Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Tue, 8 Sep 2020 16:28:35 -0700
From: Sami Tolvanen <>
To: Kees Cook <>
Cc: Masahiro Yamada <>, Will Deacon <>,
	Peter Zijlstra <>,
	Steven Rostedt <>,
	Greg Kroah-Hartman <>,
	"Paul E. McKenney" <>,
	Nick Desaulniers <>,,,,,,,,
Subject: Re: [PATCH v2 27/28] x86, relocs: Ignore L4_PAGE_OFFSET relocations

On Thu, Sep 03, 2020 at 03:47:32PM -0700, Kees Cook wrote:
> On Thu, Sep 03, 2020 at 01:30:52PM -0700, Sami Tolvanen wrote:
> > L4_PAGE_OFFSET is a constant value, so don't warn about absolute
> > relocations.
> > 
> > Signed-off-by: Sami Tolvanen <>
> Any other details on this? I assume this is an ld.lld-ism. Any idea why
> this is only a problem under LTO? (Or is this an LLVM integrated
> assembler-ism?) Regardless, yes, let's nail it down:

With the LTO v1 series, LLD generated this relocation somewhere in the section, but only with LTO:

  $ arch/x86/tools/relocs --abs-relocs vmlinux
  WARNING: Absolute relocations present
  Offset     Info     Type     Sym.Value Sym.Name
  ffffffff828e7fe0 0000000100000001 R_X86_64_64 0000000000000111

It actually looks like this might not be a problem anymore with the
current ToT kernel and the v2 series, but I'll do some more testing to
confirm this and drop the patch from v3 if it's no longer needed.


Powered by blists - more mailing lists

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.