When FRR starts, after mgmtd is initialized, backend clients connect to
it and request their config. To supply the config, mgmtd creates a
configuration transaction. At the same time, `vtysh -b` tries to read
the startup config and configure mgmtd, which also creates a
configuration transaction. If these two actions happen at the exact same
time, there's a conflict between them, because only a single
configuration translaction is allowed. Because of that, vtysh fails and
the config is completely ignored.
When starting the config reading, vtysh locks candidate and running
datastores in mgmtd. This commit adds locking of running datastore when
initializing the backend client. It allows to retry locking on the vtysh
side and read the config only when the lock is aquired instead of
failing.
This change also prevents running datastore from being changed during
initialization of backend clients. This could lead to a desynchronized
state between mgmtd and backends.
Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>
When no data requests were sent to the backends, return immediately,
instead of waiting for a timeout. This can happen if backends providing
the requested data are not connected to mgmtd.
Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>
There are places, where we can receive an existing commit transaction.
If we don't check that the request already exists, it gets overwritten
and we start having problems with transaction refcounters. Forbid having
multiple configuration sessions simultaneously.
Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>
If the transaction is not cleaned up immediately, it can be still
referenced by some threds. If it's a commit thread and it's executed
before the actual cleanup, mgmtd crashes because of the missing
commit_cfg_req.
Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>
The value should not be cleared after sending it to the first client,
otherwise other clients receive an empty string.
Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>
Like in RESTCONF GET request and NETCONF get-data request, make it
possible to request state-only, config-only, or all data.
Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>
next_phase is always curr_phase + 1. It's not necessary to maintain it
separately. Also rename curr_phase to phase.
Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>
When determining the interested backend clients for a configuration
change, don't consider disconnected clients. This fixes a crash in
`mgmt_txn_send_be_txn_create` when trying to send data to a non-existing
adapter.
Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>
Replace operation removes the current data node configuration and sets
the provided value. As current northbound code works only with one
xpath at a time, the operation only makes sense to clear the config of
a container without deleting it itself. However, the next step is to
allow passing JSON-encoded complex values to northbound operations which
will make replace operation much more useful.
Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>
Currently, there's a single operation type which doesn't return error
if the object doesn't exists. To be compatible with NETCONF/RESTCONF,
we should support differentiate between DELETE (fails when object
doesn't exist) and REMOVE (doesn't fail if the object doesn't exist).
Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>
Currently, there's no difference between CREATE and MODIFY operations.
To be compatible with NETCONF/RESTCONF, add new CREATE_EXCL operation
that throws an error if the configuration data already exists.
Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>
Currently, nb_operation enum means two different things - edit operation
type (frontend part), and callback type (backend part). These types
overlap, but they are not identical. We need to add more operation
types to support NETCONF/RESTCONF integration, so it's better to have
separate enums to identify different entities.
Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>
Add configure.ac tests for libyang functions, if not present supply the
functionality ourselves in yang.[ch]
Signed-off-by: Christian Hopps <chopps@labn.net>
Allow user to specify full YANG compatible XPath 1.0 predicates. This
allows for trimming results of generic queries using functions and other
non-key predicates from XPath 1.0
Signed-off-by: Christian Hopps <chopps@labn.net>
Adds the guts of the get-tree functionality that is called by or calls
the FE and BE code for get-tree processing.
Signed-off-by: Christian Hopps <chopps@labn.net>
The candidate yang tree should be validated before `nb_config_diff` is
called. `nb_config_diff` ignores all prohibited operations and can
provide an empty change list because of this. For example, if a user
deletes a mandatory node from the candidate datastore and tries to make
a commit, they'll receive the "No changes found to be committed!" error,
because such a change is ignored by `nb_config_diff`. Instead, mgmtd
should tell the user that their candidate datastore is not valid and
can't be commited.
Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>
At this stage, we should send a SET_CFG reply, not COMMIT_CFG reply.
Commit request is not yet initialized.
Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>
`nb_cli_apply_changes` can be called with base xpath which should be
prepended to xpaths of every change in a transaction. This base xpath is
respected by regular northbound CLI but not by mgmtd. This commit fixes
the problem.
Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>
Batch IDs are only used to verify that all messages were received and
processed by a backend. It's not necessary to do that as we use reliable
stream transport - messages can't be dropped or received out of order.
This commit also fixes possible race condition that can happen if
one backend process messages slower than other backends.
Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>
- move from client id indexed array of uints for register info
per client to a u64 bitmask.
- add bit walking FOREACH macro
Walk the client IDs whose bits are set in a mask.
Signed-off-by: Christian Hopps <chopps@labn.net>
The config is always applied fully, all batches are included. There's no
need to pass a list of applied batches as it always contains all of
them.
Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>
The code doesn't work at all. It tries to use libyang operation
metadata in a regular (not diff) data tree, and regular data trees
don't provide this data. Also, for destroy operations, it searches
for nodes in the running config, which may not have the deleted nodes
if we're not using implicit commits.
Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>
Move away from things like "lock if not locked" type code, require the
user has locked prior to geting to that point.
For now we warn if we are taking a lock we already had; however, this
should really be a failure point.
New requirements:
SETCFG -
not implicit commit - requires user has locked candidate DS and they
must unlock after
implicit commit - requires user has locked candidate and running DS
both locks will be unlocked on reply to the SETCFG
COMMITCFG -
requires user has locked candidate and running DS and they must unlock
after
rollback - this code now get both locks and then does an unlock and
early return thing on the adapter side. It needs to be un-special
cased in follow up work that would also include tests for this
functionality.
Signed-off-by: Christian Hopps <chopps@labn.net>
