ARM: Support for TLS relocations

Richard Earnshaw rearnsha@gcc.gnu.org
Tue Mar 29 16:38:00 GMT 2005


On Fri, 2005-03-25 at 16:53, Daniel Jacobowitz wrote:
> This patch implements the ARM-documented TLS relocations, except for the
> twelve-bit instruction relocations.  The first version was implemented by
> Phil, although I'm afraid he won't recognize much of it; mistakes all mine.
> 
> The implemented relocations are, for the assembler:
> 	.word foo(tlsgd)
> 	.word foo(tlsldm)
> 	.word foo(tlsldo)
> 	.word foo(gottpoff)
> 	.word foo(tpoff)
> 
> They support trailing addends, e.g. a typical usage:
> 	.word foo(tlsgd) + (. - .L2 + 8)
> 
> No linker relaxations are supported; the compiler's choice is final.
> Dynamic relocations are avoided when unneeded, though.
> 
> Tested on arm-linux, including NPTL with appropriate gcc/glibc patches.
> Is this OK for HEAD?
> 
> [Full disclosure - I actually tested a different version of this patch,
> which used a different name for the IE32 relocation operator.  After
> changing it to gottpoff I've started a new test cycle which is running now.]

I assume this has completed testing by now... and since you've said
nothing further that there weren't any problems.

This is OK provided it doesn't cause any failures for coff formats.

R.




More information about the Binutils mailing list