mm/vmalloc.c: preload a CPU with one object for split purpose

Refactor the NE_FIT_TYPE split case when it comes to an allocation of one
extra object.  We need it in order to build a remaining space.  The
preload is done per CPU in non-atomic context with GFP_KERNEL flags.

More permissive parameters can be beneficial for systems which are suffer
from high memory pressure or low memory condition.  For example on my KVM
system(4xCPUs, no swap, 256MB RAM) i can simulate the failure of page
allocation with GFP_NOWAIT flags.  Using "stress-ng" tool and starting N
workers spinning on fork() and exit(), i can trigger below trace:

<snip>
[  179.815161] stress-ng-fork: page allocation failure: order:0, mode:0x40800(GFP_NOWAIT|__GFP_COMP), nodemask=(null),cpuset=/,mems_allowed=0
[  179.815168] CPU: 0 PID: 12612 Comm: stress-ng-fork Not tainted 5.2.0-rc3+ #1003
[  179.815170] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.10.2-1 04/01/2014
[  179.815171] Call Trace:
[  179.815178]  dump_stack+0x5c/0x7b
[  179.815182]  warn_alloc+0x108/0x190
[  179.815187]  __alloc_pages_slowpath+0xdc7/0xdf0
[  179.815191]  __alloc_pages_nodemask+0x2de/0x330
[  179.815194]  cache_grow_begin+0x77/0x420
[  179.815197]  fallback_alloc+0x161/0x200
[  179.815200]  kmem_cache_alloc+0x1c9/0x570
[  179.815202]  alloc_vmap_area+0x32c/0x990
[  179.815206]  __get_vm_area_node+0xb0/0x170
[  179.815208]  __vmalloc_node_range+0x6d/0x230
[  179.815211]  ? _do_fork+0xce/0x3d0
[  179.815213]  copy_process.part.46+0x850/0x1b90
[  179.815215]  ? _do_fork+0xce/0x3d0
[  179.815219]  _do_fork+0xce/0x3d0
[  179.815226]  ? __do_page_fault+0x2bf/0x4e0
[  179.815229]  do_syscall_64+0x55/0x130
[  179.815231]  entry_SYSCALL_64_after_hwframe+0x44/0xa9
[  179.815234] RIP: 0033:0x7fedec4c738b
...
[  179.815237] RSP: 002b:00007ffda469d730 EFLAGS: 00000246 ORIG_RAX: 0000000000000038
[  179.815239] RAX: ffffffffffffffda RBX: 00007ffda469d730 RCX: 00007fedec4c738b
[  179.815240] RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000001200011
[  179.815241] RBP: 00007ffda469d780 R08: 00007fededd6e300 R09: 00007ffda47f50a0
[  179.815242] R10: 00007fededd6e5d0 R11: 0000000000000246 R12: 0000000000000000
[  179.815243] R13: 0000000000000020 R14: 0000000000000000 R15: 0000000000000000
[  179.815245] Mem-Info:
[  179.815249] active_anon:12686 inactive_anon:14760 isolated_anon:0
                active_file:502 inactive_file:61 isolated_file:70
                unevictable:2 dirty:0 writeback:0 unstable:0
                slab_reclaimable:2380 slab_unreclaimable:7520
                mapped:15069 shmem:14813 pagetables:10833 bounce:0
                free:1922 free_pcp:229 free_cma:0
<snip>

Link: http://lkml.kernel.org/r/20190606120411.8298-3-urezki@gmail.comSigned-off-by: Uladzislau Rezki (Sony) <urezki@gmail.com>
Cc: Hillf Danton <hdanton@sina.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Oleksiy Avramchenko <oleksiy.avramchenko@sonymobile.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/vmalloc.c

@@ -365,6 +365,13 @@ static LIST_HEAD(free_vmap_area_list);
  */
 static struct rb_root free_vmap_area_root = RB_ROOT;
 
+/*
+ * Preload a CPU with one object for "no edge" split case. The
+ * aim is to get rid of allocations from the atomic context, thus
+ * to use more permissive allocation masks.
+ */
+static DEFINE_PER_CPU(struct vmap_area *, ne_fit_preload_node);
+
 static __always_inline unsigned long
 va_size(struct vmap_area *va)
 {
@@ -951,9 +958,24 @@ adjust_va_to_fit_type(struct vmap_area *va,
 		 *   L V  NVA  V R
 		 * |---|-------|---|
 		 */
-		lva = kmem_cache_alloc(vmap_area_cachep, GFP_NOWAIT);
-		if (unlikely(!lva))
-			return -1;
+		lva = __this_cpu_xchg(ne_fit_preload_node, NULL);
+		if (unlikely(!lva)) {
+			/*
+			 * For percpu allocator we do not do any pre-allocation
+			 * and leave it as it is. The reason is it most likely
+			 * never ends up with NE_FIT_TYPE splitting. In case of
+			 * percpu allocations offsets and sizes are aligned to
+			 * fixed align request, i.e. RE_FIT_TYPE and FL_FIT_TYPE
+			 * are its main fitting cases.
+			 *
+			 * There are a few exceptions though, as an example it is
+			 * a first allocation (early boot up) when we have "one"
+			 * big free space that has to be split.
+			 */
+			lva = kmem_cache_alloc(vmap_area_cachep, GFP_NOWAIT);
+			if (!lva)
+				return -1;
+		}
 
 		/*
 		 * Build the remainder.
@@ -1032,7 +1054,7 @@ static struct vmap_area *alloc_vmap_area(unsigned long size,
 				unsigned long vstart, unsigned long vend,
 				int node, gfp_t gfp_mask)
 {
-	struct vmap_area *va;
+	struct vmap_area *va, *pva;
 	unsigned long addr;
 	int purged = 0;
 
@@ -1057,7 +1079,32 @@ static struct vmap_area *alloc_vmap_area(unsigned long size,
 	kmemleak_scan_area(&va->rb_node, SIZE_MAX, gfp_mask & GFP_RECLAIM_MASK);
 
 retry:
+	/*
+	 * Preload this CPU with one extra vmap_area object to ensure
+	 * that we have it available when fit type of free area is
+	 * NE_FIT_TYPE.
+	 *
+	 * The preload is done in non-atomic context, thus it allows us
+	 * to use more permissive allocation masks to be more stable under
+	 * low memory condition and high memory pressure.
+	 *
+	 * Even if it fails we do not really care about that. Just proceed
+	 * as it is. "overflow" path will refill the cache we allocate from.
+	 */
+	preempt_disable();
+	if (!__this_cpu_read(ne_fit_preload_node)) {
+		preempt_enable();
+		pva = kmem_cache_alloc_node(vmap_area_cachep, GFP_KERNEL, node);
+		preempt_disable();
+
+		if (__this_cpu_cmpxchg(ne_fit_preload_node, NULL, pva)) {
+			if (pva)
+				kmem_cache_free(vmap_area_cachep, pva);
+		}
+	}
+
 	spin_lock(&vmap_area_lock);
+	preempt_enable();
 
 	/*
 	 * If an allocation fails, the "vend" address is