On 8/27/26 20:35, Zqiang wrote:
Which tree is your test based on? (rcu tree or linux-next tree)
I'm testing on the linux-next tree(next-20260821), regarding the commit
[1].
The step5 srcu_barrier() has been intercept the srcu callback which by step4
call_srcu() insert,
so at here, the srcu_barrier should not be return, and the step4 srcu callback
should not be run.
the cleanup_srcu_struct() also should not be called.
srcu_barrier() didn't return at that point, it returned at step 9 when
srcu_invoke_callbacks() (which was queued to rcu_gp_wq at step 3) calls
the barrier cb after finishing the real five __free_bus callbacks. Steps
3 through 9 all happen within the same jiffy, so the timer armed at step
6 for the next jiffy is still pending at step 9. Also, a stale n_cbs is
left because, although srcu_invoke_callbacks()'s invoking loop may have
finished at that point, the control flow has not yet reached the
rcu_segcblist_add_len(-len) call at the end of srcu_invoke_callbacks().
[1]
https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/commit/kernel/rcu/srcutree.c?id=78a38cbf6f20bc8247e93d1149f97c12dba9fbfb
Thanks
Sunho Park
