linux-kernel - Re: [POC 09/23] livepatch: Handle race when livepatches are reloaded during a module load

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:   Wed, 22 Jan 2020 18:51:47 +0000
From:   Julien Thierry <jthierry@...hat.com>
To:     Petr Mladek <pmladek@...e.com>, Jiri Kosina <jikos@...nel.org>,
        Josh Poimboeuf <jpoimboe@...hat.com>,
        Miroslav Benes <mbenes@...e.cz>
Cc:     Joe Lawrence <joe.lawrence@...hat.com>,
        Kamalesh Babulal <kamalesh@...ux.vnet.ibm.com>,
        Nicolai Stange <nstange@...e.de>,
        live-patching@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [POC 09/23] livepatch: Handle race when livepatches are reloaded
 during a module load

Hi Petr,

On 1/17/20 3:03 PM, Petr Mladek wrote:
> klp_module_coming() might fail to load a livepatch module when
> the related livepatch gets reloaded in the meantime.
> 
> Detect this situation by adding a timestamp into struct klp_patch.
> local_clock is enough because klp_mutex must be released and taken
> several times during this scenario.
> 
> Signed-off-by: Petr Mladek <pmladek@...e.com>
> ---
>   include/linux/livepatch.h | 2 ++
>   kernel/livepatch/core.c   | 9 +++++----
>   2 files changed, 7 insertions(+), 4 deletions(-)
> 
> diff --git a/include/linux/livepatch.h b/include/linux/livepatch.h
> index a4567c17a9f2..feb33f023f9f 100644
> --- a/include/linux/livepatch.h
> +++ b/include/linux/livepatch.h
> @@ -155,6 +155,7 @@ struct klp_state {
>    * @obj_list:	dynamic list of the object entries
>    * @enabled:	the patch is enabled (but operation may be incomplete)
>    * @forced:	was involved in a forced transition
> + * ts:		timestamp when the livepatch has been loaded

Nit: Missing '@'.

>    * @free_work:	patch cleanup from workqueue-context
>    * @finish:	for waiting till it is safe to remove the patch module
>    */
> @@ -171,6 +172,7 @@ struct klp_patch {
>   	struct list_head obj_list;
>   	bool enabled;
>   	bool forced;
> +	u64 ts;
>   	struct work_struct free_work;
>   	struct completion finish;
>   };
> diff --git a/kernel/livepatch/core.c b/kernel/livepatch/core.c
> index 34e3ee2be7ef..8e693c58b736 100644
> --- a/kernel/livepatch/core.c
> +++ b/kernel/livepatch/core.c
> @@ -20,6 +20,7 @@
>   #include <linux/moduleloader.h>
>   #include <linux/completion.h>
>   #include <linux/memory.h>
> +#include <linux/sched/clock.h>
>   #include <asm/cacheflush.h>
>   #include "core.h"
>   #include "patch.h"
> @@ -854,6 +855,7 @@ static int klp_init_patch_early(struct klp_patch *patch)
>   	kobject_init(&patch->kobj, &klp_ktype_patch);
>   	patch->enabled = false;
>   	patch->forced = false;
> +	patch->ts = local_clock();
>   	INIT_WORK(&patch->free_work, klp_free_patch_work_fn);
>   	init_completion(&patch->finish);
>   
> @@ -1324,6 +1326,7 @@ int klp_module_coming(struct module *mod)
>   {
>   	char patch_name[MODULE_NAME_LEN];
>   	struct klp_patch *patch;
> +	u64 patch_ts;
>   	int ret = 0;
>   
>   	if (WARN_ON(mod->state != MODULE_STATE_COMING))
> @@ -1339,6 +1342,7 @@ int klp_module_coming(struct module *mod)
>   			continue;
>   
>   		strncpy(patch_name, patch->obj->patch_name, sizeof(patch_name));
> +		patch_ts = patch->ts;
>   		mutex_unlock(&klp_mutex);
>   
>   		ret = klp_try_load_object(patch_name, mod->name);
> @@ -1346,14 +1350,11 @@ int klp_module_coming(struct module *mod)
>   		 * The load might have failed because the patch has
>   		 * been removed in the meantime. In this case, the
>   		 * error might be ignored.
> -		 *
> -		 * FIXME: It is not fully proof. The patch might have be
> -		 * unloaded and loaded again in the mean time.
>   		 */
>   		mutex_lock(&klp_mutex);
>   		if (ret) {
>   			patch = klp_find_patch(patch_name);
> -			if (patch)
> +			if (patch && patch->ts == patch_ts)
>   				goto err;

If the timestamps differ, we have found the klp_patch, feels a bit of a 
waste to go through the list again without trying to load it right away.

Admittedly this is to solve a race condition which should not even 
happen very often...

>   			ret = 0;
>   		}
> 

Cheers,

-- 
Julien Thierry