From: Mingwei Zhang <mizhang@google.com>
To: David Matlack <dmatlack@google.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>,
Sean Christopherson <seanjc@google.com>,
Vitaly Kuznetsov <vkuznets@redhat.com>,
Wanpeng Li <wanpengli@tencent.com>,
Jim Mattson <jmattson@google.com>, Joerg Roedel <joro@8bytes.org>,
kvm@vger.kernel.org, linux-kernel@vger.kernel.org,
Ben Gardon <bgardon@google.com>,
Jing Zhang <jingzhangos@google.com>
Subject: Re: [PATCH v2 3/3] KVM: x86/mmu: Add detailed page size stats
Date: Wed, 28 Jul 2021 23:24:16 -0700 [thread overview]
Message-ID: <CAL715WJPB6NK2zY1MB+53bTH8K+_HS7r1vpcJTFioa1mqr4_qw@mail.gmail.com> (raw)
In-Reply-To: <YQAoBXm80KRjb1nY@google.com>
Hi David and Ben,
Thank you both for the suggestions. It Makes sense to me and I will
update the patch set with the change added in the next version.
Regards.
-Mingwei
On Tue, Jul 27, 2021 at 8:36 AM David Matlack <dmatlack@google.com> wrote:
>
> On Mon, Jul 26, 2021 at 10:53:57AM -0700, Mingwei Zhang wrote:
> > Existing KVM code tracks the number of large pages regardless of their
> > sizes. Therefore, when large page of 1GB (or larger) is adopted, the
> > information becomes less useful because lpages counts a mix of 1G and 2M
> > pages.
> >
> > So remove the lpages since it is easy for user space to aggregate the info.
> > Instead, provide a comprehensive page stats of all sizes from 4K to 512G.
> >
> > Suggested-by: Ben Gardon <bgardon@google.com>
> > Suggested-by: Jing Zhang <jingzhangos@google.com>
> > Signed-off-by: Mingwei Zhang <mizhang@google.com>
> > ---
> > arch/x86/include/asm/kvm_host.h | 10 +++++++++-
> > arch/x86/kvm/mmu.h | 2 ++
> > arch/x86/kvm/mmu/mmu.c | 32 +++++++++++++++++++-------------
> > arch/x86/kvm/mmu/tdp_mmu.c | 15 ++-------------
> > arch/x86/kvm/x86.c | 7 +++++--
> > 5 files changed, 37 insertions(+), 29 deletions(-)
> >
> > diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
> > index 974cbfb1eefe..2e4b6fd36e62 100644
> > --- a/arch/x86/include/asm/kvm_host.h
> > +++ b/arch/x86/include/asm/kvm_host.h
> > @@ -1206,9 +1206,17 @@ struct kvm_vm_stat {
> > u64 mmu_recycled;
> > u64 mmu_cache_miss;
> > u64 mmu_unsync;
> > - u64 lpages;
> > u64 nx_lpage_splits;
> > u64 max_mmu_page_hash_collisions;
> > + union {
> > + struct {
> > + atomic64_t pages_4k;
> > + atomic64_t pages_2m;
> > + atomic64_t pages_1g;
> > + atomic64_t pages_512g;
> > + };
> > + atomic64_t pages[4];
> > + } page_stats;
> > };
> >
> > struct kvm_vcpu_stat {
> > diff --git a/arch/x86/kvm/mmu.h b/arch/x86/kvm/mmu.h
> > index 83e6c6965f1e..ad5638815311 100644
> > --- a/arch/x86/kvm/mmu.h
> > +++ b/arch/x86/kvm/mmu.h
> > @@ -240,4 +240,6 @@ static inline bool kvm_memslots_have_rmaps(struct kvm *kvm)
> > return smp_load_acquire(&kvm->arch.memslots_have_rmaps);
> > }
> >
> > +void kvm_update_page_stats(struct kvm *kvm, int level, int count);
> > +
> > #endif
> > diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c
> > index 442cc554ebd6..7e0fc760739b 100644
> > --- a/arch/x86/kvm/mmu/mmu.c
> > +++ b/arch/x86/kvm/mmu/mmu.c
> > @@ -588,16 +588,22 @@ static bool mmu_spte_update(u64 *sptep, u64 new_spte)
> > return flush;
> > }
> >
> > +void kvm_update_page_stats(struct kvm *kvm, int level, int count)
> > +{
> > + atomic64_add(count, &kvm->stat.page_stats.pages[level - 1]);
> > +}
> > +
> > /*
> > * Rules for using mmu_spte_clear_track_bits:
> > * It sets the sptep from present to nonpresent, and track the
> > * state bits, it is used to clear the last level sptep.
> > * Returns non-zero if the PTE was previously valid.
> > */
> > -static int mmu_spte_clear_track_bits(u64 *sptep)
> > +static int mmu_spte_clear_track_bits(struct kvm *kvm, u64 *sptep)
> > {
> > kvm_pfn_t pfn;
> > u64 old_spte = *sptep;
> > + int level = sptep_to_sp(sptep)->role.level;
> >
> > if (!spte_has_volatile_bits(old_spte))
> > __update_clear_spte_fast(sptep, 0ull);
> > @@ -607,6 +613,9 @@ static int mmu_spte_clear_track_bits(u64 *sptep)
> > if (!is_shadow_present_pte(old_spte))
> > return 0;
> >
> > + if (is_last_spte(old_spte, level))
> > + kvm_update_page_stats(kvm, level, -1);
> > +
> > pfn = spte_to_pfn(old_spte);
> >
> > /*
> > @@ -984,9 +993,10 @@ static void __pte_list_remove(u64 *spte, struct kvm_rmap_head *rmap_head)
> > }
> > }
> >
> > -static void pte_list_remove(struct kvm_rmap_head *rmap_head, u64 *sptep)
> > +static void pte_list_remove(struct kvm *kvm, struct kvm_rmap_head *rmap_head,
> > + u64 *sptep)
> > {
> > - mmu_spte_clear_track_bits(sptep);
> > + mmu_spte_clear_track_bits(kvm, sptep);
> > __pte_list_remove(sptep, rmap_head);
> > }
> >
> > @@ -1119,7 +1129,7 @@ static u64 *rmap_get_next(struct rmap_iterator *iter)
> >
> > static void drop_spte(struct kvm *kvm, u64 *sptep)
> > {
> > - if (mmu_spte_clear_track_bits(sptep))
> > + if (mmu_spte_clear_track_bits(kvm, sptep))
> > rmap_remove(kvm, sptep);
> > }
> >
> > @@ -1129,7 +1139,6 @@ static bool __drop_large_spte(struct kvm *kvm, u64 *sptep)
> > if (is_large_pte(*sptep)) {
> > WARN_ON(sptep_to_sp(sptep)->role.level == PG_LEVEL_4K);
> > drop_spte(kvm, sptep);
> > - --kvm->stat.lpages;
> > return true;
> > }
> >
> > @@ -1386,7 +1395,7 @@ static bool kvm_zap_rmapp(struct kvm *kvm, struct kvm_rmap_head *rmap_head,
> > while ((sptep = rmap_get_first(rmap_head, &iter))) {
> > rmap_printk("spte %p %llx.\n", sptep, *sptep);
> >
> > - pte_list_remove(rmap_head, sptep);
> > + pte_list_remove(kvm, rmap_head, sptep);
> > flush = true;
> > }
> >
> > @@ -1421,13 +1430,13 @@ static bool kvm_set_pte_rmapp(struct kvm *kvm, struct kvm_rmap_head *rmap_head,
> > need_flush = 1;
> >
> > if (pte_write(pte)) {
> > - pte_list_remove(rmap_head, sptep);
> > + pte_list_remove(kvm, rmap_head, sptep);
> > goto restart;
> > } else {
> > new_spte = kvm_mmu_changed_pte_notifier_make_spte(
> > *sptep, new_pfn);
> >
> > - mmu_spte_clear_track_bits(sptep);
> > + mmu_spte_clear_track_bits(kvm, sptep);
> > mmu_spte_set(sptep, new_spte);
> > }
> > }
> > @@ -2232,8 +2241,6 @@ static int mmu_page_zap_pte(struct kvm *kvm, struct kvm_mmu_page *sp,
> > if (is_shadow_present_pte(pte)) {
> > if (is_last_spte(pte, sp->role.level)) {
> > drop_spte(kvm, spte);
> > - if (is_large_pte(pte))
> > - --kvm->stat.lpages;
> > } else {
> > child = to_shadow_page(pte & PT64_BASE_ADDR_MASK);
> > drop_parent_pte(child, spte);
> > @@ -2692,8 +2699,7 @@ static int mmu_set_spte(struct kvm_vcpu *vcpu, u64 *sptep,
> > trace_kvm_mmu_set_spte(level, gfn, sptep);
> >
> > if (!was_rmapped) {
> > - if (is_large_pte(*sptep))
> > - ++vcpu->kvm->stat.lpages;
> > + kvm_update_page_stats(vcpu->kvm, level, 1);
> > rmap_count = rmap_add(vcpu, sptep, gfn);
> > if (rmap_count > RMAP_RECYCLE_THRESHOLD)
> > rmap_recycle(vcpu, sptep, gfn);
> > @@ -5669,7 +5675,7 @@ static bool kvm_mmu_zap_collapsible_spte(struct kvm *kvm,
> > if (sp->role.direct && !kvm_is_reserved_pfn(pfn) &&
> > sp->role.level < kvm_mmu_max_mapping_level(kvm, slot, sp->gfn,
> > pfn, PG_LEVEL_NUM)) {
> > - pte_list_remove(rmap_head, sptep);
> > + pte_list_remove(kvm, rmap_head, sptep);
> >
> > if (kvm_available_flush_tlb_with_range())
> > kvm_flush_remote_tlbs_with_address(kvm, sp->gfn,
> > diff --git a/arch/x86/kvm/mmu/tdp_mmu.c b/arch/x86/kvm/mmu/tdp_mmu.c
> > index cba2ab5db2a0..eae404c15364 100644
> > --- a/arch/x86/kvm/mmu/tdp_mmu.c
> > +++ b/arch/x86/kvm/mmu/tdp_mmu.c
> > @@ -413,7 +413,6 @@ static void __handle_changed_spte(struct kvm *kvm, int as_id, gfn_t gfn,
> > bool was_leaf = was_present && is_last_spte(old_spte, level);
> > bool is_leaf = is_present && is_last_spte(new_spte, level);
> > bool pfn_changed = spte_to_pfn(old_spte) != spte_to_pfn(new_spte);
> > - bool was_large, is_large;
> >
> > WARN_ON(level > PT64_ROOT_MAX_LEVEL);
> > WARN_ON(level < PG_LEVEL_4K);
> > @@ -472,18 +471,8 @@ static void __handle_changed_spte(struct kvm *kvm, int as_id, gfn_t gfn,
> > return;
> > }
> >
> > - /*
> > - * Update large page stats if a large page is being zapped, created, or
> > - * is replacing an existing shadow page.
> > - */
> > - was_large = was_leaf && is_large_pte(old_spte);
> > - is_large = is_leaf && is_large_pte(new_spte);
> > - if (was_large != is_large) {
> > - if (was_large)
> > - atomic64_sub(1, (atomic64_t *)&kvm->stat.lpages);
> > - else
> > - atomic64_add(1, (atomic64_t *)&kvm->stat.lpages);
> > - }
> > + if (is_leaf != was_leaf)
> > + kvm_update_page_stats(kvm, level, is_leaf ? 1 : -1);
> >
> > if (was_leaf && is_dirty_spte(old_spte) &&
> > (!is_present || !is_dirty_spte(new_spte) || pfn_changed))
> > diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> > index 8166ad113fb2..3858d36d3c49 100644
> > --- a/arch/x86/kvm/x86.c
> > +++ b/arch/x86/kvm/x86.c
> > @@ -235,9 +235,12 @@ const struct _kvm_stats_desc kvm_vm_stats_desc[] = {
> > STATS_DESC_COUNTER(VM, mmu_recycled),
> > STATS_DESC_COUNTER(VM, mmu_cache_miss),
> > STATS_DESC_ICOUNTER(VM, mmu_unsync),
> > - STATS_DESC_ICOUNTER(VM, lpages),
> > STATS_DESC_ICOUNTER(VM, nx_lpage_splits),
> > - STATS_DESC_PCOUNTER(VM, max_mmu_page_hash_collisions)
> > + STATS_DESC_PCOUNTER(VM, max_mmu_page_hash_collisions),
> > + STATS_DESC_ICOUNTER(VM, page_stats.pages_4k),
> > + STATS_DESC_ICOUNTER(VM, page_stats.pages_2m),
> > + STATS_DESC_ICOUNTER(VM, page_stats.pages_1g),
> > + STATS_DESC_ICOUNTER(VM, page_stats.pages_512g)
>
> FYI this will make the stat names "page_stats.pages_4k",
> "page_stats.pages_2m", etc. Is that ok?
>
> If you want the stat names to be just "pages_4k", "pages_2m", etc. you
> can make the page_stats union anonymous.
>
> > };
> > static_assert(ARRAY_SIZE(kvm_vm_stats_desc) ==
> > sizeof(struct kvm_vm_stat) / sizeof(u64));
> > --
> > 2.32.0.432.gabb21c7263-goog
> >
prev parent reply other threads:[~2021-07-29 6:24 UTC|newest]
Thread overview: 15+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-07-26 17:53 [PATCH v2 0/3] Add detailed page size stats in KVM stats Mingwei Zhang
2021-07-26 17:53 ` [PATCH v2 1/3] KVM: x86/mmu: Remove redundant spte present check in mmu_set_spte Mingwei Zhang
2021-07-26 20:23 ` Ben Gardon
2021-07-29 18:16 ` Sean Christopherson
2021-07-26 17:53 ` [PATCH v2 2/3] KVM: x86/mmu: Avoid collision with !PRESENT SPTEs in TDP MMU lpage stats Mingwei Zhang
2021-07-26 21:02 ` Ben Gardon
2021-07-29 18:34 ` Sean Christopherson
2021-07-29 19:01 ` Mingwei Zhang
2021-07-26 17:53 ` [PATCH v2 3/3] KVM: x86/mmu: Add detailed page size stats Mingwei Zhang
2021-07-26 20:41 ` Ben Gardon
2021-07-26 21:06 ` Ben Gardon
2021-07-29 18:45 ` Sean Christopherson
2021-07-29 19:02 ` Mingwei Zhang
2021-07-27 15:36 ` David Matlack
2021-07-29 6:24 ` Mingwei Zhang [this message]
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=CAL715WJPB6NK2zY1MB+53bTH8K+_HS7r1vpcJTFioa1mqr4_qw@mail.gmail.com \
--to=mizhang@google.com \
--cc=bgardon@google.com \
--cc=dmatlack@google.com \
--cc=jingzhangos@google.com \
--cc=jmattson@google.com \
--cc=joro@8bytes.org \
--cc=kvm@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=pbonzini@redhat.com \
--cc=seanjc@google.com \
--cc=vkuznets@redhat.com \
--cc=wanpengli@tencent.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).