2023-04-29 05:29:16

by Randy Dunlap

Subject: [RFC PATCH] workqueue: fix cast warnings on i386

Add casts to avoid int-to-pointer-cast warings on i386 or UML for i386:

../kernel/workqueue.c: In function ‘get_work_pwq’:
../kernel/workqueue.c:713:24: warning: cast to pointer from integer of different size [-Wint-to-pointer-cast]
713 | return (void *)(data & WORK_STRUCT_WQ_DATA_MASK);
| ^
../kernel/workqueue.c: In function ‘get_work_pool’:
../kernel/workqueue.c:741:25: warning: cast to pointer from integer of different size [-Wint-to-pointer-cast]
741 | return ((struct pool_workqueue *)
| ^
../kernel/workqueue.c: In function ‘get_work_pool_id’:
../kernel/workqueue.c:763:25: warning: cast to pointer from integer of different size [-Wint-to-pointer-cast]
763 | return ((struct pool_workqueue *)
| ^

Fixes: e120153ddf86 ("workqueue: fix how cpu number is stored in work->data")
Fixes: 112202d9098a ("workqueue: rename cpu_workqueue to pool_workqueue")
Signed-off-by: Randy Dunlap <[email protected]>
Cc: Tejun Heo <[email protected]>
Cc: Lai Jiangshan <[email protected]>
---
kernel/workqueue.c | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)

diff -- a/kernel/workqueue.c b/kernel/workqueue.c
--- a/kernel/workqueue.c
+++ b/kernel/workqueue.c
@@ -710,7 +710,7 @@ static struct pool_workqueue *get_work_p
unsigned long data = atomic_long_read(&work->data);

if (data & WORK_STRUCT_PWQ)
- return (void *)(data & WORK_STRUCT_WQ_DATA_MASK);
+ return (void *)(unsigned long)(data & WORK_STRUCT_WQ_DATA_MASK);
else
return NULL;
}
@@ -739,7 +739,7 @@ static struct worker_pool *get_work_pool

if (data & WORK_STRUCT_PWQ)
return ((struct pool_workqueue *)
- (data & WORK_STRUCT_WQ_DATA_MASK))->pool;
+ (unsigned long)(data & WORK_STRUCT_WQ_DATA_MASK))->pool;

pool_id = data >> WORK_OFFQ_POOL_SHIFT;
if (pool_id == WORK_OFFQ_POOL_NONE)
@@ -761,7 +761,7 @@ static int get_work_pool_id(struct work_

if (data & WORK_STRUCT_PWQ)
return ((struct pool_workqueue *)
- (data & WORK_STRUCT_WQ_DATA_MASK))->pool->id;
+ (unsigned long)(data & WORK_STRUCT_WQ_DATA_MASK))->pool->id;

return data >> WORK_OFFQ_POOL_SHIFT;
}

2023-05-01 04:48:26

by Randy Dunlap

Subject: Re: [RFC PATCH] workqueue: fix cast warnings on i386

Hi Lai,

On 4/30/23 21:37, Lai Jiangshan wrote:
> On Sat, Apr 29, 2023 at 12:45 PM Randy Dunlap <[email protected]> wrote:
>>
>> Add casts to avoid int-to-pointer-cast warings on i386 or UML for i386:
>>
>> ../kernel/workqueue.c: In function ‘get_work_pwq’:
>> ../kernel/workqueue.c:713:24: warning: cast to pointer from integer of different size [-Wint-to-pointer-cast]
>> 713 | return (void *)(data & WORK_STRUCT_WQ_DATA_MASK);
>> | ^
>> ../kernel/workqueue.c: In function ‘get_work_pool’:
>> ../kernel/workqueue.c:741:25: warning: cast to pointer from integer of different size [-Wint-to-pointer-cast]
>> 741 | return ((struct pool_workqueue *)
>> | ^
>> ../kernel/workqueue.c: In function ‘get_work_pool_id’:
>> ../kernel/workqueue.c:763:25: warning: cast to pointer from integer of different size [-Wint-to-pointer-cast]
>> 763 | return ((struct pool_workqueue *)
>> | ^
>>
>
> Hello, Randy
>
> Both the type of the "data" and WORK_STRUCT_WQ_DATA_MASK are
> "unsigned long", so I don't think "(data & WORK_STRUCT_WQ_DATA_MASK)"
> needs to be converted to "unsigned long".
>
> WORK_STRUCT_FLAG_MASK = (1UL << WORK_STRUCT_FLAG_BITS) - 1,
> WORK_STRUCT_WQ_DATA_MASK = ~WORK_STRUCT_FLAG_MASK,
>
> This simple fix might hide the real problem.

Thanks for replying.

Do you have any suggestions for how to eliminate these warnings on i386
and ARCH=um SUBARCH=i386?

> Thanks
> Lai.
>
>> Fixes: e120153ddf86 ("workqueue: fix how cpu number is stored in work->data")
>> Fixes: 112202d9098a ("workqueue: rename cpu_workqueue to pool_workqueue")
>> Signed-off-by: Randy Dunlap <[email protected]>
>> Cc: Tejun Heo <[email protected]>
>> Cc: Lai Jiangshan <[email protected]>
>> ---
>> kernel/workqueue.c | 6 +++---
>> 1 file changed, 3 insertions(+), 3 deletions(-)
>>
>> diff -- a/kernel/workqueue.c b/kernel/workqueue.c
>> --- a/kernel/workqueue.c
>> +++ b/kernel/workqueue.c
>> @@ -710,7 +710,7 @@ static struct pool_workqueue *get_work_p
>> unsigned long data = atomic_long_read(&work->data);
>>
>> if (data & WORK_STRUCT_PWQ)
>> - return (void *)(data & WORK_STRUCT_WQ_DATA_MASK);
>> + return (void *)(unsigned long)(data & WORK_STRUCT_WQ_DATA_MASK);
>> else
>> return NULL;
>> }
>> @@ -739,7 +739,7 @@ static struct worker_pool *get_work_pool
>>
>> if (data & WORK_STRUCT_PWQ)
>> return ((struct pool_workqueue *)
>> - (data & WORK_STRUCT_WQ_DATA_MASK))->pool;
>> + (unsigned long)(data & WORK_STRUCT_WQ_DATA_MASK))->pool;
>>
>> pool_id = data >> WORK_OFFQ_POOL_SHIFT;
>> if (pool_id == WORK_OFFQ_POOL_NONE)
>> @@ -761,7 +761,7 @@ static int get_work_pool_id(struct work_
>>
>> if (data & WORK_STRUCT_PWQ)
>> return ((struct pool_workqueue *)
>> - (data & WORK_STRUCT_WQ_DATA_MASK))->pool->id;
>> + (unsigned long)(data & WORK_STRUCT_WQ_DATA_MASK))->pool->id;
>>
>> return data >> WORK_OFFQ_POOL_SHIFT;
>> }

--
~Randy

2023-05-01 04:49:29

by Lai Jiangshan

Subject: Re: [RFC PATCH] workqueue: fix cast warnings on i386

On Sat, Apr 29, 2023 at 12:45 PM Randy Dunlap <[email protected]> wrote:
>
> Add casts to avoid int-to-pointer-cast warings on i386 or UML for i386:
>
> ../kernel/workqueue.c: In function ‘get_work_pwq’:
> ../kernel/workqueue.c:713:24: warning: cast to pointer from integer of different size [-Wint-to-pointer-cast]
> 713 | return (void *)(data & WORK_STRUCT_WQ_DATA_MASK);
> | ^
> ../kernel/workqueue.c: In function ‘get_work_pool’:
> ../kernel/workqueue.c:741:25: warning: cast to pointer from integer of different size [-Wint-to-pointer-cast]
> 741 | return ((struct pool_workqueue *)
> | ^
> ../kernel/workqueue.c: In function ‘get_work_pool_id’:
> ../kernel/workqueue.c:763:25: warning: cast to pointer from integer of different size [-Wint-to-pointer-cast]
> 763 | return ((struct pool_workqueue *)
> | ^
>

Hello, Randy

Both the type of the "data" and WORK_STRUCT_WQ_DATA_MASK are
"unsigned long", so I don't think "(data & WORK_STRUCT_WQ_DATA_MASK)"
needs to be converted to "unsigned long".

WORK_STRUCT_FLAG_MASK = (1UL << WORK_STRUCT_FLAG_BITS) - 1,
WORK_STRUCT_WQ_DATA_MASK = ~WORK_STRUCT_FLAG_MASK,

This simple fix might hide the real problem.

Thanks
Lai.

> Fixes: e120153ddf86 ("workqueue: fix how cpu number is stored in work->data")
> Fixes: 112202d9098a ("workqueue: rename cpu_workqueue to pool_workqueue")
> Signed-off-by: Randy Dunlap <[email protected]>
> Cc: Tejun Heo <[email protected]>
> Cc: Lai Jiangshan <[email protected]>
> ---
> kernel/workqueue.c | 6 +++---
> 1 file changed, 3 insertions(+), 3 deletions(-)
>
> diff -- a/kernel/workqueue.c b/kernel/workqueue.c
> --- a/kernel/workqueue.c
> +++ b/kernel/workqueue.c
> @@ -710,7 +710,7 @@ static struct pool_workqueue *get_work_p
> unsigned long data = atomic_long_read(&work->data);
>
> if (data & WORK_STRUCT_PWQ)
> - return (void *)(data & WORK_STRUCT_WQ_DATA_MASK);
> + return (void *)(unsigned long)(data & WORK_STRUCT_WQ_DATA_MASK);
> else
> return NULL;
> }
> @@ -739,7 +739,7 @@ static struct worker_pool *get_work_pool
>
> if (data & WORK_STRUCT_PWQ)
> return ((struct pool_workqueue *)
> - (data & WORK_STRUCT_WQ_DATA_MASK))->pool;
> + (unsigned long)(data & WORK_STRUCT_WQ_DATA_MASK))->pool;
>
> pool_id = data >> WORK_OFFQ_POOL_SHIFT;
> if (pool_id == WORK_OFFQ_POOL_NONE)
> @@ -761,7 +761,7 @@ static int get_work_pool_id(struct work_
>
> if (data & WORK_STRUCT_PWQ)
> return ((struct pool_workqueue *)
> - (data & WORK_STRUCT_WQ_DATA_MASK))->pool->id;
> + (unsigned long)(data & WORK_STRUCT_WQ_DATA_MASK))->pool->id;
>
> return data >> WORK_OFFQ_POOL_SHIFT;
> }

2023-05-08 13:53:33

by Arnd Bergmann

Subject: Re: [RFC PATCH] workqueue: fix cast warnings on i386

On Mon, May 1, 2023, at 06:42, Randy Dunlap wrote:
> On 4/30/23 21:37, Lai Jiangshan wrote:
>> Both the type of the "data" and WORK_STRUCT_WQ_DATA_MASK are
>> "unsigned long", so I don't think "(data & WORK_STRUCT_WQ_DATA_MASK)"
>> needs to be converted to "unsigned long".
>>
>> WORK_STRUCT_FLAG_MASK = (1UL << WORK_STRUCT_FLAG_BITS) - 1,
>> WORK_STRUCT_WQ_DATA_MASK = ~WORK_STRUCT_FLAG_MASK,
>>
>> This simple fix might hide the real problem.
>
> Thanks for replying.
>
> Do you have any suggestions for how to eliminate these warnings on i386
> and ARCH=um SUBARCH=i386?

I'm pretty sure the patch I posted earlier is the correct one, see
https://lore.kernel.org/all/[email protected]/

Sorry I never followed up on this after the initial post, I have
a backlog of patches that I sent out as build fixes but that for
some reason have not made it in.

Arnd

2023-05-08 19:23:06

by Randy Dunlap

Subject: Re: [RFC PATCH] workqueue: fix cast warnings on i386

On 5/8/23 06:41, Arnd Bergmann wrote:
> On Mon, May 1, 2023, at 06:42, Randy Dunlap wrote:
>> On 4/30/23 21:37, Lai Jiangshan wrote:
>>> Both the type of the "data" and WORK_STRUCT_WQ_DATA_MASK are
>>> "unsigned long", so I don't think "(data & WORK_STRUCT_WQ_DATA_MASK)"
>>> needs to be converted to "unsigned long".
>>>
>>> WORK_STRUCT_FLAG_MASK = (1UL << WORK_STRUCT_FLAG_BITS) - 1,
>>> WORK_STRUCT_WQ_DATA_MASK = ~WORK_STRUCT_FLAG_MASK,
>>>
>>> This simple fix might hide the real problem.
>>
>> Thanks for replying.
>>
>> Do you have any suggestions for how to eliminate these warnings on i386
>> and ARCH=um SUBARCH=i386?
>
> I'm pretty sure the patch I posted earlier is the correct one, see
> https://lore.kernel.org/all/[email protected]/
>

Thanks for the reply. Your patch looks good. I appreciate the
explanation.

> Sorry I never followed up on this after the initial post, I have
> a backlog of patches that I sent out as build fixes but that for
> some reason have not made it in.

I know how that goes. I have quite a few as well.

--
~Randy