From mboxrd@z Thu Jan 1 00:00:00 1970
From: Pavel Fedin
Subject: Re: [PATCH v3 4/4] target-arm: Add the GICv2m to
the virt board
Date: Tue, 26 May 2015 16:54:17 +0300
Message-ID: <019801d097bb$75764c70$6062e550$@samsung.com>
References: <1432464666-4825-1-git-send-email-christoffer.dall@linaro.org>
<1432464666-4825-5-git-send-email-christoffer.dall@linaro.org>
<55646D0F.9040801@linaro.org>
<5564702F.2030308@linaro.org>
Mime-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: quoted-printable
Return-path:
In-reply-to: <5564702F.2030308@linaro.org>
Content-language: ru
List-Unsubscribe: ,
List-Archive:
List-Post:
List-Help:
List-Subscribe: ,
Errors-To: qemu-devel-bounces+gceq-qemu-devel=gmane.org@nongnu.org
Sender: qemu-devel-bounces+gceq-qemu-devel=gmane.org@nongnu.org
To: 'Eric Auger' , 'Peter Maydell'
Cc: 'QEMU Developers' , 'Christoffer Dall' , kvmarm@lists.cs.columbia.edu
List-Id: kvmarm@lists.cs.columbia.edu
Hi! My word...
> What is your guidance, should we introduce new memory maps for GICv3
> enabled machine or should we move to a single dynamic memory map?
IMHO there's no reason to introduce another memory map. I have already =
done test integration some time ago, and here is what i got:
--- cut ---
enum {
VIRT_FLASH,
VIRT_MEM,
VIRT_CPUPERIPHS,
VIRT_GIC_DIST,
VIRT_GIC_CPU,
VIRT_GIC_DIST_SPI =3D VIRT_GIC_CPU,
VIRT_ITS_CONTROL,
VIRT_ITS_TRANSLATION,
VIRT_LPI,
VIRT_UART,
VIRT_MMIO,
VIRT_RTC,
VIRT_FW_CFG,
VIRT_PCIE,
VIRT_GIC_V2M =3D VIRT_ITS_CONTROL,
};
--- cut ---
static const MemMapEntry a15memmap[] =3D {
/* Space up to 0x8000000 is reserved for a boot ROM */
[VIRT_FLASH] =3D { 0, 0x08000000 },
[VIRT_CPUPERIPHS] =3D { 0x08000000, 0x00020000 },
/* GIC distributor and CPU interfaces sit inside the CPU peripheral =
space */
[VIRT_GIC_DIST] =3D { 0x08000000, 0x00010000 },
[VIRT_GIC_CPU] =3D { 0x08010000, 0x00010000 }, /* =
VIRT_GIC_DIST_SPI for v3 */
[VIRT_ITS_CONTROL] =3D { 0x08020000, 0x0010000 }, /* =
VIRT_GIC_V2M for v2 */
[VIRT_ITS_TRANSLATION] =3D { 0x08030000, 0x00010000 },
[VIRT_LPI] =3D { 0x08040000, 0x00800000 },
[VIRT_UART] =3D { 0x09000000, 0x00001000 },
[VIRT_RTC] =3D { 0x09010000, 0x00001000 },
[VIRT_FW_CFG] =3D { 0x09020000, 0x0000000a },
[VIRT_MMIO] =3D { 0x0a000000, 0x00000200 },
/* ...repeating for a total of NUM_VIRTIO_TRANSPORTS, each of that =
size */
/*
* PCIE verbose map:
*
* MMIO window { 0x10000000, 0x2eff0000 },
* PIO window { 0x3eff0000, 0x00010000 },
* ECAM { 0x3f000000, 0x01000000 },
*/
[VIRT_PCIE] =3D { 0x10000000, 0x30000000 },
[VIRT_MEM] =3D { 0x40000000, 30ULL * 1024 * 1024 * 1024 },
};
--- cut ---
As you can see, it's perfectly readable and memory maps are identical. =
I even see no problem with merging ITS_CONTROL and ITS_TRANSPATION, =
because i don't see any harm in increasing v2m area.
P.S. And yes, VIRT_GIC_DIST_SPI should be VIRT_GIC_DIST_MBI instead =
(Reviewed-by: Eric Auger ), just this fragment is =
from my old integration branch, which i currently don't work on, because =
my new test environment doesn't use GICv2 at all.
Kind regards,
Pavel Fedin
Expert Engineer
Samsung Electronics Research center Russia
From mboxrd@z Thu Jan 1 00:00:00 1970
Received: from eggs.gnu.org ([2001:4830:134:3::10]:43586)
by lists.gnu.org with esmtp (Exim 4.71)
(envelope-from ) id 1YxFJK-0000kV-Da
for qemu-devel@nongnu.org; Tue, 26 May 2015 09:54:27 -0400
Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71)
(envelope-from ) id 1YxFJH-0003Pi-7d
for qemu-devel@nongnu.org; Tue, 26 May 2015 09:54:26 -0400
Received: from mailout4.w1.samsung.com ([210.118.77.14]:37157)
by eggs.gnu.org with esmtp (Exim 4.71)
(envelope-from ) id 1YxFJH-0003Mu-2C
for qemu-devel@nongnu.org; Tue, 26 May 2015 09:54:23 -0400
Received: from eucpsbgm2.samsung.com (unknown [203.254.199.245])
by mailout4.w1.samsung.com
(Oracle Communications Messaging Server 7.0.5.31.0 64bit (built May 5
2014)) with ESMTP id <0NOY00J9UMMIC950@mailout4.w1.samsung.com> for
qemu-devel@nongnu.org; Tue, 26 May 2015 14:54:18 +0100 (BST)
From: Pavel Fedin
References: <1432464666-4825-1-git-send-email-christoffer.dall@linaro.org>
<1432464666-4825-5-git-send-email-christoffer.dall@linaro.org>
<55646D0F.9040801@linaro.org>
<5564702F.2030308@linaro.org>
In-reply-to: <5564702F.2030308@linaro.org>
Date: Tue, 26 May 2015 16:54:17 +0300
Message-id: <019801d097bb$75764c70$6062e550$@samsung.com>
MIME-version: 1.0
Content-type: text/plain; charset=UTF-8
Content-transfer-encoding: quoted-printable
Content-language: ru
Subject: Re: [Qemu-devel] [PATCH v3 4/4] target-arm: Add the GICv2m to
the virt board
List-Id:
List-Unsubscribe: ,
List-Archive:
List-Post:
List-Help:
List-Subscribe: ,
To: 'Eric Auger' , 'Peter Maydell'
Cc: 'QEMU Developers' , 'Christoffer Dall' , kvmarm@lists.cs.columbia.edu
Hi! My word...
> What is your guidance, should we introduce new memory maps for GICv3
> enabled machine or should we move to a single dynamic memory map?
IMHO there's no reason to introduce another memory map. I have already =
done test integration some time ago, and here is what i got:
--- cut ---
enum {
VIRT_FLASH,
VIRT_MEM,
VIRT_CPUPERIPHS,
VIRT_GIC_DIST,
VIRT_GIC_CPU,
VIRT_GIC_DIST_SPI =3D VIRT_GIC_CPU,
VIRT_ITS_CONTROL,
VIRT_ITS_TRANSLATION,
VIRT_LPI,
VIRT_UART,
VIRT_MMIO,
VIRT_RTC,
VIRT_FW_CFG,
VIRT_PCIE,
VIRT_GIC_V2M =3D VIRT_ITS_CONTROL,
};
--- cut ---
static const MemMapEntry a15memmap[] =3D {
/* Space up to 0x8000000 is reserved for a boot ROM */
[VIRT_FLASH] =3D { 0, 0x08000000 },
[VIRT_CPUPERIPHS] =3D { 0x08000000, 0x00020000 },
/* GIC distributor and CPU interfaces sit inside the CPU peripheral =
space */
[VIRT_GIC_DIST] =3D { 0x08000000, 0x00010000 },
[VIRT_GIC_CPU] =3D { 0x08010000, 0x00010000 }, /* =
VIRT_GIC_DIST_SPI for v3 */
[VIRT_ITS_CONTROL] =3D { 0x08020000, 0x0010000 }, /* =
VIRT_GIC_V2M for v2 */
[VIRT_ITS_TRANSLATION] =3D { 0x08030000, 0x00010000 },
[VIRT_LPI] =3D { 0x08040000, 0x00800000 },
[VIRT_UART] =3D { 0x09000000, 0x00001000 },
[VIRT_RTC] =3D { 0x09010000, 0x00001000 },
[VIRT_FW_CFG] =3D { 0x09020000, 0x0000000a },
[VIRT_MMIO] =3D { 0x0a000000, 0x00000200 },
/* ...repeating for a total of NUM_VIRTIO_TRANSPORTS, each of that =
size */
/*
* PCIE verbose map:
*
* MMIO window { 0x10000000, 0x2eff0000 },
* PIO window { 0x3eff0000, 0x00010000 },
* ECAM { 0x3f000000, 0x01000000 },
*/
[VIRT_PCIE] =3D { 0x10000000, 0x30000000 },
[VIRT_MEM] =3D { 0x40000000, 30ULL * 1024 * 1024 * 1024 },
};
--- cut ---
As you can see, it's perfectly readable and memory maps are identical. =
I even see no problem with merging ITS_CONTROL and ITS_TRANSPATION, =
because i don't see any harm in increasing v2m area.
P.S. And yes, VIRT_GIC_DIST_SPI should be VIRT_GIC_DIST_MBI instead =
(Reviewed-by: Eric Auger ), just this fragment is =
from my old integration branch, which i currently don't work on, because =
my new test environment doesn't use GICv2 at all.
Kind regards,
Pavel Fedin
Expert Engineer
Samsung Electronics Research center Russia