glibc/sysdeps/powerpc/powerpc64/setjmp-common.S
Ulrich Weigand 696caf1d00 PowerPC64 ELFv2 ABI 2/6: Remove function descriptors
This patch adds support for the ELFv2 ABI feature to remove function
descriptors.  See this GCC patch for in-depth discussion:
http://gcc.gnu.org/ml/gcc-patches/2013-11/msg01141.html

This mostly involves two types of changes: updating assembler source
files to the new logic, and updating the dynamic loader.

After the refactoring in the previous patch, most of the assembler source
changes can be handled simply by providing ELFv2 versions of the
macros in sysdep.h.   One somewhat non-obvious change is in __GI__setjmp:
this used to "fall through" to the immediately following __setjmp ENTRY
point.  This is no longer safe in the ELFv2 since ENTRY defines both
a global and a local entry point, and you cannot simply fall through
to a global entry point as it requires r12 to be set up.

Also, makecontext needs to be updated to set up registers according to
the new ABI for calling into the context's start routine.

The dynamic linker changes mostly consist of removing special code
to handle function descriptors.  We also need to support the new PLT
and glink format used by the the ELFv2 linker, see:
https://sourceware.org/ml/binutils/2013-10/msg00376.html

In addition, the dynamic linker now verifies that the dynamic libraries
it loads match its own ABI.

The hack in VDSO_IFUNC_RET to "synthesize" a function descriptor
for vDSO routines is also no longer necessary for ELFv2.
2013-12-04 07:41:38 -06:00

231 lines
5.7 KiB
ArmAsm

/* setjmp for PowerPC64.
Copyright (C) 1995-2013 Free Software Foundation, Inc.
This file is part of the GNU C Library.
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, see
<http://www.gnu.org/licenses/>. */
#include <sysdep.h>
#define _ASM
#ifdef __NO_VMX__
#include <novmxsetjmp.h>
#else
#include <jmpbuf-offsets.h>
#endif
#ifndef __NO_VMX__
.section ".toc","aw"
.LC__dl_hwcap:
# ifdef SHARED
# ifdef IS_IN_rtld
/* Inside ld.so we use the local alias to avoid runtime GOT
relocations. */
.tc _rtld_local_ro[TC],_rtld_local_ro
# else
.tc _rtld_global_ro[TC],_rtld_global_ro
# endif
# else
.tc _dl_hwcap[TC],_dl_hwcap
# endif
.section ".text"
#endif
.machine "altivec"
ENTRY (setjmp)
CALL_MCOUNT 1
li r4,1 /* Set second argument to 1. */
b JUMPTARGET (GLUE(__sigsetjmp,_ent))
END (setjmp)
#if defined SHARED && !defined IS_IN_rtld && !defined __NO_VMX__
/* When called from within libc we need a special version of _setjmp
that saves r2 since the call won't go via a plt call stub. See
bugz #269. __GI__setjmp is used in csu/libc-start.c when
HAVE_CLEANUP_JMP_BUF is defined. */
ENTRY (__GI__setjmp)
std r2,40(r1) /* Save the callers TOC in the save area. */
CALL_MCOUNT 1
li r4,0 /* Set second argument to 0. */
b JUMPTARGET (GLUE(__sigsetjmp,_ent))
END (__GI__setjmp)
#endif
ENTRY (_setjmp)
CALL_MCOUNT 1
li r4,0 /* Set second argument to 0. */
b JUMPTARGET (GLUE(__sigsetjmp,_ent))
END (_setjmp)
libc_hidden_def (_setjmp)
ENTRY (__sigsetjmp)
CALL_MCOUNT 2
JUMPTARGET(GLUE(__sigsetjmp,_ent)):
#ifdef PTR_MANGLE
mr r5, r1
PTR_MANGLE (r5, r6)
std r5,(JB_GPR1*8)(3)
#else
std r1,(JB_GPR1*8)(3)
#endif
mflr r0
#if defined SHARED && !defined IS_IN_rtld
ld r5,40(r1) /* Retrieve the callers TOC. */
std r5,(JB_GPR2*8)(3)
#else
std r2,(JB_GPR2*8)(3)
#endif
std r14,((JB_GPRS+0)*8)(3)
stfd fp14,((JB_FPRS+0)*8)(3)
#ifdef PTR_MANGLE
PTR_MANGLE2 (r0, r6)
#endif
std r0,(JB_LR*8)(3)
std r15,((JB_GPRS+1)*8)(3)
stfd fp15,((JB_FPRS+1)*8)(3)
mfcr r0
std r16,((JB_GPRS+2)*8)(3)
stfd fp16,((JB_FPRS+2)*8)(3)
stw r0,((JB_CR*8)+4)(3) /* 32-bit CR. */
std r17,((JB_GPRS+3)*8)(3)
stfd fp17,((JB_FPRS+3)*8)(3)
std r18,((JB_GPRS+4)*8)(3)
stfd fp18,((JB_FPRS+4)*8)(3)
std r19,((JB_GPRS+5)*8)(3)
stfd fp19,((JB_FPRS+5)*8)(3)
std r20,((JB_GPRS+6)*8)(3)
stfd fp20,((JB_FPRS+6)*8)(3)
std r21,((JB_GPRS+7)*8)(3)
stfd fp21,((JB_FPRS+7)*8)(3)
std r22,((JB_GPRS+8)*8)(3)
stfd fp22,((JB_FPRS+8)*8)(3)
std r23,((JB_GPRS+9)*8)(3)
stfd fp23,((JB_FPRS+9)*8)(3)
std r24,((JB_GPRS+10)*8)(3)
stfd fp24,((JB_FPRS+10)*8)(3)
std r25,((JB_GPRS+11)*8)(3)
stfd fp25,((JB_FPRS+11)*8)(3)
std r26,((JB_GPRS+12)*8)(3)
stfd fp26,((JB_FPRS+12)*8)(3)
std r27,((JB_GPRS+13)*8)(3)
stfd fp27,((JB_FPRS+13)*8)(3)
std r28,((JB_GPRS+14)*8)(3)
stfd fp28,((JB_FPRS+14)*8)(3)
std r29,((JB_GPRS+15)*8)(3)
stfd fp29,((JB_FPRS+15)*8)(3)
std r30,((JB_GPRS+16)*8)(3)
stfd fp30,((JB_FPRS+16)*8)(3)
std r31,((JB_GPRS+17)*8)(3)
stfd fp31,((JB_FPRS+17)*8)(3)
#ifndef __NO_VMX__
ld r6,.LC__dl_hwcap@toc(r2)
# ifdef SHARED
/* Load _rtld-global._dl_hwcap. */
ld r6,RTLD_GLOBAL_RO_DL_HWCAP_OFFSET(r6)
# else
ld r6,0(r6) /* Load extern _dl_hwcap. */
# endif
andis. r6,r6,(PPC_FEATURE_HAS_ALTIVEC >> 16)
beq L(no_vmx)
la r5,((JB_VRS)*8)(3)
andi. r6,r5,0xf
mfspr r0,VRSAVE
stw r0,((JB_VRSAVE)*8)(3) /* 32-bit VRSAVE. */
addi r6,r5,16
beq+ L(aligned_save_vmx)
lvsr v0,0,r5
lvsl v1,0,r5
addi r6,r5,-16
# define save_misaligned_vmx(savevr,prevvr,shiftvr,tmpvr,savegpr,addgpr) \
addi addgpr,addgpr,32; \
vperm tmpvr,prevvr,savevr,shiftvr; \
stvx tmpvr,0,savegpr
/*
* We have to be careful not to corrupt the data below v20 and
* above v31. To keep things simple we just rotate both ends in
* the opposite direction to our main permute so we can use
* the common macro.
*/
/* load and rotate data below v20 */
lvx v2,0,r5
vperm v2,v2,v2,v1
save_misaligned_vmx(v20,v2,v0,v3,r5,r6)
save_misaligned_vmx(v21,v20,v0,v3,r6,r5)
save_misaligned_vmx(v22,v21,v0,v3,r5,r6)
save_misaligned_vmx(v23,v22,v0,v3,r6,r5)
save_misaligned_vmx(v24,v23,v0,v3,r5,r6)
save_misaligned_vmx(v25,v24,v0,v3,r6,r5)
save_misaligned_vmx(v26,v25,v0,v3,r5,r6)
save_misaligned_vmx(v27,v26,v0,v3,r6,r5)
save_misaligned_vmx(v28,v27,v0,v3,r5,r6)
save_misaligned_vmx(v29,v28,v0,v3,r6,r5)
save_misaligned_vmx(v30,v29,v0,v3,r5,r6)
save_misaligned_vmx(v31,v30,v0,v3,r6,r5)
/* load and rotate data above v31 */
lvx v2,0,r6
vperm v2,v2,v2,v1
save_misaligned_vmx(v2,v31,v0,v3,r5,r6)
b L(no_vmx)
L(aligned_save_vmx):
stvx 20,0,r5
addi r5,r5,32
stvx 21,0,r6
addi r6,r6,32
stvx 22,0,r5
addi r5,r5,32
stvx 23,0,r6
addi r6,r6,32
stvx 24,0,r5
addi r5,r5,32
stvx 25,0,r6
addi r6,r6,32
stvx 26,0,r5
addi r5,r5,32
stvx 27,0,r6
addi r6,r6,32
stvx 28,0,r5
addi r5,r5,32
stvx 29,0,r6
addi r6,r6,32
stvx 30,0,r5
stvx 31,0,r6
L(no_vmx):
#else
li r6,0
#endif
#if defined NOT_IN_libc && defined IS_IN_rtld
li r3,0
blr
#elif defined SHARED
b JUMPTARGET (__sigjmp_save)
#else
mflr r0
std r0,16(r1)
stdu r1,-112(r1)
cfi_adjust_cfa_offset(112)
cfi_offset(lr,16)
bl JUMPTARGET (__sigjmp_save)
nop
ld r0,112+16(r1)
addi r1,r1,112
mtlr r0
blr
#endif
END (__sigsetjmp)