netdev - Re: [PATCH] net: Make synchronize_net() be expedited only when it's really need

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:   Tue, 23 Jan 2018 18:29:47 +0300
From:   Kirill Tkhai <ktkhai@...tuozzo.com>
To:     Eric Dumazet <edumazet@...gle.com>
Cc:     Eric Dumazet <eric.dumazet@...il.com>,
        netdev <netdev@...r.kernel.org>,
        David Miller <davem@...emloft.net>,
        David Ahern <dsahern@...il.com>,
        Florian Westphal <fw@...len.de>,
        Xin Long <lucien.xin@...il.com>,
        Daniel Borkmann <daniel@...earbox.net>,
        mschiffer@...verse-factory.net, jakub.kicinski@...ronome.com,
        Vladislav Yasevich <vyasevich@...il.com>,
        Jiri Benc <jbenc@...hat.com>
Subject: Re: [PATCH] net: Make synchronize_net() be expedited only when it's
 really need

On 23.01.2018 18:12, Eric Dumazet wrote:
> On Tue, Jan 23, 2018 at 6:41 AM, Kirill Tkhai <ktkhai@...tuozzo.com> wrote:
>> Hi, Eric,
>>
>> thanks for your review.
>>
>> On 22.01.2018 20:15, Eric Dumazet wrote:
>>> On Mon, 2018-01-22 at 12:41 +0300, Kirill Tkhai wrote:
>>>> Commit be3fc413da9e "net: use synchronize_rcu_expedited()" introducing
>>>> synchronize_net() says:
>>>>
>>>>     >When we hold RTNL mutex, we would like to spend some cpu cycles but not
>>>>     >block too long other processes waiting for this mutex.
>>>>     >We also want to setup/dismantle network features as fast as possible at
>>>>     >boot/shutdown time.
>>>>     >This patch makes synchronize_net() call the expedited version if RTNL is
>>>>     >locked.
>>>>
>>>> At the time of the commit (May 23 2011) there was no possible to differ,
>>>> who is the actual owner of the mutex. Only the fact that it's locked
>>>> by someone at the moment. So (I guess) this is the only reason the generic
>>>> primitive mutex_is_locked() was used.
>>>>
>>>> But now mutex owner is available outside the locking subsystem and
>>>> __mutex_owner() may be used instead (there is an example in audit_log_start()).
>>>> So, let's make expensive synchronize_rcu_expedited() be used only
>>>> when a caller really owns rtnl_mutex().
>>>>
>>>> There are several possibilities to fix that. The first one is
>>>> to fix synchronize_net(), the second is to change rtnl_is_locked().
>>>>
>>>> I prefer the second, as it seems it's more intuitive for people
>>>> to think that rtnl_is_locked() is about current process, not
>>>> about the fact mutex is locked in general. Grep over kernel
>>>> sources just proves this fact:
>>>>
>>>> drivers/staging/rtl8723bs/os_dep/osdep_service.c:297
>>>> drivers/staging/rtl8723bs/os_dep/osdep_service.c:316
>>>>
>>>>         if (!rtnl_is_locked())
>>>>                 ret = register_netdev(pnetdev);
>>>>         else
>>>>                 ret = register_netdevice(pnetdev);
>>>>
>>>> drivers/staging/wilc1000/linux_mon.c:310
>>>>
>>>>      if (rtnl_is_locked()) {
>>>>              rtnl_unlock();
>>>>              rollback_lock = true;
>>>>      }
>>>>
>>>> Side effect of this patch is three BUGs in above examples
>>>> become fixed.
>>>>
>>>> Signed-off-by: Kirill Tkhai <ktkhai@...tuozzo.com>
>>>> ---
>>>>  net/core/rtnetlink.c |    2 +-
>>>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>>>
>>>> diff --git a/net/core/rtnetlink.c b/net/core/rtnetlink.c
>>>> index 16d644a4f974..a5ddf373ffa9 100644
>>>> --- a/net/core/rtnetlink.c
>>>> +++ b/net/core/rtnetlink.c
>>>> @@ -117,7 +117,7 @@ EXPORT_SYMBOL(rtnl_trylock);
>>>>
>>>>  int rtnl_is_locked(void)
>>>>  {
>>>> -    return mutex_is_locked(&rtnl_mutex);
>>>> +    return __mutex_owner(&rtnl_mutex) == current;
>>>>  }
>>>>  EXPORT_SYMBOL(rtnl_is_locked);
>>>>
>>>>
>>>
>>> Seems good to me, but this looks a net-next candidate to me.
>>
>> No objections. What for this may be need for net tree?! Only to fix
>> the staging drivers above. But AFAIR, staging drivers guarantees, which
>> the kernel gives, are that they may be compiled. If so, we do not need
>> this in net tree.
>>
>>> Note that this does not catch illegal uses from BH, where current is
>>> not related to our context of execution.
>>
>> It's true, but the patch is about reducing of synchronize_rcu_expedited()
>> calls.
> 
> You have not touched only this path, but all paths using ASSERT_RTNL()
> 
> This is why I think your patch would target net-next, not net tree.
> 
>> There was no an objective to limit area of the places, where
>> rtnl_is_locked() can be used. For me it looks like another logical change.
>> If we really need that, one more patch on top of this may be submitted.
>> But honestly, I can't imagine someone really needs that check.
> 
> I believe you missed ASSERT_RTNL(), used all over the place.

Not missed. I grepped all over the kernel source, and this is how BUGs
in staging drivers were found. I just can't believe we really need
this check. Ok, then how about something like this:

int rtnl_is_locked(void)
{
#ifdef CONFIG_DEBUG_KERNEL
	BUG_ON(!in_task());
#endif
        return __mutex_owner(&rtnl_mutex) == current;
}

CONFIG_DEBUG_KERNEL is because of rtnl_is_locked() is used widely,
and the check has only the debug purpose.

Kirill