File dietlibc-0.30-setjmp-longjmpg.diff of Package dietlibc

--- ia64/__longjmp.S
+++ ia64/__longjmp.S
@@ -1 +1,159 @@
-/* No longjmp yet */
+/* Copyright (C) 1999, 2000, 2001 Free Software Foundation, Inc.
+   Contributed by David Mosberger-Tang <davidm@hpl.hp.com>.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.
+
+   Note that __sigsetjmp() did NOT flush the register stack.  Instead,
+   we do it here since __longjmp() is usually much less frequently
+   invoked than __sigsetjmp(). The only difficulty is that __sigsetjmp()
+   didn't (and wouldn't be able to) save ar.rnat either.  This is a problem
+   because if we're not careful, we could end up loading random NaT bits.
+   There are two cases:
+
+	(i)  ar.bsp < ia64_rse_rnat_addr(jmpbuf.ar_bsp)
+		ar.rnat contains the desired bits---preserve ar.rnat
+		across loadrs and write to ar.bspstore
+
+	(ii) ar.bsp >= ia64_rse_rnat_addr(jmpbuf.ar_bsp)
+		The desired ar.rnat is stored in
+		ia64_rse_rnat_addr(jmpbuf.ar_bsp).  Load those
+		bits into ar.rnat after setting ar.bspstore. */
+
+#include "syscalls.h"
+
+#	define	pPos	p6	/* is rotate count positive? */
+#	define	pNeg	p7	/* is rotate count negative? */
+
+
+	/* __longjmp(__jmp_buf buf, int val) */
+
+LEAF(__longjmp)
+	alloc r8=ar.pfs,2,1,0,0
+	mov r27=ar.rsc
+	add r2=0x98,in0		// r2 <- &jmpbuf.orig_jmp_buf_addr
+	;;
+	ld8 r8=[r2],-16		// r8 <- orig_jmp_buf_addr
+	mov r10=ar.bsp
+	and r11=~0x3,r27	// clear ar.rsc.mode
+	;;
+	flushrs			// flush dirty regs to backing store (must be first in insn grp)
+	ld8 r23=[r2],8		// r23 <- jmpbuf.ar_bsp
+	sub r8=r8,in0		// r8 <- &orig_jmpbuf - &jmpbuf
+	;;
+	ld8 r25=[r2]		// r25 <- jmpbuf.ar_unat
+	extr.u r8=r8,3,6	// r8 <- (&orig_jmpbuf - &jmpbuf)/8 & 0x3f
+	;;
+	cmp.lt pNeg,pPos=r8,r0
+	mov r2=in0
+	;;
+(pPos)	mov r16=r8
+(pNeg)	add r16=64,r8
+(pPos)	sub r17=64,r8
+(pNeg)	sub r17=r0,r8
+	;;
+	mov ar.rsc=r11		// put RSE in enforced lazy mode
+	shr.u r8=r25,r16
+	add r3=8,in0		// r3 <- &jmpbuf.r1
+	shl r9=r25,r17
+	;;
+	or r25=r8,r9
+	;;
+	mov r26=ar.rnat
+	mov ar.unat=r25		// setup ar.unat (NaT bits for r1, r4-r7, and r12)
+	;;
+	ld8.fill.nta sp=[r2],16	// r12 (sp)
+	ld8.fill.nta gp=[r3],16		// r1 (gp)
+	dep r11=-1,r23,3,6	// r11 <- ia64_rse_rnat_addr(jmpbuf.ar_bsp)
+	;;
+	ld8.nta r16=[r2],16		// caller's unat
+	ld8.nta r17=[r3],16		// fpsr
+	;;
+	ld8.fill.nta r4=[r2],16	// r4
+	ld8.fill.nta r5=[r3],16		// r5 (gp)
+	cmp.geu p8,p0=r10,r11	// p8 <- (ar.bsp >= jmpbuf.ar_bsp)
+	;;
+	ld8.fill.nta r6=[r2],16	// r6
+	ld8.fill.nta r7=[r3],16		// r7
+	;;
+	mov ar.unat=r16			// restore caller's unat
+	mov ar.fpsr=r17			// restore fpsr
+	;;
+	ld8.nta r16=[r2],16		// b0
+	ld8.nta r17=[r3],16		// b1
+	;;
+(p8)	ld8 r26=[r11]		// r26 <- *ia64_rse_rnat_addr(jmpbuf.ar_bsp)
+	mov ar.bspstore=r23	// restore ar.bspstore
+	;;
+	ld8.nta r18=[r2],16		// b2
+	ld8.nta r19=[r3],16		// b3
+	;;
+	ld8.nta r20=[r2],16		// b4
+	ld8.nta r21=[r3],16		// b5
+	;;
+	ld8.nta r11=[r2],16		// ar.pfs
+	ld8.nta r22=[r3],56		// ar.lc
+	;;
+	ld8.nta r24=[r2],32		// pr
+	mov b0=r16
+	;;
+	ldf.fill.nta f2=[r2],32
+	ldf.fill.nta f3=[r3],32
+	mov b1=r17
+	;;
+	ldf.fill.nta f4=[r2],32
+	ldf.fill.nta f5=[r3],32
+	mov b2=r18
+	;;
+	ldf.fill.nta f16=[r2],32
+	ldf.fill.nta f17=[r3],32
+	mov b3=r19
+	;;
+	ldf.fill.nta f18=[r2],32
+	ldf.fill.nta f19=[r3],32
+	mov b4=r20
+	;;
+	ldf.fill.nta f20=[r2],32
+	ldf.fill.nta f21=[r3],32
+	mov b5=r21
+	;;
+	ldf.fill.nta f22=[r2],32
+	ldf.fill.nta f23=[r3],32
+	mov ar.lc=r22
+	;;
+	ldf.fill.nta f24=[r2],32
+	ldf.fill.nta f25=[r3],32
+	cmp.eq p8,p9=0,in1
+	;;
+	ldf.fill.nta f26=[r2],32
+	ldf.fill.nta f27=[r3],32
+	mov ar.pfs=r11
+	;;
+	ldf.fill.nta f28=[r2],32
+	ldf.fill.nta f29=[r3],32
+	;;
+	ldf.fill.nta f30=[r2]
+	ldf.fill.nta f31=[r3]
+(p8)	mov r8=1
+
+	mov ar.rnat=r26		// restore ar.rnat
+	;;
+	mov ar.rsc=r27		// restore ar.rsc
+(p9)	mov r8=in1
+
+	invala			// virt. -> phys. regnum mapping may change
+	mov pr=r24,-1
+	br.ret.sptk.few b0
+END(__longjmp)
--- ia64/setjmp.S
+++ ia64/setjmp.S
@@ -1 +1,177 @@
-/* No, I am not going to save 128 registers */
+/* Copyright (C) 1999, 2000, 2001, 2002 Free Software Foundation, Inc.
+   Contributed by David Mosberger-Tang <davidm@hpl.hp.com>.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.
+
+   The layout of the jmp_buf is as follows.  This is subject to change
+   and user-code should never depend on the particular layout of
+   jmp_buf!
+
+
+  	offset:	description:
+	-------	------------
+  	0x000	stack pointer (r12)	; unchangeable (see _JMPBUF_UNWINDS)
+  	0x008	r1 (gp)
+	0x010	caller's unat
+	0x018	fpsr
+  	0x020	r4
+  	0x028	r5
+  	0x030	r6
+  	0x038	r7
+  	0x040	rp (b0)
+  	0x048	b1
+  	0x050	b2
+  	0x058	b3
+  	0x060	b4
+  	0x068	b5
+  	0x070	ar.pfs
+  	0x078	ar.lc
+  	0x080	pr
+  	0x088	ar.bsp			; unchangeable (see __longjmp.S)
+  	0x090	ar.unat
+	0x098	&__jmp_buf	; address of the jmpbuf (needed to locate NaT bits in unat)
+	0x0a0	 f2
+	0x0b0	 f3
+	0x0c0	 f4
+	0x0d0	 f5
+	0x0e0	f16
+  	0x0f0	f17
+  	0x100	f18
+  	0x110	f19
+  	0x120	f20
+  	0x130	f21
+  	0x130	f22
+  	0x140	f23
+  	0x150	f24
+  	0x160	f25
+  	0x170	f26
+  	0x180	f27
+  	0x190	f28
+  	0x1a0	f29
+  	0x1b0	f30
+  	0x1c0	f31 */
+
+#include "syscalls.h"
+
+	/* The following two entry points are the traditional entry points: */
+
+LEAF(setjmp)
+	alloc r8=ar.pfs,2,0,0,0
+	mov in1=1
+	br.cond.sptk.many __sigsetjmp
+END(setjmp)
+
+LEAF(_setjmp)
+	alloc r8=ar.pfs,2,0,0,0
+	mov in1=0
+	br.cond.sptk.many __sigsetjmp
+END(_setjmp)
+
+	/* __sigsetjmp(__jmp_buf buf, int savemask) */
+
+ENTRY(__sigsetjmp)
+	.prologue ASM_UNW_PRLG_RP|ASM_UNW_PRLG_PFS, ASM_UNW_PRLG_GRSAVE(2)
+	alloc loc1=ar.pfs,2,3,2,0
+	.save ar.unat, loc2
+	mov loc2=ar.unat
+	;;
+	mov r17=ar.fpsr
+	mov r2=in0
+	add r3=8,in0
+	;;
+.mem.offset 8,0;	st8.spill.nta [r2]=sp,16	// r12 (sp)
+.mem.offset 0,0;	st8.spill.nta [r3]=gp,16	// r1 (gp)
+	;;
+	st8.nta [r2]=loc2,16		// save caller's unat
+	st8.nta [r3]=r17,16		// save fpsr
+	add r8=0xa0,in0
+	;;
+.mem.offset 8,0;	st8.spill.nta [r2]=r4,16	// r4
+.mem.offset 0,0;	st8.spill.nta [r3]=r5,16	// r5
+	add r9=0xb0,in0
+	;;
+	stf.spill.nta [r8]=f2,32
+	stf.spill.nta [r9]=f3,32
+	mov loc0=rp
+	.body
+	;;
+	stf.spill.nta [r8]=f4,32
+	stf.spill.nta [r9]=f5,32
+	mov r17=b1
+	;;
+	stf.spill.nta [r8]=f16,32
+	stf.spill.nta [r9]=f17,32
+	mov r18=b2
+	;;
+	stf.spill.nta [r8]=f18,32
+	stf.spill.nta [r9]=f19,32
+	mov r19=b3
+	;;
+	stf.spill.nta [r8]=f20,32
+	stf.spill.nta [r9]=f21,32
+	mov r20=b4
+	;;
+	stf.spill.nta [r8]=f22,32
+	stf.spill.nta [r9]=f23,32
+	mov r21=b5
+	;;
+	stf.spill.nta [r8]=f24,32
+	stf.spill.nta [r9]=f25,32
+	mov r22=ar.lc
+	;;
+	stf.spill.nta [r8]=f26,32
+	stf.spill.nta [r9]=f27,32
+	mov r24=pr
+	;;
+	stf.spill.nta [r8]=f28,32
+	stf.spill.nta [r9]=f29,32
+	;;
+	stf.spill.nta [r8]=f30
+	stf.spill.nta [r9]=f31
+
+.mem.offset 8,0;	st8.spill.nta [r2]=r6,16	// r6
+.mem.offset 0,0;	st8.spill.nta [r3]=r7,16	// r7
+	;;
+	mov r23=ar.bsp
+	mov r25=ar.unat
+	mov out0=in0
+
+	st8.nta [r2]=loc0,16		// b0
+	st8.nta [r3]=r17,16		// b1
+	mov out1=in1
+	;;
+	st8.nta [r2]=r18,16		// b2
+	st8.nta [r3]=r19,16		// b3
+	;;
+	st8.nta [r2]=r20,16		// b4
+	st8.nta [r3]=r21,16		// b5
+	;;
+	st8.nta [r2]=loc1,16		// ar.pfs
+	st8.nta [r3]=r22,16		// ar.lc
+	;;
+	st8.nta [r2]=r24,16		// pr
+	st8.nta [r3]=r23,16		// ar.bsp
+	;;
+	st8.nta [r2]=r25		// ar.unat
+	st8.nta [r3]=in0		// &__jmp_buf
+	br.call.dpnt.few rp=__sigjmp_save
+.ret0:					// force a new bundle ::q
+	mov.m ar.unat=loc2		// restore caller's unat
+	mov rp=loc0
+	mov ar.pfs=loc1
+	mov r8=0
+	br.ret.sptk.few b0
+END(__sigsetjmp)
--- ia64/syscalls.h
+++ ia64/syscalls.h
@@ -236,18 +236,38 @@
 
 #define syscall(name, sym) \
 .text; \
+.proc sym; \
 .globl sym; \
 sym: \
 	mov r15 = __NR_##name; \
-	br __unified_syscall;
+	br __unified_syscall; \
+.size sym,.-sym; \
+.endp sym;
 
-#define syscall_weak(name, sym, wsym) \
+#define syscall_weak(name, wsym, sym) \
 .text; \
+.proc sym; \
+.type wsym,@function; \
 .weak wsym; \
 wsym: \
 .globl sym; \
 sym: \
 	mov r15 = __NR_##name; \
-	br __unified_syscall;
+	br __unified_syscall; \
+.size sym,.-sym; \
+.endp sym;
+
+/* Macros to help writing .prologue directives in assembly code.  */
+#define ASM_UNW_PRLG_RP			0x8
+#define ASM_UNW_PRLG_PFS		0x4
+#define ASM_UNW_PRLG_PSP		0x2
+#define ASM_UNW_PRLG_PR			0x1
+#define ASM_UNW_PRLG_GRSAVE(ninputs)	(32+(ninputs))
+
+#define ENTRY(sym) \
+	.text; .proc sym; .globl sym; sym:
+#define LEAF(sym) ENTRY(sym)
+#define END(sym) \
+	.size sym,.-sym; .endp sym
 
 #endif