Merge tag 'sched-rt-2022-10-05' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

Pull preempt RT updates from Thomas Gleixner:
 "Introduce preempt_[dis|enable_nested() and use it to clean up various
  places which have open coded PREEMPT_RT conditionals.

  On PREEMPT_RT enabled kernels, spinlocks and rwlocks are neither
  disabling preemption nor interrupts. Though there are a few places
  which depend on the implicit preemption/interrupt disable of those
  locks, e.g. seqcount write sections, per CPU statistics updates etc.

  PREEMPT_RT added open coded CONFIG_PREEMPT_RT conditionals to
  disable/enable preemption in the related code parts all over the
  place. That's hard to read and does not really explain why this is
  necessary.

  Linus suggested to use helper functions (preempt_disable_nested() and
  preempt_enable_nested()) and use those in the affected places. On !RT
  enabled kernels these functions are NOPs, but contain a lockdep assert
  to validate that preemption is actually disabled to catch call sites
  which do not have preemption disabled.

  Clean up the affected code paths in mm, dentry and lib"

* tag 'sched-rt-2022-10-05' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  u64_stats: Streamline the implementation
  flex_proportions: Disable preemption entering the write section.
  mm/compaction: Get rid of RT ifdeffery
  mm/memcontrol: Replace the PREEMPT_RT conditionals
  mm/debug: Provide VM_WARN_ON_IRQS_ENABLED()
  mm/vmstat: Use preempt_[dis|en]able_nested()
  dentry: Use preempt_[dis|en]able_nested()
  preempt: Provide preempt_[dis|en]able_nested()

1 files changed, 2 insertions, 11 deletions

diff --git a/fs/dcache.c b/fs/dcache.c
index bb0c4d0038db..2ee8636016ee 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -2597,15 +2597,7 @@ EXPORT_SYMBOL(d_rehash);
 
 static inline unsigned start_dir_add(struct inode *dir)
 {
-	/*
-	 * The caller holds a spinlock (dentry::d_lock). On !PREEMPT_RT
-	 * kernels spin_lock() implicitly disables preemption, but not on
-	 * PREEMPT_RT.  So for RT it has to be done explicitly to protect
-	 * the sequence count write side critical section against a reader
-	 * or another writer preempting, which would result in a live lock.
-	 */
-	if (IS_ENABLED(CONFIG_PREEMPT_RT))
-		preempt_disable();
+	preempt_disable_nested();
 	for (;;) {
 		unsigned n = dir->i_dir_seq;
 		if (!(n & 1) && cmpxchg(&dir->i_dir_seq, n, n + 1) == n)
@@ -2618,8 +2610,7 @@ static inline void end_dir_add(struct inode *dir, unsigned int n,
 			       wait_queue_head_t *d_wait)
 {
 	smp_store_release(&dir->i_dir_seq, n + 2);
-	if (IS_ENABLED(CONFIG_PREEMPT_RT))
-		preempt_enable();
+	preempt_enable_nested();
 	wake_up_all(d_wait);
 }