Date: Mon, 24 Feb 2014 22:30:20 +0000 From: Karl Pielorz <kpielorz_lst@tdx.co.uk> To: =?UTF-8?Q?Roger_Pau_Monn=C3=A9?= <roger.pau@citrix.com>, freebsd-xen@freebsd.org Subject: Re: FreeBSD 10-R 8 vCPU panics at boot under XenServer (on 8 'core' CPU) Message-ID: <AAC660DF6ABDD5055357E419@study64.tdx.co.uk> In-Reply-To: <530B7F2F.2010908@citrix.com> References: <6A3B878077F7D071847052C0@Mail-PC.tdx.co.uk> <5302311E.2040700@citrix.com> <1740E0FEE5963358491F4B37@study64.tdx.co.uk> <530B7F2F.2010908@citrix.com>
next in thread | previous in thread | raw e-mail | index | archive | help
--On 24 February 2014 18:19:43 +0100 Roger Pau Monn=C3=A9=20 <roger.pau@citrix.com> wrote: > I've passed through a dual port BCE card (Broadcom NetXtreme II BCM5709) > without problems. As a test, could you try to only pass the nic or the > SAS controller to see if we can figure out if this is specific to one of > the devices? Ok, I tried several boots just passing through the LSI - all boots=20 succeeded. There's quite a variance between how long it takes to 'launch'=20 the CPU's. On a good run you'll get 'Netvsc initializing' small pause, then = Launches 5 CPU's, another small pause - and the remaining 2 launch. On a bad boot you'll get quite a long pause before the first launch, then a = few more pop in - then a very long pause before the final one(s) launch. I then changed to just passing through the NIC's - you get similar results=20 - it's markedly slower than if nothing is passed through, but not so slow=20 it fails. It's only if you pass through the NIC's and the LSI - it's *really* slow -=20 to the point that without upping the NUM_RETRIES in the patch you did it=20 often panics (unable to schedule timer). Some of these boots can take=20 minutes (but, shut down the VM and try it again and it'll complete=20 "relatively quickly" - i.e. 20-30 seconds). I can probably video some of the boots and give you a link off list to the=20 footage (that'd probably give you a better idea of the timing involved). I don't have the other machine running at the moment - so I can't do any=20 testing of what performance you actually get from the passed through=20 devices - they 'seem' to work ok from the limited testing I've done. -Karl
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?AAC660DF6ABDD5055357E419>