Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
Merge branch 'selftests-net-page_poll-allocation-error-injection'
Jakub Kicinski says:
====================
selftests: net: page_poll allocation error injection
Add a test for exercising driver memory allocation failure paths.
page pool is a bit tricky to inject errors into at the page allocator
level because of the bulk alloc and recycling, so add explicit error
injection support "in front" of the caches.
Add a test to exercise that using only the standard APIs.
This is the first useful test for the new tests with an endpoint.
There's no point testing netdevsim here, so this is also the first
HW-only test in Python.
I'm not super happy with the traffic generation using iperf3,
my initial approach was to use mausezahn. But it turned out to be
5x slower in terms of PPS. Hopefully this is good enough for now.
v1: https://lore.kernel.org/all/20240426232400.624864-1-kuba@kernel.org/
====================
Link: https://lore.kernel.org/r/20240429144426.743476-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
+214
-7
+2
net/core/page_pool.c
+2
net/core/page_pool.c
···
5
5
* Copyright (C) 2016 Red Hat, Inc.
6
6
*/
7
7
8
+
#include <linux/error-injection.h>
8
9
#include <linux/types.h>
9
10
#include <linux/kernel.h>
10
11
#include <linux/slab.h>
···
551
550
return page;
552
551
}
553
552
EXPORT_SYMBOL(page_pool_alloc_pages);
553
+
ALLOW_ERROR_INJECTION(page_pool_alloc_pages, NULL);
554
554
555
555
/* Calculate distance between two u32 values, valid if distance is below 2^(31)
556
556
* https://en.wikipedia.org/wiki/Serial_number_arithmetic#General_Solution
+1
-1
tools/testing/selftests/Makefile
+1
-1
tools/testing/selftests/Makefile
···
119
119
TARGETS_HOTPLUG += memory-hotplug
120
120
121
121
# Networking tests want the net/lib target, include it automatically
122
-
ifneq ($(filter net drivers/net,$(TARGETS)),)
122
+
ifneq ($(filter net drivers/net drivers/net/hw,$(TARGETS)),)
123
123
ifeq ($(filter net/lib,$(TARGETS)),)
124
124
INSTALL_DEP_TARGETS := net/lib
125
125
endif
+2
tools/testing/selftests/drivers/net/hw/Makefile
+2
tools/testing/selftests/drivers/net/hw/Makefile
···
9
9
hw_stats_l3.sh \
10
10
hw_stats_l3_gre.sh \
11
11
loopback.sh \
12
+
pp_alloc_fail.py \
12
13
#
13
14
14
15
TEST_FILES := \
···
17
16
#
18
17
19
18
TEST_INCLUDES := \
19
+
$(wildcard lib/py/*.py ../lib/py/*.py) \
20
20
../../../net/lib.sh \
21
21
../../../net/forwarding/lib.sh \
22
22
../../../net/forwarding/ipip_lib.sh \
+16
tools/testing/selftests/drivers/net/hw/lib/py/__init__.py
+16
tools/testing/selftests/drivers/net/hw/lib/py/__init__.py
···
1
+
# SPDX-License-Identifier: GPL-2.0
2
+
3
+
import sys
4
+
from pathlib import Path
5
+
6
+
KSFT_DIR = (Path(__file__).parent / "../../../../..").resolve()
7
+
8
+
try:
9
+
sys.path.append(KSFT_DIR.as_posix())
10
+
from net.lib.py import *
11
+
from drivers.net.lib.py import *
12
+
except ModuleNotFoundError as e:
13
+
ksft_pr("Failed importing `net` library from kernel sources")
14
+
ksft_pr(str(e))
15
+
ktap_result(True, comment="SKIP")
16
+
sys.exit(4)
+129
tools/testing/selftests/drivers/net/hw/pp_alloc_fail.py
+129
tools/testing/selftests/drivers/net/hw/pp_alloc_fail.py
···
1
+
#!/usr/bin/env python3
2
+
# SPDX-License-Identifier: GPL-2.0
3
+
4
+
import time
5
+
import os
6
+
from lib.py import ksft_run, ksft_exit, ksft_pr
7
+
from lib.py import KsftSkipEx, KsftFailEx
8
+
from lib.py import NetdevFamily, NlError
9
+
from lib.py import NetDrvEpEnv
10
+
from lib.py import cmd, tool, GenerateTraffic
11
+
12
+
13
+
def _write_fail_config(config):
14
+
for key, value in config.items():
15
+
with open("/sys/kernel/debug/fail_function/" + key, "w") as fp:
16
+
fp.write(str(value) + "\n")
17
+
18
+
19
+
def _enable_pp_allocation_fail():
20
+
if not os.path.exists("/sys/kernel/debug/fail_function"):
21
+
raise KsftSkipEx("Kernel built without function error injection (or DebugFS)")
22
+
23
+
if not os.path.exists("/sys/kernel/debug/fail_function/page_pool_alloc_pages"):
24
+
with open("/sys/kernel/debug/fail_function/inject", "w") as fp:
25
+
fp.write("page_pool_alloc_pages\n")
26
+
27
+
_write_fail_config({
28
+
"verbose": 0,
29
+
"interval": 511,
30
+
"probability": 100,
31
+
"times": -1,
32
+
})
33
+
34
+
35
+
def _disable_pp_allocation_fail():
36
+
if not os.path.exists("/sys/kernel/debug/fail_function"):
37
+
return
38
+
39
+
if os.path.exists("/sys/kernel/debug/fail_function/page_pool_alloc_pages"):
40
+
with open("/sys/kernel/debug/fail_function/inject", "w") as fp:
41
+
fp.write("\n")
42
+
43
+
_write_fail_config({
44
+
"probability": 0,
45
+
"times": 0,
46
+
})
47
+
48
+
49
+
def test_pp_alloc(cfg, netdevnl):
50
+
def get_stats():
51
+
return netdevnl.qstats_get({"ifindex": cfg.ifindex}, dump=True)[0]
52
+
53
+
def check_traffic_flowing():
54
+
stat1 = get_stats()
55
+
time.sleep(1)
56
+
stat2 = get_stats()
57
+
if stat2['rx-packets'] - stat1['rx-packets'] < 15000:
58
+
raise KsftFailEx("Traffic seems low:", stat2['rx-packets'] - stat1['rx-packets'])
59
+
60
+
61
+
try:
62
+
stats = get_stats()
63
+
except NlError as e:
64
+
if e.nl_msg.error == -95:
65
+
stats = {}
66
+
else:
67
+
raise
68
+
if 'rx-alloc-fail' not in stats:
69
+
raise KsftSkipEx("Driver does not report 'rx-alloc-fail' via qstats")
70
+
71
+
set_g = False
72
+
traffic = None
73
+
try:
74
+
traffic = GenerateTraffic(cfg)
75
+
76
+
check_traffic_flowing()
77
+
78
+
_enable_pp_allocation_fail()
79
+
80
+
s1 = get_stats()
81
+
time.sleep(3)
82
+
s2 = get_stats()
83
+
84
+
if s2['rx-alloc-fail'] - s1['rx-alloc-fail'] < 1:
85
+
raise KsftSkipEx("Allocation failures not increasing")
86
+
if s2['rx-alloc-fail'] - s1['rx-alloc-fail'] < 100:
87
+
raise KsftSkipEx("Allocation increasing too slowly", s2['rx-alloc-fail'] - s1['rx-alloc-fail'],
88
+
"packets:", s2['rx-packets'] - s1['rx-packets'])
89
+
90
+
# Basic failures are fine, try to wobble some settings to catch extra failures
91
+
check_traffic_flowing()
92
+
g = tool("ethtool", "-g " + cfg.ifname, json=True)[0]
93
+
if 'rx' in g and g["rx"] * 2 <= g["rx-max"]:
94
+
new_g = g['rx'] * 2
95
+
elif 'rx' in g:
96
+
new_g = g['rx'] // 2
97
+
else:
98
+
new_g = None
99
+
100
+
if new_g:
101
+
set_g = cmd(f"ethtool -G {cfg.ifname} rx {new_g}", fail=False).ret == 0
102
+
if set_g:
103
+
ksft_pr("ethtool -G change retval: success")
104
+
else:
105
+
ksft_pr("ethtool -G change retval: did not succeed", new_g)
106
+
else:
107
+
ksft_pr("ethtool -G change retval: did not try")
108
+
109
+
time.sleep(0.1)
110
+
check_traffic_flowing()
111
+
finally:
112
+
_disable_pp_allocation_fail()
113
+
if traffic:
114
+
traffic.stop()
115
+
time.sleep(0.1)
116
+
if set_g:
117
+
cmd(f"ethtool -G {cfg.ifname} rx {g['rx']}")
118
+
119
+
120
+
def main() -> None:
121
+
netdevnl = NetdevFamily()
122
+
with NetDrvEpEnv(__file__, nsim_test=False) as cfg:
123
+
124
+
ksft_run([test_pp_alloc], args=(cfg, netdevnl, ))
125
+
ksft_exit()
126
+
127
+
128
+
if __name__ == "__main__":
129
+
main()
+1
tools/testing/selftests/drivers/net/lib/py/__init__.py
+1
tools/testing/selftests/drivers/net/lib/py/__init__.py
+8
-2
tools/testing/selftests/drivers/net/lib/py/env.py
+8
-2
tools/testing/selftests/drivers/net/lib/py/env.py
···
2
2
3
3
import os
4
4
from pathlib import Path
5
-
from lib.py import KsftSkipEx
5
+
from lib.py import KsftSkipEx, KsftXfailEx
6
6
from lib.py import cmd, ip
7
7
from lib.py import NetNS, NetdevSimDev
8
8
from .remote import Remote
···
76
76
nsim_v4_pfx = "192.0.2."
77
77
nsim_v6_pfx = "2001:db8::"
78
78
79
-
def __init__(self, src_path):
79
+
def __init__(self, src_path, nsim_test=None):
80
80
81
81
self.env = _load_env_file(src_path)
82
82
···
88
88
self._ns_peer = None
89
89
90
90
if "NETIF" in self.env:
91
+
if nsim_test is True:
92
+
raise KsftXfailEx("Test only works on netdevsim")
91
93
self._check_env()
94
+
92
95
self.dev = ip("link show dev " + self.env['NETIF'], json=True)[0]
93
96
94
97
self.v4 = self.env.get("LOCAL_V4")
···
101
98
kind = self.env["REMOTE_TYPE"]
102
99
args = self.env["REMOTE_ARGS"]
103
100
else:
101
+
if nsim_test is False:
102
+
raise KsftXfailEx("Test does not work on netdevsim")
103
+
104
104
self.create_local()
105
105
106
106
self.dev = self._ns.nsims[0].dev
+41
tools/testing/selftests/drivers/net/lib/py/load.py
+41
tools/testing/selftests/drivers/net/lib/py/load.py
···
1
+
# SPDX-License-Identifier: GPL-2.0
2
+
3
+
import time
4
+
5
+
from lib.py import ksft_pr, cmd, ip, rand_port, wait_port_listen
6
+
7
+
class GenerateTraffic:
8
+
def __init__(self, env):
9
+
env.require_cmd("iperf3", remote=True)
10
+
11
+
self.env = env
12
+
13
+
port = rand_port()
14
+
self._iperf_server = cmd(f"iperf3 -s -p {port}", background=True)
15
+
wait_port_listen(port)
16
+
time.sleep(0.1)
17
+
self._iperf_client = cmd(f"iperf3 -c {env.addr} -P 16 -p {port} -t 86400",
18
+
background=True, host=env.remote)
19
+
20
+
# Wait for traffic to ramp up
21
+
pkt = ip("-s link show dev " + env.ifname, json=True)[0]["stats64"]["rx"]["packets"]
22
+
for _ in range(50):
23
+
time.sleep(0.1)
24
+
now = ip("-s link show dev " + env.ifname, json=True)[0]["stats64"]["rx"]["packets"]
25
+
if now - pkt > 1000:
26
+
return
27
+
pkt = now
28
+
self.stop(verbose=True)
29
+
raise Exception("iperf3 traffic did not ramp up")
30
+
31
+
def stop(self, verbose=None):
32
+
self._iperf_client.process(terminate=True)
33
+
if verbose:
34
+
ksft_pr(">> Client:")
35
+
ksft_pr(self._iperf_client.stdout)
36
+
ksft_pr(self._iperf_client.stderr)
37
+
self._iperf_server.process(terminate=True)
38
+
if verbose:
39
+
ksft_pr(">> Server:")
40
+
ksft_pr(self._iperf_server.stdout)
41
+
ksft_pr(self._iperf_server.stderr)
+4
tools/testing/selftests/net/lib/py/ksft.py
+4
tools/testing/selftests/net/lib/py/ksft.py
+10
-4
tools/testing/selftests/net/lib/py/utils.py
+10
-4
tools/testing/selftests/net/lib/py/utils.py
···
56
56
return self.process(terminate=self.terminate)
57
57
58
58
59
-
def ip(args, json=None, ns=None, host=None):
60
-
cmd_str = "ip "
59
+
def tool(name, args, json=None, ns=None, host=None):
60
+
cmd_str = name + ' '
61
61
if json:
62
-
cmd_str += '-j '
62
+
cmd_str += '--json '
63
63
cmd_str += args
64
64
cmd_obj = cmd(cmd_str, ns=ns, host=host)
65
65
if json:
···
67
67
return cmd_obj
68
68
69
69
70
+
def ip(args, json=None, ns=None, host=None):
71
+
if ns:
72
+
args = f'-netns {ns} ' + args
73
+
return tool('ip', args, json=json, host=host)
74
+
75
+
70
76
def rand_port():
71
77
"""
72
78
Get unprivileged port, for now just random, one day we may decide to check if used.
73
79
"""
74
-
return random.randint(1024, 65535)
80
+
return random.randint(10000, 65535)
75
81
76
82
77
83
def wait_port_listen(port, proto="tcp", ns=None, host=None, sleep=0.005, deadline=5):