Patchwork PowerPC: kernel: memory access violation when rtas_data_buf contents are more than 1026

login
register
mail settings
Submitter Chen Gang
Date April 24, 2013, 7:03 a.m.
Message ID <517783D8.6010701@asianux.com>
Download mbox | patch
Permalink /patch/239058/
State Rejected
Headers show

Comments

Chen Gang - April 24, 2013, 7:03 a.m.
On 2013年04月24日 14:28, Vasant Hegde wrote:
> On 04/23/2013 08:42 AM, Chen Gang wrote:
>>
>> need set '\0' for 'local_buffer'.
>>
>> SPLPAR_MAXLENGTH is 1026, RTAS_DATA_BUF_SIZE is 4096. so the contents of
>> rtas_data_buf may truncated in memcpy.
>>
>> if contents are really truncated.
>>    the splpar_strlen is more than 1026. the next while loop checking will
>>    not find the end of buffer. that will cause memory access violation.
>>
> 
> Per parameter length in ibm,get-system-parameter RTAS call is limited to
> 1026 bytes (1024 bytes of data + 2 bytes  length). And 'rtas_data_buf'
> was set to 0 (first 1026 bytes) before call RTAS call. At the worst if
> we get junk in RTAS output length field helps to exit from the while
> loop. So I don't think we need this patch.

Is get-system-parameter return the NUL terminated string ? if so, it
will no issue (just like your discription).

If it will not return NUL terminated string, please see line 326:

  "while ((*local_buffer) && (idx < splpar_strlen))"
  (when idx == 1024, *local_buffer is memory access violation).

Since we use the first 2 bytes as length, and also be sure of the real
length will never more than 1024, I suggest to:

---------------------------patch begin--------------------------------


---------------------------patch end----------------------------------

Thanks.
Vasant Hegde - April 24, 2013, 7:23 a.m.
On 04/24/2013 12:33 PM, Chen Gang wrote:
> On 2013年04月24日 14:28, Vasant Hegde wrote:
>> On 04/23/2013 08:42 AM, Chen Gang wrote:
>>>
>>> need set '\0' for 'local_buffer'.
>>>
>>> SPLPAR_MAXLENGTH is 1026, RTAS_DATA_BUF_SIZE is 4096. so the contents of
>>> rtas_data_buf may truncated in memcpy.
>>>
>>> if contents are really truncated.
>>>     the splpar_strlen is more than 1026. the next while loop checking will
>>>     not find the end of buffer. that will cause memory access violation.
>>>
>>
>> Per parameter length in ibm,get-system-parameter RTAS call is limited to
>> 1026 bytes (1024 bytes of data + 2 bytes  length). And 'rtas_data_buf'
>> was set to 0 (first 1026 bytes) before call RTAS call. At the worst if
>> we get junk in RTAS output length field helps to exit from the while
>> loop. So I don't think we need this patch.
>
> Is get-system-parameter return the NUL terminated string ? if so, it
> will no issue (just like your discription).
>

Length includes the length of the NULL. So (idx < splpar_strlen)
is safe. IMO existing code is proper.

-Vasant


> If it will not return NUL terminated string, please see line 326:
>
>    "while ((*local_buffer)&&  (idx<  splpar_strlen))"
>    (when idx == 1024, *local_buffer is memory access violation).
>
> Since we use the first 2 bytes as length, and also be sure of the real
> length will never more than 1024, I suggest to:
>
> ---------------------------patch begin--------------------------------
>
> diff --git a/arch/powerpc/kernel/lparcfg.c b/arch/powerpc/kernel/lparcfg.c
> index 801a757..f8bd7cf 100644
> --- a/arch/powerpc/kernel/lparcfg.c
> +++ b/arch/powerpc/kernel/lparcfg.c
> @@ -323,7 +323,7 @@ static void parse_system_parameter_string(struct seq_file *m)
>
>   		w_idx = 0;
>   		idx = 0;
> -		while ((*local_buffer)&&  (idx<  splpar_strlen)) {
> +		while (idx<  splpar_strlen) {
>   			workbuffer[w_idx++] = local_buffer[idx++];
>   			if ((local_buffer[idx] == ',')
>   			    || (local_buffer[idx] == '\0')) {
>
> ---------------------------patch end----------------------------------
>
> Thanks.
>
Chen Gang - April 24, 2013, 7:40 a.m.
On 2013年04月24日 15:23, Vasant Hegde wrote:
> On 04/24/2013 12:33 PM, Chen Gang wrote:
>> On 2013年04月24日 14:28, Vasant Hegde wrote:
>>> On 04/23/2013 08:42 AM, Chen Gang wrote:
>>>>
>>>> need set '\0' for 'local_buffer'.
>>>>
>>>> SPLPAR_MAXLENGTH is 1026, RTAS_DATA_BUF_SIZE is 4096. so the
>>>> contents of
>>>> rtas_data_buf may truncated in memcpy.
>>>>
>>>> if contents are really truncated.
>>>>     the splpar_strlen is more than 1026. the next while loop
>>>> checking will
>>>>     not find the end of buffer. that will cause memory access
>>>> violation.
>>>>
>>>
>>> Per parameter length in ibm,get-system-parameter RTAS call is limited to
>>> 1026 bytes (1024 bytes of data + 2 bytes  length). And 'rtas_data_buf'
>>> was set to 0 (first 1026 bytes) before call RTAS call. At the worst if
>>> we get junk in RTAS output length field helps to exit from the while
>>> loop. So I don't think we need this patch.
>>
>> Is get-system-parameter return the NUL terminated string ? if so, it
>> will no issue (just like your discription).
>>
> 
> Length includes the length of the NULL. So (idx < splpar_strlen)
> is safe. IMO existing code is proper.

OK, since it is not an issue, I will try another patches for powerpc POWER7.

  :-)

Patch

diff --git a/arch/powerpc/kernel/lparcfg.c b/arch/powerpc/kernel/lparcfg.c
index 801a757..f8bd7cf 100644
--- a/arch/powerpc/kernel/lparcfg.c
+++ b/arch/powerpc/kernel/lparcfg.c
@@ -323,7 +323,7 @@  static void parse_system_parameter_string(struct seq_file *m)
 
 		w_idx = 0;
 		idx = 0;
-		while ((*local_buffer) && (idx < splpar_strlen)) {
+		while (idx < splpar_strlen) {
 			workbuffer[w_idx++] = local_buffer[idx++];
 			if ((local_buffer[idx] == ',')
 			    || (local_buffer[idx] == '\0')) {