lkml.org 
[lkml]   [2021]   [Feb]   [8]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectRe: [PATCH] mm/hugetlb: use helper huge_page_size() to get hugepage size
From
Date
On 2021/2/9 9:36, Mike Kravetz wrote:
> On 2/8/21 5:24 PM, Miaohe Lin wrote:
>> Hi:
>> On 2021/2/9 8:45, Mike Kravetz wrote:
>>> On 2/8/21 12:24 AM, Miaohe Lin wrote:
>>>> We can use helper huge_page_size() to get the hugepage size directly to
>>>> simplify the code slightly.
>>>>
>>>> Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
>>>> ---
>>>> mm/hugetlb.c | 14 ++++++--------
>>>> 1 file changed, 6 insertions(+), 8 deletions(-)
>>>>
>>>> diff --git a/mm/hugetlb.c b/mm/hugetlb.c
>>>> index 18628f8dbfb0..6cdb59d8f663 100644
>>>> --- a/mm/hugetlb.c
>>>> +++ b/mm/hugetlb.c
>>>> @@ -3199,7 +3199,7 @@ void __init hugetlb_add_hstate(unsigned int order)
>>>> BUG_ON(order == 0);
>>>> h = &hstates[hugetlb_max_hstate++];
>>>> h->order = order;
>>>> - h->mask = ~((1ULL << (order + PAGE_SHIFT)) - 1);
>>>> + h->mask = ~(huge_page_size(h) - 1);
>>>> for (i = 0; i < MAX_NUMNODES; ++i)
>>>> INIT_LIST_HEAD(&h->hugepage_freelists[i]);
>>>> INIT_LIST_HEAD(&h->hugepage_activelist);
>>>> @@ -3474,7 +3474,7 @@ void hugetlb_report_meminfo(struct seq_file *m)
>>>> for_each_hstate(h) {
>>>> unsigned long count = h->nr_huge_pages;
>>>>
>>>> - total += (PAGE_SIZE << huge_page_order(h)) * count;
>>>> + total += huge_page_size(h) * count;
>>>>
>>>> if (h == &default_hstate)
>>>> seq_printf(m,
>>>> @@ -3487,10 +3487,10 @@ void hugetlb_report_meminfo(struct seq_file *m)
>>>> h->free_huge_pages,
>>>> h->resv_huge_pages,
>>>> h->surplus_huge_pages,
>>>> - (PAGE_SIZE << huge_page_order(h)) / 1024);
>>>> + huge_page_size(h) / SZ_1K);
>>>> }
>>>>
>>>> - seq_printf(m, "Hugetlb: %8lu kB\n", total / 1024);
>>>> + seq_printf(m, "Hugetlb: %8lu kB\n", total / SZ_1K);
>>>> }
>>>>
>>>> int hugetlb_report_node_meminfo(char *buf, int len, int nid)
>>>> @@ -3524,7 +3524,7 @@ void hugetlb_show_meminfo(void)
>>>> h->nr_huge_pages_node[nid],
>>>> h->free_huge_pages_node[nid],
>>>> h->surplus_huge_pages_node[nid],
>>>> - 1UL << (huge_page_order(h) + PAGE_SHIFT - 10));
>>>> + huge_page_size(h) >> 10);
>>>
>>> Should we change this to
>>>
>>> huge_page_size(h) / SZ_1K);
>>>> as in hugetlb_report_meminfo above? Or, is that one where it takes an
>>> additional instruction to do the divide as opposed to the shift? I would> rather add the instruction and keep everything consistent.
>>>
>>
>> Yes, it takes an additional instruction to do the divide as opposed to the shift. So I did not
>> change this. But it seems keeping everything consistent in a function is more important. So should
>> I send a V2 to change this or Andrew would kindly handle this ?
>
> I would go ahead and put together a v2 and let Andrew decide how he wants
> to handle it. You can include,
>

Will do. Thanks a lot.

> Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
>
> Thanks
>

\
 
 \ /
  Last update: 2021-02-09 02:59    [W:0.041 / U:0.996 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site