* [PATCH] tracing: 'select' GENERIC_TRACER, not 'depends'
@ 2026-11-16 22:46 Brian Norris
2026-11-16 22:58 ` sashiko-bot
2026-11-17 13:53 ` Steven Rostedt
0 siblings, 2 replies; 8+ messages in thread
From: Brian Norris @ 2026-11-16 22:46 UTC (permalink / raw)
To: Steven Rostedt, Masami Hiramatsu
Cc: linux-kernel, linux-trace-kernel, Mathieu Desnoyers, Brian Norris
GENERIC_TRACER is not a user-selectable symbol, so it's impossible to
enable FTRACE_STARTUP_TEST unless one also enables some other symbol
that 'selects' GENERIC_TRACER for us.
This very file also suggests:
All tracer options should select GENERIC_TRACER
So, use 'select', not 'depends on'.
Fixes: 60a11774b38f ("ftrace: add self-tests")
Signed-off-by: Brian Norris <briannorris@chromium.org>
---
kernel/trace/Kconfig | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/kernel/trace/Kconfig b/kernel/trace/Kconfig
index 35d043b19332..8084d46ef2a0 100644
--- a/kernel/trace/Kconfig
+++ b/kernel/trace/Kconfig
@@ -1135,7 +1135,7 @@ config FTRACE_SELFTEST
config FTRACE_STARTUP_TEST
bool "Perform a startup test on ftrace"
- depends on GENERIC_TRACER
+ select GENERIC_TRACER
select FTRACE_SELFTEST
help
This option performs a series of startup tests on ftrace. On bootup
--
2.56.1.703.gf848ed807c-goog
^ permalink raw reply related [flat|nested] 8+ messages in thread* Re: [PATCH] tracing: 'select' GENERIC_TRACER, not 'depends' 2026-11-16 22:46 [PATCH] tracing: 'select' GENERIC_TRACER, not 'depends' Brian Norris @ 2026-11-16 22:58 ` sashiko-bot 2026-11-17 13:55 ` Steven Rostedt 2026-11-17 13:53 ` Steven Rostedt 1 sibling, 1 reply; 8+ messages in thread From: sashiko-bot @ 2026-11-16 22:58 UTC (permalink / raw) To: Brian Norris; +Cc: linux-trace-kernel Thank you for your contribution! Sashiko AI review found 2 potential issues to consider: - [Low] tracing: FTRACE_STARTUP_TEST misuses select GENERIC_TRACER - [Low] tracing: inaccurate Fixes trailer commit in commit message -- commit 4d561c782a03c23e0e690cb6af78d98e81aa94d6 Author: Brian Norris <briannorris@chromium.org> > tracing: 'select' GENERIC_TRACER, not 'depends' > > GENERIC_TRACER is not a user-selectable symbol, so it's impossible to > enable FTRACE_STARTUP_TEST unless one also enables some other symbol > that 'selects' GENERIC_TRACER for us. > > This very file also suggests: > > All tracer options should select GENERIC_TRACER > > So, use 'select', not 'depends on'. > > Fixes: 60a11774b38f ("ftrace: add self-tests") > Signed-off-by: Brian Norris <briannorris@chromium.org> [Severity: Low] Commit 60a11774b38f ("ftrace: add self-tests") added FTRACE_STARTUP_TEST with "depends on TRACING" before GENERIC_TRACER existed. Should the Fixes tag instead reference 5e0a09391087 ("tracing: fix config options to not show when automatically selected"), which introduced GENERIC_TRACER and changed FTRACE_STARTUP_TEST to depend on it? > diff --git a/kernel/trace/Kconfig b/kernel/trace/Kconfig > index c41397eccee50..347658dabbe33 100644 > --- a/kernel/trace/Kconfig > +++ b/kernel/trace/Kconfig > @@ -1135,7 +1135,7 @@ config FTRACE_SELFTEST > > config FTRACE_STARTUP_TEST > bool "Perform a startup test on ftrace" > - depends on GENERIC_TRACER > + select GENERIC_TRACER > select FTRACE_SELFTEST > help > This option performs a series of startup tests on ftrace. On bootup [Severity: Low] Should FTRACE_STARTUP_TEST depend on TRACING (or select TRACING) instead of selecting GENERIC_TRACER? GENERIC_TRACER is meant only for full tracer options so that ENABLE_DEFAULT_TRACERS is hidden when a full tracer is enabled: kernel/trace/Kconfig: # All tracer options should select GENERIC_TRACER. For those options that are # enabled by all tracers (context switch and event tracer) they select TRACING. ... config ENABLE_DEFAULT_TRACERS bool "Trace process context switches and events" depends on !GENERIC_TRACER select TRACING Because FTRACE_STARTUP_TEST is a test option rather than a tracer, having it select GENERIC_TRACER disables and hides ENABLE_DEFAULT_TRACERS when no full tracer is configured, and implicitly pulls in TRACING even when all tracers are disabled. -- Sashiko AI review · https://sashiko.dev/#/patchset/20261116144652.1.Iecb63200dd245f62032a3b1f60ae78cc48868094@changeid?part=1 ^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH] tracing: 'select' GENERIC_TRACER, not 'depends' 2026-11-16 22:58 ` sashiko-bot @ 2026-11-17 13:55 ` Steven Rostedt 2026-11-17 18:23 ` Brian Norris 0 siblings, 1 reply; 8+ messages in thread From: Steven Rostedt @ 2026-11-17 13:55 UTC (permalink / raw) To: sashiko-bot; +Cc: sashiko-reviews, Brian Norris, linux-trace-kernel On Mon, 16 Nov 2026 22:58:01 +0000 sashiko-bot@kernel.org wrote: > [Severity: Low] > Should FTRACE_STARTUP_TEST depend on TRACING (or select TRACING) instead > of selecting GENERIC_TRACER? > > GENERIC_TRACER is meant only for full tracer options so that > ENABLE_DEFAULT_TRACERS is hidden when a full tracer is enabled: > > kernel/trace/Kconfig: > # All tracer options should select GENERIC_TRACER. For those options that are > # enabled by all tracers (context switch and event tracer) they select TRACING. > ... > config ENABLE_DEFAULT_TRACERS > bool "Trace process context switches and events" > depends on !GENERIC_TRACER > select TRACING > > Because FTRACE_STARTUP_TEST is a test option rather than a tracer, having > it select GENERIC_TRACER disables and hides ENABLE_DEFAULT_TRACERS when no > full tracer is configured, and implicitly pulls in TRACING even when all > tracers are disabled. Actually, the selftests should depend on TRACING and not GENERIC_TRACER. It may have been that the self tests were added before the trace events. But it should not "select" a config. The selftest are only to be selected if there is something to test. -- Steve ^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH] tracing: 'select' GENERIC_TRACER, not 'depends' 2026-11-17 13:55 ` Steven Rostedt @ 2026-11-17 18:23 ` Brian Norris 2026-11-17 18:36 ` sashiko-bot 0 siblings, 1 reply; 8+ messages in thread From: Brian Norris @ 2026-11-17 18:23 UTC (permalink / raw) To: Steven Rostedt; +Cc: sashiko-bot, sashiko-reviews, linux-trace-kernel On Tue, Nov 17, 2026 at 08:55:52AM -0500, Steven Rostedt wrote: > On Mon, 16 Nov 2026 22:58:01 +0000 > sashiko-bot@kernel.org wrote: > > > [Severity: Low] > > Should FTRACE_STARTUP_TEST depend on TRACING (or select TRACING) instead > > of selecting GENERIC_TRACER? > > > > GENERIC_TRACER is meant only for full tracer options so that > > ENABLE_DEFAULT_TRACERS is hidden when a full tracer is enabled: > > > > kernel/trace/Kconfig: > > # All tracer options should select GENERIC_TRACER. For those options that are > > # enabled by all tracers (context switch and event tracer) they select TRACING. > > ... > > config ENABLE_DEFAULT_TRACERS > > bool "Trace process context switches and events" > > depends on !GENERIC_TRACER > > select TRACING > > > > Because FTRACE_STARTUP_TEST is a test option rather than a tracer, having > > it select GENERIC_TRACER disables and hides ENABLE_DEFAULT_TRACERS when no > > full tracer is configured, and implicitly pulls in TRACING even when all > > tracers are disabled. > > Actually, the selftests should depend on TRACING and not GENERIC_TRACER. It > may have been that the self tests were added before the trace events. > > But it should not "select" a config. The selftest are only to be selected > if there is something to test. OK, I can respin with 'depends on TRACING'. However, I think the Kconfig setup for kernel/trace/ is still a bit haphazard. For example, it's reasonably common to want EVENT_TRACING but not the more invasive tracing (like FUNCTION_TRACER; and therefore no GENERIC_TRACER) -- this is how all Android kernels are built, for one. But there's no user-configurable way to enable EVENT_TRACING. We only get it implicitly if something else inadvertently selects 'FTRACE' (which selects EVENT_TRACING). For Android, that's CONFIG_KPROBES and CONFIG_HIST_TRIGGERS. ...or is that what ENABLE_DEFAULT_TRACERS is? Its name doesn't really suggest that, and it's mutually exclusive witih many other tracers, so it's a bit awkward making choices, as the whole system appears much less orthogonal than expected. Anyway, my point is that it's not exactly straightforward to enable the features you want and get the tests you want. I've appended a diff that seems a bit more user-friendly to me. It makes it more explicit in the menu how event-tracing is forced-on by many other tracers, but is user-configurable when it's not forced. I can formally submit it (in 2 patches) if you'd like. Brian --- a/kernel/trace/Kconfig +++ b/kernel/trace/Kconfig @@ -198,6 +198,7 @@ config TRACING config GENERIC_TRACER bool select TRACING + select ENABLE_DEFAULT_TRACERS # # Minimum requirements an architecture has to meet for us to @@ -595,7 +596,6 @@ config MMIOTRACE config ENABLE_DEFAULT_TRACERS bool "Trace process context switches and events" - depends on !GENERIC_TRACER select TRACING help This tracer hooks to various trace points in the kernel, @@ -1135,7 +1135,7 @@ config FTRACE_SELFTEST config FTRACE_STARTUP_TEST bool "Perform a startup test on ftrace" - depends on GENERIC_TRACER + depends on TRACING select FTRACE_SELFTEST help This option performs a series of startup tests on ftrace. On bootup ^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH] tracing: 'select' GENERIC_TRACER, not 'depends' 2026-11-17 18:23 ` Brian Norris @ 2026-11-17 18:36 ` sashiko-bot 2026-11-17 18:42 ` Brian Norris 0 siblings, 1 reply; 8+ messages in thread From: sashiko-bot @ 2026-11-17 18:36 UTC (permalink / raw) To: Brian Norris; +Cc: linux-trace-kernel Thank you for your contribution! Sashiko AI review found 2 potential issues to consider: - [High] tracing: missing Signed-off-by trailer and unformatted email reply changelog - [Low] tracing: ENABLE_DEFAULT_TRACERS stays unselected for direct TRACING selectors -- commit 3e6de7d49bb4a2d1503d2aeab72b682a01dfe176 Author: Brian Norris <briannorris@chromium.org> > tracing: 'select' GENERIC_TRACER, not 'depends' > > On Tue, Nov 17, 2026 at 08:55:52AM -0500, Steven Rostedt wrote: > > On Mon, 16 Nov 2026 22:58:01 +0000 > > sashiko-bot@kernel.org wrote: > > > > > [Severity: Low] > > > Should FTRACE_STARTUP_TEST depend on TRACING (or select TRACING) instead > > > of selecting GENERIC_TRACER? > > > > > > GENERIC_TRACER is meant only for full tracer options so that > > > ENABLE_DEFAULT_TRACERS is hidden when a full tracer is enabled: > > > > > > kernel/trace/Kconfig: > > > # All tracer options should select GENERIC_TRACER. For those options that are > > > # enabled by all tracers (context switch and event tracer) they select TRACING. > > > ... > > > config ENABLE_DEFAULT_TRACERS > > > bool "Trace process context switches and events" > > > depends on !GENERIC_TRACER > > > select TRACING > > > > > > Because FTRACE_STARTUP_TEST is a test option rather than a tracer, having > > > it select GENERIC_TRACER disables and hides ENABLE_DEFAULT_TRACERS when no > > > full tracer is configured, and implicitly pulls in TRACING even when all > > > tracers are disabled. > > > > Actually, the selftests should depend on TRACING and not GENERIC_TRACER. It > > may have been that the self tests were added before the trace events. > > > > But it should not "select" a config. The selftest are only to be selected > > if there is something to test. > > OK, I can respin with 'depends on TRACING'. > > However, I think the Kconfig setup for kernel/trace/ is still a bit > haphazard. For example, it's reasonably common to want EVENT_TRACING but > not the more invasive tracing (like FUNCTION_TRACER; and therefore no > GENERIC_TRACER) -- this is how all Android kernels are built, for one. > > But there's no user-configurable way to enable EVENT_TRACING. We only > get it implicitly if something else inadvertently selects 'FTRACE' > (which selects EVENT_TRACING). For Android, that's CONFIG_KPROBES and > CONFIG_HIST_TRIGGERS. > > ...or is that what ENABLE_DEFAULT_TRACERS is? Its name doesn't really > suggest that, and it's mutually exclusive witih many other tracers, so > it's a bit awkward making choices, as the whole system appears much less > orthogonal than expected. > > Anyway, my point is that it's not exactly straightforward to enable the > features you want and get the tests you want. > > I've appended a diff that seems a bit more user-friendly to me. It > makes it more explicit in the menu how event-tracing is forced-on by > many other tracers, but is user-configurable when it's not forced. I can > formally submit it (in 2 patches) if you'd like. > > Brian [Severity: High] This commit is missing a Signed-off-by trailer, its body is an unedited mailing-list reply rather than a commit changelog, and the subject line states the opposite of the FTRACE_STARTUP_TEST change (which switches "depends on GENERIC_TRACER" to "depends on TRACING"). > diff --git a/kernel/trace/Kconfig b/kernel/trace/Kconfig > index c41397eccee50..f49397f7c92d6 100644 > --- a/kernel/trace/Kconfig > +++ b/kernel/trace/Kconfig > @@ -198,6 +198,7 @@ config TRACING > config GENERIC_TRACER > bool > select TRACING > + select ENABLE_DEFAULT_TRACERS > > # > # Minimum requirements an architecture has to meet for us to > @@ -595,7 +596,6 @@ config MMIOTRACE > > config ENABLE_DEFAULT_TRACERS > bool "Trace process context switches and events" > - depends on !GENERIC_TRACER > select TRACING > help > This tracer hooks to various trace points in the kernel, [Severity: Low] Does ENABLE_DEFAULT_TRACERS remain unselected when TRACING is enabled by options that do not select GENERIC_TRACER? Options such as KPROBE_EVENTS and HIST_TRIGGERS select TRACING directly (which in turn selects EVENT_TRACING and CONTEXT_SWITCH_TRACER) without selecting GENERIC_TRACER. When those options are enabled with GENERIC_TRACER=n, event tracing is already forced on, but ENABLE_DEFAULT_TRACERS stays n and appears in menuconfig as a mutable toggle that has no effect on the build. Also, the comment above config TRACING still explains that GENERIC_TRACER and TRACING are separate so that ENABLE_DEFAULT_TRACERS is hidden when selected by another tracer, which no longer matches the Kconfig logic once "depends on !GENERIC_TRACER" is removed. -- Sashiko AI review · https://sashiko.dev/#/patchset/avybp3O2zxcLozKB@google.com?part=1 ^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH] tracing: 'select' GENERIC_TRACER, not 'depends' 2026-11-17 18:36 ` sashiko-bot @ 2026-11-17 18:42 ` Brian Norris 0 siblings, 0 replies; 8+ messages in thread From: Brian Norris @ 2026-11-17 18:42 UTC (permalink / raw) To: sashiko-reviews; +Cc: linux-trace-kernel On Tue, Nov 17, 2026 at 06:36:09PM +0000, sashiko-bot@kernel.org wrote: > Thank you for your contribution! Sashiko AI review found 2 potential > issues to consider: > - [High] tracing: missing Signed-off-by trailer and unformatted email > reply changelog > - [Low] tracing: ENABLE_DEFAULT_TRACERS stays unselected for direct > TRACING selectors > -- > > Anyway, my point is that it's not exactly straightforward to enable the > > features you want and get the tests you want. > > > > I've appended a diff that seems a bit more user-friendly to me. It > > makes it more explicit in the menu how event-tracing is forced-on by > > many other tracers, but is user-configurable when it's not forced. I can > > formally submit it (in 2 patches) if you'd like. Well clearly sashiko still points out issues here. But I still maintain that the config menu here is confusing and hard to use. I'll see if I can come up with a better solution, or else just accept the un-friendliness. Brian ^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH] tracing: 'select' GENERIC_TRACER, not 'depends' 2026-11-16 22:46 [PATCH] tracing: 'select' GENERIC_TRACER, not 'depends' Brian Norris 2026-11-16 22:58 ` sashiko-bot @ 2026-11-17 13:53 ` Steven Rostedt 2026-11-17 17:45 ` Brian Norris 1 sibling, 1 reply; 8+ messages in thread From: Steven Rostedt @ 2026-11-17 13:53 UTC (permalink / raw) To: Brian Norris Cc: Masami Hiramatsu, linux-kernel, linux-trace-kernel, Mathieu Desnoyers On Mon, 16 Nov 2026 14:46:53 -0800 Brian Norris <briannorris@chromium.org> wrote: > GENERIC_TRACER is not a user-selectable symbol, so it's impossible to > enable FTRACE_STARTUP_TEST unless one also enables some other symbol > that 'selects' GENERIC_TRACER for us. I believe that was the way it was intended. The selftests are to test the things that select GENERIC_TRACER. Why would we enable tests for something that isn't enabled? -- Steve > > This very file also suggests: > > All tracer options should select GENERIC_TRACER > > So, use 'select', not 'depends on'. > > Fixes: 60a11774b38f ("ftrace: add self-tests") > Signed-off-by: Brian Norris <briannorris@chromium.org> > --- ^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH] tracing: 'select' GENERIC_TRACER, not 'depends' 2026-11-17 13:53 ` Steven Rostedt @ 2026-11-17 17:45 ` Brian Norris 0 siblings, 0 replies; 8+ messages in thread From: Brian Norris @ 2026-11-17 17:45 UTC (permalink / raw) To: Steven Rostedt Cc: Masami Hiramatsu, linux-kernel, linux-trace-kernel, Mathieu Desnoyers On Tue, Nov 17, 2026 at 08:53:17AM -0500, Steven Rostedt wrote: > On Mon, 16 Nov 2026 14:46:53 -0800 > Brian Norris <briannorris@chromium.org> wrote: > > > GENERIC_TRACER is not a user-selectable symbol, so it's impossible to > > enable FTRACE_STARTUP_TEST unless one also enables some other symbol > > that 'selects' GENERIC_TRACER for us. > > I believe that was the way it was intended. The selftests are to test the > things that select GENERIC_TRACER. Why would we enable tests for something > that isn't enabled? You may have discovered in your other reply, but these tests are applicable to code used by EVENT_TRACING, which can be enabled without GENERIC_TRACER. But I do see how this still might not be exactly the right choice, especially given some of sashiko's comments. Brian ^ permalink raw reply [flat|nested] 8+ messages in thread
end of thread, other threads:[~2026-11-17 18:42 UTC | newest] Thread overview: 8+ messages (download: mbox.gz follow: Atom feed -- links below jump to the message on this page -- 2026-11-16 22:46 [PATCH] tracing: 'select' GENERIC_TRACER, not 'depends' Brian Norris 2026-11-16 22:58 ` sashiko-bot 2026-11-17 13:55 ` Steven Rostedt 2026-11-17 18:23 ` Brian Norris 2026-11-17 18:36 ` sashiko-bot 2026-11-17 18:42 ` Brian Norris 2026-11-17 13:53 ` Steven Rostedt 2026-11-17 17:45 ` Brian Norris
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox