xen, mm: allow deferred page initialization for xen pv domains

Juergen Gross noticed that commit f7f99100d8d ("mm: stop zeroing memory
during allocation in vmemmap") broke XEN PV domains when deferred struct
page initialization is enabled.

This is because the xen's PagePinned() flag is getting erased from
struct pages when they are initialized later in boot.

Juergen fixed this problem by disabling deferred pages on xen pv
domains.  It is desirable, however, to have this feature available as it
reduces boot time.  This fix re-enables the feature for pv-dmains, and
fixes the problem the following way:

The fix is to delay setting PagePinned flag until struct pages for all
allocated memory are initialized, i.e.  until after free_all_bootmem().

A new x86_init.hyper op init_after_bootmem() is called to let xen know
that boot allocator is done, and hence struct pages for all the
allocated memory are now initialized.  If deferred page initialization
is enabled, the rest of struct pages are going to be initialized later
in boot once page_alloc_init_late() is called.

xen_after_bootmem() walks page table's pages and marks them pinned.

Link: http://lkml.kernel.org/r/20180226160112.24724-2-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Acked-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Juergen Gross <jgross@suse.com>
Tested-by: Juergen Gross <jgross@suse.com>
Cc: Daniel Jordan <daniel.m.jordan@oracle.com>
Cc: Pavel Tatashin <pasha.tatashin@oracle.com>
Cc: Alok Kataria <akataria@vmware.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Laura Abbott <labbott@redhat.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Borislav Petkov <bp@suse.de>
Cc: Mathias Krause <minipli@googlemail.com>
Cc: Jinbum Park <jinb.park7@gmail.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Baoquan He <bhe@redhat.com>
Cc: Jia Zhang <zhang.jia@linux.alibaba.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Stefano Stabellini <sstabellini@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

1 files changed, 26 insertions, 12 deletions

diff --git a/arch/x86/xen/mmu_pv.c b/arch/x86/xen/mmu_pv.c
index d20763472920d..486c0a34d00b2 100644
--- a/arch/x86/xen/mmu_pv.c
+++ b/arch/x86/xen/mmu_pv.c
@@ -116,6 +116,8 @@ DEFINE_PER_CPU(unsigned long, xen_current_cr3);	 /* actual vcpu cr3 */
 
 static phys_addr_t xen_pt_base, xen_pt_size __initdata;
 
+static DEFINE_STATIC_KEY_FALSE(xen_struct_pages_ready);
+
 /*
  * Just beyond the highest usermode address.  STACK_TOP_MAX has a
  * redzone above it, so round it up to a PGD boundary.
@@ -155,11 +157,18 @@ void make_lowmem_page_readwrite(void *vaddr)
 }
 
 
+/*
+ * During early boot all page table pages are pinned, but we do not have struct
+ * pages, so return true until struct pages are ready.
+ */
 static bool xen_page_pinned(void *ptr)
 {
-	struct page *page = virt_to_page(ptr);
+	if (static_branch_likely(&xen_struct_pages_ready)) {
+		struct page *page = virt_to_page(ptr);
 
-	return PagePinned(page);
+		return PagePinned(page);
+	}
+	return true;
 }
 
 static void xen_extend_mmu_update(const struct mmu_update *update)
@@ -836,11 +845,6 @@ void xen_mm_pin_all(void)
 	spin_unlock(&pgd_lock);
 }
 
-/*
- * The init_mm pagetable is really pinned as soon as its created, but
- * that's before we have page structures to store the bits.  So do all
- * the book-keeping now.
- */
 static int __init xen_mark_pinned(struct mm_struct *mm, struct page *page,
 				  enum pt_level level)
 {
@@ -848,8 +852,18 @@ static int __init xen_mark_pinned(struct mm_struct *mm, struct page *page,
 	return 0;
 }
 
-static void __init xen_mark_init_mm_pinned(void)
+/*
+ * The init_mm pagetable is really pinned as soon as its created, but
+ * that's before we have page structures to store the bits.  So do all
+ * the book-keeping now once struct pages for allocated pages are
+ * initialized. This happens only after free_all_bootmem() is called.
+ */
+static void __init xen_after_bootmem(void)
 {
+	static_branch_enable(&xen_struct_pages_ready);
+#ifdef CONFIG_X86_64
+	SetPagePinned(virt_to_page(level3_user_vsyscall));
+#endif
 	xen_pgd_walk(&init_mm, xen_mark_pinned, FIXADDR_TOP);
 }
 
@@ -1623,14 +1637,15 @@ static inline void __set_pfn_prot(unsigned long pfn, pgprot_t prot)
 static inline void xen_alloc_ptpage(struct mm_struct *mm, unsigned long pfn,
 				    unsigned level)
 {
-	bool pinned = PagePinned(virt_to_page(mm->pgd));
+	bool pinned = xen_page_pinned(mm->pgd);
 
 	trace_xen_mmu_alloc_ptpage(mm, pfn, level, pinned);
 
 	if (pinned) {
 		struct page *page = pfn_to_page(pfn);
 
-		SetPagePinned(page);
+		if (static_branch_likely(&xen_struct_pages_ready))
+			SetPagePinned(page);
 
 		if (!PageHighMem(page)) {
 			xen_mc_batch();
@@ -2364,9 +2379,7 @@ static void __init xen_post_allocator_init(void)
 
 #ifdef CONFIG_X86_64
 	pv_mmu_ops.write_cr3 = &xen_write_cr3;
-	SetPagePinned(virt_to_page(level3_user_vsyscall));
 #endif
-	xen_mark_init_mm_pinned();
 }
 
 static void xen_leave_lazy_mmu(void)
@@ -2450,6 +2463,7 @@ static const struct pv_mmu_ops xen_mmu_ops __initconst = {
 void __init xen_init_mmu_ops(void)
 {
 	x86_init.paging.pagetable_init = xen_pagetable_init;
+	x86_init.hyper.init_after_bootmem = xen_after_bootmem;
 
 	pv_mmu_ops = xen_mmu_ops;