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

From: Wei Yang
Date: Tue Dec 11 2018 - 17:24:20 EST

On Tue, Dec 11, 2018 at 04:17:34PM +0100, Zaslonko Mikhail wrote:
>On 11.12.2018 10:49, Wei Yang wrote:
>> On Mon, Dec 10, 2018 at 02:24:51PM +0100, Michal Hocko wrote:
>>> On Mon 10-12-18 14:07:12, 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.
>>> It would be great to mention how you can end up in the situation like
>>> this(a user provided memmap or a strange HW).
>>>> 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.
>>> Why do we need to restrict this to the highest zone? In other words, why
>>> cannot we do what I was suggesting earlier [1]. What does prevent other
>>> zones to have an incomplete section boundary?
>>> [1]
>> I tried to go through the original list and make myself familiar with
>> the bug.
>> Confused why initialize the *last* end_pfn could fix this, since
>> is_mem_section_removable() will iterate on each page of a section. This
>> means we need to initialize all the pages left in the section.
>That's exactly what the fix does. We initialize all the pages left in
>the section.

You are right, I misunderstand your code.

>> One way to fix this in my mind is to record the last pfn in mem_section.
>Do you mean last initialized pfn? I guess we have agreed upon that the
>entire section should be initialized.

Well, that would be great.

>> This could be done in memory_preset(), since after that we may assume
>> the section is full. Not sure whether you would like it.

Wei Yang
Help you, Help me