Re: [PATCH v3 04/13] xen/pvcalls: implement socket command and handle events

[Boris Ostrovsky <boris.ostrovsky@oracle.com>]

On 07/31/2017 06:57 PM, Stefano Stabellini wrote:
> Send a PVCALLS_SOCKET command to the backend, use the masked
> req_prod_pvt as req_id. This way, req_id is guaranteed to be between 0
> and PVCALLS_NR_REQ_PER_RING. We already have a slot in the rsp array
> ready for the response, and there cannot be two outstanding responses
> with the same req_id.
> 
> Wait for the response by waiting on the inflight_req waitqueue and
> check for the req_id field in rsp[req_id]. Use atomic accesses and
> barriers to read the field. Note that the barriers are simple smp
> barriers (as opposed to virt barriers) because they are for internal
> frontend synchronization, not frontend<->backend communication.
> 
> Once a response is received, clear the corresponding rsp slot by setting
> req_id to PVCALLS_INVALID_ID. Note that PVCALLS_INVALID_ID is invalid
> only from the frontend point of view. It is not part of the PVCalls
> protocol.
> 
> pvcalls_front_event_handler is in charge of copying responses from the
> ring to the appropriate rsp slot. It is done by copying the body of the
> response first, then by copying req_id atomically. After the copies,
> wake up anybody waiting on waitqueue.
> 
> pvcallss_lock protects accesses to the ring.
> 
> Create a new struct sock_mapping and convert the pointer into an
> uint64_t and use it as id for the new socket to pass to the backend. The
> struct will be fully initialized later on connect or bind. In this patch
> the struct sock_mapping is empty, the fields will be added by the next
> patch.
> 
> sock->sk->sk_send_head is not used for ip sockets: reuse the field to
> store a pointer to the struct sock_mapping corresponding to the socket.
> This way, we can easily get the struct sock_mapping from the struct
> socket.
> 
> Signed-off-by: Stefano Stabellini <stefano@aporeto.com>
> CC: boris.ostrovsky@oracle.com
> CC: jgross@suse.com
> ---
>   drivers/xen/pvcalls-front.c | 119 ++++++++++++++++++++++++++++++++++++++++++++
>   drivers/xen/pvcalls-front.h |   8 +++
>   2 files changed, 127 insertions(+)
>   create mode 100644 drivers/xen/pvcalls-front.h
> 
> diff --git a/drivers/xen/pvcalls-front.c b/drivers/xen/pvcalls-front.c
> index 2afe36d..7c4a7cb 100644
> --- a/drivers/xen/pvcalls-front.c
> +++ b/drivers/xen/pvcalls-front.c
> @@ -20,6 +20,8 @@
>   #include <xen/xenbus.h>
>   #include <xen/interface/io/pvcalls.h>
>   
> +#include "pvcalls-front.h"
> +
>   #define PVCALLS_INVALID_ID UINT_MAX
>   #define PVCALLS_RING_ORDER XENBUS_MAX_RING_GRANT_ORDER
>   #define PVCALLS_NR_REQ_PER_RING __CONST_RING_SIZE(xen_pvcalls, XEN_PAGE_SIZE)
> @@ -38,11 +40,128 @@ struct pvcalls_bedata {
>   };
>   static struct xenbus_device *pvcalls_front_dev;
>   
> +struct sock_mapping {
> +	bool active_socket;
> +	struct list_head list;
> +	struct socket *sock;
> +};
> +
> +static inline int get_request(struct pvcalls_bedata *bedata, int *req_id)
> +{
> +	*req_id = bedata->ring.req_prod_pvt & (RING_SIZE(&bedata->ring) - 1);
> +	if (RING_FULL(&bedata->ring) ||
> +	    READ_ONCE(bedata->rsp[*req_id].req_id) != PVCALLS_INVALID_ID)
> +		return -EAGAIN;
> +	return 0;
> +}
> +
>   static irqreturn_t pvcalls_front_event_handler(int irq, void *dev_id)
>   {
> +	struct xenbus_device *dev = dev_id;
> +	struct pvcalls_bedata *bedata;
> +	struct xen_pvcalls_response *rsp;
> +	uint8_t *src, *dst;
> +	int req_id = 0, more = 0, done = 0;
> +
> +	if (dev == NULL)
> +		return IRQ_HANDLED;
> +
> +	bedata = dev_get_drvdata(&dev->dev);
> +	if (bedata == NULL)
> +		return IRQ_HANDLED;
> +
> +again:
> +	while (RING_HAS_UNCONSUMED_RESPONSES(&bedata->ring)) {
> +		rsp = RING_GET_RESPONSE(&bedata->ring, bedata->ring.rsp_cons);
> +
> +		req_id = rsp->req_id;
> +		dst = (uint8_t *)&bedata->rsp[req_id] + sizeof(rsp->req_id);
> +		src = (uint8_t *)rsp + sizeof(rsp->req_id);
> +		memcpy(dst, src, sizeof(*rsp) - sizeof(rsp->req_id));
> +		/*
> +		 * First copy the rest of the data, then req_id. It is
> +		 * paired with the barrier when accessing bedata->rsp.
> +		 */
> +		smp_wmb();
> +		WRITE_ONCE(bedata->rsp[req_id].req_id, rsp->req_id);
> +
> +		done = 1;
> +		bedata->ring.rsp_cons++;
> +	}
> +
> +	RING_FINAL_CHECK_FOR_RESPONSES(&bedata->ring, more);
> +	if (more)
> +		goto again;
> +	if (done)
> +		wake_up(&bedata->inflight_req);
>   	return IRQ_HANDLED;
>   }
>   
> +int pvcalls_front_socket(struct socket *sock)
> +{
> +	struct pvcalls_bedata *bedata;
> +	struct sock_mapping *map = NULL;
> +	struct xen_pvcalls_request *req;
> +	int notify, req_id, ret;
> +
> +	if (!pvcalls_front_dev)
> +		return -EACCES;
> +	/*
> +	 * PVCalls only supports domain AF_INET,
> +	 * type SOCK_STREAM and protocol 0 sockets for now.
> +	 *
> +	 * Check socket type here, AF_INET and protocol checks are done
> +	 * by the caller.
> +	 */
> +	if (sock->type != SOCK_STREAM)
> +	    return -ENOTSUPP;
> +
> +	bedata = dev_get_drvdata(&pvcalls_front_dev->dev);
> +
> +	map = kzalloc(sizeof(*map), GFP_KERNEL);
> +	if (map == NULL)
> +		return -ENOMEM;
> +	/*
> +	 * sock->sk->sk_send_head is not used for ip sockets: reuse the
> +	 * field to store a pointer to the struct sock_mapping
> +	 * corresponding to the socket. This way, we can easily get the
> +	 * struct sock_mapping from the struct socket.


Is this a safe assumption? Is it conceivable that at some point in the 
future sk_send_head might become used?


> +	 */
> +	WRITE_ONCE(sock->sk->sk_send_head, (void *)map);
> +
> +	spin_lock(&bedata->pvcallss_lock);
> +	list_add_tail(&map->list, &bedata->socket_mappings);
> +
> +	ret = get_request(bedata, &req_id);
> +	if (ret < 0) {
> +		spin_unlock(&bedata->pvcallss_lock);
> +		return ret;
What is happening to the mapping you've linked above? Should it be 
unlinked and freed? (or defer its creation until we've got a request slot?)

> +	}
> +	req = RING_GET_REQUEST(&bedata->ring, req_id);
> +	req->req_id = req_id;
> +	req->cmd = PVCALLS_SOCKET;
> +	req->u.socket.id = (uint64_t) map;
> +	req->u.socket.domain = AF_INET;
> +	req->u.socket.type = SOCK_STREAM;
> +	req->u.socket.protocol = 0;

IPPROTO_IP?


-boris


> +
> +	bedata->ring.req_prod_pvt++;
> +	RING_PUSH_REQUESTS_AND_CHECK_NOTIFY(&bedata->ring, notify);
> +	spin_unlock(&bedata->pvcallss_lock);
> +	if (notify)
> +		notify_remote_via_irq(bedata->irq);
> +
> +	wait_event(bedata->inflight_req,
> +		   READ_ONCE(bedata->rsp[req_id].req_id) == req_id);
> +
> +	ret = bedata->rsp[req_id].ret;
> +	/* read ret, then set this rsp slot to be reused */
> +	smp_mb();
> +	WRITE_ONCE(bedata->rsp[req_id].req_id, PVCALLS_INVALID_ID);
> +
> +	return ret;
> +}
> +
>   static const struct xenbus_device_id pvcalls_front_ids[] = {
>   	{ "pvcalls" },
>   	{ "" }
> diff --git a/drivers/xen/pvcalls-front.h b/drivers/xen/pvcalls-front.h
> new file mode 100644
> index 0000000..b7dabed
> --- /dev/null
> +++ b/drivers/xen/pvcalls-front.h
> @@ -0,0 +1,8 @@
> +#ifndef __PVCALLS_FRONT_H__
> +#define __PVCALLS_FRONT_H__
> +
> +#include <linux/net.h>
> +
> +int pvcalls_front_socket(struct socket *sock);
> +
> +#endif
>