On 11/18/19 11:22 AM, Kirill A. Shutemov wrote:
On Fri, Nov 15, 2019 at 11:58:00AM -0800, Andrew Morton wrote:
On Fri, 15 Nov 2019 12:58:08 +0100 Thomas Hellström (VMware) <thomas_os@xxxxxxxxxxxx> wrote:
A huge pud page can theoretically be faulted in racing with pmd_alloc()
in __handle_mm_fault(). That will lead to pmd_alloc() returning an
invalid pmd pointer. Fix this by adding a pud_trans_unstable() function
similar to pmd_trans_unstable() and check whether the pud is really stable
before using the pmd pointer.
Race:
Thread 1: Thread 2: Comment
create_huge_pud() Fallback - not taken.
create_huge_pud() Taken.
pmd_alloc() Returns an invalid pointer.
What are the user-visible runtime effects of this change?
Data corruption: kernel writes to a huge page thing it's page table.
Is a -stable backport warranted?
I believe it is.
Note that this was caught during a code audit rather than a real
experienced problem. It looks to me like the only implementation that
currently creates huge pud pagetable entries is dev_dax_huge_fault()
which doesn't appear to care much about private (COW) mappings or
write-tracking which is, I believe, a prerequisite for create_huge_pud()
falling back on thread 1, but not in thread 2.
This means (assuming that's intentional) that a stable backport
shouldn't be needed.
For the WIP huge page support for graphics memory we'll be allowing both
COW mappings and write-tracking, though, but that's still some time away.
In any case, I think this patch needs -rc testing to catch potential
pud_devmap issues before submitted to stable.
Thanks,
Thomas
Acked-by: Kirill A. Shutemov <kirill.shutemov@xxxxxxxxxxxxxxx>
