linux-kernel - Re: [PATCH 00/21] objtool: vmlinux.o and CLANG LTO support

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

Date:   Mon, 18 Jan 2021 11:22:55 -0600
From:   Josh Poimboeuf <jpoimboe@...hat.com>
To:     Sami Tolvanen <samitolvanen@...gle.com>
Cc:     X86 ML <x86@...nel.org>, LKML <linux-kernel@...r.kernel.org>,
        Peter Zijlstra <peterz@...radead.org>,
        Sedat Dilek <sedat.dilek@...il.com>,
        Kees Cook <keescook@...omium.org>,
        Nick Desaulniers <ndesaulniers@...gle.com>,
        clang-built-linux <clang-built-linux@...glegroups.com>,
        Miroslav Benes <mbenes@...e.cz>
Subject: Re: [PATCH 00/21] objtool: vmlinux.o and CLANG LTO support

On Thu, Jan 14, 2021 at 04:41:28PM -0800, Sami Tolvanen wrote:
> I can confirm that all the warnings I previously saw are now fixed,
> but I'm seeing a few new ones:
> 
> vmlinux.o: warning: objtool: balance_leaf_when_delete()+0x17d4: stack state mismatch: cfa1=7+192 cfa2=7+176
> vmlinux.o: warning: objtool: internal_move_pointers_items()+0x9f7: stack state mismatch: cfa1=7+160 cfa2=7+176

These seem legit stack state mismatches (compiler bug).  Two code
blocks, with different stack sizes, transfer control to the same
noreturn block (violating DWARF/ORC expectation that each instruction
has a deterministic stack layout).  In both cases the noreturn block has
a call to __reiserfs_panic().

https://paste.centos.org/view/081cbfc1
https://paste.centos.org/view/265968a6

> vmlinux.o: warning: objtool: strncpy_from_user()+0x181: call to do_strncpy_from_user() with UACCESS enabled
> vmlinux.o: warning: objtool: strnlen_user()+0x12b: call to do_strnlen_user() with UACCESS enabled

It's odd that Clang wouldn't inline these static single-called
functions.  I could '__always_inline' them, but is this expected
behavior?

> vmlinux.o: warning: objtool: i915_gem_execbuffer2_ioctl()+0x390: call to __ubsan_handle_negate_overflow() with UACCESS enabled

PeterZ, have you seen this one?

https://paste.centos.org/view/b4723113

> vmlinux.o: warning: objtool: .text.snd_trident_free_voice: unexpected end of section

Another weird compiler issue.  It generates obviously dead code which
jumps off the end of the function:

  f7:   b0 01                   mov    $0x1,%al
  f9:   84 c0                   test   %al,%al
  fb:   0f 84 79 05 00 00       je     67a <snd_trident_free_voice+0x67a>

https://paste.centos.org/view/a1887ae3

-- 
Josh