Live Patching
 help / color / mirror / Atom feed
* [RFC PATCH bpf-next v2 0/2] ftrace: deprecate the ftrace_enabled disable switch
@ 2026-07-31 17:53 Andrey Grodzovsky
  2026-07-31 17:53 ` [RFC PATCH bpf-next v2 1/2] ftrace: deprecate disabling via ftrace_enabled sysctl Andrey Grodzovsky
  2026-07-31 17:53 ` [RFC PATCH bpf-next v2 2/2] selftests/livepatch: update test-ftrace.sh for deprecated ftrace_enabled Andrey Grodzovsky
  0 siblings, 2 replies; 4+ messages in thread
From: Andrey Grodzovsky @ 2026-07-31 17:53 UTC (permalink / raw)
  To: bpf, linux-trace-kernel, live-patching, linux-kselftest,
	linux-doc, rostedt, mbenes
  Cc: corbet, jpoimboe, shuah, jolsa, mhiramat, stable,
	linux-open-source

This adresses a long-standing issue: kernel.ftrace_enabled=0 silently
disables BPF trampolines (fentry/fexit) and ftrace-based
kprobes/kretprobes. The write succeeds, the hook stops firing with no
error, and re-enabling silently restores it.

The solution chosen is to deny setting this knob to 0 from userspace,
thus preventing this case in the first place. Steven mentioned that the
switch became effectively useless and doesn't serve any meaningful
purpose anymore, and only creates problems for systems that rely on
ftrace, such as Livepatching and eBPF. Any attempt to set it to 0 will
fail with -EOPNOTSUPP. Reading and writing 1 remain unchanged.

Patch 1: the sysctl change plus a doc note.
Patch 2: updates the one selftest that relied on the old disable
behavior.


Changes since v1 :
  - test-ftrace.sh: keep the full original disable/reload scenario on
    kernels where kernel.ftrace_enabled=0 still works, and only fall
    back to the simple "write is refused" check on kernels that
    deprecate the knob.(Steven)

[1] https://lore.kernel.org/bpf/20260730163544.2042327-1-andrey.grodzovsky@crowdstrike.com/

Andrey Grodzovsky (2):
  ftrace: deprecate disabling via ftrace_enabled sysctl
  selftests/livepatch: update test-ftrace.sh for deprecated
    ftrace_enabled

 Documentation/trace/ftrace.rst                |  5 +++
 kernel/trace/ftrace.c                         | 19 ++++----
 .../testing/selftests/livepatch/functions.sh  | 13 ++++++
 .../selftests/livepatch/test-ftrace.sh        | 45 ++++++++++++-------
 4 files changed, 54 insertions(+), 28 deletions(-)

-- 
2.34.1


^ permalink raw reply	[flat|nested] 4+ messages in thread

* [RFC PATCH bpf-next v2 1/2] ftrace: deprecate disabling via ftrace_enabled sysctl
  2026-07-31 17:53 [RFC PATCH bpf-next v2 0/2] ftrace: deprecate the ftrace_enabled disable switch Andrey Grodzovsky
@ 2026-07-31 17:53 ` Andrey Grodzovsky
  2026-07-31 17:53 ` [RFC PATCH bpf-next v2 2/2] selftests/livepatch: update test-ftrace.sh for deprecated ftrace_enabled Andrey Grodzovsky
  1 sibling, 0 replies; 4+ messages in thread
From: Andrey Grodzovsky @ 2026-07-31 17:53 UTC (permalink / raw)
  To: bpf, linux-trace-kernel, live-patching, linux-kselftest,
	linux-doc, rostedt, mbenes
  Cc: corbet, jpoimboe, shuah, jolsa, mhiramat, stable,
	linux-open-source

Writing 0 to kernel.ftrace_enabled has not reliably disabled ftrace
for years (FTRACE_OPS_FL_PERMANENT users already block it, and more
callers rely on ftrace always being on). Refuse the write instead of
leaving it in an inconsistent "disables some, not all" state: return
-EOPNOTSUPP and log a message. Reads and enabling (writing 1) are
unaffected.

Update the docs to note the deprecation up front.

Suggested-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@crowdstrike.com>
---
 Documentation/trace/ftrace.rst |  5 +++++
 kernel/trace/ftrace.c          | 19 ++++++++-----------
 2 files changed, 13 insertions(+), 11 deletions(-)

diff --git a/Documentation/trace/ftrace.rst b/Documentation/trace/ftrace.rst
index 84f06bf0da9b..7261f25f8b4b 100644
--- a/Documentation/trace/ftrace.rst
+++ b/Documentation/trace/ftrace.rst
@@ -3313,6 +3313,11 @@ this special filter via::
 ftrace_enabled
 --------------
 
+.. note::
+   Disabling ftrace via this switch is deprecated. Writing 0 is refused
+   with -EOPNOTSUPP and logs a warning; writing 1 and reading the value
+   are unaffected.
+
 Note, the proc sysctl ftrace_enable is a big on/off switch for the
 function tracer. By default it is enabled (when function tracing is
 enabled in the kernel). If it is disabled, all function tracing is
diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c
index f93e34dd2328..82bb7356ccce 100644
--- a/kernel/trace/ftrace.c
+++ b/kernel/trace/ftrace.c
@@ -9357,7 +9357,7 @@ static void ftrace_startup_sysctl(void)
 	}
 }
 
-static void ftrace_shutdown_sysctl(void)
+static void __maybe_unused ftrace_shutdown_sysctl(void)
 {
 	int command;
 
@@ -9377,7 +9377,7 @@ static void ftrace_shutdown_sysctl(void)
 # define ftrace_shutdown_sysctl()      do { } while (0)
 #endif /* CONFIG_DYNAMIC_FTRACE */
 
-static bool is_permanent_ops_registered(void)
+static bool __maybe_unused is_permanent_ops_registered(void)
 {
 	struct ftrace_ops *op;
 
@@ -9415,15 +9415,12 @@ ftrace_enable_sysctl(const struct ctl_table *table, int write,
 		ftrace_startup_sysctl();
 
 	} else {
-		if (is_permanent_ops_registered()) {
-			ftrace_enabled = true;
-			return -EBUSY;
-		}
-
-		/* stopping ftrace calls (just send to ftrace_stub) */
-		ftrace_trace_function = ftrace_stub;
-
-		ftrace_shutdown_sysctl();
+		/*
+		 * Disabling ftrace at runtime via this knob is deprecated.
+		 */
+		ftrace_enabled = true;
+		pr_warn_once("The ftrace_enabled file is deprecated and no longer disables ftrace\n");
+		return -EOPNOTSUPP;
 	}
 
 	last_ftrace_enabled = !!ftrace_enabled;
-- 
2.34.1


^ permalink raw reply related	[flat|nested] 4+ messages in thread

* [RFC PATCH bpf-next v2 2/2] selftests/livepatch: update test-ftrace.sh for deprecated ftrace_enabled
  2026-07-31 17:53 [RFC PATCH bpf-next v2 0/2] ftrace: deprecate the ftrace_enabled disable switch Andrey Grodzovsky
  2026-07-31 17:53 ` [RFC PATCH bpf-next v2 1/2] ftrace: deprecate disabling via ftrace_enabled sysctl Andrey Grodzovsky
@ 2026-07-31 17:53 ` Andrey Grodzovsky
  2026-08-01  0:30   ` Joe Lawrence
  1 sibling, 1 reply; 4+ messages in thread
From: Andrey Grodzovsky @ 2026-07-31 17:53 UTC (permalink / raw)
  To: bpf, linux-trace-kernel, live-patching, linux-kselftest,
	linux-doc, rostedt, mbenes
  Cc: corbet, jpoimboe, shuah, jolsa, mhiramat, stable,
	linux-open-source

kernel.ftrace_enabled=0 is now refused on kernels that deprecate the
knob, so the old disable/reload flow no longer applies there. Probe
for this with ftrace_disable_supported() and keep the full original
scenario (disable, fail to load a livepatch, re-enable, load, confirm
disable is refused while loaded) on kernels where it still works;
otherwise just confirm the write is refused.

Suggested-by: Steven Rostedt <rostedt@goodmis.org>
Suggested-by: Miroslav Benes <mbenes@suse.cz>
Assisted-by: Claude:claude-sonnet-5
Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@crowdstrike.com>
---
 .../testing/selftests/livepatch/functions.sh  | 13 ++++++
 .../selftests/livepatch/test-ftrace.sh        | 45 ++++++++++++-------
 2 files changed, 41 insertions(+), 17 deletions(-)

diff --git a/tools/testing/selftests/livepatch/functions.sh b/tools/testing/selftests/livepatch/functions.sh
index 30dc677b2f45..7fa4bda70221 100644
--- a/tools/testing/selftests/livepatch/functions.sh
+++ b/tools/testing/selftests/livepatch/functions.sh
@@ -126,6 +126,19 @@ function set_ftrace_enabled() {
 	echo "livepatch: kernel.ftrace_enabled = $result" > /dev/kmsg
 }
 
+# ftrace_disable_supported() - probe whether kernel.ftrace_enabled=0
+#	can still disable ftrace on this kernel. Newer kernels deprecate
+#	the knob and always refuse the write with -EOPNOTSUPP.
+function ftrace_disable_supported() {
+	local result
+
+	sysctl -q kernel.ftrace_enabled=0 &> /dev/null
+	result=$(sysctl --values kernel.ftrace_enabled)
+	sysctl -q kernel.ftrace_enabled=1 &> /dev/null
+
+	[[ "$result" == "0" ]]
+}
+
 function cleanup() {
 	pop_config
 }
diff --git a/tools/testing/selftests/livepatch/test-ftrace.sh b/tools/testing/selftests/livepatch/test-ftrace.sh
index d2c3dea63104..cd27148510f8 100755
--- a/tools/testing/selftests/livepatch/test-ftrace.sh
+++ b/tools/testing/selftests/livepatch/test-ftrace.sh
@@ -12,29 +12,32 @@ setup_config
 # - turn ftrace_enabled OFF and verify livepatches can't load
 # - turn ftrace_enabled ON and verify livepatch can load
 # - verify that ftrace_enabled can't be turned OFF while a livepatch is loaded
+# (skipped on kernels where the sysctl is deprecated and always refuses 0)
 
 start_test "livepatch interaction with ftrace_enabled sysctl"
 
-set_ftrace_enabled 0
-load_failing_mod $MOD_LIVEPATCH
+if ftrace_disable_supported; then
 
-set_ftrace_enabled 1
-load_lp $MOD_LIVEPATCH
-if [[ "$(cat /proc/cmdline)" != "$MOD_LIVEPATCH: this has been live patched" ]] ; then
-	echo -e "FAIL\n\n"
-	die "livepatch kselftest(s) failed"
-fi
+	set_ftrace_enabled 0
+	load_failing_mod $MOD_LIVEPATCH
 
-# Check that ftrace could not get disabled when a livepatch is enabled
-set_ftrace_enabled --fail 0
-if [[ "$(cat /proc/cmdline)" != "$MOD_LIVEPATCH: this has been live patched" ]] ; then
-	echo -e "FAIL\n\n"
-	die "livepatch kselftest(s) failed"
-fi
-disable_lp $MOD_LIVEPATCH
-unload_lp $MOD_LIVEPATCH
+	set_ftrace_enabled 1
+	load_lp $MOD_LIVEPATCH
+	if [[ "$(cat /proc/cmdline)" != "$MOD_LIVEPATCH: this has been live patched" ]] ; then
+		echo -e "FAIL\n\n"
+		die "livepatch kselftest(s) failed"
+	fi
 
-check_result "livepatch: kernel.ftrace_enabled = 0
+	# Check that ftrace could not get disabled when a livepatch is enabled
+	set_ftrace_enabled --fail 0
+	if [[ "$(cat /proc/cmdline)" != "$MOD_LIVEPATCH: this has been live patched" ]] ; then
+		echo -e "FAIL\n\n"
+		die "livepatch kselftest(s) failed"
+	fi
+	disable_lp $MOD_LIVEPATCH
+	unload_lp $MOD_LIVEPATCH
+
+	check_result "livepatch: kernel.ftrace_enabled = 0
 % insmod test_modules/$MOD_LIVEPATCH.ko
 livepatch: enabling patch '$MOD_LIVEPATCH'
 livepatch: '$MOD_LIVEPATCH': initializing patching transition
@@ -60,6 +63,14 @@ livepatch: '$MOD_LIVEPATCH': completing unpatching transition
 livepatch: '$MOD_LIVEPATCH': unpatching complete
 % rmmod $MOD_LIVEPATCH"
 
+else
+
+	set_ftrace_enabled --fail 0
+	check_result "livepatch: sysctl: setting key \"kernel.ftrace_enabled\": \
+Operation not supported"
+
+fi
+
 
 # - verify livepatch can load
 # - check if traces have a patched function
-- 
2.34.1


^ permalink raw reply related	[flat|nested] 4+ messages in thread

* Re: [RFC PATCH bpf-next v2 2/2] selftests/livepatch: update test-ftrace.sh for deprecated ftrace_enabled
  2026-07-31 17:53 ` [RFC PATCH bpf-next v2 2/2] selftests/livepatch: update test-ftrace.sh for deprecated ftrace_enabled Andrey Grodzovsky
@ 2026-08-01  0:30   ` Joe Lawrence
  0 siblings, 0 replies; 4+ messages in thread
From: Joe Lawrence @ 2026-08-01  0:30 UTC (permalink / raw)
  To: Andrey Grodzovsky
  Cc: bpf, linux-trace-kernel, live-patching, linux-kselftest,
	linux-doc, rostedt, mbenes, corbet, jpoimboe, shuah, jolsa,
	mhiramat, stable, linux-open-source

On Fri, Jul 31, 2026 at 01:53:58PM -0400, Andrey Grodzovsky wrote:
> kernel.ftrace_enabled=0 is now refused on kernels that deprecate the
> knob, so the old disable/reload flow no longer applies there. Probe
> for this with ftrace_disable_supported() and keep the full original
> scenario (disable, fail to load a livepatch, re-enable, load, confirm
> disable is refused while loaded) on kernels where it still works;
> otherwise just confirm the write is refused.
> 
> Suggested-by: Steven Rostedt <rostedt@goodmis.org>
> Suggested-by: Miroslav Benes <mbenes@suse.cz>
> Assisted-by: Claude:claude-sonnet-5
> Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@crowdstrike.com>
> ---
>  .../testing/selftests/livepatch/functions.sh  | 13 ++++++
>  .../selftests/livepatch/test-ftrace.sh        | 45 ++++++++++++-------
>  2 files changed, 41 insertions(+), 17 deletions(-)
> 
> diff --git a/tools/testing/selftests/livepatch/functions.sh b/tools/testing/selftests/livepatch/functions.sh
> index 30dc677b2f45..7fa4bda70221 100644
> --- a/tools/testing/selftests/livepatch/functions.sh
> +++ b/tools/testing/selftests/livepatch/functions.sh
> @@ -126,6 +126,19 @@ function set_ftrace_enabled() {
>  	echo "livepatch: kernel.ftrace_enabled = $result" > /dev/kmsg
>  }
>  
> +# ftrace_disable_supported() - probe whether kernel.ftrace_enabled=0
> +#	can still disable ftrace on this kernel. Newer kernels deprecate
> +#	the knob and always refuse the write with -EOPNOTSUPP.
> +function ftrace_disable_supported() {
> +	local result
> +
> +	sysctl -q kernel.ftrace_enabled=0 &> /dev/null
> +	result=$(sysctl --values kernel.ftrace_enabled)
> +	sysctl -q kernel.ftrace_enabled=1 &> /dev/null
> +
> +	[[ "$result" == "0" ]]
> +}

Small nit: AFAICT, it doesn't really affect the test pattern, but it
would be a bit cleaner if this function restored the original value
instead of always turning it on, like (untested):

	orig=$(sysctl --values kernel.ftrace_enabled)
	sysctl -q kernel.ftrace_enabled=0 &> /dev/null
	result=$(sysctl --values kernel.ftrace_enabled)
	sysctl -q "kernel.ftrace_enabled=$orig" &> /dev/null

	[[ "$result" == "0" ]]

--
Joe


^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2026-08-01  0:31 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-07-31 17:53 [RFC PATCH bpf-next v2 0/2] ftrace: deprecate the ftrace_enabled disable switch Andrey Grodzovsky
2026-07-31 17:53 ` [RFC PATCH bpf-next v2 1/2] ftrace: deprecate disabling via ftrace_enabled sysctl Andrey Grodzovsky
2026-07-31 17:53 ` [RFC PATCH bpf-next v2 2/2] selftests/livepatch: update test-ftrace.sh for deprecated ftrace_enabled Andrey Grodzovsky
2026-08-01  0:30   ` Joe Lawrence

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox