On Tue, 19 May 2020 22:19:08 +0200 Sebastian Andrzej Siewior <bigeasy@xxxxxxxxxxxxx> wrote: > From: Ingo Molnar <mingo@xxxxxxxxxx> > > The various struct pagevec per CPU variables are protected by disabling > either preemption or interrupts across the critical sections. Inside > these sections spinlocks have to be acquired. > > These spinlocks are regular spinlock_t types which are converted to > "sleeping" spinlocks on PREEMPT_RT enabled kernels. Obviously sleeping > locks cannot be acquired in preemption or interrupt disabled sections. > > local locks provide a trivial way to substitute preempt and interrupt > disable instances. On a non PREEMPT_RT enabled kernel local_lock() maps > to preempt_disable() and local_lock_irq() to local_irq_disable(). > > Add swapvec_lock to protect the per-CPU lru_add_pvec and > lru_lazyfree_pvecs variables and rotate_lock to protect the per-CPU > lru_rotate_pvecs variable > > Change the relevant call sites to acquire these locks instead of using > preempt_disable() / get_cpu() / get_cpu_var() and local_irq_disable() / > local_irq_save(). > > There is neither a functional change nor a change in the generated > binary code for non PREEMPT_RT enabled non-debug kernels. > > When lockdep is enabled local locks have lockdep maps embedded. These > allow lockdep to validate the protections, i.e. inappropriate usage of a > preemption only protected sections would result in a lockdep warning > while the same problem would not be noticed with a plain > preempt_disable() based protection. > > local locks also improve readability as they provide a named scope for > the protections while preempt/interrupt disable are opaque scopeless. > > Finally local locks allow PREEMPT_RT to substitute them with real > locking primitives to ensure the correctness of operation in a fully > preemptible kernel. > No functional change. > > ... > > --- a/include/linux/swap.h > +++ b/include/linux/swap.h > @@ -12,6 +12,7 @@ > #include <linux/fs.h> > #include <linux/atomic.h> > #include <linux/page-flags.h> > +#include <linux/locallock.h> Could we please make these local_lock.h and local_lock_internal.h? Making the filenames different from everything else is just irritating! > + local_lock(swapvec_lock); It's quite peculiar that these operations appear to be pass-by-value. All other locking operations are pass-by-reference - spin_lock(&lock), not spin_lock(lock). This is what the eye expects to see and it's simply more logical - calling code shouldn't have to "know" that the locking operations are implemented as cpp macros. And we'd be in a mess if someone tried to convert these to real C functions. Which prompts the question: why were all these operations implemented in the processor anyway? afaict they could have been written in C.
