lkml.org 
[lkml]   [2017]   [Jul]   [3]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectRe: [PATCH v5 06/18] xen/pvcalls: handle commands from the frontend
From
Date
On 22/06/17 21:14, Stefano Stabellini wrote:
> When the other end notifies us that there are commands to be read
> (pvcalls_back_event), wake up the backend thread to parse the command.
>
> The command ring works like most other Xen rings, so use the usual
> ring macros to read and write to it. The functions implementing the
> commands are empty stubs for now.
>
> Signed-off-by: Stefano Stabellini <stefano@aporeto.com>
> CC: boris.ostrovsky@oracle.com
> CC: jgross@suse.com
> ---
> drivers/xen/pvcalls-back.c | 119 +++++++++++++++++++++++++++++++++++++++++++++
> 1 file changed, 119 insertions(+)
>
> diff --git a/drivers/xen/pvcalls-back.c b/drivers/xen/pvcalls-back.c
> index e4c2e46..437c2ad 100644
> --- a/drivers/xen/pvcalls-back.c
> +++ b/drivers/xen/pvcalls-back.c
> @@ -51,12 +51,131 @@ struct pvcalls_fedata {
> struct work_struct register_work;
> };
>
> +static int pvcalls_back_socket(struct xenbus_device *dev,
> + struct xen_pvcalls_request *req)
> +{
> + return 0;
> +}
> +
> +static int pvcalls_back_connect(struct xenbus_device *dev,
> + struct xen_pvcalls_request *req)
> +{
> + return 0;
> +}
> +
> +static int pvcalls_back_release(struct xenbus_device *dev,
> + struct xen_pvcalls_request *req)
> +{
> + return 0;
> +}
> +
> +static int pvcalls_back_bind(struct xenbus_device *dev,
> + struct xen_pvcalls_request *req)
> +{
> + return 0;
> +}
> +
> +static int pvcalls_back_listen(struct xenbus_device *dev,
> + struct xen_pvcalls_request *req)
> +{
> + return 0;
> +}
> +
> +static int pvcalls_back_accept(struct xenbus_device *dev,
> + struct xen_pvcalls_request *req)
> +{
> + return 0;
> +}
> +
> +static int pvcalls_back_poll(struct xenbus_device *dev,
> + struct xen_pvcalls_request *req)
> +{
> + return 0;
> +}
> +
> +static int pvcalls_back_handle_cmd(struct xenbus_device *dev,
> + struct xen_pvcalls_request *req)
> +{
> + int ret = 0;
> +
> + switch (req->cmd) {
> + case PVCALLS_SOCKET:
> + ret = pvcalls_back_socket(dev, req);
> + break;
> + case PVCALLS_CONNECT:
> + ret = pvcalls_back_connect(dev, req);
> + break;
> + case PVCALLS_RELEASE:
> + ret = pvcalls_back_release(dev, req);
> + break;
> + case PVCALLS_BIND:
> + ret = pvcalls_back_bind(dev, req);
> + break;
> + case PVCALLS_LISTEN:
> + ret = pvcalls_back_listen(dev, req);
> + break;
> + case PVCALLS_ACCEPT:
> + ret = pvcalls_back_accept(dev, req);
> + break;
> + case PVCALLS_POLL:
> + ret = pvcalls_back_poll(dev, req);
> + break;
> + default:
> + ret = -ENOTSUPP;
> + break;
> + }
> + return ret;
> +}
> +
> static void pvcalls_back_work(struct work_struct *work)
> {
> + struct pvcalls_fedata *fedata = container_of(work,
> + struct pvcalls_fedata, register_work);
> + int notify, notify_all = 0, more = 1;
> + struct xen_pvcalls_request req;
> + struct xenbus_device *dev = fedata->dev;
> +
> + while (more) {
> + while (RING_HAS_UNCONSUMED_REQUESTS(&fedata->ring)) {
> + RING_COPY_REQUEST(&fedata->ring,
> + fedata->ring.req_cons++,
> + &req);
> +
> + if (!pvcalls_back_handle_cmd(dev, &req)) {

Hmm, no response in case of not supported command?

> + RING_PUSH_RESPONSES_AND_CHECK_NOTIFY(
> + &fedata->ring, notify);
> + notify_all += notify;
> + }
> + }
> +
> + if (notify_all)
> + notify_remote_via_irq(fedata->irq);

Want to reset notify_all in above if?
Could have been an "accept" which didn't queues a response.

> +
> + RING_FINAL_CHECK_FOR_REQUESTS(&fedata->ring, more);
> + }
> }
>
> static irqreturn_t pvcalls_back_event(int irq, void *dev_id)
> {
> + struct xenbus_device *dev = dev_id;
> + struct pvcalls_fedata *fedata = NULL;
> +
> + if (dev == NULL)
> + return IRQ_HANDLED;
> +
> + fedata = dev_get_drvdata(&dev->dev);
> + if (fedata == NULL)
> + return IRQ_HANDLED;
> +
> + /*
> + * TODO: a small theoretical race exists if we try to queue work
> + * after pvcalls_back_work checked for final requests and before
> + * it returns. The queuing will fail, and pvcalls_back_work
> + * won't do the work because it is about to return. In that
> + * case, we lose the notification.
> + */
> + queue_work(fedata->wq, &fedata->register_work);

I know you like workqueues more than IRQ threads. But probably the above
TODO could be handled via an IRQ thread more easily?

I think you should either solve above race, or add a comment why it is
not problematic, or show us why an IRQ thread doesn't solve the problem.


Juergen

\
 
 \ /
  Last update: 2017-07-03 13:24    [W:0.070 / U:0.988 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site