include/linux/fs.h: disable preempt when acquire i_size_seqcount writ…

…e lock

commit 74e3d1e17b2e11d175970b85acd44f5927000ba2 upstream.

Two rt tasks bind to one CPU core.

The higher priority rt task A preempts a lower priority rt task B which
has already taken the write seq lock, and then the higher priority rt
task A try to acquire read seq lock, it's doomed to lockup.

rt task A with lower priority: call write
i_size_write                                        rt task B with higher priority: call sync, and preempt task A
  write_seqcount_begin(&inode->i_size_seqcount);    i_size_read
  inode->i_size = i_size;                             read_seqcount_begin <-- lockup here...

So disable preempt when acquiring every i_size_seqcount *write* lock will
cure the problem.

Signed-off-by: Fan Du <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Cc: Zhao Hongjiang <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

include/linux/fs.h

@@ -915,9 +915,11 @@ static inline loff_t i_size_read(const struct inode *inode)
 static inline void i_size_write(struct inode *inode, loff_t i_size)
 {
 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
+	preempt_disable();
 	write_seqcount_begin(&inode->i_size_seqcount);
 	inode->i_size = i_size;
 	write_seqcount_end(&inode->i_size_seqcount);
+	preempt_enable();
 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPT)
 	preempt_disable();
 	inode->i_size = i_size;