Date: Wed, 08 Oct 2003 16:52:47 -0400 (EDT) From: John Baldwin <jhb@FreeBSD.org> To: David Raistrick <drais@wow.atlasta.net> Cc: support@supermicro.com Subject: RE: SCSI, SMP, and Supermicro, problems! Message-ID: <XFMail.20031008165247.jhb@FreeBSD.org> In-Reply-To: <Pine.BSF.4.50L0.0310081341110.67537-100000@wow.atlasta.net>
next in thread | previous in thread | raw e-mail | index | archive | help
On 08-Oct-2003 David Raistrick wrote: > >> You can try using the attached patch. It removes the HTT option and >> HTT is on by default, but it halts the CPUs the same way that we do >> in current now. > > John, > > Thanks very much! That patch, applied to 4.8-R's /sys, does the trick. > Both local CPUs are disabled, and the system seems stable so far. > > I see you cc'd supermicro support, hopefully they'll be able to add this > to their FAQ...and perhaps they'll be able to find out where the problem > with HTT disabled lies, as well. Oh, didn't mean to. Just did a reply to all. Anyways, the real problem is that many P4 motherboard BIOS's leave the local APIC on the logical CPUs enabled so that they are eligible to receive interrupts. Due to the way we manage the Task Priority Register (TPR) on the CPUs that we know about, the system can try to route interrupts to the logical CPUs that we don't know about and aren't using. Because these CPUs aren't really awake, they don't acknowledge these interrupts and eventually the system locks up. That's the short version anyway. From what I hear, this bug of leaving the local APIC enabled is in the sample BIOS source code that a certain CPU vendor released so just about all the P4 chipsets have it. :) > Once again this is in reference to the Supermicro SuperServer 6023P-8R > with the X5DP8-G2 motherboard and FreeBSD 4.8-RELEASE. -- John Baldwin <jhb@FreeBSD.org> <>< http://www.FreeBSD.org/~jhb/ "Power Users Use the Power to Serve!" - http://www.FreeBSD.org/
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?XFMail.20031008165247.jhb>