mm, x86: add support for PUD-sized transparent hugepages

The current transparent hugepage code only supports PMDs.  This patch
adds support for transparent use of PUDs with DAX.  It does not include
support for anonymous pages.  x86 support code also added.

Most of this patch simply parallels the work that was done for huge
PMDs.  The only major difference is how the new ->pud_entry method in
mm_walk works.  The ->pmd_entry method replaces the ->pte_entry method,
whereas the ->pud_entry method works along with either ->pmd_entry or
->pte_entry.  The pagewalk code takes care of locking the PUD before
calling ->pud_walk, so handlers do not need to worry whether the PUD is
stable.

[dave.jiang@intel.com: fix SMP x86 32bit build for native_pud_clear()]
  Link: http://lkml.kernel.org/r/148719066814.31111.3239231168815337012.stgit@djiang5-desk3.ch.intel.com
[dave.jiang@intel.com: native_pud_clear missing on i386 build]
  Link: http://lkml.kernel.org/r/148640375195.69754.3315433724330910314.stgit@djiang5-desk3.ch.intel.com
Link: http://lkml.kernel.org/r/148545059381.17912.8602162635537598445.stgit@djiang5-desk3.ch.intel.com
Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Tested-by: Alexander Kapshuk <alexander.kapshuk@gmail.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Jan Kara <jack@suse.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Ross Zwisler <ross.zwisler@linux.intel.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Nilesh Choudhury <nilesh.choudhury@oracle.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

5 files changed, 370 insertions, 8 deletions

diff --git a/mm/gup.c b/mm/gup.c
index 40abe4c..1e67461 100644
--- a/mm/gup.c
+++ b/mm/gup.c
@@ -253,6 +253,13 @@ struct page *follow_page_mask(struct vm_area_struct *vma,
 			return page;
 		return no_page_table(vma, flags);
 	}
+	if (pud_devmap(*pud)) {
+		ptl = pud_lock(mm, pud);
+		page = follow_devmap_pud(vma, address, pud, flags);
+		spin_unlock(ptl);
+		if (page)
+			return page;
+	}
 	if (unlikely(pud_bad(*pud)))
 		return no_page_table(vma, flags);
 
diff --git a/mm/huge_memory.c b/mm/huge_memory.c
index f9ecc2a..85742ac 100644
--- a/mm/huge_memory.c
+++ b/mm/huge_memory.c
@@ -757,6 +757,60 @@ int vmf_insert_pfn_pmd(struct vm_area_struct *vma, unsigned long addr,
 }
 EXPORT_SYMBOL_GPL(vmf_insert_pfn_pmd);
 
+#ifdef CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
+static pud_t maybe_pud_mkwrite(pud_t pud, struct vm_area_struct *vma)
+{
+	if (likely(vma->vm_flags & VM_WRITE))
+		pud = pud_mkwrite(pud);
+	return pud;
+}
+
+static void insert_pfn_pud(struct vm_area_struct *vma, unsigned long addr,
+		pud_t *pud, pfn_t pfn, pgprot_t prot, bool write)
+{
+	struct mm_struct *mm = vma->vm_mm;
+	pud_t entry;
+	spinlock_t *ptl;
+
+	ptl = pud_lock(mm, pud);
+	entry = pud_mkhuge(pfn_t_pud(pfn, prot));
+	if (pfn_t_devmap(pfn))
+		entry = pud_mkdevmap(entry);
+	if (write) {
+		entry = pud_mkyoung(pud_mkdirty(entry));
+		entry = maybe_pud_mkwrite(entry, vma);
+	}
+	set_pud_at(mm, addr, pud, entry);
+	update_mmu_cache_pud(vma, addr, pud);
+	spin_unlock(ptl);
+}
+
+int vmf_insert_pfn_pud(struct vm_area_struct *vma, unsigned long addr,
+			pud_t *pud, pfn_t pfn, bool write)
+{
+	pgprot_t pgprot = vma->vm_page_prot;
+	/*
+	 * If we had pud_special, we could avoid all these restrictions,
+	 * but we need to be consistent with PTEs and architectures that
+	 * can't support a 'special' bit.
+	 */
+	BUG_ON(!(vma->vm_flags & (VM_PFNMAP|VM_MIXEDMAP)));
+	BUG_ON((vma->vm_flags & (VM_PFNMAP|VM_MIXEDMAP)) ==
+						(VM_PFNMAP|VM_MIXEDMAP));
+	BUG_ON((vma->vm_flags & VM_PFNMAP) && is_cow_mapping(vma->vm_flags));
+	BUG_ON(!pfn_t_devmap(pfn));
+
+	if (addr < vma->vm_start || addr >= vma->vm_end)
+		return VM_FAULT_SIGBUS;
+
+	track_pfn_insert(vma, &pgprot, pfn);
+
+	insert_pfn_pud(vma, addr, pud, pfn, pgprot, write);
+	return VM_FAULT_NOPAGE;
+}
+EXPORT_SYMBOL_GPL(vmf_insert_pfn_pud);
+#endif /* CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD */
+
 static void touch_pmd(struct vm_area_struct *vma, unsigned long addr,
 		pmd_t *pmd)
 {
@@ -887,6 +941,123 @@ out:
 	return ret;
 }
 
+#ifdef CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
+static void touch_pud(struct vm_area_struct *vma, unsigned long addr,
+		pud_t *pud)
+{
+	pud_t _pud;
+
+	/*
+	 * We should set the dirty bit only for FOLL_WRITE but for now
+	 * the dirty bit in the pud is meaningless.  And if the dirty
+	 * bit will become meaningful and we'll only set it with
+	 * FOLL_WRITE, an atomic set_bit will be required on the pud to
+	 * set the young bit, instead of the current set_pud_at.
+	 */
+	_pud = pud_mkyoung(pud_mkdirty(*pud));
+	if (pudp_set_access_flags(vma, addr & HPAGE_PUD_MASK,
+				pud, _pud,  1))
+		update_mmu_cache_pud(vma, addr, pud);
+}
+
+struct page *follow_devmap_pud(struct vm_area_struct *vma, unsigned long addr,
+		pud_t *pud, int flags)
+{
+	unsigned long pfn = pud_pfn(*pud);
+	struct mm_struct *mm = vma->vm_mm;
+	struct dev_pagemap *pgmap;
+	struct page *page;
+
+	assert_spin_locked(pud_lockptr(mm, pud));
+
+	if (flags & FOLL_WRITE && !pud_write(*pud))
+		return NULL;
+
+	if (pud_present(*pud) && pud_devmap(*pud))
+		/* pass */;
+	else
+		return NULL;
+
+	if (flags & FOLL_TOUCH)
+		touch_pud(vma, addr, pud);
+
+	/*
+	 * device mapped pages can only be returned if the
+	 * caller will manage the page reference count.
+	 */
+	if (!(flags & FOLL_GET))
+		return ERR_PTR(-EEXIST);
+
+	pfn += (addr & ~PUD_MASK) >> PAGE_SHIFT;
+	pgmap = get_dev_pagemap(pfn, NULL);
+	if (!pgmap)
+		return ERR_PTR(-EFAULT);
+	page = pfn_to_page(pfn);
+	get_page(page);
+	put_dev_pagemap(pgmap);
+
+	return page;
+}
+
+int copy_huge_pud(struct mm_struct *dst_mm, struct mm_struct *src_mm,
+		  pud_t *dst_pud, pud_t *src_pud, unsigned long addr,
+		  struct vm_area_struct *vma)
+{
+	spinlock_t *dst_ptl, *src_ptl;
+	pud_t pud;
+	int ret;
+
+	dst_ptl = pud_lock(dst_mm, dst_pud);
+	src_ptl = pud_lockptr(src_mm, src_pud);
+	spin_lock_nested(src_ptl, SINGLE_DEPTH_NESTING);
+
+	ret = -EAGAIN;
+	pud = *src_pud;
+	if (unlikely(!pud_trans_huge(pud) && !pud_devmap(pud)))
+		goto out_unlock;
+
+	/*
+	 * When page table lock is held, the huge zero pud should not be
+	 * under splitting since we don't split the page itself, only pud to
+	 * a page table.
+	 */
+	if (is_huge_zero_pud(pud)) {
+		/* No huge zero pud yet */
+	}
+
+	pudp_set_wrprotect(src_mm, addr, src_pud);
+	pud = pud_mkold(pud_wrprotect(pud));
+	set_pud_at(dst_mm, addr, dst_pud, pud);
+
+	ret = 0;
+out_unlock:
+	spin_unlock(src_ptl);
+	spin_unlock(dst_ptl);
+	return ret;
+}
+
+void huge_pud_set_accessed(struct vm_fault *vmf, pud_t orig_pud)
+{
+	pud_t entry;
+	unsigned long haddr;
+	bool write = vmf->flags & FAULT_FLAG_WRITE;
+
+	vmf->ptl = pud_lock(vmf->vma->vm_mm, vmf->pud);
+	if (unlikely(!pud_same(*vmf->pud, orig_pud)))
+		goto unlock;
+
+	entry = pud_mkyoung(orig_pud);
+	if (write)
+		entry = pud_mkdirty(entry);
+	haddr = vmf->address & HPAGE_PUD_MASK;
+	if (pudp_set_access_flags(vmf->vma, haddr, vmf->pud, entry, write))
+		update_mmu_cache_pud(vmf->vma, vmf->address, vmf->pud);
+
+unlock:
+	spin_unlock(vmf->ptl);
+}
+#endif /* CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD */
+
 void huge_pmd_set_accessed(struct vm_fault *vmf, pmd_t orig_pmd)
 {
 	pmd_t entry;
@@ -1601,6 +1772,84 @@ spinlock_t *__pmd_trans_huge_lock(pmd_t *pmd, struct vm_area_struct *vma)
 	return NULL;
 }
 
+/*
+ * Returns true if a given pud maps a thp, false otherwise.
+ *
+ * Note that if it returns true, this routine returns without unlocking page
+ * table lock. So callers must unlock it.
+ */
+spinlock_t *__pud_trans_huge_lock(pud_t *pud, struct vm_area_struct *vma)
+{
+	spinlock_t *ptl;
+
+	ptl = pud_lock(vma->vm_mm, pud);
+	if (likely(pud_trans_huge(*pud) || pud_devmap(*pud)))
+		return ptl;
+	spin_unlock(ptl);
+	return NULL;
+}
+
+#ifdef CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
+int zap_huge_pud(struct mmu_gather *tlb, struct vm_area_struct *vma,
+		 pud_t *pud, unsigned long addr)
+{
+	pud_t orig_pud;
+	spinlock_t *ptl;
+
+	ptl = __pud_trans_huge_lock(pud, vma);
+	if (!ptl)
+		return 0;
+	/*
+	 * For architectures like ppc64 we look at deposited pgtable
+	 * when calling pudp_huge_get_and_clear. So do the
+	 * pgtable_trans_huge_withdraw after finishing pudp related
+	 * operations.
+	 */
+	orig_pud = pudp_huge_get_and_clear_full(tlb->mm, addr, pud,
+			tlb->fullmm);
+	tlb_remove_pud_tlb_entry(tlb, pud, addr);
+	if (vma_is_dax(vma)) {
+		spin_unlock(ptl);
+		/* No zero page support yet */
+	} else {
+		/* No support for anonymous PUD pages yet */
+		BUG();
+	}
+	return 1;
+}
+
+static void __split_huge_pud_locked(struct vm_area_struct *vma, pud_t *pud,
+		unsigned long haddr)
+{
+	VM_BUG_ON(haddr & ~HPAGE_PUD_MASK);
+	VM_BUG_ON_VMA(vma->vm_start > haddr, vma);
+	VM_BUG_ON_VMA(vma->vm_end < haddr + HPAGE_PUD_SIZE, vma);
+	VM_BUG_ON(!pud_trans_huge(*pud) && !pud_devmap(*pud));
+
+	count_vm_event(THP_SPLIT_PMD);
+
+	pudp_huge_clear_flush_notify(vma, haddr, pud);
+}
+
+void __split_huge_pud(struct vm_area_struct *vma, pud_t *pud,
+		unsigned long address)
+{
+	spinlock_t *ptl;
+	struct mm_struct *mm = vma->vm_mm;
+	unsigned long haddr = address & HPAGE_PUD_MASK;
+
+	mmu_notifier_invalidate_range_start(mm, haddr, haddr + HPAGE_PUD_SIZE);
+	ptl = pud_lock(mm, pud);
+	if (unlikely(!pud_trans_huge(*pud) && !pud_devmap(*pud)))
+		goto out;
+	__split_huge_pud_locked(vma, pud, haddr);
+
+out:
+	spin_unlock(ptl);
+	mmu_notifier_invalidate_range_end(mm, haddr, haddr + HPAGE_PUD_SIZE);
+}
+#endif /* CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD */
+
 static void __split_huge_zero_page_pmd(struct vm_area_struct *vma,
 		unsigned long haddr, pmd_t *pmd)
 {
diff --git a/mm/memory.c b/mm/memory.c
index e721e8e..41e2a2d 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -1001,7 +1001,7 @@ static inline int copy_pmd_range(struct mm_struct *dst_mm, struct mm_struct *src
 		next = pmd_addr_end(addr, end);
 		if (pmd_trans_huge(*src_pmd) || pmd_devmap(*src_pmd)) {
 			int err;
-			VM_BUG_ON(next-addr != HPAGE_PMD_SIZE);
+			VM_BUG_ON_VMA(next-addr != HPAGE_PMD_SIZE, vma);
 			err = copy_huge_pmd(dst_mm, src_mm,
 					    dst_pmd, src_pmd, addr, vma);
 			if (err == -ENOMEM)
@@ -1032,6 +1032,18 @@ static inline int copy_pud_range(struct mm_struct *dst_mm, struct mm_struct *src
 	src_pud = pud_offset(src_pgd, addr);
 	do {
 		next = pud_addr_end(addr, end);
+		if (pud_trans_huge(*src_pud) || pud_devmap(*src_pud)) {
+			int err;
+
+			VM_BUG_ON_VMA(next-addr != HPAGE_PUD_SIZE, vma);
+			err = copy_huge_pud(dst_mm, src_mm,
+					    dst_pud, src_pud, addr, vma);
+			if (err == -ENOMEM)
+				return -ENOMEM;
+			if (!err)
+				continue;
+			/* fall through */
+		}
 		if (pud_none_or_clear_bad(src_pud))
 			continue;
 		if (copy_pmd_range(dst_mm, src_mm, dst_pud, src_pud,
@@ -1263,9 +1275,19 @@ static inline unsigned long zap_pud_range(struct mmu_gather *tlb,
 	pud = pud_offset(pgd, addr);
 	do {
 		next = pud_addr_end(addr, end);
+		if (pud_trans_huge(*pud) || pud_devmap(*pud)) {
+			if (next - addr != HPAGE_PUD_SIZE) {
+				VM_BUG_ON_VMA(!rwsem_is_locked(&tlb->mm->mmap_sem), vma);
+				split_huge_pud(vma, pud, addr);
+			} else if (zap_huge_pud(tlb, vma, pud, addr))
+				goto next;
+			/* fall through */
+		}
 		if (pud_none_or_clear_bad(pud))
 			continue;
 		next = zap_pmd_range(tlb, vma, pud, addr, next, details);
+next:
+		cond_resched();
 	} while (pud++, addr = next, addr != end);
 
 	return addr;
@@ -3490,6 +3512,30 @@ static inline bool vma_is_accessible(struct vm_area_struct *vma)
 	return vma->vm_flags & (VM_READ | VM_EXEC | VM_WRITE);
 }
 
+static int create_huge_pud(struct vm_fault *vmf)
+{
+#ifdef CONFIG_TRANSPARENT_HUGEPAGE
+	/* No support for anonymous transparent PUD pages yet */
+	if (vma_is_anonymous(vmf->vma))
+		return VM_FAULT_FALLBACK;
+	if (vmf->vma->vm_ops->huge_fault)
+		return vmf->vma->vm_ops->huge_fault(vmf);
+#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
+	return VM_FAULT_FALLBACK;
+}
+
+static int wp_huge_pud(struct vm_fault *vmf, pud_t orig_pud)
+{
+#ifdef CONFIG_TRANSPARENT_HUGEPAGE
+	/* No support for anonymous transparent PUD pages yet */
+	if (vma_is_anonymous(vmf->vma))
+		return VM_FAULT_FALLBACK;
+	if (vmf->vma->vm_ops->huge_fault)
+		return vmf->vma->vm_ops->huge_fault(vmf);
+#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
+	return VM_FAULT_FALLBACK;
+}
+
 /*
  * These routines also need to handle stuff like marking pages dirty
  * and/or accessed for architectures that don't do it in hardware (most
@@ -3605,14 +3651,41 @@ static int __handle_mm_fault(struct vm_area_struct *vma, unsigned long address,
 	};
 	struct mm_struct *mm = vma->vm_mm;
 	pgd_t *pgd;
-	pud_t *pud;
 	int ret;
 
 	pgd = pgd_offset(mm, address);
-	pud = pud_alloc(mm, pgd, address);
-	if (!pud)
+
+	vmf.pud = pud_alloc(mm, pgd, address);
+	if (!vmf.pud)
 		return VM_FAULT_OOM;
-	vmf.pmd = pmd_alloc(mm, pud, address);
+	if (pud_none(*vmf.pud) && transparent_hugepage_enabled(vma)) {
+		vmf.flags |= FAULT_FLAG_SIZE_PUD;
+		ret = create_huge_pud(&vmf);
+		if (!(ret & VM_FAULT_FALLBACK))
+			return ret;
+	} else {
+		pud_t orig_pud = *vmf.pud;
+
+		barrier();
+		if (pud_trans_huge(orig_pud) || pud_devmap(orig_pud)) {
+			unsigned int dirty = flags & FAULT_FLAG_WRITE;
+
+			vmf.flags |= FAULT_FLAG_SIZE_PUD;
+
+			/* NUMA case for anonymous PUDs would go here */
+
+			if (dirty && !pud_write(orig_pud)) {
+				ret = wp_huge_pud(&vmf, orig_pud);
+				if (!(ret & VM_FAULT_FALLBACK))
+					return ret;
+			} else {
+				huge_pud_set_accessed(&vmf, orig_pud);
+				return 0;
+			}
+		}
+	}
+
+	vmf.pmd = pmd_alloc(mm, vmf.pud, address);
 	if (!vmf.pmd)
 		return VM_FAULT_OOM;
 	if (pmd_none(*vmf.pmd) && transparent_hugepage_enabled(vma)) {
@@ -3743,13 +3816,14 @@ int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
  */
 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
 {
+	spinlock_t *ptl;
 	pmd_t *new = pmd_alloc_one(mm, address);
 	if (!new)
 		return -ENOMEM;
 
 	smp_wmb(); /* See comment in __pte_alloc */
 
-	spin_lock(&mm->page_table_lock);
+	ptl = pud_lock(mm, pud);
 #ifndef __ARCH_HAS_4LEVEL_HACK
 	if (!pud_present(*pud)) {
 		mm_inc_nr_pmds(mm);
@@ -3763,7 +3837,7 @@ int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
 	} else /* Another has populated it */
 		pmd_free(mm, new);
 #endif /* __ARCH_HAS_4LEVEL_HACK */
-	spin_unlock(&mm->page_table_lock);
+	spin_unlock(ptl);
 	return 0;
 }
 #endif /* __PAGETABLE_PMD_FOLDED */
diff --git a/mm/pagewalk.c b/mm/pagewalk.c
index 2072444..0376157 100644
--- a/mm/pagewalk.c
+++ b/mm/pagewalk.c
@@ -78,14 +78,32 @@ static int walk_pud_range(pgd_t *pgd, unsigned long addr, unsigned long end,
 
 	pud = pud_offset(pgd, addr);
 	do {
+ again:
 		next = pud_addr_end(addr, end);
-		if (pud_none_or_clear_bad(pud)) {
+		if (pud_none(*pud) || !walk->vma) {
 			if (walk->pte_hole)
 				err = walk->pte_hole(addr, next, walk);
 			if (err)
 				break;
 			continue;
 		}
+
+		if (walk->pud_entry) {
+			spinlock_t *ptl = pud_trans_huge_lock(pud, walk->vma);
+
+			if (ptl) {
+				err = walk->pud_entry(pud, addr, next, walk);
+				spin_unlock(ptl);
+				if (err)
+					break;
+				continue;
+			}
+		}
+
+		split_huge_pud(walk->vma, pud, addr);
+		if (pud_none(*pud))
+			goto again;
+
 		if (walk->pmd_entry || walk->pte_entry)
 			err = walk_pmd_range(pud, addr, next, walk);
 		if (err)
diff --git a/mm/pgtable-generic.c b/mm/pgtable-generic.c
index 71c5f91..4ed5908 100644
--- a/mm/pgtable-generic.c
+++ b/mm/pgtable-generic.c
@@ -123,6 +123,20 @@ pmd_t pmdp_huge_clear_flush(struct vm_area_struct *vma, unsigned long address,
 	flush_pmd_tlb_range(vma, address, address + HPAGE_PMD_SIZE);
 	return pmd;
 }
+
+#ifdef CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
+pud_t pudp_huge_clear_flush(struct vm_area_struct *vma, unsigned long address,
+			    pud_t *pudp)
+{
+	pud_t pud;
+
+	VM_BUG_ON(address & ~HPAGE_PUD_MASK);
+	VM_BUG_ON(!pud_trans_huge(*pudp) && !pud_devmap(*pudp));
+	pud = pudp_huge_get_and_clear(vma->vm_mm, address, pudp);
+	flush_pud_tlb_range(vma, address, address + HPAGE_PUD_SIZE);
+	return pud;
+}
+#endif
 #endif
 
 #ifndef __HAVE_ARCH_PGTABLE_DEPOSIT