]> git.ipfire.org Git - thirdparty/kernel/linux.git/commit
sched_ext: Separate bypass dispatch enabling from bypass depth tracking
authorTejun Heo <tj@kernel.org>
Fri, 6 Mar 2026 17:58:03 +0000 (07:58 -1000)
committerTejun Heo <tj@kernel.org>
Fri, 6 Mar 2026 17:58:03 +0000 (07:58 -1000)
commitaa2a0a19686c90106ade6a7e848ffbb62d55d733
tree8afbbc35bccc042033df4fdd7a93d1cbaf5abc47
parentd94d09a23340b343ffc6ad935f4a7fa90516684a
sched_ext: Separate bypass dispatch enabling from bypass depth tracking

The bypass_depth field tracks nesting of bypass operations but is also used
to determine whether the bypass dispatch path should be active. With
hierarchical scheduling, child schedulers may need to activate their parent's
bypass dispatch path without affecting the parent's bypass_depth, requiring
separation of these concerns.

Add bypass_dsp_enable_depth and bypass_dsp_claim to independently control
bypass dispatch path activation. The new enable_bypass_dsp() and
disable_bypass_dsp() functions manage this state with proper claim semantics
to prevent races. The bypass dispatch path now only activates when
bypass_dsp_enabled() returns true, which checks the new enable_depth counter.

The disable operation is carefully ordered after all tasks are moved out of
bypass DSQs to ensure they are drained before the dispatch path is disabled.
During scheduler teardown, disable_bypass_dsp() is called explicitly to ensure
cleanup even if bypass mode was never entered normally.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Andrea Righi <arighi@nvidia.com>
kernel/sched/ext.c
kernel/sched/ext_internal.h