linux-kernel - Re: [PATCH 1/1] mm, memory_hotplug: Initialize struct pages for the full memory section

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [day] [month] [year] [list]
Message-ID: <20181211221419.wr7p72u235dosb5u@master>
Date:   Tue, 11 Dec 2018 22:14:19 +0000
From:   Wei Yang <richard.weiyang@...il.com>
To:     Zaslonko Mikhail <zaslonko@...ux.bm.com>
Cc:     Wei Yang <richard.weiyang@...il.com>,
        Mikhail Zaslonko <zaslonko@...ux.ibm.com>,
        akpm@...ux-foundation.org, linux-kernel@...r.kernel.org,
        linux-mm@...ck.org, mhocko@...nel.org,
        Pavel.Tatashin@...rosoft.com, schwidefsky@...ibm.com,
        heiko.carstens@...ibm.com, gerald.schaefer@...ibm.com
Subject: Re: [PATCH 1/1] mm, memory_hotplug: Initialize struct pages for the
 full memory section

On Tue, Dec 11, 2018 at 04:23:05PM +0100, Zaslonko Mikhail wrote:
>Hello,
>
>On 11.12.2018 02:50, Wei Yang wrote:
>> On Mon, Dec 10, 2018 at 05:14:36PM +0100, Zaslonko Mikhail wrote:
>>> Hello,
>>>
>>> On 10.12.2018 16:10, Wei Yang wrote:
>>>> On Mon, Dec 10, 2018 at 02:07:12PM +0100, Mikhail Zaslonko wrote:
>>>>> If memory end is not aligned with the sparse memory section boundary, the
>>>>> mapping of such a section is only partly initialized. This may lead to
>>>>> VM_BUG_ON due to uninitialized struct page access from
>>>>> is_mem_section_removable() or test_pages_in_a_zone() function triggered by
>>>>> memory_hotplug sysfs handlers:
>>>>>
>>>>> page:000003d082008000 is uninitialized and poisoned
>>>>> page dumped because: VM_BUG_ON_PAGE(PagePoisoned(p))
>>>>> Call Trace:
>>>>> ([<0000000000385b26>] test_pages_in_a_zone+0xde/0x160)
>>>>>  [<00000000008f15c4>] show_valid_zones+0x5c/0x190
>>>>>  [<00000000008cf9c4>] dev_attr_show+0x34/0x70
>>>>>  [<0000000000463ad0>] sysfs_kf_seq_show+0xc8/0x148
>>>>>  [<00000000003e4194>] seq_read+0x204/0x480
>>>>>  [<00000000003b53ea>] __vfs_read+0x32/0x178
>>>>>  [<00000000003b55b2>] vfs_read+0x82/0x138
>>>>>  [<00000000003b5be2>] ksys_read+0x5a/0xb0
>>>>>  [<0000000000b86ba0>] system_call+0xdc/0x2d8
>>>>> Last Breaking-Event-Address:
>>>>>  [<0000000000385b26>] test_pages_in_a_zone+0xde/0x160
>>>>> Kernel panic - not syncing: Fatal exception: panic_on_oops
>>>>>
>>>>> Fix the problem by initializing the last memory section of the highest zone
>>>>> in memmap_init_zone() till the very end, even if it goes beyond the zone
>>>>> end.
>>>>>
>>>>> Signed-off-by: Mikhail Zaslonko <zaslonko@...ux.ibm.com>
>>>>> Reviewed-by: Gerald Schaefer <gerald.schaefer@...ibm.com>
>>>>> Cc: <stable@...r.kernel.org>
>>>>> ---
>>>>> mm/page_alloc.c | 15 +++++++++++++++
>>>>> 1 file changed, 15 insertions(+)
>>>>>
>>>>> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
>>>>> index 2ec9cc407216..41ef5508e5f1 100644
>>>>> --- a/mm/page_alloc.c
>>>>> +++ b/mm/page_alloc.c
>>>>> @@ -5542,6 +5542,21 @@ void __meminit memmap_init_zone(unsigned long size, int nid, unsigned long zone,
>>>>> 			cond_resched();
>>>>> 		}
>>>>> 	}
>>>>> +#ifdef CONFIG_SPARSEMEM
>>>>> +	/*
>>>>> +	 * If there is no zone spanning the rest of the section
>>>>> +	 * then we should at least initialize those pages. Otherwise we
>>>>> +	 * could blow up on a poisoned page in some paths which depend
>>>>> +	 * on full sections being initialized (e.g. memory hotplug).
>>>>> +	 */
>>>>> +	if (end_pfn == max_pfn) {
>>>>> +		while (end_pfn % PAGES_PER_SECTION) {
>>>>> +			__init_single_page(pfn_to_page(end_pfn), end_pfn, zone,
>>>>> +					   nid);
>>>>> +			end_pfn++;
>>>>> +		}
>>>>> +	}
>>>>> +#endif
>>>>
>>>> If my understanding is correct, end_pfn is not a valid range.
>>>>
>>>> memmap_init_zone() initialize the range [start_pfn, start_pfn + size). I
>>>> am afraid this will break the syntax. 
>>>>
>>>> And max_pfn is also not a valid one. For example, on x86,
>>> I used pfn_max here to check for the highest zone. What would be a better way? 
>>>
>>>> update_end_of_memory_vars() will update max_pfn, which is calculated by:
>>>>
>>>>     end_pfn = PFN_UP(start + size);
>>>>
>>>> BTW, as you mentioned this apply to hotplug case. And then why this couldn't
>>>> happen during boot up? What differ these two cases?
>>>
>>> Well, the pages left uninitialized during bootup (initial problem), but the panic itself takes 
>>> place when we try to process memory_hotplug sysfs attributes (thus triggering sysfs handlers). 
>>> You can find more details in the original thread:
>>> https://marc.info/?t=153658306400001&r=1&w=2
>>>
>> 
>> Thanks.
>> 
>> I took a look into the original thread and try to reproduce this on x86.
>> 
>> My step is:
>> 
>>   1. config page_poisoning
>>   2. use kernel parameter mem=3075M
>>   3. cat the last memory block device sysfs file removable
>>      eg. when mem is 3075, cat memory9/removable
>> 
>> I don't see the Call trace. Do I miss something to reproduce it?
>> 
>
>No you don't. I guess there might be deviations depending on the architecture (I am on s390).
>As I understand, memory block size is 384 Mb on your system and memory9 is the last block on the list?

Sorry, my calculation is not correct. The last memory_block is 23
instead of 9.

>BTW, do you have CONFIG_DEBUG_VM and CONFIG_DEBUG_VM_PGFLAGS on? 
>

Yes, I have set it:

CONFIG_DEBUG_VM=y
CONFIG_DEBUG_VM_PGFLAGS=y

And the kernel cmdline is:

BOOT_IMAGE=/vmlinuz-4.20.0-rc5+ root=UUID=98aa84d6-9ba6-4033-ab91-9ca6fe3dd74f ro \
resume=UUID=b7c21053-d9c1-4e58-8488-7d385f8ee107 console=ttyS0 \
LANG=en_US.UTF-8 mem=3075M

>
>>>>
>>>>> }
>>>>>
>>>>> #ifdef CONFIG_ZONE_DEVICE
>>>>> -- 
>>>>> 2.16.4
>>>>
>>>
>>> Thanks,
>>> Mikhail Zaslonko
>> 

-- 
Wei Yang
Help you, Help me