public inbox for systemtap@sourceware.org
 help / color / mirror / Atom feed
From: BR Chrisman <brchrisman@gmail.com>
To: Honggyu Kim <hong.gyu.kim@lge.com>
Cc: systemtap@sourceware.org
Subject: Re: [PATCH] dwflpp: insert a kprobe into an alternative line
Date: Thu, 31 Jul 2014 05:28:00 -0000	[thread overview]
Message-ID: <CAN4=B2kWmsd0GVpbhSzyGNBSB7mwsdXCfN9EWbNR-P9qZLZkbA@mail.gmail.com> (raw)
In-Reply-To: <1406766254-23556-1-git-send-email-hong.gyu.kim@lge.com>

I don't generally want systemtap to automatically change my requested
probe point.  Maybe implement as an option?

On Wed, Jul 30, 2014 at 5:24 PM, Honggyu Kim <hong.gyu.kim@lge.com> wrote:
> If a line number is given in 'statement', sometimes the line is not
> available in dwarf line number.
> Line records in dwarf may not be found for a given line number.
> In this case, alternative line numbers were suggested and exited.
>
> Rather than this, this patch makes systemtap insert a kprobe into the
> nearest line number that is available in dwarf information.
> It is implemented by replacing suggest_alternative_linenos() to
> insert_alternative_linenos().
>
> Signed-off-by: Honggyu Kim <hong.gyu.kim@lge.com>
> ---
>  dwflpp.cxx |   72 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
>  dwflpp.h   |    6 +++++
>  2 files changed, 78 insertions(+)
>
> diff --git a/dwflpp.cxx b/dwflpp.cxx
> index e53d5f2..4808f0b 100644
> --- a/dwflpp.cxx
> +++ b/dwflpp.cxx
> @@ -1776,6 +1776,72 @@ get_funcs_in_srcfile(base_func_info_map_t& funcs,
>    return matching_funcs;
>  }
>
> +void
> +dwflpp::insert_alternative_linenos(char const * srcfile,
> +                                    int lineno,
> +                                    base_func_info_map_t& funcs,
> +                                    void (* callback) (Dwarf_Addr,
> +                                                       int, void*),
> +                                    void *data)
> +{
> +  assert(cu);
> +  lines_t *cu_lines = get_cu_lines_sorted_by_lineno(srcfile);
> +
> +  // Look around lineno for linenos with LRs.
> +  int lo_try = -1;
> +  int hi_try = -1;
> +  for (size_t i = 1; i < 6; ++i)
> +    {
> +      if (lo_try == -1 && functions_have_lineno(funcs, cu_lines, lineno-i))
> +        lo_try = lineno - i;
> +      if (hi_try == -1 && functions_have_lineno(funcs, cu_lines, lineno+i))
> +        hi_try = lineno + i;
> +    }
> +
> +  int new_lineno;
> +  if (lo_try > 0 && hi_try > 0)
> +    {
> +      // pick the nearest line number
> +      if (lineno - lo_try < lineno - hi_try)
> +        new_lineno = lo_try;
> +      else
> +        new_lineno = hi_try;
> +    }
> +  else if (lo_try > 0)
> +    {
> +      new_lineno = lo_try;
> +    }
> +  else if (hi_try > 0)
> +    {
> +      new_lineno = hi_try;
> +    }
> +  else
> +    {
> +      stringstream advice;
> +      advice << _F("no line records for %s:%d [man error::dwarf]", srcfile, lineno);
> +      throw SEMANTIC_ERROR (advice.str());
> +    }
> +  clog << _F("insert a kprobe at %s:%d (no line record at %s:%d)\n",
> +             srcfile, new_lineno, srcfile, lineno);
> +
> +  // collect lines
> +  lines_t matching_lines;
> +  collect_lines_for_single_lineno(srcfile, new_lineno, false, /* is_relative */
> +                                  funcs, matching_lines);
> +  // call back with matching lines
> +  assert (!matching_lines.empty());
> +  set<Dwarf_Addr> probed_addrs;
> +  for (lines_t::iterator line  = matching_lines.begin();
> +                         line != matching_lines.end(); ++line)
> +    {
> +      int near_lineno = DWARF_LINENO(*line);
> +      Dwarf_Addr addr = DWARF_LINEADDR(*line);
> +      bool is_new_addr = probed_addrs.insert(addr).second;
> +      if (is_new_addr)
> +        callback(addr, near_lineno, data);
> +    }
> +}
> +
>  template<> void
>  dwflpp::iterate_over_srcfile_lines<void>(char const * srcfile,
>                                           const vector<int>& linenos,
> @@ -1850,7 +1916,13 @@ dwflpp::iterate_over_srcfile_lines<void>(char const * srcfile,
>        if (lineno_type == RELATIVE)
>          // just pick the first function and make it relative to that
>          lineno += current_funcs[0].decl_line;
> +#if 0
>        suggest_alternative_linenos(srcfile, lineno, current_funcs);
> +#else
> +      // This function inserts a breakpoint for a near line number
> +      // rather than suggesting near line numbers and exiting.
> +      insert_alternative_linenos(srcfile, lineno, current_funcs, callback, data);
> +#endif
>      }
>  }
>
> diff --git a/dwflpp.h b/dwflpp.h
> index 2546acf..dc7dc93 100644
> --- a/dwflpp.h
> +++ b/dwflpp.h
> @@ -559,6 +559,12 @@ private:
>    void suggest_alternative_linenos(char const * srcfile,
>                                     int lineno,
>                                     base_func_info_map_t& funcs);
> +  void insert_alternative_linenos(char const * srcfile,
> +                                   int lineno,
> +                                   base_func_info_map_t& funcs,
> +                                   void (* callback) (Dwarf_Addr,
> +                                                      int, void*),
> +                                   void *data);
>
>    static int external_function_cu_callback (Dwarf_Die* cu, external_function_query *efq);
>    static int external_function_func_callback (Dwarf_Die* func, external_function_query *efq);
> --
> 1.7.9.5
>

  reply	other threads:[~2014-07-31  5:28 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2014-07-31  0:24 Honggyu Kim
2014-07-31  5:28 ` BR Chrisman [this message]
2014-07-31 12:24   ` Frank Ch. Eigler
2014-07-31 13:28     ` Honggyu Kim

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='CAN4=B2kWmsd0GVpbhSzyGNBSB7mwsdXCfN9EWbNR-P9qZLZkbA@mail.gmail.com' \
    --to=brchrisman@gmail.com \
    --cc=hong.gyu.kim@lge.com \
    --cc=systemtap@sourceware.org \
    /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;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).