[PATCH v4 1/1] i2c: designware: use casting of u64 in clock multiplication to avoid overflow

[PATCH v4 1/1] i2c: designware: use casting of u64 in clock multiplication to avoid overflow

[Hanna Hawa]

From: Lareine Khawaly <lareine@amazon.com>

In functions i2c_dw_scl_lcnt() and i2c_dw_scl_hcnt() may have overflow
by depending on the values of the given parameters including the ic_clk.
For example in our use case where ic_clk is larger than one million,
multiplication of ic_clk * 4700 will result in 32 bit overflow.

Add cast of u64 to the calculation to avoid multiplication overflow, and
use the corresponding define for divide.

Fixes: 2373f6b9744d ("i2c-designware: split of i2c-designware.c into core and bus specific parts")
Signed-off-by: Lareine Khawaly <lareine@amazon.com>
Signed-off-by: Hanna Hawa <hhhawa@amazon.com>

---
Change Log v3->v4:
- update line length when possible
- fix change log location in the patch

Change Log v2->v3:
- Avoid changing the ic_clk parameter to u64, and do casting in the
  calculation itself instead.
- i2c_dw_clk_rate() returns unsigned long which is confusing because the
  function return the value of get_clk_rate_khz() which returns u32.
  This is not effect the overflow issue, pushed change in separated
  patch.
- use DIV_ROUND_CLOSEST_ULL instead of DIV_ROUND_CLOSEST

Change Log v1->v2:
- Update commit message and add fix tag.

 drivers/i2c/busses/i2c-designware-common.c | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/drivers/i2c/busses/i2c-designware-common.c b/drivers/i2c/busses/i2c-designware-common.c
index e0a46dfd1c15..2a669da08762 100644
--- a/drivers/i2c/busses/i2c-designware-common.c
+++ b/drivers/i2c/busses/i2c-designware-common.c
@@ -351,7 +351,8 @@ u32 i2c_dw_scl_hcnt(u32 ic_clk, u32 tSYMBOL, u32 tf, int cond, int offset)
 		 *
 		 * If your hardware is free from tHD;STA issue, try this one.
 		 */
-		return DIV_ROUND_CLOSEST(ic_clk * tSYMBOL, MICRO) - 8 + offset;
+		return DIV_ROUND_CLOSEST_ULL((u64)ic_clk * tSYMBOL, MICRO) - 8 +
+			offset;
 	else
 		/*
 		 * Conditional expression:
@@ -367,7 +368,8 @@ u32 i2c_dw_scl_hcnt(u32 ic_clk, u32 tSYMBOL, u32 tf, int cond, int offset)
 		 * The reason why we need to take into account "tf" here,
 		 * is the same as described in i2c_dw_scl_lcnt().
 		 */
-		return DIV_ROUND_CLOSEST(ic_clk * (tSYMBOL + tf), MICRO) - 3 + offset;
+		return DIV_ROUND_CLOSEST_ULL((u64)ic_clk * (tSYMBOL + tf),
+					     MICRO) - 3 + offset;
 }
 
 u32 i2c_dw_scl_lcnt(u32 ic_clk, u32 tLOW, u32 tf, int offset)
@@ -383,7 +385,8 @@ u32 i2c_dw_scl_lcnt(u32 ic_clk, u32 tLOW, u32 tf, int offset)
 	 * account the fall time of SCL signal (tf).  Default tf value
 	 * should be 0.3 us, for safety.
 	 */
-	return DIV_ROUND_CLOSEST(ic_clk * (tLOW + tf), MICRO) - 1 + offset;
+	return DIV_ROUND_CLOSEST_ULL((u64)ic_clk * (tLOW + tf), MICRO) - 1 +
+		offset;
 }
 
 int i2c_dw_set_sda_hold(struct dw_i2c_dev *dev)
-- 
2.38.1

Re: [PATCH v4 1/1] i2c: designware: use casting of u64 in clock multiplication to avoid overflow

[Andy Shevchenko]

On Tue, Dec 20, 2022 at 04:48:06PM +0000, Hanna Hawa wrote:
> From: Lareine Khawaly <lareine@amazon.com>
> 
> In functions i2c_dw_scl_lcnt() and i2c_dw_scl_hcnt() may have overflow
> by depending on the values of the given parameters including the ic_clk.
> For example in our use case where ic_clk is larger than one million,
> multiplication of ic_clk * 4700 will result in 32 bit overflow.
> 
> Add cast of u64 to the calculation to avoid multiplication overflow, and
> use the corresponding define for divide.

...

> -		return DIV_ROUND_CLOSEST(ic_clk * tSYMBOL, MICRO) - 8 + offset;
> +		return DIV_ROUND_CLOSEST_ULL((u64)ic_clk * tSYMBOL, MICRO) - 8 +
> +			offset;

Broken indentation.

...

> -		return DIV_ROUND_CLOSEST(ic_clk * (tSYMBOL + tf), MICRO) - 3 + offset;
> +		return DIV_ROUND_CLOSEST_ULL((u64)ic_clk * (tSYMBOL + tf),
> +					     MICRO) - 3 + offset;

I would still go with 'MICRO) -' part to be on the previous line despite being
over 80, this is logical split which increases readability.

> -	return DIV_ROUND_CLOSEST(ic_clk * (tLOW + tf), MICRO) - 1 + offset;
> +	return DIV_ROUND_CLOSEST_ULL((u64)ic_clk * (tLOW + tf), MICRO) - 1 +
> +		offset;

Broken indentation.

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v4 1/1] i2c: designware: use casting of u64 in clock multiplication to avoid overflow

[Andy Shevchenko]

On Tue, Dec 20, 2022 at 07:11:51PM +0200, Andy Shevchenko wrote:
> On Tue, Dec 20, 2022 at 04:48:06PM +0000, Hanna Hawa wrote:

...

> > -		return DIV_ROUND_CLOSEST(ic_clk * tSYMBOL, MICRO) - 8 + offset;
> > +		return DIV_ROUND_CLOSEST_ULL((u64)ic_clk * tSYMBOL, MICRO) - 8 +
> > +			offset;
> 
> Broken indentation.
> 
> ...
> 
> > -		return DIV_ROUND_CLOSEST(ic_clk * (tSYMBOL + tf), MICRO) - 3 + offset;
> > +		return DIV_ROUND_CLOSEST_ULL((u64)ic_clk * (tSYMBOL + tf),
> > +					     MICRO) - 3 + offset;
> 
> I would still go with 'MICRO) -' part to be on the previous line despite being
> over 80, this is logical split which increases readability.
> 
> > -	return DIV_ROUND_CLOSEST(ic_clk * (tLOW + tf), MICRO) - 1 + offset;
> > +	return DIV_ROUND_CLOSEST_ULL((u64)ic_clk * (tLOW + tf), MICRO) - 1 +
> > +		offset;
> 
> Broken indentation.

That said, can you just follow what I have said in a review of v3?

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v4 1/1] i2c: designware: use casting of u64 in clock multiplication to avoid overflow

[Hawa, Hanna]

On 12/20/2022 7:11 PM, Andy Shevchenko wrote:
> CAUTION: This email originated from outside of the organization. Do not click links or open attachments unless you can confirm the sender and know the content is safe.
> 
> 
> 
> On Tue, Dec 20, 2022 at 04:48:06PM +0000, Hanna Hawa wrote:
>> From: Lareine Khawaly <lareine@amazon.com>
>>
>> In functions i2c_dw_scl_lcnt() and i2c_dw_scl_hcnt() may have overflow
>> by depending on the values of the given parameters including the ic_clk.
>> For example in our use case where ic_clk is larger than one million,
>> multiplication of ic_clk * 4700 will result in 32 bit overflow.
>>
>> Add cast of u64 to the calculation to avoid multiplication overflow, and
>> use the corresponding define for divide.
> 
> ...
> 
>> -             return DIV_ROUND_CLOSEST(ic_clk * tSYMBOL, MICRO) - 8 + offset;
>> +             return DIV_ROUND_CLOSEST_ULL((u64)ic_clk * tSYMBOL, MICRO) - 8 +
>> +                     offset;
> 
> Broken indentation.
> 
> ...
> 
>> -             return DIV_ROUND_CLOSEST(ic_clk * (tSYMBOL + tf), MICRO) - 3 + offset;
>> +             return DIV_ROUND_CLOSEST_ULL((u64)ic_clk * (tSYMBOL + tf),
>> +                                          MICRO) - 3 + offset;
> 
> I would still go with 'MICRO) -' part to be on the previous line despite being
> over 80, this is logical split which increases readability.

Okay.. will move the 'MICRO) -' one line before
> 
>> -     return DIV_ROUND_CLOSEST(ic_clk * (tLOW + tf), MICRO) - 1 + offset;
>> +     return DIV_ROUND_CLOSEST_ULL((u64)ic_clk * (tLOW + tf), MICRO) - 1 +
>> +             offset;
> 
> Broken indentation.

Why it's broken indentation? I'm asking to know for the next time. The 
word 'offset' is not part of DIV_ROUND_CLOSEST_ULL parentheses. In wrong 
indentation the checkpatch shout about it, but it didn't happen with the 
above.

Does the below the correct indentation?

--- a/drivers/i2c/busses/i2c-designware-common.c
+++ b/drivers/i2c/busses/i2c-designware-common.c
@@ -351,7 +351,8 @@ u32 i2c_dw_scl_hcnt(u32 ic_clk, u32 tSYMBOL, u32 tf, 
int cond, int offset)
                  *
                  * If your hardware is free from tHD;STA issue, try 
this one.
                  */
-               return DIV_ROUND_CLOSEST(ic_clk * tSYMBOL, MICRO) - 8 + 
offset;
+               return DIV_ROUND_CLOSEST_ULL((u64)ic_clk * tSYMBOL, MICRO) -
+                                            8 + offset;
         else
                 /*
                  * Conditional expression:
@@ -367,7 +368,8 @@ u32 i2c_dw_scl_hcnt(u32 ic_clk, u32 tSYMBOL, u32 tf, 
int cond, int offset)
                  * The reason why we need to take into account "tf" here,
                  * is the same as described in i2c_dw_scl_lcnt().
                  */
-               return DIV_ROUND_CLOSEST(ic_clk * (tSYMBOL + tf), MICRO) 
- 3 + offset;
+               return DIV_ROUND_CLOSEST_ULL((u64)ic_clk * (tSYMBOL + 
tf), MICRO) -
+                                            3 + offset;
  }

  u32 i2c_dw_scl_lcnt(u32 ic_clk, u32 tLOW, u32 tf, int offset)
@@ -383,7 +385,8 @@ u32 i2c_dw_scl_lcnt(u32 ic_clk, u32 tLOW, u32 tf, 
int offset)
          * account the fall time of SCL signal (tf).  Default tf value
          * should be 0.3 us, for safety.
          */
-       return DIV_ROUND_CLOSEST(ic_clk * (tLOW + tf), MICRO) - 1 + offset;
+       return DIV_ROUND_CLOSEST_ULL((u64)ic_clk * (tLOW + tf), MICRO) -
+                                    1 + offset;
  }


> 
> --
> With Best Regards,
> Andy Shevchenko
> 
> 

Re: [PATCH v4 1/1] i2c: designware: use casting of u64 in clock multiplication to avoid overflow

[Andy Shevchenko]

On Tue, Dec 20, 2022 at 07:43:06PM +0200, Hawa, Hanna wrote:
> On 12/20/2022 7:11 PM, Andy Shevchenko wrote:
> > On Tue, Dec 20, 2022 at 04:48:06PM +0000, Hanna Hawa wrote:

...

> > > -             return DIV_ROUND_CLOSEST(ic_clk * tSYMBOL, MICRO) - 8 + offset;
> > > +             return DIV_ROUND_CLOSEST_ULL((u64)ic_clk * tSYMBOL, MICRO) - 8 +
> > > +                     offset;
> > 
> > Broken indentation.

...

> > > -             return DIV_ROUND_CLOSEST(ic_clk * (tSYMBOL + tf), MICRO) - 3 + offset;
> > > +             return DIV_ROUND_CLOSEST_ULL((u64)ic_clk * (tSYMBOL + tf),
> > > +                                          MICRO) - 3 + offset;
> > 
> > I would still go with 'MICRO) -' part to be on the previous line despite being
> > over 80, this is logical split which increases readability.
> 
> Okay.. will move the 'MICRO) -' one line before
> > 
> > > -     return DIV_ROUND_CLOSEST(ic_clk * (tLOW + tf), MICRO) - 1 + offset;
> > > +     return DIV_ROUND_CLOSEST_ULL((u64)ic_clk * (tLOW + tf), MICRO) - 1 +
> > > +             offset;
> > 
> > Broken indentation.
> 
> Why it's broken indentation? I'm asking to know for the next time. The word
> 'offset' is not part of DIV_ROUND_CLOSEST_ULL parentheses. In wrong
> indentation the checkpatch shout about it, but it didn't happen with the
> above.

The continuation line of the expression should go under the opening
parentheses, but you are right, the part outside DIV_ should be under
D and not as you suggested below.

But the problem is that you made illogical split while I suggested to leave
DIV_...() on one line and the rest on the other.

> Does the below the correct indentation?

No.

		return DIV_ROUND_CLOSEST_ULL((u64)ic_clk * tSYMBOL, MICRO) -
		       8 + offset;

		return DIV_ROUND_CLOSEST_ULL((u64)ic_clk * (tSYMBOL + tf), MICRO) -
		       3 + offset;

	return DIV_ROUND_CLOSEST_ULL((u64)ic_clk * (tLOW + tf), MICRO) -
	       1 + offset;

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v4 1/1] i2c: designware: use casting of u64 in clock multiplication to avoid overflow

[Andy Shevchenko]

On Tue, Dec 20, 2022 at 09:23:53PM +0200, Andy Shevchenko wrote:
> On Tue, Dec 20, 2022 at 07:43:06PM +0200, Hawa, Hanna wrote:

...

> 		return DIV_ROUND_CLOSEST_ULL((u64)ic_clk * tSYMBOL, MICRO) -
> 		       8 + offset;
> 
> 		return DIV_ROUND_CLOSEST_ULL((u64)ic_clk * (tSYMBOL + tf), MICRO) -
> 		       3 + offset;
> 
> 	return DIV_ROUND_CLOSEST_ULL((u64)ic_clk * (tLOW + tf), MICRO) -
> 	       1 + offset;

Thinking more on this, I would probably replace the order of arguments to make
it ' + offset - N' in each case. Since plus will be on the previous line and
become first it will be easier to parse the arithmetical expression.

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v4 1/1] i2c: designware: use casting of u64 in clock multiplication to avoid overflow

[Hawa, Hanna]

On 12/20/2022 9:27 PM, Andy Shevchenko wrote:
> Thinking more on this, I would probably replace the order of arguments to make
> it ' + offset - N' in each case. Since plus will be on the previous line and
> become first it will be easier to parse the arithmetical expression.

Do you want this change in the same patch? i don't think it's related 
here.. I can push separated change

Thanks,
Hanna

Re: [PATCH v4 1/1] i2c: designware: use casting of u64 in clock multiplication to avoid overflow

[Andy Shevchenko]

On Tue, Dec 20, 2022 at 09:43:25PM +0200, Hawa, Hanna wrote:
> On 12/20/2022 9:27 PM, Andy Shevchenko wrote:
> > Thinking more on this, I would probably replace the order of arguments to make
> > it ' + offset - N' in each case. Since plus will be on the previous line and
> > become first it will be easier to parse the arithmetical expression.
> 
> Do you want this change in the same patch? i don't think it's related here..
> I can push separated change

Up to you. If you think it's not suitable, then don't change.

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v4 1/1] i2c: designware: use casting of u64 in clock multiplication to avoid overflow

[Hawa, Hanna]

On 12/21/2022 6:59 PM, Andy Shevchenko wrote:
>> Do you want this change in the same patch? i don't think it's related here..
>> I can push separated change
> Up to you. If you think it's not suitable, then don't change.

Will push as separated patch.

Thanks,
Hanna