dlinfo

dlinfo

[Justin Cormack]

[-- Attachment #1: Type: text/plain, Size: 54 bytes --]

Has anyone got any plans to implement dlinfo?

Justin

[-- Attachment #2: Type: text/html, Size: 92 bytes --]

Re: dlinfo

[Rich Felker]

On Sat, Jun 29, 2013 at 04:40:17PM +0100, Justin Cormack wrote:
> Has anyone got any plans to implement dlinfo?

I'm not opposed to it, but not sure how easy it would be.
RTLD_DI_LINKMAP looks trivial (just return the argument passed in),
but the origin and search information is stuff that's probably not
saved with the current dynamic linker implementation.

Rich

Re: dlinfo

[Justin Cormack]

[-- Attachment #1: Type: text/plain, Size: 581 bytes --]

On 29 Jun 2013 16:54, "Rich Felker" <dalias@aerifal.cx> wrote:
>
> On Sat, Jun 29, 2013 at 04:40:17PM +0100, Justin Cormack wrote:
> > Has anyone got any plans to implement dlinfo?
>
> I'm not opposed to it, but not sure how easy it would be.
> RTLD_DI_LINKMAP looks trivial (just return the argument passed in),
> but the origin and search information is stuff that's probably not
> saved with the current dynamic linker implementation.

Linkmap is the only bit I need actually. NetBSD apparently only implements
that and not the rest, so it is not unprecedented.

> Rich

Justin

[-- Attachment #2: Type: text/html, Size: 781 bytes --]

Re: dlinfo

[Rich Felker]

On Sat, Jun 29, 2013 at 05:14:16PM +0100, Justin Cormack wrote:
> On 29 Jun 2013 16:54, "Rich Felker" <dalias@aerifal.cx> wrote:
> >
> > On Sat, Jun 29, 2013 at 04:40:17PM +0100, Justin Cormack wrote:
> > > Has anyone got any plans to implement dlinfo?
> >
> > I'm not opposed to it, but not sure how easy it would be.
> > RTLD_DI_LINKMAP looks trivial (just return the argument passed in),
> > but the origin and search information is stuff that's probably not
> > saved with the current dynamic linker implementation.
> 
> Linkmap is the only bit I need actually. NetBSD apparently only implements
> that and not the rest, so it is not unprecedented.

As a quick workaround then:

-D'dlinfo(x,y,z)=(*(struct link_map *)(z) = (void *)(x))'

Kidding aside, I wouldn't actually recommend this, since you'd be
encoding implementation internals (the fact that the dso handle is
actually the link_map pointer) into the application, which could badly
break if the implementation is ever changed. I'll just go ahead and
add this limited version of dlinfo; look for it soon and ping me if
you don't see it.

Rich

Re: dlinfo

[Justin Cormack]

[-- Attachment #1: Type: text/plain, Size: 1227 bytes --]

On 29 Jun 2013 17:31, "Rich Felker" <dalias@aerifal.cx> wrote:
>
> On Sat, Jun 29, 2013 at 05:14:16PM +0100, Justin Cormack wrote:
> > On 29 Jun 2013 16:54, "Rich Felker" <dalias@aerifal.cx> wrote:
> > >
> > > On Sat, Jun 29, 2013 at 04:40:17PM +0100, Justin Cormack wrote:
> > > > Has anyone got any plans to implement dlinfo?
> > >
> > > I'm not opposed to it, but not sure how easy it would be.
> > > RTLD_DI_LINKMAP looks trivial (just return the argument passed in),
> > > but the origin and search information is stuff that's probably not
> > > saved with the current dynamic linker implementation.
> >
> > Linkmap is the only bit I need actually. NetBSD apparently only
implements
> > that and not the rest, so it is not unprecedented.
>
> As a quick workaround then:
>
> -D'dlinfo(x,y,z)=(*(struct link_map *)(z) = (void *)(x))'
>
> Kidding aside, I wouldn't actually recommend this, since you'd be
> encoding implementation internals (the fact that the dso handle is
> actually the link_map pointer) into the application, which could badly
> break if the implementation is ever changed. I'll just go ahead and
> add this limited version of dlinfo; look for it soon and ping me if
> you don't see it.
>
> Rich

Thanks!

[-- Attachment #2: Type: text/html, Size: 1659 bytes --]

Re: dlinfo

[Rich Felker]

On Sat, Jun 29, 2013 at 05:34:46PM +0100, Justin Cormack wrote:
> > As a quick workaround then:
> >
> > -D'dlinfo(x,y,z)=(*(struct link_map *)(z) = (void *)(x))'
> >
> > Kidding aside, I wouldn't actually recommend this, since you'd be
> > encoding implementation internals (the fact that the dso handle is
> > actually the link_map pointer) into the application, which could badly
> > break if the implementation is ever changed. I'll just go ahead and
> > add this limited version of dlinfo; look for it soon and ping me if
> > you don't see it.
> 
> Thanks!

One more question: does RTLD_SELF need to work? That's a lot harder
and uglier, as it requires extra asm for each arch...

Rich

Re: dlinfo

[Justin Cormack]

[-- Attachment #1: Type: text/plain, Size: 815 bytes --]

On 29 Jun 2013 17:39, "Rich Felker" <dalias@aerifal.cx> wrote:
>
> On Sat, Jun 29, 2013 at 05:34:46PM +0100, Justin Cormack wrote:
> > > As a quick workaround then:
> > >
> > > -D'dlinfo(x,y,z)=(*(struct link_map *)(z) = (void *)(x))'
> > >
> > > Kidding aside, I wouldn't actually recommend this, since you'd be
> > > encoding implementation internals (the fact that the dso handle is
> > > actually the link_map pointer) into the application, which could badly
> > > break if the implementation is ever changed. I'll just go ahead and
> > > add this limited version of dlinfo; look for it soon and ping me if
> > > you don't see it.
> >
> > Thanks!
>
> One more question: does RTLD_SELF need to work? That's a lot harder
> and uglier, as it requires extra asm for each arch...
>
> Rich

I don't think so.

Justin

[-- Attachment #2: Type: text/html, Size: 1155 bytes --]

Re: dlinfo

[Rich Felker]

On Sat, Jun 29, 2013 at 05:43:12PM +0100, Justin Cormack wrote:
> On 29 Jun 2013 17:39, "Rich Felker" <dalias@aerifal.cx> wrote:
> >
> > On Sat, Jun 29, 2013 at 05:34:46PM +0100, Justin Cormack wrote:
> > > > As a quick workaround then:
> > > >
> > > > -D'dlinfo(x,y,z)=(*(struct link_map *)(z) = (void *)(x))'
> > > >
> > > > Kidding aside, I wouldn't actually recommend this, since you'd be
> > > > encoding implementation internals (the fact that the dso handle is
> > > > actually the link_map pointer) into the application, which could badly
> > > > break if the implementation is ever changed. I'll just go ahead and
> > > > add this limited version of dlinfo; look for it soon and ping me if
> > > > you don't see it.
> > >
> > > Thanks!
> >
> > One more question: does RTLD_SELF need to work? That's a lot harder
> > and uglier, as it requires extra asm for each arch...
> >
> > Rich
> 
> I don't think so.

OK. It's implemented but untested. Take a look at current git and see
if it meets your needs.

Rich

Re: dlinfo

[Justin Cormack]

On Sat, Jun 29, 2013 at 5:48 PM, Rich Felker <dalias@aerifal.cx> wrote:
> On Sat, Jun 29, 2013 at 05:43:12PM +0100, Justin Cormack wrote:
>> On 29 Jun 2013 17:39, "Rich Felker" <dalias@aerifal.cx> wrote:
>> >
>> > On Sat, Jun 29, 2013 at 05:34:46PM +0100, Justin Cormack wrote:
>> > > > As a quick workaround then:
>> > > >
>> > > > -D'dlinfo(x,y,z)=(*(struct link_map *)(z) = (void *)(x))'
>> > > >
>> > > > Kidding aside, I wouldn't actually recommend this, since you'd be
>> > > > encoding implementation internals (the fact that the dso handle is
>> > > > actually the link_map pointer) into the application, which could badly
>> > > > break if the implementation is ever changed. I'll just go ahead and
>> > > > add this limited version of dlinfo; look for it soon and ping me if
>> > > > you don't see it.
>> > >
>> > > Thanks!
>> >
>> > One more question: does RTLD_SELF need to work? That's a lot harder
>> > and uglier, as it requires extra asm for each arch...
>> >
>> > Rich
>>
>> I don't think so.
>
> OK. It's implemented but untested. Take a look at current git and see
> if it meets your needs.

Haven't been able to test yet as I have further build issues, will
report back later...

Thanks very much

Justin