[v2,1/3] KVM: MMU: Do not treat ZONE_DEVICE pages as being reserved

Commit Message

Sean Christopherson Nov. 11, 2019, 10:12 p.m. UTC

Explicitly exempt ZONE_DEVICE pages from kvm_is_reserved_pfn() and
instead manually handle ZONE_DEVICE on a case-by-case basis.  For things
like page refcounts, KVM needs to treat ZONE_DEVICE pages like normal
pages, e.g. put pages grabbed via gup().  But for flows such as setting
A/D bits or shifting refcounts for transparent huge pages, KVM needs to
to avoid processing ZONE_DEVICE pages as the flows in question lack the
underlying machinery for proper handling of ZONE_DEVICE pages.

This fixes a hang reported by Adam Borowski[*] in dev_pagemap_cleanup()
when running a KVM guest backed with /dev/dax memory, as KVM straight up
doesn't put any references to ZONE_DEVICE pages acquired by gup().

Note, Dan Williams proposed an alternative solution of doing put_page()
on ZONE_DEVICE pages immediately after gup() in order to simplify the
auditing needed to ensure is_zone_device_page() is called if and only if
the backing device is pinned (via gup()).  But that approach would break
kvm_vcpu_{un}map() as KVM requires the page to be pinned from map() 'til
unmap() when accessing guest memory, unlike KVM's secondary MMU, which
coordinates with mmu_notifier invalidations to avoid creating stale
page references, i.e. doesn't rely on pages being pinned.

[*] http://lkml.kernel.org/r/[e-mail]

Reported-by: Adam Borowski <[e-mail]>
Debugged-by: David Hildenbrand <[e-mail]>
Cc: Dan Williams <[e-mail]>
Cc: [e-mail]
Signed-off-by: Sean Christopherson <[e-mail]>
---
 arch/x86/kvm/mmu.c       |  8 ++++----
 include/linux/kvm_host.h |  1 +
 virt/kvm/kvm_main.c      | 26 +++++++++++++++++++++++---
 3 files changed, 28 insertions(+), 7 deletions(-)

Comments

Paolo Bonzini Nov. 11, 2019, 10:20 p.m. UTC | #1

On 11/11/19 23:12, Sean Christopherson wrote:
> Explicitly exempt ZONE_DEVICE pages from kvm_is_reserved_pfn() and
> instead manually handle ZONE_DEVICE on a case-by-case basis.  For things
> like page refcounts, KVM needs to treat ZONE_DEVICE pages like normal
> pages, e.g. put pages grabbed via gup().  But for flows such as setting
> A/D bits or shifting refcounts for transparent huge pages, KVM needs to
> to avoid processing ZONE_DEVICE pages as the flows in question lack the
> underlying machinery for proper handling of ZONE_DEVICE pages.
> 
> This fixes a hang reported by Adam Borowski[*] in dev_pagemap_cleanup()
> when running a KVM guest backed with /dev/dax memory, as KVM straight up
> doesn't put any references to ZONE_DEVICE pages acquired by gup().
> 
> Note, Dan Williams proposed an alternative solution of doing put_page()
> on ZONE_DEVICE pages immediately after gup() in order to simplify the
> auditing needed to ensure is_zone_device_page() is called if and only if
> the backing device is pinned (via gup()).  But that approach would break
> kvm_vcpu_{un}map() as KVM requires the page to be pinned from map() 'til
> unmap() when accessing guest memory, unlike KVM's secondary MMU, which
> coordinates with mmu_notifier invalidations to avoid creating stale
> page references, i.e. doesn't rely on pages being pinned.
> 
> [*] http://lkml.kernel.org/r/[e-mail]
> 
> Reported-by: Adam Borowski <[e-mail]>
> Debugged-by: David Hildenbrand <[e-mail]>
> Cc: Dan Williams <[e-mail]>
> Cc: [e-mail]
> Signed-off-by: Sean Christopherson <[e-mail]>
> ---
>  arch/x86/kvm/mmu.c       |  8 ++++----
>  include/linux/kvm_host.h |  1 +
>  virt/kvm/kvm_main.c      | 26 +++++++++++++++++++++++---
>  3 files changed, 28 insertions(+), 7 deletions(-)
> 
> diff --git a/arch/x86/kvm/mmu.c b/arch/x86/kvm/mmu.c
> index 24c23c66b226..bf82b1f2e834 100644
> --- a/arch/x86/kvm/mmu.c
> +++ b/arch/x86/kvm/mmu.c
> @@ -3306,7 +3306,7 @@ static void transparent_hugepage_adjust(struct kvm_vcpu *vcpu,
>  	 * here.
>  	 */
>  	if (!is_error_noslot_pfn(pfn) && !kvm_is_reserved_pfn(pfn) &&
> -	    level == PT_PAGE_TABLE_LEVEL &&
> +	    !kvm_is_zone_device_pfn(pfn) && level == PT_PAGE_TABLE_LEVEL &&
>  	    PageTransCompoundMap(pfn_to_page(pfn)) &&
>  	    !mmu_gfn_lpage_is_disallowed(vcpu, gfn, PT_DIRECTORY_LEVEL)) {
>  		unsigned long mask;
> @@ -5914,9 +5914,9 @@ static bool kvm_mmu_zap_collapsible_spte(struct kvm *kvm,
>  		 * the guest, and the guest page table is using 4K page size
>  		 * mapping if the indirect sp has level = 1.
>  		 */
> -		if (sp->role.direct &&
> -			!kvm_is_reserved_pfn(pfn) &&
> -			PageTransCompoundMap(pfn_to_page(pfn))) {
> +		if (sp->role.direct && !kvm_is_reserved_pfn(pfn) &&
> +		    !kvm_is_zone_device_pfn(pfn) &&
> +		    PageTransCompoundMap(pfn_to_page(pfn))) {
>  			pte_list_remove(rmap_head, sptep);
>  
>  			if (kvm_available_flush_tlb_with_range())
> diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h
> index a817e446c9aa..4ad1cd7d2d4d 100644
> --- a/include/linux/kvm_host.h
> +++ b/include/linux/kvm_host.h
> @@ -966,6 +966,7 @@ int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
>  void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
>  
>  bool kvm_is_reserved_pfn(kvm_pfn_t pfn);
> +bool kvm_is_zone_device_pfn(kvm_pfn_t pfn);
>  
>  struct kvm_irq_ack_notifier {
>  	struct hlist_node link;
> diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
> index b8534c6b8cf6..bc9d10a0a334 100644
> --- a/virt/kvm/kvm_main.c
> +++ b/virt/kvm/kvm_main.c
> @@ -149,10 +149,30 @@ __weak int kvm_arch_mmu_notifier_invalidate_range(struct kvm *kvm,
>  	return 0;
>  }
>  
> +bool kvm_is_zone_device_pfn(kvm_pfn_t pfn)
> +{
> +	/*
> +	 * The metadata used by is_zone_device_page() to determine whether or
> +	 * not a page is ZONE_DEVICE is guaranteed to be valid if and only if
> +	 * the device has been pinned, e.g. by get_user_pages().  WARN if the
> +	 * page_count() is zero to help detect bad usage of this helper.
> +	 */
> +	if (!pfn_valid(pfn) || WARN_ON_ONCE(!page_count(pfn_to_page(pfn))))
> +		return false;
> +
> +	return is_zone_device_page(pfn_to_page(pfn));
> +}
> +
>  bool kvm_is_reserved_pfn(kvm_pfn_t pfn)
>  {
> +	/*
> +	 * ZONE_DEVICE pages currently set PG_reserved, but from a refcounting
> +	 * perspective they are "normal" pages, albeit with slightly different
> +	 * usage rules.
> +	 */
>  	if (pfn_valid(pfn))
> -		return PageReserved(pfn_to_page(pfn));
> +		return PageReserved(pfn_to_page(pfn)) &&
> +		       !kvm_is_zone_device_pfn(pfn);
>  
>  	return true;
>  }
> @@ -1865,7 +1885,7 @@ EXPORT_SYMBOL_GPL(kvm_release_pfn_dirty);
>  
>  void kvm_set_pfn_dirty(kvm_pfn_t pfn)
>  {
> -	if (!kvm_is_reserved_pfn(pfn)) {
> +	if (!kvm_is_reserved_pfn(pfn) && !kvm_is_zone_device_pfn(pfn)) {
>  		struct page *page = pfn_to_page(pfn);
>  
>  		SetPageDirty(page);
> @@ -1875,7 +1895,7 @@ EXPORT_SYMBOL_GPL(kvm_set_pfn_dirty);
>  
>  void kvm_set_pfn_accessed(kvm_pfn_t pfn)
>  {
> -	if (!kvm_is_reserved_pfn(pfn))
> +	if (!kvm_is_reserved_pfn(pfn) && !kvm_is_zone_device_pfn(pfn))
>  		mark_page_accessed(pfn_to_page(pfn));
>  }
>  EXPORT_SYMBOL_GPL(kvm_set_pfn_accessed);
> 

Queued, thanks -- the other two will wait for after the merge window to
avoid pointless conflicts.

Paolo

Dan Williams Nov. 11, 2019, 10:39 p.m. UTC | #2

On Mon, Nov 11, 2019 at 2:12 PM Sean Christopherson
<[e-mail]> wrote:
>
> Explicitly exempt ZONE_DEVICE pages from kvm_is_reserved_pfn() and
> instead manually handle ZONE_DEVICE on a case-by-case basis.  For things
> like page refcounts, KVM needs to treat ZONE_DEVICE pages like normal
> pages, e.g. put pages grabbed via gup().  But for flows such as setting
> A/D bits or shifting refcounts for transparent huge pages, KVM needs to
> to avoid processing ZONE_DEVICE pages as the flows in question lack the
> underlying machinery for proper handling of ZONE_DEVICE pages.
>
> This fixes a hang reported by Adam Borowski[*] in dev_pagemap_cleanup()
> when running a KVM guest backed with /dev/dax memory, as KVM straight up
> doesn't put any references to ZONE_DEVICE pages acquired by gup().
>
> Note, Dan Williams proposed an alternative solution of doing put_page()
> on ZONE_DEVICE pages immediately after gup() in order to simplify the
> auditing needed to ensure is_zone_device_page() is called if and only if
> the backing device is pinned (via gup()).  But that approach would break
> kvm_vcpu_{un}map() as KVM requires the page to be pinned from map() 'til
> unmap() when accessing guest memory, unlike KVM's secondary MMU, which
> coordinates with mmu_notifier invalidations to avoid creating stale
> page references, i.e. doesn't rely on pages being pinned.
>
> [*] http://lkml.kernel.org/r/[e-mail]
>
> Reported-by: Adam Borowski <[e-mail]>
> Debugged-by: David Hildenbrand <[e-mail]>
> Cc: Dan Williams <[e-mail]>

Acked-by: Dan Williams <[e-mail]>

> Cc: [e-mail]

Perhaps add:

Fixes: 3565fce3a659 ("mm, x86: get_user_pages() for dax mappings")

...since that was the first kernel that broke KVM's assumption about
which pfn types needed to have the reference count managed.

Paolo Bonzini Nov. 11, 2019, 10:43 p.m. UTC | #3

On 11/11/19 23:39, Dan Williams wrote:
>> [*] http://lkml.kernel.org/r/[e-mail]
>>
>> Reported-by: Adam Borowski <[e-mail]>
>> Debugged-by: David Hildenbrand <[e-mail]>
>> Cc: Dan Williams <[e-mail]>
> Acked-by: Dan Williams <[e-mail]>
> 
>> Cc: [e-mail]
> Perhaps add:
> 
> Fixes: 3565fce3a659 ("mm, x86: get_user_pages() for dax mappings")
> 
> ...since that was the first kernel that broke KVM's assumption about
> which pfn types needed to have the reference count managed.
> 

Done, thanks!

Paolo

David Hildenbrand Nov. 12, 2019, 7:06 a.m. UTC | #4

> Am 11.11.2019 um 23:12 schrieb Sean Christopherson <[e-mail]>:
> 
> Explicitly exempt ZONE_DEVICE pages from kvm_is_reserved_pfn() and
> instead manually handle ZONE_DEVICE on a case-by-case basis.  For things
> like page refcounts, KVM needs to treat ZONE_DEVICE pages like normal
> pages, e.g. put pages grabbed via gup().  But for flows such as setting
> A/D bits or shifting refcounts for transparent huge pages, KVM needs to
> to avoid processing ZONE_DEVICE pages as the flows in question lack the
> underlying machinery for proper handling of ZONE_DEVICE pages.
> 
> This fixes a hang reported by Adam Borowski[*] in dev_pagemap_cleanup()
> when running a KVM guest backed with /dev/dax memory, as KVM straight up
> doesn't put any references to ZONE_DEVICE pages acquired by gup().
> 
> Note, Dan Williams proposed an alternative solution of doing put_page()
> on ZONE_DEVICE pages immediately after gup() in order to simplify the
> auditing needed to ensure is_zone_device_page() is called if and only if
> the backing device is pinned (via gup()).  But that approach would break
> kvm_vcpu_{un}map() as KVM requires the page to be pinned from map() 'til
> unmap() when accessing guest memory, unlike KVM's secondary MMU, which
> coordinates with mmu_notifier invalidations to avoid creating stale
> page references, i.e. doesn't rely on pages being pinned.
> 
> [*] http://lkml.kernel.org/r/[e-mail]
> 
> Reported-by: Adam Borowski <[e-mail]>
> Debugged-by: David Hildenbrand <[e-mail]>
> Cc: Dan Williams <[e-mail]>
> Cc: [e-mail]
> Signed-off-by: Sean Christopherson <[e-mail]>
> ---
> arch/x86/kvm/mmu.c       |  8 ++++----
> include/linux/kvm_host.h |  1 +
> virt/kvm/kvm_main.c      | 26 +++++++++++++++++++++++---
> 3 files changed, 28 insertions(+), 7 deletions(-)
> 

Thanks for taking care of this! Other KVM related code (PPC, vfio) also has a reserved check (see my series), I didn‘t have a look yet at the details, how reserved pages are treated. Will do so in the next weeks, after hollidays.

Acked-by: David Hildenbrand <[e-mail]>

> diff --git a/arch/x86/kvm/mmu.c b/arch/x86/kvm/mmu.c
> index 24c23c66b226..bf82b1f2e834 100644
> --- a/arch/x86/kvm/mmu.c
> +++ b/arch/x86/kvm/mmu.c
> @@ -3306,7 +3306,7 @@ static void transparent_hugepage_adjust(struct kvm_vcpu *vcpu,
>     * here.
>     */
>    if (!is_error_noslot_pfn(pfn) && !kvm_is_reserved_pfn(pfn) &&
> -        level == PT_PAGE_TABLE_LEVEL &&
> +        !kvm_is_zone_device_pfn(pfn) && level == PT_PAGE_TABLE_LEVEL &&
>        PageTransCompoundMap(pfn_to_page(pfn)) &&
>        !mmu_gfn_lpage_is_disallowed(vcpu, gfn, PT_DIRECTORY_LEVEL)) {
>        unsigned long mask;
> @@ -5914,9 +5914,9 @@ static bool kvm_mmu_zap_collapsible_spte(struct kvm *kvm,
>         * the guest, and the guest page table is using 4K page size
>         * mapping if the indirect sp has level = 1.
>         */
> -        if (sp->role.direct &&
> -            !kvm_is_reserved_pfn(pfn) &&
> -            PageTransCompoundMap(pfn_to_page(pfn))) {
> +        if (sp->role.direct && !kvm_is_reserved_pfn(pfn) &&
> +            !kvm_is_zone_device_pfn(pfn) &&
> +            PageTransCompoundMap(pfn_to_page(pfn))) {
>            pte_list_remove(rmap_head, sptep);
> 
>            if (kvm_available_flush_tlb_with_range())
> diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h
> index a817e446c9aa..4ad1cd7d2d4d 100644
> --- a/include/linux/kvm_host.h
> +++ b/include/linux/kvm_host.h
> @@ -966,6 +966,7 @@ int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
> void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
> 
> bool kvm_is_reserved_pfn(kvm_pfn_t pfn);
> +bool kvm_is_zone_device_pfn(kvm_pfn_t pfn);
> 
> struct kvm_irq_ack_notifier {
>    struct hlist_node link;
> diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
> index b8534c6b8cf6..bc9d10a0a334 100644
> --- a/virt/kvm/kvm_main.c
> +++ b/virt/kvm/kvm_main.c
> @@ -149,10 +149,30 @@ __weak int kvm_arch_mmu_notifier_invalidate_range(struct kvm *kvm,
>    return 0;
> }
> 
> +bool kvm_is_zone_device_pfn(kvm_pfn_t pfn)
> +{
> +    /*
> +     * The metadata used by is_zone_device_page() to determine whether or
> +     * not a page is ZONE_DEVICE is guaranteed to be valid if and only if
> +     * the device has been pinned, e.g. by get_user_pages().  WARN if the
> +     * page_count() is zero to help detect bad usage of this helper.
> +     */
> +    if (!pfn_valid(pfn) || WARN_ON_ONCE(!page_count(pfn_to_page(pfn))))
> +        return false;
> +
> +    return is_zone_device_page(pfn_to_page(pfn));
> +}
> +
> bool kvm_is_reserved_pfn(kvm_pfn_t pfn)
> {
> +    /*
> +     * ZONE_DEVICE pages currently set PG_reserved, but from a refcounting
> +     * perspective they are "normal" pages, albeit with slightly different
> +     * usage rules.
> +     */
>    if (pfn_valid(pfn))
> -        return PageReserved(pfn_to_page(pfn));
> +        return PageReserved(pfn_to_page(pfn)) &&
> +               !kvm_is_zone_device_pfn(pfn);
> 
>    return true;
> }
> @@ -1865,7 +1885,7 @@ EXPORT_SYMBOL_GPL(kvm_release_pfn_dirty);
> 
> void kvm_set_pfn_dirty(kvm_pfn_t pfn)
> {
> -    if (!kvm_is_reserved_pfn(pfn)) {
> +    if (!kvm_is_reserved_pfn(pfn) && !kvm_is_zone_device_pfn(pfn)) {
>        struct page *page = pfn_to_page(pfn);
> 
>        SetPageDirty(page);
> @@ -1875,7 +1895,7 @@ EXPORT_SYMBOL_GPL(kvm_set_pfn_dirty);
> 
> void kvm_set_pfn_accessed(kvm_pfn_t pfn)
> {
> -    if (!kvm_is_reserved_pfn(pfn))
> +    if (!kvm_is_reserved_pfn(pfn) && !kvm_is_zone_device_pfn(pfn))
>        mark_page_accessed(pfn_to_page(pfn));
> }
> EXPORT_SYMBOL_GPL(kvm_set_pfn_accessed);
> -- 
> 2.24.0
>

Patch

diff --git a/arch/x86/kvm/mmu.c b/arch/x86/kvm/mmu.c
index 24c23c66b226..bf82b1f2e834 100644
--- a/arch/x86/kvm/mmu.c
+++ b/arch/x86/kvm/mmu.c
@@ -3306,7 +3306,7 @@  static void transparent_hugepage_adjust(struct kvm_vcpu *vcpu,
 	 * here.
 	 */
 	if (!is_error_noslot_pfn(pfn) && !kvm_is_reserved_pfn(pfn) &&
-	    level == PT_PAGE_TABLE_LEVEL &&
+	    !kvm_is_zone_device_pfn(pfn) && level == PT_PAGE_TABLE_LEVEL &&
 	    PageTransCompoundMap(pfn_to_page(pfn)) &&
 	    !mmu_gfn_lpage_is_disallowed(vcpu, gfn, PT_DIRECTORY_LEVEL)) {
 		unsigned long mask;
@@ -5914,9 +5914,9 @@  static bool kvm_mmu_zap_collapsible_spte(struct kvm *kvm,
 		 * the guest, and the guest page table is using 4K page size
 		 * mapping if the indirect sp has level = 1.
 		 */
-		if (sp->role.direct &&
-			!kvm_is_reserved_pfn(pfn) &&
-			PageTransCompoundMap(pfn_to_page(pfn))) {
+		if (sp->role.direct && !kvm_is_reserved_pfn(pfn) &&
+		    !kvm_is_zone_device_pfn(pfn) &&
+		    PageTransCompoundMap(pfn_to_page(pfn))) {
 			pte_list_remove(rmap_head, sptep);
 
 			if (kvm_available_flush_tlb_with_range())
diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h
index a817e446c9aa..4ad1cd7d2d4d 100644
--- a/include/linux/kvm_host.h
+++ b/include/linux/kvm_host.h
@@ -966,6 +966,7 @@  int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
 
 bool kvm_is_reserved_pfn(kvm_pfn_t pfn);
+bool kvm_is_zone_device_pfn(kvm_pfn_t pfn);
 
 struct kvm_irq_ack_notifier {
 	struct hlist_node link;
diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
index b8534c6b8cf6..bc9d10a0a334 100644
--- a/virt/kvm/kvm_main.c
+++ b/virt/kvm/kvm_main.c
@@ -149,10 +149,30 @@  __weak int kvm_arch_mmu_notifier_invalidate_range(struct kvm *kvm,
 	return 0;
 }
 
+bool kvm_is_zone_device_pfn(kvm_pfn_t pfn)
+{
+	/*
+	 * The metadata used by is_zone_device_page() to determine whether or
+	 * not a page is ZONE_DEVICE is guaranteed to be valid if and only if
+	 * the device has been pinned, e.g. by get_user_pages().  WARN if the
+	 * page_count() is zero to help detect bad usage of this helper.
+	 */
+	if (!pfn_valid(pfn) || WARN_ON_ONCE(!page_count(pfn_to_page(pfn))))
+		return false;
+
+	return is_zone_device_page(pfn_to_page(pfn));
+}
+
 bool kvm_is_reserved_pfn(kvm_pfn_t pfn)
 {
+	/*
+	 * ZONE_DEVICE pages currently set PG_reserved, but from a refcounting
+	 * perspective they are "normal" pages, albeit with slightly different
+	 * usage rules.
+	 */
 	if (pfn_valid(pfn))
-		return PageReserved(pfn_to_page(pfn));
+		return PageReserved(pfn_to_page(pfn)) &&
+		       !kvm_is_zone_device_pfn(pfn);
 
 	return true;
 }
@@ -1865,7 +1885,7 @@  EXPORT_SYMBOL_GPL(kvm_release_pfn_dirty);
 
 void kvm_set_pfn_dirty(kvm_pfn_t pfn)
 {
-	if (!kvm_is_reserved_pfn(pfn)) {
+	if (!kvm_is_reserved_pfn(pfn) && !kvm_is_zone_device_pfn(pfn)) {
 		struct page *page = pfn_to_page(pfn);
 
 		SetPageDirty(page);
@@ -1875,7 +1895,7 @@  EXPORT_SYMBOL_GPL(kvm_set_pfn_dirty);
 
 void kvm_set_pfn_accessed(kvm_pfn_t pfn)
 {
-	if (!kvm_is_reserved_pfn(pfn))
+	if (!kvm_is_reserved_pfn(pfn) && !kvm_is_zone_device_pfn(pfn))
 		mark_page_accessed(pfn_to_page(pfn));
 }
 EXPORT_SYMBOL_GPL(kvm_set_pfn_accessed);