[lld] r350681 - [ARM][AArch64] Increase TLS alignment to reserve space for Android's TCB
Ryan Prichard via llvm-commits
llvm-commits at lists.llvm.org
Tue Jan 8 16:09:59 PST 2019
Author: rprichard
Date: Tue Jan 8 16:09:59 2019
New Revision: 350681
URL: http://llvm.org/viewvc/llvm-project?rev=350681&view=rev
Log:
[ARM][AArch64] Increase TLS alignment to reserve space for Android's TCB
ARM and AArch64 use TLS variant 1, where the first two words after the
thread pointer are reserved for the TCB, followed by the executable's TLS
segment. Both the thread pointer and the TLS segment are aligned to at
least the TLS segment's alignment.
Android/Bionic historically has not supported ELF TLS, and it has
allocated memory after the thread pointer for several Bionic TLS slots
(currently 9 but soon only 8). At least one of these allocations
(TLS_SLOT_STACK_GUARD == 5) is widespread throughout Android/AArch64
binaries and can't be changed.
To reconcile this disagreement about TLS memory layout, set the minimum
alignment for executable TLS segments to 8 words on ARM/AArch64, which
reserves at least 8 words of memory after the TP (2 for the ABI-specified
TCB and 6 for alignment padding). For simplicity, and because lld doesn't
know when it's targeting Android, increase the alignment regardless of
operating system.
Differential Revision: https://reviews.llvm.org/D53906
Modified:
lld/trunk/ELF/InputSection.cpp
lld/trunk/ELF/Writer.cpp
lld/trunk/test/ELF/aarch64-cortex-a53-843419-tlsrelax.s
lld/trunk/test/ELF/aarch64-tls-gdle.s
lld/trunk/test/ELF/aarch64-tls-iele.s
lld/trunk/test/ELF/aarch64-tls-le.s
lld/trunk/test/ELF/aarch64-tlsld-ldst.s
lld/trunk/test/ELF/arm-tls-le32.s
lld/trunk/test/ELF/arm-tls-norelax-ie-le.s
Modified: lld/trunk/ELF/InputSection.cpp
URL: http://llvm.org/viewvc/llvm-project/lld/trunk/ELF/InputSection.cpp?rev=350681&r1=350680&r2=350681&view=diff
==============================================================================
--- lld/trunk/ELF/InputSection.cpp (original)
+++ lld/trunk/ELF/InputSection.cpp Tue Jan 8 16:09:59 2019
@@ -575,6 +575,10 @@ static int64_t getTlsTpOffset() {
// Variant 1. The thread pointer points to a TCB with a fixed 2-word size,
// followed by a variable amount of alignment padding, followed by the TLS
// segment.
+ //
+ // NB: While the ARM/AArch64 ABI formally has a 2-word TCB size, lld
+ // effectively increases the TCB size to 8 words for Android compatibility.
+ // It accomplishes this by increasing the segment's alignment.
return alignTo(Config->Wordsize * 2, Out::TlsPhdr->p_align);
case EM_386:
case EM_X86_64:
Modified: lld/trunk/ELF/Writer.cpp
URL: http://llvm.org/viewvc/llvm-project/lld/trunk/ELF/Writer.cpp?rev=350681&r1=350680&r2=350681&view=diff
==============================================================================
--- lld/trunk/ELF/Writer.cpp (original)
+++ lld/trunk/ELF/Writer.cpp Tue Jan 8 16:09:59 2019
@@ -2181,11 +2181,23 @@ template <class ELFT> void Writer<ELFT>:
P->p_memsz = alignTo(P->p_memsz, Target->PageSize);
}
- // The TLS pointer goes after PT_TLS for variant 2 targets. At least glibc
- // will align it, so round up the size to make sure the offsets are
- // correct.
- if (P->p_type == PT_TLS && P->p_memsz)
+ if (P->p_type == PT_TLS && P->p_memsz) {
+ if (!Config->Shared &&
+ (Config->EMachine == EM_ARM || Config->EMachine == EM_AARCH64)) {
+ // On ARM/AArch64, reserve extra space (8 words) between the thread
+ // pointer and an executable's TLS segment by overaligning the segment.
+ // This reservation is needed for backwards compatibility with Android's
+ // TCB, which allocates several slots after the thread pointer (e.g.
+ // TLS_SLOT_STACK_GUARD==5). For simplicity, this overalignment is also
+ // done on other operating systems.
+ P->p_align = std::max<uint64_t>(P->p_align, Config->Wordsize * 8);
+ }
+
+ // The TLS pointer goes after PT_TLS for variant 2 targets. At least glibc
+ // will align it, so round up the size to make sure the offsets are
+ // correct.
P->p_memsz = alignTo(P->p_memsz, P->p_align);
+ }
}
}
Modified: lld/trunk/test/ELF/aarch64-cortex-a53-843419-tlsrelax.s
URL: http://llvm.org/viewvc/llvm-project/lld/trunk/test/ELF/aarch64-cortex-a53-843419-tlsrelax.s?rev=350681&r1=350680&r2=350681&view=diff
==============================================================================
--- lld/trunk/test/ELF/aarch64-cortex-a53-843419-tlsrelax.s (original)
+++ lld/trunk/test/ELF/aarch64-cortex-a53-843419-tlsrelax.s Tue Jan 8 16:09:59 2019
@@ -26,9 +26,9 @@ _start:
// CHECK: _start:
// CHECK-NEXT: 210ff8: 41 d0 3b d5 mrs x1, TPIDR_EL0
// CHECK-NEXT: 210ffc: 00 00 a0 d2 movz x0, #0, lsl #16
-// CHECK-NEXT: 211000: 01 02 80 f2 movk x1, #16
+// CHECK-NEXT: 211000: 01 08 80 f2 movk x1, #64
// CHECK-NEXT: 211004: 00 00 a0 d2 movz x0, #0, lsl #16
-// CHECK-NEXT: 211008: 01 02 80 f2 movk x1, #16
+// CHECK-NEXT: 211008: 01 08 80 f2 movk x1, #64
// CHECK-NEXT: 21100c: c0 03 5f d6 ret
.type v, at object
Modified: lld/trunk/test/ELF/aarch64-tls-gdle.s
URL: http://llvm.org/viewvc/llvm-project/lld/trunk/test/ELF/aarch64-tls-gdle.s?rev=350681&r1=350680&r2=350681&view=diff
==============================================================================
--- lld/trunk/test/ELF/aarch64-tls-gdle.s (original)
+++ lld/trunk/test/ELF/aarch64-tls-gdle.s Tue Jan 8 16:09:59 2019
@@ -5,15 +5,15 @@
# RUN: llvm-objdump -d %tout | FileCheck %s
# RUN: llvm-readobj -s -r %tout | FileCheck -check-prefix=RELOC %s
-#Local-Dynamic to Initial-Exec relax creates no
+#Local-Dynamic to Local-Exec relax creates no
#RELOC: Relocations [
#RELOC-NEXT: ]
-# TCB size = 0x16 and foo is first element from TLS register.
+# TCB size = 64 and foo is first element from TLS register.
# CHECK: Disassembly of section .text:
# CHECK: _start:
# CHECK: 210000: 00 00 a0 d2 movz x0, #0, lsl #16
-# CHECK: 210004: 00 02 80 f2 movk x0, #16
+# CHECK: 210004: 00 08 80 f2 movk x0, #64
# CHECK: 210008: 1f 20 03 d5 nop
# CHECK: 21000c: 1f 20 03 d5 nop
Modified: lld/trunk/test/ELF/aarch64-tls-iele.s
URL: http://llvm.org/viewvc/llvm-project/lld/trunk/test/ELF/aarch64-tls-iele.s?rev=350681&r1=350680&r2=350681&view=diff
==============================================================================
--- lld/trunk/test/ELF/aarch64-tls-iele.s (original)
+++ lld/trunk/test/ELF/aarch64-tls-iele.s Tue Jan 8 16:09:59 2019
@@ -9,13 +9,13 @@
# RELOC: Relocations [
# RELOC-NEXT: ]
-# TCB size = 0x16 and foo is first element from TLS register.
+# TCB size = 64 and foo is first element from TLS register.
# CHECK: Disassembly of section .text:
# CHECK: _start:
# CHECK-NEXT: 210000: 00 00 a0 d2 movz x0, #0, lsl #16
-# CHECK-NEXT: 210004: 80 02 80 f2 movk x0, #20
+# CHECK-NEXT: 210004: 80 08 80 f2 movk x0, #68
# CHECK-NEXT: 210008: 00 00 a0 d2 movz x0, #0, lsl #16
-# CHECK-NEXT: 21000c: 00 02 80 f2 movk x0, #16
+# CHECK-NEXT: 21000c: 00 08 80 f2 movk x0, #64
.section .tdata
.align 2
Modified: lld/trunk/test/ELF/aarch64-tls-le.s
URL: http://llvm.org/viewvc/llvm-project/lld/trunk/test/ELF/aarch64-tls-le.s?rev=350681&r1=350680&r2=350681&view=diff
==============================================================================
--- lld/trunk/test/ELF/aarch64-tls-le.s (original)
+++ lld/trunk/test/ELF/aarch64-tls-le.s Tue Jan 8 16:09:59 2019
@@ -4,7 +4,7 @@
# RUN: llvm-objdump -d %tout | FileCheck %s
# RUN: llvm-readobj -s -r %tout | FileCheck -check-prefix=RELOC %s
-#Local-Dynamic to Initial-Exec relax creates no
+#Local-Dynamic to Local-Exec relax creates no
#RELOC: Relocations [
#RELOC-NEXT: ]
@@ -17,12 +17,12 @@ _start:
add x0, x0, :tprel_hi12:v2
add x0, x0, :tprel_lo12_nc:v2
-# TCB size = 0x16 and foo is first element from TLS register.
+# TCB size = 64 and foo is first element from TLS register.
#CHECK: Disassembly of section .text:
#CHECK: _start:
#CHECK: 210000: 40 d0 3b d5 mrs x0, TPIDR_EL0
#CHECK: 210004: 00 00 40 91 add x0, x0, #0, lsl #12
-#CHECK: 210008: 00 40 00 91 add x0, x0, #16
+#CHECK: 210008: 00 00 01 91 add x0, x0, #64
#CHECK: 21000c: 40 d0 3b d5 mrs x0, TPIDR_EL0
#CHECK: 210010: 00 fc 7f 91 add x0, x0, #4095, lsl #12
#CHECK: 210014: 00 e0 3f 91 add x0, x0, #4088
@@ -36,9 +36,9 @@ v1:
.word 0
.size v1, 4
-# The current offset from the thread pointer is 20. Raise it to just below the
+# The current offset from the thread pointer is 68. Raise it to just below the
# 24-bit limit.
-.space (0xfffff8 - 20)
+.space (0xfffff8 - 68)
.type v2, at object
.globl v2
Modified: lld/trunk/test/ELF/aarch64-tlsld-ldst.s
URL: http://llvm.org/viewvc/llvm-project/lld/trunk/test/ELF/aarch64-tlsld-ldst.s?rev=350681&r1=350680&r2=350681&view=diff
==============================================================================
--- lld/trunk/test/ELF/aarch64-tlsld-ldst.s (original)
+++ lld/trunk/test/ELF/aarch64-tlsld-ldst.s Tue Jan 8 16:09:59 2019
@@ -26,27 +26,27 @@ _start: mrs x8, TPIDR_EL0
// CHECK: _start:
// CHECK-NEXT: 210000: 48 d0 3b d5 mrs x8, TPIDR_EL0
-// 0x0 + c10 = 0xc10 = tcb (16-bytes) + var0
-// CHECK-NEXT: 210004: 08 01 40 91 add x8, x8, #0, lsl #12
-// CHECK-NEXT: 210008: 14 05 c3 3d ldr q20, [x8, #3088]
-// 0x1000 + 0x820 = 0x1820 = tcb + var1
-// CHECK-NEXT: 21000c: 08 05 40 91 add x8, x8, #1, lsl #12
-// CHECK-NEXT: 210010: 00 11 44 f9 ldr x0, [x8, #2080]
-// 0x2000 + 0x428 = 0x2428 = tcb + var2
-// CHECK-NEXT: 210014: 08 09 40 91 add x8, x8, #2, lsl #12
-// CHECK-NEXT: 210018: 00 29 44 b9 ldr w0, [x8, #1064]
-// 0x3000 + 0x2c = 0x302c = tcb + var3
-// CHECK-NEXT: 21001c: 08 0d 40 91 add x8, x8, #3, lsl #12
-// CHECK-NEXT: 210020: 00 59 40 79 ldrh w0, [x8, #44]
-// 0x3000 + 0xc2e = 0x32ce = tcb + var4
-// CHECK-NEXT: 210024: 08 0d 40 91 add x8, x8, #3, lsl #12
-// CHECK-NEXT: 210028: 00 b9 70 39 ldrb w0, [x8, #3118]
+// 0x0 + c40 = 0xc40 = tcb (64-bytes) + var0
+// CHECK-NEXT: 210004: 08 01 40 91 add x8, x8, #0, lsl #12
+// CHECK-NEXT: 210008: 14 11 c3 3d ldr q20, [x8, #3136]
+// 0x1000 + 0x850 = 0x1850 = tcb + var1
+// CHECK-NEXT: 21000c: 08 05 40 91 add x8, x8, #1, lsl #12
+// CHECK-NEXT: 210010: 00 29 44 f9 ldr x0, [x8, #2128]
+// 0x2000 + 0x458 = 0x2458 = tcb + var2
+// CHECK-NEXT: 210014: 08 09 40 91 add x8, x8, #2, lsl #12
+// CHECK-NEXT: 210018: 00 59 44 b9 ldr w0, [x8, #1112]
+// 0x3000 + 0x5c = 0x305c = tcb + var3
+// CHECK-NEXT: 21001c: 08 0d 40 91 add x8, x8, #3, lsl #12
+// CHECK-NEXT: 210020: 00 b9 40 79 ldrh w0, [x8, #92]
+// 0x3000 + 0xc5e = 0x3c5e = tcb + var4
+// CHECK-NEXT: 210024: 08 0d 40 91 add x8, x8, #3, lsl #12
+// CHECK-NEXT: 210028: 00 79 71 39 ldrb w0, [x8, #3166]
-// CHECK-SYMS: 0000000000000c00 0 TLS GLOBAL DEFAULT 2 var0
-// CHECK-SYMS-NEXT: 0000000000001810 4 TLS GLOBAL DEFAULT 2 var1
-// CHECK-SYMS-NEXT: 0000000000002418 2 TLS GLOBAL DEFAULT 2 var2
-// CHECK-SYMS-NEXT: 000000000000301c 1 TLS GLOBAL DEFAULT 2 var3
-// CHECK-SYMS-NEXT: 0000000000003c1e 0 TLS GLOBAL DEFAULT 2 var4
+// CHECK-SYMS: 0000000000000c00 16 TLS GLOBAL DEFAULT 2 var0
+// CHECK-SYMS-NEXT: 0000000000001810 8 TLS GLOBAL DEFAULT 2 var1
+// CHECK-SYMS-NEXT: 0000000000002418 4 TLS GLOBAL DEFAULT 2 var2
+// CHECK-SYMS-NEXT: 000000000000301c 2 TLS GLOBAL DEFAULT 2 var3
+// CHECK-SYMS-NEXT: 0000000000003c1e 1 TLS GLOBAL DEFAULT 2 var4
.globl var0
.globl var1
@@ -59,12 +59,12 @@ _start: mrs x8, TPIDR_EL0
.type var3, at object
.section .tbss,"awT", at nobits
- .balign 16
+ .balign 64
.space 1024 * 3
var0:
.quad 0
.quad 0
- .size var1, 16
+ .size var0, 16
.space 1024 * 3
var1:
.quad 0
@@ -72,14 +72,14 @@ var1:
.space 1024 * 3
var2:
.word 0
- .size var1, 4
+ .size var2, 4
.space 1024 * 3
var3:
.hword 0
- .size var2, 2
+ .size var3, 2
.space 1024 * 3
var4:
.byte 0
- .size var3, 1
+ .size var4, 1
.space 1024 * 3
Modified: lld/trunk/test/ELF/arm-tls-le32.s
URL: http://llvm.org/viewvc/llvm-project/lld/trunk/test/ELF/arm-tls-le32.s?rev=350681&r1=350680&r2=350681&view=diff
==============================================================================
--- lld/trunk/test/ELF/arm-tls-le32.s (original)
+++ lld/trunk/test/ELF/arm-tls-le32.s Tue Jan 8 16:09:59 2019
@@ -69,9 +69,9 @@ x:
// CHECK: Disassembly of section .text:
// CHECK-NEXT: _start:
-// offset of x from Thread pointer = (TcbSize + 0x0 = 0x8)
-// CHECK-NEXT: 11000: 08 00 00 00
-// offset of z from Thread pointer = (TcbSize + 0x8 = 0x10)
-// CHECK-NEXT: 11004: 10 00 00 00
-// offset of y from Thread pointer = (TcbSize + 0x4 = 0xc)
-// CHECK-NEXT: 11008: 0c 00 00 00
+// offset of x from Thread pointer = (TcbSize + 0x0 = 0x20)
+// CHECK-NEXT: 11000: 20 00 00 00
+// offset of z from Thread pointer = (TcbSize + 0x8 = 0x28)
+// CHECK-NEXT: 11004: 28 00 00 00
+// offset of y from Thread pointer = (TcbSize + 0x4 = 0x24)
+// CHECK-NEXT: 11008: 24 00 00 00
Modified: lld/trunk/test/ELF/arm-tls-norelax-ie-le.s
URL: http://llvm.org/viewvc/llvm-project/lld/trunk/test/ELF/arm-tls-norelax-ie-le.s?rev=350681&r1=350680&r2=350681&view=diff
==============================================================================
--- lld/trunk/test/ELF/arm-tls-norelax-ie-le.s (original)
+++ lld/trunk/test/ELF/arm-tls-norelax-ie-le.s Tue Jan 8 16:09:59 2019
@@ -37,5 +37,5 @@ x2:
.type x2, %object
// CHECK: Contents of section .got:
-// x1 at offset 8 from TP, x2 at offset c from TP. Offsets include TCB size of 8
-// CHECK-NEXT: 13064 08000000 0c000000
+// x1 at offset 0x20 from TP, x2 at offset 0x24 from TP. Offsets include TCB size of 0x20
+// CHECK-NEXT: 13064 20000000 24000000
More information about the llvm-commits
mailing list