[dpdk-dev] turn on optical bypass
Hi does dpdk provides any API to turn on optical bypass functionality ? please let me know. tx, Sharath
[dpdk-dev] How to debug packet sends to virtual functions
>> I finally got things working. I apparently missed an "ifconfig up" in >> the >> guest, before starting dpdk. I'm still confused why this would be needed. Is >> dpdk unable to do a full initialization of the virtual function from the >> guest? >> > You wouldn't be able to call ifconfig on your VF device if you have bound > your guest VF device to igb_uio driver. Have you bound your VF device to > igb_uio? Or you have enabled automatic port unbinding (which is disabled by > default in recent releases)? The guest starts with loading the igbvf kernel driver, uses ifconfig up, then loads and binds to igb_uio. After that, DPDK works. If I skip any step here it doesn't work. Skipping "ifconfig" step resulted in packets being received but I couldn't send them, they just got queued up but was never sent. >> While I got one instance of DPDK running, I got a problem when starting two >> instances of DPDK running against different virtual functions. These virtual >> functions stems from the same physical interfaces. >> >> Starting them one or the other works fine. I have to adapt my static arp >> entries since they have different MAC addresses, but this is only expected. >> When starting them both however, I receive no packets. If one is running >> and currently processing packets, it will stop doing so the instance the >> second >> instance starts. >> > Are you referring to two different VM's each having a separate VF device, or > are you trying to run two primary processes on one VM with different VF > devices? It's the latter, i.e. two primary processes running DPDK on the same VM but different VF devices. The VF devices belongs to the same physical devices though. Regards Mats Liljegren
[dpdk-dev] How to debug packet sends to virtual functions
Hi Mats > The guest starts with loading the igbvf kernel driver, uses ifconfig up, > then loads and binds to igb_uio. After that, DPDK works. If I skip any step > here it doesn't work. Skipping "ifconfig" step resulted in packets being > received but I couldn't send them, they just got queued up but was never > sent. > Hm, never seen this before. Apologies if you have already mentioned this, but what NIC are you using? I'll see if I can replicate your issues. > It's the latter, i.e. two primary processes running DPDK on the same VM but > different VF devices. The VF devices belongs to the same physical devices > though. I'll try that as well. Best regards, Anatoly Burakov DPDK SW Engineer -- Intel Shannon Limited Registered in Ireland Registered Office: Collinstown Industrial Park, Leixlip, County Kildare Registered Number: 308263 Business address: Dromore House, East Park, Shannon, Co. Clare
[dpdk-dev] How to debug packet sends to virtual functions
Hi Anatoly, >> The guest starts with loading the igbvf kernel driver, uses ifconfig >> up, >> then loads and binds to igb_uio. After that, DPDK works. If I skip any step >> here it doesn't work. Skipping "ifconfig" step resulted in packets being >> received but I couldn't send them, they just got queued up but was never >> sent. >> > > Hm, never seen this before. Apologies if you have already mentioned this, but > what NIC are you using? I'll see if I can replicate your issues. Not sure if I've mentioned it, but it is a quad I350. Port 0 and 3 on host only, and 1 and 2 are being used by DPDK in a guest. The challenge was to make a ping go from port 0 to 3 and back, via a somewhat modified DPDK l2fwd example. Port 1 and 2 had two virtual functions each, and each such pair was given to an instance of this l2fwd example. >> It's the latter, i.e. two primary processes running DPDK on the same VM but >> different VF devices. The VF devices belongs to the same physical devices >> though. > > I'll try that as well. If you need more help with how we set things up, I'll be happy to help you. Regards Mats Liljegren
[dpdk-dev] Is it possible to have dpdk running with no dependency on a nic ?
Hi Stephen, Have you tried link time optimization on DPDK application? Does it decrease the I-cache miss rate evidently? thx & rgds, -Qinglai On Sun, Feb 16, 2014 at 9:02 PM, Stephen Hemminger wrote: > On Fri, 14 Feb 2014 15:11:29 -0500 > Ymo Lists wrote: > >> "Enqueuing and dequeuing items from an rte_ring using the rings-based PMD >> may be slower than using the native rings API. This is because Intel? DPDK >> Ethernet drivers make use of function pointers to call the appropriate >> enqueue or dequeue functions, while the rte_ring specific functions are >> direct function calls in the code and are often inlined by the compiler." >> >> Is that statement correct ? I would imagine that inlined code would be be >> faster than using function pointers ? > > Actually, the Intel DPDK has a bad case of inlineitis. The code for ring's > and other parts use inline on largish functions which bloats the code without > any perceivable gain in performance. The larger code causes more cache misses > which actually hurt performance. Also using GCC link time optimization helps > to reduce any need for inlining larger code bits.
[dpdk-dev] How to debug packet sends to virtual functions
>>> It's the latter, i.e. two primary processes running DPDK on the same VM but >>> different VF devices. The VF devices belongs to the same physical devices >>> though. >> >> I'll try that as well. > > If you need more help with how we set things up, I'll be happy to help you. I found the bug, it was an illegal arp entry that I had created causing the problem. So this case has been solved. I re-checked this ifconfig issue, and now I can get it to work without ifconfig in the guest. Not sure why it didn't work without it previously, but apparently, there is something else I did that made things work. Sorry about the fuzz, but thanks for your effort! Regards Mats Liljegren
[dpdk-dev] How to debug packet sends to virtual functions
> I found the bug, it was an illegal arp entry that I had created causing the > problem. So this case has been solved. > > I re-checked this ifconfig issue, and now I can get it to work without > ifconfig > in the guest. Not sure why it didn't work without it previously, but > apparently, there is something else I did that made things work. > > Sorry about the fuzz, but thanks for your effort! No problem, glad to hear it's not a DPDK bug :-) Best regards, Anatoly Burakov DPDK SW Engineer -- Intel Shannon Limited Registered in Ireland Registered Office: Collinstown Industrial Park, Leixlip, County Kildare Registered Number: 308263 Business address: Dromore House, East Park, Shannon, Co. Clare
[dpdk-dev] Question on DPDK multi-process support
Hi, I have some doubts on the DPDK multi-process support: 1) According to the Programmers Guide (pag. 109 - 20.3 Multi-process Limitations), one of the multi-process limitations is: "All Intel? DPDK processes running as a single application and using shared memory must have distinct coremask arguments. It is not possible to have a primary and secondary instance, or two secondary instances, using any of the same logical cores. Attempting to do so can cause corruption of memory pool caches, among other issues." Does it mean that I can run a number of DPDK processes at most equal to the number of my CPU logical cores? In the case I run 2 mono-lcore DPDK processes with the same coremask, what are the possible issues? 2) Can a secondary process use the rte_eth_rx_burst over a queue of a NIC port configured in the primary process? Regards, Mauro
[dpdk-dev] Is it possible to have dpdk running with no dependency on a nic ?
On Mon, 17 Feb 2014 14:49:47 +0200 jigsaw wrote: > Hi Stephen, > > Have you tried link time optimization on DPDK application? Does it > decrease the I-cache miss rate evidently? > > thx & > rgds, > -Qinglai > We use link-time-optimization without any problem. No big performance gain, but I really don't think all the inlining of big functions is necessary.
[dpdk-dev] Question on DPDK multi-process support
> Hi, > I have some doubts on the DPDK multi-process support: > > 1) According to the Programmers Guide (pag. 109 - 20.3 Multi-process > Limitations), one of the multi-process limitations is: > > "All Intel(r) DPDK processes running as a single application and using shared > memory must have distinct coremask arguments. It is not possible to have > a primary and secondary instance, or two secondary instances, using any of > the same logical cores. Attempting to do so can cause corruption of > memory pool caches, among other issues." > > Does it mean that I can run a number of DPDK processes at most equal to > the number of my CPU logical cores? In the case I run 2 mono-lcore DPDK > processes with the same coremask, what are the possible issues? > The potential issues are caused by a dependence on the lcore_id internally by Intel DPDK data structures, especially mempools. If two processes use the same lcore they will have the same lcore_id value, and will try and access the same mempool cache which is not thread-safe. This will cause mempool corruption. NOTE: this applies only to co-operating processes, i.e. those run as primary and secondary processes. There are no mempool issues with running completely independent Intel DPDK processes on the same cores, i.e. processes run using different "--file-prefix=" parameters, since those do not share any memory and data structures. > > 2) Can a secondary process use the rte_eth_rx_burst over a queue of a NIC > port configured in the primary process? > Yes. NIC ports and their data structures are shared between co-operating processes. Multiple processes doing RX on the same port is shown in the "examples/multi_process/symmetric_mp" example application.
[dpdk-dev] Is it possible to have dpdk running with no dependency on a nic ?
If in-lining is that big a concern, you could create your own wrapper function and explicitly mark it no-inline. Personally, I haven't seen any inordinate increase in i-cache miss rates because of in-lining on the applications we have - prefetchers on IA are usually capable of keeping up. However, every application is different enough that there really is no "general" rule here. -Venky -Original Message- From: dev [mailto:dev-boun...@dpdk.org] On Behalf Of Stephen Hemminger Sent: Monday, February 17, 2014 8:42 AM To: jigsaw Cc: dev at dpdk.org Subject: Re: [dpdk-dev] Is it possible to have dpdk running with no dependency on a nic ? On Mon, 17 Feb 2014 14:49:47 +0200 jigsaw wrote: > Hi Stephen, > > Have you tried link time optimization on DPDK application? Does it > decrease the I-cache miss rate evidently? > > thx & > rgds, > -Qinglai > We use link-time-optimization without any problem. No big performance gain, but I really don't think all the inlining of big functions is necessary.
