Re: [PATCH V6 3/6] mm: Introduce VM_LOCKONFAULT

On Thu, 06 Aug 2015, Vlastimil Babka wrote:

...

quoted

diff --git a/fs/proc/task_mmu.c b/fs/proc/task_mmu.c
index ca1e091..38d69fc 100644
--- a/fs/proc/task_mmu.c
+++ b/fs/proc/task_mmu.c

@@ -579,6 +579,7 @@ static void show_smap_vma_flags(struct seq_file *m, struct vm_area_struct *vma)

This function has the following comment:

Don't forget to update Documentation/ on changes.

[...]

quoted

--- a/mm/gup.c
+++ b/mm/gup.c

@@ -92,7 +92,7 @@ retry:
 		 */
 		mark_page_accessed(page);
 	}
-	if ((flags & FOLL_POPULATE) && (vma->vm_flags & VM_LOCKED)) {
+	if ((flags & FOLL_MLOCK) && (vma->vm_flags & VM_LOCKED)) {
 		/*
 		 * The preliminary mapping check is mainly to avoid the
 		 * pointless overhead of lock_page on the ZERO_PAGE

@@ -265,6 +265,9 @@ static int faultin_page(struct task_struct *tsk, struct vm_area_struct *vma,
 	unsigned int fault_flags = 0;
 	int ret;

+	/* mlock all present pages, but do not fault in new pages */
+	if ((*flags & (FOLL_POPULATE | FOLL_MLOCK)) == FOLL_MLOCK)
+		return -ENOENT;
 	/* For mm_populate(), just skip the stack guard page. */
 	if ((*flags & FOLL_POPULATE) &&
 			(stack_guard_page_start(vma, address) ||

@@ -850,7 +853,10 @@ long populate_vma_page_range(struct vm_area_struct *vma,
 	VM_BUG_ON_VMA(end   > vma->vm_end, vma);
 	VM_BUG_ON_MM(!rwsem_is_locked(&mm->mmap_sem), mm);

-	gup_flags = FOLL_TOUCH | FOLL_POPULATE;
+	gup_flags = FOLL_TOUCH | FOLL_MLOCK;
+	if ((vma->vm_flags & (VM_LOCKED | VM_LOCKONFAULT)) == VM_LOCKED)
+		gup_flags |= FOLL_POPULATE;
+
 	/*
 	 * We want to touch writable mappings with a write fault in order
 	 * to break COW, except for shared mappings because these don't COW

I think this might be breaking the populate part of
mmap(MAP_POPULATE & ~MAP_LOCKED) case, if I follow the execution
correctly (it's far from simple...)

SYSCALL_DEFINE6(mmap_pgoff... with MAP_POPULATE
  vm_mmap_pgoff(..., MAP_POPULATE...)
    do_mmap_pgoff(...MAP_POPULATE... &populate) -> populate == TRUE
    mm_populate()
      __mm_populate()
        populate_vma_page_range()

Previously, this path would have FOLL_POPULATE in gup_flags and
continue with __get_user_pages() and faultin_page() (actually
regardless of FOLL_POPULATE) which would fault in the pages.

After your patch, populate_vma_page_range() will set FOLL_MLOCK, but
since VM_LOCKED is not set, FOLL_POPULATE won't be set either.
Then faultin_page() will return on the new check:

	flags & (FOLL_POPULATE | FOLL_MLOCK)) == FOLL_MLOCK

I am on vacation atm but I will try and get to respin this series after
making sure there aren't any more FOLL flag issues.

Thanks for keeping with these :)

Eric