RFR: Fix ShFreeSet boundary case

[Roman Kennke]

Am 10.05.2018 um 17:05 schrieb Aleksey Shipilev:
> On 05/10/2018 04:42 PM, Aleksey Shipilev wrote:
>> On 05/09/2018 09:30 PM, Roman Kennke wrote:
>>> Observed with the new PLAB code. Apparently this triggers the situation fairly often because it
>>> tends to allocate region sized GCLABs.
> But this is also suspicious: we used to cap the TLAB sizes below region size to avoid fragmentation.
> This does not happen with PLABs?

Apparently not.

> Should probably adjust that part.

Probably. I suggest to experiment with the new plab stuff once it's in
and fine-tune it.

Roman