[PATCH 10/13] powerpc/64s: idle simplify KVM idle on POWER9
Gautham R Shenoy
ego at linux.vnet.ibm.com
Tue Aug 8 20:36:43 AEST 2017
Hi Nicholas,
On Sun, Aug 06, 2017 at 03:02:38AM +1000, Nicholas Piggin wrote:
> POWER9 CPUs have independent MMU contexts per thread so KVM
> does not have to bring sibling threads into real-mode when
> switching MMU mode to guest. This can simplify POWER9 sleep/wake
> paths and avoids hwsyncs.
>
> Signed-off-by: Nicholas Piggin <npiggin at gmail.com>
> ---
> arch/powerpc/include/asm/kvm_book3s_asm.h | 4 ++++
> arch/powerpc/kernel/idle_book3s.S | 8 ++-----
> arch/powerpc/kvm/book3s_hv.c | 37 ++++++++++++++++++++++++++-----
> arch/powerpc/kvm/book3s_hv_rmhandlers.S | 8 +++++++
> 4 files changed, 46 insertions(+), 11 deletions(-)
>
> diff --git a/arch/powerpc/include/asm/kvm_book3s_asm.h b/arch/powerpc/include/asm/kvm_book3s_asm.h
> index 7cea76f11c26..83596f32f50b 100644
> --- a/arch/powerpc/include/asm/kvm_book3s_asm.h
> +++ b/arch/powerpc/include/asm/kvm_book3s_asm.h
> @@ -104,6 +104,10 @@ struct kvmppc_host_state {
> u8 napping;
>
> #ifdef CONFIG_KVM_BOOK3S_HV_POSSIBLE
> + /*
> + * hwthread_req/hwthread_state pair is used to pull sibling threads
> + * out of guest on pre-ISAv3.0B CPUs where threads share MMU.
> + */
> u8 hwthread_req;
> u8 hwthread_state;
> u8 host_ipi;
> diff --git a/arch/powerpc/kernel/idle_book3s.S b/arch/powerpc/kernel/idle_book3s.S
> index e6252c5a57a4..3ab73f9223e4 100644
> --- a/arch/powerpc/kernel/idle_book3s.S
> +++ b/arch/powerpc/kernel/idle_book3s.S
> @@ -243,12 +243,6 @@ enter_winkle:
> * r3 - PSSCR value corresponding to the requested stop state.
> */
> power_enter_stop:
> -#ifdef CONFIG_KVM_BOOK3S_HV_POSSIBLE
> - /* Tell KVM we're entering idle */
> - li r4,KVM_HWTHREAD_IN_IDLE
> - /* DO THIS IN REAL MODE! See comment above. */
> - stb r4,HSTATE_HWTHREAD_STATE(r13)
> -#endif
> /*
> * Check if we are executing the lite variant with ESL=EC=0
> */
> @@ -435,6 +429,7 @@ ALT_FTR_SECTION_END_IFSET(CPU_FTR_ARCH_300)
> mr r3,r12
>
> #ifdef CONFIG_KVM_BOOK3S_HV_POSSIBLE
> +BEGIN_FTR_SECTION
> li r0,KVM_HWTHREAD_IN_KERNEL
> stb r0,HSTATE_HWTHREAD_STATE(r13)
> /* Order setting hwthread_state vs. testing hwthread_req */
> @@ -444,6 +439,7 @@ ALT_FTR_SECTION_END_IFSET(CPU_FTR_ARCH_300)
> beq 1f
> b kvm_start_guest
> 1:
> +END_FTR_SECTION_IFCLR(CPU_FTR_ARCH_300)
This would be 7 nops on power9. Should we move this to a different
function and do a bl to that?
> #endif
>
> /* Return SRR1 from power7_nap() */
> diff --git a/arch/powerpc/kvm/book3s_hv.c b/arch/powerpc/kvm/book3s_hv.c
> index 359c79cdf0cc..bb1ab14f963a 100644
> --- a/arch/powerpc/kvm/book3s_hv.c
> +++ b/arch/powerpc/kvm/book3s_hv.c
> @@ -2111,6 +2111,16 @@ static int kvmppc_grab_hwthread(int cpu)
> struct paca_struct *tpaca;
> long timeout = 10000;
>
> + /*
> + * ISA v3.0 idle routines do not set hwthread_state or test
> + * hwthread_req, so they can not grab idle threads.
> + */
> + if (cpu_has_feature(CPU_FTR_ARCH_300)) {
> + WARN_ON(1);
> + pr_err("KVM: can not control sibling threads\n");
> + return -EBUSY;
> + }
> +
> tpaca = &paca[cpu];
>
> /* Ensure the thread won't go into the kernel if it wakes */
> @@ -2145,12 +2155,26 @@ static void kvmppc_release_hwthread(int cpu)
> struct paca_struct *tpaca;
>
> tpaca = &paca[cpu];
> - tpaca->kvm_hstate.hwthread_req = 0;
> tpaca->kvm_hstate.kvm_vcpu = NULL;
> tpaca->kvm_hstate.kvm_vcore = NULL;
> tpaca->kvm_hstate.kvm_split_mode = NULL;
> }
>
> +static void kvmppc_release_hwthread_secondary(int cpu)
> +{
> + struct paca_struct *tpaca;
> +
> + if (cpu_has_feature(CPU_FTR_ARCH_300)) {
> + WARN_ON(1);
> + return;
> + }
> +
> + tpaca = &paca[cpu];
> + tpaca->kvm_hstate.hwthread_req = 0;
> + kvmppc_release_hwthread(cpu);
> +}
> +
> +
Extra blank line not needed.
> static void radix_flush_cpu(struct kvm *kvm, int cpu, struct kvm_vcpu *vcpu)
> {
> int i;
> @@ -2274,7 +2298,7 @@ static int on_primary_thread(void)
> if (kvmppc_grab_hwthread(cpu + thr)) {
> /* Couldn't grab one; let the others go */
> do {
> - kvmppc_release_hwthread(cpu + thr);
> + kvmppc_release_hwthread_secondary(cpu + thr);
> } while (--thr > 0);
> return 0;
> }
> @@ -2702,8 +2726,9 @@ static noinline void kvmppc_run_core(struct kvmppc_vcore *vc)
> kvmppc_vcore_preempt(pvc);
> spin_unlock(&pvc->lock);
> }
> - for (i = 0; i < controlled_threads; ++i)
> - kvmppc_release_hwthread(pcpu + i);
> + for (i = 1; i < controlled_threads; ++i)
> + kvmppc_release_hwthread_secondary(pcpu + i);
> + kvmppc_release_hwthread(pcpu);
> return;
> }
>
> @@ -2858,11 +2883,13 @@ static noinline void kvmppc_run_core(struct kvmppc_vcore *vc)
>
> /* Let secondaries go back to the offline loop */
> for (i = 0; i < controlled_threads; ++i) {
> - kvmppc_release_hwthread(pcpu + i);
> if (sip && sip->napped[i])
> kvmppc_ipi_thread(pcpu + i);
> cpumask_clear_cpu(pcpu + i, &vc->kvm->arch.cpu_in_guest);
> }
We are sending an IPI to the thread that has exited the guest and is
currently napping. The IPI wakes it up so that it can executes
offline loop. But we haven't released the hwthread yet, which means
that hwthread_req for this thread is still set.
The thread wakes up from nap, executes the pnv_powersave_wakeup code
where it can enter kvm_start_guest. Is this a legitimate race or am I
missing something?
> + for (i = 1; i < controlled_threads; ++i)
> + kvmppc_release_hwthread_secondary(pcpu + i);
> + kvmppc_release_hwthread(pcpu);
>
> spin_unlock(&vc->lock);
>
> diff --git a/arch/powerpc/kvm/book3s_hv_rmhandlers.S b/arch/powerpc/kvm/book3s_hv_rmhandlers.S
> index c52184a8efdf..3e024fd71fe8 100644
> --- a/arch/powerpc/kvm/book3s_hv_rmhandlers.S
> +++ b/arch/powerpc/kvm/book3s_hv_rmhandlers.S
> @@ -149,9 +149,11 @@ END_FTR_SECTION_IFSET(CPU_FTR_ARCH_207S)
> subf r4, r4, r3
> mtspr SPRN_DEC, r4
>
> +BEGIN_FTR_SECTION
> /* hwthread_req may have got set by cede or no vcpu, so clear it */
> li r0, 0
> stb r0, HSTATE_HWTHREAD_REQ(r13)
> +END_FTR_SECTION_IFCLR(CPU_FTR_ARCH_300)
>
> /*
> * For external interrupts we need to call the Linux
> @@ -314,6 +316,7 @@ kvm_novcpu_exit:
> * Relocation is off and most register values are lost.
> * r13 points to the PACA.
> * r3 contains the SRR1 wakeup value, SRR1 is trashed.
> + * This is not used by ISAv3.0B processors.
> */
> .globl kvm_start_guest
> kvm_start_guest:
> @@ -432,6 +435,9 @@ kvm_secondary_got_guest:
> * While waiting we also need to check if we get given a vcpu to run.
> */
> kvm_no_guest:
> +BEGIN_FTR_SECTION
> + twi 31,0,0
> +END_FTR_SECTION_IFSET(CPU_FTR_ARCH_300)
> lbz r3, HSTATE_HWTHREAD_REQ(r13)
> cmpwi r3, 0
> bne 53f
> @@ -2509,8 +2515,10 @@ kvm_do_nap:
> clrrdi r0, r0, 1
> mtspr SPRN_CTRLT, r0
>
> +BEGIN_FTR_SECTION
> li r0,1
> stb r0,HSTATE_HWTHREAD_REQ(r13)
> +END_FTR_SECTION_IFCLR(CPU_FTR_ARCH_300)
> mfspr r5,SPRN_LPCR
> ori r5,r5,LPCR_PECE0 | LPCR_PECE1
> BEGIN_FTR_SECTION
> --
> 2.11.0
>
More information about the Linuxppc-dev
mailing list