linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Wei Wang <wei.w.wang@intel.com>
To: virtio-dev@lists.oasis-open.org, linux-kernel@vger.kernel.org,
	qemu-devel@nongnu.org, virtualization@lists.linux-foundation.org,
	kvm@vger.kernel.org, linux-mm@kvack.org, mst@redhat.com,
	mhocko@kernel.org, akpm@linux-foundation.org,
	mawilcox@microsoft.com
Cc: david@redhat.com, penguin-kernel@I-love.SAKURA.ne.jp,
	cornelia.huck@de.ibm.com, mgorman@techsingularity.net,
	aarcange@redhat.com, amit.shah@redhat.com, pbonzini@redhat.com,
	willy@infradead.org, wei.w.wang@intel.com,
	liliang.opensource@gmail.com, yang.zhang.wz@gmail.com,
	quan.xu@aliyun.com, nilal@redhat.com, riel@redhat.com
Subject: [PATCH v18 07/10] virtio-balloon: VIRTIO_BALLOON_F_SG
Date: Wed, 29 Nov 2017 21:55:23 +0800	[thread overview]
Message-ID: <1511963726-34070-8-git-send-email-wei.w.wang@intel.com> (raw)
In-Reply-To: <1511963726-34070-1-git-send-email-wei.w.wang@intel.com>

Add a new feature, VIRTIO_BALLOON_F_SG, which enables the transfer of
balloon (i.e. inflated/deflated) pages using scatter-gather lists to the
host. A scatter-gather list is described by a vring desc.

The implementation of the previous virtio-balloon is not very efficient,
because the balloon pages are transferred to the host by one array each
time. 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 host in sgs. An sg
describes a chunk of guest physically continuous pages. With this
mechanism, 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 ~440ms resulting
in an improvement of ~89%.

TODO: optimize stage 1) by allocating/freeing a chunk of pages instead of
a single page each time.

Signed-off-by: Wei Wang <wei.w.wang@intel.com>
Signed-off-by: Liang Li <liang.z.li@intel.com>
Suggested-by: Michael S. Tsirkin <mst@redhat.com>
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
---
 drivers/virtio/virtio_balloon.c     | 230 +++++++++++++++++++++++++++++++++---
 include/uapi/linux/virtio_balloon.h |   1 +
 2 files changed, 212 insertions(+), 19 deletions(-)

diff --git a/drivers/virtio/virtio_balloon.c b/drivers/virtio/virtio_balloon.c
index 7960746..2c21c5a 100644
--- a/drivers/virtio/virtio_balloon.c
+++ b/drivers/virtio/virtio_balloon.c
@@ -32,6 +32,8 @@
 #include <linux/mm.h>
 #include <linux/mount.h>
 #include <linux/magic.h>
+#include <linux/xbitmap.h>
+#include <asm/page.h>
 
 /*
  * Balloon device works in 4K page units.  So each page is pointed to by
@@ -79,6 +81,9 @@ struct virtio_balloon {
 	/* Synchronize access/update to this struct virtio_balloon elements */
 	struct mutex balloon_lock;
 
+	/* The xbitmap used to record balloon pages */
+	struct xb page_xb;
+
 	/* The array of pfns we tell the Host about. */
 	unsigned int num_pfns;
 	__virtio32 pfns[VIRTIO_BALLOON_ARRAY_PFNS_MAX];
@@ -141,15 +146,121 @@ static void set_page_pfns(struct virtio_balloon *vb,
 					  page_to_balloon_pfn(page) + i);
 }
 
+static void kick_and_wait(struct virtqueue *vq, wait_queue_head_t wq_head)
+{
+	unsigned int len;
+
+	virtqueue_kick(vq);
+	wait_event(wq_head, virtqueue_get_buf(vq, &len));
+}
+
+static void send_one_desc(struct virtio_balloon *vb,
+			  struct virtqueue *vq,
+			  uint64_t addr,
+			  uint32_t len,
+			  bool inbuf,
+			  bool batch)
+{
+	int err;
+	unsigned int size;
+
+	/* Detach all the used buffers from the vq */
+	while (virtqueue_get_buf(vq, &size))
+		;
+
+	err = virtqueue_add_one_desc(vq, addr, len, inbuf, vq);
+	/*
+	 * This is expected to never fail: there is always at least 1 entry
+	 * available on the vq, because when the vq is full the worker thread
+	 * that adds the desc will be put into sleep until at least 1 entry is
+	 * available to use.
+	 */
+	BUG_ON(err);
+
+	/* If batching is requested, we batch till the vq is full */
+	if (!batch || !vq->num_free)
+		kick_and_wait(vq, vb->acked);
+}
+
+/*
+ * Send balloon pages in sgs to host. The balloon pages are recorded in the
+ * page xbitmap. Each bit in the bitmap corresponds to a page of PAGE_SIZE.
+ * The page xbitmap is searched for continuous "1" bits, which correspond
+ * to continuous pages, to chunk into sgs.
+ *
+ * @page_xb_start and @page_xb_end form the range of bits in the xbitmap that
+ * need to be searched.
+ */
+static void tell_host_sgs(struct virtio_balloon *vb,
+			  struct virtqueue *vq,
+			  unsigned long page_xb_start,
+			  unsigned long page_xb_end)
+{
+	unsigned long pfn_start, pfn_end;
+	uint64_t addr;
+	uint32_t len, max_len = round_down(UINT_MAX, PAGE_SIZE);
+
+	pfn_start = page_xb_start;
+	while (pfn_start < page_xb_end) {
+		pfn_start = xb_find_next_set_bit(&vb->page_xb, pfn_start,
+						 page_xb_end);
+		if (pfn_start == page_xb_end + 1)
+			break;
+		pfn_end = xb_find_next_zero_bit(&vb->page_xb,
+						pfn_start + 1,
+						page_xb_end);
+		addr = pfn_start << PAGE_SHIFT;
+		len = (pfn_end - pfn_start) << PAGE_SHIFT;
+		while (len > max_len) {
+			send_one_desc(vb, vq, addr, max_len, true, true);
+			addr += max_len;
+			len -= max_len;
+		}
+		send_one_desc(vb, vq, addr, len, true, true);
+		pfn_start = pfn_end + 1;
+	}
+
+	/*
+	 * The last few desc-s may not reach the batch size, but need a kick to
+	 * notify the device to handle them.
+	 */
+	if (vq->num_free != virtqueue_get_vring_size(vq))
+		kick_and_wait(vq, vb->acked);
+
+	xb_clear_bit_range(&vb->page_xb, page_xb_start, page_xb_end);
+}
+
+static inline int xb_set_page(struct virtio_balloon *vb,
+			       struct page *page,
+			       unsigned long *pfn_min,
+			       unsigned long *pfn_max)
+{
+	unsigned long pfn = page_to_pfn(page);
+	int ret;
+
+	*pfn_min = min(pfn, *pfn_min);
+	*pfn_max = max(pfn, *pfn_max);
+
+	do {
+		ret = xb_preload_and_set_bit(&vb->page_xb, pfn,
+					     GFP_NOWAIT | __GFP_NOWARN);
+	} while (unlikely(ret == -EAGAIN));
+
+	return ret;
+}
+
 static unsigned fill_balloon(struct virtio_balloon *vb, size_t num)
 {
 	unsigned num_allocated_pages;
 	unsigned num_pfns;
 	struct page *page;
 	LIST_HEAD(pages);
+	bool use_sg = virtio_has_feature(vb->vdev, VIRTIO_BALLOON_F_SG);
+	unsigned long pfn_max = 0, pfn_min = ULONG_MAX;
 
 	/* We can only do one array worth at a time. */
-	num = min(num, ARRAY_SIZE(vb->pfns));
+	if (!use_sg)
+		num = min(num, ARRAY_SIZE(vb->pfns));
 
 	for (num_pfns = 0; num_pfns < num;
 	     num_pfns += VIRTIO_BALLOON_PAGES_PER_PAGE) {
@@ -172,11 +283,18 @@ static unsigned fill_balloon(struct virtio_balloon *vb, size_t num)
 	vb->num_pfns = 0;
 
 	while ((page = balloon_page_pop(&pages))) {
+		if (use_sg) {
+			if (xb_set_page(vb, page, &pfn_min, &pfn_max) < 0) {
+				__free_page(page);
+				break;
+			}
+		} else {
+			set_page_pfns(vb, vb->pfns + vb->num_pfns, page);
+		}
+
 		balloon_page_enqueue(&vb->vb_dev_info, page);
 
 		vb->num_pfns += VIRTIO_BALLOON_PAGES_PER_PAGE;
-
-		set_page_pfns(vb, vb->pfns + vb->num_pfns, page);
 		vb->num_pages += VIRTIO_BALLOON_PAGES_PER_PAGE;
 		if (!virtio_has_feature(vb->vdev,
 					VIRTIO_BALLOON_F_DEFLATE_ON_OOM))
@@ -185,8 +303,12 @@ static unsigned fill_balloon(struct virtio_balloon *vb, size_t num)
 
 	num_allocated_pages = vb->num_pfns;
 	/* Did we get any? */
-	if (vb->num_pfns != 0)
-		tell_host(vb, vb->inflate_vq);
+	if (vb->num_pfns) {
+		if (use_sg)
+			tell_host_sgs(vb, vb->inflate_vq, pfn_min, pfn_max);
+		else
+			tell_host(vb, vb->inflate_vq);
+	}
 	mutex_unlock(&vb->balloon_lock);
 
 	return num_allocated_pages;
@@ -212,9 +334,12 @@ static unsigned leak_balloon(struct virtio_balloon *vb, size_t num)
 	struct page *page;
 	struct balloon_dev_info *vb_dev_info = &vb->vb_dev_info;
 	LIST_HEAD(pages);
+	bool use_sg = virtio_has_feature(vb->vdev, VIRTIO_BALLOON_F_SG);
+	unsigned long pfn_max = 0, pfn_min = ULONG_MAX;
 
-	/* We can only do one array worth at a time. */
-	num = min(num, ARRAY_SIZE(vb->pfns));
+	/* Traditionally, we can only do one array worth at a time. */
+	if (!use_sg)
+		num = min(num, ARRAY_SIZE(vb->pfns));
 
 	mutex_lock(&vb->balloon_lock);
 	/* We can't release more pages than taken */
@@ -224,7 +349,14 @@ static unsigned leak_balloon(struct virtio_balloon *vb, size_t num)
 		page = balloon_page_dequeue(vb_dev_info);
 		if (!page)
 			break;
-		set_page_pfns(vb, vb->pfns + vb->num_pfns, page);
+		if (use_sg) {
+			if (xb_set_page(vb, page, &pfn_min, &pfn_max) < 0) {
+				balloon_page_enqueue(&vb->vb_dev_info, page);
+				break;
+			}
+		} else {
+			set_page_pfns(vb, vb->pfns + vb->num_pfns, page);
+		}
 		list_add(&page->lru, &pages);
 		vb->num_pages -= VIRTIO_BALLOON_PAGES_PER_PAGE;
 	}
@@ -235,13 +367,55 @@ static unsigned leak_balloon(struct virtio_balloon *vb, size_t num)
 	 * virtio_has_feature(vdev, VIRTIO_BALLOON_F_MUST_TELL_HOST);
 	 * is true, we *have* to do it in this order
 	 */
-	if (vb->num_pfns != 0)
-		tell_host(vb, vb->deflate_vq);
+	if (vb->num_pfns) {
+		if (use_sg)
+			tell_host_sgs(vb, vb->deflate_vq, pfn_min, pfn_max);
+		else
+			tell_host(vb, vb->deflate_vq);
+	}
 	release_pages_balloon(vb, &pages);
 	mutex_unlock(&vb->balloon_lock);
 	return num_freed_pages;
 }
 
+/*
+ * The regular leak_balloon() with VIRTIO_BALLOON_F_SG needs memory allocation
+ * for xbitmap, which is not suitable for the oom case. This function does not
+ * use xbitmap to chunk pages, so it can be used by oom notifier to deflate
+ * pages when VIRTIO_BALLOON_F_SG is negotiated.
+ */
+static unsigned int leak_balloon_sg_oom(struct virtio_balloon *vb)
+{
+	unsigned int n;
+	struct page *page;
+	struct balloon_dev_info *vb_dev_info = &vb->vb_dev_info;
+	struct virtqueue *vq = vb->deflate_vq;
+	LIST_HEAD(pages);
+
+	mutex_lock(&vb->balloon_lock);
+	for (n = 0; n < oom_pages; n++) {
+		page = balloon_page_dequeue(vb_dev_info);
+		if (!page)
+			break;
+
+		list_add(&page->lru, &pages);
+		vb->num_pages -= VIRTIO_BALLOON_PAGES_PER_PAGE;
+		send_one_desc(vb, vq, virt_to_phys(page_address(page)),
+			      PAGE_SIZE, true, true);
+		release_pages_balloon(vb, &pages);
+	}
+
+	/*
+	 * The last few sgs may not reach the batch size, but need a kick to
+	 * notify the device to handle them.
+	 */
+	if (vq->num_free != virtqueue_get_vring_size(vq))
+		kick_and_wait(vq, vb->acked);
+	mutex_unlock(&vb->balloon_lock);
+
+	return n;
+}
+
 static inline void update_stat(struct virtio_balloon *vb, int idx,
 			       u16 tag, u64 val)
 {
@@ -381,7 +555,10 @@ static int virtballoon_oom_notify(struct notifier_block *self,
 		return NOTIFY_OK;
 
 	freed = parm;
-	num_freed_pages = leak_balloon(vb, oom_pages);
+	if (virtio_has_feature(vb->vdev, VIRTIO_BALLOON_F_SG))
+		num_freed_pages = leak_balloon_sg_oom(vb);
+	else
+		num_freed_pages = leak_balloon(vb, oom_pages);
 	update_balloon_size(vb);
 	*freed += num_freed_pages;
 
@@ -478,6 +655,7 @@ static int virtballoon_migratepage(struct balloon_dev_info *vb_dev_info,
 {
 	struct virtio_balloon *vb = container_of(vb_dev_info,
 			struct virtio_balloon, vb_dev_info);
+	bool use_sg = virtio_has_feature(vb->vdev, VIRTIO_BALLOON_F_SG);
 	unsigned long flags;
 
 	/*
@@ -499,16 +677,26 @@ static int virtballoon_migratepage(struct balloon_dev_info *vb_dev_info,
 	vb_dev_info->isolated_pages--;
 	__count_vm_event(BALLOON_MIGRATE);
 	spin_unlock_irqrestore(&vb_dev_info->pages_lock, flags);
-	vb->num_pfns = VIRTIO_BALLOON_PAGES_PER_PAGE;
-	set_page_pfns(vb, vb->pfns, newpage);
-	tell_host(vb, vb->inflate_vq);
-
+	if (use_sg) {
+		send_one_desc(vb, vb->inflate_vq,
+			      virt_to_phys(page_address(newpage)),
+			      PAGE_SIZE, true, false);
+	} else {
+		vb->num_pfns = VIRTIO_BALLOON_PAGES_PER_PAGE;
+		set_page_pfns(vb, vb->pfns, newpage);
+		tell_host(vb, vb->inflate_vq);
+	}
 	/* balloon's page migration 2nd step -- deflate "page" */
 	balloon_page_delete(page);
-	vb->num_pfns = VIRTIO_BALLOON_PAGES_PER_PAGE;
-	set_page_pfns(vb, vb->pfns, page);
-	tell_host(vb, vb->deflate_vq);
-
+	if (use_sg) {
+		send_one_desc(vb, vb->inflate_vq,
+			      virt_to_phys(page_address(page)),
+			      PAGE_SIZE, true, false);
+	} else {
+		vb->num_pfns = VIRTIO_BALLOON_PAGES_PER_PAGE;
+		set_page_pfns(vb, vb->pfns, page);
+		tell_host(vb, vb->deflate_vq);
+	}
 	mutex_unlock(&vb->balloon_lock);
 
 	put_page(page); /* balloon reference */
@@ -567,6 +755,9 @@ static int virtballoon_probe(struct virtio_device *vdev)
 	if (err)
 		goto out_free_vb;
 
+	if (virtio_has_feature(vdev, VIRTIO_BALLOON_F_SG))
+		xb_init(&vb->page_xb);
+
 	vb->nb.notifier_call = virtballoon_oom_notify;
 	vb->nb.priority = VIRTBALLOON_OOM_NOTIFY_PRIORITY;
 	err = register_oom_notifier(&vb->nb);
@@ -683,6 +874,7 @@ static unsigned int features[] = {
 	VIRTIO_BALLOON_F_MUST_TELL_HOST,
 	VIRTIO_BALLOON_F_STATS_VQ,
 	VIRTIO_BALLOON_F_DEFLATE_ON_OOM,
+	VIRTIO_BALLOON_F_SG,
 };
 
 static struct virtio_driver virtio_balloon_driver = {
diff --git a/include/uapi/linux/virtio_balloon.h b/include/uapi/linux/virtio_balloon.h
index 343d7dd..37780a7 100644
--- a/include/uapi/linux/virtio_balloon.h
+++ b/include/uapi/linux/virtio_balloon.h
@@ -34,6 +34,7 @@
 #define VIRTIO_BALLOON_F_MUST_TELL_HOST	0 /* Tell before reclaiming pages */
 #define VIRTIO_BALLOON_F_STATS_VQ	1 /* Memory Stats virtqueue */
 #define VIRTIO_BALLOON_F_DEFLATE_ON_OOM	2 /* Deflate balloon on OOM */
+#define VIRTIO_BALLOON_F_SG		3 /* Use sg instead of PFN lists */
 
 /* Size of a PFN in the balloon interface. */
 #define VIRTIO_BALLOON_PFN_SHIFT 12
-- 
2.7.4

  parent reply	other threads:[~2017-11-29 14:19 UTC|newest]

Thread overview: 37+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2017-11-29 13:55 [PATCH v18 00/10] Virtio-balloon Enhancement Wei Wang
2017-11-29 13:55 ` [PATCH v18 01/10] idr: add #include <linux/bug.h> Wei Wang
2017-11-30  0:58   ` Matthew Wilcox
2017-11-30  7:07     ` Michal Hocko
2017-11-30 21:49     ` Andrew Morton
2017-11-29 13:55 ` [PATCH v18 02/10] radix tree test suite: remove ARRAY_SIZE to avoid redefinition Wei Wang
2017-11-29 13:55 ` [PATCH v18 03/10] xbitmap: Introduce xbitmap Wei Wang
2017-11-29 13:55 ` [PATCH v18 04/10] xbitmap: potential improvement Wei Wang
2017-11-29 13:55 ` [PATCH v18 05/10] xbitmap: add more operations Wei Wang
2017-11-30 10:34   ` Tetsuo Handa
2017-11-30 13:35     ` Tetsuo Handa
2017-11-30 14:39       ` Matthew Wilcox
2017-12-03  1:44         ` Tetsuo Handa
2017-12-01  8:02     ` Wei Wang
2017-12-01 13:02       ` Tetsuo Handa
2017-12-01 14:13         ` Matthew Wilcox
2017-12-01 15:09         ` Wang, Wei W
2017-12-01 17:25           ` Matthew Wilcox
2017-12-03  1:50             ` Tetsuo Handa
2017-12-07 12:01               ` Wei Wang
2017-12-07 15:41                 ` Michael S. Tsirkin
2017-11-29 13:55 ` [PATCH v18 06/10] virtio_ring: add a new API, virtqueue_add_one_desc Wei Wang
2017-11-30 19:38   ` Michael S. Tsirkin
2017-12-01  8:06     ` Wei Wang
2017-11-29 13:55 ` Wei Wang [this message]
2017-11-30 10:35   ` [PATCH v18 07/10] virtio-balloon: VIRTIO_BALLOON_F_SG Tetsuo Handa
2017-11-30 16:25     ` Wang, Wei W
2017-12-01 15:38   ` Michael S. Tsirkin
2017-12-04  3:46     ` Wei Wang
2017-11-29 13:55 ` [PATCH v18 08/10] mm: support reporting free page blocks Wei Wang
2017-11-29 13:55 ` [PATCH v18 09/10] virtio-balloon: VIRTIO_BALLOON_F_FREE_PAGE_VQ Wei Wang
2017-11-29 13:55 ` [PATCH v18 10/10] virtio-balloon: don't report free pages when page poisoning is enabled Wei Wang
2017-12-01 15:49   ` Michael S. Tsirkin
2017-12-04  5:39     ` Wei Wang
2017-12-11  6:38     ` Wei Wang
2017-12-11 13:24       ` Michael S. Tsirkin
2017-12-12 12:21         ` Wei Wang

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=1511963726-34070-8-git-send-email-wei.w.wang@intel.com \
    --to=wei.w.wang@intel.com \
    --cc=aarcange@redhat.com \
    --cc=akpm@linux-foundation.org \
    --cc=amit.shah@redhat.com \
    --cc=cornelia.huck@de.ibm.com \
    --cc=david@redhat.com \
    --cc=kvm@vger.kernel.org \
    --cc=liliang.opensource@gmail.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-mm@kvack.org \
    --cc=mawilcox@microsoft.com \
    --cc=mgorman@techsingularity.net \
    --cc=mhocko@kernel.org \
    --cc=mst@redhat.com \
    --cc=nilal@redhat.com \
    --cc=pbonzini@redhat.com \
    --cc=penguin-kernel@I-love.SAKURA.ne.jp \
    --cc=qemu-devel@nongnu.org \
    --cc=quan.xu@aliyun.com \
    --cc=riel@redhat.com \
    --cc=virtio-dev@lists.oasis-open.org \
    --cc=virtualization@lists.linux-foundation.org \
    --cc=willy@infradead.org \
    --cc=yang.zhang.wz@gmail.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).