RFR(S): 8251860: ClassLoaderData::loaded_classes_do fails with "assert(ZAddress::is_marked(addr)) failed: Should be marked"

Mon Aug 31 22:22:42 UTC 2020


On 8/31/20 11:57 AM, calvin.cheung at oracle.com wrote:
> Hi Ioi,
>
> Thanks for your review!
>
> On 8/28/20 3:54 PM, Ioi Lam wrote:
>> Hi Calvin,
>>
>>  300 void ClassLoaderData::inc_keep_alive() {
>>  301   if (has_class_mirror_holder()) {
>>  302     if (!Arguments::is_dumping_archive()) {
>>  303       assert(_keep_alive > 0, "Invalid keep alive increment 
>> count");
>>  304     }
>>  305     _keep_alive++;
>>  306   }
>>  307 }
>>
>> Why is 302 needed? I think in CollectCLDClosure, you wouldn't collect 
>> any CLD whose _keep_alive count is zero.
>
> The (!cld->is_unloading()) doesn't mean that the _keep_alive count is 
> non-zero.
>
> We currently have a few tests which would hit the assert if line 302 
> is not there. One of them is the appcds/dynamicArchive/LinkClassTest. 
> While the "Parent x = new Child()" is being verified during linking, 
> the VerificationType::resolve_and_check_assignability will be called. 
> This will eventually call the SystemDictionary::register_loader but it 
> will not add a class_loader to the ClassLoaderDataGraph. Prior to this 
> patch, the _keep_alive will be incremented at 2 places:
>
> 1) java_lang_Class::set_mirror_module_field
>
>    In this case, the _keep_alive will be incremented during vm startup.
>
> 2) ClassLoaderData::ClassLoaderData
>
>    _keep_alive((has_class_mirror_holder || h_class_loader.is_null()) ? 
> 1 : 0),
>
> Neither of the above code path will be taken for the assignability check.
>
>>
>> I think MetaspaceShared::link_and_cleanup_shared_classes() can be 
>> simplified so you don't need to scan the CLDs in two passes.
>>
>> 1393  while (true) {
>>         bool has_linked = false;
>>         for (int i = 0; i < _loaded_cld->length(); i++) {
>>           ClassLoaderData* cld = _loaded_cld->at(i);
>>           for (Klass* klass = cld->klasses(); klass != NULL; klass = 
>> klass->next_link()) {
>>             if (klass->is_instance_klass()) {
>>               InstanceKlass* ik = InstanceKlass::cast(ik);
>>               if (inking_required(ik)) {
>>                 has_linked = link_class_for_cds(ik, THREAD);
>>               }
>>             }
>>           }
>>         }
>>
>>         if (!has_linked) {
>>           break;
>>         }
>>         // Class linking includes verification which may load more 
>> classes.
>>         // Keep scanning until we have linked no more classes.
>>       }
>>
> I think the has_linked assignment should be "has_linked |= 
> link_class_for_cds(ik, THREAD)" ?
>
> The above single loop is cleaner but I think it is less efficient. For 
> the simple case if there's no linking required after the first loop. 
> It will need to do another full loop to check if all classes don't 
> need to be linked in order to exit the loop. With the 2 passes 
> approach, it only needs to inspect the first element of the _klasses 
> list of each CLD to determine if we can exit the loop. I'm good with 
> the single loop since CDS dump isn't too performance critical and it 
> makes the code easier to read.
>
>>
>> Also link_class_for_cds should return the value of 
>> MetaspaceShared::try_link_class(), so it has the same behavior as 
>> before.
>
> Fixed.
>
> Here's an updated webrev:
>
>     http://cr.openjdk.java.net/~ccheung/jdk16/8251860/webrev.02/

This looks good!
Coleen
>
> thanks,
> Calvin
>>
>> The rest of the changes looks good.
>>
>> Thanks
>> - Ioi
>>
>> On 8/28/20 10:42 AM, calvin.cheung at oracle.com wrote:
>>> Hi Coleen,
>>>
>>> Thanks for your review.
>>>
>>> I've re-implemented the fix with your suggestions.
>>>
>>> Here's an updated webrev:
>>>
>>> http://cr.openjdk.java.net/~ccheung/jdk16/8251860/webrev.01/
>>>
>>> On 8/26/20 3:24 PM, Coleen Phillimore wrote:
>>>>
>>>> Hi Calvin,
>>>>
>>>> I think you should change this to call 
>>>> ClassLoaderDataGraph::loaded_cld_do(&collect_cld) and collect 
>>>> ClassLoaderData in the GrowableArray, since you then walk through 
>>>> them again to link the classes in the class loader data.
>>> With the ClassLoaderDataGraph::loaded_cld_do approach, some klasses 
>>> in the _klasses list may already in the link state. So I have to 
>>> account for that and not link those klasses again. It complicates 
>>> the fix a little bit.
>>>>
>>>> Can linking one klass cause a new class to be added to a _klasses 
>>>> list of a CLD that you've already walked?  It seems like you need 
>>>> the do() loop in the current code wrapped around walking the clds, 
>>>> then link the classes until no progress. You can fast exit when you 
>>>> find a klass linked, I believe, since the classes are added to the 
>>>> front of the _klasses list.
>>> Yes, it is possible. The updated patch will check if the first klass 
>>> in the _klass list of each CLD hasn't been linked, then it will loop 
>>> back again to go through all the CLD's _klasses lists.
>>>>
>>>> Also with this change, you can remove locked_loaded_classes_do() 
>>>> because it was added for CDS and apparently the comment is wrong.
>>>>
>>>> // This case can block but cannot do unloading (called from CDS)
>>>> void ClassLoaderDataGraph::unlocked_loaded_classes_do(KlassClosure* 
>>>> klass_closure) {
>>>>   for (ClassLoaderData* cld = _head; cld != NULL; cld = cld->next()) {
>>>>     cld->loaded_classes_do(klass_closure);
>>>>   }
>>>> }
>>>
>>> I've removed the above.
>>>
>>> thanks,
>>>
>>> Calvin
>>>
>>>>
>>>> Thanks,
>>>> Coleen
>>>>
>>>> On 8/26/20 5:35 PM, calvin.cheung at oracle.com wrote:
>>>>> bug: https://bugs.openjdk.java.net/browse/JDK-8251860
>>>>>
>>>>> webrev: http://cr.openjdk.java.net/~ccheung/jdk16/8251860/webrev.00/
>>>>>
>>>>> Please refer to the bug report for an analysis of the crash.
>>>>>
>>>>> The proposed fix is to acquire the ClassLoaderDataGraph_lock 
>>>>> before calling 
>>>>> ClassLoaderDataGraph::loaded_classes_do(&link_closure). The 
>>>>> link_closure will not do the actual linking but just to store the 
>>>>> InstanceKlass'es in an array and increment the keep_alive counter 
>>>>> of the class_loader_data of each InstanceKlass. After the call to 
>>>>> ClassLoaderDataGraph::loaded_classes_do, each InstanceKlass in the 
>>>>> array will be linked and the keep_alive counter of the 
>>>>> class_loader_data will be decremented.
>>>>>
>>>>> Testing:
>>>>>     - tested about 40 times on the MacOS host where the crash was 
>>>>> seen
>>>>>        (before the fix, the crash was seen about once in 20 runs)
>>>>>     - Tier1 - 4.
>>>>> thanks,
>>>>> Calvin
>>>>
>>