Re: [PATCH v3] mm/hmm/test: use after free in dmirror_allocate_chunk()

[Dan Williams <dan.j.williams@intel.com>]

On Mon, Sep 28, 2020 at 5:52 PM Andrew Morton <akpm@linux-foundation.org> wrote:
>
> On Sat, 26 Sep 2020 19:17:20 -0300 Jason Gunthorpe <jgg@ziepe.ca> wrote:
>
> > On Sat, Sep 26, 2020 at 03:14:02PM +0300, Dan Carpenter wrote:
> > > The error handling code does this:
> > >
> > > err_free:
> > >     kfree(devmem);
> > >         ^^^^^^^^^^^^^
> > > err_release:
> > >     release_mem_region(devmem->pagemap.range.start, range_len(&devmem->pagemap.range));
> > >                            ^^^^^^^^
> > > The problem is that when we use "devmem->pagemap.range.start" the
> > > "devmem" pointer is either NULL or freed.
> > >
> > > Neither the allocation nor the call to request_free_mem_region() has to
> > > be done under the lock so I moved those to the start of the function.
> > >
> > > Fixes: 1f9c4bb986d9 ("mm/memremap_pages: convert to 'struct range'")
> > > Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
> > > Reviewed-by: Ralph Campbell <rcampbell@nvidia.com>
> > > ---
> > > v2: The first version introduced a locking bug
> > > v3: Markus Elfring pointed out that the Fixes tag was wrong.  This bug
> > > was in the original commit and then fixed and then re-introduced.  I was
> > > quite bothered by how this bug lasted so long in the source code, but
> > > now we know.  As soon as it is introduced we fixed it.
> > >
> > > One problem with the kernel QC process is that I think everyone marks
> > > the bug as "old/dealt with" so it was only because I was added a new
> > > check for resource leaks that it was found when it was re-introduced.
> > >
> > >  lib/test_hmm.c | 44 ++++++++++++++++++++++----------------------
> > >  1 file changed, 22 insertions(+), 22 deletions(-)
> >
> > Hi Andrew,
> >
> > I don't have have any hmm related patches this cycle, can you take
> > this into your tree?
> >
> > Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
>
> Thanks.
>
> It's actually a fix against Dan Williams' -mm patch "mm/memremap_pages:
> convert to 'struct range'"

Yes, sorry, for the fix:

Acked-by: Dan Williams <dan.j.williams@intel.com>