Best wishes,
--
Ning Qu (曲宁) | Software Engineer | quning@xxxxxxxxxx | +1-408-418-6066
On Tue, Oct 15, 2013 at 3:29 AM, Kirill A. Shutemov
<kirill.shutemov@xxxxxxxxxxxxxxx> wrote:
> Ning Qu wrote:
>> We don't support huge page when page is moved from page cache to swap.
>> So in this function, we enable huge page handling in two case:
>>
>> 1) when a huge page is found in the page cache,
>> 2) or we need to alloc a huge page for page cache
>>
>> We have to refactor all the calls to shmem_getpages to simplify the job
>> of caller. Right now shmem_getpage does:
>>
>> 1) simply request a page, default as a small page
>> 2) or caller specify a flag to request either a huge page or a small page,
>> then leave the caller to decide how to use it
>>
>> Signed-off-by: Ning Qu <quning@xxxxxxxxx>
>> ---
>> mm/shmem.c | 139 +++++++++++++++++++++++++++++++++++++++++++++++--------------
>> 1 file changed, 108 insertions(+), 31 deletions(-)
>>
>> diff --git a/mm/shmem.c b/mm/shmem.c
>> index 447bd14..8fe17dd 100644
>> --- a/mm/shmem.c
>> +++ b/mm/shmem.c
>> @@ -115,15 +115,43 @@ static unsigned long shmem_default_max_inodes(void)
>> static bool shmem_should_replace_page(struct page *page, gfp_t gfp);
>> static int shmem_replace_page(struct page **pagep, gfp_t gfp,
>> struct shmem_inode_info *info, pgoff_t index);
>> +
>> static int shmem_getpage_gfp(struct inode *inode, pgoff_t index,
>> - struct page **pagep, enum sgp_type sgp, gfp_t gfp, int *fault_type);
>> + struct page **pagep, enum sgp_type sgp, gfp_t gfp, int flags,
>> + int *fault_type);
>> +
>> +#ifdef CONFIG_TRANSPARENT_HUGEPAGE_PAGECACHE
>> +static inline int shmem_getpage(struct inode *inode, pgoff_t index,
>> + struct page **pagep, enum sgp_type sgp, gfp_t gfp, int flags,
>> + int *fault_type)
>> +{
>> + int ret = 0;
>> + struct page *page = NULL;
>>
>> + if ((flags & AOP_FLAG_TRANSHUGE) &&
>> + mapping_can_have_hugepages(inode->i_mapping)) {
>
> I don't think we need ifdef here. mapping_can_have_hugepages() will be 0
> compile-time, if CONFIG_TRANSPARENT_HUGEPAGE_PAGECACHE is not defined and
> compiler should optimize out thp case.
The same problem since HPAGE_CACHE_INDEX_MASK is build bug?
>
>> @@ -1298,27 +1348,37 @@ repeat:
>> error = -ENOSPC;
>> goto unacct;
>> }
>> - percpu_counter_inc(&sbinfo->used_blocks);
>> }
>>
>> - page = shmem_alloc_page(gfp, info, index);
>> + if (must_use_thp) {
>> + page = shmem_alloc_hugepage(gfp, info, index);
>> + if (page) {
>> + count_vm_event(THP_WRITE_ALLOC);
>> + nr = hpagecache_nr_pages(page);
>
> nr = hpagecache_nr_pages(page) can be moved below if (must_use_thp).
> hpagecache_nr_pages(page) evaluates to 0 for small pages.
>
you mean something like this? If so, then fixed.
if (must_use_thp) {
page = shmem_alloc_hugepage(gfp, info, index);
if (page) {
count_vm_event(THP_WRITE_ALLOC);
} else
count_vm_event(THP_WRITE_ALLOC_FAILED);
} else {
page = shmem_alloc_page(gfp, info, index);
}
if (!page) {
error = -ENOMEM;
goto unacct;
}
nr = hpagecache_nr_pages(page);
>
> --
> Kirill A. Shutemov
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
