linux-kernel - Re: [PATCH V2 11/17] riscv: compat: Add elf.h implementation

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: <CAK8P3a0H2Nq=bFdzWGzzGuFWP85JA7=Td6_rb8GqOF3bYCRJBw@mail.gmail.com>
Date:   Mon, 10 Jan 2022 15:28:40 +0100
From:   Arnd Bergmann <arnd@...db.de>
To:     Guo Ren <guoren@...nel.org>
Cc:     Palmer Dabbelt <palmer@...belt.com>, Arnd Bergmann <arnd@...db.de>,
        Anup Patel <anup.patel@....com>,
        gregkh <gregkh@...uxfoundation.org>,
        liush <liush@...winnertech.com>, Wei Fu <wefu@...hat.com>,
        Drew Fustini <drew@...gleboard.org>,
        Wang Junqiang <wangjunqiang@...as.ac.cn>,
        Christoph Hellwig <hch@...radead.org>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        linux-riscv <linux-riscv@...ts.infradead.org>,
        linux-csky@...r.kernel.org,
        linux-s390 <linux-s390@...r.kernel.org>,
        sparclinux <sparclinux@...r.kernel.org>,
        linuxppc-dev <linuxppc-dev@...ts.ozlabs.org>,
        inux-parisc@...r.kernel.org,
        "open list:BROADCOM NVRAM DRIVER" <linux-mips@...r.kernel.org>,
        Linux ARM <linux-arm-kernel@...ts.infradead.org>,
        "the arch/x86 maintainers" <x86@...nel.org>,
        Guo Ren <guoren@...ux.alibaba.com>
Subject: Re: [PATCH V2 11/17] riscv: compat: Add elf.h implementation

On Tue, Dec 28, 2021 at 3:39 PM <guoren@...nel.org> wrote:
>
> From: Guo Ren <guoren@...ux.alibaba.com>
>
> Implement necessary type and macro for compat elf. See the code
> comment for detail.
>
> Signed-off-by: Guo Ren <guoren@...ux.alibaba.com>
> Signed-off-by: Guo Ren <guoren@...nel.org>
> Cc: Arnd Bergmann <arnd@...db.de>

This looks mostly correct,

> +/*
> + * FIXME: not sure SET_PERSONALITY for compat process is right!
> + */
> +#define SET_PERSONALITY(ex)                                               \
> +do {    if ((ex).e_ident[EI_CLASS] == ELFCLASS32)                         \
> +               set_thread_flag(TIF_32BIT);                                \
> +       else                                                               \
> +               clear_thread_flag(TIF_32BIT);                              \
> +       set_personality(PER_LINUX | (current->personality & (~PER_MASK))); \
> +} while (0)

This means the personality after exec is always set to PER_LINUX, not
PER_LINUX32, which I think is wrong: you want the PER_LINUX32
setting to stick, just like the upper bits do in the default implementation.

What the other ones do is:

| arch/parisc/include/asm/elf.h-
set_personality((current->personality & ~PER_MASK) | PER_LINUX); \

This looks like the same problem you introduce here: always forcing PER_LINUX
instead of PER_LINUX32 makes it impossible to use PER_LINUX32.

| arch/alpha/include/asm/elf.h:#define SET_PERSONALITY(EX)
                           \
| arch/alpha/include/asm/elf.h-   set_personality(((EX).e_flags &
EF_ALPHA_32BIT)         \
| arch/alpha/include/asm/elf.h-      ? PER_LINUX_32BIT : PER_LINUX)
| arch/csky/include/asm/elf.h:#define SET_PERSONALITY(ex)
set_personality(PER_LINUX)
| arch/nds32/include/asm/elf.h:#define SET_PERSONALITY(ex)
set_personality(PER_LINUX)

These look even worse: instead of forcing the lower bits to
PER_LINUX/PER_LINUX32 and
leaving the upper bits untouched, these also clear the upper bits
unconditionally.

| arch/arm64/include/asm/elf.h:#define SET_PERSONALITY(ex)
                                   \
| arch/arm64/include/asm/elf.h-   current->personality &=
~READ_IMPLIES_EXEC;                     \
| arch/x86/um/asm/elf.h:#define SET_PERSONALITY(ex) do {} while(0)
| arch/x86/include/asm/elf.h:#define set_personality_64bit()      do {
} while (0)
| arch/x86/kernel/process_64.c:static void __set_personality_ia32(void)
|         current->personality |= force_personality32;

Inconsistent: does not enforce PER_LINUX/PER_LINUX32 as the default
implementation
does, but just leaves the value untouched (other than (re)setting
READ_IMPLIES_EXEC).
I think this is harmless otherwise, as we generally ignore the lower
bits, except for the
bit of code that checks for PER_LINUX32 in override_architecture() to mangle the
output of sys_newuname() or in /proc/cpuinfo.

| arch/s390/include/asm/elf.h-    if
(personality(current->personality) != PER_LINUX32)   \
| arch/s390/include/asm/elf.h-            set_personality(PER_LINUX |
                   \
| arch/s390/include/asm/elf.h-
(current->personality & ~PER_MASK));    \
| arch/powerpc/include/asm/elf.h- if
(personality(current->personality) != PER_LINUX32)   \
| arch/powerpc/include/asm/elf.h-         set_personality(PER_LINUX |
                   \
| arch/powerpc/include/asm/elf.h-
(current->personality & (~PER_MASK)));  \
| arch/sparc/include/asm/elf_64.h-        if
(personality(current->personality) != PER_LINUX32)   \
| arch/sparc/include/asm/elf_64.h-
set_personality(PER_LINUX |             \
| arch/sparc/include/asm/elf_64.h-
(current->personality & (~PER_MASK)));  \

This is probably the behavior you want to copy.

      Arnd