[mono][sgen] Update memory load information after world is resumed

[BrzVlad]

Following recent cgroup support (#74237), mono_determine_physical_ram_size is no longer safe to call during GC.

[ghost]

Tagging subscribers to this area: @BrzVlad
See info in area-owners.md if you want to be subscribed.

Issue Details

---

Following recent cgroup support (#74237), mono_determine_physical_ram_size is no longer safe to call during GC.

Author:	BrzVlad
Assignees:	-
Labels:	area-GC-mono
Milestone:	-

[BrzVlad]

@nealef@akoeplinger Any idea if this impacts android ?

[naricc]

What part of #74237 makes it unsafe to call mono_determine_physical_ram_size during GC?

[nealef]

@nealef@akoeplinger Any idea if this impacts android ?

The new code explicitly excludes Android. That is, it uses the previous code path.

[BrzVlad]

@nariccgetCGroupMemoryLimit is the main one, it uses a lot of C api's that are unsafe.

[akoeplinger]

I didn't realize we query the ram size on each GC stw cycle, do we know if the cgroup additions have any perf impact?

We'll also need to backport this to 7.0 since we backported the cgroup change.

[naricc]

I didn't realize we query the ram size on each GC stw cycle, do we know if the cgroup additions have any perf impact?

I have not seen any perf regressions that I could attribute to this, although the microbenchmarks are so micro I am not sure how many of them exercise the GC much. There was also some missing data on some platforms around the time this was merged.

[akoeplinger]

Failures are unrelated.

[akoeplinger]

/backport to release/7.0

[github-actions]

Started backporting to release/7.0: https://github.com/dotnet/runtime/actions/runs/3174227750