On 2/17/21 11:02 AM, Andrew Morton wrote: > On Wed, 17 Feb 2021 10:49:25 -0800 Mike Kravetz <mike.krav...@oracle.com> > wrote: > >> page structs are not guaranteed to be contiguous for gigantic pages. The >> routine update_and_free_page can encounter a gigantic page, yet it assumes >> page structs are contiguous when setting page flags in subpages. >> >> If update_and_free_page encounters non-contiguous page structs, we can >> see “BUG: Bad page state in process …” errors. >> >> Non-contiguous page structs are generally not an issue. However, they can >> exist with a specific kernel configuration and hotplug operations. For >> example: Configure the kernel with CONFIG_SPARSEMEM and >> !CONFIG_SPARSEMEM_VMEMMAP. Then, hotplug add memory for the area where the >> gigantic page will be allocated. >> Zi Yan outlined steps to reproduce here [1]. >> >> [1] >> https://lore.kernel.org/linux-mm/16f7c58b-4d79-41c5-9b64-a1a1628f4...@nvidia.com/ >> >> Fixes: 944d9fec8d7a ("hugetlb: add support for gigantic page allocation at >> runtime") > > June 2014. That's a long lurk time for a bug. I wonder if some later > commit revealed it. > > I guess it doesn't matter a lot, but some -stable kernel maintainers > might wonder if they really need this fix...
I am not sure how common a CONFIG_SPARSEMEM and !CONFIG_SPARSEMEM_VMEMMAP config is. On the more popular architectures, this is not the default. But, you can build a kernel with such options. And, then you need to hotplug memory add and allocate a gigantic page there. It is unlikely to happen, but possible since Zi could force the BUG. The copy_huge_page_from_user bug requires the same non-normal configuration and is just as unlikely to occurr. But, since it can overwrite somewhat random pages I would feel better if it was fixed. -- Mike Kravetz
