[PATCH v2] KVM: x86/mmu: Update number of zapped pages even if page list is stable

[PATCH v2] KVM: x86/mmu: Update number of zapped pages even if page list is stable

[Sean Christopherson]

When zapping obsolete pages, update the running count of zapped pages
regardless of whether or not the list has become unstable due to zapping
a shadow page with its own child shadow pages.  If the VM is backed by
mostly 4kb pages, KVM can zap an absurd number of SPTEs without bumping
the batch count and thus without yielding.  In the worst case scenario,
this can cause a soft lokcup.

 watchdog: BUG: soft lockup - CPU#12 stuck for 22s! [dirty_log_perf_:13020]
   RIP: 0010:workingset_activation+0x19/0x130
   mark_page_accessed+0x266/0x2e0
   kvm_set_pfn_accessed+0x31/0x40
   mmu_spte_clear_track_bits+0x136/0x1c0
   drop_spte+0x1a/0xc0
   mmu_page_zap_pte+0xef/0x120
   __kvm_mmu_prepare_zap_page+0x205/0x5e0
   kvm_mmu_zap_all_fast+0xd7/0x190
   kvm_mmu_invalidate_zap_pages_in_memslot+0xe/0x10
   kvm_page_track_flush_slot+0x5c/0x80
   kvm_arch_flush_shadow_memslot+0xe/0x10
   kvm_set_memslot+0x1a8/0x5d0
   __kvm_set_memory_region+0x337/0x590
   kvm_vm_ioctl+0xb08/0x1040

Fixes: fbb158cb88b6 ("KVM: x86/mmu: Revert "Revert "KVM: MMU: zap pages in batch""")
Reported-by: David Matlack <dmatlack@google.com>
Reviewed-by: Ben Gardon <bgardon@google.com>
Cc: stable@vger.kernel.org
Signed-off-by: Sean Christopherson <seanjc@google.com>
---

v2:
 - Rebase to kvm/master, commit 30d7c5d60a88 ("KVM: SEV: expose...")
 - Collect Ben's review, modulo bad splat.
 - Copy+paste the correct splat and symptom. [David].

@David, I kept the unstable declaration out of the loop, mostly because I
really don't like putting declarations in loops, but also because
nr_zapped is declared out of the loop and I didn't want to change that
unnecessarily or make the code inconsistent.

 arch/x86/kvm/mmu/mmu.c | 10 ++++++----
 1 file changed, 6 insertions(+), 4 deletions(-)

diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c
index 0c839ee1282c..208c892136bf 100644
--- a/arch/x86/kvm/mmu/mmu.c
+++ b/arch/x86/kvm/mmu/mmu.c
@@ -5576,6 +5576,7 @@ static void kvm_zap_obsolete_pages(struct kvm *kvm)
 {
 	struct kvm_mmu_page *sp, *node;
 	int nr_zapped, batch = 0;
+	bool unstable;
 
 restart:
 	list_for_each_entry_safe_reverse(sp, node,
@@ -5607,11 +5608,12 @@ static void kvm_zap_obsolete_pages(struct kvm *kvm)
 			goto restart;
 		}
 
-		if (__kvm_mmu_prepare_zap_page(kvm, sp,
-				&kvm->arch.zapped_obsolete_pages, &nr_zapped)) {
-			batch += nr_zapped;
+		unstable = __kvm_mmu_prepare_zap_page(kvm, sp,
+				&kvm->arch.zapped_obsolete_pages, &nr_zapped);
+		batch += nr_zapped;
+
+		if (unstable)
 			goto restart;
-		}
 	}
 
 	/*
-- 
2.34.0.rc2.393.gf8c9666880-goog

Re: [PATCH v2] KVM: x86/mmu: Update number of zapped pages even if page list is stable

[David Matlack]

On Mon, Nov 29, 2021 at 11:52:33PM +0000, Sean Christopherson wrote:
> When zapping obsolete pages, update the running count of zapped pages
> regardless of whether or not the list has become unstable due to zapping
> a shadow page with its own child shadow pages.  If the VM is backed by
> mostly 4kb pages, KVM can zap an absurd number of SPTEs without bumping
> the batch count and thus without yielding.  In the worst case scenario,
> this can cause a soft lokcup.
> 
>  watchdog: BUG: soft lockup - CPU#12 stuck for 22s! [dirty_log_perf_:13020]
>    RIP: 0010:workingset_activation+0x19/0x130
>    mark_page_accessed+0x266/0x2e0
>    kvm_set_pfn_accessed+0x31/0x40
>    mmu_spte_clear_track_bits+0x136/0x1c0
>    drop_spte+0x1a/0xc0
>    mmu_page_zap_pte+0xef/0x120
>    __kvm_mmu_prepare_zap_page+0x205/0x5e0
>    kvm_mmu_zap_all_fast+0xd7/0x190
>    kvm_mmu_invalidate_zap_pages_in_memslot+0xe/0x10
>    kvm_page_track_flush_slot+0x5c/0x80
>    kvm_arch_flush_shadow_memslot+0xe/0x10
>    kvm_set_memslot+0x1a8/0x5d0
>    __kvm_set_memory_region+0x337/0x590
>    kvm_vm_ioctl+0xb08/0x1040
> 
> Fixes: fbb158cb88b6 ("KVM: x86/mmu: Revert "Revert "KVM: MMU: zap pages in batch""")
> Reported-by: David Matlack <dmatlack@google.com>
> Reviewed-by: Ben Gardon <bgardon@google.com>
> Cc: stable@vger.kernel.org
> Signed-off-by: Sean Christopherson <seanjc@google.com>

Reviewed-by: David Matlack <dmatlack@google.com>

> ---
> 
> v2:
>  - Rebase to kvm/master, commit 30d7c5d60a88 ("KVM: SEV: expose...")
>  - Collect Ben's review, modulo bad splat.
>  - Copy+paste the correct splat and symptom. [David].
> 
> @David, I kept the unstable declaration out of the loop, mostly because I
> really don't like putting declarations in loops, but also because
> nr_zapped is declared out of the loop and I didn't want to change that
> unnecessarily or make the code inconsistent.

Sounds good. For stable patches I definitely agree with not changing
variable declarations needlessly. And as a general rule, I get your
point on the other thread about shadowing variables and therefore
preferring to declare all variables at function scope.

I have some other ideas on this topic but it's out of scope for this
patch so I'll post it on the other thread.

> 
>  arch/x86/kvm/mmu/mmu.c | 10 ++++++----
>  1 file changed, 6 insertions(+), 4 deletions(-)
> 
> diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c
> index 0c839ee1282c..208c892136bf 100644
> --- a/arch/x86/kvm/mmu/mmu.c
> +++ b/arch/x86/kvm/mmu/mmu.c
> @@ -5576,6 +5576,7 @@ static void kvm_zap_obsolete_pages(struct kvm *kvm)
>  {
>  	struct kvm_mmu_page *sp, *node;
>  	int nr_zapped, batch = 0;
> +	bool unstable;
>  
>  restart:
>  	list_for_each_entry_safe_reverse(sp, node,
> @@ -5607,11 +5608,12 @@ static void kvm_zap_obsolete_pages(struct kvm *kvm)
>  			goto restart;
>  		}
>  
> -		if (__kvm_mmu_prepare_zap_page(kvm, sp,
> -				&kvm->arch.zapped_obsolete_pages, &nr_zapped)) {
> -			batch += nr_zapped;
> +		unstable = __kvm_mmu_prepare_zap_page(kvm, sp,
> +				&kvm->arch.zapped_obsolete_pages, &nr_zapped);
> +		batch += nr_zapped;
> +
> +		if (unstable)
>  			goto restart;
> -		}
>  	}
>  
>  	/*
> -- 
> 2.34.0.rc2.393.gf8c9666880-goog
> 

Re: [PATCH v2] KVM: x86/mmu: Update number of zapped pages even if page list is stable

[Sean Christopherson]

Very high latency ping, this is still problematic and still applies cleanly.

On Mon, Nov 29, 2021, Sean Christopherson wrote:
> When zapping obsolete pages, update the running count of zapped pages
> regardless of whether or not the list has become unstable due to zapping
> a shadow page with its own child shadow pages.  If the VM is backed by
> mostly 4kb pages, KVM can zap an absurd number of SPTEs without bumping
> the batch count and thus without yielding.  In the worst case scenario,
> this can cause a soft lokcup.
> 
>  watchdog: BUG: soft lockup - CPU#12 stuck for 22s! [dirty_log_perf_:13020]
>    RIP: 0010:workingset_activation+0x19/0x130
>    mark_page_accessed+0x266/0x2e0
>    kvm_set_pfn_accessed+0x31/0x40
>    mmu_spte_clear_track_bits+0x136/0x1c0
>    drop_spte+0x1a/0xc0
>    mmu_page_zap_pte+0xef/0x120
>    __kvm_mmu_prepare_zap_page+0x205/0x5e0
>    kvm_mmu_zap_all_fast+0xd7/0x190
>    kvm_mmu_invalidate_zap_pages_in_memslot+0xe/0x10
>    kvm_page_track_flush_slot+0x5c/0x80
>    kvm_arch_flush_shadow_memslot+0xe/0x10
>    kvm_set_memslot+0x1a8/0x5d0
>    __kvm_set_memory_region+0x337/0x590
>    kvm_vm_ioctl+0xb08/0x1040
> 
> Fixes: fbb158cb88b6 ("KVM: x86/mmu: Revert "Revert "KVM: MMU: zap pages in batch""")
> Reported-by: David Matlack <dmatlack@google.com>
> Reviewed-by: Ben Gardon <bgardon@google.com>
> Cc: stable@vger.kernel.org
> Signed-off-by: Sean Christopherson <seanjc@google.com>
> ---
> 
> v2:
>  - Rebase to kvm/master, commit 30d7c5d60a88 ("KVM: SEV: expose...")
>  - Collect Ben's review, modulo bad splat.
>  - Copy+paste the correct splat and symptom. [David].
> 
> @David, I kept the unstable declaration out of the loop, mostly because I
> really don't like putting declarations in loops, but also because
> nr_zapped is declared out of the loop and I didn't want to change that
> unnecessarily or make the code inconsistent.
> 
>  arch/x86/kvm/mmu/mmu.c | 10 ++++++----
>  1 file changed, 6 insertions(+), 4 deletions(-)
> 
> diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c
> index 0c839ee1282c..208c892136bf 100644
> --- a/arch/x86/kvm/mmu/mmu.c
> +++ b/arch/x86/kvm/mmu/mmu.c
> @@ -5576,6 +5576,7 @@ static void kvm_zap_obsolete_pages(struct kvm *kvm)
>  {
>  	struct kvm_mmu_page *sp, *node;
>  	int nr_zapped, batch = 0;
> +	bool unstable;
>  
>  restart:
>  	list_for_each_entry_safe_reverse(sp, node,
> @@ -5607,11 +5608,12 @@ static void kvm_zap_obsolete_pages(struct kvm *kvm)
>  			goto restart;
>  		}
>  
> -		if (__kvm_mmu_prepare_zap_page(kvm, sp,
> -				&kvm->arch.zapped_obsolete_pages, &nr_zapped)) {
> -			batch += nr_zapped;
> +		unstable = __kvm_mmu_prepare_zap_page(kvm, sp,
> +				&kvm->arch.zapped_obsolete_pages, &nr_zapped);
> +		batch += nr_zapped;
> +
> +		if (unstable)
>  			goto restart;
> -		}
>  	}
>  
>  	/*
> -- 
> 2.34.0.rc2.393.gf8c9666880-goog

Re: [PATCH v2] KVM: x86/mmu: Update number of zapped pages even if page list is stable

[Sean Christopherson]

On Fri, Apr 08, 2022, Sean Christopherson wrote:
> Very high latency ping, this is still problematic and still applies cleanly.

PING!  PING!  PING!  PING!

Don't make me write a script to ping you every hour :-)

> On Mon, Nov 29, 2021, Sean Christopherson wrote:
> > When zapping obsolete pages, update the running count of zapped pages
> > regardless of whether or not the list has become unstable due to zapping
> > a shadow page with its own child shadow pages.  If the VM is backed by
> > mostly 4kb pages, KVM can zap an absurd number of SPTEs without bumping
> > the batch count and thus without yielding.  In the worst case scenario,
> > this can cause a soft lokcup.
> > 
> >  watchdog: BUG: soft lockup - CPU#12 stuck for 22s! [dirty_log_perf_:13020]
> >    RIP: 0010:workingset_activation+0x19/0x130
> >    mark_page_accessed+0x266/0x2e0
> >    kvm_set_pfn_accessed+0x31/0x40
> >    mmu_spte_clear_track_bits+0x136/0x1c0
> >    drop_spte+0x1a/0xc0
> >    mmu_page_zap_pte+0xef/0x120
> >    __kvm_mmu_prepare_zap_page+0x205/0x5e0
> >    kvm_mmu_zap_all_fast+0xd7/0x190
> >    kvm_mmu_invalidate_zap_pages_in_memslot+0xe/0x10
> >    kvm_page_track_flush_slot+0x5c/0x80
> >    kvm_arch_flush_shadow_memslot+0xe/0x10
> >    kvm_set_memslot+0x1a8/0x5d0
> >    __kvm_set_memory_region+0x337/0x590
> >    kvm_vm_ioctl+0xb08/0x1040
> > 
> > Fixes: fbb158cb88b6 ("KVM: x86/mmu: Revert "Revert "KVM: MMU: zap pages in batch""")
> > Reported-by: David Matlack <dmatlack@google.com>
> > Reviewed-by: Ben Gardon <bgardon@google.com>
> > Cc: stable@vger.kernel.org
> > Signed-off-by: Sean Christopherson <seanjc@google.com>
> > ---
> > 
> > v2:
> >  - Rebase to kvm/master, commit 30d7c5d60a88 ("KVM: SEV: expose...")
> >  - Collect Ben's review, modulo bad splat.
> >  - Copy+paste the correct splat and symptom. [David].
> > 
> > @David, I kept the unstable declaration out of the loop, mostly because I
> > really don't like putting declarations in loops, but also because
> > nr_zapped is declared out of the loop and I didn't want to change that
> > unnecessarily or make the code inconsistent.
> > 
> >  arch/x86/kvm/mmu/mmu.c | 10 ++++++----
> >  1 file changed, 6 insertions(+), 4 deletions(-)
> > 
> > diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c
> > index 0c839ee1282c..208c892136bf 100644
> > --- a/arch/x86/kvm/mmu/mmu.c
> > +++ b/arch/x86/kvm/mmu/mmu.c
> > @@ -5576,6 +5576,7 @@ static void kvm_zap_obsolete_pages(struct kvm *kvm)
> >  {
> >  	struct kvm_mmu_page *sp, *node;
> >  	int nr_zapped, batch = 0;
> > +	bool unstable;
> >  
> >  restart:
> >  	list_for_each_entry_safe_reverse(sp, node,
> > @@ -5607,11 +5608,12 @@ static void kvm_zap_obsolete_pages(struct kvm *kvm)
> >  			goto restart;
> >  		}
> >  
> > -		if (__kvm_mmu_prepare_zap_page(kvm, sp,
> > -				&kvm->arch.zapped_obsolete_pages, &nr_zapped)) {
> > -			batch += nr_zapped;
> > +		unstable = __kvm_mmu_prepare_zap_page(kvm, sp,
> > +				&kvm->arch.zapped_obsolete_pages, &nr_zapped);
> > +		batch += nr_zapped;
> > +
> > +		if (unstable)
> >  			goto restart;
> > -		}
> >  	}
> >  
> >  	/*
> > -- 
> > 2.34.0.rc2.393.gf8c9666880-goog
>