[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <063D6719AE5E284EB5DD2968C1650D6DB026817E@AcuExch.aculab.com>
Date: Thu, 19 Jan 2017 17:01:05 +0000
From: David Laight <David.Laight@...LAB.COM>
To: 'Ard Biesheuvel' <ard.biesheuvel@...aro.org>,
Linus Torvalds <torvalds@...ux-foundation.org>
CC: Jessica Yu <jeyu@...hat.com>, Arnd Bergmann <arnd@...db.de>,
"Suzuki K. Poulose" <suzuki.poulose@....com>,
Rusty Russell <rusty@...tcorp.com.au>,
Will Deacon <will.deacon@....com>,
Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
Paul Mackerras <paulus@...ba.org>,
Al Viro <viro@...iv.linux.org.uk>,
Andrew Morton <akpm@...ux-foundation.org>,
ppc-dev <linuxppc-dev@...ts.ozlabs.org>
Subject: RE: [PATCH v4 3/3] modversions: treat symbol CRCs as 32 bit
quantities on 64 bit archs
From: Ard Biesheuvel
> Sent: 18 January 2017 13:53
..
> It turns out that this odd treatment of absolute symbols (i.e.,
> symbols having section number SHN_ABS) is a known issue in GNU ld
>
> https://sourceware.org/ml/binutils/2012-05/msg00019.html
...
Jeepers - that is truly f*cked.
I've even used the linker script to generate absolute symbols for the
sizes of items - they really don't want relocating (ever).
If you go back to (say) RSX11/M loads of the constants (like system
call numbers and structure offsets) would be supplied by the linker
if the assembler didn't know the value.
That linker supported (more or less) arbitrary expressions in fixups.
David
Powered by blists - more mailing lists