In order to provide the ability to do thick provisioning of RBD images,
Image.WriteSame() can be used. The `rbd` commandline utility calls the
rbd_writesame() function when an image is created with the
`--thick-provision` option.
See-also: ceph/ceph-csi#1675
Signed-off-by: Niels de Vos <ndevos@redhat.com>
Although it seems to be stable with the current Go versions,
unsafe.Pointers that are created by applying arithmetic on nil
pointers are explicitly not allowed[1], and could potentially lead to
panics by the garbage collector. To be on the safe side, this change
goes back to minimal inline wrappers for the callback registrations,
which should be zero-cost. For the callbacks themselves fortunately
no wrappers are required, because CGo happily converts void* to
uintptr arguments. The potentially problematic VoidPtr helper is
removed again.
[1] https://golang.org/pkg/unsafe/#Pointer
Signed-off-by: Sven Anderson <sven@redhat.com>
The function is named ListMetadata but returns a map of all
the metadata assigned to the rbd image. It is named ListMetadata to
match the C librbd api call.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
Like the metadata functions were recently moved to a separate file,
move TestImageMetadata to a matching metadata_test.go file.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
As rbd.go is very long, and the get-/set-/remove- metadata functions in
rbd.go make a nice logical unit, move them to a new metadata.go file.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
GetParentInfo() currently requires the caller to specify the
string length for various attributes it returns. If the
specified string length is less than required, it can return
ERANGE error. Replacing it with a new function which avoids
this.
Signed-off-by: Mudit Agarwal <muagarwa@redhat.com>
The sparsify function can be used to make an image sparse by replacing
runs of zeros with holes. This change adds a wrapper for the basic
rbd_sparsify function and accompanying tests.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
Since the index of the registered callbacks is basically only used as
void*/unsafe.Pointer, it makes more sense and code simpler, if
the index type is uintptr instead of a simple int.
Signed-off-by: Sven Anderson <sven@redhat.com>
This patch changes the name to the as-agreed-upon naming from a call.
I forgot to fix this in the original patches before merging things.
Oops.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
Previously, the code that set up the callbacks was doing a pretty
direct translation of go-function to c-function. It turns out this
is not needed due to the way we set up the "dynamic" go callbacks
via an integer based callback "registry". Instead, we can stay within
the C layer entirely for the C callback, passing it only the index
value and then return to the go layer through a static C function
that only needs the index value.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
Corresponding fields for Parent_pool and Parent_name have been deprecated from
ceph's rbd_image_info_t stuct, removing the same from go-ceph's ImageInfo struct.
Signed-off-by: Mudit Agarwal <muagarwa@redhat.com>
Added pool stat options constants.
Also implemented GetAllPoolStats() function to introduce
the functionality of rbd_pool_stats_option_add_uint64() and rbd_pool_stats_get()
Signed-off-by: Mudit Agarwal <muagarwa@redhat.com>
Added wrappers for rbd_pool_stats_create() and rbd_pool_stats_destroy() functions.
Right now, keeping these unexported.
Signed-off-by: Mudit Agarwal <muagarwa@redhat.com>
RBD_IMAGE_OPTION_CLONE_FORMAT was added in Ceph Mimic but fix for issue #295 added
it only for Ceph Nautilus, moving it to rbd/options_mimic.go
Signed-off-by: Mudit Agarwal <muagarwa@redhat.com>
Added a wrapper for function rbd_snap_get_trash_namespace() which
returns the original name of the snapshot which was moved to Trash.
Signed-off-by: Mudit Agarwal <muagarwa@redhat.com>
Added few constants used for determining snapshot namespace.
Added wrapper for function rbd_snap_get_namespace_type() which returns namespace type for a given snapshot.
Signed-off-by: Mudit Agarwal <muagarwa@redhat.com>
Add UpdateWatch function implementing rbd_update_watch.
Add Unwatch function implementing rbd_update_unwatch.
Add a higher level Watch type to encapsulate watching
the image and the watch handle.
Add basic tests to validate the callback behavior.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
In order to avoid external dependencies on implementation details,
this change replaces RBDError with the unexported rbdError. In case
some application really needs access to the integer value, it can use
the pattern
var errno interface{ Errno() int }
if errors.As(err, errno) { ... errno.Errno() ... }
Signed-off-by: Sven Anderson <sven@redhat.com>
Added wrapper for rbd_namespace_list() function which returns the names of all the namespaces
present for a given IOContext.
Signed-off-by: Mudit Agarwal muagarwa@redhat.com
rbd_namespace_create() function creates namespace for a given iocontext.
rbd_namespace_exists() function checks whether the given namespace exists or not.
rbd_namespace_remove() function removes the given namespace.
Signed-off-by: Mudit Agarwal <muagarwa@redhat.com>
Add wrapper for rbd_pool_metadata_remove() function which removes pool
metadata associated with a given key.
Signed-off-by: Mudit Agarwal <muagarwa@redhat.com>
Add wrapper for rbd_pool_metadata_set() function which sets pool metadata
associated with the given key.
Signed-off-by: Mudit Agarwal <muagarwa@redhat.com>
Add wrapper for rbd_pool_metadata_get() function which returns pool metadata
associated with the given key.
Signed-off-by: Mudit Agarwal <muagarwa@redhat.com>
Previously, the types were named RbdImageOptions and RbdImageOption.
Now they are named ImageOptions and ImageOption.
Type aliases have been added to prevent existing code from
breaking. As these types were new in v0.2, I suspect we can drop
the aliases post-v0.3.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
Image option constants are now named ImageOptionX, without an Rbd
prefix. Aliases to the old names are provided and doc comments for both.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
A previous change improved our use of pointers such that go vet no
longer had an issue with these lines but if we ever need to change the
interaction between rbd and rados like that again we can now rely on a
single call site `cephIoctx` that will return our ceph/c type given the
public Go-language level type from the rados pkg.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
The void * is (ab)used in C for passing arbitrary data to the callback,
which in our case is a integer index. However, Go tools can not tell
this is OK and throws an error running go vet. Change the wrapper
function to take a uintptr_t and cast that to void* only in the
(unchecked) C layer.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
Moving the errutil pkg to "internal/errutil" makes errutil private-like
and accessible to only other go-ceph packages. The functions it provided
were always meant to be used only by go-ceph, this just makes it more
official. This is a breaking change but it was only available to
outside users for 1 release and it is somewhat doubtful users outside
of go-ceph would have reached for these functions.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
No functionality change. Move various snapshot related functions to
their own file. This helps logically group these related funcs in one
place.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
The DiffIterate call accepts a data structure argument containing the
parameters of the image to "diff" and a callback function. This callback
is called in the C code, making use of the recently added callbacks
helper.
The callback itself is called with the offset and length of the
differing area in the image as well as a data parameter so that
a common function can distinguish or update different data for
different calls if needed (compare to a void* in C).
Signed-off-by: John Mulligan <jmulligan@redhat.com>
Checking if there are other clients interested in changes to an RBD
image is important for certain actions. For example, the Ceph-CSI
project would not want to remove an image in case it is still mounted
somewhere.
Signed-off-by: Niels de Vos <ndevos@redhat.com>
Applications will mostly want to show names of features to their users,
and not just a few bits. With the FeatureSet functions it becomes simple
for applications to map the name of a feature to bits, and the other way
around
Signed-off-by: Niels de Vos <ndevos@redhat.com>
By splitting up the features from the main rbd.go file, it becomes
easier to support new features added by newer versions of Ceph.
This also drops the Rbd-prefix from the constants, and adds backwards
compatible references.
Signed-off-by: Niels de Vos <ndevos@redhat.com>
Add a new CloneImage that makes use of rbd_clone3 and thus behaves like
the new CreateImage function.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
Add OpenImageById and OpenImageByIdReadOnly that wrap rbd_open_by_id and
rbd_open_by_id_read_only respectively.
The added test case can not currently test trivial error conditions due
to a known bug in ceph, these tests are skipped for meanwhile.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
Add a GetId function to the Image type that will fetch the id of the rdb
image. Comes with a test case.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
Similar to recent changes for rados (df46476e) there is no good reason
for the errorno to go-style error function to be exported from the
package. Decapitalize it.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
Replace most uses of Create() in the rbd test suite with a wrapper
around CreateImage named quickCreate. quickCreate retains most of
the argument signature of Create for easier refactoring of the tests
over time. New uses of quickCreate should not be added to the suite if
you can help it.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
Update the TestImageCopy function to establish clear subsections and use
the new CreateImage and OpenImage functions.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
Throughout rbd_test.go the values of 1<<22 and 22 are repeated when
created rbd images. For the common cases, replace these with variables.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
Currently, the code only provides a .Remove function on the Image type.
But this is unnecessary as the underlying api only requires an io
context and name. Make a function that matches the underlying api
better.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
In the future, CreateImage should be the canonical way one creates an
rbd image. Like the underlying ceph apis CreateImage does not return an
Image, only an error code. Open image handles should be acquired through
the OpenImage functions.
Because Create4 was fairly recently added to library there should not be
consumers in the wild. Thus we don't retain the current version for
backwards compatibility.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
Replace uses of the deprecated img.Open(...) pattern with the new open
image functions where applicable.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
Rename test case to mark the "old" deprecated open func as such.
This also adds a missing bit of coverage to the test.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
Create new OpenImage & OpenImageReadOnly functions to replace the use of
Open() on image types. This function more closely matches the underlying
of the librbd function calls.
The third argument to the functions is a snapshot name but we also
create a new constant to clearly indicate that an image should be opened
w/o a snapshot. Internally, this also ensures that a null string is
passed to the C api.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
The case of a short read is not detected correctly. This causes Read()
to not return io.EOF where it is expected.
By comparing 'len(data)' instead of unassigned 'n' (set to 0), the
number of bytes that have been read can be compared to the number of
bytes that were requested.
Signed-off-by: Niels de Vos <ndevos@redhat.com>
The case of a short read is not detected correctly. This causes ReadAt()
to not return io.EOF where it is expected.
By comparing 'len(data)' instead of unassigned 'n' (set to 0), the
number of bytes that have been read can be compared to the number of
bytes that were requested.
Signed-off-by: Niels de Vos <ndevos@redhat.com>
In order to get a higher test coverage for rbd/rbd.go, there is a need
to trigger errors for the librbd functions. This makes it possible to
test the error path. The new TestClosedImage() function tricks
RbdImage.validate() into accepting an image that is closed, but still
references the closed image.
librbd is expected to return an error when the image is closed. However,
passing a closed image does not result in stable behavior. Sometimes
librbd causes a panic, at other times there will be a hang and the test
will time-out.
The TestClosedImage() function is disabled, it will skip all testing but
remains available so that it could be enabled in the future.
Signed-off-by: Niels de Vos <ndevos@redhat.com>
In the rbd tests the setup of the rados connection was essentially copy
and paste with very little proper error checking.
This change adds a common setup function that checks all error
conditions as well as prevents the tests from blocking forever if
something is wrong with the ability of the tests to connect.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
The arguments passed to Copy() would normally be pointers, just like it
is done with other similar functions. This never worked correctly, as
the arguments were not validated correctly (not recognized) and the
error "Must specify either destination pool or destination image" always
got returned.
Modifying Image.Copy() into something usable would result in a complete
rewrite. Instead of a single Image.Copy() function that tries to cover
rbd_copy() and rbd_copy2(), provide Image.Copy() and Image.Copy2() so
that the API matches the librbd.so interface.
Signed-off-by: Niels de Vos <ndevos@redhat.com>
The return value of rbd_discard() is the number of bytes that have been
discarded, or in case of an error, a negative errno value.
It seems Image.Discard() never can have worked as intended, unless the
error was disregarded. Just like Image.Write(), Image.Discard() should
return a tuple of the number of bytes discarded and an error.
Fixes: ceph/go-ceph#123
Signed-off-by: Niels de Vos <ndevos@redhat.com>
rbd_get_features() returns a uint64_t with the features that are set in
the image. That means combining (with logical OR) features should result
in a uint64 too. Marking all features explicitly uint64 makes it easier
and cleaner to set/test features.
Signed-off-by: Niels de Vos <ndevos@redhat.com>
Keeping the _test.go code in the same package allows for the testing
of private functions. AFAWK it is more conventional as well.
See-also: ceph/go-ceph#135
Signed-off-by: Niels de Vos <ndevos@redhat.com>
We recently created errors in the rbd that match the golang convention
but left aliases for external code. Switch the test code to use the
newer more standard form instead of the "aliases."
Signed-off-by: John Mulligan <jmulligan@redhat.com>
A recent PR added new error values that were similar to the existing
naming convention but were not the normal go convention. Change all the
errors to match typical convention but leave the older names as aliases
such that older code will continue to work (for now).
Signed-off-by: John Mulligan <jmulligan@redhat.com>
