diff mbox series

[3/3] perf machine: Include data symbols in the kernel map

Message ID	20230620201818.1670753-3-namhyung@kernel.org
State	New
Headers	Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20; Sender: Namhyung Kim <namhyung@gmail.com> From: Namhyung Kim <namhyung@kernel.org> To: Arnaldo Carvalho de Melo <acme@kernel.org>, Jiri Olsa <jolsa@kernel.org> Cc: Ian Rogers <irogers@google.com>, Adrian Hunter <adrian.hunter@intel.com>, Peter Zijlstra <peterz@infradead.org>, Ingo Molnar <mingo@kernel.org>, LKML <linux-kernel@vger.kernel.org>, linux-perf-users@vger.kernel.org Subject: [PATCH 3/3] perf machine: Include data symbols in the kernel map Date: Tue, 20 Jun 2023 13:18:18 -0700 Message-ID: <20230620201818.1670753-3-namhyung@kernel.org> In-Reply-To: <20230620201818.1670753-1-namhyung@kernel.org> References: <20230620201818.1670753-1-namhyung@kernel.org> MIME-Version: 1.0 Content-Transfer-Encoding: 8bit Precedence: bulk
Series	[1/3] perf machine: Add machine->has_data_mmap field \| [1/3] perf machine: Add machine->has_data_mmap field [2/3] perf tools: Add kallsyms__get_symbol_start() [3/3] perf machine: Include data symbols in the kernel map

Commit Message

Namhyung Kim June 20, 2023, 8:18 p.m. UTC

  When perf record -d is used, it needs data mmaps to symbolize global data.
But it missed to collect kernel data maps so it cannot symbolize them.
Instead of having a separate map, just increase the kernel map size to
include the data section.

Probably we can have a separate kernel map for data, but the current
code assumes a single kernel map.  So it'd require more changes in other
places and looks error-prone.  I decided not to go that way for now.

Also it seems the kernel module size already includes the data section.

For example, my system has the following.

  $ grep -e _stext -e _etext -e _edata /proc/kallsyms
  ffffffff99800000 T _stext
  ffffffff9a601ac8 T _etext
  ffffffff9b446a00 D _edata

Size of the text section is (0x9a601ac8 - 0x99800000 = 0xe01ac8) and
size of the data section is (0x9b446a00 - 0x99800000 = 0x1c46a00).

Before:
  $ perf record -d true

  $ perf report -D | grep MMAP | head -1
  0 0 0x460 [0x60]: PERF_RECORD_MMAP -1/0: [0xffffffff99800000(0xe01ac8) @ 0xffffffff99800000]: x [kernel.kallsyms]_text
                                                               ^^^^^^^^
                                                                 here
After:
  $ perf report -D | grep MMAP | head -1
  0 0 0x460 [0x60]: PERF_RECORD_MMAP -1/0: [0xffffffff99800000(0x1c46a00) @ 0xffffffff99800000]: x [kernel.kallsyms]_text
                                                               ^^^^^^^^^

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
---
 tools/perf/util/machine.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

Comments

Ian Rogers June 22, 2023, 5:27 p.m. UTC | #1

On Tue, Jun 20, 2023 at 1:18 PM Namhyung Kim <namhyung@kernel.org> wrote:
>
> When perf record -d is used, it needs data mmaps to symbolize global data.
> But it missed to collect kernel data maps so it cannot symbolize them.
> Instead of having a separate map, just increase the kernel map size to
> include the data section.
>
> Probably we can have a separate kernel map for data, but the current
> code assumes a single kernel map.  So it'd require more changes in other
> places and looks error-prone.  I decided not to go that way for now.
>
> Also it seems the kernel module size already includes the data section.
>
> For example, my system has the following.
>
>   $ grep -e _stext -e _etext -e _edata /proc/kallsyms
>   ffffffff99800000 T _stext
>   ffffffff9a601ac8 T _etext
>   ffffffff9b446a00 D _edata
>
> Size of the text section is (0x9a601ac8 - 0x99800000 = 0xe01ac8) and
> size of the data section is (0x9b446a00 - 0x99800000 = 0x1c46a00).
>
> Before:
>   $ perf record -d true
>
>   $ perf report -D | grep MMAP | head -1
>   0 0 0x460 [0x60]: PERF_RECORD_MMAP -1/0: [0xffffffff99800000(0xe01ac8) @ 0xffffffff99800000]: x [kernel.kallsyms]_text
>                                                                ^^^^^^^^
>                                                                  here

nit: should the ^^^ be under 0xe01ac8?

> After:
>   $ perf report -D | grep MMAP | head -1
>   0 0 0x460 [0x60]: PERF_RECORD_MMAP -1/0: [0xffffffff99800000(0x1c46a00) @ 0xffffffff99800000]: x [kernel.kallsyms]_text
>                                                                ^^^^^^^^^

nit: and here under 0x1c46a00 ?

>
> Signed-off-by: Namhyung Kim <namhyung@kernel.org>

Acked-by: Ian Rogers <irogers@google.com>

Thanks,
Ian

> ---
>  tools/perf/util/machine.c | 5 ++++-
>  1 file changed, 4 insertions(+), 1 deletion(-)
>
> diff --git a/tools/perf/util/machine.c b/tools/perf/util/machine.c
> index ddc0a2130caf..e93a66f6e0b3 100644
> --- a/tools/perf/util/machine.c
> +++ b/tools/perf/util/machine.c
> @@ -1218,7 +1218,10 @@ static int machine__get_running_kernel_start(struct machine *machine,
>
>         *start = addr;
>
> -       err = kallsyms__get_function_start(filename, "_etext", &addr);
> +       if (machine->has_data_mmap)
> +               err = kallsyms__get_symbol_start(filename, "_edata", &addr);
> +       else
> +               err = kallsyms__get_function_start(filename, "_etext", &addr);
>         if (!err)
>                 *end = addr;
>
> --
> 2.41.0.185.g7c58973941-goog
>

Adrian Hunter July 11, 2023, 3:19 p.m. UTC | #2

On 20/06/23 23:18, Namhyung Kim wrote:
> When perf record -d is used, it needs data mmaps to symbolize global data.
> But it missed to collect kernel data maps so it cannot symbolize them.
> Instead of having a separate map, just increase the kernel map size to
> include the data section.
> 
> Probably we can have a separate kernel map for data, but the current
> code assumes a single kernel map.  So it'd require more changes in other
> places and looks error-prone.  I decided not to go that way for now.
> 
> Also it seems the kernel module size already includes the data section.
> 
> For example, my system has the following.
> 
>   $ grep -e _stext -e _etext -e _edata /proc/kallsyms
>   ffffffff99800000 T _stext
>   ffffffff9a601ac8 T _etext
>   ffffffff9b446a00 D _edata
> 
> Size of the text section is (0x9a601ac8 - 0x99800000 = 0xe01ac8) and
> size of the data section is (0x9b446a00 - 0x99800000 = 0x1c46a00).
> 
> Before:
>   $ perf record -d true
> 
>   $ perf report -D | grep MMAP | head -1
>   0 0 0x460 [0x60]: PERF_RECORD_MMAP -1/0: [0xffffffff99800000(0xe01ac8) @ 0xffffffff99800000]: x [kernel.kallsyms]_text
>                                                                ^^^^^^^^
>                                                                  here
> After:
>   $ perf report -D | grep MMAP | head -1
>   0 0 0x460 [0x60]: PERF_RECORD_MMAP -1/0: [0xffffffff99800000(0x1c46a00) @ 0xffffffff99800000]: x [kernel.kallsyms]_text
>                                                                ^^^^^^^^^
> 
> Signed-off-by: Namhyung Kim <namhyung@kernel.org>
> ---
>  tools/perf/util/machine.c | 5 ++++-
>  1 file changed, 4 insertions(+), 1 deletion(-)
> 
> diff --git a/tools/perf/util/machine.c b/tools/perf/util/machine.c
> index ddc0a2130caf..e93a66f6e0b3 100644
> --- a/tools/perf/util/machine.c
> +++ b/tools/perf/util/machine.c
> @@ -1218,7 +1218,10 @@ static int machine__get_running_kernel_start(struct machine *machine,
>  
>  	*start = addr;
>  
> -	err = kallsyms__get_function_start(filename, "_etext", &addr);
> +	if (machine->has_data_mmap)
> +		err = kallsyms__get_symbol_start(filename, "_edata", &addr);
> +	else
> +		err = kallsyms__get_function_start(filename, "_etext", &addr);

What is the downside of just extending it unconditionally?

>  	if (!err)
>  		*end = addr;
>

Namhyung Kim July 11, 2023, 5:30 p.m. UTC | #3

Hi Adrian,

On Tue, Jul 11, 2023 at 8:19 AM Adrian Hunter <adrian.hunter@intel.com> wrote:
>
> On 20/06/23 23:18, Namhyung Kim wrote:
> > When perf record -d is used, it needs data mmaps to symbolize global data.
> > But it missed to collect kernel data maps so it cannot symbolize them.
> > Instead of having a separate map, just increase the kernel map size to
> > include the data section.
> >
> > Probably we can have a separate kernel map for data, but the current
> > code assumes a single kernel map.  So it'd require more changes in other
> > places and looks error-prone.  I decided not to go that way for now.
> >
> > Also it seems the kernel module size already includes the data section.
> >
> > For example, my system has the following.
> >
> >   $ grep -e _stext -e _etext -e _edata /proc/kallsyms
> >   ffffffff99800000 T _stext
> >   ffffffff9a601ac8 T _etext
> >   ffffffff9b446a00 D _edata
> >
> > Size of the text section is (0x9a601ac8 - 0x99800000 = 0xe01ac8) and
> > size of the data section is (0x9b446a00 - 0x99800000 = 0x1c46a00).
> >
> > Before:
> >   $ perf record -d true
> >
> >   $ perf report -D | grep MMAP | head -1
> >   0 0 0x460 [0x60]: PERF_RECORD_MMAP -1/0: [0xffffffff99800000(0xe01ac8) @ 0xffffffff99800000]: x [kernel.kallsyms]_text
> >                                                                ^^^^^^^^
> >                                                                  here
> > After:
> >   $ perf report -D | grep MMAP | head -1
> >   0 0 0x460 [0x60]: PERF_RECORD_MMAP -1/0: [0xffffffff99800000(0x1c46a00) @ 0xffffffff99800000]: x [kernel.kallsyms]_text
> >                                                                ^^^^^^^^^
> >
> > Signed-off-by: Namhyung Kim <namhyung@kernel.org>
> > ---
> >  tools/perf/util/machine.c | 5 ++++-
> >  1 file changed, 4 insertions(+), 1 deletion(-)
> >
> > diff --git a/tools/perf/util/machine.c b/tools/perf/util/machine.c
> > index ddc0a2130caf..e93a66f6e0b3 100644
> > --- a/tools/perf/util/machine.c
> > +++ b/tools/perf/util/machine.c
> > @@ -1218,7 +1218,10 @@ static int machine__get_running_kernel_start(struct machine *machine,
> >
> >       *start = addr;
> >
> > -     err = kallsyms__get_function_start(filename, "_etext", &addr);
> > +     if (machine->has_data_mmap)
> > +             err = kallsyms__get_symbol_start(filename, "_edata", &addr);
> > +     else
> > +             err = kallsyms__get_function_start(filename, "_etext", &addr);
>
> What is the downside of just extending it unconditionally?

I don't know.. maybe some people would argue it needs the
proper protection bits other than 'x' but this patch also breaks it.
But as I said, I'm not sure if we really want to change that now.

That said, we can make it unconditional. :)

Thanks,
Namhyung

>
> >       if (!err)
> >               *end = addr;
> >
>

Adrian Hunter July 12, 2023, 5:44 a.m. UTC | #4

On 11/07/23 20:30, Namhyung Kim wrote:
> Hi Adrian,
> 
> On Tue, Jul 11, 2023 at 8:19 AM Adrian Hunter <adrian.hunter@intel.com> wrote:
>>
>> On 20/06/23 23:18, Namhyung Kim wrote:
>>> When perf record -d is used, it needs data mmaps to symbolize global data.
>>> But it missed to collect kernel data maps so it cannot symbolize them.
>>> Instead of having a separate map, just increase the kernel map size to
>>> include the data section.
>>>
>>> Probably we can have a separate kernel map for data, but the current
>>> code assumes a single kernel map.  So it'd require more changes in other
>>> places and looks error-prone.  I decided not to go that way for now.
>>>
>>> Also it seems the kernel module size already includes the data section.
>>>
>>> For example, my system has the following.
>>>
>>>   $ grep -e _stext -e _etext -e _edata /proc/kallsyms
>>>   ffffffff99800000 T _stext
>>>   ffffffff9a601ac8 T _etext
>>>   ffffffff9b446a00 D _edata
>>>
>>> Size of the text section is (0x9a601ac8 - 0x99800000 = 0xe01ac8) and
>>> size of the data section is (0x9b446a00 - 0x99800000 = 0x1c46a00).
>>>
>>> Before:
>>>   $ perf record -d true
>>>
>>>   $ perf report -D | grep MMAP | head -1
>>>   0 0 0x460 [0x60]: PERF_RECORD_MMAP -1/0: [0xffffffff99800000(0xe01ac8) @ 0xffffffff99800000]: x [kernel.kallsyms]_text
>>>                                                                ^^^^^^^^
>>>                                                                  here
>>> After:
>>>   $ perf report -D | grep MMAP | head -1
>>>   0 0 0x460 [0x60]: PERF_RECORD_MMAP -1/0: [0xffffffff99800000(0x1c46a00) @ 0xffffffff99800000]: x [kernel.kallsyms]_text
>>>                                                                ^^^^^^^^^
>>>
>>> Signed-off-by: Namhyung Kim <namhyung@kernel.org>
>>> ---
>>>  tools/perf/util/machine.c | 5 ++++-
>>>  1 file changed, 4 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/tools/perf/util/machine.c b/tools/perf/util/machine.c
>>> index ddc0a2130caf..e93a66f6e0b3 100644
>>> --- a/tools/perf/util/machine.c
>>> +++ b/tools/perf/util/machine.c
>>> @@ -1218,7 +1218,10 @@ static int machine__get_running_kernel_start(struct machine *machine,
>>>
>>>       *start = addr;
>>>
>>> -     err = kallsyms__get_function_start(filename, "_etext", &addr);
>>> +     if (machine->has_data_mmap)
>>> +             err = kallsyms__get_symbol_start(filename, "_edata", &addr);
>>> +     else
>>> +             err = kallsyms__get_function_start(filename, "_etext", &addr);
>>
>> What is the downside of just extending it unconditionally?
> 
> I don't know.. maybe some people would argue it needs the
> proper protection bits other than 'x' but this patch also breaks it.
> But as I said, I'm not sure if we really want to change that now.
> 
> That said, we can make it unconditional. :)

Might as well to start with.  Will need a big comment.

Also do we know if all arch's do it like that? Perhaps
need to fallback to _etext if _edata is not found?

> 
> Thanks,
> Namhyung
> 
>>
>>>       if (!err)
>>>               *end = addr;
>>>
>>

diff mbox series

Patch

diff --git a/tools/perf/util/machine.c b/tools/perf/util/machine.c
index ddc0a2130caf..e93a66f6e0b3 100644
--- a/tools/perf/util/machine.c
+++ b/tools/perf/util/machine.c
@@ -1218,7 +1218,10 @@  static int machine__get_running_kernel_start(struct machine *machine,
 
 	*start = addr;
 
-	err = kallsyms__get_function_start(filename, "_etext", &addr);
+	if (machine->has_data_mmap)
+		err = kallsyms__get_symbol_start(filename, "_edata", &addr);
+	else
+		err = kallsyms__get_function_start(filename, "_etext", &addr);
 	if (!err)
 		*end = addr;