linux-kernel - Re: [PATCH] x86/vm86: fix compilation warning on a unused variable

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:   Fri, 16 Dec 2016 23:19:16 -0500
From:   Jérémy Lefaure <jeremy.lefaure@....epita.fr>
To:     "Kirill A. Shutemov" <kirill@...temov.name>
Cc:     "Kirill A. Shutemov" <kirill.shutemov@...ux.intel.com>,
        x86@...nel.org, Borislav Petkov <bp@...e.de>,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH] x86/vm86: fix compilation warning on a unused variable

On Mon, 12 Dec 2016 17:52:50 +0300
"Kirill A. Shutemov" <kirill@...temov.name> wrote:

> On Thu, Dec 08, 2016 at 01:25:37PM -0500, Jérémy Lefaure wrote:
> > On Thu, 8 Dec 2016 13:50:11 +0300
> > "Kirill A. Shutemov" <kirill.shutemov@...ux.intel.com> wrote:
> >   
> > > On Wed, Dec 07, 2016 at 11:38:33PM -0500, Jérémy Lefaure wrote:  
> > > > When CONFIG_TRANSPARENT_HUGEPAGE is disabled, split_huge_pmd is a no-op
> > > > stub. In such case, vma is unused and a compiler raises a warning:
> > > > 
> > > > arch/x86/kernel/vm86_32.c: In function ‘mark_screen_rdonly’:
> > > > arch/x86/kernel/vm86_32.c:180:26: warning: unused variable ‘vma’
> > > > [-Wunused-variable]
> > > >    struct vm_area_struct *vma = find_vma(mm, 0xA0000);
> > > >                              ^~~
> > > > Adding __maybe_unused in the vma declaration fixes this warning.    
> > > 
> > > Hm. pmd_trans_huge() is zero if CONFIG_TRANSPARENT_HUGEPAGE is not set.
> > > Compiler should get rid of whole block of code under the 'if'.
> > > 
> > > Could you share your kernel config which triggers the warning?
> > > And what compiler do you use?
> > >   
> > 
> > After a `make allnoconfig`, I enable "Legacy VM86 support" and nothing
> > else. I tested with 2 compilers, gcc 4.9.2 (on debian jessie) and gcc
> > 6.2.1 (on archlinux).
> > 
> > Actually, the compiler does not raise warnings on complete build (`make
> > mrproper`, configuration and `make`) but only on partial build (`make
> > arch/x86/kernel/vm86_32.o` or `touch arch/x86/kernel/vm86_32.c &&
> > make`). So maybe it is a compiler issue ?
> > 
Sorry, forget about this part, it's false. I may have tested too
quickly and missed something.

> > The solution you propose in your other email (adding "(void)__vma;" in
> > the no-op split_huge_pmd) seems to fix the warnings on partial build.  
> 
> This doesn't make any sense, but this works too:
> 
I don't know why gcc raises a warning on that even if it is not used.
Anyway, I'm sure that the warning is reproducible. Both of your
solutions fix the issue.

> diff --git a/arch/x86/kernel/vm86_32.c b/arch/x86/kernel/vm86_32.c
> index e339717af265..63ddc598f5a9 100644
> --- a/arch/x86/kernel/vm86_32.c
> +++ b/arch/x86/kernel/vm86_32.c
> @@ -160,6 +160,7 @@ void save_v86_state(struct kernel_vm86_regs *regs, int retval)
> 
>  static void mark_screen_rdonly(struct mm_struct *mm)
>  {
> +       struct vm_area_struct *vma;
>         pgd_t *pgd;
>         p4d_t *p4d;
>         pud_t *pud;
> @@ -181,7 +182,7 @@ static void mark_screen_rdonly(struct mm_struct *mm)
>         pmd = pmd_offset(pud, 0xA0000);
> 
>         if (pmd_trans_huge(*pmd)) {
> -               struct vm_area_struct *vma = find_vma(mm, 0xA0000);
> +               vma = find_vma(mm, 0xA0000);
>                 split_huge_pmd(vma, pmd, 0xA0000);
>         }
>         if (pmd_none_or_clear_bad(pmd))