[PATCH] fanotify: allow freeze when waiting response for permission events

[PATCH] fanotify: allow freeze when waiting response for permission events

[Winston Wen]

This is a long-standing issue that uninterruptible sleep in fanotify
could make system hibernation fail if the usperspace server gets frozen
before the process waiting for the response (as reported e.g. [1][2]).

A few years ago, there was an attempt to switch to interruptible sleep
while waiting [3], but that would lead to EINTR returns from open(2)
and break userspace [4], so it's been changed to only killable [5].

And the core freezer logic had been rewritten [6][7] in v6.1, allowing
freezing in uninterrupted sleep, so we can solve this problem now.

[1] https://lore.kernel.org/lkml/1518774280-38090-1-git-send-email-t.vivek@samsung.com/
[2] https://lore.kernel.org/lkml/c1bb16b7-9eee-9cea-2c96-a512d8b3b9c7@nwra.com/
[3] https://lore.kernel.org/linux-fsdevel/20190213145443.26836-1-jack@suse.cz/
[4] https://lore.kernel.org/linux-fsdevel/d0031e3a-f050-0832-fa59-928a80ffd44b@nwra.com/
[5] https://lore.kernel.org/linux-fsdevel/20190221105558.GA20921@quack2.suse.cz/
[6] https://lore.kernel.org/lkml/20220822114649.055452969@infradead.org/
[7] https://lore.kernel.org/lkml/20230908-avoid-spurious-freezer-wakeups-v4-0-6155aa3dafae@quicinc.com/

Signed-off-by: Winston Wen <wentao@uniontech.com>
---
 fs/notify/fanotify/fanotify.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/fs/notify/fanotify/fanotify.c b/fs/notify/fanotify/fanotify.c
index 1e4def21811e..285beaf5bc09 100644
--- a/fs/notify/fanotify/fanotify.c
+++ b/fs/notify/fanotify/fanotify.c
@@ -228,8 +228,10 @@ static int fanotify_get_response(struct fsnotify_group *group,
 
 	pr_debug("%s: group=%p event=%p\n", __func__, group, event);
 
-	ret = wait_event_killable(group->fanotify_data.access_waitq,
-				  event->state == FAN_EVENT_ANSWERED);
+	ret = wait_event_state(group->fanotify_data.access_waitq,
+				  event->state == FAN_EVENT_ANSWERED,
+				  TASK_KILLABLE|TASK_FREEZABLE);
+
 	/* Signal pending? */
 	if (ret < 0) {
 		spin_lock(&group->notification_lock);
-- 
2.43.0

Re: [PATCH] fanotify: allow freeze when waiting response for permission events

[Jan Kara]

On Tue 05-03-24 14:18:04, Winston Wen wrote:
> This is a long-standing issue that uninterruptible sleep in fanotify
> could make system hibernation fail if the usperspace server gets frozen
> before the process waiting for the response (as reported e.g. [1][2]).
> 
> A few years ago, there was an attempt to switch to interruptible sleep
> while waiting [3], but that would lead to EINTR returns from open(2)
> and break userspace [4], so it's been changed to only killable [5].
> 
> And the core freezer logic had been rewritten [6][7] in v6.1, allowing
> freezing in uninterrupted sleep, so we can solve this problem now.
> 
> [1] https://lore.kernel.org/lkml/1518774280-38090-1-git-send-email-t.vivek@samsung.com/
> [2] https://lore.kernel.org/lkml/c1bb16b7-9eee-9cea-2c96-a512d8b3b9c7@nwra.com/
> [3] https://lore.kernel.org/linux-fsdevel/20190213145443.26836-1-jack@suse.cz/
> [4] https://lore.kernel.org/linux-fsdevel/d0031e3a-f050-0832-fa59-928a80ffd44b@nwra.com/
> [5] https://lore.kernel.org/linux-fsdevel/20190221105558.GA20921@quack2.suse.cz/
> [6] https://lore.kernel.org/lkml/20220822114649.055452969@infradead.org/
> [7] https://lore.kernel.org/lkml/20230908-avoid-spurious-freezer-wakeups-v4-0-6155aa3dafae@quicinc.com/
> 
> Signed-off-by: Winston Wen <wentao@uniontech.com>

Thanks and I'm glad this has finally a workable solution (keeping fingers
crossed ;)). I've added the patch to my tree.

								Honza

> ---
>  fs/notify/fanotify/fanotify.c | 6 ++++--
>  1 file changed, 4 insertions(+), 2 deletions(-)
> 
> diff --git a/fs/notify/fanotify/fanotify.c b/fs/notify/fanotify/fanotify.c
> index 1e4def21811e..285beaf5bc09 100644
> --- a/fs/notify/fanotify/fanotify.c
> +++ b/fs/notify/fanotify/fanotify.c
> @@ -228,8 +228,10 @@ static int fanotify_get_response(struct fsnotify_group *group,
>  
>  	pr_debug("%s: group=%p event=%p\n", __func__, group, event);
>  
> -	ret = wait_event_killable(group->fanotify_data.access_waitq,
> -				  event->state == FAN_EVENT_ANSWERED);
> +	ret = wait_event_state(group->fanotify_data.access_waitq,
> +				  event->state == FAN_EVENT_ANSWERED,
> +				  TASK_KILLABLE|TASK_FREEZABLE);
> +
>  	/* Signal pending? */
>  	if (ret < 0) {
>  		spin_lock(&group->notification_lock);
> -- 
> 2.43.0
> 
-- 
Jan Kara <jack@suse.com>
SUSE Labs, CR