From: Alexander Graf <agraf@suse.de>
To: Scott Wood <scottwood@freescale.com>
Cc: linuxppc-dev@lists.ozlabs.org, kvm@vger.kernel.org,
kvm-ppc@vger.kernel.org
Subject: Re: [PATCH 24/30] KVM: PPC: booke: call resched after every exit
Date: Mon, 20 Feb 2012 14:17:22 +0100 [thread overview]
Message-ID: <4B831F36-4BEB-4BDF-869D-117617B4EE44@suse.de> (raw)
In-Reply-To: <4F3EDC01.5000400@freescale.com>
On 18.02.2012, at 00:00, Scott Wood wrote:
> On 02/17/2012 11:13 AM, Alexander Graf wrote:
>> Instead of checking whether we should reschedule only when we exited
>> due to an interrupt, let's always check before entering the guest =
back
>> again. This gets the target more in line with the other archs.
>>=20
>> Signed-off-by: Alexander Graf <agraf@suse.de>
>> ---
>> arch/powerpc/kvm/booke.c | 15 ++++++++++-----
>> 1 files changed, 10 insertions(+), 5 deletions(-)
>>=20
>> diff --git a/arch/powerpc/kvm/booke.c b/arch/powerpc/kvm/booke.c
>> index bfb2092..de30b6d 100644
>> --- a/arch/powerpc/kvm/booke.c
>> +++ b/arch/powerpc/kvm/booke.c
>> @@ -572,6 +572,7 @@ int kvmppc_handle_exit(struct kvm_run *run, =
struct kvm_vcpu *vcpu,
>> unsigned int exit_nr)
>> {
>> int r =3D RESUME_HOST;
>> + int resched_needed =3D 1;
>>=20
>> /* update before a new last_exit_type is rewritten */
>> kvmppc_update_timing_stats(vcpu);
>> @@ -602,25 +603,21 @@ int kvmppc_handle_exit(struct kvm_run *run, =
struct kvm_vcpu *vcpu,
>>=20
>> switch (exit_nr) {
>> case BOOKE_INTERRUPT_MACHINE_CHECK:
>> - kvm_resched(vcpu);
>> r =3D RESUME_GUEST;
>> break;
>>=20
>> case BOOKE_INTERRUPT_EXTERNAL:
>> kvmppc_account_exit(vcpu, EXT_INTR_EXITS);
>> - kvm_resched(vcpu);
>> r =3D RESUME_GUEST;
>> break;
>>=20
>> case BOOKE_INTERRUPT_DECREMENTER:
>> kvmppc_account_exit(vcpu, DEC_EXITS);
>> - kvm_resched(vcpu);
>> r =3D RESUME_GUEST;
>> break;
>>=20
>> case BOOKE_INTERRUPT_DOORBELL:
>> kvmppc_account_exit(vcpu, DBELL_EXITS);
>> - kvm_resched(vcpu);
>> r =3D RESUME_GUEST;
>> break;
>>=20
>> @@ -869,8 +866,16 @@ int kvmppc_handle_exit(struct kvm_run *run, =
struct kvm_vcpu *vcpu,
>> BUG();
>> }
>>=20
>> - local_irq_disable();
>> + /* make sure we reschedule if we need to */
>> + while (resched_needed) {
>> + local_irq_disable();
>>=20
>> + resched_needed =3D need_resched();
>> + if (resched_needed) {
>> + local_irq_enable();
>> + cond_resched();
>> + }
>> + }
>> kvmppc_core_prepare_to_enter(vcpu);
>>=20
>> if (!(r & RESUME_HOST)) {
>=20
> kvmppc_core_prepare_to_enter can enable interrupts (and block) if =
guest
> MSR_WE is set. We may take an interrupt that wants a resched after
> waking but before interrupts are disabled again.
>=20
> We also want to check for a resched in kvmppc_vcpu_run. So, the =
resched
> check belongs in kvmppc_core_prepare_to_enter, something like:
>=20
> /* Check pending exceptions and deliver one, if possible. */
> void kvmppc_core_prepare_to_enter(struct kvm_vcpu *vcpu)
> {
> WARN_ON_ONCE(!irqs_disabled());
>=20
> while (true) {
> if (signal_pending(current))
> break;
>=20
> if (need_resched()) {
> local_irq_enable();
> cond_resched();
> local_irq_disable();
> continue;
> }
>=20
> kvmppc_core_check_exceptions(vcpu);
>=20
> if (vcpu->arch.shared->msr & MSR_WE) {
> local_irq_enable();
> kvm_vcpu_block(vcpu);
> local_irq_disable();
> =09
> kvmppc_set_exit_type(vcpu,
> EMULATED_MTMSRWE_EXITS);
> continue;
> }
>=20
> break;
> }
> }
>=20
> It would be simpler (both here and in the idle hcall) if we could just
> drop support for CONFIG_PREEMPT=3Dn. :-P
When running with CONFIG_PREEMPT=3Dn we don't have to worry about =
interrupts being enabled though, since we only preempt on known good =
checkpoints, right? While for CONFIG_PREEMPT=3Dy we can always get =
preempted, rendering most of these checks void.
So essentially we could just be lazy and do a "best effort" resched =
check, but not worry about races wrt guest entry/exit, right? And for =
real preemption modes we don't have to worry about any of the resched =
stuff IIUC, correct?
Alex
next prev parent reply other threads:[~2012-02-20 13:17 UTC|newest]
Thread overview: 44+ messages / expand[flat|nested] mbox.gz Atom feed top
2012-02-17 17:13 [PATCH 00/30] KVM: PPC: e500mc support Alexander Graf
2012-02-17 17:13 ` [PATCH 01/30] powerpc/booke: Set CPU_FTR_DEBUG_LVL_EXC on 32-bit Alexander Graf
2012-02-17 17:13 ` [PATCH 02/30] powerpc/e500: split CPU_FTRS_ALWAYS/CPU_FTRS_POSSIBLE Alexander Graf
2012-02-17 17:13 ` [PATCH 03/30] KVM: PPC: factor out lpid allocator from book3s_64_mmu_hv Alexander Graf
2012-02-17 17:13 ` [PATCH 04/30] KVM: PPC: booke: add booke-level vcpu load/put Alexander Graf
2012-02-17 17:13 ` [PATCH 05/30] KVM: PPC: booke: Move vm core init/destroy out of booke.c Alexander Graf
2012-02-17 17:13 ` [PATCH 06/30] KVM: PPC: e500: rename e500_tlb.h to e500.h Alexander Graf
2012-02-17 17:13 ` [PATCH 07/30] KVM: PPC: e500: merge <asm/kvm_e500.h> into arch/powerpc/kvm/e500.h Alexander Graf
2012-02-17 17:13 ` [PATCH 08/30] KVM: PPC: e500: clean up arch/powerpc/kvm/e500.h Alexander Graf
2012-02-17 17:13 ` [PATCH 09/30] KVM: PPC: e500: refactor core-specific TLB code Alexander Graf
2012-02-17 17:13 ` [PATCH 10/30] KVM: PPC: e500: Track TLB1 entries with a bitmap Alexander Graf
2012-02-17 17:13 ` [PATCH 11/30] KVM: PPC: e500: emulate tlbilx Alexander Graf
2012-02-17 17:13 ` [PATCH 12/30] powerpc/booke: Provide exception macros with interrupt name Alexander Graf
2012-02-17 17:13 ` [PATCH 13/30] KVM: PPC: booke: category E.HV (GS-mode) support Alexander Graf
2012-02-17 21:12 ` Scott Wood
2012-02-20 11:40 ` Alexander Graf
2012-02-17 17:13 ` [PATCH 14/30] KVM: PPC: booke: standard PPC floating point support Alexander Graf
2012-02-17 17:13 ` [PATCH 15/30] KVM: PPC: e500mc support Alexander Graf
2012-02-17 17:13 ` [PATCH 16/30] KVM: PPC: e500mc: Add doorbell emulation support Alexander Graf
2012-02-17 21:55 ` Scott Wood
2012-02-17 21:57 ` Scott Wood
2012-02-20 11:49 ` Alexander Graf
2012-02-20 15:39 ` Scott Wood
2012-02-20 15:42 ` Alexander Graf
2012-02-17 17:13 ` [PATCH 17/30] KVM: PPC: e500mc: implicitly set MSR_GS Alexander Graf
2012-02-17 17:13 ` [PATCH 18/30] KVM: PPC: e500mc: Move r1/r2 restoration very early Alexander Graf
2012-02-17 17:13 ` [PATCH 19/30] KVM: PPC: e500mc: add load inst fixup Alexander Graf
2012-02-17 23:17 ` Scott Wood
2012-02-17 17:13 ` [PATCH 20/30] KVM: PPC: rename CONFIG_KVM_E500 -> CONFIG_KVM_E500V2 Alexander Graf
2012-02-17 17:13 ` [PATCH 21/30] KVM: PPC: make e500v2 and e500mc mutually exclusive Alexander Graf
2012-02-17 22:13 ` Scott Wood
2012-02-17 17:13 ` [PATCH 22/30] KVM: PPC: booke: remove leftover debugging Alexander Graf
2012-02-17 17:13 ` [PATCH 23/30] KVM: PPC: booke: deliver program int on emulation failure Alexander Graf
2012-02-17 17:13 ` [PATCH 24/30] KVM: PPC: booke: call resched after every exit Alexander Graf
2012-02-17 23:00 ` Scott Wood
2012-02-20 13:17 ` Alexander Graf [this message]
2012-02-20 17:18 ` Scott Wood
2012-02-17 17:13 ` [PATCH 25/30] KVM: PPC: booke: BOOKE_IRQPRIO_MAX is n+1 Alexander Graf
2012-02-17 17:13 ` [PATCH 26/30] KVM: PPC: bookehv: fix exit timing Alexander Graf
2012-02-17 17:13 ` [PATCH 27/30] KVM: PPC: bookehv: remove negation for CONFIG_64BIT Alexander Graf
2012-02-17 17:13 ` [PATCH 28/30] KVM: PPC: bookehv: remove SET_VCPU Alexander Graf
2012-02-17 17:13 ` [PATCH 29/30] KVM: PPC: bookehv: disable MAS register updates early Alexander Graf
2012-02-17 17:13 ` [PATCH 30/30] KVM: PPC: bookehv: add comment about shadow_msr Alexander Graf
-- strict thread matches above, loose matches on Subject: below --
2012-02-17 16:56 [PATCH 00/30] KVM: PPC: e500mc support Alexander Graf
2012-02-17 16:56 ` [PATCH 24/30] KVM: PPC: booke: call resched after every exit Alexander Graf
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=4B831F36-4BEB-4BDF-869D-117617B4EE44@suse.de \
--to=agraf@suse.de \
--cc=kvm-ppc@vger.kernel.org \
--cc=kvm@vger.kernel.org \
--cc=linuxppc-dev@lists.ozlabs.org \
--cc=scottwood@freescale.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).