Re: [PATCH v3 1/1] app/testpmd: control passing Rx metadata to PMD
17/10/2022 10:32, Andrew Rybchenko: > On 10/6/22 21:35, Hanumanth Pothula wrote: > > Presently, Rx metadata is sent to PMD by default, leading > > to a performance drop as processing for the same in rx path > > takes extra cycles. > > > > Hence, introducing command line argument, 'nic-to-pmd-rx-metadata' > > to control passing rx metadata to PMD. By default it’s disabled. > > > > Signed-off-by: Hanumanth Pothula > > Acked-by: Andrew Rybchenko > > Applied to dpdk-next-net/main, thanks. I'm not sure this patch is really acceptable. It is disabling Rx metadata by default just for benchmarking reason because your driver is doing some processing even if metadata is not required. From a user perspective, if a command requesting metadata is entered, it won't work until we enable this new option on startup. It looks terrible. Please tell me I misunderstood something.
[PATCH v14 00/18] add support for idpf PMD in DPDK
This patchset introduced the idpf (Infrastructure Data Path Function) PMD in DPDK for Intel® IPU E2000 (Device ID: 0x1452). The Intel® IPU E2000 targets to deliver high performance under real workloads with security and isolation. Please refer to https://www.intel.com/content/www/us/en/products/network-io/infrastructure-processing-units/asic/e2000-asic.html for more information. Linux upstream is still ongoing, previous work refers to https://patchwork.ozlabs.org/project/intel-wired-lan/patch/20220128001009.721392-20-alan.br...@intel.com/. v2-v4: fixed some coding style issues and did some refactors. v5: fixed typo. v6-v9: fixed build errors and coding style issues. v11: - move shared code to common/idpf/base - Create one vport if there's no vport devargs - Refactor if conditions according to coding style - Refactor virtual channel return values - Refine dev_stop function - Refine RSS lut/key - Fix build error v12: - Refine dev_configure - Fix coding style accoding to the comments - Re-order patch - Romove dev_supported_ptypes_get v13: - refine dev_start/stop and queue_start/stop - fix timestamp offload v14: - fix wrong position for rte_validate_tx_offload Junfeng Guo (18): common/idpf: introduce common library net/idpf: add support for device initialization net/idpf: add Tx queue setup net/idpf: add Rx queue setup net/idpf: add support for device start and stop net/idpf: add support for queue start net/idpf: add support for queue stop net/idpf: add queue release net/idpf: add support for MTU configuration net/idpf: add support for basic Rx datapath net/idpf: add support for basic Tx datapath net/idpf: support parsing packet type net/idpf: add support for write back based on ITR expire net/idpf: add support for RSS net/idpf: add support for Rx offloading net/idpf: add support for Tx offloading net/idpf: add AVX512 data path for single queue model net/idpf: add support for timestamp offload MAINTAINERS |9 + doc/guides/nics/features/idpf.ini | 18 + doc/guides/nics/idpf.rst | 85 + doc/guides/nics/index.rst |1 + doc/guides/rel_notes/release_22_11.rst|6 + drivers/common/idpf/base/idpf_alloc.h | 22 + drivers/common/idpf/base/idpf_common.c| 364 +++ drivers/common/idpf/base/idpf_controlq.c | 691 drivers/common/idpf/base/idpf_controlq.h | 224 ++ drivers/common/idpf/base/idpf_controlq_api.h | 234 ++ .../common/idpf/base/idpf_controlq_setup.c| 179 + drivers/common/idpf/base/idpf_devids.h| 18 + drivers/common/idpf/base/idpf_lan_pf_regs.h | 134 + drivers/common/idpf/base/idpf_lan_txrx.h | 428 +++ drivers/common/idpf/base/idpf_lan_vf_regs.h | 114 + drivers/common/idpf/base/idpf_osdep.h | 364 +++ drivers/common/idpf/base/idpf_prototype.h | 45 + drivers/common/idpf/base/idpf_type.h | 106 + drivers/common/idpf/base/meson.build | 14 + drivers/common/idpf/base/siov_regs.h | 47 + drivers/common/idpf/base/virtchnl.h | 2866 + drivers/common/idpf/base/virtchnl2.h | 1462 + drivers/common/idpf/base/virtchnl2_lan_desc.h | 606 .../common/idpf/base/virtchnl_inline_ipsec.h | 567 drivers/common/idpf/meson.build |4 + drivers/common/idpf/version.map | 12 + drivers/common/meson.build|1 + drivers/net/idpf/idpf_ethdev.c| 1280 drivers/net/idpf/idpf_ethdev.h| 252 ++ drivers/net/idpf/idpf_logs.h | 56 + drivers/net/idpf/idpf_rxtx.c | 2308 + drivers/net/idpf/idpf_rxtx.h | 291 ++ drivers/net/idpf/idpf_rxtx_vec_avx512.c | 871 + drivers/net/idpf/idpf_rxtx_vec_common.h | 100 + drivers/net/idpf/idpf_vchnl.c | 1430 drivers/net/idpf/meson.build | 44 + drivers/net/idpf/version.map |3 + drivers/net/meson.build |1 + 38 files changed, 15257 insertions(+) create mode 100644 doc/guides/nics/features/idpf.ini create mode 100644 doc/guides/nics/idpf.rst create mode 100644 drivers/common/idpf/base/idpf_alloc.h create mode 100644 drivers/common/idpf/base/idpf_common.c create mode 100644 drivers/common/idpf/base/idpf_controlq.c create mode 100644 drivers/common/idpf/base/idpf_controlq.h create mode 100644 drivers/common/idpf/base/idpf_controlq_api.h create mode 100644 drivers/common/idpf/base/idpf_controlq_setup.c create mode 100644 drivers/common/idpf/base/idpf_devids.h create mode 100644 drivers/common/idpf/base/idpf_lan_pf_regs.h create mode 100644 drivers/common/idpf/base/idpf_lan_txrx.h create mode 100644 drivers/common/idpf/base/idpf_lan_vf_regs.h create mode 100644 drivers/common/idpf/base/idpf_osdep.h
[PATCH v14 02/18] net/idpf: add support for device initialization
Support device init and add the following dev ops: - dev_configure - dev_close - dev_infos_get Signed-off-by: Beilei Xing Signed-off-by: Xiaoyun Li Signed-off-by: Xiao Wang Signed-off-by: Wenjun Wu Signed-off-by: Junfeng Guo --- MAINTAINERS| 9 + doc/guides/nics/features/idpf.ini | 9 + doc/guides/nics/idpf.rst | 66 ++ doc/guides/nics/index.rst | 1 + doc/guides/rel_notes/release_22_11.rst | 6 + drivers/net/idpf/idpf_ethdev.c | 886 + drivers/net/idpf/idpf_ethdev.h | 189 ++ drivers/net/idpf/idpf_logs.h | 56 ++ drivers/net/idpf/idpf_vchnl.c | 468 + drivers/net/idpf/meson.build | 15 + drivers/net/idpf/version.map | 3 + drivers/net/meson.build| 1 + 12 files changed, 1709 insertions(+) create mode 100644 doc/guides/nics/features/idpf.ini create mode 100644 doc/guides/nics/idpf.rst create mode 100644 drivers/net/idpf/idpf_ethdev.c create mode 100644 drivers/net/idpf/idpf_ethdev.h create mode 100644 drivers/net/idpf/idpf_logs.h create mode 100644 drivers/net/idpf/idpf_vchnl.c create mode 100644 drivers/net/idpf/meson.build create mode 100644 drivers/net/idpf/version.map diff --git a/MAINTAINERS b/MAINTAINERS index 92b381bc30..34f8b9cc61 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -764,6 +764,15 @@ F: drivers/net/ice/ F: doc/guides/nics/ice.rst F: doc/guides/nics/features/ice.ini +Intel idpf +M: Jingjing Wu +M: Beilei Xing +T: git://dpdk.org/next/dpdk-next-net-intel +F: drivers/net/idpf/ +F: drivers/common/idpf/ +F: doc/guides/nics/idpf.rst +F: doc/guides/nics/features/idpf.ini + Intel igc M: Junfeng Guo M: Simei Su diff --git a/doc/guides/nics/features/idpf.ini b/doc/guides/nics/features/idpf.ini new file mode 100644 index 00..46aab2eb61 --- /dev/null +++ b/doc/guides/nics/features/idpf.ini @@ -0,0 +1,9 @@ +; +; Supported features of the 'idpf' network poll mode driver. +; +; Refer to default.ini for the full list of available PMD features. +; +[Features] +Linux= Y +x86-32 = Y +x86-64 = Y diff --git a/doc/guides/nics/idpf.rst b/doc/guides/nics/idpf.rst new file mode 100644 index 00..c1001d5d0c --- /dev/null +++ b/doc/guides/nics/idpf.rst @@ -0,0 +1,66 @@ +.. SPDX-License-Identifier: BSD-3-Clause +Copyright(c) 2022 Intel Corporation. + +IDPF Poll Mode Driver +== + +The [*EXPERIMENTAL*] idpf PMD (**librte_net_idpf**) provides poll mode driver support for +Intel® Infrastructure Processing Unit (Intel® IPU) E2000. + + +Linux Prerequisites +--- + +- Follow the DPDK :ref:`Getting Started Guide for Linux ` to setup the basic DPDK environment. + +- To get better performance on Intel platforms, please follow the "How to get best performance with NICs on Intel platforms" + section of the :ref:`Getting Started Guide for Linux `. + + +Pre-Installation Configuration +-- + +Runtime Config Options +~~ + +- ``vport`` (default ``0``) + + The IDPF PMD supports creation of multiple vports for one PCI device, each vport + corresponds to a single ethdev. Using the ``devargs`` parameter ``vport`` the user + can specify the vports with specific ID to be created, for example:: + +-a ca:00.0,vport=[0,2,3] + + Then idpf PMD will create 3 vports (ethdevs) for device ca:00.0. + NOTE: If the parameter is not provided, the vport 0 will be created by default. + +- ``rx_single`` (default ``0``) + + There're two queue modes supported by Intel® IPU Ethernet ES2000 Series, single queue + mode and split queue mode for Rx queue. User can choose Rx queue mode by the ``devargs`` + parameter ``rx_single``. + +-a ca:00.0,rx_single=1 + + Then idpf PMD will configure Rx queue with single queue mode. Otherwise, split queue + mode is chosen by default. + +- ``tx_single`` (default ``0``) + + There're two queue modes supported by Intel® IPU Ethernet ES2000 Series, single queue + mode and split queue mode for Tx queue. User can choose Tx queue mode by the ``devargs`` + parameter ``tx_single``. + +-a ca:00.0,tx_single=1 + + Then idpf PMD will configure Tx queue with single queue mode. Otherwise, split queue + mode is chosen by default. + + +Driver compilation and testing +-- + +Refer to the document :ref:`compiling and testing a PMD for a NIC ` +for details. + + diff --git a/doc/guides/nics/index.rst b/doc/guides/nics/index.rst index 32c7544968..b4dd5ea522 100644 --- a/doc/guides/nics/index.rst +++ b/doc/guides/nics/index.rst @@ -33,6 +33,7 @@ Network Interface Controller Drivers hns3 i40e ice +idpf igb igc ionic diff --git a/doc/guides/rel_notes/release_22_11.rst b/doc/guides/rel_notes/release_22_11.rst index 1c3daf141d..4af790de37 100644 --- a/doc/guides/rel_notes/release_22_11.rst +++ b
[PATCH v14 03/18] net/idpf: add Tx queue setup
Add support for tx_queue_setup ops.
In the single queue model, the same descriptor queue is used by SW to
post buffer descriptors to HW and by HW to post completed descriptors
to SW.
In the split queue model, "RX buffer queues" are used to pass
descriptor buffers from SW to HW while Rx queues are used only to
pass the descriptor completions, that is, descriptors that point
to completed buffers, from HW to SW. This is contrary to the single
queue model in which Rx queues are used for both purposes.
Signed-off-by: Beilei Xing
Signed-off-by: Xiaoyun Li
Signed-off-by: Junfeng Guo
---
drivers/net/idpf/idpf_ethdev.c | 13 ++
drivers/net/idpf/idpf_rxtx.c | 364 +
drivers/net/idpf/idpf_rxtx.h | 70 +++
drivers/net/idpf/idpf_vchnl.c | 2 -
drivers/net/idpf/meson.build | 1 +
5 files changed, 448 insertions(+), 2 deletions(-)
create mode 100644 drivers/net/idpf/idpf_rxtx.c
create mode 100644 drivers/net/idpf/idpf_rxtx.h
diff --git a/drivers/net/idpf/idpf_ethdev.c b/drivers/net/idpf/idpf_ethdev.c
index 112dccf5c9..11f8b4ba1c 100644
--- a/drivers/net/idpf/idpf_ethdev.c
+++ b/drivers/net/idpf/idpf_ethdev.c
@@ -11,6 +11,7 @@
#include
#include "idpf_ethdev.h"
+#include "idpf_rxtx.h"
#define IDPF_TX_SINGLE_Q "tx_single"
#define IDPF_RX_SINGLE_Q "rx_single"
@@ -37,6 +38,7 @@ static void idpf_adapter_rel(struct idpf_adapter *adapter);
static const struct eth_dev_ops idpf_eth_dev_ops = {
.dev_configure = idpf_dev_configure,
.dev_close = idpf_dev_close,
+ .tx_queue_setup = idpf_tx_queue_setup,
.dev_infos_get = idpf_dev_info_get,
};
@@ -56,6 +58,17 @@ idpf_dev_info_get(struct rte_eth_dev *dev, struct
rte_eth_dev_info *dev_info)
dev_info->max_mac_addrs = IDPF_NUM_MACADDR_MAX;
+ dev_info->default_txconf = (struct rte_eth_txconf) {
+ .tx_free_thresh = IDPF_DEFAULT_TX_FREE_THRESH,
+ .tx_rs_thresh = IDPF_DEFAULT_TX_RS_THRESH,
+ };
+
+ dev_info->tx_desc_lim = (struct rte_eth_desc_lim) {
+ .nb_max = IDPF_MAX_RING_DESC,
+ .nb_min = IDPF_MIN_RING_DESC,
+ .nb_align = IDPF_ALIGN_RING_DESC,
+ };
+
return 0;
}
diff --git a/drivers/net/idpf/idpf_rxtx.c b/drivers/net/idpf/idpf_rxtx.c
new file mode 100644
index 00..4afa0a2560
--- /dev/null
+++ b/drivers/net/idpf/idpf_rxtx.c
@@ -0,0 +1,364 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2022 Intel Corporation
+ */
+
+#include
+#include
+
+#include "idpf_ethdev.h"
+#include "idpf_rxtx.h"
+
+static int
+check_tx_thresh(uint16_t nb_desc, uint16_t tx_rs_thresh,
+ uint16_t tx_free_thresh)
+{
+ /* TX descriptors will have their RS bit set after tx_rs_thresh
+* descriptors have been used. The TX descriptor ring will be cleaned
+* after tx_free_thresh descriptors are used or if the number of
+* descriptors required to transmit a packet is greater than the
+* number of free TX descriptors.
+*
+* The following constraints must be satisfied:
+* - tx_rs_thresh must be less than the size of the ring minus 2.
+* - tx_free_thresh must be less than the size of the ring minus 3.
+* - tx_rs_thresh must be less than or equal to tx_free_thresh.
+* - tx_rs_thresh must be a divisor of the ring size.
+*
+* One descriptor in the TX ring is used as a sentinel to avoid a H/W
+* race condition, hence the maximum threshold constraints. When set
+* to zero use default values.
+*/
+ if (tx_rs_thresh >= (nb_desc - 2)) {
+ PMD_INIT_LOG(ERR, "tx_rs_thresh (%u) must be less than the "
+"number of TX descriptors (%u) minus 2",
+tx_rs_thresh, nb_desc);
+ return -EINVAL;
+ }
+ if (tx_free_thresh >= (nb_desc - 3)) {
+ PMD_INIT_LOG(ERR, "tx_free_thresh (%u) must be less than the "
+"number of TX descriptors (%u) minus 3.",
+tx_free_thresh, nb_desc);
+ return -EINVAL;
+ }
+ if (tx_rs_thresh > tx_free_thresh) {
+ PMD_INIT_LOG(ERR, "tx_rs_thresh (%u) must be less than or "
+"equal to tx_free_thresh (%u).",
+tx_rs_thresh, tx_free_thresh);
+ return -EINVAL;
+ }
+ if ((nb_desc % tx_rs_thresh) != 0) {
+ PMD_INIT_LOG(ERR, "tx_rs_thresh (%u) must be a divisor of the "
+"number of TX descriptors (%u).",
+tx_rs_thresh, nb_desc);
+ return -EINVAL;
+ }
+
+ return 0;
+}
+
+static void
+reset_split_tx_descq(struct idpf_tx_queue *txq)
+{
+ struct idpf_tx_entry *txe;
+ uint32_t i, size;
+
[PATCH v14 04/18] net/idpf: add Rx queue setup
Add support for rx_queue_setup ops.
Signed-off-by: Beilei Xing
Signed-off-by: Xiaoyun Li
Signed-off-by: Junfeng Guo
---
drivers/net/idpf/idpf_ethdev.c | 11 +
drivers/net/idpf/idpf_rxtx.c | 400 +
drivers/net/idpf/idpf_rxtx.h | 46
3 files changed, 457 insertions(+)
diff --git a/drivers/net/idpf/idpf_ethdev.c b/drivers/net/idpf/idpf_ethdev.c
index 11f8b4ba1c..0585153f69 100644
--- a/drivers/net/idpf/idpf_ethdev.c
+++ b/drivers/net/idpf/idpf_ethdev.c
@@ -38,6 +38,7 @@ static void idpf_adapter_rel(struct idpf_adapter *adapter);
static const struct eth_dev_ops idpf_eth_dev_ops = {
.dev_configure = idpf_dev_configure,
.dev_close = idpf_dev_close,
+ .rx_queue_setup = idpf_rx_queue_setup,
.tx_queue_setup = idpf_tx_queue_setup,
.dev_infos_get = idpf_dev_info_get,
};
@@ -63,12 +64,22 @@ idpf_dev_info_get(struct rte_eth_dev *dev, struct
rte_eth_dev_info *dev_info)
.tx_rs_thresh = IDPF_DEFAULT_TX_RS_THRESH,
};
+ dev_info->default_rxconf = (struct rte_eth_rxconf) {
+ .rx_free_thresh = IDPF_DEFAULT_RX_FREE_THRESH,
+ };
+
dev_info->tx_desc_lim = (struct rte_eth_desc_lim) {
.nb_max = IDPF_MAX_RING_DESC,
.nb_min = IDPF_MIN_RING_DESC,
.nb_align = IDPF_ALIGN_RING_DESC,
};
+ dev_info->rx_desc_lim = (struct rte_eth_desc_lim) {
+ .nb_max = IDPF_MAX_RING_DESC,
+ .nb_min = IDPF_MIN_RING_DESC,
+ .nb_align = IDPF_ALIGN_RING_DESC,
+ };
+
return 0;
}
diff --git a/drivers/net/idpf/idpf_rxtx.c b/drivers/net/idpf/idpf_rxtx.c
index 4afa0a2560..25dd5d85d5 100644
--- a/drivers/net/idpf/idpf_rxtx.c
+++ b/drivers/net/idpf/idpf_rxtx.c
@@ -8,6 +8,21 @@
#include "idpf_ethdev.h"
#include "idpf_rxtx.h"
+static int
+check_rx_thresh(uint16_t nb_desc, uint16_t thresh)
+{
+ /* The following constraints must be satisfied:
+* thresh < rxq->nb_rx_desc
+*/
+ if (thresh >= nb_desc) {
+ PMD_INIT_LOG(ERR, "rx_free_thresh (%u) must be less than %u",
+thresh, nb_desc);
+ return -EINVAL;
+ }
+
+ return 0;
+}
+
static int
check_tx_thresh(uint16_t nb_desc, uint16_t tx_rs_thresh,
uint16_t tx_free_thresh)
@@ -56,6 +71,87 @@ check_tx_thresh(uint16_t nb_desc, uint16_t tx_rs_thresh,
return 0;
}
+static void
+reset_split_rx_descq(struct idpf_rx_queue *rxq)
+{
+ uint16_t len;
+ uint32_t i;
+
+ if (rxq == NULL)
+ return;
+
+ len = rxq->nb_rx_desc + IDPF_RX_MAX_BURST;
+
+ for (i = 0; i < len * sizeof(struct virtchnl2_rx_flex_desc_adv_nic_3);
+i++)
+ ((volatile char *)rxq->rx_ring)[i] = 0;
+
+ rxq->rx_tail = 0;
+ rxq->expected_gen_id = 1;
+}
+
+static void
+reset_split_rx_bufq(struct idpf_rx_queue *rxq)
+{
+ uint16_t len;
+ uint32_t i;
+
+ if (rxq == NULL)
+ return;
+
+ len = rxq->nb_rx_desc + IDPF_RX_MAX_BURST;
+
+ for (i = 0; i < len * sizeof(struct virtchnl2_splitq_rx_buf_desc);
+i++)
+ ((volatile char *)rxq->rx_ring)[i] = 0;
+
+ memset(&rxq->fake_mbuf, 0x0, sizeof(rxq->fake_mbuf));
+
+ for (i = 0; i < IDPF_RX_MAX_BURST; i++)
+ rxq->sw_ring[rxq->nb_rx_desc + i] = &rxq->fake_mbuf;
+
+ /* The next descriptor id which can be received. */
+ rxq->rx_next_avail = 0;
+
+ /* The next descriptor id which can be refilled. */
+ rxq->rx_tail = 0;
+ /* The number of descriptors which can be refilled. */
+ rxq->nb_rx_hold = rxq->nb_rx_desc - 1;
+
+ rxq->bufq1 = NULL;
+ rxq->bufq2 = NULL;
+}
+
+static void
+reset_single_rx_queue(struct idpf_rx_queue *rxq)
+{
+ uint16_t len;
+ uint32_t i;
+
+ if (rxq == NULL)
+ return;
+
+ len = rxq->nb_rx_desc + IDPF_RX_MAX_BURST;
+
+ for (i = 0; i < len * sizeof(struct virtchnl2_singleq_rx_buf_desc);
+i++)
+ ((volatile char *)rxq->rx_ring)[i] = 0;
+
+ memset(&rxq->fake_mbuf, 0x0, sizeof(rxq->fake_mbuf));
+
+ for (i = 0; i < IDPF_RX_MAX_BURST; i++)
+ rxq->sw_ring[rxq->nb_rx_desc + i] = &rxq->fake_mbuf;
+
+ rxq->rx_tail = 0;
+ rxq->nb_rx_hold = 0;
+
+ if (rxq->pkt_first_seg != NULL)
+ rte_pktmbuf_free(rxq->pkt_first_seg);
+
+ rxq->pkt_first_seg = NULL;
+ rxq->pkt_last_seg = NULL;
+}
+
static void
reset_split_tx_descq(struct idpf_tx_queue *txq)
{
@@ -145,6 +241,310 @@ reset_single_tx_queue(struct idpf_tx_queue *txq)
txq->next_rs = txq->rs_thresh - 1;
}
+static int
+idpf_rx_split_bufq_setup(struct rte_eth_dev *dev, struct idpf_rx_queue *bufq,
+uint16_t queue_idx, uint16_t r
[PATCH v14 05/18] net/idpf: add support for device start and stop
Add dev ops dev_start, dev_stop and link_update.
Signed-off-by: Beilei Xing
Signed-off-by: Xiaoyun Li
Signed-off-by: Junfeng Guo
---
drivers/net/idpf/idpf_ethdev.c | 55 ++
drivers/net/idpf/idpf_rxtx.c | 20 +
2 files changed, 75 insertions(+)
diff --git a/drivers/net/idpf/idpf_ethdev.c b/drivers/net/idpf/idpf_ethdev.c
index 0585153f69..3430d00e92 100644
--- a/drivers/net/idpf/idpf_ethdev.c
+++ b/drivers/net/idpf/idpf_ethdev.c
@@ -30,17 +30,38 @@ static const char * const idpf_valid_args[] = {
};
static int idpf_dev_configure(struct rte_eth_dev *dev);
+static int idpf_dev_start(struct rte_eth_dev *dev);
+static int idpf_dev_stop(struct rte_eth_dev *dev);
static int idpf_dev_close(struct rte_eth_dev *dev);
static int idpf_dev_info_get(struct rte_eth_dev *dev,
struct rte_eth_dev_info *dev_info);
static void idpf_adapter_rel(struct idpf_adapter *adapter);
+static int
+idpf_dev_link_update(struct rte_eth_dev *dev,
+__rte_unused int wait_to_complete)
+{
+ struct rte_eth_link new_link;
+
+ memset(&new_link, 0, sizeof(new_link));
+
+ new_link.link_speed = RTE_ETH_SPEED_NUM_NONE;
+ new_link.link_duplex = RTE_ETH_LINK_FULL_DUPLEX;
+ new_link.link_autoneg = !(dev->data->dev_conf.link_speeds &
+ RTE_ETH_LINK_SPEED_FIXED);
+
+ return rte_eth_linkstatus_set(dev, &new_link);
+}
+
static const struct eth_dev_ops idpf_eth_dev_ops = {
.dev_configure = idpf_dev_configure,
+ .dev_start = idpf_dev_start,
+ .dev_stop = idpf_dev_stop,
.dev_close = idpf_dev_close,
.rx_queue_setup = idpf_rx_queue_setup,
.tx_queue_setup = idpf_tx_queue_setup,
.dev_infos_get = idpf_dev_info_get,
+ .link_update= idpf_dev_link_update,
};
static int
@@ -284,6 +305,40 @@ idpf_dev_configure(struct rte_eth_dev *dev)
return 0;
}
+static int
+idpf_dev_start(struct rte_eth_dev *dev)
+{
+ struct idpf_vport *vport = dev->data->dev_private;
+
+ if (dev->data->mtu > vport->max_mtu) {
+ PMD_DRV_LOG(ERR, "MTU should be less than %d", vport->max_mtu);
+ return -1;
+ }
+
+ vport->max_pkt_len = dev->data->mtu + IDPF_ETH_OVERHEAD;
+
+ /* TODO: start queues */
+
+ if (idpf_vc_ena_dis_vport(vport, true) != 0) {
+ PMD_DRV_LOG(ERR, "Failed to enable vport");
+ return -1;
+ }
+
+ return 0;
+}
+
+static int
+idpf_dev_stop(struct rte_eth_dev *dev)
+{
+ struct idpf_vport *vport = dev->data->dev_private;
+
+ idpf_vc_ena_dis_vport(vport, false);
+
+ /* TODO: stop queues */
+
+ return 0;
+}
+
static int
idpf_dev_close(struct rte_eth_dev *dev)
{
diff --git a/drivers/net/idpf/idpf_rxtx.c b/drivers/net/idpf/idpf_rxtx.c
index 25dd5d85d5..3528d2f2c7 100644
--- a/drivers/net/idpf/idpf_rxtx.c
+++ b/drivers/net/idpf/idpf_rxtx.c
@@ -334,6 +334,11 @@ idpf_rx_split_queue_setup(struct rte_eth_dev *dev,
uint16_t queue_idx,
if (check_rx_thresh(nb_desc, rx_free_thresh) != 0)
return -EINVAL;
+ if (rx_conf->rx_deferred_start) {
+ PMD_INIT_LOG(ERR, "Queue start is not supported currently.");
+ return -EINVAL;
+ }
+
/* Setup Rx description queue */
rxq = rte_zmalloc_socket("idpf rxq",
sizeof(struct idpf_rx_queue),
@@ -465,6 +470,11 @@ idpf_rx_single_queue_setup(struct rte_eth_dev *dev,
uint16_t queue_idx,
if (check_rx_thresh(nb_desc, rx_free_thresh) != 0)
return -EINVAL;
+ if (rx_conf->rx_deferred_start) {
+ PMD_INIT_LOG(ERR, "Queue start is not supported currently.");
+ return -EINVAL;
+ }
+
/* Setup Rx description queue */
rxq = rte_zmalloc_socket("idpf rxq",
sizeof(struct idpf_rx_queue),
@@ -569,6 +579,11 @@ idpf_tx_split_queue_setup(struct rte_eth_dev *dev,
uint16_t queue_idx,
if (check_tx_thresh(nb_desc, tx_rs_thresh, tx_free_thresh) != 0)
return -EINVAL;
+ if (tx_conf->tx_deferred_start) {
+ PMD_INIT_LOG(ERR, "Queue start is not supported currently.");
+ return -EINVAL;
+ }
+
/* Allocate the TX queue data structure. */
txq = rte_zmalloc_socket("idpf split txq",
sizeof(struct idpf_tx_queue),
@@ -691,6 +706,11 @@ idpf_tx_single_queue_setup(struct rte_eth_dev *dev,
uint16_t queue_idx,
if (check_tx_thresh(nb_desc, tx_rs_thresh, tx_free_thresh) != 0)
return -EINVAL;
+ if (tx_conf->tx_deferred_start) {
+ PMD_INIT_LOG(ERR, "Queue start is not supported currently.");
+ retur
[PATCH v14 06/18] net/idpf: add support for queue start
Add support for these device ops:
- rx_queue_start
- tx_queue_start
Signed-off-by: Beilei Xing
Signed-off-by: Xiaoyun Li
Signed-off-by: Junfeng Guo
---
drivers/net/idpf/idpf_ethdev.c | 40 ++-
drivers/net/idpf/idpf_ethdev.h | 9 +
drivers/net/idpf/idpf_rxtx.c | 237 +++--
drivers/net/idpf/idpf_rxtx.h | 6 +
drivers/net/idpf/idpf_vchnl.c | 447 +
5 files changed, 718 insertions(+), 21 deletions(-)
diff --git a/drivers/net/idpf/idpf_ethdev.c b/drivers/net/idpf/idpf_ethdev.c
index 3430d00e92..abbf519977 100644
--- a/drivers/net/idpf/idpf_ethdev.c
+++ b/drivers/net/idpf/idpf_ethdev.c
@@ -58,6 +58,8 @@ static const struct eth_dev_ops idpf_eth_dev_ops = {
.dev_start = idpf_dev_start,
.dev_stop = idpf_dev_stop,
.dev_close = idpf_dev_close,
+ .rx_queue_start = idpf_rx_queue_start,
+ .tx_queue_start = idpf_tx_queue_start,
.rx_queue_setup = idpf_rx_queue_setup,
.tx_queue_setup = idpf_tx_queue_setup,
.dev_infos_get = idpf_dev_info_get,
@@ -305,6 +307,39 @@ idpf_dev_configure(struct rte_eth_dev *dev)
return 0;
}
+static int
+idpf_start_queues(struct rte_eth_dev *dev)
+{
+ struct idpf_rx_queue *rxq;
+ struct idpf_tx_queue *txq;
+ int err = 0;
+ int i;
+
+ for (i = 0; i < dev->data->nb_tx_queues; i++) {
+ txq = dev->data->tx_queues[i];
+ if (txq == NULL || txq->tx_deferred_start)
+ continue;
+ err = idpf_tx_queue_start(dev, i);
+ if (err != 0) {
+ PMD_DRV_LOG(ERR, "Fail to start Tx queue %u", i);
+ return err;
+ }
+ }
+
+ for (i = 0; i < dev->data->nb_rx_queues; i++) {
+ rxq = dev->data->rx_queues[i];
+ if (rxq == NULL || rxq->rx_deferred_start)
+ continue;
+ err = idpf_rx_queue_start(dev, i);
+ if (err != 0) {
+ PMD_DRV_LOG(ERR, "Fail to start Rx queue %u", i);
+ return err;
+ }
+ }
+
+ return err;
+}
+
static int
idpf_dev_start(struct rte_eth_dev *dev)
{
@@ -317,7 +352,10 @@ idpf_dev_start(struct rte_eth_dev *dev)
vport->max_pkt_len = dev->data->mtu + IDPF_ETH_OVERHEAD;
- /* TODO: start queues */
+ if (idpf_start_queues(dev) != 0) {
+ PMD_DRV_LOG(ERR, "Failed to start queues");
+ return -1;
+ }
if (idpf_vc_ena_dis_vport(vport, true) != 0) {
PMD_DRV_LOG(ERR, "Failed to enable vport");
diff --git a/drivers/net/idpf/idpf_ethdev.h b/drivers/net/idpf/idpf_ethdev.h
index 84ae6641e2..96c22009e9 100644
--- a/drivers/net/idpf/idpf_ethdev.h
+++ b/drivers/net/idpf/idpf_ethdev.h
@@ -24,7 +24,9 @@
#define IDPF_DEFAULT_TXQ_NUM 16
#define IDPF_INVALID_VPORT_IDX 0x
+#define IDPF_TXQ_PER_GRP 1
#define IDPF_TX_COMPLQ_PER_GRP 1
+#define IDPF_RXQ_PER_GRP 1
#define IDPF_RX_BUFQ_PER_GRP 2
#define IDPF_CTLQ_ID -1
@@ -182,6 +184,13 @@ int idpf_vc_check_api_version(struct idpf_adapter
*adapter);
int idpf_vc_get_caps(struct idpf_adapter *adapter);
int idpf_vc_create_vport(struct idpf_adapter *adapter);
int idpf_vc_destroy_vport(struct idpf_vport *vport);
+int idpf_vc_config_rxqs(struct idpf_vport *vport);
+int idpf_vc_config_rxq(struct idpf_vport *vport, uint16_t rxq_id);
+int idpf_vc_config_txqs(struct idpf_vport *vport);
+int idpf_vc_config_txq(struct idpf_vport *vport, uint16_t txq_id);
+int idpf_switch_queue(struct idpf_vport *vport, uint16_t qid,
+ bool rx, bool on);
+int idpf_vc_ena_dis_queues(struct idpf_vport *vport, bool enable);
int idpf_vc_ena_dis_vport(struct idpf_vport *vport, bool enable);
int idpf_read_one_msg(struct idpf_adapter *adapter, uint32_t ops,
uint16_t buf_len, uint8_t *buf);
diff --git a/drivers/net/idpf/idpf_rxtx.c b/drivers/net/idpf/idpf_rxtx.c
index 3528d2f2c7..6d954afd9d 100644
--- a/drivers/net/idpf/idpf_rxtx.c
+++ b/drivers/net/idpf/idpf_rxtx.c
@@ -334,11 +334,6 @@ idpf_rx_split_queue_setup(struct rte_eth_dev *dev,
uint16_t queue_idx,
if (check_rx_thresh(nb_desc, rx_free_thresh) != 0)
return -EINVAL;
- if (rx_conf->rx_deferred_start) {
- PMD_INIT_LOG(ERR, "Queue start is not supported currently.");
- return -EINVAL;
- }
-
/* Setup Rx description queue */
rxq = rte_zmalloc_socket("idpf rxq",
sizeof(struct idpf_rx_queue),
@@ -354,6 +349,7 @@ idpf_rx_split_queue_setup(struct rte_eth_dev *dev, uint16_t
queue_idx,
rxq->rx_free_thresh = rx_free_thresh;
rxq->queue_id = vport->chunks_info.rx_start_qid + queue_idx;
rxq-
[PATCH v14 07/18] net/idpf: add support for queue stop
Add support for these device ops:
- rx_queue_stop
- tx_queue_stop
Signed-off-by: Beilei Xing
Signed-off-by: Xiaoyun Li
Signed-off-by: Junfeng Guo
---
doc/guides/nics/features/idpf.ini | 1 +
drivers/net/idpf/idpf_ethdev.c| 14 ++-
drivers/net/idpf/idpf_rxtx.c | 148 ++
drivers/net/idpf/idpf_rxtx.h | 13 +++
drivers/net/idpf/idpf_vchnl.c | 69 ++
5 files changed, 241 insertions(+), 4 deletions(-)
diff --git a/doc/guides/nics/features/idpf.ini
b/doc/guides/nics/features/idpf.ini
index 46aab2eb61..c25fa5de3f 100644
--- a/doc/guides/nics/features/idpf.ini
+++ b/doc/guides/nics/features/idpf.ini
@@ -4,6 +4,7 @@
; Refer to default.ini for the full list of available PMD features.
;
[Features]
+Queue start/stop = Y
Linux= Y
x86-32 = Y
x86-64 = Y
diff --git a/drivers/net/idpf/idpf_ethdev.c b/drivers/net/idpf/idpf_ethdev.c
index abbf519977..05f087a03c 100644
--- a/drivers/net/idpf/idpf_ethdev.c
+++ b/drivers/net/idpf/idpf_ethdev.c
@@ -59,7 +59,9 @@ static const struct eth_dev_ops idpf_eth_dev_ops = {
.dev_stop = idpf_dev_stop,
.dev_close = idpf_dev_close,
.rx_queue_start = idpf_rx_queue_start,
+ .rx_queue_stop = idpf_rx_queue_stop,
.tx_queue_start = idpf_tx_queue_start,
+ .tx_queue_stop = idpf_tx_queue_stop,
.rx_queue_setup = idpf_rx_queue_setup,
.tx_queue_setup = idpf_tx_queue_setup,
.dev_infos_get = idpf_dev_info_get,
@@ -347,22 +349,26 @@ idpf_dev_start(struct rte_eth_dev *dev)
if (dev->data->mtu > vport->max_mtu) {
PMD_DRV_LOG(ERR, "MTU should be less than %d", vport->max_mtu);
- return -1;
+ goto err_mtu;
}
vport->max_pkt_len = dev->data->mtu + IDPF_ETH_OVERHEAD;
if (idpf_start_queues(dev) != 0) {
PMD_DRV_LOG(ERR, "Failed to start queues");
- return -1;
+ goto err_mtu;
}
if (idpf_vc_ena_dis_vport(vport, true) != 0) {
PMD_DRV_LOG(ERR, "Failed to enable vport");
- return -1;
+ goto err_vport;
}
return 0;
+err_vport:
+ idpf_stop_queues(dev);
+err_mtu:
+ return -1;
}
static int
@@ -372,7 +378,7 @@ idpf_dev_stop(struct rte_eth_dev *dev)
idpf_vc_ena_dis_vport(vport, false);
- /* TODO: stop queues */
+ idpf_stop_queues(dev);
return 0;
}
diff --git a/drivers/net/idpf/idpf_rxtx.c b/drivers/net/idpf/idpf_rxtx.c
index 6d954afd9d..8d5ec41a1f 100644
--- a/drivers/net/idpf/idpf_rxtx.c
+++ b/drivers/net/idpf/idpf_rxtx.c
@@ -71,6 +71,55 @@ check_tx_thresh(uint16_t nb_desc, uint16_t tx_rs_thresh,
return 0;
}
+static void
+release_rxq_mbufs(struct idpf_rx_queue *rxq)
+{
+ uint16_t i;
+
+ if (rxq->sw_ring == NULL)
+ return;
+
+ for (i = 0; i < rxq->nb_rx_desc; i++) {
+ if (rxq->sw_ring[i] != NULL) {
+ rte_pktmbuf_free_seg(rxq->sw_ring[i]);
+ rxq->sw_ring[i] = NULL;
+ }
+ }
+}
+
+static void
+release_txq_mbufs(struct idpf_tx_queue *txq)
+{
+ uint16_t nb_desc, i;
+
+ if (txq == NULL || txq->sw_ring == NULL) {
+ PMD_DRV_LOG(DEBUG, "Pointer to rxq or sw_ring is NULL");
+ return;
+ }
+
+ if (txq->sw_nb_desc != 0) {
+ /* For split queue model, descriptor ring */
+ nb_desc = txq->sw_nb_desc;
+ } else {
+ /* For single queue model */
+ nb_desc = txq->nb_tx_desc;
+ }
+ for (i = 0; i < nb_desc; i++) {
+ if (txq->sw_ring[i].mbuf != NULL) {
+ rte_pktmbuf_free_seg(txq->sw_ring[i].mbuf);
+ txq->sw_ring[i].mbuf = NULL;
+ }
+ }
+}
+
+static const struct idpf_rxq_ops def_rxq_ops = {
+ .release_mbufs = release_rxq_mbufs,
+};
+
+static const struct idpf_txq_ops def_txq_ops = {
+ .release_mbufs = release_txq_mbufs,
+};
+
static void
reset_split_rx_descq(struct idpf_rx_queue *rxq)
{
@@ -122,6 +171,14 @@ reset_split_rx_bufq(struct idpf_rx_queue *rxq)
rxq->bufq2 = NULL;
}
+static inline void
+reset_split_rx_queue(struct idpf_rx_queue *rxq)
+{
+ reset_split_rx_descq(rxq);
+ reset_split_rx_bufq(rxq->bufq1);
+ reset_split_rx_bufq(rxq->bufq2);
+}
+
static void
reset_single_rx_queue(struct idpf_rx_queue *rxq)
{
@@ -301,6 +358,7 @@ idpf_rx_split_bufq_setup(struct rte_eth_dev *dev, struct
idpf_rx_queue *bufq,
bufq->q_set = true;
bufq->qrx_tail = hw->hw_addr + (vport->chunks_info.rx_buf_qtail_start +
queue_idx * vport->chunks_info.rx_buf_qtail_spacing);
+
[PATCH v14 08/18] net/idpf: add queue release
Add support for queue operations:
- rx_queue_release
- tx_queue_release
Signed-off-by: Beilei Xing
Signed-off-by: Xiaoyun Li
Signed-off-by: Junfeng Guo
---
drivers/net/idpf/idpf_ethdev.c | 2 +
drivers/net/idpf/idpf_rxtx.c | 81 ++
drivers/net/idpf/idpf_rxtx.h | 3 ++
3 files changed, 86 insertions(+)
diff --git a/drivers/net/idpf/idpf_ethdev.c b/drivers/net/idpf/idpf_ethdev.c
index 05f087a03c..dee1e03d43 100644
--- a/drivers/net/idpf/idpf_ethdev.c
+++ b/drivers/net/idpf/idpf_ethdev.c
@@ -63,7 +63,9 @@ static const struct eth_dev_ops idpf_eth_dev_ops = {
.tx_queue_start = idpf_tx_queue_start,
.tx_queue_stop = idpf_tx_queue_stop,
.rx_queue_setup = idpf_rx_queue_setup,
+ .rx_queue_release = idpf_dev_rx_queue_release,
.tx_queue_setup = idpf_tx_queue_setup,
+ .tx_queue_release = idpf_dev_tx_queue_release,
.dev_infos_get = idpf_dev_info_get,
.link_update= idpf_dev_link_update,
};
diff --git a/drivers/net/idpf/idpf_rxtx.c b/drivers/net/idpf/idpf_rxtx.c
index 8d5ec41a1f..053409b99a 100644
--- a/drivers/net/idpf/idpf_rxtx.c
+++ b/drivers/net/idpf/idpf_rxtx.c
@@ -171,6 +171,51 @@ reset_split_rx_bufq(struct idpf_rx_queue *rxq)
rxq->bufq2 = NULL;
}
+static void
+idpf_rx_queue_release(void *rxq)
+{
+ struct idpf_rx_queue *q = rxq;
+
+ if (q == NULL)
+ return;
+
+ /* Split queue */
+ if (q->bufq1 != NULL && q->bufq2 != NULL) {
+ q->bufq1->ops->release_mbufs(q->bufq1);
+ rte_free(q->bufq1->sw_ring);
+ rte_memzone_free(q->bufq1->mz);
+ rte_free(q->bufq1);
+ q->bufq2->ops->release_mbufs(q->bufq2);
+ rte_free(q->bufq2->sw_ring);
+ rte_memzone_free(q->bufq2->mz);
+ rte_free(q->bufq2);
+ rte_memzone_free(q->mz);
+ rte_free(q);
+ return;
+ }
+
+ /* Single queue */
+ q->ops->release_mbufs(q);
+ rte_free(q->sw_ring);
+ rte_memzone_free(q->mz);
+ rte_free(q);
+}
+
+static void
+idpf_tx_queue_release(void *txq)
+{
+ struct idpf_tx_queue *q = txq;
+
+ if (q == NULL)
+ return;
+
+ rte_free(q->complq);
+ q->ops->release_mbufs(q);
+ rte_free(q->sw_ring);
+ rte_memzone_free(q->mz);
+ rte_free(q);
+}
+
static inline void
reset_split_rx_queue(struct idpf_rx_queue *rxq)
{
@@ -392,6 +437,12 @@ idpf_rx_split_queue_setup(struct rte_eth_dev *dev,
uint16_t queue_idx,
if (check_rx_thresh(nb_desc, rx_free_thresh) != 0)
return -EINVAL;
+ /* Free memory if needed */
+ if (dev->data->rx_queues[queue_idx] != NULL) {
+ idpf_rx_queue_release(dev->data->rx_queues[queue_idx]);
+ dev->data->rx_queues[queue_idx] = NULL;
+ }
+
/* Setup Rx description queue */
rxq = rte_zmalloc_socket("idpf rxq",
sizeof(struct idpf_rx_queue),
@@ -524,6 +575,12 @@ idpf_rx_single_queue_setup(struct rte_eth_dev *dev,
uint16_t queue_idx,
if (check_rx_thresh(nb_desc, rx_free_thresh) != 0)
return -EINVAL;
+ /* Free memory if needed */
+ if (dev->data->rx_queues[queue_idx] != NULL) {
+ idpf_rx_queue_release(dev->data->rx_queues[queue_idx]);
+ dev->data->rx_queues[queue_idx] = NULL;
+ }
+
/* Setup Rx description queue */
rxq = rte_zmalloc_socket("idpf rxq",
sizeof(struct idpf_rx_queue),
@@ -630,6 +687,12 @@ idpf_tx_split_queue_setup(struct rte_eth_dev *dev,
uint16_t queue_idx,
if (check_tx_thresh(nb_desc, tx_rs_thresh, tx_free_thresh) != 0)
return -EINVAL;
+ /* Free memory if needed. */
+ if (dev->data->tx_queues[queue_idx] != NULL) {
+ idpf_tx_queue_release(dev->data->tx_queues[queue_idx]);
+ dev->data->tx_queues[queue_idx] = NULL;
+ }
+
/* Allocate the TX queue data structure. */
txq = rte_zmalloc_socket("idpf split txq",
sizeof(struct idpf_tx_queue),
@@ -754,6 +817,12 @@ idpf_tx_single_queue_setup(struct rte_eth_dev *dev,
uint16_t queue_idx,
if (check_tx_thresh(nb_desc, tx_rs_thresh, tx_free_thresh) != 0)
return -EINVAL;
+ /* Free memory if needed. */
+ if (dev->data->tx_queues[queue_idx] != NULL) {
+ idpf_tx_queue_release(dev->data->tx_queues[queue_idx]);
+ dev->data->tx_queues[queue_idx] = NULL;
+ }
+
/* Allocate the TX queue data structure. */
txq = rte_zmalloc_socket("idpf txq",
sizeof(struct idpf_tx_queue),
@@ -1102,6 +1171,18 @@ idpf_tx_queue_stop(struct rte_eth_dev *d
[PATCH v14 09/18] net/idpf: add support for MTU configuration
Add dev ops mtu_set.
Signed-off-by: Beilei Xing
Signed-off-by: Junfeng Guo
---
doc/guides/nics/features/idpf.ini | 1 +
drivers/net/idpf/idpf_ethdev.c| 14 ++
2 files changed, 15 insertions(+)
diff --git a/doc/guides/nics/features/idpf.ini
b/doc/guides/nics/features/idpf.ini
index c25fa5de3f..c8f4a0d6ed 100644
--- a/doc/guides/nics/features/idpf.ini
+++ b/doc/guides/nics/features/idpf.ini
@@ -5,6 +5,7 @@
;
[Features]
Queue start/stop = Y
+MTU update = Y
Linux= Y
x86-32 = Y
x86-64 = Y
diff --git a/drivers/net/idpf/idpf_ethdev.c b/drivers/net/idpf/idpf_ethdev.c
index dee1e03d43..94f4671057 100644
--- a/drivers/net/idpf/idpf_ethdev.c
+++ b/drivers/net/idpf/idpf_ethdev.c
@@ -35,6 +35,7 @@ static int idpf_dev_stop(struct rte_eth_dev *dev);
static int idpf_dev_close(struct rte_eth_dev *dev);
static int idpf_dev_info_get(struct rte_eth_dev *dev,
struct rte_eth_dev_info *dev_info);
+static int idpf_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
static void idpf_adapter_rel(struct idpf_adapter *adapter);
static int
@@ -68,6 +69,7 @@ static const struct eth_dev_ops idpf_eth_dev_ops = {
.tx_queue_release = idpf_dev_tx_queue_release,
.dev_infos_get = idpf_dev_info_get,
.link_update= idpf_dev_link_update,
+ .mtu_set= idpf_dev_mtu_set,
};
static int
@@ -110,6 +112,18 @@ idpf_dev_info_get(struct rte_eth_dev *dev, struct
rte_eth_dev_info *dev_info)
return 0;
}
+static int
+idpf_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu __rte_unused)
+{
+ /* mtu setting is forbidden if port is start */
+ if (dev->data->dev_started) {
+ PMD_DRV_LOG(ERR, "port must be stopped before configuration");
+ return -EBUSY;
+ }
+
+ return 0;
+}
+
static int
idpf_init_vport_req_info(struct rte_eth_dev *dev)
{
--
2.34.1
[PATCH v14 10/18] net/idpf: add support for basic Rx datapath
Add basic Rx support in split queue mode and single queue mode.
Signed-off-by: Beilei Xing
Signed-off-by: Xiaoyun Li
Signed-off-by: Junfeng Guo
---
drivers/net/idpf/idpf_ethdev.c | 2 +
drivers/net/idpf/idpf_rxtx.c | 273 +
drivers/net/idpf/idpf_rxtx.h | 7 +-
3 files changed, 281 insertions(+), 1 deletion(-)
diff --git a/drivers/net/idpf/idpf_ethdev.c b/drivers/net/idpf/idpf_ethdev.c
index 94f4671057..df2a760673 100644
--- a/drivers/net/idpf/idpf_ethdev.c
+++ b/drivers/net/idpf/idpf_ethdev.c
@@ -375,6 +375,8 @@ idpf_dev_start(struct rte_eth_dev *dev)
goto err_mtu;
}
+ idpf_set_rx_function(dev);
+
if (idpf_vc_ena_dis_vport(vport, true) != 0) {
PMD_DRV_LOG(ERR, "Failed to enable vport");
goto err_vport;
diff --git a/drivers/net/idpf/idpf_rxtx.c b/drivers/net/idpf/idpf_rxtx.c
index 053409b99a..ea499c4d37 100644
--- a/drivers/net/idpf/idpf_rxtx.c
+++ b/drivers/net/idpf/idpf_rxtx.c
@@ -1208,3 +1208,276 @@ idpf_stop_queues(struct rte_eth_dev *dev)
PMD_DRV_LOG(WARNING, "Fail to stop Tx queue %d", i);
}
}
+
+static void
+idpf_split_rx_bufq_refill(struct idpf_rx_queue *rx_bufq)
+{
+ volatile struct virtchnl2_splitq_rx_buf_desc *rx_buf_ring;
+ volatile struct virtchnl2_splitq_rx_buf_desc *rx_buf_desc;
+ uint16_t nb_refill = rx_bufq->rx_free_thresh;
+ uint16_t nb_desc = rx_bufq->nb_rx_desc;
+ uint16_t next_avail = rx_bufq->rx_tail;
+ struct rte_mbuf *nmb[rx_bufq->rx_free_thresh];
+ struct rte_eth_dev *dev;
+ uint64_t dma_addr;
+ uint16_t delta;
+ int i;
+
+ if (rx_bufq->nb_rx_hold < rx_bufq->rx_free_thresh)
+ return;
+
+ rx_buf_ring = rx_bufq->rx_ring;
+ delta = nb_desc - next_avail;
+ if (unlikely(delta < nb_refill)) {
+ if (likely(rte_pktmbuf_alloc_bulk(rx_bufq->mp, nmb, delta) ==
0)) {
+ for (i = 0; i < delta; i++) {
+ rx_buf_desc = &rx_buf_ring[next_avail + i];
+ rx_bufq->sw_ring[next_avail + i] = nmb[i];
+ dma_addr =
rte_cpu_to_le_64(rte_mbuf_data_iova_default(nmb[i]));
+ rx_buf_desc->hdr_addr = 0;
+ rx_buf_desc->pkt_addr = dma_addr;
+ }
+ nb_refill -= delta;
+ next_avail = 0;
+ rx_bufq->nb_rx_hold -= delta;
+ } else {
+ dev = &rte_eth_devices[rx_bufq->port_id];
+ dev->data->rx_mbuf_alloc_failed += nb_desc - next_avail;
+ PMD_RX_LOG(DEBUG, "RX mbuf alloc failed port_id=%u
queue_id=%u",
+ rx_bufq->port_id, rx_bufq->queue_id);
+ return;
+ }
+ }
+
+ if (nb_desc - next_avail >= nb_refill) {
+ if (likely(rte_pktmbuf_alloc_bulk(rx_bufq->mp, nmb, nb_refill)
== 0)) {
+ for (i = 0; i < nb_refill; i++) {
+ rx_buf_desc = &rx_buf_ring[next_avail + i];
+ rx_bufq->sw_ring[next_avail + i] = nmb[i];
+ dma_addr =
rte_cpu_to_le_64(rte_mbuf_data_iova_default(nmb[i]));
+ rx_buf_desc->hdr_addr = 0;
+ rx_buf_desc->pkt_addr = dma_addr;
+ }
+ next_avail += nb_refill;
+ rx_bufq->nb_rx_hold -= nb_refill;
+ } else {
+ dev = &rte_eth_devices[rx_bufq->port_id];
+ dev->data->rx_mbuf_alloc_failed += nb_desc - next_avail;
+ PMD_RX_LOG(DEBUG, "RX mbuf alloc failed port_id=%u
queue_id=%u",
+ rx_bufq->port_id, rx_bufq->queue_id);
+ }
+ }
+
+ IDPF_PCI_REG_WRITE(rx_bufq->qrx_tail, next_avail);
+
+ rx_bufq->rx_tail = next_avail;
+}
+
+uint16_t
+idpf_splitq_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts,
+ uint16_t nb_pkts)
+{
+ volatile struct virtchnl2_rx_flex_desc_adv_nic_3 *rx_desc_ring;
+ volatile struct virtchnl2_rx_flex_desc_adv_nic_3 *rx_desc;
+ uint16_t pktlen_gen_bufq_id;
+ struct idpf_rx_queue *rxq;
+ struct rte_mbuf *rxm;
+ uint16_t rx_id_bufq1;
+ uint16_t rx_id_bufq2;
+ uint16_t pkt_len;
+ uint16_t bufq_id;
+ uint16_t gen_id;
+ uint16_t rx_id;
+ uint16_t nb_rx;
+
+ nb_rx = 0;
+ rxq = rx_queue;
+
+ if (unlikely(rxq == NULL) || unlikely(!rxq->q_started))
+ return nb_rx;
+
+ rx_id = rxq->rx_tail;
+ rx_id_bufq1 = rxq->bufq1->rx_next_avail;
+ rx_id_bufq2 = rxq->bufq2->rx_next_avail;
+ rx_desc_ring = rxq->rx_ring;
+
+ wh
[PATCH v14 11/18] net/idpf: add support for basic Tx datapath
Add basic Tx support in split queue mode and single queue mode.
Signed-off-by: Beilei Xing
Signed-off-by: Xiaoyun Li
Signed-off-by: Junfeng Guo
---
drivers/net/idpf/idpf_ethdev.c | 3 +
drivers/net/idpf/idpf_ethdev.h | 1 +
drivers/net/idpf/idpf_rxtx.c | 357 +
drivers/net/idpf/idpf_rxtx.h | 10 +
4 files changed, 371 insertions(+)
diff --git a/drivers/net/idpf/idpf_ethdev.c b/drivers/net/idpf/idpf_ethdev.c
index df2a760673..6fb56e584d 100644
--- a/drivers/net/idpf/idpf_ethdev.c
+++ b/drivers/net/idpf/idpf_ethdev.c
@@ -88,6 +88,8 @@ idpf_dev_info_get(struct rte_eth_dev *dev, struct
rte_eth_dev_info *dev_info)
dev_info->max_mac_addrs = IDPF_NUM_MACADDR_MAX;
+ dev_info->tx_offload_capa = RTE_ETH_TX_OFFLOAD_MULTI_SEGS;
+
dev_info->default_txconf = (struct rte_eth_txconf) {
.tx_free_thresh = IDPF_DEFAULT_TX_FREE_THRESH,
.tx_rs_thresh = IDPF_DEFAULT_TX_RS_THRESH,
@@ -376,6 +378,7 @@ idpf_dev_start(struct rte_eth_dev *dev)
}
idpf_set_rx_function(dev);
+ idpf_set_tx_function(dev);
if (idpf_vc_ena_dis_vport(vport, true) != 0) {
PMD_DRV_LOG(ERR, "Failed to enable vport");
diff --git a/drivers/net/idpf/idpf_ethdev.h b/drivers/net/idpf/idpf_ethdev.h
index 96c22009e9..af0a8e2970 100644
--- a/drivers/net/idpf/idpf_ethdev.h
+++ b/drivers/net/idpf/idpf_ethdev.h
@@ -35,6 +35,7 @@
#define IDPF_MIN_BUF_SIZE 1024
#define IDPF_MAX_FRAME_SIZE9728
+#define IDPF_MIN_FRAME_SIZE14
#define IDPF_NUM_MACADDR_MAX 64
diff --git a/drivers/net/idpf/idpf_rxtx.c b/drivers/net/idpf/idpf_rxtx.c
index ea499c4d37..f55d2143b9 100644
--- a/drivers/net/idpf/idpf_rxtx.c
+++ b/drivers/net/idpf/idpf_rxtx.c
@@ -1365,6 +1365,148 @@ idpf_splitq_recv_pkts(void *rx_queue, struct rte_mbuf
**rx_pkts,
return nb_rx;
}
+static inline void
+idpf_split_tx_free(struct idpf_tx_queue *cq)
+{
+ volatile struct idpf_splitq_tx_compl_desc *compl_ring = cq->compl_ring;
+ volatile struct idpf_splitq_tx_compl_desc *txd;
+ uint16_t next = cq->tx_tail;
+ struct idpf_tx_entry *txe;
+ struct idpf_tx_queue *txq;
+ uint16_t gen, qid, q_head;
+ uint8_t ctype;
+
+ txd = &compl_ring[next];
+ gen = (rte_le_to_cpu_16(txd->qid_comptype_gen) &
+ IDPF_TXD_COMPLQ_GEN_M) >> IDPF_TXD_COMPLQ_GEN_S;
+ if (gen != cq->expected_gen_id)
+ return;
+
+ ctype = (rte_le_to_cpu_16(txd->qid_comptype_gen) &
+ IDPF_TXD_COMPLQ_COMPL_TYPE_M) >> IDPF_TXD_COMPLQ_COMPL_TYPE_S;
+ qid = (rte_le_to_cpu_16(txd->qid_comptype_gen) &
+ IDPF_TXD_COMPLQ_QID_M) >> IDPF_TXD_COMPLQ_QID_S;
+ q_head = rte_le_to_cpu_16(txd->q_head_compl_tag.compl_tag);
+ txq = cq->txqs[qid - cq->tx_start_qid];
+
+ switch (ctype) {
+ case IDPF_TXD_COMPLT_RE:
+ if (q_head == 0)
+ txq->last_desc_cleaned = txq->nb_tx_desc - 1;
+ else
+ txq->last_desc_cleaned = q_head - 1;
+ if (unlikely((txq->last_desc_cleaned % 32) == 0)) {
+ PMD_DRV_LOG(ERR, "unexpected desc (head = %u)
completion.",
+ q_head);
+ return;
+ }
+
+ break;
+ case IDPF_TXD_COMPLT_RS:
+ txq->nb_free++;
+ txq->nb_used--;
+ txe = &txq->sw_ring[q_head];
+ if (txe->mbuf != NULL) {
+ rte_pktmbuf_free_seg(txe->mbuf);
+ txe->mbuf = NULL;
+ }
+ break;
+ default:
+ PMD_DRV_LOG(ERR, "unknown completion type.");
+ return;
+ }
+
+ if (++next == cq->nb_tx_desc) {
+ next = 0;
+ cq->expected_gen_id ^= 1;
+ }
+
+ cq->tx_tail = next;
+}
+
+uint16_t
+idpf_splitq_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts,
+ uint16_t nb_pkts)
+{
+ struct idpf_tx_queue *txq = (struct idpf_tx_queue *)tx_queue;
+ volatile struct idpf_flex_tx_sched_desc *txr;
+ volatile struct idpf_flex_tx_sched_desc *txd;
+ struct idpf_tx_entry *sw_ring;
+ struct idpf_tx_entry *txe, *txn;
+ uint16_t nb_used, tx_id, sw_id;
+ struct rte_mbuf *tx_pkt;
+ uint16_t nb_to_clean;
+ uint16_t nb_tx = 0;
+
+ if (unlikely(txq == NULL) || unlikely(!txq->q_started))
+ return nb_tx;
+
+ txr = txq->desc_ring;
+ sw_ring = txq->sw_ring;
+ tx_id = txq->tx_tail;
+ sw_id = txq->sw_tail;
+ txe = &sw_ring[sw_id];
+
+ for (nb_tx = 0; nb_tx < nb_pkts; nb_tx++) {
+ tx_pkt = tx_pkts[nb_tx];
+
+ if (txq->nb_free <= txq->free_thresh) {
+ /* TODO: Need to refine
+* 1. free and clean: Better to decide
[PATCH v14 12/18] net/idpf: support parsing packet type
Parse packet type during receiving packets.
Signed-off-by: Wenjun Wu
Signed-off-by: Junfeng Guo
---
drivers/net/idpf/idpf_ethdev.c | 6 +
drivers/net/idpf/idpf_ethdev.h | 6 +
drivers/net/idpf/idpf_rxtx.c | 11 ++
drivers/net/idpf/idpf_rxtx.h | 5 +
drivers/net/idpf/idpf_vchnl.c | 240 +
5 files changed, 268 insertions(+)
diff --git a/drivers/net/idpf/idpf_ethdev.c b/drivers/net/idpf/idpf_ethdev.c
index 6fb56e584d..630bdabcd4 100644
--- a/drivers/net/idpf/idpf_ethdev.c
+++ b/drivers/net/idpf/idpf_ethdev.c
@@ -709,6 +709,12 @@ idpf_adapter_init(struct rte_pci_device *pci_dev, struct
idpf_adapter *adapter)
goto err_api;
}
+ ret = idpf_get_pkt_type(adapter);
+ if (ret != 0) {
+ PMD_INIT_LOG(ERR, "Failed to set ptype table");
+ goto err_api;
+ }
+
adapter->caps = rte_zmalloc("idpf_caps",
sizeof(struct virtchnl2_get_capabilities), 0);
if (adapter->caps == NULL) {
diff --git a/drivers/net/idpf/idpf_ethdev.h b/drivers/net/idpf/idpf_ethdev.h
index af0a8e2970..db9af58f72 100644
--- a/drivers/net/idpf/idpf_ethdev.h
+++ b/drivers/net/idpf/idpf_ethdev.h
@@ -39,6 +39,8 @@
#define IDPF_NUM_MACADDR_MAX 64
+#define IDPF_MAX_PKT_TYPE 1024
+
#define IDPF_VLAN_TAG_SIZE 4
#define IDPF_ETH_OVERHEAD \
(RTE_ETHER_HDR_LEN + RTE_ETHER_CRC_LEN + IDPF_VLAN_TAG_SIZE * 2)
@@ -125,6 +127,8 @@ struct idpf_adapter {
/* Max config queue number per VC message */
uint32_t max_rxq_per_msg;
uint32_t max_txq_per_msg;
+
+ uint32_t ptype_tbl[IDPF_MAX_PKT_TYPE] __rte_cache_min_aligned;
};
TAILQ_HEAD(idpf_adapter_list, idpf_adapter);
@@ -182,6 +186,7 @@ atomic_set_cmd(struct idpf_adapter *adapter, enum
virtchnl_ops ops)
struct idpf_adapter *idpf_find_adapter(struct rte_pci_device *pci_dev);
void idpf_handle_virtchnl_msg(struct rte_eth_dev *dev);
int idpf_vc_check_api_version(struct idpf_adapter *adapter);
+int idpf_get_pkt_type(struct idpf_adapter *adapter);
int idpf_vc_get_caps(struct idpf_adapter *adapter);
int idpf_vc_create_vport(struct idpf_adapter *adapter);
int idpf_vc_destroy_vport(struct idpf_vport *vport);
@@ -193,6 +198,7 @@ int idpf_switch_queue(struct idpf_vport *vport, uint16_t
qid,
bool rx, bool on);
int idpf_vc_ena_dis_queues(struct idpf_vport *vport, bool enable);
int idpf_vc_ena_dis_vport(struct idpf_vport *vport, bool enable);
+int idpf_vc_query_ptype_info(struct idpf_adapter *adapter);
int idpf_read_one_msg(struct idpf_adapter *adapter, uint32_t ops,
uint16_t buf_len, uint8_t *buf);
diff --git a/drivers/net/idpf/idpf_rxtx.c b/drivers/net/idpf/idpf_rxtx.c
index f55d2143b9..a980714060 100644
--- a/drivers/net/idpf/idpf_rxtx.c
+++ b/drivers/net/idpf/idpf_rxtx.c
@@ -1281,6 +1281,7 @@ idpf_splitq_recv_pkts(void *rx_queue, struct rte_mbuf
**rx_pkts,
volatile struct virtchnl2_rx_flex_desc_adv_nic_3 *rx_desc;
uint16_t pktlen_gen_bufq_id;
struct idpf_rx_queue *rxq;
+ const uint32_t *ptype_tbl;
struct rte_mbuf *rxm;
uint16_t rx_id_bufq1;
uint16_t rx_id_bufq2;
@@ -1300,6 +1301,7 @@ idpf_splitq_recv_pkts(void *rx_queue, struct rte_mbuf
**rx_pkts,
rx_id_bufq1 = rxq->bufq1->rx_next_avail;
rx_id_bufq2 = rxq->bufq2->rx_next_avail;
rx_desc_ring = rxq->rx_ring;
+ ptype_tbl = rxq->adapter->ptype_tbl;
while (nb_rx < nb_pkts) {
rx_desc = &rx_desc_ring[rx_id];
@@ -1347,6 +1349,10 @@ idpf_splitq_recv_pkts(void *rx_queue, struct rte_mbuf
**rx_pkts,
rxm->next = NULL;
rxm->nb_segs = 1;
rxm->port = rxq->port_id;
+ rxm->packet_type =
+ ptype_tbl[(rte_le_to_cpu_16(rx_desc->ptype_err_fflags0)
&
+ VIRTCHNL2_RX_FLEX_DESC_ADV_PTYPE_M) >>
+ VIRTCHNL2_RX_FLEX_DESC_ADV_PTYPE_S];
rx_pkts[nb_rx++] = rxm;
}
@@ -1533,6 +1539,7 @@ idpf_singleq_recv_pkts(void *rx_queue, struct rte_mbuf
**rx_pkts,
volatile union virtchnl2_rx_desc *rxdp;
union virtchnl2_rx_desc rxd;
struct idpf_rx_queue *rxq;
+ const uint32_t *ptype_tbl;
uint16_t rx_id, nb_hold;
struct rte_eth_dev *dev;
uint16_t rx_packet_len;
@@ -1551,6 +1558,7 @@ idpf_singleq_recv_pkts(void *rx_queue, struct rte_mbuf
**rx_pkts,
rx_id = rxq->rx_tail;
rx_ring = rxq->rx_ring;
+ ptype_tbl = rxq->adapter->ptype_tbl;
while (nb_rx < nb_pkts) {
rxdp = &rx_ring[rx_id];
@@ -1603,6 +1611,9 @@ idpf_singleq_recv_pkts(void *rx_queue, struct rte_mbuf
**rx_pkts,
rxm->pkt_len = rx_packet_len;
rxm->data_len = rx_packet_len;
rxm->port = rxq->port_id;
+ rxm->packet_type =
+
ptype_tbl[(
[PATCH v14 13/18] net/idpf: add support for write back based on ITR expire
Enable write back on ITR expire, then packets can be received one by
one.
Signed-off-by: Beilei Xing
Signed-off-by: Junfeng Guo
---
drivers/net/idpf/idpf_ethdev.c | 116 +
drivers/net/idpf/idpf_ethdev.h | 13
drivers/net/idpf/idpf_vchnl.c | 111 +++
3 files changed, 240 insertions(+)
diff --git a/drivers/net/idpf/idpf_ethdev.c b/drivers/net/idpf/idpf_ethdev.c
index 630bdabcd4..8045073780 100644
--- a/drivers/net/idpf/idpf_ethdev.c
+++ b/drivers/net/idpf/idpf_ethdev.c
@@ -327,6 +327,90 @@ idpf_dev_configure(struct rte_eth_dev *dev)
return 0;
}
+static int
+idpf_config_rx_queues_irqs(struct rte_eth_dev *dev)
+{
+ struct idpf_vport *vport = dev->data->dev_private;
+ struct idpf_adapter *adapter = vport->adapter;
+ struct virtchnl2_queue_vector *qv_map;
+ struct idpf_hw *hw = &adapter->hw;
+ uint32_t dynctl_reg_start;
+ uint32_t itrn_reg_start;
+ uint32_t dynctl_val, itrn_val;
+ uint16_t i;
+
+ qv_map = rte_zmalloc("qv_map",
+ dev->data->nb_rx_queues *
+ sizeof(struct virtchnl2_queue_vector), 0);
+ if (qv_map == NULL) {
+ PMD_DRV_LOG(ERR, "Failed to allocate %d queue-vector map",
+ dev->data->nb_rx_queues);
+ goto qv_map_alloc_err;
+ }
+
+ /* Rx interrupt disabled, Map interrupt only for writeback */
+
+ /* The capability flags adapter->caps->other_caps here should be
+* compared with bit VIRTCHNL2_CAP_WB_ON_ITR. The if condition should
+* be updated when the FW can return correct flag bits.
+*/
+ if (adapter->caps->other_caps != 0) {
+ dynctl_reg_start =
+ vport->recv_vectors->vchunks.vchunks->dynctl_reg_start;
+ itrn_reg_start =
+ vport->recv_vectors->vchunks.vchunks->itrn_reg_start;
+ dynctl_val = IDPF_READ_REG(hw, dynctl_reg_start);
+ PMD_DRV_LOG(DEBUG, "Value of dynctl_reg_start is 0x%x",
+ dynctl_val);
+ itrn_val = IDPF_READ_REG(hw, itrn_reg_start);
+ PMD_DRV_LOG(DEBUG, "Value of itrn_reg_start is 0x%x", itrn_val);
+ /* Force write-backs by setting WB_ON_ITR bit in DYN_CTL
+* register. WB_ON_ITR and INTENA are mutually exclusive
+* bits. Setting WB_ON_ITR bits means TX and RX Descs
+* are written back based on ITR expiration irrespective
+* of INTENA setting.
+*/
+ /* TBD: need to tune INTERVAL value for better performance. */
+ if (itrn_val != 0)
+ IDPF_WRITE_REG(hw,
+ dynctl_reg_start,
+ VIRTCHNL2_ITR_IDX_0 <<
+ PF_GLINT_DYN_CTL_ITR_INDX_S |
+ PF_GLINT_DYN_CTL_WB_ON_ITR_M |
+ itrn_val <<
+ PF_GLINT_DYN_CTL_INTERVAL_S);
+ else
+ IDPF_WRITE_REG(hw,
+ dynctl_reg_start,
+ VIRTCHNL2_ITR_IDX_0 <<
+ PF_GLINT_DYN_CTL_ITR_INDX_S |
+ PF_GLINT_DYN_CTL_WB_ON_ITR_M |
+ IDPF_DFLT_INTERVAL <<
+ PF_GLINT_DYN_CTL_INTERVAL_S);
+ }
+ for (i = 0; i < dev->data->nb_rx_queues; i++) {
+ /* map all queues to the same vector */
+ qv_map[i].queue_id = vport->chunks_info.rx_start_qid + i;
+ qv_map[i].vector_id =
+ vport->recv_vectors->vchunks.vchunks->start_vector_id;
+ }
+ vport->qv_map = qv_map;
+
+ if (idpf_vc_config_irq_map_unmap(vport, true) != 0) {
+ PMD_DRV_LOG(ERR, "config interrupt mapping failed");
+ goto config_irq_map_err;
+ }
+
+ return 0;
+
+config_irq_map_err:
+ rte_free(vport->qv_map);
+ vport->qv_map = NULL;
+
+qv_map_alloc_err:
+ return -1;
+}
+
static int
idpf_start_queues(struct rte_eth_dev *dev)
{
@@ -364,6 +448,10 @@ static int
idpf_dev_start(struct rte_eth_dev *dev)
{
struct idpf_vport *vport = dev->data->dev_private;
+ struct idpf_adapter *adapter = vport->adapter;
+ uint16_t num_allocated_vectors =
+ adapter->caps->num_allocated_vectors;
+ uint16_t req_vecs_num;
if (dev->data->mtu > vport->max_mtu) {
PMD_DRV_LOG(ERR, "MTU should be less than %d", vport->max_mtu);
@@ -372,6 +460,23 @@ idpf_dev_start(struct rte_eth_dev *dev)
vport->max_pkt_len = dev->data->mtu + IDPF_ETH_OVERHEAD;
+ req_vecs_num = IDPF_DFLT_Q_VEC_NUM;
+
[PATCH v14 14/18] net/idpf: add support for RSS
Add RSS support.
Signed-off-by: Beilei Xing
Signed-off-by: Junfeng Guo
---
drivers/net/idpf/idpf_ethdev.c | 116 +
drivers/net/idpf/idpf_ethdev.h | 26
drivers/net/idpf/idpf_vchnl.c | 97 +++
3 files changed, 239 insertions(+)
diff --git a/drivers/net/idpf/idpf_ethdev.c b/drivers/net/idpf/idpf_ethdev.c
index 8045073780..ea7d3bfd7e 100644
--- a/drivers/net/idpf/idpf_ethdev.c
+++ b/drivers/net/idpf/idpf_ethdev.c
@@ -86,6 +86,7 @@ idpf_dev_info_get(struct rte_eth_dev *dev, struct
rte_eth_dev_info *dev_info)
dev_info->max_mtu = dev_info->max_rx_pktlen - IDPF_ETH_OVERHEAD;
dev_info->min_mtu = RTE_ETHER_MIN_MTU;
+ dev_info->flow_type_rss_offloads = IDPF_RSS_OFFLOAD_ALL;
dev_info->max_mac_addrs = IDPF_NUM_MACADDR_MAX;
dev_info->tx_offload_capa = RTE_ETH_TX_OFFLOAD_MULTI_SEGS;
@@ -198,6 +199,8 @@ idpf_parse_devarg_id(char *name)
return val;
}
+#define IDPF_RSS_KEY_LEN 52
+
static int
idpf_init_vport(struct rte_eth_dev *dev)
{
@@ -218,6 +221,10 @@ idpf_init_vport(struct rte_eth_dev *dev)
vport->max_mtu = vport_info->max_mtu;
rte_memcpy(vport->default_mac_addr,
vport_info->default_mac_addr, ETH_ALEN);
+ vport->rss_algorithm = vport_info->rss_algorithm;
+ vport->rss_key_size = RTE_MIN(IDPF_RSS_KEY_LEN,
+vport_info->rss_key_size);
+ vport->rss_lut_size = vport_info->rss_lut_size;
vport->sw_idx = idx;
for (i = 0; i < vport_info->chunks.num_chunks; i++) {
@@ -275,10 +282,102 @@ idpf_init_vport(struct rte_eth_dev *dev)
return 0;
}
+static int
+idpf_config_rss(struct idpf_vport *vport)
+{
+ int ret;
+
+ ret = idpf_vc_set_rss_key(vport);
+ if (ret != 0) {
+ PMD_INIT_LOG(ERR, "Failed to configure RSS key");
+ return ret;
+ }
+
+ ret = idpf_vc_set_rss_lut(vport);
+ if (ret != 0) {
+ PMD_INIT_LOG(ERR, "Failed to configure RSS lut");
+ return ret;
+ }
+
+ ret = idpf_vc_set_rss_hash(vport);
+ if (ret != 0) {
+ PMD_INIT_LOG(ERR, "Failed to configure RSS hash");
+ return ret;
+ }
+
+ return ret;
+}
+
+static int
+idpf_init_rss(struct idpf_vport *vport)
+{
+ struct rte_eth_rss_conf *rss_conf;
+ uint16_t i, nb_q, lut_size;
+ int ret = 0;
+
+ rss_conf = &vport->dev_data->dev_conf.rx_adv_conf.rss_conf;
+ nb_q = vport->dev_data->nb_rx_queues;
+
+ vport->rss_key = rte_zmalloc("rss_key",
+vport->rss_key_size, 0);
+ if (vport->rss_key == NULL) {
+ PMD_INIT_LOG(ERR, "Failed to allocate RSS key");
+ ret = -ENOMEM;
+ goto err_alloc_key;
+ }
+
+ lut_size = vport->rss_lut_size;
+ vport->rss_lut = rte_zmalloc("rss_lut",
+sizeof(uint32_t) * lut_size, 0);
+ if (vport->rss_lut == NULL) {
+ PMD_INIT_LOG(ERR, "Failed to allocate RSS lut");
+ ret = -ENOMEM;
+ goto err_alloc_lut;
+ }
+
+ if (rss_conf->rss_key == NULL) {
+ for (i = 0; i < vport->rss_key_size; i++)
+ vport->rss_key[i] = (uint8_t)rte_rand();
+ } else if (rss_conf->rss_key_len != vport->rss_key_size) {
+ PMD_INIT_LOG(ERR, "Invalid RSS key length in RSS configuration,
should be %d",
+vport->rss_key_size);
+ ret = -EINVAL;
+ goto err_cfg_key;
+ } else {
+ rte_memcpy(vport->rss_key, rss_conf->rss_key,
+ vport->rss_key_size);
+ }
+
+ for (i = 0; i < lut_size; i++)
+ vport->rss_lut[i] = i % nb_q;
+
+ vport->rss_hf = IDPF_DEFAULT_RSS_HASH_EXPANDED;
+
+ ret = idpf_config_rss(vport);
+ if (ret != 0) {
+ PMD_INIT_LOG(ERR, "Failed to configure RSS");
+ goto err_cfg_key;
+ }
+
+ return ret;
+
+err_cfg_key:
+ rte_free(vport->rss_lut);
+ vport->rss_lut = NULL;
+err_alloc_lut:
+ rte_free(vport->rss_key);
+ vport->rss_key = NULL;
+err_alloc_key:
+ return ret;
+}
+
static int
idpf_dev_configure(struct rte_eth_dev *dev)
{
+ struct idpf_vport *vport = dev->data->dev_private;
struct rte_eth_conf *conf = &dev->data->dev_conf;
+ struct idpf_adapter *adapter = vport->adapter;
+ int ret;
if (conf->link_speeds & RTE_ETH_LINK_SPEED_FIXED) {
PMD_INIT_LOG(ERR, "Setting link speed is not supported");
@@ -324,6 +423,17 @@ idpf_dev_configure(struct rte_eth_dev *dev)
return -1;
}
+ if (adapter->caps->rss_caps != 0 && dev->data->nb_rx_queues != 0) {
+ ret = idpf_init_rss(vport);
+ if (ret != 0) {
+
[PATCH v14 15/18] net/idpf: add support for Rx offloading
Add Rx offloading support:
- support CHKSUM and RSS offload for split queue model
- support CHKSUM offload for single queue model
Signed-off-by: Beilei Xing
Signed-off-by: Xiaoyun Li
Signed-off-by: Junfeng Guo
---
doc/guides/nics/features/idpf.ini | 5 ++
drivers/net/idpf/idpf_ethdev.c| 6 ++
drivers/net/idpf/idpf_rxtx.c | 123 ++
3 files changed, 134 insertions(+)
diff --git a/doc/guides/nics/features/idpf.ini
b/doc/guides/nics/features/idpf.ini
index c8f4a0d6ed..6eefac9529 100644
--- a/doc/guides/nics/features/idpf.ini
+++ b/doc/guides/nics/features/idpf.ini
@@ -3,9 +3,14 @@
;
; Refer to default.ini for the full list of available PMD features.
;
+; A feature with "P" indicates only be supported when non-vector path
+; is selected.
+;
[Features]
Queue start/stop = Y
MTU update = Y
+L3 checksum offload = P
+L4 checksum offload = P
Linux= Y
x86-32 = Y
x86-64 = Y
diff --git a/drivers/net/idpf/idpf_ethdev.c b/drivers/net/idpf/idpf_ethdev.c
index ea7d3bfd7e..88f3db3267 100644
--- a/drivers/net/idpf/idpf_ethdev.c
+++ b/drivers/net/idpf/idpf_ethdev.c
@@ -89,6 +89,12 @@ idpf_dev_info_get(struct rte_eth_dev *dev, struct
rte_eth_dev_info *dev_info)
dev_info->flow_type_rss_offloads = IDPF_RSS_OFFLOAD_ALL;
dev_info->max_mac_addrs = IDPF_NUM_MACADDR_MAX;
+ dev_info->rx_offload_capa =
+ RTE_ETH_RX_OFFLOAD_IPV4_CKSUM |
+ RTE_ETH_RX_OFFLOAD_UDP_CKSUM|
+ RTE_ETH_RX_OFFLOAD_TCP_CKSUM|
+ RTE_ETH_RX_OFFLOAD_OUTER_IPV4_CKSUM;
+
dev_info->tx_offload_capa = RTE_ETH_TX_OFFLOAD_MULTI_SEGS;
dev_info->default_txconf = (struct rte_eth_txconf) {
diff --git a/drivers/net/idpf/idpf_rxtx.c b/drivers/net/idpf/idpf_rxtx.c
index a980714060..f15e61a785 100644
--- a/drivers/net/idpf/idpf_rxtx.c
+++ b/drivers/net/idpf/idpf_rxtx.c
@@ -1209,6 +1209,73 @@ idpf_stop_queues(struct rte_eth_dev *dev)
}
}
+#define IDPF_RX_FLEX_DESC_ADV_STATUS0_XSUM_S \
+ (RTE_BIT32(VIRTCHNL2_RX_FLEX_DESC_ADV_STATUS0_XSUM_IPE_S) | \
+RTE_BIT32(VIRTCHNL2_RX_FLEX_DESC_ADV_STATUS0_XSUM_L4E_S) | \
+RTE_BIT32(VIRTCHNL2_RX_FLEX_DESC_ADV_STATUS0_XSUM_EIPE_S) |\
+RTE_BIT32(VIRTCHNL2_RX_FLEX_DESC_ADV_STATUS0_XSUM_EUDPE_S))
+
+static inline uint64_t
+idpf_splitq_rx_csum_offload(uint8_t err)
+{
+ uint64_t flags = 0;
+
+ if (unlikely((err &
RTE_BIT32(VIRTCHNL2_RX_FLEX_DESC_ADV_STATUS0_L3L4P_S)) == 0))
+ return flags;
+
+ if (likely((err & IDPF_RX_FLEX_DESC_ADV_STATUS0_XSUM_S) == 0)) {
+ flags |= (RTE_MBUF_F_RX_IP_CKSUM_GOOD |
+ RTE_MBUF_F_RX_L4_CKSUM_GOOD);
+ return flags;
+ }
+
+ if (unlikely((err &
RTE_BIT32(VIRTCHNL2_RX_FLEX_DESC_ADV_STATUS0_XSUM_IPE_S)) != 0))
+ flags |= RTE_MBUF_F_RX_IP_CKSUM_BAD;
+ else
+ flags |= RTE_MBUF_F_RX_IP_CKSUM_GOOD;
+
+ if (unlikely((err &
RTE_BIT32(VIRTCHNL2_RX_FLEX_DESC_ADV_STATUS0_XSUM_L4E_S)) != 0))
+ flags |= RTE_MBUF_F_RX_L4_CKSUM_BAD;
+ else
+ flags |= RTE_MBUF_F_RX_L4_CKSUM_GOOD;
+
+ if (unlikely((err &
RTE_BIT32(VIRTCHNL2_RX_FLEX_DESC_ADV_STATUS0_XSUM_EIPE_S)) != 0))
+ flags |= RTE_MBUF_F_RX_OUTER_IP_CKSUM_BAD;
+
+ if (unlikely((err &
RTE_BIT32(VIRTCHNL2_RX_FLEX_DESC_ADV_STATUS0_XSUM_EUDPE_S)) != 0))
+ flags |= RTE_MBUF_F_RX_OUTER_L4_CKSUM_BAD;
+ else
+ flags |= RTE_MBUF_F_RX_OUTER_L4_CKSUM_GOOD;
+
+ return flags;
+}
+
+#define IDPF_RX_FLEX_DESC_ADV_HASH1_S 0
+#define IDPF_RX_FLEX_DESC_ADV_HASH2_S 16
+#define IDPF_RX_FLEX_DESC_ADV_HASH3_S 24
+
+static inline uint64_t
+idpf_splitq_rx_rss_offload(struct rte_mbuf *mb,
+ volatile struct virtchnl2_rx_flex_desc_adv_nic_3
*rx_desc)
+{
+ uint8_t status_err0_qw0;
+ uint64_t flags = 0;
+
+ status_err0_qw0 = rx_desc->status_err0_qw0;
+
+ if ((status_err0_qw0 &
RTE_BIT32(VIRTCHNL2_RX_FLEX_DESC_ADV_STATUS0_RSS_VALID_S)) != 0) {
+ flags |= RTE_MBUF_F_RX_RSS_HASH;
+ mb->hash.rss = (rte_le_to_cpu_16(rx_desc->hash1) <<
+ IDPF_RX_FLEX_DESC_ADV_HASH1_S) |
+ ((uint32_t)(rx_desc->ff2_mirrid_hash2.hash2) <<
+IDPF_RX_FLEX_DESC_ADV_HASH2_S) |
+ ((uint32_t)(rx_desc->hash3) <<
+IDPF_RX_FLEX_DESC_ADV_HASH3_S);
+ }
+
+ return flags;
+}
+
static void
idpf_split_rx_bufq_refill(struct idpf_rx_queue *rx_bufq)
{
@@ -1282,9 +1349,11 @@ idpf_splitq_recv_pkts(void *rx_queue, struct rte_mbuf
**rx_pkts,
uint16_t pktlen_gen_bufq_id;
struct idpf_rx_queue *rxq;
const uint32_t *ptype_tbl;
+ uint8_t status_err0_qw1;
[PATCH v14 16/18] net/idpf: add support for Tx offloading
Add Tx offloading support:
- support TSO
Signed-off-by: Beilei Xing
Signed-off-by: Xiaoyun Li
Signed-off-by: Junfeng Guo
---
doc/guides/nics/features/idpf.ini | 1 +
drivers/net/idpf/idpf_ethdev.c| 4 +-
drivers/net/idpf/idpf_rxtx.c | 128 +-
drivers/net/idpf/idpf_rxtx.h | 22 +
4 files changed, 152 insertions(+), 3 deletions(-)
diff --git a/doc/guides/nics/features/idpf.ini
b/doc/guides/nics/features/idpf.ini
index 6eefac9529..86c182021f 100644
--- a/doc/guides/nics/features/idpf.ini
+++ b/doc/guides/nics/features/idpf.ini
@@ -9,6 +9,7 @@
[Features]
Queue start/stop = Y
MTU update = Y
+TSO = P
L3 checksum offload = P
L4 checksum offload = P
Linux= Y
diff --git a/drivers/net/idpf/idpf_ethdev.c b/drivers/net/idpf/idpf_ethdev.c
index 88f3db3267..868f28d003 100644
--- a/drivers/net/idpf/idpf_ethdev.c
+++ b/drivers/net/idpf/idpf_ethdev.c
@@ -95,7 +95,9 @@ idpf_dev_info_get(struct rte_eth_dev *dev, struct
rte_eth_dev_info *dev_info)
RTE_ETH_RX_OFFLOAD_TCP_CKSUM|
RTE_ETH_RX_OFFLOAD_OUTER_IPV4_CKSUM;
- dev_info->tx_offload_capa = RTE_ETH_TX_OFFLOAD_MULTI_SEGS;
+ dev_info->tx_offload_capa =
+ RTE_ETH_TX_OFFLOAD_TCP_TSO |
+ RTE_ETH_TX_OFFLOAD_MULTI_SEGS;
dev_info->default_txconf = (struct rte_eth_txconf) {
.tx_free_thresh = IDPF_DEFAULT_TX_FREE_THRESH,
diff --git a/drivers/net/idpf/idpf_rxtx.c b/drivers/net/idpf/idpf_rxtx.c
index f15e61a785..cc296d7ab1 100644
--- a/drivers/net/idpf/idpf_rxtx.c
+++ b/drivers/net/idpf/idpf_rxtx.c
@@ -1506,6 +1506,49 @@ idpf_split_tx_free(struct idpf_tx_queue *cq)
cq->tx_tail = next;
}
+/* Check if the context descriptor is needed for TX offloading */
+static inline uint16_t
+idpf_calc_context_desc(uint64_t flags)
+{
+ if ((flags & RTE_MBUF_F_TX_TCP_SEG) != 0)
+ return 1;
+
+ return 0;
+}
+
+/* set TSO context descriptor
+ */
+static inline void
+idpf_set_splitq_tso_ctx(struct rte_mbuf *mbuf,
+ union idpf_tx_offload tx_offload,
+ volatile union idpf_flex_tx_ctx_desc *ctx_desc)
+{
+ uint16_t cmd_dtype;
+ uint32_t tso_len;
+ uint8_t hdr_len;
+
+ if (tx_offload.l4_len == 0) {
+ PMD_TX_LOG(DEBUG, "L4 length set to 0");
+ return;
+ }
+
+ hdr_len = tx_offload.l2_len +
+ tx_offload.l3_len +
+ tx_offload.l4_len;
+ cmd_dtype = IDPF_TX_DESC_DTYPE_FLEX_TSO_CTX |
+ IDPF_TX_FLEX_CTX_DESC_CMD_TSO;
+ tso_len = mbuf->pkt_len - hdr_len;
+
+ ctx_desc->tso.qw1.cmd_dtype = rte_cpu_to_le_16(cmd_dtype);
+ ctx_desc->tso.qw0.hdr_len = hdr_len;
+ ctx_desc->tso.qw0.mss_rt =
+ rte_cpu_to_le_16((uint16_t)mbuf->tso_segsz &
+IDPF_TXD_FLEX_CTX_MSS_RT_M);
+ ctx_desc->tso.qw0.flex_tlen =
+ rte_cpu_to_le_32(tso_len &
+IDPF_TXD_FLEX_CTX_MSS_RT_M);
+}
+
uint16_t
idpf_splitq_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts,
uint16_t nb_pkts)
@@ -1514,11 +1557,14 @@ idpf_splitq_xmit_pkts(void *tx_queue, struct rte_mbuf
**tx_pkts,
volatile struct idpf_flex_tx_sched_desc *txr;
volatile struct idpf_flex_tx_sched_desc *txd;
struct idpf_tx_entry *sw_ring;
+ union idpf_tx_offload tx_offload = {0};
struct idpf_tx_entry *txe, *txn;
uint16_t nb_used, tx_id, sw_id;
struct rte_mbuf *tx_pkt;
uint16_t nb_to_clean;
uint16_t nb_tx = 0;
+ uint64_t ol_flags;
+ uint16_t nb_ctx;
if (unlikely(txq == NULL) || unlikely(!txq->q_started))
return nb_tx;
@@ -1548,7 +1594,29 @@ idpf_splitq_xmit_pkts(void *tx_queue, struct rte_mbuf
**tx_pkts,
if (txq->nb_free < tx_pkt->nb_segs)
break;
- nb_used = tx_pkt->nb_segs;
+
+ ol_flags = tx_pkt->ol_flags;
+ tx_offload.l2_len = tx_pkt->l2_len;
+ tx_offload.l3_len = tx_pkt->l3_len;
+ tx_offload.l4_len = tx_pkt->l4_len;
+ tx_offload.tso_segsz = tx_pkt->tso_segsz;
+ /* Calculate the number of context descriptors needed. */
+ nb_ctx = idpf_calc_context_desc(ol_flags);
+ nb_used = tx_pkt->nb_segs + nb_ctx;
+
+ /* context descriptor */
+ if (nb_ctx != 0) {
+ volatile union idpf_flex_tx_ctx_desc *ctx_desc =
+ (volatile union idpf_flex_tx_ctx_desc *)&txr[tx_id];
+
+ if ((ol_flags & RTE_MBUF_F_TX_TCP_SEG) != 0)
+ idpf_set_splitq_tso_ctx(tx_pkt, tx_offload,
+ ctx_desc);
+
+ tx_id+
[PATCH v14 17/18] net/idpf: add AVX512 data path for single queue model
Add support of AVX512 vector data path for single queue model.
Signed-off-by: Wenjun Wu
Signed-off-by: Junfeng Guo
---
doc/guides/nics/idpf.rst| 19 +
drivers/net/idpf/idpf_ethdev.c | 3 +-
drivers/net/idpf/idpf_ethdev.h | 5 +
drivers/net/idpf/idpf_rxtx.c| 145
drivers/net/idpf/idpf_rxtx.h| 21 +
drivers/net/idpf/idpf_rxtx_vec_avx512.c | 871
drivers/net/idpf/idpf_rxtx_vec_common.h | 100 +++
drivers/net/idpf/meson.build| 28 +
8 files changed, 1191 insertions(+), 1 deletion(-)
create mode 100644 drivers/net/idpf/idpf_rxtx_vec_avx512.c
create mode 100644 drivers/net/idpf/idpf_rxtx_vec_common.h
diff --git a/doc/guides/nics/idpf.rst b/doc/guides/nics/idpf.rst
index c1001d5d0c..3039c61748 100644
--- a/doc/guides/nics/idpf.rst
+++ b/doc/guides/nics/idpf.rst
@@ -64,3 +64,22 @@ Refer to the document :ref:`compiling and testing a PMD for
a NIC tx_offload_capa =
RTE_ETH_TX_OFFLOAD_TCP_TSO |
- RTE_ETH_TX_OFFLOAD_MULTI_SEGS;
+ RTE_ETH_TX_OFFLOAD_MULTI_SEGS |
+ RTE_ETH_TX_OFFLOAD_MBUF_FAST_FREE;
dev_info->default_txconf = (struct rte_eth_txconf) {
.tx_free_thresh = IDPF_DEFAULT_TX_FREE_THRESH,
diff --git a/drivers/net/idpf/idpf_ethdev.h b/drivers/net/idpf/idpf_ethdev.h
index 8d0804f603..7d54e5db60 100644
--- a/drivers/net/idpf/idpf_ethdev.h
+++ b/drivers/net/idpf/idpf_ethdev.h
@@ -162,6 +162,11 @@ struct idpf_adapter {
uint32_t max_txq_per_msg;
uint32_t ptype_tbl[IDPF_MAX_PKT_TYPE] __rte_cache_min_aligned;
+
+ bool rx_vec_allowed;
+ bool tx_vec_allowed;
+ bool rx_use_avx512;
+ bool tx_use_avx512;
};
TAILQ_HEAD(idpf_adapter_list, idpf_adapter);
diff --git a/drivers/net/idpf/idpf_rxtx.c b/drivers/net/idpf/idpf_rxtx.c
index cc296d7ab1..9e20f2b9d3 100644
--- a/drivers/net/idpf/idpf_rxtx.c
+++ b/drivers/net/idpf/idpf_rxtx.c
@@ -4,9 +4,11 @@
#include
#include
+#include
#include "idpf_ethdev.h"
#include "idpf_rxtx.h"
+#include "idpf_rxtx_vec_common.h"
static int
check_rx_thresh(uint16_t nb_desc, uint16_t thresh)
@@ -252,6 +254,8 @@ reset_single_rx_queue(struct idpf_rx_queue *rxq)
rxq->pkt_first_seg = NULL;
rxq->pkt_last_seg = NULL;
+ rxq->rxrearm_start = 0;
+ rxq->rxrearm_nb = 0;
}
static void
@@ -2073,25 +2077,166 @@ idpf_prep_pkts(__rte_unused void *tx_queue, struct
rte_mbuf **tx_pkts,
return i;
}
+static void __rte_cold
+release_rxq_mbufs_vec(struct idpf_rx_queue *rxq)
+{
+ const uint16_t mask = rxq->nb_rx_desc - 1;
+ uint16_t i;
+
+ if (rxq->sw_ring == NULL || rxq->rxrearm_nb >= rxq->nb_rx_desc)
+ return;
+
+ /* free all mbufs that are valid in the ring */
+ if (rxq->rxrearm_nb == 0) {
+ for (i = 0; i < rxq->nb_rx_desc; i++) {
+ if (rxq->sw_ring[i] != NULL)
+ rte_pktmbuf_free_seg(rxq->sw_ring[i]);
+ }
+ } else {
+ for (i = rxq->rx_tail; i != rxq->rxrearm_start; i = (i + 1) &
mask) {
+ if (rxq->sw_ring[i] != NULL)
+ rte_pktmbuf_free_seg(rxq->sw_ring[i]);
+ }
+ }
+
+ rxq->rxrearm_nb = rxq->nb_rx_desc;
+
+ /* set all entries to NULL */
+ memset(rxq->sw_ring, 0, sizeof(rxq->sw_ring[0]) * rxq->nb_rx_desc);
+}
+
+static const struct idpf_rxq_ops def_singleq_rx_ops_vec = {
+ .release_mbufs = release_rxq_mbufs_vec,
+};
+
+static inline int
+idpf_singleq_rx_vec_setup_default(struct idpf_rx_queue *rxq)
+{
+ uintptr_t p;
+ struct rte_mbuf mb_def = { .buf_addr = 0 }; /* zeroed mbuf */
+
+ mb_def.nb_segs = 1;
+ mb_def.data_off = RTE_PKTMBUF_HEADROOM;
+ mb_def.port = rxq->port_id;
+ rte_mbuf_refcnt_set(&mb_def, 1);
+
+ /* prevent compiler reordering: rearm_data covers previous fields */
+ rte_compiler_barrier();
+ p = (uintptr_t)&mb_def.rearm_data;
+ rxq->mbuf_initializer = *(uint64_t *)p;
+ return 0;
+}
+
+int __rte_cold
+idpf_singleq_rx_vec_setup(struct idpf_rx_queue *rxq)
+{
+ rxq->ops = &def_singleq_rx_ops_vec;
+ return idpf_singleq_rx_vec_setup_default(rxq);
+}
+
void
idpf_set_rx_function(struct rte_eth_dev *dev)
{
struct idpf_vport *vport = dev->data->dev_private;
+#ifdef RTE_ARCH_X86
+ struct idpf_adapter *ad = vport->adapter;
+ struct idpf_rx_queue *rxq;
+ int i;
+
+ if (idpf_rx_vec_dev_check_default(dev) == IDPF_VECTOR_PATH &&
+ rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
+ ad->rx_vec_allowed = true;
+
+ if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_512)
+#ifdef CC_AVX512_SUPPORT
+ if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX512F) == 1
&&
+ rte_cpu
[PATCH v14 18/18] net/idpf: add support for timestamp offload
Add support for timestamp offload.
Signed-off-by: Wenjing Qiao
Signed-off-by: Junfeng Guo
---
doc/guides/nics/features/idpf.ini | 1 +
drivers/net/idpf/idpf_ethdev.c| 5 +-
drivers/net/idpf/idpf_ethdev.h| 3 ++
drivers/net/idpf/idpf_rxtx.c | 65 ++
drivers/net/idpf/idpf_rxtx.h | 90 +++
5 files changed, 163 insertions(+), 1 deletion(-)
diff --git a/doc/guides/nics/features/idpf.ini
b/doc/guides/nics/features/idpf.ini
index 86c182021f..5fbad9600c 100644
--- a/doc/guides/nics/features/idpf.ini
+++ b/doc/guides/nics/features/idpf.ini
@@ -12,6 +12,7 @@ MTU update = Y
TSO = P
L3 checksum offload = P
L4 checksum offload = P
+Timestamp offload= P
Linux= Y
x86-32 = Y
x86-64 = Y
diff --git a/drivers/net/idpf/idpf_ethdev.c b/drivers/net/idpf/idpf_ethdev.c
index 62b13b602e..84e7421d76 100644
--- a/drivers/net/idpf/idpf_ethdev.c
+++ b/drivers/net/idpf/idpf_ethdev.c
@@ -22,6 +22,8 @@ rte_spinlock_t idpf_adapter_lock;
struct idpf_adapter_list idpf_adapter_list;
bool idpf_adapter_list_init;
+uint64_t idpf_timestamp_dynflag;
+
static const char * const idpf_valid_args[] = {
IDPF_TX_SINGLE_Q,
IDPF_RX_SINGLE_Q,
@@ -93,7 +95,8 @@ idpf_dev_info_get(struct rte_eth_dev *dev, struct
rte_eth_dev_info *dev_info)
RTE_ETH_RX_OFFLOAD_IPV4_CKSUM |
RTE_ETH_RX_OFFLOAD_UDP_CKSUM|
RTE_ETH_RX_OFFLOAD_TCP_CKSUM|
- RTE_ETH_RX_OFFLOAD_OUTER_IPV4_CKSUM;
+ RTE_ETH_RX_OFFLOAD_OUTER_IPV4_CKSUM |
+ RTE_ETH_RX_OFFLOAD_TIMESTAMP;
dev_info->tx_offload_capa =
RTE_ETH_TX_OFFLOAD_TCP_TSO |
diff --git a/drivers/net/idpf/idpf_ethdev.h b/drivers/net/idpf/idpf_ethdev.h
index 7d54e5db60..ccdf4abe40 100644
--- a/drivers/net/idpf/idpf_ethdev.h
+++ b/drivers/net/idpf/idpf_ethdev.h
@@ -167,6 +167,9 @@ struct idpf_adapter {
bool tx_vec_allowed;
bool rx_use_avx512;
bool tx_use_avx512;
+
+ /* For PTP */
+ uint64_t time_hw;
};
TAILQ_HEAD(idpf_adapter_list, idpf_adapter);
diff --git a/drivers/net/idpf/idpf_rxtx.c b/drivers/net/idpf/idpf_rxtx.c
index 9e20f2b9d3..bafa007faf 100644
--- a/drivers/net/idpf/idpf_rxtx.c
+++ b/drivers/net/idpf/idpf_rxtx.c
@@ -10,6 +10,8 @@
#include "idpf_rxtx.h"
#include "idpf_rxtx_vec_common.h"
+static int idpf_timestamp_dynfield_offset = -1;
+
static int
check_rx_thresh(uint16_t nb_desc, uint16_t thresh)
{
@@ -900,6 +902,24 @@ idpf_tx_queue_setup(struct rte_eth_dev *dev, uint16_t
queue_idx,
return idpf_tx_split_queue_setup(dev, queue_idx, nb_desc,
socket_id, tx_conf);
}
+
+static int
+idpf_register_ts_mbuf(struct idpf_rx_queue *rxq)
+{
+ int err;
+ if ((rxq->offloads & RTE_ETH_RX_OFFLOAD_TIMESTAMP) != 0) {
+ /* Register mbuf field and flag for Rx timestamp */
+ err =
rte_mbuf_dyn_rx_timestamp_register(&idpf_timestamp_dynfield_offset,
+
&idpf_timestamp_dynflag);
+ if (err != 0) {
+ PMD_DRV_LOG(ERR,
+ "Cannot register mbuf field/flag for
timestamp");
+ return -EINVAL;
+ }
+ }
+ return 0;
+}
+
static int
idpf_alloc_single_rxq_mbufs(struct idpf_rx_queue *rxq)
{
@@ -993,6 +1013,13 @@ idpf_rx_queue_init(struct rte_eth_dev *dev, uint16_t
rx_queue_id)
return -EINVAL;
}
+ err = idpf_register_ts_mbuf(rxq);
+ if (err != 0) {
+ PMD_DRV_LOG(ERR, "fail to regidter timestamp mbuf %u",
+ rx_queue_id);
+ return -EIO;
+ }
+
if (rxq->bufq1 == NULL) {
/* Single queue */
err = idpf_alloc_single_rxq_mbufs(rxq);
@@ -1354,6 +1381,7 @@ idpf_splitq_recv_pkts(void *rx_queue, struct rte_mbuf
**rx_pkts,
struct idpf_rx_queue *rxq;
const uint32_t *ptype_tbl;
uint8_t status_err0_qw1;
+ struct idpf_adapter *ad;
struct rte_mbuf *rxm;
uint16_t rx_id_bufq1;
uint16_t rx_id_bufq2;
@@ -1363,9 +1391,11 @@ idpf_splitq_recv_pkts(void *rx_queue, struct rte_mbuf
**rx_pkts,
uint16_t gen_id;
uint16_t rx_id;
uint16_t nb_rx;
+ uint64_t ts_ns;
nb_rx = 0;
rxq = rx_queue;
+ ad = rxq->adapter;
if (unlikely(rxq == NULL) || unlikely(!rxq->q_started))
return nb_rx;
@@ -1376,6 +1406,9 @@ idpf_splitq_recv_pkts(void *rx_queue, struct rte_mbuf
**rx_pkts,
rx_desc_ring = rxq->rx_ring;
ptype_tbl = rxq->adapter->ptype_tbl;
+ if ((rxq->offloads & RTE_ETH_RX_OFFLOAD_TIMESTAMP) != 0)
+ rxq->hw_register_set = 1;
+
Re: [PATCH] bus/vdev: automatically add eth alias for net drivers
On Thu, Oct 20, 2022 at 1:52 PM Ferruh Yigit wrote: > >> Honestly I think the status quo is OK: > >> We have some aliases in some PMD for some historical reason > >> and everybody looks OK with that. Isn't it? > >> > > > > Well, the inconsistency bugs me a little, but if others feel the status quo > > is ok, I'm ok with that. > > In my perspective this is for cleanup, and new PMDs keep adding alias > because they are copying from existing drivers. > Except from above there is no harm to have alias. Do we have a "valid" case of adding new aliases? I don't think it is the case, so we can warn of new aliases introduction in checkpatches.sh. At worse, if a valid case is identified later, checkpatches.sh is only a warning in patchwork and maintainers will manually review this warning. -- David Marchand
Re: [RFC] pcapng: record received RSS hash in pcap file
On Tue, Jul 26, 2022 at 11:37 PM Stephen Hemminger wrote: > > There is an option for recording RSS hash with packets in the > pcapng standard. This implements this for all received packets. > > There is a corner case that can not be addressed with current > DPDK API's. If using rte_flow() and some hardware it is possible > to write a flow rule that uses another hash function like XOR. > But there is no API that records this, or provides the algorithm > info on a per-packet basis. > > The current version of wireshark does not display the recorded > hash option. But if we build it they will come. I updated the commitlog adding some info about "yet to be released 4.1" wireshark. > > Signed-off-by: Stephen Hemminger Tested-by: Ben Magistro Applied, thanks. -- David Marchand
Re: [PATCH] mempool: cache align mempool cache objects
Hi Morten,
On Wed, Oct 26, 2022 at 04:44:36PM +0200, Morten Brørup wrote:
> Add __rte_cache_aligned to the objs array.
>
> It makes no difference in the general case, but if get/put operations are
> always 32 objects, it will reduce the number of memory (or last level
> cache) accesses from five to four 64 B cache lines for every get/put
> operation.
>
> For readability reasons, an example using 16 objects follows:
>
> Currently, with 16 objects (128B), we access to 3
> cache lines:
>
> ┌┐
> │len │
> cache ││---
> line0 ││ ^
> ││ |
> ├┤ | 16 objects
> ││ | 128B
> cache ││ |
> line1 ││ |
> ││ |
> ├┤ |
> ││_v_
> cache ││
> line2 ││
> ││
> └┘
>
> With the alignment, it is also 3 cache lines:
>
> ┌┐
> │len │
> cache ││
> line0 ││
> ││
> ├┤---
> ││ ^
> cache ││ |
> line1 ││ |
> ││ |
> ├┤ | 16 objects
> ││ | 128B
> cache ││ |
> line2 ││ |
> ││ v
> └┘---
>
> However, accessing the objects at the bottom of the mempool cache is a
> special case, where cache line0 is also used for objects.
>
> Consider the next burst (and any following bursts):
>
> Current:
> ┌┐
> │len │
> cache ││
> line0 ││
> ││
> ├┤
> ││
> cache ││
> line1 ││
> ││
> ├┤
> ││
> cache ││---
> line2 ││ ^
> ││ |
> ├┤ | 16 objects
> ││ | 128B
> cache ││ |
> line3 ││ |
> ││ |
> ├┤ |
> ││_v_
> cache ││
> line4 ││
> ││
> └┘
> 4 cache lines touched, incl. line0 for len.
>
> With the proposed alignment:
> ┌┐
> │len │
> cache ││
> line0 ││
> ││
> ├┤
> ││
> cache ││
> line1 ││
> ││
> ├┤
> ││
> cache ││
> line2 ││
> ││
> ├┤
> ││---
> cache ││ ^
> line3 ││ |
> ││ | 16 objects
> ├┤ | 128B
> ││ |
> cache ││ |
> line4 ││ |
> ││_v_
> └┘
> Only 3 cache lines touched, incl. line0 for len.
I understand your logic, but are we sure that having an application that
works with bulks of 32 means that the cache will stay aligned to 32
elements for the whole life of the application?
In an application, the alignment of the cache can change if you have
any of:
- software queues (reassembly for instance)
- packet duplication (bridge, multicast)
- locally generated packets (keepalive, control protocol)
- pipeline to other cores
Even with testpmd, which work by bulk of 32, I can see that the size
of the cache filling is not aligned to 32. Right after starting the
application, we already have this:
internal cache infos:
cache_size=250
cache_count[0]=231
This is probably related to the hw rx rings size, number of queues,
number of ports.
The "250" default value for cache size in testpmd is questionable, but
with --mbcache=256, the behavior is similar.
Also, when we transmit to a NIC, the mbufs are not returned immediatly
to the pool, they may stay in the hw tx ring during some time, which is
a driver decision.
After processing traffic on cores 8 and 24 with this testpmd, I get:
cache_count[0]=231
cache_count[8]=123
cache_count[24]=122
In my opinion, it is not realistic to think that the mempool cache will
remain aligned to cachelines. In these conditions, it looks better to
keep the structure packed to avoid wasting memory.
Olivier
>
> Credits go to Olivier Matz for the nice ASCII graphics.
>
> Signed-off-by: Morten Brørup
> ---
> lib/mempool/rte_mempool.h | 6 --
> 1 file changed, 4 insertions(+), 2 deletions(-)
>
> diff --git a/lib/mempool/rte_mempool.h b/lib/mempool/rte_mempool.h
> index 1f5707f46a..3725a72951 100644
> --- a/lib/mempool/rte_mempool.h
> +++ b/lib/mempool/rte_mempool.h
> @@ -86,11 +86,13 @@ struct rte_mempool_cache {
> uint32_t size;/**< Size of the cache */
> uint32_t flushthresh; /**< Threshold before we flush excess elements */
> uint32_t len; /**< Current cache count */
> - /*
> + /**
> + * Cache objects
> + *
>* Cache is allocated to this size to allow it to overflow in certain
>* cases to avoid needless emptying of cache.
>*/
> - void *objs[RTE_MEMPOOL_CACHE_MAX_SIZE * 2]; /**< Cache objects */
> + void *objs[RTE_MEMPOOL_CACHE_MAX_SIZE * 2] __rte_cache_aligned;
> } __rte_ca
Re: [PATCH] bus/vdev: automatically add eth alias for net drivers
On 10/27/2022 8:58 AM, David Marchand wrote: On Thu, Oct 20, 2022 at 1:52 PM Ferruh Yigit wrote: Honestly I think the status quo is OK: We have some aliases in some PMD for some historical reason and everybody looks OK with that. Isn't it? Well, the inconsistency bugs me a little, but if others feel the status quo is ok, I'm ok with that. In my perspective this is for cleanup, and new PMDs keep adding alias because they are copying from existing drivers. Except from above there is no harm to have alias. Do we have a "valid" case of adding new aliases? I don't think it is the case, so we can warn of new aliases introduction in checkpatches.sh. I commented a few of them to drop alias. checkpatch can be an option, but my intention was to drop old code to reduce noise, not to add more :) OK to keep the alias if removing it will cause more trouble. At worse, if a valid case is identified later, checkpatches.sh is only a warning in patchwork and maintainers will manually review this warning.
RE: [PATCH v4 2/2] net/ice: fix vlan offload
> -Original Message-
> From: Mingjin Ye
> Sent: Thursday, October 27, 2022 1:10 AM
> To: dev@dpdk.org
> Cc: sta...@dpdk.org; Zhou, YidingX ; Ye, MingjinX
> ; Richardson, Bruce ;
> Konstantin Ananyev ; Yang, Qiming
> ; Zhang, Qi Z ; Lu, Wenzhuo
> ; Junyu Jiang ; Rong, Leyi
> ; Wisam Jaddo ; Xia, Chenbo
> ; Hemant Agrawal ;
> Jerin Jacob ; Ajit Khaparde
>
> Subject: [PATCH v4 2/2] net/ice: fix vlan offload
>
> The vlan tag and flag in Rx descriptor are not processed on vector path, then
> the upper application can't fetch the tci from mbuf.
>
> This commit add handling of vlan RX offloading.
>
> Fixes: c68a52b8b38c ("net/ice: support vector SSE in Rx")
> Fixes: ece1f8a8f1c8 ("net/ice: switch to flexible descriptor in SSE path")
> Fixes: 12443386a0b0 ("net/ice: support flex Rx descriptor RxDID22")
> Fixes: 214f452f7d5f ("net/ice: add AVX2 offload Rx")
> Fixes: 7f85d5ebcfe1 ("net/ice: add AVX512 vector path")
> Fixes: 295968d17407 ("ethdev: add namespace")
> Fixes: 808a17b3c1e6 ("net/ice: add Rx AVX512 offload path")
> Cc: sta...@dpdk.org
>
> Signed-off-by: Mingjin Ye
>
> v3:
> * Fix macros in ice_rxtx_vec_sse.c source file.
> v4:
> * Fix ice_rx_desc_to_olflags_v define in ice_rxtx_vec_sse.c source file.
> ---
Tested-by: Zhimin Huang
RE: [PATCH v4 1/2] app/testpmd: fix vlan offload of rxq
> -Original Message-
> From: Mingjin Ye
> Sent: Thursday, October 27, 2022 1:10 AM
> To: dev@dpdk.org
> Cc: sta...@dpdk.org; Zhou, YidingX ; Ye, MingjinX
> ; Singh, Aman Deep ;
> Zhang, Yuying
> Subject: [PATCH v4 1/2] app/testpmd: fix vlan offload of rxq
>
> After setting vlan offload in testpmd, the result is not updated to rxq.
> Therefore,
> the queue needs to be reconfigured after executing the "vlan offload" related
> commands.
>
> Fixes: a47aa8b97afe ("app/testpmd: add vlan offload support")
> Cc: sta...@dpdk.org
>
> Signed-off-by: Mingjin Ye
> ---
Tested-by: Zhimin Huang
RE: [PATCH v3 1/4] telemetry: support boolean type
> -Original Message- > From: Richardson, Bruce > Sent: Tuesday 25 October 2022 11:39 > To: David Marchand > Cc: Mattias Rönnblom ; dev@dpdk.org; Morten > Brørup ; Power, Ciara > > Subject: Re: [PATCH v3 1/4] telemetry: support boolean type > > On Tue, Oct 25, 2022 at 11:43:27AM +0200, David Marchand wrote: > > On Tue, Oct 25, 2022 at 11:34 AM Mattias Rönnblom > wrote: > > > > > > On 2022-10-25 11:00, David Marchand wrote: > > > > Add the boolean type RTE_TEL_BOOL_VAL for values in arrays and > dicts. > > > > > > > > Signed-off-by: David Marchand > > > > Acked-by: Morten Brørup > > > > Acked-by: Bruce Richardson > > > > Acked-by: Ciara Power > > > > --- > > > > Changes since v1: > > > > - fixed doxygen description, > > I'll let the telemetry maintainers decide on the fate of those patches. > > Thanks. > > > Well, given the lateness in the release process, if it's a choice between > getting the patches in as-is, vs not getting them in at all, I think having > them > in 22.11 is the better choice. While there is cleanup that can be done, I > don't > see this patch as making the overall code any worse. > > My 2c. > > /Bruce Agreed, it's a good addition to the library, +1 to having it in 22.11 Thanks, Ciara
Re: [PATCH] net/mana: disable driver by default
On 10/27/2022 7:13 AM, Thomas Monjalon wrote:
27/10/2022 00:57, Ferruh Yigit:
On 10/26/2022 10:53 PM, Thomas Monjalon wrote:
13/10/2022 18:30, Long Li:
Subject: [PATCH] net/mana: disable driver by default
Driver is disabled by default because its dependencies are not upstreamed
yet, code is available for development and investigation.
When all dependencies are upstreamed, driver can be enabled back.
Fixes: 517ed6e2d590 ("net/mana: add basic driver with build environment")
Signed-off-by: Ferruh Yigit
Acked-by: Long Li
Acked-by: Thomas Monjalon
Applied to dpdk-next-net/main, thanks.
Looks like we cannot dereference the doc:
doc/guides/nics/mana.rst:document isn't included in any toctree
ack
Will fix in next-net, and add mana.rst back to index to remove the warning.
Re: [PATCH] app/testpmd: fix testpmd crash when quit with mlx5 avail_thresh enabled
On 10/24/2022 8:44 AM, Spike Du wrote:
When testpmd quit with mlx5 avail_thresh enabled, a rte timer handler
delays to reconfigure rx queue to re-arm this event. However at the same
time, testpmd is destroying rx queues.
It's never a valid use case for mlx5 avail_thresh. Before testpmd quit,
user should disable avail_thresh configuration to not handle the events.
This is documented in mlx5 driver guide.
To avoid the crash in such use case, check port status, if it is not
RTE_PORT_STARTED, don't process the avail_thresh event.
Fixes: 0edfc9b08316 ("app/testpmd: add Host Shaper command")
Signed-off-by: Spike Du
Acked-by: Aman Singh
---
drivers/net/mlx5/mlx5_testpmd.c | 9 +
1 file changed, 9 insertions(+)
diff --git a/drivers/net/mlx5/mlx5_testpmd.c b/drivers/net/mlx5/mlx5_testpmd.c
index ed84583..1a9ec78 100644
--- a/drivers/net/mlx5/mlx5_testpmd.c
+++ b/drivers/net/mlx5/mlx5_testpmd.c
@@ -25,6 +25,7 @@
static uint8_t host_shaper_avail_thresh_triggered[RTE_MAX_ETHPORTS];
#define SHAPER_DISABLE_DELAY_US 10 /* 100ms */
+extern struct rte_port *ports;
This might not be needed, as we have testpmd.h included.
/**
* Disable the host shaper and re-arm available descriptor threshold event.
@@ -39,7 +40,15 @@
uint16_t port_id = port_rxq_id & 0x;
uint16_t qid = (port_rxq_id >> 16) & 0x;
struct rte_eth_rxq_info qinfo;
+ struct rte_port *port;
+ port = &ports[port_id];
+ if (port->port_status != RTE_PORT_STARTED) {
+ printf("%s port_status(%d) is incorrect, stop avail_thresh "
+ "event processing.\n",
+ __func__, port->port_status);
+ return;
+ }
printf("%s disable shaper\n", __func__);
if (rte_eth_rx_queue_info_get(port_id, qid, &qinfo)) {
printf("rx_queue_info_get returns error\n");
RE: [PATCH] mempool: cache align mempool cache objects
> From: Olivier Matz [mailto:olivier.m...@6wind.com] > Sent: Thursday, 27 October 2022 10.35 > > Hi Morten, > > On Wed, Oct 26, 2022 at 04:44:36PM +0200, Morten Brørup wrote: > > Add __rte_cache_aligned to the objs array. > > > > It makes no difference in the general case, but if get/put operations > are > > always 32 objects, it will reduce the number of memory (or last level > > cache) accesses from five to four 64 B cache lines for every get/put > > operation. > > > > For readability reasons, an example using 16 objects follows: > > > > Currently, with 16 objects (128B), we access to 3 > > cache lines: > > > > ┌┐ > > │len │ > > cache ││--- > > line0 ││ ^ > > ││ | > > ├┤ | 16 objects > > ││ | 128B > > cache ││ | > > line1 ││ | > > ││ | > > ├┤ | > > ││_v_ > > cache ││ > > line2 ││ > > ││ > > └┘ > > > > With the alignment, it is also 3 cache lines: > > > > ┌┐ > > │len │ > > cache ││ > > line0 ││ > > ││ > > ├┤--- > > ││ ^ > > cache ││ | > > line1 ││ | > > ││ | > > ├┤ | 16 objects > > ││ | 128B > > cache ││ | > > line2 ││ | > > ││ v > > └┘--- > > > > However, accessing the objects at the bottom of the mempool cache is > a > > special case, where cache line0 is also used for objects. > > > > Consider the next burst (and any following bursts): > > > > Current: > > ┌┐ > > │len │ > > cache ││ > > line0 ││ > > ││ > > ├┤ > > ││ > > cache ││ > > line1 ││ > > ││ > > ├┤ > > ││ > > cache ││--- > > line2 ││ ^ > > ││ | > > ├┤ | 16 objects > > ││ | 128B > > cache ││ | > > line3 ││ | > > ││ | > > ├┤ | > > ││_v_ > > cache ││ > > line4 ││ > > ││ > > └┘ > > 4 cache lines touched, incl. line0 for len. > > > > With the proposed alignment: > > ┌┐ > > │len │ > > cache ││ > > line0 ││ > > ││ > > ├┤ > > ││ > > cache ││ > > line1 ││ > > ││ > > ├┤ > > ││ > > cache ││ > > line2 ││ > > ││ > > ├┤ > > ││--- > > cache ││ ^ > > line3 ││ | > > ││ | 16 objects > > ├┤ | 128B > > ││ | > > cache ││ | > > line4 ││ | > > ││_v_ > > └┘ > > Only 3 cache lines touched, incl. line0 for len. > > I understand your logic, but are we sure that having an application > that > works with bulks of 32 means that the cache will stay aligned to 32 > elements for the whole life of the application? > > In an application, the alignment of the cache can change if you have > any of: > - software queues (reassembly for instance) > - packet duplication (bridge, multicast) > - locally generated packets (keepalive, control protocol) > - pipeline to other cores > > Even with testpmd, which work by bulk of 32, I can see that the size > of the cache filling is not aligned to 32. Right after starting the > application, we already have this: > > internal cache infos: > cache_size=250 > cache_count[0]=231 > > This is probably related to the hw rx rings size, number of queues, > number of ports. > > The "250" default value for cache size in testpmd is questionable, but > with --mbcache=256, the behavior is similar. > > Also, when we transmit to a NIC, the mbufs are not returned immediatly > to the pool, they may stay in the hw tx ring during some time, which is > a driver decision. > > After processing traffic on cores 8 and 24 with this testpmd, I get: > cache_count[0]=231 > cache_count[8]=123 > cache_count[24]=122 > > In my opinion, it is not realistic to think that the mempool cache will > remain aligned to cachelines. In these conditions, it looks better to > keep the structure packed to avoid wasting memory. I agree that is a special use case to only access the mempool cache in bursts of 32 objects, so the accesses are always cache line aligned. (Generalized, the burst size must not be 32; a burst size that is a multiple of RTE_CACHE_LINE_SIZE/sizeof(void*), i.e. a burst size of 8 on a 64-bit architecture, will do.) Adding a hole of 52 byte per mempool cache is nothing, considering that the mempool cache already uses 8 KB (RTE_MEMPOOL_CACHE_MAX_SIZE * 2 * sizeof(void*) = 1024 * 8 byte) for the objects. Also - assuming that memory allocations are cache line aligned - the 52 byte of
回复: 回复: [PATCH v1 3/3] examples/l3fwd-power: enable PMD power mgmt on Arm
> -邮件原件-
> 发件人: Thomas Monjalon
> 发送时间: Friday, October 21, 2022 4:42 AM
> 收件人: David Marchand
> 抄送: Hunt, David ; Ruifeng Wang
> ; dev@dpdk.org; nd ; Feifei
> Wang
> 主题: Re: 回复: [PATCH v1 3/3] examples/l3fwd-power: enable PMD power
> mgmt on Arm
>
> 11/10/2022 09:56, Feifei Wang:
> > David Marchand
> > > > On 25/08/2022 07:42, Feifei Wang wrote:
> > > > > --- a/examples/l3fwd-power/main.c
> > > > > +++ b/examples/l3fwd-power/main.c
> > > > > @@ -432,8 +432,16 @@ static void
> > > > > signal_exit_now(int sigtype)
> > > > > {
> > > > >
> > > > > - if (sigtype == SIGINT)
> > > > > + if (sigtype == SIGINT) {
> > > > > +#if defined(RTE_ARCH_ARM64)
> > >
> > > Having a arch specific behavior in the application shows that there
> > > is something wrong either in the API, or in the Arm implementation of
> the API.
> > > I don't think this is a good solution.
> > >
> > > Can't we find a better alternative? By changing the API probably?
> > Sorry I do not understand ' shows that there is something wrong either in
> the API'
>
> David means the application developer should not have to be aware of the
> arch differences.
> When you use an API, you don't check how it is implemented, and you are
> not supposed to use #ifdef.
> The API must be arch-agnostic.
Ok, Understand. Thanks for the explanation.
>
> > Here we call ' rte_power_monitor_wakeup' API is due to that we need to
> > wake up all cores from WFE instructions in arm, and then l3fwd can exit
> correctly.
> >
> > This is due to that arm arch is different from x86, if there is no
> > packets received, x86's 'UMONITOR' can automatically exit from energy-
> saving state after waiting for a period of time.
> > But arm's 'WFE' can not exit automatically. It will wait 'SEV'
> > instructions in wake_up API to wake up it.
> >
> > Finally, if user want to exit l3fwd by 'SIGINT' in arm, main core
> > should firstly call 'wake_up' API to force worker cores to exit from energy-
> saving state.
> > Otherwise, the worker will stay in the energy-saving state forever if no
> packet is received.
>
> Please find a way to have a common API,
> even if the API implementation is empty in x86 case.
Yes, I think what we need to do is not a create a new API, it is to look
for a correct location to call 'rte_power_monitor_wakeup'.
>
> > >
> > >
> > > > > + /**
> > > > > + * wake_up api does not need input parameter on Arm,
> > > > > + * so 0 is meaningless here.
> > > > > + */
> > > > > + rte_power_monitor_wakeup(0); #endif
> > > > > quit_signal = true;
> > > > > + }
>
>
回复: [PATCH v1 3/3] examples/l3fwd-power: enable PMD power mgmt on Arm
> -邮件原件-
> 发件人: Stephen Hemminger
> 发送时间: Friday, October 21, 2022 6:10 AM
> 收件人: Feifei Wang
> 抄送: David Hunt ; dev@dpdk.org; nd
> ; Ruifeng Wang
> 主题: Re: [PATCH v1 3/3] examples/l3fwd-power: enable PMD power mgmt
> on Arm
>
> On Thu, 25 Aug 2022 14:42:51 +0800
> Feifei Wang wrote:
>
> > diff --git a/examples/l3fwd-power/main.c b/examples/l3fwd-
> power/main.c
> > index 887c6eae3f..2bd0d700f0 100644
> > --- a/examples/l3fwd-power/main.c
> > +++ b/examples/l3fwd-power/main.c
> > @@ -432,8 +432,16 @@ static void
> > signal_exit_now(int sigtype)
> > {
> >
> > - if (sigtype == SIGINT)
> > + if (sigtype == SIGINT) {
> > +#if defined(RTE_ARCH_ARM64)
> > + /**
> > +* wake_up api does not need input parameter on Arm,
> > +* so 0 is meaningless here.
> > +*/
> > + rte_power_monitor_wakeup(0);
> > +#endif
> > quit_signal = true;
> > + }
> >
>
> This method is problematic. There is no guarantee that power monitor is
> async signal safe.
Agree with this. We will put 'rte_power_monitor_wakeup' out of signal_exit.
And put it after that main_lcore exit 'main_empty_poll_loop':
-
rte_eal_mp_remote_launch(main_telemetry_loop, NULL, CALL_MAIN);
%wake up all worker calls from low power state.
rte_power_monitor_wakeup(0);
if (app_mode == APP_MODE_EMPTY_POLL || app_mode == APP_MODE_TELEMETRY)
launch_timer(rte_lcore_id());
RTE_LCORE_FOREACH_WORKER(lcore_id) {
if (rte_eal_wait_lcore(lcore_id) < 0)
return -1;
}
-
[Bug 1116] Failsafe PMD: Crashes with multiple secondary processes
https://bugs.dpdk.org/show_bug.cgi?id=1116
Bug ID: 1116
Summary: Failsafe PMD: Crashes with multiple secondary
processes
Product: DPDK
Version: 21.11
Hardware: All
OS: Linux
Status: UNCONFIRMED
Severity: major
Priority: Normal
Component: ethdev
Assignee: dev@dpdk.org
Reporter: madhuker.myt...@oracle.com
Target Milestone: ---
On Hyper-V Linux VM's with failsafe/tap PMD crashes when we launch mutiple(two)
secondary processes.
Primary process able to launch well and then after when we launch secondary
processes, getting a crash on eal-intr-thread.
here are failure logs before crash:
EAL: Fail to recv reply for request
/var/run/dpdk/oracusbc/mp_socket:eal_dev_mp_request
EAL: Cannot send request to primary
EAL: Failed to send hotplug request to primary
net_failsafe: Failed to probe devargs net_tap_vsc0{*}
===
With DPDK-21.11 version.
Regards,
Madhuker.
--
You are receiving this mail because:
You are the assignee for the bug.
[PATCH] flow_classify: mark library as deprecated
This library has no maintainer and, for now, nobody expressed interest
in taking over.
Mark this experimental library as deprecated and announce plan for
removal in v23.11.
Signed-off-by: David Marchand
---
MAINTAINERS | 3 +--
doc/guides/prog_guide/flow_classify_lib.rst | 9 +
doc/guides/rel_notes/deprecation.rst| 6 ++
lib/flow_classify/rte_flow_classify.c | 3 +++
lib/meson.build | 5 -
meson_options.txt | 2 +-
6 files changed, 24 insertions(+), 4 deletions(-)
diff --git a/MAINTAINERS b/MAINTAINERS
index 2bd4a55f1b..89768d15ee 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -1448,8 +1448,7 @@ F: doc/guides/prog_guide/ipsec_lib.rst
M: Vladimir Medvedkin
F: app/test-sad/
-Flow Classify - EXPERIMENTAL
-M: Bernard Iremonger
+Flow Classify - EXPERIMENTAL - UNMAINTAINED
F: lib/flow_classify/
F: app/test/test_flow_classify*
F: doc/guides/prog_guide/flow_classify_lib.rst
diff --git a/doc/guides/prog_guide/flow_classify_lib.rst
b/doc/guides/prog_guide/flow_classify_lib.rst
index 7dae0bc8c6..ad2e10ec26 100644
--- a/doc/guides/prog_guide/flow_classify_lib.rst
+++ b/doc/guides/prog_guide/flow_classify_lib.rst
@@ -4,6 +4,15 @@
Flow Classification Library
===
+.. note::
+
+ The Flow Classification library is deprecated and will be removed in future.
+ See :doc:`../rel_notes/deprecation`.
+
+ It is disabled by default in the DPDK build.
+ To re-enable the library, remove 'flow_classify' from the "disable_libs"
+ meson option when configuring a build.
+
DPDK provides a Flow Classification library that provides the ability
to classify an input packet by matching it against a set of Flow rules.
diff --git a/doc/guides/rel_notes/deprecation.rst
b/doc/guides/rel_notes/deprecation.rst
index 05cacb3ea8..e2efa2f8b0 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -96,3 +96,9 @@ Deprecation Notices
to have another parameter ``qp_id`` to return the queue pair ID
which got error interrupt to the application,
so that application can reset that particular queue pair.
+
+* flow_classify: The flow_classify library and example have no maintainer.
+ The library is experimental and, as such, it could be removed from DPDK.
+ Its removal has been postponed to let potential users report interest
+ in maintaining it.
+ In the absence of such interest, this library will be removed in DPDK 23.11.
diff --git a/lib/flow_classify/rte_flow_classify.c
b/lib/flow_classify/rte_flow_classify.c
index e3667306e5..60ca319f24 100644
--- a/lib/flow_classify/rte_flow_classify.c
+++ b/lib/flow_classify/rte_flow_classify.c
@@ -259,6 +259,9 @@ rte_flow_classifier_create(struct
rte_flow_classifier_params *params)
struct rte_flow_classifier *cls;
int ret;
+ RTE_FLOW_CLASSIFY_LOG(WARNING,
+ "WARNING: flow_classify is deprecated and will be removed in
DPDK 23.11\n");
+
/* Check input parameters */
ret = rte_flow_classifier_check_params(params);
if (ret != 0) {
diff --git a/lib/meson.build b/lib/meson.build
index c51cdc24fa..1f802bc456 100644
--- a/lib/meson.build
+++ b/lib/meson.build
@@ -85,7 +85,10 @@ optional_libs = [
'vhost',
]
-dpdk_libs_deprecated += ['kni']
+dpdk_libs_deprecated += [
+'flow_classify',
+'kni',
+]
disabled_libs = []
opt_disabled_libs = run_command(list_dir_globs, get_option('disable_libs'),
diff --git a/meson_options.txt b/meson_options.txt
index 5de1b3ed23..ce26a8a2ec 100644
--- a/meson_options.txt
+++ b/meson_options.txt
@@ -8,7 +8,7 @@ option('developer_mode', type: 'feature', description:
'turn on additional build checks relevant for DPDK developers')
option('disable_drivers', type: 'string', value: '', description:
'Comma-separated list of drivers to explicitly disable.')
-option('disable_libs', type: 'string', value: 'kni', description:
+option('disable_libs', type: 'string', value: 'flow_classify,kni', description:
'Comma-separated list of libraries to explicitly disable. [NOTE: not
all libs can be disabled]')
option('drivers_install_subdir', type: 'string', value: 'dpdk/pmds-',
description:
'Subdirectory of libdir where to install PMDs. Defaults to using a
versioned subdirectory.')
--
2.37.3
RE: [EXT] [PATCH v2 0/4] crypto/qat: extend asymmetric crypto pmd
> This patchset extends Intel QuickAssist Technology asymmetric crypto PMD.
> Following features were added:
> - ECDH algorithm handling
> - EC point verification
>
> Depends-on: series-25284 ("crypto/qat: fix uncleared cookies in asym")
>
Series applied to dpdk-next-crypto
Thanks.
RE: [dpdk-dev v1] crypto/qat: fix of qat build request session in mp
> > Subject: [dpdk-dev v1] crypto/qat: fix of qat build request session in mp
> >
> > This patch fix the session pointer passed in set_session() when ctx has NULL
> > build request pointer in multi-processes scenario.
> >
> > Fixes: fb3b9f492205 ("crypto/qat: rework burst data path")
> > Cc: sta...@dpdk.org
> >
> > Signed-off-by: Kai Ji
> > ---
>
> Acked-by: Ciara Power
Applied to dpdk-next-crypto
Thanks.
RE: [EXT] [dpdk-dev v5] crypto/ipsec_mb: multi-process IPC request handler
> As the queue pair used in secondary process needs to be set up by > the primary process, this patch adds an IPC register function to help > secondary process to send out queue-pair setup request to primary > process via IPC request messages. A new "qp_in_used_pid" param stores > the PID to provide the ownership of the queue-pair so that only the PID > matched queue-pair can be free'd in the request. > > Signed-off-by: Kai Ji > Acked-by: Arek Kusztal > Acked-by: Pablo de Lara > --- > v5: > - minor updates and typo fix Applied to dpdk-next-crypto Thanks.
RE: [EXT] [PATCH 1/2] app/test-crypto-perf: fix number of scheduler sessions
> Subject: RE: [EXT] [PATCH 1/2] app/test-crypto-perf: fix number of scheduler
> sessions
>
> > The scheduler PMD needs 1 session header,
> > along with a session per worker.
> >
> > After the session rework, this change was made to other apps,
> > for example l2fwd-crypto, but was missed in test-crypto-perf.
> >
> > Fixes: bdce2564dbf7 ("cryptodev: rework session framework")
> > Cc: gak...@marvell.com
> >
> > Signed-off-by: Ciara Power
> Acked-by: Akhil Goyal
Series Applied to dpdk-next-crypto
Thanks.
RE: [PATCH v2] app/test: add support to run multi-seg inline IPsec test
> Subject: [PATCH v2] app/test: add support to run multi-seg inline IPsec test > > Add support to run multi-seg inline ipsec test using > new test command `inline_ipsec_sg_autotest` > > Signed-off-by: Nithin Dabilpuram > --- Acked-by: Akhil Goyal Applied to dpdk-next-crypto Thanks.
RE: [EXT] [PATCH v1] examples/fips_validation: parse block error fix
> Acked-by: Gowrishankar Muthukrishnan
>
> > When parsing request files check for file type. This fix will remove
> > dependence on command line parameter for using libjansson
> >
> > Fixes: 0f42f3d6034c ("examples/fips_validation: share callback with multiple
> > keys")
> > Cc: gmuthukri...@marvell.com
> > Signed-off-by: Brian Dooley
> > ---
Applied to dpdk-next-crypto
Thanks.
Re: [PATCH 2/2] license: add MIT license exception for gve driver
19/10/2022 20:42, Stephen Hemminger: > Tech Board and Governing Board approved license a exception > for new Google driver. More general approval for MIT > usage will be handled as needed. > > Signed-off-by: Stephen Hemminger > --- > +4. MIT 10/19/202210/18/2022 > drivers/net/gve/base/* Series applied, thanks.
Re: [PATCH] flow_classify: mark library as deprecated
On 10/27/2022 10:46 AM, David Marchand wrote: This library has no maintainer and, for now, nobody expressed interest in taking over. Mark this experimental library as deprecated and announce plan for removal in v23.11. Signed-off-by: David Marchand Acked-by: Ferruh Yigit
[PATCH v1] crypto/qat: fix null hash algorithm digest size
Add check for null hash algorithm digest size. Digest size should be 4B
or request will be rejected.
Fixes: 1703e94ac5ce ("qat: add driver for QuickAssist devices")
Cc: declan.dohe...@intel.com
Cc: sta...@dpdk.org
Signed-off-by: Brian Dooley
---
drivers/crypto/qat/qat_sym_session.c | 19 +++
1 file changed, 15 insertions(+), 4 deletions(-)
diff --git a/drivers/crypto/qat/qat_sym_session.c
b/drivers/crypto/qat/qat_sym_session.c
index d96122b208..ff4b537706 100644
--- a/drivers/crypto/qat/qat_sym_session.c
+++ b/drivers/crypto/qat/qat_sym_session.c
@@ -2040,7 +2040,12 @@ int qat_sym_cd_auth_set(struct qat_sym_session *cdesc,
hash_offset = cdesc->cd_cur_ptr-((uint8_t *)&cdesc->cd);
hash = (struct icp_qat_hw_auth_setup *)cdesc->cd_cur_ptr;
hash->auth_config.reserved = 0;
- hash->auth_config.config =
+ if (cdesc->qat_hash_alg == ICP_QAT_HW_AUTH_ALGO_NULL)
+ hash->auth_config.config =
+ ICP_QAT_HW_AUTH_CONFIG_BUILD(cdesc->auth_mode,
+ cdesc->qat_hash_alg, 4);
+ else
+ hash->auth_config.config =
ICP_QAT_HW_AUTH_CONFIG_BUILD(cdesc->auth_mode,
cdesc->qat_hash_alg, digestsize);
@@ -2408,10 +2413,16 @@ int qat_sym_cd_auth_set(struct qat_sym_session *cdesc,
/* Auth CD config setup */
hash_cd_ctrl->hash_cfg_offset = hash_offset >> 3;
hash_cd_ctrl->hash_flags = ICP_QAT_FW_AUTH_HDR_FLAG_NO_NESTED;
- hash_cd_ctrl->inner_res_sz = digestsize;
- hash_cd_ctrl->final_sz = digestsize;
hash_cd_ctrl->inner_state1_sz = state1_size;
- auth_param->auth_res_sz = digestsize;
+ if (cdesc->qat_hash_alg == ICP_QAT_HW_AUTH_ALGO_NULL) {
+ hash_cd_ctrl->inner_res_sz = 4;
+ hash_cd_ctrl->final_sz = 4;
+ auth_param->auth_res_sz = 4;
+ } else {
+ hash_cd_ctrl->inner_res_sz = digestsize;
+ hash_cd_ctrl->final_sz = digestsize;
+ auth_param->auth_res_sz = digestsize;
+ }
hash_cd_ctrl->inner_state2_sz = state2_size;
hash_cd_ctrl->inner_state2_offset = hash_cd_ctrl->hash_cfg_offset +
--
2.25.1
RE: [EXT] [PATCH v7 0/6] crypto/uadk: introduce uadk crypto driver
> Introduce a new crypto PMD for hardware accelerators based on UADK [1].
>
> UADK is a framework for user applications to access hardware accelerators.
> UADK relies on IOMMU SVA (Shared Virtual Address) feature, which share
> the same page table between IOMMU and MMU.
> Thereby user application can directly use virtual address for device dma,
> which enhances the performance as well as easy usability.
>
> [1] https://urldefense.proofpoint.com/v2/url?u=https-
> 3A__github.com_Linaro_uadk&d=DwIDAg&c=nKjWec2b6R0mOyPaz7xtfQ&r=Dn
> L7Si2wl_PRwpZ9TWey3eu68gBzn7DkPwuqhd6WNyo&m=AwapMJzvfUDL49W9f
> mtmvIGZMarGUdjRSliBhG4tep6Uh5wN2zYrZRJ4JpbrVolg&s=MxSabEjMMkeRr0L
> Z7c1gSmaJmJ1_dIdJIKrst98pKzk&e=
>
> Test:
> sudo dpdk-test --vdev=crypto_uadk (--log-level=6)
> RTE>>cryptodev_uadk_autotest
> RTE>>quit
>
> Update in v7:
> 05: fix key_size of SHA384 HMAC
>
> Update in v6:
> Akhil helped review the v5, and address all the comments from Akhil
> 01: add rel_notes info, update uadk.rst, put struct in header, etc.
> 02: move struct to header
> 04: move struct to header, remove RTE_CRYPTODEV_FF_SYM_SESSIONLESS
> 05: fixed key_size in HMAC mode
> 06: updated app/test/meson.build and uadk.rst in the mean time.
>
> Update in v5
> Patch 1 fix the build issue when uadk is installed to specific folder
> And update doc accordingly
>
> Update in v4:
> Akril suggest dpdk use pkg-config, So
> Enable uadk support x86 local build, and support pkg-config.
> Use pkg-config feature for the uadk crypto pmd.
> Add build uadk library steps in doc
> Test on both x86 and arm.
> x86 can build and install, but can not test since no device.
>
> Resend v3:
> Rebase on next/for-main, which just merged the series
> "cryptodev: rework session framework".
>
> Update in v3:
> Split patches according to Akhil's suggestions
> Please split the patches as below.
> 1. introduce driver - create files with meson.build and with probe/remove
>and device ops defined but not implemented.
>You do not need to write empty functions.
>Add basic documentation also which defines what the driver is.
>You can explain the build dependency here.
> 2. define queue structs and setup/remove APIs
> 3. Add data path
> 4. implement cipher op. Add capabilities and documentation of what is
>supported in each of the patches. Add feature flags etc.
> 5. implement auth, add capabilities and documentation
> 6. test app changes.
>
> Update in v2:
> Change uadk_supported_platform to uadk_crypto_version, which matches
> better
> than platform.
> enum uadk_crypto_version {
> UADK_CRYPTO_V2,
> UADK_CRYPTO_V3,
> };
>
> Update in v1, compared with rfc
>
> Suggested from Akhil Goyal
> Only consider crypto PMD first
> Split patch into small (individually compiled) patches.
> Update MAINTAINERS and doc/guides/cryptodevs/features/uadk.ini
>
> Zhangfei Gao (6):
> crypto/uadk: introduce uadk crypto driver
> crypto/uadk: support basic operations
> crypto/uadk: support enqueue/dequeue operations
> crypto/uadk: support cipher algorithms
> crypto/uadk: support auth algorithms
> test/crypto: support uadk PMD
>
> MAINTAINERS |6 +
> app/test/meson.build |1 +
> app/test/test_cryptodev.c |7 +
> app/test/test_cryptodev.h |1 +
> doc/guides/cryptodevs/features/uadk.ini | 55 +
> doc/guides/cryptodevs/index.rst |1 +
> doc/guides/cryptodevs/uadk.rst| 96 ++
> doc/guides/rel_notes/release_22_11.rst|6 +
> drivers/crypto/meson.build|1 +
> drivers/crypto/uadk/meson.build | 30 +
> drivers/crypto/uadk/uadk_crypto_pmd.c | 1081 +
> drivers/crypto/uadk/uadk_crypto_pmd_private.h | 79 ++
> drivers/crypto/uadk/version.map |3 +
> 13 files changed, 1367 insertions(+)
> create mode 100644 doc/guides/cryptodevs/features/uadk.ini
> create mode 100644 doc/guides/cryptodevs/uadk.rst
> create mode 100644 drivers/crypto/uadk/meson.build
> create mode 100644 drivers/crypto/uadk/uadk_crypto_pmd.c
> create mode 100644 drivers/crypto/uadk/uadk_crypto_pmd_private.h
> create mode 100644 drivers/crypto/uadk/version.map
Series Applied to dpdk-next-crypto
Welcome to dpdk.
Thanks.
RE: [PATCH v4 1/2] app/testpmd: fix vlan offload of rxq
Hi lihuisong,
This means that queue offloads need to update by recalling dev_configure
and setup target queues.
Can you tell me, where is the limitation?
Thanks,
Mingjin
> -Original Message-
> From: lihuisong (C)
> Sent: 2022年10月26日 17:53
> To: Ye, MingjinX ; dev@dpdk.org
> Cc: sta...@dpdk.org; Zhou, YidingX ; Singh, Aman
> Deep ; Zhang, Yuying
>
> Subject: Re: [PATCH v4 1/2] app/testpmd: fix vlan offload of rxq
>
>
> 在 2022/10/27 1:10, Mingjin Ye 写道:
> > After setting vlan offload in testpmd, the result is not updated to
> > rxq. Therefore, the queue needs to be reconfigured after executing the
> > "vlan offload" related commands.
> >
> > Fixes: a47aa8b97afe ("app/testpmd: add vlan offload support")
> > Cc: sta...@dpdk.org
> >
> > Signed-off-by: Mingjin Ye
> > ---
> > app/test-pmd/cmdline.c | 1 +
> > 1 file changed, 1 insertion(+)
> >
> > diff --git a/app/test-pmd/cmdline.c b/app/test-pmd/cmdline.c index
> > 17be2de402..ce125f549f 100644
> > --- a/app/test-pmd/cmdline.c
> > +++ b/app/test-pmd/cmdline.c
> > @@ -4133,6 +4133,7 @@ cmd_vlan_offload_parsed(void *parsed_result,
> > else
> > vlan_extend_set(port_id, on);
> >
> > + cmd_reconfig_device_queue(port_id, 1, 1);
> This means that queue offloads need to upadte by re-calling dev_configure
> and setup all queues, right?
> If it is, this adds a usage limitation.
> > return;
> > }
> >
[PATCH v2] examples/ipsec-secgw: support per SA HW reassembly
This add the support of hardware reassembly per SA basis.
In SA rule, new parameter reassembly_en is added to enable
HW reassembly per SA.
For example:
sa in aead_algo aead_key mode ipv4-tunnel src
dst type inline-protocol-offload port_id reassembly_en
Stats counter frag_dropped will represent the number of fragment
drop in case of reassembly failures.
Signed-off-by: Rahul Bhansali
---
Changes in v2:
- updated patch title, and moved extern of ip_reassembly_dynflag next to
ip_reassembly_dynfield_offset.
doc/guides/sample_app_ug/ipsec_secgw.rst | 12 ++-
examples/ipsec-secgw/ipsec-secgw.c | 30 +
examples/ipsec-secgw/ipsec-secgw.h | 43 +++-
examples/ipsec-secgw/ipsec.c | 2 ++
examples/ipsec-secgw/ipsec.h | 4 ++-
examples/ipsec-secgw/ipsec_worker.c | 33 +-
examples/ipsec-secgw/ipsec_worker.h | 5 +++
examples/ipsec-secgw/sa.c| 12 ++-
8 files changed, 129 insertions(+), 12 deletions(-)
diff --git a/doc/guides/sample_app_ug/ipsec_secgw.rst
b/doc/guides/sample_app_ug/ipsec_secgw.rst
index 2a1aeae7c5..4d889e8496 100644
--- a/doc/guides/sample_app_ug/ipsec_secgw.rst
+++ b/doc/guides/sample_app_ug/ipsec_secgw.rst
@@ -537,7 +537,7 @@ The SA rule syntax is shown as follows:
sa
-
+
where each options means:
@@ -793,6 +793,16 @@ where each options means:
* *esn N* N is the initial ESN value
+
+
+ * Option to enable HW reassembly per SA.
+
+ * Optional: Yes, it is disabled by default
+
+ * Syntax:
+
+ * *reassembly_en*
+
Example SA rules:
.. code-block:: console
diff --git a/examples/ipsec-secgw/ipsec-secgw.c
b/examples/ipsec-secgw/ipsec-secgw.c
index 1d9982a786..24d895451a 100644
--- a/examples/ipsec-secgw/ipsec-secgw.c
+++ b/examples/ipsec-secgw/ipsec-secgw.c
@@ -274,6 +274,7 @@ static void
print_stats_cb(__rte_unused void *param)
{
uint64_t total_packets_dropped, total_packets_tx, total_packets_rx;
+ uint64_t total_frag_packets_dropped = 0;
float burst_percent, rx_per_call, tx_per_call;
unsigned int coreid;
@@ -303,6 +304,7 @@ print_stats_cb(__rte_unused void *param)
"\nPackets received: %20"PRIu64
"\nPackets sent: %24"PRIu64
"\nPackets dropped: %21"PRIu64
+ "\nFrag Packets dropped: %16"PRIu64
"\nBurst percent: %23.2f"
"\nPackets per Rx call: %17.2f"
"\nPackets per Tx call: %17.2f",
@@ -310,21 +312,25 @@ print_stats_cb(__rte_unused void *param)
core_statistics[coreid].rx,
core_statistics[coreid].tx,
core_statistics[coreid].dropped,
+ core_statistics[coreid].frag_dropped,
burst_percent,
rx_per_call,
tx_per_call);
total_packets_dropped += core_statistics[coreid].dropped;
+ total_frag_packets_dropped +=
core_statistics[coreid].frag_dropped;
total_packets_tx += core_statistics[coreid].tx;
total_packets_rx += core_statistics[coreid].rx;
}
printf("\nAggregate statistics ==="
"\nTotal packets received: %14"PRIu64
"\nTotal packets sent: %18"PRIu64
- "\nTotal packets dropped: %15"PRIu64,
+ "\nTotal packets dropped: %15"PRIu64
+ "\nTotal frag packets dropped: %10"PRIu64,
total_packets_rx,
total_packets_tx,
- total_packets_dropped);
+ total_packets_dropped,
+ total_frag_packets_dropped);
printf("\n\n");
rte_eal_alarm_set(stats_interval * US_PER_S, print_stats_cb, NULL);
@@ -1867,7 +1873,8 @@ parse_ptype_cb(uint16_t port __rte_unused, uint16_t queue
__rte_unused,
}
static void
-port_init(uint16_t portid, uint64_t req_rx_offloads, uint64_t req_tx_offloads)
+port_init(uint16_t portid, uint64_t req_rx_offloads, uint64_t req_tx_offloads,
+ uint8_t hw_reassembly)
{
struct rte_eth_dev_info dev_info;
struct rte_eth_txconf *txconf;
@@ -1877,6 +1884,7 @@ port_init(uint16_t portid, uint64_t req_rx_offloads,
uint64_t req_tx_offloads)
struct lcore_conf *qconf;
struct rte_ether_addr ethaddr;
struct rte_eth_conf local_port_conf = port_conf;
+ struct rte_eth_ip_reassembly_params reass_capa = {0};
int ptype_supported;
ret = rte_eth_dev_info_get(portid, &dev_info);
@@ -2053,6 +2061,12 @@ port_init(uint16_t portid, uint64_t req_rx_offloads,
uint64_t req_tx_offloads)
}
}
+
+ if (hw_reassembly)
RE: [PATCH v4 1/2] app/testpmd: fix vlan offload of rxq
Hi All,
Could you please review and provide suggestions if any.
Thanks,
Mingjin
> -Original Message-
> From: Ye, MingjinX
> Sent: 2022年10月25日 1:27
> To: dev@dpdk.org
> Cc: sta...@dpdk.org; Zhou, YidingX ; Ye, MingjinX
> ; Singh, Aman Deep
> ; Zhang, Yuying
> Subject: [PATCH v4 1/2] app/testpmd: fix vlan offload of rxq
>
> After setting vlan offload in testpmd, the result is not updated to rxq.
> Therefore, the queue needs to be reconfigured after executing the "vlan
> offload" related commands.
>
> Fixes: a47aa8b97afe ("app/testpmd: add vlan offload support")
> Cc: sta...@dpdk.org
>
> Signed-off-by: Mingjin Ye
> ---
> app/test-pmd/cmdline.c | 1 +
> 1 file changed, 1 insertion(+)
>
> diff --git a/app/test-pmd/cmdline.c b/app/test-pmd/cmdline.c index
> 17be2de402..ce125f549f 100644
> --- a/app/test-pmd/cmdline.c
> +++ b/app/test-pmd/cmdline.c
> @@ -4133,6 +4133,7 @@ cmd_vlan_offload_parsed(void *parsed_result,
> else
> vlan_extend_set(port_id, on);
>
> + cmd_reconfig_device_queue(port_id, 1, 1);
> return;
> }
>
> --
> 2.34.1
Re: [PATCH 0/3] remove more experimental tags
On Tue, Oct 18, 2022 at 3:58 AM Stephen Hemminger wrote: > > Remove some of the older experimental tags for 22.11 > > Stephen Hemminger (3): > eal: remove experimental from rte_epoll_wait_interruptible > eal: make rte_log_list_types not experimental > rwlock: make trylock operations no longer experimental > > lib/eal/include/generic/rte_rwlock.h | 12 ++-- > lib/eal/include/rte_epoll.h | 1 - > lib/eal/include/rte_log.h| 4 > lib/eal/version.map | 4 ++-- > 4 files changed, 4 insertions(+), 17 deletions(-) Series applied. I added to my todolist that we need a cleanup for unneeded inclusion of rte_compat.h (the headers touched in this series are not the only one concerned). -- David Marchand
Re: [PATCH] ci: combine static and shared linking build tests
On Thu, Oct 20, 2022 at 5:34 PM Aaron Conole wrote: > > Save some cpu time and disk by testing linking against static and shared > > library in single environments. > > > > The .ci/linux-build.sh is modified so it reconfigures an existing build > > directory: an empty DEF_LIB= means that static and shared builds are > > to be tested. > > > > ABI checks, documentation generation and unit tests are disabled for > > static builds as they would be redundant with the check against > > dynamically linked binaries, if any. > > > > Note: > > - --cross-file is an option that can be passed to meson only when > > creating a build environment, > > - for some other reason, --buildtype and other non -D options are only > > accepted when setting up a build directory with meson. When > > reconfiguring, only their -D$option forms are accepted, > > > > Signed-off-by: David Marchand > Acked-by: Bruce Richardson > Acked-by: Aaron Conole Applied, thanks. -- David Marchand
Re: [PATCH] mempool: cache align mempool cache objects
On Thu, Oct 27, 2022 at 11:22:07AM +0200, Morten Brørup wrote: > > From: Olivier Matz [mailto:olivier.m...@6wind.com] > > Sent: Thursday, 27 October 2022 10.35 > > > > Hi Morten, > > > > On Wed, Oct 26, 2022 at 04:44:36PM +0200, Morten Brørup wrote: > > > Add __rte_cache_aligned to the objs array. > > > > > > It makes no difference in the general case, but if get/put operations > > are > > > always 32 objects, it will reduce the number of memory (or last level > > > cache) accesses from five to four 64 B cache lines for every get/put > > > operation. > > > > > > For readability reasons, an example using 16 objects follows: > > > > > > Currently, with 16 objects (128B), we access to 3 > > > cache lines: > > > > > > ┌┐ > > > │len │ > > > cache ││--- > > > line0 ││ ^ > > > ││ | > > > ├┤ | 16 objects > > > ││ | 128B > > > cache ││ | > > > line1 ││ | > > > ││ | > > > ├┤ | > > > ││_v_ > > > cache ││ > > > line2 ││ > > > ││ > > > └┘ > > > > > > With the alignment, it is also 3 cache lines: > > > > > > ┌┐ > > > │len │ > > > cache ││ > > > line0 ││ > > > ││ > > > ├┤--- > > > ││ ^ > > > cache ││ | > > > line1 ││ | > > > ││ | > > > ├┤ | 16 objects > > > ││ | 128B > > > cache ││ | > > > line2 ││ | > > > ││ v > > > └┘--- > > > > > > However, accessing the objects at the bottom of the mempool cache is > > a > > > special case, where cache line0 is also used for objects. > > > > > > Consider the next burst (and any following bursts): > > > > > > Current: > > > ┌┐ > > > │len │ > > > cache ││ > > > line0 ││ > > > ││ > > > ├┤ > > > ││ > > > cache ││ > > > line1 ││ > > > ││ > > > ├┤ > > > ││ > > > cache ││--- > > > line2 ││ ^ > > > ││ | > > > ├┤ | 16 objects > > > ││ | 128B > > > cache ││ | > > > line3 ││ | > > > ││ | > > > ├┤ | > > > ││_v_ > > > cache ││ > > > line4 ││ > > > ││ > > > └┘ > > > 4 cache lines touched, incl. line0 for len. > > > > > > With the proposed alignment: > > > ┌┐ > > > │len │ > > > cache ││ > > > line0 ││ > > > ││ > > > ├┤ > > > ││ > > > cache ││ > > > line1 ││ > > > ││ > > > ├┤ > > > ││ > > > cache ││ > > > line2 ││ > > > ││ > > > ├┤ > > > ││--- > > > cache ││ ^ > > > line3 ││ | > > > ││ | 16 objects > > > ├┤ | 128B > > > ││ | > > > cache ││ | > > > line4 ││ | > > > ││_v_ > > > └┘ > > > Only 3 cache lines touched, incl. line0 for len. > > > > I understand your logic, but are we sure that having an application > > that > > works with bulks of 32 means that the cache will stay aligned to 32 > > elements for the whole life of the application? > > > > In an application, the alignment of the cache can change if you have > > any of: > > - software queues (reassembly for instance) > > - packet duplication (bridge, multicast) > > - locally generated packets (keepalive, control protocol) > > - pipeline to other cores > > > > Even with testpmd, which work by bulk of 32, I can see that the size > > of the cache filling is not aligned to 32. Right after starting the > > application, we already have this: > > > > internal cache infos: > > cache_size=250 > > cache_count[0]=231 > > > > This is probably related to the hw rx rings size, number of queues, > > number of ports. > > > > The "250" default value for cache size in testpmd is questionable, but > > with --mbcache=256, the behavior is similar. > > > > Also, when we transmit to a NIC, the mbufs are not returned immediatly > > to the pool, they may stay in the hw tx ring during some time, which is > > a driver decision. > > > > After processing traffic on cores 8 and 24 with this testpmd, I get: > > cache_count[0]=231 > > cache_count[8]=123 > > cache_count[24]=122 > > > > In my opinion, it is not realistic to think that the mempool cache will > > remain aligned to cachelines. In these conditions, it looks better to > > keep the structure packed to avoid wasting memory. > > I agree that is a special use case to only access the mempool cache in > bursts of 32 objects, so the accesses are always cache line > aligned. (Generalized, the burst size must not be 32; a burst size > that is
Re: [PATCH] telemetry: make usable from C++
On Wed, Oct 12, 2022 at 2:05 PM Markus Theil wrote: > > From: Markus Theil > > Add missing include in order to make C++ compilers > happy. > We have build checks for headers (-Dcheck_includes meson option), like for C++: $ cat $BUILDDIR/buildtools/chkincs/chkincs-cpp.p/rte_telemetry.cpp #include "/home/dmarchan/dpdk/lib/telemetry/rte_telemetry.h" [917/1289] ccache c++ -Ibuildtools/chkincs/chkincs-cpp.p -Ibuildtools/chkincs -I../../../dpdk/buildtools/chkincs -Iexamples/vmdq_dcb -I../../../dpdk/examples/vmdq_dcb -I../../../dpdk/examples/common -Idrivers/bus/vdev -I../../../dpdk/drivers/bus/vdev -I. -I../../../dpdk -Iconfig -I../../../dpdk/config -Ilib/eal/include -I../../../dpdk/lib/eal/include -Ilib/eal/linux/include -I../../../dpdk/lib/eal/linux/include -Ilib/eal/x86/include -I../../../dpdk/lib/eal/x86/include -Ilib/eal/common -I../../../dpdk/lib/eal/common -Ilib/eal -I../../../dpdk/lib/eal -Ilib/kvargs -I../../../dpdk/lib/kvargs -Ilib/metrics -I../../../dpdk/lib/metrics -Ilib/telemetry -I../../../dpdk/lib/telemetry -Idrivers/bus/pci -I../../../dpdk/drivers/bus/pci -I../../../dpdk/drivers/bus/pci/linux -Ilib/pci -I../../../dpdk/lib/pci -Ilib/ring -I../../../dpdk/lib/ring -Ilib/rcu -I../../../dpdk/lib/rcu -Ilib/mempool -I../../../dpdk/lib/mempool -Ilib/mbuf -I../../../dpdk/lib/mbuf -Ilib/net -I../../../dpdk/lib/net -Ilib/meter -I../../../dpdk/lib/meter -Ilib/ethdev -I../../../dpdk/lib/ethdev -Ilib/cmdline -I../../../dpdk/lib/cmdline -Ilib/hash -I../../../dpdk/lib/hash -Ilib/timer -I../../../dpdk/lib/timer -Ilib/acl -I../../../dpdk/lib/acl -Ilib/bbdev -I../../../dpdk/lib/bbdev -Ilib/bitratestats -I../../../dpdk/lib/bitratestats -Ilib/bpf -I../../../dpdk/lib/bpf -Ilib/cfgfile -I../../../dpdk/lib/cfgfile -Ilib/compressdev -I../../../dpdk/lib/compressdev -Ilib/cryptodev -I../../../dpdk/lib/cryptodev -Ilib/distributor -I../../../dpdk/lib/distributor -Ilib/efd -I../../../dpdk/lib/efd -Ilib/eventdev -I../../../dpdk/lib/eventdev -Ilib/gpudev -I../../../dpdk/lib/gpudev -Ilib/gro -I../../../dpdk/lib/gro -Ilib/gso -I../../../dpdk/lib/gso -Ilib/ip_frag -I../../../dpdk/lib/ip_frag -Ilib/jobstats -I../../../dpdk/lib/jobstats -Ilib/kni -I../../../dpdk/lib/kni -Ilib/latencystats -I../../../dpdk/lib/latencystats -Ilib/lpm -I../../../dpdk/lib/lpm -Ilib/member -I../../../dpdk/lib/member -Ilib/pcapng -I../../../dpdk/lib/pcapng -Ilib/power -I../../../dpdk/lib/power -Ilib/rawdev -I../../../dpdk/lib/rawdev -Ilib/regexdev -I../../../dpdk/lib/regexdev -Ilib/dmadev -I../../../dpdk/lib/dmadev -Ilib/rib -I../../../dpdk/lib/rib -Ilib/reorder -I../../../dpdk/lib/reorder -Ilib/sched -I../../../dpdk/lib/sched -Ilib/security -I../../../dpdk/lib/security -Ilib/stack -I../../../dpdk/lib/stack -Ilib/vhost -I../../../dpdk/lib/vhost -Ilib/ipsec -I../../../dpdk/lib/ipsec -Ilib/fib -I../../../dpdk/lib/fib -Ilib/port -I../../../dpdk/lib/port -Ilib/pdump -I../../../dpdk/lib/pdump -Ilib/table -I../../../dpdk/lib/table -Ilib/pipeline -I../../../dpdk/lib/pipeline -Ilib/flow_classify -I../../../dpdk/lib/flow_classify -Ilib/graph -I../../../dpdk/lib/graph -Ilib/node -I../../../dpdk/lib/node -fdiagnostics-color=always -D_FILE_OFFSET_BITS=64 -Wall -Winvalid-pch -Wnon-virtual-dtor -Wextra -Werror -O2 -g -include rte_config.h -march=nehalem -DALLOW_EXPERIMENTAL_API -DALLOW_INTERNAL_API -MD -MQ buildtools/chkincs/chkincs-cpp.p/meson-generated_rte_telemetry.cpp.o -MF buildtools/chkincs/chkincs-cpp.p/meson-generated_rte_telemetry.cpp.o.d -o buildtools/chkincs/chkincs-cpp.p/meson-generated_rte_telemetry.cpp.o -c buildtools/chkincs/chkincs-cpp.p/rte_telemetry.cpp Besides, I fail to see the need for stddef.h. In which setup / case, did you get a compilation issue? -- David Marchand
RE: [PATCH] mempool: cache align mempool cache objects
> From: Olivier Matz [mailto:olivier.m...@6wind.com] > Sent: Thursday, 27 October 2022 13.43 > > On Thu, Oct 27, 2022 at 11:22:07AM +0200, Morten Brørup wrote: > > > From: Olivier Matz [mailto:olivier.m...@6wind.com] > > > Sent: Thursday, 27 October 2022 10.35 > > > > > > Hi Morten, > > > > > > On Wed, Oct 26, 2022 at 04:44:36PM +0200, Morten Brørup wrote: > > > > Add __rte_cache_aligned to the objs array. > > > > > > > > It makes no difference in the general case, but if get/put > operations > > > are > > > > always 32 objects, it will reduce the number of memory (or last > level > > > > cache) accesses from five to four 64 B cache lines for every > get/put > > > > operation. > > > > > > > > For readability reasons, an example using 16 objects follows: > > > > > > > > Currently, with 16 objects (128B), we access to 3 > > > > cache lines: > > > > > > > > ┌┐ > > > > │len │ > > > > cache ││--- > > > > line0 ││ ^ > > > > ││ | > > > > ├┤ | 16 objects > > > > ││ | 128B > > > > cache ││ | > > > > line1 ││ | > > > > ││ | > > > > ├┤ | > > > > ││_v_ > > > > cache ││ > > > > line2 ││ > > > > ││ > > > > └┘ > > > > > > > > With the alignment, it is also 3 cache lines: > > > > > > > > ┌┐ > > > > │len │ > > > > cache ││ > > > > line0 ││ > > > > ││ > > > > ├┤--- > > > > ││ ^ > > > > cache ││ | > > > > line1 ││ | > > > > ││ | > > > > ├┤ | 16 objects > > > > ││ | 128B > > > > cache ││ | > > > > line2 ││ | > > > > ││ v > > > > └┘--- > > > > > > > > However, accessing the objects at the bottom of the mempool cache > is > > > a > > > > special case, where cache line0 is also used for objects. > > > > > > > > Consider the next burst (and any following bursts): > > > > > > > > Current: > > > > ┌┐ > > > > │len │ > > > > cache ││ > > > > line0 ││ > > > > ││ > > > > ├┤ > > > > ││ > > > > cache ││ > > > > line1 ││ > > > > ││ > > > > ├┤ > > > > ││ > > > > cache ││--- > > > > line2 ││ ^ > > > > ││ | > > > > ├┤ | 16 objects > > > > ││ | 128B > > > > cache ││ | > > > > line3 ││ | > > > > ││ | > > > > ├┤ | > > > > ││_v_ > > > > cache ││ > > > > line4 ││ > > > > ││ > > > > └┘ > > > > 4 cache lines touched, incl. line0 for len. > > > > > > > > With the proposed alignment: > > > > ┌┐ > > > > │len │ > > > > cache ││ > > > > line0 ││ > > > > ││ > > > > ├┤ > > > > ││ > > > > cache ││ > > > > line1 ││ > > > > ││ > > > > ├┤ > > > > ││ > > > > cache ││ > > > > line2 ││ > > > > ││ > > > > ├┤ > > > > ││--- > > > > cache ││ ^ > > > > line3 ││ | > > > > ││ | 16 objects > > > > ├┤ | 128B > > > > ││ | > > > > cache ││ | > > > > line4 ││ | > > > > ││_v_ > > > > └┘ > > > > Only 3 cache lines touched, incl. line0 for len. > > > > > > I understand your logic, but are we sure that having an application > > > that > > > works with bulks of 32 means that the cache will stay aligned to 32 > > > elements for the whole life of the application? > > > > > > In an application, the alignment of the cache can change if you > have > > > any of: > > > - software queues (reassembly for instance) > > > - packet duplication (bridge, multicast) > > > - locally generated packets (keepalive, control protocol) > > > - pipeline to other cores > > > > > > Even with testpmd, which work by bulk of 32, I can see that the > size > > > of the cache filling is not aligned to 32. Right after starting the > > > application, we already have this: > > > > > > internal cache infos: > > > cache_size=250 > > > cache_count[0]=231 > > > > > > This is probably related to the hw rx rings size, number of queues, > > > number of ports. > > > > > > The "250" default value for cache size in testpmd is questionable, > but > > > with --mbcache=256, the behavior is similar. > > > > > > Also, when we transmit to a NIC, the mbufs are not returned > immediatly > > > to the pool, they may stay in the hw tx ring during some time, > which is > > > a driver decision. > > > > > > After processing traffic on cores 8 and 24 with this testpmd, I > get: > > > cache_count[0]=231 > > > cache_count[8]=123 > > > cache_count[24]=122 > > > > > >
RE: [PATCH] flow_classify: mark library as deprecated
> > This library has no maintainer and, for now, nobody expressed interest > in taking over. > Mark this experimental library as deprecated and announce plan for > removal in v23.11. > > Signed-off-by: David Marchand > --- Acked-by: Konstantin Ananyev > -- > 2.37.3
Re: [PATCH v3 1/1] app/testpmd: control passing Rx metadata to PMD
27/10/2022 09:34, Thomas Monjalon: > 17/10/2022 10:32, Andrew Rybchenko: > > On 10/6/22 21:35, Hanumanth Pothula wrote: > > > Presently, Rx metadata is sent to PMD by default, leading > > > to a performance drop as processing for the same in rx path > > > takes extra cycles. > > > > > > Hence, introducing command line argument, 'nic-to-pmd-rx-metadata' > > > to control passing rx metadata to PMD. By default it’s disabled. > > > > > > Signed-off-by: Hanumanth Pothula > > > > Acked-by: Andrew Rybchenko > > > > Applied to dpdk-next-net/main, thanks. > > I'm not sure this patch is really acceptable. > It is disabling Rx metadata by default just for benchmarking reason > because your driver is doing some processing even if metadata is not required. > > From a user perspective, if a command requesting metadata is entered, > it won't work until we enable this new option on startup. > It looks terrible. > > Please tell me I misunderstood something. While pulling, I see that the name is not compliant with others. I think it should start with "enable-", use hyphens and be sorted. I'll drop it from the pull for now, we can have it in -rc3.
RE: [PATCH v1] crypto/qat: fix null hash algorithm digest size
> -Original Message-
> From: Brian Dooley
> Sent: Thursday 27 October 2022 11:50
> To: Ji, Kai
> Cc: dev@dpdk.org; sta...@dpdk.org; gak...@marvell.com; Kusztal,
> ArkadiuszX ; Dooley, Brian
> ; Doherty, Declan
> Subject: [PATCH v1] crypto/qat: fix null hash algorithm digest size
>
> Add check for null hash algorithm digest size. Digest size should be 4B or
> request will be rejected.
>
> Fixes: 1703e94ac5ce ("qat: add driver for QuickAssist devices")
> Cc: declan.dohe...@intel.com
> Cc: sta...@dpdk.org
> Signed-off-by: Brian Dooley
> ---
Acked-by: Ciara Power
Re: [PATCH v4 1/2] build: allow to conditionally build apps
On Fri, Oct 14, 2022 at 9:51 AM Markus Theil wrote: > > Makes apps configurable from meson, like already > possible for drivers. > > Signed-off-by: Markus Theil With these new options, it's hard to tell which and why some application is built (or not). Can we have an output similar to libraries and drivers (see toplevel meson.build) ? -- David Marchand
Re: [PATCH v4 1/2] app/testpmd: fix vlan offload of rxq
On 10/27/2022 2:06 PM, Huang, ZhiminX wrote:
-Original Message-
From: Mingjin Ye
Sent: Thursday, October 27, 2022 1:10 AM
To: dev@dpdk.org
Cc: sta...@dpdk.org; Zhou, YidingX ; Ye, MingjinX
; Singh, Aman Deep ;
Zhang, Yuying
Subject: [PATCH v4 1/2] app/testpmd: fix vlan offload of rxq
After setting vlan offload in testpmd, the result is not updated to rxq.
Therefore,
the queue needs to be reconfigured after executing the "vlan offload" related
commands.
Fixes: a47aa8b97afe ("app/testpmd: add vlan offload support")
Cc: sta...@dpdk.org
Signed-off-by: Mingjin Ye
Acked-by: Aman Singh
---
Tested-by: Zhimin Huang
[PATCH v1] dts: fix ssh timeout and output
We were always waiting the full timeout because we always waited for
the MAGIC_PROMPT, which is only usable for cleaning the session of
previous data. That also meant that the actual prompt was appended to
the output, resulting in output being '[PEXPECT]#' when the executed
command produced empty output. Both are fixed by using the MAGIC_PROMPT
only in self._clean_session.
Fixes: e81a070a1c7d ("dts: add ssh session module")
Signed-off-by: Juraj Linkeš
---
dts/framework/remote_session/ssh_session.py | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/dts/framework/remote_session/ssh_session.py
b/dts/framework/remote_session/ssh_session.py
index f71acfb1ca..dc37e9772b 100644
--- a/dts/framework/remote_session/ssh_session.py
+++ b/dts/framework/remote_session/ssh_session.py
@@ -116,7 +116,9 @@ def send_expect_base(self, command: str, prompt: str,
timeout: float) -> str:
return before
def _clean_session(self) -> None:
+self.session.PROMPT = self.magic_prompt
self.get_output(timeout=0.01)
+self.session.PROMPT = self.session.UNIQUE_PROMPT
def _send_line(self, command: str) -> None:
if not self.is_alive():
@@ -134,7 +136,6 @@ def get_output(self, timeout: float = 15) -> str:
"""
Get all output before timeout
"""
-self.session.PROMPT = self.magic_prompt
try:
self.session.prompt(timeout)
except Exception:
@@ -143,7 +144,6 @@ def get_output(self, timeout: float = 15) -> str:
before = self._get_output()
self._flush()
-self.logger.debug(before)
return before
def _get_output(self) -> str:
--
2.30.2
RE: [PATCH v1] crypto/qat: fix null hash algorithm digest size
> > Add check for null hash algorithm digest size. Digest size should be 4B or
> > request will be rejected.
> >
> > Fixes: 1703e94ac5ce ("qat: add driver for QuickAssist devices")
> > Cc: declan.dohe...@intel.com
> > Cc: sta...@dpdk.org
> > Signed-off-by: Brian Dooley
> > ---
>
> Acked-by: Ciara Power
Applied to dpdk-next-crypto
Thanks.
RE: [PATCH v2] examples/ipsec-secgw: support per SA HW reassembly
> Subject: [PATCH v2] examples/ipsec-secgw: support per SA HW reassembly > > This add the support of hardware reassembly per SA basis. > In SA rule, new parameter reassembly_en is added to enable > HW reassembly per SA. > For example: > sa in aead_algo aead_key mode ipv4-tunnel src > dst type inline-protocol-offload port_id reassembly_en > > Stats counter frag_dropped will represent the number of fragment > drop in case of reassembly failures. > > Signed-off-by: Rahul Bhansali > --- Acked-by: Akhil Goyal Applied to dpdk-next-crypto Thanks.
Re: [PATCH] flow_classify: mark library as deprecated
27/10/2022 11:46, David Marchand: > This library has no maintainer and, for now, nobody expressed interest > in taking over. > Mark this experimental library as deprecated and announce plan for > removal in v23.11. > > Signed-off-by: David Marchand It adds enough warnings to reach potential users. Acked-by: Thomas Monjalon
Re: [PATCH v4 1/2] build: allow to conditionally build apps
On Thu, Oct 27, 2022 at 3:13 PM David Marchand wrote: > > On Fri, Oct 14, 2022 at 9:51 AM Markus Theil > wrote: > > > > Makes apps configurable from meson, like already > > possible for drivers. > > > > Signed-off-by: Markus Theil > > With these new options, it's hard to tell which and why some > application is built (or not). > Can we have an output similar to libraries and drivers (see toplevel > meson.build) ? WDYT of followup patch: https://github.com/david-marchand/dpdk/commit/apps -- David Marchand
RE: [EXT] [PATCH v2 1/1] doc: update to include bbdev snippet using literalinclude
> Adding code snippet using literalinclude so that to keep > automatically these structures in doc in sync with the > bbdev source code. > > Signed-off-by: Nicolas Chautru Applied to dpdk-next-crypto Thanks.
Re: [PATCH] mempool: cache align mempool cache objects
On Thu, Oct 27, 2022 at 02:11:29PM +0200, Morten Brørup wrote: > > From: Olivier Matz [mailto:olivier.m...@6wind.com] > > Sent: Thursday, 27 October 2022 13.43 > > > > On Thu, Oct 27, 2022 at 11:22:07AM +0200, Morten Brørup wrote: > > > > From: Olivier Matz [mailto:olivier.m...@6wind.com] > > > > Sent: Thursday, 27 October 2022 10.35 > > > > > > > > Hi Morten, > > > > > > > > On Wed, Oct 26, 2022 at 04:44:36PM +0200, Morten Brørup wrote: > > > > > Add __rte_cache_aligned to the objs array. > > > > > > > > > > It makes no difference in the general case, but if get/put > > operations > > > > are > > > > > always 32 objects, it will reduce the number of memory (or last > > level > > > > > cache) accesses from five to four 64 B cache lines for every > > get/put > > > > > operation. > > > > > > > > > > For readability reasons, an example using 16 objects follows: > > > > > > > > > > Currently, with 16 objects (128B), we access to 3 > > > > > cache lines: > > > > > > > > > > ┌┐ > > > > > │len │ > > > > > cache ││--- > > > > > line0 ││ ^ > > > > > ││ | > > > > > ├┤ | 16 objects > > > > > ││ | 128B > > > > > cache ││ | > > > > > line1 ││ | > > > > > ││ | > > > > > ├┤ | > > > > > ││_v_ > > > > > cache ││ > > > > > line2 ││ > > > > > ││ > > > > > └┘ > > > > > > > > > > With the alignment, it is also 3 cache lines: > > > > > > > > > > ┌┐ > > > > > │len │ > > > > > cache ││ > > > > > line0 ││ > > > > > ││ > > > > > ├┤--- > > > > > ││ ^ > > > > > cache ││ | > > > > > line1 ││ | > > > > > ││ | > > > > > ├┤ | 16 objects > > > > > ││ | 128B > > > > > cache ││ | > > > > > line2 ││ | > > > > > ││ v > > > > > └┘--- > > > > > > > > > > However, accessing the objects at the bottom of the mempool cache > > is > > > > a > > > > > special case, where cache line0 is also used for objects. > > > > > > > > > > Consider the next burst (and any following bursts): > > > > > > > > > > Current: > > > > > ┌┐ > > > > > │len │ > > > > > cache ││ > > > > > line0 ││ > > > > > ││ > > > > > ├┤ > > > > > ││ > > > > > cache ││ > > > > > line1 ││ > > > > > ││ > > > > > ├┤ > > > > > ││ > > > > > cache ││--- > > > > > line2 ││ ^ > > > > > ││ | > > > > > ├┤ | 16 objects > > > > > ││ | 128B > > > > > cache ││ | > > > > > line3 ││ | > > > > > ││ | > > > > > ├┤ | > > > > > ││_v_ > > > > > cache ││ > > > > > line4 ││ > > > > > ││ > > > > > └┘ > > > > > 4 cache lines touched, incl. line0 for len. > > > > > > > > > > With the proposed alignment: > > > > > ┌┐ > > > > > │len │ > > > > > cache ││ > > > > > line0 ││ > > > > > ││ > > > > > ├┤ > > > > > ││ > > > > > cache ││ > > > > > line1 ││ > > > > > ││ > > > > > ├┤ > > > > > ││ > > > > > cache ││ > > > > > line2 ││ > > > > > ││ > > > > > ├┤ > > > > > ││--- > > > > > cache ││ ^ > > > > > line3 ││ | > > > > > ││ | 16 objects > > > > > ├┤ | 128B > > > > > ││ | > > > > > cache ││ | > > > > > line4 ││ | > > > > > ││_v_ > > > > > └┘ > > > > > Only 3 cache lines touched, incl. line0 for len. > > > > > > > > I understand your logic, but are we sure that having an application > > > > that > > > > works with bulks of 32 means that the cache will stay aligned to 32 > > > > elements for the whole life of the application? > > > > > > > > In an application, the alignment of the cache can change if you > > have > > > > any of: > > > > - software queues (reassembly for instance) > > > > - packet duplication (bridge, multicast) > > > > - locally generated packets (keepalive, control protocol) > > > > - pipeline to other cores > > > > > > > > Even with testpmd, which work by bulk of 32, I can see that the > > size > > > > of the cache filling is not aligned to 32. Right after starting the > > > > application, we already have this: > > > > > > > > internal cache infos: > > > > cache_size=250 > > > > cache_count[0]=231 > > > > > > > > This is probably related to the hw rx rings size, number of queues, > > > > number of ports. > > > > > > > > The "250" default value for cache size in testpmd is questionable, > > but > > > > with --mbcache=256, the behavior is similar.
Re: [PATCH] app/testpmd: fix testpmd receive jumbo frame packets
Hi Jie,
Thanks for the patch.
On 10/25/2022 7:35 AM, Jie Wang wrote:
For NIC I40E_10G-10G_BASE_T_X722, when testpmd is configured with
link speed, it cannot receive jumbo frame packets.
Why only the jumbo frame are effected and not other pkts.
Because it has changed the link status of the ports if it was
configured with link speed.
If we configure to same link speed, will it still have issue ?
When exiting testpmd that it automatically stops packet forwarding
and stops all the ports. But it doesn't update the link status of the
ports. If stop the ports first that it will update the link status.
This patch fix the error that testpmd will update the link status of
the ports when it exits.
Fixes: d3a274ce9dee ("app/testpmd: handle SIGINT and SIGTERM")
Fixes: 284c908cc588 ("app/testpmd: request device removal interrupt")
Cc: sta...@dpdk.org
Signed-off-by: Jie Wang
---
app/test-pmd/testpmd.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/app/test-pmd/testpmd.c b/app/test-pmd/testpmd.c
index 97adafacd0..c348a3f328 100644
--- a/app/test-pmd/testpmd.c
+++ b/app/test-pmd/testpmd.c
@@ -3548,7 +3548,7 @@ pmd_test_exit(void)
}
#endif
if (ports != NULL) {
- no_link_check = 1;
+ no_link_check = 0;
RTE_ETH_FOREACH_DEV(pt_id) {
printf("\nStopping port %d...\n", pt_id);
fflush(stdout);
@@ -3675,7 +3675,7 @@ rmv_port_callback(void *arg)
need_to_start = 1;
stop_packet_forwarding();
}
- no_link_check = 1;
+ no_link_check = 0;
Well, here we are undoing a previous change done for 284c908cc588
Won't that issue come back.
stop_port(port_id);
no_link_check = org_no_link_check;
[pull-request] next-crypto 22.11 rc2
The following changes since commit 9082336048e8c5779c1b4f8930041a42e9197348: net/nfp: support new solution for tunnel decap action (2022-10-25 10:53:33 +0200) are available in the Git repository at: http://dpdk.org/git/next/dpdk-next-crypto for you to fetch changes up to ee18061c78af2997585c7a59c03bae397f16: doc: include bbdev code snippet using literalinclude (2022-10-27 19:54:50 +0530) Ali Alnubani (1): examples/l2fwd-crypto: fix typo in error message Anoob Joseph (5): crypto/cnxk: align HW accessible fields to ROC crypto/cnxk: fix failure from session rework crypto/cnxk: verify IV length during session create crypto/cnxk: remove dead code test/crypto: fix PDCP vectors Arek Kusztal (9): crypto/qat: fix uncleared cookies in asym crypto/qat: fix unnecessary session check crypto/qat: fix not set RSA lengths common/qat: read HW slice configuration crypto/qat: read HW slice configuration crypto/qat: make immutable parameters constant crypto/qat: improve asym cookies free crypto/qat: add ECDH key exchange algorithm crypto/qat: add ECDH public key verification Brian Dooley (5): crypto/qat: reallocate on OpenSSL version check examples/fips_validation: add parsing for AES GMAC examples/fips_validation: add parsing for AES CTR examples/fips_validation: fix parse block error crypto/qat: fix null hash algorithm digest size Ciara Power (6): test/crypto: fix wireless auth digest segment crypto/ipsec_mb: support remaining SGL test/crypto: add OOP SNOW3G SGL cases test/crypto: add remaining blockcipher SGL cases test/crypto-perf: fix number of scheduler sessions crypto/scheduler: fix op session in pkt size distr mode Gowrishankar Muthukrishnan (6): examples/fips_validation: add asymmetric validation examples/fips_validation: encode digest with hash OID examples/fips_validation: randomize message for conformance test crypto/cnxk: support exponent type private key examples/fips_validation: fix GMAC decryption output examples/fips_validation: validate ECDSA Hernan Vargas (28): baseband/acc100: fix ring availability calculation baseband/acc100: check AQ availability baseband/acc100: fix memory leak baseband/acc100: add LDPC encoder padding function baseband/acc100: check turbo dec/enc input baseband/acc100: add null checks baseband/acc100: enforce additional check on FCW baseband/acc100: fix ring/queue allocation baseband/acc100: fix input length for CRC24B baseband/acc100: fix clearing PF IR outside handler baseband/acc100: fix device min alignment baseband/acc100: fix null HARQ input case baseband/acc100: fix pointer after free baseband/acc100: update debug print for LDPC FCW baseband/acc100: add enqueue status baseband/acc100: support scatter-gather baseband/acc100: add HARQ index helper function baseband/acc100: enable input validation by default baseband/acc100: support LDPC transport block baseband/acc100: update LDPC enc/dec validation baseband/acc100: implement configurable queue depth baseband/acc100: add queue stop operation baseband/acc100: update uplink CB input length baseband/acc100: update log messages baseband/acc100: store FCW from first CB descriptor baseband/acc100: update device info baseband/acc100: add ring companion address baseband/acc100: configure PMON control registers Kai Ji (2): crypto/qat: fix multi-process build request session crypto/ipsec_mb: add multi-process IPC request handler Nic Chautru (16): baseband/acc100: remove unused registers baseband/acc100: refactor to segregate common code baseband/acc: rename directory from acc100 to acc baseband/acc: introduce ACC200 baseband/acc: add ACC200 HW register definitions baseband/acc: add ACC200 info get function baseband/acc: add ACC200 queue configuration baseband/acc: add LDPC processing functions baseband/acc: add LTE processing functions baseband/acc: support FFT operations baseband/acc: support interrupt baseband/acc: add device status and vf2pf comms baseband/acc: add PF configure companion function baseband/acc: simplify meson dependency baseband/acc: add descriptor index helpers test/bbdev: add FFT operations cases Nicolas Chautru (2): baseband/turbo_sw: remove Flexran SDK meson option doc: include bbdev code snippet using literalinclude Nithin Dabilpuram (1): app/test: add multi-seg inline IPsec cases Radu Nicolau (1): examples/ipsec-secgw: fix Tx checksum offload flag Rahul Bhansali (1): examples/ipsec-secgw: support per SA HW reassembly Srujana Challa
RE: [dpdk-dev v5] crypto/ipsec_mb: multi-process IPC request handler
>
> As the queue pair used in secondary process needs to be set up by
> the primary process, this patch adds an IPC register function to help
> secondary process to send out queue-pair setup request to primary
> process via IPC request messages. A new "qp_in_used_pid" param stores
> the PID to provide the ownership of the queue-pair so that only the PID
> matched queue-pair can be free'd in the request.
As I can see that patch was already merged, but still:
How we suppose to guarantee synchronization with such approach?
Let say secondary process sends an IPC message to configure
crypto-queue, and at the same moment primary process (for whatever reason)
decides to reconfigure same crypto-dev.
Is there any way to prevent such race-conditions to happen?
>
> Signed-off-by: Kai Ji
> Acked-by: Arek Kusztal
> Acked-by: Pablo de Lara
> ---
> v5:
> - minor updates and typo fix
>
> v4:
> - review comments resolved
>
> v3:
> - remove shared memzone as qp_conf params can be passed directly from
> ipc message.
>
> v2:
> - add in shared memzone for data exchange between multi-process
> ---
> drivers/crypto/ipsec_mb/ipsec_mb_ops.c | 129 -
> drivers/crypto/ipsec_mb/ipsec_mb_private.c | 24 +++-
> drivers/crypto/ipsec_mb/ipsec_mb_private.h | 38 +-
> 3 files changed, 185 insertions(+), 6 deletions(-)
>
> diff --git a/drivers/crypto/ipsec_mb/ipsec_mb_ops.c
> b/drivers/crypto/ipsec_mb/ipsec_mb_ops.c
> index cedcaa2742..bf18d692bd 100644
> --- a/drivers/crypto/ipsec_mb/ipsec_mb_ops.c
> +++ b/drivers/crypto/ipsec_mb/ipsec_mb_ops.c
> @@ -3,6 +3,7 @@
> */
>
> #include
> +#include
>
> #include
> #include
> @@ -93,6 +94,46 @@ ipsec_mb_info_get(struct rte_cryptodev *dev,
> }
> }
>
> +static int
> +ipsec_mb_secondary_qp_op(int dev_id, int qp_id,
> + const struct rte_cryptodev_qp_conf *qp_conf,
> + int socket_id, enum ipsec_mb_mp_req_type op_type)
> +{
> + int ret;
> + struct rte_mp_msg qp_req_msg;
> + struct rte_mp_msg *qp_resp_msg;
> + struct rte_mp_reply qp_resp;
> + struct ipsec_mb_mp_param *req_param;
> + struct ipsec_mb_mp_param *resp_param;
> + struct timespec ts = {.tv_sec = 1, .tv_nsec = 0};
> +
> + memset(&qp_req_msg, 0, sizeof(IPSEC_MB_MP_MSG));
> + memcpy(qp_req_msg.name, IPSEC_MB_MP_MSG, sizeof(IPSEC_MB_MP_MSG));
> + req_param = (struct ipsec_mb_mp_param *)&qp_req_msg.param;
> +
> + qp_req_msg.len_param = sizeof(struct ipsec_mb_mp_param);
> + req_param->type = op_type;
> + req_param->dev_id = dev_id;
> + req_param->qp_id = qp_id;
> + req_param->socket_id = socket_id;
> + req_param->process_id = getpid();
> + if (qp_conf) {
> + req_param->nb_descriptors = qp_conf->nb_descriptors;
> + req_param->mp_session = (void *)qp_conf->mp_session;
> + }
> +
> + qp_req_msg.num_fds = 0;
> + ret = rte_mp_request_sync(&qp_req_msg, &qp_resp, &ts);
> + if (ret) {
> + RTE_LOG(ERR, USER1, "Create MR request to primary process
> failed.");
> + return -1;
> + }
> + qp_resp_msg = &qp_resp.msgs[0];
> + resp_param = (struct ipsec_mb_mp_param *)qp_resp_msg->param;
> +
> + return resp_param->result;
> +}
> +
> /** Release queue pair */
> int
> ipsec_mb_qp_release(struct rte_cryptodev *dev, uint16_t qp_id)
> @@ -100,7 +141,10 @@ ipsec_mb_qp_release(struct rte_cryptodev *dev, uint16_t
> qp_id)
> struct ipsec_mb_qp *qp = dev->data->queue_pairs[qp_id];
> struct rte_ring *r = NULL;
>
> - if (qp != NULL && rte_eal_process_type() == RTE_PROC_PRIMARY) {
> + if (qp != NULL)
> + return 0;
> +
> + if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
> r = rte_ring_lookup(qp->name);
> rte_ring_free(r);
>
> @@ -115,6 +159,9 @@ ipsec_mb_qp_release(struct rte_cryptodev *dev, uint16_t
> qp_id)
> #endif
> rte_free(qp);
> dev->data->queue_pairs[qp_id] = NULL;
> + } else { /* secondary process */
> + return ipsec_mb_secondary_qp_op(dev->data->dev_id, qp_id,
> + NULL, 0, RTE_IPSEC_MB_MP_REQ_QP_FREE);
> }
> return 0;
> }
> @@ -222,9 +269,13 @@ ipsec_mb_qp_setup(struct rte_cryptodev *dev, uint16_t
> qp_id,
> #endif
> qp = dev->data->queue_pairs[qp_id];
> if (qp == NULL) {
> - IPSEC_MB_LOG(ERR, "Primary process hasn't configured
> device qp.");
> - return -EINVAL;
> + IPSEC_MB_LOG(DEBUG, "Secondary process setting up
> device qp.");
> + return ipsec_mb_secondary_qp_op(dev->data->dev_id,
> qp_id,
> + qp_conf, socket_id,
> RTE_IPSEC_MB_MP_REQ_QP_SET);
> }
> +
> + IPSEC_MB_LOG(ERR, "Queue pair already setup'ed.");
> + return -EINVAL;
> } else {
> /* Free memory
[PATCH] mempool: optimized debug statistics
When built with debug enabled (RTE_LIBRTE_MEMPOOL_DEBUG defined), the
performance of mempools with caches is improved as follows.
Accessing objects in the mempool is likely to increment either the
put_bulk and put_objs or the get_success_bulk and get_success_objs
debug statistics counters.
By adding an alternative set of these counters to the mempool cache
structure, accesing the dedicated debug statistics structure is avoided in
the likely cases where these counters are incremented.
The trick here is that the cache line holding the mempool cache structure
is accessed anyway, in order to update the "len" field. Updating some
debug statistics counters in the same cache line has lower performance
cost than accessing the debug statistics counters in the dedicated debug
statistics structure, i.e. in another cache line.
Running mempool_perf_autotest on a VMware virtual server shows an avg.
increase of 6.4 % in rate_persec for the tests with cache. (Only when
built with with debug enabled, obviously!)
For the tests without cache, the avg. increase in rate_persec is 0.8 %. I
assume this is noise from the test environment.
Signed-off-by: Morten Brørup
---
lib/mempool/rte_mempool.c | 7 +
lib/mempool/rte_mempool.h | 55 +++
2 files changed, 51 insertions(+), 11 deletions(-)
diff --git a/lib/mempool/rte_mempool.c b/lib/mempool/rte_mempool.c
index 21c94a2b9f..7b8c00a022 100644
--- a/lib/mempool/rte_mempool.c
+++ b/lib/mempool/rte_mempool.c
@@ -1285,6 +1285,13 @@ rte_mempool_dump(FILE *f, struct rte_mempool *mp)
sum.get_fail_objs += mp->stats[lcore_id].get_fail_objs;
sum.get_success_blks += mp->stats[lcore_id].get_success_blks;
sum.get_fail_blks += mp->stats[lcore_id].get_fail_blks;
+ /* Add the fast access statistics, if local caches exist */
+ if (mp->cache_size != 0) {
+ sum.put_bulk += mp->local_cache[lcore_id].put_bulk;
+ sum.put_objs += mp->local_cache[lcore_id].put_objs;
+ sum.get_success_bulk +=
mp->local_cache[lcore_id].get_success_bulk;
+ sum.get_success_objs +=
mp->local_cache[lcore_id].get_success_objs;
+ }
}
fprintf(f, " stats:\n");
fprintf(f, "put_bulk=%"PRIu64"\n", sum.put_bulk);
diff --git a/lib/mempool/rte_mempool.h b/lib/mempool/rte_mempool.h
index 3725a72951..d84087bc92 100644
--- a/lib/mempool/rte_mempool.h
+++ b/lib/mempool/rte_mempool.h
@@ -86,6 +86,14 @@ struct rte_mempool_cache {
uint32_t size;/**< Size of the cache */
uint32_t flushthresh; /**< Threshold before we flush excess elements */
uint32_t len; /**< Current cache count */
+#ifdef RTE_LIBRTE_MEMPOOL_DEBUG
+ uint32_t unused;
+ /* Fast access statistics, only for likely events */
+ uint64_t put_bulk; /**< Number of puts. */
+ uint64_t put_objs; /**< Number of objects successfully put.
*/
+ uint64_t get_success_bulk; /**< Successful allocation number. */
+ uint64_t get_success_objs; /**< Objects successfully allocated. */
+#endif
/**
* Cache objects
*
@@ -1327,13 +1335,19 @@ rte_mempool_do_generic_put(struct rte_mempool *mp, void
* const *obj_table,
{
void **cache_objs;
+ /* No cache provided */
+ if (unlikely(cache == NULL))
+ goto driver_enqueue;
+
+#ifdef RTE_LIBRTE_MEMPOOL_DEBUG
/* increment stat now, adding in mempool always success */
- RTE_MEMPOOL_STAT_ADD(mp, put_bulk, 1);
- RTE_MEMPOOL_STAT_ADD(mp, put_objs, n);
+ cache->put_bulk += 1;
+ cache->put_objs += n;
+#endif
- /* No cache provided or the request itself is too big for the cache */
- if (unlikely(cache == NULL || n > cache->flushthresh))
- goto driver_enqueue;
+ /* The request is too big for the cache */
+ if (unlikely(n > cache->flushthresh))
+ goto driver_enqueue_stats_incremented;
/*
* The cache follows the following algorithm:
@@ -1358,6 +1372,12 @@ rte_mempool_do_generic_put(struct rte_mempool *mp, void
* const *obj_table,
driver_enqueue:
+ /* increment stat now, adding in mempool always success */
+ RTE_MEMPOOL_STAT_ADD(mp, put_bulk, 1);
+ RTE_MEMPOOL_STAT_ADD(mp, put_objs, n);
+
+driver_enqueue_stats_incremented:
+
/* push objects to the backend */
rte_mempool_ops_enqueue_bulk(mp, obj_table, n);
}
@@ -1464,8 +1484,10 @@ rte_mempool_do_generic_get(struct rte_mempool *mp, void
**obj_table,
if (remaining == 0) {
/* The entire request is satisfied from the cache. */
- RTE_MEMPOOL_STAT_ADD(mp, get_success_bulk, 1);
- RTE_MEMPOOL_STAT_ADD(mp, get_success_objs, n);
+#ifdef RTE_LIBRTE_MEMPOOL_DEBUG
+ cache->get_success_bulk += 1;
+
[PATCH v2] mempool: optimized debug statistics
When built with debug enabled (RTE_LIBRTE_MEMPOOL_DEBUG defined), the
performance of mempools with caches is improved as follows.
Accessing objects in the mempool is likely to increment either the
put_bulk and put_objs or the get_success_bulk and get_success_objs
debug statistics counters.
By adding an alternative set of these counters to the mempool cache
structure, accessing the dedicated debug statistics structure is avoided
in the likely cases where these counters are incremented.
The trick here is that the cache line holding the mempool cache structure
is accessed anyway, in order to update the "len" field. Updating some
debug statistics counters in the same cache line has lower performance
cost than accessing the debug statistics counters in the dedicated debug
statistics structure, i.e. in another cache line.
Running mempool_perf_autotest on a VMware virtual server shows an avg.
increase of 6.4 % in rate_persec for the tests with cache. (Only when
built with debug enabled, obviously!)
For the tests without cache, the avg. increase in rate_persec is 0.8 %. I
assume this is noise from the test environment.
v2:
* Fix spelling and repeated word in commit message, caught by checkpatch.
Signed-off-by: Morten Brørup
---
lib/mempool/rte_mempool.c | 7 +
lib/mempool/rte_mempool.h | 55 +++
2 files changed, 51 insertions(+), 11 deletions(-)
diff --git a/lib/mempool/rte_mempool.c b/lib/mempool/rte_mempool.c
index 21c94a2b9f..7b8c00a022 100644
--- a/lib/mempool/rte_mempool.c
+++ b/lib/mempool/rte_mempool.c
@@ -1285,6 +1285,13 @@ rte_mempool_dump(FILE *f, struct rte_mempool *mp)
sum.get_fail_objs += mp->stats[lcore_id].get_fail_objs;
sum.get_success_blks += mp->stats[lcore_id].get_success_blks;
sum.get_fail_blks += mp->stats[lcore_id].get_fail_blks;
+ /* Add the fast access statistics, if local caches exist */
+ if (mp->cache_size != 0) {
+ sum.put_bulk += mp->local_cache[lcore_id].put_bulk;
+ sum.put_objs += mp->local_cache[lcore_id].put_objs;
+ sum.get_success_bulk +=
mp->local_cache[lcore_id].get_success_bulk;
+ sum.get_success_objs +=
mp->local_cache[lcore_id].get_success_objs;
+ }
}
fprintf(f, " stats:\n");
fprintf(f, "put_bulk=%"PRIu64"\n", sum.put_bulk);
diff --git a/lib/mempool/rte_mempool.h b/lib/mempool/rte_mempool.h
index 3725a72951..d84087bc92 100644
--- a/lib/mempool/rte_mempool.h
+++ b/lib/mempool/rte_mempool.h
@@ -86,6 +86,14 @@ struct rte_mempool_cache {
uint32_t size;/**< Size of the cache */
uint32_t flushthresh; /**< Threshold before we flush excess elements */
uint32_t len; /**< Current cache count */
+#ifdef RTE_LIBRTE_MEMPOOL_DEBUG
+ uint32_t unused;
+ /* Fast access statistics, only for likely events */
+ uint64_t put_bulk; /**< Number of puts. */
+ uint64_t put_objs; /**< Number of objects successfully put.
*/
+ uint64_t get_success_bulk; /**< Successful allocation number. */
+ uint64_t get_success_objs; /**< Objects successfully allocated. */
+#endif
/**
* Cache objects
*
@@ -1327,13 +1335,19 @@ rte_mempool_do_generic_put(struct rte_mempool *mp, void
* const *obj_table,
{
void **cache_objs;
+ /* No cache provided */
+ if (unlikely(cache == NULL))
+ goto driver_enqueue;
+
+#ifdef RTE_LIBRTE_MEMPOOL_DEBUG
/* increment stat now, adding in mempool always success */
- RTE_MEMPOOL_STAT_ADD(mp, put_bulk, 1);
- RTE_MEMPOOL_STAT_ADD(mp, put_objs, n);
+ cache->put_bulk += 1;
+ cache->put_objs += n;
+#endif
- /* No cache provided or the request itself is too big for the cache */
- if (unlikely(cache == NULL || n > cache->flushthresh))
- goto driver_enqueue;
+ /* The request is too big for the cache */
+ if (unlikely(n > cache->flushthresh))
+ goto driver_enqueue_stats_incremented;
/*
* The cache follows the following algorithm:
@@ -1358,6 +1372,12 @@ rte_mempool_do_generic_put(struct rte_mempool *mp, void
* const *obj_table,
driver_enqueue:
+ /* increment stat now, adding in mempool always success */
+ RTE_MEMPOOL_STAT_ADD(mp, put_bulk, 1);
+ RTE_MEMPOOL_STAT_ADD(mp, put_objs, n);
+
+driver_enqueue_stats_incremented:
+
/* push objects to the backend */
rte_mempool_ops_enqueue_bulk(mp, obj_table, n);
}
@@ -1464,8 +1484,10 @@ rte_mempool_do_generic_get(struct rte_mempool *mp, void
**obj_table,
if (remaining == 0) {
/* The entire request is satisfied from the cache. */
- RTE_MEMPOOL_STAT_ADD(mp, get_success_bulk, 1);
- RTE_MEMPOOL_STAT_ADD(mp, get_success_objs, n);
+#ifdef RTE_LI
DPDK Release Status Meeting 2022-10-27
Release status meeting minutes 2022-10-27 = Agenda: * Release Dates * Subtrees * Roadmaps * LTS * Defects * Opens Participants: * ARM * Canonical [No] * Debian/Microsoft * Intel * Marvell * Nvidia * Red Hat * Xilinx/AMD Release Dates - The following are the proposed current dates for 22.11: * V1 deadline: 24 August 2022 * RC1: 7 October 2022 * RC2: 28 October 2022 - moved from 24 October * RC3: 31 October 2022 * Release: 16 November 2022 Subtrees * next-net * NFP several patches merged * Indirect Aging patches under review * New version sent * New drivers in this release: * Microsoft Azure Network Adapter (MANA) PMD * Compilation will be disabled until dependencies become available * Google Virtual Ethernet (GVE) * Merged * idpf (Infrastructure Data Path Function) PMD * Under review * New updates sent * net/qdma PMD * Awaiting reply from submitter * Most likely next release * Memif driver needs reviews from maintainer * next-net-intel * ~20 patches, mainly bug fixes, waiting for merge * next-net-mlx * PR sent for RC2 * next-eventdev * PR merged for RC2 * next-virtio * PR ready for RC2 * next-crypto * 60-70 patches merged and ready for pull * New Huawei UADK driver ready for merge * Will be ready for pull on Friday 28 October * main * RC2 on Friday 28 October * Mempool changes merged * Meson version upgraded to 0.53.2. * KNI deprecation warnings added for compile and runtime * We should also upgrade the minimum required kernel version for 22.11 * Suggestions welcome * Probably the earliest LTS kernel: 4.14 * Looking for RC1 test reports from different vendors Proposed Schedule for 2023 23.03 * Proposal deadline (RFC/v1 patches): 30 December 2022 * API freeze (-rc1): 10 February 2023 * PMD features freeze (-rc2): 24 February 2023 * Built-in applications features freeze (-rc3): 3 March 2023 * Release: 10 March 2023 23.07 * Proposal deadline (RFC/v1 patches): 7 April 2023 * API freeze (-rc1): 26 May 2023 * PMD features freeze (-rc2): 16 June 2023 * Built-in applications features freeze (-rc3): 23 June 2023 * Release: 14 July 2023 23.11 * Proposal deadline (RFC/v1 patches): 11 August 2023 * API freeze (-rc1): 29 September 2023 * PMD features freeze (-rc2): 20 October 2023 * Built-in applications features freeze (-rc3): 27 October 2023 * Release: 17 November 2023 Other - * TBA LTS --- LTSes released on August 29th. Next releases will be: * 21.11.3 * 20.11.7 * 19.11.14 Discussion needed on whether we will continue the 3 year LTS cycle that was trialled for 19.11 to the 20.11 LTS cycle. * Distros * v20.11 in Debian 11 * Ubuntu 22.04 contains 21.11 Defects --- * Bugzilla links, 'Bugs', added for hosted projects * https://www.dpdk.org/hosted-projects/ Opens - * None DPDK Release Status Meetings The DPDK Release Status Meeting is intended for DPDK Committers to discuss the status of the master tree and sub-trees, and for project managers to track progress or milestone dates. The meeting occurs on every Thursday at 9:30 UTC over Jitsi on https://meet.jit.si/DPDK You don't need an invite to join the meeting but if you want a calendar reminder just send an email to "John McNamara john.mcnam...@intel.com" for the invite.
[Bug 1117] af-packet driver fix breaks rte_pktmbuf_prepend()
https://bugs.dpdk.org/show_bug.cgi?id=1117 Bug ID: 1117 Summary: af-packet driver fix breaks rte_pktmbuf_prepend() Product: DPDK Version: 21.11 Hardware: All OS: All Status: UNCONFIRMED Severity: normal Priority: Normal Component: ethdev Assignee: dev@dpdk.org Reporter: bly...@gmail.com Target Milestone: --- The following commit causes rte_pktmbuf_prepend() to fail when prepending the full headroom reserved when creating the mbuf pool. https://git.dpdk.org/dpdk-stable/commit/?h=21.11&id=d41d39bcf76900deb6aa000e929fd65734254201 Of issue is if the received frame was originally tagged (and vlan-stripped by af-socket connection) the above commit does a vlan reinsert action as part of eth_af_packet_rx(). This moves the SOF forward into the "reserved" space by four bytes. For example, if the reserved space was 128 bytes and the next stage of processing is a generic pipeline handling a variety of interface types, the generic prepend of the reserved 128-byte meta-data space will fail. This means you cannot use a common pipeline and have to have an exception now for af-packet rx handling as compared to physical NIC, tap, vhost, etc. Ideally, the initial SOF offset should be consistent relative to your design's reserved headroom, regardless of the vlan tagged status of the frame or the interface type it was received on. -- You are receiving this mail because: You are the assignee for the bug.
Re: release candidate 22.11-rc1
Hello, IBM - Power Systems DPDK 22.11.0-rc1 * Basic PF on Mellanox: No new issues or regressions were seen. * Performance: not tested. * OS: RHEL 8.5 kernel: 4.18.0-348.el8.ppc64le with gcc version 8.5.0 20210514 (Red Hat 8.5.0-10) RHEL 9.0 kernel: 5.14.0-70.13.1.el9_0.ppc64le with gcc version 11.2.1 20220127 (Red Hat 11.2.1-9) Systems tested: - IBM Power9 PowerNV 9006-22P NICs: - Mellanox Technologies MT28800 Family [ConnectX-5 Ex] - firmware version: 16.34.1002 - MLNX_OFED_LINUX-5.7-1.0.2.1 (OFED-5.7-1.0.2) - LPARs on IBM Power10 CHRP IBM,9105-22A NICs: - Mellanox Technologies MT28800 Family [ConnectX-5 Ex] - firmware version: 16.34.1002 - MLNX_OFED_LINUX-5.7-1.0.2.1 (OFED-5.7-1.0.2) Regards, Thinh Tran On 10/10/2022 8:50 PM, Thomas Monjalon wrote: A new DPDK release candidate is ready for testing: https://git.dpdk.org/dpdk/tag/?id=v22.11-rc1 There are 737 new patches in this snapshot, including many API/ABI compatibility breakages. This release won't be ABI-compatible with previous ones. Release notes: https://doc.dpdk.org/guides/rel_notes/release_22_11.html Highlights of 22.11-rc1: - LoongArch build - Intel uncore frequency control - multiple mbuf pools per Rx queue - Rx buffer split based on protocol - hardware congestion management - hairpin memory configuration - Rx/Tx descriptor dump - flow API extensions - MACsec processing offload - ShangMi crypto algorithms - baseband FFT operations - eventdev Tx queue start/stop - eventdev crypto vectorization - NitroSketch membership Some work is in progress to optimize the mempool cache. Some patches are part of -rc1, and more could be merged in -rc2. Please measure the performance of this release candidate, and check these mempool patches: https://patches.dpdk.org/project/dpdk/list/?series=25063 Please test and report issues on bugs.dpdk.org. DPDK 22.11-rc2 is expected in two weeks. Thank you everyone
Re: [EXT] [PATCH v7 0/6] crypto/uadk: introduce uadk crypto driver
On 2022/10/27 下午6:53, Akhil Goyal wrote:
Introduce a new crypto PMD for hardware accelerators based on UADK [1].
UADK is a framework for user applications to access hardware accelerators.
UADK relies on IOMMU SVA (Shared Virtual Address) feature, which share
the same page table between IOMMU and MMU.
Thereby user application can directly use virtual address for device dma,
which enhances the performance as well as easy usability.
[1] https://urldefense.proofpoint.com/v2/url?u=https-
3A__github.com_Linaro_uadk&d=DwIDAg&c=nKjWec2b6R0mOyPaz7xtfQ&r=Dn
L7Si2wl_PRwpZ9TWey3eu68gBzn7DkPwuqhd6WNyo&m=AwapMJzvfUDL49W9f
mtmvIGZMarGUdjRSliBhG4tep6Uh5wN2zYrZRJ4JpbrVolg&s=MxSabEjMMkeRr0L
Z7c1gSmaJmJ1_dIdJIKrst98pKzk&e=
Test:
sudo dpdk-test --vdev=crypto_uadk (--log-level=6)
RTE>>cryptodev_uadk_autotest
RTE>>quit
Update in v7:
05: fix key_size of SHA384 HMAC
Update in v6:
Akhil helped review the v5, and address all the comments from Akhil
01: add rel_notes info, update uadk.rst, put struct in header, etc.
02: move struct to header
04: move struct to header, remove RTE_CRYPTODEV_FF_SYM_SESSIONLESS
05: fixed key_size in HMAC mode
06: updated app/test/meson.build and uadk.rst in the mean time.
Update in v5
Patch 1 fix the build issue when uadk is installed to specific folder
And update doc accordingly
Update in v4:
Akril suggest dpdk use pkg-config, So
Enable uadk support x86 local build, and support pkg-config.
Use pkg-config feature for the uadk crypto pmd.
Add build uadk library steps in doc
Test on both x86 and arm.
x86 can build and install, but can not test since no device.
Resend v3:
Rebase on next/for-main, which just merged the series
"cryptodev: rework session framework".
Update in v3:
Split patches according to Akhil's suggestions
Please split the patches as below.
1. introduce driver - create files with meson.build and with probe/remove
and device ops defined but not implemented.
You do not need to write empty functions.
Add basic documentation also which defines what the driver is.
You can explain the build dependency here.
2. define queue structs and setup/remove APIs
3. Add data path
4. implement cipher op. Add capabilities and documentation of what is
supported in each of the patches. Add feature flags etc.
5. implement auth, add capabilities and documentation
6. test app changes.
Update in v2:
Change uadk_supported_platform to uadk_crypto_version, which matches
better
than platform.
enum uadk_crypto_version {
UADK_CRYPTO_V2,
UADK_CRYPTO_V3,
};
Update in v1, compared with rfc
Suggested from Akhil Goyal
Only consider crypto PMD first
Split patch into small (individually compiled) patches.
Update MAINTAINERS and doc/guides/cryptodevs/features/uadk.ini
Zhangfei Gao (6):
crypto/uadk: introduce uadk crypto driver
crypto/uadk: support basic operations
crypto/uadk: support enqueue/dequeue operations
crypto/uadk: support cipher algorithms
crypto/uadk: support auth algorithms
test/crypto: support uadk PMD
MAINTAINERS |6 +
app/test/meson.build |1 +
app/test/test_cryptodev.c |7 +
app/test/test_cryptodev.h |1 +
doc/guides/cryptodevs/features/uadk.ini | 55 +
doc/guides/cryptodevs/index.rst |1 +
doc/guides/cryptodevs/uadk.rst| 96 ++
doc/guides/rel_notes/release_22_11.rst|6 +
drivers/crypto/meson.build|1 +
drivers/crypto/uadk/meson.build | 30 +
drivers/crypto/uadk/uadk_crypto_pmd.c | 1081 +
drivers/crypto/uadk/uadk_crypto_pmd_private.h | 79 ++
drivers/crypto/uadk/version.map |3 +
13 files changed, 1367 insertions(+)
create mode 100644 doc/guides/cryptodevs/features/uadk.ini
create mode 100644 doc/guides/cryptodevs/uadk.rst
create mode 100644 drivers/crypto/uadk/meson.build
create mode 100644 drivers/crypto/uadk/uadk_crypto_pmd.c
create mode 100644 drivers/crypto/uadk/uadk_crypto_pmd_private.h
create mode 100644 drivers/crypto/uadk/version.map
Series Applied to dpdk-next-crypto
Welcome to dpdk.
That's Great
Thanks for your guidance and patience.
Originally, the UADK only consider ARM usage and need cross-compile on x86.
With you guidance, now the UADK support pkg-config, and can build and
install on both ARM and X86.
Thanks Akhil.
RE: [PATCH v11 15/18] net/idpf: add support for Rx offloading
> -Original Message-
> From: Andrew Rybchenko
> Sent: Tuesday, October 25, 2022 6:04 PM
> To: Guo, Junfeng ; Zhang, Qi Z
> ; Wu, Jingjing ; Xing, Beilei
>
> Cc: dev@dpdk.org; Li, Xiaoyun
> Subject: Re: [PATCH v11 15/18] net/idpf: add support for Rx offloading
>
> On 10/24/22 16:12, Junfeng Guo wrote:
> > Add Rx offloading support:
> > - support CHKSUM and RSS offload for split queue model
> > - support CHKSUM offload for single queue model
> >
> > Signed-off-by: Beilei Xing
> > Signed-off-by: Xiaoyun Li
> > Signed-off-by: Junfeng Guo
> > ---
> > doc/guides/nics/features/idpf.ini | 2 +
> > drivers/net/idpf/idpf_ethdev.c| 9 ++-
> > drivers/net/idpf/idpf_rxtx.c | 122 ++
> > 3 files changed, 132 insertions(+), 1 deletion(-)
> >
> > diff --git a/doc/guides/nics/features/idpf.ini
> > b/doc/guides/nics/features/idpf.ini
> > index d4eb9b374c..c86d9378ea 100644
> > --- a/doc/guides/nics/features/idpf.ini
> > +++ b/doc/guides/nics/features/idpf.ini
> > @@ -9,5 +9,7 @@
> > [Features]
> > Queue start/stop = Y
> > MTU update = Y
> > +L3 checksum offload = P
> > +L4 checksum offload = P
>
> RSS hash missing
>
> > Packet type parsing = Y
> > Linux= Y
> > diff --git a/drivers/net/idpf/idpf_ethdev.c
> > b/drivers/net/idpf/idpf_ethdev.c index 739cf31d65..d8cc423a23 100644
> > --- a/drivers/net/idpf/idpf_ethdev.c
> > +++ b/drivers/net/idpf/idpf_ethdev.c
> > @@ -94,7 +94,14 @@ idpf_dev_info_get(struct rte_eth_dev *dev, struct
> rte_eth_dev_info *dev_info)
> > dev_info->max_mac_addrs = IDPF_NUM_MACADDR_MAX;
> > dev_info->dev_capa =
> RTE_ETH_DEV_CAPA_RUNTIME_RX_QUEUE_SETUP |
> > RTE_ETH_DEV_CAPA_RUNTIME_TX_QUEUE_SETUP;
> > - dev_info->rx_offload_capa = 0;
> > +
> > + dev_info->rx_offload_capa =
> > + RTE_ETH_RX_OFFLOAD_IPV4_CKSUM |
> > + RTE_ETH_RX_OFFLOAD_UDP_CKSUM|
> > + RTE_ETH_RX_OFFLOAD_TCP_CKSUM|
> > + RTE_ETH_RX_OFFLOAD_OUTER_IPV4_CKSUM |
> > + RTE_ETH_RX_OFFLOAD_RSS_HASH;
>
> As I understand you know mode here and you should not report offload
> which is not supported in current mode (RSS vs single queue).
Yes, will remove RTE_ETH_RX_OFFLOAD_RSS_HASH here since RSS hash is not
supported in single queue mode currently,
So we needn't to add 'RSS hash = Y' in idpf.ini, right?
>
> > +
> > dev_info->tx_offload_capa = RTE_ETH_TX_OFFLOAD_MULTI_SEGS;
> >
> > dev_info->default_rxconf = (struct rte_eth_rxconf) {
>
> [snip]
RE: [PATCH v2] app/testpmd: fix protocol headers display for Rx buffer split
Hi All,
Could you please review and provide suggestions if any.
Thanks,
Yuan
> -Original Message-
> From: Wang, YuanX
> Sent: Tuesday, October 18, 2022 10:50 PM
> To: andrew.rybche...@oktetlabs.ru; Singh, Aman Deep
> ; Zhang, Yuying
> Cc: dev@dpdk.org; Ding, Xuan ; Tang, Yaqi
> ; Wang, YuanX
> Subject: [PATCH v2] app/testpmd: fix protocol headers display for Rx buffer
> split
>
> The "show config rxhdrs" cmd displays the configured protocol headers that
> are used for protocol-based buffer split.
> However, it shows "inner-ipv6" as "inner-ipv4".
>
> This patch fixes that by adjusting the order of condition judgments.
>
> Fixes: 52e2e7edcf48 ("app/testpmd: add protocol-based buffer split")
>
> Signed-off-by: Yuan Wang
> Tested-by: Yaqi Tang
>
> ---
> v2: add fixline.
>
> ---
> app/test-pmd/config.c | 24
> 1 file changed, 12 insertions(+), 12 deletions(-)
>
> diff --git a/app/test-pmd/config.c b/app/test-pmd/config.c index
> 0f7dbd698f..82fbbc9944 100644
> --- a/app/test-pmd/config.c
> +++ b/app/test-pmd/config.c
> @@ -4937,15 +4937,6 @@ static const char *get_ptype_str(uint32_t ptype)
> else if ((ptype & RTE_PTYPE_L2_ETHER) == RTE_PTYPE_L2_ETHER)
> return "eth";
>
> - else if ((ptype & (RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN |
> RTE_PTYPE_INNER_L4_TCP)) ==
> - (RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN |
> RTE_PTYPE_INNER_L4_TCP))
> - return "inner-ipv4-tcp";
> - else if ((ptype & (RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN |
> RTE_PTYPE_INNER_L4_UDP)) ==
> - (RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN |
> RTE_PTYPE_INNER_L4_UDP))
> - return "inner-ipv4-udp";
> - else if ((ptype & (RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN |
> RTE_PTYPE_INNER_L4_SCTP)) ==
> - (RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN |
> RTE_PTYPE_INNER_L4_SCTP))
> - return "inner-ipv4-sctp";
> else if ((ptype & (RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN |
> RTE_PTYPE_INNER_L4_TCP)) ==
> (RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN |
> RTE_PTYPE_INNER_L4_TCP))
> return "inner-ipv6-tcp";
> @@ -4955,18 +4946,27 @@ static const char *get_ptype_str(uint32_t ptype)
> else if ((ptype & (RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN |
> RTE_PTYPE_INNER_L4_SCTP)) ==
> (RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN |
> RTE_PTYPE_INNER_L4_SCTP))
> return "inner-ipv6-sctp";
> + else if ((ptype & (RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN |
> RTE_PTYPE_INNER_L4_TCP)) ==
> + (RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN |
> RTE_PTYPE_INNER_L4_TCP))
> + return "inner-ipv4-tcp";
> + else if ((ptype & (RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN |
> RTE_PTYPE_INNER_L4_UDP)) ==
> + (RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN |
> RTE_PTYPE_INNER_L4_UDP))
> + return "inner-ipv4-udp";
> + else if ((ptype & (RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN |
> RTE_PTYPE_INNER_L4_SCTP)) ==
> + (RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN |
> RTE_PTYPE_INNER_L4_SCTP))
> + return "inner-ipv4-sctp";
> else if ((ptype & RTE_PTYPE_INNER_L4_TCP) ==
> RTE_PTYPE_INNER_L4_TCP)
> return "inner-tcp";
> else if ((ptype & RTE_PTYPE_INNER_L4_UDP) ==
> RTE_PTYPE_INNER_L4_UDP)
> return "inner-udp";
> else if ((ptype & RTE_PTYPE_INNER_L4_SCTP) ==
> RTE_PTYPE_INNER_L4_SCTP)
> return "inner-sctp";
> + else if ((ptype & RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN) ==
> + RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN)
> + return "inner-ipv6";
> else if ((ptype & RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN) ==
> RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN)
> return "inner-ipv4";
> - else if ((ptype & RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN) ==
> - RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN)
> - return "inner-ipv6";
> else if ((ptype & RTE_PTYPE_INNER_L2_ETHER) ==
> RTE_PTYPE_INNER_L2_ETHER)
> return "inner-eth";
> else if ((ptype & RTE_PTYPE_TUNNEL_GRENAT) ==
> RTE_PTYPE_TUNNEL_GRENAT)
> --
> 2.25.1
Re: [PATCH v4 1/2] app/testpmd: fix vlan offload of rxq
在 2022/10/27 19:02, Ye, MingjinX 写道:
Hi lihuisong,
This means that queue offloads need to update by recalling dev_configure
and setup target queues.
Why not update queue offloads in PMD?
Can you tell me, where is the limitation?
According to other Rx/Tx offload configurations, this may not be a
limitation.
But it seems to create a dependency on user usage.
Port VLAN releated offloads are set by ethdev ops. There is no requirement
in ehedev layer that this port needs to stopped when set these offloads.
Now it depends on user does recall dev_configure and setup queues to update
queue offloads because of setting these offloads.
Thanks,
Mingjin
-Original Message-
From: lihuisong (C)
Sent: 2022年10月26日 17:53
To: Ye, MingjinX ; dev@dpdk.org
Cc: sta...@dpdk.org; Zhou, YidingX ; Singh, Aman
Deep ; Zhang, Yuying
Subject: Re: [PATCH v4 1/2] app/testpmd: fix vlan offload of rxq
在 2022/10/27 1:10, Mingjin Ye 写道:
After setting vlan offload in testpmd, the result is not updated to
rxq. Therefore, the queue needs to be reconfigured after executing the
"vlan offload" related commands.
Fixes: a47aa8b97afe ("app/testpmd: add vlan offload support")
Cc: sta...@dpdk.org
Signed-off-by: Mingjin Ye
---
app/test-pmd/cmdline.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/app/test-pmd/cmdline.c b/app/test-pmd/cmdline.c index
17be2de402..ce125f549f 100644
--- a/app/test-pmd/cmdline.c
+++ b/app/test-pmd/cmdline.c
@@ -4133,6 +4133,7 @@ cmd_vlan_offload_parsed(void *parsed_result,
else
vlan_extend_set(port_id, on);
+ cmd_reconfig_device_queue(port_id, 1, 1);
In addition, I have some comments:
1) Normally, the parsed function of testpmd command needed to re-config
port and
queue needs to check if port status is STOPED. Why don't you add this check?
If the check is not exist, queue offloads are not updated until the next
port
stop/start command is executed. Right?
2) Why is the queue-based VLAN offload API not used?
Like, rte_eth_dev_set_vlan_strip_on_queue. It seems that this kind
of API is
dedicated to do this.
This means that queue offloads need to upadte by re-calling dev_configure
and setup all queues, right?
If it is, this adds a usage limitation.
return;
}
RE: [PATCH] app/testpmd: fix testpmd receive jumbo frame packets
Hi Singh,
> -Original Message-
> From: Singh, Aman Deep
> Sent: Thursday, October 27, 2022 11:38 PM
> To: Wang, Jie1X ; dev@dpdk.org
> Cc: Yang, SteveX ; Zhang, Qi Z ;
> Yang, Qiming ; Zhang, Yuying
> ; sta...@dpdk.org
> Subject: Re: [PATCH] app/testpmd: fix testpmd receive jumbo frame packets
>
> Hi Jie,
>
> Thanks for the patch.
>
>
> On 10/25/2022 7:35 AM, Jie Wang wrote:
> > For NIC I40E_10G-10G_BASE_T_X722, when testpmd is configured with link
> > speed, it cannot receive jumbo frame packets.
>
> Why only the jumbo frame are effected and not other pkts.
>
I don't know why only the jumbo frame are effected, when set the link speed, it
will update the link status of the NIC ports.
And this phenomenon can only be reproduced on this NIC
I40E_10G-10G_BASE_T_X722, and other NICs do not have this phenomenon.
> >
> > Because it has changed the link status of the ports if it was
> > configured with link speed.
>
> If we configure to same link speed, will it still have issue ?
>
Yes, it is.
> >
> > When exiting testpmd that it automatically stops packet forwarding and
> > stops all the ports. But it doesn't update the link status of the
> > ports. If stop the ports first that it will update the link status.
> >
> > This patch fix the error that testpmd will update the link status of
> > the ports when it exits.
> >
> > Fixes: d3a274ce9dee ("app/testpmd: handle SIGINT and SIGTERM")
> > Fixes: 284c908cc588 ("app/testpmd: request device removal interrupt")
> > Cc: sta...@dpdk.org
> >
> > Signed-off-by: Jie Wang
> > ---
> > app/test-pmd/testpmd.c | 4 ++--
> > 1 file changed, 2 insertions(+), 2 deletions(-)
> >
> > diff --git a/app/test-pmd/testpmd.c b/app/test-pmd/testpmd.c index
> > 97adafacd0..c348a3f328 100644
> > --- a/app/test-pmd/testpmd.c
> > +++ b/app/test-pmd/testpmd.c
> > @@ -3548,7 +3548,7 @@ pmd_test_exit(void)
> > }
> > #endif
> > if (ports != NULL) {
> > - no_link_check = 1;
> > + no_link_check = 0;
> > RTE_ETH_FOREACH_DEV(pt_id) {
> > printf("\nStopping port %d...\n", pt_id);
> > fflush(stdout);
> > @@ -3675,7 +3675,7 @@ rmv_port_callback(void *arg)
> > need_to_start = 1;
> > stop_packet_forwarding();
> > }
> > - no_link_check = 1;
> > + no_link_check = 0;
>
> Well, here we are undoing a previous change done for 284c908cc588 Won't that
> issue come back.
>
> > stop_port(port_id);
> > no_link_check = org_no_link_check;
> >
[PATCH] net/bonding: fix slave device Rx/Tx offload configuration
Normally, the Rx/Tx offload capability of bonding interface is
the intersection of the capability of all slave devices. And
Rx/Tx offloads configuration of slave device comes from bonding
interface. But now there is a risk that slave device retains its
previous offload configurations which is not within the offload
configurations of bond interface.
Fixes: 57b156540f51 ("net/bonding: fix offloading configuration")
Cc: sta...@dpdk.org
Signed-off-by: Huisong Li
---
drivers/net/bonding/rte_eth_bond_pmd.c | 17 -
1 file changed, 4 insertions(+), 13 deletions(-)
diff --git a/drivers/net/bonding/rte_eth_bond_pmd.c
b/drivers/net/bonding/rte_eth_bond_pmd.c
index dc74852137..ca87490065 100644
--- a/drivers/net/bonding/rte_eth_bond_pmd.c
+++ b/drivers/net/bonding/rte_eth_bond_pmd.c
@@ -1741,20 +1741,11 @@ slave_configure(struct rte_eth_dev *bonded_eth_dev,
slave_eth_dev->data->dev_conf.link_speeds =
bonded_eth_dev->data->dev_conf.link_speeds;
- slave_eth_dev->data->dev_conf.txmode.offloads |=
- bonded_eth_dev->data->dev_conf.txmode.offloads;
-
- slave_eth_dev->data->dev_conf.txmode.offloads &=
- (bonded_eth_dev->data->dev_conf.txmode.offloads |
- ~internals->tx_offload_capa);
-
- slave_eth_dev->data->dev_conf.rxmode.offloads |=
- bonded_eth_dev->data->dev_conf.rxmode.offloads;
-
- slave_eth_dev->data->dev_conf.rxmode.offloads &=
- (bonded_eth_dev->data->dev_conf.rxmode.offloads |
- ~internals->rx_offload_capa);
+ slave_eth_dev->data->dev_conf.txmode.offloads =
+ bonded_eth_dev->data->dev_conf.txmode.offloads;
+ slave_eth_dev->data->dev_conf.rxmode.offloads =
+ bonded_eth_dev->data->dev_conf.rxmode.offloads;
nb_rx_queues = bonded_eth_dev->data->nb_rx_queues;
nb_tx_queues = bonded_eth_dev->data->nb_tx_queues;
--
2.33.0
[PATCH v2] net/ice: fix rx scalar path offload parse
The status_0 field of the rx descriptor is incorrectly parsed as the error
field, resulting in a parsing error of offload features.
This patch fixes parsing of wrong fields.
Fixes: daa02b5cddbb ("mbuf: add namespace to offload flags")
Fixes: dbf3c0e77a22 ("net/ice: handle Rx flex descriptor")
Cc: sta...@dpdk.org
Signed-off-by: Mingjin Ye
---
drivers/net/ice/ice_rxtx.c | 20 ++--
1 file changed, 10 insertions(+), 10 deletions(-)
diff --git a/drivers/net/ice/ice_rxtx.c b/drivers/net/ice/ice_rxtx.c
index 71e5c6f5d6..3c558b32bd 100644
--- a/drivers/net/ice/ice_rxtx.c
+++ b/drivers/net/ice/ice_rxtx.c
@@ -1485,36 +1485,36 @@ ice_rx_queue_count(void *rx_queue)
(1 << ICE_RX_FLEX_DESC_STATUS0_XSUM_EIPE_S) | \
(1 << ICE_RX_FLEX_DESC_STATUS0_XSUM_EUDPE_S) | \
(1 << ICE_RX_FLEX_DESC_STATUS0_RXE_S))
-
+#define ICE_RXD_QW1_ERROR_SHIFT19
/* Rx L3/L4 checksum */
static inline uint64_t
ice_rxd_error_to_pkt_flags(uint16_t stat_err0)
{
uint64_t flags = 0;
- /* check if HW has decoded the packet and checksum */
- if (unlikely(!(stat_err0 & (1 << ICE_RX_FLEX_DESC_STATUS0_L3L4P_S
- return 0;
+ uint64_t error_bits = (stat_err0 >> ICE_RXD_QW1_ERROR_SHIFT) & 0x7D;
- if (likely(!(stat_err0 & ICE_RX_FLEX_ERR0_BITS))) {
- flags |= (RTE_MBUF_F_RX_IP_CKSUM_GOOD |
RTE_MBUF_F_RX_L4_CKSUM_GOOD);
+ if (likely(!(error_bits & ICE_RX_FLEX_ERR0_BITS))) {
+ flags |= (RTE_MBUF_F_RX_IP_CKSUM_GOOD |
+ RTE_MBUF_F_RX_L4_CKSUM_GOOD |
+ RTE_MBUF_F_RX_OUTER_L4_CKSUM_GOOD);
return flags;
}
- if (unlikely(stat_err0 & (1 << ICE_RX_FLEX_DESC_STATUS0_XSUM_IPE_S)))
+ if (unlikely(error_bits & (1 << ICE_RX_FLEX_DESC_STATUS0_XSUM_IPE_S)))
flags |= RTE_MBUF_F_RX_IP_CKSUM_BAD;
else
flags |= RTE_MBUF_F_RX_IP_CKSUM_GOOD;
- if (unlikely(stat_err0 & (1 << ICE_RX_FLEX_DESC_STATUS0_XSUM_L4E_S)))
+ if (unlikely(error_bits & (1 << ICE_RX_FLEX_DESC_STATUS0_XSUM_L4E_S)))
flags |= RTE_MBUF_F_RX_L4_CKSUM_BAD;
else
flags |= RTE_MBUF_F_RX_L4_CKSUM_GOOD;
- if (unlikely(stat_err0 & (1 << ICE_RX_FLEX_DESC_STATUS0_XSUM_EIPE_S)))
+ if (unlikely(error_bits & (1 << ICE_RX_FLEX_DESC_STATUS0_XSUM_EIPE_S)))
flags |= RTE_MBUF_F_RX_OUTER_IP_CKSUM_BAD;
- if (unlikely(stat_err0 & (1 << ICE_RX_FLEX_DESC_STATUS0_XSUM_EUDPE_S)))
+ if (unlikely(error_bits & (1 << ICE_RX_FLEX_DESC_STATUS0_XSUM_EUDPE_S)))
flags |= RTE_MBUF_F_RX_OUTER_L4_CKSUM_BAD;
else
flags |= RTE_MBUF_F_RX_OUTER_L4_CKSUM_GOOD;
--
2.34.1
[PATCH v2] net/ice: fix rx scalar path offload parse
The status_0 field of the rx descriptor is incorrectly parsed as the error
field, which causes the parsing error of offload features.
This patch is to fixe the parsing of wrong fields.
Fixes: daa02b5cddbb ("mbuf: add namespace to offload flags")
Fixes: dbf3c0e77a22 ("net/ice: handle Rx flex descriptor")
Cc: sta...@dpdk.org
Signed-off-by: Mingjin Ye
---
drivers/net/ice/ice_rxtx.c | 20 ++--
1 file changed, 10 insertions(+), 10 deletions(-)
diff --git a/drivers/net/ice/ice_rxtx.c b/drivers/net/ice/ice_rxtx.c
index 71e5c6f5d6..3c558b32bd 100644
--- a/drivers/net/ice/ice_rxtx.c
+++ b/drivers/net/ice/ice_rxtx.c
@@ -1485,36 +1485,36 @@ ice_rx_queue_count(void *rx_queue)
(1 << ICE_RX_FLEX_DESC_STATUS0_XSUM_EIPE_S) | \
(1 << ICE_RX_FLEX_DESC_STATUS0_XSUM_EUDPE_S) | \
(1 << ICE_RX_FLEX_DESC_STATUS0_RXE_S))
-
+#define ICE_RXD_QW1_ERROR_SHIFT19
/* Rx L3/L4 checksum */
static inline uint64_t
ice_rxd_error_to_pkt_flags(uint16_t stat_err0)
{
uint64_t flags = 0;
- /* check if HW has decoded the packet and checksum */
- if (unlikely(!(stat_err0 & (1 << ICE_RX_FLEX_DESC_STATUS0_L3L4P_S
- return 0;
+ uint64_t error_bits = (stat_err0 >> ICE_RXD_QW1_ERROR_SHIFT) & 0x7D;
- if (likely(!(stat_err0 & ICE_RX_FLEX_ERR0_BITS))) {
- flags |= (RTE_MBUF_F_RX_IP_CKSUM_GOOD |
RTE_MBUF_F_RX_L4_CKSUM_GOOD);
+ if (likely(!(error_bits & ICE_RX_FLEX_ERR0_BITS))) {
+ flags |= (RTE_MBUF_F_RX_IP_CKSUM_GOOD |
+ RTE_MBUF_F_RX_L4_CKSUM_GOOD |
+ RTE_MBUF_F_RX_OUTER_L4_CKSUM_GOOD);
return flags;
}
- if (unlikely(stat_err0 & (1 << ICE_RX_FLEX_DESC_STATUS0_XSUM_IPE_S)))
+ if (unlikely(error_bits & (1 << ICE_RX_FLEX_DESC_STATUS0_XSUM_IPE_S)))
flags |= RTE_MBUF_F_RX_IP_CKSUM_BAD;
else
flags |= RTE_MBUF_F_RX_IP_CKSUM_GOOD;
- if (unlikely(stat_err0 & (1 << ICE_RX_FLEX_DESC_STATUS0_XSUM_L4E_S)))
+ if (unlikely(error_bits & (1 << ICE_RX_FLEX_DESC_STATUS0_XSUM_L4E_S)))
flags |= RTE_MBUF_F_RX_L4_CKSUM_BAD;
else
flags |= RTE_MBUF_F_RX_L4_CKSUM_GOOD;
- if (unlikely(stat_err0 & (1 << ICE_RX_FLEX_DESC_STATUS0_XSUM_EIPE_S)))
+ if (unlikely(error_bits & (1 << ICE_RX_FLEX_DESC_STATUS0_XSUM_EIPE_S)))
flags |= RTE_MBUF_F_RX_OUTER_IP_CKSUM_BAD;
- if (unlikely(stat_err0 & (1 << ICE_RX_FLEX_DESC_STATUS0_XSUM_EUDPE_S)))
+ if (unlikely(error_bits & (1 << ICE_RX_FLEX_DESC_STATUS0_XSUM_EUDPE_S)))
flags |= RTE_MBUF_F_RX_OUTER_L4_CKSUM_BAD;
else
flags |= RTE_MBUF_F_RX_OUTER_L4_CKSUM_GOOD;
--
2.34.1
[PATCH] app/testpmd: update bond port configurations when add slave
Some capabilities (like, rx_offload_capa and tx_offload_capa) of
bonding device in dev_info is zero when no slave is added. And its
capability will be updated when add a new slave device.
The capability to update dynamically can introduce some probelms if
not handled properly. For example, the reconfig() is called to initialize
bonding port configurations when create a bonding device. The global
tx_mode is assigned to dev_conf.txmode. The DEV_TX_OFFLOAD_MBUF_FAST_FREE
which is the default value of global tx_mode.offloads in testpmd is removed
from bonding device configuration because of zero rx_offload_capa.
As a result, this offload isn't set to bonding device.
Generally, port configurations of bonding device must be within the
intersection of the capability of all slave devices. If use original port
configurations, the removed capabilities because of adding a new slave may
cause failure when re-initialize bonding device.
So port configurations of bonding device also need to be updated because of
the added and removed capabilities. In addition, this also helps to ensure
consistency between testpmd and bonding device.
Signed-off-by: Huisong Li
---
app/test-pmd/testpmd.c| 40 +++
app/test-pmd/testpmd.h| 3 +-
drivers/net/bonding/bonding_testpmd.c | 2 ++
3 files changed, 44 insertions(+), 1 deletion(-)
diff --git a/app/test-pmd/testpmd.c b/app/test-pmd/testpmd.c
index 97adafacd0..7324b8865c 100644
--- a/app/test-pmd/testpmd.c
+++ b/app/test-pmd/testpmd.c
@@ -2805,6 +2805,41 @@ fill_xstats_display_info(void)
fill_xstats_display_info_for_port(pi);
}
+/*
+ * Some capabilities (like, rx_offload_capa and tx_offload_capa) of bonding
+ * device in dev_info is zero when no slave is added. And its capability of
+ * will be updated when add a new slave device. So adding a device slave need
+ * to update the port configurations of bonding device.
+ */
+static void
+update_bonding_port_dev_conf(portid_t bond_pid)
+{
+#ifdef RTE_NET_BOND
+ struct rte_port *port = &ports[bond_pid];
+ uint16_t i;
+ int ret;
+
+ ret = eth_dev_info_get_print_err(bond_pid, &port->dev_info);
+ if (ret != 0) {
+ fprintf(stderr, "Failed to get dev info for port = %u\n",
+ bond_pid);
+ return;
+ }
+
+ if (port->dev_info.tx_offload_capa & RTE_ETH_TX_OFFLOAD_MBUF_FAST_FREE)
+ port->dev_conf.txmode.offloads |=
+ RTE_ETH_TX_OFFLOAD_MBUF_FAST_FREE;
+ /* Apply Tx offloads configuration */
+ for (i = 0; i < port->dev_info.max_tx_queues; i++)
+ port->txq[i].conf.offloads = port->dev_conf.txmode.offloads;
+
+ port->dev_conf.rx_adv_conf.rss_conf.rss_hf &=
+ port->dev_info.flow_type_rss_offloads;
+#else
+ RTE_SET_USED(bond_pid);
+#endif
+}
+
int
start_port(portid_t pid)
{
@@ -2869,6 +2904,11 @@ start_port(portid_t pid)
return -1;
}
+ if (port->bond_flag == 1 && port->update_conf == 1) {
+ update_bonding_port_dev_conf(pi);
+ port->update_conf = 0;
+ }
+
/* configure port */
diag = eth_dev_configure_mp(pi, nb_rxq + nb_hairpinq,
nb_txq + nb_hairpinq,
diff --git a/app/test-pmd/testpmd.h b/app/test-pmd/testpmd.h
index 7fef96f9b1..82714119e8 100644
--- a/app/test-pmd/testpmd.h
+++ b/app/test-pmd/testpmd.h
@@ -316,7 +316,8 @@ struct rte_port {
queueid_t queue_nb; /**< nb. of queues for flow rules */
uint32_tqueue_sz; /**< size of a queue for flow rules */
uint8_t slave_flag : 1, /**< bonding slave port */
- bond_flag : 1; /**< port is bond device */
+ bond_flag : 1, /**< port is bond device */
+ update_conf : 1; /**< need to update bonding
device configuration */
struct port_template*pattern_templ_list; /**< Pattern templates. */
struct port_template*actions_templ_list; /**< Actions templates. */
struct port_table *table_list; /**< Flow tables. */
diff --git a/drivers/net/bonding/bonding_testpmd.c
b/drivers/net/bonding/bonding_testpmd.c
index 3941f4cf23..9529e16fb6 100644
--- a/drivers/net/bonding/bonding_testpmd.c
+++ b/drivers/net/bonding/bonding_testpmd.c
@@ -625,6 +625,7 @@ static void cmd_add_bonding_slave_parsed(void
*parsed_result,
slave_port_id, master_port_id);
return;
}
+ ports[master_port_id].update_conf = 1;
init_port_config();
set_port_slave_flag(slave_port_id);
}
@@ -762,6 +763,7 @@ static void cmd_create_bonded_device_parsed(void
*parsed
[Bug 1101] [dpdk-22.11.0rc1][meson test] driver-tests/eventdev_selftest_sw test failed
https://bugs.dpdk.org/show_bug.cgi?id=1101 jiang,yu (yux.ji...@intel.com) changed: What|Removed |Added CC||yux.ji...@intel.com Status|UNCONFIRMED |RESOLVED Resolution|--- |FIXED --- Comment #7 from jiang,yu (yux.ji...@intel.com) --- Fixed in http://git.dpdk.org/dpdk/commit/?id=ab059e82e12f2a9dc7561960004819de68b37110 -- You are receiving this mail because: You are the assignee for the bug.
[PATCH V2] app/testpmd: update bond port configurations when add slave
Some capabilities (like, rx_offload_capa and tx_offload_capa) of bonding
device in dev_info is zero when no slave is added. And its capability will
be updated when add a new slave device.
The capability to update dynamically may introduce some problems if not
handled properly. For example, the reconfig() is called to initialize
bonding port configurations when create a bonding device. The global
tx_mode is assigned to dev_conf.txmode. The DEV_TX_OFFLOAD_MBUF_FAST_FREE
which is the default value of global tx_mode.offloads in testpmd is removed
from bonding device configuration because of zero rx_offload_capa.
As a result, this offload isn't set to bonding device.
Generally, port configurations of bonding device must be within the
intersection of the capability of all slave devices. If use original port
configurations, the removed capabilities because of adding a new slave may
cause failure when re-initialize bonding device.
So port configurations of bonding device also need to be updated because of
the added and removed capabilities. In addition, this also helps to ensure
consistency between testpmd and bonding device.
Signed-off-by: Huisong Li
---
- v2: fix a spelling error in commit log
---
app/test-pmd/testpmd.c| 40 +++
app/test-pmd/testpmd.h| 3 +-
drivers/net/bonding/bonding_testpmd.c | 2 ++
3 files changed, 44 insertions(+), 1 deletion(-)
diff --git a/app/test-pmd/testpmd.c b/app/test-pmd/testpmd.c
index 97adafacd0..7324b8865c 100644
--- a/app/test-pmd/testpmd.c
+++ b/app/test-pmd/testpmd.c
@@ -2805,6 +2805,41 @@ fill_xstats_display_info(void)
fill_xstats_display_info_for_port(pi);
}
+/*
+ * Some capabilities (like, rx_offload_capa and tx_offload_capa) of bonding
+ * device in dev_info is zero when no slave is added. And its capability of
+ * will be updated when add a new slave device. So adding a device slave need
+ * to update the port configurations of bonding device.
+ */
+static void
+update_bonding_port_dev_conf(portid_t bond_pid)
+{
+#ifdef RTE_NET_BOND
+ struct rte_port *port = &ports[bond_pid];
+ uint16_t i;
+ int ret;
+
+ ret = eth_dev_info_get_print_err(bond_pid, &port->dev_info);
+ if (ret != 0) {
+ fprintf(stderr, "Failed to get dev info for port = %u\n",
+ bond_pid);
+ return;
+ }
+
+ if (port->dev_info.tx_offload_capa & RTE_ETH_TX_OFFLOAD_MBUF_FAST_FREE)
+ port->dev_conf.txmode.offloads |=
+ RTE_ETH_TX_OFFLOAD_MBUF_FAST_FREE;
+ /* Apply Tx offloads configuration */
+ for (i = 0; i < port->dev_info.max_tx_queues; i++)
+ port->txq[i].conf.offloads = port->dev_conf.txmode.offloads;
+
+ port->dev_conf.rx_adv_conf.rss_conf.rss_hf &=
+ port->dev_info.flow_type_rss_offloads;
+#else
+ RTE_SET_USED(bond_pid);
+#endif
+}
+
int
start_port(portid_t pid)
{
@@ -2869,6 +2904,11 @@ start_port(portid_t pid)
return -1;
}
+ if (port->bond_flag == 1 && port->update_conf == 1) {
+ update_bonding_port_dev_conf(pi);
+ port->update_conf = 0;
+ }
+
/* configure port */
diag = eth_dev_configure_mp(pi, nb_rxq + nb_hairpinq,
nb_txq + nb_hairpinq,
diff --git a/app/test-pmd/testpmd.h b/app/test-pmd/testpmd.h
index 7fef96f9b1..82714119e8 100644
--- a/app/test-pmd/testpmd.h
+++ b/app/test-pmd/testpmd.h
@@ -316,7 +316,8 @@ struct rte_port {
queueid_t queue_nb; /**< nb. of queues for flow rules */
uint32_tqueue_sz; /**< size of a queue for flow rules */
uint8_t slave_flag : 1, /**< bonding slave port */
- bond_flag : 1; /**< port is bond device */
+ bond_flag : 1, /**< port is bond device */
+ update_conf : 1; /**< need to update bonding
device configuration */
struct port_template*pattern_templ_list; /**< Pattern templates. */
struct port_template*actions_templ_list; /**< Actions templates. */
struct port_table *table_list; /**< Flow tables. */
diff --git a/drivers/net/bonding/bonding_testpmd.c
b/drivers/net/bonding/bonding_testpmd.c
index 3941f4cf23..9529e16fb6 100644
--- a/drivers/net/bonding/bonding_testpmd.c
+++ b/drivers/net/bonding/bonding_testpmd.c
@@ -625,6 +625,7 @@ static void cmd_add_bonding_slave_parsed(void
*parsed_result,
slave_port_id, master_port_id);
return;
}
+ ports[master_port_id].update_conf = 1;
init_port_config();
set_port_slave_flag(slave_port_id);
}
@@ -762,6 +763,7 @@ static voi
[PATCH v3 1/2] mempool: cache align mempool cache objects
Add __rte_cache_aligned to the objs array.
It makes no difference in the general case, but if get/put operations are
always 32 objects, it will reduce the number of memory (or last level
cache) accesses from five to four 64 B cache lines for every get/put
operation.
For readability reasons, an example using 16 objects follows:
Currently, with 16 objects (128B), we access to 3
cache lines:
┌┐
│len │
cache ││---
line0 ││ ^
││ |
├┤ | 16 objects
││ | 128B
cache ││ |
line1 ││ |
││ |
├┤ |
││_v_
cache ││
line2 ││
││
└┘
With the alignment, it is also 3 cache lines:
┌┐
│len │
cache ││
line0 ││
││
├┤---
││ ^
cache ││ |
line1 ││ |
││ |
├┤ | 16 objects
││ | 128B
cache ││ |
line2 ││ |
││ v
└┘---
However, accessing the objects at the bottom of the mempool cache is a
special case, where cache line0 is also used for objects.
Consider the next burst (and any following bursts):
Current:
┌┐
│len │
cache ││
line0 ││
││
├┤
││
cache ││
line1 ││
││
├┤
││
cache ││---
line2 ││ ^
││ |
├┤ | 16 objects
││ | 128B
cache ││ |
line3 ││ |
││ |
├┤ |
││_v_
cache ││
line4 ││
││
└┘
4 cache lines touched, incl. line0 for len.
With the proposed alignment:
┌┐
│len │
cache ││
line0 ││
││
├┤
││
cache ││
line1 ││
││
├┤
││
cache ││
line2 ││
││
├┤
││---
cache ││ ^
line3 ││ |
││ | 16 objects
├┤ | 128B
││ |
cache ││ |
line4 ││ |
││_v_
└┘
Only 3 cache lines touched, incl. line0 for len.
Credits go to Olivier Matz for the nice ASCII graphics.
v3:
* No changes. Made part of a series.
v2:
* No such version.
Signed-off-by: Morten Brørup
---
lib/mempool/rte_mempool.h | 6 --
1 file changed, 4 insertions(+), 2 deletions(-)
diff --git a/lib/mempool/rte_mempool.h b/lib/mempool/rte_mempool.h
index 1f5707f46a..3725a72951 100644
--- a/lib/mempool/rte_mempool.h
+++ b/lib/mempool/rte_mempool.h
@@ -86,11 +86,13 @@ struct rte_mempool_cache {
uint32_t size;/**< Size of the cache */
uint32_t flushthresh; /**< Threshold before we flush excess elements */
uint32_t len; /**< Current cache count */
- /*
+ /**
+* Cache objects
+*
* Cache is allocated to this size to allow it to overflow in certain
* cases to avoid needless emptying of cache.
*/
- void *objs[RTE_MEMPOOL_CACHE_MAX_SIZE * 2]; /**< Cache objects */
+ void *objs[RTE_MEMPOOL_CACHE_MAX_SIZE * 2] __rte_cache_aligned;
} __rte_cache_aligned;
/**
--
2.17.1
[PATCH v3 2/2] mempool: optimized debug statistics
When built with debug enabled (RTE_LIBRTE_MEMPOOL_DEBUG defined), the
performance of mempools with caches is improved as follows.
Accessing objects in the mempool is likely to increment either the
put_bulk and put_objs or the get_success_bulk and get_success_objs
debug statistics counters.
By adding an alternative set of these counters to the mempool cache
structure, accesing the dedicated debug statistics structure is avoided in
the likely cases where these counters are incremented.
The trick here is that the cache line holding the mempool cache structure
is accessed anyway, in order to update the "len" field. Updating some
debug statistics counters in the same cache line has lower performance
cost than accessing the debug statistics counters in the dedicated debug
statistics structure, i.e. in another cache line.
Running mempool_perf_autotest on a VMware virtual server shows an avg.
increase of 6.4 % in rate_persec for the tests with cache. (Only when
built with with debug enabled, obviously!)
For the tests without cache, the avg. increase in rate_persec is 0.8 %. I
assume this is noise from the test environment.
v3:
* Try to fix git reference by making part of a series.
* Add --in-reply-to v1 when sending email.
v2:
* Fix spelling and repeated word in commit message, caught by checkpatch.
Signed-off-by: Morten Brørup
---
lib/mempool/rte_mempool.c | 7 +
lib/mempool/rte_mempool.h | 55 +++
2 files changed, 51 insertions(+), 11 deletions(-)
diff --git a/lib/mempool/rte_mempool.c b/lib/mempool/rte_mempool.c
index 21c94a2b9f..7b8c00a022 100644
--- a/lib/mempool/rte_mempool.c
+++ b/lib/mempool/rte_mempool.c
@@ -1285,6 +1285,13 @@ rte_mempool_dump(FILE *f, struct rte_mempool *mp)
sum.get_fail_objs += mp->stats[lcore_id].get_fail_objs;
sum.get_success_blks += mp->stats[lcore_id].get_success_blks;
sum.get_fail_blks += mp->stats[lcore_id].get_fail_blks;
+ /* Add the fast access statistics, if local caches exist */
+ if (mp->cache_size != 0) {
+ sum.put_bulk += mp->local_cache[lcore_id].put_bulk;
+ sum.put_objs += mp->local_cache[lcore_id].put_objs;
+ sum.get_success_bulk +=
mp->local_cache[lcore_id].get_success_bulk;
+ sum.get_success_objs +=
mp->local_cache[lcore_id].get_success_objs;
+ }
}
fprintf(f, " stats:\n");
fprintf(f, "put_bulk=%"PRIu64"\n", sum.put_bulk);
diff --git a/lib/mempool/rte_mempool.h b/lib/mempool/rte_mempool.h
index 3725a72951..d84087bc92 100644
--- a/lib/mempool/rte_mempool.h
+++ b/lib/mempool/rte_mempool.h
@@ -86,6 +86,14 @@ struct rte_mempool_cache {
uint32_t size;/**< Size of the cache */
uint32_t flushthresh; /**< Threshold before we flush excess elements */
uint32_t len; /**< Current cache count */
+#ifdef RTE_LIBRTE_MEMPOOL_DEBUG
+ uint32_t unused;
+ /* Fast access statistics, only for likely events */
+ uint64_t put_bulk; /**< Number of puts. */
+ uint64_t put_objs; /**< Number of objects successfully put.
*/
+ uint64_t get_success_bulk; /**< Successful allocation number. */
+ uint64_t get_success_objs; /**< Objects successfully allocated. */
+#endif
/**
* Cache objects
*
@@ -1327,13 +1335,19 @@ rte_mempool_do_generic_put(struct rte_mempool *mp, void
* const *obj_table,
{
void **cache_objs;
+ /* No cache provided */
+ if (unlikely(cache == NULL))
+ goto driver_enqueue;
+
+#ifdef RTE_LIBRTE_MEMPOOL_DEBUG
/* increment stat now, adding in mempool always success */
- RTE_MEMPOOL_STAT_ADD(mp, put_bulk, 1);
- RTE_MEMPOOL_STAT_ADD(mp, put_objs, n);
+ cache->put_bulk += 1;
+ cache->put_objs += n;
+#endif
- /* No cache provided or the request itself is too big for the cache */
- if (unlikely(cache == NULL || n > cache->flushthresh))
- goto driver_enqueue;
+ /* The request is too big for the cache */
+ if (unlikely(n > cache->flushthresh))
+ goto driver_enqueue_stats_incremented;
/*
* The cache follows the following algorithm:
@@ -1358,6 +1372,12 @@ rte_mempool_do_generic_put(struct rte_mempool *mp, void
* const *obj_table,
driver_enqueue:
+ /* increment stat now, adding in mempool always success */
+ RTE_MEMPOOL_STAT_ADD(mp, put_bulk, 1);
+ RTE_MEMPOOL_STAT_ADD(mp, put_objs, n);
+
+driver_enqueue_stats_incremented:
+
/* push objects to the backend */
rte_mempool_ops_enqueue_bulk(mp, obj_table, n);
}
@@ -1464,8 +1484,10 @@ rte_mempool_do_generic_get(struct rte_mempool *mp, void
**obj_table,
if (remaining == 0) {
/* The entire request is satisfied from the cache. */
- RTE_MEMPOOL_STAT_A
[PATCH v4 1/2] mempool: cache align mempool cache objects
Add __rte_cache_aligned to the objs array.
It makes no difference in the general case, but if get/put operations are
always 32 objects, it will reduce the number of memory (or last level
cache) accesses from five to four 64 B cache lines for every get/put
operation.
For readability reasons, an example using 16 objects follows:
Currently, with 16 objects (128B), we access to 3
cache lines:
┌┐
│len │
cache ││---
line0 ││ ^
││ |
├┤ | 16 objects
││ | 128B
cache ││ |
line1 ││ |
││ |
├┤ |
││_v_
cache ││
line2 ││
││
└┘
With the alignment, it is also 3 cache lines:
┌┐
│len │
cache ││
line0 ││
││
├┤---
││ ^
cache ││ |
line1 ││ |
││ |
├┤ | 16 objects
││ | 128B
cache ││ |
line2 ││ |
││ v
└┘---
However, accessing the objects at the bottom of the mempool cache is a
special case, where cache line0 is also used for objects.
Consider the next burst (and any following bursts):
Current:
┌┐
│len │
cache ││
line0 ││
││
├┤
││
cache ││
line1 ││
││
├┤
││
cache ││---
line2 ││ ^
││ |
├┤ | 16 objects
││ | 128B
cache ││ |
line3 ││ |
││ |
├┤ |
││_v_
cache ││
line4 ││
││
└┘
4 cache lines touched, incl. line0 for len.
With the proposed alignment:
┌┐
│len │
cache ││
line0 ││
││
├┤
││
cache ││
line1 ││
││
├┤
││
cache ││
line2 ││
││
├┤
││---
cache ││ ^
line3 ││ |
││ | 16 objects
├┤ | 128B
││ |
cache ││ |
line4 ││ |
││_v_
└┘
Only 3 cache lines touched, incl. line0 for len.
Credits go to Olivier Matz for the nice ASCII graphics.
v4:
* No changes. Added reviewed- and acked-by tags.
v3:
* No changes. Made part of a series.
v2:
* No such version.
Signed-off-by: Morten Brørup
Reviewed-by: Andrew Rybchenko
Acked-by: Olivier Matz
---
lib/mempool/rte_mempool.h | 6 --
1 file changed, 4 insertions(+), 2 deletions(-)
diff --git a/lib/mempool/rte_mempool.h b/lib/mempool/rte_mempool.h
index 1f5707f46a..3725a72951 100644
--- a/lib/mempool/rte_mempool.h
+++ b/lib/mempool/rte_mempool.h
@@ -86,11 +86,13 @@ struct rte_mempool_cache {
uint32_t size;/**< Size of the cache */
uint32_t flushthresh; /**< Threshold before we flush excess elements */
uint32_t len; /**< Current cache count */
- /*
+ /**
+* Cache objects
+*
* Cache is allocated to this size to allow it to overflow in certain
* cases to avoid needless emptying of cache.
*/
- void *objs[RTE_MEMPOOL_CACHE_MAX_SIZE * 2]; /**< Cache objects */
+ void *objs[RTE_MEMPOOL_CACHE_MAX_SIZE * 2] __rte_cache_aligned;
} __rte_cache_aligned;
/**
--
2.17.1
[PATCH v4 2/2] mempool: optimized debug statistics
When built with debug enabled (RTE_LIBRTE_MEMPOOL_DEBUG defined), the
performance of mempools with caches is improved as follows.
Accessing objects in the mempool is likely to increment either the
put_bulk and put_objs or the get_success_bulk and get_success_objs
debug statistics counters.
By adding an alternative set of these counters to the mempool cache
structure, accessing the dedicated debug statistics structure is avoided in
the likely cases where these counters are incremented.
The trick here is that the cache line holding the mempool cache structure
is accessed anyway, in order to update the "len" field. Updating some
debug statistics counters in the same cache line has lower performance
cost than accessing the debug statistics counters in the dedicated debug
statistics structure, i.e. in another cache line.
Running mempool_perf_autotest on a VMware virtual server shows an avg.
increase of 6.4 % in rate_persec for the tests with cache. (Only when
built with debug enabled, obviously!)
For the tests without cache, the avg. increase in rate_persec is 0.8 %. I
assume this is noise from the test environment.
v4:
* Fix spelling and repeated word in commit message, caught by checkpatch.
v3:
* Try to fix git reference by making part of a series.
* Add --in-reply-to v1 when sending email.
v2:
* Fix spelling and repeated word in commit message, caught by checkpatch.
Signed-off-by: Morten Brørup
---
lib/mempool/rte_mempool.c | 7 +
lib/mempool/rte_mempool.h | 55 +++
2 files changed, 51 insertions(+), 11 deletions(-)
diff --git a/lib/mempool/rte_mempool.c b/lib/mempool/rte_mempool.c
index 21c94a2b9f..7b8c00a022 100644
--- a/lib/mempool/rte_mempool.c
+++ b/lib/mempool/rte_mempool.c
@@ -1285,6 +1285,13 @@ rte_mempool_dump(FILE *f, struct rte_mempool *mp)
sum.get_fail_objs += mp->stats[lcore_id].get_fail_objs;
sum.get_success_blks += mp->stats[lcore_id].get_success_blks;
sum.get_fail_blks += mp->stats[lcore_id].get_fail_blks;
+ /* Add the fast access statistics, if local caches exist */
+ if (mp->cache_size != 0) {
+ sum.put_bulk += mp->local_cache[lcore_id].put_bulk;
+ sum.put_objs += mp->local_cache[lcore_id].put_objs;
+ sum.get_success_bulk +=
mp->local_cache[lcore_id].get_success_bulk;
+ sum.get_success_objs +=
mp->local_cache[lcore_id].get_success_objs;
+ }
}
fprintf(f, " stats:\n");
fprintf(f, "put_bulk=%"PRIu64"\n", sum.put_bulk);
diff --git a/lib/mempool/rte_mempool.h b/lib/mempool/rte_mempool.h
index 3725a72951..d84087bc92 100644
--- a/lib/mempool/rte_mempool.h
+++ b/lib/mempool/rte_mempool.h
@@ -86,6 +86,14 @@ struct rte_mempool_cache {
uint32_t size;/**< Size of the cache */
uint32_t flushthresh; /**< Threshold before we flush excess elements */
uint32_t len; /**< Current cache count */
+#ifdef RTE_LIBRTE_MEMPOOL_DEBUG
+ uint32_t unused;
+ /* Fast access statistics, only for likely events */
+ uint64_t put_bulk; /**< Number of puts. */
+ uint64_t put_objs; /**< Number of objects successfully put.
*/
+ uint64_t get_success_bulk; /**< Successful allocation number. */
+ uint64_t get_success_objs; /**< Objects successfully allocated. */
+#endif
/**
* Cache objects
*
@@ -1327,13 +1335,19 @@ rte_mempool_do_generic_put(struct rte_mempool *mp, void
* const *obj_table,
{
void **cache_objs;
+ /* No cache provided */
+ if (unlikely(cache == NULL))
+ goto driver_enqueue;
+
+#ifdef RTE_LIBRTE_MEMPOOL_DEBUG
/* increment stat now, adding in mempool always success */
- RTE_MEMPOOL_STAT_ADD(mp, put_bulk, 1);
- RTE_MEMPOOL_STAT_ADD(mp, put_objs, n);
+ cache->put_bulk += 1;
+ cache->put_objs += n;
+#endif
- /* No cache provided or the request itself is too big for the cache */
- if (unlikely(cache == NULL || n > cache->flushthresh))
- goto driver_enqueue;
+ /* The request is too big for the cache */
+ if (unlikely(n > cache->flushthresh))
+ goto driver_enqueue_stats_incremented;
/*
* The cache follows the following algorithm:
@@ -1358,6 +1372,12 @@ rte_mempool_do_generic_put(struct rte_mempool *mp, void
* const *obj_table,
driver_enqueue:
+ /* increment stat now, adding in mempool always success */
+ RTE_MEMPOOL_STAT_ADD(mp, put_bulk, 1);
+ RTE_MEMPOOL_STAT_ADD(mp, put_objs, n);
+
+driver_enqueue_stats_incremented:
+
/* push objects to the backend */
rte_mempool_ops_enqueue_bulk(mp, obj_table, n);
}
@@ -1464,8 +1484,10 @@ rte_mempool_do_generic_get(struct rte_mempool *mp, void
**obj_table,
if (remaining == 0) {
/* The entire re
Re: [PATCH v4 1/2] build: allow to conditionally build apps
On 10/27/22 16:22, David Marchand wrote: On Thu, Oct 27, 2022 at 3:13 PM David Marchand wrote: On Fri, Oct 14, 2022 at 9:51 AM Markus Theil wrote: Makes apps configurable from meson, like already possible for drivers. Signed-off-by: Markus Theil With these new options, it's hard to tell which and why some application is built (or not). Can we have an output similar to libraries and drivers (see toplevel meson.build) ? WDYT of followup patch: https://github.com/david-marchand/dpdk/commit/apps LGTM
Re: [PATCH] telemetry: make usable from C++
On 10/27/22 14:04, David Marchand wrote: On Wed, Oct 12, 2022 at 2:05 PM Markus Theil wrote: From: Markus Theil Add missing include in order to make C++ compilers happy. We have build checks for headers (-Dcheck_includes meson option), like for C++: $ cat $BUILDDIR/buildtools/chkincs/chkincs-cpp.p/rte_telemetry.cpp #include "/home/dmarchan/dpdk/lib/telemetry/rte_telemetry.h" [917/1289] ccache c++ -Ibuildtools/chkincs/chkincs-cpp.p -Ibuildtools/chkincs -I../../../dpdk/buildtools/chkincs -Iexamples/vmdq_dcb -I../../../dpdk/examples/vmdq_dcb -I../../../dpdk/examples/common -Idrivers/bus/vdev -I../../../dpdk/drivers/bus/vdev -I. -I../../../dpdk -Iconfig -I../../../dpdk/config -Ilib/eal/include -I../../../dpdk/lib/eal/include -Ilib/eal/linux/include -I../../../dpdk/lib/eal/linux/include -Ilib/eal/x86/include -I../../../dpdk/lib/eal/x86/include -Ilib/eal/common -I../../../dpdk/lib/eal/common -Ilib/eal -I../../../dpdk/lib/eal -Ilib/kvargs -I../../../dpdk/lib/kvargs -Ilib/metrics -I../../../dpdk/lib/metrics -Ilib/telemetry -I../../../dpdk/lib/telemetry -Idrivers/bus/pci -I../../../dpdk/drivers/bus/pci -I../../../dpdk/drivers/bus/pci/linux -Ilib/pci -I../../../dpdk/lib/pci -Ilib/ring -I../../../dpdk/lib/ring -Ilib/rcu -I../../../dpdk/lib/rcu -Ilib/mempool -I../../../dpdk/lib/mempool -Ilib/mbuf -I../../../dpdk/lib/mbuf -Ilib/net -I../../../dpdk/lib/net -Ilib/meter -I../../../dpdk/lib/meter -Ilib/ethdev -I../../../dpdk/lib/ethdev -Ilib/cmdline -I../../../dpdk/lib/cmdline -Ilib/hash -I../../../dpdk/lib/hash -Ilib/timer -I../../../dpdk/lib/timer -Ilib/acl -I../../../dpdk/lib/acl -Ilib/bbdev -I../../../dpdk/lib/bbdev -Ilib/bitratestats -I../../../dpdk/lib/bitratestats -Ilib/bpf -I../../../dpdk/lib/bpf -Ilib/cfgfile -I../../../dpdk/lib/cfgfile -Ilib/compressdev -I../../../dpdk/lib/compressdev -Ilib/cryptodev -I../../../dpdk/lib/cryptodev -Ilib/distributor -I../../../dpdk/lib/distributor -Ilib/efd -I../../../dpdk/lib/efd -Ilib/eventdev -I../../../dpdk/lib/eventdev -Ilib/gpudev -I../../../dpdk/lib/gpudev -Ilib/gro -I../../../dpdk/lib/gro -Ilib/gso -I../../../dpdk/lib/gso -Ilib/ip_frag -I../../../dpdk/lib/ip_frag -Ilib/jobstats -I../../../dpdk/lib/jobstats -Ilib/kni -I../../../dpdk/lib/kni -Ilib/latencystats -I../../../dpdk/lib/latencystats -Ilib/lpm -I../../../dpdk/lib/lpm -Ilib/member -I../../../dpdk/lib/member -Ilib/pcapng -I../../../dpdk/lib/pcapng -Ilib/power -I../../../dpdk/lib/power -Ilib/rawdev -I../../../dpdk/lib/rawdev -Ilib/regexdev -I../../../dpdk/lib/regexdev -Ilib/dmadev -I../../../dpdk/lib/dmadev -Ilib/rib -I../../../dpdk/lib/rib -Ilib/reorder -I../../../dpdk/lib/reorder -Ilib/sched -I../../../dpdk/lib/sched -Ilib/security -I../../../dpdk/lib/security -Ilib/stack -I../../../dpdk/lib/stack -Ilib/vhost -I../../../dpdk/lib/vhost -Ilib/ipsec -I../../../dpdk/lib/ipsec -Ilib/fib -I../../../dpdk/lib/fib -Ilib/port -I../../../dpdk/lib/port -Ilib/pdump -I../../../dpdk/lib/pdump -Ilib/table -I../../../dpdk/lib/table -Ilib/pipeline -I../../../dpdk/lib/pipeline -Ilib/flow_classify -I../../../dpdk/lib/flow_classify -Ilib/graph -I../../../dpdk/lib/graph -Ilib/node -I../../../dpdk/lib/node -fdiagnostics-color=always -D_FILE_OFFSET_BITS=64 -Wall -Winvalid-pch -Wnon-virtual-dtor -Wextra -Werror -O2 -g -include rte_config.h -march=nehalem -DALLOW_EXPERIMENTAL_API -DALLOW_INTERNAL_API -MD -MQ buildtools/chkincs/chkincs-cpp.p/meson-generated_rte_telemetry.cpp.o -MF buildtools/chkincs/chkincs-cpp.p/meson-generated_rte_telemetry.cpp.o.d -o buildtools/chkincs/chkincs-cpp.p/meson-generated_rte_telemetry.cpp.o -c buildtools/chkincs/chkincs-cpp.p/rte_telemetry.cpp Besides, I fail to see the need for stddef.h. In which setup / case, did you get a compilation issue? I traced back, that we need this only for some internal patches which expose the JSON output of telemetry inside our DPDK application in order to get SFF digital optics module data without piping the telemetry socket back into our application. It can be therefore ignored and was posted by mistake. I'm archiving it.
