[Qemu-devel] [PATCH PULL 00/15] RDMA queue

[Qemu-devel] [PATCH PULL 00/15] RDMA queue

[Marcel Apfelbaum]

The following changes since commit 0abaa41d936becd914a16ee1fe2a981d96d19428:

  Merge remote-tracking branch 'remotes/ehabkost/tags/x86-next-pull-request' into staging (2018-08-17 09:46:00 +0100)

are available in the Git repository at:

  https://github.com/marcel-apf/qemu tags/rdma-pull-request

for you to fetch changes up to 21ab34c9543fe1b6d31b3edbd01a397e7e090d00:

  config: split PVRDMA from RDMA (2018-08-18 18:01:34 +0300)

----------------------------------------------------------------
Various pvrdma fixes
Split pvrdma from rdma in config

----------------------------------------------------------------
Marcel Apfelbaum (2):
  hw/pvrdma: remove not needed include
  config: split PVRDMA from RDMA

Yuval Shaia (13):
  hw/rdma: Make distinction between device init and start modes
  hw/pvrdma: Bugfix - provide the correct attr_mask to query_qp
  hw/rdma: Modify debug macros
  hw/pvrdma: Clean CQE before use
  hw/pvrdma: Make default pkey 0xFFFF
  hw/rdma: Delete useless structure RdmaRmUserMR
  hw/rdma: Do not allocate memory for non-dma MR
  hw/rdma: Reorder resource cleanup
  hw/pvrdma: Cosmetic change - indent right
  hw/rdma: Cosmetic change - move to generic function
  hw/rdma: Print backend QP number in hex format
  hw/rdma: Bugfix - Support non-aligned buffers
  hw/rdma: Add reference to pci_dev in backend_dev

 configure                   |  55 ++++++++++++++-
 hw/rdma/Makefile.objs       |   2 +-
 hw/rdma/rdma_backend.c      | 105 +++++++++++++++++++++------
 hw/rdma/rdma_backend.h      |   4 +-
 hw/rdma/rdma_backend_defs.h |   3 +-
 hw/rdma/rdma_rm.c           |  63 ++++++++---------
 hw/rdma/rdma_rm_defs.h      |  10 +--
 hw/rdma/rdma_utils.c        |   4 ++
 hw/rdma/rdma_utils.h        |  16 +++--
 hw/rdma/vmw/pvrdma.h        |   3 +
 hw/rdma/vmw/pvrdma_cmd.c    |   9 +--
 hw/rdma/vmw/pvrdma_main.c   | 137 +++++++++++++++++++++---------------
 hw/rdma/vmw/pvrdma_qp_ops.c |   5 +-
 13 files changed, 280 insertions(+), 136 deletions(-)

-- 
2.17.1

[Qemu-devel] [PATCH PULL 01/15] hw/rdma: Make distinction between device init and start modes

[Marcel Apfelbaum]

From: Yuval Shaia <yuval.shaia@oracle.com>

There are certain operations that are well considered as part of device
configuration while others are needed only when "start" command is
triggered by the guest driver. An example of device initialization step
is msix_init and example of "device start" stage is the creation of a CQ
completion handler thread.

Driver expects such distinction - implement it.

Signed-off-by: Yuval Shaia <yuval.shaia@oracle.com>
Reviewed-by: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
Message-Id: <20180805153518.2983-2-yuval.shaia@oracle.com>
Signed-off-by: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
---
 hw/rdma/rdma_backend.c      |  96 +++++++++++++++++++++------
 hw/rdma/rdma_backend.h      |   2 +
 hw/rdma/rdma_backend_defs.h |   3 +-
 hw/rdma/vmw/pvrdma_main.c   | 129 +++++++++++++++++++++---------------
 4 files changed, 155 insertions(+), 75 deletions(-)

diff --git a/hw/rdma/rdma_backend.c b/hw/rdma/rdma_backend.c
index e9ced6f9ef..647e16399f 100644
--- a/hw/rdma/rdma_backend.c
+++ b/hw/rdma/rdma_backend.c
@@ -35,6 +35,7 @@
 #define VENDOR_ERR_MR_SMALL         0x208
 
 #define THR_NAME_LEN 16
+#define THR_POLL_TO  5000
 
 typedef struct BackendCtx {
     uint64_t req_id;
@@ -91,35 +92,82 @@ static void *comp_handler_thread(void *arg)
     int rc;
     struct ibv_cq *ev_cq;
     void *ev_ctx;
+    int flags;
+    GPollFD pfds[1];
+
+    /* Change to non-blocking mode */
+    flags = fcntl(backend_dev->channel->fd, F_GETFL);
+    rc = fcntl(backend_dev->channel->fd, F_SETFL, flags | O_NONBLOCK);
+    if (rc < 0) {
+        pr_dbg("Fail to change to non-blocking mode\n");
+        return NULL;
+    }
 
     pr_dbg("Starting\n");
 
+    pfds[0].fd = backend_dev->channel->fd;
+    pfds[0].events = G_IO_IN | G_IO_HUP | G_IO_ERR;
+
+    backend_dev->comp_thread.is_running = true;
+
     while (backend_dev->comp_thread.run) {
-        pr_dbg("Waiting for completion on channel %p\n", backend_dev->channel);
-        rc = ibv_get_cq_event(backend_dev->channel, &ev_cq, &ev_ctx);
-        pr_dbg("ibv_get_cq_event=%d\n", rc);
-        if (unlikely(rc)) {
-            pr_dbg("---> ibv_get_cq_event (%d)\n", rc);
-            continue;
-        }
+        do {
+            rc = qemu_poll_ns(pfds, 1, THR_POLL_TO * (int64_t)SCALE_MS);
+        } while (!rc && backend_dev->comp_thread.run);
 
-        rc = ibv_req_notify_cq(ev_cq, 0);
-        if (unlikely(rc)) {
-            pr_dbg("Error %d from ibv_req_notify_cq\n", rc);
-        }
+        if (backend_dev->comp_thread.run) {
+            pr_dbg("Waiting for completion on channel %p\n", backend_dev->channel);
+            rc = ibv_get_cq_event(backend_dev->channel, &ev_cq, &ev_ctx);
+            pr_dbg("ibv_get_cq_event=%d\n", rc);
+            if (unlikely(rc)) {
+                pr_dbg("---> ibv_get_cq_event (%d)\n", rc);
+                continue;
+            }
 
-        poll_cq(backend_dev->rdma_dev_res, ev_cq);
+            rc = ibv_req_notify_cq(ev_cq, 0);
+            if (unlikely(rc)) {
+                pr_dbg("Error %d from ibv_req_notify_cq\n", rc);
+            }
 
-        ibv_ack_cq_events(ev_cq, 1);
+            poll_cq(backend_dev->rdma_dev_res, ev_cq);
+
+            ibv_ack_cq_events(ev_cq, 1);
+        }
     }
 
     pr_dbg("Going down\n");
 
     /* TODO: Post cqe for all remaining buffs that were posted */
 
+    backend_dev->comp_thread.is_running = false;
+
+    qemu_thread_exit(0);
+
     return NULL;
 }
 
+static void stop_comp_thread(RdmaBackendDev *backend_dev)
+{
+    backend_dev->comp_thread.run = false;
+    while (backend_dev->comp_thread.is_running) {
+        pr_dbg("Waiting for thread to complete\n");
+        sleep(THR_POLL_TO / SCALE_US / 2);
+    }
+}
+
+static void start_comp_thread(RdmaBackendDev *backend_dev)
+{
+    char thread_name[THR_NAME_LEN] = {0};
+
+    stop_comp_thread(backend_dev);
+
+    snprintf(thread_name, sizeof(thread_name), "rdma_comp_%s",
+             ibv_get_device_name(backend_dev->ib_dev));
+    backend_dev->comp_thread.run = true;
+    qemu_thread_create(&backend_dev->comp_thread.thread, thread_name,
+                       comp_handler_thread, backend_dev, QEMU_THREAD_DETACHED);
+}
+
 void rdma_backend_register_comp_handler(void (*handler)(int status,
                                         unsigned int vendor_err, void *ctx))
 {
@@ -706,7 +754,6 @@ int rdma_backend_init(RdmaBackendDev *backend_dev,
     int i;
     int ret = 0;
     int num_ibv_devices;
-    char thread_name[THR_NAME_LEN] = {0};
     struct ibv_device **dev_list;
     struct ibv_port_attr port_attr;
 
@@ -800,11 +847,8 @@ int rdma_backend_init(RdmaBackendDev *backend_dev,
     pr_dbg("interface_id=0x%" PRIx64 "\n",
            be64_to_cpu(backend_dev->gid.global.interface_id));
 
-    snprintf(thread_name, sizeof(thread_name), "rdma_comp_%s",
-             ibv_get_device_name(backend_dev->ib_dev));
-    backend_dev->comp_thread.run = true;
-    qemu_thread_create(&backend_dev->comp_thread.thread, thread_name,
-                       comp_handler_thread, backend_dev, QEMU_THREAD_DETACHED);
+    backend_dev->comp_thread.run = false;
+    backend_dev->comp_thread.is_running = false;
 
     ah_cache_init();
 
@@ -823,8 +867,22 @@ out:
     return ret;
 }
 
+
+void rdma_backend_start(RdmaBackendDev *backend_dev)
+{
+    pr_dbg("Starting rdma_backend\n");
+    start_comp_thread(backend_dev);
+}
+
+void rdma_backend_stop(RdmaBackendDev *backend_dev)
+{
+    pr_dbg("Stopping rdma_backend\n");
+    stop_comp_thread(backend_dev);
+}
+
 void rdma_backend_fini(RdmaBackendDev *backend_dev)
 {
+    rdma_backend_stop(backend_dev);
     g_hash_table_destroy(ah_hash);
     ibv_destroy_comp_channel(backend_dev->channel);
     ibv_close_device(backend_dev->context);
diff --git a/hw/rdma/rdma_backend.h b/hw/rdma/rdma_backend.h
index 3cd636dd88..3049a73962 100644
--- a/hw/rdma/rdma_backend.h
+++ b/hw/rdma/rdma_backend.h
@@ -52,6 +52,8 @@ int rdma_backend_init(RdmaBackendDev *backend_dev,
                       uint8_t backend_gid_idx, struct ibv_device_attr *dev_attr,
                       Error **errp);
 void rdma_backend_fini(RdmaBackendDev *backend_dev);
+void rdma_backend_start(RdmaBackendDev *backend_dev);
+void rdma_backend_stop(RdmaBackendDev *backend_dev);
 void rdma_backend_register_comp_handler(void (*handler)(int status,
                                         unsigned int vendor_err, void *ctx));
 void rdma_backend_unregister_comp_handler(void);
diff --git a/hw/rdma/rdma_backend_defs.h b/hw/rdma/rdma_backend_defs.h
index ff5cfc26eb..7404f64002 100644
--- a/hw/rdma/rdma_backend_defs.h
+++ b/hw/rdma/rdma_backend_defs.h
@@ -24,7 +24,8 @@ typedef struct RdmaDeviceResources RdmaDeviceResources;
 typedef struct RdmaBackendThread {
     QemuThread thread;
     QemuMutex mutex;
-    bool run;
+    bool run; /* Set by thread manager to let thread know it should exit */
+    bool is_running; /* Set by the thread to report its status */
 } RdmaBackendThread;
 
 typedef struct RdmaBackendDev {
diff --git a/hw/rdma/vmw/pvrdma_main.c b/hw/rdma/vmw/pvrdma_main.c
index 3ed7409763..6a5073974d 100644
--- a/hw/rdma/vmw/pvrdma_main.c
+++ b/hw/rdma/vmw/pvrdma_main.c
@@ -286,8 +286,78 @@ static void init_ports(PVRDMADev *dev, Error **errp)
     }
 }
 
+static void uninit_msix(PCIDevice *pdev, int used_vectors)
+{
+    PVRDMADev *dev = PVRDMA_DEV(pdev);
+    int i;
+
+    for (i = 0; i < used_vectors; i++) {
+        msix_vector_unuse(pdev, i);
+    }
+
+    msix_uninit(pdev, &dev->msix, &dev->msix);
+}
+
+static int init_msix(PCIDevice *pdev, Error **errp)
+{
+    PVRDMADev *dev = PVRDMA_DEV(pdev);
+    int i;
+    int rc;
+
+    rc = msix_init(pdev, RDMA_MAX_INTRS, &dev->msix, RDMA_MSIX_BAR_IDX,
+                   RDMA_MSIX_TABLE, &dev->msix, RDMA_MSIX_BAR_IDX,
+                   RDMA_MSIX_PBA, 0, NULL);
+
+    if (rc < 0) {
+        error_setg(errp, "Failed to initialize MSI-X");
+        return rc;
+    }
+
+    for (i = 0; i < RDMA_MAX_INTRS; i++) {
+        rc = msix_vector_use(PCI_DEVICE(dev), i);
+        if (rc < 0) {
+            error_setg(errp, "Fail mark MSI-X vector %d", i);
+            uninit_msix(pdev, i);
+            return rc;
+        }
+    }
+
+    return 0;
+}
+
+static void pvrdma_fini(PCIDevice *pdev)
+{
+    PVRDMADev *dev = PVRDMA_DEV(pdev);
+
+    pr_dbg("Closing device %s %x.%x\n", pdev->name, PCI_SLOT(pdev->devfn),
+           PCI_FUNC(pdev->devfn));
+
+    pvrdma_qp_ops_fini();
+
+    rdma_rm_fini(&dev->rdma_dev_res);
+
+    rdma_backend_fini(&dev->backend_dev);
+
+    free_dsr(dev);
+
+    if (msix_enabled(pdev)) {
+        uninit_msix(pdev, RDMA_MAX_INTRS);
+    }
+}
+
+static void pvrdma_stop(PVRDMADev *dev)
+{
+    rdma_backend_stop(&dev->backend_dev);
+}
+
+static void pvrdma_start(PVRDMADev *dev)
+{
+    rdma_backend_start(&dev->backend_dev);
+}
+
 static void activate_device(PVRDMADev *dev)
 {
+    pvrdma_start(dev);
     set_reg_val(dev, PVRDMA_REG_ERR, 0);
     pr_dbg("Device activated\n");
 }
@@ -300,7 +370,10 @@ static int unquiesce_device(PVRDMADev *dev)
 
 static int reset_device(PVRDMADev *dev)
 {
+    pvrdma_stop(dev);
+
     pr_dbg("Device reset complete\n");
+
     return 0;
 }
 
@@ -469,45 +542,6 @@ static void init_regs(PCIDevice *pdev)
     set_reg_val(dev, PVRDMA_REG_ERR, 0xFFFF);
 }
 
-static void uninit_msix(PCIDevice *pdev, int used_vectors)
-{
-    PVRDMADev *dev = PVRDMA_DEV(pdev);
-    int i;
-
-    for (i = 0; i < used_vectors; i++) {
-        msix_vector_unuse(pdev, i);
-    }
-
-    msix_uninit(pdev, &dev->msix, &dev->msix);
-}
-
-static int init_msix(PCIDevice *pdev, Error **errp)
-{
-    PVRDMADev *dev = PVRDMA_DEV(pdev);
-    int i;
-    int rc;
-
-    rc = msix_init(pdev, RDMA_MAX_INTRS, &dev->msix, RDMA_MSIX_BAR_IDX,
-                   RDMA_MSIX_TABLE, &dev->msix, RDMA_MSIX_BAR_IDX,
-                   RDMA_MSIX_PBA, 0, NULL);
-
-    if (rc < 0) {
-        error_setg(errp, "Failed to initialize MSI-X");
-        return rc;
-    }
-
-    for (i = 0; i < RDMA_MAX_INTRS; i++) {
-        rc = msix_vector_use(PCI_DEVICE(dev), i);
-        if (rc < 0) {
-            error_setg(errp, "Fail mark MSI-X vercor %d", i);
-            uninit_msix(pdev, i);
-            return rc;
-        }
-    }
-
-    return 0;
-}
-
 static void init_dev_caps(PVRDMADev *dev)
 {
     size_t pg_tbl_bytes = TARGET_PAGE_SIZE *
@@ -602,22 +636,7 @@ out:
 
 static void pvrdma_exit(PCIDevice *pdev)
 {
-    PVRDMADev *dev = PVRDMA_DEV(pdev);
-
-    pr_dbg("Closing device %s %x.%x\n", pdev->name, PCI_SLOT(pdev->devfn),
-           PCI_FUNC(pdev->devfn));
-
-    pvrdma_qp_ops_fini();
-
-    rdma_rm_fini(&dev->rdma_dev_res);
-
-    rdma_backend_fini(&dev->backend_dev);
-
-    free_dsr(dev);
-
-    if (msix_enabled(pdev)) {
-        uninit_msix(pdev, RDMA_MAX_INTRS);
-    }
+    pvrdma_fini(pdev);
 }
 
 static void pvrdma_class_init(ObjectClass *klass, void *data)
-- 
2.17.1

[Qemu-devel] [PATCH PULL 02/15] hw/pvrdma: Bugfix - provide the correct attr_mask to query_qp

[Marcel Apfelbaum]

From: Yuval Shaia <yuval.shaia@oracle.com>

Calling rdma_rm_query_qp with attr_mask equals to -1 leads to error
where backend query_qp fails to retrieve the needed QP attributes.
Fix it by providing the attr_mask we got from driver.

Signed-off-by: Yuval Shaia <yuval.shaia@oracle.com>
Message-Id: <20180805153518.2983-3-yuval.shaia@oracle.com>
Reviewed-by: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
Signed-off-by: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
---
 hw/rdma/vmw/pvrdma_cmd.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/hw/rdma/vmw/pvrdma_cmd.c b/hw/rdma/vmw/pvrdma_cmd.c
index 14255d609f..e7d6589cdc 100644
--- a/hw/rdma/vmw/pvrdma_cmd.c
+++ b/hw/rdma/vmw/pvrdma_cmd.c
@@ -524,6 +524,7 @@ static int query_qp(PVRDMADev *dev, union pvrdma_cmd_req *req,
     struct ibv_qp_init_attr init_attr;
 
     pr_dbg("qp_handle=%d\n", cmd->qp_handle);
+    pr_dbg("attr_mask=0x%x\n", cmd->attr_mask);
 
     memset(rsp, 0, sizeof(*rsp));
     rsp->hdr.response = cmd->hdr.response;
@@ -531,8 +532,8 @@ static int query_qp(PVRDMADev *dev, union pvrdma_cmd_req *req,
 
     rsp->hdr.err = rdma_rm_query_qp(&dev->rdma_dev_res, &dev->backend_dev,
                                     cmd->qp_handle,
-                                    (struct ibv_qp_attr *)&resp->attrs, -1,
-                                    &init_attr);
+                                    (struct ibv_qp_attr *)&resp->attrs,
+                                    cmd->attr_mask, &init_attr);
 
     pr_dbg("ret=%d\n", rsp->hdr.err);
     return rsp->hdr.err;
-- 
2.17.1

[Qemu-devel] [PATCH PULL 03/15] hw/rdma: Modify debug macros

[Marcel Apfelbaum]

From: Yuval Shaia <yuval.shaia@oracle.com>

- Add line counter to ease navigation in log
- Print rdma instead of pvrdma

Signed-off-by: Yuval Shaia <yuval.shaia@oracle.com>
Reviewed-by: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
Message-Id: <20180805153518.2983-4-yuval.shaia@oracle.com>
Signed-off-by: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
---
 hw/rdma/rdma_utils.c      |  4 ++++
 hw/rdma/rdma_utils.h      | 16 ++++++++++++----
 hw/rdma/vmw/pvrdma_main.c |  2 ++
 3 files changed, 18 insertions(+), 4 deletions(-)

diff --git a/hw/rdma/rdma_utils.c b/hw/rdma/rdma_utils.c
index d713f635f1..dc23f158f3 100644
--- a/hw/rdma/rdma_utils.c
+++ b/hw/rdma/rdma_utils.c
@@ -15,6 +15,10 @@
 
 #include "rdma_utils.h"
 
+#ifdef PVRDMA_DEBUG
+unsigned long pr_dbg_cnt;
+#endif
+
 void *rdma_pci_dma_map(PCIDevice *dev, dma_addr_t addr, dma_addr_t plen)
 {
     void *p;
diff --git a/hw/rdma/rdma_utils.h b/hw/rdma/rdma_utils.h
index 3dc07891bc..04c7c2ef5b 100644
--- a/hw/rdma/rdma_utils.h
+++ b/hw/rdma/rdma_utils.h
@@ -22,18 +22,26 @@
 #include "sysemu/dma.h"
 
 #define pr_info(fmt, ...) \
-    fprintf(stdout, "%s: %-20s (%3d): " fmt, "pvrdma",  __func__, __LINE__,\
+    fprintf(stdout, "%s: %-20s (%3d): " fmt, "rdma",  __func__, __LINE__,\
            ## __VA_ARGS__)
 
 #define pr_err(fmt, ...) \
-    fprintf(stderr, "%s: Error at %-20s (%3d): " fmt, "pvrdma", __func__, \
+    fprintf(stderr, "%s: Error at %-20s (%3d): " fmt, "rdma", __func__, \
         __LINE__, ## __VA_ARGS__)
 
 #ifdef PVRDMA_DEBUG
+extern unsigned long pr_dbg_cnt;
+
+#define init_pr_dbg(void) \
+{ \
+    pr_dbg_cnt = 0; \
+}
+
 #define pr_dbg(fmt, ...) \
-    fprintf(stdout, "%s: %-20s (%3d): " fmt, "pvrdma", __func__, __LINE__,\
-           ## __VA_ARGS__)
+    fprintf(stdout, "%lx %ld: %-20s (%3d): " fmt, pthread_self(), pr_dbg_cnt++, \
+            __func__, __LINE__, ## __VA_ARGS__)
 #else
+#define init_pr_dbg(void)
 #define pr_dbg(fmt, ...)
 #endif
 
diff --git a/hw/rdma/vmw/pvrdma_main.c b/hw/rdma/vmw/pvrdma_main.c
index 6a5073974d..1b1330e113 100644
--- a/hw/rdma/vmw/pvrdma_main.c
+++ b/hw/rdma/vmw/pvrdma_main.c
@@ -577,6 +577,8 @@ static void pvrdma_realize(PCIDevice *pdev, Error **errp)
     Object *memdev_root;
     bool ram_shared = false;
 
+    init_pr_dbg();
+
     pr_dbg("Initializing device %s %x.%x\n", pdev->name,
            PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn));
 
-- 
2.17.1

[Qemu-devel] [PATCH PULL 04/15] hw/pvrdma: Clean CQE before use

[Marcel Apfelbaum]

From: Yuval Shaia <yuval.shaia@oracle.com>

Next CQE is fetched from CQ ring, clean it before usage as it still
carries old CQE values.

Signed-off-by: Yuval Shaia <yuval.shaia@oracle.com>
Reviewed-by: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
Message-Id: <20180805153518.2983-5-yuval.shaia@oracle.com>
Signed-off-by: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
---
 hw/rdma/vmw/pvrdma_qp_ops.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/hw/rdma/vmw/pvrdma_qp_ops.c b/hw/rdma/vmw/pvrdma_qp_ops.c
index 99bb51111e..a8664f40c8 100644
--- a/hw/rdma/vmw/pvrdma_qp_ops.c
+++ b/hw/rdma/vmw/pvrdma_qp_ops.c
@@ -69,6 +69,7 @@ static int pvrdma_post_cqe(PVRDMADev *dev, uint32_t cq_handle,
         return -EINVAL;
     }
 
+    memset(cqe1, 0, sizeof(*cqe1));
     cqe1->wr_id = cqe->wr_id;
     cqe1->qp = cqe->qp;
     cqe1->opcode = cqe->opcode;
-- 
2.17.1

[Qemu-devel] [PATCH PULL 05/15] hw/pvrdma: Make default pkey 0xFFFF

[Marcel Apfelbaum]

From: Yuval Shaia <yuval.shaia@oracle.com>

0x7FFF is not the default pkey - fix it.

Signed-off-by: Yuval Shaia <yuval.shaia@oracle.com>
Reviewed-by: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
Message-Id: <20180805153518.2983-6-yuval.shaia@oracle.com>
Signed-off-by: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
---
 hw/rdma/vmw/pvrdma.h     | 3 +++
 hw/rdma/vmw/pvrdma_cmd.c | 2 +-
 2 files changed, 4 insertions(+), 1 deletion(-)

diff --git a/hw/rdma/vmw/pvrdma.h b/hw/rdma/vmw/pvrdma.h
index 81e0e0e99c..e2d9f93cdf 100644
--- a/hw/rdma/vmw/pvrdma.h
+++ b/hw/rdma/vmw/pvrdma.h
@@ -50,6 +50,9 @@
 #define PVRDMA_HW_VERSION    17
 #define PVRDMA_FW_VERSION    14
 
+/* Some defaults */
+#define PVRDMA_PKEY          0x7FFF
+
 typedef struct DSRInfo {
     dma_addr_t dma;
     struct pvrdma_device_shared_region *dsr;
diff --git a/hw/rdma/vmw/pvrdma_cmd.c b/hw/rdma/vmw/pvrdma_cmd.c
index e7d6589cdc..4f74b29936 100644
--- a/hw/rdma/vmw/pvrdma_cmd.c
+++ b/hw/rdma/vmw/pvrdma_cmd.c
@@ -166,7 +166,7 @@ static int query_pkey(PVRDMADev *dev, union pvrdma_cmd_req *req,
     resp->hdr.ack = PVRDMA_CMD_QUERY_PKEY_RESP;
     resp->hdr.err = 0;
 
-    resp->pkey = 0x7FFF;
+    resp->pkey = PVRDMA_PKEY;
     pr_dbg("pkey=0x%x\n", resp->pkey);
 
     return 0;
-- 
2.17.1

[Qemu-devel] [PATCH PULL 06/15] hw/rdma: Delete useless structure RdmaRmUserMR

[Marcel Apfelbaum]

From: Yuval Shaia <yuval.shaia@oracle.com>

The structure RdmaRmUserMR has no benefits, remove it an move all its
fields to struct RdmaRmMR.

Reviewed-by: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
Signed-off-by: Yuval Shaia <yuval.shaia@oracle.com>
Message-Id: <20180805153518.2983-7-yuval.shaia@oracle.com>
Signed-off-by: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
---
 hw/rdma/rdma_backend.c |  3 +--
 hw/rdma/rdma_rm.c      | 16 ++++++++--------
 hw/rdma/rdma_rm_defs.h | 10 +++-------
 3 files changed, 12 insertions(+), 17 deletions(-)

diff --git a/hw/rdma/rdma_backend.c b/hw/rdma/rdma_backend.c
index 647e16399f..52981d652d 100644
--- a/hw/rdma/rdma_backend.c
+++ b/hw/rdma/rdma_backend.c
@@ -271,8 +271,7 @@ static int build_host_sge_array(RdmaDeviceResources *rdma_dev_res,
             return VENDOR_ERR_INVLKEY | ssge[ssge_idx].lkey;
         }
 
-        dsge->addr = (uintptr_t)mr->user_mr.host_virt + ssge[ssge_idx].addr -
-                     mr->user_mr.guest_start;
+        dsge->addr = (uintptr_t)mr->virt + ssge[ssge_idx].addr - mr->start;
         dsge->length = ssge[ssge_idx].length;
         dsge->lkey = rdma_backend_mr_lkey(&mr->backend_mr);
 
diff --git a/hw/rdma/rdma_rm.c b/hw/rdma/rdma_rm.c
index 415da15efe..7403d24674 100644
--- a/hw/rdma/rdma_rm.c
+++ b/hw/rdma/rdma_rm.c
@@ -165,15 +165,15 @@ int rdma_rm_alloc_mr(RdmaDeviceResources *dev_res, uint32_t pd_handle,
         length = TARGET_PAGE_SIZE;
         addr = g_malloc(length);
     } else {
-        mr->user_mr.host_virt = host_virt;
-        pr_dbg("host_virt=0x%p\n", mr->user_mr.host_virt);
-        mr->user_mr.length = guest_length;
+        mr->virt = host_virt;
+        pr_dbg("host_virt=0x%p\n", mr->virt);
+        mr->length = guest_length;
         pr_dbg("length=%zu\n", guest_length);
-        mr->user_mr.guest_start = guest_start;
-        pr_dbg("guest_start=0x%" PRIx64 "\n", mr->user_mr.guest_start);
+        mr->start = guest_start;
+        pr_dbg("guest_start=0x%" PRIx64 "\n", mr->start);
 
-        length = mr->user_mr.length;
-        addr = mr->user_mr.host_virt;
+        length = mr->length;
+        addr = mr->virt;
     }
 
     ret = rdma_backend_create_mr(&mr->backend_mr, &pd->backend_pd, addr, length,
@@ -214,7 +214,7 @@ void rdma_rm_dealloc_mr(RdmaDeviceResources *dev_res, uint32_t mr_handle)
 
     if (mr) {
         rdma_backend_destroy_mr(&mr->backend_mr);
-        munmap(mr->user_mr.host_virt, mr->user_mr.length);
+        munmap(mr->virt, mr->length);
         res_tbl_dealloc(&dev_res->mr_tbl, mr_handle);
     }
 }
diff --git a/hw/rdma/rdma_rm_defs.h b/hw/rdma/rdma_rm_defs.h
index 226011176d..7228151239 100644
--- a/hw/rdma/rdma_rm_defs.h
+++ b/hw/rdma/rdma_rm_defs.h
@@ -55,16 +55,12 @@ typedef struct RdmaRmCQ {
     bool notify;
 } RdmaRmCQ;
 
-typedef struct RdmaRmUserMR {
-    void *host_virt;
-    uint64_t guest_start;
-    size_t length;
-} RdmaRmUserMR;
-
 /* MR (DMA region) */
 typedef struct RdmaRmMR {
     RdmaBackendMR backend_mr;
-    RdmaRmUserMR user_mr;
+    void *virt;
+    uint64_t start;
+    size_t length;
     uint32_t pd_handle;
     uint32_t lkey;
     uint32_t rkey;
-- 
2.17.1

[Qemu-devel] [PATCH PULL 07/15] hw/rdma: Do not allocate memory for non-dma MR

[Marcel Apfelbaum]

From: Yuval Shaia <yuval.shaia@oracle.com>

There is no use in the memory allocated for non-dma MR.
Delete the code that allocates it.

Signed-off-by: Yuval Shaia <yuval.shaia@oracle.com>
Reviewed-by: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
Message-Id: <20180805153518.2983-8-yuval.shaia@oracle.com>
Signed-off-by: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
---
 hw/rdma/rdma_rm.c | 52 +++++++++++++++++++----------------------------
 1 file changed, 21 insertions(+), 31 deletions(-)

diff --git a/hw/rdma/rdma_rm.c b/hw/rdma/rdma_rm.c
index 7403d24674..bf4a5c71b4 100644
--- a/hw/rdma/rdma_rm.c
+++ b/hw/rdma/rdma_rm.c
@@ -144,8 +144,6 @@ int rdma_rm_alloc_mr(RdmaDeviceResources *dev_res, uint32_t pd_handle,
     RdmaRmMR *mr;
     int ret = 0;
     RdmaRmPD *pd;
-    void *addr;
-    size_t length;
 
     pd = rdma_rm_get_pd(dev_res, pd_handle);
     if (!pd) {
@@ -158,40 +156,29 @@ int rdma_rm_alloc_mr(RdmaDeviceResources *dev_res, uint32_t pd_handle,
         pr_dbg("Failed to allocate obj in table\n");
         return -ENOMEM;
     }
+    pr_dbg("mr_handle=%d\n", *mr_handle);
 
-    if (!host_virt) {
-        /* TODO: This is my guess but not so sure that this needs to be
-         * done */
-        length = TARGET_PAGE_SIZE;
-        addr = g_malloc(length);
-    } else {
+    pr_dbg("host_virt=0x%p\n", host_virt);
+    pr_dbg("guest_start=0x%" PRIx64 "\n", guest_start);
+    pr_dbg("length=%zu\n", guest_length);
+
+    if (host_virt) {
         mr->virt = host_virt;
-        pr_dbg("host_virt=0x%p\n", mr->virt);
-        mr->length = guest_length;
-        pr_dbg("length=%zu\n", guest_length);
         mr->start = guest_start;
-        pr_dbg("guest_start=0x%" PRIx64 "\n", mr->start);
+        mr->length = guest_length;
 
-        length = mr->length;
-        addr = mr->virt;
+        ret = rdma_backend_create_mr(&mr->backend_mr, &pd->backend_pd, mr->virt,
+                                     mr->length, access_flags);
+        if (ret) {
+            pr_dbg("Fail in rdma_backend_create_mr, err=%d\n", ret);
+            ret = -EIO;
+            goto out_dealloc_mr;
+        }
     }
 
-    ret = rdma_backend_create_mr(&mr->backend_mr, &pd->backend_pd, addr, length,
-                                 access_flags);
-    if (ret) {
-        pr_dbg("Fail in rdma_backend_create_mr, err=%d\n", ret);
-        ret = -EIO;
-        goto out_dealloc_mr;
-    }
-
-    if (!host_virt) {
-        *lkey = mr->lkey = rdma_backend_mr_lkey(&mr->backend_mr);
-        *rkey = mr->rkey = rdma_backend_mr_rkey(&mr->backend_mr);
-    } else {
-        /* We keep mr_handle in lkey so send and recv get get mr ptr */
-        *lkey = *mr_handle;
-        *rkey = -1;
-    }
+    /* We keep mr_handle in lkey so send and recv get get mr ptr */
+    *lkey = *mr_handle;
+    *rkey = -1;
 
     mr->pd_handle = pd_handle;
 
@@ -214,7 +201,10 @@ void rdma_rm_dealloc_mr(RdmaDeviceResources *dev_res, uint32_t mr_handle)
 
     if (mr) {
         rdma_backend_destroy_mr(&mr->backend_mr);
-        munmap(mr->virt, mr->length);
+        pr_dbg("start=0x%" PRIx64 "\n", mr->start);
+        if (mr->start) {
+            munmap(mr->virt, mr->length);
+        }
         res_tbl_dealloc(&dev_res->mr_tbl, mr_handle);
     }
 }
-- 
2.17.1

[Qemu-devel] [PATCH PULL 08/15] hw/rdma: Reorder resource cleanup

[Marcel Apfelbaum]

From: Yuval Shaia <yuval.shaia@oracle.com>

To be consistence with allocation do the reverse order in deallocation

Signed-off-by: Yuval Shaia <yuval.shaia@oracle.com>
Reviewed-by: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
Message-Id: <20180805153518.2983-9-yuval.shaia@oracle.com>
Signed-off-by: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
---
 hw/rdma/rdma_rm.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/hw/rdma/rdma_rm.c b/hw/rdma/rdma_rm.c
index bf4a5c71b4..1f014b4ab2 100644
--- a/hw/rdma/rdma_rm.c
+++ b/hw/rdma/rdma_rm.c
@@ -543,8 +543,9 @@ void rdma_rm_fini(RdmaDeviceResources *dev_res)
     res_tbl_free(&dev_res->uc_tbl);
     res_tbl_free(&dev_res->cqe_ctx_tbl);
     res_tbl_free(&dev_res->qp_tbl);
-    res_tbl_free(&dev_res->cq_tbl);
     res_tbl_free(&dev_res->mr_tbl);
+    res_tbl_free(&dev_res->cq_tbl);
     res_tbl_free(&dev_res->pd_tbl);
+
     g_hash_table_destroy(dev_res->qp_hash);
 }
-- 
2.17.1

[Qemu-devel] [PATCH PULL 09/15] hw/pvrdma: Cosmetic change - indent right

[Marcel Apfelbaum]

From: Yuval Shaia <yuval.shaia@oracle.com>

Signed-off-by: Yuval Shaia <yuval.shaia@oracle.com>
Reviewed-by: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
Message-Id: <20180805153518.2983-10-yuval.shaia@oracle.com>
Signed-off-by: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
---
 hw/rdma/vmw/pvrdma_main.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/hw/rdma/vmw/pvrdma_main.c b/hw/rdma/vmw/pvrdma_main.c
index 1b1330e113..3d448bffc4 100644
--- a/hw/rdma/vmw/pvrdma_main.c
+++ b/hw/rdma/vmw/pvrdma_main.c
@@ -430,7 +430,7 @@ static void regs_write(void *opaque, hwaddr addr, uint64_t val, unsigned size)
             reset_device(dev);
             break;
         }
-    break;
+        break;
     case PVRDMA_REG_IMR:
         pr_dbg("Interrupt mask=0x%" PRIx64 "\n", val);
         dev->interrupt_mask = val;
@@ -439,7 +439,7 @@ static void regs_write(void *opaque, hwaddr addr, uint64_t val, unsigned size)
         if (val == 0) {
             execute_command(dev);
         }
-    break;
+        break;
     default:
         break;
     }
-- 
2.17.1

[Qemu-devel] [PATCH PULL 10/15] hw/rdma: Cosmetic change - move to generic function

[Marcel Apfelbaum]

From: Yuval Shaia <yuval.shaia@oracle.com>

To ease maintenance of struct comp_thread move all related code to
dedicated function.

Signed-off-by: Yuval Shaia <yuval.shaia@oracle.com>
Reviewed-by: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
Message-Id: <20180805153518.2983-11-yuval.shaia@oracle.com>
Signed-off-by: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
---
 hw/rdma/rdma_backend.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/hw/rdma/rdma_backend.c b/hw/rdma/rdma_backend.c
index 52981d652d..d29acc505b 100644
--- a/hw/rdma/rdma_backend.c
+++ b/hw/rdma/rdma_backend.c
@@ -146,10 +146,10 @@ static void *comp_handler_thread(void *arg)
     return NULL;
 }
 
-static void stop_comp_thread(RdmaBackendDev *backend_dev)
+static void stop_backend_thread(RdmaBackendThread *thread)
 {
-    backend_dev->comp_thread.run = false;
-    while (backend_dev->comp_thread.is_running) {
+    thread->run = false;
+    while (thread->is_running) {
         pr_dbg("Waiting for thread to complete\n");
         sleep(THR_POLL_TO / SCALE_US / 2);
     }
@@ -159,7 +159,7 @@ static void start_comp_thread(RdmaBackendDev *backend_dev)
 {
     char thread_name[THR_NAME_LEN] = {0};
 
-    stop_comp_thread(backend_dev);
+    stop_backend_thread(&backend_dev->comp_thread);
 
     snprintf(thread_name, sizeof(thread_name), "rdma_comp_%s",
              ibv_get_device_name(backend_dev->ib_dev));
@@ -876,7 +876,7 @@ void rdma_backend_start(RdmaBackendDev *backend_dev)
 void rdma_backend_stop(RdmaBackendDev *backend_dev)
 {
     pr_dbg("Stopping rdma_backend\n");
-    stop_comp_thread(backend_dev);
+    stop_backend_thread(&backend_dev->comp_thread);
 }
 
 void rdma_backend_fini(RdmaBackendDev *backend_dev)
-- 
2.17.1

[Qemu-devel] [PATCH PULL 11/15] hw/rdma: Print backend QP number in hex format

[Marcel Apfelbaum]

From: Yuval Shaia <yuval.shaia@oracle.com>

To be consistent with other prints throughout the code fix places that
print it as decimal number.

Signed-off-by: Yuval Shaia <yuval.shaia@oracle.com>
Reviewed-by: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
Message-Id: <20180805153518.2983-12-yuval.shaia@oracle.com>
Signed-off-by: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
---
 hw/rdma/rdma_rm.c           | 4 ++--
 hw/rdma/vmw/pvrdma_qp_ops.c | 4 ++--
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/hw/rdma/rdma_rm.c b/hw/rdma/rdma_rm.c
index 1f014b4ab2..859c900003 100644
--- a/hw/rdma/rdma_rm.c
+++ b/hw/rdma/rdma_rm.c
@@ -389,7 +389,7 @@ int rdma_rm_modify_qp(RdmaDeviceResources *dev_res, RdmaBackendDev *backend_dev,
     RdmaRmQP *qp;
     int ret;
 
-    pr_dbg("qpn=%d\n", qp_handle);
+    pr_dbg("qpn=0x%x\n", qp_handle);
 
     qp = rdma_rm_get_qp(dev_res, qp_handle);
     if (!qp) {
@@ -447,7 +447,7 @@ int rdma_rm_query_qp(RdmaDeviceResources *dev_res, RdmaBackendDev *backend_dev,
 {
     RdmaRmQP *qp;
 
-    pr_dbg("qpn=%d\n", qp_handle);
+    pr_dbg("qpn=0x%x\n", qp_handle);
 
     qp = rdma_rm_get_qp(dev_res, qp_handle);
     if (!qp) {
diff --git a/hw/rdma/vmw/pvrdma_qp_ops.c b/hw/rdma/vmw/pvrdma_qp_ops.c
index a8664f40c8..c668afd0ed 100644
--- a/hw/rdma/vmw/pvrdma_qp_ops.c
+++ b/hw/rdma/vmw/pvrdma_qp_ops.c
@@ -130,7 +130,7 @@ int pvrdma_qp_send(PVRDMADev *dev, uint32_t qp_handle)
     PvrdmaSqWqe *wqe;
     PvrdmaRing *ring;
 
-    pr_dbg("qp_handle=%d\n", qp_handle);
+    pr_dbg("qp_handle=0x%x\n", qp_handle);
 
     qp = rdma_rm_get_qp(&dev->rdma_dev_res, qp_handle);
     if (unlikely(!qp)) {
@@ -174,7 +174,7 @@ int pvrdma_qp_recv(PVRDMADev *dev, uint32_t qp_handle)
     PvrdmaRqWqe *wqe;
     PvrdmaRing *ring;
 
-    pr_dbg("qp_handle=%d\n", qp_handle);
+    pr_dbg("qp_handle=0x%x\n", qp_handle);
 
     qp = rdma_rm_get_qp(&dev->rdma_dev_res, qp_handle);
     if (unlikely(!qp)) {
-- 
2.17.1

[Qemu-devel] [PATCH PULL 12/15] hw/rdma: Bugfix - Support non-aligned buffers

[Marcel Apfelbaum]

From: Yuval Shaia <yuval.shaia@oracle.com>

RDMA application can provide non-aligned buffers to be registered. In
such case the DMA address passed by driver is pointing to the beginning
of the physical address of the mapped page so we can't distinguish
between two addresses from the same page.

Fix it by keeping the offset of the virtual address in mr->virt.

Signed-off-by: Yuval Shaia <yuval.shaia@oracle.com>
Reviewed-by: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
Message-Id: <20180805153518.2983-13-yuval.shaia@oracle.com>
Signed-off-by: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
---
 hw/rdma/rdma_rm.c        | 2 ++
 hw/rdma/vmw/pvrdma_cmd.c | 1 +
 2 files changed, 3 insertions(+)

diff --git a/hw/rdma/rdma_rm.c b/hw/rdma/rdma_rm.c
index 859c900003..8d59a42cd1 100644
--- a/hw/rdma/rdma_rm.c
+++ b/hw/rdma/rdma_rm.c
@@ -166,6 +166,7 @@ int rdma_rm_alloc_mr(RdmaDeviceResources *dev_res, uint32_t pd_handle,
         mr->virt = host_virt;
         mr->start = guest_start;
         mr->length = guest_length;
+        mr->virt += (mr->start & (TARGET_PAGE_SIZE - 1));
 
         ret = rdma_backend_create_mr(&mr->backend_mr, &pd->backend_pd, mr->virt,
                                      mr->length, access_flags);
@@ -203,6 +204,7 @@ void rdma_rm_dealloc_mr(RdmaDeviceResources *dev_res, uint32_t mr_handle)
         rdma_backend_destroy_mr(&mr->backend_mr);
         pr_dbg("start=0x%" PRIx64 "\n", mr->start);
         if (mr->start) {
+            mr->virt -= (mr->start & (TARGET_PAGE_SIZE - 1));
             munmap(mr->virt, mr->length);
         }
         res_tbl_dealloc(&dev_res->mr_tbl, mr_handle);
diff --git a/hw/rdma/vmw/pvrdma_cmd.c b/hw/rdma/vmw/pvrdma_cmd.c
index 4f74b29936..b74b0b48d0 100644
--- a/hw/rdma/vmw/pvrdma_cmd.c
+++ b/hw/rdma/vmw/pvrdma_cmd.c
@@ -59,6 +59,7 @@ static void *pvrdma_map_to_pdir(PCIDevice *pdev, uint64_t pdir_dma,
     }
 
     host_virt = mremap(curr_page, 0, length, MREMAP_MAYMOVE);
+    pr_dbg("mremap %p -> %p\n", curr_page, host_virt);
     if (host_virt == MAP_FAILED) {
         host_virt = NULL;
         error_report("PVRDMA: Failed to remap memory for host_virt");
-- 
2.17.1

[Qemu-devel] [PATCH PULL 13/15] hw/rdma: Add reference to pci_dev in backend_dev

[Marcel Apfelbaum]

From: Yuval Shaia <yuval.shaia@oracle.com>

The field backend_dev->dev is not initialized, fix it.

Signed-off-by: Yuval Shaia <yuval.shaia@oracle.com>
Message-Id: <20180805153518.2983-14-yuval.shaia@oracle.com>
Reviewed-by: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
Signed-off-by: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
---
 hw/rdma/rdma_backend.c    | 6 +++++-
 hw/rdma/rdma_backend.h    | 2 +-
 hw/rdma/vmw/pvrdma_main.c | 2 +-
 3 files changed, 7 insertions(+), 3 deletions(-)

diff --git a/hw/rdma/rdma_backend.c b/hw/rdma/rdma_backend.c
index d29acc505b..d7a4bbd91f 100644
--- a/hw/rdma/rdma_backend.c
+++ b/hw/rdma/rdma_backend.c
@@ -744,7 +744,7 @@ static int init_device_caps(RdmaBackendDev *backend_dev,
     return 0;
 }
 
-int rdma_backend_init(RdmaBackendDev *backend_dev,
+int rdma_backend_init(RdmaBackendDev *backend_dev, PCIDevice *pdev,
                       RdmaDeviceResources *rdma_dev_res,
                       const char *backend_device_name, uint8_t port_num,
                       uint8_t backend_gid_idx, struct ibv_device_attr *dev_attr,
@@ -756,6 +756,10 @@ int rdma_backend_init(RdmaBackendDev *backend_dev,
     struct ibv_device **dev_list;
     struct ibv_port_attr port_attr;
 
+    memset(backend_dev, 0, sizeof(*backend_dev));
+
+    backend_dev->dev = pdev;
+
     backend_dev->backend_gid_idx = backend_gid_idx;
     backend_dev->port_num = port_num;
     backend_dev->rdma_dev_res = rdma_dev_res;
diff --git a/hw/rdma/rdma_backend.h b/hw/rdma/rdma_backend.h
index 3049a73962..86e8fe8ab6 100644
--- a/hw/rdma/rdma_backend.h
+++ b/hw/rdma/rdma_backend.h
@@ -46,7 +46,7 @@ static inline uint32_t rdma_backend_mr_rkey(const RdmaBackendMR *mr)
     return mr->ibmr ? mr->ibmr->rkey : 0;
 }
 
-int rdma_backend_init(RdmaBackendDev *backend_dev,
+int rdma_backend_init(RdmaBackendDev *backend_dev, PCIDevice *pdev,
                       RdmaDeviceResources *rdma_dev_res,
                       const char *backend_device_name, uint8_t port_num,
                       uint8_t backend_gid_idx, struct ibv_device_attr *dev_attr,
diff --git a/hw/rdma/vmw/pvrdma_main.c b/hw/rdma/vmw/pvrdma_main.c
index 3d448bffc4..ca5fa8d981 100644
--- a/hw/rdma/vmw/pvrdma_main.c
+++ b/hw/rdma/vmw/pvrdma_main.c
@@ -611,7 +611,7 @@ static void pvrdma_realize(PCIDevice *pdev, Error **errp)
         goto out;
     }
 
-    rc = rdma_backend_init(&dev->backend_dev, &dev->rdma_dev_res,
+    rc = rdma_backend_init(&dev->backend_dev, pdev, &dev->rdma_dev_res,
                            dev->backend_device_name, dev->backend_port_num,
                            dev->backend_gid_idx, &dev->dev_attr, errp);
     if (rc) {
-- 
2.17.1

[Qemu-devel] [PATCH PULL 14/15] hw/pvrdma: remove not needed include

[Marcel Apfelbaum]

No need to include linux/types.h, is empty anyway.

Suggested-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
Message-Id: <20180811171534.11917-1-marcel.apfelbaum@gmail.com>
Reviewed-by: Yuval Shaia <yuval.shaia@oracle.com>
---
 hw/rdma/vmw/pvrdma_cmd.c | 1 -
 1 file changed, 1 deletion(-)

diff --git a/hw/rdma/vmw/pvrdma_cmd.c b/hw/rdma/vmw/pvrdma_cmd.c
index b74b0b48d0..4faeb21631 100644
--- a/hw/rdma/vmw/pvrdma_cmd.c
+++ b/hw/rdma/vmw/pvrdma_cmd.c
@@ -16,7 +16,6 @@
 #include "qemu/osdep.h"
 #include "qemu/error-report.h"
 #include "cpu.h"
-#include <linux/types.h>
 #include "hw/hw.h"
 #include "hw/pci/pci.h"
 #include "hw/pci/pci_ids.h"
-- 
2.17.1

[Qemu-devel] [PATCH PULL 15/15] config: split PVRDMA from RDMA

[Marcel Apfelbaum]

In some BSD systems RDMA migration is possible while
the pvrdma device can't be used because the mremap system call
is missing.

Reported-by: Rebecca Cran <rebecca@bluestop.org>
Signed-off-by: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
Message-Id: <20180816151637.24553-1-marcel.apfelbaum@gmail.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
---
 configure             | 55 ++++++++++++++++++++++++++++++++++++++++++-
 hw/rdma/Makefile.objs |  2 +-
 2 files changed, 55 insertions(+), 2 deletions(-)

diff --git a/configure b/configure
index db97930314..7d9a63636c 100755
--- a/configure
+++ b/configure
@@ -375,6 +375,7 @@ hax="no"
 hvf="no"
 whpx="no"
 rdma=""
+pvrdma=""
 gprof="no"
 debug_tcg="no"
 debug="no"
@@ -1363,6 +1364,10 @@ for opt do
   ;;
   --disable-rdma) rdma="no"
   ;;
+  --enable-pvrdma) pvrdma="yes"
+  ;;
+  --disable-pvrdma) pvrdma="no"
+  ;;
   --with-gtkabi=*) gtkabi="$optarg"
   ;;
   --disable-vte) vte="no"
@@ -1669,7 +1674,8 @@ disabled with --disable-FEATURE, default is enabled if available:
   hax             HAX acceleration support
   hvf             Hypervisor.framework acceleration support
   whpx            Windows Hypervisor Platform acceleration support
-  rdma            Enable RDMA-based migration and PVRDMA support
+  rdma            Enable RDMA-based migration
+  pvrdma          Enable PVRDMA support
   vde             support for vde network
   netmap          support for netmap network
   linux-aio       Linux AIO support
@@ -3064,6 +3070,48 @@ EOF
   fi
 fi
 
+##########################################
+# PVRDMA detection
+
+cat > $TMPC <<EOF &&
+#include <sys/mman.h>
+
+int
+main(void)
+{
+    char buf = 0;
+    void *addr = &buf;
+    addr = mremap(addr, 0, 1, MREMAP_MAYMOVE | MREMAP_FIXED);
+
+    return 0;
+}
+EOF
+
+if test "$rdma" = "yes" ; then
+    case "$pvrdma" in
+    "")
+        if compile_prog "" ""; then
+            pvrdma="yes"
+        else
+            pvrdma="no"
+        fi
+        ;;
+    "yes")
+        if ! compile_prog "" ""; then
+            error_exit "PVRDMA is not supported since mremap is not implemented"
+        fi
+        pvrdma="yes"
+        ;;
+    "no")
+        pvrdma="no"
+        ;;
+    esac
+else
+    if test "$pvrdma" = "yes" ; then
+        error_exit "PVRDMA requires rdma suppport"
+    fi
+    pvrdma="no"
+fi
 
 ##########################################
 # VNC SASL detection
@@ -5952,6 +6000,7 @@ if test "$tcg" = "yes" ; then
 fi
 echo "malloc trim support $malloc_trim"
 echo "RDMA support      $rdma"
+echo "PVRDMA support    $pvrdma"
 echo "fdt support       $fdt"
 echo "membarrier        $membarrier"
 echo "preadv support    $preadv"
@@ -6708,6 +6757,10 @@ if test "$rdma" = "yes" ; then
   echo "RDMA_LIBS=$rdma_libs" >> $config_host_mak
 fi
 
+if test "$pvrdma" = "yes" ; then
+  echo "CONFIG_PVRDMA=y" >> $config_host_mak
+fi
+
 if test "$have_rtnetlink" = "yes" ; then
   echo "CONFIG_RTNETLINK=y" >> $config_host_mak
 fi
diff --git a/hw/rdma/Makefile.objs b/hw/rdma/Makefile.objs
index 3504c39d21..bd36cbf51c 100644
--- a/hw/rdma/Makefile.objs
+++ b/hw/rdma/Makefile.objs
@@ -1,4 +1,4 @@
-ifeq ($(CONFIG_RDMA),y)
+ifeq ($(CONFIG_PVRDMA),y)
 obj-$(CONFIG_PCI) += rdma_utils.o rdma_backend.o rdma_rm.o
 obj-$(CONFIG_PCI) += vmw/pvrdma_dev_ring.o vmw/pvrdma_cmd.o \
                      vmw/pvrdma_qp_ops.o vmw/pvrdma_main.o
-- 
2.17.1

Re: [Qemu-devel] [PATCH PULL 00/15] RDMA queue

[Peter Maydell]

On 19 August 2018 at 21:22, Marcel Apfelbaum <marcel.apfelbaum@gmail.com> wrote:
> The following changes since commit 0abaa41d936becd914a16ee1fe2a981d96d19428:
>
>   Merge remote-tracking branch 'remotes/ehabkost/tags/x86-next-pull-request' into staging (2018-08-17 09:46:00 +0100)
>
> are available in the Git repository at:
>
>   https://github.com/marcel-apf/qemu tags/rdma-pull-request
>
> for you to fetch changes up to 21ab34c9543fe1b6d31b3edbd01a397e7e090d00:
>
>   config: split PVRDMA from RDMA (2018-08-18 18:01:34 +0300)
>
> ----------------------------------------------------------------
> Various pvrdma fixes
> Split pvrdma from rdma in config
>
> ----------------------------------------------------------------

Applied, thanks.

-- PMM