lockdep, mm: fix might_fault() annotation

Some code (nfs/sunrpc) uses socket ops on kernel memory while holding
the mmap_sem, this is safe because kernel memory doesn't get paged out,
therefore we'll never actually fault, and the might_fault() annotations
will generate false positives.

Reported-by: "J. Bruce Fields" <[email protected]>
Signed-off-by: Peter Zijlstra <[email protected]>
Signed-off-by: Ingo Molnar <[email protected]>
This commit is contained in:
Peter Zijlstra 2009-01-12 13:02:11 +01:00 • committed by Ingo Molnar
commit 95156f0051

View file

@ -3165,6 +3165,15 @@ void print_vma_addr(char *prefix, unsigned long ip)
#ifdef CONFIG_PROVE_LOCKING
void might_fault(void)
{
/*
* Some code (nfs/sunrpc) uses socket ops on kernel memory while
* holding the mmap_sem, this is safe because kernel memory doesn't
* get paged out, therefore we'll never actually fault, and the
* below annotations will generate false positives.
*/
if (segment_eq(get_fs(), KERNEL_DS))
return;
might_sleep();
/*
* it would be nicer only to annotate paths which are not under