[PATCH] kernel/module.c: Mark module state before set RO and NX regions

[PATCH] kernel/module.c: Mark module state before set RO and NX regions

[Jungseung Lee]

In some architectures like arm/arm64, set_memory_*() check module address
and state as well. Mark module state before set RO and NX regions for
the routine is passed.

It will fix wrong RO/NX protection for loadable kernel modules on arm/arm64.

Signed-off-by: Jungseung Lee <js07.lee@gmail.com>
---
 kernel/module.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/kernel/module.c b/kernel/module.c
index 3965511..7e7cc9f 100644
--- a/kernel/module.c
+++ b/kernel/module.c
@@ -3152,6 +3152,10 @@ static int complete_formation(struct module *mod, struct load_info *info)
 	/* This relies on module_mutex for list integrity. */
 	module_bug_finalize(info->hdr, info->sechdrs, mod);
 
+	/* Mark state as coming so strong_try_module_get() ignores us,
+	 * but kallsyms etc. can see us. */
+	mod->state = MODULE_STATE_COMING;
+
 	/* Set RO and NX regions for core */
 	set_section_ro_nx(mod->module_core,
 				mod->core_text_size,
@@ -3164,9 +3168,6 @@ static int complete_formation(struct module *mod, struct load_info *info)
 				mod->init_ro_size,
 				mod->init_size);
 
-	/* Mark state as coming so strong_try_module_get() ignores us,
-	 * but kallsyms etc. can see us. */
-	mod->state = MODULE_STATE_COMING;
 	mutex_unlock(&module_mutex);
 
 	blocking_notifier_call_chain(&module_notify_list,
-- 
1.9.1

[PATCH] kernel/module.c: Mark module state before set RO and NX regions

[Rusty Russell]

Jungseung Lee <js07.lee@gmail.com> writes:
> In some architectures like arm/arm64, set_memory_*() check module address
> and state as well. Mark module state before set RO and NX regions for
> the routine is passed.
>
> It will fix wrong RO/NX protection for loadable kernel modules on arm/arm64.

This partially reverts commit 4982223e51e8ea9d09bb33c8323b5ec1877b2b51
Author: Rusty Russell <rusty@rustcorp.com.au>
Date:   Wed May 14 10:54:19 2014 +0930

    module: set nx before marking module MODULE_STATE_COMING.

Laura Abbott (CC'd) was looking at an alternative fix for this.  Laura?

Thanks,
Rusty.

> Signed-off-by: Jungseung Lee <js07.lee@gmail.com>
> ---
>  kernel/module.c | 7 ++++---
>  1 file changed, 4 insertions(+), 3 deletions(-)
>
> diff --git a/kernel/module.c b/kernel/module.c
> index 3965511..7e7cc9f 100644
> --- a/kernel/module.c
> +++ b/kernel/module.c
> @@ -3152,6 +3152,10 @@ static int complete_formation(struct module *mod, struct load_info *info)
>  	/* This relies on module_mutex for list integrity. */
>  	module_bug_finalize(info->hdr, info->sechdrs, mod);
>  
> +	/* Mark state as coming so strong_try_module_get() ignores us,
> +	 * but kallsyms etc. can see us. */
> +	mod->state = MODULE_STATE_COMING;
> +
>  	/* Set RO and NX regions for core */
>  	set_section_ro_nx(mod->module_core,
>  				mod->core_text_size,
> @@ -3164,9 +3168,6 @@ static int complete_formation(struct module *mod, struct load_info *info)
>  				mod->init_ro_size,
>  				mod->init_size);
>  
> -	/* Mark state as coming so strong_try_module_get() ignores us,
> -	 * but kallsyms etc. can see us. */
> -	mod->state = MODULE_STATE_COMING;
>  	mutex_unlock(&module_mutex);
>  
>  	blocking_notifier_call_chain(&module_notify_list,
> -- 
> 1.9.1

[PATCH] kernel/module.c: Mark module state before set RO and NX regions

[Laura Abbott]

On 1/6/2015 8:16 PM, Rusty Russell wrote:
> Jungseung Lee <js07.lee@gmail.com> writes:
>> In some architectures like arm/arm64, set_memory_*() check module address
>> and state as well. Mark module state before set RO and NX regions for
>> the routine is passed.
>>
>> It will fix wrong RO/NX protection for loadable kernel modules on arm/arm64.
>
> This partially reverts commit 4982223e51e8ea9d09bb33c8323b5ec1877b2b51
> Author: Rusty Russell <rusty@rustcorp.com.au>
> Date:   Wed May 14 10:54:19 2014 +0930
>
>      module: set nx before marking module MODULE_STATE_COMING.
>
> Laura Abbott (CC'd) was looking at an alternative fix for this.  Laura?
>
> Thanks,
> Rusty.
>

Yes, my proposal was to stop using is_module_address and just bounds
check against the module ranges[1]. I got bogged down with other
tasks and hadn't submitted a patch for review yet.

Thanks,
Laura

[1] http://lists.infradead.org/pipermail/linux-arm-kernel/2014-December/311574.html


-- 
Qualcomm Innovation Center, Inc.
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
a Linux Foundation Collaborative Project

[PATCH] kernel/module.c: Mark module state before set RO and NX regions

[Jungseung Lee]

2015-01-08 2:55 GMT+09:00 Laura Abbott <lauraa@codeaurora.org>:
> On 1/6/2015 8:16 PM, Rusty Russell wrote:
>>
>> Jungseung Lee <js07.lee@gmail.com> writes:
>>>
>>> In some architectures like arm/arm64, set_memory_*() check module address
>>> and state as well. Mark module state before set RO and NX regions for
>>> the routine is passed.
>>>
>>> It will fix wrong RO/NX protection for loadable kernel modules on
>>> arm/arm64.
>>
>>
>> This partially reverts commit 4982223e51e8ea9d09bb33c8323b5ec1877b2b51
>> Author: Rusty Russell <rusty@rustcorp.com.au>
>> Date:   Wed May 14 10:54:19 2014 +0930
>>
>>      module: set nx before marking module MODULE_STATE_COMING.
>>
>> Laura Abbott (CC'd) was looking at an alternative fix for this.  Laura?
>>
>> Thanks,
>> Rusty.
>>
>
> Yes, my proposal was to stop using is_module_address and just bounds
> check against the module ranges[1]. I got bogged down with other
> tasks and hadn't submitted a patch for review yet.
>
I got it. I'll prepare bounds check patch for aarch32.
Thanks,

> Thanks,
> Laura
>
> [1]
> http://lists.infradead.org/pipermail/linux-arm-kernel/2014-December/311574.html
>
>
> --
> Qualcomm Innovation Center, Inc.
> Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
> a Linux Foundation Collaborative Project