| commit | ad343101c59beb0c2b0c1033f031ff76ff2237d7 | [log] [tgz] |
|---|---|---|
| author | Andrew Jeffery <andrew@aj.id.au> | Wed Feb 28 00:35:50 2018 +1030 |
| committer | Andrew Jeffery <andrew@aj.id.au> | Wed Apr 04 00:31:37 2018 +0930 |
| tree | 28c21f3cbf47ce720dd1e006bfdff7c52e45eb8d | |
| parent | fc62158caeaef009cd0055ce8cc35586581840b8 [diff] |
pnor_partition: Refactor to enforce stronger boundaries for abstractions
The RORequest and RWRequest classes did not provide a clear abstraction
over the operation of populating a window or partition associated with a
CREATE_{READ,WRITE}_WINDOW request. The role of the classes was to find
the partition for the provided offset, locate and then open its backing
file.
However, the file-descriptor for the backing file was exposed outside of
the class, as was the FFS partition struct, both of which were managed
_internal_ to the class. Thus the classes provided no encapsulation of
state and awkwardly split the tasks of managing and utilising the
resources between the callee and caller.
This commit inverts the behaviour in a fulfil() method handles the
mechanics of locating, opening, manipulating and closing the backing
file, requiring nothing of the caller. The pnor_partition reference is
managed entirely inside the Request class, derived from the offset
passed to the constructor.
Unifying the mechanics into fulfil() results in a decent reduction in
lines of code at the expense of some cyclomatic complexity. fulfil() is
publicly exposed via read() and write() wrappers on the class, and the
RORequest and RWRequest classes are removed as a result.
Change-Id: Ie587ed31f1a6db97bcb490dbcc2e27ece0b1f82c
Signed-off-by: Andrew Jeffery <andrew@aj.id.au>
Copyright 2017 IBM
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.
This repo contains the protocol definition for the host to BMC mailbox communication specification which can be found in Documentation/mbox_procotol.md.
There is also a reference implementation of a BMC mailbox daemon, the details of which can be found in Documentation/mboxd.md.
Finally there is also an implementation of a mailbox daemon control program, the details of which can be found in Documentation/mboxctl.md.
This codebase is a mix of C (due to its heritage) and C++. This is an ugly split: message logging and error handling can be vastly different inside the same codebase. The aim is to remove the split one way or the other over time and have consistent approaches to solving problems.
phosphor-mboxd is developed as part of the OpenBMC project, which also leads to integration of frameworks such as phosphor-logging. Specifically on phosphor-logging, it's noted that without care we can achieve absurd duplication or irritating splits in where errors are reported, as the C code is not capable of making use of the interfaces provided.
Message logging MUST be done to stdout or stderr, and MUST NOT be done directly via journal APIs or wrappers of the journal APIs.
Rationale:
We have two scenarios where we care about output, with the important restriction that the method must be consistent between C and C++:
In the first case it is desirable that the messages appear in the system journal. To this end, systemd will by default capture stdout and stderr of the launched binary and redirect it to the journal.
In the second case it is desirable that messages be captured by the test runner (make check) for test failure analysis, and it is undesirable for messages to appear in the system journal (as these are tests, not issues affecting the health of the system they are being executed on).
Therefore direct calls to the journal MUST be avoided for the purpose of message logging.
Note: This section specifically targets the use of phosphor-logging's log<T>(). It does not prevent the use of elog<T>().