Thursday, December 7, 2017 1:05 AM, Ferruh Yigit:
> On 12/4/2017 10:48 PM, Shahaf Shuler wrote:
> > Tuesday, December 5, 2017 12:31 AM, Ferruh Yigit:
> >> On 11/23/2017 4:08 AM, Shahaf Shuler wrote:
> >>> Since testpmd is now using the new Ethdev offloads API there is no
> >>> need for the txqflags
Thursday, December 7, 2017 12:58 AM, Ferruh Yigit:
> On 12/4/2017 10:39 PM, Shahaf Shuler wrote:
> > Tuesday, December 5, 2017 12:31 AM, Ferruh Yigit:
> >> On 11/23/2017 4:08 AM, Shahaf Shuler wrote:
> >>> Ethdev Rx/Tx offloads API has changed since:
> >>>
> >>> commit ce17eddefc20 ("ethdev: introd
> On Dec 7, 2017, at 5:01 AM, Stephen Hemminger
> wrote:
>
> Ok, went RFC hunting and the relevant one seems to be RFC 6864.
> It mandates unique id's for each datagram so TCP does send them.
>
>
Thanks for mention such the RFC, never heard about it.
Hi Maxime,
> -Original Message-
> From: Maxime Coquelin [mailto:maxime.coque...@redhat.com]
> Sent: Wednesday, December 6, 2017 4:34 PM
> To: Hu, Jiayu ; dev@dpdk.org
> Cc: y...@fridaylinux.org; Tan, Jianfeng
> Subject: Re: [dpdk-dev] [PATCH] vhost: support Generic Segmentation
> Offload
Hi all,
> -Original Message-
> From: Yang, Zhiyong
> Sent: Tuesday, December 5, 2017 2:59 PM
> To: Yang, Zhiyong ; dev@dpdk.org;
> y...@fridaylinux.org; maxime.coque...@redhat.com
> Cc: Wang, Wei W ; Tan, Jianfeng
>
> Subject: RE: [dpdk-dev] [PATCH 00/11] net/vhostpci: A new vhostpci PMD
This patch supports getting/setting input set info for
RSS/FDIR/FDIR flexible payload.
Also add some helper functions for input set configuration.
Signed-off-by: Beilei Xing
---
drivers/net/i40e/rte_pmd_i40e.c | 141 ++
drivers/net/i40e/rte_pmd_i40e.h
This patch adds command to configure input set for
RSS/flow director/flow director flexible payload.
Signed-off-by: Beilei Xing
---
app/test-pmd/cmdline.c | 237 +
1 file changed, 237 insertions(+)
diff --git a/app/test-pmd/cmdline.c b/app/test-pm
The patchset adds support RSS/FDIR/FDIR flexible payload input set
configuration for some pctype.
v4 changes:
- Change testpmd command token and print info.
v3 changes:
- Add support reset RSS/FDIR/FDIR flexible payload input set for some pctype.
v2 changes:
- Use 'static inline' to replace '
HI Fiona
> -Original Message-
> From: Trahe, Fiona [mailto:fiona.tr...@intel.com]
> Sent: 02 December 2017 00:42
> To: Verma, Shally ; dev@dpdk.org
> Cc: Athreya, Narayana Prasad ;
> Challa, Mahipal ; De Lara Guarch, Pablo
> ; Trahe, Fiona
> Subject: RE: [RFC v1] doc compression API for D
Squeeze repeated blank lines with below scripts:
for i in $(find . -name "*.[ch]"); do \
cat -s $i > /tmp/x && mv /tmp/x $i; done
Signed-off-by: Tiwei Bie
---
drivers/net/virtio/virtio_ethdev.c | 1 -
drivers/net/virtio/virtio_logs.h | 1 -
drivers/net/virtio/virtio_pci.c| 1 -
driver
RTE_VIRTIO_VPMD_RX_BURST and RTE_VIRTIO_VPMD_RX_REARM_THRESH
have been defined and used in virtio_rxtx_simple.h, but are
defined agained in virtio_rxtx_simple_*.c. It just happens to
work. So remove the redundant definitions from the *.c files.
Signed-off-by: Tiwei Bie
---
drivers/net/virtio/vir
VIRTIO_NET_CTRL_MAC_ADDR_SET is defined two times in
virtqueue.h, the second one is obviously not wanted.
Signed-off-by: Tiwei Bie
---
drivers/net/virtio/virtqueue.h | 2 --
1 file changed, 2 deletions(-)
diff --git a/drivers/net/virtio/virtqueue.h b/drivers/net/virtio/virtqueue.h
index ab466c2
Tiwei Bie (5):
net/virtio: fix vector Rx break caused by rxq flushing
net/virtio: fix typo in LRO support
net/virtio: remove a redundant macro definition for ctrl vq
net/virtio: remove redundant macro definitions for vector Rx
net/virtio: squeeze repeated blank lines
drivers/net/virtio/
Fixes: 86d59b21468a ("net/virtio: support LRO")
Fixes: ec9f3d122a58 ("net/virtio: revert not claiming LRO support")
Cc: sta...@dpdk.org
Signed-off-by: Tiwei Bie
---
drivers/net/virtio/virtio_ethdev.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/net/virtio/virtio_et
The vector Rx will be broken if backend has consumed all
the descs in the avail ring before the device is started.
Because in current implementation, vector Rx will return
immediately without refilling the avail ring if the used
ring is empty. So we have to refill the avail ring after
flushing the
> No, it is in vdev_scan.
> Look carefully the patch, especially after @@.
I overlooked. Sorry about that.
I don't have further comments. Please send v2 according to your own comments
and document update.
Thanks,
Jianfeng
On 11/30/2017 4:04 PM, Declan Doherty wrote:
> change the xmit_hash functions to handle bursts of packet instead of
> packet at a time. Updating effect tx_burst functions.
>
> Signed-off-by: Declan Doherty
> Signed-off-by: Keith Wiles
Hi Declan,
Your other patch [1] conflicts with this one, ca
On 11/30/2017 3:25 PM, Declan Doherty wrote:
> In message (http://dpdk.org/ml/archives/dev/2017-November/081557.html)
> it was noted that under congestion that the LACPDU's are dropped under
> load.
>
> This patch changes the drop logic to re-enqueue the LACPDU to the slaves
> control message queu
On 12/5/2017 1:39 AM, Tomasz Duszynski wrote:
> This patchset brings following changes:
>
> o Sync with MUSDK-17.10. Latest version of the library comes with many
> improvements and fixes thus switching to it is beneficial.
>
> o A few code and documentation updates.
>
> Changes since v2:
> o
On Thu, 7 Dec 2017 00:19:46 +
"Ananyev, Konstantin" wrote:
> > -Original Message-
> > From: dev [mailto:dev-boun...@dpdk.org] On Behalf Of Stephen Hemminger
> > Sent: Wednesday, December 6, 2017 11:16 PM
> > To: Ilya Matveychikov
> > Cc: dev@dpdk.org; Hu, Jiayu
> > Subject: Re: [dpd
On 11/24/2017 12:35 PM, Rasesh Mody wrote:
> From: Shahed Shaikh
>
> Replace rx_vxlan_port command with rx_tunnel_udp_port to support both VXLAN
> and GENEVE udp ports.
Also updates tunnel_filter command to accept "geneve" argument, can you please
separate to another patch.
And to prevent these
> -Original Message-
> From: dev [mailto:dev-boun...@dpdk.org] On Behalf Of Stephen Hemminger
> Sent: Wednesday, December 6, 2017 11:16 PM
> To: Ilya Matveychikov
> Cc: dev@dpdk.org; Hu, Jiayu
> Subject: Re: [dpdk-dev] A question about GRO neighbor packet matching
>
> On Wed, 6 Dec 201
Sorry, sent this to the wrong maintainer initially. I am cc'ing the
Broadcom bnxt maintainers in hopes that they might comment about the 54616
behavior on reset. I couldn't find any programmer guide for this chip
online but PHYs are typically well behaved. This patch does work on our
test system
From: "Charles (Chas) Williams"
If we find a Broadcom 54616, handle as a e1000_phy_none assuming that
the NIC reset has initialized the PHY to a sane state.
Signed-off-by: Chas Williams
---
drivers/net/e1000/base/e1000_82575.c | 5 +
drivers/net/e1000/base/e1000_defines.h | 1 +
2 files
On Wed, 6 Dec 2017 22:38:12 +0400
Ilya Matveychikov wrote:
> > On Dec 6, 2017, at 10:12 PM, Stephen Hemminger
> > wrote:
> >
> > On Wed, 6 Dec 2017 18:02:21 +0400
> > Ilya Matveychikov wrote:
> >
> >> Hello all,
> >>
> >>
> >> My question is about neighbor packet matching algorithm for T
On 12/4/2017 10:48 PM, Shahaf Shuler wrote:
> Tuesday, December 5, 2017 12:31 AM, Ferruh Yigit:
>> On 11/23/2017 4:08 AM, Shahaf Shuler wrote:
>>> Since testpmd is now using the new Ethdev offloads API there is no
>>> need for the txqflags configuration.
>>
>> txqflag command-line option (--txqflag
On 12/4/2017 10:39 PM, Shahaf Shuler wrote:
> Tuesday, December 5, 2017 12:31 AM, Ferruh Yigit:
>> On 11/23/2017 4:08 AM, Shahaf Shuler wrote:
>>> Ethdev Rx/Tx offloads API has changed since:
>>>
>>> commit ce17eddefc20 ("ethdev: introduce Rx queue offloads API") commit
>>> cba7f53b717d ("ethdev: i
Hi,
Thank you so much for your information.
Vivian
seat: 19009
phone: 13738006921
-Original Message-
From: Tan, Jianfeng [mailto:jianfeng@intel.com]
Sent: Wednesday, December 06, 2017 10:56 AM
To: Yuanhan Liu ; Wang, Wei 5. (NSB - CN/Hangzhou)
Cc: dev@dpdk.org
Subject: RE
Thanks, Jerin. I'll make the changes suggested below and the changes suggested
for the preceding patch in the next version of the series that is posted.
Regards,
Gabriel
> -Original Message-
> From: Jerin Jacob [mailto:jerin.ja...@caviumnetworks.com]
> Sent: Wednesday, December 6, 2017
> On Dec 6, 2017, at 10:12 PM, Stephen Hemminger
> wrote:
>
> On Wed, 6 Dec 2017 18:02:21 +0400
> Ilya Matveychikov wrote:
>
>> Hello all,
>>
>>
>> My question is about neighbor packet matching algorithm for TCP. Is it
>> correct to expect that IP packets should have continuous ID enumerat
On 12/06/2017 03:06 PM, Yuanhan Liu wrote:
On Wed, Dec 06, 2017 at 12:23:19PM +0100, Maxime Coquelin wrote:
Hi,
On 12/05/2017 07:58 AM, Wang, Wei 5. (NSB - CN/Hangzhou) wrote:
Hi, all
In DPDK document, it it described that Virtio in containers Cannot work when
there are more than VHOST_M
On Wed, 6 Dec 2017 18:02:21 +0400
Ilya Matveychikov wrote:
> Hello all,
>
>
> My question is about neighbor packet matching algorithm for TCP. Is it
> correct to expect that IP packets should have continuous ID enumeration
> (i.e. iph-next.id = iph-prev.id + 1)?
No.
> ~~~
> lib/librte_gro/gr
On 12/06/2017 04:28 PM, Victor Kaplansky wrote:
v2:
o Fixed checkpatch complains
o Added Signed-off-by
o Refined placement of guard to exclude IOMMU messages
o TODO: performance degradation measurement
See https://bugzilla.redhat.com/show_bug.cgi?id=1450680
When performing l
This counter saved the descriptor elements which are waiting to be
completed and was used to know if completion function should be
called.
This completion check can be done by other elements management
variables and we can prevent this counter management.
Remove this counter and replace the compl
Using power of 2 descriptors number makes the ring management easier
and allows to use mask operation instead of wraparound conditions.
Adjust Tx descriptor number to be power of 2 and change calculation to
use mask accordingly.
Signed-off-by: Matan Azrad
---
drivers/net/mlx4/mlx4_rxtx.c | 28 +
The previuse code took a send queue entry size for stamping from the
send queue entry pointed by completion queue entry; This 2 reads were
done per packet in completion stage.
The completion burst packets number is managed by fixed size stored in
Tx queue, so we can infer that each valid completio
The Tx queue send ring was managed by Tx block head,tail,count and mask
management variables which were used for managing the send queue remain
space and next places of empty or completed work queue entries.
This method suffered from an actual addresses recalculation per packet,
an unnecessary Tx
There is no need to check Tx queue wraparound for segments which are
not at the beginning of a Tx block. Especially relevant in a single
segment case.
Remove unnecessary aforementioned checks from Tx path.
Signed-off-by: Matan Azrad
Acked-by: Adrien Mazarguil
---
drivers/net/mlx4/mlx4_rxtx.c |
This series improves mlx4 Tx performance and fix and clean some Tx code.
1. 10% MPPS improvement for 1 queue, 1 core, 64B packets, txonly mode.
2. 20% MPPS improvement for 1 queue, 1 core, 32B*4(segs) packets, txonly mode.
V2:
Add missed function descriptions.
Accurate descriptions.
Change Tx des
At error time, the first 4 bytes of each WQE Tx block still have not
writen, so no need to stamp them because they are already stamped.
Signed-off-by: Matan Azrad
Acked-by: Adrien Mazarguil
---
drivers/net/mlx4/mlx4_rxtx.c | 22 +-
1 file changed, 1 insertion(+), 21 deletion
mlx4 Tx block can handle up to 4 data segments or control segment + up
to 3 data segments. The first data segment in each not first Tx block
must validate Tx queue wraparound and must use IO memory barrier before
writing the byte count.
The previous multi-segment code used "for" loop to iterate ov
When invalid lkey is sent to HW, HW sends an error notification in
completion function.
The previous code wouldn't crash but doesn't add any application report
in case of completion error, so application cannot know that packet
actually was dropped in case of invalid lkey.
Return back the lkey va
Hi Adrien
> -Original Message-
> From: Adrien Mazarguil [mailto:adrien.mazarg...@6wind.com]
> Sent: Wednesday, December 6, 2017 6:23 PM
> To: Matan Azrad
> Cc: dev@dpdk.org
> Subject: Re: [PATCH v2 7/8] net/mlx4: align Tx descriptors number
>
> On Wed, Dec 06, 2017 at 02:48:12PM +, M
> -Original Message-
> From: Jerin Jacob [mailto:jerin.ja...@caviumnetworks.com]
> Sent: Wednesday, December 6, 2017 9:18 AM
> To: Carrillo, Erik G
> Cc: pbhagavat...@caviumnetworks.com; dev@dpdk.org;
> nipun.gu...@nxp.com; hemant.agra...@nxp.com
> Subject: Re: [RFC PATCH v5 1/5] eventti
On Wed, Dec 06, 2017 at 02:48:13PM +, Matan Azrad wrote:
> This counter saved the descriptor elements which are waiting to be
> completted and was used to know if completion function should be
Looks like you forgot one minor change before adding my ack:
completted => completed
> called.
>
>
On Wed, Dec 06, 2017 at 02:48:12PM +, Matan Azrad wrote:
> Using power of 2 descriptors number makes the ring management easier
> and allows to use mask operation instead of wraparound conditions.
>
> Adjust Tx descriptor number to be power of 2 and change calculation to
> use mask accordingly
On Wed, Dec 06, 2017 at 02:48:10PM +, Matan Azrad wrote:
> The Tx queue send ring was managed by Tx block head,tail,count and mask
> management variables which were used for managing the send queue remain
> space and next places of empty or completed work queue entries.
>
> This method suffere
On Wed, Dec 06, 2017 at 02:48:09PM +, Matan Azrad wrote:
> mlx4 Tx block can handle up to 4 data segments or control segment + up
> to 3 data segments. The first data segment in each not first Tx block
> must validate Tx queue wraparound and must use IO memory barrier before
> writing the byte
On Tue, Dec 05, 2017 at 06:22:05PM +0100, Adrien Mazarguil wrote:
> > > > > Just for information, this "port=x" argument in mlx4 is consistent
> > > > > with the
> > > > > value found in /sys/class/net/ethX/dev_port under Linux. If we choose
> > > > > to use
> > > > > a port index (instead of a M
-Original Message-
> Date: Fri, 1 Dec 2017 14:00:56 -0600
> From: Erik Gabriel Carrillo
> To: pbhagavat...@caviumnetworks.com
> CC: dev@dpdk.org, jerin.ja...@caviumnetworks.com, nipun.gu...@nxp.com,
> hemant.agra...@nxp.com
> Subject: [RFC PATCH v5 3/5] eventtimer: add config variable for
-Original Message-
> Date: Fri, 1 Dec 2017 14:00:55 -0600
> From: Erik Gabriel Carrillo
> To: pbhagavat...@caviumnetworks.com
> CC: dev@dpdk.org, jerin.ja...@caviumnetworks.com, nipun.gu...@nxp.com,
> hemant.agra...@nxp.com
> Subject: [RFC PATCH v5 2/5] eventtimer: add common code
> X-Mai
v2:
o Fixed checkpatch complains
o Added Signed-off-by
o Refined placement of guard to exclude IOMMU messages
o TODO: performance degradation measurement
See https://bugzilla.redhat.com/show_bug.cgi?id=1450680
When performing live migration or memory hot-plugging,
the changes to the
-Original Message-
> Date: Fri, 1 Dec 2017 14:00:54 -0600
> From: Erik Gabriel Carrillo
> To: pbhagavat...@caviumnetworks.com
> CC: dev@dpdk.org, jerin.ja...@caviumnetworks.com, nipun.gu...@nxp.com,
> hemant.agra...@nxp.com
> Subject: [RFC PATCH v5 1/5] eventtimer: introduce event timer a
This counter saved the descriptor elements which are waiting to be
completted and was used to know if completion function should be
called.
This completion check can be done by other elements management
variables and we can prevent this counter management.
Remove this counter and replace the comp
The Tx queue send ring was managed by Tx block head,tail,count and mask
management variables which were used for managing the send queue remain
space and next places of empty or completed work queue entries.
This method suffered from an actual addresses recalculation per packet,
an unnecessary Tx
The previuse code took a send queue entry size for stamping from the
send queue entry pointed by completion queue entry; This 2 reads were
done per packet in completion stage.
The completion burst packets number is managed by fixed size stored in
Tx queue, so we can infer that each valid completio
Using power of 2 descriptors number makes the ring management easier
and allows to use mask operation instead of wraparound conditions.
Adjust Tx descriptor number to be power of 2 and change calculation to
use mask accordingly.
Signed-off-by: Matan Azrad
---
drivers/net/mlx4/mlx4_rxtx.c | 28 +
mlx4 Tx block can handle up to 4 data segments or control segment + up
to 3 data segments. The first data segment in each not first Tx block
must validate Tx queue wraparound and must use IO memory barrier before
writing the byte count.
The previous multi-segment code used "for" loop to iterate ov
At error time, the first 4 bytes of each WQE Tx block still have not
writen, so no need to stamp them because they are already stamped.
Signed-off-by: Matan Azrad
Acked-by: Adrien Mazarguil
---
drivers/net/mlx4/mlx4_rxtx.c | 22 +-
1 file changed, 1 insertion(+), 21 deletion
There is no need to check Tx queue wraparound for segments which are
not at the beginning of a Tx block. Especially relevant in a single
segment case.
Remove unnecessary aforementioned checks from Tx path.
Signed-off-by: Matan Azrad
Acked-by: Adrien Mazarguil
---
drivers/net/mlx4/mlx4_rxtx.c |
This series improves mlx4 Tx performance and fix and clean some Tx code.
1. 10% MPPS improvement for 1 queue, 1 core, 64B packets, txonly mode.
2. 20% MPPS improvement for 1 queue, 1 core, 32B*4(segs) packets, txonly mode.
V2:
Add missed function descriptions.
Accurate descriptions.
Change Tx des
When invalid lkey is sent to HW, HW sends an error notification in
completion function.
The previous code wouldn't crash but doesn't add any application report
in case of completion error, so application cannot know that packet
actually was dropped in case of invalid lkey.
Return back the lkey va
On Wed, Dec 06, 2017 at 07:23:11PM +0800, Tiwei Bie wrote:
> On Mon, Dec 04, 2017 at 06:02:08AM -0800, Xiao Wang wrote:
> [...]
> > diff --git a/drivers/net/virtio/virtio_rxtx.c
> > b/drivers/net/virtio/virtio_rxtx.c
> > index 6a24fde..7313bdd 100644
> > --- a/drivers/net/virtio/virtio_rxtx.c
> >
On Wed, Dec 06, 2017 at 03:55:49PM +0200, Victor Kaplansky wrote:
> When performing live migration or memory hot-plugging,
> the changes to the device and vrings made by message handler
> done independently from vring usage by PMD threads.
>
> This causes for example segfauls during live-migration
On Wed, Dec 06, 2017 at 12:23:19PM +0100, Maxime Coquelin wrote:
> Hi,
>
> On 12/05/2017 07:58 AM, Wang, Wei 5. (NSB - CN/Hangzhou) wrote:
> >
> >
> >Hi, all
> >
> >In DPDK document, it it described that Virtio in containers Cannot work when
> >there are more than VHOST_MEMORY_MAX_NREGIONS(8) hug
Hello all,
My question is about neighbor packet matching algorithm for TCP. Is it
correct to expect that IP packets should have continuous ID enumeration
(i.e. iph-next.id = iph-prev.id + 1)?
~~~
lib/librte_gro/gro_tcp4.c:check_seq_option()
...
/* check if the two packets are nei
When performing live migration or memory hot-plugging,
the changes to the device and vrings made by message handler
done independently from vring usage by PMD threads.
This causes for example segfauls during live-migration
with MQ enable, but in general virtually any request
sent by qemu changing
Hi Jasvinder,
> > > -Original Message-
> > > From: Singh, Jasvinder
> > > Sent: Thursday, November 23, 2017 11:32 AM
> > > To: dev@dpdk.org
> > > Cc: Iremonger, Bernard
> > > Subject: [PATCH 2/3] lib/librte_flow_classy: add run api for flow
> > > classification
> > >
> > > This patch ex
Hi Jasvinder,
> > > -Original Message-
> > > From: Singh, Jasvinder
> > > Sent: Thursday, November 23, 2017 11:32 AM
> > > To: dev@dpdk.org
> > > Cc: Iremonger, Bernard
> > > Subject: [PATCH 1/3] lib/librte_flow_classify: remove table id
> > > parameter from apis
> > >
> > > This patch re
If the allocation is from a single memzone, optimize
the phy-virt address conversions.
Signed-off-by: Hemant Agrawal
---
drivers/mempool/dpaa/dpaa_mempool.c | 18 +++---
drivers/mempool/dpaa/dpaa_mempool.h | 9 +
2 files changed, 24 insertions(+), 3 deletions(-)
diff --git
This is required for the optimizations w.r.t hw mempools.
They will use different kind of optimizations if the buffers
are from single contiguous memzone.
Signed-off-by: Hemant Agrawal
---
lib/librte_mempool/rte_mempool.c | 7 +--
lib/librte_mempool/rte_mempool.h | 5 +
2 files changed,
On Wed, Dec 6, 2017 at 10:20 AM, Maxime Coquelin
wrote:
> Hi,
>
> This third revision reworks the VQs destruction loop to fixes the
> of-by-one error reported by Laszlo.
>
> Having QEMU started with mq=on but guest driver not negotiating
> VIRTIO_NET_F_MQ feature ends up in the vhost device to nev
Just a reminder to all interested that there is a technical board
meeting scheduled for today, 6th December, at 3pm UTC. Agenda can be
found here: https://annuel.framapad.org/p/r.0c3cc4d1e011214183872a98f6b5c7db
and since it is rather light today, one hopes it should be a fairly
quick meeting.
All
On Wed, Dec 06, 2017 at 11:43:25AM +, Matan Azrad wrote:
> Hi Adrien
>
> > -Original Message-
> > From: Adrien Mazarguil [mailto:adrien.mazarg...@6wind.com]
> > Sent: Wednesday, December 6, 2017 12:59 PM
> > To: Matan Azrad
> > Cc: dev@dpdk.org
> > Subject: Re: [PATCH 5/8] net/mlx4: m
Hi Jasvinder
> -Original Message-
> From: Singh, Jasvinder
> Sent: Thursday, November 23, 2017 11:32 AM
> To: dev@dpdk.org
> Cc: Iremonger, Bernard
> Subject: [PATCH 3/3] doc: update documentation for flow classify lib
>
> Updates the documentation for flow classification library and sam
On Wed, Dec 06, 2017 at 11:29:38AM +, Matan Azrad wrote:
> Hi Adrien
>
> > -Original Message-
> > From: Adrien Mazarguil [mailto:adrien.mazarg...@6wind.com]
> > Sent: Wednesday, December 6, 2017 12:59 PM
> > To: Matan Azrad
> > Cc: dev@dpdk.org
> > Subject: Re: [PATCH 4/8] net/mlx4: o
Hi Adrien
> -Original Message-
> From: Adrien Mazarguil [mailto:adrien.mazarg...@6wind.com]
> Sent: Wednesday, December 6, 2017 12:59 PM
> To: Matan Azrad
> Cc: dev@dpdk.org
> Subject: Re: [PATCH 7/8] net/mlx4: align Tx descriptors number
>
> On Tue, Nov 28, 2017 at 12:19:29PM +, Mat
Hi Adrien
> -Original Message-
> From: Adrien Mazarguil [mailto:adrien.mazarg...@6wind.com]
> Sent: Wednesday, December 6, 2017 12:59 PM
> To: Matan Azrad
> Cc: dev@dpdk.org
> Subject: Re: [PATCH 5/8] net/mlx4: merge Tx queue rings management
>
> On Tue, Nov 28, 2017 at 12:19:27PM +,
On 12/06/17 10:20, Maxime Coquelin wrote:
> Hi,
>
> This third revision reworks the VQs destruction loop to fixes the
> of-by-one error reported by Laszlo.
>
> Having QEMU started with mq=on but guest driver not negotiating
> VIRTIO_NET_F_MQ feature ends up in the vhost device to never
> start. I
Hi Adrien
> -Original Message-
> From: Adrien Mazarguil [mailto:adrien.mazarg...@6wind.com]
> Sent: Wednesday, December 6, 2017 12:59 PM
> To: Matan Azrad
> Cc: dev@dpdk.org
> Subject: Re: [PATCH 4/8] net/mlx4: optimize Tx multi-segment case
>
> On Tue, Nov 28, 2017 at 12:19:26PM +,
On Mon, Dec 04, 2017 at 06:02:08AM -0800, Xiao Wang wrote:
[...]
> diff --git a/drivers/net/virtio/virtio_rxtx.c
> b/drivers/net/virtio/virtio_rxtx.c
> index 6a24fde..7313bdd 100644
> --- a/drivers/net/virtio/virtio_rxtx.c
> +++ b/drivers/net/virtio/virtio_rxtx.c
> @@ -1100,3 +1100,84 @@
>
>
Hi,
On 12/05/2017 07:58 AM, Wang, Wei 5. (NSB - CN/Hangzhou) wrote:
Hi, all
In DPDK document, it it described that Virtio in containers Cannot work when
there are more than VHOST_MEMORY_MAX_NREGIONS(8) hugepages. In another word, do
not use 2MB hugepage so far. Do you know the reason of thi
Hi Beilei,
Thanks for looking into this.
I'm testing it with TRex (tx->rx) and not (rx->tx) in a bit different
configuration and there are packets with ~40usec latency (Tx/Rx) without my
patch
So I assume the latency comes from a different place due to this change (INTR)
TRex command:
$sudo ./
Hi Akhil,
On 12/04/2017 01:19 PM, Akhil Goyal wrote:
Hi Anoob,
On 11/29/2017 9:51 AM, Anoob Joseph wrote:
Hi Akhil,
On 24-11-2017 16:19, Akhil Goyal wrote:
Hi Anoob,
On 11/24/2017 3:28 PM, Anoob wrote:
static inline void
route4_pkts(struct rt_ctx *rt_ctx, struct rte_mbuf *pkts[],
uint
Hi,
I tested Rx latency with testpmd and ixia, found that the interval
configuration works on X710.
With the default configuration, the latency is about 32us.
When RTE_LIBRTE_I40E_ITR_INTERVAL=0, the max latency is < 8us.
When RTE_LIBRTE_I40E_ITR_INTERVAL=8160, the max latency is about 8ms.
My
On Tue, Nov 28, 2017 at 12:19:29PM +, Matan Azrad wrote:
> Using power of 2 descriptors number makes the ring management easier
> and allows to use mask operation instead of wraparound conditions.
>
> Adjust Tx descriptor number to be power of 2 and change calculation to
> use mask accordingly
On Tue, Nov 28, 2017 at 12:19:30PM +, Matan Azrad wrote:
> This counter saved the descriptor elements which are waiting to be
> completted and was used to know if completion function should be
completted => completed
> called.
>
> This completion check can be done by other elements managemen
On Tue, Nov 28, 2017 at 12:19:28PM +, Matan Azrad wrote:
> The previuse code took a send queue entry size for stamping from the
> send queue entry pointed by completion queue entry; This 2 reads were
> done per packet in completion stage.
>
> The completion burst packets number is managed by f
On Tue, Nov 28, 2017 at 12:19:26PM +, Matan Azrad wrote:
> mlx4 Tx block can handle up to 4 data segments or control segment + up
> to 3 data segments. The first data segment in each not first Tx block
> must validate Tx queue wraparound and must use IO memory barrier before
> writing the byte
On Tue, Nov 28, 2017 at 12:19:27PM +, Matan Azrad wrote:
> The Tx queue send ring was managed by Tx block head,tail,count and mask
> management variables which were used for managing the send queue remain
> space and next places of empty or completted work queue entries.
completted => complete
On Tue, Nov 28, 2017 at 12:19:25PM +, Matan Azrad wrote:
> At error time, the first 4 bytes of each WQE Tx block still have not
> writen, so no need to stamp them because they are already stamped.
>
> Signed-off-by: Matan Azrad
Acked-by: Adrien Mazarguil
--
Adrien Mazarguil
6WIND
On Tue, Nov 28, 2017 at 12:19:24PM +, Matan Azrad wrote:
> There is no need to check Tx queue wraparound for segments which are
> not at the beginning of a Tx block. Especially relevant in a single
> segment case.
>
> Remove unnecessary aforementioned checks from Tx path.
>
> Signed-off-by: M
On Tue, Nov 28, 2017 at 12:19:23PM +, Matan Azrad wrote:
> When invalid lkey is sent to HW, HW sends an error notification in
> completion function.
>
> The previous code wouldn't crash but doesn't add any application report
> in case of completion error, so application cannot know that packet
Hi,
On 12/6/2017 7:30 AM, Anoob wrote:
Hi Akhil, Radu,
Please see inline.
Thanks,
Anoob
On 11/24/2017 05:33 PM, Akhil Goyal wrote:
On 11/24/2017 5:29 PM, Radu Nicolau wrote:
On 11/24/2017 11:34 AM, Akhil Goyal wrote:
Hi Radu,
On 11/24/2017 4:47 PM, Radu Nicolau wrote:
On 11/24/2017
On 11/21/2017 7:12 AM, Ferruh Yigit wrote:
To track modification:
380a7aab1ae2 ("mbuf: rename deprecated VLAN flags")
Proposed deadline for PMDs is v18.02
Signed-off-by: Ferruh Yigit
---
doc/guides/nics/todo.rst | 11 +++
1 file changed, 11 insertions(+)
diff --git a/doc/guides/nics/
06/12/2017 01:40, Ananyev, Konstantin:
> From: Thomas Monjalon [mailto:tho...@monjalon.net]
> > 05/12/2017 16:13, Ananyev, Konstantin:
> > > > Keep in mind that the owner can be an application thread.
> > > > If you prefer using a single function pointer (may help for
> > > > atomic implementation
QEMU sends VHOST_USER_SET_VRING_CALL requests for all queues
declared in QEMU command line before the guest is started.
It has the effect in DPDK vhost-user backend to allocate vrings
for all queues declared by QEMU.
If the first driver being used does not support multiqueue,
the device never chan
06/12/2017 03:52, Tan, Jianfeng:
> On 12/1/2017 8:36 AM, Thomas Monjalon wrote:
> > +int
> > +rte_vdev_add_custom_scan(rte_vdev_scan_callback callback, void *user_arg)
> > +{
> > + struct vdev_custom_scan *custom_scan;
> > +
> > + rte_spinlock_lock(&vdev_custom_scan_lock);
> > +
> > + /* chec
This patch extracts needed code for vhost_user.c to be able
to clean and free virtqueues unitary.
Signed-off-by: Maxime Coquelin
---
lib/librte_vhost/vhost.c | 22 --
lib/librte_vhost/vhost.h | 3 +++
2 files changed, 15 insertions(+), 10 deletions(-)
diff --git a/lib/librt
Hi,
This third revision reworks the VQs destruction loop to fixes the
of-by-one error reported by Laszlo.
Having QEMU started with mq=on but guest driver not negotiating
VIRTIO_NET_F_MQ feature ends up in the vhost device to never
start. Indeed, more queues are created in the vhost backend than
c
1 - 100 of 104 matches
Mail list logo
