Re: [dpdk-users] [dpdk][pktgen] - execute pktgen with mlx4 (40G) with some errors

[zhilong zheng]

Hi,

After enabling the MLX4 flag and recompiling the DPDK source code, Do you 
recompile the code of pktgen? I think you can recompile it by running make 
clean and make.

Regards,
Zhilong

> 在 2017年8月1日，11:38，Paul Tsvika  写道：
> 
> Hi Olga,
> 
> 2017-08-01 0:11 GMT+08:00 Olga Shern :
> 
>> Hi Paul,
>> 
>> Looks like your ConnectX-3 configured to be IB and not ETH
>> 
>>> pci@:04:00.0  ib0 networkMT27500 Family [ConnectX-3]
>> 
>> Use connectx_port_config script.
>> Set all the ports to Ethernet, and follow the instructions on the screen.
>> 
> 
> Thanks. I followed the script and configured  both ports to Ethernet.
> 
> ibstat info below:
> 
> CA 'mlx4_0'
>CA type: MT4099
>Number of ports: 2
>Firmware version: 2.40.7000
>Hardware version: 1
>Node GUID: 0xe41d2d0300024480
>System image GUID: 0xe41d2d0300024483
>Port 1:
>State: Down
>Physical state: Disabled
>Rate: 10
>Base lid: 0
>LMC: 0
>SM lid: 0
>Capability mask: 0x0401
>Port GUID: 0xe61d2dfffe024481
>Link layer: Ethernet
>Port 2:
>State: Active
>Physical state: LinkUp
>Rate: 40
>Base lid: 0
>LMC: 0
>SM lid: 0
>Capability mask: 0x0401
>Port GUID: 0xe61d2dfffe024482
>Link layer: Ethernet
> 
> I also change the default value from n to  CONFIG_RTE_LIBRTE_MLX4_PMD=y.
> Recompiled DPDK and it all worked.
> 
> And Below is my ifconfig info:
> 
> ens2  Link encap:Ethernet  HWaddr e4:1d:2d:02:44:81
>  UP BROADCAST MULTICAST  MTU:1500  Metric:1
>  RX packets:0 errors:0 dropped:0 overruns:0 frame:0
>  TX packets:0 errors:0 dropped:0 overruns:0 carrier:0
>  collisions:0 txqueuelen:1000
>  RX bytes:0 (0.0 B)  TX bytes:0 (0.0 B)
> 
> ens2d1Link encap:Ethernet  HWaddr e4:1d:2d:02:44:82
>  UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
>  RX packets:126 errors:0 dropped:0 overruns:0 frame:0
>  TX packets:220 errors:0 dropped:0 overruns:0 carrier:0
>  collisions:0 txqueuelen:1000
>  RX bytes:29494 (29.4 KB)  TX bytes:38063 (38.0 KB)
> 
> 
> Since I assume that I don't have to bind the device manually, I executed
> ./dpdk-devbind.py --status with the following info popped out:
> 
> Network devices using kernel driver
> ===
> :03:00.0 'Ethernet Connection X552/X557-AT 10GBASE-T 15ad' if=eno3
> drv=ixgbe unused=
> :03:00.1 'Ethernet Connection X552/X557-AT 10GBASE-T 15ad' if=eno4
> drv=ixgbe unused=
> :04:00.0 'MT27500 Family [ConnectX-3] 1003' if=ens2,ens2d1
> drv=mlx4_core unused=
> :05:00.0 'I350 Gigabit Network Connection 1521' if=eno1 drv=igb unused=
> :05:00.1 'I350 Gigabit Network Connection 1521' if=eno2 drv=igb unused=
> *Active*
> 
> 
> With these configurations, I run commands in order but still had the same
> result:
> 
> $test: modprobe -a ib_uverbs mlx4_en mlx4_core mlx4_ib
> $test: /etc/init.d/openibd restart
> $test: sudo ./app/x86_64-native-linuxapp-gcc/pktgen -l 0-6 -n 1 --proc-type
> auto --file-prefix pg -b 03:00.0 -b 03:00.1 -b 05:00.0 -b 05:00.1 -- -P -m
> "[1-5.0]"
> 
> 
 Packet Burst 32, RX Desc 512, TX Desc 1024, mbufs/port 8192, mbuf cache
> 1024
> !PANIC!: *** Did not find any ports to use ***
> PANIC in pktgen_config_ports():
> *** Did not find any ports to use ***6:
> [./app/x86_64-native-linuxapp-gcc/pktgen(_start+0x29)
> [0x447239]]
> 5: [/lib/x86_64-linux-gnu/libc.so.6(__libc_start_main+0xf0)
> [0x7f0ec0931830]]
> 4: [./app/x86_64-native-linuxapp-gcc/pktgen(main+0x580) [0x444710]]
> 3: [./app/x86_64-native-linuxapp-gcc/pktgen(pktgen_config_ports+0x1614)
> [0x46e324]]
> 2: [./app/x86_64-native-linuxapp-gcc/pktgen(__rte_panic+0xc3) [0x43ead4]]
> 1: [./app/x86_64-native-linuxapp-gcc/pktgen(rte_dump_stack+0x2b) [0x4b7ecb]]
> Aborted (core dumped)
> 
> 
> 
> Is there anything I've done wrong here?
> 
> 
> Thanks
> 
> 
> P.T
> 
> 
>> You can use  -w to specify ConnectX-3 PCI address, you don't need to use -b
>> 
>> Make sure you compile DPDK with mlx4 PMD, it is not enabled by default
>> 
>> Best Regards,
>> Olga
>> 
>> 
>> 
> 
> -- 
> P.T

Re: [dpdk-users] [dpdk][pktgen] - execute pktgen with mlx4 (40G) with some errors

[Paul Tsvika]

Hi Olga,

2017-08-01 0:11 GMT+08:00 Olga Shern :

> Hi Paul,
>
> Looks like your ConnectX-3 configured to be IB and not ETH
>
> > pci@:04:00.0  ib0 networkMT27500 Family [ConnectX-3]
>
> Use connectx_port_config script.
> Set all the ports to Ethernet, and follow the instructions on the screen.
>

Thanks. I followed the script and configured  both ports to Ethernet.

ibstat info below:

CA 'mlx4_0'
CA type: MT4099
Number of ports: 2
Firmware version: 2.40.7000
Hardware version: 1
Node GUID: 0xe41d2d0300024480
System image GUID: 0xe41d2d0300024483
Port 1:
State: Down
Physical state: Disabled
Rate: 10
Base lid: 0
LMC: 0
SM lid: 0
Capability mask: 0x0401
Port GUID: 0xe61d2dfffe024481
Link layer: Ethernet
Port 2:
State: Active
Physical state: LinkUp
Rate: 40
Base lid: 0
LMC: 0
SM lid: 0
Capability mask: 0x0401
Port GUID: 0xe61d2dfffe024482
Link layer: Ethernet

I also change the default value from n to  CONFIG_RTE_LIBRTE_MLX4_PMD=y.
Recompiled DPDK and it all worked.

And Below is my ifconfig info:

ens2  Link encap:Ethernet  HWaddr e4:1d:2d:02:44:81
  UP BROADCAST MULTICAST  MTU:1500  Metric:1
  RX packets:0 errors:0 dropped:0 overruns:0 frame:0
  TX packets:0 errors:0 dropped:0 overruns:0 carrier:0
  collisions:0 txqueuelen:1000
  RX bytes:0 (0.0 B)  TX bytes:0 (0.0 B)

ens2d1Link encap:Ethernet  HWaddr e4:1d:2d:02:44:82
  UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
  RX packets:126 errors:0 dropped:0 overruns:0 frame:0
  TX packets:220 errors:0 dropped:0 overruns:0 carrier:0
  collisions:0 txqueuelen:1000
  RX bytes:29494 (29.4 KB)  TX bytes:38063 (38.0 KB)


Since I assume that I don't have to bind the device manually, I executed
./dpdk-devbind.py --status with the following info popped out:

Network devices using kernel driver
===
:03:00.0 'Ethernet Connection X552/X557-AT 10GBASE-T 15ad' if=eno3
drv=ixgbe unused=
:03:00.1 'Ethernet Connection X552/X557-AT 10GBASE-T 15ad' if=eno4
drv=ixgbe unused=
:04:00.0 'MT27500 Family [ConnectX-3] 1003' if=ens2,ens2d1
drv=mlx4_core unused=
:05:00.0 'I350 Gigabit Network Connection 1521' if=eno1 drv=igb unused=
:05:00.1 'I350 Gigabit Network Connection 1521' if=eno2 drv=igb unused=
*Active*


With these configurations, I run commands in order but still had the same
result:

$test: modprobe -a ib_uverbs mlx4_en mlx4_core mlx4_ib
$test: /etc/init.d/openibd restart
$test: sudo ./app/x86_64-native-linuxapp-gcc/pktgen -l 0-6 -n 1 --proc-type
auto --file-prefix pg -b 03:00.0 -b 03:00.1 -b 05:00.0 -b 05:00.1 -- -P -m
"[1-5.0]"


>>> Packet Burst 32, RX Desc 512, TX Desc 1024, mbufs/port 8192, mbuf cache
1024
!PANIC!: *** Did not find any ports to use ***
PANIC in pktgen_config_ports():
*** Did not find any ports to use ***6:
[./app/x86_64-native-linuxapp-gcc/pktgen(_start+0x29)
[0x447239]]
5: [/lib/x86_64-linux-gnu/libc.so.6(__libc_start_main+0xf0)
[0x7f0ec0931830]]
4: [./app/x86_64-native-linuxapp-gcc/pktgen(main+0x580) [0x444710]]
3: [./app/x86_64-native-linuxapp-gcc/pktgen(pktgen_config_ports+0x1614)
[0x46e324]]
2: [./app/x86_64-native-linuxapp-gcc/pktgen(__rte_panic+0xc3) [0x43ead4]]
1: [./app/x86_64-native-linuxapp-gcc/pktgen(rte_dump_stack+0x2b) [0x4b7ecb]]
Aborted (core dumped)



Is there anything I've done wrong here?


Thanks


P.T


> You can use  -w to specify ConnectX-3 PCI address, you don't need to use -b
>
> Make sure you compile DPDK with mlx4 PMD, it is not enabled by default
>
> Best Regards,
> Olga
>
>
>

-- 
P.T

[dpdk-users] KNI tcp segmentation offloading

[Yuliang Li]

Hi,

I am using KNI, and I hope the TCP can use segmentation offloading.
However, when I use "ethtool -k vEth0" (the KNI virtual interface), it
shows:
tcp-segmentation-offload: off
tx-tcp-segmentation: off [fixed]
tx-tcp-ecn-segmentation: off [fixed]
tx-tcp6-segmentation: off [fixed]

and "sudo ethtool -K vEth0 tso on" gives me:
Cannot change tcp-segmentation-offload
Could not change any device features

Is there a way to enable TCP segmentation offloading with KNI?

Thanks,
-- 
Yuliang Li
PhD student
Department of Computer Science
Yale University

Re: [dpdk-users] If shared libraries are used vdev doesn't work anymore

[Thomas Monjalon]

31/07/2017 16:58, Dirk-Holger Lenz:
> If dpdk is built with 'CONFIG_RTE_BUILD_SHARED_LIB=y' then
> using the vdev feature (args: e.g. -c 3 -n 4 --vdev="crypto_openssl")
> the rte_eal_init() returns 'ERROR: failed to parse device "crypto_openssl"'.
> It looks to me that rte_eal_devargs_add() calling rte_eal_devargs_parse()
> is trying to check the device name before the shared libraries are read
> and the internal data arrays are setup.

Yes, you're right: eal_parse_args() is called before eal_plugins_init().
The fix is not small: we should split the args parsing to parse the
device arguments after loading shared libraries.

It is a release blocker.

Re: [dpdk-users] [dpdk][pktgen] - execute pktgen with mlx4 (40G) with some errors

[Olga Shern]

Hi Paul, 

Looks like your ConnectX-3 configured to be IB and not ETH 

> pci@:04:00.0  ib0 networkMT27500 Family [ConnectX-3]

Use connectx_port_config script.
Set all the ports to Ethernet, and follow the instructions on the screen.

You can use  -w to specify ConnectX-3 PCI address, you don't need to use -b

Make sure you compile DPDK with mlx4 PMD, it is not enabled by default  

Best Regards,
Olga


> -Original Message-
> From: users [mailto:users-boun...@dpdk.org] On Behalf Of Paul Tsvika
> Sent: Monday, July 31, 2017 1:14 PM
> To: users@dpdk.org
> Subject: [dpdk-users] [dpdk][pktgen] - execute pktgen with mlx4 (40G) with
> some errors
> 
> Hi,
> 
> Below is my configuration:
> 
> AOC: Mellanox MT27500 Family [ConnectX-3]
> OS: Ubuntu 16.04
> DPDK version: 17.05.1
> Pktgen: 3.3.8
> Memory: 16G
> CPU: Intel(R) Xeon(R) CPU D-1557 (12C)
> 
> I've enabled the configuration in dpdk (CONFIG_RTE_LIBRTE_MLX4_PMD=y)
> and installed Mellanox OFED
>  ww.mellanox.com%2Fpage%2Fproducts_dyn%3Fproduct_family%3D26%26
> mtag%3Dlinux_sw_drivers&data=02%7C01%7Colgas%40mellanox.com%7C4f
> e2f94155994a2a579008d4d7fce66a%7Ca652971c7d2e4d9ba6a4d149256f461b
> %7C0%7C0%7C636370928586578406&sdata=%2FPnEAOJFsGtQRY%2FP5u0Hpl
> L6kDbnhwRGIgSV%2B31w40Q%3D&reserved=0>
> driver and recompiled DPDK. It all worked.
> 
> Below is my pci network class information:
> 
> pci@:03:00.0  eno3networkEthernet Connection X552/X557-AT
> 10GBAS
> pci@:03:00.1  eno4networkEthernet Connection X552/X557-AT
> 10GBAS
> pci@:04:00.0  ib0 networkMT27500 Family [ConnectX-3]
> pci@:05:00.0  eno1networkI350 Gigabit Network Connection
> pci@:05:00.1  eno2networkI350 Gigabit Network Connection
> pci@:04:00.0  ens2d1  networkEthernet interface
> 
> And my step to run pktgen is shown in the following:
> 
> $test: modprobe -a ib_uverbs mlx4_en mlx4_core mlx4_ib
> $test: /etc/init.d/openibd restart
> $test: sudo ./app/x86_64-native-linuxapp-gcc/pktgen -l 0-6 -n 1 --proc-type
> auto --file-prefix pg -b 03:00.0 -b 03:00.1 -b 05:00.0 -b 05:00.1 -- -P -m 
> "[1-
> 5.0]"
> 
> 
> Looks like in order to use mlx4, i need to blacklist other ports so that i -b
> other ports.
> 
> 
> However, when i run the command above to execute pktgen, i got the
> following error:
> 
> >>> Packet Burst 32, RX Desc 512, TX Desc 1024, mbufs/port 8192, mbuf
> >>> cache
> 1024
> !PANIC!: *** Did not find any ports to use *** PANIC in
> pktgen_config_ports():
> *** Did not find any ports to use ***6:
> [./app/x86_64-native-linuxapp-gcc/pktgen(_start+0x29) [0x447239]]
> 5: [/lib/x86_64-linux-gnu/libc.so.6(__libc_start_main+0xf0)
> [0x7f0ec0931830]]
> 4: [./app/x86_64-native-linuxapp-gcc/pktgen(main+0x580) [0x444710]]
> 3: [./app/x86_64-native-linuxapp-gcc/pktgen(pktgen_config_ports+0x1614)
> [0x46e324]]
> 2: [./app/x86_64-native-linuxapp-gcc/pktgen(__rte_panic+0xc3) [0x43ead4]]
> 1: [./app/x86_64-native-linuxapp-gcc/pktgen(rte_dump_stack+0x2b)
> [0x4b7ecb]] Aborted (core dumped)
> 
> 
> Anyone might have related experience please provide advise.
> 
> 
> Thanks
> 
> 
> P.T
> 
> 
> --
> P.T

Re: [dpdk-users] If shared libraries are used vdev doesn't work anymore

[Dirk-Holger Lenz]

I forgot to tell that this happens with 17.08rc2 and latest clone from git.

[dpdk-users] If shared libraries are used vdev doesn't work anymore

[Dirk-Holger Lenz]

If dpdk is built with 'CONFIG_RTE_BUILD_SHARED_LIB=y' then
using the vdev feature (args: e.g. -c 3 -n 4 --vdev="crypto_openssl")
the rte_eal_init() returns 'ERROR: failed to parse device "crypto_openssl"'.
It looks to me that rte_eal_devargs_add() calling rte_eal_devargs_parse()
is trying to check the device name before the shared libraries are read
and the internal data arrays are setup.

Re: [dpdk-users] [dpdk][pktgen] - execute pktgen with mlx4 (40G) with some errors

[Paul Tsvika]

I will provide the info for the "./dpdk-devbind.py" and "ibstat"


Regards-


P.T

2017-07-31 21:40 GMT+08:00 zhilong zheng :

> No, you do not need any operation. This driver is bound to the device
> automatically.
>
> 在 2017年7月31日，21:25，Paul Tsvika  写道：
>
> Sorry I am not with my testing machine now.
> if it uses mlx4_core, should I bind the device to the driver by using the
> following command?
>
> *./dpdk-devbind.py -b mlx4_core xx:xx.x*
>
>
> Thanks
>
>
> P.T
>
> 2017-07-31 21:14 GMT+08:00 zhilong zheng :
>
>> Yes, CX3 does not use igb_uio, it uses mlx4_core as the driver and you do
>> not need to bind the device to igb_uio. Can you show the output when
>> running command “*./dpdk-devbind.py —status*” and “*ibstat*"
>>
>> 在 2017年7月31日，21:07，Paul Tsvika  写道：
>>
>> er I should
>>
>>
>>
>
>
> --
> P.T
>
>
>


-- 
P.T

Re: [dpdk-users] [dpdk][pktgen] - execute pktgen with mlx4 (40G) with some errors

[zhilong zheng]

No, you do not need any operation. This driver is bound to the device 
automatically.

> 在 2017年7月31日，21:25，Paul Tsvika  写道：
> 
> Sorry I am not with my testing machine now. 
> if it uses mlx4_core, should I bind the device to the driver by using the 
> following command?
> 
> ./dpdk-devbind.py -b mlx4_core xx:xx.x
> 
> 
> Thanks
> 
> 
> P.T
> 
> 2017-07-31 21:14 GMT+08:00 zhilong zheng  >:
> Yes, CX3 does not use igb_uio, it uses mlx4_core as the driver and you do not 
> need to bind the device to igb_uio. Can you show the output when running 
> command “./dpdk-devbind.py —status” and “ibstat"
> 
>> 在 2017年7月31日，21:07，Paul Tsvika > > 写道：
>> 
>> er I should 
> 
> 
> 
> 
> -- 
> P.T

Re: [dpdk-users] [dpdk][pktgen] - execute pktgen with mlx4 (40G) with some errors

[Paul Tsvika]

Sorry I am not with my testing machine now.
if it uses mlx4_core, should I bind the device to the driver by using the
following command?

*./dpdk-devbind.py -b mlx4_core xx:xx.x*


Thanks


P.T

2017-07-31 21:14 GMT+08:00 zhilong zheng :

> Yes, CX3 does not use igb_uio, it uses mlx4_core as the driver and you do
> not need to bind the device to igb_uio. Can you show the output when
> running command “*./dpdk-devbind.py —status*” and “*ibstat*"
>
> 在 2017年7月31日，21:07，Paul Tsvika  写道：
>
> er I should
>
>
>


-- 
P.T

Re: [dpdk-users] [dpdk][pktgen] - execute pktgen with mlx4 (40G) with some errors

[zhilong zheng]

Yes, CX3 does not use igb_uio, it uses mlx4_core as the driver and you do not 
need to bind the device to igb_uio. Can you show the output when running 
command “./dpdk-devbind.py —status” and “ibstat"

> 在 2017年7月31日，21:07，Paul Tsvika  写道：
> 
> er I should 

Re: [dpdk-users] [dpdk][pktgen] - execute pktgen with mlx4 (40G) with some errors

[Paul Tsvika]

Hi zhilong,

I did not bind the device with dpdk-devbind.py.
I am not sure if this is a must because I don't see it in the tutorial.

the command I have used to test other ethernet devices was

./dpdk-devbind.py -b igb_uio pcie_addr

Since mlx4 does not use igb_uio as its driver, if this is a must, which
driver I should take instead ?


Thanks



P.T


2017-07-31 20:54 GMT+08:00 zhilong zheng :

> Hi Paul,
>
> It seems that DPDK can only find the ethernet device with Mellanox. So you
> can check your interfaces’ link layer protocol by running command “
> *ibstat*” to ensure that they are using ethernet, but not infiniband.
>
> Hope it can help you.
>
> Regards,
> Zhilong
>
> 在 2017年7月31日，18:14，Paul Tsvika  写道：
>
> Hi,
>
> Below is my configuration:
>
> AOC: Mellanox MT27500 Family [ConnectX-3]
> OS: Ubuntu 16.04
> DPDK version: 17.05.1
> Pktgen: 3.3.8
> Memory: 16G
> CPU: Intel(R) Xeon(R) CPU D-1557 (12C)
>
> I've enabled the configuration in dpdk (CONFIG_RTE_LIBRTE_MLX4_PMD=y) and
> installed Mellanox OFED
>
>
>


-- 
P.T

Re: [dpdk-users] l3fwd rte_eth_dev_configure: err=-17

[czy12345]

The DPDK version I used is 17.02.1 . Today I also tried 17.05 and 17.05.1, the 
result were the same except the return err which was -22. 

Network devices using DPDK-compatible driver

:02:00.0 '82574L Gigabit Network Connection 10d3' drv=igb_uio 
unused=vfio-pci
:03:00.0 '82574L Gigabit Network Connection 10d3' drv=igb_uio 
unused=vfio-pci

Thanks.


At2017-07-31 13:47:41,Pavel Shirshovwrote:
> Hello,
> 
> What version of DPDK?
> What NIC on port 0?
> 
> Thanks
> 
> On Sun, Jul 30, 2017 at 12:42 AM,   wrote:
> > Hello experts,
> >
> > I have successfully ran l2fwd example. But I'm having troubles running the 
> > l3fwd example. I find that in l2fwd the nb_tx_queue parameter of function 
> > rte_eth_dev_configure is equal to 1, but in l3fwd the parameter is set to 
> > the number of lcores. And then when I use more than one lcores, the 
> > function fails. Wondering what is related with the problem?
> >
> > Thanks in advance.
> >
> > #./l3fwd -l 1 -n 4 -- -p 0x3 --config="(0,0,1),(1,0,1)" --parse-ptype
> > success
> >
> > #./l3fwd -l 1,2 -n 4 -- -p 0x3 --config="(0,0,1),(1,0,2)" --parse-ptype
> > EAL: Detected 8 lcore(s)
> > EAL: Probing VFIO support...
> > EAL: VFIO support initialized
> > EAL: PCI device :01:00.0 on NUMA socket -1
> > EAL:   probe driver: 8086:10e6 net_e1000_igb
> > EAL: PCI device :01:00.1 on NUMA socket -1
> > EAL:   probe driver: 8086:10e6 net_e1000_igb
> > EAL: PCI device :02:00.0 on NUMA socket -1
> > EAL:   probe driver: 8086:10d3 net_e1000_em
> > EAL: PCI device :03:00.0 on NUMA socket -1
> > EAL:   probe driver: 8086:10d3 net_e1000_em
> > EAL: PCI device :04:00.0 on NUMA socket -1
> > EAL:   probe driver: 8086:10d3 net_e1000_em
> > EAL: PCI device :05:00.0 on NUMA socket -1
> > EAL:   probe driver: 8086:10d3 net_e1000_em
> > EAL: PCI device :06:00.0 on NUMA socket -1
> > EAL:   probe driver: 8086:10d3 net_e1000_em
> > EAL: PCI device :07:00.0 on NUMA socket -1
> > EAL:   probe driver: 8086:10d3 net_e1000_em
> > soft parse-ptype is enabled
> > L3FWD: LPM or EM none selected, default LPM on
> > Initializing port 0 ... Creating queues: nb_rxq=1 nb_txq=2... EAL: Error - 
> > exiting with code: 1
> >   Cause: Cannot configure device: err=-17, port=0

Re: [dpdk-users] [dpdk][pktgen] - execute pktgen with mlx4 (40G) with some errors

[zhilong zheng]

Hi Paul,

It seems that DPDK can only find the ethernet device with Mellanox. So you can 
check your interfaces’ link layer protocol by running command “ibstat” to 
ensure that they are using ethernet, but not infiniband. 

Hope it can help you.

Regards,
Zhilong

> 在 2017年7月31日，18:14，Paul Tsvika  写道：
> 
> Hi,
> 
> Below is my configuration:
> 
> AOC: Mellanox MT27500 Family [ConnectX-3]
> OS: Ubuntu 16.04
> DPDK version: 17.05.1
> Pktgen: 3.3.8
> Memory: 16G
> CPU: Intel(R) Xeon(R) CPU D-1557 (12C)
> 
> I've enabled the configuration in dpdk (CONFIG_RTE_LIBRTE_MLX4_PMD=y) and
> installed Mellanox OFED

[dpdk-users] [dpdk][pktgen] - execute pktgen with mlx4 (40G) with some errors

[Paul Tsvika]

Hi,

Below is my configuration:

AOC: Mellanox MT27500 Family [ConnectX-3]
OS: Ubuntu 16.04
DPDK version: 17.05.1
Pktgen: 3.3.8
Memory: 16G
CPU: Intel(R) Xeon(R) CPU D-1557 (12C)

I've enabled the configuration in dpdk (CONFIG_RTE_LIBRTE_MLX4_PMD=y) and
installed Mellanox OFED

driver and recompiled DPDK. It all worked.

Below is my pci network class information:

pci@:03:00.0  eno3networkEthernet Connection X552/X557-AT
10GBAS
pci@:03:00.1  eno4networkEthernet Connection X552/X557-AT
10GBAS
pci@:04:00.0  ib0 networkMT27500 Family [ConnectX-3]
pci@:05:00.0  eno1networkI350 Gigabit Network Connection
pci@:05:00.1  eno2networkI350 Gigabit Network Connection
pci@:04:00.0  ens2d1  networkEthernet interface

And my step to run pktgen is shown in the following:

$test: modprobe -a ib_uverbs mlx4_en mlx4_core mlx4_ib
$test: /etc/init.d/openibd restart
$test: sudo ./app/x86_64-native-linuxapp-gcc/pktgen -l 0-6 -n 1 --proc-type
auto --file-prefix pg -b 03:00.0 -b 03:00.1 -b 05:00.0 -b 05:00.1 -- -P -m
"[1-5.0]"


Looks like in order to use mlx4, i need to blacklist other ports so that i
-b other ports.


However, when i run the command above to execute pktgen, i got the
following error:

>>> Packet Burst 32, RX Desc 512, TX Desc 1024, mbufs/port 8192, mbuf cache
1024
!PANIC!: *** Did not find any ports to use ***
PANIC in pktgen_config_ports():
*** Did not find any ports to use ***6:
[./app/x86_64-native-linuxapp-gcc/pktgen(_start+0x29) [0x447239]]
5: [/lib/x86_64-linux-gnu/libc.so.6(__libc_start_main+0xf0)
[0x7f0ec0931830]]
4: [./app/x86_64-native-linuxapp-gcc/pktgen(main+0x580) [0x444710]]
3: [./app/x86_64-native-linuxapp-gcc/pktgen(pktgen_config_ports+0x1614)
[0x46e324]]
2: [./app/x86_64-native-linuxapp-gcc/pktgen(__rte_panic+0xc3) [0x43ead4]]
1: [./app/x86_64-native-linuxapp-gcc/pktgen(rte_dump_stack+0x2b) [0x4b7ecb]]
Aborted (core dumped)


Anyone might have related experience please provide advise.


Thanks


P.T


-- 
P.T

[dpdk-users] Performance degradation after the program running for awhile

[Yu Yang]

My program is a dns server implemented with run-to-completion model, when I
start it with 11 cores, it has a normal performance(12M qps), but after
5-10 minutes, the qps decreased to 7M, I used Intel vTune Amplifier to
analysis the process and found the top 2 hotspots is *ixgbe_recv_pkts_vec*
and *rte_eth_rx_burst.  * my test environments is:

CPU: E5-2650(two sockets)
memory: 64GB
OS: ubuntu 16.04 64bits

Any advices on how to solve this strange problem?

thx