Date:   Sat,  1 Apr 2023 23:33:59 +0900
From:   "Masami Hiramatsu (Google)" <mhiramat@...nel.org>
To:     linux-trace-kernel@...r.kernel.org
Cc:     linux-kernel@...r.kernel.org, Steven Rostedt <rostedt@...dmis.org>,
        mhiramat@...nel.org, Florent Revest <revest@...omium.org>,
        Mark Rutland <mark.rutland@....com>,
        Will Deacon <will@...nel.org>
Subject: [PATCH v4 0/3] tracing: Add fprobe events

Hi,

Here is the 4th version of improve fprobe and add a basic fprobe event
support for ftrace (tracefs) and perf. Here is the previous version.

https://lore.kernel.org/all/167526695292.433354.8949652607331707144.stgit@mhiramat.roam.corp.google.com/

Since fprobe update parts are already merged to trace/for-next, this
series only has remaining 3 patches. I rebased those on the trace/for-next
and confirmed to pass the ftrace selftest.

With this fprobe events, we can continue to trace function entry/exit
even if the CONFIG_KPROBES_ON_FTRACE is not available. Since
CONFIG_KPROBES_ON_FTRACE requires the CONFIG_DYNAMIC_FTRACE_WITH_REGS,
it is not available if the architecture only supports
CONFIG_DYNAMIC_FTRACE_WITH_ARGS (e.g. arm64). And that means kprobe
events can not probe function entry/exit effectively on such architecture.
But this problem can be solved if the dynamic events supports fprobe events
because fprobe events doesn't use kprobe but ftrace via fprobe.

With this series, user can add new events on the entry and exit of kernel
functions (which can be ftraced). Unlike kprobe events, the fprobe events
can only probe the function entry and exit, the IP address will have some
offsets from the symbol address. And it can only trace the function args,
return value, and stacks. (no registers)
For probing function body, users can continue to use the kprobe events.

The fprobe events syntax is here;

 f[:[GRP/][EVENT]] FUNCTION [FETCHARGS]
 f[MAXACTIVE][:[GRP/][EVENT]] FUNCTION%return [FETCHARGS]

E.g.

 # echo 'f vfs_read $arg1'  >> dynamic_events
 # echo 'f vfs_read%return $retval'  >> dynamic_events
 # cat dynamic_events
 f:fprobes/vfs_read_entry vfs_read arg1=$arg1
 f:fprobes/vfs_read_exit vfs_read%return arg1=$retval
 # echo 1 > events/fprobes/enable
 # head -n 20 trace | tail
 #           TASK-PID     CPU#  |||||  TIMESTAMP  FUNCTION
 #              | |         |   |||||     |         |
              sh-142     [005] ...1.   448.386420: vfs_read_entry: (vfs_read+0x4/0x340) arg1=0xffff888007f7c540
              sh-142     [005] .....   448.386436: vfs_read_exit: (ksys_read+0x75/0x100 <- vfs_read) arg1=0x1
              sh-142     [005] ...1.   448.386451: vfs_read_entry: (vfs_read+0x4/0x340) arg1=0xffff888007f7c540
              sh-142     [005] .....   448.386458: vfs_read_exit: (ksys_read+0x75/0x100 <- vfs_read) arg1=0x1
              sh-142     [005] ...1.   448.386469: vfs_read_entry: (vfs_read+0x4/0x340) arg1=0xffff888007f7c540
              sh-142     [005] .....   448.386476: vfs_read_exit: (ksys_read+0x75/0x100 <- vfs_read) arg1=0x1
              sh-142     [005] ...1.   448.602073: vfs_read_entry: (vfs_read+0x4/0x340) arg1=0xffff888007f7c540
              sh-142     [005] .....   448.602089: vfs_read_exit: (ksys_read+0x75/0x100 <- vfs_read) arg1=0x1

Next things
-----------
 - Trace multiple function entry/exit (wildcard).
 - Integrate it with the function graph tracer.
 - Use ftrace_regs instead of pt_regs and remove dependency of
   CONFIG_DYNAMIC_FTRACE_WITH_REGS.
 - Support (limited) register access via ftrace_regs.
 - Support fprobe event by perf probe.
 - Support entry data accessing from exit event.
 - Support BTF for trace arguments.

This fprobe event may eventually replace the kprobe events for
function entry and exit on some archs (e.g. arm64).

Here's my current migration (kretprobe to fprobe) idea:

Phase 1. introduce fprobe events. (THIS)
Phase 2. introduce generic function graph shadow stack
Phase 3. Replace the rethook with function shadow stack
         and use ftrace_regs in fprobe handlers.
Phase 4. Extend this fprobe support to other archs.

Even if kretprobe event is replaced with fprobe event, tracefs user can
transparently use fprobe events for function entry/exit with 'p:...'
and 'r:...' syntax (for backward compatibility.)

Thank you,

---

Masami Hiramatsu (Google) (3):
      fprobe: Pass return address to the handlers
      tracing/probes: Add fprobe events for tracing function entry and exit.
      selftests/ftrace: Add fprobe related testcases


 include/linux/fprobe.h                             |   11 
 include/linux/rethook.h                            |    2 
 include/linux/trace_events.h                       |    3 
 kernel/kprobes.c                                   |    1 
 kernel/trace/Kconfig                               |   14 
 kernel/trace/Makefile                              |    1 
 kernel/trace/bpf_trace.c                           |    6 
 kernel/trace/fprobe.c                              |   17 
 kernel/trace/rethook.c                             |    3 
 kernel/trace/trace.c                               |    3 
 kernel/trace/trace.h                               |   11 
 kernel/trace/trace_fprobe.c                        | 1073 ++++++++++++++++++++
 kernel/trace/trace_probe.c                         |    4 
 kernel/trace/trace_probe.h                         |    3 
 lib/test_fprobe.c                                  |   10 
 samples/fprobe/fprobe_example.c                    |    6 
 .../ftrace/test.d/dynevent/add_remove_fprobe.tc    |   26 
 .../ftrace/test.d/dynevent/fprobe_syntax_errors.tc |   88 ++
 18 files changed, 1263 insertions(+), 19 deletions(-)
 create mode 100644 kernel/trace/trace_fprobe.c
 create mode 100644 tools/testing/selftests/ftrace/test.d/dynevent/add_remove_fprobe.tc
 create mode 100644 tools/testing/selftests/ftrace/test.d/dynevent/fprobe_syntax_errors.tc

--
Masami Hiramatsu (Google) <mhiramat@...nel.org>

Powered by blists - more mailing lists