[dpdk-dev] [PATCH 0/5] net/ena: minor fixes and docs update

[dpdk-dev] [PATCH 0/5] net/ena: minor fixes and docs update

[Michal Krawczyk]

Hi,

we would like to push fixes for part of the bugs reported in Bug 193,
together with docs update for ENAv2 about enabling WC support, release
notes and fix for assigning rte_errno negative values, as requested by
the community.

Michal Krawczyk (3):
  net/ena: remove lcore constraint from max queue number
  net/ena: fix dev init for RTE_PROC_SECONDARY
  net/ena: fix setting rte_errno to negative value

Rafal Kozik (2):
  doc: update release notes for ENA PMD
  doc: add instructions for enabling WC in ENAv2 PMD

 doc/guides/nics/ena.rst                | 16 ++++++++++++++--
 doc/guides/rel_notes/release_19_02.rst | 19 +++++++++++++++++++
 drivers/net/ena/ena_ethdev.c           | 20 ++++++++++----------
 3 files changed, 43 insertions(+), 12 deletions(-)

-- 
2.7.4

[dpdk-dev] [PATCH 1/5] net/ena: remove lcore constraint from max queue number

[Michal Krawczyk]

The number of queues was limited by number of lcores in the recent ENAv2
patch. However, some apps like symmetric_mp, are configuring multiple
queues although there is only single lcore used.

To prevent failure in that case, the lcore number constraint was
removed.

Fixes: ea93d37eb49d ("net/ena: add HW queues depth setup")

Signed-off-by: Michal Krawczyk <mk@semihalf.com>
---
 drivers/net/ena/ena_ethdev.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/drivers/net/ena/ena_ethdev.c b/drivers/net/ena/ena_ethdev.c
index 3022dda..fde5fa7 100644
--- a/drivers/net/ena/ena_ethdev.c
+++ b/drivers/net/ena/ena_ethdev.c
@@ -1687,8 +1687,7 @@ static int ena_calc_io_queue_num(struct ena_com_dev *ena_dev,
 	if (ena_dev->tx_mem_queue_type == ENA_ADMIN_PLACEMENT_POLICY_DEV)
 		io_tx_sq_num = get_feat_ctx->llq.max_llq_num;
 
-	io_queue_num = RTE_MIN(rte_lcore_count(), ENA_MAX_NUM_IO_QUEUES);
-	io_queue_num = RTE_MIN(io_queue_num, io_rx_num);
+	io_queue_num = RTE_MIN(ENA_MAX_NUM_IO_QUEUES, io_rx_num);
 	io_queue_num = RTE_MIN(io_queue_num, io_tx_sq_num);
 	io_queue_num = RTE_MIN(io_queue_num, io_tx_cq_num);
 
-- 
2.7.4

[dpdk-dev] [PATCH 2/5] net/ena: fix dev init for RTE_PROC_SECONDARY

[Michal Krawczyk]

The check for proc type in eth_ena_dev_init() should appear before
modyfing adapter structure.

Calling memset on ena_adapter from secondary process context, was
erasing all structure information, and it was causing the crash of the
main process.

Fixes: 1173fca25af9 ("ena: add polling-mode driver")
Cc: stable@dpdk.org

Signed-off-by: Michal Krawczyk <mk@semihalf.com>
---
 drivers/net/ena/ena_ethdev.c | 11 ++++++-----
 1 file changed, 6 insertions(+), 5 deletions(-)

diff --git a/drivers/net/ena/ena_ethdev.c b/drivers/net/ena/ena_ethdev.c
index fde5fa7..747390b 100644
--- a/drivers/net/ena/ena_ethdev.c
+++ b/drivers/net/ena/ena_ethdev.c
@@ -1715,19 +1715,20 @@ static int eth_ena_dev_init(struct rte_eth_dev *eth_dev)
 	static int adapters_found;
 	bool wd_state;
 
-	memset(adapter, 0, sizeof(struct ena_adapter));
-	ena_dev = &adapter->ena_dev;
-
 	eth_dev->dev_ops = &ena_dev_ops;
 	eth_dev->rx_pkt_burst = &eth_ena_recv_pkts;
 	eth_dev->tx_pkt_burst = &eth_ena_xmit_pkts;
 	eth_dev->tx_pkt_prepare = &eth_ena_prep_pkts;
-	adapter->rte_eth_dev_data = eth_dev->data;
-	adapter->rte_dev = eth_dev;
 
 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
 		return 0;
 
+	memset(adapter, 0, sizeof(struct ena_adapter));
+	ena_dev = &adapter->ena_dev;
+
+	adapter->rte_eth_dev_data = eth_dev->data;
+	adapter->rte_dev = eth_dev;
+
 	pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
 	adapter->pdev = pci_dev;
 
-- 
2.7.4

[dpdk-dev] [PATCH 3/5] net/ena: fix setting rte_errno to negative value

[Michal Krawczyk]

The rte_errno shouldn't have assigned negative error codes, so it has
to be fixed.

Fixes: b3fc5a1ae10d ("net/ena: add Tx preparation")
Cc: stable@dpdk.org

Signed-off-by: Michal Krawczyk <mk@semihalf.com>
---
 drivers/net/ena/ena_ethdev.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/net/ena/ena_ethdev.c b/drivers/net/ena/ena_ethdev.c
index 747390b..8bb05ca 100644
--- a/drivers/net/ena/ena_ethdev.c
+++ b/drivers/net/ena/ena_ethdev.c
@@ -2178,14 +2178,14 @@ eth_ena_prep_pkts(void *tx_queue, struct rte_mbuf **tx_pkts,
 		if ((ol_flags & ENA_TX_OFFLOAD_NOTSUP_MASK) != 0 ||
 				(ol_flags & PKT_TX_L4_MASK) ==
 				PKT_TX_SCTP_CKSUM) {
-			rte_errno = -ENOTSUP;
+			rte_errno = ENOTSUP;
 			return i;
 		}
 
 #ifdef RTE_LIBRTE_ETHDEV_DEBUG
 		ret = rte_validate_tx_offload(m);
 		if (ret != 0) {
-			rte_errno = ret;
+			rte_errno = -ret;
 			return i;
 		}
 #endif
@@ -2198,7 +2198,7 @@ eth_ena_prep_pkts(void *tx_queue, struct rte_mbuf **tx_pkts,
 		ret = rte_net_intel_cksum_flags_prepare(m,
 			ol_flags & ~PKT_TX_TCP_SEG);
 		if (ret != 0) {
-			rte_errno = ret;
+			rte_errno = -ret;
 			return i;
 		}
 	}
-- 
2.7.4

[dpdk-dev] [PATCH 4/5] doc: update release notes for ENA PMD

[Michal Krawczyk]

From: Rafal Kozik <rk@semihalf.com>

Add changes in ENA PMD to DPDK v19.02 release notes.

Signed-off-by: Michal Krawczyk <mk@semihalf.com>
Signed-off-by: Rafal Kozik <rk@semihalf.com>
---
 doc/guides/rel_notes/release_19_02.rst | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)

diff --git a/doc/guides/rel_notes/release_19_02.rst b/doc/guides/rel_notes/release_19_02.rst
index 57c595b..15a92d5 100644
--- a/doc/guides/rel_notes/release_19_02.rst
+++ b/doc/guides/rel_notes/release_19_02.rst
@@ -155,6 +155,25 @@ New Features
   the intel_pstate kernel driver is now supported, and automatically
   detected by the library.
 
+* **Release of the ENA PMD v2.0.0**
+
+  * Added Low Latency Queue v2 (LLQv2). This feature reduces the latency
+    of the packets by pushing the header directly through the PCI to the
+    device. This allows the NIC to start handle packet right after the doorbell
+    without waiting for DMA.
+  * Added independent configuration of HW Tx and Rx ring depths.
+  * Added support for up to 8k Rx descriptors per ring.
+  * Added additional doorbell check on Tx, to handle Tx more efficiently for big
+    bursts of packets.
+  * Added per queue statistics.
+  * Added extended statistics using xstats DPDK API.
+  * The reset routine was aligned with the DPDK API, so now it can be
+    handled as in other PMDs.
+  * Fixed out of order (OOO) completion.
+  * Fixed memory leaks due to port stops and starts in the middle of
+    traffic.
+  * Updated documentation and features list of the PMD.
+
 
 Removed Items
 -------------
-- 
2.7.4

[dpdk-dev] [PATCH 5/5] doc: add instructions for enabling WC in ENAv2 PMD

[Michal Krawczyk]

From: Rafal Kozik <rk@semihalf.com>

ENAv2 hardware provides Low Latency Queue v2 (LLQv2). It needs kernel
PCI driver to support write combining (WC). This patch add information
how to use it with igb_uio and vfio-pci drivers.

Signed-off-by: Rafal Kozik <rk@semihalf.com>
Acked-by: Michal Krawczyk <mk@semihalf.com>
---
 doc/guides/nics/ena.rst | 16 ++++++++++++++--
 1 file changed, 14 insertions(+), 2 deletions(-)

diff --git a/doc/guides/nics/ena.rst b/doc/guides/nics/ena.rst
index 9f0a650..80da4b6 100644
--- a/doc/guides/nics/ena.rst
+++ b/doc/guides/nics/ena.rst
@@ -183,10 +183,22 @@ Prerequisites
 #. Prepare the system as recommended by DPDK suite.  This includes environment
    variables, hugepages configuration, tool-chains and configuration.
 
-#. ENA PMD can operate with ``vfio-pci`` or ``igb_uio`` driver.
+#. ENA PMD can operate with ``vfio-pci``(*) or ``igb_uio`` driver.
+
+   (*) ENAv2 hardware supports Low Latency Queue v2 (LLQv2). This feature
+   reduces the latency of the packets by pushing the header directly through
+   the PCI to the device, before the DMA is even triggered. For proper work
+   kernel PCI driver must support write combining (WC). In mainline version of
+   ``igb_uio`` (in DPDK repo) it must be enabled by loding module with
+   ``wc_activate=1`` flag (example below). However, mainline's vfio-pci
+   driver in kernel doesn't have WC support yet (planed to be added).
+   If vfio-pci used user should be either turn off ENAv2 (to avoid performance
+   impact) or recompile vfio-pci driver with patch provided in
+   `amzn-github <https://github.com/amzn/amzn-drivers/tree/master/userspace/dpdk/enav2-vfio-patch>`_.
 
 #. Insert ``vfio-pci`` or ``igb_uio`` kernel module using the command
-   ``modprobe vfio-pci`` or ``modprobe igb_uio`` respectively.
+   ``modprobe vfio-pci`` or ``modprobe uio; insmod igb_uio.ko wc_activate=1``
+   respectively.
 
 #. For ``vfio-pci`` users only:
    Please make sure that ``IOMMU`` is enabled in your system,
-- 
2.7.4

Re: [dpdk-dev] [PATCH 0/5] net/ena: minor fixes and docs update

[Thomas Monjalon]

25/01/2019 09:10, Michal Krawczyk:
> Hi,
> 
> we would like to push fixes for part of the bugs reported in Bug 193,
> together with docs update for ENAv2 about enabling WC support, release
> notes and fix for assigning rte_errno negative values, as requested by
> the community.
> 
> Michal Krawczyk (3):
>   net/ena: remove lcore constraint from max queue number
>   net/ena: fix dev init for RTE_PROC_SECONDARY
>   net/ena: fix setting rte_errno to negative value
> 
> Rafal Kozik (2):
>   doc: update release notes for ENA PMD
>   doc: add instructions for enabling WC in ENAv2 PMD
> 
>  doc/guides/nics/ena.rst                | 16 ++++++++++++++--
>  doc/guides/rel_notes/release_19_02.rst | 19 +++++++++++++++++++
>  drivers/net/ena/ena_ethdev.c           | 20 ++++++++++----------
>  3 files changed, 43 insertions(+), 12 deletions(-)

This patchset is sent very late in the cycle.
After some thoughts (with Ferruh), I decide to merge it in 19.02.
I hope there will be no more request for ENA in this release,
and please avoid this situation in future.