From owner-freebsd-smp Fri Sep 28 4:53:14 2001 Delivered-To: freebsd-smp@freebsd.org Received: from digitaldaemon.com (digitaldaemon.com [63.105.9.34]) by hub.freebsd.org (Postfix) with SMTP id EDDD737B407 for ; Fri, 28 Sep 2001 04:53:10 -0700 (PDT) Received: (qmail 58897 invoked from network); 28 Sep 2001 11:51:58 -0000 Received: from 213-84-202-101.adsl.xs4all.nl (HELO jak.nl) (213.84.202.101) by jak.nl with SMTP; 28 Sep 2001 11:51:58 -0000 Message-ID: <3BB464DF.3080404@jak.nl> Date: Fri, 28 Sep 2001 13:54:07 +0200 From: Arjan Knepper Organization: JAK++ Software Development B.V. User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:0.9.2) Gecko/20010726 Netscape6/6.1 X-Accept-Language: en-us MIME-Version: 1.0 To: Bruce Evans Cc: freebsd-smp Subject: Re: Problems with accessing ports on multiple Cyclades YeP cards References: <20010928203144.O52995-100000@delplex.bde.org> Content-Type: text/plain; charset=us-ascii; format=flowed Content-Transfer-Encoding: 7bit Sender: owner-freebsd-smp@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.org Bruce Evans wrote: >On Fri, 28 Sep 2001, Arjan Knepper wrote: > >>I have a big problem with accessing the YeP serial ports in a Dell >>PowerEdge 2550 with Dual Pentium III 933Mhz. >> >>For some reason al YeP boards and serial ports ARE detected, but when I >>try to access the serial ports with cu only one board is accessable, >>trying to connect a port on the other YeP board leaves me with 2 cu >>proccesses both connecting to the same port and one of those processes >>could not be killed. >>... >>Any ideas?? >>Has it something to do with SMP? (We have a single processor PII400 >>Inetl 440 chipsset machines working with 2 Cyclades boards without problems) >>Building a Single processor kernel didn't help it even got worse because >>the machine locks-up in the boot process after detecting the third board >>and ports, diconnecting two of the three cables solved the boot-lockup. >> > >I assume that you have your own fix for the minor numbering problem >(PR 19256) or a version of FreeBSD with the fix committed. > Running 4.4-STABLE > >I suspect a problem with shared interrupts. SMP doesn't do much different >for multiple boards except it should try harder to give each board its >own interrupt, which in theory should work better. Do you use the >CY_PCI_FASTINTR option? > Yes I do. > It should be the default, but... The lockup >for the !SMP case is consistent with the following configuration: >- CY_PCI_FASTINTR not configured >- irq for YeP board shared >- some race in the driver that is more fatal when the irq is shared between > YeP boards. >I always configure CY_PCI_FASTINTR and shuffle boards in the PCI slots >to make the YeP interrupt non-shared. > The interups are not shared between the board but shared with the NIC and SCSI interface (both integrated on motherboard) For me there not much to shuffle since there are only 3 PCI slots in the the machine and all are supossed to have an YeP card. > This is easier with only 1 YeP >board, but it still required lots of shuffling on a BP6 motherboard. >(The BP6 needs an extra irq or two for the second ata controller >(HPTsomething), and its BIOS doesn't support adjusting these like it >does for the PCI slots, and the YeP irq ended up shared with an HPT >irq in most configurations; this just prevented CY_PCI_FASTINTR from >helping under FreeBSD, but it caused the boot to hang under Linux.) > Are you saying that the YeP interups shouldn't be shared at all? Not with ANY pci device? Thanks, Arjan Knepper To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-smp" in the body of the message