Re: [PATCH v3 01/10] mm/hugetlb: add cache of descriptors to resv_map for region_add

[Mike Kravetz <mike.kravetz@oracle.com>]

On 07/17/2015 02:02 AM, Naoya Horiguchi wrote:
> On Sun, Jul 12, 2015 at 09:20:59PM -0700, Mike Kravetz wrote:
>> fallocate hole punch will want to remove a specific range of
>> pages.  When pages are removed, their associated entries in
>> the region/reserve map will also be removed.  This will break
>> an assumption in the region_chg/region_add calling sequence.
>> If a new region descriptor must be allocated, it is done as
>> part of the region_chg processing.  In this way, region_add
>> can not fail because it does not need to attempt an allocation.
>>
>> To prepare for fallocate hole punch, create a "cache" of
>> descriptors that can be used by region_add if necessary.
>> region_chg will ensure there are sufficient entries in the
>> cache.  It will be necessary to track the number of in progress
>> add operations to know a sufficient number of descriptors
>> reside in the cache.  A new routine region_abort is added to
>> adjust this in progress count when add operations are aborted.
>> vma_abort_reservation is also added for callers creating
>> reservations with vma_needs_reservation/vma_commit_reservation.
>>
>> Signed-off-by: Mike Kravetz <mike.kravetz@oracle.com>
>> ---
>>   include/linux/hugetlb.h |   3 +
>>   mm/hugetlb.c            | 169 ++++++++++++++++++++++++++++++++++++++++++------
>>   2 files changed, 153 insertions(+), 19 deletions(-)
>>
>> diff --git a/include/linux/hugetlb.h b/include/linux/hugetlb.h
>> index d891f94..667cf44 100644
>> --- a/include/linux/hugetlb.h
>> +++ b/include/linux/hugetlb.h
>> @@ -35,6 +35,9 @@ struct resv_map {
>>   	struct kref refs;
>>   	spinlock_t lock;
>>   	struct list_head regions;
>> +	long adds_in_progress;
>> +	struct list_head rgn_cache;
>> +	long rgn_cache_count;
>>   };
>>   extern struct resv_map *resv_map_alloc(void);
>>   void resv_map_release(struct kref *ref);
>> diff --git a/mm/hugetlb.c b/mm/hugetlb.c
>> index a8c3087..241d16d 100644
>> --- a/mm/hugetlb.c
>> +++ b/mm/hugetlb.c
>> @@ -240,11 +240,14 @@ struct file_region {
>>   
>>   /*
>>    * Add the huge page range represented by [f, t) to the reserve
>> - * map.  Existing regions will be expanded to accommodate the
>> - * specified range.  We know only existing regions need to be
>> - * expanded, because region_add is only called after region_chg
>> - * with the same range.  If a new file_region structure must
>> - * be allocated, it is done in region_chg.
>> + * map.  In the normal case, existing regions will be expanded
>> + * to accommodate the specified range.  Sufficient regions should
>> + * exist for expansion due to the previous call to region_chg
>> + * with the same range.  However, it is possible that region_del
>> + * could have been called after region_chg and modifed the map
>> + * in such a way that no region exists to be expanded.  In this
>> + * case, pull a region descriptor from the cache associated with
>> + * the map and use that for the new range.
>>    *
>>    * Return the number of new huge pages added to the map.  This
>>    * number is greater than or equal to zero.
>> @@ -261,6 +264,27 @@ static long region_add(struct resv_map *resv, long f, long t)
>>   		if (f <= rg->to)
>>   			break;
>>   
>> +	if (&rg->link == head || t < rg->from) {
>> +		/*
>> +		 * No region exists which can be expanded to include the
>> +		 * specified range.  Pull a region descriptor from the
>> +		 * cache, and use it for this range.
>> +		 */
> 
> This comment mentions this if-block, not the VM_BUG_ON below, so it had
> better be put the above if-line.

OK, I will move and make a minor modification to the comment.

> 
>> +		VM_BUG_ON(!resv->rgn_cache_count);
> 
> resv->rgn_cache_count <= 0 might be safer.

Sure.

> ...
>> @@ -3236,11 +3360,14 @@ retry:
>>   	 * any allocations necessary to record that reservation occur outside
>>   	 * the spinlock.
>>   	 */
>> -	if ((flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED))
>> +	if ((flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED)) {
>>   		if (vma_needs_reservation(h, vma, address) < 0) {
>>   			ret = VM_FAULT_OOM;
>>   			goto backout_unlocked;
>>   		}
>> +		/* Just decrements count, does not deallocate */
>> +		vma_abort_reservation(h, vma, address);
>> +	}
> 
> This is not "abort reservation" operation, but you use "abort reservation"
> routine, which might confusing and makes future maintenance hard. I think
> this should be done in a simplified variant of vma_commit_reservation()
> (maybe just an alias of your vma_abort_reservation()) or fast path in
> vma_commit_reservation().

I am struggling a bit with the names of these routines.  The
routines in question are:

vma_needs_reservation - This is a wrapper for region_chg(), so the
	return value is the number of regions needed for the page.
	Since there is only one page, the routine effectively
	becomes a boolean.  Hence the name "needs".

vma_commit_reservation - This is a wrapper for region_add().  It
	must be called after a prior call to vma_needs_reservation
	and after actual allocation of the page.

We need a way to handle the case where vma_needs_reservation has
been called, but the page allocation is not successful.  I chose
the name vma_abort_reservation, but as noted (even in my comments)
it is not an actual abort.

I am not sure if you are suggesting vma_commit_reservation() should
handle this as a special case.  I think a separately named routine which
indicates and end of the reservation/allocation process would be
easier to understand.

What about changing the name vma_abort_reservation() to
vma_end_reservation()?  This would indicate that the reservation/
allocation process is ended.

> Thanks,
> Naoya Horiguchi

Thank you for your reviews.
-- 
Mike Kravetz

> 
>>   
>>   	ptl = huge_pte_lockptr(h, mm, ptep);
>>   	spin_lock(ptl);
>> @@ -3387,6 +3514,8 @@ int hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
>>   			ret = VM_FAULT_OOM;
>>   			goto out_mutex;
>>   		}
>> +		/* Just decrements count, does not deallocate */
>> +		vma_abort_reservation(h, vma, address);
>>   
>>   		if (!(vma->vm_flags & VM_MAYSHARE))
>>   			pagecache_page = hugetlbfs_pagecache_page(h,
>> @@ -3726,6 +3855,8 @@ int hugetlb_reserve_pages(struct inode *inode,
>>   	}
>>   	return 0;
>>   out_err:
>> +	if (!vma || vma->vm_flags & VM_MAYSHARE)
>> +		region_abort(resv_map, from, to);
>>   	if (vma && is_vma_resv_set(vma, HPAGE_RESV_OWNER))
>>   		kref_put(&resv_map->refs, resv_map_release);
>>   	return ret;
>> -- 
>> 2.1.0

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>