[PATCH] arch/sparc: additional len check in loop for prom_getbootargs

[PATCH] arch/sparc: additional len check in loop for prom_getbootargs

[Chen Gang]

  when cp >= barg_buf + BARG_LEN-2, it only break internel loop (while)
  but outside loop (for) still has effect, and "*cp++ = ' '" repeating
  so need additional checking for it.


Signed-off-by: Chen Gang <gang.chen@asianux.com>
---
 arch/sparc/prom/bootstr_32.c |    4 ++++
 1 file changed, 4 insertions(+)

diff --git a/arch/sparc/prom/bootstr_32.c b/arch/sparc/prom/bootstr_32.c
index f5ec32e..25cda1c 100644
--- a/arch/sparc/prom/bootstr_32.c
+++ b/arch/sparc/prom/bootstr_32.c
@@ -40,6 +40,10 @@ prom_getbootargs(void)
 				*cp++ = *arg++;
 			}
 			*cp++ = ' ';
+			if(cp >= barg_buf + BARG_LEN-1){
+				/* We might issue a warning here. */
+				break;
+			}
 		}
 		*cp = 0;
 		break;
-- 
1.7.9.5

Re: [PATCH] arch/sparc: additional len check in loop for prom_getbootargs

[Sam Ravnborg]

Hi Cheng.
On Thu, Nov 08, 2012 at 11:41:39AM +0800, Chen Gang wrote:
> 
>   when cp >= barg_buf + BARG_LEN-2, it only break internel loop (while)
>   but outside loop (for) still has effect, and "*cp++ = ' '" repeating
>   so need additional checking for it.
> 
> 
> Signed-off-by: Chen Gang <gang.chen@asianux.com>

I wonder how you found this bug?!?!
Anyway please consider this alternative fix:

diff --git a/arch/sparc/prom/bootstr_32.c b/arch/sparc/prom/bootstr_32.c
index f5ec32e..4ce602f 100644
--- a/arch/sparc/prom/bootstr_32.c
+++ b/arch/sparc/prom/bootstr_32.c
@@ -31,14 +31,10 @@ prom_getbootargs(void)
 			arg = (*(romvec->pv_v0bootargs))->argv[iter];
 			if (arg = NULL)
 				break;
-			while(*arg != 0) {
-				/* Leave place for space and null. */
-				if(cp >= barg_buf + BARG_LEN-2){
-					/* We might issue a warning here. */
-					break;
-				}
+			while (*arg != 0 && cp < (barg_buf + BARG_LEN - 2))
 				*cp++ = *arg++;
-			}
+
+			/* Append trailing space + null */
 			*cp++ = ' ';
 		}
 		*cp = 0;


Adding the conditional inside the while loop makes
the logic simpler. And the patch actually deletes more lines than it adds.
And please take care to follow coding style too. In particular spaces around operators.

The old code does not follow coding style - but this is no excuse.

Note - the above is not even build tested!

If you use the above code-snippet you can add my:
Acked-by: Sam Ravnborg <sam@ravnborg.org>

	Sam

Re: [PATCH] arch/sparc: additional len check in loop for prom_getbootargs

[Chen Gang]

Hell David Miller:

1) Is this patch applied ?  or need some discussion ?  or suspended ?

2) If possible: 
   A) can you tell me about our patch processing work flow "
   B) or some reference links (or document location) ?

thanks.


于 2012年11月08日 11:41, Chen Gang 写道:
> 
>   when cp >= barg_buf + BARG_LEN-2, it only break internel loop (while)
>   but outside loop (for) still has effect, and "*cp++ = ' '" repeating
>   so need additional checking for it.
> 
> 
> Signed-off-by: Chen Gang <gang.chen@asianux.com>
> ---
>  arch/sparc/prom/bootstr_32.c |    4 ++++
>  1 file changed, 4 insertions(+)
> 
> diff --git a/arch/sparc/prom/bootstr_32.c b/arch/sparc/prom/bootstr_32.c
> index f5ec32e..25cda1c 100644
> --- a/arch/sparc/prom/bootstr_32.c
> +++ b/arch/sparc/prom/bootstr_32.c
> @@ -40,6 +40,10 @@ prom_getbootargs(void)
>  				*cp++ = *arg++;
>  			}
>  			*cp++ = ' ';
> +			if(cp >= barg_buf + BARG_LEN-1){
> +				/* We might issue a warning here. */
> +				break;
> +			}
>  		}
>  		*cp = 0;
>  		break;
> 


-- 
Chen Gang

Asianux Corporation
--
To unsubscribe from this list: send the line "unsubscribe sparclinux" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] arch/sparc: additional len check in loop for prom_getbootargs

[David Miller]

I'm backlogged and often without internet access, and therefore
concentrating my limited time on kernel networking and much more
important sparc changes such as bug fixes for problems people will
actually hit.

So you'll just need to be very patient.

Re: [PATCH] arch/sparc: additional len check in loop for prom_getbootargs

[Chen Gang]

于 2012年11月09日 11:16, David Miller 写道:
> 
> I'm backlogged and often without internet access, and therefore
> concentrating my limited time on kernel networking and much more
> important sparc changes such as bug fixes for problems people will
> actually hit.
> 
> So you'll just need to be very patient.
> 
> 

ok, thanks, that is truly what I shall be.

and I suggest:

1) can write a document for patch work flow and time waiting document

2) put it in a suitable location (can be easily found by new members).

3) I think this document is useful for the new members.


thanks

-- 
Chen Gang

Asianux Corporation

Re: [PATCH] arch/sparc: additional len check in loop for prom_getbootargs

[Julian Calaby]

Hi Chen,

On Fri, Nov 9, 2012 at 2:15 PM, Chen Gang <gang.chen@asianux.com> wrote:
> Hell David Miller:
>
> 1) Is this patch applied ?  or need some discussion ?  or suspended ?
>
> 2) If possible:
>    A) can you tell me about our patch processing work flow "
>    B) or some reference links (or document location) ?

In regards to patches, you should assume that maintainers are:
1. grumpy
2. overworked
3. looking at your patch as you email them about it.

(Note that only #2 is true, but it saves everyone time and aggravation
if you assume the rest)

11 hours is not long enough for David to get to your patch, review it
and apply it. I don't believe that there are any maintainers which
would get to a patch that quickly.

If David hadn't applied the patch within a few days of it being
posted, then you might want to gently ask what's up. However, as Sam
reviewed the patch and had some comments about it, it's unlikely that
David will even look at it until you've addressed Sam's feedback.

The best way to get a patch into the kernel is to follow the rules in
Documentation/SubmittingPatches which you have, respond quickly to any
feedback, be patient, be prepared to explain why you're doing what
you're doing, and be prepared to change and resubmit your patch many
times until everyone is happy with it.

I'm sure that this is all written down in a document somewhere, but I
can't recall what or where it is.

Thanks,

-- 
Julian Calaby

Email: julian.calaby@gmail.com
Profile: http://www.google.com/profiles/julian.calaby/
.Plan: http://sites.google.com/site/juliancalaby/

Re: [PATCH] arch/sparc: additional len check in loop for prom_getbootargs

[Chen Gang]

Hi Julian Calaby:


thank you for your reply, firstly.

1) I think what you reply are valuable (at least, for me, it is true).

2) I have read Documents/SubmittingPatches (which Eric Dumazet suggested)

3) I also think, it will be better, if put part contents of your reply into Documents/SubmittingPatches as completion. 

thanks.

gchen.


于 2012年11月09日 13:06, Julian Calaby 写道:
> Hi Chen,
> 
> On Fri, Nov 9, 2012 at 2:15 PM, Chen Gang <gang.chen@asianux.com> wrote:
>> Hell David Miller:
>>
>> 1) Is this patch applied ?  or need some discussion ?  or suspended ?
>>
>> 2) If possible:
>>    A) can you tell me about our patch processing work flow "
>>    B) or some reference links (or document location) ?
> 
> In regards to patches, you should assume that maintainers are:
> 1. grumpy
> 2. overworked
> 3. looking at your patch as you email them about it.
> 
> (Note that only #2 is true, but it saves everyone time and aggravation
> if you assume the rest)
> 
> 11 hours is not long enough for David to get to your patch, review it
> and apply it. I don't believe that there are any maintainers which
> would get to a patch that quickly.
> 
> If David hadn't applied the patch within a few days of it being
> posted, then you might want to gently ask what's up. However, as Sam
> reviewed the patch and had some comments about it, it's unlikely that
> David will even look at it until you've addressed Sam's feedback.
> 
> The best way to get a patch into the kernel is to follow the rules in
> Documentation/SubmittingPatches which you have, respond quickly to any
> feedback, be patient, be prepared to explain why you're doing what
> you're doing, and be prepared to change and resubmit your patch many
> times until everyone is happy with it.
> 
> I'm sure that this is all written down in a document somewhere, but I
> can't recall what or where it is.
> 
> Thanks,
> 


-- 
Chen Gang

Asianux Corporation

Re: [PATCH] arch/sparc: additional len check in loop for prom_getbootargs

[Chen Gang]

Hello David Miller, Julian Calaby:


1) I am sorry for original mails to you:

   A) at 2012-11-09 00:33 Sam Ravnborg has already replied to me (need
additional discussion)
   B) but I did not see it, it is my fault.
   C) at 2012-11-09 11:15 I sent mail to want to get reply, and you
replied immediately.

   (the Time Location is at Beijing China)


2) sorry again, for bothering you,  and also thank Sam Ravnborg's reply.


3) next time, I shall notice, and check my own mail carefully.



Chen Gang

Re: [PATCH] arch/sparc: additional len check in loop for prom_getbootargs

[Chen Gang]

于 2012年11月09日 00:33, Sam Ravnborg 写道:
> Hi Cheng.
> On Thu, Nov 08, 2012 at 11:41:39AM +0800, Chen Gang wrote:
>>
>>   when cp >= barg_buf + BARG_LEN-2, it only break internel loop (while)
>>   but outside loop (for) still has effect, and "*cp++ = ' '" repeating
>>   so need additional checking for it.
>>
>>
>> Signed-off-by: Chen Gang <gang.chen@asianux.com>
> 
> I wonder how you found this bug?!?!

  I only find it through "code review".

    A) I grep all "strcpy" in kernel wide (about 2943 lines)

    B) I check them one by one.
         i)   when I check strcpy, also reading all relative source code.
         ii)  my goal is finding issues.
         iii) if I think it is valuable to continue reading, I will do.

    C) when find arch/sparc, I find this bug (although it is not relative with strcpy).


    It seems just starting (I have only finished checking 10 lines strcpy of 2943 lines),


> Anyway please consider this alternative fix:
> 
> diff --git a/arch/sparc/prom/bootstr_32.c b/arch/sparc/prom/bootstr_32.c
> index f5ec32e..4ce602f 100644
> --- a/arch/sparc/prom/bootstr_32.c
> +++ b/arch/sparc/prom/bootstr_32.c
> @@ -31,14 +31,10 @@ prom_getbootargs(void)
>  			arg = (*(romvec->pv_v0bootargs))->argv[iter];
>  			if (arg = NULL)
>  				break;
> -			while(*arg != 0) {
> -				/* Leave place for space and null. */
> -				if(cp >= barg_buf + BARG_LEN-2){
> -					/* We might issue a warning here. */
> -					break;
> -				}
> +			while (*arg != 0 && cp < (barg_buf + BARG_LEN - 2))
>  				*cp++ = *arg++;
> -			}
> +
> +			/* Append trailing space + null */
>  			*cp++ = ' ';
>  		}
>  		*cp = 0;
> 
> 
> Adding the conditional inside the while loop makes
> the logic simpler. And the patch actually deletes more lines than it adds.
> And please take care to follow coding style too. In particular spaces around operators.
> 
> The old code does not follow coding style - but this is no excuse.
> 

  I agree with the contents above.



> Note - the above is not even build tested!
> 

  A) I think, it will be better to give a test (although it seems obviously)
  B) but sorry for I have no relative environments now.
     i)   if testing is necessary;
     ii)    also if you have no environments, either.
     iii)       I should try. (please tell me)


> If you use the above code-snippet you can add my:
> Acked-by: Sam Ravnborg <sam@ravnborg.org>
> 

  I think it is necessary, you can do it, automatically.
  if truly need I do, please tell me.



-- 
Chen Gang

Asianux Corporation

Re: [PATCH] arch/sparc: additional len check in loop for prom_getbootargs

[Chen Gang]

于 2012年11月13日 11:25, Chen Gang 写道:
> 于 2012年11月09日 00:33, Sam Ravnborg 写道:
>> Anyway please consider this alternative fix:
>>
>> diff --git a/arch/sparc/prom/bootstr_32.c b/arch/sparc/prom/bootstr_32.c
>> index f5ec32e..4ce602f 100644
>> --- a/arch/sparc/prom/bootstr_32.c
>> +++ b/arch/sparc/prom/bootstr_32.c
>> @@ -31,14 +31,10 @@ prom_getbootargs(void)
>>  			arg = (*(romvec->pv_v0bootargs))->argv[iter];
>>  			if (arg = NULL)
>>  				break;
>> -			while(*arg != 0) {
>> -				/* Leave place for space and null. */
>> -				if(cp >= barg_buf + BARG_LEN-2){
>> -					/* We might issue a warning here. */
>> -					break;
>> -				}
>> +			while (*arg != 0 && cp < (barg_buf + BARG_LEN - 2))
>>  				*cp++ = *arg++;
>> -			}
>> +
>> +			/* Append trailing space + null */
>>  			*cp++ = ' ';
>>  		}
>>  		*cp = 0;
>>
>>
>> Adding the conditional inside the while loop makes
>> the logic simpler. And the patch actually deletes more lines than it adds.
>> And please take care to follow coding style too. In particular spaces around operators.
>>
>> The old code does not follow coding style - but this is no excuse.
>>
> 
>   I agree with the contents above.
> 

1) although I agree with what you said.

2) but for the patch code:
   after "*cp++ = ' ';", also need "break checking" for "for (; ;)" outside of "while()"

3) please check, and still it will be better to notice coding style, too.

-- 
Chen Gang

Asianux Corporation

Re: [PATCH] arch/sparc: additional len check in loop for prom_getbootargs

[David Miller]

From: Chen Gang <gang.chen@asianux.com>
Date: Tue, 13 Nov 2012 12:46:17 +0800

> 1) although I agree with what you said.
> 
> 2) but for the patch code:
>    after "*cp++ = ' ';", also need "break checking" for "for (; ;)" outside of "while()"
> 
> 3) please check, and still it will be better to notice coding style, too.

Why is every single one of your replies a set of 3 bullet points?  You
aren't giving a PowerPoint presentation.

Just talk and discuss things normally, using sentences and paragraphs
for your structure.

Everything you say and do looks very unnatural, awkward, and forced.

Re: [PATCH] arch/sparc: additional len check in loop for prom_getbootargs

[Chen Gang]

于 2012年11月13日 13:06, David Miller 写道:
> From: Chen Gang <gang.chen@asianux.com>
> Date: Tue, 13 Nov 2012 12:46:17 +0800
> 
>> 1) although I agree with what you said.
>>
>> 2) but for the patch code:
>>    after "*cp++ = ' ';", also need "break checking" for "for (; ;)" outside of "while()"
>>
>> 3) please check, and still it will be better to notice coding style, too.
> 
> Why is every single one of your replies a set of 3 bullet points?  You
> aren't giving a PowerPoint presentation.
> 
> Just talk and discuss things normally, using sentences and paragraphs
> for your structure.
> 
> Everything you say and do looks very unnatural, awkward, and forced.
> 
> 

 Excuse me,, my English is not quite well, just as you said (I need improving)


 after "*cp++ = ' ';" at line 42, also need "break checking" for "for (; ;)" outside of "while()"

 the reply from sam not check it, (the patch what I sent check it).

 this is the original source code of function prom_getbootargs in arch/sparc/prom/bootstr_32.c.
------------------------------------------------------------------------------------------------
 27         case PROM_V0:
 28                 cp = barg_buf;
 29                 /* Start from 1 and go over fd(0,0,0)kernel */
 30                 for(iter = 1; iter < 8; iter++) {
 31                         arg = (*(romvec->pv_v0bootargs))->argv[iter];
 32                         if (arg = NULL)
 33                                 break;
 34                         while(*arg != 0) {
 35                                 /* Leave place for space and null. */
 36                                 if(cp >= barg_buf + BARG_LEN-2){
 37                                         /* We might issue a warning here. */
 38                                         break;
 39                                 }
 40                                 *cp++ = *arg++;
 41                         }
 42                         *cp++ = ' ';
 43                 }
 44                 *cp = 0;
 45                 break;
---------------------------------------------------------------------------------------------------


this is the original reply from sam:
--------------------------------------------------------------------------------------------------
Hi Cheng.
On Thu, Nov 08, 2012 at 11:41:39AM +0800, Chen Gang wrote:
>
>   when cp >= barg_buf + BARG_LEN-2, it only break internel loop (while)
>   but outside loop (for) still has effect, and "*cp++ = ' '" repeating
>   so need additional checking for it.
>
>
> Signed-off-by: Chen Gang <gang.chen@asianux.com>

I wonder how you found this bug?!?!
Anyway please consider this alternative fix:

diff --git a/arch/sparc/prom/bootstr_32.c b/arch/sparc/prom/bootstr_32.c
index f5ec32e..4ce602f 100644
--- a/arch/sparc/prom/bootstr_32.c
+++ b/arch/sparc/prom/bootstr_32.c
@@ -31,14 +31,10 @@ prom_getbootargs(void)
 			arg = (*(romvec->pv_v0bootargs))->argv[iter];
 			if (arg = NULL)
 				break;
-			while(*arg != 0) {
-				/* Leave place for space and null. */
-				if(cp >= barg_buf + BARG_LEN-2){
-					/* We might issue a warning here. */
-					break;
-				}
+			while (*arg != 0 && cp < (barg_buf + BARG_LEN - 2))
 				*cp++ = *arg++;
-			}
+
+			/* Append trailing space + null */
 			*cp++ = ' ';
 		}
 		*cp = 0;


Adding the conditional inside the while loop makes
the logic simpler. And the patch actually deletes more lines than it adds.
And please take care to follow coding style too. In particular spaces around operators.

The old code does not follow coding style - but this is no excuse.

Note - the above is not even build tested!

If you use the above code-snippet you can add my:
Acked-by: Sam Ravnborg <sam@ravnborg.org>

	Sam
------------------------------------------------------------------------------------------




-- 
Chen Gang

Asianux Corporation