[PATCH -tip v9 0/5] kprobes related bugfix

[PATCH -tip v9 0/5] kprobes related bugfix

[Masami Hiramatsu]

Hi,

Here are the 9th version of the series to fix kprobes not
to optimize on the functions in irqentry text ([5/5]).
To check the irqentry text always, [4/5] makes irqentry
text section unconditional.

This involves some other cleanups on h8300, xtensa, and
cris ports, because those sources declare section 
start/end symbol as single characters, but generic
sections.h declares it as character arrays. So it leads
build errors if we apply [4/5].

Changes in v9
 - Add cris cleanup patch (3/5).

BTW, in this series I just give a minimum fix for
arch dependent code, since this is a series to fix
a bug in x86 optprobe. However, IMHO, we should
revisit these "local _stext/_etext references" and
replace it with just including asm/sections.h.
And also, I would like to expose is_kernel_text()
as a generic routine for arch dependent code
so that they can use it.

Thank you,

---

Masami Hiramatsu (5):
      h8300: mark _stext and _etext as char-arrays, not single char
      xtensa: mark _stext and _end as char-arrays, not single char
      cris: mark _stext and _end as char-arrays, not single char
      irq: Make irqentry text section unconditional
      [BUGFIX] kprobes/x86: Do not jump-optimize kprobes on irq entry code


 arch/arm/include/asm/traps.h         |    7 -------
 arch/arm64/include/asm/traps.h       |    7 -------
 arch/cris/arch-v32/mach-a3/arbiter.c |    4 ++--
 arch/cris/arch-v32/mach-fs/arbiter.c |    4 ++--
 arch/cris/kernel/traps.c             |    6 +++---
 arch/h8300/include/asm/traps.h       |    6 +++---
 arch/x86/entry/entry_64.S            |    9 ++-------
 arch/x86/kernel/kprobes/opt.c        |    9 ++++++---
 arch/x86/kernel/unwind_frame.c       |    2 --
 arch/xtensa/kernel/setup.c           |    6 +++---
 include/asm-generic/sections.h       |    4 ++++
 include/asm-generic/vmlinux.lds.h    |    8 --------
 include/linux/interrupt.h            |   14 +-------------
 13 files changed, 26 insertions(+), 60 deletions(-)

[PATCH -tip v9 0/5] kprobes related bugfix

[Masami Hiramatsu]

Hi,

Here are the 9th version of the series to fix kprobes not
to optimize on the functions in irqentry text ([5/5]).
To check the irqentry text always, [4/5] makes irqentry
text section unconditional.

This involves some other cleanups on h8300, xtensa, and
cris ports, because those sources declare section 
start/end symbol as single characters, but generic
sections.h declares it as character arrays. So it leads
build errors if we apply [4/5].

Changes in v9
 - Add cris cleanup patch (3/5).

BTW, in this series I just give a minimum fix for
arch dependent code, since this is a series to fix
a bug in x86 optprobe. However, IMHO, we should
revisit these "local _stext/_etext references" and
replace it with just including asm/sections.h.
And also, I would like to expose is_kernel_text()
as a generic routine for arch dependent code
so that they can use it.

Thank you,

---

Masami Hiramatsu (5):
      h8300: mark _stext and _etext as char-arrays, not single char
      xtensa: mark _stext and _end as char-arrays, not single char
      cris: mark _stext and _end as char-arrays, not single char
      irq: Make irqentry text section unconditional
      [BUGFIX] kprobes/x86: Do not jump-optimize kprobes on irq entry code


 arch/arm/include/asm/traps.h         |    7 -------
 arch/arm64/include/asm/traps.h       |    7 -------
 arch/cris/arch-v32/mach-a3/arbiter.c |    4 ++--
 arch/cris/arch-v32/mach-fs/arbiter.c |    4 ++--
 arch/cris/kernel/traps.c             |    6 +++---
 arch/h8300/include/asm/traps.h       |    6 +++---
 arch/x86/entry/entry_64.S            |    9 ++-------
 arch/x86/kernel/kprobes/opt.c        |    9 ++++++---
 arch/x86/kernel/unwind_frame.c       |    2 --
 arch/xtensa/kernel/setup.c           |    6 +++---
 include/asm-generic/sections.h       |    4 ++++
 include/asm-generic/vmlinux.lds.h    |    8 --------
 include/linux/interrupt.h            |   14 +-------------
 13 files changed, 26 insertions(+), 60 deletions(-)

--
Masami Hiramatsu

[PATCH -tip v9 1/5] h8300: mark _stext and _etext as char-arrays, not single char

[Masami Hiramatsu]

Mark _stext and _etext as character arrays instead of
single character, as same as include/asm-generic/sections.h
does.

Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
---
 arch/h8300/include/asm/traps.h |    6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/arch/h8300/include/asm/traps.h b/arch/h8300/include/asm/traps.h
index 15e701130b27..1c5a30ec2df8 100644
--- a/arch/h8300/include/asm/traps.h
+++ b/arch/h8300/include/asm/traps.h
@@ -33,9 +33,9 @@ extern unsigned long *_interrupt_redirect_table;
 #define TRAP2_VEC 10
 #define TRAP3_VEC 11
 
-extern char _start, _etext;
+extern char _start[], _etext[];
 #define check_kernel_text(addr) \
-	((addr >= (unsigned long)(&_start)) && \
-	 (addr <  (unsigned long)(&_etext)) && !(addr & 1))
+	((addr >= (unsigned long)(_start)) && \
+	 (addr <  (unsigned long)(_etext)) && !(addr & 1))
 
 #endif /* _H8300_TRAPS_H */

[PATCH -tip v9 2/5] xtensa: mark _stext and _end as char-arrays, not single char

[Masami Hiramatsu]

Mark _stext and _end as character arrays instead of single
character, as same as include/asm-generic/sections.h does.

Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
---
 arch/xtensa/kernel/setup.c |    6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/arch/xtensa/kernel/setup.c b/arch/xtensa/kernel/setup.c
index 33bfa5270d95..08175df7a69e 100644
--- a/arch/xtensa/kernel/setup.c
+++ b/arch/xtensa/kernel/setup.c
@@ -273,8 +273,8 @@ void __init init_arch(bp_tag_t *bp_start)
  * Initialize system. Setup memory and reserve regions.
  */
 
-extern char _end;
-extern char _stext;
+extern char _end[];
+extern char _stext[];
 extern char _WindowVectors_text_start;
 extern char _WindowVectors_text_end;
 extern char _DebugInterruptVector_literal_start;
@@ -333,7 +333,7 @@ void __init setup_arch(char **cmdline_p)
 	}
 #endif
 
-	mem_reserve(__pa(&_stext), __pa(&_end));
+	mem_reserve(__pa(_stext), __pa(_end));
 
 #ifdef CONFIG_VECTORS_OFFSET
 	mem_reserve(__pa(&_WindowVectors_text_start),

[PATCH -tip v9 3/5] cris: mark _stext and _end as char-arrays, not single char

[Masami Hiramatsu]

Mark _stext and _end as character arrays instead of single
character, as same as include/asm-generic/sections.h does.

Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
---
 arch/cris/arch-v32/mach-a3/arbiter.c |    4 ++--
 arch/cris/arch-v32/mach-fs/arbiter.c |    4 ++--
 arch/cris/kernel/traps.c             |    6 +++---
 3 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/arch/cris/arch-v32/mach-a3/arbiter.c b/arch/cris/arch-v32/mach-a3/arbiter.c
index ab5c421a4de8..735a9b0abdb8 100644
--- a/arch/cris/arch-v32/mach-a3/arbiter.c
+++ b/arch/cris/arch-v32/mach-a3/arbiter.c
@@ -227,7 +227,7 @@ static void crisv32_arbiter_config(int arbiter, int region, int unused_slots)
 	}
 }
 
-extern char _stext, _etext;
+extern char _stext[], _etext[];
 
 static void crisv32_arbiter_init(void)
 {
@@ -265,7 +265,7 @@ static void crisv32_arbiter_init(void)
 
 #ifndef CONFIG_ETRAX_KGDB
 	/* Global watch for writes to kernel text segment. */
-	crisv32_arbiter_watch(virt_to_phys(&_stext), &_etext - &_stext,
+	crisv32_arbiter_watch(virt_to_phys(_stext), _etext - _stext,
 		MARB_CLIENTS(arbiter_all_clients, arbiter_bar_all_clients),
 			      arbiter_all_write, NULL);
 #endif
diff --git a/arch/cris/arch-v32/mach-fs/arbiter.c b/arch/cris/arch-v32/mach-fs/arbiter.c
index c97f4d8120f9..047c70bdbb23 100644
--- a/arch/cris/arch-v32/mach-fs/arbiter.c
+++ b/arch/cris/arch-v32/mach-fs/arbiter.c
@@ -158,7 +158,7 @@ static void crisv32_arbiter_config(int region, int unused_slots)
 	}
 }
 
-extern char _stext, _etext;
+extern char _stext[], _etext[];
 
 static void crisv32_arbiter_init(void)
 {
@@ -190,7 +190,7 @@ static void crisv32_arbiter_init(void)
 
 #ifndef CONFIG_ETRAX_KGDB
 	/* Global watch for writes to kernel text segment. */
-	crisv32_arbiter_watch(virt_to_phys(&_stext), &_etext - &_stext,
+	crisv32_arbiter_watch(virt_to_phys(_stext), _etext - _stext,
 			      arbiter_all_clients, arbiter_all_write, NULL);
 #endif
 }
diff --git a/arch/cris/kernel/traps.c b/arch/cris/kernel/traps.c
index a01636a12a6e..d98131c45bb5 100644
--- a/arch/cris/kernel/traps.c
+++ b/arch/cris/kernel/traps.c
@@ -42,7 +42,7 @@ void (*nmi_handler)(struct pt_regs *);
 void show_trace(unsigned long *stack)
 {
 	unsigned long addr, module_start, module_end;
-	extern char _stext, _etext;
+	extern char _stext[], _etext[];
 	int i;
 
 	pr_err("\nCall Trace: ");
@@ -69,8 +69,8 @@ void show_trace(unsigned long *stack)
 		 * down the cause of the crash will be able to figure
 		 * out the call path that was taken.
 		 */
-		if (((addr >= (unsigned long)&_stext) &&
-		     (addr <= (unsigned long)&_etext)) ||
+		if (((addr >= (unsigned long)_stext) &&
+		     (addr <= (unsigned long)_etext)) ||
 		    ((addr >= module_start) && (addr <= module_end))) {
 #ifdef CONFIG_KALLSYMS
 			print_ip_sym(addr);

Re: [PATCH -tip v9 3/5] cris: mark _stext and _end as char-arrays, not single char

[Masami Hiramatsu]

On Thu,  3 Aug 2017 11:37:15 +0900
Masami Hiramatsu <mhiramat@kernel.org> wrote:

> diff --git a/arch/cris/kernel/traps.c b/arch/cris/kernel/traps.c
> index a01636a12a6e..d98131c45bb5 100644
> --- a/arch/cris/kernel/traps.c
> +++ b/arch/cris/kernel/traps.c
> @@ -42,7 +42,7 @@ void (*nmi_handler)(struct pt_regs *);
>  void show_trace(unsigned long *stack)
>  {
>  	unsigned long addr, module_start, module_end;
> -	extern char _stext, _etext;
> +	extern char _stext[], _etext[];
>  	int i;
>  
>  	pr_err("\nCall Trace: ");
> @@ -69,8 +69,8 @@ void show_trace(unsigned long *stack)
>  		 * down the cause of the crash will be able to figure
>  		 * out the call path that was taken.
>  		 */
> -		if (((addr >= (unsigned long)&_stext) &&
> -		     (addr <= (unsigned long)&_etext)) ||
> +		if (((addr >= (unsigned long)_stext) &&
> +		     (addr <= (unsigned long)_etext)) ||
>  		    ((addr >= module_start) && (addr <= module_end))) {

BTW, I would like to ask cris people, this seems to check over 1 byte,
since _etext and module_end will be placed at the next byte of the real
address area.
As same as other arch, Should it be 

(addr < (unsigned long)_etext)

or are there any other reason?

Thank you,


>  #ifdef CONFIG_KALLSYMS
>  			print_ip_sym(addr);
> 


-- 
Masami Hiramatsu <mhiramat@kernel.org>

Re: [PATCH -tip v9 3/5] cris: mark _stext and _end as char-arrays, not single char

[Jesper Nilsson]

On Wed, Aug 16, 2017 at 08:01:33PM +0900, Masami Hiramatsu wrote:
> On Thu,  3 Aug 2017 11:37:15 +0900
> Masami Hiramatsu <mhiramat@kernel.org> wrote:
> 
> > diff --git a/arch/cris/kernel/traps.c b/arch/cris/kernel/traps.c
> > index a01636a12a6e..d98131c45bb5 100644
> > --- a/arch/cris/kernel/traps.c
> > +++ b/arch/cris/kernel/traps.c
> > @@ -42,7 +42,7 @@ void (*nmi_handler)(struct pt_regs *);
> >  void show_trace(unsigned long *stack)
> >  {
> >  	unsigned long addr, module_start, module_end;
> > -	extern char _stext, _etext;
> > +	extern char _stext[], _etext[];
> >  	int i;
> >  
> >  	pr_err("\nCall Trace: ");
> > @@ -69,8 +69,8 @@ void show_trace(unsigned long *stack)
> >  		 * down the cause of the crash will be able to figure
> >  		 * out the call path that was taken.
> >  		 */
> > -		if (((addr >= (unsigned long)&_stext) &&
> > -		     (addr <= (unsigned long)&_etext)) ||
> > +		if (((addr >= (unsigned long)_stext) &&
> > +		     (addr <= (unsigned long)_etext)) ||
> >  		    ((addr >= module_start) && (addr <= module_end))) {
> 
> BTW, I would like to ask cris people, this seems to check over 1 byte,
> since _etext and module_end will be placed at the next byte of the real
> address area.
> As same as other arch, Should it be 
> 
> (addr < (unsigned long)_etext)
> 
> or are there any other reason?

As far as I can tell, this is non-intentional.

> Thank you,

> Masami Hiramatsu <mhiramat@kernel.org>

/^JN - Jesper Nilsson
-- 
               Jesper Nilsson -- jesper.nilsson@axis.com

Re: [PATCH -tip v9 3/5] cris: mark _stext and _end as char-arrays, not single char

[Jesper Nilsson]

On Thu, Aug 03, 2017 at 11:37:15AM +0900, Masami Hiramatsu wrote:
> Mark _stext and _end as character arrays instead of single
> character, as same as include/asm-generic/sections.h does.

Looks good.

Acked-by: Jesper Nilsson <jesper.nilsson@axis.com>

> Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
> ---
>  arch/cris/arch-v32/mach-a3/arbiter.c |    4 ++--
>  arch/cris/arch-v32/mach-fs/arbiter.c |    4 ++--
>  arch/cris/kernel/traps.c             |    6 +++---
>  3 files changed, 7 insertions(+), 7 deletions(-)
> 
> diff --git a/arch/cris/arch-v32/mach-a3/arbiter.c b/arch/cris/arch-v32/mach-a3/arbiter.c
> index ab5c421a4de8..735a9b0abdb8 100644
> --- a/arch/cris/arch-v32/mach-a3/arbiter.c
> +++ b/arch/cris/arch-v32/mach-a3/arbiter.c
> @@ -227,7 +227,7 @@ static void crisv32_arbiter_config(int arbiter, int region, int unused_slots)
>  	}
>  }
>  
> -extern char _stext, _etext;
> +extern char _stext[], _etext[];
>  
>  static void crisv32_arbiter_init(void)
>  {
> @@ -265,7 +265,7 @@ static void crisv32_arbiter_init(void)
>  
>  #ifndef CONFIG_ETRAX_KGDB
>  	/* Global watch for writes to kernel text segment. */
> -	crisv32_arbiter_watch(virt_to_phys(&_stext), &_etext - &_stext,
> +	crisv32_arbiter_watch(virt_to_phys(_stext), _etext - _stext,
>  		MARB_CLIENTS(arbiter_all_clients, arbiter_bar_all_clients),
>  			      arbiter_all_write, NULL);
>  #endif
> diff --git a/arch/cris/arch-v32/mach-fs/arbiter.c b/arch/cris/arch-v32/mach-fs/arbiter.c
> index c97f4d8120f9..047c70bdbb23 100644
> --- a/arch/cris/arch-v32/mach-fs/arbiter.c
> +++ b/arch/cris/arch-v32/mach-fs/arbiter.c
> @@ -158,7 +158,7 @@ static void crisv32_arbiter_config(int region, int unused_slots)
>  	}
>  }
>  
> -extern char _stext, _etext;
> +extern char _stext[], _etext[];
>  
>  static void crisv32_arbiter_init(void)
>  {
> @@ -190,7 +190,7 @@ static void crisv32_arbiter_init(void)
>  
>  #ifndef CONFIG_ETRAX_KGDB
>  	/* Global watch for writes to kernel text segment. */
> -	crisv32_arbiter_watch(virt_to_phys(&_stext), &_etext - &_stext,
> +	crisv32_arbiter_watch(virt_to_phys(_stext), _etext - _stext,
>  			      arbiter_all_clients, arbiter_all_write, NULL);
>  #endif
>  }
> diff --git a/arch/cris/kernel/traps.c b/arch/cris/kernel/traps.c
> index a01636a12a6e..d98131c45bb5 100644
> --- a/arch/cris/kernel/traps.c
> +++ b/arch/cris/kernel/traps.c
> @@ -42,7 +42,7 @@ void (*nmi_handler)(struct pt_regs *);
>  void show_trace(unsigned long *stack)
>  {
>  	unsigned long addr, module_start, module_end;
> -	extern char _stext, _etext;
> +	extern char _stext[], _etext[];
>  	int i;
>  
>  	pr_err("\nCall Trace: ");
> @@ -69,8 +69,8 @@ void show_trace(unsigned long *stack)
>  		 * down the cause of the crash will be able to figure
>  		 * out the call path that was taken.
>  		 */
> -		if (((addr >= (unsigned long)&_stext) &&
> -		     (addr <= (unsigned long)&_etext)) ||
> +		if (((addr >= (unsigned long)_stext) &&
> +		     (addr <= (unsigned long)_etext)) ||
>  		    ((addr >= module_start) && (addr <= module_end))) {
>  #ifdef CONFIG_KALLSYMS
>  			print_ip_sym(addr);

/^JN - Jesper Nilsson
-- 
               Jesper Nilsson -- jesper.nilsson@axis.com

[PATCH -tip v9 4/5] irq: Make irqentry text section unconditional

[Masami Hiramatsu]

Generate irqentry and softirqentry text sections without
any configs. This will add above extra sections, but
no performace impact.

Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
Suggested-by: Ingo Molnar <mingo@kernel.org>
---
 arch/arm/include/asm/traps.h      |    7 -------
 arch/arm64/include/asm/traps.h    |    7 -------
 arch/x86/entry/entry_64.S         |    9 ++-------
 arch/x86/kernel/unwind_frame.c    |    2 --
 include/asm-generic/sections.h    |    4 ++++
 include/asm-generic/vmlinux.lds.h |    8 --------
 include/linux/interrupt.h         |   14 +-------------
 7 files changed, 7 insertions(+), 44 deletions(-)

diff --git a/arch/arm/include/asm/traps.h b/arch/arm/include/asm/traps.h
index f555bb3664dc..683d9230984a 100644
--- a/arch/arm/include/asm/traps.h
+++ b/arch/arm/include/asm/traps.h
@@ -18,7 +18,6 @@ struct undef_hook {
 void register_undef_hook(struct undef_hook *hook);
 void unregister_undef_hook(struct undef_hook *hook);
 
-#ifdef CONFIG_FUNCTION_GRAPH_TRACER
 static inline int __in_irqentry_text(unsigned long ptr)
 {
 	extern char __irqentry_text_start[];
@@ -27,12 +26,6 @@ static inline int __in_irqentry_text(unsigned long ptr)
 	return ptr >= (unsigned long)&__irqentry_text_start &&
 	       ptr < (unsigned long)&__irqentry_text_end;
 }
-#else
-static inline int __in_irqentry_text(unsigned long ptr)
-{
-	return 0;
-}
-#endif
 
 static inline int in_exception_text(unsigned long ptr)
 {
diff --git a/arch/arm64/include/asm/traps.h b/arch/arm64/include/asm/traps.h
index 02e9035b0685..47a9066f7c86 100644
--- a/arch/arm64/include/asm/traps.h
+++ b/arch/arm64/include/asm/traps.h
@@ -37,18 +37,11 @@ void unregister_undef_hook(struct undef_hook *hook);
 
 void arm64_notify_segfault(struct pt_regs *regs, unsigned long addr);
 
-#ifdef CONFIG_FUNCTION_GRAPH_TRACER
 static inline int __in_irqentry_text(unsigned long ptr)
 {
 	return ptr >= (unsigned long)&__irqentry_text_start &&
 	       ptr < (unsigned long)&__irqentry_text_end;
 }
-#else
-static inline int __in_irqentry_text(unsigned long ptr)
-{
-	return 0;
-}
-#endif
 
 static inline int in_exception_text(unsigned long ptr)
 {
diff --git a/arch/x86/entry/entry_64.S b/arch/x86/entry/entry_64.S
index aa58155187c5..5e41396efbbf 100644
--- a/arch/x86/entry/entry_64.S
+++ b/arch/x86/entry/entry_64.S
@@ -766,13 +766,8 @@ apicinterrupt3 \num trace(\sym) smp_trace(\sym)
 #endif
 
 /* Make sure APIC interrupt handlers end up in the irqentry section: */
-#if defined(CONFIG_FUNCTION_GRAPH_TRACER) || defined(CONFIG_KASAN)
-# define PUSH_SECTION_IRQENTRY	.pushsection .irqentry.text, "ax"
-# define POP_SECTION_IRQENTRY	.popsection
-#else
-# define PUSH_SECTION_IRQENTRY
-# define POP_SECTION_IRQENTRY
-#endif
+#define PUSH_SECTION_IRQENTRY	.pushsection .irqentry.text, "ax"
+#define POP_SECTION_IRQENTRY	.popsection
 
 .macro apicinterrupt num sym do_sym
 PUSH_SECTION_IRQENTRY
diff --git a/arch/x86/kernel/unwind_frame.c b/arch/x86/kernel/unwind_frame.c
index 7574ef5f16ec..d145a0b1f529 100644
--- a/arch/x86/kernel/unwind_frame.c
+++ b/arch/x86/kernel/unwind_frame.c
@@ -84,10 +84,8 @@ static bool in_entry_code(unsigned long ip)
 	if (addr >= __entry_text_start && addr < __entry_text_end)
 		return true;
 
-#if defined(CONFIG_FUNCTION_GRAPH_TRACER) || defined(CONFIG_KASAN)
 	if (addr >= __irqentry_text_start && addr < __irqentry_text_end)
 		return true;
-#endif
 
 	return false;
 }
diff --git a/include/asm-generic/sections.h b/include/asm-generic/sections.h
index 532372c6cf15..e5da44eddd2f 100644
--- a/include/asm-generic/sections.h
+++ b/include/asm-generic/sections.h
@@ -27,6 +27,8 @@
  *	__kprobes_text_start, __kprobes_text_end
  *	__entry_text_start, __entry_text_end
  *	__ctors_start, __ctors_end
+ *	__irqentry_text_start, __irqentry_text_end
+ *	__softirqentry_text_start, __softirqentry_text_end
  */
 extern char _text[], _stext[], _etext[];
 extern char _data[], _sdata[], _edata[];
@@ -39,6 +41,8 @@ extern char __per_cpu_load[], __per_cpu_start[], __per_cpu_end[];
 extern char __kprobes_text_start[], __kprobes_text_end[];
 extern char __entry_text_start[], __entry_text_end[];
 extern char __start_rodata[], __end_rodata[];
+extern char __irqentry_text_start[], __irqentry_text_end[];
+extern char __softirqentry_text_start[], __softirqentry_text_end[];
 
 /* Start and end of .ctors section - used for constructor calls. */
 extern char __ctors_start[], __ctors_end[];
diff --git a/include/asm-generic/vmlinux.lds.h b/include/asm-generic/vmlinux.lds.h
index fffc9bdae025..438b014c192f 100644
--- a/include/asm-generic/vmlinux.lds.h
+++ b/include/asm-generic/vmlinux.lds.h
@@ -483,25 +483,17 @@
 		*(.entry.text)						\
 		VMLINUX_SYMBOL(__entry_text_end) = .;
 
-#if defined(CONFIG_FUNCTION_GRAPH_TRACER) || defined(CONFIG_KASAN)
 #define IRQENTRY_TEXT							\
 		ALIGN_FUNCTION();					\
 		VMLINUX_SYMBOL(__irqentry_text_start) = .;		\
 		*(.irqentry.text)					\
 		VMLINUX_SYMBOL(__irqentry_text_end) = .;
-#else
-#define IRQENTRY_TEXT
-#endif
 
-#if defined(CONFIG_FUNCTION_GRAPH_TRACER) || defined(CONFIG_KASAN)
 #define SOFTIRQENTRY_TEXT						\
 		ALIGN_FUNCTION();					\
 		VMLINUX_SYMBOL(__softirqentry_text_start) = .;		\
 		*(.softirqentry.text)					\
 		VMLINUX_SYMBOL(__softirqentry_text_end) = .;
-#else
-#define SOFTIRQENTRY_TEXT
-#endif
 
 /* Section used for early init (in .S files) */
 #define HEAD_TEXT  *(.head.text)
diff --git a/include/linux/interrupt.h b/include/linux/interrupt.h
index a2fddddb0d60..59ba11661b6e 100644
--- a/include/linux/interrupt.h
+++ b/include/linux/interrupt.h
@@ -18,6 +18,7 @@
 #include <linux/atomic.h>
 #include <asm/ptrace.h>
 #include <asm/irq.h>
+#include <asm/sections.h>
 
 /*
  * These correspond to the IORESOURCE_IRQ_* defines in
@@ -726,7 +727,6 @@ extern int early_irq_init(void);
 extern int arch_probe_nr_irqs(void);
 extern int arch_early_irq_init(void);
 
-#if defined(CONFIG_FUNCTION_GRAPH_TRACER) || defined(CONFIG_KASAN)
 /*
  * We want to know which function is an entrypoint of a hardirq or a softirq.
  */
@@ -734,16 +734,4 @@ extern int arch_early_irq_init(void);
 #define __softirq_entry  \
 	__attribute__((__section__(".softirqentry.text")))
 
-/* Limits of hardirq entrypoints */
-extern char __irqentry_text_start[];
-extern char __irqentry_text_end[];
-/* Limits of softirq entrypoints */
-extern char __softirqentry_text_start[];
-extern char __softirqentry_text_end[];
-
-#else
-#define __irq_entry
-#define __softirq_entry
-#endif
-
 #endif

[PATCH -tip v9 5/5] [BUGFIX] kprobes/x86: Do not jump-optimize kprobes on irq entry code

[Masami Hiramatsu]

Since the kernel segment registers are not prepared at the
entry of irq-entry code, if a kprobe on such code is
jump-optimized, accessing per-cpu variables may cause
kernel panic.
However, if the kprobe is not optimized, it kicks int3
exception and set segment registers correctly.

This checks probe-address and if it is in irq-entry code,
it prohibits optimizing such kprobes. This means we can
continuously probing such interrupt handlers by kprobes
but it is not optimized anymore.

Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
Reported-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Tested-by: Francis Deslauriers <francis.deslauriers@efficios.com>
---
 arch/x86/kernel/kprobes/opt.c |    9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/arch/x86/kernel/kprobes/opt.c b/arch/x86/kernel/kprobes/opt.c
index 69ea0bc1cfa3..4f98aad38237 100644
--- a/arch/x86/kernel/kprobes/opt.c
+++ b/arch/x86/kernel/kprobes/opt.c
@@ -39,6 +39,7 @@
 #include <asm/insn.h>
 #include <asm/debugreg.h>
 #include <asm/set_memory.h>
+#include <asm/sections.h>
 
 #include "common.h"
 
@@ -251,10 +252,12 @@ static int can_optimize(unsigned long paddr)
 
 	/*
 	 * Do not optimize in the entry code due to the unstable
-	 * stack handling.
+	 * stack handling and registers setup.
 	 */
-	if ((paddr >= (unsigned long)__entry_text_start) &&
-	    (paddr <  (unsigned long)__entry_text_end))
+	if (((paddr >= (unsigned long)__entry_text_start) &&
+	     (paddr <  (unsigned long)__entry_text_end)) ||
+	    ((paddr >= (unsigned long)__irqentry_text_start) &&
+	     (paddr <  (unsigned long)__irqentry_text_end)))
 		return 0;
 
 	/* Check there is enough space for a relative jump. */

[PATCH -tip v9 5/5] [BUGFIX] kprobes/x86: Do not jump-optimize kprobes on irq entry code

[Masami Hiramatsu]

Since the kernel segment registers are not prepared at the
entry of irq-entry code, if a kprobe on such code is
jump-optimized, accessing per-cpu variables may cause
kernel panic.
However, if the kprobe is not optimized, it kicks int3
exception and set segment registers correctly.

This checks probe-address and if it is in irq-entry code,
it prohibits optimizing such kprobes. This means we can
continuously probing such interrupt handlers by kprobes
but it is not optimized anymore.

Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
Reported-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Tested-by: Francis Deslauriers <francis.deslauriers@efficios.com>
---
 arch/x86/kernel/kprobes/opt.c |    9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/arch/x86/kernel/kprobes/opt.c b/arch/x86/kernel/kprobes/opt.c
index 69ea0bc1cfa3..4f98aad38237 100644
--- a/arch/x86/kernel/kprobes/opt.c
+++ b/arch/x86/kernel/kprobes/opt.c
@@ -39,6 +39,7 @@
 #include <asm/insn.h>
 #include <asm/debugreg.h>
 #include <asm/set_memory.h>
+#include <asm/sections.h>
 
 #include "common.h"
 
@@ -251,10 +252,12 @@ static int can_optimize(unsigned long paddr)
 
 	/*
 	 * Do not optimize in the entry code due to the unstable
-	 * stack handling.
+	 * stack handling and registers setup.
 	 */
-	if ((paddr >= (unsigned long)__entry_text_start) &&
-	    (paddr <  (unsigned long)__entry_text_end))
+	if (((paddr >= (unsigned long)__entry_text_start) &&
+	     (paddr <  (unsigned long)__entry_text_end)) ||
+	    ((paddr >= (unsigned long)__irqentry_text_start) &&
+	     (paddr <  (unsigned long)__irqentry_text_end)))
 		return 0;
 
 	/* Check there is enough space for a relative jump. */