2022-12-27 19:31:25

by SeongJae Park

Subject: [PATCH] mm/page_reporting: replace rcu_access_pointer() with rcu_dereference_protected()

Page reporting fetches pr_dev_info using rcu_access_pointer(), which is
for safely fetching a pointer that will not be dereferenced but could
concurrently updated. The code indeed does not dereference pr_dev_info
after fetcing it using rcu_access_pointer(), but it fetches the pointer
while concurrent updtes to the pointer is avoided by holding the update
side lock, page_reporting_mutex.

In the case, rcu_dereference_protected() is recommended because it
provides better readability and performance on some cases, as
rcu_dereference_protected() avoids use of READ_ONCE(). Replace the
rcu_access_pointer() calls with rcu_dereference_protected().

Signed-off-by: SeongJae Park <[email protected]>
---
mm/page_reporting.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/mm/page_reporting.c b/mm/page_reporting.c
index 79a8554f024c..079e0ebe1fb7 100644
--- a/mm/page_reporting.c
+++ b/mm/page_reporting.c
@@ -356,7 +356,7 @@ int page_reporting_register(struct page_reporting_dev_info *prdev)
mutex_lock(&page_reporting_mutex);

/* nothing to do if already in use */
- if (rcu_access_pointer(pr_dev_info)) {
+ if (rcu_dereference_protected(pr_dev_info, true)) {
err = -EBUSY;
goto err_out;
}
@@ -401,7 +401,7 @@ void page_reporting_unregister(struct page_reporting_dev_info *prdev)
{
mutex_lock(&page_reporting_mutex);

- if (rcu_access_pointer(pr_dev_info) == prdev) {
+ if (rcu_dereference_protected(pr_dev_info, true) == prdev) {
/* Disable page reporting notification */
RCU_INIT_POINTER(pr_dev_info, NULL);
synchronize_rcu();
--
2.25.1

2022-12-27 20:06:49

by Matthew Wilcox

Subject: Re: [PATCH] mm/page_reporting: replace rcu_access_pointer() with rcu_dereference_protected()

On Tue, Dec 27, 2022 at 07:21:58PM +0000, SeongJae Park wrote:
> +++ b/mm/page_reporting.c
> @@ -356,7 +356,7 @@ int page_reporting_register(struct page_reporting_dev_info *prdev)
> mutex_lock(&page_reporting_mutex);
>
> /* nothing to do if already in use */
> - if (rcu_access_pointer(pr_dev_info)) {
> + if (rcu_dereference_protected(pr_dev_info, true)) {

Pretty sure that passing a bare 'true' is an antipattern.
Instead, document _what_ lock protects us, ie:

if (rcu_dereference_protected(pr_dev_info,
lockdep_is_held(&page_reporting_mutex))) {

Obviously, we took it just one line up, but if code moves around, it
may save us.

2022-12-27 20:23:57

by SeongJae Park

Subject: Re: [PATCH] mm/page_reporting: replace rcu_access_pointer() with rcu_dereference_protected()

On Tue, 27 Dec 2022 19:55:09 +0000 Matthew Wilcox <[email protected]> wrote:

> On Tue, Dec 27, 2022 at 07:21:58PM +0000, SeongJae Park wrote:
> > +++ b/mm/page_reporting.c
> > @@ -356,7 +356,7 @@ int page_reporting_register(struct page_reporting_dev_info *prdev)
> > mutex_lock(&page_reporting_mutex);
> >
> > /* nothing to do if already in use */
> > - if (rcu_access_pointer(pr_dev_info)) {
> > + if (rcu_dereference_protected(pr_dev_info, true)) {
>
> Pretty sure that passing a bare 'true' is an antipattern.
> Instead, document _what_ lock protects us, ie:
>
> if (rcu_dereference_protected(pr_dev_info,
> lockdep_is_held(&page_reporting_mutex))) {
>
> Obviously, we took it just one line up, but if code moves around, it
> may save us.

Good point, agreed. Will do so in the next version.


Thanks,
SJ