netdev - Re: [PATCH v3 for-next 01/13] IB/core: Use SRCU when reading client_list or device

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, 14 May 2015 13:35:36 +0300
From:	Haggai Eran <haggaie@...lanox.com>
To:	Jason Gunthorpe <jgunthorpe@...idianresearch.com>
CC:	Doug Ledford <dledford@...hat.com>, <linux-rdma@...r.kernel.org>,
	<netdev@...r.kernel.org>, Liran Liss <liranl@...lanox.com>,
	Guy Shapiro <guysh@...lanox.com>,
	Shachar Raindel <raindel@...lanox.com>,
	Yotam Kenneth <yotamke@...lanox.com>,
	Matan Barak <matanb@...lanox.com>
Subject: Re: [PATCH v3 for-next 01/13] IB/core: Use SRCU when reading client_list
 or device_list

On 13/05/2015 18:57, Jason Gunthorpe wrote:
> On Wed, May 13, 2015 at 11:10:15AM +0300, Haggai Eran wrote:
> 
>>>> I guess a similar thing we can do is to rely on the context we associate
>>>> with a pair of a client and a device. If such a context exist, we don't
>>>> need to call client->add again. What do you think?
>>>
>>> I didn't look closely, isn't this enough?
>>>
>>> device_register:
>>>  mutex_lock(client_mutex);
>>>  down_write(devices_rwsem);
>>>  list_add(device_list,dev,..);
>>>  up_write(devices_rwsem);
>>>
>>>  /* Caller must prevent device_register/unregister concurrancy on the
>>>     same dev */
>>>
>>>  foreach(client_list)
>>>    .. client->add(dev,...) .. 
>>>
>>>  mutex_unlock(client_mutex)
>>>
>>> client_register:
>>>  mutex_lock(client_mutex)
>>>  list_add(client_list,new_client..)
>>>  down_read(devices_rwsem);
>>>  foreach(device_list)
>>>    .. client->add(dev,new_client,..) ..
>>>  up_read(devices_rwsem);
>>>  mutex_unlock(client_mutex)
>>>
>>> [Note, I didn't check this carefully, just intuitively seems like a
>>>  sane starting point]
>>
>> That could perhaps work for the RoCEv2 patch-set, as their deadlock
>> relates to iterating the device list. This patch set however does an
>> iteration on the client list (patch 3). Because a client remove could
>> block on this iteration, you can still get a deadlock.
> 
> Really? Gross.
> 
> Still, I think you got it right in your analysis, but we don't need RCU:
> 
> device_register:
>  mutex_lock(modify_mutex);
>  down_write(lists_rwsem);
>  list_add(device_list,dev,..);
>  up_write(lists_rwsem);
> 
>  // implied by modify_mutex: down_read(lists_rwsem)
>  foreach(client_list)
>     .. client->add(dev,...) ..
>  mutex_unlock(modify_mutex)
> 
> client_register:
>  mutex_lock(modify_mutex);
>  // implied by modify_mutex: down_read(lists_rwsem)
>  foreach(device_list)
>     .. client->add(dev,new_client...) ..
> 
>  down_write(lists_rwsem);
>  list_add(client_list,new_client..);
>  up_write(lists_rwsem);
> 
>  mutex_unlock(modify_mutex)
> 
> client_unregister:
>  mutex_lock(modify_mutex);
>  down_write(lists_rwsem);
>  list_cut(client_list,..rm_client..);
>  up_write(lists_rwsem);
> 
>  // implied by modify_mutex: down_read(lists_rwsem)
>  foreach(device_list)
>     .. client->remove(dev,rm_client...) ..
> 
>  mutex_unlock(modify_mutex)
> 
> etc. Notice the ordering.
> 

Looks good. I'll use it in the next version of the patch.

--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html