[PATCH v2] ole32: Store proxy/stub CLSIDs per process, not per apartment.

[Zebediah Figura]

On 08/02/2017 04:32 AM, Huw Davies wrote:
> On Tue, Aug 01, 2017 at 04:54:45PM -0500, Zebediah Figura wrote:
>> v2: use proper synchronization; add an MTA test (thanks Sebastian)
>> Signed-off-by: Zebediah Figura <zfigura at codeweavers.com>
>> ---
>>   dlls/ole32/compobj.c         | 48 ++++++++++++++++++++++++++++++------------
>>   dlls/ole32/compobj_private.h |  1 -
>>   dlls/ole32/tests/compobj.c   | 50 ++++++++++++++++++++++++++++++++++++++++++--
>>   3 files changed, 83 insertions(+), 16 deletions(-)
>>
>> diff --git a/dlls/ole32/compobj.c b/dlls/ole32/compobj.c
>> index 60244485249..ee904adbc38 100644
>> --- a/dlls/ole32/compobj.c
>> +++ b/dlls/ole32/compobj.c
>> @@ -170,8 +170,20 @@ struct registered_psclsid
>>       struct list entry;
>>       IID iid;
>>       CLSID clsid;
>> +    OXID apartment_id;
>>   };
>>   
>> +static struct list RegisteredPSCLSIDList = LIST_INIT(RegisteredPSCLSIDList);
> 
> Please don't use CamelCase.
> 
>> +
>> +static CRITICAL_SECTION csRegisteredPSCLSIDList;
>> +static CRITICAL_SECTION_DEBUG psclsid_cs_debug =
>> +{
>> +    0, 0, &csRegisteredPSCLSIDList,
>> +    { &psclsid_cs_debug.ProcessLocksList, &psclsid_cs_debug.ProcessLocksList },
>> +      0, 0, { (DWORD_PTR)(__FILE__ ": csRegisteredPSCLSIDList") }
>> +};
>> +static CRITICAL_SECTION csRegisteredPSCLSIDList = { &psclsid_cs_debug, -1, 0, 0, 0, 0 };
>> +
>>   /*
>>    * This is a marshallable object exposing registered local servers.
>>    * IServiceProvider is used only because it happens meet requirements
>> @@ -622,7 +634,6 @@ static APARTMENT *apartment_construct(DWORD model)
>>   
>>       list_init(&apt->proxies);
>>       list_init(&apt->stubmgrs);
>> -    list_init(&apt->psclsids);
>>       list_init(&apt->loaded_dlls);
>>       apt->ipidc = 0;
>>       apt->refs = 1;
>> @@ -1173,15 +1184,22 @@ DWORD apartment_release(struct apartment *apt)
>>               stub_manager_int_release(stubmgr);
>>           }
>>   
>> -        LIST_FOR_EACH_SAFE(cursor, cursor2, &apt->psclsids)
>> +        EnterCriticalSection(&csRegisteredPSCLSIDList);
>> +
>> +        LIST_FOR_EACH_SAFE(cursor, cursor2, &RegisteredPSCLSIDList)
>>           {
>>               struct registered_psclsid *registered_psclsid =
>>                   LIST_ENTRY(cursor, struct registered_psclsid, entry);
>>   
>> -            list_remove(&registered_psclsid->entry);
>> -            HeapFree(GetProcessHeap(), 0, registered_psclsid);
>> +            if (registered_psclsid->apartment_id == apt->oxid)
>> +            {
>> +                list_remove(&registered_psclsid->entry);
>> +                HeapFree(GetProcessHeap(), 0, registered_psclsid);
>> +            }
>>           }
>>   
>> +        LeaveCriticalSection(&csRegisteredPSCLSIDList);
>> +
>>           /* if this assert fires, then another thread took a reference to a
>>            * stub manager without taking a reference to the containing
>>            * apartment, which it must do. */
> 
> 
> It seems it's more complicated than this.  Consider two already
> existing STAs, if the PSClsid is registered in STA1 which then leaves
> the apt, a call in STA2 to CoGetPSCLSID() still succeeds.
> 
> You'll want to expand the tests to cover things like this to really figure out
> what's going on.
> 
> Also note that the MTA -> MTA test isn't very convincing, since there is only
> one MTA and hence both threads are in the same apt.
> 
> Huw.
> 
> 

I see; I'll send an updated patch.