[JITLink][AArch32] Tests for ELF::R_ARM_ABS32 and ELF::R_ARM_REL32

[sgraenitz]

Summary

Testing for Static Data relocations for both ARM and Thumb targets.

[sgraenitz]

This is looking very good already! Nice idea to check ARM and Thumb in one test. I am not sure, however, if it's worth the effort. They both produce R_ARM_ABS32 and R_ARM_REL32. The only difference is the alignment and here Thumb alone would be sufficient:

➜ llvm-objdump -r ...
RELOCATION RECORDS FOR [.text]:                   RELOCATION RECORDS FOR [.text]:
OFFSET   TYPE                     VALUE           OFFSET   TYPE                     VALUE
00000004 R_ARM_ABS32              target          00000002 R_ARM_ABS32              target
00000008 R_ARM_REL32              target          00000006 R_ARM_REL32              target
       ^ ARM is on a 4-byte boundary                     ^ Thumb is 2-byte

What we should do at some point (not necessarily in this patch) is to check whether it works on big-endian targets. Add a eb suffix to the arch in order to test it, e.g. thumbebv7-none-linux-gnueabihf

llvm/test/ExecutionEngine/JITLink/AArch32/ELF_static_data_reloc.s

17

Yes, but it's no data-relocation, so it should be out-of-scope for this test.

[sgraenitz]

llvm/test/ExecutionEngine/JITLink/AArch32/ELF_static_data_reloc.s

34

Is your intention to distinguish command-lines from comments by using // and # respectively? Because then it seems you are rotating them here? All test I know use the same comment marker for both.

[Eymay]

In D157519#4574123, @sgraenitz wrote:

This is looking very good already! Nice idea to check ARM and Thumb in one test. I am not sure, however, if it's worth the effort. They both produce R_ARM_ABS32 and R_ARM_REL32. The only difference is the alignment and here Thumb alone would be sufficient:

➜ llvm-objdump -r ...
RELOCATION RECORDS FOR [.text]:                   RELOCATION RECORDS FOR [.text]:
OFFSET   TYPE                     VALUE           OFFSET   TYPE                     VALUE
00000004 R_ARM_ABS32              target          00000002 R_ARM_ABS32              target
00000008 R_ARM_REL32              target          00000006 R_ARM_REL32              target
       ^ ARM is on a 4-byte boundary                     ^ Thumb is 2-byte

What we should do at some point (not necessarily in this patch) is to check whether it works on big-endian targets. Add a eb suffix to the arch in order to test it, e.g. thumbebv7-none-linux-gnueabihf

Agree, I'd say it is more a proof of concept. These provide the scaffolding to add tests for varying targets.

[Eymay]

Removed non-data relocation checks and standardised comment styling

[Eymay]

Thanks for pointing out.

[Eymay]

replace "//" with "#" entirely

[sgraenitz]

Thanks for the updates! Can we align the assembly in the test a little more with the (now) existing ELF_static_thumb_reloc.s? That may avoid future divergence. Otherwise, this looks good to me.

llvm/test/ExecutionEngine/JITLink/AArch32/ELF_static_data_reloc.s

14

Function addresses are not in scope for this test, so we can drop the comment as well

[Eymay]

Update the test to align with ELF_static_thumb_reloc.s test.

[Eymay]

The disassembly of both ARM and Thumb turned out to be matching but CHECK-INSTR-{ARM,THM} can be added in the future for the lines that diverge.

[sgraenitz]

Some polishing before landing this

[sgraenitz]

Drop check lines for disassembly. Data relocations don't invovle instruction opcodes.

[sgraenitz]

Data relocations have no alignment requirements. Check it by turning target into a data symbol itself and injecting an uneven address.

[sgraenitz]

Add suffixes _armv7 and _thumbv7 to temporary object file names in order to avoid collisions

[sgraenitz]

Streamline indents, new-lines, comments

[lhames]

LGTM too. Thanks @Eymay and @sgraenitz!

[Eymay]

LGTM!

llvm/test/ExecutionEngine/JITLink/AArch32/ELF_static_data_reloc.s

+# RUN: llvm-mc -triple=armv7-none-linux-gnueabi -arm-add-build-attributes -filetype=obj -o %t_armv7.o %s
+# RUN: llvm-objdump -r %t_armv7.o | FileCheck --check-prefix=CHECK-TYPE %s
+# RUN: llvm-jitlink -noexec -slab-address 0x76ff0000 -slab-allocate 10Kb -slab-page-size 4096 \
+# RUN:              -abs target=0x76bbe88f -check %s %t_armv7.o
+
+# RUN: llvm-mc -triple=thumbv7-none-linux-gnueabi -arm-add-build-attributes -filetype=obj -o %t_thumbv7.o %s
+# RUN: llvm-objdump -r %t_thumbv7.o | FileCheck --check-prefix=CHECK-TYPE %s
+# RUN: llvm-jitlink -noexec -slab-address 0x76ff0000 -slab-allocate 10Kb -slab-page-size 4096 \
+# RUN:              -abs target=0x76bbe88f -check %s %t_thumbv7.o
+
+  .data
+  .global target
+
+  .text
+  .syntax unified
+
+# CHECK-TYPE: {{[0-9a-f]+}} R_ARM_ABS32 target
+# jitlink-check: *{4}(abs32) = target
+  .global abs32
+abs32:
+  .word target
+  .size abs32, .-abs32
+
+# CHECK-TYPE: {{[0-9a-f]+}} R_ARM_REL32 target
+# jitlink-check: (rel32 + *{4}(rel32))[31:0] = target
+  .global rel32
+rel32:
+  .word target - .
+  .size rel32, .-rel32
+
+# Empty main function for jitlink to be happy
+  .globl  main
+  .type main, %function
+  .p2align  2
+main:
+  bx lr
+  .size   main, .-main