ppc64: fix setjmp/longjmp handling of TOC pointer

The TOC pointer is constant within a single dso, but needs to be saved
and restored around cross-dso calls.  The PLT stub saves it to the
caller's stack frame, and the linker adds code to the caller to restore
it.

With a local call, as within a single dso or with static linking, this
doesn't happen and the TOC pointer is always in r2.  Therefore,
setjmp/longjmp need to save/restore the TOC pointer from/to different
locations depending on whether the call to setjmp was a local or non-local
call.

It is always safe for longjmp to restore to both r2 and the caller's stack.
If the call to setjmp was local, and only r2 matters and the stack location
will be ignored, but is required by the ABI to be reserved for the TOC
pointer.  If the call was non-local, then only the stack location matters,
and whatever is restored into r2 will be clobbered anyway when the caller
reloads r2 from the stack.

A little extra care is required for sigsetjmp, because it uses setjmp
internally.  After the second return from this setjmp call, r2 will contain
the caller's TOC pointer instead of libc's TOC pointer.  We need to save
and restore the correct libc pointer before we can tail call to
__sigsetjmp_tail.

src/setjmp/powerpc64/longjmp.s

@@ -10,10 +10,14 @@ longjmp:
 	# 1) restore cr
 	ld   0,  1*8(3)
 	mtcr 0
-	# 2) restore r1-r2 (SP and TOC)
+	# 2) restore SP
 	ld   1,  2*8(3)
+	# 3) restore TOC into both r2 and the caller's stack.
+	#    Which location is required depends on whether setjmp was called
+	#    locally or non-locally, but it's always safe to restore to both.
 	ld   2,  3*8(3)
-	# 3) restore r14-r31
+	std  2,   24(1)
+	# 4) restore r14-r31
 	ld  14,  4*8(3)
 	ld  15,  5*8(3)
 	ld  16,  6*8(3)
@@ -32,7 +36,7 @@ longjmp:
 	ld  29, 19*8(3)
 	ld  30, 20*8(3)
 	ld  31, 21*8(3)
-	# 4) restore floating point registers f14-f31
+	# 5) restore floating point registers f14-f31
 	lfd 14, 22*8(3)
 	lfd 15, 23*8(3)
 	lfd 16, 24*8(3)
@@ -52,7 +56,7 @@ longjmp:
 	lfd 30, 38*8(3)
 	lfd 31, 39*8(3)
 
-	# 5) restore vector registers v20-v31
+	# 6) restore vector registers v20-v31
 	addi 3, 3, 40*8
 	lvx 20, 0, 3 ; addi 3, 3, 16
 	lvx 21, 0, 3 ; addi 3, 3, 16
@@ -67,7 +71,7 @@ longjmp:
 	lvx 30, 0, 3 ; addi 3, 3, 16
 	lvx 31, 0, 3
 
-	# 6) return r4 ? r4 : 1
+	# 7) return r4 ? r4 : 1
 	mr    3,   4
 	cmpwi cr7, 4, 0
 	bne   cr7, 1f

src/setjmp/powerpc64/setjmp.s

@@ -1,24 +1,35 @@
-	.global ___setjmp
-	.hidden ___setjmp
 	.global __setjmp
 	.global _setjmp
 	.global setjmp
 	.type   __setjmp,@function
 	.type   _setjmp,@function
 	.type   setjmp,@function
-___setjmp:
 __setjmp:
 _setjmp:
 setjmp:
+	ld 5, 24(1)   # load from the TOC slot in the caller's stack frame
+	b __setjmp_toc
+
+	.localentry __setjmp,.-__setjmp
+	.localentry _setjmp,.-_setjmp
+	.localentry setjmp,.-setjmp
+	mr 5, 2
+
+	.global __setjmp_toc
+	.hidden __setjmp_toc
+	# same as normal setjmp, except TOC pointer to save is provided in r5.
+	# r4 would normally be the 2nd parameter, but we're using r5 to simplify calling from sigsetjmp.
+	# solves the problem of knowing whether to save the TOC pointer from r2 or the caller's stack frame.
+__setjmp_toc:
 	# 0) store IP into 0, then into the jmpbuf pointed to by r3 (first arg)
 	mflr  0
 	std   0,  0*8(3)
 	# 1) store cr
 	mfcr  0
 	std   0,  1*8(3)
-	# 2) store r1-r2 (SP and TOC)
+	# 2) store SP and TOC
 	std   1,  2*8(3)
-	std   2,  3*8(3)
+	std   5,  3*8(3)
 	# 3) store r14-31
 	std  14,  4*8(3)
 	std  15,  5*8(3)

src/signal/powerpc64/sigsetjmp.s

@@ -2,29 +2,36 @@
 	.global __sigsetjmp
 	.type sigsetjmp,%function
 	.type __sigsetjmp,%function
-	.hidden ___setjmp
+	.hidden __setjmp_toc
 sigsetjmp:
 __sigsetjmp:
 	addis 2, 12, .TOC.-__sigsetjmp@ha
 	addi  2,  2, .TOC.-__sigsetjmp@l
+	ld    5, 24(1)   # load from the TOC slot in the caller's stack frame
+	b     1f
+
 	.localentry sigsetjmp,.-sigsetjmp
 	.localentry __sigsetjmp,.-__sigsetjmp
+	mr    5,  2
 
+1:
 	cmpwi cr7, 4, 0
-	beq-  cr7, ___setjmp
+	beq-  cr7, __setjmp_toc
 
-	mflr  5
-	std   5, 512(3)
-	std  16, 512+8+8(3)
+	mflr  6
+	std   6, 512(3)
+	std   2, 512+16(3)
+	std  16, 512+24(3)
 	mr   16, 3
 
-	bl ___setjmp
+	bl __setjmp_toc
 
 	mr   4,  3
 	mr   3, 16
 	ld   5, 512(3)
 	mtlr 5
-	ld  16, 512+8+8(3)
+	ld   2, 512+16(3)
+	ld  16, 512+24(3)
 
 .hidden __sigsetjmp_tail
 	b __sigsetjmp_tail