x86/mm: split vmalloc_sync_all()

Commit 3f8fd02b1b ("mm/vmalloc: Sync unmappings in
__purge_vmap_area_lazy()") introduced a call to vmalloc_sync_all() in
the vunmap() code-path.  While this change was necessary to maintain
correctness on x86-32-pae kernels, it also adds additional cycles for
architectures that don't need it.

Specifically on x86-64 with CONFIG_VMAP_STACK=y some people reported
severe performance regressions in micro-benchmarks because it now also
calls the x86-64 implementation of vmalloc_sync_all() on vunmap().  But
the vmalloc_sync_all() implementation on x86-64 is only needed for newly
created mappings.

To avoid the unnecessary work on x86-64 and to gain the performance
back, split up vmalloc_sync_all() into two functions:

	* vmalloc_sync_mappings(), and
	* vmalloc_sync_unmappings()

Most call-sites to vmalloc_sync_all() only care about new mappings being
synchronized.  The only exception is the new call-site added in the
above mentioned commit.

Shile Zhang directed us to a report of an 80% regression in reaim
throughput.

Fixes: 3f8fd02b1b ("mm/vmalloc: Sync unmappings in __purge_vmap_area_lazy()")
Reported-by: kernel test robot <oliver.sang@intel.com>
Reported-by: Shile Zhang <shile.zhang@linux.alibaba.com>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Tested-by: Borislav Petkov <bp@suse.de>
Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>	[GHES]
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: <stable@vger.kernel.org>
Link: http://lkml.kernel.org/r/20191009124418.8286-1-joro@8bytes.org
Link: https://lists.01.org/hyperkitty/list/lkp@lists.01.org/thread/4D3JPPHBNOSPFK2KEPC6KGKS6J25AIDB/
Link: http://lkml.kernel.org/r/20191113095530.228959-1-shile.zhang@linux.alibaba.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

arch/x86/mm/fault.c

@@ -190,7 +190,7 @@ static inline pmd_t *vmalloc_sync_one(pgd_t *pgd, unsigned long address)
 	return pmd_k;
 }
 
-void vmalloc_sync_all(void)
+static void vmalloc_sync(void)
 {
 	unsigned long address;
 
@@ -217,6 +217,16 @@ void vmalloc_sync_all(void)
 	}
 }
 
+void vmalloc_sync_mappings(void)
+{
+	vmalloc_sync();
+}
+
+void vmalloc_sync_unmappings(void)
+{
+	vmalloc_sync();
+}
+
 /*
  * 32-bit:
  *
@@ -319,11 +329,23 @@ out:
 
 #else /* CONFIG_X86_64: */
 
-void vmalloc_sync_all(void)
+void vmalloc_sync_mappings(void)
 {
+	/*
+	 * 64-bit mappings might allocate new p4d/pud pages
+	 * that need to be propagated to all tasks' PGDs.
+	 */
 	sync_global_pgds(VMALLOC_START & PGDIR_MASK, VMALLOC_END);
 }
 
+void vmalloc_sync_unmappings(void)
+{
+	/*
+	 * Unmappings never allocate or free p4d/pud pages.
+	 * No work is required here.
+	 */
+}
+
 /*
  * 64-bit:
  *

drivers/acpi/apei/ghes.c

@@ -171,7 +171,7 @@ int ghes_estatus_pool_init(int num_ghes)
 	 * New allocation must be visible in all pgd before it can be found by
 	 * an NMI allocating from the pool.
 	 */
-	vmalloc_sync_all();
+	vmalloc_sync_mappings();
 
 	rc = gen_pool_add(ghes_estatus_pool, addr, PAGE_ALIGN(len), -1);
 	if (rc)

include/linux/vmalloc.h

@@ -141,7 +141,8 @@ extern int remap_vmalloc_range_partial(struct vm_area_struct *vma,
 
 extern int remap_vmalloc_range(struct vm_area_struct *vma, void *addr,
 							unsigned long pgoff);
-void vmalloc_sync_all(void);
+void vmalloc_sync_mappings(void);
+void vmalloc_sync_unmappings(void);
 
 /*
  *	Lowlevel-APIs (not for driver use!)

kernel/notifier.c

@@ -519,7 +519,7 @@ NOKPROBE_SYMBOL(notify_die);
 
 int register_die_notifier(struct notifier_block *nb)
 {
-	vmalloc_sync_all();
+	vmalloc_sync_mappings();
 	return atomic_notifier_chain_register(&die_chain, nb);
 }
 EXPORT_SYMBOL_GPL(register_die_notifier);

mm/nommu.c

@@ -370,10 +370,14 @@ void vm_unmap_aliases(void)
 EXPORT_SYMBOL_GPL(vm_unmap_aliases);
 
 /*
- * Implement a stub for vmalloc_sync_all() if the architecture chose not to
+ * Implement a stub for vmalloc_sync_[un]mapping() if the architecture
- * have one.
+ * chose not to have one.
  */
-void __weak vmalloc_sync_all(void)
+void __weak vmalloc_sync_mappings(void)
+{
+}
+
+void __weak vmalloc_sync_unmappings(void)
 {
 }
 

mm/vmalloc.c

@@ -1295,7 +1295,7 @@ static bool __purge_vmap_area_lazy(unsigned long start, unsigned long end)
 	 * First make sure the mappings are removed from all page-tables
 	 * before they are freed.
 	 */
-	vmalloc_sync_all();
+	vmalloc_sync_unmappings();
 
 	/*
 	 * TODO: to calculate a flush range without looping.
@@ -3128,16 +3128,19 @@ int remap_vmalloc_range(struct vm_area_struct *vma, void *addr,
 EXPORT_SYMBOL(remap_vmalloc_range);
 
 /*
- * Implement a stub for vmalloc_sync_all() if the architecture chose not to
+ * Implement stubs for vmalloc_sync_[un]mappings () if the architecture chose
- * have one.
+ * not to have one.
  *
  * The purpose of this function is to make sure the vmalloc area
  * mappings are identical in all page-tables in the system.
  */
-void __weak vmalloc_sync_all(void)
+void __weak vmalloc_sync_mappings(void)
 {
 }
 
+void __weak vmalloc_sync_unmappings(void)
+{
+}
 
 static int f(pte_t *pte, unsigned long addr, void *data)
 {