linux-kernel - Re: modules: set mod->state to GOING before going notifiers are called

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]

Date:	Thu, 10 Mar 2016 00:08:31 -0500
From:	Jessica Yu <jeyu@...hat.com>
To:	Rusty Russell <rusty@...tcorp.com.au>
Cc:	Josh Poimboeuf <jpoimboe@...hat.com>,
	Seth Jennings <sjenning@...hat.com>,
	Jiri Kosina <jikos@...nel.org>,
	Vojtech Pavlik <vojtech@...e.com>,
	Miroslav Benes <mbenes@...e.cz>,
	Petr Mladek <pmladek@...e.com>,
	Steven Rostedt <rostedt@...dmis.org>,
	live-patching@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: modules: set mod->state to GOING before going notifiers are
 called

+++ Rusty Russell [10/03/16 13:57 +1030]:
>Jessica Yu <jeyu@...hat.com> writes:
>> In load_module(), the going notifiers are called during error handling
>> when an error occurs after the coming notifiers have already been called.
>> However, a module's state is still MODULE_STATE_COMING when the going
>> notifiers are called in the error path. To be consistent, also set
>> mod->state to MODULE_STATE_GOING before calling the going notifiers.
>>
>> Signed-off-by: Jessica Yu <jeyu@...hat.com>
>> ---
>>  kernel/module.c | 3 +++
>>  1 file changed, 3 insertions(+)
>>
>> diff --git a/kernel/module.c b/kernel/module.c
>> index 1981ae0..9e80576 100644
>> --- a/kernel/module.c
>> +++ b/kernel/module.c
>> @@ -3494,6 +3494,9 @@ static int load_module(struct load_info *info, const char __user *uargs,
>>  	return do_init_module(mod);
>>
>>   coming_cleanup:
>> +	mutex_lock(&module_mutex);
>> +	mod->state = MODULE_STATE_GOING;
>> +	mutex_unlock(&module_mutex);
>>  	blocking_notifier_call_chain(&module_notify_list,
>>  				     MODULE_STATE_GOING, mod);
>
>Actually, reviewing this patch makes me realize it is wrong.
>
>We rely on the state of the module being MODULE_STATE_COMING here:
>
>        static inline int strong_try_module_get(struct module *mod)
>        {
>        	BUG_ON(mod && mod->state == MODULE_STATE_UNFORMED);
>        	if (mod && mod->state == MODULE_STATE_COMING)
>                	return -EBUSY;
>
>We will just have to document that the notifier can be called with
>a module in MODULE_STATE_COMING if it never succeeded its
>initialization.

Ah, thanks for catching that. I think I remember seeing this conditional and
assuming it wouldn't be a problem since GOING modules would fail in
try_module_get() (as it is does not pass the module_is_live() test) and
subsequently strong_try_module_get() would also fail.. But, I think I ought to
review how module states interact before making a change like this, so, please
ignore this patch.

Jessica