Re: [PATCH] mm: Replace verify_mm_writelocked() by lockdep_assert_held_exclusive()

[Vlastimil Babka]

On 12/13/18 8:38 PM, Waiman Long wrote:
> Using down_read_trylock() to check if a task holds a write lock on
> a rwsem is not reliable. A task can hold a read lock on a rwsem and
> down_read_trylock() can fail if a writer is waiting in the wait queue.
>
> So use lockdep_assert_held_exclusive() instead which can do the right
> check when CONFIG_LOCKDEP is on.
> 
> Signed-off-by: Waiman Long 

There's already a patch in mmotm removing this completely:
https://www.ozlabs.org/~akpm/mmots/broken-out/mm-mmap-remove-verify_mm_writelocked.patch

Vlastimil

[PATCH] mm: Replace verify_mm_writelocked() by lockdep_assert_held_exclusive()

[Waiman Long]

Using down_read_trylock() to check if a task holds a write lock on
a rwsem is not reliable. A task can hold a read lock on a rwsem and
down_read_trylock() can fail if a writer is waiting in the wait queue.
So use lockdep_assert_held_exclusive() instead which can do the right
check when CONFIG_LOCKDEP is on.

Signed-off-by: Waiman Long 
---
 mm/mmap.c | 12 +---
 1 file changed, 1 insertion(+), 11 deletions(-)

diff --git a/mm/mmap.c b/mm/mmap.c
index 6c04292..62a5593 100644
--- a/mm/mmap.c
+++ b/mm/mmap.c
@@ -2962,16 +2962,6 @@ int vm_munmap(unsigned long start, size_t len)
return ret;
 }
 
-static inline void verify_mm_writelocked(struct mm_struct *mm)
-{
-#ifdef CONFIG_DEBUG_VM
-   if (unlikely(down_read_trylock(>mmap_sem))) {
-   WARN_ON(1);
-   up_read(>mmap_sem);
-   }
-#endif
-}
-
 /*
  *  this is really a simplified "do_mmap".  it only handles
  *  anonymous maps.  eventually we may be able to do some
@@ -3002,7 +2992,7 @@ static int do_brk_flags(unsigned long addr, unsigned long 
len, unsigned long fla
 * mm->mmap_sem is required to protect against another thread
 * changing the mappings in case we sleep.
 */
-   verify_mm_writelocked(mm);
+   lockdep_assert_held_exclusive(>mmap_sem);
 
/*
 * Clear old maps.  this also does some error checking for us
-- 
1.8.3.1