Hardware ASan Generating Unknown Instruction

Hi,

I am trying to execute a simple hello world program compiled like so:

path/to/compiled/clang -o test --target=aarch64-linux-gnu
-march=armv8.5-a -fsanitize=hwaddress
--sysroot=/usr/aarch64-linux-gnu/
-L/usr/lib/gcc/aarch64-linux-gnu/10.1.0/ -g test.c

However, when I look at the disassembly, there is an unknown
instruction listed at 0x2d51c:

000000000002d4c0 main:
   2d4c0: ff c3 00 d1 sub sp, sp, #48
   2d4c4: fd 7b 02 a9 stp x29, x30, [sp, #32]
   2d4c8: fd 83 00 91 add x29, sp, #32
   2d4cc: 08 01 00 d0 adrp x8, #139264
   2d4d0: 08 49 47 f9 ldr x8, [x8, #3728]
   2d4d4: 49 d0 3b d5 mrs x9, TPIDR_EL0
   2d4d8: 28 69 68 f8 ldr x8, [x9, x8]
   2d4dc: 08 7d 40 b2 orr x8, x8, #0xffffffff
   2d4e0: 08 05 00 91 add x8, x8, #1
   2d4e4: a2 13 00 d1 sub x2, x29, #4
   2d4e8: e9 03 08 aa mov x9, x8
   2d4ec: df 64 ff 97 bl #-158852
<__hwasan_check_x2_18_short>
   2d4f0: ea 03 1f 2a mov w10, wzr
   2d4f4: aa c3 1f b8 stur w10, [x29, #-4]
   2d4f8: a2 23 00 d1 sub x2, x29, #8
   2d4fc: e9 03 08 aa mov x9, x8
   2d500: da 64 ff 97 bl #-158872
<__hwasan_check_x2_18_short>
   2d504: a0 83 1f b8 stur w0, [x29, #-8]
   2d508: e2 43 00 91 add x2, sp, #16
   2d50c: e9 03 08 aa mov x9, x8
   2d510: ec 64 ff 97 bl #-158800
<__hwasan_check_x2_19_short>
   2d514: e1 0b 00 f9 str x1, [sp, #16]
   2d518: 40 00 00 d0 adrp x0, #40960
   2d51c: e0 ff e5 b2 <unknown>
   2d520: 00 40 35 91 add x0, x0, #3408
   2d524: ea 0f 00 b9 str w10, [sp, #12]
   2d528: c2 64 ff 97 bl #-158968 <printf@plt>
   2d52c: ea 0f 40 b9 ldr w10, [sp, #12]
   2d530: e0 0b 00 b9 str w0, [sp, #8]
   2d534: e0 03 0a 2a mov w0, w10
   2d538: fd 7b 42 a9 ldp x29, x30, [sp, #32]
   2d53c: ff c3 00 91 add sp, sp, #48
   2d540: c0 03 5f d6 ret

Am I compiling the program incorrectly? For reference, here's my clang version

# llvm-project/build-10.x/install/bin/clang --version
                                             [1]
clang version 10.0.1 (git@github.com:llvm/llvm-project.git
a634a80615b1e012f1a61aa0cd1e2e67ef77d0bd)
Target: x86_64-unknown-linux-gnu
Thread model: posix

Hi Derrick,

Can you send through test.c? What does clang’s assembly say (path/to/compiled/clang --target=aarch64-linux-gnu -march=armv8.5-a -fsanitize=hwaddress -S test.c; cat test.s)?

  • Mitch

I suspect that this is hitting the issue that I mentioned here:
https://reviews.llvm.org/D65857#1621335

We may need to do what I suggested there and restrict global tag entropy on non-Android Linux to 7 bits. You can try working around this issue for now by using lld as the linker (-fuse-ld=lld).

Peter

movk x0, #:prel_g3:.L.str+4294967296

Looks like it’s getting tripped up on the relocation.

Peter - was there any additional work on the prel_g3 relocation when you landed these patches?

Using lld fixes this issue.

Using lld fixes this issue.

I suspect that this is hitting the issue that I mentioned here:
https://reviews.llvm.org/D65857#1621335

We may need to do what I suggested there and restrict global tag entropy on non-Android Linux to 7 bits.

This sounds like a reasonable compromise.

Thanks for the confirmation. From the assembly that was sent on the other branch of the thread:

.set .L.str, .L.str.hwasan-3458764513820540928

-3458764513820540928 = 0xd0 << 56
i.e. a “negative” tag.

So this appears to be the issue exactly.

Peter

Should I report a new bug, or will the existing review suffice?

Can you file the bug so that we don’t lose track of the issue, please?

Peter

Will do.