From owner-freebsd-hardware Tue Sep 9 10:26:47 1997 Return-Path: Received: (from root@localhost) by hub.freebsd.org (8.8.7/8.8.7) id KAA21823 for hardware-outgoing; Tue, 9 Sep 1997 10:26:47 -0700 (PDT) Received: from biggusdiskus.flyingfox.com (biggusdiskus.flyingfox.com [206.14.52.27]) by hub.freebsd.org (8.8.7/8.8.7) with ESMTP id KAA21809 for ; Tue, 9 Sep 1997 10:26:42 -0700 (PDT) Received: (from jas@localhost) by biggusdiskus.flyingfox.com (8.8.5/8.8.5) id KAA02502; Tue, 9 Sep 1997 10:23:11 -0700 (PDT) Date: Tue, 9 Sep 1997 10:23:11 -0700 (PDT) From: Jim Shankland Message-Id: <199709091723.KAA02502@biggusdiskus.flyingfox.com> To: jkh@time.cdrom.com Subject: Re: four-port ethernet - adaptec ANA-6944A/TX Cc: freebsd-hardware@FreeBSD.ORG Sender: owner-freebsd-hardware@FreeBSD.ORG X-Loop: FreeBSD.org Precedence: bulk Jordan writes: > http://www.3am-software.com/de-970703.tar.gz > > We're trying to evaluate that driver now for possible inclusion > with the next FreeBSD release(s). Note that that version of the driver has a fairly serious bug in FreeBSD: it mallocs the tulip_softc structure, which has grown to > 4 KB in size. At the end of the tulip_softc structure are arrays of receive and transmit descriptors that must be physically contiguous. Since the structure spans a page boundary, they may not turn out to be contiguous. Matt's suggested solution was to reduce TULIP_TXDESCS (the number of transmit descriptors) from 128 to 32; this shrinks the tulip_softc structure back under 4 KB. Alternatively, the receive and transmit descriptor arrays can be allocated separately, rather then embedded in the tulip_softc structure. As long as the tulip_softc structure happens to be mapped to physically contiguous memory, everything works. But completely unrelated kernel changes can cause this to break; symptoms are that the driver wedges after transmitting about 100 packets. Jim Shankland