Re: [PATCH v4 02/16] freezer: add unsafe versions of freezable helpers for CIFS

Re: [PATCH v4 02/16] freezer: add unsafe versions of freezable helpers for CIFS

[Colin Cross]

CIFS calls wait_event_freezekillable_unsafe with a VFS lock held,
which is unsafe and will cause lockdep warnings when 6aa9707
"lockdep: check that no locks held at freeze time" is reapplied
(it was reverted in dbf520a).  CIFS shouldn't be doing this, but
it has long-running syscalls that must hold a lock but also
shouldn't block suspend.  Until CIFS freeze handling is rewritten
to use a signal to exit out of the critical section, add a new
wait_event_freezekillable_unsafe helper that will not run the
lockdep test when 6aa9707 is reapplied, and call it from CIFS.

In practice the likley result of holding the lock while freezing
is that a second task blocked on the lock will never freeze,
aborting suspend, but it is possible to manufacture a case using
the cgroup freezer, the lock, and the suspend freezer to create
a deadlock.  Silencing the lockdep warning here will allow
problems to be found in other drivers that may have a more
serious deadlock risk, and prevent new problems from being added.

Acked-by: Pavel Machek <pavel-+ZI9xUNit7I@public.gmane.org>
Signed-off-by: Colin Cross <ccross-z5hGa2qSFaRBDgjK7y7TUQ@public.gmane.org>
---
v4:
  Corrected to include CIFS wait_for_response hunk.
  The rest of this series is still at v3.

 fs/cifs/transport.c     |  2 +-
 include/linux/freezer.h | 13 +++++++++++++
 2 files changed, 14 insertions(+), 1 deletion(-)

diff --git a/fs/cifs/transport.c b/fs/cifs/transport.c
index 1a52868..e7f22f8 100644
--- a/fs/cifs/transport.c
+++ b/fs/cifs/transport.c
@@ -452,7 +452,7 @@ wait_for_response(struct TCP_Server_Info *server, struct mid_q_entry *midQ)
 {
 	int error;
 
-	error = wait_event_freezekillable(server->response_q,
+	error = wait_event_freezekillable_unsafe(server->response_q,
 				    midQ->mid_state != MID_REQUEST_SUBMITTED);
 	if (error < 0)
 		return -ERESTARTSYS;
diff --git a/include/linux/freezer.h b/include/linux/freezer.h
index 5b31e21c..d3c038e 100644
--- a/include/linux/freezer.h
+++ b/include/linux/freezer.h
@@ -212,6 +212,16 @@ static inline bool freezer_should_skip(struct task_struct *p)
 	__retval;							\
 })
 
+/* DO NOT ADD ANY NEW CALLERS OF THIS FUNCTION */
+#define wait_event_freezekillable_unsafe(wq, condition)			\
+({									\
+	int __retval;							\
+	freezer_do_not_count();						\
+	__retval = wait_event_killable(wq, (condition));		\
+	freezer_count_unsafe();						\
+	__retval;							\
+})
+
 #define wait_event_freezable(wq, condition)				\
 ({									\
 	int __retval;							\
@@ -277,6 +287,9 @@ static inline void set_freezable(void) {}
 #define wait_event_freezekillable(wq, condition)		\
 		wait_event_killable(wq, condition)
 
+#define wait_event_freezekillable_unsafe(wq, condition)			\
+		wait_event_killable(wq, condition)
+
 #endif /* !CONFIG_FREEZER */
 
 #endif	/* FREEZER_H_INCLUDED */
-- 
1.8.2.1

Re: [PATCH v4 02/16] freezer: add unsafe versions of freezable helpers for CIFS

[Jeff Layton]

On Tue,  7 May 2013 10:52:05 -0700
Colin Cross <ccross-z5hGa2qSFaRBDgjK7y7TUQ@public.gmane.org> wrote:

> CIFS calls wait_event_freezekillable_unsafe with a VFS lock held,
> which is unsafe and will cause lockdep warnings when 6aa9707
> "lockdep: check that no locks held at freeze time" is reapplied
> (it was reverted in dbf520a).  CIFS shouldn't be doing this, but
> it has long-running syscalls that must hold a lock but also
> shouldn't block suspend.  Until CIFS freeze handling is rewritten
> to use a signal to exit out of the critical section, add a new
> wait_event_freezekillable_unsafe helper that will not run the
> lockdep test when 6aa9707 is reapplied, and call it from CIFS.
> 
> In practice the likley result of holding the lock while freezing
> is that a second task blocked on the lock will never freeze,
> aborting suspend, but it is possible to manufacture a case using
> the cgroup freezer, the lock, and the suspend freezer to create
> a deadlock.  Silencing the lockdep warning here will allow
> problems to be found in other drivers that may have a more
> serious deadlock risk, and prevent new problems from being added.
> 
> Acked-by: Pavel Machek <pavel-+ZI9xUNit7I@public.gmane.org>
> Signed-off-by: Colin Cross <ccross-z5hGa2qSFaRBDgjK7y7TUQ@public.gmane.org>
> ---
> v4:
>   Corrected to include CIFS wait_for_response hunk.
>   The rest of this series is still at v3.
> 
>  fs/cifs/transport.c     |  2 +-
>  include/linux/freezer.h | 13 +++++++++++++
>  2 files changed, 14 insertions(+), 1 deletion(-)
> 
> diff --git a/fs/cifs/transport.c b/fs/cifs/transport.c
> index 1a52868..e7f22f8 100644
> --- a/fs/cifs/transport.c
> +++ b/fs/cifs/transport.c
> @@ -452,7 +452,7 @@ wait_for_response(struct TCP_Server_Info *server, struct mid_q_entry *midQ)
>  {
>  	int error;
>  
> -	error = wait_event_freezekillable(server->response_q,
> +	error = wait_event_freezekillable_unsafe(server->response_q,
>  				    midQ->mid_state != MID_REQUEST_SUBMITTED);
>  	if (error < 0)
>  		return -ERESTARTSYS;
> diff --git a/include/linux/freezer.h b/include/linux/freezer.h
> index 5b31e21c..d3c038e 100644
> --- a/include/linux/freezer.h
> +++ b/include/linux/freezer.h
> @@ -212,6 +212,16 @@ static inline bool freezer_should_skip(struct task_struct *p)
>  	__retval;							\
>  })
>  
> +/* DO NOT ADD ANY NEW CALLERS OF THIS FUNCTION */
> +#define wait_event_freezekillable_unsafe(wq, condition)			\
> +({									\
> +	int __retval;							\
> +	freezer_do_not_count();						\
> +	__retval = wait_event_killable(wq, (condition));		\
> +	freezer_count_unsafe();						\
> +	__retval;							\
> +})
> +
>  #define wait_event_freezable(wq, condition)				\
>  ({									\
>  	int __retval;							\
> @@ -277,6 +287,9 @@ static inline void set_freezable(void) {}
>  #define wait_event_freezekillable(wq, condition)		\
>  		wait_event_killable(wq, condition)
>  
> +#define wait_event_freezekillable_unsafe(wq, condition)			\
> +		wait_event_killable(wq, condition)
> +
>  #endif /* !CONFIG_FREEZER */
>  
>  #endif	/* FREEZER_H_INCLUDED */

Looks fine...

Reviewed-by: Jeff Layton <jlayton-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>