On May 16, 2025 2:18:18 PM GMT+02:00, Anup Patel <anup@xxxxxxxxxxxxxx> wrote: >On Fri, May 16, 2025 at 4:41 AM Atish Patra <atishp@xxxxxxxxxxxx> wrote: >> >> Scounteren CSR controls the direct access the hpmcounters and cycle/ >> instret/time from the userspace. It's the supervisor's responsibility >> to set it up correctly for it's user space. They hypervisor doesn't >> need to decide the policy on behalf of the supervisor. >> >> Signed-off-by: Atish Patra <atishp@xxxxxxxxxxxx> >> --- >> Changes in v3: >> - Removed the redundant declaration >> - Link to v2: https://lore.kernel.org/r/20250515-fix_scounteren_vs-v2-1-1fd8dc0693e8@xxxxxxxxxxxx >> >> Changes in v2: >> - Remove the scounteren initialization instead of just setting the TM bit. >> - Link to v1: https://lore.kernel.org/r/20250513-fix_scounteren_vs-v1-1-c1f52af93c79@xxxxxxxxxxxx >> --- >> arch/riscv/kvm/vcpu.c | 4 ---- >> 1 file changed, 4 deletions(-) >> >> diff --git a/arch/riscv/kvm/vcpu.c b/arch/riscv/kvm/vcpu.c >> index 60d684c76c58..9bfaae9a11ea 100644 >> --- a/arch/riscv/kvm/vcpu.c >> +++ b/arch/riscv/kvm/vcpu.c >> @@ -111,7 +111,6 @@ int kvm_arch_vcpu_create(struct kvm_vcpu *vcpu) >> { >> int rc; >> struct kvm_cpu_context *cntx; >> - struct kvm_vcpu_csr *reset_csr = &vcpu->arch.guest_reset_csr; >> >> spin_lock_init(&vcpu->arch.mp_state_lock); >> >> @@ -146,9 +145,6 @@ int kvm_arch_vcpu_create(struct kvm_vcpu *vcpu) >> if (kvm_riscv_vcpu_alloc_vector_context(vcpu, cntx)) >> return -ENOMEM; >> >> - /* By default, make CY, TM, and IR counters accessible in VU mode */ >> - reset_csr->scounteren = 0x7; >> - >> /* Setup VCPU timer */ >> kvm_riscv_vcpu_timer_init(vcpu); >> >> >> --- >> base-commit: 01f95500a162fca88cefab9ed64ceded5afabc12 >> change-id: 20250513-fix_scounteren_vs-fdd86255c7b7 >> -- > >Overall, this looks good. > >Reviewed-by: Anup Patel <anup@xxxxxxxxxxxxxx> > >Currently, the scounteren.TM bit is only set by the Linux SBI PMU >driver but KVM RISC-V only provides SBI PMU for guest when >Sscofpmf is available in host so we need the below hunk to >completely get rid-off scounteren initialization in KVM RISC-V. > >diff --git a/arch/riscv/kernel/head.S b/arch/riscv/kernel/head.S >index 356d5397b2a2..bdf3352acf4c 100644 >--- a/arch/riscv/kernel/head.S >+++ b/arch/riscv/kernel/head.S >@@ -131,6 +131,12 @@ secondary_start_sbi: > csrw CSR_IE, zero > csrw CSR_IP, zero > >+#ifndef CONFIG_RISCV_M_MODE >+ /* Enable time CSR */ >+ li t0, 0x2 >+ csrw CSR_SCOUNTEREN, t0 >+#endif >+ > /* Load the global pointer */ > load_global_pointer > >@@ -226,6 +232,10 @@ SYM_CODE_START(_start_kernel) > * to hand it to us. > */ > csrr a0, CSR_MHARTID >+#else >+ /* Enable time CSR */ >+ li t0, 0x2 >+ csrw CSR_SCOUNTEREN, t0 > #endif /* CONFIG_RISCV_M_MODE */ > > /* Load the global pointer */ > >I have queued this patch for Linux-6.16 with the above hunk squashed. > >Thanks, >Anup > >-- >kvm-riscv mailing list >kvm-riscv@xxxxxxxxxxxxxxxxxxx >http://lists.infradead.org/mailman/listinfo/kvm-riscv Yup, I had planned to post something like that too, but didn't get to it. Thanks, drew
