RFR: 8207345: AArch64: Trampoline generation code reads from uninitialized memory
Andrew Haley
aph at redhat.com
Mon Jul 16 17:48:23 UTC 2018
On 07/16/2018 05:53 PM, Aleksey Shipilev wrote:
> On 07/16/2018 06:42 PM, Aleksey Shipilev wrote:
>> On 07/16/2018 06:40 PM, Andrew Haley wrote:
>>> On 07/16/2018 05:38 PM, Aleksey Shipilev wrote:
>>>> Your choice, really. I used to frown about conditionals like this all around Hotspot, but then I
>>>> started to see the appeal: you can clearly see (A && B && C) here, even if A, B, C are not trivial.
>>>> But structurally, you immediately grasp that if A fails, we don't enter the block; if B fails, we
>>>> don't enter the block, if C fails, we don't enter the block. It takes some time to read through
>>>> chain of local variables to reach the same verdict.
>>>
>>> I see. I tried it, and I think your suggestions do help, and the
>>> hybrid result is best of all:
>>>
>>> // We need a trampoline if branches are far.
>>> if (far_branches()) {
>>> // We don't want to emit a trampoline if C2 is generating dummy
>>> // code during its branch shortening phase.
>>> CompileTask* task = ciEnv::current()->task();
>>> if (task != NULL && is_c2_compile(task->comp_level()) &&
>>> Compile::current->in_scratch_emit_size()) {
>>> address stub = emit_trampoline_stub(offset(), entry.target());
>>> if (stub == NULL) {
>>> return NULL; // CodeCache is full
>>> }
>>> }
>>
>> Looks good.
>
> ...except that it should be !Compile::current->in_scratch_emit_size()? (I made this mistake when
> rewriting, sorry).
LOL! It's simpler because it's actually wrong! :-)
Using the form above, it actually should be
// We need a trampoline if branches are far.
if (far_branches()) {
// We don't want to emit a trampoline if C2 is generating dummy
// code during its branch shortening phase.
CompileTask* task = ciEnv::current()->task();
if (task != NULL
&& ! (is_c2_compile(task->comp_level())
&& Compile::current()->in_scratch_emit_size())) {
address stub = emit_trampoline_stub(offset(), entry.target());
if (stub == NULL) {
return NULL; // CodeCache is full
}
}
}
i.e. we want a trampoline if we're using C1 or (we're using C2 and not in scratch emit).
Back to Plan A, I think. :-)
--
Andrew Haley
Java Platform Lead Engineer
Red Hat UK Ltd. <https://www.redhat.com>
EAC8 43EB D3EF DB98 CC77 2FAD A5CD 6035 332F A671
More information about the hotspot-compiler-dev
mailing list