On Fri, May 31, 2024 at 11:53:47PM +0000, Zhouyi Zhou wrote:

Add CFcommon.arch for the various arch's need for rcutorture.

According to [1] and [2], this patch Fixes: a6fda6dab93c ("rcutorture: Tweak kvm options") by moving x86 specific kernel option CONFIG_HYPERVISOR_GUEST to CFcommon.x86

[1] https://lore.kernel.org/all/20240427005626.1365935-1-zhouzhouyi@gmail.com/ [2] https://lore.kernel.org/all/059d36ce-6453-42be-a31e-895abd35d590@paulmck-lap...

Tested in x86_64 and PPC VM of Open Source Lab of Oregon State University.

Signed-off-by: Zhouyi Zhou zhouzhouyi@gmail.com

Hi Paul,

I tried very hard to find in Linux kernel on how to dig out the x86 specific kernel option CONFIG_HYPERVISOR_GUEST before configcheck.sh generates ConfigFragment.diags.

I can only find this functionality in scripts/kconfig/conf which travels the Kconfig hierarchy.

But the output of scripts/kconfig/conf, which is .config is also one of the input of configcheck.sh:

kvm-recheck.sh:	configcheck.sh $i/.config $i/ConfigFragment > $i/ConfigFragment.diags 2>&1

I feel some logic paradox in it ;-)

At least now we can say that you looked.

So, I pick the simplest way.

Very good, thank you!

One more thing, recent change in include/linux/bitmap.h cause the make of allmodconfig fail because of warning on both x86 platforms, I am going to do research on it.

Thank you, looking forward to seeing what you come up with.

One question below.

Thank your for your guidance Zhouyi -- tools/testing/selftests/rcutorture/bin/kvm-test-1-run.sh | 9 +++++++++ tools/testing/selftests/rcutorture/configs/rcu/CFcommon | 1 - .../selftests/rcutorture/configs/rcu/CFcommon.x86 | 1 + 3 files changed, 10 insertions(+), 1 deletion(-) create mode 100644 tools/testing/selftests/rcutorture/configs/rcu/CFcommon.x86

diff --git a/tools/testing/selftests/rcutorture/bin/kvm-test-1-run.sh b/tools/testing/selftests/rcutorture/bin/kvm-test-1-run.sh index b33cd8753689..5332224238ba 100755 --- a/tools/testing/selftests/rcutorture/bin/kvm-test-1-run.sh +++ b/tools/testing/selftests/rcutorture/bin/kvm-test-1-run.sh @@ -62,6 +62,15 @@ config_override_param () { } echo > $T/KcList +if uname -m | grep -q 86

Why not just use "uname -m" output itself for the filename extension? That way, we wouldn't need this "if" statement. Or are you looking to map both "i686" and "x86_64" to "x86"?

+# TODO: add other architecture-specific common configuration when needed +then

• if test -f $config_dir/CFcommon.x86

Do we really need this inner "if"? The config_override_param() bash function already checks for ran empty-string third argument. Or does this cause some other problem?

• then

• config_override_param "$config_dir/CFcommon.x86" KcList\
  

• 	"`cat $config_dir/CFcommon.x86 2> /dev/null`"
  

• fi

+fi config_override_param "$config_dir/CFcommon" KcList "`cat $config_dir/CFcommon 2> /dev/null`"

Also, the CFcommon.x86 check shouild follow the CFcommon check. This allows architecture-specific values to override the global ones.

Thanx, Paul

config_override_param "$config_template" KcList "`cat $config_template 2> /dev/null`" config_override_param "--gdb options" KcList "$TORTURE_KCONFIG_GDB_ARG" diff --git a/tools/testing/selftests/rcutorture/configs/rcu/CFcommon b/tools/testing/selftests/rcutorture/configs/rcu/CFcommon index 0e92d85313aa..cf0387ae5358 100644 --- a/tools/testing/selftests/rcutorture/configs/rcu/CFcommon +++ b/tools/testing/selftests/rcutorture/configs/rcu/CFcommon @@ -1,6 +1,5 @@ CONFIG_RCU_TORTURE_TEST=y CONFIG_PRINTK_TIME=y -CONFIG_HYPERVISOR_GUEST=y CONFIG_PARAVIRT=y CONFIG_KVM_GUEST=y CONFIG_KCSAN_ASSUME_PLAIN_WRITES_ATOMIC=n diff --git a/tools/testing/selftests/rcutorture/configs/rcu/CFcommon.x86 b/tools/testing/selftests/rcutorture/configs/rcu/CFcommon.x86 new file mode 100644 index 000000000000..2770560d56a0 --- /dev/null +++ b/tools/testing/selftests/rcutorture/configs/rcu/CFcommon.x86 @@ -0,0 +1 @@

+CONFIG_HYPERVISOR_GUEST=y

2.25.1