Re: [PATCH v3 5/6] mm/memory_hotplug: Move and simplify walk_memory_blocks()

[David Hildenbrand]

On 20.06.19 20:31, David Hildenbrand wrote:
> Let's move walk_memory_blocks() to the place where memory block logic
> resides and simplify it. While at it, add a type for the callback function.
> 
> Cc: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx>
> Cc: "Rafael J. Wysocki" <rafael@xxxxxxxxxx>
> Cc: David Hildenbrand <david@xxxxxxxxxx>
> Cc: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
> Cc: Stephen Rothwell <sfr@xxxxxxxxxxxxxxxx>
> Cc: Pavel Tatashin <pasha.tatashin@xxxxxxxxxx>
> Cc: Andrew Banman <andrew.banman@xxxxxxx>
> Cc: "mike.travis@xxxxxxx" <mike.travis@xxxxxxx>
> Cc: Oscar Salvador <osalvador@xxxxxxxx>
> Cc: Michal Hocko <mhocko@xxxxxxxx>
> Cc: Wei Yang <richard.weiyang@xxxxxxxxx>
> Cc: Arun KS <arunks@xxxxxxxxxxxxxx>
> Cc: Qian Cai <cai@xxxxxx>
> Signed-off-by: David Hildenbrand <david@xxxxxxxxxx>
> ---
>  drivers/base/memory.c          | 42 ++++++++++++++++++++++++++
>  include/linux/memory.h         |  3 ++
>  include/linux/memory_hotplug.h |  2 --
>  mm/memory_hotplug.c            | 55 ----------------------------------
>  4 files changed, 45 insertions(+), 57 deletions(-)
> 
> diff --git a/drivers/base/memory.c b/drivers/base/memory.c
> index c54e80fd25a8..0204384b4d1d 100644
> --- a/drivers/base/memory.c
> +++ b/drivers/base/memory.c
> @@ -44,6 +44,11 @@ static inline unsigned long pfn_to_block_id(unsigned long pfn)
>  	return base_memory_block_id(pfn_to_section_nr(pfn));
>  }
>  
> +static inline unsigned long phys_to_block_id(unsigned long phys)
> +{
> +	return pfn_to_block_id(PFN_DOWN(phys));
> +}
> +
>  static int memory_subsys_online(struct device *dev);
>  static int memory_subsys_offline(struct device *dev);
>  
> @@ -851,3 +856,40 @@ int __init memory_dev_init(void)
>  		printk(KERN_ERR "%s() failed: %d\n", __func__, ret);
>  	return ret;
>  }
> +
> +/**
> + * walk_memory_blocks - walk through all present memory blocks overlapped
> + *			by the range [start, start + size)
> + *
> + * @start: start address of the memory range
> + * @size: size of the memory range
> + * @arg: argument passed to func
> + * @func: callback for each memory section walked
> + *
> + * This function walks through all present memory blocks overlapped by the
> + * range [start, start + size), calling func on each memory block.
> + *
> + * In case func() returns an error, walking is aborted and the error is
> + * returned.
> + */
> +int walk_memory_blocks(unsigned long start, unsigned long size,
> +		       void *arg, walk_memory_blocks_func_t func)
> +{
> +	const unsigned long start_block_id = phys_to_block_id(start);
> +	const unsigned long end_block_id = phys_to_block_id(start + size - 1);
> +	struct memory_block *mem;
> +	unsigned long block_id;
> +	int ret = 0;

I *guess* the stall we are seeing is when size = 0.

(via ACPI, if info->length is 0)

if (!size)
	return 0;

... but that is just a wild guess. Will have a look after my vacation.

-- 

Thanks,

David / dhildenb