[LTP] [PATCH] network/tc01.sh: Add a regression test for tc qdisc command

[Alexey Kodanev]

Hi!
On 07.07.2021 12:22, Yang Xu wrote:
> Signed-off-by: Yang Xu <xuyang2018.jy@fujitsu.com>
> ---
>  runtest/net.tcp_cmds                   |  1 +
>  testcases/network/tcp_cmds/tc/Makefile | 13 +++++++
>  testcases/network/tcp_cmds/tc/tc01.sh  | 51 ++++++++++++++++++++++++++
>  3 files changed, 65 insertions(+)
>  create mode 100644 testcases/network/tcp_cmds/tc/Makefile
>  create mode 100755 testcases/network/tcp_cmds/tc/tc01.sh
> 
> diff --git a/runtest/net.tcp_cmds b/runtest/net.tcp_cmds
> index db47dfd5b..7e142de11 100644
> --- a/runtest/net.tcp_cmds
> +++ b/runtest/net.tcp_cmds
> @@ -12,6 +12,7 @@ netstat netstat01.sh
>  ping01 ping01.sh
>  ping02 ping02.sh
>  sendfile sendfile01.sh
> +tc01 tc01.sh
>  tcpdump tcpdump01.sh
>  telnet telnet01.sh
>  iptables iptables01.sh
> diff --git a/testcases/network/tcp_cmds/tc/Makefile b/testcases/network/tcp_cmds/tc/Makefile
> new file mode 100644
> index 000000000..60150a1ce
> --- /dev/null
> +++ b/testcases/network/tcp_cmds/tc/Makefile
> @@ -0,0 +1,13 @@
> +# SPDX-License-Identifier: GPL-2.0-or-later
> +# Copyright (C) 2009, Cisco Systems Inc.
> +# Ngie Cooper, July 2009
> +
> +top_srcdir		?= ../../../..
> +
> +include $(top_srcdir)/include/mk/env_pre.mk
> +
> +INSTALL_TARGETS		:= tc01.sh
> +
> +MAKE_TARGETS		:=
> +
> +include $(top_srcdir)/include/mk/generic_leaf_target.mk
> diff --git a/testcases/network/tcp_cmds/tc/tc01.sh b/testcases/network/tcp_cmds/tc/tc01.sh
> new file mode 100755
> index 000000000..0a241843b
> --- /dev/null
> +++ b/testcases/network/tcp_cmds/tc/tc01.sh
> @@ -0,0 +1,51 @@
> +#!/bin/sh
> +# SPDX-License-Identifier: GPL-2.0-or-later
> +# Copyright (c) 2021 FUJITSU LIMITED. All rights reserved.
> +# Author: Yang Xu<xuyang2018.jy@fujitsu.com>
> +#
> +# When using "tc qdisc add dev teql0 root teql0 command", qdisc_create()
> +# calls teql_qdisc_init() it imediately fails after check "if (m->dev == dev)"
> +# because both devices are teql0, and it does not set qdisc_priv(sch)->m
> +# leaving it zero on error path, then qdisc_create() imediately calls
> +# teql_destroy() which does not expect zero master pointer and we get OOPS
> +# on unpatched kernel.
> +#
> +# If we enable panic_on_oops, this case may crash.
> +#
> +# This kernel bug was introduced by
> +# commit 87b60cfacf9f ("net_sched: fix error recovery at qdisc creation")
> +# and has been fixed by
> +# commit 1ffbc7ea9160 ("net: sched: sch_teql: fix null-pointer dereference")
> +#
> +
> +TST_TESTFUNC="do_test"
> +TST_NEEDS_ROOT=1
> +TST_NEEDS_DRIVERS="sch_teql"
> +TST_NEEDS_CMDS="tc modprobe"
> +sys_file="/proc/sys/kernel/panic_on_oops"
> +. tst_test.sh
> +
> +do_test()
> +{
> +	[ ! -f "$sys_file" ] && tst_brk TCONF "$sys_file not enabled"

Why this check is needed and also setting panic_on_oops to
trigger the bug? If not, I would suggest to remove them from
the test.

> +	orig_value=$(cat "$sys_file")
> +	if [ $orig_value = "0" ]; then
> +		echo 1 > $sys_file
> +	fi
> +
> +	tst_res TINFO "Use tc qdisc command to trigger a null-pointer dereference"
> +	modprobe $TST_NEEDS_DRIVERS

setup()
{
	ROD modprobe $TST_NEEDS_DRIVERS
}

> +
> +	tc qdisc add dev teql0 root teql0 2>/dev/null

TST_EXP_FAIL tc qdisc add dev teql0 root teql0

> +	if [ $? -eq 0 ]; then
> +		tst_res TFAIL "tc qdisc command succeeded unexpectedly"
> +	else
> +		tst_res TPASS "the bug was not reproduced"
> +	fi
> +
> +	if [ $orig_value = "0" ]; then
> +		 echo 0 > $sys_file
> +	fi
> +}
> +
> +tst_run
>