On Fri, Mar 11, 2022 at 11:18:55AM +0000, Alexandru Elisei wrote:
> Hi,
>
Hi,
> On Wed, Mar 09, 2022 at 01:34:23PM +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. Reserve a 64kb MMIO memory region after the RTC peripheral
> > to be used by pvtime. The exact format of the structure stored by the
> > hypervisor is described in the ARM DEN0057A document.
> >
> > Signed-off-by: Sebastian Ene <sebastianene@xxxxxxxxxx>
> > ---
> > Makefile | 1 +
> > arm/aarch32/include/kvm/kvm-cpu-arch.h | 5 ++
> > arm/aarch64/arm-cpu.c | 2 +-
> > arm/aarch64/include/kvm/kvm-cpu-arch.h | 2 +
> > arm/aarch64/pvtime.c | 96 ++++++++++++++++++++++++++
> > arm/include/arm-common/kvm-arch.h | 6 +-
> > arm/kvm-cpu.c | 1 +
> > include/kvm/kvm-config.h | 1 +
> > 8 files changed, 112 insertions(+), 2 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/aarch32/include/kvm/kvm-cpu-arch.h b/arm/aarch32/include/kvm/kvm-cpu-arch.h
> > index 780e0e2..6fe0206 100644
> > --- a/arm/aarch32/include/kvm/kvm-cpu-arch.h
> > +++ b/arm/aarch32/include/kvm/kvm-cpu-arch.h
> > @@ -20,4 +20,9 @@ static inline int kvm_cpu__configure_features(struct kvm_cpu *vcpu)
> > return 0;
> > }
> >
> > +static inline int kvm_cpu__teardown_pvtime(struct kvm *kvm)
> > +{
> > + return 0;
> > +}
> > +
> > #endif /* KVM__KVM_CPU_ARCH_H */
> > diff --git a/arm/aarch64/arm-cpu.c b/arm/aarch64/arm-cpu.c
> > index d7572b7..7e4a3c1 100644
> > --- a/arm/aarch64/arm-cpu.c
> > +++ b/arm/aarch64/arm-cpu.c
> > @@ -22,7 +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;
> > - return 0;
> > + return kvm_cpu__setup_pvtime(vcpu);
> > }
> >
> > static struct kvm_arm_target target_generic_v8 = {
> > diff --git a/arm/aarch64/include/kvm/kvm-cpu-arch.h b/arm/aarch64/include/kvm/kvm-cpu-arch.h
> > index 8dfb82e..35996dc 100644
> > --- a/arm/aarch64/include/kvm/kvm-cpu-arch.h
> > +++ b/arm/aarch64/include/kvm/kvm-cpu-arch.h
> > @@ -19,5 +19,7 @@
> >
> > void kvm_cpu__select_features(struct kvm *kvm, struct kvm_vcpu_init *init);
> > int kvm_cpu__configure_features(struct kvm_cpu *vcpu);
> > +int kvm_cpu__setup_pvtime(struct kvm_cpu *vcpu);
> > +int kvm_cpu__teardown_pvtime(struct kvm *kvm);
> >
> > #endif /* KVM__KVM_CPU_ARCH_H */
> > diff --git a/arm/aarch64/pvtime.c b/arm/aarch64/pvtime.c
> > new file mode 100644
> > index 0000000..720e9de
> > --- /dev/null
> > +++ b/arm/aarch64/pvtime.c
> > @@ -0,0 +1,96 @@
> > +#include "kvm/kvm.h"
> > +#include "kvm/kvm-cpu.h"
> > +#include "kvm/util.h"
> > +
> > +#include <linux/byteorder.h>
> > +#include <linux/types.h>
> > +
> > +#define ARM_PVTIME_STRUCT_SIZE (64)
> > +
> > +static void *usr_mem;
> > +
> > +static int pvtime__alloc_region(struct kvm *kvm)
> > +{
> > + char *mem;
> > + int ret = 0;
> > +
> > + mem = mmap(NULL, ARM_PVTIME_BASE, PROT_RW,
> > + MAP_ANON_NORESERVE, -1, 0);
> > + if (mem == MAP_FAILED)
> > + return -errno;
> > +
> > + ret = kvm__register_ram(kvm, ARM_PVTIME_BASE,
> > + ARM_PVTIME_BASE, mem);
> > + if (ret) {
> > + munmap(mem, ARM_PVTIME_BASE);
> > + return ret;
> > + }
> > +
> > + usr_mem = mem;
> > + return ret;
> > +}
> > +
> > +static int pvtime__teardown_region(struct kvm *kvm)
> > +{
> > + if (usr_mem == NULL)
> > + return 0;
> > +
> > + kvm__destroy_mem(kvm, ARM_PVTIME_BASE,
> > + ARM_PVTIME_BASE, usr_mem);
> > + munmap(usr_mem, ARM_PVTIME_BASE);
> > + usr_mem = NULL;
> > + return 0;
> > +}
> > +
> > +int kvm_cpu__setup_pvtime(struct kvm_cpu *vcpu)
> > +{
> > + int ret;
> > + bool has_stolen_time;
> > + u64 pvtime_guest_addr = ARM_PVTIME_BASE + vcpu->cpu_id *
> > + ARM_PVTIME_STRUCT_SIZE;
> > + struct kvm_config *kvm_cfg = NULL;
> > + struct kvm_device_attr pvtime_attr = (struct kvm_device_attr) {
> > + .group = KVM_ARM_VCPU_PVTIME_CTRL,
> > + .attr = KVM_ARM_VCPU_PVTIME_IPA
> > + };
> > +
> > + kvm_cfg = &vcpu->kvm->cfg;
> > + if (kvm_cfg->no_pvtime)
> > + return 0;
> > +
> > + has_stolen_time = kvm__supports_extension(vcpu->kvm,
> > + KVM_CAP_STEAL_TIME);
> > + if (!has_stolen_time) {
> > + kvm_cfg->no_pvtime = true;
> > + return 0;
> > + }
> > +
> > + ret = ioctl(vcpu->vcpu_fd, KVM_HAS_DEVICE_ATTR, &pvtime_attr);
> > + if (ret) {
> > + perror("KVM_HAS_DEVICE_ATTR failed\n");
> > + goto out_err;
>
> Nitpick: here we return -1 (ioctl() return value when it fails) instead of
> returning -errno. You can set ret = -errno before the perror() call (in
> case it also fails for some reason, thus changing the value of errno).
>
> Not a big deal, but it changes the semantics of the return value for the
> function: below, for pvtime__alloc_region() we return -errno, but here we
> return -1. Someone who is debugging an error might print this return value
> and think that the ioctl() failed with error code EPERM (EPERM is 1), which
> is not a valid error code for ioctl() according to man 2 ioctl. It will
> also look rather strange for the perror to print the error message
> associated with a different error code than the error code that is returned
> from this function.
>
I will update the return code, thanks for the suggestion.
> > + }
> > +
> > + if (!usr_mem) {
> > + ret = pvtime__alloc_region(vcpu->kvm);
> > + if (ret) {
> > + perror("Failed allocating pvtime region\n");
> > + goto out_err;
> > + }
> > + }
> > +
> > + pvtime_attr.addr = (u64)&pvtime_guest_addr;
> > + ret = ioctl(vcpu->vcpu_fd, KVM_SET_DEVICE_ATTR, &pvtime_attr);
> > + if (!ret)
> > + return 0;
>
> Nitpick: same here, ret must be set to -errno before returning.
>
Right, I will set it to errno on 'ret != 0'.
> > +
> > + perror("KVM_SET_DEVICE_ATTR failed\n");
> > + pvtime__teardown_region(vcpu->kvm);
> > +out_err:
> > + return ret;
> > +}
> > +
> > +int kvm_cpu__teardown_pvtime(struct kvm *kvm)
> > +{
> > + return pvtime__teardown_region(kvm);
> > +}
> > diff --git a/arm/include/arm-common/kvm-arch.h b/arm/include/arm-common/kvm-arch.h
> > index c645ac0..43b1f77 100644
> > --- a/arm/include/arm-common/kvm-arch.h
> > +++ b/arm/include/arm-common/kvm-arch.h
> > @@ -15,7 +15,8 @@
> > * | PCI |////| plat | | | | |
> > * | I/O |////| MMIO: | Flash | virtio | GIC | PCI | DRAM
> > * | space |////| UART, | | MMIO | | (AXI) |
> > - * | |////| RTC | | | | |
> > + * | |////| RTC, | | | | |
> > + * | |////| PVTIME| | | | |
> > * +-------+----+-------+-------+--------+-----+---------+---......
> > */
> >
> > @@ -34,6 +35,9 @@
> > #define ARM_RTC_MMIO_BASE (ARM_UART_MMIO_BASE + ARM_UART_MMIO_SIZE)
> > #define ARM_RTC_MMIO_SIZE 0x10000
> >
> > +#define ARM_PVTIME_BASE (ARM_RTC_MMIO_BASE + ARM_RTC_MMIO_SIZE)
> > +#define ARM_PVTIME_SIZE SZ_64K
> > +
> > #define KVM_FLASH_MMIO_BASE (ARM_MMIO_AREA + 0x1000000)
> > #define KVM_FLASH_MAX_SIZE 0x1000000
> >
> > diff --git a/arm/kvm-cpu.c b/arm/kvm-cpu.c
> > index 84ac1e9..00660d6 100644
> > --- a/arm/kvm-cpu.c
> > +++ b/arm/kvm-cpu.c
> > @@ -144,6 +144,7 @@ void kvm_cpu__arch_nmi(struct kvm_cpu *cpu)
> >
> > void kvm_cpu__delete(struct kvm_cpu *vcpu)
> > {
> > + kvm_cpu__teardown_pvtime(vcpu->kvm);
> > free(vcpu);
> > }
> >
> > diff --git a/include/kvm/kvm-config.h b/include/kvm/kvm-config.h
> > index 6a5720c..48adf27 100644
> > --- a/include/kvm/kvm-config.h
> > +++ b/include/kvm/kvm-config.h
> > @@ -62,6 +62,7 @@ struct kvm_config {
> > bool no_dhcp;
> > bool ioport_debug;
> > bool mmio_debug;
> > + bool no_pvtime;
> > };
> >
> > #endif
> > --
>
> Tested the series, a guest is able to detect pvtime:
>
> [ 0.008661] arm-pv: using stolen time PV
>
Thanks for the review and for testing this,
> Tested-by: Alexandru Elisei <alexandru.elisei@xxxxxxx>
>
> With the two nitpicks above fixed, the patch looks correct to me:
>
> Reviewed-by: Alexandru Elisei <alexandru.elisei@xxxxxxx>
>
> Thanks,
> Alex
Cheers,
Sebastian
