Re: [virtio-dev] Re: [PATCH v7 kernel 3/5] virtio-balloon: implementation of VIRTIO_BALLOON_F_CHUNK_TRANSFER

From: Wei Wang
Date: Fri Mar 10 2017 - 05:01:27 EST


On 03/08/2017 12:01 PM, Michael S. Tsirkin wrote:
On Fri, Mar 03, 2017 at 01:40:28PM +0800, Wei Wang wrote:
From: Liang Li <liang.z.li@xxxxxxxxx>

The implementation of the current virtio-balloon is not very
efficient, because the 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 transfering pages to the host in
chunks. A chunk consists of guest physically continuous pages, and
it 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 pages). A normal chunk is formated as below:
-----------------------------------------------
| Base (52 bit) | Size (12 bit)|
-----------------------------------------------
For large size chunks, an extended chunk format is used:
-----------------------------------------------
| Base (64 bit) |
-----------------------------------------------
-----------------------------------------------
| Size (64 bit) |
-----------------------------------------------

By doing so, step 4) can also be optimized by doing address
translation and madvise() in chunks rather than page by page.

This optimization requires the negotation of a new feature bit,
VIRTIO_BALLOON_F_CHUNK_TRANSFER.

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: Liang Li <liang.z.li@xxxxxxxxx>
Signed-off-by: Wei Wang <wei.w.wang@xxxxxxxxx>
Suggested-by: Michael S. Tsirkin <mst@xxxxxxxxxx>
Cc: Michael S. Tsirkin <mst@xxxxxxxxxx>
Cc: Paolo Bonzini <pbonzini@xxxxxxxxxx>
Cc: Cornelia Huck <cornelia.huck@xxxxxxxxxx>
Cc: Amit Shah <amit.shah@xxxxxxxxxx>
Cc: Dave Hansen <dave.hansen@xxxxxxxxx>
Cc: Andrea Arcangeli <aarcange@xxxxxxxxxx>
Cc: David Hildenbrand <david@xxxxxxxxxx>
Cc: Liang Li <liliang324@xxxxxxxxx>
Cc: Wei Wang <wei.w.wang@xxxxxxxxx>
Does this pass sparse? I see some endian-ness issues here.

"pass sparse"- what does that mean?
I didn't see any complaints from "make" on my machine.

---
drivers/virtio/virtio_balloon.c | 351 ++++++++++++++++++++++++++++++++++++----
1 file changed, 323 insertions(+), 28 deletions(-)

diff --git a/drivers/virtio/virtio_balloon.c b/drivers/virtio/virtio_balloon.c
index f59cb4f..4416370 100644
--- a/drivers/virtio/virtio_balloon.c
+++ b/drivers/virtio/virtio_balloon.c
@@ -42,6 +42,10 @@
#define OOM_VBALLOON_DEFAULT_PAGES 256
#define VIRTBALLOON_OOM_NOTIFY_PRIORITY 80
+#define PAGE_BMAP_SIZE (8 * PAGE_SIZE)
+#define PFNS_PER_PAGE_BMAP (PAGE_BMAP_SIZE * BITS_PER_BYTE)
+#define PAGE_BMAP_COUNT_MAX 32
+
static int oom_pages = OOM_VBALLOON_DEFAULT_PAGES;
module_param(oom_pages, int, S_IRUSR | S_IWUSR);
MODULE_PARM_DESC(oom_pages, "pages to free on OOM");
@@ -50,6 +54,16 @@ MODULE_PARM_DESC(oom_pages, "pages to free on OOM");
static struct vfsmount *balloon_mnt;
#endif
+struct balloon_page_chunk {
+ __le64 base : 52;
+ __le64 size : 12;
+};
+
+struct balloon_page_chunk_ext {
+ __le64 base;
+ __le64 size;
+};
+
struct virtio_balloon {
struct virtio_device *vdev;
struct virtqueue *inflate_vq, *deflate_vq, *stats_vq;
@@ -67,6 +81,20 @@ struct virtio_balloon {
/* Number of balloon pages we've told the Host we're not using. */
unsigned int num_pages;
+ /* Pointer to the response header. */
+ struct virtio_balloon_resp_hdr resp_hdr;
+ /* Pointer to the start address of response data. */
+ __le64 *resp_data;
+ /* Size of response data buffer. */
+ unsigned int resp_buf_size;
+ /* Pointer offset of the response data. */
+ unsigned int resp_pos;
+ /* Bitmap used to save the pfns info */
+ unsigned long *page_bitmap[PAGE_BMAP_COUNT_MAX];
+ /* Number of split page bitmaps */
+ unsigned int nr_page_bmap;
+ /* Used to record the processed pfn range */
+ unsigned long min_pfn, max_pfn, start_pfn, end_pfn;
/*
* The pages we've told the Host we're not using are enqueued
* at vb_dev_info->pages list.
@@ -110,20 +138,180 @@ static void balloon_ack(struct virtqueue *vq)
wake_up(&vb->acked);
}
-static void tell_host(struct virtio_balloon *vb, struct virtqueue *vq)
+static inline void init_bmap_pfn_range(struct virtio_balloon *vb)
{
- struct scatterlist sg;
+ vb->min_pfn = ULONG_MAX;
+ vb->max_pfn = 0;
+}
+
+static inline void update_bmap_pfn_range(struct virtio_balloon *vb,
+ struct page *page)
+{
+ unsigned long balloon_pfn = page_to_balloon_pfn(page);
+
+ vb->min_pfn = min(balloon_pfn, vb->min_pfn);
+ vb->max_pfn = max(balloon_pfn, vb->max_pfn);
+}
+
+static void extend_page_bitmap(struct virtio_balloon *vb,
+ unsigned long nr_pfn)

what exactly does this do?

By default, only one page_bmap is allocated.
When there are too many pfns to fit into one page_bmap,
it tries to allocate more page_bmap using this extend_page_bmap() function.


+{
+ int i, bmap_count;
+ unsigned long bmap_len;
+
+ bmap_len = ALIGN(nr_pfn, BITS_PER_LONG) / BITS_PER_BYTE;
+ bmap_len = ALIGN(bmap_len, PAGE_BMAP_SIZE);
+ bmap_count = min((int)(bmap_len / PAGE_BMAP_SIZE),
+ PAGE_BMAP_COUNT_MAX);
+
+ for (i = 1; i < bmap_count; i++) {
+ vb->page_bitmap[i] = kmalloc(PAGE_BMAP_SIZE, GFP_KERNEL);
+ if (vb->page_bitmap[i])
+ vb->nr_page_bmap++;
+ else
+ break;
seems to fail silently ...

I think here it is fine whether the page_bmap extension succeeds or silently fails.
It doesn't affect the correctness of the implementation - the page_bmap is always
used through vb->nr_page_bmap. Maybe the code logic is still a little fuzzy. I plan
to re-write most of the implementation to make it clearer.

+ }
+}
+
+static void free_extended_page_bitmap(struct virtio_balloon *vb)
+{
+ int i, bmap_count = vb->nr_page_bmap;
+
+ for (i = 1; i < bmap_count; i++) {
+ kfree(vb->page_bitmap[i]);
+ vb->page_bitmap[i] = NULL;
+ vb->nr_page_bmap--;
+ }
+}
+
+static void kfree_page_bitmap(struct virtio_balloon *vb)
+{
+ int i;
+
+ for (i = 0; i < vb->nr_page_bmap; i++)
+ kfree(vb->page_bitmap[i]);
+}
+
A very confusing name. pls don't start with kfree.

Ok. I'll simply call it free_page_bmap()


+static void clear_page_bitmap(struct virtio_balloon *vb)
+{
+ int i;
+
+ for (i = 0; i < vb->nr_page_bmap; i++)
+ memset(vb->page_bitmap[i], 0, PAGE_BMAP_SIZE);
+}
+
+static void send_resp_data(struct virtio_balloon *vb, struct virtqueue *vq,
+ bool busy_wait)
+{
+ struct scatterlist sg[2];
+ struct virtio_balloon_resp_hdr *hdr = &vb->resp_hdr;
unsigned int len;
- sg_init_one(&sg, vb->pfns, sizeof(vb->pfns[0]) * vb->num_pfns);
+ len = hdr->data_len = vb->resp_pos * sizeof(__le64);
+ sg_init_table(sg, 2);
+ sg_set_buf(&sg[0], hdr, sizeof(struct virtio_balloon_resp_hdr));
+ sg_set_buf(&sg[1], vb->resp_data, len);
+
+ if (virtqueue_add_outbuf(vq, sg, 2, vb, GFP_KERNEL) == 0) {
!virtqueue_add_outbuf

Also how about making this header linear with pfns?
This way you can use 1 s/g and avoid mallocs.

Yes, I think it'll be better, thanks.


+ virtqueue_kick(vq);
+ if (busy_wait)
+ while (!virtqueue_get_buf(vq, &len)
+ && !virtqueue_is_broken(vq))
+ cpu_relax();
+ else
+ wait_event(vb->acked, virtqueue_get_buf(vq, &len));
+ vb->resp_pos = 0;
+ free_extended_page_bitmap(vb);
+ }
+}
- /* We should always be able to add one buffer to an empty queue. */
- virtqueue_add_outbuf(vq, &sg, 1, vb, GFP_KERNEL);
- virtqueue_kick(vq);
+static void do_set_resp_bitmap(struct virtio_balloon *vb, unsigned long base,
+ int size)
+{
+ /* Use the extented chunk format if the size is too large */
+ if (size > (1 << VIRTIO_BALLOON_CHUNK_SIZE_SHIFT)) {
+ struct balloon_page_chunk_ext *chunk_ext =
+ (struct balloon_page_chunk_ext *)(vb->resp_data + vb->resp_pos);
+ chunk_ext->base = cpu_to_le64(base);
+ chunk_ext->size = cpu_to_le64(size);
+ vb->resp_pos += sizeof(vb->resp_pos) / sizeof(*chunk_ext);
+ } else {
+ struct balloon_page_chunk *chunk =
+ (struct balloon_page_chunk *)(vb->resp_data + vb->resp_pos);
+ chunk->base = cpu_to_le64(base);
+ chunk->size = cpu_to_le16(size);
+ vb->resp_pos += sizeof(vb->resp_pos) / sizeof(*chunk);
+ }
+}
- /* When host has read buffer, this completes via balloon_ack */
- wait_event(vb->acked, virtqueue_get_buf(vq, &len));
+static void chunking_pages_from_bmap(struct virtio_balloon *vb,
+ struct virtqueue *vq,
+ unsigned long start_pfn, unsigned long *bitmap,
+ unsigned long len, bool busy_wait)
+{
+ unsigned long pos = 0, end = len * BITS_PER_BYTE;
+
+ while (pos < end) {
+ unsigned long one = find_next_bit(bitmap, end, pos);
+
+ if (one < end) {
+ unsigned long chunk_size, zero;
+
+ zero = find_next_zero_bit(bitmap, end, one + 1);
+ if (zero >= end)
+ chunk_size = end - one;
+ else
+ chunk_size = zero - one;
+ if (chunk_size) {
+ if ((vb->resp_pos + 2) * sizeof(__le64) >
+ vb->resp_buf_size)
why + 2? where does sizeof come from?
Any chance this can be cleaned up using structs sizeof etc?
or at least commented?
How am I supposed to figure out all this convoluted pointer math?

Right. I'll take of it in the new version.


+ send_resp_data(vb, vq, busy_wait);
+ do_set_resp_bitmap(vb, start_pfn + one, chunk_size);
+ }
+ pos = one + chunk_size;
+ } else
+ break;
+ }
+}
+static void tell_host(struct virtio_balloon *vb, struct virtqueue *vq)
+{
+ if (virtio_has_feature(vb->vdev, VIRTIO_BALLOON_F_CHUNK_TRANSFER)) {
+ int nr_pfn, nr_used_bmap, i;
+ unsigned long start_pfn, bmap_len;
+
+ start_pfn = vb->start_pfn;
+ nr_pfn = vb->end_pfn - start_pfn + 1;
+ nr_pfn = roundup(nr_pfn, BITS_PER_LONG);
+ nr_used_bmap = nr_pfn / PFNS_PER_PAGE_BMAP;
better name here including "page"?
And bmaps instead of bmap?
Agree. I'll call it page_bmaps.


+ if (nr_pfn % PFNS_PER_PAGE_BMAP)
+ nr_used_bmap++;
Pls just use macros like ALIGN or roundup?

+ bmap_len = nr_pfn / BITS_PER_BYTE;
I guess we need to round up.

+
+ for (i = 0; i < nr_used_bmap; i++) {
+ unsigned int bmap_size = PAGE_BMAP_SIZE;
+
+ if (i + 1 == nr_used_bmap)
+ bmap_size = bmap_len - PAGE_BMAP_SIZE * i;
So last one is treated speacially? why?

The last one takes the leftover only, so that chunking_pages_from_bmap()
only searches the necessary part of (rather than the entire) the page_bmap.

+ chunking_pages_from_bmap(vb, vq, start_pfn + i * PFNS_PER_PAGE_BMAP,
+ vb->page_bitmap[i], bmap_size, false);
+ }
+ if (vb->resp_pos > 0)
+ send_resp_data(vb, vq, false);
+ } else {
+ struct scatterlist sg;
+ unsigned int len;
+
+ sg_init_one(&sg, vb->pfns, sizeof(vb->pfns[0]) * vb->num_pfns);
+
+ /* We should always be able to add one buffer to an
+ * empty queue
+ */
/*
* Multi-line comments
* like this.
*/

OK.

+ virtqueue_add_outbuf(vq, &sg, 1, vb, GFP_KERNEL);
+ virtqueue_kick(vq);
+ /* When host has read buffer, this completes via balloon_ack */
+ wait_event(vb->acked, virtqueue_get_buf(vq, &len));
+ }
}
static void set_page_pfns(struct virtio_balloon *vb,
@@ -138,13 +326,58 @@ static void set_page_pfns(struct virtio_balloon *vb,
page_to_balloon_pfn(page) + i);
}
-static unsigned fill_balloon(struct virtio_balloon *vb, size_t num)
+static void set_page_bitmap(struct virtio_balloon *vb,
+ struct list_head *pages, struct virtqueue *vq)
align on ( pls.

+{
+ unsigned long pfn, pfn_limit;
+ struct page *page;
+ bool found;
+
+ vb->min_pfn = rounddown(vb->min_pfn, BITS_PER_LONG);
+ vb->max_pfn = roundup(vb->max_pfn, BITS_PER_LONG);
+ pfn_limit = PFNS_PER_PAGE_BMAP * vb->nr_page_bmap;
+
+ if (vb->nr_page_bmap == 1)
+ extend_page_bitmap(vb, vb->max_pfn - vb->min_pfn + 1);
how is 1 special?

Thanks. I think "if (vb->nr_page_bmap == 1)" should be removed.
Whenever the current amount of page_bmap is not enough to hold
the amount of pfns, it should try to entend page_bmap.

Best,
Wei