lkml.org 
[lkml]   [2019]   [Mar]   [6]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectRe: [PATCH v5 4/4] hugetlb: allow to free gigantic pages regardless of the configuration
From
Date
On 3/6/19 2:16 PM, Dave Hansen wrote:
> On 3/6/19 11:00 AM, Alexandre Ghiti wrote:
>> +static int set_max_huge_pages(struct hstate *h, unsigned long count,
>> + nodemask_t *nodes_allowed)
>> {
>> unsigned long min_count, ret;
>>
>> - if (hstate_is_gigantic(h) && !gigantic_page_supported())
>> - return h->max_huge_pages;
>> + /*
>> + * Gigantic pages allocation depends on the capability for large page
>> + * range allocation. If the system cannot provide alloc_contig_range,
>> + * allow users to free gigantic pages.
>> + */
>> + if (hstate_is_gigantic(h) && !IS_ENABLED(CONFIG_CONTIG_ALLOC)) {
>> + spin_lock(&hugetlb_lock);
>> + if (count > persistent_huge_pages(h)) {
>> + spin_unlock(&hugetlb_lock);
>> + return -EINVAL;
>> + }
>> + goto decrease_pool;
>> + }
> We talked about it during the last round and I don't seen any mention of
> it here in comments or the changelog: Why is this a goto? Why don't we
> just let the code fall through to the "decrease_pool" label? Why is
> this new block needed at all? Can't we just remove the old check and
> let it be?

I'll get rid of the goto, I don't know how to justify it properly in a
comment,
maybe because it is not necessary.
This is not a new block, this means exactly the same as before (remember
gigantic_page_supported() actually meant CONTIG_ALLOC before this series),
except that now we allow a user to free boottime allocated gigantic pages.
And no we cannot just remove the check and let it be since it would modify
the current behaviour, which is to return an error when trying to allocate
gigantic pages whereas alloc_contig_range is not defined. I thought it was
clearly commented above, I can try to make it more explicit.

\
 
 \ /
  Last update: 2019-03-06 21:09    [W:0.076 / U:27.528 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site