[PATCH] md/raid5: Fix divide type in raid5_make_request()

[PATCH] md/raid5: Fix divide type in raid5_make_request()

[Logan Gunthorpe]

0day reports a build failure on the hexagon architecture:

  ld.lld: error: undefined symbol: __hexagon_udivdi3
     referenced by raid5.c
        md/raid5.o:(raid5_make_request) in archive drivers/built-in.a
     referenced by raid5.c
        md/raid5.o:(raid5_make_request) in archive drivers/built-in.a
     did you mean: __hexagon_udivsi3
        defined in: arch/hexagon/built-in.a(lib/udivsi3.o)

This is caused by using DIV_ROUND_UP on a sector_t type.

The actual value is known to be less than 256 so a wide 64bit divide
here is not desirable. Thus cast the argument to an int to ensure it
uses a 32bit divide.

Fixes: 681fb14a7100 ("md/raid5: Pivot raid5_make_request()")
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Logan Gunthorpe <logang@deltatee.com>
---
 drivers/md/raid5.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c
index 5723a497108a..9d25696b793d 100644
--- a/drivers/md/raid5.c
+++ b/drivers/md/raid5.c
@@ -6091,7 +6091,7 @@ static bool raid5_make_request(struct mddev *mddev, struct bio * bi)
 	bi->bi_next = NULL;
 
 	bitmap_set(ctx.sectors_to_do, 0,
-		   DIV_ROUND_UP(ctx.last_sector - logical_sector,
+		   DIV_ROUND_UP((int)(ctx.last_sector - logical_sector),
 				RAID5_STRIPE_SECTORS(conf)));
 
 	pr_debug("raid456: %s, logical %llu to %llu\n", __func__,

base-commit: 57c19f921f8081c1a9444dc7f3f6b3ea43fe612e
-- 
2.30.2

Re: [PATCH] md/raid5: Fix divide type in raid5_make_request()

[Song Liu]

On Thu, Jun 23, 2022 at 9:56 AM Logan Gunthorpe <logang@deltatee.com> wrote:
>
> 0day reports a build failure on the hexagon architecture:
>
>   ld.lld: error: undefined symbol: __hexagon_udivdi3
>      referenced by raid5.c
>         md/raid5.o:(raid5_make_request) in archive drivers/built-in.a
>      referenced by raid5.c
>         md/raid5.o:(raid5_make_request) in archive drivers/built-in.a
>      did you mean: __hexagon_udivsi3
>         defined in: arch/hexagon/built-in.a(lib/udivsi3.o)
>
> This is caused by using DIV_ROUND_UP on a sector_t type.
>
> The actual value is known to be less than 256 so a wide 64bit divide
> here is not desirable. Thus cast the argument to an int to ensure it
> uses a 32bit divide.

I force pushed an update that uses DIV_ROUND_UP_SECTOR_T instead.
I guess that should also work?

Thanks,
Song

>
> Fixes: 681fb14a7100 ("md/raid5: Pivot raid5_make_request()")
> Reported-by: kernel test robot <lkp@intel.com>
> Signed-off-by: Logan Gunthorpe <logang@deltatee.com>
> ---
>  drivers/md/raid5.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c
> index 5723a497108a..9d25696b793d 100644
> --- a/drivers/md/raid5.c
> +++ b/drivers/md/raid5.c
> @@ -6091,7 +6091,7 @@ static bool raid5_make_request(struct mddev *mddev, struct bio * bi)
>         bi->bi_next = NULL;
>
>         bitmap_set(ctx.sectors_to_do, 0,
> -                  DIV_ROUND_UP(ctx.last_sector - logical_sector,
> +                  DIV_ROUND_UP((int)(ctx.last_sector - logical_sector),
>                                 RAID5_STRIPE_SECTORS(conf)));
>
>         pr_debug("raid456: %s, logical %llu to %llu\n", __func__,
>
> base-commit: 57c19f921f8081c1a9444dc7f3f6b3ea43fe612e
> --
> 2.30.2
>

Re: [PATCH] md/raid5: Fix divide type in raid5_make_request()

[Logan Gunthorpe]

On 2022-06-23 11:15, Song Liu wrote:
> On Thu, Jun 23, 2022 at 9:56 AM Logan Gunthorpe <logang@deltatee.com> wrote:
>>
>> 0day reports a build failure on the hexagon architecture:
>>
>>   ld.lld: error: undefined symbol: __hexagon_udivdi3
>>      referenced by raid5.c
>>         md/raid5.o:(raid5_make_request) in archive drivers/built-in.a
>>      referenced by raid5.c
>>         md/raid5.o:(raid5_make_request) in archive drivers/built-in.a
>>      did you mean: __hexagon_udivsi3
>>         defined in: arch/hexagon/built-in.a(lib/udivsi3.o)
>>
>> This is caused by using DIV_ROUND_UP on a sector_t type.
>>
>> The actual value is known to be less than 256 so a wide 64bit divide
>> here is not desirable. Thus cast the argument to an int to ensure it
>> uses a 32bit divide.
> 
> I force pushed an update that uses DIV_ROUND_UP_SECTOR_T instead.
> I guess that should also work?

Ah, yes, that'll also fix the bug. I chose to cast to minimize the
performance impact of an unnecessarily wide divide. But I don't have any
strong basis to suggest the performance will be worse with the wider
division. So fix it whichever way you think is best.

Logan