On Tue, Jul 29, 2014 at 03:36:57PM -0700, David Rientjes wrote:

On Tue, 29 Jul 2014, Kirill A. Shutemov wrote:

quoted

Things can go wrong if fault_around_bytes will be changed under
do_fault_around(): between fault_around_mask() and fault_around_pages().

Let's read fault_around_bytes only once during do_fault_around() and
calculate mask based on the reading.

Note: fault_around_bytes can only be updated via debug interface. Also
I've tried but was not able to trigger a bad behaviour without the
patch. So I would not consider this patch as urgent.

Signed-off-by: Kirill A. Shutemov <redacted>
---
 mm/memory.c | 17 +++++++++++------
 1 file changed, 11 insertions(+), 6 deletions(-)

diff --git a/mm/memory.c b/mm/memory.c
index 9d66bc66f338..7f4f0c41c9e9 100644
--- a/mm/memory.c
+++ b/mm/memory.c

@@ -2772,12 +2772,12 @@ static unsigned long fault_around_bytes = rounddown_pow_of_two(65536);
 
 static inline unsigned long fault_around_pages(void)
 {
-	return fault_around_bytes >> PAGE_SHIFT;
+	return ACCESS_ONCE(fault_around_bytes) >> PAGE_SHIFT;

Not sure why this is being added here, ACCESS_ONCE() would be needed 
depending on the context in which the return value is used, 
do_read_fault() won't need it.

Fair enough. I'll move it.

quoted

 }
 
-static inline unsigned long fault_around_mask(void)
+static inline unsigned long fault_around_mask(unsigned long nr_pages)
 {
-	return ~(fault_around_bytes - 1) & PAGE_MASK;
+	return ~(nr_pages * PAGE_SIZE - 1) & PAGE_MASK;
 }
 
 

This patch is corrupted because of the newline here that doesn't exist in 
linux-next.

I'll recheck.

quoted

@@ -2844,12 +2844,17 @@ late_initcall(fault_around_debugfs);
 static void do_fault_around(struct vm_area_struct *vma, unsigned long address,
 		pte_t *pte, pgoff_t pgoff, unsigned int flags)
 {
-	unsigned long start_addr;
+	unsigned long start_addr, nr_pages;
 	pgoff_t max_pgoff;
 	struct vm_fault vmf;
 	int off;
 
-	start_addr = max(address & fault_around_mask(), vma->vm_start);
+	nr_pages = fault_around_pages();
+	/* race with fault_around_bytes_set() */
+	if (unlikely(nr_pages <= 1))
+		return;

Why exactly is this unlikely if fault_around_bytes is tunable via debugfs 
to equal PAGE_SIZE?  I assume we're expecting nobody is going to be doing 
that, otherwise we'll hit the unlikely() branch here every time.

No. We hit do_fault_around() only after fault_around_pages() check in
do_read_fault(): so only in race case.

So either the unlikely or the tunable should be removed.

The problem is that fault_around_bytes isn't documented so we don't even 
know the min value without looking at the source code.

I would prefer to drop tunable, it will make code a bit simplier.
Andrew, iirc you've asked for it. Do you still think we need the handle?

I also don't see how nr_pages can be < 1.

As Andrey has pointed, the 'if' is not needed.

-- 
 Kirill A. Shutemov

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>