[PATCH 0/1] x86/kprobes: Prohibit probing of .entry_trampoline code

[PATCH 0/1] x86/kprobes: Prohibit probing of .entry_trampoline code

[Francis Deslauriers]

Hi all,

While fuzzing the Perf kprobe interface, I found that adding a probe on
the 'entry_SYSCALL_64_trampoline' symbol will crash my 4.16-rc4
kernel(661e50bc853209e41a5c14a290ca4decc43cbfd1) on a x86_64 Qemu VM.

How to reproduce:
	echo 'p:event1 entry_SYSCALL_64_trampoline' > ./kprobe_events
	echo 1 >  events/kprobes/enable
Crash log:[1]

My understanding is that the userspace CR3 register has not yet been
replaced by the kernel's CR3, when the kprobe is triggered. This means
that the kernel addresses can not be translated, thus making the
handling of the kprobe impossible.

This can be fixed by blacklisting the .entry_trampoline section. See
patch[1/1].

Here is the config I am using[2].

Thanks,

Francis Deslauriers
EfficiOS inc.

1:http://paste.ubuntu.com/p/djnpZCzQKv/
2:http://paste.ubuntu.com/p/3jrFYt6XQB/

Francis Deslauriers (1):
  x86/kprobes: Prohibit probing of .entry_trampoline code

 arch/x86/include/asm/sections.h |  1 +
 arch/x86/kernel/kprobes/core.c  | 10 +++++++++-
 arch/x86/kernel/vmlinux.lds.S   |  2 ++
 3 files changed, 12 insertions(+), 1 deletion(-)

-- 
2.7.4

[PATCH 1/1] x86/kprobes: Prohibit probing of .entry_trampoline code

[Francis Deslauriers]

.entry_trampoline is a code area that is used to ensure page table
isolation between userspace and kernelspace.

At the beginning of the execution of the trampoline, we load the
kernel's CR3 register. This has the effect of enabling the translation
of the kernel virtual addresses to physical addresses. Before this
happens most kernel addresses can not be translated because the running
process' CR3 is still used.

If a kprobe is placed on the trampoline code before that change of the
CR3 register happens the kernel crashes because int3 handling pages are
not accessible.

To fix this, add the .entry_trampoline section to the kprobe blacklist
to prohibit the probing of code before all the kernel pages are
accessible.

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
---
 arch/x86/include/asm/sections.h |  1 +
 arch/x86/kernel/kprobes/core.c  | 10 +++++++++-
 arch/x86/kernel/vmlinux.lds.S   |  2 ++
 3 files changed, 12 insertions(+), 1 deletion(-)

diff --git a/arch/x86/include/asm/sections.h b/arch/x86/include/asm/sections.h
index d6baf23..5c019d2 100644
--- a/arch/x86/include/asm/sections.h
+++ b/arch/x86/include/asm/sections.h
@@ -10,6 +10,7 @@ extern struct exception_table_entry __stop___ex_table[];
 
 #if defined(CONFIG_X86_64)
 extern char __end_rodata_hpage_align[];
+extern char __entry_trampoline_start[], __entry_trampoline_end[];
 #endif
 
 #endif	/* _ASM_X86_SECTIONS_H */
diff --git a/arch/x86/kernel/kprobes/core.c b/arch/x86/kernel/kprobes/core.c
index bd36f3c..0715f82 100644
--- a/arch/x86/kernel/kprobes/core.c
+++ b/arch/x86/kernel/kprobes/core.c
@@ -1168,10 +1168,18 @@ NOKPROBE_SYMBOL(longjmp_break_handler);
 
 bool arch_within_kprobe_blacklist(unsigned long addr)
 {
+	bool is_in_entry_trampoline_section = false;
+
+#ifdef CONFIG_X86_64
+	is_in_entry_trampoline_section =
+		(addr >= (unsigned long)__entry_trampoline_start &&
+		 addr < (unsigned long)__entry_trampoline_end);
+#endif
 	return  (addr >= (unsigned long)__kprobes_text_start &&
 		 addr < (unsigned long)__kprobes_text_end) ||
 		(addr >= (unsigned long)__entry_text_start &&
-		 addr < (unsigned long)__entry_text_end);
+		 addr < (unsigned long)__entry_text_end) ||
+		is_in_entry_trampoline_section;
 }
 
 int __init arch_init_kprobes(void)
diff --git a/arch/x86/kernel/vmlinux.lds.S b/arch/x86/kernel/vmlinux.lds.S
index 9b138a0..b854ebf 100644
--- a/arch/x86/kernel/vmlinux.lds.S
+++ b/arch/x86/kernel/vmlinux.lds.S
@@ -118,9 +118,11 @@ SECTIONS
 
 #ifdef CONFIG_X86_64
 		. = ALIGN(PAGE_SIZE);
+		VMLINUX_SYMBOL(__entry_trampoline_start) = .;
 		_entry_trampoline = .;
 		*(.entry_trampoline)
 		. = ALIGN(PAGE_SIZE);
+		VMLINUX_SYMBOL(__entry_trampoline_end) = .;
 		ASSERT(. - _entry_trampoline == PAGE_SIZE, "entry trampoline is too big");
 #endif
 
-- 
2.7.4

Re: [PATCH 1/1] x86/kprobes: Prohibit probing of .entry_trampoline code

[Masami Hiramatsu]

On Thu,  8 Mar 2018 22:18:12 -0500
Francis Deslauriers <francis.deslauriers@efficios.com> wrote:

> .entry_trampoline is a code area that is used to ensure page table
> isolation between userspace and kernelspace.
> 
> At the beginning of the execution of the trampoline, we load the
> kernel's CR3 register. This has the effect of enabling the translation
> of the kernel virtual addresses to physical addresses. Before this
> happens most kernel addresses can not be translated because the running
> process' CR3 is still used.
> 
> If a kprobe is placed on the trampoline code before that change of the
> CR3 register happens the kernel crashes because int3 handling pages are
> not accessible.
> 
> To fix this, add the .entry_trampoline section to the kprobe blacklist
> to prohibit the probing of code before all the kernel pages are
> accessible.

OK, looks good to me.

Acked-by: Masami Hiramatsu <mhiramat@kernel.org>

Ingo, could you pick it as an urgent fix?

Thanks!

> 
> Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
> ---
>  arch/x86/include/asm/sections.h |  1 +
>  arch/x86/kernel/kprobes/core.c  | 10 +++++++++-
>  arch/x86/kernel/vmlinux.lds.S   |  2 ++
>  3 files changed, 12 insertions(+), 1 deletion(-)
> 
> diff --git a/arch/x86/include/asm/sections.h b/arch/x86/include/asm/sections.h
> index d6baf23..5c019d2 100644
> --- a/arch/x86/include/asm/sections.h
> +++ b/arch/x86/include/asm/sections.h
> @@ -10,6 +10,7 @@ extern struct exception_table_entry __stop___ex_table[];
>  
>  #if defined(CONFIG_X86_64)
>  extern char __end_rodata_hpage_align[];
> +extern char __entry_trampoline_start[], __entry_trampoline_end[];
>  #endif
>  
>  #endif	/* _ASM_X86_SECTIONS_H */
> diff --git a/arch/x86/kernel/kprobes/core.c b/arch/x86/kernel/kprobes/core.c
> index bd36f3c..0715f82 100644
> --- a/arch/x86/kernel/kprobes/core.c
> +++ b/arch/x86/kernel/kprobes/core.c
> @@ -1168,10 +1168,18 @@ NOKPROBE_SYMBOL(longjmp_break_handler);
>  
>  bool arch_within_kprobe_blacklist(unsigned long addr)
>  {
> +	bool is_in_entry_trampoline_section = false;
> +
> +#ifdef CONFIG_X86_64
> +	is_in_entry_trampoline_section =
> +		(addr >= (unsigned long)__entry_trampoline_start &&
> +		 addr < (unsigned long)__entry_trampoline_end);
> +#endif
>  	return  (addr >= (unsigned long)__kprobes_text_start &&
>  		 addr < (unsigned long)__kprobes_text_end) ||
>  		(addr >= (unsigned long)__entry_text_start &&
> -		 addr < (unsigned long)__entry_text_end);
> +		 addr < (unsigned long)__entry_text_end) ||
> +		is_in_entry_trampoline_section;
>  }
>  
>  int __init arch_init_kprobes(void)
> diff --git a/arch/x86/kernel/vmlinux.lds.S b/arch/x86/kernel/vmlinux.lds.S
> index 9b138a0..b854ebf 100644
> --- a/arch/x86/kernel/vmlinux.lds.S
> +++ b/arch/x86/kernel/vmlinux.lds.S
> @@ -118,9 +118,11 @@ SECTIONS
>  
>  #ifdef CONFIG_X86_64
>  		. = ALIGN(PAGE_SIZE);
> +		VMLINUX_SYMBOL(__entry_trampoline_start) = .;
>  		_entry_trampoline = .;
>  		*(.entry_trampoline)
>  		. = ALIGN(PAGE_SIZE);
> +		VMLINUX_SYMBOL(__entry_trampoline_end) = .;
>  		ASSERT(. - _entry_trampoline == PAGE_SIZE, "entry trampoline is too big");
>  #endif
>  
> -- 
> 2.7.4
> 


-- 
Masami Hiramatsu <mhiramat@kernel.org>

[tip:x86/pti] x86/kprobes: Fix kernel crash when probing .entry_trampoline code

[tip-bot for Francis Deslauriers]

Commit-ID:  c07a8f8b08ba683ea24f3ac9159f37ae94daf47f
Gitweb:     https://git.kernel.org/tip/c07a8f8b08ba683ea24f3ac9159f37ae94daf47f
Author:     Francis Deslauriers <francis.deslauriers@efficios.com>
AuthorDate: Thu, 8 Mar 2018 22:18:12 -0500
Committer:  Ingo Molnar <mingo@kernel.org>
CommitDate: Fri, 9 Mar 2018 09:58:36 +0100

x86/kprobes: Fix kernel crash when probing .entry_trampoline code

Disable the kprobe probing of the entry trampoline:

.entry_trampoline is a code area that is used to ensure page table
isolation between userspace and kernelspace.

At the beginning of the execution of the trampoline, we load the
kernel's CR3 register. This has the effect of enabling the translation
of the kernel virtual addresses to physical addresses. Before this
happens most kernel addresses can not be translated because the running
process' CR3 is still used.

If a kprobe is placed on the trampoline code before that change of the
CR3 register happens the kernel crashes because int3 handling pages are
not accessible.

To fix this, add the .entry_trampoline section to the kprobe blacklist
to prohibit the probing of code before all the kernel pages are
accessible.

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: mathieu.desnoyers@efficios.com
Cc: mhiramat@kernel.org
Link: http://lkml.kernel.org/r/1520565492-4637-2-git-send-email-francis.deslauriers@efficios.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
---
 arch/x86/include/asm/sections.h |  1 +
 arch/x86/kernel/kprobes/core.c  | 10 +++++++++-
 arch/x86/kernel/vmlinux.lds.S   |  2 ++
 3 files changed, 12 insertions(+), 1 deletion(-)

diff --git a/arch/x86/include/asm/sections.h b/arch/x86/include/asm/sections.h
index d6baf23782bc..5c019d23d06b 100644
--- a/arch/x86/include/asm/sections.h
+++ b/arch/x86/include/asm/sections.h
@@ -10,6 +10,7 @@ extern struct exception_table_entry __stop___ex_table[];
 
 #if defined(CONFIG_X86_64)
 extern char __end_rodata_hpage_align[];
+extern char __entry_trampoline_start[], __entry_trampoline_end[];
 #endif
 
 #endif	/* _ASM_X86_SECTIONS_H */
diff --git a/arch/x86/kernel/kprobes/core.c b/arch/x86/kernel/kprobes/core.c
index bd36f3c33cd0..0715f827607c 100644
--- a/arch/x86/kernel/kprobes/core.c
+++ b/arch/x86/kernel/kprobes/core.c
@@ -1168,10 +1168,18 @@ NOKPROBE_SYMBOL(longjmp_break_handler);
 
 bool arch_within_kprobe_blacklist(unsigned long addr)
 {
+	bool is_in_entry_trampoline_section = false;
+
+#ifdef CONFIG_X86_64
+	is_in_entry_trampoline_section =
+		(addr >= (unsigned long)__entry_trampoline_start &&
+		 addr < (unsigned long)__entry_trampoline_end);
+#endif
 	return  (addr >= (unsigned long)__kprobes_text_start &&
 		 addr < (unsigned long)__kprobes_text_end) ||
 		(addr >= (unsigned long)__entry_text_start &&
-		 addr < (unsigned long)__entry_text_end);
+		 addr < (unsigned long)__entry_text_end) ||
+		is_in_entry_trampoline_section;
 }
 
 int __init arch_init_kprobes(void)
diff --git a/arch/x86/kernel/vmlinux.lds.S b/arch/x86/kernel/vmlinux.lds.S
index 9b138a06c1a4..b854ebf5851b 100644
--- a/arch/x86/kernel/vmlinux.lds.S
+++ b/arch/x86/kernel/vmlinux.lds.S
@@ -118,9 +118,11 @@ SECTIONS
 
 #ifdef CONFIG_X86_64
 		. = ALIGN(PAGE_SIZE);
+		VMLINUX_SYMBOL(__entry_trampoline_start) = .;
 		_entry_trampoline = .;
 		*(.entry_trampoline)
 		. = ALIGN(PAGE_SIZE);
+		VMLINUX_SYMBOL(__entry_trampoline_end) = .;
 		ASSERT(. - _entry_trampoline == PAGE_SIZE, "entry trampoline is too big");
 #endif
 

Re: [PATCH 0/1] x86/kprobes: Prohibit probing of .entry_trampoline code

[Masami Hiramatsu]

On Thu,  8 Mar 2018 22:18:11 -0500
Francis Deslauriers <francis.deslauriers@efficios.com> wrote:

> Hi all,
> 
> While fuzzing the Perf kprobe interface, I found that adding a probe on
> the 'entry_SYSCALL_64_trampoline' symbol will crash my 4.16-rc4
> kernel(661e50bc853209e41a5c14a290ca4decc43cbfd1) on a x86_64 Qemu VM.
> 
> How to reproduce:
> 	echo 'p:event1 entry_SYSCALL_64_trampoline' > ./kprobe_events
> 	echo 1 >  events/kprobes/enable
> Crash log:[1]
> 
> My understanding is that the userspace CR3 register has not yet been
> replaced by the kernel's CR3, when the kprobe is triggered. This means
> that the kernel addresses can not be translated, thus making the
> handling of the kprobe impossible.

Thanks for reporting!
And yes, all entry code must be nokprobe.

> 
> This can be fixed by blacklisting the .entry_trampoline section. See
> patch[1/1].
> 
> Here is the config I am using[2].
> 
> Thanks,
> 
> Francis Deslauriers
> EfficiOS inc.
> 
> 1:http://paste.ubuntu.com/p/djnpZCzQKv/
> 2:http://paste.ubuntu.com/p/3jrFYt6XQB/
> 
> Francis Deslauriers (1):
>   x86/kprobes: Prohibit probing of .entry_trampoline code
> 
>  arch/x86/include/asm/sections.h |  1 +
>  arch/x86/kernel/kprobes/core.c  | 10 +++++++++-
>  arch/x86/kernel/vmlinux.lds.S   |  2 ++
>  3 files changed, 12 insertions(+), 1 deletion(-)
> 
> -- 
> 2.7.4
> 


-- 
Masami Hiramatsu <mhiramat@kernel.org>