From mboxrd@z Thu Jan 1 00:00:00 1970 Return-path: Received: from mail2.candelatech.com ([208.74.158.173]) by bombadil.infradead.org with esmtp (Exim 4.80.1 #2 (Red Hat Linux)) id 1YoZ20-0001Zd-SA for ath10k@lists.infradead.org; Sat, 02 May 2015 15:08:42 +0000 Message-ID: <5544E861.7060402@candelatech.com> Date: Sat, 02 May 2015 08:08:17 -0700 From: Ben Greear MIME-Version: 1.0 Subject: Re: Which Firmware to use for STA (999.x crashes...) References: <5543A152.5050801@candelatech.com> <5543AA43.7050106@candelatech.com> <5543B066.2090706@candelatech.com> In-Reply-To: List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Content-Transfer-Encoding: 7bit Content-Type: text/plain; charset="us-ascii"; Format="flowed" Sender: "ath10k" Errors-To: ath10k-bounces+kvalo=adurom.com@lists.infradead.org To: richard mayers Cc: ath10k@lists.infradead.org On 05/01/2015 03:50 PM, richard mayers wrote: > I found a combination that seems to work quite well. Kernel 3.16, > drivers backported from next-20150424, and as a firmware the candela > one. > > However now I had appeared another problem. I can not read the content > of the file fw_stats. (/sys/kernel/debug/ieee80211/ath10k/). When I > tried to read it says : > > fw_stats: Connection timed out. > > Dmesg : > > ath10k_pci 0000:0c:00.0: failed to request fw stats: -110. > > I remember that I had the same problem with the first firmware I had > but I never had this problem with the candela firmware. > > Could it be the backported driver? Could easily be driver, but first: Please try stock 10.1.467 firmware and see if problem exists there. If so, then it would appear to be a driver bug. If not, then my firmware is at fault. And, try the CT beta firmware if you are trying version 13 currently: http://www.candelatech.com/downloads/ath10k-fw-beta/ Thanks, Ben > > Richard. > > 2015-05-01 19:30 GMT+02:00 richard mayers : >> I will you let know if I find differences between kernels and driver >> versions. So far as I said at the first message my best achievement >> was with candela firmware, kernel 3.16 and backported driver 3.19. >> >> The traffic generator uses around 30%, so I guess its sufficient ! >> >> Richard >> >> 2015-05-01 18:57 GMT+02:00 Ben Greear : >>> On 05/01/2015 09:43 AM, richard mayers wrote: >>>> I meant kernel. In the web page it says that: >>> >>> CT firmware should work fine with upstream kernels. The extra benefits >>> you might get from our modified kernels should be feature related and not performance >>> related. I'd be interested if you found differences in throughput using >>> various kernels... >>> >>> My 4.0 tree is buggy in at least some configurations..I'm working on bisecting it now, >>> but I would stick with our 3.17 tree for now if you are using our kernels. >>> >>>> To take advantage of all of the CT firmware features, please consider >>>> using one of these kernels. Both kernels are rebased fairly often to >>>> keep all CT patches at the top of the git changelog. >>>> >>>> For this measurements I am actually using old laptops. I started the >>>> experiments with a board running OpenWrt, but the results were quite >>>> bad. So I thought that was due to the CPU and I moved to laptops, >>>> however I tried all the lenovo laptops I had at my disposition and >>>> none of them were listing the network card (a lot of people had the >>>> same problem with ). Finally, I found that the cards were recognized >>>> in old Dell laptops, they have a dual core 1.66Ghz, Do you think is >>>> that enough ? I will also try to modify the sockets buffers. >>>> >>>> I read a paper were they report throughputs of 800-900 Mbit/s using >>>> the same network card and running the tests in normal environments. >>> >>> We have seen 950Mbps on one golden Netlink AP, but cannot reliably >>> reproduce that good throughput. >>> >>> Your CPU is probably sufficient...but you might check 'top' >>> to make sure your traffic generator is not at 100% CPU. >>> >>> Thanks, >>> Ben >>> >>>> >>>> Thanks ! >>>> >>>> Richard >>>> >>>> 2015-05-01 18:30 GMT+02:00 Ben Greear : >>>>> On 05/01/2015 09:21 AM, richard mayers wrote: >>>>>> Sorry for not providing details, I wanted to focus more in the >>>>>> firmware crash. Therefore you recommend me to use the Candela >>>>>> firmware. If I just want to run throughput, delay, and fairness >>>>>> measurements do you think is okay not using the kernel provided by >>>>>> them ?. >>>>> >>>>> >>>>> You mean firmware not provided by them? >>>>> >>>>> I'd be happy to see comparisons between my CT firmware and others, >>>>> but if official firmware just crashes, then I guess you cannot do >>>>> that. >>>>> >>>>> If you can crash CT firmware, let me know...I have a chance of >>>>> fixing it. I have no chance of fixing upstream firmware, but >>>>> maybe you can get some help from QCA folks. >>>>> >>>>> With UDP, we often see > 700Mbps throughput, but most of these >>>>> types of tests we do in isolation chambers cabled up. >>>>> >>>>> Over-the-air is probably more like 600Mbps in a clean environment. >>>>> >>>>> You typically have to increase kernel socket buffers to get best >>>>> throughput, and CPU on the host starts to matter at those speeds >>>>> as well. >>>>> >>>>> Antenna matter a lot...just play around with it. >>>>> >>>>> Thanks, >>>>> Ben >>>>> >>>>>> >>>>>> Anyway, for the throughput I do as follows: >>>>>> >>>>>> Tool : Iperf. (iperf -c server_ip -b 1000M -i 1) >>>>>> Protocol : UDP >>>>>> Channel : Over the air, with three antennas. >>>>>> AP: hostapd, channel 149, 80Mhz. With the same kernel, driver, >>>>>> firmware, and NIC than the Station. >>>>>> STA: I get the maximum throughput with MCS = 5/6, when I go further to >>>>>> 7-9 the throughput drops dramatically to 50Mbit/s. >>>>>> >>>>>> I think the antennas should be separated each other 12 cm, I am wrong >>>>>> ? can this have a high impact on the throughput ? >>>>>> >>>>>> Richard >>>>>> >>>>>> 2015-05-01 17:52 GMT+02:00 Ben Greear : >>>>>>> Candela firmware should work just fine for stations. >>>>>>> >>>>>>> And, when reporting throughput, you should also provide details >>>>>>> about how you tested (tool, protocol, over-the-air or not, AP >>>>>>> involved, etc). >>>>>>> >>>>>>> Thanks, >>>>>>> Ben >>>>>>> >>>>>>> >>>>>>> On 05/01/2015 08:38 AM, richard mayers wrote: >>>>>>>> Hi everyone! >>>>>>>> >>>>>>>> I need to do some 802.11ac measurements. For that I am using the >>>>>>>> Compex wle900vx. >>>>>>>> >>>>>>>> I had a set up in which everything was working quite well, however I >>>>>>>> had a maximum throughput of 380Mbits/s ( with all the debug options >>>>>>>> enabled ) >>>>>>>> >>>>>>>> In this first set up I had been using ubuntu 14.04 LTS, kernel version >>>>>>>> 3.16, for the driver I used the latest stable backport (3.19) and >>>>>>>> finally as a firmware the commercial firmware from Candela Tech >>>>>>>> (latest version) in both the AP and STA. >>>>>>>> >>>>>>>> However, I read in the wiki that the latest firmware version for AP is >>>>>>>> the 10.2.4.48, and the Stations should use 999.999.0.63. For this >>>>>>>> second set up I changed almost everything: >>>>>>>> >>>>>>>> - Upgraded kernel to 4.0.1 >>>>>>>> - Backported drivers from the git Modinfo output -> (backported from >>>>>>>> Linux (v3.2-rc1-243765-ga2ff7c6) using backports >>>>>>>> backports-20150424-2-g982) >>>>>>>> >>>>>>>> AP : 10.2.4.48 firmware >>>>>>>> STA : Following the wiki : 999.999.0.63 >>>>>>>> >>>>>>>> During boot the firmware is loaded without any problem: >>>>>>>> >>>>>>>> ath10k_pci 0000:0c:00.0: qca988x hw2.0 (0x4100016c, 0x043202ff) fw >>>>>>>> 999.999.0.636 api 2 htt 2.1 wmi 1 cal otp max_sta 16 >>>>>>>> >>>>>>>> However when I associate with the AP, the firmware crashes. >>>>>>>> >>>>>>>> [11831.954992] ath10k_pci 0000:0c:00.0: firmware crashed! (uuid n/a) >>>>>>>> [11831.955031] ath10k_pci 0000:0c:00.0: qca988x hw2.0 (0x4100016c, >>>>>>>> 0x043202ff) fw 999.999.0.636 api 2 htt 2.1 wmi 1 cal otp max_sta 16 >>>>>>>> [11831.955039] ath10k_pci 0000:0c:00.0: debug 0 debugfs 0 tracing 0 >>>>>>>> dfs 0 testmode 0 >>>>>>>> [11831.957059] ath10k_pci 0000:0c:00.0: firmware register dump: >>>>>>>> [11831.957068] ath10k_pci 0000:0c:00.0: [00]: 0x4100016C 0x00000000 >>>>>>>> 0x009C4521 0x00000000 >>>>>>>> [11831.957076] ath10k_pci 0000:0c:00.0: [04]: 0x009C4521 0x00060530 >>>>>>>> 0x00000019 0x00955A00 >>>>>>>> [11831.957084] ath10k_pci 0000:0c:00.0: [08]: 0x00306862 0x00000000 >>>>>>>> 0x0040CC94 0x00000020 >>>>>>>> [11831.957091] ath10k_pci 0000:0c:00.0: [12]: 0x00000000 0x00000000 >>>>>>>> 0x00958360 0x0095836B >>>>>>>> [11831.957098] ath10k_pci 0000:0c:00.0: [16]: 0x80991091 0x0040AD94 >>>>>>>> 0x0043155C 0x00000000 >>>>>>>> [11831.957106] ath10k_pci 0000:0c:00.0: [20]: 0x0000FFFF 0x00000000 >>>>>>>> 0x0043155C 0x00000000 >>>>>>>> [11831.957114] ath10k_pci 0000:0c:00.0: [24]: 0x809A0978 0x0040AD94 >>>>>>>> 0x00439484 0xC1A18834 >>>>>>>> [11831.957121] ath10k_pci 0000:0c:00.0: [28]: 0x809AD1A2 0x0040ADE4 >>>>>>>> 0x00439484 0x0043FB1C >>>>>>>> [11831.957129] ath10k_pci 0000:0c:00.0: [32]: 0x809B35A3 0x0000001A >>>>>>>> 0x00440A50 0x0000000E >>>>>>>> [11831.957136] ath10k_pci 0000:0c:00.0: [36]: 0x00000000 0x00000000 >>>>>>>> 0x00000000 0x00000000 >>>>>>>> [11831.957143] ath10k_pci 0000:0c:00.0: [40]: 0x00000000 0x00000000 >>>>>>>> 0x00000000 0x00000094 >>>>>>>> [11831.957151] ath10k_pci 0000:0c:00.0: [44]: 0x00439DC8 0x00000000 >>>>>>>> 0x00000000 0x00400000 >>>>>>>> [11831.957158] ath10k_pci 0000:0c:00.0: [48]: 0x809AE0B4 0x0040AE04 >>>>>>>> 0x00400000 0x0043FB1C >>>>>>>> [11831.957166] ath10k_pci 0000:0c:00.0: [52]: 0x00000001 0x00000000 >>>>>>>> 0x00423410 0x00400000 >>>>>>>> [11831.957173] ath10k_pci 0000:0c:00.0: [56]: 0x809AE17E 0x0040AE44 >>>>>>>> 0x0040FE6C 0x0040D310 >>>>>>>> [11832.046908] ath10k_pci 0000:0c:00.0: failed to synchronize monitor >>>>>>>> vdev 1 stop: -108 >>>>>>>> [11832.046915] ath10k_pci 0000:0c:00.0: failed to stop monitor vdev: -108 >>>>>>>> >>>>>>>> I have seen that in the past a lot of people had problems with this >>>>>>>> firmware but since years have passed maybe the way of solving have >>>>>>>> evolved. >>>>>>>> >>>>>>>> Apart from that I also got the 00:03:07:12:34:56 MAC address issue. >>>>>>>> >>>>>>>> Any ideas guys? Should I use the candelatech firmware for the stations ? >>>>>>>> Richard >>>>>>>> >>>>>>>> _______________________________________________ >>>>>>>> ath10k mailing list >>>>>>>> ath10k@lists.infradead.org >>>>>>>> http://lists.infradead.org/mailman/listinfo/ath10k >>>>>>>> >>>>>>> >>>>>>> >>>>>>> -- >>>>>>> Ben Greear >>>>>>> Candela Technologies Inc http://www.candelatech.com >>>>>>> >>>>>> >>>>> >>>>> >>>>> -- >>>>> Ben Greear >>>>> Candela Technologies Inc http://www.candelatech.com >>>>> >>>> >>> >>> >>> -- >>> Ben Greear >>> Candela Technologies Inc http://www.candelatech.com >>> > > _______________________________________________ > ath10k mailing list > ath10k@lists.infradead.org > http://lists.infradead.org/mailman/listinfo/ath10k > -- Ben Greear Candela Technologies Inc http://www.candelatech.com _______________________________________________ ath10k mailing list ath10k@lists.infradead.org http://lists.infradead.org/mailman/listinfo/ath10k