Code Review Request, JDK-8219991 : New fix of the deadlock in sun.security.ssl.SSLSocketImpl

Daniel Fuchs daniel.fuchs at oracle.com
Tue May 7 08:26:54 UTC 2019


Hi Xuelei,

looks good to me as well.

best regards,

-- daniel

On 05/05/2019 16:18, Xuelei Fan wrote:
> All good catches!
> 
> I made the update accordingly.  Here is the new webrev:
>    http://cr.openjdk.java.net/~xuelei/8219991/webrev.03/
> 
> Thanks,
> Xuelei
> 
> On 5/3/2019 11:27 PM, Alan Bateman wrote:
>> On 04/05/2019 02:23, Xuelei Fan wrote:
>>> Hi,
>>>
>>> There is a surprise in the regression test.  I made the update, and 
>>> now the testing passed.  Here is the new webrev:
>>>     http://cr.openjdk.java.net/~xuelei/8219991/webrev.01/
>> I assume you can save two volatile writes by not initializing 
>> isClosing and hasDepleted to false (their initial value).
>>
>> If readLockedDeplete were to throw an exception (and there several 
>> opportunities for exceptions in that metho) then read would unwind 
>> without unlocking. You may need to structure it to reduce that 
>> possibility.
>>
>> -Alan
>>
>>



More information about the security-dev mailing list