Re: [PATCH] mm: introduce kv[mz]alloc helpers

Re: [PATCH] mm: introduce kv[mz]alloc helpers

[Tetsuo Handa]

On 2017/01/04 23:20, Michal Hocko wrote:
> OK, so I've checked the open coded implementations and converted most of
> them. There are few which are either confused and need some special
> handling or need double checking.
> 
> diff --git a/drivers/md/bcache/util.h b/drivers/md/bcache/util.h
> index cf2cbc211d83..9dc0f0ff0321 100644
> --- a/drivers/md/bcache/util.h
> +++ b/drivers/md/bcache/util.h
> @@ -44,10 +44,7 @@ struct closure;
>  	(heap)->size = (_size);						\
>  	_bytes = (heap)->size * sizeof(*(heap)->data);			\
>  	(heap)->data = NULL;						\
> -	if (_bytes < KMALLOC_MAX_SIZE)					\
> -		(heap)->data = kmalloc(_bytes, (gfp));			\
> -	if ((!(heap)->data) && ((gfp) & GFP_KERNEL))			\
> -		(heap)->data = vmalloc(_bytes);				\
> +	(heap)->data = kvmalloc(_bytes, (gfp) & GFP_KERNEL);		\
>  	(heap)->data;							\
>  })
>  
> @@ -138,10 +135,7 @@ do {									\
>  	(fifo)->front = (fifo)->back = 0;				\
>  	(fifo)->data = NULL;						\
>  									\
> -	if (_bytes < KMALLOC_MAX_SIZE)					\
> -		(fifo)->data = kmalloc(_bytes, (gfp));			\
> -	if ((!(fifo)->data) && ((gfp) & GFP_KERNEL))			\
> -		(fifo)->data = vmalloc(_bytes);				\
> +	(fifo)->data = kvmalloc(_bytes, (gfp) & GFP_KERNEL);		\
>  	(fifo)->data;							\
>  })

These macros are doing strange checks.
((gfp) & GFP_KERNEL) means any bit in GFP_KERNEL is set.
((gfp) & GFP_KERNEL) == GFP_KERNEL might make sense. Actually,
all callers seems to be passing GFP_KERNEL to these macros.

Kent, how do you want to correct this? You want to apply
a patch that removes gfp argument before applying this patch?
Or, you want Michal to directly overwrite by this patch?

Michal, "(fifo)->data = NULL;" line will become redundant
and "(gfp) & GFP_KERNEL" will become "GFP_KERNEL".

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH] mm: introduce kv[mz]alloc helpers

[Michal Hocko]

On Thu 05-01-17 19:40:10, Tetsuo Handa wrote:
> On 2017/01/04 23:20, Michal Hocko wrote:
> > OK, so I've checked the open coded implementations and converted most of
> > them. There are few which are either confused and need some special
> > handling or need double checking.
> > 
> > diff --git a/drivers/md/bcache/util.h b/drivers/md/bcache/util.h
> > index cf2cbc211d83..9dc0f0ff0321 100644
> > --- a/drivers/md/bcache/util.h
> > +++ b/drivers/md/bcache/util.h
> > @@ -44,10 +44,7 @@ struct closure;
> >  	(heap)->size = (_size);						\
> >  	_bytes = (heap)->size * sizeof(*(heap)->data);			\
> >  	(heap)->data = NULL;						\
> > -	if (_bytes < KMALLOC_MAX_SIZE)					\
> > -		(heap)->data = kmalloc(_bytes, (gfp));			\
> > -	if ((!(heap)->data) && ((gfp) & GFP_KERNEL))			\
> > -		(heap)->data = vmalloc(_bytes);				\
> > +	(heap)->data = kvmalloc(_bytes, (gfp) & GFP_KERNEL);		\
> >  	(heap)->data;							\
> >  })
> >  
> > @@ -138,10 +135,7 @@ do {									\
> >  	(fifo)->front = (fifo)->back = 0;				\
> >  	(fifo)->data = NULL;						\
> >  									\
> > -	if (_bytes < KMALLOC_MAX_SIZE)					\
> > -		(fifo)->data = kmalloc(_bytes, (gfp));			\
> > -	if ((!(fifo)->data) && ((gfp) & GFP_KERNEL))			\
> > -		(fifo)->data = vmalloc(_bytes);				\
> > +	(fifo)->data = kvmalloc(_bytes, (gfp) & GFP_KERNEL);		\
> >  	(fifo)->data;							\
> >  })
> 
> These macros are doing strange checks.
> ((gfp) & GFP_KERNEL) means any bit in GFP_KERNEL is set.
> ((gfp) & GFP_KERNEL) == GFP_KERNEL might make sense. Actually,
> all callers seems to be passing GFP_KERNEL to these macros.

Yes the code is confused. I've seen worse when going through the drivers
code...

> Kent, how do you want to correct this? You want to apply
> a patch that removes gfp argument before applying this patch?
> Or, you want Michal to directly overwrite by this patch?

I would just get rid of it here as init_heap has just one caller with
GFP_KERNEL and __init_fifo has GFP_KERNEL users as well. But if it is
preferable to clean up this first then I can do that.
 
> Michal, "(fifo)->data = NULL;" line will become redundant
> and "(gfp) & GFP_KERNEL" will become "GFP_KERNEL".

true. will remove it.

-- 
Michal Hocko
SUSE Labs

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH] mm: introduce kv[mz]alloc helpers

[Kent Overstreet]

On Thu, Jan 05, 2017 at 07:40:10PM +0900, Tetsuo Handa wrote:
> On 2017/01/04 23:20, Michal Hocko wrote:
> > OK, so I've checked the open coded implementations and converted most of
> > them. There are few which are either confused and need some special
> > handling or need double checking.
> > 
> > diff --git a/drivers/md/bcache/util.h b/drivers/md/bcache/util.h
> > index cf2cbc211d83..9dc0f0ff0321 100644
> > --- a/drivers/md/bcache/util.h
> > +++ b/drivers/md/bcache/util.h
> > @@ -44,10 +44,7 @@ struct closure;
> >  	(heap)->size = (_size);						\
> >  	_bytes = (heap)->size * sizeof(*(heap)->data);			\
> >  	(heap)->data = NULL;						\
> > -	if (_bytes < KMALLOC_MAX_SIZE)					\
> > -		(heap)->data = kmalloc(_bytes, (gfp));			\
> > -	if ((!(heap)->data) && ((gfp) & GFP_KERNEL))			\
> > -		(heap)->data = vmalloc(_bytes);				\
> > +	(heap)->data = kvmalloc(_bytes, (gfp) & GFP_KERNEL);		\
> >  	(heap)->data;							\
> >  })
> >  
> > @@ -138,10 +135,7 @@ do {									\
> >  	(fifo)->front = (fifo)->back = 0;				\
> >  	(fifo)->data = NULL;						\
> >  									\
> > -	if (_bytes < KMALLOC_MAX_SIZE)					\
> > -		(fifo)->data = kmalloc(_bytes, (gfp));			\
> > -	if ((!(fifo)->data) && ((gfp) & GFP_KERNEL))			\
> > -		(fifo)->data = vmalloc(_bytes);				\
> > +	(fifo)->data = kvmalloc(_bytes, (gfp) & GFP_KERNEL);		\
> >  	(fifo)->data;							\
> >  })
> 
> These macros are doing strange checks.
> ((gfp) & GFP_KERNEL) means any bit in GFP_KERNEL is set.
> ((gfp) & GFP_KERNEL) == GFP_KERNEL might make sense. Actually,
> all callers seems to be passing GFP_KERNEL to these macros.
> 
> Kent, how do you want to correct this? You want to apply
> a patch that removes gfp argument before applying this patch?
> Or, you want Michal to directly overwrite by this patch?
> 
> Michal, "(fifo)->data = NULL;" line will become redundant
> and "(gfp) & GFP_KERNEL" will become "GFP_KERNEL".

Please just go ahead and replace all that crap with a single call to kvmalloc().

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>