This is my third attempt to send padata cpumasks patchset.
The patchset includes fixes of all unclear things Steffen noted in previous two patchsets.
Changes:
1) Make two cpumasks in padata instead of one. The first cpumask is used by parallel workers and
another is used by the workers doing serialization. Two distinguish cpumasks perform to build
configuration where CPUs used by parallel and serial workers aren't intersect. It significantly
improves performance.
Each padata instance now includes notifier chain which can be used by users interested in instance's
cpumask(serial or parallel) change. If one of cpumask is changed an event is generated.
2) Add sysfs primitives to padata. Each padata instance contains kobject which can be embedded to any
proper sysfs hierarchy. Padata kobject can be used to change or show serial or parallel cpumask.
3) Add sysfs representation to pcrypt. Pcrypt now creates /sys/kernel/pcrypt/[pencrypt|pdecrypt] during
module loading phase. pencrypt and pdecrypt directories are represented by kobjects of padata instances
that belongs to pencrypt and pdecrypt respectively. Using this sysfs interface user can change and read
serial and parallel cpumasks of both instances.
--
W.B.R.
Dan Kruchinin
--
