[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20241004082117.GH18071@noisy.programming.kicks-ass.net>
Date: Fri, 4 Oct 2024 10:21:17 +0200
From: Peter Zijlstra <peterz@...radead.org>
To: Josh Poimboeuf <jpoimboe@...nel.org>
Cc: x86@...nel.org, linux-kernel@...r.kernel.org,
Ard Biesheuvel <ardb@...nel.org>
Subject: Re: [PATCH] objtool: Detect non-relocated text references
On Thu, Oct 03, 2024 at 05:31:10PM -0700, Josh Poimboeuf wrote:
> When kernel IBT is enabled, objtool detects all text references in order
> to determine which functions can be indirectly branched to.
>
> In text, such references look like one of the following:
>
> mov $0x0,%rax R_X86_64_32S .init.text+0x7e0a0
> lea 0x0(%rip),%rax R_X86_64_PC32 autoremove_wake_function-0x4
>
> Either way the function pointer is denoted by a relocation, so objtool
> just reads that.
>
> However there are some "lea xxx(%rip)" cases which don't use relocations
> because they're referencing code in the same translation unit. Objtool
> doesn't have visibility to those.
>
> The only currently known instances of that are a few hand-coded asm text
> references which don't actually need ENDBR. So it's not actually a
> problem at the moment.
>
> However if we enable -fpie, the compiler would start generating them and
> there would definitely be bugs in the IBT sealing.
>
> Detect non-relocated text references and handle them appropriately.
>
> [ Note: I removed the manual static_call_tramp check -- that should
> already be handled by the noendbr check. ]
>
> Reported-by: Ard Biesheuvel <ardb@...nel.org>
> Signed-off-by: Josh Poimboeuf <jpoimboe@...nel.org>
> ---
Acked-by: Peter Zijlstra (Intel) <peterz@...radead.org>
Powered by blists - more mailing lists