Hi,
On Fri, Feb 18, 2022 at 02:03:11PM +0000, Sebastian Ene wrote:
> On Thu, Feb 17, 2022 at 03:43:15PM +0000, Alexandru Elisei wrote:
> > Hi,
> >
> > Some general comments while I familiarize myself with the stolen time spec.
> >
> > On Thu, Feb 17, 2022 at 03:08:53PM +0000, Sebastian Ene wrote:
> > > This patch adds support for stolen time by sharing a memory region
> > > with the guest which will be used by the hypervisor to store the stolen
> > > time information. The exact format of the structure stored by the
> > > hypervisor is described in the ARM DEN0057A document.
> > >
> > > Signed-off-by: Sebastian Ene <sebastianene@xxxxxxxxxx>
> > > ---
> >
> > It is customary to describe the changes you have made from the previous version,
> > to make it easier for the people who want to review your code.
> >
>
> Hello,
>
> Thanks for the feedback. I will include the changelog in the next
> patches.
>
> > > Makefile | 1 +
> > > arm/aarch64/arm-cpu.c | 1 +
> > > arm/aarch64/include/kvm/kvm-cpu-arch.h | 1 +
> > > arm/aarch64/pvtime.c | 84 ++++++++++++++++++++++++++
> > > arm/include/arm-common/kvm-arch.h | 4 ++
> > > arm/kvm-cpu.c | 14 ++---
> > > 6 files changed, 98 insertions(+), 7 deletions(-)
> > > create mode 100644 arm/aarch64/pvtime.c
> > >
> > > diff --git a/Makefile b/Makefile
> > > index f251147..e9121dc 100644
> > > --- a/Makefile
> > > +++ b/Makefile
> > > @@ -182,6 +182,7 @@ ifeq ($(ARCH), arm64)
> > > OBJS += arm/aarch64/arm-cpu.o
> > > OBJS += arm/aarch64/kvm-cpu.o
> > > OBJS += arm/aarch64/kvm.o
> > > + OBJS += arm/aarch64/pvtime.o
> > > ARCH_INCLUDE := $(HDRS_ARM_COMMON)
> > > ARCH_INCLUDE += -Iarm/aarch64/include
> > >
> > > diff --git a/arm/aarch64/arm-cpu.c b/arm/aarch64/arm-cpu.c
> > > index d7572b7..326fb20 100644
> > > --- a/arm/aarch64/arm-cpu.c
> > > +++ b/arm/aarch64/arm-cpu.c
> > > @@ -22,6 +22,7 @@ static void generate_fdt_nodes(void *fdt, struct kvm *kvm)
> > > static int arm_cpu__vcpu_init(struct kvm_cpu *vcpu)
> > > {
> > > vcpu->generate_fdt_nodes = generate_fdt_nodes;
> > > + kvm_cpu__setup_pvtime(vcpu);
> > > return 0;
> > > }
> > >
> > > diff --git a/arm/aarch64/include/kvm/kvm-cpu-arch.h b/arm/aarch64/include/kvm/kvm-cpu-arch.h
> > > index 8dfb82e..b57d6e6 100644
> > > --- a/arm/aarch64/include/kvm/kvm-cpu-arch.h
> > > +++ b/arm/aarch64/include/kvm/kvm-cpu-arch.h
> > > @@ -19,5 +19,6 @@
> > >
> > > void kvm_cpu__select_features(struct kvm *kvm, struct kvm_vcpu_init *init);
> > > int kvm_cpu__configure_features(struct kvm_cpu *vcpu);
> > > +void kvm_cpu__setup_pvtime(struct kvm_cpu *vcpu);
> > >
> > > #endif /* KVM__KVM_CPU_ARCH_H */
> > > diff --git a/arm/aarch64/pvtime.c b/arm/aarch64/pvtime.c
> > > new file mode 100644
> > > index 0000000..c09fd89
> > > --- /dev/null
> > > +++ b/arm/aarch64/pvtime.c
> > > @@ -0,0 +1,84 @@
> > > +#include "kvm/kvm.h"
> > > +#include "kvm/kvm-cpu.h"
> > > +#include "kvm/util.h"
> > > +
> > > +#include <linux/byteorder.h>
> > > +#include <linux/types.h>
> > > +
> > > +struct pvtime_data_priv {
> > > + bool is_supported;
> > > + char *usr_mem;
> > > +};
> > > +
> > > +static struct pvtime_data_priv pvtime_data = {
> > > + .is_supported = true,
> > > + .usr_mem = NULL
> > > +};
> > > +
> > > +static int pvtime__alloc_region(struct kvm *kvm)
> > > +{
> > > + char *mem;
> > > + int ret = 0;
> > > +
> > > + mem = mmap(NULL, AARCH64_PVTIME_IPA_MAX_SIZE, PROT_RW,
> > > + MAP_ANON_NORESERVE, -1, 0);
> > > + if (mem == MAP_FAILED)
> > > + return -ENOMEM;
> > > +
> > > + ret = kvm__register_dev_mem(kvm, AARCH64_PVTIME_IPA_START,
> > > + AARCH64_PVTIME_IPA_MAX_SIZE, mem);
> > > + if (ret) {
> > > + munmap(mem, AARCH64_PVTIME_IPA_MAX_SIZE);
> > > + return ret;
> > > + }
> > > +
> > > + pvtime_data.usr_mem = mem;
> > > + return ret;
> > > +}
> > > +
> > > +static int pvtime__teardown_region(struct kvm *kvm)
> > > +{
> > > + if (pvtime_data.usr_mem == NULL)
> > > + return 0;
> > > +
> > > + kvm__destroy_mem(kvm, AARCH64_PVTIME_IPA_START,
> > > + AARCH64_PVTIME_IPA_MAX_SIZE, pvtime_data.usr_mem);
> > > + munmap(pvtime_data.usr_mem, AARCH64_PVTIME_IPA_MAX_SIZE);
> > > + pvtime_data.usr_mem = NULL;
> > > + return 0;
> > > +}
> > > +
> > > +void kvm_cpu__setup_pvtime(struct kvm_cpu *vcpu)
> > > +{
> > > + int ret;
> > > + u64 pvtime_guest_addr = AARCH64_PVTIME_IPA_START + vcpu->cpu_id *
> > > + AARCH64_PVTIME_SIZE;
> > > + struct kvm_device_attr pvtime_attr = (struct kvm_device_attr) {
> > > + .group = KVM_ARM_VCPU_PVTIME_CTRL,
> > > + .addr = KVM_ARM_VCPU_PVTIME_IPA
> > > + };
> > > +
> > > + if (!pvtime_data.is_supported)
> > > + return;
> > > +
> > > + ret = ioctl(vcpu->vcpu_fd, KVM_HAS_DEVICE_ATTR, &pvtime_attr);
> > > + if (ret)
> > > + goto out_err;
> > > +
> > > + if (!pvtime_data.usr_mem) {
> > > + ret = pvtime__alloc_region(vcpu->kvm);
> > > + if (ret)
> > > + goto out_err;
> > > + }
> > > +
> > > + pvtime_attr.addr = (u64)&pvtime_guest_addr;
> > > + ret = ioctl(vcpu->vcpu_fd, KVM_SET_DEVICE_ATTR, &pvtime_attr);
> > > + if (!ret)
> > > + return;
> > > +
> > > + pvtime__teardown_region(vcpu->kvm);
> > > +out_err:
> > > + pvtime_data.is_supported = false;
> > > +}
> > > +
> > > +dev_exit(pvtime__teardown_region);
> > > diff --git a/arm/include/arm-common/kvm-arch.h b/arm/include/arm-common/kvm-arch.h
> > > index c645ac0..865bd68 100644
> > > --- a/arm/include/arm-common/kvm-arch.h
> > > +++ b/arm/include/arm-common/kvm-arch.h
> > > @@ -54,6 +54,10 @@
> > > #define ARM_PCI_MMIO_SIZE (ARM_MEMORY_AREA - \
> > > (ARM_AXI_AREA + ARM_PCI_CFG_SIZE))
> > >
> > > +#define AARCH64_PVTIME_IPA_MAX_SIZE SZ_64K
> > > +#define AARCH64_PVTIME_IPA_START (ARM_MEMORY_AREA - \
> > > + AARCH64_PVTIME_IPA_MAX_SIZE)
> >
> > This overlaps with the ARM_PCI_MMIO_AREA. If you want to change the memory
> > layout for a guest, there's a handy ASCII map at the top of this file. That
> > should also be updated to reflect the modified layout.
>
> Right, when pvtime is supported it ovelaps that region. While this
> feature is supported only for arm64, does it make sense to update the ASCII
> map in this header which is arm-common ? Also, probably it makes sense to move
> all of the definitions to pvtime.c ? What do you think ?
kvmtool uses the same memory layout for arm and arm64. If you want to have
different memory layouts between arm and arm64, then the memory layout from
this file should be moved to arm/aarch{32,64}/include/kvm/kvm-arch.h and
modified there.
I'm not 100% sure how to proceed here. pvtime requires only 64KiB of memory
out of a 1GiB - 256MiB region, so making ARM_PCI_MMIO_SIZE 64KiB smaller
for both architectures is not a big deal as far as I can tell.
Having both arm and arm64 use the same memory layout means that fixes
benefit both architectures (there had been several memory layout fixes in
the past, where regions were overlapping).
On the other hand, KVM for arm does not exist anymore, and hasn't for quite
some time, and changing the memory layout for an arm64 feature means that
we risk breaking arm support, which is becoming harder and harder to test.
And this will become increasingly likely as new arm64 features that change
the memory layout are added.
To be honest, I think the best person to answer this is Will, as the
maintainer of kvmtool. I have been going back and forth about this, I think
I slightly prefer making the change here, to the common layout.
>
> >
> > Why do you want to put it below RAM? Is there a requirement to have it there or
> > was the location chosen for other reasons?
> >
>
> I don't have a strong argument for this placement but I am open to
> suggestions if you would like to move it somewhere else. For example, we
> are placing pvtime in the same region in crosvm.
It looks ok to me to have it there, clumped with the other reserved
regions.
>
> > > +#define AARCH64_PVTIME_SIZE (64)
> >
> > This looks like something that should go in pvtime.c, as it's not relevant to
> > the memory layout.
> >
>
> I agree, I will move this. Thanks.
>
> > >
> > > #define ARM_LOMAP_MAX_MEMORY ((1ULL << 32) - ARM_MEMORY_AREA)
> > > #define ARM_HIMAP_MAX_MEMORY ((1ULL << 40) - ARM_MEMORY_AREA)
> > > diff --git a/arm/kvm-cpu.c b/arm/kvm-cpu.c
> > > index 6a2408c..84ac1e9 100644
> > > --- a/arm/kvm-cpu.c
> > > +++ b/arm/kvm-cpu.c
> > > @@ -116,6 +116,13 @@ struct kvm_cpu *kvm_cpu__arch_init(struct kvm *kvm, unsigned long cpu_id)
> > > die("Unable to find matching target");
> > > }
> > >
> > > + /* Populate the vcpu structure. */
> > > + vcpu->kvm = kvm;
> > > + vcpu->cpu_id = cpu_id;
> > > + vcpu->cpu_type = vcpu_init.target;
> > > + vcpu->cpu_compatible = target->compatible;
> > > + vcpu->is_running = true;
> > > +
> > > if (err || target->init(vcpu))
> > > die("Unable to initialise vcpu");
> > >
> > > @@ -125,13 +132,6 @@ struct kvm_cpu *kvm_cpu__arch_init(struct kvm *kvm, unsigned long cpu_id)
> > > vcpu->ring = (void *)vcpu->kvm_run +
> > > (coalesced_offset * PAGE_SIZE);
> > >
> > > - /* Populate the vcpu structure. */
> > > - vcpu->kvm = kvm;
> > > - vcpu->cpu_id = cpu_id;
> > > - vcpu->cpu_type = vcpu_init.target;
> > > - vcpu->cpu_compatible = target->compatible;
> > > - vcpu->is_running = true;
> > > -
> >
> > Why this change?
> >
>
> I will add a comment to describe why I moved this. During pvtime setup,
> in vcpu initialization I need to access the kvm structure from the vcpu.
Yeah, figured that out after I sent the email. Would you mind turning this
change into a separate patch, something like "arm: Populate the VCPU struct
before target->init()"? Please feel free to word the patch however you see
fit.
Thanks,
Alex
>
> > Thanks,
> > Alex
> >
>
> Thanks for the review,
> Sebastian
>
> > > if (kvm_cpu__configure_features(vcpu))
> > > die("Unable to configure requested vcpu features");
> > >
> > > --
> > > 2.35.1.265.g69c8d7142f-goog
> > >
_______________________________________________
kvmarm mailing list
kvmarm@xxxxxxxxxxxxxxxxxxxxx
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
