qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: Peter Xu <peterx@redhat.com>
To: Chuang Xu <xuchuangxclwt@bytedance.com>
Cc: qemu-devel@nongnu.org, mst@redhat.com, sgarzare@redhat.com,
	richard.henderson@linaro.org, pbonzini@redhat.com,
	david@kernel.org, philmd@linaro.org, farosas@suse.de,
	Jason Wang <jasowang@redhat.com>
Subject: Re: [RFC v1 1/2] vhost: eliminate duplicate dirty_bitmap sync when log shared by multiple devices
Date: Tue, 9 Dec 2025 15:47:34 -0500	[thread overview]
Message-ID: <aTiK5opcOOrsFem9@x1.local> (raw)
In-Reply-To: <20251208120952.37563-2-xuchuangxclwt@bytedance.com>

On Mon, Dec 08, 2025 at 08:09:51PM +0800, Chuang Xu wrote:
> From: xuchuangxclwt <xuchuangxclwt@bytedance.com>
> 
> Although logs can now be shared among multiple vhost devices,
> live migration still performs repeated vhost_log_sync for each
> vhost device during bitmap_sync, which increases the time required
> for bitmap_sync and makes it more difficult for dirty pages to converge.
> 
> Attempt to eliminate these duplicate sync.
> 
> Signed-off-by: Chuang Xu <xuchuangxclwt@bytedance.com>

It looks reasonable from migration POV, but I don't know the details.

Please remember to copy Jason (I added for this email) when repost.

Thanks,

> ---
>  hw/virtio/vhost.c         | 30 ++++++++++++++++++++++--------
>  include/hw/virtio/vhost.h |  1 +
>  2 files changed, 23 insertions(+), 8 deletions(-)
> 
> diff --git a/hw/virtio/vhost.c b/hw/virtio/vhost.c
> index 266a11514a..d397ca327f 100644
> --- a/hw/virtio/vhost.c
> +++ b/hw/virtio/vhost.c
> @@ -268,14 +268,6 @@ static int vhost_sync_dirty_bitmap(struct vhost_dev *dev,
>      return 0;
>  }
>  
> -static void vhost_log_sync(MemoryListener *listener,
> -                          MemoryRegionSection *section)
> -{
> -    struct vhost_dev *dev = container_of(listener, struct vhost_dev,
> -                                         memory_listener);
> -    vhost_sync_dirty_bitmap(dev, section, 0x0, ~0x0ULL);
> -}
> -
>  static void vhost_log_sync_range(struct vhost_dev *dev,
>                                   hwaddr first, hwaddr last)
>  {
> @@ -287,6 +279,27 @@ static void vhost_log_sync_range(struct vhost_dev *dev,
>      }
>  }
>  
> +static void vhost_log_sync(MemoryListener *listener,
> +                          MemoryRegionSection *section)
> +{
> +    struct vhost_dev *dev = container_of(listener, struct vhost_dev,
> +                                         memory_listener);
> +    struct vhost_log *log = dev->log;
> +
> +    if (log && log->refcnt > 1) {
> +        /*
> +         * When multiple devices use same log, we implement the logic of
> +         * vhost_log_sync just like what we do in vhost_log_put.
> +         */
> +        log->sync_cnt = (log->sync_cnt + 1) % log->refcnt;
> +        if (!log->sync_cnt) {
> +            vhost_log_sync_range(dev, 0, dev->log_size * VHOST_LOG_CHUNK - 1);
> +        }
> +    } else {
> +        vhost_sync_dirty_bitmap(dev, section, 0x0, ~0x0ULL);
> +    }
> +}
> +
>  static uint64_t vhost_get_log_size(struct vhost_dev *dev)
>  {
>      uint64_t log_size = 0;
> @@ -383,6 +396,7 @@ static struct vhost_log *vhost_log_get(VhostBackendType backend_type,
>          ++log->refcnt;
>      }
>  
> +    log->sync_cnt = 0;
>      return log;
>  }
>  
> diff --git a/include/hw/virtio/vhost.h b/include/hw/virtio/vhost.h
> index 08bbb4dfe9..43bf1c2150 100644
> --- a/include/hw/virtio/vhost.h
> +++ b/include/hw/virtio/vhost.h
> @@ -50,6 +50,7 @@ typedef unsigned long vhost_log_chunk_t;
>  struct vhost_log {
>      unsigned long long size;
>      int refcnt;
> +    int sync_cnt;
>      int fd;
>      vhost_log_chunk_t *log;
>  };
> -- 
> 2.20.1
> 

-- 
Peter Xu



  reply	other threads:[~2025-12-09 20:48 UTC|newest]

Thread overview: 10+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-12-08 12:09 [RFC v1 0/2] migration: reduce bitmap sync time and make dirty pages converge much more easily Chuang Xu
2025-12-08 12:09 ` [RFC v1 1/2] vhost: eliminate duplicate dirty_bitmap sync when log shared by multiple devices Chuang Xu
2025-12-09 20:47   ` Peter Xu [this message]
2025-12-10  6:51     ` Jason Wang
2025-12-10 13:52       ` Chuang Xu
2025-12-08 12:09 ` [RFC v1 2/2] migration: merge fragmented clear_dirty ioctls Chuang Xu
2025-12-09 21:10   ` Peter Xu
2025-12-10 14:18     ` Chuang Xu
2025-12-10 19:45       ` Peter Xu
2025-12-09 21:06 ` [RFC v1 0/2] migration: reduce bitmap sync time and make dirty pages converge much more easily Peter Xu

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=aTiK5opcOOrsFem9@x1.local \
    --to=peterx@redhat.com \
    --cc=david@kernel.org \
    --cc=farosas@suse.de \
    --cc=jasowang@redhat.com \
    --cc=mst@redhat.com \
    --cc=pbonzini@redhat.com \
    --cc=philmd@linaro.org \
    --cc=qemu-devel@nongnu.org \
    --cc=richard.henderson@linaro.org \
    --cc=sgarzare@redhat.com \
    --cc=xuchuangxclwt@bytedance.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).