[PATCH] hv_balloon: avoid touching uninitialized struct page during tail onlining

Sasha Levin sashal at kernel.org
Mon Jan 7 17:56:50 UTC 2019


On Mon, Jan 07, 2019 at 02:44:30PM +0100, Vitaly Kuznetsov wrote:
>David Hildenbrand <david at redhat.com> writes:
>> On 04.01.19 15:19, Vitaly Kuznetsov wrote:
>>> Hyper-V memory hotplug protocol has 2M granularity and in Linux x86 we use
>>> 128M. To deal with it we implement partial section onlining by registering
>>> custom page onlining callback (hv_online_page()). Later, when more memory
>>> arrives we try to online the 'tail' (see hv_bring_pgs_online()).
>>>
>>> It was found that in some cases this 'tail' onlining causes issues:
>>>
>>>  BUG: Bad page state in process kworker/0:2  pfn:109e3a
>>>  page:ffffe08344278e80 count:0 mapcount:1 mapping:0000000000000000 index:0x0
>>>  flags: 0xfffff80000000()
>>>  raw: 000fffff80000000 dead000000000100 dead000000000200 0000000000000000
>>>  raw: 0000000000000000 0000000000000000 0000000000000000 0000000000000000
>>>  page dumped because: nonzero mapcount
>>>  ...
>>>  Workqueue: events hot_add_req [hv_balloon]
>>>  Call Trace:
>>>   dump_stack+0x5c/0x80
>>>   bad_page.cold.112+0x7f/0xb2
>>>   free_pcppages_bulk+0x4b8/0x690
>>>   free_unref_page+0x54/0x70
>>>   hv_page_online_one+0x5c/0x80 [hv_balloon]
>>>   hot_add_req.cold.24+0x182/0x835 [hv_balloon]
>>>   ...
>>>
>>> Turns out that we now have deferred struct page initialization for memory
>>> hotplug so e.g. memory_block_action() in drivers/base/memory.c does
>>> pages_correctly_probed() check and in that check it avoids inspecting
>>> struct pages and checks sections instead. But in Hyper-V balloon driver we
>>> do PageReserved(pfn_to_page()) check and this is now wrong.
>>>
>>> Switch to checking online_section_nr() instead.
>>>
>>> Signed-off-by: Vitaly Kuznetsov <vkuznets at redhat.com>
>>> ---
>>>  drivers/hv/hv_balloon.c | 10 ++++++----
>>>  1 file changed, 6 insertions(+), 4 deletions(-)
>>>
>>> diff --git a/drivers/hv/hv_balloon.c b/drivers/hv/hv_balloon.c
>>> index 5301fef16c31..7c6349a50ef1 100644
>>> --- a/drivers/hv/hv_balloon.c
>>> +++ b/drivers/hv/hv_balloon.c
>>> @@ -888,12 +888,14 @@ static unsigned long handle_pg_range(unsigned long pg_start,
>>>  			pfn_cnt -= pgs_ol;
>>>  			/*
>>>  			 * Check if the corresponding memory block is already
>>> -			 * online by checking its last previously backed page.
>>> -			 * In case it is we need to bring rest (which was not
>>> -			 * backed previously) online too.
>>> +			 * online. It is possible to observe struct pages still
>>> +			 * being uninitialized here so check section instead.
>>> +			 * In case the section is online we need to bring the
>>> +			 * rest of pfns (which were not backed previously)
>>> +			 * online too.
>>>  			 */
>>>  			if (start_pfn > has->start_pfn &&
>>> -			    !PageReserved(pfn_to_page(start_pfn - 1)))
>>> +			    online_section_nr(pfn_to_section_nr(start_pfn)))
>>>  				hv_bring_pgs_online(has, start_pfn, pgs_ol);
>>>
>>>  		}
>>>
>>
>> I wonder if you should use pfn_to_online_page() and check for PageOffline().
>>
>> (I guess online_section_nr() should also do the trick)
>
>I'm worried a bit about racing with mm code here as we're not doing
>mem_hotplug_begin()/done() so I'd slightly prefer keeping
>online_section_nr() (pfn_to_online_page() also uses it but then it gets
>to the particular struct page). Moreover, with pfn_to_online_page() we
>will be looking at some other pfn - because the start_pfn is definitelly
>offline (pre-patch we were looking at start_pfn-1). Just looking at the
>whole section seems cleaner.
>
>P.S. I still think about bringing mem_hotplug_begin()/done() to
>hv_balloon but that's going to be a separate discussion, here I want to
>have a small fix backportable to stable.

This should probably be marked for stable then :)

--
Thanks,
Sasha


More information about the devel mailing list