Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Tue, 21 Feb 2017 15:29:01 +0900
From: Ho-Eun Ryu <hoeun.ryu@...il.com>
To: Ard Biesheuvel <ard.biesheuvel@...aro.org>
Cc: kernel-hardening@...ts.openwall.com,
 "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
 Arnd Bergmann <arnd@...db.de>,
 Kees Cook <keescook@...omium.org>,
 Ingo Molnar <mingo@...nel.org>,
 "linux-arch@...r.kernel.org" <linux-arch@...r.kernel.org>
Subject: Re: [RFC 1/7] arch: add __ro_mostly_after_init
 section marker


> On 19 Feb 2017, at 8:24 PM, Ard Biesheuvel <ard.biesheuvel@...aro.org> wrote:
> 
> On 19 February 2017 at 10:04, Hoeun Ryu <hoeun.ryu@...il.com> wrote:
>> After `__ro_after_init` marker is included in kernel, many kernel data
>> objects can be read-only-after-init. But there are many other places that
>> would be good to read-only-after-init but `__ro_after_init` can not be simply
>> applicable to them because they should be writable at some points, which are
>> during module_init/exit or dynamic de/registration for a specific subsystem.
>> `__ro_mostly_after_init` is basically the same to `__ro_after_init`. The
>> section is mapped as read-only after kernel init. The different thing is
>> this section is temporarily mapped as read-write during module_init/exit and
>> de/registration of a subsystem using set_ro_mostly_after_init_rw/ro pair.
>> Use `__ro_mostly_after_init` as a way to mark such memory instead when
>> `__ro_after_init` is not applicable because the memory should be writable
>> at the described points of time. They are read-only right after kernel init
>> and writable temporarily only during module_init/exit and dynamic
>> de/registration for a subsystem.
>> 
>> Signed-off-by: Hoeun Ryu <hoeun.ryu@...il.com>
>> ---
>> include/asm-generic/sections.h    |  1 +
>> include/asm-generic/vmlinux.lds.h | 10 ++++++++++
>> include/linux/cache.h             | 11 +++++++++++
>> 3 files changed, 22 insertions(+)
>> 
>> diff --git a/include/asm-generic/sections.h b/include/asm-generic/sections.h
>> index 4df64a1..16a6f21 100644
>> --- a/include/asm-generic/sections.h
>> +++ b/include/asm-generic/sections.h
>> @@ -34,6 +34,7 @@ extern char __bss_start[], __bss_stop[];
>> extern char __init_begin[], __init_end[];
>> extern char _sinittext[], _einittext[];
>> extern char __start_data_ro_after_init[], __end_data_ro_after_init[];
>> +extern char __start_data_ro_mostly_after_init[], __end_data_ro_mostly_after_init[];
>> extern char _end[];
>> extern char __per_cpu_load[], __per_cpu_start[], __per_cpu_end[];
>> extern char __kprobes_text_start[], __kprobes_text_end[];
>> diff --git a/include/asm-generic/vmlinux.lds.h b/include/asm-generic/vmlinux.lds.h
>> index 4e09b28..cc5f44e 100644
>> --- a/include/asm-generic/vmlinux.lds.h
>> +++ b/include/asm-generic/vmlinux.lds.h
>> @@ -265,6 +265,15 @@
>>        __end_data_ro_after_init = .;
>> #endif
>> 
>> +#ifndef RO_MOSTLY_AFTER_INIT_DATA
>> +#define RO_MOSTLY_AFTER_INIT_DATA(align)                               \
>> +       . = ALIGN(align);                                               \
>> +       VMLINUX_SYMBOL(__start_data_ro_mostly_after_init) = .;          \
>> +       *(.data..ro_mostly_after_init)                                  \
>> +       . = ALIGN(align);                                               \
>> +       VMLINUX_SYMBOL(__end_data_ro_mostly_after_init) = .;
>> +#endif
>> +
>> /*
>>  * Read only Data
>>  */
>> @@ -275,6 +284,7 @@
>>                *(.rodata) *(.rodata.*)                                 \
>>                RO_AFTER_INIT_DATA      /* Read only after init */      \
>>                KEEP(*(__vermagic))     /* Kernel version magic */      \
>> +               RO_MOSTLY_AFTER_INIT_DATA(align)                        \
> 
> You can't really drop this in the middle of a section like this. On
> arm64, we try very hard to use a minimal segment alignment of 64 KB
> (of 2 MB if DEBUG_ALIGN_RODATA=y), to ensure that the TLB footprint of
> the kernel image is minimized.
> 
> So this should be a separate section in the arm64 linker script.
> 

Could we achieve that using ALIGN(SECTION_SIZE) not ALIGN(PAGE_SIZE) for RO_DATA ?

>>                . = ALIGN(8);                                           \
>>                VMLINUX_SYMBOL(__start___tracepoints_ptrs) = .;         \
>>                KEEP(*(__tracepoints_ptrs)) /* Tracepoints: pointer array */ \
>> diff --git a/include/linux/cache.h b/include/linux/cache.h
>> index 1be04f8..fd1cb9b 100644
>> --- a/include/linux/cache.h
>> +++ b/include/linux/cache.h
>> @@ -30,6 +30,17 @@
>> #define __ro_after_init __attribute__((__section__(".data..ro_after_init")))
>> #endif
>> 
>> +/*
>> + * __ro_mostly_after_init is almost like __ro_after_init.
>> + * but __ro_mostly_after_init section is temporarily writable only during
>> + * module_init/exit or dynamic de/registeration of a subsystem using
>> + * set_ro_mostly_after_init_rw/ro pair.
>> + */
>> +#ifndef __ro_mostly_after_init
>> +#define __ro_mostly_after_init \
>> +       __attribute__((__section__(".data..ro_mostly_after_init")))
>> +#endif
>> +
>> #ifndef ____cacheline_aligned
>> #define ____cacheline_aligned __attribute__((__aligned__(SMP_CACHE_BYTES)))
>> #endif
>> --
>> 2.7.4

Powered by blists - more mailing lists

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.