kvm: add support for guest physical bits
v2 cover kvm: add support for guest physical bits
The matching kernel bits are here:
https://lore.kernel.org/kvm/[email protected]/T/
Gerd Hoffmann (2):
[debug] log kvm supported cpuid
kvm: add support for guest physical bits
target/i386/cpu.h | 1 + target/i386/cpu.c | 1 + target/i386/kvm/kvm.c | 31 +++++++++++++++++++++++++++++++ 3 files changed, 33 insertions(+)
Signature · 1 line
2.44.0
PATCH v2 1/2 log kvm supported cpuid
--- target/i386/kvm/kvm.c | 14 ++++++++++++++ 1 file changed, 14 insertions(+)
target/i386/kvm/kvm.c +14 -0
@@ -276,6 +276,20 @@ static struct kvm_cpuid2 *try_get_cpuid(KVMState *s, int max)exit (1 ); } } + + { +int i; + +for (i =0 ; i < cpuid->nent; i++) { +fprintf (stderr,"cpuid: %8x/%d - %8x %8x %8x %8x\n" , + cpuid->entries[i].function, + cpuid->entries[i].index, + cpuid->entries[i].eax, + cpuid->entries[i].ebx, + cpuid->entries[i].ecx, + cpuid->entries[i].edx); + } + }return cpuid; }
Signature · 1 line
2.44.0
Re: [PATCH v2 1/2] [debug] log kvm supported cpuid
Quoted 1 line
target/i386/kvm/kvm.c | 14 ++++++++++++++
Oops, that was not meant to be posted.
Please ignore and look at patch 2/2.
thanks,
Gerd
PATCH v2 2/2 kvm: add support for guest physical bits
In reply to Gerd HoffmannQuery kvm for supported guest physical address bits, in cpuid function 80000008, eax[23:16]. Usually this is identical to host physical address bits. With NPT or EPT being used this might be restricted to 48 (max 4-level paging address space size) even if the host cpu supports more physical address bits.
When set pass this to the guest, using cpuid too. Guest firmware can use this to figure how big the usable guest physical address space is, so PCI bar mapping are actually reachable.
--- target/i386/cpu.h | 1 + target/i386/cpu.c | 1 + target/i386/kvm/kvm.c | 17 +++++++++++++++++ 3 files changed, 19 insertions(+)
target/i386/cpu.h +1 -0
@@ -2026,6 +2026,7 @@ struct ArchCPU {/* Number of physical address bits supported */ uint32_t phys_bits; +uint32_t guest_phys_bits;/* in order to simplify APIC support, we leave this pointer to the user */
target/i386/cpu.c +1 -0
@@ -6570,6 +6570,7 @@ void cpu_x86_cpuid(CPUX86State *env, uint32_t index, uint32_t count,if (env->features[FEAT_8000_0001_EDX ] &CPUID_EXT2_LM ) {/* 64 bit processor */ *eax |= (cpu_x86_virtual_addr_width (env) <<8 ); + *eax |= (cpu->guest_phys_bits <<16 ); } *ebx = env->features[FEAT_8000_0008_EBX ];if (cs->nr_cores * cs->nr_threads >1 ) {
target/i386/kvm/kvm.c +17 -0
@@ -238,6 +238,15 @@ static int kvm_get_tsc(CPUState *cs)return 0 ; } +/* return cpuid fn 8000_0008 eax[23:16] aka GuestPhysBits */ +static int kvm_get_guest_phys_bits (KVMState *s) +{ +uint32_t eax; + + eax =kvm_arch_get_supported_cpuid (s,0x80000008 ,0 ,R_EAX ); +return (eax >>16 ) &0xff ; +} +static inline void do_kvm_synchronize_tsc (CPUState *cpu,run_on_cpu_data arg) {kvm_get_tsc (cpu);
@@ -1730,6 +1739,7 @@ int kvm_arch_init_vcpu(CPUState *cs)X86CPU *cpu =X86_CPU (cs);CPUX86State *env = &cpu->env;uint32_t limit, i, j, cpuid_i; +uint32_t guest_phys_bits;uint32_t unused;struct kvm_cpuid_entry2 *c;uint32_t signature[3 ];
@@ -1765,6 +1775,13 @@ int kvm_arch_init_vcpu(CPUState *cs) env->apic_bus_freq =KVM_APIC_BUS_FREQUENCY ; + guest_phys_bits =kvm_get_guest_phys_bits (cs->kvm_state); +if (guest_phys_bits && + (cpu->guest_phys_bits ==0 || + cpu->guest_phys_bits > guest_phys_bits)) { + cpu->guest_phys_bits = guest_phys_bits; + } +/* * kvm_hyperv_expand_features() is called here for the second time in case * KVM_CAP_SYS_HYPERV_CPUID is not supported. While we can't possibly handle
Signature · 1 line
2.44.0
Quoted 70 lines · Gerd Hoffmann
Query kvm for supported guest physical address bits, in cpuid function 80000008, eax[23:16]. Usually this is identical to host physical address bits. With NPT or EPT being used this might be restricted to 48 (max 4-level paging address space size) even if the host cpu supports more physical address bits.
When set pass this to the guest, using cpuid too. Guest firmware can use this to figure how big the usable guest physical address space is, so PCI bar mapping are actually reachable.
Signed-off-by Gerd Hoffmann <[email protected]>--- target/i386/cpu.h | 1 + target/i386/cpu.c | 1 + target/i386/kvm/kvm.c | 17 +++++++++++++++++ 3 files changed, 19 insertions(+)target/i386/cpu.h +1
-0@@ -2026,6 +2026,7 @@ struct ArchCPU {/* Number of physical address bits supported */ uint32_t phys_bits; +uint32_t guest_phys_bits;/* in order to simplify APIC support, we leave this pointer to the user */ target/i386/cpu.c +1
-0@@ -6570,6 +6570,7 @@ void cpu_x86_cpuid(CPUX86State *env, uint32_t index, uint32_t count,if (env->features[FEAT_8000_0001_EDX ] &CPUID_EXT2_LM ) {/* 64 bit processor */ *eax |= (cpu_x86_virtual_addr_width (env) <<8 ); + *eax |= (cpu->guest_phys_bits <<16 ); } *ebx = env->features[FEAT_8000_0008_EBX ];if (cs->nr_cores * cs->nr_threads >1 ) {target/i386/kvm/kvm.c +10
-0@@ -238,6 +238,15 @@ static int kvm_get_tsc(CPUState *cs)return 0 ; } +/* return cpuid fn 8000_0008 eax[23:16] aka GuestPhysBits */ +static int kvm_get_guest_phys_bits (KVMState *s) +{ +uint32_t eax; + + eax =kvm_arch_get_supported_cpuid (s,0x80000008 ,0 ,R_EAX ); +return (eax >>16 ) &0xff ; +} +static inline void do_kvm_synchronize_tsc (CPUState *cpu,run_on_cpu_data arg) {kvm_get_tsc (cpu);@@ -1730,6 +1739,7 @@ int kvm_arch_init_vcpu(CPUState *cs)X86CPU *cpu =X86_CPU (cs);CPUX86State *env = &cpu->env;uint32_t limit, i, j, cpuid_i; +uint32_t guest_phys_bits;uint32_t unused;struct kvm_cpuid_entry2 *c;uint32_t signature[3 ];
target/i386/kvm/kvm.c int kvm_arch_init_vcpu(CPUState *cs) · line 1783
Quoted 9 lines
target/i386/kvm/kvm.c +6
-0@@ -1765,6 +1775,13 @@ int kvm_arch_init_vcpu(CPUState *cs) env->apic_bus_freq =KVM_APIC_BUS_FREQUENCY ; + guest_phys_bits =kvm_get_guest_phys_bits (cs->kvm_state); +if (guest_phys_bits && + (cpu->guest_phys_bits ==0 || + cpu->guest_phys_bits > guest_phys_bits)) { + cpu->guest_phys_bits = guest_phys_bits; + }
suggest to move this added code block to kvm_cpu_realizefn().
Main code in kvm_arch_init_vcpu() is to consume the data in cpu or env->features to configure KVM specific configuration via KVM ioctls.
The preparation work of initializing the required data usually not happen in kvm_arch_init_vcpu();
Quoted 3 lines
target/i386/kvm/kvm.c +0
-0/* * kvm_hyperv_expand_features() is called here for the second time in case * KVM_CAP_SYS_HYPERV_CPUID is not supported. While we can't possibly handle
Paolo Bonzini
maintainerQuoted 22 lines · Gerd Hoffmann
Query kvm for supported guest physical address bits, in cpuid function 80000008, eax[23:16]. Usually this is identical to host physical address bits. With NPT or EPT being used this might be restricted to 48 (max 4-level paging address space size) even if the host cpu supports more physical address bits.
When set pass this to the guest, using cpuid too. Guest firmware can use this to figure how big the usable guest physical address space is, so PCI bar mapping are actually reachable.
Signed-off-by Gerd Hoffmann <[email protected]>--- target/i386/cpu.h | 1 + target/i386/cpu.c | 1 + target/i386/kvm/kvm.c | 17 +++++++++++++++++ 3 files changed, 19 insertions(+)target/i386/cpu.h +0
-0
target/i386/cpu.h
Quoted 6 lines
target/i386/cpu.h +6
-0+ guest_phys_bits =kvm_get_guest_phys_bits (cs->kvm_state); +if (guest_phys_bits && + (cpu->guest_phys_bits ==0 || + cpu->guest_phys_bits > guest_phys_bits)) { + cpu->guest_phys_bits = guest_phys_bits; + }
Like Xiaoyao mentioned, the right place for this is kvm_cpu_realizefn, after host_cpu_realizefn returns. It should also be conditional on cpu->host_phys_bits. It also makes sense to:
- make kvm_get_guest_phys_bits() return bits 7:0 if bits 23:16 are zero
- here, set cpu->guest_phys_bits only if it is not equal to cpu->phys_bits (this undoes the previous suggestion, but I think it's cleaner)
- add a property in x86_cpu_properties[] to allow configuration with TCG.
Paolo
Quoted 7 lines
+
/*
*kvm_hyperv_expand_features()is called here for the second time in case
*KVM_CAP_SYS_HYPERV_CPUIDis not supported. While we can't possibly handle
--
2.44.0
Hi,
Quoted 6 lines
Quoted 6 lines
target/i386/cpu.h +2
-0+ guest_phys_bits =kvm_get_guest_phys_bits (cs->kvm_state); + if (guest_phys_bits &&
Quoted 8 lines
target/i386/cpu.h
Quoted 4 lines
target/i386/cpu.h +4
-0+ (cpu->guest_phys_bits == 0 || + cpu->guest_phys_bits > guest_phys_bits)) { + cpu->guest_phys_bits = guest_phys_bits; + }Like Xiaoyao mentioned, the right place for this is
kvm_cpu_realizefn, afterhost_cpu_realizefnreturns. It should also be conditional oncpu->host_phys_bits.
Ok.
Quoted 7 lines
It also makes sense to:
- make
kvm_get_guest_phys_bits()return bits 7:0 if bits 23:16 are zero- here, set
cpu->guest_phys_bitsonly if it is not equal tocpu->phys_bits(this undoes the previous suggestion, but I think it's cleaner)
Not sure about that.
I think it would be good to have a backward compatibility story. Currently neither the kernel nor qemu set guest_phys_bits. So if the firmware finds guest_phys_bits == 0 it does not know whenever ...
(a) kernel or qemu being too old, or
(b) no restrictions apply, it is safe to go with phys_bits.
One easy option would be to always let qemu pass through guest_phys_bits from the kernel, even in case it is equal to phys_bits.
Quoted 1 line
- add a property in x86_cpu_properties[] to allow configuration with TCG.
Was thinking about configuration too. Not sure it is a good idea to add yet another phys-bits config option to the mix of options we already have ...
In case host_phys_bits=true qemu could simply usemin(kernel guest-phys-bits,host-phys-bits-limit)
For the host_phys_bits=false case it would probably be best to just not set guest_phys_bits.
take care,
Gerd
Paolo Bonzini
maintainerQuoted 34 lines · Gerd Hoffmann
Hi,
Quoted 6 lines
Quoted 6 lines
target/i386/cpu.h +2
-0+ guest_phys_bits =kvm_get_guest_phys_bits (cs->kvm_state); + if (guest_phys_bits &&Quoted 8 lines
target/i386/cpu.h
Quoted 4 lines
target/i386/cpu.h +4
-0+ (cpu->guest_phys_bits == 0 || + cpu->guest_phys_bits > guest_phys_bits)) { + cpu->guest_phys_bits = guest_phys_bits; + }Like Xiaoyao mentioned, the right place for this is
kvm_cpu_realizefn, afterhost_cpu_realizefnreturns. It should also be conditional oncpu->host_phys_bits.Ok.
Quoted 7 lines
It also makes sense to:
- make
kvm_get_guest_phys_bits()return bits 7:0 if bits 23:16 are zero- here, set
cpu->guest_phys_bitsonly if it is not equal tocpu->phys_bits(this undoes the previous suggestion, but I think it's cleaner)Not sure about that.
I think it would be good to have a backward compatibility story. Currently neither the kernel nor qemu set
guest_phys_bits. So if the firmware findsguest_phys_bits== 0 it does not know whenever ...
Quoted 5 lines
(a) kernel or qemu being too old, or
(b) no restrictions apply, it is safe to go withphys_bits.One easy option would be to always let qemu pass through
guest_phys_bitsfrom the kernel, even in case it is equal tophys_bits.
Ah, I see - you would like to be able to use all 52 bits (instead of going for a safer 46 or 48) and therefore you need to have nonzero guest_phys_bits even if it's equal to phys_bits. While on an old kernel, you would pass forward 0.
Quoted 5 lines
Quoted 1 line
- add a property in x86_cpu_properties[] to allow configuration with TCG.
Was thinking about configuration too. Not sure it is a good idea to add yet another phys-bits config option to the mix of options we already have ...
I think it's nice that you can use TCG to test various cases, which requires a new property.
Quoted 2 lines
In case host_phys_bits=true qemu could simply use
min(kernel guest-phys-bits,host-phys-bits-limit)
Yes, that works.
Paolo
Quoted 2 lines
For the host_phys_bits=false case it would probably be best to just not set
guest_phys_bits.
Quoted 9 lines · Gerd Hoffmann
Query kvm for supported guest physical address bits, in cpuid function 80000008, eax[23:16]. Usually this is identical to host physical address bits. With NPT or EPT being used this might be restricted to 48 (max 4-level paging address space size) even if the host cpu supports more physical address bits.
When set pass this to the guest, using cpuid too. Guest firmware can use this to figure how big the usable guest physical address space is, so PCI bar mapping are actually reachable.
If this patch is applied, do you have plans to implement it in OVMF/Seabios?
Thanks,
Tao
Quoted 78 lines
Signed-off-by Gerd Hoffmann <[email protected]>--- target/i386/cpu.h | 1 + target/i386/cpu.c | 1 + target/i386/kvm/kvm.c | 17 +++++++++++++++++ 3 files changed, 19 insertions(+)target/i386/cpu.h +1
-0@@ -2026,6 +2026,7 @@ struct ArchCPU {/* Number of physical address bits supported */ uint32_t phys_bits; +uint32_t guest_phys_bits;/* in order to simplify APIC support, we leave this pointer to the user */ target/i386/cpu.c +1
-0@@ -6570,6 +6570,7 @@ void cpu_x86_cpuid(CPUX86State *env, uint32_t index, uint32_t count,if (env->features[FEAT_8000_0001_EDX ] &CPUID_EXT2_LM ) {/* 64 bit processor */ *eax |= (cpu_x86_virtual_addr_width (env) <<8 ); + *eax |= (cpu->guest_phys_bits <<16 ); } *ebx = env->features[FEAT_8000_0008_EBX ];if (cs->nr_cores * cs->nr_threads >1 ) {target/i386/kvm/kvm.c +17
-0@@ -238,6 +238,15 @@ static int kvm_get_tsc(CPUState *cs)return 0 ; } +/* return cpuid fn 8000_0008 eax[23:16] aka GuestPhysBits */ +static int kvm_get_guest_phys_bits (KVMState *s) +{ +uint32_t eax; + + eax =kvm_arch_get_supported_cpuid (s,0x80000008 ,0 ,R_EAX ); +return (eax >>16 ) &0xff ; +} +static inline void do_kvm_synchronize_tsc (CPUState *cpu,run_on_cpu_data arg) {kvm_get_tsc (cpu);@@ -1730,6 +1739,7 @@ int kvm_arch_init_vcpu(CPUState *cs)X86CPU *cpu =X86_CPU (cs);CPUX86State *env = &cpu->env;uint32_t limit, i, j, cpuid_i; +uint32_t guest_phys_bits;uint32_t unused;struct kvm_cpuid_entry2 *c;uint32_t signature[3 ];@@ -1765,6 +1775,13 @@ int kvm_arch_init_vcpu(CPUState *cs) env->apic_bus_freq =KVM_APIC_BUS_FREQUENCY ; + guest_phys_bits =kvm_get_guest_phys_bits (cs->kvm_state); +if (guest_phys_bits && + (cpu->guest_phys_bits ==0 || + cpu->guest_phys_bits > guest_phys_bits)) { + cpu->guest_phys_bits = guest_phys_bits; + } +/* * kvm_hyperv_expand_features() is called here for the second time in case * KVM_CAP_SYS_HYPERV_CPUID is not supported. While we can't possibly handle Signature · 1 line
2.44.0
Quoted 7 lines · Tao Su
Quoted 6 lines · Gerd Hoffmann
Query kvm for supported guest physical address bits, in cpuid function 80000008, eax[23:16]. Usually this is identical to host physical address bits. With NPT or EPT being used this might be restricted to 48 (max 4-level paging address space size) even if the host cpu supports more physical address bits.
Quoted 6 lines
Quoted 3 lines
When set pass this to the guest, using cpuid too. Guest firmware can use this to figure how big the usable guest physical address space is, so PCI bar mapping are actually reachable.
If this patch is applied, do you have plans to implement it in OVMF/Seabios?
Yes. ovmf test patches:
https://github.com/kraxel/edk2/commits/devel/guest-phys-bits/
take care,
Gerd