Changeset 84468 in vbox for trunk/src/VBox/Devices/Network

Timestamp:

May 23, 2020 7:17:02 AM (5 years ago)

Author:

vboxsync

svn:sync-xref-src-repo-rev:

138183

Message:

Network/DevVirtioNet_1.0.cpp: Removed force interrupt flag from API and calls since it is no longer a necessary workardound option due to fixing how avail ring flags are read.

File:

• trunk/src/VBox/Devices/Network/DevVirtioNet_1_0.cpp (modified) (4 diffs)

trunk/src/VBox/Devices/Network/DevVirtioNet_1_0.cpp

@@ -1600 +1600 @@
                   rc);
 
-    /** @todo   WHY *must* we *force* notifying guest that we filled its Rx buffer(s)?
-     *          If we don't notify the guest, it doesn't detect it and stalls, even though
-     *          guest is responsible for setting the used-ring flag in the Rx queue that tells
-     *          us to skip the notification interrupt! Obviously forcing the interrupt is
-     *          non-optimal performance-wise and seems to contradict the Virtio spec.
-     *          Is that a bug in the linux virtio_net.c driver? */
-
-    virtioCoreQueueSync(pDevIns, &pThis->Virtio, RXQIDX(idxRxQueue), /* fForce */ true);
+    virtioCoreQueueSync(pDevIns, &pThis->Virtio, RXQIDX(idxRxQueue));
 
     return VINF_SUCCESS;
@@ -2063 +2056 @@
 
     virtioCoreR3QueuePut(pDevIns, &pThis->Virtio, CTRLQIDX, pReturnSegBuf, pDescChain, true);
-    virtioCoreQueueSync(pDevIns, &pThis->Virtio, CTRLQIDX, false);
+    virtioCoreQueueSync(pDevIns, &pThis->Virtio, CTRLQIDX);
 
     for (int i = 0; i < cSegs; i++)
@@ -2321 +2314 @@
 
             /* Update used ring idx and notify guest that we've transmitted the data it sent */
-            virtioCoreQueueSync(pVirtio->pDevInsR3, pVirtio, idxTxQueue, false);
+            virtioCoreQueueSync(pVirtio->pDevInsR3, pVirtio, idxTxQueue);
         }
 
@@ -2764 +2757 @@
 
     LEAVE_CRITICAL_SECTION;
-    AssertRelease(!pThisCC->pDrvBase);
     return rc;
 }