On Wed, Jul 13, 2016 at 04:51:02AM -0700, Jeff Shaw wrote:
> The correct mailing list dev at dpdk.org, not dev at dpkg.org.
Nice catch!
Reviewed-by: Yuanhan Liu
--yliu
On Wed, Jul 13, 2016 at 08:54:08AM -0700, Rich Lane wrote:
> On Wednesday, July 13, 2016, Yuanhan Liu
> wrote:
>
> On Wed, Jul 13, 2016 at 10:34:07AM +0300, Ilya Maximets wrote:
> > This scenario fixed somehow, I agree. But this patch still needed to
> protect
> > vhost from untr
Hi Konstantin,
> > > > > It is quite safe to move the barrier before DEQUEUE because after
> > > > > the DEQUEUE there is nothing really that we would want to protect
> with a
> > > read barrier.
> > > >
> > > > I don't think so.
> > > > If you remove barrier after DEQUEUE(), that means on system
On 14.07.2016 04:42, Yuanhan Liu wrote:
> On Wed, Jul 13, 2016 at 08:54:08AM -0700, Rich Lane wrote:
>> On Wednesday, July 13, 2016, Yuanhan Liu
>> wrote:
>>
>> On Wed, Jul 13, 2016 at 10:34:07AM +0300, Ilya Maximets wrote:
>> > This scenario fixed somehow, I agree. But this patch still n
Hi Jan,
On Wednesday 13 July 2016 11:04 PM, Jan Viktorin wrote:
> On Wed, 13 Jul 2016 11:20:43 +0200
> Jan Viktorin wrote:
>
>> Hello Shreyansh,
>>
>> On Tue, 12 Jul 2016 11:31:10 +0530
>> Shreyansh Jain wrote:
>>
>>> Introduce a RTE_INIT macro used to mark an init function as a constructor.
>>
Flow bifurcation is a mechanism which depends the advanced
Ethernet device to split traffic between queues. It provides
the capability to let the kernel driver and DPDK driver co-exist
and take their advantages.
It is achieved by using SRIOV and NIC's advanced filtering. This
patch describes it and
In current implementation vhost will crash with segmentation fault
if malicious or buggy virtio application breaks addresses of descriptors.
Before commit 0823c1cb0a73 this crash was reproducible even with
normal DPDK application that tries to change number of virtqueues
dynamically inside VM.
Fi
2016-07-13 22:58, Damjan Marion:
> I have issues with linking application to 16.07-rc2.
>
> Looks like reason is constructor function in include file,
> so our unit test apps are failing to link as they are not linked with dpdk
> libs.
> (and they should not be as they are not calling any dpdk fu
> -Original Message-
> From: Shaw, Jeffrey B
> Sent: Wednesday, July 13, 2016 12:51 PM
> To: dev at dpdk.org; Butler, Siobhan A ;
> Mcnamara, John
> Subject: [PATCH] doc: fix mailing list address typo.
>
> The correct mailing list dev at dpdk.org, not dev at dpkg.org.
>
> Signed-off-by
Hi,All,
There are 4 build errors reported by our dpdk daily report,and can you take a
look on these issues.
Jan is working on ena issues,and I will help to test it.
OS, Kernel , GCC , ICC
Affected Targets
Build errors
comments
FC18_64,3.6.10-4,4.7.2,14.0.0
x86_64-ivshmem-linuxapp-gcc
CC
Hi Olivier,
Thank you for working on this.
We tried to enable checksum offload in OVS-DPDK and couldn't proceed due to the
performance impact.
I assume this patch will fix that issue by enabling checksum offloading with
vectorization ON at Rx side.
Few questions,
1) Is there any plan to extend t
> On 14 Jul 2016, at 10:20, Thomas Monjalon
> wrote:
>
> 2016-07-13 22:58, Damjan Marion:
>> I have issues with linking application to 16.07-rc2.
>>
>> Looks like reason is constructor function in include file,
>> so our unit test apps are failing to link as they are not linked with dpdk
>> l
This patch set contains
1)Error handling fixes in pdump library.
2)Support of server and client socket path command line options in pdump tool.
3)Default socket path name fixes in pdump library doc.
Reshma Pattan (3):
pdump: fix error handlings
app/pdump: add new command line options for sock
2016-07-14 09:36, Damjan Marion:
>
> > On 14 Jul 2016, at 10:20, Thomas Monjalon
> > wrote:
> >
> > 2016-07-13 22:58, Damjan Marion:
> >> I have issues with linking application to 16.07-rc2.
> >>
> >> Looks like reason is constructor function in include file,
> >> so our unit test apps are fai
Hi Thomas,
Would you check who can work on "dpdk16.07-rc2 compile error on FC18 & UB1204"?
For "dpdk16.07-rc2 compile error on suse11", I saw document shows we support
suse12+. Seems it is not issue for this release.
Agree?
Thanks & Regards,
Yu Liu
From: Gu, YongjieX
Sent: Thursday, July 14, 2
2016-07-14 11:34, Liu, Yu Y:
> Hi Thomas,
>
> Would you check who can work on "dpdk16.07-rc2 compile error on FC18 &
> UB1204"?
I think I've already fixed it:
http://dpdk.org/ml/archives/dev/2016-July/043918.html
> For "dpdk16.07-rc2 compile error on suse11", I saw document shows we sup
Hi Juhamatti,
>
> Hi Konstantin,
>
> > > > > > It is quite safe to move the barrier before DEQUEUE because
> > > > > > after the DEQUEUE there is nothing really that we would want
> > > > > > to protect
> > with a
> > > > read barrier.
> > > > >
> > > > > I don't think so.
> > > > > If you remo
> On 14 Jul 2016, at 13:30, Thomas Monjalon
> wrote:
>
> 2016-07-14 09:36, Damjan Marion:
linking fails with:
dpdk/include/rte_spinlock.h:103: undefined reference to
`rte_cpu_get_flag_enabled?
Is there any chance that this one is moved to some .c file, so it is
Someone to review this patch please?
It can be integrated in RC3 if we are sure it doesn't break anything.
2016-07-07 15:26, Yong Wang:
> The offset of the 2nd mmap when mapping the region after msix_bar
> needs to take region address into consideration. This is exposed
> when using vfio-pci to m
From: Damjan Marion
Having constructor function in the header gile is generaly
bad idea, as it will eventually be implanted to 3rd party
library.
In this case it is causing linking issues with 3rd party
libraries when application is not linked to dpdk, due to missing
symbol called by constructor
'noon,
In rte_eth_stats_get() after doing the driver callout to populate struct
rte_eth_stats, the rx_nombuf member is overwritten with
dev->data->rx_mbuf_alloc_failed even though some drivers will have
filled rx_nombuf with a value from elsewhere. This makes assignment of
rx_nombuf from withi
2016-07-01 15:10, Bernard Iremonger:
> add release note for live migration of a VM with SRIOV VF
[...]
> +* **Added support for live migration of a VM with SRIOV VF.**
> +
> + Live migration of a VM with Virtio and VF PMD's using the bonding PMD.
Is there some new code for this feature?
I am not
2016-07-14 14:29, Remy Horton:
> 'noon,
>
> In rte_eth_stats_get() after doing the driver callout to populate struct
> rte_eth_stats, the rx_nombuf member is overwritten with
> dev->data->rx_mbuf_alloc_failed even though some drivers will have
> filled rx_nombuf with a value from elsewhere. Thi
Hi Thomas,
> Subject: Re: [dpdk-dev] [PATCH] doc: update release notes
>
> 2016-07-01 15:10, Bernard Iremonger:
> > add release note for live migration of a VM with SRIOV VF
> [...]
> > +* **Added support for live migration of a VM with SRIOV VF.**
> > +
> > + Live migration of a VM with Virt
Hello.
How about deleting rx_nombuf from rte_eth_stats?
Do you think this counter is necessary? It just shows enormous numbers in case
of a lack of processing speed.
But we already have imissed counter which shows real number of packets, dropped
for the same reason.
> 14 2016 ?., ? 16:37,
> Someone to review this patch please?
> It can be integrated in RC3 if we are sure it doesn't break anything.
>
> 2016-07-07 15:26, Yong Wang:
> > The offset of the 2nd mmap when mapping the region after msix_bar
> > needs to take region address into consideration. This is exposed when
> > using
2016-07-14 13:45, Iremonger, Bernard:
> > 2016-07-01 15:10, Bernard Iremonger:
> > > add release note for live migration of a VM with SRIOV VF
> > [...]
> > > +* **Added support for live migration of a VM with SRIOV VF.**
> > > +
> > > + Live migration of a VM with Virtio and VF PMD's using the bo
On Thu, Jul 14, 2016 at 02:50:51PM +, Burakov, Anatoly wrote:
> > Someone to review this patch please?
> > It can be integrated in RC3 if we are sure it doesn't break anything.
> >
> > 2016-07-07 15:26, Yong Wang:
> > > The offset of the 2nd mmap when mapping the region after msix_bar
> > > ne
Hi Thomas,
> Subject: Re: [dpdk-dev] [PATCH] doc: update release notes
>
> 2016-07-14 13:45, Iremonger, Bernard:
> > > 2016-07-01 15:10, Bernard Iremonger:
> > > > add release note for live migration of a VM with SRIOV VF
> > > [...]
> > > > +* **Added support for live migration of a VM with SR
Hi,
On 01/07/16 11:19, Sergio Gonzalez Monroy wrote:
> On 01/07/2016 11:16, Elo, Matias (Nokia - FI/Espoo) wrote:
>>> -Original Message-
>>> From: Sergio Gonzalez Monroy [mailto:sergio.gonzalez.monroy at intel.com]
>>> Sent: Friday, July 01, 2016 1:05 PM
>>> To: Elo, Matias (Nokia - FI/Esp
From: Reshma Pattan
Zero packets can be returned by rte_eth_rx_burst() and
rte_distributor_returned_pkts() inside lcore_rx(), so
for zero packet scenario instead of proceeding to
next operations we should continue to the next iteration of the
loop to avoid unnecessary processing overhead which is
On 14/07/2016 14:51, Igor Ryzhov wrote:
[..]
> How about deleting rx_nombuf from rte_eth_stats? Do you think this
> counter is necessary? It just shows enormous numbers in case of a
> lack of processing speed. But we already have imissed counter which
> shows real number of packets, dropped for th
On Thu, 14 Jul 2016 10:57:55 +0530
Shreyansh jain wrote:
> Hi Jan,
>
> On Wednesday 13 July 2016 11:04 PM, Jan Viktorin wrote:
> > On Wed, 13 Jul 2016 11:20:43 +0200
> > Jan Viktorin wrote:
> >
> >> Hello Shreyansh,
> >>
> >> On Tue, 12 Jul 2016 11:31:10 +0530
> >> Shreyansh Jain wrote:
> >
This patchset takes the existing lpm large routing table information and
converts it from a header file included at compile time to a resource linked
in. This improves things in two ways:
1. Improves DPDK build time
2. Removes approx 1 million lines of code from our LOC counts, as the header
fil
Since we now have the ability to store resource data directly inside
the test binary, take the test_lpm_routes.h header file and make it a
test resource. Later commits will then switch the test C file over to use
this rather than including it directly.
Signed-off-by: Bruce Richardson
---
app/tes
The large routing table header file in only needed by the performance
tests and not by the functional tests to remove the include of
test_lpm_routes.h from test_lpm.c.
Fixes: 62dac65a3265 ("app/test: isolate lpm performance cases")
Signed-off-by: Bruce Richardson
---
app/test/test_lpm.c | 2 +-
The one route that was different to the others in the test_lpm_routes.h
file was the entry "{0, 8}" which was the only route without a "U" after
the IP part. Add in the extra "U" to that entry so that it can be used
as a check character when parsing routes manually.
Signed-off-by: Bruce Richardson
Change the lpm autotest to use the routes data from the resource data
stored in the binary rather than including it directly into the C file
as a C header. This speeds up compile and link time, without changing
the test results.
Signed-off-by: Bruce Richardson
---
app/test/test_lpm_perf.c | 75 +
Change the file extension of the test_lpm_routes file from .h to .dat.
This makes the lines-of-code counts for DPDK more realistic as they are not
affected by the huge counts from the lpm data.
Signed-off-by: Bruce Richardson
---
app/test/Makefile | 2 +-
app/te
On Tue, 12 Jul 2016 11:30:25 +0200
Maxime Coquelin wrote:
> The error is reported using test build script:
I recommend to note that the error is reported only by GCC 6+.
>
> $ scripts/test-build.sh x86_64-native-linuxapp-gcc
> ...
> drivers/net/virtio/virtio_user_ethdev.c: In function
> ?virt
On Tue, 12 Jul 2016 11:31:22 +0530
Shreyansh Jain wrote:
> Now that hotplug has been moved to eal, there is no reason to keep the device
> type in this layer.
>
> Signed-off-by: David Marchand
> Signed-off-by: Shreyansh Jain
> ---
> app/test/virtual_pmd.c| 2 +-
> drivers
On Tue, 12 Jul 2016 11:31:20 +0530
Shreyansh Jain wrote:
> Hotplug which deals with resources should come from the layer that already
> handles them, i.e. EAL.
>
> For both attach and detach operations, 'name' is used to select the bus
> that will handle the request.
>
> Signed-off-by: David Ma
On Tue, 12 Jul 2016 11:31:21 +0530
Shreyansh Jain wrote:
> Remove bus logic from ethdev hotplug by using eal for this.
>
> Current api is preserved:
> - the last port that has been created is tracked to return it to the
> application when attaching,
> - the internal device name is reused when
On Tue, 12 Jul 2016 11:31:17 +0530
Shreyansh Jain wrote:
> eal is a better place than crypto / ethdev for naming resources.
s/for naming/to name/
What is meant by "resources" here?
> Add a helper in eal and make use of it in crypto / ethdev.
>
> Signed-off-by: David Marchand
> Signed-off-by:
Our testing suggests minimal (in some cases zero) impact to core-bound
forwarding throughput as measured by testpmd. Throughput increase is
observed in l3fwd as now the vpmd can be used with hw_ip_checksum
enabled and without needing '--parse-ptype'.
The benefits to applications using this functio
From: Damjan Marion
The ptype is decoded from the rx descriptor and stored
in the packet type field in the mbuf using the same function
as the non-vector driver.
Signed-off-by: Damjan Marion
Signed-off-by: Jeff Shaw
---
drivers/net/i40e/i40e_rxtx.c | 566 +-
From: Damjan Marion
Decode the checksum flags from the rx descriptor, setting
the appropriate bit in the mbuf ol_flags field when the flag
indicates a bad checksum.
Signed-off-by: Damjan Marion
Signed-off-by: Jeff Shaw
---
drivers/net/i40e/i40e_rxtx_vec.c | 48 +++-
On Tue, 12 Jul 2016 11:31:18 +0530
Shreyansh Jain wrote:
> This helper updates a pci device object with latest information it can
> find.
> It will be used mainly for hotplug code.
>
> Signed-off-by: David Marchand
> Signed-off-by: Shreyansh Jain
> ---
> lib/librte_eal/bsdapp/eal/eal_pci.c
Disable fastpath interrupts and remove unneeded delay in
bnx2x_interrupt_action(). This patch fixes and prevents performance
degradation for BNX2X PMD.
Fixes: 540a2110 ("bnx2x: driver core")
Signed-off-by: Rasesh Mody
---
drivers/net/bnx2x/bnx2x.c|2 +-
drivers/net/bnx2x/bnx2x_ethde
This patch fixes following error:
EAL: Detected 36 lcore(s)
EAL: Probing VFIO support...
PMD: bnxt_rte_pmd_init() called for (null)
EAL: PCI device :08:00.0 on NUMA socket 0
EAL: probe driver: 14e4:16a1 rte_bnx2x_pmd
EAL: PCI device :08:00.1 on NUMA socket 0
EAL: pr
On Tue, 12 Jul 2016 11:31:11 +0530
Shreyansh Jain wrote:
> Preparing for getting rid of rte_cryptodev_driver, here are two wrappers
> that can be used by pci drivers that assume a 1 to 1 association between
> pci resource and upper interface.
>
> Signed-off-by: David Marchand
> Signed-off-by: S
On Tue, 12 Jul 2016 11:31:12 +0530
Shreyansh Jain wrote:
> Preparing for getting rid of eth_drv, here are two wrappers that can be
> used by pci drivers that assume a 1 to 1 association between pci resource and
> upper interface.
>
> Signed-off-by: David Marchand
> Signed-off-by: Shreyansh Jain
On Tue, 12 Jul 2016 11:31:16 +0530
Shreyansh Jain wrote:
> Now that virtio pci driver is initialized in a constructor, iopl() stuff
> happens early enough so that interrupt thread can be created right after
> plugin loading.
> This way, chelsio driver should be happy again [1].
>
> [1] http://dp
On Tue, 12 Jul 2016 11:31:15 +0530
Shreyansh Jain wrote:
> Now that all pdev are pci drivers, we don't need to register ethdev drivers
> through a dedicated channel.
>
> Signed-off-by: David Marchand
> Signed-off-by: Shreyansh Jain
> ---
Reviewed-by: Jan Viktorin
On Tue, 12 Jul 2016 11:31:14 +0530
Shreyansh Jain wrote:
> Now that all pdev are pci drivers, we don't need to register crypto drivers
> through a dedicated channel.
>
> Signed-off-by: David Marchand
> Signed-off-by: Shreyansh Jain
> ---
Reviewed-by: Jan Viktorin
On Tue, 12 Jul 2016 11:31:06 +0530
Shreyansh Jain wrote:
> These lists can be initialized once and for all at build time.
> With this, those lists are only manipulated in a common place
> (and we could even make them private).
>
> A nice side effect is that pci drivers can now register in constr
On Tue, 12 Jul 2016 11:31:08 +0530
Shreyansh Jain wrote:
> Pure coding style, but it might make it easier later if we want to move
> fields in rte_cryptodev_driver and eth_driver structures.
>
> Signed-off-by: David Marchand
> Signed-off-by: Shreyansh Jain
> ---
Reviewed-by: Jan Viktorin
On Tue, 12 Jul 2016 11:31:09 +0530
Shreyansh Jain wrote:
> rte_eal_dev_init is declared in both eal_private.h and rte_dev.h since its
> introduction.
> This function has been exported in ABI, so remove it from eal_private.h
>
> Fixes: e57f20e05177 ("eal: make vdev init path generic for both virt
On Tue, 12 Jul 2016 11:31:07 +0530
Shreyansh Jain wrote:
> This information is not used and just adds noise.
>
> Signed-off-by: David Marchand
> Signed-off-by: Shreyansh Jain
> ---
Reviewed-by: Jan Viktorin
On Wed, 13 Jul 2016 23:24:09 +0200
Thomas Monjalon wrote:
> The tests are registered with their command name by adding a structure
> to a list. The structure of each test was declared in each test file
> and passed to the register macro.
> This rework generate the structure inside the register ma
Hello,
Recently I tried to use bulk rx function to reduce CPU usage of
rte_eth_rx_burst.
However application performance with i40e_recv_pkts_bulk_alloc was
significantly worse than with i40e_recv_pkts. (3m less PPS, 0.5 IPC on
receiving core)
Quick investigation revealed two problems:
- First
Added prefetch of first packet payload cacheline in i40e_rx_scan_hw_ring
Added prefetch of second mbuf cacheline in i40e_rx_alloc_bufs
Signed-off-by: Vladyslav Buslov
---
drivers/net/i40e/i40e_rxtx.c | 7 +--
1 file changed, 5 insertions(+), 2 deletions(-)
diff --git a/drivers/net/i40e/i40e
Hello Damjan,
thank you for the patch. It makes sense to me.
Next time, please CC the appropriate maintainers.
(See the MAINTAINERS file in the root of the DPDK source tree.)
In the subject after "spinlock:" you should start with a lower case
letter, so "move constructor..."
On Thu, 14 Jul 2016
Dear Jan,
Thank you for your comments. A bit too much overhead to submit simple patch
so let?s forget about it. I will just add it as it is to our private
collection of patches.
If anybody wants to pick it from here, please do...
Thanks,
Damjan
> On 14 Jul 2016, at 20:03, Jan Viktorin wrote
Moving the rte_rtm_init() constructor routine out of the
header file and into a new rte_spinlock.c for all archs/platforms.
Having constructor routines in a header file is not a good
place to keep these types of functions.
The problem is with linking 3rd party libraries when
an application is not
Thanks Keith for continuing work.
2016-07-14 14:31, Keith Wiles:
> Moving the rte_rtm_init() constructor routine out of the
> header file and into a new rte_spinlock.c for all archs/platforms.
> Having constructor routines in a header file is not a good
> place to keep these types of functions.
>
> On 15 Jul 2016, at 00:06, Thomas Monjalon
> wrote:
>
> 2016-07-14 18:10, Damjan Marion:
>> Dear Jan,
>>
>> Thank you for your comments. A bit too much overhead to submit simple patch
>> so let?s forget about it. I will just add it as it is to our private
>> collection of patches.
>
> These
> On Jul 14, 2016, at 2:59 PM, Thomas Monjalon
> wrote:
>
> Thanks Keith for continuing work.
>
> 2016-07-14 14:31, Keith Wiles:
>> Moving the rte_rtm_init() constructor routine out of the
>> header file and into a new rte_spinlock.c for all archs/platforms.
>> Having constructor routines in a
The offset of the 2nd mmap() when mapping the region after msix_bar
needs to take region address into consideration as mmap() takes
address that is resource-relative instead of bar-relative. This is
exposed when binding vmxnet3 to vfio-pci.
Fixes: 90a1633b2347 ("eal/linux: allow to map BARs with
When binding a device to igb_uio with intr_conf.rxq set to 1, nb_efd
is 1 (for link event) but rte_intr_dp_is_en() will still return true.
rte_intr_dp_is_en() should also consider intr_handle type in addition
to nb_efd.
Signed-off-by: Yong Wang
---
lib/librte_eal/linuxapp/eal/eal_interrupts.c |
70 matches
Mail list logo
