> On Mar 29, 2016, at 03:37, Ilya Dryomov <idryomov@xxxxxxxxx> wrote:
>
> On Fri, Mar 25, 2016 at 4:05 AM, Yan, Zheng <zyan@xxxxxxxxxx> wrote:
>>
>>> On Mar 24, 2016, at 21:55, Ilya Dryomov <idryomov@xxxxxxxxx> wrote:
>>>
>>> On Thu, Mar 24, 2016 at 2:18 PM, Yan, Zheng <zyan@xxxxxxxxxx> wrote:
>>>>
>>>>> On Mar 24, 2016, at 18:26, Ilya Dryomov <idryomov@xxxxxxxxx> wrote:
>>>
>>> [ snip ]
>>>
>>>>> TBH I still can't wrap my head around lockless layout/namespace pointer
>>>>> updates and how that can ever work reliably…
>>>>
>>>> see Documentation/RCU/rcuref.txt. How RCU is used in my code is similar to the second example.
>>>
>>> It's the higher-level cephfs code and installing new namespaces in the
>>> middle of whatever else that might be going on that I'm worried about.
>>> See my other mail, I'd like to hear what exactly is being achieved with
>>> this approach and have you considered reusing i_ceph_lock (or others,
>>> of which there are plenty).
>>
>> Updating namespace pointer is protected by i_ceph_lock, For read
>> access of the namespace, using i_ceph_lock does not provide any extra
>> guarantee. This approach provides lockless access of namespace, it’s
>> convenient for functions that do not know i_ceph_lock.
>
> Could you name some of these functions?
ceph_osdc_{new_request,readpages,write pages}
>
> Thanks,
>
> Ilya
--
To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
