Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 31 Jan 2008 06:11:52 -0500
From:      John Baldwin <jhb@freebsd.org>
To:        Bruce M Simpson <bms@incunabulum.net>
Cc:        freebsd-stable@freebsd.org
Subject:   Re: kldload: Unsupported file type
Message-ID:  <200801310611.52552.jhb@freebsd.org>
In-Reply-To: <47A15C62.3020201@incunabulum.net>
References:  <479FE898.1030801@incunabulum.net> <200801300904.35750.jhb@freebsd.org> <47A15C62.3020201@incunabulum.net>

next in thread | previous in thread | raw e-mail | index | archive | help
On Thursday 31 January 2008 12:28:02 am Bruce M Simpson wrote:
> John Baldwin wrote:
> >>> It is printed whenever a kernel module is loaded.
> >>> The modules load OK. Nothing special or different about them.
> >>>
> >>>       
> ...
> > The kernel is a link_elf type object I believe, so you have to have it.
> >
> >   
> 
> That follows (I was reading this the other day 'cause we don't support 
> weak ELF symbols in the kernel for C++) however, why is the message 
> being triggered now?
> 
> Could it be ET_REL ?
> There have been no major changes to linking for the 6.3 buildkernel 
> target IIRC.
> 
> 
> BTW only my amd64 system appears to be affected.

The problem is that .ko's on amd64 are handled by link_elf_obj.c and not 
link_elf.c, thus if link_elf.c is first in the list of linker file handlers, 
then every .ko on amd64 is first going to try link_elf.c which fails and 
emits the error and then get loaded successfully by link_elf_obj.c.  Probably 
what should happen is that the linker error message should be cached somehow 
and only print out the last error if the overall load fails.

-- 
John Baldwin



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200801310611.52552.jhb>