[04/14] perf map: Add map__objdump_2rip()

  Sometimes we want to convert an address in objdump output to
map-relative address to match with a sample data.  Let's add
map__objdump_2rip() for that.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
---
 tools/perf/util/map.c | 20 ++++++++++++++++++++
 tools/perf/util/map.h |  3 +++
 2 files changed, 23 insertions(+)
  

On Fri, Feb 2, 2024 at 2:05 PM Namhyung Kim <namhyung@kernel.org> wrote:
>
> Sometimes we want to convert an address in objdump output to
> map-relative address to match with a sample data.  Let's add
> map__objdump_2rip() for that.

Hi Namhyung,

I think the naming can be better here. Aren't the objdump addresses
DSO relative offsets? Is the relative IP relative to the map or the
DSO?

Thanks,
Ian

> Cc: Adrian Hunter <adrian.hunter@intel.com>
> Signed-off-by: Namhyung Kim <namhyung@kernel.org>
> ---
>  tools/perf/util/map.c | 20 ++++++++++++++++++++
>  tools/perf/util/map.h |  3 +++
>  2 files changed, 23 insertions(+)
>
> diff --git a/tools/perf/util/map.c b/tools/perf/util/map.c
> index 54c67cb7ecef..66542864b7b5 100644
> --- a/tools/perf/util/map.c
> +++ b/tools/perf/util/map.c
> @@ -594,6 +594,26 @@ u64 map__objdump_2mem(struct map *map, u64 ip)
>         return ip + map__reloc(map);
>  }
>
> +u64 map__objdump_2rip(struct map *map, u64 ip)
> +{
> +       const struct dso *dso = map__dso(map);
> +
> +       if (!dso->adjust_symbols)
> +               return ip;
> +
> +       if (dso->rel)
> +               return ip + map__pgoff(map);
> +
> +       /*
> +        * kernel modules also have DSO_TYPE_USER in dso->kernel,
> +        * but all kernel modules are ET_REL, so won't get here.
> +        */
> +       if (dso->kernel == DSO_SPACE__USER)
> +               return ip - dso->text_offset;
> +
> +       return map__map_ip(map, ip + map__reloc(map));
> +}
> +
>  bool map__contains_symbol(const struct map *map, const struct symbol *sym)
>  {
>         u64 ip = map__unmap_ip(map, sym->start);
> diff --git a/tools/perf/util/map.h b/tools/perf/util/map.h
> index 49756716cb13..65e2609fa1b1 100644
> --- a/tools/perf/util/map.h
> +++ b/tools/perf/util/map.h
> @@ -132,6 +132,9 @@ u64 map__rip_2objdump(struct map *map, u64 rip);
>  /* objdump address -> memory address */
>  u64 map__objdump_2mem(struct map *map, u64 ip);
>
> +/* objdump address -> rip */
> +u64 map__objdump_2rip(struct map *map, u64 ip);
> +
>  struct symbol;
>  struct thread;
>
> --
> 2.43.0.594.gd9cf4e227d-goog
>
  

Hi Ian,

On Fri, Feb 2, 2024 at 5:42 PM Ian Rogers <irogers@google.com> wrote:
>
> On Fri, Feb 2, 2024 at 2:05 PM Namhyung Kim <namhyung@kernel.org> wrote:
> >
> > Sometimes we want to convert an address in objdump output to
> > map-relative address to match with a sample data.  Let's add
> > map__objdump_2rip() for that.
>
> Hi Namhyung,
>
> I think the naming can be better here. Aren't the objdump addresses
> DSO relative offsets? Is the relative IP relative to the map or the
> DSO?

AFAIK the objdump addresses are DSO-relative and rip is to map.
They are mostly the same but sometimes different due to kASLR
for the kernel.

>
> > Cc: Adrian Hunter <adrian.hunter@intel.com>
> > Signed-off-by: Namhyung Kim <namhyung@kernel.org>
> > ---
> >  tools/perf/util/map.c | 20 ++++++++++++++++++++
> >  tools/perf/util/map.h |  3 +++
> >  2 files changed, 23 insertions(+)
> >
> > diff --git a/tools/perf/util/map.c b/tools/perf/util/map.c
> > index 54c67cb7ecef..66542864b7b5 100644
> > --- a/tools/perf/util/map.c
> > +++ b/tools/perf/util/map.c
> > @@ -594,6 +594,26 @@ u64 map__objdump_2mem(struct map *map, u64 ip)
> >         return ip + map__reloc(map);
> >  }
> >
> > +u64 map__objdump_2rip(struct map *map, u64 ip)
> > +{
> > +       const struct dso *dso = map__dso(map);
> > +
> > +       if (!dso->adjust_symbols)
> > +               return ip;
> > +
> > +       if (dso->rel)
> > +               return ip + map__pgoff(map);
> > +
> > +       /*
> > +        * kernel modules also have DSO_TYPE_USER in dso->kernel,
> > +        * but all kernel modules are ET_REL, so won't get here.
> > +        */

Hmm.. This comment is not true anymore.  Will remove
in other places too.

Thanks,
Namhyung


> > +       if (dso->kernel == DSO_SPACE__USER)
> > +               return ip - dso->text_offset;
> > +
> > +       return map__map_ip(map, ip + map__reloc(map));
> > +}
> > +
> >  bool map__contains_symbol(const struct map *map, const struct symbol *sym)
> >  {
> >         u64 ip = map__unmap_ip(map, sym->start);
> > diff --git a/tools/perf/util/map.h b/tools/perf/util/map.h
> > index 49756716cb13..65e2609fa1b1 100644
> > --- a/tools/perf/util/map.h
> > +++ b/tools/perf/util/map.h
> > @@ -132,6 +132,9 @@ u64 map__rip_2objdump(struct map *map, u64 rip);
> >  /* objdump address -> memory address */
> >  u64 map__objdump_2mem(struct map *map, u64 ip);
> >
> > +/* objdump address -> rip */
> > +u64 map__objdump_2rip(struct map *map, u64 ip);
> > +
> >  struct symbol;
> >  struct thread;
> >
> > --
> > 2.43.0.594.gd9cf4e227d-goog
> >
  

On Tue, Feb 6, 2024 at 3:04 PM Namhyung Kim <namhyung@kernel.org> wrote:
>
> Hi Ian,
>
> On Fri, Feb 2, 2024 at 5:42 PM Ian Rogers <irogers@google.com> wrote:
> >
> > On Fri, Feb 2, 2024 at 2:05 PM Namhyung Kim <namhyung@kernel.org> wrote:
> > >
> > > Sometimes we want to convert an address in objdump output to
> > > map-relative address to match with a sample data.  Let's add
> > > map__objdump_2rip() for that.
> >
> > Hi Namhyung,
> >
> > I think the naming can be better here. Aren't the objdump addresses
> > DSO relative offsets? Is the relative IP relative to the map or the
> > DSO?
>
> AFAIK the objdump addresses are DSO-relative and rip is to map.
> They are mostly the same but sometimes different due to kASLR
> for the kernel.

Perhaps we need to use names like map_rip for mapping relative and
dso_rip to clean this up, or to add a different mapping_type to the
enum. For non-kernel maps addresses for map are either the whole
virtual address space (identity) or relative to a dso:
https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/util/map.h?h=perf-tools-next#n115
https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/util/map.h?h=perf-tools-next#n20
The dso addresses should work for objdump so perhaps the kernel
addresses need map__pgoff fixing?

Thanks,
Ian

> >
> > > Cc: Adrian Hunter <adrian.hunter@intel.com>
> > > Signed-off-by: Namhyung Kim <namhyung@kernel.org>
> > > ---
> > >  tools/perf/util/map.c | 20 ++++++++++++++++++++
> > >  tools/perf/util/map.h |  3 +++
> > >  2 files changed, 23 insertions(+)
> > >
> > > diff --git a/tools/perf/util/map.c b/tools/perf/util/map.c
> > > index 54c67cb7ecef..66542864b7b5 100644
> > > --- a/tools/perf/util/map.c
> > > +++ b/tools/perf/util/map.c
> > > @@ -594,6 +594,26 @@ u64 map__objdump_2mem(struct map *map, u64 ip)
> > >         return ip + map__reloc(map);
> > >  }
> > >
> > > +u64 map__objdump_2rip(struct map *map, u64 ip)
> > > +{
> > > +       const struct dso *dso = map__dso(map);
> > > +
> > > +       if (!dso->adjust_symbols)
> > > +               return ip;
> > > +
> > > +       if (dso->rel)
> > > +               return ip + map__pgoff(map);
> > > +
> > > +       /*
> > > +        * kernel modules also have DSO_TYPE_USER in dso->kernel,
> > > +        * but all kernel modules are ET_REL, so won't get here.
> > > +        */
>
> Hmm.. This comment is not true anymore.  Will remove
> in other places too.
>
> Thanks,
> Namhyung
>
>
> > > +       if (dso->kernel == DSO_SPACE__USER)
> > > +               return ip - dso->text_offset;
> > > +
> > > +       return map__map_ip(map, ip + map__reloc(map));
> > > +}
> > > +
> > >  bool map__contains_symbol(const struct map *map, const struct symbol *sym)
> > >  {
> > >         u64 ip = map__unmap_ip(map, sym->start);
> > > diff --git a/tools/perf/util/map.h b/tools/perf/util/map.h
> > > index 49756716cb13..65e2609fa1b1 100644
> > > --- a/tools/perf/util/map.h
> > > +++ b/tools/perf/util/map.h
> > > @@ -132,6 +132,9 @@ u64 map__rip_2objdump(struct map *map, u64 rip);
> > >  /* objdump address -> memory address */
> > >  u64 map__objdump_2mem(struct map *map, u64 ip);
> > >
> > > +/* objdump address -> rip */
> > > +u64 map__objdump_2rip(struct map *map, u64 ip);
> > > +
> > >  struct symbol;
> > >  struct thread;
> > >
> > > --
> > > 2.43.0.594.gd9cf4e227d-goog
> > >
  

On Tue, Feb 6, 2024 at 3:34 PM Ian Rogers <irogers@google.com> wrote:
>
> On Tue, Feb 6, 2024 at 3:04 PM Namhyung Kim <namhyung@kernel.org> wrote:
> >
> > Hi Ian,
> >
> > On Fri, Feb 2, 2024 at 5:42 PM Ian Rogers <irogers@google.com> wrote:
> > >
> > > On Fri, Feb 2, 2024 at 2:05 PM Namhyung Kim <namhyung@kernel.org> wrote:
> > > >
> > > > Sometimes we want to convert an address in objdump output to
> > > > map-relative address to match with a sample data.  Let's add
> > > > map__objdump_2rip() for that.
> > >
> > > Hi Namhyung,
> > >
> > > I think the naming can be better here. Aren't the objdump addresses
> > > DSO relative offsets? Is the relative IP relative to the map or the
> > > DSO?
> >
> > AFAIK the objdump addresses are DSO-relative and rip is to map.
> > They are mostly the same but sometimes different due to kASLR
> > for the kernel.
>
> Perhaps we need to use names like map_rip for mapping relative and
> dso_rip to clean this up, or to add a different mapping_type to the
> enum. For non-kernel maps addresses for map are either the whole
> virtual address space (identity) or relative to a dso:
> https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/util/map.h?h=perf-tools-next#n115
> https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/util/map.h?h=perf-tools-next#n20
> The dso addresses should work for objdump so perhaps the kernel
> addresses need map__pgoff fixing?

I'm not sure about the vDSO case.

By the way, I need to take a look if we can make this objdump-rip
thing simpler as you mentioned.  My feeling is that it can be done
but I'd like to do it in a separate work and to move this forward.

Thanks,
Namhyung
  

On Wed, Feb 7, 2024 at 11:04 AM Namhyung Kim <namhyung@kernel.org> wrote:
>
> On Tue, Feb 6, 2024 at 3:34 PM Ian Rogers <irogers@google.com> wrote:
> >
> > On Tue, Feb 6, 2024 at 3:04 PM Namhyung Kim <namhyung@kernel.org> wrote:
> > >
> > > Hi Ian,
> > >
> > > On Fri, Feb 2, 2024 at 5:42 PM Ian Rogers <irogers@google.com> wrote:
> > > >
> > > > On Fri, Feb 2, 2024 at 2:05 PM Namhyung Kim <namhyung@kernel.org> wrote:
> > > > >
> > > > > Sometimes we want to convert an address in objdump output to
> > > > > map-relative address to match with a sample data.  Let's add
> > > > > map__objdump_2rip() for that.
> > > >
> > > > Hi Namhyung,
> > > >
> > > > I think the naming can be better here. Aren't the objdump addresses
> > > > DSO relative offsets? Is the relative IP relative to the map or the
> > > > DSO?
> > >
> > > AFAIK the objdump addresses are DSO-relative and rip is to map.
> > > They are mostly the same but sometimes different due to kASLR
> > > for the kernel.
> >
> > Perhaps we need to use names like map_rip for mapping relative and
> > dso_rip to clean this up, or to add a different mapping_type to the
> > enum. For non-kernel maps addresses for map are either the whole
> > virtual address space (identity) or relative to a dso:
> > https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/util/map.h?h=perf-tools-next#n115
> > https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/util/map.h?h=perf-tools-next#n20
> > The dso addresses should work for objdump so perhaps the kernel
> > addresses need map__pgoff fixing?
>
> I'm not sure about the vDSO case.
>
> By the way, I need to take a look if we can make this objdump-rip
> thing simpler as you mentioned.  My feeling is that it can be done
> but I'd like to do it in a separate work and to move this forward.

Makes sense. Could we add a comment to the header file definition to
capture some of this? Perhaps a "to be removed" to avoid later patches
adding dependencies to the function.

Thanks,
Ian

> Thanks,
> Namhyung
  

On Wed, Feb 7, 2024 at 11:56 AM Ian Rogers <irogers@google.com> wrote:
>
> On Wed, Feb 7, 2024 at 11:04 AM Namhyung Kim <namhyung@kernel.org> wrote:
> >
> > On Tue, Feb 6, 2024 at 3:34 PM Ian Rogers <irogers@google.com> wrote:
> > >
> > > On Tue, Feb 6, 2024 at 3:04 PM Namhyung Kim <namhyung@kernel.org> wrote:
> > > >
> > > > Hi Ian,
> > > >
> > > > On Fri, Feb 2, 2024 at 5:42 PM Ian Rogers <irogers@google.com> wrote:
> > > > >
> > > > > On Fri, Feb 2, 2024 at 2:05 PM Namhyung Kim <namhyung@kernel.org> wrote:
> > > > > >
> > > > > > Sometimes we want to convert an address in objdump output to
> > > > > > map-relative address to match with a sample data.  Let's add
> > > > > > map__objdump_2rip() for that.
> > > > >
> > > > > Hi Namhyung,
> > > > >
> > > > > I think the naming can be better here. Aren't the objdump addresses
> > > > > DSO relative offsets? Is the relative IP relative to the map or the
> > > > > DSO?
> > > >
> > > > AFAIK the objdump addresses are DSO-relative and rip is to map.
> > > > They are mostly the same but sometimes different due to kASLR
> > > > for the kernel.
> > >
> > > Perhaps we need to use names like map_rip for mapping relative and
> > > dso_rip to clean this up, or to add a different mapping_type to the
> > > enum. For non-kernel maps addresses for map are either the whole
> > > virtual address space (identity) or relative to a dso:
> > > https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/util/map.h?h=perf-tools-next#n115
> > > https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/util/map.h?h=perf-tools-next#n20
> > > The dso addresses should work for objdump so perhaps the kernel
> > > addresses need map__pgoff fixing?
> >
> > I'm not sure about the vDSO case.
> >
> > By the way, I need to take a look if we can make this objdump-rip
> > thing simpler as you mentioned.  My feeling is that it can be done
> > but I'd like to do it in a separate work and to move this forward.
>
> Makes sense. Could we add a comment to the header file definition to
> capture some of this? Perhaps a "to be removed" to avoid later patches
> adding dependencies to the function.

Sure, will add that in v6.

Thanks,
Namhyung