From: Ben Gardon <bgardon@google.com>
To: linux-kernel@vger.kernel.org, kvm@vger.kernel.org
Cc: Paolo Bonzini <pbonzini@redhat.com>, Peter Xu <peterx@redhat.com>,
Sean Christopherson <seanjc@google.com>,
Peter Shier <pshier@google.com>,
Yulei Zhang <yulei.kernel@gmail.com>,
Wanpeng Li <kernellwp@gmail.com>,
Xiao Guangrong <xiaoguangrong.eric@gmail.com>,
Kai Huang <kai.huang@intel.com>,
Keqian Zhu <zhukeqian1@huawei.com>,
David Hildenbrand <david@redhat.com>,
Ben Gardon <bgardon@google.com>
Subject: [PATCH v5 7/7] KVM: x86/mmu: Lazily allocate memslot rmaps
Date: Tue, 18 May 2021 10:34:14 -0700 [thread overview]
Message-ID: <20210518173414.450044-8-bgardon@google.com> (raw)
In-Reply-To: <20210518173414.450044-1-bgardon@google.com>
If the TDP MMU is in use, wait to allocate the rmaps until the shadow
MMU is actually used. (i.e. a nested VM is launched.) This saves memory
equal to 0.2% of guest memory in cases where the TDP MMU is used and
there are no nested guests involved.
Signed-off-by: Ben Gardon <bgardon@google.com>
---
arch/x86/include/asm/kvm_host.h | 2 ++
arch/x86/kvm/mmu.h | 7 ++++-
arch/x86/kvm/mmu/mmu.c | 14 +++++++---
arch/x86/kvm/mmu/tdp_mmu.c | 6 +++--
arch/x86/kvm/mmu/tdp_mmu.h | 4 +--
arch/x86/kvm/x86.c | 46 +++++++++++++++++++++++++++++++++
6 files changed, 71 insertions(+), 8 deletions(-)
diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
index fc75ed49bfee..7b65f82ade1c 100644
--- a/arch/x86/include/asm/kvm_host.h
+++ b/arch/x86/include/asm/kvm_host.h
@@ -1868,4 +1868,6 @@ static inline int kvm_cpu_get_apicid(int mps_cpu)
int kvm_cpu_dirty_log_size(void);
+int alloc_all_memslots_rmaps(struct kvm *kvm);
+
#endif /* _ASM_X86_KVM_HOST_H */
diff --git a/arch/x86/kvm/mmu.h b/arch/x86/kvm/mmu.h
index af09c47b1aa2..e987c9af82b6 100644
--- a/arch/x86/kvm/mmu.h
+++ b/arch/x86/kvm/mmu.h
@@ -234,7 +234,12 @@ void kvm_mmu_pre_destroy_vm(struct kvm *kvm);
static inline bool kvm_memslots_have_rmaps(struct kvm *kvm)
{
- return kvm->arch.memslots_have_rmaps;
+ /*
+ * Ensure that threads reading memslots_have_rmaps in various
+ * lock contexts see the value before trying to dereference
+ * the memslot rmap pointers.
+ */
+ return smp_load_acquire(&kvm->arch.memslots_have_rmaps);
}
#endif
diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c
index 1e0daabc83ca..2ac7bec515a1 100644
--- a/arch/x86/kvm/mmu/mmu.c
+++ b/arch/x86/kvm/mmu/mmu.c
@@ -3294,6 +3294,10 @@ static int mmu_alloc_shadow_roots(struct kvm_vcpu *vcpu)
}
}
+ r = alloc_all_memslots_rmaps(vcpu->kvm);
+ if (r)
+ return r;
+
write_lock(&vcpu->kvm->mmu_lock);
r = make_mmu_pages_available(vcpu);
if (r < 0)
@@ -5481,9 +5485,13 @@ void kvm_mmu_init_vm(struct kvm *kvm)
{
struct kvm_page_track_notifier_node *node = &kvm->arch.mmu_sp_tracker;
- kvm_mmu_init_tdp_mmu(kvm);
-
- kvm->arch.memslots_have_rmaps = true;
+ if (!kvm_mmu_init_tdp_mmu(kvm))
+ /*
+ * No smp_load/store wrappers needed here as we are in
+ * VM init and there cannot be any memslots / other threads
+ * accessing this struct kvm yet.
+ */
+ kvm->arch.memslots_have_rmaps = true;
node->track_write = kvm_mmu_pte_write;
node->track_flush_slot = kvm_mmu_invalidate_zap_pages_in_memslot;
diff --git a/arch/x86/kvm/mmu/tdp_mmu.c b/arch/x86/kvm/mmu/tdp_mmu.c
index 95eeb5ac6a8a..ea00c9502ba1 100644
--- a/arch/x86/kvm/mmu/tdp_mmu.c
+++ b/arch/x86/kvm/mmu/tdp_mmu.c
@@ -14,10 +14,10 @@ static bool __read_mostly tdp_mmu_enabled = false;
module_param_named(tdp_mmu, tdp_mmu_enabled, bool, 0644);
/* Initializes the TDP MMU for the VM, if enabled. */
-void kvm_mmu_init_tdp_mmu(struct kvm *kvm)
+bool kvm_mmu_init_tdp_mmu(struct kvm *kvm)
{
if (!tdp_enabled || !READ_ONCE(tdp_mmu_enabled))
- return;
+ return false;
/* This should not be changed for the lifetime of the VM. */
kvm->arch.tdp_mmu_enabled = true;
@@ -25,6 +25,8 @@ void kvm_mmu_init_tdp_mmu(struct kvm *kvm)
INIT_LIST_HEAD(&kvm->arch.tdp_mmu_roots);
spin_lock_init(&kvm->arch.tdp_mmu_pages_lock);
INIT_LIST_HEAD(&kvm->arch.tdp_mmu_pages);
+
+ return true;
}
static __always_inline void kvm_lockdep_assert_mmu_lock_held(struct kvm *kvm,
diff --git a/arch/x86/kvm/mmu/tdp_mmu.h b/arch/x86/kvm/mmu/tdp_mmu.h
index 5fdf63090451..b046ab5137a1 100644
--- a/arch/x86/kvm/mmu/tdp_mmu.h
+++ b/arch/x86/kvm/mmu/tdp_mmu.h
@@ -80,12 +80,12 @@ int kvm_tdp_mmu_get_walk(struct kvm_vcpu *vcpu, u64 addr, u64 *sptes,
int *root_level);
#ifdef CONFIG_X86_64
-void kvm_mmu_init_tdp_mmu(struct kvm *kvm);
+bool kvm_mmu_init_tdp_mmu(struct kvm *kvm);
void kvm_mmu_uninit_tdp_mmu(struct kvm *kvm);
static inline bool is_tdp_mmu_enabled(struct kvm *kvm) { return kvm->arch.tdp_mmu_enabled; }
static inline bool is_tdp_mmu_page(struct kvm_mmu_page *sp) { return sp->tdp_mmu_page; }
#else
-static inline void kvm_mmu_init_tdp_mmu(struct kvm *kvm) {}
+static inline bool kvm_mmu_init_tdp_mmu(struct kvm *kvm) { return false; }
static inline void kvm_mmu_uninit_tdp_mmu(struct kvm *kvm) {}
static inline bool is_tdp_mmu_enabled(struct kvm *kvm) { return false; }
static inline bool is_tdp_mmu_page(struct kvm_mmu_page *sp) { return false; }
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index 7cbaa92687f7..28dc8bdd0c8a 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -10931,6 +10931,8 @@ static int memslot_rmap_alloc(struct kvm_memory_slot *slot,
int lpages = gfn_to_index(slot->base_gfn + npages - 1,
slot->base_gfn, level) + 1;
+ WARN_ON(slot->arch.rmap[i]);
+
slot->arch.rmap[i] = kvcalloc(lpages, sz, GFP_KERNEL_ACCOUNT);
if (!slot->arch.rmap[i]) {
memslot_rmap_free(slot);
@@ -10941,6 +10943,50 @@ static int memslot_rmap_alloc(struct kvm_memory_slot *slot,
return 0;
}
+int alloc_all_memslots_rmaps(struct kvm *kvm)
+{
+ struct kvm_memslots *slots;
+ struct kvm_memory_slot *slot;
+ int r, i;
+
+ /*
+ * Check if memslots alreday have rmaps early before acquiring
+ * the slots_arch_lock below.
+ */
+ if (kvm_memslots_have_rmaps(kvm))
+ return 0;
+
+ mutex_lock(&kvm->slots_arch_lock);
+
+ /*
+ * Read memslots_have_rmaps again, under the slots arch lock,
+ * before allocating the rmaps
+ */
+ if (kvm_memslots_have_rmaps(kvm)) {
+ mutex_unlock(&kvm->slots_arch_lock);
+ return 0;
+ }
+
+ for (i = 0; i < KVM_ADDRESS_SPACE_NUM; i++) {
+ slots = __kvm_memslots(kvm, i);
+ kvm_for_each_memslot(slot, slots) {
+ r = memslot_rmap_alloc(slot, slot->npages);
+ if (r) {
+ mutex_unlock(&kvm->slots_arch_lock);
+ return r;
+ }
+ }
+ }
+
+ /*
+ * Ensure that memslots_have_rmaps becomes true strictly after
+ * all the rmap pointers are set.
+ */
+ smp_store_release(&kvm->arch.memslots_have_rmaps, true);
+ mutex_unlock(&kvm->slots_arch_lock);
+ return 0;
+}
+
static int kvm_alloc_memslot_metadata(struct kvm *kvm,
struct kvm_memory_slot *slot,
unsigned long npages)
--
2.31.1.751.gd2f1c929bd-goog
next prev parent reply other threads:[~2021-05-18 17:35 UTC|newest]
Thread overview: 9+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-05-18 17:34 [PATCH v5 0/7] Lazily allocate memslot rmaps Ben Gardon
2021-05-18 17:34 ` [PATCH v5 1/7] KVM: x86/mmu: Deduplicate rmap freeing Ben Gardon
2021-05-18 17:34 ` [PATCH v5 2/7] KVM: x86/mmu: Factor out allocating memslot rmap Ben Gardon
2021-05-18 17:34 ` [PATCH v5 3/7] KVM: mmu: Refactor memslot copy Ben Gardon
2021-05-18 17:34 ` [PATCH v5 4/7] KVM: mmu: Add slots_arch_lock for memslot arch fields Ben Gardon
2021-05-18 17:34 ` [PATCH v5 5/7] KVM: x86/mmu: Add a field to control memslot rmap allocation Ben Gardon
2021-05-18 17:34 ` [PATCH v5 6/7] KVM: x86/mmu: Skip rmap operations if rmaps not allocated Ben Gardon
2021-05-18 17:34 ` Ben Gardon [this message]
2021-05-24 12:43 ` [PATCH v5 0/7] Lazily allocate memslot rmaps Paolo Bonzini
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=20210518173414.450044-8-bgardon@google.com \
--to=bgardon@google.com \
--cc=david@redhat.com \
--cc=kai.huang@intel.com \
--cc=kernellwp@gmail.com \
--cc=kvm@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=pbonzini@redhat.com \
--cc=peterx@redhat.com \
--cc=pshier@google.com \
--cc=seanjc@google.com \
--cc=xiaoguangrong.eric@gmail.com \
--cc=yulei.kernel@gmail.com \
--cc=zhukeqian1@huawei.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).