linux-kernel - Re: [linux-next:master 12681/13861] drivers/i2c/i2c-core-base.o: warning: objtool: __i2c_transfer+0x120: stack state mismatch: reg1[24]=-1+0 reg2[24]=-2-24

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <35f7eac8-a970-5fd3-25f4-55f417e8e15e@loongson.cn>
Date: Tue, 8 Apr 2025 17:32:02 +0800
From: Tiezhu Yang <yangtiezhu@...ngson.cn>
To: Josh Poimboeuf <jpoimboe@...nel.org>
Cc: Philip Li <philip.li@...el.com>, kernel test robot <lkp@...el.com>,
 Guenter Roeck <linux@...ck-us.net>, oe-kbuild-all@...ts.linux.dev,
 Andrew Morton <akpm@...ux-foundation.org>,
 Linux Memory Management List <linux-mm@...ck.org>,
 Alessandro Carminati <acarmina@...hat.com>,
 Peter Zijlstra <peterz@...radead.org>, linux-kernel@...r.kernel.org,
 loongarch@...ts.linux.dev
Subject: Re: [linux-next:master 12681/13861] drivers/i2c/i2c-core-base.o:
 warning: objtool: __i2c_transfer+0x120: stack state mismatch: reg1[24]=-1+0
 reg2[24]=-2-24

On 04/08/2025 02:29 PM, Josh Poimboeuf wrote:
> On Tue, Apr 08, 2025 at 10:45:51AM +0800, Tiezhu Yang wrote:
>> So this is a run-time bug rather than a compile-time warning, it should
>> put the option "-fno-shrink-wrap" outside CONFIG_OBJTOOL in
>> arch/loongarch/Makefile as a workaround, like this:
>
> If loongarch folks agree it's a compiler bug, it should be reported to
> GCC, so the issue is better understood (and can get fixed).
>
> Without understanding the root cause, we don't know if -fno-shrink-wrap
> fixes it, or just makes this particular occurrence go away.

OK, thank you. I have discussed offline with the developers
Rui Wang and Lulu Cheng who are familiar with compiler, the
root cause may be that if a jump label's control flow path
exactly matches the caller's epilogue, the compiler may omit
restoring saved registers, it needs to be confirmed by GCC
developers.

By the way, add an empty inline assembly can also work around
the problem, like this:

diff --git a/arch/loongarch/include/asm/jump_label.h 
b/arch/loongarch/include/asm/jump_label.h
index 8a924bd69d19..dbc105e62380 100644
--- a/arch/loongarch/include/asm/jump_label.h
+++ b/arch/loongarch/include/asm/jump_label.h
@@ -34,6 +34,7 @@ static __always_inline bool arch_static_branch(struct 
static_key * const key, co
         return false;

  l_yes:
+       asm volatile("");
         return true;
  }

@@ -47,6 +48,7 @@ static __always_inline bool 
arch_static_branch_jump(struct static_key * const ke
         return false;

  l_yes:
+       asm volatile("");
         return true;
  }

We will fix this issue once the root cause is clear.

Thanks,
Tiezhu