From: John Snow <jsnow@redhat.com>
To: Max Reitz <mreitz@redhat.com>, qemu-devel@nongnu.org
Cc: kwolf@redhat.com, famz@redhat.com, armbru@redhat.com,
vsementsov@parallels.com, stefanha@redhat.com
Subject: Re: [Qemu-devel] [PATCH v14 18/19] block: Resize bitmaps on bdrv_truncate
Date: Mon, 23 Feb 2015 14:12:53 -0500 [thread overview]
Message-ID: <54EB7BB5.2030403@redhat.com> (raw)
In-Reply-To: <54EB5D66.2060304@redhat.com>
On 02/23/2015 12:03 PM, Max Reitz wrote:
> On 2015-02-20 at 18:07, John Snow wrote:
>> Signed-off-by: John Snow <jsnow@redhat.com>
>> ---
>> block.c | 20 ++++++++++++++++++++
>> include/block/block.h | 1 +
>> include/qemu/hbitmap.h | 10 ++++++++++
>> util/hbitmap.c | 47
>> +++++++++++++++++++++++++++++++++++++++++++++++
>> 4 files changed, 78 insertions(+)
>>
>> diff --git a/block.c b/block.c
>> index f3a6dd4..59a8ec9 100644
>> --- a/block.c
>> +++ b/block.c
>> @@ -3514,6 +3514,7 @@ int bdrv_truncate(BlockDriverState *bs, int64_t
>> offset)
>> ret = drv->bdrv_truncate(bs, offset);
>> if (ret == 0) {
>> ret = refresh_total_sectors(bs, offset >> BDRV_SECTOR_BITS);
>> + bdrv_dirty_bitmap_truncate(bs);
>> if (bs->blk) {
>> blk_dev_resize_cb(bs->blk);
>> }
>> @@ -5524,6 +5525,25 @@ BdrvDirtyBitmap
>> *bdrv_reclaim_dirty_bitmap(BlockDriverState *bs,
>> return parent;
>> }
>> +static void dirty_bitmap_truncate(BdrvDirtyBitmap *bitmap, uint64_t
>> size)
>> +{
>> + assert(!bdrv_dirty_bitmap_frozen(bitmap));
>
> This assertion is valid because the only way to make a bitmap frozen is
> by starting a drive-backup block job on the BDS, and doing so blocks the
> HMP block_resize command (and I don't know other ways to invoke
> bdrv_truncate() online).
>
> (That means: It's fine. I just had to think about whether it is.)
>
>> + hbitmap_truncate(bitmap->bitmap, size);
>> +}
>> +
>> +void bdrv_dirty_bitmap_truncate(BlockDriverState *bs)
>> +{
>> + BdrvDirtyBitmap *bitmap;
>> + uint64_t size = bdrv_nb_sectors(bs);
>> +
>> + QLIST_FOREACH(bitmap, &bs->dirty_bitmaps, list) {
>> + if (bdrv_dirty_bitmap_frozen(bitmap)) {
>> + continue;
>> + }
>> + dirty_bitmap_truncate(bitmap, size);
>> + }
>> +}
>> +
>> void bdrv_release_dirty_bitmap(BlockDriverState *bs, BdrvDirtyBitmap
>> *bitmap)
>> {
>> BdrvDirtyBitmap *bm, *next;
>> diff --git a/include/block/block.h b/include/block/block.h
>> index f6a50ae..aa6912d 100644
>> --- a/include/block/block.h
>> +++ b/include/block/block.h
>> @@ -442,6 +442,7 @@ BdrvDirtyBitmap
>> *bdrv_create_dirty_bitmap(BlockDriverState *bs,
>> uint32_t granularity,
>> const char *name,
>> Error **errp);
>> +void bdrv_dirty_bitmap_truncate(BlockDriverState *bs);
>> int bdrv_dirty_bitmap_create_successor(BlockDriverState *bs,
>> BdrvDirtyBitmap *bitmap,
>> Error **errp);
>> diff --git a/include/qemu/hbitmap.h b/include/qemu/hbitmap.h
>> index c19c1cb..a75157e 100644
>> --- a/include/qemu/hbitmap.h
>> +++ b/include/qemu/hbitmap.h
>> @@ -65,6 +65,16 @@ struct HBitmapIter {
>> HBitmap *hbitmap_alloc(uint64_t size, int granularity);
>> /**
>> + * hbitmap_truncate:
>> + * @hb: The bitmap to change the size of.
>> + * @size: The number of elements to change the bitmap to accommodate.
>> + *
>> + * truncate or grow an existing bitmap to accommodate a new number of
>> elements.
>> + * This may invalidate existing HBitmapIterators.
>> + */
>> +void hbitmap_truncate(HBitmap *hb, uint64_t size);
>> +
>> +/**
>> * hbitmap_merge:
>> * @a: The bitmap to store the result in.
>> * @b: The bitmap to merge into @a.
>> diff --git a/util/hbitmap.c b/util/hbitmap.c
>> index 962ff29..d17b850 100644
>> --- a/util/hbitmap.c
>> +++ b/util/hbitmap.c
>> @@ -90,6 +90,9 @@ struct HBitmap {
>> * bitmap will still allocate HBITMAP_LEVELS arrays.
>> */
>> unsigned long *levels[HBITMAP_LEVELS];
>> +
>> + /* The length of each levels[] array. */
>> + uint64_t sizes[HBITMAP_LEVELS];
>> };
>> /* Advance hbi to the next nonzero word and return it. hbi->pos
>> @@ -384,6 +387,7 @@ HBitmap *hbitmap_alloc(uint64_t size, int
>> granularity)
>> hb->granularity = granularity;
>> for (i = HBITMAP_LEVELS; i-- > 0; ) {
>> size = MAX((size + BITS_PER_LONG - 1) >> BITS_PER_LEVEL, 1);
>> + hb->sizes[i] = size;
>> hb->levels[i] = g_new0(unsigned long, size);
>> }
>> @@ -396,6 +400,49 @@ HBitmap *hbitmap_alloc(uint64_t size, int
>> granularity)
>> return hb;
>> }
>> +void hbitmap_truncate(HBitmap *hb, uint64_t size)
>> +{
>> + bool truncate;
>> + unsigned i;
>> + uint64_t num_elements = size;
>> +
>> + /* Size comes in as logical elements, adjust for granularity. */
>> + size = (size + (1ULL << hb->granularity) - 1) >> hb->granularity;
>> + assert(size <= ((uint64_t)1 << HBITMAP_LOG_MAX_SIZE));
>> + truncate = size < hb->size;
>> +
>> + if (size == hb->size) {
>> + /* A hard day's work */
>> + return;
>> + }
>> +
>> + hb->size = size;
>> + for (i = HBITMAP_LEVELS; i-- > 0; ) {
>
> By now I've heard that this is a relatively wide-spread pattern; but I
> still don't like it because I find it pretty hard to figure out that it
> does what it's supposed to if you've never seen it before (and with "by
> now I've heard" I mean that I read some blog post where someone
> mentioned that he didn't like it).
>
> At least it's not "i --> 0", so I guess I should be fine with it.
>
> (and with "by now" I mean that I'm the one responsible for it not being
> "-->" in hbitmap_merge())
>
A copy-paste loop from hbitmap_alloc is why this pattern keeps emerging.
It's not necessarily how I'd do it, either.
>> + size = MAX((size + BITS_PER_LONG - 1) >> BITS_PER_LEVEL, 1);
>> + if (hb->sizes[i] == size) {
>> + continue;
>> + }
>> + hb->sizes[i] = size;
>> + hb->levels[i] = g_realloc_n(hb->levels[i], size,
>> sizeof(unsigned long));
>
> How about initializing the new memory in case of truncate == false? I'm
> not sure whether to initialize it to 0 or 1, though. Probably 0.
>
Good call; I had actually thought that g_realloc did initialize the
memory, but looking at the docs again: nope.
>> + }
>> + assert(size == 1);
>> +
>> + /* Clear out any "extra space" we may have that the user didn't
>> request:
>> + * It may have garbage data in it, now. */
>> + if (truncate) {
>> + /* Due to granularity fuzziness, we may accidentally reset
>> some of
>> + * the last bits that are actually valid. So, record the
>> current value,
>> + * reset the "dead range," then re-set the one sector we care
>> about. */
>> + bool status = hbitmap_get(hb, num_elements - 1);
>> + uint64_t fix_count = (hb->size << hb->granularity) -
>> num_elements;
>> + hbitmap_reset(hb, num_elements, fix_count);
>
> This can fail an assertion (or at least I believe it's this function, it
> may be the hbitmap_set() below, too (but status should be false, so I
> don't think that)):
>
> ./qemu-img create -f raw test.img 64M; echo
> "{'execute':'qmp_capabilities'}{'execute':'block-dirty-bitmap-add','arguments':{'node':'drv','name':'foo'}}{'execute':'human-monitor-command','arguments':{'command-line':'block_resize
> drv 32M'}}" | x86_64-softmmu/qemu-system-x86_64 -drive
> if=none,id=drv,file=test.img,format=raw -qmp stdio -machine accel=qtest
> -nodefaults -display none
> Formatting 'test.img', fmt=raw size=67108864
> {"QMP": {"version": {"qemu": {"micro": 50, "minor": 2, "major": 2},
> "package": ""}, "capabilities": []}}
> {"return": {}}
> {"return": {}}
> qemu-system-x86_64: util/hbitmap.c:149: hbitmap_iter_init: Assertion
> `pos < hb->size' failed.
>
> It's probably because fix_count == 0 and therefore num_elements is not
> an index into the bitmap (which hbitmap_reset() and subsequently
> hb_count_between() don't take into account).
>
> Max
>
Ah, indeed. If you hit a perfect even multiple where there are no bits
at all to clean, you wind up with a one-past-the-end, even though the
operation should be a no-op.
>> + if (status) {
>> + hbitmap_set(hb, num_elements - 1, 1);
>> + }
>> + }
>> +}
>> +
>> +
>> /**
>> * Given HBitmaps A and B, let A := A (BITOR) B.
>> * Bitmap B will not be modified.
>
next prev parent reply other threads:[~2015-02-23 19:13 UTC|newest]
Thread overview: 28+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-02-20 23:07 [Qemu-devel] [PATCH v14 00/19] block: incremental backup series John Snow
2015-02-20 23:07 ` [Qemu-devel] [PATCH v14 01/19] qapi: Add optional field "name" to block dirty bitmap John Snow
2015-02-20 23:07 ` [Qemu-devel] [PATCH v14 02/19] qmp: Ensure consistent granularity type John Snow
2015-02-20 23:07 ` [Qemu-devel] [PATCH v14 03/19] qmp: Add block-dirty-bitmap-add and block-dirty-bitmap-remove John Snow
2015-02-27 18:44 ` Eric Blake
2015-02-20 23:07 ` [Qemu-devel] [PATCH v14 04/19] block: Introduce bdrv_dirty_bitmap_granularity() John Snow
2015-02-27 20:42 ` Eric Blake
2015-02-20 23:07 ` [Qemu-devel] [PATCH v14 05/19] hbitmap: add hbitmap_merge John Snow
2015-02-20 23:07 ` [Qemu-devel] [PATCH v14 06/19] qmp: Add block-dirty-bitmap-enable and block-dirty-bitmap-disable John Snow
2015-02-20 23:07 ` [Qemu-devel] [PATCH v14 07/19] block: Add bitmap successors John Snow
2015-02-20 23:07 ` [Qemu-devel] [PATCH v14 08/19] qmp: Add support of "dirty-bitmap" sync mode for drive-backup John Snow
2015-02-20 23:07 ` [Qemu-devel] [PATCH v14 09/19] qmp: add block-dirty-bitmap-clear John Snow
2015-02-20 23:07 ` [Qemu-devel] [PATCH v14 10/19] qapi: Add transaction support to block-dirty-bitmap operations John Snow
2015-02-23 15:48 ` Max Reitz
2015-02-20 23:07 ` [Qemu-devel] [PATCH v14 11/19] qmp: Add dirty bitmap status fields in query-block John Snow
2015-02-20 23:07 ` [Qemu-devel] [PATCH v14 12/19] block: add BdrvDirtyBitmap documentation John Snow
2015-02-20 23:07 ` [Qemu-devel] [PATCH v14 13/19] block: Ensure consistent bitmap function prototypes John Snow
2015-02-20 23:07 ` [Qemu-devel] [PATCH v14 14/19] iotests: add invalid input incremental backup tests John Snow
2015-02-20 23:07 ` [Qemu-devel] [PATCH v14 15/19] iotests: add simple incremental backup case John Snow
2015-02-20 23:07 ` [Qemu-devel] [PATCH v14 16/19] iotests: add transactional incremental backup test John Snow
2015-02-20 23:07 ` [Qemu-devel] [PATCH v14 17/19] iotests: add incremental backup failure recovery test John Snow
2015-02-20 23:07 ` [Qemu-devel] [PATCH v14 18/19] block: Resize bitmaps on bdrv_truncate John Snow
2015-02-23 17:03 ` Max Reitz
2015-02-23 19:12 ` John Snow [this message]
2015-02-20 23:07 ` [Qemu-devel] [PATCH v14 19/19] docs: incremental backup documentation John Snow
2015-02-20 23:55 ` Eric Blake
2015-02-21 0:09 ` John Snow
2015-02-23 17:07 ` [Qemu-devel] [PATCH v14 00/19] block: incremental backup series Max Reitz
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=54EB7BB5.2030403@redhat.com \
--to=jsnow@redhat.com \
--cc=armbru@redhat.com \
--cc=famz@redhat.com \
--cc=kwolf@redhat.com \
--cc=mreitz@redhat.com \
--cc=qemu-devel@nongnu.org \
--cc=stefanha@redhat.com \
--cc=vsementsov@parallels.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).