[<prev] [next>] [day] [month] [year] [list]
Message-Id: <1672211275-31393-1-git-send-email-chensong_2000@189.cn>
Date: Wed, 28 Dec 2022 15:07:55 +0800
From: Song Chen <chensong_2000@....cn>
To: rostedt@...dmis.org, mhiramat@...nel.org, arnd@...db.de
Cc: linux-kernel@...r.kernel.org, linux-trace-kernel@...r.kernel.org,
linux-arch@...r.kernel.org, Song Chen <chensong_2000@....cn>
Subject: [PATCH v4 0/2] reorganize trace_peobe_tmpl.h
Hi,
In the first patchset, my intension was to remove trace_probe_tmpl.h,
just keep a unified process_fetch_insn and process_fetch_insn_bottom
in trace_probe.c for all *probes. is_uprobe was the most important
move to approach. If it was not accepted, the whole patchset couldn't
stand.
Then I change my goal to at least no warnings or errors when impelenting
a new feature based on trace probe if it includes trace_probe_tmpl.h by
provideing default behaviors. What's more, it also removes redundant
code in kprobe and eprobe.
---
v2:
1, reorganize patchset.
v3:
1, mark nokprobe_inline for get_event_field.
2, remove warnings reported from kernel test robot.
3, fix errors reported from kernel test robot.
v4:
1, reset changes in v3(2) and v3(3), they are not reasonable fix.
2, fixed errors by adding "#ifdef CONFIG_HAVE_REGS_AND_STACK_ACCESS_API".
Song Chen (2):
kernel/trace: Introduce trace_probe_print_args and use it in *probes
kernel/trace: Provide default impelentations defined in
trace_probe_tmpl.h
kernel/trace/trace_eprobe.c | 146 +-----------------------------
kernel/trace/trace_events_synth.c | 7 +-
kernel/trace/trace_kprobe.c | 106 +---------------------
kernel/trace/trace_probe.c | 27 ++++++
kernel/trace/trace_probe.h | 2 +
kernel/trace/trace_probe_kernel.h | 143 +++++++++++++++++++++++++++--
kernel/trace/trace_probe_tmpl.h | 28 ------
kernel/trace/trace_uprobe.c | 2 +-
8 files changed, 174 insertions(+), 287 deletions(-)
--
2.25.1
Powered by blists - more mailing lists