cmd/internal/obj, cmd/link, runtime: increase stack limit to accommodate larger frames on ppc64x

Larger stack frames mean nosplit functions use more stack and so the limit
needs to increase.

The change to test/nosplit.go is a bit ugly but I can't really think of a
way to make it nicer.

Change-Id: I2616b58015f0b62abbd62951575fcd0d2d8643c2
Reviewed-on: https://go-review.googlesource.com/16504Reviewed-by: Russ Cox <rsc@golang.org>

src/cmd/internal/obj/stack.go

@@ -41,7 +41,7 @@ const (
 	STACKSYSTEM = 0
 	StackSystem = STACKSYSTEM
 	StackBig    = 4096
-	StackGuard  = 640*stackGuardMultiplier + StackSystem
+	StackGuard  = 720*stackGuardMultiplier + StackSystem
 	StackSmall  = 128
 	StackLimit  = StackGuard - StackSystem - StackSmall
 )

src/runtime/stack.go

@@ -90,7 +90,7 @@ const (
 
 	// The stack guard is a pointer this many bytes above the
 	// bottom of the stack.
-	_StackGuard = 640*sys.StackGuardMultiplier + _StackSystem
+	_StackGuard = 720*sys.StackGuardMultiplier + _StackSystem
 
 	// After a stack split check the SP is allowed to be this
 	// many bytes below the stack guard.  This saves an instruction

test/nosplit.go

@@ -115,10 +115,15 @@ main 132
 main 136
 
 # A nosplit leaf can use the whole 128-CallSize bytes available on entry.
-main 112 nosplit
-main 116 nosplit
-main 120 nosplit
-main 124 nosplit
+# (CallSize is 32 on ppc64)
+main 96 nosplit
+main 100 nosplit; REJECT ppc64 ppc64le
+main 104 nosplit; REJECT ppc64 ppc64le
+main 108 nosplit; REJECT ppc64 ppc64le
+main 112 nosplit; REJECT ppc64 ppc64le
+main 116 nosplit; REJECT ppc64 ppc64le
+main 120 nosplit; REJECT ppc64 ppc64le
+main 124 nosplit; REJECT ppc64 ppc64le
 main 128 nosplit; REJECT
 main 132 nosplit; REJECT
 main 136 nosplit; REJECT
@@ -126,11 +131,16 @@ main 136 nosplit; REJECT
 # Calling a nosplit function from a nosplit function requires
 # having room for the saved caller PC and the called frame.
 # Because ARM doesn't save LR in the leaf, it gets an extra 4 bytes.
-# Because ppc64 doesn't save LR in the leaf, it gets an extra 8 bytes.
-main 112 nosplit call f; f 0 nosplit
-main 116 nosplit call f; f 0 nosplit
-main 120 nosplit call f; f 0 nosplit; REJECT amd64
-main 124 nosplit call f; f 0 nosplit; REJECT amd64 386
+# Because arm64 doesn't save LR in the leaf, it gets an extra 8 bytes.
+# ppc64 doesn't save LR in the leaf, but CallSize is 32, so it gets 24 fewer bytes than amd64.
+main 96 nosplit call f; f 0 nosplit
+main 100 nosplit call f; f 0 nosplit; REJECT ppc64 ppc64le
+main 104 nosplit call f; f 0 nosplit; REJECT ppc64 ppc64le
+main 108 nosplit call f; f 0 nosplit; REJECT ppc64 ppc64le
+main 112 nosplit call f; f 0 nosplit; REJECT ppc64 ppc64le
+main 116 nosplit call f; f 0 nosplit; REJECT ppc64 ppc64le
+main 120 nosplit call f; f 0 nosplit; REJECT ppc64 ppc64le amd64
+main 124 nosplit call f; f 0 nosplit; REJECT ppc64 ppc64le amd64 386
 main 128 nosplit call f; f 0 nosplit; REJECT
 main 132 nosplit call f; f 0 nosplit; REJECT
 main 136 nosplit call f; f 0 nosplit; REJECT
@@ -138,24 +148,28 @@ main 136 nosplit call f; f 0 nosplit; REJECT
 # Calling a splitting function from a nosplit function requires
 # having room for the saved caller PC of the call but also the
 # saved caller PC for the call to morestack.
-# Again the ARM and ppc64 work in less space.
-main 104 nosplit call f; f 0 call f
-main 108 nosplit call f; f 0 call f
-main 112 nosplit call f; f 0 call f; REJECT amd64
-main 116 nosplit call f; f 0 call f; REJECT amd64
-main 120 nosplit call f; f 0 call f; REJECT amd64 386
-main 124 nosplit call f; f 0 call f; REJECT amd64 386
+# RISC architectures differ in the same way as before.
+main 96 nosplit call f; f 0 call f
+main 100 nosplit call f; f 0 call f; REJECT ppc64 ppc64le
+main 104 nosplit call f; f 0 call f; REJECT ppc64 ppc64le
+main 108 nosplit call f; f 0 call f; REJECT ppc64 ppc64le
+main 112 nosplit call f; f 0 call f; REJECT ppc64 ppc64le amd64
+main 116 nosplit call f; f 0 call f; REJECT ppc64 ppc64le amd64
+main 120 nosplit call f; f 0 call f; REJECT ppc64 ppc64le amd64 386
+main 124 nosplit call f; f 0 call f; REJECT ppc64 ppc64le amd64 386
 main 128 nosplit call f; f 0 call f; REJECT
 main 132 nosplit call f; f 0 call f; REJECT
 main 136 nosplit call f; f 0 call f; REJECT
 
 # Indirect calls are assumed to be splitting functions.
-main 104 nosplit callind
-main 108 nosplit callind
-main 112 nosplit callind; REJECT amd64
-main 116 nosplit callind; REJECT amd64
-main 120 nosplit callind; REJECT amd64 386
-main 124 nosplit callind; REJECT amd64 386
+main 96 nosplit callind
+main 100 nosplit callind; REJECT ppc64 ppc64le
+main 104 nosplit callind; REJECT ppc64 ppc64le
+main 108 nosplit callind; REJECT ppc64 ppc64le
+main 112 nosplit callind; REJECT ppc64 ppc64le amd64
+main 116 nosplit callind; REJECT ppc64 ppc64le amd64
+main 120 nosplit callind; REJECT ppc64 ppc64le amd64 386
+main 124 nosplit callind; REJECT ppc64 ppc64le amd64 386
 main 128 nosplit callind; REJECT
 main 132 nosplit callind; REJECT
 main 136 nosplit callind; REJECT
@@ -284,16 +298,16 @@ TestCases:
 				name := m[1]
 				size, _ := strconv.Atoi(m[2])
 
-				// The limit was originally 128 but is now 512.
+				// The limit was originally 128 but is now 592.
 				// Instead of rewriting the test cases above, adjust
 				// the first stack frame to use up the extra bytes.
 				if i == 0 {
-					size += 512 - 128
+					size += 592 - 128
 					// Noopt builds have a larger stackguard.
 					// See ../cmd/dist/buildruntime.go:stackGuardMultiplier
 					for _, s := range strings.Split(os.Getenv("GO_GCFLAGS"), " ") {
 						if s == "-N" {
-							size += 640
+							size += 720
 						}
 					}
 				}