Re: [PATCH] ata: libata-scsi: fix sloppy result type of ata_ioc32()

[Damien Le Moal <damien.lemoal@opensource.wdc.com>]

On 6/22/22 05:37, Sergey Shtylyov wrote:
> On 6/21/22 10:14 PM, Sergey Shtylyov wrote:
> [...]
>>>>>> While ata_ioc32() returns *int*, its result gets assigned to and compared
>>>>>> with the *unsigned long* variable 'val' in ata_sas_scsi_ioctl(), its only
>>>>>> caller, which implies a problematic implicit cast -- fix that by returning
>>>>>> *unsigned long* instead.
>>>>>>
>>>>>> Found by Linux Verification Center (linuxtesting.org) with the SVACE static
>>>>>> analysis tool.
>>>>>>
>>>>>> Signed-off-by: Sergey Shtylyov <s.shtylyov@omp.ru>
>>>>>>
>>>>>> ---
>>>>>> This patch is against the 'for-next' branch of Damien's 'libata.git' repo.
>>>>>>
>>>>>>  drivers/ata/libata-scsi.c |    2 +-
>>>>>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>>>>>
>>>>>> Index: libata/drivers/ata/libata-scsi.c
>>>>>> ===================================================================
>>>>>> --- libata.orig/drivers/ata/libata-scsi.c
>>>>>> +++ libata/drivers/ata/libata-scsi.c
>>>>>> @@ -539,7 +539,7 @@ int ata_task_ioctl(struct scsi_device *s
>>>>>>  	return rc;
>>>>>>  }
>>>>>>  
>>>>>> -static int ata_ioc32(struct ata_port *ap)
>>>>>> +static unsigned long ata_ioc32(struct ata_port *ap)
>>>>>>  {
>>>>>>  	if (ap->flags & ATA_FLAG_PIO_DMA)
>>>>>>  		return 1;
>>>>
>>>>> Actually, this should be a bool I think and the ioctl code cleaned to use
>>>>
>>>>    By the ioctl code you mean ata_sas_scsi_ioctl()?
>>>
>>> yes.
>>>
>>>>> that type since the val argument of the ioctl is also used as a bool.
>>>>
>>>>    As for HDIO_SET_32BIT, that's prolly OK but what to do with HDIO_GET_32BIT
>>>> (it calls put_user() on *unsigned long*)?
>>>
>>> Something like this should work fine:
>>>
>>> diff --git a/drivers/ata/libata-scsi.c b/drivers/ata/libata-scsi.c
>>> index 86dbb1cdfabd..ec7f79cbb135 100644
>>> --- a/drivers/ata/libata-scsi.c
>>> +++ b/drivers/ata/libata-scsi.c
>> [...]
>>> @@ -571,16 +572,16 @@ int ata_sas_scsi_ioctl(struct ata_port *ap, struct
>>> scsi_device *scsidev,
>>>                 return put_user(val, (unsigned long __user *)arg);
>>>
>>>         case HDIO_SET_32BIT:
>>
>>    Hmm, I told you this *case* is prolly OK -- it was HDIO_GET_32BIT *case* that
>> I was concerned about... So you mean that HDIO_GET_32BIT handling should remain
>> intact?
>>
>>> -               val = (unsigned long) arg;
>>> +               pio32 = !!((unsigned long) arg);
> 
>    No, this one won't do -- it changes the behavior in case ATA_PFLAG_PIO32CHANGE
> isn't set... :-/
> 
>>>                 rc = 0;
>>>                 spin_lock_irqsave(ap->lock, flags);
>>>                 if (ap->pflags & ATA_PFLAG_PIO32CHANGE) {
>>> -                       if (val)
>>> +                       if (pio32)
>>>                                 ap->pflags |= ATA_PFLAG_PIO32;
>>>                         else
>>>                                 ap->pflags &= ~ATA_PFLAG_PIO32;
>>>                 } else {
>>> -                       if (val != ata_ioc32(ap))
>>> +                       if (pio32 != ata_ioc32(ap))
>>>                                 rc = -EINVAL;
>>>                 }
>>>                 spin_unlock_irqrestore(ap->lock, flags);
>>
>>    Not really sure this is worth it... *unsigned long* result type for
>> ata_ioc32() seems simpler.
> 
>    Actually, even just modifying ata_ioc32() to return 'bool' produces
> a seemingly correct code. Note that ata_ioc32() is inlined in any case...

If there are no issues with the bool type conversion, I would really
prefer that rather than the unsigned long route. The latter is really
about silencing a static analyzer rather than ideal code. Given the name
of the function, returning an unsigned long is really strange.

> 
> MBR, Sergey


-- 
Damien Le Moal
Western Digital Research