Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / include / xen / interface / io / ring.h
diff --git a/kernel/include/xen/interface/io/ring.h b/kernel/include/xen/interface/io/ring.h
new file mode 100644 (file)
index 0000000..7d28aff
--- /dev/null
@@ -0,0 +1,272 @@
+/******************************************************************************
+ * ring.h
+ *
+ * Shared producer-consumer ring macros.
+ *
+ * Tim Deegan and Andrew Warfield November 2004.
+ */
+
+#ifndef __XEN_PUBLIC_IO_RING_H__
+#define __XEN_PUBLIC_IO_RING_H__
+
+typedef unsigned int RING_IDX;
+
+/* Round a 32-bit unsigned constant down to the nearest power of two. */
+#define __RD2(_x)  (((_x) & 0x00000002) ? 0x2                 : ((_x) & 0x1))
+#define __RD4(_x)  (((_x) & 0x0000000c) ? __RD2((_x)>>2)<<2    : __RD2(_x))
+#define __RD8(_x)  (((_x) & 0x000000f0) ? __RD4((_x)>>4)<<4    : __RD4(_x))
+#define __RD16(_x) (((_x) & 0x0000ff00) ? __RD8((_x)>>8)<<8    : __RD8(_x))
+#define __RD32(_x) (((_x) & 0xffff0000) ? __RD16((_x)>>16)<<16 : __RD16(_x))
+
+/*
+ * Calculate size of a shared ring, given the total available space for the
+ * ring and indexes (_sz), and the name tag of the request/response structure.
+ * A ring contains as many entries as will fit, rounded down to the nearest
+ * power of two (so we can mask with (size-1) to loop around).
+ */
+#define __CONST_RING_SIZE(_s, _sz)                             \
+       (__RD32(((_sz) - offsetof(struct _s##_sring, ring)) /   \
+               sizeof(((struct _s##_sring *)0)->ring[0])))
+
+/*
+ * The same for passing in an actual pointer instead of a name tag.
+ */
+#define __RING_SIZE(_s, _sz)                                           \
+       (__RD32(((_sz) - (long)&(_s)->ring + (long)(_s)) / sizeof((_s)->ring[0])))
+
+/*
+ * Macros to make the correct C datatypes for a new kind of ring.
+ *
+ * To make a new ring datatype, you need to have two message structures,
+ * let's say struct request, and struct response already defined.
+ *
+ * In a header where you want the ring datatype declared, you then do:
+ *
+ *     DEFINE_RING_TYPES(mytag, struct request, struct response);
+ *
+ * These expand out to give you a set of types, as you can see below.
+ * The most important of these are:
+ *
+ *     struct mytag_sring      - The shared ring.
+ *     struct mytag_front_ring - The 'front' half of the ring.
+ *     struct mytag_back_ring  - The 'back' half of the ring.
+ *
+ * To initialize a ring in your code you need to know the location and size
+ * of the shared memory area (PAGE_SIZE, for instance). To initialise
+ * the front half:
+ *
+ *     struct mytag_front_ring front_ring;
+ *     SHARED_RING_INIT((struct mytag_sring *)shared_page);
+ *     FRONT_RING_INIT(&front_ring, (struct mytag_sring *)shared_page,
+ *                    PAGE_SIZE);
+ *
+ * Initializing the back follows similarly (note that only the front
+ * initializes the shared ring):
+ *
+ *     struct mytag_back_ring back_ring;
+ *     BACK_RING_INIT(&back_ring, (struct mytag_sring *)shared_page,
+ *                   PAGE_SIZE);
+ */
+
+#define DEFINE_RING_TYPES(__name, __req_t, __rsp_t)                    \
+                                                                       \
+/* Shared ring entry */                                                        \
+union __name##_sring_entry {                                           \
+    __req_t req;                                                       \
+    __rsp_t rsp;                                                       \
+};                                                                     \
+                                                                       \
+/* Shared ring page */                                                 \
+struct __name##_sring {                                                        \
+    RING_IDX req_prod, req_event;                                      \
+    RING_IDX rsp_prod, rsp_event;                                      \
+    uint8_t  pad[48];                                                  \
+    union __name##_sring_entry ring[1]; /* variable-length */          \
+};                                                                     \
+                                                                       \
+/* "Front" end's private variables */                                  \
+struct __name##_front_ring {                                           \
+    RING_IDX req_prod_pvt;                                             \
+    RING_IDX rsp_cons;                                                 \
+    unsigned int nr_ents;                                              \
+    struct __name##_sring *sring;                                      \
+};                                                                     \
+                                                                       \
+/* "Back" end's private variables */                                   \
+struct __name##_back_ring {                                            \
+    RING_IDX rsp_prod_pvt;                                             \
+    RING_IDX req_cons;                                                 \
+    unsigned int nr_ents;                                              \
+    struct __name##_sring *sring;                                      \
+};
+
+/*
+ * Macros for manipulating rings.
+ *
+ * FRONT_RING_whatever works on the "front end" of a ring: here
+ * requests are pushed on to the ring and responses taken off it.
+ *
+ * BACK_RING_whatever works on the "back end" of a ring: here
+ * requests are taken off the ring and responses put on.
+ *
+ * N.B. these macros do NO INTERLOCKS OR FLOW CONTROL.
+ * This is OK in 1-for-1 request-response situations where the
+ * requestor (front end) never has more than RING_SIZE()-1
+ * outstanding requests.
+ */
+
+/* Initialising empty rings */
+#define SHARED_RING_INIT(_s) do {                                      \
+    (_s)->req_prod  = (_s)->rsp_prod  = 0;                             \
+    (_s)->req_event = (_s)->rsp_event = 1;                             \
+    memset((_s)->pad, 0, sizeof((_s)->pad));                           \
+} while(0)
+
+#define FRONT_RING_INIT(_r, _s, __size) do {                           \
+    (_r)->req_prod_pvt = 0;                                            \
+    (_r)->rsp_cons = 0;                                                        \
+    (_r)->nr_ents = __RING_SIZE(_s, __size);                           \
+    (_r)->sring = (_s);                                                        \
+} while (0)
+
+#define BACK_RING_INIT(_r, _s, __size) do {                            \
+    (_r)->rsp_prod_pvt = 0;                                            \
+    (_r)->req_cons = 0;                                                        \
+    (_r)->nr_ents = __RING_SIZE(_s, __size);                           \
+    (_r)->sring = (_s);                                                        \
+} while (0)
+
+/* Initialize to existing shared indexes -- for recovery */
+#define FRONT_RING_ATTACH(_r, _s, __size) do {                         \
+    (_r)->sring = (_s);                                                        \
+    (_r)->req_prod_pvt = (_s)->req_prod;                               \
+    (_r)->rsp_cons = (_s)->rsp_prod;                                   \
+    (_r)->nr_ents = __RING_SIZE(_s, __size);                           \
+} while (0)
+
+#define BACK_RING_ATTACH(_r, _s, __size) do {                          \
+    (_r)->sring = (_s);                                                        \
+    (_r)->rsp_prod_pvt = (_s)->rsp_prod;                               \
+    (_r)->req_cons = (_s)->req_prod;                                   \
+    (_r)->nr_ents = __RING_SIZE(_s, __size);                           \
+} while (0)
+
+/* How big is this ring? */
+#define RING_SIZE(_r)                                                  \
+    ((_r)->nr_ents)
+
+/* Number of free requests (for use on front side only). */
+#define RING_FREE_REQUESTS(_r)                                         \
+    (RING_SIZE(_r) - ((_r)->req_prod_pvt - (_r)->rsp_cons))
+
+/* Test if there is an empty slot available on the front ring.
+ * (This is only meaningful from the front. )
+ */
+#define RING_FULL(_r)                                                  \
+    (RING_FREE_REQUESTS(_r) == 0)
+
+/* Test if there are outstanding messages to be processed on a ring. */
+#define RING_HAS_UNCONSUMED_RESPONSES(_r)                              \
+    ((_r)->sring->rsp_prod - (_r)->rsp_cons)
+
+#define RING_HAS_UNCONSUMED_REQUESTS(_r)                               \
+    ({                                                                 \
+       unsigned int req = (_r)->sring->req_prod - (_r)->req_cons;      \
+       unsigned int rsp = RING_SIZE(_r) -                              \
+                          ((_r)->req_cons - (_r)->rsp_prod_pvt);       \
+       req < rsp ? req : rsp;                                          \
+    })
+
+/* Direct access to individual ring elements, by index. */
+#define RING_GET_REQUEST(_r, _idx)                                     \
+    (&((_r)->sring->ring[((_idx) & (RING_SIZE(_r) - 1))].req))
+
+#define RING_GET_RESPONSE(_r, _idx)                                    \
+    (&((_r)->sring->ring[((_idx) & (RING_SIZE(_r) - 1))].rsp))
+
+/* Loop termination condition: Would the specified index overflow the ring? */
+#define RING_REQUEST_CONS_OVERFLOW(_r, _cons)                          \
+    (((_cons) - (_r)->rsp_prod_pvt) >= RING_SIZE(_r))
+
+/* Ill-behaved frontend determination: Can there be this many requests? */
+#define RING_REQUEST_PROD_OVERFLOW(_r, _prod)               \
+    (((_prod) - (_r)->rsp_prod_pvt) > RING_SIZE(_r))
+
+
+#define RING_PUSH_REQUESTS(_r) do {                                    \
+    wmb(); /* back sees requests /before/ updated producer index */    \
+    (_r)->sring->req_prod = (_r)->req_prod_pvt;                                \
+} while (0)
+
+#define RING_PUSH_RESPONSES(_r) do {                                   \
+    wmb(); /* front sees responses /before/ updated producer index */  \
+    (_r)->sring->rsp_prod = (_r)->rsp_prod_pvt;                                \
+} while (0)
+
+/*
+ * Notification hold-off (req_event and rsp_event):
+ *
+ * When queueing requests or responses on a shared ring, it may not always be
+ * necessary to notify the remote end. For example, if requests are in flight
+ * in a backend, the front may be able to queue further requests without
+ * notifying the back (if the back checks for new requests when it queues
+ * responses).
+ *
+ * When enqueuing requests or responses:
+ *
+ *  Use RING_PUSH_{REQUESTS,RESPONSES}_AND_CHECK_NOTIFY(). The second argument
+ *  is a boolean return value. True indicates that the receiver requires an
+ *  asynchronous notification.
+ *
+ * After dequeuing requests or responses (before sleeping the connection):
+ *
+ *  Use RING_FINAL_CHECK_FOR_REQUESTS() or RING_FINAL_CHECK_FOR_RESPONSES().
+ *  The second argument is a boolean return value. True indicates that there
+ *  are pending messages on the ring (i.e., the connection should not be put
+ *  to sleep).
+ *
+ *  These macros will set the req_event/rsp_event field to trigger a
+ *  notification on the very next message that is enqueued. If you want to
+ *  create batches of work (i.e., only receive a notification after several
+ *  messages have been enqueued) then you will need to create a customised
+ *  version of the FINAL_CHECK macro in your own code, which sets the event
+ *  field appropriately.
+ */
+
+#define RING_PUSH_REQUESTS_AND_CHECK_NOTIFY(_r, _notify) do {          \
+    RING_IDX __old = (_r)->sring->req_prod;                            \
+    RING_IDX __new = (_r)->req_prod_pvt;                               \
+    wmb(); /* back sees requests /before/ updated producer index */    \
+    (_r)->sring->req_prod = __new;                                     \
+    mb(); /* back sees new requests /before/ we check req_event */     \
+    (_notify) = ((RING_IDX)(__new - (_r)->sring->req_event) <          \
+                (RING_IDX)(__new - __old));                            \
+} while (0)
+
+#define RING_PUSH_RESPONSES_AND_CHECK_NOTIFY(_r, _notify) do {         \
+    RING_IDX __old = (_r)->sring->rsp_prod;                            \
+    RING_IDX __new = (_r)->rsp_prod_pvt;                               \
+    wmb(); /* front sees responses /before/ updated producer index */  \
+    (_r)->sring->rsp_prod = __new;                                     \
+    mb(); /* front sees new responses /before/ we check rsp_event */   \
+    (_notify) = ((RING_IDX)(__new - (_r)->sring->rsp_event) <          \
+                (RING_IDX)(__new - __old));                            \
+} while (0)
+
+#define RING_FINAL_CHECK_FOR_REQUESTS(_r, _work_to_do) do {            \
+    (_work_to_do) = RING_HAS_UNCONSUMED_REQUESTS(_r);                  \
+    if (_work_to_do) break;                                            \
+    (_r)->sring->req_event = (_r)->req_cons + 1;                       \
+    mb();                                                              \
+    (_work_to_do) = RING_HAS_UNCONSUMED_REQUESTS(_r);                  \
+} while (0)
+
+#define RING_FINAL_CHECK_FOR_RESPONSES(_r, _work_to_do) do {           \
+    (_work_to_do) = RING_HAS_UNCONSUMED_RESPONSES(_r);                 \
+    if (_work_to_do) break;                                            \
+    (_r)->sring->rsp_event = (_r)->rsp_cons + 1;                       \
+    mb();                                                              \
+    (_work_to_do) = RING_HAS_UNCONSUMED_RESPONSES(_r);                 \
+} while (0)
+
+#endif /* __XEN_PUBLIC_IO_RING_H__ */