From mboxrd@z Thu Jan  1 00:00:00 1970
Received: from eggs.gnu.org ([2001:4830:134:3::10]:59487)
	by lists.gnu.org with esmtp (Exim 4.71)
	(envelope-from <benh@kernel.crashing.org>) id 1VdHpC-0005Zy-QI
	for qemu-devel@nongnu.org; Mon, 04 Nov 2013 05:56:07 -0500
Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71)
	(envelope-from <benh@kernel.crashing.org>) id 1VdHp8-0003Rr-BR
	for qemu-devel@nongnu.org; Mon, 04 Nov 2013 05:56:02 -0500
Message-ID: <1383562538.4776.52.camel@pasglop>
From: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Date: Mon, 04 Nov 2013 21:55:38 +1100
In-Reply-To: <00DBC3F8-D221-4E59-B0D0-38E1A4C015CC@suse.de>
References: <1383301276-29566-1-git-send-email-aik@ozlabs.ru>
	<00DBC3F8-D221-4E59-B0D0-38E1A4C015CC@suse.de>
Content-Type: text/plain; charset="UTF-8"
Mime-Version: 1.0
Content-Transfer-Encoding: 7bit
Subject: Re: [Qemu-devel] [PATCH] spapr: make sure RMA is in first mode of
 first memory node
List-Id: <qemu-devel.nongnu.org>
List-Unsubscribe: <https://lists.nongnu.org/mailman/options/qemu-devel>,
	<mailto:qemu-devel-request@nongnu.org?subject=unsubscribe>
List-Archive: <http://lists.nongnu.org/archive/html/qemu-devel>
List-Post: <mailto:qemu-devel@nongnu.org>
List-Help: <mailto:qemu-devel-request@nongnu.org?subject=help>
List-Subscribe: <https://lists.nongnu.org/mailman/listinfo/qemu-devel>,
	<mailto:qemu-devel-request@nongnu.org?subject=subscribe>
To: Alexander Graf <agraf@suse.de>
Cc: Alexey Kardashevskiy <aik@ozlabs.ru>, "list@suse.de:PReP" <qemu-ppc@nongnu.org>, QEMU Developers <qemu-devel@nongnu.org>, Nikunj A Dadhania <nikunj@linux.vnet.ibm.com>

On Mon, 2013-11-04 at 11:44 +0100, Alexander Graf wrote:
> On 01.11.2013, at 11:21, Alexey Kardashevskiy <aik@ozlabs.ru> wrote:
> 
> > SLOF gets really confused if RTAS/device-tree and everything else
> > what SLOF can use is not in the very first block of the very first
> > memory node.
> > 
> > This makes sure that the RMA area is where SLOF expects it to be.
> > 
> > Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
> > Cc: Nikunj A Dadhania <nikunj@linux.vnet.ibm.com>
> > Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
> > ---
> > hw/ppc/spapr.c | 8 +++++++-
> > 1 file changed, 7 insertions(+), 1 deletion(-)
> > 
> > diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
> > index 09dc635..09a5d94 100644
> > --- a/hw/ppc/spapr.c
> > +++ b/hw/ppc/spapr.c
> > @@ -1113,7 +1113,7 @@ static void ppc_spapr_init(QEMUMachineInitArgs *args)
> >     int i;
> >     MemoryRegion *sysmem = get_system_memory();
> >     MemoryRegion *ram = g_new(MemoryRegion, 1);
> > -    hwaddr rma_alloc_size;
> > +    hwaddr rma_alloc_size, node0_size;
> >     uint32_t initrd_base = 0;
> >     long kernel_size = 0, initrd_size = 0;
> >     long load_limit, rtas_limit, fw_size;
> > @@ -1154,6 +1154,12 @@ static void ppc_spapr_init(QEMUMachineInitArgs *args)
> >             spapr->rma_size = MIN(spapr->rma_size, 0x10000000);
> >         }
> >     }
> > +    /*
> > +     * SLOF gets confused if RMA resides not in the first block
> > +     * of the first memory node so let's fix it.
> > +     */
> > +    node0_size = (nb_numa_nodes > 1) ? node_mem[0] : ram_size;
> > +    spapr->rma_size = MIN(spapr->rma_size, node0_size);
> 
> So if I create a NUMA node of 4MB that will be my RMA? That sounds pretty broken, especially on 970.
> 
> Why does SLOF have any issues with NUMA memory nodes? It can just ignore them, no?

Because the only way SLOF knows about the RMA is by using the first
"reg" entry of the first memory node and that's *all* SLOF knows about.

If we start putting things like the DT, SLOF itself, etc... outside of
that region, it will crash.

So we "constrain" things to the rma that way.

Creating 4M nodes makes no sense anyway

Cheers,
Ben.