[PATCH 1/2] x86/insn: Fix some potential undefined behavior.

[PATCH 1/2] x86/insn: Fix some potential undefined behavior.

[Ian Rogers]

From: Numfor Mbiziwo-Tiapo <nums@google.com>

If insn_init is given a NULL kaddr and 0 buflen then validate_next will
perform arithmetic on NULL, add a guard to avoid this.

Don't perform unaligned loads in __get_next and __peek_nbyte_next as
these are forms of undefined behavior.

These problems were identified using the undefined behavior sanitizer
(ubsan) with the tools version of the code and perf test. Part of this
patch was previously posted here:
https://lore.kernel.org/lkml/20190724184512.162887-4-nums@google.com/

Signed-off-by: Ian Rogers <irogers@google.com>
Signed-off-by: Numfor Mbiziwo-Tiapo <nums@google.com>
---
 arch/x86/lib/insn.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/arch/x86/lib/insn.c b/arch/x86/lib/insn.c
index 404279563891..57236940de46 100644
--- a/arch/x86/lib/insn.c
+++ b/arch/x86/lib/insn.c
@@ -17,13 +17,13 @@
 
 /* Verify next sizeof(t) bytes can be on the same instruction */
 #define validate_next(t, insn, n)	\
-	((insn)->next_byte + sizeof(t) + n <= (insn)->end_kaddr)
+	((insn)->end_kaddr != 0 && (insn)->next_byte + sizeof(t) + n <= (insn)->end_kaddr)
 
 #define __get_next(t, insn)	\
-	({ t r = *(t*)insn->next_byte; insn->next_byte += sizeof(t); r; })
+	({ t r; memcpy(&r, insn->next_byte, sizeof(t)); insn->next_byte += sizeof(t); r; })
 
 #define __peek_nbyte_next(t, insn, n)	\
-	({ t r = *(t*)((insn)->next_byte + n); r; })
+	({ t r; memcpy(&r, (insn)->next_byte + n, sizeof(t)); r; })
 
 #define get_next(t, insn)	\
 	({ if (unlikely(!validate_next(t, insn, 0))) goto err_out; __get_next(t, insn); })
-- 
2.28.0.1011.ga647a8990f-goog

[PATCH 2/2] tools/x86: Fix some potential undefined behavior

[Ian Rogers]

From: Numfor Mbiziwo-Tiapo <nums@google.com>

If insn_init is given a NULL kaddr and 0 buflen then validate_next will
perform arithmetic on NULL, add a guard to avoid this.

Don't perform unaligned loads in __get_next and __peek_nbyte_next as
these are forms of undefined behavior.

These problems were identified using the undefined behavior sanitizer
(ubsan) with  perf test. Part of this patch was previously posted here:
https://lore.kernel.org/lkml/20190724184512.162887-4-nums@google.com/

Signed-off-by: Ian Rogers <irogers@google.com>
Signed-off-by: Numfor Mbiziwo-Tiapo <nums@google.com>
---
 tools/arch/x86/lib/insn.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/tools/arch/x86/lib/insn.c b/tools/arch/x86/lib/insn.c
index 0151dfc6da61..e8874a8cac2c 100644
--- a/tools/arch/x86/lib/insn.c
+++ b/tools/arch/x86/lib/insn.c
@@ -17,13 +17,13 @@
 
 /* Verify next sizeof(t) bytes can be on the same instruction */
 #define validate_next(t, insn, n)	\
-	((insn)->next_byte + sizeof(t) + n <= (insn)->end_kaddr)
+	((insn)->end_kaddr != 0 && (insn)->next_byte + sizeof(t) + n <= (insn)->end_kaddr)
 
 #define __get_next(t, insn)	\
-	({ t r = *(t*)insn->next_byte; insn->next_byte += sizeof(t); r; })
+	({ t r; memcpy(&r, insn->next_byte, sizeof(t)); insn->next_byte += sizeof(t); r; })
 
 #define __peek_nbyte_next(t, insn, n)	\
-	({ t r = *(t*)((insn)->next_byte + n); r; })
+	({ t r; memcpy(&r, (insn)->next_byte + n, sizeof(t)); r; })
 
 #define get_next(t, insn)	\
 	({ if (unlikely(!validate_next(t, insn, 0))) goto err_out; __get_next(t, insn); })
-- 
2.28.0.1011.ga647a8990f-goog

Re: [PATCH 2/2] tools/x86: Fix some potential undefined behavior

[Masami Hiramatsu]

Hi,

Please merge the change on arch/x86/lib/insn.c and tools/arch/x86/lib/insn.c
to a single patch for bisecting.

Thank you,

On Wed, 14 Oct 2020 23:21:48 -0700
Ian Rogers <irogers@google.com> wrote:

> From: Numfor Mbiziwo-Tiapo <nums@google.com>
> 
> If insn_init is given a NULL kaddr and 0 buflen then validate_next will
> perform arithmetic on NULL, add a guard to avoid this.
> 
> Don't perform unaligned loads in __get_next and __peek_nbyte_next as
> these are forms of undefined behavior.
> 
> These problems were identified using the undefined behavior sanitizer
> (ubsan) with  perf test. Part of this patch was previously posted here:
> https://lore.kernel.org/lkml/20190724184512.162887-4-nums@google.com/
> 
> Signed-off-by: Ian Rogers <irogers@google.com>
> Signed-off-by: Numfor Mbiziwo-Tiapo <nums@google.com>
> ---
>  tools/arch/x86/lib/insn.c | 6 +++---
>  1 file changed, 3 insertions(+), 3 deletions(-)
> 
> diff --git a/tools/arch/x86/lib/insn.c b/tools/arch/x86/lib/insn.c
> index 0151dfc6da61..e8874a8cac2c 100644
> --- a/tools/arch/x86/lib/insn.c
> +++ b/tools/arch/x86/lib/insn.c
> @@ -17,13 +17,13 @@
>  
>  /* Verify next sizeof(t) bytes can be on the same instruction */
>  #define validate_next(t, insn, n)	\
> -	((insn)->next_byte + sizeof(t) + n <= (insn)->end_kaddr)
> +	((insn)->end_kaddr != 0 && (insn)->next_byte + sizeof(t) + n <= (insn)->end_kaddr)
>  
>  #define __get_next(t, insn)	\
> -	({ t r = *(t*)insn->next_byte; insn->next_byte += sizeof(t); r; })
> +	({ t r; memcpy(&r, insn->next_byte, sizeof(t)); insn->next_byte += sizeof(t); r; })
>  
>  #define __peek_nbyte_next(t, insn, n)	\
> -	({ t r = *(t*)((insn)->next_byte + n); r; })
> +	({ t r; memcpy(&r, (insn)->next_byte + n, sizeof(t)); r; })
>  
>  #define get_next(t, insn)	\
>  	({ if (unlikely(!validate_next(t, insn, 0))) goto err_out; __get_next(t, insn); })
> -- 
> 2.28.0.1011.ga647a8990f-goog
> 


-- 
Masami Hiramatsu <mhiramat@kernel.org>

Re: [PATCH 1/2] x86/insn: Fix some potential undefined behavior.

[Ingo Molnar]

* Ian Rogers <irogers@google.com> wrote:

> From: Numfor Mbiziwo-Tiapo <nums@google.com>
> 
> If insn_init is given a NULL kaddr and 0 buflen then validate_next will
> perform arithmetic on NULL, add a guard to avoid this.
> 
> Don't perform unaligned loads in __get_next and __peek_nbyte_next as
> these are forms of undefined behavior.

So, 'insn' is a kernel structure, usually allocated on the kernel stack. 
How could these fields ever be unaligned?

> 
> These problems were identified using the undefined behavior sanitizer
> (ubsan) with the tools version of the code and perf test. Part of this
> patch was previously posted here:
> https://lore.kernel.org/lkml/20190724184512.162887-4-nums@google.com/
> 
> Signed-off-by: Ian Rogers <irogers@google.com>
> Signed-off-by: Numfor Mbiziwo-Tiapo <nums@google.com>
> ---
>  arch/x86/lib/insn.c | 6 +++---
>  1 file changed, 3 insertions(+), 3 deletions(-)
> 
> diff --git a/arch/x86/lib/insn.c b/arch/x86/lib/insn.c
> index 404279563891..57236940de46 100644
> --- a/arch/x86/lib/insn.c
> +++ b/arch/x86/lib/insn.c
> @@ -17,13 +17,13 @@
>  
>  /* Verify next sizeof(t) bytes can be on the same instruction */
>  #define validate_next(t, insn, n)	\
> -	((insn)->next_byte + sizeof(t) + n <= (insn)->end_kaddr)
> +	((insn)->end_kaddr != 0 && (insn)->next_byte + sizeof(t) + n <= (insn)->end_kaddr)
>  
>  #define __get_next(t, insn)	\
> -	({ t r = *(t*)insn->next_byte; insn->next_byte += sizeof(t); r; })
> +	({ t r; memcpy(&r, insn->next_byte, sizeof(t)); insn->next_byte += sizeof(t); r; })
>  
>  #define __peek_nbyte_next(t, insn, n)	\
> -	({ t r = *(t*)((insn)->next_byte + n); r; })
> +	({ t r; memcpy(&r, (insn)->next_byte + n, sizeof(t)); r; })
>  
>  #define get_next(t, insn)	\
>  	({ if (unlikely(!validate_next(t, insn, 0))) goto err_out; __get_next(t, insn); })

Is there any code generation side effect of this change to the resulting 
code?

Thanks,

	Ingo

Re: [PATCH 1/2] x86/insn: Fix some potential undefined behavior.

[Masami Hiramatsu]

On Wed, 14 Oct 2020 23:21:47 -0700
Ian Rogers <irogers@google.com> wrote:

> From: Numfor Mbiziwo-Tiapo <nums@google.com>
> 
> If insn_init is given a NULL kaddr and 0 buflen then validate_next will
> perform arithmetic on NULL, add a guard to avoid this.

Maybe we should check the kaddr and end_kaddr existence in insn_init().
At least end_kaddr != 0 can be checked in insn_init() because it will
not be updated, right?
Actually, I expected that the caller checked it, but if you concern
the case, insn_init() must return -EINVAL for such case.

> Don't perform unaligned loads in __get_next and __peek_nbyte_next as
> these are forms of undefined behavior.

Hm, I thought x86 could handle such unaligned access. Of course, recent
change made the insn.c complied on another arch, so maybe it should
be treated. 

BTW (and OT), would we have to take care all of those unaligned
access? ( e.g. int *p = somewhere; while (*p++) ... ?)

Also, you have to update the copy of this file under tools/ too.

Thank you,

> 
> These problems were identified using the undefined behavior sanitizer
> (ubsan) with the tools version of the code and perf test. Part of this
> patch was previously posted here:
> https://lore.kernel.org/lkml/20190724184512.162887-4-nums@google.com/
> 
> Signed-off-by: Ian Rogers <irogers@google.com>
> Signed-off-by: Numfor Mbiziwo-Tiapo <nums@google.com>
> ---
>  arch/x86/lib/insn.c | 6 +++---
>  1 file changed, 3 insertions(+), 3 deletions(-)
> 
> diff --git a/arch/x86/lib/insn.c b/arch/x86/lib/insn.c
> index 404279563891..57236940de46 100644
> --- a/arch/x86/lib/insn.c
> +++ b/arch/x86/lib/insn.c
> @@ -17,13 +17,13 @@
>  
>  /* Verify next sizeof(t) bytes can be on the same instruction */
>  #define validate_next(t, insn, n)	\
> -	((insn)->next_byte + sizeof(t) + n <= (insn)->end_kaddr)
> +	((insn)->end_kaddr != 0 && (insn)->next_byte + sizeof(t) + n <= (insn)->end_kaddr)
>  
>  #define __get_next(t, insn)	\
> -	({ t r = *(t*)insn->next_byte; insn->next_byte += sizeof(t); r; })
> +	({ t r; memcpy(&r, insn->next_byte, sizeof(t)); insn->next_byte += sizeof(t); r; })
>  
>  #define __peek_nbyte_next(t, insn, n)	\
> -	({ t r = *(t*)((insn)->next_byte + n); r; })
> +	({ t r; memcpy(&r, (insn)->next_byte + n, sizeof(t)); r; })
>  
>  #define get_next(t, insn)	\
>  	({ if (unlikely(!validate_next(t, insn, 0))) goto err_out; __get_next(t, insn); })
> -- 
> 2.28.0.1011.ga647a8990f-goog
> 


-- 
Masami Hiramatsu <mhiramat@kernel.org>

Re: [PATCH 1/2] x86/insn: Fix some potential undefined behavior.

[Peter Zijlstra]

On Wed, Oct 14, 2020 at 11:21:47PM -0700, Ian Rogers wrote:
> From: Numfor Mbiziwo-Tiapo <nums@google.com>
> 
> If insn_init is given a NULL kaddr and 0 buflen then validate_next will
> perform arithmetic on NULL, add a guard to avoid this.

How is this a problem? NULL is (void *)0, you can do arithmetic on that
just fine.

Is UBSAN taking drugs again?

> Don't perform unaligned loads in __get_next and __peek_nbyte_next as
> these are forms of undefined behavior.

Fair enough; that could actually be a problem when we start to
cross-build this stuff. A RISC hosted version of the x86 decoder could
indeed trip this up.


But also, these are two changes in one patch.