RFR(S) 8017498: JVM crashes when native code calls sigaction(sig) where sig>=0x20
Calvin Cheung
calvin.cheung at oracle.com
Tue Jul 9 16:10:23 PDT 2013
Thanks for your review.
Calvin
On 7/9/2013 3:08 PM, Karen Kinnear wrote:
> Calvin,
>
> The changes look good - thanks for fixing this.
> I agree with David - the likelihood of hanging depends on what you
> overwrite, so not hanging is good but not sufficient. That said, the
> new test is still useful.
>
> thanks,
> Karen
>
> On Jul 8, 2013, at 6:57 PM, Calvin Cheung wrote:
>
>> Hi David,
>>
>> Thanks for your review.
>> I've updated the webrev at the same location:
>> http://cr.openjdk.java.net/~ccheung/8017498/webrev/
>>
>> I've re-run the vm.quick, vm.signal, and regression test on linux_64.
>> I've also contacted the original submitter of this bug.
>>
>> thanks,
>> Calvin
>>
>> On 7/7/2013 6:23 PM, David Holmes wrote:
>>> HI Calvin,
>>>
>>> On 6/07/2013 5:53 AM, Calvin Cheung wrote:
>>>> Please review a small fix:
>>>> http://cr.openjdk.java.net/~ccheung/8017498/webrev/
>>>>
>>>> Bug:
>>>> http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=8017498
>>>> https://jbs.oracle.com/bugs/browse/JDK-8017498
>>>>
>>>> Summary:
>>>> I couldn't reproduce the crash as stated in the bug report.
>>>> However, I saw hang in the native code when sig value is much
>>>> greater than 32, e.g. 132. Fix suggested by David H. is to add
>>>> another check for the sig value is less than MAXSIGNUM in jsig.c.
>>> I don't think this is quite enough. I think the second clause:
>>>
>>> 119 } else if (jvm_signal_installing) {
>>>
>>> would also need to be
>>>
>>> 119 } else if (sig < MAXSIGNUM && jvm_signal_installing) {
>>>
>>>> With the fix, the hang is gone and the regression test is for
>>>> testing the hang condition.
>>> We need to reach out to the original reporter and ensure that this fixes their problem.
>>>
>>> Thanks,
>>> David
>>>
>>>> Testing:
>>>> JPRT
>>>> jtreg test on test/runtime with the new regression test
>>>> Aurora adhoc vm.signal and vm.quick tests on linux
>>>> (in progress, no failure so far)
>>>>
>>>> thanks,
>>>> Calvin
More information about the hotspot-runtime-dev
mailing list