[PATCH net-next v3 0/3] selftest: net: Add selftest for netpoll

[PATCH net-next v3 0/3] selftest: net: Add selftest for netpoll

[Breno Leitao]

I am submitting a new selftest for the netpoll subsystem specifically
targeting the case where the RX is polling in the TX path, which is
a case that we don't have any test in the tree today. This is done when
netpoll_poll_dev() called, and this test creates a scenario when that is
probably.

The test does the following:

 1) Configuring a single RX/TX queue to increase contention on the
    interface.
 2) Generating background traffic to saturate the network, mimicking
    real-world congestion.
 3) Sending netconsole messages to trigger netpoll polling and monitor
    its behavior.
 4) Using dynamic netconsole targets via configfs, with the ability to
    delete and recreate targets during the test.
 5) Running bpftrace in parallel to verify that netpoll_poll_dev() is
    called when expected. If it is called, then the test passes,
    otherwise the test is marked as skipped.

In order to achieve it, I stole Jakub's bpftrace helper from [1], and
did some small changes that I found useful to use the helper.

So, this patchset basically contains:

 1) The code stolen from Jakub
 2) Improvements on bpftrace() helper 
 3) The selftest itself

Link: https://lore.kernel.org/all/20250421222827.283737-22-kuba@kernel.org/ [1]

---
Changes in v3:
- Make pylint happy (Simon)
- Remove the unnecessary patch in bpftrace to raise an exception when it
  fails. (Jakub)
- Improved the bpftrace code (Willem)
- Stop sending messages if bpftrace is not alive anymore.
- Link to v2: https://lore.kernel.org/r/20250625-netpoll_test-v2-0-47d27775222c@debian.org

Changes in v2:
- Stole Jakub's helper to run bpftrace
- Removed the DEBUG option and moved logs to logging
- Change the code to have a higher chance of calling netpoll_poll_dev().
  In my current configuration, it is hitting multiple times during the
  test.
- Save and restore TX/RX queue size (Jakub)
- Link to v1: https://lore.kernel.org/r/20250620-netpoll_test-v1-1-5068832f72fc@debian.org

---
Breno Leitao (2):
      selftests: drv-net: Strip '@' prefix from bpftrace map keys
      selftests: net: add netpoll basic functionality test

Jakub Kicinski (1):
      selftests: drv-net: add helper/wrapper for bpftrace

 tools/testing/selftests/drivers/net/Makefile       |   1 +
 .../selftests/drivers/net/lib/py/__init__.py       |   3 +-
 .../testing/selftests/drivers/net/netpoll_basic.py | 345 +++++++++++++++++++++
 tools/testing/selftests/net/lib/py/utils.py        |  35 +++
 4 files changed, 383 insertions(+), 1 deletion(-)
---
base-commit: 8efa26fcbf8a7f783fd1ce7dd2a409e9b7758df0
change-id: 20250612-netpoll_test-a1324d2057c8

Best regards,
--  
Breno Leitao <leitao@debian.org>

[PATCH net-next v3 1/3] selftests: drv-net: add helper/wrapper for bpftrace

[Breno Leitao]

From: Jakub Kicinski <kuba@kernel.org>

bpftrace is very useful for low level driver testing. perf or trace-cmd
would also do for collecting data from tracepoints, but they require
much more post-processing.

Add a wrapper for running bpftrace and sanitizing its output.
bpftrace has JSON output, which is great, but it prints loose objects
and in a slightly inconvenient format. We have to read the objects
line by line, and while at it return them indexed by the map name.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Breno Leitao <leitao@debian.org>
---
 .../selftests/drivers/net/lib/py/__init__.py       |  3 +-
 tools/testing/selftests/net/lib/py/utils.py        | 33 ++++++++++++++++++++++
 2 files changed, 35 insertions(+), 1 deletion(-)

diff --git a/tools/testing/selftests/drivers/net/lib/py/__init__.py b/tools/testing/selftests/drivers/net/lib/py/__init__.py
index 9ed1d8f70524a..98829a0f7a02c 100644
--- a/tools/testing/selftests/drivers/net/lib/py/__init__.py
+++ b/tools/testing/selftests/drivers/net/lib/py/__init__.py
@@ -14,7 +14,8 @@ try:
     from net.lib.py import EthtoolFamily, NetdevFamily, NetshaperFamily, \
         NlError, RtnlFamily
     from net.lib.py import CmdExitFailure
-    from net.lib.py import bkg, cmd, defer, ethtool, fd_read_timeout, ip, \
+    from net.lib.py import bkg, cmd, bpftrace, defer, ethtool, \
+        fd_read_timeout, ip, \
         rand_port, tool, wait_port_listen
     from net.lib.py import fd_read_timeout
     from net.lib.py import KsftSkipEx, KsftFailEx, KsftXfailEx
diff --git a/tools/testing/selftests/net/lib/py/utils.py b/tools/testing/selftests/net/lib/py/utils.py
index 34470d65d871a..760ccf6fccccc 100644
--- a/tools/testing/selftests/net/lib/py/utils.py
+++ b/tools/testing/selftests/net/lib/py/utils.py
@@ -185,6 +185,39 @@ def ethtool(args, json=None, ns=None, host=None):
     return tool('ethtool', args, json=json, ns=ns, host=host)
 
 
+def bpftrace(expr, json=None, ns=None, host=None, timeout=None):
+    """
+    Run bpftrace and return map data (if json=True).
+    The output of bpftrace is inconvenient, so the helper converts
+    to a dict indexed by map name, e.g.:
+     {
+       "@":     { ... },
+       "@map2": { ... },
+     }
+    """
+    cmd_arr = ['bpftrace']
+    # Throw in --quiet if json, otherwise the output has two objects
+    if json:
+        cmd_arr += ['-f', 'json', '-q']
+    if timeout:
+        expr += ' interval:s:' + str(timeout) + ' { exit(); }'
+    cmd_arr += ['-e', expr]
+    cmd_obj = cmd(cmd_arr, ns=ns, host=host, shell=False)
+    if json:
+        # bpftrace prints objects as lines
+        ret = {}
+        for l in cmd_obj.stdout.split('\n'):
+            if not l.strip():
+                continue
+            one = _json.loads(l)
+            if one.get('type') != 'map':
+                continue
+            for k, v in one["data"].items():
+                ret[k] = v
+        return ret
+    return cmd_obj
+
+
 def rand_port(type=socket.SOCK_STREAM):
     """
     Get a random unprivileged port.

-- 
2.47.1

[PATCH net-next v3 2/3] selftests: drv-net: Strip '@' prefix from bpftrace map keys

[Breno Leitao]

The '@' prefix in bpftrace map keys is specific to bpftrace and can be
safely removed when processing results. This patch modifies the bpftrace
utility to strip the '@' from map keys before storing them in the result
dictionary, making the keys more consistent with Python conventions.

Signed-off-by: Breno Leitao <leitao@debian.org>
---
 tools/testing/selftests/net/lib/py/utils.py | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/tools/testing/selftests/net/lib/py/utils.py b/tools/testing/selftests/net/lib/py/utils.py
index 760ccf6fccccc..33c23a928ed1d 100644
--- a/tools/testing/selftests/net/lib/py/utils.py
+++ b/tools/testing/selftests/net/lib/py/utils.py
@@ -213,6 +213,8 @@ def bpftrace(expr, json=None, ns=None, host=None, timeout=None):
             if one.get('type') != 'map':
                 continue
             for k, v in one["data"].items():
+                if k.startswith('@'):
+                    k = k.lstrip('@')
                 ret[k] = v
         return ret
     return cmd_obj

-- 
2.47.1

[PATCH net-next v3 3/3] selftests: net: add netpoll basic functionality test

[Breno Leitao]

Add a basic selftest for the netpoll polling mechanism, specifically
targeting the netpoll poll() side.

The test creates a scenario where network transmission is running at
maximum speed, and netpoll needs to poll the NIC. This is achieved by:

  1. Configuring a single RX/TX queue to create contention
  2. Generating background traffic to saturate the interface
  3. Sending netconsole messages to trigger netpoll polling
  4. Using dynamic netconsole targets via configfs
  5. Delete and create new netconsole targets after some messages
  6. Start a bpftrace in parallel to make sure netpoll_poll_dev() is
     called
  7. If bpftrace exists and netpoll_poll_dev() was called, stop.

The test validates a critical netpoll code path by monitoring traffic
flow and ensuring netpoll_poll_dev() is called when the normal TX path
is blocked.

This addresses a gap in netpoll test coverage for a path that is
tricky for the network stack.

Signed-off-by: Breno Leitao <leitao@debian.org>
---
 tools/testing/selftests/drivers/net/Makefile       |   1 +
 .../testing/selftests/drivers/net/netpoll_basic.py | 345 +++++++++++++++++++++
 2 files changed, 346 insertions(+)

diff --git a/tools/testing/selftests/drivers/net/Makefile b/tools/testing/selftests/drivers/net/Makefile
index bd309b2d39095..9bd84d6b542e5 100644
--- a/tools/testing/selftests/drivers/net/Makefile
+++ b/tools/testing/selftests/drivers/net/Makefile
@@ -16,6 +16,7 @@ TEST_PROGS := \
 	netcons_fragmented_msg.sh \
 	netcons_overflow.sh \
 	netcons_sysdata.sh \
+	netpoll_basic.py \
 	ping.py \
 	queues.py \
 	stats.py \
diff --git a/tools/testing/selftests/drivers/net/netpoll_basic.py b/tools/testing/selftests/drivers/net/netpoll_basic.py
new file mode 100755
index 0000000000000..f523d5a1c707e
--- /dev/null
+++ b/tools/testing/selftests/drivers/net/netpoll_basic.py
@@ -0,0 +1,345 @@
+#!/usr/bin/env python3
+# SPDX-License-Identifier: GPL-2.0
+# Author: Breno Leitao <leitao@debian.org>
+"""
+ This test aims to evaluate the netpoll polling mechanism (as in
+ netpoll_poll_dev()). It presents a complex scenario where the network
+ attempts to send a packet but fails, prompting it to poll the NIC from within
+ the netpoll TX side.
+
+ This has been a crucial path in netpoll that was previously untested. Jakub
+ suggested using a single RX/TX queue, pushing traffic to the NIC, and then
+ sending netpoll messages (via netconsole) to trigger the poll.
+
+ In parallel, bpftrace is used to detect if netpoll_poll_dev() was called. If
+ so, the test passes, otherwise it will be skipped. This test is very dependent on
+ the driver and environment, given we are trying to trigger a tricky scenario.
+"""
+
+import errno
+import logging
+import os
+import random
+import string
+import threading
+import time
+
+from lib.py import (
+    bpftrace,
+    ethtool,
+    GenerateTraffic,
+    ksft_exit,
+    ksft_pr,
+    ksft_run,
+    KsftFailEx,
+    KsftSkipEx,
+    NetDrvEpEnv,
+)
+
+# Configure logging
+logging.basicConfig(
+    level=logging.INFO,
+    format="%(asctime)s - %(levelname)s - %(message)s",
+)
+
+NETCONSOLE_CONFIGFS_PATH: str = "/sys/kernel/config/netconsole"
+NETCONS_REMOTE_PORT: int = 6666
+NETCONS_LOCAL_PORT: int = 1514
+# Max number of netcons messages to send. Each iteration will setup
+# netconsole and send 10 messages
+ITERATIONS: int = 20
+# MAPS contains the information coming from bpftrace
+# it will have only one key: @hits, which tells the number of times
+# netpoll_poll_dev() was called
+MAPS: dict[str, int] = {}
+# Thread to run bpftrace in parallel
+BPF_THREAD: threading.Thread = None
+# Time bpftrace will be running in parallel.
+BPFTRACE_TIMEOUT: int = 15
+
+
+def ethtool_read_rx_tx_queue(interface_name: str) -> tuple[int, int]:
+    """
+    Read the number of RX and TX queues using ethtool. This will be used
+    to restore it after the test
+    """
+    rx_queue = 0
+    tx_queue = 0
+
+    try:
+        ethtool_result = ethtool(f"-g {interface_name}").stdout
+        for line in ethtool_result.splitlines():
+            if line.startswith("RX:"):
+                rx_queue = int(line.split()[1])
+            if line.startswith("TX:"):
+                tx_queue = int(line.split()[1])
+    except IndexError as exception:
+        raise KsftSkipEx(
+            f"Failed to read RX/TX queues numbers: {exception}. Not going to mess with them."
+        ) from exception
+
+    if not rx_queue or not tx_queue:
+        raise KsftSkipEx(
+            "Failed to read RX/TX queues numbers. Not going to mess with them."
+        )
+    return rx_queue, tx_queue
+
+
+def ethtool_set_rx_tx_queue(interface_name: str, rx_val: int, tx_val: int) -> None:
+    """Set the number of RX and TX queues to 1 using ethtool"""
+    try:
+        # This don't need to be reverted, since interfaces will be deleted after test
+        ethtool(f"-G {interface_name} rx {rx_val} tx {tx_val}")
+    except Exception as exception:
+        raise KsftSkipEx(
+            f"Failed to configure RX/TX queues: {exception}. Ethtool not available?"
+        ) from exception
+
+
+def netcons_generate_random_target_name() -> str:
+    """Generate a random target name starting with 'netcons'"""
+    random_suffix = "".join(random.choices(string.ascii_lowercase + string.digits, k=8))
+    return f"netcons_{random_suffix}"
+
+
+def netcons_create_target(
+    config_data: dict[str, str],
+    target_name: str,
+) -> None:
+    """Create a netconsole dynamic target against the interfaces"""
+    logging.debug("Using netconsole name: %s", target_name)
+    try:
+        os.makedirs(f"{NETCONSOLE_CONFIGFS_PATH}/{target_name}", exist_ok=True)
+        logging.debug(
+            "Created target directory: %s/%s", NETCONSOLE_CONFIGFS_PATH, target_name
+        )
+    except OSError as exception:
+        if exception.errno != errno.EEXIST:
+            raise KsftFailEx(
+                f"Failed to create netconsole target directory: {exception}"
+            ) from exception
+
+    try:
+        for key, value in config_data.items():
+            path = f"{NETCONSOLE_CONFIGFS_PATH}/{target_name}/{key}"
+            logging.debug("Writing %s to %s", key, path)
+            with open(path, "w", encoding="utf-8") as file:
+                # Always convert to string to write to file
+                file.write(str(value))
+                file.close()
+
+        # Read all configuration values for debugging purposes
+        for debug_key in config_data.keys():
+            with open(
+                f"{NETCONSOLE_CONFIGFS_PATH}/{target_name}/{debug_key}",
+                "r",
+                encoding="utf-8",
+            ) as file:
+                content = file.read()
+                logging.debug(
+                    "%s/%s/%s : %s",
+                    NETCONSOLE_CONFIGFS_PATH,
+                    target_name,
+                    debug_key,
+                    content,
+                )
+
+    except Exception as exception:
+        raise KsftFailEx(
+            f"Failed to configure netconsole target: {exception}"
+        ) from exception
+
+
+def netcons_configure_target(
+    cfg: NetDrvEpEnv, interface_name: str, target_name: str
+) -> None:
+    """Configure netconsole on the interface with the given target name"""
+    config_data = {
+        "extended": "1",
+        "dev_name": interface_name,
+        "local_port": NETCONS_LOCAL_PORT,
+        "remote_port": NETCONS_REMOTE_PORT,
+        "local_ip": cfg.addr_v["4"] if cfg.addr_ipver == "4" else cfg.addr_v["6"],
+        "remote_ip": (
+            cfg.remote_addr_v["4"] if cfg.addr_ipver == "4" else cfg.remote_addr_v["6"]
+        ),
+        "remote_mac": "00:00:00:00:00:00",  # Not important for this test
+        "enabled": "1",
+    }
+
+    netcons_create_target(config_data, target_name)
+    logging.debug(
+        "Created netconsole target: %s on interface %s", target_name, interface_name
+    )
+
+
+def netcons_delete_target(name: str) -> None:
+    """Delete a netconsole dynamic target"""
+    target_path = f"{NETCONSOLE_CONFIGFS_PATH}/{name}"
+    try:
+        if os.path.exists(target_path):
+            os.rmdir(target_path)
+    except OSError as exception:
+        raise KsftFailEx(
+            f"Failed to delete netconsole target: {exception}"
+        ) from exception
+
+
+def netcons_load_module() -> None:
+    """Try to load the netconsole module"""
+    os.system("modprobe netconsole")
+
+
+def bpftrace_call() -> None:
+    """Call bpftrace to find how many times netpoll_poll_dev() is called.
+    Output is saved in the global variable `maps`"""
+
+    # This is going to update the global variable, that will be seen by the
+    # main function
+    global MAPS  # pylint: disable=W0603
+
+    # This will be passed to bpftrace as in bpftrace -e "expr"
+    expr = "kprobe:netpoll_poll_dev { @hits = count(); }"
+
+    MAPS = bpftrace(expr, timeout=BPFTRACE_TIMEOUT, json=True)
+    logging.debug("BPFtrace output: %s", MAPS)
+
+
+def bpftrace_start():
+    """Start a thread to call `call_bpf` in a parallel thread"""
+    global BPF_THREAD  # pylint: disable=W0603
+
+    BPF_THREAD = threading.Thread(target=bpftrace_call)
+    BPF_THREAD.start()
+    if not BPF_THREAD.is_alive():
+        raise KsftSkipEx("BPFtrace thread is not alive. Skipping test")
+
+
+def bpftrace_stop() -> None:
+    """Stop the bpftrace thread"""
+    if BPF_THREAD:
+        BPF_THREAD.join()
+
+
+def bpftrace_any_hit(join: bool) -> bool:
+    """Check if netpoll_poll_dev() was called by checking the global variable `maps`"""
+    if BPF_THREAD.is_alive():
+        if join:
+            # Wait for bpftrace to finish
+            BPF_THREAD.join()
+        else:
+            # bpftrace is still running, so, we will not check the result yet
+            return False
+
+    logging.debug("MAPS coming from bpftrace = %s", MAPS)
+    if "hits" not in MAPS.keys():
+        raise KsftFailEx(f"bpftrace failed to run!?: {MAPS}")
+
+    return MAPS["hits"] > 0
+
+
+def do_netpoll_flush_monitored(cfg: NetDrvEpEnv, ifname: str, target_name: str) -> None:
+    """Print messages to the console, trying to trigger a netpoll poll"""
+    # Start bpftrace in parallel, so, it is watching
+    # netpoll_poll_dev() while we are sending netconsole messages
+    bpftrace_start()
+
+    do_netpoll_flush(cfg, ifname, target_name)
+
+    if bpftrace_any_hit(join=True):
+        ksft_pr("netpoll_poll_dev() was called. Success")
+        return
+
+    raise KsftSkipEx("netpoll_poll_dev() was not called. Skipping test")
+
+
+def do_netpoll_flush(cfg: NetDrvEpEnv, ifname: str, target_name: str) -> None:
+    """Print messages to the console, trying to trigger a netpoll poll"""
+    netcons_configure_target(cfg, ifname, target_name)
+    retry = 0
+
+    for i in range(int(ITERATIONS)):
+        if not BPF_THREAD.is_alive():
+            # bpftrace is done, stop sending messages
+            break
+
+        msg = f"netcons test #{i}"
+        with open("/dev/kmsg", "w", encoding="utf-8") as kmsg:
+            for j in range(10):
+                try:
+                    kmsg.write(f"{msg}-{j}\n")
+                except OSError as exception:
+                    # in some cases, kmsg can be busy, so, we will retry
+                    time.sleep(1)
+                    retry += 1
+                    if retry < 5:
+                        logging.info("Failed to write to kmsg. Retrying")
+                        # Just retry a few times
+                        continue
+                    raise KsftFailEx(
+                        f"Failed to write to kmsg: {exception}"
+                    ) from exception
+
+            if bpftrace_any_hit(join=False):
+                # Check if netpoll_poll_dev() was called, but do not wait for it
+                # to finish.
+                ksft_pr("netpoll_poll_dev() was called. Success")
+                return
+
+        # Every 5 iterations, toggle netconsole
+        netcons_delete_target(target_name)
+        netcons_configure_target(cfg, ifname, target_name)
+        # If we sleep here, we will have a better chance of triggering
+        # This number is based on a few tests I ran while developing this test
+        time.sleep(0.4)
+
+
+def test_netpoll(cfg: NetDrvEpEnv) -> None:
+    """
+    Test netpoll by sending traffic to the interface and then sending
+    netconsole messages to trigger a poll
+    """
+
+    target_name = netcons_generate_random_target_name()
+    ifname = cfg.dev["ifname"]
+    traffic = None
+    original_queues = ethtool_read_rx_tx_queue(ifname)
+
+    try:
+        # Set RX/TX queues to 1 to force congestion
+        ethtool_set_rx_tx_queue(ifname, 1, 1)
+
+        traffic = GenerateTraffic(cfg)
+        do_netpoll_flush_monitored(cfg, ifname, target_name)
+    finally:
+        if traffic:
+            traffic.stop()
+
+        # Revert RX/TX queues
+        ethtool_set_rx_tx_queue(ifname, original_queues[0], original_queues[1])
+        netcons_delete_target(target_name)
+        bpftrace_stop()
+
+
+def test_check_dependencies() -> None:
+    """Check if the dependencies are met"""
+    if not os.path.exists(NETCONSOLE_CONFIGFS_PATH):
+        raise KsftSkipEx(
+            f"Directory {NETCONSOLE_CONFIGFS_PATH} does not exist. CONFIG_NETCONSOLE_DYNAMIC might not be set."  # pylint: disable=C0301
+        )
+
+
+def main() -> None:
+    """Main function to run the test"""
+    netcons_load_module()
+    test_check_dependencies()
+    with NetDrvEpEnv(__file__, nsim_test=True) as cfg:
+        ksft_run(
+            [test_netpoll],
+            args=(cfg,),
+        )
+    ksft_exit()
+
+
+if __name__ == "__main__":
+    main()

-- 
2.47.1

Re: [PATCH net-next v3 3/3] selftests: net: add netpoll basic functionality test

[Jakub Kicinski]

On Fri, 27 Jun 2025 10:03:11 -0700 Breno Leitao wrote:
> +    raise KsftSkipEx("netpoll_poll_dev() was not called. Skipping test")

As discussed offline SKIPing is not an option for SW tests.
-- 
pw-bot: cr

Re: [PATCH net-next v3 3/3] selftests: net: add netpoll basic functionality test

[Willem de Bruijn]

Breno Leitao wrote:
> Add a basic selftest for the netpoll polling mechanism, specifically
> targeting the netpoll poll() side.
> 
> The test creates a scenario where network transmission is running at
> maximum speed, and netpoll needs to poll the NIC. This is achieved by:
> 
>   1. Configuring a single RX/TX queue to create contention
>   2. Generating background traffic to saturate the interface
>   3. Sending netconsole messages to trigger netpoll polling
>   4. Using dynamic netconsole targets via configfs
>   5. Delete and create new netconsole targets after some messages
>   6. Start a bpftrace in parallel to make sure netpoll_poll_dev() is
>      called
>   7. If bpftrace exists and netpoll_poll_dev() was called, stop.
> 
> The test validates a critical netpoll code path by monitoring traffic
> flow and ensuring netpoll_poll_dev() is called when the normal TX path
> is blocked.
> 
> This addresses a gap in netpoll test coverage for a path that is
> tricky for the network stack.
> 
> Signed-off-by: Breno Leitao <leitao@debian.org>
> ---
>  tools/testing/selftests/drivers/net/Makefile       |   1 +
>  .../testing/selftests/drivers/net/netpoll_basic.py | 345 +++++++++++++++++++++
>  2 files changed, 346 insertions(+)
> 
> diff --git a/tools/testing/selftests/drivers/net/Makefile b/tools/testing/selftests/drivers/net/Makefile
> index bd309b2d39095..9bd84d6b542e5 100644
> --- a/tools/testing/selftests/drivers/net/Makefile
> +++ b/tools/testing/selftests/drivers/net/Makefile
> @@ -16,6 +16,7 @@ TEST_PROGS := \
>  	netcons_fragmented_msg.sh \
>  	netcons_overflow.sh \
>  	netcons_sysdata.sh \
> +	netpoll_basic.py \
>  	ping.py \
>  	queues.py \
>  	stats.py \
> diff --git a/tools/testing/selftests/drivers/net/netpoll_basic.py b/tools/testing/selftests/drivers/net/netpoll_basic.py
> new file mode 100755
> index 0000000000000..f523d5a1c707e
> --- /dev/null
> +++ b/tools/testing/selftests/drivers/net/netpoll_basic.py
> @@ -0,0 +1,345 @@
> +#!/usr/bin/env python3
> +# SPDX-License-Identifier: GPL-2.0
> +# Author: Breno Leitao <leitao@debian.org>
> +"""
> + This test aims to evaluate the netpoll polling mechanism (as in
> + netpoll_poll_dev()). It presents a complex scenario where the network
> + attempts to send a packet but fails, prompting it to poll the NIC from within
> + the netpoll TX side.
> +
> + This has been a crucial path in netpoll that was previously untested. Jakub
> + suggested using a single RX/TX queue, pushing traffic to the NIC, and then
> + sending netpoll messages (via netconsole) to trigger the poll.
> +
> + In parallel, bpftrace is used to detect if netpoll_poll_dev() was called. If
> + so, the test passes, otherwise it will be skipped. This test is very dependent on
> + the driver and environment, given we are trying to trigger a tricky scenario.
> +"""
> +
> +import errno
> +import logging
> +import os
> +import random
> +import string
> +import threading
> +import time
> +
> +from lib.py import (
> +    bpftrace,
> +    ethtool,
> +    GenerateTraffic,
> +    ksft_exit,
> +    ksft_pr,
> +    ksft_run,
> +    KsftFailEx,
> +    KsftSkipEx,
> +    NetDrvEpEnv,
> +)
> +
> +# Configure logging
> +logging.basicConfig(
> +    level=logging.INFO,
> +    format="%(asctime)s - %(levelname)s - %(message)s",
> +)
> +
> +NETCONSOLE_CONFIGFS_PATH: str = "/sys/kernel/config/netconsole"
> +NETCONS_REMOTE_PORT: int = 6666
> +NETCONS_LOCAL_PORT: int = 1514
> +# Max number of netcons messages to send. Each iteration will setup
> +# netconsole and send 10 messages
> +ITERATIONS: int = 20
> +# MAPS contains the information coming from bpftrace
> +# it will have only one key: @hits, which tells the number of times
> +# netpoll_poll_dev() was called

nit: no longer has ampersand prefix

> +MAPS: dict[str, int] = {}
> +# Thread to run bpftrace in parallel
> +BPF_THREAD: threading.Thread = None
> +# Time bpftrace will be running in parallel.
> +BPFTRACE_TIMEOUT: int = 15
> +
> +
> +def ethtool_read_rx_tx_queue(interface_name: str) -> tuple[int, int]:
> +    """
> +    Read the number of RX and TX queues using ethtool. This will be used
> +    to restore it after the test
> +    """
> +    rx_queue = 0
> +    tx_queue = 0
> +
> +    try:
> +        ethtool_result = ethtool(f"-g {interface_name}").stdout
> +        for line in ethtool_result.splitlines():
> +            if line.startswith("RX:"):
> +                rx_queue = int(line.split()[1])
> +            if line.startswith("TX:"):
> +                tx_queue = int(line.split()[1])

Does this work on devices that use combined?

> +    except IndexError as exception:
> +        raise KsftSkipEx(
> +            f"Failed to read RX/TX queues numbers: {exception}. Not going to mess with them."
> +        ) from exception
> +
> +    if not rx_queue or not tx_queue:
> +        raise KsftSkipEx(
> +            "Failed to read RX/TX queues numbers. Not going to mess with them."
> +        )
> +    return rx_queue, tx_queue
> +
> +
> +def ethtool_set_rx_tx_queue(interface_name: str, rx_val: int, tx_val: int) -> None:
> +    """Set the number of RX and TX queues to 1 using ethtool"""
> +    try:
> +        # This don't need to be reverted, since interfaces will be deleted after test
> +        ethtool(f"-G {interface_name} rx {rx_val} tx {tx_val}")
> +    except Exception as exception:
> +        raise KsftSkipEx(
> +            f"Failed to configure RX/TX queues: {exception}. Ethtool not available?"
> +        ) from exception
> +
> +
> +def netcons_generate_random_target_name() -> str:
> +    """Generate a random target name starting with 'netcons'"""
> +    random_suffix = "".join(random.choices(string.ascii_lowercase + string.digits, k=8))
> +    return f"netcons_{random_suffix}"
> +
> +
> +def netcons_create_target(
> +    config_data: dict[str, str],
> +    target_name: str,
> +) -> None:
> +    """Create a netconsole dynamic target against the interfaces"""
> +    logging.debug("Using netconsole name: %s", target_name)
> +    try:
> +        os.makedirs(f"{NETCONSOLE_CONFIGFS_PATH}/{target_name}", exist_ok=True)
> +        logging.debug(
> +            "Created target directory: %s/%s", NETCONSOLE_CONFIGFS_PATH, target_name
> +        )
> +    except OSError as exception:
> +        if exception.errno != errno.EEXIST:
> +            raise KsftFailEx(
> +                f"Failed to create netconsole target directory: {exception}"
> +            ) from exception
> +
> +    try:
> +        for key, value in config_data.items():
> +            path = f"{NETCONSOLE_CONFIGFS_PATH}/{target_name}/{key}"
> +            logging.debug("Writing %s to %s", key, path)
> +            with open(path, "w", encoding="utf-8") as file:
> +                # Always convert to string to write to file
> +                file.write(str(value))
> +                file.close()
> +
> +        # Read all configuration values for debugging purposes
> +        for debug_key in config_data.keys():
> +            with open(
> +                f"{NETCONSOLE_CONFIGFS_PATH}/{target_name}/{debug_key}",
> +                "r",
> +                encoding="utf-8",
> +            ) as file:
> +                content = file.read()
> +                logging.debug(
> +                    "%s/%s/%s : %s",
> +                    NETCONSOLE_CONFIGFS_PATH,
> +                    target_name,
> +                    debug_key,
> +                    content,
> +                )
> +
> +    except Exception as exception:
> +        raise KsftFailEx(
> +            f"Failed to configure netconsole target: {exception}"
> +        ) from exception
> +
> +
> +def netcons_configure_target(
> +    cfg: NetDrvEpEnv, interface_name: str, target_name: str
> +) -> None:
> +    """Configure netconsole on the interface with the given target name"""
> +    config_data = {
> +        "extended": "1",
> +        "dev_name": interface_name,
> +        "local_port": NETCONS_LOCAL_PORT,
> +        "remote_port": NETCONS_REMOTE_PORT,
> +        "local_ip": cfg.addr_v["4"] if cfg.addr_ipver == "4" else cfg.addr_v["6"],
> +        "remote_ip": (
> +            cfg.remote_addr_v["4"] if cfg.addr_ipver == "4" else cfg.remote_addr_v["6"]
> +        ),
> +        "remote_mac": "00:00:00:00:00:00",  # Not important for this test
> +        "enabled": "1",
> +    }
> +
> +    netcons_create_target(config_data, target_name)
> +    logging.debug(
> +        "Created netconsole target: %s on interface %s", target_name, interface_name
> +    )
> +
> +
> +def netcons_delete_target(name: str) -> None:
> +    """Delete a netconsole dynamic target"""
> +    target_path = f"{NETCONSOLE_CONFIGFS_PATH}/{name}"
> +    try:
> +        if os.path.exists(target_path):
> +            os.rmdir(target_path)
> +    except OSError as exception:
> +        raise KsftFailEx(
> +            f"Failed to delete netconsole target: {exception}"
> +        ) from exception
> +
> +
> +def netcons_load_module() -> None:
> +    """Try to load the netconsole module"""
> +    os.system("modprobe netconsole")
> +
> +
> +def bpftrace_call() -> None:
> +    """Call bpftrace to find how many times netpoll_poll_dev() is called.
> +    Output is saved in the global variable `maps`"""
> +
> +    # This is going to update the global variable, that will be seen by the
> +    # main function
> +    global MAPS  # pylint: disable=W0603
> +
> +    # This will be passed to bpftrace as in bpftrace -e "expr"
> +    expr = "kprobe:netpoll_poll_dev { @hits = count(); }"
> +
> +    MAPS = bpftrace(expr, timeout=BPFTRACE_TIMEOUT, json=True)
> +    logging.debug("BPFtrace output: %s", MAPS)
> +
> +
> +def bpftrace_start():
> +    """Start a thread to call `call_bpf` in a parallel thread"""
> +    global BPF_THREAD  # pylint: disable=W0603
> +
> +    BPF_THREAD = threading.Thread(target=bpftrace_call)
> +    BPF_THREAD.start()
> +    if not BPF_THREAD.is_alive():
> +        raise KsftSkipEx("BPFtrace thread is not alive. Skipping test")
> +
> +
> +def bpftrace_stop() -> None:
> +    """Stop the bpftrace thread"""
> +    if BPF_THREAD:
> +        BPF_THREAD.join()
> +
> +
> +def bpftrace_any_hit(join: bool) -> bool:
> +    """Check if netpoll_poll_dev() was called by checking the global variable `maps`"""
> +    if BPF_THREAD.is_alive():
> +        if join:
> +            # Wait for bpftrace to finish
> +            BPF_THREAD.join()
> +        else:
> +            # bpftrace is still running, so, we will not check the result yet
> +            return False
> +
> +    logging.debug("MAPS coming from bpftrace = %s", MAPS)
> +    if "hits" not in MAPS.keys():
> +        raise KsftFailEx(f"bpftrace failed to run!?: {MAPS}")
> +
> +    return MAPS["hits"] > 0
> +
> +
> +def do_netpoll_flush_monitored(cfg: NetDrvEpEnv, ifname: str, target_name: str) -> None:
> +    """Print messages to the console, trying to trigger a netpoll poll"""
> +    # Start bpftrace in parallel, so, it is watching
> +    # netpoll_poll_dev() while we are sending netconsole messages
> +    bpftrace_start()
> +
> +    do_netpoll_flush(cfg, ifname, target_name)
> +
> +    if bpftrace_any_hit(join=True):
> +        ksft_pr("netpoll_poll_dev() was called. Success")
> +        return
> +
> +    raise KsftSkipEx("netpoll_poll_dev() was not called. Skipping test")
> +
> +
> +def do_netpoll_flush(cfg: NetDrvEpEnv, ifname: str, target_name: str) -> None:
> +    """Print messages to the console, trying to trigger a netpoll poll"""
> +    netcons_configure_target(cfg, ifname, target_name)
> +    retry = 0
> +
> +    for i in range(int(ITERATIONS)):
> +        if not BPF_THREAD.is_alive():
> +            # bpftrace is done, stop sending messages
> +            break
> +
> +        msg = f"netcons test #{i}"
> +        with open("/dev/kmsg", "w", encoding="utf-8") as kmsg:
> +            for j in range(10):
> +                try:
> +                    kmsg.write(f"{msg}-{j}\n")
> +                except OSError as exception:
> +                    # in some cases, kmsg can be busy, so, we will retry
> +                    time.sleep(1)
> +                    retry += 1
> +                    if retry < 5:
> +                        logging.info("Failed to write to kmsg. Retrying")
> +                        # Just retry a few times
> +                        continue
> +                    raise KsftFailEx(
> +                        f"Failed to write to kmsg: {exception}"
> +                    ) from exception
> +
> +            if bpftrace_any_hit(join=False):
> +                # Check if netpoll_poll_dev() was called, but do not wait for it
> +                # to finish.
> +                ksft_pr("netpoll_poll_dev() was called. Success")
> +                return
> +
> +        # Every 5 iterations, toggle netconsole
> +        netcons_delete_target(target_name)
> +        netcons_configure_target(cfg, ifname, target_name)
> +        # If we sleep here, we will have a better chance of triggering
> +        # This number is based on a few tests I ran while developing this test
> +        time.sleep(0.4)
> +
> +
> +def test_netpoll(cfg: NetDrvEpEnv) -> None:
> +    """
> +    Test netpoll by sending traffic to the interface and then sending
> +    netconsole messages to trigger a poll
> +    """
> +
> +    target_name = netcons_generate_random_target_name()
> +    ifname = cfg.dev["ifname"]
> +    traffic = None
> +    original_queues = ethtool_read_rx_tx_queue(ifname)
> +
> +    try:
> +        # Set RX/TX queues to 1 to force congestion
> +        ethtool_set_rx_tx_queue(ifname, 1, 1)
> +
> +        traffic = GenerateTraffic(cfg)
> +        do_netpoll_flush_monitored(cfg, ifname, target_name)
> +    finally:
> +        if traffic:
> +            traffic.stop()
> +
> +        # Revert RX/TX queues
> +        ethtool_set_rx_tx_queue(ifname, original_queues[0], original_queues[1])
> +        netcons_delete_target(target_name)
> +        bpftrace_stop()
> +
> +
> +def test_check_dependencies() -> None:
> +    """Check if the dependencies are met"""
> +    if not os.path.exists(NETCONSOLE_CONFIGFS_PATH):
> +        raise KsftSkipEx(
> +            f"Directory {NETCONSOLE_CONFIGFS_PATH} does not exist. CONFIG_NETCONSOLE_DYNAMIC might not be set."  # pylint: disable=C0301
> +        )
> +
> +
> +def main() -> None:
> +    """Main function to run the test"""
> +    netcons_load_module()
> +    test_check_dependencies()
> +    with NetDrvEpEnv(__file__, nsim_test=True) as cfg:
> +        ksft_run(
> +            [test_netpoll],
> +            args=(cfg,),
> +        )
> +    ksft_exit()
> +
> +
> +if __name__ == "__main__":
> +    main()
> 
> -- 
> 2.47.1
> 

Re: [PATCH net-next v3 3/3] selftests: net: add netpoll basic functionality test

[Breno Leitao]

Hello Willem,

On Sat, Jun 28, 2025 at 10:57:20AM -0400, Willem de Bruijn wrote:
> Breno Leitao wrote:
> > +NETCONSOLE_CONFIGFS_PATH: str = "/sys/kernel/config/netconsole"
> > +NETCONS_REMOTE_PORT: int = 6666
> > +NETCONS_LOCAL_PORT: int = 1514
> > +# Max number of netcons messages to send. Each iteration will setup
> > +# netconsole and send 10 messages
> > +ITERATIONS: int = 20
> > +# MAPS contains the information coming from bpftrace
> > +# it will have only one key: @hits, which tells the number of times
> > +# netpoll_poll_dev() was called
> 
> nit: no longer has ampersand prefix

Good catch. I will update.

> > +def ethtool_read_rx_tx_queue(interface_name: str) -> tuple[int, int]:
> > +    """
> > +    Read the number of RX and TX queues using ethtool. This will be used
> > +    to restore it after the test
> > +    """
> > +    rx_queue = 0
> > +    tx_queue = 0
> > +
> > +    try:
> > +        ethtool_result = ethtool(f"-g {interface_name}").stdout
> > +        for line in ethtool_result.splitlines():
> > +            if line.startswith("RX:"):
> > +                rx_queue = int(line.split()[1])
> > +            if line.startswith("TX:"):
> > +                tx_queue = int(line.split()[1])
> 
> Does this work on devices that use combined?

Not sure. This is suppossed to work mostly on netdevsim (for now).

Since I am not familiar with combined TX/RX, I've looked at ethtool
code, and it seems RX and TX wil always be printed?

This is what I found when `-g` is passed to ethtool.

	static int dump_ring(const struct ethtool_ringparam *ering)
	{
		fprintf(stdout,
			"Pre-set maximums:\n"
			"RX:            %u\n"
			"RX Mini:       %u\n"
			"RX Jumbo:      %u\n"
			"TX:            %u\n",
			ering->rx_max_pending,
			ering->rx_mini_max_pending,
			ering->rx_jumbo_max_pending,
			ering->tx_max_pending);

		fprintf(stdout,
			"Current hardware settings:\n"
			"RX:            %u\n"
			"RX Mini:       %u\n"
			"RX Jumbo:      %u\n"
			"TX:            %u\n",
			ering->rx_pending,
			ering->rx_mini_pending,
			ering->rx_jumbo_pending,
			ering->tx_pending);

		fprintf(stdout, "\n");
		return 0;
	}


Thanks for the review,
--breno

Re: [PATCH net-next v3 3/3] selftests: net: add netpoll basic functionality test

[Breno Leitao]

On Fri, Jun 27, 2025 at 11:38:54AM -0700, Jakub Kicinski wrote:
> On Fri, 27 Jun 2025 10:03:11 -0700 Breno Leitao wrote:
> > +    raise KsftSkipEx("netpoll_poll_dev() was not called. Skipping test")
> 
> As discussed offline SKIPing is not an option for SW tests.

Sure, I will move it to failure.

Unfortunately the expected path didn't hit in vmtest. I am still trying
to reproduce the failure on my side, but no luck. It hits from 10 to 16
times per run. 

Do you want me to send it as a failure, or, wait until we get something
better that pass 100% of the time?

Thanks
--breno

Re: [PATCH net-next v3 3/3] selftests: net: add netpoll basic functionality test

[Willem de Bruijn]

Breno Leitao wrote:
> Hello Willem,
> 
> On Sat, Jun 28, 2025 at 10:57:20AM -0400, Willem de Bruijn wrote:
> > Breno Leitao wrote:
> > > +NETCONSOLE_CONFIGFS_PATH: str = "/sys/kernel/config/netconsole"
> > > +NETCONS_REMOTE_PORT: int = 6666
> > > +NETCONS_LOCAL_PORT: int = 1514
> > > +# Max number of netcons messages to send. Each iteration will setup
> > > +# netconsole and send 10 messages
> > > +ITERATIONS: int = 20
> > > +# MAPS contains the information coming from bpftrace
> > > +# it will have only one key: @hits, which tells the number of times
> > > +# netpoll_poll_dev() was called
> > 
> > nit: no longer has ampersand prefix
> 
> Good catch. I will update.
> 
> > > +def ethtool_read_rx_tx_queue(interface_name: str) -> tuple[int, int]:
> > > +    """
> > > +    Read the number of RX and TX queues using ethtool. This will be used
> > > +    to restore it after the test
> > > +    """
> > > +    rx_queue = 0
> > > +    tx_queue = 0
> > > +
> > > +    try:
> > > +        ethtool_result = ethtool(f"-g {interface_name}").stdout
> > > +        for line in ethtool_result.splitlines():
> > > +            if line.startswith("RX:"):
> > > +                rx_queue = int(line.split()[1])
> > > +            if line.startswith("TX:"):
> > > +                tx_queue = int(line.split()[1])
> > 
> > Does this work on devices that use combined?
> 
> Not sure. This is suppossed to work mostly on netdevsim (for now).

Okay. Given that the test targets that. LGTM.
 
> Since I am not familiar with combined TX/RX, I've looked at ethtool
> code, and it seems RX and TX wil always be printed?
> 
> This is what I found when `-g` is passed to ethtool.

I'm also a bit confused about how combined is supposed to work. This
was discussed or documented somewhere recently, but I can't seem to
find an authoritative reference.

To my intuition, "tx N rx M" is equivalent to "combined min(N, M)"
plus the remainder of the larger ones. So "tx 1 rx 1" is equivalent
to "combined 1". But not sure if this is true for all drivers.

Anyway, as said, targeting netdevsim, so can leave out of scope.