qemu-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Qemu-devel] [PATCH v3 for-4.0 4/7] libvhost-user: Support recording inf


From: elohimes
Subject: [Qemu-devel] [PATCH v3 for-4.0 4/7] libvhost-user: Support recording inflight I/O in shared memory
Date: Thu, 3 Jan 2019 18:18:16 +0800

From: Xie Yongji <address@hidden>

This patch adds support for VHOST_USER_GET_SHM_SIZE and
VHOST_USER_SET_SHM_FD message to get shared memory from qemu.
Then we maintain a "bitmap" of all descriptors in
the shared memory for each queue to record inflight I/O.

Signed-off-by: Xie Yongji <address@hidden>
Signed-off-by: Zhang Yu <address@hidden>
---
 contrib/libvhost-user/libvhost-user.c | 221 +++++++++++++++++++++++++-
 contrib/libvhost-user/libvhost-user.h |  33 ++++
 2 files changed, 248 insertions(+), 6 deletions(-)

diff --git a/contrib/libvhost-user/libvhost-user.c 
b/contrib/libvhost-user/libvhost-user.c
index 23bd52264c..f18f5e6e62 100644
--- a/contrib/libvhost-user/libvhost-user.c
+++ b/contrib/libvhost-user/libvhost-user.c
@@ -53,6 +53,18 @@
             _min1 < _min2 ? _min1 : _min2; })
 #endif
 
+/* Round number down to multiple */
+#define ALIGN_DOWN(n, m) ((n) / (m) * (m))
+
+/* Round number up to multiple */
+#define ALIGN_UP(n, m) ALIGN_DOWN((n) + (m) - 1, (m))
+
+/* Align each region to cache line size in shared memory */
+#define SHM_ALIGNMENT 64
+
+/* The version of shared memory */
+#define SHM_VERSION 1
+
 #define VHOST_USER_HDR_SIZE offsetof(VhostUserMsg, payload.u64)
 
 /* The version of the protocol we support */
@@ -100,6 +112,8 @@ vu_request_to_string(unsigned int req)
         REQ(VHOST_USER_POSTCOPY_ADVISE),
         REQ(VHOST_USER_POSTCOPY_LISTEN),
         REQ(VHOST_USER_POSTCOPY_END),
+        REQ(VHOST_USER_GET_SHM_SIZE),
+        REQ(VHOST_USER_SET_SHM_FD),
         REQ(VHOST_USER_MAX),
     };
 #undef REQ
@@ -890,6 +904,41 @@ vu_check_queue_msg_file(VuDev *dev, VhostUserMsg *vmsg)
     return true;
 }
 
+static int
+vu_check_queue_inflights(VuDev *dev, VuVirtq *vq)
+{
+    int i = 0;
+
+    if ((dev->protocol_features &
+        VHOST_USER_PROTOCOL_F_SLAVE_SHMFD) == 0) {
+        return 0;
+    }
+
+    if (unlikely(!vq->shm)) {
+        return -1;
+    }
+
+    vq->used_idx = vq->vring.used->idx;
+    vq->inflight_num = 0;
+    for (i = 0; i < vq->vring.num; i++) {
+        if (vq->shm->inflight[i] == 0) {
+            continue;
+        }
+
+        vq->inflight_desc[vq->inflight_num++] = i;
+        vq->inuse++;
+    }
+    vq->shadow_avail_idx = vq->last_avail_idx = vq->inuse + vq->used_idx;
+
+    /* in case of I/O hang after reconnecting */
+    if (eventfd_write(vq->kick_fd, 1) ||
+        eventfd_write(vq->call_fd, 1)) {
+        return -1;
+    }
+
+    return 0;
+}
+
 static bool
 vu_set_vring_kick_exec(VuDev *dev, VhostUserMsg *vmsg)
 {
@@ -925,6 +974,10 @@ vu_set_vring_kick_exec(VuDev *dev, VhostUserMsg *vmsg)
                dev->vq[index].kick_fd, index);
     }
 
+    if (vu_check_queue_inflights(dev, &dev->vq[index])) {
+        vu_panic(dev, "Failed to check inflights for vq: %d\n", index);
+    }
+
     return false;
 }
 
@@ -1215,6 +1268,115 @@ vu_set_postcopy_end(VuDev *dev, VhostUserMsg *vmsg)
     return true;
 }
 
+static int
+vu_setup_shm(VuDev *dev)
+{
+    int i;
+    char *addr = (char *)dev->shm_info.addr;
+    uint64_t size = 0;
+    uint32_t vq_size = ALIGN_UP(dev->shm_info.vq_size, dev->shm_info.align);
+
+    if (dev->shm_info.version != SHM_VERSION) {
+        DPRINT("Invalid version for shm: %d", dev->shm_info.version);
+        return -1;
+    }
+
+    if (dev->shm_info.dev_size != 0) {
+        DPRINT("Invalid dev_size for shm: %d", dev->shm_info.dev_size);
+        return -1;
+    }
+
+    if (dev->shm_info.vq_size != sizeof(VuVirtqShm)) {
+        DPRINT("Invalid vq_size for shm: %d", dev->shm_info.vq_size);
+        return -1;
+    }
+
+    for (i = 0; i < VHOST_MAX_NR_VIRTQUEUE; i++) {
+        size += vq_size;
+        if (size > dev->shm_info.mmap_size) {
+            break;
+        }
+        dev->vq[i].shm = (VuVirtqShm *)addr;
+        addr += vq_size;
+    }
+
+    return 0;
+}
+
+static bool
+vu_get_shm_size(VuDev *dev, VhostUserMsg *vmsg)
+{
+    if (vmsg->size != sizeof(vmsg->payload.shm)) {
+        vu_panic(dev, "Invalid get_shm_size message:%d", vmsg->size);
+        vmsg->size = 0;
+        return true;
+    }
+
+    vmsg->payload.shm.dev_size = 0;
+    vmsg->payload.shm.vq_size = sizeof(VuVirtqShm);
+    vmsg->payload.shm.align = SHM_ALIGNMENT;
+    vmsg->payload.shm.version = SHM_VERSION;
+
+    DPRINT("send shm dev_size: %"PRId32"\n", vmsg->payload.shm.dev_size);
+    DPRINT("send shm vq_size: %"PRId32"\n", vmsg->payload.shm.vq_size);
+    DPRINT("send shm align: %"PRId32"\n", vmsg->payload.shm.align);
+    DPRINT("send shm version: %"PRId32"\n", vmsg->payload.shm.version);
+
+    return true;
+}
+
+static bool
+vu_set_shm_fd(VuDev *dev, VhostUserMsg *vmsg)
+{
+    int fd;
+    uint64_t mmap_size, mmap_offset;
+    void *rc;
+
+    if (vmsg->fd_num != 1 ||
+        vmsg->size != sizeof(vmsg->payload.shm)) {
+        vu_panic(dev, "Invalid set_shm_fd message size:%d fds:%d",
+                 vmsg->size, vmsg->fd_num);
+        return false;
+    }
+
+    fd = vmsg->fds[0];
+    mmap_size = vmsg->payload.shm.mmap_size;
+    mmap_offset = vmsg->payload.shm.mmap_offset;
+    DPRINT("set_shm_fd mmap_size: %"PRId64"\n", mmap_size);
+    DPRINT("set_shm_fd mmap_offset: %"PRId64"\n", mmap_offset);
+    DPRINT("set_shm_fd dev_size: %"PRId32"\n", vmsg->payload.shm.dev_size);
+    DPRINT("set_shm_fd vq_size: %"PRId32"\n", vmsg->payload.shm.vq_size);
+    DPRINT("set_shm_fd align: %"PRId32"\n", vmsg->payload.shm.align);
+    DPRINT("set_shm_fd version: %"PRId32"\n", vmsg->payload.shm.version);
+
+    rc = mmap(0, mmap_size, PROT_READ | PROT_WRITE, MAP_SHARED,
+              fd, mmap_offset);
+
+    close(fd);
+
+    if (rc == MAP_FAILED) {
+        vu_panic(dev, "set_shm_fd mmap error: %s", strerror(errno));
+        return false;
+    }
+
+    if (dev->shm_info.addr) {
+        munmap(dev->shm_info.addr, dev->shm_info.mmap_size);
+    }
+    dev->shm_info.addr = rc;
+    dev->shm_info.mmap_size = mmap_size;
+    dev->shm_info.dev_size = vmsg->payload.shm.dev_size;
+    dev->shm_info.vq_size = vmsg->payload.shm.vq_size;
+    dev->shm_info.align = vmsg->payload.shm.align;
+    dev->shm_info.version = vmsg->payload.shm.version;
+
+    if (vu_setup_shm(dev)) {
+        vu_panic(dev, "setup shm failed");
+        return false;
+    }
+
+    return false;
+}
+
 static bool
 vu_process_message(VuDev *dev, VhostUserMsg *vmsg)
 {
@@ -1292,6 +1454,10 @@ vu_process_message(VuDev *dev, VhostUserMsg *vmsg)
         return vu_set_postcopy_listen(dev, vmsg);
     case VHOST_USER_POSTCOPY_END:
         return vu_set_postcopy_end(dev, vmsg);
+    case VHOST_USER_GET_SHM_SIZE:
+        return vu_get_shm_size(dev, vmsg);
+    case VHOST_USER_SET_SHM_FD:
+        return vu_set_shm_fd(dev, vmsg);
     default:
         vmsg_close_fds(vmsg);
         vu_panic(dev, "Unhandled request: %d", vmsg->request);
@@ -1359,8 +1525,13 @@ vu_deinit(VuDev *dev)
             close(vq->err_fd);
             vq->err_fd = -1;
         }
+        vq->shm = NULL;
     }
 
+    if (dev->shm_info.addr) {
+        munmap(dev->shm_info.addr, dev->shm_info.mmap_size);
+        dev->shm_info.addr = NULL;
+    }
 
     vu_close_log(dev);
     if (dev->slave_fd != -1) {
@@ -1829,12 +2000,6 @@ virtqueue_map_desc(VuDev *dev,
     *p_num_sg = num_sg;
 }
 
-/* Round number down to multiple */
-#define ALIGN_DOWN(n, m) ((n) / (m) * (m))
-
-/* Round number up to multiple */
-#define ALIGN_UP(n, m) ALIGN_DOWN((n) + (m) - 1, (m))
-
 static void *
 virtqueue_alloc_element(size_t sz,
                                      unsigned out_num, unsigned in_num)
@@ -1935,9 +2100,44 @@ vu_queue_map_desc(VuDev *dev, VuVirtq *vq, unsigned int 
idx, size_t sz)
     return elem;
 }
 
+static int
+vu_queue_inflight_get(VuDev *dev, VuVirtq *vq, int desc_idx)
+{
+    if ((dev->protocol_features &
+        VHOST_USER_PROTOCOL_F_SLAVE_SHMFD) == 0) {
+        return 0;
+    }
+
+    if (unlikely(!vq->shm)) {
+        return -1;
+    }
+
+    vq->shm->inflight[desc_idx] = 1;
+
+    return 0;
+}
+
+static int
+vu_queue_inflight_put(VuDev *dev, VuVirtq *vq, int desc_idx)
+{
+    if ((dev->protocol_features &
+        VHOST_USER_PROTOCOL_F_SLAVE_SHMFD) == 0) {
+        return 0;
+    }
+
+    if (unlikely(!vq->shm)) {
+        return -1;
+    }
+
+    vq->shm->inflight[desc_idx] = 0;
+
+    return 0;
+}
+
 void *
 vu_queue_pop(VuDev *dev, VuVirtq *vq, size_t sz)
 {
+    int i;
     unsigned int head;
     VuVirtqElement *elem;
 
@@ -1946,6 +2146,12 @@ vu_queue_pop(VuDev *dev, VuVirtq *vq, size_t sz)
         return NULL;
     }
 
+    if (unlikely(vq->inflight_num > 0)) {
+        i = (--vq->inflight_num);
+        elem = vu_queue_map_desc(dev, vq, vq->inflight_desc[i], sz);
+        return elem;
+    }
+
     if (vu_queue_empty(dev, vq)) {
         return NULL;
     }
@@ -1976,6 +2182,8 @@ vu_queue_pop(VuDev *dev, VuVirtq *vq, size_t sz)
 
     vq->inuse++;
 
+    vu_queue_inflight_get(dev, vq, head);
+
     return elem;
 }
 
@@ -2121,4 +2329,5 @@ vu_queue_push(VuDev *dev, VuVirtq *vq,
 {
     vu_queue_fill(dev, vq, elem, len, 0);
     vu_queue_flush(dev, vq, 1);
+    vu_queue_inflight_put(dev, vq, elem->index);
 }
diff --git a/contrib/libvhost-user/libvhost-user.h 
b/contrib/libvhost-user/libvhost-user.h
index 4aa55b4d2d..fdfda688d2 100644
--- a/contrib/libvhost-user/libvhost-user.h
+++ b/contrib/libvhost-user/libvhost-user.h
@@ -53,6 +53,7 @@ enum VhostUserProtocolFeature {
     VHOST_USER_PROTOCOL_F_CONFIG = 9,
     VHOST_USER_PROTOCOL_F_SLAVE_SEND_FD = 10,
     VHOST_USER_PROTOCOL_F_HOST_NOTIFIER = 11,
+    VHOST_USER_PROTOCOL_F_SLAVE_SHMFD = 12,
 
     VHOST_USER_PROTOCOL_F_MAX
 };
@@ -91,6 +92,8 @@ typedef enum VhostUserRequest {
     VHOST_USER_POSTCOPY_ADVISE  = 28,
     VHOST_USER_POSTCOPY_LISTEN  = 29,
     VHOST_USER_POSTCOPY_END     = 30,
+    VHOST_USER_GET_SHM_SIZE = 31,
+    VHOST_USER_SET_SHM_FD = 32,
     VHOST_USER_MAX
 } VhostUserRequest;
 
@@ -138,6 +141,15 @@ typedef struct VhostUserVringArea {
     uint64_t offset;
 } VhostUserVringArea;
 
+typedef struct VhostUserShm {
+    uint64_t mmap_size;
+    uint64_t mmap_offset;
+    uint32_t dev_size;
+    uint32_t vq_size;
+    uint32_t align;
+    uint32_t version;
+} VhostUserShm;
+
 #if defined(_WIN32)
 # define VU_PACKED __attribute__((gcc_struct, packed))
 #else
@@ -163,6 +175,7 @@ typedef struct VhostUserMsg {
         VhostUserLog log;
         VhostUserConfig config;
         VhostUserVringArea area;
+        VhostUserShm shm;
     } payload;
 
     int fds[VHOST_MEMORY_MAX_NREGIONS];
@@ -234,9 +247,19 @@ typedef struct VuRing {
     uint32_t flags;
 } VuRing;
 
+typedef struct VuVirtqShm {
+    char inflight[VIRTQUEUE_MAX_SIZE];
+} VuVirtqShm;
+
 typedef struct VuVirtq {
     VuRing vring;
 
+    VuVirtqShm *shm;
+
+    uint16_t inflight_desc[VIRTQUEUE_MAX_SIZE];
+
+    uint16_t inflight_num;
+
     /* Next head to pop */
     uint16_t last_avail_idx;
 
@@ -279,11 +302,21 @@ typedef void (*vu_set_watch_cb) (VuDev *dev, int fd, int 
condition,
                                  vu_watch_cb cb, void *data);
 typedef void (*vu_remove_watch_cb) (VuDev *dev, int fd);
 
+typedef struct VuDevShmInfo {
+    void *addr;
+    uint64_t mmap_size;
+    uint32_t dev_size;
+    uint32_t vq_size;
+    uint32_t align;
+    uint32_t version;
+} VuDevShmInfo;
+
 struct VuDev {
     int sock;
     uint32_t nregions;
     VuDevRegion regions[VHOST_MEMORY_MAX_NREGIONS];
     VuVirtq vq[VHOST_MAX_NR_VIRTQUEUE];
+    VuDevShmInfo shm_info;
     int log_call_fd;
     int slave_fd;
     uint64_t log_size;
-- 
2.17.1




reply via email to

[Prev in Thread] Current Thread [Next in Thread]