Re: [Intel-wired-lan] [PATCH] i40e: fix MMIO write access to an invalid page in i40e_clear_hw

[Kyungwook Boo <bookyungwook@gmail.com>]

On 25. 3. 6. 16:59, Przemek Kitszel wrote:
> On 3/6/25 06:25, Kyungwook Boo wrote:
>> In i40e_clear_hw(), when the device sends a specific input(e.g., 0),
>> an integer underflow in the num_{pf,vf}_int variables can occur,
>> leading to MMIO write access to an invalid page.
>>
>> To fix this, we change the type of the unsigned integer variables
>> num_{pf,vf}_int to signed integers. Additionally, in the for-loop where the
>> integer underflow occurs, we also change the type of the loop variable i to
>> a signed integer.
>>
>> Signed-off-by: Kyungwook Boo <bookyungwook@gmail.com>
>> Signed-off-by: Loktionov, Aleksandr <aleksandr.loktionov@intel.com>
> 
> when Alex said "make sure I signed too" he meant:
> "make sure the variable @i is signed too", not the Sign-off ;)
> 
> (please wait 24h for the next submission, and also put "iwl-next" after
> the "PATCH" word)
> 
>> Signed-off-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
> 
> I didn't signed that either

Oh.. I totally misunderstood the comment.
I apologize for mistakenly adding the sign.

>> Link: https://lore.kernel.org/lkml/ffc91764-1142-4ba2-91b6-8c773f6f7095@gmail.com/T/
>> ---
>>   drivers/net/ethernet/intel/i40e/i40e_common.c | 10 +++++-----
>>   1 file changed, 5 insertions(+), 5 deletions(-)
>>
>> diff --git a/drivers/net/ethernet/intel/i40e/i40e_common.c b/drivers/net/ethernet/intel/i40e/i40e_common.c
>> index 370b4bddee44..9a73cb94dc5e 100644
>> --- a/drivers/net/ethernet/intel/i40e/i40e_common.c
>> +++ b/drivers/net/ethernet/intel/i40e/i40e_common.c
>> @@ -817,8 +817,8 @@ int i40e_pf_reset(struct i40e_hw *hw)
>>   void i40e_clear_hw(struct i40e_hw *hw)
>>   {
>>       u32 num_queues, base_queue;
>> -    u32 num_pf_int;
>> -    u32 num_vf_int;
>> +    s32 num_pf_int;
>> +    s32 num_vf_int;
>>       u32 num_vfs;
>>       u32 i, j;
> 
> It's fine to move the declaration of @i into the for loop, but
> you have to remove it here, otherwise it's shadowing, which we
> avoid.
> 
>>       u32 val;
>> @@ -848,18 +848,18 @@ void i40e_clear_hw(struct i40e_hw *hw)
>>       /* stop all the interrupts */
>>       wr32(hw, I40E_PFINT_ICR0_ENA, 0);
>>       val = 0x3 << I40E_PFINT_DYN_CTLN_ITR_INDX_SHIFT;
>> -    for (i = 0; i < num_pf_int - 2; i++)
>> +    for (s32 i = 0; i < num_pf_int - 2; i++)
>>           wr32(hw, I40E_PFINT_DYN_CTLN(i), val);
>>         /* Set the FIRSTQ_INDX field to 0x7FF in PFINT_LNKLSTx */
>>       val = eol << I40E_PFINT_LNKLST0_FIRSTQ_INDX_SHIFT;
>>       wr32(hw, I40E_PFINT_LNKLST0, val);
>> -    for (i = 0; i < num_pf_int - 2; i++)
>> +    for (s32 i = 0; i < num_pf_int - 2; i++)
>>           wr32(hw, I40E_PFINT_LNKLSTN(i), val);
>>       val = eol << I40E_VPINT_LNKLST0_FIRSTQ_INDX_SHIFT;
>>       for (i = 0; i < num_vfs; i++)
>>           wr32(hw, I40E_VPINT_LNKLST0(i), val);
>> -    for (i = 0; i < num_vf_int - 2; i++)
>> +    for (s32 i = 0; i < num_vf_int - 2; i++)
>>           wr32(hw, I40E_VPINT_LNKLSTN(i), val);
>>         /* warn the HW of the coming Tx disables */
> 

Thank you for reviewing the patch.
I will correct the patch and resubmit it.

Best,
Kyungwook Boo