From mboxrd@z Thu Jan  1 00:00:00 1970
From: marc.zyngier@arm.com (Marc Zyngier)
Date: Thu, 31 Mar 2016 10:06:01 +0100
Subject: [RFC PATCH 00/45] KVM: arm/arm64: Rework virtual GIC emulation
In-Reply-To: <20160330195555.GL4126@cbox>
References: <1458871508-17279-1-git-send-email-andre.przywara@arm.com>
 <56FA7F29.4060005@arm.com> <56FBBBC2.6010602@arm.com>
 <56FBC16A.7040404@arm.com> <20160330195555.GL4126@cbox>
Message-ID: <56FCE879.2060604@arm.com>
To: linux-arm-kernel@lists.infradead.org
List-Id: linux-arm-kernel.lists.infradead.org

On 30/03/16 20:55, Christoffer Dall wrote:
> On Wed, Mar 30, 2016 at 01:07:06PM +0100, Marc Zyngier wrote:
>> On 30/03/16 12:42, Vladimir Murzin wrote:
>>> On 29/03/16 14:12, Vladimir Murzin wrote:
>>>> Hi Andre,
>>>>
>>>> On 25/03/16 02:04, Andre Przywara wrote:
>>>>> Please have a look at the series, review it and give the code some
>>>>> serious testing (and possibly debugging). All feedback is appreciated.
>>>>
>>>> I see that with the new vgic implementation kvmtool starts throwing:
>>>>
>>>> kvm [1273]: Unable to register VGICv3 redist MMIO regions
>>>>
>>>
>>> It comes from kvm_io_bus_register_dev()
>>>
>>> if (bus->dev_count - bus->ioeventfd_count > NR_IOBUS_DEVS - 1)
>>> 	return -ENOSPC;
>>>
>>> with bus->dev_count being 1000
>>
>> Ouch. That's a consequence of having one "device" per actual register
>> set, I believe. So either we bump this limit up by a factor 10 (at
>> least), or we switch back to the old way of handling access to the GIC
>> regions (big "catch-all", and further demuxing).
>>
>> Thoughts?
>>
> Since we have to have code to iterate through the individual register
> regions for the userland MMIO accesses anyway, I think we should just
> revert back to the old way.

That's what I was thinking as well. We could use a binary search to
speed-up the "in-region" search, just like the kvm_io_bus stuff does.

Thanks,

	M.
-- 
Jazz is not dead. It just smells funny...