/mm/memcontrol.c
@@ -5640,10 +5640,14 @@ static enum mc_target_type get_mctgt_type(struct vm_area_struct *vma,
 
 	if (pte_present(ptent))
 		page = mc_handle_present_pte(vma, addr, ptent);
+	else if (pte_none_mostly(ptent))
+		/*
+		 * PTE markers should be treated as a none pte here, separated
+		 * from other swap handling below.
+		 */
+		page = mc_handle_file_pte(vma, addr, ptent);
 	else if (is_swap_pte(ptent))
 		page = mc_handle_swap_pte(vma, ptent, &ent);
-	else if (pte_none(ptent))
-		page = mc_handle_file_pte(vma, addr, ptent);
 
 	if (!page && !ent.val)
 		return ret;
diff --git a/mm/memory.c b/mm/memory.c
index 90212057a546..9743c8b74bf2 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -100,6 +100,8 @@ struct page *mem_map;
 EXPORT_SYMBOL(mem_map);
 #endif
 
+static vm_fault_t do_fault(struct vm_fault *vmf);
+
 /*
  * A number of key systems in x86 including ioremap() rely on the assumption
  * that high_memory defines the upper bound on direct map memory, then end
@@ -1415,6 +1417,8 @@ again:
 			if (!should_zap_page(details, page))
 				continue;
 			rss[mm_counter(page)]--;
+		} else if (is_pte_marker_entry(entry)) {
+			/* By default, simply drop all pte markers when zap */
 		} else if (is_hwpoison_entry(entry)) {
 			if (!should_zap_cows(details))
 				continue;
@@ -3555,6 +3559,23 @@ static inline bool should_try_to_free_swap(struct page *page,
 		page_count(page) == 2;
 }
 
+static vm_fault_t handle_pte_marker(struct vm_fault *vmf)
+{
+	swp_entry_t entry = pte_to_swp_entry(vmf->orig_pte);
+	unsigned long marker = pte_marker_get(entry);
+
+	/*
+	 * PTE markers should always be with file-backed memories, and the
+	 * marker should never be empty.  If anything weird happened, the best
+	 * thing to do is to kill the process along with its mm.
+	 */
+	if (WARN_ON_ONCE(vma_is_anonymous(vmf->vma) || !marker))
+		return VM_FAULT_SIGBUS;
+
+	/* TODO: handle pte markers */
+	return 0;
+}
+
 /*
  * We enter with non-exclusive mmap_lock (to exclude vma changes,
  * but allow concurrent faults), and pte mapped but not yet locked.
@@ -3592,6 +3613,8 @@ vm_fault_t do_swap_page(struct vm_fault *vmf)
 			ret = vmf->page->pgmap->ops->migrate_to_ram(vmf);
 		} else if (is_hwpoison_entry(entry)) {
 			ret = VM_FAULT_HWPOISON;
+		} else if (is_pte_marker_entry(entry)) {
+			ret = handle_pte_marker(vmf);
 		} else {
 			print_bad_pte(vma, vmf->address, vmf->orig_pte, NULL);
 			ret = VM_FAULT_SIGBUS;
diff --git a/mm/mincore.c b/mm/mincore.c
index f4f627325e12..fa200c14185f 100644
--- a/mm/mincore.c
+++ b/mm/mincore.c
@@ -122,7 +122,8 @@ static int mincore_pte_range(pmd_t *pmd, unsigned long addr, unsigned long end,
 	for (; addr != end; ptep++, addr += PAGE_SIZE) {
 		pte_t pte = *ptep;
 
-		if (pte_none(pte))
+		/* We need to do cache lookup too for pte markers */
+		if (pte_none_mostly(pte))
 			__mincore_unmapped_range(addr, addr + PAGE_SIZE,
 						 vma, vec);
 		else if (pte_present(pte))
diff --git a/mm/mprotect.c b/mm/mprotect.c
index 20a46f21cca8..e84694267b0f 100644
--- a/mm/mprotect.c
+++ b/mm/mprotect.c
@@ -193,6 +193,9 @@ static unsigned long change_pte_range(struct mmu_gather *tlb,
 					newpte = pte_swp_mksoft_dirty(newpte);
 				if (pte_swp_uffd_wp(oldpte))
 					newpte = pte_swp_mkuffd_wp(newpte);
+			} else if (is_pte_marker_entry(entry)) {
+				/* Skip it, the same as none pte */
+				continue;
 			} else {
 				newpte = oldpte;
 			}