ldscript.powerpc*: Only put .dynamic in PT_DYNAMIC

[jrtc27]

Summary

Currently there are a few output sections left as implicitly using
:kernel :dynamic before :kernel on its own is used again, which means
they end up in both the PT_LOAD and the PT_DYNAMIC segments, causing
kldxref to treat the kernel as invalid. Thus, hoist the :kernel to the
very next section to ensure only .dynamic is in PT_DYNAMIC.

Whilst here, sync ldscript.powerpc64le with ldscript.powerpc64 to pick
up various fixes that were presumably made between the start of the
powerpc64le port and it being committed and got missed.

[jhb]

Was this broken with the pre-libelf kldxref as well?

[jrtc27]

In D43066#981966, @jhb wrote:

Was this broken with the pre-libelf kldxref as well?

I don't have a system to test on, and don't hugely want to deal with QEMU for it, but it's a new check in the libelf kldxref, and since .dynamic is terminated by a DT_NULL entry it should have stopped at the right point in the too big segment. So I believe this is a latent oddity in the kernel that is newly exposed by kldxref.

[imp]

sys/conf/ldscript.powerpc

87

maybe we should simplify based on this comment?

[jhb]

In D43066#981967, @jrtc27 wrote:

In D43066#981966, @jhb wrote:

Was this broken with the pre-libelf kldxref as well?

I don't have a system to test on, and don't hugely want to deal with QEMU for it, but it's a new check in the libelf kldxref, and since .dynamic is terminated by a DT_NULL entry it should have stopped at the right point in the too big segment. So I believe this is a latent oddity in the kernel that is newly exposed by kldxref.

Hmm, I think it's the check I added that the SHT_DYNAMIC section has the same start/size as the PT_DYNAMIC segment. I could perhaps relax that to allow a larger PT_DYNAMIC segment size to support older kernels without your fix.

[jhb]

In D43066#982359, @jhb wrote:

In D43066#981967, @jrtc27 wrote:

In D43066#981966, @jhb wrote:

Was this broken with the pre-libelf kldxref as well?

I don't have a system to test on, and don't hugely want to deal with QEMU for it, but it's a new check in the libelf kldxref, and since .dynamic is terminated by a DT_NULL entry it should have stopped at the right point in the too big segment. So I believe this is a latent oddity in the kernel that is newly exposed by kldxref.

Hmm, I think it's the check I added that the SHT_DYNAMIC section has the same start/size as the PT_DYNAMIC segment. I could perhaps relax that to allow a larger PT_DYNAMIC segment size to support older kernels without your fix.

I've relaxed the check for powerpc in D43123. I think we still want this fix regardless.