Re: [PATCH v4] pvcalls-front: Avoid get_free_pages(GFP_KERNEL) under spinlock

[Boris Ostrovsky]

On 12/1/18 1:33 AM, Wen Yang wrote:
> The problem is that we call this with a spin lock held.
> The call tree is:
> pvcalls_front_accept() holds bedata->socket_lock.
>     -> create_active()
>         -> __get_free_pages() uses GFP_KERNEL
>
> The create_active() function is only called from pvcalls_front_accept()
> with a spin_lock held, The allocation is not allowed to sleep and
> GFP_KERNEL is not sufficient.
>
> This issue was detected by using the Coccinelle software.
>
> v2: Add a function doing the allocations which is called
>     outside the lock and passing the allocated data to
>     create_active().
>
> v3: Use the matching deallocators i.e., free_page()
>     and free_pages(), respectively.
>
> v4: It would be better to pre-populate map (struct sock_mapping),
>     rather than introducing one more new struct.
>
> Suggested-by: Juergen Gross <jgross@xxxxxxxx>
> Suggested-by: Boris Ostrovsky <boris.ostrovsky@xxxxxxxxxx>
> Suggested-by: Stefano Stabellini <sstabellini@xxxxxxxxxx>
> Signed-off-by: Wen Yang <wen.yang99@xxxxxxxxxx>
> CC: Julia Lawall <julia.lawall@xxxxxxx>
> CC: Boris Ostrovsky <boris.ostrovsky@xxxxxxxxxx>
> CC: Juergen Gross <jgross@xxxxxxxx>
> CC: Stefano Stabellini <sstabellini@xxxxxxxxxx>
> CC: xen-devel@xxxxxxxxxxxxxxxxxxxx
> CC: linux-kernel@xxxxxxxxxxxxxxx
> ---
>  drivers/xen/pvcalls-front.c | 57 ++++++++++++++++++++++++++++++-------
>  1 file changed, 46 insertions(+), 11 deletions(-)
>
> diff --git a/drivers/xen/pvcalls-front.c b/drivers/xen/pvcalls-front.c
> index 77224d8f3e6f..555c9abdf58f 100644
> --- a/drivers/xen/pvcalls-front.c
> +++ b/drivers/xen/pvcalls-front.c
> @@ -335,13 +335,18 @@ int pvcalls_front_socket(struct socket *sock)
>  	return ret;
>  }
>  
> -static int create_active(struct sock_mapping *map, int *evtchn)
> +static void free_active_ring(struct sock_mapping *map)
>  {
> -	void *bytes;
> -	int ret = -ENOMEM, irq = -1, i;
> +	if (!map)
> +		return;
> +	free_pages((unsigned long)map->active.data.in,
> +			map->active.ring->ring_order);
> +	free_page((unsigned long)map->active.ring);
> +}
>  
> -	*evtchn = -1;
> -	init_waitqueue_head(&map->active.inflight_conn_req);
> +static int alloc_active_ring(struct sock_mapping *map)
> +{
> +	void *bytes;
>  
>  	map->active.ring = (struct pvcalls_data_intf *)
>  		__get_free_page(GFP_KERNEL | __GFP_ZERO);
> @@ -352,6 +357,26 @@ static int create_active(struct sock_mapping *map, int *evtchn)
>  					PVCALLS_RING_ORDER);
>  	if (bytes == NULL)
>  		goto out_error;
> +	map->active.data.in = bytes;
> +	map->active.data.out = bytes +
> +		XEN_FLEX_RING_SIZE(PVCALLS_RING_ORDER);
> +
> +	return 0;
> +
> +out_error:
> +	free_active_ring(map);
> +	return -ENOMEM;
> +}
> +
> +static int create_active(struct sock_mapping *map, int *evtchn)
> +{
> +	void *bytes;
> +	int ret = -ENOMEM, irq = -1, i;
> +
> +	*evtchn = -1;
> +	init_waitqueue_head(&map->active.inflight_conn_req);
> +
> +	bytes = map->active.data.in;

Why is this needed?

I may not be be reading the diff correctly, but your patch appears to be
whitespace-damaged and I can't apply it.

>  	for (i = 0; i < (1 << PVCALLS_RING_ORDER); i++)
>  		map->active.ring->ref[i] = gnttab_grant_foreign_access(
>  			pvcalls_front_dev->otherend_id,
> @@ -361,10 +386,6 @@ static int create_active(struct sock_mapping *map, int *evtchn)
>  		pvcalls_front_dev->otherend_id,
>  		pfn_to_gfn(virt_to_pfn((void *)map->active.ring)), 0);
>  
> -	map->active.data.in = bytes;
> -	map->active.data.out = bytes +
> -		XEN_FLEX_RING_SIZE(PVCALLS_RING_ORDER);
> -
>  	ret = xenbus_alloc_evtchn(pvcalls_front_dev, evtchn);
>  	if (ret)
>  		goto out_error;
> @@ -385,8 +406,7 @@ static int create_active(struct sock_mapping *map, int *evtchn)
>  out_error:
>  	if (*evtchn >= 0)
>  		xenbus_free_evtchn(pvcalls_front_dev, *evtchn);
> -	free_pages((unsigned long)map->active.data.in, PVCALLS_RING_ORDER);
> -	free_page((unsigned long)map->active.ring);
> +	free_active_ring(map);


I think that since you are allocating the data outside of this call it
should also be freed outside, when create_active() fails.

-boris


>  	return ret;
>  }
>  
> @@ -406,11 +426,17 @@ int pvcalls_front_connect(struct socket *sock, struct sockaddr *addr,
>  		return PTR_ERR(map);
>  
>  	bedata = dev_get_drvdata(&pvcalls_front_dev->dev);
> +	ret = alloc_active_ring(map);
> +	if (ret < 0) {
> +		pvcalls_exit_sock(sock);
> +		return ret;
> +	}
>  
>  	spin_lock(&bedata->socket_lock);
>  	ret = get_request(bedata, &req_id);
>  	if (ret < 0) {
>  		spin_unlock(&bedata->socket_lock);
> +		free_active_ring(map);
>  		pvcalls_exit_sock(sock);
>  		return ret;
>  	}
> @@ -780,12 +806,20 @@ int pvcalls_front_accept(struct socket *sock, struct socket *newsock, int flags)
>  		}
>  	}
>  
> +	ret = alloc_active_ring(map);
> +	if (ret < 0) {
> +		clear_bit(PVCALLS_FLAG_ACCEPT_INFLIGHT,
> +				(void *)&map->passive.flags);
> +		pvcalls_exit_sock(sock);
> +		return ret;
> +	}
>  	spin_lock(&bedata->socket_lock);
>  	ret = get_request(bedata, &req_id);
>  	if (ret < 0) {
>  		clear_bit(PVCALLS_FLAG_ACCEPT_INFLIGHT,
>  			  (void *)&map->passive.flags);
>  		spin_unlock(&bedata->socket_lock);
> +		free_active_ring(map);
>  		pvcalls_exit_sock(sock);
>  		return ret;
>  	}
> @@ -794,6 +828,7 @@ int pvcalls_front_accept(struct socket *sock, struct socket *newsock, int flags)
>  		clear_bit(PVCALLS_FLAG_ACCEPT_INFLIGHT,
>  			  (void *)&map->passive.flags);
>  		spin_unlock(&bedata->socket_lock);
> +		free_active_ring(map);
>  		pvcalls_exit_sock(sock);
>  		return -ENOMEM;
>  	}