From: "Michael S. Tsirkin" <mst@redhat.com>
To: Paolo Bonzini <pbonzini@redhat.com>
Cc: Igor Mammedov <imammedo@redhat.com>,
linux-kernel@vger.kernel.org, kvm@vger.kernel.org
Subject: Re: [PATCH 3/5] vhost: support upto 509 memory regions
Date: Wed, 17 Jun 2015 21:11:10 +0200 [thread overview]
Message-ID: <20150617205949-mutt-send-email-mst@redhat.com> (raw)
In-Reply-To: <5581A496.5060503@redhat.com>
On Wed, Jun 17, 2015 at 06:47:18PM +0200, Paolo Bonzini wrote:
>
>
> On 17/06/2015 18:41, Michael S. Tsirkin wrote:
> > On Wed, Jun 17, 2015 at 06:38:25PM +0200, Paolo Bonzini wrote:
> >>
> >>
> >> On 17/06/2015 18:34, Michael S. Tsirkin wrote:
> >>> On Wed, Jun 17, 2015 at 06:31:32PM +0200, Paolo Bonzini wrote:
> >>>>
> >>>>
> >>>> On 17/06/2015 18:30, Michael S. Tsirkin wrote:
> >>>>> Meanwhile old tools are vulnerable to OOM attacks.
> >>>>
> >>>> For each vhost device there will be likely one tap interface, and I
> >>>> suspect that it takes way, way more than 16KB of memory.
> >>>
> >>> That's not true. We have a vhost device per queue, all queues
> >>> are part of a single tap device.
> >>
> >> s/tap/VCPU/ then. A KVM VCPU also takes more than 16KB of memory.
> >
> > That's up to you as a kvm maintainer :)
>
> Not easy, when the CPU alone requires three (albeit non-consecutive)
> pages for the VMCS, the APIC access page and the EPT root.
>
> > People are already concerned about vhost device
> > memory usage, I'm not happy to define our user/kernel interface
> > in a way that forces even more memory to be used up.
>
> So, the questions to ask are:
>
> 1) What is the memory usage like immediately after vhost is brought up,
> apart from these 16K?
About 24K, but most are iov pool arrays are kept around as an optimization
to avoid kmalloc on data path. Below 1K tracks persistent state.
Recently people have been complaining about these pools
so I've been thinking about switching to a per-cpu array,
or something similar.
> 2) Is there anything in vhost that allocates a user-controllable amount
> of memory?
Definitely not in vhost-net.
> 3) What is the size of the data structures that support one virtqueue
> (there are two of them)?
Around 256 bytes.
> Does it depend on the size of the virtqueues?
No.
> 4) Would it make sense to share memory regions between multiple vhost
> devices? Would it be hard to implement?
It's not trivial. It would absolutely require userspace ABI
extensions.
> It would also make memory
> operations O(1) rather than O(#cpus).
>
> Paolo
We'd save the kmalloc/memcpy/kfree, that is true.
But we'd still need to flush all VQs so it's still O(#cpus),
we'd just be doing less stuff in that O(#cpus).
--
MST
next prev parent reply other threads:[~2015-06-17 19:11 UTC|newest]
Thread overview: 80+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-06-16 16:33 [PATCH 0/5] vhost: support upto 509 memory regions Igor Mammedov
2015-06-16 16:33 ` [PATCH 1/5] vhost: use binary search instead of linear in find_region() Igor Mammedov
2015-06-16 21:07 ` Michael S. Tsirkin
2015-06-16 21:13 ` Igor Mammedov
2015-06-16 16:33 ` [PATCH 2/5] vhost: extend memory regions allocation to vmalloc Igor Mammedov
2015-06-16 16:33 ` [PATCH 3/5] vhost: support upto 509 memory regions Igor Mammedov
2015-06-16 21:14 ` Michael S. Tsirkin
2015-06-16 22:00 ` Igor Mammedov
2015-06-17 6:34 ` Michael S. Tsirkin
2015-06-17 7:28 ` Igor Mammedov
2015-06-17 7:39 ` Michael S. Tsirkin
2015-06-17 8:54 ` Igor Mammedov
2015-06-17 10:11 ` Michael S. Tsirkin
2015-06-17 10:37 ` Igor Mammedov
2015-06-17 10:46 ` Michael S. Tsirkin
2015-06-17 11:48 ` Igor Mammedov
2015-06-17 11:51 ` Michael S. Tsirkin
2015-06-17 12:23 ` Igor Mammedov
2015-06-17 13:13 ` Michael S. Tsirkin
2015-06-17 13:20 ` Paolo Bonzini
2015-06-17 14:32 ` Michael S. Tsirkin
2015-06-17 15:12 ` Igor Mammedov
2015-06-17 15:38 ` Michael S. Tsirkin
2015-06-17 16:09 ` Igor Mammedov
2015-06-17 16:30 ` Michael S. Tsirkin
2015-06-17 16:31 ` Paolo Bonzini
2015-06-17 16:34 ` Michael S. Tsirkin
2015-06-17 16:38 ` Paolo Bonzini
2015-06-17 16:41 ` Michael S. Tsirkin
2015-06-17 16:47 ` Paolo Bonzini
2015-06-17 17:32 ` Igor Mammedov
2015-06-17 19:11 ` Michael S. Tsirkin [this message]
2015-06-17 17:30 ` Igor Mammedov
2015-06-18 9:12 ` Igor Mammedov
2015-06-18 9:50 ` Michael S. Tsirkin
2015-06-18 10:03 ` Paolo Bonzini
2015-06-18 11:39 ` Igor Mammedov
2015-06-18 11:41 ` Michael S. Tsirkin
2015-06-18 11:50 ` Paolo Bonzini
2015-06-18 13:19 ` Michael S. Tsirkin
2015-06-18 13:46 ` Paolo Bonzini
2015-06-18 14:47 ` Michael S. Tsirkin
2015-06-18 15:54 ` Igor Mammedov
2015-06-18 16:02 ` Paolo Bonzini
2015-06-19 7:56 ` Michael S. Tsirkin
2015-06-19 7:57 ` Paolo Bonzini
2015-06-19 8:05 ` Michael S. Tsirkin
2015-06-19 8:52 ` Paolo Bonzini
2015-06-19 10:14 ` Michael S. Tsirkin
2015-06-19 10:44 ` Paolo Bonzini
2015-06-19 13:34 ` Michael S. Tsirkin
2015-06-19 13:34 ` Michael S. Tsirkin
2015-06-19 15:19 ` Paolo Bonzini
2015-06-19 15:19 ` Paolo Bonzini
2015-06-19 16:20 ` Michael S. Tsirkin
2015-06-19 16:20 ` Michael S. Tsirkin
2015-06-19 16:26 ` Paolo Bonzini
2015-06-19 16:26 ` Paolo Bonzini
2015-06-19 16:33 ` Michael S. Tsirkin
2015-06-19 16:33 ` Michael S. Tsirkin
2015-06-19 16:44 ` Paolo Bonzini
2015-06-19 16:44 ` Paolo Bonzini
2015-06-22 7:10 ` Igor Mammedov
2015-06-22 7:10 ` Igor Mammedov
2015-06-22 9:45 ` Paolo Bonzini
2015-06-22 9:45 ` Paolo Bonzini
2015-06-19 16:45 ` Michael S. Tsirkin
2015-06-19 16:45 ` Michael S. Tsirkin
2015-06-19 16:50 ` Paolo Bonzini
2015-06-19 16:50 ` Paolo Bonzini
2015-06-18 12:02 ` Igor Mammedov
2015-06-17 8:53 ` Paolo Bonzini
2015-06-16 16:33 ` [PATCH 4/5] vhost: add per VQ memory region caching Igor Mammedov
2015-06-16 16:33 ` [PATCH 5/5] vhost: translate_desc: optimization for desc.len < region size Igor Mammedov
2015-06-16 21:11 ` Michael S. Tsirkin
2015-06-16 21:16 ` [PATCH 0/5] vhost: support upto 509 memory regions Michael S. Tsirkin
2015-06-16 22:19 ` Igor Mammedov
2015-06-17 6:31 ` Michael S. Tsirkin
2015-06-17 7:33 ` Igor Mammedov
2015-06-17 7:40 ` Michael S. Tsirkin
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=20150617205949-mutt-send-email-mst@redhat.com \
--to=mst@redhat.com \
--cc=imammedo@redhat.com \
--cc=kvm@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=pbonzini@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.