On 2022/05/18 4:43, Sergey Shtylyov wrote:
> Hello!
>
> On 5/16/22 2:29 PM, Damien Le Moal wrote:
>
>>> The code multiplying the # of cylinders/heads/sectors in ata_id_n_sectors()
>>> to get a disk capacity implicitly uses the *int* type for that calculation
>>> and casting the result to 'u64' before returning ensues a sign extension.
>>> Explicitly casting the 'u16' typed multipliers to 'u32' results in avoiding
>>> a sign extension instruction and so in a more compact code...
>>>
>>> Found by Linux Verification Center (linuxtesting.org) with the SVACE static
>>> analysis tool.
>>>
>>> Signed-off-by: Sergey Shtylyov <s.shtylyov@xxxxxx>
>>>
>>> ---
>>> This patch is against the 'for-next' branch of Damien's 'libata.git' repo.
>>>
>>> drivers/ata/libata-core.c | 10 ++++++----
>>> 1 file changed, 6 insertions(+), 4 deletions(-)
>>>
>>> Index: libata/drivers/ata/libata-core.c
>>> ===================================================================
>>> --- libata.orig/drivers/ata/libata-core.c
>>> +++ libata/drivers/ata/libata-core.c
>>> @@ -1107,11 +1107,13 @@ static u64 ata_id_n_sectors(const u16 *i
>>> return ata_id_u32(id, ATA_ID_LBA_CAPACITY);
>>> } else {
>>> if (ata_id_current_chs_valid(id))
>>> - return id[ATA_ID_CUR_CYLS] * id[ATA_ID_CUR_HEADS] *
>>> - id[ATA_ID_CUR_SECTORS];
>>> + return (u32)id[ATA_ID_CUR_CYLS] *
>>> + (u32)id[ATA_ID_CUR_HEADS] *
>>> + (u32)id[ATA_ID_CUR_SECTORS];
>>> else
>>
>> While at it, you can drop this useless "else". The 2 else above this one are
>> actually also useless...
>
> OK. But I think it's all a matter of a separate patch. I don't want to touch
> the LBA branches in this same patch...
OK.
>
>>> - return id[ATA_ID_CYLS] * id[ATA_ID_HEADS] *
>>> - id[ATA_ID_SECTORS];
>>> + return (u32)id[ATA_ID_CYLS] *
>>> + (u32)id[ATA_ID_HEADS] *
>>> + (u32)id[ATA_ID_SECTORS];
>>
>> Given that the function returns an u64, I would cast everything to u64. That
>
> I don't think this is a good idea. Looking at the produced x86 32-bit code,
> gcc produces an extra (3rd) multiplication instruction for no value.
>
>> will avoid overflows too, which was possible before,
>
> No, it wasn't possible. Any possible CHS capacity always fits into 32 bits --
> max # of sectors per track is 255, max # of heads is only 16.
> What actually seems to make sense is changing the order of multiplications
> to first multiply # of sectors by # of heads and than multiply that by # of
> cylinders...
OK.
>
>> eventhough no problems seem
>> to have been reported...
>
> Because there's not problem. :-)
> The current CHS capacity is stored in the words 57-58 (so 32-bit) and we
> could read it from there instead of the multiplications... BUT I do remember
> the disks (IIRC Fujitsu... but I'm not sure now -- that was back in 90s!)
> that had totally wrong value in these words... so the code we have now is
> a good thing! :-)
>
>> Who uses CHS these days :)
>
> Indeed, the CHS days are long gone... :-)
>
> [...]
>
> MBR, Sergey
--
Damien Le Moal
Western Digital Research
