ac.preferred_zoneref->zone passed to alloc_flags_nofragment() can be NULL.
'zone' pointer unconditionally derefernced in alloc_flags_nofragment().
Bail out on NULL zone to avoid potential crash.
Currently we don't see any crashes only because alloc_flags_nofragment()
has another bug which allows compiler to optimize away all accesses to
'zone'.
Fixes: 6bb154504f8b ("mm, page_alloc: spread allocations across zones before introducing fragmentation")
Signed-off-by: Andrey Ryabinin <[email protected]>
---
mm/page_alloc.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 933bd42899e8..2b2c7065102f 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -3461,6 +3461,9 @@ alloc_flags_nofragment(struct zone *zone, gfp_t gfp_mask)
alloc_flags |= ALLOC_KSWAPD;
#ifdef CONFIG_ZONE_DMA32
+ if (!zone)
+ return alloc_flags;
+
if (zone_idx(zone) != ZONE_NORMAL)
goto out;
--
2.21.0
Commit 0a79cdad5eb2 ("mm: use alloc_flags to record if kswapd can wake")
removed setting of the ALLOC_NOFRAGMENT flag. Bring it back.
Fixes: 0a79cdad5eb2 ("mm: use alloc_flags to record if kswapd can wake")
Signed-off-by: Andrey Ryabinin <[email protected]>
---
mm/page_alloc.c | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 2b2c7065102f..a85b8252c5ad 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -3465,7 +3465,7 @@ alloc_flags_nofragment(struct zone *zone, gfp_t gfp_mask)
return alloc_flags;
if (zone_idx(zone) != ZONE_NORMAL)
- goto out;
+ return alloc_flags;
/*
* If ZONE_DMA32 exists, assume it is the one after ZONE_NORMAL and
@@ -3474,9 +3474,9 @@ alloc_flags_nofragment(struct zone *zone, gfp_t gfp_mask)
*/
BUILD_BUG_ON(ZONE_NORMAL - ZONE_DMA32 != 1);
if (nr_online_nodes > 1 && !populated_zone(--zone))
- goto out;
+ return alloc_flags;
-out:
+ alloc_flags |= ALLOC_NOFRAGMENT;
#endif /* CONFIG_ZONE_DMA32 */
return alloc_flags;
}
--
2.21.0
On Tue, Apr 23, 2019 at 03:08:06PM +0300, Andrey Ryabinin wrote:
> Commit 0a79cdad5eb2 ("mm: use alloc_flags to record if kswapd can wake")
> removed setting of the ALLOC_NOFRAGMENT flag. Bring it back.
>
> Fixes: 0a79cdad5eb2 ("mm: use alloc_flags to record if kswapd can wake")
> Signed-off-by: Andrey Ryabinin <[email protected]>
Acked-by: Mel Gorman <[email protected]>
--
Mel Gorman
SUSE Labs
On Tue, Apr 23, 2019 at 03:08:05PM +0300, Andrey Ryabinin wrote:
> ac.preferred_zoneref->zone passed to alloc_flags_nofragment() can be NULL.
> 'zone' pointer unconditionally derefernced in alloc_flags_nofragment().
> Bail out on NULL zone to avoid potential crash.
> Currently we don't see any crashes only because alloc_flags_nofragment()
> has another bug which allows compiler to optimize away all accesses to
> 'zone'.
>
> Fixes: 6bb154504f8b ("mm, page_alloc: spread allocations across zones before introducing fragmentation")
> Signed-off-by: Andrey Ryabinin <[email protected]>
Acked-by: Mel Gorman <[email protected]>
--
Mel Gorman
SUSE Labs
On Tue, 23 Apr 2019 15:08:06 +0300 Andrey Ryabinin <[email protected]> wrote:
> Commit 0a79cdad5eb2 ("mm: use alloc_flags to record if kswapd can wake")
> removed setting of the ALLOC_NOFRAGMENT flag. Bring it back.
What are the runtime effects of this fix?
On Tue, Apr 23, 2019 at 12:01:43PM -0700, Andrew Morton wrote:
> On Tue, 23 Apr 2019 15:08:06 +0300 Andrey Ryabinin <[email protected]> wrote:
>
> > Commit 0a79cdad5eb2 ("mm: use alloc_flags to record if kswapd can wake")
> > removed setting of the ALLOC_NOFRAGMENT flag. Bring it back.
>
> What are the runtime effects of this fix?
The runtime effect is that ALLOC_NOFRAGMENT behaviour is restored so
that allocations are spread across local zones to avoid fragmentation
due to mixing pageblocks as long as possible.
--
Mel Gorman
SUSE Labs
On Wed, 24 Apr 2019 10:04:03 +0100 Mel Gorman <[email protected]> wrote:
> On Tue, Apr 23, 2019 at 12:01:43PM -0700, Andrew Morton wrote:
> > On Tue, 23 Apr 2019 15:08:06 +0300 Andrey Ryabinin <[email protected]> wrote:
> >
> > > Commit 0a79cdad5eb2 ("mm: use alloc_flags to record if kswapd can wake")
> > > removed setting of the ALLOC_NOFRAGMENT flag. Bring it back.
> >
> > What are the runtime effects of this fix?
>
> The runtime effect is that ALLOC_NOFRAGMENT behaviour is restored so
> that allocations are spread across local zones to avoid fragmentation
> due to mixing pageblocks as long as possible.
OK, thanks. Is this worth a -stable backport?
On Wed, Apr 24, 2019 at 03:46:24PM -0700, Andrew Morton wrote:
> On Wed, 24 Apr 2019 10:04:03 +0100 Mel Gorman <[email protected]> wrote:
>
> > On Tue, Apr 23, 2019 at 12:01:43PM -0700, Andrew Morton wrote:
> > > On Tue, 23 Apr 2019 15:08:06 +0300 Andrey Ryabinin <[email protected]> wrote:
> > >
> > > > Commit 0a79cdad5eb2 ("mm: use alloc_flags to record if kswapd can wake")
> > > > removed setting of the ALLOC_NOFRAGMENT flag. Bring it back.
> > >
> > > What are the runtime effects of this fix?
> >
> > The runtime effect is that ALLOC_NOFRAGMENT behaviour is restored so
> > that allocations are spread across local zones to avoid fragmentation
> > due to mixing pageblocks as long as possible.
>
> OK, thanks. Is this worth a -stable backport?
Yes, but only for 5.0 obviously and both should be included if that is
the case. I did not push for it initially as problems in this area are
hard for a general user to detect and people have not complained about
5.0's fragmentation handling.
--
Mel Gorman
SUSE Labs
On Thu, 25 Apr 2019 00:40:53 +0100 Mel Gorman <[email protected]> wrote:
> On Wed, Apr 24, 2019 at 03:46:24PM -0700, Andrew Morton wrote:
> > On Wed, 24 Apr 2019 10:04:03 +0100 Mel Gorman <[email protected]> wrote:
> >
> > > On Tue, Apr 23, 2019 at 12:01:43PM -0700, Andrew Morton wrote:
> > > > On Tue, 23 Apr 2019 15:08:06 +0300 Andrey Ryabinin <[email protected]> wrote:
> > > >
> > > > > Commit 0a79cdad5eb2 ("mm: use alloc_flags to record if kswapd can wake")
> > > > > removed setting of the ALLOC_NOFRAGMENT flag. Bring it back.
> > > >
> > > > What are the runtime effects of this fix?
> > >
> > > The runtime effect is that ALLOC_NOFRAGMENT behaviour is restored so
> > > that allocations are spread across local zones to avoid fragmentation
> > > due to mixing pageblocks as long as possible.
> >
> > OK, thanks. Is this worth a -stable backport?
>
> Yes, but only for 5.0 obviously and both should be included if that is
> the case. I did not push for it initially as problems in this area are
> hard for a general user to detect and people have not complained about
> 5.0's fragmentation handling.
Ah, OK. 0a79cdad5eb2 didn't have a -stable tag so I suppose we can
leave this patch un-stabled.
If they went and backported 0a79cdad5eb2 anyway, let's hope the scripts
are smart enough to catch this patch's Fixes: link.