RFR: 8016538: volatile double access via Unsafe.cpp is not atomic

[Christian Thalinger]

On Jul 10, 2013, at 3:46 AM, David Holmes <david.holmes at oracle.com> wrote:

> On 10/07/2013 2:03 PM, Yumin Qi wrote:
>> This is from comments I put on jbs:
>> The same test did not fail on solaris-i586, from disassembled code:
>> 0xfe9605ff: Unsafe_GetDoubleVolatile+0x00c7: addl 0x00000014(%ebp),%eax
>> 0xfe960602: Unsafe_GetDoubleVolatile+0x00ca: fldl (%eax)
>> 0xfe960604: Unsafe_GetDoubleVolatile+0x00cc: fstpl 0xffffffe0(%ebp)
>> 0xfe960607: Unsafe_GetDoubleVolatile+0x00cf: fldl 0xffffffe0(%ebp)
>> 0xfe96060a: Unsafe_GetDoubleVolatile+0x00d2: fstpl 0xffffffd0(%ebp)
>> 0xfe96060d: Unsafe_GetDoubleVolatile+0x00d5: movl 0x00000004(%edi),%eax
>> 0xfe960610: Unsafe_GetDoubleVolatile+0x00d8: movl %eax,0xffffffd8(%ebp)
>> 0xfe960613: Unsafe_GetDoubleVolatile+0x00db: movl 0x00000008(%edi),%ecx
>> 0xfe960616: Unsafe_GetDoubleVolatile+0x00de: cmpl $0x00000000,(%ecx)
>> 0xfe960619: Unsafe_GetDoubleVolatile+0x00e1: je
>> Unsafe_GetDoubleVolatile+0x100 [ 0xfe960638, .+0x1f ]
>> 
>> It is using fldl/fstpl for return *p;
> 
> Interesting. Thanks,

But this is more a compile issue than a Linux issue, right?  What if someone builds on Solaris or Mac OS X with plain GCC?

-- Chris

> 
> David
> 
>> 
>> Thanks
>> Yumin
>> 
>> On 7/9/2013 5:40 PM, David Holmes wrote:
>>> Hi Yumin,
>>> 
>>> On 10/07/2013 4:32 AM, Yumin Qi wrote:
>>>> Hi,
>>>> 
>>>>   Can i have your codereview of this change:
>>>>   Unsafe_Get[SET]DoubleVolatile are not atomic on 32bit x86 linux
>>>> caused test for
>>>> org.openjdk.concurrent.torture.tests.atomicity.primitives.reflect.DoubleAtomicityTest
>>>> 
>>>> failed.
>>>>   The fix is using same treatment for jdouble as for jlong. Tests
>>>> passed on failed test case, vm quick tests. It only happens on linux
>>>> 32bit platform, on solaris, the inlined code be compiled into code of
>>>> using fldl/fstpl which are atomic instructions.
>>> 
>>> Why is this not also a problem on Solaris? What does the Solaris
>>> compiler generate for
>>> 
>>> inline jdouble  OrderAccess::load_acquire(volatile jdouble* p) {
>>> return *p; }
>>> 
>>>>   Contributed by: dholmes
>>> 
>>> Guess that means I can't Review it :)
>>> 
>>> David
>>> -----
>>> 
>>>>   URL: http://cr.openjdk.java.net/~minqi/8016538/01
>>>> 
>>>>   Thanks
>>>>   Yumin
>>