Re: [dpdk-dev] [PATCH 0/2] remove use of weak functions from libraries
On Mon, May 27, 2019 at 04:57:15PM -0400, Aaron Conole wrote: > Bruce Richardson writes: > > > On Mon, May 27, 2019 at 04:13:53PM +0200, David Marchand wrote: > >>Hello, > >>On Wed, Apr 10, 2019 at 3:45 PM Bruce Richardson > >><[1]bruce.richard...@intel.com> wrote: > >> > >> Weak functions don't work well with static library builds as the > >> linker > >> always picks the first version of a function irrespective of whether > >> it is > >> weak or not. The solution to this is to use the "whole-archive" flag > >> when > >> linking, but this has the nasty side-effect that it causes the > >> resulting > >> binary to be larger than it needs to be. > >> A further problem with this approach of using "whole-archive" is > >> that one > >> either needs to link all libraries with this flag or track what > >> libraries > >> need it or not - the latter being especially a problem for apps not > >> using > >> the DPDK build system itself (i.e. most apps not shipped with DPDK > >> itself). > >> For meson builds this information needs to make its way all the way > >> through > >> to the pkgconfig file generated - not a trivial undertaking. > >> Thankfully, though, the use of weak functions is limited to use for > >> multiple functions within a single library, meaning that when the > >> lib is > >> being built, the build system itself can know whether the "weak" > >> function > >> is needed or not. This allows us to change the weak function to a > >> conditionally compiled regular function used in fallback case. This > >> makes > >> the need for "whole-archive" flag, and any special linking measures > >> for the > >> library, go away. > >> [This set does not touch the drivers, only the libraries, since > >> there are > >> other special linker flags needed for drivers in general, making the > >> problem less severe for driver .a files.] > >> > >>Was there something preventing this patchset from getting merged ? > >>-- > >>David Marchand > >> > > I believe Aaron Conole had some changes in the same area and was looking to > > roll these changes into a bigger patchset of his. Aaron, can you please > > confirm? > > Yes - Sorry the patches are in my queue. Maybe it just makes sense to > merge these though? > Funnily enough, I've no objections to that :-)
[dpdk-dev] [PATCH] net/ena: update Rx checksum errors on Rx path
Rx checksum flags and input errors shouldn't be updated on Tx, as it
would work only for packets forwarding.
The ierrors statistic should be updated on Rx, right after checking
Rx checksum flags if the Rx checksum offload is enabled.
Signed-off-by: Michal Krawczyk
---
drivers/net/ena/ena_ethdev.c | 8 +++-
1 file changed, 3 insertions(+), 5 deletions(-)
diff --git a/drivers/net/ena/ena_ethdev.c b/drivers/net/ena/ena_ethdev.c
index b6651fc0f..2cc4a169f 100644
--- a/drivers/net/ena/ena_ethdev.c
+++ b/drivers/net/ena/ena_ethdev.c
@@ -2105,8 +2105,10 @@ static uint16_t eth_ena_recv_pkts(void *rx_queue, struct
rte_mbuf **rx_pkts,
ena_rx_mbuf_prepare(mbuf_head, &ena_rx_ctx);
if (unlikely(mbuf_head->ol_flags &
- (PKT_RX_IP_CKSUM_BAD | PKT_RX_L4_CKSUM_BAD)))
+ (PKT_RX_IP_CKSUM_BAD | PKT_RX_L4_CKSUM_BAD))) {
+ rte_atomic64_inc(&rx_ring->adapter->drv_stats->ierrors);
++rx_ring->rx_stats.bad_csum;
+ }
mbuf_head->hash.rss = ena_rx_ctx.hash;
@@ -2334,10 +2336,6 @@ static uint16_t eth_ena_xmit_pkts(void *tx_queue, struct
rte_mbuf **tx_pkts,
/* Set TX offloads flags, if applicable */
ena_tx_mbuf_prepare(mbuf, &ena_tx_ctx, tx_ring->offloads);
- if (unlikely(mbuf->ol_flags &
-(PKT_RX_L4_CKSUM_BAD | PKT_RX_IP_CKSUM_BAD)))
- rte_atomic64_inc(&tx_ring->adapter->drv_stats->ierrors);
-
rte_prefetch0(tx_pkts[(sent_idx + 4) & ring_mask]);
/* Process first segment taking into
--
2.14.1
Re: [dpdk-dev] [PATCH] app/compress-perf: add info about socket id used by app and by the driver
Added info about socket id used by compression perf test process and by the driver Signed-off-by: Tomasz Cel Signed-off-by: Fiona Trahe Acked-by: Artur Trybula < arturx.tryb...@intel.com> ---
Re: [dpdk-dev] [PATCH v2 03/10] net/bnxt: implement vector mode driver
On 5/24/2019 3:49 PM, Lance Richardson wrote:
> Introduce vector mode support for the bnxt pmd.
>
> Signed-off-by: Lance Richardson
> Signed-off-by: Ajit Khaparde
> ---
> v2:
> * Squashed with v1 patch 4 ("fix double counting VLAN tags").
> * Dropped two unnecessary coding style changes from bnxt_txr.h.
>
> config/common_base | 1 +
> drivers/net/bnxt/Makefile| 1 +
> drivers/net/bnxt/bnxt_ethdev.c | 95 +-
> drivers/net/bnxt/bnxt_ring.h | 3 +-
> drivers/net/bnxt/bnxt_rxq.c | 5 +
> drivers/net/bnxt/bnxt_rxq.h | 4 +
> drivers/net/bnxt/bnxt_rxr.h | 9 +-
> drivers/net/bnxt/bnxt_rxtx_vec_sse.c | 481 +++
> drivers/net/bnxt/bnxt_txr.h | 5 +
> drivers/net/bnxt/meson.build | 5 +
> 10 files changed, 600 insertions(+), 9 deletions(-)
> create mode 100644 drivers/net/bnxt/bnxt_rxtx_vec_sse.c
>
> diff --git a/config/common_base b/config/common_base
> index 6b96e0e80..1bbb7c10b 100644
> --- a/config/common_base
> +++ b/config/common_base
> @@ -212,6 +212,7 @@ CONFIG_RTE_LIBRTE_BNX2X_DEBUG_PERIODIC=n
> # Compile burst-oriented Broadcom BNXT PMD driver
> #
> CONFIG_RTE_LIBRTE_BNXT_PMD=y
> +CONFIG_RTE_LIBRTE_BNXT_INC_VECTOR=n
I think this compile flag can be omitted, it should be possible to detect if
vector PMD can be run dynamically. And indeed meson build is doing something
similar. Please check 'ice' driver for sample.
>
> #
> # Compile burst-oriented Chelsio Terminator (CXGBE) PMD
> diff --git a/drivers/net/bnxt/Makefile b/drivers/net/bnxt/Makefile
> index 8be3cb0e4..9e006b5d1 100644
> --- a/drivers/net/bnxt/Makefile
> +++ b/drivers/net/bnxt/Makefile
> @@ -41,6 +41,7 @@ SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_vnic.c
> SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_irq.c
> SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_util.c
> SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += rte_pmd_bnxt.c
> +SRCS-$(CONFIG_RTE_LIBRTE_BNXT_INC_VECTOR) += bnxt_rxtx_vec_sse.c
You need a x86 protect for this, since DPDK supports multiple arch
<...>
Re: [dpdk-dev] [PATCH v2 03/10] net/bnxt: implement vector mode driver
On 5/24/2019 3:49 PM, Lance Richardson wrote:
> @@ -1597,6 +1659,8 @@
>
> bnxt_txq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
>
> static int bnxt_mtu_set_op(struct rte_eth_dev *eth_dev, uint16_t new_mtu)
> {
> + uint32_t new_pkt_size = new_mtu + ETHER_HDR_LEN + ETHER_CRC_LEN +
> + VLAN_TAG_SIZE * BNXT_NUM_VLANS;
'ETHER_HDR_LEN' & 'ETHER_CRC_LEN' are renamed to
'RTE_ETHER_HDR_LEN' & 'RTE_ETHER_CRC_LEN'
I am mostly fixing these while applying but since there will be a new version,
can you please rebase on top of latest next-net, to be able to catch these.
[dpdk-dev] Some sad news on Rami Rosen
Hi folks, I’m sorry to inform the DPDK Community that Rami Rosen has passed away. Rami was well known in Software Defined Networking circles were he literally wrote the book on Linux Kernel Networking [1]. He was a frequent contributor to DPDK documentation submitting and acknowledging patches, and was well-known and liked in the community [2]. Most recently, he was due to speak at DPDK Summit China next month on Developing CI/CD for Huawei Smartnic DPDK PMD using DTS [3]. I was looking forward to meeting him there. His funeral was held in Israel yesterday at the Kiryat Tiv'on Cemetery, he was a good friend and he will be sorely missed by all who knew him. Ar dheis Dé go raibh a anam, Ray K [1] https://www.apress.com/gp/book/9781430261964 [2] https://www.openhub.net/people?query=rami%20rosen [3] https://dpdkchina2019.sched.com/
Re: [dpdk-dev] [PATCH v2 09/10] net/bnxt: HWRM version update
On 5/24/2019 3:49 PM, Lance Richardson wrote: > From: Ajit Khaparde > > Update HWRM API to version 1.10.0.74 There are three commits related: "update HWRM API" (to version 1.10.0.19) "update HWRM version" (to 1.10.0.48) "HWRM version update" (to version 1.10.0.74) Is "HWRM API" & "HWRM version" different things? I assume they are not different and these commits are updating "HWRM API version" gradually, if so can you please use a consistent commit title? Thanks, ferruh > > Signed-off-by: Ajit Khaparde > Reviewed-by: Lance Richardson <...>
Re: [dpdk-dev] [PATCH v2 10/10] net/bnxt: update release notes for bnxt
On 5/24/2019 3:49 PM, Lance Richardson wrote: > From: Ajit Khaparde > > Update release doc briefly describing updates to bnxt PMD. It is better to keep the doc update with the feature patch, instead of a separate doc patch. Can you please distribute content of this patch into relevant patches? > > Signed-off-by: Ajit Khaparde > --- > doc/guides/rel_notes/release_19_08.rst | 6 ++ > 1 file changed, 6 insertions(+) > > diff --git a/doc/guides/rel_notes/release_19_08.rst > b/doc/guides/rel_notes/release_19_08.rst > index b9510f93a..2ff9d4b1c 100644 > --- a/doc/guides/rel_notes/release_19_08.rst > +++ b/doc/guides/rel_notes/release_19_08.rst > @@ -54,6 +54,12 @@ New Features > Also, make sure to start the actual text at the margin. > = > > +* **Updated the bnxt PMD.** > + > + Updated the bnxt PMD. The major enhancements include: > + * Update HWRM version to 1.10.74 > + * Added vector RX/TX We have x86(SSE/AVX2/AVX512), Arm (neon) & PPC (altivec) vectorization support in DPDK, can you please detail what is added. > + * Performance optimizations in non-vector Tx path > > Removed Items > - >
Re: [dpdk-dev] [PATCH v2 00/10] bnxt patchset
On 5/24/2019 3:49 PM, Lance Richardson wrote: > This patchset bsae on top of the previous submission adds the following: > 1) Support for vector mode TX and RX. > 2) HWRM API update (split into multiple patches). > 3) Fixes for RSS reta update and query. > > It also updates the release notes. > > v2: > * Squashed patches 3 and 4 from v1 patchset. > * Added Meson build support for vector mode PMD. > * Dropped two unnecessary code style changes from patch 3. > > Ajit Khaparde (5): > net/bnxt: fix RSS reta indirection table update > net/bnxt: update HWRM API > net/bnxt: update HWRM version > net/bnxt: HWRM version update > net/bnxt: update release notes for bnxt > > Lance Richardson (5): > net/bnxt: move tx bd checking to header file > net/bnxt: compute and store scattered RX status > net/bnxt: implement vector mode driver > net/bnxt: use reta update mask and translate qid to grp id > net/bnxt: fix reta query op There are warnings from "./devtools/checkpatches.sh" & "./devtools/check-git-log.sh" Can you please address them in next version? It is acceptable to have some level of "checkpatches.sh" in the base code update (HWRM update code for your case) but please sure DPDK specific files don't have any warning. Thanks, ferruh
Re: [dpdk-dev] AF_XDP PMD cannot use with software NICs
On 25-May-19 1:45 AM, Yutaro Hayakawa wrote: Hello, In current AF_XDP PMD implementation, we cannot use it with software NICs like VETH. Because it allocates umem with RTE_MEMZONE_IOVA_CONTIG. Since AF_XDP itself is usable with any network devices, I think it is better to have option to take umem with RTE_MEMZONE_SIZE_HINT_ONLY. It is also useful for testing perpose or using DPDK from containers. I'm not sure i understand the problem. IOVA_CONTIG has no relation to SIZE_HINT_ONLY - the former forces the allocation to be IOVA-contiguous. The latter allows to allocate from other page sizes when the requested one is not available, e.g. if you request memory with RTE_MEMZONE_1GB and RTE_MEMZONE_SIZE_HINT_ONLY, it will try to allocate from 1GB pages first, and if that fails, will allocate from any other page size available - but says nothing about whether that memory has to be IOVA-contiguous. What is the problem with allocating IOVA-contiguous memory for AF_XDP and VETH? Any thought or future plan for this problem? Thanks, Yutaro -- Thanks, Anatoly
Re: [dpdk-dev] [PATCH 0/3] fix 32-bit meson builds
On Mon, 2019-05-27 at 17:15 +0100, Bruce Richardson wrote: > This set fixes some issues seen on the automated CI system with > building > on 32-bit Linux distro's using meson and ninja. The fixes are to > disable > unsupported parts of the build, and switch the 32-bit builds to > always > having large file support, so that make and meson are consistent in > that > regard. > > Bruce Richardson (3): > net/nfp: disable nfp for 32-bit meson builds > build: enable large file support on 32-bit > build: remove unnecessary large file support defines > > drivers/common/dpaax/dpaax_iova_table.c | 4 ++-- > drivers/net/nfp/meson.build | 2 +- > examples/vhost_crypto/meson.build | 1 - > examples/vhost_scsi/meson.build | 1 - > lib/librte_eal/linux/eal/eal_memalloc.c | 1 - > lib/librte_eal/linux/eal/eal_memory.c | 1 - > lib/librte_vhost/Makefile | 2 +- > mk/arch/arm/rte.vars.mk | 2 +- > mk/arch/i686/rte.vars.mk| 2 +- > 9 files changed, 6 insertions(+), 10 deletions(-) Series-acked-by: Luca Boccassi -- Kind regards, Luca Boccassi
Re: [dpdk-dev] [PATCH v3] cryptodev: make xform key pointer constant
Hi Akhil, This patch was sent in 19.05 but I just found it can be applied to 19.08 RC0 cleanly. Do you think I need to send a v4 (with no change) for 19.08? Regards, Fan > -Original Message- > From: Akhil Goyal [mailto:akhil.go...@nxp.com] > Sent: Friday, March 29, 2019 4:02 PM > To: Thomas Monjalon > Cc: Zhang, Roy Fan ; dev@dpdk.org; Trahe, Fiona > ; Kusztal, ArkadiuszX > Subject: Re: [dpdk-dev] [PATCH v3] cryptodev: make xform key pointer > constant > > > > On 3/29/2019 9:18 PM, Akhil Goyal wrote: > > > > On 3/29/2019 7:49 PM, Thomas Monjalon wrote: > >> 29/03/2019 15:07, Akhil Goyal: > >>> On 3/20/2019 7:02 PM, Fan Zhang wrote: > drivers/common/cpt/cpt_ucode.h | 14 ++--- > drivers/crypto/aesni_gcm/aesni_gcm_pmd.c| 2 +- > drivers/crypto/aesni_mb/rte_aesni_mb_pmd.c | 2 +- > drivers/crypto/openssl/rte_openssl_pmd.c| 8 +-- > drivers/crypto/qat/qat_sym_session.c| 8 +-- > drivers/crypto/qat/qat_sym_session.h| 4 +- > drivers/crypto/snow3g/rte_snow3g_pmd.c | 23 ++-- > drivers/crypto/snow3g/rte_snow3g_pmd_private.h | 1 + > drivers/crypto/virtio/virtio_cryptodev.c| 30 +++--- > drivers/crypto/virtio/virtio_cryptodev.h| 2 + > drivers/net/ixgbe/ixgbe_ipsec.c | 35 +--- > drivers/net/ixgbe/ixgbe_ipsec.h | 3 +- > drivers/net/softnic/rte_eth_softnic_cli.c | 74 > -- > --- > drivers/net/softnic/rte_eth_softnic_internals.h | 4 ++ > examples/ip_pipeline/cli.c | 74 > - > examples/ip_pipeline/cryptodev.c| 2 - > examples/ip_pipeline/pipeline.h | 6 ++ > examples/l2fwd-crypto/main.c| 40 + > lib/librte_cryptodev/rte_crypto_sym.h | 12 ++-- > 19 files changed, 190 insertions(+), 154 deletions(-) > > > >>> I believe this patch should be applied to master. > >>> This has changes for net and virtio pmds. > >> It's all related to crypto code, you can take it in the crypto tree. > > Ok.. > > Applied to dpdk-next-crypto > This patch is taken back. As the changes introduce API breakage which needs > 1 release advance deprecation notice. > > > > > Thanks. > >>> for librte_cryptodev, and l2fwd-crypto > >>> > >>> Acked-by: Akhil Goyal > >>
[dpdk-dev] [PATCH] crypto/qat: set message field to zero in sym SGL case
Signed-off-by: Gosia Bakota
---
drivers/crypto/qat/qat_sym.c |2 ++
1 files changed, 2 insertions(+), 0 deletions(-)
mode change 100644 => 100755 drivers/crypto/qat/qat_sym.c
diff --git a/drivers/crypto/qat/qat_sym.c b/drivers/crypto/qat/qat_sym.c
old mode 100644
new mode 100755
index 8801ca5..7515a55
--- a/drivers/crypto/qat/qat_sym.c
+++ b/drivers/crypto/qat/qat_sym.c
@@ -530,6 +530,8 @@
qat_req->comn_mid.dest_data_addr =
cookie->qat_sgl_dst_phys_addr;
}
+ qat_req->comn_mid.src_length = 0;
+ qat_req->comn_mid.dst_length = 0;
} else {
qat_req->comn_mid.src_data_addr = src_buf_start;
qat_req->comn_mid.dest_data_addr = dst_buf_start;
--
1.7.0.7
[dpdk-dev] [PATCH 1/2] net/mlx5: fix testpmd crash on null-flow drv opts
mlx5 implements mlx5_flow_null_drv_ops to be used when a specific
flow typei/driver is not available or invalid.
This routines return error without modifying the rte_flow_error
parameter passed to them which causes testpmd, for example, to crash.
This commit addresses the issue by modifying the rte_flow_error
parameter in theses routines.
Fixes: 0c76d1c9a18d ("net/mlx5: add abstraction for multiple flow drivers")
Fixes: 684dafe795d0 ("net/mlx5: add flow query abstraction interface")
Cc: sta...@dpdk.org
Signed-off-by: Moti Haimovsky
---
drivers/net/mlx5/mlx5_flow.c | 29 +++--
1 file changed, 15 insertions(+), 14 deletions(-)
diff --git a/drivers/net/mlx5/mlx5_flow.c b/drivers/net/mlx5/mlx5_flow.c
index 9887018..e5a8e33 100644
--- a/drivers/net/mlx5/mlx5_flow.c
+++ b/drivers/net/mlx5/mlx5_flow.c
@@ -1694,19 +1694,20 @@ uint32_t mlx5_flow_adjust_priority(struct rte_eth_dev
*dev, int32_t priority,
const struct rte_flow_attr *attr __rte_unused,
const struct rte_flow_item items[] __rte_unused,
const struct rte_flow_action actions[] __rte_unused,
- struct rte_flow_error *error __rte_unused)
+ struct rte_flow_error *error)
{
- rte_errno = ENOTSUP;
- return -rte_errno;
+ return rte_flow_error_set(error, ENOTSUP,
+ RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL, NULL);
}
static struct mlx5_flow *
flow_null_prepare(const struct rte_flow_attr *attr __rte_unused,
const struct rte_flow_item items[] __rte_unused,
const struct rte_flow_action actions[] __rte_unused,
- struct rte_flow_error *error __rte_unused)
+ struct rte_flow_error *error)
{
- rte_errno = ENOTSUP;
+ rte_flow_error_set(error, ENOTSUP,
+ RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL, NULL);
return NULL;
}
@@ -1716,19 +1717,19 @@ uint32_t mlx5_flow_adjust_priority(struct rte_eth_dev
*dev, int32_t priority,
const struct rte_flow_attr *attr __rte_unused,
const struct rte_flow_item items[] __rte_unused,
const struct rte_flow_action actions[] __rte_unused,
- struct rte_flow_error *error __rte_unused)
+ struct rte_flow_error *error)
{
- rte_errno = ENOTSUP;
- return -rte_errno;
+ return rte_flow_error_set(error, ENOTSUP,
+ RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL, NULL);
}
static int
flow_null_apply(struct rte_eth_dev *dev __rte_unused,
struct rte_flow *flow __rte_unused,
- struct rte_flow_error *error __rte_unused)
+ struct rte_flow_error *error)
{
- rte_errno = ENOTSUP;
- return -rte_errno;
+ return rte_flow_error_set(error, ENOTSUP,
+ RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL, NULL);
}
static void
@@ -1748,10 +1749,10 @@ uint32_t mlx5_flow_adjust_priority(struct rte_eth_dev
*dev, int32_t priority,
struct rte_flow *flow __rte_unused,
const struct rte_flow_action *actions __rte_unused,
void *data __rte_unused,
- struct rte_flow_error *error __rte_unused)
+ struct rte_flow_error *error)
{
- rte_errno = ENOTSUP;
- return -rte_errno;
+ return rte_flow_error_set(error, ENOTSUP,
+ RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL, NULL);
}
/* Void driver to protect from null pointer reference. */
--
1.8.3.1
[dpdk-dev] [PATCH 0/2] net/mlx5: remove TCF support from PMD
Today, it is possible to offload an interface flow rules to the hardware using DPDK flow commands. With mlx5 it is also possible to offload a limited set of flow rules to the mlxsw (or E-switch) using the same DPDK flow commands. A 'transfer' attribute was added to the flow rule creation command in order to distinguish between configuring port flows and E-switch flows. The commands destined for the E-switch are transposed to TC-flower rules and are send, as Netlink messages, to the mlx5 driver, or more precisely to the netdev which represent the mlxsw port. With the addition to the PMD of E-switch configuration via DR (direct verbs rules) it is now possible to configure the E-switch using these commands instead of using TC-Flower messages. Doing so will allow us to remove the TCF support and the dependency of libmnl from the PMD. The purpose of this RFC is to propose configuring the E-switch flow filtering using DR, to remove the TCF support from the PMD and to remove the dependency of the PMD in libmnl. As for today VLAN insertion or removal is not supported in DR, this support will be added in separate commits. Moti Haimovsky (2): net/mlx5: fix testpmd crash on null-flow drv opts net/mlx5: remove TCF support from PMD doc/build-sdk-meson.txt |2 +- doc/guides/nics/mlx5.rst | 19 - doc/guides/platform/bluefield.rst |4 - drivers/net/mlx5/Makefile | 303 -- drivers/net/mlx5/meson.build | 123 +- drivers/net/mlx5/mlx5.c | 32 - drivers/net/mlx5/mlx5.h |3 - drivers/net/mlx5/mlx5_flow.c | 43 +- drivers/net/mlx5/mlx5_flow.h | 25 - drivers/net/mlx5/mlx5_flow_tcf.c | 6382 - mk/rte.app.mk |2 +- 11 files changed, 24 insertions(+), 6914 deletions(-) delete mode 100644 drivers/net/mlx5/mlx5_flow_tcf.c -- 1.8.3.1
Re: [dpdk-dev] [PATCH] crypto/qat: set message field to zero in sym SGL case
> -Original Message- > From: Bakota, Gosia > Sent: Tuesday, May 28, 2019 11:07 AM > To: dev@dpdk.org > Cc: akhil.go...@nxp.com; Trahe, Fiona ; Bakota, Gosia > > Subject: [PATCH] crypto/qat: set message field to zero in sym SGL case > > Signed-off-by: Gosia Bakota Nack Checkpatch issue - will send a v2
[dpdk-dev] [PATCH v2 1/5] net/nfp: disable nfp for 32-bit meson builds
>From comment in the build config for the make build-system:
"NFP PMD is not supported on 32-bit"
Therefore disable NFP for 32-bit meson builds as well as for 32-bit
make ones.
Signed-off-by: Bruce Richardson
Acked-by: Luca Boccassi
---
drivers/net/nfp/meson.build | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/net/nfp/meson.build b/drivers/net/nfp/meson.build
index ea51ac3..8c87c5b 100644
--- a/drivers/net/nfp/meson.build
+++ b/drivers/net/nfp/meson.build
@@ -1,7 +1,7 @@
# SPDX-License-Identifier: BSD-3-Clause
# Copyright(c) 2018 Intel Corporation
-if not is_linux
+if not is_linux or not dpdk_conf.get('RTE_ARCH_64')
build = false
endif
sources = files('nfpcore/nfp_cpp_pcie_ops.c',
--
2.7.4
[dpdk-dev] [PATCH v2 2/5] build: enable large file support on 32-bit
By default on 32-bit systems, file offsets are given as 32-bit values
which prevents support for large files. While this is unlikely to be
a problem, enabling large file support globally makes "make" and
"meson" builds consistent, since meson always enables large file
support, and without this change, the size of "struct stat" fields
will be different between the two builds.
The only location where this appears to be significant is in the
dpaax common code, where a printf needs to be updated for 32-bit
builds.
Signed-off-by: Bruce Richardson
Acked-by: Luca Boccassi
---
drivers/common/dpaax/dpaax_iova_table.c | 4 ++--
mk/arch/arm/rte.vars.mk | 2 +-
mk/arch/i686/rte.vars.mk| 2 +-
3 files changed, 4 insertions(+), 4 deletions(-)
diff --git a/drivers/common/dpaax/dpaax_iova_table.c
b/drivers/common/dpaax/dpaax_iova_table.c
index 2dd38a9..ae0af09 100644
--- a/drivers/common/dpaax/dpaax_iova_table.c
+++ b/drivers/common/dpaax/dpaax_iova_table.c
@@ -99,7 +99,7 @@ read_memory_node(unsigned int *count)
goto cleanup;
}
- DPAAX_DEBUG("Size of device-tree mem node: %lu", statbuf.st_size);
+ DPAAX_DEBUG("Size of device-tree mem node: %" PRIu64, statbuf.st_size);
if (statbuf.st_size > MEM_NODE_FILE_LEN) {
DPAAX_DEBUG("More memory nodes available than assumed.");
DPAAX_DEBUG("System may not work properly!");
@@ -118,7 +118,7 @@ read_memory_node(unsigned int *count)
*/
*count = (statbuf.st_size / 16);
if ((*count) <= 0 || (statbuf.st_size % 16 != 0)) {
- DPAAX_DEBUG("Invalid memory node values or count. (size=%lu)",
+ DPAAX_DEBUG("Invalid memory node values or count. (size=%"
PRIu64 ")",
statbuf.st_size);
goto cleanup;
}
diff --git a/mk/arch/arm/rte.vars.mk b/mk/arch/arm/rte.vars.mk
index 27b1147..dc8c10a 100644
--- a/mk/arch/arm/rte.vars.mk
+++ b/mk/arch/arm/rte.vars.mk
@@ -4,7 +4,7 @@
ARCH ?= arm
CROSS ?=
-CPU_CFLAGS ?= -marm -munaligned-access
+CPU_CFLAGS ?= -marm -munaligned-access -D_FILE_OFFSET_BITS=64
CPU_LDFLAGS ?=
CPU_ASFLAGS ?= -felf
diff --git a/mk/arch/i686/rte.vars.mk b/mk/arch/i686/rte.vars.mk
index 3f48f67..c867883 100644
--- a/mk/arch/i686/rte.vars.mk
+++ b/mk/arch/i686/rte.vars.mk
@@ -24,7 +24,7 @@ ARCH ?= i386
ARCH_DIR := x86
CROSS ?=
-CPU_CFLAGS ?= -m32
+CPU_CFLAGS ?= -m32 -D_FILE_OFFSET_BITS=64
CPU_LDFLAGS ?= -melf_i386
CPU_ASFLAGS ?= -felf
--
2.7.4
[dpdk-dev] [PATCH v2 0/5] fix 32-bit meson builds
This set fixes some issues seen on the automated CI system with building on 32-bit Linux distro's using meson and ninja. The fixes are to disable unsupported parts of the build, and switch the 32-bit builds to always having large file support, so that make and meson are consistent in that regard. V2: add two additional patches to fix clang builds. This allows the test-meson-builds.sh script to run successfully on 32-bit Ubuntu 16.04 systems. Bruce Richardson (5): net/nfp: disable nfp for 32-bit meson builds build: enable large file support on 32-bit build: remove unnecessary large file support defines eal: mark unused function in 32-bit builds build: add libatomic dependency for 32-bit clang compile drivers/common/dpaax/dpaax_iova_table.c | 4 ++-- drivers/event/octeontx/meson.build | 5 + drivers/event/opdl/meson.build | 5 + drivers/net/nfp/meson.build | 2 +- examples/vhost_crypto/meson.build | 1 - examples/vhost_scsi/meson.build | 1 - lib/librte_eal/linux/eal/eal_memalloc.c | 1 - lib/librte_eal/linux/eal/eal_memory.c | 2 +- lib/librte_rcu/meson.build | 5 + lib/librte_vhost/Makefile | 2 +- mk/arch/arm/rte.vars.mk | 2 +- mk/arch/i686/rte.vars.mk| 2 +- 12 files changed, 22 insertions(+), 10 deletions(-) -- 2.7.4
[dpdk-dev] [PATCH v2 3/5] build: remove unnecessary large file support defines
Since we now always use _FILE_OFFSET_BITS=64 flag when building
DPDK, we can remove the Makefile and C-file #defines setting it
individually for parts of the build.
Signed-off-by: Bruce Richardson
Acked-by: Luca Boccassi
---
examples/vhost_crypto/meson.build | 1 -
examples/vhost_scsi/meson.build | 1 -
lib/librte_eal/linux/eal/eal_memalloc.c | 1 -
lib/librte_eal/linux/eal/eal_memory.c | 1 -
lib/librte_vhost/Makefile | 2 +-
5 files changed, 1 insertion(+), 5 deletions(-)
diff --git a/examples/vhost_crypto/meson.build
b/examples/vhost_crypto/meson.build
index 8e9860f..2485f3b 100644
--- a/examples/vhost_crypto/meson.build
+++ b/examples/vhost_crypto/meson.build
@@ -9,7 +9,6 @@
build = dpdk_conf.has('RTE_LIBRTE_VHOST')
allow_experimental_apis = true
deps += ['vhost', 'cryptodev']
-cflags += ['-D_FILE_OFFSET_BITS=64']
sources = files(
'main.c'
)
diff --git a/examples/vhost_scsi/meson.build b/examples/vhost_scsi/meson.build
index 2972e4d..77e5201 100644
--- a/examples/vhost_scsi/meson.build
+++ b/examples/vhost_scsi/meson.build
@@ -15,7 +15,6 @@ if not cc.has_header('linux/virtio_scsi.h')
endif
deps += 'vhost'
-cflags += ['-D_FILE_OFFSET_BITS=64']
sources = files(
'scsi.c', 'vhost_scsi.c'
)
diff --git a/lib/librte_eal/linux/eal/eal_memalloc.c
b/lib/librte_eal/linux/eal/eal_memalloc.c
index 1e9ebb8..b1849a2 100644
--- a/lib/librte_eal/linux/eal/eal_memalloc.c
+++ b/lib/librte_eal/linux/eal/eal_memalloc.c
@@ -2,7 +2,6 @@
* Copyright(c) 2017-2018 Intel Corporation
*/
-#define _FILE_OFFSET_BITS 64
#include
#include
#include
diff --git a/lib/librte_eal/linux/eal/eal_memory.c
b/lib/librte_eal/linux/eal/eal_memory.c
index 416dad8..40a0f68 100644
--- a/lib/librte_eal/linux/eal/eal_memory.c
+++ b/lib/librte_eal/linux/eal/eal_memory.c
@@ -3,7 +3,6 @@
* Copyright(c) 2013 6WIND S.A.
*/
-#define _FILE_OFFSET_BITS 64
#include
#include
#include
diff --git a/lib/librte_vhost/Makefile b/lib/librte_vhost/Makefile
index 5dd3189..8623e91 100644
--- a/lib/librte_vhost/Makefile
+++ b/lib/librte_vhost/Makefile
@@ -11,7 +11,7 @@ EXPORT_MAP := rte_vhost_version.map
LIBABIVER := 4
CFLAGS += -DALLOW_EXPERIMENTAL_API
-CFLAGS += $(WERROR_FLAGS) -I$(SRCDIR) -O3 -D_FILE_OFFSET_BITS=64
+CFLAGS += $(WERROR_FLAGS) -I$(SRCDIR) -O3
CFLAGS += -I vhost_user
CFLAGS += -fno-strict-aliasing
LDLIBS += -lpthread
--
2.7.4
[dpdk-dev] [PATCH v2 4/5] eal: mark unused function in 32-bit builds
The get_socket_mem_size() function is only used in 64-bit builds,
causing clang to warn about it for 32-bit builds. Add the __rte_unused
attribute to the function to silence the warning.
Signed-off-by: Bruce Richardson
---
lib/librte_eal/linux/eal/eal_memory.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/lib/librte_eal/linux/eal/eal_memory.c
b/lib/librte_eal/linux/eal/eal_memory.c
index 40a0f68..1853ace 100644
--- a/lib/librte_eal/linux/eal/eal_memory.c
+++ b/lib/librte_eal/linux/eal/eal_memory.c
@@ -1088,6 +1088,7 @@ remap_needed_hugepages(struct hugepage_file *hugepages,
int n_pages)
return 0;
}
+__rte_unused /* function is unused on 32-bit builds */
static inline uint64_t
get_socket_mem_size(int socket)
{
--
2.7.4
[dpdk-dev] [PATCH v2 5/5] build: add libatomic dependency for 32-bit clang compile
When compiling with clang on 32-bit platforms, we are missing copies
of 64-bit atomic functions. We can solve this by linking against
libatomic for the drivers and libs which need those atomic ops.
Signed-off-by: Bruce Richardson
---
drivers/event/octeontx/meson.build | 5 +
drivers/event/opdl/meson.build | 5 +
lib/librte_rcu/meson.build | 5 +
3 files changed, 15 insertions(+)
diff --git a/drivers/event/octeontx/meson.build
b/drivers/event/octeontx/meson.build
index 2b74bb6..3f83be6 100644
--- a/drivers/event/octeontx/meson.build
+++ b/drivers/event/octeontx/meson.build
@@ -11,3 +11,8 @@ sources = files('ssovf_worker.c',
)
deps += ['common_octeontx', 'mempool_octeontx', 'bus_vdev', 'pmd_octeontx']
+
+# for clang 32-bit compiles we need libatomic for 64-bit atomic ops
+if cc.get_id() == 'clang' and dpdk_conf.get('RTE_ARCH_64') == false
+ ext_deps += cc.find_library('atomic')
+endif
diff --git a/drivers/event/opdl/meson.build b/drivers/event/opdl/meson.build
index cc6029c..e1cfb2d 100644
--- a/drivers/event/opdl/meson.build
+++ b/drivers/event/opdl/meson.build
@@ -9,3 +9,8 @@ sources = files(
'opdl_test.c',
)
deps += ['bus_vdev']
+
+# for clang 32-bit compiles we need libatomic for 64-bit atomic ops
+if cc.get_id() == 'clang' and dpdk_conf.get('RTE_ARCH_64') == false
+ ext_deps += cc.find_library('atomic')
+endif
diff --git a/lib/librte_rcu/meson.build b/lib/librte_rcu/meson.build
index 0c2d5a2..43edb87 100644
--- a/lib/librte_rcu/meson.build
+++ b/lib/librte_rcu/meson.build
@@ -5,3 +5,8 @@ allow_experimental_apis = true
sources = files('rte_rcu_qsbr.c')
headers = files('rte_rcu_qsbr.h')
+
+# for clang 32-bit compiles we need libatomic for 64-bit atomic ops
+if cc.get_id() == 'clang' and dpdk_conf.get('RTE_ARCH_64') == false
+ ext_deps += cc.find_library('atomic')
+endif
--
2.7.4
[dpdk-dev] [PATCH 0/2] net/mlx5: remove TCF support from PMD
Today, it is possible to offload an interface flow rules to the hardware using DPDK flow commands. With mlx5 it is also possible to offload a limited set of flow rules to the mlxsw (or E-switch) using the same DPDK flow commands. A 'transfer' attribute was added to the flow rule creation command in order to distinguish between configuring port flows and E-switch flows. The commands destined for the E-switch are transposed to TC-flower rules and are send, as Netlink messages, to the mlx5 driver, or more precisely to the netdev which represent the mlxsw port. With the addition to the PMD of E-switch configuration via DR (direct verbs rules) it is now possible to configure the E-switch using these commands instead of using TC-Flower messages. Doing so will allow us to remove the TCF support and the dependency of libmnl from the PMD. The purpose of this RFC is to propose configuring the E-switch flow filtering using DR, to remove the TCF support from the PMD and to remove the dependency of the PMD in libmnl. As for today VLAN insertion or removal is not supported in DR, this support will be added in separate commits. Moti Haimovsky (2): net/mlx5: fix testpmd crash on null-flow drv opts net/mlx5: remove TCF support from PMD doc/build-sdk-meson.txt |2 +- doc/guides/nics/mlx5.rst | 19 - doc/guides/platform/bluefield.rst |4 - drivers/net/mlx5/Makefile | 303 -- drivers/net/mlx5/meson.build | 123 +- drivers/net/mlx5/mlx5.c | 32 - drivers/net/mlx5/mlx5.h |3 - drivers/net/mlx5/mlx5_flow.c | 43 +- drivers/net/mlx5/mlx5_flow.h | 25 - drivers/net/mlx5/mlx5_flow_tcf.c | 6382 - mk/rte.app.mk |2 +- 11 files changed, 24 insertions(+), 6914 deletions(-) delete mode 100644 drivers/net/mlx5/mlx5_flow_tcf.c -- 1.8.3.1
[dpdk-dev] [PATCH 1/2] net/mlx5: fix testpmd crash on null-flow drv opts
mlx5 implements mlx5_flow_null_drv_ops to be used when a specific
flow typei/driver is not available or invalid.
This routines return error without modifying the rte_flow_error
parameter passed to them which causes testpmd, for example, to crash.
This commit addresses the issue by modifying the rte_flow_error
parameter in theses routines.
Fixes: 0c76d1c9a18d ("net/mlx5: add abstraction for multiple flow drivers")
Fixes: 684dafe795d0 ("net/mlx5: add flow query abstraction interface")
Cc: sta...@dpdk.org
Signed-off-by: Moti Haimovsky
---
drivers/net/mlx5/mlx5_flow.c | 29 +++--
1 file changed, 15 insertions(+), 14 deletions(-)
diff --git a/drivers/net/mlx5/mlx5_flow.c b/drivers/net/mlx5/mlx5_flow.c
index 9887018..e5a8e33 100644
--- a/drivers/net/mlx5/mlx5_flow.c
+++ b/drivers/net/mlx5/mlx5_flow.c
@@ -1694,19 +1694,20 @@ uint32_t mlx5_flow_adjust_priority(struct rte_eth_dev
*dev, int32_t priority,
const struct rte_flow_attr *attr __rte_unused,
const struct rte_flow_item items[] __rte_unused,
const struct rte_flow_action actions[] __rte_unused,
- struct rte_flow_error *error __rte_unused)
+ struct rte_flow_error *error)
{
- rte_errno = ENOTSUP;
- return -rte_errno;
+ return rte_flow_error_set(error, ENOTSUP,
+ RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL, NULL);
}
static struct mlx5_flow *
flow_null_prepare(const struct rte_flow_attr *attr __rte_unused,
const struct rte_flow_item items[] __rte_unused,
const struct rte_flow_action actions[] __rte_unused,
- struct rte_flow_error *error __rte_unused)
+ struct rte_flow_error *error)
{
- rte_errno = ENOTSUP;
+ rte_flow_error_set(error, ENOTSUP,
+ RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL, NULL);
return NULL;
}
@@ -1716,19 +1717,19 @@ uint32_t mlx5_flow_adjust_priority(struct rte_eth_dev
*dev, int32_t priority,
const struct rte_flow_attr *attr __rte_unused,
const struct rte_flow_item items[] __rte_unused,
const struct rte_flow_action actions[] __rte_unused,
- struct rte_flow_error *error __rte_unused)
+ struct rte_flow_error *error)
{
- rte_errno = ENOTSUP;
- return -rte_errno;
+ return rte_flow_error_set(error, ENOTSUP,
+ RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL, NULL);
}
static int
flow_null_apply(struct rte_eth_dev *dev __rte_unused,
struct rte_flow *flow __rte_unused,
- struct rte_flow_error *error __rte_unused)
+ struct rte_flow_error *error)
{
- rte_errno = ENOTSUP;
- return -rte_errno;
+ return rte_flow_error_set(error, ENOTSUP,
+ RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL, NULL);
}
static void
@@ -1748,10 +1749,10 @@ uint32_t mlx5_flow_adjust_priority(struct rte_eth_dev
*dev, int32_t priority,
struct rte_flow *flow __rte_unused,
const struct rte_flow_action *actions __rte_unused,
void *data __rte_unused,
- struct rte_flow_error *error __rte_unused)
+ struct rte_flow_error *error)
{
- rte_errno = ENOTSUP;
- return -rte_errno;
+ return rte_flow_error_set(error, ENOTSUP,
+ RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL, NULL);
}
/* Void driver to protect from null pointer reference. */
--
1.8.3.1
Re: [dpdk-dev] [PATCH v2 03/10] net/bnxt: implement vector mode driver
On Tue, May 28, 2019 at 5:08 AM Ferruh Yigit wrote:
>
> On 5/24/2019 3:49 PM, Lance Richardson wrote:
> > @@ -1597,6 +1659,8 @@
> >
> > bnxt_txq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
> >
> > static int bnxt_mtu_set_op(struct rte_eth_dev *eth_dev, uint16_t new_mtu)
> > {
> > + uint32_t new_pkt_size = new_mtu + ETHER_HDR_LEN + ETHER_CRC_LEN +
> > + VLAN_TAG_SIZE * BNXT_NUM_VLANS;
>
> 'ETHER_HDR_LEN' & 'ETHER_CRC_LEN' are renamed to
> 'RTE_ETHER_HDR_LEN' & 'RTE_ETHER_CRC_LEN'
>
> I am mostly fixing these while applying but since there will be a new version,
> can you please rebase on top of latest next-net, to be able to catch these.
Sure. Will rebase on next-net and resend after addressing feedback.
Thanks,
Lance
Re: [dpdk-dev] [PATCH v2 4/5] eal: mark unused function in 32-bit builds
On Tue, 2019-05-28 at 12:07 +0100, Bruce Richardson wrote:
> The get_socket_mem_size() function is only used in 64-bit builds,
> causing clang to warn about it for 32-bit builds. Add the
> __rte_unused
> attribute to the function to silence the warning.
>
> Signed-off-by: Bruce Richardson <
> bruce.richard...@intel.com
> >
> ---
> lib/librte_eal/linux/eal/eal_memory.c | 1 +
> 1 file changed, 1 insertion(+)
>
> diff --git a/lib/librte_eal/linux/eal/eal_memory.c
> b/lib/librte_eal/linux/eal/eal_memory.c
> index 40a0f68..1853ace 100644
> --- a/lib/librte_eal/linux/eal/eal_memory.c
> +++ b/lib/librte_eal/linux/eal/eal_memory.c
> @@ -1088,6 +1088,7 @@ remap_needed_hugepages(struct hugepage_file
> *hugepages, int n_pages)
> return 0;
> }
>
> +__rte_unused /* function is unused on 32-bit builds */
> static inline uint64_t
> get_socket_mem_size(int socket)
> {
Acked-by: Luca Boccassi
--
Kind regards,
Luca Boccassi
Re: [dpdk-dev] [PATCH v2 5/5] build: add libatomic dependency for 32-bit clang compile
On Tue, 2019-05-28 at 12:07 +0100, Bruce Richardson wrote:
> When compiling with clang on 32-bit platforms, we are missing copies
> of 64-bit atomic functions. We can solve this by linking against
> libatomic for the drivers and libs which need those atomic ops.
>
> Signed-off-by: Bruce Richardson <
> bruce.richard...@intel.com
> >
> ---
> drivers/event/octeontx/meson.build | 5 +
> drivers/event/opdl/meson.build | 5 +
> lib/librte_rcu/meson.build | 5 +
> 3 files changed, 15 insertions(+)
>
> diff --git a/drivers/event/octeontx/meson.build
> b/drivers/event/octeontx/meson.build
> index 2b74bb6..3f83be6 100644
> --- a/drivers/event/octeontx/meson.build
> +++ b/drivers/event/octeontx/meson.build
> @@ -11,3 +11,8 @@ sources = files('ssovf_worker.c',
> )
>
> deps += ['common_octeontx', 'mempool_octeontx', 'bus_vdev',
> 'pmd_octeontx']
> +
> +# for clang 32-bit compiles we need libatomic for 64-bit atomic ops
> +if cc.get_id() == 'clang' and dpdk_conf.get('RTE_ARCH_64') == false
> + ext_deps += cc.find_library('atomic')
> +endif
> diff --git a/drivers/event/opdl/meson.build
> b/drivers/event/opdl/meson.build
> index cc6029c..e1cfb2d 100644
> --- a/drivers/event/opdl/meson.build
> +++ b/drivers/event/opdl/meson.build
> @@ -9,3 +9,8 @@ sources = files(
> 'opdl_test.c',
> )
> deps += ['bus_vdev']
> +
> +# for clang 32-bit compiles we need libatomic for 64-bit atomic ops
> +if cc.get_id() == 'clang' and dpdk_conf.get('RTE_ARCH_64') == false
> + ext_deps += cc.find_library('atomic')
> +endif
> diff --git a/lib/librte_rcu/meson.build b/lib/librte_rcu/meson.build
> index 0c2d5a2..43edb87 100644
> --- a/lib/librte_rcu/meson.build
> +++ b/lib/librte_rcu/meson.build
> @@ -5,3 +5,8 @@ allow_experimental_apis = true
>
> sources = files('rte_rcu_qsbr.c')
> headers = files('rte_rcu_qsbr.h')
> +
> +# for clang 32-bit compiles we need libatomic for 64-bit atomic ops
> +if cc.get_id() == 'clang' and dpdk_conf.get('RTE_ARCH_64') == false
> + ext_deps += cc.find_library('atomic')
> +endif
Acked-by: Luca Boccassi
--
Kind regards,
Luca Boccassi
[dpdk-dev] [PATCH 00/27] sched: feature enhancements
This patchset refactors the dpdk qos sched library to add following features to enhance the scheduler functionality. 1. flexibile configuration of the pipe traffic classes and queues; Currently, each pipe has 16 queues hardwired into 4 TCs scheduled with strict priority, and each TC has exactly with 4 queues that are scheduled with Weighted Fair Queuing (WFQ). Instead of hardwiring queues to traffic class within the specific pipe, the new implementation allows more flexible/configurable split of pipe queues between strict priority (SP) and best-effort (BE) traffic classes along with the support of more number of traffic classes i.e. max 16. All the high priority TCs (TC1, TC2, ...) have exactly 1 queue, while the lowest priority BE TC, has 1, 4 or 8 queues. This is justified by the fact that all the high priority TCs are fully provisioned (small to medium traffic rates), while most of the traffic fits into the BE class, which is typically oversubscribed. Furthermore, this change allows to use less than 16 queues per pipe when not all the 16 queues are needed. Therefore, no memory will be allocated to the queues that are not needed. 2. Subport level configuration of pipe nodes; Currently, all parameters for the pipe nodes (subscribers) configuration are part of the port level structure which forces all groups of subscribers (i.e. pipes) in different subports to have similar configurations in terms of their number, queue sizes, traffic-classes, etc. The new implementation moves pipe nodes configuration parameters from port level to subport level structure. Therefore, different subports of the same port can have different configuration for the pipe nodes (subscribers), for examples- number of pipes, queue sizes, queues to traffic-class mapping, etc. Jasvinder Singh (27): sched: update macros for flexible config sched: update subport and pipe data structures sched: update internal data structures sched: update port config api sched: update port free api sched: update subport config api sched: update pipe profile add api sched: update pipe config api sched: update pkt read and write api sched: update subport and tc queue stats sched: update port memory footprint api sched: update packet enqueue api sched: update grinder pipe and tc cache sched: update grinder next pipe and tc functions sched: update pipe and tc queues prefetch sched: update grinder wrr compute function sched: modify credits update function sched: update mbuf prefetch function sched: update grinder schedule function sched: update grinder handle function sched: update packet dequeue api sched: update sched queue stats api test/sched: update unit test net/softnic: update softnic tm function examples/qos_sched: update qos sched sample app examples/ip_pipeline: update ip pipeline sample app sched: code cleanup app/test/test_sched.c | 39 +- doc/guides/rel_notes/deprecation.rst |6 - doc/guides/rel_notes/release_19_08.rst|7 +- drivers/net/softnic/rte_eth_softnic.c | 131 ++ drivers/net/softnic/rte_eth_softnic_cli.c | 285 ++- .../net/softnic/rte_eth_softnic_internals.h |4 +- drivers/net/softnic/rte_eth_softnic_tm.c | 89 +- examples/ip_pipeline/cli.c| 85 +- examples/ip_pipeline/tmgr.c | 21 +- examples/ip_pipeline/tmgr.h |3 - examples/qos_sched/app_thread.c | 11 +- examples/qos_sched/cfg_file.c | 282 ++- examples/qos_sched/init.c | 110 +- examples/qos_sched/main.h |6 +- examples/qos_sched/profile.cfg| 59 +- examples/qos_sched/profile_ov.cfg | 47 +- examples/qos_sched/stats.c| 152 +- lib/librte_pipeline/rte_table_action.c|1 - lib/librte_pipeline/rte_table_action.h|4 +- lib/librte_sched/Makefile |2 +- lib/librte_sched/meson.build |2 +- lib/librte_sched/rte_sched.c | 2015 ++--- lib/librte_sched/rte_sched.h | 147 +- lib/librte_sched/rte_sched_common.h | 33 + 24 files changed, 2314 insertions(+), 1227 deletions(-) -- 2.20.1
[dpdk-dev] [PATCH 01/27] sched: update macros for flexible config
From: Jasvinder Singh
Update existing macros and add new one for best-effort traffic class
queues to allow configuration flexiblity for pipe traffic classes and
queues, and subport level configuration of the pipe parameters.
Signed-off-by: Jasvinder Singh
Signed-off-by: Abraham Tovar
Signed-off-by: Lukasz Krakowiak
---
doc/guides/rel_notes/deprecation.rst | 6 -
doc/guides/rel_notes/release_19_08.rst | 7 -
lib/librte_sched/Makefile | 2 +-
lib/librte_sched/meson.build | 2 +-
lib/librte_sched/rte_sched.h | 37 +++---
5 files changed, 35 insertions(+), 19 deletions(-)
diff --git a/doc/guides/rel_notes/deprecation.rst
b/doc/guides/rel_notes/deprecation.rst
index 098d24381..a408270f5 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -99,12 +99,6 @@ Deprecation Notices
to one it means it represents IV, when is set to zero it means J0 is used
directly, in this case 16 bytes of J0 need to be passed.
-* sched: To allow more traffic classes, flexible mapping of pipe queues to
- traffic classes, and subport level configuration of pipes and queues
- changes will be made to macros, data structures and API functions defined
- in "rte_sched.h". These changes are aligned to improvements suggested in the
- RFC https://mails.dpdk.org/archives/dev/2018-November/120035.html.
-
* metrics: The function ``rte_metrics_init`` will have a non-void return
in order to notify errors instead of calling ``rte_exit``.
diff --git a/doc/guides/rel_notes/release_19_08.rst
b/doc/guides/rel_notes/release_19_08.rst
index b9510f93a..210f32e7f 100644
--- a/doc/guides/rel_notes/release_19_08.rst
+++ b/doc/guides/rel_notes/release_19_08.rst
@@ -83,6 +83,11 @@ API Changes
Also, make sure to start the actual text at the margin.
=
+* sched: To allow more traffic classes, flexible mapping of pipe queues to
+ traffic classes, and subport level configuration of pipes and queues
+ changes are made to public macros, data structures and API functions defined
+ in "rte_sched.h".
+
ABI Changes
---
@@ -170,7 +175,7 @@ The libraries prepended with a plus sign were incremented
in this version.
librte_rcu.so.1
librte_reorder.so.1
librte_ring.so.2
- librte_sched.so.2
+ + librte_sched.so.3
librte_security.so.2
librte_stack.so.1
librte_table.so.3
diff --git a/lib/librte_sched/Makefile b/lib/librte_sched/Makefile
index 644fd9d15..3d7f410e1 100644
--- a/lib/librte_sched/Makefile
+++ b/lib/librte_sched/Makefile
@@ -18,7 +18,7 @@ LDLIBS += -lrte_timer
EXPORT_MAP := rte_sched_version.map
-LIBABIVER := 2
+LIBABIVER := 3
#
# all source are stored in SRCS-y
diff --git a/lib/librte_sched/meson.build b/lib/librte_sched/meson.build
index 8e989e5f6..59d43c6d8 100644
--- a/lib/librte_sched/meson.build
+++ b/lib/librte_sched/meson.build
@@ -1,7 +1,7 @@
# SPDX-License-Identifier: BSD-3-Clause
# Copyright(c) 2017 Intel Corporation
-version = 2
+version = 3
sources = files('rte_sched.c', 'rte_red.c', 'rte_approx.c')
headers = files('rte_sched.h', 'rte_sched_common.h',
'rte_red.h', 'rte_approx.h')
diff --git a/lib/librte_sched/rte_sched.h b/lib/librte_sched/rte_sched.h
index 9c55a787d..cf7695f27 100644
--- a/lib/librte_sched/rte_sched.h
+++ b/lib/librte_sched/rte_sched.h
@@ -52,7 +52,7 @@ extern "C" {
* multiple connections of same traffic class belonging to
* the same user;
* - Weighted Round Robin (WRR) is used to service the
- * queues within same pipe traffic class.
+ * queues within same pipe lowest priority (best-effort) traffic class.
*
*/
@@ -66,26 +66,43 @@ extern "C" {
#include "rte_red.h"
#endif
-/** Number of traffic classes per pipe (as well as subport).
- * Cannot be changed.
+/** Maximum number of queues per pipe.
+ * Note that the multiple queues (power of 2) can only be assigned to
+ * lowest priority (best-effort) traffic class. Other higher priority traffic
+ * classes can only have one queue.
+ *
+ * Can not change.
+ */
+#define RTE_SCHED_QUEUES_PER_PIPE16
+
+/** Number of WRR queues for lowest priority (best-effort) traffic class per
+ * pipe.
*/
-#define RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE4
+#define RTE_SCHED_WRR_QUEUES_PER_PIPE8
-/** Number of queues per pipe traffic class. Cannot be changed. */
+/** Number of traffic classes per pipe (as well as subport). */
#define RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS4
+#define RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE\
+(RTE_SCHED_QUEUES_PER_PIPE - RTE_SCHED_WRR_QUEUES_PER_PIPE + 1)
-/** Number of queues per pipe. */
-#define RTE_SCHED_QUEUES_PER_PIPE \
- (RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE * \
- RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS)
+/** Maximum number of subports that can be defined per port.
+ * Compile-time configurable.
+ */
Re: [dpdk-dev] [PATCH 0/2] Add ABI Version Testing to app/test
On Tue, May 28, 2019 at 12:51:56PM +0100, Ray Kinsella wrote: > This patchset adds ABI Version Testing functionality to the app/test unit > test framework. > > The patchset is intended to address two issues previously raised during ML > conversations on ABI Stability; > 1. How do we unit test still supported previous ABI Versions. > 2. How to we unit test inline functions from still supported previous ABI > Versions. > > The more obvious way to achieve both of the above is to simply archive > pre-built binaries compiled against previous versions of DPDK for use unit > testing previous ABI Versions, and while this should still be done as an > additional check, this approach does not scale well, must every DPDK > developer have a local copy of these binaries to test with, before > upstreaming changes? > > Instead starting with rte_lpm, I did the following:- > > * I reproduced mostly unmodified unit tests from previous ABI Versions, > in this case v2.0 and v16.04 > * I reproduced the rte_lpm interface header from these previous ABI > Versions,including the inline functions and remapping symbols to > appropriate versions. > * I added support for multiple abi versions to the app/test unit test > framework to allow users to switch between abi versions (set_abi_version), > without further polluting the already long list of unit tests available in > app/test. > > The intention here is that, in future as developers need to depreciate > APIs, their associated unit tests may move into the ABI Version testing > mechanism of the app/test instead of being replaced by the latest set of > unit tests as would be the case today. > > ToDo: > * Refactor the v2.0 and v16.04 unit tests to separate functional and > performance test cases. > * Add support for trigger ABI Version unit tests from the app/test command > line. > While I admire the goal, given the amount of code that seems to be involved here, I'm not sure if the "test" binary is the place to put this. I think it might be better as a separate ABI compatiblity test app. A separate question is whether this is really necessary to ensure ABI compatibility? Do other projects do this? Is the info from the abi compatiblity checker script already in DPDK, or from other already-available tools not sufficient? /Bruce
[dpdk-dev] [PATCH 04/27] sched: update port config api
From: Jasvinder Singh
Update the scheduler port configuration api implementation to allow
configuration flexiblity for pipe traffic classes and queues, and
subport level configuration of the pipe parameters..
Signed-off-by: Jasvinder Singh
Signed-off-by: Abraham Tovar
Signed-off-by: Lukasz Krakowiak
---
lib/librte_sched/rte_sched.c | 189 +++
1 file changed, 15 insertions(+), 174 deletions(-)
diff --git a/lib/librte_sched/rte_sched.c b/lib/librte_sched/rte_sched.c
index 8256ac407..8f11b4597 100644
--- a/lib/librte_sched/rte_sched.c
+++ b/lib/librte_sched/rte_sched.c
@@ -388,8 +388,6 @@ pipe_profile_check(struct rte_sched_pipe_params *params,
static int
rte_sched_port_check_params(struct rte_sched_port_params *params)
{
- uint32_t i;
-
if (params == NULL)
return -1;
@@ -407,40 +405,10 @@ rte_sched_port_check_params(struct rte_sched_port_params
*params)
/* n_subports_per_port: non-zero, limited to 16 bits, power of 2 */
if (params->n_subports_per_port == 0 ||
- params->n_subports_per_port > 1u << 16 ||
+ params->n_subports_per_port > RTE_SCHED_SUBPORTS_PER_PORT ||
!rte_is_power_of_2(params->n_subports_per_port))
return -6;
- /* n_pipes_per_subport: non-zero, power of 2 */
- if (params->n_pipes_per_subport == 0 ||
- !rte_is_power_of_2(params->n_pipes_per_subport))
- return -7;
-
- /* qsize: non-zero, power of 2,
-* no bigger than 32K (due to 16-bit read/write pointers)
-*/
- for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i++) {
- uint16_t qsize = params->qsize[i];
-
- if (qsize == 0 || !rte_is_power_of_2(qsize))
- return -8;
- }
-
- /* pipe_profiles and n_pipe_profiles */
- if (params->pipe_profiles == NULL ||
- params->n_pipe_profiles == 0 ||
- params->n_pipe_profiles > RTE_SCHED_PIPE_PROFILES_PER_PORT)
- return -9;
-
- for (i = 0; i < params->n_pipe_profiles; i++) {
- struct rte_sched_pipe_params *p = params->pipe_profiles + i;
- int status;
-
- status = pipe_profile_check(p, params->rate);
- if (status != 0)
- return status;
- }
-
return 0;
}
@@ -524,36 +492,6 @@ rte_sched_port_get_memory_footprint(struct
rte_sched_port_params *params)
return size0 + size1;
}
-static void
-rte_sched_port_config_qsize(struct rte_sched_port *port)
-{
- /* TC 0 */
- port->qsize_add[0] = 0;
- port->qsize_add[1] = port->qsize_add[0] + port->qsize[0];
- port->qsize_add[2] = port->qsize_add[1] + port->qsize[0];
- port->qsize_add[3] = port->qsize_add[2] + port->qsize[0];
-
- /* TC 1 */
- port->qsize_add[4] = port->qsize_add[3] + port->qsize[0];
- port->qsize_add[5] = port->qsize_add[4] + port->qsize[1];
- port->qsize_add[6] = port->qsize_add[5] + port->qsize[1];
- port->qsize_add[7] = port->qsize_add[6] + port->qsize[1];
-
- /* TC 2 */
- port->qsize_add[8] = port->qsize_add[7] + port->qsize[1];
- port->qsize_add[9] = port->qsize_add[8] + port->qsize[2];
- port->qsize_add[10] = port->qsize_add[9] + port->qsize[2];
- port->qsize_add[11] = port->qsize_add[10] + port->qsize[2];
-
- /* TC 3 */
- port->qsize_add[12] = port->qsize_add[11] + port->qsize[2];
- port->qsize_add[13] = port->qsize_add[12] + port->qsize[3];
- port->qsize_add[14] = port->qsize_add[13] + port->qsize[3];
- port->qsize_add[15] = port->qsize_add[14] + port->qsize[3];
-
- port->qsize_sum = port->qsize_add[15] + port->qsize[3];
-}
-
static void
rte_sched_port_log_pipe_profile(struct rte_sched_port *port, uint32_t i)
{
@@ -660,84 +598,33 @@ rte_sched_pipe_profile_convert(struct
rte_sched_pipe_params *src,
}
}
-static void
-rte_sched_port_config_pipe_profile_table(struct rte_sched_port *port,
- struct rte_sched_port_params *params)
-{
- uint32_t i;
-
- for (i = 0; i < port->n_pipe_profiles; i++) {
- struct rte_sched_pipe_params *src = params->pipe_profiles + i;
- struct rte_sched_pipe_profile *dst = port->pipe_profiles + i;
-
- rte_sched_pipe_profile_convert(src, dst, params->rate);
- rte_sched_port_log_pipe_profile(port, i);
- }
-
- port->pipe_tc3_rate_max = 0;
- for (i = 0; i < port->n_pipe_profiles; i++) {
- struct rte_sched_pipe_params *src = params->pipe_profiles + i;
- uint32_t pipe_tc3_rate = src->tc_rate[3];
-
- if (port->pipe_tc3_rate_max < pipe_tc3_rate)
- port->pipe_tc3_rate_max = pipe_tc3_rate;
- }
-}
-
struct rte_sched_port *
rte_sched_port_config(struct rte_sched_port_params *params)
{
struct rte_sched_port *port = NULL;
-
[dpdk-dev] [PATCH 03/27] sched: update internal data structures
From: Jasvinder Singh
Update internal data structures of the scheduler to allow configuration
flexiblity for pipe traffic classes and queues, and subport level
configuration of the pipe parameters.
Signed-off-by: Jasvinder Singh
Signed-off-by: Abraham Tovar
Signed-off-by: Lukasz Krakowiak
---
lib/librte_sched/rte_sched.c | 182 +--
1 file changed, 130 insertions(+), 52 deletions(-)
diff --git a/lib/librte_sched/rte_sched.c b/lib/librte_sched/rte_sched.c
index a60ddf97e..8256ac407 100644
--- a/lib/librte_sched/rte_sched.c
+++ b/lib/librte_sched/rte_sched.c
@@ -37,6 +37,7 @@
#define RTE_SCHED_TB_RATE_CONFIG_ERR (1e-7)
#define RTE_SCHED_WRR_SHIFT 3
+#define RTE_SCHED_TRAFFIC_CLASS_BE
(RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE - 1)
#define RTE_SCHED_GRINDER_PCACHE_SIZE (64 / RTE_SCHED_QUEUES_PER_PIPE)
#define RTE_SCHED_PIPE_INVALIDUINT32_MAX
#define RTE_SCHED_BMP_POS_INVALID UINT32_MAX
@@ -46,6 +47,73 @@
*/
#define RTE_SCHED_TIME_SHIFT 8
+struct rte_sched_strict_priority_class {
+ struct rte_sched_queue *queue;
+ struct rte_mbuf **qbase;
+ uint32_t qindex;
+ uint16_t qsize;
+};
+
+struct rte_sched_best_effort_class {
+ struct rte_sched_queue *queue[RTE_SCHED_WRR_QUEUES_PER_PIPE];
+ struct rte_mbuf **qbase[RTE_SCHED_WRR_QUEUES_PER_PIPE];
+ uint32_t qindex[RTE_SCHED_WRR_QUEUES_PER_PIPE];
+ uint16_t qsize[RTE_SCHED_WRR_QUEUES_PER_PIPE];
+ uint32_t qmask;
+ uint32_t qpos;
+
+ /* WRR */
+ uint16_t wrr_tokens[RTE_SCHED_WRR_QUEUES_PER_PIPE];
+ uint16_t wrr_mask[RTE_SCHED_WRR_QUEUES_PER_PIPE];
+ uint8_t wrr_cost[RTE_SCHED_WRR_QUEUES_PER_PIPE];
+};
+
+enum grinder_state {
+ e_GRINDER_PREFETCH_PIPE = 0,
+ e_GRINDER_PREFETCH_TC_QUEUE_ARRAYS,
+ e_GRINDER_PREFETCH_MBUF,
+ e_GRINDER_READ_MBUF
+};
+
+struct rte_sched_grinder {
+ /* Pipe cache */
+ uint16_t pcache_qmask[RTE_SCHED_GRINDER_PCACHE_SIZE];
+ uint32_t pcache_qindex[RTE_SCHED_GRINDER_PCACHE_SIZE];
+ uint32_t pcache_w;
+ uint32_t pcache_r;
+
+ /* Current pipe */
+ enum grinder_state state;
+ uint32_t productive;
+ uint32_t pindex;
+ struct rte_sched_subport *subport;
+ struct rte_sched_pipe *pipe;
+ struct rte_sched_pipe_profile *pipe_params;
+
+ /* TC cache */
+ uint8_t tccache_qmask[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE];
+ uint32_t tccache_qindex[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE];
+ uint32_t tccache_w;
+ uint32_t tccache_r;
+
+ /* Current TC */
+ uint32_t tc_index;
+ struct rte_sched_strict_priority_class sp;
+ struct rte_sched_best_effort_class be;
+ struct rte_sched_queue *queue[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE];
+ struct rte_mbuf **qbase[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE];
+ uint32_t qindex[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE];
+ uint16_t qsize;
+ uint32_t qmask;
+ uint32_t qpos;
+ struct rte_mbuf *pkt;
+
+ /* WRR */
+ uint16_t wrr_tokens[RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS];
+ uint16_t wrr_mask[RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS];
+ uint8_t wrr_cost[RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS];
+};
+
struct rte_sched_subport {
/* Token bucket (TB) */
uint64_t tb_time; /* time of last update */
@@ -71,7 +139,41 @@ struct rte_sched_subport {
/* Statistics */
struct rte_sched_subport_stats stats;
-};
+
+ /* Subport Pipes*/
+ uint32_t n_subport_pipes;
+
+ uint16_t qsize[RTE_SCHED_QUEUES_PER_PIPE];
+ uint32_t n_pipe_profiles;
+ uint32_t pipe_tc_be_rate_max;
+#ifdef RTE_SCHED_RED
+ struct rte_red_config
red_config[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE][RTE_COLORS];
+#endif
+
+ /* Scheduling loop detection */
+ uint32_t pipe_loop;
+ uint32_t pipe_exhaustion;
+
+ /* Bitmap */
+ struct rte_bitmap *bmp;
+ uint32_t grinder_base_bmp_pos[RTE_SCHED_PORT_N_GRINDERS]
__rte_aligned_16;
+
+ /* Grinders */
+ struct rte_sched_grinder grinder[RTE_SCHED_PORT_N_GRINDERS];
+ uint32_t busy_grinders;
+
+ /* Queue base calculation */
+ uint32_t qsize_add[RTE_SCHED_QUEUES_PER_PIPE];
+ uint32_t qsize_sum;
+
+ struct rte_sched_pipe *pipe;
+ struct rte_sched_queue *queue;
+ struct rte_sched_queue_extra *queue_extra;
+ struct rte_sched_pipe_profile *pipe_profiles;
+ uint8_t *bmp_array;
+ struct rte_mbuf **queue_array;
+ uint8_t memory[0] __rte_cache_aligned;
+} __rte_cache_aligned;
struct rte_sched_pipe_profile {
/* Token bucket (TB) */
@@ -84,8 +186,12 @@ struct rte_sched_pipe_profile {
uint32_t tc_credits_per_period[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE];
uint8_t tc_ov_weight;
+ /* Strict priority and best effort traffic class queues */
+ uint8_t n_sp_q
[dpdk-dev] [PATCH 02/27] sched: update subport and pipe data structures
From: Jasvinder Singh
Update public data structures for subport and pipe to allow configuration
flexiblity for pipe traffic classes and queues, and subport level
configuration of the pipe parameters.
Signed-off-by: Jasvinder Singh
Signed-off-by: Abraham Tovar
Signed-off-by: Lukasz Krakowiak
---
app/test/test_sched.c| 2 +-
examples/qos_sched/init.c| 2 +-
lib/librte_sched/rte_sched.h | 76 ++--
3 files changed, 49 insertions(+), 31 deletions(-)
diff --git a/app/test/test_sched.c b/app/test/test_sched.c
index 4eed8dbde..460eb53ec 100644
--- a/app/test/test_sched.c
+++ b/app/test/test_sched.c
@@ -40,7 +40,7 @@ static struct rte_sched_pipe_params pipe_profile[] = {
.tc_rate = {305175, 305175, 305175, 305175},
.tc_period = 40,
- .wrr_weights = {1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
1},
+ .wrr_weights = {1, 1, 1, 1, 1, 1, 1, 1},
},
};
diff --git a/examples/qos_sched/init.c b/examples/qos_sched/init.c
index 37c2b95fd..f44a07cd6 100644
--- a/examples/qos_sched/init.c
+++ b/examples/qos_sched/init.c
@@ -186,7 +186,7 @@ static struct rte_sched_pipe_params
pipe_profiles[RTE_SCHED_PIPE_PROFILES_PER_PO
.tc_ov_weight = 1,
#endif
- .wrr_weights = {1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
1},
+ .wrr_weights = {1, 1, 1, 1, 1, 1, 1, 1},
},
};
diff --git a/lib/librte_sched/rte_sched.h b/lib/librte_sched/rte_sched.h
index cf7695f27..71728f725 100644
--- a/lib/librte_sched/rte_sched.h
+++ b/lib/librte_sched/rte_sched.h
@@ -117,6 +117,33 @@ extern "C" {
#define RTE_SCHED_FRAME_OVERHEAD_DEFAULT 24
#endif
+/*
+ * Pipe configuration parameters. The period and credits_per_period
+ * parameters are measured in bytes, with one byte meaning the time
+ * duration associated with the transmission of one byte on the
+ * physical medium of the output port, with pipe or pipe traffic class
+ * rate (measured as percentage of output port rate) determined as
+ * credits_per_period divided by period. One credit represents one
+ * byte.
+ */
+struct rte_sched_pipe_params {
+ /* Pipe token bucket */
+ uint32_t tb_rate; /**< Rate (measured in bytes per second) */
+ uint32_t tb_size; /**< Size (measured in credits) */
+
+ /* Pipe traffic classes */
+ uint32_t tc_rate[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE];
+ /**< Traffic class rates (measured in bytes per second) */
+ uint32_t tc_period;
+ /**< Enforcement period (measured in milliseconds) */
+#ifdef RTE_SCHED_SUBPORT_TC_OV
+ uint8_t tc_ov_weight; /**< Weight Traffic class 3 oversubscription */
+#endif
+
+ /* Pipe queues */
+ uint8_t wrr_weights[RTE_SCHED_WRR_QUEUES_PER_PIPE]; /**< WRR weights */
+};
+
/*
* Subport configuration parameters. The period and credits_per_period
* parameters are measured in bytes, with one byte meaning the time
@@ -128,14 +155,32 @@ extern "C" {
*/
struct rte_sched_subport_params {
/* Subport token bucket */
- uint32_t tb_rate;/**< Rate (measured in bytes per
second) */
- uint32_t tb_size;/**< Size (measured in credits) */
+ uint32_t tb_rate; /**< Rate (measured in bytes per second) */
+ uint32_t tb_size; /**< Size (measured in credits) */
/* Subport traffic classes */
uint32_t tc_rate[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE];
/**< Traffic class rates (measured in bytes per second) */
uint32_t tc_period;
/**< Enforcement period for rates (measured in milliseconds) */
+
+ uint32_t n_subport_pipes;/**< Number of subport_pipes */
+ uint16_t qsize[RTE_SCHED_QUEUES_PER_PIPE];
+ /**< Packet queue size for each traffic class.
+* All queues which are not needed, have zero size. All the pipes
+* within the same subport share the similar configuration for the
+* queues.
+*/
+ struct rte_sched_pipe_params *pipe_profiles;
+ /**< Pipe profile table.
+* Every pipe is configured using one of the profiles from this table.
+*/
+ uint32_t n_pipe_profiles; /**< Profiles in the pipe profile table */
+#ifdef RTE_SCHED_RED
+ struct rte_red_params
+ red_params[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE][RTE_COLORS];
+ /**< RED parameters */
+#endif
};
/** Subport statistics */
@@ -158,33 +203,6 @@ struct rte_sched_subport_stats {
#endif
};
-/*
- * Pipe configuration parameters. The period and credits_per_period
- * parameters are measured in bytes, with one byte meaning the time
- * duration associated with the transmission of one byte on the
- * physical medium of the output port, with pipe or pipe traffic class
- * rate (measured as percentage of output port rate) determined as
- * credits_per_period divided by period. One credit represents one
- * byte.
- */
-struct rte_sched_pipe_par
[dpdk-dev] [PATCH 06/27] sched: update subport config api
From: Jasvinder Singh
Update suport configuration api implementation of the scheduler to allow
configuration flexiblity for pipe traffic classes and queues, and subport
level configuration of the pipe parameters.
Signed-off-by: Jasvinder Singh
Signed-off-by: Abraham Tovar
Signed-off-by: Lukasz Krakowiak
---
lib/librte_sched/rte_sched.c | 325 ++-
1 file changed, 283 insertions(+), 42 deletions(-)
diff --git a/lib/librte_sched/rte_sched.c b/lib/librte_sched/rte_sched.c
index 39a6165e3..020c028fd 100644
--- a/lib/librte_sched/rte_sched.c
+++ b/lib/librte_sched/rte_sched.c
@@ -495,24 +495,72 @@ rte_sched_port_get_array_base(struct
rte_sched_port_params *params, enum rte_sch
return base;
}
-uint32_t
-rte_sched_port_get_memory_footprint(struct rte_sched_port_params *params)
+static uint32_t
+rte_sched_subport_get_array_base(struct rte_sched_subport_params *params,
+ enum rte_sched_subport_array array)
{
- uint32_t size0, size1;
- int status;
+ uint32_t n_subport_pipes = params->n_subport_pipes;
+ uint32_t n_subport_queues = RTE_SCHED_QUEUES_PER_PIPE * n_subport_pipes;
- status = rte_sched_port_check_params(params);
- if (status != 0) {
- RTE_LOG(NOTICE, SCHED,
- "Port scheduler params check failed (%d)\n", status);
+ uint32_t size_pipe = n_subport_pipes * sizeof(struct rte_sched_pipe);
+ uint32_t size_queue = n_subport_queues * sizeof(struct rte_sched_queue);
+ uint32_t size_queue_extra
+ = n_subport_queues * sizeof(struct rte_sched_queue_extra);
+ uint32_t size_pipe_profiles = RTE_SCHED_PIPE_PROFILES_PER_SUBPORT *
+ sizeof(struct rte_sched_pipe_profile);
+ uint32_t size_bmp_array =
+ rte_bitmap_get_memory_footprint(n_subport_queues);
+ uint32_t size_per_pipe_queue_array, size_queue_array;
- return 0;
+ uint32_t base, i;
+
+ size_per_pipe_queue_array = 0;
+ for (i = 0; i < RTE_SCHED_QUEUES_PER_PIPE; i++) {
+ size_per_pipe_queue_array += params->qsize[i] * sizeof(struct
rte_mbuf *);
}
+ size_queue_array = n_subport_pipes * size_per_pipe_queue_array;
- size0 = sizeof(struct rte_sched_port);
- size1 = rte_sched_port_get_array_base(params,
e_RTE_SCHED_PORT_ARRAY_TOTAL);
+ base = 0;
- return size0 + size1;
+ if (array == e_RTE_SCHED_SUBPORT_ARRAY_PIPE)
+ return base;
+ base += RTE_CACHE_LINE_ROUNDUP(size_pipe);
+
+ if (array == e_RTE_SCHED_SUBPORT_ARRAY_QUEUE)
+ return base;
+ base += RTE_CACHE_LINE_ROUNDUP(size_queue);
+
+ if (array == e_RTE_SCHED_SUBPORT_ARRAY_QUEUE_EXTRA)
+ return base;
+ base += RTE_CACHE_LINE_ROUNDUP(size_queue_extra);
+
+ if (array == e_RTE_SCHED_SUBPORT_ARRAY_PIPE_PROFILES)
+ return base;
+ base += RTE_CACHE_LINE_ROUNDUP(size_pipe_profiles);
+
+ if (array == e_RTE_SCHED_SUBPORT_ARRAY_BMP_ARRAY)
+ return base;
+ base += RTE_CACHE_LINE_ROUNDUP(size_bmp_array);
+
+ if (array == e_RTE_SCHED_SUBPORT_ARRAY_QUEUE_ARRAY)
+ return base;
+ base += RTE_CACHE_LINE_ROUNDUP(size_queue_array);
+
+ return base;
+}
+
+static void
+rte_sched_subport_config_qsize(struct rte_sched_subport *subport)
+{
+ uint32_t i;
+
+ subport->qsize_add[0] = 0;
+
+ for (i = 1; i < RTE_SCHED_QUEUES_PER_PIPE; i++)
+ subport->qsize_add[i] =
+ subport->qsize_add[i-1] + subport->qsize[i-1];
+
+ subport->qsize_sum = subport->qsize_add[15] + subport->qsize[15];
}
static void
@@ -621,6 +669,120 @@ rte_sched_pipe_profile_convert(struct
rte_sched_pipe_params *src,
}
}
+static int
+rte_sched_subport_check_params(struct rte_sched_subport_params *params,
+ uint32_t rate)
+{
+ uint32_t i, j;
+
+ /* Check user parameters */
+ if (params == NULL)
+ return -1;
+
+ if (params->tb_rate == 0 || params->tb_rate > rate)
+ return -2;
+
+ if (params->tb_size == 0)
+ return -3;
+
+ for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i++)
+ if (params->tc_rate[i] > params->tb_rate) {
+ printf("traffic class %u, tc_rate %u, tb_rate %u\n", i,
+ params->tc_rate[i], params->tb_rate);
+ return -4;
+ }
+ if (params->tc_period == 0)
+ return -6;
+
+ /* n_subport_pipes: non-zero, power of 2 */
+ if (params->n_subport_pipes == 0 ||
+ !rte_is_power_of_2(params->n_subport_pipes))
+ return -7;
+
+ /* qsize: power of 2, if non-zero
+* no bigger than 32K (due to 16-bit read/write pointers)
+*/
+ for (i = 0, j = 0; i < RTE_SCHED_QUEUES_PER_PIPE; i++) {
+ uint
[dpdk-dev] [PATCH 07/27] sched: update pipe profile add api
From: Jasvinder Singh
Update the pipe profile add api implementation of the scheduler to allow
configuration flexiblity for pipe traffic classes and queues, and subport
level configuration of the pipe parameters.
Signed-off-by: Jasvinder Singh
Signed-off-by: Abraham Tovar
Signed-off-by: Lukasz Krakowiak
---
lib/librte_sched/rte_sched.c | 257 +++
lib/librte_sched/rte_sched.h | 3 +
2 files changed, 205 insertions(+), 55 deletions(-)
diff --git a/lib/librte_sched/rte_sched.c b/lib/librte_sched/rte_sched.c
index 020c028fd..c1079cdaa 100644
--- a/lib/librte_sched/rte_sched.c
+++ b/lib/librte_sched/rte_sched.c
@@ -365,44 +365,49 @@ rte_sched_port_qsize(struct rte_sched_port *port,
uint32_t qindex)
static int
pipe_profile_check(struct rte_sched_pipe_params *params,
- uint32_t rate)
+ uint32_t rate, uint16_t *qsize)
{
uint32_t i;
/* Pipe parameters */
if (params == NULL)
- return -10;
+ return -11;
/* TB rate: non-zero, not greater than port rate */
if (params->tb_rate == 0 ||
params->tb_rate > rate)
- return -11;
+ return -12;
/* TB size: non-zero */
if (params->tb_size == 0)
- return -12;
+ return -13;
/* TC rate: non-zero, less than pipe rate */
- for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i++) {
- if (params->tc_rate[i] == 0 ||
- params->tc_rate[i] > params->tb_rate)
- return -13;
+ for (i = 0; i < RTE_SCHED_TRAFFIC_CLASS_BE; i++) {
+ if ((qsize[i] == 0 && params->tc_rate[i] != 0) ||
+ (qsize[i] != 0 && (params->tc_rate[i] == 0 ||
+ params->tc_rate[i] > params->tb_rate)))
+ return -14;
}
+ if (params->tc_rate[RTE_SCHED_TRAFFIC_CLASS_BE] == 0)
+ return -15;
/* TC period: non-zero */
if (params->tc_period == 0)
- return -14;
+ return -16;
#ifdef RTE_SCHED_SUBPORT_TC_OV
/* TC3 oversubscription weight: non-zero */
if (params->tc_ov_weight == 0)
- return -15;
+ return -17;
#endif
/* Queue WRR weights: non-zero */
- for (i = 0; i < RTE_SCHED_QUEUES_PER_PIPE; i++) {
- if (params->wrr_weights[i] == 0)
- return -16;
+ for (i = 0; i < RTE_SCHED_WRR_QUEUES_PER_PIPE; i++) {
+ uint32_t qindex = RTE_SCHED_TRAFFIC_CLASS_BE + i;
+ if ((qsize[qindex] != 0 && params->wrr_weights[i] == 0) ||
+ (qsize[qindex] == 0 && params->wrr_weights[i] != 0))
+ return -18;
}
return 0;
@@ -549,6 +554,120 @@ rte_sched_subport_get_array_base(struct
rte_sched_subport_params *params,
return base;
}
+static void
+rte_sched_pipe_queues_config(struct rte_sched_subport *subport,
+ struct rte_sched_pipe_profile *dst)
+{
+ uint32_t i;
+
+ /* Queues: strict priority */
+ for (i = 0; i < RTE_SCHED_TRAFFIC_CLASS_BE; i++)
+ if (subport->qsize[i])
+ dst->n_sp_queues++;
+
+ /* Queues: best effort */
+ for (i = 0; i < RTE_SCHED_WRR_QUEUES_PER_PIPE; i++)
+ if (subport->qsize[RTE_SCHED_TRAFFIC_CLASS_BE + i])
+ dst->n_be_queues++;
+}
+
+static void
+rte_sched_pipe_wrr_queues_config(struct rte_sched_pipe_params *src,
+ struct rte_sched_pipe_profile *dst)
+{
+ uint32_t wrr_cost[RTE_SCHED_WRR_QUEUES_PER_PIPE];
+
+ if (dst->n_be_queues == 1) {
+ dst->wrr_cost[0] = (uint8_t) src->wrr_weights[0];
+
+ return;
+ }
+
+ if (dst->n_be_queues == 2) {
+ uint32_t lcd;
+ wrr_cost[0] = src->wrr_weights[0];
+ wrr_cost[1] = src->wrr_weights[1];
+
+ lcd = rte_get_lcd(wrr_cost[0], wrr_cost[1]);
+
+ wrr_cost[0] = lcd / wrr_cost[0];
+ wrr_cost[1] = lcd / wrr_cost[1];
+
+ dst->wrr_cost[0] = (uint8_t) wrr_cost[0];
+ dst->wrr_cost[1] = (uint8_t) wrr_cost[1];
+
+ return;
+ }
+
+ if (dst->n_be_queues == 4) {
+ uint32_t lcd, lcd1, lcd2;
+
+ wrr_cost[0] = src->wrr_weights[0];
+ wrr_cost[1] = src->wrr_weights[1];
+ wrr_cost[2] = src->wrr_weights[2];
+ wrr_cost[3] = src->wrr_weights[3];
+
+ lcd1 = rte_get_lcd(wrr_cost[0], wrr_cost[1]);
+ lcd2 = rte_get_lcd(wrr_cost[2], wrr_cost[3]);
+ lcd = rte_get_lcd(lcd1, lcd2);
+
+ wrr_cost[0] = lcd / wrr_cost[0];
+ wrr_cost[1] = lcd / wrr_cost[1];
+ wrr_cost[2] = lcd / wrr_cost[2];
+ wrr_cost[3] = lcd / wrr_cost[3];
+
+
[dpdk-dev] [PATCH 05/27] sched: update port free api
From: Jasvinder Singh
Update the scheduler port free api implementation to allow configuration
flexiblity for pipe traffic classes and queues, and subport level
configuration of the pipe parameters.
Signed-off-by: Jasvinder Singh
Signed-off-by: Abraham Tovar
Signed-off-by: Lukasz Krakowiak
---
lib/librte_sched/rte_sched.c | 61
1 file changed, 48 insertions(+), 13 deletions(-)
diff --git a/lib/librte_sched/rte_sched.c b/lib/librte_sched/rte_sched.c
index 8f11b4597..39a6165e3 100644
--- a/lib/librte_sched/rte_sched.c
+++ b/lib/librte_sched/rte_sched.c
@@ -316,6 +316,29 @@ rte_sched_port_queues_per_subport(struct rte_sched_port
*port)
#endif
+static inline uint32_t
+rte_sched_subport_queues(struct rte_sched_subport *subport)
+{
+ return RTE_SCHED_QUEUES_PER_PIPE * subport->n_subport_pipes;
+}
+
+static inline struct rte_mbuf **
+rte_sched_subport_qbase(struct rte_sched_subport *subport, uint32_t qindex)
+{
+ uint32_t pindex = qindex >> 4;
+ uint32_t qpos = qindex & 0xF;
+
+ return (subport->queue_array + pindex *
+ subport->qsize_sum + subport->qsize_add[qpos]);
+}
+
+static inline uint16_t
+rte_sched_subport_qsize(struct rte_sched_subport *subport, uint32_t qindex)
+{
+ uint32_t qpos = qindex & 0xF;
+ return subport->qsize[qpos];
+}
+
static inline uint32_t
rte_sched_port_queues_per_port(struct rte_sched_port *port)
{
@@ -647,28 +670,40 @@ rte_sched_port_config(struct rte_sched_port_params
*params)
void
rte_sched_port_free(struct rte_sched_port *port)
{
- uint32_t qindex;
- uint32_t n_queues_per_port;
+ uint32_t n_subport_queues;
+ uint32_t qindex, i;
/* Check user parameters */
if (port == NULL)
return;
- n_queues_per_port = rte_sched_port_queues_per_port(port);
+ for (i = 0; i < port->n_subports_per_port; i++) {
+ struct rte_sched_subport *s = port->subports[i];
- /* Free enqueued mbufs */
- for (qindex = 0; qindex < n_queues_per_port; qindex++) {
- struct rte_mbuf **mbufs = rte_sched_port_qbase(port, qindex);
- uint16_t qsize = rte_sched_port_qsize(port, qindex);
- struct rte_sched_queue *queue = port->queue + qindex;
- uint16_t qr = queue->qr & (qsize - 1);
- uint16_t qw = queue->qw & (qsize - 1);
+ if (s == NULL)
+ continue;
+
+ n_subport_queues = rte_sched_subport_queues(s);
+
+ /* Free enqueued mbufs */
+ for (qindex = 0; qindex < n_subport_queues; qindex++) {
+ struct rte_mbuf **mbufs =
+ rte_sched_subport_qbase(s, qindex);
+ uint16_t qsize = rte_sched_subport_qsize(s, qindex);
+ if (qsize != 0) {
+ struct rte_sched_queue *queue =
+ s->queue + qindex;
+ uint16_t qr = queue->qr & (qsize - 1);
+ uint16_t qw = queue->qw & (qsize - 1);
+
+ for (; qr != qw; qr = (qr + 1) & (qsize - 1))
+ rte_pktmbuf_free(mbufs[qr]);
+ }
+ }
- for (; qr != qw; qr = (qr + 1) & (qsize - 1))
- rte_pktmbuf_free(mbufs[qr]);
+ rte_bitmap_free(s->bmp);
}
- rte_bitmap_free(port->bmp);
rte_free(port);
}
--
2.20.1
[dpdk-dev] [PATCH 09/27] sched: update pkt read and write api
From: Jasvinder Singh
Update run time packet read and write api implementation of the scheduler
to allow configuration flexiblity for pipe traffic classes and queues, and
subport level configuration of the pipe parameters..
Signed-off-by: Jasvinder Singh
Signed-off-by: Abraham Tovar
Signed-off-by: Lukasz Krakowiak
---
lib/librte_sched/rte_sched.c | 32 +---
lib/librte_sched/rte_sched.h | 8
2 files changed, 21 insertions(+), 19 deletions(-)
diff --git a/lib/librte_sched/rte_sched.c b/lib/librte_sched/rte_sched.c
index 4b1959bb4..d28d2d203 100644
--- a/lib/librte_sched/rte_sched.c
+++ b/lib/librte_sched/rte_sched.c
@@ -1351,17 +1351,15 @@ rte_sched_port_pipe_profile_add(struct rte_sched_port
*port,
static inline uint32_t
rte_sched_port_qindex(struct rte_sched_port *port,
+ struct rte_sched_subport *s,
uint32_t subport,
uint32_t pipe,
- uint32_t traffic_class,
uint32_t queue)
{
return ((subport & (port->n_subports_per_port - 1)) <<
- (port->n_pipes_per_subport_log2 + 4)) |
- ((pipe & (port->n_pipes_per_subport - 1)) << 4) |
- ((traffic_class &
- (RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE - 1)) << 2) |
- (queue & (RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS - 1));
+ (port->n_max_subport_pipes_log2 + 4)) |
+ ((pipe & (s->n_subport_pipes - 1)) << 4) |
+ (queue & (RTE_SCHED_QUEUES_PER_PIPE - 1));
}
void
@@ -1371,9 +1369,9 @@ rte_sched_port_pkt_write(struct rte_sched_port *port,
uint32_t traffic_class,
uint32_t queue, enum rte_color color)
{
- uint32_t queue_id = rte_sched_port_qindex(port, subport, pipe,
- traffic_class, queue);
- rte_mbuf_sched_set(pkt, queue_id, traffic_class, (uint8_t)color);
+ struct rte_sched_subport *s = port->subports[subport];
+ uint32_t qindex = rte_sched_port_qindex(port, s, subport, pipe, queue);
+ rte_mbuf_sched_set(pkt, qindex, traffic_class, (uint8_t)color);
}
void
@@ -1382,13 +1380,17 @@ rte_sched_port_pkt_read_tree_path(struct rte_sched_port
*port,
uint32_t *subport, uint32_t *pipe,
uint32_t *traffic_class, uint32_t *queue)
{
- uint32_t queue_id = rte_mbuf_sched_queue_get(pkt);
+ struct rte_sched_subport *s;
+ uint32_t qindex = rte_mbuf_sched_queue_get(pkt);
+ uint32_t tc_id = rte_mbuf_sched_traffic_class_get(pkt);
+
+ *subport = (qindex >> (port->n_max_subport_pipes_log2 + 4)) &
+ (port->n_subports_per_port - 1);
- *subport = queue_id >> (port->n_pipes_per_subport_log2 + 4);
- *pipe = (queue_id >> 4) & (port->n_pipes_per_subport - 1);
- *traffic_class = (queue_id >> 2) &
- (RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE - 1);
- *queue = queue_id & (RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS - 1);
+ s = port->subports[*subport];
+ *pipe = (qindex >> 4) & (s->n_subport_pipes - 1);
+ *traffic_class = tc_id;
+ *queue = qindex & (RTE_SCHED_QUEUES_PER_PIPE - 1);
}
enum rte_color
diff --git a/lib/librte_sched/rte_sched.h b/lib/librte_sched/rte_sched.h
index da5790fc4..635b59550 100644
--- a/lib/librte_sched/rte_sched.h
+++ b/lib/librte_sched/rte_sched.h
@@ -402,9 +402,9 @@ rte_sched_queue_read_stats(struct rte_sched_port *port,
* @param pipe
* Pipe ID within subport
* @param traffic_class
- * Traffic class ID within pipe (0 .. 3)
+ * Traffic class ID within pipe (0 .. 8)
* @param queue
- * Queue ID within pipe traffic class (0 .. 3)
+ * Queue ID within pipe traffic class (0 .. 15)
* @param color
* Packet color set
*/
@@ -429,9 +429,9 @@ rte_sched_port_pkt_write(struct rte_sched_port *port,
* @param pipe
* Pipe ID within subport
* @param traffic_class
- * Traffic class ID within pipe (0 .. 3)
+ * Traffic class ID within pipe (0 .. 8)
* @param queue
- * Queue ID within pipe traffic class (0 .. 3)
+ * Queue ID within pipe traffic class (0 .. 15)
*
*/
void
--
2.20.1
[dpdk-dev] [PATCH 08/27] sched: update pipe config api
From: Jasvinder Singh
Update the pipe configuration api implementation of the scheduler to allow
configuration flexiblity for pipe traffic classes and queues, and subport
level configuration of the pipe parameters.
Signed-off-by: Jasvinder Singh
Signed-off-by: Abraham Tovar
Signed-off-by: Lukasz Krakowiak
---
lib/librte_sched/rte_sched.c | 58 ++--
lib/librte_sched/rte_sched.h | 2 +-
2 files changed, 36 insertions(+), 24 deletions(-)
diff --git a/lib/librte_sched/rte_sched.c b/lib/librte_sched/rte_sched.c
index c1079cdaa..4b1959bb4 100644
--- a/lib/librte_sched/rte_sched.c
+++ b/lib/librte_sched/rte_sched.c
@@ -1206,38 +1206,43 @@ rte_sched_pipe_config(struct rte_sched_port *port,
if (port == NULL ||
subport_id >= port->n_subports_per_port ||
- pipe_id >= port->n_pipes_per_subport ||
- (!deactivate && profile >= port->n_pipe_profiles))
+ pipe_id >= port->subports[subport_id]->n_subport_pipes ||
+ (!deactivate && profile >=
+ port->subports[subport_id]->n_pipe_profiles))
return -1;
/* Check that subport configuration is valid */
- s = port->subport + subport_id;
+ s = port->subports[subport_id];
if (s->tb_period == 0)
return -2;
- p = port->pipe + (subport_id * port->n_pipes_per_subport + pipe_id);
+ p = s->pipe + pipe_id;
/* Handle the case when pipe already has a valid configuration */
if (p->tb_time) {
- params = port->pipe_profiles + p->profile;
+ params = s->pipe_profiles + p->profile;
#ifdef RTE_SCHED_SUBPORT_TC_OV
- double subport_tc3_rate = (double) s->tc_credits_per_period[3]
+ double subport_tc_be_rate =
+ (double) s->tc_credits_per_period[
+ RTE_SCHED_TRAFFIC_CLASS_BE]
/ (double) s->tc_period;
- double pipe_tc3_rate = (double) params->tc_credits_per_period[3]
+ double pipe_tc_be_rate =
+ (double) params->tc_credits_per_period[
+ RTE_SCHED_TRAFFIC_CLASS_BE]
/ (double) params->tc_period;
- uint32_t tc3_ov = s->tc_ov;
+ uint32_t tc_be_ov = s->tc_ov;
/* Unplug pipe from its subport */
s->tc_ov_n -= params->tc_ov_weight;
- s->tc_ov_rate -= pipe_tc3_rate;
- s->tc_ov = s->tc_ov_rate > subport_tc3_rate;
+ s->tc_ov_rate -= pipe_tc_be_rate;
+ s->tc_ov = s->tc_ov_rate > subport_tc_be_rate;
- if (s->tc_ov != tc3_ov) {
+ if (s->tc_ov != tc_be_ov) {
RTE_LOG(DEBUG, SCHED,
- "Subport %u TC3 oversubscription is OFF (%.4lf
>= %.4lf)\n",
- subport_id, subport_tc3_rate, s->tc_ov_rate);
+ "Subport %u Best effort TC oversubscription is
OFF (%.4lf >= %.4lf)\n",
+ subport_id, subport_tc_be_rate, s->tc_ov_rate);
}
#endif
@@ -1250,34 +1255,41 @@ rte_sched_pipe_config(struct rte_sched_port *port,
/* Apply the new pipe configuration */
p->profile = profile;
- params = port->pipe_profiles + p->profile;
+ params = s->pipe_profiles + p->profile;
/* Token Bucket (TB) */
p->tb_time = port->time;
p->tb_credits = params->tb_size / 2;
/* Traffic Classes (TCs) */
+ p->n_sp_queues = params->n_sp_queues;
+ p->n_be_queues = params->n_be_queues;
p->tc_time = port->time + params->tc_period;
for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i++)
- p->tc_credits[i] = params->tc_credits_per_period[i];
+ if (s->qsize[i])
+ p->tc_credits[i] = params->tc_credits_per_period[i];
#ifdef RTE_SCHED_SUBPORT_TC_OV
{
/* Subport TC3 oversubscription */
- double subport_tc3_rate = (double) s->tc_credits_per_period[3]
+ double subport_tc_be_rate =
+ (double) s->tc_credits_per_period[
+ RTE_SCHED_TRAFFIC_CLASS_BE]
/ (double) s->tc_period;
- double pipe_tc3_rate = (double) params->tc_credits_per_period[3]
+ double pipe_tc_be_rate =
+ (double) params->tc_credits_per_period[
+ RTE_SCHED_TRAFFIC_CLASS_BE]
/ (double) params->tc_period;
- uint32_t tc3_ov = s->tc_ov;
+ uint32_t tc_be_ov = s->tc_ov;
s->tc_ov_n += params->tc_ov_weight;
- s->tc_ov_rate += pipe_tc3_rate;
- s->tc_ov = s->tc_ov_rate > subport_tc3_rate;
+ s->tc_ov
[dpdk-dev] [PATCH 10/27] sched: update subport and tc queue stats
From: Jasvinder Singh
Update subport and tc queue stats implementation of scheduler to allow
configuration flexiblity for pipe traffic classes and queues, and subport
level configuration of the pipe parameters.
Signed-off-by: Jasvinder Singh
Signed-off-by: Abraham Tovar
Signed-off-by: Lukasz Krakowiak
---
lib/librte_sched/rte_sched.c | 46 +++-
1 file changed, 24 insertions(+), 22 deletions(-)
diff --git a/lib/librte_sched/rte_sched.c b/lib/librte_sched/rte_sched.c
index d28d2d203..86f2bdf51 100644
--- a/lib/librte_sched/rte_sched.c
+++ b/lib/librte_sched/rte_sched.c
@@ -1412,7 +1412,7 @@ rte_sched_subport_read_stats(struct rte_sched_port *port,
stats == NULL || tc_ov == NULL)
return -1;
- s = port->subport + subport_id;
+ s = port->subports[subport_id];
/* Copy subport stats and clear */
memcpy(stats, &s->stats, sizeof(struct rte_sched_subport_stats));
@@ -1468,10 +1468,10 @@ rte_sched_port_queue_is_empty(struct rte_sched_port
*port, uint32_t qindex)
#ifdef RTE_SCHED_COLLECT_STATS
static inline void
-rte_sched_port_update_subport_stats(struct rte_sched_port *port, uint32_t
qindex, struct rte_mbuf *pkt)
+rte_sched_port_update_subport_stats(struct rte_sched_subport *s,
+ struct rte_mbuf *pkt)
{
- struct rte_sched_subport *s = port->subport + (qindex /
rte_sched_port_queues_per_subport(port));
- uint32_t tc_index = (qindex >> 2) & 0x3;
+ uint32_t tc_index = rte_mbuf_sched_traffic_class_get(pkt);
uint32_t pkt_len = pkt->pkt_len;
s->stats.n_pkts_tc[tc_index] += 1;
@@ -1480,31 +1480,31 @@ rte_sched_port_update_subport_stats(struct
rte_sched_port *port, uint32_t qindex
#ifdef RTE_SCHED_RED
static inline void
-rte_sched_port_update_subport_stats_on_drop(struct rte_sched_port *port,
- uint32_t qindex,
- struct rte_mbuf *pkt, uint32_t
red)
+rte_sched_port_update_subport_stats_on_drop(struct rte_sched_subport *subport,
+ struct rte_mbuf *pkt,
+ uint32_t red)
#else
static inline void
-rte_sched_port_update_subport_stats_on_drop(struct rte_sched_port *port,
- uint32_t qindex,
- struct rte_mbuf *pkt,
__rte_unused uint32_t red)
+rte_sched_port_update_subport_stats_on_drop(struct rte_sched_subport *subport,
+ struct rte_mbuf *pkt,
+ __rte_unused uint32_t red)
#endif
{
- struct rte_sched_subport *s = port->subport + (qindex /
rte_sched_port_queues_per_subport(port));
- uint32_t tc_index = (qindex >> 2) & 0x3;
+ uint32_t tc_index = rte_mbuf_sched_traffic_class_get(pkt);
uint32_t pkt_len = pkt->pkt_len;
- s->stats.n_pkts_tc_dropped[tc_index] += 1;
- s->stats.n_bytes_tc_dropped[tc_index] += pkt_len;
+ subport->stats.n_pkts_tc_dropped[tc_index] += 1;
+ subport->stats.n_bytes_tc_dropped[tc_index] += pkt_len;
#ifdef RTE_SCHED_RED
- s->stats.n_pkts_red_dropped[tc_index] += red;
+ subport->stats.n_pkts_red_dropped[tc_index] += red;
#endif
}
static inline void
-rte_sched_port_update_queue_stats(struct rte_sched_port *port, uint32_t
qindex, struct rte_mbuf *pkt)
+rte_sched_port_update_queue_stats(struct rte_sched_subport *subport,
+ uint32_t qindex, struct rte_mbuf *pkt)
{
- struct rte_sched_queue_extra *qe = port->queue_extra + qindex;
+ struct rte_sched_queue_extra *qe = subport->queue_extra + qindex;
uint32_t pkt_len = pkt->pkt_len;
qe->stats.n_pkts += 1;
@@ -1513,17 +1513,19 @@ rte_sched_port_update_queue_stats(struct rte_sched_port
*port, uint32_t qindex,
#ifdef RTE_SCHED_RED
static inline void
-rte_sched_port_update_queue_stats_on_drop(struct rte_sched_port *port,
+rte_sched_port_update_queue_stats_on_drop(struct rte_sched_subport *subport,
uint32_t qindex,
- struct rte_mbuf *pkt, uint32_t
red)
+ struct rte_mbuf *pkt,
+ int32_t red)
#else
static inline void
-rte_sched_port_update_queue_stats_on_drop(struct rte_sched_port *port,
+rte_sched_port_update_queue_stats_on_drop(struct rte_sched_subport *subport,
uint32_t qindex,
- struct rte_mbuf *pkt,
__rte_unused uint32_t red)
+ struct rte_mbuf *pkt,
+ __rte_unused uint32_t red)
#endif
{
- struct rte_sched_queue_extra *qe = port->queue_extra + qindex;
+ struct rte_sche
[dpdk-dev] [PATCH 13/27] sched: update grinder pipe and tc cache
From: Jasvinder Singh
Update grinder pipe and tc cache population of scheduler to allow
configuration flexiblity for pipe traffic classes and queues, and
subport level configuration of the pipe parameters.
Signed-off-by: Jasvinder Singh
Signed-off-by: Abraham Tovar
Signed-off-by: Lukasz Krakowiak
---
lib/librte_sched/rte_sched.c | 47
1 file changed, 21 insertions(+), 26 deletions(-)
diff --git a/lib/librte_sched/rte_sched.c b/lib/librte_sched/rte_sched.c
index 6cfa73761..32a5121cd 100644
--- a/lib/librte_sched/rte_sched.c
+++ b/lib/librte_sched/rte_sched.c
@@ -2174,9 +2174,10 @@ grinder_pipe_exists(struct rte_sched_port *port,
uint32_t base_pipe)
#endif /* RTE_SCHED_OPTIMIZATIONS */
static inline void
-grinder_pcache_populate(struct rte_sched_port *port, uint32_t pos, uint32_t
bmp_pos, uint64_t bmp_slab)
+grinder_pcache_populate(struct rte_sched_subport *subport, uint32_t pos,
+ uint32_t bmp_pos, uint64_t bmp_slab)
{
- struct rte_sched_grinder *grinder = port->grinder + pos;
+ struct rte_sched_grinder *grinder = subport->grinder + pos;
uint16_t w[4];
grinder->pcache_w = 0;
@@ -2205,34 +2206,28 @@ grinder_pcache_populate(struct rte_sched_port *port,
uint32_t pos, uint32_t bmp_
}
static inline void
-grinder_tccache_populate(struct rte_sched_port *port, uint32_t pos, uint32_t
qindex, uint16_t qmask)
+grinder_tccache_populate(struct rte_sched_subport *subport, uint32_t pos,
+ uint32_t qindex, uint16_t qmask)
{
- struct rte_sched_grinder *grinder = port->grinder + pos;
- uint8_t b[4];
+ struct rte_sched_grinder *grinder = subport->grinder + pos;
+ uint32_t i;
+ uint8_t b;
grinder->tccache_w = 0;
grinder->tccache_r = 0;
- b[0] = (uint8_t) (qmask & 0xF);
- b[1] = (uint8_t) ((qmask >> 4) & 0xF);
- b[2] = (uint8_t) ((qmask >> 8) & 0xF);
- b[3] = (uint8_t) ((qmask >> 12) & 0xF);
-
- grinder->tccache_qmask[grinder->tccache_w] = b[0];
- grinder->tccache_qindex[grinder->tccache_w] = qindex;
- grinder->tccache_w += (b[0] != 0);
-
- grinder->tccache_qmask[grinder->tccache_w] = b[1];
- grinder->tccache_qindex[grinder->tccache_w] = qindex + 4;
- grinder->tccache_w += (b[1] != 0);
-
- grinder->tccache_qmask[grinder->tccache_w] = b[2];
- grinder->tccache_qindex[grinder->tccache_w] = qindex + 8;
- grinder->tccache_w += (b[2] != 0);
+ for (i = 0; i < RTE_SCHED_TRAFFIC_CLASS_BE; i++) {
+ b = (uint8_t) ((qmask >> i) & 0x1);
+ grinder->tccache_qmask[grinder->tccache_w] = b;
+ grinder->tccache_qindex[grinder->tccache_w] = qindex + i;
+ grinder->tccache_w += (b != 0);
+ }
- grinder->tccache_qmask[grinder->tccache_w] = b[3];
- grinder->tccache_qindex[grinder->tccache_w] = qindex + 12;
- grinder->tccache_w += (b[3] != 0);
+ b = (uint8_t) (qmask >> (RTE_SCHED_TRAFFIC_CLASS_BE));
+ grinder->tccache_qmask[grinder->tccache_w] = b;
+ grinder->tccache_qindex[grinder->tccache_w] = qindex +
+ RTE_SCHED_TRAFFIC_CLASS_BE;
+ grinder->tccache_w += (b != 0);
}
static inline int
@@ -2304,7 +2299,7 @@ grinder_next_pipe(struct rte_sched_port *port, uint32_t
pos)
port->grinder_base_bmp_pos[pos] = bmp_pos;
/* Install new pipe group into grinder's pipe cache */
- grinder_pcache_populate(port, pos, bmp_pos, bmp_slab);
+ grinder_pcache_populate(port->subport, pos, bmp_pos, bmp_slab);
pipe_qmask = grinder->pcache_qmask[0];
pipe_qindex = grinder->pcache_qindex[0];
@@ -2318,7 +2313,7 @@ grinder_next_pipe(struct rte_sched_port *port, uint32_t
pos)
grinder->pipe_params = NULL; /* to be set after the pipe structure is
prefetched */
grinder->productive = 0;
- grinder_tccache_populate(port, pos, pipe_qindex, pipe_qmask);
+ grinder_tccache_populate(port->subport, pos, pipe_qindex, pipe_qmask);
grinder_next_tc(port, pos);
/* Check for pipe exhaustion */
--
2.20.1
[dpdk-dev] [PATCH 16/27] sched: update grinder wrr compute function
From: Jasvinder Singh
Update weighted round robin function for best-effort traffic class
queues of the scheduler to allow configuration flexiblity for pipe traffic
classes and queues, and subport level configuration of the pipe parameters.
Signed-off-by: Jasvinder Singh
Signed-off-by: Abraham Tovar
Signed-off-by: Lukasz Krakowiak
---
lib/librte_sched/rte_sched.c| 135 +---
lib/librte_sched/rte_sched_common.h | 41 +
2 files changed, 125 insertions(+), 51 deletions(-)
diff --git a/lib/librte_sched/rte_sched.c b/lib/librte_sched/rte_sched.c
index 07939c04f..a9b5f7bf8 100644
--- a/lib/librte_sched/rte_sched.c
+++ b/lib/librte_sched/rte_sched.c
@@ -2316,73 +2316,106 @@ grinder_next_pipe(struct rte_sched_subport *subport,
uint32_t pos)
return 1;
}
-
static inline void
-grinder_wrr_load(struct rte_sched_port *port, uint32_t pos)
+grinder_wrr_load(struct rte_sched_subport *subport, uint32_t pos)
{
- struct rte_sched_grinder *grinder = port->grinder + pos;
+ struct rte_sched_grinder *grinder = subport->grinder + pos;
struct rte_sched_pipe *pipe = grinder->pipe;
struct rte_sched_pipe_profile *pipe_params = grinder->pipe_params;
- uint32_t tc_index = grinder->tc_index;
- uint32_t qmask = grinder->qmask;
- uint32_t qindex;
-
- qindex = tc_index * 4;
-
- grinder->wrr_tokens[0] = ((uint16_t) pipe->wrr_tokens[qindex]) <<
RTE_SCHED_WRR_SHIFT;
- grinder->wrr_tokens[1] = ((uint16_t) pipe->wrr_tokens[qindex + 1]) <<
RTE_SCHED_WRR_SHIFT;
- grinder->wrr_tokens[2] = ((uint16_t) pipe->wrr_tokens[qindex + 2]) <<
RTE_SCHED_WRR_SHIFT;
- grinder->wrr_tokens[3] = ((uint16_t) pipe->wrr_tokens[qindex + 3]) <<
RTE_SCHED_WRR_SHIFT;
-
- grinder->wrr_mask[0] = (qmask & 0x1) * 0x;
- grinder->wrr_mask[1] = ((qmask >> 1) & 0x1) * 0x;
- grinder->wrr_mask[2] = ((qmask >> 2) & 0x1) * 0x;
- grinder->wrr_mask[3] = ((qmask >> 3) & 0x1) * 0x;
+ uint32_t qmask = grinder->be.qmask;
+ uint32_t qindex = grinder->be.qindex[0];
+ uint32_t i;
- grinder->wrr_cost[0] = pipe_params->wrr_cost[qindex];
- grinder->wrr_cost[1] = pipe_params->wrr_cost[qindex + 1];
- grinder->wrr_cost[2] = pipe_params->wrr_cost[qindex + 2];
- grinder->wrr_cost[3] = pipe_params->wrr_cost[qindex + 3];
+ for (i = 0; i < pipe->n_be_queues; i++) {
+ grinder->be.wrr_tokens[i] =
+ ((uint16_t) pipe->wrr_tokens[qindex + i]) <<
RTE_SCHED_WRR_SHIFT;
+ grinder->be.wrr_mask[i] = ((qmask >> i) & 0x1) * 0x;
+ grinder->be.wrr_cost[i] = pipe_params->wrr_cost[qindex + i];
+ }
}
static inline void
-grinder_wrr_store(struct rte_sched_port *port, uint32_t pos)
+grinder_wrr_store(struct rte_sched_subport *subport, uint32_t pos)
{
- struct rte_sched_grinder *grinder = port->grinder + pos;
+ struct rte_sched_grinder *grinder = subport->grinder + pos;
struct rte_sched_pipe *pipe = grinder->pipe;
uint32_t tc_index = grinder->tc_index;
- uint32_t qindex;
-
- qindex = tc_index * 4;
+ uint32_t i;
- pipe->wrr_tokens[qindex] = (grinder->wrr_tokens[0] &
grinder->wrr_mask[0])
- >> RTE_SCHED_WRR_SHIFT;
- pipe->wrr_tokens[qindex + 1] = (grinder->wrr_tokens[1] &
grinder->wrr_mask[1])
- >> RTE_SCHED_WRR_SHIFT;
- pipe->wrr_tokens[qindex + 2] = (grinder->wrr_tokens[2] &
grinder->wrr_mask[2])
- >> RTE_SCHED_WRR_SHIFT;
- pipe->wrr_tokens[qindex + 3] = (grinder->wrr_tokens[3] &
grinder->wrr_mask[3])
- >> RTE_SCHED_WRR_SHIFT;
+ if (tc_index == RTE_SCHED_TRAFFIC_CLASS_BE)
+ for (i = 0; i < pipe->n_be_queues; i++)
+ pipe->wrr_tokens[i] =
+ (grinder->be.wrr_tokens[i] &
grinder->be.wrr_mask[i]) >>
+ RTE_SCHED_WRR_SHIFT;
}
static inline void
-grinder_wrr(struct rte_sched_port *port, uint32_t pos)
+grinder_wrr(struct rte_sched_subport *subport, uint32_t pos)
{
- struct rte_sched_grinder *grinder = port->grinder + pos;
+ struct rte_sched_grinder *grinder = subport->grinder + pos;
+ struct rte_sched_pipe *pipe = grinder->pipe;
+ uint32_t n_be_queues = pipe->n_be_queues;
uint16_t wrr_tokens_min;
- grinder->wrr_tokens[0] |= ~grinder->wrr_mask[0];
- grinder->wrr_tokens[1] |= ~grinder->wrr_mask[1];
- grinder->wrr_tokens[2] |= ~grinder->wrr_mask[2];
- grinder->wrr_tokens[3] |= ~grinder->wrr_mask[3];
+ if (n_be_queues == 1) {
+ grinder->be.wrr_tokens[0] |= ~grinder->be.wrr_mask[0];
+ grinder->be.qpos = 0;
+ wrr_tokens_min = grinder->be.wrr_tokens[0];
+ grinder->be.wrr_tokens[0] -= wrr_tokens_min;
+ return;
+ }
+
+ if (n_be_queues == 2) {
+
[dpdk-dev] [PATCH 14/27] sched: update grinder next pipe and tc functions
From: Jasvinder Singh
Update implementation of grinder next pipe and tc functions to allow
configuration flexiblity for pipe traffic classes and queues, and subport
level configuration of the pipe parameters.
Signed-off-by: Jasvinder Singh
Signed-off-by: Abraham Tovar
Signed-off-by: Lukasz Krakowiak
---
lib/librte_sched/rte_sched.c | 115 ---
1 file changed, 53 insertions(+), 62 deletions(-)
diff --git a/lib/librte_sched/rte_sched.c b/lib/librte_sched/rte_sched.c
index 32a5121cd..74a8e0a71 100644
--- a/lib/librte_sched/rte_sched.c
+++ b/lib/librte_sched/rte_sched.c
@@ -345,24 +345,6 @@ rte_sched_port_queues_per_port(struct rte_sched_port *port)
return RTE_SCHED_QUEUES_PER_PIPE * port->n_pipes_per_subport *
port->n_subports_per_port;
}
-static inline struct rte_mbuf **
-rte_sched_port_qbase(struct rte_sched_port *port, uint32_t qindex)
-{
- uint32_t pindex = qindex >> 4;
- uint32_t qpos = qindex & 0xF;
-
- return (port->queue_array + pindex *
- port->qsize_sum + port->qsize_add[qpos]);
-}
-
-static inline uint16_t
-rte_sched_port_qsize(struct rte_sched_port *port, uint32_t qindex)
-{
- uint32_t tc = (qindex >> 2) & 0x3;
-
- return port->qsize[tc];
-}
-
static int
pipe_profile_check(struct rte_sched_pipe_params *params,
uint32_t rate, uint16_t *qsize)
@@ -2120,13 +2102,14 @@ grinder_schedule(struct rte_sched_port *port, uint32_t
pos)
#ifdef SCHED_VECTOR_SSE4
static inline int
-grinder_pipe_exists(struct rte_sched_port *port, uint32_t base_pipe)
+grinder_pipe_exists(struct rte_sched_subport *subport, uint32_t base_pipe)
{
__m128i index = _mm_set1_epi32(base_pipe);
- __m128i pipes = _mm_load_si128((__m128i *)port->grinder_base_bmp_pos);
+ __m128i pipes =
+ _mm_load_si128((__m128i *)subport->grinder_base_bmp_pos);
__m128i res = _mm_cmpeq_epi32(pipes, index);
- pipes = _mm_load_si128((__m128i *)(port->grinder_base_bmp_pos + 4));
+ pipes = _mm_load_si128((__m128i *)(subport->grinder_base_bmp_pos + 4));
pipes = _mm_cmpeq_epi32(pipes, index);
res = _mm_or_si128(res, pipes);
@@ -2139,10 +2122,10 @@ grinder_pipe_exists(struct rte_sched_port *port,
uint32_t base_pipe)
#elif defined(SCHED_VECTOR_NEON)
static inline int
-grinder_pipe_exists(struct rte_sched_port *port, uint32_t base_pipe)
+grinder_pipe_exists(struct rte_sched_subport *subport, uint32_t base_pipe)
{
uint32x4_t index, pipes;
- uint32_t *pos = (uint32_t *)port->grinder_base_bmp_pos;
+ uint32_t *pos = (uint32_t *)subport->grinder_base_bmp_pos;
index = vmovq_n_u32(base_pipe);
pipes = vld1q_u32(pos);
@@ -2159,12 +2142,12 @@ grinder_pipe_exists(struct rte_sched_port *port,
uint32_t base_pipe)
#else
static inline int
-grinder_pipe_exists(struct rte_sched_port *port, uint32_t base_pipe)
+grinder_pipe_exists(struct rte_sched_subport *subport, uint32_t base_pipe)
{
uint32_t i;
for (i = 0; i < RTE_SCHED_PORT_N_GRINDERS; i++) {
- if (port->grinder_base_bmp_pos[i] == base_pipe)
+ if (subport->grinder_base_bmp_pos[i] == base_pipe)
return 1;
}
@@ -2231,47 +2214,54 @@ grinder_tccache_populate(struct rte_sched_subport
*subport, uint32_t pos,
}
static inline int
-grinder_next_tc(struct rte_sched_port *port, uint32_t pos)
+grinder_next_tc(struct rte_sched_subport *subport, uint32_t pos)
{
- struct rte_sched_grinder *grinder = port->grinder + pos;
+ struct rte_sched_grinder *grinder = subport->grinder + pos;
+ struct rte_sched_pipe *pipe = grinder->pipe;
struct rte_mbuf **qbase;
uint32_t qindex;
uint16_t qsize;
+ uint32_t i;
if (grinder->tccache_r == grinder->tccache_w)
return 0;
qindex = grinder->tccache_qindex[grinder->tccache_r];
- qbase = rte_sched_port_qbase(port, qindex);
- qsize = rte_sched_port_qsize(port, qindex);
+ grinder->tc_index =
+ (qindex < RTE_SCHED_TRAFFIC_CLASS_BE) ?
+ qindex : RTE_SCHED_TRAFFIC_CLASS_BE;
- grinder->tc_index = (qindex >> 2) & 0x3;
- grinder->qmask = grinder->tccache_qmask[grinder->tccache_r];
- grinder->qsize = qsize;
+ qbase = rte_sched_subport_qbase(subport, qindex);
+ if (grinder->tc_index < pipe->n_sp_queues) {
+ qsize = rte_sched_subport_qsize(subport, qindex);
- grinder->qindex[0] = qindex;
- grinder->qindex[1] = qindex + 1;
- grinder->qindex[2] = qindex + 2;
- grinder->qindex[3] = qindex + 3;
+ grinder->sp.qindex = qindex;
+ grinder->sp.queue = subport->queue + qindex;
+ grinder->sp.qbase = qbase;
+ grinder->sp.qsize = qsize;
- grinder->queue[0] = port->queue + qindex;
- grinder->queue[1] = port->queue + qindex + 1;
- grinder->queue[2] = port-
[dpdk-dev] [PATCH 15/27] sched: update pipe and tc queues prefetch
From: Jasvinder Singh
Update pipe and tc queues prefetch functions of scheduler to allow
configuration flexiblity for pipe traffic classes and queues, and subport
level configuration of the pipe parameters..
Signed-off-by: Jasvinder Singh
Signed-off-by: Abraham Tovar
Signed-off-by: Lukasz Krakowiak
---
lib/librte_sched/rte_sched.c | 42 +---
1 file changed, 25 insertions(+), 17 deletions(-)
diff --git a/lib/librte_sched/rte_sched.c b/lib/librte_sched/rte_sched.c
index 74a8e0a71..07939c04f 100644
--- a/lib/librte_sched/rte_sched.c
+++ b/lib/librte_sched/rte_sched.c
@@ -2389,34 +2389,42 @@ grinder_wrr(struct rte_sched_port *port, uint32_t pos)
#define grinder_evict(port, pos)
static inline void
-grinder_prefetch_pipe(struct rte_sched_port *port, uint32_t pos)
+grinder_prefetch_pipe(struct rte_sched_subport *subport, uint32_t pos)
{
- struct rte_sched_grinder *grinder = port->grinder + pos;
+ struct rte_sched_grinder *grinder = subport->grinder + pos;
rte_prefetch0(grinder->pipe);
- rte_prefetch0(grinder->queue[0]);
+ rte_prefetch0(grinder->sp.queue);
}
static inline void
grinder_prefetch_tc_queue_arrays(struct rte_sched_port *port, uint32_t pos)
{
- struct rte_sched_grinder *grinder = port->grinder + pos;
- uint16_t qsize, qr[4];
+ struct rte_sched_grinder *grinder = port->subport->grinder + pos;
+ struct rte_sched_pipe *pipe = grinder->pipe;
+ struct rte_sched_queue *queue;
+ uint32_t tc_index = grinder->tc_index, i;
+ uint16_t qsize, qr[RTE_SCHED_WRR_QUEUES_PER_PIPE];
+
+ if (tc_index < pipe->n_sp_queues) {
+ queue = grinder->sp.queue;
+ qsize = grinder->sp.qsize;
+ qr[0] = queue->qr & (qsize - 1);
- qsize = grinder->qsize;
- qr[0] = grinder->queue[0]->qr & (qsize - 1);
- qr[1] = grinder->queue[1]->qr & (qsize - 1);
- qr[2] = grinder->queue[2]->qr & (qsize - 1);
- qr[3] = grinder->queue[3]->qr & (qsize - 1);
+ rte_prefetch0(grinder->sp.qbase + qr[0]);
+ return;
+ }
+
+ for (i = 0; i < pipe->n_be_queues; i++) {
+ queue = grinder->be.queue[i];
+ qsize = grinder->be.qsize[i];
+ qr[i] = queue->qr & (qsize - 1);
- rte_prefetch0(grinder->qbase[0] + qr[0]);
- rte_prefetch0(grinder->qbase[1] + qr[1]);
+ rte_prefetch0(grinder->be.qbase[i] + qr[i]);
+ }
grinder_wrr_load(port, pos);
grinder_wrr(port, pos);
-
- rte_prefetch0(grinder->qbase[2] + qr[2]);
- rte_prefetch0(grinder->qbase[3] + qr[3]);
}
static inline void
@@ -2447,7 +2455,7 @@ grinder_handle(struct rte_sched_port *port, uint32_t pos)
case e_GRINDER_PREFETCH_PIPE:
{
if (grinder_next_pipe(port->subport, pos)) {
- grinder_prefetch_pipe(port, pos);
+ grinder_prefetch_pipe(port->subport, pos);
port->busy_grinders++;
grinder->state = e_GRINDER_PREFETCH_TC_QUEUE_ARRAYS;
@@ -2508,7 +2516,7 @@ grinder_handle(struct rte_sched_port *port, uint32_t pos)
/* Look for another active pipe */
if (grinder_next_pipe(port->subport, pos)) {
- grinder_prefetch_pipe(port, pos);
+ grinder_prefetch_pipe(port->subport, pos);
grinder->state = e_GRINDER_PREFETCH_TC_QUEUE_ARRAYS;
return result;
--
2.20.1
[dpdk-dev] [PATCH 21/27] sched: update packet dequeue api
From: Jasvinder Singh
Update packet dequeue api implementation of scheduler to allow configuration
flexiblity for pipe traffic classes and queues, and subport level
configuration of the pipe parameters.
Signed-off-by: Jasvinder Singh
Signed-off-by: Abraham Tovar
Signed-off-by: Lukasz Krakowiak
---
lib/librte_sched/rte_sched.c | 45
1 file changed, 35 insertions(+), 10 deletions(-)
diff --git a/lib/librte_sched/rte_sched.c b/lib/librte_sched/rte_sched.c
index 0c5479426..116e6a627 100644
--- a/lib/librte_sched/rte_sched.c
+++ b/lib/librte_sched/rte_sched.c
@@ -2677,6 +2677,7 @@ rte_sched_port_time_resync(struct rte_sched_port *port)
uint64_t cycles = rte_get_tsc_cycles();
uint64_t cycles_diff = cycles - port->time_cpu_cycles;
uint64_t bytes_diff;
+ uint32_t i;
/* Compute elapsed time in bytes */
bytes_diff = rte_reciprocal_divide(cycles_diff << RTE_SCHED_TIME_SHIFT,
@@ -2689,20 +2690,21 @@ rte_sched_port_time_resync(struct rte_sched_port *port)
port->time = port->time_cpu_bytes;
/* Reset pipe loop detection */
- port->pipe_loop = RTE_SCHED_PIPE_INVALID;
+ for (i = 0; i < port->n_subports_per_port; i++)
+ port->subports[i]->pipe_loop = RTE_SCHED_PIPE_INVALID;
}
static inline int
-rte_sched_port_exceptions(struct rte_sched_port *port, int second_pass)
+rte_sched_port_exceptions(struct rte_sched_subport *subport, int second_pass)
{
int exceptions;
/* Check if any exception flag is set */
- exceptions = (second_pass && port->busy_grinders == 0) ||
- (port->pipe_exhaustion == 1);
+ exceptions = (second_pass && subport->busy_grinders == 0) ||
+ (subport->pipe_exhaustion == 1);
/* Clear exception flags */
- port->pipe_exhaustion = 0;
+ subport->pipe_exhaustion = 0;
return exceptions;
}
@@ -2710,7 +2712,9 @@ rte_sched_port_exceptions(struct rte_sched_port *port,
int second_pass)
int
rte_sched_port_dequeue(struct rte_sched_port *port, struct rte_mbuf **pkts,
uint32_t n_pkts)
{
- uint32_t i, count;
+ struct rte_sched_subport *subport;
+ uint32_t subport_id = port->subport_id;
+ uint32_t i, n_subports = 0, count;
port->pkts_out = pkts;
port->n_pkts_out = 0;
@@ -2719,10 +2723,31 @@ rte_sched_port_dequeue(struct rte_sched_port *port,
struct rte_mbuf **pkts, uint
/* Take each queue in the grinder one step further */
for (i = 0, count = 0; ; i++) {
- count += grinder_handle(port, port->subport,
- i & (RTE_SCHED_PORT_N_GRINDERS - 1));
- if ((count == n_pkts) ||
- rte_sched_port_exceptions(port, i >=
RTE_SCHED_PORT_N_GRINDERS)) {
+ subport = port->subports[subport_id];
+
+ count += grinder_handle(port, subport, i &
+ (RTE_SCHED_PORT_N_GRINDERS - 1));
+ if (count == n_pkts) {
+ subport_id++;
+
+ if (subport_id == port->n_subports_per_port)
+ subport_id = 0;
+
+ port->subport_id = subport_id;
+ break;
+ }
+
+ if (rte_sched_port_exceptions(subport, i >=
RTE_SCHED_PORT_N_GRINDERS)) {
+ i = 0;
+ subport_id++;
+ n_subports++;
+ }
+
+ if (subport_id == port->n_subports_per_port)
+ subport_id = 0;
+
+ if (n_subports == port->n_subports_per_port) {
+ port->subport_id = subport_id;
break;
}
}
--
2.20.1
[dpdk-dev] [PATCH 18/27] sched: update mbuf prefetch function
From: Jasvinder Singh
Update mbuf prefetch function of the scheduler grinder to allow configuration
flexiblity for pipe traffic classes and queues, and subport level
configuration of the pipe parameters.
Signed-off-by: Jasvinder Singh
Signed-off-by: Abraham Tovar
Signed-off-by: Lukasz Krakowiak
---
lib/librte_sched/rte_sched.c | 42
1 file changed, 33 insertions(+), 9 deletions(-)
diff --git a/lib/librte_sched/rte_sched.c b/lib/librte_sched/rte_sched.c
index ba344f0a1..06d89e3fd 100644
--- a/lib/librte_sched/rte_sched.c
+++ b/lib/librte_sched/rte_sched.c
@@ -2478,19 +2478,43 @@ grinder_prefetch_tc_queue_arrays(struct rte_sched_port
*port, uint32_t pos)
}
static inline void
-grinder_prefetch_mbuf(struct rte_sched_port *port, uint32_t pos)
+grinder_prefetch_mbuf(struct rte_sched_subport *subport, uint32_t pos)
{
- struct rte_sched_grinder *grinder = port->grinder + pos;
- uint32_t qpos = grinder->qpos;
- struct rte_mbuf **qbase = grinder->qbase[qpos];
- uint16_t qsize = grinder->qsize;
- uint16_t qr = grinder->queue[qpos]->qr & (qsize - 1);
+ struct rte_sched_grinder *grinder = subport->grinder + pos;
+ struct rte_mbuf **qbase;
+ uint32_t tc_index = grinder->tc_index;
+ uint32_t qpos;
+ uint16_t qsize, qr;
+
+ if (tc_index < RTE_SCHED_TRAFFIC_CLASS_BE) {
+ qbase = grinder->sp.qbase;
+ qsize = grinder->sp.qsize;
+ qr = grinder->sp.queue->qr & (qsize - 1);
+
+ grinder->pkt = qbase[qr];
+ rte_prefetch0(grinder->pkt);
+
+ if (unlikely((qr & 0x7) == 7)) {
+ uint16_t qr_next =
+ (grinder->sp.queue->qr + 1) & (qsize - 1);
+
+ rte_prefetch0(qbase + qr_next);
+ }
+
+ return;
+ }
+
+ qpos = grinder->be.qpos;
+ qbase = grinder->be.qbase[qpos];
+ qsize = grinder->be.qsize[qpos];
+ qr = grinder->be.queue[qpos]->qr & (qsize - 1);
grinder->pkt = qbase[qr];
rte_prefetch0(grinder->pkt);
if (unlikely((qr & 0x7) == 7)) {
- uint16_t qr_next = (grinder->queue[qpos]->qr + 1) & (qsize - 1);
+ uint16_t qr_next =
+ (grinder->be.queue[qpos]->qr + 1) & (qsize - 1);
rte_prefetch0(qbase + qr_next);
}
@@ -2529,7 +2553,7 @@ grinder_handle(struct rte_sched_port *port, uint32_t pos)
case e_GRINDER_PREFETCH_MBUF:
{
- grinder_prefetch_mbuf(port, pos);
+ grinder_prefetch_mbuf(port->subport, pos);
grinder->state = e_GRINDER_READ_MBUF;
return 0;
@@ -2544,7 +2568,7 @@ grinder_handle(struct rte_sched_port *port, uint32_t pos)
/* Look for next packet within the same TC */
if (result && grinder->qmask) {
grinder_wrr(port->subport, pos);
- grinder_prefetch_mbuf(port, pos);
+ grinder_prefetch_mbuf(port->subport, pos);
return 1;
}
--
2.20.1
[dpdk-dev] [PATCH 22/27] sched: update sched queue stats api
From: Jasvinder Singh
Update queue stats read api implementation of the scheduler to allow
configuration flexiblity for pipe traffic classes and queues, and subport
level configuration of the pipe parameters.
Signed-off-by: Jasvinder Singh
Signed-off-by: Abraham Tovar
Signed-off-by: Lukasz Krakowiak
---
lib/librte_sched/rte_sched.c | 41 +++-
1 file changed, 22 insertions(+), 19 deletions(-)
diff --git a/lib/librte_sched/rte_sched.c b/lib/librte_sched/rte_sched.c
index 116e6a627..563161713 100644
--- a/lib/librte_sched/rte_sched.c
+++ b/lib/librte_sched/rte_sched.c
@@ -306,16 +306,6 @@ enum rte_sched_subport_array {
e_RTE_SCHED_SUBPORT_ARRAY_TOTAL,
};
-#ifdef RTE_SCHED_COLLECT_STATS
-
-static inline uint32_t
-rte_sched_port_queues_per_subport(struct rte_sched_port *port)
-{
- return RTE_SCHED_QUEUES_PER_PIPE * port->n_pipes_per_subport;
-}
-
-#endif
-
static inline uint32_t
rte_sched_subport_queues(struct rte_sched_subport *subport)
{
@@ -340,9 +330,14 @@ rte_sched_subport_qsize(struct rte_sched_subport *subport,
uint32_t qindex)
}
static inline uint32_t
-rte_sched_port_queues_per_port(struct rte_sched_port *port)
+rte_sched_port_queues(struct rte_sched_port *port)
{
- return RTE_SCHED_QUEUES_PER_PIPE * port->n_pipes_per_subport *
port->n_subports_per_port;
+ uint32_t n_queues = 0, i;
+
+ for (i = 0; i < port->n_subports_per_port; i++)
+ n_queues += rte_sched_subport_queues(port->subports[i]);
+
+ return n_queues;
}
static int
@@ -1366,18 +1361,25 @@ rte_sched_queue_read_stats(struct rte_sched_port *port,
struct rte_sched_queue_stats *stats,
uint16_t *qlen)
{
+ uint32_t subport_id, qindex;
+ struct rte_sched_subport *s;
struct rte_sched_queue *q;
struct rte_sched_queue_extra *qe;
/* Check user parameters */
if ((port == NULL) ||
- (queue_id >= rte_sched_port_queues_per_port(port)) ||
+ (queue_id >= rte_sched_port_queues(port)) ||
(stats == NULL) ||
(qlen == NULL)) {
return -1;
}
- q = port->queue + queue_id;
- qe = port->queue_extra + queue_id;
+
+ subport_id = (queue_id >> (port->n_max_subport_pipes_log2 + 4)) &
+ (port->n_subports_per_port - 1);
+ s = port->subports[subport_id];
+ qindex = ((1 << (port->n_max_subport_pipes_log2 + 4)) - 1) & queue_id;
+ q = s->queue + qindex;
+ qe = s->queue_extra + qindex;
/* Copy queue stats and clear */
memcpy(stats, &qe->stats, sizeof(struct rte_sched_queue_stats));
@@ -1392,9 +1394,10 @@ rte_sched_queue_read_stats(struct rte_sched_port *port,
#ifdef RTE_SCHED_DEBUG
static inline int
-rte_sched_port_queue_is_empty(struct rte_sched_port *port, uint32_t qindex)
+rte_sched_port_queue_is_empty(struct rte_sched_subport *subport,
+ uint32_t qindex)
{
- struct rte_sched_queue *queue = port->queue + qindex;
+ struct rte_sched_queue *queue = subport->queue + qindex;
return queue->qr == queue->qw;
}
@@ -1535,7 +1538,7 @@ rte_sched_port_set_queue_empty_timestamp(struct
rte_sched_port *port __rte_unuse
#ifdef RTE_SCHED_DEBUG
static inline void
-debug_check_queue_slab(struct rte_sched_port *port, uint32_t bmp_pos,
+debug_check_queue_slab(struct rte_sched_subport *subport, uint32_t bmp_pos,
uint64_t bmp_slab)
{
uint64_t mask;
@@ -1547,7 +1550,7 @@ debug_check_queue_slab(struct rte_sched_port *port,
uint32_t bmp_pos,
panic = 0;
for (i = 0, mask = 1; i < 64; i++, mask <<= 1) {
if (mask & bmp_slab) {
- if (rte_sched_port_queue_is_empty(port, bmp_pos + i)) {
+ if (rte_sched_port_queue_is_empty(subport, bmp_pos +
i)) {
printf("Queue %u (slab offset %u) is empty\n",
bmp_pos + i, i);
panic = 1;
}
--
2.20.1
[dpdk-dev] [PATCH 17/27] sched: modify credits update function
From: Jasvinder Singh
Modify credits update function of the scheduler grinder to allow configuration
flexiblity for pipe traffic classes and queues, and subport level
configuration of the pipe parameters.
Signed-off-by: Jasvinder Singh
Signed-off-by: Abraham Tovar
Signed-off-by: Lukasz Krakowiak
---
lib/librte_sched/rte_sched.c | 87 +---
1 file changed, 52 insertions(+), 35 deletions(-)
diff --git a/lib/librte_sched/rte_sched.c b/lib/librte_sched/rte_sched.c
index a9b5f7bf8..ba344f0a1 100644
--- a/lib/librte_sched/rte_sched.c
+++ b/lib/librte_sched/rte_sched.c
@@ -1857,13 +1857,15 @@ rte_sched_port_enqueue(struct rte_sched_port *port,
struct rte_mbuf **pkts,
#ifndef RTE_SCHED_SUBPORT_TC_OV
static inline void
-grinder_credits_update(struct rte_sched_port *port, uint32_t pos)
+grinder_credits_update(struct rte_sched_port *port,
+ struct rte_sched_subport *subport, uint32_t pos)
{
- struct rte_sched_grinder *grinder = port->grinder + pos;
- struct rte_sched_subport *subport = grinder->subport;
+ struct rte_sched_grinder *grinder = subport->grinder + pos;
struct rte_sched_pipe *pipe = grinder->pipe;
struct rte_sched_pipe_profile *params = grinder->pipe_params;
+ uint32_t n_sp_queues = pipe->n_sp_queues;
uint64_t n_periods;
+ uint32_t i;
/* Subport TB */
n_periods = (port->time - subport->tb_time) / subport->tb_period;
@@ -1879,19 +1881,23 @@ grinder_credits_update(struct rte_sched_port *port,
uint32_t pos)
/* Subport TCs */
if (unlikely(port->time >= subport->tc_time)) {
- subport->tc_credits[0] = subport->tc_credits_per_period[0];
- subport->tc_credits[1] = subport->tc_credits_per_period[1];
- subport->tc_credits[2] = subport->tc_credits_per_period[2];
- subport->tc_credits[3] = subport->tc_credits_per_period[3];
+ for (i = 0; i < n_sp_queues; i++)
+ subport->tc_credits[i] =
subport->tc_credits_per_period[i];
+
+ subport->tc_credits[RTE_SCHED_TRAFFIC_CLASS_BE] =
+
subport->tc_credits_per_period[RTE_SCHED_TRAFFIC_CLASS_BE];
+
subport->tc_time = port->time + subport->tc_period;
}
/* Pipe TCs */
if (unlikely(port->time >= pipe->tc_time)) {
- pipe->tc_credits[0] = params->tc_credits_per_period[0];
- pipe->tc_credits[1] = params->tc_credits_per_period[1];
- pipe->tc_credits[2] = params->tc_credits_per_period[2];
- pipe->tc_credits[3] = params->tc_credits_per_period[3];
+ for (i = 0; i < n_sp_queues; i++)
+ pipe->tc_credits[i] = params->tc_credits_per_period[i];
+
+ pipe->tc_credits[RTE_SCHED_TRAFFIC_CLASS_BE] =
+
params->tc_credits_per_period[RTE_SCHED_TRAFFIC_CLASS_BE];
+
pipe->tc_time = port->time + params->tc_period;
}
}
@@ -1899,26 +1905,34 @@ grinder_credits_update(struct rte_sched_port *port,
uint32_t pos)
#else
static inline uint32_t
-grinder_tc_ov_credits_update(struct rte_sched_port *port, uint32_t pos)
+grinder_tc_ov_credits_update(struct rte_sched_port *port,
+ struct rte_sched_subport *subport, uint32_t pos)
{
- struct rte_sched_grinder *grinder = port->grinder + pos;
- struct rte_sched_subport *subport = grinder->subport;
+ struct rte_sched_grinder *grinder = subport->grinder + pos;
+ struct rte_sched_pipe *pipe = grinder->pipe;
uint32_t tc_ov_consumption[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE];
- uint32_t tc_ov_consumption_max;
+ uint32_t tc_consumption = 0, tc_ov_consumption_max;
uint32_t tc_ov_wm = subport->tc_ov_wm;
+ uint32_t n_sp_queues = pipe->n_sp_queues;
+ uint32_t i;
if (subport->tc_ov == 0)
return subport->tc_ov_wm_max;
+ for (i = 0; i < n_sp_queues; i++) {
+ tc_ov_consumption[i] =
+ subport->tc_credits_per_period[i] -
subport->tc_credits[i];
+ tc_consumption += tc_ov_consumption[i];
+ }
- tc_ov_consumption[0] = subport->tc_credits_per_period[0] -
subport->tc_credits[0];
- tc_ov_consumption[1] = subport->tc_credits_per_period[1] -
subport->tc_credits[1];
- tc_ov_consumption[2] = subport->tc_credits_per_period[2] -
subport->tc_credits[2];
- tc_ov_consumption[3] = subport->tc_credits_per_period[3] -
subport->tc_credits[3];
+ tc_ov_consumption[RTE_SCHED_TRAFFIC_CLASS_BE] =
+ subport->tc_credits_per_period[RTE_SCHED_TRAFFIC_CLASS_BE] -
+ subport->tc_credits[RTE_SCHED_TRAFFIC_CLASS_BE];
- tc_ov_consumption_max = subport->tc_credits_per_period[3] -
- (tc_ov_consumption[0] + tc_ov_consumption[1] +
tc_ov_consumption[2]);
+ tc_ov_consumption_max =
+ subport->tc_credits_per_pe
[dpdk-dev] [PATCH 20/27] sched: update grinder handle function
From: Jasvinder Singh
Update grinder handle function implementation of the scheduler to allow
configuration flexiblity for pipe traffic classes and queues, and subport
level configuration of the pipe parameters.
Signed-off-by: Jasvinder Singh
Signed-off-by: Abraham Tovar
Signed-off-by: Lukasz Krakowiak
---
lib/librte_sched/rte_sched.c | 49 +++-
1 file changed, 26 insertions(+), 23 deletions(-)
diff --git a/lib/librte_sched/rte_sched.c b/lib/librte_sched/rte_sched.c
index 34b85af20..0c5479426 100644
--- a/lib/librte_sched/rte_sched.c
+++ b/lib/librte_sched/rte_sched.c
@@ -2495,7 +2495,7 @@ grinder_wrr(struct rte_sched_subport *subport, uint32_t
pos)
}
-#define grinder_evict(port, pos)
+#define grinder_evict(subport, pos)
static inline void
grinder_prefetch_pipe(struct rte_sched_subport *subport, uint32_t pos)
@@ -2507,9 +2507,9 @@ grinder_prefetch_pipe(struct rte_sched_subport *subport,
uint32_t pos)
}
static inline void
-grinder_prefetch_tc_queue_arrays(struct rte_sched_port *port, uint32_t pos)
+grinder_prefetch_tc_queue_arrays(struct rte_sched_subport *subport, uint32_t
pos)
{
- struct rte_sched_grinder *grinder = port->subport->grinder + pos;
+ struct rte_sched_grinder *grinder = subport->grinder + pos;
struct rte_sched_pipe *pipe = grinder->pipe;
struct rte_sched_queue *queue;
uint32_t tc_index = grinder->tc_index, i;
@@ -2532,8 +2532,8 @@ grinder_prefetch_tc_queue_arrays(struct rte_sched_port
*port, uint32_t pos)
rte_prefetch0(grinder->be.qbase[i] + qr[i]);
}
- grinder_wrr_load(port->subport, pos);
- grinder_wrr(port->subport, pos);
+ grinder_wrr_load(subport, pos);
+ grinder_wrr(subport, pos);
}
static inline void
@@ -2580,9 +2580,9 @@ grinder_prefetch_mbuf(struct rte_sched_subport *subport,
uint32_t pos)
}
static inline uint32_t
-grinder_handle(struct rte_sched_port *port, uint32_t pos)
+grinder_handle(struct rte_sched_port *port,
+ struct rte_sched_subport *subport, uint32_t pos)
{
- struct rte_sched_subport *subport = port->subport;
struct rte_sched_grinder *grinder = subport->grinder + pos;
switch (grinder->state) {
@@ -2603,9 +2603,9 @@ grinder_handle(struct rte_sched_port *port, uint32_t pos)
{
struct rte_sched_pipe *pipe = grinder->pipe;
- grinder->pipe_params = port->pipe_profiles + pipe->profile;
- grinder_prefetch_tc_queue_arrays(port, pos);
- grinder_credits_update(port, port->subport, pos);
+ grinder->pipe_params = subport->pipe_profiles + pipe->profile;
+ grinder_prefetch_tc_queue_arrays(subport, pos);
+ grinder_credits_update(port, subport, pos);
grinder->state = e_GRINDER_PREFETCH_MBUF;
return 0;
@@ -2626,38 +2626,40 @@ grinder_handle(struct rte_sched_port *port, uint32_t
pos)
result = grinder_schedule(port, subport, pos);
/* Look for next packet within the same TC */
- if (result && grinder->qmask) {
- grinder_wrr(port->subport, pos);
- grinder_prefetch_mbuf(port->subport, pos);
+ if (result &&
+ (grinder->tc_index == RTE_SCHED_TRAFFIC_CLASS_BE) &&
+ (grinder->be.qmask)) {
+ grinder_wrr(subport, pos);
+ grinder_prefetch_mbuf(subport, pos);
return 1;
}
- grinder_wrr_store(port->subport, pos);
+ grinder_wrr_store(subport, pos);
/* Look for another active TC within same pipe */
- if (grinder_next_tc(port->subport, pos)) {
- grinder_prefetch_tc_queue_arrays(port, pos);
+ if (grinder_next_tc(subport, pos)) {
+ grinder_prefetch_tc_queue_arrays(subport, pos);
grinder->state = e_GRINDER_PREFETCH_MBUF;
return result;
}
if (grinder->productive == 0 &&
- port->pipe_loop == RTE_SCHED_PIPE_INVALID)
- port->pipe_loop = grinder->pindex;
+ subport->pipe_loop == RTE_SCHED_PIPE_INVALID)
+ subport->pipe_loop = grinder->pindex;
- grinder_evict(port, pos);
+ grinder_evict(subport, pos);
/* Look for another active pipe */
- if (grinder_next_pipe(port->subport, pos)) {
- grinder_prefetch_pipe(port->subport, pos);
+ if (grinder_next_pipe(subport, pos)) {
+ grinder_prefetch_pipe(subport, pos);
grinder->state = e_GRINDER_PREFETCH_TC_QUEUE_ARRAYS;
return result;
}
/* No active pi
[dpdk-dev] [PATCH 19/27] sched: update grinder schedule function
From: Jasvinder Singh
Update grinder schedule function of the scheduler to allow configuration
flexiblity for pipe traffic classes and queues, and subport level
configuration of the pipe parameters.
Signed-off-by: Jasvinder Singh
Signed-off-by: Abraham Tovar
Signed-off-by: Lukasz Krakowiak
---
lib/librte_sched/rte_sched.c | 120 ++-
1 file changed, 90 insertions(+), 30 deletions(-)
diff --git a/lib/librte_sched/rte_sched.c b/lib/librte_sched/rte_sched.c
index 06d89e3fd..34b85af20 100644
--- a/lib/librte_sched/rte_sched.c
+++ b/lib/librte_sched/rte_sched.c
@@ -2007,10 +2007,10 @@ grinder_credits_update(struct rte_sched_port *port,
#ifndef RTE_SCHED_SUBPORT_TC_OV
static inline int
-grinder_credits_check(struct rte_sched_port *port, uint32_t pos)
+grinder_credits_check(struct rte_sched_port *port,
+ struct rte_sched_subport *subport, uint32_t pos)
{
- struct rte_sched_grinder *grinder = port->grinder + pos;
- struct rte_sched_subport *subport = grinder->subport;
+ struct rte_sched_grinder *grinder = subport->grinder + pos;
struct rte_sched_pipe *pipe = grinder->pipe;
struct rte_mbuf *pkt = grinder->pkt;
uint32_t tc_index = grinder->tc_index;
@@ -2030,7 +2030,7 @@ grinder_credits_check(struct rte_sched_port *port,
uint32_t pos)
if (!enough_credits)
return 0;
- /* Update port credits */
+ /* Update subport credits */
subport->tb_credits -= pkt_len;
subport->tc_credits[tc_index] -= pkt_len;
pipe->tb_credits -= pkt_len;
@@ -2042,10 +2042,10 @@ grinder_credits_check(struct rte_sched_port *port,
uint32_t pos)
#else
static inline int
-grinder_credits_check(struct rte_sched_port *port, uint32_t pos)
+grinder_credits_check(struct rte_sched_port *port,
+ struct rte_sched_subport *subport, uint32_t pos)
{
- struct rte_sched_grinder *grinder = port->grinder + pos;
- struct rte_sched_subport *subport = grinder->subport;
+ struct rte_sched_grinder *grinder = subport->grinder + pos;
struct rte_sched_pipe *pipe = grinder->pipe;
struct rte_mbuf *pkt = grinder->pkt;
uint32_t tc_index = grinder->tc_index;
@@ -2054,11 +2054,18 @@ grinder_credits_check(struct rte_sched_port *port,
uint32_t pos)
uint32_t subport_tc_credits = subport->tc_credits[tc_index];
uint32_t pipe_tb_credits = pipe->tb_credits;
uint32_t pipe_tc_credits = pipe->tc_credits[tc_index];
- uint32_t pipe_tc_ov_mask1[] = {UINT32_MAX, UINT32_MAX, UINT32_MAX,
pipe->tc_ov_credits};
- uint32_t pipe_tc_ov_mask2[] = {0, 0, 0, UINT32_MAX};
- uint32_t pipe_tc_ov_credits = pipe_tc_ov_mask1[tc_index];
+ uint32_t pipe_tc_ov_mask1[RTE_SCHED_TRAFFIC_CLASS_BE];
+ uint32_t pipe_tc_ov_mask2[RTE_SCHED_TRAFFIC_CLASS_BE] = {0};
+ uint32_t pipe_tc_ov_credits, i;
int enough_credits;
+ for (i = 0; i < RTE_DIM(pipe_tc_ov_mask1); i++)
+ pipe_tc_ov_mask1[i] = UINT32_MAX;
+
+ pipe_tc_ov_mask1[RTE_SCHED_TRAFFIC_CLASS_BE] = pipe->tc_ov_credits;
+ pipe_tc_ov_mask2[RTE_SCHED_TRAFFIC_CLASS_BE] = UINT32_MAX;
+ pipe_tc_ov_credits = pipe_tc_ov_mask1[tc_index];
+
/* Check pipe and subport credits */
enough_credits = (pkt_len <= subport_tb_credits) &&
(pkt_len <= subport_tc_credits) &&
@@ -2081,17 +2088,47 @@ grinder_credits_check(struct rte_sched_port *port,
uint32_t pos)
#endif /* RTE_SCHED_SUBPORT_TC_OV */
-
-static inline int
-grinder_schedule(struct rte_sched_port *port, uint32_t pos)
+static inline void
+grinder_schedule_sp(struct rte_sched_port *port,
+ struct rte_sched_subport *subport, uint32_t pos)
{
- struct rte_sched_grinder *grinder = port->grinder + pos;
- struct rte_sched_queue *queue = grinder->queue[grinder->qpos];
+ struct rte_sched_grinder *grinder = subport->grinder + pos;
+ struct rte_sched_queue *queue = grinder->sp.queue;
struct rte_mbuf *pkt = grinder->pkt;
uint32_t pkt_len = pkt->pkt_len + port->frame_overhead;
- if (!grinder_credits_check(port, pos))
- return 0;
+ /* Advance port time */
+ port->time += pkt_len;
+
+ /* Send packet */
+ port->pkts_out[port->n_pkts_out++] = pkt;
+ queue->qr++;
+
+ if (queue->qr == queue->qw) {
+ uint32_t qindex = grinder->sp.qindex;
+
+ rte_bitmap_clear(subport->bmp, qindex);
+ rte_sched_port_set_queue_empty_timestamp(port, subport, qindex);
+ }
+
+ /* Reset pipe loop detection */
+ subport->pipe_loop = RTE_SCHED_PIPE_INVALID;
+ grinder->productive = 1;
+}
+
+static inline void
+grinder_schedule_be(struct rte_sched_port *port,
+ struct rte_sched_subport *subport, uint32_t pos)
+{
+ struct rte_sched_grinder *grinder = subport->grinder + pos;
+ struct rte_mbuf *pkt = grinder->pkt;
+ struct rte_sched_queue
[dpdk-dev] [PATCH 23/27] test/sched: update unit test
From: Jasvinder Singh
Update unit test to allow configuration flexiblity for pipe traffic
classes and queues, and subport level configuration of the pipe
parameters.
Signed-off-by: Jasvinder Singh
Signed-off-by: Abraham Tovar
Signed-off-by: Lukasz Krakowiak
---
app/test/test_sched.c | 37 -
1 file changed, 20 insertions(+), 17 deletions(-)
diff --git a/app/test/test_sched.c b/app/test/test_sched.c
index 460eb53ec..ec94861da 100644
--- a/app/test/test_sched.c
+++ b/app/test/test_sched.c
@@ -20,40 +20,43 @@
#define SUBPORT 0
#define PIPE1
#define TC 2
-#define QUEUE 3
-
-static struct rte_sched_subport_params subport_param[] = {
- {
- .tb_rate = 125000,
- .tb_size = 100,
-
- .tc_rate = {125000, 125000, 125000, 125000},
- .tc_period = 10,
- },
-};
+#define QUEUE 2
static struct rte_sched_pipe_params pipe_profile[] = {
{ /* Profile #0 */
.tb_rate = 305175,
.tb_size = 100,
- .tc_rate = {305175, 305175, 305175, 305175},
+ .tc_rate = {305175, 305175, 305175, 305175,
+ 305175, 305175, 305175, 305175, 305175},
.tc_period = 40,
.wrr_weights = {1, 1, 1, 1, 1, 1, 1, 1},
},
};
+static struct rte_sched_subport_params subport_param[] = {
+ {
+ .tb_rate = 125000,
+ .tb_size = 100,
+
+ .tc_rate = {125000, 125000, 125000, 125000,
+ 125000, 125000, 125000, 125000,
125000},
+ .tc_period = 10,
+ .n_subport_pipes = 1024,
+ .qsize = {32, 32, 32, 32, 32, 32, 32, 32,
+ 32, 32, 32, 32, 32, 32, 32, 32},
+ .pipe_profiles = pipe_profile,
+ .n_pipe_profiles = 1,
+ },
+};
+
static struct rte_sched_port_params port_param = {
.socket = 0, /* computed */
.rate = 0, /* computed */
.mtu = 1522,
.frame_overhead = RTE_SCHED_FRAME_OVERHEAD_DEFAULT,
.n_subports_per_port = 1,
- .n_pipes_per_subport = 1024,
- .qsize = {32, 32, 32, 32},
- .pipe_profiles = pipe_profile,
- .n_pipe_profiles = 1,
};
#define NB_MBUF 32
@@ -131,7 +134,7 @@ test_sched(void)
err = rte_sched_subport_config(port, SUBPORT, subport_param);
TEST_ASSERT_SUCCESS(err, "Error config sched, err=%d\n", err);
- for (pipe = 0; pipe < port_param.n_pipes_per_subport; pipe ++) {
+ for (pipe = 0; pipe < subport_param[0].n_subport_pipes; pipe++) {
err = rte_sched_pipe_config(port, SUBPORT, pipe, 0);
TEST_ASSERT_SUCCESS(err, "Error config sched pipe %u,
err=%d\n", pipe, err);
}
--
2.20.1
[dpdk-dev] [PATCH 24/27] net/softnic: update softnic tm function
From: Jasvinder Singh
Update softnic tm function to allow configuration flexiblity for
pipe traffic classes and queues, and subport level configuration
of the pipe parameters.
Signed-off-by: Jasvinder Singh
Signed-off-by: Abraham Tovar
Signed-off-by: Lukasz Krakowiak
---
drivers/net/softnic/rte_eth_softnic.c | 131
drivers/net/softnic/rte_eth_softnic_cli.c | 286 --
.../net/softnic/rte_eth_softnic_internals.h | 4 +-
drivers/net/softnic/rte_eth_softnic_tm.c | 89 +++---
4 files changed, 442 insertions(+), 68 deletions(-)
diff --git a/drivers/net/softnic/rte_eth_softnic.c
b/drivers/net/softnic/rte_eth_softnic.c
index 32b001fd3..9d0168549 100644
--- a/drivers/net/softnic/rte_eth_softnic.c
+++ b/drivers/net/softnic/rte_eth_softnic.c
@@ -28,6 +28,19 @@
#define PMD_PARAM_TM_QSIZE1"tm_qsize1"
#define PMD_PARAM_TM_QSIZE2"tm_qsize2"
#define PMD_PARAM_TM_QSIZE3"tm_qsize3"
+#define PMD_PARAM_TM_QSIZE4"tm_qsize4"
+#define PMD_PARAM_TM_QSIZE5"tm_qsize5"
+#define PMD_PARAM_TM_QSIZE6"tm_qsize6"
+#define PMD_PARAM_TM_QSIZE7"tm_qsize7"
+#define PMD_PARAM_TM_QSIZE8"tm_qsize8"
+#define PMD_PARAM_TM_QSIZE9"tm_qsize9"
+#define PMD_PARAM_TM_QSIZE10 "tm_qsize10"
+#define PMD_PARAM_TM_QSIZE11 "tm_qsize11"
+#define PMD_PARAM_TM_QSIZE12 "tm_qsize12"
+#define PMD_PARAM_TM_QSIZE13 "tm_qsize13"
+#define PMD_PARAM_TM_QSIZE14 "tm_qsize14"
+#define PMD_PARAM_TM_QSIZE15 "tm_qsize15"
+
static const char * const pmd_valid_args[] = {
PMD_PARAM_FIRMWARE,
@@ -39,6 +52,18 @@ static const char * const pmd_valid_args[] = {
PMD_PARAM_TM_QSIZE1,
PMD_PARAM_TM_QSIZE2,
PMD_PARAM_TM_QSIZE3,
+ PMD_PARAM_TM_QSIZE4,
+ PMD_PARAM_TM_QSIZE5,
+ PMD_PARAM_TM_QSIZE6,
+ PMD_PARAM_TM_QSIZE7,
+ PMD_PARAM_TM_QSIZE8,
+ PMD_PARAM_TM_QSIZE9,
+ PMD_PARAM_TM_QSIZE10,
+ PMD_PARAM_TM_QSIZE11,
+ PMD_PARAM_TM_QSIZE12,
+ PMD_PARAM_TM_QSIZE13,
+ PMD_PARAM_TM_QSIZE14,
+ PMD_PARAM_TM_QSIZE15,
NULL
};
@@ -434,6 +459,18 @@ pmd_parse_args(struct pmd_params *p, const char *params)
p->tm.qsize[1] = SOFTNIC_TM_QUEUE_SIZE;
p->tm.qsize[2] = SOFTNIC_TM_QUEUE_SIZE;
p->tm.qsize[3] = SOFTNIC_TM_QUEUE_SIZE;
+ p->tm.qsize[4] = SOFTNIC_TM_QUEUE_SIZE;
+ p->tm.qsize[5] = SOFTNIC_TM_QUEUE_SIZE;
+ p->tm.qsize[6] = SOFTNIC_TM_QUEUE_SIZE;
+ p->tm.qsize[7] = SOFTNIC_TM_QUEUE_SIZE;
+ p->tm.qsize[8] = SOFTNIC_TM_QUEUE_SIZE;
+ p->tm.qsize[9] = SOFTNIC_TM_QUEUE_SIZE;
+ p->tm.qsize[10] = SOFTNIC_TM_QUEUE_SIZE;
+ p->tm.qsize[11] = SOFTNIC_TM_QUEUE_SIZE;
+ p->tm.qsize[12] = SOFTNIC_TM_QUEUE_SIZE;
+ p->tm.qsize[13] = SOFTNIC_TM_QUEUE_SIZE;
+ p->tm.qsize[14] = SOFTNIC_TM_QUEUE_SIZE;
+ p->tm.qsize[15] = SOFTNIC_TM_QUEUE_SIZE;
/* Firmware script (optional) */
if (rte_kvargs_count(kvlist, PMD_PARAM_FIRMWARE) == 1) {
@@ -504,6 +541,88 @@ pmd_parse_args(struct pmd_params *p, const char *params)
goto out_free;
}
+ if (rte_kvargs_count(kvlist, PMD_PARAM_TM_QSIZE4) == 1) {
+ ret = rte_kvargs_process(kvlist, PMD_PARAM_TM_QSIZE4,
+ &get_uint32, &p->tm.qsize[4]);
+ if (ret < 0)
+ goto out_free;
+ }
+
+ if (rte_kvargs_count(kvlist, PMD_PARAM_TM_QSIZE5) == 1) {
+ ret = rte_kvargs_process(kvlist, PMD_PARAM_TM_QSIZE5,
+ &get_uint32, &p->tm.qsize[5]);
+ if (ret < 0)
+ goto out_free;
+ }
+
+ if (rte_kvargs_count(kvlist, PMD_PARAM_TM_QSIZE6) == 1) {
+ ret = rte_kvargs_process(kvlist, PMD_PARAM_TM_QSIZE6,
+ &get_uint32, &p->tm.qsize[6]);
+ if (ret < 0)
+ goto out_free;
+ }
+
+ if (rte_kvargs_count(kvlist, PMD_PARAM_TM_QSIZE7) == 1) {
+ ret = rte_kvargs_process(kvlist, PMD_PARAM_TM_QSIZE7,
+ &get_uint32, &p->tm.qsize[7]);
+ if (ret < 0)
+ goto out_free;
+ }
+ if (rte_kvargs_count(kvlist, PMD_PARAM_TM_QSIZE8) == 1) {
+ ret = rte_kvargs_process(kvlist, PMD_PARAM_TM_QSIZE8,
+ &get_uint32, &p->tm.qsize[8]);
+ if (ret < 0)
+ goto out_free;
+ }
+ if (rte_kvargs_count(kvlist, PMD_PARAM_TM_QSIZE9) == 1) {
+ ret = rte_kvargs_p
[dpdk-dev] [PATCH 25/27] examples/qos_sched: update qos sched sample app
From: Jasvinder Singh
Update qos sched sample app to allow configuration flexibility for
pipe traffic classes and queues, and subport level configuration
of the pipe parameters.
Signed-off-by: Jasvinder Singh
Signed-off-by: Abraham Tovar
Signed-off-by: Lukasz Krakowiak
---
examples/qos_sched/app_thread.c | 11 +-
examples/qos_sched/cfg_file.c | 282 +-
examples/qos_sched/init.c | 108 +++-
examples/qos_sched/main.h | 6 +-
examples/qos_sched/profile.cfg| 59 +--
examples/qos_sched/profile_ov.cfg | 47 -
examples/qos_sched/stats.c| 175 +++---
7 files changed, 432 insertions(+), 256 deletions(-)
diff --git a/examples/qos_sched/app_thread.c b/examples/qos_sched/app_thread.c
index e14b275e3..25a8d42a0 100644
--- a/examples/qos_sched/app_thread.c
+++ b/examples/qos_sched/app_thread.c
@@ -20,13 +20,11 @@
* QoS parameters are encoded as follows:
* Outer VLAN ID defines subport
* Inner VLAN ID defines pipe
- * Destination IP 0.0.XXX.0 defines traffic class
* Destination IP host (0.0.0.XXX) defines queue
* Values below define offset to each field from start of frame
*/
#define SUBPORT_OFFSET 7
#define PIPE_OFFSET9
-#define TC_OFFSET 20
#define QUEUE_OFFSET 20
#define COLOR_OFFSET 19
@@ -39,11 +37,10 @@ get_pkt_sched(struct rte_mbuf *m, uint32_t *subport,
uint32_t *pipe,
*subport = (rte_be_to_cpu_16(pdata[SUBPORT_OFFSET]) & 0x0FFF) &
(port_params.n_subports_per_port - 1); /* Outer VLAN
ID*/
*pipe = (rte_be_to_cpu_16(pdata[PIPE_OFFSET]) & 0x0FFF) &
- (port_params.n_pipes_per_subport - 1); /* Inner VLAN ID
*/
- *traffic_class = (pdata[QUEUE_OFFSET] & 0x0F) &
- (RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE - 1); /*
Destination IP */
- *queue = ((pdata[QUEUE_OFFSET] >> 8) & 0x0F) &
- (RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS - 1) ; /*
Destination IP */
+ (subport_params[*subport].n_subport_pipes - 1); /*
Inner VLAN ID */
+ *queue = active_queues[(pdata[QUEUE_OFFSET] >> 8) % n_active_queues];
+ *traffic_class = (*queue > (RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE - 1) ?
+ (RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE - 1) : *queue); /*
Destination IP */
*color = pdata[COLOR_OFFSET] & 0x03;/* Destination IP */
return 0;
diff --git a/examples/qos_sched/cfg_file.c b/examples/qos_sched/cfg_file.c
index 76c4b..6d3674e53 100644
--- a/examples/qos_sched/cfg_file.c
+++ b/examples/qos_sched/cfg_file.c
@@ -24,7 +24,6 @@ int
cfg_load_port(struct rte_cfgfile *cfg, struct rte_sched_port_params
*port_params)
{
const char *entry;
- int j;
if (!cfg || !port_params)
return -1;
@@ -37,93 +36,6 @@ cfg_load_port(struct rte_cfgfile *cfg, struct
rte_sched_port_params *port_params
if (entry)
port_params->n_subports_per_port = (uint32_t)atoi(entry);
- entry = rte_cfgfile_get_entry(cfg, "port", "number of pipes per
subport");
- if (entry)
- port_params->n_pipes_per_subport = (uint32_t)atoi(entry);
-
- entry = rte_cfgfile_get_entry(cfg, "port", "queue sizes");
- if (entry) {
- char *next;
-
- for(j = 0; j < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; j++) {
- port_params->qsize[j] = (uint16_t)strtol(entry, &next,
10);
- if (next == NULL)
- break;
- entry = next;
- }
- }
-
-#ifdef RTE_SCHED_RED
- for (j = 0; j < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; j++) {
- char str[32];
-
- /* Parse WRED min thresholds */
- snprintf(str, sizeof(str), "tc %d wred min", j);
- entry = rte_cfgfile_get_entry(cfg, "red", str);
- if (entry) {
- char *next;
- int k;
- /* for each packet colour (green, yellow, red) */
- for (k = 0; k < RTE_COLORS; k++) {
- port_params->red_params[j][k].min_th
- = (uint16_t)strtol(entry, &next, 10);
- if (next == NULL)
- break;
- entry = next;
- }
- }
-
- /* Parse WRED max thresholds */
- snprintf(str, sizeof(str), "tc %d wred max", j);
- entry = rte_cfgfile_get_entry(cfg, "red", str);
- if (entry) {
- char *next;
- int k;
- /* for each packet colour (green, yellow, red) */
- for (k = 0; k < RTE_COLORS; k++) {
-
[dpdk-dev] [PATCH 26/27] examples/ip_pipeline: update ip pipeline sample app
From: Jasvinder Singh
Update ip pipeline sample app to allow configuration flexiblity
for pipe traffic classes and queues, and subport level configuration
of the pipe parameters.
Signed-off-by: Jasvinder Singh
Signed-off-by: Abraham Tovar
Signed-off-by: Lukasz Krakowiak
---
examples/ip_pipeline/cli.c | 85 +-
examples/ip_pipeline/tmgr.c| 21 +++
examples/ip_pipeline/tmgr.h| 3 -
lib/librte_pipeline/rte_table_action.c | 1 -
lib/librte_pipeline/rte_table_action.h | 4 +-
5 files changed, 53 insertions(+), 61 deletions(-)
diff --git a/examples/ip_pipeline/cli.c b/examples/ip_pipeline/cli.c
index bcf62fbf5..2f0ee2ec1 100644
--- a/examples/ip_pipeline/cli.c
+++ b/examples/ip_pipeline/cli.c
@@ -377,8 +377,11 @@ cmd_swq(char **tokens,
static const char cmd_tmgr_subport_profile_help[] =
"tmgr subport profile\n"
"\n"
-" \n"
-" \n";
+" "
+"\n"
+" \n"
+" pps \n"
+" qsize ";
static void
cmd_tmgr_subport_profile(char **tokens,
@@ -389,7 +392,7 @@ cmd_tmgr_subport_profile(char **tokens,
struct rte_sched_subport_params p;
int status, i;
- if (n_tokens != 10) {
+ if (n_tokens != 34) {
snprintf(out, out_size, MSG_ARG_MISMATCH, tokens[0]);
return;
}
@@ -410,11 +413,32 @@ cmd_tmgr_subport_profile(char **tokens,
return;
}
- if (parser_read_uint32(&p.tc_period, tokens[9]) != 0) {
+ if (parser_read_uint32(&p.tc_period, tokens[14]) != 0) {
snprintf(out, out_size, MSG_ARG_INVALID, "tc_period");
return;
}
+ if (strcmp(tokens[15], "pps") != 0) {
+ snprintf(out, out_size, MSG_ARG_NOT_FOUND, "pps");
+ return;
+ }
+
+ if (parser_read_uint32(&p.n_subport_pipes, tokens[16]) != 0) {
+ snprintf(out, out_size, MSG_ARG_INVALID, "n_subport_pipes");
+ return;
+ }
+
+ if (strcmp(tokens[17], "qsize") != 0) {
+ snprintf(out, out_size, MSG_ARG_NOT_FOUND, "qsize");
+ return;
+ }
+
+ for (i = 0; i < RTE_SCHED_QUEUES_PER_PIPE; i++)
+ if (parser_read_uint16(&p.qsize[i], tokens[18 + i]) != 0) {
+ snprintf(out, out_size, MSG_ARG_INVALID, "qsize");
+ return;
+ }
+
status = tmgr_subport_profile_add(&p);
if (status != 0) {
snprintf(out, out_size, MSG_CMD_FAIL, tokens[0]);
@@ -425,10 +449,11 @@ cmd_tmgr_subport_profile(char **tokens,
static const char cmd_tmgr_pipe_profile_help[] =
"tmgr pipe profile\n"
"\n"
-" \n"
+" "
+"\n"
" \n"
" \n"
-" \n";
+" \n";
static void
cmd_tmgr_pipe_profile(char **tokens,
@@ -439,7 +464,7 @@ cmd_tmgr_pipe_profile(char **tokens,
struct rte_sched_pipe_params p;
int status, i;
- if (n_tokens != 27) {
+ if (n_tokens != 24) {
snprintf(out, out_size, MSG_ARG_MISMATCH, tokens[0]);
return;
}
@@ -460,20 +485,20 @@ cmd_tmgr_pipe_profile(char **tokens,
return;
}
- if (parser_read_uint32(&p.tc_period, tokens[9]) != 0) {
+ if (parser_read_uint32(&p.tc_period, tokens[14]) != 0) {
snprintf(out, out_size, MSG_ARG_INVALID, "tc_period");
return;
}
#ifdef RTE_SCHED_SUBPORT_TC_OV
- if (parser_read_uint8(&p.tc_ov_weight, tokens[10]) != 0) {
+ if (parser_read_uint8(&p.tc_ov_weight, tokens[15]) != 0) {
snprintf(out, out_size, MSG_ARG_INVALID, "tc_ov_weight");
return;
}
#endif
- for (i = 0; i < RTE_SCHED_QUEUES_PER_PIPE; i++)
- if (parser_read_uint8(&p.wrr_weights[i], tokens[11 + i]) != 0) {
+ for (i = 0; i < RTE_SCHED_WRR_QUEUES_PER_PIPE; i++)
+ if (parser_read_uint8(&p.wrr_weights[i], tokens[16 + i]) != 0) {
snprintf(out, out_size, MSG_ARG_INVALID, "wrr_weights");
return;
}
@@ -489,8 +514,6 @@ static const char cmd_tmgr_help[] =
"tmgr \n"
" rate \n"
" spp \n"
-" pps \n"
-" qsize\n"
" fo \n"
" mtu \n"
" cpu \n";
@@ -504,9 +527,8 @@ cmd_tmgr(char **tokens,
struct tmgr_port_params p;
char *name;
struct tmgr_port *tmgr_port;
- int i;
- if (n_tokens != 19) {
+ if (n_tokens != 12) {
snprintf(out, out_size, MSG_ARG_MISMATCH, tokens[0]);
return;
}
@@ -533,53 +555,32 @@ cmd_tmgr(char **tokens,
return;
}
- if (strcmp(tokens[6], "pps") != 0) {
- snprintf(out, out_size, MSG_ARG_NOT_FOUND, "pps");
- return;
- }
-
- if (parser_read_uint32(&p.n_pipes_per_subport, tokens[7]) != 0) {
- snprintf(out, out_size, MSG_
[dpdk-dev] [PATCH 27/27] sched: code cleanup
From: Jasvinder Singh
Remove redundant macros and fields from the data structures.
Signed-off-by: Jasvinder Singh
Signed-off-by: Abraham Tovar
Signed-off-by: Lukasz Krakowiak
---
lib/librte_sched/rte_sched.c | 53
lib/librte_sched/rte_sched.h | 18
2 files changed, 71 deletions(-)
diff --git a/lib/librte_sched/rte_sched.c b/lib/librte_sched/rte_sched.c
index 563161713..79731af8e 100644
--- a/lib/librte_sched/rte_sched.c
+++ b/lib/librte_sched/rte_sched.c
@@ -100,18 +100,7 @@ struct rte_sched_grinder {
uint32_t tc_index;
struct rte_sched_strict_priority_class sp;
struct rte_sched_best_effort_class be;
- struct rte_sched_queue *queue[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE];
- struct rte_mbuf **qbase[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE];
- uint32_t qindex[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE];
- uint16_t qsize;
- uint32_t qmask;
- uint32_t qpos;
struct rte_mbuf *pkt;
-
- /* WRR */
- uint16_t wrr_tokens[RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS];
- uint16_t wrr_mask[RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS];
- uint8_t wrr_cost[RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS];
};
struct rte_sched_subport {
@@ -215,7 +204,6 @@ struct rte_sched_pipe {
/* TC oversubscription */
uint32_t tc_ov_credits;
uint8_t tc_ov_period_id;
- uint8_t reserved[3];
} __rte_cache_aligned;
struct rte_sched_queue {
@@ -233,18 +221,10 @@ struct rte_sched_queue_extra {
struct rte_sched_port {
/* User parameters */
uint32_t n_subports_per_port;
- uint32_t n_pipes_per_subport;
- uint32_t n_pipes_per_subport_log2;
int socket;
uint32_t rate;
uint32_t mtu;
uint32_t frame_overhead;
- uint16_t qsize[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE];
- uint32_t n_pipe_profiles;
- uint32_t pipe_tc3_rate_max;
-#ifdef RTE_SCHED_RED
- struct rte_red_config
red_config[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE][RTE_COLORS];
-#endif
/* Timing */
uint64_t time_cpu_cycles; /* Current CPU time measured in CPU cyles
*/
@@ -252,50 +232,17 @@ struct rte_sched_port {
uint64_t time;/* Current NIC TX time measured in bytes
*/
struct rte_reciprocal inv_cycles_per_byte; /* CPU cycles per byte */
- /* Scheduling loop detection */
- uint32_t pipe_loop;
- uint32_t pipe_exhaustion;
-
- /* Bitmap */
- struct rte_bitmap *bmp;
- uint32_t grinder_base_bmp_pos[RTE_SCHED_PORT_N_GRINDERS]
__rte_aligned_16;
-
/* Grinders */
- struct rte_sched_grinder grinder[RTE_SCHED_PORT_N_GRINDERS];
- uint32_t busy_grinders;
struct rte_mbuf **pkts_out;
uint32_t n_pkts_out;
uint32_t subport_id;
uint32_t n_max_subport_pipes_log2; /* Max number of subport pipes */
- /* Queue base calculation */
- uint32_t qsize_add[RTE_SCHED_QUEUES_PER_PIPE];
- uint32_t qsize_sum;
-
/* Large data structures */
- struct rte_sched_subport *subport;
- struct rte_sched_pipe *pipe;
- struct rte_sched_queue *queue;
- struct rte_sched_queue_extra *queue_extra;
- struct rte_sched_pipe_profile *pipe_profiles;
- uint8_t *bmp_array;
- struct rte_mbuf **queue_array;
struct rte_sched_subport *subports[RTE_SCHED_SUBPORTS_PER_PORT];
- uint8_t memory[0] __rte_cache_aligned;
} __rte_cache_aligned;
-enum rte_sched_port_array {
- e_RTE_SCHED_PORT_ARRAY_SUBPORT = 0,
- e_RTE_SCHED_PORT_ARRAY_PIPE,
- e_RTE_SCHED_PORT_ARRAY_QUEUE,
- e_RTE_SCHED_PORT_ARRAY_QUEUE_EXTRA,
- e_RTE_SCHED_PORT_ARRAY_PIPE_PROFILES,
- e_RTE_SCHED_PORT_ARRAY_BMP_ARRAY,
- e_RTE_SCHED_PORT_ARRAY_QUEUE_ARRAY,
- e_RTE_SCHED_PORT_ARRAY_TOTAL,
-};
-
enum rte_sched_subport_array {
e_RTE_SCHED_SUBPORT_ARRAY_PIPE = 0,
e_RTE_SCHED_SUBPORT_ARRAY_QUEUE,
diff --git a/lib/librte_sched/rte_sched.h b/lib/librte_sched/rte_sched.h
index 28b589309..5d6828e2e 100644
--- a/lib/librte_sched/rte_sched.h
+++ b/lib/librte_sched/rte_sched.h
@@ -81,7 +81,6 @@ extern "C" {
#define RTE_SCHED_WRR_QUEUES_PER_PIPE8
/** Number of traffic classes per pipe (as well as subport). */
-#define RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS4
#define RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE\
(RTE_SCHED_QUEUES_PER_PIPE - RTE_SCHED_WRR_QUEUES_PER_PIPE + 1)
@@ -95,10 +94,6 @@ extern "C" {
/** Maximum number of pipe profiles that can be defined per subport.
* Compile-time configurable.
*/
-#ifndef RTE_SCHED_PIPE_PROFILES_PER_PORT
-#define RTE_SCHED_PIPE_PROFILES_PER_PORT 256
-#endif
-
#ifndef RTE_SCHED_PIPE_PROFILES_PER_SUBPORT
#define RTE_SCHED_PIPE_PROFILES_PER_SUBPORT 256
#endif
@@ -229,19 +224,6 @@ struct rte_sched_port_params {
uint32_t frame_overhead; /**< Framing overhead per packet
* (measured in bytes
[dpdk-dev] [PATCH v2 0/2] net/mlx5: remove TCF support from PMD
Today, it is possible to offload an interface flow rules to the hardware using DPDK flow commands. With mlx5 it is also possible to offload a limited set of flow rules to the mlxsw (or E-switch) using the same DPDK flow commands. A 'transfer' attribute was added to the flow rule creation command in order to distinguish between configuring port flows and E-switch flows. The commands destined for the E-switch are transposed to TC-flower rules and are send, as Netlink messages, to the mlx5 driver, or more precisely to the netdev which represent the mlxsw port. With the addition to the PMD of E-switch configuration via DR (direct verbs rules) it is now possible to configure the E-switch using these commands instead of using TC-Flower messages. Doing so will allow us to remove the TCF support and the dependency of libmnl from the PMD. The purpose of this RFC is to propose configuring the E-switch flow filtering using DR, to remove the TCF support from the PMD and to remove the dependency of the PMD in libmnl. As for today VLAN insertion or removal is not supported in DR, this support will be added in separate commits. Moti Haimovsky (2): net/mlx5: fix testpmd crash on null-flow drv opts net/mlx5: remove TCF support from PMD doc/build-sdk-meson.txt |2 +- doc/guides/nics/mlx5.rst | 19 - doc/guides/platform/bluefield.rst |4 - drivers/net/mlx5/Makefile | 303 -- drivers/net/mlx5/meson.build | 123 +- drivers/net/mlx5/mlx5.c | 32 - drivers/net/mlx5/mlx5.h |3 - drivers/net/mlx5/mlx5_flow.c | 43 +- drivers/net/mlx5/mlx5_flow.h | 25 - drivers/net/mlx5/mlx5_flow_tcf.c | 6382 - mk/rte.app.mk |2 +- 11 files changed, 24 insertions(+), 6914 deletions(-) delete mode 100644 drivers/net/mlx5/mlx5_flow_tcf.c -- 1.8.3.1
[dpdk-dev] [PATCH v2 1/2] net/mlx5: fix testpmd crash on null-flow drv opts
mlx5 implements mlx5_flow_null_drv_ops to be used when a specific
flow typei/driver is not available or invalid.
This routines return error without modifying the rte_flow_error
parameter passed to them which causes testpmd, for example, to crash.
This commit addresses the issue by modifying the rte_flow_error
parameter in theses routines.
Fixes: 0c76d1c9a18d ("net/mlx5: add abstraction for multiple flow drivers")
Fixes: 684dafe795d0 ("net/mlx5: add flow query abstraction interface")
Cc: sta...@dpdk.org
Signed-off-by: Moti Haimovsky
---
drivers/net/mlx5/mlx5_flow.c | 29 +++--
1 file changed, 15 insertions(+), 14 deletions(-)
diff --git a/drivers/net/mlx5/mlx5_flow.c b/drivers/net/mlx5/mlx5_flow.c
index 9887018..e5a8e33 100644
--- a/drivers/net/mlx5/mlx5_flow.c
+++ b/drivers/net/mlx5/mlx5_flow.c
@@ -1694,19 +1694,20 @@ uint32_t mlx5_flow_adjust_priority(struct rte_eth_dev
*dev, int32_t priority,
const struct rte_flow_attr *attr __rte_unused,
const struct rte_flow_item items[] __rte_unused,
const struct rte_flow_action actions[] __rte_unused,
- struct rte_flow_error *error __rte_unused)
+ struct rte_flow_error *error)
{
- rte_errno = ENOTSUP;
- return -rte_errno;
+ return rte_flow_error_set(error, ENOTSUP,
+ RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL, NULL);
}
static struct mlx5_flow *
flow_null_prepare(const struct rte_flow_attr *attr __rte_unused,
const struct rte_flow_item items[] __rte_unused,
const struct rte_flow_action actions[] __rte_unused,
- struct rte_flow_error *error __rte_unused)
+ struct rte_flow_error *error)
{
- rte_errno = ENOTSUP;
+ rte_flow_error_set(error, ENOTSUP,
+ RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL, NULL);
return NULL;
}
@@ -1716,19 +1717,19 @@ uint32_t mlx5_flow_adjust_priority(struct rte_eth_dev
*dev, int32_t priority,
const struct rte_flow_attr *attr __rte_unused,
const struct rte_flow_item items[] __rte_unused,
const struct rte_flow_action actions[] __rte_unused,
- struct rte_flow_error *error __rte_unused)
+ struct rte_flow_error *error)
{
- rte_errno = ENOTSUP;
- return -rte_errno;
+ return rte_flow_error_set(error, ENOTSUP,
+ RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL, NULL);
}
static int
flow_null_apply(struct rte_eth_dev *dev __rte_unused,
struct rte_flow *flow __rte_unused,
- struct rte_flow_error *error __rte_unused)
+ struct rte_flow_error *error)
{
- rte_errno = ENOTSUP;
- return -rte_errno;
+ return rte_flow_error_set(error, ENOTSUP,
+ RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL, NULL);
}
static void
@@ -1748,10 +1749,10 @@ uint32_t mlx5_flow_adjust_priority(struct rte_eth_dev
*dev, int32_t priority,
struct rte_flow *flow __rte_unused,
const struct rte_flow_action *actions __rte_unused,
void *data __rte_unused,
- struct rte_flow_error *error __rte_unused)
+ struct rte_flow_error *error)
{
- rte_errno = ENOTSUP;
- return -rte_errno;
+ return rte_flow_error_set(error, ENOTSUP,
+ RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL, NULL);
}
/* Void driver to protect from null pointer reference. */
--
1.8.3.1
[dpdk-dev] [PATCH 11/27] sched: update port memory footprint api
From: Jasvinder Singh
Update port memory footprint api implementation of scheduler to allow
configuration flexiblity for pipe traffic classes and queues, and subport
level configuration of the pipe parameters.
Signed-off-by: Jasvinder Singh
Signed-off-by: Abraham Tovar
Signed-off-by: Lukasz Krakowiak
---
lib/librte_sched/rte_sched.c | 86 +---
lib/librte_sched/rte_sched.h | 7 ++-
2 files changed, 25 insertions(+), 68 deletions(-)
diff --git a/lib/librte_sched/rte_sched.c b/lib/librte_sched/rte_sched.c
index 86f2bdf51..e2a49633d 100644
--- a/lib/librte_sched/rte_sched.c
+++ b/lib/librte_sched/rte_sched.c
@@ -440,66 +440,6 @@ rte_sched_port_check_params(struct rte_sched_port_params
*params)
return 0;
}
-static uint32_t
-rte_sched_port_get_array_base(struct rte_sched_port_params *params, enum
rte_sched_port_array array)
-{
- uint32_t n_subports_per_port = params->n_subports_per_port;
- uint32_t n_pipes_per_subport = params->n_pipes_per_subport;
- uint32_t n_pipes_per_port = n_pipes_per_subport * n_subports_per_port;
- uint32_t n_queues_per_port = RTE_SCHED_QUEUES_PER_PIPE *
n_pipes_per_subport * n_subports_per_port;
-
- uint32_t size_subport = n_subports_per_port * sizeof(struct
rte_sched_subport);
- uint32_t size_pipe = n_pipes_per_port * sizeof(struct rte_sched_pipe);
- uint32_t size_queue = n_queues_per_port * sizeof(struct
rte_sched_queue);
- uint32_t size_queue_extra
- = n_queues_per_port * sizeof(struct rte_sched_queue_extra);
- uint32_t size_pipe_profiles
- = RTE_SCHED_PIPE_PROFILES_PER_PORT * sizeof(struct
rte_sched_pipe_profile);
- uint32_t size_bmp_array =
rte_bitmap_get_memory_footprint(n_queues_per_port);
- uint32_t size_per_pipe_queue_array, size_queue_array;
-
- uint32_t base, i;
-
- size_per_pipe_queue_array = 0;
- for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i++) {
- size_per_pipe_queue_array += RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS
- * params->qsize[i] * sizeof(struct rte_mbuf *);
- }
- size_queue_array = n_pipes_per_port * size_per_pipe_queue_array;
-
- base = 0;
-
- if (array == e_RTE_SCHED_PORT_ARRAY_SUBPORT)
- return base;
- base += RTE_CACHE_LINE_ROUNDUP(size_subport);
-
- if (array == e_RTE_SCHED_PORT_ARRAY_PIPE)
- return base;
- base += RTE_CACHE_LINE_ROUNDUP(size_pipe);
-
- if (array == e_RTE_SCHED_PORT_ARRAY_QUEUE)
- return base;
- base += RTE_CACHE_LINE_ROUNDUP(size_queue);
-
- if (array == e_RTE_SCHED_PORT_ARRAY_QUEUE_EXTRA)
- return base;
- base += RTE_CACHE_LINE_ROUNDUP(size_queue_extra);
-
- if (array == e_RTE_SCHED_PORT_ARRAY_PIPE_PROFILES)
- return base;
- base += RTE_CACHE_LINE_ROUNDUP(size_pipe_profiles);
-
- if (array == e_RTE_SCHED_PORT_ARRAY_BMP_ARRAY)
- return base;
- base += RTE_CACHE_LINE_ROUNDUP(size_bmp_array);
-
- if (array == e_RTE_SCHED_PORT_ARRAY_QUEUE_ARRAY)
- return base;
- base += RTE_CACHE_LINE_ROUNDUP(size_queue_array);
-
- return base;
-}
-
static uint32_t
rte_sched_subport_get_array_base(struct rte_sched_subport_params *params,
enum rte_sched_subport_array array)
@@ -899,22 +839,36 @@ rte_sched_subport_get_memory_footprint(struct
rte_sched_port *port,
}
uint32_t
-rte_sched_port_get_memory_footprint(struct rte_sched_port_params *params)
+rte_sched_port_get_memory_footprint(struct rte_sched_port_params *port_params,
+ struct rte_sched_subport_params *subport_params)
{
- uint32_t size0, size1;
+ uint32_t size0 = 0, size1 = 0, i;
int status;
- status = rte_sched_port_check_params(params);
+ status = rte_sched_port_check_params(port_params);
if (status != 0) {
RTE_LOG(NOTICE, SCHED,
- "Port scheduler params check failed (%d)\n", status);
+ "Port scheduler port params check failed (%d)\n",
status);
+
+ return 0;
+ }
+
+ status = rte_sched_subport_check_params(subport_params,
+ port_params->rate);
+ if (status != 0) {
+ RTE_LOG(NOTICE, SCHED,
+ "Port scheduler subport params check failed (%d)\n",
status);
return 0;
}
size0 = sizeof(struct rte_sched_port);
- size1 = rte_sched_port_get_array_base(params,
- e_RTE_SCHED_PORT_ARRAY_TOTAL);
+
+ for (i = 0; i < port_params->n_subports_per_port; i++) {
+ struct rte_sched_subport_params *sp = &subport_params[i];
+ size1 += rte_sched_subport_get_array_base(sp,
+ e_RTE_SCHED_SUBPORT_ARRAY_TOTAL);
+ }
return size0 + size1;
}
diff --git a/lib/librte_sched/rte_sc
[dpdk-dev] [PATCH 12/27] sched: update packet enqueue api
From: Jasvinder Singh
Update packet enqueue api implementation of the scheduler to allow
configuration flexiblity for pipe traffic classes and queues, and subport
level configuration of the pipe parameters.
Signed-off-by: Jasvinder Singh
Signed-off-by: Abraham Tovar
Signed-off-by: Lukasz Krakowiak
---
lib/librte_sched/rte_sched.c | 203 +++
1 file changed, 133 insertions(+), 70 deletions(-)
diff --git a/lib/librte_sched/rte_sched.c b/lib/librte_sched/rte_sched.c
index e2a49633d..6cfa73761 100644
--- a/lib/librte_sched/rte_sched.c
+++ b/lib/librte_sched/rte_sched.c
@@ -1494,7 +1494,11 @@ rte_sched_port_update_queue_stats_on_drop(struct
rte_sched_subport *subport,
#ifdef RTE_SCHED_RED
static inline int
-rte_sched_port_red_drop(struct rte_sched_port *port, struct rte_mbuf *pkt,
uint32_t qindex, uint16_t qlen)
+rte_sched_port_red_drop(struct rte_sched_subport *subport,
+ struct rte_mbuf *pkt,
+ uint32_t qindex,
+ uint16_t qlen,
+ uint64_t time)
{
struct rte_sched_queue_extra *qe;
struct rte_red_config *red_cfg;
@@ -1502,23 +1506,24 @@ rte_sched_port_red_drop(struct rte_sched_port *port,
struct rte_mbuf *pkt, uint3
uint32_t tc_index;
enum rte_color color;
- tc_index = (qindex >> 2) & 0x3;
+ tc_index = rte_mbuf_sched_traffic_class_get(pkt);
color = rte_sched_port_pkt_read_color(pkt);
- red_cfg = &port->red_config[tc_index][color];
+ red_cfg = &subport->red_config[tc_index][color];
if ((red_cfg->min_th | red_cfg->max_th) == 0)
return 0;
- qe = port->queue_extra + qindex;
+ qe = subport->queue_extra + qindex;
red = &qe->red;
- return rte_red_enqueue(red_cfg, red, qlen, port->time);
+ return rte_red_enqueue(red_cfg, red, qlen, time);
}
static inline void
-rte_sched_port_set_queue_empty_timestamp(struct rte_sched_port *port, uint32_t
qindex)
+rte_sched_port_set_queue_empty_timestamp(struct rte_sched_port *port,
+ struct rte_sched_subport *subport, uint32_t qindex)
{
- struct rte_sched_queue_extra *qe = port->queue_extra + qindex;
+ struct rte_sched_queue_extra *qe = subport->queue_extra + qindex;
struct rte_red *red = &qe->red;
rte_red_mark_queue_empty(red, port->time);
@@ -1526,10 +1531,23 @@ rte_sched_port_set_queue_empty_timestamp(struct
rte_sched_port *port, uint32_t q
#else
-#define rte_sched_port_red_drop(port, pkt, qindex, qlen) 0
-
-#define rte_sched_port_set_queue_empty_timestamp(port, qindex)
+static inline int rte_sched_port_red_drop(
+ struct rte_sched_subport *subport __rte_unused,
+ struct rte_mbuf *pkt __rte_unused,
+ uint32_t qindex __rte_unused,
+ uint16_t qlen __rte_unused,
+ uint64_t time __rte_unused)
+{
+ return 0;
+}
+static inline void
+rte_sched_port_set_queue_empty_timestamp(struct rte_sched_port *port
__rte_unused,
+ struct rte_sched_subport *subport __rte_unused,
+ uint32_t qindex __rte_unused)
+{
+ return;
+}
#endif /* RTE_SCHED_RED */
#ifdef RTE_SCHED_DEBUG
@@ -1561,6 +1579,17 @@ debug_check_queue_slab(struct rte_sched_port *port,
uint32_t bmp_pos,
#endif /* RTE_SCHED_DEBUG */
+static inline struct rte_sched_subport *
+rte_sched_port_subport_prefetch0(struct rte_sched_port *port,
+ struct rte_mbuf *pkt)
+{
+ uint32_t qindex = rte_mbuf_sched_queue_get(pkt);
+ uint32_t subport_id = (qindex >> (port->n_max_subport_pipes_log2 + 4)) &
+ (port->n_subports_per_port - 1);
+
+ return port->subports[subport_id];
+}
+
static inline uint32_t
rte_sched_port_enqueue_qptrs_prefetch0(struct rte_sched_port *port,
struct rte_mbuf *pkt)
@@ -1571,53 +1600,55 @@ rte_sched_port_enqueue_qptrs_prefetch0(struct
rte_sched_port *port,
#endif
uint32_t qindex = rte_mbuf_sched_queue_get(pkt);
- q = port->queue + qindex;
+ uint32_t subport_id = (qindex >> (port->n_max_subport_pipes_log2 + 4)) &
+ (port->n_subports_per_port - 1);
+ struct rte_sched_subport *s = port->subports[subport_id];
+ uint32_t queue_id = ((1 << (port->n_max_subport_pipes_log2 + 4)) - 1) &
+ qindex;
+
+ q = s->queue + queue_id;
rte_prefetch0(q);
#ifdef RTE_SCHED_COLLECT_STATS
- qe = port->queue_extra + qindex;
+ qe = s->queue_extra + queue_id;
rte_prefetch0(qe);
#endif
- return qindex;
+ return queue_id;
}
static inline void
-rte_sched_port_enqueue_qwa_prefetch0(struct rte_sched_port *port,
+rte_sched_port_enqueue_qwa_prefetch0(struct rte_sched_subport *subport,
uint32_t qindex, struct rte_mbuf **qbase)
{
struct rte_sched_queue *q;
struct rte_mbuf **q_qw;
uint16_t qsize;
-
Re: [dpdk-dev] [PATCH 0/2] Add ABI Version Testing to app/test
Hi Bruce, There was a bit of a misfire on the patch submission - it came from the wrong email a/c and the ML (rightly) rejected it. Let me submit the patch properly and the feedback can begin in earnest then. Ray K On 28/05/2019 13:08, Bruce Richardson wrote: > On Tue, May 28, 2019 at 12:51:56PM +0100, Ray Kinsella wrote: >> This patchset adds ABI Version Testing functionality to the app/test unit >> test framework. >> >> The patchset is intended to address two issues previously raised during ML >> conversations on ABI Stability; >> 1. How do we unit test still supported previous ABI Versions. >> 2. How to we unit test inline functions from still supported previous ABI >> Versions. >> >> The more obvious way to achieve both of the above is to simply archive >> pre-built binaries compiled against previous versions of DPDK for use unit >> testing previous ABI Versions, and while this should still be done as an >> additional check, this approach does not scale well, must every DPDK >> developer have a local copy of these binaries to test with, before >> upstreaming changes? >> >> Instead starting with rte_lpm, I did the following:- >> >> * I reproduced mostly unmodified unit tests from previous ABI Versions, >> in this case v2.0 and v16.04 >> * I reproduced the rte_lpm interface header from these previous ABI >> Versions,including the inline functions and remapping symbols to >> appropriate versions. >> * I added support for multiple abi versions to the app/test unit test >> framework to allow users to switch between abi versions (set_abi_version), >> without further polluting the already long list of unit tests available in >> app/test. >> >> The intention here is that, in future as developers need to depreciate >> APIs, their associated unit tests may move into the ABI Version testing >> mechanism of the app/test instead of being replaced by the latest set of >> unit tests as would be the case today. >> >> ToDo: >> * Refactor the v2.0 and v16.04 unit tests to separate functional and >> performance test cases. >> * Add support for trigger ABI Version unit tests from the app/test command >> line. >> > While I admire the goal, given the amount of code that seems to be involved > here, I'm not sure if the "test" binary is the place to put this. I think > it might be better as a separate ABI compatiblity test app. > > A separate question is whether this is really necessary to ensure ABI > compatibility? Do other projects do this? Is the info from the abi > compatiblity checker script already in DPDK, or from other > already-available tools not sufficient? > > /Bruce >
[dpdk-dev] [PATCH] eventdev: optimize Rx adapter event enqueue
Setup event when the Rx queue is added to the
adapter in place of generating the event when it is
being enqueued to the event device.
Signed-off-by: Nikhil Rao
---
lib/librte_eventdev/rte_event_eth_rx_adapter.c | 39 ++
1 file changed, 15 insertions(+), 24 deletions(-)
This patch depends on
http://patchwork.dpdk.org/patch/53614/
diff --git a/lib/librte_eventdev/rte_event_eth_rx_adapter.c
b/lib/librte_eventdev/rte_event_eth_rx_adapter.c
index a201668..c51b9f3 100644
--- a/lib/librte_eventdev/rte_event_eth_rx_adapter.c
+++ b/lib/librte_eventdev/rte_event_eth_rx_adapter.c
@@ -198,11 +198,8 @@ struct eth_rx_queue_info {
int queue_enabled; /* True if added */
int intr_enabled;
uint16_t wt;/* Polling weight */
- uint8_t event_queue_id; /* Event queue to enqueue packets to */
- uint8_t sched_type; /* Sched type for events */
- uint8_t priority; /* Event priority */
- uint32_t flow_id; /* App provided flow identifier */
uint32_t flow_id_mask; /* Set to ~0 if app provides flow id else 0 */
+ uint64_t event;
};
static struct rte_event_eth_rx_adapter **event_eth_rx_adapter;
@@ -758,10 +755,8 @@ static uint16_t rxa_gcd_u16(uint16_t a, uint16_t b)
struct rte_eth_event_enqueue_buffer *buf =
&rx_adapter->event_enqueue_buffer;
struct rte_event *ev = &buf->events[buf->count];
- int32_t qid = eth_rx_queue_info->event_queue_id;
- uint8_t sched_type = eth_rx_queue_info->sched_type;
- uint8_t priority = eth_rx_queue_info->priority;
- uint32_t flow_id;
+ uint64_t event = eth_rx_queue_info->event;
+ uint32_t flow_id_mask = eth_rx_queue_info->flow_id_mask;
struct rte_mbuf *m = mbufs[0];
uint32_t rss_mask;
uint32_t rss;
@@ -803,17 +798,9 @@ static uint16_t rxa_gcd_u16(uint16_t a, uint16_t b)
rss = do_rss ?
rxa_do_softrss(m, rx_adapter->rss_key_be) :
m->hash.rss;
- flow_id =
- eth_rx_queue_info->flow_id &
- eth_rx_queue_info->flow_id_mask;
- flow_id |= rss & ~eth_rx_queue_info->flow_id_mask;
- ev->flow_id = flow_id;
- ev->op = RTE_EVENT_OP_NEW;
- ev->sched_type = sched_type;
- ev->queue_id = qid;
- ev->event_type = RTE_EVENT_TYPE_ETH_RX_ADAPTER;
- ev->sub_event_type = 0;
- ev->priority = priority;
+ ev->event = event;
+ ev->flow_id = (rss & ~flow_id_mask) |
+ (ev->flow_id & flow_id_mask);
ev->mbuf = m;
ev++;
}
@@ -1705,6 +1692,7 @@ static uint16_t rxa_gcd_u16(uint16_t a, uint16_t b)
int pollq;
int intrq;
int sintrq;
+ struct rte_event *qi_ev;
if (rx_queue_id == -1) {
uint16_t nb_rx_queues;
@@ -1721,16 +1709,19 @@ static uint16_t rxa_gcd_u16(uint16_t a, uint16_t b)
sintrq = rxa_shared_intr(dev_info, rx_queue_id);
queue_info = &dev_info->rx_queue[rx_queue_id];
- queue_info->event_queue_id = ev->queue_id;
- queue_info->sched_type = ev->sched_type;
- queue_info->priority = ev->priority;
queue_info->wt = conf->servicing_weight;
+ qi_ev = (struct rte_event *)&queue_info->event;
+ qi_ev->event = ev->event;
+ qi_ev->op = RTE_EVENT_OP_NEW;
+ qi_ev->event_type = RTE_EVENT_TYPE_ETH_RX_ADAPTER;
+ qi_ev->sub_event_type = 0;
+
if (conf->rx_queue_flags &
RTE_EVENT_ETH_RX_ADAPTER_QUEUE_FLOW_ID_VALID) {
- queue_info->flow_id = ev->flow_id;
queue_info->flow_id_mask = ~0;
- }
+ } else
+ qi_ev->flow_id = 0;
rxa_update_queue(rx_adapter, dev_info, rx_queue_id, 1);
if (rxa_polled_queue(dev_info, rx_queue_id)) {
--
1.8.3.1
[dpdk-dev] [PATCH 0/2] Add ABI Version Testing to app/test
This patchset adds ABI Version Testing functionality to the app/test unit test framework. The patchset is intended to address two issues previously raised during ML conversations on ABI Stability; 1. How do we unit test still supported previous ABI Versions. 2. How to we unit test inline functions from still supported previous ABI Versions. The more obvious way to achieve both of the above is to simply archive pre-built binaries compiled against previous versions of DPDK for use unit testing previous ABI Versions, and while this should still be done as an additional check, this approach does not scale well, must every DPDK developer have a local copy of these binaries to test with, before upstreaming changes? Instead starting with rte_lpm, I did the following:- * I reproduced mostly unmodified unit tests from previous ABI Versions, in this case v2.0 and v16.04 * I reproduced the rte_lpm interface header from these previous ABI Versions,including the inline functions and remapping symbols to appropriate versions. * I added support for multiple abi versions to the app/test unit test framework to allow users to switch between abi versions (set_abi_version), without further polluting the already long list of unit tests available in app/test. The intention here is that, in future as developers need to depreciate APIs, their associated unit tests may move into the ABI Version testing mechanism of the app/test instead of being replaced by the latest set of unit tests as would be the case today. ToDo: * Refactor the v2.0 and v16.04 unit tests to separate functional and performance test cases. * Add support for trigger ABI Version unit tests from the app/test command line. Ray Kinsella (2): app/test: Add ABI Version Testing functionality app/test: LPMv4 ABI Version Testing app/test/Makefile | 12 +- app/test/commands.c| 131 ++- app/test/meson.build |5 + app/test/test.c|2 + app/test/test.h| 52 +- app/test/test_lpm.c|1 + app/test/test_lpm_perf.c | 293 +-- app/test/test_lpm_routes.c | 287 +++ app/test/test_lpm_routes.h | 25 + app/test/v16.04/dcompat.h | 23 + app/test/v16.04/rte_lpm.h | 463 +++ app/test/v16.04/rte_lpm_neon.h | 119 +++ app/test/v16.04/rte_lpm_sse.h | 120 +++ app/test/v16.04/test_lpm.c | 1405 app/test/v16.04/test_v1604.c | 14 + app/test/v2.0/dcompat.h| 23 + app/test/v2.0/rte_lpm.h| 443 ++ app/test/v2.0/test_lpm.c | 1306 + app/test/v2.0/test_v20.c | 14 + 19 files changed, 4420 insertions(+), 318 deletions(-) create mode 100644 app/test/test_lpm_routes.c create mode 100644 app/test/test_lpm_routes.h create mode 100644 app/test/v16.04/dcompat.h create mode 100644 app/test/v16.04/rte_lpm.h create mode 100644 app/test/v16.04/rte_lpm_neon.h create mode 100644 app/test/v16.04/rte_lpm_sse.h create mode 100644 app/test/v16.04/test_lpm.c create mode 100644 app/test/v16.04/test_v1604.c create mode 100644 app/test/v2.0/dcompat.h create mode 100644 app/test/v2.0/rte_lpm.h create mode 100644 app/test/v2.0/test_lpm.c create mode 100644 app/test/v2.0/test_v20.c -- 2.17.1
[dpdk-dev] [PATCH 1/2] app/test: Add ABI Version Testing functionality
This patchset adds ABI Version Testing functionality to the app/test
unit test framework, comprised of
1. The TEST_DPDK_ABI_VERSION_* and REISTER_TEST_ABI_VERSION macros to
register abi versions with infrastructure.
2. The MAP_ABI_SYMBOL_VERSION macro to remap symbols based on their ABI
Version.
3. The set_abi_version command, to switch between ABI Versions.
Signed-off-by: Ray Kinsella
---
app/test/commands.c | 131 ++--
app/test/test.c | 2 +
app/test/test.h | 52 +++---
3 files changed, 159 insertions(+), 26 deletions(-)
diff --git a/app/test/commands.c b/app/test/commands.c
index 8d5a03a95..06fc33ee5 100644
--- a/app/test/commands.c
+++ b/app/test/commands.c
@@ -50,12 +50,22 @@
//
+static uint8_t test_abi_version = TEST_DPDK_ABI_VERSION_DEFAULT;
+
+static struct test_abi_version_list abi_version_list =
+ TAILQ_HEAD_INITIALIZER(abi_version_list);
+
static struct test_commands_list commands_list =
TAILQ_HEAD_INITIALIZER(commands_list);
-void
-add_test_command(struct test_command *t)
+void add_abi_version(struct test_abi_version *av)
+{
+ TAILQ_INSERT_TAIL(&abi_version_list, av, next);
+}
+
+void add_test_command(struct test_command *t, uint8_t abi_version)
{
+ t->abi_version = abi_version;
TAILQ_INSERT_TAIL(&commands_list, t, next);
}
@@ -63,6 +73,12 @@ struct cmd_autotest_result {
cmdline_fixed_string_t autotest;
};
+cmdline_parse_token_string_t
+cmd_autotest_autotest[TEST_DPDK_ABI_VERSION_MAX] = {
+ [0 ... TEST_DPDK_ABI_VERSION_MAX-1] =
+ TOKEN_STRING_INITIALIZER(struct cmd_autotest_result, autotest, "")
+};
+
static void cmd_autotest_parsed(void *parsed_result,
__attribute__((unused)) struct cmdline *cl,
__attribute__((unused)) void *data)
@@ -72,7 +88,8 @@ static void cmd_autotest_parsed(void *parsed_result,
int ret = 0;
TAILQ_FOREACH(t, &commands_list, next) {
- if (!strcmp(res->autotest, t->command))
+ if (!strcmp(res->autotest, t->command)
+ && t->abi_version == test_abi_version)
ret = t->callback();
}
@@ -86,10 +103,6 @@ static void cmd_autotest_parsed(void *parsed_result,
fflush(stdout);
}
-cmdline_parse_token_string_t cmd_autotest_autotest =
- TOKEN_STRING_INITIALIZER(struct cmd_autotest_result, autotest,
-"");
-
cmdline_parse_inst_t cmd_autotest = {
.f = cmd_autotest_parsed, /* function to call */
.data = NULL, /* 2nd arg of func */
@@ -244,6 +257,53 @@ cmdline_parse_inst_t cmd_quit = {
//
+struct cmd_set_abi_version_result {
+ cmdline_fixed_string_t set;
+ cmdline_fixed_string_t abi_version_name;
+};
+
+static void cmd_set_abi_version_parsed(
+ void *parsed_result,
+ __attribute__((unused)) struct cmdline *cl,
+ __attribute__((unused)) void *data)
+{
+ struct test_abi_version *av;
+ struct cmd_set_abi_version_result *res = parsed_result;
+
+ TAILQ_FOREACH(av, &abi_version_list, next) {
+ if (!strcmp(res->abi_version_name, av->version_name)) {
+
+ printf("abi version set to %s\n", av->version_name);
+ test_abi_version = av->version_id;
+ cmd_autotest.tokens[0] =
+ (void *)&cmd_autotest_autotest[av->version_id];
+ }
+ }
+
+ fflush(stdout);
+}
+
+cmdline_parse_token_string_t cmd_set_abi_version_set =
+ TOKEN_STRING_INITIALIZER(struct cmd_set_abi_version_result, set,
+ "set_abi_version");
+
+cmdline_parse_token_string_t cmd_set_abi_version_abi_version =
+ TOKEN_STRING_INITIALIZER(struct cmd_set_abi_version_result,
+ abi_version_name, NULL);
+
+cmdline_parse_inst_t cmd_set_abi_version = {
+ .f = cmd_set_abi_version_parsed, /* function to call */
+ .data = NULL, /* 2nd arg of func */
+ .help_str = "set abi version: ",
+ .tokens = {/* token list, NULL terminated */
+ (void *)&cmd_set_abi_version_set,
+ (void *)&cmd_set_abi_version_abi_version,
+ NULL,
+ },
+};
+
+//
+
struct cmd_set_rxtx_result {
cmdline_fixed_string_t set;
cmdline_fixed_string_t mode;
@@ -259,7 +319,7 @@ static void cmd_set_rxtx_parsed(void *parsed_result, struct
cmdline *cl,
cmdline_parse_token_string_t cmd_set_rxtx_set =
TOKEN_STRING_INITIALIZER(struct cmd_set_rxtx_result, set,
-"set_rxtx_mode");
+ "set_rxtx_mode");
cmdline_parse_token_string_t cmd_set_rxtx_mode =
TOKEN_STRING_INITIALIZER(s
[dpdk-dev] [PATCH 2/2] app/test: LPMv4 ABI Version Testing
This second patch adds the LPM ABI Version Unit Tests, comprised of
1. Registering DPDK v2.0 and DPDK v16.04 ABI Versions with the
infrastructure.
2. Forward Porting the DPDK v2.0 and DPDK v16.04 LPM Unit Test
cases, remapping the LPM Library symbols to the appropriate versions.
3. Refactoring the lpm perf routes table to make this
functionality available to the v2.0 and v16.04 unit tests, forwarding
porting this code also from v2.0 etc would have increased the DPDK
codebase several MLoC.q
Signed-off-by: Ray Kinsella
---
app/test/Makefile | 12 +-
app/test/meson.build |5 +
app/test/test_lpm.c|1 +
app/test/test_lpm_perf.c | 293 +--
app/test/test_lpm_routes.c | 287 +++
app/test/test_lpm_routes.h | 25 +
app/test/v16.04/dcompat.h | 23 +
app/test/v16.04/rte_lpm.h | 463 +++
app/test/v16.04/rte_lpm_neon.h | 119 +++
app/test/v16.04/rte_lpm_sse.h | 120 +++
app/test/v16.04/test_lpm.c | 1405
app/test/v16.04/test_v1604.c | 14 +
app/test/v2.0/dcompat.h| 23 +
app/test/v2.0/rte_lpm.h| 443 ++
app/test/v2.0/test_lpm.c | 1306 +
app/test/v2.0/test_v20.c | 14 +
16 files changed, 4261 insertions(+), 292 deletions(-)
create mode 100644 app/test/test_lpm_routes.c
create mode 100644 app/test/test_lpm_routes.h
create mode 100644 app/test/v16.04/dcompat.h
create mode 100644 app/test/v16.04/rte_lpm.h
create mode 100644 app/test/v16.04/rte_lpm_neon.h
create mode 100644 app/test/v16.04/rte_lpm_sse.h
create mode 100644 app/test/v16.04/test_lpm.c
create mode 100644 app/test/v16.04/test_v1604.c
create mode 100644 app/test/v2.0/dcompat.h
create mode 100644 app/test/v2.0/rte_lpm.h
create mode 100644 app/test/v2.0/test_lpm.c
create mode 100644 app/test/v2.0/test_v20.c
diff --git a/app/test/Makefile b/app/test/Makefile
index 68d6b4fbc..5899eb8b9 100644
--- a/app/test/Makefile
+++ b/app/test/Makefile
@@ -78,6 +78,10 @@ SRCS-y += test_ring.c
SRCS-y += test_ring_perf.c
SRCS-y += test_pmd_perf.c
+#ABI Version Testing
+SRCS-$(CONFIG_RTE_BUILD_SHARED_LIB) += v2.0/test_v20.c
+SRCS-$(CONFIG_RTE_BUILD_SHARED_LIB) += v16.04/test_v1604.c
+
ifeq ($(CONFIG_RTE_LIBRTE_TABLE),y)
SRCS-y += test_table.c
SRCS-$(CONFIG_RTE_LIBRTE_PIPELINE) += test_table_pipeline.c
@@ -107,7 +111,6 @@ SRCS-y += test_logs.c
SRCS-y += test_memcpy.c
SRCS-y += test_memcpy_perf.c
-
SRCS-$(CONFIG_RTE_LIBRTE_MEMBER) += test_member.c
SRCS-$(CONFIG_RTE_LIBRTE_MEMBER) += test_member_perf.c
@@ -122,11 +125,18 @@ SRCS-$(CONFIG_RTE_LIBRTE_HASH) += test_hash_multiwriter.c
SRCS-$(CONFIG_RTE_LIBRTE_HASH) += test_hash_readwrite.c
SRCS-$(CONFIG_RTE_LIBRTE_HASH) += test_hash_readwrite_lf.c
+SRCS-$(CONFIG_RTE_LIBRTE_LPM) += test_lpm_routes.c
SRCS-$(CONFIG_RTE_LIBRTE_LPM) += test_lpm.c
SRCS-$(CONFIG_RTE_LIBRTE_LPM) += test_lpm_perf.c
SRCS-$(CONFIG_RTE_LIBRTE_LPM) += test_lpm6.c
SRCS-$(CONFIG_RTE_LIBRTE_LPM) += test_lpm6_perf.c
+#LPM ABI Testing
+ifeq ($(CONFIG_RTE_BUILD_SHARED_LIB),y)
+SRCS-$(CONFIG_RTE_LIBRTE_LPM) += v2.0/test_lpm.c
+SRCS-$(CONFIG_RTE_LIBRTE_LPM) += v16.04/test_lpm.c
+endif
+
SRCS-y += test_debug.c
SRCS-y += test_errno.c
SRCS-y += test_tailq.c
diff --git a/app/test/meson.build b/app/test/meson.build
index 83391cef0..628f4e1ff 100644
--- a/app/test/meson.build
+++ b/app/test/meson.build
@@ -4,6 +4,8 @@
test_sources = files('commands.c',
'packet_burst_generator.c',
'sample_packet_forward.c',
+ 'v2.0/test_v20.c',
+ 'v16.04/test_v1604.c',
'test.c',
'test_acl.c',
'test_alarm.c',
@@ -63,6 +65,9 @@ test_sources = files('commands.c',
'test_lpm6.c',
'test_lpm6_perf.c',
'test_lpm_perf.c',
+ 'test_lpm_routes.c',
+ 'v2.0/test_lpm.c',
+ 'v16.04/test_lpm.c',
'test_malloc.c',
'test_mbuf.c',
'test_member.c',
diff --git a/app/test/test_lpm.c b/app/test/test_lpm.c
index 5d697dd0f..bfa702677 100644
--- a/app/test/test_lpm.c
+++ b/app/test/test_lpm.c
@@ -1277,6 +1277,7 @@ test_lpm(void)
int status, global_status = 0;
for (i = 0; i < NUM_LPM_TESTS; i++) {
+ printf("# test %02d\n", i);
status = tests[i]();
if (status < 0) {
printf("ERROR: LPM Test %u: FAIL\n", i);
diff --git a/app/test/test_lpm_perf.c b/app/test/test_lpm_perf.c
index 3b98ce0c8..a6b8b35c2 100644
--- a/app/test/test_lpm_perf.c
+++ b/app/test/test_lpm_perf.c
@@ -5,7 +5,6 @@
#include
#include
#include
-#include
#include
#include
@@ -13,6 +12,7 @@
#include
#include
+#include "test_lpm_routes.h"
#include "test.h"
#include "test_xmmt_ops.h"
@@ -27,295 +27,6 @@
#define BATCH_SIZE (1 << 12)
#define BULK_SIZE 32
-#define MAX_RULE_NUM (120)
-
-struct route_rule {
- uint32_t ip;
- uint8_t depth;
-};
-
-struct route_rule large_r
[dpdk-dev] [PATCH v3 0/2] net/mlx5: remove TCF support from PMD
Today, it is possible to offload an interface flow rules to the hardware using DPDK flow commands. With mlx5 it is also possible to offload a limited set of flow rules to the mlxsw (or E-switch) using the same DPDK flow commands. A 'transfer' attribute was added to the flow rule creation command in order to distinguish between configuring port flows and E-switch flows. The commands destined for the E-switch are transposed to TC-flower rules and are send, as Netlink messages, to the mlx5 driver, or more precisely to the netdev which represent the mlxsw port. With the addition to the PMD of E-switch configuration via DR (direct verbs rules) it is now possible to configure the E-switch using these commands instead of using TC-Flower messages. Doing so will allow us to remove the TCF support and the dependency of libmnl from the PMD. The purpose of this RFC is to propose configuring the E-switch flow filtering using DR, to remove the TCF support from the PMD and to remove the dependency of the PMD in libmnl. As for today VLAN insertion or removal is not supported in DR, this support will be added in separate commits. Moti Haimovsky (2): net/mlx5: fix crashing testpmd on null drv opts net/mlx5: remove TCF support from PMD doc/build-sdk-meson.txt |2 +- doc/guides/nics/mlx5.rst | 19 - doc/guides/platform/bluefield.rst |4 - drivers/net/mlx5/Makefile | 303 -- drivers/net/mlx5/meson.build | 123 +- drivers/net/mlx5/mlx5.c | 32 - drivers/net/mlx5/mlx5.h |3 - drivers/net/mlx5/mlx5_flow.c | 43 +- drivers/net/mlx5/mlx5_flow.h | 25 - drivers/net/mlx5/mlx5_flow_tcf.c | 6382 - mk/rte.app.mk |2 +- 11 files changed, 24 insertions(+), 6914 deletions(-) delete mode 100644 drivers/net/mlx5/mlx5_flow_tcf.c -- 1.8.3.1
[dpdk-dev] [PATCH v3 1/2] net/mlx5: fix crashing testpmd on null drv opts
mlx5 implements mlx5_flow_null_drv_ops to be used when a specific
flow typei/driver is not available or invalid.
This routines return error without modifying the rte_flow_error
parameter passed to them which causes testpmd, for example, to crash.
This commit addresses the issue by modifying the rte_flow_error
parameter in theses routines.
Fixes: 0c76d1c9a18d ("net/mlx5: add abstraction for multiple flow drivers")
Fixes: 684dafe795d0 ("net/mlx5: add flow query abstraction interface")
Cc: sta...@dpdk.org
Signed-off-by: Moti Haimovsky
---
v3:
* Modified patch subject.
v2:
* Fixed checkpatch warnings.
---
drivers/net/mlx5/mlx5_flow.c | 29 +++--
1 file changed, 15 insertions(+), 14 deletions(-)
diff --git a/drivers/net/mlx5/mlx5_flow.c b/drivers/net/mlx5/mlx5_flow.c
index 9887018..e5a8e33 100644
--- a/drivers/net/mlx5/mlx5_flow.c
+++ b/drivers/net/mlx5/mlx5_flow.c
@@ -1694,19 +1694,20 @@ uint32_t mlx5_flow_adjust_priority(struct rte_eth_dev
*dev, int32_t priority,
const struct rte_flow_attr *attr __rte_unused,
const struct rte_flow_item items[] __rte_unused,
const struct rte_flow_action actions[] __rte_unused,
- struct rte_flow_error *error __rte_unused)
+ struct rte_flow_error *error)
{
- rte_errno = ENOTSUP;
- return -rte_errno;
+ return rte_flow_error_set(error, ENOTSUP,
+ RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL, NULL);
}
static struct mlx5_flow *
flow_null_prepare(const struct rte_flow_attr *attr __rte_unused,
const struct rte_flow_item items[] __rte_unused,
const struct rte_flow_action actions[] __rte_unused,
- struct rte_flow_error *error __rte_unused)
+ struct rte_flow_error *error)
{
- rte_errno = ENOTSUP;
+ rte_flow_error_set(error, ENOTSUP,
+ RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL, NULL);
return NULL;
}
@@ -1716,19 +1717,19 @@ uint32_t mlx5_flow_adjust_priority(struct rte_eth_dev
*dev, int32_t priority,
const struct rte_flow_attr *attr __rte_unused,
const struct rte_flow_item items[] __rte_unused,
const struct rte_flow_action actions[] __rte_unused,
- struct rte_flow_error *error __rte_unused)
+ struct rte_flow_error *error)
{
- rte_errno = ENOTSUP;
- return -rte_errno;
+ return rte_flow_error_set(error, ENOTSUP,
+ RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL, NULL);
}
static int
flow_null_apply(struct rte_eth_dev *dev __rte_unused,
struct rte_flow *flow __rte_unused,
- struct rte_flow_error *error __rte_unused)
+ struct rte_flow_error *error)
{
- rte_errno = ENOTSUP;
- return -rte_errno;
+ return rte_flow_error_set(error, ENOTSUP,
+ RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL, NULL);
}
static void
@@ -1748,10 +1749,10 @@ uint32_t mlx5_flow_adjust_priority(struct rte_eth_dev
*dev, int32_t priority,
struct rte_flow *flow __rte_unused,
const struct rte_flow_action *actions __rte_unused,
void *data __rte_unused,
- struct rte_flow_error *error __rte_unused)
+ struct rte_flow_error *error)
{
- rte_errno = ENOTSUP;
- return -rte_errno;
+ return rte_flow_error_set(error, ENOTSUP,
+ RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL, NULL);
}
/* Void driver to protect from null pointer reference. */
--
1.8.3.1
Re: [dpdk-dev] checklist for DPDK on Windows
Should we have a page with an up-to-date status of the work? Would be good to have an overview of the progress, even if it's not as simple as a progress bar :) > From: Ranjit Menon > > 15/03/2019 00:04, Thomas Monjalon: > >> Hi, > >> > >> Below is a list of directories, files or functions which we need to > >> check to make basic DPDK works on Windows. > >> If something is missing, please complete. > >> > >> The goal of this list is to make a status of what is already done, > >> and plan what should be done next. It will help to share the workload > >> among all volunteers during the next months. > >> > >> buildtools/pmdinfogen > - not ported to Windows > >> usertools/dpdk-devbind.py > - not ported to Windows > >> drivers/bus/vdev > - not ported to Windows > >> drivers/bus/pci > - ported to Windows in draft repo (using ICC) > >> lib/librte_pci > - ported to Windows in draft repo (using ICC) > >> lib/librte_cmdline > - ported to Windows in draft repo (using ICC) > >> lib/librte_kvargs > - ported to Windows in draft repo (using ICC) > >> lib/librte_ring > - ported to Windows in draft repo (using ICC) > >> lib/librte_mempool > - ported to Windows in draft repo (using ICC) > >> lib/librte_mbuf > - ported to Windows in draft repo (using ICC) > >> lib/librte_net > - ported to Windows in draft repo (using ICC) > >> lib/librte_eal/common/include/rte_errno.h > >> lib/librte_eal/common/include/rte_string_fns.h > >> lib/librte_eal/common/include/rte_lcore.h > >> lib/librte_eal/common/arch/x86/rte_cpuflags.c > >> lib/librte_eal/common/arch/x86/rte_cycles.c > >> lib/librte_eal/common/eal_common_options.c > >> lib/librte_eal/common/eal_common_thread.c > >> lib/librte_eal/common/eal_common_proc.c > - most 'common' [c/h] files ported to Windows in draft repo (using ICC) > >> lib/librte_eal/windows/eal/eal.c > >> eal_create_runtime_dir() > >> rte_eal_iopl_init() > - part of file port available in "Helloworld" patch > >> lib/librte_eal/windows/eal/eal_alarm.c > - ported to Windows in draft repo (using ICC) > >> lib/librte_eal/windows/eal/eal_cpuflags.c > - not ported to Windows > >> lib/librte_eal/windows/eal/eal_debug.c > - part of file port available in "Helloworld" patch > >> lib/librte_eal/windows/eal/eal_dev.c > - not ported to Windows > >> lib/librte_eal/windows/eal/eal_interrupts.c > - ported to Windows in draft repo (using ICC) > >> lib/librte_eal/windows/eal/eal_lcore.c > - part of file port available in "Helloworld" patch > >> lib/librte_eal/windows/eal/eal_log.c > - ported to Windows in draft repo (using ICC) > >> lib/librte_eal/windows/eal/eal_memory.c > - ported to Windows in draft repo (using ICC) > >> lib/librte_eal/windows/eal/eal_thread.c > - part of file port available in "Helloworld" patch > >> lib/librte_eal/windows/eal/eal_timer.c > - ported to Windows in draft repo (using ICC) > >> > >> Please check this list and mention what is done or in progress. > >> The best would be to reference some patches or commits to help > >> progress together as a community, thanks.
Re: [dpdk-dev] [PATCH] Fix off-by-one errors in snprintf in various files
On 5/10/19 11:28 AM, Burakov, Anatoly wrote: On 10-May-19 3:53 PM, Michael Santana wrote: snprintf guarantees to always correctly place a null terminator in the buffer string. So manually placing a null terminator in a buffer right after a call to snprintf is redundant code. Additionally, there is no need to use 'sizeof(buffer) - 1' in snprintf as this means we are not using the last character in the buffer. 'sizeof(buffer)' is enough. Cc: sta...@dpdk.org Signed-off-by: Michael Santana --- LGTM Acked-by: Anatoly Burakov ping
Re: [dpdk-dev] [PATCH 0/2] Add ABI Version Testing to app/test
Someone kindly approved it, and saved me sending it again from the right email a/c, thank you. On 28/05/2019 13:08, Bruce Richardson wrote: > On Tue, May 28, 2019 at 12:51:56PM +0100, Ray Kinsella wrote: >> This patchset adds ABI Version Testing functionality to the app/test unit >> test framework. >> >> The patchset is intended to address two issues previously raised during ML >> conversations on ABI Stability; >> 1. How do we unit test still supported previous ABI Versions. >> 2. How to we unit test inline functions from still supported previous ABI >> Versions. >> >> The more obvious way to achieve both of the above is to simply archive >> pre-built binaries compiled against previous versions of DPDK for use unit >> testing previous ABI Versions, and while this should still be done as an >> additional check, this approach does not scale well, must every DPDK >> developer have a local copy of these binaries to test with, before >> upstreaming changes? >> >> Instead starting with rte_lpm, I did the following:- >> >> * I reproduced mostly unmodified unit tests from previous ABI Versions, >> in this case v2.0 and v16.04 >> * I reproduced the rte_lpm interface header from these previous ABI >> Versions,including the inline functions and remapping symbols to >> appropriate versions. >> * I added support for multiple abi versions to the app/test unit test >> framework to allow users to switch between abi versions (set_abi_version), >> without further polluting the already long list of unit tests available in >> app/test. >> >> The intention here is that, in future as developers need to depreciate >> APIs, their associated unit tests may move into the ABI Version testing >> mechanism of the app/test instead of being replaced by the latest set of >> unit tests as would be the case today. >> >> ToDo: >> * Refactor the v2.0 and v16.04 unit tests to separate functional and >> performance test cases. >> * Add support for trigger ABI Version unit tests from the app/test command >> line. >> > While I admire the goal, given the amount of code that seems to be involved > here, I'm not sure if the "test" binary is the place to put this. I think > it might be better as a separate ABI compatiblity test app. I did think about that also - the test binary, similar to testpmd is very busy. I sought to mitigate that with set_abi_version command, mitigating unit test name collisions etc. I would have a huge concern about putting it into a separate binary, as these tcs would quickly become forgotten about. > > A separate question is whether this is really necessary to ensure ABI > compatibility? I would argue that we currently have no idea if ABI versioned functions actually work? Why offer backward compatibility, if we don't test it? > Do other projects do this? The C++ stdlib is many ways similar to DPDK, ABI compatibility is a well understood problem over there. See the following presentation Slide 20. https://accu.org/content/conf2015/JonathanWakely-What%20Is%20An%20ABI%20And%20Why%20Is%20It%20So%20Complicated.pdf They have a small corpus of TC's for this https://github.com/gcc-mirror/gcc/tree/master/libstdc%2B%2B-v3/testsuite/[backward,abi] > Is the info from the abi > compatiblity checker script already in DPDK, or from other > already-available tools not sufficient? Well this just tells you that you ABI has changed. We don't have anything at the moment that tells us that the ABI compatible functions actually work ... ? > > /Bruce >
Re: [dpdk-dev] [PATCH 07/27] sched: update pipe profile add api
On Tue, 28 May 2019 14:05:33 +0200
Lukasz Krakowiak wrote:
>
> static int
> pipe_profile_check(struct rte_sched_pipe_params *params,
> - uint32_t rate)
> + uint32_t rate, uint16_t *qsize)
> {
> uint32_t i;
>
> /* Pipe parameters */
> if (params == NULL)
> - return -10;
> + return -11;
Having used this before, and suffered from the error handling.
Please change this to do proper logging and use a normal convention
for error numbers. This is not in fast path, it should be more
user friendly.
Re: [dpdk-dev] [PATCH v2 1/2] test/hash: use existing lcore API
> On May 27, 2019, at 5:15 AM, David Marchand wrote:
>
> Hello Dharmik,
Hi David,
>
> On Wed, May 22, 2019 at 9:38 PM Dharmik Thakkar
> wrote:
> > On May 22, 2019, at 12:17 PM, David Marchand
> > wrote:
> > On Wed, May 22, 2019 at 6:27 PM Dharmik Thakkar
> > wrote:
> > > @@ -738,10 +736,9 @@ struct {
> > > rte_eal_remote_launch(test_rwc_reader,
> > > (void
> > > *)(uintptr_t)read_type,
> > >
> > > enabled_core_ids[i]);
> > > - rte_eal_mp_wait_lcore();
> > >
> > > for (i = 1; i <= rwc_core_cnt[n]; i++)
> > > - if (lcore_config[i].ret < 0)
> > > + if (rte_eal_wait_lcore(i) < 0)
> > if (rte_eal_wait_lcore(enabled_core_ids[i]) < 0)
> >
> > (There are similar changes in other functions too.
> > I realize that this is a separate issue than what the patch is aimed for.
> > If you see fit, please integrate it, else I will put out a patch once your
> > patch has been merged.)
> >
> > Indeed, reproduced.
> >
>
> Did you have a chance to look at this ?
> These loops could be reworked by using RTE_LCORE_FOREACH_SLAVE() (so that it
> automatically skips the master lcore) removing the need for
> enabled_core_ids[].
>
Yes, I have tried to use RTE_LCORE_FOREACH_SLAVE() and eliminate
enabled_core_ids[]. However, the code becomes quite complicated to read with if
statements
within RTE_LCORE_FOREACH_SLAVE(), to break the loop (since not all slave cores
are checked for). This is a snip from test_hash_multi_add_lookup():
/* snip */
RTE_LCORE_FOREACH_SLAVE(core_id) {
if (i < rwc_core_cnt[n])
rte_eal_remote_launch(test_rwc_reader,
(void
*)(uintptr_t)read_type,
core_id);
else if (i >= rwc_core_cnt[n] &&
i < (rwc_core_cnt[n] +
rwc_core_cnt[m])) {
rte_eal_remote_launch(test_rwc_multi_writer,
(void
*)(uintptr_t)pos_core,
core_id);
pos_core++;
} else
break;
i++;
}
>
> --
> David Marchand
[dpdk-dev] [RFC] net/mlx5 add hairpin support
The hairpin feature acts as "bump on the wire", meaning that a packet that is received from the wire can be modified using offloaded action and then sent back to the wire without application intervention which can save CPU cycles. The hairpin is the inverse function of loopback in which application sends a packet then it is received again by the application without being sent to the wire. The hairpin can be used by a number of different NVF, for example load balancer, gateway and so on. In current version the hairpin will work only on routing from the wire port to the same wire port. This can later be expanded to work between different ports. The hairpin will be composed of 2 flows one ingress flow, which will be detected as hairpin flow by setting transfer flag = 1 and action port. The second rule if required, will be normal egress flow. In order to combine between the flows, the application can set metadata on the ingress which will be copied to the egress were it can be matched. Signed-off-by: Ori Kam Thanks Ori Kam
Re: [dpdk-dev] [PATCH v2 09/10] net/bnxt: HWRM version update
WIll do. Thanks, Lance On Tue, May 28, 2019 at 5:12 AM Ferruh Yigit wrote: > > On 5/24/2019 3:49 PM, Lance Richardson wrote: > > From: Ajit Khaparde > > > > Update HWRM API to version 1.10.0.74 > > There are three commits related: > "update HWRM API" (to version 1.10.0.19) > "update HWRM version" (to 1.10.0.48) > "HWRM version update" (to version 1.10.0.74) > > Is "HWRM API" & "HWRM version" different things? > I assume they are not different and these commits are updating "HWRM API > version" gradually, if so can you please use a consistent commit title? > > Thanks, > ferruh > > > > > Signed-off-by: Ajit Khaparde > > Reviewed-by: Lance Richardson > > <...> > > >
Re: [dpdk-dev] [PATCH v2 00/10] bnxt patchset
On Tue, May 28, 2019 at 5:19 AM Ferruh Yigit wrote: ... > There are warnings from > "./devtools/checkpatches.sh" & "./devtools/check-git-log.sh" > Can you please address them in next version? > > It is acceptable to have some level of "checkpatches.sh" in the base code > update > (HWRM update code for your case) but please sure DPDK specific files don't > have > any warning. > > Thanks, > ferruh Sorry about that. I had run checkpatches.sh (which had warnings only for HWRM API updates), but not check-git-log.sh. Thanks, Lance
[dpdk-dev] [PATCH v3 0/8] bnxt patchset
This patchset bsae on top of the previous submission adds the following: 1) Support for vector mode TX and RX. 2) HWRM API update (split into multiple patches). 3) Fixes for RSS reta update and query. It also updates the release notes. v2: * Squashed patches 3 and 4 from v1 patchset. * Added Meson build support for vector mode PMD. * Dropped two unnecessary code style changes from patch 3. v3: * Squashed three RSS RETA fix patches into one. * Eliminated RTE_LIBRTE_BNXT_INC_VECTOR configuration flag. * Addressed commit log issues from check-git-log.sh. * Made subject line consistent for HWRM API update patches. * Separated release notes updates to keep update with associated patch (except for the first update for which the associated patch has already been applied). Ajit Khaparde (5): net/bnxt: update release notes for bnxt net/bnxt: fix RSS RETA indirection table ops net/bnxt: update HWRM API to version 1.10.0.19 net/bnxt: update HWRM API to version 1.10.0.48 net/bnxt: update HWRM API to version 1.10.0.74 Lance Richardson (3): net/bnxt: move Tx bd checking to header file net/bnxt: compute and store scattered Rx status net/bnxt: implement vector mode driver doc/guides/rel_notes/release_19_08.rst |6 + drivers/net/bnxt/Makefile |3 + drivers/net/bnxt/bnxt_ethdev.c | 201 +- drivers/net/bnxt/bnxt_hwrm.c |1 + drivers/net/bnxt/bnxt_ring.h |3 +- drivers/net/bnxt/bnxt_rxq.c|5 + drivers/net/bnxt/bnxt_rxq.h|4 + drivers/net/bnxt/bnxt_rxr.h|9 +- drivers/net/bnxt/bnxt_rxtx_vec_sse.c | 481 ++ drivers/net/bnxt/bnxt_txr.c| 15 - drivers/net/bnxt/bnxt_txr.h| 20 + drivers/net/bnxt/hsi_struct_def_dpdk.h | 7573 drivers/net/bnxt/meson.build |4 + 13 files changed, 7219 insertions(+), 1106 deletions(-) create mode 100644 drivers/net/bnxt/bnxt_rxtx_vec_sse.c -- 2.17.1
[dpdk-dev] [PATCH v3 2/8] net/bnxt: move Tx bd checking to header file
To allow sharing of tx_bds_in_hw() and bnxt_tx_avail() between
vector-mode and non-vector transmit functions, move these functions
into bnxt_txr.h.
Signed-off-by: Lance Richardson
Reviewed-by: Ajit Khaparde
Reviewed-by: Maxime Coquelin
---
drivers/net/bnxt/bnxt_txr.c | 15 ---
drivers/net/bnxt/bnxt_txr.h | 15 +++
2 files changed, 15 insertions(+), 15 deletions(-)
diff --git a/drivers/net/bnxt/bnxt_txr.c b/drivers/net/bnxt/bnxt_txr.c
index 9de12e0d0..16598ba63 100644
--- a/drivers/net/bnxt/bnxt_txr.c
+++ b/drivers/net/bnxt/bnxt_txr.c
@@ -103,21 +103,6 @@ int bnxt_init_tx_ring_struct(struct bnxt_tx_queue *txq,
unsigned int socket_id)
return 0;
}
-static inline uint32_t bnxt_tx_bds_in_hw(struct bnxt_tx_queue *txq)
-{
- return ((txq->tx_ring->tx_prod - txq->tx_ring->tx_cons) &
- txq->tx_ring->tx_ring_struct->ring_mask);
-}
-
-static inline uint32_t bnxt_tx_avail(struct bnxt_tx_queue *txq)
-{
- /* Tell compiler to fetch tx indices from memory. */
- rte_compiler_barrier();
-
- return ((txq->tx_ring->tx_ring_struct->ring_size -
-bnxt_tx_bds_in_hw(txq)) - 1);
-}
-
static uint16_t bnxt_start_xmit(struct rte_mbuf *tx_pkt,
struct bnxt_tx_queue *txq,
uint16_t *coal_pkts,
diff --git a/drivers/net/bnxt/bnxt_txr.h b/drivers/net/bnxt/bnxt_txr.h
index f802d5080..13ca04676 100644
--- a/drivers/net/bnxt/bnxt_txr.h
+++ b/drivers/net/bnxt/bnxt_txr.h
@@ -37,6 +37,21 @@ struct bnxt_sw_tx_bd {
unsigned short nr_bds;
};
+static inline uint32_t bnxt_tx_bds_in_hw(struct bnxt_tx_queue *txq)
+{
+ return ((txq->tx_ring->tx_prod - txq->tx_ring->tx_cons) &
+ txq->tx_ring->tx_ring_struct->ring_mask);
+}
+
+static inline uint32_t bnxt_tx_avail(struct bnxt_tx_queue *txq)
+{
+ /* Tell compiler to fetch tx indices from memory. */
+ rte_compiler_barrier();
+
+ return ((txq->tx_ring->tx_ring_struct->ring_size -
+bnxt_tx_bds_in_hw(txq)) - 1);
+}
+
void bnxt_free_tx_rings(struct bnxt *bp);
int bnxt_init_one_tx_ring(struct bnxt_tx_queue *txq);
int bnxt_init_tx_ring_struct(struct bnxt_tx_queue *txq, unsigned int
socket_id);
--
2.17.1
[dpdk-dev] [PATCH v3 3/8] net/bnxt: compute and store scattered Rx status
In preparation for a bnxt vector-mode driver, compute and store
scattered_rx status for the device when started.
Signed-off-by: Lance Richardson
Reviewed-by: Ajit Khaparde
Reviewed-by: Maxime Coquelin
---
drivers/net/bnxt/bnxt_ethdev.c | 23 +++
1 file changed, 23 insertions(+)
diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index b0244d6d1..5b3932b7b 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -624,6 +624,27 @@ static int bnxt_dev_lsc_intr_setup(struct rte_eth_dev
*eth_dev)
return 0;
}
+/*
+ * Determine whether the current configuration requires support for scattered
+ * receive; return 1 if scattered receive is required and 0 if not.
+ */
+static int bnxt_scattered_rx(struct rte_eth_dev *eth_dev)
+{
+ uint16_t buf_size;
+ int i;
+
+ for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
+ struct bnxt_rx_queue *rxq = eth_dev->data->rx_queues[i];
+
+ buf_size = (uint16_t)(rte_pktmbuf_data_room_size(rxq->mb_pool) -
+ RTE_PKTMBUF_HEADROOM);
+ if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len +
+ VLAN_TAG_SIZE * BNXT_NUM_VLANS > buf_size)
+ return 1;
+ }
+ return 0;
+}
+
static int bnxt_dev_start_op(struct rte_eth_dev *eth_dev)
{
struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
@@ -642,6 +663,8 @@ static int bnxt_dev_start_op(struct rte_eth_dev *eth_dev)
if (rc)
goto error;
+ eth_dev->data->scattered_rx = bnxt_scattered_rx(eth_dev);
+
bnxt_link_update_op(eth_dev, 1);
if (rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
--
2.17.1
[dpdk-dev] [PATCH v3 4/8] net/bnxt: implement vector mode driver
Introduce SSE vector mode support for the bnxt pmd.
Signed-off-by: Lance Richardson
Signed-off-by: Ajit Khaparde
---
v2:
* Squashed with v1 patch 4 ("fix double counting VLAN tags").
* Dropped two unnecessary coding style changes from bnxt_txr.h.
v3:
* Eliminated vector mode configration flag.
* Added RTE_ARCH_X86 checks around vector mode code.
* Added release notes item for SSE vector mode.
* Added RTE_ prefix to ETHER_HDR_LEN and ETHER_CRC_LEN.
doc/guides/rel_notes/release_19_08.rst | 1 +
drivers/net/bnxt/Makefile | 3 +
drivers/net/bnxt/bnxt_ethdev.c | 98 -
drivers/net/bnxt/bnxt_ring.h | 3 +-
drivers/net/bnxt/bnxt_rxq.c| 5 +
drivers/net/bnxt/bnxt_rxq.h| 4 +
drivers/net/bnxt/bnxt_rxr.h| 9 +-
drivers/net/bnxt/bnxt_rxtx_vec_sse.c | 481 +
drivers/net/bnxt/bnxt_txr.h| 5 +
drivers/net/bnxt/meson.build | 4 +
10 files changed, 603 insertions(+), 10 deletions(-)
create mode 100644 drivers/net/bnxt/bnxt_rxtx_vec_sse.c
diff --git a/doc/guides/rel_notes/release_19_08.rst
b/doc/guides/rel_notes/release_19_08.rst
index 2b66f41f5..7e5ca3c34 100644
--- a/doc/guides/rel_notes/release_19_08.rst
+++ b/doc/guides/rel_notes/release_19_08.rst
@@ -58,6 +58,7 @@ New Features
Updated the bnxt PMD. The major enhancements include:
* Performance optimizations in non-vector Tx path
+ * Added support for SSE vector mode
Removed Items
-
diff --git a/drivers/net/bnxt/Makefile b/drivers/net/bnxt/Makefile
index 8be3cb0e4..21a565063 100644
--- a/drivers/net/bnxt/Makefile
+++ b/drivers/net/bnxt/Makefile
@@ -41,6 +41,9 @@ SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_vnic.c
SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_irq.c
SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_util.c
SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += rte_pmd_bnxt.c
+ifeq ($(CONFIG_RTE_ARCH_X86), y)
+SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_rxtx_vec_sse.c
+endif
#
# Export include files
diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index 5b3932b7b..bccba445a 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -638,13 +638,73 @@ static int bnxt_scattered_rx(struct rte_eth_dev *eth_dev)
buf_size = (uint16_t)(rte_pktmbuf_data_room_size(rxq->mb_pool) -
RTE_PKTMBUF_HEADROOM);
- if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len +
- VLAN_TAG_SIZE * BNXT_NUM_VLANS > buf_size)
+ if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len > buf_size)
return 1;
}
return 0;
}
+static eth_rx_burst_t
+bnxt_receive_function(__rte_unused struct rte_eth_dev *eth_dev)
+{
+#ifdef RTE_ARCH_X86
+ /*
+* Vector mode receive can be enabled only if scatter rx is not
+* in use and rx offloads are limited to VLAN stripping and
+* CRC stripping.
+*/
+ if (!eth_dev->data->scattered_rx &&
+ !(eth_dev->data->dev_conf.rxmode.offloads &
+ ~(DEV_RX_OFFLOAD_VLAN_STRIP |
+ DEV_RX_OFFLOAD_KEEP_CRC |
+ DEV_RX_OFFLOAD_JUMBO_FRAME |
+ DEV_RX_OFFLOAD_IPV4_CKSUM |
+ DEV_RX_OFFLOAD_UDP_CKSUM |
+ DEV_RX_OFFLOAD_TCP_CKSUM |
+ DEV_RX_OFFLOAD_OUTER_IPV4_CKSUM |
+ DEV_RX_OFFLOAD_VLAN_FILTER))) {
+ PMD_DRV_LOG(INFO, "Using vector mode receive for port %d\n",
+ eth_dev->data->port_id);
+ return bnxt_recv_pkts_vec;
+ }
+ PMD_DRV_LOG(INFO, "Vector mode receive disabled for port %d\n",
+ eth_dev->data->port_id);
+ PMD_DRV_LOG(INFO,
+ "Port %d scatter: %d rx offload: %" PRIX64 "\n",
+ eth_dev->data->port_id,
+ eth_dev->data->scattered_rx,
+ eth_dev->data->dev_conf.rxmode.offloads);
+#endif
+ return bnxt_recv_pkts;
+}
+
+static eth_tx_burst_t
+bnxt_transmit_function(__rte_unused struct rte_eth_dev *eth_dev)
+{
+#ifdef RTE_ARCH_X86
+ /*
+* Vector mode receive can be enabled only if scatter tx is not
+* in use and tx offloads other than VLAN insertion are not
+* in use.
+*/
+ if (!eth_dev->data->scattered_rx &&
+ !(eth_dev->data->dev_conf.txmode.offloads &
+ ~DEV_TX_OFFLOAD_VLAN_INSERT)) {
+ PMD_DRV_LOG(INFO, "Using vector mode transmit for port %d\n",
+ eth_dev->data->port_id);
+ return bnxt_xmit_pkts_vec;
+ }
+ PMD_DRV_LOG(INFO, "Vector mode transmit disabled for port %d\n",
+ eth_dev->data->port_id);
+ PMD_DRV_LOG(INFO,
+ "Port %d scatter: %d tx offload: %" PRIX64 "\n",
+ eth_dev->data->port_id,
+ eth_dev->data->scattered_rx,
[dpdk-dev] [PATCH v3 1/8] net/bnxt: update release notes for bnxt
From: Ajit Khaparde
Update release doc briefly describing updates to bnxt PMD,
including transmit optimization changes in the following
commits:
commit 5c26b1706862 ("net/bnxt: support bulk free of Tx mbufs")
commit f8dea49ef89b ("net/bnxt: optimize Tx batching code")
Signed-off-by: Ajit Khaparde
Signed-off-by: Lance Richardson
---
doc/guides/rel_notes/release_19_08.rst | 4
1 file changed, 4 insertions(+)
diff --git a/doc/guides/rel_notes/release_19_08.rst
b/doc/guides/rel_notes/release_19_08.rst
index 5ebb564b1..2b66f41f5 100644
--- a/doc/guides/rel_notes/release_19_08.rst
+++ b/doc/guides/rel_notes/release_19_08.rst
@@ -54,6 +54,10 @@ New Features
Also, make sure to start the actual text at the margin.
=
+* **Updated the bnxt PMD.**
+
+ Updated the bnxt PMD. The major enhancements include:
+ * Performance optimizations in non-vector Tx path
Removed Items
-
--
2.17.1
[dpdk-dev] [PATCH v3 5/8] net/bnxt: fix RSS RETA indirection table ops
From: Ajit Khaparde
We are trying to update the indirection table for all the VNICs.
We should update the table only for the default vnic0.
Fix the reta update function to only update table entries that are
selected by the update mask. Translate queue number to firmware
group ID when updating an entry.
Fix reta query op to only return table entries as identfied by the
provided mask. Translate firmware group IDs to queue numbers.
Removed extraneous code from bnxt_reta_query_op().
Fixes: d819382543f3 ("net/bnxt: add RSS redirection table operations")
Signed-off-by: Ajit Khaparde
Signed-off-by: Lance Richardson
Reviewed-by: Rahul Gupta
Reviewed-by: Lance Richardson
Reviewed-by: Ajit Khaparde
Cc: sta...@dpdk.org
---
v3:
* Squashed three RSS RETA fix commits into one.
drivers/net/bnxt/bnxt_ethdev.c | 92 +-
drivers/net/bnxt/bnxt_hwrm.c | 1 +
2 files changed, 70 insertions(+), 23 deletions(-)
diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index bccba445a..30dff72a6 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -618,12 +618,6 @@ static void bnxt_print_link_info(struct rte_eth_dev
*eth_dev)
eth_dev->data->port_id);
}
-static int bnxt_dev_lsc_intr_setup(struct rte_eth_dev *eth_dev)
-{
- bnxt_print_link_info(eth_dev);
- return 0;
-}
-
/*
* Determine whether the current configuration requires support for scattered
* receive; return 1 if scattered receive is required and 0 if not.
@@ -977,30 +971,72 @@ static void bnxt_allmulticast_disable_op(struct
rte_eth_dev *eth_dev)
bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
}
+/* Return bnxt_rx_queue pointer corresponding to a given rxq. */
+static struct bnxt_rx_queue *bnxt_qid_to_rxq(struct bnxt *bp, uint16_t qid)
+{
+ if (qid >= bp->rx_nr_rings)
+ return NULL;
+
+ return bp->eth_dev->data->rx_queues[qid];
+}
+
+/* Return rxq corresponding to a given rss table ring/group ID. */
+static uint16_t bnxt_rss_to_qid(struct bnxt *bp, uint16_t fwr)
+{
+ unsigned int i;
+
+ for (i = 0; i < bp->rx_nr_rings; i++) {
+ if (bp->grp_info[i].fw_grp_id == fwr)
+ return i;
+ }
+
+ return INVALID_HW_RING_ID;
+}
+
static int bnxt_reta_update_op(struct rte_eth_dev *eth_dev,
struct rte_eth_rss_reta_entry64 *reta_conf,
uint16_t reta_size)
{
struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
- struct bnxt_vnic_info *vnic;
+ struct bnxt_vnic_info *vnic = &bp->vnic_info[0];
+ uint16_t tbl_size = HW_HASH_INDEX_SIZE;
+ uint16_t idx, sft;
int i;
+ if (!vnic->rss_table)
+ return -EINVAL;
+
if (!(dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG))
return -EINVAL;
- if (reta_size != HW_HASH_INDEX_SIZE) {
+ if (reta_size != tbl_size) {
PMD_DRV_LOG(ERR, "The configured hash table lookup size "
"(%d) must equal the size supported by the hardware "
- "(%d)\n", reta_size, HW_HASH_INDEX_SIZE);
+ "(%d)\n", reta_size, tbl_size);
return -EINVAL;
}
- /* Update the RSS VNIC(s) */
- for (i = 0; i < bp->max_vnics; i++) {
- vnic = &bp->vnic_info[i];
- memcpy(vnic->rss_table, reta_conf, reta_size);
- bnxt_hwrm_vnic_rss_cfg(bp, vnic);
+
+ for (i = 0; i < reta_size; i++) {
+ struct bnxt_rx_queue *rxq;
+
+ idx = i / RTE_RETA_GROUP_SIZE;
+ sft = i % RTE_RETA_GROUP_SIZE;
+
+ if (!(reta_conf[idx].mask & (1ULL << sft)))
+ continue;
+
+ rxq = bnxt_qid_to_rxq(bp, reta_conf[idx].reta[sft]);
+ if (!rxq) {
+ PMD_DRV_LOG(ERR, "Invalid ring in reta_conf.\n");
+ return -EINVAL;
+ }
+
+ vnic->rss_table[i] =
+ vnic->fw_grp_ids[reta_conf[idx].reta[sft]];
}
+
+ bnxt_hwrm_vnic_rss_cfg(bp, vnic);
return 0;
}
@@ -1010,8 +1046,8 @@ static int bnxt_reta_query_op(struct rte_eth_dev *eth_dev,
{
struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
struct bnxt_vnic_info *vnic = &bp->vnic_info[0];
- struct rte_intr_handle *intr_handle
- = &bp->pdev->intr_handle;
+ uint16_t tbl_size = HW_HASH_INDEX_SIZE;
+ uint16_t idx, sft, i;
/* Retrieve from the default VNIC */
if (!vnic)
@@ -1019,18 +1055,28 @@ static int bnxt_reta_query_op(struct rte_eth_dev
*eth_dev,
if (!vnic->rss_table)
return -EINVAL;
- if (reta_size != HW_HASH_INDEX_SIZE) {
+ if (reta_size != tbl_size) {
[dpdk-dev] [PATCH v3 6/8] net/bnxt: update HWRM API to version 1.10.0.19
From: Ajit Khaparde
Update HWRM API to version 1.10.0.19
Signed-off-by: Ajit Khaparde
Reviewed-by: Lance Richardson
Signed-off-by: Lance Richardson
---
v3:
* Fixed headline in commit log.
drivers/net/bnxt/hsi_struct_def_dpdk.h | 3112 +++-
1 file changed, 2514 insertions(+), 598 deletions(-)
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h
b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index e80057936..ea9a7d40e 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -1,5 +1,5 @@
/* SPDX-License-Identifier: BSD-3-Clause
- * Copyright (c) 2014-2018 Broadcom Limited
+ * Copyright (c) 2014-2019 Broadcom Limited
* All rights reserved.
*
* DO NOT MODIFY!!! This file is automatically generated.
@@ -68,9 +68,9 @@ struct hwrm_resp_hdr {
/* RoCE slow path command */
#define TLV_TYPE_ROCE_SP_COMMAND UINT32_C(0x3)
/* RoCE slow path command to query CC Gen1 support. */
-#define TLV_TYPE_QUERY_ROCE_CC_GEN1 UINT32_C(0xcommand 0x0005)
+#define TLV_TYPE_QUERY_ROCE_CC_GEN1 UINT32_C(0x4)
/* RoCE slow path command to modify CC Gen1 support. */
-#define TLV_TYPE_MODIFY_ROCE_CC_GEN1 UINT32_C(0xcommand 0x0005)
+#define TLV_TYPE_MODIFY_ROCE_CC_GEN1 UINT32_C(0x5)
/* Engine CKV - The device's serial number. */
#define TLV_TYPE_ENGINE_CKV_DEVICE_SERIAL_NUMBER UINT32_C(0x8001)
/* Engine CKV - Per-function random nonce data. */
@@ -366,6 +366,7 @@ struct cmd_nums {
#define HWRM_TUNNEL_DST_PORT_QUERYUINT32_C(0xa0)
#define HWRM_TUNNEL_DST_PORT_ALLOCUINT32_C(0xa1)
#define HWRM_TUNNEL_DST_PORT_FREE UINT32_C(0xa2)
+ #define HWRM_STAT_CTX_ENG_QUERY UINT32_C(0xaf)
#define HWRM_STAT_CTX_ALLOC UINT32_C(0xb0)
#define HWRM_STAT_CTX_FREEUINT32_C(0xb1)
#define HWRM_STAT_CTX_QUERY UINT32_C(0xb2)
@@ -442,6 +443,14 @@ struct cmd_nums {
#define HWRM_FW_IPC_MSG UINT32_C(0x110)
#define HWRM_CFA_REDIRECT_TUNNEL_TYPE_INFOUINT32_C(0x111)
#define HWRM_CFA_REDIRECT_QUERY_TUNNEL_TYPE UINT32_C(0x112)
+ /* Experimental */
+ #define HWRM_CFA_FLOW_AGING_TIMER_RESET UINT32_C(0x113)
+ /* Experimental */
+ #define HWRM_CFA_FLOW_AGING_CFG UINT32_C(0x114)
+ /* Experimental */
+ #define HWRM_CFA_FLOW_AGING_QCFG UINT32_C(0x115)
+ /* Experimental */
+ #define HWRM_CFA_FLOW_AGING_QCAPS UINT32_C(0x116)
/* Engine CKV - Ping the device and SRT firmware to get the public key.
*/
#define HWRM_ENGINE_CKV_HELLO UINT32_C(0x12d)
/* Engine CKV - Get the current allocation status of keys provisioned
in the key vault. */
@@ -591,61 +600,79 @@ struct cmd_nums {
struct ret_codes {
uint16_terror_code;
/* Request was successfully executed by the HWRM. */
- #define HWRM_ERR_CODE_SUCCESSUINT32_C(0x0)
+ #define HWRM_ERR_CODE_SUCCESS UINT32_C(0x0)
/* The HWRM failed to execute the request. */
- #define HWRM_ERR_CODE_FAIL UINT32_C(0x1)
+ #define HWRM_ERR_CODE_FAIL UINT32_C(0x1)
/*
* The request contains invalid argument(s) or input
* parameters.
*/
- #define HWRM_ERR_CODE_INVALID_PARAMS UINT32_C(0x2)
+ #define HWRM_ERR_CODE_INVALID_PARAMSUINT32_C(0x2)
/*
* The requester is not allowed to access the requested
* resource. This error code shall be provided in a
* response to a request to query or modify an existing
* resource that is not accessible by the requester.
*/
- #define HWRM_ERR_CODE_RESOURCE_ACCESS_DENIED UINT32_C(0x3)
+ #define HWRM_ERR_CODE_RESOURCE_ACCESS_DENIEDUINT32_C(0x3)
/*
* The HWRM is unable to allocate the requested resource.
* This code only applies to requests for HWRM resource
* allocations.
*/
- #define HWRM_ERR_CODE_RESOURCE_ALLOC_ERROR UINT32_C(0x4)
+ #define HWRM_ERR_CODE_RESOURCE_ALLOC_ERROR UINT32_C(0x4)
/*
* Invalid combination of flags is specified in the
* request.
*/
- #define HWRM_ERR_CODE_INVALID_FLAGS UINT32_C(0x5)
+ #define HWRM_ERR_CODE_INVALID_FLAGS UINT32_C(0x5)
/*
* Invalid combination of enables fields is specified in
* the request.
*/
- #define HWRM_ERR_CODE_INVALID_ENABLESUINT32_C(0x6)
+ #define HWRM_ERR_CODE_INVALID_ENABLES UINT32_C(0x6)
/*
* Request contains a required TLV that is not supported by
* the installed version o
[dpdk-dev] [PATCH v3 7/8] net/bnxt: update HWRM API to version 1.10.0.48
From: Ajit Khaparde
Update HWRM version to 1.10.0.48
Signed-off-by: Ajit Khaparde
Reviewed-by: Lance Richardson
Signed-off-by: Lance Richardson
---
v3:
* Fixed headline in commit log.
drivers/net/bnxt/hsi_struct_def_dpdk.h | 3156 +---
1 file changed, 2874 insertions(+), 282 deletions(-)
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h
b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index ea9a7d40e..00a8ff87e 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -260,6 +260,7 @@ struct cmd_nums {
*/
uint16_treq_type;
#define HWRM_VER_GET UINT32_C(0x0)
+ #define HWRM_ERROR_RECOVERY_QCFG UINT32_C(0xc)
#define HWRM_FUNC_DRV_IF_CHANGE UINT32_C(0xd)
#define HWRM_FUNC_BUF_UNRGTR UINT32_C(0xe)
#define HWRM_FUNC_VF_CFG UINT32_C(0xf)
@@ -372,6 +373,8 @@ struct cmd_nums {
#define HWRM_STAT_CTX_QUERY UINT32_C(0xb2)
#define HWRM_STAT_CTX_CLR_STATS UINT32_C(0xb3)
#define HWRM_PORT_QSTATS_EXT UINT32_C(0xb4)
+ #define HWRM_PORT_PHY_MDIO_WRITE UINT32_C(0xb5)
+ #define HWRM_PORT_PHY_MDIO_READ UINT32_C(0xb6)
#define HWRM_FW_RESET UINT32_C(0xc0)
#define HWRM_FW_QSTATUS UINT32_C(0xc1)
#define HWRM_FW_HEALTH_CHECK UINT32_C(0xc2)
@@ -397,6 +400,8 @@ struct cmd_nums {
#define HWRM_WOL_FILTER_QCFG UINT32_C(0xf2)
#define HWRM_WOL_REASON_QCFG UINT32_C(0xf3)
/* Experimental */
+ #define HWRM_CFA_METER_QCAPS UINT32_C(0xf4)
+ /* Experimental */
#define HWRM_CFA_METER_PROFILE_ALLOC UINT32_C(0xf5)
/* Experimental */
#define HWRM_CFA_METER_PROFILE_FREE UINT32_C(0xf6)
@@ -407,6 +412,8 @@ struct cmd_nums {
/* Experimental */
#define HWRM_CFA_METER_INSTANCE_FREE UINT32_C(0xf9)
/* Experimental */
+ #define HWRM_CFA_METER_INSTANCE_CFG UINT32_C(0xfa)
+ /* Experimental */
#define HWRM_CFA_VFR_ALLOCUINT32_C(0xfd)
/* Experimental */
#define HWRM_CFA_VFR_FREE UINT32_C(0xfe)
@@ -451,6 +458,34 @@ struct cmd_nums {
#define HWRM_CFA_FLOW_AGING_QCFG UINT32_C(0x115)
/* Experimental */
#define HWRM_CFA_FLOW_AGING_QCAPS UINT32_C(0x116)
+ /* Experimental */
+ #define HWRM_CFA_CTX_MEM_RGTR UINT32_C(0x117)
+ /* Experimental */
+ #define HWRM_CFA_CTX_MEM_UNRGTR UINT32_C(0x118)
+ /* Experimental */
+ #define HWRM_CFA_CTX_MEM_QCTX UINT32_C(0x119)
+ /* Experimental */
+ #define HWRM_CFA_CTX_MEM_QCAPSUINT32_C(0x11a)
+ /* Experimental */
+ #define HWRM_CFA_COUNTER_QCAPSUINT32_C(0x11b)
+ /* Experimental */
+ #define HWRM_CFA_COUNTER_CFG UINT32_C(0x11c)
+ /* Experimental */
+ #define HWRM_CFA_COUNTER_QCFG UINT32_C(0x11d)
+ /* Experimental */
+ #define HWRM_CFA_COUNTER_QSTATS UINT32_C(0x11e)
+ /* Experimental */
+ #define HWRM_CFA_TCP_FLAG_PROCESS_QCFGUINT32_C(0x11f)
+ /* Experimental */
+ #define HWRM_CFA_EEM_QCAPSUINT32_C(0x120)
+ /* Experimental */
+ #define HWRM_CFA_EEM_CFG UINT32_C(0x121)
+ /* Experimental */
+ #define HWRM_CFA_EEM_QCFG UINT32_C(0x122)
+ /* Experimental */
+ #define HWRM_CFA_EEM_OP UINT32_C(0x123)
+ /* Experimental */
+ #define HWRM_CFA_ADV_FLOW_MGNT_QCAPS UINT32_C(0x124)
/* Engine CKV - Ping the device and SRT firmware to get the public key.
*/
#define HWRM_ENGINE_CKV_HELLO UINT32_C(0x12d)
/* Engine CKV - Get the current allocation status of keys provisioned
in the key vault. */
@@ -469,6 +504,8 @@ struct cmd_nums {
#define HWRM_ENGINE_CKV_RNG_GET UINT32_C(0x134)
/* Engine CKV - Generate and encrypt a new AES key. */
#define HWRM_ENGINE_CKV_KEY_GEN UINT32_C(0x135)
+ /* Engine CKV - Configure a label index with a label value. */
+ #define HWRM_ENGINE_CKV_KEY_LABEL_CFG UINT32_C(0x136)
/* Engine - Query the available queue groups configuration. */
#define HWRM_ENGINE_QG_CONFIG_QUERY UINT32_C(0x13c)
/* Engine - Query the queue groups assigne
[dpdk-dev] [PATCH v3 8/8] net/bnxt: update HWRM API to version 1.10.0.74
From: Ajit Khaparde
Update HWRM API to version 1.10.0.74
Signed-off-by: Ajit Khaparde
Reviewed-by: Lance Richardson
Signed-off-by: Lance Richardson
---
v3:
* Fixed headline in commit log.
doc/guides/rel_notes/release_19_08.rst |1 +
drivers/net/bnxt/hsi_struct_def_dpdk.h | 1399
2 files changed, 1169 insertions(+), 231 deletions(-)
diff --git a/doc/guides/rel_notes/release_19_08.rst
b/doc/guides/rel_notes/release_19_08.rst
index 7e5ca3c34..ea8fbcdfd 100644
--- a/doc/guides/rel_notes/release_19_08.rst
+++ b/doc/guides/rel_notes/release_19_08.rst
@@ -59,6 +59,7 @@ New Features
Updated the bnxt PMD. The major enhancements include:
* Performance optimizations in non-vector Tx path
* Added support for SSE vector mode
+ * Updated HWRM API to version 1.10.0.74
Removed Items
-
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h
b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index 00a8ff87e..305ee09bc 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -1,5 +1,5 @@
/* SPDX-License-Identifier: BSD-3-Clause
- * Copyright (c) 2014-2019 Broadcom Limited
+ * Copyright (c) 2014-2019 Broadcom Inc.
* All rights reserved.
*
* DO NOT MODIFY!!! This file is automatically generated.
@@ -27,7 +27,8 @@ struct hwrm_cmd_hdr {
/*
* The target ID of the command:
* * 0x0-0xFFF8 - The function ID
-* * 0xFFF8-0xFFFE - Reserved for internal processors
+* * 0xFFF8-0xFFFC, 0xFFFE - Reserved for internal processors
+* * 0xFFFD - Reserved for user-space HWRM interface
* * 0x - HWRM
*/
uint16_ttarget_id;
@@ -221,8 +222,14 @@ struct hwrm_short_input {
#define HWRM_SHORT_INPUT_SIGNATURE_SHORT_CMD UINT32_C(0x4321)
#define HWRM_SHORT_INPUT_SIGNATURE_LAST \
HWRM_SHORT_INPUT_SIGNATURE_SHORT_CMD
- /* Reserved for future use. */
- uint16_tunused_0;
+ /* The target ID of the command */
+ uint16_ttarget_id;
+ /* Default target_id (0x0) to maintain compatibility with old driver */
+ #define HWRM_SHORT_INPUT_TARGET_ID_DEFAULT UINT32_C(0x0)
+ /* Reserved for user-space HWRM interface */
+ #define HWRM_SHORT_INPUT_TARGET_ID_TOOLS UINT32_C(0xfffd)
+ #define HWRM_SHORT_INPUT_TARGET_ID_LAST \
+ HWRM_SHORT_INPUT_TARGET_ID_TOOLS
/* This value indicates the length of the request. */
uint16_tsize;
/*
@@ -394,6 +401,8 @@ struct cmd_nums {
#define HWRM_FWD_RESP UINT32_C(0xd2)
#define HWRM_FWD_ASYNC_EVENT_CMPL UINT32_C(0xd3)
#define HWRM_OEM_CMD UINT32_C(0xd4)
+ /* Tells the fw to run PRBS test on a given port and lane. */
+ #define HWRM_PORT_PRBS_TEST UINT32_C(0xd5)
#define HWRM_TEMP_MONITOR_QUERY UINT32_C(0xe0)
#define HWRM_WOL_FILTER_ALLOC UINT32_C(0xf0)
#define HWRM_WOL_FILTER_FREE UINT32_C(0xf1)
@@ -486,6 +495,8 @@ struct cmd_nums {
#define HWRM_CFA_EEM_OP UINT32_C(0x123)
/* Experimental */
#define HWRM_CFA_ADV_FLOW_MGNT_QCAPS UINT32_C(0x124)
+ /* Experimental */
+ #define HWRM_CFA_TFLIBUINT32_C(0x125)
/* Engine CKV - Ping the device and SRT firmware to get the public key.
*/
#define HWRM_ENGINE_CKV_HELLO UINT32_C(0x12d)
/* Engine CKV - Get the current allocation status of keys provisioned
in the key vault. */
@@ -506,6 +517,8 @@ struct cmd_nums {
#define HWRM_ENGINE_CKV_KEY_GEN UINT32_C(0x135)
/* Engine CKV - Configure a label index with a label value. */
#define HWRM_ENGINE_CKV_KEY_LABEL_CFG UINT32_C(0x136)
+ /* Engine CKV - Query a label */
+ #define HWRM_ENGINE_CKV_KEY_LABEL_QCFGUINT32_C(0x137)
/* Engine - Query the available queue groups configuration. */
#define HWRM_ENGINE_QG_CONFIG_QUERY UINT32_C(0x13c)
/* Engine - Query the queue groups assigned to a function. */
@@ -587,6 +600,19 @@ struct cmd_nums {
/* Experimental */
#define HWRM_PCIE_QSTATS UINT32_C(0x204)
/* Experimental */
+ #define HWRM_MFG_FRU_WRITE_CONTROLUINT32_C(0x205)
+ /* Returns the current value of a free running counter from the device.
*/
+ #define HWRM_MFG_TIMERS_QUERY UINT32_C(0x206)
+ /* Experimental */
+ #define HWRM_MFG_OTP_CFG UINT32_C(0x207)
+ /* Experimental */
+ #define HWRM_MFG_OTP_QCFG UINT32_C(0x208)
+ /*
+* Tells the fw to run the DMA read from the host and DMA write
+
[dpdk-dev] [PATCH] net/i40e: remove queue_stats_mapping_set
This driver was inserting its own stub for queue_stats_mapping which
did nothing but cause this device to return a different errno
than every other device driver. All devices that don't implement
queue stats mapping should return the same error.
Maybe the plan originally was to implement something, if that
ever happens, just put in the right code.
Fixes: 4861cde46116 ("i40e: new poll mode driver")
Signed-off-by: Stephen Hemminger
---
drivers/net/i40e/i40e_ethdev.c | 16
1 file changed, 16 deletions(-)
diff --git a/drivers/net/i40e/i40e_ethdev.c b/drivers/net/i40e/i40e_ethdev.c
index cab440f5a8f6..5bbcbb4fc608 100644
--- a/drivers/net/i40e/i40e_ethdev.c
+++ b/drivers/net/i40e/i40e_ethdev.c
@@ -237,10 +237,6 @@ static int i40e_dev_xstats_get_names(struct rte_eth_dev
*dev,
struct rte_eth_xstat_name *xstats_names,
unsigned limit);
static void i40e_dev_stats_reset(struct rte_eth_dev *dev);
-static int i40e_dev_queue_stats_mapping_set(struct rte_eth_dev *dev,
- uint16_t queue_id,
- uint8_t stat_idx,
- uint8_t is_rx);
static int i40e_fw_version_get(struct rte_eth_dev *dev,
char *fw_version, size_t fw_size);
static void i40e_dev_info_get(struct rte_eth_dev *dev,
@@ -456,7 +452,6 @@ static const struct eth_dev_ops i40e_eth_dev_ops = {
.xstats_get_names = i40e_dev_xstats_get_names,
.stats_reset = i40e_dev_stats_reset,
.xstats_reset = i40e_dev_stats_reset,
- .queue_stats_mapping_set = i40e_dev_queue_stats_mapping_set,
.fw_version_get = i40e_fw_version_get,
.dev_infos_get= i40e_dev_info_get,
.dev_supported_ptypes_get = i40e_dev_supported_ptypes_get,
@@ -3431,17 +3426,6 @@ i40e_dev_xstats_get(struct rte_eth_dev *dev, struct
rte_eth_xstat *xstats,
return count;
}
-static int
-i40e_dev_queue_stats_mapping_set(__rte_unused struct rte_eth_dev *dev,
-__rte_unused uint16_t queue_id,
-__rte_unused uint8_t stat_idx,
-__rte_unused uint8_t is_rx)
-{
- PMD_INIT_FUNC_TRACE();
-
- return -ENOSYS;
-}
-
static int
i40e_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size)
{
--
2.20.1
Re: [dpdk-dev] [PATCH 00/12] rxq q_errors[] statistics fixes
On 3/4/2019 11:18 AM, David Marchand wrote: > According to the api, the q_errors[] per queue statistic is for reception > errors not transmit errors. > This is a first cleanup on statistics before looking at oerrors. > The fix is correct according the documentation but it was waiting for a solution to capture the Tx queue errors which we are removing this support from some PMDs because they were storing this information into wrong field. Since the affected stats are Tx queue error stats, and all Tx errors still can be stored on 'oerrors' I am for getting the fix, although we need to figure out how to store Tx queue error stats. For series, Reviewed-by: Ferruh Yigit Series applied to dpdk-next-net/master, thanks.
[dpdk-dev] [PATCH v2 00/11] A new net PMD - hinic
This patch set adds support of a new net PMD for Huawei Intelligent nic. This patch provides supoort for basic RX/TX and the contorl path needed for it. Later on new features will be added like VLAN, VFs, etc. Basic features: 1. Basic device operations: probe, initialization, start/stop, configure, info get. 2. RX/TX queue operations: setup/release, start/stop. 3. RX/TX. Stats: 1. statistics and extended statistics. --- v2: - Fix arm64 compilation issue. - Fix some checkpatch checks issues - Fix patches thread issue. - Fit to the newest rte_ prefix patch Ziyang Xuan (11): net/hinic/base: add registers for Huawei Hi1822 NIC net/hinic/base: add command channels code net/hinic/base: add mgmt module interactive code net/hinic/base: add code about hardware operation net/hinic/base: add eqs and context code net/hinic/base: add code for nic business net/hinic/base: add various headers net/hinic: add hinic PMD build and doc files net/hinic: add RX module net/hinic: add TX module net/hinic: add support for basic device operations MAINTAINERS |9 + config/common_base|5 + doc/guides/nics/features/hinic.ini| 32 + doc/guides/nics/hinic.rst | 49 + drivers/net/Makefile |1 + drivers/net/hinic/Makefile| 94 + drivers/net/hinic/base/hinic_compat.h | 252 ++ drivers/net/hinic/base/hinic_csr.h| 164 ++ drivers/net/hinic/base/hinic_ctx_def.h| 195 ++ drivers/net/hinic/base/hinic_logs.c | 16 + drivers/net/hinic/base/hinic_logs.h | 31 + drivers/net/hinic/base/hinic_pmd_api_cmd.c| 1040 drivers/net/hinic/base/hinic_pmd_api_cmd.h| 271 +++ drivers/net/hinic/base/hinic_pmd_cfg.c| 269 ++ drivers/net/hinic/base/hinic_pmd_cfg.h| 264 ++ drivers/net/hinic/base/hinic_pmd_cmdq.c | 902 +++ drivers/net/hinic/base/hinic_pmd_cmdq.h | 190 ++ drivers/net/hinic/base/hinic_pmd_dpdev.h | 149 ++ drivers/net/hinic/base/hinic_pmd_eqs.c| 725 ++ drivers/net/hinic/base/hinic_pmd_eqs.h| 94 + drivers/net/hinic/base/hinic_pmd_hw.h | 49 + drivers/net/hinic/base/hinic_pmd_hw_mgmt.h| 85 + drivers/net/hinic/base/hinic_pmd_hwdev.c | 1544 drivers/net/hinic/base/hinic_pmd_hwdev.h | 207 ++ drivers/net/hinic/base/hinic_pmd_hwif.c | 542 + drivers/net/hinic/base/hinic_pmd_hwif.h | 90 + drivers/net/hinic/base/hinic_pmd_mgmt.c | 619 + drivers/net/hinic/base/hinic_pmd_mgmt.h | 125 + .../net/hinic/base/hinic_pmd_mgmt_interface.h | 503 drivers/net/hinic/base/hinic_pmd_nic.h| 85 + drivers/net/hinic/base/hinic_pmd_niccfg.c | 1408 +++ drivers/net/hinic/base/hinic_pmd_niccfg.h | 333 +++ drivers/net/hinic/base/hinic_pmd_nicio.c | 920 +++ drivers/net/hinic/base/hinic_pmd_nicio.h | 53 + drivers/net/hinic/base/hinic_pmd_qp.c | 26 + drivers/net/hinic/base/hinic_pmd_qp.h | 76 + drivers/net/hinic/base/hinic_pmd_wq.c | 164 ++ drivers/net/hinic/base/hinic_pmd_wq.h | 52 + drivers/net/hinic/base/hinic_port_cmd.h | 494 drivers/net/hinic/base/hinic_qe_def.h | 461 drivers/net/hinic/base/meson.build| 63 + drivers/net/hinic/hinic_pmd_dpdev.c | 718 ++ drivers/net/hinic/hinic_pmd_ethdev.c | 2157 + drivers/net/hinic/hinic_pmd_ethdev.h | 106 + drivers/net/hinic/hinic_pmd_rx.c | 919 +++ drivers/net/hinic/hinic_pmd_rx.h | 135 ++ drivers/net/hinic/hinic_pmd_tx.c | 1070 drivers/net/hinic/hinic_pmd_tx.h | 97 + drivers/net/hinic/meson.build | 18 + drivers/net/hinic/rte_pmd_hinic_version.map |4 + mk/rte.app.mk |1 + 51 files changed, 17876 insertions(+) create mode 100644 doc/guides/nics/features/hinic.ini create mode 100644 doc/guides/nics/hinic.rst create mode 100644 drivers/net/hinic/Makefile create mode 100644 drivers/net/hinic/base/hinic_compat.h create mode 100644 drivers/net/hinic/base/hinic_csr.h create mode 100644 drivers/net/hinic/base/hinic_ctx_def.h create mode 100644 drivers/net/hinic/base/hinic_logs.c create mode 100644 drivers/net/hinic/base/hinic_logs.h create mode 100644 drivers/net/hinic/base/hinic_pmd_api_cmd.c create mode 100644 drivers/net/hinic/base/hinic_pmd_api_cmd.h create mode 100644 drivers/net/hinic/base/hinic_pmd_cfg.c create mode 100644 drivers/net/hinic/base/hinic_pmd_cfg.h create mode 100644 drivers/net/hinic/base/hinic_pmd_cmdq.c create mode 100644 drivers/net/hinic/base/hinic_pmd_cmdq.h create mode 100644 drivers/net/hinic/base/hinic_pmd_dpdev.h create mode 100644 drivers/net/hinic/base/hinic_pmd_eqs.c creat
[dpdk-dev] [PATCH v2 01/11] net/hinic/base: add registers for Huawei Hi1822 NIC
Add the registers that comprise the Huawei Hi1822 NIC. There is no functionality in this patch. Signed-off-by: Ziyang Xuan --- drivers/net/hinic/base/hinic_csr.h | 164 + 1 file changed, 164 insertions(+) create mode 100644 drivers/net/hinic/base/hinic_csr.h diff --git a/drivers/net/hinic/base/hinic_csr.h b/drivers/net/hinic/base/hinic_csr.h new file mode 100644 index 0..55af3a861 --- /dev/null +++ b/drivers/net/hinic/base/hinic_csr.h @@ -0,0 +1,164 @@ +/* SPDX-License-Identifier: BSD-3-Clause + * Copyright(c) 2017 Huawei Technologies Co., Ltd + */ + +#ifndef _HINIC_CSR_H_ +#define _HINIC_CSR_H_ + +#define HINIC_CSR_GLOBAL_BASE_ADDR 0x4000 + +/* HW interface registers */ +#define HINIC_CSR_FUNC_ATTR0_ADDR 0x0 +#define HINIC_CSR_FUNC_ATTR1_ADDR 0x4 +#define HINIC_CSR_FUNC_ATTR2_ADDR 0x8 +#define HINIC_CSR_FUNC_ATTR4_ADDR 0x10 +#define HINIC_CSR_FUNC_ATTR5_ADDR 0x14 + +#define HINIC_FUNC_CSR_MAILBOX_DATA_OFF0x80 +#define HINIC_FUNC_CSR_MAILBOX_CONTROL_OFF 0x0100 +#define HINIC_FUNC_CSR_MAILBOX_INT_OFFSET_OFF 0x0104 +#define HINIC_FUNC_CSR_MAILBOX_RESULT_H_OFF0x0108 +#define HINIC_FUNC_CSR_MAILBOX_RESULT_L_OFF0x010C + +#define HINIC_CSR_DMA_ATTR_TBL_BASE0xC80 + +#define HINIC_ELECTION_BASE0x200 + +#define HINIC_CSR_DMA_ATTR_TBL_STRIDE 0x4 +#define HINIC_CSR_DMA_ATTR_TBL_ADDR(idx) \ + (HINIC_CSR_DMA_ATTR_TBL_BASE\ + + (idx) * HINIC_CSR_DMA_ATTR_TBL_STRIDE) + +#define HINIC_PPF_ELECTION_STRIDE 0x4 +#define HINIC_CSR_MAX_PORTS4 +#define HINIC_CSR_PPF_ELECTION_ADDR\ + (HINIC_CSR_GLOBAL_BASE_ADDR + HINIC_ELECTION_BASE) + +#define HINIC_CSR_GLOBAL_MPF_ELECTION_ADDR \ + (HINIC_CSR_GLOBAL_BASE_ADDR + HINIC_ELECTION_BASE + \ + HINIC_CSR_MAX_PORTS * HINIC_PPF_ELECTION_STRIDE) + +/* MSI-X registers */ +#define HINIC_CSR_MSIX_CTRL_BASE 0x2000 +#define HINIC_CSR_MSIX_CNT_BASE0x2004 + +#define HINIC_CSR_MSIX_STRIDE 0x8 + +#define HINIC_CSR_MSIX_CTRL_ADDR(idx) \ + (HINIC_CSR_MSIX_CTRL_BASE + (idx) * HINIC_CSR_MSIX_STRIDE) + +#define HINIC_CSR_MSIX_CNT_ADDR(idx) \ + (HINIC_CSR_MSIX_CNT_BASE + (idx) * HINIC_CSR_MSIX_STRIDE) + +/* EQ registers */ +#define HINIC_AEQ_MTT_OFF_BASE_ADDR0x200 +#define HINIC_CEQ_MTT_OFF_BASE_ADDR0x400 + +#define HINIC_EQ_MTT_OFF_STRIDE0x40 + +#define HINIC_CSR_AEQ_MTT_OFF(id) \ + (HINIC_AEQ_MTT_OFF_BASE_ADDR + (id) * HINIC_EQ_MTT_OFF_STRIDE) + +#define HINIC_CSR_CEQ_MTT_OFF(id) \ + (HINIC_CEQ_MTT_OFF_BASE_ADDR + (id) * HINIC_EQ_MTT_OFF_STRIDE) + +#define HINIC_CSR_EQ_PAGE_OFF_STRIDE 8 + +#define HINIC_AEQ_HI_PHYS_ADDR_REG(q_id, pg_num) \ + (HINIC_CSR_AEQ_MTT_OFF(q_id) + \ + (pg_num) * HINIC_CSR_EQ_PAGE_OFF_STRIDE) + +#define HINIC_AEQ_LO_PHYS_ADDR_REG(q_id, pg_num) \ + (HINIC_CSR_AEQ_MTT_OFF(q_id) + \ + (pg_num) * HINIC_CSR_EQ_PAGE_OFF_STRIDE + 4) + +#define HINIC_CEQ_HI_PHYS_ADDR_REG(q_id, pg_num) \ + (HINIC_CSR_CEQ_MTT_OFF(q_id) + \ + (pg_num) * HINIC_CSR_EQ_PAGE_OFF_STRIDE) + +#define HINIC_CEQ_LO_PHYS_ADDR_REG(q_id, pg_num) \ + (HINIC_CSR_CEQ_MTT_OFF(q_id) + \ + (pg_num) * HINIC_CSR_EQ_PAGE_OFF_STRIDE + 4) + +#define HINIC_EQ_HI_PHYS_ADDR_REG(type, q_id, pg_num) \ + ((u32)((type == HINIC_AEQ) ? \ + HINIC_AEQ_HI_PHYS_ADDR_REG(q_id, pg_num) : \ + HINIC_CEQ_HI_PHYS_ADDR_REG(q_id, pg_num))) + +#define HINIC_EQ_LO_PHYS_ADDR_REG(type, q_id, pg_num) \ + ((u32)((type == HINIC_AEQ) ? \ + HINIC_AEQ_LO_PHYS_ADDR_REG(q_id, pg_num) : \ + HINIC_CEQ_LO_PHYS_ADDR_REG(q_id, pg_num))) + +#define HINIC_AEQ_CTRL_0_ADDR_BASE 0xE00 +#define HINIC_AEQ_CTRL_1_ADDR_BASE 0xE04 +#define HINIC_AEQ_CONS_IDX_0_ADDR_BASE 0xE08 +#define HINIC_AEQ_CONS_IDX_1_ADDR_BASE 0xE0C + +#define HINIC_EQ_OFF_STRIDE0x80 + +#define HINIC_CSR_AEQ_CTRL_0_ADDR(idx) \ + (HINIC_AEQ_CTRL_0_ADDR_BASE + (idx) * HINIC_EQ_OFF_STRIDE) + +#define HINIC_CSR_AEQ_CTRL_1_ADDR(idx) \ + (HINIC_AEQ_CTRL_1_ADDR_BASE + (idx) * HINIC_EQ_OFF_STRIDE) + +#define HINIC_CSR_AEQ_CONS_IDX_ADDR(idx) \ + (HINIC_AEQ_CONS_IDX_0_ADDR_BASE + (idx) * HINIC_EQ_OFF_STRIDE) + +#define HIN
[dpdk-dev] [PATCH v2 02/11] net/hinic/base: add command channels code
Admin channels include api channel and command queue,
Api channel is for mgmt module. And command queue is
for ucode module.
Signed-off-by: Ziyang Xuan
---
drivers/net/hinic/base/hinic_pmd_api_cmd.c | 1040
drivers/net/hinic/base/hinic_pmd_api_cmd.h | 271 +
drivers/net/hinic/base/hinic_pmd_cmdq.c| 902 +
drivers/net/hinic/base/hinic_pmd_cmdq.h| 190
4 files changed, 2403 insertions(+)
create mode 100644 drivers/net/hinic/base/hinic_pmd_api_cmd.c
create mode 100644 drivers/net/hinic/base/hinic_pmd_api_cmd.h
create mode 100644 drivers/net/hinic/base/hinic_pmd_cmdq.c
create mode 100644 drivers/net/hinic/base/hinic_pmd_cmdq.h
diff --git a/drivers/net/hinic/base/hinic_pmd_api_cmd.c
b/drivers/net/hinic/base/hinic_pmd_api_cmd.c
new file mode 100644
index 0..a2a11fa1c
--- /dev/null
+++ b/drivers/net/hinic/base/hinic_pmd_api_cmd.c
@@ -0,0 +1,1040 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2017 Huawei Technologies Co., Ltd
+ */
+
+#include "hinic_pmd_dpdev.h"
+
+#define API_CMD_CHAIN_CELL_SIZE_SHIFT 6U
+
+#define API_CMD_CELL_DESC_SIZE 8
+#define API_CMD_CELL_DATA_ADDR_SIZE8
+
+#define API_CHAIN_NUM_CELLS32
+#define API_CHAIN_CELL_SIZE128
+#define API_CHAIN_RSP_DATA_SIZE128
+
+#define API_CHAIN_CELL_ALIGNMENT 8
+
+#define API_CMD_TIMEOUT1
+
+#define API_CMD_BUF_SIZE 2048UL
+
+#define API_CMD_NODE_ALIGN_SIZE512UL
+#define API_PAYLOAD_ALIGN_SIZE 64
+
+#define API_CHAIN_RESP_ALIGNMENT 64ULL
+
+#define COMPLETION_TIMEOUT_DEFAULT 1000UL
+#define POLLING_COMPLETION_TIMEOUT_DEFAULT 1000U
+
+#define API_CMD_RESPONSE_DATA_PADDR(val) be64_to_cpu(*((u64 *)(val)))
+
+#define READ_API_CMD_PRIV_DATA(id, token) (((id) << 16) + (token))
+#define WRITE_API_CMD_PRIV_DATA(id)((id) << 16)
+
+#define MASKED_IDX(chain, idx) ((idx) & ((chain)->num_cells - 1))
+
+#undef SIZE_4BYTES
+#undef SIZE_8BYTES
+#define SIZE_4BYTES(size) (ALIGN((u32)(size), 4U) >> 2)
+#define SIZE_8BYTES(size) (ALIGN((u32)(size), 8U) >> 3)
+
+enum api_cmd_data_format {
+ SGL_DATA = 1,
+};
+
+enum api_cmd_type {
+ API_CMD_WRITE_TYPE = 0,
+ API_CMD_READ_TYPE = 1,
+};
+
+enum api_cmd_bypass {
+ NOT_BYPASS = 0,
+ BYPASS = 1,
+};
+
+enum api_cmd_resp_aeq {
+ NOT_TRIGGER = 0,
+ TRIGGER = 1,
+};
+
+static u8 xor_chksum_set(void *data)
+{
+ int idx;
+ u8 checksum = 0;
+ u8 *val = (u8 *)data;
+
+ for (idx = 0; idx < 7; idx++)
+ checksum ^= val[idx];
+
+ return checksum;
+}
+
+static void set_prod_idx(struct hinic_api_cmd_chain *chain)
+{
+ enum hinic_api_cmd_chain_type chain_type = chain->chain_type;
+ struct hinic_hwif *hwif = chain->hwdev->hwif;
+ u32 hw_prod_idx_addr = HINIC_CSR_API_CMD_CHAIN_PI_ADDR(chain_type);
+ u32 prod_idx = chain->prod_idx;
+
+ hinic_hwif_write_reg(hwif, hw_prod_idx_addr, prod_idx);
+}
+
+static u32 get_hw_cons_idx(struct hinic_api_cmd_chain *chain)
+{
+ u32 addr, val;
+
+ addr = HINIC_CSR_API_CMD_STATUS_0_ADDR(chain->chain_type);
+ val = hinic_hwif_read_reg(chain->hwdev->hwif, addr);
+
+ return HINIC_API_CMD_STATUS_GET(val, CONS_IDX);
+}
+
+static void dump_api_chain_reg(struct hinic_api_cmd_chain *chain)
+{
+ u32 addr, val;
+
+ addr = HINIC_CSR_API_CMD_STATUS_0_ADDR(chain->chain_type);
+ val = hinic_hwif_read_reg(chain->hwdev->hwif, addr);
+
+ PMD_DRV_LOG(ERR, "chain type: 0x%x", chain->chain_type);
+ PMD_DRV_LOG(ERR, "chain hw cpld error: 0x%x",
+ HINIC_API_CMD_STATUS_GET(val, CPLD_ERR));
+ PMD_DRV_LOG(ERR, "chain hw check error: 0x%x",
+ HINIC_API_CMD_STATUS_GET(val, CHKSUM_ERR));
+ PMD_DRV_LOG(ERR, "chain hw current fsm: 0x%x",
+ HINIC_API_CMD_STATUS_GET(val, FSM));
+ PMD_DRV_LOG(ERR, "chain hw current ci: 0x%x",
+ HINIC_API_CMD_STATUS_GET(val, CONS_IDX));
+
+ addr = HINIC_CSR_API_CMD_CHAIN_PI_ADDR(chain->chain_type);
+ val = hinic_hwif_read_reg(chain->hwdev->hwif, addr);
+ PMD_DRV_LOG(ERR, "Chain hw current pi: 0x%x", val);
+}
+
+/**
+ * chain_busy - check if the chain is still processing last requests
+ * @chain: chain to check
+ **/
+static int chain_busy(struct hinic_api_cmd_chain *chain)
+{
+ struct hinic_api_cmd_cell_ctxt *ctxt =
+ &chain->cell_ctxt[chain->prod_idx];
+
+ switch (chain->chain_type) {
+ case HINIC_API_CMD_WRITE_ASYNC_TO_MGMT_CPU:
+ case HINIC_API_CMD_PMD_WRITE_TO_MGMT:
+ chain->cons_idx = get_hw_cons_idx(chain);
+ if (chain->cons_idx == MASKED_IDX(chain, chain->prod_idx + 1)) {
+ PMD_DRV_LOG(ERR, "API CMD chain %d is busy, cons_idx:
%d, prod_idx: %d",
+
[dpdk-dev] [PATCH v2 03/11] net/hinic/base: add mgmt module interactive code
Add the structures, functionalities for interaction with
mgmt module.
Signed-off-by: Ziyang Xuan
---
drivers/net/hinic/base/hinic_pmd_hw_mgmt.h| 85 +++
drivers/net/hinic/base/hinic_pmd_mgmt.c | 619 ++
drivers/net/hinic/base/hinic_pmd_mgmt.h | 125
.../net/hinic/base/hinic_pmd_mgmt_interface.h | 503 ++
4 files changed, 1332 insertions(+)
create mode 100644 drivers/net/hinic/base/hinic_pmd_hw_mgmt.h
create mode 100644 drivers/net/hinic/base/hinic_pmd_mgmt.c
create mode 100644 drivers/net/hinic/base/hinic_pmd_mgmt.h
create mode 100644 drivers/net/hinic/base/hinic_pmd_mgmt_interface.h
diff --git a/drivers/net/hinic/base/hinic_pmd_hw_mgmt.h
b/drivers/net/hinic/base/hinic_pmd_hw_mgmt.h
new file mode 100644
index 0..5f3b12b7d
--- /dev/null
+++ b/drivers/net/hinic/base/hinic_pmd_hw_mgmt.h
@@ -0,0 +1,85 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2017 Huawei Technologies Co., Ltd
+ */
+
+#ifndef _HINIC_PMD_HW_MGMT_H_
+#define _HINIC_PMD_HW_MGMT_H_
+
+/* show each drivers only such as nic_service_cap,
+ * toe_service_cap structure, but not show service_cap
+ */
+enum hinic_service_type {
+ SERVICE_T_NIC = 0,
+ SERVICE_T_MAX = 7,
+
+ /* Only used for interruption resource management,
+* mark the request module
+*/
+ SERVICE_T_INTF = (1 << 15),
+ SERVICE_T_CQM= (1 << 16),
+};
+
+enum intr_type {
+ INTR_TYPE_MSIX,
+ INTR_TYPE_MSI,
+ INTR_TYPE_INT,
+ /* PXE,OVS need single thread processing, synchronization
+* messages must use poll wait mechanism interface
+*/
+ INTR_TYPE_NONE,
+};
+
+struct nic_service_cap {
+ /* PF resources */
+ u16 max_sqs;
+ u16 max_rqs;
+
+ /* VF resources, VF obtain them through the MailBox mechanism from
+* corresponding PF
+*/
+ u16 vf_max_sqs;
+ u16 vf_max_rqs;
+
+ bool lro_en;/* LRO feature enable bit */
+ u8 lro_sz; /* LRO context space: n*16B */
+ u8 tso_sz; /* TSO context space: n*16B */
+};
+
+/* Defines the IRQ information structure*/
+struct irq_info {
+ u16 msix_entry_idx; /* IRQ corresponding index number */
+ u32 irq_id; /* the IRQ number from OS */
+};
+
+/* Define the version information structure*/
+struct dev_version_info {
+ u8 up_ver; /* uP version, directly read from uP
+ * is not configured to file
+ */
+ u8 ucode_ver;/* The microcode version,
+ * read through the CMDq from microcode
+ */
+ u8 cfg_file_ver; /* uP configuration file version */
+ u8 sdk_ver; /* SDK driver version */
+ u8 hw_ver; /* Hardware version */
+};
+
+/* Obtain service_cap.nic_cap.dev_nic_cap.max_sqs */
+u16 hinic_func_max_qnum(void *hwdev);
+
+u16 hinic_global_func_id(void *hwdev); /* func_attr.glb_func_idx */
+
+enum func_type {
+ TYPE_PF,
+ TYPE_VF,
+ TYPE_PPF,
+};
+
+enum hinic_msix_state {
+ HINIC_MSIX_ENABLE,
+ HINIC_MSIX_DISABLE,
+};
+
+enum func_type hinic_func_type(void *hwdev);
+
+#endif /* _HINIC_PMD_HW_MGMT_H_ */
diff --git a/drivers/net/hinic/base/hinic_pmd_mgmt.c
b/drivers/net/hinic/base/hinic_pmd_mgmt.c
new file mode 100644
index 0..4825adb73
--- /dev/null
+++ b/drivers/net/hinic/base/hinic_pmd_mgmt.c
@@ -0,0 +1,619 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2017 Huawei Technologies Co., Ltd
+ */
+
+#include "hinic_pmd_dpdev.h"
+
+static void hinic_mgmt_recv_msg_handler(struct hinic_msg_pf_to_mgmt
*pf_to_mgmt,
+ struct hinic_recv_msg *recv_msg,
+ void *param);
+
+#define BUF_OUT_DEFAULT_SIZE 1
+
+#define MAX_PF_MGMT_BUF_SIZE 2048UL
+
+#define MGMT_MSG_SIZE_MIN 20
+#define MGMT_MSG_SIZE_STEP 16
+#defineMGMT_MSG_RSVD_FOR_DEV 8
+
+#define MGMT_MSG_TIMEOUT 5000/* millisecond */
+
+#define SYNC_MSG_ID_MASK 0x1FF
+#define ASYNC_MSG_ID_MASK 0x1FF
+#define ASYNC_MSG_FLAG 0x200
+
+#define MSG_NO_RESP0x
+
+#define MAX_MSG_SZ 2016
+
+#define MSG_SZ_IS_VALID(in_size) ((in_size) <= MAX_MSG_SZ)
+
+#define SYNC_MSG_ID(pf_to_mgmt)((pf_to_mgmt)->sync_msg_id)
+
+#define SYNC_MSG_ID_INC(pf_to_mgmt)(SYNC_MSG_ID(pf_to_mgmt) = \
+ (SYNC_MSG_ID(pf_to_mgmt) + 1) & SYNC_MSG_ID_MASK)
+
+#define ASYNC_MSG_ID(pf_to_mgmt) ((pf_to_mgmt)->async_msg_id)
+
+#define ASYNC_MSG_ID_INC(pf_to_mgmt) (ASYNC_MSG_ID(pf_to_mgmt) = \
+ ((ASYNC_MSG_ID(pf_to_mgmt) + 1) & ASYNC_MSG_ID_MASK) \
+ | ASYNC_MSG_FLAG)
+
+#define HINIC_SEQ_ID_MAX_VAL 42
+#define HINIC_MSG_SEG_LEN 48
+
+/**
+ * m
[dpdk-dev] [PATCH v2 04/11] net/hinic/base: add code about hardware operation
Add code for hardware operation, including configuration,
query and so on.
Signed-off-by: Ziyang Xuan
---
drivers/net/hinic/base/hinic_pmd_cfg.c | 269
drivers/net/hinic/base/hinic_pmd_cfg.h | 264
drivers/net/hinic/base/hinic_pmd_hw.h| 49 +
drivers/net/hinic/base/hinic_pmd_hwdev.c | 1544 ++
drivers/net/hinic/base/hinic_pmd_hwdev.h | 207 +++
drivers/net/hinic/base/hinic_pmd_hwif.c | 542
drivers/net/hinic/base/hinic_pmd_hwif.h | 90 ++
7 files changed, 2965 insertions(+)
create mode 100644 drivers/net/hinic/base/hinic_pmd_cfg.c
create mode 100644 drivers/net/hinic/base/hinic_pmd_cfg.h
create mode 100644 drivers/net/hinic/base/hinic_pmd_hw.h
create mode 100644 drivers/net/hinic/base/hinic_pmd_hwdev.c
create mode 100644 drivers/net/hinic/base/hinic_pmd_hwdev.h
create mode 100644 drivers/net/hinic/base/hinic_pmd_hwif.c
create mode 100644 drivers/net/hinic/base/hinic_pmd_hwif.h
diff --git a/drivers/net/hinic/base/hinic_pmd_cfg.c
b/drivers/net/hinic/base/hinic_pmd_cfg.c
new file mode 100644
index 0..3af05198e
--- /dev/null
+++ b/drivers/net/hinic/base/hinic_pmd_cfg.c
@@ -0,0 +1,269 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2017 Huawei Technologies Co., Ltd
+ */
+
+#include "hinic_pmd_dpdev.h"
+
+static void parse_dev_cap(struct hinic_hwdev *dev,
+ struct hinic_dev_cap *dev_cap,
+ enum func_type type);
+
+bool hinic_support_nic(struct hinic_hwdev *hwdev, struct nic_service_cap *cap)
+{
+ if (!hwdev)
+ return false;
+
+ if (!IS_NIC_TYPE(hwdev))
+ return false;
+
+ if (cap)
+ memcpy(cap, &hwdev->cfg_mgmt->svc_cap.nic_cap, sizeof(*cap));
+
+ return true;
+}
+
+static void hinic_parse_shared_res_cap(struct service_cap *cap,
+ struct hinic_dev_cap *dev_cap,
+ __rte_unused enum func_type type)
+{
+ struct host_shared_resource_cap *shared_cap = &cap->shared_res_cap;
+
+ shared_cap->host_pctxs = dev_cap->host_pctx_num;
+
+ if (dev_cap->host_sf_en)
+ cap->sf_en = true;
+ else
+ cap->sf_en = false;
+
+ shared_cap->host_cctxs = dev_cap->host_ccxt_num;
+ shared_cap->host_scqs = dev_cap->host_scq_num;
+ shared_cap->host_srqs = dev_cap->host_srq_num;
+ shared_cap->host_mpts = dev_cap->host_mpt_num;
+
+ PMD_DRV_LOG(INFO, "Get share resource capability:");
+ PMD_DRV_LOG(INFO, "host_pctxs: 0x%x, host_cctxs: 0x%x, host_scqs: 0x%x,
host_srqs: 0x%x, host_mpts: 0x%x",
+ shared_cap->host_pctxs, shared_cap->host_cctxs,
+ shared_cap->host_scqs, shared_cap->host_srqs,
+ shared_cap->host_mpts);
+}
+
+static void hinic_parse_l2nic_res_cap(struct service_cap *cap,
+ struct hinic_dev_cap *dev_cap,
+ enum func_type type)
+{
+ struct nic_service_cap *nic_cap = &cap->nic_cap;
+
+ if (type == TYPE_PF || type == TYPE_PPF) {
+ nic_cap->max_sqs = dev_cap->nic_max_sq + 1;
+ nic_cap->max_rqs = dev_cap->nic_max_rq + 1;
+ nic_cap->vf_max_sqs = dev_cap->nic_vf_max_sq + 1;
+ nic_cap->vf_max_rqs = dev_cap->nic_vf_max_rq + 1;
+ } else {
+ nic_cap->max_sqs = dev_cap->nic_max_sq;
+ nic_cap->max_rqs = dev_cap->nic_max_rq;
+ nic_cap->vf_max_sqs = 0;
+ nic_cap->vf_max_rqs = 0;
+ }
+
+ if (dev_cap->nic_lro_en)
+ nic_cap->lro_en = true;
+ else
+ nic_cap->lro_en = false;
+
+ nic_cap->lro_sz = dev_cap->nic_lro_sz;
+ nic_cap->tso_sz = dev_cap->nic_tso_sz;
+
+ PMD_DRV_LOG(INFO, "Get l2nic resource capability:");
+ PMD_DRV_LOG(INFO, "max_sqs: 0x%x, max_rqs: 0x%x, vf_max_sqs: 0x%x,
vf_max_rqs: 0x%x",
+ nic_cap->max_sqs, nic_cap->max_rqs,
+ nic_cap->vf_max_sqs, nic_cap->vf_max_rqs);
+}
+
+static int get_cap_from_fw(struct hinic_hwdev *dev, enum func_type type)
+{
+ int err;
+ u16 in_len, out_len;
+ struct hinic_dev_cap dev_cap;
+
+ memset(&dev_cap, 0, sizeof(dev_cap));
+ in_len = sizeof(dev_cap);
+ out_len = in_len;
+ dev_cap.mgmt_msg_head.resp_aeq_num = HINIC_AEQ1;
+ err = hinic_msg_to_mgmt_sync(dev, HINIC_MOD_CFGM, HINIC_CFG_NIC_CAP,
+&dev_cap, in_len, &dev_cap, &out_len, 0);
+ if (err || dev_cap.mgmt_msg_head.status || !out_len) {
+ PMD_DRV_LOG(ERR, "Get capability from FW failed, err: %d,
status: %d, out_len: %d",
+ err, dev_cap.mgmt_msg_head.status, out_len);
+ return -EFAULT;
+ }
+
+ parse_dev_cap(dev, &dev_cap, type);
+ return 0;
+}
+
+static int get_dev_cap(struct hinic_hwdev *dev)
+{
+
[dpdk-dev] [PATCH v2 05/11] net/hinic/base: add eqs and context code
Add code for eq. Aeq is a kind queue for mgmt asynchronous message
and mgmt command response message.
Signed-off-by: Ziyang Xuan
---
drivers/net/hinic/base/hinic_ctx_def.h | 195 ++
drivers/net/hinic/base/hinic_pmd_dpdev.h | 149 +
drivers/net/hinic/base/hinic_pmd_eqs.c | 725 +++
drivers/net/hinic/base/hinic_pmd_eqs.h | 94 +++
4 files changed, 1163 insertions(+)
create mode 100644 drivers/net/hinic/base/hinic_ctx_def.h
create mode 100644 drivers/net/hinic/base/hinic_pmd_dpdev.h
create mode 100644 drivers/net/hinic/base/hinic_pmd_eqs.c
create mode 100644 drivers/net/hinic/base/hinic_pmd_eqs.h
diff --git a/drivers/net/hinic/base/hinic_ctx_def.h
b/drivers/net/hinic/base/hinic_ctx_def.h
new file mode 100644
index 0..510adf7e3
--- /dev/null
+++ b/drivers/net/hinic/base/hinic_ctx_def.h
@@ -0,0 +1,195 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2017 Huawei Technologies Co., Ltd
+ */
+
+#ifndef _HINIC_CTX_DEF_H_
+#define _HINIC_CTX_DEF_H_
+
+#ifdef __cplusplus
+#if __cplusplus
+extern "C"{
+#endif
+#endif /* __cplusplus */
+
+#define MASKED_SQ_IDX(sq, idx) ((idx) & (sq)->wq->mask)
+
+#define HINIC_Q_CTXT_MAX 42
+
+/* performance: ci addr RTE_CACHE_SIZE(64B) alignment */
+#define HINIC_CI_Q_ADDR_SIZE (64)
+
+#define CI_TABLE_SIZE(num_qps, pg_sz) \
+ (ALIGN((num_qps) * HINIC_CI_Q_ADDR_SIZE, pg_sz))
+
+#define HINIC_CI_VADDR(base_addr, q_id)((u8 *)(base_addr) + \
+ (q_id) * HINIC_CI_Q_ADDR_SIZE)
+
+#define HINIC_CI_PADDR(base_paddr, q_id) ((base_paddr) + \
+ (q_id) * HINIC_CI_Q_ADDR_SIZE)
+
+#define Q_CTXT_SIZE48
+#define TSO_LRO_CTXT_SIZE 240
+
+#define SQ_CTXT_OFFSET(max_sqs, max_rqs, q_id) \
+ (((max_rqs) + (max_sqs)) * TSO_LRO_CTXT_SIZE \
+ + (q_id) * Q_CTXT_SIZE)
+
+#define RQ_CTXT_OFFSET(max_sqs, max_rqs, q_id) \
+ (((max_rqs) + (max_sqs)) * TSO_LRO_CTXT_SIZE \
+ + (max_sqs) * Q_CTXT_SIZE + (q_id) * Q_CTXT_SIZE)
+
+#define SQ_CTXT_SIZE(num_sqs) ((u16)(sizeof(struct hinic_qp_ctxt_header) \
+ + (num_sqs) * sizeof(struct hinic_sq_ctxt)))
+
+#define RQ_CTXT_SIZE(num_rqs) ((u16)(sizeof(struct hinic_qp_ctxt_header) \
+ + (num_rqs) * sizeof(struct hinic_rq_ctxt)))
+
+#define SQ_CTXT_CEQ_ATTR_CEQ_ID_SHIFT 8
+#define SQ_CTXT_CEQ_ATTR_GLOBAL_SQ_ID_SHIFT13
+#define SQ_CTXT_CEQ_ATTR_EN_SHIFT 23
+#define SQ_CTXT_CEQ_ATTR_ARM_SHIFT 31
+
+#define SQ_CTXT_CEQ_ATTR_CEQ_ID_MASK 0x1FU
+#define SQ_CTXT_CEQ_ATTR_GLOBAL_SQ_ID_MASK 0x3FFU
+#define SQ_CTXT_CEQ_ATTR_EN_MASK 0x1U
+#define SQ_CTXT_CEQ_ATTR_ARM_MASK 0x1U
+
+#define SQ_CTXT_CEQ_ATTR_SET(val, member) (((val) & \
+ SQ_CTXT_CEQ_ATTR_##member##_MASK) \
+ << SQ_CTXT_CEQ_ATTR_##member##_SHIFT)
+
+#define SQ_CTXT_CI_IDX_SHIFT 11
+#define SQ_CTXT_CI_OWNER_SHIFT 23
+
+#define SQ_CTXT_CI_IDX_MASK0xFFFU
+#define SQ_CTXT_CI_OWNER_MASK 0x1U
+
+#define SQ_CTXT_CI_SET(val, member)(((val) & \
+ SQ_CTXT_CI_##member##_MASK) \
+ << SQ_CTXT_CI_##member##_SHIFT)
+
+#define SQ_CTXT_WQ_PAGE_HI_PFN_SHIFT 0
+#define SQ_CTXT_WQ_PAGE_PI_SHIFT 20
+
+#define SQ_CTXT_WQ_PAGE_HI_PFN_MASK0xFU
+#define SQ_CTXT_WQ_PAGE_PI_MASK0xFFFU
+
+#define SQ_CTXT_WQ_PAGE_SET(val, member) (((val) & \
+ SQ_CTXT_WQ_PAGE_##member##_MASK) \
+ << SQ_CTXT_WQ_PAGE_##member##_SHIFT)
+
+#define SQ_CTXT_PREF_CACHE_THRESHOLD_SHIFT 0
+#define SQ_CTXT_PREF_CACHE_MAX_SHIFT 14
+#define SQ_CTXT_PREF_CACHE_MIN_SHIFT 25
+
+#define SQ_CTXT_PREF_CACHE_THRESHOLD_MASK 0x3FFFU
+#define SQ_CTXT_PREF_CACHE_MAX_MASK0x7FFU
+#define SQ_CTXT_PREF_CACHE_MIN_MASK0x7FU
+
+#define SQ_CTXT_PREF_WQ_PFN_HI_SHIFT 0
+#define SQ_CTXT_PREF_CI_SHIFT 20
+
+#define SQ_CTXT_PREF_WQ_PFN_HI_MASK0xFU
+#define SQ_CTXT_PREF_CI_MASK 0xFFFU
+
+#define SQ_CTXT_PREF_SET(val, member) (((val) & \
+ SQ_CTXT_PREF_##member##_MASK) \
+
[dpdk-dev] [PATCH v2 06/11] net/hinic/base: add code for nic business
Add code for nic business, including qps structures, qps configuration,
wqs configuration for qps, nic business configuration functionalities.
Signed-off-by: Ziyang Xuan
---
drivers/net/hinic/base/hinic_pmd_nic.h| 85 ++
drivers/net/hinic/base/hinic_pmd_niccfg.c | 1408 +
drivers/net/hinic/base/hinic_pmd_niccfg.h | 333 +
drivers/net/hinic/base/hinic_pmd_nicio.c | 920 ++
drivers/net/hinic/base/hinic_pmd_nicio.h | 53 +
drivers/net/hinic/base/hinic_pmd_qp.c | 26 +
drivers/net/hinic/base/hinic_pmd_qp.h | 76 ++
drivers/net/hinic/base/hinic_pmd_wq.c | 164 +++
drivers/net/hinic/base/hinic_pmd_wq.h | 52 +
9 files changed, 3117 insertions(+)
create mode 100644 drivers/net/hinic/base/hinic_pmd_nic.h
create mode 100644 drivers/net/hinic/base/hinic_pmd_niccfg.c
create mode 100644 drivers/net/hinic/base/hinic_pmd_niccfg.h
create mode 100644 drivers/net/hinic/base/hinic_pmd_nicio.c
create mode 100644 drivers/net/hinic/base/hinic_pmd_nicio.h
create mode 100644 drivers/net/hinic/base/hinic_pmd_qp.c
create mode 100644 drivers/net/hinic/base/hinic_pmd_qp.h
create mode 100644 drivers/net/hinic/base/hinic_pmd_wq.c
create mode 100644 drivers/net/hinic/base/hinic_pmd_wq.h
diff --git a/drivers/net/hinic/base/hinic_pmd_nic.h
b/drivers/net/hinic/base/hinic_pmd_nic.h
new file mode 100644
index 0..7bea29424
--- /dev/null
+++ b/drivers/net/hinic/base/hinic_pmd_nic.h
@@ -0,0 +1,85 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2017 Huawei Technologies Co., Ltd
+ */
+
+#ifndef _HINIC_PMD_NIC_H_
+#define _HINIC_PMD_NIC_H_
+
+#define HINIC_FLUSH_QUEUE_TIMEOUT 3000
+
+struct hinic_hwdev;
+struct hinic_wq;
+
+struct hinic_sq {
+ struct hinic_wq *wq;
+ volatile u16*cons_idx_addr;
+ void __iomem*db_addr;
+
+ u16 q_id;
+ u16 owner;
+ u16 sq_depth;
+};
+
+struct hinic_rq {
+ struct hinic_wq *wq;
+ volatile u16*pi_virt_addr;
+ dma_addr_t pi_dma_addr;
+
+ u16 irq_id;
+ u16 msix_entry_idx;
+ u16 q_id;
+ u16 rq_depth;
+};
+
+struct hinic_qp {
+ struct hinic_sq sq;
+ struct hinic_rq rq;
+};
+
+struct vf_data_storage {
+ u8 vf_mac_addr[ETH_ALEN];
+ bool registered;
+ bool pf_set_mac;
+ u16 pf_vlan;
+ u8 pf_qos;
+
+ bool link_forced;
+ bool link_up; /* only valid if VF link is forced */
+};
+
+struct hinic_nic_io {
+ struct hinic_hwdev *hwdev;
+
+ u16 global_qpn;
+ u8 link_status;
+
+ struct hinic_wq *sq_wq;
+ struct hinic_wq *rq_wq;
+
+ u16 max_qps;
+ u16 num_qps;
+
+ u16 num_sqs;
+ u16 num_rqs;
+
+ u16 sq_depth;
+ u16 rq_depth;
+
+ u16 rq_buf_size;
+ u16 vhd_mode;
+
+ struct hinic_qp *qps;
+ /* sq ci mem base addr of the function*/
+ void*ci_vaddr_base;
+ dma_addr_t ci_dma_base;
+
+ struct hinic_event event;
+ void*event_handle;
+
+ u16 max_vfs;
+ u16 num_vfs;
+ u8 vf_link_mode;
+ struct vf_data_storage *vf_infos;
+};
+
+#endif /* _HINIC_PMD_NIC_H_ */
diff --git a/drivers/net/hinic/base/hinic_pmd_niccfg.c
b/drivers/net/hinic/base/hinic_pmd_niccfg.c
new file mode 100644
index 0..6da2172b4
--- /dev/null
+++ b/drivers/net/hinic/base/hinic_pmd_niccfg.c
@@ -0,0 +1,1408 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2017 Huawei Technologies Co., Ltd
+ */
+
+#include "hinic_pmd_dpdev.h"
+
+#define l2nic_msg_to_mgmt_sync(hwdev, cmd, buf_in, \
+ in_size, buf_out, out_size) \
+ hinic_msg_to_mgmt_sync(hwdev, HINIC_MOD_L2NIC, cmd, \
+ buf_in, in_size,\
+ buf_out, out_size, 0)
+
+int hinic_init_function_table(void *hwdev, u16 rx_buf_sz)
+{
+ struct hinic_function_table function_table;
+ u16 out_size = sizeof(function_table);
+ int err;
+
+ if (!hwdev) {
+ PMD_DRV_LOG(ERR, "Hwdev is NULL");
+ return -EINVAL;
+ }
+
+ memset(&function_table, 0, sizeof(function_table));
+ function_table.mgmt_msg_head.resp_aeq_num = HINIC_AEQ1;
+ function_table.func_id = hinic_global_func_id(hwdev);
+ function_table.mtu = 0x3FFF;/* default, max mtu */
+ function_table.rx_wqe_buf_size = rx_buf_sz;
+
+ err = hinic_msg_to_mgmt_sync(hwdev, HINIC
[dpdk-dev] [PATCH v2 07/11] net/hinic/base: add various headers
Add various headers that define mgmt commands, cmdq commands,
rx data structures, tx data structures and basic defines for
use in the code.
Signed-off-by: Ziyang Xuan
---
drivers/net/hinic/base/hinic_compat.h | 252
drivers/net/hinic/base/hinic_port_cmd.h | 494
drivers/net/hinic/base/hinic_qe_def.h | 461 ++
drivers/net/hinic/hinic_pmd_ethdev.h| 106 +
drivers/net/hinic/hinic_pmd_rx.h| 135 +++
drivers/net/hinic/hinic_pmd_tx.h| 97 +
6 files changed, 1545 insertions(+)
create mode 100644 drivers/net/hinic/base/hinic_compat.h
create mode 100644 drivers/net/hinic/base/hinic_port_cmd.h
create mode 100644 drivers/net/hinic/base/hinic_qe_def.h
create mode 100644 drivers/net/hinic/hinic_pmd_ethdev.h
create mode 100644 drivers/net/hinic/hinic_pmd_rx.h
create mode 100644 drivers/net/hinic/hinic_pmd_tx.h
diff --git a/drivers/net/hinic/base/hinic_compat.h
b/drivers/net/hinic/base/hinic_compat.h
new file mode 100644
index 0..12bf0a07b
--- /dev/null
+++ b/drivers/net/hinic/base/hinic_compat.h
@@ -0,0 +1,252 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2017 Huawei Technologies Co., Ltd
+ */
+
+#ifndef _HINIC_COMPAT_H_
+#define _HINIC_COMPAT_H_
+
+#ifdef __cplusplus
+#if __cplusplus
+extern "C"{
+#endif
+#endif /* __cplusplus */
+
+#include
+#include
+
+#include
+#include
+#include
+#include
+#include
+#include
+#include
+#include
+#include
+#include
+
+typedef uint8_t u8;
+typedef int8_ts8;
+typedef uint16_t u16;
+typedef uint32_t u32;
+typedef int32_t s32;
+typedef uint64_t u64;
+
+#ifndef dma_addr_t
+typedef uint64_t dma_addr_t;
+#endif
+
+#ifndef gfp_t
+#define gfp_t unsigned
+#endif
+
+#ifndef bool
+#define bool int
+#endif
+
+#ifndef FALSE
+#define FALSE (0)
+#endif
+
+#ifndef TRUE
+#define TRUE (1)
+#endif
+
+#ifndef false
+#define false (0)
+#endif
+
+#ifndef true
+#define true (1)
+#endif
+
+#ifndef NULL
+#define NULL ((void *)0)
+#endif
+
+#define HINIC_ERROR(-1)
+#define HINIC_OK (0)
+
+#ifndef BIT
+#define BIT(n) (1 << (n))
+#endif
+
+#define upper_32_bits(n) ((u32)(((n) >> 16) >> 16))
+#define lower_32_bits(n) ((u32)(n))
+
+/* Returns X / Y, rounding up. X must be nonnegative to round correctly. */
+#define DIV_ROUND_UP(X, Y) (((X) + ((Y) - 1)) / (Y))
+
+/* Returns X rounded up to the nearest multiple of Y. */
+#define ROUND_UP(X, Y) (DIV_ROUND_UP(X, Y) * (Y))
+
+#undef ALIGN
+#define ALIGN(x, a) RTE_ALIGN(x, a)
+
+#define PTR_ALIGN(p, a)((typeof(p))ALIGN((unsigned long)(p),
(a)))
+
+/* Reported driver name. */
+#define HINIC_DRIVER_NAME "net_hinic"
+
+extern int hinic_logtype;
+
+#define PMD_DRV_LOG(level, fmt, args...) \
+ rte_log(RTE_LOG_ ## level, hinic_logtype, \
+ HINIC_DRIVER_NAME": " fmt "\n", ##args)
+
+#define HINIC_ASSERT_EN
+
+#ifdef HINIC_ASSERT_EN
+#define HINIC_ASSERT(exp) \
+ do {\
+ if (!(exp)) { \
+ rte_panic("line%d\tassert \"" #exp "\" failed\n", \
+ __LINE__);\
+ } \
+ } while (0)
+#else
+#define HINIC_ASSERT(exp) do {} while (0)
+#endif
+
+#define HINIC_BUG_ON(x) HINIC_ASSERT(!(x))
+
+/* common definition */
+#ifndef ETH_ALEN
+#define ETH_ALEN 6
+#endif
+#define ETH_HLEN 14
+#define ETH_CRC_LEN4
+#define VLAN_PRIO_SHIFT13
+#define VLAN_N_VID 4096
+
+/* bit order interface */
+#define cpu_to_be16(o) rte_cpu_to_be_16(o)
+#define cpu_to_be32(o) rte_cpu_to_be_32(o)
+#define cpu_to_be64(o) rte_cpu_to_be_64(o)
+#define cpu_to_le32(o) rte_cpu_to_le_32(o)
+#define be16_to_cpu(o) rte_be_to_cpu_16(o)
+#define be32_to_cpu(o) rte_be_to_cpu_32(o)
+#define be64_to_cpu(o) rte_be_to_cpu_64(o)
+#define le32_to_cpu(o) rte_le_to_cpu_32(o)
+
+/* virt memory and dma phy memory */
+#define __iomem
+#define __force
+#define GFP_KERNEL RTE_MEMZONE_IOVA_CONTIG
+#define PAGE_SHIFT 12
+#define PAGE_SIZE RTE_PGSIZE_4K
+#define HINIC_MEM_ALLOC_ALIGNE_MIN 8
+
+static inline int hinic_test_bit(int nr, volatile unsigned long *addr)
+{
+ int res;
+
+ rte_mb();
+ res = ((*addr) & (1UL << nr)) != 0;
+ rte_mb();
+ return res;
+}
+
+static inline void hinic_set_bit(unsigned int nr, volatile unsigned long *addr)
+{
+ __sync_fetch_and_or(addr, (1UL << nr));
+}
+
+static inline void hinic_clear_bit(int nr, volatile unsigned long *addr)
+{
+ __sync_fetch_and_and(addr, ~(1UL << nr));
+}
+
+static inline int hinic_test_and_clear_bit(int nr, volatile unsigned long
*addr)
+{
+ unsigned long mask = (1UL << nr);
+
+ return __sync_fetch_and_and(addr, ~mask) & mask;
+}
+
+static inline int hinic_test_and_set_bit(int nr, volatile unsigned long *addr)
+{
+ unsigned long mask = (1UL << nr);
+
+ return __sync_fetch_and_or
[dpdk-dev] [PATCH v2 08/11] net/hinic: add hinic PMD build and doc files
Add build and doc files, and common code files. Signed-off-by: Ziyang Xuan --- MAINTAINERS | 9 + config/common_base | 5 + doc/guides/nics/features/hinic.ini | 32 ++ doc/guides/nics/hinic.rst | 49 ++ drivers/net/Makefile| 1 + drivers/net/hinic/Makefile | 94 drivers/net/hinic/base/hinic_logs.c | 16 + drivers/net/hinic/base/hinic_logs.h | 31 ++ drivers/net/hinic/base/meson.build | 63 +++ drivers/net/hinic/hinic_pmd_dpdev.c | 718 drivers/net/hinic/meson.build | 18 + mk/rte.app.mk | 1 + 12 files changed, 1037 insertions(+) create mode 100644 doc/guides/nics/features/hinic.ini create mode 100644 doc/guides/nics/hinic.rst create mode 100644 drivers/net/hinic/Makefile create mode 100644 drivers/net/hinic/base/hinic_logs.c create mode 100644 drivers/net/hinic/base/hinic_logs.h create mode 100644 drivers/net/hinic/base/meson.build create mode 100644 drivers/net/hinic/hinic_pmd_dpdev.c create mode 100644 drivers/net/hinic/meson.build diff --git a/MAINTAINERS b/MAINTAINERS index 15d0829c5..b76340c28 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -577,6 +577,15 @@ F: drivers/net/enic/ F: doc/guides/nics/enic.rst F: doc/guides/nics/features/enic.ini +Huawei hinic +M: Xiaoyun Wang +M: Ziyang Xuan +M: Guoyang Zhou +M: Rami Rosen +F: drivers/net/hinic/ +F: doc/guides/nics/hinic.rst +F: doc/guides/nics/features/hinic.ini + Intel e1000 M: Wenzhuo Lu T: git://dpdk.org/next/dpdk-next-net-intel diff --git a/config/common_base b/config/common_base index 6b96e0e80..8d8ba10d5 100644 --- a/config/common_base +++ b/config/common_base @@ -276,6 +276,11 @@ CONFIG_RTE_LIBRTE_E1000_DEBUG_TX=n CONFIG_RTE_LIBRTE_E1000_DEBUG_TX_FREE=n CONFIG_RTE_LIBRTE_E1000_PF_DISABLE_STRIP_CRC=n +# +# Compile burst-oriented HINIC PMD driver +# +CONFIG_RTE_LIBRTE_HINIC_PMD=y + # # Compile burst-oriented IXGBE PMD driver # diff --git a/doc/guides/nics/features/hinic.ini b/doc/guides/nics/features/hinic.ini new file mode 100644 index 0..d3b9e35be --- /dev/null +++ b/doc/guides/nics/features/hinic.ini @@ -0,0 +1,32 @@ +; +; Supported features of the 'hinic' network poll mode driver. +; +; Refer to default.ini for the full list of available PMD features. +; +[Features] +Speed capabilities = Y +Link status = Y +Link status event= Y +Free Tx mbuf on demand = Y +Queue start/stop = Y +Jumbo frame = N +Scattered Rx = Y +TSO = Y +Promiscuous mode = Y +Unicast MAC filter = Y +Multicast MAC filter = Y +RSS hash = Y +RSS key update = Y +RSS reta update = Y +Inner RSS= Y +CRC offload = Y +L3 checksum offload = Y +L4 checksum offload = Y +Inner L3 checksum= Y +Inner L4 checksum= Y +Basic stats = Y +Extended stats = Y +Stats per queue = Y +BSD nic_uio = Y +Linux UIO= Y +x86-64 = Y diff --git a/doc/guides/nics/hinic.rst b/doc/guides/nics/hinic.rst new file mode 100644 index 0..43e002ba1 --- /dev/null +++ b/doc/guides/nics/hinic.rst @@ -0,0 +1,49 @@ +.. SPDX-License-Identifier: BSD-3-Clause +Copyright(c) 2017 Huawei Technologies Co., Ltd + + +Hinic Poll Mode Driver +== + +The hinic PMD (librte_pmd_hinic) provides poll mode driver support for +Huawei Intelligent PCIE Network Interface Card. + +Prerequisites +- + +- Follow the DPDK :ref:`Getting Started Guide for Linux ` to setup the basic DPDK environment. + +Requires firmware 1.6.2.5 + + +Pre-Installation Configuration +-- + +Config File Options +~~~ + +The following options can be modified in the ``config`` file. + +- ``CONFIG_RTE_LIBRTE_HINIC_PMD`` (default ``y``) + + + +Runtime Config Options +~~ + +None + +Driver compilation and testing +-- + +Refer to the document :ref:`compiling and testing a PMD for a NIC ` +for details. + +Sample Application Notes + + + +Limitations or Known issues +--- +Jumbo frames is not supported yet. +Build with Clang/ICC is not supported yet. diff --git a/drivers/net/Makefile b/drivers/net/Makefile index 3a72cf38c..606b27456 100644 --- a/drivers/net/Makefile +++ b/drivers/net/Makefile @@ -29,6 +29,7 @@ DIRS-$(CONFIG_RTE_LIBRTE_ENETC_PMD) += enetc DIRS-$(CONFIG_RTE_LIBRTE_ENIC_PMD) += enic DIRS-$(CONFIG_RTE_LIBRTE_PMD_FAILSAFE) += failsafe DIRS-$(CONFIG_RTE_LIBRTE_FM10K_PMD) += fm10k +DIRS-$(CONFIG_RTE_LIBRTE_HINIC_PMD) += hinic DIRS-$(CONFIG_RTE_LIBRTE_I40E_PMD) += i40e DIRS-$(CONFIG_RTE_LIBRTE_IAVF_PMD) += iavf DIRS-$(CONFIG_RTE_LIBRTE_ICE_PMD) += ice diff --git a/drivers/net/hinic/Makefile b/drivers/net/hinic/Makefile new file mode 100644 index 0..6e19acf08 --- /dev/null +++ b/drivers/net/hinic/Makefile @@ -0,0 +1,94 @@ +# SPD
[dpdk-dev] [PATCH v2 11/11] net/hinic: add support for basic device operations
Add hinic PMD initialization and ethernet operatioins code.
Signed-off-by: Ziyang Xuan
---
drivers/net/hinic/Makefile |2 +-
drivers/net/hinic/hinic_pmd_ethdev.c| 2157 +++
drivers/net/hinic/rte_pmd_hinic_version.map |4 +
3 files changed, 2162 insertions(+), 1 deletion(-)
create mode 100644 drivers/net/hinic/hinic_pmd_ethdev.c
create mode 100644 drivers/net/hinic/rte_pmd_hinic_version.map
diff --git a/drivers/net/hinic/Makefile b/drivers/net/hinic/Makefile
index f488fa430..2c90ab679 100644
--- a/drivers/net/hinic/Makefile
+++ b/drivers/net/hinic/Makefile
@@ -76,7 +76,7 @@ SRCS-$(CONFIG_RTE_LIBRTE_HINIC_PMD) += hinic_pmd_qp.c
SRCS-$(CONFIG_RTE_LIBRTE_HINIC_PMD) += hinic_pmd_wq.c
SRCS-$(CONFIG_RTE_LIBRTE_HINIC_PMD) += hinic_pmd_dpdev.c
-#SRCS-$(CONFIG_RTE_LIBRTE_HINIC_PMD) += hinic_pmd_ethdev.c
+SRCS-$(CONFIG_RTE_LIBRTE_HINIC_PMD) += hinic_pmd_ethdev.c
SRCS-$(CONFIG_RTE_LIBRTE_HINIC_PMD) += hinic_pmd_rx.c
SRCS-$(CONFIG_RTE_LIBRTE_HINIC_PMD) += hinic_pmd_tx.c
# this lib depends upon:
diff --git a/drivers/net/hinic/hinic_pmd_ethdev.c
b/drivers/net/hinic/hinic_pmd_ethdev.c
new file mode 100644
index 0..3ce9d257b
--- /dev/null
+++ b/drivers/net/hinic/hinic_pmd_ethdev.c
@@ -0,0 +1,2157 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2017 Huawei Technologies Co., Ltd
+ */
+
+#include
+#include
+#include
+#include
+#include
+#include
+#include
+#include
+#include
+#include
+#include
+
+#include "hinic_pmd_ethdev.h"
+#include "hinic_pmd_tx.h"
+#include "hinic_pmd_rx.h"
+
+#define HINIC_MIN_RX_BUF_SIZE 1024
+
+#define HINIC_MAX_MAC_ADDRS1
+#define EQ_MSIX_RESEND_TIMER_CLEAR 1
+
+/* Hinic PMD parameters */
+#define ETH_HINIC_FW_VER "check_fw_version"
+
+static const char *const valid_params[] = {
+ ETH_HINIC_FW_VER,
+ NULL};
+
+
+/** Driver-specific log messages type. */
+int hinic_logtype;
+
+static int check_fw_ver = 1;
+
+static int hinic_dev_init(struct rte_eth_dev *eth_dev);
+static int hinic_dev_uninit(struct rte_eth_dev *dev);
+static int hinic_init_mac_addr(struct rte_eth_dev *eth_dev);
+static void hinic_deinit_mac_addr(struct rte_eth_dev *eth_dev);
+static int hinic_rx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx,
+uint16_t nb_desc, unsigned int socket_id,
+__rte_unused const struct rte_eth_rxconf *rx_conf,
+struct rte_mempool *mp);
+static int hinic_tx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx,
+uint16_t nb_desc, unsigned int socket_id,
+__rte_unused const struct rte_eth_txconf *tx_conf);
+
+static const struct eth_dev_ops hinic_pmd_ops = {
+ .dev_configure = hinic_dev_configure,
+ .dev_infos_get = hinic_dev_infos_get,
+ .rx_queue_setup= hinic_rx_queue_setup,
+ .tx_queue_setup= hinic_tx_queue_setup,
+ .dev_start = hinic_dev_start,
+ .link_update = hinic_link_update,
+ .rx_queue_release = hinic_rx_queue_release,
+ .tx_queue_release = hinic_tx_queue_release,
+ .dev_stop = hinic_dev_stop,
+ .dev_close = hinic_dev_close,
+ .promiscuous_enable= hinic_dev_promiscuous_enable,
+ .promiscuous_disable = hinic_dev_promiscuous_disable,
+ .rss_hash_update = hinic_rss_hash_update,
+ .rss_hash_conf_get = hinic_rss_conf_get,
+ .reta_update = hinic_rss_indirtbl_update,
+ .reta_query= hinic_rss_indirtbl_query,
+ .stats_get = hinic_dev_stats_get,
+ .stats_reset = hinic_dev_stats_reset,
+ .xstats_get= hinic_dev_xstats_get,
+ .xstats_reset = hinic_dev_xstats_reset,
+ .xstats_get_names = hinic_dev_xstats_get_names,
+ .fw_version_get= hinic_fw_version_get,
+};
+
+static int hinic_check_fw_ver_param(__rte_unused const char *key,
+ const char *value,
+ __rte_unused void *opaque)
+{
+ int num = -1;
+ char *end = NULL;
+
+ while (isblank(*value))
+ value++;
+
+ num = strtoul(value, &end, 10);
+ if ((*end == '-') || errno)
+ return -1;
+ check_fw_ver = num;
+
+ return 0;
+}
+
+static int
+hinic_pci_verify_fw_ver(struct rte_eth_dev *eth_dev,
+ struct rte_devargs *devargs)
+{
+ struct rte_kvargs *kvlist;
+ const char *hinic_fw_ver_arg = ETH_HINIC_FW_VER;
+ int ret = HINIC_OK;
+ char ver_str[64] = {0};
+
+ if (hinic_fw_version_get(eth_dev, ver_str, 64) != HINIC_OK)
+ PMD_DRV_LOG(ERR, "Failed to get FW versi
[dpdk-dev] [PATCH v2 09/11] net/hinic: add RX module
Add code for RX module.
Signed-off-by: Ziyang Xuan
---
drivers/net/hinic/Makefile | 2 +-
drivers/net/hinic/hinic_pmd_rx.c | 919 +++
2 files changed, 920 insertions(+), 1 deletion(-)
create mode 100644 drivers/net/hinic/hinic_pmd_rx.c
diff --git a/drivers/net/hinic/Makefile b/drivers/net/hinic/Makefile
index 6e19acf08..dd40216ec 100644
--- a/drivers/net/hinic/Makefile
+++ b/drivers/net/hinic/Makefile
@@ -77,7 +77,7 @@ SRCS-$(CONFIG_RTE_LIBRTE_HINIC_PMD) += hinic_pmd_wq.c
SRCS-$(CONFIG_RTE_LIBRTE_HINIC_PMD) += hinic_pmd_dpdev.c
#SRCS-$(CONFIG_RTE_LIBRTE_HINIC_PMD) += hinic_pmd_ethdev.c
-#SRCS-$(CONFIG_RTE_LIBRTE_HINIC_PMD) += hinic_pmd_rx.c
+SRCS-$(CONFIG_RTE_LIBRTE_HINIC_PMD) += hinic_pmd_rx.c
#SRCS-$(CONFIG_RTE_LIBRTE_HINIC_PMD) += hinic_pmd_tx.c
# this lib depends upon:
DEPDIRS-$(CONFIG_RTE_LIBRTE_HINIC_PMD) += lib/librte_eal lib/librte_ether
diff --git a/drivers/net/hinic/hinic_pmd_rx.c b/drivers/net/hinic/hinic_pmd_rx.c
new file mode 100644
index 0..27e723764
--- /dev/null
+++ b/drivers/net/hinic/hinic_pmd_rx.c
@@ -0,0 +1,919 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2017 Huawei Technologies Co., Ltd
+ */
+
+#include
+#include
+#include
+#ifdef __ARM64_NEON__
+#include
+#endif
+
+#include "hinic_pmd_ethdev.h"
+#include "hinic_pmd_rx.h"
+
+#ifdef HINIC_XSTAT_RXBUF_INFO
+static void hinic_rxq_buffer_done_count(struct hinic_rxq *rxq)
+{
+ u16 sw_ci, avail_pkts = 0, hit_done = 0, cqe_hole = 0;
+ u32 status;
+ volatile struct hinic_rq_cqe *rx_cqe;
+
+ for (sw_ci = 0; sw_ci < rxq->q_depth; sw_ci++) {
+ rx_cqe = &rxq->rx_cqe[sw_ci];
+
+ /* test current ci is done */
+ status = rx_cqe->status;
+ if (!HINIC_GET_RX_DONE_BE(status)) {
+ if (hit_done) {
+ cqe_hole++;
+ hit_done = 0;
+ }
+ continue;
+ }
+
+ avail_pkts++;
+ hit_done = 1;
+ }
+
+ rxq->rxq_stats.rx_avail = avail_pkts;
+ rxq->rxq_stats.rx_hole = cqe_hole;
+}
+#endif
+
+void hinic_rxq_get_stats(struct hinic_rxq *rxq, struct hinic_rxq_stats *stats)
+{
+ if (!rxq || !stats)
+ return;
+
+#ifdef HINIC_XSTAT_RXBUF_INFO
+ rxq->rxq_stats.rx_mbuf = (rxq->q_depth)
+ - HINIC_GET_RQ_FREE_WQEBBS(rxq);
+
+ hinic_rxq_buffer_done_count(rxq);
+#endif
+#ifdef HINIC_XSTAT_MBUF_USE
+ rxq->rxq_stats.left_mbuf = rxq->rxq_stats.alloc_mbuf
+ - rxq->rxq_stats.free_mbuf;
+#endif
+ memcpy(stats, &rxq->rxq_stats, sizeof(rxq->rxq_stats));
+}
+
+void hinic_rxq_stats_reset(struct hinic_rxq *rxq)
+{
+ struct hinic_rxq_stats *rxq_stats;
+
+ if (rxq == NULL)
+ return;
+
+ rxq_stats = &rxq->rxq_stats;
+ memset(rxq_stats, 0, sizeof(*rxq_stats));
+}
+
+/* mbuf alloc and free */
+static inline struct rte_mbuf *hinic_rte_rxmbuf_alloc(struct rte_mempool *mp)
+{
+ struct rte_mbuf *m;
+
+ m = rte_mbuf_raw_alloc(mp);
+ return m;
+}
+
+static int hinic_rx_alloc_cqe(struct hinic_rxq *rxq)
+{
+ size_t cqe_mem_size;
+
+ /* allocate continuous cqe memory for saving number of memory zone */
+ cqe_mem_size = sizeof(struct hinic_rq_cqe) * rxq->q_depth;
+ rxq->cqe_start_vaddr = dma_zalloc_coherent(rxq->nic_dev, cqe_mem_size,
+ &rxq->cqe_start_paddr,
+ GFP_KERNEL);
+ if (!rxq->cqe_start_vaddr) {
+ PMD_DRV_LOG(ERR, "Allocate cqe dma memory failed");
+ return -ENOMEM;
+ }
+
+ rxq->rx_cqe = (struct hinic_rq_cqe *)rxq->cqe_start_vaddr;
+
+ return HINIC_OK;
+}
+
+static void hinic_rx_free_cqe(struct hinic_rxq *rxq)
+{
+ size_t cqe_mem_size;
+
+ cqe_mem_size = sizeof(struct hinic_rq_cqe) * rxq->q_depth;
+ dma_free_coherent(rxq->nic_dev, cqe_mem_size,
+ rxq->cqe_start_vaddr,
+ rxq->cqe_start_paddr);
+ rxq->cqe_start_vaddr = NULL;
+}
+
+static int hinic_rx_fill_wqe(struct hinic_rxq *rxq)
+{
+ hinic_nic_dev *nic_dev = rxq->nic_dev;
+ struct hinic_rq_wqe *rq_wqe;
+ dma_addr_t buf_dma_addr, cqe_dma_addr;
+ u16 pi = 0;
+ int rq_wqe_len;
+ int i;
+
+ buf_dma_addr = 0;
+ cqe_dma_addr = rxq->cqe_start_paddr;
+ for (i = 0; i < rxq->q_depth; i++) {
+ rq_wqe = (struct hinic_rq_wqe *)
+ hinic_get_rq_wqe(nic_dev->hwdev, rxq->q_id, &pi);
+ if (!rq_wqe) {
+ PMD_DRV_LOG(ERR, "Get rq wqe failed");
+ break;
+ }
+
+ hinic_prepare_rq_wqe(rq_wqe, pi, buf_dma_addr, cqe_dma_addr);
+ cqe_dma_addr += sizeof(s
[dpdk-dev] [PATCH v2 10/11] net/hinic: add TX module
Add code for TX module
Signed-off-by: Ziyang Xuan
---
drivers/net/hinic/Makefile |2 +-
drivers/net/hinic/hinic_pmd_tx.c | 1070 ++
2 files changed, 1071 insertions(+), 1 deletion(-)
create mode 100644 drivers/net/hinic/hinic_pmd_tx.c
diff --git a/drivers/net/hinic/Makefile b/drivers/net/hinic/Makefile
index dd40216ec..f488fa430 100644
--- a/drivers/net/hinic/Makefile
+++ b/drivers/net/hinic/Makefile
@@ -78,7 +78,7 @@ SRCS-$(CONFIG_RTE_LIBRTE_HINIC_PMD) += hinic_pmd_wq.c
SRCS-$(CONFIG_RTE_LIBRTE_HINIC_PMD) += hinic_pmd_dpdev.c
#SRCS-$(CONFIG_RTE_LIBRTE_HINIC_PMD) += hinic_pmd_ethdev.c
SRCS-$(CONFIG_RTE_LIBRTE_HINIC_PMD) += hinic_pmd_rx.c
-#SRCS-$(CONFIG_RTE_LIBRTE_HINIC_PMD) += hinic_pmd_tx.c
+SRCS-$(CONFIG_RTE_LIBRTE_HINIC_PMD) += hinic_pmd_tx.c
# this lib depends upon:
DEPDIRS-$(CONFIG_RTE_LIBRTE_HINIC_PMD) += lib/librte_eal lib/librte_ether
DEPDIRS-$(CONFIG_RTE_LIBRTE_HINIC_PMD) += lib/librte_mempool lib/librte_mbuf
diff --git a/drivers/net/hinic/hinic_pmd_tx.c b/drivers/net/hinic/hinic_pmd_tx.c
new file mode 100644
index 0..8aa3e5367
--- /dev/null
+++ b/drivers/net/hinic/hinic_pmd_tx.c
@@ -0,0 +1,1070 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2017 Huawei Technologies Co., Ltd
+ */
+
+#include
+#include
+#include
+#include
+#include
+#ifdef __ARM64_NEON__
+#include
+#endif
+
+#include "hinic_pmd_ethdev.h"
+#include "hinic_pmd_tx.h"
+
+/* packet header and tx offload info */
+#define VXLANLEN(8)
+#define MAX_PLD_OFFSET (221)
+#define MAX_SINGLE_SGE_SIZE (65536)
+#define TSO_ENABLE (1)
+
+#define HINIC_NONTSO_PKT_MAX_SGE (17) /* non-tso max sge 17 */
+#define HINIC_NONTSO_SEG_NUM_INVALID(num) ((num) > HINIC_NONTSO_PKT_MAX_SGE)
+
+#define HINIC_TSO_PKT_MAX_SGE (127)/* tso max sge 127 */
+#define HINIC_TSO_SEG_NUM_INVALID(num) ((num) > HINIC_TSO_PKT_MAX_SGE)
+
+#define HINIC_TX_CKSUM_OFFLOAD_MASK ( \
+ PKT_TX_IP_CKSUM | \
+ PKT_TX_TCP_CKSUM | \
+ PKT_TX_UDP_CKSUM | \
+ PKT_TX_SCTP_CKSUM | \
+ PKT_TX_OUTER_IP_CKSUM | \
+ PKT_TX_TCP_SEG)
+
+/* sizeof(struct hinic_sq_bufdesc) == 16, shift 4 */
+#define HINIC_BUF_DESC_SIZE(nr_descs) \
+ (SIZE_8BYTES(((u32)nr_descs) << 4))
+
+/* tx offload info */
+struct hinic_tx_offload_info {
+ u8 outer_l2_len;
+ u8 outer_l3_type;
+ u8 outer_l3_len;
+
+ u8 inner_l2_len;
+ u8 inner_l3_type;
+ u8 inner_l3_len;
+
+ u8 tunnel_length;
+ u8 tunnel_type;
+ u8 inner_l4_type;
+ u8 inner_l4_len;
+
+ u8 payload_offset;
+ u8 inner_l4_tcp_udp;
+};
+
+/* tx sge info */
+struct hinic_wqe_info {
+ u16 pi;
+ u16 owner;
+ u16 around;
+ u16 seq_wqebbs;
+ u16 sge_cnt;
+ u16 cpy_mbuf_cnt;
+};
+
+static inline void hinic_sq_wqe_cpu_to_be32(void *data, int nr_wqebb)
+{
+ int i;
+#ifndef __ARM64_NEON__
+ __m128i *wqe_line = (__m128i *)data;
+ __m128i shuf_mask = _mm_set_epi8(12, 13, 14, 15, 8, 9, 10,
+11, 4, 5, 6, 7, 0, 1, 2, 3);
+
+ for (i = 0; i < nr_wqebb; i++) {
+ /* convert 64B wqebb using 4 SSE instructions */
+ wqe_line[0] = _mm_shuffle_epi8(wqe_line[0], shuf_mask);
+ wqe_line[1] = _mm_shuffle_epi8(wqe_line[1], shuf_mask);
+ wqe_line[2] = _mm_shuffle_epi8(wqe_line[2], shuf_mask);
+ wqe_line[3] = _mm_shuffle_epi8(wqe_line[3], shuf_mask);
+ wqe_line += 4;
+ }
+#else
+ uint8x16_t *wqe_line = (uint8x16_t *)data;
+ const uint8x16_t shuf_mask = {3, 2, 1, 0, 7, 6, 5, 4, 11, 10,
+ 9, 8, 15, 14, 13, 12};
+
+ for (i = 0; i < nr_wqebb; i++) {
+ wqe_line[0] = vqtbl1q_u8(wqe_line[0], shuf_mask);
+ wqe_line[1] = vqtbl1q_u8(wqe_line[1], shuf_mask);
+ wqe_line[2] = vqtbl1q_u8(wqe_line[2], shuf_mask);
+ wqe_line[3] = vqtbl1q_u8(wqe_line[3], shuf_mask);
+ wqe_line += 4;
+ }
+#endif
+}
+
+static inline void hinic_sge_cpu_to_be32(void *data, int nr_sge)
+{
+ int i;
+#ifndef __ARM64_NEON__
+ __m128i *sge_line = (__m128i *)data;
+ __m128i shuf_mask = _mm_set_epi8(12, 13, 14, 15, 8, 9, 10,
+11, 4, 5, 6, 7, 0, 1, 2, 3);
+
+ for (i = 0; i < nr_sge; i++) {
+ /* convert 16B sge using 1 SSE instructions */
+ *sge_line = _mm_shuffle_epi8(*sge_line, shuf_mask);
+ sge_line++;
+ }
+#else
+ uint8x16_t *sge_line = (uint8x16_t *)data;
+ const uint8x16_t shuf_mask = {3, 2, 1, 0, 7, 6, 5, 4, 11, 10,
+ 9, 8, 15, 14, 13, 12};
+
+ for (i = 0; i < nr_sge; i++) {
+ *sge_line = vqtbl1q_u8(*sge_line, shuf_mask);
+
Re: [dpdk-dev] [PATCH v2 08/11] net/hinic: add hinic PMD build and doc files
On Wed, 29 May 2019 11:50:36 +0800
Ziyang Xuan wrote:
> +Huawei hinic
> +M: Xiaoyun Wang
> +M: Ziyang Xuan
> +M: Guoyang Zhou
> +M: Rami Rosen
You should remove Rami, he will be missed.
> diff --git a/drivers/net/hinic/base/hinic_logs.c
> b/drivers/net/hinic/base/hinic_logs.c
> new file mode 100644
> index 0..1e9b5ad92
> --- /dev/null
> +++ b/drivers/net/hinic/base/hinic_logs.c
> @@ -0,0 +1,16 @@
> +/* SPDX-License-Identifier: BSD-3-Clause
> + * Copyright(c) 2017 Huawei Technologies Co., Ltd
> + */
> +
> +
> +#include "hinic_logs.h"
> +
> +int hinic_logtype;
> +
> +RTE_INIT(hinic_init_log)
> +{
> + hinic_logtype = rte_log_register("pmd.net.hinic");
> + if (hinic_logtype >= 0)
> + rte_log_set_level(hinic_logtype, RTE_LOG_INFO);
> +}
> +
You could just put this in some other file
> diff --git a/drivers/net/hinic/base/hinic_logs.h
> b/drivers/net/hinic/base/hinic_logs.h
> new file mode 100644
> index 0..ed84a37a6
> --- /dev/null
> +++ b/drivers/net/hinic/base/hinic_logs.h
> @@ -0,0 +1,31 @@
> +/* SPDX-License-Identifier: BSD-3-Clause
> + * Copyright(c) 2017 Huawei Technologies Co., Ltd
> + */
> +
> +#ifndef _HINIC_LOGS_H_
> +#define _HINIC_LOGS_H_
> +
> +#ifdef __cplusplus
> +#if __cplusplus
> +extern "C"{
> +#endif
> +#endif /* __cplusplus */
> +#include
Why all the C++ guards, this is a private file.
> +error_cflags = ['-Wno-unused-value',
> + '-Wno-unused-but-set-variable',
> + '-Wno-unused-variable',
> +]
Please fix these errors instead of masking them
> +/**
> + * Atomically writes the link status information into global
> + * struct rte_eth_dev.
> + */
> +int hinic_dev_atomic_write_link_status(struct rte_eth_dev *dev,
> +struct rte_eth_link *link)
> +{
> + struct rte_eth_link *dst = &dev->data->dev_link;
> + struct rte_eth_link *src = link;
> +
> + if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst,
> + *(uint64_t *)src) == 0)
> + return HINIC_ERROR;
> +
> + return HINIC_OK;
> +}
There already is code in rte_ethdev in current versions of DPDK to do this.
[dpdk-dev] [PATCH] app/test-crypto-perf: fix csv format
From: Ankur Dwivedi
The format for printing float is incorrect for the following fields
in pmd-cyclecount test: Cycles/Op,Cycles/Enq,Cycles/Deq.
Currently, the format is %.f3. This format will round off the number to
the nearest integer and append a 3 after that.
This patch changes the format to %.3f. This will print the number as a
floating point with a precision of 3 fractional digits.
Fixes: 96dfeb609be1 ("app/crypto-perf: add new PMD benchmarking mode")
Signed-off-by: Ankur Dwivedi
Signed-off-by: Anoob Joseph
---
app/test-crypto-perf/cperf_test_pmd_cyclecount.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/app/test-crypto-perf/cperf_test_pmd_cyclecount.c
b/app/test-crypto-perf/cperf_test_pmd_cyclecount.c
index 92af5ec..f08a226 100644
--- a/app/test-crypto-perf/cperf_test_pmd_cyclecount.c
+++ b/app/test-crypto-perf/cperf_test_pmd_cyclecount.c
@@ -16,7 +16,7 @@
#define PRETTY_HDR_FMT "%12s%12s%12s%12s%12s%12s%12s%12s%12s%12s\n\n"
#define PRETTY_LINE_FMT "%12u%12u%12u%12u%12u%12u%12u%12.0f%12.0f%12.0f\n"
#define CSV_HDR_FMT "%s,%s,%s,%s,%s,%s,%s,%s,%s,%s\n"
-#define CSV_LINE_FMT "%10u;%10u;%u;%u;%u;%u;%u;%.f3;%.f3;%.f3\n"
+#define CSV_LINE_FMT "%10u;%10u;%u;%u;%u;%u;%u;%.3f;%.3f;%.3f\n"
struct cperf_pmd_cyclecount_ctx {
uint8_t dev_id;
--
2.7.4
Re: [dpdk-dev] [PATCH v3] cryptodev: make xform key pointer constant
> -Original Message- > From: Zhang, Roy Fan > Sent: Tuesday, May 28, 2019 3:22 PM > To: Akhil Goyal ; Thomas Monjalon > > Cc: dev@dpdk.org; Trahe, Fiona ; Kusztal, ArkadiuszX > > Subject: RE: [dpdk-dev] [PATCH v3] cryptodev: make xform key pointer constant > > Hi Akhil, > > This patch was sent in 19.05 but I just found it can be applied to 19.08 RC0 > cleanly. > Do you think I need to send a v4 (with no change) for 19.08? > No need to send a v4, I will take v3. Planning to take some patches next week. Will let you know in case of some issue.
Re: [dpdk-dev] [PATCH v3 4/8] net/bnxt: implement vector mode driver
On 5/28/19 9:23 PM, Lance Richardson wrote: diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c index 5b3932b7b..bccba445a 100644 --- a/drivers/net/bnxt/bnxt_ethdev.c +++ b/drivers/net/bnxt/bnxt_ethdev.c @@ -638,13 +638,73 @@ static int bnxt_scattered_rx(struct rte_eth_dev *eth_dev) buf_size = (uint16_t)(rte_pktmbuf_data_room_size(rxq->mb_pool) - RTE_PKTMBUF_HEADROOM); - if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len + - VLAN_TAG_SIZE * BNXT_NUM_VLANS > buf_size) + if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len > buf_size) return 1; } return 0; } It seems you forgot to squash this in patch 3.
