Regression still in next for drop migrate type checks

Regression still in next for drop migrate type checks

[Tony Lindgren]

Hi all,

Looks like I'm still carrying patch "mm: distinguish CMA and MOVABLE
isolation in  has_unmovable_pages" from Michal [1] for commit e1d753dff0fa
("mm: drop migrate type checks from has_unmovable_pages") and Linux next
has been broken for CMA for few weeks now as noted in the discussion
also at [1].

What's the status of this regression? How come it's been known
broken for two weeks and still not fixed or reverted in next?
This is clearly is not suitable for "wider testing" ;)

For reference, the symptoms I'm seeing is that at least loadable
modules using fail to work CMA produce the following errors:

# dmesg | grep -B1 cma
[    0.000000] Memory policy: Data cache writealloc
[    0.000000] cma: Reserved 16 MiB at 0xbe800000
--
[    0.000000] Inode-cache hash table entries: 65536 (order: 6, 262144 bytes)
[    0.000000] Memory: 997000K/1046528K available (9216K kernel code, 815K rwdata, 2464K rodata, 1024K init, 7558K bss, 33144K reserved, 16384K cma-reserved, 243712K highmem)
--
[   22.436920] twl_rtc 48070000.i2c:twl@48:rtc: rtc core: registered 48070000.i2c:twl@48:rtc as rtc0
[   23.593933] cma: cma_alloc: alloc failed, req-size: 32 pages, ret: -16
--
[   24.768920] ehci-omap 4a064c00.ehci: new USB bus registered, assigned bus number 1
[   24.828338] cma: cma_alloc: alloc failed, req-size: 1 pages, ret: -16
--
[   27.030517] DSS: OMAP DSS rev 4.0
[   27.068237] cma: cma_alloc: alloc failed, req-size: 1 pages, ret: -16
--
[   27.651489] random: crng init done
[   27.738494] cma: cma_alloc: alloc failed, req-size: 129 pages, ret: -16

Regards,

Tony


[1] https://lkml.org/lkml/2017/10/19/364

Re: Regression still in next for drop migrate type checks

[Andrew Morton]

On Fri, 3 Nov 2017 09:00:12 -0700 Tony Lindgren <tony@atomide.com> wrote:

> Looks like I'm still carrying patch "mm: distinguish CMA and MOVABLE
> isolation in  has_unmovable_pages" from Michal [1] for commit e1d753dff0fa
> ("mm: drop migrate type checks from has_unmovable_pages") and Linux next
> has been broken for CMA for few weeks now as noted in the discussion
> also at [1].
> 
> What's the status of this regression? How come it's been known
> broken for two weeks and still not fixed or reverted in next?
> This is clearly is not suitable for "wider testing" ;)

Travel to Europe and subsequent bustage in the -next tree held things
up.  Today's merge looks good so I expect to get this into Stephen
today.

In future, please feel free to send such hot fixes to Stephen.

Re: Regression still in next for drop migrate type checks

[Tony Lindgren]

* Andrew Morton <akpm@linux-foundation.org> [171103 17:34]:
> On Fri, 3 Nov 2017 09:00:12 -0700 Tony Lindgren <tony@atomide.com> wrote:
> 
> > Looks like I'm still carrying patch "mm: distinguish CMA and MOVABLE
> > isolation in  has_unmovable_pages" from Michal [1] for commit e1d753dff0fa
> > ("mm: drop migrate type checks from has_unmovable_pages") and Linux next
> > has been broken for CMA for few weeks now as noted in the discussion
> > also at [1].
> > 
> > What's the status of this regression? How come it's been known
> > broken for two weeks and still not fixed or reverted in next?
> > This is clearly is not suitable for "wider testing" ;)
> 
> Travel to Europe and subsequent bustage in the -next tree held things
> up.  Today's merge looks good so I expect to get this into Stephen
> today.

OK thanks for the update.

> In future, please feel free to send such hot fixes to Stephen.

Sure.

Tony

Re: Regression still in next for drop migrate type checks

[Michal Hocko]

On Fri 03-11-17 09:00:12, Tony Lindgren wrote:
> Hi all,
> 
> Looks like I'm still carrying patch "mm: distinguish CMA and MOVABLE
> isolation in  has_unmovable_pages" from Michal [1] for commit e1d753dff0fa
> ("mm: drop migrate type checks from has_unmovable_pages") and Linux next
> has been broken for CMA for few weeks now as noted in the discussion
> also at [1].
> 
> What's the status of this regression? How come it's been known
> broken for two weeks and still not fixed or reverted in next?
> This is clearly is not suitable for "wider testing" ;)

This is the first time I hear about the regression. Joonsoo has pointed
out that there might be a problem but he didn't have any specific
failure in hands. The fix has been posted [1] and Andrew has picked it
up few days ago. Could you try to apply that patch? It will show up in
linux-next as soon as Andrew releases mmotm.

[1] http://lkml.kernel.org/r/20171019122118.y6cndierwl2vnguj@dhcp22.suse.cz
-- 
Michal Hocko
SUSE Labs

Re: Regression still in next for drop migrate type checks

[Michal Hocko]

On Sat 04-11-17 09:25:00, Michal Hocko wrote:
> On Fri 03-11-17 09:00:12, Tony Lindgren wrote:
> > Hi all,
> > 
> > Looks like I'm still carrying patch "mm: distinguish CMA and MOVABLE
> > isolation in  has_unmovable_pages" from Michal [1] for commit e1d753dff0fa
> > ("mm: drop migrate type checks from has_unmovable_pages") and Linux next
> > has been broken for CMA for few weeks now as noted in the discussion
> > also at [1].
> > 
> > What's the status of this regression? How come it's been known
> > broken for two weeks and still not fixed or reverted in next?
> > This is clearly is not suitable for "wider testing" ;)
> 
> This is the first time I hear about the regression. Joonsoo has pointed
> out that there might be a problem but he didn't have any specific
> failure in hands. The fix has been posted [1] and Andrew has picked it
> up few days ago. Could you try to apply that patch? It will show up in
> linux-next as soon as Andrew releases mmotm.

Just noticed that Andrew has already released mmotm with the patch
included. I would appreciate if you could just apply this patch on top
of failing kernel to double check and have your Tested-by.

Thanks!
-- 
Michal Hocko
SUSE Labs

Re: Regression still in next for drop migrate type checks

[Tony Lindgren]

* Michal Hocko <mhocko@kernel.org> [171104 08:26]:
> On Fri 03-11-17 09:00:12, Tony Lindgren wrote:
> > Hi all,
> > 
> > Looks like I'm still carrying patch "mm: distinguish CMA and MOVABLE
> > isolation in  has_unmovable_pages" from Michal [1] for commit e1d753dff0fa
> > ("mm: drop migrate type checks from has_unmovable_pages") and Linux next
> > has been broken for CMA for few weeks now as noted in the discussion
> > also at [1].
> > 
> > What's the status of this regression? How come it's been known
> > broken for two weeks and still not fixed or reverted in next?
> > This is clearly is not suitable for "wider testing" ;)
> 
> This is the first time I hear about the regression. Joonsoo has pointed
> out that there might be a problem but he didn't have any specific
> failure in hands. The fix has been posted [1] and Andrew has picked it
> up few days ago. Could you try to apply that patch? It will show up in
> linux-next as soon as Andrew releases mmotm.

Yes thanks I can confirm that next-20171106 works for me. So as far as
I'm concerned, we're good to go.

Regards,

Tony


> [1] http://lkml.kernel.org/r/20171019122118.y6cndierwl2vnguj@dhcp22.suse.cz

Re: Regression still in next for drop migrate type checks

[Michal Hocko]

On Mon 06-11-17 06:38:04, Tony Lindgren wrote:
> * Michal Hocko <mhocko@kernel.org> [171104 08:26]:
> > On Fri 03-11-17 09:00:12, Tony Lindgren wrote:
> > > Hi all,
> > > 
> > > Looks like I'm still carrying patch "mm: distinguish CMA and MOVABLE
> > > isolation in  has_unmovable_pages" from Michal [1] for commit e1d753dff0fa
> > > ("mm: drop migrate type checks from has_unmovable_pages") and Linux next
> > > has been broken for CMA for few weeks now as noted in the discussion
> > > also at [1].
> > > 
> > > What's the status of this regression? How come it's been known
> > > broken for two weeks and still not fixed or reverted in next?
> > > This is clearly is not suitable for "wider testing" ;)
> > 
> > This is the first time I hear about the regression. Joonsoo has pointed
> > out that there might be a problem but he didn't have any specific
> > failure in hands. The fix has been posted [1] and Andrew has picked it
> > up few days ago. Could you try to apply that patch? It will show up in
> > linux-next as soon as Andrew releases mmotm.
> 
> Yes thanks I can confirm that next-20171106 works for me. So as far as
> I'm concerned, we're good to go.

Can I assume your Tested-by?
-- 
Michal Hocko
SUSE Labs

Re: Regression still in next for drop migrate type checks

[Tony Lindgren]

* Michal Hocko <mhocko@kernel.org> [171104 08:37]:
> On Sat 04-11-17 09:25:00, Michal Hocko wrote:
> > On Fri 03-11-17 09:00:12, Tony Lindgren wrote:
> > > Hi all,
> > > 
> > > Looks like I'm still carrying patch "mm: distinguish CMA and MOVABLE
> > > isolation in  has_unmovable_pages" from Michal [1] for commit e1d753dff0fa
> > > ("mm: drop migrate type checks from has_unmovable_pages") and Linux next
> > > has been broken for CMA for few weeks now as noted in the discussion
> > > also at [1].
> > > 
> > > What's the status of this regression? How come it's been known
> > > broken for two weeks and still not fixed or reverted in next?
> > > This is clearly is not suitable for "wider testing" ;)
> > 
> > This is the first time I hear about the regression. Joonsoo has pointed
> > out that there might be a problem but he didn't have any specific
> > failure in hands. The fix has been posted [1] and Andrew has picked it
> > up few days ago. Could you try to apply that patch? It will show up in
> > linux-next as soon as Andrew releases mmotm.
> 
> Just noticed that Andrew has already released mmotm with the patch
> included. I would appreciate if you could just apply this patch on top
> of failing kernel to double check and have your Tested-by.

Yes Linux next now works for me like I replied earlier. Probably already
too late for a Tested-by as it's been applied :)

Tony

Re: Regression still in next for drop migrate type checks

[Tony Lindgren]

* Michal Hocko <mhocko@kernel.org> [171106 14:44]:
> On Mon 06-11-17 06:38:04, Tony Lindgren wrote:
> > * Michal Hocko <mhocko@kernel.org> [171104 08:26]:
> > > On Fri 03-11-17 09:00:12, Tony Lindgren wrote:
> > > > Hi all,
> > > > 
> > > > Looks like I'm still carrying patch "mm: distinguish CMA and MOVABLE
> > > > isolation in  has_unmovable_pages" from Michal [1] for commit e1d753dff0fa
> > > > ("mm: drop migrate type checks from has_unmovable_pages") and Linux next
> > > > has been broken for CMA for few weeks now as noted in the discussion
> > > > also at [1].
> > > > 
> > > > What's the status of this regression? How come it's been known
> > > > broken for two weeks and still not fixed or reverted in next?
> > > > This is clearly is not suitable for "wider testing" ;)
> > > 
> > > This is the first time I hear about the regression. Joonsoo has pointed
> > > out that there might be a problem but he didn't have any specific
> > > failure in hands. The fix has been posted [1] and Andrew has picked it
> > > up few days ago. Could you try to apply that patch? It will show up in
> > > linux-next as soon as Andrew releases mmotm.
> > 
> > Yes thanks I can confirm that next-20171106 works for me. So as far as
> > I'm concerned, we're good to go.
> 
> Can I assume your Tested-by?

Yes sounds like it already got added.

Regards,

Tony