Re: [PATCH v3 1/3]nbd: fix might_sleep warning on socket shutdown

From: Markus Pargmann
Date: Wed Jun 29 2016 - 03:18:33 EST


On Friday 24 June 2016 13:09:34 Pranay Kr. Srivastava wrote:
> spinlocked ranges should be small and not contain calls into huge
> subfunctions. Fix my mistake and just get the pointer to the socket
> instead of doing everything with spinlock held.
>
> Reported-by: Mikulas Patocka <mikulas@xxxxxxxxxxxxx>
> Signed-off-by: Markus Pargmann <mpa@xxxxxxxxxxxxxx>
>
> Changelog:
> Pranay Kr. Srivastava<pranjas@xxxxxxxxx>:
>
> 1) Use spin_lock instead of irq version for sock_shutdown.
>
> 2) Use system work queue to actually trigger the shutdown of
> socket. This solves the issue when kernel_sendmsg is currently
> blocked while a timeout occurs.
>
> Signed-off-by: Pranay Kr. Srivastava <pranjas@xxxxxxxxx>
> ---
> drivers/block/nbd.c | 69 ++++++++++++++++++++++++++++++++++-------------------
> 1 file changed, 44 insertions(+), 25 deletions(-)
>
> diff --git a/drivers/block/nbd.c b/drivers/block/nbd.c
> index 56f7f5d..586d946 100644
> --- a/drivers/block/nbd.c
> +++ b/drivers/block/nbd.c
> @@ -39,6 +39,7 @@
> #include <asm/types.h>
>
> #include <linux/nbd.h>
> +#include <linux/workqueue.h>
>
> struct nbd_device {
> u32 flags;
> @@ -69,6 +70,10 @@ struct nbd_device {
> #if IS_ENABLED(CONFIG_DEBUG_FS)
> struct dentry *dbg_dir;
> #endif
> + /*
> + *This is specifically for calling sock_shutdown, for now.
> + */

Please fix the indentation of this comment.

> + struct work_struct ws_shutdown;
> };
>
> #if IS_ENABLED(CONFIG_DEBUG_FS)
> @@ -95,6 +100,11 @@ static int max_part;
> */
> static DEFINE_SPINLOCK(nbd_lock);
>
> +/*
> + * Shutdown function for nbd_dev work struct.
> + */
> +static void nbd_ws_func_shutdown(struct work_struct *);

You could as well put the function implementation here. No need for a
function signature.

> +
> static inline struct device *nbd_to_dev(struct nbd_device *nbd)
> {
> return disk_to_dev(nbd->disk);
> @@ -172,39 +182,35 @@ static void nbd_end_request(struct nbd_device *nbd, struct request *req)
> */
> static void sock_shutdown(struct nbd_device *nbd)
> {
> - spin_lock_irq(&nbd->sock_lock);
> -
> - if (!nbd->sock) {
> - spin_unlock_irq(&nbd->sock_lock);
> - return;
> - }
> + struct socket *sock;
>
> - dev_warn(disk_to_dev(nbd->disk), "shutting down socket\n");
> - kernel_sock_shutdown(nbd->sock, SHUT_RDWR);
> - sockfd_put(nbd->sock);
> + spin_lock(&nbd->sock_lock);
> + sock = nbd->sock;
> nbd->sock = NULL;
> - spin_unlock_irq(&nbd->sock_lock);
> + spin_unlock(&nbd->sock_lock);
> +
> + if (!sock)
> + return;
>
> del_timer(&nbd->timeout_timer);
> + dev_warn(disk_to_dev(nbd->disk), "shutting down socket\n");
> + kernel_sock_shutdown(sock, SHUT_RDWR);
> + sockfd_put(sock);
> }
>
> static void nbd_xmit_timeout(unsigned long arg)
> {
> struct nbd_device *nbd = (struct nbd_device *)arg;
> - unsigned long flags;
>
> if (list_empty(&nbd->queue_head))
> return;
> -
> - spin_lock_irqsave(&nbd->sock_lock, flags);
> -
> nbd->timedout = true;
> -
> - if (nbd->sock)
> - kernel_sock_shutdown(nbd->sock, SHUT_RDWR);
> -
> - spin_unlock_irqrestore(&nbd->sock_lock, flags);
> -
> + schedule_work(&nbd->ws_shutdown);
> + /*
> + * Make sure sender thread sees nbd->timedout.
> + */
> + smp_wmb();
> + wake_up(&nbd->waiting_wq);
> dev_err(nbd_to_dev(nbd), "Connection timed out, shutting down connection\n");
> }
>
> @@ -574,8 +580,8 @@ static int nbd_thread_send(void *data)
> while (!kthread_should_stop() || !list_empty(&nbd->waiting_queue)) {
> /* wait for something to do */
> wait_event_interruptible(nbd->waiting_wq,
> - kthread_should_stop() ||
> - !list_empty(&nbd->waiting_queue));
> + kthread_should_stop() ||
> + !list_empty(&nbd->waiting_queue));

This is unrelated, please remove.

>
> /* extract request */
> if (list_empty(&nbd->waiting_queue))
> @@ -583,12 +589,16 @@ static int nbd_thread_send(void *data)
>
> spin_lock_irq(&nbd->queue_lock);
> req = list_entry(nbd->waiting_queue.next, struct request,
> - queuelist);
> + queuelist);

Unrelated as well.

> list_del_init(&req->queuelist);
> spin_unlock_irq(&nbd->queue_lock);
>
> - /* handle request */

Unrelated.

> nbd_handle_req(nbd, req);
> + if (nbd->timedout) {
> + req->errors++;
> + nbd_end_request(nbd, req);
> + } else
> + nbd_handle_req(nbd, req);

This does not change anything to avoid the spinlock sock_shutdown issue.
Please split in a separate patch.

Also this is already handled in nbd_handle_req(). For !nbd->sock the
same code is executed.


Otherwise the patch looks good.

Best Regards,

Markus

> }
>
> nbd->task_send = NULL;
> @@ -668,6 +678,7 @@ static void nbd_reset(struct nbd_device *nbd)
> set_capacity(nbd->disk, 0);
> nbd->flags = 0;
> nbd->xmit_timeout = 0;
> + INIT_WORK(&nbd->ws_shutdown, nbd_ws_func_shutdown);
> queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD, nbd->disk->queue);
> del_timer_sync(&nbd->timeout_timer);
> }
> @@ -802,11 +813,11 @@ static int __nbd_ioctl(struct block_device *bdev, struct nbd_device *nbd,
> error = nbd_thread_recv(nbd, bdev);
> nbd_dev_dbg_close(nbd);
> kthread_stop(thread);
> + sock_shutdown(nbd);
>
> mutex_lock(&nbd->tx_lock);
> nbd->task_recv = NULL;
>
> - sock_shutdown(nbd);
> nbd_clear_que(nbd);
> kill_bdev(bdev);
> nbd_bdev_reset(bdev);
> @@ -862,6 +873,14 @@ static const struct block_device_operations nbd_fops = {
> .compat_ioctl = nbd_ioctl,
> };
>
> +static void nbd_ws_func_shutdown(struct work_struct *ws_nbd)
> +{
> + struct nbd_device *nbd_dev = container_of(ws_nbd, struct nbd_device,
> + ws_shutdown);
> +
> + sock_shutdown(nbd_dev);
> +}
> +
> #if IS_ENABLED(CONFIG_DEBUG_FS)
>
> static int nbd_dbg_tasks_show(struct seq_file *s, void *unused)
>

--
Pengutronix e.K. | |
Industrial Linux Solutions | http://www.pengutronix.de/ |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |

Attachment: signature.asc
Description: This is a digitally signed message part.