[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Qemu-block] [PATCH RFC for-2.6 1/3] HBitmap: Introduce "meta" bitma
From: |
Fam Zheng |
Subject: |
Re: [Qemu-block] [PATCH RFC for-2.6 1/3] HBitmap: Introduce "meta" bitmap to track bit changes |
Date: |
Wed, 30 Dec 2015 19:07:14 +0800 |
User-agent: |
Mutt/1.5.21 (2010-09-15) |
On Wed, 12/30 13:53, Vladimir Sementsov-Ogievskiy wrote:
> On 07.12.2015 08:59, Fam Zheng wrote:
> >The meta bitmap will have the same size and granularity as the tracked
> >bitmap, and upon each bit toggle, the corresponding bit in the meta
> >bitmap, at an identical position, will be set.
> >
> >Signed-off-by: Fam Zheng <address@hidden>
> >---
> > include/qemu/hbitmap.h | 7 +++++++
> > util/hbitmap.c | 22 ++++++++++++++++++++++
> > 2 files changed, 29 insertions(+)
> >
> >diff --git a/include/qemu/hbitmap.h b/include/qemu/hbitmap.h
> >index bb94a00..09a6b06 100644
> >--- a/include/qemu/hbitmap.h
> >+++ b/include/qemu/hbitmap.h
> >@@ -181,6 +181,13 @@ void hbitmap_iter_init(HBitmapIter *hbi, const HBitmap
> >*hb, uint64_t first);
> > */
> > unsigned long hbitmap_iter_skip_words(HBitmapIter *hbi);
> >+/* hbitmap_create_meta
> >+ * @hb: The HBitmap to operate on.
> >+ *
> >+ * Create a "meta" hbitmap to track dirtiness of the bits in this HBitmap.
> >+ */
> >+HBitmap *hbitmap_create_meta(HBitmap *hb);
> >+
> > /**
> > * hbitmap_iter_next:
> > * @hbi: HBitmapIter to operate on.
> >diff --git a/util/hbitmap.c b/util/hbitmap.c
> >index 50b888f..3ad406e 100644
> >--- a/util/hbitmap.c
> >+++ b/util/hbitmap.c
> >@@ -81,6 +81,9 @@ struct HBitmap {
> > */
> > int granularity;
> >+ /* A meta dirty bitmap to track the dirtiness of bits in this HBitmap.
> >*/
> >+ HBitmap *meta;
> >+
> > /* A number of progressively less coarse bitmaps (i.e. level 0 is the
> > * coarsest). Each bit in level N represents a word in level N+1 that
> > * has a set bit, except the last level where each bit represents the
> >@@ -232,6 +235,7 @@ static inline bool hb_set_elem(unsigned long *elem,
> >uint64_t start, uint64_t las
> > /* The recursive workhorse (the depth is limited to HBITMAP_LEVELS)... */
> > static void hb_set_between(HBitmap *hb, int level, uint64_t start,
> > uint64_t last)
> > {
> >+ uint64_t save_start = start;
> > size_t pos = start >> BITS_PER_LEVEL;
> > size_t lastpos = last >> BITS_PER_LEVEL;
> > bool changed = false;
> >@@ -252,6 +256,9 @@ static void hb_set_between(HBitmap *hb, int level,
> >uint64_t start, uint64_t last
> > }
> > }
> > changed |= hb_set_elem(&hb->levels[level][i], start, last);
> >+ if (hb->meta && level == HBITMAP_LEVELS - 1 && changed) {
> >+ hbitmap_set(hb->meta, save_start, last - save_start + 1);
> >+ }
>
> I thing now, that the same may be accomplished for BdrvDirtyBitmap,
> all we need is return "changed" status from hb_set_between and then
> from hbitmap_set.
That is true, but it makes further optimizing to *really* only set the toggled
meta bits much more difficult (i.e. when only a few bits between start and last
are changed).
I haven't written any code for that optimization, but I did base my other
persistent dirty bitmap work on v2 of this series. It would be great if we can
harmonize on this, so we both have a common base of block dirty bitmap.
I can post v2 to the list very soon, if the idea is okay for you; or if you've
your preferred way, we can take a look together. What do you think?
Fam
>
> > /* If there was any change in this layer, we may have to update
> > * the one above.
> >@@ -298,6 +305,7 @@ static inline bool hb_reset_elem(unsigned long *elem,
> >uint64_t start, uint64_t l
> > /* The recursive workhorse (the depth is limited to HBITMAP_LEVELS)... */
> > static void hb_reset_between(HBitmap *hb, int level, uint64_t start,
> > uint64_t last)
> > {
> >+ uint64_t save_start = start;
> > size_t pos = start >> BITS_PER_LEVEL;
> > size_t lastpos = last >> BITS_PER_LEVEL;
> > bool changed = false;
> >@@ -336,6 +344,10 @@ static void hb_reset_between(HBitmap *hb, int level,
> >uint64_t start, uint64_t la
> > lastpos--;
> > }
> >+ if (hb->meta && level == HBITMAP_LEVELS - 1 && changed) {
> >+ hbitmap_set(hb->meta, save_start, last - save_start + 1);
> >+ }
> >+
> > if (level > 0 && changed) {
> > hb_reset_between(hb, level - 1, pos, lastpos);
> > }
> >@@ -384,6 +396,9 @@ void hbitmap_free(HBitmap *hb)
> > for (i = HBITMAP_LEVELS; i-- > 0; ) {
> > g_free(hb->levels[i]);
> > }
> >+ if (hb->meta) {
> >+ hbitmap_free(hb->meta);
> >+ }
> > g_free(hb);
> > }
> >@@ -493,3 +508,10 @@ bool hbitmap_merge(HBitmap *a, const HBitmap *b)
> > return true;
> > }
> >+
> >+HBitmap *hbitmap_create_meta(HBitmap *hb)
> >+{
> >+ assert(!hb->meta);
> >+ hb->meta = hbitmap_alloc(hb->size, hb->granularity);
> >+ return hb->meta;
> >+}
>
>
> --
> Best regards,
> Vladimir
> * now, @virtuozzo.com instead of @parallels.com. Sorry for this inconvenience.
>
[Qemu-block] [PATCH RFC for-2.6 2/3] tests: Add test code for meta bitmap, Fam Zheng, 2015/12/07
[Qemu-block] [PATCH RFC for-2.6 3/3] block: Support meta dirty bitmap, Fam Zheng, 2015/12/07
Re: [Qemu-block] [PATCH RFC for-2.6 0/3] block: Add meta dirty bitmap for migration/persistence, Vladimir Sementsov-Ogievskiy, 2015/12/07
Re: [Qemu-block] [Qemu-devel] [PATCH RFC for-2.6 0/3] block: Add meta dirty bitmap for migration/persistence, John Snow, 2015/12/07