linux-kernel - Re: [PATCH 3/3] x86: Disable running 32bit processes if ia32

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]

Message-ID: <CAMzpN2gvCE6UqC7kk61UCM9y7AGBHoczSL6YTcmLsdJGWnwhPQ@mail.gmail.com>
Date:   Thu, 8 Jun 2023 00:37:37 -0400
From:   Brian Gerst <brgerst@...il.com>
To:     Nikolay Borisov <nik.borisov@...e.com>
Cc:     x86@...nel.org, linux-kernel@...r.kernel.org, mhocko@...e.com,
        jslaby@...e.cz
Subject: Re: [PATCH 3/3] x86: Disable running 32bit processes if ia32_disabled
 is passed

On Wed, Jun 7, 2023 at 3:41 AM Nikolay Borisov <nik.borisov@...e.com> wrote:
>
> In addition to disabling 32bit syscall interface let's also disable the
> ability to run 32bit processes altogether. This is achieved by setting
> the GDT_ENTRY_DEFAULT_USER32_CS descriptor to not present which would
> cause 32 bit processes to trap with a #NP exception. Furthermore,
> forbid loading compat processes as well.
>
> Signed-off-by: Nikolay Borisov <nik.borisov@...e.com>
> ---
>  arch/x86/include/asm/elf.h   | 5 +++--
>  arch/x86/kernel/cpu/common.c | 8 ++++++++
>  2 files changed, 11 insertions(+), 2 deletions(-)
>
> diff --git a/arch/x86/include/asm/elf.h b/arch/x86/include/asm/elf.h
> index 18fd06f7936a..406245bc0fb0 100644
> --- a/arch/x86/include/asm/elf.h
> +++ b/arch/x86/include/asm/elf.h
> @@ -148,9 +148,10 @@ do {                                               \
>  #define elf_check_arch(x)                      \
>         ((x)->e_machine == EM_X86_64)
>
> +extern bool ia32_disabled;
>  #define compat_elf_check_arch(x)                                       \
> -       (elf_check_arch_ia32(x) ||                                      \
> -        (IS_ENABLED(CONFIG_X86_X32_ABI) && (x)->e_machine == EM_X86_64))
> +       (!ia32_disabled && (elf_check_arch_ia32(x) ||                   \
> +        (IS_ENABLED(CONFIG_X86_X32_ABI) && (x)->e_machine == EM_X86_64)))
>
>  static inline void elf_common_init(struct thread_struct *t,
>                                    struct pt_regs *regs, const u16 ds)
> diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c
> index 71f8b55f70c9..ddc301c09419 100644
> --- a/arch/x86/kernel/cpu/common.c
> +++ b/arch/x86/kernel/cpu/common.c
> @@ -2359,6 +2359,11 @@ void microcode_check(struct cpuinfo_x86 *prev_info)
>  }
>  #endif
>
> +static void remove_user32cs_from_gdt(void * __unused)
> +{
> +       get_current_gdt_rw()[GDT_ENTRY_DEFAULT_USER32_CS].p = 0;
> +}
> +
>  /*
>   * Invoked from core CPU hotplug code after hotplug operations
>   */
> @@ -2368,4 +2373,7 @@ void arch_smt_update(void)
>         cpu_bugs_smt_update();
>         /* Check whether IPI broadcasting can be enabled */
>         apic_smt_update();
> +       if (ia32_disabled)
> +               on_each_cpu(remove_user32cs_from_gdt, NULL, 1);
> +
>  }

Disabling USER32_CS isn't really necessary, as simply running 32-bit
user code (in a normally 64-bit process) doesn't pose much of a threat
to the kernel with 32-bit syscalls disabled.

Wine, for example, is moving to a model where the main code runs in
64-bit mode, uses only 64-bit unix libraries, and the 32->64 bit
transitions are done entirely in userspace.  It will still need the
ability to use 32-bit code segments, but won't need 32-bit unix
libraries or syscalls to run 32-bit Windows code.

Brian Gerst