From: Xiao Guangrong <xiaoguangrong@cn.fujitsu.com>
To: Avi Kivity <avi@redhat.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>,
LKML <linux-kernel@vger.kernel.org>,
KVM list <kvm@vger.kernel.org>
Subject: [PATCH 3/5] KVM: MMU: gather remote tlb flush which occurs during page zapped
Date: Sun, 30 May 2010 20:39:49 +0800 [thread overview]
Message-ID: <4C025C95.5000307@cn.fujitsu.com> (raw)
In-Reply-To: <4C025BDC.1020304@cn.fujitsu.com>
Using kvm_mmu_prepare_zap_page() and kvm_mmu_zap_page() instead of
kvm_mmu_zap_page() that can reduce remote tlb flush IPI
Have tested with xp/vista64, fedora12 32/64 guests, and it not broken
Signed-off-by: Xiao Guangrong <xiaoguangrong@cn.fujitsu.com>
---
arch/x86/kvm/mmu.c | 75 +++++++++++++++++++----------------------------
arch/x86/kvm/mmutrace.h | 2 +-
2 files changed, 31 insertions(+), 46 deletions(-)
diff --git a/arch/x86/kvm/mmu.c b/arch/x86/kvm/mmu.c
index 0c957bf..e2b1020 100644
--- a/arch/x86/kvm/mmu.c
+++ b/arch/x86/kvm/mmu.c
@@ -1199,7 +1199,8 @@ static void kvm_unlink_unsync_page(struct kvm *kvm, struct kvm_mmu_page *sp)
--kvm->stat.mmu_unsync;
}
-static int kvm_mmu_zap_page(struct kvm *kvm, struct kvm_mmu_page *sp);
+static int kvm_mmu_prepare_zap_page(struct kvm *kvm, struct kvm_mmu_page *sp);
+static void kvm_mmu_commit_zap_page(struct kvm *kvm);
#define for_each_gfn_sp(kvm, sp, gfn, pos, n) \
hlist_for_each_entry_safe(sp, pos, n, \
@@ -1221,7 +1222,7 @@ static int __kvm_sync_page(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
bool clear_unsync)
{
if (sp->role.cr4_pae != !!is_pae(vcpu)) {
- kvm_mmu_zap_page(vcpu->kvm, sp);
+ kvm_mmu_prepare_zap_page(vcpu->kvm, sp);
return 1;
}
@@ -1232,7 +1233,7 @@ static int __kvm_sync_page(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
}
if (vcpu->arch.mmu.sync_page(vcpu, sp)) {
- kvm_mmu_zap_page(vcpu->kvm, sp);
+ kvm_mmu_prepare_zap_page(vcpu->kvm, sp);
return 1;
}
@@ -1249,6 +1250,7 @@ static int kvm_sync_page_transient(struct kvm_vcpu *vcpu,
ret = __kvm_sync_page(vcpu, sp, false);
if (!ret)
mmu_convert_notrap(sp);
+ kvm_mmu_commit_zap_page(vcpu->kvm);
return ret;
}
@@ -1271,13 +1273,13 @@ static void kvm_sync_pages(struct kvm_vcpu *vcpu, gfn_t gfn)
WARN_ON(s->role.level != PT_PAGE_TABLE_LEVEL);
if ((s->role.cr4_pae != !!is_pae(vcpu)) ||
(vcpu->arch.mmu.sync_page(vcpu, s))) {
- kvm_mmu_zap_page(vcpu->kvm, s);
+ kvm_mmu_prepare_zap_page(vcpu->kvm, s);
continue;
}
kvm_unlink_unsync_page(vcpu->kvm, s);
flush = true;
}
-
+ kvm_mmu_commit_zap_page(vcpu->kvm);
if (flush)
kvm_mmu_flush_tlb(vcpu);
}
@@ -1363,6 +1365,7 @@ static void mmu_sync_children(struct kvm_vcpu *vcpu,
kvm_sync_page(vcpu, sp);
mmu_pages_clear_parents(&parents);
}
+ kvm_mmu_commit_zap_page(vcpu->kvm);
cond_resched_lock(&vcpu->kvm->mmu_lock);
kvm_mmu_pages_init(parent, &parents, &pages);
}
@@ -1551,7 +1554,7 @@ static int mmu_zap_unsync_children(struct kvm *kvm,
struct kvm_mmu_page *sp;
for_each_sp(pages, sp, parents, i) {
- kvm_mmu_zap_page(kvm, sp);
+ kvm_mmu_prepare_zap_page(kvm, sp);
mmu_pages_clear_parents(&parents);
zapped++;
}
@@ -1565,7 +1568,7 @@ static int kvm_mmu_prepare_zap_page(struct kvm *kvm, struct kvm_mmu_page *sp)
{
int ret;
- trace_kvm_mmu_zap_page(sp);
+ trace_kvm_mmu_prepare_zap_page(sp);
++kvm->stat.mmu_shadow_zapped;
ret = mmu_zap_unsync_children(kvm, sp);
kvm_mmu_page_unlink_children(kvm, sp);
@@ -1601,33 +1604,6 @@ static void kvm_mmu_commit_zap_page(struct kvm *kvm)
}
}
-static int kvm_mmu_zap_page(struct kvm *kvm, struct kvm_mmu_page *sp)
-{
- int ret;
-
- trace_kvm_mmu_zap_page(sp);
- ++kvm->stat.mmu_shadow_zapped;
- ret = mmu_zap_unsync_children(kvm, sp);
- kvm_mmu_page_unlink_children(kvm, sp);
- kvm_mmu_unlink_parents(kvm, sp);
- kvm_flush_remote_tlbs(kvm);
- if (!sp->role.invalid && !sp->role.direct)
- unaccount_shadowed(kvm, sp->gfn);
- if (sp->unsync)
- kvm_unlink_unsync_page(kvm, sp);
- if (!sp->root_count) {
- /* Count self */
- ret++;
- kvm_mmu_free_page(kvm, sp);
- } else {
- sp->role.invalid = 1;
- list_move(&sp->link, &kvm->arch.active_mmu_pages);
- kvm_reload_remote_mmus(kvm);
- }
- kvm_mmu_reset_last_pte_updated(kvm);
- return ret;
-}
-
/*
* Changing the number of mmu pages allocated to the vm
* Note: if kvm_nr_mmu_pages is too small, you will get dead lock
@@ -1652,8 +1628,9 @@ void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned int kvm_nr_mmu_pages)
page = container_of(kvm->arch.active_mmu_pages.prev,
struct kvm_mmu_page, link);
- used_pages -= kvm_mmu_zap_page(kvm, page);
+ used_pages -= kvm_mmu_prepare_zap_page(kvm, page);
}
+ kvm_mmu_commit_zap_page(kvm);
kvm_nr_mmu_pages = used_pages;
kvm->arch.n_free_mmu_pages = 0;
}
@@ -1677,9 +1654,10 @@ restart:
pgprintk("%s: gfn %lx role %x\n", __func__, gfn,
sp->role.word);
r = 1;
- if (kvm_mmu_zap_page(kvm, sp))
+ if (kvm_mmu_prepare_zap_page(kvm, sp))
goto restart;
}
+ kvm_mmu_commit_zap_page(kvm);
return r;
}
@@ -1692,9 +1670,10 @@ restart:
for_each_gfn_indirect_valid_sp(kvm, sp, gfn, node, nn) {
pgprintk("%s: zap %lx %x\n",
__func__, gfn, sp->role.word);
- if (kvm_mmu_zap_page(kvm, sp))
+ if (kvm_mmu_prepare_zap_page(kvm, sp))
goto restart;
}
+ kvm_mmu_commit_zap_page(kvm);
}
static void page_header_update_slot(struct kvm *kvm, void *pte, gfn_t gfn)
@@ -2121,8 +2100,10 @@ static void mmu_free_roots(struct kvm_vcpu *vcpu)
sp = page_header(root);
--sp->root_count;
- if (!sp->root_count && sp->role.invalid)
- kvm_mmu_zap_page(vcpu->kvm, sp);
+ if (!sp->root_count && sp->role.invalid) {
+ kvm_mmu_prepare_zap_page(vcpu->kvm, sp);
+ kvm_mmu_commit_zap_page(vcpu->kvm);
+ }
vcpu->arch.mmu.root_hpa = INVALID_PAGE;
spin_unlock(&vcpu->kvm->mmu_lock);
return;
@@ -2135,10 +2116,11 @@ static void mmu_free_roots(struct kvm_vcpu *vcpu)
sp = page_header(root);
--sp->root_count;
if (!sp->root_count && sp->role.invalid)
- kvm_mmu_zap_page(vcpu->kvm, sp);
+ kvm_mmu_prepare_zap_page(vcpu->kvm, sp);
}
vcpu->arch.mmu.pae_root[i] = INVALID_PAGE;
}
+ kvm_mmu_commit_zap_page(vcpu->kvm);
spin_unlock(&vcpu->kvm->mmu_lock);
vcpu->arch.mmu.root_hpa = INVALID_PAGE;
}
@@ -2792,7 +2774,7 @@ restart:
*/
pgprintk("misaligned: gpa %llx bytes %d role %x\n",
gpa, bytes, sp->role.word);
- if (kvm_mmu_zap_page(vcpu->kvm, sp))
+ if (kvm_mmu_prepare_zap_page(vcpu->kvm, sp))
goto restart;
++vcpu->kvm->stat.mmu_flooded;
continue;
@@ -2827,6 +2809,7 @@ restart:
++spte;
}
}
+ kvm_mmu_commit_zap_page(vcpu->kvm);
kvm_mmu_audit(vcpu, "post pte write");
spin_unlock(&vcpu->kvm->mmu_lock);
if (!is_error_pfn(vcpu->arch.update_pte.pfn)) {
@@ -2860,9 +2843,10 @@ void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu)
sp = container_of(vcpu->kvm->arch.active_mmu_pages.prev,
struct kvm_mmu_page, link);
- kvm_mmu_zap_page(vcpu->kvm, sp);
+ kvm_mmu_prepare_zap_page(vcpu->kvm, sp);
++vcpu->kvm->stat.mmu_recycled;
}
+ kvm_mmu_commit_zap_page(vcpu->kvm);
}
int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t cr2, u32 error_code)
@@ -3001,9 +2985,10 @@ void kvm_mmu_zap_all(struct kvm *kvm)
spin_lock(&kvm->mmu_lock);
restart:
list_for_each_entry_safe(sp, node, &kvm->arch.active_mmu_pages, link)
- if (kvm_mmu_zap_page(kvm, sp))
+ if (kvm_mmu_prepare_zap_page(kvm, sp))
goto restart;
+ kvm_mmu_commit_zap_page(kvm);
spin_unlock(&kvm->mmu_lock);
kvm_flush_remote_tlbs(kvm);
@@ -3015,7 +3000,7 @@ static int kvm_mmu_remove_some_alloc_mmu_pages(struct kvm *kvm)
page = container_of(kvm->arch.active_mmu_pages.prev,
struct kvm_mmu_page, link);
- return kvm_mmu_zap_page(kvm, page);
+ return kvm_mmu_prepare_zap_page(kvm, page);
}
static int mmu_shrink(int nr_to_scan, gfp_t gfp_mask)
@@ -3040,7 +3025,7 @@ static int mmu_shrink(int nr_to_scan, gfp_t gfp_mask)
kvm_freed = kvm;
}
nr_to_scan--;
-
+ kvm_mmu_commit_zap_page(kvm);
spin_unlock(&kvm->mmu_lock);
srcu_read_unlock(&kvm->srcu, idx);
}
diff --git a/arch/x86/kvm/mmutrace.h b/arch/x86/kvm/mmutrace.h
index 42f07b1..3aab0f0 100644
--- a/arch/x86/kvm/mmutrace.h
+++ b/arch/x86/kvm/mmutrace.h
@@ -190,7 +190,7 @@ DEFINE_EVENT(kvm_mmu_page_class, kvm_mmu_unsync_page,
TP_ARGS(sp)
);
-DEFINE_EVENT(kvm_mmu_page_class, kvm_mmu_zap_page,
+DEFINE_EVENT(kvm_mmu_page_class, kvm_mmu_prepare_zap_page,
TP_PROTO(struct kvm_mmu_page *sp),
TP_ARGS(sp)
--
1.6.1.2
next prev parent reply other threads:[~2010-05-30 12:43 UTC|newest]
Thread overview: 15+ messages / expand[flat|nested] mbox.gz Atom feed top
2010-05-30 12:36 [PATCH 1/5] KVM: MMU: introduce some macros to cleanup hlist traverseing Xiao Guangrong
2010-05-30 12:37 ` [PATCH 2/5] KVM: MMU: split the operations of kvm_mmu_zap_page() Xiao Guangrong
2010-05-30 13:16 ` Avi Kivity
2010-05-31 2:13 ` Xiao Guangrong
2010-05-31 11:05 ` Avi Kivity
2010-06-01 2:29 ` Xiao Guangrong
2010-06-01 8:08 ` Avi Kivity
2010-05-30 12:39 ` Xiao Guangrong [this message]
2010-05-30 12:40 ` [PATCH 4/5] KVM: MMU: traverse sp hlish safely Xiao Guangrong
2010-05-30 12:42 ` [PATCH 5/5] KVM: MMU: reduce remote tlb flush in kvm_mmu_pte_write() Xiao Guangrong
2010-05-30 13:12 ` [PATCH 1/5] KVM: MMU: introduce some macros to cleanup hlist traverseing Avi Kivity
2010-05-31 2:00 ` Xiao Guangrong
2010-05-31 11:14 ` Avi Kivity
2010-06-01 2:38 ` Xiao Guangrong
2010-06-01 8:09 ` Avi Kivity
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=4C025C95.5000307@cn.fujitsu.com \
--to=xiaoguangrong@cn.fujitsu.com \
--cc=avi@redhat.com \
--cc=kvm@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=mtosatti@redhat.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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.