scx_lavd: cur_logical_clk update will be lost when concurrent updating

[archerbroler]

In advance_cur_logical_clk, when multiple CPU run lavd_running the same time, the cur_logical_clk will lost update.
Each CPU will read the same cur_logical_clk, and update with their own taskc->vdeadline_log_clk. Then cur_logical_clk result will be the largest one, which make the cur_logical_clk's value not meaningful.
I think update to cur_logical_clk' should use somethin like this:

__sync_fetch_and_add(&cur_logical_clk, taskc->vdeadline_delta_ns);

Perhaps this is a reason to remove nr_queued as I mentioned in #893

[multics69]

That's a good finding. I should update it using CAS. Will fix it soon. Thank you @archerbroler !