Re: [PATCH -mm v2] mm/page_isolation: fix potential warning from user

[Qian Cai]

> On Jan 20, 2020, at 9:01 AM, David Hildenbrand <david@xxxxxxxxxx> wrote:
> 
> On 20.01.20 14:56, Qian Cai wrote:
>> 
>> 
>>> On Jan 20, 2020, at 8:38 AM, David Hildenbrand <david@xxxxxxxxxx> wrote:
>>> 
>>> On 20.01.20 14:30, David Hildenbrand wrote:
>>>> On 20.01.20 14:19, Qian Cai wrote:
>>>>> It makes sense to call the WARN_ON_ONCE(zone_idx(zone) == ZONE_MOVABLE)
>>>>> from start_isolate_page_range(), but should avoid triggering it from
>>>>> userspace, i.e, from is_mem_section_removable() because it could be a
>>>>> DoS if warn_on_panic is set.
>>>>> 
>>>>> While at it, simplify the code a bit by removing an unnecessary jump
>>>>> label and a local variable, so set_migratetype_isolate() could really
>>>>> return a bool.
>>>>> 
>>>>> Suggested-by: Michal Hocko <mhocko@xxxxxxxxxx>
>>>>> Signed-off-by: Qian Cai <cai@xxxxxx>
>>>>> ---
>>>>> 
>>>>> v2: Improve the commit log.
>>>>>   Warn for all start_isolate_page_range() users not just offlining.
>>>>> 
>>>>> mm/page_alloc.c     | 11 ++++-------
>>>>> mm/page_isolation.c | 30 +++++++++++++++++-------------
>>>>> 2 files changed, 21 insertions(+), 20 deletions(-)
>>>>> 
>>>>> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
>>>>> index 621716a25639..3c4eb750a199 100644
>>>>> --- a/mm/page_alloc.c
>>>>> +++ b/mm/page_alloc.c
>>>>> @@ -8231,7 +8231,7 @@ struct page *has_unmovable_pages(struct zone *zone, struct page *page,
>>>>> 		if (is_migrate_cma(migratetype))
>>>>> 			return NULL;
>>>>> 
>>>>> -		goto unmovable;
>>>>> +		return page;
>>>>> 	}
>>>>> 
>>>>> 	for (; iter < pageblock_nr_pages; iter++) {
>>>>> @@ -8241,7 +8241,7 @@ struct page *has_unmovable_pages(struct zone *zone, struct page *page,
>>>>> 		page = pfn_to_page(pfn + iter);
>>>>> 
>>>>> 		if (PageReserved(page))
>>>>> -			goto unmovable;
>>>>> +			return page;
>>>>> 
>>>>> 		/*
>>>>> 		 * If the zone is movable and we have ruled out all reserved
>>>>> @@ -8261,7 +8261,7 @@ struct page *has_unmovable_pages(struct zone *zone, struct page *page,
>>>>> 			unsigned int skip_pages;
>>>>> 
>>>>> 			if (!hugepage_migration_supported(page_hstate(head)))
>>>>> -				goto unmovable;
>>>>> +				return page;
>>>>> 
>>>>> 			skip_pages = compound_nr(head) - (page - head);
>>>>> 			iter += skip_pages - 1;
>>>>> @@ -8303,12 +8303,9 @@ struct page *has_unmovable_pages(struct zone *zone, struct page *page,
>>>>> 		 * is set to both of a memory hole page and a _used_ kernel
>>>>> 		 * page at boot.
>>>>> 		 */
>>>>> -		goto unmovable;
>>>>> +		return page;
>>>>> 	}
>>>>> 	return NULL;
>>>>> -unmovable:
>>>>> -	WARN_ON_ONCE(zone_idx(zone) == ZONE_MOVABLE);
>>>>> -	return pfn_to_page(pfn + iter);
>>>>> }
>>>>> 
>>>>> #ifdef CONFIG_CONTIG_ALLOC
>>>>> diff --git a/mm/page_isolation.c b/mm/page_isolation.c
>>>>> index e70586523ca3..31f5516f5d54 100644
>>>>> --- a/mm/page_isolation.c
>>>>> +++ b/mm/page_isolation.c
>>>>> @@ -15,12 +15,12 @@
>>>>> #define CREATE_TRACE_POINTS
>>>>> #include <trace/events/page_isolation.h>
>>>>> 
>>>>> -static int set_migratetype_isolate(struct page *page, int migratetype, int isol_flags)
>>>>> +static bool set_migratetype_isolate(struct page *page, int migratetype,
>>>>> +				    int isol_flags)
>>>> 
>>>> Why this change?
>>>> 
>>>>> {
>>>>> -	struct page *unmovable = NULL;
>>>>> +	struct page *unmovable = ERR_PTR(-EBUSY);
>>>> 
>>>> Also, why this change?
>>>> 
>>>>> 	struct zone *zone;
>>>>> 	unsigned long flags;
>>>>> -	int ret = -EBUSY;
>>>>> 
>>>>> 	zone = page_zone(page);
>>>>> 
>>>>> @@ -49,21 +49,25 @@ static int set_migratetype_isolate(struct page *page, int migratetype, int isol_
>>>>> 									NULL);
>>>>> 
>>>>> 		__mod_zone_freepage_state(zone, -nr_pages, mt);
>>>>> -		ret = 0;
>>>>> 	}
>>>>> 
>>>>> out:
>>>>> 	spin_unlock_irqrestore(&zone->lock, flags);
>>>>> -	if (!ret)
>>>>> +
>>>>> +	if (!unmovable) {
>>>>> 		drain_all_pages(zone);
>>>>> -	else if ((isol_flags & REPORT_FAILURE) && unmovable)
>>>>> -		/*
>>>>> -		 * printk() with zone->lock held will guarantee to trigger a
>>>>> -		 * lockdep splat, so defer it here.
>>>>> -		 */
>>>>> -		dump_page(unmovable, "unmovable page");
>>>>> -
>>>>> -	return ret;
>>>>> +	} else {
>>>>> +		WARN_ON_ONCE(zone_idx(zone) == ZONE_MOVABLE);
>>>>> +
>>>>> +		if ((isol_flags & REPORT_FAILURE) && !IS_ERR(unmovable))
>>>>> +			/*
>>>> 
>>>> Why this change? (!IS_ERR)
>>>> 
>>>> 
>>>> Some things here look unrelated - or I am missing something :)
>>>> 
>>> 
>>> FWIW, I'd prefer this change without any such cleanups (e.g., I don't
>>> like returning a bool from this function and the IS_ERR handling, makes
>>> the function harder to read than before)
>> 
>> What is Michal or Andrewâs opinion? BTW, a bonus point to return a bool
>> is that it helps the code robustness in general, as UBSAN will be able to
>> catch any abuse.
>> 
> 
> A return type of bool on a function that does not test a property
> ("has_...", "is"...") is IMHO confusing.

That is fine. It could be renamed to set_migratetype_is_isolate() or
is_set_migratetype_isolate() which seems pretty minor because we
have no consistency in the naming of this in linux kernel at all, i.e.,
many existing bool function names without those test of properties. 

> 
> If we have an int, it is clear that "0" means "success". With a bool
> (true/false), it is not clear.
> 
> -- 
> Thanks,
> 
> David / dhildenb
>