[PATCH] io-wq: ensure work->task_pid is cleared on init

[PATCH] io-wq: ensure work->task_pid is cleared on init

[Jens Axboe]

We use ->task_pid for exit cancellation, but we need to ensure it's
cleared to zero for io_req_work_grab_env() to do the right thing.

Fixes: 36282881a795 ("io-wq: add io_wq_cancel_pid() to cancel based on a specific pid")
Signed-off-by: Jens Axboe <axboe@kernel.dk>

---

diff --git a/fs/io-wq.h b/fs/io-wq.h
index ccc7d84af57d..9e9419c08bc1 100644
--- a/fs/io-wq.h
+++ b/fs/io-wq.h
@@ -88,6 +88,7 @@ struct io_wq_work {
 		(work)->creds = NULL;			\
 		(work)->fs = NULL;			\
 		(work)->flags = 0;			\
+		(work)->task_pid = 0;			\
 	} while (0)					\
 
 typedef void (get_work_fn)(struct io_wq_work *);

-- 
Jens Axboe

Re: [PATCH] io-wq: ensure work->task_pid is cleared on init

[Bart Van Assche]

On 2/25/20 10:55 AM, Jens Axboe wrote:
> We use ->task_pid for exit cancellation, but we need to ensure it's
> cleared to zero for io_req_work_grab_env() to do the right thing.

How about initializing .task_pid with this (totally untested) patch?

diff --git a/fs/io-wq.h b/fs/io-wq.h
index ccc7d84af57d..b8d3287cec57 100644
--- a/fs/io-wq.h
+++ b/fs/io-wq.h
@@ -80,15 +80,7 @@ struct io_wq_work {
  };

  #define INIT_IO_WORK(work, _func)			\
-	do {						\
-		(work)->list.next = NULL;		\
-		(work)->func = _func;			\
-		(work)->files = NULL;			\
-		(work)->mm = NULL;			\
-		(work)->creds = NULL;			\
-		(work)->fs = NULL;			\
-		(work)->flags = 0;			\
-	} while (0)					\
+	do { *(work) = (struct io_wq_work){ .func = _func }; } while (0)

  typedef void (get_work_fn)(struct io_wq_work *);
  typedef void (put_work_fn)(struct io_wq_work *);

Thanks,

Bart.

Re: [PATCH] io-wq: ensure work->task_pid is cleared on init

[Jens Axboe]

On 2/25/20 12:38 PM, Bart Van Assche wrote:
> On 2/25/20 10:55 AM, Jens Axboe wrote:
>> We use ->task_pid for exit cancellation, but we need to ensure it's
>> cleared to zero for io_req_work_grab_env() to do the right thing.
> 
> How about initializing .task_pid with this (totally untested) patch?
> 
> diff --git a/fs/io-wq.h b/fs/io-wq.h
> index ccc7d84af57d..b8d3287cec57 100644
> --- a/fs/io-wq.h
> +++ b/fs/io-wq.h
> @@ -80,15 +80,7 @@ struct io_wq_work {
>   };
> 
>   #define INIT_IO_WORK(work, _func)			\
> -	do {						\
> -		(work)->list.next = NULL;		\
> -		(work)->func = _func;			\
> -		(work)->files = NULL;			\
> -		(work)->mm = NULL;			\
> -		(work)->creds = NULL;			\
> -		(work)->fs = NULL;			\
> -		(work)->flags = 0;			\
> -	} while (0)					\
> +	do { *(work) = (struct io_wq_work){ .func = _func }; } while (0)
> 
>   typedef void (get_work_fn)(struct io_wq_work *);
>   typedef void (put_work_fn)(struct io_wq_work *);

That's not a bad idea and would be more future proof, in case more fields
are added at a later date.

-- 
Jens Axboe