Re: [Qemu-devel] [PATCH v10 3/6] virtio-balloon: VIRTIO_BALLOON_F_PAGE_CHUNKS

[Wang, Wei W]

On 05/06/2017 06:29 AM, Michael S. Tsirkin wrote:
> On Thu, May 04, 2017 at 04:50:12PM +0800, Wei Wang wrote:
> > Add a new feature, VIRTIO_BALLOON_F_PAGE_CHUNKS, which enables the
> > transfer of the ballooned (i.e. inflated/deflated) pages in chunks to
> > the host.
> >
> > The implementation of the previous virtio-balloon is not very
> > efficient, because the ballooned pages are transferred to the host one
> > by one. Here is the breakdown of the time in percentage spent on each
> > step of the balloon inflating process (inflating 7GB of an 8GB idle
> > guest).
> >
> > 1) allocating pages (6.5%)
> > 2) sending PFNs to host (68.3%)
> > 3) address translation (6.1%)
> > 4) madvise (19%)
> >
> > It takes about 4126ms for the inflating process to complete.
> > The above profiling shows that the bottlenecks are stage 2) and stage
> > 4).
> >
> > This patch optimizes step 2) by transferring pages to the host in
> > chunks. A chunk consists of guest physically continuous pages.
> > When the pages are packed into a chunk, they are converted into
> > balloon page size (4KB) pages. A chunk is offered to the host via a
> > base PFN (i.e. the start PFN of those physically continuous
> > pages) and the size (i.e. the total number of the 4KB balloon size
> > pages). A chunk is formatted as below:
> > --------------------------------------------------------
> > |                 Base (52 bit)        | Rsvd (12 bit) |
> > --------------------------------------------------------
> > --------------------------------------------------------
> > |                 Size (52 bit)        | Rsvd (12 bit) |
> > --------------------------------------------------------
> >
> > By doing so, step 4) can also be optimized by doing address
> > translation and madvise() in chunks rather than page by page.
> >
> > With this new feature, the above ballooning process takes ~590ms
> > resulting in an improvement of ~85%.
> >
> > TODO: optimize stage 1) by allocating/freeing a chunk of pages instead
> > of a single page each time.
> >
> > Signed-off-by: Wei Wang <address@hidden>
> > Signed-off-by: Liang Li <address@hidden>
> > Suggested-by: Michael S. Tsirkin <address@hidden>
> 
> 
> This is much cleaner, thanks. It might be even better to have wrappers that 
> put
> array and its size in a struct and manage that struct, but I won't require 
> this for
> submission.

OK, thanks. Would this be your suggestion:

struct virtio_balloon_page_struct { 
        unsigned int page_bmap_num;
        unsigned long *page_bmap[VIRTIO_BALLOON_PAGE_BMAP_MAX_NUM];
}

Best,
Wei