diff options
Diffstat (limited to 'sysdeps/ia64/dl-trampoline.S')
| -rw-r--r-- | sysdeps/ia64/dl-trampoline.S | 538 |
1 files changed, 538 insertions, 0 deletions
diff --git a/sysdeps/ia64/dl-trampoline.S b/sysdeps/ia64/dl-trampoline.S new file mode 100644 index 0000000000..87b684b656 --- /dev/null +++ b/sysdeps/ia64/dl-trampoline.S @@ -0,0 +1,538 @@ +/* PLT trampolines. ia64 version. + Copyright (C) 2005-2014 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> +#undef ret + +/* + This code is used in dl-runtime.c to call the `_dl_fixup' function + and then redirect to the address it returns. `_dl_fixup()' takes two + arguments, however _dl_profile_fixup() takes five. + + The ABI specifies that we will never see more than 8 input + registers to a function call, thus it is safe to simply allocate + those, and simpler than playing stack games. */ + +/* Used to save and restore 8 incoming fp registers */ +#define RESOLVE_FRAME_SIZE (16*8) + +ENTRY(_dl_runtime_resolve) + { .mmi + .prologue + .save ar.pfs, r40 + alloc loc0 = ar.pfs, 8, 6, 2, 0 + /* Use the 16 byte scratch area. r2 will start at f8 and + r3 will start at f9. */ + adds r2 = -(RESOLVE_FRAME_SIZE - 16), r12 + adds r3 = -(RESOLVE_FRAME_SIZE - 32), r12 + } + { .mii + .fframe RESOLVE_FRAME_SIZE + adds r12 = -RESOLVE_FRAME_SIZE, r12 + .save rp, loc1 + mov loc1 = b0 + .body + mov loc2 = r8 /* preserve struct value register */ + ;; + } + { .mii + mov loc3 = r9 /* preserve language specific register */ + mov loc4 = r10 /* preserve language specific register */ + mov loc5 = r11 /* preserve language specific register */ + } + { .mmi + stf.spill [r2] = f8, 32 + stf.spill [r3] = f9, 32 + mov out0 = r16 + ;; + } + { .mmi + stf.spill [r2] = f10, 32 + stf.spill [r3] = f11, 32 + shl out1 = r15, 4 + ;; + } + { .mmi + stf.spill [r2] = f12, 32 + stf.spill [r3] = f13, 32 + /* Relocation record is 24 byte. */ + shladd out1 = r15, 3, out1 + ;; + } + { .mmb + stf.spill [r2] = f14 + stf.spill [r3] = f15 + br.call.sptk.many b0 = _dl_fixup + } + { .mii + /* Skip the 16byte scratch area. */ + adds r2 = 16, r12 + adds r3 = 32, r12 + mov b6 = ret0 + ;; + } + { .mmi + ldf.fill f8 = [r2], 32 + ldf.fill f9 = [r3], 32 + mov b0 = loc1 + ;; + } + { .mmi + ldf.fill f10 = [r2], 32 + ldf.fill f11 = [r3], 32 + mov gp = ret1 + ;; + } + { .mmi + ldf.fill f12 = [r2], 32 + ldf.fill f13 = [r3], 32 + mov ar.pfs = loc0 + ;; + } + { .mmi + ldf.fill f14 = [r2], 32 + ldf.fill f15 = [r3], 32 + .restore sp /* pop the unwind frame state */ + adds r12 = RESOLVE_FRAME_SIZE, r12 + ;; + } + { .mii + mov r9 = loc3 /* restore language specific register */ + mov r10 = loc4 /* restore language specific register */ + mov r11 = loc5 /* restore language specific register */ + } + { .mii + mov r8 = loc2 /* restore struct value register */ + ;; + } + /* An alloc is needed for the break system call to work. + We don't care about the old value of the pfs register. */ + { .mmb + .prologue + .body + alloc r2 = ar.pfs, 0, 0, 8, 0 + br.sptk.many b6 + ;; + } +END(_dl_runtime_resolve) + + +/* The fourth argument to _dl_profile_fixup and the third one to + _dl_call_pltexit are a pointer to La_ia64_regs: + + 8byte r8 + 8byte r9 + 8byte r10 + 8byte r11 + 8byte in0 + 8byte in1 + 8byte in2 + 8byte in3 + 8byte in4 + 8byte in5 + 8byte in6 + 8byte in7 + 16byte f8 + 16byte f9 + 16byte f10 + 16byte f11 + 16byte f12 + 16byte f13 + 16byte f14 + 16byte f15 + 8byte ar.unat + 8byte sp + + The fifth argument to _dl_profile_fixup is a pointer to long int. + The fourth argument to _dl_call_pltexit is a pointer to + La_ia64_retval: + + 8byte r8 + 8byte r9 + 8byte r10 + 8byte r11 + 16byte f8 + 16byte f9 + 16byte f10 + 16byte f11 + 16byte f12 + 16byte f13 + 16byte f14 + 16byte f15 + + Since stack has to be 16 byte aligned, the stack allocation is in + 16byte increment. Before calling _dl_profile_fixup, the stack will + look like + + psp new frame_size + +16 La_ia64_regs + sp scratch + + */ + +#define PLTENTER_FRAME_SIZE (4*8 + 8*8 + 8*16 + 2*8 + 16) +#define PLTEXIT_FRAME_SIZE (PLTENTER_FRAME_SIZE + 4*8 + 8*16) + +#ifndef PROF +ENTRY(_dl_runtime_profile) + { .mii + .prologue + .save ar.pfs, r40 + alloc loc0 = ar.pfs, 8, 12, 8, 0 + .vframe loc10 + mov loc10 = r12 + .save rp, loc1 + mov loc1 = b0 + } + { .mii + .save ar.unat, r17 + mov r17 = ar.unat + .save ar.lc, loc6 + mov loc6 = ar.lc + mov loc11 = gp + } + { .mii + .body + /* There is a 16 byte scratch area. r2 will start at r8 and + r3 will start at r9 for La_ia64_regs. */ + adds r2 = -(PLTENTER_FRAME_SIZE - 16), r12 + adds r3 = -(PLTENTER_FRAME_SIZE - 24), r12 + adds r12 = -PLTENTER_FRAME_SIZE, r12 + ;; + } + { .mmi + st8 [r2] = r8, 16; + st8 [r3] = r9, 16; + mov out2 = b0 /* needed by _dl_fixup_profile */ + ;; + } + { .mmi + st8 [r2] = r10, 16; + st8 [r3] = r11, 16; + adds out3 = 16, r12 /* pointer to La_ia64_regs */ + ;; + } + { .mmi + .mem.offset 0, 0 + st8.spill [r2] = in0, 16 + .mem.offset 8, 0 + st8.spill [r3] = in1, 16 + mov out4 = loc10 /* pointer to new frame size */ + ;; + } + { .mmi + .mem.offset 0, 0 + st8.spill [r2] = in2, 16 + .mem.offset 8, 0 + st8.spill [r3] = in3, 16 + mov loc2 = r8 /* preserve struct value register */ + ;; + } + { .mmi + .mem.offset 0, 0 + st8.spill [r2] = in4, 16 + .mem.offset 8, 0 + st8.spill [r3] = in5, 16 + mov loc3 = r9 /* preserve language specific register */ + ;; + } + { .mmi + .mem.offset 0, 0 + st8 [r2] = in6, 16 + .mem.offset 8, 0 + st8 [r3] = in7, 24 /* adjust for f9 */ + mov loc4 = r10 /* preserve language specific register */ + ;; + } + { .mii + mov r18 = ar.unat /* save it in La_ia64_regs */ + mov loc7 = out3 /* save it for _dl_call_pltexit */ + mov loc5 = r11 /* preserve language specific register */ + } + { .mmi + stf.spill [r2] = f8, 32 + stf.spill [r3] = f9, 32 + mov out0 = r16 /* needed by _dl_fixup_profile */ + ;; + } + { .mii + mov ar.unat = r17 /* restore it for function call */ + mov loc8 = r16 /* save it for _dl_call_pltexit */ + nop.i 0x0 + } + { .mmi + stf.spill [r2] = f10, 32 + stf.spill [r3] = f11, 32 + shl out1 = r15, 4 + ;; + } + { .mmi + stf.spill [r2] = f12, 32 + stf.spill [r3] = f13, 32 + /* Relocation record is 24 byte. */ + shladd out1 = r15, 3, out1 + ;; + } + { .mmi + stf.spill [r2] = f14, 32 + stf.spill [r3] = f15, 24 + mov loc9 = out1 /* save it for _dl_call_pltexit */ + ;; + } + { .mmb + st8 [r2] = r18 /* store ar.unat */ + st8 [r3] = loc10 /* store sp */ + br.call.sptk.many b0 = _dl_profile_fixup + } + { .mii + /* Skip the 16byte scratch area, 4 language specific GRs and + 8 incoming GRs to restore incoming fp registers. */ + adds r2 = (4*8 + 8*8 + 16), r12 + adds r3 = (4*8 + 8*8 + 32), r12 + mov b6 = ret0 + ;; + } + { .mmi + ldf.fill f8 = [r2], 32 + ldf.fill f9 = [r3], 32 + mov gp = ret1 + ;; + } + { .mmi + ldf.fill f10 = [r2], 32 + ldf.fill f11 = [r3], 32 + mov r8 = loc2 /* restore struct value register */ + ;; + } + { .mmi + ldf.fill f12 = [r2], 32 + ldf.fill f13 = [r3], 32 + mov r9 = loc3 /* restore language specific register */ + ;; + } + { .mmi + ldf.fill f14 = [r2], 32 + ldf.fill f15 = [r3], 32 + mov r10 = loc4 /* restore language specific register */ + ;; + } + { .mii + ld8 r15 = [loc10] /* load the new frame size */ + mov r11 = loc5 /* restore language specific register */ + ;; + cmp.eq p6, p7 = -1, r15 + ;; + } + { .mii +(p7) cmp.eq p8, p9 = 0, r15 +(p6) mov b0 = loc1 +(p6) mov ar.lc = loc6 + } + { .mib + nop.m 0x0 +(p6) mov ar.pfs = loc0 +(p6) br.cond.dptk.many .Lresolved + ;; + } + + /* At this point, the stack looks like + + +psp free + +16 La_ia64_regs + sp scratch + + We need to keep the current stack and call the resolved + function by copying the r15 byte from sp + PLTENTER_FRAME_SIZE + + 16 (scratch area) to sp + 16 (scratch area). Since stack + has to be 16byte aligned, we around r15 up to 16byte. */ + + { .mbb +(p9) adds r15 = 15, r15 +(p8) br.cond.dptk.many .Lno_new_frame + nop.b 0x0 + ;; + } + { .mmi + and r15 = -16, r15 + ;; + /* We don't copy the 16byte scatch area. Prepare r16/r17 as + destination. */ + sub r16 = r12, r15 + sub r17 = r12, r15 + ;; + } + { .mii + adds r16 = 16, r16 + adds r17 = 24, r17 + sub r12 = r12, r15 /* Adjust stack */ + ;; + } + { .mii + nop.m 0x0 + shr r15 = r15, 4 + ;; + adds r15 = -1, r15 + ;; + } + { .mii + /* Skip the 16byte scatch area. Prepare r2/r3 as source. */ + adds r2 = 16, loc10 + adds r3 = 24, loc10 + mov ar.lc = r15 + ;; + } +.Lcopy: + { .mmi + ld8 r18 = [r2], 16 + ld8 r19 = [r3], 16 + nop.i 0x0 + ;; + } + { .mmb + st8 [r16] = r18, 16 + st8 [r17] = r19, 16 + br.cloop.sptk.few .Lcopy + } +.Lno_new_frame: + { .mii + mov out0 = in0 + mov out1 = in1 + mov out2 = in2 + } + { .mii + mov out3 = in3 + mov out4 = in4 + mov out5 = in5 + } + { .mib + mov out6 = in6 + mov out7 = in7 + /* Call the resolved function */ + br.call.sptk.many b0 = b6 + } + { .mii + /* Prepare stack for _dl_call_pltexit. Loc10 has the original + stack pointer. */ + adds r12 = -PLTEXIT_FRAME_SIZE, loc10 + adds r2 = -(PLTEXIT_FRAME_SIZE - 16), loc10 + adds r3 = -(PLTEXIT_FRAME_SIZE - 24), loc10 + ;; + } + { .mmi + /* Load all possible return values into buffer. */ + st8 [r2] = r8, 16 + st8 [r3] = r9, 16 + mov out0 = loc8 + ;; + } + { .mmi + st8 [r2] = r10, 16 + st8 [r3] = r11, 24 + mov out1 = loc9 + ;; + } + { .mmi + stf.spill [r2] = f8, 32 + stf.spill [r3] = f9, 32 + mov out2 = loc7 /* Pointer to La_ia64_regs */ + ;; + } + { .mmi + stf.spill [r2] = f10, 32 + stf.spill [r3] = f11, 32 + adds out3 = 16, r12 /* Pointer to La_ia64_retval */ + ;; + } + { .mmi + stf.spill [r2] = f12, 32 + stf.spill [r3] = f13, 32 + /* We need to restore gp for _dl_call_pltexit. */ + mov gp = loc11 + ;; + } + { .mmb + stf.spill [r2] = f14 + stf.spill [r3] = f15 + br.call.sptk.many b0 = _dl_call_pltexit + } + { .mmi + /* Load all the non-floating and floating return values. Skip + the 16byte scratch area. */ + adds r2 = 16, r12 + adds r3 = 24, r12 + nop.i 0x0 + ;; + } + { .mmi + ld8 r8 = [r2], 16 + ld8 r9 = [r3], 16 + nop.i 0x0 + ;; + } + { .mmi + ld8 r10 = [r2], 16 + ld8 r11 = [r3], 24 + nop.i 0x0 + ;; + } + { .mmi + ldf.fill f8 = [r2], 32 + ldf.fill f9 = [r3], 32 + mov ar.lc = loc6 + ;; + } + { .mmi + ldf.fill f10 = [r2], 32 + ldf.fill f11 = [r3], 32 + mov ar.pfs = loc0 + ;; + } + { .mmi + ldf.fill f12 = [r2], 32 + ldf.fill f13 = [r3], 32 + mov b0 = loc1 + ;; + } + { .mmi + ldf.fill f14 = [r2] + ldf.fill f15 = [r3] + /* We know that the previous stack pointer, loc10, isn't 0. + We use it to reload p7. */ + cmp.ne p7, p0 = 0, loc10 + ;; + } +.Lresolved: + { .mmb + .restore sp + mov r12 = loc10 +(p7) br.ret.sptk.many b0 + ;; + } + /* An alloc is needed for the break system call to work. We + don't care about the old value of the pfs register. After + this alloc, we can't use any rotating registers. Otherwise + assembler won't be happy. This has to be at the end. */ + { .mmb + .prologue + .body + alloc r2 = ar.pfs, 0, 0, 8, 0 + br.sptk.many b6 + ;; + } +END(_dl_runtime_profile) +#endif |