vGICv3 support
v7 cover vGICv3 support
This series introduces support for GICv3 by KVM. Software emulation is currently not supported.
Differences from v6:
- Wrap own GIC type definitions on top of KVM ones. Fixed build on non-ARM-Linux hosts
Differences from v5:
- Fixed various checkpatch.pl style warnings
- Removed TODO in gicv3_init_irqs_and_mmio(), relevant memory API patch included
- gicv3_init_irqs_and_mmio() now takes 3 arguments instead of 4. It is more convenient to pass MMIO descriptors as array
Differences from v4:
- Do not reintroduce several constants shared with GICv2, reuse them instead.
- Added gicv3_init_irqs_and_mmio() in base class, to be used by both software emulation and KVM code. Avoids code duplication.
- Do not add NULL msi-parent phandle to PCI device in the FDT
- Removed a couple of stale things from virt.c
Differences from v3:
- Fixed stupid build breakage in patch 0002
- Rebased on top of current master, patch 0003 adjusted according to kvm_irqchip_create() changes
- Added assertion against uninitialized kernel_irqchip_type
- Removed kernel_irqchip_type initialization from models which do not use KVM vGIC
Differences from v2:
- Removed some unrelated and unnecessary changes from virt machine, occasionally slipped in; some of them caused qemu to crash on ARM32.
- Fixed build for ARM32; vGICv3 code requires definitions which are present only in ARM64 kernel
Differences from v1:
- Base class included, taken from the series by Shlomo Pongratz:
http://lists.nongnu.org/archive/html/qemu-devel/2015-06/msg01512.html
The code is refactored as little as possible in order to simplify further addition of software emulation:
- Minor fixes in code style and comments, according to old reviews
- Removed REV_V3 definition because it's currently not used, and it does not add any meaning to number 3.
- Removed reserved regions for MBI and ITS (except for 'virt' machine memory map). These should go to separate classes when implemented.
- Improved commit messages
- vGIC patches restructured
- Use 'gicversion' option instead of virt-v3 machine
Pavel Fedin (5):
Merge memory_region_init_reservation() into memory_region_init_io() Extract some reusable vGIC code
Introduce irqchip type specification for KVM
Initial implementation of vGICv3
Add gicversion option to virt machine
Shlomo Pongratz (1):
Implement GIC-500 base class
hw/arm/vexpress.c | 1 + hw/arm/virt.c | 140 +++++++++++++++++---- hw/intc/Makefile.objs | 4 + hw/intc/arm_gic_kvm.c | 84 ++++++------- hw/intc/arm_gicv3_common.c | 249 +++++++++++++++++++++++++++++++++++++ hw/intc/arm_gicv3_kvm.c | 155 +++++++++++++++++++++++ hw/intc/gicv3_internal.h | 156 +++++++++++++++++++++++ hw/intc/vgic_common.h | 43 +++++++ include/exec/memory.h | 14 ++- include/hw/arm/fdt.h | 2 +- include/hw/arm/virt.h | 6 +- include/hw/boards.h | 1 + include/hw/intc/arm_gicv3_common.h | 112 +++++++++++++++++ include/sysemu/kvm.h | 3 +- kvm-all.c | 2 +- memory.c | 10 +- stubs/kvm.c | 2 +- target-arm/kvm-consts.h | 6 + target-arm/kvm.c | 13 +- 19 files changed, 917 insertions(+), 86 deletions(-) create mode 100644 hw/intc/arm_gicv3_common.c create mode 100644 hw/intc/arm_gicv3_kvm.c create mode 100644 hw/intc/gicv3_internal.h create mode 100644 hw/intc/vgic_common.h create mode 100644 include/hw/intc/arm_gicv3_common.h
Signature · 1 line
1.9.5.msysgit.0
PATCH v7 1/6 Merge memory_region_init_reservation() into memory_region_init_io()
Just speficying ops = NULL in some cases can be more convenient than having two functions. GICv3 code is going to use this. GICv2 code can be refactored in a similar way, killing some code duplication.
--- include/exec/memory.h | 14 +++++++++++--- memory.c | 10 +--------- 2 files changed, 12 insertions(+), 12 deletions(-)
include/exec/memory.h +11 -3
@@ -436,6 +436,9 @@ void memory_region_init_alias(MemoryRegion *mr,* memory_region_init_rom_device: Initialize a ROM memory region. Writes are * handled via callbacks. * +* If NULL callbacks pointer is given, then I/O space is not supposed to be +* handled by QEMU itself. Any access via the memory API will cause an abort(). +* * @mr: the #MemoryRegion to be initialized. * @owner: the object that tracks the region's reference count * @ops: callbacks for write access handling.
@@ -458,16 +461,21 @@ void memory_region_init_rom_device(MemoryRegion *mr,* A reservation region primariy serves debugging purposes. It claims I/O * space that is not supposed to be handled by QEMU itself. Any access via * the memory API will cause an abort(). +* This function is deprecated. Use memory_region_init_io() with NULL +* callbacks instead. * * @mr: the #MemoryRegion to be initialized * @owner: the object that tracks the region's reference count * @name: used for debugging; not visible to the user or ABI * @size: size of the region. */ -void memory_region_init_reservation (MemoryRegion *mr, -struct Object *owner, +static inline void memory_region_init_reservation (MemoryRegion *mr, +Object *owner,const char *name, -uint64_t size); +uint64_t size) +{ +memory_region_init_io (mr, owner,NULL , mr, name, size); +}/** * memory_region_init_iommu: Initialize a memory region that translates
memory.c +1 -9
@@ -1187,7 +1187,7 @@ void memory_region_init_io(MemoryRegion *mr, uint64_t size) {memory_region_init (mr, owner, name, size); - mr->ops = ops; + mr->ops = ops ? ops : &unassigned_mem_ops; mr->opaque = opaque; mr->terminates = true; }
@@ -1307,14 +1307,6 @@ void memory_region_init_iommu(MemoryRegion *mr,notifier_list_init (&mr->iommu_notify); } -void memory_region_init_reservation (MemoryRegion *mr, -Object *owner, -const char *name, -uint64_t size) -{ -memory_region_init_io (mr, owner, &unassigned_mem_ops, mr, name, size); -} -static void memory_region_finalize (Object *obj) {MemoryRegion *mr =MEMORY_REGION (obj);
Signature · 1 line
1.9.5.msysgit.0
Re: [Qemu-devel] [PATCH v7 1/6] Merge memory_region_init_reservation() into memory_region_init_io()
Quoted 1 line · Pavel Fedin
Just speficying ops = NULL in some cases can be more convenient than having
"specifying"
Quoted 15 lines
two functions. GICv3 code is going to use this. GICv2 code can be refactored in a similar way, killing some code duplication.
Signed-off-by Pavel Fedin <[email protected]>Acked-by Paolo Bonzini <[email protected]>--- include/exec/memory.h | 14 +++++++++++--- memory.c | 10 +--------- 2 files changed, 12 insertions(+), 12 deletions(-)include/exec/memory.h +0
-0
include/exec/memory.h void memory_region_init_alias(MemoryRegion *mr, · line 440
Quoted 5 lines
include/exec/memory.h +2
-0@@ -436,6 +436,9 @@ void memory_region_init_alias(MemoryRegion *mr,* memory_region_init_rom_device: Initialize a ROM memory region. Writes are * handled via callbacks. * +* If NULL callbacks pointer is given, then I/O space is not supposed to be +* handled by QEMU itself. Any access via the memory API will cause an abort().
I know the memory_region_init_reservation() documentation says this, but is it true? Looking at the memory.c and exec.c code it seems like we'd return MEMTX_DECODE_ERROR for callers who care about memory transaction statuses, behave as RAZ/WI for the bulk of callers that don't, and maybe print some debug logging, but I can't see a relevant call to abort()...
Paolo?
(For the avoidance of doubt, if this is an error in the doc text I don't insist on it being fixed before this patch is applied.)
thanks
-- PMM
PATCH v7 2/6 Implement GIC-500 base class
In reply to Pavel FedinFrom: Shlomo Pongratz <[email protected]>
This class is to be used by both software and KVM implementations of GICv3
--- hw/intc/Makefile.objs | 1 + hw/intc/arm_gicv3_common.c | 249 +++++++++++++++++++++++++++++++++++++ hw/intc/gicv3_internal.h | 156 +++++++++++++++++++++++ include/hw/intc/arm_gicv3_common.h | 112 +++++++++++++++++ 4 files changed, 518 insertions(+) create mode 100644 hw/intc/arm_gicv3_common.c create mode 100644 hw/intc/gicv3_internal.h create mode 100644 include/hw/intc/arm_gicv3_common.h
hw/intc/Makefile.objs +1 -0
@@ -12,6 +12,7 @@ common-obj-$(CONFIG_IOAPIC) += ioapic_common.o common-obj-$(CONFIG_ARM_GIC) += arm_gic_common.o common-obj-$(CONFIG_ARM_GIC) += arm_gic.o common-obj-$(CONFIG_ARM_GIC) += arm_gicv2m.o +common-obj-$(CONFIG_ARM_GIC) += arm_gicv3_common.o common-obj-$(CONFIG_OPENPIC) += openpic.o obj-$(CONFIG_APIC) += apic.o apic_common.o
hw/intc/arm_gicv3_common.c new file +249 -0
249 lines not on this page · raw message
hw/intc/gicv3_internal.h new file +156 -0
156 lines not on this page · raw message
include/hw/intc/arm_gicv3_common.h new file +112 -0
112 lines not on this page · raw message
Signature · 1 line
1.9.5.msysgit.0
Re: [Qemu-devel] [PATCH v7 2/6] Implement GIC-500 base class
Quoted 27 lines · Pavel Fedin
From: Shlomo Pongratz <[email protected]>
This class is to be used by both software and KVM implementations of GICv3
+++
b/include/hw/intc/arm_gicv3_common.hinclude/hw/intc/arm_gicv3_common.h new file +20
-0@@ -0,0 +1,112 @@ +/* +* ARM GIC support +* +* Copyright (c) 2012 Linaro Limited +* Copyright (c) 2015 Huawei. +* Written by Peter Maydell +* Extended to 64 cores by Shlomo Pongratz +* +* This program is free software; you can redistribute it and/or modify +* it under the terms of the GNU General Public License as published by +* the Free Software Foundation, either version 2 of the License, or +* (at your option) any later version. +* +* This program is distributed in the hope that it will be useful, +* but WITHOUT ANY WARRANTY; without even the implied warranty of +* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +* GNU General Public License for more details. +* +* You should have received a copy of the GNU General Public License along +* with this program; if not, see <http://www.gnu.org/licenses/>.
include/hw/intc/arm_gicv3_common.h line 30
Quoted 10 lines
include/hw/intc/arm_gicv3_common.h new file +10
-0+*/ + +#ifndef HW_ARM_GICV3_COMMON_H +#define HW_ARM_GICV3_COMMON_H + +#include "hw/sysbus.h" +#include "hw/intc/arm_gic_common.h" + +/* Maximum number of possible interrupts, determined by the GIC architecture */ +#define GICV3_MAXIRQ 1020
So how do LPIs work? They have IDs above 1023.
include/hw/intc/arm_gicv3_common.h line 31
Quoted 1 line
include/hw/intc/arm_gicv3_common.h new file +1
-0+#define GICV3_NCPU 64
Where does '64' come from as a maximum limit?
include/hw/intc/arm_gicv3_common.h line 39
Quoted 8 lines
include/hw/intc/arm_gicv3_common.h new file +8
-0+ +typedef struct gicv3_irq_state { + /* The enable bits are only banked for per-cpu interrupts. */ + uint64_t enabled; + uint64_t pending; + uint64_t active; + uint64_t level; + uint64_t group;
Why are these uint64_t ?
Quoted 11 lines
include/hw/intc/arm_gicv3_common.h new file +11
-0+bool edge_trigger;/* true: edge-triggered, false: level-triggered */ +} gicv3_irq_state; + +typedef struct gicv3_sgi_state { +uint64_t pending[GICV3_NCPU ]; +}gicv3_sgi_state ; + +typedef struct GICv3State { + /*< private >*/ + SysBusDevice parent_obj; + /*< public >*/
include/hw/intc/arm_gicv3_common.h line 60
Quoted 10 lines
include/hw/intc/arm_gicv3_common.h new file +10
-0+ +qemu_irq parent_irq[GICV3_NCPU ]; +qemu_irq parent_fiq[GICV3_NCPU ]; +/* GICD_CTLR; for a GIC with the security extensions the NS banked version +* of this register is just an alias of bit 1 of the S banked version. +*/ +uint32_t ctlr; +/* Sim GICC_CTLR; again, the NS banked version is just aliases of bits of +* the S banked register, so our state only needs to store the S version. +*/
"Sim" ?
include/hw/intc/arm_gicv3_common.h line 73
Quoted 13 lines
include/hw/intc/arm_gicv3_common.h new file +13
-0+uint32_t cpu_ctlr[GICV3_NCPU ]; +bool cpu_enabled[GICV3_NCPU ]; + +gicv3_irq_state irq_state[GICV3_MAXIRQ ]; +uint64_t irq_target[GICV3_MAXIRQ ]; +uint8_t priority1[GIC_INTERNAL ][GICV3_NCPU ]; +uint8_t priority2[GICV3_MAXIRQ -GIC_INTERNAL ]; +uint16_t last_active[GICV3_MAXIRQ ][GICV3_NCPU ]; +/* For each SGI on the target CPU, we store 64 bits +* indicating which source CPUs have made this SGI +* pending on the target CPU. These correspond to +* the bytes in the GIC_SPENDSGIR* registers as +* read by the target CPU.
This comment doesn't make any sense. You can't fit 64 bits into a byte, and GIC_SPENDSGIR<n> only hold 8 set-pending bits per SGI.
Quoted 7 lines
include/hw/intc/arm_gicv3_common.h new file +7
-0+*/ +gicv3_sgi_state sgi_state[GIC_NR_SGIS ]; + +uint16_t priority_mask[GICV3_NCPU ]; +uint16_t running_irq[GICV3_NCPU ]; +uint16_t running_priority[GICV3_NCPU ]; +uint16_t current_pending[GICV3_NCPU ];
include/hw/intc/arm_gicv3_common.h line 90
Quoted 10 lines
include/hw/intc/arm_gicv3_common.h new file +10
-0+ +uint32_t cpu_mask;/* For redistributor */ +uint32_t num_cpu; +MemoryRegion iomem_dist;/* Distributor */ +MemoryRegion iomem_lpi;/* Redistributor */ +uint32_t num_irq; +uint32_t revision; +bool security_levels; +int dev_fd;/* kvm device fd if backed by kvm vgic support */ +} GICv3State;
This whole struct reads like "we just took the GICv2 state and changed it as little as possible beyond bumping the NCPU define a bit". That doesn't make me very confident that it's actually correct for GICv3...
thanks
-- PMM
Re: [Qemu-devel] [PATCH v7 2/6] Implement GIC-500 base class
Hello!
Quoted 1 line
So how do LPIs work? They have IDs above 1023.
Currently we don't have LPIs. Shlomo's SW emulation did not include them.
Quoted 3 lines
Quoted 1 line
+#define
GICV3_NCPU64Where does '64' come from as a maximum limit?
We don't use Aff2 field as far as i know. It's Shlomo's limitation, not my one.
Quoted 4 lines
This whole struct reads like "we just took the GICv2 state and changed it as little as possible beyond bumping the NCPU define a bit". That doesn't make me very confident that it's actually correct for GICv3...
At least, with complete Shlomo's SW emulation code, it works.
And could you be not so strict on this file? It's actually Shlomo's work, not my one, i include it only because i need to base on something. He will fix up the rest of SW emulation stuff.
Kind regards,
Pavel Fedin
Expert Engineer
Samsung Electronics Research center Russia
Re: [Qemu-devel] [PATCH v7 2/6] Implement GIC-500 base class
Quoted 9 lines · Pavel Fedin
Quoted 4 lines
This whole struct reads like "we just took the GICv2 state and changed it as little as possible beyond bumping the NCPU define a bit". That doesn't make me very confident that it's actually correct for GICv3...
At least, with complete Shlomo's SW emulation code, it works. And could you be not so strict on this file? It's actually Shlomo's
work, not my one, i include it only because i need to base on something. He will fix up the rest of SW emulation stuff.
Data structures are important -- they're the foundation on which everything else rests (and at some point they get baked into VM migration formats, though we aren't at that point yet). It's worth getting them right from the start.
thanks
-- PMM
Re: [Qemu-devel] [PATCH v7 2/6] Implement GIC-500 base class
In reply to Pavel FedinQuoted 1 line · Pavel Fedin
And could you be not so strict on this file?
OK, so having thought about this this week, this is what I suggest.
For this patch series we should retain this class hierarchy, but remove all the fields corresponding to state information which isn't used by the KVM subclass (and all the accessor macro definitions, and that GICV3_NCPU definition, and so on). We should have a migration information struct which says "not migratable".
That gives us the right structure into which we can add the correct state fields when we get the design right (and for whichever of emulation and KVM state save/load gets there first).
We do need to make sure we get right:
* the class hierarchy
* the API to the rest of QEMU (meaning of GPIO pins, memory regions, any qdev/qom properties)
(and in the other parts of the patchset, also that we get the right memory map for the virt board with the GICv3 in it and that the mechanism for selecting GICv2 or GICv3 or "whatever you've got" is good.)
thanks
-- PMM
Re: [Qemu-devel] [PATCH v7 2/6] Implement GIC-500 base class
Hello!
Quoted 6 lines
For this patch series we should retain this class hierarchy, but remove all the fields corresponding to state information which isn't used by the KVM subclass (and all the accessor macro definitions, and that
GICV3_NCPUdefinition, and so on). We should have a migration information struct which says "not migratable".
Ok. I am wating for my GICv2 refactor review which includes memory API change.
Kind regards,
Pavel Fedin
Expert Engineer
Samsung Electronics Research center Russia
Re: [Qemu-devel] [PATCH v7 2/6] Implement GIC-500 base class
In reply to Peter MaydellHello!
Quoted 10 lines
include/hw/intc/arm_gicv3_common.h line 39
Quoted 8 lines
include/hw/intc/arm_gicv3_common.h new file +8
-0+ +typedef struct gicv3_irq_state { + /* The enable bits are only banked for per-cpu interrupts. */ + uint64_t enabled; + uint64_t pending; + uint64_t active; + uint64_t level; + uint64_t group;Why are these
uint64_t?
I studied the code for a little bit more. These are bitmasks, where bit represents a CPU. Take a look at gicv3_internal.h, GIC_SET/CLEAR/TEST_xxx macros. 'cm' stands for 'cpu mask'. This is currently not used by my code, this is mainly for Shlomo's SW emulation. However, there is some usage in common class initialization code.
And i will fix up all badly understood comments.
Kind regards,
Pavel Fedin
Expert Engineer
Samsung Electronics Research center Russia
Re: [Qemu-devel] [PATCH v7 2/6] Implement GIC-500 base class
In reply to Peter MaydellHi,
See inline.
Quoted 41 lines · Peter Maydell
Quoted 3 lines · On 24 July 2015 at 10:55, Pavel Fedin <[email protected] <javascript:;>>…
From: Shlomo Pongratz <[email protected] <javascript:;>>
This class is to be used by both software and KVM implementations of
GICv3
Quoted 13 lines
+++
b/include/hw/intc/arm_gicv3_common.hinclude/hw/intc/arm_gicv3_common.h new file +9
-0@@ -0,0 +1,112 @@ +/* +* ARM GIC support +* +* Copyright (c) 2012 Linaro Limited +* Copyright (c) 2015 Huawei. +* Written by Peter Maydell +* Extended to 64 cores by Shlomo Pongratz +* +* This program is free software; you can redistribute it and/or modify include/hw/intc/arm_gicv3_common.h line 19
Quoted 10 lines
include/hw/intc/arm_gicv3_common.h new file +10
-0+* it under the terms of the GNU General Public License as published by +* the Free Software Foundation, either version 2 of the License, or +* (at your option) any later version. +* +* This program is distributed in the hope that it will be useful, +* but WITHOUT ANY WARRANTY; without even the implied warranty of +* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +* GNU General Public License for more details. +* +* You should have received a copy of the GNU General Public License along
include/hw/intc/arm_gicv3_common.h
Quoted 10 lines
include/hw/intc/arm_gicv3_common.h new file +10
-0+* with this program; if not, see <http://www.gnu.org/licenses/>. +*/ + +#ifndef HW_ARM_GICV3_COMMON_H +#define HW_ARM_GICV3_COMMON_H + +#include "hw/sysbus.h" +#include "hw/intc/arm_gic_common.h" + +/* Maximum number of possible interrupts, determined by the GIC architecture */
Quoted 4 lines
include/hw/intc/arm_gicv3_common.h line 30
Quoted 1 line
include/hw/intc/arm_gicv3_common.h new file +1
-0+#define GICV3_MAXIRQ 1020So how do LPIs work? They have IDs above 1023.
Currently I don't use LPI as the virt virtual machine doesn't use it. There is no point to write something I can't test.
Quoted 5 lines
include/hw/intc/arm_gicv3_common.h line 31
Quoted 2 lines
include/hw/intc/arm_gicv3_common.h new file +1
-0+#define GICV3_NCPU 64Where does '64' come from as a maximum limit?
The goal is 128 (the GIC500 limitation) but since the largest builtin is uint64_t this will require the usage of bitfiled library. Therefore I thought it should be delayed to future version after the logic part is debugged.
Quoted 12 lines
include/hw/intc/arm_gicv3_common.h line 39
Quoted 9 lines
include/hw/intc/arm_gicv3_common.h new file +8
-0+ +typedef struct gicv3_irq_state { + /* The enable bits are only banked for per-cpu interrupts. */ + uint64_t enabled; + uint64_t pending; + uint64_t active; + uint64_t level; + uint64_t group;Why are these
uint64_t?
These are bitfields. which were enlarged from 8 (old limit) to 64 (new limit)
Quoted 23 lines
Review line 40
Quoted 2 lines
(quoted diff) +1
-0+ bool edge_trigger; /* true: edge-triggered, false: level-triggered*/
Quoted 4 lines
include/hw/intc/arm_gicv3_common.h new file +4
-0+} gicv3_irq_state; + +typedef struct gicv3_sgi_state { + uint64_t pending[GICV3_NCPU];include/hw/intc/arm_gicv3_common.h
Quoted 10 lines
include/hw/intc/arm_gicv3_common.h new file +10
-0+} gicv3_sgi_state; + +typedef struct GICv3State { + /*< private >*/ + SysBusDevice parent_obj; + /*< public >*/ + + qemu_irq parent_irq[GICV3_NCPU]; + qemu_irq parent_fiq[GICV3_NCPU]; + /* GICD_CTLR; for a GIC with the security extensions the NS bankedversion
Review line 55
Quoted 1 line
(quoted diff) +1
-0+ * of this register is just an alias of bit 1 of the S bankedversion.
include/hw/intc/arm_gicv3_common.h
Quoted 3 lines
include/hw/intc/arm_gicv3_common.h new file +3
-0+*/ +uint32_t ctlr; +/* Sim GICC_CTLR; again, the NS banked version is just aliases of
Quoted 7 lines
bits of
Review line 59
Quoted 1 line
(quoted diff) +1
-0+ * the S banked register, so our state only needs to store the Sversion.
Review line 60
Quoted 1 line
(quoted diff) +1
-0+ */"Sim" ?
Should be simulate as I only support system registers and not memory mapped registers, but keeping this variable makes the code more close to GICv2 code.
Quoted 15 lines
Quoted 4 lines
include/hw/intc/arm_gicv3_common.h new file +3
-0+uint32_t cpu_ctlr[GICV3_NCPU ]; +bool cpu_enabled[GICV3_NCPU ]; +include/hw/intc/arm_gicv3_common.h line 73
Quoted 10 lines
include/hw/intc/arm_gicv3_common.h new file +10
-0+gicv3_irq_state irq_state[GICV3_MAXIRQ ]; +uint64_t irq_target[GICV3_MAXIRQ ]; +uint8_t priority1[GIC_INTERNAL ][GICV3_NCPU ]; +uint8_t priority2[GICV3_MAXIRQ -GIC_INTERNAL ]; +uint16_t last_active[GICV3_MAXIRQ ][GICV3_NCPU ]; +/* For each SGI on the target CPU, we store 64 bits +* indicating which source CPUs have made this SGI +* pending on the target CPU. These correspond to +* the bytes in the GIC_SPENDSGIR* registers as +* read by the target CPU.
Quoted 4 lines
This comment doesn't make any sense. You can't fit 64 bits into a byte, and GIC_SPENDSGIR<n> only hold 8 set-pending bits per SGI.
Correct, the comment is wrong and should be removed as spi_mending was replaced by a structure.
The usage of a structure and not by a uint64_t vector is because there is no serialization macro for it.
Quoted 15 lines
Quoted 15 lines
include/hw/intc/arm_gicv3_common.h new file +14
-0+*/ +gicv3_sgi_state sgi_state[GIC_NR_SGIS ]; + +uint16_t priority_mask[GICV3_NCPU ];u + uint16_t running_irq[GICV3_NCPU]; +uint16_t running_priority[GICV3_NCPU ]; +uint16_t current_pending[GICV3_NCPU ]; + +uint32_t cpu_mask;/* For redistributor */ +uint32_t num_cpu; +MemoryRegion iomem_dist;/* Distributor */ +MemoryRegion iomem_lpi;/* Redistributor */ +uint32_t num_irq; +uint32_t revision;
Quoted 9 lines
include/hw/intc/arm_gicv3_common.h line 90
Quoted 3 lines
include/hw/intc/arm_gicv3_common.h new file +3
-0+bool security_levels; +int dev_fd;/* kvm device fd if backed by kvm vgic support */ +} GICv3State;This whole struct reads like "we just took the GICv2 state and changed it as little as possible beyond bumping the NCPU define a bit". That doesn't make me very confident that it's actually correct for GICv3...
I admit that the intention of the first is to make as little changes a possible and yet to be able to boot 64 Linux kernel (which was achieved). Note that the kernel identified and used the GICv3 driver.
Quoted 4 lines
thanks
-- PMM
PATCH v7 3/6 Extract some reusable vGIC code
In reply to Pavel FedinThese functions are useful also for vGICv3 implementation. Make them accessible from within other modules.
Actually kvm_dist_get() and kvm_dist_put() could also be made reusable, but they would require two extra parameters (s->dev_fd and s->num_cpu) as well as lots of typecasts of 's' to DeviceState * and back to GICState *. This makes the code very ugly so i decided to stop at this point. I tried also an approach with making a base class for all possible GICs, but it would contain only three variables (dev_fd, cpu_num and irq_num), and accessing them through the rest of the code would be again tedious (either ugly casts or qemu-style separate object pointer). So i disliked it too.
--- hw/intc/arm_gic_kvm.c | 84 ++++++++++++++++++++++++--------------------------- hw/intc/vgic_common.h | 43 ++++++++++++++++++++++++++ 2 files changed, 82 insertions(+), 45 deletions(-) create mode 100644 hw/intc/vgic_common.h
hw/intc/arm_gic_kvm.c +39 -45
@@ -23,6 +23,7 @@#include "sysemu/kvm.h" #include "kvm_arm.h" #include "gic_internal.h" +#include "vgic_common.h" //#define DEBUG_GIC_KVM
@@ -52,7 +53,7 @@ typedef struct KVMARMGICClass {void (*parent_reset)(DeviceState *dev); } KVMARMGICClass; -static void kvm_arm_gic_set_irq (void *opaque,int irq,int level) +void kvm_arm_gic_set_irq (uint32_t num_irq,int irq,int level) {/* Meaning of the 'irq' parameter: * [0..N-1] : external interrupts
@@ -63,10 +64,9 @@ static void kvm_arm_gic_set_irq(void *opaque, int irq, int level)* has separate fields in the irq number for type, * CPU number and interrupt number. */ -GICState *s = (GICState *)opaque;int kvm_irq, irqtype, cpu; -if (irq < (s->num_irq -GIC_INTERNAL )) { +if (irq < (num_irq -GIC_INTERNAL )) {/* External interrupt. The kernel numbers these like the GIC * hardware, with external interrupt IDs starting after the * internal ones.
@@ -77,7 +77,7 @@ static void kvm_arm_gic_set_irq(void *opaque, int irq, int level) }else {/* Internal interrupt: decode into (cpu, interrupt id) */ irqtype =KVM_ARM_IRQ_TYPE_PPI ; - irq -= (s->num_irq -GIC_INTERNAL ); + irq -= (num_irq -GIC_INTERNAL ); cpu = irq /GIC_INTERNAL ; irq %=GIC_INTERNAL ; }
@@ -87,12 +87,19 @@ static void kvm_arm_gic_set_irq(void *opaque, int irq, int level)kvm_set_irq (kvm_state, kvm_irq, !!level); } +static void kvm_arm_gicv2_set_irq (void *opaque,int irq,int level) +{ +GICState *s = (GICState *)opaque; + +kvm_arm_gic_set_irq (s->num_irq, irq, level); +} +static bool kvm_arm_gic_can_save_restore (GICState *s) {return s->dev_fd >=0 ; } -static bool kvm_gic_supports_attr (GICState *s,int group,int attrnum) +bool kvm_gic_supports_attr (int dev_fd,int group,int attrnum) {struct kvm_device_attr attr = { .group = group,
@@ -100,14 +107,14 @@ static bool kvm_gic_supports_attr(GICState *s, int group, int attrnum) .flags = 0, }; -if (s->dev_fd ==-1 ) { +if (dev_fd ==-1 ) {return false; } -return kvm_device_ioctl (s->dev_fd,KVM_HAS_DEVICE_ATTR , &attr) ==0 ; +return kvm_device_ioctl (dev_fd,KVM_HAS_DEVICE_ATTR , &attr) ==0 ; } -static void kvm_gic_access (GICState *s,int group,int offset, +void kvm_gic_access (int dev_fd,int group,int offset,int cpu,uint32_t *val,bool write) {struct kvm_device_attr attr;
@@ -130,7 +137,7 @@ static void kvm_gic_access(GICState *s, int group, int offset, type =KVM_GET_DEVICE_ATTR ; } - err =kvm_device_ioctl (s->dev_fd, type, &attr); + err =kvm_device_ioctl (dev_fd, type, &attr);if (err <0 ) {fprintf (stderr,"KVM_{SET/GET}_DEVICE_ATTR failed: %s\n" ,strerror (-err));
@@ -138,20 +145,6 @@ static void kvm_gic_access(GICState *s, int group, int offset, } } -static void kvm_gicd_access (GICState *s,int offset,int cpu, -uint32_t *val,bool write) -{ -kvm_gic_access (s,KVM_DEV_ARM_VGIC_GRP_DIST_REGS , - offset, cpu, val, write); -} - -static void kvm_gicc_access (GICState *s,int offset,int cpu, -uint32_t *val,bool write) -{ -kvm_gic_access (s,KVM_DEV_ARM_VGIC_GRP_CPU_REGS , - offset, cpu, val, write); -} -#define for_each_irq_reg (_ctr, _max_irq, _field_width) \ for (_ctr = 0; _ctr < ((_max_irq) / (32 / (_field_width))); _ctr++)
@@ -291,7 +284,7 @@ static void kvm_dist_get(GICState *s, uint32_t offset, int width, irq = i * regsz; cpu =0 ;while ((cpu < s->num_cpu && irq <GIC_INTERNAL ) || cpu ==0 ) { -kvm_gicd_access (s, offset, cpu, ®, false); +kvm_gicd_access (s->dev_fd, offset, cpu, ®, false);for (j =0 ; j < regsz; j++) { field =extract32 (reg, j * width, width);translate_fn (s, irq + j, cpu, &field, false);
@@ -324,7 +317,7 @@ static void kvm_dist_put(GICState *s, uint32_t offset, int width,translate_fn (s, irq + j, cpu, &field, true); reg =deposit32 (reg, j * width, width, field); } -kvm_gicd_access (s, offset, cpu, ®, true); +kvm_gicd_access (s->dev_fd, offset, cpu, ®, true); cpu++; }
@@ -355,10 +348,10 @@ static void kvm_arm_gic_put(GICState *s)/* s->ctlr -> GICD_CTLR */ reg = s->ctlr; -kvm_gicd_access (s,0x0 ,0 , ®, true); +kvm_gicd_access (s->dev_fd,0x0 ,0 , ®, true);/* Sanity checking on GICD_TYPER and s->num_irq, s->num_cpu */ -kvm_gicd_access (s,0x4 ,0 , ®, false); +kvm_gicd_access (s->dev_fd,0x4 ,0 , ®, false); num_irq = ((reg &0x1f ) +1 ) *32 ; num_cpu = ((reg &0xe0 ) >>5 ) +1 ;
@@ -416,24 +409,24 @@ static void kvm_arm_gic_put(GICState *s)for (cpu =0 ; cpu < s->num_cpu; cpu++) {/* s->cpu_ctlr[cpu] -> GICC_CTLR */ reg = s->cpu_ctlr[cpu]; -kvm_gicc_access (s,0x00 , cpu, ®, true); +kvm_gicc_access (s->dev_fd,0x00 , cpu, ®, true);/* s->priority_mask[cpu] -> GICC_PMR */ reg = (s->priority_mask[cpu] &0xff ); -kvm_gicc_access (s,0x04 , cpu, ®, true); +kvm_gicc_access (s->dev_fd,0x04 , cpu, ®, true);/* s->bpr[cpu] -> GICC_BPR */ reg = (s->bpr[cpu] &0x7 ); -kvm_gicc_access (s,0x08 , cpu, ®, true); +kvm_gicc_access (s->dev_fd,0x08 , cpu, ®, true);/* s->abpr[cpu] -> GICC_ABPR */ reg = (s->abpr[cpu] &0x7 ); -kvm_gicc_access (s,0x1c , cpu, ®, true); +kvm_gicc_access (s->dev_fd,0x1c , cpu, ®, true);/* s->apr[n][cpu] -> GICC_APRn */ for (i =0 ; i <4 ; i++) { reg = s->apr[i][cpu]; -kvm_gicc_access (s,0xd0 + i *4 , cpu, ®, true); +kvm_gicc_access (s->dev_fd,0xd0 + i *4 , cpu, ®, true); } } }
@@ -454,11 +447,11 @@ static void kvm_arm_gic_get(GICState *s)*/ /* GICD_CTLR -> s->ctlr */ -kvm_gicd_access (s,0x0 ,0 , ®, false); +kvm_gicd_access (s->dev_fd,0x0 ,0 , ®, false); s->ctlr = reg;/* Sanity checking on GICD_TYPER -> s->num_irq, s->num_cpu */ -kvm_gicd_access (s,0x4 ,0 , ®, false); +kvm_gicd_access (s->dev_fd,0x4 ,0 , ®, false); s->num_irq = ((reg &0x1f ) +1 ) *32 ; s->num_cpu = ((reg &0xe0 ) >>5 ) +1 ;
@@ -469,7 +462,7 @@ static void kvm_arm_gic_get(GICState *s) }/* GICD_IIDR -> ? */ -kvm_gicd_access (s,0x8 ,0 , ®, false); +kvm_gicd_access (s->dev_fd,0x8 ,0 , ®, false);/* Clear all the IRQ settings */ for (i =0 ; i < s->num_irq; i++) {
@@ -507,24 +500,24 @@ static void kvm_arm_gic_get(GICState *s)for (cpu =0 ; cpu < s->num_cpu; cpu++) {/* GICC_CTLR -> s->cpu_ctlr[cpu] */ -kvm_gicc_access (s,0x00 , cpu, ®, false); +kvm_gicc_access (s->dev_fd,0x00 , cpu, ®, false); s->cpu_ctlr[cpu] = reg;/* GICC_PMR -> s->priority_mask[cpu] */ -kvm_gicc_access (s,0x04 , cpu, ®, false); +kvm_gicc_access (s->dev_fd,0x04 , cpu, ®, false); s->priority_mask[cpu] = (reg &0xff );/* GICC_BPR -> s->bpr[cpu] */ -kvm_gicc_access (s,0x08 , cpu, ®, false); +kvm_gicc_access (s->dev_fd,0x08 , cpu, ®, false); s->bpr[cpu] = (reg &0x7 );/* GICC_ABPR -> s->abpr[cpu] */ -kvm_gicc_access (s,0x1c , cpu, ®, false); +kvm_gicc_access (s->dev_fd,0x1c , cpu, ®, false); s->abpr[cpu] = (reg &0x7 );/* GICC_APRn -> s->apr[n][cpu] */ for (i =0 ; i <4 ; i++) { -kvm_gicc_access (s,0xd0 + i *4 , cpu, ®, false); +kvm_gicc_access (s->dev_fd,0xd0 + i *4 , cpu, ®, false); s->apr[i][cpu] = reg; } }
@@ -569,7 +562,7 @@ static void kvm_arm_gic_realize(DeviceState *dev, Error **errp)* ... */ i += (GIC_INTERNAL * s->num_cpu); -qdev_init_gpio_in (dev, kvm_arm_gic_set_irq, i); +qdev_init_gpio_in (dev, kvm_arm_gicv2_set_irq, i);for (i =0 ; i < s->num_irq -GIC_INTERNAL ; i++) {qemu_irq irq =qdev_get_gpio_in (dev, i);
@@ -596,15 +589,16 @@ static void kvm_arm_gic_realize(DeviceState *dev, Error **errp)return ; } -if (kvm_gic_supports_attr (s,KVM_DEV_ARM_VGIC_GRP_NR_IRQS ,0 )) { +if (kvm_gic_supports_attr (s->dev_fd,KVM_DEV_ARM_VGIC_GRP_NR_IRQS ,0 )) {uint32_t numirqs = s->num_irq; -kvm_gic_access (s,KVM_DEV_ARM_VGIC_GRP_NR_IRQS ,0 ,0 , &numirqs,1 ); +kvm_gic_access (s->dev_fd,KVM_DEV_ARM_VGIC_GRP_NR_IRQS ,0 ,0 , + &numirqs,1 ); }/* Tell the kernel to complete VGIC initialization now */ -if (kvm_gic_supports_attr (s,KVM_DEV_ARM_VGIC_GRP_CTRL , +if (kvm_gic_supports_attr (s->dev_fd,KVM_DEV_ARM_VGIC_GRP_CTRL ,KVM_DEV_ARM_VGIC_CTRL_INIT )) { -kvm_gic_access (s,KVM_DEV_ARM_VGIC_GRP_CTRL , +kvm_gic_access (s->dev_fd,KVM_DEV_ARM_VGIC_GRP_CTRL ,KVM_DEV_ARM_VGIC_CTRL_INIT ,0 ,0 ,1 ); }
hw/intc/vgic_common.h new file +43 -0
43 lines not on this page · raw message
Signature · 1 line
1.9.5.msysgit.0
PATCH v7 4/6 Introduce irqchip type specification for KVM
In reply to Pavel FedinThis patch introduces kernel_irqchip_type member in Machine class, which is passed to kvm_arch_irqchip_create. Machine models which can use vGIC now use it in order to supply correct GIC type for KVM capability verification. The variable is defined as int in order to be architecture-agnostic for potential future uses by other architectures.
--- hw/arm/vexpress.c | 1 + hw/arm/virt.c | 3 +++ include/hw/boards.h | 1 + include/sysemu/kvm.h | 3 ++- kvm-all.c | 2 +- stubs/kvm.c | 2 +- target-arm/kvm-consts.h | 6 ++++++ target-arm/kvm.c | 13 +++++++++++-- 8 files changed, 26 insertions(+), 5 deletions(-)
hw/arm/vexpress.c +1 -0
@@ -556,6 +556,7 @@ static void vexpress_common_init(MachineState *machine)const hwaddr *map = daughterboard->motherboard_map;int i; + machine->kernel_irqchip_type =QEMU_GIC_TYPE_V2 ; daughterboard->init (vms, machine->ram_size, machine->cpu_model, pic);/*
hw/arm/virt.c +3 -0
@@ -945,6 +945,9 @@ static void virt_instance_init(Object *obj)"Set on/off to enable/disable the ARM " "Security Extensions (TrustZone)" , NULL); + +/* Default GIC type is v2 */ + vms->parent.kernel_irqchip_type =QEMU_GIC_TYPE_V2 ; }static void virt_class_init (ObjectClass *oc,void *data)
include/hw/boards.h +1 -0
@@ -127,6 +127,7 @@ struct MachineState {char *accel;bool kernel_irqchip_allowed;bool kernel_irqchip_required; +int kernel_irqchip_type;int kvm_shadow_mem;char *dtb;char *dumpdtb;
include/sysemu/kvm.h +2 -1
@@ -434,6 +434,7 @@ void kvm_init_irq_routing(KVMState *s);/** * kvm_arch_irqchip_create: * @KVMState: The KVMState pointer +* @type: irqchip type, architecture-specific * * Allow architectures to create an in-kernel irq chip themselves. *
@@ -441,7 +442,7 @@ void kvm_init_irq_routing(KVMState *s);* 0: irq chip was not created * > 0: irq chip was created */ -int kvm_arch_irqchip_create (KVMState *s); +int kvm_arch_irqchip_create (KVMState *s,int type);/** * kvm_set_one_reg - set a register value in KVM via KVM_SET_ONE_REG ioctl
kvm-all.c +1 -1
@@ -1395,7 +1395,7 @@ static void kvm_irqchip_create(MachineState *machine, KVMState *s)/* First probe and see if there's a arch-specific hook to create the * in-kernel irqchip for us */ - ret =kvm_arch_irqchip_create (s); + ret =kvm_arch_irqchip_create (s, machine->kernel_irqchip_type);if (ret ==0 ) { ret =kvm_vm_ioctl (s,KVM_CREATE_IRQCHIP ); }
stubs/kvm.c +1 -1
@@ -1,7 +1,7 @@#include "qemu-common.h" #include "sysemu/kvm.h" -int kvm_arch_irqchip_create (KVMState *s) +int kvm_arch_irqchip_create (KVMState *s,int type) {return 0 ; }
target-arm/kvm-consts.h +6 -0
@@ -39,6 +39,12 @@ MISMATCH_CHECK(CP_REG_SIZE_U64, KVM_REG_SIZE_U64) MISMATCH_CHECK(CP_REG_ARM, KVM_REG_ARM) MISMATCH_CHECK(CP_REG_ARCH_MASK, KVM_REG_ARCH_MASK) +#define QEMU_GIC_TYPE_V2 5 +#define QEMU_GIC_TYPE_V3 7 + +MISMATCH_CHECK(QEMU_GIC_TYPE_V2, KVM_DEV_TYPE_ARM_VGIC_V2) +MISMATCH_CHECK(QEMU_GIC_TYPE_V3, KVM_DEV_TYPE_ARM_VGIC_V3) +#define QEMU_PSCI_0_1_FN_BASE 0x95c1ba5e#define QEMU_PSCI_0_1_FN (n) (QEMU_PSCI_0_1_FN_BASE + (n))#define QEMU_PSCI_0_1_FN_CPU_SUSPEND QEMU_PSCI_0_1_FN(0)
target-arm/kvm.c +11 -2
@@ -583,19 +583,28 @@ void kvm_arch_init_irq_routing(KVMState *s) { } -int kvm_arch_irqchip_create (KVMState *s) +int kvm_arch_irqchip_create (KVMState *s,int type) {int ret; +/* Failure here means forgotten initialization of +* machine->kernel_irqchip_type in model code +*/ +assert (type !=0 ); +/* If we can create the VGIC using the newer device control API, we * let the device do this when it initializes itself, otherwise we * fall back to the old API */ - ret =kvm_create_device (s,KVM_DEV_TYPE_ARM_VGIC_V2 , true); + ret =kvm_create_device (s, type, true);if (ret ==0 ) {return 1 ; } +/* Fallback will create VGIC v2 */ +if (type !=KVM_DEV_TYPE_ARM_VGIC_V2 ) { +return ret; + }return 0 ; }
Signature · 1 line
1.9.5.msysgit.0
Re: [Qemu-devel] [PATCH v7 4/6] Introduce irqchip type specification for KVM
Quoted 5 lines · Pavel Fedin
This patch introduces
kernel_irqchip_typemember in Machine class, which is passed tokvm_arch_irqchip_create. Machine models which can use vGIC now use it in order to supply correct GIC type for KVM capability verification. The variable is defined as int in order to be architecture-agnostic for potential future uses by other architectures.
This doesn't look right. The board model should just create the GIC device that it wants, and it should be the job of the GIC device itself to coordinate with the kvm.c code if it has to do so to pass the right kind of irqchip type.
thanks
-- PMM
Re: [Qemu-devel] [PATCH v7 4/6] Introduce irqchip type specification for KVM
Hello!
Quoted 10 lines
Quoted 5 lines
This patch introduces
kernel_irqchip_typemember in Machine class, which is passed tokvm_arch_irqchip_create. Machine models which can use vGIC now use it in order to supply correct GIC type for KVM capability verification. The variable is defined as int in order to be architecture-agnostic for potential future uses by other architectures.This doesn't look right. The board model should just create the GIC device that it wants, and it should be the job of the GIC device itself to coordinate with the
kvm.ccode if it has to do so to pass the right kind of irqchip type.
Unfortunately, this is how qemu is designed. kvm_irqchip_create() is called long before machine model's init code (machvirt_init() in our case) is called. So, if we want to check for the right thing, we have to know it before machine model actually starts to instantiate its components. I don't see how to change this in a more or less simple way. This routine, except capability testing, does one more important thing on old kernels. There it actually creates vGICv2 using KVM_CREATE_IRQCHIP ioctl. And, on non-ARM architectures, this ioctl also works, and this is the only place where creation is done.
Kind regards,
Pavel Fedin
Expert Engineer
Samsung Electronics Research center Russia
Re: [Qemu-devel] [PATCH v7 4/6] Introduce irqchip type specification for KVM
Quoted 10 lines · Pavel Fedin
Unfortunately, this is how qemu is designed.
kvm_irqchip_create()is
called long before machine model's init code (machvirt_init()in our case) is called. So, if we want to check for the right thing, we have to know it before machine model actually starts to instantiate its components.I don't see how to change this in a more or less simple way. This
routine, except capability testing, does one more important thing on old kernels. There it actually creates vGICv2 usingKVM_CREATE_IRQCHIPioctl.
Right, but all we really need to do in kvm_irqchip_create() is confirm whether we can use the new API later; we don't actually need to care whether we will later be creating a GICv2 or a GICv3 (or whether the kernel supports GICv3).
Rather than having kvm_irqchip_create() pile up a list of every irqchip type we might want later, I think it's probably better to just have it do the check for KVM_CAP_DEVICE_CTRL. If that check fails, we are on an old kernel, and should return 0 so we do the old-style KVM_CREATE_IRQCHIP. If it succeeds, then we're on a new kernel, and can postpone doing any GIC init until the GIC device object is created. Then the GIC device object should just try to do the kvm_create_device() for the relevant kind of GIC. If we're trying to create a GICv3 on a kernel that only knows about a GICv2 then this will fail at this point and we can display a suitable error message.
Christoffer: was there a particular reason you wrote kvm_irqchip_create() to do a test device-create as well as check the capability bit?
thanks
-- PMM
PATCH v7 5/6 Initial implementation of vGICv3
In reply to Pavel FedinGet/put routines are missing, live migration is not possible.
--- hw/intc/Makefile.objs | 3 + hw/intc/arm_gicv3_kvm.c | 155 ++++++++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 158 insertions(+) create mode 100644 hw/intc/arm_gicv3_kvm.c
hw/intc/Makefile.objs +3 -0
@@ -17,6 +17,9 @@ common-obj-$(CONFIG_OPENPIC) += openpic.o obj-$(CONFIG_APIC) += apic.o apic_common.o obj-$(CONFIG_ARM_GIC_KVM) += arm_gic_kvm.o +ifeq ($(ARCH), aarch64) # Only 64-bit KVM can use these +obj-$(CONFIG_ARM_GIC_KVM) += arm_gicv3_kvm.o +endif obj-$(CONFIG_STELLARIS) += armv7m_nvic.o obj-$(CONFIG_EXYNOS4) += exynos4210_gic.o exynos4210_combiner.o obj-$(CONFIG_GRLIB) += grlib_irqmp.o
hw/intc/arm_gicv3_kvm.c new file +155 -0
155 lines not on this page · raw message
Signature · 1 line
1.9.5.msysgit.0
Re: [Qemu-devel] [PATCH v7 5/6] Initial implementation of vGICv3
Hi. I have been reading the GIC specs over the last few days and I think I'm getting closer to understanding what we need to do with this series to get it to a committable state (without tying it up with GICv3 emulation). I'll send some mail about that tomorrow probably, when I've figured out the details. In the meantime, some review comments below (mostly minor stuff).
Quoted 1 line · Pavel Fedin
Get/put routines are missing, live migration is not possible.
This commit message could do with being made a bit less terse.
Quoted 12 lines
Signed-off-by Pavel Fedin <[email protected]>--- hw/intc/Makefile.objs | 3 + hw/intc/arm_gicv3_kvm.c | 155 ++++++++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 158 insertions(+) create mode 100644 hw/intc/arm_gicv3_kvm.chw/intc/Makefile.objs +0
-0
hw/intc/Makefile.objs common-obj-$(CONFIG_OPENPIC) += openpic.o · line 21
Quoted 5 lines
hw/intc/Makefile.objs +2
-0@@ -17,6 +17,9 @@ common-obj-$(CONFIG_OPENPIC) += openpic.o obj-$(CONFIG_APIC) += apic.o apic_common.o obj-$(CONFIG_ARM_GIC_KVM) += arm_gic_kvm.o +ifeq ($(ARCH), aarch64) # Only 64-bit KVM can use these +obj-$(CONFIG_ARM_GIC_KVM) += arm_gicv3_kvm.o
Does it actually fail to compile in a 32-bit KVM config? Anyway, probably better to be consistent with how target-arm/Makefile.objs enables kvm64.o:
obj-$(call land,$(CONFIG_ARM_GIC_KVM),$(TARGET_AARCH64)) += kvm64.o
hw/intc/Makefile.objs common-obj-$(CONFIG_OPENPIC) += openpic.o
Quoted 10 lines
hw/intc/Makefile.objs +10
-0+static void kvm_arm_gicv3_reset(DeviceState *dev) +{ + GICv3State *s = ARM_GICV3_COMMON(dev); + KVMARMGICv3Class *kgc = KVM_ARM_GICV3_GET_CLASS(s); + + DPRINTF("Reset\n"); + + kgc->parent_reset(dev); + kvm_arm_gicv3_put(s); +}
If we don't currently do anything in reset then does the GIC just go wrong on a VM reset?
Quoted 13 lines
hw/intc/Makefile.objs +13
-0+static void kvm_arm_gicv3_realize(DeviceState *dev, Error **errp) +{ + GICv3State *s = KVM_ARM_GICV3(dev); + KVMARMGICv3Class *kgc = KVM_ARM_GICV3_GET_CLASS(s); + Error *local_err = NULL; + int ret; + + DPRINTF("kvm_arm_gicv3_realize\n"); + + kgc->parent_realize(dev, &local_err); + if (local_err) { + error_propagate(errp, local_err); + return;
hw/intc/Makefile.objs
Quoted 10 lines
hw/intc/Makefile.objs +10
-0+ } + + gicv3_init_irqs_and_mmio(s, kvm_arm_gicv3_set_irq, NULL); + + /* Try to create the device via the device control API */ + s->dev_fd = -1; + ret = kvm_create_device(kvm_state, KVM_DEV_TYPE_ARM_VGIC_V3, false); + if (ret >= 0) { + s->dev_fd = ret; + } else if (ret != -ENODEV && ret != -ENOTSUP) {
Why aren't ENODEV and ENOTSUP fatal errors like other errnos?
hw/intc/Makefile.objs
Quoted 5 lines
hw/intc/Makefile.objs +5
-0+ error_setg_errno(errp, -ret, "error creating in-kernel VGIC"); + return; + } + + if (kvm_gic_supports_attr(s->dev_fd, KVM_DEV_ARM_VGIC_GRP_NR_IRQS, 0)) {
Is there any kernel which supports GICv3 but does not support this attribute? I would hope not, in which case we can skip the conditional check for support.
hw/intc/Makefile.objs
Quoted 9 lines
hw/intc/Makefile.objs +9
-0+ uint32_t numirqs = s->num_irq; + DPRINTF("KVM_DEV_ARM_VGIC_GRP_NR_IRQS = %u\n", numirqs); + kvm_gic_access(s->dev_fd, KVM_DEV_ARM_VGIC_GRP_NR_IRQS, + 0, 0, &numirqs, 1); + } + + /* Tell the kernel to complete VGIC initialization now */ + if (kvm_gic_supports_attr(s->dev_fd, KVM_DEV_ARM_VGIC_GRP_CTRL, + KVM_DEV_ARM_VGIC_CTRL_INIT)) {
Ditto.
hw/intc/Makefile.objs
Quoted 10 lines
hw/intc/Makefile.objs +10
-0+ DPRINTF("KVM_DEV_ARM_VGIC_CTRL_INIT\n"); + kvm_gic_access(s->dev_fd, KVM_DEV_ARM_VGIC_GRP_CTRL, + KVM_DEV_ARM_VGIC_CTRL_INIT, 0, 0, 1); + } + + kvm_arm_register_device(&s->iomem_dist, -1, KVM_DEV_ARM_VGIC_GRP_ADDR, + KVM_VGIC_V3_ADDR_TYPE_DIST, s->dev_fd); + kvm_arm_register_device(&s->iomem_lpi, -1, KVM_DEV_ARM_VGIC_GRP_ADDR, + KVM_VGIC_V3_ADDR_TYPE_REDIST, s->dev_fd); +}
thanks
-- PMM
Re: [Qemu-devel] [PATCH v7 5/6] Initial implementation of vGICv3
Hello!
Quoted 4 lines · Pavel Fedin
Quoted 1 line
Get/put routines are missing, live migration is not possible.
This commit message could do with being made a bit less terse.
Ok. Subject was self-explanatory, so i didn't have much to add.
Quoted 5 lines
Quoted 5 lines
hw/intc/arm_gicv3_kvm.c new file +5
-0+static void kvm_arm_gicv3_reset (DeviceState *dev) +{ +GICv3State *s =ARM_GICV3_COMMON (dev); +KVMARMGICv3Class *kgc =KVM_ARM_GICV3_GET_CLASS (s); +
Quoted 8 lines
hw/intc/arm_gicv3_kvm.c line 78
Quoted 5 lines
hw/intc/arm_gicv3_kvm.c new file +5
-0+DPRINTF ("Reset\n" ); + + kgc->parent_reset (dev); +kvm_arm_gicv3_put (s); +}If we don't currently do anything in reset then does the GIC just go wrong on a VM reset?
No it doesn't, reset works.
Quoted 8 lines
hw/intc/arm_gicv3_kvm.c line 102
Quoted 6 lines
hw/intc/arm_gicv3_kvm.c new file +6
-0+/* Try to create the device via the device control API */ + s->dev_fd =-1 ; + ret =kvm_create_device (kvm_state,KVM_DEV_TYPE_ARM_VGIC_V3 , false); +if (ret >=0 ) { + s->dev_fd = ret; + }else if (ret != -ENODEV && ret != -ENOTSUP ) {Why aren't ENODEV and ENOTSUP fatal errors like other errnos?
Stupid leftover from vGICv2 code, copypasted. I have already removed it in my repo, just decided to delay the respin until i know the fate of this: https://lists.gnu.org/archive/html/qemu-devel/2015-07/msg05778.html
Quoted 21 lines
hw/intc/arm_gicv3_kvm.c line 107
Quoted 6 lines
hw/intc/arm_gicv3_kvm.c new file +5
-0+error_setg_errno (errp, -ret,"error creating in-kernel VGIC" ); +return ; + } + +if (kvm_gic_supports_attr (s->dev_fd,KVM_DEV_ARM_VGIC_GRP_NR_IRQS ,0 )) {Is there any kernel which supports GICv3 but does not support this attribute? I would hope not, in which case we can skip the conditional check for support.
hw/intc/arm_gicv3_kvm.c line 116
Quoted 9 lines
hw/intc/arm_gicv3_kvm.c new file +9
-0+uint32_t numirqs = s->num_irq; +DPRINTF ("KVM_DEV_ARM_VGIC_GRP_NR_IRQS = %u\n" , numirqs); +kvm_gic_access (s->dev_fd,KVM_DEV_ARM_VGIC_GRP_NR_IRQS , +0 ,0 , &numirqs,1 ); + } + +/* Tell the kernel to complete VGIC initialization now */ +if (kvm_gic_supports_attr (s->dev_fd,KVM_DEV_ARM_VGIC_GRP_CTRL , +KVM_DEV_ARM_VGIC_CTRL_INIT )) {
Quoted 1 line
Ditto.
I intentionally put some tracing to these conditions. On my system KVM_DEV_ARM_VGIC_GRP_NR_IRQS is supported and KVM_DEV_ARM_VGIC_CTRL_INIT is not. So will it always be this way?
Kind regards,
Pavel Fedin
Expert Engineer
Samsung Electronics Research center Russia
Re: [Qemu-devel] [PATCH v7 5/6] Initial implementation of vGICv3
Quoted 17 lines · Pavel Fedin
Hello!
Quoted 6 lines · Pavel Fedin
Quoted 5 lines
hw/intc/arm_gicv3_kvm.c new file +5
-0+static void kvm_arm_gicv3_reset (DeviceState *dev) +{ +GICv3State *s =ARM_GICV3_COMMON (dev); +KVMARMGICv3Class *kgc =KVM_ARM_GICV3_GET_CLASS (s); +Quoted 8 lines
hw/intc/arm_gicv3_kvm.c line 78
Quoted 5 lines
hw/intc/arm_gicv3_kvm.c new file +5
-0+DPRINTF ("Reset\n" ); + + kgc->parent_reset (dev); +kvm_arm_gicv3_put (s); +}If we don't currently do anything in reset then does the GIC just go wrong on a VM reset?
Quoted 1 line
No it doesn't, reset works.
So who resets the KVM internal GIC state and how?
Quoted 20 lines
Quoted 20 lines
hw/intc/arm_gicv3_kvm.c line 107
Quoted 5 lines
hw/intc/arm_gicv3_kvm.c new file +5
-0+error_setg_errno (errp, -ret,"error creating in-kernel VGIC" ); +return ; + } + +if (kvm_gic_supports_attr (s->dev_fd,KVM_DEV_ARM_VGIC_GRP_NR_IRQS ,0 )) {Is there any kernel which supports GICv3 but does not support this attribute? I would hope not, in which case we can skip the conditional check for support.
Quoted 9 lines
hw/intc/arm_gicv3_kvm.c new file +9
-0+uint32_t numirqs = s->num_irq; +DPRINTF ("KVM_DEV_ARM_VGIC_GRP_NR_IRQS = %u\n" , numirqs); +kvm_gic_access (s->dev_fd,KVM_DEV_ARM_VGIC_GRP_NR_IRQS , +0 ,0 , &numirqs,1 ); + } + +/* Tell the kernel to complete VGIC initialization now */ +if (kvm_gic_supports_attr (s->dev_fd,KVM_DEV_ARM_VGIC_GRP_CTRL , +KVM_DEV_ARM_VGIC_CTRL_INIT )) {
Quoted 5 lines
Quoted 1 line
Ditto.
I intentionally put some tracing to these conditions. On my
systemKVM_DEV_ARM_VGIC_GRP_NR_IRQSis supported andKVM_DEV_ARM_VGIC_CTRL_INITis not. So will it always be this way?
That is weird. I thought that the kernel devs had agreed that all new VGIC code should require explicit initialization (ie that lazy init-on-first-time-it's-needed was only there for legacy guests and the old GICv2). The lazy-init stuff was a big source of bugs where things weren't inited at the point where you expected, which is why it should have gone away...
-- PMM
Re: [Qemu-devel] [PATCH v7 5/6] Initial implementation of vGICv3
Quoted 11 lines · Peter Maydell
Quoted 3 lines · Pavel Fedin
I intentionally put some tracing to these conditions. On my
systemKVM_DEV_ARM_VGIC_GRP_NR_IRQSis supported andKVM_DEV_ARM_VGIC_CTRL_INITis not. So will it always be this way?That is weird. I thought that the kernel devs had agreed that all new VGIC code should require explicit initialization (ie that lazy init-on-first-time-it's-needed was only there for legacy guests and the old GICv2). The lazy-init stuff was a big source of bugs where things weren't inited at the point where you expected, which is why it should have gone away...
I just checked with Marc, and he agreed that all kernels with GICv3 support should support VGIC_CTRL_INIT. What kernel are you running?
thanks
-- PMM
Re: [Qemu-devel] [PATCH v7 5/6] Initial implementation of vGICv3
Hello!
Quoted 11 lines
Quoted 10 lines
Quoted 3 lines
I intentionally put some tracing to these conditions. On my
systemKVM_DEV_ARM_VGIC_GRP_NR_IRQSis supported andKVM_DEV_ARM_VGIC_CTRL_INITis not. So will it always be this way?That is weird. I thought that the kernel devs had agreed that all new VGIC code should require explicit initialization (ie that lazy init-on-first-time-it's-needed was only there for legacy guests and the old GICv2). The lazy-init stuff was a big source of bugs where things weren't inited at the point where you expected, which is why it should have gone away...
Quoted 3 lines
I just checked with Marc, and he agreed that all kernels with GICv3 support should support
VGIC_CTRL_INIT. What kernel are you running?
v3.18 with backported GICv3 support. It is a choice of our HW vendor and their current official kernel.
Kind regards,
Pavel Fedin
Expert Engineer
Samsung Electronics Research center Russia
Re: [Qemu-devel] [PATCH v7 5/6] Initial implementation of vGICv3
Quoted 13 lines · Pavel Fedin
Hello!
Quoted 11 lines
Quoted 10 lines
Quoted 3 lines
I intentionally put some tracing to these conditions. On my
systemKVM_DEV_ARM_VGIC_GRP_NR_IRQSis supported andKVM_DEV_ARM_VGIC_CTRL_INITis not. So will it always be this way?That is weird. I thought that the kernel devs had agreed that all new VGIC code should require explicit initialization (ie that lazy init-on-first-time-it's-needed was only there for legacy guests and the old GICv2). The lazy-init stuff was a big source of bugs where things weren't inited at the point where you expected, which is why it should have gone away...
Quoted 6 lines
Quoted 3 lines
I just checked with Marc, and he agreed that all kernels with GICv3 support should support
VGIC_CTRL_INIT. What kernel are you running?v3.18 with backported GICv3 support. It is a choice of our
HW vendor and their current official kernel.
Sounds like their backporting is buggy, then. They need to fix it, I don't think we need to support broken branches in QEMU.
thanks
-- PMM
Re: [Qemu-devel] [PATCH v7 5/6] Initial implementation of vGICv3
Hello!
Quoted 10 lines
Quoted 6 lines
Quoted 3 lines
I just checked with Marc, and he agreed that all kernels with GICv3 support should support
VGIC_CTRL_INIT. What kernel are you running?v3.18 with backported GICv3 support. It is a choice of our
HW vendor and their current official kernel.Sounds like their backporting is buggy, then. They need to fix it, I don't think we need to support broken branches in QEMU.
Ok, i will recheck it.
So, final conclusion: both conditions should be removed and their code should be executed unconditionally. Correct ?
Kind regards,
Pavel Fedin
Expert Engineer
Samsung Electronics Research center Russia
Re: [Qemu-devel] [PATCH v7 5/6] Initial implementation of vGICv3
Quoted 9 lines · Pavel Fedin
Hello!
Quoted 7 lines
Quoted 6 lines
Quoted 3 lines
I just checked with Marc, and he agreed that all kernels with GICv3 support should support
VGIC_CTRL_INIT. What kernel are you running?v3.18 with backported GICv3 support. It is a choice of our
HW vendor and their current official kernel.
Quoted 7 lines
Quoted 3 lines
Sounds like their backporting is buggy, then. They need to fix it, I don't think we need to support broken branches in QEMU.
Ok, i will recheck it.
So, final conclusion: both conditions should be removed and their
code should be executed unconditionally. Correct ?
Yes.
thanks
-- PMM
Re: [Qemu-devel] [PATCH v7 5/6] Initial implementation of vGICv3
In reply to Peter MaydellHello!
Quoted 6 lines
hw/intc/Makefile.objs common-obj-$(CONFIG_OPENPIC) += openpic.o · line 21
Quoted 4 lines
obj-$(CONFIG_APIC) += apic.o apic_common.o
obj-$(CONFIG_ARM_GIC_KVM) += arm_gic_kvm.ohw/intc/Makefile.objs +2
-0+ifeq ($(ARCH), aarch64) # Only 64-bit KVM can use these +obj-$(CONFIG_ARM_GIC_KVM) += arm_gicv3_kvm.oDoes it actually fail to compile in a 32-bit KVM config?
I forgot to reply to this. Yes, it does, because KVM_VGIC_V3_ADDR_TYPE_xxx definitions are available only for 64 bits.
Kind regards,
Pavel Fedin
Expert Engineer
Samsung Electronics Research center Russia
PATCH v7 6/6 Add gicversion option to virt machine
In reply to Pavel FedinSet kernel_irqchip_type according to value of the option and pass it around where necessary. Instantiate devices and fdt nodes according to the choice.
max_cpus for virt machine increased to 64. GICv2 compatibility check happens inside arm_gic_common_realize().
--- hw/arm/virt.c | 137 ++++++++++++++++++++++++++++++++++++++++++-------- include/hw/arm/fdt.h | 2 +- include/hw/arm/virt.h | 6 ++- 3 files changed, 121 insertions(+), 24 deletions(-)
hw/arm/virt.c +115 -22
@@ -48,6 +48,7 @@#include "hw/arm/sysbus-fdt.h" #include "hw/platform-bus.h" #include "hw/arm/fdt.h" +#include "qapi/visitor.h" /* Number of external interrupt lines to configure the GIC with */ #define NUM_IRQS 256
@@ -106,7 +107,12 @@ static const MemMapEntry a15memmap[] = { /* GIC distributor and CPU interfaces sit inside the CPU peripheral space */ [VIRT_GIC_DIST] = { 0x08000000, 0x00010000 }, [VIRT_GIC_CPU] = { 0x08010000, 0x00010000 }, - [VIRT_GIC_V2M] = { 0x08020000, 0x00001000 }, + [VIRT_GIC_V2M] = { 0x08020000, 0x00010000 }, +/* On v3 VIRT_GIC_DIST_MBI and VIRT_ITS_CONTROL take place of +* VIRT_GIC_CPU and VIRT_GIC_V2M respectively + */ + [VIRT_ITS_TRANSLATION] = { 0x08030000, 0x00010000 }, + [VIRT_LPI] = { 0x08040000, 0x00800000 }, [VIRT_UART] = { 0x09000000, 0x00001000 }, [VIRT_RTC] = { 0x09010000, 0x00001000 }, [VIRT_FW_CFG] = { 0x09020000, 0x0000000a },
@@ -256,10 +262,13 @@ static void fdt_add_timer_nodes(const VirtBoardInfo *vbi)* they are edge-triggered. */ ARMCPU *armcpu; +uint32_t max;uint32_t irqflags =GIC_FDT_IRQ_FLAGS_EDGE_LO_HI ; +/* Argument is 32 bit but 8 bits are reserved for flags */ + max = (vbi->smp_cpus >=24 ) ?24 : vbi->smp_cpus; irqflags =deposit32 (irqflags,GIC_FDT_IRQ_PPI_CPU_START , -GIC_FDT_IRQ_PPI_CPU_WIDTH , (1 << vbi->smp_cpus) -1 ); +GIC_FDT_IRQ_PPI_CPU_WIDTH , (1 << max) -1 );qemu_fdt_add_subnode (vbi->fdt,"/timer" );
@@ -283,6 +292,18 @@ static void fdt_add_cpu_nodes(const VirtBoardInfo *vbi) {int cpu; +/* +* From Documentation/devicetree/bindings/arm/cpus.txt +* On ARM v8 64-bit systems value should be set to 2, +* that corresponds to the MPIDR_EL1 register size. +* If MPIDR_EL1[63:32] value is equal to 0 on all CPUs +* in the system, #address-cells can be set to 1, since +* MPIDR_EL1[63:32] bits are not used for CPUs +* identification. +* +* Now GIC500 doesn't support affinities 2 & 3 so currently +* #address-cells can stay 1 until future GIC +*/ qemu_fdt_add_subnode (vbi->fdt,"/cpus" );qemu_fdt_setprop_cell (vbi->fdt,"/cpus" ,"#address-cells" ,0x1 );qemu_fdt_setprop_cell (vbi->fdt,"/cpus" ,"#size-cells" ,0x0 );
@@ -319,25 +340,36 @@ static void fdt_add_v2m_gic_node(VirtBoardInfo *vbi)qemu_fdt_setprop_cell (vbi->fdt,"/intc/v2m" ,"phandle" , vbi->v2m_phandle); } -static void fdt_add_gic_node (VirtBoardInfo *vbi) +static void fdt_add_gic_node (VirtBoardInfo *vbi,int type) { vbi->gic_phandle =qemu_fdt_alloc_phandle (vbi->fdt);qemu_fdt_setprop_cell (vbi->fdt,"/" ,"interrupt-parent" , vbi->gic_phandle);qemu_fdt_add_subnode (vbi->fdt,"/intc" ); -/* 'cortex-a15-gic' means 'GIC v2' */ -qemu_fdt_setprop_string (vbi->fdt,"/intc" ,"compatible" , -"arm,cortex-a15-gic" );qemu_fdt_setprop_cell (vbi->fdt,"/intc" ,"#interrupt-cells" ,3 );qemu_fdt_setprop (vbi->fdt,"/intc" ,"interrupt-controller" ,NULL ,0 ); -qemu_fdt_setprop_sized_cells (vbi->fdt,"/intc" ,"reg" , -2 , vbi->memmap[VIRT_GIC_DIST ].base, -2 , vbi->memmap[VIRT_GIC_DIST ].size, -2 , vbi->memmap[VIRT_GIC_CPU ].base, -2 , vbi->memmap[VIRT_GIC_CPU ].size);qemu_fdt_setprop_cell (vbi->fdt,"/intc" ,"#address-cells" ,0x2 );qemu_fdt_setprop_cell (vbi->fdt,"/intc" ,"#size-cells" ,0x2 );qemu_fdt_setprop (vbi->fdt,"/intc" ,"ranges" ,NULL ,0 ); +if (type ==KVM_DEV_TYPE_ARM_VGIC_V3 ) { +qemu_fdt_setprop_string (vbi->fdt,"/intc" ,"compatible" , +"arm,gic-v3" ); +qemu_fdt_setprop_sized_cells (vbi->fdt,"/intc" ,"reg" , +2 , vbi->memmap[VIRT_GIC_DIST ].base, +2 , vbi->memmap[VIRT_GIC_DIST ].size, +2 , vbi->memmap[VIRT_LPI ].base, +2 , vbi->memmap[VIRT_LPI ].size); + }else { +/* 'cortex-a15-gic' means 'GIC v2' */ +qemu_fdt_setprop_string (vbi->fdt,"/intc" ,"compatible" , +"arm,cortex-a15-gic" ); +qemu_fdt_setprop_sized_cells (vbi->fdt,"/intc" ,"reg" , +2 , vbi->memmap[VIRT_GIC_DIST ].base, +2 , vbi->memmap[VIRT_GIC_DIST ].size, +2 , vbi->memmap[VIRT_GIC_CPU ].base, +2 , vbi->memmap[VIRT_GIC_CPU ].size); + } +qemu_fdt_setprop_cell (vbi->fdt,"/intc" ,"phandle" , vbi->gic_phandle); }
@@ -360,20 +392,35 @@ static void create_v2m(VirtBoardInfo *vbi, qemu_irq *pic)fdt_add_v2m_gic_node (vbi); } -static void create_gic (VirtBoardInfo *vbi,qemu_irq *pic) +static void create_gic (VirtBoardInfo *vbi,qemu_irq *pic,int type) {/* We create a standalone GIC v2 */ DeviceState *gicdev;SysBusDevice *gicbusdev; -const char *gictype ="arm_gic" ; +const char *gictype;int i; -if (kvm_irqchip_in_kernel ()) { - gictype ="kvm-arm-gic" ; +if (type ==KVM_DEV_TYPE_ARM_VGIC_V3 ) { +/* TODO: Software emulation is not implemented yet */ +if (!kvm_irqchip_in_kernel ()) { +fprintf (stderr,"KVM is currently required for GICv3 emulation\n" ); +exit (1 ); + } + gictype ="kvm-arm-gicv3" ; + }else { + gictype =kvm_irqchip_in_kernel () ?"kvm-arm-gic" :"arm_gic" ; } gicdev =qdev_create (NULL , gictype); -qdev_prop_set_uint32 (gicdev,"revision" ,2 ); + +for (i =0 ; i < vbi->smp_cpus; i++) { +CPUState *cpu =qemu_get_cpu (i); +CPUARMState *env = cpu->env_ptr; + env->nvic = gicdev; + } + +qdev_prop_set_uint32 (gicdev,"revision" , + type ==KVM_DEV_TYPE_ARM_VGIC_V3 ?3 :2 );qdev_prop_set_uint32 (gicdev,"num-cpu" , smp_cpus);/* Note that the num-irq property counts both internal and external * interrupts; there are always 32 of the former (mandated by GIC spec).
@@ -382,7 +429,11 @@ static void create_gic(VirtBoardInfo *vbi, qemu_irq *pic)qdev_init_nofail (gicdev); gicbusdev =SYS_BUS_DEVICE (gicdev);sysbus_mmio_map (gicbusdev,0 , vbi->memmap[VIRT_GIC_DIST ].base); -sysbus_mmio_map (gicbusdev,1 , vbi->memmap[VIRT_GIC_CPU ].base); +if (type ==KVM_DEV_TYPE_ARM_VGIC_V3 ) { +sysbus_mmio_map (gicbusdev,1 , vbi->memmap[VIRT_LPI ].base); + }else { +sysbus_mmio_map (gicbusdev,1 , vbi->memmap[VIRT_GIC_CPU ].base); + }/* Wire the outputs from each CPU's generic timer to the * appropriate GIC PPI inputs, and the GIC's IRQ output to
@@ -409,9 +460,11 @@ static void create_gic(VirtBoardInfo *vbi, qemu_irq *pic) pic[i] =qdev_get_gpio_in (gicdev, i); } -fdt_add_gic_node (vbi); +fdt_add_gic_node (vbi, type); -create_v2m (vbi, pic); +if (type ==KVM_DEV_TYPE_ARM_VGIC_V2 ) { +create_v2m (vbi, pic); + } }static void create_uart (const VirtBoardInfo *vbi,qemu_irq *pic)
@@ -715,7 +768,10 @@ static void create_pcie(const VirtBoardInfo *vbi, qemu_irq *pic)qemu_fdt_setprop_cells (vbi->fdt, nodename,"bus-range" ,0 , nr_pcie_buses -1 ); -qemu_fdt_setprop_cells (vbi->fdt, nodename,"msi-parent" , vbi->v2m_phandle); +if (vbi->v2m_phandle) { +qemu_fdt_setprop_cells (vbi->fdt, nodename,"msi-parent" , + vbi->v2m_phandle); + }qemu_fdt_setprop_sized_cells (vbi->fdt, nodename,"reg" ,2 , base_ecam,2 , size_ecam);
@@ -875,7 +931,7 @@ static void machvirt_init(MachineState *machine)create_flash (vbi); -create_gic (vbi, pic); +create_gic (vbi, pic, machine->kernel_irqchip_type);create_uart (vbi, pic);
@@ -933,6 +989,37 @@ static void virt_set_secure(Object *obj, bool value, Error **errp) vms->secure = value; } +static void virt_get_gic_version (Object *obj,Visitor *v,void *opaque, +const char *name,Error **errp) +{ +MachineState *ms =MACHINE (obj); +int32_t value = (ms->kernel_irqchip_type ==KVM_DEV_TYPE_ARM_VGIC_V3 ) ? +3 :2 ; + +visit_type_int32 (v, &value, name, errp); +} + +static void virt_set_gic_version (Object *obj,Visitor *v,void *opaque, +const char *name,Error **errp) +{ +MachineState *ms =MACHINE (obj); +int32_t value; + +visit_type_int32 (v, &value, name, errp); + +switch (value) { +case 3 : + ms->kernel_irqchip_type =QEMU_GIC_TYPE_V3 ; +break ; +case 2 : + ms->kernel_irqchip_type =QEMU_GIC_TYPE_V2 ; +break ; +default : +error_report ("Only GICv2 and GICv3 supported currently\n" ); +exit (1 ); + } +} +static void virt_instance_init (Object *obj) {VirtMachineState *vms =VIRT_MACHINE (obj);
@@ -948,6 +1035,11 @@ static void virt_instance_init(Object *obj)/* Default GIC type is v2 */ vms->parent.kernel_irqchip_type =QEMU_GIC_TYPE_V2 ; +object_property_add (obj,"gicversion" ,"int" , virt_get_gic_version, + virt_set_gic_version,NULL ,NULL ,NULL ); +object_property_set_description (obj,"gicversion" , +"Set GIC version. " +"Valid values are 2 and 3" ,NULL ); }static void virt_class_init (ObjectClass *oc,void *data)
@@ -957,7 +1049,8 @@ static void virt_class_init(ObjectClass *oc, void *data) mc->name =TYPE_VIRT_MACHINE ; mc->desc ="ARM Virtual Machine" , mc->init = machvirt_init; - mc->max_cpus =8 ; +/* With gic3 full implementation (with bitops) rase the lmit to 128 */ + mc->max_cpus =64 ; mc->has_dynamic_sysbus = true; mc->block_default_type =IF_VIRTIO ; mc->no_cdrom =1 ;
include/hw/arm/fdt.h +1 -1
@@ -29,6 +29,6 @@#define GIC_FDT_IRQ_FLAGS_LEVEL_LO 8#define GIC_FDT_IRQ_PPI_CPU_START 8 -#define GIC_FDT_IRQ_PPI_CPU_WIDTH 8 +#define GIC_FDT_IRQ_PPI_CPU_WIDTH 24#endif
include/hw/arm/virt.h +5 -1
@@ -46,6 +46,11 @@ enum {VIRT_CPUPERIPHS ,VIRT_GIC_DIST ,VIRT_GIC_CPU , +VIRT_GIC_V2M , +VIRT_GIC_DIST_MBI =VIRT_GIC_CPU , +VIRT_ITS_CONTROL =VIRT_GIC_V2M , +VIRT_ITS_TRANSLATION , +VIRT_LPI ,VIRT_UART ,VIRT_MMIO , VIRT_RTC,
@@ -54,7 +59,6 @@ enum {VIRT_PCIE_MMIO ,VIRT_PCIE_PIO ,VIRT_PCIE_ECAM , -VIRT_GIC_V2M , VIRT_PLATFORM_BUS, };
Signature · 1 line
1.9.5.msysgit.0
Re: [Qemu-devel] [PATCH v7 6/6] Add gicversion option to virt machine
Quoted 8 lines · Pavel Fedin
Set
kernel_irqchip_typeaccording to value of the option and pass it around where necessary. Instantiate devices and fdt nodes according to the choice.
max_cpusfor virt machine increased to 64. GICv2 compatibility check happens insidearm_gic_common_realize().gicdev = qdev_create(NULL, gictype);
hw/arm/virt.c static void create_v2m(VirtBoardInfo *vbi, qemu_irq *pic) · line 420
Quoted 7 lines
hw/arm/virt.c +6
-1-qdev_prop_set_uint32 (gicdev,"revision" ,2 ); + +for (i =0 ; i < vbi->smp_cpus; i++) { +CPUState *cpu =qemu_get_cpu (i); +CPUARMState *env = cpu->env_ptr; + env->nvic = gicdev; + }
We definitely need to come up with a something cleaner than this (which is ugly for two reasons -- firstly for borrowing the nvic pointer and secondly because we shouldn't just be putting random pointers between the two like this, ideally; and we definitely don't want the board code to have to do it at this level.)
-- PMM
Re: [Qemu-devel] [PATCH v7 6/6] Add gicversion option to virt machine
Hello!
Quoted 11 lines
hw/arm/virt.c static void create_v2m(VirtBoardInfo *vbi, qemu_irq *pic) · line 420
Quoted 8 lines
gicdev = qdev_create(NULL, gictype);hw/arm/virt.c +6
-1-qdev_prop_set_uint32 (gicdev,"revision" ,2 ); + +for (i =0 ; i < vbi->smp_cpus; i++) { +CPUState *cpu =qemu_get_cpu (i); +CPUARMState *env = cpu->env_ptr; + env->nvic = gicdev; + }We definitely need to come up with a something cleaner than this (which is ugly for two reasons
This could be done:
a) as property
b) as global variable because 'gicdev' is a single of its kind.
But, actually, this is currently only for TCG, which needs it in order to forward system register accesses to GICv3 code. Would it be OK if i just omit this assignment ?
Kind regards,
Pavel Fedin
Expert Engineer
Samsung Electronics Research center Russia
Re: [Qemu-devel] [PATCH v7 6/6] Add gicversion option to virt machine
Quoted 14 lines · Pavel Fedin
Hello!
Quoted 11 lines
hw/arm/virt.c static void create_v2m(VirtBoardInfo *vbi, qemu_irq *pic) · line 420
Quoted 8 lines
gicdev = qdev_create(NULL, gictype);hw/arm/virt.c +6
-1-qdev_prop_set_uint32 (gicdev,"revision" ,2 ); + +for (i =0 ; i < vbi->smp_cpus; i++) { +CPUState *cpu =qemu_get_cpu (i); +CPUARMState *env = cpu->env_ptr; + env->nvic = gicdev; + }We definitely need to come up with a something cleaner than this (which is ugly for two reasons
Quoted 7 lines
This could be done:
a) as property
b) as global variable because 'gicdev' is a single of its kind.But, actually, this is currently only for TCG, which needs it in
order to forward system register accesses to GICv3 code. Would it be OK if i just omit this assignment ?
Yes, I think so.
I'm surprised we tell the CPU about the GIC pointer for the system register stuff -- I was expecting that we'd give the GIC a CPU pointer. (We could in theory implement some equivalent of the architected protocol between the redistributors and the CPU interfaces, but I think that would be overkill.)
-- PMM
Re: [Qemu-devel] [PATCH v7 6/6] Add gicversion option to virt machine
Quoted 26 lines · Peter Maydell
Quoted 14 lines · On 3 August 2015 at 08:19, Pavel Fedin <[email protected] <javascript:;>>…
Hello!
Quoted 11 lines
hw/arm/virt.c static void create_v2m(VirtBoardInfo *vbi, qemu_irq *pic) · line 420
Quoted 8 lines
gicdev = qdev_create(NULL, gictype);hw/arm/virt.c +6
-1-qdev_prop_set_uint32 (gicdev,"revision" ,2 ); + +for (i =0 ; i < vbi->smp_cpus; i++) { +CPUState *cpu =qemu_get_cpu (i); +CPUARMState *env = cpu->env_ptr; + env->nvic = gicdev; + }We definitely need to come up with a something cleaner than this (which is ugly for two reasons
Quoted 7 lines
This could be done:
a) as property
b) as global variable because 'gicdev' is a single of its kind.But, actually, this is currently only for TCG, which needs it in
order to forward system register accesses to GICv3 code. Would it be OK if i just omit this assignment ?Yes, I think so.
Quoted 8 lines
I'm surprised we tell the CPU about the GIC pointer for the system register stuff -- I was expecting that we'd give the GIC a CPU pointer. (We could in theory implement some equivalent of the architected protocol between the redistributors and the CPU interfaces, but I think that would be overkill.)
-- PMM
The problem is that each function added as a system's instruction helper to target-arm/cpu64.c has the signature of void set(CPUARMState *env, ARMCPRegInfo *ri, uint64_t value) or uint64_t get(CPUARMState *env, ARMCPRegInfo *ri)
I just mimicked the way armv7m_nvic_XXXX API works.
So in a sense the CPU must be familiar with the GIC (as an opaque object of course).
Best regards,
S.P.
Re: [Qemu-devel] [PATCH v7 6/6] Add gicversion option to virt machine
Quoted 9 lines · Shlomo Pongratz
Quoted 5 lines · Peter Maydell
I'm surprised we tell the CPU about the GIC pointer for the system register stuff -- I was expecting that we'd give the GIC a CPU pointer. (We could in theory implement some equivalent of the architected protocol between the redistributors and the CPU interfaces, but I think that would be overkill.)
Quoted 5 lines
The problem is that each function added as a system's instruction helper to
target-arm/cpu64.chas the signature of voidset(CPUARMState *env, ARMCPRegInfo *ri, uint64_t value)oruint64_tget(CPUARMState *env, ARMCPRegInfo *ri)
I just mimicked the wayarmv7m_nvic_XXXXAPI works.
The v7M NVIC is a terrible example to copy -- it is ancient code mostly written before QEMU acquired various useful abstraction layers, and has received very little maintenance since then.
Quoted 2 lines
So in a sense the CPU must be familiar with the GIC (as an opaque object of course).
If the GIC just registers a set of system register information with the CPU then the CPU doesn't need to know that it's the GIC providing those system register functions. (You can use define_arm_cp_regs_with_opaque() to define registers such that your callbacks can fish an opaque data pointer back out of ri->opaque, so you can get back to the GIC data structures.)
-- PMM