zyppe/anolis-glibc
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
anolis-glibc/glibc-2.28-Refactor-code-and-fix-bug-in-_dl_runtime_.patch
ticat_fp d91eae1237 LoongArch: Sync loongarch64 code to lnd.36 
Signed-off-by: ticat_fp <fanpeng@loongson.cn>
2024-05-29 10:24:08 +08:00

306 lines
9.7 KiB
Diff
Raw Blame History

From 27a004c9777340afd86fc0d129f6ffad508bf090 Mon Sep 17 00:00:00 2001
From: caiyinyu <caiyinyu@loongson.cn>
Date: Tue, 11 Jul 2023 16:09:55 +0800
Subject: [PATCH 12/14] glibc-2.28: Refactor code and fix bug in
_dl_runtime_resolve.
Change-Id: I4907e6643ef25b87d7862e957ce9bf6d201da816
Signed-off-by: ticat_fp <fanpeng@loongson.cn>
---
sysdeps/loongarch/dl-machine.h | 8 +-
sysdeps/loongarch/dl-trampoline.S | 7 ++
sysdeps/loongarch/dl-trampoline.h | 159 +++++++++++++-----------------
sysdeps/loongarch/sys/asm.h | 9 ++
4 files changed, 90 insertions(+), 93 deletions(-)
diff --git a/sysdeps/loongarch/dl-machine.h b/sysdeps/loongarch/dl-machine.h
index 6e9c6258..ff520a07 100644
--- a/sysdeps/loongarch/dl-machine.h
+++ b/sysdeps/loongarch/dl-machine.h
@@ -381,9 +381,13 @@ elf_machine_runtime_setup (struct link_map *l, struct r_scope_elem *scope[],
/* If using PLTs, fill in the first two entries of .got.plt. */
if (l->l_info[DT_JMPREL])
{
- extern void _dl_runtime_resolve (void) __attribute__ ((visibility ("hidden")));
+
+#if HAVE_LOONGARCH_VEC_ASM && !defined __loongarch_soft_float
extern void _dl_runtime_resolve_lasx (void) __attribute__ ((visibility ("hidden")));
extern void _dl_runtime_resolve_lsx (void) __attribute__ ((visibility ("hidden")));
+#endif
+ extern void _dl_runtime_resolve (void) __attribute__ ((visibility ("hidden")));
+
ElfW(Addr) *gotplt = (ElfW(Addr) *) D_PTR (l, l_info[DT_PLTGOT]);
/* If a library is prelinked but we have to relocate anyway,
we have to be able to undo the prelinking of .got.plt.
@@ -391,11 +395,13 @@ elf_machine_runtime_setup (struct link_map *l, struct r_scope_elem *scope[],
if (gotplt[1])
l->l_mach.plt = gotplt[1] + l->l_addr;
+#if HAVE_LOONGARCH_VEC_ASM && !defined __loongarch_soft_float
if (SUPPORT_LASX)
gotplt[0] = (ElfW(Addr)) &_dl_runtime_resolve_lasx;
else if (SUPPORT_LSX)
gotplt[0] = (ElfW(Addr)) &_dl_runtime_resolve_lsx;
else
+#endif
gotplt[0] = (ElfW(Addr)) &_dl_runtime_resolve;
gotplt[1] = (ElfW(Addr)) l;
diff --git a/sysdeps/loongarch/dl-trampoline.S b/sysdeps/loongarch/dl-trampoline.S
index 5f627a63..78d741f3 100644
--- a/sysdeps/loongarch/dl-trampoline.S
+++ b/sysdeps/loongarch/dl-trampoline.S
@@ -16,16 +16,23 @@
License along with the GNU C Library. If not, see
<http://www.gnu.org/licenses/>. */
+#include <sysdep.h>
+#include <sys/asm.h>
+
+#if HAVE_LOONGARCH_VEC_ASM && !defined __loongarch_soft_float
#define USE_LASX
#define _dl_runtime_resolve _dl_runtime_resolve_lasx
#include "dl-trampoline.h"
+#undef FRAME_SIZE
#undef USE_LASX
#undef _dl_runtime_resolve
#define USE_LSX
#define _dl_runtime_resolve _dl_runtime_resolve_lsx
#include "dl-trampoline.h"
+#undef FRAME_SIZE
#undef USE_LSX
#undef _dl_runtime_resolve
+#endif
#include "dl-trampoline.h"
diff --git a/sysdeps/loongarch/dl-trampoline.h b/sysdeps/loongarch/dl-trampoline.h
index 96f41f1d..9a6d9b6c 100644
--- a/sysdeps/loongarch/dl-trampoline.h
+++ b/sysdeps/loongarch/dl-trampoline.h
@@ -17,31 +17,24 @@
License along with the GNU C Library. If not, see
<http://www.gnu.org/licenses/>. */
-#include <sysdep.h>
-#include <sys/asm.h>
-
/* Assembler veneer called from the PLT header code for lazy loading.
The PLT header passes its own args in t0-t2. */
-
-#ifdef __loongarch_soft_float
-# define FRAME_SIZE (-((-10 * SZREG) & ALMASK))
+#ifdef USE_LASX
+# define FRAME_SIZE (-((-9 * SZREG - 8 * SZFREG - 8 * SZXREG) & ALMASK))
+#elif defined USE_LSX
+# define FRAME_SIZE (-((-9 * SZREG - 8 * SZFREG - 8 * SZVREG) & ALMASK))
+#elif !defined __loongarch_soft_float
+# define FRAME_SIZE (-((-9 * SZREG - 8 * SZFREG) & ALMASK))
#else
-# define FRAME_SIZE (-((-10 * SZREG - 8 * 256) & ALMASK))
+# define FRAME_SIZE (-((-9 * SZREG) & ALMASK))
#endif
ENTRY (_dl_runtime_resolve, 3)
- # Save arguments to stack.
-
-#ifdef __loongarch64
- li.d t3, -FRAME_SIZE
- add.d sp, sp, t3
-#elif defined __loongarch32
- li.w t3, -FRAME_SIZE
- add.w sp, sp, t3
-#endif
+ /* Save arguments to stack. */
+ ADDI sp, sp, -FRAME_SIZE
- REG_S ra, sp, 9*SZREG
+ REG_S ra, sp, 0*SZREG
REG_S a0, sp, 1*SZREG
REG_S a1, sp, 2*SZREG
REG_S a2, sp, 3*SZREG
@@ -51,55 +44,45 @@ ENTRY (_dl_runtime_resolve, 3)
REG_S a6, sp, 7*SZREG
REG_S a7, sp, 8*SZREG
-#ifndef __loongarch_soft_float
- FREG_S fa0, sp, 10*SZREG + 0*SZFREG
- FREG_S fa1, sp, 10*SZREG + 1*SZFREG
- FREG_S fa2, sp, 10*SZREG + 2*SZFREG
- FREG_S fa3, sp, 10*SZREG + 3*SZFREG
- FREG_S fa4, sp, 10*SZREG + 4*SZFREG
- FREG_S fa5, sp, 10*SZREG + 5*SZFREG
- FREG_S fa6, sp, 10*SZREG + 6*SZFREG
- FREG_S fa7, sp, 10*SZREG + 7*SZFREG
#ifdef USE_LASX
- xvst xr0, sp, 10*SZREG + 0*256
- xvst xr1, sp, 10*SZREG + 1*256
- xvst xr2, sp, 10*SZREG + 2*256
- xvst xr3, sp, 10*SZREG + 3*256
- xvst xr4, sp, 10*SZREG + 4*256
- xvst xr5, sp, 10*SZREG + 5*256
- xvst xr6, sp, 10*SZREG + 6*256
- xvst xr7, sp, 10*SZREG + 7*256
+ xvst xr0, sp, 9*SZREG + 8*SZFREG + 0*SZXREG
+ xvst xr1, sp, 9*SZREG + 8*SZFREG + 1*SZXREG
+ xvst xr2, sp, 9*SZREG + 8*SZFREG + 2*SZXREG
+ xvst xr3, sp, 9*SZREG + 8*SZFREG + 3*SZXREG
+ xvst xr4, sp, 9*SZREG + 8*SZFREG + 4*SZXREG
+ xvst xr5, sp, 9*SZREG + 8*SZFREG + 5*SZXREG
+ xvst xr6, sp, 9*SZREG + 8*SZFREG + 6*SZXREG
+ xvst xr7, sp, 9*SZREG + 8*SZFREG + 7*SZXREG
#elif defined USE_LSX
- vst vr0, sp, 10*SZREG + 0*128
- vst vr1, sp, 10*SZREG + 1*128
- vst vr2, sp, 10*SZREG + 2*128
- vst vr3, sp, 10*SZREG + 3*128
- vst vr4, sp, 10*SZREG + 4*128
- vst vr5, sp, 10*SZREG + 5*128
- vst vr6, sp, 10*SZREG + 6*128
- vst vr7, sp, 10*SZREG + 7*128
-#endif
+ vst vr0, sp, 9*SZREG + 8*SZFREG + 0*SZVREG
+ vst vr1, sp, 9*SZREG + 8*SZFREG + 1*SZVREG
+ vst vr2, sp, 9*SZREG + 8*SZFREG + 2*SZVREG
+ vst vr3, sp, 9*SZREG + 8*SZFREG + 3*SZVREG
+ vst vr4, sp, 9*SZREG + 8*SZFREG + 4*SZVREG
+ vst vr5, sp, 9*SZREG + 8*SZFREG + 5*SZVREG
+ vst vr6, sp, 9*SZREG + 8*SZFREG + 6*SZVREG
+ vst vr7, sp, 9*SZREG + 8*SZFREG + 7*SZVREG
+#elif !defined __loongarch_soft_float
+ FREG_S fa0, sp, 9*SZREG + 0*SZFREG
+ FREG_S fa1, sp, 9*SZREG + 1*SZFREG
+ FREG_S fa2, sp, 9*SZREG + 2*SZFREG
+ FREG_S fa3, sp, 9*SZREG + 3*SZFREG
+ FREG_S fa4, sp, 9*SZREG + 4*SZFREG
+ FREG_S fa5, sp, 9*SZREG + 5*SZFREG
+ FREG_S fa6, sp, 9*SZREG + 6*SZFREG
+ FREG_S fa7, sp, 9*SZREG + 7*SZFREG
#endif
- # Update .got.plt and obtain runtime address of callee.
-#ifdef __loongarch64
- slli.d a1, t1, 1
+ /* Update .got.plt and obtain runtime address of callee */
+ SLLI a1, t1, 1
or a0, t0, zero
- add.d a1, a1, t1
+ ADD a1, a1, t1
la a2, _dl_fixup
jirl ra, a2, 0
or t1, v0, zero
-#elif defined __loongarch32
- slli.w a1, t1, 1
- or a0, t0, zero
- add.w a1, a1, t1
- la a2, _dl_fixup
- jirl ra, a2, 0
- or t1, v0, zero
-#endif
- # Restore arguments from stack.
- REG_L ra, sp, 9*SZREG
+ /* Restore arguments from stack. */
+ REG_L ra, sp, 0*SZREG
REG_L a0, sp, 1*SZREG
REG_L a1, sp, 2*SZREG
REG_L a2, sp, 3*SZREG
@@ -109,45 +92,37 @@ ENTRY (_dl_runtime_resolve, 3)
REG_L a6, sp, 7*SZREG
REG_L a7, sp, 8*SZREG
-#ifndef __loongarch_soft_float
- FREG_L fa0, sp, 10*SZREG + 0*SZFREG
- FREG_L fa1, sp, 10*SZREG + 1*SZFREG
- FREG_L fa2, sp, 10*SZREG + 2*SZFREG
- FREG_L fa3, sp, 10*SZREG + 3*SZFREG
- FREG_L fa4, sp, 10*SZREG + 4*SZFREG
- FREG_L fa5, sp, 10*SZREG + 5*SZFREG
- FREG_L fa6, sp, 10*SZREG + 6*SZFREG
- FREG_L fa7, sp, 10*SZREG + 7*SZFREG
#ifdef USE_LASX
- xvld xr0, sp, 10*SZREG + 0*256
- xvld xr1, sp, 10*SZREG + 1*256
- xvld xr2, sp, 10*SZREG + 2*256
- xvld xr3, sp, 10*SZREG + 3*256
- xvld xr4, sp, 10*SZREG + 4*256
- xvld xr5, sp, 10*SZREG + 5*256
- xvld xr6, sp, 10*SZREG + 6*256
- xvld xr7, sp, 10*SZREG + 7*256
+ xvld xr0, sp, 9*SZREG + 8*SZFREG + 0*SZXREG
+ xvld xr1, sp, 9*SZREG + 8*SZFREG + 1*SZXREG
+ xvld xr2, sp, 9*SZREG + 8*SZFREG + 2*SZXREG
+ xvld xr3, sp, 9*SZREG + 8*SZFREG + 3*SZXREG
+ xvld xr4, sp, 9*SZREG + 8*SZFREG + 4*SZXREG
+ xvld xr5, sp, 9*SZREG + 8*SZFREG + 5*SZXREG
+ xvld xr6, sp, 9*SZREG + 8*SZFREG + 6*SZXREG
+ xvld xr7, sp, 9*SZREG + 8*SZFREG + 7*SZXREG
#elif defined USE_LSX
- vld vr0, sp, 10*SZREG + 0*128
- vld vr1, sp, 10*SZREG + 1*128
- vld vr2, sp, 10*SZREG + 2*128
- vld vr3, sp, 10*SZREG + 3*128
- vld vr4, sp, 10*SZREG + 4*128
- vld vr5, sp, 10*SZREG + 5*128
- vld vr6, sp, 10*SZREG + 6*128
- vld vr7, sp, 10*SZREG + 7*128
-#endif
-#endif
-
-#ifdef __loongarch64
- li.d t3, FRAME_SIZE
- add.d sp, sp, t3
-#elif defined __loongarch32
- li.w t3, FRAME_SIZE
- addi.w sp, sp, FRAME_SIZE
+ vld vr0, sp, 9*SZREG + 8*SZFREG + 0*SZVREG
+ vld vr1, sp, 9*SZREG + 8*SZFREG + 1*SZVREG
+ vld vr2, sp, 9*SZREG + 8*SZFREG + 2*SZVREG
+ vld vr3, sp, 9*SZREG + 8*SZFREG + 3*SZVREG
+ vld vr4, sp, 9*SZREG + 8*SZFREG + 4*SZVREG
+ vld vr5, sp, 9*SZREG + 8*SZFREG + 5*SZVREG
+ vld vr6, sp, 9*SZREG + 8*SZFREG + 6*SZVREG
+ vld vr7, sp, 9*SZREG + 8*SZFREG + 7*SZVREG
+#elif !defined __loongarch_soft_float
+ FREG_L fa0, sp, 9*SZREG + 0*SZFREG
+ FREG_L fa1, sp, 9*SZREG + 1*SZFREG
+ FREG_L fa2, sp, 9*SZREG + 2*SZFREG
+ FREG_L fa3, sp, 9*SZREG + 3*SZFREG
+ FREG_L fa4, sp, 9*SZREG + 4*SZFREG
+ FREG_L fa5, sp, 9*SZREG + 5*SZFREG
+ FREG_L fa6, sp, 9*SZREG + 6*SZFREG
+ FREG_L fa7, sp, 9*SZREG + 7*SZFREG
#endif
+ ADDI sp, sp, FRAME_SIZE
- # Invoke the callee.
+ /* Invoke the callee. */
jirl zero, t1, 0
END (_dl_runtime_resolve)
diff --git a/sysdeps/loongarch/sys/asm.h b/sysdeps/loongarch/sys/asm.h
index 734e45ae..e80c6245 100644
--- a/sysdeps/loongarch/sys/asm.h
+++ b/sysdeps/loongarch/sys/asm.h
@@ -9,8 +9,17 @@
# define PTRLOG 3
# define SZREG 8
# define SZFREG 8
+# define SZVREG 16
+# define SZXREG 32
# define REG_L ld.d
# define REG_S st.d
+# define SRLI srli.d
+# define SLLI slli.d
+# define ADDI addi.d
+# define ADD add.d
+# define SUB sub.d
+# define BSTRINS bstrins.d
+# define LI li.d
# define FREG_L fld.d
# define FREG_S fst.d
#elif defined __loongarch32
--
2.33.0
Reference in a new issue View git blame Copy permalink