On Thu, Dec 14, 2023 at 3:45 PM Andrew Jones <ajones@xxxxxxxxxxxxxxxx> wrote:
>
> When the SBI STA extension exists we can use it to implement
> paravirt steal-time support. Fill in the empty pv-time functions
> with an SBI STA implementation and add the Kconfig knobs allowing
> it to be enabled.
>
> Signed-off-by: Andrew Jones <ajones@xxxxxxxxxxxxxxxx>
> ---
> arch/riscv/Kconfig | 19 ++++++++++
> arch/riscv/kernel/paravirt.c | 67 ++++++++++++++++++++++++++++++++++--
> 2 files changed, 83 insertions(+), 3 deletions(-)
>
> diff --git a/arch/riscv/Kconfig b/arch/riscv/Kconfig
> index 95a2a06acc6a..b99fd8129edf 100644
> --- a/arch/riscv/Kconfig
> +++ b/arch/riscv/Kconfig
> @@ -724,6 +724,25 @@ config COMPAT
>
> If you want to execute 32-bit userspace applications, say Y.
>
> +config PARAVIRT
> + bool "Enable paravirtualization code"
> + depends on RISCV_SBI
> + help
> + This changes the kernel so it can modify itself when it is run
> + under a hypervisor, potentially improving performance significantly
> + over full virtualization.
> +
> +config PARAVIRT_TIME_ACCOUNTING
> + bool "Paravirtual steal time accounting"
> + depends on PARAVIRT
> + help
> + Select this option to enable fine granularity task steal time
> + accounting. Time spent executing other tasks in parallel with
> + the current vCPU is discounted from the vCPU power. To account for
> + that, there can be a small performance impact.
> +
> + If in doubt, say N here.
> +
> config RELOCATABLE
> bool "Build a relocatable kernel"
> depends on MMU && 64BIT && !XIP_KERNEL
> diff --git a/arch/riscv/kernel/paravirt.c b/arch/riscv/kernel/paravirt.c
> index 141dbcc36fa2..b09dfd81bcd2 100644
> --- a/arch/riscv/kernel/paravirt.c
> +++ b/arch/riscv/kernel/paravirt.c
> @@ -6,12 +6,21 @@
> #define pr_fmt(fmt) "riscv-pv: " fmt
>
> #include <linux/cpuhotplug.h>
> +#include <linux/compiler.h>
> +#include <linux/errno.h>
> #include <linux/init.h>
> #include <linux/jump_label.h>
> +#include <linux/kconfig.h>
> +#include <linux/kernel.h>
> +#include <linux/percpu-defs.h>
> #include <linux/printk.h>
> #include <linux/static_call.h>
> #include <linux/types.h>
>
> +#include <asm/barrier.h>
> +#include <asm/page.h>
> +#include <asm/sbi.h>
> +
> struct static_key paravirt_steal_enabled;
> struct static_key paravirt_steal_rq_enabled;
>
> @@ -31,24 +40,76 @@ static int __init parse_no_stealacc(char *arg)
>
> early_param("no-steal-acc", parse_no_stealacc);
>
> +DEFINE_PER_CPU(struct sbi_sta_struct, steal_time) __aligned(64);
> +
> static bool __init has_pv_steal_clock(void)
> {
> + if (sbi_spec_version >= sbi_mk_version(2, 0) &&
> + sbi_probe_extension(SBI_EXT_STA) > 0) {
> + pr_info("SBI STA extension detected\n");
> + return true;
> + }
> +
> return false;
> }
>
> -static int pv_time_cpu_online(unsigned int cpu)
> +static int sbi_sta_steal_time_set_shmem(unsigned long lo, unsigned long hi,
> + unsigned long flags)
> {
> + struct sbiret ret;
> +
> + ret = sbi_ecall(SBI_EXT_STA, SBI_EXT_STA_STEAL_TIME_SET_SHMEM,
> + lo, hi, flags, 0, 0, 0);
> + if (ret.error) {
> + if (lo == SBI_STA_SHMEM_DISABLE && hi == SBI_STA_SHMEM_DISABLE)
> + pr_warn("Failed to disable steal-time shmem");
> + else
> + pr_warn("Failed to set steal-time shmem");
> + return sbi_err_map_linux_errno(ret.error);
> + }
> +
> return 0;
> }
>
> +static int pv_time_cpu_online(unsigned int cpu)
> +{
> + struct sbi_sta_struct *st = this_cpu_ptr(&steal_time);
> + phys_addr_t pa = __pa(st);
> + unsigned long lo = (unsigned long)pa;
> + unsigned long hi = IS_ENABLED(CONFIG_32BIT) ? upper_32_bits((u64)pa) : 0;
> +
> + return sbi_sta_steal_time_set_shmem(lo, hi, 0);
> +}
> +
> static int pv_time_cpu_down_prepare(unsigned int cpu)
> {
> - return 0;
> + return sbi_sta_steal_time_set_shmem(SBI_STA_SHMEM_DISABLE,
> + SBI_STA_SHMEM_DISABLE, 0);
> }
>
> static u64 pv_time_steal_clock(int cpu)
> {
> - return 0;
> + struct sbi_sta_struct *st = per_cpu_ptr(&steal_time, cpu);
> + u32 sequence;
> + u64 steal;
> +
> + if (IS_ENABLED(CONFIG_32BIT)) {
> + /*
> + * Check the sequence field before and after reading the steal
> + * field. Repeat the read if it is different or odd.
> + */
> + do {
> + sequence = READ_ONCE(st->sequence);
> + virt_rmb();
> + steal = READ_ONCE(st->steal);
> + virt_rmb();
> + } while ((le32_to_cpu(sequence) & 1) ||
> + sequence != READ_ONCE(st->sequence));
Actually, we should be doing this sequence for both RV64 and RV32
because for RV64 the steal time value is valid only when sequence is
an even number.
> + } else {
> + steal = READ_ONCE(st->steal);
> + }
> +
> + return le64_to_cpu(steal);
> }
>
> int __init pv_time_init(void)
> --
> 2.43.0
>
Regards,
Anup
