From: Kevin Wolf <kwolf@redhat.com>
To: qemu-block@nongnu.org
Cc: kwolf@redhat.com, qemu-devel@nongnu.org
Subject: [Qemu-devel] [PULL 03/31] block: Let bdrv_open_inherit() return the snapshot
Date: Wed, 25 May 2016 19:39:28 +0200 [thread overview]
Message-ID: <1464197996-4581-4-git-send-email-kwolf@redhat.com> (raw)
In-Reply-To: <1464197996-4581-1-git-send-email-kwolf@redhat.com>
From: Max Reitz <mreitz@redhat.com>
If bdrv_open_inherit() creates a snapshot BDS and *pbs is NULL, that
snapshot BDS should be returned instead of the BDS under it.
This has worked so far because (nearly) all users of BDRV_O_SNAPSHOT use
blk_new_open() to create the BDS tree. bdrv_append() (which is called by
bdrv_append_temp_snapshot()) redirects pointers from parents (i.e. the
BB in this case) to the newly appended child (i.e. the overlay),
therefore, while bdrv_open_inherit() did not return the root BDS, the BB
still pointed to it.
The only instance where BDRV_O_SNAPSHOT is used but blk_new_open() is
not is in blockdev_init() if no BDS tree is created, and instead
blk_new() is used and the flags are stored in the BB root state.
However, qmp_blockdev_change_medium() filters the BDRV_O_SNAPSHOT flag
before invoking bdrv_open(), so it will not have any effect.
In any case, it would be nicer if bdrv_open_inherit() could just always
return the root of the BDS tree that has been created.
To this end, bdrv_append_temp_snapshot() now returns the snapshot BDS
instead of just appending it on top of the snapshotted BDS. Also, it
calls bdrv_ref() before bdrv_append() (which bdrv_open_inherit() has to
undo if not returning the overlay).
Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
---
block.c | 47 ++++++++++++++++++++++++++++++++++++++++-------
1 file changed, 40 insertions(+), 7 deletions(-)
diff --git a/block.c b/block.c
index f85c5a2..17ee088 100644
--- a/block.c
+++ b/block.c
@@ -1422,8 +1422,10 @@ done:
return c;
}
-static int bdrv_append_temp_snapshot(BlockDriverState *bs, int flags,
- QDict *snapshot_options, Error **errp)
+static BlockDriverState *bdrv_append_temp_snapshot(BlockDriverState *bs,
+ int flags,
+ QDict *snapshot_options,
+ Error **errp)
{
/* TODO: extra byte is a hack to ensure MAX_PATH space on Windows. */
char *tmp_filename = g_malloc0(PATH_MAX + 1);
@@ -1439,7 +1441,6 @@ static int bdrv_append_temp_snapshot(BlockDriverState *bs, int flags,
/* Get the required size from the image */
total_size = bdrv_getlength(bs);
if (total_size < 0) {
- ret = total_size;
error_setg_errno(errp, -total_size, "Could not get image size");
goto out;
}
@@ -1479,12 +1480,19 @@ static int bdrv_append_temp_snapshot(BlockDriverState *bs, int flags,
goto out;
}
+ /* bdrv_append() consumes a strong reference to bs_snapshot (i.e. it will
+ * call bdrv_unref() on it), so in order to be able to return one, we have
+ * to increase bs_snapshot's refcount here */
+ bdrv_ref(bs_snapshot);
bdrv_append(bs_snapshot, bs);
+ g_free(tmp_filename);
+ return bs_snapshot;
+
out:
QDECREF(snapshot_options);
g_free(tmp_filename);
- return ret;
+ return NULL;
}
/*
@@ -1704,17 +1712,42 @@ static int bdrv_open_inherit(BlockDriverState **pbs, const char *filename,
}
QDECREF(options);
- *pbs = bs;
/* For snapshot=on, create a temporary qcow2 overlay. bs points to the
* temporary snapshot afterwards. */
if (snapshot_flags) {
- ret = bdrv_append_temp_snapshot(bs, snapshot_flags, snapshot_options,
- &local_err);
+ BlockDriverState *snapshot_bs;
+ snapshot_bs = bdrv_append_temp_snapshot(bs, snapshot_flags,
+ snapshot_options, &local_err);
snapshot_options = NULL;
if (local_err) {
+ ret = -EINVAL;
goto close_and_fail;
}
+ if (!*pbs) {
+ /* We are not going to return bs but the overlay on top of it
+ * (snapshot_bs); thus, we have to drop the strong reference to bs
+ * (which we obtained by calling bdrv_new()). bs will not be
+ * deleted, though, because the overlay still has a reference to it.
+ */
+ bdrv_unref(bs);
+ bs = snapshot_bs;
+ } else {
+ /* We are not going to return snapshot_bs, so we have to drop the
+ * strong reference to it (which was returned by
+ * bdrv_append_temp_snapshot()). snapshot_bs will not be deleted,
+ * though, because bdrv_append_temp_snapshot() made all parental
+ * references to bs (*pbs) point to snapshot_bs.
+ * In fact, if *pbs was not NULL, we are not going to return any new
+ * BDS. But we do not need to decrement bs's refcount here as is
+ * done above, because with a non-NULL *pbs this function never even
+ * had a strong reference to bs. */
+ bdrv_unref(snapshot_bs);
+ }
+ }
+
+ if (!*pbs) {
+ *pbs = bs;
}
return 0;
--
1.8.3.1
next prev parent reply other threads:[~2016-05-25 17:40 UTC|newest]
Thread overview: 34+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-05-25 17:39 [Qemu-devel] [PULL 00/31] Block layer patches Kevin Wolf
2016-05-25 17:39 ` [Qemu-devel] [PULL 01/31] block: Fix bdrv_next() memory leak Kevin Wolf
2016-06-06 8:41 ` Paolo Bonzini
2016-05-25 17:39 ` [Qemu-devel] [PULL 02/31] block: Drop useless bdrv_new() call Kevin Wolf
2016-05-25 17:39 ` Kevin Wolf [this message]
2016-05-25 17:39 ` [Qemu-devel] [PULL 04/31] tests: Drop BDS from test-throttle.c Kevin Wolf
2016-05-25 17:39 ` [Qemu-devel] [PULL 05/31] block: Drop blk_new_with_bs() Kevin Wolf
2016-05-25 17:39 ` [Qemu-devel] [PULL 06/31] block: Drop bdrv_new_root() Kevin Wolf
2016-05-25 17:39 ` [Qemu-devel] [PULL 07/31] block: Make bdrv_open() return a BDS Kevin Wolf
2016-05-25 17:39 ` [Qemu-devel] [PULL 08/31] block: Assert !bs->refcnt in bdrv_close() Kevin Wolf
2016-05-25 17:39 ` [Qemu-devel] [PULL 09/31] block: Drop bdrv_parent_cb_...() from bdrv_close() Kevin Wolf
2016-05-25 17:39 ` [Qemu-devel] [PULL 10/31] block: Drop errp parameter from blk_new() Kevin Wolf
2016-05-25 17:39 ` [Qemu-devel] [PULL 11/31] block: Introduce bdrv_replace_child() Kevin Wolf
2016-05-25 17:39 ` [Qemu-devel] [PULL 12/31] block: Make bdrv_drain() use bdrv_drained_begin/end() Kevin Wolf
2016-05-25 17:39 ` [Qemu-devel] [PULL 13/31] block: Fix reconfiguring graph with drained nodes Kevin Wolf
2016-05-25 17:39 ` [Qemu-devel] [PULL 14/31] block: Propagate .drained_begin/end callbacks Kevin Wolf
2016-05-25 17:39 ` [Qemu-devel] [PULL 15/31] dma-helpers: change interface to byte-based Kevin Wolf
2016-05-25 17:39 ` [Qemu-devel] [PULL 16/31] dma-helpers: change BlockBackend to opaque value in DMAIOFunc Kevin Wolf
2016-05-25 17:39 ` [Qemu-devel] [PULL 17/31] block: Rename blk_write_zeroes() Kevin Wolf
2016-05-25 17:39 ` [Qemu-devel] [PULL 18/31] block: keep a list of block jobs Kevin Wolf
2016-05-25 17:39 ` [Qemu-devel] [PULL 19/31] block: Cancel jobs first in bdrv_close_all() Kevin Wolf
2016-05-25 17:39 ` [Qemu-devel] [PULL 20/31] block: Default to enabled write cache in blk_new() Kevin Wolf
2016-05-25 17:39 ` [Qemu-devel] [PULL 21/31] block: Convert block job core to BlockBackend Kevin Wolf
2016-05-25 17:39 ` [Qemu-devel] [PULL 22/31] block: Make blk_co_preadv/pwritev() public Kevin Wolf
2016-05-25 17:39 ` [Qemu-devel] [PULL 23/31] stream: Use BlockBackend for I/O Kevin Wolf
2016-05-25 17:39 ` [Qemu-devel] [PULL 24/31] mirror: Allow target that already has a BlockBackend Kevin Wolf
2016-05-25 17:39 ` [Qemu-devel] [PULL 25/31] mirror: Use BlockBackend for I/O Kevin Wolf
2016-05-25 17:39 ` [Qemu-devel] [PULL 26/31] backup: Don't leak BackupBlockJob in error path Kevin Wolf
2016-05-25 17:39 ` [Qemu-devel] [PULL 27/31] backup: Pack Notifier within BackupBlockJob Kevin Wolf
2016-05-25 17:39 ` [Qemu-devel] [PULL 28/31] backup: Remove bs parameter from backup_do_cow() Kevin Wolf
2016-05-25 17:39 ` [Qemu-devel] [PULL 29/31] backup: Use BlockBackend for I/O Kevin Wolf
2016-05-25 17:39 ` [Qemu-devel] [PULL 30/31] commit: " Kevin Wolf
2016-05-25 17:39 ` [Qemu-devel] [PULL 31/31] blockjob: Remove BlockJob.bs Kevin Wolf
2016-05-26 14:06 ` [Qemu-devel] [PULL 00/31] Block layer patches Peter Maydell
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=1464197996-4581-4-git-send-email-kwolf@redhat.com \
--to=kwolf@redhat.com \
--cc=qemu-block@nongnu.org \
--cc=qemu-devel@nongnu.org \
/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).