[AArch64] Don't generate GOT entry for large model TLS LE relocation

[Jiong Wang]

On 23/12/15 18:10, Marcus Shawcroft wrote:
> On 23 December 2015 at 16:55, Jiong Wang <jiong.wang@foss.arm.com> wrote:
>
>> This patch revert the following commit
>> commit b7a944fea3a0194d81f6de4d958f3a1d2c6ad03a
>> Author: Renlin Li <renlin.li@arm.com>
>> Date:   Fri Oct 2 17:02:53 2015 +0100
> I have not looked at this in detail, I'm out of the office at the
> moment, but I did notice that this is more than a revert!
>
> Please split this into two patches.
>
> 1) The revert request.
> 2) The new functionality you want to add.
>
> I think the revert is safe because it is a relaxation, leaving the
> original sequence in place should be fine, also it only affects large
> model and the default model is small. I think the patch being reverted
> here may have had test cases submitted later as a separate patch?

Yes. the patch being reverted was trying the fix bug exposed by testcase

  ld-aarch64/tls-relax-large-gd-le

which is submitted in the next commit. So my concern was a pure revert 
of the
patch will actually cause regression on above testcase. Thus I 
integrated the
correct fix from my understanding to make sure no regression will happen.

>
> The revert part of this is IMO going to be suitable for 2.26, however
> I don;t think we should be taking the new functionality part on 2.26,
> at least not until it has baked on the trunk for a while.
>
> Cheers
> /Marcus