Here is what I see connecting from a bbb on nfs root via a 100Mb switch to another host:
tcpbench tcp 5.2 Mbps udp 11.6 Mbps 990 pps iperf tcp 5.29 Mbits/sec udp 1.05 Mbits/sec I'm not sure why udp differs here. Connecting from linux that was preinstalled on the emmc: iperf tcp 92.5 Mbits/sec udp 1.05 Mbits/sec Perhaps the switch frontend/ALE configuration is not set up correctly? I'm sure you must have read a bit about that when looking into adding support for the other port. It looks like there is not a clean seperation of registers/interrupts/dma for the different ports which is annoying. The driver in OpenBSD does not attempt to do interrupt mitigation either. Though perhaps with such a large difference something more fundamental like the processor cache/pmap/frequency handling could be implicated? On Tue, Nov 04, 2014 at 08:42:20AM +0100, Mérindol Bernard wrote: > Hi Jonathan, > > Thank for your answer. > > I have re build my OpenBSD with Artheros Phy support > > com0: console > cpsw0 at omap0: version 1.12 (0), address d4:94:a1:84:28:f2 > atphy0 at cpsw0 phy 0: F1 10/100/1000 PHY, rev. 4 > atphy1 at cpsw0 phy 1: F1 10/100/1000 PHY, rev. 4 > scsibus0 at sdmmc0: 2 targets, initiator 0 > sd0 at scsibus0 targ 1 lun 0: <SD/MMC, Drive #01, > SCSI2 0/direct fixed > > > But the result is same. I try to install last version of NetBSD to test. > > If you or an other person have BeaglBoard with OpenBSD installed is it > possible to send to me the result of this command. > > On OpenBsd > nc -l 1234 > /dev/null > > on other host (with XXBSD example OS X); > mkfile 100m CentM > > date;nc <@IP of your BeagleBoard> 1234 < CentM;date > > If you use Linux try with this command to create 100M file. > dd if=/dev/zero of=CentM bs=100M count=1 > > Thank > Bernard Merindol > > PS: send also the result of ifconfig cpsw0 > > > On 3 nov. 2014, at 14:13, Jonathan Gray <j...@jsg.id.au> wrote: > > > > > > You have a different phy to the beaglebone boards. > > It seems to be an Atheros phy that would be handled by atphy(4) > > if it were added to the kernel configuration. > > > > I doubt that would make much of a difference though. > > > > rev 1.5 in NetBSD which programs the multicast filter could be > > interesting > > > > http://cvsweb.netbsd.org/bsdweb.cgi/src/sys/arch/arm/omap/if_cpsw.c?only_with_tag=MAIN > > > > "Make cpsw driver work without uboot support > > > > On some eval boards such as BeagleBone, the cpsw device is initialized > > rightly by the uboot of the boards so that the cpsw driver doesn't need > > to do some initializations but works fine. > > > > The patch adds initializations to make the driver work solely. It also > > adds support for 1000BaseT (RGMII) PHY that is equipped on some boards, > > e.g., CKB-3352." > > > > It seems the phy might need special handling according to > > http://processors.wiki.ti.com/index.php/AM335x_General_Purpose_EVM_HW_User_Guide#Gbit_Ethernet > > > > "Note: The GP EVM PCB does not support external delay for the clock > > signals on the RGMII interface. The AR8031A PHY can be setup to use > > internal clock delay mode once booting is complete. Therefore there may > > be problems in function/performance when booting in Gb mode or when > > running Gb Ethernet Mode before the AR8031A PHY is configured properly. > > See the EVM Errata." > > > > Which sounds rather like the cpsw specific hack in > > http://cvsweb.netbsd.org/bsdweb.cgi/~checkout~/src/sys/dev/mii/micphy.c?rev=1.3 > > > > On Mon, Nov 03, 2014 at 12:01:39PM +0100, Mérindol Bernard wrote: > >> Hi All, > >> > >> I works with AM335x Starter Kit from TI, I modify the cpsw driver to works > >> with 2 ports (cpsw0 and cpsw1). > >> > >> Before this modification I have tested the performance on native driver > >> with this command. > >> > >> on OBSD > >> nc -l 1234 > /dev/null > >> > >> > >> on host connected by switch 100Mbit/s full duplex (MAC OX 10.10) > >> > >> nc 192.168.169.62 1234 < CentM > >> > >> where CentM il file with 100MByte (created with makefile 100m CentM). > >> > >> On BSD5.5 or 5.6 the time for this transfert is 93s = 8,6 Mbit/s > >> > >> I have try the same transfert with Linux (on same board) the time is 9s = > >> 89Mbit/s > >> > >> The performance of transfert on OBSD is poor why ? > >> > >> Driver ? > >> Kernel ? > >> Or mistake on my configuration. > >> > >> I need ARM card with 2 ethernets (giga) Port to configure in routing from > >> port 1 to port 2. I hope 200Mbit/s (linux performance) for this routing. > >> > >> Thank for your help. > >> Bernard > >> _________________________________ > >> result of ifconfig > >> # ifconfig cpsw0 > >> cpsw0: flags=28843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST,NOINET6> mtu 1500 > >> lladdr d4:94:a1:84:28:f2 > >> priority: 0 > >> groups: egress > >> media: Ethernet 100baseTX full-duplex > >> status: active > >> inet 192.168.169.62 netmask 0xffffff00 broadcast 192.168.169.255 > >> > >> > >> __________________________________ > >> result of dmessg > >> # dmesg > >> OpenBSD 5.6-current (GENERIC-OMAP) #0: Tue Oct 28 01:01:45 AEDT 2014 > >> j...@armv7.jsg.id.au:/usr/src/sys/arch/armv7/compile/GENERIC-OMAP > >> real mem = 268435456 (256MB) > >> avail mem = 255725568 (243MB) > >> warning: no entropy supplied by boot loader > >> mainbus0 at root > >> cpu0 at mainbus0: ARM Cortex A8 R3 rev 2 (ARMv7 core) > >> cpu0: DC enabled IC enabled WB disabled EABT branch prediction enabled > >> cpu0: 32KB(64b/l,4way) I-cache, 32KB(64b/l,4way) wr-back D-cache > >> omap0 at mainbus0: BeagleBone > >> prcm0 at omap0 rev 0.2 > >> sitaracm0 at omap0: control module, rev 1.0 > >> intc0 at omap0 rev 5.0 > >> edma0 at omap0 rev 0.0 > >> dmtimer0 at omap0 rev 3.1 > >> dmtimer1 at omap0 rev 3.1 > >> omdog0 at omap0 rev 0.1 > >> omgpio0 at omap0: rev 0.1 > >> gpio0 at omgpio0: 32 pins > >> omgpio1 at omap0: rev 0.1 > >> gpio1 at omgpio1: 32 pins > >> omgpio2 at omap0: rev 0.1 > >> gpio2 at omgpio2: 32 pins > >> omgpio3 at omap0: rev 0.1 > >> gpio3 at omgpio3: 32 pins > >> omap0: device tiiic unit 0 not configured > >> omap0: device tiiic unit 1 not configured > >> omap0: device tiiic unit 2 not configured > >> ommmc0 at omap0 > >> sdmmc0 at ommmc0 > >> ommmc1 at omap0 > >> sdmmc1 at ommmc1 > >> com0 at omap0: ti16750, 64 byte fifo > >> com0: console > >> cpsw0 at omap0: version 1.12 (0), address d4:94:a1:84:28:f2 > >> ukphy0 at cpsw0 phy 0: Generic IEEE 802.3u media interface, rev. 4: OUI > >> 0x001374, model 0x0007 > >> ukphy1 at cpsw0 phy 1: Generic IEEE 802.3u media interface, rev. 4: OUI > >> 0x001374, model 0x0007 > >> scsibus0 at sdmmc0: 2 targets, initiator 0 > >> sd0 at scsibus0 targ 1 lun 0: <SD/MMC, Drive #01, > SCSI2 0/direct fixed > >> sd0: 3781MB, 512 bytes/sector, 7744512 sectors > >> sdmmc1: can't enable card > >> vscsi0 at root > >> scsibus1 at vscsi0: 256 targets > >> softraid0 at root > >> scsibus2 at softraid0: 256 targets > >> boot device: sd0 > >> root on sd0a (853680ee25c0e223.a) swap on sd0b dump on sd0b > >> WARNING: CHECK AND RESET THE DATE! > >> > >> > >> __________________________________________ > >> Bernard M. Merindol, Grenoble France > >> bern...@merindol.org GSM: +33-6-08-75-03-52 > >> > > __________________________________________ > Bernard M. Merindol, Grenoble France > bern...@merindol.org GSM: +33-6-08-75-03-52 >
