mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
* [PATCH] x86/cfi: mark cfi_debug as __maybe_unused
@ 2026-11-16 17:30 Ali Nasrollahi
  2026-11-17 18:35 ` Peter Zijlstra
  0 siblings, 1 reply; 4+ messages in thread
From: Ali Nasrollahi @ 2026-11-16 17:30 UTC (permalink / raw)
  To: x86; +Cc: linux-kernel, llvm, kees, peterz, nathan, Ali Nasrollahi

    When CONFIG_FINEIBT is disabled, cfi_debug is never read, but
    alternative_instructions() still resets it after calling
    __apply_fineibt(). Clang 23 diagnoses this as -Wunused-but-set-global
    with W=1, causing a build failure when CONFIG_WERROR=y. Mark cfi_debug
    as __maybe_unused to account for configurations without FineIBT.

Signed-off-by: Ali Nasrollahi <A.Nasrolahi01@gmail.com>
---
 arch/x86/kernel/alternative.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/arch/x86/kernel/alternative.c b/arch/x86/kernel/alternative.c
index efbc11f8c336..cc2bdca06f81 100644
--- a/arch/x86/kernel/alternative.c
+++ b/arch/x86/kernel/alternative.c
@@ -1195,7 +1195,7 @@ void __init_or_module apply_seal_endbr(s32 *start, s32 *end) { }
 #endif
 
 enum cfi_mode cfi_mode __ro_after_init = __CFI_DEFAULT;
-static bool cfi_debug __ro_after_init;
+static bool cfi_debug __ro_after_init __maybe_unused;
 
 #ifdef CONFIG_FINEIBT_BHI
 bool cfi_bhi __ro_after_init = false;
-- 
2.56.1


^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [PATCH] x86/cfi: mark cfi_debug as __maybe_unused
  2026-11-16 17:30 [PATCH] x86/cfi: mark cfi_debug as __maybe_unused Ali Nasrollahi
@ 2026-11-17 18:35 ` Peter Zijlstra
  2026-11-19  6:05   ` Ali Nasrollahi
  0 siblings, 1 reply; 4+ messages in thread
From: Peter Zijlstra @ 2026-11-17 18:35 UTC (permalink / raw)
  To: Ali Nasrollahi; +Cc: x86, linux-kernel, llvm, kees, nathan

On Mon, Nov 16, 2026 at 09:00:27PM +0330, Ali Nasrollahi wrote:
>     When CONFIG_FINEIBT is disabled, cfi_debug is never read, but
>     alternative_instructions() still resets it after calling
>     __apply_fineibt(). Clang 23 diagnoses this as -Wunused-but-set-global
>     with W=1, causing a build failure when CONFIG_WERROR=y. Mark cfi_debug
>     as __maybe_unused to account for configurations without FineIBT.
> 
> Signed-off-by: Ali Nasrollahi <A.Nasrolahi01@gmail.com>

Does something like so also work?

---
diff --git a/arch/x86/kernel/alternative.c b/arch/x86/kernel/alternative.c
index efbc11f8c336..9bfc36566e17 100644
--- a/arch/x86/kernel/alternative.c
+++ b/arch/x86/kernel/alternative.c
@@ -1195,9 +1195,9 @@ void __init_or_module apply_seal_endbr(s32 *start, s32 *end) { }
 #endif
 
 enum cfi_mode cfi_mode __ro_after_init = __CFI_DEFAULT;
-static bool cfi_debug __ro_after_init;
 
 #ifdef CONFIG_FINEIBT_BHI
+static bool cfi_debug __ro_after_init;
 bool cfi_bhi __ro_after_init = false;
 #endif
 
@@ -1894,6 +1894,7 @@ static void __init_or_module __apply_fineibt(s32 *start_retpoline, s32 *end_retp
 		break;
 	}
 
+	cfi_debug = false;
 err:
 	pr_err("Something went horribly wrong trying to rewrite the CFI implementation.\n");
 }
@@ -2281,7 +2282,6 @@ void __init alternative_instructions(void)
 
 	__apply_fineibt(__retpoline_sites, __retpoline_sites_end,
 			__cfi_sites, __cfi_sites_end, true);
-	cfi_debug = false;
 
 	/*
 	 * Rewrite the retpolines, must be done before alternatives since

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [PATCH] x86/cfi: mark cfi_debug as __maybe_unused
  2026-11-17 18:35 ` Peter Zijlstra
@ 2026-11-19  6:05   ` Ali Nasrollahi
  2026-11-19  9:10     ` Peter Zijlstra
  0 siblings, 1 reply; 4+ messages in thread
From: Ali Nasrollahi @ 2026-11-19  6:05 UTC (permalink / raw)
  To: Peter Zijlstra; +Cc: x86, linux-kernel, llvm, kees, nathan

On 26/11/17 07:35PM, Peter Zijlstra wrote:
> On Mon, Nov 16, 2026 at 09:00:27PM +0330, Ali Nasrollahi wrote:
> >     When CONFIG_FINEIBT is disabled, cfi_debug is never read, but
> >     alternative_instructions() still resets it after calling
> >     __apply_fineibt(). Clang 23 diagnoses this as -Wunused-but-set-global
> >     with W=1, causing a build failure when CONFIG_WERROR=y. Mark cfi_debug
> >     as __maybe_unused to account for configurations without FineIBT.
> > 
> > Signed-off-by: Ali Nasrollahi <A.Nasrolahi01@gmail.com>
> 
> Does something like so also work?
> 
> ---
> diff --git a/arch/x86/kernel/alternative.c b/arch/x86/kernel/alternative.c
> index efbc11f8c336..9bfc36566e17 100644
> --- a/arch/x86/kernel/alternative.c
> +++ b/arch/x86/kernel/alternative.c
> @@ -1195,9 +1195,9 @@ void __init_or_module apply_seal_endbr(s32 *start, s32 *end) { }
>  #endif
>  
>  enum cfi_mode cfi_mode __ro_after_init = __CFI_DEFAULT;
> -static bool cfi_debug __ro_after_init;
>  
>  #ifdef CONFIG_FINEIBT_BHI
> +static bool cfi_debug __ro_after_init;
>  bool cfi_bhi __ro_after_init = false;
>  #endif
>  
> @@ -1894,6 +1894,7 @@ static void __init_or_module __apply_fineibt(s32 *start_retpoline, s32 *end_retp
>  		break;
>  	}
>  
> +	cfi_debug = false;
>  err:
>  	pr_err("Something went horribly wrong trying to rewrite the CFI implementation.\n");
>  }
> @@ -2281,7 +2282,6 @@ void __init alternative_instructions(void)
>  
>  	__apply_fineibt(__retpoline_sites, __retpoline_sites_end,
>  			__cfi_sites, __cfi_sites_end, true);
> -	cfi_debug = false;
>  
>  	/*
>  	 * Rewrite the retpolines, must be done before alternatives since

Thank you for your review.

Regarding this:

>  #ifdef CONFIG_FINEIBT_BHI
> +static bool cfi_debug __ro_after_init;
>  bool cfi_bhi __ro_after_init = false;
>  #endif

FINEIBT can be enabled while FINEIBT_BHI is disabled, for example when
rust is enabled. In that case, the declaration is compiled out while
the uses remain, which naturally caused my build to fail with an
undeclared cfi_debug. Maybe you meant to put it under under
CONFIG_FINEIBT instead?

Additionally, I kept my original patch limited to the warning fix
because I wasn't sure about the effects of changing the reset. With your
change the CFI_OFF, CFI_KCFI and CFI_FINEIBT cases return before
reaching the new assignment. Wouldn't this change the behviour? or
perhaps that is expected.

Cheers,
    -- Ali

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [PATCH] x86/cfi: mark cfi_debug as __maybe_unused
  2026-11-19  6:05   ` Ali Nasrollahi
@ 2026-11-19  9:10     ` Peter Zijlstra
  0 siblings, 0 replies; 4+ messages in thread
From: Peter Zijlstra @ 2026-11-19  9:10 UTC (permalink / raw)
  To: Ali Nasrollahi; +Cc: x86, linux-kernel, llvm, kees, nathan

On Thu, Nov 19, 2026 at 09:35:30AM +0330, Ali Nasrollahi wrote:
> On 26/11/17 07:35PM, Peter Zijlstra wrote:
> > On Mon, Nov 16, 2026 at 09:00:27PM +0330, Ali Nasrollahi wrote:
> > >     When CONFIG_FINEIBT is disabled, cfi_debug is never read, but
> > >     alternative_instructions() still resets it after calling
> > >     __apply_fineibt(). Clang 23 diagnoses this as -Wunused-but-set-global
> > >     with W=1, causing a build failure when CONFIG_WERROR=y. Mark cfi_debug
> > >     as __maybe_unused to account for configurations without FineIBT.
> > > 
> > > Signed-off-by: Ali Nasrollahi <A.Nasrolahi01@gmail.com>
> > 
> > Does something like so also work?
> > 
> > ---
> > diff --git a/arch/x86/kernel/alternative.c b/arch/x86/kernel/alternative.c
> > index efbc11f8c336..9bfc36566e17 100644
> > --- a/arch/x86/kernel/alternative.c
> > +++ b/arch/x86/kernel/alternative.c
> > @@ -1195,9 +1195,9 @@ void __init_or_module apply_seal_endbr(s32 *start, s32 *end) { }
> >  #endif
> >  
> >  enum cfi_mode cfi_mode __ro_after_init = __CFI_DEFAULT;
> > -static bool cfi_debug __ro_after_init;
> >  
> >  #ifdef CONFIG_FINEIBT_BHI
> > +static bool cfi_debug __ro_after_init;
> >  bool cfi_bhi __ro_after_init = false;
> >  #endif
> >  
> > @@ -1894,6 +1894,7 @@ static void __init_or_module __apply_fineibt(s32 *start_retpoline, s32 *end_retp
> >  		break;
> >  	}
> >  
> > +	cfi_debug = false;
> >  err:
> >  	pr_err("Something went horribly wrong trying to rewrite the CFI implementation.\n");
> >  }
> > @@ -2281,7 +2282,6 @@ void __init alternative_instructions(void)
> >  
> >  	__apply_fineibt(__retpoline_sites, __retpoline_sites_end,
> >  			__cfi_sites, __cfi_sites_end, true);
> > -	cfi_debug = false;
> >  
> >  	/*
> >  	 * Rewrite the retpolines, must be done before alternatives since
> 
> Thank you for your review.
> 
> Regarding this:
> 
> >  #ifdef CONFIG_FINEIBT_BHI
> > +static bool cfi_debug __ro_after_init;
> >  bool cfi_bhi __ro_after_init = false;
> >  #endif
> 
> FINEIBT can be enabled while FINEIBT_BHI is disabled, for example when

Yeah, no, that's me not being able to read. I mean to stick that under
CONFIG_FINEIBT.



^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2026-11-19  9:10 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2026-11-16 17:30 [PATCH] x86/cfi: mark cfi_debug as __maybe_unused Ali Nasrollahi
2026-11-17 18:35 ` Peter Zijlstra
2026-11-19  6:05   ` Ali Nasrollahi
2026-11-19  9:10     ` Peter Zijlstra

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

all inboxes | Powered by JetHome®