runtime: grow the heap incrementally

Currently, we map and grow the heap a whole arena (64MB) at a time.
Unfortunately, in order to fix #32828, we need to switch from
scavenging inline with allocation back to scavenging on heap growth,
but heap-growth scavenging happens in large jumps because we grow the
heap in large jumps.

In order to prepare for better heap-growth scavenging, this CL
separates mapping more space for the heap from actually "growing" it
(tracking the new space with spans). Instead, growing the heap keeps
track of the "current arena" it's growing into. It track that with new
spans as needed, and only maps more arena space when the current arena
is inadequate. The effect to the user is the same, but this will let
us scavenge on much smaller increments of heap growth.

There are two slightly subtleties to this change:

1. If an allocation requires mapping a new arena and that new arena
   isn't contiguous with the current arena, we don't want to lose the
   unused space in the current arena, so we have to immediately track
   that with a span.

2. The mapped space must be accounted as released and idle, even
   though it isn't actually tracked in a span.

For #32828, since this makes heap-growth scavenging far more
effective, especially at small heap sizes. For example, this change is
necessary for TestPhysicalMemoryUtilization to pass once we remove
inline scavenging.

Change-Id: I300e74a0534062467e4ce91cdc3508e5ef9aa73a
Reviewed-on: https://go-review.googlesource.com/c/go/+/189957
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>

src/runtime/export_test.go

@@ -345,6 +345,9 @@ func ReadMemStatsSlow() (base, slow MemStats) {
 			slow.HeapReleased += uint64(i.span().released())
 		}
 
+		// Unused space in the current arena also counts as released space.
+		slow.HeapReleased += uint64(mheap_.curArena.end - mheap_.curArena.base)
+
 		getg().m.mallocing--
 	})
 

src/runtime/mheap.go

@@ -185,6 +185,12 @@ type mheap struct {
 	// simply blocking GC (by disabling preemption).
 	sweepArenas []arenaIdx
 
+	// curArena is the arena that the heap is currently growing
+	// into. This should always be physPageSize-aligned.
+	curArena struct {
+		base, end uintptr
+	}
+
 	_ uint32 // ensure 64-bit alignment of central
 
 	// central free lists for small size classes.
@@ -1249,29 +1255,74 @@ HaveSpan:
 // h must be locked.
 func (h *mheap) grow(npage uintptr) bool {
 	ask := npage << _PageShift
-	v, size := h.sysAlloc(ask)
-	if v == nil {
-		print("runtime: out of memory: cannot allocate ", ask, "-byte block (", memstats.heap_sys, " in use)\n")
-		return false
-	}
 
-	// Create a fake "in use" span and free it, so that the
-	// right accounting and coalescing happens.
+	nBase := round(h.curArena.base+ask, physPageSize)
+	if nBase > h.curArena.end {
+		// Not enough room in the current arena. Allocate more
+		// arena space. This may not be contiguous with the
+		// current arena, so we have to request the full ask.
+		av, asize := h.sysAlloc(ask)
+		if av == nil {
+			print("runtime: out of memory: cannot allocate ", ask, "-byte block (", memstats.heap_sys, " in use)\n")
+			return false
+		}
+
+		if uintptr(av) == h.curArena.end {
+			// The new space is contiguous with the old
+			// space, so just extend the current space.
+			h.curArena.end = uintptr(av) + asize
+		} else {
+			// The new space is discontiguous. Track what
+			// remains of the current space and switch to
+			// the new space. This should be rare.
+			if size := h.curArena.end - h.curArena.base; size != 0 {
+				h.growAddSpan(unsafe.Pointer(h.curArena.base), size)
+			}
+			// Switch to the new space.
+			h.curArena.base = uintptr(av)
+			h.curArena.end = uintptr(av) + asize
+		}
+
+		// The memory just allocated counts as both released
+		// and idle, even though it's not yet backed by spans.
+		//
+		// The allocation is always aligned to the heap arena
+		// size which is always > physPageSize, so its safe to
+		// just add directly to heap_released. Coalescing, if
+		// possible, will also always be correct in terms of
+		// accounting, because s.base() must be a physical
+		// page boundary.
+		memstats.heap_released += uint64(asize)
+		memstats.heap_idle += uint64(asize)
+
+		// Recalculate nBase
+		nBase = round(h.curArena.base+ask, physPageSize)
+	}
+
+	// Grow into the current arena.
+	v := h.curArena.base
+	h.curArena.base = nBase
+	h.growAddSpan(unsafe.Pointer(v), nBase-v)
+	return true
+}
+
+// growAddSpan adds a free span when the heap grows into [v, v+size).
+// This memory must be in the Prepared state (not Ready).
+//
+// h must be locked.
+func (h *mheap) growAddSpan(v unsafe.Pointer, size uintptr) {
 	s := (*mspan)(h.spanalloc.alloc())
 	s.init(uintptr(v), size/pageSize)
 	h.setSpans(s.base(), s.npages, s)
 	s.state = mSpanFree
-	memstats.heap_idle += uint64(size)
-	// (*mheap).sysAlloc returns untouched/uncommitted memory.
+	// [v, v+size) is always in the Prepared state. The new span
+	// must be marked scavenged so the allocator transitions it to
+	// Ready when allocating from it.
 	s.scavenged = true
-	// s is always aligned to the heap arena size which is always > physPageSize,
-	// so its totally safe to just add directly to heap_released. Coalescing,
-	// if possible, will also always be correct in terms of accounting, because
-	// s.base() must be a physical page boundary.
-	memstats.heap_released += uint64(size)
+	// This span is both released and idle, but grow already
+	// updated both memstats.
 	h.coalesce(s)
 	h.free.insert(s)
-	return true
 }
 
 // Free the span back into the heap.