This is the mail archive of the
mailing list for the binutils project.
Re: [PATCH] ld/ld.texinfo: Do not constrain dynamic linker lazy binding behaviour.
- From: Florian Weimer <fweimer at redhat dot com>
- To: "Carlos O'Donell" <carlos at redhat dot com>, binutils at sourceware dot org, Nick Clifton <nickc at redhat dot com>
- Cc: Suprateeka R Hegde <hegdesmailbox at gmail dot com>
- Date: Thu, 7 Jul 2016 18:11:21 +0200
- Subject: Re: [PATCH] ld/ld.texinfo: Do not constrain dynamic linker lazy binding behaviour.
- Authentication-results: sourceware.org; auth=none
- References: <firstname.lastname@example.org>
On 07/07/2016 06:09 PM, Carlos O'Donell wrote:
The resolution of function symbols under lazy binding should
not be constrained by the binutils documentation.
The lazy binding can theoretically happen at any time between
load of the shared object and the function call. Under some
conditions it may be desirable to have it happen earlier than
the function call time.
For the sake of the argument consider a future optimization where
we might be able to do a bulk resolution of all symbols in a given
shared object, and so lazy resolution of _all_ symbols for a given
shared object happens at the first call to a function from the
shared object rather than at each call.
No regressions on x86_64, and pdf of the manual looks good.
OK to checkin?
Maybe also mention that binding per relocation can happen multiple times?
I think this is true even today because the PLT update is racy, and it's
observable through IFUNC resolvers.