Re: [PATCH v2] ftrace: Introduce PERMANENT ftrace_ops flag
From: Miroslav Benes <mbenes@suse.cz>
Date: 2019-10-15 11:23:15
Also in:
lkml
Hi Miroslav, Maybe we should add a test to verify this new behavior? See sample version below (lightly tested). We can add to this one, or patch seperately if you prefer.
Thanks a lot, Joe. It looks nice. I'll include it in v3. One question below.
quoted hunk ↗ jump to hunk
-->8-- -->8-- -->8-- -->8-- -->8-- -->8-- -->8-- -->8-- -->8-- -->8--quoted
From c8c9f22e3816ca4c90ab7e7159d2ce536eaa5fad Mon Sep 17 00:00:00 2001From: Joe Lawrence <joe.lawrence@redhat.com> Date: Mon, 14 Oct 2019 18:25:01 -0400 Subject: [PATCH] selftests/livepatch: test interaction with ftrace_enabled Since livepatching depends upon ftrace handlers to implement "patched" functionality, verify that the ftrace_enabled sysctl value interacts with livepatch registration as expected. Signed-off-by: Joe Lawrence <joe.lawrence@redhat.com> --- tools/testing/selftests/livepatch/Makefile | 3 +- .../testing/selftests/livepatch/functions.sh | 18 +++++ .../selftests/livepatch/test-ftrace.sh | 65 +++++++++++++++++++ 3 files changed, 85 insertions(+), 1 deletion(-) create mode 100755 tools/testing/selftests/livepatch/test-ftrace.shdiff --git a/tools/testing/selftests/livepatch/Makefile b/tools/testing/selftests/livepatch/Makefile index fd405402c3ff..1886d9d94b88 100644 --- a/tools/testing/selftests/livepatch/Makefile +++ b/tools/testing/selftests/livepatch/Makefile@@ -4,6 +4,7 @@ TEST_PROGS_EXTENDED := functions.sh TEST_PROGS := \ test-livepatch.sh \ test-callbacks.sh \ - test-shadow-vars.sh + test-shadow-vars.sh \ + test-ftrace.sh include ../lib.mkdiff --git a/tools/testing/selftests/livepatch/functions.sh b/tools/testing/selftests/livepatch/functions.sh index 79b0affd21fb..556252efece0 100644 --- a/tools/testing/selftests/livepatch/functions.sh +++ b/tools/testing/selftests/livepatch/functions.sh@@ -52,6 +52,24 @@ function set_dynamic_debug() { EOF } +function push_ftrace_enabled() { + FTRACE_ENABLED=$(sysctl --values kernel.ftrace_enabled) +}
Shouldn't we call push_ftrace_enabled() somewhere at the beginning of the test script? set_dynamic_debug() calls its push_dynamic_debug() directly, but set_ftrace_enabled() is different, because it is called more than once in the script. One could argue that ftrace_enabled has to be true at the beginning of testing anyway, but I think it would be cleaner. Btw, we should probably guarantee that ftrace_enabled is true when livepatch selftests are invoked.
+function pop_ftrace_enabled() {
+ if [[ -n "$FTRACE_ENABLED" ]]; then
+ sysctl kernel.ftrace_enabled="$FTRACE_ENABLED"
+ fi
+}
+# set_ftrace_enabled() - save the current ftrace_enabled config and tweak
+# it for the self-tests. Set a script exit trap
+# that restores the original value.
+function set_ftrace_enabled() {
+ local sysctl="$1"
+ trap pop_ftrace_enabled EXIT INT TERM HUP
+ result=$(sysctl kernel.ftrace_enabled="$1" 2>&1 | paste --serial --delimiters=' ')
+ echo "livepatch: $result" > /dev/kmsg
+}
+
# loop_until(cmd) - loop a command until it is successful or $MAX_RETRIES,
# sleep $RETRY_INTERVAL between attempts
# cmd - command and its arguments to runMiroslav