Commit 873f10cf authored by Fangyu Yu's avatar Fangyu Yu Committed by Anup Patel
Browse files

RISC-V: KVM: Read HGEIP CSR on the correct cpu 



When executing kvm_riscv_vcpu_aia_has_interrupts, the vCPU may have
migrated and the IMSIC VS-file have not been updated yet, currently
the HGEIP CSR should be read from the imsic->vsfile_cpu ( the pCPU
before migration ) via on_each_cpu_mask, but this will trigger an
IPI call and repeated IPI within a period of time is expensive in
a many-core systems.

Just let the vCPU execute and update the correct IMSIC VS-file via
kvm_riscv_vcpu_aia_imsic_update may be a simple solution.

Fixes: 4cec89db ("RISC-V: KVM: Move HGEI[E|P] CSR access to IMSIC virtualization")
Signed-off-by: default avatarFangyu Yu <fangyu.yu@linux.alibaba.com>
Reviewed-by: default avatarGuo Ren <guoren@kernel.org>
Reviewed-by: default avatarAnup Patel <anup@brainfault.org>
Tested-by: default avatarAnup Patel <anup@brainfault.org>
Link: https://lore.kernel.org/r/20251016012659.82998-1-fangyu.yu@linux.alibaba.com


Signed-off-by: default avatarAnup Patel <anup@brainfault.org>
parent ea138a60

arch/riscv/kvm/aia_imsic.c

+14 −2
Original line number Diff line number Diff line
@@ -689,8 +689,20 @@ bool kvm_riscv_vcpu_aia_imsic_has_interrupt(struct kvm_vcpu *vcpu) 
	 */



	read_lock_irqsave(&imsic->vsfile_lock, flags);

	if (imsic->vsfile_cpu > -1)

	if (imsic->vsfile_cpu > -1) {

		/*

		 * This function is typically called from kvm_vcpu_block() via

		 * kvm_arch_vcpu_runnable() upon WFI trap. The kvm_vcpu_block()

		 * can be preempted and the blocking VCPU might resume on a

		 * different CPU. This means it is possible that current CPU

		 * does not match the imsic->vsfile_cpu hence this function

		 * must check imsic->vsfile_cpu before accessing HGEIP CSR.

		 */

		if (imsic->vsfile_cpu != vcpu->cpu)

			ret = true;

		else

			ret = !!(csr_read(CSR_HGEIP) & BIT(imsic->vsfile_hgei));

	}

	read_unlock_irqrestore(&imsic->vsfile_lock, flags);



	return ret;