Re: [PATCH v3 1/7] mm: Remove BUG_ON mmap_sem not held from xxx_trans_huge_lock()

From: Thomas HellstrÃm (VMware)
Date: Thu Oct 03 2019 - 07:32:26 EST


Hi, Kirill,

On 10/3/19 1:02 PM, Kirill A. Shutemov wrote:
On Wed, Oct 02, 2019 at 03:47:24PM +0200, Thomas HellstrÃm (VMware) wrote:
From: Thomas Hellstrom <thellstrom@xxxxxxxxxx>

The caller needs to make sure that the vma is not torn down during the
lock operation and can also use the i_mmap_rwsem for file-backed vmas.
Remove the BUG_ON. We could, as an alternative, add a test that either
vma->vm_mm->mmap_sem or vma->vm_file->f_mapping->i_mmap_rwsem are held.

Cc: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
Cc: Matthew Wilcox <willy@xxxxxxxxxxxxx>
Cc: Will Deacon <will.deacon@xxxxxxx>
Cc: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
Cc: Rik van Riel <riel@xxxxxxxxxxx>
Cc: Minchan Kim <minchan@xxxxxxxxxx>
Cc: Michal Hocko <mhocko@xxxxxxxx>
Cc: Huang Ying <ying.huang@xxxxxxxxx>
Cc: JÃrÃme Glisse <jglisse@xxxxxxxxxx>
Cc: Kirill A. Shutemov <kirill@xxxxxxxxxxxxx>
Signed-off-by: Thomas Hellstrom <thellstrom@xxxxxxxxxx>
The patch looks good to me:

Acked-by: Kirill A. Shutemov <kirill.shutemov@xxxxxxxxxxxxxxx>

But I looked at usage at pagewalk.c and it is inconsitent. The walker
takes ptl before calling ->pud_entry(), but not for ->pmd_entry().

It should be fixed: do not take the lock before ->pud_entry(). The
callback must take care of it.

Looks like we have single ->pud_entry() implementation the whole kernel.
It should be trivial to fix.

Could you do this?

I could probably fix that. There are some comments in the patch introducing that code as to why it was done that way, though, but I don't remember offhand what the arguments were.

But there seems to be more races WRT puds. See my next email. Perhaps this should be fixed as part of a larger audit of the huge_pud code?

/Thomas