linux-kernel - Re: [PATCH v2 0/3] tools/perf: Fix perf probe crash by clang DWARF5 file

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Tue, 1 Nov 2022 17:03:51 -0700
From:   Namhyung Kim <namhyung@...nel.org>
To:     "Masami Hiramatsu (Google)" <mhiramat@...nel.org>
Cc:     Arnaldo Carvalho de Melo <acme@...nel.org>,
        Peter Zijlstra <peterz@...radead.org>,
        Ingo Molnar <mingo@...hat.com>,
        Mark Rutland <mark.rutland@....com>,
        Alexander Shishkin <alexander.shishkin@...ux.intel.com>,
        Jiri Olsa <jolsa@...nel.org>, linux-perf-users@...r.kernel.org,
        linux-kernel@...r.kernel.org, Steven Rostedt <rostedt@...dmis.org>
Subject: Re: [PATCH v2 0/3] tools/perf: Fix perf probe crash by clang DWARF5 file

Hi Masami,

On Tue, Nov 1, 2022 at 6:48 AM Masami Hiramatsu (Google)
<mhiramat@...nel.org> wrote:
>
> Hi,
>
> Here is the 2nd version of the patches for perf probe which improves the
> robustness against clang DWARF5 file.
>
> Since the Clang generates a bit different DWARF5 file, the perf probe
> crashes or failes to analyze it. There are actually fragile code against
> it, so I fixed it ([1/3]) to avoid crash by SEGV. And make it accepts
> Clang's DWARF5 file ([2/3],[3/3]).
>
> Without this series, the perf probe crashes with the DWARF5 file
> which generated by clang as below;
>
>   $ ./perf probe -k $BIN_PATH/vmlinux -s $SRC_PATH -L vfs_read:10
>   Segmentation fault
>
> This series fixes it to handle such file correctly;
>
>   $ ./perf probe -k $BIN_PATH/vmlinux -s $SRC_PATH -L vfs_read:10
>   <vfs_read@...C_PATH/fs/read_write.c:10>
>
>        11         ret = rw_verify_area(READ, file, pos, count);
>        12         if (ret)
>                            return ret;
>
>
> On the DWARF5 specification, Sec 2.14, there is
> "The value 0 indicates that no source file has been specified."
> for DW_AT_decl_file, but clang generated DWARF5 will use the value 0.
>
> This issue is discussed on DWARF std ML;
> https://www.mail-archive.com/dwarf-discuss@lists.dwarfstd.org/msg00884.html
>
> And suggested that removing this part from the specification.
> http://wiki.dwarfstd.org/index.php?title=DWARF5_Line_Table_File_Numbers
>
> So as far as I understand, this is out of standard at this moment,
> but the standard itself has a discussion on this part. And maybe updated
> as currently clang does in the next release/revision.
>
> Thank you,
>
> ---
>
> Masami Hiramatsu (Google) (3):
>       tools/perf: Fix to avoid crashing if DW_AT_decl_file is NULL
>       tools/perf: Fix to use dwarf_attr_integrate for generic attr accessor
>       tools/perf: Fix to get declared file name from clang DWARF5

Acked-by: Namhyung Kim <namhyung@...nel.org>

Thanks,
Namhyung


>
>
>  tools/perf/util/dwarf-aux.c    |   58 ++++++++++++++++++++++++++++------------
>  tools/perf/util/dwarf-aux.h    |    3 ++
>  tools/perf/util/probe-finder.c |   37 +++++++++++++++++---------
>  3 files changed, 68 insertions(+), 30 deletions(-)
>
> --
> Masami Hiramatsu (Google) <mhiramat@...nel.org>