Re: [RESEND PATCH v2] of: fix kmemleak crash caused by imbalance in early memory reservation

[Guenter Roeck]

On 3/6/19 5:39 AM, Rob Herring wrote:
> On Tue, Mar 5, 2019 at 8:12 PM Guenter Roeck <linux@xxxxxxxxxxxx> wrote:
> > On Tue, Feb 12, 2019 at 04:12:24PM -0600, Rob Herring wrote:
> > > On Tue, Feb 12, 2019 at 3:50 PM Stephen Rothwell <sfr@xxxxxxxxxxxxxxxx> wrote:
> > > > Hi all,
> > > >
> > > > On Tue, 12 Feb 2019 10:03:09 -0600 Rob Herring <robh+dt@xxxxxxxxxx> wrote:
> > > > > On Mon, Feb 11, 2019 at 10:47 AM Marc Gonzalez <marc.w.gonzalez@xxxxxxx> wrote:
> > > > > > On 04/02/2019 15:37, Marc Gonzalez wrote:
> > > > > >
> > > > > > > Cc: stable@xxxxxxxxxxxxxxx # 3.15+
> > > > > > > Fixes: 3f0c820664483 ("drivers: of: add initialization code for dynamic reserved memory")
> > > > > > > Acked-by: Marek Szyprowski <m.szyprowski@xxxxxxxxxxx>
> > > > > > > Acked-by: Prateek Patel <prpatel@xxxxxxxxxx>
> > > > > > > Tested-by: Marc Gonzalez <marc.w.gonzalez@xxxxxxx>
> > > > > > > Signed-off-by: Mike Rapoport <rppt@xxxxxxxxxxxxx>
> > > > > > > ---
> > > > > > > Resend with DT CCed to reach robh's patch queue
> > > > > > > I added CC: stable, Fixes, and Prateek's ack
> > > > > > > Trim recipients list to minimize inconvenience
> > > > > >
> > > > > > I'm confused over commit 3532b3b554a216f30edb841d29eef48521bdc592 in linux-next
> > > > > > "memblock: drop __memblock_alloc_base()"
> > > > > >
> > > > > > It's definitely going to conflict with the proposed patch
> > > > > > over drivers/of/of_reserved_mem.c
> > > > > >
> > > > > > Rob, what's the next step then?
> > > > >
> > > > > Rebase it on top of what's in linux-next and apply it to the tree
> > > > > which has the above dependency. I'm guessing that is Andrew Morton's
> > > > > tree.
> > > >
> > > > Yeah, that is in Andrew's "post linux-next" patch series, so if you
> > > > rebase it on top of linux-next and then send it to Andrew with some
> > > > explanation.
> > > >
> > > > ...
> > > >
> > > > Actually, if it is intended for the stable trees, then presumably it is
> > > > intended to go to Linus for the current release?  In which case, the
> > > > patch in Andrew's tree will have to be changed to cope after your patch
> > > > appears in Linus' tree (and therefore, linux-next).
> > >
> > > At this point in the cycle, I wasn't planning to send this for 5.0.
> > > It's not fixing something introduced in 5.0 and it is a debug feature.
> > Hi Rob,
> >
> > this may be a debug feature, but we do test our kernels with it enabled,
> > and the problem does affect our 4.19 branch (chromeos-4.19). Are you
> > suggesting that we should backport the fix into our branch and not send
> > the backport to -stable ?
>
> No, not at all. Just that I wasn't going to add it to the probable
> last 5.0-rc and would wait.
>
> However, it's complicated by other memblock changes in 5.1 and not a
> trivial backport. One of the versions on the list should be easier to
> backport than what's in mainline (or going to be).

We went ahead and applied a backport of an older version of the patch series
to chromeos-4.19. We'll see how well that works, but so far it looks like
it fixes our problem.

Guenter