From: Sean Christopherson <seanjc@google.com>
To: isaku.yamahata@intel.com
Cc: linux-kernel@vger.kernel.org, kvm@vger.kernel.org,
Paolo Bonzini <pbonzini@redhat.com>,
Thomas Gleixner <tglx@linutronix.de>,
Marc Zyngier <maz@kernel.org>, Will Deacon <will@kernel.org>,
isaku.yamahata@gmail.com, Kai Huang <kai.huang@intel.com>,
Chao Gao <chao.gao@intel.com>,
Atish Patra <atishp@atishpatra.org>,
Shaokun Zhang <zhangshaokun@hisilicon.com>,
Daniel Lezcano <daniel.lezcano@linaro.org>,
Huang Ying <ying.huang@intel.com>,
Huacai Chen <chenhuacai@kernel.org>,
Dave Hansen <dave.hansen@linux.intel.com>,
Borislav Petkov <bp@alien8.de>
Subject: Re: [PATCH v5 09/30] KVM: Drop kvm_count_lock and instead protect kvm_usage_count with kvm_lock
Date: Wed, 12 Oct 2022 20:14:16 +0000 [thread overview]
Message-ID: <Y0cgGEr8nKpOBLrQ@google.com> (raw)
In-Reply-To: <92836b09c8e0f19f8e506008e45993881d22b6d1.1663869838.git.isaku.yamahata@intel.com>
On Thu, Sep 22, 2022, isaku.yamahata@intel.com wrote:
> From: Isaku Yamahata <isaku.yamahata@intel.com>
>
> Because kvm_count_lock unnecessarily complicates the KVM locking convention
> Drop kvm_count_lock and instead protect kvm_usage_count with kvm_lock for
> simplicity. kvm_arch_hardware_enable/disable() callbacks depend on
> non-preemptiblity with the spin lock. Add preempt_disable/enable()
> around hardware enable/disable callback to keep the assumption.
There's the other "minor" wrinkle that prior to patch 7, "KVM: Rename and move
CPUHP_AP_KVM_STARTING to ONLINE section, kvm_online_cpu() was called with IRQs
disabled and couldn't sleep, i.e. couldn't acquire a mutex. That's very important
to capture in the changelog.
> Because kvm_suspend() and kvm_resume() is called with interrupt disabled,
> they don't need preempt_disable/enable() pair.
>
> Opportunistically add some comments on locking.
>
> Suggested-by: Sean Christopherson <seanjc@google.com>
> Signed-off-by: Isaku Yamahata <isaku.yamahata@intel.com>
...
> @@ -5028,13 +5029,20 @@ static int kvm_online_cpu(unsigned int cpu)
> if (kvm_usage_count) {
> WARN_ON_ONCE(atomic_read(&hardware_enable_failed));
>
> + /*
> + * arch callback kvm_arch_hardware_eanble() assumes that
s/eanble/enable
Though even better would be to avoid function names entirely.
> + * preemption is disabled for historical reason. Disable
> + * preemption until all arch callbacks are fixed.
> + */
Probably better to put this comment above to the WARN_ON_ONCE() in hardware_enable_nolock()
since that's where the oddity and dependency on arch behavior lies. And then it
can be turned into a FIXME, e.g.
/*
* FIXME: drop the "preemption disabled" requirement here and in the
* disable path once all arch code plays nice with preemption.
*/
> + preempt_disable();
> hardware_enable_nolock(NULL);
> + preempt_enable();
> if (atomic_read(&hardware_enable_failed)) {
> atomic_set(&hardware_enable_failed, 0);
> ret = -EIO;
> }
> }
> - raw_spin_unlock(&kvm_count_lock);
> + mutex_unlock(&kvm_lock);
> return ret;
> }
>
> @@ -5042,6 +5050,8 @@ static void hardware_disable_nolock(void *junk)
> {
> int cpu = raw_smp_processor_id();
>
> + WARN_ON_ONCE(preemptible());
> +
> if (!cpumask_test_cpu(cpu, cpus_hardware_enabled))
> return;
> cpumask_clear_cpu(cpu, cpus_hardware_enabled);
> @@ -5050,10 +5060,18 @@ static void hardware_disable_nolock(void *junk)
>
> static int kvm_offline_cpu(unsigned int cpu)
> {
> - raw_spin_lock(&kvm_count_lock);
> - if (kvm_usage_count)
> + mutex_lock(&kvm_lock);
> + if (kvm_usage_count) {
> + /*
> + * arch callback kvm_arch_hardware_disable() assumes that
> + * preemption is disabled for historical reason. Disable
> + * preemption until all arch callbacks are fixed.
> + */
I vote to drop this comment and instead document everything in the enable FIXME
(see above).
> + preempt_disable();
> hardware_disable_nolock(NULL);
> - raw_spin_unlock(&kvm_count_lock);
> + preempt_enable();
> + }
> + mutex_unlock(&kvm_lock);
> return 0;
> }
...
> @@ -5708,15 +5728,27 @@ static void kvm_init_debug(void)
>
> static int kvm_suspend(void)
> {
> - if (kvm_usage_count)
> + /*
> + * The caller ensures that CPU hotplug is disabled by
> + * cpu_hotplug_disable() and other CPUs are offlined. No need for
> + * locking.
Disabling CPU hotplug prevents racing with kvm_online_cpu()/kvm_offline_cpu(), but
doesn't prevent racing with hardware_enable_all()/hardware_disable_all().
And the lockdep doesn't mesh with the comment, which explains why kvm_lock doesn't
_need_ to be held, but not why kvm_lock _can't_ be held.
Maybe this?
/*
* Secondary CPUs and CPU hotplug are disabled across the suspend/resume
* callbacks, i.e. no need to acquire kvm_lock to ensure the usage count
* is stable. Assert that kvm_lock is not held as a paranoid sanity
* check that the system isn't suspended when KVM is enabling hardware.
*/
> + */
> + lockdep_assert_not_held(&kvm_lock);
> +
> + if (kvm_usage_count) {
> + /*
> + * Because kvm_suspend() is called with interrupt disabled, no
> + * need to disable preemption.
> + */
Add a lockdep and drop the comment, e.g. below the lockdep_assert_not_held(), add
lockdep_assert_irqs_disabled();
That covers the "why doesn't this disable preemption" _and_ enforces that IRQs are
indeed disabled.
> hardware_disable_nolock(NULL);
> + }
> return 0;
> }
>
> static void kvm_resume(void)
> {
> if (kvm_usage_count) {
> - lockdep_assert_not_held(&kvm_count_lock);
> + lockdep_assert_not_held(&kvm_lock);
> hardware_enable_nolock(NULL);
> }
> }
> --
> 2.25.1
>
next prev parent reply other threads:[~2022-10-12 20:14 UTC|newest]
Thread overview: 45+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-09-22 18:20 [PATCH v5 00/30] KVM: hardware enable/disable reorganize isaku.yamahata
2022-09-22 18:20 ` [PATCH v5 01/30] KVM: x86: Drop kvm_user_return_msr_cpu_online() isaku.yamahata
2022-09-22 18:20 ` [PATCH v5 02/30] KVM: x86: Use this_cpu_ptr() instead of per_cpu_ptr(smp_processor_id()) isaku.yamahata
2022-09-22 18:20 ` [PATCH v5 03/30] KVM: x86: Move check_processor_compatibility from init ops to runtime ops isaku.yamahata
2022-10-11 19:53 ` Sean Christopherson
2022-09-22 18:20 ` [PATCH v5 04/30] Partially revert "KVM: Pass kvm_init()'s opaque param to additional arch funcs" isaku.yamahata
2022-09-22 18:20 ` [PATCH v5 05/30] KVM: Provide more information in kernel log if hardware enabling fails isaku.yamahata
2022-10-12 19:45 ` Sean Christopherson
2022-09-22 18:20 ` [PATCH v5 06/30] KVM: arm64: Simplify the CPUHP logic isaku.yamahata
2022-09-22 18:20 ` [PATCH v5 07/30] KVM: Rename and move CPUHP_AP_KVM_STARTING to ONLINE section isaku.yamahata
2022-09-22 18:20 ` [PATCH v5 08/30] KVM: Do compatibility checks on hotplugged CPUs isaku.yamahata
2022-09-22 18:20 ` [PATCH v5 09/30] KVM: Drop kvm_count_lock and instead protect kvm_usage_count with kvm_lock isaku.yamahata
2022-10-12 20:14 ` Sean Christopherson [this message]
2022-09-22 18:20 ` [PATCH v5 10/30] KVM: Add arch hooks when VM is added/deleted isaku.yamahata
2022-10-04 0:16 ` Isaku Yamahata
2022-10-12 20:43 ` Sean Christopherson
2022-09-22 18:20 ` [PATCH v5 11/30] KVM: Add arch hook for reboot event isaku.yamahata
2022-10-12 20:47 ` Sean Christopherson
2022-09-22 18:20 ` [PATCH v5 12/30] KVM: Add arch hook for suspend isaku.yamahata
2022-09-22 18:20 ` [PATCH v5 13/30] KVM: Add arch hook for resume event isaku.yamahata
2022-09-22 18:20 ` [PATCH v5 14/30] KVM: Add arch hook for cpu online event isaku.yamahata
2022-09-22 18:20 ` [PATCH v5 15/30] KVM: Add arch hook for cpu offline event isaku.yamahata
2022-09-22 18:20 ` [PATCH v5 16/30] KVM: Remove on_each_cpu(hardware_disable_nolock) in kvm_exit() isaku.yamahata
2022-10-12 20:50 ` Sean Christopherson
2022-09-22 18:20 ` [PATCH v5 17/30] KVM: Move out KVM arch PM hooks and hardware enable/disable logic isaku.yamahata
2022-10-12 21:10 ` Sean Christopherson
2022-09-22 18:20 ` [PATCH v5 18/30] KVM: kvm_arch.c: Remove _nolock post fix isaku.yamahata
2022-09-22 18:20 ` [PATCH v5 19/30] KVM: kvm_arch.c: Remove a global variable, hardware_enable_failed isaku.yamahata
2022-09-22 18:20 ` [PATCH v5 20/30] KVM: Introduce an arch wrapper to check all processor compatibility isaku.yamahata
2022-09-22 18:20 ` [PATCH v5 21/30] KVM: x86: Duplicate arch callbacks related to pm events and compat check isaku.yamahata
2022-09-22 18:20 ` [PATCH v5 22/30] KVM: x86: Move TSC fixup logic to KVM arch resume callback isaku.yamahata
2022-09-22 18:20 ` [PATCH v5 23/30] KVM: Eliminate kvm_arch_post_init_vm() isaku.yamahata
2022-09-22 18:20 ` [PATCH v5 24/30] KVM: Add config to not compile kvm_arch.c isaku.yamahata
2022-09-22 18:20 ` [PATCH v5 25/30] KVM: x86: Delete kvm_arch_hardware_enable/disable() isaku.yamahata
2022-09-22 18:20 ` [PATCH v5 26/30] KVM: x86: Make x86 processor compat check callback empty isaku.yamahata
2022-09-22 18:20 ` [PATCH v5 27/30] RFC: KVM: powerpc: Move processor compatibility check to hardware setup isaku.yamahata
2022-09-23 6:58 ` Michael Ellerman
2022-09-27 0:40 ` Isaku Yamahata
2022-09-22 18:20 ` [PATCH v5 28/30] KVM: Eliminate kvm_arch_check_processor_compat() isaku.yamahata
2022-09-22 18:20 ` [PATCH v5 29/30] RFC: KVM: x86: Remove cpus_hardware_enabled and related sanity check isaku.yamahata
2022-09-22 18:20 ` [PATCH v5 30/30] RFC: KVM: " isaku.yamahata
2022-10-13 0:25 ` [PATCH v5 00/30] KVM: hardware enable/disable reorganize Sean Christopherson
2022-10-14 4:04 ` Sean Christopherson
2022-11-02 18:02 ` Paolo Bonzini
2022-11-02 18:59 ` Sean Christopherson
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=Y0cgGEr8nKpOBLrQ@google.com \
--to=seanjc@google.com \
--cc=atishp@atishpatra.org \
--cc=bp@alien8.de \
--cc=chao.gao@intel.com \
--cc=chenhuacai@kernel.org \
--cc=daniel.lezcano@linaro.org \
--cc=dave.hansen@linux.intel.com \
--cc=isaku.yamahata@gmail.com \
--cc=isaku.yamahata@intel.com \
--cc=kai.huang@intel.com \
--cc=kvm@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=maz@kernel.org \
--cc=pbonzini@redhat.com \
--cc=tglx@linutronix.de \
--cc=will@kernel.org \
--cc=ying.huang@intel.com \
--cc=zhangshaokun@hisilicon.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).