Date: Fri, 15 Jun 2018 16:02:23 +0900 (JST) From: Mori Hiroki <yamori813@yahoo.co.jp> To: Adrian Chadd <adrian.chadd@gmail.com>, "josef.lar@gmail.com" <josef.lar@gmail.com> Cc: "freebsd-mips@freebsd.org" <freebsd-mips@freebsd.org> Subject: Re: Routing performance (Buffalo WZR-HP-G301NH) Message-ID: <672991.33614.qm@web101701.mail.ssk.yahoo.co.jp> In-Reply-To: <161784.64603.qm@web101702.mail.ssk.yahoo.co.jp> References: <69b4cfd1-1664-ebbf-5734-7be38a70e938@gmail.com> <CAJ-Vmo=BMTVVcyTpY8NqcsuLKdGtXnQSbfduWqgejdveNtbukw@mail.gmail.com> <161784.64603.qm@web101702.mail.ssk.yahoo.co.jp>
next in thread | previous in thread | raw e-mail | index | archive | help
Hi=0A=0AI remember=A0WZR-HP-AG301H. This target arge hints was very worry.= =0A=0AIn Zrouter=A0boards/Buffalo/WZR-HP-G301NH/board.hints=0A=0A#hint.arge= .0.media=3D1000=A0=0Ahint.arge.0.media=3D100=A0=0Ahint.arge.0.fduplex=3D1= =A0=0Ahint.arge.0.phymask=3D0x0=0A=0A# This is RTL8366RB revision setting= =0Ahint.arge.0.miimode=3D3 =A0 # RGMII=0Ahint.arge.1.miimode=3D3 =A0 # RGMI= I=0Ahint.arge.0.pll_1000=3D0x1f000000=0Ahint.arge.1.pll_1000=3D0x100=0A=0Ah= int.arge.1.phymask=3D0x10=0Ahint.arge.1.mdio=3Dmdioproxy0 =A0 =A0 =A0 =A0 = =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0= =A0 =A0 =A0=0Ahint.arge.1.eeprommac=3D0xbfff120c=0A=0Ahint.rtl8366rb.0.phy= 4cpu=3D1=0A=0AThis setting is work only=A0100baseTX.=0A=0Aifconfig and ethe= rswich command say arge is=A0100baseTX.=0A=0Apll_1000 value get from dd-wrt= .=A0=0A=0A=0AHiroki Mori=0A=0A=0A----- Original Message -----=0A> From: Mor= i Hiroki <yamori813@yahoo.co.jp>=0A> To: Adrian Chadd <adrian.chadd@gmail.c= om>; "josef.lar@gmail.com" <josef.lar@gmail.com>=0A> Cc: "freebsd-mips@free= bsd.org" <freebsd-mips@freebsd.org>=0A> Date: 2018/6/15, Fri 09:09=0A> Subj= ect: Re: Routing performance (Buffalo WZR-HP-G301NH)=0A> =0A> Hi=0A> =0A> I= add iperf3 profile to ZRouter and check network performance.=0A> =0A> This= is result.=0A> =0A> WZR-HP-G300NH=0A> =0A> AR9132=0A> =0A> =0A> [=A0 4] = =A0 0.00-10.04=A0 sec =A0 118 MBytes=A0 98.2 Mbits/sec=A0 =A0 0 =A0 =A0 =A0= =A0 =A0 =A0 sender=0A> [=A0 4] =A0 0.00-10.04=A0 sec =A0 118 MBytes=A0 98.= 2 Mbits/sec=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 receiver=0A> =0A> WZR-HP-AG3= 00H=0A> =0A> AR7161=0A> =0A> =0A> [=A0 4] =A0 0.00-10.03=A0 sec =A0 222 MBy= tes =A0 186 Mbits/sec=A0 =A0 0 =A0 =A0 =A0 =A0 =A0 =A0 sender=0A> [=A0 4] = =A0 0.00-10.03=A0 sec =A0 222 MBytes =A0 186 Mbits/sec=A0 =A0 =A0 =A0 =A0 = =A0 =A0 =A0 =A0 receiver=0A> =0A> I seem=A0AR9132 is slower than=A0AR7161= =0A> =0A> Also you can tcp tuning on sysctl.=0A> =0A> Hiroki Mori=0A> =0A> = =0A> ----- Original Message -----=0A>> From: Adrian Chadd <adrian.chadd@gm= ail.com>=0A>> To: josef.lar@gmail.com=0A>> Cc: "freebsd-mips@freebsd.org.= Robert Pera" =0A> <mips@freebsd.org>=0A>> Date: 2018/6/12, Tue 06:39=0A>>= Subject: Re: Routing performance (Buffalo WZR-HP-G301NH)=0A>> =0A>> Hi!= =0A>> =0A>> Ok, so!=0A>> =0A>> * the seiral console hurts a lot. Don't pr= int stuff on the console=0A>> whilst testing.=0A>> * I've done a writeup = on this when I dug into it a couple years ago.=0A>> https://adrianchadd.bl= ogspot.com/=0A>> =0A>> =0A>> =0A>> -adrian=0A>> =0A>> On Fri, 4 May 2018 = at 02:22, Josef Larsson <josef.lar@gmail.com> =0A> wrote:=0A>>> =0A>>> =A0 = Hi,=0A>>> =0A>>> =A0 I would like to use the Buffalo WZR-HP-G301NH as a rou= ter with pf. I =0A> use=0A>>> =A0 a custom built ZRouter image for this. Th= is works in practice, but I =0A> am=0A>>> =A0 not happy with the performanc= e (I basically can't download faster =0A> than=0A>>> =A0 80 mbit/s).=0A>>> = =0A>>> =A0 I have a couple of concerns:=0A>>> =0A>>> =A0 1. According to ps= aux, there is an interrupt load at about 9 - 10 %=0A>>> =A0 during idle co= nditions. Is this reasonable? It seems high to me.=0A>>> =A0 2. When I am r= unning "pmcstat -TS instructions -w1", there =0A> seems =0A>> to be a=0A>>= > =A0 lot of action going on in uart_ar71xx_probe. Is this reasonable? Ther= e=0A>>> =A0 are some unknown functions, and I am suspect that the function = names =0A> are=0A>>> =A0 not resolved correctly. Also, apb_attach is invoke= d a lot according=0A>>> =A0 pmcstat. While the sampling percentage for this= function is less than =0A> 2=0A>>> =A0 %, I still find it odd that an atta= ch function is invoked at idle...=0A>>> =A0 3. When downloading a large fil= e, the system becomes pretty much=0A>>> =A0 unresponsive. When a "ps aux" c= all actually goes through, =0A> one can =0A>> see=0A>>> =A0 that the CPU i= s busy with ~100 % interrupt handling. This is what I=0A>>> =A0 would like = to profile with pmcstat, but the arge functions do not show=0A>>> =A0 up as= functions with high load, which is not really what I expect at=0A>>> =A0 t= his point, since the interrupt definitely has to do with the LAN =0A> load.= =0A>>> =A0 Any suggestions on how to profile the arge driver?=0A>>> =0A>>> = =A0 Best regards,=0A>>> =0A>>> =A0 Josef Larsson=0A>>> =0A>>> =0A>> ______= _________________________________________=0A>> freebsd-mips@freebsd.org ma= iling list=0A>> https://lists.freebsd.org/mailman/listinfo/freebsd-mips=0A= >> To unsubscribe, send any mail to =0A>> "freebsd-mips-unsubscribe@freeb= sd.org"=0A>> =0A> _______________________________________________=0A> freeb= sd-mips@freebsd.org mailing list=0A> https://lists.freebsd.org/mailman/list= info/freebsd-mips=0A> To unsubscribe, send any mail to =0A> "freebsd-mips-u= nsubscribe@freebsd.org"=0A>
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?672991.33614.qm>