[PATCH] D22961: [ELF] - Linkerscript: restrict moving location counter backwards.

[Rui Ueyama via llvm-commits]

Looks like we should do something for them, but can you describe what we
should do? I can't find a simple rule to describe the behavior of GNU
linkers.

On Mon, Aug 8, 2016 at 10:35 AM, Davide Italiano <dccitaliano at gmail.com>
wrote:

> davide added a comment.
>
> In https://reviews.llvm.org/D22961#508592, @grimar wrote:
>
> > In https://reviews.llvm.org/D22961#508415, @emaste wrote:
> >
> > > > Problem here that BSD script is not full, there are orphan sections.
> gold and ld use heuristic to place orphan sections
> > >
> > > >  somewhere in the middle. LLD does not do that and we produce broken
> output here:
> > >
> > >
> > > We'll want to address this in FreeBSD's linker script in general so
> that we don't have the orphan sections. I agree that a warning/error would
> be desirable here to aid in identifying and correcting these cases. I'm not
> sure off hand how to handle the __start_<section> symbol generation when
> including these sections in the linker script though.
> >
> >
> > Just in case - I had to add next sections explicitly to BSD script.
> >
> >   set_sysctl_set             : { *(set_sysctl_set) }
> >   set_sysinit_set            : { *(set_sysinit_set) }
> >   set_sysuninit_set          : { *(set_sysuninit_set) }
> >   set_modmetadata_set        : { *(set_modmetadata_set) }
> >   set_ah_chips               : { *(set_ah_chips) }
> >   set_ah_rfs                 : { *(set_ah_rfs) }
> >   set_kbddriver_set          : { *(set_kbddriver_set) }
> >   set_cons_set               : { *(set_cons_set) }
> >   usb_host_id                : { *(usb_host_id) }
> >   set_vt_drv_set             : { *(set_vt_drv_set) }
> >   set_sdt_providers_set      : { *(set_sdt_providers_set) }
> >   set_sdt_probes_set         : { *(set_sdt_probes_set) }
> >   set_sdt_argtypes_set       : { *(set_sdt_argtypes_set) }
> >   set_kdb_dbbe_set           : { *(set_kdb_dbbe_set) }
> >   set_ratectl_set            : { *(set_ratectl_set) }
> >   set_crypto_set             : { *(set_crypto_set) }
> >   set_ieee80211_ioctl_getset : { *(set_ieee80211_ioctl_getset) }
> >   set_ieee80211_ioctl_setset : { *(set_ieee80211_ioctl_setset) }
> >   set_scanner_set            : { *(set_scanner_set) }
> >   set_videodriver_set        : { *(set_videodriver_set) }
> >   set_scterm_set             : { *(set_scterm_set) }
> >   set_scrndr_set             : { *(set_scrndr_set) }
> >   set_vga_set                : { *(set_vga_set) }
> >   kern_conf                  : { *(kern_conf) }
> >   set_pcpu                   : { *(set_pcpu) }
> >
> >
> > + Also there one more unprocessed: .SUNW_ctf but it is not allocatable
> so I am fine it is added as orphan for me now. Generally for non
> allocatable sections it is probably ok to be orphans I think.
>
>
> I hit the same problem internally, on a codebase I don't own but I have to
> support. Sorry, I can't share it, but it's a kernel and it's very similar
> to the linker script provided by FreeBSD.
> Setting the wrong VA causes the loader to be very confused. I think we
> should support this particular feature of ld.bfd in a bug-by-bug compatible
> fashion.
> If FreeBSD can change his linker script, fine, great for them. But not
> everybody has the same luxury.
>
>
> https://reviews.llvm.org/D22961
>
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20160808/b1b315d9/attachment.html>