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 <kvm@vger.kernel.org>
Subject: Re: [PATCH v2 19/22] KVM: MMU: lockless walking shadow page table
Date: Wed, 29 Jun 2011 19:50:41 +0800 [thread overview]
Message-ID: <4E0B1191.6040100@cn.fujitsu.com> (raw)
In-Reply-To: <4E0B09EA.1040304@redhat.com>
On 06/29/2011 07:18 PM, Avi Kivity wrote:
> On 06/29/2011 02:16 PM, Xiao Guangrong wrote:
>> >> @@ -1767,6 +1874,14 @@ static void kvm_mmu_commit_zap_page(struct kvm *kvm,
>> >>
>> >> kvm_flush_remote_tlbs(kvm);
>> >>
>> >> + if (atomic_read(&kvm->arch.reader_counter)) {
>> >> + kvm_mmu_isolate_pages(invalid_list);
>> >> + sp = list_first_entry(invalid_list, struct kvm_mmu_page, link);
>> >> + list_del_init(invalid_list);
>> >> + call_rcu(&sp->rcu, free_pages_rcu);
>> >> + return;
>> >> + }
>> >> +
>> >
>> > I think we should do this unconditionally. The cost of ping-ponging the shared cache line containing reader_counter will increase with large smp counts. On the other hand, zap_page is very rare, so it can be a little slower. Also, less code paths = easier to understand.
>> >
>>
>> On soft mmu, zap_page is very frequently, it can cause performance regression in my test.
>
> Any idea what the cause of the regression is? It seems to me that simply deferring freeing shouldn't have a large impact.
>
I guess it is because the page is freed too frequently, i have done the test, it shows
about 3219 pages is freed per second
Kernbench performance comparing:
the origin way: 3m27.723
free all shadow page in rcu context: 3m30.519
next prev parent reply other threads:[~2011-06-29 11:48 UTC|newest]
Thread overview: 62+ messages / expand[flat|nested] mbox.gz Atom feed top
2011-06-22 14:27 [PATCH v2 0/22] KVM: optimize for MMIO handled Xiao Guangrong
2011-06-22 14:28 ` [PATCH v2 01/22] KVM: MMU: fix walking shadow page table Xiao Guangrong
2011-06-22 17:13 ` Marcelo Tosatti
2011-06-23 2:05 ` Xiao Guangrong
2011-06-27 6:35 ` Xiao Guangrong
2011-06-22 14:28 ` [PATCH v2 02/22] KVM: MMU: do not update slot bitmap if spte is nonpresent Xiao Guangrong
2011-06-22 14:29 ` [PATCH v2 03/22] KVM: x86: fix broken read emulation spans a page boundary Xiao Guangrong
2011-06-29 8:21 ` Avi Kivity
2011-06-29 10:53 ` Xiao Guangrong
2011-06-29 11:19 ` Avi Kivity
2011-06-22 14:29 ` [PATCH v2 04/22] KVM: x86: introduce vcpu_gva_to_gpa to cleanup the code Xiao Guangrong
2011-06-29 8:24 ` Avi Kivity
2011-06-29 10:56 ` Xiao Guangrong
2011-06-29 11:09 ` Avi Kivity
2011-06-29 11:26 ` Xiao Guangrong
2011-06-29 11:26 ` Avi Kivity
2011-06-29 11:48 ` Gleb Natapov
2011-06-22 14:30 ` [PATCH v2 05/22] KVM: x86: abstract the operation for read/write emulation Xiao Guangrong
2011-06-29 8:37 ` Avi Kivity
2011-06-29 10:59 ` Xiao Guangrong
2011-06-22 14:30 ` [PATCH v2 06/22] KVM: x86: cleanup the code of " Xiao Guangrong
2011-06-22 14:31 ` [PATCH v2 07/22] KVM: MMU: cache mmio info on page fault path Xiao Guangrong
2011-06-29 8:48 ` Avi Kivity
2011-06-29 11:09 ` Xiao Guangrong
2011-06-29 11:10 ` Avi Kivity
2011-06-22 14:31 ` [PATCH v2 08/22] KVM: MMU: optimize to handle dirty bit Xiao Guangrong
2011-06-22 14:31 ` [PATCH v2 09/22] KVM: MMU: cleanup for FNAME(fetch) Xiao Guangrong
2011-06-22 14:32 ` [PATCH v2 10/22] KVM: MMU: rename 'pt_write' to 'emulate' Xiao Guangrong
2011-06-22 14:32 ` [PATCH v2 11/22] KVM: MMU: count used shadow pages on prepareing path Xiao Guangrong
2011-06-22 14:32 ` [PATCH v2 12/22] KVM: MMU: split kvm_mmu_free_page Xiao Guangrong
2011-06-22 14:33 ` [PATCH v2 13/22] KVM: MMU: remove bypass_guest_pf Xiao Guangrong
2011-06-22 14:33 ` [PATCH v2 14/22] KVM: MMU: filter out the mmio pfn from the fault pfn Xiao Guangrong
2011-06-22 14:34 ` [PATCH v2 15/22] KVM: MMU: abstract some functions to handle " Xiao Guangrong
2011-06-22 14:34 ` [PATCH v2 16/22] KVM: MMU: introduce the rules to modify shadow page table Xiao Guangrong
2011-06-22 14:34 ` [PATCH v2 17/22] KVM: MMU: clean up spte updating and clearing Xiao Guangrong
2011-06-22 14:35 ` [PATCH 18/22] KVM: MMU: do not need atomicly to set/clear spte Xiao Guangrong
2011-06-22 14:35 ` [PATCH v2 19/22] KVM: MMU: lockless walking shadow page table Xiao Guangrong
2011-06-29 9:16 ` Avi Kivity
2011-06-29 11:16 ` Xiao Guangrong
2011-06-29 11:18 ` Avi Kivity
2011-06-29 11:50 ` Xiao Guangrong [this message]
2011-06-29 12:18 ` Avi Kivity
2011-06-29 12:28 ` Xiao Guangrong
2011-06-29 12:27 ` Avi Kivity
2011-06-29 12:39 ` Xiao Guangrong
2011-06-29 13:01 ` Avi Kivity
2011-06-29 13:05 ` Xiao Guangrong
2011-06-22 14:35 ` [PATCH v2 20/22] KVM: MMU: reorganize struct kvm_shadow_walk_iterator Xiao Guangrong
2011-06-22 14:36 ` [PATCH v2 21/22] KVM: MMU: mmio page fault support Xiao Guangrong
2011-06-22 21:59 ` Marcelo Tosatti
2011-06-23 3:19 ` Xiao Guangrong
2011-06-23 6:40 ` Xiao Guangrong
2011-06-23 14:21 ` Marcelo Tosatti
2011-06-23 17:55 ` Xiao Guangrong
2011-06-23 20:13 ` Marcelo Tosatti
2011-06-24 2:04 ` Xiao Guangrong
2011-06-26 8:42 ` Avi Kivity
2011-06-27 11:00 ` [PATCH v3 " Xiao Guangrong
2011-06-29 9:22 ` [PATCH v2 " Avi Kivity
2011-06-29 12:28 ` Xiao Guangrong
2011-06-22 14:36 ` [PATCH v2 22/22] KVM: MMU: trace mmio page fault Xiao Guangrong
2011-06-29 9:23 ` [PATCH v2 0/22] KVM: optimize for MMIO handled 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=4E0B1191.6040100@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 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).