From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([208.118.235.92]:49941) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1TiTkI-0004dM-B3 for qemu-devel@nongnu.org; Tue, 11 Dec 2012 12:35:59 -0500 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1TiTkF-0007Fc-T1 for qemu-devel@nongnu.org; Tue, 11 Dec 2012 12:35:54 -0500 Date: Tue, 11 Dec 2012 11:35:31 -0600 From: Scott Wood References: <1355186215.5334.21@snotra> In-Reply-To: (from agraf@suse.de on Tue Dec 11 02:10:14 2012) Message-ID: <1355247331.13481.5@snotra> MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii"; delsp=Yes; format=Flowed Content-Disposition: inline Content-Transfer-Encoding: quoted-printable Subject: Re: [Qemu-devel] [Qemu-ppc] [PATCH 16/19] openpic: add Shared MSI support List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Alexander Graf Cc: "qemu-ppc@nongnu.org List" , qemu-devel qemu-devel On 12/11/2012 02:10:14 AM, Alexander Graf wrote: >=20 >=20 > On 11.12.2012, at 01:36, Scott Wood wrote: >=20 > > On 12/08/2012 07:44:39 AM, Alexander Graf wrote: > >> The OpenPIC allows MSI access through shared MSI registers. =20 > Implement > >> them for the MPC8544 MPIC, so we can support MSIs. > >> Signed-off-by: Alexander Graf > >> --- > >> hw/openpic.c | 150 =20 > ++++++++++++++++++++++++++++++++++++++++++++++++++-------- > >> 1 files changed, 130 insertions(+), 20 deletions(-) > >> diff --git a/hw/openpic.c b/hw/openpic.c > >> index f2f152f..f71d668 100644 > >> --- a/hw/openpic.c > >> +++ b/hw/openpic.c > >> @@ -38,6 +38,7 @@ > >> #include "pci.h" > >> #include "openpic.h" > >> #include "sysbus.h" > >> +#include "msi.h" > >> //#define DEBUG_OPENPIC > >> @@ -52,6 +53,7 @@ > >> #define MAX_TMR 4 > >> #define VECTOR_BITS 8 > >> #define MAX_IPI 4 > >> +#define MAX_MSI 8 > >> #define VID 0x03 /* MPIC version ID */ > >> /* OpenPIC capability flags */ > >> @@ -62,6 +64,8 @@ > >> #define OPENPIC_GLB_REG_SIZE 0x10F0 > >> #define OPENPIC_TMR_REG_START 0x10F0 > >> #define OPENPIC_TMR_REG_SIZE 0x220 > >> +#define OPENPIC_MSI_REG_START 0x1600 > >> +#define OPENPIC_MSI_REG_SIZE 0x200 > >> #define OPENPIC_SRC_REG_START 0x10000 > >> #define OPENPIC_SRC_REG_SIZE (MAX_IRQ * 0x20) > >> #define OPENPIC_CPU_REG_START 0x20000 > >> @@ -126,6 +130,12 @@ > >> #define IDR_P1_SHIFT 1 > >> #define IDR_P0_SHIFT 0 > >> +#define MSIIR_OFFSET 0x140 > >> +#define MSIIR_SRS_SHIFT 29 > >> +#define MSIIR_SRS_MASK (0x7 << MSIIR_SRS_SHIFT) > >> +#define MSIIR_IBS_SHIFT 24 > >> +#define MSIIR_IBS_MASK (0x1f << MSIIR_IBS_SHIFT) > > > > FWIW, if you want to model newer MPICs such as on p4080, they have =20 > multiple banks of MSIs, so you may not want to hardcode one bank. >=20 > The OpenPIC code was suffering a lot from attempts to generalize =20 > different implementations without implementing them. >=20 > If we want to add support for p4080 MPICs later, we add a new model =20 > to the emulation and make the nr of msi banks a parameter, like the =20 > patch set does for all the other raven/mpc8544 differences. That way =20 > we don't get into the current mess of a halfway accurate emulation =20 > unless we really want it. So because the old code made a mess of it, we're saying "abstraction is =20 bad" in general? All I'm saying is this should be done with a runtime data structure (of =20 which there could be more than one) rather than #defines. -Scott=