Unfinished/untested PLT trampoline code.

diff --git a/sysdeps/arm/nacl/dl-trampoline.S b/sysdeps/arm/nacl/dl-trampoline.S
index 052b61ab1f9a503f79f250e3a6473fa3d17d0188..24d4c08df4d2ddf1f4bbb5f80a664971f48ca4e5 100644 (file)
--- a/sysdeps/arm/nacl/dl-trampoline.S
+++ b/sysdeps/arm/nacl/dl-trampoline.S
@@ -1,11 +1,91 @@
-/* XXX temporary stubs */
+/* PLT trampolines.  ARM/NaCl version.
+   Copyright (C) 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>
+
+       .syntax unified
        .text
+
+@ Change &GOT[n+3] into 8*n.  Note relocs are 8 bytes each.
+.macro compute_reloc_arg pltgot, got2
+       sub r1, \pltgot, \got2  @ r1 = &GOT[n+3] - &GOT[2] = 4*(n-1)
+       sub r1, r1, #4          @ r1 = 4*n
+       add r1, r1, r1          @ r1 *= 2 = 8*n
+.endm
+
+       CFI_SECTIONS
        .globl _dl_runtime_resolve
-       .type _dl_runtime_resolve, #function
+       .type _dl_runtime_resolve, %function
        .p2align 4
 _dl_runtime_resolve:
-       sfi_trap
+       cfi_startproc
+       cfi_adjust_cfa_offset (8)
+
+       @ We get called with:
+       @       lr contains the return address from this call
+       @       stack[1] contains &GOT[n+3] (pointer to function)
+       @       stack[0] points to &GOT[2]
+
+       ldr ip, [sp]            @ ip gets &GOT[2]
+
+       @ Save the argument registers and the return address.
+       @ r4 doesn't need to be saved, but it makes the total
+       @ adjustment to sp (including the two words pushed by
+       @ the PLT code) an even eight words, so sp stays aligned.
+       push {r0-r4, lr}
+       cfi_adjust_cfa_offset (24)
+       cfi_rel_offset (r0, 0)
+       cfi_rel_offset (r1, 4)
+       cfi_rel_offset (r2, 8)
+       cfi_rel_offset (r3, 12)
+       cfi_rel_offset (r4, 16)
+       cfi_rel_offset (lr, 20)
+
+       ldr r1, [sp, #28]       @ r1 gets &GOT[n+3]
+
+       @ Get the 'struct link_map *' for first argument to _dl_fixup.
+       sfi_breg ip, ldr r0, [\B, #-4]
+
+       @ Get the reloc offset for the second argument to _dl_fixup.
+       compute_reloc_arg r1, ip
+
+       @ This does the real work, and returns the real call target.
+       sfi_bl _dl_fixup
+       mov ip, r0
+
+       @ Restore the saved registers.
+       pop {r0-r4, lr}
+       cfi_adjust_cfa_offset (-24)
+       cfi_restore (r0)
+       cfi_restore (r1)
+       cfi_restore (r2)
+       cfi_restore (r3)
+       cfi_restore (r4)
+       cfi_restore (lr)
+
+       @ Now compensate for the two words pushed by the PLT code.
+       sfi_sp add sp, #8
+       cfi_adjust_cfa_offset (-8)
+
+       @ Finally, jump to the newfound call target.
+       sfi_bx ip
+
+       cfi_endproc
        .size _dl_runtime_resolve, .-_dl_runtime_resolve
 
 #ifndef PROF
@@ -13,7 +93,11 @@ _dl_runtime_resolve:
        .type _dl_runtime_profile, #function
        .p2align 4
 _dl_runtime_profile:
+       cfi_startproc
+       cfi_adjust_cfa_offset (8)
+       @ XXX tbd
        sfi_trap
+       cfi_endproc
        .size _dl_runtime_profile, .-_dl_runtime_profile
 #endif
        .previous