All of lore.kernel.org
 help / color / mirror / Atom feed
From: Max Reitz <mreitz@redhat.com>
To: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>,
	qemu-block@nongnu.org, qemu-devel@nongnu.org
Cc: kwolf@redhat.com, armbru@redhat.com, eblake@redhat.com,
	jsnow@redhat.com, famz@redhat.com, den@openvz.org,
	stefanha@redhat.com, pbonzini@redhat.com
Subject: Re: [Qemu-devel] [PATCH 07/22] qcow2-bitmap: introduce auto-loading bitmaps
Date: Fri, 21 Oct 2016 21:56:01 +0200	[thread overview]
Message-ID: <524e8350-025f-6ab0-763c-447fdaf72277@redhat.com> (raw)
In-Reply-To: <5c32e4bd-a947-8b7b-f55d-f9b84c03d9e6@virtuozzo.com>

[-- Attachment #1: Type: text/plain, Size: 3943 bytes --]

On 21.10.2016 13:59, Vladimir Sementsov-Ogievskiy wrote:
> 07.10.2016 22:25, Max Reitz пишет:
>> On 30.09.2016 12:53, Vladimir Sementsov-Ogievskiy wrote:
>>> Auto loading bitmaps are bitmaps in Qcow2, with AUTO flag set. They are
>>> loaded at image open and becomes BdrvDirtyBitmap's for corresponding
>>> drive. These bitmaps are deleted from Qcow2 image after loading to avoid
>>> conflicts.
>>>
>>> Extra data in bitmaps is not supported for now.
>>>
>>> Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
>>> ---
>>>   block/qcow2-bitmap.c | 518
>>> ++++++++++++++++++++++++++++++++++++++++++++++++++-
>>>   block/qcow2.c        |   5 +
>>>   block/qcow2.h        |   3 +
>>>   3 files changed, 525 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/block/qcow2-bitmap.c b/block/qcow2-bitmap.c
>>> index cd18b07..760a047 100644
>>> --- a/block/qcow2-bitmap.c
>>> +++ b/block/qcow2-bitmap.c
>> [...]
>>
>>> +static int directory_update(BlockDriverState *bs, uint8_t *new_dir,
>>> +                            size_t new_size, uint32_t new_nb_bitmaps)
>>> +{
>>> +    BDRVQcow2State *s = bs->opaque;
>>> +    int ret;
>>> +    int64_t new_offset = 0;
>>> +    uint64_t old_offset = s->bitmap_directory_offset;
>>> +    uint64_t old_size = s->bitmap_directory_size;
>>> +    uint32_t old_nb_bitmaps = s->nb_bitmaps;
>>> +    uint64_t old_autocl = s->autoclear_features;
>>> +
>>> +    if (new_size > QCOW_MAX_DIRTY_BITMAP_DIRECTORY_SIZE) {
>>> +        return -EINVAL;
>>> +    }
>>> +
>>> +    if ((new_size == 0) != (new_nb_bitmaps == 0)) {
>>> +        return -EINVAL;
>>> +    }
>>> +
>>> +    if (new_nb_bitmaps > 0) {
>>> +        new_offset = directory_write(bs, new_dir, new_size);
>>> +        if (new_offset < 0) {
>>> +            return new_offset;
>>> +        }
>>> +
>>> +        ret = bdrv_flush(bs);
>>> +        if (ret < 0) {
>>> +            goto fail;
>>> +        }
>>> +    }
>>> +    s->bitmap_directory_offset = new_offset;
>>> +    s->bitmap_directory_size = new_size;
>>> +    s->nb_bitmaps = new_nb_bitmaps;
>>> +
>>> +    ret = update_header_sync(bs);
>>> +    if (ret < 0) {
>>> +        goto fail;
>>> +    }
>>> +
>>> +    if (old_size) {
>>> +        qcow2_free_clusters(bs, old_offset, old_size,
>>> QCOW2_DISCARD_ALWAYS);
>>> +    }
>>> +
>>> +    return 0;
>>> +
>>> +fail:
>>> +    if (new_offset > 0) {
>>> +        qcow2_free_clusters(bs, new_offset, new_size,
>>> QCOW2_DISCARD_ALWAYS);
>>> +        s->bitmap_directory_offset = old_offset;
>>> +        s->bitmap_directory_size = old_size;
>>> +        s->nb_bitmaps = old_nb_bitmaps;
>>> +        s->autoclear_features = old_autocl;
>> What if bdrv_flush() in update_header_sync() failed? Then you cannot be
>> sure what bitmap directory the header is actually pointing to. In that
>> case, it would be wrong to assume it's still pointing to the old one and
>> freeing the new one.
>>
>> Max
> 
> Hmm.. Good point. Any suggestions?
> 
> I'm trying to achieve some kind of transaction - if function failed file
> is not changed.. But now I understand that because of flush ability to
> fail I can't achieve this..
> 
> Has write and flush any guaranties in case of fail? Would it be
> old-or-new state, or some mixed state is possible?

Anything is possible, I think. However, if flushing the data fails, it
still means writing it was successful -- you just don't know whether it
has been written to the disk successfully.

The most correct way out would be to signal image corruption, but that
isn't what we should do, I think.

If the write operation was successful but flushing was not, I'd always
assume the data hopefully gets flushed later and continue as if it had
been written to disk successfully. Therefore, I think it would be best
for update_header_sync() to just ignore bdrv_flush()'s return value.

Max


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 480 bytes --]

  reply	other threads:[~2016-10-21 19:56 UTC|newest]

Thread overview: 100+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2016-09-30 10:53 [Qemu-devel] [PATCH v7 00/22] qcow2: persistent dirty bitmaps Vladimir Sementsov-Ogievskiy
2016-09-30 10:53 ` [Qemu-devel] [PATCH 01/22] hbitmap: improve dirty iter Vladimir Sementsov-Ogievskiy
2016-10-01 13:52   ` Max Reitz
2016-09-30 10:53 ` [Qemu-devel] [PATCH 02/22] tests: add hbitmap iter test Vladimir Sementsov-Ogievskiy
2016-10-01 14:02   ` Max Reitz
2016-09-30 10:53 ` [Qemu-devel] [PATCH 03/22] block: fix bdrv_dirty_bitmap_granularity signature Vladimir Sementsov-Ogievskiy
2016-09-30 10:53 ` [Qemu-devel] [PATCH 04/22] block/dirty-bitmap: add deserialize_ones func Vladimir Sementsov-Ogievskiy
2016-09-30 10:53 ` [Qemu-devel] [PATCH 05/22] qcow2-bitmap: structs and consts Vladimir Sementsov-Ogievskiy
2016-10-01 14:34   ` Max Reitz
2016-10-01 14:56     ` Max Reitz
2016-10-07 13:11     ` Vladimir Sementsov-Ogievskiy
2016-10-11 11:50     ` Vladimir Sementsov-Ogievskiy
2016-10-12 18:20       ` Max Reitz
2016-09-30 10:53 ` [Qemu-devel] [PATCH 06/22] qcow2: add dirty bitmaps extension Vladimir Sementsov-Ogievskiy
2016-10-01 14:46   ` Max Reitz
2016-10-11 12:09     ` Vladimir Sementsov-Ogievskiy
2016-10-12 18:21       ` Max Reitz
2016-10-13 12:18         ` Vladimir Sementsov-Ogievskiy
2016-09-30 10:53 ` [Qemu-devel] [PATCH 07/22] qcow2-bitmap: introduce auto-loading bitmaps Vladimir Sementsov-Ogievskiy
2016-10-01 16:26   ` Max Reitz
2016-10-14 18:44     ` Vladimir Sementsov-Ogievskiy
2016-10-15 17:03       ` Max Reitz
2016-10-15 17:22         ` Vladimir Sementsov-Ogievskiy
2016-10-20 12:22     ` Vladimir Sementsov-Ogievskiy
2016-10-21 19:49       ` Max Reitz
2016-10-07 19:25   ` Max Reitz
2016-10-21 11:59     ` Vladimir Sementsov-Ogievskiy
2016-10-21 19:56       ` Max Reitz [this message]
2016-09-30 10:53 ` [Qemu-devel] [PATCH 08/22] block/dirty-bitmap: add autoload field to BdrvDirtyBitmap Vladimir Sementsov-Ogievskiy
2016-10-07 17:05   ` Max Reitz
2016-09-30 10:53 ` [Qemu-devel] [PATCH 09/22] block: introduce persistent dirty bitmaps Vladimir Sementsov-Ogievskiy
2016-10-07 17:54   ` Max Reitz
2016-10-11 13:11     ` Vladimir Sementsov-Ogievskiy
2016-10-12 18:24       ` Max Reitz
2016-10-07 19:28   ` Max Reitz
2016-10-12 11:38     ` Vladimir Sementsov-Ogievskiy
2016-10-12 12:30       ` Vladimir Sementsov-Ogievskiy
2016-10-12 18:25         ` Max Reitz
2016-09-30 10:53 ` [Qemu-devel] [PATCH 10/22] block/dirty-bitmap: add bdrv_dirty_bitmap_next() Vladimir Sementsov-Ogievskiy
2016-10-07 18:11   ` Max Reitz
2016-09-30 10:53 ` [Qemu-devel] [PATCH 11/22] qcow2-bitmap: add qcow2_store_persistent_bitmaps() Vladimir Sementsov-Ogievskiy
2016-10-07 19:24   ` Max Reitz
2016-10-13 16:48     ` Vladimir Sementsov-Ogievskiy
2016-10-15 16:40       ` Max Reitz
2016-10-17 17:19     ` Vladimir Sementsov-Ogievskiy
2016-10-21 19:44       ` Max Reitz
2016-10-21 21:04         ` Eric Blake
2016-10-17 17:57   ` Vladimir Sementsov-Ogievskiy
2016-10-17 17:58     ` [Qemu-devel] DROP THIS " Vladimir Sementsov-Ogievskiy
2016-09-30 10:53 ` [Qemu-devel] [PATCH 12/22] qcow2-bitmap: add IN_USE flag Vladimir Sementsov-Ogievskiy
2016-10-07 19:44   ` Max Reitz
2016-10-21 15:34     ` Vladimir Sementsov-Ogievskiy
2016-10-21 19:58       ` Max Reitz
2016-10-24 10:32         ` Vladimir Sementsov-Ogievskiy
2016-10-24 11:35           ` Vladimir Sementsov-Ogievskiy
2016-10-24 17:08             ` Max Reitz
2016-10-24 17:18               ` Max Reitz
2016-10-25 10:53                 ` Vladimir Sementsov-Ogievskiy
2016-10-26  9:04                   ` Vladimir Sementsov-Ogievskiy
2016-10-26  9:21                     ` Vladimir Sementsov-Ogievskiy
2016-10-26 12:13                       ` Vladimir Sementsov-Ogievskiy
2016-10-26 13:02                         ` Vladimir Sementsov-Ogievskiy
2016-10-26 15:28                     ` Max Reitz
2016-11-07 16:12                   ` Vladimir Sementsov-Ogievskiy
2016-11-07 16:18                     ` Max Reitz
2016-10-24 16:54           ` Max Reitz
2016-09-30 10:53 ` [Qemu-devel] [PATCH 13/22] qcow2-bitmap: check constraints Vladimir Sementsov-Ogievskiy
2016-10-07 19:54   ` Max Reitz
2016-09-30 10:53 ` [Qemu-devel] [PATCH 14/22] qcow2: delete bitmaps on truncate Vladimir Sementsov-Ogievskiy
2016-10-07 19:58   ` Max Reitz
2016-09-30 10:53 ` [Qemu-devel] [PATCH 15/22] qcow2-bitmap: add autoclear bit Vladimir Sementsov-Ogievskiy
2016-10-07 20:11   ` Max Reitz
2016-10-24 14:25     ` Vladimir Sementsov-Ogievskiy
2016-10-24 17:21       ` Max Reitz
2016-09-30 10:53 ` [Qemu-devel] [PATCH 16/22] qmp: add persistent flag to block-dirty-bitmap-add Vladimir Sementsov-Ogievskiy
2016-10-07 19:52   ` Eric Blake
2016-10-24 14:44     ` Vladimir Sementsov-Ogievskiy
2016-10-10 16:08   ` Max Reitz
2016-10-24 15:12     ` Vladimir Sementsov-Ogievskiy
2016-10-24 17:30       ` Max Reitz
2016-10-25 11:05         ` Vladimir Sementsov-Ogievskiy
2016-09-30 10:53 ` [Qemu-devel] [PATCH 17/22] qmp: add autoload parameter " Vladimir Sementsov-Ogievskiy
2016-10-07 19:53   ` Eric Blake
2016-10-10 16:25   ` Max Reitz
2016-10-24 15:55     ` Vladimir Sementsov-Ogievskiy
2016-09-30 10:53 ` [Qemu-devel] [PATCH 18/22] qapi: add md5 checksum of last dirty bitmap level to query-block Vladimir Sementsov-Ogievskiy
2016-10-10 16:44   ` Max Reitz
2016-10-10 17:03     ` Max Reitz
2016-10-10 19:22       ` Eric Blake
2016-09-30 10:53 ` [Qemu-devel] [PATCH 19/22] iotests: test qcow2 persistent dirty bitmap Vladimir Sementsov-Ogievskiy
2016-10-10 17:04   ` Max Reitz
2016-09-30 10:53 ` [Qemu-devel] [PATCH 20/22] qcow2-dirty-bitmap: refcounts Vladimir Sementsov-Ogievskiy
2016-10-10 17:59   ` Max Reitz
2016-09-30 10:53 ` [Qemu-devel] [PATCH 21/22] specs/qcow2: fix bitmap granularity qemu-specific note Vladimir Sementsov-Ogievskiy
2016-10-07 20:18   ` Eric Blake
2016-11-09 16:43     ` Vladimir Sementsov-Ogievskiy
2016-09-30 10:53 ` [Qemu-devel] [PATCH 22/22] specs/qcow2: do not use wording 'bitmap header' Vladimir Sementsov-Ogievskiy
2016-10-07 20:20   ` Eric Blake
2016-10-01 13:37 ` [Qemu-devel] [PATCH v7 00/22] qcow2: persistent dirty bitmaps Max Reitz
2016-10-13 18:11   ` John Snow

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=524e8350-025f-6ab0-763c-447fdaf72277@redhat.com \
    --to=mreitz@redhat.com \
    --cc=armbru@redhat.com \
    --cc=den@openvz.org \
    --cc=eblake@redhat.com \
    --cc=famz@redhat.com \
    --cc=jsnow@redhat.com \
    --cc=kwolf@redhat.com \
    --cc=pbonzini@redhat.com \
    --cc=qemu-block@nongnu.org \
    --cc=qemu-devel@nongnu.org \
    --cc=stefanha@redhat.com \
    --cc=vsementsov@virtuozzo.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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.