Re: [PATCH] mm, dump_page: do not crash with bad compound_page()

[John Hubbard <jhubbard@nvidia.com>]

On 8/4/20 11:48 AM, Matthew Wilcox wrote:
> On Tue, Aug 04, 2020 at 11:39:43AM -0700, John Hubbard wrote:
>> +static inline int __compound_mapcount(struct page *page)
>> +{
>> +	page = compound_head(page);
>> +	return atomic_read(compound_mapcount_ptr(page)) + 1;
>> +}
> 
> I'd suggest instead:
> 
> static inline int head_mapcount(struct page *head)
> {
> 	return atomic_read(compound_mapcount_ptr(head)) + 1;
> }
> 
>> +static inline int dump_page_compound_mapcount(struct page *page)
>> +{
>> +	if (WARN_ON_ONCE(!PageCompound(page)))
>> +		return 0;
>> +	return __compound_mapcount(page);
>>   }
> 
> And just dropping this ... it shouldn't be used outside mm/debug.c anyway.

Yes, that's cleaner. I'll send a v2 that is not a reply, so that Andrew can
spot it more easily. OK to add your "Suggested-by:" to that?

> 
> Thinking about it, we'll get the hint later that this is not to be trusted
> when the flags from the page do not include 'head'.
> 

Good point. I think this will work out pretty well, then.

thanks,
-- 
John Hubbard
NVIDIA