qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: Paolo Bonzini <pbonzini@redhat.com>
To: Peter Lieven <pl@kamp.de>, qemu-devel@nongnu.org
Cc: kwolf@redhat.com, peter.maydell@linaro.org, mst@redhat.com,
	dgilbert@redhat.com, mreitz@redhat.com, kraxel@redhat.com
Subject: Re: [Qemu-devel] [PATCH 00/15] optimize Qemu RSS usage
Date: Tue, 28 Jun 2016 13:37:02 +0200	[thread overview]
Message-ID: <db2ee39f-a46b-0e91-15e1-cfd20fe82e60@redhat.com> (raw)
In-Reply-To: <1467104499-27517-1-git-send-email-pl@kamp.de>



On 28/06/2016 11:01, Peter Lieven wrote:
> I recently found that Qemu is using several hundred megabytes of RSS memory
> more than older versions such as Qemu 2.2.0. So I started tracing
> memory allocation and found 2 major reasons for this.
> 
> 1) We changed the qemu coroutine pool to have a per thread and a global release
>    pool. The choosen poolsize and the changed algorithm could lead to up to
>    192 free coroutines with just a single iothread. Each of the coroutines
>    in the pool each having 1MB of stack memory.

But the fix, as you correctly note, is to reduce the stack size.  It
would be nice to compile block-obj-y with -Wstack-usage=2048 too.

> 2) Between Qemu 2.2.0 and 2.3.0 RCU was introduced which lead to delayed freeing
>    of memory. This lead to higher heap allocations which could not effectively
>    be returned to kernel (most likely due to fragmentation).

I agree that some of the exec.c allocations need some care, but I would
prefer to use a custom free list or lazy allocation instead of mmap.

Changing allocations to use mmap also is not really useful if you do it
for objects that are never freed (as in patches 8-9-10-15 at least, and
probably 11 too which is one of the most contentious).

In other words, the effort tracking down the allocation is really,
really appreciated.  But the patches look like you only had a hammer at
hand, and everything looked like a nail. :)

Paolo

> The following series is what I came up with. Beside the coroutine patches I changed
> some allocations to forcibly use mmap. All these allocations are not repeatly made
> during runtime so the impact of using mmap should be neglectible.
> 
> There are still some big malloced allocations left which cannot be easily changed
> (e.g. the pixman buffers in VNC). So it might an idea to set a lower mmap threshold for
> malloc since this threshold seems to be in the order of several Megabytes on modern systems.
> 
> Peter Lieven (15):
>   coroutine-ucontext: mmap stack memory
>   coroutine-ucontext: add a switch to monitor maximum stack size
>   coroutine-ucontext: reduce stack size to 64kB
>   coroutine: add a knob to disable the shared release pool
>   util: add a helper to mmap private anonymous memory
>   exec: use mmap for subpages
>   qapi: use mmap for QmpInputVisitor
>   virtio: use mmap for VirtQueue
>   loader: use mmap for ROMs
>   vmware_svga: use mmap for scratch pad
>   qom: use mmap for bigger Objects
>   util: add a function to realloc mmapped memory
>   exec: use mmap for PhysPageMap->nodes
>   vnc-tight: make the encoding palette static
>   vnc: use mmap for VncState
> 
>  configure                 | 33 ++++++++++++++++++--
>  exec.c                    | 11 ++++---
>  hw/core/loader.c          | 16 +++++-----
>  hw/display/vmware_vga.c   |  3 +-
>  hw/virtio/virtio.c        |  5 +--
>  include/qemu/mmap-alloc.h |  7 +++++
>  include/qom/object.h      |  1 +
>  qapi/qmp-input-visitor.c  |  5 +--
>  qom/object.c              | 20 ++++++++++--
>  ui/vnc-enc-tight.c        | 21 ++++++-------
>  ui/vnc.c                  |  5 +--
>  ui/vnc.h                  |  1 +
>  util/coroutine-ucontext.c | 66 +++++++++++++++++++++++++++++++++++++--
>  util/mmap-alloc.c         | 27 ++++++++++++++++
>  util/qemu-coroutine.c     | 79 ++++++++++++++++++++++++++---------------------
>  15 files changed, 225 insertions(+), 75 deletions(-)
> 

  parent reply	other threads:[~2016-06-28 11:37 UTC|newest]

Thread overview: 78+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2016-06-28  9:01 [Qemu-devel] [PATCH 00/15] optimize Qemu RSS usage Peter Lieven
2016-06-28  9:01 ` [Qemu-devel] [PATCH 01/15] coroutine-ucontext: mmap stack memory Peter Lieven
2016-06-28 10:02   ` Peter Maydell
2016-06-28 10:21     ` Peter Lieven
2016-06-28 11:04   ` Paolo Bonzini
2016-06-28  9:01 ` [Qemu-devel] [PATCH 02/15] coroutine-ucontext: add a switch to monitor maximum stack size Peter Lieven
2016-06-28  9:01 ` [Qemu-devel] [PATCH 03/15] coroutine-ucontext: reduce stack size to 64kB Peter Lieven
2016-06-28 10:54   ` Paolo Bonzini
2016-06-28 10:57     ` Dr. David Alan Gilbert
2016-06-28 11:17       ` Peter Lieven
2016-06-28 11:35         ` Dr. David Alan Gilbert
2016-06-28 12:09           ` Peter Lieven
2016-06-28 14:20             ` Dr. David Alan Gilbert
2016-06-30  6:34               ` Peter Lieven
2016-06-28 11:13     ` Peter Lieven
2016-06-28 11:26       ` Paolo Bonzini
2016-06-28  9:01 ` [Qemu-devel] [PATCH 04/15] coroutine: add a knob to disable the shared release pool Peter Lieven
2016-06-28 10:41   ` Paolo Bonzini
2016-06-28 10:47     ` Peter Lieven
2016-06-28  9:01 ` [Qemu-devel] [PATCH 05/15] util: add a helper to mmap private anonymous memory Peter Lieven
2016-10-16  2:10   ` Michael S. Tsirkin
2016-10-18 13:50     ` Alex Bennée
2016-06-28  9:01 ` [Qemu-devel] [PATCH 06/15] exec: use mmap for subpages Peter Lieven
2016-06-28 10:48   ` Paolo Bonzini
2016-06-28  9:01 ` [Qemu-devel] [PATCH 07/15] qapi: use mmap for QmpInputVisitor Peter Lieven
2016-06-28  9:29   ` Dr. David Alan Gilbert
2016-06-28  9:39     ` Peter Lieven
2016-06-28 10:10       ` Daniel P. Berrange
2016-06-28 10:17         ` Dr. David Alan Gilbert
2016-06-28 10:21           ` Daniel P. Berrange
2016-06-28 14:10           ` Eric Blake
2016-06-28 11:36   ` Paolo Bonzini
2016-06-28 14:14     ` Eric Blake
2016-06-30 14:12   ` Markus Armbruster
2016-07-04  9:02     ` Paolo Bonzini
2016-07-04 11:18       ` Markus Armbruster
2016-07-04 11:36         ` Peter Lieven
2016-07-04 11:42         ` Paolo Bonzini
2016-06-28  9:01 ` [Qemu-devel] [PATCH 08/15] virtio: use mmap for VirtQueue Peter Lieven
2016-06-28  9:01 ` [Qemu-devel] [PATCH 09/15] loader: use mmap for ROMs Peter Lieven
2016-06-28 10:41   ` Paolo Bonzini
2016-06-28 11:26     ` Peter Lieven
2016-07-04  7:30     ` Peter Lieven
2016-06-28  9:01 ` [Qemu-devel] [PATCH 10/15] vmware_svga: use mmap for scratch pad Peter Lieven
2016-06-28  9:01 ` [Qemu-devel] [PATCH 11/15] qom: use mmap for bigger Objects Peter Lieven
2016-06-28 10:08   ` Daniel P. Berrange
2016-06-28 10:10   ` Peter Maydell
2016-06-28 10:19     ` Peter Lieven
2016-06-28 10:42   ` Paolo Bonzini
2016-06-28 10:49     ` Peter Lieven
2016-06-30 14:15       ` Markus Armbruster
2016-06-28  9:01 ` [Qemu-devel] [PATCH 12/15] util: add a function to realloc mmapped memory Peter Lieven
2016-06-28  9:01 ` [Qemu-devel] [PATCH 13/15] exec: use mmap for PhysPageMap->nodes Peter Lieven
2016-06-28 10:43   ` Paolo Bonzini
2016-06-28 10:48     ` Peter Lieven
2016-07-11  9:31     ` Peter Lieven
2016-07-11  9:44       ` Peter Lieven
2016-07-11 10:37       ` Paolo Bonzini
2016-07-12 14:34         ` Peter Lieven
2016-07-13 10:27           ` Paolo Bonzini
2016-07-14 14:47             ` Peter Lieven
2016-06-28  9:01 ` [Qemu-devel] [PATCH 14/15] vnc-tight: make the encoding palette static Peter Lieven
2016-06-28 11:12   ` Paolo Bonzini
2016-06-28 11:18     ` Peter Lieven
2016-06-28  9:01 ` [Qemu-devel] [PATCH 15/15] vnc: use mmap for VncState Peter Lieven
2016-06-28 11:37 ` Paolo Bonzini [this message]
2016-06-28 12:14   ` [Qemu-devel] [PATCH 00/15] optimize Qemu RSS usage Peter Lieven
2016-06-28 12:29     ` Paolo Bonzini
2016-06-28 12:33       ` Peter Lieven
2016-06-28 12:56         ` Paolo Bonzini
2016-06-28 12:56         ` Dr. David Alan Gilbert
2016-06-28 14:43           ` Peter Lieven
2016-06-28 14:52             ` Peter Lieven
2016-10-12 21:18 ` Michael R. Hines
2016-10-18 10:47   ` Peter Lieven
2016-10-19 17:40     ` Michael R. Hines
2016-10-31 22:00     ` Michael R. Hines
2016-11-01 22:02       ` Michael R. Hines

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=db2ee39f-a46b-0e91-15e1-cfd20fe82e60@redhat.com \
    --to=pbonzini@redhat.com \
    --cc=dgilbert@redhat.com \
    --cc=kraxel@redhat.com \
    --cc=kwolf@redhat.com \
    --cc=mreitz@redhat.com \
    --cc=mst@redhat.com \
    --cc=peter.maydell@linaro.org \
    --cc=pl@kamp.de \
    --cc=qemu-devel@nongnu.org \
    /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).