[pr/110] RFR: 8317812: [Lilliput] Make C2 LoadNKlassCompactHeader more robust

[Aleksey Shipilev]

On Tue, 10 Oct 2023 19:08:35 GMT, Aleksey Shipilev <shade at openjdk.org> wrote:

>> Lilliput's C2 code for generating LoadNKlass currently assumes that the disp of the incoming address is klass_offset_in_bytes. It then extracts the base register and loads from the mark_offset_in_bytes instead.
>> Sometimes (apparently very rarely) it happens that C2 emits code that pre-adds obj+klass_offset into a register, and uses that as base, but with offset 0. In this case we would trip the assert or crash in release builds.
>> 
>> - [x] specjvm derby (which triggered the bug)
>> - [x] tier1 x86_64 +UCOH
>> - [x] tier1 aarch64 +UCOH
>
> src/hotspot/cpu/x86/x86_64.ad line 5336:
> 
>> 5334:   format %{ "movl    $dst, $mem\t# compressed klass ptr" %}
>> 5335:   ins_encode %{
>> 5336:     Register index = $mem$$index != 4 ? $mem$$index$$Register : noreg;
> 
> I don't understand what`$mem$$index != 4` is supposed to check. Can't we just pass the whole register to `load_nklass_compact_c2`? Is it already `noreg` when index is not set?

Or is unset index `-1`, not `4`?

-------------

PR Review Comment: https://git.openjdk.org/lilliput/pull/111#discussion_r1353197046