[Skiboot] [PATCH] Revert "fsp/console: Allocate irq for each hvc console"
stewart at linux.vnet.ibm.com
Mon Feb 6 15:21:12 AEDT 2017
Alistair Popple <alistair at popple.id.au> writes:
> On Wed, 23 Nov 2016 05:03:22 PM Samuel Mendoza-Jonas wrote:
>> On Wed, 2016-11-23 at 16:52 +1100, Samuel Mendoza-Jonas wrote:
>> > On Tue, 2016-11-22 at 10:49 +0530, Pridhiviraj Paidipeddi wrote:
>> > > This reverts commit 583c8203dcb26b42cea81e4734ea926dae05dbb9.
>> > Stewart, please don't merge this yet.
>> > Without the patch to add irqs for hvc consoles, we'll get the original
>> > lockup problem in Linux 4.2+ that this aimed to solve. We'll need to
>> > work out something to make older kernels happy.
>> Ah as I page this back into my brain we should be saved by commit fd6b71fc
>> which ignores unresponsive consoles (eg. hvc1). Still, without this patch
>> host serial will be unusable on Firenze systems.
>> > >
>> > > Below is the WARNING in pre 4.2 linux kernels which is raised in firenze
>> > > systems due to interrupts mapping failure.
>> > >
>> > > [ 0.947741] irq: irq-62==>hwirq-0x3e mapping failed: -22
>> > > [ 0.947793] ------------[ cut here ]------------
>> > > [ 0.947838] WARNING: at kernel/irq/irqdomain.c:485
>> > >
>> > > So this commit 583c8203dcb26b42cea81e4734ea926dae05dbb9 is causing
>> > > the above kernel WARNING(found by git-bisect).
>> > >
>> > > Signed-off-by: Pridhiviraj Paidipeddi <ppaidipe at linux.vnet.ibm.com>
>> > > ---
>> > > -
>> > > - opal_event = dt_find_by_name(opal_node, "event");
>> > > - if (opal_event)
>> > > - dt_add_property_cells(fs_node, "interrupt-parent",
>> > > - opal_event->phandle);
> I think the correct fix is probably just to revert those lines. That should
> avoid the warning but not result in any functional change, although I will
> have to check all the kernel sides.
Any further thoughts? Or a patch perchance?
OPAL Architect, IBM.
More information about the Skiboot