Hi,
Recently I came across a 82546EB Dual-port Gigabit Ethernet (Copper) NIC
and tried to include it in our current DPDK setup. However, the card
doesn't seem to be supported (I was unable to bind the igb_uio driver).
There was a post in mailing list earlier this year stating that this
particular
I do not think a newer version will effect the performance, but you can try it.
git clone git://github.com/Pktgen/Pktgen-DPDK
This one is 2.2.5 and DPDK 1.5.0
Keith Wiles, Principal Technologist for Networking member of the CTO office,
Wind River
mobile 940.213.5533
[Powering 30 Years of Innov
Give this a try, if that does not work then something else is going on here. I
am trying to make sure we do not cross the QPI for any reason putting the RX/TX
queues related to a port on the same core.
sudo ./app/build/pktgen -c 3ff -n 3 $BLACK-LIST -- -p 0xf0 -P -m
"[2:4].0, [6:8].1, [3:5].2, [
Hi,
I am encountering a strange behavior of rte_ring_sc_dequeue, though I'm not
yet sure what causes this.
I have a code:
rc = rte_ring_sc_dequeue(fwdp->rxtx_rings->xmit_ring, &rpackets);
At first dequeue, rpackets gets a correct address of an rte_mbuf, however
at the second dequeue it returns
Gratitude to reply.
FCP and Full-offload iSCSI will be nice, for example Qlogic 16Gbps FCP.
But IO scenario is very different from IP scenario as my understanding.
March
BTW: Do you have people at Bay Area?
PS:
Already finis
Normally when I see this problem it means the the lcores are not mapped
correctly. What can happen is you have a Rx and a TX on the same physical core
or two RX/TX on the same physical core.
Make sure you have a Rx or Tx running on a single core look at the
cpu_layout.py output and verify the c
BTW, the configuration looks fine, but you need to make sure the lcores are not
split between two different CPU sockets. You can use the
dpdk/tools/cpu_layout.py to do dump out the system configuration.
Keith Wiles, Principal Technologist for Networking member of the CTO office,
Wind River
mob
Sorry I miss-typed the speed of my machine it is 2.4Ghz not 3.4GHz, but that
should not change the problem here.
I am not sure how to determine your machine has a problem other then starting
up one port at a time and see if the rate drops when you start up the forth
port.
Keith Wiles, Principa
On Tue, Nov 19, 2013 at 4:18 PM, Wiles, Roger Keith
wrote:
> Give this a try, if that does not work then something else is going on here.
> I am trying to make sure we do not cross the QPI for any reason putting the
> RX/TX queues related to a port on the same core.
>
> sudo ./app/build/pktgen -c
How do you have Pktgen configured in this case?
On my westmere dual socket 3.4Ghz machine I can send 20G on a single NIC 82599x
two ports. My machine has a PCIe bug that does not allow me to send on more
then 3 ports at wire rate. I get close to 40G 64 byte packets, but the forth
port does is a
We should be able to enable RSS with one Rx queue.
RSS hash can be useful independently of the number of queues.
Applications can use RSS hash to identify different flows.
Signed-off-by: Maxime Leroy
---
lib/librte_pmd_e1000/igb_rxtx.c |7 ++-
lib/librte_pmd_ixgbe/ixgbe_rxtx.c |7 +
As explained in rte_ethdev.h, ETH_MQ_RX_NONE allows to not choose RSS, DCB
or VMDQ mode to select the rx queues.
But the igb/ixgbe code always select RSS mode with ETH_MQ_RX_NONE. This patch
fixes this incoherence between the api and the source code.
Signed-off-by: Maxime Leroy
---
lib/librte_p
On Tue, Nov 19, 2013 at 12:24 PM, Wiles, Roger Keith
wrote:
> Normally when I see this problem it means the the lcores are not mapped
> correctly. What can happen is you have a Rx and a TX on the same physical
> core or two RX/TX on the same physical core.
>
> Make sure you have a Rx or Tx running
On Tue, Nov 19, 2013 at 11:54 AM, Wiles, Roger Keith
wrote:
>
> BTW, the configuration looks fine, but you need to make sure the lcores are
> not split between two different CPU sockets. You can use the
> dpdk/tools/cpu_layout.py to do dump out the system configuration.
>
>
> Keith Wiles, Princi
On Tue, Nov 19, 2013 at 11:31 AM, Wiles, Roger Keith
wrote:
> How do you have Pktgen configured in this case?
>
> On my westmere dual socket 3.4Ghz machine I can send 20G on a single NIC
> 82599x two ports. My machine has a PCIe bug that does not allow me to send
> on more then 3 ports at wire rat
so on the list if you prefer to do so ;) did you try a "perf top" on
your system under load?
On 19/11/2013 10:56, March Jane wrote:
>
> Gratitude to reply.
>
> FCP and Full-offload iSCSI will be nice, for example Qlogic 16Gbps FCP.
>
> But IO scenario is v
Hi All,
I have two NICs (82599) x two ports that are used as packet generators. I
want to generate full line-rate packets (40Gbps), but Pktgen-DPDK does not
seem to be able to do it when two port in a NIC are used simultaneously.
Does anyone know how to generate 40Gbps without replicating packets
? Nowadays dilemmas in enterprise-class storage system & motivation --
The model of most storage system is ?Front-end cards + CPU + back-end
magnetic media hard drive?, in such system, the hard drive is very slow in
terms of its capability in random accessing - roughly 200IO per secon
18 matches
Mail list logo
