[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <a15d43b5-6806-9f12-6b5e-d17c44bd953c@virtuozzo.com>
Date: Tue, 14 Nov 2017 20:25:41 +0300
From: Kirill Tkhai <ktkhai@...tuozzo.com>
To: Eric Dumazet <eric.dumazet@...il.com>
Cc: davem@...emloft.net, vyasevic@...hat.com,
kstewart@...uxfoundation.org, pombredanne@...b.com,
vyasevich@...il.com, mark.rutland@....com,
gregkh@...uxfoundation.org, adobriyan@...il.com, fw@...len.de,
nicolas.dichtel@...nd.com, xiyou.wangcong@...il.com,
roman.kapl@...go.com, paul@...l-moore.com, dsahern@...il.com,
daniel@...earbox.net, lucien.xin@...il.com,
mschiffer@...verse-factory.net, rshearma@...cade.com,
linux-kernel@...r.kernel.org, netdev@...r.kernel.org,
ebiederm@...ssion.com, avagin@...tuozzo.com, gorcunov@...tuozzo.com
Subject: Re: [PATCH] net: Convert net_mutex into rw_semaphore and down read it
on net->init/->exit
On 14.11.2017 20:07, Eric Dumazet wrote:
> On Tue, 2017-11-14 at 16:53 +0300, Kirill Tkhai wrote:
>> Curently mutex is used to protect pernet operations list. It makes
>> cleanup_net() to execute ->exit methods of the same operations set,
>> which was used on the time of ->init, even after net namespace is
>> unlinked from net_namespace_list.
>>
> ...
>
>> The solution is to convert net_mutex to the rw_semaphore. Then,
>> pernet_operations::init/::exit methods, modifying the net-related data,
>> will require down_read() locking only, while down_write() will be used
>> for changing pernet_list.
>>
> ...
>
>> This patch requires commit 76f8507f7a64 "locking/rwsem: Add down_read_killable()"
>> from Linus tree (not in net-next yet).
>
>
> Looks great, thanks for doing this.
>
> I wonder if the down_read_killable() is really needed, maybe a
> down_read() would not be blocking the thread too long after this change.
I'm afraid there are spin_unlock() or similar actions leading to preempt_enable(),
which are made under the mutex/new rw_sem. They may be preempted and leave a cpu
for some time. So the possibility to break the wait may be useful in some way...
Powered by blists - more mailing lists