normalize_image_name(): fix for slash-ended basedirAfter commit f11090d, when basedir already ends with '/' character,
this function stops working correctly.
Ideally, we have to check if basedir ends with slash, add it if missing
then compare the beginning of image file name with that string.
Practically, that would mean copying the basedir string over to
some new place, adding slash, etc.
So, we'd better compare the strings without the slash...
snapshot-merge: fix/improve message about deltasI just noticed that ploop snapshot-merge with -n prints
somewhat confusing information about merged deltas, i.e.
it always prints "child_delta -> parent_delta", although
in case of "-n new_delta" it shoulld be something like
"child_delta parent_delta -> new_delta".
Also, all delta files are printed as absolute paths,
making the output a bit cluttered. This patch makes use of
normalize_image_na...
ploop(8): move restore-descriptor description... to the appropriate subsection (under "Misc commands").
Previously it was in "Ballooning" subsection.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
ploop_replace_image(): fix wrt levelThe code very incorrectly assumed that dd.xml idx == level.
This is surely not true in case dd.xml contains non-linear
chain of snapshots/deltas.
Fix by treating index and level separately. While at it,
also enable the case of replacing an unused delta on a
running ploop device.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
find_image_idx_by_file(): normalize imageInstead of requiring from a caller to normalize image file name,
let the function itself do it.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
ploop_get_delta_attr_str(): addThis is currently only used to get file name of image file
for a given level.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
find_level_by_delta(): rework/improveFirst and most important change in find_delta_by_level() is it now
compares files by the device and inode numbers, not by names.
Also:
* return SYSEXIT_* codes
* found level is returned via last argument
This way we can simplify its callers a bit.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Return back multi-level snapshot mergeMulti-level snapshot merge functionality was always available:
1. merge_image() function -- but it is not part of public ploop lib API
2. ploop merge with multiple deltas or levels -- but it is not using
DiskDescriptor.xml
This commit brings back the multi-level merge functionality.
Now, if ploop_merge_param.guid2 is set, ploop_merge_snapshot()
tries to merge all snapshots in the guid...
ploop_merge_snapshot_by_guid(): fix error messageFix message "Unable to merge snapshot %s: it has %d children":
* merge snapshot -> merge to snapshot
* Unable to -> Can't
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
ploop_di_merge_image: fix error messages1. Message "Unable to merge snapshot %s: it has %d children"
is wrong, because:
- it talks about guid but it should be parent guid
- it says "merge snapshot" but we are not merging it --
we are merging TO it.
Note this message can never be seen because we do the very same check
earlier in ploop_merge_snapshot_by_guid().
2. Message "Unable to merge image %s" is wrong,
it sh...
ploop(8): expand description of snapshot-mergeThere is a certain incomprehension in what is the difference between
snapshot-delete and snapshot-merge. This patch tries to improve
snapshot-merge description by telling what exactly is going on during
this operation. Note that snapshot-delete is already described.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
check_snapshot_mount(): add is_device_inuse() checkTemporary snapshots are supposed to be removed as long as they are
not in use anymore. There is a way to mark the temporary snapshot
as being used -- open its corresponding ploop device (see holder_fd
parameter to ploop_create_temporary_snapshot()). Other places dealing
with temporary snapshots check that the snapshot is not in use (see
merge_temporary_snapshots() that calls is_snapshot_in_use(...
Rework check_snapshot_mount() v2This commit tries to make check_snapshot_mount() easier.
The motivation behind the change is it was used incorrectly
from do_delete_snapshot().
In do_delete_snapshot() we only want to check one snapshot
and we don't care about the parent one. So, let's simplify
it to only check one snapshot.
In ploop_merge_snapshot_by_guid() we want to check
two snapshots -- the parent one and the child one. ...
ploop_merge_snapshot_by_guid: fix ending messageWe should always print the name of the parent snapshot,
i.e. the one that was merged into a child one and which
UUID remains. Otherwise it's not clear which snapshot
was merged into which.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Introduce make_images_list_by_guids()1. Modify existing make_images_list() function, adding the end_guid
parameter. If end_guid == NULL it works exactly as before.
2. Rename it to make_images_list_by_guids()
3. Add make_images_list() wrapper for compatibility
4. Document the new function.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
ploop_merge_snapshot_by_guid(): untangleUntangle the PLOOP_MERGE_WITH_PARENT/PLOOP_MERGE_WITH_CHILD mess
so now it is more readable and shorter (no duplicated code).
The only downside is we can no longer distinguish between
"no parent guid found" and "can't merge base image" errors,
so let's modify the error message accordingly.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
mv get_child_by_uuid -> find_child_by_guidRename
ploop_get_child_by_uuid()
to
ploop_find_child_by_guid()
This is done for readability and uniformity with
ploop_find_parent_by_guid()
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
ploop_di_merge_image(): remove last argumentThis function is only called from ploop_merge_snapshot_by_guid(),
and in there we already know the file name corresponds to child_guid.
Remove it, making the code just a little bit less cryptic.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
ploop tool: usage formatting fixes1. Unify parameters presentation -- should be uppercase and unquoted,
e.g. UUID (not uuid, not <uuid).
2. When describing a parameter, use "NAME := description" style.
3. Don't use tabs inside strings printed by usage_*() functions.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
ploop tsnapshot(): fix usageWrong usage (from snapshot not tsnapshot) was printed in case of
invalid command line.
This is a fix to commit 785f4da.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
ploop_replace_image(): rm param NULL checkLet's not check param argument for being NULL, as
1 It should never be NULL anyway
2. Since commit fafab72 we are dereferencing it before doing this check
(in keep assignment) so the check is totally useless now.
This is an addition to commit fafab72.
Reported by Coverity, CID 1287120.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
ploop_umount(): don't return -1, use SYSEXIT_*The return value from this function is returned by a few functions
that expect to return SYSEXIT_* values, such as ploop_delete_snapshot()
and ploop_merge_snapshot().
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
ploop replace: implement offline replaceAn offline replace is just a way to replace a file name in
DiskDescriptor.xml with some other file name. In this mode,
same checks are performed in a new delta. Flag --keep-name
is also supported, in such case ploop replace works like
"mv -f" but with bells and whistles (all the checks etc.)
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
ploop_replace_image(): optimization for keep_nameIn case KEEP_NAME flag is set, we need to rename the new image file
to the current (old) one. But as in-kernel ploop driver doesn't allow
to remove (or rename(), as it comes out) a delta file which is in use
unless its hardlink count is > 1, we have to make a trick here:
- create a hardlink to the file
- rename it
- remove the hardlink
It just occured to me we can avoid doing this trick if ...
ploop replace: add long option --keep-nameThis is the same as recently introduced -k.
Man page and usage() modified accordingly.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
ploop_replace_image(): always unlock dd.xmlSpotted a few cases when we return error right away, without unlocking
DiskDescriptor.xml. Fix those.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
normalize_image_name(): fix for corner caseChances to hit this bug are pretty slim, but I just did it.
For the following input parameters:
basedir="/vz/private/104/root.hdd"
image="/vz/private/104/root.hdd.{86be108e-d30e-4054-b6d4-412d4d0c2ce6}"
function errorneously returned
{86be108e-d30e-4054-b6d4-412d4d0c2ce6}
Apparently, we need to add a check for '/' right after basedir. Note
we do this check right after strncmp which retu...
setver.sh: also set version in Makefile.incAs per previous commit, now we need to set version (and, optionally,
release) in Makefile.inc. Logic is similar to that for the spec file,
aside the special treating for RELEASE variable:
- release of "1" becomes empty
- non-empty release is prepended with a dash
Fix trap statements accordingly, due to more complex rollback procedure.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Makefile.inc: specify package versionDue to a new build environment we won't use a spec file from the
same repository as the sources, so we can't rely on .spec file
to hold the version of the package. Therefore, let's specify
package version (and optional RELEASE which should usually be empty)
in Makefile.inc which is included into other Makefiles.
lib/Makefile determines library version from the package version the
same way as b...
ploop_replace_image(): add check for keep_nameWhen KEEP_NAME flag is set, do a check that a file could be
renamed to old file. to error out early if it couldn't.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
ploop replace: implement -kCurrently, replacing a delta means changing its file name. This commit
adds a flag to keep the same file name. If this flag is set, after
successful replace a new delta is renamed to the old one (removing it!).
This makes it possible to replace a delta without modifying dd.xml
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
[style] ploop_replace_image: reuse oldfile varDo not mess with oldfile variable while writing conf_tmp dd.xml file.
Use oldfile when printing "Replacing ..." log message, so that oldfile
will look exactly like specified from the command line (if specified).
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
setting library version from ploop.speclets do not use ploop.spec to set library version
* spec is not part of API
* spec is not stored in same sources tree
Signed-off-by: Igor Sukhih <igor@parallels.com>
print_sys_block_ploop(): compile fixThis is a fix to commit 6d00f65 -- I accidentally commited old version
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
ploop merge: obsolete form with DD.xmlIt's not quite clear how ploop tool ended up with two commands doing the
same thing, but it did:
- ploop merge <dd.xml>
- ploop snapshot-merge <dd.xml>
are doing the same thing.
Now, "ploop merge" was originally a low-level ploop command, while
"ploop snapshot-merge" is a form to be used with dd.xml.
Let's make it this way to avoid user confusion and code repetition.
Now, if ploop merge is...
Remove delta_fopsThis came from a prehistoric implementation. Currently we only
use standard filesystems operations (and do not use update_size() op),
so this abstraction layer is no longer required.
Acked-by: Igor Sukhih <igor@parallels.com>
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
print_sys_block_ploop(): print less infoThis function is called when EBUSY is returned from ADD_DELTA
or REPLACE_DELTA ploop ioctl, in order to figure out why delta
is busy (presumably it is used by another ploop device).
As we don't need to print the whole contents of /sys/block/ploop*,
but rather some specific info from pdelta. According to Maxim Patlasov,
having pdelta/{image,ro,io} is enough.
This helps to reduce the output, as...
Kir Kolyshkin
Igor Sukhih