> -Original Message-
> From: Rong, Leyi
> Sent: Monday, September 23, 2019 2:03 PM
> To: Lu, Wenzhuo ; Zhang, Qi Z
> ; Ye, Xiaolong
> Cc: dev@dpdk.org; Rong, Leyi
> Subject: [PATCH v4] net/iavf: enable AVX2 for iavf
>
> This patch enables AVX data path for iavf PMD.
>
> Signed-off-by
Remove unused code.
Signed-off-by: Paul M Stillwell Jr
Signed-off-by: Qi Zhang
---
drivers/net/ice/base/ice_common.c| 4
drivers/net/ice/base/ice_common.h| 2 --
drivers/net/ice/base/ice_flex_pipe.c | 1 -
drivers/net/ice/base/ice_flow.h | 3 ---
4 files changed, 10 deletions(
Add support for IPv6 traffic class (tc) field for switch rule.
Correct ice_ipv6_hdr based on the IPv6 Protocol using bitfields.
Add big/little endian convert for tc field before it is inserted,
since tc is only one byte and also does not have a byte-aligned
offset.
Signed-off-by: Junfeng Guo
Sign
Previously, switch code would use only pre-reserved index sl
ots at the end of each field vector for recipe result index
locations. This patch adds code that detects other internal
empty index slots that could potentially be used. For each
recipe that is added, a determ ination is made as to whethe
ice_cp_bitmap() already exists and should be used instead of
using ice_memcpy(). Note, there are a couple comments that sugges
t using a bitmap-specific copy function, but those are not correct
since the source block of memory is not a bitmap.
Signed-off-by: Bruce Allan
Signed-off-by: Paul M Sti
Remove the error log message when attempting to download a p
ackage that has an unsupported version.
Signed-off-by: Dan Nowlin
Signed-off-by: Paul M Stillwell Jr
Signed-off-by: Qi Zhang
---
drivers/net/ice/base/ice_flex_pipe.c | 16
1 file changed, 4 insertions(+), 12 deletion
As title says, fix an alignment issue.
Fixes: 51d04e4933e3 ("net/ice/base: add flexible pipeline module")
Cc: sta...@dpdk.org
Signed-off-by: Tony Nguyen
Signed-off-by: Paul M Stillwell Jr
Signed-off-by: Qi Zhang
---
drivers/net/ice/base/ice_flex_pipe.c | 5 +++--
1 file changed, 3 insertions(
As title says, fix an alignment issue.
Fixes: 51d04e4933e3 ("net/ice/base: add flexible pipeline module")
Cc: sta...@dpdk.org
Signed-off-by: Tony Nguyen
Signed-off-by: Paul M Stillwell Jr
Signed-off-by: Qi Zhang
---
drivers/net/ice/base/ice_flex_pipe.c | 5 +++--
1 file changed, 3 insertions(
Before this patch, IPv4 and UDP inner hash rule will be over
written by later rules after RSS initialization phase. This is
because the PTYPE bitmap table cover some PTYPEs belong to another
PTGs. And some PTYPEs are reserved. Remove these PTYPEs in TCP,
UDP, SCTP and ipv4 bitmap table.
Fixes: aa1
Add two bytes to meet the requirement of 4 byte alignment for dummy
packet for creating switch rule for PPPoE.
Fixes: 032b6c617a96 ("net/ice/base: add support for GTP and PPPoE protocols")
Signed-off-by: Junfeng Guo
Signed-off-by: Paul M Stillwell Jr
Signed-off-by: Qi Zhang
---
drivers/net/ic
Macros no longer be used and can be removed
Signed-off-by: Bruce Allan
Signed-off-by: Paul M Stillwell Jr
Signed-off-by: Qi Zhang
---
drivers/net/ice/base/ice_type.h | 5 -
1 file changed, 5 deletions(-)
diff --git a/drivers/net/ice/base/ice_type.h b/drivers/net/ice/base/ice_type.h
index
Before this patch, RSS tunneled rules can not be destroyed at runtime.
This is because it can not find the existing matching profile for tunnels.
segs[0] should always be zero and all matched, segs[1] for inner part.
It only construct one segment.
This patch modifies construct segment in ice_rem_r
Add GTPU qfi field support for flow director. Note that for GTPU pkt,
only qfi field (6 bits) can be set for FD. The supported GTPU pkts are
defined as:
ICE_FLTR_PTYPE_NONF_IPV4_GTPU_IPV4_UDP,
ICE_FLTR_PTYPE_NONF_IPV4_GTPU_IPV4_TCP,
ICE_FLTR_PTYPE_NONF_IPV4_GTPU_IPV4_ICMP,
ICE_FLTR_PTYPE_NONF_IPV4_
Add parameter "symm" to rss configuration APIs.
When symm is 1, Symmetric Teoplitz Hash can be enabled by
configuring GLQF_HSYMM properly.
NOTE:
Symmetric Teoplitz hash will work only if hash schema of
VSIQF_HASH_CTL be configured to 01b and it is assumed be enabled
in PMD.
Signed-off-by: Qi Zhan
Before this patch, if set rule for IPv4 first and then set rule
for TCP with IPv4. The first rule for inner IP will be overwritten
by TCP rule. This is because MAC_IPV6_TUN_MAC_IPV4_PAY using the
same ptgs PTG_TUN_INNER_IPV4_OTHER with MAC_IPV4_TUN_MAC_IPV4_PAY,
this ptype should not in TCP bitmap.
ice_memdup() is preferred over an alloc immediately followed by a copy.
Signed-off-by: Bruce Allan
Signed-off-by: Paul M Stillwell Jr
Signed-off-by: Qi Zhang
---
drivers/net/ice/base/ice_flex_pipe.c | 5 ++---
drivers/net/ice/base/ice_switch.c| 7 ++-
2 files changed, 4 insertions(+),
Update FW API minor version to align to current value advertised
by FW in NVM images.
Signed-off-by: Kevin Scott
Signed-off-by: Paul M Stillwell Jr
Signed-off-by: Qi Zhang
---
drivers/net/ice/base/ice_controlq.h | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/net/ic
Add dest MAC address support so that this field can be matched when
we set Flow Director filter with dst addr for MAC.
Signed-off-by: Junfeng Guo
Signed-off-by: Paul M Stillwell Jr
Signed-off-by: Qi Zhang
---
drivers/net/ice/base/ice_fdir.c | 19 +++
1 file changed, 19 insertio
The code in ice_sched_cleanup_all checks whether the port info is NULL
prior to calling ice_sched_clear_port.
More importantly, it also checks whether the port structure has been
initialized by checking its port_state field as well.
Signed-off-by: Jacob Keller
Signed-off-by: Paul M Stillwell Jr
Add fdir queue region support.
Signed-off-by: Qi Zhang
Signed-off-by: Paul M Stillwell Jr
---
drivers/net/ice/base/ice_fdir.c | 3 +++
drivers/net/ice/base/ice_fdir.h | 2 ++
2 files changed, 5 insertions(+)
diff --git a/drivers/net/ice/base/ice_fdir.c b/drivers/net/ice/base/ice_fdir.c
index e
Properly associate switch recipes to profiles. Previous code was
using the wrong bitfield for updating the associations, which was
causing other PFs to not properly identify and use existing
recipes. This sometimes resulted in rules not being added when it
should have been possible.
Signed-off-by:
Add two ptypes(MAC_PPPOE_IPV4_SCTP and MAC_PPPOE_IPV6_SCTP) in
sctp ptype bitmap to enable rss.
Signed-off-by: Zhirun Yan
Signed-off-by: Paul M Stillwell Jr
Signed-off-by: Qi Zhang
---
drivers/net/ice/base/ice_flow.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/n
When declaring the ice_prot_ext, and ice_prot_id_tbl structure,
we can use a fixed length array instead of a variable length one
which helps us catch future code changes that might desynchronize
the enum ice_protocol_type and the structs. This change also
necessitates removing the last member of t
Enable getting value from input to set up flow director counters,
so that the FDIR counters can count none, packets only, bytes only
or both packets and bytes as demanded.
Signed-off-by: Junfeng Guo
Signed-off-by: Paul M Stillwell Jr
Signed-off-by: Qi Zhang
---
drivers/net/ice/base/ice_fdir.c
Add support for switch rule about single-VLAN-PPPoE. Note that double
VLAN is not supported by the hardware at this point, therefore only
single-VLAN support for PPPoE is added.
Signed-off-by: Junfeng Guo
Signed-off-by: Paul M Stillwell Jr
Signed-off-by: Qi Zhang
---
drivers/net/ice/base/ice_p
The DPDK PF doesn't support SRIOV so remove the related iavf host
code.
Signed-off-by: Paul M Stillwell Jr
Signed-off-by: Qi Zhang
---
drivers/net/ice/base/ice_flow.c | 128
1 file changed, 128 deletions(-)
diff --git a/drivers/net/ice/base/ice_flow.c b
There could be an error returned from ice_fill_adv_dummy_packet()
so we need to propagate that to the caller. Additionally, the
call to ice_flow_xtract_pkt_flags() could also return an error so
we need to propagate it as well.
Also add in the correct offsets for GENEVE and VXLAN_GPE to the
dummy p
Add non-word aligned field support for ipv4 with ttl, tos and proto.
All these fields are one byte within one word. In order to match
bytes within the IPv4 header for flow director we need to use a mask.
Signed-off-by: Junfeng Guo
Signed-off-by: Paul M Stillwell Jr
Signed-off-by: Qi Zhang
---
Remove redundant empty lines
Signed-off-by: Paul M Stillwell Jr
Signed-off-by: Qi Zhang
---
drivers/net/ice/base/ice_adminq_cmd.h| 111 ---
drivers/net/ice/base/ice_bitops.h| 2 -
drivers/net/ice/base/ice_common.c| 25 ---
drivers/net/ice/b
Add non-word aligned field support for IPv6 with hlim, tc and proto.
All these fields are one byte within one word. In order to match
bytes within the IPv6 header for flow director we need to use a mask.
Signed-off-by: Junfeng Guo
Signed-off-by: Paul M Stillwell Jr
Signed-off-by: Qi Zhang
---
Before this patch, the logic of protocol header checking only support
non-tunneled packet. This patch remove the inner protocol in L3/L4 RSS
seg hdr mask and change the protocol header validation to reflect this.
So, for ice_add_rss_cfg(), the last parameter addl_hdrs could specify
the protocol hea
Add VXLAN tunnel training packets to flow director and change the
interface to support tunnel packets.
Signed-off-by: Henry Tieman
Signed-off-by: Paul M Stillwell Jr
Signed-off-by: Qi Zhang
---
drivers/net/ice/base/ice_fdir.c | 266
drivers/net/ice/base
The patchset depends on the first batch
http://patchwork.dpdk.org/project/dpdk/list/?series=6158&state=*
Key Features:
1) Add tunnel support for fdir
2) Add non-word aligned field support for fdir
3) Add dest mac field support for fdir
4) Add flow count support for fdir
5) Add queue region suppor
If reserve virtio header room by function rte_pktmbuf_prepend, both
segment data length and packet length of mbuf will be increased.
Data length will be equal to descriptor length, while packet length
should be decreased as virtio-net header won't be taken into packet.
Thus will cause mismatch in m
On Fri, Sep 20, 2019 at 12:36:33AM +0800, Marvin Liu wrote:
> Simplify flush shadow used ring function names as all shadow rings are
> reflect to used rings. No need to emphasize ring type.
I think the old name "flush_shadow_used_ring" is more readable
than the new name "flush_shadow". In the new
This patch enables AVX data path for iavf PMD.
Signed-off-by: Leyi Rong
---
v4:
- Romoves redundant IAVF_NO_VECTOR_FLAGS define.
v3:
- Adds release note.
v2:
- Fixes build error on aarch64.
---
doc/guides/rel_notes/release_19_11.rst | 3 +
drivers/net/iavf/Makefile | 21 +
d
On Fri, Sep 20, 2019 at 12:36:30AM +0800, Marvin Liu wrote:
> Burst enqueue function will first check whether descriptors are cache
> aligned. It will also check prerequisites in the beginning. Burst
> enqueue function not support chained mbufs, single packet enqueue
> function will handle it.
>
>
> -Original Message-
> From: Li, Xiaoyun
> Sent: Monday, September 9, 2019 11:27 AM
> To: Wu, Jingjing ; Wiles, Keith
> ; Maslekar,
> Omkar ; Liang, Cunming
> Cc: dev@dpdk.org; Li, Xiaoyun
> Subject: [PATCH v4 3/4] raw/ntb: add enqueue and dequeue functions
>
> Introduce enqueue and
On Fri, Sep 20, 2019 at 12:36:29AM +0800, Marvin Liu wrote:
> Add macro for unifying Clang/ICC/GCC unroll pragma format. Burst
> functions were contained of several small loops which optimized by
> compiler’s loop unrolling pragma.
>
> Signed-off-by: Marvin Liu
>
> diff --git a/lib/librte_vhost/
> -Original Message-
> From: Yang, Qiming
> Sent: Monday, September 23, 2019 11:26
> To: Rong, Leyi ; Wang, Haiyue ;
> Lu, Wenzhuo
> ; Zhang, Qi Z ; Ye, Xiaolong
>
> Cc: dev@dpdk.org
> Subject: RE: [dpdk-dev] [PATCH v4 3/6] net/ice: add protocol extraction
> support for per Rx queue
>
> static int
> -ntb_xstats_reset(struct rte_rawdev *dev __rte_unused,
> - const uint32_t ids[] __rte_unused,
> - uint32_t nb_ids __rte_unused)
> +ntb_xstats_reset(struct rte_rawdev *dev,
> + const uint32_t ids[],
> + uint32_t nb_ids)
> {
> -
Hi
> -Original Message-
> From: Wu, Jingjing
> Sent: Monday, September 23, 2019 10:50
> To: Li, Xiaoyun ; Wiles, Keith ;
> Maslekar, Omkar ; Liang, Cunming
>
> Cc: dev@dpdk.org
> Subject: RE: [PATCH v4 1/4] raw/ntb: setup ntb queue
>
> <...>
> > +static void
> > +ntb_rxq_release(struct n
> -Original Message-
> From: dev [mailto:dev-boun...@dpdk.org] On Behalf Of Leyi Rong
> Sent: Thursday, September 19, 2019 2:26 PM
> To: Wang, Haiyue ; Lu, Wenzhuo
> ; Zhang, Qi Z ; Ye, Xiaolong
>
> Cc: dev@dpdk.org
> Subject: [dpdk-dev] [PATCH v4 3/6] net/ice: add protocol extraction s
<...>
> +static void
> +ntb_rxq_release(struct ntb_rx_queue *rxq)
> +{
> + if (!rxq) {
> + NTB_LOG(ERR, "Pointer to rxq is NULL");
> + return;
> + }
> +
> + ntb_rxq_release_mbufs(rxq);
> +
> + rte_free(rxq->sw_ring);
> + rte_free(rxq);
It' better to free
> -Original Message-
> From: Di, ChenxuX
> Sent: Thursday, September 19, 2019 10:48 AM
> To: dev@dpdk.org
> Cc: Yang, Qiming ; Di, ChenxuX
>
> Subject: [PATCH v4 0/5] drivers/net: release port upon close
>
> Set RTE_ETH_DEV_CLOSE_REMOVE upon probe so all the private resources
> for the
Hi,
We are evaluating dpdk on windows10 64 bit desktop version ( linux runs fine ).
Build system is setup on PC, but I see that dpdk is supported on windows server
2016/2019 only.
Can we run dpdk on windows 10 ?
Best regards,
srinivas
Hi All,
While testing 18.11.3-rc1 on Azure we faced following problems:
1. VERIFY-DPDK-OVS - This test case builds OVS (OpenVSwitch) with DPDK support
on Ubuntu distro (16.04 and 18.04) and tests if the OVS DPDK ports can be
created. This testcase aborted, after diagnosis we found out that ins
> -Original Message-
> From: Matan Azrad
> Sent: Sunday, September 22, 2019 9:59
> To: Slava Ovsiienko ; dev@dpdk.org
> Cc: Raslan Darawsheh
> Subject: RE: [PATCH] net/mlx5: fix transmit doorbell register write barrier
>
> Hi Slava
>
> Questions inline.
>
[.. skipped ...]
> > + /*
From: Dekel Peled
> In the process of recovery from error CQE, when using vectorized Rx burst,
> the initialization of data length in mbufs was not done.
> As a result the wrong length was left written in mbuf, causing memory
> overwrite or wrong error report.
>
> This patch fixes the initializ
In the process of recovery from error CQE, when using vectorized Rx
burst, the initialization of data length in mbufs was not done.
As a result the wrong length was left written in mbuf, causing
memory overwrite or wrong error report.
This patch fixes the initialization of mbuf data length during
From: Kevin Traynor
> On 30/05/2019 11:20, Matan Azrad wrote:
> > Add support for data-path Rx and Tx completions with error handling:
> >
> > 1. Detect the error.
> > 2. Do not crash.
> > 3. Report it in statistics counters.
> > 4. Dump debug information to system log file.
> > 5. Recover the er
From: Viacheslav Ovsiienko
> The txq_uar_init() routine uses the uninitialized uar_mmap_offset field in
> 32-bit configurations due to this field is initialized after txq_uar_init()
> call.
>
> Fixes: 120dc4a7dcd3 ("net/mlx5: remove device register remap")
>
> Signed-off-by: Viacheslav Ovsiie
52 matches
Mail list logo
