llvm dev - Nov 2021 - [ASAN] [Regression] Stack pointer corruption on ARMv7

Hello,

I discovered a stack pointer corruption regression in LLVM 13 causing stack leak
with ASAN enabled when targeting cortex-a9.

The issue is reproducible with the file in the attachment. Also published
online[1]. The arguments to compile are as follows:

clang -o error.o -c -target arm-gnu-linux-eabi -mcpu=cortex-a9
-fsanitize=address error.c

The problematic part is the loop in the `test` function. When `func` is called,
LLVM does not adjust the stack back, causing fast stack overflow:

.LBB2_5:                                @ %for.cond
        ...
        sub     sp, sp, #8
        str     r12, [sp]
        bl      func
.LBB2_7:                                @ %if.end
        b       .LBB2_5

LLVM 12.0.1 does not have this issue. I will post the issue to the bugtracker
when GitHub Issues land or bugzilla returns. Can anyone confirm that the issue
is still present in main?

Best regards,
Vitaly

[1] https://godbolt.org/z/GjxqPfMW6

-------------- next part --------------
A non-text attachment was scrubbed...
Name: error.c
Type: application/octet-stream
Size: 360 bytes
Desc: not available
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20211128/30616bda/attachment.obj>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: Message signed with OpenPGP
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20211128/30616bda/attachment.sig>

The issue is present in main and is caused by
d88f96dff3f192fc0c1bf57f7810b95a709b3591.
Added Tim and Jessica to CC.

Differential: https://reviews.llvm.org/D102613

Best regards,
Vitaly
> On 28 Nov 2021, at 20:09, Vit “vit9696” U <vit9696 at avp.su> wrote:
> 
> Hello,
> 
> I discovered a stack pointer corruption regression in LLVM 13 causing stack
leak with ASAN enabled when targeting cortex-a9.
> 
> The issue is reproducible with the file in the attachment. Also published
online[1]. The arguments to compile are as follows:
> 
> clang -o error.o -c -target arm-gnu-linux-eabi -mcpu=cortex-a9
-fsanitize=address error.c
> 
> The problematic part is the loop in the `test` function. When `func` is
called, LLVM does not adjust the stack back, causing fast stack overflow:
> 
> .LBB2_5:                                @ %for.cond
>        ...
>        sub     sp, sp, #8
>        str     r12, [sp]
>        bl      func
> .LBB2_7:                                @ %if.end
>        b       .LBB2_5
> 
> LLVM 12.0.1 does not have this issue. I will post the issue to the
bugtracker when GitHub Issues land or bugzilla returns. Can anyone confirm that
the issue is still present in main?
> 
> Best regards,
> Vitaly
> 
> [1] https://godbolt.org/z/GjxqPfMW6
> 
> <error.c>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: Message signed with OpenPGP
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20211129/98917d30/attachment.sig>