I have some thoughts for the future:
1a. The traffic generator is specified per-node, so maybe we could also
change the binding to be for the whole lifetime of the TG node,
1b. But the same is true for the SUT node as well, right? After we do
the port update (with kernel driver), we can just bind to DPDK driver.
With SUT in the mix, this looks like a change for a different patch,
2. We could add a symlink to the devbind script with the target being in
the dts directory. This way we don't have to go outside the dts
directory and if DTS ever become a python package, we could just copy
the script to the appropriate place. This is also something we don't
really need to do.
And also two minor comments. A lot of suggestions for separate patches
overall. :-)
On 19. 9. 2024 20:16, jspew...@iol.unh.edu wrote:
From: Jeremy Spewock <jspew...@iol.unh.edu>
The DTS framework in its current state supports binding ports to
different drivers on the SUT node but not the TG node. The TG node
already has the information that it needs about the different drivers
that it has available in the configuration file, but it did not
previously have access to the devbind script, so it did not use that
information for anything.
This patch moves the location of the tmp directory as well as the method
for binding ports into the node class rather than the SUT node class and
adds an abstract method for getting the path to the devbind script into
the node class. Then, binding ports to the correct drivers is moved into
the build target setup and run on both nodes.
Bugzilla ID: 1420
Signed-off-by: Jeremy Spewock <jspew...@iol.unh.edu>
---
With the two minor comments,
Reviewed-by: Juraj Linkeš <juraj.lin...@pantheon.tech>
diff --git a/dts/framework/testbed_model/node.py
b/dts/framework/testbed_model/node.py
@@ -58,8 +65,10 @@ class Node(ABC):
lcores: list[LogicalCore]
ports: list[Port]
_logger: DTSLogger
+ _remote_tmp_dir: PurePath
_other_sessions: list[OSSession]
_test_run_config: TestRunConfiguration
+ _path_to_devbind_script: PurePath | None
A note on the naming. We have _remote_tmp_dir and
_path_to_devbind_script. Both are pointing to a remote file/dir, but
only one has the _remote prefix. They should probably be unified.
I've thought a bit about what the right name is. Dropping the prefix
makes sense; sut_node.tmp_dir should mean the tmp dir on the SUT node
(which would make it remote from the execution host's point of view, but
not the node's view; the file is local to SUT node). This could be a
good separate patch (improving the remote/local naming scheme to make it
consistent and as sensible as possible).
diff --git a/dts/framework/utils.py b/dts/framework/utils.py
@@ -29,6 +29,8 @@
from .exception import ConfigurationError, InternalError
REGEX_FOR_PCI_ADDRESS: str = "/[0-9a-fA-F]{4}:[0-9a-fA-F]{2}:[0-9a-fA-F]{2}.[0-9]{1}/"
+#: Path to DPDK directory on the host where DTS is being run.
+LOCAL_DPDK_DIR: PurePath = PurePath(__file__).parents[2]
Local paths can be just pathlib.Path. PurePaths are for path
manipulations only (useful for remote paths in RemoteSessions,
OSSessions and Nodes), but for local existing paths we should use Path.
The OSSession and subclasses need a bit of an update in this regard -
use Path for local paths and PurePaths for remote ones. We added this
into our pre-built DPDK patch.
def expand_range(range_str: str) -> list[int]:
