d/numam-dpdk
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity
377af5f38e
numam-dpdk/drivers/net/vmxnet3/vmxnet3_rxtx.c

1105 lines
30 KiB
C
Raw Normal View History

vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
/*-
* BSD LICENSE
remove trailing whitespaces This commit removes trailing whitespace from lines in files. Almost all files are affected, as the BSD license copyright header had trailing whitespace on 4 lines in it [hence the number of files reporting 8 lines changed in the diffstat]. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com> Acked-by: Neil Horman <nhorman@tuxdriver.com> [Thomas: remove spaces before tabs in libs] [Thomas: remove more trailing spaces in non-C files] Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-03 23:42:50 +00:00
*
vmxnet3: move to drivers/net/ Move vmxnet3 PMD to drivers/net directory. As part of the move, rename the "vmxnet3" subdirectory, containing the original FreeBSD drivers, from "vmxnet3" to the more standard name "base", to indicate it contains the base drivers used for the implementation. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com> Acked-by: John McNamara <john.mcnamara@intel.com> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2015-05-15 15:56:59 +00:00
* Copyright(c) 2010-2015 Intel Corporation. All rights reserved.
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
* All rights reserved.
remove trailing whitespaces This commit removes trailing whitespace from lines in files. Almost all files are affected, as the BSD license copyright header had trailing whitespace on 4 lines in it [hence the number of files reporting 8 lines changed in the diffstat]. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com> Acked-by: Neil Horman <nhorman@tuxdriver.com> [Thomas: remove spaces before tabs in libs] [Thomas: remove more trailing spaces in non-C files] Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-03 23:42:50 +00:00
*
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
remove trailing whitespaces This commit removes trailing whitespace from lines in files. Almost all files are affected, as the BSD license copyright header had trailing whitespace on 4 lines in it [hence the number of files reporting 8 lines changed in the diffstat]. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com> Acked-by: Neil Horman <nhorman@tuxdriver.com> [Thomas: remove spaces before tabs in libs] [Thomas: remove more trailing spaces in non-C files] Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-03 23:42:50 +00:00
*
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
* * Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* * Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in
* the documentation and/or other materials provided with the
* distribution.
* * Neither the name of Intel Corporation nor the names of its
* contributors may be used to endorse or promote products derived
* from this software without specific prior written permission.
remove trailing whitespaces This commit removes trailing whitespace from lines in files. Almost all files are affected, as the BSD license copyright header had trailing whitespace on 4 lines in it [hence the number of files reporting 8 lines changed in the diffstat]. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com> Acked-by: Neil Horman <nhorman@tuxdriver.com> [Thomas: remove spaces before tabs in libs] [Thomas: remove more trailing spaces in non-C files] Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-03 23:42:50 +00:00
*
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
* A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
* OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
* SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
* LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#include <sys/queue.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <errno.h>
#include <stdint.h>
#include <stdarg.h>
#include <unistd.h>
#include <inttypes.h>
#include <rte_byteorder.h>
#include <rte_common.h>
#include <rte_cycles.h>
#include <rte_log.h>
#include <rte_debug.h>
#include <rte_interrupts.h>
#include <rte_pci.h>
#include <rte_memory.h>
#include <rte_memzone.h>
#include <rte_launch.h>
#include <rte_eal.h>
#include <rte_per_lcore.h>
#include <rte_lcore.h>
#include <rte_atomic.h>
#include <rte_branch_prediction.h>
#include <rte_mempool.h>
#include <rte_malloc.h>
#include <rte_mbuf.h>
#include <rte_ether.h>
#include <rte_ethdev.h>
#include <rte_prefetch.h>
vmxnet3: add Rx check offloads Only supports IPv4 so far. Signed-off-by: Yong Wang <yongwang@vmware.com>
2014-11-05 01:49:41 +00:00
#include <rte_ip.h>
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
#include <rte_udp.h>
#include <rte_tcp.h>
#include <rte_sctp.h>
#include <rte_string_fns.h>
#include <rte_errno.h>
vmxnet3: move to drivers/net/ Move vmxnet3 PMD to drivers/net directory. As part of the move, rename the "vmxnet3" subdirectory, containing the original FreeBSD drivers, from "vmxnet3" to the more standard name "base", to indicate it contains the base drivers used for the implementation. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com> Acked-by: John McNamara <john.mcnamara@intel.com> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2015-05-15 15:56:59 +00:00
#include "base/vmxnet3_defs.h"
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
#include "vmxnet3_ring.h"
#include "vmxnet3_logs.h"
#include "vmxnet3_ethdev.h"
vmxnet3: remove unnecessary inlining No reason to inline large functions. Compiler will decide already based on optimization level. Also register array should be const. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Yong Wang <yongwang@vmware.com>
2015-07-09 18:24:16 +00:00
static const uint32_t rxprod_reg[2] = {VMXNET3_REG_RXPROD, VMXNET3_REG_RXPROD2};
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
vmxnet3: remove unnecessary inlining No reason to inline large functions. Compiler will decide already based on optimization level. Also register array should be const. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Yong Wang <yongwang@vmware.com>
2015-07-09 18:24:16 +00:00
static int vmxnet3_post_rx_bufs(vmxnet3_rx_queue_t*, uint8_t);
static void vmxnet3_tq_tx_complete(vmxnet3_tx_queue_t *);
vmxnet3: fix build with debug Functions for queue dump are not used and cause compilation error if CONFIG_RTE_LIBRTE_VMXNET3_DEBUG_DRIVER is enabled. Fixed by disabling them. Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com> Acked-by: Bruce Richardson <bruce.richardson@intel.com>
2014-07-02 13:11:47 +00:00
#ifdef RTE_LIBRTE_VMXNET3_DEBUG_DRIVER_NOT_USED
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
static void vmxnet3_rxq_dump(struct vmxnet3_rx_queue *);
static void vmxnet3_txq_dump(struct vmxnet3_tx_queue *);
#endif
vmxnet3: fix build with debug Functions for queue dump are not used and cause compilation error if CONFIG_RTE_LIBRTE_VMXNET3_DEBUG_DRIVER is enabled. Fixed by disabling them. Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com> Acked-by: Bruce Richardson <bruce.richardson@intel.com>
2014-07-02 13:11:47 +00:00
#ifdef RTE_LIBRTE_VMXNET3_DEBUG_DRIVER_NOT_USED
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
static void
vmxnet3_rxq_dump(struct vmxnet3_rx_queue *rxq)
{
uint32_t avail = 0;
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
if (rxq == NULL)
return;
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
PMD_RX_LOG(DEBUG,
net/vmxnet3: fix dump format Remove 0x prefix for %p format to prevent double 0x in logs Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com> Acked-by: Yong Wang <yongwang@vmware.com>
2016-06-23 10:52:54 +00:00
"RXQ: cmd0 base : %p cmd1 base : %p comp ring base : %p.",
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
rxq->cmd_ring[0].base, rxq->cmd_ring[1].base, rxq->comp_ring.base);
PMD_RX_LOG(DEBUG,
vmxnet3: fix double spacing of log messages The debug log macro's already include newline, no need to double space the output. Note: other drivers have the same problem Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:13 +00:00
"RXQ: cmd0 basePA : 0x%lx cmd1 basePA : 0x%lx comp ring basePA : 0x%lx.",
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
(unsigned long)rxq->cmd_ring[0].basePA,
(unsigned long)rxq->cmd_ring[1].basePA,
(unsigned long)rxq->comp_ring.basePA);
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
avail = vmxnet3_cmd_ring_desc_avail(&rxq->cmd_ring[0]);
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
PMD_RX_LOG(DEBUG,
vmxnet3: fix double spacing of log messages The debug log macro's already include newline, no need to double space the output. Note: other drivers have the same problem Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:13 +00:00
"RXQ:cmd0: size=%u; free=%u; next2proc=%u; queued=%u",
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
(uint32_t)rxq->cmd_ring[0].size, avail,
rxq->comp_ring.next2proc,
rxq->cmd_ring[0].size - avail);
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
avail = vmxnet3_cmd_ring_desc_avail(&rxq->cmd_ring[1]);
vmxnet3: fix double spacing of log messages The debug log macro's already include newline, no need to double space the output. Note: other drivers have the same problem Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:13 +00:00
PMD_RX_LOG(DEBUG, "RXQ:cmd1 size=%u; free=%u; next2proc=%u; queued=%u",
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
(uint32_t)rxq->cmd_ring[1].size, avail, rxq->comp_ring.next2proc,
rxq->cmd_ring[1].size - avail);
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
}
static void
vmxnet3_txq_dump(struct vmxnet3_tx_queue *txq)
{
uint32_t avail = 0;
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
if (txq == NULL)
return;
net/vmxnet3: fix dump format Remove 0x prefix for %p format to prevent double 0x in logs Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com> Acked-by: Yong Wang <yongwang@vmware.com>
2016-06-23 10:52:54 +00:00
PMD_TX_LOG(DEBUG, "TXQ: cmd base : %p comp ring base : %p data ring base : %p.",
vmxnet3: leverage data ring on Tx path Data_ring is a pre-mapped guest ring buffer that vmxnet3 backend has access to directly without a need for buffer address mapping and unmapping during packet transmission. It is useful in reducing device emulation cost on the tx path. There are some additional cost though on the guest driver for packet copy and overall it's a win. This patch leverages the data_ring for packets with a length less than or equal to the data_ring entry size (128B). For larger packet, we won't use the data_ring as that requires one extra tx descriptor and it's not clear if doing this will be beneficial. Performance results show that this patch significantly boosts vmxnet3 64B tx performance (pkt rate) for l2fwd application on a Ivy Bridge server by >20% at which point we start to hit some bottleneck on the rx side. Signed-off-by: Yong Wang <yongwang@vmware.com>
2014-11-05 01:49:43 +00:00
txq->cmd_ring.base, txq->comp_ring.base, txq->data_ring.base);
PMD_TX_LOG(DEBUG, "TXQ: cmd basePA : 0x%lx comp ring basePA : 0x%lx data ring basePA : 0x%lx.",
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
(unsigned long)txq->cmd_ring.basePA,
vmxnet3: leverage data ring on Tx path Data_ring is a pre-mapped guest ring buffer that vmxnet3 backend has access to directly without a need for buffer address mapping and unmapping during packet transmission. It is useful in reducing device emulation cost on the tx path. There are some additional cost though on the guest driver for packet copy and overall it's a win. This patch leverages the data_ring for packets with a length less than or equal to the data_ring entry size (128B). For larger packet, we won't use the data_ring as that requires one extra tx descriptor and it's not clear if doing this will be beneficial. Performance results show that this patch significantly boosts vmxnet3 64B tx performance (pkt rate) for l2fwd application on a Ivy Bridge server by >20% at which point we start to hit some bottleneck on the rx side. Signed-off-by: Yong Wang <yongwang@vmware.com>
2014-11-05 01:49:43 +00:00
(unsigned long)txq->comp_ring.basePA,
(unsigned long)txq->data_ring.basePA);
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
avail = vmxnet3_cmd_ring_desc_avail(&txq->cmd_ring);
vmxnet3: fix double spacing of log messages The debug log macro's already include newline, no need to double space the output. Note: other drivers have the same problem Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:13 +00:00
PMD_TX_LOG(DEBUG, "TXQ: size=%u; free=%u; next2proc=%u; queued=%u",
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
(uint32_t)txq->cmd_ring.size, avail,
txq->comp_ring.next2proc, txq->cmd_ring.size - avail);
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
}
#endif
vmxnet3: remove unnecessary inlining No reason to inline large functions. Compiler will decide already based on optimization level. Also register array should be const. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Yong Wang <yongwang@vmware.com>
2015-07-09 18:24:16 +00:00
static void
vmxnet3: fix stop/restart This change makes vmxnet3 consistent with other pmds in terms of dev_stop behavior: rather than releasing tx/rx rings, it only resets the ring structure and release the pending mbufs. Verified with various tests (test-pmd and pktgen) over vmxnet3 that dev stop/restart works fine. Signed-off-by: Yong Wang <yongwang@vmware.com>
2014-11-05 01:49:40 +00:00
vmxnet3_cmd_ring_release_mbufs(vmxnet3_cmd_ring_t *ring)
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
{
while (ring->next2comp != ring->next2fill) {
/* No need to worry about tx desc ownership, device is quiesced by now. */
vmxnet3_buf_info_t *buf_info = ring->buf_info + ring->next2comp;
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
if (buf_info->m) {
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
rte_pktmbuf_free(buf_info->m);
buf_info->m = NULL;
buf_info->bufPA = 0;
buf_info->len = 0;
}
vmxnet3_cmd_ring_adv_next2comp(ring);
}
vmxnet3: fix stop/restart This change makes vmxnet3 consistent with other pmds in terms of dev_stop behavior: rather than releasing tx/rx rings, it only resets the ring structure and release the pending mbufs. Verified with various tests (test-pmd and pktgen) over vmxnet3 that dev stop/restart works fine. Signed-off-by: Yong Wang <yongwang@vmware.com>
2014-11-05 01:49:40 +00:00
}
static void
vmxnet3_cmd_ring_release(vmxnet3_cmd_ring_t *ring)
{
vmxnet3_cmd_ring_release_mbufs(ring);
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
rte_free(ring->buf_info);
vmxnet3: fix crash on stop The cmd_ring_release can be called twice if queue has already been released. This cause crash on shutdown. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-08-26 00:41:02 +00:00
ring->buf_info = NULL;
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
}
void
vmxnet3_dev_tx_queue_release(void *txq)
{
vmxnet3_tx_queue_t *tq = txq;
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
vmxnet3: fix stop/restart This change makes vmxnet3 consistent with other pmds in terms of dev_stop behavior: rather than releasing tx/rx rings, it only resets the ring structure and release the pending mbufs. Verified with various tests (test-pmd and pktgen) over vmxnet3 that dev stop/restart works fine. Signed-off-by: Yong Wang <yongwang@vmware.com>
2014-11-05 01:49:40 +00:00
if (tq != NULL) {
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
/* Release the cmd_ring */
vmxnet3_cmd_ring_release(&tq->cmd_ring);
}
}
void
vmxnet3_dev_rx_queue_release(void *rxq)
{
int i;
vmxnet3_rx_queue_t *rq = rxq;
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
vmxnet3: fix stop/restart This change makes vmxnet3 consistent with other pmds in terms of dev_stop behavior: rather than releasing tx/rx rings, it only resets the ring structure and release the pending mbufs. Verified with various tests (test-pmd and pktgen) over vmxnet3 that dev stop/restart works fine. Signed-off-by: Yong Wang <yongwang@vmware.com>
2014-11-05 01:49:40 +00:00
if (rq != NULL) {
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
/* Release both the cmd_rings */
remove trailing whitespaces This commit removes trailing whitespace from lines in files. Almost all files are affected, as the BSD license copyright header had trailing whitespace on 4 lines in it [hence the number of files reporting 8 lines changed in the diffstat]. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com> Acked-by: Neil Horman <nhorman@tuxdriver.com> [Thomas: remove spaces before tabs in libs] [Thomas: remove more trailing spaces in non-C files] Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-03 23:42:50 +00:00
for (i = 0; i < VMXNET3_RX_CMDRING_SIZE; i++)
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
vmxnet3_cmd_ring_release(&rq->cmd_ring[i]);
}
}
vmxnet3: fix stop/restart This change makes vmxnet3 consistent with other pmds in terms of dev_stop behavior: rather than releasing tx/rx rings, it only resets the ring structure and release the pending mbufs. Verified with various tests (test-pmd and pktgen) over vmxnet3 that dev stop/restart works fine. Signed-off-by: Yong Wang <yongwang@vmware.com>
2014-11-05 01:49:40 +00:00
static void
vmxnet3_dev_tx_queue_reset(void *txq)
{
vmxnet3_tx_queue_t *tq = txq;
struct vmxnet3_cmd_ring *ring = &tq->cmd_ring;
struct vmxnet3_comp_ring *comp_ring = &tq->comp_ring;
vmxnet3: leverage data ring on Tx path Data_ring is a pre-mapped guest ring buffer that vmxnet3 backend has access to directly without a need for buffer address mapping and unmapping during packet transmission. It is useful in reducing device emulation cost on the tx path. There are some additional cost though on the guest driver for packet copy and overall it's a win. This patch leverages the data_ring for packets with a length less than or equal to the data_ring entry size (128B). For larger packet, we won't use the data_ring as that requires one extra tx descriptor and it's not clear if doing this will be beneficial. Performance results show that this patch significantly boosts vmxnet3 64B tx performance (pkt rate) for l2fwd application on a Ivy Bridge server by >20% at which point we start to hit some bottleneck on the rx side. Signed-off-by: Yong Wang <yongwang@vmware.com>
2014-11-05 01:49:43 +00:00
struct vmxnet3_data_ring *data_ring = &tq->data_ring;
vmxnet3: fix stop/restart This change makes vmxnet3 consistent with other pmds in terms of dev_stop behavior: rather than releasing tx/rx rings, it only resets the ring structure and release the pending mbufs. Verified with various tests (test-pmd and pktgen) over vmxnet3 that dev stop/restart works fine. Signed-off-by: Yong Wang <yongwang@vmware.com>
2014-11-05 01:49:40 +00:00
int size;
if (tq != NULL) {
/* Release the cmd_ring mbufs */
vmxnet3_cmd_ring_release_mbufs(&tq->cmd_ring);
}
/* Tx vmxnet rings structure initialization*/
ring->next2fill = 0;
ring->next2comp = 0;
ring->gen = VMXNET3_INIT_GEN;
comp_ring->next2proc = 0;
comp_ring->gen = VMXNET3_INIT_GEN;
size = sizeof(struct Vmxnet3_TxDesc) * ring->size;
size += sizeof(struct Vmxnet3_TxCompDesc) * comp_ring->size;
vmxnet3: leverage data ring on Tx path Data_ring is a pre-mapped guest ring buffer that vmxnet3 backend has access to directly without a need for buffer address mapping and unmapping during packet transmission. It is useful in reducing device emulation cost on the tx path. There are some additional cost though on the guest driver for packet copy and overall it's a win. This patch leverages the data_ring for packets with a length less than or equal to the data_ring entry size (128B). For larger packet, we won't use the data_ring as that requires one extra tx descriptor and it's not clear if doing this will be beneficial. Performance results show that this patch significantly boosts vmxnet3 64B tx performance (pkt rate) for l2fwd application on a Ivy Bridge server by >20% at which point we start to hit some bottleneck on the rx side. Signed-off-by: Yong Wang <yongwang@vmware.com>
2014-11-05 01:49:43 +00:00
size += sizeof(struct Vmxnet3_TxDataDesc) * data_ring->size;
vmxnet3: fix stop/restart This change makes vmxnet3 consistent with other pmds in terms of dev_stop behavior: rather than releasing tx/rx rings, it only resets the ring structure and release the pending mbufs. Verified with various tests (test-pmd and pktgen) over vmxnet3 that dev stop/restart works fine. Signed-off-by: Yong Wang <yongwang@vmware.com>
2014-11-05 01:49:40 +00:00
memset(ring->base, 0, size);
}
static void
vmxnet3_dev_rx_queue_reset(void *rxq)
{
int i;
vmxnet3_rx_queue_t *rq = rxq;
struct vmxnet3_cmd_ring *ring0, *ring1;
struct vmxnet3_comp_ring *comp_ring;
int size;
if (rq != NULL) {
/* Release both the cmd_rings mbufs */
for (i = 0; i < VMXNET3_RX_CMDRING_SIZE; i++)
vmxnet3_cmd_ring_release_mbufs(&rq->cmd_ring[i]);
}
ring0 = &rq->cmd_ring[0];
ring1 = &rq->cmd_ring[1];
comp_ring = &rq->comp_ring;
/* Rx vmxnet rings structure initialization */
ring0->next2fill = 0;
ring1->next2fill = 0;
ring0->next2comp = 0;
ring1->next2comp = 0;
ring0->gen = VMXNET3_INIT_GEN;
ring1->gen = VMXNET3_INIT_GEN;
comp_ring->next2proc = 0;
comp_ring->gen = VMXNET3_INIT_GEN;
size = sizeof(struct Vmxnet3_RxDesc) * (ring0->size + ring1->size);
size += sizeof(struct Vmxnet3_RxCompDesc) * comp_ring->size;
memset(ring0->base, 0, size);
}
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
void
vmxnet3_dev_clear_queues(struct rte_eth_dev *dev)
{
unsigned i;
PMD_INIT_FUNC_TRACE();
for (i = 0; i < dev->data->nb_tx_queues; i++) {
struct vmxnet3_tx_queue *txq = dev->data->tx_queues[i];
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
if (txq != NULL) {
txq->stopped = TRUE;
vmxnet3: fix stop/restart This change makes vmxnet3 consistent with other pmds in terms of dev_stop behavior: rather than releasing tx/rx rings, it only resets the ring structure and release the pending mbufs. Verified with various tests (test-pmd and pktgen) over vmxnet3 that dev stop/restart works fine. Signed-off-by: Yong Wang <yongwang@vmware.com>
2014-11-05 01:49:40 +00:00
vmxnet3_dev_tx_queue_reset(txq);
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
}
}
for (i = 0; i < dev->data->nb_rx_queues; i++) {
struct vmxnet3_rx_queue *rxq = dev->data->rx_queues[i];
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
if (rxq != NULL) {
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
rxq->stopped = TRUE;
vmxnet3: fix stop/restart This change makes vmxnet3 consistent with other pmds in terms of dev_stop behavior: rather than releasing tx/rx rings, it only resets the ring structure and release the pending mbufs. Verified with various tests (test-pmd and pktgen) over vmxnet3 that dev stop/restart works fine. Signed-off-by: Yong Wang <yongwang@vmware.com>
2014-11-05 01:49:40 +00:00
vmxnet3_dev_rx_queue_reset(rxq);
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
}
}
}
vmxnet3: support TSO This commit adds vmxnet3 TSO support. Verified with test-pmd (set fwd csum) that both tso and non-tso pkts can be successfully transmitted and all segmentes for a tso pkt are correct on the receiver side. Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-01-13 02:08:36 +00:00
static int
vmxnet3_unmap_pkt(uint16_t eop_idx, vmxnet3_tx_queue_t *txq)
{
int completed = 0;
struct rte_mbuf *mbuf;
/* Release cmd_ring descriptor and free mbuf */
eal: add assert macro for debug The macro RTE_VERIFY always checks a condition. It is optimized with "unlikely" hint. While this macro is well suited for test applications, it is preferred in libraries and examples to enable such check in debug mode. That's why the macro RTE_ASSERT is introduced to call RTE_VERIFY only if built with debug logs enabled. A lot of assert macros were duplicated and enabled with a specific flag. Removing these #ifdef allows to test these code branches more easily and avoid dead code pitfalls. The ENA_ASSERT is kept (in debug mode only) because it has more parameters to log. Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2016-04-22 12:21:26 +00:00
RTE_ASSERT(txq->cmd_ring.base[eop_idx].txd.eop == 1);
vmxnet3: support TSO This commit adds vmxnet3 TSO support. Verified with test-pmd (set fwd csum) that both tso and non-tso pkts can be successfully transmitted and all segmentes for a tso pkt are correct on the receiver side. Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-01-13 02:08:36 +00:00
mbuf = txq->cmd_ring.buf_info[eop_idx].m;
if (mbuf == NULL)
rte_panic("EOP desc does not point to a valid mbuf");
rte_pktmbuf_free(mbuf);
txq->cmd_ring.buf_info[eop_idx].m = NULL;
while (txq->cmd_ring.next2comp != eop_idx) {
/* no out-of-order completion */
eal: add assert macro for debug The macro RTE_VERIFY always checks a condition. It is optimized with "unlikely" hint. While this macro is well suited for test applications, it is preferred in libraries and examples to enable such check in debug mode. That's why the macro RTE_ASSERT is introduced to call RTE_VERIFY only if built with debug logs enabled. A lot of assert macros were duplicated and enabled with a specific flag. Removing these #ifdef allows to test these code branches more easily and avoid dead code pitfalls. The ENA_ASSERT is kept (in debug mode only) because it has more parameters to log. Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2016-04-22 12:21:26 +00:00
RTE_ASSERT(txq->cmd_ring.base[txq->cmd_ring.next2comp].txd.cq == 0);
vmxnet3: support TSO This commit adds vmxnet3 TSO support. Verified with test-pmd (set fwd csum) that both tso and non-tso pkts can be successfully transmitted and all segmentes for a tso pkt are correct on the receiver side. Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-01-13 02:08:36 +00:00
vmxnet3_cmd_ring_adv_next2comp(&txq->cmd_ring);
completed++;
}
/* Mark the txd for which tcd was generated as completed */
vmxnet3_cmd_ring_adv_next2comp(&txq->cmd_ring);
return completed + 1;
}
vmxnet3: remove unnecessary inlining No reason to inline large functions. Compiler will decide already based on optimization level. Also register array should be const. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Yong Wang <yongwang@vmware.com>
2015-07-09 18:24:16 +00:00
static void
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
vmxnet3_tq_tx_complete(vmxnet3_tx_queue_t *txq)
{
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
int completed = 0;
vmxnet3_comp_ring_t *comp_ring = &txq->comp_ring;
struct Vmxnet3_TxCompDesc *tcd = (struct Vmxnet3_TxCompDesc *)
(comp_ring->base + comp_ring->next2proc);
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
while (tcd->gen == comp_ring->gen) {
vmxnet3: support TSO This commit adds vmxnet3 TSO support. Verified with test-pmd (set fwd csum) that both tso and non-tso pkts can be successfully transmitted and all segmentes for a tso pkt are correct on the receiver side. Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-01-13 02:08:36 +00:00
completed += vmxnet3_unmap_pkt(tcd->txdIdx, txq);
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
vmxnet3_comp_ring_adv_next2proc(comp_ring);
tcd = (struct Vmxnet3_TxCompDesc *)(comp_ring->base +
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
comp_ring->next2proc);
}
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
vmxnet3: fix double spacing of log messages The debug log macro's already include newline, no need to double space the output. Note: other drivers have the same problem Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:13 +00:00
PMD_TX_LOG(DEBUG, "Processed %d tx comps & command descs.", completed);
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
}
uint16_t
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
vmxnet3_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts,
uint16_t nb_pkts)
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
{
uint16_t nb_tx;
vmxnet3_tx_queue_t *txq = tx_queue;
vmxnet3: support multi-segment transmit Change sending loop to support multi-segment mbufs. The VMXNET3 api has start-of-packet and end-packet flags, so it is not hard to send multi-segment mbuf's. Also, update descriptor in 32 bit value rather than toggling bitfields which is slower and error prone. Based on code in earlier driver, and the Linux kernel driver. Add a compiler barrier to make sure that update of earlier descriptor are completed prior to update of generation bit on start of packet. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Yong Wang <yongwang@vmware.com>
2015-07-09 18:24:12 +00:00
struct vmxnet3_hw *hw = txq->hw;
vmxnet3: rework Tx Clean up txNumDeferred usage. Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-01-13 02:08:34 +00:00
Vmxnet3_TxQueueCtrl *txq_ctrl = &txq->shared->ctrl;
uint32_t deferred = rte_le_to_cpu_32(txq_ctrl->txNumDeferred);
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
vmxnet3: leverage data ring on Tx path Data_ring is a pre-mapped guest ring buffer that vmxnet3 backend has access to directly without a need for buffer address mapping and unmapping during packet transmission. It is useful in reducing device emulation cost on the tx path. There are some additional cost though on the guest driver for packet copy and overall it's a win. This patch leverages the data_ring for packets with a length less than or equal to the data_ring entry size (128B). For larger packet, we won't use the data_ring as that requires one extra tx descriptor and it's not clear if doing this will be beneficial. Performance results show that this patch significantly boosts vmxnet3 64B tx performance (pkt rate) for l2fwd application on a Ivy Bridge server by >20% at which point we start to hit some bottleneck on the rx side. Signed-off-by: Yong Wang <yongwang@vmware.com>
2014-11-05 01:49:43 +00:00
if (unlikely(txq->stopped)) {
vmxnet3: fix double spacing of log messages The debug log macro's already include newline, no need to double space the output. Note: other drivers have the same problem Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:13 +00:00
PMD_TX_LOG(DEBUG, "Tx queue is stopped.");
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
return 0;
}
/* Free up the comp_descriptors aggressively */
vmxnet3_tq_tx_complete(txq);
nb_tx = 0;
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
while (nb_tx < nb_pkts) {
vmxnet3: support multi-segment transmit Change sending loop to support multi-segment mbufs. The VMXNET3 api has start-of-packet and end-packet flags, so it is not hard to send multi-segment mbuf's. Also, update descriptor in 32 bit value rather than toggling bitfields which is slower and error prone. Based on code in earlier driver, and the Linux kernel driver. Add a compiler barrier to make sure that update of earlier descriptor are completed prior to update of generation bit on start of packet. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Yong Wang <yongwang@vmware.com>
2015-07-09 18:24:12 +00:00
Vmxnet3_GenericDesc *gdesc;
vmxnet3_buf_info_t *tbi;
uint32_t first2fill, avail, dw2;
struct rte_mbuf *txm = tx_pkts[nb_tx];
struct rte_mbuf *m_seg = txm;
vmxnet3: restore Tx data ring support Tx data ring support was removed in a previous change that added multi-seg transmit. This change adds it back. According to the original commit (2e849373), 64B pkt rate with l2fwd improved by ~20% on an Ivy Bridge server at which point we start to hit some bottleneck on the rx side. I also re-did the same test on a different setup (Haswell processor, ~2.3GHz clock rate) on top of the master and still observed ~17% performance gains. Fixes: 7ba5de417e3c ("vmxnet3: support multi-segment transmit") Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-01-13 02:08:33 +00:00
int copy_size = 0;
vmxnet3: support TSO This commit adds vmxnet3 TSO support. Verified with test-pmd (set fwd csum) that both tso and non-tso pkts can be successfully transmitted and all segmentes for a tso pkt are correct on the receiver side. Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-01-13 02:08:36 +00:00
bool tso = (txm->ol_flags & PKT_TX_TCP_SEG) != 0;
/* # of descriptors needed for a packet. */
unsigned count = txm->nb_segs;
avail = vmxnet3_cmd_ring_desc_avail(&txq->cmd_ring);
if (count > avail) {
/* Is command ring full? */
if (unlikely(avail == 0)) {
PMD_TX_LOG(DEBUG, "No free ring descriptors");
txq->stats.tx_ring_full++;
txq->stats.drop_total += (nb_pkts - nb_tx);
break;
}
vmxnet3: support multi-segment transmit Change sending loop to support multi-segment mbufs. The VMXNET3 api has start-of-packet and end-packet flags, so it is not hard to send multi-segment mbuf's. Also, update descriptor in 32 bit value rather than toggling bitfields which is slower and error prone. Based on code in earlier driver, and the Linux kernel driver. Add a compiler barrier to make sure that update of earlier descriptor are completed prior to update of generation bit on start of packet. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Yong Wang <yongwang@vmware.com>
2015-07-09 18:24:12 +00:00
vmxnet3: support TSO This commit adds vmxnet3 TSO support. Verified with test-pmd (set fwd csum) that both tso and non-tso pkts can be successfully transmitted and all segmentes for a tso pkt are correct on the receiver side. Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-01-13 02:08:36 +00:00
/* Command ring is not full but cannot handle the
* multi-segmented packet. Let's try the next packet
* in this case.
*/
PMD_TX_LOG(DEBUG, "Running out of ring descriptors "
"(avail %d needed %d)", avail, count);
txq->stats.drop_total++;
if (tso)
txq->stats.drop_tso++;
vmxnet3: support multi-segment transmit Change sending loop to support multi-segment mbufs. The VMXNET3 api has start-of-packet and end-packet flags, so it is not hard to send multi-segment mbuf's. Also, update descriptor in 32 bit value rather than toggling bitfields which is slower and error prone. Based on code in earlier driver, and the Linux kernel driver. Add a compiler barrier to make sure that update of earlier descriptor are completed prior to update of generation bit on start of packet. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Yong Wang <yongwang@vmware.com>
2015-07-09 18:24:12 +00:00
rte_pktmbuf_free(txm);
vmxnet3: support TSO This commit adds vmxnet3 TSO support. Verified with test-pmd (set fwd csum) that both tso and non-tso pkts can be successfully transmitted and all segmentes for a tso pkt are correct on the receiver side. Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-01-13 02:08:36 +00:00
nb_tx++;
vmxnet3: support multi-segment transmit Change sending loop to support multi-segment mbufs. The VMXNET3 api has start-of-packet and end-packet flags, so it is not hard to send multi-segment mbuf's. Also, update descriptor in 32 bit value rather than toggling bitfields which is slower and error prone. Based on code in earlier driver, and the Linux kernel driver. Add a compiler barrier to make sure that update of earlier descriptor are completed prior to update of generation bit on start of packet. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Yong Wang <yongwang@vmware.com>
2015-07-09 18:24:12 +00:00
continue;
}
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
vmxnet3: support TSO This commit adds vmxnet3 TSO support. Verified with test-pmd (set fwd csum) that both tso and non-tso pkts can be successfully transmitted and all segmentes for a tso pkt are correct on the receiver side. Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-01-13 02:08:36 +00:00
/* Drop non-TSO packet that is excessively fragmented */
if (unlikely(!tso && count > VMXNET3_MAX_TXD_PER_PKT)) {
PMD_TX_LOG(ERR, "Non-TSO packet cannot occupy more than %d tx "
"descriptors. Packet dropped.", VMXNET3_MAX_TXD_PER_PKT);
txq->stats.drop_too_many_segs++;
txq->stats.drop_total++;
rte_pktmbuf_free(txm);
nb_tx++;
continue;
vmxnet3: support multi-segment transmit Change sending loop to support multi-segment mbufs. The VMXNET3 api has start-of-packet and end-packet flags, so it is not hard to send multi-segment mbuf's. Also, update descriptor in 32 bit value rather than toggling bitfields which is slower and error prone. Based on code in earlier driver, and the Linux kernel driver. Add a compiler barrier to make sure that update of earlier descriptor are completed prior to update of generation bit on start of packet. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Yong Wang <yongwang@vmware.com>
2015-07-09 18:24:12 +00:00
}
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
net/vmxnet3: coding style changes Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-08-29 19:18:46 +00:00
if (txm->nb_segs == 1 &&
rte_pktmbuf_pkt_len(txm) <= VMXNET3_HDR_COPY_SIZE) {
vmxnet3: restore Tx data ring support Tx data ring support was removed in a previous change that added multi-seg transmit. This change adds it back. According to the original commit (2e849373), 64B pkt rate with l2fwd improved by ~20% on an Ivy Bridge server at which point we start to hit some bottleneck on the rx side. I also re-did the same test on a different setup (Haswell processor, ~2.3GHz clock rate) on top of the master and still observed ~17% performance gains. Fixes: 7ba5de417e3c ("vmxnet3: support multi-segment transmit") Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-01-13 02:08:33 +00:00
struct Vmxnet3_TxDataDesc *tdd;
tdd = txq->data_ring.base + txq->cmd_ring.next2fill;
copy_size = rte_pktmbuf_pkt_len(txm);
rte_memcpy(tdd->data, rte_pktmbuf_mtod(txm, char *), copy_size);
}
vmxnet3: support multi-segment transmit Change sending loop to support multi-segment mbufs. The VMXNET3 api has start-of-packet and end-packet flags, so it is not hard to send multi-segment mbuf's. Also, update descriptor in 32 bit value rather than toggling bitfields which is slower and error prone. Based on code in earlier driver, and the Linux kernel driver. Add a compiler barrier to make sure that update of earlier descriptor are completed prior to update of generation bit on start of packet. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Yong Wang <yongwang@vmware.com>
2015-07-09 18:24:12 +00:00
/* use the previous gen bit for the SOP desc */
dw2 = (txq->cmd_ring.gen ^ 0x1) << VMXNET3_TXD_GEN_SHIFT;
first2fill = txq->cmd_ring.next2fill;
do {
/* Remember the transmit buffer for cleanup */
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
tbi = txq->cmd_ring.buf_info + txq->cmd_ring.next2fill;
vmxnet3: support multi-segment transmit Change sending loop to support multi-segment mbufs. The VMXNET3 api has start-of-packet and end-packet flags, so it is not hard to send multi-segment mbuf's. Also, update descriptor in 32 bit value rather than toggling bitfields which is slower and error prone. Based on code in earlier driver, and the Linux kernel driver. Add a compiler barrier to make sure that update of earlier descriptor are completed prior to update of generation bit on start of packet. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Yong Wang <yongwang@vmware.com>
2015-07-09 18:24:12 +00:00
/* NB: the following assumes that VMXNET3 maximum
vmxnet3: support TSO This commit adds vmxnet3 TSO support. Verified with test-pmd (set fwd csum) that both tso and non-tso pkts can be successfully transmitted and all segmentes for a tso pkt are correct on the receiver side. Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-01-13 02:08:36 +00:00
* transmit buffer size (16K) is greater than
* maximum size of mbuf segment size.
*/
vmxnet3: support multi-segment transmit Change sending loop to support multi-segment mbufs. The VMXNET3 api has start-of-packet and end-packet flags, so it is not hard to send multi-segment mbuf's. Also, update descriptor in 32 bit value rather than toggling bitfields which is slower and error prone. Based on code in earlier driver, and the Linux kernel driver. Add a compiler barrier to make sure that update of earlier descriptor are completed prior to update of generation bit on start of packet. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Yong Wang <yongwang@vmware.com>
2015-07-09 18:24:12 +00:00
gdesc = txq->cmd_ring.base + txq->cmd_ring.next2fill;
vmxnet3: restore Tx data ring support Tx data ring support was removed in a previous change that added multi-seg transmit. This change adds it back. According to the original commit (2e849373), 64B pkt rate with l2fwd improved by ~20% on an Ivy Bridge server at which point we start to hit some bottleneck on the rx side. I also re-did the same test on a different setup (Haswell processor, ~2.3GHz clock rate) on top of the master and still observed ~17% performance gains. Fixes: 7ba5de417e3c ("vmxnet3: support multi-segment transmit") Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-01-13 02:08:33 +00:00
if (copy_size)
gdesc->txd.addr = rte_cpu_to_le_64(txq->data_ring.basePA +
net/vmxnet3: coding style changes Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-08-29 19:18:46 +00:00
txq->cmd_ring.next2fill *
sizeof(struct Vmxnet3_TxDataDesc));
vmxnet3: restore Tx data ring support Tx data ring support was removed in a previous change that added multi-seg transmit. This change adds it back. According to the original commit (2e849373), 64B pkt rate with l2fwd improved by ~20% on an Ivy Bridge server at which point we start to hit some bottleneck on the rx side. I also re-did the same test on a different setup (Haswell processor, ~2.3GHz clock rate) on top of the master and still observed ~17% performance gains. Fixes: 7ba5de417e3c ("vmxnet3: support multi-segment transmit") Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-01-13 02:08:33 +00:00
else
gdesc->txd.addr = rte_mbuf_data_dma_addr(m_seg);
vmxnet3: support multi-segment transmit Change sending loop to support multi-segment mbufs. The VMXNET3 api has start-of-packet and end-packet flags, so it is not hard to send multi-segment mbuf's. Also, update descriptor in 32 bit value rather than toggling bitfields which is slower and error prone. Based on code in earlier driver, and the Linux kernel driver. Add a compiler barrier to make sure that update of earlier descriptor are completed prior to update of generation bit on start of packet. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Yong Wang <yongwang@vmware.com>
2015-07-09 18:24:12 +00:00
gdesc->dword[2] = dw2 | m_seg->data_len;
gdesc->dword[3] = 0;
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
vmxnet3: support multi-segment transmit Change sending loop to support multi-segment mbufs. The VMXNET3 api has start-of-packet and end-packet flags, so it is not hard to send multi-segment mbuf's. Also, update descriptor in 32 bit value rather than toggling bitfields which is slower and error prone. Based on code in earlier driver, and the Linux kernel driver. Add a compiler barrier to make sure that update of earlier descriptor are completed prior to update of generation bit on start of packet. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Yong Wang <yongwang@vmware.com>
2015-07-09 18:24:12 +00:00
/* move to the next2fill descriptor */
vmxnet3_cmd_ring_adv_next2fill(&txq->cmd_ring);
vmxnet3: add vlan Tx offload Signed-off-by: Yong Wang <yongwang@vmware.com>
2014-11-05 01:49:39 +00:00
vmxnet3: support multi-segment transmit Change sending loop to support multi-segment mbufs. The VMXNET3 api has start-of-packet and end-packet flags, so it is not hard to send multi-segment mbuf's. Also, update descriptor in 32 bit value rather than toggling bitfields which is slower and error prone. Based on code in earlier driver, and the Linux kernel driver. Add a compiler barrier to make sure that update of earlier descriptor are completed prior to update of generation bit on start of packet. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Yong Wang <yongwang@vmware.com>
2015-07-09 18:24:12 +00:00
/* use the right gen for non-SOP desc */
dw2 = txq->cmd_ring.gen << VMXNET3_TXD_GEN_SHIFT;
} while ((m_seg = m_seg->next) != NULL);
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
vmxnet3: support TSO This commit adds vmxnet3 TSO support. Verified with test-pmd (set fwd csum) that both tso and non-tso pkts can be successfully transmitted and all segmentes for a tso pkt are correct on the receiver side. Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-01-13 02:08:36 +00:00
/* set the last buf_info for the pkt */
tbi->m = txm;
vmxnet3: support multi-segment transmit Change sending loop to support multi-segment mbufs. The VMXNET3 api has start-of-packet and end-packet flags, so it is not hard to send multi-segment mbuf's. Also, update descriptor in 32 bit value rather than toggling bitfields which is slower and error prone. Based on code in earlier driver, and the Linux kernel driver. Add a compiler barrier to make sure that update of earlier descriptor are completed prior to update of generation bit on start of packet. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Yong Wang <yongwang@vmware.com>
2015-07-09 18:24:12 +00:00
/* Update the EOP descriptor */
gdesc->dword[3] |= VMXNET3_TXD_EOP | VMXNET3_TXD_CQ;
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
vmxnet3: support multi-segment transmit Change sending loop to support multi-segment mbufs. The VMXNET3 api has start-of-packet and end-packet flags, so it is not hard to send multi-segment mbuf's. Also, update descriptor in 32 bit value rather than toggling bitfields which is slower and error prone. Based on code in earlier driver, and the Linux kernel driver. Add a compiler barrier to make sure that update of earlier descriptor are completed prior to update of generation bit on start of packet. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Yong Wang <yongwang@vmware.com>
2015-07-09 18:24:12 +00:00
/* Add VLAN tag if present */
gdesc = txq->cmd_ring.base + first2fill;
if (txm->ol_flags & PKT_TX_VLAN_PKT) {
gdesc->txd.ti = 1;
gdesc->txd.tci = txm->vlan_tci;
}
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
vmxnet3: support TSO This commit adds vmxnet3 TSO support. Verified with test-pmd (set fwd csum) that both tso and non-tso pkts can be successfully transmitted and all segmentes for a tso pkt are correct on the receiver side. Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-01-13 02:08:36 +00:00
if (tso) {
uint16_t mss = txm->tso_segsz;
eal: add assert macro for debug The macro RTE_VERIFY always checks a condition. It is optimized with "unlikely" hint. While this macro is well suited for test applications, it is preferred in libraries and examples to enable such check in debug mode. That's why the macro RTE_ASSERT is introduced to call RTE_VERIFY only if built with debug logs enabled. A lot of assert macros were duplicated and enabled with a specific flag. Removing these #ifdef allows to test these code branches more easily and avoid dead code pitfalls. The ENA_ASSERT is kept (in debug mode only) because it has more parameters to log. Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2016-04-22 12:21:26 +00:00
RTE_ASSERT(mss > 0);
vmxnet3: support TSO This commit adds vmxnet3 TSO support. Verified with test-pmd (set fwd csum) that both tso and non-tso pkts can be successfully transmitted and all segmentes for a tso pkt are correct on the receiver side. Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-01-13 02:08:36 +00:00
gdesc->txd.hlen = txm->l2_len + txm->l3_len + txm->l4_len;
gdesc->txd.om = VMXNET3_OM_TSO;
gdesc->txd.msscof = mss;
deferred += (rte_pktmbuf_pkt_len(txm) - gdesc->txd.hlen + mss - 1) / mss;
} else if (txm->ol_flags & PKT_TX_L4_MASK) {
vmxnet3: add Tx L4 checksum offload Support TCP/UDP checksum offload. Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-01-13 02:08:35 +00:00
gdesc->txd.om = VMXNET3_OM_CSUM;
gdesc->txd.hlen = txm->l2_len + txm->l3_len;
switch (txm->ol_flags & PKT_TX_L4_MASK) {
case PKT_TX_TCP_CKSUM:
gdesc->txd.msscof = gdesc->txd.hlen + offsetof(struct tcp_hdr, cksum);
break;
case PKT_TX_UDP_CKSUM:
gdesc->txd.msscof = gdesc->txd.hlen + offsetof(struct udp_hdr, dgram_cksum);
break;
default:
PMD_TX_LOG(WARNING, "requested cksum offload not supported %#llx",
txm->ol_flags & PKT_TX_L4_MASK);
abort();
}
vmxnet3: support TSO This commit adds vmxnet3 TSO support. Verified with test-pmd (set fwd csum) that both tso and non-tso pkts can be successfully transmitted and all segmentes for a tso pkt are correct on the receiver side. Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-01-13 02:08:36 +00:00
deferred++;
vmxnet3: add Tx L4 checksum offload Support TCP/UDP checksum offload. Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-01-13 02:08:35 +00:00
} else {
gdesc->txd.hlen = 0;
gdesc->txd.om = VMXNET3_OM_NONE;
gdesc->txd.msscof = 0;
vmxnet3: support TSO This commit adds vmxnet3 TSO support. Verified with test-pmd (set fwd csum) that both tso and non-tso pkts can be successfully transmitted and all segmentes for a tso pkt are correct on the receiver side. Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-01-13 02:08:36 +00:00
deferred++;
vmxnet3: add Tx L4 checksum offload Support TCP/UDP checksum offload. Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-01-13 02:08:35 +00:00
}
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
vmxnet3: support multi-segment transmit Change sending loop to support multi-segment mbufs. The VMXNET3 api has start-of-packet and end-packet flags, so it is not hard to send multi-segment mbuf's. Also, update descriptor in 32 bit value rather than toggling bitfields which is slower and error prone. Based on code in earlier driver, and the Linux kernel driver. Add a compiler barrier to make sure that update of earlier descriptor are completed prior to update of generation bit on start of packet. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Yong Wang <yongwang@vmware.com>
2015-07-09 18:24:12 +00:00
/* flip the GEN bit on the SOP */
rte_compiler_barrier();
gdesc->dword[2] ^= VMXNET3_TXD_GEN;
vmxnet3: support TSO This commit adds vmxnet3 TSO support. Verified with test-pmd (set fwd csum) that both tso and non-tso pkts can be successfully transmitted and all segmentes for a tso pkt are correct on the receiver side. Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-01-13 02:08:36 +00:00
txq_ctrl->txNumDeferred = rte_cpu_to_le_32(deferred);
vmxnet3: support multi-segment transmit Change sending loop to support multi-segment mbufs. The VMXNET3 api has start-of-packet and end-packet flags, so it is not hard to send multi-segment mbuf's. Also, update descriptor in 32 bit value rather than toggling bitfields which is slower and error prone. Based on code in earlier driver, and the Linux kernel driver. Add a compiler barrier to make sure that update of earlier descriptor are completed prior to update of generation bit on start of packet. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Yong Wang <yongwang@vmware.com>
2015-07-09 18:24:12 +00:00
nb_tx++;
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
}
vmxnet3: rework Tx Clean up txNumDeferred usage. Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-01-13 02:08:34 +00:00
PMD_TX_LOG(DEBUG, "vmxnet3 txThreshold: %u", rte_le_to_cpu_32(txq_ctrl->txThreshold));
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
vmxnet3: rework Tx Clean up txNumDeferred usage. Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-01-13 02:08:34 +00:00
if (deferred >= rte_le_to_cpu_32(txq_ctrl->txThreshold)) {
txq_ctrl->txNumDeferred = 0;
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
/* Notify vSwitch that packets are available. */
VMXNET3_WRITE_BAR0_REG(hw, (VMXNET3_REG_TXPROD + txq->queue_id * VMXNET3_REG_ALIGN),
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
txq->cmd_ring.next2fill);
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
}
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
return nb_tx;
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
}
/*
* Allocates mbufs and clusters. Post rx descriptors with buffer details
* so that device can receive packets in those buffers.
net/vmxnet3: coding style changes Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-08-29 19:18:46 +00:00
* Ring layout:
* Among the two rings, 1st ring contains buffers of type 0 and type 1.
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
* bufs_per_pkt is set such that for non-LRO cases all the buffers required
* by a frame will fit in 1st ring (1st buf of type0 and rest of type1).
* 2nd ring contains buffers of type 1 alone. Second ring mostly be used
* only for LRO.
*/
vmxnet3: remove unnecessary inlining No reason to inline large functions. Compiler will decide already based on optimization level. Also register array should be const. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Yong Wang <yongwang@vmware.com>
2015-07-09 18:24:16 +00:00
static int
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
vmxnet3_post_rx_bufs(vmxnet3_rx_queue_t *rxq, uint8_t ring_id)
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
{
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
int err = 0;
uint32_t i = 0, val = 0;
struct vmxnet3_cmd_ring *ring = &rxq->cmd_ring[ring_id];
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
vmxnet3: improve Rx performance This patch includes two small performance optimizations on the rx path: (1) It adds unlikely hints on various infrequent error paths to the compiler to make branch prediction more efficient. (2) It also moves a constant assignment out of the pkt polling loop. This saves one branching per packet. Performance evaluation configs: - On the DPDK-side, it's running some l3 forwarding app inside a VM on ESXi with one core assigned for polling. - On the client side, pktgen/dpdk is used to generate 64B tcp packets at line rate (14.8M PPS). Performance results on a Nehalem box (4cores@2.8GHzx2) shown below. CPU usage is collected factoring out the idle loop cost. - Before the patch, ~900K PPS with 65% CPU of a core used for DPDK. - After the patch, only 45% of a core used, while maintaining the same packet rate. Signed-off-by: Yong Wang <yongwang@vmware.com>
2014-11-05 01:49:42 +00:00
if (ring_id == 0) {
/* Usually: One HEAD type buf per packet
* val = (ring->next2fill % rxq->hw->bufs_per_pkt) ?
* VMXNET3_RXD_BTYPE_BODY : VMXNET3_RXD_BTYPE_HEAD;
*/
/* We use single packet buffer so all heads here */
val = VMXNET3_RXD_BTYPE_HEAD;
} else {
/* All BODY type buffers for 2nd ring */
val = VMXNET3_RXD_BTYPE_BODY;
}
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
while (vmxnet3_cmd_ring_desc_avail(ring) > 0) {
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
struct Vmxnet3_RxDesc *rxd;
struct rte_mbuf *mbuf;
vmxnet3_buf_info_t *buf_info = &ring->buf_info[ring->next2fill];
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
rxd = (struct Vmxnet3_RxDesc *)(ring->base + ring->next2fill);
/* Allocate blank mbuf for the current Rx Descriptor */
mbuf: add raw allocation function Many drivers provide their own implementation of rte_mbuf_raw_alloc(), duplicating the code. Introduce a new public function in rte_mbuf to allocate a raw mbuf (uninitialized). Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
2016-05-11 14:43:46 +00:00
mbuf = rte_mbuf_raw_alloc(rxq->mp);
vmxnet3: improve Rx performance This patch includes two small performance optimizations on the rx path: (1) It adds unlikely hints on various infrequent error paths to the compiler to make branch prediction more efficient. (2) It also moves a constant assignment out of the pkt polling loop. This saves one branching per packet. Performance evaluation configs: - On the DPDK-side, it's running some l3 forwarding app inside a VM on ESXi with one core assigned for polling. - On the client side, pktgen/dpdk is used to generate 64B tcp packets at line rate (14.8M PPS). Performance results on a Nehalem box (4cores@2.8GHzx2) shown below. CPU usage is collected factoring out the idle loop cost. - Before the patch, ~900K PPS with 65% CPU of a core used for DPDK. - After the patch, only 45% of a core used, while maintaining the same packet rate. Signed-off-by: Yong Wang <yongwang@vmware.com>
2014-11-05 01:49:42 +00:00
if (unlikely(mbuf == NULL)) {
vmxnet3: remove redundant function names in log Signed-off-by: Huawei Xie <huawei.xie@intel.com> Acked-by: Yuanhan Liu <yuanhan.liu@linux.intel.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-02-10 16:12:39 +00:00
PMD_RX_LOG(ERR, "Error allocating mbuf");
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
rxq->stats.rx_buf_alloc_failure++;
err = ENOMEM;
break;
}
/*
* Load mbuf pointer into buf_info[ring_size]
* buf_info structure is equivalent to cookie for virtio-virtqueue
*/
buf_info->m = mbuf;
buf_info->len = (uint16_t)(mbuf->buf_len -
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
RTE_PKTMBUF_HEADROOM);
net/vmxnet3: coding style changes Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-08-29 19:18:46 +00:00
buf_info->bufPA = rte_mbuf_data_dma_addr_default(mbuf);
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
/* Load Rx Descriptor with the buffer's GPA */
rxd->addr = buf_info->bufPA;
/* After this point rxd->addr MUST not be NULL */
rxd->btype = val;
rxd->len = buf_info->len;
/* Flip gen bit at the end to change ownership */
rxd->gen = ring->gen;
vmxnet3_cmd_ring_adv_next2fill(ring);
i++;
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
}
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
/* Return error only if no buffers are posted at present */
if (vmxnet3_cmd_ring_desc_avail(ring) >= (ring->size - 1))
return -err;
else
return i;
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
}
vmxnet3: support RSS and refactor Rx offload Refactor the logic to compute receive offload flags to a simpler function. And add support for putting RSS flow hash into packet. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Signed-off-by: Bill Hong <bhong@brocade.com> Acked-by: Yong Wang <yongwang@vmware.com>
2015-07-09 18:24:14 +00:00
/* Receive side checksum and other offloads */
static void
vmxnet3_rx_offload(const Vmxnet3_RxCompDesc *rcd, struct rte_mbuf *rxm)
{
/* Check for RSS */
if (rcd->rssType != VMXNET3_RCD_RSS_TYPE_NONE) {
rxm->ol_flags |= PKT_RX_RSS_HASH;
rxm->hash.rss = rcd->rssHash;
}
/* Check packet type, checksum errors, etc. Only support IPv4 for now. */
if (rcd->v4) {
struct ether_hdr *eth = rte_pktmbuf_mtod(rxm, struct ether_hdr *);
struct ipv4_hdr *ip = (struct ipv4_hdr *)(eth + 1);
if (((ip->version_ihl & 0xf) << 2) > (int)sizeof(struct ipv4_hdr))
vmxnet3: replace some offload flags with packet type To unify packet types among all PMDs, bit masks of packet type for 'ol_flags' are replaced by unified packet type. To avoid breaking ABI compatibility, all the changes would be enabled by RTE_NEXT_ABI. Signed-off-by: Helin Zhang <helin.zhang@intel.com> Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
2015-07-09 16:31:29 +00:00
rxm->packet_type = RTE_PTYPE_L3_IPV4_EXT;
vmxnet3: support RSS and refactor Rx offload Refactor the logic to compute receive offload flags to a simpler function. And add support for putting RSS flow hash into packet. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Signed-off-by: Bill Hong <bhong@brocade.com> Acked-by: Yong Wang <yongwang@vmware.com>
2015-07-09 18:24:14 +00:00
else
vmxnet3: replace some offload flags with packet type To unify packet types among all PMDs, bit masks of packet type for 'ol_flags' are replaced by unified packet type. To avoid breaking ABI compatibility, all the changes would be enabled by RTE_NEXT_ABI. Signed-off-by: Helin Zhang <helin.zhang@intel.com> Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
2015-07-09 16:31:29 +00:00
rxm->packet_type = RTE_PTYPE_L3_IPV4;
vmxnet3: support RSS and refactor Rx offload Refactor the logic to compute receive offload flags to a simpler function. And add support for putting RSS flow hash into packet. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Signed-off-by: Bill Hong <bhong@brocade.com> Acked-by: Yong Wang <yongwang@vmware.com>
2015-07-09 18:24:14 +00:00
if (!rcd->cnc) {
if (!rcd->ipc)
rxm->ol_flags |= PKT_RX_IP_CKSUM_BAD;
if ((rcd->tcp || rcd->udp) && !rcd->tuc)
rxm->ol_flags |= PKT_RX_L4_CKSUM_BAD;
}
}
}
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
/*
* Process the Rx Completion Ring of given vmxnet3_rx_queue
* for nb_pkts burst and return the number of packets received
*/
uint16_t
vmxnet3_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
{
uint16_t nb_rx;
uint32_t nb_rxd, idx;
uint8_t ring_idx;
vmxnet3_rx_queue_t *rxq;
Vmxnet3_RxCompDesc *rcd;
vmxnet3_buf_info_t *rbi;
Vmxnet3_RxDesc *rxd;
struct rte_mbuf *rxm = NULL;
struct vmxnet3_hw *hw;
nb_rx = 0;
ring_idx = 0;
nb_rxd = 0;
idx = 0;
rxq = rx_queue;
hw = rxq->hw;
rcd = &rxq->comp_ring.base[rxq->comp_ring.next2proc].rcd;
vmxnet3: improve Rx performance This patch includes two small performance optimizations on the rx path: (1) It adds unlikely hints on various infrequent error paths to the compiler to make branch prediction more efficient. (2) It also moves a constant assignment out of the pkt polling loop. This saves one branching per packet. Performance evaluation configs: - On the DPDK-side, it's running some l3 forwarding app inside a VM on ESXi with one core assigned for polling. - On the client side, pktgen/dpdk is used to generate 64B tcp packets at line rate (14.8M PPS). Performance results on a Nehalem box (4cores@2.8GHzx2) shown below. CPU usage is collected factoring out the idle loop cost. - Before the patch, ~900K PPS with 65% CPU of a core used for DPDK. - After the patch, only 45% of a core used, while maintaining the same packet rate. Signed-off-by: Yong Wang <yongwang@vmware.com>
2014-11-05 01:49:42 +00:00
if (unlikely(rxq->stopped)) {
vmxnet3: fix double spacing of log messages The debug log macro's already include newline, no need to double space the output. Note: other drivers have the same problem Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:13 +00:00
PMD_RX_LOG(DEBUG, "Rx queue is stopped.");
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
return 0;
}
while (rcd->gen == rxq->comp_ring.gen) {
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
if (nb_rx >= nb_pkts)
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
break;
vmxnet3: improve Rx performance This patch includes two small performance optimizations on the rx path: (1) It adds unlikely hints on various infrequent error paths to the compiler to make branch prediction more efficient. (2) It also moves a constant assignment out of the pkt polling loop. This saves one branching per packet. Performance evaluation configs: - On the DPDK-side, it's running some l3 forwarding app inside a VM on ESXi with one core assigned for polling. - On the client side, pktgen/dpdk is used to generate 64B tcp packets at line rate (14.8M PPS). Performance results on a Nehalem box (4cores@2.8GHzx2) shown below. CPU usage is collected factoring out the idle loop cost. - Before the patch, ~900K PPS with 65% CPU of a core used for DPDK. - After the patch, only 45% of a core used, while maintaining the same packet rate. Signed-off-by: Yong Wang <yongwang@vmware.com>
2014-11-05 01:49:42 +00:00
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
idx = rcd->rxdIdx;
ring_idx = (uint8_t)((rcd->rqID == rxq->qid1) ? 0 : 1);
rxd = (Vmxnet3_RxDesc *)rxq->cmd_ring[ring_idx].base + idx;
eal: add assert macro for debug The macro RTE_VERIFY always checks a condition. It is optimized with "unlikely" hint. While this macro is well suited for test applications, it is preferred in libraries and examples to enable such check in debug mode. That's why the macro RTE_ASSERT is introduced to call RTE_VERIFY only if built with debug logs enabled. A lot of assert macros were duplicated and enabled with a specific flag. Removing these #ifdef allows to test these code branches more easily and avoid dead code pitfalls. The ENA_ASSERT is kept (in debug mode only) because it has more parameters to log. Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2016-04-22 12:21:26 +00:00
RTE_SET_USED(rxd); /* used only for assert when enabled */
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
rbi = rxq->cmd_ring[ring_idx].buf_info + idx;
vmxnet3: improve Rx performance This patch includes two small performance optimizations on the rx path: (1) It adds unlikely hints on various infrequent error paths to the compiler to make branch prediction more efficient. (2) It also moves a constant assignment out of the pkt polling loop. This saves one branching per packet. Performance evaluation configs: - On the DPDK-side, it's running some l3 forwarding app inside a VM on ESXi with one core assigned for polling. - On the client side, pktgen/dpdk is used to generate 64B tcp packets at line rate (14.8M PPS). Performance results on a Nehalem box (4cores@2.8GHzx2) shown below. CPU usage is collected factoring out the idle loop cost. - Before the patch, ~900K PPS with 65% CPU of a core used for DPDK. - After the patch, only 45% of a core used, while maintaining the same packet rate. Signed-off-by: Yong Wang <yongwang@vmware.com>
2014-11-05 01:49:42 +00:00
PMD_RX_LOG(DEBUG, "rxd idx: %d ring idx: %d.", idx, ring_idx);
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
eal: add assert macro for debug The macro RTE_VERIFY always checks a condition. It is optimized with "unlikely" hint. While this macro is well suited for test applications, it is preferred in libraries and examples to enable such check in debug mode. That's why the macro RTE_ASSERT is introduced to call RTE_VERIFY only if built with debug logs enabled. A lot of assert macros were duplicated and enabled with a specific flag. Removing these #ifdef allows to test these code branches more easily and avoid dead code pitfalls. The ENA_ASSERT is kept (in debug mode only) because it has more parameters to log. Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2016-04-22 12:21:26 +00:00
RTE_ASSERT(rcd->len <= rxd->len);
RTE_ASSERT(rbi->m);
vmxnet3: get rid of debug ifdefs By defining macro as a stub it is possible to get rid of #ifdef's in the actual code. Always evaluate the argument (even in the stub) so that there are no extra unused variable errors. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Yong Wang <yongwang@vmware.com>
2015-07-09 18:24:15 +00:00
vmxnet3: improve Rx performance This patch includes two small performance optimizations on the rx path: (1) It adds unlikely hints on various infrequent error paths to the compiler to make branch prediction more efficient. (2) It also moves a constant assignment out of the pkt polling loop. This saves one branching per packet. Performance evaluation configs: - On the DPDK-side, it's running some l3 forwarding app inside a VM on ESXi with one core assigned for polling. - On the client side, pktgen/dpdk is used to generate 64B tcp packets at line rate (14.8M PPS). Performance results on a Nehalem box (4cores@2.8GHzx2) shown below. CPU usage is collected factoring out the idle loop cost. - Before the patch, ~900K PPS with 65% CPU of a core used for DPDK. - After the patch, only 45% of a core used, while maintaining the same packet rate. Signed-off-by: Yong Wang <yongwang@vmware.com>
2014-11-05 01:49:42 +00:00
/* Get the packet buffer pointer from buf_info */
rxm = rbi->m;
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
vmxnet3: improve Rx performance This patch includes two small performance optimizations on the rx path: (1) It adds unlikely hints on various infrequent error paths to the compiler to make branch prediction more efficient. (2) It also moves a constant assignment out of the pkt polling loop. This saves one branching per packet. Performance evaluation configs: - On the DPDK-side, it's running some l3 forwarding app inside a VM on ESXi with one core assigned for polling. - On the client side, pktgen/dpdk is used to generate 64B tcp packets at line rate (14.8M PPS). Performance results on a Nehalem box (4cores@2.8GHzx2) shown below. CPU usage is collected factoring out the idle loop cost. - Before the patch, ~900K PPS with 65% CPU of a core used for DPDK. - After the patch, only 45% of a core used, while maintaining the same packet rate. Signed-off-by: Yong Wang <yongwang@vmware.com>
2014-11-05 01:49:42 +00:00
/* Clear descriptor associated buf_info to be reused */
rbi->m = NULL;
rbi->bufPA = 0;
/* Update the index that we received a packet */
rxq->cmd_ring[ring_idx].next2comp = idx;
/* For RCD with EOP set, check if there is frame error */
vmxnet3: support jumbo frames Add support for linking multi-segment buffers together to handle Jumbo packets. The vmxnet3 API supports having header and body buffer types. What this patch does is fill the primary ring completely with header buffers and the secondary ring with body buffers. This allows for non-jumbo frames to only use one mbuf (from primary ring); and jumbo frames will have first mbuf from primary ring and following mbufs from other ring. This could be optimized in future if the DPDK had API to supply different sized mbufs (two pools) into driver. Signed-off-by: Stephen Hemminger <shemming@brocade.com> Acked-by: Remy Horton <remy.horton@intel.com> Release note addition: Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2016-03-04 18:08:00 +00:00
if (unlikely(rcd->eop && rcd->err)) {
vmxnet3: improve Rx performance This patch includes two small performance optimizations on the rx path: (1) It adds unlikely hints on various infrequent error paths to the compiler to make branch prediction more efficient. (2) It also moves a constant assignment out of the pkt polling loop. This saves one branching per packet. Performance evaluation configs: - On the DPDK-side, it's running some l3 forwarding app inside a VM on ESXi with one core assigned for polling. - On the client side, pktgen/dpdk is used to generate 64B tcp packets at line rate (14.8M PPS). Performance results on a Nehalem box (4cores@2.8GHzx2) shown below. CPU usage is collected factoring out the idle loop cost. - Before the patch, ~900K PPS with 65% CPU of a core used for DPDK. - After the patch, only 45% of a core used, while maintaining the same packet rate. Signed-off-by: Yong Wang <yongwang@vmware.com>
2014-11-05 01:49:42 +00:00
rxq->stats.drop_total++;
rxq->stats.drop_err++;
if (!rcd->fcs) {
rxq->stats.drop_fcs++;
PMD_RX_LOG(ERR, "Recv packet dropped due to frame err.");
vmxnet3: fix vlan Rx stripping Shouldn't reset vlan_tci to 0 if a valid VLAN tag is stripped. Signed-off-by: Yong Wang <yongwang@vmware.com>
2014-11-05 01:49:38 +00:00
}
vmxnet3: improve Rx performance This patch includes two small performance optimizations on the rx path: (1) It adds unlikely hints on various infrequent error paths to the compiler to make branch prediction more efficient. (2) It also moves a constant assignment out of the pkt polling loop. This saves one branching per packet. Performance evaluation configs: - On the DPDK-side, it's running some l3 forwarding app inside a VM on ESXi with one core assigned for polling. - On the client side, pktgen/dpdk is used to generate 64B tcp packets at line rate (14.8M PPS). Performance results on a Nehalem box (4cores@2.8GHzx2) shown below. CPU usage is collected factoring out the idle loop cost. - Before the patch, ~900K PPS with 65% CPU of a core used for DPDK. - After the patch, only 45% of a core used, while maintaining the same packet rate. Signed-off-by: Yong Wang <yongwang@vmware.com>
2014-11-05 01:49:42 +00:00
PMD_RX_LOG(ERR, "Error in received packet rcd#:%d rxd:%d",
(int)(rcd - (struct Vmxnet3_RxCompDesc *)
rxq->comp_ring.base), rcd->rxdIdx);
rte_pktmbuf_free_seg(rxm);
goto rcd_done;
}
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
vmxnet3: improve Rx performance This patch includes two small performance optimizations on the rx path: (1) It adds unlikely hints on various infrequent error paths to the compiler to make branch prediction more efficient. (2) It also moves a constant assignment out of the pkt polling loop. This saves one branching per packet. Performance evaluation configs: - On the DPDK-side, it's running some l3 forwarding app inside a VM on ESXi with one core assigned for polling. - On the client side, pktgen/dpdk is used to generate 64B tcp packets at line rate (14.8M PPS). Performance results on a Nehalem box (4cores@2.8GHzx2) shown below. CPU usage is collected factoring out the idle loop cost. - Before the patch, ~900K PPS with 65% CPU of a core used for DPDK. - After the patch, only 45% of a core used, while maintaining the same packet rate. Signed-off-by: Yong Wang <yongwang@vmware.com>
2014-11-05 01:49:42 +00:00
/* Initialize newly received packet buffer */
rxm->port = rxq->port_id;
rxm->nb_segs = 1;
rxm->next = NULL;
rxm->pkt_len = (uint16_t)rcd->len;
rxm->data_len = (uint16_t)rcd->len;
rxm->data_off = RTE_PKTMBUF_HEADROOM;
vmxnet3: support RSS and refactor Rx offload Refactor the logic to compute receive offload flags to a simpler function. And add support for putting RSS flow hash into packet. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Signed-off-by: Bill Hong <bhong@brocade.com> Acked-by: Yong Wang <yongwang@vmware.com>
2015-07-09 18:24:14 +00:00
rxm->ol_flags = 0;
rxm->vlan_tci = 0;
vmxnet3: improve Rx performance This patch includes two small performance optimizations on the rx path: (1) It adds unlikely hints on various infrequent error paths to the compiler to make branch prediction more efficient. (2) It also moves a constant assignment out of the pkt polling loop. This saves one branching per packet. Performance evaluation configs: - On the DPDK-side, it's running some l3 forwarding app inside a VM on ESXi with one core assigned for polling. - On the client side, pktgen/dpdk is used to generate 64B tcp packets at line rate (14.8M PPS). Performance results on a Nehalem box (4cores@2.8GHzx2) shown below. CPU usage is collected factoring out the idle loop cost. - Before the patch, ~900K PPS with 65% CPU of a core used for DPDK. - After the patch, only 45% of a core used, while maintaining the same packet rate. Signed-off-by: Yong Wang <yongwang@vmware.com>
2014-11-05 01:49:42 +00:00
vmxnet3: support jumbo frames Add support for linking multi-segment buffers together to handle Jumbo packets. The vmxnet3 API supports having header and body buffer types. What this patch does is fill the primary ring completely with header buffers and the secondary ring with body buffers. This allows for non-jumbo frames to only use one mbuf (from primary ring); and jumbo frames will have first mbuf from primary ring and following mbufs from other ring. This could be optimized in future if the DPDK had API to supply different sized mbufs (two pools) into driver. Signed-off-by: Stephen Hemminger <shemming@brocade.com> Acked-by: Remy Horton <remy.horton@intel.com> Release note addition: Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2016-03-04 18:08:00 +00:00
/*
* If this is the first buffer of the received packet,
* set the pointer to the first mbuf of the packet
* Otherwise, update the total length and the number of segments
* of the current scattered packet, and update the pointer to
* the last mbuf of the current packet.
*/
if (rcd->sop) {
eal: add assert macro for debug The macro RTE_VERIFY always checks a condition. It is optimized with "unlikely" hint. While this macro is well suited for test applications, it is preferred in libraries and examples to enable such check in debug mode. That's why the macro RTE_ASSERT is introduced to call RTE_VERIFY only if built with debug logs enabled. A lot of assert macros were duplicated and enabled with a specific flag. Removing these #ifdef allows to test these code branches more easily and avoid dead code pitfalls. The ENA_ASSERT is kept (in debug mode only) because it has more parameters to log. Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2016-04-22 12:21:26 +00:00
RTE_ASSERT(rxd->btype == VMXNET3_RXD_BTYPE_HEAD);
vmxnet3: support jumbo frames Add support for linking multi-segment buffers together to handle Jumbo packets. The vmxnet3 API supports having header and body buffer types. What this patch does is fill the primary ring completely with header buffers and the secondary ring with body buffers. This allows for non-jumbo frames to only use one mbuf (from primary ring); and jumbo frames will have first mbuf from primary ring and following mbufs from other ring. This could be optimized in future if the DPDK had API to supply different sized mbufs (two pools) into driver. Signed-off-by: Stephen Hemminger <shemming@brocade.com> Acked-by: Remy Horton <remy.horton@intel.com> Release note addition: Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2016-03-04 18:08:00 +00:00
if (unlikely(rcd->len == 0)) {
eal: add assert macro for debug The macro RTE_VERIFY always checks a condition. It is optimized with "unlikely" hint. While this macro is well suited for test applications, it is preferred in libraries and examples to enable such check in debug mode. That's why the macro RTE_ASSERT is introduced to call RTE_VERIFY only if built with debug logs enabled. A lot of assert macros were duplicated and enabled with a specific flag. Removing these #ifdef allows to test these code branches more easily and avoid dead code pitfalls. The ENA_ASSERT is kept (in debug mode only) because it has more parameters to log. Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2016-04-22 12:21:26 +00:00
RTE_ASSERT(rcd->eop);
vmxnet3: support jumbo frames Add support for linking multi-segment buffers together to handle Jumbo packets. The vmxnet3 API supports having header and body buffer types. What this patch does is fill the primary ring completely with header buffers and the secondary ring with body buffers. This allows for non-jumbo frames to only use one mbuf (from primary ring); and jumbo frames will have first mbuf from primary ring and following mbufs from other ring. This could be optimized in future if the DPDK had API to supply different sized mbufs (two pools) into driver. Signed-off-by: Stephen Hemminger <shemming@brocade.com> Acked-by: Remy Horton <remy.horton@intel.com> Release note addition: Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2016-03-04 18:08:00 +00:00
PMD_RX_LOG(DEBUG,
"Rx buf was skipped. rxring[%d][%d])",
ring_idx, idx);
rte_pktmbuf_free_seg(rxm);
goto rcd_done;
}
rxq->start_seg = rxm;
vmxnet3_rx_offload(rcd, rxm);
} else {
struct rte_mbuf *start = rxq->start_seg;
eal: add assert macro for debug The macro RTE_VERIFY always checks a condition. It is optimized with "unlikely" hint. While this macro is well suited for test applications, it is preferred in libraries and examples to enable such check in debug mode. That's why the macro RTE_ASSERT is introduced to call RTE_VERIFY only if built with debug logs enabled. A lot of assert macros were duplicated and enabled with a specific flag. Removing these #ifdef allows to test these code branches more easily and avoid dead code pitfalls. The ENA_ASSERT is kept (in debug mode only) because it has more parameters to log. Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2016-04-22 12:21:26 +00:00
RTE_ASSERT(rxd->btype == VMXNET3_RXD_BTYPE_BODY);
vmxnet3: support jumbo frames Add support for linking multi-segment buffers together to handle Jumbo packets. The vmxnet3 API supports having header and body buffer types. What this patch does is fill the primary ring completely with header buffers and the secondary ring with body buffers. This allows for non-jumbo frames to only use one mbuf (from primary ring); and jumbo frames will have first mbuf from primary ring and following mbufs from other ring. This could be optimized in future if the DPDK had API to supply different sized mbufs (two pools) into driver. Signed-off-by: Stephen Hemminger <shemming@brocade.com> Acked-by: Remy Horton <remy.horton@intel.com> Release note addition: Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2016-03-04 18:08:00 +00:00
start->pkt_len += rxm->data_len;
start->nb_segs++;
rxq->last_seg->next = rxm;
}
rxq->last_seg = rxm;
if (rcd->eop) {
net/vmxnet3: fix VLAN tag placed in wrong mbuf in chain The VLAN tag information should be stored in the first mbuf of a chain of buffers, not in the last one. Fixes: 9fd5e98b62e4 ("vmxnet3: support RSS and refactor Rx offload") Signed-off-by: John Guzik <john@shieldxnetworks.com> Acked-by: Yong Wang <yongwang@vmware.com>
2016-04-12 23:08:04 +00:00
struct rte_mbuf *start = rxq->start_seg;
/* Check for hardware stripped VLAN tag */
if (rcd->ts) {
start->ol_flags |= (PKT_RX_VLAN_PKT | PKT_RX_VLAN_STRIPPED);
start->vlan_tci = rte_le_to_cpu_16((uint16_t)rcd->tci);
}
rx_pkts[nb_rx++] = start;
vmxnet3: support jumbo frames Add support for linking multi-segment buffers together to handle Jumbo packets. The vmxnet3 API supports having header and body buffer types. What this patch does is fill the primary ring completely with header buffers and the secondary ring with body buffers. This allows for non-jumbo frames to only use one mbuf (from primary ring); and jumbo frames will have first mbuf from primary ring and following mbufs from other ring. This could be optimized in future if the DPDK had API to supply different sized mbufs (two pools) into driver. Signed-off-by: Stephen Hemminger <shemming@brocade.com> Acked-by: Remy Horton <remy.horton@intel.com> Release note addition: Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2016-03-04 18:08:00 +00:00
rxq->start_seg = NULL;
}
vmxnet3: add Rx check offloads Only supports IPv4 so far. Signed-off-by: Yong Wang <yongwang@vmware.com>
2014-11-05 01:49:41 +00:00
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
rcd_done:
vmxnet3: improve Rx performance This patch includes two small performance optimizations on the rx path: (1) It adds unlikely hints on various infrequent error paths to the compiler to make branch prediction more efficient. (2) It also moves a constant assignment out of the pkt polling loop. This saves one branching per packet. Performance evaluation configs: - On the DPDK-side, it's running some l3 forwarding app inside a VM on ESXi with one core assigned for polling. - On the client side, pktgen/dpdk is used to generate 64B tcp packets at line rate (14.8M PPS). Performance results on a Nehalem box (4cores@2.8GHzx2) shown below. CPU usage is collected factoring out the idle loop cost. - Before the patch, ~900K PPS with 65% CPU of a core used for DPDK. - After the patch, only 45% of a core used, while maintaining the same packet rate. Signed-off-by: Yong Wang <yongwang@vmware.com>
2014-11-05 01:49:42 +00:00
rxq->cmd_ring[ring_idx].next2comp = idx;
net/vmxnet3: coding style changes Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-08-29 19:18:46 +00:00
VMXNET3_INC_RING_IDX_ONLY(rxq->cmd_ring[ring_idx].next2comp,
rxq->cmd_ring[ring_idx].size);
vmxnet3: improve Rx performance This patch includes two small performance optimizations on the rx path: (1) It adds unlikely hints on various infrequent error paths to the compiler to make branch prediction more efficient. (2) It also moves a constant assignment out of the pkt polling loop. This saves one branching per packet. Performance evaluation configs: - On the DPDK-side, it's running some l3 forwarding app inside a VM on ESXi with one core assigned for polling. - On the client side, pktgen/dpdk is used to generate 64B tcp packets at line rate (14.8M PPS). Performance results on a Nehalem box (4cores@2.8GHzx2) shown below. CPU usage is collected factoring out the idle loop cost. - Before the patch, ~900K PPS with 65% CPU of a core used for DPDK. - After the patch, only 45% of a core used, while maintaining the same packet rate. Signed-off-by: Yong Wang <yongwang@vmware.com>
2014-11-05 01:49:42 +00:00
/* It's time to allocate some new buf and renew descriptors */
vmxnet3_post_rx_bufs(rxq, ring_idx);
if (unlikely(rxq->shared->ctrl.updateRxProd)) {
VMXNET3_WRITE_BAR0_REG(hw, rxprod_reg[ring_idx] + (rxq->queue_id * VMXNET3_REG_ALIGN),
rxq->cmd_ring[ring_idx].next2fill);
}
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
vmxnet3: improve Rx performance This patch includes two small performance optimizations on the rx path: (1) It adds unlikely hints on various infrequent error paths to the compiler to make branch prediction more efficient. (2) It also moves a constant assignment out of the pkt polling loop. This saves one branching per packet. Performance evaluation configs: - On the DPDK-side, it's running some l3 forwarding app inside a VM on ESXi with one core assigned for polling. - On the client side, pktgen/dpdk is used to generate 64B tcp packets at line rate (14.8M PPS). Performance results on a Nehalem box (4cores@2.8GHzx2) shown below. CPU usage is collected factoring out the idle loop cost. - Before the patch, ~900K PPS with 65% CPU of a core used for DPDK. - After the patch, only 45% of a core used, while maintaining the same packet rate. Signed-off-by: Yong Wang <yongwang@vmware.com>
2014-11-05 01:49:42 +00:00
/* Advance to the next descriptor in comp_ring */
vmxnet3_comp_ring_adv_next2proc(&rxq->comp_ring);
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
vmxnet3: improve Rx performance This patch includes two small performance optimizations on the rx path: (1) It adds unlikely hints on various infrequent error paths to the compiler to make branch prediction more efficient. (2) It also moves a constant assignment out of the pkt polling loop. This saves one branching per packet. Performance evaluation configs: - On the DPDK-side, it's running some l3 forwarding app inside a VM on ESXi with one core assigned for polling. - On the client side, pktgen/dpdk is used to generate 64B tcp packets at line rate (14.8M PPS). Performance results on a Nehalem box (4cores@2.8GHzx2) shown below. CPU usage is collected factoring out the idle loop cost. - Before the patch, ~900K PPS with 65% CPU of a core used for DPDK. - After the patch, only 45% of a core used, while maintaining the same packet rate. Signed-off-by: Yong Wang <yongwang@vmware.com>
2014-11-05 01:49:42 +00:00
rcd = &rxq->comp_ring.base[rxq->comp_ring.next2proc].rcd;
nb_rxd++;
if (nb_rxd > rxq->cmd_ring[0].size) {
net/vmxnet3: coding style changes Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-08-29 19:18:46 +00:00
PMD_RX_LOG(ERR, "Used up quota of receiving packets,"
vmxnet3: improve Rx performance This patch includes two small performance optimizations on the rx path: (1) It adds unlikely hints on various infrequent error paths to the compiler to make branch prediction more efficient. (2) It also moves a constant assignment out of the pkt polling loop. This saves one branching per packet. Performance evaluation configs: - On the DPDK-side, it's running some l3 forwarding app inside a VM on ESXi with one core assigned for polling. - On the client side, pktgen/dpdk is used to generate 64B tcp packets at line rate (14.8M PPS). Performance results on a Nehalem box (4cores@2.8GHzx2) shown below. CPU usage is collected factoring out the idle loop cost. - Before the patch, ~900K PPS with 65% CPU of a core used for DPDK. - After the patch, only 45% of a core used, while maintaining the same packet rate. Signed-off-by: Yong Wang <yongwang@vmware.com>
2014-11-05 01:49:42 +00:00
" relinquish control.");
break;
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
}
}
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
return nb_rx;
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
}
/*
* Create memzone for device rings. malloc can't be used as the physical address is
* needed. If the memzone is already created, then this function returns a ptr
* to the old one.
*/
static const struct rte_memzone *
ring_dma_zone_reserve(struct rte_eth_dev *dev, const char *ring_name,
uint16_t queue_id, uint32_t ring_size, int socket_id)
{
char z_name[RTE_MEMZONE_NAMESIZE];
const struct rte_memzone *mz;
eal: deprecate rte_snprintf The function rte_snprintf serves no useful purpose. It is the same as snprintf() for all valid inputs. Deprecate it and replace all uses in current code. Leave the tests for the deprecated function in place. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-24 18:15:28 +00:00
snprintf(z_name, sizeof(z_name), "%s_%s_%d_%d",
net/vmxnet3: coding style changes Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-08-29 19:18:46 +00:00
dev->driver->pci_drv.driver.name, ring_name,
dev->data->port_id, queue_id);
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
mz = rte_memzone_lookup(z_name);
if (mz)
return mz;
return rte_memzone_reserve_aligned(z_name, ring_size,
net/vmxnet3: coding style changes Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-08-29 19:18:46 +00:00
socket_id, 0, VMXNET3_RING_BA_ALIGN);
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
}
int
vmxnet3_dev_tx_queue_setup(struct rte_eth_dev *dev,
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
uint16_t queue_idx,
uint16_t nb_desc,
unsigned int socket_id,
net/vmxnet3: coding style changes Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-08-29 19:18:46 +00:00
__rte_unused const struct rte_eth_txconf *tx_conf)
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
{
vmxnet3: leverage data ring on Tx path Data_ring is a pre-mapped guest ring buffer that vmxnet3 backend has access to directly without a need for buffer address mapping and unmapping during packet transmission. It is useful in reducing device emulation cost on the tx path. There are some additional cost though on the guest driver for packet copy and overall it's a win. This patch leverages the data_ring for packets with a length less than or equal to the data_ring entry size (128B). For larger packet, we won't use the data_ring as that requires one extra tx descriptor and it's not clear if doing this will be beneficial. Performance results show that this patch significantly boosts vmxnet3 64B tx performance (pkt rate) for l2fwd application on a Ivy Bridge server by >20% at which point we start to hit some bottleneck on the rx side. Signed-off-by: Yong Wang <yongwang@vmware.com>
2014-11-05 01:49:43 +00:00
struct vmxnet3_hw *hw = dev->data->dev_private;
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
const struct rte_memzone *mz;
struct vmxnet3_tx_queue *txq;
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
struct vmxnet3_cmd_ring *ring;
struct vmxnet3_comp_ring *comp_ring;
vmxnet3: leverage data ring on Tx path Data_ring is a pre-mapped guest ring buffer that vmxnet3 backend has access to directly without a need for buffer address mapping and unmapping during packet transmission. It is useful in reducing device emulation cost on the tx path. There are some additional cost though on the guest driver for packet copy and overall it's a win. This patch leverages the data_ring for packets with a length less than or equal to the data_ring entry size (128B). For larger packet, we won't use the data_ring as that requires one extra tx descriptor and it's not clear if doing this will be beneficial. Performance results show that this patch significantly boosts vmxnet3 64B tx performance (pkt rate) for l2fwd application on a Ivy Bridge server by >20% at which point we start to hit some bottleneck on the rx side. Signed-off-by: Yong Wang <yongwang@vmware.com>
2014-11-05 01:49:43 +00:00
struct vmxnet3_data_ring *data_ring;
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
int size;
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
PMD_INIT_FUNC_TRACE();
vmxnet3: fix Tx flags check Now that vmxnet3 supports TCP/UDP checksum offload, let's update the default txq flags to allow such offloads. Also fixed the tx queue setup check to allow TCP/UDP checksum and only error out if SCTP checksum is requested. Fixes: f598fd063bb1 ("vmxnet3: add Tx L4 checksum offload") Reported-by: Heng Ding <hengx.ding@intel.com> Signed-off-by: Yong Wang <yongwang@vmware.com>
2016-03-28 22:35:55 +00:00
if ((tx_conf->txq_flags & ETH_TXQ_FLAGS_NOXSUMSCTP) !=
vmxnet3: add Tx L4 checksum offload Support TCP/UDP checksum offload. Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-01-13 02:08:35 +00:00
ETH_TXQ_FLAGS_NOXSUMSCTP) {
PMD_INIT_LOG(ERR, "SCTP checksum offload not supported");
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
return -EINVAL;
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
}
net/vmxnet3: coding style changes Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-08-29 19:18:46 +00:00
txq = rte_zmalloc("ethdev_tx_queue", sizeof(struct vmxnet3_tx_queue),
RTE_CACHE_LINE_SIZE);
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
if (txq == NULL) {
vmxnet3: fix double spacing of log messages The debug log macro's already include newline, no need to double space the output. Note: other drivers have the same problem Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:13 +00:00
PMD_INIT_LOG(ERR, "Can not allocate tx queue structure");
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
return -ENOMEM;
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
}
txq->queue_id = queue_idx;
txq->port_id = dev->data->port_id;
txq->shared = &hw->tqd_start[queue_idx];
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
txq->hw = hw;
txq->qid = queue_idx;
txq->stopped = TRUE;
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
ring = &txq->cmd_ring;
comp_ring = &txq->comp_ring;
vmxnet3: leverage data ring on Tx path Data_ring is a pre-mapped guest ring buffer that vmxnet3 backend has access to directly without a need for buffer address mapping and unmapping during packet transmission. It is useful in reducing device emulation cost on the tx path. There are some additional cost though on the guest driver for packet copy and overall it's a win. This patch leverages the data_ring for packets with a length less than or equal to the data_ring entry size (128B). For larger packet, we won't use the data_ring as that requires one extra tx descriptor and it's not clear if doing this will be beneficial. Performance results show that this patch significantly boosts vmxnet3 64B tx performance (pkt rate) for l2fwd application on a Ivy Bridge server by >20% at which point we start to hit some bottleneck on the rx side. Signed-off-by: Yong Wang <yongwang@vmware.com>
2014-11-05 01:49:43 +00:00
data_ring = &txq->data_ring;
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
/* Tx vmxnet ring length should be between 512-4096 */
if (nb_desc < VMXNET3_DEF_TX_RING_SIZE) {
vmxnet3: fix double spacing of log messages The debug log macro's already include newline, no need to double space the output. Note: other drivers have the same problem Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:13 +00:00
PMD_INIT_LOG(ERR, "VMXNET3 Tx Ring Size Min: %u",
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
VMXNET3_DEF_TX_RING_SIZE);
remove trailing whitespaces This commit removes trailing whitespace from lines in files. Almost all files are affected, as the BSD license copyright header had trailing whitespace on 4 lines in it [hence the number of files reporting 8 lines changed in the diffstat]. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com> Acked-by: Neil Horman <nhorman@tuxdriver.com> [Thomas: remove spaces before tabs in libs] [Thomas: remove more trailing spaces in non-C files] Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-03 23:42:50 +00:00
return -EINVAL;
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
} else if (nb_desc > VMXNET3_TX_RING_MAX_SIZE) {
vmxnet3: fix double spacing of log messages The debug log macro's already include newline, no need to double space the output. Note: other drivers have the same problem Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:13 +00:00
PMD_INIT_LOG(ERR, "VMXNET3 Tx Ring Size Max: %u",
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
VMXNET3_TX_RING_MAX_SIZE);
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
return -EINVAL;
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
} else {
remove trailing whitespaces This commit removes trailing whitespace from lines in files. Almost all files are affected, as the BSD license copyright header had trailing whitespace on 4 lines in it [hence the number of files reporting 8 lines changed in the diffstat]. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com> Acked-by: Neil Horman <nhorman@tuxdriver.com> [Thomas: remove spaces before tabs in libs] [Thomas: remove more trailing spaces in non-C files] Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-03 23:42:50 +00:00
ring->size = nb_desc;
ring->size &= ~VMXNET3_RING_SIZE_MASK;
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
}
vmxnet3: leverage data ring on Tx path Data_ring is a pre-mapped guest ring buffer that vmxnet3 backend has access to directly without a need for buffer address mapping and unmapping during packet transmission. It is useful in reducing device emulation cost on the tx path. There are some additional cost though on the guest driver for packet copy and overall it's a win. This patch leverages the data_ring for packets with a length less than or equal to the data_ring entry size (128B). For larger packet, we won't use the data_ring as that requires one extra tx descriptor and it's not clear if doing this will be beneficial. Performance results show that this patch significantly boosts vmxnet3 64B tx performance (pkt rate) for l2fwd application on a Ivy Bridge server by >20% at which point we start to hit some bottleneck on the rx side. Signed-off-by: Yong Wang <yongwang@vmware.com>
2014-11-05 01:49:43 +00:00
comp_ring->size = data_ring->size = ring->size;
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
/* Tx vmxnet rings structure initialization*/
ring->next2fill = 0;
ring->next2comp = 0;
ring->gen = VMXNET3_INIT_GEN;
comp_ring->next2proc = 0;
comp_ring->gen = VMXNET3_INIT_GEN;
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
size = sizeof(struct Vmxnet3_TxDesc) * ring->size;
size += sizeof(struct Vmxnet3_TxCompDesc) * comp_ring->size;
vmxnet3: leverage data ring on Tx path Data_ring is a pre-mapped guest ring buffer that vmxnet3 backend has access to directly without a need for buffer address mapping and unmapping during packet transmission. It is useful in reducing device emulation cost on the tx path. There are some additional cost though on the guest driver for packet copy and overall it's a win. This patch leverages the data_ring for packets with a length less than or equal to the data_ring entry size (128B). For larger packet, we won't use the data_ring as that requires one extra tx descriptor and it's not clear if doing this will be beneficial. Performance results show that this patch significantly boosts vmxnet3 64B tx performance (pkt rate) for l2fwd application on a Ivy Bridge server by >20% at which point we start to hit some bottleneck on the rx side. Signed-off-by: Yong Wang <yongwang@vmware.com>
2014-11-05 01:49:43 +00:00
size += sizeof(struct Vmxnet3_TxDataDesc) * data_ring->size;
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
mz = ring_dma_zone_reserve(dev, "txdesc", queue_idx, size, socket_id);
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
if (mz == NULL) {
vmxnet3: fix double spacing of log messages The debug log macro's already include newline, no need to double space the output. Note: other drivers have the same problem Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:13 +00:00
PMD_INIT_LOG(ERR, "ERROR: Creating queue descriptors zone");
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
return -ENOMEM;
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
}
memset(mz->addr, 0, mz->len);
/* cmd_ring initialization */
ring->base = mz->addr;
ring->basePA = mz->phys_addr;
/* comp_ring initialization */
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
comp_ring->base = ring->base + ring->size;
comp_ring->basePA = ring->basePA +
(sizeof(struct Vmxnet3_TxDesc) * ring->size);
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
vmxnet3: leverage data ring on Tx path Data_ring is a pre-mapped guest ring buffer that vmxnet3 backend has access to directly without a need for buffer address mapping and unmapping during packet transmission. It is useful in reducing device emulation cost on the tx path. There are some additional cost though on the guest driver for packet copy and overall it's a win. This patch leverages the data_ring for packets with a length less than or equal to the data_ring entry size (128B). For larger packet, we won't use the data_ring as that requires one extra tx descriptor and it's not clear if doing this will be beneficial. Performance results show that this patch significantly boosts vmxnet3 64B tx performance (pkt rate) for l2fwd application on a Ivy Bridge server by >20% at which point we start to hit some bottleneck on the rx side. Signed-off-by: Yong Wang <yongwang@vmware.com>
2014-11-05 01:49:43 +00:00
/* data_ring initialization */
data_ring->base = (Vmxnet3_TxDataDesc *)(comp_ring->base + comp_ring->size);
data_ring->basePA = comp_ring->basePA +
(sizeof(struct Vmxnet3_TxCompDesc) * comp_ring->size);
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
/* cmd_ring0 buf_info allocation */
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
ring->buf_info = rte_zmalloc("tx_ring_buf_info",
add prefix to cache line macros CACHE_LINE_SIZE is a macro defined in machine/param.h in FreeBSD and conflicts with DPDK macro version. Adding RTE_ prefix to avoid conflicts. CACHE_LINE_MASK and CACHE_LINE_ROUNDUP are also prefixed. Signed-off-by: Sergio Gonzalez Monroy <sergio.gonzalez.monroy@intel.com> [Thomas: updated on HEAD, including PPC]
2014-11-19 12:26:06 +00:00
ring->size * sizeof(vmxnet3_buf_info_t), RTE_CACHE_LINE_SIZE);
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
if (ring->buf_info == NULL) {
vmxnet3: fix double spacing of log messages The debug log macro's already include newline, no need to double space the output. Note: other drivers have the same problem Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:13 +00:00
PMD_INIT_LOG(ERR, "ERROR: Creating tx_buf_info structure");
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
return -ENOMEM;
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
}
/* Update the data portion with txq */
dev->data->tx_queues[queue_idx] = txq;
return 0;
}
int
vmxnet3_dev_rx_queue_setup(struct rte_eth_dev *dev,
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
uint16_t queue_idx,
uint16_t nb_desc,
unsigned int socket_id,
net/vmxnet3: coding style changes Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-08-29 19:18:46 +00:00
__rte_unused const struct rte_eth_rxconf *rx_conf,
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
struct rte_mempool *mp)
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
{
const struct rte_memzone *mz;
struct vmxnet3_rx_queue *rxq;
net/vmxnet3: coding style changes Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-08-29 19:18:46 +00:00
struct vmxnet3_hw *hw = dev->data->dev_private;
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
struct vmxnet3_cmd_ring *ring0, *ring1, *ring;
struct vmxnet3_comp_ring *comp_ring;
int size;
uint8_t i;
char mem_name[32];
PMD_INIT_FUNC_TRACE();
net/vmxnet3: coding style changes Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-08-29 19:18:46 +00:00
rxq = rte_zmalloc("ethdev_rx_queue", sizeof(struct vmxnet3_rx_queue),
RTE_CACHE_LINE_SIZE);
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
if (rxq == NULL) {
vmxnet3: fix double spacing of log messages The debug log macro's already include newline, no need to double space the output. Note: other drivers have the same problem Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:13 +00:00
PMD_INIT_LOG(ERR, "Can not allocate rx queue structure");
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
return -ENOMEM;
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
}
rxq->mp = mp;
rxq->queue_id = queue_idx;
rxq->port_id = dev->data->port_id;
rxq->shared = &hw->rqd_start[queue_idx];
rxq->hw = hw;
rxq->qid1 = queue_idx;
rxq->qid2 = queue_idx + hw->num_rx_queues;
rxq->stopped = TRUE;
ring0 = &rxq->cmd_ring[0];
ring1 = &rxq->cmd_ring[1];
comp_ring = &rxq->comp_ring;
/* Rx vmxnet rings length should be between 256-4096 */
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
if (nb_desc < VMXNET3_DEF_RX_RING_SIZE) {
vmxnet3: fix double spacing of log messages The debug log macro's already include newline, no need to double space the output. Note: other drivers have the same problem Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:13 +00:00
PMD_INIT_LOG(ERR, "VMXNET3 Rx Ring Size Min: 256");
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
return -EINVAL;
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
} else if (nb_desc > VMXNET3_RX_RING_MAX_SIZE) {
vmxnet3: fix double spacing of log messages The debug log macro's already include newline, no need to double space the output. Note: other drivers have the same problem Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:13 +00:00
PMD_INIT_LOG(ERR, "VMXNET3 Rx Ring Size Max: 4096");
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
return -EINVAL;
} else {
ring0->size = nb_desc;
ring0->size &= ~VMXNET3_RING_SIZE_MASK;
ring1->size = ring0->size;
}
comp_ring->size = ring0->size + ring1->size;
/* Rx vmxnet rings structure initialization */
ring0->next2fill = 0;
ring1->next2fill = 0;
ring0->next2comp = 0;
ring1->next2comp = 0;
ring0->gen = VMXNET3_INIT_GEN;
ring1->gen = VMXNET3_INIT_GEN;
comp_ring->next2proc = 0;
comp_ring->gen = VMXNET3_INIT_GEN;
size = sizeof(struct Vmxnet3_RxDesc) * (ring0->size + ring1->size);
size += sizeof(struct Vmxnet3_RxCompDesc) * comp_ring->size;
mz = ring_dma_zone_reserve(dev, "rxdesc", queue_idx, size, socket_id);
if (mz == NULL) {
vmxnet3: fix double spacing of log messages The debug log macro's already include newline, no need to double space the output. Note: other drivers have the same problem Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:13 +00:00
PMD_INIT_LOG(ERR, "ERROR: Creating queue descriptors zone");
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
return -ENOMEM;
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
}
memset(mz->addr, 0, mz->len);
/* cmd_ring0 initialization */
ring0->base = mz->addr;
ring0->basePA = mz->phys_addr;
/* cmd_ring1 initialization */
ring1->base = ring0->base + ring0->size;
ring1->basePA = ring0->basePA + sizeof(struct Vmxnet3_RxDesc) * ring0->size;
/* comp_ring initialization */
vmxnet3: leverage data ring on Tx path Data_ring is a pre-mapped guest ring buffer that vmxnet3 backend has access to directly without a need for buffer address mapping and unmapping during packet transmission. It is useful in reducing device emulation cost on the tx path. There are some additional cost though on the guest driver for packet copy and overall it's a win. This patch leverages the data_ring for packets with a length less than or equal to the data_ring entry size (128B). For larger packet, we won't use the data_ring as that requires one extra tx descriptor and it's not clear if doing this will be beneficial. Performance results show that this patch significantly boosts vmxnet3 64B tx performance (pkt rate) for l2fwd application on a Ivy Bridge server by >20% at which point we start to hit some bottleneck on the rx side. Signed-off-by: Yong Wang <yongwang@vmware.com>
2014-11-05 01:49:43 +00:00
comp_ring->base = ring1->base + ring1->size;
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
comp_ring->basePA = ring1->basePA + sizeof(struct Vmxnet3_RxDesc) *
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
ring1->size;
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
/* cmd_ring0-cmd_ring1 buf_info allocation */
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
for (i = 0; i < VMXNET3_RX_CMDRING_SIZE; i++) {
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
ring = &rxq->cmd_ring[i];
ring->rid = i;
snprintf(mem_name, sizeof(mem_name), "rx_ring_%d_buf_info", i);
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
net/vmxnet3: coding style changes Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-08-29 19:18:46 +00:00
ring->buf_info = rte_zmalloc(mem_name,
ring->size * sizeof(vmxnet3_buf_info_t),
RTE_CACHE_LINE_SIZE);
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
if (ring->buf_info == NULL) {
vmxnet3: fix double spacing of log messages The debug log macro's already include newline, no need to double space the output. Note: other drivers have the same problem Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:13 +00:00
PMD_INIT_LOG(ERR, "ERROR: Creating rx_buf_info structure");
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
return -ENOMEM;
}
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
}
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
/* Update the data portion with rxq */
dev->data->rx_queues[queue_idx] = rxq;
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
return 0;
}
/*
* Initializes Receive Unit
* Load mbufs in rx queue in advance
*/
int
vmxnet3_dev_rxtx_init(struct rte_eth_dev *dev)
{
vmxnet3: remove useless adapter wrapper The adapter struct is just a wrapper around the vmxnet3_hw structure. Eliminate the wrapper and get rid of the macro used to access and needlessly cast the private data. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:39:58 +00:00
struct vmxnet3_hw *hw = dev->data->dev_private;
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
int i, ret;
uint8_t j;
PMD_INIT_FUNC_TRACE();
for (i = 0; i < hw->num_rx_queues; i++) {
vmxnet3_rx_queue_t *rxq = dev->data->rx_queues[i];
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
for (j = 0; j < VMXNET3_RX_CMDRING_SIZE; j++) {
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
/* Passing 0 as alloc_num will allocate full ring */
ret = vmxnet3_post_rx_bufs(rxq, j);
if (ret <= 0) {
net/vmxnet3: coding style changes Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-08-29 19:18:46 +00:00
PMD_INIT_LOG(ERR,
"ERROR: Posting Rxq: %d buffers ring: %d",
i, j);
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
return -ret;
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
}
net/vmxnet3: coding style changes Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-08-29 19:18:46 +00:00
/*
* Updating device with the index:next2fill to fill the
* mbufs for coming packets.
*/
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
if (unlikely(rxq->shared->ctrl.updateRxProd)) {
VMXNET3_WRITE_BAR0_REG(hw, rxprod_reg[j] + (rxq->queue_id * VMXNET3_REG_ALIGN),
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
rxq->cmd_ring[j].next2fill);
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
}
}
rxq->stopped = FALSE;
vmxnet3: support jumbo frames Add support for linking multi-segment buffers together to handle Jumbo packets. The vmxnet3 API supports having header and body buffer types. What this patch does is fill the primary ring completely with header buffers and the secondary ring with body buffers. This allows for non-jumbo frames to only use one mbuf (from primary ring); and jumbo frames will have first mbuf from primary ring and following mbufs from other ring. This could be optimized in future if the DPDK had API to supply different sized mbufs (two pools) into driver. Signed-off-by: Stephen Hemminger <shemming@brocade.com> Acked-by: Remy Horton <remy.horton@intel.com> Release note addition: Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2016-03-04 18:08:00 +00:00
rxq->start_seg = NULL;
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
}
for (i = 0; i < dev->data->nb_tx_queues; i++) {
struct vmxnet3_tx_queue *txq = dev->data->tx_queues[i];
vmxnet3: cleanup style and indentation This driver had several style problems, the worst of which was botched indentation. Fix almost all the problems reported by checkpatch. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:37:04 +00:00
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
txq->stopped = FALSE;
}
return 0;
}
static uint8_t rss_intel_key[40] = {
0x6D, 0x5A, 0x56, 0xDA, 0x25, 0x5B, 0x0E, 0xC2,
0x41, 0x67, 0x25, 0x3D, 0x43, 0xA3, 0x8F, 0xB0,
0xD0, 0xCA, 0x2B, 0xCB, 0xAE, 0x7B, 0x30, 0xB4,
0x77, 0xCB, 0x2D, 0xA3, 0x80, 0x30, 0xF2, 0x0C,
0x6A, 0x42, 0xB7, 0x3B, 0xBE, 0xAC, 0x01, 0xFA,
};
/*
* Configure RSS feature
*/
int
vmxnet3_rss_configure(struct rte_eth_dev *dev)
{
vmxnet3: remove useless adapter wrapper The adapter struct is just a wrapper around the vmxnet3_hw structure. Eliminate the wrapper and get rid of the macro used to access and needlessly cast the private data. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:39:58 +00:00
struct vmxnet3_hw *hw = dev->data->dev_private;
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
struct VMXNET3_RSSConf *dev_rss_conf;
struct rte_eth_rss_conf *port_rss_conf;
ethdev: more RSS flags - i40e RSS flags have been added (and enlarged to 64-bit) - A new configuration of 'uint8_t rss_key_len' has been added in 'struct rte_eth_rss_conf' to support different length of RSS keys. - In each PMD, only the supported flags are masked. Signed-off-by: Helin Zhang <helin.zhang@intel.com> Signed-off-by: Jing Chen <jing.d.chen@intel.com> Acked-by: Cunming Liang <cunming.liang@intel.com> Acked-by: Jijiang Liu <jijiang.liu@intel.com> Acked-by: Jingjing Wu <jingjing.wu@intel.com> Acked-by: Heqing Zhu <heqing.zhu@intel.com> Tested-by: Waterman Cao <waterman.cao@intel.com>
2014-06-05 05:08:50 +00:00
uint64_t rss_hf;
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
uint8_t i, j;
PMD_INIT_FUNC_TRACE();
vmxnet3: remove useless adapter wrapper The adapter struct is just a wrapper around the vmxnet3_hw structure. Eliminate the wrapper and get rid of the macro used to access and needlessly cast the private data. Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2014-06-13 01:39:58 +00:00
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
dev_rss_conf = hw->rss_conf;
port_rss_conf = &dev->data->dev_conf.rx_adv_conf.rss_conf;
/* loading hashFunc */
dev_rss_conf->hashFunc = VMXNET3_RSS_HASH_FUNC_TOEPLITZ;
/* loading hashKeySize */
dev_rss_conf->hashKeySize = VMXNET3_RSS_MAX_KEY_SIZE;
net/vmxnet3: coding style changes Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-08-29 19:18:46 +00:00
/* loading indTableSize: Must not exceed VMXNET3_RSS_MAX_IND_TABLE_SIZE (128)*/
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
dev_rss_conf->indTableSize = (uint16_t)(hw->num_rx_queues * 4);
if (port_rss_conf->rss_key == NULL) {
/* Default hash key */
port_rss_conf->rss_key = rss_intel_key;
}
/* loading hashKey */
net/vmxnet3: coding style changes Signed-off-by: Yong Wang <yongwang@vmware.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
2016-08-29 19:18:46 +00:00
memcpy(&dev_rss_conf->hashKey[0], port_rss_conf->rss_key,
dev_rss_conf->hashKeySize);
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
/* loading indTable */
for (i = 0, j = 0; i < dev_rss_conf->indTableSize; i++, j++) {
if (j == dev->data->nb_rx_queues)
j = 0;
dev_rss_conf->indTable[i] = j;
}
/* loading hashType */
dev_rss_conf->hashType = 0;
ethdev: more RSS flags - i40e RSS flags have been added (and enlarged to 64-bit) - A new configuration of 'uint8_t rss_key_len' has been added in 'struct rte_eth_rss_conf' to support different length of RSS keys. - In each PMD, only the supported flags are masked. Signed-off-by: Helin Zhang <helin.zhang@intel.com> Signed-off-by: Jing Chen <jing.d.chen@intel.com> Acked-by: Cunming Liang <cunming.liang@intel.com> Acked-by: Jijiang Liu <jijiang.liu@intel.com> Acked-by: Jingjing Wu <jingjing.wu@intel.com> Acked-by: Heqing Zhu <heqing.zhu@intel.com> Tested-by: Waterman Cao <waterman.cao@intel.com>
2014-06-05 05:08:50 +00:00
rss_hf = port_rss_conf->rss_hf & VMXNET3_RSS_OFFLOAD_ALL;
if (rss_hf & ETH_RSS_IPV4)
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
dev_rss_conf->hashType |= VMXNET3_RSS_HASH_TYPE_IPV4;
ethdev: unification of RSS offload types RSS offload types were defined separately for 1/10G and 40G NICs, and have no relationship with flow types. The modifications are to unify all RSS offload types for all PMDs. Unified RSS offload types have new and common names which can be used for any PMD or applications, and decouple from specific hardwares. Signed-off-by: Helin Zhang <helin.zhang@intel.com> Acked-by: Jingjing Wu <jingjing.wu@intel.com> [Thomas: merge with fm10k] Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2015-02-04 07:16:32 +00:00
if (rss_hf & ETH_RSS_NONFRAG_IPV4_TCP)
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
dev_rss_conf->hashType |= VMXNET3_RSS_HASH_TYPE_TCP_IPV4;
ethdev: more RSS flags - i40e RSS flags have been added (and enlarged to 64-bit) - A new configuration of 'uint8_t rss_key_len' has been added in 'struct rte_eth_rss_conf' to support different length of RSS keys. - In each PMD, only the supported flags are masked. Signed-off-by: Helin Zhang <helin.zhang@intel.com> Signed-off-by: Jing Chen <jing.d.chen@intel.com> Acked-by: Cunming Liang <cunming.liang@intel.com> Acked-by: Jijiang Liu <jijiang.liu@intel.com> Acked-by: Jingjing Wu <jingjing.wu@intel.com> Acked-by: Heqing Zhu <heqing.zhu@intel.com> Tested-by: Waterman Cao <waterman.cao@intel.com>
2014-06-05 05:08:50 +00:00
if (rss_hf & ETH_RSS_IPV6)
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
dev_rss_conf->hashType |= VMXNET3_RSS_HASH_TYPE_IPV6;
ethdev: unification of RSS offload types RSS offload types were defined separately for 1/10G and 40G NICs, and have no relationship with flow types. The modifications are to unify all RSS offload types for all PMDs. Unified RSS offload types have new and common names which can be used for any PMD or applications, and decouple from specific hardwares. Signed-off-by: Helin Zhang <helin.zhang@intel.com> Acked-by: Jingjing Wu <jingjing.wu@intel.com> [Thomas: merge with fm10k] Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
2015-02-04 07:16:32 +00:00
if (rss_hf & ETH_RSS_NONFRAG_IPV6_TCP)
vmxnet3: import new vmxnet3 poll mode driver implementation Poll Mode Driver for Paravirtual VMXNET3 NIC. As a PMD, the VMXNET3 driver provides the packet reception and transmission callbacks, vmxnet3_recv_pkts and vmxnet3_xmit_pkts. It does not support scattered packet reception as part of vmxnet3_recv_pkts and vmxnet3_xmit_pkts. Also, it does not support scattered packet reception as part of the device operations supported. The VMXNET3 PMD handles all the packet buffer memory allocation and resides in guest address space and it is solely responsible to free that memory when not needed. The packet buffers and features to be supported are made available to hypervisor via VMXNET3 PCI configuration space BARs. During RX/TX, the packet buffers are exchanged by their GPAs, and the hypervisor loads the buffers with packets in the RX case and sends packets to vSwitch in the TX case. The VMXNET3 PMD is compiled with vmxnet3 device headers. The interface is similar to that of the other PMDs available in the Intel(R) DPDK API. The driver pre-allocates the packet buffers and loads the command ring descriptors in advance. The hypervisor fills those packet buffers on packet arrival and write completion ring descriptors, which are eventually pulled by the PMD. After reception, the Intel(R) DPDK application frees the descriptors and loads new packet buffers for the coming packets. The interrupts are disabled and there is no notification required. This keeps performance up on the RX side, even though the device provides a notification feature. In the transmit routine, the Intel(R) DPDK application fills packet buffer pointers in the descriptors of the command ring and notifies the hypervisor. In response the hypervisor takes packets and passes them to the vSwitch. It writes into the completion descriptors ring. The rings are read by the PMD in the next transmit routine call and the buffers and descriptors are freed from memory. Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
2014-02-10 15:27:26 +00:00
dev_rss_conf->hashType |= VMXNET3_RSS_HASH_TYPE_TCP_IPV6;
return VMXNET3_SUCCESS;
}
Reference in New Issue Copy Permalink