linux-kernel - Re: [PATCH v2] efistub: Only link libstub to final vmlinux

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <CAMj1kXH-rK0bRyHXdJ-crAyMyvJHApH0WR7_8Qd8vrSPBLK+yg@mail.gmail.com>
Date: Fri, 10 Oct 2025 20:40:26 -0700
From: Ard Biesheuvel <ardb@...nel.org>
To: Huacai Chen <chenhuacai@...nel.org>
Cc: Tiezhu Yang <yangtiezhu@...ngson.cn>, Josh Poimboeuf <jpoimboe@...nel.org>, 
	loongarch@...ts.linux.dev, linux-arm-kernel@...ts.infradead.org, 
	linux-riscv@...ts.infradead.org, linux-efi@...r.kernel.org, 
	linux-kbuild@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH v2] efistub: Only link libstub to final vmlinux

On Fri, 10 Oct 2025 at 19:54, Huacai Chen <chenhuacai@...nel.org> wrote:
>
> On Sat, Oct 11, 2025 at 9:13 AM Tiezhu Yang <yangtiezhu@...ngson.cn> wrote:
> >
> > On 2025/10/11 上午12:25, Ard Biesheuvel wrote:
> > ...
> > > Why do we need both (1) and (2)?
> >
> > Not both, either (1) or (2).
> > Which one do you prefer? Or any other suggestions?
> >
> > Taking all of the considerations in balance, we should decide
> > what is the proper way.
> As a summary, there are three methods:
> (1) Only link libstub with vmlinux.o during the final vmlinux link.
> (2) Remove the attribute __noreturn for real_kernel_entry() and add while (1).
> (3) Ignore "__efistub_" prefix in objtool.
>
> Josh prefers method (1), I prefer method (3) but also accept method
> (1) if it is not only specific to loongarch.
>

This is a false positive warning in objtool, which complains about a
function that falls through, even though that can never happen in
reality.

To me, it is not acceptable to modify how vmlinux.o is constructed
also for other architectures, in order to hide some of its constituent
parts from objtool, which do not use objtool to begin with.


If you are not willing to fix objtool, I suggest fixing the loongarch
code like this:

--- a/drivers/firmware/efi/libstub/loongarch.c
+++ b/drivers/firmware/efi/libstub/loongarch.c
@@ -10,7 +10,7 @@
 #include "efistub.h"
 #include "loongarch-stub.h"

-typedef void __noreturn (*kernel_entry_t)(bool efi, unsigned long cmdline,
+typedef void (*kernel_entry_t)(bool efi, unsigned long cmdline,
                                          unsigned long systab);

 efi_status_t check_platform_features(void)
@@ -81,4 +81,6 @@

        real_kernel_entry(true, (unsigned long)cmdline_ptr,
                          (unsigned long)efi_system_table);
+
+       return EFI_LOAD_ERROR;
 }