Split in-kernel devices
cover Split in-kernel devices
This is the final version of the in-kernel devices support for kvm. Series do not yet include any switch to disable it. But I had that last patch floating around, and as soon as we set up on the best method to do it (special machine, cmd line switch, whatever), I can resend.
I believe this model provides really ellegant devices, and it is much easier to understand than previous code.
PATCH 1/8 introduce VMSTATE_U64
Slightly modified version of a patch already included in qemu-kvm:
This is a patch actually written by Juan, which, according to him, he plans on posting to qemu.git. Problem is that linux defines u64 in a way that is type-uncompatible with uint64_t.
I am including it here, because it is a dependency to my patch series that follows.
--- hw/hw.h | 24 ++++++++++++++++++++++++ savevm.c | 23 +++++++++++++++++++++++ 2 files changed, 47 insertions(+), 0 deletions(-)
hw/hw.h +24 -0
@@ -332,6 +332,11 @@ extern const VMStateInfo vmstate_info_uint16;extern const VMStateInfo vmstate_info_uint32;extern const VMStateInfo vmstate_info_uint64; +#ifdef __linux__ +#include <linux/types.h> +extern const VMStateInfo vmstate_info_u64; +#endif +extern const VMStateInfo vmstate_info_timer;extern const VMStateInfo vmstate_info_ptimer;extern const VMStateInfo vmstate_info_buffer;
@@ -388,6 +393,16 @@ extern const VMStateInfo vmstate_info_unused_buffer; .offset = vmstate_offset_array(_state, _field, _type, _num), \ } +#define VMSTATE_ARRAY_UNSAFE (_field, _state, _num, _version, _info, _type) {\ + .name = (stringify(_field)), \ + .version_id = (_version), \ + .num = (_num), \ + .info = &(_info), \ + .size = sizeof(_type), \ + .flags = VMS_ARRAY, \ + .offset = offsetof(_state, _field) \ +} +#define VMSTATE_ARRAY_TEST (_field, _state, _num, _test, _info, _type) {\ .name = (stringify(_field)), \ .field_exists = (_test), \
@@ -580,6 +595,15 @@ extern const VMStateDescription vmstate_i2c_slave;#define VMSTATE_UINT64 (_f, _s) \ VMSTATE_UINT64_V(_f, _s, 0) +/* This is needed because on linux __u64 is unsigned long long +and on glibc uint64_t is unsigned long on 64 bits */ +#ifdef __linux__ +#define VMSTATE_U64_V (_f, _s, _v) \ + VMSTATE_SINGLE(_f, _s, _v, vmstate_info_u64, __u64) +#define VMSTATE_U64 (_f, _s) \ + VMSTATE_U64_V(_f, _s, 0) +#endif +#define VMSTATE_UINT8_EQUAL (_f, _s) \ VMSTATE_SINGLE(_f, _s, 0, vmstate_info_uint8_equal, uint8_t)
savevm.c +23 -0
@@ -856,6 +856,29 @@ const VMStateInfo vmstate_info_uint64 = { .put = put_uint64, }; +/* 64 bit linux kernel unsigned int */ + +#ifdef __linux__ +static int get_u64 (QEMUFile *f,void *pv,size_t size) +{ +__u64 *v = pv; +qemu_get_be64s (f, (uint64_t *)v); +return 0 ; +} + +static void put_u64 (QEMUFile *f,void *pv,size_t size) +{ +__u64 *v = pv; +qemu_put_be64s (f, (uint64_t *)v); +} + +const VMStateInfo vmstate_info_u64 = { + .name ="__u64" , + .get = get_u64, + .put = put_u64, +}; +#endif /* __linux__ */ +/* 8 bit int. See that the received value is the same than the one in the field */
Signature · 1 line
1.6.2.5
PATCH 2/8 Provide ioapic-kvm
This patch provides the file ioapic-kvm.c, which implements a schim over the kvm in-kernel IO APIC.
--- Makefile.target | 2 + hw/ioapic-kvm.c | 90 +++++++++++++++++++++++++++++++++++++++++++++++++++++++ hw/pc.c | 7 ++++- hw/pc.h | 2 + kvm-all.c | 20 ++++++++++++ kvm.h | 4 ++ 6 files changed, 124 insertions(+), 1 deletions(-) create mode 100644 hw/ioapic-kvm.c
Makefile.target +2 -0
@@ -197,6 +197,8 @@ obj-i386-y += usb-uhci.o vmmouse.o vmport.o vmware_vga.o hpet.o obj-i386-y += device-hotplug.o pci-hotplug.o smbios.o wdt_ib700.o obj-i386-y += ne2000-isa.o +obj-i386-$(CONFIG_KVM) += ioapic-kvm.o + # shared objects obj-ppc-y = ppc.o ide/core.o ide/qdev.o ide/isa.o ide/pci.o ide/macio.o obj-ppc-y += ide/cmd646.o
hw/ioapic-kvm.c new file +90 -0
90 lines not on this page · raw message
hw/pc.c +6 -1
@@ -39,6 +39,7 @@#include "ide.h" #include "loader.h" #include "elf.h" +#include "kvm.h" /* output Bochs bios info messages */ //#define DEBUG_BIOS
@@ -1200,7 +1201,11 @@ static void pc_init1(ram_addr_t ram_size,register_ioport_write (0x92 ,1 ,1 , ioport92_write,NULL );if (pci_enabled) { - isa_irq_state->ioapic =ioapic_init (); +if (kvm_enabled () &&kvm_irqchip_in_kernel ()) { + isa_irq_state->ioapic =kvm_ioapic_init (); + }else { + isa_irq_state->ioapic =ioapic_init (); + } } pit =pit_init (0x40 ,isa_reserve_irq (0 ));pcspk_init (pit);
hw/pc.h +2 -0
@@ -48,6 +48,8 @@ void ioapic_set_irq(void *opaque, int vector, int level);void apic_reset_irq_delivered (void );int apic_get_irq_delivered (void ); +qemu_irq *kvm_ioapic_init (void ); +/* i8254.c */ #define PIT_FREQ 1193182
kvm-all.c +20 -0
@@ -411,6 +411,26 @@ int kvm_check_extension(KVMState *s, unsigned int extension)return ret; } +#ifdef KVM_CAP_IRQCHIP +int kvm_set_irqchip (struct kvm_irqchip *chip) +{ +if (!kvm_state->irqchip_in_kernel) { +return 0 ; + } + +return kvm_vm_ioctl (kvm_state,KVM_SET_IRQCHIP , chip); +} + +int kvm_get_irqchip (struct kvm_irqchip *chip) +{ +if (!kvm_state->irqchip_in_kernel) { +return 0 ; + } + +return kvm_vm_ioctl (kvm_state,KVM_GET_IRQCHIP , chip); +} +#endif +int kvm_init (int smp_cpus) { static const char upgrade_note[] =
kvm.h +4 -0
@@ -16,6 +16,7 @@#include "config.h" #include "qemu-queue.h" +#include <linux/kvm.h> #ifdef CONFIG_KVM extern int kvm_allowed;
@@ -63,6 +64,9 @@ int kvm_update_guest_debug(CPUState *env, unsigned long reinject_trap);int kvm_pit_in_kernel (void );int kvm_irqchip_in_kernel (void ); +int kvm_set_irqchip (struct kvm_irqchip *chip); +int kvm_get_irqchip (struct kvm_irqchip *chip); +/* internal API */ struct KVMState ;
Signature · 1 line
1.6.2.5
PATCH 3/8 provide apic_set_irq_delivered
i8259 chip will use it, so provide it, and export it through pc.h
--- hw/apic.c | 5 +++++ hw/pc.h | 1 + 2 files changed, 6 insertions(+), 0 deletions(-)
hw/apic.c +5 -0
@@ -388,6 +388,11 @@ void apic_reset_irq_delivered(void) apic_irq_delivered =0 ; } +void apic_set_irq_delivered (void ) +{ + apic_irq_delivered =1 ; +} +int apic_get_irq_delivered (void ) {return apic_irq_delivered;
hw/pc.h +1 -0
@@ -46,6 +46,7 @@ int apic_get_interrupt(CPUState *env);qemu_irq *ioapic_init (void );void ioapic_set_irq (void *opaque,int vector,int level);void apic_reset_irq_delivered (void ); +void apic_set_irq_delivered (void );int apic_get_irq_delivered (void );qemu_irq *kvm_ioapic_init (void );
Signature · 1 line
1.6.2.5
PATCH 4/8 provide i8259-kvm
This patch provides the file i8259-kvm.c, which implements a schim over the kvm in-kernel PIC.
--- Makefile.target | 2 +- hw/i8259-kvm.c | 112 +++++++++++++++++++++++++++++++++++++++++++++++++++++++ hw/pc.c | 8 +++- hw/pc.h | 1 + kvm-all.c | 26 ++++++++++++- kvm.h | 2 + 6 files changed, 147 insertions(+), 4 deletions(-) create mode 100644 hw/i8259-kvm.c
Makefile.target +1 -1
@@ -197,7 +197,7 @@ obj-i386-y += usb-uhci.o vmmouse.o vmport.o vmware_vga.o hpet.o obj-i386-y += device-hotplug.o pci-hotplug.o smbios.o wdt_ib700.o obj-i386-y += ne2000-isa.o -obj-i386-$(CONFIG_KVM) += ioapic-kvm.o +obj-i386-$(CONFIG_KVM) += ioapic-kvm.o i8259-kvm.o # shared objects obj-ppc-y = ppc.o ide/core.o ide/qdev.o ide/isa.o ide/pci.o ide/macio.o
hw/i8259-kvm.c new file +112 -0
112 lines not on this page · raw message
hw/pc.c +7 -1
@@ -1154,8 +1154,14 @@ static void pc_init1(ram_addr_t ram_size, } cpu_irq =qemu_allocate_irqs (pic_irq_request,NULL ,1 ); - i8259 =i8259_init (cpu_irq[0 ]); isa_irq_state =qemu_mallocz (sizeof (*isa_irq_state)); + +if (kvm_enabled () &&kvm_irqchip_in_kernel ()) { + i8259 =kvm_i8259_init (cpu_irq[0 ]); + }else { + i8259 =i8259_init (cpu_irq[0 ]); + } + isa_irq_state->i8259 = i8259; isa_irq =qemu_allocate_irqs (isa_irq_handler, isa_irq_state,24 );
hw/pc.h +1 -0
@@ -33,6 +33,7 @@ uint32_t pic_intack_read(PicState2 *s);void pic_info (Monitor *mon);void irq_info (Monitor *mon); +qemu_irq *kvm_i8259_init (qemu_irq parent_irq);/* APIC */ typedef struct IOAPICState IOAPICState ; void apic_deliver_irq(uint8_t dest, uint8_t dest_mode,
kvm-all.c +24 -2
@@ -411,7 +411,6 @@ int kvm_check_extension(KVMState *s, unsigned int extension)return ret; } -#ifdef KVM_CAP_IRQCHIP int kvm_set_irqchip (struct kvm_irqchip *chip) {if (!kvm_state->irqchip_in_kernel) {
@@ -429,7 +428,30 @@ int kvm_get_irqchip(struct kvm_irqchip *chip)return kvm_vm_ioctl (kvm_state,KVM_GET_IRQCHIP , chip); } -#endif + +int kvm_set_irq (int irq,int level,int *status) +{ +struct kvm_irq_level event; +int r; + +if (!kvm_state->irqchip_in_kernel) { +return 0 ; + } + + event.level = level; + event.irq = irq; + + r =kvm_vm_ioctl (kvm_state,KVM_IRQ_LINE_STATUS , &event); + +if (r <0 ) +return 0 ; + +if (status) { + *status = event.status; + } + +return 1 ; +}int kvm_init (int smp_cpus) {
kvm.h +2 -0
@@ -67,6 +67,8 @@ int kvm_irqchip_in_kernel(void);int kvm_set_irqchip (struct kvm_irqchip *chip);int kvm_get_irqchip (struct kvm_irqchip *chip); +int kvm_set_irq (int irq,int level,int *status); +/* internal API */ struct KVMState ;
Signature · 1 line
1.6.2.5
PATCH 5/8 Don't call apic functions directly from kvm code
It is actually not necessary to call a tpr function to save and load cr8, as cr8 is part of the processor state, and thus, it is much easier to just add it to CPUState.
As for apic base, wrap kvm usages, so we can call either the qemu device, or the in kernel version.
--- target-i386/cpu.h | 1 + target-i386/kvm.c | 25 +++++++++++++++++++------ 2 files changed, 20 insertions(+), 6 deletions(-)
target-i386/cpu.h +1 -0
@@ -606,6 +606,7 @@ typedef struct CPUX86State {SegmentCache idt;/* only base and limit are used */ target_ulong cr[5 ];/* NOTE: cr1 is unused */ +target_ulong cr8;int32_t a20_mask;/* FPU state */
target-i386/kvm.c +19 -6
@@ -338,6 +338,19 @@ static void get_seg(SegmentCache *lhs, const struct kvm_segment *rhs) | (rhs->avl * DESC_AVL_MASK); } +static void kvm_set_apic_base (CPUState *env,uint64_t val) +{ +if (!kvm_irqchip_in_kernel ()) +cpu_set_apic_base (env, val); +} + +static uint64_t kvm_get_apic_base (CPUState *env) +{ +if (!kvm_irqchip_in_kernel ()) +return cpu_get_apic_base (env); +return 0 ; +} +static void kvm_getput_reg (__u64 *kvm_reg,target_ulong *qemu_reg,int set) {if (set)
@@ -447,8 +460,8 @@ static int kvm_put_sregs(CPUState *env) sregs.cr3 = env->cr[3 ]; sregs.cr4 = env->cr[4 ]; - sregs.cr8 =cpu_get_apic_tpr (env); - sregs.apic_base =cpu_get_apic_base (env); + sregs.cr8 = env->cr8; + sregs.apic_base =kvm_get_apic_base (env); sregs.efer = env->efer;
@@ -546,7 +559,7 @@ static int kvm_get_sregs(CPUState *env) env->cr[3 ] = sregs.cr3; env->cr[4 ] = sregs.cr4; -cpu_set_apic_base (env, sregs.apic_base); +kvm_set_apic_base (env, sregs.apic_base); env->efer = sregs.efer;//cpu_set_apic_tpr(env, sregs.cr8);
@@ -742,7 +755,7 @@ int kvm_arch_pre_run(CPUState *env, struct kvm_run *run) run->request_interrupt_window =0 ;dprintf ("setting tpr\n" ); - run->cr8 =cpu_get_apic_tpr (env); + run->cr8 = env->cr8;return 0 ; }
@@ -754,8 +767,8 @@ int kvm_arch_post_run(CPUState *env, struct kvm_run *run)else env->eflags &= ~IF_MASK ; -cpu_set_apic_tpr (env, run->cr8); -cpu_set_apic_base (env, run->apic_base); + env->cr8 = run->cr8; +kvm_set_apic_base (env, run->apic_base);return 0 ; }
Signature · 1 line
1.6.2.5
PATCH 6/8 provide apic-kvm
This patch provides the file apic-kvm.c, which implements a schim over the kvm in-kernel APIC.
--- Makefile.target | 2 +- hw/apic-kvm.c | 157 +++++++++++++++++++++++++++++++++++++++++++++++++++++ hw/pc.c | 6 ++- hw/pc.h | 2 + kvm.h | 3 + target-i386/cpu.h | 4 ++ target-i386/kvm.c | 23 ++++++++- 7 files changed, 194 insertions(+), 3 deletions(-) create mode 100644 hw/apic-kvm.c
Makefile.target +1 -1
@@ -197,7 +197,7 @@ obj-i386-y += usb-uhci.o vmmouse.o vmport.o vmware_vga.o hpet.o obj-i386-y += device-hotplug.o pci-hotplug.o smbios.o wdt_ib700.o obj-i386-y += ne2000-isa.o -obj-i386-$(CONFIG_KVM) += ioapic-kvm.o i8259-kvm.o +obj-i386-$(CONFIG_KVM) += ioapic-kvm.o i8259-kvm.o apic-kvm.o # shared objects obj-ppc-y = ppc.o ide/core.o ide/qdev.o ide/isa.o ide/pci.o ide/macio.o
hw/apic-kvm.c new file +157 -0
157 lines not on this page · raw message
hw/pc.c +5 -1
@@ -1018,7 +1018,11 @@ static CPUState *pc_new_cpu(const char *cpu_model)if ((env->cpuid_features &CPUID_APIC ) || smp_cpus >1 ) { env->cpuid_apic_id = env->cpu_index;/* APIC reset callback resets cpu */ -apic_init (env); +if (kvm_enabled () &&kvm_irqchip_in_kernel ()) { +kvm_apic_init (env); + }else { +apic_init (env); + } }else {qemu_register_reset ((QEMUResetHandler *)cpu_reset, env); }
hw/pc.h +2 -0
@@ -41,6 +41,8 @@ void apic_deliver_irq(uint8_t dest, uint8_t dest_mode, uint8_t vector_num, uint8_t polarity, uint8_t trigger_mode);int apic_init (CPUState *env); +int kvm_apic_init (CPUState *env); +int apic_accept_pic_intr (CPUState *env);void apic_deliver_pic_intr (CPUState *env,int level);int apic_get_interrupt (CPUState *env);
kvm.h +3 -0
@@ -99,6 +99,9 @@ int kvm_arch_init(KVMState *s, int smp_cpus);int kvm_arch_init_vcpu (CPUState *env); +int kvm_set_lapic (CPUState *env,struct kvm_lapic_state *s); +int kvm_get_lapic (CPUState *env,struct kvm_lapic_state *s); +struct kvm_guest_debug ;struct kvm_debug_exit_arch ;
target-i386/cpu.h +4 -0
@@ -855,6 +855,10 @@ void cpu_set_apic_tpr(CPUX86State *env, uint8_t val);uint8_t cpu_get_apic_tpr (CPUX86State *env);#endif +/* hw/apic-kvm.c */ +void kvm_apic_set_base (CPUX86State *env,uint64_t val); +uint64_t kvm_apic_get_base (CPUX86State *env); +/* hw/pc.c */ void cpu_smm_update (CPUX86State *env);uint64_t cpu_get_tsc (CPUX86State *env);
target-i386/kvm.c +22 -1
@@ -342,13 +342,16 @@ static void kvm_set_apic_base(CPUState *env, uint64_t val) {if (!kvm_irqchip_in_kernel ())cpu_set_apic_base (env, val); +else +kvm_apic_set_base (env, val); }static uint64_t kvm_get_apic_base (CPUState *env) {if (!kvm_irqchip_in_kernel ())return cpu_get_apic_base (env); -return 0 ; +else +return kvm_apic_get_base (env); }static void kvm_getput_reg (__u64 *kvm_reg,target_ulong *qemu_reg,int set)
@@ -970,3 +973,21 @@ void kvm_arch_update_guest_debug(CPUState *env, struct kvm_guest_debug *dbg) } }#endif /* KVM_CAP_SET_GUEST_DEBUG */ + +#ifdef KVM_CAP_IRQCHIP +int kvm_set_lapic (CPUState *env,struct kvm_lapic_state *s) +{ +if (!kvm_irqchip_in_kernel ()) +return 0 ; + +return kvm_vcpu_ioctl (env,KVM_SET_LAPIC , s); +} + +int kvm_get_lapic (CPUState *env,struct kvm_lapic_state *s) +{ +if (!kvm_irqchip_in_kernel ()) +return 0 ; + +return kvm_vcpu_ioctl (env,KVM_GET_LAPIC , s); +} +#endif
Signature · 1 line
1.6.2.5
PATCH 7/8 Initialize in-kernel irqchip
Now that we have all devices set up, this patch initializes the irqchip. This is dependant on the io-thread, since we need someone to pull ourselves out of the halted state.
I believe this should be the default when we are running over the io-thread. Later on, I plan to post a patch that makes it optional.
--- kvm-all.c | 20 ++++++++++++++++++++ 1 files changed, 20 insertions(+), 0 deletions(-)
kvm-all.c +20 -0
@@ -453,6 +453,22 @@ int kvm_set_irq(int irq, int level, int *status)return 1 ; } +static int kvm_create_irqchip (KVMState *s) +{ +int ret =0 ; +#if defined(CONFIG_IOTHREAD ) +if (!kvm_check_extension (s,KVM_CAP_IRQCHIP )) +return -1 ; + + ret =kvm_vm_ioctl (s,KVM_CREATE_IRQCHIP ); +if (ret <0 ) +return ret; + + s->irqchip_in_kernel =1 ; +#endif +return ret; +} +int kvm_init (int smp_cpus) { static const char upgrade_note[] =
@@ -539,6 +555,10 @@ int kvm_init(int smp_cpus) }#endif + ret =kvm_create_irqchip (s); +if (ret <0 ) + goto err; + ret =kvm_arch_init (s, smp_cpus);if (ret <0 ) goto err;
Signature · 1 line
1.6.2.5
PATCH 8/8 Do GSI routing
To follow correctly what our bios ACPI tables say, we have to be able to program our irqchips with GSI routing mappings. This support is already in qemu-kvm
--- kvm-all.c | 6 ++- kvm.h | 3 + target-i386/kvm.c | 118 +++++++++++++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 125 insertions(+), 2 deletions(-)
kvm-all.c +4 -2
@@ -66,6 +66,8 @@ struct KVMState#endif int irqchip_in_kernel;int pit_in_kernel; + +KVMArchState *arch_state; };static KVMState *kvm_state;
@@ -559,12 +561,12 @@ int kvm_init(int smp_cpus)if (ret <0 ) goto err; + kvm_state = s; + ret =kvm_arch_init (s, smp_cpus);if (ret <0 ) goto err; - kvm_state = s; -return 0 ; err:
kvm.h +3 -0
@@ -72,7 +72,10 @@ int kvm_set_irq(int irq, int level, int *status);/* internal API */ struct KVMState ; +struct KVMArchState ; +typedef struct KVMState KVMState ; +typedef struct KVMArchState KVMArchState ;int kvm_ioctl (KVMState *s,int type, ...);
target-i386/kvm.c +118 -0
134 lines not on this page · raw message
Signature · 1 line
1.6.2.5