[PATCH v3] mm: alloc_pages_bulk: support both simple and full-featured API

[Yunsheng Lin]

On 2025/4/20 19:21, Leon Romanovsky wrote:

...

>>
>> diff --git a/drivers/vfio/pci/mlx5/cmd.c b/drivers/vfio/pci/mlx5/cmd.c
>> index 11eda6b207f1..fb094527715f 100644
>> --- a/drivers/vfio/pci/mlx5/cmd.c
>> +++ b/drivers/vfio/pci/mlx5/cmd.c
>> @@ -446,8 +446,6 @@ static int mlx5vf_add_migration_pages(struct mlx5_vhca_data_buffer *buf,
>>  		if (ret)
>>  			goto err_append;
>>  		buf->allocated_length += filled * PAGE_SIZE;
>> -		/* clean input for another bulk allocation */
>> -		memset(page_list, 0, filled * sizeof(*page_list));
>>  		to_fill = min_t(unsigned int, to_alloc,
>>  				PAGE_SIZE / sizeof(*page_list));
> 
> If it is possible, let's drop this hunk to reduce merge conflicts.
> The whole mlx5vf_add_migration_pages() is planned to be rewritten.
> https://lore.kernel.org/linux-rdma/076a3991e663fe07c1a5395f5805c514b63e4d94.1744825142.git.leon@kernel.org/

It seems mlx5vf_add_migration_pages() is changed to use the pattern
of passing 'page_array + allocated' and 'nr_pages - allocated' in the
above patch, so I think it is ok to drop the above hunk.

Hi, Andrew
Do you want me to resend this patch without the above hunk or it is
possible that you can drop the above hunk when committing if there
is no other comment need fixing?

> 
> Thanks
> 
>