linux-kernel - Re: [PATCH 1/2] Allow constructor name selection by architecture.

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: <CAE1uF8dCy3m3TO4xa6Cc3i-CM3QYUFADZH_6o+hgy6MSPEWnyQ@mail.gmail.com>
Date:	Mon, 1 Apr 2013 18:21:51 -0400
From:	"George G. Davis" <gdavis@...sta.com>
To:	Sam Ravnborg <sam@...nborg.org>
Cc:	Vincent Sanders <vincent.sanders@...labora.co.uk>,
	Arnd Bergmann <arnd@...db.de>, linux-kernel@...r.kernel.org,
	Vincent Sanders <vince@...labora.co.uk>,
	Andrew Morton <akpm@...ux-foundation.org>,
	linux-arm-kernel@...ts.infradead.org,
	Russell King - ARM Linux <linux@....linux.org.uk>
Subject: Re: [PATCH 1/2] Allow constructor name selection by architecture.

On Mon, Apr 1, 2013 at 5:58 PM, Sam Ravnborg <sam@...nborg.org> wrote:
> On Mon, Apr 01, 2013 at 05:47:38PM -0400, George G. Davis wrote:
>> On Jun 6, 2012, at 6:12 AM, Russell King - ARM Linux wrote:
>> > On Tue, May 29, 2012 at 10:06:14AM +0100, Russell King - ARM Linux wrote:
>> >> On Mon, May 28, 2012 at 10:30:05PM +0200, Sam Ravnborg wrote:
>> >>> On Mon, May 28, 2012 at 07:33:37PM +0100, Vincent Sanders wrote:
>> >>>> From: Vincent Sanders <vince@...labora.co.uk>
>> >>>>
>> >>>> The constructor symbol name is different between platforms. Allow this
>> >>>> to be selected by configuration and set suitable default values.
>> >>>>
>> >>>> Signed-off-by: Vincent Sanders <vincent.sanders@...labora.co.uk>
>> >>>> ---
>> >>>> include/asm-generic/vmlinux.lds.h |    6 +++---
>> >>>> init/Kconfig                      |    6 ++++++
>> >>>> kernel/module.c                   |    2 +-
>> >>>> 3 files changed, 10 insertions(+), 4 deletions(-)
>> >>>>
>> >>>> diff --git a/include/asm-generic/vmlinux.lds.h b/include/asm-generic/vmlinux.lds.h
>> >>>> index 8aeadf6..fd34808 100644
>> >>>> --- a/include/asm-generic/vmlinux.lds.h
>> >>>> +++ b/include/asm-generic/vmlinux.lds.h
>> >>>> @@ -471,9 +471,9 @@
>> >>>>  }
>> >>>>
>> >>>> #ifdef CONFIG_CONSTRUCTORS
>> >>>> -#define KERNEL_CTORS()  . = ALIGN(8);                      \
>> >>>> -                        VMLINUX_SYMBOL(__ctors_start) = .; \
>> >>>> -                        *(.ctors)                          \
>> >>>> +#define KERNEL_CTORS()  . = ALIGN(8);                                   \
>> >>>> +                        VMLINUX_SYMBOL(__ctors_start) = .;              \
>> >>>> +                        *(CONFIG_CONSTRUCTORS_NAME)                     \
>> >>>>                  VMLINUX_SYMBOL(__ctors_end) = .;
>> >>>
>> >>> What is wrong with adding both "standard" names for ctors uncnditionally?
>> >>> Like this:
>> >>>>                  *(.ctors)                          \
>> >>>> +                        *(.init_array)                     \
>> >>
>> >> That doesn't get rid of CONFIG_CONSTRUCTORS_NAME, because it's needed
>> >> in the module code.  Do you have a suggestion to solve that as well?
>> >
>> > Ping.
>>
>> Pinging this back to life.  I'd like to see GCOV for ARM eABI finally make it upstream.
>> So, any objections to this?  Should it be resubmitted?
>
> Why is CONFIG_CONSTRUCTORS_NAME needed in module code?

Because ARM eABI uses the .init_array section for C++ constructors,
rather than .ctors.
So the patch defined CONFIG_CONSTRUCTORS_NAME to set the correct section name
for ARM eABI while leaving non-ARM-eABI as before, .ctors.  Here are the changes
from the patch for reference:

diff --git a/init/Kconfig b/init/Kconfig
index 6cfd71d..52181a1 100644
--- a/init/Kconfig
+++ b/init/Kconfig
@@ -20,6 +20,12 @@ config CONSTRUCTORS
        bool
        depends on !UML

+config CONSTRUCTORS_NAME
+       string
+       depends on CONSTRUCTORS
+       default ".init_array" if ARM && AEABI
+       default ".ctors"
+
 config HAVE_IRQ_WORK
        bool

diff --git a/kernel/module.c b/kernel/module.c
index 78ac6ec..e5fad5e 100644
--- a/kernel/module.c
+++ b/kernel/module.c
@@ -2600,7 +2600,7 @@ static void find_module_sections(struct module
*mod, struct load_info *info)
        mod->unused_gpl_crcs = section_addr(info, "__kcrctab_unused_gpl");
 #endif
 #ifdef CONFIG_CONSTRUCTORS
-       mod->ctors = section_objs(info, ".ctors",
+       mod->ctors = section_objs(info, CONFIG_CONSTRUCTORS_NAME,
                                  sizeof(*mod->ctors), &mod->num_ctors);
 #endif

W/o the above, GCOV does not work on ARM eABI for kernel modules.
Meanwhile, it still works as before for non-ARM-eABI kernel modules.

Thanks!

--
Regards,
George
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/