From owner-freebsd-hardware Tue Jul 25 12: 7:24 2000 Delivered-To: freebsd-hardware@freebsd.org Received: from portnoy.lbl.gov (portnoy.lbl.gov [131.243.2.11]) by hub.freebsd.org (Postfix) with ESMTP id 709D237B81A; Tue, 25 Jul 2000 12:07:14 -0700 (PDT) (envelope-from jin@portnoy.lbl.gov) Received: (from jin@localhost) by portnoy.lbl.gov (8.10.0/8.10.0) id e6PJ7DN26979; Tue, 25 Jul 2000 12:07:13 -0700 (PDT) Date: Tue, 25 Jul 2000 12:07:13 -0700 (PDT) From: Jin Guojun (DSD staff) Message-Id: <200007251907.e6PJ7DN26979@portnoy.lbl.gov> To: bugs@freebsd.org, ken@kdm.org Subject: Re: SysKonnect and Intel gig boards Cc: freebsd-hardware@freebsd.org, jhartley@netrail.net, mjacob@feral.com Sender: owner-freebsd-hardware@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.org The recent two versions of CURRENT are not very usable. CURRENT-20000723 installation floppy is not runnable. So, I backed to previous version CURRENT-20000722 and have a few problems: (1) recompiled kernel is not loadable. I re-config and recompiled several times. (2) The default kernel (installed one) has serious kernel issue -- [109] sleepy.lbl.gov: top top: nlist failed [110] sleepy.lbl.gov: vmstat 1 vmstat: undefined symbols: _cp_time _kmemstatistics _bucket _zlist [111] sleepy.lbl.gov: uname -a FreeBSD sleepy.lbl.gov 5.0-20000722-CURRENT FreeBSD 5.0-20000722-CURRENT #0: Sat Jul 22 12:28:30 GMT 2000 root@usw2.freebsd.org:/usr/src/sys/compile/GENERIC i386 So, there is no progress and result can be reported at this time. I will try next version of CURRENT when it comes out. Thanks, -Jin ------------- original message ----------- On Mon, Jul 10, 2000 at 18:30:11 -0600, Kenneth D. Merry wrote: > On Mon, Jul 10, 2000 at 16:56:16 -0700, Jin Guojun wrote: > > (2) chewing significant CPU: (tested on completely idle machines) > > 500 MHz P-III + single NetGear A620 I/O > > takes 47% CPU time > > > > 750 MHz AMD K7 + single NetGear A620 I/O > > takes less 10% CPU > > > > 500 MHz P-III + single SysKonnect I/O > > takes less 5% CPU > > What benchmark are you running and what size packets are you using? > > I think I mentioned to you before that the driver is probably tuned for > 9000 byte packets, and you might need to tweak things somewhat to get > better performance with 1500 byte packets. Another parameter to try is enabling transmit flow control. The attached patch should do it for the -current driver, I think. I haven't really tested this to see what effect it has, and it'll only work with Tigon II boards. (The Netgear board is a Tigon II board.) Ken -- Kenneth Merry ken@kdm.org --0OAP2g/MAC+5xKAE Content-Type: text/plain; charset=us-ascii Content-Disposition: attachment; filename="if_ti.c.flow_control.20000721" ==== //depot/FreeBSD-current/src/sys/pci/if_ti.c#27 - /a/ken/perforce/FreeBSD-cu rrent/src/sys/pci/if_ti.c ==== *** /tmp/tmp.61653.0 Fri Jul 21 16:50:09 2000 --- /a/ken/perforce/FreeBSD-current/src/sys/pci/if_ti.c Fri Jul 21 16:49:57 2000 *************** *** 2320,2327 **** --- 2320,2329 ---- case IFM_AUTO: CSR_WRITE_4(sc, TI_GCR_GLINK, TI_GLNK_PREF|TI_GLNK_1000MB| TI_GLNK_FULL_DUPLEX|TI_GLNK_RX_FLOWCTL_Y| + TI_GLNK_TX_FLOWCTL_Y| TI_GLNK_AUTONEGENB|TI_GLNK_ENB); CSR_WRITE_4(sc, TI_GCR_LINK, TI_LNK_100MB|TI_LNK_10MB| + TI_LNK_TX_FLOWCTL_Y|TI_LNK_RX_FLOWCTL_Y| TI_LNK_FULL_DUPLEX|TI_LNK_HALF_DUPLEX| TI_LNK_AUTONEGENB|TI_LNK_ENB); TI_DO_CMD(TI_CMD_LINK_NEGOTIATION, To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hardware" in the body of the message