[PATCH] RISC-V: KVM: Add arch-specific tracepoints

Yuhang.Chen yhchen312 at gmail.com
Wed Jul 29 23:50:48 PDT 2026


Hi Anup,

Thanks for the review.

I've addressed your comments in v2.

On Tue, Jul 28, 2026 at 09:05:39PM +0530, Anup Patel wrote:
> Drop kvm_guest_fault, kvm_sbi_ecall, kvm_wait_riscv, and
> kvm_csr_access tracepoints. Instead add common
> kvm_vcpu_exit tracepoint for all vcpu exits just before the
> switch case in kvm_riscv_vcpu_exit() function.

Done. kvm_vcpu_exit now reports the full trap context
(sepc/scause/stval/htval/htinst) and is placed just before the switch
case, after the CAUSE_IRQ_FLAG early-return, so it fires only for
synchronous traps.

> Drop the kvm_irq_line and kvm_timer_update_irq tracepoints.
> Instead add common kvm_vcpu_irq tracepoint called from
> kvm_riscv_vcpu_set/unset_interrupt() function.

Done. kvm_vcpu_irq is called from kvm_riscv_vcpu_set_interrupt() and
kvm_riscv_vcpu_unset_interrupt(), covering all callers.

> The kvm_mmio_emulate tracepoint is fine.

Kept kvm_mmio_emulate unchanged.

v2 follows in this thread, with example trace output captured from QEMU
included in the commit message.

Thanks,
Yuhang



More information about the linux-riscv mailing list