From: Masami Hiramatsu (Google) <mhiramat@kernel.org>
To: Steven Rostedt <rostedt@goodmis.org>
Cc: linux-kernel@vger.kernel.org, linux-trace-kernel@vger.kernel.org,
Masami Hiramatsu <mhiramat@kernel.org>,
Mark Rutland <mark.rutland@arm.com>,
Mathieu Desnoyers <mathieu.desnoyers@efficios.com>,
Andrew Morton <akpm@linux-foundation.org>,
Alexei Starovoitov <alexei.starovoitov@gmail.com>,
Florent Revest <revest@chromium.org>,
Martin KaFai Lau <martin.lau@linux.dev>,
bpf <bpf@vger.kernel.org>, Sven Schnelle <svens@linux.ibm.com>,
Alexei Starovoitov <ast@kernel.org>, Jiri Olsa <jolsa@kernel.org>,
Arnaldo Carvalho de Melo <acme@kernel.org>,
Daniel Borkmann <daniel@iogearbox.net>,
Alan Maguire <alan.maguire@oracle.com>,
Peter Zijlstra <peterz@infradead.org>,
Thomas Gleixner <tglx@linutronix.de>, Guo Ren <guoren@kernel.org>
Subject: Re: [PATCH v2 24/27] function_graph: Use static_call and branch to optimize entry function
Date: Mon, 3 Jun 2024 12:11:07 +0900 [thread overview]
Message-ID: <20240603121107.42f98858ebb790805f75c9b1@kernel.org> (raw)
In-Reply-To: <20240602033834.997761817@goodmis.org>
On Sat, 01 Jun 2024 23:38:08 -0400
Steven Rostedt <rostedt@goodmis.org> wrote:
> From: "Steven Rostedt (Google)" <rostedt@goodmis.org>
>
> In most cases function graph is used by a single user. Instead of calling
> a loop to call function graph callbacks in this case, call the function
> entry callback directly.
>
> Add a static_key that will be used to set the function graph logic to
> either do the loop (when more than one callback is registered) or to call
> the callback directly if there is only one registered callback.
I understand this works, but my concern is that, if we use fprobe
and function_graph at the same time, does it always loop on both gops?
I mean if those are the subops of one ftrace_ops, ftrace_trampoline
will always call the same function_graph_enter() for both gops, and loop
on the gops list.
For example, if there are 2 fgraph_ops, one has "vfs_*" filter and
another has "sched_*" filter, those does not cover each other.
Are there any way to solve this issue? I think my previous series
calls function_graph_enter_ops() directly from trampoline (If it works
correctly...)
Thank you,
>
> Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
> ---
> kernel/trace/fgraph.c | 77 ++++++++++++++++++++++++++++++++++++-------
> 1 file changed, 66 insertions(+), 11 deletions(-)
>
> diff --git a/kernel/trace/fgraph.c b/kernel/trace/fgraph.c
> index 4d566a0a741d..7c3b0261b1bb 100644
> --- a/kernel/trace/fgraph.c
> +++ b/kernel/trace/fgraph.c
> @@ -11,6 +11,7 @@
> #include <linux/jump_label.h>
> #include <linux/suspend.h>
> #include <linux/ftrace.h>
> +#include <linux/static_call.h>
> #include <linux/slab.h>
>
> #include <trace/events/sched.h>
> @@ -511,6 +512,10 @@ static struct fgraph_ops fgraph_stub = {
> .retfunc = ftrace_graph_ret_stub,
> };
>
> +static struct fgraph_ops *fgraph_direct_gops = &fgraph_stub;
> +DEFINE_STATIC_CALL(fgraph_func, ftrace_graph_entry_stub);
> +DEFINE_STATIC_KEY_TRUE(fgraph_do_direct);
> +
> /**
> * ftrace_graph_stop - set to permanently disable function graph tracing
> *
> @@ -636,21 +641,34 @@ int function_graph_enter(unsigned long ret, unsigned long func,
> if (offset < 0)
> goto out;
>
> - for_each_set_bit(i, &fgraph_array_bitmask,
> - sizeof(fgraph_array_bitmask) * BITS_PER_BYTE) {
> - struct fgraph_ops *gops = fgraph_array[i];
> - int save_curr_ret_stack;
> -
> - if (gops == &fgraph_stub)
> - continue;
> +#ifdef CONFIG_HAVE_STATIC_CALL
> + if (static_branch_likely(&fgraph_do_direct)) {
> + int save_curr_ret_stack = current->curr_ret_stack;
>
> - save_curr_ret_stack = current->curr_ret_stack;
> - if (ftrace_ops_test(&gops->ops, func, NULL) &&
> - gops->entryfunc(&trace, gops))
> - bitmap |= BIT(i);
> + if (static_call(fgraph_func)(&trace, fgraph_direct_gops))
> + bitmap |= BIT(fgraph_direct_gops->idx);
> else
> /* Clear out any saved storage */
> current->curr_ret_stack = save_curr_ret_stack;
> + } else
> +#endif
> + {
> + for_each_set_bit(i, &fgraph_array_bitmask,
> + sizeof(fgraph_array_bitmask) * BITS_PER_BYTE) {
> + struct fgraph_ops *gops = fgraph_array[i];
> + int save_curr_ret_stack;
> +
> + if (gops == &fgraph_stub)
> + continue;
> +
> + save_curr_ret_stack = current->curr_ret_stack;
> + if (ftrace_ops_test(&gops->ops, func, NULL) &&
> + gops->entryfunc(&trace, gops))
> + bitmap |= BIT(i);
> + else
> + /* Clear out any saved storage */
> + current->curr_ret_stack = save_curr_ret_stack;
> + }
> }
>
> if (!bitmap)
> @@ -1155,6 +1173,8 @@ void fgraph_update_pid_func(void)
> gops = container_of(op, struct fgraph_ops, ops);
> gops->entryfunc = ftrace_pids_enabled(op) ?
> fgraph_pid_func : gops->saved_func;
> + if (ftrace_graph_active == 1)
> + static_call_update(fgraph_func, gops->entryfunc);
> }
> }
> }
> @@ -1209,6 +1229,32 @@ static void init_task_vars(int idx)
> read_unlock(&tasklist_lock);
> }
>
> +static void ftrace_graph_enable_direct(bool enable_branch)
> +{
> + trace_func_graph_ent_t func = NULL;
> + int i;
> +
> + for_each_set_bit(i, &fgraph_array_bitmask,
> + sizeof(fgraph_array_bitmask) * BITS_PER_BYTE) {
> + func = fgraph_array[i]->entryfunc;
> + fgraph_direct_gops = fgraph_array[i];
> + }
> + if (WARN_ON_ONCE(!func))
> + return;
> +
> + static_call_update(fgraph_func, func);
> + if (enable_branch)
> + static_branch_disable(&fgraph_do_direct);
> +}
> +
> +static void ftrace_graph_disable_direct(bool disable_branch)
> +{
> + if (disable_branch)
> + static_branch_disable(&fgraph_do_direct);
> + static_call_update(fgraph_func, ftrace_graph_entry_stub);
> + fgraph_direct_gops = &fgraph_stub;
> +}
> +
> int register_ftrace_graph(struct fgraph_ops *gops)
> {
> int command = 0;
> @@ -1235,7 +1281,11 @@ int register_ftrace_graph(struct fgraph_ops *gops)
>
> ftrace_graph_active++;
>
> + if (ftrace_graph_active == 2)
> + ftrace_graph_disable_direct(true);
> +
> if (ftrace_graph_active == 1) {
> + ftrace_graph_enable_direct(false);
> register_pm_notifier(&ftrace_suspend_notifier);
> ret = start_graph_tracing();
> if (ret)
> @@ -1292,6 +1342,11 @@ void unregister_ftrace_graph(struct fgraph_ops *gops)
>
> ftrace_shutdown_subops(&graph_ops, &gops->ops, command);
>
> + if (ftrace_graph_active == 1)
> + ftrace_graph_enable_direct(true);
> + else if (!ftrace_graph_active)
> + ftrace_graph_disable_direct(false);
> +
> if (!ftrace_graph_active) {
> ftrace_graph_return = ftrace_stub_graph;
> ftrace_graph_entry = ftrace_graph_entry_stub;
> --
> 2.43.0
>
>
--
Masami Hiramatsu (Google) <mhiramat@kernel.org>
next prev parent reply other threads:[~2024-06-03 3:11 UTC|newest]
Thread overview: 41+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-06-02 3:37 [PATCH v2 00/27] function_graph: Allow multiple users for function graph tracing Steven Rostedt
2024-06-02 3:37 ` [PATCH v2 01/27] function_graph: Convert ret_stack to a series of longs Steven Rostedt
2024-06-02 3:37 ` [PATCH v2 02/27] fgraph: Use BUILD_BUG_ON() to make sure we have structures divisible by long Steven Rostedt
2024-06-02 3:37 ` [PATCH v2 03/27] function_graph: Add an array structure that will allow multiple callbacks Steven Rostedt
2024-06-02 3:37 ` [PATCH v2 04/27] function_graph: Allow multiple users to attach to function graph Steven Rostedt
2024-06-02 3:37 ` [PATCH v2 05/27] function_graph: Handle tail calls for stack unwinding Steven Rostedt
2024-06-02 3:37 ` [PATCH v2 06/27] function_graph: Remove logic around ftrace_graph_entry and return Steven Rostedt
2024-06-02 3:37 ` [PATCH v2 07/27] ftrace/function_graph: Pass fgraph_ops to function graph callbacks Steven Rostedt
2024-06-02 3:37 ` [PATCH v2 08/27] ftrace: Allow function_graph tracer to be enabled in instances Steven Rostedt
2024-06-02 3:37 ` [PATCH v2 09/27] ftrace: Allow ftrace startup flags to exist without dynamic ftrace Steven Rostedt
2024-06-02 3:37 ` [PATCH v2 10/27] ftrace: Add subops logic to allow one ops to manage many Steven Rostedt
2024-06-03 1:33 ` Masami Hiramatsu
2024-06-03 2:06 ` Steven Rostedt
2024-06-03 2:46 ` Masami Hiramatsu
2024-06-03 14:54 ` Steven Rostedt
2024-06-03 17:05 ` Steven Rostedt
2024-06-02 3:37 ` [PATCH v2 11/27] ftrace: Allow subops filtering to be modified Steven Rostedt
2024-06-03 2:37 ` Masami Hiramatsu
2024-06-03 14:52 ` Steven Rostedt
2024-06-03 23:12 ` Masami Hiramatsu
2024-06-02 3:37 ` [PATCH v2 12/27] function_graph: Have the instances use their own ftrace_ops for filtering Steven Rostedt
2024-06-02 3:37 ` [PATCH v2 13/27] function_graph: Add pid tracing back to function graph tracer Steven Rostedt
2024-06-02 3:37 ` [PATCH v2 14/27] function_graph: Use a simple LRU for fgraph_array index number Steven Rostedt
2024-06-02 3:37 ` [PATCH v2 15/27] function_graph: Add "task variables" per task for fgraph_ops Steven Rostedt
2024-06-02 3:38 ` [PATCH v2 16/27] function_graph: Move set_graph_function tests to shadow stack global var Steven Rostedt
2024-06-02 3:38 ` [PATCH v2 17/27] function_graph: Move graph depth stored data " Steven Rostedt
2024-06-02 3:38 ` [PATCH v2 18/27] function_graph: Move graph notrace bit " Steven Rostedt
2024-06-02 3:38 ` [PATCH v2 19/27] function_graph: Implement fgraph_reserve_data() and fgraph_retrieve_data() Steven Rostedt
2024-06-02 3:38 ` [PATCH v2 20/27] function_graph: Add selftest for passing local variables Steven Rostedt
2024-06-02 3:38 ` [PATCH v2 21/27] ftrace: Add multiple fgraph storage selftest Steven Rostedt
2024-06-02 3:38 ` [PATCH v2 22/27] function_graph: Use for_each_set_bit() in __ftrace_return_to_handler() Steven Rostedt
2024-06-02 3:38 ` [PATCH v2 23/27] function_graph: Use bitmask to loop on fgraph entry Steven Rostedt
2024-06-02 3:38 ` [PATCH v2 24/27] function_graph: Use static_call and branch to optimize entry function Steven Rostedt
2024-06-03 3:11 ` Masami Hiramatsu [this message]
2024-06-03 15:00 ` Steven Rostedt
2024-06-03 15:07 ` Steven Rostedt
2024-06-03 23:08 ` Masami Hiramatsu
2024-06-02 3:38 ` [PATCH v2 25/27] function_graph: Use static_call and branch to optimize return function Steven Rostedt
2024-06-02 3:38 ` [PATCH v2 26/27] selftests/ftrace: Add function_graph tracer to func-filter-pid test Steven Rostedt
2024-06-02 3:38 ` [PATCH v2 27/27] selftests/ftrace: Add fgraph-multi.tc test Steven Rostedt
2024-06-02 3:44 ` [PATCH v2 00/27] function_graph: Allow multiple users for function graph tracing Steven Rostedt
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20240603121107.42f98858ebb790805f75c9b1@kernel.org \
--to=mhiramat@kernel.org \
--cc=acme@kernel.org \
--cc=akpm@linux-foundation.org \
--cc=alan.maguire@oracle.com \
--cc=alexei.starovoitov@gmail.com \
--cc=ast@kernel.org \
--cc=bpf@vger.kernel.org \
--cc=daniel@iogearbox.net \
--cc=guoren@kernel.org \
--cc=jolsa@kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-trace-kernel@vger.kernel.org \
--cc=mark.rutland@arm.com \
--cc=martin.lau@linux.dev \
--cc=mathieu.desnoyers@efficios.com \
--cc=peterz@infradead.org \
--cc=revest@chromium.org \
--cc=rostedt@goodmis.org \
--cc=svens@linux.ibm.com \
--cc=tglx@linutronix.de \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox