The build was failing sometimes when building with multiple
parallel jobs:
# rm build/build/app/test/*res*
# make -j6
objcopy: 'resource.tmp': No such file
The reason is that each resource was built from the same temporary file.
The failure is seen because of a race condition when remo
It is better to keep the line "Fixes:" longer than 75 characters
than splitting.
Signed-off-by: Thomas Monjalon
---
scripts/check-git-log.sh | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/scripts/check-git-log.sh b/scripts/check-git-log.sh
index 153f378..4253d14 100755
--- a
Hi Markos,
> -Original Message-
> From: dev [mailto:dev-bounces at dpdk.org] On Behalf Of Markos Chandras
> Sent: Thursday, June 23, 2016 5:26 PM
> To: dev at dpdk.org
> Cc: Markos Chandras
> Subject: [dpdk-dev] [PATCH] e1000/base: Add missing braces to the 'if'
> statements
>
> Add the
Hi Thomas,
> -Original Message-
> From: Thomas Monjalon [mailto:thomas.monjalon at 6wind.com]
> Sent: Thursday, June 23, 2016 8:22 PM
> To: Lu, Wenzhuo
> Cc: dev at dpdk.org
> Subject: Re: [dpdk-dev] [PATCH v3] ethdev: fix DCB config issue on ixgbe
>
> 2016-06-23 01:04, Lu, Wenzhuo:
> >
Hi Stephen,
> -Original Message-
> From: Stephen Hemminger [mailto:stephen at networkplumber.org]
> Sent: Friday, June 24, 2016 4:51 AM
> To: Tan, Jianfeng
> Cc: dev at dpdk.org; yuanhan.liu at linux.intel.com; Xie, Huawei
> Subject: Re: [dpdk-dev] [PATCH] net/virtio-user: fix implicit int
Hi Yigit,
> -Original Message-
> From: Yigit, Ferruh
> Sent: Thursday, June 23, 2016 5:02 PM
> To: Tan, Jianfeng; dev at dpdk.org
> Cc: Xie, Huawei; rich.lane at bigswitch.com; yuanhan.liu at linux.intel.com;
> mst at redhat.com; nakajima.yoshihiro at lab.ntt.co.jp; p.fedin at
> samsung.c
According to 'rte_eth_stats' structure comments, 'imissed'
should represent RX error counting, but currently 'imissed' is
used to count TX error.
The patch replaces 'imissed' by 'oerrors'.
Fixes: ee584e9710b9 ("vhost: add driver on top of the library")
Signed-off-by: Tetsuya Mukawa
---
drivers/n
On Fri, Jun 24, 2016 at 11:04:20AM +0900, Tetsuya Mukawa wrote:
> According to 'rte_eth_stats' structure comments, 'imissed'
> should represent RX error counting, but currently 'imissed' is
> used to count TX error.
> The patch replaces 'imissed' by 'oerrors'.
>
> Fixes: ee584e9710b9 ("vhost: add
On Tue, Jun 21, 2016 at 09:15:03AM -0400, Aaron Conole wrote:
> Yuanhan Liu writes:
>
> > On Fri, Jun 17, 2016 at 11:32:36AM -0400, Aaron Conole wrote:
> >> Prior to this commit, the only way to add a vhost-user socket to the
> >> system is by relying on librte_vhost to open the unix domain socke
On 2016/06/24 11:17, Yuanhan Liu wrote:
> On Fri, Jun 24, 2016 at 11:04:20AM +0900, Tetsuya Mukawa wrote:
>> According to 'rte_eth_stats' structure comments, 'imissed'
>> should represent RX error counting, but currently 'imissed' is
>> used to count TX error.
>> The patch replaces 'imissed' by 'oe
On 6/23/16, 11:22 PM, "dev on behalf of Thomas Monjalon" wrote:
>> David Hunt (2):
>> mempool: support mempool handler operations
>> app/test: test mempool handler
>> mbuf: make default mempool ops configurable at build
>
>Applied, thanks for the nice feature
>
>I'm sorry David, the revisi
Hi,
One of my customers intends to buy HPE Ethernet 10Gb 2-port
560FLR-SFP+ Adapter
(http://www8.hp.com/h20195/v2/getpdf.aspx/c04111435.pdf?ver=8) for
running a DPDK based app.
I have never tested my app with the above NIC (always used X520 to test my app)
If someone has already tried with this N
2016-06-24 00:43, Lu, Wenzhuo:
> Thanks for this patch. But normally the code in the base directory is synced
> from the kernel driver. So we don't change it if there's no critical issue.
> It's easy for us to maintain it. Thanks.
I think a build error is critical enough.
>
> On Tue, Jun 21, 2016 at 09:15:03AM -0400, Aaron Conole wrote:
> > Yuanhan Liu writes:
> >
> > > On Fri, Jun 17, 2016 at 11:32:36AM -0400, Aaron Conole wrote:
> > >> Prior to this commit, the only way to add a vhost-user socket to the
> > >> system is by relying on librte_vhost to open the uni
On Fri, Jun 24, 2016 at 07:43:29AM +, Loftus, Ciara wrote:
> >
> > On Tue, Jun 21, 2016 at 09:15:03AM -0400, Aaron Conole wrote:
> > > Yuanhan Liu writes:
> > >
> > > > On Fri, Jun 17, 2016 at 11:32:36AM -0400, Aaron Conole wrote:
> > > >> Prior to this commit, the only way to add a vhost-use
Mainly on updating vhost-user part: we now support client mode.
Also refine some words, and add a bit more explanation.
And made an emphatic statement that you are suggested to use vhost-user
instead of vhost-cuse, because we have enhanced vhost-user a lot since
v2.2 (Actually, I doubt there are a
Hi,
I'm trying to add a feature to?DPDK and I'm having a hard time printing a
backtrace.
I tried using this[1] functions for printing, but it does not print more than
one function. Maybe it lacks the symbols it needs.
I tried compiling with "-rdynamic", but it breaks the compilation with an erro
On 6/23/2016 3:38 PM, Pablo de Lara wrote:
> Using gcc 6.1, in some cases, kni fails to compile
> because of unused variables:
>
> build/lib/librte_eal/linuxapp/kni/ixgbe_main.c:82:19:
> error: ?ixgbe_copyright?
> defined but not used [-Werror=unused-const-variable=]
> static const char ixgbe_copy
This patch-set add the support for floating VEB in i40e.
All the VFs VSIs can decide whether to connect to the legacy VEB/VEPA or
the floating VEB. When connect to the floating VEB a new floating VEB is
created. Now all the VFs need to connect to floating VEB or legacy VEB,
cannot connect to both o
Add the new floating VEB related arguments option in the devarg.
Using this parameter, all the applications can decide whether to use legacy
VEB/VEPA or floating VEB.
To enable this feature, the user should pass a devargs parameter to the
EAL like "-w 84:00.0,enable_floating_veb=1", and the applica
This patch add the support for floating VEB in i40e.
All the VFs VSIs can decide whether to connect to the legacy VEB/VEPA or
the floating VEB. When connect to the floating VEB a new floating VEB is
created. Now all the VFs need to connect to floating VEB or legacy VEB,
cannot connect to both of th
Hi Thomas, Markos,
> -Original Message-
> From: Thomas Monjalon [mailto:thomas.monjalon at 6wind.com]
> Sent: Friday, June 24, 2016 3:13 PM
> To: Lu, Wenzhuo
> Cc: dev at dpdk.org; Markos Chandras
> Subject: Re: [dpdk-dev] [PATCH] e1000/base: Add missing braces to the 'if'
> statements
>
Enhance mlx5 with a data path that bypasses Verbs.
The first half of this patchset removes support for functionality completely
rewritten in the second half (scatter/gather, inline send), while the data
path is refactored without Verbs.
The PMD remains usable during the transition.
This patchset
Fixes: 701c8d80c820 ("pci: support class id probing")
Signed-off-by: Nelio Laranjeiro
---
drivers/crypto/qat/rte_qat_cryptodev.c | 5 +
drivers/net/mlx4/mlx4.c| 18 ++
drivers/net/mlx5/mlx5.c| 24
drivers/net/nfp/nfp_n
Except for the first time when memory registration occurs, the lkey is
always cached. Since memory registration is slow and performs system calls,
performance can be improved by moving that code to its own function outside
of the data path so only the lookup code is left in the original inlined
fun
This is done in preparation of bypassing Verbs entirely for the data path
as a performance improvement. TX gather cannot be maintained during the
transition and will be reimplemented later.
Signed-off-by: Nelio Laranjeiro
Signed-off-by: Adrien Mazarguil
---
drivers/net/mlx5/mlx5_ethdev.c | 2
This is done in preparation of bypassing Verbs entirely for the data path
as a performance improvement. RX scatter cannot be maintained during the
transition and will be reimplemented later.
Signed-off-by: Nelio Laranjeiro
Signed-off-by: Adrien Mazarguil
---
drivers/net/mlx5/mlx5_ethdev.c | 31
There is no scatter/gather support anymore, CONFIG_RTE_LIBRTE_MLX5_SGE_WR_N
has no purpose and can be removed.
Signed-off-by: Nelio Laranjeiro
Signed-off-by: Adrien Mazarguil
---
config/common_base | 1 -
doc/guides/nics/mlx5.rst | 7 ---
drivers/net/mlx5/Makefile| 4
Inline TX will be fully managed by the PMD after Verbs is bypassed in the
data path. Remove the current code until then.
Signed-off-by: Nelio Laranjeiro
Signed-off-by: Adrien Mazarguil
---
config/common_base | 1 -
doc/guides/nics/mlx5.rst | 10 --
drivers/net/mlx5/Makefile
To keep the data path as efficient as possible, move fields only useful to
the control path into new structure txq_ctrl.
Signed-off-by: Nelio Laranjeiro
Signed-off-by: Adrien Mazarguil
---
drivers/net/mlx5/mlx5.c| 21 +++--
drivers/net/mlx5/mlx5_ethdev.c | 28 +++---
drivers/net/mlx5/
To keep the data path as efficient as possible, move fields only useful to
the control path into new structure rxq_ctrl.
Signed-off-by: Nelio Laranjeiro
Signed-off-by: Adrien Mazarguil
---
drivers/net/mlx5/mlx5.c | 6 +-
drivers/net/mlx5/mlx5_fdir.c | 8 +-
drivers/net/mlx5/mlx5_rxq.c
The latest version of Mellanox OFED exposes hardware definitions necessary
to implement data path operation bypassing Verbs. Update the minimum
version requirement to MLNX_OFED >= 3.3 and clean up compatibility checks
for previous releases.
Signed-off-by: Nelio Laranjeiro
Signed-off-by: Adrien Ma
These structures and macros extend those exposed by libmlx5 (in mlx5_hw.h)
to let the PMD manage work queue and completion queue elements directly.
Signed-off-by: Nelio Laranjeiro
Signed-off-by: Adrien Mazarguil
---
drivers/net/mlx5/mlx5_prm.h | 163
The intent is to replace the remaining compile-time options and environment
variables with a common mean of runtime configuration. This commit only
adds the kvargs handling code, subsequent commits will update the rest.
Signed-off-by: Nelio Laranjeiro
Signed-off-by: Adrien Mazarguil
---
drivers
These wrappers are meant to prevent code duplication later.
Signed-off-by: Nelio Laranjeiro
Signed-off-by: Adrien Mazarguil
---
drivers/net/mlx5/mlx5.h| 2 ++
drivers/net/mlx5/mlx5_ethdev.c | 34 --
drivers/net/mlx5/mlx5_txq.c| 2 +-
3 files changed
Bypass Verbs to improve RX performance.
Signed-off-by: Nelio Laranjeiro
Signed-off-by: Yaacov Hazan
Signed-off-by: Adrien Mazarguil
Signed-off-by: Vasily Philipov
---
drivers/net/mlx5/mlx5_ethdev.c | 4 +-
drivers/net/mlx5/mlx5_fdir.c | 2 +-
drivers/net/mlx5/mlx5_rxq.c| 303 +++
Bypass Verbs to improve Tx performance.
Signed-off-by: Nelio Laranjeiro
Signed-off-by: Yaacov Hazan
Signed-off-by: Adrien Mazarguil
---
drivers/net/mlx5/Makefile | 5 -
drivers/net/mlx5/mlx5_ethdev.c | 10 +-
drivers/net/mlx5/mlx5_mr.c | 4 +-
drivers/net/mlx5/mlx5_rxtx.c | 361
Mini (compressed) CQEs are returned by the NIC when PCI back pressure is
detected, in which case the first CQE64 contains common packet information
followed by a number of CQE8 providing the rest, followed by a matching
number of empty CQE64 entries to be used by software for decompression.
Before
From: Adrien Mazarguil
Replacing the variable countdown (which depends on the number of
descriptors) with a fixed relative threshold known at compile time improves
performance by reducing the TX queue structure footprint and the amount of
code to manage completions during a burst.
Completions ar
From: Yaacov Hazan
Implement send inline feature which copies packet data directly into WQEs
for improved latency. The maximum packet size and the minimum number of Tx
queues to qualify for inline send are user-configurable.
This feature is effective when HW causes a performance bottleneck.
Sig
This feature enables the TX burst function to emit up to 5 packets using
only two WQEs on devices that support it. Saves PCI bandwidth and improves
performance.
Signed-off-by: Nelio Laranjeiro
Signed-off-by: Adrien Mazarguil
Signed-off-by: Olga Shern
---
doc/guides/nics/mlx5.rst | 10 +
From: Adrien Mazarguil
Signed-off-by: Adrien Mazarguil
---
drivers/net/mlx5/mlx5_txq.c | 5 +
1 file changed, 5 insertions(+)
diff --git a/drivers/net/mlx5/mlx5_txq.c b/drivers/net/mlx5/mlx5_txq.c
index 4f17fb0..bae9f3d 100644
--- a/drivers/net/mlx5/mlx5_txq.c
+++ b/drivers/net/mlx5/mlx5_t
From: Adrien Mazarguil
The space necessary to store segmented packets cannot be known in advance
and must be verified for each of them.
Signed-off-by: Adrien Mazarguil
Signed-off-by: Nelio Laranjeiro
---
drivers/net/mlx5/mlx5_rxtx.c | 144 +++
1 file ch
From: Adrien Mazarguil
Compared to its previous incarnation, the software limit on the number of
mbuf segments is no more (previously MLX5_PMD_SGE_WR_N, set to 4 by
default) hence no need for linearization code and related buffers that
permanently consumed a non negligible amount of memory to han
From: Adrien Mazarguil
Since commit "mlx5: resurrect Tx gather support", older GCC versions (such
as 4.8.5) may complain about the following:
mlx5_rxtx.c: In function `mlx5_tx_burst':
mlx5_rxtx.c:705:25: error: `wqe' may be used uninitialized in this
function [-Werror=maybe-uninitialized]
From: Adrien Mazarguil
Toggling RX checksum offloads is already done at initialization time. This
code does not belong in rxq_rehash().
Signed-off-by: Adrien Mazarguil
Signed-off-by: Nelio Laranjeiro
---
drivers/net/mlx5/mlx5_rxq.c | 11 ---
1 file changed, 11 deletions(-)
diff --git
From: Adrien Mazarguil
The primary purpose of rxq_rehash() function is to stop and restart
reception on a queue after re-posting buffers. This may fail if the array
that temporarily stores existing buffers for reuse cannot be allocated.
Update rxq_rehash() to work on the target queue directly (n
This commit brings back Rx scatter and related support by the MTU update
function. The maximum number of segments per packet is not a fixed value
anymore (previously MLX5_PMD_SGE_WR_N, set to 4 by default) as it caused
performance issues when fewer segments were actually needed as well as
limitatio
> -Original Message-
> From: dev [mailto:dev-bounces at dpdk.org] On Behalf Of Catalin Vasile
> Sent: Friday, June 24, 2016 9:10 AM
> To: dev at dpdk.org
> Subject: [dpdk-dev] backtracing from within the code
>
> Hi,
>
> I'm trying to add a feature to?DPDK and I'm having a hard time prin
On Fri, Jun 24, 2016 at 04:29:38PM +0800, Zhe Tao wrote:
> Add the new floating VEB related arguments option in the devarg.
> Using this parameter, all the applications can decide whether to use legacy
> VEB/VEPA or floating VEB.
> To enable this feature, the user should pass a devargs parameter to
On Fri, Jun 24, 2016 at 12:44:18AM +0200, Thomas Monjalon wrote:
> It is better to keep the line "Fixes:" longer than 75 characters
> than splitting.
>
> Signed-off-by: Thomas Monjalon
Definite +1
Acked-by: Bruce Richardson
I can repeat this problem by "export EXTRA_CFLAGS="-O0 -g"" on Power8. But
I'm not sure why this happens. The "-O3 -g" option works properly. I'll
investigate more.
-Original Message-
From: Dumitrescu, Cristian [mailto:cristian.dumitre...@intel.com]
Sent: 2016?6?24? 1:26
To: N?lio Laranje
On Thu, Jun 16, 2016 at 01:59:46PM +, Zhang, Helin wrote:
>
>
> > -Original Message-
> > From: Xing, Beilei
> > Sent: Thursday, June 16, 2016 9:36 PM
> > To: Zhang, Helin
> > Cc: dev at dpdk.org; Xing, Beilei
> > Subject: [PATCH v4] e1000: configure VLAN TPID
> >
> > This patch ena
2016-06-24 09:25, Dumitrescu, Cristian:
> From: dev [mailto:dev-bounces at dpdk.org] On Behalf Of Catalin Vasile
> > I'm trying to add a feature to DPDK and I'm having a hard time printing a
> > backtrace.
> > I tried using this[1] functions for printing, but it does not print more
> > than one
>
The verbose option was available but not advertised.
Fixes: 6e38dfe21389 ("scripts: add verbose test build option")
Signed-off-by: Thomas Monjalon
---
scripts/test-build.sh | 1 +
1 file changed, 1 insertion(+)
diff --git a/scripts/test-build.sh b/scripts/test-build.sh
index 31d5583..7a9f623 1
> -Original Message-
> From: dev [mailto:dev-bounces at dpdk.org] On Behalf Of Yari Adan
> Petralanda
> Sent: Friday, June 17, 2016 12:00 PM
> To: dev at dpdk.org
> Subject: [dpdk-dev] [PATCH v2] hash: new function to retrieve a key given its
> position
>
> The function rte_hash_get_key_
On Thu, Jun 16, 2016 at 12:19:05PM -0700, Nelson Escobar wrote:
> For performance reasons, this patch uses 2 VIC RQs per RQ presented to
> DPDK.
>
> The VIC requires that each descriptor be marked as either a start of
> packet (SOP) descriptor or a non-SOP descriptor. A one RQ solution
> requires
On Fri, Jun 17, 2016 at 06:32:06AM +, Harish Patil wrote:
> >
> >Fixes: 540a211084a7 ("bnx2x: driver core")
> >
> >Signed-off-by: Chas Williams <3chas3 at gmail.com>
> >---
> > drivers/net/bnx2x/bnx2x_rxtx.c | 1 -
> > 1 file changed, 1 deletion(-)
> >
> >diff --git a/drivers/net/bnx2x/bnx2x_rxt
On Thu, Jun 16, 2016 at 10:22:46PM -0700, John Daley wrote:
> Pull in common VNIC code which enables querying for max egress
> packet size.
>
With this patch applied is the user able to query the max packet size, or is it
just that the driver is able to do so for use by other functions?
/Bruce
On Thu, Jun 16, 2016 at 10:22:47PM -0700, John Daley wrote:
> The max MTU is set to the max egress packet size allowed by the VIC
> minus the size of a an IPv4 L2 header with .1Q (18 bytes).
>
I think a bit more detail might be needed here. For example:
* What was the MTU set to by default befor
On Fri, Jun 24, 2016 at 11:59:19AM +0100, Bruce Richardson wrote:
> On Thu, Jun 16, 2016 at 10:22:47PM -0700, John Daley wrote:
> > The max MTU is set to the max egress packet size allowed by the VIC
> > minus the size of a an IPv4 L2 header with .1Q (18 bytes).
> >
>
> I think a bit more detail
On Thu, Jun 16, 2016 at 10:22:49PM -0700, John Daley wrote:
> Signed-off-by: John Daley
> ---
This patch should be squashed into the previous one.
/Bruce
On Mon, Jun 20, 2016 at 11:19:19AM +0800, Jianbo Liu wrote:
> On 17 June 2016 at 22:06, Jerin Jacob
> wrote:
> > made second cache line access behavior same as IA
> >
> > Signed-off-by: Jerin Jacob
> > ---
> > drivers/net/ixgbe/ixgbe_rxtx_vec_neon.c | 8
> > 1 file changed, 4 insertion
Hi Zhe,
On 6/24/2016 9:29 AM, Zhe Tao wrote:
> Add the new floating VEB related arguments option in the devarg.
> Using this parameter, all the applications can decide whether to use legacy
> VEB/VEPA or floating VEB.
> To enable this feature, the user should pass a devargs parameter to the
> EAL
On Fri, 24 Jun 2016 04:55:39 +
"Wiles, Keith" wrote:
> On 6/23/16, 11:22 PM, "dev on behalf of Thomas Monjalon" dpdk.org on behalf of thomas.monjalon at 6wind.com> wrote:
>
> >> David Hunt (2):
> >> mempool: support mempool handler operations
> >> app/test: test mempool handler
> >> m
The build was failing sometimes when building with multiple
parallel jobs:
# rm build/build/app/test/*res*
# make -j6
objcopy: 'resource.tmp': No such file
The reason is that each resource was built from the same temporary file.
The failure is seen because of a race condition when remo
2016-06-24 13:20, Jan Viktorin:
> thanks for the patchset. I am sorry, I didn't have any time for DPDK this week
> and didn't test it before applying. The current master produces the following
> error in my regular builds:
>
> INSTALL-LIB librte_eal.a
> == Build lib/librte_ring
> CC rte_ring.o
On Mon, Jun 20, 2016 at 12:27:46PM -0700, John Daley wrote:
> Negative array index write using variable pos as an index to array
> enic->fdir.nodes. Fixed by add array index check.
>
> Fixes: fefed3d1e62c ("enic: new driver") Coverity ID 13270
> Signed-off-by: John Daley
> ---
>
Applied to dpdk-
On Tue, Jun 21, 2016 at 02:05:57PM +0200, Jan Medala wrote:
> As requested, big patch splitted into logical pieces for easier review.
> Improved style and fixed icc compiler issues.
>
Thanks for the patch split. However, many of the patches don't have a commit
message describing (at a high level)
On Tue, Jun 21, 2016 at 04:58:20PM -0500, Ajit Khaparde wrote:
> This patch adds support for Cumulus+ Ethernet adapters.
> These Cumulus+ Ethernet adapters support 10Gb/25Gb/40Gb/50Gb speeds.
>
> Signed-off-by: Ajit Khaparde
Applied to dpdk-next-net/rel_16_07
/Bruce
On Wed, Jun 22, 2016 at 10:53:51AM +0800, Beilei Xing wrote:
> In current i40e codebase, if single VLAN header is added in a packet,
> it's treated as inner VLAN. Generally, a single VLAN header is
> treated as the outer VLAN header. So change corresponding register
> for single VLAN.
>
> Fixes: 1
Yuanhan Liu writes:
> On Fri, Jun 24, 2016 at 07:43:29AM +, Loftus, Ciara wrote:
>> >
>> > On Tue, Jun 21, 2016 at 09:15:03AM -0400, Aaron Conole wrote:
>> > > Yuanhan Liu writes:
>> > >
>> > > > On Fri, Jun 17, 2016 at 11:32:36AM -0400, Aaron Conole wrote:
>> > > >> Prior to this commit, t
The PCI test was failing because some fake devices had no PCI class.
Fixes: 1dbba1650c89 ("app/test: remove real PCI ids")
Signed-off-by: Thomas Monjalon
---
app/test/test_pci_sysfs/bus/pci/devices/:01:02.0/class | 1 +
app/test/test_pci_sysfs/bus/pci/devices/:02:ab.0/class | 1 +
2 fil
Sorry about this, it will need a v7, a bug was introduced in the
previous version to fix a checkpatch warning. kvargs were no more
interpreted.
I will resend a v7 after verifying everything.
Regards,
--
N?lio Laranjeiro
6WIND
On Fri, 24 Jun 2016 13:24:56 +0200
Thomas Monjalon wrote:
> 2016-06-24 13:20, Jan Viktorin:
> > thanks for the patchset. I am sorry, I didn't have any time for DPDK this
> > week
> > and didn't test it before applying. The current master produces the
> > following
> > error in my regular builds
Enhance mlx5 with a data path that bypasses Verbs.
The first half of this patchset removes support for functionality completely
rewritten in the second half (scatter/gather, inline send), while the data
path is refactored without Verbs.
The PMD remains usable during the transition.
This patchset
Fixes: 701c8d80c820 ("pci: support class id probing")
Signed-off-by: Nelio Laranjeiro
---
drivers/crypto/qat/rte_qat_cryptodev.c | 5 +
drivers/net/mlx4/mlx4.c| 18 ++
drivers/net/mlx5/mlx5.c| 24
drivers/net/nfp/nfp_n
Except for the first time when memory registration occurs, the lkey is
always cached. Since memory registration is slow and performs system calls,
performance can be improved by moving that code to its own function outside
of the data path so only the lookup code is left in the original inlined
fun
This is done in preparation of bypassing Verbs entirely for the data path
as a performance improvement. TX gather cannot be maintained during the
transition and will be reimplemented later.
Signed-off-by: Nelio Laranjeiro
Signed-off-by: Adrien Mazarguil
---
drivers/net/mlx5/mlx5_ethdev.c | 2
This is done in preparation of bypassing Verbs entirely for the data path
as a performance improvement. RX scatter cannot be maintained during the
transition and will be reimplemented later.
Signed-off-by: Nelio Laranjeiro
Signed-off-by: Adrien Mazarguil
---
drivers/net/mlx5/mlx5_ethdev.c | 31
There is no scatter/gather support anymore, CONFIG_RTE_LIBRTE_MLX5_SGE_WR_N
has no purpose and can be removed.
Signed-off-by: Nelio Laranjeiro
Signed-off-by: Adrien Mazarguil
---
config/common_base | 1 -
doc/guides/nics/mlx5.rst | 7 ---
drivers/net/mlx5/Makefile| 4
Inline TX will be fully managed by the PMD after Verbs is bypassed in the
data path. Remove the current code until then.
Signed-off-by: Nelio Laranjeiro
Signed-off-by: Adrien Mazarguil
---
config/common_base | 1 -
doc/guides/nics/mlx5.rst | 10 --
drivers/net/mlx5/Makefile
To keep the data path as efficient as possible, move fields only useful to
the control path into new structure txq_ctrl.
Signed-off-by: Nelio Laranjeiro
Signed-off-by: Adrien Mazarguil
---
drivers/net/mlx5/mlx5.c| 21 +++--
drivers/net/mlx5/mlx5_ethdev.c | 28 +++---
drivers/net/mlx5/
To keep the data path as efficient as possible, move fields only useful to
the control path into new structure rxq_ctrl.
Signed-off-by: Nelio Laranjeiro
Signed-off-by: Adrien Mazarguil
---
drivers/net/mlx5/mlx5.c | 6 +-
drivers/net/mlx5/mlx5_fdir.c | 8 +-
drivers/net/mlx5/mlx5_rxq.c
The latest version of Mellanox OFED exposes hardware definitions necessary
to implement data path operation bypassing Verbs. Update the minimum
version requirement to MLNX_OFED >= 3.3 and clean up compatibility checks
for previous releases.
Signed-off-by: Nelio Laranjeiro
Signed-off-by: Adrien Ma
These structures and macros extend those exposed by libmlx5 (in mlx5_hw.h)
to let the PMD manage work queue and completion queue elements directly.
Signed-off-by: Nelio Laranjeiro
Signed-off-by: Adrien Mazarguil
---
drivers/net/mlx5/mlx5_prm.h | 163
The intent is to replace the remaining compile-time options and environment
variables with a common mean of runtime configuration. This commit only
adds the kvargs handling code, subsequent commits will update the rest.
Signed-off-by: Nelio Laranjeiro
Signed-off-by: Adrien Mazarguil
---
drivers
These wrappers are meant to prevent code duplication later.
Signed-off-by: Nelio Laranjeiro
Signed-off-by: Adrien Mazarguil
---
drivers/net/mlx5/mlx5.h| 2 ++
drivers/net/mlx5/mlx5_ethdev.c | 34 --
drivers/net/mlx5/mlx5_txq.c| 2 +-
3 files changed
Bypass Verbs to improve RX performance.
Signed-off-by: Nelio Laranjeiro
Signed-off-by: Yaacov Hazan
Signed-off-by: Adrien Mazarguil
Signed-off-by: Vasily Philipov
---
drivers/net/mlx5/mlx5_ethdev.c | 4 +-
drivers/net/mlx5/mlx5_fdir.c | 2 +-
drivers/net/mlx5/mlx5_rxq.c| 303 +++
Bypass Verbs to improve Tx performance.
Signed-off-by: Nelio Laranjeiro
Signed-off-by: Yaacov Hazan
Signed-off-by: Adrien Mazarguil
---
drivers/net/mlx5/Makefile | 5 -
drivers/net/mlx5/mlx5_ethdev.c | 10 +-
drivers/net/mlx5/mlx5_mr.c | 4 +-
drivers/net/mlx5/mlx5_rxtx.c | 361
Mini (compressed) CQEs are returned by the NIC when PCI back pressure is
detected, in which case the first CQE64 contains common packet information
followed by a number of CQE8 providing the rest, followed by a matching
number of empty CQE64 entries to be used by software for decompression.
Before
From: Adrien Mazarguil
Replacing the variable countdown (which depends on the number of
descriptors) with a fixed relative threshold known at compile time improves
performance by reducing the TX queue structure footprint and the amount of
code to manage completions during a burst.
Completions ar
From: Yaacov Hazan
Implement send inline feature which copies packet data directly into WQEs
for improved latency. The maximum packet size and the minimum number of Tx
queues to qualify for inline send are user-configurable.
This feature is effective when HW causes a performance bottleneck.
Sig
This feature enables the TX burst function to emit up to 5 packets using
only two WQEs on devices that support it. Saves PCI bandwidth and improves
performance.
Signed-off-by: Nelio Laranjeiro
Signed-off-by: Adrien Mazarguil
Signed-off-by: Olga Shern
---
doc/guides/nics/mlx5.rst | 10 +
From: Adrien Mazarguil
Signed-off-by: Adrien Mazarguil
---
drivers/net/mlx5/mlx5_txq.c | 5 +
1 file changed, 5 insertions(+)
diff --git a/drivers/net/mlx5/mlx5_txq.c b/drivers/net/mlx5/mlx5_txq.c
index 4f17fb0..bae9f3d 100644
--- a/drivers/net/mlx5/mlx5_txq.c
+++ b/drivers/net/mlx5/mlx5_t
From: Adrien Mazarguil
The space necessary to store segmented packets cannot be known in advance
and must be verified for each of them.
Signed-off-by: Adrien Mazarguil
Signed-off-by: Nelio Laranjeiro
---
drivers/net/mlx5/mlx5_rxtx.c | 144 +++
1 file ch
From: Adrien Mazarguil
Compared to its previous incarnation, the software limit on the number of
mbuf segments is no more (previously MLX5_PMD_SGE_WR_N, set to 4 by
default) hence no need for linearization code and related buffers that
permanently consumed a non negligible amount of memory to han
From: Adrien Mazarguil
Since commit "mlx5: resurrect Tx gather support", older GCC versions (such
as 4.8.5) may complain about the following:
mlx5_rxtx.c: In function `mlx5_tx_burst':
mlx5_rxtx.c:705:25: error: `wqe' may be used uninitialized in this
function [-Werror=maybe-uninitialized]
From: Adrien Mazarguil
Toggling RX checksum offloads is already done at initialization time. This
code does not belong in rxq_rehash().
Signed-off-by: Adrien Mazarguil
Signed-off-by: Nelio Laranjeiro
---
drivers/net/mlx5/mlx5_rxq.c | 11 ---
1 file changed, 11 deletions(-)
diff --git
From: Adrien Mazarguil
The primary purpose of rxq_rehash() function is to stop and restart
reception on a queue after re-posting buffers. This may fail if the array
that temporarily stores existing buffers for reuse cannot be allocated.
Update rxq_rehash() to work on the target queue directly (n
1 - 100 of 145 matches
Mail list logo